Report

Microbial Biosynthesis of Alkanes

See allHide authors and affiliations

Science  30 Jul 2010:
Vol. 329, Issue 5991, pp. 559-562
DOI: 10.1126/science.1187936

Abstract

Alkanes, the major constituents of gasoline, diesel, and jet fuel, are naturally produced by diverse species; however, the genetics and biochemistry behind this biology have remained elusive. Here we describe the discovery of an alkane biosynthesis pathway from cyanobacteria. The pathway consists of an acyl–acyl carrier protein reductase and an aldehyde decarbonylase, which together convert intermediates of fatty acid metabolism to alkanes and alkenes. The aldehyde decarbonylase is related to the broadly functional nonheme diiron enzymes. Heterologous expression of the alkane operon in Escherichia coli leads to the production and secretion of C13 to C17 mixtures of alkanes and alkenes. These genes and enzymes can now be leveraged for the simple and direct conversion of renewable raw materials to fungible hydrocarbon fuels.

Efforts to transition from fossil fuels to renewable alternatives have focused on the conversion of renewable biomass to “drop-in” compatible fuels and chemicals (13). Routes to renewable hydrocarbons are emerging, but to date, these require expensive chemical hydrogenation. Alkanes, observed throughout nature, are produced directly from fatty acid metabolites—for example, as plant cuticular waxes (4), as insect pheromones (5), and with unknown functions in numerous organisms (69). Biochemical studies of alkane biosynthesis have focused on eukaryotic systems, with most evidence supporting a decarbonylation of fatty aldehydes as the primary mechanism (10, 11). Although cer1 from Arabidopsis thaliana has been proposed as a candidate gene encoding this activity (12), no studies conclusively associate any gene with these biochemical activities.

Alkanes have been reported in a diversity of microorganisms, but some results remain controversial (13, 14). From our assessment, the most consistent reports are from the cyanobacteria (9, 15, 16) and natural habitats dominated by cyanobacteria (17). Heptadecane is the most abundant alkane reported in these photoautotrophic bacteria, an observation consistent with the “n – 1” rule for alkanes, resulting from decarbonylation of typically even-numbered fatty aldehydes. Because cyanobacteria are phylogenetically homogeneous, with more than 50 sequenced genomes publicly available, our search began with comparative biochemistry and genomics. Eleven cyanobacterial strains of known sequence were photoautotrophically grown, and their culture extracts were evaluated for hydrocarbon production (Table 1). Ten of these strains produced alkanes, mainly heptadecane and pentadecane, along with alkenes, presumably derived from unsaturated fatty aldehydes. However, one strain, Synechococcus sp. PCC7002, did not produce alkanes. On the assumption that an alkane biosynthesis pathway was not present in Synechococcus sp. PCC7002, we undertook a subtractive genome analysis. The 10 genomes of the alkane-producing cyanobacteria were intersected, and the PCC7002 genome was subtracted by using a 40% sequence identity cut-off to select orthologs. Seventeen genes common to the 10 producing strains remained, and 10 of these already had assigned functions (table S2). Two of the remaining hypothetical proteins stood out as likely candidates for alkane biosynthesis. Representative of these were open reading frames orf1593 and orf1594 from S. elongatus PCC7942. PCC7942_orf1594 belongs to the short-chain dehydrogenase or reductase family, whereas PCC7942_orf1593 shared similarity to the ferritin-like or ribonucleotide reductase–like family. Because alkane biosynthesis was predicted to proceed via the decarbonylation of fatty aldehydes, it was possible that PCC7942_orf1594 catalyzed the reduction of a fatty acid intermediate, such as an acyl carrier protein (ACP) or coenzyme A (CoA) acyl-thioester, to form a fatty aldehyde (18, 19), and that PCC7942_orf1593, which is similar to enzymes that catalyze radical-based chemical reactions such as ribonucleotide reductase R2 (20), was the decarbonylase. PCC7942_orf1593 and orf1594 orthologs appear to occur only in cyanobacteria and likely form a conserved operon, because they are found adjacent to each other in a majority of cyanobacterial genomes (table S3). Cyanothece sp. PCC7424 was the only additional strain identified that lacked orthologs to these sequences.

Table 1

Occurrence of alkanes in selected cyanobacteria with fully sequenced genomes (21).

View this table:

To test the hypothesis that the S. elongatus PCC7942_orf1593 and orf1594 family proteins were necessary and sufficient for alkane biosynthesis, a genetic “knock in” and “knock out” strategy was taken. PCC7942_orf1593 and orf1594 were expressed both separately and together in Escherichia coli, and extracts of these cells were evaluated for the production of hydrocarbons. Although E. coli produced no detectable hydrocarbons, extracts of PCC7942_orf1594-expressing cells contained substantial quantities of even-chain fatty aldehydes and fatty alcohols (Fig. 1A), whereas coexpression of both PCC7942_orf1593 and orf1594 resulted in the production of odd-chain alkanes and alkenes, as well as even-chain fatty aldehydes and fatty alcohols (Fig. 1A). Thus, PCC7942_orf1593 and orf1594 are sufficient for in vivo alkane biosynthesis, and fatty aldehydes are likely the biosynthetic intermediates. Expression of PCC7942_orf1593 alone was indistinguishable from the E. coli negative control, which was expected, because E. coli does not naturally produce fatty aldehydes. The production of fatty alcohols upon PCC7942_orf1594 expression can be attributed to intrinsic activities in E. coli that reduce fatty aldehydes to fatty alcohols, as is observed upon expression of fatty acyl-CoA reductases (19). Although it is possible that fatty alcohols may be biosynthetic intermediates, for mechanistic reasons it seems unlikely, and in vitro experiments support this hypothesis (see below). Fifteen additional PCC7942_orf1593 orthologs from various cyanobacteria were evaluated (table S3), and all conferred alkane production when expressed in E. coli together with PCC7942_orf1594. Alkane profiles of selected strains in Fig. 1B show that the “recombinant hydrocarbon” mixtures are primarily made up of pentadecane and heptadecene. Strains with the highest titers, e.g., coexpressing the orf1593 orthologs from Nostoc punctiforme PCC73102, produced a mixture of tridecane, pentadecene, pentadecane, and heptadecene, typically at a ratio of 10:10:40:40. Alkane titers were over 300 mg/liter when a modified mineral medium was used (21), and more than 80% of the hydrocarbons were found outside the cells (fig. S2).

Fig. 1

Demonstration that a two-gene pathway widespread in cyanobacteria is responsible for alkane biosynthesis. (A) E. coli MG1655 does not produce hydrocarbons. When S. elongatus PCC7942_orf1594 was expressed, E. coli produced fatty aldehydes and fatty alcohols. The coexpression of S. elongatus PCC7942_orf1593 and orf1594 led to alkane, alkene, fatty aldehyde, and fatty alcohol production. (B) For comparison, selected PCC7942_orf1593 orthologs from other cyanobacteria were coexpressed with PCC7942_orf1594 in E. coli, and all produced similar compounds [only alkanes and alkenes are shown; fig. S1 shows an example of a detailed gas chromatography–mass spectrometry (GC-MS) trace of a recombinant hydrocarbons mixture]. (C) Synechocystis sp. PCC6803 has the ability to synthesize the alkane heptadecane. (D) Deletion of the two adjacent Synechocystis sp. PCC6803 orfs sll0208 and sll0209, S. elongatus PCC7942_orf1593 and 1594 orthologs, abolished heptadecane biosynthesis. GC-MS traces are shown. (E) Model of the alkane biosynthesis pathway in cyanobacteria consisting of a fatty aldehyde–generating acyl-ACP reductase (e.g., PCC7942_orf1594 or sll0209) and a fatty aldehyde decarbonylase (e.g., PCC7942_orf1593 or sll0208). The PCC7942_orf1593 orthologs in (B) are from the following strains: N. punctiforme PCC73102, Thermosynechococcus elongatus BP-1, Synechococcus sp. Ja-3-3Ab, P. marinus MIT9313, P. marinus NATL2A, and Synechococcus sp. RS9117, which has two paralogs (RS9117-1 and -2).

In S. elongatus PCC7942, orf1593 and orf1594 appear to be part of a larger operon that contains a gene encoding a subunit of the acetyl-CoA carboxylase (accA), which is essential for growth. To avoid any polar effects of deleting the genes, we replaced the PC7942_orf1593 and 1594 orthologs in Synechocystis sp. PCC6803 (where the orthologous genes apparently form only a bicistronic operon) with a kanamycin-resistance cassette. This replacement abolished the presence of alkanes (Fig. 1, C and D) in the extracts of photoautoptrophically grown cells. Thus, these genes are both necessary for alkane biosynthesis in cyanobacteria and sufficient to confer alkane biosynthesis in a heterologous host, such as E. coli. Our findings further support the model of odd-chain alkane biosynthesis through the decarbonylation of even-chain fatty aldehydes.

The proposed pathway for alkane biosynthesis is depicted in Fig. 1E. To evaluate this model, we undertook a preliminary in vitro characterization of the purified recombinant enzymes (21). Acyl-acyl carrier protein (acyl-ACP) and acyl-coenzyme A (acyl-CoA), the major activated forms of fatty acids in bacteria, were evaluated as substrates for PCC7942_orf1594. When incubated with acyl-ACP or acyl-CoA (e.g., oleoyl-ACP and oleoyl-CoA), PCC7942_orf1594 catalyzed the reduced nicotinamide adenine dinucleotide phosphate (NADP+) (NADPH)–dependent reduction to the corresponding fatty aldehyde (e.g., cis-9-octadecenal) and required divalent cations, such as magnesium for catalysis. Fatty alcohols were not observed in the in vitro reaction. Although both fatty acyl thioesters were substrates, the Michaelis constant (KM) for acyl-ACP was 8 ± 2 μM, and that for acyl-CoA was 130 ± 30 μM (Fig. 2, A and B). Accordingly, we named this enzyme acyl-ACP reductase (AAR), because it appears that acyl ACP is kinetically preferred and is likely the in vivo substrate. This hypothesis is supported by our finding that PCC7942_orf 1593 and orf1594 coexpression in an E. coli strain devoid of acyl-CoAs (the major acyl-CoA synthetase gene fadD was deleted) led to very similar levels of hydrocarbon production, comparable to those in an E. coli wild-type strain (Fig. 2C). This substrate selectivity distinguishes AAR from fatty acyl-CoA reductases (FAR) that specifically reduce acyl-CoAs but not acyl-ACPs to fatty aldehydes (19).

Fig. 2

Demonstration that S. elongatus PCC7942_orf1594 encodes an acyl-ACP reductase. (A and B) Michaelis-Menten plots for PCC7942_orf1594 protein with oleoyl-ACP and oleoyl-CoA as substrates. NADPH concentration was 2 mM. (C) Comparison of hydrocarbon production of E. coli wild-type and an E. coli fadD deletion strain. Both strains express S. elongatus PCC7942_orf1593 and orf1594. fadD encodes the major acyl-CoA synthetase of E. coli. As this strain lacks the ability to synthesize acyl-CoAs, hydrocarbons are derived directly from acyl-ACP.

Comparison of PCC7942_orf1593 to the sequence database suggested that the cyanobacterial aldehyde decarbonylases are members of the ferritin-like or ribonucleotide reductase–like family of nonheme diiron enzymes (20). This was confirmed by a crystal structure of the PCC7942_orf1593 ortholog from Prochlorococcus marinus MIT9313, PMT1231, solved by the Joint Center of Structural Genomics (protein database entry PDB|2OC5|A) but without assigning an enzymatic function. We confirmed that PMT1231, like PCC7942_orf1593, is an active aldehyde decarbonylase, because it conferred alkane biosynthesis to E. coli when coexpressed with PCC7942_orf1594 (Fig. 1B). The three-dimensional structure shows similarities to the 8 α-helical bundle of the second subunit of E. coli ribonucleotide reductase (R2) (22) (Fig. 3, A and B). In the solved structures, both proteins have two irons coordinated to histidine and aspartate or glutamate residues (Fig. 3, C and D). In R2, the diiron is part of a diiron(III)–tyrosyl radical cofactor that is essential to the ribonucleotide reduction process as a radical-chain initiator (20). The radical is then transduced to the catalytic R1 subunit. Homology modeling revealed that this tyrosine residue in R2 (Tyr122) was replaced by phenylalanine residues in the aldehyde decarbonylases (e.g., Phe80 in PMT1231) (Fig. 3, C and D), which suggested that decarbonylation does not include a diiron(III)–tyrosyl radical cofactor. To examine the possibility that another conserved tyrosine might form such a cofactor in the aldehyde decarbonylases, we replaced a tyrosine, which was conserved in all cyanobacterial decarbonylases and appeared to be within 6 Å of the diiron (e.g., Tyr135 in PMT1231), with phenylalanine. This mutation had no detectable effect on alkane biosynthesis activity in vivo (fig. S3), which suggested that decarbonylation proceeds through a mechanism different from ribonucleotide reductase R2. A different class of R2s, exemplified by the R2 from Chlamydia trachomatis, is a manganese/iron and not a diiron protein, and it does not use a diiron(III)-tyrosyl but a stable manganese(IV)/iron(III) cofactor for radical initiation (23). At present, we cannot exclude that active aldehyde decarbonylases are manganese/iron proteins—heterologous expression in E. coli might result in mismetallation. Aldehyde decarbonylases are also considerably smaller than ribonucleotide reductase R2s (220 to 250 versus 300 to 400 amino acids) and lack the C-terminal region that, in R2, interacts with the R1 subunit.

Fig. 3

Comparison of the similar three-dimensional structures of a cyanobacterial aldehyde decarbonylase and ribonucleotide reductase R2 from E. coli. (A) Aldehyde decarbonylase from P. marinus MIT9313 from PDB file 2OC5. (B) Ribonucleotide reductase R2 from E. coli from PDB file 1RIB. (C) Active site of the P. marinus MIT9313 aldehyde decarbonylase. (D) Active site of the E. coli ribonucleotide reductase R2. Amino acid residues that coordinate the metal centers are shown. The active site of E. coli ribonucleotide reductase R2 also shows a tyrosine, which is part of the diiron(III)–tyrosyl radical cofactor and which is replaced with a phenylalanine in the aldehyde decarbonylase structure. The structure of P. marinus MIT9313 aldehyde decarbonylase shows two iron atoms bound, but it cannot be excluded that the active enzyme is a manganese/iron protein (see text). Structures were viewed in PyMOL.

To evaluate in vitro decarbonylation, we tested the purified aldehyde decarbonylase from Nostoc punctiforme PCC73102, NpunR1711, which had shown the highest in vivo activity (Fig. 1B), with octadecanal as substrate in the presence or absence of different cofactors. As a number of nonheme diiron enzymes require ferredoxin, ferredoxin reductase, and reducing equivalents for activity (24, 25), we tested commercially available spinach ferredoxin and ferredoxin reductase and NADPH. Indeed, in vitro decarbonylation of octadecanal to heptadecane was only observed in the presence of ferredoxin, ferredoxin reductase, and NADPH, and omitting any one of these cofactors completely abolished in vitro activity of the decarbonylase (fig. S4). As coexpression of a cyanobacterial acyl-ACP reductase and aldehyde decarbonylase is sufficient for in vivo alkane biosynthesis in E. coli (see above), ferredoxin and ferredoxin reductase components essential for decarbonylation must be provided at least to some extent by endogenous E. coli proteins.

The genes and enzymes described here provide a foundation for the deeper understanding and further development of this pathway. The ability to biologically convert renewable carbohydrate selectively to fuel-grade alkanes without hydrogenation is an important step toward the goal of low-cost renewable transportation fuels.

Supporting Online Material

www.sciencemag.org/cgi/content/full/329/5991/559/DC1

Materials and Methods

Figs. S1 to S4

Tables S1 to S3

References

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank M. Alibhai for homology modeling, S. Brubaker for bioinformatics work, T. Baron for technical assistance, and C. Chang for help with preparing Fig. 3. The patent applications WO 2009/140695 and WO 2009/140696 are relevant to this paper. All authors have a financial interest in LS9, Inc. We would like to dedicate this paper to the memory of C. Richard Hutchinson, who died on 5 January 2010. “Hutch” was an important mentor for A.S. and S.B.d.C. He will be missed.
View Abstract

Navigate This Article