EGSB reactors performance

Hydrogen production rates in arabinose- and glucose-fed reactors (R_arab and R_gluc) are shown in Figures 2 and 3, respectively. Only H₂ and CO₂ were detected in the gas phase; methane was not produced during all operation time. During start-up (period I), hydrogen production rates of approximately 0.3 L H₂ L^-1d^-1 were observed in both reactors. This corresponds to hydrogen yields of roughly 0.2 and 0.3 mol H₂ per mol of substrate consumed, for R_arab and R_gluc respectively (Table 1). In period II, the increase in arabinose and glucose inlet concentration to 16.6 mM and 13.8 mM, respectively, resulted in hydrogen yields of about 0.80 mol H₂ per mole of substrate in both R_arab and R_gluc (Table 1). Maximum hydrogen production rates in period II were of 1.36 ± 0.04 and 1.12 ± 0.07 LH₂ L^-1 d^-1 in R_arab and R_gluc, respectively. Substrate was completely consumed in both reactors and the main by-products formed were butyrate, acetate and lactate (Figures 2 and 3). In operation period III, substrate concentrations fed to R_arab and R_gluc were increased to 33.3 mM of arabinose and 27.7 mM of glucose, respectively. As a result of this increase, there was a temporary raise in arabinose/glucose concentration in the effluent but, after 13 days of acclimation to the higher substrate loads, virtually all glucose and an average of 79% arabinose were used in the reactors (Table 1) Steady state hydrogen production rates of 3.26 ± 0.16 and 2.06 ± 0.06 L H₂ L^-1 d^-1 were observed in R_arab and R_gluc, respectively (Figures 2 and 3). During period III, R_gluc showed a stable hydrogen yield of about 0.75 mol H₂ per mole of substrate consumed. Hydrogen yield in R_arab was significantly higher, that is. 1.10 mol H₂ per mole of substrate consumed. Lactate concentration in R_gluc increased sharply during period III of operation reaching values of approximately 20 mM (Figure 3). An increase in lactate concentration was also observed in R_arab, but did not exceed 11 mM (Figure 2). Estimation of the theoretical reduced form of nicotinamide adenine dinucleotide (NADH) production from glucose and arabinose, considering the main catabolic pathways (that is, Embden-Meyerhof for glucose and a combination of pentose phosphate and Embden-Meyerhof pathways for arabinose (Figure 1)), demonstrates that a higher reducing power was potentially formed in R_gluc than in R_arab. Estimated NADH concentration in R_gluc was 42 mM after three days of operation, while in R_arab was 37 mM after five days of operation.

Bacterial community composition dynamics in EGSB reactors

DGGE profiles generated for sludge samples withdrawn from R_arab and R_gluc (Figure 4) show that bacterial composition in both reactors' sludge at the end of periods II (Day 27) and III (Day 41) are identical. Differences in substrate composition did not affect the bacterial community in reactors R_arab and R_gluc and similarity index between Arab/Gluc samples at the end of the operation was as high as 94%. Predominant DGGE bands in R_gluc and R_arab were identical to the ones present in the inoculum used in this study and for which the phylogeny had been previously assessed [6]. Two of the predominant DGGE bands showed high similarity (> 99%) with the hydrogen-producing Thermoanaerobacterium thermosaccharolyticum. Members of the Klebsiella, Bacillus and Sporolactobacillus genera, detected in the inoculum sludge, were also predominant in R_gluc and R_arab.

Effect of hydrogen partial pressure and pH on batch hydrogen production from arabinose and glucose

The effect of the hydrogen partial pressure, while using arabinose and glucose as substrates, was studied in batch experiments at pH 5.5 (equivalent to pH 5.0 at 70°C). Assays were performed allowing the accumulation of hydrogen in the headspace (no headspace flushing, NHF), or preventing hydrogen accumulation in the headspace (headspace flushing, HF). Subsequently, HF assays were performed at pH 7.0 (that is, pH 6.5 at 70°C) to study the effect of pH increase in hydrogen production. Substrates were added at the beginning of the experiment and a second addition was performed after complete depletion of the first load.

In the NHF (pH 5.5), maximum hydrogen concentration in the gas was achieved 44 and 20 h after the second addition of arabinose or glucose addition, respectively (Figure 5a, b). At this point, hydrogen partial pressure in both arabinose and glucose assays was roughly 1.2 × 10⁴ Pa (at 70°C), which corresponds to a dissolved hydrogen concentration of 105 μM. From this point on, hydrogen production was not significant, even though only 35% of arabinose and 13% of glucose were present at the end of the experiment. Identical hydrogen yield, that is, 0.7 mol H₂ per mole of substrate, was obtained for NHF arabinose and glucose experiments (Table 2).

Hydrogen production from arabinose could be increased in assays in which hydrogen partial pressure in the headspace was kept low (HF). A cumulative hydrogen production of 1. To -1.7 × 10⁴ Pa (at 70°C) was attained in HF (pH 5.5) arabinose experiments (Figure 5c). This value is significantly higher than the one obtained in NHF experiments (P< 0.01: t-test), and corresponds to an increase of about 40% in hydrogen pressure. However, the highest increment in hydrogen cumulative production (that is, 62%) was observed in HF arabinose assays performed at pH 7.0 (cumulative hydrogen pressure of 2.8 × 10⁴ Pa at 70°C (Figure 5f)). Arabinose was totally consumed in HF assays at pH 7.0, while a fraction substrate (approximately 1%) was not used in HF assays at pH 5.5 (Figure 5c, e). Nevertheless, non-consumed arabinose in HF at pH 5.5 was considerably lower than in NHF assays (Figure 5a, c). Hydrogen yields in HF arabinose experiments at pH 5.5 and pH 7.0 were 0.76 and 1.15 mol H₂ per mole of substrate consumed, respectively (Table 2).

Hydrogen production values in HF and NHF glucose experiments at pH 5.0 were not significantly different. However, cumulative hydrogen production from glucose in HF experiments at pH 7.0 was significantly higher (P < 0.001: t-test) than at pH 5.5 (Figure 5d, f). Hydrogen cumulative pressure in HF glucose assays at pH 7 was of 2.6 × 10⁴ Pa (at 70°C, Figure 5f). Glucose was totally consumed in HF assays both at pH 5.5 and pH 7.0. Hydrogen yields in HF glucose experiments at pH 5.5 and pH 7.0 were 0.6 and 1.4 mol H₂ per mole of substrate consumed, respectively (Table 2).

At pH 5.5 approximately 20 mM of ethanol was produced from both substrates. At pH 7 ethanol formation did not exceed 15 mM (Figure 6). Acetate formation from both substrates at pH 7 achieved approximately 14 mM (Figure 6). In the case of glucose a decrease in 40% of lactate formation was also observed in incubations at pH 7.

Continuous hydrogen production in EGSB reactors

The fact that the microbial communities' composition in the reactors did not change along the three operational periods (Figure 4), suggests that the higher concentration of lactate produced in R_gluc during period III is related to metabolic changes and is not a consequence of bacterial community shifts. Two of the predominant DGGE bands present in the reactors sludge could be affiliated with Thermoanaaerobacterim thermosacharolyticum (similarity higher than 99%). A draft genome of T. thermosacharolyticum (Joint Genome Institute) allowed a search of genes that encode metabolic enzymes involved in pyruvate conversion. A L-lactate dehydrogenase (EC 1.1.1.27) was present indicating the possibility of pyruvate reduction to lactate. Some genes codifying subunits of enzymes related to pyruvate-ferredoxin oxidoreductases and NADH oxidoreductases were also found but a complete picture of the mechanisms involved in pyruvate conversion to acetyl-CoA cannot be retrieved. Clones corresponding to other predominant DGGE bands present in reactors sludge exhibited highest sequence identity with Klebsiella sp. (99%) and Bacillus coagullans (99%). All these microorganisms are able to produce hydrogen and lactate, among other products, from a variety of carbon sources [21–23]. No genomic information is available for these species and physiological information is sometimes contradictory. For instance, the presence of Bacillus coagullans in hydrogen producing reactors has been associated to the increase of lactate production [24, 25], but also to optimized hydrogen production [23, 26].

The main possible reactions for the fermentation of arabinose and glucose, and the calculated Gibbs free energy of global reactions are shown in Table 3 (equations (3) to (10)) (only the reactions yielding experimentally detected soluble fermentation products in R_gluc and R_arab are represented). From a thermodynamic point of view, lactate formation from glucose and arabinose is less favorable than the formation of butyrate or ethanol. However, in continuous processes lactate was one of the main soluble fermentation products present in both reactors, especially in R_gluc at higher influent, substrate concentration (27.7 mM). This might be related to the need of recycling reducing power from NADH. It has been proposed that thermophiles usually possess some escape routes to dispose of reductants in order to prevent obstructions in their metabolic flux. A possible route for this is the production of more reduced organic compounds like lactate, acetone and butanol [9, 23]. A switch to lactate formation in Thermoanaerobacterium sp. was observed as a mechanism of reductant disposal and NAD(P)H oxidation [9, 27].

Hydrogen partial pressure and pH influence on hydrogen production yields

The metabolic pathways of hydrogen formation are sensitive to hydrogen concentrations and are subject to end-product inhibition. Results from this study showed that hydrogen production from arabinose and glucose is indeed higher when hydrogen is not allowed to accumulate in the headspace. Keeping low hydrogen partial pressure caused an increase in hydrogen production that could be mainly related to enhanced sugar utilization under these conditions. Nevertheless, in arabinose assays substrate was never completely depleted, not even when hydrogen was removed from the headspace. This can indicate that limiting factors other than P H₂, such as liquid by-products inhibition, might be involved in hydrogen production from arabinose.

It has been reported that thermophilic hydrogen producing microorganisms could be inhibited by the presence of hydrogen, even when at very low partial pressure (from 0.1 × 10⁴ to 7.5 × 10⁴ Pa) [28]. Values of hydrogen partial pressure of 2 × 10³ Pa, 1.6 × 10³ Pa and 1.0 × 10⁴ Pa were described as inhibitory for hydrogen production with Thermotoga maritima, Pyrococcus furiosus and Caldicellulosiruptor saccharolyticus, respectively [29]. In the present study hydrogen production by extreme thermophile mixed cultures using glucose and arabinose was inhibited at a P H₂ similar to the one reported for C. saccharolyticus (that is, 1.2 × 10⁴ Pa at 70°C).

Higher cumulative hydrogen production and yields were obtained at pH 7.0, either using glucose or arabinose as substrate. Lower hydrogen production at pH 5.5 was coupled to high ethanol and low acetate production (Figure 6). The present study suggests that, at extreme thermophilic conditions, maintenance of neutral pH (around 6.5 at 70°C) can aid preventing hydrogen losses by avoiding the production of more reduced organic compounds (such as lactic acid, acetone, butanol, and so on).

Overall, the results presented in this study show that both pH and hydrogen partial pressure affect hydrogen production efficiencies by extreme thermophilic mixed cultures. However, pH influenced hydrogen production in a greater extent than hydrogen partial pressure, both when using glucose or arabinose as substrate. Different soluble fermentation products' composition was observed in batch experiments and in continuous reactors. This can be related with the accumulation of soluble fermentation products happening in the batch assays, which can lead to different environmental conditions and, therefore, induce different metabolic pathways [30–32]. Nevertheless, batch results can give valuable insights for improving hydrogen production in continuous process.

Continuous hydrogen production in EGSB reactors

Batch experiments

Medium composition and substrates

Assays were performed in 70 mL serum bottles containing 18 mL of buffered medium. Phosphate-buffered medium (20 mM) and bicarbonate-buffered medium (Stams et al. 1993) were used for experiments at pH 5.5 and pH 7, respectively. Bottles with phosphate-buffered medium were flushed with N₂ (100%), while bottles with bicarbonate-buffered medium were equilibrated with a mixture of N₂:CO₂ (80:20%). Both media were supplemented with trace elements, salts and vitamins according to the procedure described by Stams et al. [33]; yeast extract was added to a final concentration of 0.5 g L^-1. Medium was reduced with 0.8 mM sodium sulfide (Na₂S.9H₂O) and inoculated with 0.4 g of granular sludge. Arabinose (13 mM) and glucose (11 mM) were used as the main carbon source. Bottles were incubated in the dark at 70°C without shaking. After substrate depletion, a second pulse of 13 mM arabinose or 11 mM glucose was added and incubation extended.

Effect of hydrogen partial pressure

The effect of hydrogen partial pressure on hydrogen production from arabinose and glucose was investigated in batch mode at pH 5.5. Two series of batch experiments were performed: in series NHF (no headspace flushing), hydrogen was allowed to accumulate in the gas phase, while in series HF (headspace flushing) hydrogen was removed from the bottles' headspace and replaced by 100% N₂. All experiments were performed in triplicate and included controls without substrate. Sugars consumption, production of hydrogen gas and soluble fermentation products were monitored. Dissolved hydrogen concentration was calculated using the Henry's law at 70°C: K_H*Pi, where K_H is the Henry's law constant for hydrogen (8.7 × 10^-9 M/Pa at 70°C).

Effect of pH

The effect of pH on hydrogen production from arabinose and glucose fermentation was studied in two series of batch experiments, one at pH 7.0 and the other at pH 5.5. Incubation was done at 70°C and all the experiments were performed in triplicate. Sugars consumption, formation of hydrogen gas and soluble fermentation products were monitored and dissolved hydrogen concentration was calculated using the Henry's law at 70°C.

Analytical methods

Hydrogen concentration in the gas phase was determined by gas chromatography (GC) using a Hayesep Q column (80/100 mesh) and thermal conductivity detector Varian 3300 Gas Chromoatograph, (Varian, Walnut Creek, USA)) with nitrogen (30 mL minute^-1) as the carrier gas. The injector, detector and column temperatures were 120, 170, and 35°C respectively. Methane and carbon dioxide content of the gas phase from batch experiments and EGSB reactors was determined by gas chromatography using a Porapack Q (100 to 180 mesh) column, with helium as the carrier gas at 30 mL minute^-1, and a thermal conductivity detector. Temperatures of the detector, injector and oven were 110, 110 and 35°C, respectively. In the EGSB reactors gas flow rate was measured by a Ritter Milligascounter (Dr. Ing. Ritter Apparatebau GmbH, Bochum, Germany). Volatile fatty acids (VFA), ethanol, lactic acid, L-arabinose and glucose were determined by high performance liquid chromatography using an HPLC (Jasco, Tokyo, Japan) with a Chrompack column (6.5 × 30 mm²); sulfuric acid (0.01 N) at a flow rate of 0.7 mL minute^-1 was used as mobile phase. Column temperature was set at 60°C. Detection of VFA, lactic acid, ethanol, arabinose, glucose was made sequentially using a UV detector at 210 nm and a RI detector.

PCR-DGGE

Representative granular sludge samples were collected from R_arab and R_gluc and stored at -18°C. Total genomic DNA was extracted from approximately 500 μL of sample by using the FastDNA SPIN kit for soil (Qbiogene, Carlsbad, CA, USA). 16S rRNA gene fragments of approximately 450 bp were amplified for DGGE analysis by PCR using a Taq DNA polymerase kit (Life Technologies, Gaithersburg, MD, USA) using the primer set 954GC-f and 1369-r, as previously described by Nubel et al. [34]. The size of the obtained PCR products was checked by comparison with appropriate size and mass standard (MBI Fermentas, Vilnius, Lithuania), by electrophoresis on an 1% (w/v) agarose gel and ethidium bromide staining. Gels ran at a constant voltage of 100 V in an agarose gel electrophoresis system (Mupid-EX, Seraing, Belgium). Nucleic acids were detected using an UV transilluminator (BioRad, Hercules, CA, USA).

DGGE analysis of the amplicons was done by using the DCode system (Bio-Rad). PCR products were electrophoresed in a 0.5 × Trisacetate-EDTA buffer for 16 h at 85 V and 60°C on polyacrylamide gel (8%) containing a linear gradient ranging from 30% to 60% denaturant. Silver staining of DGGE gels was performed as previously described [35]. DGGE gels were scanned at 400 dpi and the DGGE profiles compared using the Bionumerics 5.0 software package (Applied Maths, Gent, Belgium). Similarity indices (Si) of the compared profiles were calculated from the densitometric curves of the scanned DGGE profiles by using the Pearson product-moment correlation [36].

Gibb's Free energy calculations

Standard Gibb's free energy at 25°C (ΔG^o) was calculated using standard Gibb's free energy of formation values (ΔG^o_f) obtained from the literature [19, 20] or calculated using the group addition method [19].