To discover mutants of PirXI that enhance xylose metabolism, we designed and constructed small focused mutant libraries followed by in vivo screening for better variants. Recent structural and biochemical information was used to select target positions for mutagenesis, focusing on residues that surround the metal-binding sites. Replacing the second shell residues might have an effect on metal binding or reactivity and thereby influence the activity. The fully conserved metal-coordinating residues in PirXI are Glu233, Glu269, Asp297 and Asp340 for site M1 and Glu269, His272, Asp308 and Asp310 for the catalytic metal M2. For the first library (LibM1), we targeted residues that are in close proximity of the M2 site. Five residues were selected: three (Val270, Ala273 and Thr274) that lie on the same helix as the metal binding residues (His272 and Glu269), and two (Trp307 and Thr309) that are on a nearby loop (Fig. 1). Amino acid diversity to be introduced at each position was selected based on phylogenetic diversity, in silico-predicted stabilities of the mutants, and visual inspection of the predicted mutant structures. For phylogenetic input, a multiple sequence alignment was performed on 22 different class I and 100 class II XI sequences. Considering conservation scores and similarities between amino acid properties, the library diversity was decided. For example, residue Ala270 is fully conserved throughout all class II enzymes and therefore the diversity at this position was restricted to alanine and glycine to avoid extreme modifications. Changes in free energy of folding (ΔΔGfold) of mutants relative to the wild-type enzyme were predicted using FoldX calculations . Large decreases in predicted stability were used to dismiss mutations from the library design. The resulting LibM1 library included 1008 different variants (Table 1).
To construct the LibM1 mutant library, at each target position a minimum number of partially undefined codons covering the selected set of amino acid substitutions was chosen using a spreadsheet implementing the CodonFinder routine . The codons were selected in such a way that all the desired amino acids (including the wild-type residues) are incorporated at balanced coverage without introduction of undesired codons like stop codons. The 1008 LibM1 mutant library was covered by 8 partially undefined codons (Table 1).
Library DNA was obtained by generating gene fragments using PCR and subsequent cloning into E. coli–yeast shuttle expression vector (pRS426-URA) as described in “Methods”. Transformation of the library DNA to E. coli resulted in over 8000 clones, of which pooled plasmid DNA was transformed to S. cerevisiae strain DS75543, producing 6000 colonies. Considering the library size of 1008, these numbers are sufficient for near full library coverage . Prior to yeast transformation, library diversity was confirmed by sequencing a mixture of plasmids isolated from the mixed collection of E. coli transformants. The sequencing results showed that all expected bases were incorporated at the correct positions, indicating sufficient library quality to proceed to screening.
Screening for improved xylose utilization
Library LibM1 was screened by growth competition of S. cerevisiae strain DS75543 transformed with library plasmids DNA. The entire collection of yeast transformants was inoculated into xylose medium and cells were cultivated with serial transfers to fresh xylose medium. Faster growing cells, which over time became dominant, were assumed to harbor an improved PirXI. Screening was performed both under aerobic and anaerobic (oxygen-limited) conditions, each in duplicate, as different variants can be expected depending on the metabolic status of the cells. For anaerobic growth, the cultures were kept oxygen-limited as described in “Methods”. The effect of limited oxygen availability was reflected in the final cell densities (OD600) of the cultures, which were ~ 3 and > 20 for anaerobic and aerobic conditions, respectively. Anaerobic cultures initially required 8–9 days before growth occurred. A reduced lag time and/or increased growth rate was observed after multiple transfers with all four selection cultures, also in comparison to a control culture harboring only wild-type PirXI.
Both aerobic and anaerobic cultures were harvested after the 10th transfer and plasmids were isolated to evaluate the selected PirXI genes. Sequencing showed that all four cultures, i.e., both the aerobic duplicates and anaerobic duplicates, contained only one PirXI variant, which carried the mutations V270A and A273G.
Effect of V270A–A273G PirXI on xylose utilization
The consistent selection of the V270A–A273G variant from library LibM1 suggested that the screening method was reliable and sensitive. Nevertheless, it is possible that other events such as genomic mutations or variations in expression level were responsible for the improved growth of yeast carrying the V270A–A273G mutant PirXI. The replicon of the pRS416 vector used in this work is derived from yeast plasmid 2µ, and plasmid copy numbers can vary from culture to culture . Furthermore, in laboratory evolution of S. cerevisiae for growth on xylose, cells may acquire diverse chromosomal mutations that cause improved growth [35, 36]. To prove that in our case the selected mutations in the PirXI structural gene caused improved growth on xylose, the mutations were reconstructed by site-directed mutagenesis in the original PirXI gene and cloned in a vector that was not subjected to previous selection. S. cerevisiae DS75543 cells were then transformed with the freshly prepared constructs and their growth performance in a 96-well plate was monitored. We have repeated this process several times and consistently observed that the cells containing the mutant PirXI grow on xylose better compared to those containing wild-type PirXI (Fig. 2a). We have also observed that general growth performance of the cells slightly differs between experiments as well as between clones picked from a single transformation experiment. Figure 2b shows a high variability in growth between 30 independent transformants despite their identical genotype. Nevertheless, the results clearly indicate that the V270A–A273G mutant PirXI improves growth on xylose as compared to wild type, especially in the earlier phases of growth (Fig. 2b). On average, the mutant cultures started to grow earlier and more quickly reached their final density.
In 96-well plates, oxygen availability may not be well controlled and results could be influenced by evaporation. Therefore, a comparison between the wild-type and V270A–A273G XI variants was also performed using shake flask cultures with replicates inoculated with pre-cultures from independent transformants. The resulting growth curves (Fig. 2c) confirm that the mutated PirXI is beneficial for growth on xylose. The specific growth rates (µ) were calculated from the exponential part of the curves, using the following equation for fitting: lnX = lnX0 + µ(t − t0), where X is the measured OD600 and µ is the rate. The average growth rates of the wild type and the mutant are 0.13 ± 0.01 and 0.18 ± 0.01 h−1, respectively. These results show that the observed improved growth is due to the V270A–A273G mutations in PirXI, not by unidentified mutations elsewhere on the plasmid or in the chromosome of the selected transformants.
Yeast cells selected for good growth on xylose show high overexpression of PirXI [12, 15], which may be a metabolic burden for the cells and trigger selection of mutants with a higher activity:expression ratio in competition experiments. To examine if the PirXI mutations affected enzyme expression, we studied XI levels in cells grown on xylose. The specific activity with 100 mM xylose measured with the cell-free extracts were 0.94 U/mg and 0.54 U/mg for the wild type and the mutant, respectively. SDS-PAGE gels revealed that the expression levels for the wild-type and the mutant enzyme were similar (Fig. 3).
The uncertainty of in vivo metal binding properties of PirXI and metal content of the yeast cytoplasm makes it difficult to define a metal composition for assays that gives results reporting on in vivo performance. When we measured the PirXI activity of extracts of S. cerevisiae DS75543 cells without metal addition, the results indicated that the activity of the wild-type enzyme was almost twofold higher compared to the mutant. This unexpected observation could be caused by changes in PirXI metal composition during enzyme preparation and dilution, e.g., due to binding of metals released from organelles such as vacuoles or from changes in metal–protein interactions.
To examine if the individual mutations in the PirXI variant are both necessary for improved growth, we constructed the single mutants V270A and A273G and examined the effect on growth on xylose, particularly on the early growth phase. The growth curves indicate that the V270A mutation has a larger effect, showing much earlier initiation of exponential growth (Fig. 4), but also cells containing the A273G mutant PirXI showed a slight growth improvement compared to the cells expressing the wild-type enzyme.
Kinetic properties of PirXI V270A–A273G
We expected that the positive effect of the V270A–A273G mutations on xylose-supported growth would be due to improved catalytic parameters, i.e., increased catalytic rate (kcat) or a better substrate affinity (reduced KM). Since PirXI can be activated by different divalent metals and activities depend on the type of metal that is bound , we measured XI activities with metals that the enzyme potentially binds in vivo as previously found by metal analysis (Mg2+, Mn2+ or Ca2+) . With none of these metals, the in vitro activities revealed an increased kcat or decreased KM for the mutant enzyme in comparison to wild type. In contrast, the wild type performed better in the presence of all metals tested, showing slightly higher catalytic rates and substrate affinities (Table 2). Especially, with Mn2+, the activity of the mutant decreased 50% compared to the wild type. This result indicates that an increase in specific activity with these metals, at least individually, is not responsible for the improved growth on xylose of yeast expressing V270A–A273G PirXI.
Besides the metal-dependence of the isomerase, metal affinities were considered as a possible cause of improved in vivo enzyme performance. We estimated metal affinities of the wild-type and the mutant enzyme by measuring the activation constant (Kact) for each metal. This constant represents the metal concentration giving half-maximal enzyme activity. Since xylose isomerase requires two metals for activity, the value depends on the binding sites with the lowest affinity if both sites must be occupied. For measuring Kact with Mg2+ and Mn2+, 100 mM xylose was used as substrate. In case of Ca2+, 400 mM xylose was used since the KM,xylose of the PirXI-Ca2+ is very high (Table 2). The data showed that the V270A–A273G mutant showed slightly higher affinity for Ca2+ and Mn2+, whereas the wild-type enzyme has slightly higher affinity for Mg2+ (Table 2). However, the differences are small and do not indicate a shift in metal affinity as the cause of improved growth.
Metal affinity was also examined by measuring the effect of metal addition on PirXI thermostability since metal binding can stabilize metalloenzymes . Effects on apparent melting temperatures were measured in the presence of different concentrations of metals using thermal shift assays (Fig. 5). The results showed that the apo forms of wild-type and V270A–A273G PirXI have a similar thermostability. Interestingly, whereas Tm,app of the wild-type PirXI increased with metal concentration according to a hyperbolic saturation-like curve, the Tm,app of the mutant enzyme was constant up to ca. 200 µM of Mn2+ or Ca2+ added, with an increase at higher metal concentrations (Fig. 5a, c). In contrast, when Mg2+ was added, the thermostability of the mutant enzyme was not increased even at concentrations that were saturating for enzyme activity (Fig. 5b). The difference between the metal-concentration dependence of mid points of thermal shift assays and Kact values measured in the presence of substrate suggests that substrate influences metal binding, as also observed when examining X-ray structures of the enzyme with different combinations of ligands . Only in the presence of the substrate xylose both metal-binding sites in the crystal structures were occupied, whereas in xylitol- or glycerol-bound enzyme only the M1 metal site was occupied with a metal ion.
The metal content of yeast cells is complex and consists of both free metal ions and metal ions bound to macromolecules . In general, in vivo metal binding by metalloproteins is controlled by mechanisms such as intracellular metal homeostasis, localization of protein folding, and activities of metal transporters and metallochaperones . In a previous study, we showed that changes in intracellular metal composition affect metal composition of PirXI, which in turn influences catalytic performance . PirXI isolated from yeast grown on xylose is mostly bound with Ca2+, which barely activates the enzyme. Therefore, a large portion of PirXI does not contribute to in vivo conversion of xylose. In contrast, the smaller fraction of PirXI that is bound with the strongly activating Mn2+ contributes most to the in vivo enzyme activity .
In view of the complex metal composition of S. cerevisiae, we measured the activities of the wild type and the V270A–A273G mutant in the presence of varying concentrations of Mn2+ and a fixed high concentration (1 mM) of Ca2+ (Fig. 6). As expected, both variants showed higher activity with increasing concentration of Mn2+. Interestingly, the degree to which Mn2+ influenced the activity was different between the wild-type and the mutant enzyme. At low concentrations of Mn2+ (10–100 µM) and in the presence of 1 mM Ca2+ the mutant enzyme showed slightly higher activity. This indicates that the activation of the mutant enzyme by Mn2+ in the presence of a high concentration of Ca2+ is improved. Even though the activity of the mutant is lower in the presence of Mn2+ or Ca2+ alone, at certain low Mn2+/Ca2+ concentration ratios, the V270A–A273G mutant enzyme is better activated than the wild type. These results suggest that differences in in vivo metal activation may be responsible for the improved growth of yeast cells expressing the V270A–A273G mutant PirXI.
Crystal structures of PirXI wild type and V270A–A273G
To examine possible structural changes in the V270A–A273G PirXI, we solved and compared crystal structures of this variant and the wild-type enzyme purified from yeast (Fig. 7). The overall structures of the wild type and mutant enzyme are very similar and confirmed the mutations. In the mutant structure, the side chain of Phe280 moves ~ 0.5 Å in the direction of Ala270. Due to the decrease in hydrophobicity and size the surrounding waters also shift towards Ala270. Mutation A273G shows no effect on the structure. There is no significant difference between the two structures to explain the improved in vivo performance of the mutant.
The enzyme crystals were prepared without removal or addition of metal ions so that only intrinsic metal ions are present. When the metals ions were refined as Mg2+, the Fo–Fc map showed unaccounted electron density at the metal positions, suggesting the presence of heavier ions. In an anomalous electron density map a clear signal was observed at the M1 and M2a positions with σ levels of 4.8 and 3.5, respectively. Mg2+ ions do not have an anomalous signal at the in-house used wavelength of 1.54 Å. However, a comparison with anomalous maps of previously determined structures  shows similar peak heights in the Ca–xylose structure of PirXI (PDB code 5NH8). Therefore, metal ions at the M1 and M2a positions were refined as Ca2+ with 100% occupancy resulting in a flat Fo–Fc map in both the wild-type and the double-mutant structures. The temperature factors (B-factors) of the two Ca2+ ions are 11.9 and 16.4 for the wild type and 11.4 and 19.3 for the mutant, which are lower than those of the surrounding residues. The distances of the coordinating side chains to the M1 ion in the wild type PirXI and the mutant enzyme isolated from yeast are similar to those in the wild-type Ca–xylose structure reported earlier (5NH8). These results indicate that most of the metal-binding sites of PirXI isolated from yeast are occupied by poorly activating Ca2+ ions, both in the wild-type and in the PirXI V270A–A273G mutant enzyme. Other metal ions, such as Mg 2+, Fe2+, Mn2+ or Co2+, may be bound with low occupancy.
Construction and screening of library LibM2
A second library design for discovery of better xylose isomerase mutants focused on mutations in a stretch of six residues flanking the substrate binding site. In this case, to avoid the risk of improved growth by chromosomal mutations, we compared growth properties of yeast clones transformed with the library DNA (Table 3). The growth of library colonies on solid medium containing xylose as sole carbon source was monitored by visual inspection. Plasmid DNA was isolated from suspected positive (larger) clones, retransformed to yeast and rescreened.
Library design again included selection of target positions and diversity to be introduced at each position (Fig. 8). The residues at the six target positions (Ser141, Thr142, Ala143, Asn144, Val145 and Gly147) at the C5 side of the substrate interact with the substrate either directly or indirectly. Therefore, it was expected that modifying these residues can improve substrate binding and the catalytic rate. Residue Thr142 is fully conserved throughout all known xylose isomerase sequences. In the structure it is connected to O5 of the substrate via a water molecule. To keep this interaction, we limited the diversity at this position to Thr and Ser. In a previous study, the PirXI T142S mutation was discovered to improve the growth of yeast on xylose . We preserved Phe146 as it is fully conserved and it plays an important role in keeping the active site hydrophobic. Together with Trp189 and other hydrophobic aromatic residues (Trp50 and Phe61) this promotes the hydride shift by shielding the hydride from solvent [40,41,42]. The resulting library consists of 3584 variants (Table 3).
The library was constructed using the same strategy as for library LibM1. The initial E. coli transformation yielded ca. 8000 colonies and the diversity was confirmed by sequencing a plasmid mixture obtained from pooled transformants. The subsequent transformation to S. cerevisiae DS75543 also resulted in over 8000 colonies. For identification of clones showing improved xylose utilization, transformed cells were washed from glucose plates and spread on xylose plates (see “Methods” for details). Many cells did not grow at all or started to grow very slow, causing visible differences between individual colonies, also for wild type. The latter indicated that other factors than PirXI activity influenced colony growth, for example the physiological status of transformed cells at the moment of plating. After three rounds of screening and retransformation, 46 colonies which showed superior growth were selected. To identify the best variant, growth in xylose containing liquid medium was measured using 96-well plates and compared to wild type. Most of the 46 variants reproducibly showed improved growth. The xylose isomerase genes from the 24 best growing variants were sequenced, revealing 10 different variants, one of which was wild type. The sequences that appeared most frequent (4–8 times) were reconstructed in a clean background and the effects on growth on xylose were evaluated after transformation to fresh DS75543 cells. Among these mutants, variant S1 (S141N–T142S–A143S–G147A) consistently showed the biggest improvement of growth on xylose when several independent cultures were tested. As shown with variant V270A–A273G, the most significant effect of mutant S1 also appears to be on the earlier start of the growth while showing a slightly increased exponential growth rate (Fig. 9).
Kinetic properties of PirXI S1
The reconstructed PirXI mutant S1 was purified from E. coli and its activity was measured after reconstitution with different metals. As with variant V270A–A273G, the Michaelis–Menten kinetic parameters measured in the presence of Mg2+, Mn2+ or Ca2+ revealed reduced kcat values as compared to wild type (Table 2). The KM for xylose was also several fold higher in the presence of Mg2+ or Mn2+ compared to the wild type. Furthermore, the Kact values indicate that a shift in metal affinity does not promote better in vivo performance of the mutant enzyme as the affinity towards the most activating metal Mn2+ decreased, while the affinity towards Ca2+, which poorly activates the enzyme, increased.
Performance of other xylose isomerase mutants
The results described above indicate that both libraries yielded PirXI mutants that caused accelerated growth on xylose. However, their properties exhibited disconnection between in vivo performance and in vitro catalytic properties. We further explored the ambiguous relation between in vivo and in vitro enzyme properties by studying PirXI mutants discovered independently in previous studies, using different S. cerevisiae host strains [9, 14]. Using directed evolution with random mutant libraries, Lee et al. discovered PirXI variant E15D–T142S which increased the growth rate on xylose from 0.01 h−1 up to 0.06 h−1 . Activities of these enzymes have only been measured with cell lysates, making a comparison difficult, but KM values appear high. Later, Katahira et al. discovered that mutations at position N338, especially substitution N338C, improved growth of yeast on xylose. This mutation was effective not only in PirXI but also in related XIs . It was reported that yeast cells carrying the N338C variant of PirXI consumed xylose 3–4 times faster than cells carrying wild-type PirXI. The catalytic properties of the mutant enzymes from these studies have not been described. Very recently, when our study was nearing completion, Seike et al. described mutations in XI from Lachnoclostridium phytofermentans (LpXI) that enhanced d-xylose metabolism . The most effective mutations were T63I and V162A. The corresponding positions in PirXI are distant from the active site and the mutations were not examined here.
We constructed the two earlier PirXI mutants [9, 14], E15D–T142S and N338C, and confirmed that the variants are beneficial for PirXI-mediated growth of strain DS75543 on d-xylose as well (Fig. 10). Subsequently, we expressed the mutants in E. coli, purified the enzymes and measured kinetic parameters. For this, activities were determined with Mg2+, Mn2+ or Ca2+ added to the apoenzyme and the activation constants were also determined (Table 2). As with the new mutants described in the current paper, Michaelis–Menten parameters and metal affinities did not reflect the positive effect of the mutations on growth, with the exception of an increased kcat of the N338C mutant in the presence of Mg2+ and Mn2+. However, this enzyme also has a higher KM. This result shows that the disconnection between in vivo performance and in vitro properties of PirXI is not dependent on the screening strain or selection conditions.