Hydrolysis and fermentation steps of a pretreated sawmill mixed feedstock for bioethanol production in a wood biorefinery

10 The aim was to demonstrate the feasibility of second-generation bioethanol production using 11 for the very first time a sawmill mixed feedstock comprising four softwood species, 12 representative of biomass resource in Auvergne-Rhône-Alpes region (France). The feedstock 13 was subjected to a microwave-assisted water/ethanol Organosolv pretreatment. The 14 investigation focused on the enzymatic hydrolysis of this pretreated sawmill feedstock (PSF) 15 using Cellic ® Ctec2 as the enzyme, followed by fermentation of the resulting sugar solution 16 using Saccharomyces cerevisiae strain. The cellulose-rich PSF with 71% w/w cellulose 17 content presented a high saccharification yield (up to 80%), which made it perfect for 18 subsequent fermentation; this yield was predicted vs. time up to 5.2% w/v PSF loading using 19 a mathematical model fitted only on data at 1.5%. Finally, high PSF loading (7.5%) and 20 scaleup were shown to impair the saccharification yield, but alcoholic fermentation could still 21 be carried out up to 80% of the theoretical glucose-to-ethanol conversion yield. 22 *


Hydrolysis and fermentation steps of a pretreated sawmill mixed feedstock for bioethanol production in a wood biorefinery
Maarouf Abdou-Alio, O Tugui, L Rusu, Agnès Pons, Christophe Vial

Introduction
3 Biofuels, in particular second-generation biofuels, have gained interest from academic 4 research, government, and large companies nowadays. Lignocellulosic biofuel production is 5 indeed widely accepted in the society as it is perceived as non-competitive with agri-food 6 (Bryngemark, 2019). Also, these biofuels represent a renewable option to replace the 7 depleting oil supply and can help mitigate the climate change impacts resulting from fossil 8 fuels. For low-cost biofuel production, lignocellulosic biomass is a potential candidate 9 feedstock because its price is estimated as being the lowest compared to starch which is 10 presently used to produce bioethanol (Chovau et al., 2013). 11 Auvergne Rhône-Alpes is one of the largest forest regions in France with a forest density 12 estimated at 439 10 6 m 3 . As a consequence, the local wood processing industries generate a 13 significant amount of sawdust or other wastes that are not valorized otherwise than being 14 burnt for heat in this region. An alternative way of valorization of this waste is the production 15 of second-generation bioethanol in order to develop a reliable and sustainable regional energy 16 model that fits well with the energy transition policy and the fight against climate change. 17 This could have the advantage of promoting the valorization of local wood waste with a view 18 to setting up a forest biorefinery. Beyond the regional impact, large-scale second-generation 19 biofuel facilities could also contribute to enhance the security of energy supply and, more 20 generally, to strengthen the world economy (Sarks et al., 2014). 21 To increase lignocellulose digestibility, numerous lignocellulose pretreatment methods have 22 been matured. Among all the thermomechanical and chemical processes, Organosolv 23 pretreatment can be considered to be an eco-friendly and attractive method to remove lignin 24 from lignocellulose materials using pure or diluted organic solvents in order to purify 1 cellulose (Akgul and Kirci, 2009;Zhao et al., 2009). Thus, Organosolv pretreatments are very 2 efficient techniques for the production of second-generation bioethanol because they can 3 provide both cellulose and lignin with high purity, and minimize at the same time the 4 production of fermentation inhibitors (Mupondwa et al., 2017). 5 As a subsequent step to pretreatment, hydrolysis transforms the cellulose and the remaining 6 hemicelluloses into fermentable sugars. The enzymatic hydrolysis step has been recognized as 7 a major techno-economic bottleneck in the whole wood-to-ethanol bioconversion process. (iii) a high sugar concentration leads to severe inhibition effects of the finished products. 4 Similarly, Qiu et al. (2017) concluded that increasing initial solid concentrations above 2% 5 w/w reduced significantly the conversion rate. 6 To circumvent these issues, process optimization though mathematical modeling is the key.  Wojtusik et al., 2016). It must be pointed out that the complete mechanism of enzymatic 18 hydrolysis is still unknown, and could also depend on the type and purity of the substrate and 19 of the enzymes. 20 In this work, the goal was, therefore, to hydrolyse and ferment an industrial and readily 21 available sawmill feedstock previously pretreated through an "Organosolv pretreatment" 22 based on an original microwave heating (Alio et al., 2019). This process presents the 23 simultaneous advantage to produce high quality lignin that can be used in the closed-loop 24 biorefinery concept with potentially high added-value coproducts. So, this study focussed first 25 on glucose concentration and cellulose-to-glucose conversion yield as a function of the 1 operating parameters in a batch enzymatic hydrolysis process; from these data, a model aimed 2 to predict the glucose concentration, and the cellulose-to-glucose conversion yield could be 3 developed. Then, the feasibility of bioethanol production by fermentation using the obtained 4 enzymatic hydrolysate was studied using the yeast Saccharomyces cerevisiae. Finally, the 5 scale-up from flasks to batch bioreactors of the hydrolysis and fermentation steps for the 6 production of second-generation bioethanol was investigated. The initial composition of the sawdust mixture was determined to be 44.3% ± 0.5% glucan 14 (dry wt.), 25.6% ± 0.2% hemicellulose (dry wt.), 26% ± 3% Klason lignin (dry wt.), ash 15 content 0.3% ±0.2% (dry wt.), and 3.2% ± 0.2% extractives (dry wt.) (Alio et al., 2019). 16 The pretreated mixed sawdust was subjected to a microwave-assisted Organosolv 17 pretreatment in which four parameters were studied to optimize the fractionation of the wood, 18 including cellulose recovery yield and cellulose purity, lignin recovery yield, and the absence 19 of formation of inhibitors; these parameters were, namely: (i) the concentration of sulfuric 20 acid (H2SO4) as a catalyst; (ii) the ethanol/water ratio in the extraction solvent; (iii) the 21 treatment temperature; (iv) the process time. The optimal conditions determined 22 experimentally on sawdust could be summarized as follows: an ethanol/water ratio of 60:40 23 with 0.25% H2SO4 for one-hour extraction at 175 • C. These conditions made it possible to remove 50% of the lignin while preserving 82% ± 3% of the initial cellulose with a purity of 1 71% ± 3% w/w. The composition of the pretreated sawdust mixture was then: 70.6% w/w 2 cellulose, 9.8% w/w hemicellulose, and 19.6% w/w lignin, as determined by strong acid 3 hydrolysis (Alio et al., 2019). 4 5 The enzymatic hydrolysis was carried out using an enzyme preparation: Cellic ® Ctec 2 6 (Novozymes, Denmark). The activity on filter paper fibers (FPU) was determined as 7 described by Ghose (1987), and the protein content as described by Bradford (1976 12 All enzymatic hydrolysis runs were performed according to the NREL standard procedure 13 (Selig, 2008). All experiments were carried out using the pretreated sawdust mixture obtained 14 through the best pretreatment previously described. Depending on the assays, pretreated 15 sawdust mixture solids could be oven-dried (110°C for 24h) or not. Enzymatic hydrolysis of 16 pretreated mixture was performed in 150 mL Erlenmeyer flasks under batch conditions at 17 50°C in a shaker water bath (Julabo SW22, France) with a mixing speed of 180 rpm according 18 to Mukasekuru et al. (2018). Hydrolysis was carried out in a 50 mM acetate buffer solution at 19 pH 4.8 using the cocktail of enzyme mentioned above and streptomycin antibiotics were 20 added to prevent contamination. Different biomass loading concentrations were tested: 1.5 %, 21 3.75 %, 4.5%, and 5.2 % w/v. These were selected because the homogeneity of the liquid- 22 solid dispersion could be maintained, despite the increase in viscosity of this dispersion in this 23 range, which was not the case at higher solid substrate loadings. The carbohydrate profile in the hydrolysate was determined using HPLC, as described below in section 2.6. The 1 enzymatic hydrolysis yield was calculated using the following equation:

Enzymatic hydrolysis
where is the concentration of glucose determined by HPLC (g/L), represents 4 the volume of the liquid (L), 0.90 is the correction factor for the conversion of cellulose-to-5 glucose, is the mass fraction of cellulose (g cellulose/g solid) and is the mass 6 of dry solids (g). 7 1 For the scale-up experiments, enzymatic hydrolysis was conducted as described in section 2.3 2 using a biomass concentration of 7.5% w/v in a 1000 mL glass bottle with a working volume 3 of 800 mL; this higher substrate loading was allowed by the improved mixing conditions, so 4 as to maintain spatially homogenous reaction rate and concentrations for sampling purpose. 5 Then, the fermentation step was carried out in a 500 mL bioreactor (Infors HT, Multitron 2) 6 with a working volume of 400 mL. The hydrolysate obtained after the enzymatic hydrolysis 7 was completed as described in section 2.4. Only (E.H) was used as the culture medium. samples. After centrifugation (Thermo scientific, France) for 5 minutes at 10,000 g, the 16 supernatant was filtered using a 0.2 μm cellulose acetate filter (Chromafil, Germany). The 17 concentrations of glucose, cellobiose, xylose, ethanol, and by-products (formic acid, levulenic 18 acid, and acetic acid) were measured using a high-performance liquid chromatography 19 (HPLC) device (1260 Infinity Quaternary LC system, Agilent Technology, USA). This was 20 equipped with two ionic exclusion columns in series (Rezex ROA 300×7.8 mm, Phenomenex, 21 USA). The mobile phase was a solution of 5 mM sulfuric acid at 0.7 mL/min flowrate. 22 Products detection was done using a refractometer (HP 1100 series, Agilent Technologies, 23 USA). 1 All the results reported for batch hydrolysis and fermentation are the mean values of at least 2 two replicates based on two batches of repeated experiments under the same conditions. Three 3 samples were taken for each time point. The mean and standard deviation were calculated by 4 Excel © Sheets version 1902 (Microsoft © Office package). 5 2.8. Enzymatic hydrolysis kinetic modeling 6 As described in section 1, various models had been proposed in the literature to describe this 7 process. A rapid analysis of preliminary data had shown that the rate of glucose production 8 could not be fitted by a Lineweaver-Burk plot, which suggested a more complex mechanism 9 than classical or modified Michaelian kinetics as a function of substrate content. Similarly, 10 glucose production rate increased with substrate loading, so that the effect of enzyme mass 11 transfer resistance never seemed to become predominant, despite the increase in dispersion 12 viscosity. Actually, the initial reaction rate appeared to be nearly proportional to the initial The transformation rate of the cellulose 1 by the pair EG/CBH considering the (competitive) 2 inhibition effect by the products (cellobiose and glucose) is given by:

Data analysis and number of samples replicate
Then, the rate of transformation of cellobiose into glucose 2 by BGL is described by a 5 Michaelian kinetics: where , , , E1b and E2 represent the respective concentrations of substrate, cellobiose and 8 glucose and of the EG/BGL and BGL enzyme species (g.kg −1 ), the subscript corresponding 9 to bound enzymes. In these equations, , 1 represent the inhibition constants of glucose 10 in reaction i and of cellobiose (g.kg −1 ), respectively; 2 designates the enzyme Michaelis 11 constant in 2 (g.kg −1 ), and finally, k1r and k2r are the respective kinetic constant of r1 and r2, 12 and RS the substrate reactivity. Adsorption was modelled using a Langmuir isotherm, which 13 assumes that equilibrium is rapidly reached and can be described by 15 where the amount of free enzyme E1 is related to the amount of bound enzyme E1b by the 16 enzyme conservation law and two additional parameters: K1ad the dissociation constant for the 17 enzyme/substrate complex, and E1max the maximum mass of enzyme that can adsorb onto 18 cellulose, both expressed in g enzyme/g cellulose. Equation (4) also expresses the relationship 1 experimental data. hydrolysate. When the cellulose-to-glucose conversion yield was calculated (Fig. 1Erreur !   14 Source du renvoi introuvable.), it was found that only 12.4% ± 0.7% of the cellulose present 15 in the SG pretreated sawdust mixture was converted to glucose within 8 days of incubation. 16 Reducing the substrate particle size to a smaller granulometry (IG) doubled cellulose-to- obtained in this study highlight that a reduction in particle size significantly enhanced the 3 hydrolysis rate of cellulose. However, the grinding process may be costly because of its 4 important power requirements that counterbalance its beneficial effects on cellulose-to-  (Table 1), the costs were dramatically increased when particle 9 size was reduced below 2 mm. However, for both the Miscanthus and the switchgrass, the 10 optimal particle size ranged between 4 and 6 mm, and the corresponding total cost (grinding, 11 storage and transportation) were about $55 and $61 per ton for Miscanthus and switchgrass, 12 respectively. size to cut the cost of the grinding process, at 1.5% (w/v) substrate concentration and an 18 enzyme loading of 50 FPU/g of pretreated substrate. Regarding these results for both cases, it 19 can be noticed that the enzymatic hydrolysis was strongly affected by the humidity level. So, 20 a low conversion rate was observed with the dried pretreated substrate with a maximum 21 conversion yield reached after 8 days of 12.4 ± 0.5%, while a value of 56% ± 2% was reached 22 for the wet pretreated substrate in the same period. There is no doubt that this difference was 23 due to the low digestibility of the dried substrate, linked itself to a structure modification of the substrate as a direct impact of the drying process. According to Kang  Organosolv pretreatment that did not significantly delignify the raw sawdust material, as a 14 significant fraction of the lignin remained in the resulting pretreated substrate (39% ± 1% 15 w/w of the raw feedstocks lignin content, which represents 20% ± 2% w/w of the pretreated 16 substrate). High lignin content is, indeed, known to hinder enzymatic hydrolysis through the 17 nonproductive binding of cellulase enzymes. Likewise, as a non-cellulose component, 18 hemicellulose is generally considered as a physical hindrance in enzymatic hydrolysis of 19 cellulose, and prevents the access of cellulase from cellulose surface (Qiu et al., 2017). 20 Finally, it emerges from experimental data that there is no need to dry the pretreated substrate, 21 which saves energy and enhances hydrolysis rate at the same time, and that hydrolysis must 22 be carried out rapidly after pretreatment to prevent natural drying.  3.75%, 4.5%, and 5.2% w/v) with a particle size between 0.5 and 1 mm, at 50 FPU/g to. 4 Basically, it can be noted that, in general, the higher the amount of hydrolyzed substrate, the 5 higher the amount of glucose produced, as expected. However, a surprising result was that the 6 conversion yield first increased rapidly when the substrate loading was increased from 1.5% to 7 3.75%, and then became nearly constant, independent from substrate loading (Fig. 2). This 8 means that the hydrolysis rate was nearly proportional to the initial substrate loading, even 9 though it decreased over time when substrate consumption proceeded. Thus, a cellulose-to-10 glucose conversion yield about 77 ± 3% could be reached when wet substrate loading was 11 between 3.75% and 5.2% (w/v). As the conversion yield was not significantly different for 12 3.75%, 4.5%, and 5.2% (w/v) substrate proportion, the consequence is that the glucose 13 concentration increased nearly proportionally to the substrate content up to 32 ± 1 g/L at 5.2% 14 (w/v) substrate loading, corresponding to a conversion yield of 80% ± 2% after 12 days of 15 hydrolysis. 16 Similar outcomes on the evolution of glucose concentration had been already reported by  18 investigated the effect of substrate loading during enzymatic hydrolysis using a pretreated 19 wheat straw by phosphoric acid and hydrogen peroxide. Four substrate loadings (2,10,15,20 and 20%) were tested, and it was pointed out that the highest substrate consistency led to the 21 highest glucose concentration in the final hydrolysate. However, the cellulose-to-glucose 22 conversion yield decreased slightly when the substrate loading increased. 23 Finally, contrary to expectations which predicted an increase in glucose concentration coupled 24 to a decrease in yield when substrate content was increased, it was found that the lowest substrate loading investigated, 1.5% (w/v), lead simultaneously to the lowest conversion yield 1 (57% ± 3%) and the lowest concentration of glucose (6.7 ± 0.3 g/L), whereas the best results 2 were achieved for the highest proportion of substrate (5.2% w/v) in terms of yield and glucose 3 content. Thus, by increasing pretreated substrate loading, higher concentrations of 4 fermentable sugars are available and then, higher ethanol concentration can be achieved. Due 5 to mixing issues resulting from the high viscosity of the cellulose suspension, experiments in 6 150 mL Erlenmeyer flasks had to be limited to 5.2% (w/v) substrate loading, but a higher 7 concentration will be tested in the scale-up assays in section 3.4, as better mixing condition 8 could be achieved.  10 To produce bioethanol, the upstream enzymatic hydrolysis process must provide, as much as  strain of Saccharomyces cerevisiae led us to reach an alcoholic fermentation yield close to 10 80% of the theoretical yield, which seems consistent with the absence of inhibiting 11 compounds produced during the pretreatment and hydrolysis steps.  13 The enzymatic hydrolysis scale-up was carried out in a 800 mL working volume under the 14 same conditions applied to the other pretreated substrate loadings, except that better mixing 15 conditions allowed to increase solid loading up to 7.5 % (w/v) pretreated sawdust mixture 16 substrate. This value is closer to the solids loading range proposed by Stenberg et al. (2000) 17 and Wingren et al. (2003). In these assays, the final glucose concentration was measured at 18 about 34.5 g/L, which represents a cellulose-to-glucose conversion yield of 59 ± 2% for up to 19 12 days (data not shown). This value is lower than in 150 mL flasks, which can probably be 20 explained by the high viscosity of the 7.5% (w/v) substrate suspension. The fermentation step 21 was then performed in a 500 mL bioreactor using the obtained hydrolysate enriched with 22 vitamins and minerals. The fermentation assays were carried out for 24 hrs.; the produced 23 ethanol and the decrease in glucose contents were monitored during the fermentation.

Scale-up of the enzymatic hydrolysis and fermentation processes
According to Fig. 4, approximately all the glucose was consumed after only 8 hours. The 1 produced ethanol in fermentation broth reached nearly 16 ± 2 g/L at the same time, which 2 represents almost 80% of the theoretical glucose-to-ethanol conversion yield. 3 According to the information provided by only for the kinetics of chemical steps. This highlights that more complex models are 4 probably required and that the respective influences of mixing conditions and dispersion 5 rheology must be studied for scale-up purpose. 6

7
The feasibility of lignocellulosic bioethanol production using a sawmill mixed feedstock of 8 softwood species was assessed. Enzymatic hydrolysis, applied to an undried microwave-9 assisted water/ethanol Organosolv pretreated substrate (71% w/w cellulose) with particle size 10 higher than 0.5 mm, reached 80% saccharification yield with up to 5.2% (w/v) substrate 11 loading. Higher loading decreased this yield, but higher ethanol concentration could be 12 achieved from subsequent fermentation using Saccharomyces cerevisiae, e.g. 16 g/L when 13 substrate loading was 7.5% (w/v). As an outcome, this process which tends to a closed-loop 14 biorefinery, is promising for regions where only mixed softwood feedstock is available. 15 Acknowledgments 16 This work was funded by Auvergne Rhône-Alpes Regional Council, and the European Regional 17 Development Fund (FEDER/ERDF) to promote the valorization of local feedstocks (wood- 18 wastes) on a regional scale.