Engineering the outcome of cofermentation processes by altering the feedstock sugar-toprotein ratio

Engineering the outcome of cofermentation processes by 1 altering the feedstock sugar-to-protein ratio 2 R. Bevilacqua*, M. Mauricio-Iglesias, S. Balboa, J.M. Lema, M. Carballa 3 CRETUS, Department of Chemical Engineering, Universidade de Santiago de Compostela, 15782 4 Santiago de Compostela, Spain – [email protected] 5 ABSTRACT 6 This work investigates the impact of the sugar-to-protein (STP) ratio on the outcome of their 7 anaerobic cofermentation in terms of substrate conversion and product selectivity. For this 8 purpose, a continuous stirred tank reactor was operated at pH 7 and fed with casein and glucose 9 at different STP ratios (0.25, 0.50, 0.75, 1.00 and 2.00 in COD basis). Casein conversion was 10 unaffected by glucose presence as long as the ratio was lower or equal to 1. In this range of STP 11 ratio, n-butyric and n-valeric acid production was promoted due to the occurrence and progressive 12 intensification of chain elongation processes. Conversely, STP ratios greater than 1 are associated 13 with lower amino acids consumption, inhibition of the elongation metabolism and lower volatile 14 fatty acids production due to the formation of alternative end products (ethanol, lactate and 15 formate) and unidentified compounds. Interestingly, these negative effects are reversible, as 16 lowering the sugar-to-protein ratio allows to recover protein acidification degree, process 17 productivity and the chain elongation. Overall, this work successfully demonstrates that sugar18 protein cofermentation processes can be steered by adjusting their proportions in the feedstock. 19 KEYWORDS: amino acids; biorefinery; chain elongation; feedstock composition; glucose; volatile 20 fatty acids 21 22 1 INTRODUCTION 23 Several studies1–5 highlighted the potential of mixing different substrates to enhance the 24 production of volatile fatty acids (VFAs) during mixed-culture fermentation (MCF) processes. The 25 positive effect observed during the cofermentation of proteic streams with those rich in sugars is 26 generally associated with a better balancing of micronutrients and carbon/nitrogen proportions, 27 dilution of potentially toxic or inhibitory compounds, and/or an increase in hydrolysis rate due to 28 the higher biomass yields achieved 6. 29 However, most literature examples dealing with anaerobic cofermentation are case studies 30 involving mixtures of specific waste and wastewaters, as in this example studying waste activated 31 sludge and potato peel waste5. Thus, the application of the resulting knowledge is limited to those 32 specific substrates and the conclusions are not valid for the conversion of a generic mixture of 33 proteins and carbohydrates into VFAs. Besides, all these previous results appear not to be 34 conclusive concerning the influence of mixing different organic fractions. For example, Breure et 35 al.9 observed that the presence of a sugar (e.g. glucose) can partially inhibit the hydrolysis of 36 proteins, and consequently their conversion into VFAs, when the two fractions loading were 37 similar, whereas lower sugar loads did not show negative effects on protein fermentation. 38 Conversely, Ma et al.5 determined that increasing the carbohydrate fraction in the feedstock 39 favours the consumption of proteins, with this synergistic effect being maintained even when 40 carbohydrates were dominant over proteins. In disagreement with the two previous results, 41 Tommaso et al.10 observed that even minimal glucose presence induces a decrease in the 42 degradation rate of the chosen model protein, bovine serum albumin. 43 Besides the conversion efficiency, sugars and proteins feature different VFA selectivity. The 44 fermentation of sugars (e.g. glucose) mainly yields acetic, propionic and butyric acid11 on 45 proportions that can be steered through pH adjustments12, similarly to lactose fermentation13. 46 Conversely, proteins selectivity heavily depends on their composition,14 given the potential 47 combination occurring from the mix of the 20 main amino acids (AAs). Acetic acid tends to be the 48 main product at neutral and alkaline conditions whereas low pH favours the conversion to longer 49 chain VFAs15. Moreover, branched chain VFAs and n-valeric acid are mostly obtained through the 50 fermentation of specific AAs rather than from sugars16. This substrate dependence suggests that 51 the VFA distribution of a cofermentation process could be steered based on the feeding 52 proportions between the two organic fractions. Yet, the literature concerning this effect is again 53 contradictory. For example, supplementing gelatin fermentation with either glucose or lactose in 54 equal proportions was associated to an increased production of n-butyric acid and ethanol9. 55 Instead, Zhou et al.17 observed an increase in acetic and propionic concentrations when 56 progressively feeding greater proportions of carbohydrate-rich corn straw to the sludge-degrading 57 reactor. In another case study, limiting the sugar fraction in the feeding mixture seemed to favour 58 the formation of n-valeric acid5. 59 The aforementioned information points out the need of a universal parameter to better 60 understand the interaction between proteins and carbohydrates in a cofermentation process in 61 order to engineer the process towards the desired outcome. Therefore, the present study 62 proposes the sugar-to-protein ratio (STP), measured in chemical oxygen demand (COD) basis, as 63 the parameter to assess and understand such interaction. The use of model protein and sugar 64 compounds (i.e. casein and glucose, respectively) aims at facilitating results interpretation as well 65 as their extrapolation to a generic protein-carbohydrate cofermentation process. 66 2 MATERIALS AND METHODS 67 2.1 Feedstock composition 68 Casein peptone (A2208,0500 PanReac) and D(+)-glucose anhydrous (131341.1211 PanReac) were 69 the model compounds used in this study. Protein concentration was fixed at 7.50 g/L throughout 70 the experiment, while glucose concentration was progressively increased from 1.87 g/L to 14.96 71 g/L. The feedstock solution was supplemented with macroand micro-nutrients, as described in 72 Bevilacqua et al.18 , and it was maintained refrigerated throughout the experiment (4°C). 73 2.2 Continuous reactors operation 74 The continuous stirred tank reactor (CSTR) of 1 L used in the present study was the same as 75 described in Bevilacqua et al. 18 The pH was set at 7.0 for the whole duration of the experiment, 76 while the reactor was maintained at 25 ºC through a temperature-controlled room. Being a CSTR, 77 the hydraulic retention time (HRT) and the solids retention time were both equal to 1.5 d. The 78 main difference between the two studies was glucose being included in the feedstock at increasing 79 concentrations in order to test several STP ratios (in COD basis): 0.25, 0.50, 0.75, 1.00, 2.00. Each 80 resulting STP ratio (Table 1) was maintained for at least 40 days, in order to evaluate its impact on 81 the cofermentation process after reaching a steady-state operation. 82 Table 1. Operational conditions of the different phases of the cofermentation reactor. STP: sugar83 to-protein ratio (COD basis); OLR: organic loading rate (g COD/L·d). 84 Phase STP ratio Casein OLR Glucose OLR I 0.25 5.33 1.33 II 0.50 5.33 2.67 III 0.75 5.33 4.00 IV 1.00 5.33 5.33 V 2.00 5.33 10.7 85 The reactor performance was monitored as described in Bevilacqua et al.18 . In brief, the pH was 86 controlled at the 7.0 setpoint via a multiparametric analyser (CHEMITEC, Italy) and NaOH 3M 87 additions. VFA and Total Ammonia Nitrogen (TAN) concentrations were determined twice a week, 88 while COD (total and soluble) and solids concentrations were measured once a week. Amino acid 89 (AA) analysis was performed on samples specifically selected from steady state periods of 90 operation. 91 2.3 Analytical methods 92 The analytical methods used are previously described in Bevilacqua et al. 15,18. A summary is 93 included in Supplementary Information. 94 2.4 Microbial community analyses 95 At each sugar-to-protein ratio, three biomass samples were taken, corresponding to three 96 consecutive weeks of stable operation. Genomic DNA from 1 mL homogenized samples was 97 extracted by triplicate using the Nucleospin Microbial DNA extraction kit (Machery-Nagel), 98 according to the instructions of the manufacturer. The replica from eachsample were pooled 99 together after quantification, ensuring quality control and normalization with Nanodrop and Qubit 100 fluorometer (Thermo Fisher Scientific Waltham, MA, USA). The V3-V4 hypervariable region for 101 Bacteria was amplified using Bakt_341F (5’ CCT ACG GGN GGC WGC AG 3’) and Bakt_805R (5’ GAC 102 TAC HVG GGT ATC TAA TCC 3’)19. DNA metabarcoding analyses of the region were carried out by 103 AllGenetics & Biology SL (www.allgenetics.eu) in an Illumina MiSeq platform. 104 Bioinformatic analyses were performed using the Microbial Genomics module (version 21.1) 105 workflow of the CLC Genomics workbench (version 21.0.3). Raw sequences were filtered to 106 remove low-quality reads and then clustered into Operational Taxonomic Units (OTUs) at 97% 107 cutoff for sequence similarity and classified against the non-redundant version SILVA SSU 108 reference taxonomy (release 132; http://www.arb-silva.de)20 . Only the most abundant bacterial 109 OTUs (above 1 % of the total observed OTUs) were considered for further analysis. 110 Microbial abundance from phyla to genus level was analyzed, log-transformed and the statistical 111 significance was determined for p < 0.05 by permutational multivariate analysis of variance 112 (PERMANOVA), including Bonferroni correction. Alpha diversity was estimated from the 113 rarefaction analysis using the resulting phylogenetic tree of OTUs generated by the MUSCLE 114 algorithm, with a maximum sampling depth to 26,618 reads. Beta diversity was measured by Bray115 Curtis distances between each pair of samples applying principal coordinate analysis (PCoA) to the 116 distance matrices. Significance was, likewise, assessed by PERMANOVA. 117 118 2.5 Calculations 119 Acidification degree was the parameter chosen to describe substrate conversion (in COD basis), 120 while ammonification was also used as a proxy to monitor protein conversion to VFA, as amino 121 acid fermentation is always related to NH4+ release. In addition, balances between AA 122 consumption and VFA production were established to verify protein conversion stoichiometry. 123 More details can be found in Supplementary Information. 124 3. RESULTS AND DISCUSSION 125 3.1 Cofermentation reactor operation 126 The cofermentation reactor was continuously operated for 344 days (Fig.1a). The first 56 days 127 were jointly considered as a phase of start-up and acclimation to glucose presence (1.33 g 128 COD/L·d), given that the inoculum was used to degrade only proteins during a previous 129 experiment18. To inhibit methanogenesis, which began to occur at day 42, sodium 2130 bromoethanesulphonate (BES, 137502, SigmaAldrich) was added to the reactor feedstock at a 131 concentration of 0.5 g/L starting from day 45. At day 344, the reactor was stopped due to Covid132 19 lockdown and restrictions on research activity and its content was stored at 4°C. The operation 133 was then resumed after two months (Fig. 1b) by acclimating the stored biomass at the original 134 conditions of pH, temperature and nitrogen sparging. The reactor was operated in batch mode for 135 the first 10 days by adding a diluted feedstock pulse to the vessel, in order to safely reactivate the 136 biomass activity. After having detected the occurrence of VFA production (Fig. 1d), continuous 137 feeding started at an hydraulic retention time (HRT) equal to 3 d (STP 1.00), to avoid potential 138 washout of the biomass. After one week it was lowered to 2 d, and finally set at the original value 139 of 1.5 d at day 24. On day 45, glucose concentration was increased to achieve the highest STP 140 value (2.00). The reactor operation was then finalised at day 88. 141 Biomass concentration rapidly grew from 0.6 to 1.0 g VSS/L when exposed at the lowest glucose 142 loading (STP 0.25), compatibly with the higher yields associated with sugar substrates21. Increasing 143 the STP ratio further favoured biomass growth, reaching 1.4 g VSS/L and 2.8 g VSS/L at STP ratios 144 of 1.00 and 2.00, respectively. 145 Methanisation was successfully inhibited from day 50 on, since no difference was detected 146 between the total COD concentrations in the reactor influent and effluent (Fig. 1a and b). The 147 difference between total and soluble COD in the effluents matched the biomass concentrations 148 achieved in the reactor. The overall concentration of VFA (COD basis) increased progressively with 149 the application of higher STP ratios, peaking at approximately 10 g COD/L (STP 1.00). VFA 150 production was 20% lower (8 g COD/L) after the reactor operation was resumed at the same 151 conditions (Fig. 1b), suggesting that the interruption and subsequent storage might have affected 152 the microbial population. Soluble COD concentration was systematically higher than the VFA-COD 153 concentration, suggesting the presence of non-converted substrate, alternative end products (e.g. 154 ethanol) and/or unidentified products. As glucose could not be detected in the reactor effluents, 155 only protein can account for the non-converted substrate. 156 As expected, global VFA production increased at higher STP ratios. However, the effect of STP ratio 157 on individual VFA production was acid-dependant (Fig. 1c and d). Acetic, n-butyric and n-valeric 158 acids were the main products for most of the reactor original operation (≥750 mg/L), progressively 159 increasing with the STP ratio. Interestingly, n-valeric acid production peaked at 1500 mg/L when 160 applying an STP value of 0.75, becoming the VFA with the highest concentration. Acetic acid 161 replaced it at STP 1.00, reaching 2000 mg/L. In comparison, n-butyric acid concentration grew 162 more steadily, stabilising at a final concentration of 1500 mg/L at STP 1.00. Conversely, iso-butyric 163 and iso-valeric acid production decreased from 330 to 250 mg/L and from 650 to 500 mg/L 164 respectively when applying an STP value greater than 0.25. n-Caproic acid was only detected for a 165 limited amount of time (STP 0.50) and only in small concentrations (≤ 150 mg/L). During the 166 resumed operation, the increase in STP ratio especially favoured acetic and propionic production 167 (≥2200 mg/L) in detriment of all the other VFAs, whose concentrations were equal or lower than 168 500 mg/L. Lactate, formate and ethanol production was not observed during the original 169 experiment and at variable concentrations during the resumed operation (data not shown). 170 To assess the impact of STP ratio on casein-glucose cofermentation, several steady-state periods 171 were identified: day 56 – 119, day 142 – 232, day 249 – 295 and day 312 – 344 for STP ratios of 172 0.25, 0.50, 0.75 and 1.00, respectively. For the resumed operation, the selected stable periods 173 were day 24 – 45 and day 52 – 88 for STP ratios of 1.00 and 2.00, respectively. 174 175 Figure 1. COD balance (a, original operation; b, resumed operation: ▲ Influent total COD; ● 176 Effluent total COD; □ Effluent soluble COD; ◇ VFAs COD) and individual VFA concentrations in the 177 cofermentation reactor (c, original operation; d, resumed operation: ● Acetic; ◆ Propionic; ▲ Iso178 Butyric; x n-Butyric; ⁕ Iso-Valeric; ■ n-Valeric). The vertical black lines indicate the change in the 179 STP ratio. 180 3.2 The influence of STP ratio on protein conversion and amino acid consumption 181 Glucose consumption was complete regardless of the STP ratio, while casein consumption was 182 above 60% based on the ammonification parameter, except for the STP ratio of 2.00 (Fig. 2). Given 183 3.4 The influence of STP on the microbial community structure 275 A total of 296,998 reads were obtained after trimming and quality filtering, ranging from 46,105 to 276 32,006 with an average of 36,512 reads per sample, identifying 466 different OTUs (Table S1). In 277 addition, the rarefaction curves obtained by the normalization of OTUs count for individual 278 biological replicates reached plateau (Fig. S1), pointing out an adequate sample sequencing depth 279 (26,618 reads). 280 Only OTUs with a minimum combined abundance of 1% were used for further analysis, resulting in 281 132 OTUs distributed in 16 classes among 10 phyla (Fig. 6). Firmicutes and Actinobacteria were the 282 dominant phyla in all the samples (above 65%), particularly at STP ratio of 1.0 (above 97%). Both 283 phyla, together with Proteobacteria and Bacteroidetes, are obligated or facultatively anaerobic 284 bacteria well known by their ability to decompose polysaccharides and proteinaceous substrates 285 to produce VFAs27. 286 287 Figure 6. Bacterial classes with a total abundance higher than 1% at different STP ratios. The 288 vertical black lines indicate the change in the STP ratio. Class abundances are colored according to 289 the phyla they belong to. 290 However, a more clear influence of increasing glucose loads was observed at class level (Fig. 6). 291 Overall, Bacteroidia abundance shows a decreasing trend, while the presence of Actinobacteria 292 and Erysipelotrichia is favored. Interestingly, the microbial community composition was similar up 293 to STP values of 0.75, but significant changes occurred when this ratio was increased to 1 and 294 further to 2. This pattern was also confirmed by Beta diversity analysis (Fig. 7), where all samples 295 belonging to STP ratios below 1.0 clustered together (moreover, PERMANOVA analyses showed no 296 significant differences among STP ratios of 0.25, 0.50 and 0.75 (pseudo-f statistic 1.99, 6.90, 1.3, p297 values > 0.1)) and separately from those belonging to STP ratios of 1 and 2, respectively. 298 299 Figure 7. Principal component analysis (PCoA) showing the differences on the community 300 composition related to increasing glucose loads. Points represent each sample and are coloured 301 according to the STP value: 0.25 (light green), 0.50 (medium green), 0.75 (dark green), 1.0 (blue) 302 and 2.0 (red). 303 Changing STP ratio from 0.75 to 1.0 lead to a very significant increase of Erysipelotrichia 304 abundance (Fig. 6) in detriment of Bacteroidia and Negativicutes (Fig. S2). In addition, a decrease 305 in Alpha diversity was observed (Fig. 8). Increasing further the STP ratio from 1.0 to 2.0 favored 306 the presence of Actinobacteria, Bacteroidia and Clostridia in detriment of Coriobacteriia (Fig. S2), 307 and also the overall diversity of the microbial community increases (Fig. 8). Combining these 308 results with the different product selectivities observed during the reactor operation (Fig. 4), we 309 could speculate the positive link between Actinobacteria and propionic acid production as well as 310 the link between Coriobacteriia and the production of longer chain VFA (butyric and valeric acids). 311 312 Figure 8. Summary of alpha diversity statistics shown as boxplot. A) Shannon entropy index; B) 313 Simpson index. Statistical significance was measured by Kruskal-Wallis test. 314 3.5 CE can be recovered by lowering the STP ratio 315 To verify whether the CE process could be recovered by lowering the glucose loading, a parallel 316 cofermentation reactor was inoculated with biomass taken from the main reactor on day 45 of the 317 resumed operation (STP 1.00) and operated at an STP ratio of 0.50 (Fig. 9). 318 319 320 Figure 9. Operation of the parallel reactor at an STP ratio of 0.50 to assess CE process recovery (a, 321 COD balance: ▲ Influent total COD; ● Effluent total COD; □ Effluent soluble COD; ◇ VFAs COD; b, 322 VFA concentrations: ● Acetic; ◆ Propionic; ▲ Iso-Butyric; x n-Butyric; ⁕ Iso-Valeric; ■ n-Valeric). 323 The vertical black lines separate the acclimation phase from the steady-state operation. 324 Both the total and the soluble COD of the reactor effluent decreased compatibly with the lower 325 STP applied to the reactor (Fig. 9a). Based on the VFA production (COD basis), it was possible to 326 identify two operational periods: from the start up to day 30 (acclimation stage) and from day 30 327 to 43 (steady-state operation). Interestingly, the values of all COD parameters were similar to 328 those previously obtained at STP 0.50 (Fig. 1a), providing the first proof concerning the 329 reversibility of excessive sugar supplementation. 330 In terms of VFA production (Fig. 9b), the acclimation period was associated with a decrease in 331 acetic and propionic acid concentrations, whereas the other VFAs remained mostly stable. In 332 contrast, except for propionic and iso-butyric acid, VFAs production increased between day 30 and 333 43. In particular, n-butyric, n-valeric acid and iso-valeric generation showed a two-fold increase 334 which, coupled with the absence of lactate and ethanol in the reactor effluents, further confirms 335 the reversibility of the effects caused by STP ratios greater than 1.00. Also, the balance of valeric 336 acids (Fig. 10) indicates that CE process was recovered during the steady-state period, as proline 337 consumption alone is not able to justify n-valeric acid production. This balance also highlights the 338 occurrence of isomerisation from the iso to the n-form during the acclimation step. 339 Comparing these results with those described in the previous sections, it was hypothesised that 340 the increased availability of glucose associated with higher STP ratios might be making further 341 conversion of ethanol and lactate into VFAs less appealing to the microbial community due to 342 kinetic limitations associated with high OLRs (16 g COD/L·d at STP ratio 2.00). Besides, the absence 343 of substrate limitations might be making specialised metabolic pathways, such as CE, less 344 appealing from a bioenergetics point of view. Still, the disruptive effect caused by the operation 345 interruption cannot be completely discarded, as it might have accelerated the disappearance of 346 the CE process at the highest STP ratios by altering the microbial community in first place. Besides, 347 the VFA concentrations were not strictly the same as during the original experiment at STP 0.50 348 (Fig. 1c), suggesting that longer operation time might be required to fully recover the previously 349 obtained steady state. 350 Figure 10. Iso and n-valeric acid balance in the parallel reactor at an STP ratio of 0.50: ■ Proline; ■ 351 n-valeric acid; ■ Isoleucine; ■ Leucine; ■ Iso-valeric acid. AA concentrations are expressed as 352 VFA equivalents according to the stoichiometry described by Regueira et al.16 353 4. CONCLUSIONS 354 This study successfully investigated the interactions between amino acids and glucose during their 355 cofermentation in order to understand the impact of the STP ratio on substrate consumption, 356 acidification degree, product selectivity and microbial community structure. In particular, the main 357 findings are: 358 • STP ratios equal or lower than 1.00 do not affect the extent of protein conversion, but 359 excessive sugar loading hinders AA consumption and favours the production of alternative 360 end products. 361 • The products distribution can be steered towards the production of n-butyric and n-valeric 362 acid by increasing the sugar proportion up to the optimal STP ratio of 1.00, which promotes 363 the occurrence of CE processes. 364 • The increasing load of glucose affected microbial community composition, especially at the 365 highest STP ratio tested (2.0). Overall, the presence of Actinobacteria and Erysipelotrichia 366 was favored in detriment of Bacteroidia abundance. 367 • The changes produced by excessive sugar loadings are reversible, as lowering the STP ratio 368 allows to recover the longer chain VFA production to a certain extent. 369 CONFLICTS OF INTEREST 370 There are no conflicts of interest to declare. 371 ACKNOWLEDGEMENTS 372 This project has received funding from the European Union’s ERA-IB programme under grant 373 agreement number PCIN-2016-102 (BIOCHEM project). The authors belong to a Galician 374 Competitive Research Group (GRC), co-funded by ERDF (UE). 375 REFERENCES 376 1 H. Rughoonundun, R. Mohee and M. T. Holtzapple, Influence of carbon-to-nitrogen ratio on 377 the mixed-acid fermentation of wastewater sludge and pretreated bagasse, Bioresour. 378 Technol., 2012, 112, 91–97. 379 2 Á. Val Del Río, T. Palmeiro-Sanchez, M. Figueroa, A. Mosquera-Corral, J. L. Campos and R. 380 Méndez, Anaerobic digestion of aerobic granular biomass: effects of thermal pre-treatment 381 and addition of primary sludge, J. Chem. Technol. Biotechnol., 2014, 89, 690–697. 382 3 A. F. Duque, C. S. S. Oliveira, I. T. D. Carmo, A. R. Gouveia, F. Pardelha, A. M. Ramos and M. 383 A. M. Reis, Response of a three-stage process for PHA production by mixed microbial 384 cultures to feedstock shift: impact on polymer composition, N. Biotechnol., 2014, 31, 276– 385 288. 386 4 Z. Guo, A. Zhou, C. Yang, B. Liang, T. Sangeetha, Z. He, L. Wang, W. Cai, A. Wang and W. Liu, 387 Enhanced short chain fatty acids production from waste activated sludge conditioning with 388 typical agricultural residues: carbon source composition regulates community functions, 389 Biotechnol. Biofuels, 2015, 8, 192. 390 5 H. Ma, H. Liu, L. Zhang, M. Yang, B. Fu and H. Liu, Novel insight into the relationship 391 between organic substrate composition and volatile fatty acids distribution in acidogenic co392 fermentation, Biotechnol. Biofuels, 2017, 10, 137. 393 6 W. Fang, X. Zhang, P. Zhang, J. Wan, H. Guo, D. S. M. Ghasimi, X. C. Morera and T. Zhang, 394 Overview of key operation factors and strategies for improving fermentative volatile fatty 395 acid production and product regulation from sewage sludge, J. Environ. Sci., 2020, 87, 93– 396 111. 397 7 P. Kehrein, M. van Loosdrecht, P. Osseweijer, M. Garfí, J. Dewulf and J. Posada, A critical 398 review of resource recovery from municipal wastewater treatment plants – market supply 399 potentials, technologies and bottlenecks, Environ. Sci. Water Res. Technol., 2020, 6, 877– 400 910. 401 8 U. Jayakrishnan, D. Deka and G. Das, Enhancing the volatile fatty acid production from agro402 industrial waste streams through sludge pretreatment, Environ. Sci. Water Res. Technol., 403 2019, 5, 334–345. 404 9 A. M. Breure, K. A. Mooijman and J. G. van Andel, Protein degradation in anaerobic 405 digestion: influence of volatile fatty acids and carbohydrates on hydrolysis and acidogenic 406 fermentation of gelatin, Appl. Microbiol. Biotechnol., 1986, 24, 426–431. 407 10 G. Tommaso, R. Ribeiro, M. B. A. Varesche, M. Zaiat and E. Foresti, Influence of multiple 408 substrates on anaerobic protein degradation in a packed-bed bioreactor, Water Sci. 409 Technol., 2003, 48, 23–31. 410 11 R. González-Cabaleiro, J. M. Lema and J. Rodríguez, Metabolic Energy-Based Modelling 411 Explains Product Yielding in Anaerobic Mixed Culture Fermentations, PLoS One, 2015, 10, 1– 412 17. 413 12 M. F. Temudo, R. Kleerebezem and M. C. M. van Loosdrecht, Influence of the pH on (open) 414 mixed culture fermentation of glucose: A chemostat study, Biotechnol. Bioeng., 2007, 98, 415 69. 416 13 A. R. Gouveia, E. B. Freitas, C. F. Galinha, G. Carvalho, A. F. Duque and M. A. M. Reis, 417 Dynamic change of pH in acidogenic fermentation of cheese whey towards 418 polyhydroxyalkanoates production: Impact on performance and microbial population, N. 419 Biotechnol., 2017, 37, 108–116. 420 14 R. Bevilacqua, A. Regueira, M. Mauricio-Iglesias, J. M. Lema and M. Carballa, Protein 421 composition determines the preferential consumption of amino acids during anaerobic 422 mixed-culture fermentation, Water Res., 2020, 183, 115958. 423 15 R. Bevilacqua, A. Regueira, M. Mauricio-Iglesias, J. M. Lema and M. Carballa, Steering the 424 conversion of protein residues to volatile fatty acids by adjusting pH, Bioresour. Technol., 425 2021, 320, 124315. 426 16 A. Regueira, J. M. Lema, M. Carballa and M. Mauricio-Iglesias, Metabolic modeling for 427 predicting VFA production from protein-rich substrates by mixed-culture fermentation, 428 Biotechnol. Bioeng., 2020, 117, 73–84. 429 17 A. Zhou, Z. Guo, C. Yang, F. Kong, W. Liu and A. Wang, Volatile fatty acids productivity by 430 anaerobic co-digesting waste activated sludge and corn straw: Effect of feedstock 431 proportion, J. Biotechnol., 2013, 168, 234–239. 432 18 R. Bevilacqua, A. Regueira, M. Mauricio-Iglesias, J. M. Lema and M. Carballa, Understanding 433 the effect of trace elements supplementation on volatile fatty acids production from 434 proteins, J. Environ. Chem. Eng., , DOI:10.1016/j.jece.2021.105934. 435 19 D. P. R. Herlemann, M. Labrenz, K. Jürgens, S. Bertilsson, J. J. Waniek and A. F. Andersson, 436 Transitions in bacterial communities along the 2000 km salinity gradient of the Baltic Sea, 437 ISME J., 2011, 5, 1571–1579. 438 20 C. Quast, E. Pruesse, P. Yilmaz, J. Gerken, T. Schweer, P. Yarza, J. Peplies and F. O. Glöckner, 439 The SILVA ribosomal RNA gene database project: improved data processing and web-based 440 tools, Nucleic Acids Res., 2013, 41, D590–D596. 441 21 G. Tchobanoglous, Laurie M. Franklin, E. Burton, H. D. Stensel. Metcalf & Eddy, Wastewater 442 engineering : treatment and reuse, Fourth edition / revised by George Tchobanoglous, 443 Franklin L. Burton, H. David Stensel. Boston : McGraw-Hill, [2003] ©2003, 2014. 444 22 A. Aguilar, C. Casas, J. Lafuente and J. M. Lema, Kinetic modelling of isomerization and 445 anaerobic degradation of nand i-butyrate, J. Ferment. Bioeng., 1990, 69, 261–264. 446 23 A. Regueira, R. Bevilacqua, J. M. Lema, M. Carballa and M. Mauricio-Iglesias, A metabolic 447 model for targeted volatile fatty acids production by cofermentation of carbohydrates and 448 proteins, Bioresour. Technol., 2020, 298, 122535. 449 24 A. Regueira, J. L. Rombouts, S. A. Wahl, M. Mauricio-Iglesias, J. M. Lema and R. 450 Kleerebezem, Resource allocation explains lactic acid production in mixed-culture anaerobic 451 fermentations, Biotechnol. Bioeng., 2021, 118, 745–758. 452 25 L. T. Angenent, H. Richter, W. Buckel, C. M. Spirito, K. J. J. Steinbusch, C. M. Plugge, D. P. B. 453 T. B. Strik, T. I. M. Grootscholten, C. J. N. Buisman and H. V. M. Hamelers, Chain Elongation 454 with Reactor Microbiomes: Open-Culture Biotechnology to Produce Biochemicals, Environ. 455 Sci. Technol., 2016, 50, 2796–2810. 456 26 S. Liang and C. Wan, Carboxylic acid production from Brewer’s spent grain via mixed culture 457 fermentation, Bioresour. Technol., 2015, 182, 179–183. 458 27 L. Levén, A. R. B. Eriksson and A. Schnürer, Effect of process temperature on bacterial and 459