Bioprocess intensification for acetone-butanol-ethanol fermentation from brewer's spent grain: Fed-batch strategies coupled with in-situ gas stripping

Biomass and Bioenergy 156 (2022) 106327 Available online 23 December 2021 0961-9534/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Bioprocess intensification for acetone-butanol-ethanol fermentation from brewer’s spent grain: Fed-batch strategies coupled with in-situ gas stripping Pedro E. Plaza a , b , M´ onica Coca a , b , Susana Lucas Yagüe a , b , Gloria Guti´ errez a , c , Eloísa Roch´ on d , M. Teresa García-Cubero a , b , * a Institute of Sustainable Processes, University of Valladolid, Dr. Mergelina, s/n, 47011, Valladolid, Spain b Department of Chemical Engineering and Environmental Technology, University of Valladolid, c/Dr. Mergelina s/n, 47011 Valladolid, Spain c Department of Systems Engineering and Automatic Control, University of Valladolid, Dr. Mergelina s/n, 47011, Valladolid, Spain d Dpto. Bioingeniería, Facultad de Ingeniería, Universidad de la República, J. Herrera y Reissig 565, CP 11300, Montevideo, Uruguay ARTICLE INFO Keywords: Brewer’s spent grain Fed-batch ABE fermentation In-situ gas-stripping Clostridium beijerinckii Mathematical modelling ABSTRACT A fed-batch ABE fermentation process coupled with in-situ gas-stripping to mitigate butanol toxicity was investigated. Two feeding strategies were compared: pulses of sugars and continuous feeding of the liquid released in the dilute acid pretreatment of brewer’s spent grain. The concentrations of butanol (13.2 g L -1 total and 50 g L -1 average in the condensates for glucose pulse feeding) were higher than those obtained under batch conditions, showing that in-situ gas-stripping can relieve butanol toxicity. The continuous feeding of the fedbatch reactor produced similar butanol concentrations (10.2 g L -1 total and 65 g L -1 in the condensates), more stable concentrations of solvents in the condensates and enhanced monosaccharides uptake (99.1%) in comparison with the pulse feeding strategy. The efficient utilization of the enzymatic and pretreatment hydrolysates in the same fed-batch reactor is an integrated approach that could reduce capital and operating costs. The mathematical model proposed showed good performance to predict concentrations in the fermentation broth. 1. Introduction Butanol is a bulk chemical used in the production of paints, rubbers and resins and can be considered an advanced biofuel [1]. Butanol presents some advantages over ethanol, such as a higher calorific value (29 vs 21 MJ L -1 ) similar to gasoline (32 MJ/L), among others [2]. The butanol market was estimated at 4560 million USD in 2020, with a prevision of 7150 million USD for 2028 [3]. Butanol can be obtained by means of chemical synthesis from petroleum derived products [4] or through fermentation processes. Nowadays, the emphasis is focused on the use of renewable raw materials from low cost lignocellulosic residues from agricultural and agro-industrial processes. The biochemical route is based on the use of microorganisms from the Clostridium genus [5]. These microorganisms produce acetone, butanol and ethanol at different ratios, depending of the strain used and the fermentation conditions; although it is frequent to find a mass ratio of 3:6:1 acetone:butanol:ethanol (ABE) [6]. C. beijerinckii was used in this study due to its capacity for pH auto-regulation and lack of strain degeneration compared to other Clostridium strains [7,8]. The performance of ABE fermentation can be affected by several factors, such as the initial pH, the substrate concentration [9], the presence of toxic compounds [10] and inhibition by solvents production [11]. The majority of these factors can be solved through detoxification processes and the conditioning of the substrate, except the strong inhibition caused by solvents. It has been demonstrated that product inhibition, especially by butanol, is the main reason for the low concentrations of solvents in ABE fermentation and one of the principal limitations for the industrial production of butanol [12]. Solvent production ceases, due to the hydrophobic nature of butanol, when the butanol concentration reaches values between 8 and 15 g L -1 , depending on the microorganism and strain [13]. Overcoming product inhibition is one of the main challenges to make the separation process energetically competitive. Considering the low butanol titers and the high boiling point of butanol (118 ◦C), reducing the energy consumption of the separation process is a critical point, as the purification of butanol through conventional distillation requires large quantities of energy (79.5 MJ kg -1 butanol [14]), making the * Corresponding author. Institute of Sustainable Processes, University of Valladolid. Dr. Mergelina, s/n, 47011, Valladolid, Spain. E-mail addresses: [email protected], [email protected] (M.T. García-Cubero). Contents lists available at ScienceDirect Biomass and Bioenergy journal homepage: www.elsevier.com/locate/biombioe https://doi.org/10.1016/j.biombioe.2021.106327 Received 23 June 2021; Received in revised form 16 November 2021; Accepted 7 December 2021 Biomass and Bioenergy 156 (2022) 106327 2 process energetically non-viable. The integration of in-situ product recovery processes with ABE fermentation continuously removes solvents from the broth, reducing butanol toxicity. The recovered product has a higher butanol concentration, which reduces downstream processing costs. Integrated recovery processes must minimize costs, not affect microorganisms negatively, have a high affinity, selectivity, robustness and ease of implementation [15]. Some of the more commonly investigated technologies are liquid-liquid extraction [16], pervaporation [17], and perstraction [18]. Although these alternatives can reach high concentrations of solvents at low or moderate energy consumptions, they present some operational problems that made the gas-stripping separation process more attractive. The gas-stripping separation strategy removes volatile components from the fermentation broth without harming the microorganisms [19], while improving the product yield, solvent productivity and also reducing waste stream volumes [2]. In this process, the fermentation off-gas is fed into the bioreactor to drag the volatile components, which are later recovered in a condenser. Total ABE solvents concentrations can reach 32 g L -1 using Clostridium acetobutylicum DSM 792 [9]. Thus, a concentrated condensate is obtained, which could reduce downstream energy consumption to 14–31 MJ kg -1 butanol [16]. ABE fermentation, coupled with in-situ gas-stripping, can be carried out in batch mode [20], but the solvents production is limited as the microorganism depletes the monosaccharides without reaching product inhibition. In this case, a fed-batch operation strategy can assure longer term operation, as it can avoid both the substrate and product inhibition, enabling the operation time to be extended and increasing productivity. The feeding is usually carried out through pulses with high concentrations of monosaccharides [21]. The objective of this study was to analyze different feeding strategies for the fed-batch ABE fermentation process coupled with in-situ gasstripping to alleviate butanol toxicity and obtain condensates concentrated in butanol. Feeding with pulses of concentrated sugar solutions and the continuous feeding of the liquid fraction obtained after dilute acid pretreatment of brewer’s spent grain (BSG) have been compared to select the most suitable to increase butanol production. The enzymatic hydrolysate and the pretreatment liquid fraction obtained from a lignocellulosic agro-industrial residue, such as BSG, were used as substrates to investigate an efficient strategy for integrating the whole process hydrolysates and reducing waste streams. Moreover, a mathematical model was developed to describe the integrated fermentationrecovery process. This work compares different fed-batch feeding strategies in an ABE fermentation process with in-situ solvent recovery by gas-stripping, such as pulse feeding of monosaccharide solutions (glucose, xylose and a mixture of both glucose and xylose) and continuous feeding of the liquid produce in the pretreatment step in order to increase butanol titer. To the best of our knowledge, there are no previous published articles analyzing this continuous feeding strategy that allows the integration of the different process hydrolysates. 2. Materials and methods 2.1. Hydrolysates to be fermented from BSG The hydrolysates used in the ABE fermentation process were obtained through a dilute sulfuric acid pretreatment and further enzymatic hydrolysis of BSG. BSG was kindly donated by a local brewery. The fresh BSG was stored at −20 ◦C before use. Prior to the experimental runs, the BSG was dried in an oven at 45 ◦C. The BSG was used unmilled in the whole process. The chemical composition of BSG is shown in Table 1. The BSG was pretreated under dilute acid conditions (15% w/w solids load, 0.05 g H 2 SO 4 g -1 DM, 121 ◦C, 30 min), as established in previous works [22]. This pretreatment produced a liquid hydrolysate (Table 1) rich in monosaccharides that can be valorized by feeding the fed-batch ABE fermentation process. The enzymatic hydrolysis was performed on a Labfors Biofors HT (Infors, Switzerland) bioreactor using the pretreated BSG, type II water and a commercial enzyme cocktail composed by Cellic CTec2 (120 FPU mL -1 , cellulases and β-glucosidases). The enzyme was kindly provided by Novozymes (Denmark). The Table 1 Composition of BSG, pretreated BSG, pretreatment liquid and enzymatic hydrolysate. Solids BSG (% w/w DM) Pretreated BSG (%w/w DM) Glucan 17.3 ±0.3 23.2 ±0.9 Xylan 14.0 ±0.3 10.6 ±0.5 Arabinan 6.6 ±0.1 2.6 ±0.3 Acid insoluble lignin 18.9 ±0.6 26.2 ±0.4 Total lignin 25.5 ±0.6 30.6 ±0.6 Total ash 3.7 ±0.0 3.8 ±0.1 Detoxified Liquids Pretreatment (g L -1 ) Enzymatic hydrolysate (g L -1 ) Glucose 6.9 ±0.4 38.3 ±1.5 Xylose 15.3 ±0.9 5.1 ±0.3 Arabinose 8.9 ±0.6 n.d Acetic acid 0.9 ±0.0 0.6 ±0.0 Furfural n.d. n.d. HMF n.d. n.d. Phenolic compounds 0.6 ±0.0 0.6 ±0.0 n.d.: not detected. Fig. 1. Schematic diagram of the fed-batch fermentation process coupled with in-situ gas-stripping. P.E. Plaza et al. Biomass and Bioenergy 156 (2022) 106327 3 experimental conditions (15% w/w DM and 15 FPU g -1 DM, 50 ◦C, pH 4.8, 80 rpm) were determined in a previous work [10]. The enzymatic hydrolysis produced a monosaccharide rich liquid stream that was later used as substrate for the ABE fermentation process. According to previous results [10], the pretreatment liquid and the enzymatic hydrolysate were subjected to a detoxification process. The detoxification was carried out with activated charcoal at a solid/liquid ratio of 1.5% w/v on a rotary shaker at 35 ◦C, 135 rpm for 1 h. The mixture was then vacuum filtered, the pH was adjusted to 5.5 and the detoxified liquids were used for the fermentation process. All experiments were carried out in triplicate. 2.2. Microorganism The microorganism Clostridium beijerinckii DSM 6422 was obtained from the German collection of microorganisms (DSM, Leibniz, Germany). The strain was maintained on Reinforced Clostridial Medium, RCM (Fluka, Sigma-Aldrich, Spain) in Hungate tubes (18 ×150 mm), in spore form and cold stored at 4 ◦C under anaerobic conditions. The inoculum was grown as previously explained in other studies [22]. 2.3. Fed-batch fermentation coupled with in-situ gas-stripping The fed-batch fermentation coupled with in-situ gas-stripping (Fig. 1) was carried out in a 2 L bioreactor (Biostat Bioplus), containing 0.7 L of BSG enzymatic hydrolysate (total monosaccharides concentration about 43 g/L) as substrate for ABE fermentation by C. beijerinckii DSM 6422. After sterilization subjecting the substrate to 90 ◦C during 20 min inside the reactor, a vitamin solution (0.001 g L -1 PABA and 0.00001 g L -1 biotin), a salt solution (0.40 g L -1 MgSO 4 , 0.015 g L -1 MnSO 4 , 0.05 g L -1 FeSO 4 and 0.01 g L -1 NaCl, 0.06 g L -1 KCl) and acetate buffer solution (0.50 g L -1 KH 2 PO 4 , 0.50 g L -1 K 2 HPO 4 and 2.20 g L -1 ammonium acetate) were added to the medium and the preculture was then inoculated at 10% (v/v) and flushed with free O 2 nitrogen. The pH was adjusted to 6.3 ±0.1. The temperature was selected at 35 ◦C and the stirring was set at 50 rpm. Prior to the fermentation, the gas-stripping equipment (condenser and gas lines) was flushed with O 2 free nitrogen. The fermentation process was started in batch mode for 32 h using the enzymatic hydrolysate as substrate. At this moment, the gas-stripping started, at a flowrate of 1.0 vvm, by recycling the fermentation off-gas (a mixture of CO 2 and H 2 ) using a peristaltic pump (Masterflex Quickload) and 18 size Tygon pump tubing (Cole-Parmer). The vapors from the gas-stripping process were cooled in a condenser at 0 ◦C, using a glycerol-water 30% v/v solution in a refrigerated circulating bath (Fisher Scientific Isotemp). The condensed solvents were collected in a flask immersed inside the refrigerated circulating bath. Samples were taken periodically from the reactor for monosaccharides, organic acids, solvents and optical density analysis and from the condenser of the gasstripping process for solvent concentration analysis. The integrated fermentation process was carried out with two different feeding strategies to compare their performances. In the first set of experiments, concentrated monosaccharide solutions (one experiment was performed with a solution of 600 g L -1 of glucose, another with a solution of 600 g L -1 of xylose and a third with a mixture of both sugars with 450 g L -1 glucose and 150 g L -1 xylose) were added in pulses to elevate the sugar concentration without significantly increasing the volume inside the reactor. In the second set of experiments, the pretreatment liquid, rich in pentose sugars, was used for continuous feeding to valorize the pentoses in this hydrolysate. The continuous feeding rate (10.6 mL h -1 ) was adjusted to compensate for the uptake rate of monosaccharides by the microorganism. The feeding solution was added with a peristaltic pump (Watson Marlow 520S). An antifoam (Antifoam 204, Sigma) was added to control foam formation, if necessary. Two different yields and productivities have been evaluated at 120 h: the first one was based on the total concentration of solvents produced, considering the solvents collected in the gas stripping condensate and the solvents remaining in the fermentation broth, while the second one was evaluated taking into account only the solvents collected in the gas stripping condensate. All experiments were carried out in triplicate. 2.4. Analytical methods The chemical composition of the BSG was analyzed as described in a previous work [22]. The concentration of monosaccharides, solvents, organic acids, and potential inhibitors in the liquids were measured by HPLC as described in a previous work [22]. The biomass concentration was determined through the optical density analyzed using a spectrophotometer (Hitachi U-2000) at 600 nm. 2.5. Data analysis An ANOVA variance analysis was performed to determine statistical differences at a confidence level of 95% (p <0.05). A Tukey multiple range test was performed using Statgraphics Centurion XVIII. 2.6. Process modeling of the continuous fed-batch fermentation A mathematical model was proposed to describe the fed-batch fermentation under continuous feeding coupled with in-situ gasstripping. The model was based on Monod kinetic considering a term of product inhibition and cell death [9]. Individual mass balances for the concentration of cells (X), glucose (S G ), xylose (S X ), arabinose (S A ) and butanol (P) in the bioreactor are described by the following equations: dX dt =( μ mGSG KSG +SG + μ mX SX KSX +SX + μ mASA KSA +SA)*X*(1−P Kp)a−kd*X−Qin V*X (1) dSG dt =Qin V(SG,in −SG)−(( μ mGSG KSG +SG)*(1−P Kp)a *X YX/SG)(2) dSX dt =Qin V(SX,in −SX)−(( μ mX SX KSX +SX)*(1−P Kp)a *X YX/SX)(3) dSA dt =Qin V(SA,in −SA)−(( μ mASA KSA +SA)*(1−P Kp)a *X YX/SA)(4) dV dt =Qin (6) In Eqs. (1) to (6), X is the cell concentration in the bioreactor (dry weight, g L -1 ); μ m is the maximum specific growth rate (h −1 ) for glucose ( μ mG ), xylose ( μ mX ) and arabinose ( μ mA ); S (g L -1 ) is the substrate dP dt =(1−P Kp)a *X*[( μ mGSG KSG +SG)(YP/SG YX/SG)+( μ mX SX KSX +SX)(YP/SX YX/SX)+( μ mASA KSA +SA)(YP/SA YX/SA)]− − kSa*P−Qin V*P(5) P.E. Plaza et al. Biomass and Bioenergy 156 (2022) 106327 4 concentration in the fermentation broth for glucose (S G ), xylose (S X ) and arabinose (S A ); S in (g/L) is the substrate concentration in the hydrolysate fed for glucose (S G,in ), xylose (S X,in ) and arabinose (S A,in ). K S (g L -1 ) is the substrate saturation parameter for glucose (K SG ), xylose (K SX ) and arabinose (K SA ). P is the butanol concentration in the fermentation broth (g L -1 ); K p is the product concentration at which no cell growth occurs (g/L); α is the degree of product inhibition and k d is the specific growth rate (h −1 ). Y X/S (g g -1 ) is the cell yield coefficient for glucose (Y X/SG ), xylose (Y X/SX ) and arabinose (Y X/SA ). Y P/S (g g -1 ) is the butanol yield coefficient corresponding to glucose (Y P/SG ), xylose (Y P/SX ) and arabinose (Y P/SA ). k s a (h −1 ) is the butanol removal rate parameter, evaluated from batch stripping experiments, considering that the rate of removal of solvents, r P follows the equation: rP=dP dt = − ksa⋅P(7) The variation of the reaction volume (V) due to gas-stripping was considered negligible in comparison to the inlet flow rate (Q in ). The software EcosimPro© was used to estimate the kinetic parameters that produce the best fit between the experimental data and the results predicted by the model. The parameter estimation was carried out by dynamic optimization using Sequential Quadratic Programming (SQP). The objective function to be minimized considered the sum of square residuals (differences between the measured and the predicted concentrations by the model). 3. Results and discussion In order to obtain hydrolysates rich in monosaccharides to be fermented to butanol, BSG was subjected to a dilute acid pretreatment and the pretreated BSG was subjected to enzymatic hydrolysis. The pretreated BSG presented a composition of 23.2 ±0.9% w/w DM glucan, 10.6 ±0.5% w/w DM xylan and 2.6 ±0.3% w/w DM arabinan and a total lignin content of 30.6 ±0.6% w/w DM (Table 1). The pretreatment produced an hydrolysate with a content of monosaccharides after detoxification of 6.9 ±0.4 g L -1 glucose, 15.3 ±0.9 g L -1 xylose, 8.9 ± 0.6 g L -1 arabinose and a low concentration of inhibitors (0.9 ±0.0 g L -1 acetic acid, 0.6 ±0.0 g L -1 phenolic compounds). The enzymatic hydrolysis was carried out at a solids load of 15% w/w DM in order to obtain a hydrolysate more concentrated in monosaccharides [10]. After detoxification, the monosaccharides content was still high (38.3 ±1.5 g L -1 glucose, 5.1 ±0.3 g L -1 xylose), and the concentration of phenolic compounds (0.6 ±0.0 g L -1 ) is lower than that can produce inhibition of the fermentation process [10] (Table 1). 3.1. Fed-batch fermentation coupled with gas-stripping and pulse feeding In this first set of experiments, fed-batch ABE fermentation runs coupled with gas-stripping and pulse feeding were carried out using C. beijerinckii DSM 6422. The enzymatic hydrolysate from pretreated BSG was used as a substrate for the microorganism. The pulse feeding consisted of a solution of glucose, xylose or a mixture of both at a high concentration (600 g L -1 ). The gas-stripping system was started at 32 h, when the butanol concentration in the bioreactor reached 5–6 g L -1 to avoid product inhibition [9]. The pulses were added when the concentration of monosaccharides inside the reactor reached values between 5 and 15 g L -1 , with the aim of increasing the concentration to about 30–40 g L -1 and extending the operation time of the process. The experiments were conducted for 168 h at a gas flow rate of 1.0 vvm, selected from previous experimental runs (data not shown). There was no pH control in these experiments, as the microorganism can regulate the pH. The initial pH was around 6.3 ±0.1 and the final pH was around 4.6 ±0.2. The results are shown in Figs. 2–4 and Tables 2 and 3. The monosaccharides uptake was high for the three feeding strategies (85.3% for the glucose pulses, 66.2% for the xylose pulses and 64.4% for the pulses of the solution of glucose and xylose). The pulse feeding strategies based on the addition of xylose, or the solution of glucose and xylose, presented lower sugar consumptions than the glucose pulse feeding. It can also be observed that the microorganism showed a strong preference for glucose, leaving xylose unconsumed when the pulses of glucose were fed (Fig. 2A). Glucose was consumed at a higher rate when the mixture of glucose and xylose was fed (leaving 70% of the xylose unconsumed, Fig. 4A). These results are probably due to catabolite repression [23], stressing the fact that feeding xylose at the same time as glucose had a moderate effect over the performance of the process. On the other hand, the microorganism showed that, in the absence of glucose in the feeding, it can metabolize xylose as substrate, reaching a good production of solvents (11.2 g butanol L -1 and 16.3 g ABE L -1 ), yields and productivities (Tables 2 and 3). In all the experiments, the uptake of monosaccharides ceased between 120 and 168 h, leaving sugars unconsumed in the fermentation broth. This can be due to the lack of nutrients, since they were not introduced in the feeding pulses [9], to culture degeneration caused by inhibitors [24] and/or the accumulation of metabolites [25]. The production of organic acids, mainly butyric and acetic acid, was low throughout these experiments. Butyric acid was not detected whereas acetic acid reached maximum concentrations of 0.8 ±0.1 g L -1 . The concentration of butanol in the reactor declined when the gasstripping started at 32 h in the three cases (Figs. 2B, 3B and 4B), reducing product inhibition and reaching total concentrations between 11.2 and 13.2 g butanol L -1 and between 16.3 and 19.3 g ABE L -1 . Compared with the results reached under batch fermentation (6 g butanol/L and 8 g ABE/L [22]), fed-batch fermentation coupled with in-situ gas-stripping considerably improved the solvent concentrations, because product toxicity was relieved. Acetone and ethanol concentrations were similar throughout the three experiments (concentrations of acetone and ethanol about 5 g L -1 and 1 g L -1 , respectively), producing mainly butanol under these conditions. The results agreed with data provided by the scientific literature (Table 3). Lu et al. [25] reported similar butanol and ABE concentrations (total concentrations of 13.5 g butanol L -1 and 17.7 g ABE L -1 ) by batch fermentation coupled with in-situ gas-stripping using wood pulping hydrolysate as fermentation substrate and an adaptive mutant strain of C. beijerinckii (C. beijerinckii CC101) with a higher butanol tolerance. Wechgama et al. [26] analyzed a batch fermentation process by C. beijerinckii TISTR 1461 coupled with in-situ gas-stripping, reporting total concentrations of 14.1 g butanol L -1 and 18.9 g ABE L -1 using sugarcane molasses as substrate. Cai et al. [24] obtained total concentrations of 18.6 g butanol L -1 and 28.3 g ABE L -1 , with a fed-batch strategy, introducing pulses of corn stover bagasse (CSB) hydrolysate (210 g L -1 glucose, 78 g L -1 xylose) to increase the concentrations of monosaccharides in the fermentation broth from 5 to above 30 g L -1 . CSB hydrolysates were vacuum concentrated before feeding. The microorganism used by Cai et al. [24], C. acetobutylicum ABE-P 1201 derived from ATCC 824 by evolutionary engineering, also showed a preference for glucose over xylose, evidencing catabolite repression. It should be noted that ABE mass ratios at 120 h for total concentrations were 2.4:7.2:0.4 for the glucose pulse feeding run, 2.8:6.9:0.4 for the xylose pulse feeding experiment and 3:6.1:0.9 for the glucose and xylose pulse feeding. Results showed that the pulse feeding strategies did not significantly improve the butanol mass ratio in comparison with the results obtained under batch fermentation (2.9:7.0:0.2 [22]). The concentration of solvents in the liquids recovered from the condenser of the gas-stripping equipment (Figs. 2C, 3C and 4C) showed the concentration declined with time, obtaining less concentrated condensates with the time course of stripping and observing a sharp decline between 120 and 168 h, which coincided with the cease in the uptake of monosaccharides. Acetic and butyric acids were not detected in the condensates, as the gas-stripping process only removes acetone, butanol, ethanol and water. The solvent concentrations observed at 48 h were high for the three strategies (between 50.5 and 71.5 g of butanol L -1 and between 68.9 and 97.9 g of ABE L -1 ) with very low concentrations of P.E. Plaza et al. Biomass and Bioenergy 156 (2022) 106327 5 ethanol. However, butanol concentrations were not high enough to observe phase separation, which can occur when the butanol titer is higher than its solubility in water, about 77 g of butanol L -1 at 20 ◦C [19]. Phase separation results in a more energy-efficient butanol recovery process [27]. Lu et al. [25] obtained maximum concentrations in condensates of 78 g butanol L -1 and 97 g ABE L -1 (not reaching phase separation), declining to 39 g butanol L -1 and 53 g ABE L -1 at the end of the process (75 h) using a gas flow rate of 0.25 vvm. Roch´ on et al. [9] obtained condensates with a concentration high enough for phase separation using a gas flow rate of 0.4 vvm, which allows higher concentrations but lower volumes to be obtained. Under these conditions, an organic phase with 444.8 g butanol L -1 and 499.9 g ABE L -1 and an aqueous phase with 79.1 g butanol L -1 and 126.2 g ABE L -1 were obtained at the end of the process. Cai et al. [24] reached concentrations of 135 g butanol L -1 and 220 g ABE L -1 in the first condensate, declining to 75 g butanol L -1 and 110 g ABE L -1 in the final condensate after 210 h Fig. 2. Results for pulse feeding with glucose solution (600 g L -1 ). A: Monosaccharides uptake and biomass production. B: Production of solvents. C: Concentration of solvents in the liquid recovered from the gas-stripping condenser. Figure B: Total: Total concentration (continuous line), Reactor: Concentration inside the reactor (dashed line). Feeding pulses pointed out with arrows on figure A. P.E. Plaza et al. Biomass and Bioenergy 156 (2022) 106327 6 using a gas flow rate of 1.3 vvm. Although the solvent concentration is much higher, the descending trend observed in the condensate concentrations was similar to that observed in this study. The concentration of solvents in the condensate is highly dependent on the concentration of solvents in the reactor [24,25]. In Figs. 2–4, it can be clearly observed that the concentrations of solvents in the condensates decrease as the concentration of solvents in the fermenter diminished. It should be noted that the composition of the condensates presents a higher proportion of butanol than the total concentrations, indicating a higher selectivity for butanol than acetone and ethanol in the gas-stripping process. The ABE mass ratios in the condensates (calculated at 120 h) were 1.9:7.7:0.3 for the glucose pulse feeding, 2.3:7.3:0.4 for the xylose pulse feeding and 2.6:7.0:0.4 for the pulse feeding of glucose and xylose. The higher proportion of butanol showed that the gas-stripping process is more selective towards butanol. The in-situ gas-stripping process was capable of recovering large quantities of solvents from the reactor. The removal of solvents in the reactor results in recoveries of butanol between 61.5 and 72.9 g in condensates/100 g total produced and ABE between 58.0 Fig. 3. Results for pulse feeding with xylose solution (600 g L -1 ). A: Monosaccharides uptake and biomass production. B: Production of solvents. C: Concentration of solvents in the liquid recovered from the gas-stripping condenser. Figure B: Total: Total concentration (continuous line), Reactor: Concentration inside the reactor (dashed line). Feeding pulses pointed out with arrows on figure A. P.E. Plaza et al. Biomass and Bioenergy 156 (2022) 106327 7 and 69.1 g in condensates/100 g total produced (Table 2). Other authors [25] obtained somewhat lower recoveries (48% butanol, 39% acetone, 23% ethanol) at a lower gas recycled rate than that used in our study. Two yields and productivities were calculated from the experimental data: considering the total concentrations of solvents and from the concentration of solvents in the condensates. The results were very similar for the three pulse feeding strategies. The yields calculated from the total concentrations (Table 3) were similar to those reported by other authors such as Roch´ on et al. [9] (0.16–0.18 g g -1 for butanol) and lower than those obtained by Lu et al. [25] (0.23 g g -1 for butanol and Fig. 4. Results for pulse feeding with mixture of glucose and xylose (75% glucose, 25% xylose, 600 g L -1 ). A: Monosaccharides uptake and biomass production. B: Production of solvents. C: Concentration of solvents in the liquid recovered from the gas-stripping condenser. Figure B: Total: Total concentration (Continuous line), Reactor: Concentration inside the reactor (Dashed line). Feeding pulses pointed out with arrows on figure A. P.E. Plaza et al. Biomass and Bioenergy 156 (2022) 106327 8 0.33 g g -1 for ABE). The productivities reached lower values than those obtained by Roch´ on et al. [9] using C. acetobutylicum DSM 792 and sugarcane-sweet sorghum industrial juices (0.13 g butanol L -1 h -1 and 0.22 g ABE L -1 h -1 ). The yields referring to the concentrations in the condensates (Table 2) were 0.76–0.93 g g -1 for butanol and 0.98–1.27 g g -1 for ABE at 120 h. The productivity referring to the condensates reached values of 0.38–0.46 g L -1 h -1 for butanol and 0.55–0.60 g L -1 h -1 for ABE. These values are much higher than those obtained under batch conditions (0.04–0.07 g L -1 h -1 for butanol and 0.06–0.09 g L -1 h -1 for ABE) [22] due to the high concentration of solvents in the condensed liquids. To sum up, the three pulse feeding strategies produced higher concentrations of solvents than those obtained in batch fermentation processes. Catabolite repression was observed in all the experiments, as the microorganism metabolizes glucose before xylose. In the experimental run where pulses of xylose were fed, the microorganism consumed xylose at a higher rate only when the glucose was depleted. Therefore, in the absence of glucose, C. beijerinckii DSM 6422 can metabolize xylose, reaching an adequate production of solvents, yields and productivities, demonstrating that pulse feeding of xylose is as beneficial for butanol production as pulse feeding of glucose. This result opens the door to fedbatch feeding of hemicellulose hydrolysates, which are rich in xylose and arabinose. 3.2. Fed-batch fermentation coupled with gas-stripping and continuous feeding of the hemicellulosic hydrolysate The liquid obtained in the dilute acid pretreatment of BSG was used to continuously feed the bioreactor during the gas-stripping process. The objective of this strategy is to valorize the liquid fraction of the pretreatment by the production of butanol in the same bioreactor used to ferment the enzymatic hydrolysate, hence reducing capital costs. The fermentation of this stream, rich in pentoses, increases the global yields of the process [10]. For the industrial implementation of the ABE process, the hemicellulosic hydrolysate stream should be integrated with the enzymatic hydrolysate to be converted into butanol. There is no pH control used in these experiments, as the microorganism regulates the pH itself. The initial pH was around 6.3 ±0.1 and the final pH was around 4.6 ±0.2. The results are summarized in Fig. 5 and Tables 2 and 3 The concentration of monosaccharides in the pretreatment liquid (Table 1) is low for pulse feeding, as higher concentrations are required. Table 2 Influence of the feeding strategy on yields, productivities and recovery of solvents. Parameters calculated at 120 h. Feeding strategy Y butanol (g g -1 ) Y ABE (g g -1 ) Q butanol (g L -1 h -1 ) Q ABE (g L -1 h -1 ) Butanol recovery (g in condensates/100 g total produced) ABE recovery (g in condensates/100 g total produced) Pulses of glucose 0.76 0.98 0.46 0.60 61.5 62.3 Pulses of xylose 0.93 1.27 0.42 0.57 72.9 69.1 Pulses of glucose +xylose 0.81 1.16 0.38 0.55 65.1 58.0 Continuous feeding of pretreatment liquid 0.95 1.16 0.55 0.67 49.6 50.8 Butanol and ABE yields (Y Butanol , Y ABE ) expressed as g g -1 sugars consumed; butanol and ABE productivities (Q Butanol , Q ABE ) expressed as g L -1 h -1 . Yields and productivities calculated using the concentration of solvents in condensed liquids from gas-stripping. Table 3 ABE fermentation coupled with in-situ gas-stripping. Comparison of solvent concentrations, yields and productivities. Parameters calculated at 120 h. Substrate Operation mode Type of feeding Microorganism Total butanol (g L -1 ) Total ABE (g L -1 ) Y butanol Y ABE Q butanol Q ABE Reference Wood pulping hydrolysate a Batch C. beijerinckii CC101 13.5 17.7 0.23 0.33 0.13 0.17 [25] Sugarcane-sweet sorghum juices a Batch C. acetobutylicum DSM 792 10.5 17.5 0.18 n.a. 0.10 0.13 [9] Fed-batch Pulses of concentrated sugarcane-sweet sorghum juices C. acetobutylicum DSM 792 18.6 31.8 0.16 n.a. 0.13 0.22 Surgacane molasses a Batch C. beijerinckii TISTR 1461 14.1 18.9 0.39 n.a. 0.29 n.a [26] Corn stover bagasse b Fed-batch Pulses of concentrated corn stover bagasse C. acetobutylicum ABE-P 1201 18.6 28.3 0.19 0.29 0.09 0.13 [24] P2 medium a Batch C. acetobutylicum JB200 19.8 31.8 0.25 0.40 0.41 0.66 [16] Brewer’s spent grain a Fed-batch Pulses of concentrated glucose C. beijerinckii DSM 6422 13.2 18.0 0.17 0.24 0.11 0.15 This study Fed-batch Pulses of concentrated xylose C. beijerinckii DSM 6422 11.2 16.3 0.20 0.29 0.09 0.13 Fed-batch Pulses of concentrated glucose and xylose C. beijerinckii DSM 6422 11.8 19.3 0.20 0.27 0.08 0.13 Fed-batch Continuous feeding of pretreatment liquids C. beijerinckii DSM 6422 10.2 13.7 0.14 0.20 0.08 0.12 Butanol and ABE yields (Y Butanol , Y ABE ) expressed as g g -1 sugars consumed, calculated using the total production of solvents. Butanol and ABE productivities (Q Butanol , Q ABE ) expressed as g L -1 h -1 , calculated using the total production of solvents. a Continuous stripping. b Intermittent stripping; n.a.: not available. P.E. Plaza et al. Biomass and Bioenergy 156 (2022) 106327 9 One alternative is the vacuum concentration of this stream in order to make it suitable for pulse feeding [24]. However, this option would increase the energy consumption of the process, counteracting the energy savings of the gas-stripping process. Instead of this, a continuous feeding strategy was used, feeding a flow equivalent to the quantity of monosaccharides consumed by the microorganism per hour. The continuous feeding and the gas-stripping were started at 32 h when the butanol concentration was about 7 g L -1 . The feeding flow rate (10.6 mL h -1 ) was calculated through batch fermentation experiments coupled with in-situ gas-stripping (data not shown). The results show that the co-fermentation strategy improved monosaccharides uptake to 99.1%. The feeding was stopped at 120 h because the maximum functional Fig. 5. Results for continuous feeding with pretreatment liquids. A: Monosaccharides uptake and biomass production. B: Production of solvents. C: Concentration of solvents in the liquid recovered from the gas-stripping condenser. Figure B: Total: Total concentration (Continuous line), Reactor: Concentration inside the reactor (Dashed line). Continuous feeding start pointed out with arrow in figure A. P.E. Plaza et al.