- Email: [email protected]
Improved bioconversion of crude glycerol to hydrogen by statistical optimization of media components
Renewable Energy 75 (2015) 583e589
Contents lists available at ScienceDirect
Renewable Energy journal homepage: www.elsevier.com/locate/renene
Improved bioconversion of crude glycerol to hydrogen by statistical optimization of media components Rahul Mangayil a, *, Tommi Aho a, b, Matti Karp a, Ville Santala a a b
Department of Chemistry and Bioengineering, Tampere University of Technology, Tampere, Finland Department of Signal Processing, Tampere University of Technology, Tampere, Finland
a r t i c l e i n f o
a b s t r a c t
Article history: Received 12 December 2013 Accepted 18 October 2014 Available online
Bioconversion of crude glycerol to hydrogen has gained importance as it addresses both sustainable energy production and waste disposal issues. Until recently, statistical optimizations of crude glycerol bioconversion to hydrogen have been greatly focused on pure strains. In this study, biohydrogen production from crude glycerol by an enriched microbial culture (predominated with Clostridium species) was improved by statistical optimization of media components. PlacketteBurman design identified MgCl2.6H2O and KCl with negative effect on hydrogen production and selected NH4Cl, K2HPO4 and KH2PO4 as significant variables. BoxeBehnken design indicated the optimal region beyond design area and studies were continued by ridge analysis. Central composite face centered design envisaged a maximal hydrogen yield of 1.41 mol-H2/mol-glycerolconsumed at concentrations 4.40 g/L and 2.27 g/L for NH4Cl and KH2PO4 respectively. Confirmation experiment with the optimized media (NH4Cl, 4.40 g/L; K2HPO4, 1.6 g/L; KH2PO4, 2.27 g/L; MgCl2.6H2O, 1.0 g/L; KCl, 1.0 g/L; Na-acetate.3H2O, 1.0 g/L and tryptone, 2.0 g/L) revealed an excellent correlation between predicted and experimental hydrogen yield. Optimization of media components by design of experiments enhanced hydrogen yield by 29%. © 2014 Elsevier Ltd. All rights reserved.
Keywords: Biohydrogen Crude glycerol Optimization Response surface methodology
1. Introduction Concern over increasing environmental pollution and the need for a sustainable energy source have led to an increased global biodiesel production for its use as a diesel substitute. This increased biodiesel production has resulted in severe waste disposal crisis due to the production of proportionally equivalent amounts of crude glycerol as the by-product [1]. Since crude glycerol is an excellent carbon source, bioconversion of crude glycerol has been an interesting area of study. Moreover, the process also offers a sustainable method for the disposal of biodiesel waste. Bioconversion of crude glycerol to hydrogen (H2) [2,3], 1,3-propanediol [4], ethanol [5], and lactic acid [6] have been extensively studied. Fermentative H2 production from crude glycerol has gained much importance due to high H2 content and is a promising substrate for sustainable energy production with high energy content [7]. Selembo et al. [8] studied H2 production, in batch type fermentation, from crude glycerol using mixed inoculum, reporting
* Corresponding author. Department of Chemistry and Bioengineering, Tampere University of Technology, P.O. Box 541, 33101 Tampere, Finland. Tel.: þ358 3 3115 2052; fax: þ358 3 3115 2869. E-mail address: rahul.mangayil@tut.fi (R. Mangayil). http://dx.doi.org/10.1016/j.renene.2014.10.051 0960-1481/© 2014 Elsevier Ltd. All rights reserved.
a yield of 0.31 mol-H2/mol-glycerol. Efficient bioconversion of crude glycerol to H2 was reported by Ngo et al. [9], wherein pretreated crude glycerol was used as the substrate, assisted with N2 sparging and pH control, using Thermotoga neapolitana, reporting a yield of 2.73 mol-H2/mol-glycerolconsumed. In a recent publication, Jitrwung et al. [10] investigated the effect of optimizing media optimization on bioconversion of crude glycerol to H2 by Enterobacter aerogenes, reporting a yield of 0.84 mol-H2/mol-glycerol. Eco-biotechnological application for improved bioconversion of crude glycerol has been recently studied. Comparison of crude glycerol utilization from different bio-diesel plants and its bioconversion to H2 by a functional mesophilic microbial consortium enriched from activated sludge was investigated by Varrone et al. [11]. The authors reported a H2 yield of 0.90 mol-H2/molglycerol from minimal media supplement with 15 g/L crude glycerol and a 97% degradation efficiency of the sole carbon source. Implementing statistical optimization for improved biohydrogen production from crude glycerol by mixed microbial inoculum has gained importance. Varrone et al. [12] have reported on improved H2 production from crude glycerol using statistical approach in optimizing the culture conditions. They investigated the biohydrogen production efficiency of mesophilic mixed microbial culture on crude glycerol as the carbon source, reporting a yield of 0.96 mol-H2/mol-glycerol. Apart from culture conditions such as
584
R. Mangayil et al. / Renewable Energy 75 (2015) 583e589
temperature, pH, inoculums size and optimal substrate concentration, microorganisms require appropriate concentrations of media components for optimal metabolism during fermentation process [13,14]. N2 is an important media component, assisting in the synthesis of proteins, nucleic acids and enzymes, significant for the microbial growth and hydrogenase enzyme activity [14,15]. Optimal phosphate concentration positively affects the H2 production due to its buffering capacity. High phosphate and potassium ion concentrations are reported to cause an increased cytoplasmic osmotic pressure, thus negatively affecting the microbial growth [15,16]. Magnesium, an important component for cell wall and cell membrane synthesis and an enzyme cofactor, also plays a pivotal role in microbial growth and metabolism [17]. Our previous study focused on enriching H2 producing mixed microbial culture. This study aims in improving the earlier reported H2 yield by optimizing the current media composition, without altering the original media components, using design of experiments (DoE) and response surface methodology (RSM). DoE and RSM are efficient approaches for such optimization tasks, as shown by previous studies [12e14]. In this study, PlacketteBurman design was used to screen the significant media components [18]. The path of steepest ascent [19], BoxeBehnken [20], ridge analysis [21e24] and Central Composite Face Centered Design (CCD) [25] models were used to optimize the concentrations of screened media components. Finally, the validation of the model was performed at the predicted optimal concentrations. 2. Materials and methods 2.1. Glycerol source and culture conditions Industrial glycerol was kindly provided by Savon Siemen Oy (Iisalmi, Finland). The crude glycerol contained 45% (v/v) glycerol and 30% (v/v) methanol with an alkaline pH (~12). The inoculum used for this study was an enriched microbial consortium (mainly comprised of Clostridium sporogenes strain CL3 (Accession no: JF836014.1), Clostridium subterminale isolate DSM 758 (Accession no: EU857637) and uncultured bacterium clone (Accession no: FJ512181.1)) [3] from activated sludge enriched in modified HM100 medium (NH4Cl, 1.0 g/L; K2HPO4, 0.3 g/L; KH2PO4, 0.3 g/L; MgCl2.6H2O, 2.0 g/L; KCl, 4.0 g/L; Na-acetate.3H2O, 1.0 g/L; tryptone, 2.0 g/L; Na-dithionite, 0.5 g/L and resazurin, 0.002 g/L) amended with crude glycerol [3]. The pre-inoculum was prepared by inoculating 10% of the enriched microbial community in 120 ml serum bottles with working volume of 50 ml HM100 medium containing pure glycerol (5 g/L). The pre-inoculum was grown at 150 rpm with an initial pH of 6.5 and cultivation temperature at 40 C. For optimization studies, ammonium chloride (NH4Cl), dipotassium phosphate (K2HPO4), potassium di-hydrogen phosphate (KH2PO4), magnesium chloride hexa-hydrate (MgCl2.6H2O) and potassium chloride (KCl) were chosen. The media was prepared in similar way, varying the selected variables and keeping the remaining media components in concentrations as of the original media composition. The DoE studies were conducted in batch type fermentation with 120 ml serum bottles with a working volume of 50 ml sterile anoxic crude glycerol (1 g/L) amended enrichment medium. The investigations were performed as triplicate experiments at 40 C and 150 rpm with a growth period of 72 h. 2.2. Analytical methods Crude glycerol, volatile fatty acids and alcohols were analyzed by high performance liquid chromatography (HPLC; LC-20AD, Shimadzu, Japan). The HPLC was equipped with a 300 mm 8 mm Shodex SUGAR (SH1011) column and a refractive
index detector (RID-10A). The HPLC samples were prepared as described previously [18]. The proportions of H2 gas were measured using gas chromatograph (GC-2014, Shimadzu GC), fitted with a thermal conductivity detector and a 2 m 2 mm PORAPAK column. N2 was used as the carrier gas with a flow rate of 20 ml/ min. The temperatures of column, detector and oven were maintained at 80 C, 110 C and 80 C respectively. Measurements for each sample were repeated twice and averaged. Substrate blank (i.e. cultivation without the crude glycerol) was included in all experiment sets to deduct H2 produced from tryptone in the growth media. The H2 yield values were calculated as described previously [26]. The presence of methanol in crude glycerol and its utilization by the microbial community was tested. From growth-curve test and end-metabolite distribution it was observed that the functional community did not utilize methanol (data not shown). Hence, methanol was excluded from the H2 yield calculations. 2.3. Optimization procedure 2.3.1. PlacketteBurman design PlacketteBurman design [18] was chosen to screen and identify variables that had significant influence on the bioconversion of crude glycerol to H2. PlacketteBurman design is based on the first order polynomial model:
Y ¼ b0 þ
X
bi Xi
(1)
where Y is the response, b0 is the model intercept and bi is the linear coefficient, and Xi is the level of independent variable. Five media components, NH4Cl, K2HPO4, KH2PO4, MgCl2.6H2O and KCl, were selected to investigate the trivial components influencing H2 production. Based on PlacketteBurman design, each factor was prepared in two levels: 1 for low level and þ1 for high level (Table 1). The main effect of each variable to the response was then calculated as the difference between the averages of the high level (þ1) and the low level (1) response measurements of that variable. In this study, five assigned variables were screened in eight experimental designs. The level of each factor used in the experimental design along with the response and details of the linear model constructed using the PlacketteBurman design is reported in Tables 1 and 2 respectively. The analysis of variance (ANOVA) was performed to select the variable with significant effect (p value > 0.05) on the response. The results presented were average H2 produced from triplicate experiments with standard deviations. The Design-Expert (version 8.0.7.1, Stat-Ease Inc., MN, USA) was used for statistical analysis. 2.3.2. Path of steepest ascent The identified significant variables from PlacketteBurman design were subjected to steepest ascent approach. This optimization step helps to move the experimental region from the design center towards the direction of an optimal response. The path of steepest ascent is based on the first order polynomial model (Eq. (1)) obtained from the PlacketteBurman design and provides information on the test range of selected variables for further optimization steps [19]. Table 3 reports the concentration range of the selected variables to move the design area towards optimal response. The average yield, with standard deviations, was calculated from triplicate experiments for each design points. 2.3.3. BoxeBehnken design For further optimization of the selected media components for enhanced H2 production, a three-variable BoxeBehnken design
R. Mangayil et al. / Renewable Energy 75 (2015) 583e589
585
Table 1 PlacketteBurman experimental design with five independent variables along with their response, YH2 (mol-H2/mol-glycerolconsumed). NH4Cl
Run
1 2 3 4 5 6 7 8
K2HPO4
KH2PO4
MgCl2.6H2O
YH2
Code X1
X2 (g/L)
Code X2
X3 (g/L)
Code X3
X4 (g/L)
Code X4
X4 (g/L)
Code X4
2.00 0.50 2.00 0.50 2.00 0.50 2.00 0.50
þ1 1 þ1 1 þ1 1 þ1 1
1.00 1.00 0.10 0.10 1.00 1.00 0.10 0.10
þ1 þ1 1 1 þ1 þ1 1 1
1.00 0.10 0.10 1.00 1.00 0.10 0.10 1.00
þ1 1 1 þ1 þ1 1 1 þ1
4.00 4.00 4.00 4.00 1.00 1.00 1.00 1.00
þ1 þ1 þ1 þ1 1 1 1 1
8.00 1.00 8.00 1.00 1.00 8.00 1.00 8.00
þ1 1 þ1 1 1 þ1 1 þ1
[20] with three replicates at the center point was performed. The design and the levels of variable are summarized in Table 4. The experimental design consisted of 15 runs including three center points. Each experimental design run was performed in triplicates and the data produced were averaged with standard deviations. The optimal point for H2 production was predicted using a second order polynomial function that was fitted to correlate the relationship between the predicted response and the variables:
Y ¼ b0 þ
KCl
X1 (g/L)
X
bi Xi þ
X
bii Xi2 þ
X
bij Xi Xj
0.37 0.32 0.15 0.32 0.72 0.35 0.57 0.35
± ± ± ± ± ± ± ±
0.14 0.04 0.02 0.04 0.04 0.04 0.07 0.03
that the predicted response for H2 production was slightly over the theoretical maximum at the end of the search path. Table 6 details the ridge analysis design for the selected variables along with distance from the center point and predicted response (yhat). All the trial runs were performed in triplicates, along with a substrate blank and the results were averaged with standard deviations. The design points in the ridge analysis were determined using RSM package [Extra1] in R software environment for statistical computing and graphics [27].
(2)
where Y is the predicted response; b0 , a constant; bi , the linear coefficients; bii , the squared coefficients; and bij , the cross-product coefficients. The details of the second order polynomial model constructed using the BoxeBehnken design are provided in Table 5. The DesignExpert 8.0.7.1 (Stat-Ease Inc., MN, USA) demo version was used for the regression and graphical analysis of the experimental data. ANOVA was performed in order to evaluate the statistical significance of the model.
2.3.4. Ridge analysis Ridge analysis [21,22] was applied in determining the search path on the second order polynomial surface that was constructed with the help of BoxeBehnken design. Ridge analysis, non-linear analog to the steepest ascent method, uses the fitted second order polynomial function (Eq. (2)) for optimizing the response. This method iteratively moves specified step sizes in the search space and, in each step, allows to identify the point at which the response is maximal [23,24]. In order to span the search path long enough, i.e. to include the maximal response, ridge analysis was designed so Table 2 Details of the linear model constructed using the PlacketteBurman design. Code
Variable
Low level (g/L)
High level (g/L)
Coefficient
F Value
p-Value (Prob > F)
X1 X2 X3 X4 X5
NH4Cl K2HPO4 KH2PO4 MgCl2.6H2O KCl
0.50 0.10 0.10 1.00 1.00
2.00 1.00 1.00 4.00 8.00
0.059 0.047 0.048 0.10 0.088
48.74 30.32 31.76 148.39 106.19
0.0199 0.0314 0.0301 0.0067 0.0093
Table 3 Steepest ascent experimental design to move the experimental area towards optimal response, YH2 (mol-H2/mol-glycerolconsumed). Run
NH4Cl (g/L)
K2HPO4 (g/L)
KH2PO4 (g/L)
YH2
1 2 3 4 5 6
1.25 2 2.8 3.6 4.4 5.2
0.6 1.6 2.6 3.7 4.7 5.6
0.6 1.6 2.7 3.8 4.8 5.9
0.43 1.04 0.87 0.49 0.47 0.44
± ± ± ± ± ±
0.10 0.10 0.32 0.18 0.10 0.06
2.3.5. Central composite face-centered design Based on the experimental results from ridge analysis runs, NH4Cl and KH2PO4 were selected for CCD, keeping the other media components at optimized concentrations. For the design model, three different coded levels for each factor were used: 1 (low), 0 (center) and þ1 (high). The designs of all 13 experiments are listed in Table 7. Table 8 provides the details of the second order polynomial model constructed using the CCD. Experiments were performed in triplicates and the data produced were the mean value of the repeats with standard deviations. Inclusion of a substrate blank assisted in calculating the H2 produced exclusively from crude glycerol. The Design-Expert (version 8.0.7.1, Stat-Ease Inc., MN, USA) was used for building the model and statistical analysis for the experimental data using ANOVA. 3. Results and discussion 3.1. Screening of important media components for H2 production from crude glycerol Nutrients play a pivot role in microbial cell growth and bioconversion. The relative significance of five media components, Table 4 BoxeBehnken experimental design with three independent variables along with their response, YH2 (mol-H2/mol-glycerolconsumed). Run
NH4Cl X1 (g/L)
Code X1
X2 (g/L)
Code X2
X3 (g/L)
Code X3
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
0.50 3.50 0.50 3.50 0.50 3.50 0.50 3.50 2.00 2.00 2.00 2.00 2.00 2.00 2.00
1 þ1 1 þ1 1 þ1 1 þ1 0 0 0 0 0 0 0
1.10 1.10 2.10 2.10 1.60 1.60 1.60 1.60 1.10 2.10 1.10 2.10 1.60 1.60 1.60
1 1 þ1 þ1 0 0 0 0 1 þ1 1 þ1 0 0 0
1.60 1.60 1.60 1.60 1.10 1.10 2.10 2.10 1.10 1.10 2.10 2.10 1.60 1.60 1.60
0 0 0 0 1 1 þ1 þ1 1 1 þ1 þ1 0 0 0
K2HPO4
KH2PO4
YH2
0.56 0.90 0.57 1.09 0.83 1.14 0.77 1.35 0.76 0.94 0.91 0.89 1.04 1.05 1.07
± ± ± ± ± ± ± ± ± ± ± ± ± ± ±
0.05 0.00 0.02 0.06 0.02 0.06 0.04 0.03 0.02 0.01 0.03 0.07 0.02 0.03 0.02
586
R. Mangayil et al. / Renewable Energy 75 (2015) 583e589
Table 5 Details of the second order polynomial model constructed using the BoxeBehnken design.
Table 8 Details of the second order polynomial model constructed using the Central composite face-centered design.
Code
Variable
Low level (g/L)
High level (g/L)
Coefficient
F Value
p-Value (Prob > F)
Code
Variable
Low level (g/L)
High level (g/L)
Coefficient
F Value
p-Value (Prob > F)
X1 X2 X3
NH4Cl K2HPO4 KH2PO4
0.50 1.10 1.10
3.50 2.10 2.10
0.22 0.030 0.046
1940.10 36.13 85.26
0.0001 0.0018 0.0003
X1 X2
NH4Cl KH2PO4
2.95 1.52
5.15 3.52
0.099 0.065
75.36 32.83
0.0001 0.0007
3.2. Path of steepest ascent Table 6 Ridge analysis design response, YH2 (mol-H2/mol-glycerolconsumed). a
b
Run
Distance
NH4Cl (g/L)
K2HPO4 (g/L)
KH2PO4 (g/L)
yhat
YH2
1 2 3 4 5 6
0.00 1.15 2.30 3.45 4.60 5.75
2.00 3.33 4.05 4.63 5.17 5.71
1.60 1.64 1.62 1.58 1.55 1.51
1.60 1.96 2.53 3.08 3.64 4.20
1.05 1.28 1.57 1.96 2.46 3.09
1.03 1.09 1.39 1.19 1.01 0.81
a b
± ± ± ± ± ±
0.03 0.03 0.10 0.09 0.03 0.04
Distance of each run from the center point. Predicted response.
3.3. BoxeBehnken design
NH4Cl, K2HPO4, KH2PO4, MgCl2.6H2O and KCl, on the bioconversion of crude glycerol to H2 was studied by PlacketteBurman design. Table 1 shows the experimental design and Table 2 presents the details of the linear model constructed using the PlacketteBurman design. It could be noted from Table 1 that the maximal response was obtained at Run 5, with a H2 yield of 0.72 ± 0.04 mol-H2/molglycerolconsumed. ANOVA results confirmed that the model was significant with an F-value of 73.08 and there could be only 1.36% chance that the F-value could be due to noise. The signal to noise ratio for the model was calculated to be 27.46. The R2 value of 0.99 and the predicted R2 value of 0.91 indicate a reasonable agreement and indicate a reliable model. The ANOVA results also show that all the 5 variables tested had an effect on the response. As indicated by the model coefficients, NH4Cl, K2HPO4 and KH2PO4 displayed a positive effect on H2 yield, whereas MgCl2.6H2O and KCl had a negative effect for the response. In the view of the results from PlacketteBurman; NH4Cl, K2HPO4 and KH2PO4 were significant model terms that asserted a positive effect on H2 production from crude glycerol. Therefore, these media components were selected and MgCl2.6H2O and KCl were kept at low concentrations (1 g/L) in subsequent experiments.
Table 7 Central composite face-centered design experimental design with the selected variables along with their response, YH2 (mol-H2/mol-glycerolconsumed). Run
1 2 3 4 5 6 7 8 9 10 11 12 13
NH4Cl (g/L)
The steepest ascent optimization step was used to determine the advantageous adjustment direction of the selected variables [28]. Based on the linear model (Eq. (1)) obtained from the PlacketteBurman design. The concentrations of NH4Cl, K2HPO4 and KH2PO4 were increased in five steps as shown in Table 3. The results indicated that Run 2 (Table 3) gave the maximum H2 yield of 1.04 ± 0.1 mol-H2/mol-glycerolconsumed and thus was chosen as the center point for the next optimization step.
KH2PO4 (g/L)
X1
Code X1
X2
Code X2
2.95 2.95 5.15 5.15 2.95 5.15 4.05 4.05 4.05 4.05 4.05 4.05 4.05
1 1 þ1 þ1 1 þ1 0 0 0 0 0 0 0
1.52 3.52 1.52 3.52 2.52 2.52 1.52 3.52 2.52 2.52 2.52 2.52 2.52
1 þ1 1 þ1 0 0 1 þ1 0 0 0 0 0
YH2
1.09 0.93 1.29 1.18 1.15 1.28 1.31 1.19 1.41 1.41 1.40 1.39 1.35
± ± ± ± ± ± ± ± ± ± ± ± ±
0.08 0.04 0.11 0.10 0.12 0.27 0.06 0.06 0.03 0.13 0.00 0.02 0.01
BoxeBehnken design further explored the optimization of the three selected variables for maximization of H2 production. Table 4 represents the experimental design for the three variables with 15 runs including 3 runs at the center point, along with their response. Table 5 presents the details of the second order polynomial model constructed using the BoxeBehnken design. A second order polynomial model for coded factors from the multiple regression analysis on the experimental data was obtained to be as:
Y ¼ 1:05 þ 0:22X1 þ 0:046X2 þ 0:030X3 þ 0:044X1 X2 þ 0:067X1 X3 0:048X2 X3 0:063X12 0:21X22 þ 0:031X32 (3) where, Y is the predicted H2 yield; X1 , X2 , and X3 are the coded values for NH4Cl, K2HPO4 and KH2PO4 respectively. ANOVA results indicate that the model is significant (Fvalue ¼ 347.6) with an insignificant lack of fit (p > 0.1) of 0.59. The R2 value of 0.99 was in good agreement with the predicted R2 value of 0.98, with a signal to noise ratio of 67.5. All the variables showed both individual and interacting significant effect on the response. The model was significant, but the response surface graphs indicated that the optimal region was beyond the design range (data not shown). This was due to the concentration range chosen in the steepest ascent. From Table 2 in PlacketteBurman design it could be noted that among the three selected factors, the significance (pvalue) followed the order NH4Cl < KH2PO4 < K2HPO4. Increasing amounts of phosphates in the growth medium can contribute to a reduction in microbial growth and H2 production due to high cytoplasmic osmotic pressure [29]. Thus, a decline in H2 yields in the steepest ascent test maybe due to high concentrations of either K2HPO4 or KH2PO4. In order to investigate the optimal concentration ranges that would include the maximal response, the adjustment direction was next determined by ridge analysis. 3.4. Ridge analysis Ridge analysis was chosen as the next step to determine the design range that would include the maximal response. This method uses the second order polynomial model from BoxeBehnken design (Eq. (3)). The approach used in here was to design the experimental points to start in the center of the previous BoxeBehnken design and to end at the point in which the predicted
R. Mangayil et al. / Renewable Energy 75 (2015) 583e589
response exceeded the theoretical maximum for H2 production from glycerol. The experimental design and predicted and observed H2 yields were as shown in Table 6. Trial runs designed based on Eq. (3) indicated that the concentration variations were significant for NH4Cl and KH2PO4. The concentration of K2HPO4 remained around 1.60 g/L (Runs 1e4). Results indicated that H2 yield increased from Runs 1 to 3, after which a decline in the response was observed. As observed earlier, in PlacketteBurman design, K2HPO4 showed the least significance among the variables that positively influenced the H2 yield (Table 2). From the data obtained from the ridge analysis experiments, it was decided that the concentration of K2HPO4 can be kept at 1.60 g/L in subsequent optimization steps. The data also point towards further optimization of NH4Cl and KH2PO4. 3.5. Central composite face-centered design NH4Cl and KH2PO4 were subjected to final optimization step by CCD, keeping the other media component concentrations as 1.6 g/L of K2HPO4, 1.0 g/L of MgCl2.6H2O, 1.0 g/L of KCl, 1.0 g/L of Na-acetate.3H2O and 2.0 g/L of tryptone. The CCD consisted of 13 runs including 5 runs at the center point as shown in Table 7. Table 8 presents the details of the second order polynomial model. The H2 yield ranged from 0.93 ± 0.04 mol-H2/mol-glycerolconsumed (Run 2) to 1.41 ± 0.03 mol-H2/mol-glycerolconsumed (Run 9). Multiple regression analysis was applied to the experimental data and a second order polynomial model was obtained:
Y ¼ 1:39 þ 0:099X1 0:065X2 þ 0:016X1 X2 0:15X12 0:12X22 (4) where, Y is the predicted H2 yield; X1 and X2 are the coded values of NH4Cl and KH2PO4 respectively. ANOVA results indicated a high determination coefficient (R2 ¼ 0.97) and a high adjusted determination coefficient (Adj R2 ¼ 0.96) implying a highly significant model. The model was significant (F-value ¼ 64.3) with an insignificant lack of fit (0.30). The results suggest that the two variables, individually and in the square of terms, were significant to the response. Threedimensional response surface and two-dimensional contour graphs were plotted using the Design-Expert 8.0.7.1 (Stat-Ease Inc.,
587
MN, USA) demo version software in order to understand the optimal levels of the variables tested. Fig. 1(A and B) illustrates the contour curve and corresponding 3-dimensional response surface plots on X1 (NH4Cl) and X2 (KH2PO4). Elliptical contours are obtained when there is a significant interaction between the variables tested [30]. The ANOVA suggested that the two variables were interdependent and the interaction effect between NH4Cl and KH2PO4 ðX1 X2 Þ were insignificant (p-value ¼ 0.29). H2 yield increased as the concentration of NH4Cl was increased from 2.95 to 4.05 g/L and then decreased with a further increase in the concentration level. Optimal levels of N2 in growth media are beneficial for microbial growth and H2 production and vary with the inoculum type [15,31,32]. Similar effect was observed for KH2PO4 (1.52e2.52 g/L), indicating an optimal buffering condition in the medium [29,31]. The optimal concentrations of NH4Cl and KH2PO4, obtained from partial differentiation of Eq. (4), were 4.40 g/L and 2.27 g/L respectively. At these conditions, the H2 yield was predicted to be 1.41 mol-H2/mol-glycerolconsumed (Fig. 1). 3.6. Validation of the model For confirmation experiments, an optimized HM100 medium (NH4Cl, 4.40 g/L; K2HPO4, 1.6 g/L; KH2PO4, 2.27 g/L; MgCl2.6H2O, 1.0 g/L; KCl, 1.0 g/L; Na-acetate.3H2O, 1.0 g/L; tryptone, 2.0 g/L; Nadithionite, 0.05 g/L and resazurin, 0.002 g/L) containing crude glycerol was prepared. The initial pH was adjusted to 6.5 and the cultivation temperature was 40 C. The batch experiments in serum bottles were conducted in triplicate experiments with a substrate blank and the data obtained were averaged. Acetate, ethanol and butyrate were the major metabolic end products produced. The H2 yield under the optimized conditions was found to be 1.42 ± 0.15 mol-H2/mol-glycerolconsumed. The H2 yield from the validation experiment indicated an excellent correlation between the predicted and experimental values. The verification experiment revealed only 0.6% difference between the predicted (1.41 mol-H2/mol-glycerolconsumed) and experimental (1.42 ± 0.15 mol-H2/mol-glycerolconsumed) values. In our previous study, H2 yield of 1.1 ± 0.1 mol-H2/mol-glycerolconsumed was obtained with modified HM100 medium (NH4Cl 1.0 g/L, K2HPO4 0.3 g/ L, KH2PO4 0.3 g/L, MgCl2.6H2O 2.0 g/L, KCl 4.0 g/L, Na-acetate.3H2O 1.0 g/L, tryptone 1.0 g/L, cysteineeHCl 0.5 g/L and resazurin 0.002 g/
Fig. 1. (A) Two dimensional contour plot and (B) three dimensional response surface plot showing the effects NH4Cl and KH2PO4 on YH2.
588
R. Mangayil et al. / Renewable Energy 75 (2015) 583e589
Table 9 Overview of recent studies on mesophilic biohydrogen production from crude glycerol by pure strains and mixed cultures. Inocula
Statistical methods P.B.
Pure cultures K. pneumoniae E. aerogenes E. aerogenes E. aerogenes KKU-S1 Halanaerobium saccharolyticum subsp. saccharolyticum Mixed cultures isolated from Wheat soil Activated sludge Wastewater Anaerobic digested sludge Activated sludge
S.A.
B.B.
R.A.
C.C.D.
Y.E. (g/L)
T.E. (g/L)
YH2
Ref.
Yes No No Yes No
Yes No No Yes Yes
0.53 0.84 0.85 0.12 3.0
[2] [10] [33] [34] [35]
No No No Yes No
Yes No No Yes No
0.31 0.90 0.96 0.41 1.42
[8] [11] [12] [36] cur.st
F.F.
X X X X NA
NA
NA
NA
X NA
NA
NA NA X NA X
NA NA X NA X
NA NA X NA X
NA NA
NA NA
NA NA
NA X
NA X
NA
P.B. e PlacketteBurman design; S.A. e Path of steepest accent; B.B. e BoxeBehnken design; R.A. e Ridge analysis; C.C:D. e Central composite design; Y.E. e Yeast extract; T.E. e Trace elements; Ref. e Reference and NA e Not applicable (Statistical optimizations not conducted).
L) in batch fermentation experiments [3]. The maximum H2 yield of 1.42 ± 0.15 mol-H2/mol-glycerolconsumed achieved after medium optimization by RSM was 29% higher than our previous reported yield [3]. This confirms RSM analysis as a powerful technique to optimize the bioconversion of crude glycerol using mixed microbial inoculum.
was the sole phosphate source, but the H2 production remained similar when used along with KH2PO4 (PH/KH 0.67). Similar to the observations from the current study and by Jitrwung and Yargeau (2011) suggests that adequate phosphate levels in culture medium are crucial for cell growth and optimal H2 production. 4. Conclusions
3.7. Comparison of previously reported statistically optimized media Increasing volume of residual glycerol as a waste from biodiesel production plants has been widely exploited as an alternative source for the production of industrial important high value compounds such as H2, 1,3-propanediol, ethanol and other organic acids. Table 9 summarizes the statistical optimization methods and results from recent studies on biohydrogen production from crude glycerol under mesophilic conditions. However, it is worth to note that a direct comparison is impossible due to different operational parameters, inocula, crude glycerol contents, substrate concentration and use of additional media supplements. Only very few studies [12] have focused on statistical optimization for improved mesophilic biohydrogen production from crude glycerol by mixed microbial consortia. However, for pure strains more studies have been conducted. One of the first studies dealing with statistical media optimization for improved biohydrogen production from crude glycerol was reported by Liu and Fang (2007) using Klebsiella pneumoniae DSM 2026 [2]. The authors reported a H2 yield of 0.4 and 0.53 mol-H2/mol-glycerol from 20.4 g/L glycerol with an optimized medium containing 5.7 g/L KCl, 13.8 g/L NH4Cl, 1.5 g/L CaCl2 supplemented with 3 g/L yeast extract from batch and 5 L stir tank bioreactor, respectively. Jitrwung and Yargeau (2011) have investigated on optimizing media components for improved biohydrogen production from crude glycerol by E. aerogenes [33]. The optimization was performed using BoxeBehnken design and response surface methodology on medium components (NH4NO3, FeSO4, and Na2HPO4) and O2, reporting a yield of 0.85 mol H2/mol glycerol from 18 g/L of crude glycerol. Recently, Jitrwung et al. (2013) have conducted optimization of salts (MgSO4, Na2EDTA, CaCl2, Na2HPO4, and KH2PO4) present in the culture media aiming to shorten the microbial growth lag phase [10]. Authors reported a positive effect of Mgþ ions (0.2 g/L MgSO4) on microbial cell growth and observed no effect on hydrogen production. Here the higher media concentration of MgCl2 (2 g/L) had a negative effect on hydrogen production and improved H2 production was observed when Mgþ ion concentration (1 g/L MgCl2) was kept at low levels. A short lag phase was observed when Na2HPO4
The current study was successful in applying the DoE to enriched microbial consortia in order to screen and identify the optimal concentrations of media components. The PlacketteBurman identified NH4Cl and KH2PO4 as variables that positively influenced the H2 production. A H2 yield of 1.41 mol-H2/mol-glycerolconsumed was predicted by response surface methodology with CCD using an optimized media (NH4Cl, 4.40 g/L; K2HPO4, 1.6 g/L; KH2PO4, 2.27 g/L; MgCl2.6H2O, 1.0 g/L; KCl, 1.0 g/L; Na-acetate.3H2O, 1.0 g/L and tryptone, 2.0 g/L). High correlation was observed between predicted and experimental data with a variance of 0.58%. In the present study, under optimized conditions, a 29% improvement in H2 yield is reported by a functional microbial community compared to previously obtained result [3]. Acknowledgments The research was funded by the Maj and Tor Nessling Foundation (Project no: 2012356) and The Academy of Finland (Project no's. 126974 and 139830). References [1] Yang F, Hanna M, Sun R. Value-added uses for crude glycerol e a byproduct of biodiesel production. Biotechnol Biofuels 2012;5:13. [2] Liu F, Fang B. Optimization of bio-hydrogen production from biodiesel wastes by Klebsiella pneumonia. Biotechnol J 2007;2:374e80. [3] Mangayil R, Karp M, Santala V. Bioconversion of crude glycerol from biodiesel production to hydrogen. Int J Hydrogen Energy 2012;37:12198e204. [4] Hiremath A, Kannabiran M, Rangaswamy V. 1,3-Propanediol production from crude glycerol from jatropha biodiesel process. New Biotechnol 2011;28: 19e23. [5] Oh BR, Seo JW, Heo SY, Hong WK, Luo LH, Joe MH, et al. Efficient production of ethanol from crude glycerol by a Klebsiella pneumoniae mutant strain. Bioresour Technol 2011;102:3918e22. [6] Almeida JRM, F avaro LCL, Quirino BF. Biodiesel biorefinery: opportunities and challenges for microbial production of fuels and chemicals from glycerol waste. Biotechnol Biofuels 2012;5:48. [7] Sarma SJ, Brar SK, Sydney EB, Le Bihan Y, Buelna G, Soccol CR. Microbial hydrogen production by bioconversion of crude glycerol: a review. Int J Hydrogen Energy 2012;37:6473e90. [8] Selembo PA, Perez JM, Lloyd WA, Logan BE. Enhanced hydrogen and 1,3propanediol production from glycerol by fermentation using mixed cultures. Biotechnol Bioeng 2009;104:1098e106.
R. Mangayil et al. / Renewable Energy 75 (2015) 583e589 [9] Ngo TA, Kim MS, Sim SJ. High-yield biohydrogen production from biodiesel manufacturing waste by Thermotoga neapolitana. Int J Hydrogen Energy 2011;36:5836e42. [10] Jitrwung R, Verrett J, Yargeau V. Optimization of selected salts concentration for improved biohydrogen production from biodiesel-based glycerol using Enterobacter aerogenes. Renew Energy 2013;50:222e6. [11] Varrone C, Rosa S, Fiocchetti F, Giussani B, Izzo G, Massini G, et al. Enrichment of activated sludge for enhanced hydrogen production from crude glycerol. Int J Hydrogen Energy 2013;38:1319e31. [12] Varrone C, Giussani B, Izzo G, Massini G, Marone A, Signorini A, et al. Statistical optimization of biohydrogen and ethanol production from crude glycerol by microbial mixed culture. Int J Hydrogen Energy 2012;37:16479e88. [13] Saraphirom P, Reungsang A. Optimization of biohydrogen production from sweet sorghum syrup using statistical methods. Int J Hydrogen Energy 2010;35:13435e44. [14] Sittijunda S, Reungsang A. Media optimization for biohydrogen production from waste glycerol by anaerobic thermophilic mixed cultures. Int J Hydrogen Energy 2012;37:15473e82. [15] Wang J, Wan W. Factors influencing fermentative hydrogen production: a review. Int J Hydrogen Energy 2009;34:799e811. [16] Lin CY, Lay CH. Effects of carbonate and phosphate concentrations on hydrogen production using anaerobic sewage sludge microflora. Int J Hydrogen Energy 2004;29:275e81. [17] Wang XJ, Ren NQ, Sheng Xiang W, QianGuo W. Influence of gaseous endproducts inhibition and nutrient limitations on the growth and hydrogen production by hydrogen-producing fermentative bacterial B49. Int J Hydrogen Energy 2007;32:748e54. [18] Plackett RL, Burman JP. The design of optimum multifactorial experiments. Biometrika 1946;33:305e25. [19] Wang J, Wan W. Experimental design methods for fermentative hydrogen production: a review. Int J Hydrogen Energy 2009;34:235e44. [20] Box GEP, Behnkin EW. Some new three level designs for the study of quantitative variables. Technometrics 1960;2:455e75. [21] Hoerl AE. Optimum solution of many variables equations. Chem Eng Prog 1959;55:69e78. [22] Draper NR. Ridge analysis of response surfaces. Technometrics 1963;5: 469e79.
589
[23] Russell VL. Response-surface methods in R, using rsm. J Stat Softw 2009;32: 1e17. [24] Khuri AI, Mukhopadhyay S. Response surface methodology. WIREs Comput Stat 2010;2:128e49. [25] Ihaka R, Gentleman RR. a language for data analysis and graphics. J Comp Graph Stat 1996;5:299e314. [26] Mangayil R, Santala V, Karp M. Fermentative hydrogen production from different sugars by Citrobacter sp. CMC-1 in batch culture. Int J Hydrogen Energy 2011;36:15187e94. € m C, Wold S. Design of [27] Eriksson L, Johansson E, Kettaneh-Wold N, Wikstro experiments: principles and applications. 3rd ed. Umetrics Academy; 2008. [28] Long C, Cui J, Liu Z, Liu Y, Long M, Hu Z. Statistical optimization of fermentative hydrogen production from xylose by newly isolated Enterobacter sp. CN1. Int J Hydrogen Energy 2010;35:6657e64. [29] Pan CM, Fan YT, Xing Y, Hou HW, Zhang ML. Statistical optimization of process parameters on biohydrogen production from glucose by Clostridium sp. Fanp2. Bioresour Technol 2008;99:3146e54. [30] Xu H, Sun LP, Shi YZ, Wu YH, Zhang B, Zhao DQ. Optimization of cultivation conditions for extracellular polysaccharide and mycelium biomass by Morchella esculenta As51620. Biochem Eng J 2008;39:66e73. [31] Bisaillon A, Turcot J, Hallenbeck PC. The effect of nutrient limitation on hydrogen production by batch cultures of Escherichia coli. Int J Hydrogen Energy 2006;31:1504e8. [32] Salerno MB, Park W, Zuo Y, Logan BE. Inhibition of biohydrogen production by ammonia. Water Res 2006;40:1167e72. [33] Jitrwung R, Yargeau V. Optimization of media composition for the production of biohydrogen from waste glycerol. Int J Hydrogen Energy 2011;36:9602e11. [34] Reungsang A, Sittijunda S, Angelidaki I. Simultaneous production of hydrogen and ethanol from waste glycerol by Enterobacter aerogenes KKU-S1. Int J Hydrogen Energy 2013;38:1813e25. € A, Santala V, Karp M. Non-sterile process for biohydrogen and 1,3[35] Kivisto propanediol production from raw glycerol. Int J Hydrogen Energy 2013;38: 11749e55. [36] Seifert K, Waligorska M, Wojtowski M, Laniecki M. Hydrogen generation from glycerol in batch fermentation process. Int J Hydrogen Energy 2009;34: 3671e8.
Contents lists available at ScienceDirect
Renewable Energy journal homepage: www.elsevier.com/locate/renene
Improved bioconversion of crude glycerol to hydrogen by statistical optimization of media components Rahul Mangayil a, *, Tommi Aho a, b, Matti Karp a, Ville Santala a a b
Department of Chemistry and Bioengineering, Tampere University of Technology, Tampere, Finland Department of Signal Processing, Tampere University of Technology, Tampere, Finland
a r t i c l e i n f o
a b s t r a c t
Article history: Received 12 December 2013 Accepted 18 October 2014 Available online
Bioconversion of crude glycerol to hydrogen has gained importance as it addresses both sustainable energy production and waste disposal issues. Until recently, statistical optimizations of crude glycerol bioconversion to hydrogen have been greatly focused on pure strains. In this study, biohydrogen production from crude glycerol by an enriched microbial culture (predominated with Clostridium species) was improved by statistical optimization of media components. PlacketteBurman design identified MgCl2.6H2O and KCl with negative effect on hydrogen production and selected NH4Cl, K2HPO4 and KH2PO4 as significant variables. BoxeBehnken design indicated the optimal region beyond design area and studies were continued by ridge analysis. Central composite face centered design envisaged a maximal hydrogen yield of 1.41 mol-H2/mol-glycerolconsumed at concentrations 4.40 g/L and 2.27 g/L for NH4Cl and KH2PO4 respectively. Confirmation experiment with the optimized media (NH4Cl, 4.40 g/L; K2HPO4, 1.6 g/L; KH2PO4, 2.27 g/L; MgCl2.6H2O, 1.0 g/L; KCl, 1.0 g/L; Na-acetate.3H2O, 1.0 g/L and tryptone, 2.0 g/L) revealed an excellent correlation between predicted and experimental hydrogen yield. Optimization of media components by design of experiments enhanced hydrogen yield by 29%. © 2014 Elsevier Ltd. All rights reserved.
Keywords: Biohydrogen Crude glycerol Optimization Response surface methodology
1. Introduction Concern over increasing environmental pollution and the need for a sustainable energy source have led to an increased global biodiesel production for its use as a diesel substitute. This increased biodiesel production has resulted in severe waste disposal crisis due to the production of proportionally equivalent amounts of crude glycerol as the by-product [1]. Since crude glycerol is an excellent carbon source, bioconversion of crude glycerol has been an interesting area of study. Moreover, the process also offers a sustainable method for the disposal of biodiesel waste. Bioconversion of crude glycerol to hydrogen (H2) [2,3], 1,3-propanediol [4], ethanol [5], and lactic acid [6] have been extensively studied. Fermentative H2 production from crude glycerol has gained much importance due to high H2 content and is a promising substrate for sustainable energy production with high energy content [7]. Selembo et al. [8] studied H2 production, in batch type fermentation, from crude glycerol using mixed inoculum, reporting
* Corresponding author. Department of Chemistry and Bioengineering, Tampere University of Technology, P.O. Box 541, 33101 Tampere, Finland. Tel.: þ358 3 3115 2052; fax: þ358 3 3115 2869. E-mail address: rahul.mangayil@tut.fi (R. Mangayil). http://dx.doi.org/10.1016/j.renene.2014.10.051 0960-1481/© 2014 Elsevier Ltd. All rights reserved.
a yield of 0.31 mol-H2/mol-glycerol. Efficient bioconversion of crude glycerol to H2 was reported by Ngo et al. [9], wherein pretreated crude glycerol was used as the substrate, assisted with N2 sparging and pH control, using Thermotoga neapolitana, reporting a yield of 2.73 mol-H2/mol-glycerolconsumed. In a recent publication, Jitrwung et al. [10] investigated the effect of optimizing media optimization on bioconversion of crude glycerol to H2 by Enterobacter aerogenes, reporting a yield of 0.84 mol-H2/mol-glycerol. Eco-biotechnological application for improved bioconversion of crude glycerol has been recently studied. Comparison of crude glycerol utilization from different bio-diesel plants and its bioconversion to H2 by a functional mesophilic microbial consortium enriched from activated sludge was investigated by Varrone et al. [11]. The authors reported a H2 yield of 0.90 mol-H2/molglycerol from minimal media supplement with 15 g/L crude glycerol and a 97% degradation efficiency of the sole carbon source. Implementing statistical optimization for improved biohydrogen production from crude glycerol by mixed microbial inoculum has gained importance. Varrone et al. [12] have reported on improved H2 production from crude glycerol using statistical approach in optimizing the culture conditions. They investigated the biohydrogen production efficiency of mesophilic mixed microbial culture on crude glycerol as the carbon source, reporting a yield of 0.96 mol-H2/mol-glycerol. Apart from culture conditions such as
584
R. Mangayil et al. / Renewable Energy 75 (2015) 583e589
temperature, pH, inoculums size and optimal substrate concentration, microorganisms require appropriate concentrations of media components for optimal metabolism during fermentation process [13,14]. N2 is an important media component, assisting in the synthesis of proteins, nucleic acids and enzymes, significant for the microbial growth and hydrogenase enzyme activity [14,15]. Optimal phosphate concentration positively affects the H2 production due to its buffering capacity. High phosphate and potassium ion concentrations are reported to cause an increased cytoplasmic osmotic pressure, thus negatively affecting the microbial growth [15,16]. Magnesium, an important component for cell wall and cell membrane synthesis and an enzyme cofactor, also plays a pivotal role in microbial growth and metabolism [17]. Our previous study focused on enriching H2 producing mixed microbial culture. This study aims in improving the earlier reported H2 yield by optimizing the current media composition, without altering the original media components, using design of experiments (DoE) and response surface methodology (RSM). DoE and RSM are efficient approaches for such optimization tasks, as shown by previous studies [12e14]. In this study, PlacketteBurman design was used to screen the significant media components [18]. The path of steepest ascent [19], BoxeBehnken [20], ridge analysis [21e24] and Central Composite Face Centered Design (CCD) [25] models were used to optimize the concentrations of screened media components. Finally, the validation of the model was performed at the predicted optimal concentrations. 2. Materials and methods 2.1. Glycerol source and culture conditions Industrial glycerol was kindly provided by Savon Siemen Oy (Iisalmi, Finland). The crude glycerol contained 45% (v/v) glycerol and 30% (v/v) methanol with an alkaline pH (~12). The inoculum used for this study was an enriched microbial consortium (mainly comprised of Clostridium sporogenes strain CL3 (Accession no: JF836014.1), Clostridium subterminale isolate DSM 758 (Accession no: EU857637) and uncultured bacterium clone (Accession no: FJ512181.1)) [3] from activated sludge enriched in modified HM100 medium (NH4Cl, 1.0 g/L; K2HPO4, 0.3 g/L; KH2PO4, 0.3 g/L; MgCl2.6H2O, 2.0 g/L; KCl, 4.0 g/L; Na-acetate.3H2O, 1.0 g/L; tryptone, 2.0 g/L; Na-dithionite, 0.5 g/L and resazurin, 0.002 g/L) amended with crude glycerol [3]. The pre-inoculum was prepared by inoculating 10% of the enriched microbial community in 120 ml serum bottles with working volume of 50 ml HM100 medium containing pure glycerol (5 g/L). The pre-inoculum was grown at 150 rpm with an initial pH of 6.5 and cultivation temperature at 40 C. For optimization studies, ammonium chloride (NH4Cl), dipotassium phosphate (K2HPO4), potassium di-hydrogen phosphate (KH2PO4), magnesium chloride hexa-hydrate (MgCl2.6H2O) and potassium chloride (KCl) were chosen. The media was prepared in similar way, varying the selected variables and keeping the remaining media components in concentrations as of the original media composition. The DoE studies were conducted in batch type fermentation with 120 ml serum bottles with a working volume of 50 ml sterile anoxic crude glycerol (1 g/L) amended enrichment medium. The investigations were performed as triplicate experiments at 40 C and 150 rpm with a growth period of 72 h. 2.2. Analytical methods Crude glycerol, volatile fatty acids and alcohols were analyzed by high performance liquid chromatography (HPLC; LC-20AD, Shimadzu, Japan). The HPLC was equipped with a 300 mm 8 mm Shodex SUGAR (SH1011) column and a refractive
index detector (RID-10A). The HPLC samples were prepared as described previously [18]. The proportions of H2 gas were measured using gas chromatograph (GC-2014, Shimadzu GC), fitted with a thermal conductivity detector and a 2 m 2 mm PORAPAK column. N2 was used as the carrier gas with a flow rate of 20 ml/ min. The temperatures of column, detector and oven were maintained at 80 C, 110 C and 80 C respectively. Measurements for each sample were repeated twice and averaged. Substrate blank (i.e. cultivation without the crude glycerol) was included in all experiment sets to deduct H2 produced from tryptone in the growth media. The H2 yield values were calculated as described previously [26]. The presence of methanol in crude glycerol and its utilization by the microbial community was tested. From growth-curve test and end-metabolite distribution it was observed that the functional community did not utilize methanol (data not shown). Hence, methanol was excluded from the H2 yield calculations. 2.3. Optimization procedure 2.3.1. PlacketteBurman design PlacketteBurman design [18] was chosen to screen and identify variables that had significant influence on the bioconversion of crude glycerol to H2. PlacketteBurman design is based on the first order polynomial model:
Y ¼ b0 þ
X
bi Xi
(1)
where Y is the response, b0 is the model intercept and bi is the linear coefficient, and Xi is the level of independent variable. Five media components, NH4Cl, K2HPO4, KH2PO4, MgCl2.6H2O and KCl, were selected to investigate the trivial components influencing H2 production. Based on PlacketteBurman design, each factor was prepared in two levels: 1 for low level and þ1 for high level (Table 1). The main effect of each variable to the response was then calculated as the difference between the averages of the high level (þ1) and the low level (1) response measurements of that variable. In this study, five assigned variables were screened in eight experimental designs. The level of each factor used in the experimental design along with the response and details of the linear model constructed using the PlacketteBurman design is reported in Tables 1 and 2 respectively. The analysis of variance (ANOVA) was performed to select the variable with significant effect (p value > 0.05) on the response. The results presented were average H2 produced from triplicate experiments with standard deviations. The Design-Expert (version 8.0.7.1, Stat-Ease Inc., MN, USA) was used for statistical analysis. 2.3.2. Path of steepest ascent The identified significant variables from PlacketteBurman design were subjected to steepest ascent approach. This optimization step helps to move the experimental region from the design center towards the direction of an optimal response. The path of steepest ascent is based on the first order polynomial model (Eq. (1)) obtained from the PlacketteBurman design and provides information on the test range of selected variables for further optimization steps [19]. Table 3 reports the concentration range of the selected variables to move the design area towards optimal response. The average yield, with standard deviations, was calculated from triplicate experiments for each design points. 2.3.3. BoxeBehnken design For further optimization of the selected media components for enhanced H2 production, a three-variable BoxeBehnken design
R. Mangayil et al. / Renewable Energy 75 (2015) 583e589
585
Table 1 PlacketteBurman experimental design with five independent variables along with their response, YH2 (mol-H2/mol-glycerolconsumed). NH4Cl
Run
1 2 3 4 5 6 7 8
K2HPO4
KH2PO4
MgCl2.6H2O
YH2
Code X1
X2 (g/L)
Code X2
X3 (g/L)
Code X3
X4 (g/L)
Code X4
X4 (g/L)
Code X4
2.00 0.50 2.00 0.50 2.00 0.50 2.00 0.50
þ1 1 þ1 1 þ1 1 þ1 1
1.00 1.00 0.10 0.10 1.00 1.00 0.10 0.10
þ1 þ1 1 1 þ1 þ1 1 1
1.00 0.10 0.10 1.00 1.00 0.10 0.10 1.00
þ1 1 1 þ1 þ1 1 1 þ1
4.00 4.00 4.00 4.00 1.00 1.00 1.00 1.00
þ1 þ1 þ1 þ1 1 1 1 1
8.00 1.00 8.00 1.00 1.00 8.00 1.00 8.00
þ1 1 þ1 1 1 þ1 1 þ1
[20] with three replicates at the center point was performed. The design and the levels of variable are summarized in Table 4. The experimental design consisted of 15 runs including three center points. Each experimental design run was performed in triplicates and the data produced were averaged with standard deviations. The optimal point for H2 production was predicted using a second order polynomial function that was fitted to correlate the relationship between the predicted response and the variables:
Y ¼ b0 þ
KCl
X1 (g/L)
X
bi Xi þ
X
bii Xi2 þ
X
bij Xi Xj
0.37 0.32 0.15 0.32 0.72 0.35 0.57 0.35
± ± ± ± ± ± ± ±
0.14 0.04 0.02 0.04 0.04 0.04 0.07 0.03
that the predicted response for H2 production was slightly over the theoretical maximum at the end of the search path. Table 6 details the ridge analysis design for the selected variables along with distance from the center point and predicted response (yhat). All the trial runs were performed in triplicates, along with a substrate blank and the results were averaged with standard deviations. The design points in the ridge analysis were determined using RSM package [Extra1] in R software environment for statistical computing and graphics [27].
(2)
where Y is the predicted response; b0 , a constant; bi , the linear coefficients; bii , the squared coefficients; and bij , the cross-product coefficients. The details of the second order polynomial model constructed using the BoxeBehnken design are provided in Table 5. The DesignExpert 8.0.7.1 (Stat-Ease Inc., MN, USA) demo version was used for the regression and graphical analysis of the experimental data. ANOVA was performed in order to evaluate the statistical significance of the model.
2.3.4. Ridge analysis Ridge analysis [21,22] was applied in determining the search path on the second order polynomial surface that was constructed with the help of BoxeBehnken design. Ridge analysis, non-linear analog to the steepest ascent method, uses the fitted second order polynomial function (Eq. (2)) for optimizing the response. This method iteratively moves specified step sizes in the search space and, in each step, allows to identify the point at which the response is maximal [23,24]. In order to span the search path long enough, i.e. to include the maximal response, ridge analysis was designed so Table 2 Details of the linear model constructed using the PlacketteBurman design. Code
Variable
Low level (g/L)
High level (g/L)
Coefficient
F Value
p-Value (Prob > F)
X1 X2 X3 X4 X5
NH4Cl K2HPO4 KH2PO4 MgCl2.6H2O KCl
0.50 0.10 0.10 1.00 1.00
2.00 1.00 1.00 4.00 8.00
0.059 0.047 0.048 0.10 0.088
48.74 30.32 31.76 148.39 106.19
0.0199 0.0314 0.0301 0.0067 0.0093
Table 3 Steepest ascent experimental design to move the experimental area towards optimal response, YH2 (mol-H2/mol-glycerolconsumed). Run
NH4Cl (g/L)
K2HPO4 (g/L)
KH2PO4 (g/L)
YH2
1 2 3 4 5 6
1.25 2 2.8 3.6 4.4 5.2
0.6 1.6 2.6 3.7 4.7 5.6
0.6 1.6 2.7 3.8 4.8 5.9
0.43 1.04 0.87 0.49 0.47 0.44
± ± ± ± ± ±
0.10 0.10 0.32 0.18 0.10 0.06
2.3.5. Central composite face-centered design Based on the experimental results from ridge analysis runs, NH4Cl and KH2PO4 were selected for CCD, keeping the other media components at optimized concentrations. For the design model, three different coded levels for each factor were used: 1 (low), 0 (center) and þ1 (high). The designs of all 13 experiments are listed in Table 7. Table 8 provides the details of the second order polynomial model constructed using the CCD. Experiments were performed in triplicates and the data produced were the mean value of the repeats with standard deviations. Inclusion of a substrate blank assisted in calculating the H2 produced exclusively from crude glycerol. The Design-Expert (version 8.0.7.1, Stat-Ease Inc., MN, USA) was used for building the model and statistical analysis for the experimental data using ANOVA. 3. Results and discussion 3.1. Screening of important media components for H2 production from crude glycerol Nutrients play a pivot role in microbial cell growth and bioconversion. The relative significance of five media components, Table 4 BoxeBehnken experimental design with three independent variables along with their response, YH2 (mol-H2/mol-glycerolconsumed). Run
NH4Cl X1 (g/L)
Code X1
X2 (g/L)
Code X2
X3 (g/L)
Code X3
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
0.50 3.50 0.50 3.50 0.50 3.50 0.50 3.50 2.00 2.00 2.00 2.00 2.00 2.00 2.00
1 þ1 1 þ1 1 þ1 1 þ1 0 0 0 0 0 0 0
1.10 1.10 2.10 2.10 1.60 1.60 1.60 1.60 1.10 2.10 1.10 2.10 1.60 1.60 1.60
1 1 þ1 þ1 0 0 0 0 1 þ1 1 þ1 0 0 0
1.60 1.60 1.60 1.60 1.10 1.10 2.10 2.10 1.10 1.10 2.10 2.10 1.60 1.60 1.60
0 0 0 0 1 1 þ1 þ1 1 1 þ1 þ1 0 0 0
K2HPO4
KH2PO4
YH2
0.56 0.90 0.57 1.09 0.83 1.14 0.77 1.35 0.76 0.94 0.91 0.89 1.04 1.05 1.07
± ± ± ± ± ± ± ± ± ± ± ± ± ± ±
0.05 0.00 0.02 0.06 0.02 0.06 0.04 0.03 0.02 0.01 0.03 0.07 0.02 0.03 0.02
586
R. Mangayil et al. / Renewable Energy 75 (2015) 583e589
Table 5 Details of the second order polynomial model constructed using the BoxeBehnken design.
Table 8 Details of the second order polynomial model constructed using the Central composite face-centered design.
Code
Variable
Low level (g/L)
High level (g/L)
Coefficient
F Value
p-Value (Prob > F)
Code
Variable
Low level (g/L)
High level (g/L)
Coefficient
F Value
p-Value (Prob > F)
X1 X2 X3
NH4Cl K2HPO4 KH2PO4
0.50 1.10 1.10
3.50 2.10 2.10
0.22 0.030 0.046
1940.10 36.13 85.26
0.0001 0.0018 0.0003
X1 X2
NH4Cl KH2PO4
2.95 1.52
5.15 3.52
0.099 0.065
75.36 32.83
0.0001 0.0007
3.2. Path of steepest ascent Table 6 Ridge analysis design response, YH2 (mol-H2/mol-glycerolconsumed). a
b
Run
Distance
NH4Cl (g/L)
K2HPO4 (g/L)
KH2PO4 (g/L)
yhat
YH2
1 2 3 4 5 6
0.00 1.15 2.30 3.45 4.60 5.75
2.00 3.33 4.05 4.63 5.17 5.71
1.60 1.64 1.62 1.58 1.55 1.51
1.60 1.96 2.53 3.08 3.64 4.20
1.05 1.28 1.57 1.96 2.46 3.09
1.03 1.09 1.39 1.19 1.01 0.81
a b
± ± ± ± ± ±
0.03 0.03 0.10 0.09 0.03 0.04
Distance of each run from the center point. Predicted response.
3.3. BoxeBehnken design
NH4Cl, K2HPO4, KH2PO4, MgCl2.6H2O and KCl, on the bioconversion of crude glycerol to H2 was studied by PlacketteBurman design. Table 1 shows the experimental design and Table 2 presents the details of the linear model constructed using the PlacketteBurman design. It could be noted from Table 1 that the maximal response was obtained at Run 5, with a H2 yield of 0.72 ± 0.04 mol-H2/molglycerolconsumed. ANOVA results confirmed that the model was significant with an F-value of 73.08 and there could be only 1.36% chance that the F-value could be due to noise. The signal to noise ratio for the model was calculated to be 27.46. The R2 value of 0.99 and the predicted R2 value of 0.91 indicate a reasonable agreement and indicate a reliable model. The ANOVA results also show that all the 5 variables tested had an effect on the response. As indicated by the model coefficients, NH4Cl, K2HPO4 and KH2PO4 displayed a positive effect on H2 yield, whereas MgCl2.6H2O and KCl had a negative effect for the response. In the view of the results from PlacketteBurman; NH4Cl, K2HPO4 and KH2PO4 were significant model terms that asserted a positive effect on H2 production from crude glycerol. Therefore, these media components were selected and MgCl2.6H2O and KCl were kept at low concentrations (1 g/L) in subsequent experiments.
Table 7 Central composite face-centered design experimental design with the selected variables along with their response, YH2 (mol-H2/mol-glycerolconsumed). Run
1 2 3 4 5 6 7 8 9 10 11 12 13
NH4Cl (g/L)
The steepest ascent optimization step was used to determine the advantageous adjustment direction of the selected variables [28]. Based on the linear model (Eq. (1)) obtained from the PlacketteBurman design. The concentrations of NH4Cl, K2HPO4 and KH2PO4 were increased in five steps as shown in Table 3. The results indicated that Run 2 (Table 3) gave the maximum H2 yield of 1.04 ± 0.1 mol-H2/mol-glycerolconsumed and thus was chosen as the center point for the next optimization step.
KH2PO4 (g/L)
X1
Code X1
X2
Code X2
2.95 2.95 5.15 5.15 2.95 5.15 4.05 4.05 4.05 4.05 4.05 4.05 4.05
1 1 þ1 þ1 1 þ1 0 0 0 0 0 0 0
1.52 3.52 1.52 3.52 2.52 2.52 1.52 3.52 2.52 2.52 2.52 2.52 2.52
1 þ1 1 þ1 0 0 1 þ1 0 0 0 0 0
YH2
1.09 0.93 1.29 1.18 1.15 1.28 1.31 1.19 1.41 1.41 1.40 1.39 1.35
± ± ± ± ± ± ± ± ± ± ± ± ±
0.08 0.04 0.11 0.10 0.12 0.27 0.06 0.06 0.03 0.13 0.00 0.02 0.01
BoxeBehnken design further explored the optimization of the three selected variables for maximization of H2 production. Table 4 represents the experimental design for the three variables with 15 runs including 3 runs at the center point, along with their response. Table 5 presents the details of the second order polynomial model constructed using the BoxeBehnken design. A second order polynomial model for coded factors from the multiple regression analysis on the experimental data was obtained to be as:
Y ¼ 1:05 þ 0:22X1 þ 0:046X2 þ 0:030X3 þ 0:044X1 X2 þ 0:067X1 X3 0:048X2 X3 0:063X12 0:21X22 þ 0:031X32 (3) where, Y is the predicted H2 yield; X1 , X2 , and X3 are the coded values for NH4Cl, K2HPO4 and KH2PO4 respectively. ANOVA results indicate that the model is significant (Fvalue ¼ 347.6) with an insignificant lack of fit (p > 0.1) of 0.59. The R2 value of 0.99 was in good agreement with the predicted R2 value of 0.98, with a signal to noise ratio of 67.5. All the variables showed both individual and interacting significant effect on the response. The model was significant, but the response surface graphs indicated that the optimal region was beyond the design range (data not shown). This was due to the concentration range chosen in the steepest ascent. From Table 2 in PlacketteBurman design it could be noted that among the three selected factors, the significance (pvalue) followed the order NH4Cl < KH2PO4 < K2HPO4. Increasing amounts of phosphates in the growth medium can contribute to a reduction in microbial growth and H2 production due to high cytoplasmic osmotic pressure [29]. Thus, a decline in H2 yields in the steepest ascent test maybe due to high concentrations of either K2HPO4 or KH2PO4. In order to investigate the optimal concentration ranges that would include the maximal response, the adjustment direction was next determined by ridge analysis. 3.4. Ridge analysis Ridge analysis was chosen as the next step to determine the design range that would include the maximal response. This method uses the second order polynomial model from BoxeBehnken design (Eq. (3)). The approach used in here was to design the experimental points to start in the center of the previous BoxeBehnken design and to end at the point in which the predicted
R. Mangayil et al. / Renewable Energy 75 (2015) 583e589
response exceeded the theoretical maximum for H2 production from glycerol. The experimental design and predicted and observed H2 yields were as shown in Table 6. Trial runs designed based on Eq. (3) indicated that the concentration variations were significant for NH4Cl and KH2PO4. The concentration of K2HPO4 remained around 1.60 g/L (Runs 1e4). Results indicated that H2 yield increased from Runs 1 to 3, after which a decline in the response was observed. As observed earlier, in PlacketteBurman design, K2HPO4 showed the least significance among the variables that positively influenced the H2 yield (Table 2). From the data obtained from the ridge analysis experiments, it was decided that the concentration of K2HPO4 can be kept at 1.60 g/L in subsequent optimization steps. The data also point towards further optimization of NH4Cl and KH2PO4. 3.5. Central composite face-centered design NH4Cl and KH2PO4 were subjected to final optimization step by CCD, keeping the other media component concentrations as 1.6 g/L of K2HPO4, 1.0 g/L of MgCl2.6H2O, 1.0 g/L of KCl, 1.0 g/L of Na-acetate.3H2O and 2.0 g/L of tryptone. The CCD consisted of 13 runs including 5 runs at the center point as shown in Table 7. Table 8 presents the details of the second order polynomial model. The H2 yield ranged from 0.93 ± 0.04 mol-H2/mol-glycerolconsumed (Run 2) to 1.41 ± 0.03 mol-H2/mol-glycerolconsumed (Run 9). Multiple regression analysis was applied to the experimental data and a second order polynomial model was obtained:
Y ¼ 1:39 þ 0:099X1 0:065X2 þ 0:016X1 X2 0:15X12 0:12X22 (4) where, Y is the predicted H2 yield; X1 and X2 are the coded values of NH4Cl and KH2PO4 respectively. ANOVA results indicated a high determination coefficient (R2 ¼ 0.97) and a high adjusted determination coefficient (Adj R2 ¼ 0.96) implying a highly significant model. The model was significant (F-value ¼ 64.3) with an insignificant lack of fit (0.30). The results suggest that the two variables, individually and in the square of terms, were significant to the response. Threedimensional response surface and two-dimensional contour graphs were plotted using the Design-Expert 8.0.7.1 (Stat-Ease Inc.,
587
MN, USA) demo version software in order to understand the optimal levels of the variables tested. Fig. 1(A and B) illustrates the contour curve and corresponding 3-dimensional response surface plots on X1 (NH4Cl) and X2 (KH2PO4). Elliptical contours are obtained when there is a significant interaction between the variables tested [30]. The ANOVA suggested that the two variables were interdependent and the interaction effect between NH4Cl and KH2PO4 ðX1 X2 Þ were insignificant (p-value ¼ 0.29). H2 yield increased as the concentration of NH4Cl was increased from 2.95 to 4.05 g/L and then decreased with a further increase in the concentration level. Optimal levels of N2 in growth media are beneficial for microbial growth and H2 production and vary with the inoculum type [15,31,32]. Similar effect was observed for KH2PO4 (1.52e2.52 g/L), indicating an optimal buffering condition in the medium [29,31]. The optimal concentrations of NH4Cl and KH2PO4, obtained from partial differentiation of Eq. (4), were 4.40 g/L and 2.27 g/L respectively. At these conditions, the H2 yield was predicted to be 1.41 mol-H2/mol-glycerolconsumed (Fig. 1). 3.6. Validation of the model For confirmation experiments, an optimized HM100 medium (NH4Cl, 4.40 g/L; K2HPO4, 1.6 g/L; KH2PO4, 2.27 g/L; MgCl2.6H2O, 1.0 g/L; KCl, 1.0 g/L; Na-acetate.3H2O, 1.0 g/L; tryptone, 2.0 g/L; Nadithionite, 0.05 g/L and resazurin, 0.002 g/L) containing crude glycerol was prepared. The initial pH was adjusted to 6.5 and the cultivation temperature was 40 C. The batch experiments in serum bottles were conducted in triplicate experiments with a substrate blank and the data obtained were averaged. Acetate, ethanol and butyrate were the major metabolic end products produced. The H2 yield under the optimized conditions was found to be 1.42 ± 0.15 mol-H2/mol-glycerolconsumed. The H2 yield from the validation experiment indicated an excellent correlation between the predicted and experimental values. The verification experiment revealed only 0.6% difference between the predicted (1.41 mol-H2/mol-glycerolconsumed) and experimental (1.42 ± 0.15 mol-H2/mol-glycerolconsumed) values. In our previous study, H2 yield of 1.1 ± 0.1 mol-H2/mol-glycerolconsumed was obtained with modified HM100 medium (NH4Cl 1.0 g/L, K2HPO4 0.3 g/ L, KH2PO4 0.3 g/L, MgCl2.6H2O 2.0 g/L, KCl 4.0 g/L, Na-acetate.3H2O 1.0 g/L, tryptone 1.0 g/L, cysteineeHCl 0.5 g/L and resazurin 0.002 g/
Fig. 1. (A) Two dimensional contour plot and (B) three dimensional response surface plot showing the effects NH4Cl and KH2PO4 on YH2.
588
R. Mangayil et al. / Renewable Energy 75 (2015) 583e589
Table 9 Overview of recent studies on mesophilic biohydrogen production from crude glycerol by pure strains and mixed cultures. Inocula
Statistical methods P.B.
Pure cultures K. pneumoniae E. aerogenes E. aerogenes E. aerogenes KKU-S1 Halanaerobium saccharolyticum subsp. saccharolyticum Mixed cultures isolated from Wheat soil Activated sludge Wastewater Anaerobic digested sludge Activated sludge
S.A.
B.B.
R.A.
C.C.D.
Y.E. (g/L)
T.E. (g/L)
YH2
Ref.
Yes No No Yes No
Yes No No Yes Yes
0.53 0.84 0.85 0.12 3.0
[2] [10] [33] [34] [35]
No No No Yes No
Yes No No Yes No
0.31 0.90 0.96 0.41 1.42
[8] [11] [12] [36] cur.st
F.F.
X X X X NA
NA
NA
NA
X NA
NA
NA NA X NA X
NA NA X NA X
NA NA X NA X
NA NA
NA NA
NA NA
NA X
NA X
NA
P.B. e PlacketteBurman design; S.A. e Path of steepest accent; B.B. e BoxeBehnken design; R.A. e Ridge analysis; C.C:D. e Central composite design; Y.E. e Yeast extract; T.E. e Trace elements; Ref. e Reference and NA e Not applicable (Statistical optimizations not conducted).
L) in batch fermentation experiments [3]. The maximum H2 yield of 1.42 ± 0.15 mol-H2/mol-glycerolconsumed achieved after medium optimization by RSM was 29% higher than our previous reported yield [3]. This confirms RSM analysis as a powerful technique to optimize the bioconversion of crude glycerol using mixed microbial inoculum.
was the sole phosphate source, but the H2 production remained similar when used along with KH2PO4 (PH/KH 0.67). Similar to the observations from the current study and by Jitrwung and Yargeau (2011) suggests that adequate phosphate levels in culture medium are crucial for cell growth and optimal H2 production. 4. Conclusions
3.7. Comparison of previously reported statistically optimized media Increasing volume of residual glycerol as a waste from biodiesel production plants has been widely exploited as an alternative source for the production of industrial important high value compounds such as H2, 1,3-propanediol, ethanol and other organic acids. Table 9 summarizes the statistical optimization methods and results from recent studies on biohydrogen production from crude glycerol under mesophilic conditions. However, it is worth to note that a direct comparison is impossible due to different operational parameters, inocula, crude glycerol contents, substrate concentration and use of additional media supplements. Only very few studies [12] have focused on statistical optimization for improved mesophilic biohydrogen production from crude glycerol by mixed microbial consortia. However, for pure strains more studies have been conducted. One of the first studies dealing with statistical media optimization for improved biohydrogen production from crude glycerol was reported by Liu and Fang (2007) using Klebsiella pneumoniae DSM 2026 [2]. The authors reported a H2 yield of 0.4 and 0.53 mol-H2/mol-glycerol from 20.4 g/L glycerol with an optimized medium containing 5.7 g/L KCl, 13.8 g/L NH4Cl, 1.5 g/L CaCl2 supplemented with 3 g/L yeast extract from batch and 5 L stir tank bioreactor, respectively. Jitrwung and Yargeau (2011) have investigated on optimizing media components for improved biohydrogen production from crude glycerol by E. aerogenes [33]. The optimization was performed using BoxeBehnken design and response surface methodology on medium components (NH4NO3, FeSO4, and Na2HPO4) and O2, reporting a yield of 0.85 mol H2/mol glycerol from 18 g/L of crude glycerol. Recently, Jitrwung et al. (2013) have conducted optimization of salts (MgSO4, Na2EDTA, CaCl2, Na2HPO4, and KH2PO4) present in the culture media aiming to shorten the microbial growth lag phase [10]. Authors reported a positive effect of Mgþ ions (0.2 g/L MgSO4) on microbial cell growth and observed no effect on hydrogen production. Here the higher media concentration of MgCl2 (2 g/L) had a negative effect on hydrogen production and improved H2 production was observed when Mgþ ion concentration (1 g/L MgCl2) was kept at low levels. A short lag phase was observed when Na2HPO4
The current study was successful in applying the DoE to enriched microbial consortia in order to screen and identify the optimal concentrations of media components. The PlacketteBurman identified NH4Cl and KH2PO4 as variables that positively influenced the H2 production. A H2 yield of 1.41 mol-H2/mol-glycerolconsumed was predicted by response surface methodology with CCD using an optimized media (NH4Cl, 4.40 g/L; K2HPO4, 1.6 g/L; KH2PO4, 2.27 g/L; MgCl2.6H2O, 1.0 g/L; KCl, 1.0 g/L; Na-acetate.3H2O, 1.0 g/L and tryptone, 2.0 g/L). High correlation was observed between predicted and experimental data with a variance of 0.58%. In the present study, under optimized conditions, a 29% improvement in H2 yield is reported by a functional microbial community compared to previously obtained result [3]. Acknowledgments The research was funded by the Maj and Tor Nessling Foundation (Project no: 2012356) and The Academy of Finland (Project no's. 126974 and 139830). References [1] Yang F, Hanna M, Sun R. Value-added uses for crude glycerol e a byproduct of biodiesel production. Biotechnol Biofuels 2012;5:13. [2] Liu F, Fang B. Optimization of bio-hydrogen production from biodiesel wastes by Klebsiella pneumonia. Biotechnol J 2007;2:374e80. [3] Mangayil R, Karp M, Santala V. Bioconversion of crude glycerol from biodiesel production to hydrogen. Int J Hydrogen Energy 2012;37:12198e204. [4] Hiremath A, Kannabiran M, Rangaswamy V. 1,3-Propanediol production from crude glycerol from jatropha biodiesel process. New Biotechnol 2011;28: 19e23. [5] Oh BR, Seo JW, Heo SY, Hong WK, Luo LH, Joe MH, et al. Efficient production of ethanol from crude glycerol by a Klebsiella pneumoniae mutant strain. Bioresour Technol 2011;102:3918e22. [6] Almeida JRM, F avaro LCL, Quirino BF. Biodiesel biorefinery: opportunities and challenges for microbial production of fuels and chemicals from glycerol waste. Biotechnol Biofuels 2012;5:48. [7] Sarma SJ, Brar SK, Sydney EB, Le Bihan Y, Buelna G, Soccol CR. Microbial hydrogen production by bioconversion of crude glycerol: a review. Int J Hydrogen Energy 2012;37:6473e90. [8] Selembo PA, Perez JM, Lloyd WA, Logan BE. Enhanced hydrogen and 1,3propanediol production from glycerol by fermentation using mixed cultures. Biotechnol Bioeng 2009;104:1098e106.
R. Mangayil et al. / Renewable Energy 75 (2015) 583e589 [9] Ngo TA, Kim MS, Sim SJ. High-yield biohydrogen production from biodiesel manufacturing waste by Thermotoga neapolitana. Int J Hydrogen Energy 2011;36:5836e42. [10] Jitrwung R, Verrett J, Yargeau V. Optimization of selected salts concentration for improved biohydrogen production from biodiesel-based glycerol using Enterobacter aerogenes. Renew Energy 2013;50:222e6. [11] Varrone C, Rosa S, Fiocchetti F, Giussani B, Izzo G, Massini G, et al. Enrichment of activated sludge for enhanced hydrogen production from crude glycerol. Int J Hydrogen Energy 2013;38:1319e31. [12] Varrone C, Giussani B, Izzo G, Massini G, Marone A, Signorini A, et al. Statistical optimization of biohydrogen and ethanol production from crude glycerol by microbial mixed culture. Int J Hydrogen Energy 2012;37:16479e88. [13] Saraphirom P, Reungsang A. Optimization of biohydrogen production from sweet sorghum syrup using statistical methods. Int J Hydrogen Energy 2010;35:13435e44. [14] Sittijunda S, Reungsang A. Media optimization for biohydrogen production from waste glycerol by anaerobic thermophilic mixed cultures. Int J Hydrogen Energy 2012;37:15473e82. [15] Wang J, Wan W. Factors influencing fermentative hydrogen production: a review. Int J Hydrogen Energy 2009;34:799e811. [16] Lin CY, Lay CH. Effects of carbonate and phosphate concentrations on hydrogen production using anaerobic sewage sludge microflora. Int J Hydrogen Energy 2004;29:275e81. [17] Wang XJ, Ren NQ, Sheng Xiang W, QianGuo W. Influence of gaseous endproducts inhibition and nutrient limitations on the growth and hydrogen production by hydrogen-producing fermentative bacterial B49. Int J Hydrogen Energy 2007;32:748e54. [18] Plackett RL, Burman JP. The design of optimum multifactorial experiments. Biometrika 1946;33:305e25. [19] Wang J, Wan W. Experimental design methods for fermentative hydrogen production: a review. Int J Hydrogen Energy 2009;34:235e44. [20] Box GEP, Behnkin EW. Some new three level designs for the study of quantitative variables. Technometrics 1960;2:455e75. [21] Hoerl AE. Optimum solution of many variables equations. Chem Eng Prog 1959;55:69e78. [22] Draper NR. Ridge analysis of response surfaces. Technometrics 1963;5: 469e79.
589
[23] Russell VL. Response-surface methods in R, using rsm. J Stat Softw 2009;32: 1e17. [24] Khuri AI, Mukhopadhyay S. Response surface methodology. WIREs Comput Stat 2010;2:128e49. [25] Ihaka R, Gentleman RR. a language for data analysis and graphics. J Comp Graph Stat 1996;5:299e314. [26] Mangayil R, Santala V, Karp M. Fermentative hydrogen production from different sugars by Citrobacter sp. CMC-1 in batch culture. Int J Hydrogen Energy 2011;36:15187e94. € m C, Wold S. Design of [27] Eriksson L, Johansson E, Kettaneh-Wold N, Wikstro experiments: principles and applications. 3rd ed. Umetrics Academy; 2008. [28] Long C, Cui J, Liu Z, Liu Y, Long M, Hu Z. Statistical optimization of fermentative hydrogen production from xylose by newly isolated Enterobacter sp. CN1. Int J Hydrogen Energy 2010;35:6657e64. [29] Pan CM, Fan YT, Xing Y, Hou HW, Zhang ML. Statistical optimization of process parameters on biohydrogen production from glucose by Clostridium sp. Fanp2. Bioresour Technol 2008;99:3146e54. [30] Xu H, Sun LP, Shi YZ, Wu YH, Zhang B, Zhao DQ. Optimization of cultivation conditions for extracellular polysaccharide and mycelium biomass by Morchella esculenta As51620. Biochem Eng J 2008;39:66e73. [31] Bisaillon A, Turcot J, Hallenbeck PC. The effect of nutrient limitation on hydrogen production by batch cultures of Escherichia coli. Int J Hydrogen Energy 2006;31:1504e8. [32] Salerno MB, Park W, Zuo Y, Logan BE. Inhibition of biohydrogen production by ammonia. Water Res 2006;40:1167e72. [33] Jitrwung R, Yargeau V. Optimization of media composition for the production of biohydrogen from waste glycerol. Int J Hydrogen Energy 2011;36:9602e11. [34] Reungsang A, Sittijunda S, Angelidaki I. Simultaneous production of hydrogen and ethanol from waste glycerol by Enterobacter aerogenes KKU-S1. Int J Hydrogen Energy 2013;38:1813e25. € A, Santala V, Karp M. Non-sterile process for biohydrogen and 1,3[35] Kivisto propanediol production from raw glycerol. Int J Hydrogen Energy 2013;38: 11749e55. [36] Seifert K, Waligorska M, Wojtowski M, Laniecki M. Hydrogen generation from glycerol in batch fermentation process. Int J Hydrogen Energy 2009;34: 3671e8.