An open-source multiple-bioreactor system for replicable gas-fermentation experiments: Nitrate feed results in stochastic inhibition events, but improves ethanol production of Clostridium ljungdahlii with CO2 and H2

The pH-value in fermentation broth has a large impact on the metabolic flux and growth behavior of acetogens. A decreasing pH level throughout time due to undissociated acetic acid accumulation is anticipated under uncontrolled pH conditions such as in bottle experiment. As a result, the impact of changes in the metabolism (e.g., due to a genetic modification) might remain unclear or even unrevealed. In contrast, pH-controlled conditions can be easily achieved in commercially available bioreactors. However, their acquisition is costly and their operation is time consuming, and therefore the experiment is often limited to a single bioreactor run. Here, we present a self-built, relatively cheap, and easy to handle open-source multiple-bioreactor system (MBS) consisting of six pH-controlled bioreactors at a 1-L scale. The functionality of the MBS was tested in three experiments by cultivating the acetogen Clostridium ljungdahlii with CO2 and H2 at steady-state conditions (=chemostat). The experiments were addressing the questions: (1) does the MBS provide replicable data for gas-fermentation experiments?; (2) does feeding acetate alter the production rate of ethanol; and (3) does feeding nitrate influence the product spectrum under controlled pH conditions with CO2 and H2? We applied four different periods in each experiment ranging from pH 6.0 to pH 4.5. Our data show high reproducibility for gas-fermentation experiments with C. ljungdahlii, using the MBS. We found that feeding acetate did not improve ethanol production, but rather impaired growth and reduced acetate production. Using nitrate as sole N-source, on the other hand, enhanced biomass production even at a low pH. However, we observed differences in growth, acetate, and ethanol production rates between triplicate bioreactors (n=3). We explained the different performances because of stochastic inhibition events, which we observed through the accumulation of nitrite, and which led to complete crashes at different operating times. One of these bioreactors recovered after the crash and showed enhanced ethanol production rates while simultaneously producing less acetate. The MBS offers a great opportunity to perform bench-scale bioreactor experiments at steady-state conditions with replicates, which is especially attractive for academia.


Introduction 40
An increasing world population will likely lead to growing energy demands. To meet these demands 41 in a sustainable way, we need to rethink the status quo of a fossil-based economy and transition into 42 a renewable-based and circular economy. Furthermore, we have to mitigate the apparent climate 43 effects of anthropogenic greenhouse gas emissions, such as carbon dioxide (CO 2 ), which are caused 44 preliminary by industry, agriculture, and transportation. Biotechnology offers potential to contribute 45 to climate-friendly and economically feasible solutions. One promising solution is synthesis gas 46 (syngas) fermentation with microbes (Mohammadi et al., 2011). For syngas fermentation, mixtures of 47 the gases CO 2 , hydrogen (H 2 ), and carbon monoxide (CO) are converted into products, such as 48 acetate and ethanol, by acetogenic bacteria (Dürre, 2017). This process provides a promising way to 49 produce chemicals and biofuels with a reduced CO 2 -footprint ( the impact of cultivation parameters can be investigated within one study, these studies often are 68 difficult to compare with each other, because very different bioreactor architectures and process 69 parameters are used (Asimakopoulos et al., 2018). Furthermore, because of the complexity, these 70 setups are not suitable to perform preliminary experiments with genetically engineered strains. These 71 issues can be partly overcome by utilizing commercially available bioreactor (chemostat) systems. 72 However, these systems are costly, and therefore often not available to laboratories that do not focus 73 on bioprocess engineering. Consequently, genetically engineered strains are typically not studied in 74 fermentations beyond the serum bottle size, which leaves a gap between the construction of these 75 strains and the investigation under controlled fermentation conditions. 76 To close this gap, we developed a cost-efficient, open-source multiple-bioreactor system (MBS) that 77 can be built from off-the-shelf components at a considerably lower cost compared to the cost of 78 commercial bioreactor systems. We give all information on purchasing the required parts, the 79 assembly of the MBS, the process control elements (e.g., stirring, pH, temperature), and further 80 improvement ideas. We tested our MBS with C. ljungdahlii and CO 2 and H 2 as substrate under 81 controlled pH conditions by addressing three questions: (1) does the MBS provide replicable data for 82 gas-fermentation experiments; (2) does feeding acetate alter the production rate of ethanol; and (3) 83 does feeding nitrate influence the product spectrum? 84 We selected the first question to test the reproducibility of our system, while the latter two questions 85 address the physiology of the microbe. Acetogens such as C. ljungdahlii are known to be highly 86 dependent on the pH in the fermentation broth. Since their main fermentation product is acetate, 87 which acts (in the form of the undissociated acetic acid) as a weak acid, no pH control in serum 88 bottles would intrinsically lower the pH of the medium during growth, but can be controlled in 89 bioreactors such as in our MBS (Drake et al., 2008). 90 Our second question addresses acetate as a fermentation product directly. It has been demonstrated 91 that feeding acetate from the first stage to the second stage in a two-stage bioreactor system with 92 syngas including CO led to increased ethanol production rates with C. ljungdahlii (Richter et al., 93 2013). Others have shown that feeding acetate to a single bioreactor with CO 2 and H 2 also led to 94 increased ethanol production rates with C. autoethanogenum (Mock et al., 2015). In both cases the 95 pH was an important parameter because ethanol production is thermodynamically triggered at a low 96 pH . 97 Our third question addresses the finding from a recent study in which nitrate was used as an 98 alternative electron acceptor by C. ljungdahlii, while it also served as sole nitrogen source (N-source) 99 in ammonium-free medium (Emerson et al., 2019). The co-utilization of CO 2 and nitrate enhanced 100 the autotrophic biomass formation with CO 2 and H 2 compared to standard cultivation conditions with 101 ammonium as the sole N-source. Contrarily, ethanol production was strongly reduced under nitrate 102 conditions with CO 2 and H 2 . The authors discussed that nitrate reduction consumes electrons, which 103 would be no longer available for the production of ethanol. At the same time, nitrate reduction led to 104 an accumulation of ammonium, which resulted in a continuous increase of the pH from 6.0 to 8.0 105 during the batch cultivations in serum bottles (Emerson et al., 2019), which would intrinsically 106 prevent ethanol production. These two questions on the physiology address growth conditions in 107 which the medium pH has a considerable impact on the production of biomass and fermentation 108 products, such as acetate and ethanol, and therefore are perfectly suited to demonstrate the 109 functionality of our MBS. 110 The pH of the TE was adjusted to 6.0 after adding NTA. The solution was autoclaved and stored at 124

Materials and Methods
4°C. Wolfe's vitamin solution was prepared aerobically containing (per liter): 2 mg biotin; 2 mg folic 125 acid; 10 mg pyridoxine-hydrochloride; 5 mg thiamin-HCl; 5 mg riboflavin; 5 mg nicotinic acid; 5 mg 126 calcium pantothenate; 5 mg p-aminobenzoic acid; 5 mg lipoic acid; and 0.1 mg cobalamin. The 127 vitamin solution was sterilized using a sterile filter (0.2 µm), sparged with N 2 through a sterile filter, 128 and stored at 4°C. The 50x fructose/MES solution contained (per 100 mL): 25 g fructose; and 10 g 129 MES. The pH was adjusted to 6.0 by adding KOH. The solution was sterilized, sparged with N 2 130 through a sterile filter, and stored at room temperature. The reducing agent was prepared under 100% 131 N 2 in a glove box (UniLab Pro Eco, MBraun, Germany) and contained (per 100 mL): 0.9 g NaOH; 4 132 g cysteine-HCl; and 2.17 g/L Na 2 S (60 weight%). Anaerobic water was used for the preparation of 133 the reducing agent. The reducing agent was autoclaved and stored at 4°C. 134 For all bioreactor experiments, the standard PETC medium for the initial batch phase was 135 supplemented with 0.5 g L -1 yeast extract and autoclaved inside the bioreactor vessel with an open 136 off-gas line to enable pressure balance. The autoclaved bioreactors were slowly cooled down at room 137 temperature overnight with an attached sterile filter at the off-gas line. After transferring each 138 bioreactor to the MBS frame, the medium was continuously sparged with a gas mixture of CO 2 and 139 H 2 (20:80 vol%) through a sterile filter. After 1 h, vitamins and reducing agent were added through 140 the sampling port. N 2 gas was applied through a sterile filter to flush the sampling port after each 141 addition of media components. Subsequently, each bioreactor was inoculated with 5 mL of an 142 exponential heterotrophically grown PETC culture (OD 600 0.5-0.8). Feed bottles containing 4 L of 143 PETC medium with additions, as described below, were prepared for continuous mode, autoclaved, 144 and stored overnight with an attached sterile filter on the off-gas line. The bottles were sparged with 145 N 2 for 2 h through a sterile filter. Vitamins and reducing agents were added under sterile conditions. 146 A gas bag with N 2 gas was attached with a sterile filter to balance the pressure in the feed bottle 147 during the bioreactor run. Standard PETC medium for continuous mode did not contain yeast extract 148 and was adjusted to the respective pH of the period. For the first experiment we used standard PETC 149 medium. For the second experiment, we added 100 mM Na-acetate or 100 mM NaCl (to give a 150 similar ion strength). For the third experiment, we supplied no NH 4 Cl, but we supplied the equivalent 151 amount of nitrogen as NaNO 3 (18.7 mM). 152

Bioreactor setup and standard operating conditions 153
Six 1-L self-built bioreactors (Fig. 1, Results S1) with a working volume of 0.5 L were used for the 154 three experimental bioreactor runs (Fig. 2, Fig. 3, Fig. S4). The cultivation temperature was 37°C 155 and the agitation was set to 300 rpm. The gas flow rate was adjusted to 30 mL min -1 prior to 156 inoculation. For continuous-mode operating conditions, the medium feed rate was measured to be 157 0.10 mL min -1 , which resulted in a 3.5-day hydraulic retention time (HRT), and which represents 1.7 158 HRT periods within each period of six days. To establish microbial growth in the MBS after one 159 inoculation event for each bioreactor, we operated the MBS in batch mode for three days before 160 switching to continuous mode. The pH was set to 6.0 during the batch mode and the first 6 days 161 (Period I) in continuous mode. Subsequently, the pH setting was lowered stepwise in 6 days to a pH 162 of 5.5, 5.0, and 4.5 (Period II-IV). While the pH of the feed medium was adjusted to the anticipated 163 pH of the period, we did not use the acid feed to actively adjust the pH within the bioreactor. Instead, 164 we let the pH decrease to the set value by the microbial production of undissociated acetic acid to 165 avoid a pH shock. This took approximately two days of each 6-day period ( Table 1). For the first 166 experimental bioreactor run (Fig. 2), only base pumps were connected to the bioreactors. For the 167 second experimental bioreactor run (Fig. 3), acid and base pumps were connected to the bioreactors 168 with nitrate feed and only base pumps to the bioreactors with ammonium feed. 169 170

Sampling and analyses
Bioreactors were sampled once or twice per day. A pre-sample of 3 mL of cell suspension was 172 discarded, before transferring 2 mL of cell suspension into 2-mL reaction tubes (main sample). The 173 multi-channel pump was switched off during the sampling procedure. Cell growth was monitored by 174 measuring the optical density at 600 nm (OD 600 ) (Nanophotometer NP80, Implen, Germany). For 175 OD 600 -values larger than 0.5, dilutions with 100 mM phosphate-buffered saline (PBS) at pH=7.4 176 were prepared. Nitrate and nitrite concentrations were qualitatively monitored using test stripes 177 (Quantofix nitrate/nitrite, Macherey-Nagel, Germany). A correlation between cellular dry weight 178 (CDW) and OD 600 was calculated by harvesting 50 mL of culture sample from every bioreactor, 179 centrifugation of the samples at 3428 relative centrifugal force (rcf) (Eppendorf centrifuge 5920R) 180 for 12 min at room temperature (RT) and, subsequently, drying the pellet at 65°C for 3 days. The 181 CDW for an OD 600 of 1 was determined to be 0.24 g L -1 for cultures grown in PETC medium and 182 0.29 g L -1 for cultures grown in PETC medium with nitrate instead of ammonium. 183 Acetate and ethanol concentrations were analyzed via a high pressure liquid chromatography (HPLC) 184 (LC20, Shimadzu, Japan) system that was equipped with an Aminex HPX-87H column and operated 185 with 5 mM sulfuric acid as eluent. The flow was 0.6 mL min -1 (LC-20AD). The oven temperature 186 was 65°C (CTO-20AC). The sample rack of the HPLC was constantly cooled to 15°C in the 187 autosampler unit (SIL-20AC HT ). For HPLC sample preparation, all culture samples were centrifuged 188 for 3 min at 15871 rcf (Centrifuge 5424, Eppendorf, Germany) in 1.5-mL reaction tubes. 750 µl of 189 the supernatant was transferred into clean reaction tubes and stored at -20°C until use. Frozen 190 samples were thawed at 30°C and 250 revolutions per minute (rpm) for 10 min (Thermomixer C, 191 Eppendorf, Germany). The samples were centrifuged again and 500 µl of the supernatant was 192 transferred into short thread HPLC/GC vials (glass vial ND9, VWR, Germany) and sealed with short 193 screw caps, which contained rubber septa (6 mm for ND9, VWR, Germany). New standards for 194 acetate and ethanol were prepared for every analysis (retention time of acetate was 14.8 min and 195 retention time of ethanol was 21.5 min). All samples were randomized. 196

Results 197
Designing and testing the functionality of the MBS 198 We based our experiments in this study on a versatile self-built multiple-bioreactor system (MBS). 199 The MBS (Fig. 1, Results S1, Fig. S1 and S2, To show high comparability and reproducibility of our MBS, we grew C. ljungdahlii simultaneously 204 as triplicates with standard PETC medium and CO 2 and H 2 (ammonium, bioreactors 1/2/3) during the 205 first experiment (Fig. 2, Fig. S3). We observed that growth was similar in the triplicate bioreactors 206 during the cultivation for 27.6 days. During the initial batch mode, the average OD 600 increased to 207 0.58 ± 0.01 ( Fig. 2A). After switching to continuous mode, the average OD 600 increased further to 208 values of 0.82 ± 0.04 during Period I. For Periods II, III, and IV, the average OD 600 for the 209 bioreactors constantly decreased to values of 0.69 ± 0.01, 0.51 ± 0.05, and 0.18 ± 0.06 ( Fig. 2A, 210 Table 1). As expected, the pH of each bioreactor was decreasing during all periods by microbial 211 acetate production. The simultaneous and constant decrease of OD 600 indicated reduced growth rates 212 of C. ljungdahlii at a lower pH level in our system. In batch mode, the acetate production rates 213 increased with increasing OD 600 , but then considerably dropped after switching to continuous mode 214 ( Fig. 2A). The acetate production rates increased again to an average value of 73.1 ± 2.1 mmol-C L -1 215 d -1 for Period I (Fig. 2B). The acetate production rates decreased to average values of 60.1 ± 3.4 216 mmol-C L -1 d -1 , and 53.7 ± 3.3 mmol-C L -1 d -1 for Periods II and III, respectively. For Period IV, the 217 acetate production rate had only an average value of 29.2 ± 10.0 mmol-C L -1 d -1 (Fig. 2B). Ethanol 218 production rates in mmol-C L -1 d -1 were negligible during batch mode, but slowly increased after 219 switching to continuous mode. The highest ethanol production rates were observed for Period II with 220 average values of 22.5 ± 0.5 mmol-C L -1 d -1 . During the Periods III and IV, the ethanol production 221 rates kept decreasing to average values of 18.7 ± 4.8 mmol-C L -1 d -1 and 3.4 ± 1.4 mmol-C L -1 d -1 , 222 respectively (Fig. 2B). The results showed high reproducibility with small standard deviation for all 223 tested parameters using the MBS. 224 Feeding additional acetate to continuous and pH-controlled gas fermentation of C. ljungdahlii 225 with CO 2 and H 2 226 During the second experiment, we investigated the impact of feeding acetate on the production of 227 ethanol from CO 2 and H 2 during an operating period of 27.2 days (Results S2, Fig. S4). For this, 228 three bioreactors (Na-acetate, bioreactors 4/5/6) were fed with PETC medium that was supplemented 229 with 100 mM Na-acetate, while three bioreactors (NaCl, bioreactors 7/8/9) were fed with PETC 230 medium that was supplemented with 100 mM NaCl to achieve a similar ionic strength. We found 231 again that our triplicate bioreactors were reproducible in terms of growth. However, feeding acetate 232 did not result in the expected increased ethanol production rates. Instead, the additional acetate led to 233 impaired growth, lower OD 600 -values, and reduced production rates of acetate and ethanol (Fig. S4, 234 Table S2). However, feeding additional NaCl compared to our first experiment with standard PETC 235 medium had a positive stimulating impact on ethanol production rates at lower pH ( Fig. S5 and S7, 236 Table S2). Growth (OD 600 ) and acetate production were not influenced by the higher NaCl content in 237 the medium and showed similarities to our preliminary experiment with standard PETC medium 238 (Fig. 2, Table 1). The second experiment shows that acetate did not, but that NaCl did improve 239 ethanol production of C. ljungdahlii with CO 2 and H 2 . 240 Applying nitrate as a sole N-source to a continuous and pH-controlled gas fermentation of 241 C. ljungdahlii with CO 2 and H 2 242 During our third experiment with an operating period of 27.5 days, we investigated the impact of 243 nitrate as an alternative N-source on the production of ethanol from CO 2 and H 2 (Fig. 3). For this 244 experiment, three bioreactors (nitrate, bioreactors 10/11/12) were fed with PETC medium containing 245 nitrate instead of ammonium at an equivalent molar amount of nitrogen (=18.7 mM). We found an 246 increasing pH due to ammonium production in preliminary bottle experiments in nitrate-containing 247 PETC medium (Fig. S7). A pH increase was also observed in the nitrate bottle experiments of 248 Emerson et al. (2019). Despite the pH-control in our experiment, all bioreactors with nitrate feed 249 showed remarkable differences in growth, pH, acetate production, and ethanol production rates. 250 Therefore, we report individual data for each bioreactor and highlight lowest and highest values (Fig.  251 3, Table 2). We use the data of the first experiment (ammonium, bioreactor 1/2/3) as the control in 252 which ammonium served as the sole N-source (Fig. 2, Table 1). Unexpectedly, we observed a pH-253 buffering effect in all bioreactors with nitrate feed during the fermentation. This was most likely due 254 to an interplay between the produced acetate and ammonium by the microbes. Overall, the pH was 255 slowly decreasing in all bioreactors with nitrate feed (Fig. 3A), and we did not measure increasing 256 pH values. 257 During the initial batch mode, two of the bioreactors with nitrate feed (bioreactor 10 and 11) reached 258 OD 600 -values of 0.58, while one of the bioreactors with nitrate feed (bioreactor 12) stagnated after two days of cultivation in batch mode and reached an OD 600 of 0.23 (Fig. 3A). After switching to 260 continuous mode, all three bioreactors with nitrate feed reached similar OD 600 -values of ~1.2 during 261 the end of Period I, which were 25-30% higher compared to the OD 600 of the bioreactors with 262 ammonium feed (Table 1, Fig. 2, Fig. 3A). The highest observed OD 600 for the bioreactors with 263 nitrate feed were 1.29 on day 21 during Period IV for bioreactor 10, 1.36 on day 9 during Period II 264 for bioreactor 11, and 1.34 on day 7 during Period I for bioreactor 12 (Fig. 3, Table 2). In 265 comparison, the bioreactors with ammonium feed had the highest average OD 600 -value of 0.90 ± 0.02 266 on day 4 for Period I (Fig. 2, Table 1). 267 Another noticeable difference was that the OD 600 -values did not remain stable for all bioreactors with 268 nitrate feed during the experiment. Each bioreactor showed fluctuating values at higher OD 600 before 269 it crashed at different time points during the operating period. For instance, bioreactor 10 underwent 270 a crash in biomass growth on day 8 which continued for the rest of Period II with continuously 271 decreasing OD 600 (Fig. 3A). At this point, the pH stagnated at a pH of 5.8 for bioreactor 10. 272 However, after switching to Period III, bioreactor 10 recovered in OD 600 back to a value of 1.30 and 273 reached a pH of 5.0 on day 17 of the operating period. During Period IV the OD 600 was stable for 274 bioreactor 10 (Fig. 3A). On day 11 of the operating period, bioreactor 11 underwent a similar crash 275 in biomass growth but, contrarily, did not recover back to high OD 600 -values until the end of the 276 operating period. The pH of this bioreactor 11 remained stable at pH 5.5 from day 11 of the operating 277 period independent of the following periods (Fig. 3A). A similar behavior was observed for 278 bioreactor 12, but the abrupt crash in OD 600 occurred on day 21 during Period III. Bioreactor 12 also 279 did not recover from the crash until the end of the operating period (Fig. 3A). It is noteworthy, that 280 we detected nitrate and nitrite in culture samples of bioreactors with nitrate feed undergoing a crash, 281 while neither nitrate nor nitrite were detectable in actively growing or recovering bioreactors with 282 nitrate feed. This indicates a high uptake rate for nitrate by the microbes from the feed medium, and 283 an immediate conversion of the nitrate to ammonium via nitrite as an intermediate. 284 The acetate production rates of all bioreactors with nitrate feed somewhat followed the OD 600 profile, 285 and reached the highest values that we observed in all our experiments with a maximum value of 129 286 mmol-C L -1 d -1 for bioreactor 10 during Period I (Fig. 3B). The acetate production rate considerably 287 decreased at the time point of the OD 600 crashes for the three bioreactors with nitrate feed (Fig. 3B). 288 Ethanol production rates were negligible during batch mode for all bioreactors with nitrate feed and 289 slowly increased with decreasing pH during the different periods, after switching to continuous 290 mode, and also considerably dropped after the OD 600 crashes (Fig. 3B). We observed high ethanol 291 production rates in each of the three bioreactors with nitrate feed before the respective OD 600 crashes 292 with values of 62.0 mmol-C L -1 d -1 for bioreactor 10, 29.9 mmol-C L -1 d -1 for bioreactor 11, and 30.6 293 mmol-C L -1 d -1 for bioreactor 12 (Table 2, Fig. 3). While for bioreactors 11 and 12 the ethanol 294 production rates did not recover after the crashes, for bioreactor 10 the ethanol production rate 295 increased with increasing OD 600 after the crash, and reached its maximum during Period IV, which is 296 the highest value for the entire study (Fig. 3B, Table 2). This value is ~2.5-fold higher compared to 297 the highest ethanol production rate observed for C. ljungdahlii growing with ammonium (Fig. 2,  298 Table 2). Acetate production rates, on the other hand, remained low after the recovery of bioreactor 299 10, which led to the highest measured ethanol/acetate ratio of ~3.8 with CO 2 and H 2 in this study. To 300 our knowledge it is also the highest ethanol/acetate ratio for published studies with acetogens and 301 CO 2 and H 2 , because Mock et al. (2019) achieved a ratio of ~1:1. We found that all three bioreactors 302 with nitrate feed behaved differently and underwent stochastic crashes in the OD 600 at different time 303 points that were most likely connected to a simultaneous accumulation of nitrite. The recovery of one 304 bioreactor from this crash influenced the ethanol and acetate production rates. 305 Finally, to further test the impact of pH changes on the bioreactors with nitrate feed, we manually 306 adjusted the pH by feeding acid to all bioreactors with nitrate feed to decrease and then keep the pH 307 at pH 4.5 on day 24 of the operating period (Fig. 3, indicated with a star symbol). This intervention 308 immediately resulted in a second crash of the OD 600 and production rates for bioreactor 10, while 309 bioreactors 11 and 12 already were at low OD 600 -values with low acetate and ethanol production rates 310 at that time point. 311

Discussion 312
Our MBS resulted in reproducible gas-fermentation experiments with C. ljungdahlii 313 The MBS was successfully tested to cultivate C. ljungdahlii with CO 2 and H 2 in triplicates under 314 various pH conditions during three experiments for four conditions (total of 12 bioreactors). The 315 highly comparable growth behavior of the three replicate bioreactors under batch and continuous 316 conditions for three conditions (i.e., Na-acetate, NaCl, and ammonium conditions) confirm a high 317 stability of our MBS (Fig. S4A, S5A, and S6A). We did observe minor differences in the ethanol 318 and acetate production rates between replicates, which were connected to the same medium feed 319 bottle under continuous conditions (Fig. S4B, S5B, and S6B). These differences in single replicates 320 may lead to different production rates, even in controlled bioreactors, and may result from slightly 321 varying gassing or medium feed rates, variations in the pH control, or small but varying diffusion of 322 oxygen into individual bioreactors. This finding clearly indicates the need for replicates during strain 323 characterization and pre-selection in lab-scale bioreactor experiments before scaling up to larger 324 fermentations. With our MBS, we can combine experiments at steady-state conditions for replicates, 325 which safes time in generating statistically relevant data sets. Our future work to further optimize the 326 MBS will target the additional integration of analytic equipment to calculate gas consumption and 327 carbon uptake rates. We had sampled the inlet and outlet gases during all experiments, but our 328 current setup was not sufficient to obtain reliable results. Additional equipment, such as mass-flow 329 controllers, will fill this gap and further increase the data quality during future experiments. 330 Acetate feed does not improve ethanol production in C. ljungdahlii, but salt stress does at low 331 pH 332 For our second experiment, we addressed the question whether external acetate feed would increase 333 the ethanol production from CO 2 and H 2 for C. ljungdahlii, which had been demonstrated for 334 C. autoethanogenum (Mock et al., 2015). Both strains are closely related. In contrast to the relatively 335 high ethanol production rates for C. autoethanogenum, we observed a negative impact of the acetate 336 feed on growth and production rates of acetate and ethanol (Results S2, Fig. S4). The most important 337 differences for our experiment were that for the previous study by Mock et al. (2015): (1) a sterile 338 filtration system were used, which allowed the retention of 90% of all microbial cells in the 339 bioreactor; (2) a 2-fold higher dilution rate for the medium was applied, which resulted in a 2.3-fold 340 higher feeding rate of acetate; and (3) a ~12-fold higher gas feeding rate of 350 mL min -1 was 341 applied. These differences resulted in a 18-fold higher maximum biomass concentration (based on 342 cellular dry weight) for the previous study compared to our experiment (during Period II). 343 Regardless, we had anticipated a higher ethanol production rate for C. ljungdahlii with external 344 acetate feed. Therefore, further experiments are required why this did not occur. 345 For the control, we had supplemented the standard PETC medium with 100 mM NaCl to give a 346 similar ion strength compared to Na-acetate. This resulted in increased ethanol production rates and 347 ethanol/acetate ratios of 0.8 and 1.2 at a lower pH level during Periods III and IV, respectively (Results S2, Table S2, Fig. S5 and S7). This is considerably higher than the ethanol/acetate ratios of 349 ~0.1-0.4 for all other periods in this experiment. While it is likely that the increased salt 350 concentration in the feed medium resulted in a cellular stress response, which has been demonstrated 351 before (Philips et al., 2017), the exact nature of this response in our experiment remains elusive and 352 requires further experimentation. Others have found increased ethanol production rates from syngas 353 with C. ljungdahlii in response to cellular stress, which was induced by a low pH in combination 354 with sulfur limitation (Martin et al., 2016), or oxygen exposure (Whitham et al., 2015). The feeding 355 of NaCl at low pH might have interacting effects on ethanol production from CO 2 and H 2 . 356  363 S8), we also expected enhanced cell growth in our bioreactor experiment. Our data confirm that the 364 use of nitrate as sole N-source is enhancing CO 2 and H 2 -dependent growth of C. ljungdahlii by up to 365 50% (based on OD 600 ) in continuous mode (Fig. 3, Table 2). Emerson et al. (2019) observed 42% 366 increased growth rates for bottles experiments with CO 2 and H 2 , while the pH increased from 6.0 to 367 8.0. We observed a similar increase in the pH-value and a 200% increased final OD 600 in our 368 preliminary bottle experiments (Fig. S8). Our bioreactors with nitrate feed had high OD 600 -values at 369 low pH values, whereas all our other bioreactors showed a correlation between low pH and low 370 OD 600 . We had not anticipated this observation, since acetate production is becoming 371 thermodynamically limited at lower pH . Consequently, less acetate is produced 372 from acetyl-CoA and, in turn, less ATP is available for the Wood-Ljungdahl pathway (Schuchmann 373 and Müller, 2014). One possible explanation for this observation is that the depleting pool of ATP at 374 low pH is refilled with ATP generated through the reduction of nitrate and concomitant redirection of 375 reducing equivalents. This ATP can then be used for biomass formation. 376 Our data show that the highest OD 600 in our bioreactors with nitrate feed remained between an OD 600 377 of 1.29 and an OD 600 of 1.36 during different periods ( Table 2). This indicates that ATP was not the 378 limiting factor for growth for the bioreactors with nitrate feed and nitrate reduction was sufficient to 379 regenerate redox cofactors. Ethanol formation was not observed in our bottle experiments and in the 380 experiments by Emerson et al. (2019). This led to the hypothesis by the authors that C. ljungdahlii 381 predominantly shifts electrons into nitrate reduction rather than towards ethanol formation. The 382 generated ammonium is responsible for the increasing pH-value. In contrast, we demonstrated for all 383 bioreactors with nitrate feed that ethanol production is still possible under pH-controlled conditions 384 (Fig. 3). We believe that ethanol formation was absent in our bottle experiments due to the increasing 385 pH-value from ammonia. This is proof that observations with bottles should be followed up with pH-386 controlled bioreactors. 387 Nitrite accumulation indicated a metabolic crash of C. ljungdahlii followed by a high ethanol 388 production selectivity after recovering 389 All three bioreactors with nitrate feed showed different performance behavior during the continuous 390 mode. All these bioreactors underwent a crash at different time points during the cultivation. These 391 crashes were stochastic, because we had already observed the reproducible nature of our MBS. For each of the three bioreactors, we measured an accumulation of nitrite and nitrate at the time point 393 when the crash occurred and afterwards. Before the crashes, we were not able to detect nitrate in any 394 sample. Therefore, we assume that the applied nitrate feed rate of 0.11 mmol h -1 (18.7 mM x 0.10 mL 395 min -1 ) was lower than the metabolic uptake rate for nitrate of C. ljungdahlii. Our results indicate that 396 an accumulation of nitrite and nitrate is harmful to the microbes and leads to an abrupt halt of the 397 metabolism for yet unknown reasons. A complete physiological characterization of the nitrate 398 metabolism of C. ljungdahlii, or any other acetogen, is still missing in literature. Emerson et al. 399 (2019) described that once the applied nitrate was depleted, the culture halted acetate production and 400 crashed (as measured by the OD 600 ). The authors explained the crash through an abrupt end of the 401 ATP supply, which is critical to maintain high cell densities for C. autoethanogenum (Valgepea et 402 al., 2017). However, the bottle cultures of Emerson et al. (2019) did not crash completely. The OD 600 403 decreased by 50% but recovered after a short lag phase, indicating that the remaining CO 2 and H 2 404 was further consumed. An accumulation of nitrite was neither observed during the crash in these 405 experiments nor in our own preliminary bottle experiments (Fig. S8) For recovering the culture, we assume that the inhibiting compounds have to be washed out of the 415 system to a certain critical threshold. In addition, some removal of the inhibiting compounds due to 416 the recovering activity of the culture would also contribute. For bioreactor 10, we observed a constant 417 decrease of the OD 600 and acetate and ethanol production rates after the crash in Period I. However, 418 on day 13-14 the decrease started to reach a valley, which indicates that the microbial growth was 419 able to catch up with the dilution of our continuous process. For this bioreactor, the pH in Period II 420 was still high enough to support sufficient growth, and after ~2 HRT periods the growth rate of the 421 microbes exceeded the dilution rate again, which is indicated by an increasing OD 600 in the bioreactor 422 (Fig. 3A). The recovery of this bioreactor 10 in growth as well as in acetate and ethanol production 423 rates indicates that: (1) the nitrate reduction pathway is not per se inhibited at low pH; and (2) the 424 reduction of nitrate and the production of ethanol is possible simultaneously, and that most likely the 425 low pH triggers a thermodynamic shift towards ethanol production . However, it 426 remains elusive why the ethanol/acetate ratio in bioreactor 10 reached a nearly ten-fold higher value 427 after recovering from the crash in the presence of nitrate compared to the bioreactors with ammonium 428 feed (Fig. 3B, Table 2). 429 In contrast, the crash occurred for bioreactor 11 in Period II. While the OD 600 immediately decreased 430 after the crash, the acetate and ethanol production rates remained somewhat constant until the switch 431 to Period III. However, this bioreactor never recovered from the crash in terms of OD 600 . We believe 432 that the lower pH levels during Period III for bioreactor 11 prevented the growth recovery, which for 433 bioreactor 10 took place at the higher pH level of Period II. We had found reduced growth conditions 434 for all other bioreactors at the lower pH levels (Fig. 3, Fig. S4 were not given. We observed higher acetate production rates for bioreactor 10 and 11 before the 445 crash compared to those of the bioreactors with ammonium feed (Fig. 2B, Fig. 3B). Bioreactor 12 446 did not reach a similarly high acetate concentration, but the crash occurred at the beginning of Period 447 IV at the lower pH of 4.5. Intrinsically, the extracellular undissociated acetic acid concentration 448 would be higher as a key to trigger the crash event. However, more work is necessary to ascertain the 449 mechanisms of the stochastic crashes. 450 Nitrate reduction offers a great potential to further optimize gas fermentation of C. ljungdahlii. 451 Because ATP limitation is one of the highest burdens to overcome for acetogens ( The 1-L bioreactor vessel  570  consisted of a double-walled glass vessel and a customized lid, while it was placed on a multi-stirring  571 plate with up to six bioreactors. The bioreactor temperature was maintained through a water 572 circulation unit at 37°C. The autoclavable lid offered connections for 5x GL14 and 1x GL25. A set of 573 stainless-steel tubing was used for the gas-in/-out lines and for the medium feed-out line. The three-574 way valve at the medium feed-out line was required for sampling using a 5-mL syringe. The pH and 575 bioreactor medium temperature was tracked via a pH/pt1000-electrode that was connected to a multi-576 parameter instrument. The multi-parameter instrument controlled and triggered two mini pumps (for 577 base and acid) at programmable conditions. For continuous mode, the feed medium to each 578 bioreactor was pumped via a single multi-channel pump from the feed tank into the bioreactor. The 579 same pump was used to transfer the effluent from each bioreactor into the effluent tank. Sterile CO 2 580 and H 2 gas (20:80 vol-%) was sparged into the system through stainless-steel tubing with an attached 581 sparger. The gas-out line was connected to a 100-mL serum bottle to serve as a water trap before the 582 outgoing gas passed an airlock. The 1x, 2x, and 6x next to each unit in the figure describe the 583 quantity, which is required to operate six bioreactors simultaneously. Abbreviations: A/B, Acid 584 and/or base feed line; E, pH/pt1000 electrode; Fa, medium feed-out line; Fi, medium feed-in line; Ga, 585 gas-out line; Gi, gas-in line; GL14, screw joint connection size 14; GL25, screw joint connection size 586 25; rpm, revolutions per minute; SB, stirring bar; 3WV, three-way valve. Blue lines indicate liquid 587 transfer, red lines contain gas, and dotted black lines provide electric power or signals. 588  acetate and ethanol production rates in mmol-C L -1 d -1 (B). The bioreactors with nitrate feed were 597 grown in ammonium-free PETC medium supplemented with 18.7 mM Na-nitrate. The horizontal 598 dotted lines indicate the continuous process in which medium of different pH was fed to each 599 bioreactor. The red arrows and circles indicate the crash in OD 600 of each bioreactor with nitrate feed 600 at different time points. The star symbol describes the time point were the pH was lowered manually 601 by adding HCl to the system until a pH of 4.5 was reached. Period: I, pH=6.0; II, pH=5.5, III, 602 pH=5.0; and IV, pH=4.5. 603 Tables 605 Table 1 606