Screening of randomly mutagenized xylanase libraries
The different steps of the engineering strategy are summarized in Figure . The initial phase of this work involved the use of error-prone PCR (epPCR) to generate random biodiversity. In preliminary work, we observed that more than 10 base mutations/kb produced >70% inactive clones. Therefore, a progressive strategy employing three successive rounds of epPCR was preferred, with moderate mutational charge (5 to 7 base mutations/kb) at each stage. The results of activity screening (where activity can generally be considered to be the product of both expression levels and specific activity) at each round are summarized in Table . Regarding the first round of screening, this work has already been reported by Song et al
]. Although the best mutant from this first round, designated Tx-Xyn-AF7, displays a wild-type amino acid sequence, its DNA sequence contains two mutations (at nucleotide positions 27 and 516) that cause approximately twofold higher expression of the recombinant enzyme. Therefore, the sequence encoding Tx-Xyn-AF7 was used as the template for the second round of epPCR.
Flowchart of the in vitro evolutional process. The best-performing mutants, used as parental input for a subsequent round of evolution, are boxed and mutants are designated according to the point mutations that characterize them.
Summary of directed evolution for improvement of Thermobacillus xylanilyticus xylanase (Tx-Xyn) xylanase activity
DNA sequence analysis of ten library clones, taken from the second-generation library, revealed an average mutation rate of 5.4 base substitutions/kb and a transition/transversion ratio of approximately 1.4, indicating that the mutations were relatively unbiased in this respect. A total of 4,333 clones were screened on intact wheat straw (In-WS), and the 4 most active clones (>4CV) were selected, using the activity of Tx-Xyn-AF7-bearing clones as the base case for comparison. DNA sequencing revealed that all four clones were characterized by single amino acid changes. Two clones were mutated at position 3 (Y3L and Y3H), while two others were mutated at independent, but neighboring locations (W109R and Y111H).
Examination of the three-dimensional structure of Tx-Xyn revealed that Y3 lies in the distal glycon part of the active site cleft, while W109 and Y111 are situated nearby and in the thumb region, respectively; thus all three residues are potentially important for enzyme function. For this reason, at this stage in the experiment it was decided to focus on these mutations for the creation of further mutant libraries. However, to ensure that all of the possible permutations would be present in the third generation, recombination was achieved using site-directed mutagenesis. Consequently, five double mutants (Y3L-W109R, Y3L-Y111H, Y3H-W109R, Y3H-Y111H and W109R-Y111H) and two triple mutants (Y3L-W109R-Y111H and Y3H-W109R-Y111H) were created. Together with the other four original single mutants, these were used as parental templates for the next round of epPCR, which led to the creation of a fourth generation (Figure ).
To efficiently challenge clones present in the fourth library, the microtiter plate assay was modified by replacing In-WS with xylanase-depleted wheat straw (Dpl-WS). The principle behind this was to select clones that produce enzymes that can actually hydrolyze arabinoxylans that are inaccessible or resistant to wild-type xylanase. The key features and performance descriptors of this modified assay are summarized in Table . Overall, the CV value for individual wells of Tx-Xyn-AF7 control varied between 8% to 11%, indicating that this screen was sufficiently reliable for library screening.
Characteristics of intact wheat straw (In-WS) and xylanase-depleted wheat straw (Dpl-WS) and summary of the two screening assays
DNA sequence analysis of a randomly picked sample of fourth-generation library clones revealed an average mutation rate of 7.2 nucleotide substitutions/kb. Likewise, functional screening using the modified Dpl-WS assay indicated that 0.6% of screened clones presented activities that were significantly higher (>5CV) than the mean value of the activity of Tx-Xyn-AF7 clones. Therefore, the top 30 clones were isolated and used for subsequent rounds of DNA recombination.
Optimization of mutant xylanases using DNA recombination
To further increment the functional fitness of the enzymes expressed by the candidate clones obtained from random mutagenesis, the staggered extension process (StEP) DNA shuffling approach was adopted, because it offers a much simpler procedure than classical DNA shuffling [29
]. This method was used to successively create fifth, sixth and seventh-generation libraries. To appreciate the impact of the iterative use of StEP on overall library fitness, Figure shows the relative performance of fourth-generation to sixth-generation libraries. At each generational increment, library fitness increased in accordance with expectations [30
]. The results of statistical analyses performed on the three successive libraries (fifth, sixth and seventh generations) that were created using this method are summarized in Table .
Figure 3 Iterative improvement of enzyme fitness after screening on xylanase-depleted wheat straw (Dpl-WS). The x-axis represents clones in a microtiter plate, randomly selected from a wild-type control series (open triangles, using Thermobacillus xylanilyticus (more ...)
Mutational frequency in the fifth to seventh generations
For the initial round of DNA shuffling, 30 clones were used as parental input. After DNA shuffling, the library was submitted to screening using the modified Dpl-WS assay. This step allowed the selection of seven hits whose activities were significantly higher (>7CV) than the mean value of the activity of Tx-Xyn-AF7 clones. DNA sequencing revealed that these 7 clones contained 11 point mutations, including Y111H and some new amino acid substitutions (Figure ). As before, the seven mutants were used as parental input for two further rounds (sixth and seventh) of DNA shuffling.
After the creation of the seventh-generation library, the experiment was stopped, because DNA sequencing of the highest performing seventh-generation clones showed that five mutational combinations out of a total of seven had already been identified in the sixth generation (Figure ). This observation suggested that the evolutionary itinerary had almost reached an end, with very little new biodiversity being introduced.
Among the seven best performing seventh-generation clones, Y6H-Y111H and Y6H-S27T-Y111H displayed the highest activity increase (>8 CV) in the screening, compared to that of wild-type control (Tx-Xyn-AF7). In addition, among the six amino acid substitutions that were detected in clones obtained from DNA shuffling, Y111H was present in every template and the frequency of Y6H and S27T increased from the fifth generation to the seventh generation (Table ). Consequently, we decided to focus on clones containing these three amino acid changes for enzyme production and characterization. Overall mutants that were retained for characterization included Y6H-Y111H, S27T-Y111H and Y6H-S27T-Y111H from the seventh-generation screening and the single mutants Y111H, Y6H and S27T.
Site-saturation mutagenesis (SSM) at positions 3 and 111
Among the second-generation clones, selected for higher activity on In-WS, two amino acid positions, 3 and 111, were pinpointed as potentially interesting locations. Therefore, in addition to the use of Y3H and Y111H as parental templates for further random mutagenesis and DNA shuffling, SSM was performed to investigate the importance of these two residues with respect to enzyme activity on recalcitrant arabinoxylan (AX) in wheat straw (that is, Dpl-WS). In each case a library was created and 288 clones were screened using the modified Dpl-WS assay. This number of clones was sufficient to ensure a 99.87% probability that all possible amino acid variants were present [33
]. Additionally, a random sample of each library was submitted to DNA sequence analysis in order to control the success of the experiment.
Figure shows the results of the screening of the two site-saturation libraries. Overall, the Y111N (N represents any amino acid) library provides a larger population of improved clones, though both libraries contain a small minority of clones that display activities that are above the value of μ + 4σ of wild-type control (where σ is standard deviation and μ is mean value). Three highest performing clones were selected from each library and analyzed by DNA sequencing. All three clones from the Y3N library displayed the same Y3W mutation, whereas two clones from the Y111N library were phenotypically and genotypically identical (encoding the mutation Y111S) and one displayed an Y111T mutation. In view of these results, three individual clones encoding Y3W, Y111S and Y111T were retained for further characterization.
Figure 4 Xylanase-depleted wheat straw (Dpl-WS) screening of site-saturation libraries. Filled squares and open circles correspond to site-saturation mutagenesis (SSM) performed at positions 3 and 111, respectively. The 288 clones of each library are positioned (more ...)
Characterization of key properties of the Tx-Xyn mutants
Since the screening of mutant enzyme libraries obeys the maxim 'you get what you screen for', the mutants selected in this work were only improved with respect to the hydrolysis of wheat straw. Hence, other important properties such as thermostability could have been negatively affected. Consequently, the thermostability of each mutant was assessed (Table ). Although the thermostability of some mutants at 60°C was clearly affected (for example, that of Y6H and Y6H-Y111H), all of the enzymes were sufficiently stable to enable the measurement of kinetic properties without any major modifications to the protocols that were routinely used to characterize wild-type Tx-Xyn. It is also noteworthy that all of the mutants were highly stable at 50°C, since measured activity remained stable over a 6 h incubation period.
Thermostability of Thermobacillus xylanilyticus xylanase (Tx-Xyn) and mutants thereof
Each of the mutants was characterized with regard to its ability to hydrolyze birchwood xylan (BWX) and low-viscosity wheat arabinoxylan (LVWAX). According to our findings (data not shown), BWX is devoid of α-L-arabinosyl substitutions, and LVWAX displays an A/X ratio of 0.54. Concerning wild-type Tx-Xyn, its turnover number and performance constant were higher for LVWAX, though the apparent KM value was lower on BWX. This tendency was also displayed by the majority of the mutants (Table ). Regarding the apparent values of KM, all of the mutants displayed improved affinity for BWX, but this was not the case for LVWAX. Notably, Y111H was the mutant that displayed the best affinity for BWX, while its affinity for LVWAX was unaltered. However, the rate constant for Y111H-mediated hydrolysis of BWX was lowered when compared to that of the wild-type enzyme, but was improved on LVWAX. Intriguingly, the opposite was true for Y111T, for which the value of kcat was 48% greater than that of Tx-Xyn on BWX, but identical to that of Tx-Xyn on LVWAX. When Y111H was combined with other mutations (for example, S27T-Y111H or Y6H-Y111H), its influence on the performance constant appeared to be dominant, annulling the improved activity on BWX, displayed by the single mutants S27T and Y6H.
Kinetic parameters of Thermobacillus xylanilyticus xylanase (Tx-Xyn) and mutants for hydrolyses involving either birchwood xylan (BWX) or LVWAX
Assessment of the impact of Tx-Xyn mutants on wheat straw
To further evaluate the altered properties of the different mutants, their activities on the original wheat straw samples (In-WS and Dpl-WS) were examined. Reactions were performed using pure preparations of wild-type and mutant xylanases either alone or in the combination with Accellerase 1500 (a cellulase cocktail). The results of HPAEC-PAD analyses performed on the reaction supernatants are shown in Figure , which show the conversion of total xylose and glucose (that is, %tot. xyl
and %tot. glu
, w/w) in the straw residues. The soluble sugar yields are summarized in Additional files 1
Figure 5 Percentage conversion of total sugars in xylanase-depleted wheat straw (Dpl-WS) and intact wheat straw (In-WS) by Thermobacillus xylanilyticus xylanase (Tx-Xyn) alone or in combination with Accellerase 1500 . (A) Conversion of total xylose in reactions (more ...)
The hydrolysis of Dpl-WS revealed that all of the mutants could release further amounts of soluble xylose equivalents and that their performance was superior to that of wild-type Tx-Xyn. The mutants S27T and Y111T produced the most outstanding results, because these could release 2.3-fold and 2.1-fold more xylose equivalents from Dpl-WS than Tx-Xyn. The lowest performers were Y111H and Y3W, which yielded 35% and 46% more xylose equivalents, respectively (Figure ). However, it should be noted that even the best variant S27T could only release 2.5% tot. xyl of Dpl-WS (5.5 g xylose per kg wheat straw), which is evidence of the recalcitrance of this substrate.
For the hydrolysis of In-WS (pH 5.8), wild-type Tx-Xyn released 43.7 g equivalent xylose per kg wheat straw. This represents 4.4% of the dry weight and 16.7% of total xylan (16.7%tot. xyl) content. Similar results were obtained for the mutants Y6H, Y6H-Y111H, Y6H-S27T-Y111H and Y3W, but five other mutants yielded higher amounts (18.6% to 20.4%tot. xyl) of soluble xylose equivalents, with the best mutant being Y111T (Figure ).
The five mutants displaying improved activity on In-WS, were further selected to investigate synergy with cellulases on In-WS, operating at the optimum pH for Accellerase (pH 5.0). Likewise, suitable control reactions at pH 5.0 were performed using only mutant xylanases, or wild-type Tx-Xyn. All controls revealed that the different xylanases displayed reduced hydrolytic capacity, compared to their activity at pH 5.8 (Figure ). According to its manufacturer, Accellerase 1500 principally contains endoglucanase and β-glucosidase activities. In our trials, Accellerase alone was able to solubilize 7.3%tot. xyl and 18.9%tot. glu In-WS (Figure ). However, in combination with xylanases, higher yields of xylose and glucose were measured, which were greater than the sum of the yields of Accellerase and xylanase alone, clearly revealing synergistic interactions between the enzyme participants. The mixture of wild-type Tx-Xyn and Accellerase solubilized 24.5%tot. xyl and 23.6%tot. glu of In-WS (Figure ). However, significantly the different mutants were able to improve on this performance, solubilizing 27.4 to 29.0%tot. xyl and 24.9 to 26.4%tot. glu from In-WS.