The XI from C. phytofermentans was the first prokaryotic XI that showed high activity upon expression in Saccharomyces cerevisiae, both in laboratory and industrial strains . However, the industrial strain expressing the codon-optimized version of the gene could only ferment D-xylose to ethanol after further evolutionary adaptation in D-xylose medium. Though the rate of D-xylose utilization by the evolved strain was much too low to allow industrially viable ethanol production from lignocellulosic feedstocks, the work provided a starting point for the development of strains using a bacterial XI that was less inhibited by xylitol than the fungal Piromyces XI. In the present work we combined rational metabolic engineering based on expression of the C. phytofermentans XI with systematic evolutionary engineering, and developed a robust industrial S. cerevisiae strain that efficiently converts D-xylose to ethanol with high yield and productivity.
Rational metabolic engineering alone was not able to establish efficient D-xylose (or L-arabinose) utilization capacity in the Ethanol Red strain. The recombinant strain HDY.GUF5 failed to show significant D-xylose (or L-arabinose) fermentation. Efficient rational engineering strategies rely on the complete understanding of the metabolic network, as well as its regulation in response to the dynamic environmental conditions to which the engineered strain is exposed . Because of the complexity and still limited understanding of the biological and regulatory network of D-xylose metabolism in recombinant S. cerevisiae strains, rational approaches have faced huge challenges to eliminate the factors that limit efficient D-xylose fermentation . Several of these factors have been identified [31, 52]. Most of these requirements have been addressed in the strain HDY.GUF5, which include overexpression of the PPP genes, XKS1 and the hexose/pentose transporter encoding gene HXT7, as well as codon optimization of some of the genes based on the highly efficient glycolytic codon usage of yeast . Although expression of the same codon-optimized XI in a laboratory strain established moderate D-xylose fermentation , the industrial strain used in this study as well as previously , was not able to metabolize D-xylose. This is likely due to the difference in the genetic background of the strains, although the precise mechanism remains unclear .
Combining metabolic engineering with evolutionary engineering alone or together with random mutagenesis has been proven successful for developing strains with improved D-xylose fermentation efficiency [20, 29, 51, 54, 55]. In addition, genome shuffling has also been used in combination with metabolic engineering and evolutionary adaptation, for improving D-xylose utilization capacity in different S. cerevisiae strains [56, 57]. In the present paper, we successfully exploited a combinatorial approach using all three random strain improvement strategies described above, in order to improve D-xylose fermentation efficiency of the recombinant industrial strain HDY.GUF5.
We first started with random mutagenesis of the recombinant strain to generate very diverse genetic variation that might establish initial D-xylose fermentation capacity. Selection of mutants with a significant D-xylose anaerobic fermentation rate is a challenging task, because likely multiple mutations are required . In addition, a previous study reported that direct selection of a mutant S. cerevisiae population capable of anaerobic D-xylose utilization, was unsuccessful . Therefore, we first selected clones from a heavily-mutagenized population that were able to grow at least to some extent on D-xylose medium as a sole carbon source. Since strong random mutagenesis likely results in both beneficial and deleterious mutations, we presumed that genetic recombination of the mutants obtained, with the original industrial strain by genome shuffling, and selection for D-xylose utilization capacity, would result in enrichment of beneficial and loss of unfavorable mutations . After only one step of genome shuffling, the whole shuffled culture already demonstrated a significantly improved rate of D-xylose fermentation. However, attempts to isolate single cell clones from this shuffled culture with better D-xylose utilization rate, compared to that of the best mutant strain M315, failed. Thus, we decided to enrich the clones with most rapid D-xylose utilization and at the same time further improve their rate of D-xylose utilization, through adaptive evolution in D-xylose medium. The selection of clones with a shorter lag phase on D-xylose and a higher D-xylose utilization rate is most obvious. However, clones that utilize a larger part of the D-xylose will be able to undergo more proliferation cycles and therefore will tend to be present in higher amounts and thus also preferentially transferred to the next culture. Since the cultures were semi-anaerobic, the D-xylose is largely converted to ethanol and therefore these clones will likely also have a higher ethanol yield.
After only two transfers in D-xylose medium under semi-anaerobic conditions, the D-xylose fermentation rate already increased dramatically. Subsequent serial transfers resulted in further gradual improvement of D-xylose utilization. In previous studies, evolutionary adaptation under aerobic conditions followed by gradual transition to anaerobic conditions, was necessary to obtain strains with anaerobic D-xylose utilization capacity. In addition, several generations were required to obtain strains with efficient D-xylose fermentation capacity [28, 55, 59]. In our study, even though the best isolate GS1.11-26 was isolated after 11 serial transfers, clones with high D-xylose utilization capacity could already be isolated after only 2 transfers. The rapid improvement in the rate of D-xylose fermentation might be explained by the presence of a suitable combination of important genetic changes introduced by the mutagenesis and genome shuffling, and sustaining rapid improvement of D-xylose utilization by a repetitive subsequent genetic modification, such as amplification of the XylA gene or another crucial genetic element, and/or rapid enrichment of clones with a superior combination of mutant alleles. An important genetic change might have been generated also in the second culture of the evolutionary adaptation, which was characterized by a sharp rise in CO2 evolution at the end, and a dramatic increase in the rate of fermentation when this culture was transferred to the next batch. During evolutionary engineering, expansions and contractions of different subpopulations can occur [60, 61] and an individual cell with a beneficial mutation, providing a relative fitness advantage, can develop into a dominant subpopulation after several generations in serial batch transfer experiments . In our work, high variability in the rate of D-xylose fermentation was observed among individual clones isolated from intermediate cultures in the evolutionary adaptation process. However, isolates from the last culture showed a very similar fermentation performance, although not precisely the same, suggesting that the fitter clones finally conquered and dominated the culture.
The best strain, GS1.11-26, showed a reproducible and stable D-xylose fermentation rate and was further characterized both in laboratory medium and in three industrially relevant lignocellulosic feedstocks. In synthetic medium with D-xylose as a sole carbon source, the GS1.11-26 strain showed a maximum specific D-xylose consumption rate at least 15 times higher than the previous industrial strain BWY10Xyl expressing the same codon-optimized C. phytofermentans XI . The GS1.11-26 strain also accomplished complete attenuation of D-xylose with an ethanol yield of 0.46 g/g D-xylose, whereas the previous strain BWY10Xyl left a substantial amount of D-xylose unfermented. Moreover, the yield of ethanol obtained with GS1.11-26 was higher than the yield obtained with the best strain reported recently . As a consequence, GS1.11-26 exhibited the highest D-xylose to ethanol conversion yield than any other recombinant strain of S. cerevisiae reported so far. The high ethanol yield can also be explained by the very low xylitol yield, which is remarkable since the GRE3 gene had not been deleted nor was it inactivated in the strain development programme. The low xylitol yield, in the absence of GRE3 inactivation, might be due to the inherently higher metabolic flux in the industrial bioethanol production strain Ethanol Red, compared to the previously used strain backgrounds.
The GS1.11-26 strain also performed very well in lignocellulose hydrolysates both in SHF and SSF. The yield of ethanol per g consumed sugars, was slightly higher in all the lignocellulose hydrolysates compared to that in synthetic and YP medium. This is probably due to the lower amount of xylitol and glycerol formed, and is consistent with previous results [46, 63]. In SHF, it reached high maximum D-xylose consumption rates of 1.1 g/g inoculum DW/h and it showed partial co-fermentation of glucose and D-xylose during separate hydrolysis and fermentation. We have used a parameter for calculation of the specific sugar consumption rate based on the initial inoculum density, since the whole slurry was used for the fermentation experiment and since it is difficult to estimate the biomass during the fermentation process.
The ethanol yield from glucose and D-xylose in lignocellulose hydrolysates was also close to maximum and final ethanol titers between 3.9 and 5.8% (v/v) were reached, depending on the type of hydrolysate. The GS1.11-26 strain maintained a high level of tolerance like its parent Ethanol Red in inhibitor rich spruce hydrolysate and to individual inhibitors HMF and furfural. However, the strain did not retain the same high ethanol tolerance as the original Ethanol Red parent strain, though it was still able to accumulate more than 15% ethanol in very high-gravity fermentation (YP + 330 g/L glucose). The relatively high tolerance of GS1.11-26 to inhibitors, like HMF and furfural, found in spruce hydrolysate, but its lower tolerance to other stresses, like ethanol and acetic acid, can be explained by the fact that, after the genome shuffling step spruce hydrolysate was used as selective medium. The cells that were able to grow in spruce hydrolysate were further used for the evolutionary adaptation. This result demonstrates the importance of the selection conditions during evolutionary engineering, which is in agreement with the principle, “you get what you screened for” . GS1.11-26 showed reduced tolerance to acetic acid compared to the parent HDY.GUF5. This can be explained by different mechanisms underlying tolerance to various inhibitors . In S. cerevisiae, the tolerance mechanism to HMF and furfural, is similar, but distinct from that of acetic acid .
SSF is an interesting process for production of ethanol from lignocellulosic feedstocks, e.g. because it strongly reduces feedback-inhibition on enzymatic hydrolysis by the liberated monosaccharides and also reduces the danger of contamination. SSF performed at higher temperature (39°C) was also shown to increase the final yield of ethanol, because of a better compromise between the temperature optima of the enzymes and the yeast . In this respect, GS1.11-26 performed very well with almost complete attenuation of both glucose and D-xylose in about 96 h at 39°C. In a previously reported SSF of Arundo hydrolysate using Ethanol Red , there was also some D-xylose consumption, but only as a result of D-xylose reduction to xylitol. In our study, the final ethanol concentration for GS1.11-26 was 20.3 g/L, corresponding to an ethanol yield of 0.29, to be compared to the previously reported values for Ethanol Red of 15.3 g/L and 0.22 g ethanol/g total sugar. The ethanol yield thus increased by about 32% due to the efficient D-xylose conversion. The increase in the ethanol yield was in fact slightly higher than the increase expected from the D-xylose conversion alone, possibly due to removal of D-xylose inhibition on enzymatic hydrolysis.
Cell extracts of strain GS1.11-26 displayed 17-fold higher XI activity compared to cell extracts of the parent strain. However, there were no mutations in the XylA gene. Increased XI activity without any mutations in the XylA gene has also been reported recently [50, 51]. The high XI activity was explained by integration into the genome in multiple copies of the plasmid carrying the XylA gene . Although the recombinant strain in our work has been constructed through chromosomal integration of the XylA gene, it is possible that multiple chromosomal amplifications of the gene have occurred during the evolutionary adaptation process. The precise mechanism behind the establishment of the high XI activity in GS1.11-26 is currently being investigated and will be reported elsewhere.
Overexpression of XI in the parent strain did not increase the D-xylose consumption rate, as opposed to overexpression in the mutant strain M315 obtained after the mutagenesis step. This indicates that rapid D-xylose consumption requires a synergistic interaction between high XI activity and one or more mutations in the genome, which is in agreement with another report, in which high D-xylose assimilation capacity could only be attributed partially to the high activity of XI .
This clearly shows that the generation of other genetic changes, e.g. as obtained in our work by chemical mutagenesis, is essential for development of a pentose-utilizing strain with high performance. On the other hand, the random mutagenesis steps also resulted in unfavorable effects on other properties, such as reduced aerobic growth rate in glucose and a reduced glucose fermentation rate. This has also been reported previously during selection of a recombinant S. cerevisiae strain for anaerobic growth in D-xylose. In that report, strains exhibiting significant improvement in anaerobic D-xylose utilization also showed a reduced aerobic growth rate in glucose . We do not know whether the reduced glucose consumption rate or reduced aerobic growth rate in glucose, are trade-offs for the high D-xylose utilization capacity. Future research will have to show whether these negative side-effects are due to background mutations in the strain, which can be lost without affecting its high performance for D-xylose utilization, or whether they are causally linked to the high D-xylose fermentation rate. This will have important implications for further improvement of the strain for efficient co-fermentation of glucose, D-xylose and L-arabinose. In spite of this, the GS1.11-26 shows highly promising potential for further development of an all-round robust yeast strain for efficient fermentation of various lignocellulose hydrolysates. Moreover, it already contains the genes for additional utilization of L-arabinose and should be easily evolved also for efficient fermentation of this pentose sugar.