3.1. Batch fermentation
The growth kinetics of E. mundtii QU 25 on glucose, cellobiose, and xylose were studied by batch experiments. The batch fermentations were conducted using the medium supplemented with varying concentrations of sugars (10, 20, 50, 100 and 150 g∙L-1) and the growth kinetic parameters were determined during the early exponential growth phase with no product inhibition occurred. The optimized values of kinetic parameters determined by nonlinear regression are listed in Table 1. The maximum growth rates (mmax) of the studied strain were 1.20 h-1, 0.99 h-1, and 0.62 h-1 for glucose, cellobiose, and xylose, respectively, which were within similar range for other lactic acid producing strains using different substrates [23-28]. The higher mmax and Ks for glucose than xylose and cellobiose implied the preferred carbon source of the strain QU 25, which was further confirmed by the residual sugars in CSFTR.
The cell growth and lactic acid formation were critically affected by the product inhibition (Kip), while the substrate inhibition (Kis) had relatively small adverse impacts on the studied sugars. A linear relationship was determined between substrate consumption and lactic acid formation with a yield coefficient of 1.0025 for all three sugars (details not shown), which is similar to the calculated yield of lactic acid based on stoichiometry. The yield of lactic acid was determined from the stoichiometric equations and batch kinetic date with values of 0.950 g·g-1 for glucose, 1.015 g·g-1 for xylose, and 0.925 g·g-1 for cellobiose. The cell yields were at low range (0.028-0.056 g·g-1) which could be reasonable if consider the high product yields (0.925‒1.015 g·g-1). The model parameters were evaluated through dynamic simulations of batch fermentation; and the results of sugar utilization, lactic acid production, and cell growth from three single sugars were demonstrated in Fig.3. The simulation results (lines) all fit well with experimental results (symbols).
In addition, dynamic simulations were also carried out on batch fermentation with mixed sugars (Fig.4). The impacts of CCR was clearly shown in G100X50 (Fig.4(a)), of which the majority of xylose was not consumed as glucose was the preferential carbon source in the synthetic hydrolysate. Large amount of xylose remained in the fermentation broth even when glucose was almost completely converted into lactic acid and cell biomass. On the other hand, CCR was not observed when C100X50 was applied (Fig.4(b)) as both sugars were utilized by the fermentation strain simultaneously during fermentation. However, large amount of cellobiose was not utilized during the testing period when similar amount of cells was produced in the batch system. Lactic acid concentration was slightly higher in C100X50 than in G100X50.
Simulation results of the duel sugar batch systems showed reasonable fits with the experimental results. The simulation of C100X50 was carried out by the dynamic model described in the methodology section, and without inclusion of any inhibiting factor from the co-fermenting sugars. In order to characterize CCR, the simulation of G100X50 was carried out by introducing an inhibiting coefficient from glucose to xylose in Equation (6). This simulation approach was designed only for this application as the specific range of glucose concentrations to induce CCR was not clarified. The sensitivity of the inhibiting functions needs to be validated with more comprehensive experiments, which is beyond the scope of this study. As the main focus of this work is to investigate the fermentation of cellobiose and xylose, we tend to study the inhibiting kinetics of CCR in greater details elsewhere.
3.2. Continuous co-fermentation
This experiment investigated the feature of continuous co-fermentation process for complete utilization of hexose and pentose. G100X60 and C100X50 were initially tested at a dilution rate of 0.2 h-1 for process control and comparison (Table 2). For G100X60, when the continuous fermentation reached a steady state, a cell concentration of 1.99 g·L-1 was achieved and much higher amount of glucose was consumed than xylose (27.5 g·L-1 over 2.64 g·L-1, respectively). The ratio of consumed glucose to xylose was 10.4:1, which exhibited an obvious CCR for xylose utilization. E. mundtii QU 25 growing on G100X60 produced 24.2 g·L-1 lactic acid with a yield of 0.803 g·g-1 and productivity of 4.84 g·L-1·h-1; and small amount of by-product (0.151 g·L-1 acetic acid) was detected. When glucose was replaced by cellobiose (C100X60) at the same dilution rate of 0.20 h-1, the cell concentration increased to 2.42 g·L-1 at steady state. A cellobiose consumption of 21.8 g·L-1 was achieved with a higher xylose consumption of 11.7 g·L-1 than at G100X50. The ratio of consumed cellobiose to xylose reduced to 1.86:1, suggesting the relief of CCR for xylose consumption. The continuous fermentation process feeding with C100X60 demonstrated a homofermentative lactic acid production of 27.5 g·L-1 with yield of 0.820 g·g-1 and productivity of 5.49 g·L-1·h-1.
To confirm the beneficial effects of C100X60co-fermentation over G100X60, the sugar mixture was changed back to G100X60 after the test of C100X50. Similar parameters were obtained comparing with the first cycle (Table 2). The mixed sugars should be the decisive factor in continuous co-fermentation. Feeding with C100X60 achieved simultaneous sugar utilizations in continuous co-fermentation at a high productivity.
3.3. Effects of dilution rates on continuous co-fermentation
Dilution rate is a critical control parameter for maximizing productivity in continuous fermentation process. In this study, continuous co-fermentations of C100X60 were performed at increased dilution rates from 0.05 to 0.25 h-1 (0.05 h-1, intervals) and the results were presented in Table 3 (top). The cell formation increased with the dilution rate between 0.05 h-1 to 0.20 h-1, which achieved a maximum value of 2.57 g·L-1 at 0.20 h-1. Further increase in dilution rate to 0.25 h-1 resulted in decreased cell concentration to 1.74 g·L-1. Total sugar consumption was ranged 28.5-33.7 g·L-1 and lactic acid production was 20.8-27.9 g·L-1 at all dilution rates. The consumption ratio of cellobiose to xylose was lower than 1.97:1, among which reduced to 1.84:1 at dilution rate of 0.20 h-1. The highest productivity of 5.37 g·L-1·h-1 with 26.9 g·L-1 lactic acid was obtained at the dilution rate of 0.20 h-1, which was considered optimal for C100X60 utilizations by E. mundtii QU 25 in continuous mode. The highest productivity obtained in this study was much higher than 0.635 g·L-1·h-1 in batch experiments [13]. However, high residual cellobiose of 77.2 g·L-1 and xylose of 50.3 g·L-1 were observed and hence further investigations were carried out to decrease the residual sugars.
C50X30 was investigated at similar fashion of continuous co-fermentation for increasing dilution rates of 0.05-0.35 h-1 as the data presented in Table 3 (bottom). The cell concentration increased with dilution rate from 1.06 g·L-1 (at D=0.05 h-1) to 2.42 g·L-1 (at D=0.25 h-1); but further increase of dilution rates to 0.30 h-1 and 0.35 h-1 resulted in lower cell concentrations of 2.29 and 2.16 g·L-1, respectively. The limiting cell retention time (inverse of the dilution rates) of the strain in the CFSTR was approximately 3-4 hours. Consumption ratio of cellobiose to xylose was almost similar at dilution rates of 0.05-0.20 h-1 ranged 2.15-2.43:1, respectively. When the dilution rate was higher than 0.20 h-1, the sugar consumption ratio increased dramatically from 2.17:1 (at D=0.20 h-1) to 3.29:1 (at D=0.25 h-1), implying a critical condition among cell concentration, sugar compositions, and increase of CCR for xylose utilization. The lactic acid productivity increased with the increase of dilution rate until D=0.30 h-1, of which a maximum value of 6.52 g·L-1·h-1 was reached.
As the co-fermentation experiments were carried out continuously with changes of influent conditions adjusted over well-controlled retention time, the experiment results served well as examples of dynamic simulations of the mathematical model. The dynamic records of the experiments and the corresponding simulation results were presented in Fig.5. Effluent cellulose, xylose, lactic acid, and cell concentrations were shown in four different rows of the subfigures; and the two columns represent the influent sugars combinations of C100X50 (left) and C50X30 (right). D1- D7 represents the tested dilution rates (0.05 – 0.35 h-1) and the data with multiple points at the transition phase were taken when steady states at a dilution factor were achieved.
In general, the model described reasonably well the dynamic status of the measured parameters, especially on the predictions of immediate changes of the substrates and products concentrations between batch and continuous modes. However, some limitations were also discovered for further improvement of the experiments and model structure. In the experiments, cellobiose, xylose, and lactic acid concentrations were relatively at constant levels after the first dilution rate (D1), but obvious increases of cell concentration were observed from D1 to D4 (onward) for both experiments. Before the cell started to be diluted, regional peaks can be found at specific dilution rates, i.e., at the end of D4 for C100X50 and D5 for C50X30, respectively. In fact, sugar consumptions and lactic acid productions also followed this pattern but the covered ranges were less significant. The changes in cell concentrations over dilution rates were not predicted by the simulation.
Based on Mono-kinetics, cell concentrations increase with the related reproduction on the substrate. This increase is compensated by cell decay and “wash-off” effects due to high dilution rates in CSFTR. As the cell yield coefficients determined in the batch system were quite low, the simulated cell concentration in the process should be a continuous decrease over the increase of dilution rates. This uncertainty between the model and experiment results may be due to the incomplete cell suspension in the fermentation broth or other uncharacterized factors in the model. The fermentor used in this study is a typical cylindrical column container with mechanical stirrer installed through the reactor from the top. The CSFTR was controlled by pumping the same amount of liquid in and out of the system. The cell samples were collected through a sampling pipe link to the bottom of reactor. When performing long term continuous experiment, the fermenting cells may not be completely suspended in the fermentation broth, or consistently discharged with the liquid effluent. Slightly higher cell concentration may exist in bottom part of the jar which resulted in the inconsistency. Regardless, clarification of this issue need further investigation and was not significant when the fermentation cells were completely retained in the fermentation process, as detailed in the next section.
3.4. Continuous fermentation with cell recycle (CF/CR)
Controlling cell concentration through cell recycle has demonstrated to be an efficient technique to obtain high cell density and lactic acid productivity in continuous fermentation with single carbon, i.e., glucose [17] and starch [18]. The CF/CR process were performed after receiving concentrated cell concentration from 4-L reactor, with mMRS medium containing C50X30 at pH of 7.0 and dilution rate of 0.2 h–1 (Fig. 6). Approximately 15-fold higher cells (33.6 g·L-1) and 2-fold lower residual xylose concentration (4.71 g·L-1) in the fermentation broth were achieved in comparison to the processes with no cell recycle. Sugar consumption ratio of cellobiose to xylose was 1.92:1 in CF/CR compared 2.17:1 in conventional mode under the same dilution rate. High lactic acid concentration of 65.2 g·L-1 and productivity of 13.03 g·L-1·h-1 were obtained with slightly lower lactic acid yield (0.854 g·g-1), compared to 22.9 g·L-1, 4.57 g·L-1·h-1, and 0.871 g·g-1 without cell recycle, respectively. Minimal byproducts of 0.02-1.97 g·L-1 acetic acid, 0.26-1.93 g·L-1 formic acid, and 0-1.65 g·L-1 ethanol were produced. The fermentation strategy demonstrated outstanding productivity, end-product concentrations, and consumptions of mixed sugars toward more feasible applications. Abdel-Rahman et al. also showed that a high lactic acid yield of 0.912 g·g-1 and productivity of 6.49 g·L-1·h-1 at a cellobiose/xylose continuous fermentation with cell recycle at a dilution rate of 0.1 h-1 [19]
The dynamic changes of cellobiose, xylose, lactic acid, and cell concentrations in the CF/CR process (symbols, including the operation at batch mode) and the simulation results (lines) were presented in Fig.6(a) through Fig.6(d), respectively. Significant consumptions of cellobiose and xylose were shown at the beginning phase of batch system (from 0 to 24th hour), which associated with corresponding growth of lactic acid production and slight increase of cells. Xylose was not completely consumed and approximately 10 g·L-1 residual sugar exited the system at 24th-36th hour before cell injection. The concentrated cells were introduced 15 hours before the process was changed to the continuous mode. The CF/CR was functioned properly with consistent reduction of residual sugars and increase of lactic acid/cell concentrations.
The simulation results showed outstanding characterization of the process conditions over the whole experiments. It accurately predicted the consumption of cellulose at batch mode and the overall statuses of the components in the continuous mode. While no measurements were conducted during the transition period (from the 39th to the 54th hours), the model simulated the degradation of sugars due to significant increase of cells, as no additional sugars were introduced in the reactor. During the transition period, the cell concentration declined considerably due to decay, and then increased again in the continuous process when sugars were again introduced in the CF/CR. Although the CCR on the xylose consumption during the batch mode was not simulated (Fig.6(b), hours 24-36), this model showed high sensitivity handling the flow condition changes, cell growth, and cell retention problems.
3.5. Importance of cell retention time (CRT)
In summary of all the experiment results collected in this study for C50X30, the relationships of cellobiose, xylose, lactic acid, and cell concentrations were plotted in Fig.7. CRTs did show critical impacts to the continuous process. With the increase of CRT, the fermentation strain with a high density is more effective utilizing the sugars and may be more robust reflecting to the changing properties of the hydrolysate. The benefits of CRT control have been demonstrated in many biological systems, i.e., to regulate the consumption rates of various carbon sources [20] and real-time gas phase monitoring for optimal cells metabolisms [21].
Meanwhile, it should be emphasized the difference of the commonly applied control parameter dilution factors (1/HRT) used in the conventional fermentation process over the factor for cell retention (1/CRT). As the cells have been recovered from the liquid stream, the CRT of the CF/CR must be longer than HRT, and this expression applies to all the other biological systems such as bioaugmentation or fed-batch fermentation.
To better visualize the potential applications of the long CRT operation, we derived the steady state expressions of the sugars and cell via Equations (7) and (8) as follows:
The results of the steady state expressions were plotted against the experiment results in Fig.7, which also summarize the potential benefits and issues of the model. The model clearly demonstrated the possibility of CF/CR operation for process control and optimization. For instance, the important wash-off period at low CRT operation on residual sugars and delayed cell growth was shown in the model, suggesting a potential issue of continuous operation that was not observed in batch experiments. As the fermentation experiments at short CRT were prepared with the well-inoculated seed strains, the challenges of low CRT operation were not observed in this work. The important degradation of cell due to cell decay and complete consumption of xylose at extremely long CRT, has not been characterized due to experimental design. The impacts of rapid sugar uptake and delayed growth, or CCR were not characterized due to current model structure and limited simulation parameters. Those uncertain points would rely on further investigations of the system as future works.