Formate-Dependent Growth and Homoacetogenic Fermentation by a Bacterium from Human Feces: Description of Bryantella formatexigens gen. nov., sp. nov.

Abstract

Formate stimulates growth of a new bacterium from human feces. With high formate, it ferments glucose to acetate via the Wood-Ljungdahl pathway. The original isolate fermented vegetable cellulose and carboxymethylcellulose, but it lost this ability after storage at −76°C. 16S rRNA gene sequencing identifies it as a distinct line within the Clostridium coccoides supra-generic rRNA grouping. We propose naming it Bryantella formatexigens gen. nov., sp. nov.

Plant cell wall polysaccharides in human diets are not digested by host enzymes (3). The cellulose and hemicellulose in vegetables and fruits are digested in the colon (2, 11, 27) and are fermented by the colonic microbial community to molar ratios of ca. 56 acetate: 22 propionate: 22 butyrate (4, 32) and H₂, CO₂, and CH₄ (32). Bacteria that digest filter paper (FP) or Avicel are relatively unimportant in the human colon. Betian et al. (1) and Wedekind et al. (31) showed that the frequency of individuals that harbor them is low and, when present, their concentrations are ca. 0.001 times the concentration of all anaerobic bacteria. We hypothesized that bacteria that use amorphous cellulose found in vegetables, but not the crystalline cellulose in FP, are present in the colon. A goal of this study was to enumerate and isolate human colonic bacteria that use vegetable cellulose.

MPN study.

We used a purified cellulose preparation from cabbage for most-probable-number (MPN) comparisons of the concentrations of bacteria that use FP cellulose, vegetable cellulose, and starch in human fecal suspensions. A modification of the method described by Ehle et al. (8) was used to prepare a cellulose-enriched fiber preparation (VCP) from fresh white cabbage (29). The hydrolysis of 1 g of VCP with 2 N HCl solubilized 287 mg of reducing sugar (glucose equivalent) (22). The insoluble residue contained 138 mg of glucose equivalents when measured by the anthrone procedure (26). Distilled water suspensions of the powder were ball milled for 18 h at 25°C prior to addition to media.

MPN analyses were run concurrently with 0.8% VCP, 0.5% corn starch (CS), or 1-cm by 5-cm strips of Whatman number 1 FP with 10 subjects and concurrently with VCP and FP cellulose with 15 subjects. The medium (B1C) contained (per liter): NaHCO₃, 5.0 g; K₂HPO₄, 0.3 g; KH₂PO₄, 0.3 g; (NH₄)₂SO₄, 0.3 g; NH₄Cl, 1 g; NaCl, 0.61 g; MgSO₄ · 7H₂O, 0.15 g; CaCl₂ · 2H₂O, 80 mg; MnSO₄ · H₂O, 4.5 mg; FeSO₄ · 7H₂O, 3.0 mg; CoSO₄ · 7H₂O, 1.8 mg; ZnSO₄ · 7H₂O, 1.8 mg; CuSO₄ · 5H₂O, 100 μg; AlK(SO₄)₂ · 12H₂O, 180 μg; Na₂MoO₄ · 2H₂O, 100 μg; H₃BO₃, 100 μg; Na₂SeO₄, 1.9 mg; NiCl₂ · 6H₂O, 92 μg; nitrilotriacetic acid, 15 mg; thiamine · HCl, 2.0 mg; d-pantothenic acid, 2.0 mg; nicotinamide, 2.0 mg; riboflavin, 2.0 mg; pyridoxine · HCl, 2.0 mg; biotin, 10.0 mg; cyanocobalamin, 20 μg; p-aminobenzoic acid, 100 μg; folic acid, 50 μg; cysteine · HCl · H₂O, 0.5 g; rumen fluid, 100 ml; sodium acetate, 2.5 g; sodium formate, 2.5 g; trypticase, 2.0 g. Resazurin (1 mg/liter) was added as an oxidation reduction potential indicator. The pH was adjusted to 7 with NaOH prior to gassing with 100% CO₂ and the addition of NaHCO₃. After dispensing into serum bottles and autoclaving under CO₂, a sterile solution containing 0.125 g each of cysteine and sodium sulfide/ml (30 μl per ml of medium) was added prior to inoculation. Incubation was at 37°C.

Table 1 shows the results of MPN analyses of enema samples of patients presenting for flexible sigmoidoscopy. Five tubes of B1C plus the indicated substrate were inoculated with each dilution of the enema samples. Disappearance of substrate in inoculated tubes indicated the presence of hydrolytic bacteria. FP hydrolysis was found in only 3 out of 15 subjects. The concentrations of FP-digesting bacteria were appreciably lower than those for bacteria that used CS or VCP. Our results on the frequency and concentrations of FP-digesting bacteria confirm those of Betian et al. (1) and Wedekind et al. (31). VCP-digesting bacteria were found in 11 of 15 subjects at ca. 16 times higher concentrations than those of FP-digesting bacteria. CS-digesting bacteria were present in all subjects examined, and their concentrations, similar to those previously found by colony enumeration (30), were the highest of the polysaccharide-digesting populations (ca. 45 times higher than those for VCP-digesting bacteria).

TABLE 1.

MPN concentrations of bacteria in human fecal suspensions that use vegetable cellulose, FP cellulose, or starch^a

Substrate	Total no. of subjects	No. positive	Bacteria concn (per g of dry feces) ± SD
Vegetable cellulose	15	11	Log10 9.66 ± 0.70
FP	15	3	Log10 8.45 ± 0.40
Starch	10	10	Log10 11.31 ± 0.41

^aSamples with more than 4 × 10⁶ bacteria per ml of fecal specimen. The dilutions necessary to detect samples with less than 4 × 10⁶ bacteria per ml were not inoculated into MPN tubes.

Isolation.

We isolated a novel gram-positive bacterium after enrichment of 0.5 ml of 10⁻⁶ to 10⁻⁹ dilutions (five tubes per dilution) of one human fecal suspension (28) in B1C plus 0.8% VCP. Hydrolysis was monitored by observation of the disappearance of insoluble VCP. The sample yielded an MPN of 1.1 × 10¹⁰ hydrolytic bacteria per g of dry feces. Transfers were made from a tube with 0.5 ml of the 10⁻⁹ dilution to the same medium with VCP. Growth was also obtained when 2.0% carboxymethylcellulose type 4M6F (CMC) (Hercules Inc., Wilmington, Del.) replaced 0.8% VCP. Transfers from the VCP medium were diluted and plated on WM medium with 2% agar and 0.6% CMC in anaerobic roll tubes (17). WM medium, with mineral concentrations based on those used by McInerney et al. (15), contained (per liter): NaHCO₃, 3.5 g; KH₂PO₄, 0.5 g; NH₄Cl, 1 g; NaCl, 0.4 g; MgCl₂ · 6H₂O, 0.33 g; CaCl₂ · 2H₂O, 50 mg; FeCl₂ · 4H₂O, 1.5 mg; CoCl₂ · 6H₂O, 0.2 mg; ZnSO₄ · 7H₂O, 0.1 mg; MnCl₂ · 4H₂O, 0.03 mg; CuCl₂ · 2H₂O, 0.01 mg; Na₂MoO₄ · 2H₂O, 0.03 mg; H₃BO₃, 0.3 mg; Na₂SeO₄, 1.9 mg; NiCl₂ · 6H₂O, 0.02 mg; thiamine · HCl, 2.0 mg; D-pantothenic acid, 2.0 mg; nicotinamide, 2.0 mg; riboflavin, 2.0 mg; pyridoxine · HCl, 2.0 mg; biotin, 10.0 mg; cyanocobalamin, 20 μg; p-aminobenzoic acid, 100 μg; folic acid, 50 μg; cysteine · HCl · H₂O, 0.5 g; sodium acetate, 1.0 g; isobutyric acid, 0.54 ml; 2-methylbutyric, valeric, and isovaleric acids (0.6 ml each); casein hydrolysate, 2.0 g (Type I, No. C-9386; Sigma Chemical Co., St. Louis, Mo.); and resazurin, 1 mg. Adjustment of pH, the gas phase, NaHCO₃ addition, autoclaving, addition of cysteine and sodium sulfide, and incubation were the same as for B1C.

Pure-culture features.

A culture derived from the transfer of an isolated colony to B1C plus 0.6% CMC was replated on WM plus 2% agar with 0.6% CMC. An isolated colony was transferred and grown on B1C with 2.0% CMC. The isolate, strain I-52^T, consisted of gram-positive, nonmotile short rods (ca. 1.2 by 0.7 μm) that grew mainly in pairs and short chains (Fig. 1). It grew in carbohydrate-containing medium. Poor growth occurred in the absence of formate. Strain I-52^T grew for 48 h in B1C medium with 18.5 mM glucose to an optical density at 660 nm (OD₆₆₀) (1-cm light path) greater than 2.0 with 2.5 or 25.0 mM formate. With 0.0 and 0.25 mM formate, the OD₆₆₀ was 0.32 and 0.46, respectively. The inoculum was from an unwashed culture grown with 30 mM formate. Vanillate (24 mM), but not methanol (78 mM) or 80% H₂-20% CO₂, substituted for formate in the glucose-containing medium. In the presence of 54 mM formate, the isolate grew with added glucose, CMC, vegetable cellulose, stachyose, sucrose, lactose, maltose, galactose, mannose, or xylose. It did not grow with formate and Avicel, FP, lactate, starch, pectin, vanillate, syringate, methanol, 80% H₂-20% CO₂, or ethanol. No growth occurred with 37 mM formate in the absence of carbohydrates.

FIG. 1.

Morphology of strain I-52^T. The strain was grown 18 h in B1C. A wet mount was examined with Nomarski optics. The bar represents 10 μm.

Strain I-52^T was further characterized by using the commercially available API Rapid ID32A (bioMérieux, Inc., Durham, N.C.) system according to the manufacturer's directions. Activity was detected for α-arabinosidase, α-galactosidase, β-galactosidase, β-galactosidase-6-phosphate, α-glucosidase, β-glucosidase, β-glucuronidase, and N-acetyl-β-glucosaminidase. No activity was detected for alkaline phosphatase, arginine arylamidase, arginine dihydrolase, alanine arylamidase, α-fucosidase, glutamic acid decarboxylase, glutamyl glutamic acid arylamidase, glycine arylamidase, histidine arylamidase, leucine arylamidase, leucyl glycine arylamidase, phosphoamidase, phenyl alanine arylamidase, proline arylamidase, pyroglutamic acid arylamidase, serine arylamidase, tyrosine arylamidase, or urease. The organism was indole negative and did not reduce nitrate to nitrite.

The dependence of strain I-52^T growth on the addition of CMC to B1C broth was apparent from cell density measurements. It grew to a maximal OD₆₆₀ of 1.0 (1.8-cm light path) in 24 h in a medium with 2% CMC added as the sole carbohydrate in B1C with added formate. With VCP containing 2 mg of cellulose/ml (26) instead of CMC, 70% (1.4 mg) of the cellulose disappeared after growth for 96 h (26). No cellulose utilization occurred when FP or Avicel was used as a growth substrate. Unfortunately, recent transfers from stock cultures maintained at −76°C for several years did not grow with CMC or the VCP. Other properties of the transfers were identical to those of the original isolate.

Fermentation.

A brief report of the fermentation of the isolate was presented in previous publications (33, 34). Although 2.5 mM formate was sufficient for good growth of strain I-52^T, increasing the formate concentration ca. 20-fold dramatically altered the nature of the products formed from glucose (Table 2). Either without added formate or with 4.8 mM formate, succinate, lactate, and acetate are major products (Table 2). With 48 mM formate, acetate production increased considerably at the expense of succinate and lactate formation (Table 2). No H₂ was detected. Fermentations of glucose were conducted with added NaH¹⁴CO₃ or H¹⁴COONa (18). Schmidt degradation of the acetate produced showed that fermentations with NaH¹⁴CO₃ produced acetate with ¹⁴C almost entirely in the carboxyl group of acetate (Table 3). The methyl group contained almost all of the ¹⁴C in acetate when H¹⁴COONa replaced NaH¹⁴CO₃ (Table 3). If CO₂ was the precursor of both carbons of the third acetate, then one third of all the carbon in the acetate formed by a homoacetogenic fermentation of glucose would be labeled by NaH¹⁴CO₃. If the methyl of the acetate is formed from added formate and not from CO₂, then 1/6 of the acetate carbons (16.7%) would be labeled by H¹⁴COONa and 1/6 (16.7%) would be labeled by NaH¹⁴CO₃. The percentage of the total acetate C that was labeled when H¹⁴COONa was used was 18% of the acetate carbon and 21% when the labeled substrate was NaH¹⁴CO₃. This is consistent with the results of Table 3 that show that the methyl group of labeled acetate is produced from H¹⁴COONa and the carboxyl group is from NaH¹⁴CO₃. Incorporation of H¹⁴COONa into the methyl group and NaH¹⁴CO₃ into the carboxyl group of acetate establishes the operation of the Wood-Ljungdahl pathway in strain I-52^T.

TABLE 2.

Fermentation of glucose with different concentrations of formate^a

Amt of added formate (mM)	Fermentation with the following substrates and products (mM):							C recovery (%)	O/Rb
	Used			Formed
	Glucose	Formate	CO2	Succinate	Acetate	Lactate	CO2
0	14.3	0.0	5.8	11.6	10.0	7.0	0.0	95	1.00
4.8	23.2	4.8	9.3	18.6	23.1	13.0	2.8	106	1.03
48	25.1	27.2	2.2	4.4	70.3	1.6	13.6	98	1.00

^aCultures, inoculated with an unwashed inoculum, were incubated in B1C medium with 28 mM glucose without formate, except as indicated in the table, for 48 h. After gas analysis (23), cultures were acidified and analyzed by high-performance liquid chromatography (9, 18). The amounts of the acid products in the medium immediately after inoculation were similarly determined and were subtracted from the final values. The following equations were used to estimate CO₂ utilization and production: 3 C₆H₁₂O₆ + 4 HCO₂H → 10 CH₃COOH + 2 CO₂ + 2 H₂O; 3 C₆H₁₂O₆ + 2 CO₂ → 4 HOOC(CH₂)₂COOH + 2 CH₃COOH + 2 H₂O; and C₆H₁₂O₆ → 2 CH₃CHOHCOOH.

^bO/R, oxidation equivalents/reduction equivalents.

TABLE 3.

Labeling of acetate by H¹⁴COOH and ¹⁴CO₂^a

Radiolabeled substrate	Sp act of:
	Acetate	Acetate methyl	Acetate carboxyl
H14COOH	762	593	32
14CO2	858	35	682

^aCultures were incubated for 48 h in 10 ml of B1C medium with a solution of 2 μCi of NaH¹⁴CO₃ or 0.5 μCi of Na¹⁴COOH, 30 mM glucose, and 25 mM formate.

Interspecies H₂ transfer between strain I-52^T and Methanobrevibacter smithii strain PS (DSM861^T) did not occur. I-52^T grew by itself with glucose and formate or in a coculture with M. smithii without formate but not with glucose alone. Using a washed inoculum, the OD₆₆₀ (1-cm light path) of cultures grown for 48 h were 0.06, 1.78, and 1.67 with glucose alone, glucose plus formate, and glucose plus methanogen, respectively. Although the addition of the methanogen allowed growth in the absence of formate, analysis of the fermentation products indicate incomplete interspecies H₂ transfer (32), i.e., only a small amount of methane was formed and succinate was a major product. No lactate was formed when the methanogen was present. Some interspecies transfer may have occurred at the expense of lactate but not succinate production. However, the major influence of the methanogen appeared to be the production of a nutrient that substituted for the low formate requirement for growth, and methanogenesis did not use electrons used for succinate formation by strain I-52^T.

The homoacetogenic fermentation of glucose by strain I-52^T with high concentrations of formate is reminiscent of that of Syntrophococcus sucromutans (13). S. sucromutans uses fructose when supplied with formate as an electron-accepting cosubstrate (13). Growth was also obtained with fructose without formate when S. sucromutans was cocultured with M. smithii strain PS. Apparently, S. sucromutans and strain I-52^T cannot reduce CO₂ to formate or produce formate from other nutrients. No formic dehydrogenase was detected in S. sucromutans, although an active CO dehydrogenase was found (5, 14). Formate is probably also used by both organisms for biosynthetic pathways, e.g., purine synthesis (10).

S. sucromutans cannot generate electron acceptors for metabolism of carbohydrates to acetate. It needs added formate or methoxy groups to produce intermediates of the Wood-Ljungdahl reactions that act as electron acceptors for carbohydrate metabolism. Acrylate side chains of benzenoid compounds or interspecies transfer of H₂ can substitute for formate or methoxy groups. In contrast, strain I-52^T produces its own electron sink reactions. It forms acetate, lactate, and succinate from glucose when supplied with low concentrations of formate. However, in the presence of high concentrations of formate, like S. sucromutans, it apparently produces intermediates of the Wood-Ljungdahl reactions and a homoacetogenic fermentation.

Significant steady-state concentrations of formate are not produced by the fermentation of the microbial community of the human colon (35). Added H¹³COOH is mainly converted to ¹³CO₂, and some ¹³C is incorporated into the methyl group of acetate either by direct incorporation or after conversion of formate to CO₂ (35). The batch culture results in this study suggest that strain I-52^T would produce lactate, succinate, and acetate in the colonic environment. Drake (6) pointed out that acetogenesis by most acetogens is conditional and depends on the availability of a reductant and a terminal electron acceptor, including CO_2. The acetogen Peptostreptococcus productus U-1 produces lactate, succinate, and acetate under CO₂-limited conditions, and CO₂ enrichment increases acetate formation and decreases the formation of lactate and succinate (7, 19). Batch growth of strain I-52^T may be initiated with a homoacetogenic fermentation, with transition to other products when formate drops to much lower levels than the added 2.5 mM. Continuous cultures with varying levels of formate should reveal the transition point and aid in examining the mechanism of regulation of carbon flow to the alternative electron sink pathways.

Molecular characterization.

The G+C content of the DNA was determined as described previously (16), except the methanol content of the chromatographic buffer was 8% and the temperature was 37°C. The G+C content was 50.3 mol%. An almost complete fragment of the 16S rRNA gene (ca. 1,450 bases) of strain I-52^T was amplified from DNA by PCR using universal primers pA (positions 8 to 28; Escherichia coli numbering) and pH* (positions 1542 to 1522) and was directly sequenced using a Taq DyeDeoxy terminator cycle sequencing kit (Applied Biosystems, Foster City, Calif.) and an automatic DNA sequencer (model 373A; Applied Biosystems). The 16S rRNA gene fragments were generated by PCR and were sequenced as described by Hutson et al. (12). The closest known relatives of the new isolate were determined by database searches using the program FASTA (24). These sequences and those of other known related strains were retrieved from GenBank and were aligned with the newly determined sequences using the program DNATools (25). The resulting multiple-sequence alignment was corrected manually by using the program GeneDoc (20). A phylogenetic tree was constructed according to the neighbor-joining method with the programs DNATools and TREEVIEW (21), and the stability of the groupings was estimated by bootstrap analysis (1,000 replications) with the same programs. The 16S rRNA gene sequence of strain I-52^T has been deposited in GenBank under accession number AJ318527. Sequence database searches revealed that strain I-52^T was phylogenetically a member of the Clostridium subphylum of the gram-positive bacteria. Upon treeing analysis, the new isolate formed a hitherto unknown line of descent within the Clostridium coccoides rRNA group of organisms (Fig. 2). The sequence similarity to the nearest phylogenetic relative, C. xylanolyticum, was 92.5%. Sequence similarity comparisons revealed that strain I-52^T was only distantly related to other species within the C. coccoides group, with sequence divergence values of 7% or greater shown with all presently described members of this supra-generic grouping.

FIG. 2.

Unrooted tree showing the phylogenetic relationships of strain I-52^T within the C. coccoides rRNA grouping. The tree, constructed by using the neighbor-joining method, was based on a comparison of ca. 1,300 nucleotides. Bootstrap values, expressed as a percentage of 1,000 replications, are given at the branching points.

Morphologically, the short rod-shaped strain I-52^T somewhat resembles S. sucromutans but differs from the latter in staining gram positive and not requiring large amounts of rumen fluid for growth (13). Phylogenetically, strain I-52^T, like S. sucromutans, is a member of the C. coccoides rRNA supra-generic grouping. However, both sequence divergence (>10%) and treeing analysis show these organisms are phylogenetically only distantly related. Strain I-52^T forms a distinct line of descent and does not display a particularly close or a statistically significant phylogenetic affinity with any described species.

In addition to being distinct from S. sucromutans, strain I-52^T is phenotypically different from the plethora of other genera found within the C. coccoides rRNA complex. For example, in addition to its unusual homoacetogenic fermentation, strain I-52^T can be distinguished from Clostridium spp. and Sporobacterium in not producing endospores, from Lachnospira by the absence of curved cellular shapes, from Roseburia and Butyrivibrio in end products of glucose metabolism (i.e., not producing butyric acid), and in being nonmotile. It differs from Coprococcus and Ruminococcus in cellular morphology and in end products of glucose metabolism. Strain I-52^T is also metabolically distinct from the numerous putative Eubacterium species currently found within the C. coccoides rRNA complex. The putative Eubacterium species invariably displayed >10% sequence divergence with strain I-52^T and therefore cannot be considered members of the same genus. It is clear that the novel acetogen reported here is both phenotypically and phylogenetically incompatible with all recognized genera within the C. coccoides rRNA cluster, and it merits classification as a new genus. Therefore, we propose the unknown rod-shaped bacterium be classified as a new genus and species, Bryantella formatexigens.

Description of Bryantella.

Bryantella (Bry. an. tel'la. N.L. fem. n., named after the American microbiologist Marvin P. Bryant in recognition of his outstanding contributions to the microbial ecology of anaerobic ecosystems). It consists of short rod-shaped cells. Gram-positive, nonmotile, and non-spore forming. Anaerobic. Catalase and oxidase negative. Does not require rumen fluid for growth. Acetate is the sole product of glucose fermentation when grown in the presence of high concentrations of formate. Glucose fermentation with low concentrations of formate yield succinate, lactate, and acetate. Indole negative. Nitrate is not reduced. The G+C content of DNA is 50.3 mol%. The type species is Bryantella formatexigens.

Description of Bryantella formatexigens sp. nov.

Bryantella formatexigens (for.mat.ex′i.gens. N.L. neut. n. formatum formate, L. part. adj. exigens demanding, N.L. adj. formatexigens formate-demanding).

Cells consist of nonmotile, short rods (ca. 1.2 × 0.7 μm) which occur mainly in pairs and short chains. Gram positive. Strictly anaerobic chemoorganotroph. Catalase-negative. Does not require rumen fluid for growth. Cellulolytic but may lose this activity upon prolonged storage at −76°C. Acetate is the sole product of glucose fermentation when grown in the presence of high concentrations of formate. Glucose fermentation with low concentrations of formate yield succinate, lactate, and acetate. In the presence of formate (54 mM), strain I-52^T grows with added glucose, vegetable cellulose preparation, carboxymethylcellulose, stachyose, sucrose, lactose, maltose, galactose, mannose, or xylose. Does not grow with formate and FP cellulose, Avicel, lactate, starch, pectin, vanillate, syringate, methanol, ethanol, or H₂-CO₂. No growth with 0.25% formate in the absence of carbohydrates. By using the commercially available API Rapid ID32A system, activity is detected for α-arabinosidase, α-galactosidase, β-galactosidase, β-galactosidase-6-phosphate, α-glucosidase, β-glucosidase, β-glucuronidase, and N-acetyl-β-glucosaminidase. No activity is detected for alkaline phosphatase, arginine arylamidase, arginine dihydrolase, alanine arylamidase, α-fucosidase, glutamic acid decarboxylase, glutamyl glutamic acid arylamidase, glycine arylamidase, histidine arylamidase, leucine arylamidase, leucyl glycine arylamidase, phosphoamidase, phenyl alanine arylamidase, proline arylamidase, pyroglutamic acid arylamidase, serine arylamidase, tyrosine arylamidase, or urease. Indole negative. Nitrate is not reduced. The G+C content of DNA is 50.3 mol%. The type strain is I-52^T = DSM 14469^T = CCUG 46960^T. Isolated from human feces.