Simultaneous saccharification and fermentation (SSF)
As a first step towards consolidated bioprocessing, itaconic acid production was assessed in a SSF setup to evaluate the capability of the engineered U. maydis Δcyp3 ΔPria1::Petef Δfuz7 PetefmttA to produce itaconic acid under glucose limiting conditions. Thereby, fermentation supernatants of two different cellulolytic fungi (T. reesei RUT-C30 (RFP1) and Penicillium verruculosum M28-10) were compared as source for cellulases. These fungi have been found compatible with the presence of itaconic acid in a previous screening (49). Since nitrogen limitation is a prerequisite for itaconic acid production with U. maydis, the residual NH4+-concentration present in the cellulase-containing supernatants had to be monitored (50). The supernatants were diluted accordingly to reach a final NH4+-concentration equivalent to the 0.8 g/L NH4Cl, typically present in itaconic acid production medium for U. maydis (26, 28, 29). This resulted in a final cellulase titer of 2.2 FPU/mL for the cultivations containing T. reesei supernatant and 0.6 FPU/mL for the cultivations containing P. verruculosum supernatant. These values corresponded to 18 and 5 FPU/g cellulose, respectively, and are close to the cellulase loading of 10 FPU/g cellulose, typically used for separate cellulose hydrolysis (51). Since cellulose digestibility has a major impact on cellulose hydrolysis, 120 g/L of highly digestible Sigmacell as well as 120 g/L recalcitrant α-cellulose were tested as substrates (52). All media were buffered to pH 6.5 using 100 mM MES buffer.
As shown in Fig. 2, both T. reesei and P. verruculosum supernatants enabled itaconic acid production from cellulose by U. maydis during SSF. Using the T. reesei supernatant, similar itaconic acid concentrations as in a reference culture containing 50 g/L of glucose instead of cellulose were reached. The highest itaconic acid concentration achieved during the SSF was 21 g/L using Sigmacell as substrate, which is 4-fold higher than previously demonstrated for a SSF using wildtype U. maydis MB215, and close to the values achieved with A. terreus (Table 1). Even using the relatively recalcitrant α-cellulose, a similar titer of 19 g/L itaconic acid was produced. Although the itaconic acid production using the P. verruculosum supernatant was generally lower than using T. reesei supernatant, especially with α-cellulose, the achieved titers were still considerable in relation to the almost 4-fold lower cellulase activity of 0.6 FPU/mL. Remarkably, the itaconic acid yield based on the consumed amount of glucose equivalents from cellulose was essentially identical to the yield achieved using glucose in the reference and also comparable to yields achieved with A. terreus using purified cellulose hydrolysate (Table 1).
From the pH profile, it can be seen that in most of the cultures the pH dropped to about 3.5 after 69 h of cultivation. Since it is known that U. maydis stops itaconic acid production at such low pH values (53), it is possible that higher itaconic acid titers could have been achieved using a higher buffer concentration or active pH control, which would also have further improved the cellulose hydrolysis. This hypothesis was confirmed by comparing the itaconic acid production of glucose-supplemented SSF cultures either buffered with MES or with CaCO3 (Supplementary Fig. S1). In case of the MES-buffered culture, itaconic acid production stopped before the exhaustion of glucose when reaching pH 3.5. In contrast, the CaCO3-buffered culture that maintained pH between 6 and 7 continued itaconic acid production after exhaustion of the added glucose and further converted cellulose into itaconic acid. An added benefit of the CaCO3 addition is an in situ precipitation of the product as calcium itaconate, which alleviates product inhibition and facilitates downstream processing (28). Calcium salt precipitation still belongs to the most mature and widely applied downstream technology for industrial organic acid production (28, 54).
Consolidated bioprocessing (CBP) with co-cultures of T. reesei and U. maydis
As a next step, itaconic acid production was assessed in a CBP setup with a sequential co-culture of U. maydis and T. reesei. First, T. reesei was grown for one week in pure culture to produce sufficient cellulase enzymes, whereafter U. maydis was added to an OD of 0.67. To prevent a termination of itaconic acid production due to a decreasing pH, the medium was buffered with 33 g/L CaCO3. The performance of the CBP culture was directly compared to a corresponding SSF culture using undiluted sterile filtered supernatant of the same T. reesei pre-culture used for CBP. A schematical representation of the experimental setup is depicted in Figure 3.
As can be seen in Figure 4A, up to 10.5 g/L of itaconic acid was produced in the co-culture CBP. The SSF in contrast produced up to 15.3 g/L of itaconic acid. Thus, the CBP was clearly less effective in producing itaconic acid than the SSF. This could be already expected since two organisms have to share the same resources. The cellulose consumption (Figure 4E) was clearly higher in the CBP compared to the SSF, especially during the initial 72h growth phase. Since all conditions are identical between CBP and SSF except for the presence of living T. reesei cells, this increase in cellulose consumption can be clearly attributed to T. reesei. In turn, also the itaconic acid yield was affected by the CBP in comparison to the SSF. While a yield of 0.381 g/g Glucose was reached in the SSF, only 0.134 g/g Glucose were achieved in the CBP.
On the positive side, this increased cellulose consumption also demonstrates an enhanced cellulose hydrolysis performance with increased metabolic substrate demand. A recent publication on cellulosic malic acid production demonstrated that an increase in metabolic activity can drastically enhance cellulose hydrolysis without increasing cellulase concentration (55). Therefore, the main challenge for optimizing the CBP setup is to channel the substrate consumption towards U. maydis and minimize the activity of T. reesei during the itaconic acid production phase. Still, some low residual activity of T. reesei could be beneficial, since the cellulase activity was more stable in the CBP compared to the SSF (Fig. 4C, D). This was likely due to constant re-synthesis of degraded cellulases by T. reesei.
The population dynamics between U. maydis and T. reesei were not investigated in detail. Still, microscopic examination of the samples suggested a relatively balanced population ratio between T. reesei and U. maydis towards the end of the culture. There were always cells visible from both T. reesei and U. maydis in randomly chosen fields of view (Figure 5). The nitrogen supply in the cultures had to be limited in order to induce itaconic acid production by U. maydis. As a result of this limited nitrogen availability and competition with cellulase synthesis by T. reesei, both organisms could only grow until the shared nitrogen pool was exhausted. Hence, the U. maydis cell number per field of view was clearly lower in CBP compared to SSF, which might explain the lower itaconic acid productivity. If the nitrogen consumption of T. reesei in the CBP would have been compensated by appropriate addition of NH4+, similar U. maydis cell densities and itaconic acid productivities might have been reached as in the SSF.
The major benefit of CBP in contrast to SSF is that no enzymes had to be added and that cellulose can directly be converted into itaconic acid. Since the targeted substrates are cellulosic waste streams, which have very low cost, the production yield is less important. It has to be evaluated whether the economic benefits of the CBP process can compensate the yield loss. The outcome of this will most likely depend stongly on the price of cellulase enzyme production and cellulosic substrate.
Influence of inoculation time
During the proof of principle CBP described above, T. reesei was cultured for one week in pure culture to produce sufficient cellulase enzymes before adding U. maydis. However, a one week pre-cultivation phase considerably lowers the average productivity of the co-culture compared to the SSF scenario (Table 2). To further optimize the co-culture and increase the productivity, the effect of inoculation delay between T. reesei and U. maydis on the CBP performance was studied. Four different additional inoculation delays were tested: 0 days (direct co-inoculation at the beginning), 3, 4, and 5 days. Because cellulase production by T. reesei usually starts after 2 days, it was expected that preliminary inoculation of U. maydis would strongly affect cellulase production due to competition for nitrogen. Therefore, 1 and 2 days delay were not tested. The experiment was performed in medium containing 5 g/L glucose for initial growth acceleration and 30 g/L α-cellulose. Furthermore, the experiment was carried out as fed-batch with regular feeding of α-cellulose powder for maximization of itaconic acid production.
When T. reesei and U. maydis were co-inoculated, the culture did not produce any cellulase enzymes nor any itaconic acid. Instead, exclusively U. maydis grew and consumed all glucose before T. reesei was able to germinate, thereby preventing the production of cellulases that would have enabled further growth of both organisms on cellulose. Because of the limited initial glucose concentration, a nitrogen limitation could not be reached, which explains the lack of itaconic acid production. Co-inoculation with vital cells of T. reesei instead of T. reesei spores might solve this problem since a starting quantity of cellulases would be present that would allow the co-culture to grow on cellulose instead of collapsing.
When U. maydis was added to the T. reesei culture after cellulase production had already started, itaconic acid production was successful. The influence of the different inoculation delays on both cellulase and itaconic acid production was surprisingly low (Fig. 6). Cellulase production of T. reesei is directly proportional to the available concentration of the nitrogen source. Because of this, it would have been expected that earlier inoculation of U. maydis would reduce cellulase production because of the competition for nitrogen. In this case, due to an earlier limitation of nitrogen also an earlier onset of itaconic acid production would have been expected. However, this was clearly not the case, suggesting only a very low growth and nitrogen consumption by U. maydis. To analyse the growth of U. maydis, a differential fluorescent staining technique was developed to clearly discriminate U. maydis cells from T. reesei and thereby allow for manual cell counting of the U. maydis population (see material and methods). For the cultures with 3 and 4 days inoculation delay, U. maydis only grew very slowly before the first cellulose feed. Starting from an initial inoculation density of 0.8∙107 cells/mL (corresponding to a measured OD of 1.14), only a cell density of 2.7∙107 and 1.7∙107 cells/mL within 46 h and 25 h was reached, respectively, although the medium contained still all nutrients necessary for growth (Fig. 6 E, F, G). The extent of growth inhibition becomes clearer in comparison to the growth of U. maydis in the instantly co-inoculated experiment. Here, U. maydis was able to grow to a cell density of 25∙107 cells/mL in 24 h with just 5 g/L of glucose. For the experiments with inoculation delay, significant growth of U. maydis was only evident after increasing the carbon supply by feeding additional cellulose (Supplementary figure S2). These phenomena may be explained by a higher affinity for sugars of T. reesei compared to U. maydis, enabling the former to out-compete the latter under sugar limitation. For reference, we observed µmax values of 0.17 and 0.21 for T. reesei and U. maydis, respectively, when grown in pure culture at pH 7 in MES buffered glucose media. Hence, under non-limited conditions U. maydis is the faster growing of both organisms. By feeding cellulose, the sugar concentration is increased, which enabled U. maydis to grow much faster.
Additionally, besides the temporal effect of the inoculation delay, there was also an unexpected viability effect that should have caused a growth benefit for U. maydis in the early inoculated experiments compared to the late inoculated experiments. The U. maydis inoculum was prepared from a YPD medium grown pre-culture that was washed twice with bi-distilled water and then stored as aqueous cell suspension at 0°C for the different inoculation time points, so that the same stock could be used for all tested conditions. The viability of the aqueous inoculum was monitored for each inoculation time point by always inoculating a YPD medium filled flask in parallel to the CBP cultures and recording growth using online scattered light measurements. By observing an increase in the lag time with the age of the aqueous inoculum, it became clear that the viable cell density dropped significantly over time (Supplementary Figure S3). The increase in lag time by more than one doubling time (3.5 h) suggests at least a 2-fold difference in viable cell density. Despite the drop in viability, there was no impact on CBP performance. Thus, the sugar supply rate was the key factor determining population dynamics and sugar partitioning between T. reesei and U. maydis, while neither the inoculation delay nor inoculation density of U. maydis had a significant effect.
Also regarding itaconic acid productivity, the sugar supply rate (and thus, the cellulose hydrolysis rate) was most likely the limiting factor. For a glucose-based itaconic acid production reference, typically a cell density of 50∙107 cells/mL and a productivity of 0.77 g/L/h are reached under the investigated conditions (Supplementary Figure S4). Since the determined U. maydis cell density during CBP ranged from 10∙107 to 30∙107 cells/mL, a theoretical itaconic acid production capacity of 0.16 to 0.46 g/L/h was present in the CBP. The fact that only a maximum productivity of 0.10 g/L/h was reached in the fed-batch CBP indicates that the cells were not producing itaconate at maximum capacity, likely due to the abovementioned competition for glucose.
As envisaged, the total average productivity (including the cellulase production phase) in this fed-batch experiment was indeed higher than in the previous batch experiment with 7 day inoculation delay (Table 2). However, this effect was not related exclusively to the smaller inoculation delay or an earlier onset of itaconic acid production. Instead, the productivity was generally slightly higher and was sustained for a longer period, so that the influence of the cellulase production phase duration on the total average itaconic acid productivity was minimized. This was due to the regular feeding of cellulose and thus mainly a benefit of fed-batch fermentation in contrast to batch fermentation. The key factor controlling the start of itaconic acid production was the time point of the first cellulose feeding, which in all cases was synchronized to 5 days after start of the experiment and thereby likely also synchronized the itaconic acid production. This feeding regime was chosen because a preliminary mass feeding of cellulose would have compromised cellulase production by T. reesei. An earlier start of itaconic acid production could therefore be at the expense of cellulase activity.
Detailed online process monitoring during co-culture CBP using online respiratory analysis
Up to now, consortium based CBP has been proven successful only in academic research. One major obstacle for industrial application of such processes is a lack of suitable and established process control techniques. The substrate consumption for example is very difficult to assess in complex cultures containing solid cellulose particles. Here, the respiration rate of the fed-batch CBP was monitored online as a direct measure for microbial activity. As described in earlier studies (49, 56), cellulose consumption and thereby suitable time points for feeding fresh substrate could be estimated online based on the cumulative oxygen consumption. This way, intermittent starvation of the culture could be prevented (Figure 7A, C).
The use of online respiration measurement further allowed to estimate the product formation online. Assuming the pure aerobic combustion of glucose as only carbon source for maintenance and biomass formation, the respiratory quotient can be considered to be 1 (Figure 1, pathway C). During phases of significant formation of partially oxidized products such as itaconic acid, which is typically produced at a yield of 0.4 g/g in U. maydis, the RQ should be even below 1. Pure itaconic acid formation would cause an RQ of 0.67 (Figure 1, pathway A). However, since CaCO3 was used for buffering of the cultures, the production of 1 mol itaconic acid would release an additional mole of CO2 by reacting with the carbonate, thereby increasing the RQ above 1 (Figure 1, pathways A-B, blue values). This effect is depicted more clearly in Supplementary Figure S5, comparing the theoretical RQ in absence or in presence of CaCO3 in relation to the itaconic acid yield. Hence, an RQ above 1 in presence of CaCO3 indicates that itaconic acid production is ongoing, while a drop of RQ close to 1 indicates cessation of itaconic acid production. During the experiment, RQ values well above the theoretically expected values have been measured, because the used strain still produces considerable amounts of reduced by-products such as glycolipids, which result in an increased RQ (25, 57). An explanation for this phenomenon is illustrated in Figure S5.
The profile of the RQ over time thereby serves as an indicator for the time-resolved itaconic acid yield during the process (Fig 7B). While the global average itaconic acid yield was only 0.16 g/g, this averaged value likely results from dynamic fluctuations of phases with high yield (0.4 g/g) and phases without itaconic acid production. The data suggest that the actual itaconic acid yield was highly dependent on the cellulose concentration and thus sugar supply rate. After each feeding, the RQ went up to a maximum of about 1.3 before gradually decreasing to values close to or even below 1, where itaconic acid production likely stopped (Fig 7B). When a surplus of substrate was available after a cellulose pulse, the yield approached values typically observed with glucose based fermentations, but then dropped as the substrate depleted. Thus, when the substrate supply was high, itaconic acid production was the dominant process, while respiration of T. reesei and U. maydis became more dominant with low substrate supply. This implies that a certain threshold substrate supply rate has to be sustained in order to enable itaconic acid production and that the maximum itaconic acid yield is only achieved under carbon non-limited conditions. This observation fits well to the observed influence of sugar supply rate on substrate partitioning and population dynamics between T. reesei and U. maydis observed at the early phase of the fermentation.
Assuming i) a baseline RQ of 1 for both T. reesei and U. maydis, ii) that any surplus CO2 is derived from acid reaction with CaCO3 and iii) that itaconic acid is the only acid formed, the amount of itaconic acid produced can be directly estimated from the difference between cumulated CO2 evolution and the cumulated O2 uptake. Fig. 7D shows a comparison of itaconic acid production and online estimation based on respiration measurement. The axes are scaled in a way that 130 g/L itaconic acid equals 1 mol/L CO2 difference, hence 1 mol CO2 per mol itaconic acid (130 g/mol). Although the online estimation did not fit exactly to the HPLC measurements, the method can give an approximation and a clear trend of itaconic acid production. Underestimation can be explained by either production of reduced storage molecules, CO2 fixation or non-respirative oxidation reactions while over-estimation is most likely a result of other organic acid by-products, which have been observed in the CBP cultures but not during itaconic acid production from glucose (Supplementary Table S6). The online estimation suggests that the time resolved itaconic productivity (the slopes in Fig. 7D) reached values up to 2.2 mmol/L/h (or 0.29 g/L/h) after each cellulose feeding, which implies that the actual itaconic acid production potential present in the CPB was considerably higher than the achieved averaged value of 0.07 g/L/h. These data suggest that if the duration of the feeding intervals of the cellulose are reduced, the space-time-yield could be increased up to four-fold, reducing the total process time to 240 h. The best way to optimize the process would be an RQ-controlled automatic feeding of cellulose powder.
Techno-economic perspective
Nieder-Heitmann et al. performed a techno-economic analysis to compare sugar cane bagasse-based itaconic acid production via SHF toclassical glucose-based production. Assuming an overall conversion yield of cellulose-rich sugarcane bagasse to itaconic acid between 0.086 and 0.188 g/g, a titer of 147 g/L and a productivity of 1.15 g/L/h, cellulosic itaconic acid would be already on-par or superior to glucose based itaconic acid (58). While the calculated minimum selling price was most sensitive to yield and productivity, the titer was less important. It was shown that major economic burdens of the cellulosic SHF process were the high investment costs related to additional production facilities for separate hydrolysis and enzyme production as well as the energy requirement for the concentration of raw hydrolysates to the high sugar concentrations necessary for efficient itaconic acid production with A. terreus. Both of these major cost drivers were neutralized in the here presented SSF and CBP processes. Additional cost savings can be expected for the downstream processing as the itaconic acid precipitates in situ during fermentation, which makes product isolation less energy intense (54). While the overall cellulose-to-itaconic acid conversion yield of 0.17 g/g in the fed-batch CBP already fits well into the economically viable scenario of Nieder-Heitmann et al., the productivity still needs to be significantly improved. However, a detailed techno-economic calculation for our case will be necessary to determine the minimum viable productivity, as cost savings from lower energy demand and lower investment costs will likely compensate to some degree for a lower productivity.