Biohydrogen Production from Hydrolysates of Selected Tropical Biomass Wastes with Clostridium Butyricum

Biohydrogen production has received widespread attention from researchers in industry and academic fields. Response surface methodology (RSM) was applied to evaluate the effects of several key variables in anaerobic fermentation of glucose with Clostridium butyrium, and achieved the highest production rate and yield of hydrogen. Highest H2 yield of 2.02 mol H2/mol-glucose was achieved from 24 h bottle fermentation of glucose at 35 °C, while the composition of medium was (g/L): 15.66 glucose, 6.04 yeast extract, 4 tryptone, 3 K2HPO4, 3 KH2PO4, 0.05 L-cysteine, 0.05 MgSO4·7H2O, 0.1 MnSO4·H2O and 0.3 FeSO4·7H2O, which was very different from that for cell growth. Sugarcane bagasse and Jatropha hulls were selected as typical tropical biomass wastes to produce sugars via a two-step acid hydrolysis for hydrogen production. Under the optimized fermentation conditions, H2 yield (mol H2/mol-total reducing sugar) was 2.15 for glucose, 2.06 for bagasse hydrolysate and 1.95 for Jatropha hull hydrolysate in a 3L fermenter for 24 h at 35 °C, with H2 purity of 49.7–64.34%. The results provide useful information and basic data for practical use of tropical plant wastes to produce hydrogen.


Results and Discussion
Glucose and hydrolysates were fermented with C. butyrium in both bottles (batch) and a fermenter. Single-factor bottle experiments were conducted to find best culture medium for hydrogen production from glucose with results given in supplementary materials ( Figure S1). Linear relation between dry cell weight (DCW) and optical density at 650 nm (OD 650 ) was plotted in supplementary materials ( Figure S2). Both H 2 yield and DCW versus     Tables 1-3, respectively. Study on the path of steepest ascent could be observed in Table 4. Central composite design (CCD), experimental results and analysis were summarized in Tables 5 and 6. Response surface plot and corresponding contour, canonical analysis for RSM were given in Fig. 1 and Table 7, respectively. Products versus time in bottle fermentation of hydrolysates of bagasse and Jatropha hulls were showed in Figs 2 and 3. Composition analysis of bagasse and Jatropha hulls was given in Table 8. Products produced from glucose and hydrolysates in both bottles and fermenter were presented in Table 9. Table 10 Compares hydrogen yield in this study with reported data. All the experiments were repeated more than two times and the standard deviation     30.47%) with the initial medium. Microbial growth was also studied under different variable concentrations in culture medium. Regression curve presents well-linear relationship between DCW and OD 650 ( Figure S2). It was found that the optimal concentrations of variables for microbial growth and H 2 yield were different ( Figure S3). Obviously, the optimal concentrations of glucose, L-cysteine, MgSO 4 ·7H 2 O and FeSO 4 ·7H 2 O were 30, 0.1, 0.15 and 0.1 g/L for DCW, respectively, while they were 15, 0.05, 0.05 and 0.3 g/L for H 2 production ( Figure S3). The most optimal culture medium for the strain growth was not the maximum for hydrogen production.
PB design. PB design was used to identify variables that had significant effects on H 2 production. The medium compositions were: glucose, yeast extract, tryptone, K 2 HPO 4 , KH 2 PO 4 , L-cysteine, MgSO 4 ·7H 2 O and FeSO 4 ·7H 2 O (Table 1). Both high and low levels (1 and −1) of each variable were chosen based on the above preliminary single-factor experiments. Table 2 gave the PB experiment results in 24 h batch fermentation at 35 °C with 130 rpm, while the statistical analysis of PB experiment data was summarized in Table 3. The significance of each variable was determined by their corresponding p-values. Factors evidencing p-values of less than 0.05 were considered to be significant effects on the response 22 , and were subsequently studied in further optimization studies. In this case, X 1 (glucose; p = 0.0015) and X 2 (yeast extract; p = 0.0181) were statistically significant in the production of H 2 (p < 0.05). X 1 had the greatest negative impact on H 2 yield. On the other side, X 2 had positive effect on H 2 yield. X 3 (tryptone), X 5 (KH 2 PO 4 ), X 7 (MgSO 4 ·7H 2 O) and X 8 (FeSO 4 ·7H 2 O) were set at their high levels according to the positive effects although they were insignificant to H 2 yield. Factors such as X 4 (K 2 HPO 4 ) and X 6 (L-cysteine) with negative effects were selected on their low levels. To approach the optimum response, a fitted first-order model equation for H 2 yield (Y) was obtained from the PB design experiments: The coefficient of each variable in Eq. (1) represents the effect weight of the variable on H 2 yield. The quality of the fitted polynomial model equation was expressed by the coefficient of determination (R 2 ). The obtained R 2 was 0.9858, indicating that 98.58% variability in the response could be explained by the model, but only 1.42% variability was not explained. Meanwhile, the value of adjusted determination coefficient (Adj R 2 = 0.9478) was very high, which advocates a high significance of the model 23 . These results show that the response equation provides a suitable model for the PB design experiments.
Path of the steepest ascent. Experiments were started from zero level in PB design. If t-value was positive, climbing direction would increase and vice versa 24 . Overall, step size was determined by the above t-values  Table 8. Structural carbohydrates and lignin in bagasse and Jatropha hulls a . a All data were given by mean ± standard deviation (n = 3).
( Table 3) and single-factor experiments. According to t-values (Table 3) and single-factor results, step size of 2 and 0.5 g/L was set for glucose and yeast extract, respectively. Glucose shows decreasing trend but an increasing trend for yeast extract. The concentrations of other factors were steadfast in all trials at their optimal levels. It was observed that H 2 yield increased along the path from test 1 to 3, reached the peak of 1.78, and decreased from test 4 to 7. This means that the optimal level was close to that in test 3 (16.0 g/L glucose and 6.0 g/L yeast extract). In Table 4, the highest H 2 yield was 1.78 mol H 2 /mol-glucose for test 3. In these experiments, H 2 yield was remarkably improved. This suggests that the steepest ascent method was an effective technique to determine an optimal level. However, the optimal values for these two variables need to be determined via the following CCD.

RSM and CCD.
A series of experiments were carried out via RSM to obtain an optimal combination of glucose and yeast extract. Table 5 gave five different coding levels, full experimental design and the results of CCD, with the center point set based on the results in steepest ascent test (Table 4). Table 6 summarizes the significance test of regression coefficients. Using multivariate regression analysis of the observed data (Table 6), the obtained model illustrates the relationship between glucose (X 1 ) and yeast extract (X 2 ) corresponding to H 2 yield (Y) as below: where Y was the predicted H 2 yield, X 1 and X 2 were the coded values of glucose and yeast extract. The regression model (Eq. 2) can be used to predict the range of H 2 production for various levels of the selected variables. Determination and adjusted determination coefficients (R 2 and Adj R 2 ) were calculated as 91.72% and 85.8% for the regression, demonstrating the agreement between the experimental and predicted H 2 yields that provides a good estimation of the response within the range of process conditions. Analysis of variance (ANOVA) is essential to test the significance and adequacy of model. High t-value reveals an adequate explanation of the variation of data regarding to their mean value. The t-test and p-values served to examine the significance of each coefficient, which also illustrated the interaction strength among independent variables 25 . The t-test of each significant variable level was given in Table 6. In this model term, the linear and quadratic of X 1 , X 1 X 1 and X 2 X 2 (p < 0.05) were recognized as the significant factors 26 , suggesting that H 2 yield was directly related to these two main factors (glucose and yeast extract). However, the interaction between glucose and yeast extract (X 1 X 2 , p > 0.05) seemed to be less significant on H 2 production. F-values of the model and lack of fit were 170.16 and 2.03, respectively (Table 7), while model p-values and p-values of lack of fit were < 0.0001 and 0.2528, suggesting that the model was good fit to the experimental data.
Three dimensional (3D) response surface and corresponding contour of glucose and yeast extract were presented in Fig. 1. The response surface (Fig. 1a) Table 9. Comparisons of product yields from glucose, hydrolysates of bagasse and Jatropha hulls for H 2 production for 24 h fermentation at 35 °C with 130 rpm stirring/shaking in both reactors a . a All data were given by mean ± standard deviation (n = 3). were well-defined. The contour plot was almost circular, suggesting that the interaction of glucose and yeast extract had a less significant effect on H 2 production. In the surface plot, H 2 yield achieved the highest value within the studied region. Canonical correlation analysis (CCA) is a multivariate technique focused on determining the relationship among groups of variables in a datum set 27 . CCA was further conducted (Table 7), at X 1 = − 0.3402 (glucose, 15.66 g/L) and X 2 = 0.0804 (yeast extract, 6.04 g/L), the predicted maximum H 2 yield corresponding to these values was 2.02 mol H 2 /mol-glucose.  21 . The optimized media for growth and H 2 production were different. All the above experiments were conducted in sealed 100 mL bottles. The obtained optimized results were further used for the production of H 2 from actual biomass wastes below.

Raw biomass
Hydrogen from hydrolysates of bagasse and Jatropha hulls. Bagasse and Jatropha hulls mainly consist of cellulose and hemicellulose and lignin 11,18 . The chemical compositions of bagasse and Jatropha hulls were analyzed according to the technical report from US National Renewable Energy Laboratory (NREL) 28,29 (Table 8). Bagasse consisted of 25.17 wt% hemicellulose and 42.05 wt% cellulose, but 13.37 wt% hemicellulose and 36.95 wt% cellulose for Jatropha hulls. Bagasse and Jatropha hulls were hydrolyzed by two-step dilute acid hydrolysis in an autoclave. After reaction, the hydrolysates were neutralized, concentrated, detoxificated, and fermented under the optimized medium obtained from glucose. Figures 2 and 3 illustrated liquid and gas products vs. time in the bottle fermentation of hydrolysates from bagasse and Jatropha hulls, respectively. For the fermentation of bagasse hydrolysate (Fig. 2), as time increased, yields of H 2 , acetic acid and butyric acid increased with the decreasing TRS. These liquid products rose very slowly before 4 h as the strain was in the lag phase for a short adaptation to the new environment 30 . In Fig. 2a, after 5 h, acetic acid and butyric acid grew rapidly to 3.25 and 2.48 g/L at 20 h (vs. 1.26 and 0.62 g/L at 5 h), and slightly rose further to 3.35 and 2.68 g/L at 24 h, respectively. Acetate and butyrate rose slightly after 15 h and stopped increasing after 20 h with the consumption of TRS because the accumulation of acetate and butyrate caused a sharp drop in culture pH and inhibited the growth of bacteria. 2-Propanol also presented similar rising trend. Similarly, H 2 production (Fig. 2b) rose sharply from 3.05 to 115.87 mL at time from 4 to 22 h, and changed little until 24 h. At the same time, total gas rose much higher from 5.29 at 4 h to 228.9 mL at 22 h because CO 2 was produced. For the fermentation of hydrolysate of Jatropha hulls (Fig. 3a), acetic acid and butyric acid increased rapidly to 3.87 and 3.59 g/L at 22 h from 2.16 and 0.17 g/L at 5 h. However, more butyric acid (3.52 g/L) produced at 24 h from the hydrolysate of Jatropha hulls than that of bagasse (2.68 g/L). The initial acetic acid concentration from the hydrolysate of Jatropha hulls (1.96 g/L) was higher than that of bagasse (1.06 g/L) that may inhibit the further production of acetic acid. More acetic acid was produced from the hydrolysate of bagasse (2.28 g/L) than Jatropha hulls (1.97 g/L) at 24 h (Table 9). Contrary, the production of butyric acid from the hydrolysate Jatropha hulls (3.43 g/L) was more than that of bagasse (2.47 g/L). The growth of strains was likely to be inhibited by the acetate produced when grown on media containing glucose 31 . So, less H 2 was produced from hydrolysate of Jatropha hulls (110.4 vs. 115.87 mL for bagasse hydrolysate at 24 h) via acetic route discussed below.
Sugars were also fermented in a 3L fermenter for 24 h at 35 °C with 130 rpm stirring (Table 9). H 2 yield from glucose is 2.15 (mol H 2 /mol-glucose), which is slightly higher than the experiment in bottle fermentation (2.02 mol H 2 /mol-glucose) because hydrogen pressure in bottle fermentation increased as gas accumulated in the fixed 100-mL bottle that may prevent hydrogen production. But, no pressure rose for the fermenter since it was connected on-line to an empty gas bag. Previous study shown that partial pressure of H 2 (pH 2 ) was an extremely important factor for continuous H 2 synthesis 32 . The hydrogen evolution rate and yield were improved by 10% and 15%, respectively, by reducing the H 2 partial pressure through pumping out produced gas in a fermentor with silicone rubber 33 . So, high pressure prevented hydrogen production in bottle fermentation. The hydrogen production also affected by its concentration. As hydrogen concentration increased, H 2 synthesis decreased and metabolic pathways shifted towards the production of more reduced substrates, such as lactate, ethanol, acetone, butanol or alanine 32 . On the other hand, acids produced in the bottles without neutralization may also hamper the production of H 2 . Acetate accumulation caused a sharp drop of culture pH and subsequent inhibition of bacterial hydrogen production, that were reported previously 34 . For fermenter experiments, H 2 yield from hydrolysates of bagasse and Jatropha hulls is 2.06 and 1.95 (mol H 2 /mol-TRS), which reached 95.8% and 90.7% of that from pure glucose, respectively.
Generally, glucose or other reducing sugars (e.g., xylose) were the preferred carbon source for fermentation with overall reactions as below 35 : According to Eq. (3), 4 mol H 2 /mol-glucose was obtained with acetic acid and CO 2 as by-products. However, if butyric acid was formed as by-product, only 2 mol H 2 /mol-glucose was produced (Eq. 4). The ratio of acetic/ butyric acids obtained could be related to the production of H 2 . Previous studies showed that the increase of the ratio was accompanied by increased production of H 2 35 . In Table 9, it was well-confirmed in fermenter that acetic/ butyric acid ratio was 0.67, 1.03 and 1.15 (g/g) for Jatropha hulls, bagasse and glucose, corresponding to their H 2 yield (mol/mol-TRS) of 1.95, 2.06 and 2.15, respectively. As comparison, bottle experiments with lower acetic/ butyric acid ratio (g/g) of 0.57, 0.9 and 1.09 for hydrolysates of Jatropha hulls, bagasse and glucose had lower H 2 yield (mol/mol-TRS) of 1.89, 1.99 and 2.03 (Table 9). Gas is composed of 49.7-64.34% H 2 and 35.66-50.29% CO 2 from anaerobic fermentation in bottles and fermenter. H 2 from glucose (64.34%) had much higher concentration than those from hydrolysates of bagasse (53.92%) and Jatropha hulls (49.7%) in fermenter. The concentrations of H 2 produced from glucose and bagasse in fermenters were higher than those in bottles (64.34% and 53.92% vs. 52.69% and 50.16%). However, H 2 produced from Jatropha hulls (49.7%) in fermenter was less pure than that in bottles (56.62%). /mol-glucose was achieved for 24 h in a well-controlled fermenter. The hydrolysates of bagasse and Jatropha hulls from two-step dilute acid hydrolysis were further successfully fermented to hydrogen with yields of 2.06 and 1.95 (mol H 2 /mol-total reducing sugars) in fermenter, respectively. Higher H 2 yield from glucose and bagasse hydrolysate was due to their high acetic/butyric acid ratio via acetic route for hydrogen production. It is also found that optimized condition for H 2 production was not the best condition for microbial growth. Tropical biomass wastes as inexpensive raw materials can be effectively produce biohydrogen. Microorganism and batch (bottle) fermentation. C. butyricum (CICC 20763) used for producing H 2 was purchased from China Center of Industrial Culture Collection (CICC, Beijing). The culture was maintained on corn meal agar stab at 4 °C. The seed medium composed of (g/L): 20 glucose, 5 yeast extract, 5 tryptone and 3 NaCl with pH 6.5. To make the seed cultivation, a loop of C. butyricum from a fresh slant tube was inoculated into a serum bottle (250 mL) containing 50 mL seed medium and then incubated in a rotary shaker (ZWY-2102C, Zhicheng Analytical Instrument Manufacturing Co., Ltd., Shanghai) at 35 °C with 130 rpm for 12 h.

Materials. Substrates
The initial medium for anaerobic fermentation was composed of (g/L): 25 glucose, 10 tryptone, 5 Na 2 HPO 4 , 3 NaCl and 0.05 FeSO 4 ·7H 2 O. A 4% (v/v) of the seed culture was inoculated into the medium for fermentation. After sterilization at 121 °C for 20 min in an autoclave (HVE-50, Hirayama Manufacturing Corp., Tokyo), substrates with C. butyricum (about 20 mL, initial pH 6.5) were loaded into 100-mL bottles in an anaerobic incubator (YQX-II, CIMO Medical Instrument Manufacturing Co., Ltd., Shanghai) to maintain anaerobic environment by nitrogen purge (99.999% N 2 , Meisel Gas Products Co., Ltd., Kunming, Yunnan) and sealed by rubber-aluminium cap. The bottles with samples were put in an incubator shaker (ZWY-2102C) at 130 rpm and 35 °C for batch fermentation. After fermentation (up to 24 h), gas was collected in a bag (500 mL) by a 50 mL syringe (Agilent Inc., Palo Alto, CA) and its volume (e.g., up to 233 mL at 24 h) was measured by the same syringe. Gas was analyzed by gas chromatograph (GC) and liquid sample by high performance liquid chromatograph (HPLC), ultraviolet (UV)-visible spectrophotometer and biosensor introduced in detail below. PB design. PB design was applied to select factors that significantly influenced H 2 yield, based on the first-order (linear) model as below 22 : i i 0 where Y was H 2 yield as response; X i was coded independent factor and β i was linear coefficient; β 0 was intercept value. Each variable had two-levels, high and low, coded by (+1) and (−1), respectively. Path of steepest ascent method. The method of steepest ascent given by Box and Wilson is a procedure for moving sequentially along the direction of the maximum increase in response 39 . The direction in which H 2 yield increased most rapidly was that of steepest ascent, started from the zero level of variables (significant factors) in PB design, while the step size was decided by the estimated coefficient ratio from Eq. 5, together with practical experience. This test was disused until response no longer increased. As a result, the steepest ascent method allowed factors approaching the optimal level and gave a more limited region for RSM optimization 40 . CCD and RSM. CCD was employed to optimize the two most significant factors (glucose and yeast extract) for the maximum H 2 yield, screened by PB design. These two independent factors were studied at five different levels (−1.41, −1, 0, +1 and +1.41), and a set of 13 experiments were carried out ( Table 5). The behavior of the experiment was explained by the following second-order polynomial equation 23 : where Y was the predict response, β 0 was intercept, X i and X j were input variables which influenced the response Y, β i was linear coefficient, β ii was quadratic coefficient, and β ij was interaction coefficient. The second-order polynomial coefficients were also calculated.
Composition analysis of bagasse and Jatropha hulls. Composition analysis of bagasse and Jatropha hulls was conducted according to the standard US NREL protocols 28,29 . All the raw materials were dried at 45 °C until achieving constant weight. Biomass sample (2-10 g, oven dry weight, ODW) was extracted in a Soxhlet extractor with deionized water and ethanol at boiling points for 12-24 h, respectively. The extracted solutions were dried in a flask to remove water and ethanol at 45 to 60 °C by a rotary evaporator. The extracted solid biomass (0.3 ± 0.01g, dried) was hydrolyzed in a 25-mL tube containing 4.98 ± 0.01 g concentrated sulfuric acid (72 wt%) stirred by a magnetic bar at 30 °C water bath for 1 h, and then diluted to 4 wt% sulfuric acid solution by adding 84 ± 0.01 mL deionized water. The diluted solution with biomass sample was put in a 100 mL bottle sealed by a rubber-aluminum cap and autoclaved at 121 °C for 60 ± 1min (HVE-50) for second-step hydrolysis, at the same time, a reference standard sugar mixture (glucose, xylose, galactose, arabinose and mannose) with similar concentrations to the sugars in biomass sample was also put in the autoclave to calibrate the decomposition of each sugar. The filtrated supernatant was collected for acid-soluble lignin determination by using UV-Visible spectrophotometer at 240 nm wavelength (UV-1800, Shimadzu, Kyoto). The supernatant was neutralized to pH 5.0-6.0 with calcium carbonate for sugar analysis described below. The solid residual separated from the hydrolysate by filtration (0.22 μ m pore size) was oven-dried for acid-insoluble lignin determination. Crucibles containing residues were placed in a muffle furnace (4-10, Ever light medical equipment Co., Ltd, Beijing) at 575 ± 25 °C for 24 h to determine ash weight.
Hydrolysis of Jatropha hulls and bagasse. Two-step dilute acid hydrolysis was conducted in a 500-mL high-pressure autoclave (FCFD05-30, Yantai Jianbang Chemical Mechanical Co. Ltd., Shandong). In the first-step hydrolysis, bagasse or Jatropha hulls (20 g, dry weight) mixed with 2 wt% sulphuric acid solution with solid-liquid ratio (SLR) of 1:10 (g/g) was hydrolyzed at 130 °C for 1 h 41 . After reaction, solid residue was filtrated (0.22 μ m pore size), washed and dried for second-step hydrolysis in 4 wt% sulphuric acid solution with 15 g solid residue, SLR of 1:10 (g/g) at 150 °C for 1 h 11 . Hydrolysates from both steps were mixed together, and neutralized to pH 6.0 with calcium hydroxide. After filtration (0.22 μ m pore size), the solution was concentrated to 15.66 g/L TRS by a rotary evaporator at 60-70 °C. The concentrated solutions (100 mL) were further detoxicated with 2 wt% activated carbon at 50 °C for 2 h, and used for the following fermentations. The hydrolysates were fermented under the optimized medium obtained from glucose in bottles because they mainly contained hexose and pentose sugars with minor of acids, furfural, hybrid protein and other toxic components that were neutralized and detoxified.
Fermentation in a fermenter. Fermentation of glucose and hydrolysates (about 1 L) with C. butyricum was also carried out in a 3L fermenter (LiFlus GX, Hanil Science Industrial Co. Ltd., Incheon, South Korea) at 35 °C for 24 h with 130 rpm stirring. Its pH was controlled at 6.5 by adding 1 M sodium hydroxide or 1 M hydrochloric acid. Gas generated (up to 8.8 L) during the fermentation process was collected into a gas bag (10 L) on-line and its volume was measured after fermentation using a wet gas meter (LMF-1, Shanghai A.K. Instruments Co., Ltd.), where water was saturated by NaHCO 3 to avoid absorption of CO 2 . Gas and liquid products were analyzed described below.
Analytical methods. Gas  where P (0.8 × 10 5 Pa in Kunming at an altitude of 1,896 m, calculated by Barometric formula 42 ) is atmospheric pressure; V H2 is gas volume (m 3 ); R is perfect gas constant (8.3144, m 3 ·Pa·mol −1 · K −1 ); T is gas temperature (298 K), and N is mole number of substrate (e.g., glucose, TRS, glycerol] before fermentation (mol). Liquid samples (glucose, xylose, acetic acid and butyric acid) after fermentation were measured by HPLC (LC-20A, Shimadzu) equipped with a refractive index detector (RID, Shimadzu) setting at 55 °C and Aminex HPX-87H column (Bio-Rad, Hercules, CA). Running temperature was 60 °C and 0.005 M H 2 SO 4 was applied as mobile phase with a flow rate of 0.6 mL/min. Each product was calibrated by its standard solutions with five different concentrations (e.g. 1, 2, 3, 4, 5 g/L). Similarly, sugars from the composition analysis of biomass were analyzed by the HPLC and Hi-Plex Pb column. The mobile phase was de-ionized water at 0.5 mL min −1 . Temperatures of the detector and the column oven were 55 °C and 70 °C, respectively. Concentrations of the monomeric sugar standards were 0.05, 0.1, 0.5, 1, 1.5 and 2 g L −1 . All the standard calibration curves for HPLC analysis obtained with R 2 > 0.998 were used for quantitative calculation.
Microbial mass (DCW, g/L) was determined through measuring the absorbance of broth at 650 nm (OD 650 ) by the UV-visible spectrophotometer. One unit of optical density (OD 650 ) was estimated to be equal to 0.36 g/L DCW (y = 0.2835x + 0.0846, R 2 = 0.9974, Figure S2). Besides cross-checked by HPLC, glucose in liquid samples after fermentation was mainly quickly determined using a biosensor (SBA-40, Institute of Biology, Shandong Academy of Sciences, Jinan) calibrated with glucose standard solution (100 mg/dL). TRS in the hydrolysates after two-step hydrolysis or fermentation was determined by the UV spectrophotometer using dinitrosalicylic acid (DNS) method 43 calibrated with five different concentrations (0, 0.2, 0.4, 0.6, 0.8 and 1.0 g/L) with R 2 > 0.9999.