Mutagenic Organized Recombination Process by Homologous In Vivo Grouping (MORPHING) for Directed Enzyme Evolution

Approaches that depend on directed evolution require reliable methods to generate DNA diversity so that mutant libraries can focus on specific target regions. We took advantage of the high frequency of homologous DNA recombination in Saccharomyces cerevisiae to develop a strategy for domain mutagenesis aimed at introducing and in vivo recombining random mutations in defined segments of DNA. Mutagenic Organized Recombination Process by Homologous IN vivo Grouping (MORPHING) is a one-pot random mutagenic method for short protein regions that harnesses the in vivo recombination apparatus of yeast. Using this approach, libraries can be prepared with different mutational loads in DNA segments of less than 30 amino acids so that they can be assembled into the remaining unaltered DNA regions in vivo with high fidelity. As a proof of concept, we present two eukaryotic-ligninolytic enzyme case studies: i) the enhancement of the oxidative stability of a H2O2-sensitive versatile peroxidase by independent evolution of three distinct protein segments (Leu28-Gly57, Leu149-Ala174 and Ile199-Leu268); and ii) the heterologous functional expression of an unspecific peroxygenase by exclusive evolution of its native 43-residue signal sequence.


Introduction
In the past two decades directed evolution strategies have had a huge impact on protein engineering and synthetic biology [1][2][3][4][5]. Combining directed evolution with new computational and hybrid approaches has allowed researchers to design ''smart'' mutant libraries to address bottlenecks in enzyme functionality, helping to maintain the balance between activity and stability, or even creating novel catalytic activities [6,7]. The use of non-adaptive evolution involving neutral genetic drift has been added to this arsenal of techniques to create polymorphic populations with the aim of enhancing protein robustness and substrate promiscuity [8][9][10][11].
Although the majority of the vast protein sequence space is probably non-functional, with 20 n possible permutations (if we exclude the introduction of non-natural amino acids), it is still far from being fully explored [12]. Advances in the field involve the use of ultrahigh-throughput screening methods and the construction of focused mutant libraries to restrict the sequence space [13,14]. Most of the available methods used to create targeted libraries are based on computational studies that identify a limited number of positions for saturation mutagenesis, combinatorial and/or iterative [15][16][17]. A more recent approach is to introduce ancestral consensum mutations that have been identified by phylogenetic analysis and ancestral inference in extant enzymes [18]. Despite the wide array of focused evolution methods, there remains a need for consistent domain mutagenesis/recombination strategies targeting specific protein subsets for random mutagenesis and recombination, while conserving the remaining protein regions. Although this kind of focused-indiscriminate approach has received little attention in the literature [19,20], it can effectively unmask structural determinants of specific enzymatic attributes, which can then be optimized using the aforementioned methods.
Escherichia coli is by far the most common host in directed evolution experiments of prokaryotic proteins. However, broad differences with eukaryotic cells (missing chaperones, different codon usage, lack of posttranslational modifications) preclude the use of this bacteria to engineer eukaryotic enzymes which mostly end up in misfolding and inclusion bodies formation. Alternatively, Saccharomyces cerevisiae is the model organism of choice for in vitro evolution of eukaryotic genes, permitting the development of comprehensive synthetic biology and metabolic engineering studies, particularly when dealing with the production of fuels and chemicals [21][22][23][24]. In recent years, a range of methods have been described in this yeast to construct mutant libraries with different mutational bias, to integrate multiple DNA fragments for creating combinatorial libraries or to assemble expression cassettes that generate fully autonomous artificial pathways [25][26][27][28]. The high frequency of homologous DNA recombination in S. cerevisiae permits to simply shuffle foreign genes creating multiple crossover events, to repair linearized vectors for in vivo cloning or to promote the molecular evolution of multigenic phenotypes [29][30][31][32][33].
Here, we present a simple, rapid and reliable random domain mutagenesis/recombination method for short fragments that is based on the physiological properties of S. cerevisiae. Mutagenic Organized Recombination Process by Homologous IN vivo Grouping (MORPHING) randomly introduces mutations in specific protein segments using overlapping areas to favor in vivo splicing and recombination in yeast. The versatility of this method was evaluated in two case studies of ligninolytic oxidoreductases. First, we used MORPHING to enhance the oxidative stability of a versatile peroxidase (VP) from the basidiomycete Pleurotus eryngii. VP has three different catalytic sites for the oxidation of low-, medium-and high-redox potential compounds, which makes the enzyme extremely fragile in the presence of catalytic concentrations of H 2 O 2 [34]. The second enzyme studied was the unspecific peroxygenase (UPO) from the edible mushroom Agrocybe aegerita, a heme-thiolate peroxidase with special catalytic capacity that includes oxygen transfer reactions [35]. Despite its importance in organic synthesis, heterologous functional expression and directed evolution of this enzyme has not yet been reported. We used MORPHING to exclusively target the native UPO signal peptide for evolution towards functional expression in yeast.

Materials and Methods
VP from Pleurotus eryngii (the R4 mutant) was used as the parental type in the construction of the library. This R4 mutant was engineered for secretion by 4 rounds of directed evolution, resulting in expression levels of 22 mg/L [36]. The UPO1 variant (12C12) was generated by directed evolution in a previous study to functionally express the upo1 gene (clone C1A-2) from A. aegerita [37] in S. cerevisiae (unpublished material). ABTS (2,29-azino-bis(3ethylbenzothiazoline-6-sulfonic acid)), Taq polymerase, and the S. cerevisiae transformation kit were purchased from Sigma-Aldrich (Madrid, Spain). The iProof High Fidelity DNA polymerase was purchased from Bio-Rad (USA). The Zymoprep Yeast Plasmid Miniprep Kit and Zymoclean Gel DNA Recovery Kit were obtained from Zymo Research (Orange, CA, USA), while NBD (5-nitro-1,3-benzodioxole) was purchased from TCI America (USA). The Escherichia coli XL2-Blue competent cells and the GeneMorph II Kit (mutazyme II polymerase) were from Stratagene (La Jolla, CA, USA), and the uracil independent and ampicillin resistance pJRoC30 shuttle vector was obtained from the California Institute of Technology (CALTECH, USA). The protease-deficient S. cerevisiae strain BJ5465 (a ura3-52 trp1 leu2D1 his3D200 pep4::HIS3 prb1D1.6R can1 GAL) was obtained from LGCPromochem (Barcelona, Spain), the NucleoSpin Plasmid kit was purchased from Macherey-Nagel (Germany), and the restriction enzymes BamHI and XhoI from New England Biolabs (Hertfordshire, UK). All chemicals were of reagent-grade purity.

MORPHING Protocol
All the PCR products generated were cleaned, concentrated and loaded onto a low melting-point preparative agarose gel and purified using the Zymoclean Gel DNA Recovery Kit (Zymo Research). The PCR products were cloned by replacing the corresponding parental gene in pJRoC30. To remove the parental gene, the plasmid was linearized with BamHI and XhoI. The pJRoC30-R4 variant was used as a template to construct MORPHING libraries of VP, while the pJRoC30-12C12 was used as the template to construct MORPHING libraries of UPO signal peptide.
For VP MORPHING, the whole gene was fragmented into three different segments in individual PCR reactions. Each segment contained homologous overhangs of ,50 bp that overlapped one another to promote in vivo cloning in yeast. The targeted regions were subjected to random mutagenesis while the remaining segments were amplified by high-fidelity polymerases ( Figure S1). The distal His environment region was selected to tune mutational loads. Small mutant libraries of around 500 clones were screened to optimize the mutagenic conditions. A similar protocol was followed for UPO MORPHING but in this case the UPO gene was split into two segments, one containing the signal peptide and the other the mature protein.
iii) Reassembly of the whole gene: The whole gene was reassembled in vivo and recombined by transformation into S. cerevisiae cells using the Yeast Transformation Kit. The DNA transformation mixture contained the linearized plasmid (200 ng) mixed with the targeted region, as well as the segments upstream and downstream of those regions (400 ng per segment). Transformed cells were plated on SC drop-out plates and incubated for 3 days at 30uC. Subsequently, the mutant libraries were subjected to the HTP-protocol for oxidative stability, as described below.
iii) Reassembly of the whole gene: The whole gene was reassembled in vivo and recombined by transformation into S. cerevisiae cells using the Yeast Transformation Kit. The DNA transformation mixture contained the linearized plasmid (100 ng) mixed with the targeted region, as well as the mature protein (200 ng per segment). Transformed cells were plated on SC dropout plates and incubated for 3 days at 30uC. Subsequently, the mutant libraries were subjected to the HTP-protocol to assess activity as described below.

Directed evolution of whole UPO gene
The whole UPO gene including its signal peptide was subjected to one round of directed evolution by In vivo Assembly of Mutant libraries (IvAM) [27].

In vivo recombination of mutant libraries in S.
cerevisiae. Taq polymerase library and Mutazyme library were added in equimolar concentrations (200 ng each) to the linearized vector (100 ng). Transformed cells were plated on SC drop-out plates and incubated for 3 days at 30uC. Thereafter, the mutant libraries were subjected to the HTP-protocol to assay activity as described below.

Construction of fusion gene (evolved signal peptide plus the native upo) by In Vivo Overlap Extension
(IVOE). the native upo was amplified by high-fidelity PCR in a final volume of 50 mL containing: DNA template (0.2 ng/mL), 0.5 mM Forward primer SP* F, 0.5 mM Reverse primer RMLC R (Table S1), 1 mM dNTPs (0.25 mM each), 3% (v/v) dimethylsulfoxide (DMSO), and 0.02 U/mL iProof polymerase. High fidelity PCR was carried out on the gradient thermocycler under the following conditions: 98uC for 30 s (1 cycle); 98uC for 10 s, 52uC for 25 s, 72uC for 40 s (28 cycles); and 72uC for 15 min (1 cycle). PCR fragments corresponding to the SP* and the native upo (200 ng each) were recombined together with the linearized vector (100 ng) by IVOE [25].

Combinatorial saturation mutagenesis experiments in VP
Reaction mixtures were prepared in a final volume of 50 mL containing: DNA template (0.2 ng/mL), 0.2 mM Forward primer, 0.2 mM Reverse primer (RMLN F/MET SAT R primers and MET SAT F/RMLC R for first PCR and second PCR, respectively, (Table S1)), 0.8 mM dNTPs (0.2 mM each), 3% (v/v) dimethylsulfoxide (DMSO), and 0.02 U/mL iProof polymerase. High fidelity PCRs were carried out on the gradient thermocycler under the following conditions: 98uC for 30 s (1 cycle); 98uC for 10 s, 53uC for 30 s, 72uC for 1 min (28 cycles); and 72uC for 10 min (1 cycle). The whole gene was reassembled in vivo and recombined by transformation into S. cerevisiae cells using the Yeast Transformation Kit. The DNA transformation mixture was composed of the linearized plasmid (200 ng) mixed with the mutated fragments (400 ng per fragment). Transformed cells were plated on SC drop-out plates and incubated for 3 days at 30uC. Thereafter, the mutant libraries were subjected to the HTPprotocol described below.
High-throughput oxidative stability assay of VP Individual clones were selected and cultured in sterile 96-well plates (Greiner Bio-One GmbH, Germany) containing 50 mL per well of SC minimal medium. In each plate, column number 6 was inoculated with the parental R4 mutant as an internal standard, and well-H1 (containing minimal medium supplemented with uracil) was inoculated with untransformed S. cerevisiae as a negative control. Plates were wrapped in parafilm to prevent evaporation and incubated at 30uC, 225 rpm and 80% relative humidity in a humidity shaker (Minitron-INFORS, Biogen, Spain). After 48 h, 160 mL of expression medium was added to each well and the plates were incubated for a further 24 h. The plates (master plates) were centrifuged for 15 min at 3000 rpm and 4uC (Eppendorf 5810R centrifuge, Germany) and the master plates were duplicated with the help of a robot (Liquid Handler Quadra 96-320, Tomtec, Hamden, CT, USA) by transferring 20 mL of supernatant into two replica plates: the initial activity plate (IA plate) and the residual activity plate (RA plate). Next, 180 mL of stability buffer (20 mM sodium tartrate buffer, pH 5.0: Buffer A) was added to the IA plates and 180 mL of incubation solution (Buffer A containing 0.3 mM H 2 O 2 ) was added to the RA plates using a Multidrop robot (Multidrop Combi, ThermoFischer Scientific, Vantaa, Finland). Both plates were briefly stirred and incubated at room temperature for 60 min, such that the activity assessed in the RA plates was reduced by 2/3rds with respect to the initial activity of the parental type. The supernatants (20 mL) were transferred from both RA and IA plates to new plates to measure the residual and initial activity values by adding ABTS in specific buffers: 180 mL of 100 mM sodium tartrate buffer [pH 3.5] containing 2 mM ABTS and 0.1 mM H 2 O 2 to estimate of residual activity; and 180 mL of 100 mM sodium tartrate buffer [pH 3.5] containing 2 mM ABTS and 0.13 mM H 2 O 2 to estimate the initial activity. The plates were stirred briefly and the absorption at 418 nm (e ABTS N+ = 36,000 M 21 cm 21 ) was recorded (end-point mode, t 0 ) on a plate reader (SPECTRAMax Plus 384, Molecular Devices, Sunnyvale, CA). The plates were then incubated at room temperature until a green color developed and the absorption was measured again (t 1 ). The relative activities were calculated from the difference between the absorption value after incubation and that of the initial measurement normalized to the parental type in the corresponding plate (Dt 1 -t 0 ). Oxidative stability values were calculated as the ratio between residual activity and the initial activity values (RA/IA). To rule out false positives, two consecutive re-screenings were carried out. Moreover, a third re-screening was performed to determine the increase in the apparent half-life of each selected variant (t 1/2 H 2 O 2 , expressed in minutes) relative to the parental R4 in different molar ratios [ First re-screening: Aliquots of 5 mL of the best clones were removed from the master plates and used to inoculate 50 mL of minimal medium in new 96-well plates. Columns 1 and 12, and rows A and H, were not used to prevent the appearance of false positives. After incubating for 24 h at 30uC, 225 rpm, and 80% relative humidity, 5 mL was transferred to the adjacent wells and incubated for a further 24 h. Finally, 160 mL of expression medium was added and the plates were incubated for another 24 h. Accordingly, each mutant was grown in 4 wells. The parental types were subjected to the same procedure (row D, wells 7-11) and the plates were assessed using the same protocols as those used for the screening described above.
Second re-screening: An aliquot from the wells with the best clones in the first re-screening was inoculated in 3 mL of YPD and incubated at 30uC and 225 rpm for 24 h, recovering the plasmids from these cultures (Zymoprep Yeast Plasmid Miniprep Kit). As the product of the zymoprep was very impure and the concentration of DNA extracted very low, the zymoprep mixtures containing shuttle vectors were transformed into super-competent E. coli cells (XL2-Blue, Stratagene) and plated on LB/amp plates. Single colonies were picked and used to inoculate 5 mL LB/amp media, and they were grown overnight at 37uC and 225 rpm. The plasmids were then extracted (NucleoSpin Plasmid kit, Macherey-Nagel, Germany) and S. cerevisiae was transformed with plasmids from the best mutants as well as with the parental type. Five colonies for each mutant were selected and re-screened as described above.
Third re-screening (determination of t 1/2 H 2 O 2 ): A single colony from the S. cerevisiae clone containing the parental R4, the new mutants and untransformed yeast were picked from a SC drop-out plate (SC supplemented with uracil for untransformed cells), used to inoculate 5 mL of minimal medium, and incubated for 48 h at 30uC and 225 rpm (Minitron-INFORS, Biogen, Spain). An aliquot of cells was removed and used to inoculate a final volume of 5 mL of minimal medium in a 50 mL falcon tube (optical density, OD 600 = 0.3), and they were incubated until two growth phases had been completed (6-8 h, OD 600 = 1). Thereafter, 9 mL of expression medium (500 mg/L bovine hemoglobin) was inoculated with 1 mL of this preculture in a 100 mL flask (OD 600 = 0.1). After incubating for ,48 h at 30uC and 225 rpm (maximal VP activity; OD 600 = 25-30), the cells were separated by centrifugation for 15 min at 3000 rpm and 4uC (Eppendorf 5810R Centrifuge, Germany), and the supernatants were collected and stored at 4uC. The protein concentration was estimated from supernatants using the Bio-Rad protein assay kit, (Bio-Rad, USA). VP apparent concentration was calculated as the difference between the total protein content of yeast expressing VP and that in its absence (from non-transformed yeast cells -lacking VP gene-). High-throughput screening assay for UPO secretion Individual clones were selected and inoculated in sterile 96-well plates (Greiner Bio-One GmbH, Germany) containing 50 mL per well of SC minimal medium. In each plate, column number 6 was inoculated with the corresponding parental type, and one well (H1-control) was inoculated with untransformed S. cerevisiae cells in minimal medium containing uracil. The plates were wrapped in parafilm to prevent evaporation and incubated at 30uC, 225 rpm and 80% relative humidity in a humidity shaker. After 48 h, 160 mL of expression medium was added to each well and the plates were incubated for 48 h. The plates (master plates) were centrifuged (Eppendorf 5810R Centrifuge, Germany) for 15 min at 3000 rpm and 4uC, and the supernatants (20 mL) were transferred from the master plate to two replica plates by a robot (Liquid Handler Quadra 96-320, Tomtec, Hamden, CT, USA), adding the reaction mixture (180 mL) together with ABTS or NBD to each replica plate. Both colorimetric assays were used for properly detecting secretion levels improvements regardless of the substrate used (ABTS is typically used for assessing peroxidative activity whereas NBD for peroxygenase activity The plates were then incubated at room temperature until a green (ABTS) or yellow (NBD) color developed, and the absorption was measured again. The values were normalized against the parental type in the corresponding plate. To rule out false positives, two rescreenings were carried out as described above for VP.

Thermostability assay (T 50 )
Appropriate dilutions of the supernatants were prepared such that aliquots (20 mL) produced a linear response in kinetic mode. A gradient profile was constructed using a thermocycler (Mycycler, Bio-Rad, USA) for the selected mutants and the parental type, using 50 mL for each point in a gradient scale ranging from 30 to 80uC. After a 10 min incubation, samples were removed and chilled on ice for 10 min. Thereafter, 20 mL samples were removed and incubated for 5 min at room temperature. Finally, 180 mL of 100 mM sodium tartrate buffer [pH 3.5], 2 mM ABTS and 0.1 mM H 2 O 2 was added to the samples to measure activities.
The thermostability values were calculated as the ratio between the residual activity at different temperature points and the initial activity at room temperature. The T 50 value was determined as the transition midpoint of the inactivation curve of the protein as a function of temperature, which in our case was defined as the temperature at which the enzyme lost 50% of its initial activity after a 10 min incubation.

Protein and homology modeling
The crystal structure of VPL2 from P. eryngii at 2.8 Å resolution (1 Å = 0.1 nm, PDB ID: 3FJW) was used to generate a model to map the new mutations found with the help of the PyMOL Molecular Visualization System (Schrödinger). A homology model was generated by carrying out a structural alignment in PyMOL with the following crystal structures (PDB IDs are indicated): 3FJW, native VP from P. eryngii; 1IYN, recombinant chloroplastic ascorbate peroxidase (ApX) from N. tabacum expressed in E. coli; 3M5Q, native manganese peroxidase (MnP) isozyme 1 from P. chrysosporium; 1H3J, native peroxidase from C. cinerea (CiP); and 1W4W, recombinant horseradish peroxidase C1A from horseradish (HRP) expressed in E. coli. IDENTITY and SIMILARITY percentages were obtained using SEQUENCE SIMILARITY AND IDENTITY Software (http://imed.med.ucm.es/Tools/sias. html).

Results and Discussion
MORPHING is a method of generating DNA diversity based on the high frequency of homologous recombination of S. cerevisiae. In a single step, this approach allows us to assemble delimited randomly mutagenized regions with the remaining, unaltered fragments of a gene. Unlike most evolution methods focused in restricted areas, mutations are randomly generated and they do not depend on the engineering of a set of spiked/degenerate synthetic oligonucleotides. By MORPHING, small segments are targeted and subjected to error-prone PCR with defined mutational frequencies, while the remaining portions of the gene are amplified using high-fidelity polymerases ( Figure 1). Errorprone PCR methods have the drawback of codon bias although they can be modified by alternating between different polymerases in successive generations of evolution. Indeed, standard Taq polymerases (with a transition/transversion ratio [T s /T v ] ranging from 2.9 to 0.8, [38]) were employed in our mutagenic experiments, although mutational bias may be altered by combining this protocol with other well-known polymerases and mutational strategies [27,36,39]. The pool of mutated/conserved fragments and the linearized plasmid are subjected to one-pot repair and cloned in vivo, giving rise to a complete autonomously replicating plasmid upon transformation in yeast, without the need for additional PCR reactions or ligation/amplification steps. The number of crossover events (n+1, where n is the number of fragments) between segments is directly proportional to the number of segments assembled, allowing several regions to be studied alone or in a combinatorial manner. Depending on the distance between mutations, crossover events can occur between the different mutations in the target fragment(s), mediated by the S. cerevisiae recombination machinery, fostering enrichment. The success of this method is facilitated by the high fidelity DNA splicing of fragments through the small overhangs with overlapping sequences of ,50 bp that flank each segment. These overhangs ensure the in vivo reconstitution of the whole gene with random mutations only in the segments specifically targeted. Under these rules, up to six recombination events between fragments can be created without significantly affecting transformation efficiencies (,10 5 clones per transformation reaction can be obtained, which are good enough to screen mutant libraries).

Engineering oxidative stability
We first used this protocol to engineer oxidative stability in an evolved VP variant, the VP-R4 mutant generated in a previous directed evolution experiment to enhance its functional expression in yeast and its stability [36]. The sensitivity of VP (EC 1.11.1.16) to peroxides is the highest reported for any peroxidase to date being strongly inhibited in the presence of catalytic concentrations of H 2 O 2 due to a mechanism-based phenomenon known as suicide inactivation that is common to all peroxidases [40]. The inherent fragility of VP is explained by its complex structure. With a redox potential of over +1.2 V, three different catalytic sites (the heme domain, a catalytic tryptophan located at the protein's surface and the Mn 2+ binding site), and two access channels, the dense production and traffic of free radicals jeopardizes the enzyme's stability and function, (Figure 2) [34,36].
To identify what are potentially the most H 2 O 2 -sensitive regions of the VP, we performed a multiple structural alignment using high-and medium-redox potential peroxidases with improved oxidative stability [39,[41][42][43], (Figure 3). After careful examination of the model, three different regions (of 26, 30 and 69 amino acids each, excluding the recombination areas) in the vicinity of the heme group were targeted for random mutagenesis and recombination, (Figures S1, S2). The first region subjected to MORPHING was the distal His environment (L28-G57) that contains the H 2 O 2 binding site within a helix that is highly conserved in all high-redox potential peroxidases. The second target region was the proximal His environment (L149-A174) located on the opposite side to the distal His, in the surroundings of the heme domain. The third region was the Met environment (I199-L268), containing three of the five putative oxidizable Met in the VP-R4 variant. Several mutational loads were assayed for each region to construct independent mutant libraries and then explore them for oxidative stability. Mutational loads were adjusted by modifying the PCR conditions in order to introduce 1 to 5 mutations per segment (including the crossover areas). The frequencies of mutation were estimated from different landscapes of mutant libraries (500 clones each), calculating the number of clones with ,10% of the parental enzyme activity, and they were further verified by DNA sequencing of a random sample of mutants including active and non-active variants ( Figure 4). Overall, an average value of T s of 65% (A«G, 52%; T«C, 13%) and T v of 35% (T«A, 17.4%; A«C, 8.8%; G«C, 8.8%) was observed in the libraries under study (i.e. T s /T v ratio,1.8). Mutational frequencies of 0.5 nucleotide changes/100 bp were obtained with T s GRA when [MnCl 2 ]#0.1 mM, regardless of the DNA template concentration. In segments as short as 30 residues long (e.g., the distal His environment: L28-G57), we obtained high mutational loads at [MnCl 2 ].0.1 mM (frequencies of ,3 mutations/100 bp). Among the high mutational variants, the 1E11 (2T s , 3T v ) and 2G4 (3T s , 2T v ) mutants incorporated 5 mutations each that inactivated the protein due to the highly conserved nature of this region ( Figure 5).
After the initial screening and three consecutive re-screenings, several mutants with an improved increment in apparent half-life (Dt 1/2 vs H 2 O 2 ) were identified. We found two beneficial mutations at the L28-G57 segment. The 3G10 variant showed a 2.1-fold improvement in stability vs H 2 O 2 , with a significant Dt 1/2 vs H 2 O 2 of 18 min with respect to the parental type and a 5.5uC increase in the T 50 (the temperature at which the enzyme retains 50% of its activity after a 10 min incubation, Figure 6). Only one mutation, GAG E40K AAG , was found in 3G10 and the same mutation was introduced in 2F5 and 5D1 variants. This mutational redundancy highlights the role of this specific alteration in generating oxidative stability in VP and significantly, the highly stable CiP contains a Lys residue at the same position [39], Figure 5. In VP, E40 is one of three acidic residues that form the Mn 2+ binding site ( Figure 2) and it is plausible that closing some of the protein inlets involved in the generation of free radicals may be beneficial for VP stability, albeit at the cost of compromising this catalytic site. The ACT T45A GCT mutation was discovered in both the 1C12 and 2G8 mutants, the latter of which also contained the silent GCT A61A GCC mutation ( Figure 5). While the T45A mutation did not alter thermostability (T 50 = 59uC), it conferred a 1.3-fold improvement in stability and it was associated with a Dt 1/2 vs H 2 O 2 of ,10 min with respect to the parental type ( Figure 6). The same amino acid (V53A according to CiP numbering) was introduced into the evolved CiP although without improving thermostability, Figure 5 [39]. It is likely that the only effect of introducing an Ala at this position relates with the accessibility of H 2 O 2 to the internal protein structure. The CCT P141A GCT mutation in the 5A9 variant arose in the L149-A174 segment (proximal His environment). Situated at the heme entrance, Pro141 is a highly conserved residue in all fungal peroxidases. However, the P141A mutation resulted in a 1.4-fold improvement in oxidative stability, producing a Dt 1/2 vs H 2 O 2 of ,5 min with respect to the parental type and no change in thermostability ( Figure 6). P141A is located in a strictly conserved region and according to our model, the substitution of Ala by Pro may widen the heme channel making easier the traffic of oxidized species and therefore limiting their harmful residence time in the inner protein structure. To further study the possible synergies between beneficial mutations E40K, T45A and P141A, a triple variant was constructed by site-directed mutagenesis doubling the Dt 1/2 vs H 2 O 2 to 45 min (data not shown). This mutant was the departure point to engineer an oxidative stable VP variant by successive rounds of in vitro evolution (unpublished material).
No mutants were discovered in the I199-L268 segment (the Met environment), suggesting that the three oxidizable methionines in this area are not involved in the oxidative stabilization of VP. This result was corroborated by subjecting Met262 and Met265 to combinatorial saturation mutagenesis and screening. No improved variants were identified in this way and ,95% of the clones were inactive in this mutagenic landscape, indicating that these residues are very sensitive and do not tolerate changes ( Figure S3). Similarly, Met152 in the proximal His environment was not mutated.
In this first enzyme case study, MORPHING effectively identified new structural determinants, such as the P141A mutation, that are important for oxidative stability. Significantly, the identification of the E40K and T45A mutations was in good agreement with previous findings [39], thereby validating this approach. Although each mutated segment was analyzed independently by constructing small high-quality libraries with different mutational loads, it is also feasible that different combinations of mutated segments can be prepared using the yeast's recombination machinery.

Enhancing functional UPO expression
MORPHING was also used to assess whether the secretion levels of an unspecific peroxygenase (UPO) could be enhanced in S. cerevisiae. UPO (EC 1.11.2.1) is a new, potentially ligninolytic, peroxidase that has attracted much research attention due to its versatility and applicability in a variety of synthetic processes [35,44]. Its peroxygenative (oxygen-transfer) activity is of particular importance as UPO can behave as a self-sufficient monooxygenase to mediate regio-and enantio-selective oxyfunctionalizations that are essential for organic synthesis. Among the array of oxygen transfer reactions catalyzed by UPO are brominations, sulphoxidations, N-oxidations, aromatic peroxygenations, alkyl hydroxylations, double bond epoxidations and ether cleavages. Like many other ligninolytic oxidoreductases, UPO is not readily expressed in heterologous hosts so that it can be tailored by directed evolution and therefore, we recently addressed this problem by subjecting the whole UPO gene to several rounds of random mutagenesis, recombination and screening in S. cerevisiae (unpublished data). To further enhance the secretion of this protein, we used MORPHING to independently evolve the 43 amino acid UPO signal peptide in order to enrich the signal leader in beneficial mutations without altering the biochemical properties of the enzyme. To reliably compare the two directed evolution strategies, we exposed the signal peptide alone and the entire UPO gene (including its signal leader) to one round of directed evolution in two parallel experiments. Mutational rates for the leader library and the full-gene library were 0.5 and 0.1 nucleotide changes/ 100 bp, respectively. Both libraries were screened with the help of an ad-hoc dual-colorimetric assay to estimate the enzymes peroxidase (with ABTS) and peroxygenase (with NBD) activities. The mutagenic landscape generated by MORPHING revealed an increased tolerance to mutations in the signal peptide than in the whole UPO gene. Indeed, we found that when random mutagenesis of the leader and the whole UPO gene was compared, 30% and 49% of clones had ,10% of parental enzyme activity on ABTS as a substrate, respectively (Figure 7). These results are consistent with the fact that mutations in the leader sequence only affect secretion, whereas those in the whole UPO gene may also compromise the catalytic properties. While no variants carrying mutations in the leader were identified in the whole UPO gene after one round of evolution, several independent but almost consecutive mutations were observed in three independent beneficial variants after MORPHING of the leader sequence ( TTC F12Y TAC , GCG A14V GTG and AGG R15G GGG ). Each of these mutations individually enhanced secretion by ,20% with respect to the parental type. Although MORPHING was useful to unmask these beneficial mutations for secretion, the eukaryotic machinery of S. cerevisiae was unable to join these positions by homologous recombination due to their proximity. Therefore, we constructed a signal peptide containing the full set of these mutations by conventional site-directed mutagenesis looking for a synergic effect in secretion levels. The evolved signal peptide also included the beneficial mutation GCC A21D GAC (discovered in the earlier stages of evolution). Finally, the signal peptide of native UPO was replaced by the evolved signal peptide F12Y-A14V-R15G-A21D (see Materials and Methods for details), resulting in a 27-fold increase in total secretion compared to the native UPO fused to its original leader (,260.11 ABTS U/L and 5465.7 ABTS U/L, respectively). F12Y-A14V-R15G-A21D mutations are located at the hydrophobic core of the leader and they may exert a beneficial effect on secretion by promoting a more suited interaction with the signal recognition particle during translocation to the endoplasmic reticulum.
This second case study demonstrates that the use of our focused mutagenesis method to direct the evolution of signal leaders is a suitable approach to promote heterologous functional expression of complex eukaryotic genes in yeast. Targeting mutational loads to leader sequences is a simple means of detecting mutations that are beneficial for secretion and that can be subsequently combined in a single signal peptide to generate potential synergies along the S. cerevisiae secretory pathway.

Conclusions
The random domain mutagenesis/recombination method presented here is a reliable one-pot approach for the construction of focused mutant libraries of eukaryotic genes in S. cerevisiae. In general terms, MORPHING allows the researcher to focus exclusively on the random introduction of mutations and their recombination in restricted region/s, while protecting critical domains from mutagenesis. The selection of the target regions to be evolved is as important as the mutational loads chosen for each mutant library, which can be easily varied to enrich the target segments in beneficial mutations.
The two case studies presented here validate the versatility of our method by tackling two distinct problems. While MORPH-ING proved useful to explore limited targeted regions, allowing us to identify several structural determinants of H 2 O 2 inhibition in VP that could be applied to other high redox-potential fungal peroxidases, it also effectively decoupled secretion and catalytic activity for functional UPO expression. This approach can also be used to explore other complex problems, such as to alter substrate specificity or enantio-selectivity by subjecting several segments of the same gene to random mutagenesis, promoting their in vivo assembly in one transformation step. Indeed, this strategy is currently being used by our group to evolve a fungal aryl alcohol oxidase for the selective oxidation of different alcohols. Apart from structure-function relationship studies, MORPHING can be also useful when structural information is absent, e.g., for the evolution of leader peptides for secretion, for the modification of promoters, or in the evolution of unknown regions of biochemical relevance that have been revealed by conventional directed evolution.
Additional advantages of this method include the reduction of the sequence space to be explored (good results can be achieved with small libraries of 400-500 clones), the conservation of certain catalytic properties while improving other traits, and the discovery of new structural/catalytic determinants that can be further optimized using focused saturation mutagenesis. The combination of MORPHING with classical directed evolution and semirational approaches, or with neutral genetic drift, may lead to the development of new adaptive pathways to engineer more robust eukaryotic enzymes in yeast [8,13,14]. Figure S1 VP MORPHING. Three different regions of VP were targeted for random mutagenesis and recombination (L28-G57, L149-A174, and I199-L268). The VP gene is shown in blue, the a-factor prepro-leader to promote secretion in yeast in red and the shuttle vector in green. The areas of crossover between the fragments are represented by crosses. The overlapping areas between segments were created by superimposing PCR reactions in defined regions (see also Figure S2 and Table S1).