Genetic Diversity of Hydrogen-Producing Bacteria in an Acidophilic Ethanol-H₂-Coproducing System, Analyzed Using the [Fe]-Hydrogenase Gene

Abstract

Hydrogen gas (H₂) produced by bacterial fermentation of biomass can be a sustainable energy source. The ability to produce H₂ gas during anaerobic fermentation was previously thought to be restricted to a few species within the genera Clostridium and Enterobacter. This work reports genomic evidence for the presence of novel H₂-producing bacteria (HPB) in acidophilic ethanol-H₂-coproducing communities that were enriched using molasses wastewater. The majority of the enriched dominant populations in the acidophilic ethanol-H₂-coproducing system were affiliated with low-G+C-content gram-positive bacteria, Bacteroidetes, and Actinobacteria, based on the 16S rRNA gene. However, PCR primers designed to specifically target bacterial hydA yielded 17 unique hydA sequences whose amino acid sequences differed from those of known HPB. The putative ethanol-H₂-coproducing bacteria comprised 11 novel phylotypes closely related to Ethanoligenens harbinense, Clostridium thermocellum, and Clostridium saccharoperbutylacetonicum. Furthermore, analysis of the alcohol dehydrogenase isoenzyme also pointed to an E. harbinense-like organism, which is known to have a high conversion rate of carbohydrate to H₂ and ethanol. We also found six novel HPB that were associated with lactate-, propionate-, and butyrate-oxidizing bacteria in the acidophilic H₂-producing sludge. Thus, the microbial ecology of mesophilic and acidophilic H₂ fermentation involves many other bacteria in addition to Clostridium and Enterobacter.

Hydrogen gas (H₂), a clean fuel whose only reaction product with oxygen is H₂O, may become an important component of the global energy economy (7, 48). Because the cost of H₂ produced by electrolysis and thermochemical decomposition of water is high, most H₂ is produced today almost exclusively by reforming natural gas, oil, and coal (34), an approach that is neither renewable nor carbon neutral. Furthermore, reforming takes place at high temperature and pressure and produces carbon monoxide, which fouls H₂ fuel cells (54). In contrast, biological H₂ production occurs at ambient conditions, does not release carbon monoxide, and uses renewable organic matter as the H₂ source (10, 46). Fermentative H₂ production from wastewater and other biomass promises to be an economical and sustainable technology if conversion efficiencies can be increased (5, 48). Use of a natural mixed inoculum, such as sewage sludge, and continuous operation can maximize the cost advantages (27, 28, 36, 39, 47).

The conventional wisdom is that fermentative H₂-producing bacteria (HPB) are restricted to a few genera, such as Clostridium and Enterobacter (20, 32), which lose the ability to produce H₂ at pHs below 5 to 5.5 (19). However, a few studies find significant H₂ production by anaerobic sludge at low pH values of 4.0 to 4.5, and the soluble organic fermentation products are mainly ethanol and acetic acid (37, 38). These findings break with the conventional concept that H₂ production occurs by butyric-type and mixed-acid-type fermentation at pH values higher than 5.5 (15, 19). They imply that unknown and novel HPB are present in the acidophilic communities.

Until now, most research on communities involving HPB has been based on culture-dependent isolation and 16S rRNA gene analysis. However, functional genes have been exploited as molecular markers to monitor environmental microorganisms, and they offer advantages over phylogenetic studies solely based on 16S rRNA genes (21, 44). Hydrogenases (H₂ases) catalyze the reversible formation of H₂ from protons and electrons according to the half-reaction H₂↔ 2H⁺ + 2e⁻ (33, 45). According to the metal cofactors at the active sites, H₂ases are classified into three major groups: NiFe or NiFeSe, Fe only, and metal free (49). NiFe varieties are usually found in microorganisms that consume H₂ (49). Wawer and Muyzer (1995) reported the genetic diversity of Desulfovibrio spp. in environmental samples by use of [NiFe]-H₂ase gene fragments and demonstrated that genes encoding H₂ases could be biomarkers for genus- or species-specific detection of bacteria (50). Fe-only H₂ases have been identified in a small group of microbes, where they often catalyze the reduction of protons as terminal electron acceptors to produce H₂ (1, 8, 35). Hence, genes encoding [Fe]-H₂ases (hyd) can be specific biomarkers of HPB. They have been divided into hydA (large-subunit) and hydB (small-subunit) classes. Chang et al. specifically detected clostridia in an anaerobic H₂ biofermentation system by use of [Fe]-hydA (9). However, specific detection of all fermentative HPB in microbial communities based on universal [Fe]-hydA primers has not been reported. A metal-free H₂ase has been discovered in some methanogens (49), but it does not produce H₂.

Here we use a range of DNA-, RNA-, and enzyme-directed tools to understand the diversity and dynamics of H₂-producing communities that are enriched under acidophilic conditions. In particular, this study exploits unique PCR primers targeting hydA to document the distribution and diversity of HPB in acidophilic communities. It also reports that diverse ethanol-H₂-coproducing bacteria (EHCB) and fatty acid-oxidizing HPB are enriched in the acidophilic sludge.

MATERIALS AND METHODS

Reactor operation.

Two continuous stirred-tank reactors (CSTRs), coded A and B, were operated. The liquid volume was 3 liters, the temperature was 35 ± 1°C, and the pH was maintained at 4.5 to 5.0 by a pH control system. Molasses wastewater from a beet sugar refinery was used as a substrate, and the inoculum was obtained from domestic sewage (Wenchang domestic sewage treatment plant, Harbin, China). The organic loading rate of the reactors was increased from 5 kg chemical oxygen demand (COD)·m⁻³·day⁻¹ to 30 kg COD/m³·day during the experiment by increasing the influent COD in five steps, with the hydraulic retention time held constant at 6 h.

Chemical analyses.

H₂ was measured using a gas chromatograph (SC-2; Shanghai Analytical Instruments Co., Ltd.) equipped with a thermal conductivity detector and a molecular sieve column (Porapak Q 50/80) with nitrogen as the carrier gas. Carbon dioxide and methane were measured similarly, except that a different column was used (Porapak Q 80/100), with helium as the carrier gas. In each case, the size of the injected sample was 0.5 ml. We analyzed volatile fatty acids (VFAs) and ethanol with a gas chromatograph (Agilent 6890N) equipped with a flame ionization detector and a fused-silica capillary column (DB-FFAP), with helium as the carrier gas. Procedures described in Standard Methods for the Examination of Water and Wastewater, 18th ed. (16), were used to determine the COD and levels of volatile suspended solids (VSS).

DNA and RNA extraction.

Samples of anaerobic activated sludge were collected from the two reactors at days 0, 7, 14, 21, 28, 35, and 42. Genomic DNA and total RNA were extracted and purified using the PowerSoil DNA isolation kit and the PowerSoil total-RNA isolation kit (Mo Bio Laboratories, Inc.) according to the manufacturer's instructions.

PCR amplification of the 16S rRNA gene.

PCR amplification was carried out in 50-μl volumes containing 12.5 μM each primer, 200 μM each deoxyribonucleoside triphosphate, 5 μl of 10× PCR buffer (100 mM Tris-HCl, 15 mM MgCl₂, 500 mM KCl [pH 8.3]), 0.5 U of Taq DNA polymerase (Promega, Madison, WI), 100 ng of the DNA extract, and sterile deionized water to make up the total volume of 50 μl. The 16S rRNA primers used for denaturing gradient gel electrophoresis (DGGE) analysis were BSF338 (5′-ACTCCTACGGGAGGCAGCAG-3′; E. coli 16S rRNA positions 338 to 354), which was attached to a GC clamp (CGCCCGCCGCGCCCCGCGCCCGTCCCGCCGCCCCCGCCCG) at the 5′ terminus, and BSR534 (5′-ATTACCGCGGCTGCTGG-3′; E. coli 16S rRNA positions 517 to 534) (11, 38). The primers used for 16S rRNA gene clone libraries were BSF341 (5′-CCTACGGGAGGCAGCAG-3′; E. coli 16S rRNA positions 341 to 357) and BSR926 (5′-CCGTCAATTYYTTTRAGTTT-3′; E. coli 16S rRNA positions 917 to 926). The samples were amplified in a GenAmp PCR system 9700 (Perkin-Elmer Applied Biosystems, Foster City, CA) programmed as follows: initial denaturation of DNA for 5 min at 94°C; 30 cycles of 1 min at 94°C, 30 s at 60°C (55°C for BSF341 and BSR926), decreasing 0.1°C per cycle to 57°C (52°C for BSF341 and BSR926), and 30 s at 72°C; and extension of incomplete products for 30 min at 72°C. PCR products were examined by electrophoresis on a 2% (wt/vol) agarose gel containing ethidium bromide (0.5 μg/ml). Blank controls were carried out for all steps.

RT-PCR amplification of hydA.

Amino acid sequences of genes known to encode [Fe]-H₂ase (33 sequences of 14 genera) from the GenBank database were aligned by the ClustalW program and identified by two conserved regions, ADLTIMEE and EVMACPGGCI (see Fig. S1 and Table S1 in the supplemental material). According to the consensus region of [Fe]-H₂ase amino acid sequences, two degenerate universal primers, hydF1 (5′-GCCGACCTKACMATMATGGA-3′) and hydR1 (5′-ATRCARCCRCCSGGRCAGGCCAT-3′), were designed. The phosphoric acid of the 5′ end of the primers was removed to avoid improper annealing between the primers and the DNA template. For hydA DGGE analysis, hydA segments were recovered from activated sludge by 5′-phosphate-modified primers in the first amplification; then a second amplification was carried out using the regular primer hydF1 (5′ end not modified), which was attached to a GC clamp. Reverse transcription-PCR (RT-PCR) was performed using the Promega (Madison, WI) Reverse Transcription System according to the manufacturer's instructions. The reverse transcription samples were amplified under the following conditions: initial denaturation for 5 min at 94°C; 10 cycles of 1 min at 94°C, 90 s at 47°C, and 90 s at 72°C; 20 cycles of 1 min at 94°C, 90 s at 55°C, and 90 s at 72°C; and extension of incomplete products for 30 min at 72°C. The PCR products were examined by electrophoresis on a 1% (wt/vol) agarose gel containing ethidium bromide (0.5 μg/ml). Blank controls were carried out for all steps.

DGGE analysis.

DGGE was performed with a DCode universal mutation detection system (Bio-Rad Laboratories, Hercules, CA). A double-gradient gel was used for analyzing amplified 16S rRNA gene products. A second gradient of 6 to 12% polyacrylamide (acrylamide/bisacrylamide ratio, 37.5:1) together with a 30 to 60% denaturing gradient was superimposed (11, 31). One hundred percent denaturation corresponds to 7 M urea and 40% (vol/vol) deionized formamide. DGGE analysis for the RT-PCR products of the hydA gene used 8% (wt/vol) polyacrylamide gels with a denaturing gradient ranging from 30 to 60%. The gradient gel was cast with a gradient delivery system (model 475; Bio-Rad). Approximately 1 μg of PCR or RT-PCR products per lane was loaded onto DGGE gels.

Electrophoresis was run for 15 min at 25 V and for 3.5 h at 200 V (for the 16S rRNA gene) or 11 h at 75 V (for hydA) in 1× Tris-acetate-EDTA buffer maintained at 60°C. The gels were silver stained according to the method of Bassam et al. (3). Prominent DGGE bands were selected and excised for nucleotide sequencing. The gel was crushed in 50 μl TE buffer (10 mM Tris-HCl, 1 mM EDTA [pH 8.0]), and the mixture was allowed to equilibrate overnight at 4°C. After the slurry was centrifuged at 5,000 × g for 1 min, we took 1 μl of buffer containing DNA as the template for a PCR performed under the conditions described above for sludge samples, except that the forward primer lacked the GC clamp. The PCR products were purified with a Gel Recovery purification kit (Watson Biotechnologies Inc., Shanghai, China) and cloned into Escherichia coli JM109 using the pGEM-T plasmid vector system (Promega, Madison, WI) in accordance with the manufacturer's instructions. Ten clones from each band were randomly chosen for reamplification with the GC clamp. Five microliters of reamplification product from each clone was subjected to DGGE analysis as described above for sludge samples in order to check the purity and to confirm the melting behavior of the band recovered. If the bands from the clones were identical with the DGGE parents' bands, these clones from the same band were sequenced to estimate the numbers of 16S rRNA and hydA sequences comigrating on the DGGE band.

Clone library and sequencing analysis.

The PCR or RT-PCR products were purified, ligated into vector pCR2.1 using a TOPO TA cloning kit (Invitrogen, Carlsbad, CA), and cloned into chemically competent One Shot Escherichia coli cells, provided with the cloning kit, according to the manufacturer's instructions. From these transformants, clone libraries of the 16S rRNA and hydA genes from the 42-day sample of each reactor were constructed. For each reactor, five libraries were constructed from independent PCR or RT-PCR amplifications of the extracted DNA and RNA. Fifty clones for the 16S rRNA gene and 30 clones for the hydA gene were randomly picked and identified from each library. Thus, in total, 250 16S rRNA gene clones and 150 hydA gene clones from each sample were analyzed. Positive clones were screened by colony PCR with vector-specific primers as described previously (40). The cloned PCR fragments were sequenced with an ABI Prism model 3730 automatic sequencer (Perkin-Elmer, Foster City, CA).

Phylogenetic analysis.

16S rRNA sequences were analyzed against the GenBank database and Ribosomal Database Project II (RDP II; http://rdp.cme.msu.edu). All sequences were examined for chimerism using the CHECK_CHIMERA program at RDP II and BELLEROPHON (http://foo.maths.uq.edu.au/∼huber/bellerophon.pl) (22). The partial sequences of the hydA gene were aligned with the same region of the most closely related strains available in the GenBank database by using the ClustalW function of the BioEdit package (17). Neighbor-joining phylogenetic trees of 16S rRNA and hydA partial sequences were constructed with the molecular evolutionary genetics analysis package (MEGA, version 3.1) and the Jukes-Cantor algorithm (25). A bootstrap analysis with 1,000 replicates was carried out to check the robustness of the tree.

Isoenzyme activity assay.

Bacteria from as much as 1 ml of anaerobic activated sludge were harvested by centrifugation and resuspended in phosphate-buffered saline, pH 7.4, consisting of 8 g of NaCl, 0.2 g of KCl, 1.44 g of Na₂HPO₄, and 0.24 g of K₂HPO₄ per liter of distilled water. The pellets were washed three times in phosphate-buffered saline and then mixed with extraction buffer (100 mM Tris-HCl [pH 8.0], 5 mM dithiothreitol, and 0.1% β-mercaptoethanol) at 4°C. After the cells were broken by sonication (Sonics; VC130PB) and centrifugation, the proteins were analyzed in the supernatant by electrophoresis. The protein concentrations of crude extracts were estimated by the Bradford method, using a Coomassie blue-based reagent and bovine serum albumin (New England Biolabs Inc.) as standards (6).

Native gels containing 7.5% polyacrylamide were prepared in 1× TBE (90 mM Tris-borate, 1 mM EDTA [pH 8.0]), with each lane loading 3 mg of protein. Electrophoresis was carried out in a Protean II xi cell (Bio-Rad) equipped with a power supply (model PAC1000; Bio-Rad) under constant conditions (4°C, 64 mA) for 3 h in 1× running buffer (25 mM Tris, 192 mM glycine). After electrophoresis, the gels were soaked in an alcohol dehydrogenase (ADH) activity staining solution of 0.6 mM NAD⁺, 0.4 mM nitroblue tetrazolium, 0.15 mM phenazine methanosulfonate, 1 mM Tris-HCl (pH 7.5), and 4% ethanol for 30 min at 37°C in the dark. After staining, the gels were washed in distilled water.

Statistical analysis.

Phylotypes from clone libraries were determined with the DOTUR program; sequences were grouped into phylotypes at a cutoff value (≥99% similarity) using the furthest-neighbor algorithm with 0.001 precision (41). Collector's curves of observed and estimated richness and Shannon and Simpson diversity indices were calculated using the DOTUR and EstimateS 7.5 (http://viceroy.eeb.uconn.edu/estimates) programs. The differences between the clone libraries of the two reactors were determined using the ∫-LIBSHUFF program (42).

Nucleotide sequence accession numbers.

The sequences reported in this paper have been deposited in the GenBank database under accession numbers DQ074125 to DQ074147, DQ298807 to DQ298832 [hydA], DQ464458 to DQ464598, and DQ482726 to DQ482730.

RESULTS

Enrichment culture of HPB.

Figure 1 shows the progress of the fermentation reactions for the 42-day experiments in the two CSTRs. The fermentation gas was composed of CO₂ and H₂, without CH₄. Figure 1A and B show that the H₂ production rate and H₂ conversion efficiency were too low to measure before 7 days. After 14 days, the H₂ production rate and H₂ conversion efficiency increased quickly. The H₂ conversion efficiency began to level off around day 28. On day 42, the H₂ production rates reached 0.45 liter·g of VSS⁻¹·day⁻¹ (reactor A) and 0.48 liter·g of VSS⁻¹·day⁻¹ (reactor B). The H₂ conversion efficiencies from glucose were 0.18 liter of H₂·g⁻¹ of COD removed (reactor A) and 0.21 liter of H₂·g⁻¹ of COD removed (reactor B), which convert to 0.13 and 0.15 g of oxygen demand (OD) in H₂·g⁻¹ of OD in glucose.

FIG. 1.

Fermentation products in the two reactors. (A and B) H₂ production in reactors A and B, respectively. •, H₂ production rate; ○, H₂ conversion efficiency. (C and D) Concentrations of VFAs and alcohol and pHs in reactors A and B, respectively. ○, acetic acid; ▪, propionic acid; □, butyric acid; ▴, valeric acid; •, ethanol; ▵, lactate; ▿, pH. Error bars, means ± standard deviations obtained from three replications.

The main soluble organic fermentative products (Fig. 1C and D) were ethanol, acetate, propionate, butyrate, valerate, and lactate; the contents of VFAs and ethanol had the same patterns in the two duplicate reactors. At the beginning of the experiment, the millimole ratios of ethanol, acetate, propionate, butyrate, valerate, and lactate were 0.7:8.2:3.7:1.6:0.8:1.1 (reactor A [Fig. 1C]) and 0.4:7.8:4.2:1.2:0.3:0.9 (reactor B [Fig. 1D]); thus, the main liquid end products were acetate, propionate, and butyrate. Butyrate levels had begun to increase gradually after 7 days and then somewhat decreased. After 14 days, ethanol and acetate levels increased quickly, becoming the dominant products long before day 42, when the millimole ratios of ethanol, acetate, propionate, butyrate, valerate, and lactate were 29.1:22.5:3.0:3.0:1.8:4.8 (reactor A) and 32.7:27.5:2.3:4.7:0.9:3.5 (reactor B). Therefore, the bacterial communities progressed to ethanol-type fermentation that produced ethanol and acetate as its main liquid end products, and the pH values were 4.52 and 4.45 by the end of the experiment (Fig. 1C and D).

Bacterial community structure.

Preliminary experiments indicated that Archaea and Fungi were not found in the microbial community by PCR amplification with universal primers of small-subunit rRNA (data not shown). Detailed results for the changes in bacterial community structure as assessed by DGGE of the 16S rRNA gene are shown in Fig. 2. In summary, the number of clearly distinguishable DGGE bands of the 16S rRNA gene increased markedly between days 14 and 21, when the fermentation type was becoming established (Fig. 1). After 21 days, new dominant bands appeared and some previously prominent bands disappeared, but the bands within the DGGE profiles of the 16S rRNA gene were relatively constant after day 28.

FIG. 2.

DGGE profiles of the PCR-amplified V3 regions of the 16S rRNA genes in the microbial communities from two reactors. (A) Reactor A; (B) reactor B. Lanes are labeled with the time of sampling. Arrows indicate the DGGE bands selected for cloning and sequencing.

Fourteen prominent bands from the 16S rRNA gene DGGE gel were excised and sequenced. The phylotype (or operational taxonomic unit [OTU]) was determined by using 99% minimum similarity as the threshold (13). The results show that one band often was not made up of a unique phylotype, and the sequences of the 14 bands fell into 23 distinct OTUs. Comparison with the 16S rRNA genes derived from cultivated strains in the GenBank and RDP II databases showed that 11 of the 23 OTUs (47.8%) had sequence similarities of >97% with known strains (Table 1). The closest affiliations were those with Saccharothrix, Clostridium, Leuconostoc, Bifidobacterium, Lactobacillus, Lactococcus, Megasphaera, Mitsuokella, and Atopobium spp. After 28 days, the intensities of bands 9, 12, and 14 from the 16S rRNA gene DGGE became strong; they were affiliated with Megasphaera elsdenii (accession no. U95028; 100% similarity), Acetanaerobacterium elongatum (AY487928; 96.5%), and Ethanoligenens harbinense (AY295777; 96.6%). Since these bands became important when ethanol-type H₂ production was established, they are the putative H₂-producing microorganisms.

TABLE 1.

Phylogenic affiliations of DGGE bands based on the V3 region of the 16S rRNA gene

Banda	Accession no.	Size (bp)	Most closely related bacterial sequence		Similarity (%)
			Species and strain	Accession no.
1-1	DQ074125	173	Anaerofustis stercorihominis WAL 14563	AJ518871	89.0
1-2	DQ074126	177	Saccharothrix australiensis NRRL 11239	AF114803	97.2
1-3	DQ074127	172	Clostridium acetobutylicum NCIMB 8052	X81021	98.3
2-1	DQ074128	172	Paenibacillus hongkongensis HKU3	AF433165	95.3
2-2	DQ074129	173	Anaerofustis stercorihominis WAL 14563	AJ518871	88.4
2-3	DQ074130	177	Brevibacillus thermoruber DSM 7064	Z26921	91.0
3-1	DQ074131	197	Leuconostoc pseudomesenteroides RO1	AF515228	98.5
4-1	DQ074132	177	Bifidobacterium minimum T20	AY174103	100
5-1	DQ074133	192	Bulleidia moorei AHP 13983	AY044915	91.7
5-2	DQ074134	198	Megasphaera paucivorans VTT E-032341	DQ223730	95.5
6-1	DQ074135	197	Lactobacillus satsumensis	AB154519	99.0
			Lactobacillus sp. strain 71	AY681129
6-2	DQ074136	198	Lactococcus lactis Akira2	AY348313	97.5
6-3	DQ074137	192	Prevotella albensis M384	AJ011683	92.2
7-1	DQ074142	192	Prevotella albensis M384	AJ011683	93.8
8-1	DQ074143	198	Megasphaera hominis	L79909	100
9-1	DQ074144	198	Megasphaera elsdenii S2	U95028	100
			Megasphaera elsdenii S3	U95029
10-1	DQ074145	199	Mitsuokella multacida NCTC 10934	X81878	99.0
10-2	DQ074146	197	Lactobacillus zeae ATCC15820	D86516	100
			Lactobacillus casei BJ G23-3	AY244628
			Lactobacillus rhamnosus GG ATCC 53103	AY370682
11-1	DQ074147	192	Prevotella ruminicola L16	AY699286	95.3
12-1	DQ074138	174	Acetanaerobacterium elongatum Z7	AY487928	96.5
13-1	DQ074139	198	Selenomonas lacticifex DSM20757	AF373024	94.4
13-2	DQ074140	179	Atopobium vaginae VA14183_00	AF325325	98.9
14-1	DQ074141	174	Ethanoligenens harbinense YUAN-3	AY295777	96.6

^aThe number before the hyphen represents the band excised from the DGGE gel. The number after the hyphen represents a different clone of the same band.

Two hundred twenty-seven and 208 positive clones containing 16S rRNA gene inserts were obtained from 10 clone libraries from the A and B sludge specimens, respectively, at day 42, whereas 59 and 87 OTUs were obtained for the same two reactors by the DOTUR program (Fig. 3). Coverages were 87.2% and 84.1% for reactors A and B, respectively. Twenty OTUs (33.9%) from reactor A and 10 OTUs (11.5%) from reactor B had sequence similarities of >97% with known strains (see Table S2 in the supplemental material). Only nine OTUs affiliated with putative HPB (Ethanoligenens sp. and Megasphaera sp.) were identified by use of 16S rRNA gene clone libraries from both reactors. Phylogenetic analysis indicated that the dominant populations could be divided into five groups (Fig. 4): low-G+C-content gram-positive bacteria, Bacteroidetes, Actinobacteria, a few Alphaproteobacteria, and unclassified bacteria. Relative abundances indicated that low-G+C-content gram-positive bacteria (53.0% in reactor A and 63.7% in reactor B), Bacteroidetes (39.7% and 15.4%, respectively), and Actinobacteria (7.0% and 14.4%, respectively) constituted the majority (Fig. 5). The microbial community appeared more diverse in reactor B than in reactor A, based on Shannon and Simpson (1/D) diversity indices, which were, respectively, 2.8 and 6.3 in reactor A and 4.2 and 50.0 in reactor B. Only 15 OTUs were shared by the two libraries, and the community structures of the two libraries were distinct according to the ∫-LIBSHUFF program (P < 0.05). We obtained a total of 131 OTUs in both reactors.

FIG. 3.

Collector's curves of observed and estimated phylotype richness of 16S rRNA gene clone libraries by the DOTUR program. (A) Reactor A; (B) reactor B. Estimator curves include observed OTUs (bottom), Chao1 (middle), and ACE (abundance-based coverage estimator) (top). Phylotypes were defined using the 99% OTU cutoff.

FIG. 4.

Phylogenetic tree derived from 16S rRNA gene clone libraries. Each row represents a different OTU (phylotype). The phyla are color coded as follows, from top to bottom: green, low-G+C-content gram-positive bacteria; blue, Actinobacteria; black, unclassified; yellow, Alphaproteobacteria; red, Bacteroidetes. The relative abundances of phylotypes from the 16S rRNA gene clone libraries of the two reactors are shown on the right in grayscale values. Letters above the abundance graph correspond to two different enrichments.

FIG. 5.

Relative abundances of sequences from the 16S rRNA gene libraries of two reactors (A and B). The sequence frequencies are grouped according to phylum. “Unknown” sequences are unclassified.

Despite differences between the two communities overall, the dominant H₂-producing populations in the 16S rRNA gene libraries and the DGGE profiles were identical. Putative H₂-producing populations were affiliated with Ethanoligenens and Megasphaera spp. by both methods.

Distribution of hydA clones.

Compelling evidence for the enrichment of HPB was obtained by the identification of hydA. DGGE profiles of hydA expression showed that the H₂-producing populations of the two reactors shifted over time (Fig. 6). RT-PCR amplification of hydA was faint on day 0. After 7 days, the intensity and diversity of bands increased. Five prominent bands from the DGGE gel of hydA were excised, reamplified, cloned, and sequenced. The amino acid sequences of the bands were affiliated with the hydA genes of Syntrophomonas wolfei-like, M. elsdenii-like, C. thermocellum-like, and E. harbinense-like organisms (see Table S3 in the supplemental material).

FIG. 6.

DGGE profiles of partial hydA genes in the microbial communities from two reactors (A and B, respectively). Lanes are labeled with the time of sampling. Arrows with numbers indicate the DGGE bands selected for cloning and sequencing. The initial RNA template was standardized for RT-PCR. The PCR product was loaded for every lane and was not standardized.

One hundred thirty and 132 positive clones containing hydA inserts were obtained from 10 clone libraries from sludge specimens of both reactors at day 42, whereas 11 and 10 hydA sequences were identified from the two reactors by the DOTUR program (based on a 1% cutoff). Coverages were 98.4% and 99.2% for reactors A and B, respectively. A phylogenetic tree was constructed based on partial [Fe]-H₂ase amino acid sequences from the DGGE bands and clone libraries. All hydA sequences fell into six clusters, affiliated with Clostridium saccharoperbutylacetonicum-like hydA (group I; 48.9% similarity with a known hydA gene), Clostridium thermocellum-like hydA (group II; 58.6% and 60.9% similarities with known hydA genes), Syntrophomonas wolfei-like hydA (group III; 57.1%, 60%, and 60.6% similarities), Syntrophobacter fumaroxidans-like hydA (group IV; 61.9% similarity), Megasphaera elsdenii-like hydA (group V; 77.8% similarity), and E. harbinense-like hydA (group VI; 62.9%, 63.3%, 65.2%, 71.9%, 85.1%, and 90.5% similarities), respectively (Fig. 7). Comparison of the community structures indicates that the H₂-producing communities had no distinct difference based on ∫-LIBSHUFF (P > 0.05). The Shannon and Simpson (1/D) diversity indices of HPB were, respectively, 1.9 and 5.0 for reactor A and 2.1 and 7.7 for reactor B. Thus, based on hydA, the community in reactor B was slightly more diverse. We found 4 phylotypes (of the 21 phylotypes) affiliated with the hydA genes of C. thermocellum (58.6%), S. wolfei (60.6%), M. elsdenii (77.8%), and E. harbinense (65.2%) in the DGGE bands and the two clone libraries (see Table S3 in the supplemental material). Therefore, all 17 unique hydA sequences appeared in both reactors.

FIG. 7.

Phylogenetic tree derived from alignments of partial amino acid sequences of [Fe]-H₂ases, including the sequences of DGGE bands and clone libraries and sequences from the database. GenBank accession numbers are given in parentheses. Bootstrap values of >50% for neighbor joining are shown (percentages of 1,000 resamplings). Bar indicates 1% divergence. ○, sequences from the hydA clone library of reactor A; •, sequences from the clone library of reactor B; □, bands excised from the DGGE gel (see Fig. 6). Numbers after the hyphen (b1 to b5) represent bands. Different clusters of hydA derived from the two reactors are highlighted by different colors.

Although these HPB were previously undescribed, they are affiliated with E. harbinense, C. thermocellum, S. wolfei, M. elsdenii, C. saccharoperbutylacetonicum, and S. fumaroxidans. E. harbinense-like (46.1%) and C. thermocellum-like (38.5%) organisms accounted for 84.6% of the organisms from reactor A; together, they accounted for 75.8% (E. harbinense-like organism, 39.4%; C. thermocellum-like organism, 36.4%) of those from reactor B (Fig. 8). Among these putative HPB, E. harbinense, C. thermocellum, and C. saccharoperbutylacetonicum were identified as EHCB according to previous studies (24, 51, 52, 53), but M. elsdenii, S. fumaroxidans, and S. wolfei could produce H₂ using lactate, propionate, and butyrate as the sole carbon sources according to previous descriptions (12, 14, 18, 29, 30).

FIG. 8.

Clone libraries of partial hydA genes from reactors A and B. Bar graph shows percentages of Clostridium thermocellum-like (C. t), Syntrophomonas wolfei-like (S. w), Ethanoligenens harbinense-like (E. h), Megasphaera elsdenii-like (M. e), Clostridium saccharoperbutylacetonicum-like (C. s), and Syntrophobacter fumaroxidans-like (S. f) organisms in each library. The number of clones is given in parentheses above each bar.

EHCB.

ADH isoenzyme analysis revealed EHCB in the acidophilic sludge. Zymograms (Fig. 9) show three types of ADH (I, II, and III) that are monomeric enzymes present at some time during the operation of the CSTRs. On day 7, ADH-II appeared, and it existed thereafter except in reactor B on day 21. From day 21 to day 28, ADH-III appeared, and its position coincided with the ADH of E. harbinense YUAN-3 (Fig. 9, last lane), which is an EHCB. ADH-II (reactor B on day 21) and ADH-III (both reactors on day 35) disappeared in the zymogram due to insufficient loading of samples. Moreover, the results for liquid fermentation products indicate that ethanol levels in both reactors increased quickly, and ethanol became the dominant product from the 14th to the 28th day. The DGGE profiles of the 16S rRNA gene also show that bands associated with E. harbinense strengthened gradually after 14 days. These results are consistent with the ADH analysis. Although a semiquantitative conclusion based on end point PCR could be biased, hydA clone libraries showed that C. thermocellum was a dominant EHCB among the HPB (Fig. 8). Therefore, ADH-II appeared to be affiliated with C. thermocellum. These results suggest that populations associated with ADH-II and ADH-III were key for H₂ production.

FIG. 9.

Zymograms of ADHs from anaerobic sludge. (A) Reactor A; (B) reactor B. Lanes are labeled with the time of sampling. Roman numerals on the left represent zymotypes. The specificities of ADHs were defined by their activities toward the substrate ethanol.

DISCUSSION

HPB were enriched from anaerobic sludge at pH 4.5, and the microbial communities in two duplicate reactors showed ethanol-H₂ coproduction after about 14 days. This finding was consistent with previous investigations (37, 38), in which a pH of 4.0 to 4.5 led to ethanol-H₂ coproduction in an unbuffered system. The H₂ yields were about 0.13 and 0.15 g of OD in H₂·g⁻¹ of OD in glucose (1.6 and 1.8 mol of H₂·mol⁻¹ of glucose) in the two reactors, values typical of H₂ yields found in fermentative H₂ production from food waste, glucose, and cellulose as substrates at a pH of >5 or 5.5 (23, 43, 46). However, a simple comparison is unreasonable, due to differences in inocula, substrates, and operation conditions. Rigorous operations should be carried out to compare H₂ yields for different fermentation types in the future.

16S rRNA gene sequences from clone libraries indicated that the putative HPB were affiliated with nine phylotypes of Ethanoligenens and Megasphaera spp. Clone libraries of the hydA gene confirmed that the HPB were previously undescribed strains affiliated with E. harbinense and M. elsdenii, along with C. thermocellum, S. wolfei, S. fumaroxidans, and C. saccharoperbutylacetonicum (17 phylotypes). Only Ethanoligenens and Clostridium were identified as putative HPB in a previous analysis of an ethanol-H₂-coproducing community based on 16S rRNA (with 97% similarity as the threshold) (38). Thus, direct amplification of hydA yielded a more complete view of the bacterial populations associated with H₂ production than that provided by 16S rRNA gene analysis.

A striking result is that not all HPB identified by hydA sequences had been described by previous studies based on the 16S rRNA gene. Previous investigations (2, 15, 23, 43, 46) indicated that the dominant populations from hydrogen-producing sludge were Clostridium, Bacillus, and Staphylococcus groups of low-G+C-content gram-positive bacteria; putative HPB were affiliated with Clostridium, Thermoanaerobacterium, Desulfotomaculum, and Thermotogales spp. In contrast, the majority of the dominant populations based on hydA were affiliated with low-G+C-content gram-positive bacteria, Bacteroidetes, and Actinobacteria groups. Furthermore, all these HPB were uncultivated but affiliated with E. harbinense, C. thermocellum, C. saccharoperbutylacetonicum, S. wolfei, S. fumaroxidans, and M. elsdenii.

These results suggest that mesophilic EHCB are enriched by acidophilic conditions. Eleven novel phylotypes closely related to E. harbinense, C. thermocellum, and C. saccharoperbutylacetonicum were the putative EHCB, and the ADH isoenzyme analysis also pointed to an E. harbinense-like organism, which has a high rate of conversion of carbohydrates to H₂ and ethanol (53). E. harbinense was also found to be a dominant EHCB in a previous analysis on an ethanol-H₂-coproducing community based on the 16S rRNA gene (38). Previous work also has isolated a few EHCB, including mesophilic E. harbinense, thermophilic C. thermocellum, and Thermoanaerobium brockii (4, 26, 38, 53). Low pHs favor Ethanoligenens (38, 52, 53), but acidophilic and mesophilic clostridia that are EHCB have not been described previously.

In addition, M. elsdenii-like, S. fumaroxidans-like, and S. wolfei-like organisms, which are fatty acid-oxidizing bacteria, were among the HPB in the acidophilic sludge. They consume short-chain fatty acids, produce H₂, and play a role in buffering against a pH drop (12, 14, 18, 29, 30). These fatty acid-oxidizing HPB were not found in a previous analysis of an ethanol-H₂-coproducing community based on 16S rRNA genes (38). Thus, abundant EHCB (11 phylotypes) and fatty acid-oxidizing HPB (6 phylotypes) may play a significant, but previously unrecognized, role during H₂ production of the acidophilic sludge.

The sampling interval of 7 days affected the times when strains could “appear” or “disappear” in the DGGE or zymogram. For example, ADH-II (reactor B on day 21) and ADH-III (both reactors on day 35) disappeared from the zymogram, possibly due to insufficient loading of sample DNA. Because long-term monitoring was difficult by several techniques, the experimental results were not exhaustive. The results of analyses could be complicated due to any time lapse. Therefore, continuous monitoring could offset the lapses from some time points.

Two conserved regions, ADLTIMEE and EVMACPGGCI, were found in the large subunit of [Fe]-H₂ases by sequence alignment of 14 genera. The ADLTIMEE region is located in the first 50 to 100 amino acid residues of the active-site domain (hydrogen cluster) (35). EVMACPGGCI is located in the C-terminal end of the active-site domain. Both conserved regions belong to α-helix and β-sheet structures (35). The amino acid residues between the ADLTIMEE and EVMACPGGCI regions cover two-thirds of the active-site domain of the large subunit of the [Fe]-H₂ases. Therefore, hydA segments of 500 to 600 bp, which are amplified using the pair of degenerate primers designed by us, represent important genetic information on the hydrogen cluster. This approach can be useful for elucidating the relationships between hydA gene expression, the H₂ production rate, and the effects of ecological factors on microbial community structure. It will facilitate the identification of novel unknown HPB in the natural environment or in communities by mining of the hydA gene.