Patent application title: Industrial Fatty Acid Engineering General System for Modifying Fatty Acids
Inventors:
IPC8 Class: AC12P740FI
USPC Class:
1 1
Class name:
Publication date: 2020-06-11
Patent application number: 20200181656
Abstract:
Compositions and methods for a hybrid biological and chemical process
utilizing chemotrophic microorganisms that converts syngas and/or gaseous
CO2 and/or a mixture of CO2 gas and H2 gas into one or more desaturated
hydrocarbons, unsaturated fatty acids, hydroxy acids, or diacids.Claims:
1.-20. (canceled)
21. A composition comprising a bacterial cell in a culture medium that is in contact with a feedstock that comprises syngas and/or gaseous CO.sub.2 and/or other C1 compounds and/or inorganic carbon and/or liquid, gaseous, and/or solid products of pyrolytic reactions, wherein said bacterial cell chemoautotrophically converts said feedstock into organic compounds comprising one or more lipids or hydrocarbons comprising fatty alcohols, alkenes, unsaturated fatty acids, diacids, or hydroxy acids, wherein the bacterial cell comprises one or more exogenous nucleic acid sequence encoding an enzyme selected from the group consisting of a fatty acid aldehyde acyl-ACP reductase, a fatty acid aldehyde decarbonylase, a thioesterase, a fatty acyl-CoA/fatty acyl-ACP reductase, a fatty acid hydroxylase, a desaturase, a cytochrome P450 enzyme, or a NADH or NADPH cytochrome P450 reductase, and wherein the bacterial cell is selected from Ralstonia sp.; Alcaligenes sp.; and Cupriavidus sp.
22. The composition of claim 21, wherein the bacterial cell is chosen from Rhodococcus DSM 43205, Rhodococcus DSM 43206, Rhodococcus DSM 3346, Cupriavidus DSM531, or Cupriavidus DSM 541.
23. The composition of claim 21, wherein the bacterial cell comprises at least a first and a second exogenous nucleic acid sequence but no more than five exogenous nucleic acid sequences.
24. The composition of claim 21, wherein the exogenous nucleic acid sequence consists of a first, second, and a third exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid aldehyde acyl-ACP reductase, the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase, and the third exogenous nucleic acid sequence encodes a thioesterase.
25. The composition of claim 21, wherein the exogenous nucleic acid sequence consists of a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid aldehyde acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase.
26. The composition of claim 21, wherein the one or more lipids or hydrocarbons comprise at least one organic molecule having a carbon chain length of at least 8 carbon atoms and at least one carbon-carbon double bond.
27. The composition of claim 21, wherein the one or more lipids comprise at least one hydroxy acid molecule having a carbon chain length of at least 6 carbon atoms.
28. The composition of claim 21, wherein the one or more lipids comprise at least one diacid acid molecule having a carbon chain length of at least 6 carbon atoms.
29. The composition of claim 21, wherein the one or more lipids or hydrocarbons comprise a mixture of lipids having at least one unsaturated fatty acid molecule having a carbon chain length from 8 carbon atoms to 30 carbon atoms.
30. The composition of claim 21, wherein the one or more hydrocarbons comprise a mixture of hydrocarbons having at least one desaturated hydrocarbon molecule having a carbon chain length from 8 carbon atoms to 30 carbon atoms.
31. The composition of claim 21, wherein the one or more lipids comprise a mixture of lipids having at least one hydroxy acid molecule having a carbon chain length from 8 carbon atoms to 24 carbon atoms.
32. The composition of claim 21, wherein the one or more lipids comprise a mixture of lipids having at least one dicarboxylic acid molecule having a carbon chain length from 6 carbon atoms to 18 carbon atoms.
33. The composition of claim 21, wherein the one or more lipids or hydrocarbons comprise a quantity of at least one alkene, alkyne, hydroxy acid, dicarboxylic acid, and/or unsaturated fatty acid at a level higher than the quantity of the alkene, alkyne, hydroxy acid, dicarboxylic acid, and/or unsaturated fatty acid in the same bacterial cell not comprising the at least first exogenous nucleic acid sequence.
34. The composition of claim 21, wherein the bacterial cell produces and/or secretes at least 60% of one or more lipids or hydrocarbons by weight.
35. The composition of claim 21, wherein less than 10% by weight of the organic compounds produced is organic acid with carbon chain length of four carbons or less.
36. The composition of claim 21, wherein said feedstock comprises gaseous H2 that is utilized by the bacterial cell as an energy source for said chemoautotrophic conversion into organic compounds and that is generated or recycled using renewable, alternative, or conventional sources of power selected from at least one of photovoltaics, solar thermal, wind power, hydroelectric, nuclear, geothermal, enhanced geothermal, ocean thermal, ocean wave power, and tidal power.
37. The composition of claim 21, where said feedstock comprises syngas that is generated from lignocellulosic energy crops, crop residue, bagasse, saw dust, forestry residue, food waste, municipal solid waste, biogas, landfill gas, or stranded natural gas.
38. The composition of claim 21, wherein said gaseous feedstock comprises CO.sub.2, H.sub.2 and/or CO; and O.sub.2.
39. The composition of claim 21, wherein said exogenous nucleic acid sequence consists of an exogenous nucleic acid sequence that encodes a thioesterase.
40. The composition of claim 21, wherein said culture medium does not comprise an organic carbon source as a primary source of carbon for microbial growth or biosynthesis.
Description:
RELATED APPLICATIONS
[0001] This application claims priority under 35 U.S.C. .sctn. 119(e) to U.S. Provisional Patent Application No. 61/536,056, filed Sep. 19, 2011 and entitled "INDUSTRIAL FATTY ACID ENGINEERING SYSTEM." This application is also a continuation-in-part of International Patent Application No. PCT/US2011/34218, filed Apr. 27, 2011, and entitled "USE OF OXYHYDROGEN MICROORGANISMS FOR NON-PHOTOSYNTHETIC CARBON CAPTURE AND CONVERSION OF INORGANIC AND/OR C1 CARBON SOURCES INTO USEFUL ORGANIC COMPOUNDS," which is a continuation-in-part of International Patent Application No. PCT/US2010/001402, filed May 12, 2010, and entitled "BIOLOGICAL AND CHEMICAL PROCESS UTILIZING CHEMOAUTOTROPHIC MICROORGANISMS FOR THE CHEMOSYNTHETIC FIXATION OF CARBON DIOXIDE AND/OR OTHER INORGANIC CARBON SOURCES INTO ORGANIC COMPOUNDS, AND THE GENERATION OF ADDITIONAL USEFUL PRODUCTS," which is a continuation-in-part of U.S. patent application Ser. No. 12/613,550, filed Nov. 6, 2009, and entitled "BIOLOGICAL AND CHEMICAL PROCESS UTILIZING CHEMOAUTOTROPHIC MICROORGANISMS FOR THE CHEMOSYNTHETIC FIXATION OF CARBON DIOXIDE AND/OR OTHER INORGANIC CARBON SOURCES INTO ORGANIC COMPOUNDS, AND THE GENERATION OF ADDITIONAL USEFUL PRODUCTS," which claims the benefit of U.S. Provisional Patent Application No. 61/111,794, filed Nov. 6, 2008, and entitled, "BIOLOGICAL AND CHEMICAL PROCESS UTILIZING CHEMOAUTOTROPHIC MICROORGANISMS FOR THE RECYCLING OF CARBON FROM CARBON DIOXIDE AND OTHER INORGANIC CARBON SOURCES THROUGH CHEMOSYNTHESIS INTO BIOFUEL AND ADDITIONAL USEFUL PRODUCTS." Each of these applications is incorporated herein by reference in its entirety for all purposes.
FIELD OF INVENTION
[0002] This disclosure relates to compositions capable of producing and methods of the producing oils, fuels, and oleochemicals through cultivating bacteria that grow on carbon-containing gas such as syngas, producer gas, CO2, carbon monoxide and mixtures of the same containing hydrogen gas. This disclosure further relates to methods of fixing carbon from gas input into useful organic molecules such as diacids, hydroxy fatty acids, unsaturated fatty acids. The bacteria of the invention can be genetically engineered for use in the methods or other aspects of the invention described herein.
BACKGROUND
[0003] Sustainable and renewable sources of organic molecules are needed to help reduce the amount of carbon dioxide emissions in the atmosphere, as well as to reduce global energy consumption based upon coal, oil, and natural gas economies. Increased demand in the global economy has placed increasing pressure on traditional hydrocarbon feedstocks for the production of organic molecules. Many industries, including plastics and chemical manufacturers, rely heavily on the availability of fossil hydrocarbon sources as a feedstock for their manufacturing processes. Cost-effective alternatives to current sources of supply could help mitigate the upward pressure on fossil resource demand and raw material costs.
[0004] Biologic systems that fix carbon through natural biochemical metabolic processes are known. Algal systems have been developed to create hydrocarbon or petroleum replacements through photosynthetic reactions, as well as heterotrophic reactions utilizing fixed carbon feedstocks such as sugar that indirectly depend upon photosynthesis, however insufficient yields limit the effectiveness, economic feasibility, practicality and commercial adoption of these algal technologies. Bacterial cells have also been genetically engineered to process sugar feedstocks into useful organic compounds in heterotrophic fermentation systems; however, there are significant drawbacks for these systems as well.
[0005] Heterotrophic fermentations are vulnerable to contamination because heterotrophic microorganisms that can grow on fixed carbon nutrients and compete with a production strain are far more ubiquitous in the surface environment. Heterotrophic technologies also generally suffer limitations in terms of food versus fuel conflict and negative environmental impacts.
[0006] Gas-to-liquid (GTL) technologies have the benefit of allowing the utilization of waste carbon sources--including highly lignocellulosic waste through the conversion to synthesis gas (syngas) via gasification, as well as waste CO2 through the provision of reduced hydrogen--in the production of organic chemicals. Syngas is a mix of gases that generally contains H.sub.2, CO, and CO.sub.2 as major components, which can be generated through steam reforming of methane and/or liquid petroleum gas or biogas or through gasification of any organic material, including but not limited to biomass, waste organic matter, various polymers, and coal. Many gasification processes are available for the production of syngas. A number of gasification processes subject the carbonaceous feedstock to partial oxidation at high temperatures (500-1500.degree. C.), with the oxygen supply restricted to prevent complete combustion, producing syngas with varying composition depending on feedstock and reaction conditions such that the ratio of H.sub.2:CO can range from 0.5:1 to 3:1. The hydrogen component of syngas can be raised through the reaction of CO with steam in the water gas shift reaction with a concomitant increase in CO.sub.2 in the syngas mix.
[0007] Some major technologies for syngas conversion to chemicals include chemical catalytic processes such as the Fischer-Tropsch (F-T) as well as processes for the synthesis of methanol or other mixed alcohols, and biological syngas fermentation processes. F-T has been worked on for almost one hundred years and relies on metal-based, inorganic catalysts for the conversion of syngas into longer chain hydrocarbons. Difficulties with F-T include: a wide chain length distribution of products resulting in the need to reprocess short chain length products such as methane and LPG and/or the need to perform additional costly post-processing steps on long chain waxes and tars such as hydrocracking; high catalyst sensitivity to syngas impurities such as sulfur containing compounds, tars, and particulates, generally necessitating multiple costly gas clean up steps; relatively low flexibility in terms of accommodating various ratios of syngas constituents i.e. H2:CO, and low tolerance of CO.sub.2, often resulting in additional costly syngas conditioning steps such as water gas shift and CO.sub.2 removal; the actual F-T step being relatively high temperature and pressure resulting in costly compression and heating requirements; the wide distribution of products generally necessitating the storage, handling, and transport of a wide array of products which is often uneconomic except for relatively large scale operations; F-T products (e.g. diesel, jet fuel, naphtha, waxes) being relatively low in value at current (2011) prices compared to many different higher value oils, lipids, and oleochemicals that can be produced biologically. The difficulties with F-T generally also apply to other chemical conversion processes such as methanol synthesis.
[0008] The gasification of biomass to generate syngas has a long history going back to World War II where biomass gasification was used for running modified automobiles, boats, buses, and trucks. Presently, a number of biomass gasification technologies are at, or near commercialization (able to gasify 10,000 or more tons of biomass per year), and are generally used for the production of heat and/or electricity. The synthesis of chemicals or fuels from syngas generated via biomass gasification is at an earlier stage of development, and is generally pre-commercial.
[0009] Using syngas and/or CO.sub.2 and/or renewable H.sub.2 in a gas bioprocess enables the utilization of cheaper and more flexible sources of energy and/or carbon for the biological synthesis of sustainable chemicals and fuels than is possible through heterotrophic or phototrophic synthesis. In a syngas bioprocess, syngas acts as both a carbon and energy source for the microbial culture. Some of the advantages of syngas bioprocesses include: the production of a relatively narrow range of carbon chain length distribution compared to F-T; lower sensitivity to syngas impurities; greater tolerance of varying ratios of H2:CO and the presence of CO2; ability to operate at much closer to ambient temperature and pressure; ability to produce various higher value oleochemical products.
[0010] A bioprocess based upon a gaseous feedstock such as syngas can allow for far lower negative environmental and food production impacts in the biological synthesis of liquid fuels and/or chemicals than the highly land and water intensive heterotrophic or phototrophic-based technologies. However, current biological GTL technologies generally yield relatively short chain alcohols, or other short chain organic compounds, as products, which have relatively low energy density and infrastructure compatibility and limited applications. The syngas-growing microorganisms used in current biological GTL technologies are generally inappropriate for the synthesis of mid- to long-carbon chain lipid-based chemicals. Furthermore the types of microorganisms used in current biological GTL technologies such as Clostridia have a relatively low tolerance for their short carbon chain gas fermentation products such as ethanol, butanol, or acetic acid, which limits titers and complicates product recovery, hurting the overall economics of the GTL process.
[0011] There is a need to identify a set of microorganisms that can grow in conventional and scalable contained reaction vessels and that produce commercially viable sets of organic carbon chains of at least eight carbon atoms long in a commercially feasible method. There is a need to identify microorganisms not limited metabolically by typical fixed carbon inputs such as sugar, and a microorganism that can additionally utilize syngas, producer gas, as well as a wide array of abiotic sources of carbon and energy for the synthesis of drop-in fuels and chemicals, leading to a feedstock flexibility that far exceeds comparable heterotrophic systems. There is a need to identify and use microorganisms that can utilize electron donors such as hydrogen, present in syngas, producer gas, as well as readily generated through a wide array of abiotic renewable energy technologies, for growth and carbon fixation.
[0012] The targeting of fatty acids produced through fatty acid biosynthesis to short chain lengths has been achieved in heterotrophic microorganisms. This has been accomplished through the use of thioesterases to change populations of fatty acids C8-C14 and the over-expression of thioesterases to increase shorter chain length fatty acids. Examples in the prior art include C8-C14 thioesterase expression to produce shorter chain lengths in U.S. Pat. No. 7,883,882 Renewable chemical production from novel fatty acid feedstocks, Franklin et al. Solazyme, p. 58.
[0013] However there is a need to target the production of shorter chain length fatty acids in microorganisms that are capable of growing and producing lipids chemotrophically on syngas or H.sub.2/CO.sub.2 gas mixes to enable microbial GTL production of lipids with targeted, mid-length carbon chains.
[0014] Dicarboxylic acids (Diacids) such as dodecanedoic acid (n=10) arc used in production of nylon (nylon-6,12), polyamides, coatings, adhesives, greases, polyesters, dyes, detergents, flame retardants and fragrances. Diacids can be produced by fermentation of long-chain alkanes by Candida tropicalis (Kroha K, Infom 2004, 15, 568). Traumatic acid, monounsaturated dodecanedoic acid (10E-dodeca-1,12-dicarboxylic acid) has been produced from plant tissues English J et al., Science 1939, 90, 329. Pyrococcus furiosus produces an array of dicarboxylic acids (Carballeira, 1997). The total amount of dicarboxylic acids comprises only 3.4% of the total, however, this could be boosted by various literature methods.
[0015] There is a need for a biological, non-heterotrophic means of producing diacids from low-cost or sustainable syngas feedstocks.
[0016] Nutritionally important n-3 fatty acids include .alpha.-linolenic acid (ALA), eicosapentaenoic acid (EPA), and docosahexaenoic acid (DHA), all of which are polyunsaturated. N-3 fatty acids that are important in human physiology are .alpha.-linolenic acid (18:3, n-3; ALA), eicosapentaenoic acid (20:5, n-3; EPA), and docosahexaenoic acid (22:6, n-3; DHA). These three polyunsaturates have either 3, 5, or 6 double bonds in a carbon chain of 18, 20, or 22 carbon atoms, respectively. As with most naturally produced fatty acids, all double bonds are in the cis-configuration.
[0017] A fatty acid desaturase is an enzyme that removes two hydrogen atoms from a fatty acid, creating a carbon/carbon double bond. These desaturases are classified as delta--indicating that the double bond is created at a fixed position from the carboxyl group of a fatty acid (for example, .DELTA.9 desaturase creates a double bond at the 9th position from the carboxyl end). omega (e.g. .omega.3desaturase)--indicating the double bond is created between the third and fourth carbon from the methyl end of the fatty acid. In the biosynthesis of essential fatty acids, an elongase alternates with different desaturases (for example, .DELTA.6desaturase) repeatedly inserting an ethyl group, then forming a double bond.
[0018] Most polyunsaturated oils come from fish and there is a need for alternate, and particularly microbial sources of polyunsaturated fatty acids, given depleting fish stocks and increasing pollution in the oceans.
SUMMARY OF THE INVENTION
[0019] The present invention allows microorganisms to be engineered to convert CO.sub.2 gas and/or syngas and/or producer gas to higher value mid- to long-carbon chain length oleochemicals or monomers. The present technology allows the development of new genetically enhanced strains of microorganisms that can be used for syngas bioprocessing within biological gas-to-liquid (GTL) processes to produce and/or secrete various relatively long chain organic compounds that are drop-in, and are currently only produced in bulk from petroleum or higher plants.
[0020] The present invention relates to the engineering of microorganisms, including but not limited to hydrogen oxidizing, carbon monoxide oxidizing, and knallgas microorganisms, with a natural capability to grow and synthesize biomass on gaseous carbon sources such as syngas and/or CO.sub.2, such that the engineered microorganisms synthesize targeted products, including chemicals and fuels, under gas cultivation. The microorganisms and methods of the present invention enable low cost synthesis of chemicals and fuels, which can compete on price with petrochemicals and higher-plant derived oleochemicals and monomers, and which will generally have a substantially lower price than oleochemicals produced through heterotrophic or phototrophic synthesis.
[0021] The invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more lipids. In some embodiments, the composition comprises a microorganism, wherein the microorganism is a hydrogen-oxidizing chemotrophic microorganism. In some embodiments, the composition comprises a microorganism, wherein the microorganism is a carbon monoxide-oxidizing microorganism. In some embodiments, the composition comprises a microorganism, wherein the microorganism is a knallgas microorganism. In some embodiments, the composition comprises a microorganism, wherein the microorganism is chosen from the genera Rhodococcus or Gordonia. In some embodiments, the composition comprises a microorganism, wherein the microorganism is Rhodococcus opacus. In some embodiments, the composition comprises a microorganism, wherein the microorganism is Rhodococcus opacus (DSM 43205) or Rhodococcus sp (DSM 3346). In some embodiments, the composition comprises a microorganism, wherein the microorganism is chosen from the genera Ralstonia or Cupriavidus. In some embodiments, the composition comprises a microorganism, wherein the microorganism is Cupriavidus necator.
[0022] In some embodiments, the composition comprises a microorganism wherein the microorganism can naturally grow on H2/CO2 and/or syngas, and wherein the microorganism can naturally accumulate lipid to 50% or more of the cell biomass by weight. In some embodiments the microorganisms have a native ability to send a high flux of carbon down the fatty acid biosynthesis pathway. In some embodiments the microorganism exhibiting these traits is Rhodococcus opacus (DSM 43205 or DSM 43206).
[0023] In some embodiments, the composition comprises a microorganism that can naturally grow on H2/CO2 and/or syngas, and wherein the microorganism can naturally accumulate polyhydroxybutyrate (PHB) or polyhydroxyalkanoate (PHA) to 50% or more of the cell biomass by weight. In some embodiments the microorganisms have a native ability to direct a high flux of carbon through the acetyl-CoA metabolic intermediate, which can lead into fatty acid biosynthesis, along with a number of other synthetic pathways including PHA and PHB synthesis. A microorganism is considered to direct a high flux of carbon through acetyl-CoA if a product of a synthesis pathway going through the acetyl-CoA metabolic intermediate, including but not limited to polyhydroxybutyrate (PHB) or polyhydroxyalkanoate (PHA), can represent 50% or more of the cell biomass by weight. In some embodiments the microorganism exhibiting these traits is Cupriavidus necator (DSM 531 or DSM 541).
[0024] In some embodiments, the invention relates to a non-naturally occurring microorganism capable of converting syngas or other gaseous carbon sources into targeted oleochemical and/or monomer products, where the wild-type microorganism is capable of growing on syngas or other gaseous carbon sources, but is either not capable of synthesizing said targeted oleochemical and/or monomer products, or is capable of synthesizing the targeted oleochemicals and/or monomers, but is not capable of synthesizing the targeted biochemical products at the concentration and/or efficiency of the non-natural microorganism. In such microorganisms, one or more proteins or enzymes are expressed in the microorganism, thereby modifying, extending, diverting, enhancing, promoting, or otherwise altering the lipid biosynthesis pathway or its regulation for the synthesis and/or enhanced synthesis of a targeted lipid-based product, oleochemical, monomer, or hydrocarbon.
[0025] In some embodiments, the invention relates to a non-naturally occurring microorganism capable of converting syngas or other gaseous carbon sources into targeted oleochemical and monomer products, where the wild-type microorganism is capable of growing on syngas or other gaseous carbon sources and is capable of synthesizing said targeted oleochemical and monomer products, but the non-naturally occurring microorganism is capable of synthesizing the targeted biochemical products at a higher concentration and/or efficiency than the wild-type microorganism due to the overexpression and/or underexpression of one or more proteins or enzymes.
[0026] In some embodiments, the invention relates to compositions comprising one or more bacterial cells that consist of one, two, or three exogenous nucleic acid sequences where said bacteria can grow using syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas as a source of carbon and/or energy.
[0027] In some embodiments, the invention relates to compositions comprising one or more bacterial cells of Rhodococcus opacus (DSM 43205) that consist of zero, one, two, or three exogenous nucleic acid sequences.
[0028] In some embodiments one, two, or three exogenous nucleic acid sequences encode one or more thioesterase proteins.
[0029] In some embodiments the source of thioesterase is inherent to the production organisms. In some embodiments the source of thioesterase is Rhodococcus opacus B4. In some embodiments the thioesterase is derived from bacteria or plants other than the host microorganism.
[0030] In some embodiments, the invention relates to compositions comprising one or more bacterial cells that consist of two exogenous nucleic acid sequences that encode the following proteins: fatty acid acyl-ACP reductase, a fatty acid aldehyde decarbonylase, where said bacteria can grow using syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas as a source of carbon and/or energy.
[0031] In some embodiments, the invention relates to compositions comprising one or more bacterial cells that consist of three exogenous nucleic acid sequences that encode the following proteins: fatty acid acyl-ACP reductase, a fatty acid aldehyde decarbonylase, and a thioesterase, where said bacteria can grow using syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas as a source of carbon and/or energy.
[0032] In some embodiments, the bacterial cell produces and/or secretes one or more lipids in an amount that is greater than the amount of lipids produced and/or secreted by the same cell not comprising the exogenous nucleic acid sequence.
[0033] In some embodiments, the bacterial cell produces and/or secretes one or more lipids having a given carbon chain length, where the amount of said lipid produced and/or secreted is greater than the amount produced and/or secreted by the same cell not comprising the exogenous nucleic acid sequence.
[0034] In some embodiments, the bacterial cell produces and/or secretes one or more lipid molecules in an amount that is less than the amount of lipids produced and/or secreted by the same cell not comprising the exogenous nucleic acid sequence.
[0035] In some embodiments, the bacterial cell produces and/or secretes one or more hydrocarbons in an amount that is greater than the amount of hydrocarbons produced and/or secreted by the same cell not comprising the exogenous nucleic acid sequence.
[0036] In some embodiments, the bacterial cell or compositions comprising the bacterial cell comprise at least one exogenous nucleic acid sequence that is integrated into the genome of the cell.
[0037] In some embodiments, the invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more unsaturated hydrocarbons, wherein the microorganism comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase. In some embodiments the microorganism is Rhodococcus opacus.
[0038] In some embodiments the invention relates to a method of producing a lipid or mixture of lipids in a microorganism population comprising the cell or the composition described herein, wherein the method comprises: culturing a population of microorganisms comprising the cell or the composition described herein in a feedstock comprising syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas.
[0039] In some embodiments, the invention relates to a method of producing a lipid or mixture of lipids, wherein the method comprises: culturing a population of bacterial cells comprising the cell or the composition described herein in a feedstock comprising syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas. In some embodiments, the microorganism population comprises a bacterial strain of Rhodococcus opacus. In some embodiments, the microorganism population comprises a bacterial strain of Rhodococcus opacus (DSM 43205 or 43206).
[0040] In some embodiments, the invention relates to a method of producing a lipid or mixture of lipids, wherein the method comprises: culturing a population of bacterial cells comprising the cell or the composition described herein in a feedstock comprising methanol, a common impurity of syngas, with or without the addition of syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas. In some embodiments, the microorganism population comprises a bacterial strain of Rhodococcus opacus. In some embodiments, the microorganism population comprises a bacterial strain of Rhodococcus opacus (DSM 43205).
[0041] In some embodiments, the invention relates to a method of producing a lipid or mixture of lipids, wherein the method comprises: culturing a population of bacterial cells comprising the cell or the composition described herein in a feedstock comprising syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas. In some embodiments, the microorganism population comprises a bacterial strain of Cupriavidus necator.
[0042] In some embodiments, the molecule is chosen from one or more alkene, alkyne, unsaturated fatty acid, hydroxyacid and/or dicarboxylic acid (diacid). In some embodiments, the method produces a lipid or mixture of lipids at a quantity higher than the quantity of lipid or mixture of lipids in the same bacterial cell population not comprising the exogenous nucleic acids described herein. In some embodiments the one or more lipids comprise a quantity of at least one alkene, alkyne, unsaturated fatty acid, hydroxyacid and/or di acid at a level higher than the quantity of the alkene, alkyne, unsaturated fatty acid, hydroxyacid and/or diacid in the same microorganism not comprising the exogenous nucleic acid sequences.
[0043] In some embodiments of the invention, the invention relates to a method of producing and/or secreting a lipid or mixture of lipids by culturing a population of microorganisms comprising a bacterial cell described herein, wherein the exogenous nucleic acid sequences are operably linked to a promoter that is inducible in response to a first stimulus, and wherein the method further comprises: culturing the population of bacterial cells for a first period of time in the presence of a first stimulus to produce one or more lipids chosen from an alkene, alkyne, unsaturated fatty acid, hydroxyacid and/or diacid.
[0044] In some embodiments, the invention relates to a method of producing one or more diacids, comprising exposing a bacterial cell to syngas and/or gaseous CO2 or a mixture of gaseous CO2 and gaseous H2; wherein the bacterial cell is capable of fixing gaseous CO2 into one or more diacids, and wherein the microorganism comprises at least a first exogenous nucleic acid and a second exogenous nucleic acid, wherein the first exogenous nucleic acid encodes cytochrome P450 enzyme and the second exogenous nucleic acid encodes NADPH cytochrome P450 reductase or NADH cytochrome P450 reductase. In some embodiments, the first and second exogenous nucleic acids are heterologous nucleic acid sequences. In some embodiments, the bacterial cell comprises at least a first, a second, and a third exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes a cytochrome P450 enzyme, the second exogenous nucleic acid sequence encodes a NADPH cytochrome P450 reductase, and the third exogenous nucleic acid sequence encodes a thioesterase. In some embodiments, the bacterial cell comprises at least a first exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes a thioesterase. In some embodiments, the bacterial cell comprises no more than five exogenous nucleic acid sequences that encode a lipid pathway enzyme. In some embodiments, the composition comprises a microorganism, wherein the microorganism is Rhodococcus opacus (DSM 43205 or 43206) or Rhodococcus sp (DSM 3346). In some embodiments, the composition comprises a microorganism, wherein the microorganism is chosen from the genera Ralstonia or Cupriavidus. In some embodiments, the composition comprises a microorganism, wherein the microorganism is Cupriavidus necator. In some embodiments the microorganism is from the suborder corynebacterineae or the family burkholderiaceae. In some embodiments the microorganism through its native machinery produces a complement of fatty acids described in the Fatty Acid Output section below.
[0045] In some embodiments, the invention relates to a method of producing one or more hydroxyacid, diacid, or unsaturated fatty acid, or any combination thereof comprising exposing a bacterial cell to syngas and/or gaseous CO2 or a mixture of gaseous CO2 and gaseous H2; wherein the bacterial cell is capable of fixing gaseous CO2 into one or more lipids; wherein the lipids are recovered from the bioreactor and fed to a second bioreactor wherein the lipids are postprocessed to generate hydroxyacid, diacid, and/or unsaturated fatty acids via a second microorganism such as but not limited to Candida tropicalis. In some embodiments, the bacterial cell comprises at least a first exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes a thioesterase. In some embodiments, the composition comprises a bacterial cell, wherein the bacteria is Rhodococcus opacus (DSM 43205 or 43206) or Rhodococcus sp (DSM 3346). In some embodiments, the bacterial cell is chosen from the genera Ralstonia or Cupriavidus. In some embodiments, the bacterial cell is Cupriavidus necator. In some embodiments the bacterial cell is from the suborder corynebacterineae or the family burkholderiaceae. In some embodiments the bacterial cell through its native machinery produces a complement of fatty acids described in the Fatty Acid Output section below.
[0046] In some embodiments, the invention relates to a method of producing one or more unsaturated fatty acids, comprising exposing a bacterial cell to syngas and/or gaseous CO2 or a mixture of gaseous CO2 and gaseous H2; wherein the bacterial cell is capable of fixing gaseous CO2 into one or more unsaturated fatty acids and wherein the microorganism comprises at least a first exogenous nucleic acid, wherein the first exogenous nucleic acid encodes a desaturase that introduces double bonds to fatty acids. In some embodiments, the first exogenous nucleic acids is a heterologous nucleic acid sequence. In some embodiments, the bacterial cell comprises at least a first, and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes a desaturase, the second exogenous nucleic acid sequence encodes a thioesterase. In some embodiments, the composition comprises a microorganism, wherein the microorganism is Rhodococcus opacus (DSM 43205 or 43206) or Rhodococcus sp (DSM 3346). In some embodiments, the composition comprises a microorganism, wherein the microorganism is chosen from the genera Ralstonia or Cupriavidus. In some embodiments, the composition comprises a microorganism, wherein the microorganism is Cupriavidus necator. In some embodiments the microorganism is from the suborder corynebacterineae or the family burkholderiaceae. In some embodiments the microorganism through its native machinery produces a complement of fatty acids described in the Fatty Acid Output section below.
[0047] In some embodiments, the invention relates to a method of producing one or more hydroxy fatty acids (hydroxy acids), comprising exposing a bacterial cell to syngas and/or gaseous CO2 or a mixture of gaseous CO2 and gaseous H2; wherein the bacterial cell is capable of fixing gaseous CO2 into one or more hydroxy acids and wherein the microorganism comprises at least a first exogenous nucleic acid, wherein the first exogenous nucleic acid encodes a P450-dependent fatty acid hydroxylase that introduces hydroxyl groups at positions along the fatty acid chain. In some embodiments, the first exogenous nucleic acids is a heterologous nucleic acid sequence. In some embodiments, the bacterial cell comprises at least a first, and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes a P450-dependent fatty acid hydroxylase, the second exogenous nucleic acid sequence encodes a thioesterase. In some embodiments, the composition comprises a microorganism, wherein the microorganism is Rhodococcus opacus (DSM 43205 or 43206) or Rhodococcus sp (DSM 3346). In some embodiments, the composition comprises a microorganism, wherein the microorganism is chosen from the genera Ralstonia or Cupriavidus. In some embodiments, the composition comprises a microorganism, wherein the microorganism is Cupriavidus necator. In some embodiments the microorganism is from the suborder corynebacterineae or the family burkholderiaceae. In some embodiments the microorganism through its native machinery produces a complement of fatty acids described in the Fatty Acid Output section below.
[0048] In some embodiments, the invention relates to a method of manufacturing one or more lipids, comprising (a) culturing a cell described herein in a reaction vessel or bioreactor in the presence of syngas and/or gaseous CO2 or a mixture of gaseous CO2 and gaseous H2, wherein the cell produces and/or secretes one or more lipids in an quantity equal to or greater than at least 10% of the cell's total dry cellular mass; and (b) separating the one or more lipids from reaction vessel. In some embodiments, the method further comprises purifying the one or more lipids after separation from the reaction vessel or bioreactor.
[0049] In some embodiments, the invention relates to a method of producing a alkene or alkyne in a bacterial cell comprising at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes a fatty acid acyl-ACP reductase and the second exogenous nucleic acid encodes a fatty acid aldehyde decarbonylase.
[0050] In some embodiments, the bacterial cell producing a alkene or alkyne comprises at least a first, a second, and a third exogenous nucleic acid sequences, wherein the first exogenous nucleic acid sequence encodes a fatty acid acyl-ACP reductase and the second exogenous nucleic acid encodes a fatty acid aldehyde decarbonylase, and the third exogenous nucleic acid encodes a thioesterase.
[0051] In some embodiments, the invention relates to a bioreactor comprising the composition or bacterial cells described herein.
[0052] In some embodiments, the invention relates to a system for the production of one or more lipids or mixture of lipids, comprising a bioreactor, which comprises: (a) a microorganism population comprising a cell described herein; and (b) an inlet connected to a feedstock source allowing delivery of a feedstock comprising syngas and/or gaseous CO2 or a mixture of gaseous CO2 and gaseous H2.
[0053] In some embodiments, the invention relates to the population of fatty acids being modified to produce molecules of desired carbon chain length by incorporation of one or more thioesterases.
[0054] In some embodiments, the invention relates to the population of fatty acids being modified to add additional carboxylic acid (--COOH) groups using exogenous enzymes.
[0055] In some embodiments, the invention relates to the population of fatty acids being modified to add hydroxyl groups (--OH) using the exogenous enzymes (hydroxylases).
[0056] In some embodiments, the invention relates to the population of fatty acids being modified to add desaturation through the incorporation of one or more double bonds, using the exogenous enzymes (desaturases).
BRIEF DESCRIPTION OF THE DRAWINGS
[0057] Non-limiting embodiments of the present invention will be described by way of example with reference to the accompanying figures, which are schematic and are not intended to be drawn to scale. For purposes of clarity, not every component is labeled in every figure, nor is every component of each embodiment of the invention shown where illustration is not necessary to allow those of ordinary skill in the art to understand the invention. In the figures:
[0058] FIG. 1 describes the taxonomic names afforded to the chemoautotrophic and oleaginous microorganisms used in selected embodiments of the invention.
[0059] FIG. 2 shows the 16S rRNA gene based-rooted phylogenetic tree of gordoniaceae, mycobacteriaceae, nocardiaceae and burkholderiaceae.
[0060] Bar, 0.01% estimated sequence divergence.
[0061] FIG. 3 shows the sequence similarity of Rhodococcus opacus (DSM 43205) 16S rRNA gene (NR_026186.1) to members of the family gordoniaceae, mycobacteriaceae, nocardiaceae and burkholderiaceae. The Genbank accession numbers, DNA length and % identity of analyzed genes are indicated.
[0062] FIG. 4 describes the nucleotide sequence alignment of the 16S rRNA genes.
[0063] FIG. 5 demonstrates the growth of chemotrophic and oleaginous microorganisms on different carbon sources. Bacterial growth was measured using optical density (OD) detection at 650 nm after the indicated days (in parentheses). Media and growth conditions described in the Examples section below. ND, not done.
[0064] FIG. 6 describes the measured lipid content of microorganisms on heterotrophic and chemoautotrophic growth conditions as a percentage of total cellular dry matter (CDM). Cells were grown under conditions described in FIG. 5, harvested after 72 hr (unless otherwise indicated) and analyzed by gas chromatography. For CDM, total dry weight was determined gravimetrically.
[0065] FIG. 7 describes the fatty acid profile of R. opacus (DSM 44193) under heterotrophic growth conditions. Cells were harvested after 72 hr and analyzed by gas chromatography.
[0066] FIG. 8 describes the fatty acid profile R. opacus (DSM43205) under heterotrophic (A) and chemoautotrophic (B) growth conditions. Cells were harvested after 72 hours of growth and analyzed by gas chromatography.
[0067] FIG. 9 describes the fatty acid profile Rhodococcus sp. (DSM 3346) under heterotrophic (A) chemoautotrophic (B) growth conditions. Cells were harvested after 72 hr and analyzed by gas chromatography.
[0068] FIG. 10 describes shuttle vectors (A) and genetic elements (B) for transformation and gene expression of in chemoautotrophic and oleaginous microorganisms. MCS: multiple cloning site.
[0069] FIG. 11 describes the map of the plasmids pSeqCO1 (A; SEQ ID: 01), pSeqCO2 (B; SEQ ID: 02), pVer1 (C; SEQ ID: 03) and pVer2 (D; SEQ ID: 04) described in FIG. 10. The genetic elements are indicated.
[0070] FIG. 12 describes the transformation of chemoautotrophic and oleaginous microorganisms with shuttle vectors described in FIG. 10.
[0071] FIG. 13 describes the growth of Cupriavidus necator (DSM531) transformed with the plasmid (Y) pSeqCO2 (SEQ ID:2) and untransformed (N) on different kanamycin concentrations. Single colony of transformants and control were grown LB medium (per 1 L: 10 g Bacto-tryptone, 5 g yeast extract, 10 g NaCl pH=7.0) at 30.degree. C. in the indicated kanamycin concentrations. The growth was measured using O.D650 after the indicated number of days.
[0072] FIG. 14 describes the restriction map related to the cloning strategy of FadDR and Fad genes into pSeqCO2 plasmid transformed for the experiment. Genes from the cyanobacterium (Synechocystis sp. PCC 6803) used in the experiment were FadR (SEQ ID: 08) and FAD (SEQ ID: 09) driven by the Synechocystis sp. Rubisco large subunit promoter (SEQ ID: 10) were cloned into pSeqCO2 plasmid using the indicated restriction sites to give pSeqCO2::FUEL.
[0073] FIG. 15 describes the production of hydrocarbons in Cupriavidus necator transformed with pSeqCO2::FUEL (Cn_FUEL2.1) and empty vector (Cn-P). GC chromatogram of hydrocarbon (indicated in red) extracted from transformants grown in 50 ml LB media under previously identified conditions.
[0074] FIG. 16 describes the hydrocarbons specific products and distribution (percentage in parentheses) from Cupriavidus necator transformed with pSeqCO2::FUEL (Cn_FUEL2.1 and Cn_FUEL2.2) and empty vector (Cn-P).
[0075] FIG. 17 shows the molecular structure of two hydrocarbons that were produced in relative abundance in the Cupriavidus necator transformed with pSeqCO2::FUEL (Cn_FUEL2.1 and Cn_FUEL2.2).
[0076] FIG. 18 describes the effect of pSeqCO2::FUEL (Cn_FUEL2.1 and 2.2) and empty vector (Cn-P) on the fatty acids distribution under the experimental conditions described previously.
[0077] FIG. 19 describes the modification of the fatty acid chain length by the enzymatic action of thioesterase (TE) in oleaginous bacteria.
[0078] FIG. 20 describes the modification of the fatty acid chain length by the enzymatic action of fatty acyl-ACP thioesterase (TE) in burkholderiaceae.
[0079] FIG. 21 describes the similarity of Rhodococcus opacus (B4) thioesterases protein sequence (YP_002784058.1) to other organisms. The Genbank accession numbers, amino acid length and % identity of analyzed proteins are indicated.
[0080] FIG. 22 describes the chemoautotrophic growth of Cupriavidus necator transformed with pSeqCO2::FUEL (Cn-FUEL2.1), empty vector (Cn-P) and untransformed (Cn). Bacterial growth was measured at O.D650 after 12 days.
DETAILED DESCRIPTION
[0081] Various terms relating to the methods and other aspects of the present invention are used throughout the specification and claims. Such terms are to be given their ordinary meaning in the art unless otherwise indicated. Other specifically defined terms arc to be construed in a manner consistent with the definition provided herein.
[0082] As used in this specification and the appended claims, the singular forms "a," "an," and "the" include plural referents unless the content clearly dictates otherwise.
[0083] The term "about" as used herein when referring to a measurable value such as an amount, a temporal duration, and the like, is meant to encompass variations of +20%, .+-.10%, +5%, +1%, or +0.1% from the specified value, as such variations are appropriate to perform the disclosed methods.
[0084] The terms "amino acid" refer to a molecule containing both an amine group and a carboxyl group that are bound to a carbon, which is, designated the .alpha.-carbon. Suitable amino acids include, without limitation, both the D- and L-isomers of the naturally occurring amino acids, as well as non-naturally occurring amino acids prepared by organic synthesis or other metabolic routes. In some embodiments, a single "amino acid" might have multiple sidechain moieties, as available per an extended aliphatic or aromatic backbone scaffold. Unless the context specifically indicates otherwise, the term amino acid, as used herein, is intended to include amino acid analogs.
[0085] The term "biomass" refers to a material produced by growth and/or propagation of cells. Biomass may contain cells and/or intracellular contents as well as extracellular material, includes, but is not limited to, compounds secreted by a cell.
[0086] The term "bioreactor" or "fermentor" refers to a closed or partially closed vessel in which cells are grown and maintained. The cells may be, but are not necessarily held in liquid suspension. In some embodiments rather than being held in liquid suspension, cells may alternatively be growing and/or maintained in contact with, on, or within another non-liquid substrate including but not limited to a solid growth support material.
[0087] The term "catalyst" refers to a chemical actor, such as a molecule or macromolecular structure, which accelerates the speed at which a chemical reaction occurs where a reactant or reactants is converted into a product or products, while the catalyst is not turned into a product itself, or otherwise changed or consumed at the completion of the chemical reaction. After a catalyst participates in one chemical reaction, because it is unchanged, it may participate in further chemical reactions, acting on additional reactants to create additional products. To accelerate a chemical reaction a catalyst decreases the activation energy barrier across the reaction path allowing it to occur at a colder temperature, or faster at a given temperature. In this way a more rapid approach of the system to chemical equilibrium may be achieved. Catalysts subsume enzymes, which are protein catalysts.
[0088] The term "cellulosic material" refers to any material with a high amount of cellulose, which is a polysaccharide having the formula (C6H10O5)n, that generally consists of a linear chain of hundreds to thousands of .beta.(1.fwdarw.4) linked D-glucose monomers. Sources of cellulosic material include but are not limited to cardboard, cotton, corn stover, paper, lumber chips, sawdust, sugar beet pulp, sugar cane bagasses, and switchgrass.
[0089] The term "CoA" or "coenzyme A" refers to an organic cofactor for condensing enzymes involved in fatty acid synthesis and oxidation, pyruvate oxidation, acetyl or other acyl group transfer, and in other acetylation.
[0090] The term "cofactor" subsumes all molecules needed by an enzyme to perform its catalytic activity. In some embodiments, the cofactor is any molecule apart from the substrate.
[0091] A "conservative amino acid substitution" is one in which the amino acid residue is replaced with an amino acid residue having a similar side chain. Families of amino acid residues having similar side chains have been defined in the art. These families include amino acids with basic side chains (e.g., K, R, H), acidic side chains (e.g., D, E), uncharged polar side chains (e.g., G, N, Q, S, T, Y, C, H), nonpolar side chains (e.g., G, A, V, L, I, P, F, M, W), beta-branched side chains (e.g., T, V, I) and aromatic side chains (e.g., Y, F, W, H). Thus, a predicted nonessential amino acid residue in an amino acid sequence encoded by an exogenous nucleic acid sequence, for example, is replaced with another amino acid residue from the same side chain family. Other examples of acceptable substitutions are substitutions based on isosteric considerations (e.g. norleucine for methionine) or other biochemical properties (e.g. 2-thienylalanine for phenylalanine).
[0092] As used herein, "enzyme fragment" is meant to refer to a fragment of an enzyme that includes the sequences sufficient to function substantially similar to the function of the wild-type enzyme upon which the fragment sequence is based. Fragments are generally 10 or more amino acids in length. Some preferred lengths of fatty acid reductase arc at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, at least 85, at least 90, at least 95, at least 100, at least 105, at least 110, at least 115, at least 120, at least 125, at least 130, at least 135, at least 140, at least 145, at least 150, at least 155, at least 160, at least 165, at least 170, at least 175, at least 180, at least 185, at least 190, at least 195, at least 200, at least 205, at least 210 at least 215, at least 220, at least 225, least 230 at least 235, at least 240, at least 245, at least 250, at least 255, at least 260, at least 265, at least 270, at least 275, at least 280, at least 285, at least 290, at least 295, at least 300, at least 305, at least 310, at least 315, at least 320, at least 325, at least 330, at least 335, at least 340, at least 345, at least 350, at least 355, at least 360, at least 365, at least 370, at least 375, at least 380, at least 385, at least 390, at least 395, at least 400, at least 405, at least 410, at least 415, at least 420, at least 425, or at least 430. Some preferred lengths of fatty acid reductase fragments are 15 or fewer, 20 or fewer, 25 or fewer, 30 or fewer, 35 or fewer, 40 or fewer, 45 or fewer, 50 or fewer, 55 or fewer, 60 or fewer, 65 or fewer, 70 or fewer, 75 or fewer, 80 or fewer, 85 or fewer, 90 or fewer, 95 or fewer, 100 or fewer, 105 or fewer, 110 or fewer, 115 or fewer, 120 or fewer, 125 or fewer, 130 or fewer, 135 or fewer, 140 or fewer, 145 or fewer, 150 or fewer, 155 or fewer, 160 or fewer, 165 or fewer, 170 or fewer, 175 or fewer, 180 or fewer, 185 or fewer, 190 or fewer, 195 or fewer, 200 or fewer, 205 or fewer, 210 or fewer, 215 or fewer, 220 or fewer, 225 or fewer, 230 or fewer, 235 or fewer, 240 or fewer, 245 or fewer, 250 or fewer, 255 or fewer, 260 or fewer, 265 or fewer, 270 or fewer, 275 or fewer, 280 or fewer, 285 or fewer, 290 or fewer, 295 or fewer, 300 or fewer, 305 or fewer, 310 or fewer, 315 or fewer, 320 or fewer, 325 or fewer, 330 or fewer, 335 or fewer, 340 or fewer, 345 or fewer, 350 or fewer, 355 or fewer, 360 or fewer, 365 or fewer, 370 or fewer, 375 or fewer, 380 or fewer, 385 or fewer, 390 or fewer, 395 or fewer, 400 or fewer, 415 or fewer, 420 or fewer, 425 or fewer, 430 or fewer, or 435 or fewer. Some preferred lengths of fatty acid decarbonylase are at least 10, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, at least 85, at least 90, at least 95, at least 100, at least 105, at least 110, at least 115, at least 120, at least 125, at least 130, at least 135, at least 140, at least 145, at least 150, at least 155, at least 160, at least 165, at least 170, at least 175, at least 180, at least 185, at least 190, at least 195, at least 200, at least 205, at least 210 at least 215, at least 220, at least 225, least 230 at least 235, at least 240, at least 245, at least 250, at least 255, at least 260, at least 265, at least 270, at least 275, at least 280, at least 285, at least 290, at least 295, at least 300, at least 305, at least 310, at least 315, at least 320, at least 325, at least 330, at least 335, at least 340, at least 345, at least 350, at least 355, at least 360, at least 365, at least 370, at least 375, at least 380, at least 385, at least 390, at least 395, at least 400, at least 405, at least 410, at least 415, or at least 420 amino acids long. In some embodiments, the lengths of the fatty acid decarbonylase fragments are 15 or fewer, amino acids, 20 or fewer, 25 or fewer, 30 or fewer, 35 or fewer, 40 or fewer, 45 or fewer, 50 or fewer, 55 or fewer, 60 or fewer, 65 or fewer, 70 or fewer, 75 or fewer, 80 or fewer, 85 or fewer, 90 or fewer, 95 or fewer, 100 or fewer, 105 or fewer, 110 or fewer, 115 or fewer, 120 or fewer, 125 or fewer, 130 or fewer, 135 or fewer, 140 or fewer, 145 or fewer, 150 or fewer, 155 or fewer, 160 or fewer, 165 or fewer, 170 or fewer, 175 or fewer, 180 or fewer, 185 or fewer, 190 or fewer, 195 or fewer, 200 or fewer, 205 or fewer, 210 or fewer, 215 or fewer, 220 or fewer, 225 or fewer, 230 or fewer, 235 or fewer, 240 or fewer, 245 or fewer, 250 or fewer, 255 or fewer, 260 or fewer, 265 or fewer, 270 or fewer, 275 or fewer, 280 or fewer, 285 or fewer, 290 or fewer, 295 or fewer, 300 or fewer, 305 or fewer, 310 or fewer, 315 or fewer, 320 or fewer, 325 or fewer, 330 or fewer, 335 or fewer, 340 or fewer, 345 or fewer, 350 or fewer, 355 or fewer, 360 or fewer, 365 or fewer, 370 or fewer, 375 or fewer, 380 or fewer, 385 or fewer, 390 or fewer, 395 or fewer, 400 or fewer, 415 or fewer, 422 or fewer. Some preferred lengths of thioesterase fragments are at least 10 amino acids, at least 15, at least 20, at least 25, at least 30, at least 35, at least 40, at least 45, at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, at least 80, at least 85, at least 90, at least 95, at least 100, at least 105, at least 110, at least 115, at least 120, at least 125, at least 130, at least 135, at least 140, at least 145, at least 150, at least 155, at least 160, at least 165, at least 170, at least 175, at least 180, at least 185, at least 190, at least 195, at least 200, at least 205, at least 210 at least 215, at least 220, at least 225, least 230 at least 235, at least 240, at least 245, at least 250 or at least 255. Some preferred lengths of thioesterase fragments are 15 or fewer, 20 or fewer, 25 or fewer, 30 or fewer, 35 or fewer, 40 or fewer, 45 or fewer, 50 or fewer, 55 or fewer, 60 or fewer, 65 or fewer, 70 or fewer, 75 or fewer, 80 or fewer, 85 or fewer, 90 or fewer, 95 or fewer, 100 or fewer, 105 or fewer, 110 or fewer, 115 or fewer, 120 or fewer, 125 or fewer, 130 or fewer, 135 or fewer, 140 or fewer, 145 or fewer, 150 or fewer, 155 or fewer, 160 or fewer, 165 or fewer, 170 or fewer, 175 or fewer, 180 or fewer, 185 or fewer, 190 or fewer, 195 or fewer, 200 or fewer, 205 or fewer, 210 or fewer, 215 or fewer, 220 or fewer, 225 or fewer, 230 or fewer, 235 or fewer, 240 or fewer, 245 or fewer, 250 or fewer, 255 or fewer or 260 or fewer. As used in the paragraph herein reference to preferred fragment sizes are intended to refer to all permutation of ranges between at least and less than such as ranges may be any number set forth as an "at least" size to any number set forth as an "less than t" size in order to provide a range of sizes such as 20-400, 20-30, 40-100, etc.
[0093] The terms "exogenous gene" means a nucleic acid that has been recombinantly introduced into a cell, which encodes the synthesis of RNA and/or protein. In some embodiments, the exogenous gene is introduced by transformation. In some embodiments, the exogenous gene is introduced into the cell by electroporation. A transformed cell may be referred to as a recombinant cell, into which additional exogenous gene(s) may be introduced. The exogenous gene put into the host species may be taken from a different species (this is called heterologous), or it may naturally occur within the same species (this is homologous as defined below). Therefore, exogenous genes subsume homologous genes that are integrated within or introduced to regions of the genome, episome, or plasmid that differ from the locations where the gene naturally occurs. Multiple copies of the exogenous gene may be introduced into the cell. An exogenous gene may be present in more than one copy within the host cell or transformed cell. In some embodiments, the microorganism comprises between and including 1 and 10,000 copies of the nucleic acid that encodes an exogenous protein. In some embodiments, the microorganism comprises between and including 1 and 1,000 copies of the nucleic acid that encodes an exogenous protein. In some embodiments, the microorganism comprises between and including 1 and 10,000 copies of the nucleic acid that encodes an exogenous protein. In some embodiments, the microorganism comprises between and including 1 and 1,000 copies of the nucleic acid that encodes an exogenous protein. In some embodiments, the microorganism comprises between and including 1 and 500 copies of the nucleic acid that encodes an exogenous protein. In some embodiments, the exogenous gene is maintained by a cell as an insertion into the genome or as an episomal molecule. In some embodiments, the microorganism comprises no more than 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, or 1000 copies of the one or more nucleic acids that encode one or more exogenous proteins.
[0094] As used herein, the term "expressible form" refers to gene constructs that contain the necessary regulatory elements operably linked to a coding sequence that encodes an enzyme or fragment thereof capable of conferring enzymatic activity to a cell, such that when present in the cell, the coding sequence will be expressed. In some embodiments of the invention, the composition comprising the microorganisms or bacterial cells of the present invention comprise no more than ten expressible forms of exogenous nucleic acid sequences. In some embodiments of the invention, the composition comprising the microorganisms or bacterial cells of the present invention comprise no more than nine expressible forms of exogenous nucleic acid sequences. In some embodiments of the invention, the composition comprising the microorganisms or bacterial cells of the present invention comprise no more than eight expressible forms of exogenous nucleic acid sequences. In some embodiments of the invention, the composition comprising the microorganisms or bacterial cells of the present invention comprise no more than seven expressible forms of exogenous nucleic acid sequences. In some embodiments of the invention, the composition comprising the microorganisms or bacterial cells of the present invention comprise no more than six expressible forms of exogenous nucleic acid sequences. In some embodiments of the invention, the composition comprising the microorganisms or bacterial cells of the present invention comprise no more than five expressible forms of exogenous nucleic acid sequences. In some embodiments of the invention, the composition comprising the microorganisms or bacterial cells of the present invention comprise no more than four expressible forms of exogenous nucleic acid sequences. In some embodiments of the invention, the composition comprising the microorganisms or bacterial cells of the present invention comprise no more than three expressible forms of exogenous nucleic acid sequences. In some embodiments of the invention, the composition comprising the microorganisms or bacterial cells of the present invention comprise no more than two expressible forms of exogenous nucleic acid sequences. In some embodiments of the invention, the composition comprising the microorganisms or bacterial cells of the present invention comprise no more than one expressible form of an exogenous nucleic acid sequences.
[0095] SEQ ID NO:1 refers to Sequesco plasmid sequence 1.
[0096] SEQ ID NO:2 refers to Sequesco plasmid sequence 2.
[0097] SEQ ID NO: 3 refers to Sequesco plasmid Ver1 plasmid sequence.
[0098] SEQ ID NO:4 refers to Sequesco plasmid Ver2 plasmid sequence.
[0099] SEQ ID NO:5 refers to cyanobacterium FadR.
[0100] SEQ ID NO:6 refers to cyanobacterium FAD.
[0101] SEQ ID NO:7 refers to cyanobacterium Rubisco large subunit promoter
[0102] SEQ ID NO:8, refers to the 16S rRNA sequence from the genus Rhodococcus opacus DSM43205
[0103] SEQ ID NO:9 refers to the 16S rRNA sequence from the genus Rhodococcus opacus B4.
[0104] SEQ ID NO:10 refers to the 16S rRNA sequence from the genus Ralstonia.
[0105] SEQ ID NO:11 refers to Rhodococcus opacus TE
[0106] The terms "fatty acyl-ACP thioesterase" (TE) mean an enzyme that catalyzes the cleavage of a fatty acid from an acyl carrier protein (ACP) during lipid synthesis.
[0107] The terms "fatty acyl-ACP/acyl-CoA reductase" (FadR) refers to an enzyme catalyzing the reaction that produces a fatty aldehyde from an acyl-ACP or acyl-CoA molecule by reduction.
[0108] The terms "fatty aldehyde decarbonylase" (FAD) refers to an enzyme catalyzing the reaction that produces an alkane from a fatty aldehyde molecule by decarbonylization.
[0109] As used herein, the term "functional fragment" is meant to refer to a fragment of any polypeptide or amino acid sequence that is encoded by an exogenous nucleic acid sequence of the present invention which retains its ability to function like the amino acid sequence to which the fragment is homologous. Functional fragments of enzymes are at least about 5 amino acids in length derived from enzyme and may comprise non-wild-type amino acid sequences. One having ordinary skill in the art can readily determine whether a protein or peptide is a functional fragment of a particular amino acid sequence by examining its sequence and testing its ability to function in a fashion similar to that function of the amino acid sequence upon which the fragment is based. Truncated versions of exogenous proteins may be prepared and tested using routine methods and readily available starting material. As used herein, the term "functional fragment" is also meant to refer to peptides, polypeptides, amino acid sequence linked by non-peptidal bonds, or proteins which comprise an amino acid sequence that is identical or substantially homologous to at least a portion of the exogenous amino acid sequence and which are capable of functioning in a similar function to the exogenous amino acid sequence to which the fragment is homologous. The term "substantially homologous" refers to an amino acid sequence that has conservative substitutions. One having ordinary skill in the art can produce functional fragments of the FadD, FAD, thioesterase, cytochrome P450 enzyme, desaturase, and hydroxylase amino acid sequences following the disclosure provided herein and well known techniques. The functional fragments thus identified may be used and formulated in place of full length FadD, FAD, thioesterase, cytochrome P450 enzyme, desaturase, and hydroxylase without undue experimentation.
[0110] The term "gasification" refers to a generally high temperature (>700.degree. C.) process that converts carbonaceous materials into a mixture of gases including hydrogen, carbon monoxide, and carbon dioxide called syngas or producer gas. The process generally involves partial combustion and/or the application of externally generated heat along with the controlled addition of oxygen and/or steam.
[0111] As used herein, "homologous" refers to the sequences homology between two nucleic acid sequences or two amino acid sequences. Two nucleic acid sequences or two amino acid sequences that are sufficiently homologous to retain immunogenic function are "homologues." Sequence homology for nucleotides and amino acids may be determined using FASTA, BLAST and Gapped BLAST (Altschul et al., Nuc. Acids Res., 1997, 25, 3389, which is incorporated herein by reference in its entirety) and PAUP* 4.0b10 software (D. L. Swofford, Sinauer Associates, Massachusetts). "Percentage of similarity" is calculated using PAUP* 4.0b10 software (D. L. Swofford, Sinauer Associates, Massachusetts). The average similarity of the enzymatic sequence or 16S rRNA sequence is calculated compared to all sequences in the phylogenic tree. Briefly, the BLAST algorithm, which stands for Basic Local Alignment Search Tool is suitable for determining sequence similarity (Altschul et al., J. Mol. Biol., 1990, 215, 403410, which is incorporated herein by reference in its entirety). Software for performing BLAST analyses is publicly available though the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/). This algorithm involves first identifying high scoring sequence pair (HSPs) by identifying short words of length W in the query sequence that either match or satisfy some positive-valued threshold score T when aligned with a word of the same length in a database sequence. T is referred to as the neighborhood word score threshold (Altschul et al., supra). These initial neighborhood word hits act as seeds for initiating searches to find HSPs containing them. The word hits are extended in both directions along each sequence for as far as the cumulative alignment score can be increased. Extension for the word hits in each direction arc halted when: 1) the cumulative alignment score falls off by the quantity X from its maximum achieved value; 2) the cumulative score goes to zero or below, due to the accumulation of one or more negative-scoring residue alignments; or 3) the end of either sequence is reached. The Blast algorithm parameters W, T and X determine the sensitivity and speed of the alignment. The Blast program uses as defaults a word length (W) of 11, the BLOSUM62 scoring matrix (see Henikoff et al., Proc. Natl. Acad. Sci. USA, 1992, 89, 10915-10919, which is incorporated herein by reference in its entirety) alignments (B) of 50, expectation (E) of 10, M=5, N=4, and a comparison of both strands. The BLAST algorithm (Karlin et al., Proc. Natl. Acad. Sci. USA, 1993, 90, 5873-5787, which is incorporated herein by reference in its entirety) and Gapped BLAST perform a statistical analysis of the similarity between two sequences. One measure of similarity provided by the BLAST algorithm is the smallest sum probability (P(N)), which provides an indication of the probability by which a match between two nucleotide sequences would occur by chance. For example, a nucleic acid is considered similar to another if the smallest sum probability in comparison of the test nucleic acid to the other nucleic acid is less than about 1, preferably less than about 0.1, more preferably less than about 0.01, and most preferably less than about 0.001.
[0112] The term "hydrocarbon" refers to a molecule composed exclusively of carbon and hydrogen atoms with the carbons bonded covalently in a branched, cyclic, linear, or partially cyclic chain and with hydrogen atoms covalently bonded to the carbons such that the chemical octet rule for the carbons is generally satisfied. In some hydrocarbons there may occur some number of double or triple bonds between adjacent carbon atoms in the chain. Thus, the label hydrocarbon subsumes branched, cyclic, linear, branched, or partially cyclic alkanes (also called paraffins), alkenes (also called olefins), and alkynes. The structure of hydrocarbon molecules range from the smallest, methane (CH4), a primary component of natural gas, to high molecular weight complex molecules including asphaltenes present in bitumens crude oil, and petroleum. Other examples include dodecane (C12), hexadecane (C16), or octadecane (C18) etc. Hydrocarbons of the present invention may be in gaseous, liquid, or solid phases, either as singly or in multiply coexisting phases. In some embodiments, the hydrocarbons are selected from one or more of the following: linear, branched, cyclic, or partially cyclic alkanes, alkenes, alkynes, lipids, and paraffin. In some embodiments the hydrocarbon are selected from one or more of the following: octane, squalene Spiro[4.5]decane, Bicyclo[10.8.0]eicosane, cis,cis-1,6-Dimethylspiro[4.5]decane, 1,19-Eicosadiene, Cyclooctacosane, Bicyclo[10.8.0]eicosane, 1-Pentadecyne, 1-Pentadecyne, Heptacosyl acetate, 5-Cyclohexyl-1-pentene, 1-Hexadecyne and Cyclodecacyclotetradecene, -eicosahydro.
[0113] The term "hydrophobic fraction" gives the fraction of matter that has low solubility in water and greater solubility in a hydrophobic phase than in an aqueous phase. In some embodiments, the hydrophobic fraction is non-polar. In some embodiments, the genetically modified bacterial cells described herein increase the hydrophobic fraction of hydrocarbons in a cell as compared to the same cell that is not genetically modified.
[0114] The term "improve lipid yield" refers to an increase in the lipid production of an organism through any means. In some embodiments, the increase is caused by raising the cell dry weight density of a microbial culture and/or raising the fraction of cell mass that is composed of lipid and/or reducing the cell doubling time and/or the biomass doubling time, resulting in an overall increase in the lipid production rate per unit volume.
[0115] The term "knallgas" refers to the mixture of molecular hydrogen and oxygen gas. A "knallgas microorganism" is a microbe that can use hydrogen as an electron donor and oxygen as an electron acceptor in the generation of intracellular energy carriers such as Adenosine-5'-triphosphate (ATP). The terms "oxyhydrogen" and "oxyhydrogen microorganism" can be used synonymously with "knallgas" and "knallgas microorganism" respectively.
[0116] The term "lignocellulosic material" is any material composed of cellulose, hemicellulose, and lignin where the carbohydrate polymers (cellulose and hemicelluloses) are tightly bound to lignin. Lignocellulosic materials subsume agricultural residues (including corn stover and sugarcane bagasse), most biomass energy crops, wood residues (including sawmill and paper mill discards), and a substantial fraction of municipal waste.
[0117] The terms "lipids" refers to category of molecules that can be dissolved in nonpolar solvents (such as chloroform and/or ether) and which also have low or no solubility in water. The hydrophobic character of lipids molecules typically results from the presence of long chain hydrocarbon sections within the molecule. Lipids subsume the following molecule types: hydrocarbons, fatty acids (saturated and unsaturated), fatty alcohols, fatty aldehydes, hydroxy acids, diacids, monoglycerides, diglycerides, triglycerides, phospholipids, sphingolipids, sterols such as cholesterol and steroid hormones, fat-soluble vitamins (such as vitamins A, D, E and K), polyketides, terpenoids, and waxes.
[0118] The term "lipid modification enzyme" corresponds to an enzyme that catalyzes a reaction changing a lipid's covalent bonds such as TE, FAR, FadR, FAD, fatty aldehyde reductase, lipase, cytochrome P450 enzyme, desaturase, or hydroxylase. Any enzyme that catalyzes a reaction step or steps in lipid synthesis, catabolism, or modification, including carrier proteins, is called a "lipid pathway enzyme".
[0119] The term "lysate" refers to the liquid containing a mixture and/or a solution of cell contents that result from cell lysis. In some embodiments, the methods of the present invention comprise a purification of hydrocarbons or mixture of hydrocarbons in a cellular lysate. In some embodiments, the methods of the present invention comprise a purification of lipids and/or hydrocarbons and/or a mixture of hydrocarbons in a cellular lysate.
[0120] The term "lysis" refers to the rupture of the plasma membrane and if present the cell wall of a cell such that a significant amount of intracellular material escapes to the extracellular space. Lysis can be performed using electrochemical, mechanical, osmotic, thermal, or viral means. In some embodiments, the methods of the present invention comprise performing a lysis of cells or microorganisms described herein in order to separate a hydrocarbon or mixture of hydrocarbons from the contents of a bioreactor. In some embodiments, the methods of the present invention comprise performing a lysis of cells or microorganisms described herein in order to separate a lipid or mixture of lipids from the contents of a bioreactor.
[0121] The terms "microorganism" and "microbe" mean microscopic single celled life forms.
[0122] The term "molecule" means any distinct or distinguishable structural unit of matter comprising one or more atoms, and includes for example hydrocarbons, lipids, polypeptides and polynucleotides.
[0123] The term "oleaginous" refers to something that is rich in oil or produces oil in high quantities.
[0124] The term "organic compound" refers to any gaseous, liquid, or solid chemical compounds which contain carbon atoms with the following exceptions that are consider inorganic: carbides, carbonates, simple oxides of carbon, cyanides, and allotropes of pure carbon such as diamond and graphite.
[0125] "Promoter" is a control DNA sequence that regulates transcription. For purposes of the invention, a promoter may includes nucleic acid sequences near the start site of transcription that are required for proper function of the promoter, as for example, a TATA element for a promoter of polymerase II type. Promoters of the present invention can include distal enhancer or repressor elements that may lie in positions up to many thousands of base pairs away from the start site of transcription. The term "inducible promoter" refers to an operable linkage between a promoter and a nucleic acid where the promoter's mediation of nucleic acid transcription is sensitive to a specific stimulus. In some embodiments, the inducible promoter requires a cofactor which can be added to the environment of the composition comprising the nucleic acid sequence that contains the inducible promoter. An "operable linkage" refers to an operative connection between nucleic acid sequences, such as for example between a control sequence (e.g. a promoter) and another sequence that codes for a protein i.e. a coding sequence. If a promoter can regulate transcription of an exogenous gene then it is in operable linkage with the gene.
[0126] The term "syngas" (from synthetic gas or synthesis gas) refers to a gas mixture that contains various proportions of hydrogen, carbon monoxide, and carbon dioxide, and which typically also includes a variety of impurities such as methane, hydrogen sulfide, condensable gases, and tars. "Producer gas" is a related term that generally refers to gas mixes similar to syngas except for the presence of a large N2 component that results from using air directly in the gasification process.
[0127] Bacterial Species
[0128] The invention relates to chemotrophic bacterial strains that comprise one or more exogenous nucleic acid sequences. The present invention results from the discovery that chemotrophic bacteria and particular related microorganisms provide unforeseen advantages in the economic and large scale production of chemicals, oils, fuels, and other hydrocarbon or lipid substances from gaseous and waste carbon feedstocks, and also from the discovery of genetic techniques and systems for modifying these microorganisms for improved performance in these applications. The lipids and other biochemicals synthesized by the microorganisms of the present invention can be applied to uses including but not limited to petrochemical substitutes, monomers, feedstock for the production of polymers, lubricants, as ingredients in animal feed, food, personal care, and cosmetic products. In some embodiments of the present invention enzymatic and chemical processes can be utilized to produce alkenes, alkynes, hydroxy acids, diacids, and unsaturated fatty acids. Some embodiments enable the production of renewable hydrocarbons. In addition, the present invention gives methods for culturing and/or modifying chemotrophic bacteria for improved lipid yield and/or lower production costs. In some embodiments the genetically modified bacteria produce more of a certain type or types of lipid molecules as compared to the same bacteria that is not genetically modified.
[0129] The present invention relates to compositions comprising and methods of using genetically modified microorganisms to produce and/or secrete carbon-based products from conversion of gaseous carbon feedstocks including but not limited to syngas or producer gas. The present invention relates to methods and mechanisms to confer production and/or secretion of carbon-based products of interest including but not limited to ethylene, chemicals, monomers, polymers, alkenes, alkynes, hydroxy acids, diacids, unsaturated fatty acids, hydrocarbons, isoprenoids, proteins, polysaccharides, nutraceutical or pharmaceutical products or intermediates thereof in obligate or facultative chemotrophic organisms such that these organisms convert carbon dioxide and/or other forms of inorganic carbon and/or syngas and/or other C1 compounds such as methanol and/or the liquid, gaseous, and solid products of pyrolytic reactions such as pyrolysis oil, into carbon-based products of interest, and in particular the use of such organisms for the commercial production of ethylene, chemicals, monomers, polymers, alkenes, alkynes, hydroxy acids, diacids, unsaturated fatty acids, hydrocarbons, isoprenoids, proteins, polysaccharides, nutraceutical or pharmaceutical products or intermediates thereof.
[0130] Chemoautotrophs are capable of performing chemosynthetic reactions that fix CO2, and/or other forms of inorganic carbon, to organic compounds, using the potential energy stored in inorganic chemicals to drive the reaction, rather than radiant energy from light as in microorganisms performing photosynthesis [Shively et al, 1998; Smith et al, 1967; Hugler et al, 2005; Hugker et al., 2005; Scott and Cavanaugh, 2007]. Carbon fixing biochemical pathways that occur in chemoautotrophs include the reductive tricarboxylic acid cycle, the Calvin-Benson-Bassham cycle [Jessup Shively, Geertje van Kaulen, Wim Meijer, Annu. Rev. Microbiol., 1998, 191-230], and the Wood-Ljungdahl pathway [Ljungdahl, 1986; Gottschalk, 1989; Lee, 2008; Fischer, 2008].
[0131] The invention relates to compositions comprising and methods of using chemoautotrophic metabolism to produce ATP for the support of ATP consuming synthetic reactions and cellular maintenance, without the co-production of methane or short chain organic acids such as acetic or butyric acid, by means of energy conserving reactions for the production of ATP using inorganic electron donors, including but not limited to the oxyhydrogen reaction.
[0132] The production of hydrocarbons or other lipids with carbon chain lengths longer than C4 is most commonly and efficiently accomplished biologically through fatty acid biosynthesis [Fischer, Klein-Marcuschamer, Stephanolpoulos, Metabolic Engineering (2008) 10, 295-304]. The initial molecule entering into the fatty acid biosynthesis pathway is acetyl-coenzyme A (acetyl-CoA), a central metabolite from which many high value biochemicals can be derived. In some embodiments, the invention utilizes microorganisms with a naturally occurring pathway for the conversion of CO, CO.sub.2 and/or H.sub.2 to acetyl-CoA. In some embodiments, the invention utilizes microorganisms that can fix CO and/or CO.sub.2 through the reductive tricarboxylic acid cycle, the Calvin-Benson-Bassham cycle, and/or the Wood-Ljungdahl pathway. In some embodiments the invention utilizes microorganisms the fix C1 compounds through a methanotropic pathway. In some embodiments the microorganisms naturally produce enzymes that catalyze the fixation of gaseous inorganic carbon to produce acetyl-CoA, utilizing gaseous electron donors such as are present in syngas as reducing agents, with such enzymatic proteins including but not limited to acetyl-CoA synthase, acetyl-CoA synthase disulfide reductase, cobalamide corrinoid/iron-sulfur protein, carbon monoxide dehydrogenase, hydrogenase, and methyltransferase.
[0133] Unlike methanogenic, acetogenic and solventogenic pathways, present in methanogens and acetogens respectively, which can produce short chain organic compounds (C1-C4) with net ATP production or zero net consumption, fatty acid synthesis involves net ATP consumption. For example the following gives the net reaction for synthesis of Palmitic acid (C16) starting from Acetyl-CoA:
8Acetyl-CoA+7ATP+H.sub.2O+14NADPH+14H+->Palmitic acid+8CoA+14NADP++7ADP+7Pi
[0134] A drawback with using an obligate methanogen or acetogen in a GTL process for the production of lipids, is the obligate use of CO2 as an electron acceptor for the production of ATP that is needed for fatty acid synthesis. If H2 is the electron donor, the ATP produced per H2 consumed in an acetogen or methanogen is relatively low: one ATP per 4H2 for methane [Thauer, R. K., Kaster, A. K., Seedorf, H., Buckel, W. & Hedderich, R. Methanogenic archaea: ecologically relevant differences in energy conservation. Nat Rev Microbiol 6, 579-591, doi:nrmicro1931 [pii]] or acetic acid production, and one ATP per 10H2 for butyric acid production [Papoutsakis, Biotechnology & Bioengineering (1984) 26, 174-187; Heise, Muller, Gottschalk, J. of Bacteriology (1989) 5473-5478; Lee, Park, Jang, Nielsen, Kim, Jung, Biotechnology & Bioengineering (2008) 101, 2, 209-228]. In some embodiments, the invention relates to a microorganism or compositions comprising a microorganism, wherein the microorganism produces ATP from an inorganic electron donor such as but not limited to H2 without synthesis of methane or short chain organic acids.
[0135] Hydrogen-oxidizing microorganisms that use more electronegative electron acceptors in energy conserving reactions for ATP production, such as but not limited to hydrogenotrophic oxyhydrogen or knallgas microbes that link the oxyhydrogen reaction, 2H2+O2->2H2O, to ATP production, can produce more ATP per H2 consumed than acetogens or methanogens. For example knallgas microorganisms can produce up to two ATP per H2 consumed [Bongers, J. Bacteriology, (October 1970) 145-151], which is eight times more ATP produced per H2 consumed than what can be produced in microorganisms undergoing methanogenesis or acetogenesis. For this reason using microorganisms that can utilize more electronegative electron acceptors in the production of ATP, such as but not limited to knallgas microbes, in fatty acid biosynthesis from syngas or H2, can be more efficient for supporting fatty acid biosynthesis than using the acetogens or methanogens that are currently used in biological GTL technologies. In some embodiments, the invention relates to a microorganism or compositions comprising a microorganism, wherein the microorganism is a knallgas microbe and comprises at least one or more exogenous nucleic acid sequences that encodes one or more enzymes to enable fixation of a carbon-containing gas feedstock, including but not limited to syngas or producer gas, into useful carbon-based products of interest including but not limited to ethylene, chemicals, monomers, polymers, alkenes, alkynes, hydroxy acids, diacids, unsaturated fatty acids, hydrocarbons, isoprenoids, proteins, polysaccharides, nutraceutical or pharmaceutical products or intermediates thereof. The invention relates to a genetically modified microorganism and compositions comprising such a microorganism, wherein the microorganism comprises one or more exogenous genes and wherein the microorganism grows on carbon-containing gas or utilizes a gaseous feedstock selected from syngas, CO2, H2, CO, or mixtures of gas comprising one or more gases selected from syngas, CO2, H2, or CO.
[0136] The invention relates to a cell and compositions comprising a cell of the class Actinobacteria comprising at least one exogenous gene. The invention also relates to cells and compositions comprising cells of the family of Nocardiaceae comprising at least one exogenous gene. The invention relates to cells and compositions comprising cells of Corynebacterium, Gordonia, Rhodococcus, Mycobacterium and Tsukamurella comprising at least one exogenous gene. In some embodiments, the invention relate to cells of the family of Nocardiaceae comprising an exogenous gene, wherein the cell is not a cell of the genus Mycobacterium. In some embodiments, the invention provides a cell and compositions comprising a cell of the genus Rhodococcus comprising an exogenous gene, and in some embodiments the cell is a strain of the species Rhodococcus sp., Rhodococcus opacus, Rhodococcus aurantiacus; Rhodococcus baikonurensis; Rhodococcus boritolerans; Rhodococcus equi; Rhodococcus coprophilus; Rhodococcus corynebacterioides; Nocardia corynebacterioides (synonym: Nocardia corynebacterioides); Rhodococcus erythropolis; Rhodococcus fascians; Rhodococcus globerulus; Rhodococcus gordoniae; Rhodococcus jostii Rhodococcus koreensis; Rhodococcus kroppenstedtii; Rhodococcus maanshanensis; Rhodococcus marinonascens; Rhodococcus opacus; Rhodococcus percolatus; Rhodococcus phenolicus; Rhodococcus polyvorum; Rhodococcus pyridinivorans; Rhodococcus rhodochrous; Rhodococcus rhodnii; (synonym: Nocardia rhodnii); Rhodococcus ruber (synonym: Streptothrix rubra); Rhodococcus sp. RHA1; Rhodococcus triatomae; Rhodococcus tukisamuensis; Rhodococcus wratislaviensis (synonym: Tsukamurella wratislaviensis); Rhodococcus yunnanensis; Rhodococcus zopfii. In some embodiments the cell comprising one or more exogenous genes is strain Rhodococcus opacus DSM number 43205 or 43206. In some embodiments the cell comprising one or more exogenous genes is strain Rhodococcus sp. DSM number 3346. In some embodiments, the invention provides cells and compositions comprising a cell of the genus Rhodococcus comprising an exogenous gene, wherein the cell or composition comprising a cell of Rhodococcus is non-infectious to animals and/or plants. In some embodiments, the invention provides cells and compositions comprising a cell of the genus Rhodococcus comprising an exogenous gene, wherein the Rhodococcus cell or composition comprising a Rhodococcus cell is non-infectious to humans. In some embodiments, the invention provides cells and compositions comprising a cell of the genus Rhodococcus comprising an exogenous gene, wherein the Rhodococcus cell or composition comprising a Rhodococcus cell is non-infectious to plants. In some embodiments, the invention provides cells and compositions comprising cells of the genus Rhodococcus comprising an exogenous gene, wherein, if the cell is from Rhodococcus equi or Rhodococcus fascians species, the species is non-infectious to animals and/or plants. In some embodiments, the invention relates to a Rhodococcus cell or composition comprising a Rhodococcus cell, wherein the cell is not a species selected from Rhodococcus equi or Rhodococcus fascians.
[0137] In some embodiments, the invention relates to a Rhodococcus cell or composition comprising a Rhodococcus cell, wherein the cell is incapable of producing any acrylic acid or acrylamide. In some embodiments, the invention relates to a Rhodococcus cell or composition comprising a Rhodococcus cell, wherein the cell produces less than 10, 9, 8, 7, 6, 5, 4, 3, 2, or 1% of its weight of total dry cellular matter in acrylamide or acrylic/methylacrylic acid. In some embodiments, the invention relates to a Rhodococcus cell or composition comprising a Rhodococcus cell, wherein the cell is not from the species Rhodococcus rhodochrous. In some embodiments, the invention relates to Rhodococcus cell or composition comprising a Rhodococcus cell, wherein the cell is incapable of producing 10-hydroxy-12-octadecenoic acid. In some embodiments, the invention relates to a Rhodococcus cell or composition comprising a Rhodococcus cell, wherein the cell is unable to produce more than 10, 9, 8, 7, 6, 5, 4, 3, 2, or 1% of its weight of total dry cellular matter in 10-hydroxy-12-octadecenoic acid. In some embodiments, the invention relates to Rhodococcus cell or composition comprising a Rhodococcus cell, wherein the cell is incapable of producing optically-active 4-amino-3-hydroxybutyric acid. In some embodiments, the invention relates to a Rhodococcus cell or composition comprising a Rhodococcus cell, wherein the cell is unable to produce more than 10, 9, 8, 7, 6, 5, 4, 3, 2, or 1% of its weight of total dry cellular matter in optically-active 4-amino-3-hydroxybutyric acid.
[0138] In some embodiments, the cell or compositions comprising one of more cells is not E. coli. In some embodiments, the cell or compositions comprising one of more cells is from the genus Rhodococcus but is not for the species equi. In some embodiments, the cell of the present invention is not pathogenic to animals or plants. In some embodiments, the cell of the present invention is not pathogenic to humans. In some embodiments, the cell or compositions comprising one of more cells is from the genus Ralstonia. In some embodiments, the cell or compositions comprising one of more cells is from the species Ralstonia eutropha. In some embodiments the cell comprising one or more exogenous genes is strain Cupriavidus necator DSM number 531 or 541.
[0139] In some embodiments, the cell or compositions comprising the one or more cells have a 16S rRNA sequence with at least 50, 60, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98 or 99% nucleotide homology to one or more of SEQ ID NOs: 11 or 12. In some embodiments, the cell or compositions comprising the one or more cells have a 16S rRNA sequence with at least 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98 or 99% nucleotide homology to one or more of SEQ ID NOs: 11. In some embodiments, the cell or compositions comprising the one or more cells have a 16S rRNA sequence with at least 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98 or 99% nucleotide homology to one or more of SEQ ID NOs: 12. In some embodiments, the cell or compositions comprising the one or more cells have a 16S rRNA sequence with at least 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98 or 99% nucleotide homology to one or more of SEQ ID NOs: 13.
[0140] In some embodiments the microorganism of the claimed invention is not dependent upon light to grow and/or metabolize and/or synthesize lipid molecules. In some embodiments, the microorganism of the claimed invention does not require any type of sugar to grow and/or metabolize and/or synthesize lipid molecules. In some embodiments, the microorganism of the claimed invention does not require any type of organic compound to grow and/or metabolize and/or synthesize lipid molecules. In some embodiments, the microorganism of the claimed invention does not require any type of fixed carbon to grow and/or metabolize and/or synthesize lipid molecules. In some embodiments, the microorganism can grow and/or metabolize lipids in a slightly anaerobic or extremely anaerobic environment. In some embodiments, the microorganism of the claimed invention is a facultative microorganism
[0141] Microbial culturing in the present invention is performed both for the sake of implementing genetic modifications, and for production of organic compounds, and specifically lipids and/or hydrocarbons (e.g., alkenes, alkynes, unsaturated fatty acids, or triacylglyccrols, hydroxy acids, diacids). Microbial culturing with the aim of genetic manipulation is generally performed at a small benchtop scale and often under conditions that select for genetically modified traits. Microbial culturing aimed at the commercial production of organic compounds and specifically lipids and/or hydrocarbons is typically performed in bioreactors at much greater scale (e.g., 500 L, 1,000 L 5,000 L, 10,000 L, 50,000 L, 100,000 L, 1,000,000 L bioreactor volumes and higher). In certain embodiments the chemoautotrophs of the present invention are grown in a liquid media inside a bioreactor using the methods of the invention. In some embodiments, the bioreactor containing the microorganisms is constructed of opaque materials that keep the culture in darkness. Bioreactors constructed out of opaque materials such as steel or reinforced concrete can be designed to have extremely big working volumes. In some embodiments of the present invention steel fermenters 50,000 liter and greater in volume are utilized. In some embodiments of the present invention egg-shape or cylindrical digesters 3,000,000 liters and greater in volume are utilized. In some embodiments, the bioreactor comprising the microorganism does not allow light to penetrate its interior.
[0142] The bioreactor or fermentor is used to culture cells through the various phases of their physiological cycle. A bioreactor is utilized for the cultivation of cells, which may be maintained at particular phases in their growth curve. The use of bioreactors is advantageous in many ways for cultivating chemoautotrophic growth. For certain embodiments, oleaginous cell mass, which is used to produce oleochemicals, is grown to high densities in liquid suspension. Generally the control of growth conditions including control of dissolved carbon dioxide, oxygen, and other gases such as hydrogen, as well as other dissolved nutrients, trace elements, temperature and pH, is facilitated in a bioreactor.
[0143] Nutrient media as well as gases can be added to the bioreactor as either a batch addition, or periodically, or in response to a detected depletion or programmed set point, or continuously over the period the culture is grown and/or maintained. For certain embodiments, the bioreactor at inoculation is filled with a starting batch of nutrient media and/or gases at the beginning of growth, and no additional nutrient media and/or gases are added after inoculation. For certain embodiments, nutrient media and/or gases are added periodically after inoculation. For certain embodiments, nutrient media and/or gas is added after inoculation in response to a detected depletion of nutrient and/or gas. For certain embodiments, nutrient media and/or gas is added continuously after inoculation.
[0144] For certain embodiments the bioreactors have mechanisms to enable mixing of the nutrient media that include but are not limited to spinning stir bars, blades, impellers, or turbines, spinning, rocking, or turning vessels, gas lifts and sparging. The culture media may be mixed continuously or intermittently.
[0145] The ports that are standard in bioreactors may be utilized to deliver, or withdraw, gases, liquids, solids, and/or slurries, into the bioreactor vessel enclosing the microbes of the present invention. Many bioreactors have multiple ports for different purposes (e.g. ports for media addition, gas addition, probes for pH and DO, sampling), and a given port may be used for various purposes during the course of a fermentation run. As an example, a port might be used to add nutrient media to the bioreactor at one point in time and at another time might be used for sampling. Preferably, the multiple use of a sampling port can be performed without introducing contamination or invasive species into the growth environment. A valve or other actuator enabling control of the sample flow or continuous sampling can be provided to a sampling port. For certain embodiments the bioreactors are equipped with at least one port suitable for culture inoculation that can additionally serve other uses including the addition of media or gas. Bioreactors ports enable control of the gas composition and flow rate into the culture environment. For example the ports can be used as gas inlets into the bioreactor through which gases are pumped. For some embodiments gases that may be pumped into a bioreactor include syngas, producer gas, hydrogen gas, CO2, air, air/CO2 mixtures, ammonia, nitrogen, noble gases, such as argon, as well as other gases. In some embodiments that CO.sub.2 may come from sources including but are not limited to: CO.sub.2 from the gasification of organic matter; CO.sub.2 from the calcination of limestone, CaCO.sub.3, to produce quicklime, CaO; CO.sub.2 from methane steam reforming, such as the CO.sub.2 byproduct from ammonia or hydrogen production; combustion; CO.sub.2 byproduct of sugar fermentation; CO.sub.2 byproduct from sodium phosphate production; geologically or geothermally produced CO.sub.2. Raising the gas flow rate into a bioreactor can enhance mixing of the culture and produce turbulence if the gas inlet is positioned under the surface of the liquid media such that gas bubbles or sparges up through the media. In some embodiments, a bioreactor comprises gas outlet ports for gas escape and pressure release. In some embodiments, gas inlets and outlets are preferably equipped with check valves to prevent gas backflow.
[0146] The present invention relates to bioreactors that comprise a cell, which comprises at least one exogenous nucleic acid sequences that encodes a lipid pathway enzyme. The present invention relates to a system of at least one bioreactor that comprise a cell, which comprises at least one exogenous nucleic acid sequences that encodes a lipid pathway enzyme. In some embodiments, the system comprises two or more, three or more, or four or more bioreactors, at least one of which comprise a cell, which comprises at least one exogenous nucleic acid sequences that encodes a lipid pathway enzyme. In some embodiments, the system of bioreactors comprises at least a first and second bioreactor, wherein the first bioreactor comprises a cell, which comprises at least one exogenous nucleic acid sequences that encodes a lipid pathway enzyme; and wherein the second bioreactor comprises a microorganism derived from a different species, wherein the microorganism from a different species comprises at least one exogenous nucleic acid sequence that encodes a lipid pathway enzyme. In some embodiments, the system of bioreactors comprises a first bioreactor that comprises the cell of the present invention and a second bioreactor comprising a microalgal, yeast, or bacterial cell.
[0147] In some embodiments, the cells of the present invention are capable of producing desaturated alkanes between 8 and 18 carbon atoms long at greater than 18 grams per liter volume of culture per three day period. In some embodiments, the cells of the present invention are capable of producing desaturated hydrocarbons between 8 and 18 carbon atoms long at greater than or equal to 18 grams per liter volume of culture per three day period, wherein the desatruated hydrocarbons are desaturated at a carbon position other than carbon-9.
[0148] Genetic Modifications
[0149] The present invention relates to methods of modifying a bacterial cell to express one or more exogenous nucleic acid sequences that encodes one or more enzymes to enable fixation of a carbon-containing gas feedstock into useful carbon-based products of interest in an amount greater than an amount of carbon-based products produced by the same bacterial cell that does not express the exogenous nucleic acid sequences. Methods of selecting and manufacturing nucleic acid sequences for modification of bacterial cells are known and can be performed by transformation, electroporation, phage infection of bacteria, or other techniques for nucleic acid transfer generally known in the art. Standard recombinant DNA and molecular cloning techniques useful for the invention are well known in the art and are described by Sambrook, J., Fritsch, E. F. and Maniatis, T. Molecular Cloning: A Laboratory Manual; Cold Spring Harbor Laboratory Press: Cold Spring Harbor, (1989) (Maniatis) and by T. J. Silhavy, M. L. Bennan, and L. W. Enquist, Experiments with Gene Fusions, Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. (1984) and by Ausubel, F. M. et al., Current Protocols in Molecular Biology, pub. by Greene Publishing Assoc. and Wiley-Interscience (1987), all of which are incorporated by reference in their entireties.
[0150] The invention relates to genetic constructs comprising one or more exogenous genes that encode one or more amino acid sequences to enable fixation of a carbon-containing gas feedstock, including but not limited to syngas or producer gas, into useful carbon-based products of interest in an amount greater than an amount of carbon-based products produced by the same bacterial cell that does not express the exogenous nucleic acid sequence or sequences. Another aspect of the present invention relates to compositions that comprise at least one bacterial cell, which comprises at least one nucleic acid sequence that encodes at least one exogenous amino acid sequence that functions as a fatty acid acyl-ACP reductase, a fatty acid aldehyde decarbonylase and/or a thioesterase. In some embodiments, the bacterial cell is transformed with one or more, two or more, three or more, four or more, or five or more exogenous nucleic acid sequences that encode one or more amino acid sequences to enable fixation of a carbon-containing gas feedstock, including but not limited to syngas or producer gas, into useful carbon-based products of interest in an amount greater than an amount of carbon-based products produced by the same bacterial cell that does not express the exogenous nucleic acid sequence or sequences. According to the present invention, genetic material that encodes the enzyme is delivered to a bacterial cell in an expressible form. The genetic material, DNA or RNA, is taken up by the cells of the invention and expressed. The enzyme or enzymes that are thereby produced can biochemically modify lipid molecules to remove or add hydroxyl groups, remove or add carbonyl groups, remove or add carbon-carbon double bonds, remove or add carbon-carbon triple bonds, remove or add aldehyde groups, remove or add hydroxy groups, remove or add carboxylic acid groups, or remove or add ester groups to lipid molecules in lipid.
[0151] In some embodiments, the genetic constructs of the present invention comprise DNA, RNA, or combinations of both DNA and RNA. In some embodiments, the genetic construct of the present invention is a plasmid. It will be appreciated that, in some embodiments, the plasmid contains a variety of open reading frames (ORFs) encoding proteins of many diverse functions, including those enzymes that enable hydrocarbon or lipid modification, glutathione-S transferase (GST) activity, origins of replication, multiple cloning sites, promoters, and/or termination sequences. It is contemplated therefore that a host cell transformed with the plasmid will demonstrate the ability to modify a variety of lipids as well as maintain its copy number in the cytoplasm of the cell. The glutathione-S transferases (GSTs) represent a large group of detoxification enzymes. GSTs catalyze the conjugation of glutathione, homoglutathione and other glutathione-like analog via sulfhydryl group, to a large range of hydrophobic, electrophilic compounds. The conjugation can result in detoxification of these compounds. GST genes are found in both prokaryotic (e.g., E. coli) and eukaryotic organisms (e.g., yeast, plant and human). Although the homologies between the GSTs from prokaryotes and eukaryotes were low, many of the residues assigned to be important for the enzymatic function or structure in the eukaryotes were found to be conserved in prokaryotic GSTs (Nishida et al., J. Biol Chem 269:32536-32541 (1994)). It has been suggested that bacterial GST may represent a defense against the effects of antibiotics (Piccolomini et al., J Gen Microbiol 135:3119-3125 (1989)). Accordingly it is contemplated that a host strain transformed with the plasmid will have the ability detoxify harmful compounds via conjugation of those compounds to glutathione.
[0152] In some embodiments, the instant plasmid additionally encodes a variety of maintenance proteins, useful for maintaining, stabilizing and replicating the plasmid. It is contemplated that these genes may be used in conjunction with other bacterial plasmids deficient in these functions for the increased stabilization or robust maintenance of the plasmid. In some embodiments, the plasmid comprises maintenance proteins of particular interest including the REP origin of replication (encoded by ORF 38) the TRA proteins (TRAI, TRAJ and TRAK, encoded by ORF's 23, 24 and 25 respectively) and the VAG proteins (VAGD and VAGC, encoded by ORF's 33 and 34 respectively). The tra gene family is known to be involved in plasmid conjugation, a process that promotes DNA transfer from a donor to a recipient cell mediated by physical contact (Firth et al, Escherichia coli and Salmonella: Cellular and Molecular Biology, ASM press (1996)). Among tra gene products, TraI and TraK proteins are reported to be required for efficient plasmid site-specific recombination (Paterson et al. J. Bacteriol 181:2572-2583 (1999)). Furthermore, TraI is required for conjugal DNA transfer. Fukuda and Ohtsubo (Genes Cells 2:735-751 (1997)) reported that TraI has the activity of site- and strand-specific nicking of the supercoiled plasmid DNA. TraJ, traJ gene product, regulates transcription originating at the tra operon promoter P.sub.traY. (Firth et al., Escherichia coli and Salmonella: Cellular and Molecular Biology, ASM press (1996)). The stabilization proteins VAGC and VAGD encoded by vagC and vagD are involved in the maintaining the plasmid as an autonomous replicating unit. Bacterial maintenance proteins of particular interest on the pSeq and pVer plasmids include.
TABLE-US-00001 SEQ ID: 01 TCGCGCGTTT CGGTGATGAC GGTGAAAACC TCTGACACAT GCAGCTCCCG GAGACGGTCA CAGCTTGTCT GTAAGCGGAT GCCGGGAGCA GACAAGCCCG AGCGCGCAAA GCCACTACTG CCACTTTTGG AGACTGTGTA CGTCGAGGGC CTCTGCCAGT GTCGAACAGA CATTCGCCTA CGGCCCTCGT CTGTTCGGGC TCAGGGCGCG TCAGCGGGTG TTGGCGGGTG TCGGGGCTGG CTTAACTATG CGGCATCAGA GCAGATTGTA CTGAGAGTGC ACCATATGCG GTGTGAAATA AGTCCCGCGC AGTCGCCCAC AACCGCCCAC AGCCCCGACC GAATTGATAC GCCGTAGTCT CGTCTAACAT GACTCTCACG TGGTATACGC CACACTTTAT CCGCACAGAT GCGTAAGGAG AAAATACCGC ATCAGGCGCC ATTCGCCATT CAGGCTGCGC AACTGTTGGG AAGGGCGATC GGTGCGGGCC TCTTCGCTAT GGCGTGTCTA CGCATTCCTC TTTTATGGCG TAGTCCGCGG TAAGCGGTAA GTCCGACGCG TTGACAACCC TTCCCGCTAG CCACGCCCGG AGAAGCGATA TACGCCAGCT GGCGAAAGGG GGATGTGCTG CAAGGCGATT AAGTTGGGTA ACGCCAGGGT TTTCCCAGTC ACGACGTTGT AAAACGACGG CCAGTGCCAA ATGCGGTCGA CCGCTTTCCC CCTACACGAC GTTCCGCTAA TTCAACCCAT TGCGGTCCCA AAAGGGTCAG TGCTGCAACA TTTTGCTGCC GGTCACGGTT GCTTGCATGC CTGCAGGTCG ACGGGCCCGG GATCCGATGC TCTTCCGCTA AGATCTGCCG CGGCCGCGTC CTCAGAAGAA CTCGTCAAGA AGGCGATAGA CGAACGTACG GACGTCCAGC TGCCCGGGCC CTAGGCTACG AGAAGGCGAT TCTAGACGGC GCCGGCGCAG GAGTCTTCTT GAGCAGTTCT TCCGCTATCT AGGCGATGCG CTGCGAATCG GGAGCGGCGA TACCGTAAAG CACGAGGAAG CGGTCAGCCC ATTCGCCGCC AAGCTCTTCA GCAATATCAC GGGTAGCCAA TCCGCTACGC GACGCTTAGC CCTCGCCGCT ATGGCATTTC GTGCTCCTTC GCCAGTCGGG TAAGCGGCGG TTCGAGAAGT CGTTATAGTG CCCATCGGTT CGCTATGTCC TGATAGCGGT CCGCCACACC CAGCCGGCCA CAGTCGATGA ATCCAGAAAA GCGGCCATTT TCCACCATGA TATTCGGCAA GCAGGCATCG GCGATACAGG ACTATCGCCA GGCGGTGTGG GTCGGCCGGT GTCAGCTACT TAGGTCTTTT CGCCGGTAAA AGGTGGTACT ATAAGCCGTT CGTCCGTAGC CCATGGGTCA CGACGAGATC CTCGCCGTCG GGCATGCGCG CCTTGAGCCT GGCGAACAGT TCGGCTGGCG CGAGCCCCTG ATGCTCTTCG TCCAGATCAT GGTACCCAGT GCTGCTCTAG GAGCGGCAGC CCGTACGCGC GGAACTCGGA CCGCTTGTCA AGCCGACCGC GCTCGGGGAC TACGAGAAGC AGGTCTAGTA CCTGATCGAC AAGACCGGCT TCCATCCGAG TACGTGCTCG CTCGATGCGA TGTTTCGCTT GGTGGTCGAA TGGGCAGGTA GCCGGATCAA GCGTATGCAG GGACTAGCTG TTCTGGCCGA AGGTAGGCTC ATGCACGAGC GAGCTACGCT ACAAAGCGAA CCACCAGCTT ACCCGTCCAT CGGCCTAGTT CGCATACGTC CCGCCGCATT GCATCAGCCA TGATGGATAC TTTCTCGGCA GGAGCAAGGT GGGATGACAG GAGATCCTGC CCCGGCACTT CGCCCAATAG CAGCCAGTCC GGCGGCGTAA CGTAGTCGGT ACTACCTATG AAAGAGCCGT CCTCGTTCCA CCCTACTGTC CTCTAGGACG GGGCCGTGAA GCGGGTTATC GTCGGTCAGG CTTCCCGCTT CAGTGACAAC GTCGAGCACA GCTGCGCAAG GAACGCCCGT CGTGGCCAGC CACGATAGCC GCGCTGCCTC GTCCTGCAGT TCATTCAGGG GAAGGGCGAA GTCACTGTTG CAGCTCGTGT CGACGCGTTC CTTGCGGGCA GCACCGGTCG GTGCTATCGG CGCGACGGAG CAGGACGTCA AGTAAGTCCC CACCGGACAG GTCGGTCTTG ACAAAAAGAA CCGGGCGCCC CTGCGCTGAC AGCCGGAACA CGGCGGCATC AGAGCAGCCG ATTGTCTGTT GTGCCCAGTC GTGGCCTGTC CAGCCAGAAC TGTTTTTCTT GGCCCGCGGG GACGCGACTG TCGGCCTTGT GCCGCCGTAG TCTCGTCGGC TAACAGACAA CACGGGTCAG ATAGCCGAAT AGCCTCTCCA CCCAAGCGGC CGGAGAACCT GCGTGCAATC CATCTTGTTC AATCATGATA TCCCTTAATT AACCGTTAAC ACTAGTTCAG TATCGGCTTA TCGGAGAGGT GGGTTCGCCG GCCTCTTGGA CGCACGTTAG GTAGAACAAG TTAGTACTAT AGGGAATTAA TTGGCAATTG TGATCAAGTC TCCATCTCGC CGTGTATGCG GGCCTGACGG ATCAACGTTC CCACCGAGCC AGTCGAGATG TTCATCTGGT CGGCGATCTG CCGGTACTTC AAACCTTGTT AGGTAGAGCG GCACATACGC CCGGACTGCC TAGTTGCAAG GGTGGCTCGG TCAGCTCTAC AAGTAGACCA GCCGCTAGAC GGCCATGAAG TTTGGAACAA TGCGCAGTTC CACAGCCTTC TTGCGGCGTT CCTGCGCACG AGCGATGTAG TCGCCTCGGT CTTCGGCGAC GAGCCGTTTG ATGGTGCTTT TCGAGACGCC ACGCGTCAAG GTGTCGGAAG AACGCCGCAA GGACGCGTGC TCGCTACATC AGCGGAGCCA GAAGCCGCTG CTCGGCAAAC TACCACGAAA AGCTCTGCGG GAACTTGTCA GCCAACTCCT GCGCGGTCTG CGTGCGACGC ATCACGCGTT CTGCAGCACC CATCAGTCCG TCCCCTCTGC TGCTGCGAAC AGTGCCGATC CTTGAACAGT CGGTTGAGGA CGCGCCAGAC GCACGCTGCG TAGTGCGCAA GACGTCGTGG GTAGTCAGGC AGGGGAGACG ACGACGCTTG TCACGGCTAG GATCGACCTT CTTGAGCTTC GGCCGCGGCG CGGTGGCGTT CTTCCGTACC GCTTCCGTTT TTGCGCTGCT GCTCACTTTG CCGCGGCGTG CCTGGATTTT CTAGCTGGAA GAACTCGAAG CCGGCGCCGC GCCACCGCAA GAAGGCATGG CGAAGGCAAA AACGCGACGA CGAGTGAAAC GGCGCCGCAC GGACCTAAAA CGAGAACTCG GCGGCGGTGA AGGTGCGGTG GGTCCAGTGG GCGACTGATT TGCCGATCTG CTCGGCCTCG GCCCGACTCA TGGGGCCGAT CCCGTCGTTG GCTCTTGAGC CGCCGCCACT TCCACGCCAC CCAGGTCACC CGCTGACTAA ACGGCTAGAC GAGCCGGAGC CGGGCTGAGT ACCCCGGCTA GGGCAGCAAC GCGTCGAGGG TGAAGTTGGT CAGGGCGGTG AAGTCGGTGA CCATCTGCCG CCACACAGTG ATCGACGGGT AGTTCTGTTT CCGGATCTCG CGGTAGGCCC CGCAGCTCCC ACTTCAACCA GTCCCGCCAC TTCAGCCACT GGTAGACGGC GGTGTGTCAC TAGCTGCCCA TCAAGACAAA GGCCTAGAGC GCCATCCGGG ATTCCCGGGT GCGGTCGAAC AGTTCGACGT TCCGGCCCGT TTCGGTCCTG ACCTGTGTCT TGCGGCCGTA GTCCGGTGGG GCGGGGAAAC GGTCACCGAG TAAGGGCCCA CGCCAGCTTG TCAAGCTGCA AGGCCGGGCA AAGCCAGGAC TGGACACAGA ACGCCGGCAT CAGGCCACCC CGCCCCTTTG CCAGTGGCTC CGCTTTTGCG AGGCCTTTGA GCGAGTACGG ATCCGAGGGA CCCCAGACCG TCGTCCAGTG CGGGTGGATC GGGTTCTGGG TGAGCTGCTG CGCGTAGCCC GCGAAAACGC TCCGGAAACT CGCTCATGCC TAGGCTCCCT GGGGTCTGGC AGCAGGTCAC GCCCACCTAG CCCAAGACCC ACTCGACGAC GCGCATCGGG TGATCGGCGC CGACCACCGA GGCGATCAGC CCCTGGTTCA CCCGGTCGTA GAGCCGCAGC GGGCCCTGTC GGGCTGCCTG GAGGGTGTAG ACCGGGCTTT ACTAGCCGCG GCTGGTGGCT CCGCTAGTCG GGGACCAAGT GGGCCAGCAT CTCGGCGTCG CCCGGGACAG CCCGACGGAC CTCCCACATC TGGCCCGAAA CGAGCAGCCA CCACAGGTGC GCGTGCTCGG TCGCGGGATT GATCGTCATC ACGGTCGGAT CGGGCAGATC CGCGTTACGT GCGGCCCACT GCGCCTGGTC GCTCGTCGGT GGTGTCCACG CGCACGAGCC AGCGCCCTAA CTAGCAGTAG TGCCAGCCTA GCCCGTCTAG GCGCAATGCA CGCCGGGTGA CGCGGACCAG GTCGTCCACG TCGAGCACCA AGCCCAACCT GATCGACGGG GTGCGGGCCG CAATGTAGCG GCGGGTGAGC GCCTCCGCGC GCGGCTGCGG CCACTGCCCG CAGCAGGTGC AGCTCGTGGT TCGGGTTGGA CTAGCTGCCC CACGCCCGGC GTTACATCGC CGCCCACTCG CGGAGGCGCG CGCCGACGCC GGTGACGGGC TCCCGGACGT AGTCATCCGT CGCGTGCGGG TATTTGAACC GCCAGCGGTC CAACCAGGCG TCAACAGCAG CGGTCATGAC CGCCAAGCTA GGGCCGGATC AGGGCCTGCA TCAGTAGGCA GCGCACGCCC ATAAACTTGG CGGTCGCCAG GTTGGTCCGC AGTTGTCGTC GCCAGTACTG GCGGTTCGAT CCCGGCCTAG TGTACCGATC GGGGGAGGCG CGCCGCAAAT TATTTAAGAG TCTCGCTAGC AAACCATGTC AGGTGTTGCG GTGGGTTCCG GGTAAACCTC CACCCGAATT ACATGGCTAG CCCCCTCCGC GCGGCGTTTA ATAAATTCTC AGAGCGATCG TTTGGTACAG TCCACAACGC CACCCAAGGC CCATTTGGAG GTGGGCTTAA ATTTAAGAGT CTCGCTAGCT AAGCCCTATC TGATGCTGCG CGGGGGGTCC TTCGCACTGA ATCTCAAAGG TGGCCGGCTG AATTTCGTCG CGCGAAAACC TAAATTCTCA GAGCGATCGA TTCGGGATAG ACTACGACGC GCCCCCCAGG AAGCGTGACT TAGAGTTTCC ACCGGCCGAC TTAAAGCAGC GCGCTTTTGG TCCCTGGACA GTTCTGGAAT TCAGCAAGAG GTGTGTCTGA ACTTCGGTGT TTTTTTGGGG GGTGACTCCA GCGGGGTGGG CACAACGCGA ACAGAGACCT AGGGACCTGT CAAGACCTTA AGTCGTTCTC CACACAGACT TGAAGCCACA AAAAAACCCC CCACTGAGGT CGCCCCACCC GTGTTGCGCT TGTCTCTGGA TGTGTGTACG ACGGCGGGAG GTAAGTCGGG TACGGCTCGG ACTGCGGTAG AGCAACCGTC GAATCGATTT CGAGCAGAGC GAGCAGAGCA AGATATTCCA ACACACATGC TGCCGCCCTC CATTCAGCCC ATGCCGAGCC TGACGCCATC TCGTTGGCAG CTTAGCTAAA GCTCGTCTCG CTCGTCTCGT TCTATAAGGT AAACTCCGGG GTTCCTCGGC GGCCTCCCCC GTCTGTTTGC TCAACCGAGG GAGACCTGGC GGTCCCGCGT TTCCGGACGC GCGGGACCGC CTACCGCTCG TTTGAGGCCC CAAGGAGCCG CCGGAGGGGG CAGACAAACG AGTTGGCTCC CTCTGGACCG CCAGGGCGCA AAGGCCTGCG CGCCCTGGCG GATGGCGAGC AGAGCGGAAG AGCATCTAGA TGCATTCGCG AGGTACCGAG CTCGAATTCG TAATCATGGT CATAGCTGTT TCCTGTGTGA AATTGTTATC CGCTCACAAT TCTCGCCTTC TCGTAGATCT ACGTAAGCGC TCCATGGCTC GAGCTTAAGC ATTAGTACCA GTATCGACAA AGGACACACT TTAACAATAG GCGAGTGTTA
TCCACACAAC ATACGAGCCG GAAGCATAAA GTGTAAAGCC TGGGGTGCCT AATGAGTGAG CTAACTCACA TTAATTGCGT TGCGCTCACT GCCCGCTTTC AGGTGTGTTG TATGCTCGGC CTTCGTATTT CACATTTCGG ACCCCACGGA TTACTCACTC GATTGAGTGT AATTAACGCA ACGCGAGTGA CGGGCGAAAG CAGTCGGGAA ACCTGTCGTG CCAGCTGCAT TAATGAATCG GCCAACGCGC GGGGAGAGGC GGTTTGCGTA TTGGGCGCTC TTCCGCTTCC TCGCTCACTG GTCAGCCCTT TGGACAGCAC GGTCGACGTA ATTACTTAGC CGGTTGCGCG CCCCTCTCCG CCAAACGCAT AACCCGCGAG AAGGCGAAGG AGCGAGTGAC ACTCGCTGCG CTCGGTCGTT CGGCTGCGGC GAGCGGTATC AGCTCACTCA AAGGCGGTAA TACGGTTATC CACAGAATCA GGGGATAACG CAGGAAAGAA TGAGCGACGC GAGCCAGCAA GCCGACGCCG CTCGCCATAG TCGAGTGAGT TTCCGCCATT ATGCCAATAG GTGTCTTAGT CCCCTATTGC GTCCTTTCTT CATGTGAGCA AAAGGCCAGC AAAAGGCCAG GAACCGTAAA AAGGCCGCGT TGCTGGCGTT TTTCCATAGG CTCCGCCCCC CTGACGAGCA TCACAAAAAT GTACACTCGT TTTCCGGTCG TTTTCCGGTC CTTGGCATTT TTCCGGCGCA ACGACCGCAA AAAGGTATCC GAGGCGGGGG GACTGCTCGT AGTGTTTTTA CGACGCTCAA GTCAGAGGTG GCGAAACCCG ACAGGACTAT AAAGATACCA GGCGTTTCCC CCTGGAAGCT CCCTCGTGCG CTCTCCTGTT CCGACCCTGC GCTGCGAGTT CAGTCTCCAC CGCTTTGGGC TGTCCTGATA TTTCTATGGT CCGCAAAGGG GGACCTTCGA GGGAGCACGC GAGAGGACAA GGCTGGGACG CGCTTACCGG ATACCTGTCC GCCTTTCTCC CTTCGGGAAG CGTGGCGCTT TCTCATAGCT CACGCTGTAG GTATCTCAGT TCGGTGTAGG TCGTTCGCTC GCGAATGGCC TATGGACAGG CGGAAAGAGG GAAGCCCTTC GCACCGCGAA AGAGTATCGA GTGCGACATC CATAGAGTCA AGCCACATCC AGCAAGCGAG CAAGCTGGGC TGTGTGCACG AACCCCCCGT TCAGCCCGAC CGCTGCGCCT TATCCGGTAA CTATCGTCTT GAGTCCAACC CGGTAAGACA CGACTTATCG GTTCGACCCG ACACACGTGC TTGGGGGGCA AGTCGGGCTG GCGACGCGGA ATAGGCCATT GATAGCAGAA CTCAGGTTGG GCCATTCTGT GCTGAATAGC CCACTGGCAG CAGCCACTGG TAACAGGATT AGCAGAGCGA GGTATGTAGG CGGTGCTACA GAGTTCTTGA AGTGGTGGCC TAACTACGGC TACACTAGAA GGTGACCGTC GTCGGTGACC ATTGTCCTAA TCGTCTCGCT CCATACATCC GCCACGATGT CTCAAGAACT TCACCACCGG ATTGATGCCG ATGTGATCTT GGACAGTATT TGGTATCTGC GCTCTGCTGA AGCCAGTTAC CTTCGGAAAA AGAGTTGGTA GCTCTTGATC CGGCAAACAA ACCACCGCTG GTAGCGGTGG CCTGTCATAA ACCATAGACG CGAGACGACT TCGGTCAATG GAAGCCTTTT TCTCAACCAT CGAGAACTAG GCCGTTTGTT TGGTGGCGAC CATCGCCACC TTTTTTTGTT TGCAAGCAGC AGATTACGCG CAGAAAAAAA GGATCTCAAG AAGATCCTTT GATCTTTTCT ACGGGGTCTG ACGCTCAGTG GAACGAAAAC AAAAAAACAA ACGTTCGTCG TCTAATGCGC GTCTTTTTTT CCTAGAGTTC TTCTAGGAAA CTAGAAAAGA TGCCCCAGAC TGCGAGTCAC CTTGCTTTTG TCACGTTAAG GGATTTTGGT CATGAGATTA TCAAAAAGGA TCTTCACCTA GATCCTTTTA AATTAAAAAT GAAGTTTTAA ATCAATCTAA AGTATATATG AGTGCAATTC CCTAAAACCA GTACTCTAAT AGTTTTTCCT AGAAGTGGAT CTAGGAAAAT TTAATTTTTA CTTCAAAATT TAGTTAGATT TCATATATAC AGTAAACTTG GTCTGACAGT TACCAATGCT TAATCAGTGA GGCACCTATC TCAGCGATCT GTCTATTTCG TTCATCCATA GTTGCCTGAC TCCCCGTCGT TCATTTGAAC CAGACTGTCA ATGGTTACGA ATTAGTCACT CCGTGGATAG AGTCGCTAGA CAGATAAAGC AAGTAGGTAT CAACGGACTG AGGGGCAGCA GTAGATAACT ACGATACGGG AGGGCTTACC ATCTGGCCCC AGTGCTGCAA TGATACCGCG AGACCCACGC TCACCGGCTC CAGATTTATC AGCAATAAAC CATCTATTGA TGCTATGCCC TCCCGAATGG TAGACCGGGG TCACGACGTT ACTATGGCGC TCTGGGTGCG AGTGGCCGAG GTCTAAATAG TCGTTATTTG CAGCCAGCCG GAAGGGCCGA GCGCAGAAGT GGTCCTGCAA CTTTATCCGC CTCCATCCAG TCTATTAATT GTTGCCGGGA AGCTAGAGTA AGTAGTTCGC GTCGGTCGGC CTTCCCGGCT CGCGTCTTCA CCAGGACGTT GAAATAGGCG GAGGTAGGTC AGATAATTAA CAACGGCCCT TCGATCTCAT TCATCAAGCG CAGTTAATAG TTTGCGCAAC GTTGTTGCCA TTGCTACAGG CATCGTGGTG TCACGCTCGT CGTTTGGTAT GGCTTCATTC AGCTCCGGTT CCCAACGATC GTCAATTATC AAACGCGTTG CAACAACGGT AACGATGTCC GTAGCACCAC AGTGCGAGCA GCAAACCATA CCGAAGTAAG TCGAGGCCAA GGGTTGCTAG AAGGCGAGTT ACATGATCCC CCATGTTGTG CAAAAAAGCG GTTAGCTCCT TCGGTCCTCC GATCGTTGTC AGAAGTAAGT TGGCCGCAGT GTTATCACTC TTCCGCTCAA TGTACTAGGG GGTACAACAC GTTTTTTCGC CAATCGAGGA AGCCAGGAGG CTAGCAACAG TCTTCATTCA ACCGGCGTCA CAATAGTGAG ATGGTTATGG CAGCACTGCA TAATTCTCTT ACTGTCATGC CATCCGTAAG ATGCTTTTCT GTGACTGGTG AGTACTCAAC CAAGTCATTC TGAGAATAGT TACCAATACC GTCGTGACGT ATTAAGAGAA TGACAGTACG GTAGGCATTC TACGAAAAGA CACTGACCAC TCATGAGTTG GTTCAGTAAG ACTCTTATCA GTATGCGGCG ACCGAGTTGC TCTTGCCCGG CGTCAATACG GGATAATACC GCGCCACATA GCAGAACTTT AAAAGTGCTC ATCATTGGAA AACGTTCTTC CATACGCCGC TGGCTCAACG AGAACGGGCC GCAGTTATGC CCTATTATGG CGCGGTGTAT CGTCTTGAAA TTTTCACGAG TAGTAACCTT TTGCAAGAAG GGGGCGAAAA CTCTCAAGGA TCTTACCGCT GTTGAGATCC AGTTCGATGT AACCCACTCG TGCACCCAAC TGATCTTCAG CATCTTTTAC TTTCACCAGC CCCCGCTTTT GAGAGTTCCT AGAATGGCGA CAACTCTAGG TCAAGCTACA TTGGGTGAGC ACGTGGGTTG ACTAGAAGTC GTAGAAAATG AAAGTGGTCG GTTTCTGGGT GAGCAAAAAC AGGAAGGCAA AATGCCGCAA AAAAGGGAAT AAGGGCGACA CGGAAATGTT GAATACTCAT ACTCTTCCTT TTTCAATATT CAAAGACCCA CTCGTTTTTG TCCTTCCGTT TTACGGCGTT TTTTCCCTTA TTCCCGCTGT GCCTTTACAA CTTATGAGTA TGAGAAGGAA AAAGTTATAA ATTGAAGCAT TTATCAGGGT TATTGTCTCA TGAGCGGATA CATATTTGAA TGTATTTAGA AAAATAAACA AATAGGGGTT CCGCGCACAT TTCCCCGAAA TAACTTCGTA AATAGTCCCA ATAACAGAGT ACTCGCCTAT GTATAAACTT ACATAAATCT TTTTATTTGT TTATCCCCAA GGCGCGTGTA AAGGGGCTTT AGTGCCACCT GACGTCTAAG AAACCATTAT TATCATGACA TTAACCTATA AAAATAGGCG TATCACGAGG CCCTTTCGTC TCACGGTGGA CTGCAGATTC TTTGGTAATA ATAGTACTGT AATTGGATAT TTTTATCCGC ATAGTGCTCC GGGAAAGCAG SEQ ID: 02 GGGGAGCCGC GCCGAAGGCG TGGGGGAACC CCGCAGGGGT GCCCTTCTTT GGGCACCAAA GAACTAGATA TAGGGCGAAA TGCGAAAGAC TTAAAAATCA CCCCTCGGCG CGGCTTCCGC ACCCCCTTGG GGCGTCCCCA CGGGAAGAAA CCCGTGGTTT CTTGATCTAT ATCCCGCTTT ACGCTTTCTG AATTTTTAGT ACAACTTAAA AAAGGGGGGT ACGCAACAGC TCATTGCGGC ACCCCCCGCA ATAGCTCATT GCGTAGGTTA AAGAAAATCT GTAATTGACT GCCACTTTTA TGTTGAATTT TTTCCCCCCA TGCGTTGTCG AGTAACGCCG TGGGGGGCGT TATCGAGTAA CGCATCCAAT TTCTTTTAGA CATTAACTGA CGGTGAAAAT CGCAACGCAT AATTGTTGTC GCGCTGCCGA AAAGTTGCAG CTGATTGCGC ATGGTGCCGC AACCGTGCGG CACCCTACCG CATGGAGATA AGCATGGCCA GCGTTGCGTA TTAACAACAG CGCGACGGCT TTTCAACGTC GACTAACGCG TACCACGGCG TTGGCACGCC GTGGGATGGC GTACCTCTAT TCGTACCGGT CGCAGTCCAG AGAAATCGGC ATTCAAGCCA AGAACAAGCC CGGTCACTGG GTGCAAACGG AACGCAAAGC GCATGAGGCG TGGGCCGGGC TTATTGCGAG GCGTCAGGTC TCTTTAGCCG TAAGTTCGGT TCTTGTTCGG GCCAGTGACC CACGTTTGCC TTGCGTTTCG CGTACTCCGC ACCCGGCCCG AATAACGCTC GAAACCCACG GCGGCAATGC TGCTGCATCA CCTCGTGGCG CAGATGGGCC ACCAGAACGC CGTGGTGGTC AGCCAGAAGA CACTTTCCAA GCTCATCGGA CTTTGGGTGC CGCCGTTACG ACGACGTAGT GGAGCACCGC GTCTACCCGG TGGTCTTGCG GCACCACCAG TCGGTCTTCT GTGAAAGGTT CGAGTAGCCT CGTTCTTTGC GGACGGTCCA ATACGCAGTC AAGGACTTGG TGGCCGAGCG CTGGATCTCC GTCGTGAAGC TCAACGGCCC CGGCACCGTG TCGGCCTACG GCAAGAAACG CCTGCCAGGT TATGCGTCAG TTCCTGAACC ACCGGCTCGC GACCTAGAGG CAGCACTTCG AGTTGCCGGG GCCGTGGCAC AGCCGGATGC TGGTCAATGA CCGCGTGGCG TGGGGCCAGC CCCGCGACCA GTTGCGCCTG TCGGTGTTCA GTGCCGCCGT GGTGGTTGAT CACGACGACC AGGACGAATC ACCAGTTACT GGCGCACCGC ACCCCGGTCG GGGCGCTGGT CAACGCGGAC AGCCACAAGT CACGGCGGCA CCACCAACTA GTGCTGCTGG TCCTGCTTAG GCTGTTGGGG CATGGCGACC TGCGCCGCAT CCCGACCCTG TATCCGGGCG AGCAGCAACT ACCGACCGGC CCCGGCGAGG AGCCGCCCAG CCAGCCCGGC CGACAACCCC GTACCGCTGG ACGCGGCGTA GGGCTGGGAC ATAGGCCCGC TCGTCGTTGA TGGCTGGCCG GGGCCGCTCC TCGGCGGGTC GGTCGGGCCG ATTCCGGGCA TGGAACCAGA CCTGCCAGCC TTGACCGAAA CGGAGGAATG GGAACGGCGC GGGCAGCAGC GCCTGCCGAT GCCCGATGAG CCGTGTTTTC TAAGGCCCGT ACCTTGGTCT GGACGGTCGG
AACTGGCTTT GCCTCCTTAC CCTTGCCGCG CCCGTCGTCG CGGACGGCTA CGGGCTACTC GGCACAAAAG TGGACGATGG CGAGCCGTTG GAGCCGCCGA CACGGGTCAC GCTGCCGCGC CGGTAGCACT TGGGTTGCGC AGCAACCCGT AAGTGCGCTG TTCCAGACTA ACCTGCTACC GCTCGGCAAC CTCGGCGGCT GTGCCCAGTG CGACGGCGCG GCCATCGTGA ACCCAACGCG TCGTTGGGCA TTCACGCGAC AAGGTCTGAT TCGGCTGTAG CCGCCTCGCC GCCCTATACC TTGTCTGCCT CCCCGCGTTG CGTCGCGGTG CATGGAGCCG GGCCACCTCG ACCTGAATGG AAGCCGGCGG AGCCGACATC GGCGGAGCGG CGGGATATGG AACAGACGGA GGGGCGCAAC GCAGCGCCAC GTACCTCGGC CCGGTGGAGC TGGACTTACC TTCGGCCGCC CACCTCGCTA ACGGATTCAC CGTTTTTATC AGGCTCTGGG AGGCAGAATA AATGATCATA TCGTCAATTA TTACCTCCAC GGGGAGAGCC TGAGCAAACT GTGGAGCGAT TGCCTAAGTG GCAAAAATAG TCCGAGACCC TCCGTCTTAT TTACTAGTAT AGCAGTTAAT AATGGAGGTG CCCCTCTCGG ACTCGTTTGA GGCCTCAGGC ATTTGAGAAG CACACGGTCA CACTGCTTCC GGTAGTCAAT AAACCGGTAA ACCAGCAATA GACATAAGCG GCTATTTAAC GACCCTGCCC CCGGAGTCCG TAAACTCTTC GTGTGCCAGT GTGACGAAGG CCATCAGTTA TTTGGCCATT TGGTCGTTAT CTGTATTCGC CGATAAATTG CTGGGACGGG TGAACCGACG ACCGGGTCGA ATTTGCTTTC GAATTTCTGC CATTCATCCG CTTATTATCA CTTATTCAGG CGTAGCACCA GGCGTTTAAG GGCACCAATA ACTTGGCTGC TGGCCCAGCT TAAACGAAAG CTTAAAGACG GTAAGTAGGC GAATAATAGT GAATAAGTCC GCATCGTGGT CCGCAAATTC CCGTGGTTAT ACTGCCTTAA AAAAATTACG CCCCGCCCTG CCACTCATCG CAGTCGGCCT ATTGGTTAAA AAATGAGCTG ATTTAACAAA AATTTAACGC GAATTTTAAC TGACGGAATT TTTTTAATGC GGGGCGGGAC GGTGAGTAGC GTCAGCCGGA TAACCAATTT TTTACTCGAC TAAATTGTTT TTAAATTGCG CTTAAAATTG AAAATATTAA CGCTTACAAT TTCCATTCGC CATTCAGGCT GCGCAACTGT TGGGAAGGGC GATCGGTGCG GGCCTCTTCG CTATTACGCC AGCTGGCGAA TTTTATAATT GCGAATGTTA AAGGTAAGCG GTAAGTCCGA CGCGTTGACA ACCCTTCCCG CTAGCCACGC CCGGAGAAGC GATAATGCGG TCGACCGCTT AGGGGGATGT GCTGCAAGGC GATTAAGTTG GGTAACGCCA GGGTTTTCCC AGTCACGACG TTGTAAAACG ACGGCCAGTG AGCGCGCGTA ATACGACTCA TCCCCCTACA CGACGTTCCG CTAATTCAAC CCATTGCGGT CCCAAAAGGG TCAGTGCTGC AACATTTTGC TGCCGGTCAC TCGCGCGCAT TATGCTGAGT CTATAGGGCG AATTGGAGCT CCACCGCGGT GGCGGCCGCT CTAGAACTAG TGGATCCCCC GGGCTGCAGG AATTCGATAT CAAGCTTATC GATACCGTCG GATATCCCGC TTAACCTCGA GGTGGCGCCA CCGCCGGCGA GATCTTGATC ACCTAGGGGG CCCGACGTCC TTAAGCTATA GTTCGAATAG CTATGGCAGC ACCTCGAGGG GGGGCCCGGT ACCCAGCTTT TGTTCCCTTT AGTGAGGGTT AATTGCGCGC TTGGCGTAAT CATGGTCATA GCTGTTTCCT GTGTGAAATT TGGAGCTCCC CCCCGGGCCA TGGGTCGAAA ACAAGGGAAA TCACTCCCAA TTAACGCGCG AACCGCATTA GTACCAGTAT CGACAAAGGA CACACTTTAA GTTATCCGCT CACAATTCCA CACAACATAC GAGCCGGAAG CATAAAGTGT AAAGCCTGGG GTGCCTAATG AGTGAGCTAA CTCACATTAA TTGCGTTGCG CAATAGGCGA GTGTTAAGGT GTGTTGTATG CTCGGCCTTC GTATTTCACA TTTCGGACCC CACGGATTAC TCACTCGATT GAGTGTAATT AACGCAACGC CTCACTGCCC GCTTTCCAGT CGGGAAACCT GTCGTGCCAG CTGCATTAAT GAATCGGCCA ACGCGCGGGG AGAGGCGGTT TGCGTATTGG GCGCATGCAT GAGTGACGGG CGAAAGGTCA GCCCTTTGGA CAGCACGGTC GACGTAATTA CTTAGCCGGT TGCGCGCCCC TCTCCGCCAA ACGCATAACC CGCGTACGTA AAAAACTGTT GTAATTCATT AAGCATTCTG CCGACATGGA AGCCATCACA AACGGCATGA TGAACCTGAA TCGCCAGCGG CATCAGCACC TTGTCGCCTT TTTTTGACAA CATTAAGTAA TTCGTAAGAC GGCTGTACCT TCGGTAGTGT TTGCCGTACT ACTTGGACTT AGCGGTCGCC GTAGTCGTGG AACAGCGGAA GCGTATAATA TTTGCCCATG GGGGTGGGCG AAGAACTCCA GCATGAGATC CCCGCGCTGG AGGATCATCC AGCCGGCGTC CCGGAAAACG ATTCCGAAGC CGCATATTAT AAACGGGTAC CCCCACCCGC TTCTTGAGGT CGTACTCTAG GGGCGCGACC TCCTAGTAGG TCGGCCGCAG GGCCTTTTGC TAAGGCTTCG CCAACCTTTC ATAGAAGGCG GCGGTGGAAT CGAAATCTCG TGATGGCAGG TTGGGCGTCG CTTGGTCGGT CATTTCGAAC CCCAGAGTCC CGCTCAGAAG GGTTGGAAAG TATCTTCCGC CGCCACCTTA GCTTTAGAGC ACTACCGTCC AACCCGCAGC GAACCAGCCA GTAAAGCTTG GGGTCTCAGG GCGAGTCTTC AACTCGTCAA GAAGGCGATA GAAGGCGATG CGCTGCGAAT CGGGAGCGGC GATACCGTAA AGCACGAGGA AGCGGTCAGC CCATTCGCCG CCAAGCTCTT TTGAGCAGTT CTTCCGCTAT CTTCCGCTAC GCGACGCTTA GCCCTCGCCG CTATGGCATT TCGTGCTCCT TCGCCAGTCG GGTAAGCGGC GGTTCGAGAA CAGCAATATC ACGGGTAGCC AACGCTATGT CCTGATAGCG GTCCGCCACA CCCAGCCGGC CACAGTCGAT GAATCCAGAA AAGCGGCCAT TTTCCACCAT GTCGTTATAG TGCCCATCGG TTGCGATACA GGACTATCGC CAGGCGGTGT GGGTCGGCCG GTGTCAGCTA CTTAGGTCTT TTCGCCGGTA AAAGGTGGTA GATATTCGGC AAGCAGGCAT CGCCATGGGT CACGACGAGA TCCTCGCCGT CGGGCATGCG CGCCTTGAGC CTGGCGAACA GTTCGGCTGG CGCGAGCCCC CTATAAGCCG TTCGTCCGTA GCGGTACCCA GTGCTGCTCT AGGAGCGGCA GCCCGTACGC GCGGAACTCG GACCGCTTGT CAAGCCGACC GCGCTCGGGG TGATGCTCTT CGTCCAGATC ATCCTGATCG ACAAGACCGG CTTCCATCCG AGTACGTGCT CGCTCGATGC GATGTTTCGC TTGGTGGTCG AATGGGCAGG ACTACGAGAA GCAGGTCTAG TAGGACTAGC TGTTCTGGCC GAAGGTAGGC TCATGCACGA GCGAGCTACG CTACAAAGCG AACCACCAGC TTACCCGTCC TAGCCGGATC AAGCGTATGC AGCCGCCGCA TTGCATCAGC CATGATGGAT ACTTTCTCGG CAGGAGCAAG GTGAGATGAC AGGAGATCCT GCCCCGGCAC ATCGGCCTAG TTCGCATACG TCGGCGGCGT AACGTAGTCG GTACTACCTA TGAAAGAGCC GTCCTCGTTC CACTCTACTG TCCTCTAGGA CGGGGCCGTG TTCGCCCAAT AGCAGCCAGT CCCTTCCCGC TTCAGTGACA ACGTCGAGCA CAGCTGCGCA AGGAACGCCC GTCGTGGCCA GCCACGATAG CCGCGCTGCC AAGCGGGTTA TCGTCGGTCA GGGAAGGGCG AAGTCACTGT TGCAGCTCGT GTCGACGCGT TCCTTGCGGG CAGCACCGGT CGGTGCTATC GGCGCGACGG TCGTCCTGCA GTTCATTCAG GGCACCGGAC AGGTCGGTCT TGACAAAAAG AACCGGGCGC CCCTGCGCTG ACAGCCGGAA CACGGCGGCA TCAGAGCAGC AGCAGGACGT CAAGTAAGTC CCGTGGCCTG TCCAGCCAGA ACTGTTTTTC TTGGCCCGCG GGGACGCGAC TGTCGGCCTT GTGCCGCCGT AGTCTCGTCG CGATTGTCTG TTGTGCCCAG TCATAGCCGA ATAGCCTCTC CACCCAAGCG GCCGGAGAAC CTGCGTGCAA TCCATCTTGT TCAATCATGC GAAACGATCC GCTAACAGAC AACACGGGTC AGTATCGGCT TATCGGAGAG GTGGGTTCGC CGGCCTCTTG GACGCACGTT AGGTAGAACA AGTTAGTACG CTTTGCTAGG TCATCCTGTC TCTTGATCAG ATCTTGATCC CCTGCGCCAT CAGATCCTTG GCGGCAAGAA AGCCATCCAG TTTACTTTGC AGGGCTTCCC AACCTTACCA AGTAGGACAG AGAACTAGTC TAGAACTAGG GGACGCGGTA GTCTAGGAAC CGCCGTTCTT TCGGTAGGTC AAATGAAACG TCCCGAAGGG TTGGAATGGT GAGGGCGCCC CAGCTGGCAA TTCCGGTTCG CTTGCTGTCC ATAAAACCGC CCAGTCTAGC TATCGCCATG TAAGCCCACT GCAAGCTACC TGCTTTCTCT CTCCCGCGGG GTCGACCGTT AAGGCCAAGC GAACGACAGG TATTTTGGCG GGTCAGATCG ATAGCGGTAC ATTCGGGTGA CGTTCGATGG ACGAAAGAGA TTGCGCTTGC GTTTTCCCTT GTCCAGATAG CCCAGTAGCT GACATTCATC CCAGGTGGCA CTTTTCGGGG AAATGTGCGC GCCCGCGTTC CTGCTGGCGC AACGCGAACG CAAAAGGGAA CAGGTCTATC GGGTCATCGA CTGTAAGTAG GGTCCACCGT GAAAAGCCCC TTTACACGCG CGGGCGCAAG GACGACCGCG TGGGCCTGTT TCTGGCGCTG GACTTCCCGC TGTTCCGTCA GCAGCTTTTC GCCCACGGCC TTGATGATCG CGGCGGCCTT GGCCTGCATA TCCCGATTCA ACCCGGACAA AGACCGCGAC CTGAAGGGCG ACAAGGCAGT CGTCGAAAAG CGGGTGCCGG AACTACTAGC GCCGCCGGAA CCGGACGTAT AGGGCTAAGT ACGGCCCCAG GGCGTCCAGA ACGGGCTTCA GGCGCTCCCG AAGGTCTCGG GCCGTCTCTT GGGCTTGATC GGCCTTCTTG CGCATCTCAC GCGCTCCTGC TGCCGGGGTC CCGCAGGTCT TGCCCGAAGT CCGCGAGGGC TTCCAGAGCC CGGCAGAGAA CCCGAACTAG CCGGAAGAAC GCGTAGAGTG CGCGAGGACG GGCGGCCTGT AGGGCAGGCT CATACCCCTG CCGAACCGCT TTTGTCAGCC GGTCGGCCAC GGCTTCCGGC GTCTCAACGC GCTTTGAGAT TCCCAGCTTT CCGCCGGACA TCCCGTCCGA GTATGGGGAC GGCTTGGCGA AAACAGTCGG CCAGCCGGTG CCGAAGGCCG CAGAGTTGCG CGAAACTCTA AGGGTCGAAA TCGGCCAATC CCTGCGGTGC ATAGGCGCGT GGCTCGACCG CTTGCGGGCT GATGGTGACG TGGCCCACTG GTGGCCGCTC CAGGGCCTCG TAGAACGCCT AGCCGGTTAG
GGACGCCACG TATCCGCGCA CCGAGCTGGC GAACGCCCGA CTACCACTGC ACCGGGTGAC CACCGGCGAG GTCCCGGAGC ATCTTGCGGA GAATGCGCGT GTGACGTGCC TTGCTGCCCT CGATGCCCCG TTGCAGCCCT AGATCGGCCA CAGCGGCCGC AAACGTGGTC TGGTCGCGGG TCATCTGCGC CTTACGCGCA CACTGCACGG AACGACGGGA GCTACGGGGC AACGTCGGGA TCTAGCCGGT GTCGCCGGCG TTTGCACCAG ACCAGCGCCC AGTAGACGCG TTTGTTGCCG ATGAACTCCT TGGCCGACAG CCTGCCGTCC TGCGTCAGCG GCACCACGAA CGCGGTCATG TGCGGGCTGG TTTCGTCACG GTGGATGCTG AAACAACGGC TACTTGAGGA ACCGGCTGTC GGACGGCAGG ACGCAGTCGC CGTGGTGCTT GCGCCAGTAC ACGCCCGACC AAAGCAGTGC CACCTACGAC GCCGTCACGA TGCGATCCGC CCCGTACTTG TCCGCCAGCC ACTTGTGCGC CTTCTCGAAG AACGCCGCCT GCTGTTCTTG GCTGGCCGAC TTCCACCATT CGGCAGTGCT ACGCTAGGCG GGGCATGAAC AGGCGGTCGG TGAACACGCG GAAGAGCTTC TTGCGGCGGA CGACAAGAAC CGACCGGCTG AAGGTGGTAA CCGGGCTGGC CGTCATGACG TACTCGACCG CCAACACAGC GTCCTTGCGC CGCTTCTCTG GCAGCAACTC GCGCAGTCGG CCCATCGCTT CATCGGTGCT GGCCCGACCG GCAGTACTGC ATGAGCTGGC GGTTGTGTCG CAGGAACGCG GCGAAGAGAC CGTCGTTGAG CGCGTCAGCC GGGTAGCGAA GTAGCCACGA GCTGGCCGCC CAGTGCTCGT TCTCTGGCGT CCTGCTGGCG TCAGCGTTGG GCGTCTCGCG CTCGCGGTAG GCGTGCTTGA GACTGGCCGC CACGTTGCCC CGACCGGCGG GTCACGAGCA AGAGACCGCA GGACGACCGC AGTCGCAACC CGCAGAGCGC GAGCGCCATC CGCACGAACT CTGACCGGCG GTGCAACGGG ATTTTCGCCA GCTTCTTGCA TCGCATGATC GCGTATGCCG CCATGCCTGC CCCTCCCTTT TGGTGTCCAA CCGGCTCGAC GGGGGCAGCG CAAGGCGGTG TAAAAGCGGT CGAAGAACGT AGCGTACTAG CGCATACGGC GGTACGGACG GGGAGGGAAA ACCACAGGTT GGCCGAGCTG CCCCCGTCGC GTTCCGCCAC CCTCCGGCGG GCCACTCAAT GCTTGAGTAT ACTCACTAGA CTTTGCTTCG CAAAGTCGTG ACCGCCTACG GCGGCTGCGG CGCCCTACGG GCTTGCTCTC GGAGGCCGCC CGGTGAGTTA CGAACTCATA TGAGTGATCT GAAACGAAGC GTTTCAGCAC TGGCGGATGC CGCCGACGCC GCGGGATGCC CGAACGAGAG CGGGCTTCGC CCTGCGCGGT CGCTGCGCTC CCTTGCCAGC CCGTGGATAT GTGGACGATG GCCGCGAGCG GCCACCGGCT GGCTCGCTTC GCTCGGCCCG GCCCGAAGCG GGACGCGCCA GCGACGCGAG GGAACGGTCG GGCACCTATA CACCTGCTAC CGGCGCTCGC CGGTGGCCGA CCGAGCGAAG CGAGCCGGGC TGGACAACCC TGCTGGACAA GCTGATGGAC AGGCTGCGCC TGCCCACGAG CTTGACCACA GGGATTGCCC ACCGGCTACC CAGCCTTCGA CCACATACCC ACCTGTTGGG ACGACCTGTT CGACTACCTG TCCGACGCGG ACGGGTGCTC GAACTGGTGT CCCTAACGGG TGGCCGATGG GTCGGAAGCT GGTGTATGGG ACCGGCTCCA ACTGCGCGGC CTGCGGCCTT GCCCCATCAA TTTTTTTAAT TTTCTCTGGG GAAAAGCCTC CGGCCTGCGG CCTGCGCGCT TCGCTTGCCG TGGCCGAGGT TGACGCGCCG GACGCCGGAA CGGGGTAGTT AAAAAAATTA AAAGAGACCC CTTTTCGGAG GCCGGACGCC GGACGCGCGA AGCGAACGGC GTTGGACACC AAGTGGAAGG CGGGTCAAGG CTCGCGCAGC GACCGCGCAG CGGCTTGGCC TTGACGCGCC TGGAACGACC CAAGCCTATG CGAGTGGGGG CAACCTGTGG TTCACCTTCC GCCCAGTTCC GAGCGCGTCG CTGGCGCGTC GCCGAACCGG AACTGCGCGG ACCTTGCTGG GTTCGGATAC GCTCACCCCC CAGTCGAAGG CGAAGCCCGC CCGCCTGCCC CCCGAGCCTC ACGGCGGCGA GTGCGGGGGT TCCAAGGGGG CAGCGCCACC TTGGGCAAGG CCGAAGGCCG GTCAGCTTCC GCTTCGGGCG GGCGGACGGG GGGCTCGGAG TGCCGCCGCT CACGCCCCCA AGGTTCCCCC GTCGCGGTGG AACCCGTTCC GGCTTCCGGC CGCAGTCGAT CAACAAGCCC CGGAGGGGCC ACTTTTTGCC GGAGGCGTCA GCTAGTTGTT CGGGGCCTCC CCGGTGAAAA ACGGCCTC SEQ ID: 03 GGGGAGCCGC GCCGAAGGCG TGGGGGAACC CCGCAGGGGT GCCCTTCTTT GGGCACCAAA GAACTAGATA TAGGGCGAAA TGCGAAAGAC TTAAAAATCA CCCCTCGGCG CGGCTTCCGC ACCCCCTTGG GGCGTCCCCA CGGGAAGAAA CCCGTGGTTT CTTGATCTAT ATCCCGCTTT ACGCTTTCTG AATTTTTAGT ACAACTTAAA AAAGGGGGGT ACGCAACAGC TCATTGCGGC ACCCCCCGCA ATAGCTCATT GCGTAGGTTA AAGAAAATCT GTAATTGACT GCCACTTTTA TGTTGAATTT TTTCCCCCCA TGCGTTGTCG AGTAACGCCG TGGGGGGCGT TATCGAGTAA CGCATCCAAT TTCTTTTAGA CATTAACTGA CGGTGAAAAT CGCAACGCAT AATTGTTGTC GCGCTGCCGA AAAGTTGCAG CTGATTGCGC ATGGTGCCGC AACCGTGCGG CACCCTACCG CATGGAGATA AGCATGGCCA GCGTTGCGTA TTAACAACAG CGCGACGGCT TTTCAACGTC GACTAACGCG TACCACGGCG TTGGCACGCC GTGGGATGGC GTACCTCTAT TCGTACCGGT CGCAGTCCAG AGAAATCGGC ATTCAAGCCA AGAACAAGCC CGGTCACTGG GTGCAAACGG AACGCAAAGC GCATGAGGCG TGGGCCGGGC TTATTGCGAG GCGTCAGGTC TCTTTAGCCG TAAGTTCGGT TCTTGTTCGG GCCAGTGACC CACGTTTGCC TTGCGTTTCG CGTACTCCGC ACCCGGCCCG AATAACGCTC GAAACCCACG GCGGCAATGC TGCTGCATCA CCTCGTGGCG CAGATGGGCC ACCAGAACGC CGTGGTGGTC AGCCAGAAGA CACTTTCCAA GCTCATCGGA CTTTGGGTGC CGCCGTTACG ACGACGTAGT GGAGCACCGC GTCTACCCGG TGGTCTTGCG GCACCACCAG TCGGTCTTCT GTGAAAGGTT CGAGTAGCCT CGTTCTTTGC GGACGGTCCA ATACGCAGTC AAGGACTTGG TGGCCGAGCG CTGGATCTCC GTCGTGAAGC TCAACGGCCC CGGCACCGTG TCGGCCTACG GCAAGAAACG CCTGCCAGGT TATGCGTCAG TTCCTGAACC ACCGGCTCGC GACCTAGAGG CAGCACTTCG AGTTGCCGGG GCCGTGGCAC AGCCGGATGC TGGTCAATGA CCGCGTGGCG TGGGGCCAGC CCCGCGACCA GTTGCGCCTG TCGGTGTTCA GTGCCGCCGT GGTGGTTGAT CACGACGACC AGGACGAATC ACCAGTTACT GGCGCACCGC ACCCCGGTCG GGGCGCTGGT CAACGCGGAC AGCCACAAGT CACGGCGGCA CCACCAACTA GTGCTGCTGG TCCTGCTTAG GCTGTTGGGG CATGGCGACC TGCGCCGCAT CCCGACCCTG TATCCGGGCG AGCAGCAACT ACCGACCGGC CCCGGCGAGG AGCCGCCCAG CCAGCCCGGC CGACAACCCC GTACCGCTGG ACGCGGCGTA GGGCTGGGAC ATAGGCCCGC TCGTCGTTGA TGGCTGGCCG GGGCCGCTCC TCGGCGGGTC GGTCGGGCCG ATTCCGGGCA TGGAACCAGA CCTGCCAGCC TTGACCGAAA CGGAGGAATG GGAACGGCGC GGGCAGCAGC GCCTGCCGAT GCCCGATGAG CCGTGTTTTC TAAGGCCCGT ACCTTGGTCT GGACGGTCGG AACTGGCTTT GCCTCCTTAC CCTTGCCGCG CCCGTCGTCG CGGACGGCTA CGGGCTACTC GGCACAAAAG TGGACGATGG CGAGCCGTTG GAGCCGCCGA CACGGGTCAC GCTGCCGCGC CGGTAGCACT TGGGTTGCGC AGCAACCCGT AAGTGCGCTG TTCCAGACTA ACCTGCTACC GCTCGGCAAC CTCGGCGGCT GTGCCCAGTG CGACGGCGCG GCCATCGTGA ACCCAACGCG TCGTTGGGCA TTCACGCGAC AAGGTCTGAT TCGGCTGTAG CCGCCTCGCC GCCCTATACC TTGTCTGCCT CCCCGCGTTG CGTCGCGGTG CATGGAGCCG GGCCACCTCG ACCTGAATGG AAGCCGGCGG AGCCGACATC GGCGGAGCGG CGGGATATGG AACAGACGGA GGGGCGCAAC GCAGCGCCAC GTACCTCGGC CCGGTGGAGC TGGACTTACC TTCGGCCGCC CACCTCGCTA ACGGATTCAC CGTTTTTATC AGGCTCTGGG AGGCAGAATA AATGATCATA TCGTCAATTA TTACCTCCAC GGGGAGAGCC TGAGCAAACT GTGGAGCGAT TGCCTAAGTG GCAAAAATAG TCCGAGACCC TCCGTCTTAT TTACTAGTAT AGCAGTTAAT AATGGAGGTG CCCCTCTCGG ACTCGTTTGA GGCCTCAGGC ATTTGAGAAG CACACGGTCA CACTGCTTCC GGTAGTCAAT AAACCGGTAA ACCAGCAATA GACATAAGCG GCTATTTAAC GACCCTGCCC CCGGAGTCCG TAAACTCTTC GTGTGCCAGT GTGACGAAGG CCATCAGTTA TTTGGCCATT TGGTCGTTAT CTGTATTCGC CGATAAATTG CTGGGACGGG TGAACCGACG ACCGGGTCGA ATTTGCTTTC GAATTTCTGC CATTCATCCG CTTATTATCA CTTATTCAGG CGTAGCACCA GGCGTTTAAG GGCACCAATA ACTTGGCTGC TGGCCCAGCT TAAACGAAAG CTTAAAGACG GTAAGTAGGC GAATAATAGT GAATAAGTCC GCATCGTGGT CCGCAAATTC CCGTGGTTAT ACTGCCTTAA AAAAATTACG CCCCGCCCTG CCACTCATCG CAGTCGGCCT ATTGGTTAAA AAATGAGCTG ATTTAACAAA AATTTAACGC GAATTTTAAC TGACGGAATT TTTTTAATGC GGGGCGGGAC GGTGAGTAGC GTCAGCCGGA TAACCAATTT TTTACTCGAC TAAATTGTTT TTAAATTGCG CTTAAAATTG AAAATATTAA CGCTTACAAT TTCCATTCGC CATTCAGGCT GCGCAACTGT TGGGAAGGGC GATCGGTGCG GGCCTCTTCG CTATTACGCC AGCTGGCGAA TTTTATAATT GCGAATGTTA AAGGTAAGCG GTAAGTCCGA CGCGTTGACA ACCCTTCCCG CTAGCCACGC CCGGAGAAGC GATAATGCGG TCGACCGCTT AGGGGGATGT GCTGCAAGGC GATTAAGTTG GGTAACGCCA GGGTTTTCCC AGTCACGACG TTGTAAAACG ACGGCCAGTG AGCGCGCGTA ATACGACTCA TCCCCCTACA CGACGTTCCG CTAATTCAAC CCATTGCGGT CCCAAAAGGG TCAGTGCTGC AACATTTTGC TGCCGGTCAC TCGCGCGCAT TATGCTGAGT CTATAGGGCG AATTGGAGCT CCACCGCGGT GGCGGCCGCT CTAGAACTAG TGGATCCCCC GGGCTGCAGG AATTCGATAT CAAGCTTATC GATACCGTCG GATATCCCGC TTAACCTCGA GGTGGCGCCA CCGCCGGCGA GATCTTGATC ACCTAGGGGG
CCCGACGTCC TTAAGCTATA GTTCGAATAG CTATGGCAGC ACGGGCCCGG GATCCGATGC TCTTCCGCTA AGATCTTTTA CTAGTTCAGT CCATCTCGCC GTGTATGCGG GCCTGACGGA TCAACGTTCC CACCGAGCCA TGCCCGGGCC CTAGGCTACG AGAAGGCGAT TCTAGAAAAT GATCAAGTCA GGTAGAGCGG CACATACGCC CGGACTGCCT AGTTGCAAGG GTGGCTCGGT GTCGAGATGT TCATCTGGTC GGCGATCTGC CGGTACTTCA AACCTTGTTT GCGCAGTTCC ACAGCCTTCT TGCGGCGTTC CTGCGCACGA GCGATGTAGT CAGCTCTACA AGTAGACCAG CCGCTAGACG GCCATGAAGT TTGGAACAAA CGCGTCAAGG TGTCGGAAGA ACGCCGCAAG GACGCGTGCT CGCTACATCA CGCCTCGGTC TTCGGCGACG AGCCGTTTGA TGGTGCTTTT CGAGACGCCG AACTTGTCAG CCAACTCCTG CGCGGTCTGC GTGCGACGCA TCACGCGTTC GCGGAGCCAG AAGCCGCTGC TCGGCAAACT ACCACGAAAA GCTCTGCGGC TTGAACAGTC GGTTGAGGAC GCGCCAGACG CACGCTGCGT AGTGCGCAAG TGCAGCACCC ATCAGTCCGT CCCCTCTGCT GCTGCGAACA GTGCCGATCG ATCGACCTTC TTGAGCTTCG GCCGCGGCGC GGTGGCGTTC TTCCGTACCG ACGTCGTGGG TAGTCAGGCA GGGGAGACGA CGACGCTTGT CACGGCTAGC TAGCTGGAAG AACTCGAAGC CGGCGCCGCG CCACCGCAAG AAGGCATGGC CTTCCGTTTT TGCGCTGCTG CTCACTTTGC CGCGGCGTGC CTGGATTTTC GAGAACTCGG CGGCGGTGAA GGTGCGGTGG GTCCAGTGGG CGACTGATTT GAAGGCAAAA ACGCGACGAC GAGTGAAACG GCGCCGCACG GACCTAAAAG CTCTTGAGCC GCCGCCACTT CCACGCCACC CAGGTCACCC GCTGACTAAA GCCGATCTGC TCGGCCTCGG CCCGACTCAT GGGGCCGATC CCGTCGTTGG CGTCGAGGGT GAAGTTGGTC AGGGCGGTGA AGTCGGTGAC CATCTGCCGC CGGCTAGACG AGCCGGAGCC GGGCTGAGTA CCCCGGCTAG GGCAGCAACC GCAGCTCCCA CTTCAACCAG TCCCGCCACT TCAGCCACTG GTAGACGGCG CACACAGTGA TCGACGGGTA GTTCTGTTTC CGGATCTCGC GGTAGGCCCA TTCCCGGGTG CGGTCGAACA GTTCGACGTT CCGGCCCGTT TCGGTCCTGA GTGTGTCACT AGCTGCCCAT CAAGACAAAG GCCTAGAGCG CCATCCGGGT AAGGGCCCAC GCCAGCTTGT CAAGCTGCAA GGCCGGGCAA AGCCAGGACT CCTGTGTCTT GCGGCCGTAG TCCGGTGGGG CGGGGAAACG GTCACCGAGC GCTTTTGCGA GGCCTTTGAG CGAGTACGGA TCCGAGGGAC CCCAGACCGT GGACACAGAA CGCCGGCATC AGGCCACCCC GCCCCTTTGC CAGTGGCTCG CGAAAACGCT CCGGAAACTC GCTCATGCCT AGGCTCCCTG GGGTCTGGCA CGTCCAGTGC GGGTGGATCG GGTTCTGGGT GAGCTGCTGC GCGTAGCCCT GATCGGCGCC GACCACCGAG GCGATCAGCC CCTGGTTCAC CCGGTCGTAG GCAGGTCACG CCCACCTAGC CCAAGACCCA CTCGACGACG CGCATCGGGA CTAGCCGCGG CTGGTGGCTC CGCTAGTCGG GGACCAAGTG GGCCAGCATC AGCCGCAGCG GGCCCTGTCG GGCTGCCTGG AGGGTGTAGA CCGGGCTTTC GAGCAGCCAC CACAGGTGCG CGTGCTCGGT CGCGGGATTG ATCGTCATCA TCGGCGTCGC CCGGGACAGC CCGACGGACC TCCCACATCT GGCCCGAAAG CTCGTCGGTG GTGTCCACGC GCACGAGCCA GCGCCCTAAC TAGCAGTAGT CGGTCGGATC GGGCAGATCC GCGTTACGTG CGGCCCACTG CGCCTGGTCG TCGTCCACGT CGAGCACCAA GCCCAACCTG ATCGACGGGG TGCGGGCCGC GCCAGCCTAG CCCGTCTAGG CGCAATGCAC GCCGGGTGAC GCGGACCAGC AGCAGGTGCA GCTCGTGGTT CGGGTTGGAC TAGCTGCCCC ACGCCCGGCG AATGTAGCGG CGGGTGAGCG CCTCCGCGCG CGGCTGCGGC CACTGCCCGT CCCGGACGTA GTCATCCGTC GCGTGCGGGT ATTTGAACCG CCAGCGGTCC TTACATCGCC GCCCACTCGC GGAGGCGCGC GCCGACGCCG GTGACGGGCA GGGCCTGCAT CAGTAGGCAG CGCACGCCCA TAAACTTGGC GGTCGCCAGG AACCAGGCGT CAACAGCAGC GGTCATGACC GCCAAGCTAG GGCCGGATCT GTACCGATCG GGGGAGGCGC GCCGCAAATT ATTTAAGAGT CTCGCTAGCA TTGGTCCGCA GTTGTCGTCG CCAGTACTGG CGGTTCGATC CCGGCCTAGA CATGGCTAGC CCCCTCCGCG CGGCGTTTAA TAAATTCTCA GAGCGATCGT AACCATGTCA GGTGTTGCGG TGGGTTCCGG GTAAACCTCC ACCCGAATTA TTTAAGAGTC TCGCTAGCTA AGCCCTATCT GATGCTGCGC GGGGGGTCCT TTGGTACAGT CCACAACGCC ACCCAAGGCC CATTTGGAGG TGGGCTTAAT AAATTCTCAG AGCGATCGAT TCGGGATAGA CTACGACGCG CCCCCCAGGA TCGCACTGAA TCTCAAAGGT GGCCGGCTGA ATTTCGTCGC GCGAAAACCT CCCTGGACAG TTCTGGAATT CAGCAAGAGG TGTGTCTGAA CTTCGGTGTT AGCGTGACTT AGAGTTTCCA CCGGCCGACT TAAAGCAGCG CGCTTTTGGA GGGACCTGTC AAGACCTTAA GTCGTTCTCC ACACAGACTT GAAGCCACAA TTTTTGGGGG GTGACTCCAG CGGGGTGGGC ACAACGCGAA CAGAGACCTT GTGTGTACGA CGGCGGGAGG TAAGTCGGGT ACGGCTCGGA CTGCGGTAGA AAAAACCCCC CACTGAGGTC GCCCCACCCG TGTTGCGCTT GTCTCTGGAA CACACATGCT GCCGCCCTCC ATTCAGCCCA TGCCGAGCCT GACGCCATCT GCAACCGTCG AATCGATTTC GAGCAGAGCG AGCAGAGCAA GATATTCCAA AACTCCGGGG TTCCTCGGCG GCCTCCCCCG TCTGTTTGCT CAACCGAGGG CGTTGGCAGC TTAGCTAAAG CTCGTCTCGC TCGTCTCGTT CTATAAGGTT TTGAGGCCCC AAGGAGCCGC CGGAGGGGGC AGACAAACGA GTTGGCTCCC AGACCTGGCG GTCCCGCGTT TCCGGACGCG CGGGACCGCC TACCGCTCGA GAGCGGAAGA GCATCTAGAT GCATTCGCGA GGTACCCAGC TTTTGTTCCC TCTGGACCGC CAGGGCGCAA AGGCCTGCGC GCCCTGGCGG ATGGCGAGCT CTCGCCTTCT CGTAGATCTA CGTAAGCGCT CCATGGGTCG AAAACAAGGG TTTAGTGAGG GTTAATTGCG CGCTTGGCGT AATCATGGTC ATAGCTGTTT CCTGTGTGAA ATTGTTATCC GCTCACAATT CCACACAACA TACGAGCCGG AAATCACTCC CAATTAACGC GCGAACCGCA TTAGTACCAG TATCGACAAA GGACACACTT TAACAATAGG CGAGTGTTAA GGTGTGTTGT ATGCTCGGCC AAGCATAAAG TGTAAAGCCT GGGGTGCCTA ATGAGTGAGC TAACTCACAT TAATTGCGTT GCGCTCACTG CCCGCTTTCC AGTCGGGAAA CCTGTCGTGC TTCGTATTTC ACATTTCGGA CCCCACGGAT TACTCACTCG ATTGAGTGTA ATTAACGCAA CGCGAGTGAC GGGCGAAAGG TCAGCCCTTT GGACAGCACG CAGCTGCATT AATGAATCGG CCAACGCGCG GGGAGAGGCG GTTTGCGTAT TGGGCGCATG CATAAAAACT GTTGTAATTC ATTAAGCATT CTGCCGACAT GTCGACGTAA TTACTTAGCC GGTTGCGCGC CCCTCTCCGC CAAACGCATA ACCCGCGTAC GGAAGCCATC ACAAACGGCA TGATGAACCT GAATCGCCAG CGGCATCAGC ACCTTGTCGC CTTGCGTATA ATATTTGCCC ATGGGGGTGG GCGAAGAACT CCTTCGGTAG TGTTTGCCGT ACTACTTGGA CTTAGCGGTC GCCGTAGTCG TGGAACAGCG GAACGCATAT TATAAACGGG TACCCCCACC CGCTTCTTGA CCAGCATGAG ATCCCCGCGC TGGAGGATCA TCCAGCCGGC GTCCCGGAAA ACGATTCCGA AGCCCAACCT TTCATAGAAG GCGGCGGTGG AATCGAAATC GGTCGTACTC TAGGGGCGCG ACCTCCTAGT AGGTCGGCCG CAGGGCCTTT TGCTAAGGCT TCGGGTTGGA AAGTATCTTC CGCCGCCACC TTAGCTTTAG TCGTGATGGC AGGTTGGGCG TCGCTTGGTC GGTCATTTCG AACCCCAGAG TCCCGCTCAG AAGAACTCGT CAAGAAGGCG ATAGAAGGCG ATGCGCTGCG AGCACTACCG TCCAACCCGC AGCGAACCAG CCAGTAAAGC TTGGGGTCTC AGGGCGAGTC TTCTTGAGCA GTTCTTCCGC TATCTTCCGC TACGCGACGC AATCGGGAGC GGCGATACCG TAAAGCACGA GGAAGCGGTC AGCCCATTCG CCGCCAAGCT CTTCAGCAAT ATCACGGGTA GCCAACGCTA TGTCCTGATA TTAGCCCTCG CCGCTATGGC ATTTCGTGCT CCTTCGCCAG TCGGGTAAGC GGCGGTTCGA GAAGTCGTTA TAGTGCCCAT CGGTTGCGAT ACAGGACTAT GCGGTCCGCC ACACCCAGCC GGCCACAGTC GATGAATCCA GAAAAGCGGC CATTTTCCAC CATGATATTC GGCAAGCAGG CATCGCCATG GGTCACGACG CGCCAGGCGG TGTGGGTCGG CCGGTGTCAG CTACTTAGGT CTTTTCGCCG GTAAAAGGTG GTACTATAAG CCGTTCGTCC GTAGCGGTAC CCAGTGCTGC AGATCCTCGC CGTCGGGCAT GCGCGCCTTG AGCCTGGCGA ACAGTTCGGC TGGCGCGAGC CCCTGATGCT CTTCGTCCAG ATCATCCTGA TCGACAAGAC TCTAGGAGCG GCAGCCCGTA CGCGCGGAAC TCGGACCGCT TGTCAAGCCG ACCGCGCTCG GGGACTACGA GAAGCAGGTC TAGTAGGACT AGCTGTTCTG CGGCTTCCAT CCGAGTACGT GCTCGCTCGA TGCGATGTTT CGCTTGGTGG TCGAATGGGC AGGTAGCCGG ATCAAGCGTA TGCAGCCGCC GCATTGCATC GCCGAAGGTA GGCTCATGCA CGAGCGAGCT ACGCTACAAA GCGAACCACC AGCTTACCCG TCCATCGGCC TAGTTCGCAT ACGTCGGCGG CGTAACGTAG AGCCATGATG GATACTTTCT CGGCAGGAGC AAGGTGAGAT GACAGGAGAT CCTGCCCCGG CACTTCGCCC AATAGCAGCC AGTCCCTTCC CGCTTCAGT TCGGTACTAC CTATGAAAGA GCCGTCCTCG TTCCACTCTA CTGTCCTCTA GGACGGGGCC GTGAAGCGGG TTATCGTCGG TCAGGGAAGG GCGAAGTCAC ACAACGTCGA GCACAGCTGC GCAAGGAACG CCCGTCGTGG CCAGCCACGA TAGCCGCGCT GCCTCGTCCT GCAGTTCATT CAGGGCACCG GACAGGTCGG TGTTGCAGCT CGTGTCGACG CGTTCCTTGC GGGCAGCACC GGTCGGTGCT ATCGGCGCGA CGGAGCAGGA CGTCAAGTAA GTCCCGTGGC CTGTCCAGCC TCTTGACAAA AAGAACCGGG CGCCCCTGCG CTGACAGCCG GAACACGGCG GCATCAGAGC AGCCGATTGT CTGTTGTGCC CAGTCATAGC CGAATAGCCT AGAACTGTTT TTCTTGGCCC GCGGGGACGC GACTGTCGGC CTTGTGCCGC CGTAGTCTCG TCGGCTAACA GACAACACGG GTCAGTATCG GCTTATCGGA CTCCACCCAA GCGGCCGGAG
AACCTGCGTG CAATCCATCT TGTTCAATCA TGCGAAACGA TCCTCATCCT GTCTCTTGAT CAGATCTTGA TCCCCTGCGC GAGGTGGGTT CGCCGGCCTC TTGGACGCAC GTTAGGTAGA ACAAGTTAGT ACGCTTTGCT AGGAGTAGGA CAGAGAACTA GTCTAGAACT AGGGGACGCG CATCAGATCC TTGGCGGCAA GAAAGCCATC CAGTTTACTT TGCAGGGCTT CCCAACCTTA CCAGAGGGCG CCCCAGCTGG CAATTCCGGT TCGCTTGCTG GTAGTCTAGG AACCGCCGTT CTTTCGGTAG GTCAAATGAA ACGTCCCGAA GGGTTGGAAT GGTCTCCCGC GGGGTCGACC GTTAAGGCCA AGCGAACGAC TCCATAAAAC CGCCCAGTCT AGCTATCGCC ATGTAAGCCC ACTGCAAGCT ACCTGCTTTC TCTTTGCGCT TGCGTTTTCC CTTGTCCAGA TAGCCCAGTA AGGTATTTTG GCGGGTCAGA TCGATAGCGG TACATTCGGG TGACGTTCGA TGGACGAAAG AGAAACGCGA ACGCAAAAGG GAACAGGTCT ATCGGGTCAT GCTGACATTC ATCCCAGGTG GCACTTTTCG GGGAAATGTG CGCGCCCGCG TTCCTGCTGG CGCTGGGCCT GTTTCTGGCG CTGGACTTCC CGCTGTTCCG CGACTGTAAG TAGGGTCCAC CGTGAAAAGC CCCTTTACAC GCGCGGGCGC AAGGACGACC GCGACCCGGA CAAAGACCGC GACCTGAAGG GCGACAAGGC TCAGCAGCTT TTCGCCCACG GCCTTGATGA TCGCGGCGGC CTTGGCCTGC ATATCCCGAT TCAACGGCCC CAGGGCGTCC AGAACGGGCT TCAGGCGCTC AGTCGTCGAA AAGCGGGTGC CGGAACTACT AGCGCCGCCG GAACCGGACG TATAGGGCTA AGTTGCCGGG GTCCCGCAGG TCTTGCCCGA AGTCCGCGA CCGAAGGTCT CGGGCCGTCT CTTGGGCTTG ATCGGCCTTC TTGCGCATCT CACGCGCTCC TGCGGCGGCC TGTAGGGCAG GCTCATACCC CTGCCGAACC GGCTTCCAGA GCCCGGCAGA GAACCCGAAC TAGCCGGAAG AACGCGTAGA GTGCGCGAGG ACGCCGCCGG ACATCCCGTC CGAGTATGGG GACGGCTTGG GCTTTTGTCA GCCGGTCGGC CACGGCTTCC GGCGTCTCAA CGCGCTTTGA GATTCCCAGC TTTTCGGCCA ATCCCTGCGG TGCATAGGCG CGTGGCTCGA CGAAAACAGT CGGCCAGCCG GTGCCGAAGG CCGCAGAGTT GCGCGAAACT CTAAGGGTCG AAAAGCCGGT TAGGGACGCC ACGTATCCGC GCACCGAGCT CCGCTTGCGG GCTGATGGTG ACGTGGCCCA CTGGTGGCCG CTCCAGGGCC TCGTAGAACG CCTGAATGCG CGTGTGACGT GCCTTGCTGC CCTCGATGCC GGCGAACGCC CGACTACCAC TGCACCGGGT GACCACCGGC GAGGTCCCGG AGCATCTTGC GGACTTACGC GCACACTGCA CGGAACGACG GGAGCTACGG CCGTTGCAGC CCTAGATCGG CCACAGCGGC CGCAAACGTG GTCTGGTCGC GGGTCATCTG CGCTTTGTTG CCGATGAACT CCTTGGCCGA CAGCCTGCCG GGCAACGTCG GGATCTAGCC GGTGTCGCCG GCGTTTGCAC CAGACCAGCG CCCAGTAGAC GCGAAACAAC GGCTACTTGA GGAACCGGCT GTCGGACGGC TCCTGCGTCA GCGGCACCAC GAACGCGGTC ATGTGCGGGC TGGTTTCGTC ACGGTGGATG CTGGCCGTCA CGATGCGATC CGCCCCGTAC TTGTCCGCCA AGGACGCAGT CGCCGTGGTG CTTGCGCCAG TACACGCCCG ACCAAAGCAG TGCCACCTAC GACCGGCAGT GCTACGCTAG GCGGGGCATG AACAGGCGGT GCCACTTGTG CGCCTTCTCG AAGAACGCCG CCTGCTGTTC TTGGCTGGCC GACTTCCACC ATTCCGGGCT GGCCGTCATG ACGTACTCGA CCGCCAACAC CGGTGAACAC GCGGAAGAGC TTCTTGCGGC GGACGACAAG AACCGACCGG CTGAAGGTGG TAAGGCCCGA CCGGCAGTAC TGCATGAGCT GGCGGTTGTG AGCGTCCTTG CGCCGCTTCT CTGGCAGCAA CTCGCGCAGT CGGCCCATCG CTTCATCGGT GCTGCTGGCC GCCCAGTGCT CGTTCTCTGG CGTCCTGCTG TCGCAGGAAC GCGGCGAAGA GACCGTCGTT GAGCGCGTCA GCCGGGTAGC GAAGTAGCCA CGACGACCGG CGGGTCACGA GCAAGAGACC GCAGGACGAC GCGTCAGCGT TGGGCGTCTC GCGCTCGCGG TAGGCGTGCT TGAGACTGGC CGCCACGTTG CCCATTTTCG CCAGCTTCTT GCATCGCATG ATCGCGTATG CGCAGTCGCA ACCCGCAGAG CGCGAGCGCC ATCCGCACGA ACTCTGACCG GCGGTGCAAC GGGTAAAAGC GGTCGAAGAA CGTAGCGTAC TAGCGCATAC CCGCCATGCC TGCCCCTCCC TTTTGGTGTC CAACCGGCTC GACGGGGGCA GCGCAAGGCG GTGCCTCCGG CGGGCCACTC AATGCTTGAG TATACTCACT GGCGGTACGG ACGGGGAGGG AAAACCACAG GTTGGCCGAG CTGCCCCCGT CGCGTTCCGC CACGGAGGCC GCCCGGTGAG TTACGAACTC ATATGAGTGA AGACTTTGCT TCGCAAAGTC GTGACCGCCT ACGGCGGCTG CGGCGCCCTA CGGGCTTGCT CTCCGGGCTT CGCCCTGCGC GGTCGCTGCG CTCCCTTGCC TCTGAAACGA AGCGTTTCAG CACTGGCGGA TGCCGCCGAC GCCGCGGGAT GCCCGAACGA GAGGCCCGAA GCGGGACGCG CCAGCGACGC GAGGGAACGG SEQ ID: 04 GGGGAGCCGC GCCGAAGGCG TGGGGGAACC CCGCAGGGGT GCCCTTCTTT GGGCACCAAA GAACTAGATA TAGGGCGAAA TGCGAAAGAC TTAAAAATCA CCCCTCGGCG CGGCTTCCGC ACCCCCTTGG GGCGTCCCCA CGGGAAGAAA CCCGTGGTTT CTTGATCTAT ATCCCGCTTT ACGCTTTCTG AATTTTTAGT ACAACTTAAA AAAGGGGGGT ACGCAACAGC TCATTGCGGC ACCCCCCGCA ATAGCTCATT GCGTAGGTTA AAGAAAATCT GTAATTGACT GCCACTTTTA TGTTGAATTT TTTCCCCCCA TGCGTTGTCG AGTAACGCCG TGGGGGGCGT TATCGAGTAA CGCATCCAAT TTCTTTTAGA CATTAACTGA CGGTGAAAAT CGCAACGCAT AATTGTTGTC GCGCTGCCGA AAAGTTGCAG CTGATTGCGC ATGGTGCCGC AACCGTGCGG CACCCTACCG CATGGAGATA AGCATGGCCA GCGTTGCGTA TTAACAACAG CGCGACGGCT TTTCAACGTC GACTAACGCG TACCACGGCG TTGGCACGCC GTGGGATGGC GTACCTCTAT TCGTACCGGT CGCAGTCCAG AGAAATCGGC ATTCAAGCCA AGAACAAGCC CGGTCACTGG GTGCAAACGG AACGCAAAGC GCATGAGGCG TGGGCCGGGC TTATTGCGAG GCGTCAGGTC TCTTTAGCCG TAAGTTCGGT TCTTGTTCGG GCCAGTGACC CACGTTTGCC TTGCGTTTCG CGTACTCCGC ACCCGGCCCG AATAACGCTC GAAACCCACG GCGGCAATGC TGCTGCATCA CCTCGTGGCG CAGATGGGCC ACCAGAACGC CGTGGTGGTC AGCCAGAAGA CACTTTCCAA GCTCATCGGA CTTTGGGTGC CGCCGTTACG ACGACGTAGT GGAGCACCGC GTCTACCCGG TGGTCTTGCG GCACCACCAG TCGGTCTTCT GTGAAAGGTT CGAGTAGCCT CGTTCTTTGC GGACGGTCCA ATACGCAGTC AAGGACTTGG TGGCCGAGCG CTGGATCTCC GTCGTGAAGC TCAACGGCCC CGGCACCGTG TCGGCCTACG GCAAGAAACG CCTGCCAGGT TATGCGTCAG TTCCTGAACC ACCGGCTCGC GACCTAGAGG CAGCACTTCG AGTTGCCGGG GCCGTGGCAC AGCCGGATGC TGGTCAATGA CCGCGTGGCG TGGGGCCAGC CCCGCGACCA GTTGCGCCTG TCGGTGTTCA GTGCCGCCGT GGTGGTTGAT CACGACGACC AGGACGAATC ACCAGTTACT GGCGCACCGC ACCCCGGTCG GGGCGCTGGT CAACGCGGAC AGCCACAAGT CACGGCGGCA CCACCAACTA GTGCTGCTGG TCCTGCTTAG GCTGTTGGGG CATGGCGACC TGCGCCGCAT CCCGACCCTG TATCCGGGCG AGCAGCAACT ACCGACCGGC CCCGGCGAGG AGCCGCCCAG CCAGCCCGGC CGACAACCCC GTACCGCTGG ACGCGGCGTA GGGCTGGGAC ATAGGCCCGC TCGTCGTTGA TGGCTGGCCG GGGCCGCTCC TCGGCGGGTC GGTCGGGCCG ATTCCGGGCA TGGAACCAGA CCTGCCAGCC TTGACCGAAA CGGAGGAATG GGAACGGCGC GGGCAGCAGC GCCTGCCGAT GCCCGATGAG CCGTGTTTTC TAAGGCCCGT ACCTTGGTCT GGACGGTCGG AACTGGCTTT GCCTCCTTAC CCTTGCCGCG CCCGTCGTCG CGGACGGCTA CGGGCTACTC GGCACAAAAG TGGACGATGG CGAGCCGTTG GAGCCGCCGA CACGGGTCAC GCTGCCGCGC CGGTAGCACT TGGGTTGCGC AGCAACCCGT AAGTGCGCTG TTCCAGACTA ACCTGCTACC GCTCGGCAAC CTCGGCGGCT GTGCCCAGTG CGACGGCGCG GCCATCGTGA ACCCAACGCG TCGTTGGGCA TTCACGCGAC AAGGTCTGAT TCGGCTGTAG CCGCCTCGCC GCCCTATACC TTGTCTGCCT CCCCGCGTTG CGTCGCGGTG CATGGAGCCG GGCCACCTCG ACCTGAATGG AAGCCGGCGG AGCCGACATC GGCGGAGCGG CGGGATATGG AACAGACGGA GGGGCGCAAC GCAGCGCCAC GTACCTCGGC CCGGTGGAGC TGGACTTACC TTCGGCCGCC CACCTCGCTA ACGGATTCAC CGTTTTTATC AGGCTCTGGG AGGCAGAATA AATGATCATA TCGTCAATTA TTACCTCCAC GGGGAGAGCC TGAGCAAACT GTGGAGCGAT TGCCTAAGTG GCAAAAATAG TCCGAGACCC TCCGTCTTAT TTACTAGTAT AGCAGTTAAT AATGGAGGTG CCCCTCTCGG ACTCGTTTGA GGCCTCAGGC ATTTGAGAAG CACACGGTCA CACTGCTTCC GGTAGTCAAT AAACCGGTAA ACCAGCAATA GACATAAGCG GCTATTTAAC GACCCTGCCC CCGGAGTCCG TAAACTCTTC GTGTGCCAGT GTGACGAAGG CCATCAGTTA TTTGGCCATT TGGTCGTTAT CTGTATTCGC CGATAAATTG CTGGGACGGG TGAACCGACG ACCGGGTCGA ATTTGCTTTC GAATTTCTGC CATTCATCCG CTTATTATCA CTTATTCAGG CGTAGCACCA GGCGTTTAAG GGCACCAATA ACTTGGCTGC TGGCCCAGCT TAAACGAAAG CTTAAAGACG GTAAGTAGGC GAATAATAGT GAATAAGTCC GCATCGTGGT CCGCAAATTC CCGTGGTTAT ACTGCCTTAA AAAAATTACG CCCCGCCCTG CCACTCATCG CAGTCGGCCT ATTGGTTAAA AAATGAGCTG ATTTAACAAA AATTTAACGC GAATTTTAAC TGACGGAATT TTTTTAATGC GGGGCGGGAC GGTGAGTAGC GTCAGCCGGA TAACCAATTT TTTACTCGAC TAAATTGTTT TTAAATTGCG CTTAAAATTG AAAATATTAA CGCTTACAAT TTCCATTCGC CATTCAGGCT GCGCAACTGT TGGGAAGGGC GATCGGTGCG GGCCTCTTCG CTATTACGCC AGCTGGCGAA TTTTATAATT GCGAATGTTA AAGGTAAGCG GTAAGTCCGA CGCGTTGACA ACCCTTCCCG
CTAGCCACGC CCGGAGAAGC GATAATGCGG TCGACCGCTT AGGGGGATGT GCTGCAAGGC GATTAAGTTG GGTAACGCCA GGGTTTTCCC AGTCACGACG TTGTAAAACG ACGGCCAGTG AGCGCGCGTA ATACGACTCA TCCCCCTACA CGACGTTCCG CTAATTCAAC CCATTGCGGT CCCAAAAGGG TCAGTGCTGC AACATTTTGC TGCCGGTCAC TCGCGCGCAT TATGCTGAGT CTATAGGGCG AATTGGAGCT CCACCGCGGT GGCGGCCGCT CTAGAACTAG TGGATCCCCC GGGCTGCAGG AATTCGATAT CAAGCTTTTA CGCCCCGCCC GATATCCCGC TTAACCTCGA GGTGGCGCCA CCGCCGGCGA GATCTTGATC ACCTAGGGGG CCCGACGTCC TTAAGCTATA GTTCGAAAAT GCGGGGCGGG TGCCACTCAT CGCAGTACTG TTGTAATTCA TTAAGCATTC TGCCGACATG GAAGCCATCA CAAACGGCAT GATGAACCTG AATCGCCAGC GGCATCAGCA ACGGTGAGTA GCGTCATGAC AACATTAAGT AATTCGTAAG ACGGCTGTAC CTTCGGTAGT GTTTGCCGTA CTACTTGGAC TTAGCGGTCG CCGTAGTCGT CCTTGTCGCC TTGCGTATAA TATTTGCCCA TGGTGAAAAC GGGGGCGAAG AAGTTGTCCA TATTGGCCAC GTTTAAATCA AAACTGGTGA AACTCACCCA GGAACAGCGG AACGCATATT ATAAACGGGT ACCACTTTTG CCCCCGCTTC TTCAACAGGT ATAACCGGTG CAAATTTAGT TTTGACCACT TTGAGTGGGT GGGATTGGCT GAGACGAAAA ACATATTCTC AATAAACCCT TTAGGGAAAT AGGCCAGGTT TTCACCGTAA CACGCCACAT CTTGCGAATA TATGTGTAGA CCCTAACCGA CTCTGCTTTT TGTATAAGAG TTATTTGGGA AATCCCTTTA TCCGGTCCAA AAGTGGCATT GTGCGGTGTA GAACGCTTAT ATACACATCT AACTGCCGGA AATCGTCGTG GTATTCACTC CAGAGCGATG AAAACGTTTC AGTTTGCTCA TGGAAAACGG TGTAACAAGG GTGAACACTA TCCCATATCA TTGACGGCCT TTAGCAGCAC CATAAGTGAG GTCTCGCTAC TTTTGCAAAG TCAAACGAGT ACCTTTTGCC ACATTGTTCC CACTTGTGAT AGGGTATAGT CCAGCTCACC GTCTTTCATT GCCATACGAA ATTCCGGATG AGCATTCATC AGGCGGGCAA GAATGTGAAT AAAGGCCGGA TAAAACTTGT GCTTATTTTT GGTCGAGTGG CAGAAAGTAA CGGTATGCTT TAAGGCCTAC TCGTAAGTAG TCCGCCCGTT CTTACACTTA TTTCCGGCCT ATTTTGAACA CGAATAAAAA CTTTACGGTC TTTAAAAAGG CCGTAATATC CAGCTGAACG GTCTGGTTAT AGGTACATTG AGCAACTGAC TGAAATGCCT CAAAATGTTC TTTACGATGC GAAATGCCAG AAATTTTTCC GGCATTATAG GTCGACTTGC CAGACCAATA TCCATGTAAC TCGTTGACTG ACTTTACGGA GTTTTACAAG AAATGCTACG CATTGGGATA TATCAACGGT GGTATATCCA GTGATTTTTT TCTCCATATG GTTAACCTTA ATTAAGGGGT CGACGGGCCC GGGATCCGAT GCTCTTCCGC GTAACCCTAT ATAGTTGCCA CCATATAGGT CACTAAAAAA AGAGGTATAC CAATTGGAAT TAATTCCCCA GCTGCCCGGG CCCTAGGCTA CGAGAAGGCG TAAGATCTTT TACTAGTTCA GTCCATCTCG CCGTGTATGC GGGCCTGACG GATCAACGTT CCCACCGAGC CAGTCGAGAT GTTCATCTGG TCGGCGATCT ATTCTAGAAA ATGATCAAGT CAGGTAGAGC GGCACATACG CCCGGACTGC CTAGTTGCAA GGGTGGCTCG GTCAGCTCTA CAAGTAGACC AGCCGCTAGA GCCGGTACTT CAAACCTTGT TTGCGCAGTT CCACAGCCTT CTTGCGGCGT TCCTGCGCAC GAGCGATGTA GTCGCCTCGG TCTTCGGCGA CGAGCCGTTT CGGCCATGAA GTTTGGAACA AACGCGTCAA GGTGTCGGAA GAACGCCGCA AGGACGCGTG CTCGCTACAT CAGCGGAGCC AGAAGCCGCT GCTCGGCAAA GATGGTGCTT TTCGAGACGC CGAACTTGTC AGCCAACTCC TGCGCGGTCT GCGTGCGACG CATCACGCGT TCTGCAGCAC CCATCAGTCC GTCCCCTCTG CTACCACGAA AAGCTCTGCG GCTTGAACAG TCGGTTGAGG ACGCGCCAGA CGCACGCTGC GTAGTGCGCA AGACGTCGTG GGTAGTCAGG CAGGGGAGAC CTGCTGCGAA CAGTGCCGAT CGATCGACCT TCTTGAGCTT CGGCCGCGGC GCGGTGGCGT TCTTCCGTAC CGCTTCCGTT TTTGCGCTGC TGCTCACTTT GACGACGCTT GTCACGGCTA GCTAGCTGGA AGAACTCGAA GCCGGCGCCG CGCCACCGCA AGAAGGCATG GCGAAGGCAA AAACGCGACG ACGAGTGAAA GCCGCGGCGT GCCTGGATTT TCGAGAACTC GGCGGCGGTG AAGGTGCGGT GGGTCCAGTG GGCGACTGAT TTGCCGATCT GCTCGGCCTC GGCCCGACTC CGGCGCCGCA CGGACCTAAA AGCTCTTGAG CCGCCGCCAC TTCCACGCCA CCCAGGTCAC CCGCTGACTA AACGGCTAGA CGAGCCGGAG CCGGGCTGAG ATGGGGCCGA TCCCGTCGTT GGCGTCGAGG GTGAAGTTGG TCAGGGCGGT GAAGTCGGTG ACCATCTGCC GCCACACAGT GATCGACGGG TAGTTCTGTT TACCCCGGCT AGGGCAGCAA CCGCAGCTCC CACTTCAACC AGTCCCGCCA CTTCAGCCAC TGGTAGACGG CGGTGTGTCA CTAGCTGCCC ATCAAGACAA TCCGGATCTC GCGGTAGGCC CATTCCCGGG TGCGGTCGAA CAGTTCGACG TTCCGGCCCG TTTCGGTCCT GACCTGTGTC TTGCGGCCGT AGTCCGGTGG AGGCCTAGAG CGCCATCCGG GTAAGGGCCC ACGCCAGCTT GTCAAGCTGC AAGGCCGGGC AAAGCCAGGA CTGGACACAG AACGCCGGCA TCAGGCCACC GGCGGGGAAA CGGTCACCGA GCGCTTTTGC GAGGCCTTTG AGCGAGTACG GATCCGAGGG ACCCCAGACC GTCGTCCAGT GCGGGTGGAT CGGGTTCTGG CCGCCCCTTT GCCAGTGGCT CGCGAAAACG CTCCGGAAAC TCGCTCATGC CTAGGCTCCC TGGGGTCTGG CAGCAGGTCA CGCCCACCTA GCCCAAGACC GTGAGCTGCT GCGCGTAGCC CTGATCGGCG CCGACCACCG AGGCGATCAG CCCCTGGTTC ACCCGGTCGT AGAGCCGCAG CGGGCCCTGT CGGGCTGCCT CACTCGACGA CGCGCATCGG GACTAGCCGC GGCTGGTGGC TCCGCTAGTC GGGGACCAAG TGGGCCAGCA TCTCGGCGTC GCCCGGGACA GCCCGACGGA GGAGGGTGTA GACCGGGCTT TCGAGCAGCC ACCACAGGTG CGCGTGCTCG GTCGCGGGAT TGATCGTCAT CACGGTCGGA TCGGGCAGAT CCGCGTTACG CCTCCCACAT CTGGCCCGAA AGCTCGTCGG TGGTGTCCAC GCGCACGAGC CAGCGCCCTA ACTAGCAGTA GTGCCAGCCT AGCCCGTCTA GGCGCAATGC TGCGGCCCAC TGCGCCTGGT CGTCGTCCAC GTCGAGCACC AAGCCCAACC TGATCGACGG GGTGCGGGCC GCAATGTAGC GGCGGGTGAG CGCCTCCGCG ACGCCGGGTG ACGCGGACCA GCAGCAGGTG CAGCTCGTGG TTCGGGTTGG ACTAGCTGCC CCACGCCCGG CGTTACATCG CCGCCCACTC GCGGAGGCGC CGCGGCTGCG GCCACTGCCC GTCCCGGACG TAGTCATCCG TCGCGTGCGG GTATTTGAAC CGCCAGCGGT CCAACCAGGC GTCAACAGCA GCGGTCATGA GCGCCGACGC CGGTGACGGG CAGGGCCTGC ATCAGTAGGC AGCGCACGCC CATAAACTTG GCGGTCGCCA GGTTGGTCCG CAGTTGTCGT CGCCAGTACT CCGCCAAGCT AGGGCCGGAT CTGTACCGAT CGGGGGAGGC GCGCCGCAAA TTATTTAAGA GTCTCGCTAG CAAACCATGT CAGGTGTTGC GGTGGGTTCC GGCGGTTCGA TCCCGGCCTA GACATGGCTA GCCCCCTCCG CGCGGCGTTT AATAAATTCT CAGAGCGATC GTTTGGTACA GTCCACAACG CCACCCAAGG GGGTAAACCT CCACCCGAAT TATTTAAGAG TCTCGCTAGC TAAGCCCTAT CTGATGCTGC GCGGGGGGTC CTTCGCACTG AATCTCAAAG GTGGCCGGCT CCCATTTGGA GGTGGGCTTA ATAAATTCTC AGAGCGATCG ATTCGGGATA GACTACGACG CGCCCCCCAG GAAGCGTGAC TTAGAGTTTC CACCGGCCGA GAATTTCGTC GCGCGAAAAC CTCCCTGGAC AGTTCTGGAA TTCAGCAAGA GGTGTGTCTG AACTTCGGTG TTTTTTTGGG GGGTGACTCC AGCGGGGTGG CTTAAAGCAG CGCGCTTTTG GAGGGACCTG TCAAGACCTT AAGTCGTTCT CCACACAGAC TTGAAGCCAC AAAAAAACCC CCCACTGAGG TCGCCCCACC GCACAACGCG AACAGAGACC TTGTGTGTAC GACGGCGGGA GGTAAGTCGG GTACGGCTCG GACTGCGGTA GAGCAACCGT CGAATCGATT TCGAGCAGAG CGTGTTGCGC TTGTCTCTGG AACACACATG CTGCCGCCCT CCATTCAGCC CATGCCGAGC CTGACGCCAT CTCGTTGGCA GCTTAGCTAA AGCTCGTCTC CGAGCAGAGC AAGATATTCC AAAACTCCGG GGTTCCTCGG CGGCCTCCCC CGTCTGTTTG CTCAACCGAG GGAGACCTGG CGGTCCCGCG TTTCCGGACG GCTCGTCTCG TTCTATAAGG TTTTGAGGCC CCAAGGAGCC GCCGGAGGGG GCAGACAAAC GAGTTGGCTC CCTCTGGACC GCCAGGGCGC AAAGGCCTGC CGCGGGACCG CCTACCGCTC GAGAGCGGAA GAGCATCTAG ATGCATTCGC GAGGTACCCA GCTTTTGTTC CCTTTAGTGA GGGTTAATTG CGCGCTTGGC GCGCCCTGGC GGATGGCGAG CTCTCGCCTT CTCGTAGATC TACGTAAGCG CTCCATGGGT CGAAAACAAG GGAAATCACT CCCAATTAAC GCGCGAACCG GTAATCATGG TCATAGCTGT TTCCTGTGTG AAATTGTTAT CCGCTCACAA TTCCACACAA CATACGAGCC GGAAGCATAA AGTGTAAAGC CTGGGGTGCC CATTAGTACC AGTATCGACA AAGGACACAC TTTAACAATA GGCGAGTGTT AAGGTGTGTT GTATGCTCGG CCTTCGTATT TCACATTTCG GACCCCACGG TAATGAGTGA GCTAACTCAC ATTAATTGCG TTGCGCTCAC TGCCCGCTTT CCAGTCGGGA AACCTGTCGT GCCAGCTGCA TTAATGAATC GGCCAACGCG ATTACTCACT CGATTGAGTG TAATTAACGC AACGCGAGTG ACGGGCGAAA GGTCAGCCCT TTGGACAGCA CGGTCGACGT AATTACTTAG CCGGTTGCGC CGGGGAGAGG CGGTTTGCGT ATTGGGCGCA TGCATAAAAA CTGTTGTAAT TCATTAAGCA TTCTGCCGAC ATGGAAGCCA TCACAAACGG CATGATGAAC GCCCCTCTCC GCCAAACGCA TAACCCGCGT ACGTATTTTT GACAACATTA AGTAATTCGT AAGACGGCTG TACCTTCGGT AGTGTTTGCC GTACTACTTG CTGAATCGCC AGCGGCATCA GCACCTTGTC GCCTTGCGTA TAATATTTGC CCATGGGGGT GGGCGAAGAA CTCCAGCATG AGATCCCCGC GCTGGAGGAT GACTTAGCGG TCGCCGTAGT CGTGGAACAG CGGAACGCAT ATTATAAACG GGTACCCCCA CCCGCTTCTT GAGGTCGTAC
TCTAGGGGCG CGACCTCCTA CATCCAGCCG GCGTCCCGGA AAACGATTCC GAAGCCCAAC CTTTCATAGA AGGCGGCGGT GGAATCGAAA TCTCGTGATG GCAGGTTGGG CGTCGCTTGG GTAGGTCGGC CGCAGGGCCT TTTGCTAAGG CTTCGGGTTG GAAAGTATCT TCCGCCGCCA CCTTAGCTTT AGAGCACTAC CGTCCAACCC GCAGCGAACC TCGGTCATTT CGAACCCCAG AGTCCCGCTC AGAAGAACTC GTCAAGAAGG CGATAGAAGG CGATGCGCTG CGAATCGGGA GCGGCGATAC CGTAAAGCAC AGCCAGTAAA GCTTGGGGTC TCAGGGCGAG TCTTCTTGAG CAGTTCTTCC GCTATCTTCC GCTACGCGAC GCTTAGCCCT CGCCGCTATG GCATTTCGTG GAGGAAGCGG TCAGCCCATT CGCCGCCAAG CTCTTCAGCA ATATCACGGG TAGCCAACGC TATGTCCTGA TAGCGGTCCG CCACACCCAG CCGGCCACAG CTCCTTCGCC AGTCGGGTAA GCGGCGGTTC GAGAAGTCGT TATAGTGCCC ATCGGTTGCG ATACAGGACT ATCGCCAGGC GGTGTGGGTC GGCCGGTGTC TCGATGAATC CAGAAAAGCG GCCATTTTCC ACCATGATAT TCGGCAAGCA GGCATCGCCA TGGGTCACGA CGAGATCCTC GCCGTCGGGC ATGCGCGCCT AGCTACTTAG GTCTTTTCGC CGGTAAAAGG TGGTACTATA AGCCGTTCGT CCGTAGCGGT ACCCAGTGCT GCTCTAGGAG CGGCAGCCCG TACGCGCGGA TGAGCCTGGC GAACAGTTCG GCTGGCGCGA GCCCCTGATG CTCTTCGTCC AGATCATCCT GATCGACAAG ACCGGCTTCC ATCCGAGTAC GTGCTCGCTC ACTCGGACCG CTTGTCAAGC CGACCGCGCT CGGGGACTAC GAGAAGCAGG TCTAGTAGGA CTAGCTGTTC TGGCCGAAGG TAGGCTCATG CACGAGCGAG GATGCGATGT TTCGCTTGGT GGTCGAATGG GCAGGTAGCC GGATCAAGCG TATGCAGCCG CCGCATTGCA TCAGCCATGA TGGATACTTT CTCGGCAGGA CTACGCTACA AAGCGAACCA CCAGCTTACC CGTCCATCGG CCTAGTTCGC ATACGTCGGC GGCGTAACGT AGTCGGTACT ACCTATGAAA GAGCCGTCCT GCAAGGTGAG ATGACAGGAG ATCCTGCCCC GGCACTTCGC CCAATAGCAG CCAGTCCCTT CCCGCTTCAG TGACAACGTC GAGCACAGCT GCGCAAGGAA CGTTCCACTC TACTGTCCTC TAGGACGGGG CCGTGAAGCG GGTTATCGTC GGTCAGGGAA GGGCGAAGTC ACTGTTGCAG CTCGTGTCGA CGCGTTCCTT CGCCCGTCGT GGCCAGCCAC GATAGCCGCG CTGCCTCGTC CTGCAGTTCA TTCAGGGCAC CGGACAGGTC GGTCTTGACA AAAAGAACCG GGCGCCCCTG GCGGGCAGCA CCGGTCGGTG CTATCGGCGC GACGGAGCAG GACGTCAAGT AAGTCCCGTG GCCTGTCCAG CCAGAACTGT TTTTCTTGGC CCGCGGGGAC CGCTGACAGC CGGAACACGG CGGCATCAGA GCAGCCGATT GTCTGTTGTG CCCAGTCATA GCCGAATAGC CTCTCCACCC AAGCGGCCGG AGAACCTGCG GCGACTGTCG GCCTTGTGCC GCCGTAGTCT CGTCGGCTAA CAGACAACAC GGGTCAGTAT CGGCTTATCG GAGAGGTGGG TTCGCCGGCC TCTTGGACGC TGCAATCCAT CTTGTTCAAT CATGCGAAAC GATCCTCATC CTGTCTCTTG ATCAGATCTT GATCCCCTGC GCCATCAGAT CCTTGGCGGC AAGAAAGCCA ACGTTAGGTA GAACAAGTTA GTACGCTTTG CTAGGAGTAG GACAGAGAAC TAGTCTAGAA CTAGGGGACG CGGTAGTCTA GGAACCGCCG TTCTTTCGGT TCCAGTTTAC TTTGCAGGGC TTCCCAACCT TACCAGAGGG CGCCCCAGCT GGCAATTCCG GTTCGCTTGC TGTCCATAAA ACCGCCCAGT CTAGCTATCG AGGTCAAATG AAACGTCCCG AAGGGTTGGA ATGGTCTCCC GCGGGGTCGA CCGTTAAGGC CAAGCGAACG ACAGGTATTT TGGCGGGTCA GATCGATAGC CCATGTAAGC CCACTGCAAG CTACCTGCTT TCTCTTTGCG CTTGCGTTTT CCCTTGTCCA GATAGCCCAG TAGCTGACAT TCATCCCAGG TGGCACTTTT GGTACATTCG GGTGACGTTC GATGGACGAA AGAGAAACGC GAACGCAAAA GGGAACAGGT CTATCGGGTC ATCGACTGTA AGTAGGGTCC ACCGTGAAAA CGGGGAAATG TGCGCGCCCG CGTTCCTGCT GGCGCTGGGC CTGTTTCTGG CGCTGGACTT CCCGCTGTTC CGTCAGCAGC TTTTCGCCCA CGGCCTTGAT GCCCCTTTAC ACGCGCGGGC GCAAGGACGA CCGCGACCCG GACAAAGACC GCGACCTGAA GGGCGACAAG GCAGTCGTCG AAAAGCGGGT GCCGGAACTA GATCGCGGCG GCCTTGGCCT GCATATCCCG ATTCAACGGC CCCAGGGCGT CCAGAACGGG CTTCAGGCGC TCCCGAAGGT CTCGGGCCGT CTCTTGGGCT CTAGCGCCGC CGGAACCGGA CGTATAGGGC TAAGTTGCCG GGGTCCCGCA GGTCTTGCCC GAAGTCCGCG AGGGCTTCCA GAGCCCGGCA GAGAACCCGA TGATCGGCCT TCTTGCGCAT CTCACGCGCT CCTGCGGCGG CCTGTAGGGC AGGCTCATAC CCCTGCCGAA CCGCTTTTGT CAGCCGGTCG GCCACGGCTT ACTAGCCGGA AGAACGCGTA GAGTGCGCGA GGACGCCGCC GGACATCCCG TCCGAGTATG GGGACGGCTT GGCGAAAACA GTCGGCCAGC CGGTGCCGAA CCGGCGTCTC AACGCGCTTT GAGATTCCCA GCTTTTCGGC CAATCCCTGC GGTGCATAGG CGCGTGGCTC GACCGCTTGC GGGCTGATGG TGACGTGGCC GGCCGCAGAG TTGCGCGAAA CTCTAAGGGT CGAAAAGCCG GTTAGGGACG CCACGTATCC GCGCACCGAG CTGGCGAACG CCCGACTACC ACTGCACCGG CACTGGTGGC CGCTCCAGGG CCTCGTAGAA CGCCTGAATG CGCGTGTGAC GTGCCTTGCT GCCCTCGATG CCCCGTTGCA GCCCTAGATC GGCCACAGCG GTGACCACCG GCGAGGTCCC GGAGCATCTT GCGGACTTAC GCGCACACTG CACGGAACGA CGGGAGCTAC GGGGCAACGT CGGGATCTAG CCGGTGTCGC GCCGCAAACG TGGTCTGGTC GCGGGTCATC TGCGCTTTGT TGCCGATGAA CTCCTTGGCC GACAGCCTGC CGTCCTGCGT CAGCGGCACC ACGAACGCGG CGGCGTTTGC ACCAGACCAG CGCCCAGTAG ACGCGAAACA ACGGCTACTT GAGGAACCGG CTGTCGGACG GCAGGACGCA GTCGCCGTGG TGCTTGCGCC TCATGTGCGG GCTGGTTTCG TCACGGTGGA TGCTGGCCGT CACGATGCGA TCCGCCCCGT ACTTGTCCGC CAGCCACTTG TGCGCCTTCT CGAAGAACGC AGTACACGCC CGACCAAAGC AGTGCCACCT ACGACCGGCA GTGCTACGCT AGGCGGGGCA TGAACAGGCG GTCGGTGAAC ACGCGGAAGA GCTTCTTGCG CGCCTGCTGT TCTTGGCTGG CCGACTTCCA CCATTCCGGG CTGGCCGTCA TGACGTACTC GACCGCCAAC ACAGCGTCCT TGCGCCGCTT CTCTGGCAGC GCGGACGACA AGAACCGACC GGCTGAAGGT GGTAAGGCCC GACCGGCAGT ACTGCATGAG CTGGCGGTTG TGTCGCAGGA ACGCGGCGAA GAGACCGTCG AACTCGCGCA GTCGGCCCAT CGCTTCATCG GTGCTGCTGG CCGCCCAGTG CTCGTTCTCT GGCGTCCTGC TGGCGTCAGC GTTGGGCGTC TCGCGCTCGC TTGAGCGCGT CAGCCGGGTA GCGAAGTAGC CACGACGACC GGCGGGTCAC GAGCAAGAGA CCGCAGGACG ACCGCAGTCG CAACCCGCAG AGCGCGAGCG GGTAGGCGTG CTTGAGACTG GCCGCCACGT TGCCCATTTT CGCCAGCTTC TTGCATCGCA TGATCGCGTA TGCCGCCATG CCTGCCCCTC CCTTTTGGTG CCATCCGCAC GAACTCTGAC CGGCGGTGCA ACGGGTAAAA GCGGTCGAAG AACGTAGCGT ACTAGCGCAT ACGGCGGTAC GGACGGGGAG GGAAAACCAC TCCAACCGGC TCGACGGGGG CAGCGCAAGG CGGTGCCTCC GGCGGGCCAC TCAATGCTTG AGTATACTCA CTAGACTTTG CTTCGCAAAG TCGTGACCGC AGGTTGGCCG AGCTGCCCCC GTCGCGTTCC GCCACGGAGG CCGCCCGGTG AGTTACGAAC TCATATGAGT GATCTGAAAC GAAGCGTTTC AGCACTGGCG CTACGGCGGC TGCGGCGCCC TACGGGCTTG CTCTCCGGGC TTCGCCCTGC GCGGTCGCTG CGCTCCCTTG CCAGCCCGTG GATATGTGGA CGATGGCCGC GATGCCGCCG ACGCCGCGGG ATGCCCGAAC GAGAGGCCCG AAGCGGGACG CGCCAGCGAC GCGAGGGAAC GGTCGGGCAC CTATACACCT GCTACCGGCG GAGCGGCCAC CGGCTGGCTC GCTTCGCTCG GCCCGTGGAC AACCCTGCTG GACAAGCTGA TGGACAGGCT GCGCCTGCCC ACGAGCTTGA CCACAGGGAT CTCGCCGGTG GCCGACCGAG CGAAGCGAGC CGGGCACCTG TTGGGACGAC CTGTTCGACT ACCTGTCCGA CGCGGACGGG TGCTCGAACT GGTGTCCCTA TGCCCACCGG CTACCCAGCC TTCGACCACA TACCCACCGG CTCCAACTGC GCGGCCTGCG GCCTTGCCCC ATCAATTTTT TTAATTTTCT CTGGGGAAAA ACGGGTGGCC GATGGGTCGG AAGCTGGTGT ATGGGTGGCC GAGGTTGACG CGCCGGACGC CGGAACGGGG TAGTTAAAAA AATTAAAAGA GACCCCTTTT GCCTCCGGCC TGCGGCCTGC GCGCTTCGCT TGCCGGTTGG ACACCAAGTG GAAGGCGGGT CAAGGCTCGC GCAGCGACCG CGCAGCGGCT TGGCCTTGAC CGGAGGCCGG ACGCCGGACG CGCGAAGCGA ACGGCCAACC TGTGGTTCAC CTTCCGCCCA GTTCCGAGCG CGTCGCTGGC GCGTCGCCGA ACCGGAACTG GCGCCTGGAA CGACCCAAGC CTATGCGAGT GGGGGCAGTC GAAGGCGAAG CCCGCCCGCC TGCCCCCCGA GCCTCACGGC GGCGAGTGCG GGGGTTCCAA CGCGGACCTT GCTGGGTTCG GATACGCTCA CCCCCGTCAG CTTCCGCTTC GGGCGGGCGG ACGGGGGGCT CGGAGTGCCG CCGCTCACGC CCCCAAGGTT GGGGGCAGCG CCACCTTGGG CAAGGCCGAA GGCCGCGCAG TCGATCAACA AGCCCCGGAG GGGCCACTTT TTGCCGGAG CCCCCGTCGC GGTGGAACCC GTTCCGGCTT CCGGCGCGTC AGCTAGTTGT TCGGGGCCTC CCCGGTGAAA AACGGCCTC SEQ ID: 05 MPELAVRTEF DYSSEIYKDA YSRINAIVIE GEQEAYSNYL QMAELLPEDK EELTRLAKME NRHKKGFQAC GNNLQVNPDM PYAQEFFAGL HGNFQHAFSE GKVVTCLLIQ ALIIEAFAIA AYNIYIPVAD DFARKITEGV VKDEYTHLNY GEEWLKANFA TAKEELEQAN KENLPLVWKM LNQVQGDAKV LGMEKEALVE DFMISYGEAL SNIGFSTREI MRMSSYGLAG V SEQ ID: 06 MFGLIGHLTS LEHAQAVAED LGYPEYANQG LDFWCSAPPQ
VVDNFQVKSV TGQVIEGKYV ESCFLPEMLT QRRIKAAIRK ILNAMALAQK VGLDITALGG FSSIVFEEFN LKQNNQVRNV ELDFQRFTTG NTHTAYVICR QVESGAKQLG IDLSQATVAV CGATGDIGSA VCRWLDSKHQ VKELLLIARN RQRLENLQEE LGRGKIMDLE TALPQADIIV WVASMPKGVE IAGEMLKKPC LIVDGGYPKN LDTRVKADGV HILKGGIVEH SLDITWEIMK IVEMDIPSRQ MFACFAEAIL LEFEGWRTNF SWGRNQISVN KMEAIGEASV KHGFCPLVAL SEQ ID: 07 CAGTCAATGG AGAGCATTGC CATAAGTAAA GGCATCCCCT GCGTGATAAG ATTACCTTCA GAAAACAGAT AGTTGCTGGG TTATCGCAGA TTTTTCTCGC GTCAGTTACC TCTCGTAACG GTATTCATTT CCGTAGGGGA CGCACTATTC TAATGGAAGT CTTTTGTCTA TCAACGACCC AATAGCGTCT AAAAAGAGCG AACCAAATAA CTGTAAATAA TAACTGTCTC TGGGGCGACG GTAGGCTTTA TATTGCCAAA TTTCGCCCGT GGGAGAAAGC TAGGCTATTC AATGTTTATG TTGGTTTATT GACATTTATT ATTGACAGAG ACCCCGCTGC CATCCGAAAT ATAACGGTTT AAAGCGGGCA CCCTCTTTCG ATCCGATAAG TTACAAATAC GAGGACTCCT SEQ ID: 08 CCTGGCTCAG GACGAACGCT GGCGGCGTGC TTAACACATG CAAGTCGAGC GGTAAGGCCC TTCGGGGTAC ACGAGCGGCG AACGGGTGAG TAACACGTGG GGACCGAGTC CTGCTTGCGA CCGCCGCACG AATTGTGTAC GTTCAGCTCG CCATTCCGGG AAGCCCCATG TGCTCGCCGC TTGCCCACTC ATTGTGCACC GTGATCTGCC CTGCACTTCG GGATAAGCCT GGGAAACTGG GTCTAATACC GGATATGACC TTCGGCTGCA TGGCTGAGGG TGGAAAGGTT TACTGGTGCA CACTAGACGG GACGTGAAGC CCTATTCGGA CCCTTTGACC CAGATTATGG CCTATACTGG AAGCCGACGT ACCGACTCCC ACCTTTCCAA ATGACCACGT GGATGGGCCC GCGGCCTATC AGCTTGTTGG TGGGGTAATG GCCTACCAAG GCGACGACGG GTAGCCGACC TGAGAGGGTG ACCGGCCACA CTGGGACTGA CCTACCCGGG CGCCGGATAG TCGAACAACC ACCCCATTAC CGGATGGTTC CGCTGCTGCC CATCGGCTGG ACTCTCCCAC TGGCCGGTGT GACCCTGACT GACACGGCCC AGACTCCTAC GGGAGGCAGC AGTGGGGAAT ATTGCACAAT GGGCGAAAGC CTGATGCAGC GACGCCGCGT GAGGGATGAC GGCCTTCGGG CTGTGCCGGG TCTGAGGATG CCCTCCGTCG TCACCCCTTA TAACGTGTTA CCCGCTTTCG GACTACGTCG CTGCGGCGCA CTCCCTACTG CCGGAAGCCC TTGTAAACCT CTTTCAGCAG GGACGAAGCG AAAGTGACGG TACCTGCAGA AGAAGCACCG GCCAACTACG TGCCAGCAGC CGCGGTAATA CGTAGGGTGC AACATTTGGA GAAAGTCGTC CCTGCTTCGC TTTCACTGCC ATGGACGTCT TCTTCGTGGC CGGTTGATGC ACGGTCGTCG GCGCCATTAT GCATCCCACG AAGCGTTGTC CGGAATTACT GGGCGTAAAG AGCTCGTAGG CGGTTTGTCG CGTCGTCTGT GAAAACTCAN AGCTCAACCT CGAGCTTGCA GGCGATACGG TTCGCAACAG GCCTTAATGA CCCGCATTTC TCGAGCATCC GCCAAACAGC GCAGCAGACA CTTTTGAGTN TCGAGTTGGA GCTCGAACGT CCGCTATGCC GCAGACTTGA GTACTGCAGG GGAGACTGGA ATTCCTGGTG TAGCGGTGAA ATGCGCAGAT ATCAGGAGGA ACACCGGTGG CGAAGGCGGG TCTCTGGGCA CGTCTGAACT CATGACGTCC CCTCTGACCT TAAGGACCAC ATCGCCACTT TACGCGTCTA TAGTCCTCCT TGTGGCCACC GCTTCCGCCC AGAGACCCGT GTAACTGACG CTGAGGAGCG AAAGCGTGGG TAGCAAACAG GATTAGATAC CCTGGTAGTC CACGCCGTAA ACGGTGGGCG CTAGGTGTGG GTTTCCTTCC CATTGACTGC GACTCCTCGC TTTCGCACCC ATCGTTTGTC CTAATCTATG GGACCATCAG GTGCGGCATT TGCCACCCGC GATCCACACC CAAAGGAAGG ACGGGATCCG TGCCGTAGTT AACGCATTAA GCGCCCCGCC TGGGGAGTAC GGCCGCAAGG TTAAAACTCA AAGGAATTGA CGGGGGCCCG CACAAGCGGC TGCCCTAGGC ACGGCATCAA TTGCGTAATT CGCGGGGCGG ACCCCTCATG CCGGCGTTCC AATTTTGAGT TTCCTTAACT GCCCCCGGGC GTGTTCGCCG GGAGCATGTG GATTAATTCG ATGCAACGCG AAGAACCTTA CCTGGGTTTG ACATATACCG GAAAGCCGTA GAGATACCGC CCCCCTTGTG GTCGGTATAC CCTCGTACAC CTAATTAAGC TACGTTGCGC TTCTTGGAAT GGACCCAAAC TGTATATGGC CTTTCGGCAT CTCTATGGCG GGGGGAACAC CAGCCATATG AGGTGGTGCA TGGCTGTCGT CAGCTCGTGT CGTGAGATGT TGGGTTAAGT CCCGCAACGA GCGCAACCCT TGTCTTATGT TGCCAGCACG TAATGGTGGG TCCACCACGT ACCGACAGCA GTCGAGCACA GCACTCTACA ACCCAATTCA GGGCGTTGCT CGCGTTGGGA ACAGAATACA ACGGTCGTGC ATTACCACCC GACTCGTAAG AGACTGCCGG GGTCAACTCG GAGGAAGGTG GGGACGACGT CAAGTCATCA TGCCCCTTAT GTCCAGGGCT TCACACATGC TACAATGGCC CTGAGCATTC TCTGACGGCC CCAGTTGAGC CTCCTTCCAC CCCTGCTGCA GTTCAGTAGT ACGGGGAATA CAGGTCCCGA AGTGTGTACG ATGTTACCGG GGTACAGAGG GCTGCGATAC CGTGAGGTGG AGCGAATCCC TTAAAGCCGG TCTCAGTTCG GATCGGGGTC TGCAACTCGA CCCCGTGAAG TCGGAGTCGC CCATGTCTCC CGACGCTATG GCACTCCACC TCGCTTAGGG AATTTCGGCC AGAGTCAAGC CTAGCCCCAG ACGTTGAGCT GGGGCACTTC AGCCTCAGCG TAGTAATCGC AGATCAGCAA CGCTGCGGTG AATACGTTCC CGGGCCTTGT ACACACCGCC CGTCACGTCA TGAAAGTCGG TAACACCCGA AGCCGGTGGC ATCATTAGCG TCTAGTCGTT GCGACGCCAC TTATGCAAGG GCCCGGAACA TGTGTGGCGG GCAGTGCAGT ACTTTCAGCC ATTGTGGGCT TCGGCCACCG CTAACCCCTT GTGGGAGGGA GCCGTCGAAG GTGGGATCGG CGATTGGGAC GAAGTCGTAA CAAGGTAGCC GTACCGGAAG GGATTGGGGA ACACCCTCCC TCGGCAGCTT CCACCCTAGC CGCTAACCCT GCTTCAGCAT TGTTCCATCG GCATGGCCTT CC SEQ ID: 09 TCAACGGAGA GTTTGATCCT GGCTCAGGAC GAACGCTGGC GGCGTGCTTA ACACATGCAA GTCGAGCGGT AAGGCCCTTC GGGGTACACG AGCGGCGAAC AGTTGCCTCT CAAACTAGGA CCGAGTCCTG CTTGCGACCG CCGCACGAAT TGTGTACGTT CAGCTCGCCA TTCCGGGAAG CCCCATGTGC TCGCCGCTTG GGGTGAGTAA CACGTGGGTG ATCTGCCCTG CACTTCGGGA TAAGCCTGGG AAACTGGGTC TAATACCGGA TATGACCTTC GGCTGCATGG CCGTTGGTGG CCCACTCATT GTGCACCCAC TAGACGGGAC GTGAAGCCCT ATTCGGACCC TTTGACCCAG ATTATGGCCT ATACTGGAAG CCGACGTACC GGCAACCACC AAAGGTTTAC TGGTGCAGGA TGGGCCCGCG GCCTATCAGC TTGTTGGTGG GGTAATGGCC TACCAAGGCG ACGACGGGTA GCCGACCTGA GAGGGTGACC TTTCCAAATG ACCACGTCCT ACCCGGGCGC CGGATAGTCG AACAACCACC CCATTACCGG ATGGTTCCGC TGCTGCCCAT CGGCTGGACT CTCCCACTGG GGCCACACTG GGACTGAGAC ACGGCCCAGA CTCCTACGGG AGGCAGCAGT GGGGAATATT GCACAATGGG CGAAAGCCTG ATGCAGCGAC GCCGCGTGAG CCGGTGTGAC CCTGACTCTG TGCCGGGTCT GAGGATGCCC TCCGTCGTCA CCCCTTATAA CGTGTTACCC GCTTTCGGAC TACGTCGCTG CGGCGCACTC GGATGACGGC CTTCGGGTTG TAAACCTCTT TCAGCAGGGA CGAAGCGAAA GTGACGGTAC CTGCAGAAGA AGCACCGGCC AACTACGTGC CAGCAGCCGC CCTACTGCCG GAAGCCCAAC ATTTGGAGAA AGTCGTCCCT GCTTCGCTTT CACTGCCATG GACGTCTTCT TCGTGGCCGG TTGATGCACG GTCGTCGGCG GGTAATACGT AGGGTGCAAG CGTTGTCCGG AATTACTGGG CGTAAAGAGC TCGTAGGCGG TTTGTCGCGT CGTCTGTGAA AACTCGAGGC TCAACCTCGA CCATTATGCA TCCCACGTTC GCAACAGGCC TTAATGACCC GCATTTCTCG AGCATCCGCC AAACAGCGCA GCAGACACTT TTGAGCTCCG AGTTGGAGCT GCTTGCAGGC GATACGGGCA GACTTGAGTA CTGCAGGGGA GACTGGAATT CCTGGTGTAG CGGTGAAATG CGCAGATATC AGGAGGAACA CCGGTGGCGA CGAACGTCCG CTATGCCCGT CTGAACTCAT GACGTCCCCT CTGACCTTAA GGACCACATC GCCACTTTAC GCGTCTATAG TCCTCCTTGT GGCCACCGCT AGGCGGGTCT CTGGGCAGTA ACTGACGCTG AGGAGCGAAA GCGTGGGTAG CGAACAGGAT TAGATACCCT GGTAGTCCAC GCCGTAAACG GTGGGCGCTA TCCGCCCAGA GACCCGTCAT TGACTGCGAC TCCTCGCTTT CGCACCCATC GCTTGTCCTA ATCTATGGGA CCATCAGGTG CGGCATTTGC CACCCGCGAT GGTGTGGGTT TCCTTCCACG GGATCCGTGC CGTAGCTAAC GCATTAAGCG CCCCGCCTGG GGAGTACGGC CGCAAGGCTA AAACTCAAAG GAATTGACGG CCACACCCAA AGGAAGGTGC CCTAGGCACG GCATCGATTG CGTAATTCGC GGGGCGGACC CCTCATGCCG GCGTTCCGAT TTTGAGTTTC CTTAACTGCC GGGCCCGCAC AAGCGGCGGA GCATGTGGAT TAATTCGATG CAACGCGAAG AACCTTACCT GGGTTTGACA TATACCGGAA AGCTGCAGAG ATGTGGCCCC CCCGGGCGTG TTCGCCGCCT CGTACACCTA ATTAAGCTAC GTTGCGCTTC TTGGAATGGA CCCAAACTGT ATATGGCCTT TCGACGTCTC TACACCGGGG CCTTGTGGTC GGTATACAGG TGGTGCATGG CTGTCGTCAG CTCGTGTCGT GAGATGTTGG GTTAAGTCCC GCAACGAGCG CAACCCTTGT CTTATGTTGC GGAACACCAG CCATATGTCC ACCACGTACC GACAGCAGTC GAGCACAGCA CTCTACAACC CAATTCAGGG CGTTGCTCGC GTTGGGAACA GAATACAACG CAGCACGTAA TGGTGGGGAC TCGTAAGAGA CTGCCGGGGT CAACTCGGAG GAAGGTGGGG ACGACGTCAA GTCATCATGC CCCTTATGTC CAGGGCTTCA GTCGTGCATT ACCACCCCTG AGCATTCTCT GACGGCCCCA GTTGAGCCTC CTTCCACCCC TGCTGCAGTT CAGTAGTACG GGGAATACAG GTCCCGAAGT CACATGCTAC AATGGCCGGT ACAGAGGGCT GCGATACCGT
GAGGTGGAGC GAATCCCTTA AAGCCGGTCT CAGTTCGGAT CGGGGTCTGC AACTCGACCC GTGTACGATG TTACCGGCCA TGTCTCCCGA CGCTATGGCA CTCCACCTCG CTTAGGGAAT TTCGGCCAGA GTCAAGCCTA GCCCCAGACG TTGAGCTGGG CGTGAAGTCG GAGTCGCTAG TAATCGCAGA TCAGCAACGC TGCGGTGAAT ACGTTCCCGG GCCTTGTACA CACCGCCCGT CACGTCATGA AAGTCGGTAA GCACTTCAGC CTCAGCGATC ATTAGCGTCT AGTCGTTGCG ACGCCACTTA TGCAAGGGCC CGGAACATGT GTGGCGGGCA GTGCAGTACT TTCAGCCATT CACCCGAAGC CGGTGGCCTA ACCCCTCGTG GGAGGGAGCC GTCGAAGGTG GGATCGGCGA TTGGGACGAA GTCGTAACAA GGTAGCCGTA CCGGAAGGTG GTGGGCTTCG GCCACCGGAT TGGGGAGCAC CCTCCCTCGG CAGCTTCCAC CCTAGCCGCT AACCCTGCTT CAGCATTGTT CCATCGGCAT GGCCTTCCAC CGGCTGGATC ACCTCCTTTC TGCCGACCTA GTGGAGGAAA GA SEQ ID: 10 ACGTGGCGGC ATGCCTTACA CATGCAAGTC GAACGGCAGC GCGGACTTCG GTCTGGCGGC GAGTGGCGAA CGGGTGAGTA ATACATCGGA ACGTACCCTG TGCACCGCCG TACGGAATGT GTACGTTCAG CTTGCCGTCG CGCCTGAAGC CAGACCGCCG CTCACCGCTT GCCCACTCAT TATGTAGCCT TGCATGGGAC TTGTGGGGGA TAACTAGTCG AAAGATTAGC TAATACCGCA TACGACCTGA GGGTGAAAGT GGGGGACCGC AAGGCCTCAC GCAGCAGGAG CGGCCGATGT AACACCCCCT ATTGATCAGC TTTCTAATCG ATTATGGCGT ATGCTGGACT CCCACTTTCA CCCCCTGGCG TTCCGGAGTG CGTCGTCCTC GCCGGCTACA CTGATTAGCT AGTTGGTGGG GTAAAGGCCC ACCAAGGCGA CGATCAGTAG CTGGTCTGAG AGGACGATCA GCCACACTGG GACTGAGACA CGGCCCAGAC GACTAATCGA TCAACCACCC CATTTCCGGG TGGTTCCGCT GCTAGTCATC GACCAGACTC TCCTGCTAGT CGGTGTGACC CTGACTCTGT GCCGGGTCTG TCCTACGGGA GGCAGCAGTG GGGAATTTTG GACAATGGGG GCAACCCTGA TCCAGCAATG CCGCGTGTGT GAAGAAGGCC TTCGGGTTGT AAAGCACTTT AGGATGCCCT CCGTCGTCAC CCCTTAAAAC CTGTTACCCC CGTTGGGACT AGGTCGTTAC GGCGCACACA CTTCTTCCGG AAGCCCAACA TTTCGTGAAA TGTCCGGAAA GAAATCGCGC TGGTTAATAC CTGCGTGATG ACGGTACCGG AAGAATAAGC ACCGGCTAAC TACGTGCCAG CAGCCGCGGT AATACGTAGG ACAGGCCTTT CTTTAGCGCG ACCAATTATG GACGCACTAC TGCCATGGCC TTCTTATTCG TGGCCGATTG ATGCACGGTC GTCGGCGCCA TTATGCATCC GTGCGAGCGT TAATCGGAAT TACTGGGCGT AAAGCGTGCG CAGGCGGTTT TGTAAGACAG GCGTGAAATC CCCGGGCTTA ACCTGGGAAT TGCGCTTGTG CACGCTCGCA ATTAGCCTTA ATGACCCGCA TTTCGCACGC GTCCGCCAAA ACATTCTGTC CGCACTTTAG GGGCCCGAAT TGGACCCTTA ACGCGAACAC ACTGCAAGGC TAGAGTGCGT CAGAGGGGGG TAGAATTCCA CGTGTAGCAG TGAAATGCGT AGAGATGTGG AGGAATACCG ATGGCGAAGG CGAGCCCCCT TGACGTTCCG ATCTCACGCA GTCTCCCCCC ATCTTAAGGT GCACATCGTC ACTTTACGCA TCTCTACACC TCCTTATGGC TACCGCTTCC GCTCGGGGGA GGACCTTGAC TGACGCTCAT GCACGAAAGC GTGGGGAGCA AACAGGATTA GATACCCTGG TAGTCCACGC CCTAAACGAT GTCAACTAGT TGTTGGGATT CCTGGAACTG ACTGCGAGTA CGTGCTTTCG CACCCCTCGT TTGTCCTAAT CTATGGGACC ATCAGGTGCG GGATTTGCTA CAGTTGATCA ACAACCCTAA CATTTTCTCA GTAACGTAGC TAACGCGTGA AGTTGACCGC CTGGGGAGTA CGGCTGCAAG ATTAAAACTC AAAGGAATTG ACGGGGACCC GCACAAGCGG GTAAAAGAGT CATTGCATCG ATTGCGCACT TCAACTGGCG GACCCCTCAT GCCGACGTTC TAATTTTGAG TTTCCTTAAC TGCCCCTGGG CGTGTTCGCC TGGATGATGT GGATTAATTC GATGCAACGC GAAAAACCTT ACCTACCCTT GACATGCCCT AACGAAGCAG AGATGCATTA GTGCCCGCAA AGGGAAAGTG ACCTACTACA CCTAATTAAG CTACGTTGCG CTTTTTGGAA TGGATGGGAA CTGTACGGGA TTGCTTCGTC TCTACGTAAT CACGGGCGTT TCCCTTTCAC GGACACAGGT GCTGCATGGC TGTCGTCAGC TCGTGTCGTG AGATGTTGGG TTAAGTCCCG CAACGAGCGC AACCCTTGTC TCTAGTTGCC TACGCAAGAG CCTGTGTCCA CGACGTACCG ACAGCAGTCG AGCACAGCAC TCTACAACCC AATTCAGGGC GTTGCTCGCG TTGGGAACAG AGATCAACGG ATGCGTTCTC CACTCTAGAG AGACTGCCGG TGACAAACCG GAGGAAGGTG GGGATGACGT CAAGTCCTCA TGGCCCTTAT GGGTAGGGCT TCACACGTCA TACAATGGTG GTGAGATCTC TCTGACGGCC ACTGTTTGGC CTCCTTCCAC CCCTACTGCA GTTCAGGAGT ACCGGGAATA CCCATCCCGA AGTGTGCAGT ATGTTACCAC CGTACAGAGG GTTGCCAACC CGCGAGGGGG AGCTAATCCC AGAAAACGCA TCGTAGTCCG GATCGTAGTC TGCAACTCGA CTACGTGAAG CTGGAATCGC GCATGTCTCC CAACGGTTGG GCGCTCCCCC TCGATTAGGG TCTTTTGCGT AGCATCAGGC CTAGCATCAG ACGTTGAGCT GATGCACTTC GACCTTAGCG TAGTAATCGC GGATCAGCAT GCCGCGGTGA ATACGTTCCC GGGTCTTGTA CACACCGCCC GTCACACCAT GGGAGTGGGT TTTGCCAGAA GTAGTTAGCC ATCATTAGCG CCTAGTCGTA CGGCGCCACT TATGCAAGGG CCCAGAACAT GTGTGGCGGG CAGTGTGGTA CCCTCACCCA AAACGGTCTT CATCAATCGG TAACCGCAAG GAGGGCGATT ACCACGGCAG GGTTCATGAC TGGGGTGAAG TCGTAACAAG GTATTGGCGT TCCTCCCGCT AATGGTGCCG TCCCAAGTAC TGACCCCACT TCAGCATTGT TCCA SEQ ID 11 MASIEDILEL EALEKDIFRG AVHPSVLKRT FGGQVAGQSL VSAVRTVDER FEVHSLHGYF LRPGNPTEPT VYLVDRIRDG RSFCTRRVTG IQDGKAIFTM SASFHSQDEG IEHQDTMPSV PEPEELVDAQ TVEEMAATDL YREWKEWDVR IVPAGCTGKT PGIAAKQRVW MRYRNKLPDD QVFHICTLAY LSDMTLLGAS KVPHPGVVTQ TASLDHAMWF LRPFRADEWL LYDQTSPSAG FGRALTQGRM FDRKGTMVAA VVQEGLTRIQ RDQDQRDIET GNMA
[0153] In some embodiments, the cell comprises a plasmid that contains one or more exogenous nucleic acid sequences encoding enzymes or proteins that include but are not limited to one or more of the following: an acyl carrier protein, a TE, a FAR, a FadR, a FAD, a fatty aldehyde reductase, a cytochrome P450 enzyme, a NADH or NADPH cytochrome P450 reductase, a desaturase, a hydroxylase, and an antibiotic resistance enabling protein; wherein the plasmid is at least 20, 30, 40, 50, 60, 70, 80, 90, 95, 96, 97, 98, or 99% homologous to SEQ ID NO:1, SEQ TD NO:2, SEQ ID NO:3, or SEQ TD NO:4.
[0154] In one embodiment, the exogenous gene encodes a FadR. In some cases, the reductase encoded by the exogenous gene catalyzes the reduction of an 8 to 18-carbon fatty acyl-CoA to a corresponding aldehyde. In one embodiment, the reductase encoded by the exogenous gene catalyzes the reduction of a 12-carbon fatty acyl-CoA to dodecanal.
[0155] In some embodiments, the invention relates to a bacterial cell or a compositions comprising at least one bacterial cell that comprises at least a first and a second exogenous nucleic acid sequence, wherein the first nucleic acid sequence encodes a FadR or a functional fragment of a FadR and the second exogenous nucleic acid sequence encodes a fatty acyl-CoA ligase or a functional fragment thereof. In some embodiments, the functional fragments of the enzymes encoded by the one or more exogenous nucleic acid sequences are at least 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 91, 92, 93, 94, 95, 96, 97, 98, or 99% homologous to the nucleic acid sequences that encode the full-length amino acid sequence upon which the functional fragment is based. Any enzyme disclosed in this application and part of the invention may be replaced with a functional fragment. Any composition or cell disclosed in the application may be used in any disclosed method of this application.
[0156] In some embodiments, the genetic constructs contain sequences directing transcription and translation of the relevant exogenous (either heterologous or homologous) gene, a selectable marker, and sequences allowing autonomous replication or chromosomal integration. Suitable vectors comprise a region 5' of the gene which harbors transcriptional initiation controls and a region 3' of the DNA fragment which controls transcriptional termination. It is most preferred when both control regions are derived from genes homologous to the transformed host cell, although it is to be understood that such control regions need not be derived from the genes native to the specific species chosen as a production host. In some cells the exogenous gene is coding sequence and is in operable linkage with a promoter, and in some embodiments the promoter is derived from a gene endogenous to a species of the genus Rhodococcus or Ralstonia. Initiation control regions or promoters, which are useful to drive expression of the instant ORFs in the desired host cell are numerous and familiar to those skilled in the art. Virtually any promoter capable of driving these genes is suitable for the present invention including but not limited to CYC1, HIS3, GAL1, GAL10, ADH1, PGK, PHOS, GAPDH, ADC1, TRP1, URA3, LEU2, ENO; and lac, ara, tet, trp, IPL, IPR, T7, tac, and trc as well as the amy, apr, npr promoters and various phage promoters useful for expression in the lipid-producing bacteria of the present invention. In other embodiments the promoter is upregulated in response to reduction or elimination of a cofactor in the culture media of the cell, such as at least a 3-fold upregulation as determined by transcript abundance in a cell when the cell is exposed to extracellular environment changes from containing at least 10 mM or 5 mM cofactor to containing no cofactor.
[0157] Termination control regions may also be derived from various genes native to the preferred hosts. Optionally, the genetic constructs of the present invention do not comprise a termination control region.
[0158] In some embodiments, the bacterial cell or the composition comprising the bacterial cell comprises at least one genetic construct, which comprises one or more coding sequences. In some embodiments, the invention relates to the bacterial cell or the composition comprising at least one bacterial cell wherein the at least one cell comprises two or more genetic constructs, three or more genetic constructs, or four or more genetic constructs, each comprising one or more coding sequences. In some embodiments, the coding sequences of the claimed invention encode at least one protein that modifies or accelerates lipid production in the host cell. In some embodiments the coding sequence encodes at least one protein that alters the levels of individual lipids or hydrocarbons produced by the cell as compared to the same cell not modified by an exogenous nucleic acid sequence. In some embodiments, the coding sequence may encode at least one protein that alters the amount of one specific lipid or hydrocarbon molecule of the cell as compared to the same cell not modified by the nucleic acid. For example, in one embodiment, the one or more exogenous nucleic acid sequence encodes at least one lipid pathway enzyme that causes an increase in the ratio of C14:C16:C18 unsaturated lipids or desaturated hydrocarbons produced or secreted by the cell as compared to the C14:C16:C18 unsaturated lipids or desaturated hydrocarbons produced or secreted by the same cell not transformed with the nucleic acid sequence that encodes the lipid pathway enzyme. In one embodiment, the one or more exogenous nucleic acid sequence encodes at least one lipid pathway enzyme that causes a decrease in the ratio of C14:C16:C18 unsaturated lipids or desaturated hydrocarbons produced or secreted by the cell as compared to the C14:C16:C18 unsaturated lipids or desaturated hydrocarbons produced or secreted by the same cell not transformed with the nucleic acid sequence that encodes the lipid pathway enzyme. In some embodiments, the one or more cells comprising one or more exogenous nucleic acid sequences produces more hydrocarbon than the same one or more cells not transformed or modified with the one or more exogenous nucleic acid sequences.
[0159] In some embodiments, the bacterial cell produces and/or secretes one or more unsaturated lipids or hydrocarbons in a ratio greater than the ratio of unsaturated lipids or hydrocarbons produced and/or secreted by the same cell not comprising the one or more exogenous nucleic acid sequences. In some embodiments, the bacterial cell produces and/or secretes one or more unsaturated lipids or hydrocarbons, wherein at least 50% of the one or more lipids or hydrocarbons have 8 to 18 carbon atoms. In some embodiments, the bacterial cell produces and/or secretes one or more lipids or hydrocarbons, wherein at least 60% of the one or more lipids or hydrocarbons have 8 to 18 carbon atoms. In some embodiments, the bacterial cell produces and/or secretes one or more lipids or hydrocarbons, wherein at least 70% of the one or more lipids or hydrocarbons have 8 to 18 carbon atoms. In some embodiments, the bacterial cell produces and/or secretes one or more lipids or hydrocarbons, wherein at least 75% of the one or more lipids or hydrocarbons have 8 to 18 carbon atoms. In some embodiments, the bacterial cell produces and/or secretes one or more lipids or hydrocarbons, wherein at least 80% of the one or more lipids or hydrocarbons have 8 to 18 carbon atoms.
[0160] In some embodiments, the invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more unsaturated hydrocarbons, wherein the microorganism comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase, wherein the one or more hydrocarbons have a carbon chain length of at least 8 carbon atoms. In some embodiments, The invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more unsaturated hydrocarbons, wherein the microorganism comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase; wherein the one or more hydrocarbons comprise a mixture of hydrocarbon molecules having a carbon chain length from 8 carbon atoms to 18 carbon atoms. In some embodiments, the invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more unsaturated lipids, wherein the microorganism comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase; wherein the one or more lipids comprise a quantity of at least one alkene, or alkyne at a level higher than the quantity of the alkene, alkyne in the same microorganism not comprising the heterologous nucleic acid sequences. In some embodiments, the invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more unsaturated lipids, wherein the microorganism comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase; wherein the microorganism produces and/or secretes at least 60% of one or more lipids by weight.
[0161] In some embodiments, the invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more unsaturated lipids, wherein the microorganism comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase; wherein the microorganism produces and/or secretes at least 65% of one or more lipids by weight. In some embodiments, the invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more unsaturated lipids, wherein the microorganism comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase; wherein the microorganism produces and/or secretes at least 70% of one or more hydrocarbons by weight.
[0162] In some embodiments, the invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more unsaturated lipids, wherein the microorganism comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase; wherein the microorganism produces and/or secretes at least 75% of one or more lipids by weight. In some embodiments, the invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more unsaturated lipids, wherein the microorganism comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase; wherein the microorganism produces and/or secretes at least 80% of one or more lipids by weight.
[0163] In some embodiments, the invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more unsaturated lipids, wherein the microorganism comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase; wherein the microorganism produces and/or secretes at least 85% of one or more lipids by weight. In some embodiments, the invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more unsaturated hydrocarbons, wherein the microorganism comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase; wherein less than 10% by weight of the hydrocarbons produced is methane. In some embodiments, the invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more organic compounds, wherein the microorganism comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase; wherein less than 10% by weight of the organic compounds produced are organic acids with carbon chain length of four carbons or less.
[0164] In some embodiments, the invention relates to a composition comprising a microorganism that converts syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas into one or more unsaturated hydrocarbons, wherein the microorganism comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase; wherein the hydrocarbons produced comprise a mixture of at least two hydrocarbons having a carbon backbone from 8 to 18 carbon atoms.
[0165] The present invention also relates to a bacterial cell comprising at least two exogenous nucleic acid sequences, wherein the at least two exogenous nucleic acid sequences encode fatty acid acyl-ACP reductase and fatty acid aldehyde decarbonylase, and wherein the cell converts gaseous CO2 and/or gaseous H2 and/or syngas into lipids. In some embodiments, the invention relates to a bacterial cell comprising at least two exogenous nucleic acid sequences, wherein the at least two exogenous nucleic acid sequences encode fatty acid acyl-ACP reductase and fatty acid aldehyde decarbonylase, and wherein the cell converts gaseous CO2 and/or gaseous H2 and/or syngas into lipid; wherein the cell produces and/or secretes at least 75% of one or more hydrocarbons by weight. In some embodiments, the invention elates to a bacterial cell comprising at least two exogenous nucleic acid sequences, wherein the at least two exogenous nucleic acid sequences encode fatty acid acyl-ACP reductase and fatty acid aldehyde decarbonylase, and wherein the cell converts gaseous CO2 and/or gaseous H2 and/or syngas into lipid; wherein the cell produces and/or secretes at least 75% of one or more hydrocarbons by weight when cultured at least 42 degrees Celsius for at least 1 hour. In some embodiments, the bacterial cell is cultured without exposure to light.
[0166] In some embodiments, the invention relates to a bacterial cell comprising at least two exogenous nucleic acid sequences, wherein the at least two exogenous nucleic acid sequences encode fatty acid acyl-ACP reductase and fatty acid aldehyde decarbonylase, and wherein the cell converts gaseous CO2 and/or gaseous H2 and/or syngas into a hydrocarbon or mixture of hydrocarbons, and/or other lipids; wherein the cell is a strain of Rhodococcus opacus.
[0167] In some embodiments, the invention relates to a bacterial cell comprising at least two exogenous nucleic acid sequences, wherein the at least two exogenous nucleic acid sequences encode fatty acid aldehyde acyl-ACP and fatty acid aldehyde decarbonylase, and wherein the cell converts gaseous CO2 and/or gaseous H2 and/or syngas into a hydrocarbon or mixture of hydrocarbons, and/or other lipids; wherein the cell is a strain of Cupriavidus necator.
[0168] In some embodiments, the invention relates to a bacterial cell comprising a first, a second, and a third exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase, the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase, and the third exogenous nucleic acid sequence encodes a thioesterase; and wherein the cell converts gaseous CO2 and/or gaseous H2 and/or syngas into a lipid or mixture of lipids. In some embodiments, the bacterial cell comprises no more than eight exogenous nucleic acids that encode a lipid pathway enzyme. In some embodiments, the bacterial cell comprises no more than seven exogenous nucleic acids that encode a lipid pathway enzyme. In some embodiments, the bacterial cell comprises no more than six exogenous nucleic acids that encode a lipid pathway enzyme. In some embodiments, the bacterial cell comprises no more than five exogenous nucleic acids that encode a lipid pathway enzyme. In some embodiments, the bacterial cell comprises no more than four exogenous nucleic acids that encode a lipid pathway enzyme. In some embodiments, the bacterial cell comprises no more than three exogenous nucleic acids that encode a lipid pathway enzyme. In some embodiments, the bacterial cell comprises no more than two exogenous nucleic acids that encode a lipid pathway enzyme. In some embodiments, the bacterial cell comprises no more than one exogenous nucleic acid that encodes a lipid pathway enzyme. In some embodiments, the bacterial cell comprises no more than eight exogenous nucleic acids that encode a protein. In some embodiments, the bacterial cell comprises no more than seven exogenous nucleic acids that encode a protein. In some embodiments, the bacterial cell comprises no more than six exogenous nucleic acids that encode a protein. In some embodiments, the bacterial cell comprises no more than five exogenous nucleic acids that encode a protein. In some embodiments, the bacterial cell comprises no more than four exogenous nucleic acids that encode a protein. In some embodiments, the bacterial cell comprises no more than three exogenous nucleic acids that encode a protein. In some embodiments, the bacterial cell comprises no more than two exogenous nucleic acids that encode a protein. In some embodiments, the bacterial cell comprises no more than one exogenous nucleic acid that encodes a protein.
[0169] In some embodiments, the molecule is chosen from one or more alkene, alkyne, unsaturated fatty acid, hydroxyacid and/or dicarboxylic acid (diacid). In some embodiments, the method produces a lipid or mixture of lipids at a quantity higher than the quantity of lipid or mixture of lipids in the same bacterial cell population not comprising the exogenous nucleic acids described herein. In some embodiments the one or more lipids comprise a quantity of at least one alkene, alkyne, unsaturated fatty acid, hydroxyacid and/or diacid at a level higher than the quantity of the alkene, alkyne, unsaturated fatty acid, hydroxyacid and/or diacid in the same microorganism not comprising the exogenous nucleic acid sequences. In some embodiments, the method comprises a population of microorganisms or bacterial cell described herein that produces and/or secretes lipids of a weight equal to or greater than 10% of the total percentage of cellular dry matter. In some embodiment, the method comprises a population of microorganisms or bacterial cell described herein that produces and/or secretes lipids of a weight equal to or greater than 20% of the total percentage of cellular dry matter. In some embodiment, the method comprises a population of microorganisms or bacterial cell described herein that produces and/or secretes lipids of a weight equal to or greater than 30% of the total percentage of cellular dry matter. In some embodiments, the method comprises a population of microorganisms or bacterial cell described herein that produces and/or secretes lipids of a weight equal to or greater than 40% of the total percentage of cellular dry matter. In some embodiment, the method comprises a population of microorganisms or bacterial cell described herein that produces and/or secretes lipids of a weight equal to or greater than 50% of the total percentage of cellular dry matter. In some embodiments, the method comprises a population of microorganisms or bacterial cells described herein that produces and/or secretes lipids of a weight equal to or greater than 60% of the total percentage of cellular dry matter. In some embodiments, the method comprises a population of microorganisms or bacterial cells described herein that produces and/or secretes lipids of a weight equal to or greater than 70% of the total percentage of cellular dry matter. In some embodiments, the method comprises a population of microorganisms or bacterial cell described herein that produces of secretes lipids of a weight equal to or greater than 75% of the total percentage of cellular dry matter. In some embodiment, the method comprises a population of microorganisms or bacterial cell described herein that produces of secretes lipids of a weight equal to or greater than 80% of the total percentage of cellular dry matter. In some embodiments, the method comprises a population of microorganisms or bacterial cell described herein that produces of secretes lipids of a weight equal to or greater than 85% of the total percentage of cellular dry matter. In some embodiments, the bacterial cell or composition comprising the bacterial cell produces and/or secretes at least 10% of the total percentage of the cellular dry matter or 10% by weight. In some embodiment, the method comprises a population of microorganisms comprising a bacterial cell described herein that produces or secretes lipids, wherein at least 5% of the lipids have carbon backbones from 8 to 18 carbon atoms in length. In some embodiment, the method comprises a population of microorganisms comprising a bacterial cell described herein that produces or secretes lipids, wherein at least 10% of the lipids have carbon backbones from 8 to 18 carbon atoms in length. In some embodiments, the method comprises a population of microorganisms comprising a bacterial cell described herein that produces or secretes lipids, wherein at least 15% of the lipids have carbon backbones from 8 to 18 carbon atoms in length. In some embodiments, the method comprises a population of microorganisms comprising a bacterial cell described herein that produces or secretes lipids, wherein at least 20% of the lipids have carbon backbones from 8 to 18 carbon atoms in length.
[0170] In some embodiments of the invention, the invention relates to a method of fixing carbon from a gaseous feedstock containing carbonaceous molecules, wherein the method comprises the step of exposing a composition comprising a bacterial cell to syngas and/or gaseous CO2 and/or gaseous H2; wherein the bacterial cell comprises at least one exogenous nucleic acid sequence. In some embodiments the exogenous nucleic acid sequences are fatty acid acyl-ACP reductase or a fatty acid aldehyde decarbonylase. In some embodiments of the method, the bacterial cell comprises at least a first and a second exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes a fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase. In some embodiments, the bacterial cell is Rhodococcus opacus or the population of microorganisms comprises a Rhodococcus cell. In some embodiments, the bacterial cell is Cupriavidus necator or the population of microorganisms comprises a Cupriavidus cell. In some embodiments, the bacterial cell comprises at least a first, a second, and a third exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes a fatty acid acyl-ACP reductase, the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase, and the third exogenous nucleic acid sequence encodes a thioesterase. In some embodiments, the bacterial cell comprises no more than five exogenous nucleic acid sequences that encode a lipid pathway enzyme. In some embodiments the bacterial cell comprises at least a first and a second exogenous nucleic acid sequence but no more than five exogenous nucleic acid sequences, wherein the first exogenous nucleic acid sequence encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase.
[0171] In some embodiments, the invention relates to a method of producing one or more alkene, alkyne, unsaturated fatty acid, hydroxyacid and/or diacid or any combination thereof comprising exposing a bacterial cell to syngas and/or gaseous CO2 or a mixture of gaseous CO2 and gaseous H2; wherein the bacterial cell is capable of fixing gaseous CO2 into one or more alkene, alkyne, unsaturated fatty acid, hydroxy acid and/or diacid and wherein the microorganism comprises at least a first exogenous nucleic acid and a second exogenous nucleic acid, wherein the first exogenous nucleic acid encodes fatty acid acyl-ACP reductase and the second exogenous nucleic acid encodes fatty acid aldehyde decarbonylase. In some embodiments, the first and second exogenous nucleic acids are heterologous nucleic acid sequences. In some embodiments, the bacterial cell comprises at least a first, a second, and a third exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes a fatty acid acyl-ACP reductase, the second exogenous nucleic acid sequence encodes a fatty acid aldehyde decarbonylase, and the third exogenous nucleic acid sequence encodes a thioesterase. In some embodiments, the bacterial cell comprises at least a first exogenous nucleic acid sequence, wherein the first exogenous nucleic acid sequence encodes a thioesterase. In some embodiments, the bacterial cell comprises no more than five exogenous nucleic acid sequences that encode a lipid pathway enzyme.
[0172] In some embodiments, the one or more exogenous nucleic acid sequence encodes at least one lipid pathway enzyme that causes an increase in the ratio of C12:C14:C16 lipids or hydrocarbons produced or secreted by the cell as compared to the C12:C14:C16 lipids or hydrocarbons produced or secreted by the same cell not transformed with the nucleic acid sequence that encodes the lipid pathway enzyme. In one embodiment, the one or more exogenous nucleic acid sequence encodes at least one lipid pathway enzyme that causes a decrease in the ratio of C12:C14:C16 lipids or hydrocarbons produced or secreted by the cell as compared to the C12:C14:C16 lipids or hydrocarbons produced or secreted by the same cell not transformed with the nucleic acid sequence that encodes the lipid pathway enzyme. In some embodiments, the one or more exogenous nucleic acid sequence encodes at least one lipid pathway enzyme that causes an increase in the ratio of C13:C15:C17 lipids or hydrocarbons produced or secreted by the cell as compared to the C13:C15:C17 lipids or hydrocarbons produced or secreted by the same cell not transformed with the nucleic acid sequence that encodes the lipid pathway enzyme. In one embodiment, the one or more exogenous nucleic acid sequence encodes at least one lipid pathway enzyme that causes a decrease in the ratio of C13:C15:C17 lipids or hydrocarbons produced or secreted by the cell as compared to the C13:C15:C17 lipids or hydrocarbons produced or secreted by the same cell not transformed with the nucleic acid sequence that encodes the lipid pathway enzyme. In some embodiments, the one or more exogenous nucleic acid sequence encodes at least one lipid pathway enzyme that causes an increase in the ratio of odd-carbon numbered lipids or hydrocarbons produced or secreted by the cell as compared to the odd-carbon numbered lipids or hydrocarbons produced or secreted by the same cell not transformed with the nucleic acid sequence that encodes the lipid pathway enzyme. In some embodiments, the one or more exogenous nucleic acid sequence encodes at least one lipid pathway enzyme that causes a decrease in the ratio of odd-carbon numbered lipids or hydrocarbons produced or secreted by the cell as compared to the odd-carbon numbered lipids or hydrocarbons produced or secreted by the same cell not transformed with the nucleic acid sequence that encodes the lipid pathway enzyme. In one embodiment, the one or more exogenous nucleic acid sequence encodes at least one lipid pathway enzyme that causes a decrease in the ratio of even:odd carbon numbered lipids or hydrocarbons produced or secreted by the cell as compared to the ratio of even:odd carbon numbered lipids or hydrocarbons produced or secreted by the same cell not transformed with the nucleic acid sequence that encodes the one or more lipid pathway enzymes. In one embodiment, the one or more exogenous nucleic acid sequence encodes at least one lipid pathway enzyme that causes a increase in the ratio of even:odd carbon numbered lipids or hydrocarbons produced or secreted by the cell as compared to the ratio of even:odd carbon numbered lipids or hydrocarbons produced or secreted by the same cell not transformed with the nucleic acid sequence that encodes the one or more lipid pathway enzymes.
[0173] In some embodiments the exogenous gene or genes codes for enzymes or proteins including but not limited to one or more of the following: an acyl carrier protein, a TE, a FAR, a FadR, a FAD, a fatty aldehyde reductase, a cytochrome P450 enzyme, a NADH or NADPH cytochrome P450 reductase, a desaturase, a hydroxylase, and an antibiotic resistance enabling protein. In some embodiments, the coding sequence comprises an exogenous nucleic acid sequence that encodes a TE that catalyzes hydrolysis of one or more fatty acyl-ACP substrates with chain lengths ranging over C8, C9, C10, C11, C12, C13, C14, C15, C16, C17, or C18. In some embodiments, the cell comprises a plasmid that contains one or more exogenous nucleic acid sequences that encode an amino acid sequence for an enzyme or protein such as but not limited to one or more of the following: an acyl carrier protein, a TE, a FAR, a FadR, a FAD, a fatty aldehyde reductase, a cytochrome P450 enzyme, a NADH or NADPH cytochrome P450 reductase, a desaturase, a hydroxylase, and an antibiotic resistance enabling protein. In some embodiments, the one or more exogenous nucleic acid sequences comprise SEQ ID NO:5 or a functional fragment thereof that is at least 70%, 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, or 99% homologous to SEQ ID NO:5. In some embodiments, the one or more exogenous nucleic acid sequences comprise SEQ ID NO:6 or a functional fragment thereof that is at least 70%, 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, or 99% homologous to SEQ ID NO:6.
[0174] In further embodiments, at least one coding sequence of the at least one exogenous nucleic acid sequence encodes a lipid pathway enzyme. In some embodiments, the at least one coding sequence of the at least one exogenous nucleic acid sequence encodes a lipid modification enzyme. In some embodiments, the composition or cell comprises a nucleic acid that encodes at least one fatty acid decarbonylase, at least one fatty acid reductase, a thioesterase, or any combination of any one more full-length lipid pathway enzymes or functional fragments thereof. In some embodiments the one or more exogenous nucleic acid sequences are integrated into the genome of the cell. In some embodiments, the one or more exogenous nucleic acid sequences are on an episomal plasmid within the transformed host cell.
[0175] Methods of Isolation and Purification
[0176] Following the methods of the present invention microorganisms are grown and maintained for the production of lipids in a medium containing a gaseous carbon source, such as but not limited to syngas or producer gas, in the absence of light; such growth is known as chemotrophic growth. In some embodiments, the invention relates to methods of cultivating oleaginous cells for the large-scale production of oil and/or fuel. In some embodiments, the invention relates to methods of cultivating oleaginous cells in bioreactors 50,000 liters or greater in volume, which are conventionally constructed out of low cost, sturdy, and opaque materials such as steel or reinforced concrete or earthworks. The size, depth, and construction of such bioreactors dictate that the cells will be grown in near or total darkness. In some embodiments, the oleaginous microorganisms are cultured for the synthesis of lipids in accordance with the methods of the present invention in a medium containing gaseous inorganic carbon, such as but not limited to syngas or producer gas, as the primary or sole carbon source, and without any exposure to light. This type of growth is known as chemoautotrophic growth.
[0177] To give an illustration, a bioreactor containing nutrient medium is inoculated with of oleaginous bacterial cells; generally there will follow a lag phase prior to the cells beginning to double. After the lag phase, the cell doubling time decreases and the culture goes into the logarithmic phase. The logarithmic phase is eventually followed by an increase of the doubling time that, while not intending to be limited by theory, is thought to result from either a depletion of nutrients including nitrogen sources, or a rise in the concentration of inhibitory chemicals, or quorum sensing by the microbes. The growth slows down and then ceases when the culture goes into the stationary phase. In order to harvest cell mass with high lipid content, the culture is generally harvested late in the logarithmic phase or in the stationary phase. In some embodiments, the cells are harvested in logarithmic phase. In some embodiments, the cells are harvested in stationary phase. The accumulation of lipid can generally be triggered by the depletion of the nitrogen source or another key nutrient excepting the carbon or the energy source (e.g. hydrogen). This signals the cells to store lipids produced from the excess carbon and energy sources. Optimization of lipid production and the targeting of specific lipid distributions can be achieved by control of bioreactor conditions and/or nutrient levels and/or through genetic modifications of the cells. In some embodiments the lipid production and distribution of lipid molecules produced is optimized through one or more of the following: control of bioreactor conditions, control of nutrient levels, genetic modifications of the cells.
[0178] The synthesis of lipids by the microbes disclosed in the present invention can happen during the logarithmic phase and afterwards during the stationary phase when cell doubling has stopped provided there is an ample supply of carbon and energy sources,
[0179] In some embodiments, microorganisms grown using conditions described herein and known in the art comprise at least 20% lipid content by weight, but under chemotrophic conditions, comprise at least 10% lipid content by weight. In some embodiments, under chemotrophic conditions, the microorganisms of the present invention comprise at least about 10, 15, 20, 25, 30, 35, or 40% by weight of lipids, at least about 50% by weight, or at least about 60% by weight of lipids. Improved lipid yield and/or lower production costs can be achieved by controlling process parameters. In certain embodiments, a bacterium is grown in a nutrient media and/or gas mix having a nitrogen, oxygen, phosphorous, or sulfur limitation, while a gaseous carbon and energy source such as syngas is provided in excess. Lipid yield is generally higher in microbial cultures grown with a nitrogen limitation versus microbial cultures grown without nitrogen limitation. In certain embodiments, lipid yield rises by at least: 10%, 50%, 100%, 200%, 500%, or 1000%. The microbial growth can occur with nutrient limitation for a part or for all of the fermentation run. Feeding an excess of energy and carbon source to a population of oleaginous microbes, but little or no nitrogen, can produce a rise in cellular lipid content. In some embodiments, microbial growth occurs on limited amounts of nitrogen or in the complete absence of nitrogen.
[0180] Genes are well known in the art that code for cofactors useful in the present invention, or that are involved in synthesizing such cofactors.
[0181] In another embodiment, genes that code for cofactors useful in the present invention, or that are involved in synthesizing such cofactors, are put in oleaginous bacteria, using the constructs and methods such as described above. Lipid yield is improved in another embodiment by growing an oleaginous bacteria with one or more lipid pathway enzyme cofactor(s) added to the culture environment. The lipid yield is generally improved in the presence of a certain concentration of the cofactor(s) compared to lipid yield without supplemental cofactor(s). In some embodiments, the cofactor(s) are delivered to the culture by having a microbe (e.g., bacteria) present in the culture that contains an exogenous gene coding for the cofactor(s) at a concentration sufficient to increase lipid yield as compared to the lipid yield of the microbe in the absence of the cofactor. Cofactor(s) may also be delivered to a culture by having a microbe (e.g., bacteria) present in the culture that contains an exogenous gene that coding for a protein involved in the cofactor synthesis. In some embodiments, any vitamin needed for the proper function of a lipid pathway enzyme including biotin and/or pantothenate is included in the culture environment.
[0182] The specific examples of bioreactors, culture conditions, heterotrophic and chemotrophic growth, maintenance, and lipid production methods described herein can be combined in any suitable manner to improve efficiencies of microbial growth and lipid and/or protein production.
[0183] In another aspect of the invention, the invention relates to a method of producing a molecule or mixture of molecules in a microorganism population comprising the cell or the composition described herein, wherein the method comprises: culturing a population of microorganisms comprising the cell or the composition described herein in a feedstock comprising syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas.
[0184] In another aspect of the invention, the invention relates to a method of producing a hydrocarbon or mixture of hydrocarbons in a microorganism population comprising the cell or the composition described herein, wherein the method comprises: culturing a population of microorganisms comprising the cell or the composition described herein in a feedstock comprising syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas.
[0185] In another aspect of the invention, the invention relates to a method of producing a lipid or mixture of lipids in a microorganism population comprising the cell or the composition described herein, wherein the method comprises: culturing a population of microorganisms comprising the cell or the composition described herein in a feedstock comprising syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas.
[0186] In another aspect of the invention, the invention relates to a method of producing an alkene or mixture of alkenes in a microorganism population comprising the cell or the composition described herein, wherein the method comprises: culturing a population of microorganisms comprising the cell or the composition described herein in a feedstock comprising syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas.
[0187] In another aspect of the invention, the invention relates to a method of producing an alkyne or mixture of alkynes in a microorganism population comprising the cell or the composition described herein, wherein the method comprises: culturing a population of microorganisms comprising the cell or the composition described herein in a feedstock comprising syngas and/or gaseous CO2 and/or a mixture of CO2 gas and H2 gas.
[0188] In some embodiments, the methods of the claimed invention do not rely on desulfonation to produce and/or secrete one or more hydrocarbons. In some embodiments, an exogenous nucleic acid is introduced into the cells of the claimed invention to silence or disrupt transcription of endogenous genes of the cell that encode enzymes capable of desulfonation of commercial surfactants under conditions and for a time period sufficient for growth of the cell with a gaseous feedstock comprising a gas comprising carbon.
[0189] In some embodiments, the feedstock does not include linoleic acid.
[0190] The following documents are incorporated herein by reference in their entirety for all purposes: U.S. Provisional Patent Application No. 61/328,184, filed Apr. 27, 2010 and entitled "USE OF OXYHYDROGEN MICROORGANISMS FOR NON-PHOTOSYNTHETIC CARBON CAPTURE AND CONVERSION OF INORGANIC CARBON SOURCES INTO USEFUL ORGANIC COMPOUNDS"; International Patent Application Serial No. PCT/US2010/001402, filed May 12, 2010, entitled "BIOLOGICAL AND CHEMICAL PROCESS UTILIZING CHEMOAUTOTROPHIC MICROORGANISMS FOR THE CHEMOSYTHETIC FIXATION OF CARBON DIOXIDE AND/OR OTHER INORGANIC CARBON SOURCES INTO ORGANIC COMPOUNDS, AND THE GENERATION OF ADDITIONAL USEFUL PRODUCTS"; and U.S. Patent Application Publication No. 2010/0120104, filed Nov. 6, 2009, entitled "BIOLOGICAL AND CHEMICAL PROCESS UTILIZING CHEMOAUTOTROPHIC MICROORGANISMS FOR THE CHEMOSYTHETIC FIXATION OF CARBON DIOXIDE AND/OR OTHER INORGANIC CARBON SOURCES INTO ORGANIC COMPOUNDS, AND THE GENERATION OF ADDITIONAL USEFUL PRODUCTS.
[0191] Doan T T P, Carlsson A S, Hamberg M, Bulow L, Stymne S, Olsson P, Functional expression of five Arabidopsis fatty acyl-CoA reductase genes in Escherichia coli, J Plant Phys 166(2008):787-96.
[0192] Kavanagh K L, Jornvall H, Persson B, Oppermann U, The SDR superfamily: functional and structural diversity within a family of metabolic and regulatory enzymes, Cell Mol Life Sci 65 (2008) 3895-3906.
[0193] Labesse G, Vidal-Cros A, Chomilier J, Gaudry M, Mornon J-P, Structural comparisons lead to the definition of a new superfamily of NAD(P)(H)-accepting oxidoreductases: the single-domain reductases/epimerases/dehydrogenases (the `RED` family), Biochem J (1994) 304:95-99.
[0194] The following examples are intended to illustrate certain embodiments of the present invention, but do not exemplify the full scope of the invention. Various publications, including patents, published applications, technical articles and scholarly articles are cited throughout the specification. Each of these cited publications is incorporated by reference herein, in its entirety.
Example 1
[0195] Characterization of organisms sharing high 16SrRNA sequence similarity.
[0196] To identify organisms closely related to R. opacus strain (DSM43205), a basic local alignment search (BLASTR) with the BLASTN programs search of nucleotide databases using the 16S rRNA (NR_026186.1) was carried out. The phylogenetic relationships, based on the 16S rRNA gene sequence homology, between the tested strain and the reference strains of the suborder corynebacterineae (corynebacterium, gordoniaceae, mycobacteriaceae and nocardiaceae) and the family burkholderiaceae (genus cupriavidus and ralstonia) are shown in FIG. 2. The nocardiaceae are related and form two clusters of organisms: clusture 1 that contains 20 organisms from the genus nocardia and rhodococcus and cluster 2 that contains 3 R. opacus strains (DSM43205, GM14 and DSM43206). The gordoniaceae, mycobacteriaceae and burkholderiaceae form 3 separated groups (1, 2 and 3). The gram positive chemoautotroph lipid accumulating strain R. opacus (DSM43205; NR_026186.1) exhibits high sequence similarity to cluster 1 (94.3-99.1%) and to the gram positive groups 1 and 2 (92.7-93.5% and 93.3-93.6% respectively) (FIGS. 3 and 4). The sequence similarity to the gram-negative chemoautotroph poly(3-hydroxybutyrate) (PHB) accumulating strains in group 3 is 73.7%.
Example 2
[0197] Plasmid Design and Construction
[0198] To generate an E. coli Rhodococci shuttle vector suitable for electroporation, the plasmid pSeqCO1 (SEQ ID: 01) was constructed with the genetic elements described in FIG. 10A. pSeqCO1 consists of the replication gene operon, ampicillin and kanamycin resistance genes, LacZ operon and the multiple cloning site as described in FIG. 10B and FIG. 11A. For replication in Rhodococci, the DNA fragment of the repAB operon (1744 bp downsteam from the XhoI restriction site in the native pKNR01 plasmid of the bacteria Rhodococcus opacus B4; Na et al. 2005, J Biosci Bioeng. 99: 408-414) was synthesized with the restriction sites KpnI and SalI and cloned into PUC18 digested with KpnI and SalI. The resultant vector was digested with SpeI and BglII and ligated with the PCR product of the Kanamycin resistance gene from pBBR1MCS-2 (Kovach et al. 1995 Gene 166: 175-176) digested with the engineered restriction sites SpeI and BglII to give pSeqCO1.
[0199] To generate an E. coli-cupriavidus shuttle vector suitable for electroporation and bacterial conjugation, the plasmid pSeqCO2 (SEQ ID: 02) was used with the genetic elements described in FIG. 10A. pSeqCO2 (SEQ ID: 02; FIGS. 10 and 11B) is the plasmid pBBR1MCS-2 described in Kovach et al. (1995 Gene 166: 175-176) that contains the IncQ like replication gene, Mob gene that mobilized when the RK2 transfer functions are provided in trans, kanamycin resistance gene, LacZ operon and the multiple cloning site as described in FIG. 10B and FIG. 11B.
[0200] Pver1 (SEQ ID: 03; FIGS. 10 and 11C) is an E. coli-cupriavidus-Rhodococci shuttle vector suitable for electroporation and bacterial conjugation. The plasmid was generated by cloning the repAB operon (described in pSeqCO1) into pSeqCO2 using the KpnI and SalI restriction sites.
[0201] Pver2 (SEQ ID: 04; FIGS. 10 and 11D) is an E. coli-cupriavidus-Rhodococci shuttle vector suitable for electroporation and bacterial conjugation. The plasmid was generated by cloning the synthesized chloramphenicol gene (Alton and Vapnek Nature 1979 282: 864-869) with the engineered restriction sites SalI and HindIII into Pver1.
[0202] The genes FadDR (SEQ ID: 05) and Fad (SEQ ID: 06) and the rbcLXS promoter (SEQ ID: 7) were PCR amplified from the cyanobacterium Synechocystis sp. PCC 6803 genome and cloned into gateway plasmid to give pFUEL. A 4 kBp XhoI BamHI fragment that contains FadDR, Fad and rbcLXS was rescued from pFUEL and cloned into pSeqCO2 digested XhoI BamHI with to give pSeqCO2::FUEL (FIG. 14).
Example 3
[0203] Microorganism Transformation
[0204] Transformation of Rhodococci was carried out using the plasmids pSeqCO1 and pVer1 (FIG. 12) as described below.
[0205] Rhodococci competent cells were prepared by incubating a single colony 2 ml NB medium (5 g/L peptone, 1 g/L meat extract, 2 g/L yeast extract, 5 g/L NaCl; pH=7.0.+-.0.2) at 30.degree. C. overnight. One ml was inoculated to 50 ml NB medium supplemented with 0.85% (w/v) glycine and 1% (w/v) sucrose in a 250 ml baffled Erlenmeyer Flask and incubated to a cell density of O.D600=0.5. Cells were collected by centrifugation at 3,000.times.g for 10 min at 4.degree. C. and washed 3 times with 50 ml (each) of sterile ice-cold double distilled water (ddH2O). Cells were concentrated 20-fold by re-suspending the collected cells in 2.5 ml of ddH2O and 400 .mu.l aliquots stored in 1.5 ml tube at -70.degree. C. Electroporation was carried out by thawing the competent cells on ice and mixing with the plasmid DNA (final concentration 0.1-0.25 .mu.g/ml). The competent cells and plasmid DNA mixture was incubated at 40.degree. C. for 5 min, transferred into 0.2 cm width and electroporated using a single-pulse electroporation (10 kV/cm, 600.OMEGA., 25 .mu.F and 3-5 ms pulse time). The pulsed cells were regenerated at 30.degree. C. for 4 h (DSM 44193) and 6 h (DSM 43205) in the presence of 600 .mu.l NB. Transformants were selected after cultivation for 3-4 days at 30.degree. C. on NB-agar plate containing kanamycin (75 .mu.g/ml). As shown in FIG. 12, the plasmids pSeqCO1 and pVcr1 confer resistance to kanamycin (75 .mu.g/ml) in transformed R. opacus strains (44193 and 43205). Untransformed R. opacus strains (44193 and 43205) (NC) were sensitive to the concentration described above.
[0206] Transformation of genus cupriavidus was carried out using the plasmids pSeqCO2 (FIG. 12) as described below.
[0207] Cupriavidus necator (DSM531) competent cells were prepared by incubating a single colony in 5 ml NR medium (10 g/1 polypeptone, 10 g/1 yeast extract, 5 g/1 beef extract and 5 g/1 ammonium sulfate; pH 7.0) at 30.degree. C. overnight. The pre-culture was inoculated into 100 ml of fresh NR medium and incubated to a cell density of O.D600=0.8. Cells were collected by centrifugation at 3,000.times.g for 10 min at 4.degree. C. and washed 3 times with 50 ml (each) of sterile ice-cold ddH2O. The collected cells were re-suspended in 400 .mu.l of 10% (v/v) sterile glycerol in sterile ice-cold ddH2O and stored in 50 .mu.l aliquots at -70.degree. C.
[0208] For electroporation, the competent cells were thawed on ice, transferred into 0.2 cm width of ice cold cuvette and gently mixed with 1 .mu.g of plasmid DNA. Cells were electroporated using a single-pulse electroporation (11.5 kV/cm, 25 .mu.F and 5 ms pulse time). The pulsed cells were transferred into 1 ml of fresh NR medium and culture for 2 h at 30.degree. C. Transformants were selected after cultivation for 48 h at 30.degree. C. on NR-agar plate containing kanamycin (200 .mu.g/ml). As shown in FIG. 12, the plasmid pSeqCO2 confers resistance to kanamycin (200 .mu.g/ml) in transformed Cupriavidus necator (DSM531). Untransformed Cupriavidus necator (DSM531) cells (NC) were sensitive to the concentration described above.
Example 4
[0209] Inoculation and Growth Conditions
[0210] Organisms from the genus rhodococcus and from the genus cupriavidus were tested for their ability to grow on different carbon sources (FIG. 5). Colonies from strains grown on LB agar plates at 30.degree. C. were transferred into flasks containing 10% (v/v) of the indicated media for 3-20 days at 30.degree. C. and 250 rpm. R. opacus strain DSM 44193 exhibited growth only under heterotrophic growth conditions as measured by optical density (OD) at 650 nm on MSM medium (1 L Medium A: 9 g Na2HPO412H2O, 1.5 g H2PO4, 1.0 g NH4Cl and 0.2 g MgSO4.7H2O per 1 L; 10 ml Medium B: 50 mg Ferric ammonium citrate and 100 mg CaCl2 per 100 ml; 10 ml Medium C: 5 g NaHCO.sub.3 per 100 ml; and 1 ml Trace Mineral Solution: 100 mg ZnSO4.7H2O, 30 mg MnCl2. 4H2O, 300 mg H3BO3, 200 mg COCL2.6H2O, 10 mg CuC12.2H2O, 20 mg NiCl2.6H2O and 30 mg Na2MoO4.2H2O per 1 L) supplemented with 40 g/L glucose. R. opacus strain DSM 43205 showed identical growth rates under heterotrophic conditions reaching O.D=9.0. Strain DSM 43205 was also able to grow on chemoautotrophic conditions (MSM medium supplemented with 66.7% H2, 9.5% CO2, 5% 02 and 18.8% N2) and heterotrophically on a single carbon compound as the solely carbon source (MSM medium supplemented with 25 g/1 methanol). Rhodococcus sp. (DSM 3346) exhibited growth under heterotrophic conditions and chemoautotrophic conditions (DSMZ Medium 81: 1 L of Mineral Medium for chemolithotrophic growth: 2.9 g Na2HPO4.2H2O, 2.3 g KH2PO4, 1.0 g NH4Cl, 0.5 g MgSO4.7H2O, 0.5 g NaHCO.sub.3, 0.01 g CaCl.2H2O and 0.05 g Fe(NH4) citrate per 1 L; and 5 ml Trace Mineral Solution, supplemented with 80% H2, 10% CO2 and 10%02). Cupriavidus necator (DSM 531) was able to grow under heterotrophic and chemoautotrophic conditions (media described for Strain DSM 43205) (FIG. 5 and FIG. 22). Cupriavidus necator (DSM 531) transformed with pSeqCO2 was able to grow on LB media supplemented with 300 400 and 500 .mu.g/ml kanamycin exhibiting O.D600 of 1.47, 1.52 and 1.51 respectively (FIG. 13). Untransformed cells exhibited growth on control (LB only) and some growth on 300 .mu.g/ml kanamycin while no growth was detected on 400 and 500 .mu.g/ml kanamycin.
Example 5
[0211] Lipid Profiles
[0212] Production of Fatty Acid
[0213] Under heterotrophic growth conditions strains DSM 44193, DSM 43205, DSM 3346 and DSM 531 produce lipid (FIG. 6). Lipid content determined by gas chromatography analysis of cells harvested after 72 hr (unless otherwise indicated) showed over 19% of cellular dry matter (CDM) determined gravimetrically for strains DSM 44193, DSM 43205 and DSM 3346. The lipid content of DSM 43205 was higher than 10% of under chemoautotrophic conditions. Under heterotrophic growth conditions DSM 44193 produces 32%, 26% and 21% of 16, 17 and 18-carbon fatty acid respectively (FIG. 7). DSM43205 produces similar amounts of 16, 17 and 18-carbon fatty acid (30%, 24% and 32% respectively) (FIG. 8A). Chemoautotrophic growth condition significantly reduces the 17-carbon fatty acid abundance (6%) and maintains similar levels of 16 and 18-carbon fatty acid (36% and 27% respectively) (FIG. 8B). DSM3346 exhibits similar fatty acid distribution of 16, 17 and 18-carbon fatty acid (39%, 24% and 25% respectively) (FIG. 9A) under heterotrophic growth. Chemoautotrophic growth condition significantly increases the 16-carbon fatty acid levels (66%) and reduces the 17 and 18-carbon fatty acid levels (4%, 14%)(FIG. 9B).
Example 6
[0214] Production of Hydrocarbons
[0215] To redirect carbon flux from fatty acid toward hydrocarbon biosynthesis, the genes Fatty acyl-CoA/Fatty acyl-ACP reductase (FadR) and Fatty aldehyde decarbonylase (FAD) from the decarbonylation pathway of cyanobacteria (indicated in red) were expressed in Cupriavidus necator (DSM 531).
[0216] The plasmid pSeqCO2::FUEL (FIG. 14) described in the text was introduced into Cupriavidus necator (DSM 531) as described above and 2 independent transformants (Cn-FUEL2.1 and Cn-FUEL2.2) were selected. One hundred ml of Cn-FUEL2.1, Cn-FUEL2.2 and control cells (empty plasmid: Cn-P) were incubated on LB medium with 400 .mu.g/ml kanamycin for 30 hr. Cells were harvested at 3,000.times.g for 10 min at 4.degree. C. and pellet was analyzed by GC/MS. Cn-FUEL2.1 (FIG. 15A) and Cn-FUEL2.2 showed a specific peak at 45.00 min compared to control Cn-P (FIG. 15B) indicating the presence of hydrocarbons in the engineered strains. Cn-FUEL2.1, Cn-FUEL2.2 produced high levels (over 2%) of unique molecules such as: Spiro[4.5]decane, Bicyclo[10.8.0]eicosane, cis,cis-1,6-Dimethylspiro[4.5]decane, 1,19-Eicosadiene, Cyclooctacosane, Bicyclo[10.8.0]eicosane, 1-Pentadecyne, 1-Pentadecyne, Heptacosyl acetate, 5-Cyclohexyl-1-pentene, 1-Hexadecyne and Cyclodecacyclotetradecene, -eicosahydro (FIGS. 16 and 17).
[0217] The effect of the production of hydrocarbons on fatty acid distribution is shown in FIG. 18. The fatty acids profile of 2 independent control experiments (Cn-P) shows predominantly 16-carbon (63% and 61%) and 18-carbon (33% and 32%) fatty acids. In contrast, Cn-FUEL2.1 and Cn-FUEL2.2 exhibit significantly lower levels of 16-carbon (29%, 33% respectively) and 18-carbon (3% and 2% respectively) fatty acids. Cn-FUEL2.1 and Cn-FUEL2.2 show a significant increase in the 15-carbon fatty acid (50% and 45% respectively) compared to 0.08% and 0.09% in the control strains Cn-P.
Example 7
[0218] Purification
[0219] Purification Hydrocarbons
[0220] To produce hydrocarbons in bacteria, genes from the decarbonylation pathway of cyanobacteria, including but not limited to, the FadR (SEQ ID: 05) and FAD (SEQ ID: 06) genes will be cloned into pVer2 (SEQ ID: 04) to give pVer2::FUEL. Bacteria, including but not limited to, R. opacus strain (DSM43205) will be transformed with the plasmid pVer2::FUEL by electroporation and grown in 100 ml LB medium supplemented with 75 .mu.g/ml kanamycin for 30 hr. The cells (2.times.50 ml) will be harvested at 3,000.times.g for 10 min at 4.degree. C. and the pellet and the supernatant further analyzed. Analysis of hydrocarbons from the cell pellet will be carried out in 25 mm.times.150 mm glass tube in the presence of 50 .mu.L of Eicosane standard (approx 200 .mu.g/ml) and 50 .mu.l lipid standard (.about.200 ug/ml). Pellet will be extracted with 5 mL chloroform, 10 ml methanol, 4 ml phosphate buffer (phosphate buffer reagent: 50 mM, pH 7.4, 8.7 g K2HPO4 in 1 L water, and about 2.5 ml 6N HCl to adjust pH=7.4, and 50 ml chloroform per 1 L buffer). The mixture will be vortexed for 30 sec, sonicated for 2 min and incubated in dark for at least 3 hr. Phases will be separated in the presence of 5 mL chloroform and 5 ml ddH2O, vortexed and spun down 2000 rpm for 1 min. The bottom layer will be transferred with a glass Pasteur pipette to clean 16 mm.times.125 mm glass tube with Teflon-lined screw top and dried under N2. The dried extract will be re-suspended in hexane and analyzed by Gas Chromatography for the presence of hydrocarbons, including but not limited to 1-Hexadecyne.
[0221] Purification of Fatty Acids
[0222] To modify the fatty acid distribution in bacteria, thioesterases that regulate the fatty acid chain length, including but not limited to the YP_002784058.1 gene will be cloned into pVer2 (SEQ ID: 04) to give pVer2::TE. Bacteria, including but not limited to, R. opacus strain (DSM43205) will be transformed with the plasmid pVer2::TE by electroporation and grown in 100 ml LB medium supplemented with 75 .mu.g/ml kanamycin for 30 hr. The cells (2.times.50 ml) will be harvested at 3,000.times.g for 10 min at 4.degree. C. and the pellet and the supernatant further analyzed. Analysis of fatty acids from the cell pellet will be carried out in 25 mm.times.150 mm glass tube in the presence of 500 .mu.L Eicosane standard (approx 200 .mu.g/mL) and 50 .mu.L lipid standard (.about.200 ug/ml). Pellet will be extracted with 5 ml chloroform, 10 ml methanol, 4 ml phosphate buffer (phosphate buffer reagent: 50 mM, pH 7.4, 8.7 g K2HPO4 in 1 L water, and about 2.5 mL 6N HCl to adjust pH=7.4, and 50 ml chloroform per 1 L buffer). The mixture will be vortexed for 30 sec, sonicated for 2 min and incubated in dark for at least 3 hr. Phases will be separated in the presence of 5 ml chloroform and 5 ml ddH2O, vortexed and spun down 2000 rpm for 1 min. The bottom layer will be transferred with a glass Pasteur pipette to clean 16 mm.times.125 mm glass tube with Teflon-lined screw top and dried under N2. The dried extract will be re-suspended 1.5 ml of a 10:1:1 mixture of Methanol:CHCl3:concentrated HCl, vortexed and incubated in 60.degree. C. for 14-16 hr (overnight). The extracts will be cooled and 2 ml of ddH2O and 2 ml of hexane will be added, vortexed and centrifuged for 5 min at 2000 rpm for phase separation. The top hexane layer will be transferred to clean 16 mm tube. An additional two hexane extractions (vortex, centrifugation and phase separation) will be carried out in the extract tube. The hexane extracts will be dried in a GC vial and analyzed by Gas Chromatography for the presence of fatty acids, including but not limited to dodecanoic acid.
Example 8
[0223] Dicarboxylic acids with targeted chain length. Genetically engineer bacteria from the suborder corynebacterineae or the family burkholderiaceae to express thioesterases which yield different length fatty acids. For example the YP_002784058.1 gene mentioned previously or:
TABLE-US-00002 UniProt Entry Protein name Organism C length FATB_GOSHI Myristoyl-acyl carrier Gossypium 16:0 protein thioesterase hirsutum FATB_UMBCA Lauroyl-acyl carrier Umbelliularia 12:0 protein thioesterase californica FATB_CINCA Myristoyl-acyl carrier Cinnamomum 14:0 protein thioesterase camphora FATA_CORSA Oleoyl-acyl carrier Coriandrum 18:0 protein thioesterase sativum FATB_CUPHO Myristyl-acyl carrier Cyphea 16:0 protein thioesterase hookeriana
[0224] Thioesterases generating shorter chain fatty acids (e.g., C10:0 or C12:0) will be identified and incorporated into the bacteria from the suborder corynebacterineae and the family burkholderiaceae.
[0225] Extract the resulting lipids and provide as the sole source of carbon to a culture of Candida tropicalis ATCC 20336, which contains the relevant enzymatic pathways to produce the alpha, omega-dicarboxylic acids. Identify and purify dicarboxylic acid end products from second culture.
Example 8
[0226] Dicarboxylic Acids with Targeted Chain Length
[0227] As in previous example, engineer host strain bacteria from the suborder corynebacterineae or the family burkholderiaceae to express thioesterases which yield targeted chain length fatty acids.
[0228] Also, engineer into host strain cytochrome P450 pathway from Candida tropicalis, including the CYP52A genes with NADPH cytochrome P450 reductase to generate dicarboxylic acid from the fatty acids. Craft et al. have identified the key genes for generation of alpha, omega-dicarboxylic acids in Candida tropicalis: CYP52A13, CYP52A14, CYP52A17, CYP52A18, and CYP52A12 along with the corresponding reductase (Craft 2003).
[0229] Perform a single culture, which generates appropriate length fatty acids, then modifies to attach a second carboxylic acid.
Example 9
[0230] Dicarboxylic Acids.
[0231] Culture the hyperthermophilic archaeon Pyrococcus furiosus in order to generate the dicarboxylic acids described in Carballeira et al. (Carballeira 1997). Determine genetic machinery for generating these dicarboxylic acids, and compare P furiosus genome with the bacteria from the suborder corynebacterineae and the family burkholderiaceae genomes. Move the relevant genetic modules from the P furiosus into the bacteria from the suborder corynebacterineae and the family burkholderiaceae in order to post-process lipids into dicarboxylic acids. This can be combined with genes which produce shorter fatty acids through the appropriate thioesterases.
Example 10
[0232] Hydroxy-acids. For generating omega-hydroxylated fatty acids, incorporate Vicia sativa P450-dependent fatty acid omega hydroxylase into the bacteria from the suborder corynebacterineae and the family burkholderiaceae cell line. This enzyme hydroxylates myristic acid (C14), lauric acid (C12), palmitic acid (C16), but not oleic acid (C18).
Example 11
[0233] Hydroxy-acids. For generating in-chain hydroxylated fatty acids, incorporate CYP81B1 (H tuberosus) or CYP709C1 (unknown) fatty acid hydroxylases into the bacteria from the suborder corynebacterineae and the family burkholderiaceae cell line. The CYP81B1 enzyme omega-1 and omega-5 mono-hydroxylates capric (C10:0), lauric (C12:0), and myristic (C14:0) (Pompon 1996). The CYP709C1 gene hydroxylates the omega-1 and omega-2 positions independent of chain length (Kandel 2005).
[0234] Specific preferred embodiments of the present invention have been described here in sufficient detail to enable those skilled in the art to practice the full scope of invention. However it is to be understood that many possible variations of the present invention, which have not been specifically described, still fall within the scope of the present invention and the appended claims. Hence these descriptions given herein are added only by way of example and are not intended to limit, in any way, the scope of this invention. More generally, those skilled in the art will readily appreciate that all parameters, dimensions, materials, and configurations described herein are meant to be exemplary and that the actual parameters, dimensions, materials, and/or configurations will depend upon the specific application or applications for which the teachings of the present invention is/are used. Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific embodiments of the invention described herein. It is, therefore, to be understood that the foregoing embodiments are presented by way of example only and that, within the scope of the appended claims and equivalents thereto, the invention may be practiced otherwise than as specifically described and claimed. The present invention is directed to each individual feature, system, article, material, kit, and/or method described herein. In addition, any combination of two or more such features, systems, articles, materials, kits, and/or methods, if such features, systems, articles, materials, kits, and/or methods are not mutually inconsistent, is included within the scope of the present invention.
[0235] The indefinite articles "a" and "an," as used herein in the specification and in the claims, unless clearly indicated to the contrary, should be understood to mean "at least one."
[0236] The phrase "and/or," as used herein in the specification and in the claims, should be understood to mean "either or both" of the elements so conjoined, i.e., elements that are conjunctively present in some cases and disjunctively present in other cases. Other elements may optionally be present other than the elements specifically identified by the "and/or" clause, whether related or unrelated to those elements specifically identified unless clearly indicated to the contrary. Thus, as a non-limiting example, a reference to "A and/or B," when used in conjunction with open-ended language such as "comprising" can refer, in one embodiment, to A without B (optionally including elements other than B); in another embodiment, to B without A (optionally including elements other than A); in yet another embodiment, to both A and B (optionally including other elements); etc.
[0237] As used herein in the specification and in the claims, "or" should be understood to have the same meaning as "and/or" as defined above. For example, when separating items in a list, "or" or "and/or" shall be interpreted as being inclusive, i.e., the inclusion of at least one, but also including more than one, of a number or list of elements, and, optionally, additional unlisted items. Only terms clearly indicated to the contrary, such as "only one of" or "exactly one of," or, when used in the claims, "consisting of," will refer to the inclusion of exactly one element of a number or list of elements. In general, the term "or" as used herein shall only be interpreted as indicating exclusive alternatives (i.e. "one or the other but not both") when preceded by terms of exclusivity, such as "either," "one of," "only one of," or "exactly one of." "Consisting essentially of," when used in the claims, shall have its ordinary meaning as used in the field of patent law.
[0238] In the claims, as well as in the specification above, all transitional phrases such as "comprising," "including," "carrying," "having," "containing," "involving," "holding," and the like are to be understood to be open-ended, i.e., to mean including but not limited to. Only the transitional phrases "consisting of" and "consisting essentially of" shall be closed or semi-closed transitional phrases, respectively.
Sequence CWU
1
1
41110560DNAArtificial SequenceDescription of Artificial Sequence Synthetic
polynucleotide 1tcgcgcgttt cggtgatgac ggtgaaaacc tctgacacat
gcagctcccg gagacggtca 60cagcttgtct gtaagcggat gccgggagca gacaagcccg
agcgcgcaaa gccactactg 120ccacttttgg agactgtgta cgtcgagggc ctctgccagt
gtcgaacaga cattcgccta 180cggccctcgt ctgttcgggc tcagggcgcg tcagcgggtg
ttggcgggtg tcggggctgg 240cttaactatg cggcatcaga gcagattgta ctgagagtgc
accatatgcg gtgtgaaata 300agtcccgcgc agtcgcccac aaccgcccac agccccgacc
gaattgatac gccgtagtct 360cgtctaacat gactctcacg tggtatacgc cacactttat
ccgcacagat gcgtaaggag 420aaaataccgc atcaggcgcc attcgccatt caggctgcgc
aactgttggg aagggcgatc 480ggtgcgggcc tcttcgctat ggcgtgtcta cgcattcctc
ttttatggcg tagtccgcgg 540taagcggtaa gtccgacgcg ttgacaaccc ttcccgctag
ccacgcccgg agaagcgata 600tacgccagct ggcgaaaggg ggatgtgctg caaggcgatt
aagttgggta acgccagggt 660tttcccagtc acgacgttgt aaaacgacgg ccagtgccaa
atgcggtcga ccgctttccc 720cctacacgac gttccgctaa ttcaacccat tgcggtccca
aaagggtcag tgctgcaaca 780ttttgctgcc ggtcacggtt gcttgcatgc ctgcaggtcg
acgggcccgg gatccgatgc 840tcttccgcta agatctgccg cggccgcgtc ctcagaagaa
ctcgtcaaga aggcgataga 900cgaacgtacg gacgtccagc tgcccgggcc ctaggctacg
agaaggcgat tctagacggc 960gccggcgcag gagtcttctt gagcagttct tccgctatct
aggcgatgcg ctgcgaatcg 1020ggagcggcga taccgtaaag cacgaggaag cggtcagccc
attcgccgcc aagctcttca 1080gcaatatcac gggtagccaa tccgctacgc gacgcttagc
cctcgccgct atggcatttc 1140gtgctccttc gccagtcggg taagcggcgg ttcgagaagt
cgttatagtg cccatcggtt 1200cgctatgtcc tgatagcggt ccgccacacc cagccggcca
cagtcgatga atccagaaaa 1260gcggccattt tccaccatga tattcggcaa gcaggcatcg
gcgatacagg actatcgcca 1320ggcggtgtgg gtcggccggt gtcagctact taggtctttt
cgccggtaaa aggtggtact 1380ataagccgtt cgtccgtagc ccatgggtca cgacgagatc
ctcgccgtcg ggcatgcgcg 1440ccttgagcct ggcgaacagt tcggctggcg cgagcccctg
atgctcttcg tccagatcat 1500ggtacccagt gctgctctag gagcggcagc ccgtacgcgc
ggaactcgga ccgcttgtca 1560agccgaccgc gctcggggac tacgagaagc aggtctagta
cctgatcgac aagaccggct 1620tccatccgag tacgtgctcg ctcgatgcga tgtttcgctt
ggtggtcgaa tgggcaggta 1680gccggatcaa gcgtatgcag ggactagctg ttctggccga
aggtaggctc atgcacgagc 1740gagctacgct acaaagcgaa ccaccagctt acccgtccat
cggcctagtt cgcatacgtc 1800ccgccgcatt gcatcagcca tgatggatac tttctcggca
ggagcaaggt gggatgacag 1860gagatcctgc cccggcactt cgcccaatag cagccagtcc
ggcggcgtaa cgtagtcggt 1920actacctatg aaagagccgt cctcgttcca ccctactgtc
ctctaggacg gggccgtgaa 1980gcgggttatc gtcggtcagg cttcccgctt cagtgacaac
gtcgagcaca gctgcgcaag 2040gaacgcccgt cgtggccagc cacgatagcc gcgctgcctc
gtcctgcagt tcattcaggg 2100gaagggcgaa gtcactgttg cagctcgtgt cgacgcgttc
cttgcgggca gcaccggtcg 2160gtgctatcgg cgcgacggag caggacgtca agtaagtccc
caccggacag gtcggtcttg 2220acaaaaagaa ccgggcgccc ctgcgctgac agccggaaca
cggcggcatc agagcagccg 2280attgtctgtt gtgcccagtc gtggcctgtc cagccagaac
tgtttttctt ggcccgcggg 2340gacgcgactg tcggccttgt gccgccgtag tctcgtcggc
taacagacaa cacgggtcag 2400atagccgaat agcctctcca cccaagcggc cggagaacct
gcgtgcaatc catcttgttc 2460aatcatgata tcccttaatt aaccgttaac actagttcag
tatcggctta tcggagaggt 2520gggttcgccg gcctcttgga cgcacgttag gtagaacaag
ttagtactat agggaattaa 2580ttggcaattg tgatcaagtc tccatctcgc cgtgtatgcg
ggcctgacgg atcaacgttc 2640ccaccgagcc agtcgagatg ttcatctggt cggcgatctg
ccggtacttc aaaccttgtt 2700aggtagagcg gcacatacgc ccggactgcc tagttgcaag
ggtggctcgg tcagctctac 2760aagtagacca gccgctagac ggccatgaag tttggaacaa
tgcgcagttc cacagccttc 2820ttgcggcgtt cctgcgcacg agcgatgtag tcgcctcggt
cttcggcgac gagccgtttg 2880atggtgcttt tcgagacgcc acgcgtcaag gtgtcggaag
aacgccgcaa ggacgcgtgc 2940tcgctacatc agcggagcca gaagccgctg ctcggcaaac
taccacgaaa agctctgcgg 3000gaacttgtca gccaactcct gcgcggtctg cgtgcgacgc
atcacgcgtt ctgcagcacc 3060catcagtccg tcccctctgc tgctgcgaac agtgccgatc
cttgaacagt cggttgagga 3120cgcgccagac gcacgctgcg tagtgcgcaa gacgtcgtgg
gtagtcaggc aggggagacg 3180acgacgcttg tcacggctag gatcgacctt cttgagcttc
ggccgcggcg cggtggcgtt 3240cttccgtacc gcttccgttt ttgcgctgct gctcactttg
ccgcggcgtg cctggatttt 3300ctagctggaa gaactcgaag ccggcgccgc gccaccgcaa
gaaggcatgg cgaaggcaaa 3360aacgcgacga cgagtgaaac ggcgccgcac ggacctaaaa
cgagaactcg gcggcggtga 3420aggtgcggtg ggtccagtgg gcgactgatt tgccgatctg
ctcggcctcg gcccgactca 3480tggggccgat cccgtcgttg gctcttgagc cgccgccact
tccacgccac ccaggtcacc 3540cgctgactaa acggctagac gagccggagc cgggctgagt
accccggcta gggcagcaac 3600gcgtcgaggg tgaagttggt cagggcggtg aagtcggtga
ccatctgccg ccacacagtg 3660atcgacgggt agttctgttt ccggatctcg cggtaggccc
cgcagctccc acttcaacca 3720gtcccgccac ttcagccact ggtagacggc ggtgtgtcac
tagctgccca tcaagacaaa 3780ggcctagagc gccatccggg attcccgggt gcggtcgaac
agttcgacgt tccggcccgt 3840ttcggtcctg acctgtgtct tgcggccgta gtccggtggg
gcggggaaac ggtcaccgag 3900taagggccca cgccagcttg tcaagctgca aggccgggca
aagccaggac tggacacaga 3960acgccggcat caggccaccc cgcccctttg ccagtggctc
cgcttttgcg aggcctttga 4020gcgagtacgg atccgaggga ccccagaccg tcgtccagtg
cgggtggatc gggttctggg 4080tgagctgctg cgcgtagccc gcgaaaacgc tccggaaact
cgctcatgcc taggctccct 4140ggggtctggc agcaggtcac gcccacctag cccaagaccc
actcgacgac gcgcatcggg 4200tgatcggcgc cgaccaccga ggcgatcagc ccctggttca
cccggtcgta gagccgcagc 4260gggccctgtc gggctgcctg gagggtgtag accgggcttt
actagccgcg gctggtggct 4320ccgctagtcg gggaccaagt gggccagcat ctcggcgtcg
cccgggacag cccgacggac 4380ctcccacatc tggcccgaaa cgagcagcca ccacaggtgc
gcgtgctcgg tcgcgggatt 4440gatcgtcatc acggtcggat cgggcagatc cgcgttacgt
gcggcccact gcgcctggtc 4500gctcgtcggt ggtgtccacg cgcacgagcc agcgccctaa
ctagcagtag tgccagccta 4560gcccgtctag gcgcaatgca cgccgggtga cgcggaccag
gtcgtccacg tcgagcacca 4620agcccaacct gatcgacggg gtgcgggccg caatgtagcg
gcgggtgagc gcctccgcgc 4680gcggctgcgg ccactgcccg cagcaggtgc agctcgtggt
tcgggttgga ctagctgccc 4740cacgcccggc gttacatcgc cgcccactcg cggaggcgcg
cgccgacgcc ggtgacgggc 4800tcccggacgt agtcatccgt cgcgtgcggg tatttgaacc
gccagcggtc caaccaggcg 4860tcaacagcag cggtcatgac cgccaagcta gggccggatc
agggcctgca tcagtaggca 4920gcgcacgccc ataaacttgg cggtcgccag gttggtccgc
agttgtcgtc gccagtactg 4980gcggttcgat cccggcctag tgtaccgatc gggggaggcg
cgccgcaaat tatttaagag 5040tctcgctagc aaaccatgtc aggtgttgcg gtgggttccg
ggtaaacctc cacccgaatt 5100acatggctag ccccctccgc gcggcgttta ataaattctc
agagcgatcg tttggtacag 5160tccacaacgc cacccaaggc ccatttggag gtgggcttaa
atttaagagt ctcgctagct 5220aagccctatc tgatgctgcg cggggggtcc ttcgcactga
atctcaaagg tggccggctg 5280aatttcgtcg cgcgaaaacc taaattctca gagcgatcga
ttcgggatag actacgacgc 5340gccccccagg aagcgtgact tagagtttcc accggccgac
ttaaagcagc gcgcttttgg 5400tccctggaca gttctggaat tcagcaagag gtgtgtctga
acttcggtgt ttttttgggg 5460ggtgactcca gcggggtggg cacaacgcga acagagacct
agggacctgt caagacctta 5520agtcgttctc cacacagact tgaagccaca aaaaaacccc
ccactgaggt cgccccaccc 5580gtgttgcgct tgtctctgga tgtgtgtacg acggcgggag
gtaagtcggg tacggctcgg 5640actgcggtag agcaaccgtc gaatcgattt cgagcagagc
gagcagagca agatattcca 5700acacacatgc tgccgccctc cattcagccc atgccgagcc
tgacgccatc tcgttggcag 5760cttagctaaa gctcgtctcg ctcgtctcgt tctataaggt
aaactccggg gttcctcggc 5820ggcctccccc gtctgtttgc tcaaccgagg gagacctggc
ggtcccgcgt ttccggacgc 5880gcgggaccgc ctaccgctcg tttgaggccc caaggagccg
ccggaggggg cagacaaacg 5940agttggctcc ctctggaccg ccagggcgca aaggcctgcg
cgccctggcg gatggcgagc 6000agagcggaag agcatctaga tgcattcgcg aggtaccgag
ctcgaattcg taatcatggt 6060catagctgtt tcctgtgtga aattgttatc cgctcacaat
tctcgccttc tcgtagatct 6120acgtaagcgc tccatggctc gagcttaagc attagtacca
gtatcgacaa aggacacact 6180ttaacaatag gcgagtgtta tccacacaac atacgagccg
gaagcataaa gtgtaaagcc 6240tggggtgcct aatgagtgag ctaactcaca ttaattgcgt
tgcgctcact gcccgctttc 6300aggtgtgttg tatgctcggc cttcgtattt cacatttcgg
accccacgga ttactcactc 6360gattgagtgt aattaacgca acgcgagtga cgggcgaaag
cagtcgggaa acctgtcgtg 6420ccagctgcat taatgaatcg gccaacgcgc ggggagaggc
ggtttgcgta ttgggcgctc 6480ttccgcttcc tcgctcactg gtcagccctt tggacagcac
ggtcgacgta attacttagc 6540cggttgcgcg cccctctccg ccaaacgcat aacccgcgag
aaggcgaagg agcgagtgac 6600actcgctgcg ctcggtcgtt cggctgcggc gagcggtatc
agctcactca aaggcggtaa 6660tacggttatc cacagaatca ggggataacg caggaaagaa
tgagcgacgc gagccagcaa 6720gccgacgccg ctcgccatag tcgagtgagt ttccgccatt
atgccaatag gtgtcttagt 6780cccctattgc gtcctttctt catgtgagca aaaggccagc
aaaaggccag gaaccgtaaa 6840aaggccgcgt tgctggcgtt tttccatagg ctccgccccc
ctgacgagca tcacaaaaat 6900gtacactcgt tttccggtcg ttttccggtc cttggcattt
ttccggcgca acgaccgcaa 6960aaaggtatcc gaggcggggg gactgctcgt agtgttttta
cgacgctcaa gtcagaggtg 7020gcgaaacccg acaggactat aaagatacca ggcgtttccc
cctggaagct ccctcgtgcg 7080ctctcctgtt ccgaccctgc gctgcgagtt cagtctccac
cgctttgggc tgtcctgata 7140tttctatggt ccgcaaaggg ggaccttcga gggagcacgc
gagaggacaa ggctgggacg 7200cgcttaccgg atacctgtcc gcctttctcc cttcgggaag
cgtggcgctt tctcatagct 7260cacgctgtag gtatctcagt tcggtgtagg tcgttcgctc
gcgaatggcc tatggacagg 7320cggaaagagg gaagcccttc gcaccgcgaa agagtatcga
gtgcgacatc catagagtca 7380agccacatcc agcaagcgag caagctgggc tgtgtgcacg
aaccccccgt tcagcccgac 7440cgctgcgcct tatccggtaa ctatcgtctt gagtccaacc
cggtaagaca cgacttatcg 7500gttcgacccg acacacgtgc ttggggggca agtcgggctg
gcgacgcgga ataggccatt 7560gatagcagaa ctcaggttgg gccattctgt gctgaatagc
ccactggcag cagccactgg 7620taacaggatt agcagagcga ggtatgtagg cggtgctaca
gagttcttga agtggtggcc 7680taactacggc tacactagaa ggtgaccgtc gtcggtgacc
attgtcctaa tcgtctcgct 7740ccatacatcc gccacgatgt ctcaagaact tcaccaccgg
attgatgccg atgtgatctt 7800ggacagtatt tggtatctgc gctctgctga agccagttac
cttcggaaaa agagttggta 7860gctcttgatc cggcaaacaa accaccgctg gtagcggtgg
cctgtcataa accatagacg 7920cgagacgact tcggtcaatg gaagcctttt tctcaaccat
cgagaactag gccgtttgtt 7980tggtggcgac catcgccacc tttttttgtt tgcaagcagc
agattacgcg cagaaaaaaa 8040ggatctcaag aagatccttt gatcttttct acggggtctg
acgctcagtg gaacgaaaac 8100aaaaaaacaa acgttcgtcg tctaatgcgc gtcttttttt
cctagagttc ttctaggaaa 8160ctagaaaaga tgccccagac tgcgagtcac cttgcttttg
tcacgttaag ggattttggt 8220catgagatta tcaaaaagga tcttcaccta gatcctttta
aattaaaaat gaagttttaa 8280atcaatctaa agtatatatg agtgcaattc cctaaaacca
gtactctaat agtttttcct 8340agaagtggat ctaggaaaat ttaattttta cttcaaaatt
tagttagatt tcatatatac 8400agtaaacttg gtctgacagt taccaatgct taatcagtga
ggcacctatc tcagcgatct 8460gtctatttcg ttcatccata gttgcctgac tccccgtcgt
tcatttgaac cagactgtca 8520atggttacga attagtcact ccgtggatag agtcgctaga
cagataaagc aagtaggtat 8580caacggactg aggggcagca gtagataact acgatacggg
agggcttacc atctggcccc 8640agtgctgcaa tgataccgcg agacccacgc tcaccggctc
cagatttatc agcaataaac 8700catctattga tgctatgccc tcccgaatgg tagaccgggg
tcacgacgtt actatggcgc 8760tctgggtgcg agtggccgag gtctaaatag tcgttatttg
cagccagccg gaagggccga 8820gcgcagaagt ggtcctgcaa ctttatccgc ctccatccag
tctattaatt gttgccggga 8880agctagagta agtagttcgc gtcggtcggc cttcccggct
cgcgtcttca ccaggacgtt 8940gaaataggcg gaggtaggtc agataattaa caacggccct
tcgatctcat tcatcaagcg 9000cagttaatag tttgcgcaac gttgttgcca ttgctacagg
catcgtggtg tcacgctcgt 9060cgtttggtat ggcttcattc agctccggtt cccaacgatc
gtcaattatc aaacgcgttg 9120caacaacggt aacgatgtcc gtagcaccac agtgcgagca
gcaaaccata ccgaagtaag 9180tcgaggccaa gggttgctag aaggcgagtt acatgatccc
ccatgttgtg caaaaaagcg 9240gttagctcct tcggtcctcc gatcgttgtc agaagtaagt
tggccgcagt gttatcactc 9300ttccgctcaa tgtactaggg ggtacaacac gttttttcgc
caatcgagga agccaggagg 9360ctagcaacag tcttcattca accggcgtca caatagtgag
atggttatgg cagcactgca 9420taattctctt actgtcatgc catccgtaag atgcttttct
gtgactggtg agtactcaac 9480caagtcattc tgagaatagt taccaatacc gtcgtgacgt
attaagagaa tgacagtacg 9540gtaggcattc tacgaaaaga cactgaccac tcatgagttg
gttcagtaag actcttatca 9600gtatgcggcg accgagttgc tcttgcccgg cgtcaatacg
ggataatacc gcgccacata 9660gcagaacttt aaaagtgctc atcattggaa aacgttcttc
catacgccgc tggctcaacg 9720agaacgggcc gcagttatgc cctattatgg cgcggtgtat
cgtcttgaaa ttttcacgag 9780tagtaacctt ttgcaagaag ggggcgaaaa ctctcaagga
tcttaccgct gttgagatcc 9840agttcgatgt aacccactcg tgcacccaac tgatcttcag
catcttttac tttcaccagc 9900ccccgctttt gagagttcct agaatggcga caactctagg
tcaagctaca ttgggtgagc 9960acgtgggttg actagaagtc gtagaaaatg aaagtggtcg
gtttctgggt gagcaaaaac 10020aggaaggcaa aatgccgcaa aaaagggaat aagggcgaca
cggaaatgtt gaatactcat 10080actcttcctt tttcaatatt caaagaccca ctcgtttttg
tccttccgtt ttacggcgtt 10140ttttccctta ttcccgctgt gcctttacaa cttatgagta
tgagaaggaa aaagttataa 10200attgaagcat ttatcagggt tattgtctca tgagcggata
catatttgaa tgtatttaga 10260aaaataaaca aataggggtt ccgcgcacat ttccccgaaa
taacttcgta aatagtccca 10320ataacagagt actcgcctat gtataaactt acataaatct
ttttatttgt ttatccccaa 10380ggcgcgtgta aaggggcttt agtgccacct gacgtctaag
aaaccattat tatcatgaca 10440ttaacctata aaaataggcg tatcacgagg ccctttcgtc
tcacggtgga ctgcagattc 10500tttggtaata atagtactgt aattggatat ttttatccgc
atagtgctcc gggaaagcag 10560210288DNAArtificial SequenceDescription of
Artificial Sequence Synthetic polynucleotide 2ggggagccgc gccgaaggcg
tgggggaacc ccgcaggggt gcccttcttt gggcaccaaa 60gaactagata tagggcgaaa
tgcgaaagac ttaaaaatca cccctcggcg cggcttccgc 120acccccttgg ggcgtcccca
cgggaagaaa cccgtggttt cttgatctat atcccgcttt 180acgctttctg aatttttagt
acaacttaaa aaaggggggt acgcaacagc tcattgcggc 240accccccgca atagctcatt
gcgtaggtta aagaaaatct gtaattgact gccactttta 300tgttgaattt tttcccccca
tgcgttgtcg agtaacgccg tggggggcgt tatcgagtaa 360cgcatccaat ttcttttaga
cattaactga cggtgaaaat cgcaacgcat aattgttgtc 420gcgctgccga aaagttgcag
ctgattgcgc atggtgccgc aaccgtgcgg caccctaccg 480catggagata agcatggcca
gcgttgcgta ttaacaacag cgcgacggct tttcaacgtc 540gactaacgcg taccacggcg
ttggcacgcc gtgggatggc gtacctctat tcgtaccggt 600cgcagtccag agaaatcggc
attcaagcca agaacaagcc cggtcactgg gtgcaaacgg 660aacgcaaagc gcatgaggcg
tgggccgggc ttattgcgag gcgtcaggtc tctttagccg 720taagttcggt tcttgttcgg
gccagtgacc cacgtttgcc ttgcgtttcg cgtactccgc 780acccggcccg aataacgctc
gaaacccacg gcggcaatgc tgctgcatca cctcgtggcg 840cagatgggcc accagaacgc
cgtggtggtc agccagaaga cactttccaa gctcatcgga 900ctttgggtgc cgccgttacg
acgacgtagt ggagcaccgc gtctacccgg tggtcttgcg 960gcaccaccag tcggtcttct
gtgaaaggtt cgagtagcct cgttctttgc ggacggtcca 1020atacgcagtc aaggacttgg
tggccgagcg ctggatctcc gtcgtgaagc tcaacggccc 1080cggcaccgtg tcggcctacg
gcaagaaacg cctgccaggt tatgcgtcag ttcctgaacc 1140accggctcgc gacctagagg
cagcacttcg agttgccggg gccgtggcac agccggatgc 1200tggtcaatga ccgcgtggcg
tggggccagc cccgcgacca gttgcgcctg tcggtgttca 1260gtgccgccgt ggtggttgat
cacgacgacc aggacgaatc accagttact ggcgcaccgc 1320accccggtcg gggcgctggt
caacgcggac agccacaagt cacggcggca ccaccaacta 1380gtgctgctgg tcctgcttag
gctgttgggg catggcgacc tgcgccgcat cccgaccctg 1440tatccgggcg agcagcaact
accgaccggc cccggcgagg agccgcccag ccagcccggc 1500cgacaacccc gtaccgctgg
acgcggcgta gggctgggac ataggcccgc tcgtcgttga 1560tggctggccg gggccgctcc
tcggcgggtc ggtcgggccg attccgggca tggaaccaga 1620cctgccagcc ttgaccgaaa
cggaggaatg ggaacggcgc gggcagcagc gcctgccgat 1680gcccgatgag ccgtgttttc
taaggcccgt accttggtct ggacggtcgg aactggcttt 1740gcctccttac ccttgccgcg
cccgtcgtcg cggacggcta cgggctactc ggcacaaaag 1800tggacgatgg cgagccgttg
gagccgccga cacgggtcac gctgccgcgc cggtagcact 1860tgggttgcgc agcaacccgt
aagtgcgctg ttccagacta acctgctacc gctcggcaac 1920ctcggcggct gtgcccagtg
cgacggcgcg gccatcgtga acccaacgcg tcgttgggca 1980ttcacgcgac aaggtctgat
tcggctgtag ccgcctcgcc gccctatacc ttgtctgcct 2040ccccgcgttg cgtcgcggtg
catggagccg ggccacctcg acctgaatgg aagccggcgg 2100agccgacatc ggcggagcgg
cgggatatgg aacagacgga ggggcgcaac gcagcgccac 2160gtacctcggc ccggtggagc
tggacttacc ttcggccgcc cacctcgcta acggattcac 2220cgtttttatc aggctctggg
aggcagaata aatgatcata tcgtcaatta ttacctccac 2280ggggagagcc tgagcaaact
gtggagcgat tgcctaagtg gcaaaaatag tccgagaccc 2340tccgtcttat ttactagtat
agcagttaat aatggaggtg cccctctcgg actcgtttga 2400ggcctcaggc atttgagaag
cacacggtca cactgcttcc ggtagtcaat aaaccggtaa 2460accagcaata gacataagcg
gctatttaac gaccctgccc ccggagtccg taaactcttc 2520gtgtgccagt gtgacgaagg
ccatcagtta tttggccatt tggtcgttat ctgtattcgc 2580cgataaattg ctgggacggg
tgaaccgacg accgggtcga atttgctttc gaatttctgc 2640cattcatccg cttattatca
cttattcagg cgtagcacca ggcgtttaag ggcaccaata 2700acttggctgc tggcccagct
taaacgaaag cttaaagacg gtaagtaggc gaataatagt 2760gaataagtcc gcatcgtggt
ccgcaaattc ccgtggttat actgccttaa aaaaattacg 2820ccccgccctg ccactcatcg
cagtcggcct attggttaaa aaatgagctg atttaacaaa 2880aatttaacgc gaattttaac
tgacggaatt tttttaatgc ggggcgggac ggtgagtagc 2940gtcagccgga taaccaattt
tttactcgac taaattgttt ttaaattgcg cttaaaattg 3000aaaatattaa cgcttacaat
ttccattcgc cattcaggct gcgcaactgt tgggaagggc 3060gatcggtgcg ggcctcttcg
ctattacgcc agctggcgaa ttttataatt gcgaatgtta 3120aaggtaagcg gtaagtccga
cgcgttgaca acccttcccg ctagccacgc ccggagaagc 3180gataatgcgg tcgaccgctt
agggggatgt gctgcaaggc gattaagttg ggtaacgcca 3240gggttttccc agtcacgacg
ttgtaaaacg acggccagtg agcgcgcgta atacgactca 3300tccccctaca cgacgttccg
ctaattcaac ccattgcggt cccaaaaggg tcagtgctgc 3360aacattttgc tgccggtcac
tcgcgcgcat tatgctgagt ctatagggcg aattggagct 3420ccaccgcggt ggcggccgct
ctagaactag tggatccccc gggctgcagg aattcgatat 3480caagcttatc gataccgtcg
gatatcccgc ttaacctcga ggtggcgcca ccgccggcga 3540gatcttgatc acctaggggg
cccgacgtcc ttaagctata gttcgaatag ctatggcagc 3600acctcgaggg ggggcccggt
acccagcttt tgttcccttt agtgagggtt aattgcgcgc 3660ttggcgtaat catggtcata
gctgtttcct gtgtgaaatt tggagctccc ccccgggcca 3720tgggtcgaaa acaagggaaa
tcactcccaa ttaacgcgcg aaccgcatta gtaccagtat 3780cgacaaagga cacactttaa
gttatccgct cacaattcca cacaacatac gagccggaag 3840cataaagtgt aaagcctggg
gtgcctaatg agtgagctaa ctcacattaa ttgcgttgcg 3900caataggcga gtgttaaggt
gtgttgtatg ctcggccttc gtatttcaca tttcggaccc 3960cacggattac tcactcgatt
gagtgtaatt aacgcaacgc ctcactgccc gctttccagt 4020cgggaaacct gtcgtgccag
ctgcattaat gaatcggcca acgcgcgggg agaggcggtt 4080tgcgtattgg gcgcatgcat
gagtgacggg cgaaaggtca gccctttgga cagcacggtc 4140gacgtaatta cttagccggt
tgcgcgcccc tctccgccaa acgcataacc cgcgtacgta 4200aaaaactgtt gtaattcatt
aagcattctg ccgacatgga agccatcaca aacggcatga 4260tgaacctgaa tcgccagcgg
catcagcacc ttgtcgcctt tttttgacaa cattaagtaa 4320ttcgtaagac ggctgtacct
tcggtagtgt ttgccgtact acttggactt agcggtcgcc 4380gtagtcgtgg aacagcggaa
gcgtataata tttgcccatg ggggtgggcg aagaactcca 4440gcatgagatc cccgcgctgg
aggatcatcc agccggcgtc ccggaaaacg attccgaagc 4500cgcatattat aaacgggtac
ccccacccgc ttcttgaggt cgtactctag gggcgcgacc 4560tcctagtagg tcggccgcag
ggccttttgc taaggcttcg ccaacctttc atagaaggcg 4620gcggtggaat cgaaatctcg
tgatggcagg ttgggcgtcg cttggtcggt catttcgaac 4680cccagagtcc cgctcagaag
ggttggaaag tatcttccgc cgccacctta gctttagagc 4740actaccgtcc aacccgcagc
gaaccagcca gtaaagcttg gggtctcagg gcgagtcttc 4800aactcgtcaa gaaggcgata
gaaggcgatg cgctgcgaat cgggagcggc gataccgtaa 4860agcacgagga agcggtcagc
ccattcgccg ccaagctctt ttgagcagtt cttccgctat 4920cttccgctac gcgacgctta
gccctcgccg ctatggcatt tcgtgctcct tcgccagtcg 4980ggtaagcggc ggttcgagaa
cagcaatatc acgggtagcc aacgctatgt cctgatagcg 5040gtccgccaca cccagccggc
cacagtcgat gaatccagaa aagcggccat tttccaccat 5100gtcgttatag tgcccatcgg
ttgcgataca ggactatcgc caggcggtgt gggtcggccg 5160gtgtcagcta cttaggtctt
ttcgccggta aaaggtggta gatattcggc aagcaggcat 5220cgccatgggt cacgacgaga
tcctcgccgt cgggcatgcg cgccttgagc ctggcgaaca 5280gttcggctgg cgcgagcccc
ctataagccg ttcgtccgta gcggtaccca gtgctgctct 5340aggagcggca gcccgtacgc
gcggaactcg gaccgcttgt caagccgacc gcgctcgggg 5400tgatgctctt cgtccagatc
atcctgatcg acaagaccgg cttccatccg agtacgtgct 5460cgctcgatgc gatgtttcgc
ttggtggtcg aatgggcagg actacgagaa gcaggtctag 5520taggactagc tgttctggcc
gaaggtaggc tcatgcacga gcgagctacg ctacaaagcg 5580aaccaccagc ttacccgtcc
tagccggatc aagcgtatgc agccgccgca ttgcatcagc 5640catgatggat actttctcgg
caggagcaag gtgagatgac aggagatcct gccccggcac 5700atcggcctag ttcgcatacg
tcggcggcgt aacgtagtcg gtactaccta tgaaagagcc 5760gtcctcgttc cactctactg
tcctctagga cggggccgtg ttcgcccaat agcagccagt 5820cccttcccgc ttcagtgaca
acgtcgagca cagctgcgca aggaacgccc gtcgtggcca 5880gccacgatag ccgcgctgcc
aagcgggtta tcgtcggtca gggaagggcg aagtcactgt 5940tgcagctcgt gtcgacgcgt
tccttgcggg cagcaccggt cggtgctatc ggcgcgacgg 6000tcgtcctgca gttcattcag
ggcaccggac aggtcggtct tgacaaaaag aaccgggcgc 6060ccctgcgctg acagccggaa
cacggcggca tcagagcagc agcaggacgt caagtaagtc 6120ccgtggcctg tccagccaga
actgtttttc ttggcccgcg gggacgcgac tgtcggcctt 6180gtgccgccgt agtctcgtcg
cgattgtctg ttgtgcccag tcatagccga atagcctctc 6240cacccaagcg gccggagaac
ctgcgtgcaa tccatcttgt tcaatcatgc gaaacgatcc 6300gctaacagac aacacgggtc
agtatcggct tatcggagag gtgggttcgc cggcctcttg 6360gacgcacgtt aggtagaaca
agttagtacg ctttgctagg tcatcctgtc tcttgatcag 6420atcttgatcc cctgcgccat
cagatccttg gcggcaagaa agccatccag tttactttgc 6480agggcttccc aaccttacca
agtaggacag agaactagtc tagaactagg ggacgcggta 6540gtctaggaac cgccgttctt
tcggtaggtc aaatgaaacg tcccgaaggg ttggaatggt 6600gagggcgccc cagctggcaa
ttccggttcg cttgctgtcc ataaaaccgc ccagtctagc 6660tatcgccatg taagcccact
gcaagctacc tgctttctct ctcccgcggg gtcgaccgtt 6720aaggccaagc gaacgacagg
tattttggcg ggtcagatcg atagcggtac attcgggtga 6780cgttcgatgg acgaaagaga
ttgcgcttgc gttttccctt gtccagatag cccagtagct 6840gacattcatc ccaggtggca
cttttcgggg aaatgtgcgc gcccgcgttc ctgctggcgc 6900aacgcgaacg caaaagggaa
caggtctatc gggtcatcga ctgtaagtag ggtccaccgt 6960gaaaagcccc tttacacgcg
cgggcgcaag gacgaccgcg tgggcctgtt tctggcgctg 7020gacttcccgc tgttccgtca
gcagcttttc gcccacggcc ttgatgatcg cggcggcctt 7080ggcctgcata tcccgattca
acccggacaa agaccgcgac ctgaagggcg acaaggcagt 7140cgtcgaaaag cgggtgccgg
aactactagc gccgccggaa ccggacgtat agggctaagt 7200acggccccag ggcgtccaga
acgggcttca ggcgctcccg aaggtctcgg gccgtctctt 7260gggcttgatc ggccttcttg
cgcatctcac gcgctcctgc tgccggggtc ccgcaggtct 7320tgcccgaagt ccgcgagggc
ttccagagcc cggcagagaa cccgaactag ccggaagaac 7380gcgtagagtg cgcgaggacg
ggcggcctgt agggcaggct catacccctg ccgaaccgct 7440tttgtcagcc ggtcggccac
ggcttccggc gtctcaacgc gctttgagat tcccagcttt 7500ccgccggaca tcccgtccga
gtatggggac ggcttggcga aaacagtcgg ccagccggtg 7560ccgaaggccg cagagttgcg
cgaaactcta agggtcgaaa tcggccaatc cctgcggtgc 7620ataggcgcgt ggctcgaccg
cttgcgggct gatggtgacg tggcccactg gtggccgctc 7680cagggcctcg tagaacgcct
agccggttag ggacgccacg tatccgcgca ccgagctggc 7740gaacgcccga ctaccactgc
accgggtgac caccggcgag gtcccggagc atcttgcgga 7800gaatgcgcgt gtgacgtgcc
ttgctgccct cgatgccccg ttgcagccct agatcggcca 7860cagcggccgc aaacgtggtc
tggtcgcggg tcatctgcgc cttacgcgca cactgcacgg 7920aacgacggga gctacggggc
aacgtcggga tctagccggt gtcgccggcg tttgcaccag 7980accagcgccc agtagacgcg
tttgttgccg atgaactcct tggccgacag cctgccgtcc 8040tgcgtcagcg gcaccacgaa
cgcggtcatg tgcgggctgg tttcgtcacg gtggatgctg 8100aaacaacggc tacttgagga
accggctgtc ggacggcagg acgcagtcgc cgtggtgctt 8160gcgccagtac acgcccgacc
aaagcagtgc cacctacgac gccgtcacga tgcgatccgc 8220cccgtacttg tccgccagcc
acttgtgcgc cttctcgaag aacgccgcct gctgttcttg 8280gctggccgac ttccaccatt
cggcagtgct acgctaggcg gggcatgaac aggcggtcgg 8340tgaacacgcg gaagagcttc
ttgcggcgga cgacaagaac cgaccggctg aaggtggtaa 8400ccgggctggc cgtcatgacg
tactcgaccg ccaacacagc gtccttgcgc cgcttctctg 8460gcagcaactc gcgcagtcgg
cccatcgctt catcggtgct ggcccgaccg gcagtactgc 8520atgagctggc ggttgtgtcg
caggaacgcg gcgaagagac cgtcgttgag cgcgtcagcc 8580gggtagcgaa gtagccacga
gctggccgcc cagtgctcgt tctctggcgt cctgctggcg 8640tcagcgttgg gcgtctcgcg
ctcgcggtag gcgtgcttga gactggccgc cacgttgccc 8700cgaccggcgg gtcacgagca
agagaccgca ggacgaccgc agtcgcaacc cgcagagcgc 8760gagcgccatc cgcacgaact
ctgaccggcg gtgcaacggg attttcgcca gcttcttgca 8820tcgcatgatc gcgtatgccg
ccatgcctgc ccctcccttt tggtgtccaa ccggctcgac 8880gggggcagcg caaggcggtg
taaaagcggt cgaagaacgt agcgtactag cgcatacggc 8940ggtacggacg gggagggaaa
accacaggtt ggccgagctg cccccgtcgc gttccgccac 9000cctccggcgg gccactcaat
gcttgagtat actcactaga ctttgcttcg caaagtcgtg 9060accgcctacg gcggctgcgg
cgccctacgg gcttgctctc ggaggccgcc cggtgagtta 9120cgaactcata tgagtgatct
gaaacgaagc gtttcagcac tggcggatgc cgccgacgcc 9180gcgggatgcc cgaacgagag
cgggcttcgc cctgcgcggt cgctgcgctc ccttgccagc 9240ccgtggatat gtggacgatg
gccgcgagcg gccaccggct ggctcgcttc gctcggcccg 9300gcccgaagcg ggacgcgcca
gcgacgcgag ggaacggtcg ggcacctata cacctgctac 9360cggcgctcgc cggtggccga
ccgagcgaag cgagccgggc tggacaaccc tgctggacaa 9420gctgatggac aggctgcgcc
tgcccacgag cttgaccaca gggattgccc accggctacc 9480cagccttcga ccacataccc
acctgttggg acgacctgtt cgactacctg tccgacgcgg 9540acgggtgctc gaactggtgt
ccctaacggg tggccgatgg gtcggaagct ggtgtatggg 9600accggctcca actgcgcggc
ctgcggcctt gccccatcaa tttttttaat tttctctggg 9660gaaaagcctc cggcctgcgg
cctgcgcgct tcgcttgccg tggccgaggt tgacgcgccg 9720gacgccggaa cggggtagtt
aaaaaaatta aaagagaccc cttttcggag gccggacgcc 9780ggacgcgcga agcgaacggc
gttggacacc aagtggaagg cgggtcaagg ctcgcgcagc 9840gaccgcgcag cggcttggcc
ttgacgcgcc tggaacgacc caagcctatg cgagtggggg 9900caacctgtgg ttcaccttcc
gcccagttcc gagcgcgtcg ctggcgcgtc gccgaaccgg 9960aactgcgcgg accttgctgg
gttcggatac gctcaccccc cagtcgaagg cgaagcccgc 10020ccgcctgccc cccgagcctc
acggcggcga gtgcgggggt tccaaggggg cagcgccacc 10080ttgggcaagg ccgaaggccg
gtcagcttcc gcttcgggcg ggcggacggg gggctcggag 10140tgccgccgct cacgccccca
aggttccccc gtcgcggtgg aacccgttcc ggcttccggc 10200cgcagtcgat caacaagccc
cggaggggcc actttttgcc ggaggcgtca gctagttgtt 10260cggggcctcc ccggtgaaaa
acggcctc 10288312758DNAArtificial
SequenceDescription of Artificial Sequence Synthetic polynucleotide
3ggggagccgc gccgaaggcg tgggggaacc ccgcaggggt gcccttcttt gggcaccaaa
60gaactagata tagggcgaaa tgcgaaagac ttaaaaatca cccctcggcg cggcttccgc
120acccccttgg ggcgtcccca cgggaagaaa cccgtggttt cttgatctat atcccgcttt
180acgctttctg aatttttagt acaacttaaa aaaggggggt acgcaacagc tcattgcggc
240accccccgca atagctcatt gcgtaggtta aagaaaatct gtaattgact gccactttta
300tgttgaattt tttcccccca tgcgttgtcg agtaacgccg tggggggcgt tatcgagtaa
360cgcatccaat ttcttttaga cattaactga cggtgaaaat cgcaacgcat aattgttgtc
420gcgctgccga aaagttgcag ctgattgcgc atggtgccgc aaccgtgcgg caccctaccg
480catggagata agcatggcca gcgttgcgta ttaacaacag cgcgacggct tttcaacgtc
540gactaacgcg taccacggcg ttggcacgcc gtgggatggc gtacctctat tcgtaccggt
600cgcagtccag agaaatcggc attcaagcca agaacaagcc cggtcactgg gtgcaaacgg
660aacgcaaagc gcatgaggcg tgggccgggc ttattgcgag gcgtcaggtc tctttagccg
720taagttcggt tcttgttcgg gccagtgacc cacgtttgcc ttgcgtttcg cgtactccgc
780acccggcccg aataacgctc gaaacccacg gcggcaatgc tgctgcatca cctcgtggcg
840cagatgggcc accagaacgc cgtggtggtc agccagaaga cactttccaa gctcatcgga
900ctttgggtgc cgccgttacg acgacgtagt ggagcaccgc gtctacccgg tggtcttgcg
960gcaccaccag tcggtcttct gtgaaaggtt cgagtagcct cgttctttgc ggacggtcca
1020atacgcagtc aaggacttgg tggccgagcg ctggatctcc gtcgtgaagc tcaacggccc
1080cggcaccgtg tcggcctacg gcaagaaacg cctgccaggt tatgcgtcag ttcctgaacc
1140accggctcgc gacctagagg cagcacttcg agttgccggg gccgtggcac agccggatgc
1200tggtcaatga ccgcgtggcg tggggccagc cccgcgacca gttgcgcctg tcggtgttca
1260gtgccgccgt ggtggttgat cacgacgacc aggacgaatc accagttact ggcgcaccgc
1320accccggtcg gggcgctggt caacgcggac agccacaagt cacggcggca ccaccaacta
1380gtgctgctgg tcctgcttag gctgttgggg catggcgacc tgcgccgcat cccgaccctg
1440tatccgggcg agcagcaact accgaccggc cccggcgagg agccgcccag ccagcccggc
1500cgacaacccc gtaccgctgg acgcggcgta gggctgggac ataggcccgc tcgtcgttga
1560tggctggccg gggccgctcc tcggcgggtc ggtcgggccg attccgggca tggaaccaga
1620cctgccagcc ttgaccgaaa cggaggaatg ggaacggcgc gggcagcagc gcctgccgat
1680gcccgatgag ccgtgttttc taaggcccgt accttggtct ggacggtcgg aactggcttt
1740gcctccttac ccttgccgcg cccgtcgtcg cggacggcta cgggctactc ggcacaaaag
1800tggacgatgg cgagccgttg gagccgccga cacgggtcac gctgccgcgc cggtagcact
1860tgggttgcgc agcaacccgt aagtgcgctg ttccagacta acctgctacc gctcggcaac
1920ctcggcggct gtgcccagtg cgacggcgcg gccatcgtga acccaacgcg tcgttgggca
1980ttcacgcgac aaggtctgat tcggctgtag ccgcctcgcc gccctatacc ttgtctgcct
2040ccccgcgttg cgtcgcggtg catggagccg ggccacctcg acctgaatgg aagccggcgg
2100agccgacatc ggcggagcgg cgggatatgg aacagacgga ggggcgcaac gcagcgccac
2160gtacctcggc ccggtggagc tggacttacc ttcggccgcc cacctcgcta acggattcac
2220cgtttttatc aggctctggg aggcagaata aatgatcata tcgtcaatta ttacctccac
2280ggggagagcc tgagcaaact gtggagcgat tgcctaagtg gcaaaaatag tccgagaccc
2340tccgtcttat ttactagtat agcagttaat aatggaggtg cccctctcgg actcgtttga
2400ggcctcaggc atttgagaag cacacggtca cactgcttcc ggtagtcaat aaaccggtaa
2460accagcaata gacataagcg gctatttaac gaccctgccc ccggagtccg taaactcttc
2520gtgtgccagt gtgacgaagg ccatcagtta tttggccatt tggtcgttat ctgtattcgc
2580cgataaattg ctgggacggg tgaaccgacg accgggtcga atttgctttc gaatttctgc
2640cattcatccg cttattatca cttattcagg cgtagcacca ggcgtttaag ggcaccaata
2700acttggctgc tggcccagct taaacgaaag cttaaagacg gtaagtaggc gaataatagt
2760gaataagtcc gcatcgtggt ccgcaaattc ccgtggttat actgccttaa aaaaattacg
2820ccccgccctg ccactcatcg cagtcggcct attggttaaa aaatgagctg atttaacaaa
2880aatttaacgc gaattttaac tgacggaatt tttttaatgc ggggcgggac ggtgagtagc
2940gtcagccgga taaccaattt tttactcgac taaattgttt ttaaattgcg cttaaaattg
3000aaaatattaa cgcttacaat ttccattcgc cattcaggct gcgcaactgt tgggaagggc
3060gatcggtgcg ggcctcttcg ctattacgcc agctggcgaa ttttataatt gcgaatgtta
3120aaggtaagcg gtaagtccga cgcgttgaca acccttcccg ctagccacgc ccggagaagc
3180gataatgcgg tcgaccgctt agggggatgt gctgcaaggc gattaagttg ggtaacgcca
3240gggttttccc agtcacgacg ttgtaaaacg acggccagtg agcgcgcgta atacgactca
3300tccccctaca cgacgttccg ctaattcaac ccattgcggt cccaaaaggg tcagtgctgc
3360aacattttgc tgccggtcac tcgcgcgcat tatgctgagt ctatagggcg aattggagct
3420ccaccgcggt ggcggccgct ctagaactag tggatccccc gggctgcagg aattcgatat
3480caagcttatc gataccgtcg gatatcccgc ttaacctcga ggtggcgcca ccgccggcga
3540gatcttgatc acctaggggg cccgacgtcc ttaagctata gttcgaatag ctatggcagc
3600acgggcccgg gatccgatgc tcttccgcta agatctttta ctagttcagt ccatctcgcc
3660gtgtatgcgg gcctgacgga tcaacgttcc caccgagcca tgcccgggcc ctaggctacg
3720agaaggcgat tctagaaaat gatcaagtca ggtagagcgg cacatacgcc cggactgcct
3780agttgcaagg gtggctcggt gtcgagatgt tcatctggtc ggcgatctgc cggtacttca
3840aaccttgttt gcgcagttcc acagccttct tgcggcgttc ctgcgcacga gcgatgtagt
3900cagctctaca agtagaccag ccgctagacg gccatgaagt ttggaacaaa cgcgtcaagg
3960tgtcggaaga acgccgcaag gacgcgtgct cgctacatca cgcctcggtc ttcggcgacg
4020agccgtttga tggtgctttt cgagacgccg aacttgtcag ccaactcctg cgcggtctgc
4080gtgcgacgca tcacgcgttc gcggagccag aagccgctgc tcggcaaact accacgaaaa
4140gctctgcggc ttgaacagtc ggttgaggac gcgccagacg cacgctgcgt agtgcgcaag
4200tgcagcaccc atcagtccgt cccctctgct gctgcgaaca gtgccgatcg atcgaccttc
4260ttgagcttcg gccgcggcgc ggtggcgttc ttccgtaccg acgtcgtggg tagtcaggca
4320ggggagacga cgacgcttgt cacggctagc tagctggaag aactcgaagc cggcgccgcg
4380ccaccgcaag aaggcatggc cttccgtttt tgcgctgctg ctcactttgc cgcggcgtgc
4440ctggattttc gagaactcgg cggcggtgaa ggtgcggtgg gtccagtggg cgactgattt
4500gaaggcaaaa acgcgacgac gagtgaaacg gcgccgcacg gacctaaaag ctcttgagcc
4560gccgccactt ccacgccacc caggtcaccc gctgactaaa gccgatctgc tcggcctcgg
4620cccgactcat ggggccgatc ccgtcgttgg cgtcgagggt gaagttggtc agggcggtga
4680agtcggtgac catctgccgc cggctagacg agccggagcc gggctgagta ccccggctag
4740ggcagcaacc gcagctccca cttcaaccag tcccgccact tcagccactg gtagacggcg
4800cacacagtga tcgacgggta gttctgtttc cggatctcgc ggtaggccca ttcccgggtg
4860cggtcgaaca gttcgacgtt ccggcccgtt tcggtcctga gtgtgtcact agctgcccat
4920caagacaaag gcctagagcg ccatccgggt aagggcccac gccagcttgt caagctgcaa
4980ggccgggcaa agccaggact cctgtgtctt gcggccgtag tccggtgggg cggggaaacg
5040gtcaccgagc gcttttgcga ggcctttgag cgagtacgga tccgagggac cccagaccgt
5100ggacacagaa cgccggcatc aggccacccc gcccctttgc cagtggctcg cgaaaacgct
5160ccggaaactc gctcatgcct aggctccctg gggtctggca cgtccagtgc gggtggatcg
5220ggttctgggt gagctgctgc gcgtagccct gatcggcgcc gaccaccgag gcgatcagcc
5280cctggttcac ccggtcgtag gcaggtcacg cccacctagc ccaagaccca ctcgacgacg
5340cgcatcggga ctagccgcgg ctggtggctc cgctagtcgg ggaccaagtg ggccagcatc
5400agccgcagcg ggccctgtcg ggctgcctgg agggtgtaga ccgggctttc gagcagccac
5460cacaggtgcg cgtgctcggt cgcgggattg atcgtcatca tcggcgtcgc ccgggacagc
5520ccgacggacc tcccacatct ggcccgaaag ctcgtcggtg gtgtccacgc gcacgagcca
5580gcgccctaac tagcagtagt cggtcggatc gggcagatcc gcgttacgtg cggcccactg
5640cgcctggtcg tcgtccacgt cgagcaccaa gcccaacctg atcgacgggg tgcgggccgc
5700gccagcctag cccgtctagg cgcaatgcac gccgggtgac gcggaccagc agcaggtgca
5760gctcgtggtt cgggttggac tagctgcccc acgcccggcg aatgtagcgg cgggtgagcg
5820cctccgcgcg cggctgcggc cactgcccgt cccggacgta gtcatccgtc gcgtgcgggt
5880atttgaaccg ccagcggtcc ttacatcgcc gcccactcgc ggaggcgcgc gccgacgccg
5940gtgacgggca gggcctgcat cagtaggcag cgcacgccca taaacttggc ggtcgccagg
6000aaccaggcgt caacagcagc ggtcatgacc gccaagctag ggccggatct gtaccgatcg
6060ggggaggcgc gccgcaaatt atttaagagt ctcgctagca ttggtccgca gttgtcgtcg
6120ccagtactgg cggttcgatc ccggcctaga catggctagc cccctccgcg cggcgtttaa
6180taaattctca gagcgatcgt aaccatgtca ggtgttgcgg tgggttccgg gtaaacctcc
6240acccgaatta tttaagagtc tcgctagcta agccctatct gatgctgcgc ggggggtcct
6300ttggtacagt ccacaacgcc acccaaggcc catttggagg tgggcttaat aaattctcag
6360agcgatcgat tcgggataga ctacgacgcg ccccccagga tcgcactgaa tctcaaaggt
6420ggccggctga atttcgtcgc gcgaaaacct ccctggacag ttctggaatt cagcaagagg
6480tgtgtctgaa cttcggtgtt agcgtgactt agagtttcca ccggccgact taaagcagcg
6540cgcttttgga gggacctgtc aagaccttaa gtcgttctcc acacagactt gaagccacaa
6600tttttggggg gtgactccag cggggtgggc acaacgcgaa cagagacctt gtgtgtacga
6660cggcgggagg taagtcgggt acggctcgga ctgcggtaga aaaaaccccc cactgaggtc
6720gccccacccg tgttgcgctt gtctctggaa cacacatgct gccgccctcc attcagccca
6780tgccgagcct gacgccatct gcaaccgtcg aatcgatttc gagcagagcg agcagagcaa
6840gatattccaa aactccgggg ttcctcggcg gcctcccccg tctgtttgct caaccgaggg
6900cgttggcagc ttagctaaag ctcgtctcgc tcgtctcgtt ctataaggtt ttgaggcccc
6960aaggagccgc cggagggggc agacaaacga gttggctccc agacctggcg gtcccgcgtt
7020tccggacgcg cgggaccgcc taccgctcga gagcggaaga gcatctagat gcattcgcga
7080ggtacccagc ttttgttccc tctggaccgc cagggcgcaa aggcctgcgc gccctggcgg
7140atggcgagct ctcgccttct cgtagatcta cgtaagcgct ccatgggtcg aaaacaaggg
7200tttagtgagg gttaattgcg cgcttggcgt aatcatggtc atagctgttt cctgtgtgaa
7260attgttatcc gctcacaatt ccacacaaca tacgagccgg aaatcactcc caattaacgc
7320gcgaaccgca ttagtaccag tatcgacaaa ggacacactt taacaatagg cgagtgttaa
7380ggtgtgttgt atgctcggcc aagcataaag tgtaaagcct ggggtgccta atgagtgagc
7440taactcacat taattgcgtt gcgctcactg cccgctttcc agtcgggaaa cctgtcgtgc
7500ttcgtatttc acatttcgga ccccacggat tactcactcg attgagtgta attaacgcaa
7560cgcgagtgac gggcgaaagg tcagcccttt ggacagcacg cagctgcatt aatgaatcgg
7620ccaacgcgcg gggagaggcg gtttgcgtat tgggcgcatg cataaaaact gttgtaattc
7680attaagcatt ctgccgacat gtcgacgtaa ttacttagcc ggttgcgcgc ccctctccgc
7740caaacgcata acccgcgtac ggaagccatc acaaacggca tgatgaacct gaatcgccag
7800cggcatcagc accttgtcgc cttgcgtata atatttgccc atgggggtgg gcgaagaact
7860ccttcggtag tgtttgccgt actacttgga cttagcggtc gccgtagtcg tggaacagcg
7920gaacgcatat tataaacggg tacccccacc cgcttcttga ccagcatgag atccccgcgc
7980tggaggatca tccagccggc gtcccggaaa acgattccga agcccaacct ttcatagaag
8040gcggcggtgg aatcgaaatc ggtcgtactc taggggcgcg acctcctagt aggtcggccg
8100cagggccttt tgctaaggct tcgggttgga aagtatcttc cgccgccacc ttagctttag
8160tcgtgatggc aggttgggcg tcgcttggtc ggtcatttcg aaccccagag tcccgctcag
8220aagaactcgt caagaaggcg atagaaggcg atgcgctgcg agcactaccg tccaacccgc
8280agcgaaccag ccagtaaagc ttggggtctc agggcgagtc ttcttgagca gttcttccgc
8340tatcttccgc tacgcgacgc aatcgggagc ggcgataccg taaagcacga ggaagcggtc
8400agcccattcg ccgccaagct cttcagcaat atcacgggta gccaacgcta tgtcctgata
8460ttagccctcg ccgctatggc atttcgtgct ccttcgccag tcgggtaagc ggcggttcga
8520gaagtcgtta tagtgcccat cggttgcgat acaggactat gcggtccgcc acacccagcc
8580ggccacagtc gatgaatcca gaaaagcggc cattttccac catgatattc ggcaagcagg
8640catcgccatg ggtcacgacg cgccaggcgg tgtgggtcgg ccggtgtcag ctacttaggt
8700cttttcgccg gtaaaaggtg gtactataag ccgttcgtcc gtagcggtac ccagtgctgc
8760agatcctcgc cgtcgggcat gcgcgccttg agcctggcga acagttcggc tggcgcgagc
8820ccctgatgct cttcgtccag atcatcctga tcgacaagac tctaggagcg gcagcccgta
8880cgcgcggaac tcggaccgct tgtcaagccg accgcgctcg gggactacga gaagcaggtc
8940tagtaggact agctgttctg cggcttccat ccgagtacgt gctcgctcga tgcgatgttt
9000cgcttggtgg tcgaatgggc aggtagccgg atcaagcgta tgcagccgcc gcattgcatc
9060gccgaaggta ggctcatgca cgagcgagct acgctacaaa gcgaaccacc agcttacccg
9120tccatcggcc tagttcgcat acgtcggcgg cgtaacgtag agccatgatg gatactttct
9180cggcaggagc aaggtgagat gacaggagat cctgccccgg cacttcgccc aatagcagcc
9240agtcccttcc cgcttcagtt cggtactacc tatgaaagag ccgtcctcgt tccactctac
9300tgtcctctag gacggggccg tgaagcgggt tatcgtcggt cagggaaggg cgaagtcaca
9360caacgtcgag cacagctgcg caaggaacgc ccgtcgtggc cagccacgat agccgcgctg
9420cctcgtcctg cagttcattc agggcaccgg acaggtcggt gttgcagctc gtgtcgacgc
9480gttccttgcg ggcagcaccg gtcggtgcta tcggcgcgac ggagcaggac gtcaagtaag
9540tcccgtggcc tgtccagcct cttgacaaaa agaaccgggc gcccctgcgc tgacagccgg
9600aacacggcgg catcagagca gccgattgtc tgttgtgccc agtcatagcc gaatagccta
9660gaactgtttt tcttggcccg cggggacgcg actgtcggcc ttgtgccgcc gtagtctcgt
9720cggctaacag acaacacggg tcagtatcgg cttatcggac tccacccaag cggccggaga
9780acctgcgtgc aatccatctt gttcaatcat gcgaaacgat cctcatcctg tctcttgatc
9840agatcttgat cccctgcgcg aggtgggttc gccggcctct tggacgcacg ttaggtagaa
9900caagttagta cgctttgcta ggagtaggac agagaactag tctagaacta ggggacgcgc
9960atcagatcct tggcggcaag aaagccatcc agtttacttt gcagggcttc ccaaccttac
10020cagagggcgc cccagctggc aattccggtt cgcttgctgg tagtctagga accgccgttc
10080tttcggtagg tcaaatgaaa cgtcccgaag ggttggaatg gtctcccgcg gggtcgaccg
10140ttaaggccaa gcgaacgact ccataaaacc gcccagtcta gctatcgcca tgtaagccca
10200ctgcaagcta cctgctttct ctttgcgctt gcgttttccc ttgtccagat agcccagtaa
10260ggtattttgg cgggtcagat cgatagcggt acattcgggt gacgttcgat ggacgaaaga
10320gaaacgcgaa cgcaaaaggg aacaggtcta tcgggtcatg ctgacattca tcccaggtgg
10380cacttttcgg ggaaatgtgc gcgcccgcgt tcctgctggc gctgggcctg tttctggcgc
10440tggacttccc gctgttccgc gactgtaagt agggtccacc gtgaaaagcc cctttacacg
10500cgcgggcgca aggacgaccg cgacccggac aaagaccgcg acctgaaggg cgacaaggct
10560cagcagcttt tcgcccacgg ccttgatgat cgcggcggcc ttggcctgca tatcccgatt
10620caacggcccc agggcgtcca gaacgggctt caggcgctca gtcgtcgaaa agcgggtgcc
10680ggaactacta gcgccgccgg aaccggacgt atagggctaa gttgccgggg tcccgcaggt
10740cttgcccgaa gtccgcgacc gaaggtctcg ggccgtctct tgggcttgat cggccttctt
10800gcgcatctca cgcgctcctg cggcggcctg tagggcaggc tcatacccct gccgaaccgg
10860cttccagagc ccggcagaga acccgaacta gccggaagaa cgcgtagagt gcgcgaggac
10920gccgccggac atcccgtccg agtatgggga cggcttgggc ttttgtcagc cggtcggcca
10980cggcttccgg cgtctcaacg cgctttgaga ttcccagctt ttcggccaat ccctgcggtg
11040cataggcgcg tggctcgacg aaaacagtcg gccagccggt gccgaaggcc gcagagttgc
11100gcgaaactct aagggtcgaa aagccggtta gggacgccac gtatccgcgc accgagctcc
11160gcttgcgggc tgatggtgac gtggcccact ggtggccgct ccagggcctc gtagaacgcc
11220tgaatgcgcg tgtgacgtgc cttgctgccc tcgatgccgg cgaacgcccg actaccactg
11280caccgggtga ccaccggcga ggtcccggag catcttgcgg acttacgcgc acactgcacg
11340gaacgacggg agctacggcc gttgcagccc tagatcggcc acagcggccg caaacgtggt
11400ctggtcgcgg gtcatctgcg ctttgttgcc gatgaactcc ttggccgaca gcctgccggg
11460caacgtcggg atctagccgg tgtcgccggc gtttgcacca gaccagcgcc cagtagacgc
11520gaaacaacgg ctacttgagg aaccggctgt cggacggctc ctgcgtcagc ggcaccacga
11580acgcggtcat gtgcgggctg gtttcgtcac ggtggatgct ggccgtcacg atgcgatccg
11640ccccgtactt gtccgccaag gacgcagtcg ccgtggtgct tgcgccagta cacgcccgac
11700caaagcagtg ccacctacga ccggcagtgc tacgctaggc ggggcatgaa caggcggtgc
11760cacttgtgcg ccttctcgaa gaacgccgcc tgctgttctt ggctggccga cttccaccat
11820tccgggctgg ccgtcatgac gtactcgacc gccaacaccg gtgaacacgc ggaagagctt
11880cttgcggcgg acgacaagaa ccgaccggct gaaggtggta aggcccgacc ggcagtactg
11940catgagctgg cggttgtgag cgtccttgcg ccgcttctct ggcagcaact cgcgcagtcg
12000gcccatcgct tcatcggtgc tgctggccgc ccagtgctcg ttctctggcg tcctgctgtc
12060gcaggaacgc ggcgaagaga ccgtcgttga gcgcgtcagc cgggtagcga agtagccacg
12120acgaccggcg ggtcacgagc aagagaccgc aggacgacgc gtcagcgttg ggcgtctcgc
12180gctcgcggta ggcgtgcttg agactggccg ccacgttgcc cattttcgcc agcttcttgc
12240atcgcatgat cgcgtatgcg cagtcgcaac ccgcagagcg cgagcgccat ccgcacgaac
12300tctgaccggc ggtgcaacgg gtaaaagcgg tcgaagaacg tagcgtacta gcgcataccc
12360gccatgcctg cccctccctt ttggtgtcca accggctcga cgggggcagc gcaaggcggt
12420gcctccggcg ggccactcaa tgcttgagta tactcactgg cggtacggac ggggagggaa
12480aaccacaggt tggccgagct gcccccgtcg cgttccgcca cggaggccgc ccggtgagtt
12540acgaactcat atgagtgaag actttgcttc gcaaagtcgt gaccgcctac ggcggctgcg
12600gcgccctacg ggcttgctct ccgggcttcg ccctgcgcgg tcgctgcgct cccttgcctc
12660tgaaacgaag cgtttcagca ctggcggatg ccgccgacgc cgcgggatgc ccgaacgaga
12720ggcccgaagc gggacgcgcc agcgacgcga gggaacgg
12758415158DNAArtificial SequenceDescription of Artificial Sequence
Synthetic polynucleotide 4ggggagccgc gccgaaggcg tgggggaacc
ccgcaggggt gcccttcttt gggcaccaaa 60gaactagata tagggcgaaa tgcgaaagac
ttaaaaatca cccctcggcg cggcttccgc 120acccccttgg ggcgtcccca cgggaagaaa
cccgtggttt cttgatctat atcccgcttt 180acgctttctg aatttttagt acaacttaaa
aaaggggggt acgcaacagc tcattgcggc 240accccccgca atagctcatt gcgtaggtta
aagaaaatct gtaattgact gccactttta 300tgttgaattt tttcccccca tgcgttgtcg
agtaacgccg tggggggcgt tatcgagtaa 360cgcatccaat ttcttttaga cattaactga
cggtgaaaat cgcaacgcat aattgttgtc 420gcgctgccga aaagttgcag ctgattgcgc
atggtgccgc aaccgtgcgg caccctaccg 480catggagata agcatggcca gcgttgcgta
ttaacaacag cgcgacggct tttcaacgtc 540gactaacgcg taccacggcg ttggcacgcc
gtgggatggc gtacctctat tcgtaccggt 600cgcagtccag agaaatcggc attcaagcca
agaacaagcc cggtcactgg gtgcaaacgg 660aacgcaaagc gcatgaggcg tgggccgggc
ttattgcgag gcgtcaggtc tctttagccg 720taagttcggt tcttgttcgg gccagtgacc
cacgtttgcc ttgcgtttcg cgtactccgc 780acccggcccg aataacgctc gaaacccacg
gcggcaatgc tgctgcatca cctcgtggcg 840cagatgggcc accagaacgc cgtggtggtc
agccagaaga cactttccaa gctcatcgga 900ctttgggtgc cgccgttacg acgacgtagt
ggagcaccgc gtctacccgg tggtcttgcg 960gcaccaccag tcggtcttct gtgaaaggtt
cgagtagcct cgttctttgc ggacggtcca 1020atacgcagtc aaggacttgg tggccgagcg
ctggatctcc gtcgtgaagc tcaacggccc 1080cggcaccgtg tcggcctacg gcaagaaacg
cctgccaggt tatgcgtcag ttcctgaacc 1140accggctcgc gacctagagg cagcacttcg
agttgccggg gccgtggcac agccggatgc 1200tggtcaatga ccgcgtggcg tggggccagc
cccgcgacca gttgcgcctg tcggtgttca 1260gtgccgccgt ggtggttgat cacgacgacc
aggacgaatc accagttact ggcgcaccgc 1320accccggtcg gggcgctggt caacgcggac
agccacaagt cacggcggca ccaccaacta 1380gtgctgctgg tcctgcttag gctgttgggg
catggcgacc tgcgccgcat cccgaccctg 1440tatccgggcg agcagcaact accgaccggc
cccggcgagg agccgcccag ccagcccggc 1500cgacaacccc gtaccgctgg acgcggcgta
gggctgggac ataggcccgc tcgtcgttga 1560tggctggccg gggccgctcc tcggcgggtc
ggtcgggccg attccgggca tggaaccaga 1620cctgccagcc ttgaccgaaa cggaggaatg
ggaacggcgc gggcagcagc gcctgccgat 1680gcccgatgag ccgtgttttc taaggcccgt
accttggtct ggacggtcgg aactggcttt 1740gcctccttac ccttgccgcg cccgtcgtcg
cggacggcta cgggctactc ggcacaaaag 1800tggacgatgg cgagccgttg gagccgccga
cacgggtcac gctgccgcgc cggtagcact 1860tgggttgcgc agcaacccgt aagtgcgctg
ttccagacta acctgctacc gctcggcaac 1920ctcggcggct gtgcccagtg cgacggcgcg
gccatcgtga acccaacgcg tcgttgggca 1980ttcacgcgac aaggtctgat tcggctgtag
ccgcctcgcc gccctatacc ttgtctgcct 2040ccccgcgttg cgtcgcggtg catggagccg
ggccacctcg acctgaatgg aagccggcgg 2100agccgacatc ggcggagcgg cgggatatgg
aacagacgga ggggcgcaac gcagcgccac 2160gtacctcggc ccggtggagc tggacttacc
ttcggccgcc cacctcgcta acggattcac 2220cgtttttatc aggctctggg aggcagaata
aatgatcata tcgtcaatta ttacctccac 2280ggggagagcc tgagcaaact gtggagcgat
tgcctaagtg gcaaaaatag tccgagaccc 2340tccgtcttat ttactagtat agcagttaat
aatggaggtg cccctctcgg actcgtttga 2400ggcctcaggc atttgagaag cacacggtca
cactgcttcc ggtagtcaat aaaccggtaa 2460accagcaata gacataagcg gctatttaac
gaccctgccc ccggagtccg taaactcttc 2520gtgtgccagt gtgacgaagg ccatcagtta
tttggccatt tggtcgttat ctgtattcgc 2580cgataaattg ctgggacggg tgaaccgacg
accgggtcga atttgctttc gaatttctgc 2640cattcatccg cttattatca cttattcagg
cgtagcacca ggcgtttaag ggcaccaata 2700acttggctgc tggcccagct taaacgaaag
cttaaagacg gtaagtaggc gaataatagt 2760gaataagtcc gcatcgtggt ccgcaaattc
ccgtggttat actgccttaa aaaaattacg 2820ccccgccctg ccactcatcg cagtcggcct
attggttaaa aaatgagctg atttaacaaa 2880aatttaacgc gaattttaac tgacggaatt
tttttaatgc ggggcgggac ggtgagtagc 2940gtcagccgga taaccaattt tttactcgac
taaattgttt ttaaattgcg cttaaaattg 3000aaaatattaa cgcttacaat ttccattcgc
cattcaggct gcgcaactgt tgggaagggc 3060gatcggtgcg ggcctcttcg ctattacgcc
agctggcgaa ttttataatt gcgaatgtta 3120aaggtaagcg gtaagtccga cgcgttgaca
acccttcccg ctagccacgc ccggagaagc 3180gataatgcgg tcgaccgctt agggggatgt
gctgcaaggc gattaagttg ggtaacgcca 3240gggttttccc agtcacgacg ttgtaaaacg
acggccagtg agcgcgcgta atacgactca 3300tccccctaca cgacgttccg ctaattcaac
ccattgcggt cccaaaaggg tcagtgctgc 3360aacattttgc tgccggtcac tcgcgcgcat
tatgctgagt ctatagggcg aattggagct 3420ccaccgcggt ggcggccgct ctagaactag
tggatccccc gggctgcagg aattcgatat 3480caagctttta cgccccgccc gatatcccgc
ttaacctcga ggtggcgcca ccgccggcga 3540gatcttgatc acctaggggg cccgacgtcc
ttaagctata gttcgaaaat gcggggcggg 3600tgccactcat cgcagtactg ttgtaattca
ttaagcattc tgccgacatg gaagccatca 3660caaacggcat gatgaacctg aatcgccagc
ggcatcagca acggtgagta gcgtcatgac 3720aacattaagt aattcgtaag acggctgtac
cttcggtagt gtttgccgta ctacttggac 3780ttagcggtcg ccgtagtcgt ccttgtcgcc
ttgcgtataa tatttgccca tggtgaaaac 3840gggggcgaag aagttgtcca tattggccac
gtttaaatca aaactggtga aactcaccca 3900ggaacagcgg aacgcatatt ataaacgggt
accacttttg cccccgcttc ttcaacaggt 3960ataaccggtg caaatttagt tttgaccact
ttgagtgggt gggattggct gagacgaaaa 4020acatattctc aataaaccct ttagggaaat
aggccaggtt ttcaccgtaa cacgccacat 4080cttgcgaata tatgtgtaga ccctaaccga
ctctgctttt tgtataagag ttatttggga 4140aatcccttta tccggtccaa aagtggcatt
gtgcggtgta gaacgcttat atacacatct 4200aactgccgga aatcgtcgtg gtattcactc
cagagcgatg aaaacgtttc agtttgctca 4260tggaaaacgg tgtaacaagg gtgaacacta
tcccatatca ttgacggcct ttagcagcac 4320cataagtgag gtctcgctac ttttgcaaag
tcaaacgagt accttttgcc acattgttcc 4380cacttgtgat agggtatagt ccagctcacc
gtctttcatt gccatacgaa attccggatg 4440agcattcatc aggcgggcaa gaatgtgaat
aaaggccgga taaaacttgt gcttattttt 4500ggtcgagtgg cagaaagtaa cggtatgctt
taaggcctac tcgtaagtag tccgcccgtt 4560cttacactta tttccggcct attttgaaca
cgaataaaaa ctttacggtc tttaaaaagg 4620ccgtaatatc cagctgaacg gtctggttat
aggtacattg agcaactgac tgaaatgcct 4680caaaatgttc tttacgatgc gaaatgccag
aaatttttcc ggcattatag gtcgacttgc 4740cagaccaata tccatgtaac tcgttgactg
actttacgga gttttacaag aaatgctacg 4800cattgggata tatcaacggt ggtatatcca
gtgatttttt tctccatatg gttaacctta 4860attaaggggt cgacgggccc gggatccgat
gctcttccgc gtaaccctat atagttgcca 4920ccatataggt cactaaaaaa agaggtatac
caattggaat taattcccca gctgcccggg 4980ccctaggcta cgagaaggcg taagatcttt
tactagttca gtccatctcg ccgtgtatgc 5040gggcctgacg gatcaacgtt cccaccgagc
cagtcgagat gttcatctgg tcggcgatct 5100attctagaaa atgatcaagt caggtagagc
ggcacatacg cccggactgc ctagttgcaa 5160gggtggctcg gtcagctcta caagtagacc
agccgctaga gccggtactt caaaccttgt 5220ttgcgcagtt ccacagcctt cttgcggcgt
tcctgcgcac gagcgatgta gtcgcctcgg 5280tcttcggcga cgagccgttt cggccatgaa
gtttggaaca aacgcgtcaa ggtgtcggaa 5340gaacgccgca aggacgcgtg ctcgctacat
cagcggagcc agaagccgct gctcggcaaa 5400gatggtgctt ttcgagacgc cgaacttgtc
agccaactcc tgcgcggtct gcgtgcgacg 5460catcacgcgt tctgcagcac ccatcagtcc
gtcccctctg ctaccacgaa aagctctgcg 5520gcttgaacag tcggttgagg acgcgccaga
cgcacgctgc gtagtgcgca agacgtcgtg 5580ggtagtcagg caggggagac ctgctgcgaa
cagtgccgat cgatcgacct tcttgagctt 5640cggccgcggc gcggtggcgt tcttccgtac
cgcttccgtt tttgcgctgc tgctcacttt 5700gacgacgctt gtcacggcta gctagctgga
agaactcgaa gccggcgccg cgccaccgca 5760agaaggcatg gcgaaggcaa aaacgcgacg
acgagtgaaa gccgcggcgt gcctggattt 5820tcgagaactc ggcggcggtg aaggtgcggt
gggtccagtg ggcgactgat ttgccgatct 5880gctcggcctc ggcccgactc cggcgccgca
cggacctaaa agctcttgag ccgccgccac 5940ttccacgcca cccaggtcac ccgctgacta
aacggctaga cgagccggag ccgggctgag 6000atggggccga tcccgtcgtt ggcgtcgagg
gtgaagttgg tcagggcggt gaagtcggtg 6060accatctgcc gccacacagt gatcgacggg
tagttctgtt taccccggct agggcagcaa 6120ccgcagctcc cacttcaacc agtcccgcca
cttcagccac tggtagacgg cggtgtgtca 6180ctagctgccc atcaagacaa tccggatctc
gcggtaggcc cattcccggg tgcggtcgaa 6240cagttcgacg ttccggcccg tttcggtcct
gacctgtgtc ttgcggccgt agtccggtgg 6300aggcctagag cgccatccgg gtaagggccc
acgccagctt gtcaagctgc aaggccgggc 6360aaagccagga ctggacacag aacgccggca
tcaggccacc ggcggggaaa cggtcaccga 6420gcgcttttgc gaggcctttg agcgagtacg
gatccgaggg accccagacc gtcgtccagt 6480gcgggtggat cgggttctgg ccgccccttt
gccagtggct cgcgaaaacg ctccggaaac 6540tcgctcatgc ctaggctccc tggggtctgg
cagcaggtca cgcccaccta gcccaagacc 6600gtgagctgct gcgcgtagcc ctgatcggcg
ccgaccaccg aggcgatcag cccctggttc 6660acccggtcgt agagccgcag cgggccctgt
cgggctgcct cactcgacga cgcgcatcgg 6720gactagccgc ggctggtggc tccgctagtc
ggggaccaag tgggccagca tctcggcgtc 6780gcccgggaca gcccgacgga ggagggtgta
gaccgggctt tcgagcagcc accacaggtg 6840cgcgtgctcg gtcgcgggat tgatcgtcat
cacggtcgga tcgggcagat ccgcgttacg 6900cctcccacat ctggcccgaa agctcgtcgg
tggtgtccac gcgcacgagc cagcgcccta 6960actagcagta gtgccagcct agcccgtcta
ggcgcaatgc tgcggcccac tgcgcctggt 7020cgtcgtccac gtcgagcacc aagcccaacc
tgatcgacgg ggtgcgggcc gcaatgtagc 7080ggcgggtgag cgcctccgcg acgccgggtg
acgcggacca gcagcaggtg cagctcgtgg 7140ttcgggttgg actagctgcc ccacgcccgg
cgttacatcg ccgcccactc gcggaggcgc 7200cgcggctgcg gccactgccc gtcccggacg
tagtcatccg tcgcgtgcgg gtatttgaac 7260cgccagcggt ccaaccaggc gtcaacagca
gcggtcatga gcgccgacgc cggtgacggg 7320cagggcctgc atcagtaggc agcgcacgcc
cataaacttg gcggtcgcca ggttggtccg 7380cagttgtcgt cgccagtact ccgccaagct
agggccggat ctgtaccgat cgggggaggc 7440gcgccgcaaa ttatttaaga gtctcgctag
caaaccatgt caggtgttgc ggtgggttcc 7500ggcggttcga tcccggccta gacatggcta
gccccctccg cgcggcgttt aataaattct 7560cagagcgatc gtttggtaca gtccacaacg
ccacccaagg gggtaaacct ccacccgaat 7620tatttaagag tctcgctagc taagccctat
ctgatgctgc gcggggggtc cttcgcactg 7680aatctcaaag gtggccggct cccatttgga
ggtgggctta ataaattctc agagcgatcg 7740attcgggata gactacgacg cgccccccag
gaagcgtgac ttagagtttc caccggccga 7800gaatttcgtc gcgcgaaaac ctccctggac
agttctggaa ttcagcaaga ggtgtgtctg 7860aacttcggtg tttttttggg gggtgactcc
agcggggtgg cttaaagcag cgcgcttttg 7920gagggacctg tcaagacctt aagtcgttct
ccacacagac ttgaagccac aaaaaaaccc 7980cccactgagg tcgccccacc gcacaacgcg
aacagagacc ttgtgtgtac gacggcggga 8040ggtaagtcgg gtacggctcg gactgcggta
gagcaaccgt cgaatcgatt tcgagcagag 8100cgtgttgcgc ttgtctctgg aacacacatg
ctgccgccct ccattcagcc catgccgagc 8160ctgacgccat ctcgttggca gcttagctaa
agctcgtctc cgagcagagc aagatattcc 8220aaaactccgg ggttcctcgg cggcctcccc
cgtctgtttg ctcaaccgag ggagacctgg 8280cggtcccgcg tttccggacg gctcgtctcg
ttctataagg ttttgaggcc ccaaggagcc 8340gccggagggg gcagacaaac gagttggctc
cctctggacc gccagggcgc aaaggcctgc 8400cgcgggaccg cctaccgctc gagagcggaa
gagcatctag atgcattcgc gaggtaccca 8460gcttttgttc cctttagtga gggttaattg
cgcgcttggc gcgccctggc ggatggcgag 8520ctctcgcctt ctcgtagatc tacgtaagcg
ctccatgggt cgaaaacaag ggaaatcact 8580cccaattaac gcgcgaaccg gtaatcatgg
tcatagctgt ttcctgtgtg aaattgttat 8640ccgctcacaa ttccacacaa catacgagcc
ggaagcataa agtgtaaagc ctggggtgcc 8700cattagtacc agtatcgaca aaggacacac
tttaacaata ggcgagtgtt aaggtgtgtt 8760gtatgctcgg ccttcgtatt tcacatttcg
gaccccacgg taatgagtga gctaactcac 8820attaattgcg ttgcgctcac tgcccgcttt
ccagtcggga aacctgtcgt gccagctgca 8880ttaatgaatc ggccaacgcg attactcact
cgattgagtg taattaacgc aacgcgagtg 8940acgggcgaaa ggtcagccct ttggacagca
cggtcgacgt aattacttag ccggttgcgc 9000cggggagagg cggtttgcgt attgggcgca
tgcataaaaa ctgttgtaat tcattaagca 9060ttctgccgac atggaagcca tcacaaacgg
catgatgaac gcccctctcc gccaaacgca 9120taacccgcgt acgtattttt gacaacatta
agtaattcgt aagacggctg taccttcggt 9180agtgtttgcc gtactacttg ctgaatcgcc
agcggcatca gcaccttgtc gccttgcgta 9240taatatttgc ccatgggggt gggcgaagaa
ctccagcatg agatccccgc gctggaggat 9300gacttagcgg tcgccgtagt cgtggaacag
cggaacgcat attataaacg ggtaccccca 9360cccgcttctt gaggtcgtac tctaggggcg
cgacctccta catccagccg gcgtcccgga 9420aaacgattcc gaagcccaac ctttcataga
aggcggcggt ggaatcgaaa tctcgtgatg 9480gcaggttggg cgtcgcttgg gtaggtcggc
cgcagggcct tttgctaagg cttcgggttg 9540gaaagtatct tccgccgcca ccttagcttt
agagcactac cgtccaaccc gcagcgaacc 9600tcggtcattt cgaaccccag agtcccgctc
agaagaactc gtcaagaagg cgatagaagg 9660cgatgcgctg cgaatcggga gcggcgatac
cgtaaagcac agccagtaaa gcttggggtc 9720tcagggcgag tcttcttgag cagttcttcc
gctatcttcc gctacgcgac gcttagccct 9780cgccgctatg gcatttcgtg gaggaagcgg
tcagcccatt cgccgccaag ctcttcagca 9840atatcacggg tagccaacgc tatgtcctga
tagcggtccg ccacacccag ccggccacag 9900ctccttcgcc agtcgggtaa gcggcggttc
gagaagtcgt tatagtgccc atcggttgcg 9960atacaggact atcgccaggc ggtgtgggtc
ggccggtgtc tcgatgaatc cagaaaagcg 10020gccattttcc accatgatat tcggcaagca
ggcatcgcca tgggtcacga cgagatcctc 10080gccgtcgggc atgcgcgcct agctacttag
gtcttttcgc cggtaaaagg tggtactata 10140agccgttcgt ccgtagcggt acccagtgct
gctctaggag cggcagcccg tacgcgcgga 10200tgagcctggc gaacagttcg gctggcgcga
gcccctgatg ctcttcgtcc agatcatcct 10260gatcgacaag accggcttcc atccgagtac
gtgctcgctc actcggaccg cttgtcaagc 10320cgaccgcgct cggggactac gagaagcagg
tctagtagga ctagctgttc tggccgaagg 10380taggctcatg cacgagcgag gatgcgatgt
ttcgcttggt ggtcgaatgg gcaggtagcc 10440ggatcaagcg tatgcagccg ccgcattgca
tcagccatga tggatacttt ctcggcagga 10500ctacgctaca aagcgaacca ccagcttacc
cgtccatcgg cctagttcgc atacgtcggc 10560ggcgtaacgt agtcggtact acctatgaaa
gagccgtcct gcaaggtgag atgacaggag 10620atcctgcccc ggcacttcgc ccaatagcag
ccagtccctt cccgcttcag tgacaacgtc 10680gagcacagct gcgcaaggaa cgttccactc
tactgtcctc taggacgggg ccgtgaagcg 10740ggttatcgtc ggtcagggaa gggcgaagtc
actgttgcag ctcgtgtcga cgcgttcctt 10800cgcccgtcgt ggccagccac gatagccgcg
ctgcctcgtc ctgcagttca ttcagggcac 10860cggacaggtc ggtcttgaca aaaagaaccg
ggcgcccctg gcgggcagca ccggtcggtg 10920ctatcggcgc gacggagcag gacgtcaagt
aagtcccgtg gcctgtccag ccagaactgt 10980ttttcttggc ccgcggggac cgctgacagc
cggaacacgg cggcatcaga gcagccgatt 11040gtctgttgtg cccagtcata gccgaatagc
ctctccaccc aagcggccgg agaacctgcg 11100gcgactgtcg gccttgtgcc gccgtagtct
cgtcggctaa cagacaacac gggtcagtat 11160cggcttatcg gagaggtggg ttcgccggcc
tcttggacgc tgcaatccat cttgttcaat 11220catgcgaaac gatcctcatc ctgtctcttg
atcagatctt gatcccctgc gccatcagat 11280ccttggcggc aagaaagcca acgttaggta
gaacaagtta gtacgctttg ctaggagtag 11340gacagagaac tagtctagaa ctaggggacg
cggtagtcta ggaaccgccg ttctttcggt 11400tccagtttac tttgcagggc ttcccaacct
taccagaggg cgccccagct ggcaattccg 11460gttcgcttgc tgtccataaa accgcccagt
ctagctatcg aggtcaaatg aaacgtcccg 11520aagggttgga atggtctccc gcggggtcga
ccgttaaggc caagcgaacg acaggtattt 11580tggcgggtca gatcgatagc ccatgtaagc
ccactgcaag ctacctgctt tctctttgcg 11640cttgcgtttt cccttgtcca gatagcccag
tagctgacat tcatcccagg tggcactttt 11700ggtacattcg ggtgacgttc gatggacgaa
agagaaacgc gaacgcaaaa gggaacaggt 11760ctatcgggtc atcgactgta agtagggtcc
accgtgaaaa cggggaaatg tgcgcgcccg 11820cgttcctgct ggcgctgggc ctgtttctgg
cgctggactt cccgctgttc cgtcagcagc 11880ttttcgccca cggccttgat gcccctttac
acgcgcgggc gcaaggacga ccgcgacccg 11940gacaaagacc gcgacctgaa gggcgacaag
gcagtcgtcg aaaagcgggt gccggaacta 12000gatcgcggcg gccttggcct gcatatcccg
attcaacggc cccagggcgt ccagaacggg 12060cttcaggcgc tcccgaaggt ctcgggccgt
ctcttgggct ctagcgccgc cggaaccgga 12120cgtatagggc taagttgccg gggtcccgca
ggtcttgccc gaagtccgcg agggcttcca 12180gagcccggca gagaacccga tgatcggcct
tcttgcgcat ctcacgcgct cctgcggcgg 12240cctgtagggc aggctcatac ccctgccgaa
ccgcttttgt cagccggtcg gccacggctt 12300actagccgga agaacgcgta gagtgcgcga
ggacgccgcc ggacatcccg tccgagtatg 12360gggacggctt ggcgaaaaca gtcggccagc
cggtgccgaa ccggcgtctc aacgcgcttt 12420gagattccca gcttttcggc caatccctgc
ggtgcatagg cgcgtggctc gaccgcttgc 12480gggctgatgg tgacgtggcc ggccgcagag
ttgcgcgaaa ctctaagggt cgaaaagccg 12540gttagggacg ccacgtatcc gcgcaccgag
ctggcgaacg cccgactacc actgcaccgg 12600cactggtggc cgctccaggg cctcgtagaa
cgcctgaatg cgcgtgtgac gtgccttgct 12660gccctcgatg ccccgttgca gccctagatc
ggccacagcg gtgaccaccg gcgaggtccc 12720ggagcatctt gcggacttac gcgcacactg
cacggaacga cgggagctac ggggcaacgt 12780cgggatctag ccggtgtcgc gccgcaaacg
tggtctggtc gcgggtcatc tgcgctttgt 12840tgccgatgaa ctccttggcc gacagcctgc
cgtcctgcgt cagcggcacc acgaacgcgg 12900cggcgtttgc accagaccag cgcccagtag
acgcgaaaca acggctactt gaggaaccgg 12960ctgtcggacg gcaggacgca gtcgccgtgg
tgcttgcgcc tcatgtgcgg gctggtttcg 13020tcacggtgga tgctggccgt cacgatgcga
tccgccccgt acttgtccgc cagccacttg 13080tgcgccttct cgaagaacgc agtacacgcc
cgaccaaagc agtgccacct acgaccggca 13140gtgctacgct aggcggggca tgaacaggcg
gtcggtgaac acgcggaaga gcttcttgcg 13200cgcctgctgt tcttggctgg ccgacttcca
ccattccggg ctggccgtca tgacgtactc 13260gaccgccaac acagcgtcct tgcgccgctt
ctctggcagc gcggacgaca agaaccgacc 13320ggctgaaggt ggtaaggccc gaccggcagt
actgcatgag ctggcggttg tgtcgcagga 13380acgcggcgaa gagaccgtcg aactcgcgca
gtcggcccat cgcttcatcg gtgctgctgg 13440ccgcccagtg ctcgttctct ggcgtcctgc
tggcgtcagc gttgggcgtc tcgcgctcgc 13500ttgagcgcgt cagccgggta gcgaagtagc
cacgacgacc ggcgggtcac gagcaagaga 13560ccgcaggacg accgcagtcg caacccgcag
agcgcgagcg ggtaggcgtg cttgagactg 13620gccgccacgt tgcccatttt cgccagcttc
ttgcatcgca tgatcgcgta tgccgccatg 13680cctgcccctc ccttttggtg ccatccgcac
gaactctgac cggcggtgca acgggtaaaa 13740gcggtcgaag aacgtagcgt actagcgcat
acggcggtac ggacggggag ggaaaaccac 13800tccaaccggc tcgacggggg cagcgcaagg
cggtgcctcc ggcgggccac tcaatgcttg 13860agtatactca ctagactttg cttcgcaaag
tcgtgaccgc aggttggccg agctgccccc 13920gtcgcgttcc gccacggagg ccgcccggtg
agttacgaac tcatatgagt gatctgaaac 13980gaagcgtttc agcactggcg ctacggcggc
tgcggcgccc tacgggcttg ctctccgggc 14040ttcgccctgc gcggtcgctg cgctcccttg
ccagcccgtg gatatgtgga cgatggccgc 14100gatgccgccg acgccgcggg atgcccgaac
gagaggcccg aagcgggacg cgccagcgac 14160gcgagggaac ggtcgggcac ctatacacct
gctaccggcg gagcggccac cggctggctc 14220gcttcgctcg gcccgtggac aaccctgctg
gacaagctga tggacaggct gcgcctgccc 14280acgagcttga ccacagggat ctcgccggtg
gccgaccgag cgaagcgagc cgggcacctg 14340ttgggacgac ctgttcgact acctgtccga
cgcggacggg tgctcgaact ggtgtcccta 14400tgcccaccgg ctacccagcc ttcgaccaca
tacccaccgg ctccaactgc gcggcctgcg 14460gccttgcccc atcaattttt ttaattttct
ctggggaaaa acgggtggcc gatgggtcgg 14520aagctggtgt atgggtggcc gaggttgacg
cgccggacgc cggaacgggg tagttaaaaa 14580aattaaaaga gacccctttt gcctccggcc
tgcggcctgc gcgcttcgct tgccggttgg 14640acaccaagtg gaaggcgggt caaggctcgc
gcagcgaccg cgcagcggct tggccttgac 14700cggaggccgg acgccggacg cgcgaagcga
acggccaacc tgtggttcac cttccgccca 14760gttccgagcg cgtcgctggc gcgtcgccga
accggaactg gcgcctggaa cgacccaagc 14820ctatgcgagt gggggcagtc gaaggcgaag
cccgcccgcc tgccccccga gcctcacggc 14880ggcgagtgcg ggggttccaa cgcggacctt
gctgggttcg gatacgctca cccccgtcag 14940cttccgcttc gggcgggcgg acggggggct
cggagtgccg ccgctcacgc ccccaaggtt 15000gggggcagcg ccaccttggg caaggccgaa
ggccgcgcag tcgatcaaca agccccggag 15060gggccacttt ttgccggagc ccccgtcgcg
gtggaacccg ttccggcttc cggcgcgtca 15120gctagttgtt cggggcctcc ccggtgaaaa
acggcctc 151585231PRTArtificial
SequenceDescription of Artificial Sequence Synthetic polypeptide
5Met Pro Glu Leu Ala Val Arg Thr Glu Phe Asp Tyr Ser Ser Glu Ile1
5 10 15Tyr Lys Asp Ala Tyr Ser
Arg Ile Asn Ala Ile Val Ile Glu Gly Glu 20 25
30Gln Glu Ala Tyr Ser Asn Tyr Leu Gln Met Ala Glu Leu
Leu Pro Glu 35 40 45Asp Lys Glu
Glu Leu Thr Arg Leu Ala Lys Met Glu Asn Arg His Lys 50
55 60Lys Gly Phe Gln Ala Cys Gly Asn Asn Leu Gln Val
Asn Pro Asp Met65 70 75
80Pro Tyr Ala Gln Glu Phe Phe Ala Gly Leu His Gly Asn Phe Gln His
85 90 95Ala Phe Ser Glu Gly Lys
Val Val Thr Cys Leu Leu Ile Gln Ala Leu 100
105 110Ile Ile Glu Ala Phe Ala Ile Ala Ala Tyr Asn Ile
Tyr Ile Pro Val 115 120 125Ala Asp
Asp Phe Ala Arg Lys Ile Thr Glu Gly Val Val Lys Asp Glu 130
135 140Tyr Thr His Leu Asn Tyr Gly Glu Glu Trp Leu
Lys Ala Asn Phe Ala145 150 155
160Thr Ala Lys Glu Glu Leu Glu Gln Ala Asn Lys Glu Asn Leu Pro Leu
165 170 175Val Trp Lys Met
Leu Asn Gln Val Gln Gly Asp Ala Lys Val Leu Gly 180
185 190Met Glu Lys Glu Ala Leu Val Glu Asp Phe Met
Ile Ser Tyr Gly Glu 195 200 205Ala
Leu Ser Asn Ile Gly Phe Ser Thr Arg Glu Ile Met Arg Met Ser 210
215 220Ser Tyr Gly Leu Ala Gly Val225
2306340PRTArtificial SequenceDescription of Artificial Sequence
Synthetic polypeptide 6Met Phe Gly Leu Ile Gly His Leu Thr Ser Leu
Glu His Ala Gln Ala1 5 10
15Val Ala Glu Asp Leu Gly Tyr Pro Glu Tyr Ala Asn Gln Gly Leu Asp
20 25 30Phe Trp Cys Ser Ala Pro Pro
Gln Val Val Asp Asn Phe Gln Val Lys 35 40
45Ser Val Thr Gly Gln Val Ile Glu Gly Lys Tyr Val Glu Ser Cys
Phe 50 55 60Leu Pro Glu Met Leu Thr
Gln Arg Arg Ile Lys Ala Ala Ile Arg Lys65 70
75 80Ile Leu Asn Ala Met Ala Leu Ala Gln Lys Val
Gly Leu Asp Ile Thr 85 90
95Ala Leu Gly Gly Phe Ser Ser Ile Val Phe Glu Glu Phe Asn Leu Lys
100 105 110Gln Asn Asn Gln Val Arg
Asn Val Glu Leu Asp Phe Gln Arg Phe Thr 115 120
125Thr Gly Asn Thr His Thr Ala Tyr Val Ile Cys Arg Gln Val
Glu Ser 130 135 140Gly Ala Lys Gln Leu
Gly Ile Asp Leu Ser Gln Ala Thr Val Ala Val145 150
155 160Cys Gly Ala Thr Gly Asp Ile Gly Ser Ala
Val Cys Arg Trp Leu Asp 165 170
175Ser Lys His Gln Val Lys Glu Leu Leu Leu Ile Ala Arg Asn Arg Gln
180 185 190Arg Leu Glu Asn Leu
Gln Glu Glu Leu Gly Arg Gly Lys Ile Met Asp 195
200 205Leu Glu Thr Ala Leu Pro Gln Ala Asp Ile Ile Val
Trp Val Ala Ser 210 215 220Met Pro Lys
Gly Val Glu Ile Ala Gly Glu Met Leu Lys Lys Pro Cys225
230 235 240Leu Ile Val Asp Gly Gly Tyr
Pro Lys Asn Leu Asp Thr Arg Val Lys 245
250 255Ala Asp Gly Val His Ile Leu Lys Gly Gly Ile Val
Glu His Ser Leu 260 265 270Asp
Ile Thr Trp Glu Ile Met Lys Ile Val Glu Met Asp Ile Pro Ser 275
280 285Arg Gln Met Phe Ala Cys Phe Ala Glu
Ala Ile Leu Leu Glu Phe Glu 290 295
300Gly Trp Arg Thr Asn Phe Ser Trp Gly Arg Asn Gln Ile Ser Val Asn305
310 315 320Lys Met Glu Ala
Ile Gly Glu Ala Ser Val Lys His Gly Phe Cys Pro 325
330 335Leu Val Ala Leu
3407410DNAUnknownDescription of Unknown Cyanobacterium Rubisco large
subunit promoter polynucleotide 7cagtcaatgg agagcattgc cataagtaaa
ggcatcccct gcgtgataag attaccttca 60gaaaacagat agttgctggg ttatcgcaga
tttttctcgc gtcagttacc tctcgtaacg 120gtattcattt ccgtagggga cgcactattc
taatggaagt cttttgtcta tcaacgaccc 180aatagcgtct aaaaagagcg aaccaaataa
ctgtaaataa taactgtctc tggggcgacg 240gtaggcttta tattgccaaa tttcgcccgt
gggagaaagc taggctattc aatgtttatg 300ttggtttatt gacatttatt attgacagag
accccgctgc catccgaaat ataacggttt 360aaagcgggca ccctctttcg atccgataag
ttacaaatac gaggactcct 41082962DNARhodococcus
opacusmodified_base(1070)..(1070)a, c, t or
gmodified_base(1170)..(1170)a, c, t or g 8cctggctcag gacgaacgct
ggcggcgtgc ttaacacatg caagtcgagc ggtaaggccc 60ttcggggtac acgagcggcg
aacgggtgag taacacgtgg ggaccgagtc ctgcttgcga 120ccgccgcacg aattgtgtac
gttcagctcg ccattccggg aagccccatg tgctcgccgc 180ttgcccactc attgtgcacc
gtgatctgcc ctgcacttcg ggataagcct gggaaactgg 240gtctaatacc ggatatgacc
ttcggctgca tggctgaggg tggaaaggtt tactggtgca 300cactagacgg gacgtgaagc
cctattcgga ccctttgacc cagattatgg cctatactgg 360aagccgacgt accgactccc
acctttccaa atgaccacgt ggatgggccc gcggcctatc 420agcttgttgg tggggtaatg
gcctaccaag gcgacgacgg gtagccgacc tgagagggtg 480accggccaca ctgggactga
cctacccggg cgccggatag tcgaacaacc accccattac 540cggatggttc cgctgctgcc
catcggctgg actctcccac tggccggtgt gaccctgact 600gacacggccc agactcctac
gggaggcagc agtggggaat attgcacaat gggcgaaagc 660ctgatgcagc gacgccgcgt
gagggatgac ggccttcggg ctgtgccggg tctgaggatg 720ccctccgtcg tcacccctta
taacgtgtta cccgctttcg gactacgtcg ctgcggcgca 780ctccctactg ccggaagccc
ttgtaaacct ctttcagcag ggacgaagcg aaagtgacgg 840tacctgcaga agaagcaccg
gccaactacg tgccagcagc cgcggtaata cgtagggtgc 900aacatttgga gaaagtcgtc
cctgcttcgc tttcactgcc atggacgtct tcttcgtggc 960cggttgatgc acggtcgtcg
gcgccattat gcatcccacg aagcgttgtc cggaattact 1020gggcgtaaag agctcgtagg
cggtttgtcg cgtcgtctgt gaaaactcan agctcaacct 1080cgagcttgca ggcgatacgg
ttcgcaacag gccttaatga cccgcatttc tcgagcatcc 1140gccaaacagc gcagcagaca
cttttgagtn tcgagttgga gctcgaacgt ccgctatgcc 1200gcagacttga gtactgcagg
ggagactgga attcctggtg tagcggtgaa atgcgcagat 1260atcaggagga acaccggtgg
cgaaggcggg tctctgggca cgtctgaact catgacgtcc 1320cctctgacct taaggaccac
atcgccactt tacgcgtcta tagtcctcct tgtggccacc 1380gcttccgccc agagacccgt
gtaactgacg ctgaggagcg aaagcgtggg tagcaaacag 1440gattagatac cctggtagtc
cacgccgtaa acggtgggcg ctaggtgtgg gtttccttcc 1500cattgactgc gactcctcgc
tttcgcaccc atcgtttgtc ctaatctatg ggaccatcag 1560gtgcggcatt tgccacccgc
gatccacacc caaaggaagg acgggatccg tgccgtagtt 1620aacgcattaa gcgccccgcc
tggggagtac ggccgcaagg ttaaaactca aaggaattga 1680cgggggcccg cacaagcggc
tgccctaggc acggcatcaa ttgcgtaatt cgcggggcgg 1740acccctcatg ccggcgttcc
aattttgagt ttccttaact gcccccgggc gtgttcgccg 1800ggagcatgtg gattaattcg
atgcaacgcg aagaacctta cctgggtttg acatataccg 1860gaaagccgta gagataccgc
cccccttgtg gtcggtatac cctcgtacac ctaattaagc 1920tacgttgcgc ttcttggaat
ggacccaaac tgtatatggc ctttcggcat ctctatggcg 1980gggggaacac cagccatatg
aggtggtgca tggctgtcgt cagctcgtgt cgtgagatgt 2040tgggttaagt cccgcaacga
gcgcaaccct tgtcttatgt tgccagcacg taatggtggg 2100tccaccacgt accgacagca
gtcgagcaca gcactctaca acccaattca gggcgttgct 2160cgcgttggga acagaataca
acggtcgtgc attaccaccc gactcgtaag agactgccgg 2220ggtcaactcg gaggaaggtg
gggacgacgt caagtcatca tgccccttat gtccagggct 2280tcacacatgc tacaatggcc
ctgagcattc tctgacggcc ccagttgagc ctccttccac 2340ccctgctgca gttcagtagt
acggggaata caggtcccga agtgtgtacg atgttaccgg 2400ggtacagagg gctgcgatac
cgtgaggtgg agcgaatccc ttaaagccgg tctcagttcg 2460gatcggggtc tgcaactcga
ccccgtgaag tcggagtcgc ccatgtctcc cgacgctatg 2520gcactccacc tcgcttaggg
aatttcggcc agagtcaagc ctagccccag acgttgagct 2580ggggcacttc agcctcagcg
tagtaatcgc agatcagcaa cgctgcggtg aatacgttcc 2640cgggccttgt acacaccgcc
cgtcacgtca tgaaagtcgg taacacccga agccggtggc 2700atcattagcg tctagtcgtt
gcgacgccac ttatgcaagg gcccggaaca tgtgtggcgg 2760gcagtgcagt actttcagcc
attgtgggct tcggccaccg ctaacccctt gtgggaggga 2820gccgtcgaag gtgggatcgg
cgattgggac gaagtcgtaa caaggtagcc gtaccggaag 2880ggattgggga acaccctccc
tcggcagctt ccaccctagc cgctaaccct gcttcagcat 2940tgttccatcg gcatggcctt
cc 296293042DNARhodococcus
opacus 9tcaacggaga gtttgatcct ggctcaggac gaacgctggc ggcgtgctta acacatgcaa
60gtcgagcggt aaggcccttc ggggtacacg agcggcgaac agttgcctct caaactagga
120ccgagtcctg cttgcgaccg ccgcacgaat tgtgtacgtt cagctcgcca ttccgggaag
180ccccatgtgc tcgccgcttg gggtgagtaa cacgtgggtg atctgccctg cacttcggga
240taagcctggg aaactgggtc taataccgga tatgaccttc ggctgcatgg ccgttggtgg
300cccactcatt gtgcacccac tagacgggac gtgaagccct attcggaccc tttgacccag
360attatggcct atactggaag ccgacgtacc ggcaaccacc aaaggtttac tggtgcagga
420tgggcccgcg gcctatcagc ttgttggtgg ggtaatggcc taccaaggcg acgacgggta
480gccgacctga gagggtgacc tttccaaatg accacgtcct acccgggcgc cggatagtcg
540aacaaccacc ccattaccgg atggttccgc tgctgcccat cggctggact ctcccactgg
600ggccacactg ggactgagac acggcccaga ctcctacggg aggcagcagt ggggaatatt
660gcacaatggg cgaaagcctg atgcagcgac gccgcgtgag ccggtgtgac cctgactctg
720tgccgggtct gaggatgccc tccgtcgtca ccccttataa cgtgttaccc gctttcggac
780tacgtcgctg cggcgcactc ggatgacggc cttcgggttg taaacctctt tcagcaggga
840cgaagcgaaa gtgacggtac ctgcagaaga agcaccggcc aactacgtgc cagcagccgc
900cctactgccg gaagcccaac atttggagaa agtcgtccct gcttcgcttt cactgccatg
960gacgtcttct tcgtggccgg ttgatgcacg gtcgtcggcg ggtaatacgt agggtgcaag
1020cgttgtccgg aattactggg cgtaaagagc tcgtaggcgg tttgtcgcgt cgtctgtgaa
1080aactcgaggc tcaacctcga ccattatgca tcccacgttc gcaacaggcc ttaatgaccc
1140gcatttctcg agcatccgcc aaacagcgca gcagacactt ttgagctccg agttggagct
1200gcttgcaggc gatacgggca gacttgagta ctgcagggga gactggaatt cctggtgtag
1260cggtgaaatg cgcagatatc aggaggaaca ccggtggcga cgaacgtccg ctatgcccgt
1320ctgaactcat gacgtcccct ctgaccttaa ggaccacatc gccactttac gcgtctatag
1380tcctccttgt ggccaccgct aggcgggtct ctgggcagta actgacgctg aggagcgaaa
1440gcgtgggtag cgaacaggat tagataccct ggtagtccac gccgtaaacg gtgggcgcta
1500tccgcccaga gacccgtcat tgactgcgac tcctcgcttt cgcacccatc gcttgtccta
1560atctatggga ccatcaggtg cggcatttgc cacccgcgat ggtgtgggtt tccttccacg
1620ggatccgtgc cgtagctaac gcattaagcg ccccgcctgg ggagtacggc cgcaaggcta
1680aaactcaaag gaattgacgg ccacacccaa aggaaggtgc cctaggcacg gcatcgattg
1740cgtaattcgc ggggcggacc cctcatgccg gcgttccgat tttgagtttc cttaactgcc
1800gggcccgcac aagcggcgga gcatgtggat taattcgatg caacgcgaag aaccttacct
1860gggtttgaca tataccggaa agctgcagag atgtggcccc cccgggcgtg ttcgccgcct
1920cgtacaccta attaagctac gttgcgcttc ttggaatgga cccaaactgt atatggcctt
1980tcgacgtctc tacaccgggg ccttgtggtc ggtatacagg tggtgcatgg ctgtcgtcag
2040ctcgtgtcgt gagatgttgg gttaagtccc gcaacgagcg caacccttgt cttatgttgc
2100ggaacaccag ccatatgtcc accacgtacc gacagcagtc gagcacagca ctctacaacc
2160caattcaggg cgttgctcgc gttgggaaca gaatacaacg cagcacgtaa tggtggggac
2220tcgtaagaga ctgccggggt caactcggag gaaggtgggg acgacgtcaa gtcatcatgc
2280cccttatgtc cagggcttca gtcgtgcatt accacccctg agcattctct gacggcccca
2340gttgagcctc cttccacccc tgctgcagtt cagtagtacg gggaatacag gtcccgaagt
2400cacatgctac aatggccggt acagagggct gcgataccgt gaggtggagc gaatccctta
2460aagccggtct cagttcggat cggggtctgc aactcgaccc gtgtacgatg ttaccggcca
2520tgtctcccga cgctatggca ctccacctcg cttagggaat ttcggccaga gtcaagccta
2580gccccagacg ttgagctggg cgtgaagtcg gagtcgctag taatcgcaga tcagcaacgc
2640tgcggtgaat acgttcccgg gccttgtaca caccgcccgt cacgtcatga aagtcggtaa
2700gcacttcagc ctcagcgatc attagcgtct agtcgttgcg acgccactta tgcaagggcc
2760cggaacatgt gtggcgggca gtgcagtact ttcagccatt cacccgaagc cggtggccta
2820acccctcgtg ggagggagcc gtcgaaggtg ggatcggcga ttgggacgaa gtcgtaacaa
2880ggtagccgta ccggaaggtg gtgggcttcg gccaccggat tggggagcac cctccctcgg
2940cagcttccac cctagccgct aaccctgctt cagcattgtt ccatcggcat ggccttccac
3000cggctggatc acctcctttc tgccgaccta gtggaggaaa ga
3042102924DNARalstonia sp. 10acgtggcggc atgccttaca catgcaagtc gaacggcagc
gcggacttcg gtctggcggc 60gagtggcgaa cgggtgagta atacatcgga acgtaccctg
tgcaccgccg tacggaatgt 120gtacgttcag cttgccgtcg cgcctgaagc cagaccgccg
ctcaccgctt gcccactcat 180tatgtagcct tgcatgggac ttgtggggga taactagtcg
aaagattagc taataccgca 240tacgacctga gggtgaaagt gggggaccgc aaggcctcac
gcagcaggag cggccgatgt 300aacaccccct attgatcagc tttctaatcg attatggcgt
atgctggact cccactttca 360ccccctggcg ttccggagtg cgtcgtcctc gccggctaca
ctgattagct agttggtggg 420gtaaaggccc accaaggcga cgatcagtag ctggtctgag
aggacgatca gccacactgg 480gactgagaca cggcccagac gactaatcga tcaaccaccc
catttccggg tggttccgct 540gctagtcatc gaccagactc tcctgctagt cggtgtgacc
ctgactctgt gccgggtctg 600tcctacggga ggcagcagtg gggaattttg gacaatgggg
gcaaccctga tccagcaatg 660ccgcgtgtgt gaagaaggcc ttcgggttgt aaagcacttt
aggatgccct ccgtcgtcac 720cccttaaaac ctgttacccc cgttgggact aggtcgttac
ggcgcacaca cttcttccgg 780aagcccaaca tttcgtgaaa tgtccggaaa gaaatcgcgc
tggttaatac ctgcgtgatg 840acggtaccgg aagaataagc accggctaac tacgtgccag
cagccgcggt aatacgtagg 900acaggccttt ctttagcgcg accaattatg gacgcactac
tgccatggcc ttcttattcg 960tggccgattg atgcacggtc gtcggcgcca ttatgcatcc
gtgcgagcgt taatcggaat 1020tactgggcgt aaagcgtgcg caggcggttt tgtaagacag
gcgtgaaatc cccgggctta 1080acctgggaat tgcgcttgtg cacgctcgca attagcctta
atgacccgca tttcgcacgc 1140gtccgccaaa acattctgtc cgcactttag gggcccgaat
tggaccctta acgcgaacac 1200actgcaaggc tagagtgcgt cagagggggg tagaattcca
cgtgtagcag tgaaatgcgt 1260agagatgtgg aggaataccg atggcgaagg cgagccccct
tgacgttccg atctcacgca 1320gtctcccccc atcttaaggt gcacatcgtc actttacgca
tctctacacc tccttatggc 1380taccgcttcc gctcggggga ggaccttgac tgacgctcat
gcacgaaagc gtggggagca 1440aacaggatta gataccctgg tagtccacgc cctaaacgat
gtcaactagt tgttgggatt 1500cctggaactg actgcgagta cgtgctttcg cacccctcgt
ttgtcctaat ctatgggacc 1560atcaggtgcg ggatttgcta cagttgatca acaaccctaa
cattttctca gtaacgtagc 1620taacgcgtga agttgaccgc ctggggagta cggctgcaag
attaaaactc aaaggaattg 1680acggggaccc gcacaagcgg gtaaaagagt cattgcatcg
attgcgcact tcaactggcg 1740gacccctcat gccgacgttc taattttgag tttccttaac
tgcccctggg cgtgttcgcc 1800tggatgatgt ggattaattc gatgcaacgc gaaaaacctt
acctaccctt gacatgccct 1860aacgaagcag agatgcatta gtgcccgcaa agggaaagtg
acctactaca cctaattaag 1920ctacgttgcg ctttttggaa tggatgggaa ctgtacggga
ttgcttcgtc tctacgtaat 1980cacgggcgtt tccctttcac ggacacaggt gctgcatggc
tgtcgtcagc tcgtgtcgtg 2040agatgttggg ttaagtcccg caacgagcgc aacccttgtc
tctagttgcc tacgcaagag 2100cctgtgtcca cgacgtaccg acagcagtcg agcacagcac
tctacaaccc aattcagggc 2160gttgctcgcg ttgggaacag agatcaacgg atgcgttctc
cactctagag agactgccgg 2220tgacaaaccg gaggaaggtg gggatgacgt caagtcctca
tggcccttat gggtagggct 2280tcacacgtca tacaatggtg gtgagatctc tctgacggcc
actgtttggc ctccttccac 2340ccctactgca gttcaggagt accgggaata cccatcccga
agtgtgcagt atgttaccac 2400cgtacagagg gttgccaacc cgcgaggggg agctaatccc
agaaaacgca tcgtagtccg 2460gatcgtagtc tgcaactcga ctacgtgaag ctggaatcgc
gcatgtctcc caacggttgg 2520gcgctccccc tcgattaggg tcttttgcgt agcatcaggc
ctagcatcag acgttgagct 2580gatgcacttc gaccttagcg tagtaatcgc ggatcagcat
gccgcggtga atacgttccc 2640gggtcttgta cacaccgccc gtcacaccat gggagtgggt
tttgccagaa gtagttagcc 2700atcattagcg cctagtcgta cggcgccact tatgcaaggg
cccagaacat gtgtggcggg 2760cagtgtggta ccctcaccca aaacggtctt catcaatcgg
taaccgcaag gagggcgatt 2820accacggcag ggttcatgac tggggtgaag tcgtaacaag
gtattggcgt tcctcccgct 2880aatggtgccg tcccaagtac tgaccccact tcagcattgt
tcca 292411284PRTRhodococcus opacus 11Met Ala Ser Ile
Glu Asp Ile Leu Glu Leu Glu Ala Leu Glu Lys Asp1 5
10 15Ile Phe Arg Gly Ala Val His Pro Ser Val
Leu Lys Arg Thr Phe Gly 20 25
30Gly Gln Val Ala Gly Gln Ser Leu Val Ser Ala Val Arg Thr Val Asp
35 40 45Glu Arg Phe Glu Val His Ser Leu
His Gly Tyr Phe Leu Arg Pro Gly 50 55
60Asn Pro Thr Glu Pro Thr Val Tyr Leu Val Asp Arg Ile Arg Asp Gly65
70 75 80Arg Ser Phe Cys Thr
Arg Arg Val Thr Gly Ile Gln Asp Gly Lys Ala 85
90 95Ile Phe Thr Met Ser Ala Ser Phe His Ser Gln
Asp Glu Gly Ile Glu 100 105
110His Gln Asp Thr Met Pro Ser Val Pro Glu Pro Glu Glu Leu Val Asp
115 120 125Ala Gln Thr Val Glu Glu Met
Ala Ala Thr Asp Leu Tyr Arg Glu Trp 130 135
140Lys Glu Trp Asp Val Arg Ile Val Pro Ala Gly Cys Thr Gly Lys
Thr145 150 155 160Pro Gly
Ile Ala Ala Lys Gln Arg Val Trp Met Arg Tyr Arg Asn Lys
165 170 175Leu Pro Asp Asp Gln Val Phe
His Ile Cys Thr Leu Ala Tyr Leu Ser 180 185
190Asp Met Thr Leu Leu Gly Ala Ser Lys Val Pro His Pro Gly
Val Val 195 200 205Thr Gln Thr Ala
Ser Leu Asp His Ala Met Trp Phe Leu Arg Pro Phe 210
215 220Arg Ala Asp Glu Trp Leu Leu Tyr Asp Gln Thr Ser
Pro Ser Ala Gly225 230 235
240Phe Gly Arg Ala Leu Thr Gln Gly Arg Met Phe Asp Arg Lys Gly Thr
245 250 255Met Val Ala Ala Val
Val Gln Glu Gly Leu Thr Arg Ile Gln Arg Asp 260
265 270Gln Asp Gln Arg Asp Ile Glu Thr Gly Asn Met Ala
275 280121460DNARhodococcus
opacusmodified_base(799)..(799)a, c, t or g 12ggcggcgtgc ttaacacatg
caagtcgagc ggtaaggccc ttcggggtac acgagcggcg 60aacgggtgag taacacgtgg
gtgatctgcc ctgcacttcg ggataagcct gggaaactgg 120gtctaatacc ggatatgacc
ttcggctgca tggctgaggg tggaaaggtt tactggtgca 180ggatgggccc gcggcctatc
agcttgttgg tggggtaatg gcctaccaag gcgacgacgg 240gtagccgacc tgagagggtg
accggccaca ctgggactga gacacggccc agactcctac 300gggaggcagc agtggggaat
attgcacaat gggcgaaagc ctgatgcagc gacgccgcgt 360gagggatgac ggccttcggg
ttgtaaacct ctttcagcag ggacgaagcg aaagtgacgg 420tacctgcaga agaagcaccg
gccaactacg tgccagcagc cgcggtaata cgtagggtgc 480aagcgttgtc cggaattact
gggcgtaaag agctcgtagg cggtttgtcg cgtcgtctgt 540gaaaactcac agctcaacct
cgagcttgca ggcgatacgg gcagacttga gtactgcagg 600ggagactgga attcctggtg
tagcggtgaa atgcgcagat atcaggagga acaccggtgg 660cgaaggcggg tctctgggca
gtaactgacg ctgaggagcg aaagcgtggg tagcaaacag 720gattagatac cctggtagtc
cacgccgtaa acggtgggcg ctaggtgtgg gtttccttcc 780acgggatccg tgccgtagnt
aacgcattaa gcgccccgcc tggggagtac ggccgcaagg 840ttaaaactca aaggaattga
cgggggcccg cacaagcggc ggagcatgtg gattaattcg 900atgcaacgcg aagaacctta
cctgggtttg acatataccg gaaagccgta gagataccgc 960cccccttgtg gtcggtatac
aggtggtgca tggctgtcgt cagctcgtgt cgtgagatgt 1020tgggttaagt cccgcaacga
gcgcaaccct tgtcttatgt tgccagcacg taatggtggg 1080gactcgtaag agactgccgg
ggtcaactcg gaggaaggtg gggacgacgt caagtcatca 1140tgccccttat gtccagggct
tcacacatgc tacaatggcc ggtacagagg gctgcgatac 1200cgtgaggtgg agcgaatccc
ttaaagccgg tctcagttcg gatcggggtc tgcaactcga 1260ccccgtgaag tcggagtcgc
tagtaatcgc agatcagcaa cgctgcggtg aatacgttcc 1320cgggccttgt acacaccgcc
cgtcacgtca tgaaagtcgg taacacccga agccggtggc 1380ctaacccctt gtgggaggga
gccgtcgaag gtgggatcgg cgattgggac gaagtcgtaa 1440caaggtagcc gtaccggaag
1460131473DNARhodococcus
opacusmodified_base(562)..(562)a, c, t or g 13aggacgaacg ctggcggcgt
gcttaacaca tgcaagtcga gcggtaaggc ccttcggggt 60acacgagcgg cgaacgggtg
agtaacacgt gggtgatctg ccctgcactt cgggataagc 120ctgggaaact gggtctaata
ccggatatga ccttcggctg catggctgag ggtggaaagg 180tttactggtg caggatgggc
ccgcggccta tcagcttgtt ggtggggtaa tggcctacca 240agccgacgac gggtagccga
cctgagaggg tgaccggcca cactgggact gagacacggc 300ccagactcct acgggaggca
gcagtgggga atattgcaca atgggcgaaa gcctgatgca 360gcgacgccgc gtgagggatg
acggccttcg ggttgtaaac ctctttcagc agggacgaag 420cgaaagtgac ggtacctgca
gaagaagcac cggccaacta cgtgccagca gccgcggtaa 480tacgtagggt gcaagcgttg
tccggaatta ctgggcgtaa agagctcgta ggcggtttgt 540cgcgtcgtct gtgaaaactc
anagctcaac ctcgagcttg caggcgatac gggcagactt 600gagtactgca ggggagactg
gaattcctgg tgtagcggtg aaatgcgcag atatcaggag 660gaacaccggt ggcgaaggcg
ggtctctggg cagtaactga cgctgaggag cgaaagcgtg 720ggtagcaaac aggattagat
accctggtag tccacgccgt aaacggtggg cgctaggtgt 780gggtttcctt ccacgggatc
cgtgccgtag ctaacgcatt aagcgccccg cctggggagt 840acggccgcaa ggctaaaact
caaaggaatt gacgggggcc cgcacaagcg gcggagcatg 900tggattaatt cgatgcaacg
cgaagaacct tacctgggtt tgacatatac cggaaagccg 960tagagatacg gccccccttg
tggtcggtat acaggtggtg catggctgtc gtcagctcgt 1020gtcgtgagat gttgggttaa
gtcccgcaac gagcgcaacc cttgtcttat gttgccagca 1080cgtaatggtg gggactcgta
agagactgcc ggggtcaact cggaggaagg tggggacgac 1140gtcaagtcat catgcccctt
atgtccaggg cttcacacat gctacaatgg ccggtacaga 1200gggctgcgat accgtgaggt
ggagcgaatc ccttaaagcc ggtctcagtt cggatcgggg 1260tctgcaactc gaccccgtga
agtcggagtc gctagtaatc gcagatcagc aacgctgcgg 1320tgaatacgtt cccgggcctt
gtacacaccg cccgtcacgt catgaaagtc ggtaacaccc 1380gaagccggtg gcctaacccc
ttgtgggagg gagccgtcga aggtgggatc ggcgattggg 1440acgaagtcgt aacaaggtag
ccgtaccgga agg 1473141462DNACupriavidus
necator 14acgtggcggc atgccttaca catgcaagtc gaacggcagc gcggacttcg
gtctggcggc 60gagtggcgaa cgggtgagta atacatcgga acgtaccctg ttgtggggga
taactagtcg 120aaagattagc taataccgca tacgacctga gggtgaaagt gggggaccgc
aaggcctcac 180gcagcaggag cggccgatgt ctgattagct agttggtggg gtaaaggccc
accaaggcga 240cgatcagtag ctggtctgag aggacgatca gccacactgg gactgagaca
cggcccagac 300tcctacggga ggcagcagtg gggaattttg gacaatgggg gcaaccctga
tccagcaatg 360ccgcgtgtgt gaagaaggcc ttcgggttgt aaagcacttt tgtccggaaa
gaaatcgcgc 420tggttaatac ctgcgtgatg acggtaccgg aagaataagc accggctaac
tacgtgccag 480cagccgcggt aatacgtagg gtgcgagcgt taatcggaat tactgggcgt
aaagcgtgcg 540caggcggttt tgtaagacag gcgtgaaatc cccgggctta acctgggaat
tgcgcttgtg 600actgcaaggc tagagtgcgt cagagggggg tagaattcca cgtgtagcag
tgaaatgcgt 660agagatgtgg aggaataccg atggcgaagg cgagccccct ggaccttgac
tgacgctcat 720gcacgaaagc gtggggagca aacaggatta gataccctgg tagtccacgc
cctaaacgat 780gtcaactagt tgttgggatt cattttctca gtaacgtagc taacgcgtga
agttgaccgc 840ctggggagta cggctgcaag attaaaactc aaaggaattg acggggaccc
gcacaagcgg 900tggatgatgt ggattaattc gatgcaacgc gaaaaacctt acctaccctt
gacatgccct 960aacgaagcag agatgcatta gtgcccgcaa agggaaagtg ggacacaggt
gctgcatggc 1020tgtcgtcagc tcgtgtcgtg agatgttggg ttaagtcccg caacgagcgc
aacccttgtc 1080tctagttgcc tacgcaagag cactctagag agactgccgg tgacaaaccg
gaggaaggtg 1140gggatgacgt caagtcctca tggcccttat gggtagggct tcacacgtca
tacaatggtg 1200cgtacagagg gttgccaacc cgcgaggggg agctaatccc agaaaacgca
tcgtagtccg 1260gatcgtagtc tgcaactcga ctacgtgaag ctggaatcgc tagtaatcgc
ggatcagcat 1320gccgcggtga atacgttccc gggtcttgta cacaccgccc gtcacaccat
gggagtgggt 1380tttgccagaa gtagttagcc taaccgcaag gagggcgatt accacggcag
ggttcatgac 1440tggggtgaag tcgtaacaag gt
1462151479DNARalstonia sp. 15agtttgatcc tggctcagat tgaacgctgg
cggcatgcct tacacatgca agtcgaacgg 60cagcgcggac ttcggtctgg cggcgagtgg
cgaacgggtg agtaatacat cggaacgtac 120cctgttgtgg gggataacta gtcgaaagat
tagctaatac cgcatacgac ctgagggtga 180aagcggggga ccgtaaggcc tcgcgcagca
ggagcggccg atgtctgatt agctagttgg 240tggggtaaag gcccaccaag gcgacgatca
gtagctggtc tgagaggacg atcagccaca 300ctgggactga gacacggccc agactcctac
gggaggcagc agtggggaat tttggacaat 360gggggcaacc ctgatccagc aatgccgcgt
gtgtgaagaa ggccttcggg ttgtaaagca 420cttttgtccg gaaagaaaac gctctggtta
atacctggag tggatgacgg taccggaaga 480ataagcaccg gctaactacg tgccagcagc
cgcggtaata cgtagggtgc gagcgttaat 540cggaattact gggcgtaaag cgtgcgcagg
cggttttgta agacaggcgt gaaatccccg 600agctcaactt gggaattgcg cttgtgactg
caaggctaga gtatgtcaga ggggggtaga 660attccacgtg tagcagtgaa atgcgtagag
atgtggagga ataccgatgg cgaaggcagc 720cccctgggac gtcactgacg ctcatgcacg
aaagcgtggg gagcaaacag gattagatac 780cctggtagtc cacgccctaa acgatgtcaa
ctagttgttg gggattcatt tcttcagtaa 840cgtagctaac gcgtgaagtt gaccgcctgg
ggagtacggt cgcaagatta aaactcaaag 900gaattgacgg ggacccgcac aagcggtgga
tgatgtggat taattcgatg caacgcgaaa 960aaccttacct acccttgaca tgccactaac
gaagcagaga tgcatcaggt gcccgaaagg 1020gaaagtggac acaggtgctg catggctgtc
gtcagctcgt gtcgtgagat gttgggttaa 1080gtcccgcaac gagcgcaacc cttatcttta
gttgctacgc aagggcactc tagagagact 1140gccggtgaca aaccggagga aggtggggat
gacgtcaagt cctcatggcc cttatgggta 1200gggcttcaca cgtcatacaa tggtgcgtac
agagggttgc caacccgcga gggggagcta 1260atcccagaaa acgcatcgta gtccggatcg
cagtctgcaa ctcgactgcg tgaagctgga 1320atcgctagta atcgcggatc agcatgccgc
ggtgaatacg ttcccgggtc ttgtacacac 1380cgcccgtcac accatgggag tgggttttgc
cagaagtagt tagcctaacc gcaaggaggg 1440cgattaccac ggcagggttc atgactgggg
tgaagtcgt 1479161486DNAGordonia alkanivorans
16gctcaggacg aacgctggcg gcgtgcttaa cacatgcaag tcgaacggaa aggcccagct
60tgctgggtac tcgagtggcg aacgggtgag taacacgtgg gtgatctgcc ctgaactttg
120ggataagcct gggaaactgg gtctaatacc ggatatgacc ttggagtgca tgctctgggg
180tggaaagctt ttgcggttca ggatgggccc gcggcctatc agcttgttgg tggggtaatg
240gcctaccaag gcgacgacgg gtagccgacc tgagagggtg atcggccaca ctgggactga
300gacacggccc agactcctac gggaggcagc agtggggaat attgcacaat gggcgcaagc
360ctgatgcagc gacgccgcgt gagggatgac ggccttcggg ttgtaaacct ctttcaccag
420ggacgaagcg caagtgacgg tacctggaga agaagcaccg gccaactacg tgccagcagc
480cgcggtaata cgtagggtgc gagcgttgtc cggaattact gggcgtaaag agctcgtagg
540cggtttgtcg cgtcgtctgt gaaattctgc aactcaattg taggcgtgca ggcgatacgg
600gcagacttga gtactacagg ggagactgga attcctggtg tagcggtgaa atgcgcagat
660atcaggagga acaccggtgg cgaaggcggg tctctgggta gtaactgacg ctgaggagcg
720aaagcgtggg tagcgaacag gattagatac cctggtagtc cacgccgtaa acggtgggta
780ctaggtgtgg ggctcatttc acgagttccg tgccgtagct aacgcattaa gtaccccgcc
840tggggagtac ggccgcaagg ctaaaactca aaggaattga cgggggcccg cacaagcggc
900ggagcatgtg gattaattcg atgcaacgcg aagaacctta cctgggtttg acatacacca
960gacgcatgta gagatacatg ttcccttgtg gttggtgtac aggtggtgca tggctgtcgt
1020cagctcgtgt cgtgagatgt tgggttaagt cccgcaacga gcgcaaccct tgtcctgtat
1080tgccagcggg ttatgccggg gacttgcagg agactgccgg ggtcaactcg gaggaaggtg
1140gggatgacgt caagtcatca tgccccttat gtccagggct tcacacatgc tacaatggct
1200ggtacagagg gctgcgatac cgtgaggtgg agcgaatccc ttaaagccag tctcagttcg
1260gattggggtc tgcaactcga ccccatgaag tcggagtcgc tagtaatcgc agatcagcaa
1320cgctgcggtg aatacgttcc cgggccttgt acacaccgcc cgtcacgtca tgaaagtcgg
1380taacacccga agccggtggc ctaacccctt gtgggaggga gctgtcgaag gtgggatcgg
1440cgattgggac gaagtcgtaa caaggtagcc gtaccggaag gtgcgg
1486171491DNAGordonia sp 17gatcatggct caggacgaac gctggcggcg tgcttaacac
atgcaagtcg aacggaaagg 60cccgcttgcg ggtactcgag tggcgaacgg gtgagtaaca
cgtgggtgat ctgccctgga 120ctctgggata agcctgggaa actgggtcta ataccggata
tgaccttaca tcgcatggtg 180tttggtggaa agcttttgcg gttcaggatg ggcccgcggc
ctatcagctt gttggtgggg 240taatggccta ccaaggcgac gacgggtagc cgacctgaga
gggtgatcgg ccacactggg 300actgagacac ggcccagact cctacgggag gcagcagtgg
ggaatattgc acaatgggcg 360caagcctgat gcagcgacgc cgcgtgaggg atgacggcct
tcgggttgta aacctctttc 420accagggacg aagcgcaagt gacggtacct ggagaagaag
caccggccaa ctacgtgcca 480gcagccgcgg taatacgtag ggtgcgagcg ttgtccggaa
ttactgggcg taaagagctc 540gtaggcggtt tgtcgcgtcg tctgtgaaat tctgcaactc
aattgtaggc gtgcaggcga 600tacgggcaga cttgagtact acaggggaga ctggaattcc
tggtgtagcg gtgaaatgcg 660cagatatcag gaggaacacc ggtggcgaag gcgggtctct
gggtagtaac tgacgctgag 720gagcgaaagc gtgggtagcg aacaggatta gataccctgg
tagtccacgc cgtaaacggt 780gggtactagg tgtggggctc atttcacgag ttccgtgccg
tagctaacgc attaagtacc 840ccgcctgggg agtacggccg caaggctaaa actcaaagga
attgacgggg gcccgcacaa 900gcggcggagc atgtggatta attcgatgca acgcgaagaa
ccttacctgg gtttgacata 960caccagaaag ctatagagat atagcccccc ttgtggttgg
tgtacaggtg gtgcatggct 1020gtcgtcagct cgtgtcgtga gatgttgggt taagtcccgc
aacgagcgca acccttgtcc 1080tgtattgcca gcgggttatg ccggggactt gcaggagact
gccggggtca actcggagga 1140aggtggggat gacgtcaagt catcatgccc cttatgtcca
gggcttcaca catgctacaa 1200tggctggtac agagggctgc gataccgtga ggtggagcga
atcccttaaa gccagtctca 1260gttcggattg gggtctgcaa ctcgacccca tgaagtcgga
gtcgctagta atcgcagatc 1320agcaacgctg cggtgaatac gttcccgggc cttgtacaca
ccgcccgtca cgtcatgaaa 1380gtcggtaaca cccgaagccg gtggcctaac cccttgtggg
agggagctgt cgaaggtggg 1440atcggcgatt gggacgaagt cgtaacaagg tagccgtacc
ggaaggtgcg g 1491181505DNAMycobacterium fortuitum 18ttgatcctgg
ctcaggacga acgctggcgg cgtgcttaac acatgcaagt cgaacggaaa 60ggcccttcgg
ggtactcgag tggcgaacgg gtgagtaaca cgtgggtgat ctgccctgca 120ctttgggata
agcctgggaa actgggtcta ataccgaata tgaccacgcg cttcatggtg 180tgtggtggaa
agcttttgcg gtgtgggatg ggcccgcggc ctatcagctt gttggtgggg 240taatggccta
ccaaggcgac gacgggtagc cggcctgaga gggtgaccgg ccacactggg 300actgagatac
ggcccagact cctacgggag gcagcagtgg ggaatattgc acaatgggcg 360caagcctgat
gcagcgacgc cgcgtgaggg atgacggcct tcgggttgta aacctctttc 420aatagggacg
aagcgcaagt gacggtacct atagaagaag gaccggccaa ctacgtgcca 480gcagccgcgg
taatacgtag ggtccgagcg ttgtccggaa ttactgggcg taaagagctc 540gtaggtggtt
tgtcgcgttg ttcgtgaaaa ctcacagctt aactgtgggc gtgcgggcga 600tacgggcaga
ctagagtact gcaggggaga ctggaattcc tggtgtagcg gtggaatgcg 660cagatatcag
gaggaacacc ggtggcgaag gcgggtctct gggcagtaac tgacgctgag 720gagcgaaagc
gtggggagcg aacaggatta gataccctgg tagtccacgc cgtaaacggt 780gggtactagg
tgtgggtttc cttccttggg atccgtgccg tagctaacgc attaagtacc 840ccgcctgggg
agtacggccg caaggctaaa actcaaagga attgacgggg gcccgcacaa 900gcggcggagc
atgtggatta attcgatgca acgcgaagaa ccttacctgg gtttgacatg 960cacaggacga
ctgcagagat gtggtttccc ttgtggcctg tgtgcaggtg gtgcatggct 1020gtcgtcagct
cgtgtcgtga gatgttgggt taagtcccgc aacgagcgca acccttgtct 1080catgttgcca
gcacgttatg gtggggactc gtgagagact gccggggtca actcggagga 1140aggtggggat
gacgtcaagt catcatgccc cttatgtcca gggcttcaca catgctacaa 1200tggccggtac
aaagggctgc gatgccgtga ggtggagcga atcctttcaa agccggtctc 1260agttcggatc
ggggtctgca actcgacccc gtgaagtcgg agtcgctagt aatcgcagat 1320cagcaacgct
gcggtgaata cgttcccggg ccttgtacac accgcccgtc acgtcatgaa 1380agtcggtaac
acccgaagcc ggtggcctaa cccttgtgga gggagccgtc gaaggtggga 1440tcggcgattg
ggacgaagtc gtaacaaggt agccgtaccg gaaggtgcgg ctggatcacc 1500tcctt
1505191460DNAMycobacterium parafortuitum 19cgaacgctgg cggcgtgctt
aacacatgca agtcgaacgg aaaggccctt cggggtactc 60gagtggcgaa cgggtgagta
acacgtgggt gatctgccct gcactttggg ataagcctgg 120gaaactgggt ctaataccga
atatgatcat tggcttcctg gctggtggtg gaaagctttt 180gcggtgtggg atgggcccgc
ggcctatcag cttgttggtg gggtaatggc ctaccaaggc 240gacgacgggt agccggcctg
agagggtgac cggccacact gggactgaga tacggcccag 300actcctacgg gaggcagcag
tggggaatat tgcacaatgg gcgcaagcct gatgcagcga 360cgccgcgtga gggatgacgg
ccttcgggtt gtaaacctct ttcgccaggg acgaagcgca 420agtgacggta cctggagaag
aaggaccggc caactacgtg ccagcagccg cggtaatacg 480tagggtccga gcgttgtccg
gaattactgg gcgtaaagag ctcgtaggtg gtttgtcgcg 540ttgttcgtga aaactcacag
cttaactgtg ggcgtgcggg cgatacgggc agactagagt 600actgcagggg agactggaat
tcctggtgta gcggtggaat gcgcagatat caggaggaac 660accggtggcg aaggcgggtc
tctgggcagt aactgacgct gaggagcgaa agcgtgggga 720gcgaacagga ttagataccc
tggtagtcca cgccgtaaac ggtgggtact aggtgtgggt 780ttccttcctt gggatccgtg
ccgtagctaa cgcattaagt accccgcctg gggagtacgg 840ccgcaaggct aaaactcaaa
gaaattgacg ggggcccgca caagcggcgg agcatgtgga 900ttaattcgat gcaacgcgaa
gaaccttacc tgggtttgac atgcacagga cgccggcaga 960gatgtcggtt cccttgtggc
ctgtgtgcag gtggtgcatg gctgtcgtca gctcgtgtcg 1020tgagatgttg ggttaagtcc
cgcaacgagc gcaacccttg tctcatgttg ccagcacgta 1080atggtgggga ctcgtgagag
actgccgggg tcaactcgga ggaaggtggg gatgacgtca 1140agtcatcatg ccccttatgt
ccagggcttc acacatgcta caatggccgg tacaaagggc 1200tgcgatgccg tgaggtggag
cgaatccttt caaagccggt ctcagttcgg atcggggtct 1260gcaactcgac cccgtgaagt
cggagtcgct agtaatcgca gatcagcaac gctgcggtga 1320atacgttccc gggccttgta
cacaccgccc gtcacgtcat gaaagtcggt aacacccgaa 1380gccggtggcc taaccccttg
tgggagggag ccgtcgaagg tgggatcggc gattgggacg 1440aagtcgtaac aaggtagccg
1460201480DNAMycobacterium
sphagni 20gagtttgatc ctggctcagg acgaacgctg gcggcgtgct taacacatgc
aagtcgaacg 60gaaaggccct tcggggtact cgagtggcga acgggtgagt aacacgtggg
tgatctgccc 120tgcactttgg gataagcctg ggaaactggg tctaataccg aataggaccg
catgcttcat 180ggtgtgtggt ggaaagcttt tgcggtgtgg gatgggcccg cggcctatca
gcttgttggt 240ggggtaatgg cctaccaagg cgacgacggg tagccggcct gagagggtgt
ccggccacac 300tgggactgag atacggccca gactcctacg ggaggcagca gtggggaata
ttgcacaatg 360ggcgcaagcc tgatgcagcg acgccgcgtg agggatgacg gccttcgggt
tgtaaacctc 420tttcagcagg gacgaagcgc aagtgacggt acctgtagaa gaagcaccgg
ccaactacgt 480gccagcagcc gcggtaatac gtagggtgcg agcgttgtcc ggaattactg
ggcgtaaaga 540gctcgtaggt ggtttgtcgc gttgttcgtg aaaactcaca gctcaactgt
gggcgtgcgg 600gcgatacggg cagacttgag tactgcaggg gagactggaa ttcctggtgt
agcggtggaa 660tgcgcagata tcaggaggaa caccggtggc gaaggcgggt ctctgggcag
taactgacgc 720tgaggagcga aagcgtgggg agcgaacagg attagatacc ctggtagtcc
acgccgtaaa 780cggtgggtac taggtgtggg tttccttcct tgggatccgt gccgtagcta
acgcattaag 840taccccgcct ggggagtacg gccgcaaggc taaaactcaa agaaattgac
gggggcccgc 900acaagcggcg gagcatgtgg attaattcga tgcaacgcga agaaccttac
ctgggtttga 960catgcacagg acgccggcag agatgtcggt tcccttgtgg cctgtgtgca
ggtggtgcat 1020ggctgtcgtc agctcgtgtc gtgagatgtt gggttaagtc ccgcaacgag
cgcaaccctt 1080gtctcatgtt gccagcacgt aatggtgggg actcgtgaga gactgccggg
gtcaactcgg 1140aggaaggtgg ggatgacgtc aagtcatcat gccccttatg tccagggctt
cacacatgct 1200acaatggccg gtacaaaggg ctgcgatgcc gtgaggtgga gcgaatcctt
tcaaagccgg 1260tctcagttcg gatcggggtc tgcaactcga ccccgtgaag tcggagtcgc
tagtaatcgc 1320agatcagcaa cgctgcggtg aatacgttcc cgggccttgt acacaccgcc
cgtcacgtca 1380tgaaagtcgg taacacccga agccggtggc ctaacccctt gtgggaggga
gccgtcgaag 1440gtgggatcgg cgattgggac gaagtcgtaa caaggtagcc
1480211494DNANocardia farcinica 21gacgaacgct ggcggcgtgc
ttaacacatg caagtcgagc ggtaaggccc ttcggggtac 60acgagcggcg aacgggtgag
taacacgtgg gtgatctgcc ctgtacttcg ggataagcct 120gggaaactgg gtctaatacc
ggatatgacc ttacatcgca tggtgtttgg tggaaagatt 180tatcggtaca ggatgggccc
gcggcctatc agcttgttgg tggggtaatg gcctaccaag 240gcgacgacgg gtagccggcc
tgagagggcg accggccaca ctgggactga gacacggccc 300agactcctac gggaggcagc
agtggggaat attgcacaat gggcgaaagc ctgatgcagc 360gacgccgcgt gagggatgac
ggccttcggg ttgtaaacct ctttcgacag ggacgaagcg 420caagtgacgg tacctgtaga
agaagcaccg gccaactacg tgccagcagc cgcggtaata 480cgtagggtgc gagcgttgtc
cggaattact gggcgtaaag agcttgtagg cggtttgtcg 540cgtcgtccgt gaaaacttgg
ggctcaaccc caagcttgcg ggcgatacgg gcagacttga 600gtactgcagg ggagactgga
attcctggtg tagcggtgaa atgcgcagat atcaggagga 660acaccggtgg cgaaggcggg
tctctgggca gtaactgacg ctgagaagcg aaagcgtggg 720tagcgaacag gattagatac
cctggtagtc cacgccgtaa acggtgggcg ctaggtgtgg 780gtttccttcc acgggatccg
tgccgtagct aacgcattaa gcgccccgcc tggggagtac 840ggccgcaagg ctaaaactca
aaggaattga cgggggcccg cacaagcggc ggagcatgtg 900gattaattcg atgcaacgcg
aagaacctta cctgggtttg acatacaccg gaaacctgca 960gagatgtagg cccccttgtg
gtcggtgtac aggtggtgca tggctgtcgt cagctcgtgt 1020cgtgagatgt tgggttaagt
cccgcaacga gcgcaaccct tgtcctgtgt tgccagcgcg 1080ttatggcggg gactcgcagg
agactgccgg ggtcaactcg gaggaaggtg gggacgacgt 1140caagtcatca tgccccttat
gtccagggct tcacacatgc tacaatggcc ggtacagagg 1200gctgcgatac cgtgaggtgg
agcgaatccc ttaaagccgg tctcagttcg gatcggggtc 1260tgcaactcga ccccgtgaag
ttggagtcgc tagtaatcgc agatcagcaa cgctgcggtg 1320aatacgttcc cgggccttgt
acacaccgcc cgtcacgtca tgaaagtcgg taacacccga 1380agccggtggc ctaacccctt
gtgggaggga gccgtcgaag gtgggatcgg cgattgggac 1440gaagtcgtaa caaggtagcc
gtaccggaag gtgcggctgg atcacctcct ttct 1494221513DNANocardia sp.
22gagtttgatc ctggctcagg acgaacgctg gcggcgtgct taacacatgc aagtcgagcg
60gtaaggccct tcggggtaca cgagcggcga acgggtgagt aacacgtggg tgatctgccc
120tgtacttcgg gataagcctg ggaaactggg tctaataccg gatatgacct tacatcgcat
180ggtgtttggt ggaaagattt atcggtacag gatgggcccg cggcctatca gcttgttggt
240ggggtaatgg cctaccaagg cgacgacggg tagccggcct gagagggcga ccggccacac
300tgggactgag acacggccca gactcctacg ggaggcagca gtggggaata ttgcacaatg
360ggcgaaagcc tgatgcagcg acgccgcgtg agggatgacg gccttcgggt tgtaaacctc
420tttcgacagg gacgaagcgc aagtgacggt acctgtagaa gaagcaccgg ccaactacgt
480gccagcagcc gcggtaatac gtagggtgcg agcgttgtcc ggaattactg ggcgtaaaga
540gcttgtaggc ggtttgtcgc gtcgtccgtg aaaacttggg gctcaacccc aagcttgcgg
600gcgatacggg cagacttgag tactgcaggg gagactggaa ttcctggtgt agcggtgaaa
660tgcgcagata tcaggaggaa caccggtggc gaaggcgggt ctctgggcag taaccgacgc
720tgagaagcga aagcgtgggt agcgaacagg attagatacc ctggtagtcc acgccgtaaa
780cggtgggcgc taggtgtggg tttccttcca cgggatccgt gccgtagcta acgcattaag
840cgccccgcct ggggagtacg gccgcaaggc taaaactcaa aggaattgac gggggcccgc
900acaagcggcg gagcatgtgg attaattcga tgcaacgcga agaaccttac ctgggtttga
960catacaccgg aaacctgcag agatgtaggc ccccttgtgg tcggtgtaca ggtggtgcat
1020ggccgtcgtc agctcgtgtc gtgagatgtt gggttaagtc ccgcaacgag cgcaaccctt
1080gtcctgtgtt gccagcgcgt tatggcgggg actcgcagga gactgccggg gtcaactcgg
1140aggaaggtgg ggacgacgtc aagtcatcat gccccttatg tccagggctt cacacatgct
1200acaatggccg gtacagaggg ctgcgatacc gtgaggtgga gcgaatccct taaagccggt
1260ctcagttcgg atcggggtct gcaactcgac cccgtgaagt tggagtcgct agtaatcgca
1320gatcagcaac gctgcggtga atacgttccc gggccttgta cacaccgccc gtcacgtcat
1380gaaagtcggt aacacccgaa gccggtggcc taaccccttg tgggagggag ccgtcgaagg
1440tgggatcggc gattgggacg aagtcgtaac aaggtagccg taccggaagg tgcggctgga
1500tcacctcctt tct
1513231490DNARhodococcus rhodochrous 23gagtttgaat ctggctcagg acgaacgctg
gcggcgtgct taacacatgc aagtcgaacg 60atgaagccca gcttgctggg tggattagtg
gcgaacgggt gagtaacacg tgggtgatct 120gccctgcact ctgggataag cctgggaaac
tgggtctaat accggatatg acctcttgct 180gcatggcgag gggtggaaag tttttcggtg
caggatgagc ccgcggccta tcagcttgtt 240ggtggggtaa tggcctacca aggcgacgac
gggtagccgg cctgagaggg cgaccggcca 300cactgggact gagacacggc ccagactcct
acgggaggca gcagtgggga atattgcaca 360atgggcgaaa gcctgatgca gcgacgccgc
gtgagggatg acggccttcg ggttgtaaac 420ctctttcagc agggacgaag cgaaagtgac
ggtacctgca gaagaagcac cggccaacta 480cgtgccagca gccgcggtaa tacgtagggt
gcgagcgttg tccggaatta ctgggcgtaa 540agagctcgta ggcggtttgt cgcgtcgtct
gtgaaatccc gcagctcaac tgcgggcttg 600caggcgatac gggcagactc gagtactgca
ggggagactg gaattcctgg tgtagcggtg 660aaatgcgcag atatcaggag gaacaccggt
ggcgaaggcg ggtctctggg cagtaactga 720cgctgaggag cgaaagcgtg ggtagcgaac
aggattagat accctggtag tccacgccgt 780aaacggtggg cgctaggtgt gggtttcctt
ccacgggatc cgtgccgtag ccaacgcatt 840aagcgccccg cctggggagt acggccgcaa
ggctaaaact caaaggaatt gacgggggcc 900cgcacaagcg gcggagcatg tggattaatt
cgatgcaacg cgaagaacct tacctgggtt 960tgacatgtac cggacgactg cagagatgtg
gtttcccttg tggccggtag acaggtggtg 1020catggctgtc gtcagctcgt gtcgtgagat
gttgggttaa gtcccgcaac gagcgcaacc 1080cttgtcctgt gttgccagca cgtaatggtg
gggactcgca ggagactgcc ggggtcaact 1140cggaggaagg tggggacgac gtcaagtcat
catgcccctt atgtccaggg cttcacacat 1200gctacaatgg tcggtacaga gggctgcgat
accgtgaggt ggagcgaatc ccttaaagcc 1260ggtctcagtt cggatcgggg tctgcaactc
gaccccgtga agtcggagtc gctagtaatc 1320gcagatcagc aacgctgcgg tgaatacgtt
cccgggcctt gtacacaccg cccgtcacgt 1380catgaaagtc ggtaacaccc gaagccggtg
gcctaacccc ttgtgggagg gagccgtcga 1440aggtgggatc ggcgattggg acgaagtcgt
aacaaggtag ccgtaccgga 1490241481DNARhodococcus coprophilus
24cctggctcag gacgaacgct ggcggcgtgc ttaacacatg caagtcgaac gatgatgccc
60agcttgctgg gcggattagt ggcgaacggg tgagtaacac gtgggtgatc tgccctgcac
120ttcgggataa gcctgggaaa ctgggtctaa taccggatat gaccatggga tgcatgtcct
180gtggtggaaa ggtttactgg tgcaggatga gcccgcggcc tatcagcttg ttggtggggt
240aatggcctac caaggcgacg acgggtagcc ggcctgagag ggcgaccggc cacactggga
300ctgagacacg gcccagactc ctacgggagg cagcagtggg gaatattgca caatgggcga
360aagcctgatg cagcgacgcc gcgtgaggga tgacggcctt cgggttgtaa acctctttca
420gcagggacga agcgcaagtg actgtacctg cagaagaagc accggctaac tacgtgccag
480cagccgcggt aatacgtagg gtgcgagcgt tgtccggaat tactgggcgt aaagagttcg
540taggcggttt gtcgcgtcgt gtgtgaaatc ccgcagctca actgcgggct tgcaggcgat
600acgggcagac ttgagtactg caggggagac tggaattcct ggtgtagcgg tgaaatgcgc
660agatatcagg aggaacaccg gtggcgaagg cgggtctctg ggcagtaact gacgctgagg
720aacgaaagcg tgggtagcga acaggattag ataccctggt agtccacgcc gtaaacggtg
780ggcgctaggt gtgggtttcc ttccacggga tccgtgccgt agctaacgca ttaagcgccc
840cgcctgggga gtacggccgc aaggctaaaa ctcaaaggaa ttgacggggg cccgcacaag
900cggcggagca tgtggattaa ttcgatgcaa cgcgaagaac cttacctggg tttgacatat
960accggacgac tgcagagatg tggtttccct tgtggtcggt atacaggtgg tgcatggctg
1020tcgtcagctc gtgtcgtgag atgttgggtt aagtcccgca acgagcgcaa cccttgtctt
1080atgttgccag cacgtaatgg gggggactcg taagagactg ccggggtcaa ctcggaggaa
1140ggtggggacg acgtcaagtc atcatgcccc ttatgtccag ggcttcacac atgctacaat
1200ggtcggtaca gagggctgcg ataccgtgag gtggagcgaa tcccttaaag ccggtctcag
1260ttcggatcgg ggtctgcaac tcgaccccgt gaagtcggag tcgctagtaa tcgcagatca
1320gcaacgctgc ggtgaatacg ttcccgggcc ttgtacacac cgcccgtcac gtcatgaaag
1380tcggtaacac ccgaagccgg tggcctaacc ccttgtggga gggagccgtc gaaggtggga
1440tcggcgattg ggacgaagtc gtaacaaggt agccgtaccg g
1481251486DNARhodococcus triatomae 25ggcggcgtgc ttaacacatg caagtcgagc
ggtaaggcct ttcggggtac acgagcggcg 60aacgggtgag taacacgtgg gtgatctgcc
ctgcactctg ggataagcct gggaaactgg 120gtctaatacc ggatatgact accggctgca
tggtctggtg gtggaaagat ttatcggtgc 180aggatgggcc cgcggcctat cagcttgttg
gtggggtaat ggcctaccaa ggcgacgacg 240ggtagccgac ctgagagggt gaccggccac
actgggactg agacacggcc cagactccta 300cgggaggcag cagtggggaa tattgcacaa
tgggcgaaag cctgatgcag cgacgccgcg 360tgagggatga cggccttcgg gttgtaaacc
tctttcaaca gggacgaagc gcaagtgacg 420gtacctgtag aagaagcacc ggccaactac
gtgccagcag ccgcggtaat acgtagggtg 480cgagcgttgt ccggaattac tgggcgtaaa
gagctcgtag gcggtttgtc gcgtcgtctg 540tgaaaaccag cagctcaact gctggcttgc
aggcgatacg ggcagacttg agtactgcag 600gggagactgg aattcctggt gtagcggtga
aatgcgcaga tatcaggagg aacaccggtg 660gcgaaggcgg gtctctgggc agtaactgac
gctgaggagc gaaagcgtgg gtagcgaaca 720ggattagata ccctggtagt ccacgccgta
aacggtgggc gctaggtgtg ggtttccttc 780cacgggatcc gtgccgtagc taacgcatta
agcgccccgc ctggggagta cggccgcaag 840gctaaaactc aaaggaattg acgggggccc
gcacaagcgg cggagcatgt ggattaattc 900gatgcaacgc gaagaacctt acctgggttt
gacatacacc ggaaagccgt agagatacgg 960ccccccttgt ggtcggtgta caggtggtgc
atggctgtcg tcagctcgtg tcgtgagatg 1020ttgggttaag tcccgcaacg agcgcaaccc
ttgtcctgtg ttgccagcac gtaatggtgg 1080ggactcgcag gagactgccg gggtcaactc
ggaggaaggt ggggacgacg tcaagtcatc 1140atgcccctta tgtccagggc ttcacacatg
ctacaatggc cggtacagag ggctgcgata 1200ccgtgaggtg gagcgaatcc cttaaagccg
gtctcagttc ggatcggggt ctgcaactcg 1260accccgtgaa gtcggagtcg ctagtaatcg
cagatcagca acgctgcggt gaatacgttc 1320ccgggccttg tacacaccgc ccgtcacgtc
atgaaagtcg gtaacacccg aagccggtgg 1380cctaacccct tgtgggaggg agccgtcgaa
ggtgggatcg gcgattggga cgaagtcgta 1440acaaggtagc cgtaccggaa ggtgcggctg
gatcacttcc tttcta 1486261507DNANocardia coeliaca
26tttgatcctg gctcaggacg aacgctggcg gcgtgcttaa cacatgcaag tcgagcggta
60aggcctttcg gggtacacga gcggcgaacg ggtgagtaac acgtgggtga tctgccctgc
120acttcgggat aagcctggga aactgggtct aataccggat atgacctcag gttgcatgac
180ttggggtgga aagatttatc ggtgcaggat gggcccgcgg cctatcagct tgttggtggg
240gtaatggcct accaaggcga cgacgggtag ccgacctgag agggtgaccg gccacactgg
300gactgagaca cggcccagac tcctacggga ggcagcagtg gggaatattg cacaatgggc
360gaaagcctga tgcagcgacg ccgcgtgagg gatgacggcc ttcgggttgt aaacctcttt
420cagcagggac gaagcgcaag tgacggtacc tgcagaagaa gcaccggcta actacgtgcc
480agcagccgcg gtaatacgta gggtgcaagc gttgtccgga attactgggc gtaaagagtt
540cgtaggcggt ttgtcgcgtc gtttgtgaaa accagcagct caactgctgg cttgcaggcg
600atacgggcag acttgagtac tgcaggggag actggaattc ctggtgtagc ggtgaaatgc
660gcagatatca ggaggaacac cggtggcgaa ggcgggtctc tgggcagtaa ctgacgctga
720ggaacgaaag cgtgggtagc gaacaggatt agataccctg gtagtccacg ccgtaaacgg
780tgggcgctag gtgtgggttc cttccacgga atccgtgccg tagctaacgc attaagcgcc
840ccgcctgggg agtacggccg caaggctaaa actcaaagga attgacgggg gcccgcacaa
900gcggcggagc atgtggatta attcgatgca acgcgaagaa ccttacctgg gtttgacata
960taccggaaag ctgcagagat gtggcccccc ttgtggtcgg tatacaggtg gtgcatggct
1020gtcgtcagct cgtgtcgtga gatgttgggt taagtcccgc aacgagcgca acccctatct
1080tatgttgcca gcacgttatg gtggggactc gtaagagact gccggggtca actcggagga
1140aggtggggac gacgtcaagt catcatgccc cttatgtcca gggcttcaca catgctacaa
1200tggccagtac agagggctgc gagaccgtga ggtggagcga atcccttaaa gctggtctca
1260gttcggatcg gggtctgcaa ctcgaccccg tgaagtcgga gtcgctagta atcgcagatc
1320agcaacgctg cggtgaatac gttcccgggc cttgtacaca ccgcccgtca cgtcatgaaa
1380gtcggtaaca cccgaagccg gtggcttaac cccttgtggg agggagccgt cgaaggtggg
1440atcggcgatt gggacgaagt cgtaacaagg tagccgtacc ggaaggtgcg gctggatcac
1500ctccttt
1507271507DNANocardia globerula 27gtttgatcct ggctcaggac gaacgctggc
ggcgtgctta acacatgcaa gtcgagcggt 60aaggcctttc ggggtacacg agcggcgaac
gggtgagtaa cacgtgggtg atctgccctg 120cacttcggga taagcctggg aaactgggtc
taataccgga tatgacctcc tatcgcatgg 180tgggtggtgg aaagatttat cggtgcagga
tgggcccgcg gcctatcagc ttgttggtgg 240ggtaatggcc taccaaggcg acgacgggta
gccgacctga gagggtgacc ggccacactg 300ggactgagac acggcccaga ctcctacggg
aggcagcagt ggggaatatt gcacaatggg 360cgaaagcctg atgcagcgac gccgcgtgag
ggacgacggc cttcgggttg taaacctctt 420tcagcaggga cgaagcgcaa gtgacggtac
ctgcagaaga agcaccggct aactacgtgc 480cagcagccgc ggtaatacgt agggtgcaag
cgttgtccgg aattactggg cgtaaagagt 540tcgtaggcgg tttgtcacgt cgtttgtgaa
aactcacagc tcaactgtga gcctgcaggc 600gatacgggca gacttgagta ctgcagggga
gactggaatt cctggtgtag cggtgaaatg 660cgcagatatc aggaggaaca ccggtggcga
aggcgggtct ctgggcagta actgacgctg 720aggaacgaaa gcgtgggtag cgaacaggat
tagataccct ggtagtccac gccgtaaacg 780gtgggcgcta ggtgtgggtt ccttccacgg
aatccgtgcc gtagctaacg cattaagcgc 840cccgcctggg gagtacggcc gcaaggctaa
aactcaaagg aattgacggg ggcccgcaca 900agcggcggag catgtggatt aattcgatgc
aacgcgaaga accttacctg ggtttgacat 960ataccggaaa gccgtagaga tacggccccc
cttgtggtcg gtatacaggt ggtgcatggc 1020tgtcgtcagc tcgtgtcgtg agatgttggg
ttaagtcccg caacgagcgc aacccctatc 1080ttatgttgcc agcacgttat ggtggggact
cgtaagagac tgccggggtc aactcggagg 1140aaggtgggga cgacgtcaag tcatcatgcc
ccttatgtcc agggcttcac acatgctaca 1200atggccagta cagagggctg cgagaccgtg
aggtggagcg aatcccttaa agctggtctc 1260agttcggatc ggggtctgca actcgacccc
gtgaagtcgg agtcgctagt aatcgcagat 1320cagcaacgct gcggtgaata cgttcccggg
ccttgtacac accgcccgtc acgtcatgaa 1380agtcggtaac acccgaagcc ggtggcttaa
ccccttgtgg gagggagccg tcgaaggtgg 1440gatcggcgat tgggacgaag tcgtaacaag
gtagccgtac cggaaggtgc ggctggatca 1500cctcctt
1507281508DNARhodococcus equi
28gagtttgatc ctggctcagg acgaacgctg gcggcgtgct taacacatgc aagtcgagcg
60gtagggccct tcggggtaca cgagcggcga acgggtgagt aacacgtggg tgatctgccc
120tgcacttcgg gataagcttg ggaaactggg tctaataccg gatatgagcc tctactgcat
180ggtggaggtt ggaaaggttt actggtgcag gatgggcccg cggcctatca gcttgttggt
240ggggtaatgg cctaccaagg cgacgacggg tagccggcct gagagggcga ccggccacac
300tgggactgag acacggccca gactcctacg ggaggcagca gtggggaata ttgcacaatg
360ggcgaaagcc tgatgcagcg acgccgcgtg agggatgacg gccttcgggt tgtaaacctc
420tttcagcagg gacgaagcga gagtgacggt acctgcagaa gaagcaccgg ccaactacgt
480gccagcagcc gcggtaatac gtagggtgcg agcgttgtcc ggaattactg ggcgtaaaga
540gctcgtaggc ggtttgtcgc gtcgtcggtg aaaaccagca gctcaactgc tggcttgcag
600gcgatacggg cagacttgag tactgcaggg gagactggaa ttcctggtgt agcggtgaaa
660tgcgcagata tcaggaggaa caccggtggc gaaggcgggt ctctgggcag taactgacgc
720tgaggagcga aagcgtgggt agcgaacagg attagatacc ctggtagtcc acgccgtaaa
780cggtgggcgc taggtgtggg tttccttcca cgggatccgt gccgtagcta acgcattaag
840cgccccgcct ggggagtacg gccgcaaggc taaaactcaa aggaattgac gggggcccgc
900acaagcggcg gagcatgtgg attaattcga tgcaacgcga agaaccttac ctgggtttga
960catataccgg aaagccgtag agatacggcc ccccttgtgg tcggtataca ggtggtgcat
1020ggctgtcgtc agctcgtgtc gtgagatgtt gggttaagtc ccgcaacgag cgcaaccctt
1080gtcctgtgtt gccagcacgt aatggtgggg actcgcagga gaccgccggg gtcaactcgg
1140aggaaggtgg ggacgacgtc aagtcatcat gccccttatg tccagggctt cacacatgct
1200acaatggccg gtacagaggg ctgcgatacc gtgaggtgga gcgaatccct taaagccggt
1260ctcagttcgg atcggggtct gcaactcgac cccgtgaagt cggagtcgct agtaatcgca
1320gatcagcaac gctgcggtga atacgttccc gggccttgta cacaccgccc gtcacgtcat
1380gaaagtcggt aacacccgaa gccggtggcc taacccttgt ggagggagcc gtcgaaggtg
1440ggatcggcga ttgggacgaa gtcgtaacaa ggtagccgta ccggaaggtg cggctggatc
1500acctcctt
1508291442DNARhodococcus sp. 29ctggctcagg acgaacgctg gcggcgtgct
taacacatgc aagtcgagcg gtaaggccct 60tcggggtaca cgagcggcga acgggtgagt
aacacgtggg tgatctgccc tgcacttcgg 120gataagcctg ggaaactggg tctaataccg
gatatgacct tcggctgcat ggctgagggt 180ggaaaggttt actggtgcag gatgagcccg
cggcctatca gcttgttggt ggggtaatgg 240cctaccaagg cgacgacggg tagccgacct
gagagggtga ccggccacac tgggactgag 300acacggccca gactcctacg ggaggcagca
gtggggaata ttgcacaatg ggcgaaagcc 360tgatgcagcg acgccgcgtg agggatgacg
gccttcgggt tgtaaacctc tttcagcagg 420gacgaagcga aagtgacggt acctgcagaa
gaagcaccgg ctaactacgt gccagcagcc 480gcggtaatac gtagggtgca agcgttgtcc
ggaattactg ggcgtaaaga gttcgtaggc 540ggtttgtcgc gtcgtctgtg aaaactcaca
gctcaactgt gagcttgcag gcgatacggg 600cagacttgag tactgcaggg gagactggaa
ttcctggtgt agcggtgaaa tgcgcagata 660tcaggaggaa caccggtggc gaaggcgggt
ctctgggcag taactgacgc tgaggaacga 720aagcgtgggt agcaaacagg attagatacc
ctggtagtcc acgccgtaaa cggtgggcgc 780taggtgtggg ttccttccac gggatctgtg
ccgtagctaa cgcattaagc gccccgcctg 840gggagtacgg ccgcaaggct aaaactcaaa
ggaattgacg ggggcccgca caagcggcgg 900agcatgtgga ttaattcgat gcaacgcgaa
gaaccttacc tgggtttgac atataccgga 960aagccgtaga gatacggccc cccttgtggt
cggtatacag gtggtgcatg gctgtcgtca 1020gctcgtgtcg tgagatgttg ggttaagtcc
cgcaacgagc gcaacccttg tcttatgttg 1080ccagcacgta atggtgggga ctcgtaagag
actgccgggg tcaactcgga ggaaggtggg 1140gacgacgtca agtcatcatg ccccttatgt
ccagggcttc acacatgcta caatggccag 1200tacagagggc tgcgaaccgt gaggtggagc
gaatccctta aagcyggtct cagttcggat 1260cggggtctgc aactcgaccc cgtgaagtcg
gagtcgctag taatcgcaga tcagcaacgc 1320tgcggtgaat acgttcccgg gccttgtaca
caccgcccgt cacgtcatga aagtcggtaa 1380cacccgaagc cggtggccta accccttgtg
ggagggagcc gtcgaaggtg ggatcggcga 1440tt
1442301474DNARhodococcus sp.
30agagtttgat cctggctcag gacgaacgct ggcggcgtgc ttaacacatg caagtcgagc
60ggtaaggccc ttcggggtac acgagcggcg aacgggtgag taacacgtgg gtgatctgcc
120ctgcacttcg ggataagcct gggaaactgg gtctaatacc ggatatgacc ttcggctgca
180tggctgaggg tggaaaggtt tactggtgca ggatgggccc gcggcctatc agcttgttgg
240tggggtaatg gcctaccaag gcgacgacgg gtagccgacc tgagagggtg accggccaca
300ctgggactga gacacggccc agactcctac gggaggcagc agtggggaat attgcacaat
360gggcgaaagc ctgatgcagc gacgccgcgt gagggatgac ggccttcggg ttgtaaacct
420ctttcagcag ggacgaagcg aaagtgacgg tacctgcaga agaagcaccg gctaactacg
480tgccagcagc cgcggtaata cgtagggtgc aagcgttgtc cggaattact gggcgtaaag
540agttcgtagg cggtttgtcg cgtcgtttgt gaaaactcam rgctcaactg tgagcttgca
600ggcgatacgg gcagacttga gtactgcagg ggagactgga attcctggtg tagcggtgaa
660atgcgcagat atcaggagga acaccggtgg cgaaggcggg tctctgggca gtaactgacg
720ctgaggaacg aaagcgtggg tagcaaacag gattagatac cctggtagtc cacgccgtaa
780acggtgggcg ctaggtgtgg gttccttcca cgggatctgt gccgtagcta acgcattaag
840cgccccgcct ggggagtacg gccgcaaggc taaaactcaa aggaattgac gggggcccgc
900acaagcggcg gagcatgtgg attaattcga tgcaacgcga agaaccttac ctgggtttga
960catataccgg aaagccgtag agatacggcc ccccttgtgg tcggtataca ggtggtgcat
1020ggctgtcgtc agctcgtgtc gtgagatgtt gggttaagtc ccgcaacgag cgcaaccctt
1080gtcttatgtt gccagcacgt aatggtgggg actcgtaaga gactgccggg gtcaactcgg
1140aggaaggtgg ggacgacgtc aagtcatcat gccccttatg tccagggctt cacacatgct
1200acaatggcca gtacagaggg ctgcgagacc gtgaggtgga gcgaatccct taaagctggt
1260ctcagttcgg atcggggtct gcaactcgac cccgtgaagt cggagtcgct agtaatcgca
1320gatcagcaac gctgcggtga atacgttccc gggccttgta cacaccgccc gtcacgtcat
1380gaaagtcggt aacacccgaa gccggtggcc taaccccttg tgggagggag ccgtcgaagg
1440tgggatcggc gattgggacg aagtcgtaac aagg
1474311437DNARhodococcus jostii 31aggacgaacg ctggcggcgt gcttaacaca
tgcaagtcga gcggtaaggc ccttcggggt 60acacgagcgg cgaacgggtg agtaacacgt
gggtgatctg ccctgcactt cgggataagc 120ctgggaaact gggtctaata ccggatatga
ccttcggctg catggctgag ggtggaaagg 180tttactggtg caggatgggc ccgcggccta
tcagcttgtt ggtggggtaa tggcctacca 240aggcgacgac gggtagccga cctgagaggg
tgaccggcca cactgggact gagacacggc 300ccagactcct acgggaggca gcagtgggga
atattgcaca atgggcgaaa gcctgatgca 360gcgacgccgc gtgagggatg acggccttcg
ggttgtaaac ctctttcagc agggacgaag 420cgaaagtgac ggtacctgca gaagaagcac
cggctaacta cgtgccagca gccgcggtaa 480tacgtagggt gcaagcgttg tccggaatta
ctgggcgtaa agagttcgta ggcggtttgt 540cgcgtcgttt gtgaaaactc acagctcaac
tgtgagcctg caggcgatac gggcagactt 600gagtactgca ggggagactg gaattcctgg
tgtagcggtg aaatgcgcag atatcaggag 660gaacaccggt ggcgaaggcg ggtctctggg
cagtaactga cgctgaggaa cgaaagcgtg 720ggtagcaaac aggattagat accctggtag
tccacgccgt aaacggtggg cgctaggtgt 780gggttccttc cacgggatct gtgccgtagc
taacgcatta agcgccccgc ctggggagta 840cggccgcaag gctaaaactc aaaggaattg
acgggggccc gcacaagcgg cggagcatgt 900ggattaattc gatgcaacgc gaagaacctt
acctgggttt gacatatacc ggaaagccgt 960agagatacgg ccccccttgt ggtcggtata
caggtggtgc atggctgtcg tcagctcgtg 1020tcgtgagatg ttgggttaag tcccgcaacg
agcgcaaccc ttgtcttatg ttgccagcac 1080gtaatggtgg ggactcgtaa gagactgccg
gggtcaactc ggaggaaggt ggggacgacg 1140tcaagtcatc atgcccctta tgtccagggc
ttcacacatg ctacaatggc cagtacagag 1200ggctgcgaga ccgtgaggtg gagcgaatcc
cttaaagctg gtctcagttc ggatcggggt 1260ctgcaactcg accccgtgaa gtcggagtcg
ctagtaatcg cagatcagca acgctgcggt 1320gaatacgttc ccgggccttg tacacaccgc
ccgtcacgtc atgaaagtcg gtaacacccg 1380aagccggtgg cctaacccct tgtgggaggg
agccgtcgaa ggtgggatcg gcgattg 1437321485DNARhodococcus
opacusmodified_base(812)..(812)a, c, t or g 32gatcctggct caggacgaac
gctggcggcg tgcttaacac atgcaagtcg agcggtaagg 60cccttcgggg tacacgagcg
gcgaacgggt gagtaacacg tgggtgatct gccctgcact 120tcgggataag cctgggaaac
tgggtctaat accggatatg accttcggct gcatggctga 180gggtggaaag gtttactggt
gcaggatggg cccgcggcct atcagcttgt tggtggggta 240atggcctacc aaggcgacga
cgggtagccg acctgagagg gtgaccggcc acactgggac 300tgagacacgg cccagactcc
tacgggaggc agcagtgggg aatattgcac aatgggcgaa 360agcctgatgc agcgacgccg
cgtgagggat gacggccttc gggttgtaaa cctctttcag 420cagggacgaa gcgaaagtga
cggtacctgc agaagaagca ccggccaact acgtgccagc 480agccgcggta atacgtaggg
tgcaagcgtt gtccggaatt actgggcgta aagagttcgt 540aggcggtttg tcgcgtcgtc
tgtgaaaact caaagctcaa cctcgagcct gcaggcgata 600cgggcagact tgagtactgc
aggggagact ggaattcctg gtgtagcggt gaaatgcgca 660gatatcagga ggaacaccgg
tggcgaaggc gggtctctgg gcagtaactg acgctgagga 720acgaaagcgt gggtagcgaa
caggattaga taccctggta gtccacgccg taaacggtgg 780gcgctaggtg tgggtttcct
tccacgggat cngtgccgta gctaacgcat taagcgcccc 840gcctggggag tacggccgca
aggctaaaac tcaaaggaat tgacgggggc ccgcacaagc 900ggcggagcat gtggattaat
tcgatgcaac gcgaagaacc ttacctgggt ttgacatata 960ccggaaagcc gtagagatac
ggcccccctt gtggtcggta tacaggtggt gcatggctgt 1020cgtcagctcg tgtcgtgaga
tgttgggtta agtcccgcaa cgagcgcaac ccttgtctta 1080tgttgccagc acgtaatggt
ggggactcgt aagagactgc cggggtcaac tcggaggaag 1140gtggggacga cgtcaagtca
tcatgcccct tatgtccagg gcttcacaca tgctacaatg 1200gccggtacag agggctgcga
taccgtgagg tggagcgaat cccttaaagc tggtctcagt 1260tcggatcggg gtctgcaact
cgaccccgtg aagtcggagt cgctagtaat cgcagatcag 1320caacgctgcg gtgaatacgt
tcccgggcct tgtacacacc gcccgtcacg tcatgaaagt 1380cggtaacacc cgaagccggt
ggcctaaccc cttgtgggag ggagccgtcg aaggtgggat 1440cggcgattgg gacgaagtcg
taacaaggta gccgtaccgg aaggt 1485331510DNARhodococcus
imtechensis 33ttgatcctgg ctcaggacga acgctggcgg cgtgcttaac acatgcaagt
cgagcggtaa 60ggcccttcgg ggtacacgag cggcgaacgg gtgagtaaca cgtgggtgat
ctgccctgca 120cttcgggata agcctgggaa actgggtcta ataccggata tgaccttcgg
ctgcatggct 180gagggtggaa aggtttactg gtgcaggatg ggcccgcggc ctatcagctt
gttggtgggg 240taatggccta ccaaggcgac gacgggtagc cgacctgaga gggtgaccgg
ccacactggg 300actgagacac ggcccagact cctacgggag gcagcagtgg ggaatattgc
acaatgggcg 360aaagcctgat gcagcgacgc cgcgtgaggg atgacggcct tcgggttgta
aacctctttc 420agcagggacg aagcgaaagt gacggtacct gcagaagaag caccggccaa
ctacgtgcca 480tcagccgcgg taatacgtag ggtgcaagcg ttgtccggaa ttactgggcg
taaagagctc 540gtaggcggtt tgtcgtgtcg tctgtgaaaa ctcgaggctc aacctcgagc
ttgcaggcga 600tacgggcaga cttgagtact gcaggggaga ctggaattcc tggtgtagcg
gtgaaatgcg 660cagatatcag gaggaacacc ggtggcgaag gcgggtctct gggcagtaac
tgacgctgag 720gagcgaaagc gtggaaaccg aacaggatta gataccctgg tagtccacgc
cgtaaacggt 780gggcgctagg tgtgggtttc cttccacggg atccgtgccg tagctaacgc
attaagcgcc 840ccgcctgggg agtacggccg caaggctaaa actcaaagga attgacgggg
gcccgcacaa 900gcggcggagc atgtggatta attcgatgca acgcgaagaa ccttacctgg
gtttgacata 960taccggaaag ccgtagagat acggcccccc ttgtggtcgg tatacaggtg
gtgcatggct 1020gtcgtcagct cgtgtcgtaa gatgttgggt taagtcccgc aacgagcgca
acccttgtct 1080tatgttgcca gcacgtaatg gtggggactc gtaagagact gccggggtca
actcggagga 1140aggtggggac gacgtcaagt catcatgccc cttatgtcca gggcttcaca
catgctacaa 1200tggccagtac agagggctgc gagaccgtga ggtggagcga atcccttaaa
gctggtctca 1260gttcggatcg gggtctgcaa ctcgaccccg tgaagtcgga gtcgctagta
atcgcagatc 1320agcaacgctg cggtgaatac gttcccaggc cttgtacaca ccgcccgtca
cgtcatgaaa 1380gtcggtaaca cccgaagccg gtggcctaac cccttgtggg agggagccgt
cgaaggtggg 1440atcggcgatt gggacgaagt cgtaacaagg tagccgtacc ggaaggtgcg
gctggaaact 1500gccgaggggg
1510341473DNARhodococcus koreensis 34gacgaacgct ggcggcgtgc
ttaacacatg caagtcgagc ggtaaggccc ttcggggtac 60acgagcggcg aacgggtgag
taacacgtgg gtgatctgcc ctgcacttcg ggataagcct 120gggaaactgg gtctaatacc
ggatatgacc aaggactgca tggtttttgg tggaaaggtt 180tactggtgca ggatgggccc
gcggcctatc agcttgttgg tggggtaatg gcctaccaag 240gcgacgacgg gtagccgacc
tgagagggtg accggccaca ctgggactga gacacggccc 300agactcctac gggaggcagc
agtggggaat attgcacaat gggcgaaagc ctgatgcagc 360gacgccgcgt gagggatgac
ggccttcggg ttgtaaacct ctttcagcag ggacgaagcg 420agagtgacgg tacctgcaga
agaagcaccg gccaactacg tgccagcagc cgcggtaata 480cgtagggtgc aagcgttgtc
cggaattact gggcgtaaag agctcgtagg cggtttgtcg 540cgtcgtctgt gaaaactcga
ggctcaacct cgagcttgca ggcgatacgg gcagacttga 600gtactgcagg ggagactgga
attcctggtg tagcggtgaa atgcgcagat atcaggagga 660acaccggtgg cgaaggcggg
tctctgggca gtaactgacg ctgaggagcg aaagcgtggg 720tagcgaacag gattagatac
cctggtagtc cacgccgtaa acggtgggcg ctaggtgtgg 780gttccttcca cgggatccgt
gccgtagcta acgcattaag cgccccgcct ggggagtacg 840gccgcaaggc taaaactcaa
aggaattgac gggggcccgc acaagcggcg gagcatgtgg 900attaattcga tgcaacgcga
agaaccttac ctgggtttga catataccgg aaagccgtag 960agatacggcc ccccttgtgg
tcggtataca ggtggtgcat ggctgtcgtc agctcgtgtc 1020gtgagatgtt gggttaagtc
ccgcaacgag cgcaaccctt gtcttatgtt gccagcacgt 1080aatggtgggg actcgtaaga
gactgccggg gtcaactcgg aggaaggtgg ggacgacgtc 1140aagtcatcat gccccttatg
tccagggctt cacacatgct acaatggcca gtacagaggg 1200ctgcgagacc gtgaggtgga
gcgaatccct taaagctggt ctcagttcgg atcggggtct 1260gcaactcgac cccgtgaagt
cggagtcgct agtaatcgca gatcagcaac gctgcggtga 1320atacgttccc gggccttgta
cacaccgccc gtcacgtcat gaaagtcggt aacacccgaa 1380gccggtggcc taaccccttg
tgggagggag ccgtcgaagg tgggatcggc gattgggacg 1440aagtcgtaac aaggtagccg
taccggaagg tgc 1473351450DNARhodococcus
sp. 35tcctggctca ggacgaacgc tggcggcgtg cttaacacat gcaagtcgag cggtaaggcc
60cttcggggta cacgagcggc gaacgggtga gtaacacgtg ggtgatctgc cctgcacttc
120gggataagcc tgggaaactg ggtctaatac cggatatgac cttcggctgc atggctgttg
180gtggaaaggt ttactggtgc aggatgggcc cgcggcctat cagcttgttg gtggggtaat
240ggcctaccaa ggcgacgacg ggtagccgac ctgagagggt gaccggccac actgggactg
300agacacggcc cagactccta cgggaggcag cagtggggaa tattgcacaa tgggcgaaag
360cctgatgcag cgacgccgcg tgagggatga cggccttcgg gttgtaaacc tctttcagca
420gggacgaagc gagagtgacg gtacctgcag aagaagcacc ggccaactac gtgccagcag
480ccgcggtaat acgtagggtg caagcgttgt ccggaattac tgggcgtaaa gagctcgtag
540gcggtttgtc gcgtcgtctg tgaaaactcg aggctcaacc tcgagcttgc aggcgatacg
600ggcagacttg agtactgcag gggagactgg aattcctggt gtagcggtga aatgcgcaga
660tatcaggagg aacaccggtg gcgaaggcgg gtctctgggc agtaactgac gctgaggagc
720gaaagcgtgg gtagcgaaca ggattagata ccctggtagt ccacgccgta aacggtgggc
780gctaggtgtg ggtttccttc cacgggatcc gtgccgtagc taacgcatta agcgccccgc
840ctggggagta cggccgcaag gctaaaactc aaaggaattg acgggggccc gcacaagcgg
900cggagcatgt ggattaattc gatgcaacgc gaagaacctt acctgggttt gacatatacc
960ggaaagccgt agagatacgg ccccccttgt ggtcggtata caggtggtgc atggctgtcg
1020tcagctcgtg tcgtgagatg ttgggttaag tcccgcaacg agcgcaaccc ttgtcttatg
1080ttgccagcac gtaatggtgg ggactcgtaa gagactgccg gggtcaactc ggaggaaggt
1140ggggacgacg tcaagtcatc atgcccctta tgtccagggc ttcacacatg ctacaatggc
1200cggtacagag ggctgcgata ccgtgaggtg gagcgaatcc cttaaagccg gtctcagttc
1260ggatcggggt ctgcaactcg accccgtgaa gtcggagtcg ctagtaatcg cagatcagca
1320acgctgcggt gaatacgttc ccgggccttg tacacaccgc ccgtcacgtc atgaaagtcg
1380gtaacacccg aagccggtgg cctaacccct cgtgggaggg agccgtcgaa ggtgggatcg
1440gcgattggga
1450361471DNARhodococcus opacus 36gacgaacgct ggcggcgtgc ttaacacatg
caagtcgagc ggtaaggccc ttcggggtac 60acgagcggcg aacgggtgag taacacgtgg
gtgatctgcc ctgcacttcg ggataagcct 120gggaaactgg gtctaatacc ggatatgacc
ttcggctgca tggctgaggg tggaaaggtt 180tactggtgca ggatgggccc gcggcctatc
agcttgttgg tggggtaatg gcctaccaag 240gcgacgacgg gtagccgacc tgagagggtg
accggccaca ctgggactga gacacggccc 300agactcctac gggaggcagc agtggggaat
attgcacaat gggcgaaagc ctgatgcagc 360gacgccgcgt gagggatgac ggccttcggg
ttgtaaacct ctttcagcag ggacgaagcg 420agagtgacgg tacctgcaga agaagcaccg
gccaactacg tgccagcagc cgcggtaata 480cgtagggtgc aagcgttgtc cggaattact
gggcgtaaag agctcgtagg cggtttgtcg 540cgtcgtctgt gaaaactcga ggctcaacct
cgagcttgca ggcgatacgg gcagacttga 600gtactgcagg ggagactgga attcctggtg
tagcggtgaa atgcgcagat atcaggagga 660acaccggtgg cgaaggcggg tctctgggca
gtaactgacg ctgaggagcg aaagcgtggg 720tagcgaacag gattagatac cctggtagtc
cacgccgtaa acggtgggcg ctaggtgtgg 780gtttccttcc acgggatccg tgccgtagct
aacgcattaa gcgccccgcc tggggagtac 840ggccgcaagg ctaaaactca aaggaattga
cgggggcccg cacaagcggc ggagcatgtg 900gattaattcg atgcaacgcg aagaacctta
cctgggtttg acatataccg gaaagccgta 960gagatacggc cccccttgtg gtcggtatac
aggtggtgca tggctgtcgt cagctcgtgt 1020cgtgagatgt tgggttaagt cccgcaacga
gcgcaaccct tgtcttatgt tgccagcacg 1080taatggtggg gactcgtaag agactgccgg
ggtcaactcg gaggaaggtg gggacgacgt 1140caagtcatca tgccccttat gtccagggct
tcacacatgc tacaatggcc ggtacagagg 1200gctgcgatac cgtgaggtgg agcgaatccc
ttaaagccgg tctcagttcg gatcggggtc 1260tgcaactcga ccccgtgaag tcggagtcgc
tagtaatcgc agatcagcaa cgctgcggtg 1320aatacgttcc cgggccttgt acacaccgcc
cgtcacgtca tgaaagtcgg taacacccga 1380agccggtggc ctaacccctc gtgggaggga
gccgtcgaag gtgggatcgg cgattgggac 1440gaagtcgtaa caaggtagcc gtaccggaag g
1471371482DNARhodococcus sp.
37gagtttgatc ctggctcagg acgaacgctg gcggcgtgct taacacatgc aagtcgagcg
60gtaaggccct tcggggtaca cgagcggcga acgggtgagt aacacgtggg tgatctgccc
120tgcacttcgg gataagcctg ggaaactggg tctaataccg gatatgacct tcggctgcat
180ggctgagggt ggaaaggttt actggtgcag gatgggcccg cggcctatca gcttgttggt
240ggggtaatgg cctaccaagg cgacgacggg tagccgacct gagagggtga ccggccacac
300tgggactgag acacggccca gactcctacg ggaggcagca gtggggaata ttgcacaatg
360ggcgaaagcc tgatgcagcg acgccgcgtg agggatgacg accttcgggt tgtaaacctc
420tttcagcagg gacgaagcga aagtgacggt acctgcagaa gaagcaccgg ccaactacgt
480gccagcagcc gcggtaatac gtagggtgca agcgttgtcc ggaattactg ggcgtaaaga
540gctcgtaggc ggtttgtcgc gtcgtctgtg aaaactcgag gctcaacctc gagcttgcag
600gcgatacggg cagacttgag tactgcaggg gagactggaa ttcctggtgt agcggtgaaa
660tgcgcagata tcaggaggaa caccggtggc gaaggcgggt ctctgggcag taactgacgc
720tgaggagcga aagcgtgggt agcgaacagg attagatacc ctggtagtcc acgccgtaaa
780cggtgggcgc taggtgtggg tttccttcca cgggatccgt gccgtagcta acgcattaag
840cgccccgcct ggggagtacg gccgcaaggc taaaactcaa aggaattgac gggggcccgc
900acaagcggcg gagcatgtgg attaattcga tgcaacgcga agaaccttac ctgggtttga
960catataccgg aaagccgtag agatacggcc ccccttgtgg tcggtataca ggtggtgcat
1020ggctgtcgtc agctcgtgtc gtgagatgtt gggttaagtc ccgcaacgag cgcaaccctt
1080gtcttatgtt gccagcacgt aatggtgggg actcgtaaga gactgccggg gtcaactcgg
1140aggaaggtgg ggacgacgtc aagtcatcat gccccttatg tccagggctt cacacatgct
1200acaatggccg gtacagaggg ctgcgatacc gtgaggtgga gcgaatccct taaagccggt
1260ctcagttcgg atcggggtct gcaactcgac cccgtgaagt cggagtcgct agtaatcgca
1320gatcagcaac gctgcggtga atacgttccc gggccttgta cacaccgccc gtcacgtcat
1380gaaagtcggt aacacccgaa gccggtggcc taaccccttg tgggagggag ccgtcgaagg
1440tgggatcggc gattgggacg aagtcgtaac aaggtagccg ta
1482381446DNARhodococcus sp. 38gcggcgtgct taacacatgc aagtcgagcg
gtaaggccct tcggggtaca cgagcggcga 60acgggtgagt aacacgtggg tgatctgccc
tgcacttcgg gataagcctg ggaaactggg 120tctaataccg gatatgacct tcggctgcat
ggctgagggt ggaaaggttt actggtgcag 180gatgggcccg cggcctatca gcttgttggt
ggggtaatgg cctaccaagg cgacgacggg 240tagccgacct gagagggtga ccggccacac
tgggactgag acacggccca gactcctacg 300ggaggcagca gtggggaata ttgcacaatg
ggcgaaagcc tgatgcagcg acgccgcgtg 360agggatgacg gccttcgggt tgtaaacctc
tttcagcagg gacgaagcga aagtgacggt 420acctgcagaa gaagcaccgg ccaactacgt
gccagcagcc gcggtaatac gtagggtgca 480agcgttgtcc ggaattactg ggcgtaaaga
gctcgtaggc ggtttgtcgc gtcgtctgtg 540aaaactcgag gctcaacctc gagcttgcag
gcgatacggg cagacttgag tactgcaggg 600gagactggaa ttcctggtgt agcggtgaaa
tgcgcagata tcaggaggaa caccggtggc 660gaaggcgggt ctctgggcag taactgacgc
tgaggggcga aagcgtgggt agcgaacagg 720attagatacc ctggtagtcc acgccgtaaa
cggtgggcgc taggtgtggg tttccttcca 780cgggatccgt gccgtagcta acgcattaag
cgccccgcct ggggagtacg gccgcaaggc 840taaaactcaa aggaattgac gggggcccgc
acaagcggcg gagcatgtgg attaattcga 900tgcaacgcga agaaccttac ctgggtttga
catataccgg aaagccgtag agatacggcc 960ccccttgtgg tcggtataca ggtggtgcat
ggctgtcgtc agctcgtgtc gtgagatgtt 1020gggttaagtc ccgcaacgag cgcaaccctt
gtcttatgtt gccagcacgt aatggtgggg 1080actcgtaaga gactgccggg gtcaactcgg
aggaaggtgg ggacgacgtc aagtcatcat 1140gccccttatg tccagggctt cacacatgct
acaatggccg gtacagaggg ctgcgatacc 1200gtgaggtgga gcgaatccct taaagccggt
ctcagttcgg atcggggtct gcaactcgac 1260cccgtgaagt cggagtcgct agtaatcgca
gatcagcaac gctgcggtga atacgttccc 1320gggccttgta cacaccgccc gtcacgtcat
gaaagtcggt aacacccgaa gccagtggcc 1380taaccccttg tgggagggag ccgtcgaagg
tgggatcggc gattgggacg aagtcgtaac 1440aaggta
1446391497DNARhodococcus wratislaviensis
39cctggctcag gacgaacgct ggcggcgtgc ttaacacatg caagtcgagc ggtaaggccc
60ttcggggtac acgagcggcg aacgggtgag taacacgtgg gtgatctgcc ctgcacttcg
120ggataagcct gggaaactgg gtctaatacc ggatatgacc ttcggctgca tggctgaggg
180tggaaaggtt tactggtgca ggatgggccc gcggcctatc agcttgttgg tggggtaatg
240gcctaccaag gcgacgacgg gtagccgacc tgagagggtg accggccaca ctgggactga
300gacacggccc agactcctac gggaggcagc agtggggaat attgcacaat gggcgaaagc
360ctgatgcagc gacgccgcgt gagggatgac ggccttcggg ttgtaaacct ctttcagcag
420ggacgaagcg aaagtgacgg tacctgcaga agaagcaccg gccaactacg tgccagcagc
480cgcggtaata cgtagggtgc aagcgttgtc cggaattact gggcgtaaag agctcgtagg
540cggtttgtcg cgtcgtctgt gaaaactcga ggctcaacct cgagcttgca ggcgatacgg
600gcagacttga gtactgcagg ggagactgga attcctggtg tagcggtgaa atgcgcagat
660atcaggagga acaccggtgg cgaaggcggg tctctgggca gtaactgacg ctgaggagcg
720aaagcgtggg tagcgaacag gattagatac cctggtagtc cacgccgtaa acggtgggcg
780ctaggtgtgg gtttccttcc acgggatccg tgccgtagct aacgcattaa gcgccccgcc
840tggggagtac ggccgcaagg ctaaaactca aaggaattga cgggggcccg cacaagcggc
900ggagcatgtg gattaattcg atgcaacgcg aagaacctta cctgggtttg acatataccg
960gaaagccgta gagatacggc cccccttgtg gtcggtatac aggtggtgca tggctgtcgt
1020cagctcgtgt cgtgagatgt tgggttaagt cccgcaacga gcgcaaccct tgtcttatgt
1080tgccagcacg taatggtggg gactcgtaag agactgccgg ggtcaactcg gaggaaggtg
1140gggacgacgt caagtcatca tgccccttat gtccagggct tcacacatgc tacaatggcc
1200ggtacagagg gctgcgatac cgtgaggtgg agcgaatccc ttaaagccgg tctcagttcg
1260gatcggggtc tgcaactcga ccccgtgaag tcggagtcgc tagtaatcgc agatcagcaa
1320cgctgcggtg aatacgttcc cgggccttgt acacaccgcc cgtcacgtca tgaaagtcgg
1380taacacccga agccggtggc ctaacccctt gtgggaggga gccgtcgaag gtgggatcgg
1440cgattgggac gaagtcgtaa caaggtagcc gtaccggaag gtgcggctgg atcacct
1497401481DNARhodococcus opacusmodified_base(570)..(570)a, c, t, g,
unknown or other 40cctggctcag gacgaacgct ggcggcgtgc ttaacacatg caagtcgagc
ggtaaggccc 60ttcggggtac acgagcggcg aacgggtgag taacacgtgg gtgatctgcc
ctgcacttcg 120ggataagcct gggaaactgg gtctaatacc ggatatgacc ttcggctgca
tggctgaggg 180tggaaaggtt tactggtgca ggatgggccc gcggcctatc agcttgttgg
tggggtaatg 240gcctaccaag gcgacgacgg gtagccgacc tgagagggtg accggccaca
ctgggactga 300gacacggccc agactcctac gggaggcagc agtggggaat attgcacaat
gggcgaaagc 360ctgatgcagc gacgccgcgt gagggatgac ggccttcggg ttgtaaacct
ctttcagcag 420ggacgaagcg aaagtgacgg tacctgcaga agaagcaccg gccaactacg
tgccagcagc 480cgcggtaata cgtagggtgc aagcgttgtc cggaattact gggcgtaaag
agctcgtagg 540cggtttgtcg cgtcgtctgt gaaaactcan agctcaacct cgagcttgca
ggcgatacgg 600gcagacttga gtactgcagg ggagactgga attcctggtg tagcggtgaa
atgcgcagat 660atcaggagga acaccggtgg cgaaggcggg tctctgggca gtaactgacg
ctgaggagcg 720aaagcgtggg tagcaaacag gattagatac cctggtagtc cacgccgtaa
acggtgggcg 780ctaggtgtgg gtttccttcc acgggatccg tgccgtagtt aacgcattaa
gcgccccgcc 840tggggagtac ggccgcaagg ttaaaactca aaggaattga cgggggcccg
cacaagcggc 900ggagcatgtg gattaattcg atgcaacgcg aagaacctta cctgggtttg
acatataccg 960gaaagccgta gagataccgc cccccttgtg gtcggtatac aggtggtgca
tggctgtcgt 1020cagctcgtgt cgtgagatgt tgggttaagt cccgcaacga gcgcaaccct
tgtcttatgt 1080tgccagcacg taatggtggg gactcgtaag agactgccgg ggtcaactcg
gaggaaggtg 1140gggacgacgt caagtcatca tgccccttat gtccagggct tcacacatgc
tacaatggcc 1200ggtacagagg gctgcgatac cgtgaggtgg agcgaatccc ttaaagccgg
tctcagttcg 1260gatcggggtc tgcaactcga ccccgtgaag tcggagtcgc tagtaatcgc
agatcagcaa 1320cgctgcggtg aatacgttcc cgggccttgt acacaccgcc cgtcacgtca
tgaaagtcgg 1380taacacccga agccggtggc ctaacccctt gtgggaggga gccgtcgaag
gtgggatcgg 1440cgattgggac gaagtcgtaa caaggtagcc gtaccggaag g
1481411521DNARhodococcus opacus 41tcaacggaga gtttgatcct
ggctcaggac gaacgctggc ggcgtgctta acacatgcaa 60gtcgagcggt aaggcccttc
ggggtacacg agcggcgaac gggtgagtaa cacgtgggtg 120atctgccctg cacttcggga
taagcctggg aaactgggtc taataccgga tatgaccttc 180ggctgcatgg ccgttggtgg
aaaggtttac tggtgcagga tgggcccgcg gcctatcagc 240ttgttggtgg ggtaatggcc
taccaaggcg acgacgggta gccgacctga gagggtgacc 300ggccacactg ggactgagac
acggcccaga ctcctacggg aggcagcagt ggggaatatt 360gcacaatggg cgaaagcctg
atgcagcgac gccgcgtgag ggatgacggc cttcgggttg 420taaacctctt tcagcaggga
cgaagcgaaa gtgacggtac ctgcagaaga agcaccggcc 480aactacgtgc cagcagccgc
ggtaatacgt agggtgcaag cgttgtccgg aattactggg 540cgtaaagagc tcgtaggcgg
tttgtcgcgt cgtctgtgaa aactcgaggc tcaacctcga 600gcttgcaggc gatacgggca
gacttgagta ctgcagggga gactggaatt cctggtgtag 660cggtgaaatg cgcagatatc
aggaggaaca ccggtggcga aggcgggtct ctgggcagta 720actgacgctg aggagcgaaa
gcgtgggtag cgaacaggat tagataccct ggtagtccac 780gccgtaaacg gtgggcgcta
ggtgtgggtt tccttccacg ggatccgtgc cgtagctaac 840gcattaagcg ccccgcctgg
ggagtacggc cgcaaggcta aaactcaaag gaattgacgg 900gggcccgcac aagcggcgga
gcatgtggat taattcgatg caacgcgaag aaccttacct 960gggtttgaca tataccggaa
agctgcagag atgtggcccc ccttgtggtc ggtatacagg 1020tggtgcatgg ctgtcgtcag
ctcgtgtcgt gagatgttgg gttaagtccc gcaacgagcg 1080caacccttgt cttatgttgc
cagcacgtaa tggtggggac tcgtaagaga ctgccggggt 1140caactcggag gaaggtgggg
acgacgtcaa gtcatcatgc cccttatgtc cagggcttca 1200cacatgctac aatggccggt
acagagggct gcgataccgt gaggtggagc gaatccctta 1260aagccggtct cagttcggat
cggggtctgc aactcgaccc cgtgaagtcg gagtcgctag 1320taatcgcaga tcagcaacgc
tgcggtgaat acgttcccgg gccttgtaca caccgcccgt 1380cacgtcatga aagtcggtaa
cacccgaagc cggtggccta acccctcgtg ggagggagcc 1440gtcgaaggtg ggatcggcga
ttgggacgaa gtcgtaacaa ggtagccgta ccggaaggtg 1500cggctggatc acctcctttc t
1521
User Contributions:
Comment about this patent or add new information about this topic: