The results of the laboratory fermentations confirm that CO2 is inhibitory to the ethanol fermentation. This inhibition manifests itself in two ways: a lower rate of glucose consumption and ethanol accumulation at higher CO2 levels, associated with a lower peak cell count at 2.5 and 3.5 atm and an increased rate of cell death at

3.5 atm. The lower rate would lead to incomplete conversion of available glucose at the higher CO2 levels if the fermentations were terminated at 45 h. Letting the fermentations continue to 65 h or more, there was little or no difference in the ultimate conversion of glucose except possibly at 3.5 atm. Two reservations accompany these conclusions. 1) Each condition was only tested once, so that no statistical test can be applied. The effects described here are greater, usually substantially greater, than the intrinsic uncertainties of the analytical methods. The reproducibility of the laboratory fermentation itself, however, was not tested due to limitations of time and resources. 2) Not all the carbon can be accounted for at 3.5 atm and in the air-supplemented run, though this may reflect a simple setup error. There was no indication of increased evaporation or acid production. The calculation of the percentage completion was based on the final ethanol concentration and was not influenced by any error in carbon recovery or amount of carbohydrate added. Similar slowing of the fermentation and decreases in the peak cell number have been observed at lower temperature by workers in the brewing industry (6-9). Slowing of industrial alcohol fermentation of molasses by CO2 was observed by Ukrainian workers (28). However, if the concentration rather than the partial pressure of carbon dioxide is the determining factor (6), it is somewhat surprising to observe appreciable slowing of the fermentation at pressures as low as

1.5 atmospheres, since the solubility of carbon dioxide is so much decreased at 32 °С. This slowing of the fermentation must arise in part from causes other than decreased cell count, since cell count was not decreased at 1.5 atm.

The laboratory experiments were run as fed-batch fermentations with pure glucose as the carbon source. This design was chosen to avoid any uncertainties associated with the use of complex substrates and enzymes in situ. The remainder of the medium was based on a well-characterized defined medium with ammonium chloride and yeast extract added as sources of nitrogen. Although we intended the lab runs to simulate the most important features of plant conditions, there were some noticeable differences. We were unable to get good fermentation rates or cell counts until we increased the level of N in the media. We now think it likely that the lower nitrogen media led to N-limitation of the inoculum; the runs employing the lower-N media are not reported here. Even with the higher N level, we could not match the cell count, the retention of cell vitality or the fermentation rate of the industrial runs. This is possibly due to the leaner semi-defined media used in the lab runs in comparison with the rich, complex ground-corn medium employed in the plant. Also, the base medium we used is not a particularly good match for the inorganic components of the com mash. Supplementation with higher levels of N, lipids, or potassium did not improve the performance in shake flask experiments (not shown). The lab runs were a good match for the industrial runs in the critical variables of pH and temperature, particularly during the latter part of the fermentation when the CO2 inhibition manifests itself.

Glycerol production represents an economically significant diversion of carbon, about 4-5%, from the production of ethanol (18 and our results). Glycerol production under plant conditions is largely growth associated. Under lab conditions, glycerol production was largely non-growth-associated; this may be related to the fact that cell growth in the industrial fermentation was nitrogen limited, while growth in the main set of lab runs was not, or it may be related to the higher rate of cell death in the lab runs. The limiting factor in the lab fermentations was not identified. Glycerol formation was least in the 3.5 atmosphere lab run, and greatest at 0.5 atmospheres. This is contrary to Oura’s hypothesis (18) but in accord with the results of Kuriyama et al. and Bur’yan & Volodrez (19,20). These two papers employed semiaerobic continuous fermentations while our experiments were anaerobic batch fermentations. Oura based his suggestion of increased glycerol production at elevated CO2 on the requirement of pyruvate carboxylase for CO2 in

—°— log viable — glycerol -o— ethanol/10 ■— C02, mM count Figure 3b. Product Levels, Air Supplement 0—23 h

time, h

g/l. h —■— g/vc. h

Figure 3c. Glycerol Productivity, Air Supplement 0-23 h

order to produce oxaloacetate, which is necessary for succinate production.

However the Km of yeast pyruvate carboxylase for KHCO3 is in the range 2-8 mM (29), so the enzyme should be easily saturated with CO2. Kuriyama et al. suggest pyruvate dehydrogenase as the site sensitive to CO2 but their results and ours may also be related to inhibition of pyruvate carboxylase by high CO2 since Foster & Davis (30) found that high CO2 inhibited Rhizopus pyruvate carboxylase. Although high levels of CO2 suppressed some of the glycerol production the concomitant decrease in fermentation rate means that increased CO2 levels would not be a good strategy for increasing yield of conventional batch fermentations under industrial conditions. Oura demonstrated that low levels of aeration in a continuous fermentation led to decreased glycerol production (31). We were able to manipulate the rate of glycerol production with air supplementation, both in the lab and in the plant. The air-supplementation in each case was calculated to be just sufficient to permit the cells to retain redox balance without producing glycerol, assuming 50% oxygen uptake. In the lab, oxygen eliminated the initial spike of growth-associated glycerol productivity and decreased the non-growth-associated productivity on a per-cell basis but the accompanying higher cell count resulted no decrease in overall production. The oxygen uptake in the lab fermentation was substantially more than 50%, but decreases in the flow rate compensated for this so that the total oxygen uptake through the run was close to the intended amount. In the plant, oxygen decreased the growth-associated glycerol production on a per-cell basis but did not eliminate it. Additional glycerol was produced later in the run, resulting in no net improvement. In the lab, aeration was continued until 53 hours and then discontinued without resulting in an increase in glycerol production. It is plausible
to think that an aeration schedule could be developed to decrease glycerol production, though more experimentation would be required to establish whether this could be achieved.

We were unable to influence the carbon dioxide level or the fermentation kinetics in the plant by operating the fermentor at reduced pressure. This might have been more effective with a larger pump which could keep up with the peak CO2 production. The light aeration in the air-supplemented fermentation resulted in a small decrease in carbon dioxide level in the industrial fermentor, but no clear-cut improvement in fermentation kinetics. Other experiments including increased CO2 back pressure and nitrogen sparge were considered and rejected due to cost or complexity.

Conclusions

Carbon dioxide levels as low as 1.5 atmospheres, the range prevailing in the shallow fermentors at Morris Ag-Energy, slows ethanol fermentation appreciably. In these experiments this effect became noticeable after the fermentation was about half complete.

Carbon dioxide at 2.5 and 3.5 atmospheres decreased the peak yeast cell count slightly and at 3.5 atmospheres was associated with more rapid cell death.

Higher carbon dioxide was accompanied by decreased glycerol production.

Air supplementation of the fermentation decreased glycerol productivity on a per — cell basis and led to higher cell counts. These effects combined to leave the overall glycerol production essentially unchanged.

Acknowledgments

This research was supported by a grant from the Minnesota Com Research and Promotion Council, with additional support from AURI (Minnesota’s Agricultural Utilization Research Institute), Morris Ag-Energy, and the Minnesota Department of Public Service. We thank Duaine Flanders and Richard W. Fulmer for encouragement and helpful advice, Gerald Bachmeier for arranging plant access, Nancy Goebel for help with assays and lab fermentations, and Sterling Keller and Craig Bremmon for help with instrumentation.