Protein engineering to improve coenzyme recycling in the metabolic steps catalyzed by XR and XDH
Biochemical constraints dictate that anaerobic conversion of xylose into ethanol is possible only when XR and XDH have matching coenzyme specificities [19, 21, 43]. The xylose pathway from Pichia stipitis which has served as point of departure for the construction of numerous xylose-fermenting strains of S. cerevisiae [4, 6] does not fulfill this requirement well. Its XDH is strictly specific for NAD+  while the XR strongly prefers NADPH over NADH [45, 46]. Using known natural enzymes, the assembly of a chimeric pathway in which XR and XDH show exactly comparable utilization of NADP(H) and NAD(H) appears to be currently out of reach.
Protein engineering has therefore been pursued to make the coenzyme specificity of XR or XDH better compatible with that of the corresponding partner enzyme of the xylose pathway. Following the early studies by Metzger and Hollenberg , Makino and coworkers succeeded in creating a NADP(H)-dependent version of P. stipitis XDH through rational design . A notable feature of the best improved multiple mutant of XDH was a catalytic efficiency for the NADP+-dependent reaction that exceeded about 3.8-fold the corresponding efficiency of the wild-type using NAD+. While XDH from yeast and fungal sources is typically a Zn2+-dependent enzyme evolutionary related to medium-chain dehydrogenases/reductases [48, 49], bacterial polyol dehydrogenases possessing activity with xylitol do not use an active-site metal in catalysis and are found with the short-chain dehydrogenase/reductase superfamily of proteins . Ehrensberger and Wilson  determined a 1.9 Å crystal structure of XDH from Gluconobacter oxidans based on which they were able to convert the NAD+-dependent wild-type enzyme into a strictly NADP+-specific variant that had retained about 14% of the original catalytic efficiency for oxidation of xylitol.
Successful creation of a highly active XR mutant featuring a substantially lower preference for NADPH than the wild-type enzyme has strongly benefited from crystal structures of the enzyme from C. tenuis bound with NADP(H)  and NAD(H) . Some of the mutations found to be useful in CtXR  were later also introduced at homologous positions of the amino acid sequence of XR from P. stipitis [28, 29, 54] (see later). Selection of the K274R-N276D doubly mutated CtXR for the in vivo experiments reported herein was based on a detailed steady-state kinetic characterization of a series of single and multiple-site variants of CtXR  and included analysis of mixed coenzyme utilization in the presence of physiological concentrations of NADPH and NADH . The double mutant eliminates the 33-fold preference of the wild-type for reaction with NADPH; however, it is clearly not a perfectly NADH-dependent enzyme. With the desired application for xylose fermentation in mind, it was important to also consider the possible effect of the mutated XR on the fluxional efficiency of the xylose pathway. The K274R-N276D double mutant was expected from its kinetic parameters to substitute the wild-type enzyme in the NAD(H)-dependent conversion of xylose without introducing an extra kinetic bottleneck.
Metabolic consequences of altering the coenzyme preference of XR in a xylose-fermenting strain of S. cerevisiae
The discussion will focus on physiological effects observed in stable xylose-fermenting strains of S. cerevisiae where the relevant genes were integrated into the yeast genome. Note, however, that preliminary reports have been published in which xylose fermentation by yeast strains expressing mutated P. stipitis XR or XDH from multi-copy plasmid vectors was investigated. They support the general idea that enhanced recycling of NADH [28, 29] or NADPH [30, 31] in the XR-XDH pathway helps decreasing xylitol formation and can eventually increase the ethanol yield. However, two independent studies, in which exactly the same mutants of P. stipitis XDH were examined, reached opposite conclusions regarding the effect on ethanol yield resulting from the usage of NADP+ instead of NAD+ in the XDH step [30, 31]. These results emphasize the possible ambiguity in tracing back changes in strain physiology to the modification of the cosubstrate specificity of XR or XDH.
Therefore, the relevant phenotypes of the two isogenic yeast strains constructed in this work were carefully analyzed. Gene expression under control of the TDH3 promoter yielded levels of specific activity for XR (utilizing NADH), XDH, and XK that were about half those obtained by other groups who used the phosphoglycerate kinase 1 promoter for expressing the genes of the Pichia stipitis xylose pathway along with the endogenous XK gene [17, 55, 56]. (The comparison is relevant because purified CtXR  and GmXDH  display similar specific activities as the corresponding P. stipitis enzymes [44–46].) The observed ratio of the specific activities of XR-NADH, XDH, and XK was 1: ≈5–7: ≈10 and lies within the window of operation recommended by Hahn-Hägerdal and co-workers [55, 58].
We were concerned about the difference in specific XR and XK activities found in strains BP000 and BP10001 that was substantially larger than expected from the estimated experimental error of 15 – 20% for the entire procedure of cell disruption and activity measurement. A gene copy number effect can be ruled out considering that (1) chromosomal integration of the three overexpressed genes occurred in a single step; and (2) unlike XR and XK, the specific activity of XDH was identical in both strains. However, for the purpose of strain comparison for xylose fermentation it may be noted that the specific uptake rates for the xylose substrate were very similar in BP000 and BP10001. We therefore regarded the two yeast strains as a suitable system for examining metabolic consequences resulting from the change in XR coenzyme specificity. The unknown source of variation in the specific enzyme activities was not further pursued.
The 52% decrease in xylitol yield resulting from the genetic replacement of wild-type CtXR by the K274R-N276D double mutant is quite significant in comparison to the success other metabolic engineering strategies have had in suppressing xylitol formation [for comprehensive reviews, see [1–5]], not only in terms of the magnitude of the effect but also because it was accompanied by similar changes in ethanol yield (42% increase) and glycerol yield (57% decrease). The acetate yield in bioreactor cultivations of the two strains was not affected within limits of the experimental error. Therefore, alteration of XR coenzyme specificity appears to have caused a global metabolic response, which contributes to a comprehensive improvement of the distribution of fermentation products.
It is interesting to bring into comparison these data with results of a detailed study by Jeppson et al.  who examined the effect of substituting wild-type XR from P. stipitis by a Lys270→Met mutant thereof, which according to studies of Lee and co-workers  exhibits a 17-fold higher K
m for NADPH than the native enzyme. The analogous site-directed replacement in CtXR, Lys274→Met, caused improvement of the coenzyme selectivity, NADPH compared to NADH, from a value of 33 in the wild-type enzyme to 5.5 in the mutant . However, it was also accompanied by a more substantial, 20-fold decrease in catalytic efficiency for the NADH-dependent reduction of xylose .
Despite the expected strong impairment of XR physiological function resulting from the mutation Lys270→Met , the yeast strain harboring a single gene copy for P. stipitis mutant XR consumed xylose in batch fermentations as fast as the isogenic control strain that contained native XR, and it produced less xylitol (0.17 vs. 0.29; 42% decrease) and more ethanol (0.36 vs. 0.31; 16% increase) . Formation of acetate and glycerol was, however, enhanced by about 40% in the mutant XR strain under these conditions. Interestingly, the effect of altered cosubstrate specificity of P. stipitis XR on ethanol yield was not clearly visible in strains that harbored two copies of the respective XR gene and hence consumed xylose about 1.5-fold faster than the corresponding single-copy strains.
In a continuous culture that used a mixed sugar substrate (10 g/L glucose, 10 g/L xylose), the strain harboring a single gene copy for the K270M mutant produced 8% more ethanol (0.40 g/g) and 41% less xylitol than the corresponding control strain. It was suggested from results of metabolic flux analysis that xylose conversion by the K270M mutant took place exclusively via NADH-dependent reaction while the wild-type form of P. stipitis XR showed balanced utilization of NADH and NADPH under these conditions (see later). Unfortunately, significant differences in physiological parameters for the native XR strains BP000 (Y
EtOH/xylose = 0.24; q
xylose = 0.06 h-1, where Y is a yield coefficient and q is the specific uptake rate) and TMB3001 (Y
EtOH/xylose = 0.31; q
xylose = 0.145 h-1 ) set a limit to the quantitative evaluation of the possible benefit, particularly on Y
EtOH/xylose, originating from the use of the K274R-N276D double mutant of CtXR (this work) compared to the K270M mutant of the P. stipitis enzyme .
Notwithstanding, if we assume that quantitative information about XR performance under in vivo conditions can be gleaned from the results of relevant in vitro assays [40, 41], the CtXR double mutant is expected to be a much superior catalyst with regard to both coenzyme selectivity and efficiency. Unfortunately, the large preference for NADPH seen with isolated preparations of native P. stipitis XR [45, 46] is very difficult to reconcile with the suggestion from metabolic flux analysis that a very substantial fraction of xylose (≈ 50%) is consumed by the enzyme in vivo via the NADH-dependent pathway [56, 60]. Therefore, while further systematic integration of XR protein engineering into the development of novel xylose-fermenting strains of S. cerevisiae would seem to be a promising approach, it also requires that the apparent conflict in findings for in vitro and in vivo experiments be sorted out in future studies.