High-rate hydrogen production from galactose in an upflow anaerobic sludge blanket reactor (UASBr)

Abstract

High-rate hydrogen production from galactose and rapid granule formation were achieved in a mesophilic (37 °C) upflow anaerobic sludge blank reactor (UASBr). Hybrid immobilized cells were seeded to a CSTR fed with galactose (15 g L⁻¹) at a hydraulic retention time (HRT) of 12 h for 12 days, after which the mixed liquor from the CSTR was transferred to the UASBr. Rapid granule formation was observed after 5 days of initial UASBr operation at 6 h HRT and further enhanced at 3 h HRT. The peak hydrogen production rate (HPR) and hydrogen yield (HY) of 32.7 L L⁻¹ d⁻¹ and 1.95 mol mol⁻¹ galactose were attained at the HRT of 3 h with an organic loading rate (OLR) of 120 g L⁻¹ d⁻¹. Microbial community analysis characterized by pyrosequencing of 16S rRNA genes showed that the proportion of Clostridium increased during the HRT at 3 h, whereas Bacilli proportions dominated at 6 h HRT.

---

1. Introduction

Depletion of fossil fuel reservoirs and rapid industrialization has led to the need for the development of sustainable and clean energy sources. Hydrogen (H₂) is a highly efficient and clean energy carrier that can be used as an alternative to fossil fuels.¹ Dark hydrogen fermentation is a promising biological H₂ production method with several advantages, including a relatively high hydrogen production rate (HPR), wide range of organic feedstock utilization and operation under mild temperature without light.^2,3 Galactose has been employed as a promising substrate for fermentative hydrogen production since it could be readily available from red algae hydrolysate. Moreover, it is a structural analog of glucose that is readily utilized by hydrogen producers.¹

Dark hydrogen fermentation by various reactor configurations has recently been reviewed, and it has been proposed that this process could be conducted in a UASBr because of its easy operation and granule formation.⁴ Anaerobic granules are aggregates of bacterial population along with some EPS (extra polymeric substances), which doesn't deform during reduced hydrodynamic shear and settle efficiently. Granule formation is essential towards high production performance since it provides a stable microbial population. In general, the development of hydrogen producing granules (HPG) in UASBr is complex process and usually takes months to form matured and stable granules, due to the slow/self-agitation of reactor during the low retention time.⁵ Besides, the contact between the microbial biomass and the substrate is improperly maintained, thus avoid the formation of granules efficient and faster. However, matured granules act as a good barrier to withstand biomass washout, organic shock and other process disturbances.⁶ Hence, the formation of HPG is considered a key factor in enhancing the efficiency of the UASBr, which may significantly shorten the start-up period of the reactor.⁷ Additionally, the syntrophic relationship between the microbial species in the granules would provide the shear stability against the harsh conditions and also act as a feed for other microbial community due to the formed extra cellular enzymatic activity.⁸ Nevertheless, the syntrophic population maintains the mutual cooperation between the species, which can perform the interspecies metabolic transfer and sustaining heterogeneous community.^9,10

In recent years, the start-up time for HPG formation has been reduced by various approaches such as adding the mixed liquor from a continuous stirred tank (CSTR) effluent,^4,7,11 acid-shock pretreatment to reduce the zeta potential and thereby induce granules formation,¹² addition of calcium ions,¹³ and high-recirculation of the reactor during the initial stage.⁶ In spite of the existence of high shear forces in the CSTR, the self-flocculation process enabled by the efficient mixing of the system and better contact between the microorganisms and substrate.¹¹

Characterization of hydrogen producing granules was investigated earlier to determine the microbial insights and morphological changes that contribute to the higher production performance.¹⁴ Most studies have used 16S rDNA based analysis (PCR-DGGE sequencing) to reveal the dominant microflora in the granules. For example, Hung et al. (2011) reported the interaction of Clostridium with other facultative anaerobes (such as Klebsiella and Streptococcus) in a self-formed granule during the hydrogen fermentation of sucrose.^9,15 Next generation DNA sequencing analyses including 454-pyrosequencing have been utilized for analysis of bacterial diversity of mixed cultures because they are more advanced and provide more insight and data than conventional techniques such as fluorescence in situ hybridization (FISH) and PCR-DGGE.^10,16,17 However, to the best of our knowledge, no studies have investigated changes in microbial dynamics during the formation of hydrogen producing granules (HPG) during continuous hydrogen fermentation using 454 pyrosequencing analysis. Thus, the present study was conducted to monitor changes in the microbial diversity of the high-rate HPG at 2 different HRTs (6 and 3 h) in a UASBr fed with galactose, which is an attractive carbohydrate source and analog to C6 sugar glucose provided as feedstock. In addition, the microbial diversity of the granules was characterized by scanning electron microscopy (SEM) and 454-pyrosequencing to reveal the major changes and factors contributing to the enhanced performance.

2. Materials and methods

2.1. Seed inoculum preparation

Seed sludge was taken from an anaerobic digester at a local wastewater treatment plant. The characteristics of the sludge were total chemical oxygen demand (T-COD): 22.6 g COD per L, total solids (TS): 35.9 g L⁻¹; volatile suspended solids (VSS): 23.6 g L⁻¹, pH: 6.8. The sludge was pretreated at 90 °C for 30 min to suppress the hydrogen consuming bacteria, then dried in an oven at 100 °C overnight. The dried seed sludge was finely ground and used for preparation of hybrid immobilized cells as previously described.¹ Hybrid immobilized cells were prepared by adding 20% w/v of the dried seed sludge into a solution containing sodium-alginate (2%), silicon dioxide (SiO₂-1%) and chitosan (1%) for entrapment and encapsulation. The mixed suspension was subsequently extruded into sterile calcium chloride (2%) solution to entrap cells in alginate beads. The formed beads, which had an average size of 5–6 mm, were further mixed with a fresh solution of 2% calcium chloride for 2 h, then washed three times with sterile distilled water and dried, after which they were stored in a refrigerator at 4 °C until use. Galactose was used as a sole carbon source in this study. Modified Endo medium with the following feed ingredients was used: 15 g L⁻¹ galactose; 3 g L⁻¹ NH₄HCO₃; 0.125 g L⁻¹ KH₂PO₄; 0.100 g L⁻¹ MgCl₂·6H₂O; 0.015 g L⁻¹ MnSO₄·6H₂O; 0.025 g L⁻¹ FeSO₄·7H₂O; 0.005 g L⁻¹ CuSO₄·5H₂O; 0.001 g L⁻¹ CoCl₂·5H₂O; 6.72 g L⁻¹ NaHCO₃.

2.2. Experimental setup

A continuous stirred tank reactor (CSTR) with a working volume of 3.0 L was operated at a temperature of 35 ± 1 °C. The reactor was seeded with 500 g of hybrid immobilized cells with a concentration of 20 g L⁻¹ (VSS). The hydraulic retention time (HRT) and pH of the reactor were maintained at 12 h and 5.5, respectively. The pH was maintained at above 5.5 by the addition of 3 N NaOH via a peristaltic pump. A mechanical stirrer installed on top of the reactor agitated the mixture at 150 rpm.

After 12 days of CSTR operation, 500 mL of the mixed liquor was transferred to an upflow anaerobic sludge blanket reactor (UASBr) with a working volume of 2.1 L (780 mm liquid depth and 58 mm inner diameter). A high-rate recirculation about 5.6 (32.8 mL min⁻¹)-fold higher than the initial flow rate of 5.8 mL min⁻¹ at 6 h HRT was adopted as the operational strategy for the first 5 days to induce rapid granule formation,⁶ after which recirculation was terminated and the flow rate changed from 6 h to 3 h HRT upon attaining pseudo-steady state performance with stable H₂ production (±15%) for at least 5 consecutive days. The UASBr was operated for 43 days, with 6 h HRT maintained for 18 days after post-start up, followed by 3 h HRT for 23 days. The pH of the reactor was maintained at 5.5 to 6.0 by adding a sufficient amount of 3 g L⁻¹ NH₄HCO₃ and 6.72 g L⁻¹ NaHCO₃ buffering agents into the feed medium.¹⁸

2.3. Analytical methods

The hydrogen, carbon dioxide, and nitrogen content of the biogas samples were analyzed by gas chromatography (SRI Instruments Model 310C, USA) using a thermal conductivity detector (TCD) and a 1.8 m × 3.2 mm stainless-steel column packed with silica gel and high purity nitrogen carrier gas. The temperatures of the injector, column, and detector were maintained at 80 °C for H₂ analysis and 90 °C for CO₂ and N₂ analysis. Biogas volume was measured using a wet gas meter (JH-LMF-1, Shanghai Jinghao International Trade Co., Ltd.). The organic acids contents were quantified by high performance liquid chromatography (Waters 717, USA) with Aminex-87H (BioRad Laboratories, USA) and an ultraviolet detector (Waters 2487, USA) at 210 nm. The pH, chemical oxygen demand (COD), and solids (total solids, TS; volatile suspended solids, VSS) were measured in accordance with the standard methods.¹⁹ Detailed procedures were explained in our previous study.²⁰ The COD mass balance was computed as described elsewhere.²¹

Extracellular polymeric substances (EPS) were extracted using the NaCl method as previously described.²² The extracted EPS was then analyzed for protein (PN) using Lowry's method with bovine serum albumin as an internal standard. Polysaccharides (PS) were measured using phenol-sulfuric acid with glucose as an internal standard.

2.4. Microbial community analysis

Microbial changes under different HRT operational conditions were assessed by 454 pyrosequencing. Total genomic DNA collected at HRTs of 6 h (day 20) and 3 h (day 30) were extracted using the Mobio PowerSoil DNA extraction kit (Solana Beach, CA, USA) according to the manufacturer's instructions. Bacterial 16S rRNA gene fragments were PCR-amplified using the 27F (5′-AGAGTTTGATCMTGGCTCAG-3′) and 518R (5′-ATTTACCGCGGCTGCTGG-3′) primer set. 454 pyrosequencing of the PCR amplicons was conducted by Macrogen Inc. (Seoul, Korea) according to the manufacturer's instructions (454 Life Science, Branford, USA). Sequencing results were analyzed as described in our earlier study.¹⁶

2.5. Scanning electron microscopy (SEM)

Morphological observation (images) of the granules was conducted using a Hitachi S-4300 scanning electron microscope (SEM) (Hitachi, Japan) with a voltage of 150 kV and 10 mm working distance. Prior to analysis, samples were coated with platinum at a thickness of 20 nm using a Hitachi E-1030 ion sputterer (Hitachi, Japan).

3. Results and discussion

3.1. Hydrogen fermentation in CSTR

The hydrogen production performance of the immobilized cells was evaluated during continuous operation to promote seed formation for successful UASBr operation from galactose. Previous investigations indicated that hydrogen producers can be rapidly self-flocculated and form flocs during CSTR operation, despite the existence of high shear forces by the applied mixing conditions.⁴ Therefore, this research was initially designed to promote self-flocculated hydrogen producers in a CSTR, thereby transferring the mixed liquor procured from the CSTR for the successful start-up of the UASBr operation. The hydraulic retention time (HRT) of the CSTR operation was fixed at 12 h.²⁰

The hydrogen production performance of the CSTR is shown in Fig. 1. The reactor was operated in batch mode for 48 h to induce the initial growth of the hydrogen producers, then switched to continuous mode at an HRT of 12 h. The operational pH was maintained at above 5.5. The hydrogen production progressed rapidly after 5 days of operation, reaching a hydrogen production rate (HPR) of 4.1 L L⁻¹ d⁻¹ and a hydrogen yield (HY) of 0.91 mol H₂ per mol galactose_added. Steady performance of the reactor (where the HPR showed less than 10% variation) was observed from day 6 to 12 with the average HPR and HY of 4.70 ± 0.19 L L⁻¹ d⁻¹ and 1.09 ± 0.04 mol H₂ per mol galactose_added, respectively. Addition of the immobilized cells would result in rapid start-up because of the aggregation of biomass to the immobilized cells.^16,21,23 The reactor operation is summarized in Table 1.

Fig. 1 Hydrogen production performances of CSTR operation and soluble metabolic products formation during CSTR operation.

Table 1 Summary of reactor performance

Reactor type	CSTR	UASBr
HRT (h)	12	6	3
Pseudo-steady state period	Day 6 to day 12	Day 10 to day 20	Day 35 to day 43
HPR (L L^−1 d^−1)	4.70 ± 0.19	15.36 ± 1.15	32.72 ± 2.08
HY (mol per mol galactoseadded)	1.09 ± 0.04	1.81 ± 0.13	1.95 ± 0.12
Sugar removal (%)	99.3 ± 0.20	99.8 ± 0.18	99.6 ± 0.41

---

Fig. 1 shows the daily variations in the production of soluble metabolic products (SMPs) during CSTR operation. n-Butyric (n-HBu) and acetic (HAc) acids were the dominant metabolites, with concentrations of 4294 ± 121 mg L⁻¹ and 2240 ± 102 mg L⁻¹, followed by lactic acid (HLa) and i-butyric (i-HBu) acid, with values of 576 ± 54 mg L⁻¹ and 564 ± 70 mg L⁻¹, respectively. Formic acid (HFo), propionic acid (HPr) and ethanol (EtOH) were minor SMPs, with values ranging from 151 to 386 mg L⁻¹, respectively. The total SMPs production was 8211 ± 150 mg L⁻¹ during steady-state operation. Our recent studies showed that the dominant butyrate mediated hydrogen fermentation pathway favored efficient hydrogen production from galactose.^16,20 The average biomass concentration was 564 ± 0.43 g VSS per L. Notably, the galactose utilization rate was >99% in this study, indicating that galactose is the preferred substrate for hydrogen producers in a CSTR. Indeed, this value was 33% higher than that observed for suspended cells operation.²⁰ Moreover, the HY 1.09 ± 0.04 mol H₂ per mol galactose_added followed a similar pattern, with an increase to 13% occurring relative to suspended cells operation at the same HRT of 12 h when using galactose feedstock. The observed variations are attributed to the preference of immobilized cells as an inoculum source in a CSTR. The COD mass balance of the system showed a closure value of 96%, validating the accuracy of the analytical data and the measured hydrogen production in a CSTR.

3.2. Granules formation in the UASBr

The UASBr was inoculated with 500 mL of the 12 day old mixed liquor from the CSTR and operated at 6 h HRT. The high-rate recirculation of 32.8 mL min⁻¹ during the first five days facilitated microbial aggregation and initiated granule formation.^5,7 Recirculation was then stopped to allow mature granule formation. On day 20, the HRT was reduced to 3 h. The HPG developed rapidly at an HRT of 3 h up to almost one third of the working volume (Fig. 2), which implies that transferring the mixed liquor from the CSTR and initial high-rate recirculation followed by HRT reduction is an efficient strategy for the rapid formation of HPG using galactose as a carbon source.

Fig. 2 Photographs showing the granule formation at various days in UASBr.

Fig. 2 show the morphology of granules over 40 days of UASBr operation. The granules were appeared in white color, which reflected the lower sulfide content as mentioned in previous studies.^9,24 During the initial stage (day 2), the formation of granules was very low and they were seen as suspended biomass. However, the SEM image showed filamentous bacterial populations. The filamentous structure was attributed to the direct addition of the 12 day old CSTR mixed liquor because the addition of flocculated bacterial cells procured from CSTR enables rapid granulation and stable hydrogen production behavior, which is attributed to the presence of the self-aggregated hydrogen producing flocs.^4,7 From day 5, rod shaped cells increased, indicating the accumulation of hydrogen producers. The structures were quite similar to the granule structures observed in previous studies.^9,25 In the later stages, the bacterial population attached together and formed a layer like structure because of their adhesion properties and communication between cells. This self-aggregation likely improved the stability of biomass inside the reactor system and the production performances over time. Syntrophic relationship between biofilm forming and hydrogen producing bacterial populations attribute to the rapid formation of granules. More details of the microbiota are discussed in further section.

EPS are secreted by diversified microbiomes present in anaerobic reactors.²⁶ The major constituents of EPS, carbohydrates and proteins, act as building blocks for microbial granules formation and maturation.^27,28 The secreted EPS would facilitate the attachment of the self-flocculated biomass and initiate the cell to cell communications between the microbial cells for the development/maturation of the granules.²⁹ Once the bio-granulation formed the EPS will be helpful for maintaining the structure of the granules under short HRT during the continuous operation. However, when exceeding the threshold dilution range, the changes in EPS content leads to the loss of the active biomass and detachment of microbial cells which affected the shape, size and other surface properties of the hydrogen producing granules.³⁰ Moreover, the EPS are highly sensitive to the metal toxicity, in which it can block the access of the micro nutrients to the bacterial cells, which in turn negatively affected the hydrogen production performances.³¹ As explained earlier by Lee et al.³² the increased secretion of EPS at long SRT lowering of the hydrogen production performances, whereas short SRT operation leads to the enhanced hydrogen production with low EPS productivity.

Earlier reports showed that the secretion of EPS by hydrogen producers (acidogen) was higher than that of acetogens and methanogens.³³ Fang et al.⁹ reported that the EPS content of HPG primarily consisted of carbohydrates and proteins, with negligible levels of lipids, uronic, and nucleic acids. Moreover, the PN/PS ratio could act as a key parameter to elucidate sludge surface characteristics such as surface charges, hydrophobicity and viscous nature as explained by Lu et al.²⁶ In this study, high EPS contents were clearly observed in the early stages of granule formation. On day 2, the total protein (PN) and carbohydrate contents (PS) were found to be 96 mg per g VSS and 217 mg per g VSS, respectively, with a PN/PS ratio of 0.47 on day 2. PN and PS decreased with further operation as shown in (Fig. 3). The PN/PS ratio increased slightly to 0.50 on day 5, then remained in the range of 0.32 to 0.40 from day 10 to 14 during operation at 6 h HRT. During further operation at 3 h HRT from day 20 to 40, the PN/PS ratio decreased to 0.24, demonstrating that variations in the distribution of excreted products are attributed to changes in the microbial community dynamics, as discussed below.

Fig. 3 EPS content from the UASBr on various days.

Moreover, as shown in the SEM images presented in Fig. 2, the granules appeared slimy during later stages (from 30–40 days of operation), and appeared to have high levels of disintegration and adherent layers. These results are consistent with those of a previous study by Zhang et al.,²⁷ who found that changes in the PN/PS ratio from 0.5 to 0.2 led to successful granule formation with efficient hydrogen production. Moreover, the excessive secretion of EPS may have a negative impact on sludge settling and affect mass transfer between the substrate and microorganisms (S/I ratio).^28,33

3.3. Performance of the UASBr

Fig. 4 shows the daily variation in biogas production rate (BPR) and HPR of the UASBr. The biogas consisted of a mixture of H₂ and CO₂, with no detectable amount of methane, revealing the effectiveness of the acclimatized enriched hydrogen producing cultures. The H₂ content increased after the first 5 days of operation from 40% to over 50% with changes in organic loading rate (OLR) by shortening the HRT. This could have occurred in response to fluctuations in hydrogen content during the initial start-up period due to adaptation of the microbes to changes in the reactor environment.^7,21 The stable BPR and H₂ content after 10 days of reactor operation indicated the effectiveness of the start-up strategy of mixed culture addition, followed by high-rate recirculation. After terminating the recirculation, the H₂ content ranged from 51 to 60% of the biogas. Table 1 summarizes the overall performance of the bioreactor obtained under steady state conditions at HRTs of 6 and 3 h.

Fig. 4 Production performances (BPR, HPR and HY) at various HRTs in UASBr.

As shown in Fig. 4 and Table 1, the HPR increased from 15.3 to 32.7 L L⁻¹ d⁻¹ when the HRT was reduced from 6 to 3 h. Hydrogen yield (HY) also showed a similar pattern as HPR, with values increasing from 1.82 to 1.95 mol per mol galactose_added, respectively. The peak HPR of 32.7 L L⁻¹ d⁻¹ and HY of 1.95 mol per mol galactose_added obtained at 3 h HRT demonstrated that low HRT operation is suitable for efficient and stable hydrogen production performance from galactose. As reported by ref. 34, 3 h HRT would provide better biomass retention and wash-out of other non-beneficial hydrogen producers, which aids in elevating hydrogen production performances.

Table 3 summarizes the start-up period of HPG formation and hydrogen production performances obtained from the organic feedstock employed during fermentative hydrogen production. These findings demonstrate that the HPG formed in the present study were efficient at handling the high OLR and low HRT with stable hydrogen production. Furthermore, granules formation occurred within 5 days, which was shorter than in previous studies.^4–7,11 The formed HPG are matured further during the low HRT is attributed by the changes in the upflow regimes and OLR, which increased the velocity, and aided in the substrate diffusion to the microorganisms for better motion of the gas bubbles released from the HPG with effective mixing in the UASBr. As explained earlier³⁵ increased OLR and the substrate velocity as a function of HRT reduction accelerated the granules formation in a short span of time with improved biogas productivity. Jung et al.¹¹ also reported that rapid-granules formation occurred in a UASBr fed with coffee drink wastewater in a short span of 5.5 days following the transfer of 1.5 L of 7 day old post-operated CSTR mixed liquor. However, the present study achieved a higher HPR with a larger OLR (120 g L⁻¹ d⁻¹). The reason for the surpassing performance of this study might be attributed by the start-up strategy, hybrid immobilization, microbial composition of the seed sludge, and substrate.

Table 2 COD mass balance of the CSTR and UASBr

Reactor type	HRT (h)	CODsub,in^a (mg COD per L per h)	CODsub,res^b (mg COD per L per h)	CODSMP^c (mg COD per L per h)	CODH2^d (mg COD per L per h)	CODbio^e (mg COD per L per h)	CODsum^f (mg COD per L per h)	COD balance^g (%)
a CODsub,in: mg COD per L per h of influent substrate (galactose 15 g L^−1 × 1.07 (g per g COD) = 16 g per L COD), calculated by (substrate concentration (mg COD per L) × feeding rate (L h^−1)).b CODsub,res: mg COD per L per h of residual substrate in the effluent, calculated by (CODsub,in × (1 − substrate utilization)).c CODSMP: mg COD per L per h of soluble metabolic products (SMP), calculated by (SMP concentration (mg COD per L) × feeding rate (L h^−1)).d CODH2: g COD per h of H2 evolved, calculated by (mol H2 per h × 16 g COD per g per H2).e CODBio: g COD per h of biomass in the effluent, calculated by (mg cell per L × feeding rate (L h^−1) × 1.42 mg COD per mg VSS per L), assuming that cell formula is C5H7O2N.f CODsum: g COD per h, sum of residual substrate + SMP + biomass + H2.g COD balance (%): [CODsum]/[CODsub,in] × 100.
CSTR	12	1333	5	1094	121	47	1289	96.7
UASBr	6	2666	5	2057	405	127	2594	97.2
	3	5333	21	4081	863	190	5152	96.6

---

Table 3 Comparison of H₂-producing UASBr with granule formation^a

Inoculum	Substrate	Initial HRT (h)	Start-up period (days)	HPR (L L^−1 d^−1)	HY (mol H2 per mol monosugaradded)	Reference
a CDWW: coffee drink wastewater.
Heat-treated anaerobic digester sludge	Glucose	5	20	7.68	1.54	6
Sewage sludge	Sucrose	24	52	6	0.75	5
Heat-treated anaerobic digester sludge	CDWW	8	10	2.59	1.29	11
Heat-treated anaerobic digester sludge	CDWW	6	5.5	2.76	1.78	4
Heat-treated anaerobic digester sludge	Xylose	6	30	9.9	2.98	7
Heat-treated anaerobic digester sludge	Galactose	6	5	32.7	1.95	This study

---

3.4. Soluble metabolic products, biomass production and COD balance

Generation of metabolic end products could be a key indicator of the success of H₂ production if either followed an acetogenic or solventogenic pathway.^17,20 The daily production of organic acid and ethanol in the UASBr at 6 and 3 h HRT are provided in Fig. 5. The major organic acids detected were butyric (HBu, i-HBu) and acetic (Hac) acid, while lower levels of propionic (HPr), formic (HFo) and lactic (HLa) acid were found. The total organic acids varied remarkably from 6500 to 7000 mg L⁻¹ as the HRT changed from 6 to 3 h. Butyric acid was the dominant metabolite, being present at levels of 4500 to 5300 mg L⁻¹, regardless of the applied HRTs. As demonstrated earlier, the butyric acid pathway thermodynamically favored the high butyric acid content, which generally coincided with the high H₂ production.^36,37

Fig. 5 Soluble metabolic acids distribution under various HRT operations in UASBr.

Acetic acid was the next most dominant pathway in the investigated reactor. The acetic acid levels ranged from 2200 to 3000 mg L⁻¹, and remained relatively stable throughout the operation. Propionic acid is considered an undesired metabolite.^17,38 In this study, it was detected at levels of 100–450 mg L⁻¹ (at 6 h HRT) and 70–160 mg L⁻¹ (at 3 h HRT). Low propionic acid production at lower HRT would result from the washout of hydrogen consumers and enrichment of hydrogen producers. Lactic acid was detected at very low levels of 39–236 mg L⁻¹ at the 6 h HRT and was almost undetectable at the end of the experimental period because of the selective dominance of hydrogen producing microbes. Formic acid was detected at 25–270 mg L⁻¹ in the beginning of the operation, and then decreased to 50–105 mg L⁻¹ when the HRT decreased to 3 h. The major solvent detected was ethanol at levels of 400 to 1280 mg L⁻¹. Ethanol production increased when HRT was changed from 6 to 3 h in the first 2 days, but then returned to 44–800 mg L⁻¹.

Biomass concentration in terms of volatile suspended solids (VSS) with respect to the change in HRT is shown in Fig. 5. The biomass concentration ranged from 400–670 mg L⁻¹ at HRTs of 6 h and 3 h, respectively, which is similar to the value reported by Jung et al. (2013). Under the 6 h HRT, the UASBr reached pseudo-steady state from day 10 to day 20, while at a the 3 h HRT steady state was attained from day 35 to day 43. Under these conditions, the effluent galactose concentration was under 200 mg L⁻¹.

The COD mass balance was computed based on the distribution of soluble metabolites, microbial biomass, and hydrogen (Table 2). The COD mass balance of the system showed a good balance between substrate utilization, microbial growth and stable hydrogen production, and the value the closure over 97% indicating the analytical methods and data were pretty accurate. The limited variation of less than 3% during COD recovery could be due to the marginal error of the determination methods used.³⁹

3.5. Microbial community analysis of UASBr

The microbial population was significantly influenced by changing the HRT from 6 to 3 h. The relative abundance of the species richness and diversity indices is shown in Table 4. The operational taxonomic units (OTUs) of both HRTs were similar, with values of 20 and 21, indicating the species richness of bacteria in the HPGs. However, the Chao index (21.5) at 6 h HRT showed a slightly higher value than 3 h HRT (20). The Shannon and Simpson values also showed that the species richness was influenced by changes in the flow rate of the reactor. The microbial community in the HPGs obtained at 3 h HRT had a higher Simpson's index value (3.04) and lower Shannon's value (0.825), demonstrating that 6 h HRT resulted in higher microbial diversity. The shorter HRT (3 h) would result in a population shift in the mixed culture because of the wash-out of bacterial cells at the higher dilution rate.

Table 4 Community richness and diversity of HPGs obtained at HRT of 6 and 3 h

HRT (h)	OTU	Chao	Shannon	Simpson	Goods coverage
6	21	21.5	2.96	0.830	0.99
3	20	20	3.05	0.825	0.99

---

Fig. 6 shows the distributions of microbial diversity based on phylum (a) and class (b) level abundance. Microbial diversity differed significantly between 6 and 3 h HRT. At the phylum level (Fig. 6a), the content of Firmicutes in the HPG at 6 h HRT was 60.24%, while this dramatically increased to 87.07% at 3 h HRT. However, the levels of Actinobacteria and Proteobacteria decreased significantly from 21.59% and 10.95% to zero and 2.69%, respectively, whereas the levels of Bacteroidetes increased from 7.00% to 9.16% as the HRT shifted from 6 to 3 h. At a HRT of 3 h, Firmicutes was the major phyla (Fig. 6a) of the bacterial populations, with a relative abundance of over 87%. This abundance was much higher than that of Bacteroidetes (9.1%), Proteobacteria (2.69%) and unclassified bacteria (1.8%), indicating the selective enrichment of the Firmicutes population and efficient hydrogen production performance. This is likely because bacteria within Firmicutes, such as Bacilli and Clostridium, are considered key species for the biohydrogen production.⁴⁰

Fig. 6 Relative abundance of bacterial populations (a) phylum level and (b) class level.

Interestingly, at the class level (Fig. 6b), the dominant bacterial groups were Bacilli, Clostridia and Micrococcales, whereas those with the lowest abundance were Gammaproteobacteria and Bacteroidia. Specifically, Bacilli and Clostridia were dominant at 3 h HRT with a peak abundance of 46% and 41%, while they were present at 27.7% and 32.54% at 6 h HRT. At 3 h HRT, the bacterial classes Micrococcales and Gammaproteobacteria showed the lowest abundance of 0.0% and 2.69%, which clearly indicates that the microbial diversity composition of the HPGs is significantly affected by changes in flow rate during continuous operation. A similar phenomenon was reported by Park et al.,¹⁶ that the relative abundance of Bacilli and Clostridia under a short HRT of 12 to 8 h showed a high hydrogen yield (2.1 mol H₂ per mol galactose) in a CSTR type bioreactor. Unlike earlier studies of UASBr operation as summarized in Table 2, the HPGs formed in the present study mainly consisted of effective proportions of Bacilli and Clostridia, resulting in a stronger ability to retain active biomass in the system at the lower HRT of 3 h.

Fig. 7 illustrates the distribution of microbial communities in the HPGs based on the family and genus level. As shown in Fig. 7a, Sporolactobacillaceae, Clostridiaceae, Arthrobacter, Enterobacteriaceae, and Prevotellaceae were the predominant populations, while there were negligible levels of Lactobacillaceae, Moraxellaceae, Veillonellaceae and unclassified bacteria. The remarkable improvement of the Clostridiaceae, Sporolactobacillaceae, and Prevotellaceae family under 3 h HRT showed an elevated improvement in HPR from 15.36 to 32.72 L L⁻¹ d⁻¹, demonstrating that these populations are responsible for the rapid-granules formation and high-rate hydrogen production.

Fig. 7 Relative abundance of bacterial populations (a) family level and (b) genus level.

It has been reported bacterial species within Clostridia such as Clostridium pasteurianum and other uncultured-Clostridium are biohydrogen producers.^41,42 These organisms were also frequently detected in the present study. A high proportion (23.8%) of uncultured-Clostridium was detected at 3 h HRT, which followed a similar butyrate-mediated hydrogen production pathway as other pure cultures of Clostridium sp. In general, Clostridial spp. exhibit butyrate-type hydrogen fermentation with the formation of acetate, lactate and ethanol, which is concordant with the distribution of soluble metabolic products observed in this study, with butyrate and acetate being found in the greatest abundance, followed by ethanol and lactate. The operational pH was maintained in the range of 5.5 to 6.0 inside the UASBr, which is in the optimal range for growth of hydrogen producers.^9,43

Two other important bacterial families observed in the HPGs were Sporolactobacillaceae and Prevotellaceae, which are responsible for granules formation. As indicated in earlier studies, both Prevotellaceae and Sporolactobacillus have the ability to form biofilm layers and undergo microbial assisted granular formation.^9,44 Since the former are hydrogen-sulfide producing bacteria and the latter are lactate producing bacteria, these populations could also compete for substrate utilization, which affects the hydrogen yield. However, these populations might maintain a syntrophic relationship with hydrogen-producing bacteria and aid in biomass retention under short HRTs because of their granule forming ability.^45,46 Understanding the syntrophic association between granular forming microorganisms and hydrogen producing microorganisms would enable improvement of reactor stability and performance.

The microbial community analysis implied that the monitoring the microbial population changes during the continuous operation is prerequisite for effective assessment of the hydrogen producing performances. The Changes in HRT influences the hydrogen production performances and microbial community structure. The increased Clostridium proportion at 3 h HRT attributed high hydrogen yield with the reduction of the significant proportions of unclassified Actinobacteria and Klebsiella populations. Nevertheless, the Sporolactobacillus and Prevotellaceae populations were also increased at the condition with stable hydrogen production performances in this study.

3.6. Significance of this study

Granules formation in a continuous hydrogen producing system is momentous for efficient and stable hydrogen production performances, due to their ability to retain active biomass under severe washout at low HRT. In general, UASBr was effectively used to treat the low to high strength wastewater for the generation of methane, due to their effective hold up of the active biomass and better biogas production rates, however the granules formation is a tedious process, usually takes days to months for the matured granules and bio-film formation with the co-existence of syntrophic microbial populations.⁴⁷ The hydrogen production performances and the duration of the granules formation in UASBr operation are summarized in Table 3, unlike the methane producers, the HPG formed in a short span of time 5 to 52 days (Table 3). The adoption of engineering approaches such as high-rate recirculation, addition of mixed liquor flocs procured by initial CSTR operation and the addition of chemical agents are suggested to reduce the start-up phase and promotes the rapid HPG formation of the UASBr operation.

According to the results obtained from this research, addition of 12 days old mixed liquor flocs obtained from the CSTR effluent induced the rapid HPG in a UASBr after 5 days of high-rate recirculation. Further operational parameters such as changes in OLR and HRT aided on the development of granular bio-film with the stable hydrogen production performances. HRT greatly affected the microbial community dynamics, biomass production, soluble metabolic products and hydrogen production performances in the continuous UASBr operation. High OLR with low HRT ensured the maximum hydrogen production. The peak HPR of 32.7 L L⁻¹ d⁻¹ was obtained at low HRT of 3 h with an OLR of 120 g L⁻¹ d⁻¹. Consequently, transferring the mixed liquor from CSTR, initial high-rate recirculation followed by HRT reduction is a useful strategy for adequate hydrogen production from galactose feedstock.

4. Conclusions

This study demonstrated the rapid-granule formation and high-rate hydrogen production from galactose in a UASBr. The start-up strategy of transferring the mixed liquor from a CSTR and initial high-rate recirculation, followed by HRT reduction is an efficient strategy for the rapid formation of H₂ producing granules from galactose. Granules with lower EPS content (138 mg per g VSS) formed at a shorter HRT of 3 h HRT, prompting hydrogen production. Microbial community analysis clearly indicated that both hydrogen-producing and granule forming bacteria are required for rapid granule formation and high-rate hydrogen production. Bacterial populations such as Prevotellaceae and Sporolactobacillus would nurture biofilm formation along with Clostridiaceae and attribute to the higher production observed in this study.

Acknowledgements

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (No. 2014R1A2A2A04005475).

References

1. G. Kumar, P. Sivagurunathan, T. Kobayashi, K. Q. Xu and S.-H. Kim, Algal Res., 2015, 11, 326–333.
2. G. Kumar, P. Bakonyi, S. Periyasamy, S. H. Kim, N. Nemestóthy and K. Bélafi-Bakó, Renewable Sustainable Energy Rev., 2015, 44, 728–737.
3. P. Sivagurunathan, G. Kumar, P. Bakonyi, S.-H. Kim, T. Kobayashi, K. Q. Xu, G. Lakner, G. Tóth, N. Nemestóthy and K. Bélafi-Bakó, Int. J. Hydrogen Energy, 2016, 41, 3820–3836.
4. K.-W. Jung, D.-H. Kim and H.-S. Shin, Int. J. Hydrogen Energy, 2011, 36, 1466–1473.
5. F.-Y. Chang and C.-Y. Lin, Int. J. Hydrogen Energy, 2004, 29, 33–39.
6. K.-W. Jung, S.-K. Cho, Y.-M. Yun, H.-S. Shin and D.-H. Kim, Int. J. Hydrogen Energy, 2013, 38, 9097–9103.
7. K.-W. Jung, D.-H. Kim and H.-S. Shin, Int. J. Hydrogen Energy, 2013, 38, 7253–7258.
8. R. A. Wirtz and R. R. Dague, Water Environ. Res., 1996, 68, 883–892.
9. H. H. P. Fang, H. Liu and T. Zhang, Biotechnol. Bioeng., 2002, 78, 44–52.
10. A. Chojnacka, M. K. Błaszczyk, P. Szczęsny, K. Nowak, M. Sumińska, K. Tomczyk-Żak, U. Zielenkiewicz and A. Sikora, Bioresour. Technol., 2011, 102, 10057–10064.
11. K.-W. Jung, D.-H. Kim and H.-S. Shin, Int. J. Hydrogen Energy, 2010, 35, 13370–13378.
12. D.-W. Liang, S. S. Shayegan, W. J. Ng and J. He, Biochem. Eng. J., 2010, 49, 119–125.
13. K.-S. Lee, Y.-S. Lo, Y.-C. Lo, P.-J. Lin and J.-S. Chang, Enzyme Microb. Technol., 2004, 35, 605–612.
14. H. H. Fang and H. Liu, Bioresour. Technol., 2002, 82, 87–93.
15. C.-H. Hung, C.-H. Cheng, D.-W. Guan, S.-T. Wang, S.-C. Hsu, C.-M. Liang and C.-Y. Lin, Int. J. Hydrogen Energy, 2011, 36, 8704–8711.
16. J.-H. Park, G. Kumar, J.-H. Park, H.-D. Park and S.-H. Kim, Bioresour. Technol., 2015, 188, 109–116.
17. I. Moreno-Andrade, J. Carrillo-Reyes, S. G. Santiago and M. C. Bujanos-Adame, Int. J. Hydrogen Energy, 2015, 40, 17246–17252.
18. C.-Y. Lin, S.-Y. Wu, P.-J. Lin, J.-S. Chang, C.-H. Hung, K.-S. Lee, F.-Y. Chang, C.-Y. Chu, C.-H. Cheng, C.-H. Lay and A. C. Chang, Int. J. Hydrogen Energy, 2010, 35, 13452–13457.
19. A. W. W. A. American Public Health Association, and Water Environment Federation: Standard methods for the examination of water and wastewater, 20th edn, APHA, AWWA, and WEF, Washington, D.C., USA, 1998.
20. G. Kumar, J.-H. Park, M.-S. Kim, D.-H. Kim and S.-H. Kim, Int. J. Hydrogen Energy, 2014, 39, 20625–20631.
21. P. Sivagurunathan, B. Sen and C.-Y. Lin, Appl. Energy, 2015, 147, 1–9.
22. G.-Y. Zhen, X.-Q. Lu, Y.-Y. Li and Y.-C. Zhao, Bioresour. Technol., 2013, 136, 654–663.
23. S. Y. Wu, C. H. Hung, C. Y. Lin, P. J. Lin, K. S. Lee, C. N. Lin, F. Y. Chang and J. S. Chang, Int. J. Hydrogen Energy, 2008, 33, 1542–1549.
24. C. E. Hernández-Mendoza and G. Buitrón, J. Chem. Technol. Biotechnol., 2014, 89, 143–149.
25. S. Y. Wu, C. H. Hung, C. N. Lin, H. W. Chen, A. S. Lee and J. S. Chang, Biotechnol. Bioeng., 2006, 93, 934–946.
26. X. Lu, G. Zhen, A. L. Estrada, M. Chen, J. Ni, T. Hojo, K. Kubota and Y.-Y. Li, Bioresour. Technol., 2015, 180, 264–273.
27. Z.-P. Zhang, K.-Y. Show, J.-H. Tay, D. T. Liang, D.-J. Lee and W.-J. Jiang, Biotechnol. Bioeng., 2007, 96, 1040–1050.
28. W.-W. Li and H.-Q. Yu, Bioresour. Technol., 2011, 102, 8653–8660.
29. Z.-P. Zhang, S. S. Adav, K.-Y. Show, J.-H. Tay, D. T. Liang, D.-J. Lee and A. Su, Biotechnol. Bioeng., 2008, 101, 926–936.
30. F.-Y. Chang and C.-Y. Lin, Water Sci. Technol., 2006, 54, 105–112.
31. C. Li and H. H. P. Fang, Chemosphere, 2007, 67, 668–673.
32. D.-Y. Lee, Y.-Y. Li, Y.-K. Oh, M.-S. Kim and T. Noike, Int. J. Hydrogen Energy, 2009, 34, 1244–1252.
33. X. S. Jia, H. Furumai and H. H. P. Fang, Environ. Technol., 1996, 17, 283–291.
34. J. I. Horiuchi, T. Shimizu, K. Tada, T. Kanno and M. Kobayashi, Bioresour. Technol., 2002, 82, 209–213.
35. A. Francese, P. Córdoba, J. Durán and F. Siñeriz, World J. Microbiol. Biotechnol., 1998, 14, 337–341.
36. P. Sivagurunathan, G. Kumar, J.-H. Park, J.-H. Park, H.-D. Park, J.-J. Yoon and S.-H. Kim, Int. J. Hydrogen Energy, 2016, 41, 4393–4403.
37. H.-S. Lee, M. B. Salerno and B. E. Rittmann, Environ. Sci. Technol., 2008, 42, 2401–2407.
38. P. Sivagurunathan, B. Sen and C.-Y. Lin, Int. J. Hydrogen Energy, 2014, 39, 19232–19241.
39. B. E. Rittmann and P. L. McCarty, Environmental Biotechnology: Principles and Applications, Tata McGraw Hill Education Private Limited, 2012.
40. J.-H. Park, S.-H. Lee, H.-J. Ju, S.-H. Kim, J.-J. Yoon and H.-D. Park, Renewable Energy, 2016, 86, 889–894.
41. Y. Ueno, T. Kawai, S. Sato, S. Otsuka and M. Morimoto, J. Biosci. Bioeng., 1995, 79, 395–397.
42. J. Wu and C. Lin, Water Sci. Technol., 2004, 49, 223–228.
43. J.-H. Park, S.-H. Lee, J.-J. Yoon, S.-H. Kim and H.-D. Park, Bioresour. Technol., 2014, 157, 98–106.
44. J. Washio, T. Sato, T. Koseki and N. Takahashi, J. Med. Microbiol., 2005, 54, 889–895.
45. S. Anna, Z. Urszula, B. Mieczysław and J. Marcin, in Lactic Acid Bacteria – R & D for Food, Health and Livestock Purposes, InTech, 2013, pp. 487–514.
46. M. Iosif, M. Carsten and S. Heidrun, Fermentative Hydrogen Production by Molasses, Effect of Hydraulic Retention Time, Organic Loading Rate and Microbial Dynamics, ed. Dragica Minic, ISBN: 978-953-51-0812-2, InTech, 2012,  DOI:10.5772/47750.
47. T. Abbasi and S. A. Abbasi, Renewable Sustainable Energy Rev., 2012, 16, 1696–1708.

---