Biohydrogen Production from Agro-industrial Wastes Using Clostridiumbeijerinckii and Isolated Bacteria as Inoculum

The objective of this work was to evaluate the potential of biohydrogen production in agro-industrial residues, which were soft drink wastewater (SDW), corn steep liquor (CSL), cheese whey (CW), and expired Guaraná soft drink, using one model strain Clostridiumbeijerinckii ATCC 8260 and newly isolated Clostridiumbutyricum DEBB-B348. The agro-industrial residues were characterized, and all of them contained significant concentrations of carbohydrates such as glucose, fructose, lactose, and maltose, and in the case of CSL and CW they also contained significant concentrations of amino acids. After performing subsequent experimental designs, the significant factors were cheese whey concentration, corn steep liquor concentration, and fermentation time for C. beijerinckii, and corn steep liquor concentration and fermentation time for C. butyricum (p ≤ 0.05), with R2 of 0.950 and 0.895, respectively. The maximum hydrogen volume production was 18.5 ± 1.68 mL and 27.4 ± 1.84 mL for each strain, respectively. It is noteworthy the high yield of hydrogen produced with C.beijerinckii (235 mL H2/g COD removed) and C.butyricum (310 mL H2/g COD removed). It was also noted that CSL and/or CW-based culture media do not need to be supplemented with cysteine-HCl. These results indicate a potential hydrogen production process utilizing less expensive substrates.


Introduction
Alternative renewable energies are a transversal area that encompasses many of the sustainable development goals, such as renewable energy, innovation and infrastructure, sustainable cities and communities, responsible consumption, climate action, and life on land. The unacceptable environmental impacts caused by the large use of fossil fuels have boosted research on sustainable biofuels with reduced environmental impacts [1]. Among the existing biofuels, hydrogen ("bioH 2 ") stands out for not producing greenhouse gases during its combustion, for being produced from a variety of energy sources, for being safely stored and transported, and for being converted into electricity [2]. Hydrogen may play an important role in building a more sustainable society, especially in the transportation sector where CO 2 capture from vehicles is virtually impossible.
The renewable production of hydrogen is derived from the dissociation of H 2 O (commercial technology), the thermochemical processing of biomass, and the anaerobic fermentation of organic raw materials. Fermentative hydrogen production ("bioH 2 ") is generated under milder conditions and using (agro) industrial byproducts.
Several types of microorganisms are capable of producing biohydrogen, such as cyanobacteria (Synechocystis sp., Desertifilum sp., Synechococcus sp., Phormidium corium, Synechocystis sp., Oscillatoria sp., Anabaena sp.) [3], microalgae (Scenedesmus obliquus, Chlamydomonas reinhardtii, Chlorella fusca, Platymonas subcordiformis, Chlorella vulgaris) [4], and pure strains or microbial consortia of facultative and anaerobic bacteria [1]. Producing genres involve Bacillus, Citrobacter, Enterobacter, Escherichia, and Rhodopseudomonas [5,6]. The choice of microorganism used for the dark fermentation is of extreme importance as it will define the yield of hydrogen from organic sources. Clostridium deserves special attention due to the highest hydrogen yields when compared to other genera and its tendency to dominate dark fermentation in cases of co-culture [7]. Many species of this genus are capable of producing biohydrogen, such as Clostridium acetobutylicum, Clostridium butyricum, Clostridium beijerinckii, and Clostridium tyrobutyricum [8]. Considering the different characteristics of the microorganisms selected for hydrogen production, the conditions under which the fermentation process will occur must be optimized and specified. Besides the physical factors affecting the bioH 2 production such as pH and temperature, the substrate utilized should provide sufficient nutrients for better microbial growth and activity. Since the agro-industrial residues and wastewater are easily available and represent organic rich substrates generated in large volumes [9], they could be suitable substrates for biohydrogen production [10].
The use of agro-industrial residues as sources for bioH 2 production significantly reduces the costs of production and provides more adequate disposal for these wastes. Recent studies have proven that the use of substrates such as cassava processing wastewater, corn steep liquor (CSL), beverage wastewater, mushroom farm waste, and others are suitable for biohydrogen production in significant quantities [11][12][13][14]. The possibility of using regional wastes for the production of a renewable source of energy has a great impact on local communities, economies, and the environment. The objective of this work was to evaluate the biohydrogen production potential of four agro-industrial residues (soft drink wastewater, corn steep liquor, cheese whey, and expired Guaraná soft drink) using one model strain of C. beijerinckii (ATCC 8260) and newly isolated C. butyricum (DEBB-B348).

Physico-chemical Characterization of Wastewaters
Soft drink wastewater (SDW) and expired Guaraná were donated by the company Ambev (Almirante Tamandaré, Paraná, Brazil). Cheese whey (CW) was provided by the company Anila (Fernandes Pinheiro, Paraná, Brazil), and CSL was provided by the company Ingredion (Balsa Nova, Paraná, Brazil). The industrial wastewaters were stored at − 20 °C until use.
Cations and anions were analyzed by ion chromatography (761 Compact IC, Metrohm AG) using Metrosep C 3250/4.0 and Metrosep A Supp5 250/4.0 columns, respectively. For the quantification of cations, a mobile phase composed of 3.5 mM HNO 3 at a flow rate of 0.9 mL/min was used, while for anions a mobile phase containing 3.2 mM Na 2 CO 3 and 1.0 mM NaHCO 3 at a flow rate of 0.7 mL min −1 was used. Run times were 25 and 30 min, respectively.
The determination of the chemical oxygen demand (COD) and the total nitrogen content were done using Standard Methods for the Examination of Water and Wastewater 1992 [15]. All reagents were obtained from Sigma-Aldrich.
The free amino acid content of CSL and CW was analyzed by an automated SYKAM S433 amino acid analyzer (Eresing, Germany) using the ninhydrin method. Three solutions were used in the identification reactions: buffer A, B, and a ninhydrin reagent. Each liter of buffer A was composed of 11.8 g tri-sodium citrate dihydrate, 6.0 g citric acid, 65 mL methanol, 0.5 g phenol, and 6.5 mL hydrochloric acid (32% v/v). Each liter of buffer B was composed of 19.6 g tri-sodium citrate dihydrate, 3.1 g sodium hydroxide, and 5.0 g boric acid. The pH of buffers A and B were 3.45 and 10.85 respectively. In the case of the ninhydrin reagent, each liter was composed of 600 mL methanol, 20.0 g ninhydrin, 2.0 g phenol, 0.90 g hydrindantin, 400 mL sodium/potassium acetate buffer (4 M), and 35 mL ethylene glycol monomethyl ether. The operating conditions were the following: the flow rates in the gradient and in the amino module were 0.45 mL/min and 0.25 mL/min, respectively, and the reaction temperature in the reactor was 130 °C. This method was also used to determine the amino acid kinetics during the fermentation of the best bioH 2 production condition.

Inoculum
C. beijerinckii ATCC 8260 and an isolated strain from a consortium obtained from sugarcane cultivation soil (LPBAH3) [16] were first compared for their ability to produce biohydrogen from a combination of the industrial wastewaters. Isolation of a new strain was carried out by successive cultivation in Man, Rogosa, & Sharpe (MRS) medium by the pour plate method and incubation in an anaerobic jar. Upon isolation, the strain was submitted to identification. The 16S rRNA gene sequences of the reference strains retrieved from NCBI (National Center for Biotechnology Information, MA, USA) were aligned using the online version of MAFFT program, version 7, with the option Auto (FFT-NS-1, FFT-NS-2, FFT-NS-i, or L-INS-i). A neighborjoining phylogenetic tree was constructed using the MEGA X 10.1 computer [17] based on the MSA file by MAFFT. The evolutionary distances were computed by the maximum composite likelihood method [18] and maximum-parsimony [19]. The robustness of individual branches was estimated by bootstrapping with 1000 replicates [20]. The isolated strain was identified as C. butyricum (DEBB-B348). C. beijerinckii and C. butyricum (DEBB-B348) were kept in a medium composed of soft drink wastewater containing 3% CSL (v/v). The pH of the medium was adjusted to 7.0 using 35% NaOH solution. Procedures according to Balch et al. [21] were adopted to guarantee an anaerobic environment. Oxygen was removed from the medium by boiling (100 to 105 °C) under anoxic conditions under an argon atmosphere. Reduction of redox potential was ensured by the addition of NaHCO 3 (1.0 g/L) and cysteine-HCl (0.5 g/L) at 85 °C and 65 °C, respectively (for inoculants only). All experiments were carried out in Hungate tubes with a working volume of 6 mL and a total volume 16 mL. The tubes were sealed with bakelite caps and autoclavable rubber stoppers. To keep the strains active, they were subcultured weekly by transferring 1 mL of inoculum to fresh medium and incubated at 37 °C.

Experimental Design, Optimization, and Statistical Analysis
A Plackett-Burman design for six parameters with a total of 12 runs was used for each strain. The parameters studied were (i) , and (vi) cheese whey (CW) 0-40% (v/v). The Pareto diagram was used to determine the effects of significance in the experiments performed. All statistical analyses were done with a 95% significance level.

Kinetics of bioH 2 and Volatile Fatty Acids
The kinetics of hydrogen and volatile fatty acid production were carried out in optimized media under anaerobic conditions. All experiments were conducted in Hungate tubes with a working volume of 6 mL that were sealed with bakelite caps and autoclavable rubber stoppers. The pH of the medium was adjusted to 7.0 using 35% NaOH solution. The inoculum rate was 10%, and the temperature of incubation was 37 °C. Incubation was carried out in an ORION 502 incubator.
Quantification of hydrogen production and volatile fatty acids was carried periodically at 8 h and 16 h intervals for C. butyricum and C. beijerinckii, respectively. A glass syringe was used to collect the produced gas, which was analyzed in a 490 Micro GC System gas chromatograph (Agilent) equipped with two columns (Molsieve 5 Å and PoraPLOT U) and a thermal conductivity detector (TCD). In the Molsieve 5 Å column, the injection temperature was 110 °C, with an injection time of 20 ms, column temperature of 90 °C, and an initial pressure of 190 kPa. In the PoraPLOT U (PPU) column, the injection temperature was 110 °C, with a column temperature of 90 °C and an initial pressure of 150 kPa. The time for each run was 1.2 min. The mobile phase used was argon gas with a purity of 99.99%.
(1) The qualitative and quantitative analysis of volatile fatty acids and substrate consumption were carried out in a HPLC Agilent 1260 Infinity Quaternary LC with RI detector and Hi-Plex-H column. The detector and column temperature were 50 °C and 60 °C, respectively. The mobile phase used was 5 mM H 2 SO 4 at a flow rate of 0.6 mL/min, and the injection volume was 10 μL. Prior to injection (1 mL), the liquid samples were submitted to centrifugation at 6000 rpm for 10 min in microcentrifuge Thermo Scientific 75,002,411 and microfiltration in cellulose acetate membranes (0.22 µm). All reagents used in the chromatography were HPLC grade.

Physico-chemical Characterization of Wastewaters
The physicochemical composition of the three studied agroindustrial residues (CSL, SDW, and CW) was different in terms of carbohydrates, organic acids, nitrogen, cations, and anions ( Table 2). The expired refrigerant "Guaraná" was not characterized since it was not significant. The CW was the only waste that presented lactose in detectable quantities (4.55 g L −1 ), while glucose, fructose, and maltose were undetectable. Similar results have been documented in whey used for hydrogen production by Escherichia coli, but with a higher content of lactose (15 g L −1 ) [22,23]. Both CSL and soft drink wastewater contained glucose and fructose; however, CSL exhibited much higher concentrations (> 35 g L −1 for each carbohydrate, versus 0.7 g L −1 and 3.98 g L −1 of glucose and fructose, respectively, in SDW). The CSL used in this study is richer in the cited carbohydrates than the used by Martinez-Burgos et al. [24] for the production of bioH 2 . It is important to highlight that CSL also presented an elevated concentration of lactic acid (92.5 g L −1 ) which can be used as a second carbon source for some lactic acid bacteria and species of the Clostridium genus [25,26].
According to the organic matter composition, all effluents showed high COD. The COD of CW and CSL was similar to that described in the literature [27][28][29][30]. SDW, however, presented lower values for COD in comparison to the scientific literature [13,31,32] and was more similar to beverage wastewater [33].
The presence and concentration of Mg, Fe, Na, and K have a significant effect on hydrogen productivity, especially the first two [34]. Mg, Na, and K were observed in high concentrations in all wastewaters. The concentration of Mg observed in SDW, although lower than in CW and CSL, is enough to stimulate biohydrogen production metabolism. Fe was not detected, and its supplementation could enhance the productivity obtained in this study [35][36][37]. Ca was another cation observed at high concentrations for all the three wastewaters. Mn and Cu, which are widely recognized as enzymes cofactors, were not identified. Supplementation with Mn and Cu could be considered in future studies.
CSL and CW showed up as important sources of nitrogen (10.5 and 3.6, respectively), values comparable to other research [23,38]. This content is sufficient to support biomass production without the need to supplement with other nitrogen sources, such as yeast extract, meat extract, and ammonium sulphate [22,39], which have a significant impact on the process economy. This macro nutrient is important mainly in microbial growth and indirectly in hydrogen production, since the energy vector is associated with microbial growth [11]. The amino acid composition of CSL and CW showed that L-cysteine, a recurring supplement for hydrogen production fermentations [37,40,41], is present in CW and CSL in significant amounts. The presence of methionine, alanine, histidine, and lysine in ranges from 1.0 to 10.0 g L −1 may also have a positive impact on biohydrogen production. The synergetic effect of these amino acids with ferric oxide was described to increase the volumetric hydrogen produced 1.3 times [42,43]. The higher amino acid concentrations found in this work were methionine and valine for CSL and histidine and arginine for CW. According to Hofer et al. [26], phenylalanine and isoleucine should be primarily free in the effluents, while arginine, histidine, and lysine are essentially bound in proteins or peptides. Thus, arginine, histidine, and lysine concentrations in CW may be underestimated.
The initial pH is relevant for biohydrogen production. The optimum pH for Clostridium is approximately 6.0 [44,45], while fermentation is severely impaired under pH 4.0 or over pH 12.0 [14]. Although all the fermentations were adjusted to pH 7.0, it is interesting to consider the amount of alkali necessary to correct the pH of CSL and CW. Despite the chemical composition of SDW not being remarkably interesting in terms of carbohydrates, cations, and anions, it could be used as a dilutant to facilitate pH correction.

Clostridium butyricum Identification
According to the BLAST analysis and phylogenetic tree, the isolate strain was identified as C. butyricum with 99% similarity. This strain was registered in GenBank under the accession number for nucleotide sequence SUB7395486 Clostridium MT435521. The phylogenetic tree constructed with the neighbor-joining method revealed that the C. butyricum was included in the same subcluster with Clostridium saccharoperbutylacetonicum, Clostridium puniceum, Clostridium saccharobutylicum, Clostridium chromiireducens, Clostridium diolis, C. beijerinckii, and Clostridium neonatale.

Biohydrogen Production Optimization
The analysis of variance (ANOVA) p values of Plackett-Burman designs were 0.009 and 0.000 for C. beijerinckii and C. butyricum, respectively. This means that at least one of the tested variables has a significant effect (p ≤ 0.05) on hydrogen production. Posteriorly, in the Pareto diagram (Fig. 1), it was shown that among the five variables tested, only %CSL and fermentation time significantly affected the production of hydrogen (p ≤ 0.05) for C. butyricum strain. For C. beijerinckii, %CW, %CSL, and fermentation time were significant (p ≤ 0.05). Previous works demonstrate that the concentration of CSL, whey, and the fermentation time affect hydrogen production via dark fermentation [24,29], which can be explained by the fact that both substrates besides being nitrogen sources also contain significant amounts of carbon sources by which they affect microbial growth and indirectly hydrogen production [7]. In the case of time, it is another key Fig. 1 Pareto chart from Plackett-Burman design for C. butyricum and C. beijerinckii variable that must be controlled, as microbial growth is dependent on the incubation time [46].
The inoculum rate was not significant (p ≤ 0.05), which may have been because large inoculum volumes were tested; an inoculum rate of 10% was then used for all experiments. It was expected that due to its large amount sugar, the expired soft drink would be significant in all tests performed. However, this variable was not significant (p ≤ 0.05). Soft drinks were added to the medium in a concentration of 20%, and even though it was diluted, the antimicrobial properties of the finished product were still outstanding. According to Kregiel [47], soft drinks contain chemical preservatives and acids that play an important role in avoiding microbial growth. Sorbates and benzoates present in the soft drink act together to enhance antimicrobial effectiveness against bacteria and other microorganisms. Notwithstanding, the author still emphasizes that this combination causes the inhibition of amino acid uptake and destroys the internal proton level of microbial cells.
On the other hand, the contour surface graphs ( Fig. 2A,B) showed that the optimal hydrogen production is reached with 12-14% CSL, at 37.5-48 h time of fermentation and 38.5-45% CW for C. beijerinckii. In the case of C. butyricum, the optimum hydrogen production was 10-14% CSL and 36-52 h (Fig. 2C).

Mathematical Models
Two second-order polynomial models that simulate the production of biohydrogen under the conditions described are presented in Eqs. 2 (C. beijerinckii) and 3 (C. butyricum).
The mathematical models presented an R 2 of 0.95 and 0.895 for Eqs. 2 and 3, respectively, which indicates that 95% and 89.5% of hydrogen production are governed by the variables selected. Also, the values of adjusted R 2 for C. beijerinckii and C. butyricum were adj R 2 = 94.50 and adj R 2 = 81.15, respectively. According to Hye et al. and Martinez-Burgos et al. [25,48], mathematical models with R 2 ≥ 80% are adequate to simulate the behavior of the response variables.
On the other hand, the models are valid in both cases, since the three assumptions are satisfied: (1) The residuals fit a normal distribution (normality assumption). (2) The residues are completely dispersed; they do not present trends (independence assumption). In fact, the Durbin-Watson statistic values ( 2.0 and 1.79 for Eqs. 2 and 3, respectively) and the homoscedasticity were respected (the p value of the Bartlett test was equal to 0.137 and 0.872 for Eqs. 2 and 3, respectively) (homoscedasticity assumption).
According to the mathematical model (Eq. 2), the optimal conditions for the production of biohydrogen with C. beijerinckii were CSL = 12%, t = 48 h, and CW = 40%. The experimental and predicted hydrogen volumes at the optimum condition were 18.5 ± 1.68 mL and 16.87 mL, respectively. In the case of C. butyricum, the conditions were CSL = 12% and t = 48 h. The experimental and predicted biohydrogen volumes for these strains were 27.4 ± 1.84 mL and 26.5 mL, respectively. In both cases, there was no

Evaluation of Biohydrogen Production Under Optimum Conditions
Under the optimum conditions selected, the maximum volume of hydrogen produced by the bacteria C. butyricum and C. beijerinckii was 27.47 ± 1.8 mL and 19.05 ± 0.70 mL, respectively (Fig. 3A). It was also shown that after 32 h and 48 h for C. butyricum and C. beijerinckii, respectively, the production of hydrogen did not improve so the fermentation process could be stopped at this time. It is noteworthy that the yield of hydrogen produced with C. butyricum (310 mL H 2 /g COD removed ) was approximately 40% greater than that obtained as C. beijerinckii (235 mL H 2 /g COD removed ). The maximum production rate was 0.12 L H 2 /L h and 0.08 L H 2 /L h for C. butyricum and C. beijerinckii, respectively. These yields are higher than those reported in other studies that had been supplemented with some compounds. Krishnan et al. [49] supplemented the effluent with Fe 2+ , peptone, and Na 2 HPO 4 + 2H 2 O and obtained a yield of 215 mL H 2 /g COD removed . Zampol et al. [50] supplemented the medium with KH 2 PO 4 , K 2 HPO 4 , MgSO 4 , NaCl, CaCl 2 , yeast extract, iron citrate and reported a maximum yield of 137 mL H 2 /g COD removed . Ozkan et al. [51] supplemented the effluent with basal medium and obtained a yield of 115.6 mL H 2 /g COD removed .
In both cases, the main carbon sources such as glucose and fructose were consumed almost completely (Figs. 4 and 5). In the case of fermentation with C. beijerinckii, a slight increase in the concentration of fructose and glucose in the medium was observed after 40 h of fermentation (Fig. 4).  This could be explained as SDW may contain significant amounts of sucrose which is hydrolyzed by the microorganism's extracellular enzymes to glucose and fructose [52,53].
In addition, it was observed that the microorganisms could use lactic acid as an alternative carbon source. The metabolites produced in the fermentation were mainly butyric and acetic acid, with predominance of the first. It is noteworthy that the production of the latter was greater with C. butyricum which could be an explanation of the greater production of hydrogen with this strain; acetate generates two times more H 2 than butyrate [54]. Also, it was observed that part of the acetic acid and the butyric acid produced was consumed in the same fermentation, as these can also be used as alternative carbon sources. Martinez-Burgos et al. and Matsumoto et al. [25,55] showed that acetic and butyric acid can be used as carbon sources by some microorganisms in dark fermentation. Likewise, both microorganisms produced traces of propionic acid. However, the production of this metabolite is not desirable in dark fermentation because stoichiometrically hydrogen is needed for its production [8,11].
On the other hand, it was observed that for both microorganisms studied, there was no significant difference (p value ≤ 0.05) between the hydrogen production between the cysteine-HCl-supplemented and not supplemented media. Thus, the volumes of hydrogen produced by the C. beijerinckii strain with and without cysteine-HCl supplementation were 19.20 ± 0.63 and 18.04 ± 1.11 mL, respectively. In the case of the C. butyricum strain, the volumes of hydrogen production with and without amino acid supplementation were 28.52 ± 1.02 mL and 29.10 ± 0.73 mL. This indicates that the CSL can supply the amino acid demands of the microorganisms, and its use could decrease the costs of the culture media destined for the production of hydrogen via biological production.
The amino acid profile during the C. butyricum fermentation (Fig. 6) shows a significant consumption in the amino acids cysteine and threonine available in the medium. In the case of cysteine, it went from 10.85 ± 1.90 µmol mL −1 to 4.5 ± 0.42 µmol mL −1 at the beginning and end of fermentation, respectively. For threonine, it went from 11.85 ± 2.33 µmol mL −1 to 4.5 ± 0.44 µmol mL −1 at the end of fermentation. The consumption of these amino acids could be explained since amino acids can be used as enzymatic cofactors of hydrogenases in dark fermentation [42,43]. The methionine and valine concentrations were maintained above the concentration of the other amino acids, and most of them showed similar concentrations in the beginning and end of the fermentation time. Sharma and Melkania [42,43] tested the effect of methionine, alanine, histidine, cysteine, and lysine in the production of hydrogen by coculture of E. coli and Enterobacter aerogenes in municipal solid waste. The best productivities were with the addition of alanine and cysteine, and the butyrate concentration was increased during the fermentation, favoring the butyrate hydrogenogenic pathway. A similar increase in the butyric acid occurred in this work, possibly explaining why cysteine was consumed. Several authors indicate that cysteine supplementation in the medium enhances biohydrogen production [36,37,40,41]. The bioavailability of each amino acid also indicates which can be used and directed for the metabolic pathways. Hofer et al. [26] found that the content of threonine, tryptophan, and phenylalanine was more than 60% free in CSL, while cysteine and valine were around 35% free in the solution. The cysteine and threonine found in this work were also likely available for C. butyricum dark fermentation.

Conclusion
This work showed that the species C. beijerinckii ATCC 8260 and the identified strain C. butyricum DEBB-B348 are capable of producing hydrogen from soft drink wastewater, corn steep liquor, and cheese whey. High yields were obtained without the need for supplementation of the fermentation medium. The mathematical models reached by the RCCD showed R 2 of 0.950 and 0.895, with fermentation time and corn steep liquor concentration being the most significant factors. The carbohydrate consumption and organic acid formation kinetics also indicate that the butyrate metabolic pathway was favored. The results show an interesting alternative to the production of hydrogen through dark fermentation, giving a sustainable alternative for agro-industrial waste disposal and generating an eco-friendly fuel.

Funding
The authors want to thank the National Council of Technological and Scientific Development (CNPq) and Coordination for the Improvement of Higher Education Personnel (CAPES), Brazil, for the financial support. Fig. 6 Behavior of the consumption of amino acids in dark fermentation by C. butyricum