|Home | About | Journals | Submit | Contact Us | Français|
Homoacetogens produce acetate from H2 and CO2 via the Wood-Ljungdahl pathway. Some homoacetogens have been isolated from the rumen, but these organisms are expected to be only part of the full diversity present. To survey the presence of rumen homoacetogens, we analyzed sequences of formyltetrahydrofolate synthetase (FTHFS), a key enzyme of the Wood-Ljungdahl pathway. A total of 275 partial sequences of genes encoding FTHFS were PCR amplified from rumen contents of a cow, two sheep, and a deer. Phylogenetic trees were constructed using these FTHFS gene sequences and the translated amino acid sequences, together with other sequences from public databases and from novel nonhomoacetogenic bacteria isolated from the rumen. Over 90% of the FTHFS sequences fell into 34 clusters defined with good bootstrap support. Few rumen-derived FTHFS sequences clustered with sequences of known homoacetogens. Conserved residues were identified in the deduced FTHFS amino acid sequences from known homoacetogens, and their presence in the other sequences was used to determine a “homoacetogen similarity” (HS) score. A homoacetogen FTHFS profile hidden Markov model (HoF-HMM) was used to assess the homology of rumen and homoacetogen FTHFS sequences. Many clusters had low HS scores and HoF-HMM matches, raising doubts about whether the sequences originated from homoacetogens. In keeping with these findings, FTHFS sequences from nonhomoacetogenic bacterial isolates grouped in these clusters with low scores. However, sequences that formed 10 clusters containing no known isolates but representing 15% of our FTHFS sequences from rumen samples had high HS scores and HoF-HMM matches and so could represent novel homoacetogens.
Feed ingested by ruminant animals is fermented in the rumen by a complex community of microbes. This community produces, among other products, the volatile fatty acids acetate, propionate, and butyrate, which are absorbed across the rumen wall and satisfy a large part of the animals' carbon and energy requirements. Hydrogen gas (H2) is also formed and is the major precursor of the methane (CH4) formed in ruminant animals. This ruminant-derived CH4 is a contributor to global greenhouse gas emissions (46) and also represents an energy loss for the animals (34). Proposed ruminant greenhouse gas mitigation strategies include using feeds that produce less CH4 and more volatile fatty acids (31). Alternative strategies include interventions that slow or halt methanogenesis by vaccination, using natural inhibitors found in plants, and supplementing feed with fats and oils or small-molecule inhibitors (31, 32). In the absence of methanogenesis, accumulation of H2 could lead to a decrease in the rate of feed fermentation (31, 53) and hence a decrease in animal productivity. Other microbes that use H2 without producing methane could be valuable in conjunction with intervention strategies that inhibit methanogens. This possibility has sparked interest in possible inoculation of ruminants with alternative H2 users.
Bacteria that use the Wood-Ljungdahl pathway to produce acetate from CO2 are metabolically (6) and phylogenetically (48) diverse and are designated “homoacetogens.” Homoacetogens grow with H2 or other suitable electron donors, such as formate or sugars, plus CO2 as a terminal electron acceptor, heterotrophically with organic substrates such as sugars and methoxylated compounds, or mixotrophically with, e.g., H2 and organic substrates. Homoacetogens have been reported to occur in a normally functioning rumen, but they are unlikely to compete with methanogens for H2 (24, 25, 34). However, homoacetogens could play an important role in the disposal of H2 if methanogens are not established in or are eliminated from the rumen (11, 17). At present, it is not clear whether resident rumen homoacetogens could fulfill the H2 disposal role or whether homoacetogens would have to be added to the rumen to take over this role from the methanogens.
Cultivation-based enumeration techniques have shown that the sizes of rumen acetogen populations range from undetectable to 1.2 × 109 per g of rumen contents and that the prevalence of these acetogens depends on diet, animal age, and time of sampling (5, 7, 23, 24). Several homoacetogens, including Acetitomaculum ruminis (15), Eubacterium limosum (14, 17), Blautia schinkii, and Blautia producta (11), have been isolated from ruminants. Homoacetogens have also been isolated from the kangaroo forestomach, whose function is analogous to that of the rumen, which suggests that homoacetogenesis may play a role in hydrogen removal in the low-methane-emission forestomach (37).
Because homoacetogens occur in different lineages of bacteria (48), traditional 16S rRNA gene-based surveys provide little information on their prevalence. The formyltetrahydrofolate synthetase (FTHFS) gene (fhs) has been used as a functional marker for homoacetogens, as the enzyme that it encodes catalyzes a key step in the reductive acetogenesis pathway (26). The structure of the enzyme of the homoacetogen Moorella thermoacetica has been reported, and putative functional features have been identified (27, 41, 42). FTHFS sequences from true homoacetogens differ from their homologs in sulfate-reducing bacteria and in other bacteria that degrade purines and amino acids via the glycine synthase-glycine reductase pathway (12, 21, 22, 26). At present, only a limited number of FTHFS sequences have been deposited in databases, and the vast majority of them are partial sequences retrieved from complex microbial communities. FTHFS sequences have been surveyed in sludge (39, 43, 54), termites (40, 44), salt marsh plant roots (21), horse manure (22), cow manure, freshwater sediment, rice field soil, and sewage (54), but so far only one study has investigated bovine ruminal FTHFS sequences (30). The rumen FTHFS sequences had low levels of similarity to the FTHFS sequences of known homoacetogens and could be sequences of novel homoacetogens. To our knowledge, no bacteria with these unique FTHFS sequences have been identified.
The aims of this study were to assess the diversity of FTHFS gene sequences retrieved from rumen samples and to screen novel rumen isolates for the presence of FTHFS genes and test their ability to grow as homoacetogens. We used alignments of FTHFS sequences to define a homoacetogen similarity score based on the presence of diagnostic amino acids and developed a hidden Markov model to assess the likelihood that FTHFS sequences of unknown origin are sequences from true homoacetogens that are able to use H2 or alternative electron donors for reductive acetogenesis.
The reference strains Acetobacterium woodii WB1 (= DSM 1030), A. ruminis 139B (= DSM 5522), Clostridium magnum WoBdP1 (= DSM 2767), Oxobacter pfennigii V5-2 (= DSM 3222), and Sporomusa ovata H1 (= DSM 2662) were obtained from the DSMZ (Braunschweig, Germany). Rumen isolates Clostridium sp. strain CA6, Eubacterium sp. strain SA11, and Blautia sp. strains Ser5 and Ser8 were obtained from the AgResearch culture collection, as were 51 strains of heterotrophic bacteria from the rumen of a ryegrass-clover pasture-fed wether Romney sheep (18).
The use of animals was approved by the AgResearch Grasslands Animal Ethics Committee and complied with the AgResearch Code of Ethical Conduct for the Use of Animals in Research, Testing and Teaching, as prescribed in the Animal Welfare Act of 1999 and its amendments. Rumen contents were collected from ruminally fistulated or slaughtered animals. The animals used were four Romney wether sheep fed ryegrass-clover pasture, silage (ChaffHage; The Great Hage Company, Reporoa, New Zealand), or concentrate pellets (containing 30% [wt/wt] barley grain, 15% [wt/wt] maize grain, 15% [wt/wt] alfalfa [Medicago sativa] meal, 15% [wt/wt] palm kernel extract, 10% [wt/wt] molasses, 5% [wt/wt] fish meal, and 5% [wt/wt] rumen-protected fat), four Suffolk-Romney cross ewe sheep fed willow (Salix spp.) fodder, four nonlactating Friesian-Jersey cross cows fed ryegrass-clover pasture or silage (ChaffHage), two nonlactating Friesian-Jersey cross cows fed alfalfa hay, and four castrated red deer (Cervus elaphus) bucks fed pasture ryegrass-clover or silage (ChaffHage). Portions (50 g) of samples of total rumen contents were immediately frozen, lyophilized, and then homogenized using a 100-W household coffee grinder (Russell Hobbs, Mordialloc, Victoria, Australia).
Total genomic DNA was extracted from 100 mg of freeze-dried and homogenized rumen contents or ryegrass (Lolium perenne) with a stool kit used according to the manufacturer's recommendations (Qiagen, Hilden, Germany). DNA was extracted from pure cultures with a Fast DNA kit (QBiogene, Carlsbad, CA) or InstaGene matrix (Bio-Rad, Hercules, CA).
Primers FTHFSf (5′-TTYACWGGHGAYTTCCATGC-3′) and FTHFSr (5′-GTATTGDGTYTTRGCCATACA-3′) were designed by Leaphart and Lovell (22). Known homoacetogen FTHFS sequences (20-22, 44) were aligned in BioEdit (16) with ClustalW. New primers that targeted conserved areas in these aligned sequences were designed and designated FTHFS rt f (5′-TGGGMDAARGGYRGHBWDGGYGG-3′) and FTHFS rt mod r (5′-CCRCCHWVDCYRCCYTTHKCCC-3′). The primers were used in various combinations (Fig. (Fig.1)1) to obtain products that were designated FTHFS PCR A (touchdown and 53°C PCR protocols), FTHFS PCR B (53°C PCR protocol), and FTHFS PCR C (touchdown and 53°C PCR protocols). The specificities of the primer combinations and PCR protocols were tested using DNA extracted from reference strains of homoacetogens and a rumen sample from a ryegrass-clover pasture-fed Romney wether sheep.
Each 25-μl PCR mixture consisted of 1× Taq buffer, 1.5 mM MgCl2 (FTHFS PCR A and C) or 2 mM MgCl2 (FTHFS PCR B), 1 U Taq DNA polymerase (Roche, Auckland, New Zealand), each deoxynucleoside triphosphate at a concentration of 0.2 mM, 0.4 mg ml−1 bovine serum albumin (Invitrogen, Carlsbad, CA), 0.5 pmol of each primer, and 1 μl template DNA (10-fold serially diluted in order to semiquantitatively detect FTHFS genes in samples). The previously described touchdown FTHFS PCR protocol (22) was modified as follows: initial denaturation at 94°C for 2 min, followed by nine cycles of denaturation at 94°C for 30 s, annealing at 63°C for 30 s (decreased by 1°C per cycle to 55°C), and elongation at 72°C for 1 min. After the touchdown protocol, 25 cycles with an annealing temperature of 55°C were carried out, followed by a final extension for 10 min at 72°C. As an alternative, a 53°C FTHFS PCR protocol was carried out as follows: initial denaturation at 94°C for 2 min, followed by 30 cycles of denaturation at 94°C for 30 s, annealing at 53°C for 30 s, and elongation at 72°C for 1 min. This was followed by a final extension for 10 min at 72°C. 16S rRNA genes were amplified as described previously (18).
A ryegrass-clover pasture-fed sheep (sheep 5472, Romney wether, 8 years old, s2gra clones), a concentrate-fed sheep (sheep 5580, Romney wether, 8 months old, s0con clones), a silage-fed cow (cow 714, Friesian-Jersey cross, 6 years old, c4sil clones), and a silage-fed deer (deer 2, castrated red deer buck, 10 years old, d2sil clones) were used for clone library analysis. Amplified FTHFS gene PCR products from the highest template dilution that resulted in a clear band when samples were analyzed by agarose gel electrophoresis were cloned in Escherichia coli using the pGEM-T Easy vector system (Promega, Alexandria, NSW, Australia) or a TOPO TA cloning kit (Invitrogen). The presence of inserts was confirmed by performing colony PCR with primers M13 reverse (5′-AG GGATAACAATTTCACACAGG-3′) and GEM2945f (5′-CTGCAAGGCGATTAAGTTGGG-3′) for products cloned in pGEM-T or primers M13 forward (5′-CCCAGTCAGGACGTTGTAAAACG-3′) and TOP168r (6′-ATGTTGTGTGGAATTGTGAGCGG-3′) for products cloned in TOPO TA. Colony PCR products were purified using the Wizard SV gel and PCR clean-up system (Promega) and were sequenced with primers M13 forward and M13 reverse (Allan Wilson Centre Genome Sequencing Service, Massey University, Palmerston North, New Zealand).
BLAST (1) was used to find sequences similar to the sequences that we amplified from rumen samples. Approximately 17% of the FTHFS PCR B products and 4% of the FTHFS PCR C products resulted in non-FTHFS BLAST matches, whereas all hits for FTHFS PCR A products were specific for FTHFS. Additionally, FTHFS sequences (21, 22, 35, 36, 38-40, 44, 50, 54) were extracted from the GenBank database (3) using the search terms “FTHFS,” “fhs,” and “formyltetrahydrofolate synthetase.” An overview of the sequences is given in Table S1 in the supplemental material. FTHFS sequences were translated and aligned with ClustalW in version 4 of the MEGA software package (47). The alignment and translation were corrected manually based on previously identified conserved or functionally important areas of FTHFS (26, 42). Phylogenetic trees of FTHFS sequences that covered the full region amplified by FTHFS PCR A were constructed based on both distance and character methods and a variety of treeing algorithms available in the ARB (28), PHYLIP (10), and MEGA software packages. For deduced amino acid sequences, a PAM distance matrix was used together with neighbor-joining (ARB and MEGA), maximum parsimony (ARB), and maximum likelihood (PHYLIP) algorithms. Additionally, a Poisson distance matrix was used together with neighbor-joining, unweighted-pair group method using average linkage (UPGMA), and minimum evolution treeing algorithms (MEGA), and the Kimura distance method was used in combination with neighbor joining (ARB). For nucleotide sequences Olsen (ARB) and Jukes-Cantor (ARB and MEGA) distance matrices were used in combination with neighbor joining. Where possible, confidence values were calculated using 500 to 1,000 bootstrapped trees. Nucleotide sequence-based trees were also constructed using 250 bp from both the 5′ and 3′ ends of the FTHFS genes to see whether 1,100-bp partial sequences clustered differently from the partial sequences and therefore could be chimeric. To describe the treeing results, we use the term “cluster” rather than operational taxonomic unit or clade, as the groups do not necessarily have equivalent taxonomic ranks. Sequences that did not fall into stable or coherent clusters in this analysis or did not result in BLAST matches with FTHFS genes were omitted from further analysis. A total of 76 possible chimeric sequences were identified this way; 4 of them were from our study, and the remainder were from previously reported studies or GenBank entries. Tree building was repeated without these potentially chimeric sequences. Sequences obtained using the FTHFS PCR B and C protocols were inserted into nucleotide trees using the parsimony function in ARB.
Thirty-five peptide sequences from verified homoacetogens were aligned, and 40 positions were identified at which the amino acids found in verified homoacetogens were more prevalent than the amino acids in the FTHFS of nonhomoacetogens (Table (Table1).1). The number of matches (m) at these 40 positions with residues typical of FTHFS from homoacetogens was used to calculate “homoacetogen similarity” (HS) scores for individual sequences (HSi) and for coherent clusters (HSc), as follows: HSi = m/40 × 100 and HSc = , where n is the number of sequences in a cluster. HSc is therefore simply the mean percentage of residues typical of homoacetogenic FTHFS found at the 40 diagnostic positions in all sequences in a coherent cluster.
Profile hidden Markov models (HMM) can be used to create models of gene or protein families and can assist in finding homologous relationships between sequences with known functions. Profile HMMs do not use pairwise comparison but look at position-specific information, i.e., conservation within each column of an alignment. By using a profile HMM, database sequences can be analyzed and a score can be assigned to sequences based on their similarity to a given profile HMM. For construction and calibration of HMM profiles, the HMMER software package (http://hmmer.wustl.edu) (8) was used with default parameters (commands “hmmbuild” and “hmmcalibrate”). For each sequence an HMMER bit score is calculated with an E-value, which serves as an indicator of the number of false-positive matches expected at or above this bit score. As the E-value varies with the size of the query set, we used the bit score to rank our sequences. A homoacetogen FTHFS profile HMM (HoF-HMM, available from us) was built using in silico-translated and manually corrected alignments of FTHFS protein sequences covering positions 137 to 487 in the M. thermoacetica sequence (27), produced using ClustalW (MEGA software). The sequences in these alignments were exclusively FTHFS sequences from known homoacetogens (GenBank accession numbers AF295701 to AF295709, AJ494822 to AJ494824, AY162313, AY162314, AY254548, DQ152900 to DQ152908, J02911, GU124153 to GU124155, and GU124175 to GU124180). FTHFS sequences were tested and scored (command “hmmsearch”) against HoF-HMM to assess the similarity of query FTHFS sequences to FTHFS sequences of homoacetogens. Additionally, FTHFS sequences were also tested against an HMM for the FTHFS gene family available in Pfam (PF01268) constructed by R. D. Finn and A. Bateman (http://pfam.sanger.ac.uk/family?acc=PF01268; accessed 9 September 2009).
Rumen isolates (18) that produced an FTHFS PCR product were grown in RM02 medium [containing (per liter of water) 10 mmol KH2PO4, 5 mmol (NH4)2SO4, 20 mmol KCl, 3 mmol CaCl2, 4 mmol MgCl2, 1 ml trace element solution SL10 (52), 1 ml selenite-tungstate solution (49), 50 mmol NaHCO3, 3 mmol l-cysteine-HCl·H2O, 0.2 mg resazurin, 50 ml clarified rumen fluid with vitamins (18), and 1 g yeast extract; gassed with 100% CO2; pH 6.4] with H2 or 5 mM glucose as a substrate to see whether they were capable of reductive acetogenesis. Control tubes contained no added H2 or glucose. Hydrogen-containing substrate tubes were pressurized with H2-CO2 (4:1, vol/vol) at 170 kPa. Growth was followed by measuring the optical density at 600 nm (Ultrospec 1100 pro; Amersham, Little Chalfont, United Kingdom) during incubation at 39°C. The production of acetate and other fermentation end products by each isolate was assessed with an LC10Ai high-performance liquid chromatograph equipped with an RID10A detector (Shimadzu, Kyoto, Japan) and a Resex 8 u 8% H organic acid 300- by 780-mm column (00H-0138-KO; Phenomenex, Torrance, CA) maintained at 45°C, using 5 mM H2SO4 at a flow rate of 0.8 ml min−1 as the mobile phase (9). All isolates were able to grow in the medium when fermentable sugars were supplied as growth substrates.
Sequences obtained in this study have been deposited in the GenBank database under accession numbers GU124152 to GU124473 (see Table S1 in the supplemental material).
Different primer combinations were used to amplify different regions of the FTHFS gene, designated regions A, B, and C (Fig. (Fig.1).1). No single primer combination could amplify the FTHFS gene from all homoacetogen reference strains (Table (Table2).2). Only FTHFS PCR C amplified a partial FTHFS sequence from the rumen homoacetogen A. ruminis 139B. Partial FTHFS gene sequences were also determined for Clostridium sp. CA6, O. pfennigii V5-2, and the homoacetogens A. woodii WB1, Blautia spp. Ser8 and Ser5, and Eubacterium sp. SA11. These new sequences were later used to develop scoring methods to classify FTHFS sequences.
Partial FTHFS genes were amplified by PCR from rumen contents from a variety of ruminants (deer, sheep, and cows) fed different diets (summer pasture, winter pasture, silage, grain concentrate pellets, willow, and hay). 16S rRNA gene PCR products could be amplified from 10- to 1,000-fold-more-dilute template DNA than FTHFS gene PCR products. This indicates that bacteria with FTHFS are quite abundant in the rumen. The ratio of FTHFS gene PCR products to 16S rRNA gene PCR products was highest in sheep fed summer pasture, winter pasture, or grain concentrate pellets and in silage-fed deer. This implies that these animals may harbor greater numbers of FTHFS-positive bacteria. Amplification of partial FTHFS gene sequences from rumen samples resulted in similar PCR product intensities at similar dilutions regardless of whether FTHFS PCR A, B, or C was used.
The products generated using one of the primer combinations (PCR A) appeared to have a toxic effect on E. coli, which resulted in cloning efficiencies that were up to 90% lower and in fewer colonies per plate, especially with the TOPO TA system, compared with the results obtained for the products generated using the PCR B and C primer combinations or with PCR-amplified 16S rRNA gene products. Omission of IPTG from the plates on which the recombinants were cultured in an attempt to minimize expression did not increase the number of colonies. It is possible that the FTHFS products might interfere with one-carbon metabolism in E. coli. We added 200 μl of 1 mM methanol, 1 mM folate, or 5 mM formate to each plate, but this did not increase the number of recombinants obtained.
The BLAST matches for the PCR-amplified FTHFS sequences were predominantly putative FTHFS sequences from uncultured organisms. Our rumen-derived FTHFS sequences were not similar to the FTHFS sequences of plants, such as spinach (GenBank accession number M83940). We could not amplify FTHFS PCR products from ryegrass (L. perenne), which further suggests that the rumen FTHFS sequences are not likely to be of plant origin. All phylogenetic tree construction methods resulted in similar tree topologies. Clusters 1, 8, 9, 18, and 19 had low bootstrap support when most treeing methods were used, but they were obtained with both maximum likelihood and maximum parsimony methods. A representative tree based on partial FTHFS sequences (350 amino acid residues) is shown in Fig. Fig.2.2. Over 90% (703/780) of FTHFS sequences fell into 34 clusters that were generally defined with good bootstrap support (Fig. (Fig.2;2; see Table S2 in the supplemental material) and were consistently found with all treeing methods.
The cloned sequences obtained from the three different PCRs (PCRs A, B, and C) were diverse and were spread across the 34 clusters described above (see Table S2 in the supplemental material). A few cloned sequences from PCR C segregated into two clusters that contained no other FTHFS sequences from other studies (see Table S2 in the supplemental material). Due to the diversity and the large number of clusters, it was not possible to determine whether the different primer combinations resulted in a bias of amplicons toward certain FTHFS sequence clusters or whether particular FTHFS sequences were more dominant in any one animal (see Table S2 in the supplemental material).
FTHFS sequences from known homoacetogens did not form a stable monophyletic cluster, and less than 4% of the FTHFS sequences amplified from rumen contents grouped with them. Most rumen FTHFS sequences did not group with sequences from cultured isolates, including nonhomoacetogens. Distinctive subclusters were identified in the Blautia-associated, Butyrivibrio, rumen 5, rumen 9, and other clusters (data not shown). The FTHFS sequences from known homoacetogenic Blautia isolates formed a distinct subcluster within the Blautia-associated cluster. All Blautia-associated sequences had a 4-amino-acid insertion between residues 233 and 234, a trait shared with the termite treponeme cluster FTHFS sequences (44). Numbering was based on the M. thermoacetica FTHFS sequence (27).
FTHFS PCR products were obtained from 21 of 51 new heterotrophic, anaerobic rumen isolates that had not been specifically enriched under conditions that selected for homoacetogens (18). These isolates were tested for the ability to produce acetate from H2 and CO2 or glucose. None of the isolates grew on H2 and CO2 or produced acetate as the main fermentation end product from glucose, suggesting that these bacteria are not capable of reductive acetogenesis. The homoacetogens Eubacterium sp. SA11 and Blautia spp. Ser5 and Ser8 did produce acetate as their main end product from H2 and CO2 (Table (Table2)2) and glucose, confirming that these strains are capable of reductive acetogenesis via the Wood-Ljungdahl pathway.
As most sequences amplified directly from rumen contents did not cluster with FTHFS sequences from known homoacetogens or sequences from FTHFS-containing sulfate-reducing or purinolytic bacteria, it was difficult to determine whether these sequences could have originated from true homoacetogens. FTHFS sequences in several of the clusters were less than 50% similar to sequences in other clusters, raising the question of whether they were indeed FTHFS sequences. However, all sequences tested did display a significant match with the Pfam HMM for the FTHFS gene family, but this did not allow differentiation of FTHFS sequences from homoacetogens from other FTHFS and FTHFS-like sequences.
Lovell and Leaphart identified residues that distinguish FTHFS of true homoacetogens from other FTHFS (26). Peptide sequences from 21 verified homoacetogens were aligned, and conserved residues were identified by visual inspection (Table (Table1).1). Using this enlarged data set, we identified 40 positions at which the amino acids found in homoacetogens were more prevalent than the amino acids in FTHFS of nonhomoacetogens (Table (Table1).1). The number of positions at which there is a residue typically found in homoacetogen FTHFS (Table (Table1)1) can be determined for any FTHFS sequence. To determine whether a cluster contains sequences of FTHFS that may be used in reductive acetogenesis, the mean percentage of FTHFS residues typical of homoacetogens in each sequence of a cluster was calculated. We refer to this percentage here as the HSc score (Fig. (Fig.22).
Clusters that contained FTHFS sequences from known homoacetogens had high HSc scores (>94%). Several clusters that did not contain FTHFS sequences from known homoacetogen isolates also had high HSc scores (≥90%), indicating that they may contain sequences of as-yet-unknown homoacetogens (Fig. (Fig.2).2). Clusters that contained FTHFS sequences from nonhomoacetogenic isolates that were identified as members of the phyla Firmicutes and Bacteroidetes and belonged to a number of known and potentially new genera (18) all had low HSc scores (<60%).
As the selection of the residues used to calculate the HSc score was subjective, we used an FTHFS profile HMM to assess homology relationships between FTHFS sequences from known homoacetogens and FTHFS sequences amplified from the rumen and other environments. FTHFS sequences were scored against the homoacetogen FTHFS profile HMM (HoF-HMM) to estimate their similarities to FTHFS sequences from known homoacetogens (Fig. (Fig.3).3). The significance of the matches of translated FTHFS sequences was expressed as an HMMER bit score and a corresponding E-value with the HoF-HMM. The higher the bit score, the better the match with the profile HMM was. The E-value gives an indication of the number of false-positive matches expected for a given bit score. The bit scores of FTHFS sequences obtained from GenBank that we tested against the HoF-HMM ranged from 409.1 to 791.2. These matches can be considered significant matches. The E-values ranged from 5.8e-121 to 5.3e-236, indicating that false positives were unlikely. Homoacetogen FTHFS sequences from known homoacetogens and rumen clone sequences from clusters with high HSc scores had the best overall HoF-HMM bit score rankings, ranging from 732.8 to 791.2. FTHFS sequences that originated from nonhomoacetogenic Clostridium sp. CA6, Treponema azotonutricium ZAS-9, and the sulfate-reducing isolate BG9 had HoF-HMM bit scores that fell into the range of scores for homoacetogens. Cloned sequences that did not group with sequences of known homoacetogens, sulfate-reducing bacteria, or purinolytic bacteria and that fell into clusters with low HSc scores had intermediate HoF-HMM bit scores ranging from 409.1 to 657.3. There were also several clusters of FTHFS sequences that did not group with the FTHFS sequences of known homoacetogens, had high HSc scores, and had HoF-HMM scores that fell between those of known homoacetogens and those of nonhomoacetogens.
In previous studies, FTHFS sequences from known homoacetogens have been found to cluster together, separate from the FTHFS sequences of purinolytic bacteria and the coherent cluster of FTHFS sequences from sulfate-reducing bacteria (26). This could imply that sequences that cluster outside this homoacetogen cluster are not from homoacetogens or that they may be from as-yet-uncultured homoacetogens. In the absence of cultures, this remains to be proven. To complicate matters further, some FTHFS sequences from nonhomoacetogens, such as Treponema ZAS-9 (44) and Clostridium sp. CA6 (this study), have been found to cluster with homologs from homoacetogens. This may be because of a recent lesion elsewhere in the reductive acetogenesis pathway that renders these isolates incapable of producing acetate, because they have a requirement for electron donors other than H2, or because the FTHFS genes have recently been acquired by horizontal gene transfer. After constructing phylogenetic trees with a large number of FTHFS sequences from isolates and FTHFS sequences amplified from environmental DNA, we found that FTHFS sequences from known homoacetogens no longer formed a coherent cluster. Also, many sequences that were loosely affiliated (low bootstrap values) with the homoacetogens no longer clustered with these FTHFS sequences. This can also be explained, as bootstrap values supporting homoacetogen FTHFS clusters in previous studies were sometimes low (26, 35) and the introduction of new sequences changes the topology of a tree in regions where grouping is weak. Studies of libraries of cloned FTHFS sequences from human feces (35), anaerobic sludge (43), termites (40), and different ruminants (this study) showed that many PCR-amplified FTHFS sequences do not cluster with the FTHFS sequences from known homoacetogens or other cultured isolates. This raises the issue of which organisms these putative FTHFS sequences originate from and what function they have.
Isolates of heterotrophic rumen bacteria which possessed an FTHFS-like gene did not produce acetate as their main end product from H2 and CO2 or glucose. These isolates grew well on glucose, whereas their growth on hydrogen and their production of acetate under these conditions were no greater than the background growth and production of acetate in medium with no added substrate. This indicates that these isolates are not capable of reductive acetogenesis. The presence of FTHFS in these 21 of 51 random heterotrophic isolates belonging to multiple genera of the phyla Firmicutes and Bacteroidetes suggests that FTHFS is widely distributed in rumen bacteria but that it does not play a role in reductive acetogenesis in these organisms. The FTHFS sequences from these isolates did not cluster with those from known homoacetogens, and they possessed few of the amino acid residues characteristic of FTHFS from known homoacetogens. Thus, the FTHFS-like genes in these isolates may be xenologs, analogs, or homologs of the FTHFS gene. FTHFS is not used exclusively in the Wood-Ljungdahl pathway; it is also used as a methyltransferase in purine and glycine degradation and in the metabolism of some sulfate-reducing bacteria (12). The complexity of feed entering the rumen offers opportunities for bacteria with these types of metabolism to grow in the rumen. However, none of the rumen-derived FTHFS sequences clustered with the sequences of sulfate reducers or Treponema spp.
An FTHFS gene could be amplified from A. ruminis only using the FTHFS PCR C primer combination. This indicates that several primer combinations may be required to obtain insight into the total homoacetogen community. Some homoacetogen FTHFS genes may not be targeted by current FTHFS primers. Thus, there is no universal PCR for the FTHFS gene. Every primer combination assessed in this study amplified FTHFS gene sequences that clustered with those of nonhomoacetogenic Butyrivibrio strains. Therefore, the FTHFS PCR cannot be restricted to true homoacetogens. This means that the FTHFS gene is not suitable for real-time PCR or specific quantification of homoacetogens. However, as certain homoacetogens possess characteristic residues and motifs, it may be possible to develop strain-specific FTHFS PCR protocols in the future.
LigH, an ortho-demethylating enzyme of Sphingomonas paucimobilis involved in tetrahydrofolate-mediated C1 metabolism, is approximately 60% similar to FTHFS of M. thermoacetica (26, 29). This raises the question of whether other FTHFS sequences that have low levels of similarity to the sequences from known homoacetogens are sequences from “authentic” homoacetogens. We found that the levels of sequence identity between putative FTHFS sequences, such as those in the termite treponeme cluster (40), which contains homoacetogens, and FTHTS-like sequences in the Bacteroidetes cluster, which contains no known homoacetogens, were in some cases less than 50%. We believe that the HS score and HoF-HMM can be used to determine whether sequences originate from true homoacetogens. A limitation of the HS score and HoF-HMM is that the calculations are based on available sequences of known homoacetogens. Thus, there is a danger that unidentified FTHFS sequences from true but as-yet-uncultured homoacetogens that contain residues not found in the proteins from known homoacetogens are not included in the models, resulting in HS and HoF-HMM scores that are lower than they should be. However, as 40 residues were taken into account to calculate the HS score and a wide range of FTHFS sequences from known homoacetogens were used, we believe that both the HS score and the homoacetogen FTHFS profile HMM are useful indicators of likelihood. An advantage of the HoF-HMM is that it provides a fast way to score large numbers of FTHFS sequences from databases with regard to their likelihood of originating from homoacetogens. However, unlike the HS score, it is not possible to tell which residues are responsible for the score in the HoF-HMM.
We found that clusters that contained known homoacetogens had a high HSc score, whereas clusters that contained FTHFS sequences known to originate from nonhomoacetogenic isolates had a low HSc score. The HoF-HMM bit scores agreed with these HSc scores. Therefore, clusters of rumen FTHFS sequences that do not group with sequences from known homoacetogens but have a high HSc score (≥90%) and a high HoF-HMM bit score could represent novel homoacetogens. We found 10 clusters that fell into this category, representing 15% (FTHFS PCR A) or 9% (FTHFS PCR B and C) of our sequences that were amplified directly from rumen contents. Hence, most of the FTHFS sequences detected in the rumen samples studied probably did not originate from homoacetogens.
FTHFS activity has been described for a wide variety of bacteria, including nonhomoacetogens, such as Proteus vulgaris and Bacteroides sp. (51). Some nonhomoacetogenic FTHFS-containing isolates, such as T. azotonutricium ZAS-9, sulfate-reducing bacterium BG9, and Clostridium sp. CA6, may have lost the ability to grow as homoacetogens recently due, for example, to a mutation in the FTHFS gene or another gene that may be essential for growth with H2, or they may require electron donors other than H2. Substantiation of both the HoF-HMM and the HS score will ultimately require isolation of the microbes and testing whether they are capable of using the Wood-Ljungdahl pathway.
To the best of our knowledge, this analysis was the most comprehensive analysis of rumen content FTHFS sequences to date and the first study that sought to differentiate FTHFS sequences from a potentially functional perspective. Other genes involved in the Wood-Ljungdahl pathway, such as the gene encoding acetyl-coenzyme A (acetyl-CoA) synthase, have been assessed for use in surveying homoacetogens (13). However, like FTHFS, acetyl-CoA synthase is present in nonhomoacetogens (12, 45). Investigation of new rumen homoacetogens, especially homoacetogens with FTHFS that fall into clusters with high HSc scores and that have high scores when the HoF-HMM is used, would be valuable. The majority of the rumen FTHFS sequences with high HS scores and HoF-HMM profile matches probably represent novel groups of homoacetogens. If rumen methanogens are successfully inhibited or eliminated in an effort to reduce greenhouse gas emissions, we will need to know whether these resident rumen homoacetogens are able take over the role of H2 disposal or if it will be necessary to supplement the rumen with alternative hydrogen users.
We thank Marek Kirs for designing primer FTHFS rt f, Nikki Kenters for providing DNA and fermentation end product data for isolates of rumen bacteria, Jeyamalar Jeyanathan for providing DNA extracted from rumen contents, Bryan Treloar for high-performance liquid chromatography analysis of samples, and Graeme Attwood and Bill Kelly for critically reviewing the manuscript.
This work was funded by a contract with the Pastoral Greenhouse Gas Research Consortium.
Published ahead of print on 29 January 2010.
†Supplemental material for this article may be found at http://aem.asm.org/.