Online measurement of dissolved carbon monoxide concentrations reveals critical operating conditions in gas fermentation experiments

Syngas fermentation is one possible contributor to the reduction of greenhouse gas emissions. The conversion of industrial waste gas streams containing CO or H2, which are usually combusted, directly reduces the emission of CO2 into the atmosphere. Additionally, other carbon‐containing waste streams can be gasified, making them accessible for microbial conversion into platform chemicals. However, there is still a lack of detailed process understanding, as online monitoring of dissolved gas concentrations is currently not possible. Several studies have demonstrated growth inhibition of Clostridium ljungdahlii at high CO concentrations in the headspace. However, growth is not inhibited by the CO concentration in the headspace, but by the dissolved carbon monoxide tension (DCOT). The DCOT depends on the CO concentration in the headspace, CO transfer rate, and biomass concentration. Hence, the measurement of the DCOT is a superior method to investigate the toxic effects of CO on microbial fermentation. Since CO is a component of syngas, a detailed understanding is crucial. In this study, a newly developed measurement setup is presented that allows sterile online measurement of the DCOT. In an abiotic experiment, the functionality of the measurement principle was demonstrated for various CO concentrations in the gas supply (0%–40%) and various agitation rates (300–1100 min−1). In continuous stirred tank reactor fermentation experiments, the measurement showed reliable results. The production of ethanol and 2,3‐butanediol increased with increasing DCOT. Moreover, a critical DCOT was identified, leading to the inhibition of the culture. Thus, the reported online measurement method is beneficial for process understanding. In future processes, it can be used for closed‐loop fermentation control.


| INTRODUCTION
Gas fermentation is, among others, one possibility to reduce carbon dioxide emissions and to provide platform chemicals from renewable resources or industrial and urban waste streams. Gaseous sources containing CO, CO 2 , and H 2 , from now on referred to as syngas, can either originate from industrial plants, waste streams, or lignocellulosic biomass (McKendry, 2002;Y. Sun et al., 2015). Several reviews have investigated various aspects of syngas fermentation, and a number of demonstration and pilot plants are already under construction or in operation (De Tissera et al., 2017;Takors et al., 2018;Teixeira et al., 2018). Thus, commercially operating plants that contribute to a circular and sustainable economy are expected in the near future (De Tissera et al., 2017;Phillips et al., 2017;Takors et al., 2018).
In most studies investigating syngas fermentation, CO in combination with H 2 and/or CO 2 was used (Hurst & Lewis, 2010;Jack et al., 2019;Mohammadi et al., 2016;X. Sun et al., 2019;Takors et al., 2018;Valgepea et al., 2018). CO is the energetically most beneficial substrate, compared to the consumption of CO and H 2 or CO 2 , and H 2 mixtures (Phillips et al., 2017;Wilkins & Atiyeh, 2011). However, in nature, CO mostly occurs at rather low concentrations and can become highly toxic for most organisms (Ragsdale, 2004). Several studies have investigated CO as a substrate for the cultivation of acetogenic microorganisms. CO is the only component of syngas, which can be a source of carbon and energy at the same time, and is the energetically most feasible gaseous substrate for acetogenic organisms (Phillips et al., 2017). CO is widely available from humanmade processes, as it can originate from industrial plants such as steel mills, or can be produced via gasification of carbon-containing waste streams. Different studies have investigated the influence of varying partial or headspace pressures on the cultivation of Clostridium ljungdahlii. For example, Stoll et al. (2019) found that in fermentations using a gas mixture of CO, CO 2 , H 2 , and N 2 with an elevated headspace pressure between 1 and 7 bars, the maximum growth rate decreased. In fermentation experiments with a pressure of 7 bar, no hydrogen or carbon dioxide was consumed, and cultivation conditions were described as unstable. It is assumed that the decreasing consumption of hydrogen results from the toxicity of CO on hydrogenases (Menon & Ragsdale, 1996). Jack et al. (2019) tested different CO to H 2 ratios in C. ljungdahlii cultivation experiments and reported higher ethanol production with an increasing CO to H 2 partial pressure ratio as well as higher 2,3-butanediol (2,3-BDO) production at a higher CO partial pressure. The production of more reduced products, such as ethanol and 2,3-BDO, in comparison to acetate results from more available energy in cultivation experiments with CO (Jack et al., 2019). These results are in accordance with the data provided by Najafpour and Younesi (2006), showing an increase in ethanol production with increasing initial headspace pressures.
The same phenomenon has also been observed for other acetogenic Clostridia. Hurst and Lewis (2010) reported on C. carboxidivorans cultivation experiments using a CO partial pressure range of 0.35−2 atm, where increasing ethanol to acetate ratio was measured with increasing CO partial pressure.
Syngas fermentation at a laboratory scale is usually conducted either in serum bottles or in gas continuous or gas and liquid continuous stirred tank reactors (CSTR). Serum bottles are advantageous as several experiments can be conducted in parallel owing to the low technical effort required. However, the information obtained from serum bottle experiments is limited, as it relies on offline sampling.
Especially during gas fermentation, gas and liquid sampling result in changing experimental conditions. In contrast, stirred tank reactors are limited in the number of parallel experiments, but offer a wide range of online monitoring opportunities, such as pH and redox potential. Another key parameters of fermentation, well known from aerobic fermentation, is the dissolved gas concentration (Knoll et al., 2005). In aerobic fermentation, dissolved oxygen tension (DOT) is widely used for fermentation control. Oxygen can be the limiting substrate in aerobic fermentations when the DOT is too low. This may result in decreased productivity or a switch in the metabolism toward a different product spectrum (Knoll et al., 2005).
Because the same phenomenon likely applies to CO as a carbon and energy source in syngas fermentation, quantification of the dissolved carbon monoxide tension (DCOT) is crucial for process understanding and development. As mentioned above, several groups have reported influence on the cultivation performance and the product spectrum with varying gas partial pressures, resulting in varying dissolved gas concentrations (Abubackar et al., 2012;Cotter et al., 2009;Hurst & Lewis, 2010). Offline analysis of the DCOT using a myoglobin assay was reported (Ungerman & Heindel, 2008). This method is based on sampling from the liquid phase and the following reaction of CO with myoglobin. Subsequently, the saturation of myoglobin was measured. The method does not require any additional equipment installed within the fermentation vessel; however, manual labor is needed for sampling and sample processing. In contrast to the offline measurement of the DCOT, online DCOT monitoring would be superior to reduce error-prone manual labor and to increase the available data points (Ungerman & Heindel, 2008).
Measurement of the DOT is usually performed using electrodes mounted into the fermentation vessel, with the sensitive part immersed into the liquid phase of the fermenter. In 1961, Phillips and Johnson published an article reviewing different methods for online determination of DOT, one of them is referred to as the "tubing method." This method is based on the diffusion of oxygen into a Teflon tube mounted in the liquid phase of the fermentation vessel.
The tube is constantly ventilated using a carrier gas, and the oxygen concentration at the end of the tube is measured with a gas analyzer (Phillips & Johnson, 1961).
In this study, a method for the DCOT measurement, based on the principle published by Phillips and Johnson (1961) has been set up and established in CSTR cultivation experiments with C. ljungdahlii using CO as the sole carbon and energy source. To the best of our knowledge, no online measurement method for DCOT has yet been established (De Tissera et al., 2017). 254 | 2 | MATERIALS AND METHODS 2.1 | Bacterial strains and culture media C. ljungdahlii (DSMZ 13528) was received as an actively growing culture in ATCC 1754 medium containing 5 g L −1 fructose. Cryogenic stocks were prepared by adding 10% vol/vol of anoxic dimethyl sulfoxide (DMSO) to a C. ljungdahlii culture in the exponential growth phase at an optical density (OD) of 0.8. DMSO was then added to the cultivation flask and thoroughly mixed with the culture. Subsequently, 2 ml of the culture was transferred into a sealed anoxic 5 ml glass vial using a syringe and needle. The culture was then stored at −80°C.
In all the fermentation experiments, ATCC 1754 medium was used as described previously (Tanner et al., 1993). Cysteine was used as the sole reducing agent and added from a sterile, anaerobic stock solution to the anaerobic media stock just before inoculation. The final cysteine concentration in the medium was 0.75 g L −1 . For precultures, 5 g L −1 of fructose was used as the sole source of carbon, and 100 mM Bis-Tris buffer was added to it. All the fermentation experiments were performed with CO as the sole source of carbon and energy. The initial pH value for all the cultivation experiments was 6.0. A detailed description of the media preparation is provided in Data S1.

| Precultures
Two sequential precultures were grown in ATCC 1754 medium with the addition of 20 mM Bis-Tris buffer and 5 g L −1 of fructose. For the first preculture, a 100 ml serum bottle (Glasgerätebau Ochs) containing 20 ml medium was inoculated with 1 ml cryogenic stock and cultivated for 24 h. The second preculture was grown in three simultaneously cultivated 250 ml serum bottles (Glasgerätebau Ochs) containing 45 ml fresh medium and 5 ml of the first preculture for 24 h before inoculation of the CSTR. Precultures were grown on an orbital benchtop shaker (Professional incubating orbital mini shaker; Talboys) with a shaking diameter of 3 mm and a shaking frequency of 220 min −1 at 37°C.

| CSTR experiments
CSTR experiments were inoculated from an actively growing preculture at an OD of 0.8 in a 1:10 ratio. Initially, the CSTR was run in a liquid batch and gas continuous mode (from now on referred to as the semi-batch mode). Subsequently, a continuous liquid feed was applied additionally to the continuous gas supply (from now on referred to as continuous operation mode). All the CSTR experiments were performed in a 2 L benchtop bioreactor (Applikon Biotechnology) with an operating volume of 1.5 L at 37°C. Three baffles and two Rushton impellers (46-mm diameter) were installed. The agitation rate was varied between 300 and 1200 min −1 ; detailed information is provided for each fermentation experiment. The gas supply for the fermentation experiments was set to 0.03 vvm (45 ml min −1 ) and was mixed with nitrogen (99.999%; Praxair) and CO (99.995%; Praxair) using two mass flow controllers (MDR) in combination with an MDR controller unit (MiniTerm; EASE-Products). CO (20%-40%) was used as the sole carbon and energy source for all fermentation experiments. Samples were taken regularly using a sampling system (Super Safe Sampler; Infors HT). The fermenter was equipped with a pH probe (EasyFerm Plus PHI K8 225; Hamilton), and the pH was constantly maintained at 6 using 1 M NaOH. The fermenter exhaust gas was diluted 1:10 with nitrogen (99.999%; Praxair) behind the exhaust gas cooler using MDR for flow control (EL-Flow; Bronkhorst High-Tech; Figure 1), and measured using an exhaust gas analyzer (X-Stream XEK; Emerson). For the continuous liquid operation mode, a peristaltic pump (FIXO ISM 835; Ismatec) was used to control the medium feed to maintain a dilution rate (D) of 0.02 h −1 . The liquid effluent was controlled by the fermenter weight to maintain a constant filling volume. Therefore, the fermenter was placed on a scale (SBC 51; Scaletec). Details of the DCOT measurement are depicted in Figure 1 and the results section. Figure 2 shows the calibration procedure. In the fermentation presented in Figure

| Offline analysis
A 2-ml sample from the liquid phase was taken regularly and the OD was determined at 600 nm using a spectrometer (Genesys 20; Thermo Fisher Scientific). The remaining sample was centrifuged for 5 min at relative centrifugal force = 18,000g (Rotina 35R; Hettich Zentrifugen), and the supernatant was frozen for further use. High-performance liquid chromatography (HPLC) analysis (Prominence HPLC; Shimadzu) was performed after filtering each sample (0.2 µm cellulose acetate membrane; VRW). An organic acid column (ROA-Organic Acid H+; Phenomenex Inc.) was used with 5 mM H 2 SO 4 as the eluent at a flow rate of 0.8 ml min −1 and a temperature of 60°C. A refractive index detector (RID-10A; Shimadzu) was used for the detection of fructose, acetate, ethanol, and 2,3-BDO. Compared to the setup presented by Phillips and Johnson, the tube length was reduced by a factor of 17, because of the 10 times thinner tube wall in the herewith presented method. Reducing the tube length is essential to maintain sufficient mixing and prevent inhomogeneous areas in the stirred liquid phase. The applied CO sensor measures CO concentrations ranging from 0 to 300 ppm.

| Measurement of DCOT
Thus, the measurement range is 100 times lower than the measurement range of the oxygen sensor used by Phillips and Johnson.
Furthermore, no minimum gas flow rate past the sensor is required, and therefore, the tube length can be reduced. By reducing the flow rate of the carrier gas through the measurement tube, the retention time in the tube can be prolonged, and the tube length can be further reduced (Phillips & Johnson, 1961). The setup described in Figure 1 was first evaluated in an abiotic experiment. The DCOT was F I G U R E 1 Schematic outline of the dissolved carbon monoxide tension measurement. Nitrogen (green dots) was constantly flushed through a silicon tube. CO (blue diamonds) diffused from the fermentation broth through the wall of the tube into the tube lumen and was transported to a CO sensor. Tube length: 72.5 cm, inner diameter: 2 mm, outer diameter: 2.6 mm, material: silicon (VMQ) (Deutsch & Neumann) carrier gas: N 2 (ultrahigh purity), volumetric flow: 15 ml min −1 , total fermenter volume: 2 L, and liquid volume: measured at five different agitation rates (300-1100 min −1 ) and four different CO concentrations in the gas supply ( Figure 2). As the experiment was performed under sterile conditions, no CO was consumed in the liquid phase. Hence, the DCOT is independent of the agitation rate. As shown in Figure 2, the CO concentration in the gas supply correlates with the CO concentration at the end of the measurement tube with an R 2 value of 0.995. As the experiment was performed under sterile conditions, the partial pressure of CO in the gas supply correlates with the DCOT as an equilibrium state was reached. The equation shown in Figure 2 allows the calculation of the DCOT in percent (%) in relation to the gas supply concentration. The maximum CO concentration used in the experiments was 40%.
Therefore, the DCOT is set to 100% when 40% CO is used in the gas supply. This is in accordance with the procedure used for DOT quantification in aerobic fermentation, and in most cases, a DOT of 100% results from aeration using ambient air, which contains 21% oxygen.
3.2 | CSTR experiment with a sudden increase in CO concentration in the gas supply previously (Köpke et al., 2011). After 286 h, the CO tension in the gas supply increased from 20% to 40%. For 48 h, DCOT slowly increased before suddenly rising from 16% to 100% (Figure 3a). Until then, the base addition, OD, ethanol concentration, and 2,3-BDO concentrations increased, while the acetate concentration slightly increased (Figure 3).
In the semi-batch mode, the maximum growth rate was determined to be approximately 0.041 h −1 (Figure S1), based on the OD ( Figure 3b). Therefore, a dilution rate ( As reported by other groups, the agitation rate during the initial batch phase was increased stepwise from 300 to 700 min −1 to prevent CO poisoning, since 20% CO was added via the gas supply The DCOT plotted in Figure 3a started at a value of 50%. After inoculation, the DCOT decreased until reaching a value of approximately 0% after 48 h. In the beginning, when low agitation rates were applied, fluctuations in the DCOT measurement were observed (compare Figures S2-S4). The fluctuations are due to gas bubbles adhering to the tube, thus altering the CO transfer into the tube. Increasing the agitation rate increases the shear forces along the tube, which decreases the adherence of gas bubbles. When the agitation rate is increased, an increasing DCOT is expected. However, in Figures S2-S4 representing the initial semi-batch phase, a decreasing DCOT was measured with an increasing agitation rate. This phenomenon possibly resulted from fewer gas bubbles adhering to the tube with an increasing shear force when a higher agitation rate was applied ( Figures   S2-S4). This phenomenon was not observed at agitation rates above 700 min −1 , as shown in Figure 4, which will be discussed in detail later. The products plotted in Figure 3c show a drastic change after the CO increase. The acetate concentration decreased slightly, and within 48 h, the ethanol and 2,3-BDO concentrations increased to 6 and 1.2 g L −1 , respectively. Increased production of ethanol has been reported for an increasing CO partial pressure in serum bottles, thus proving the results shown in Figure 3   Once the CO concentration in the gas supply was increased, the DCOT increased. An increase in the DCOT resulted in more alcohol production and finally led to a collapse of the metabolic activity once a DCOT of 16% was reached.

| Influence of agitation rate on DCOT in a CSTR experiment
To investigate the influence of an increase in k L a instead of the CO concentration in the gas supply, the CO concentration in the gas supply was maintained at 20% and the agitation rate was varied in the next fermentation (Figure 4). In addition, an exhaust gas analysis The online data of the exhaust gas analyzer verifies the CO-CO 2 conversion reported by several studies previously (Abubackar et al., 2011;Phillips et al., 1993;Wilkins & Atiyeh, 2011). With a decreasing CO concentration in the exhaust gas, the CO 2 concentration in the exhaust gas increases. The highest CO conversion rate of 85% was measured at an agitation rate of 1000 min −1 . The measured conversion rates match the conversion rates reported elsewhere (Phillips et al., 1993;Richter et al., 2013).
The DCOT plotted in Figure 4a decreased immediately after inoculation to approximately 0% after approximately 80 h. As mentioned before, a low agitation rate resulted in a noisy signal of the DCOT, as observed for agitation rates of 300 and 500 min −1 during the first 80 h, and between 240 and 455 h for an agitation rate of 500 min −1 . In addition, the DCOT signal remained at approximately 0%. A DCOT of approximately 0% indicates a mass transfer limitation. However, CO was continuously supplied and consumed. The base addition plotted in Figure 4b depends on the COTR, which is adjusted by the agitation rate. An increasing COTR indicates an increase in metabolic activity. Therefore, the linear base addition indicates a constant metabolic activity, further proving the gas transfer limited state of the culture. For 600 h, hardly any ethanol was detected (neglecting the semi-batch phase). Thereafter, the agitation rate exceeded 800 min −1 , and small amounts of ethanol were measured in offline samples. After 740 h, the agitation rate was set to 1100 min −1 and straight away to 1300 min −1 (compared to Figure 4d), with a sudden increase in ethanol production and a slight increase in the DCOT. 2,3-BDO was not detected in any sample. As the setup was not capable of maintaining a high agitation rate, it was decreased to 1000 min −1 shortly after. Nevertheless, the DCOT increased, and ethanol production indicated a change in the metabolic state of the culture that has been observed for an increasing CO concentration in the gas supply, as shown in Figure 3.
The results of the fermentation experiment in Figure 4 show the impact of the agitation rate on the CO fermentation experiment.
At agitation rates exceeding 900 rpm, slight alcohol production was measured. However, the impact of the agitation rate was far less extensive than that of the CO concentration in the gas supply ( Figure 3).

| Influence of a stepwise increase in the CO concentration in the gas supply on the DCOT in a CSTR experiment
To confirm the results of the previous fermentation experiments and to investigate the influence of the adaption time on higher CO concentrations, the CO concentration in the gas supply was increased stepwise from 20% to 40%. In contrast to the fermentation shown in Figure 3, an intermediate level of 30% CO in the gas supply was selected to allow adjustment in the cell concentration. The results are shown in Figure 5. Initially, the OD was approximately 0.1, while acetate and fructose carryover from the preculture was observed.
Immediately after inoculation, as discussed for the previous fermentation experiments, the OD, base addition, and acetate concentration started increasing (Figure 5b,c). After 90 h, the continuous mode was started and 20% CO was maintained in the gas supply for another 48 h. Thereafter, the CO concentration in the gas supply was adjusted to 30%. The DCOT increases to approximately 6% and remained at this level for 48 h. After 48 h, the CO concentration in the gas supply was set to 40%. Subsequently, the base addition de- and 5. Furthermore, the acetate concentration in Figure 3 decreased less after the increase to 40% CO, compared to the acetate concentration in Figure 5. As shown in Figure 3, the decrease in the acetate concentration increased when the DCOT increased. In Figure 5, the acetate concentration decreased immediately after the increase in the CO concentration to 40%. The different responses to the increase in CO concentration possibly resulted from the different histories of the culture. In Figure 3, C. ljungdahlii was cultivated at 20% previously and did not show any stress signals. In Figure 5, C. ljungdahlii was cultivated on 30% CO before the CO concentration in the gas phase was increased to 40% CO. During the cultivation of 30% CO, C. ljungdahlii was already exposed to an increased DCOT, and the production of alcohol during this period indicates that the culture was already stressed. This difference in cultural history is probably the reason for the different reactions of the cultures.
The CO fermentation experiment with a gradually increasing CO concentration in the gas supply ( Figure 5) confirms the data obtained from the fermentation experiment shown in Figure 3. A CO concentration in the gas supply exceeding 20% resulted in a DCOT significantly above 0%. While cultivation with 30% CO was possible under the given settings, 40% CO in the gas supply resulted in a gradual increase in the DCOT and a metabolic collapse at a DCOT of 16%.

| Evaluation of the dissolved CO measurement method
For the first time, a method to measure the online DCOT in fermentation experiments is reported. Knowledge of the DCOT over time is crucial for a detailed process understanding. As is commonly done for the DOT in aerobic fermentation, the online measured DCOT should be used for process control. As suggested by the online results, a DCOT control via the agitation rate or the CO concentration in the gas supply could be used to enhance alcohol production in syngas fermentation. DCOT control, analogous to the DOT control in aerobic fermentation, is only possible if continuous online data of the DCOT is available. This is a clear advantage of the newly reported online measurement method in comparison to the previously reported offline measurement methods. Offline measurement of the DCOT does not allow the implementation of a DCOT control.
The noninvasive online measurement principle of the DCOT presented in this study offers several advantages over offline measurement methods currently available (Kundu et al., 2003;Riggs & Heindel, 2006;Ungerman & Heindel, 2008). conditions before it becomes apparent in the exhaust gas analysis; thus, it is highly beneficial for fermentation control strategies ( Figure 5). In particular, when aiming at the production of alcohol, the DCOT signal is the only reliable online signal as no further base addition is needed to maintain a constant pH during alcohol production, as shown in Figures 3-5. In all experiments, the DCOT indicated changes in the cultivation performance before the exhaust gas analysis revealed any changes (Figures 4 and 5). This is a big advantage over the offline determination of the DCOT. Solely relying on offline sampling bears a high risk of missing crucial turning points of the fermentation process. As changes in the DCOT result in changes of metabolic activity, online monitoring of the DCOT can be used for fermentation control, as it has been performed for aerobic fermentation for decades (Phillips & Johnson, 1961). Fermentation experiments showed a noisy DCOT at low agitation rates (below 500 min −1 ). This probably resulted from gas bubbles adhering to the tube because of low shear forces. This can be circumvented by readjusting the measurement tube to the same height as the impeller.
This measure would increase the shear force, removing gas bubbles adhering to the measurement tube. The addition of an antifoam agent ( Figure 3) altered the measured DCOT values. It was shown, that the installation of a hose clamp onto the stirrer shaft above the liquid phase prevented excessive foam formation. Therefore, the addition of an antifoam agent can be prevented.
In the two fermentation experiments, the DCOT exceeded 16% after increasing the CO concentration in the gas supply to 40%, followed by a halt of the metabolic activity (Figures 3 and 5). With increasing DCOT, the ethanol concentration, as well as the 2,3-BDO concentration, increased. The toxicity of the products can be neglected as concentrations exceeding the concentrations in the presented fermentation experiments reported previously (Ramió-Pujol et al., 2018;Richter, Molitor, Wei et al., 2016). Therefore, a critical threshold of the DCOT can be assumed. Exceeding the threshold value prevents further metabolic activity.
The described measurement shows the need to monitor the DCOT during syngas fermentation. As stated in the various publications, the DCOT can influence the outcome of fermentation, although no online measurement tool has been available to date. As shown in the fermentation experiments, the DCOT can be used to control C. ljungdahlii towards either acetate or alcohol production.
Further experiments will aim at a DCOT control that is similar to well-known aerobic fermentation setups, where the DOT is controlled at a steady level using gas flow and agitation rate, and in some cases pressure (Knabben et al., 2010;Knoll et al., 2007).

| CONCLUSION
A method for online monitoring of the DCOT has been adapted from reports on DOT measurements and has been successfully implemented and validated. The results showed the feasibility of DCOT monitoring in syngas fermentation to detect the metabolic state of the cultivation before being detected by any other online signal. The method was applied to demonstrate a metabolic shift towards alcohol production when the DCOT increased, and a halt of the metabolic activity when the DCOT exceeded 16%. The DCOT measurement showed changes in the metabolic activity before any other online measure, including exhaust gas analysis. The presented method is easy to apply and faster in response compared to alternative methods based on offline samples.