The time course for the ethanol fermentation process showing residual glucose concentration, ethanol concentration in the reactor and cell density measured as OD600 is shown in Fig. 3.
The fermentation process can be separated into various stages: Stage I (hr 0–3) was the aerobic batch growth stage, in which aeration at 1vvm with 600 rpm agitation was used to achieve rapid cell growth; Stage II (hr 3–10) was the batch fermentation stage, in which the aeration and mixing rates were reduced to 0.2 vvm and 300 rpm respectively, and MSU circulation was turned on with the stripping process activated; Stage III (hr 10–12) was a fed batch fermentation stage, where 300 ml of concentrated feed solution was added as a single batch; and Stage IV (hrs 12 to 28) was the continuous fed-batch stage where the concentrated sugar medium feeding rates were 1.8ml/min (12-26.45hr) and 2.8 ml/min (hr 26.45-28.0). The fermentation was terminated after 28.0 hr. The variation in key parameters are discussed below, by stage.
Stage I: Aerobic stage (0–3 h)
The OD600 is an approximate measure of the cell concentration in the fermentation broth. Therefore, in a simple batch culture the health of the culture, can be inferred from changes in OD600. Over the aerobic growth stage where the broth was aerated with 2.5 L/min of air, the primary aim was to encourage cell growth to build a robust culture for rapid production of ethanol. During this stage, the OD600 increased to approximately three times the initial value, consistent with other studies using P. thermoglucosidasius[21, 22]. A low concentration of ethanol was produced during this stage because the microorganism is Crabtree-negative[23] and does not express fermentative alcohol dehydrogenase in an oxygen rich environment[24]. The liquid recirculation with the MSU was not operational during the aerobic stage; therefore, no data for the ethanol concentration in the condensate was collected.
Stage II: Batch fermentation stage ( 3–10 h)
During the batch fermentation stage (Stage II), the aeration rate in the fermenter was reduced to 0.2 vvm, and the agitation rate was reduced to 300 rpm to maintain microaerophilic conditions (required to induce fermentative growth), while the liquid circulation between the MSU and fermenter was started. The OD600 continued to increase up to ~ 8 h and then decreased over the next 2 h due to carbon source exhaustion, and further at 12 h due to the dilution effect of the batch feed addition at 10 h (Fig. 3). Post fermentation analysis showed that the glucose in the bioreactor had been exhausted by 8 h so the subsequent reduction in OD600 might also reflect a degree of cell lysis which is typically observed after rapid sugar starvation of P thermoglucosidasius NCIMB 11955. The initial increase in cell density could also partially reflect the reduction in liquid volume in the system due to evaporation from the MSU. Over 95% of the maximum theoretical expected ethanol (at 90% assumed yield) at stage II was recovered and collected in bottles (Table 1 and Fig. 4). This stage demonstrates that hot microbubble stripping had no evident harmful effects on cells recirculating in the fermentation broth, complementing microbubble research in biological systems under isothermal conditions [25].
The ethanol concentration in the bioreactor increased from 2.8 to 10.6 g/L during the 3–8 h period (Fig. 3) which allows estimation of an approximate stripping rate. From the measured glucose utilisation rate (average 5.22 g/L.h) and assuming an ethanol yield of 90% of the theoretical maximum gives an ethanol productivity of 2.40 g/L.h (Fig. 4). Based on the total volume within the system and allowing for with the remaining ethanol in the bioreactor gives an effective stripping rate of 0.83 g/L.h for stage II (based on the actual recovery, or 0.88 g/L.h based on assumed yields), which is significantly less than the stripping rates produced in our previous study with pure ethanol-water mixtures (~ 10–20 g/L.h) [26]. In addition to the low concentration driving force for mass transfer, this result is likely due to a combination of several factors. The maximum stripping gas temperature used in this study was 75°C compared to 120°C in our previous work. Additionally, the antifoam agents, salts, cell secretions, waste bread media etc. all affect the mass transfer rate by altering the conditions at the gas-liquid interface [27, 28].
The concentration of ethanol in the first recovered condensate (at 8 h) was high in comparison to the subsequent values (12–24 h), especially considering the low concentration of ethanol in the broth (Fig. 3). This is consistent with previous work[16] and is likely due to the heating of the condenser from the hot, humid outlet of the MSU to the condenser’s steady-state temperature [16]. After this point, and continuing into the start of the fed-batch fermentation stage, the decrease in concentration is due to the decrease in liquid ethanol concentration, as this is a key parameter in the determination of the driving force for mass transfer [29].
Stage III: Fed-batch stage (10–12 h)
At 10 h, a 300 ml batch of concentrated waste bread sugar media was added which increased the sugar content in the reactor to 26.8 g/L. Fermentation of these additional sugars was continued until 12 h when the sugar concentration had decreased to 17.7 g/L. It was also observed that the OD600 dropped from 8.1 at 10 h to 7.1 at 12 h, while ethanol in the fermentation reactor decreased from 8.4 to 5.8 g/l. (Fig. 3) over the same period. The drop in OD can be explained mainly from the effect of dilution by the feed addition, but the reduction in ethanol concentration suggests that the rate of removal exceeded that of production. As the cells had accidentally been starved of sugars at the end of previous stage, it is likely that the surviving cells were gradually recovering and their lower specific sugar consumption rate might have allowed for partially aerobic growth; therefore, this stage would be expected to produce a lower yield of ethanol and was too short to generate meaningful data.
Stage IV and V: Continuous fed-batch stage (12–28 h)
At 12 h, a continuous concentrated waste bread sugar media feed at 1.8 ml/min was initiated marking the transition to the continuous fed-batch fermentation stage, which ran overnight with no sampling and data recording. The lack of regular decanting of bottle B during this period could have resulted in considerable ethanol losses. It is notable that the glucose data measured after 12 h suggested that the cells were recovering from the brief glucose starvation period and, by 24 h, the cell density had increased (note that this was now in a larger volume so not strictly comparable to the value after 8 h). However, the glucose concentration in the culture had increased, indicating that glucose supply was exceeding demand, and by 24 h the concentration exceeded the known toxicity limits for P. thermoglucosidasius, above which the maximum growth rate reduces. Based on the sugars metabolised, at 26 h approximately 54% of the expected ethanol was recovered after removal by the ethanol stripping system. This assumes that cells had returned to fully fermentative growth by 12h.
From the previous analysis of metabolic rates that could be supported by ethanol stripping in this system it was possible that the continuous fed-batch glucose feed rate of 27 g/L.h exceeded the gas-stripping capacity (although it should be noted that this will increase with an increase in ethanol concentration in the reactor). However, after 24 h of culture the ethanol concentration was still below inhibitory levels in the bioreactor despite glucose accumulating. The fact that the ethanol concentration in the condenser bottle after 12 h was relatively low, suggests that production was limited by the effects of the earlier glucose starvation, combined with subsequent glucose toxicity, rather than limited stripping and subsequent ethanol toxicity. Nevertheless, it is notable that the glucose concentration at 28.0 h was lower than that at 24 h suggesting that the spike in glucose concentrations had been transient, and was recovering by 26.5 h. Notably, this was accompanied by an increase in ethanol concentration in the reactor (Fig. 3) and increases in ethanol stripping (Fig. 4). Despite the circuitous route to get to this point, the data points between 24 and 28.0 h act to demonstrate the effectiveness of ex-situ gas-stripping for increasing the productivity of ethanol. Cells were growing at a high cell density with a high production rate of ethanol, as witnessed by the concentrations in the condensate bottles. Yet the ethanol concentration in the culture was below the levels which start to affect the growth of P thermoglucosidasius. The ethanol productivity at stage IV was 1.43 g/L.h, while the effective stripping rate of ethanol was 0.83 g/L.h. These values were affected by the lack of sample collection overnight leading to ethanol losses, which were unaccounted for. Additionally, they probably represent an average value of continued low production after 12h, followed by higher production and stripping after a return to full anaerobic metabolism. Over the period 25 to 28 h the ethanol concentration in the condensate bottle A was very high, ranging between 46–111 g/L, and was therefore, in some cases, more concentrated than the product stream of a traditional fermentation process (87–95 g/L) using baker’s yeast [27, 30].
At 26.45 h the continuous feed was increased from 1.8 ml/min to 2.8 ml/min (56% increase in feeding rate). Interestingly, ethanol productivity remained relatively high and the glucose concentration in the bioreactor fell, suggesting that the culture may have adapted to tolerate the high glucose concentration, a phenomenon which has been reported elsewhere [31]. However, little useful data on the limits of ethanol extraction was obtained beyond this point, so the experiment was terminated soon afterwards. The ethanol productivity for stage V was 4.72 g/L.h while the ethanol stripping rate was 4.80 g/L.h, indicating that the post-starvation recovery and possible adaptation of the bacterial cells was complete, and they were fermenting rapidly. Considering the starting and added sugars during the entire fermentation, we would have expected the ethanol concentration to increase in the reactor and collection vessels to a maximum of 8.24% v/v if all the broth and feed glucose was metabolised (7.42%, if 90% of maximum theoretical).
Figure 4: (a) Ethanol productivity and effective stripping rate by stage calculated based on sugar consumption, assuming a yield of 90% of the theoretical maximum. As glucose was exhausted after 8 h both the average (3–10 h) and actual (3–8 h) productivity are presented for stage II. Stripping rate could only be estimated over the period, 3–10 h. (b) ethanol concentration of individually collected samples from the MSU condenser trap (bottle A) and from the bioreactor off-gas (note that bottle A was emptied into a chilled storage collation bottle after each measurement to avoid losses due to constant gas flow, whereas the concentration in the off-gas was accumulative).
Substrate Mass Balance
Based on the measured volumes of all the constituent components of the system, a sugar mass balance was performed to evaluate the overall productivity of the system. It was assumed that ethanol yield from the fermentation was 90% of the theoretical maximum, a yield typical of industrial bioethanol yeasts from glucose [30]. The overall volume (starting volume with fed broth and removal accounted for) at the end of the experiment, was 4.57 L, with the microbubble removal of an accounted amount of 0.43 L ethanolic liquids collected in bottles, thereby leaving behind an assumed 4.14 L in the reactor. It must be noted that there would have been some additional liquid losses due to evaporation, which are not accounted for in this mass balance, especially with Table 1 showing that the total ethanol accounted for after 28.0 h was only around 51% of that expected when all the added, remaining and consumed sugars had been considered (assuming 90% of the theoretical maximum ethanol yield). This implies an unaccounted “loss” of around 49% of the ethanol, assuming all sugars were converted to alcohol. The total amount of sugar left at the end of fermentation out of a total of 583.35g was 212.26 g, which represents a glucose consumption of 371 g (Table 1). Assuming a yield of 90% of the theoretical maximum based on consumed sugars, this would therefore have produced 170 g of ethanol, which over the entire final volume was equivalent to an overall ethanol concentration of 4.72% (v/v, 37.24 g/L) without stripping, with an overall ethanol productivity of 1.49 g/L.h, taking into account that the start point was at 3 h as this was when the process was switched to fermentative conditions. In comparison, if the initial sugar at 3 h (26.12 g/L, in 2.5 L) was consumed in a batch process by 8hr, the ethanol concentration would have been 1.52% (v/v), corresponding to an overall ethanol productivity of 2.40 g/L.h. The final stage (stage V) of fed-batch fermentation with microbubble stripping showed a consumption and production rate above that which would be possible over a complete batch process and is therefore a significant process improvement. The range of microorganisms tabulated by Azhar et al.[30] had ethanol productivities in the range of 0.17–1.38 g/L.h for batch cultures and the fed-batch process had an overall ethanol productivity of 3.46 g/L.h using a wild strain of S. cerevisiae. Elsewhere, a fermentation with gas stripping process using pretreated sugarcane bagasse as substrate and Kluyveromyces sp. IIPE453 as production organism had a maximum ethanol generation rate of 1.25 g/L.h [32].), Thus, the ethanol generation in both stages II and IV were operating at rates competitive with or higher than other microorganisms. The ethanol recovered at each sampling stage is shown in Fig. 4(b), showing incremental ethanol recovery as fermentation period increases.
Despite the operational issues encountered, it is evident that a fed-batch system with ex-situ hot microbubble stripping can increase ethanol production by P. thermoglucosidasius compared to a simple batch fermentation without ethanol stripping, which would be subject to product inhibition. With tuning of the initial batch fermentation time and fed-batch sugar addition rate, errors in which compromised the rate of metabolism of the organism in this study, a more complete coupled process could be undertaken that would identify the optimum ethanol and sugar steady state concentrations of the coupled processes.
Table 1
Ethanol yields from mass balance after 28 h of fermentation.
*Glucose in the fermentation broth (g), 4.57 L @ 28.0 h | **Ethanol yield expected @ 28 h (90% theoretical) | Ethanol recovered (g) (87.16 g = 51% of the expected) |
total added (t = 0 h) | unused | consumed | total ethanol (g) | ethanol %(v/v) | ethanol (g/L) | condenser (bottle A) | vapour trap (bottle B, 200 ml) | vapour trap (500 ml) | vapour trap (15 ml) | remaining in the broth |
583.35 | 212.26 | 371.29 | 170.28 | 4.72 | 37.24 | 22.19 | 1.00 | 3.62 | 0.32 | 60.03 |
* Final volume of reactor after (total of starting batch, fed-batch and batch additions) |
** Ethanol yields are assumed to be around 90% for P. thermoglucosidasius and industrial yeasts |
Ethanol Mass Balance
Figure 4(b) shows the ethanol concentration of samples collected from the condenser system linked to the MSU and the trap collecting the vapour leaving the bioreactor at each time point over the 5 stages of fermentation. This data was collected to allow a mass balance of ethanol/substrate over each of the stages and the entire process. The ethanol concentration in the vapour traps linked to the MSU condensate bottle and the bioreactor outlet were continuous measurements (i.e. the liquid in the trap was not changed during the experiment) and both the volume and ethanol concentration increased throughout the run.
A relatively slow increase in the volume of the post-MSU condensate liquid trap (30 ml over the entire experiment) demonstrated that the condenser system linked to the MSU captured most of the ethanol and water vapour (results not shown.) The losses from this vapour trap bottle were not measured as this was vented to the atmosphere, which may have reduced the total ethanol recovery figure. The trap placed on the gas outlet of the bioreactor increased by 50 ml over the course of the experiment with a final ethanol concentration of 6.6 g/l (Fig. 4b), giving an average increase of 1.8 ml/h. The smaller, subsequent trapping bottle increased from 15 ml to 19.5 ml. Figure 4b shows that a significant proportion of the ethanol in the post-bioreactor trap was generated in the last few hours of operation (stage V). During the fermentation period of 24–28 h, approx. 8.5 g of ethanol was collected by the condenser (bottle A) and the subsequent 200 ml trap (bottle B), while approx. 2.5 g of ethanol was recovered by the 500 ml and the 15 ml traps connected to the vapour exiting the bioreactor. Although the total volumes recovered from the post-bioreactor trap were small compared to the MSU condensate bottle, it should be noted that the bioreactor gas stream exited via a standard bioreactor condenser cooled by a small chiller unit, which would preferentially recondense water vapour together with ethanol vapour[33]. So the efficiency of this reflux system was relatively poor under conditions of high productivity, and the selective loss of ethanol compared to the liquid phase concentration in the bioreactor, is evident.
Together with the ethanol recovered in the trap after the MSU condensers, that in the post-bioreactor traps can be used to calculate the total ethanol recovered during the experiment. From the sugar mass balance, it has been shown that 170 g (4.72% v/v) of ethanol could have been produced (Table 1) with only 51% of this ethanol accounted for, representing an apparent ethanol loss of 49%. From the problems encountered at the end of the batch phase and start of the fed-batch phase of the fermentation, we know that some of these “losses” are due to aerobic growth (with no ethanol production) in the recovery phase. However, actual ethanol losses are not unexpected, as the concentrations in the liquid traps were high, relative to that in the bioreactor, and the smell of ethanol was evident in the fermentation room. Some ethanol losses during sampling, weighing/measuring and transferring to the collation bottle are also to be expected, in addition to the fact that the final trap bottles were gassed out to the atmosphere. Over the extended time of the experiment, these combined losses will be significant.
Driving force for mass transfer
The calculated average ethanol generation rates with operating parameters can be used to estimate the average ethanol concentration in the MSU (not measured), based on our previous work [26]. The design equation is,
$${V}_{F}{\widehat{P}}_{E}={R}_{cir}\left({C}_{F}-{C}_{MSU}\right)$$
1
where \({V}_{F}\) is the fermenter volume, \({\widehat{P}}_{E}\) is the ethanol productivity, \({R}_{cir}\) is the liquid circulation rate between fermenter and the MSU, and \({C}_{F}\) and \({C}_{MSU}\) are the ethanol concentrations in the fermenter and MSU respectively. The average ethanol concentrations calculated using Eq. (1) are shown in Table 2.
Table 2
Comparison of ethanol concentration in the bioreactor (measured) and MSU (estimated)
Stage | \({\varvec{C}}_{\varvec{F}}\) (g/L, measured) | \({\varvec{C}}_{\varvec{M}\varvec{S}\varvec{U}}\) (g/L, estimated) |
Stage II | 7.27 | 7.10 |
Stage III | 7.10 | 6.83 |
Stage IV | 10.45 | 10.24 |
Stage V | 14.80 | 13.95 |
Table 2 shows that the average ethanol concentrations in the bioreactor and MSU were close at all times. Note that Eq. (1) is designed for steady-state operation and these estimates are averages over the entire stage rather than instantaneous values. This demonstrates that the circulation rate set between the two vessels was sufficient to maintain a concentration driving inside the MSU to promote rapid mass transfer.
Moving forward, even though we have been able to demonstrate ethanol production above the minimum economic levels of 4% v/v suggested by Lynd et al [7] with the aid of microbubble extraction, the repeatability and robustness of these results should be investigated, and in a manner where large data gaps are not present to allow for the analysis of what caused the bottleneck in the process. Some tuning should also be undertaken, as the ethanol concentration in the bioreactor for large parts of the investigation (up to 26 h) was lower than the optimum of 1.6% (v/v, 12.3 g/L). Additionally, the sugar addition rate should be reduced in line with the glucose consumption rate or continually adjusted using a control system to remain under the toxicity level. Finally, while this process is beneficial in that it produced ethanol from a waste product (waste-bread), the effects of using pretreated lignocellulose should be investigated and optimised as necessary.