Redesigning Aldolase Stereoselectivity by Homologous Grafting

The 2-deoxy-d-ribose-5-phosphate aldolase (DERA) offers access to highly desirable building blocks for organic synthesis by catalyzing a stereoselective C-C bond formation between acetaldehyde and certain electrophilic aldehydes. DERA´s potential is particularly highlighted by the ability to catalyze sequential, highly enantioselective aldol reactions. However, its synthetic use is limited by the absence of an enantiocomplementary enzyme. Here, we introduce the concept of homologous grafting to identify stereoselectivity-determining amino acid positions in DERA. We identified such positions by structural analysis of the homologous aldolases 2-keto-3-deoxy-6-phosphogluconate aldolase (KDPG) and the enantiocomplementary enzyme 2-keto-3-deoxy-6-phosphogalactonate aldolase (KDPGal). Mutation of these positions led to a slightly inversed enantiopreference of both aldolases to the same extent. By transferring these sequence motifs onto DERA we achieved the intended change in enantioselectivity.


Introduction
Protein engineering offers the opportunity to alter protein properties such as stability, substrate specificity, and conversion rates, respectively.However, addressing stereoselectivity still remains challenging for such endeavors, and this holds especially true in cases with no enantiocomplementary enzyme being available.In a summary of successful changes of enzyme enantioselectivity, Faber, Kazlauskas and coworkers described that these were achieved in most cases by altering the substrate binding mode [1].Notably, even a single amino acid exchange can be sufficient to invert stereoselectivity as demonstrated, e.g., for thiamine diphosphatedependent enzymes [2,3].The key step for rational protein design is the identification of these positions for mutagenesis.According to a study of Kazlauskas et al., amino acids close to the active site are often the more effective points of attack for changing enantioselectivity [4].A fruitful approach to identify these stereoselectivity-determining positions is the analysis of enzyme families to expose conserved amino acids as summarized by Pleiss [5].However, understanding the molecular mechanism of an enzyme´s enantiopreference is still challenging, even when a crystal structure is known.Grafting of specific patterns from related enzymes can be an alternative to generate a protein with the desired function [6,7].Referring to enzyme selectivity, Pazmiño et al. could identify amino acid positions, where variations led to an improved enantioselectivity by comparison of two sequence-related cyclopentanone monooxygenases, with one protein structure being modeled [8].
Enantiocomplementarity is essential for organic synthesis endeavors: enantiomerically pure building blocks are required in both enantiomeric forms.In particular, the aldol motif is a key to many synthetic endeavors.Besides conventional aldol reactions, the use of aldolases as biocatalysts became popular during the last two decades.Especially the 2-deoxy-D-ribose-5-phosphate aldolase (DERA) is of interest because it catalyzes aldol reactions between the nucleophilic acetaldehyde and certain electrophilic aldehydes.Interestingly, DERA catalyzes sequential aldol reactions facilitating access to lactols with up to two newly formed stereogenic centers in a single sequence [9][10][11][12][13].However, the synthetic use of DERA is limited by the lack of a known natural enantiocomplementary DERA.The chemoselectivity of DERA was already enhanced by Jennewein et al., allowing DERA to accept chloroacetaldehyde as a substrate [14], and Wong and coworkers improved DERA´s use of non-phosphorylated electrophiles [15].However, so far no alteration of DERA´s enantio-or diastereoselectivity has been attempted.Thus, the stereoinversion of DERA is optimal for applying and testing the grafting approach from homologous aldolases.
We identified amino acid positions responsible for enantioselectivity and activity by comparing homologous aldolases, the enantiocomplementary pyruvate-dependent 2-keto-3-deoxy-6-phosphogluconate aldolase (KDPG) and 2-keto-3-deoxy-6-phosphogalactonate aldolase (KDPGal).All three aldolases are (αβ) 8 -barrel (TIM-barrel) proteins and, as class I aldolases, form an intermediate Schiff base between a particular lysine in their active site and their nucleophilic substrate during catalysis.While using a common electrophile, glyceraldehyde-3-phosphate, these aldolases merely differ by their specificity for a certain nucleophile, acetaldehyde for DERA and pyruvate for KDPG and KDPGal (Fig 1).In our homologous grafting approach we used geometric analyses and molecular dynamics (MD) simulations of KDPG and KDPGal to identify their determinants of stereoselectivity, which were then transferred onto DERA, leading to the intended change in its enantioselectivity.

Molecular Dynamics Simulations
The crystallographic protein structures of KDPG (PDB ID: 1eun [16]) and KDPGal (PDB ID: 2v81 [17]) were obtained from the Protein Data Bank PDB [18].For KDPG, which forms a trimer in the crystallographic structure, only chain A was used.Crystallographic waters assigned to the prepared protein chains were retained for the simulation.Sulfate ions were removed.Selenomethionines were replaced by methionines, retaining the crystallographic conformation.Protein hydrogen atoms and side chain conformations of asparagines, glutamines, and histidines were assigned by REDUCE [19].Hydrogens of crystal waters were added by the xLEaP program from the AMBER 9 package of molecular simulations programs [20].
MD simulations (MD) were performed with the AMBER 9 package of molecular simulation programs [20] using the Cornell et al. force field [21] with modifications introduced by Hornak et al. (ff99SB) [22].Protein structures including crystallographic waters were solvated in a truncated octahedron of TIP3P water [23] such that the distance between the edges of the box and the closest solute atom was at least 11 Å.Periodic boundary conditions were applied using the particle mesh Ewald (PME) method [24] to treat long-range electrostatic interactions.Bond lengths of bonds to hydrogen atoms were constrained by SHAKE [25,26].The time step for all MD simulations was 2 fs, and a direct-space non-bonded cutoff of 8 Å was applied.After minimization the system was heated over 50 ps from 100 K to 300 K using canonical ensemble (NVT) MD.Then, the solvent density was adjusted over 50 ps using isothermal-isobaric ensemble (NPT) MD.Positional restraints applied during thermalization (with a force constant of 5 kcal mol -1 Å -2 ) were reduced in a stepwise manner over 50 ps followed by 50 ps of unrestrained canonical ensemble (NVT) MD at 300 K with a time constant of 2 ps for heat bath coupling to a Berendsen thermostat.Production runs were carried out for 10 ns; snapshots of the structure were saved every 1 ps.Interatomic distances were calculated for all saved snapshots by the PTRAJ program from the AMBER 9 package of molecular simulations programs [20].

Biological Procedures
DNA manipulation: Enzymes, PCR reagents, and oligonucleotides were purchased from standard suppliers.PCR reactions were performed in 100 μl thin-wall tubes using a VWR Doppio gradient-thermocycler.DNA amounts were determined using a Thermo Fischer Scientific NanoDrop 2000c spectrophotometer.Commercially available kits were applied for DNA isolation and purification.DNA sequencing was conducted at SEQLAB or GATC Biotech.All primers are listed in S1 Table.
Cloning of KDPG EC aldolase.The gene eda EC encoding for KDPG from E. coli was amplified from the isolated genomic DNA of the strain DH5α using the primer pair edaEC_N-deI_fw and edaEC_XhoI_rv together with Pfu polymerase (10 min 95°C; 35x[45 s 95°C; 45 s 62°C; 45 s 72°C]; 10 min 72°C).The amplified DNA products were restricted by restriction endonucleases NdeI and XhoI and ligated into the likewise restricted and dephosphorylated vector pET22b (Novagen).The aldolase gene was expressed cytosolically with a C-terminally fused hexa-histidine tag.The obtained vector is referred to as pET22b::eda.
Generation of KDPG EC -variants.According to the method of Tillett and coworkers a sequence and ligation independent cloning was carried out to obtain the mutants [27].Two pairwise PCRs were carried out using Phusion polymerase (ThermoScientific Fischer) and the primer pairs edaEC_chimI_fw and edaEC_chimI_rv_t as well as edaEC_chimI_fw_t and edaEC_chimI_rv, respectively (30 s 98°C; 25x[10 s 98°C; 30 s 55-65°C; 2 min 72°C]; 10 min 72°C; template: pET22b::eda).After the PCR the two mixes were combined and used for transformation of chemically competent E. coli DH5α.To introduce mutation T161V the same procedure was used with edaEC_T161V_fw and edaEC_T161V_rv_t as well as edaEC_T161V_fw_t and edaEC_T161V_rv, respectively, on the templates pET22b::eda or pET22b::eda_chimI (see above) to obtain both the single and the double mutant.
Cloning of KDPGal EC aldolase.The procedure for the gene encoding for KDPGal EC was exactly the same as for the gene encoding for KDPG EC but with the primer pair dgoaEC_N-deI_fw and dgoaEC_XhoI_rv.The obtained vector is referred to as pET22b::dgoa.
Generation of a DERA EC library.The gene deoC EC encoding E. coli DERA EC was amplified (10 min 95°C; 35x[45 s 95°C; 45 s 61°C; 45 s 72°C]; 10 min 72°C) from E. coli K12 chromosomal DNA using deocEC_NdeI_fwd and deocEC_EcoRI_rev as primers (S1 Table #59 and #60).The ligation with pET-21a(+) was done using the sticky ends generated by NdeI and EcoRI.The resulting construct is pET-21a::deoc EC wt [31].A DERA enzyme library consisting of 49 variants was constructed by exchanging the amino acid codons for T18, L20, and A203 within the wildtype (wt) gene to codons for 12 amino acids (R, N, D, C, G, H, I, L, F, S, Y, V, except for L20N and L20C) and deleting the amino acid codons for G204 and G205 as well as pairwise combinations of these single mutations (S1 Fig).
General Protocol for QuikChange PCR: PCR reactions were performed in a total volume of 50 μl containing 10 μl 5x Phusion HF buffer, 2 U Phusion TM polymerase, ca.30 ng template DNA, 200 μM dNTPs, 0.2 μM of each primer, and water.The applied temperature protocol was 30 sec at 95°C, 16 cycles of 30 sec at 95°C, 1 min at 50-65°C, and 6 min at 68°C, with a final step of 10 min at 68°C.Amplification was confirmed by agarose gel electrophoresis (1% agarose gels).PCR products were incubated with 10 U DpnI at 37°C overnight to digest template DNA.
General Protocol for Round-The-Horn PCR (RTH PCR): PCR reactions were performed in a total volume of 20 μl containing 4 μl 5x Phusion HF buffer, 0.4 U Phusion TM polymerase, ca.30 ng template DNA, 200 μM dNTPs, 6.25 nM of each primer, and water.The applied temperature protocol was 30 sec at 98°C, 25 cycles of 10.2 sec at 98°C, 30 sec at 55-70°C, and 2 min at 72°C, with a final step of 10 min at 70°C.Amplification was confirmed by agarose gel electrophoresis (1% agarose gels).Ligation with 2 μl T4 ligation buffer and 1 μl T4 ligase was performed at room temperature for 1 h.The ligated PCR products were incubated with 10 U DpnI at 37°C overnight to digest template DNA.
General Procedure for Plasmid Preparation: Chemically competent E. coli DH5α cells were transformed with the PCR products, followed by selection on LB agar plates supplemented with Ampicillin.Inoculation of a 5 ml overnight culture with LB medium supplemented with Ampicillin by a single colony enabled plasmid preparation.
Frozen cells (up to 12.5 g) were thawed by suspending in purification buffer [20% cells (w/v)] and disrupted using a FRENCH press Cell Disruptor Cell (Thermo Electron, Oberhausen GmbH).The crude extract was centrifuged for 20 min at 20,000 rcf and 4°C to obtain the cell free crude extract, which was applied to a 5 ml Ni-NTA column equilibrated with purification buffer (Qiagen, Hilden).The chromatographic elution was carried out using an ÄKTA Purifier System (GE Healthcare, Munich): 6 CV purification buffer with 0 mM imidazole; 6 CV purification buffer with 30 mM imidazole, and elution with 6 CV purification buffer with 250 mM imidazole.The eluate was concentrated at 4°C using Vivaspin 20 MWCO 10 kDa ultra-concentrators (Sartorius Stedim Biotech S.A., France).Imidazole was finally removed by PD-10 desalting columns (GE Healthcare, Munich) equilibrated on purification buffer.Up to 150 mg purified protein was obtained from 1 l TB-medium.Purification buffer: 20 mM potassium phosphate, pH 6.5.Elution buffer concentrate: 20 mM potassium phosphate, 2 M imidazole/ HCl, pH 6.5.Purification samples of KDPG and KDPGal aldolase were analyzed by subjecting equal amounts of protein (determined by Bradford assay [33]) to the SDS-PAGE using colloidal Coomassie stain [34]

(S2 Fig).
Expression and purification of DERA wt and variants.Recombinant expression of DERA was performed in 3 l baffled Fernbach flasks using E. coli BL21(DE3) cells.Chemically competent cells were transformed with the respective DERA-vectors, followed by selection on LB plates supplemented with ampicillin. 1 l terrific broth (TB) with 100 μg ml -1 ampicillin was inoculated with a 5 ml overnight culture prepared from a single colony and grown at 37°C.After incubation at 25°C and 120 rpm for 8 h, expression of deoC gene was induced by the addition of 0.1 mM IPTG.After prolonged incubation of 15 h at 25°C, cells were harvested.For disruption using Ultrasonic Desintegrator Sonoplus HP 2070 (Bandelin, Berlin), the cells were suspended 20% (w/v) in triethanol amine buffer (100 mM, pH 7).Centrifugation at 18,000 rcf and 4°C resulted in cell free crude extract.For purification 2 ml of the cell free crude extract were loaded on a Q Sepharose HP column (5 ml) applied in an ÄKTA Purifier System (GE Healthcare, Munich) and eluted with triethanol amine buffer (100 mM, pH 7).The eluate was concentrated at 4°C using Vivaspin 20 MWCO 10 kDa ultra-concentrators (Sartorius Stedim Biotech S.A., France).

Protocol of DERA-Catalyzed Reactions for Library Screenings
The DERA catalyzed aldol reaction of acetaldehyde and propanal to (R)-or (S)-hydroxypentanal was performed in analytical scale, followed by a derivatization with 2,4-dinitrophenylhydrazine (2,4-DNPH).The products were extracted and measured by chiral HPLC.A stock solution, containing 7 μl of propanal (0.10 mmol), 7 μl of acetaldehyde (0.12 mmol, 1.3 eq.), and 14 μl dimethyl sulfoxide (0.2 mmol, 2.0 eq.), were pipetted to 0.5 ml of triethanolamine buffer (0.1 M, pH 7).After addition of 20 μl enzyme solution (cell free crude extract), the tubes were shaken for 16 h at 25°C and 200 rpm.For the derivatization step a solution of 40 mg 2,4-DNPH (0.2 mmol) and 7 μl conc.HCl (0.31 mmol) in 0.5 ml dimethyl sulfoxide was added to the DERA reaction mixture and stirred for 2 h at 50°C, 300 rpm.Extraction and sample preparation for HPLC was done according to Dick et al. [35] Samples without enzyme as well as a reaction with E. coli cells transformed with an empty vector were used as negative controls.

Kinetic measurements with DERA
The activity was measured for the retro-aldol reaction of the natural substrate 2-deoxy-Dribose-5-phosphate to glyceraldehyde-3-phosphate, which is converted by auxiliary enzymes to glycerol-3-phosphate under NADH consumption.The reaction is monitored on a photometer at 340 nm.The assay was done according to Dick et al. [36].

Identification of Key-Residues for KDPG and KDPGal
DERA is the only aldolase using acetaldehyde as natural nucleophile (donor).Thus, there is no enantiocomplementary enzyme available.We therefore set out to learn from the homologous aldolases KDPG and KDPGal from E. coli how these enzymes accomplish enantioselectivity and use this knowledge to invert the enantioselectivity of DERA.Like DERA, these enzymes have an (αβ) 8 -architecture and belong to the type I aldolases.In contrast to DERA, KDPG and KDPGal aldolase utilize pyruvate rather than acetaldehyde as nucleophile.However, they naturally catalyze the hydrolysis of the respective 2-keto-3-deoxy-hexose acid to pyruvate and glyceraldehyde-3-phosphate, which is the same electrophile as in the natural DERA reaction (Fig 1).Our hypothesis is that the active sites of DERA as well as of KDPG and KDPGal consist of one part that is adapted to the nucleophile and another part that is adapted to the electrophile (Fig 2).The molecular basis of stereoselectivity should be inherent to the electrophilic part of the active site, which is presumably similar in both classes of aldolases, because the nucleophile adopts the very same position independent of the kind of pyruvate-dependent aldolase.Thus, the first step in designing an enantiocomplementary DERA was to identify stereodeterminants in the electrophilic part of the active site of the pyruvate-dependent aldolases.The identified stereodeterminants shall then be transferred onto DERA to alter its stereoselectivity.
Slight conformational variations within the active site may affect the respective enantioselectivity.Hence, MD simulations were conducted to account for the dynamic behavior of the active sites.The MD simulations were carried out over 10 ns with snapshots extracted every 1 ps.The stereoselectivity-determining region was defined as all atoms within 6 Å radius of the newly formed stereogenic center (Fig 3E).The geometry of the active site was analyzed by calculating the distance of particular atoms to the ε-amino group of the active lysine as a reference point as to where the substrate would be present.
Analyses of the MD simulations revealed that most measured distances in the active sites of KDPG and KDPGal are similar.However, some amino acids at homologous sequence positions have significantly different distances to the ε-amino group of the active lysine, namely I12/V20 (in β 1 -sheet), E37/E45 (in β 2 -sheet), and V154/T161 (in β 7 -sheet), in the order KDPGal/KDPG aldolase respectively (Fig 3A -3D and 3F).The glutamate residues E37/E45 are part of the essential proton shuttling system enabling the formation of the nucleophilic enamine species.Hence, we assumed that major interactions are formed between these glutamates and the nucleophile, but not to the newly formed stereogenic center.Thus, we focused our design efforts on the two remaining sites, I12/V20 and V154/T161.[17].The substrate is highlighted in stick representation.We hypothesize that the blue part of the protein is adapted to fix the nucleophilic carbonyl (here pyruvate), whereas the orange part is adapted to the electrophilic carbonyl (here glyceraldehyde-3-phosphate) and, thus, provides the basis for stereoselectivity.doi:10.1371/journal.pone.0156525.g002 Site V154/T161 was previously shown to affect stereoselectivity in KDPG and KDPGal aldolase, respectively [17,[38][39][40][41]. To investigate this, we introduced V154T KDPGal and T161V KDPG single point mutations.So far, the I12/V20 site has not been investigated as to its influence on the stereoselectivity.However, crystal structures (Fig 3) suggest that single point mutations of this site may not be sufficient to affect stereoselectivity because the β 1 -sheet places V20 KDPG even closer to the active lysine than the more extended homologous I12 KDPGal .In particular, P19 KDPG caused a kink in the β 1 -sheet, that way pushing V20 KDPG deeper into the active site pocket than I12 KDPGal .Thus, to test the influence of the I12/V20 site, we transferred the entire β 1 -sheet of KDPG to KDPGal (chimβ 1;KDPGal : V17L, V18I, P19A, V20I, I21L) and vice versa (chimβ 1;KDPG : L9V, I10V, A11P, I12P, L13I).Finally, combinations of the single point mutations and the β-sheet transfer were introduced (chimβ 1 T161V KDPG and chimβ 1 V154T KDPGal ).
To further support the suggested mutation sites, we conducted phylogenetic analyses of bacterial KDPG and KDPGal aldolases (Fig 4).The I12/V20 site found by MD simulations is virtually fully conserved throughout all sequences, whilst the degree of conservation in the surrounding patches varies strongly.The P19 in KDPG mentioned above is highly conserved as is the homologous position A11 KDPGal.This supports our hypothesis that single amino acid exchanges at the I12/V20 site alone are insufficient to affect the stereoselectivity.

Altering Stereoselectivity in Pyruvate-Dependent Aldolases
The effect of the introduced mutations on the aldolases´stereoselectivity was investigated by enzymatic reactions in synthetic (aldol) direction using pyruvic acid (2) as nucleophile and propanal (1) as electrophile (Fig 5).This approach has various advantages over measuring the aldolases´enzymatic activity using their natural substrates (2-keto-3-deoxy-gluconate/galactonate-6-phosphate).First, conducting the assay in aldol direction yields an effective enantiomeric excess (ee) under competitive conditions and not an apparent value from comparative steady-state kinetics.Second, using the simplified substrate propanal leads to a simplified product portfolio, whereas the natural electrophile glyceraldehyde-3-phosphate would yield mixtures of lactones, furanoses, and pyranoses including the respective diastereomers.
Either the mutation in the β 7 -sheet or the transferred strand β 1 affected the stereoselectivity (decrease of ee by up to -49%ee; Table 1).Combining both modifications strongly enhanced the effect and resulted in a slightly reversed stereoselectivity of KDPG (ee = -5%) and KDPGal (ee = 10%).Comparing these results with those from Walters et al., the enantioselectivities [log (E)] could be reproduced with the same tendency for the known mutants.However, for the uncompetitive assay used by Walters et al. the selectivities were much more optimistic [45,46].

Development of an Aldol-Screening for DERA
One challenge was to develop a comparable screening method for KDPG/KDPGal aldolases and DERA.A screening in aldol direction should be applied to measure the ee of the newly formed stereogenic center (in comparison to a retro-aldol based assay [47,48]).In order to design a comparable assay we used the same electrophile propanal (1) as above (Fig 5).The choice of the nucleophile is restricted by the corresponding enzymes, pyruvate for KDPG/ KDPGal and acetaldehyde for DERA (Fig 6).The stereoselectivity of the respective DERA-catalyzed aldol reaction was assessed in a model reaction by the generation of both enantiomers of 3-hydroxypentanal (6).For KDPG and KDPGal aldolase, the products 4 were analyzed by gas chromatography using a chiral stationary phase.To be able to analyze the unstable aldol products 6 of DERA, they were derivatized.The formed hydrazones 7 [49][50][51] were extracted and measured by HPLC.Because this procedure may cause lot-specific differences during Table 1.Results of the stereoselectivity assay with the variants of the pyruvate-dependent aldolases KDPG and KDPGal.Own experiments have been carried out as triplicates from different enzyme-lots; standard-deviations are given for each ee.A negative control without enzyme did not show any product formation.Complex mutations: chimβ 1;KDPGal : V17L, V18I, P19A, V20I, I21L and chimβ 1;KDPG : L9V, I10V, A11P, I12P, L13I.

Transfer of Key-Residues onto Homologous DERA
As both sites I12/V20 (β 1 -sheet) and V154/T161 (β 7 -sheet) were confirmed as important determinants of KDPG stereoselectivity, we transferred them to corresponding sites in DERA.These corresponding sites were identified, based on a structural alignment, as T18 and L20 (both β 1 -sheet) as well as A203 (β 7 -sheet), respectively (Fig 3A).For a first generation of DERA variants, the respective sites were mutated using the degenerate codon NDT, that way including twelve different amino acids, in order to exhaustively explore the effect of these positions on stereoselectivity.
The results of the first generation of DERA variants show that T18 and L20 in the β 1 -sheet have a strong influence on DERA stereoselectivity ( Fig 7).Especially T18S (ee = 20%) and T18R (ee = 23%) led to a decreased enantiomeric excess.These results support our hypothesis that the homologous sites are transferable and indicate that T18 is more important for stereoselectivity than L20.
Of the variants of the A203 position in the β 7 -sheet, only A203S and A203G showed sufficient activity to perform the aldol screening.The two variants contained amino acids of similar or smaller size than the original alanine.A203G showed the largest effect on stereoselectivity compared to DERA wt (ee = 88%); however, the overall effect is moderate compared to effects due to variants of the T18 position.With reference to these results, a second generation of DERA variants was constructed.Double mutations of the positions T18 and A203 were created.As there is no smaller amino acid than glycine, which resulted in the largest effect on stereoselectivity in A203G DERA, we aimed at further expanding the binding pocket for the electrophile by deleting the amino acids G204 and G205.Structural models generated from the DERA crystal structure by deleting these residues, changing A203 to glycine, and minimization indicated that these deletions (ΔG204, ΔG204 Δ205) should result in a retraction of the loop at the end of the β 7 -sheet (Fig 8).Furthermore, to study the effects of double mutations, combinations of the A203G and T18 variants were investigated.
The combination of A203G and T18S did not lead to a further decreased enantiomeric excess (ee = 54%).The deletion of G204 and G205 in combination with A203G, respectively, did not lead to a change in enantiomeric excess either.Finally, the combination of A203G, one or two glycine deletions, and mutation in position T18 to serine decreased the enantioselectivity significantly (ee = 63-64%, Fig 9), although this value is higher than that obtained for the A203G T18S variant above (ee = 54%).
Michaelis-Menten kinetics on the natural substrate of the DERA of the variant with the highest change in enantioselectivity T18S as well as the wildtype enzyme were performed.The results in Table 2 show an only ~3-fold decreased catalytic efficiency for T18S compared to the wildtype, leading to a catalytic efficiency of ~36 s -1 mM -1 .

Conclusions
Enzymes are attractive catalysts because of their strict stereoselectivity.In some cases natural enantiocomplementary versions are available.Unfortunately, if an enantiocomplementary enzyme is missing, synthetic applications are severely restricted.Although DERA is such an example, this enzyme is still highly attractive for synthetic purposes [11,53].Various endeavors, such as rational enzyme design and directed evolution approaches, were employed for different enzymes to invert their stereoselectivity [54][55][56][57][58][59][60].For the pyruvate dependent aldolases very convincing approaches for changing the stereoselectivity have been conducted [39,40,61].However, to our knowledge, no alteration of the stereoselectivity of DERA has been described so far.
Here, we present an approach that aims at transferring the stereoselectivity-determining parts from homologous, pyruvate-dependent, and enantiocomplementary aldolases KDPG and KDPGal onto DERA.In order to apply this promising strategy, the stereoselectivity-determining sites of KDPG and KDPGal were identified using MD simulations, geometric, and phylogenetic analyses.For the pyruvate-dependent aldolases, we demonstrated the importance of I12/V20 (β 1 -sheet) and V154/T161 (β 7 -sheet) as stereoselectivity-determining sites, using a novel assay working in aldol instead of retro-aldol direction of the catalyzed reaction to elucidate the stereochemistry of the respective products.The advantage of the novel assay is that it is unbiased by non-competing effects contrary to previous assays exploiting the retro-aldol reaction in separate reactions of the two enantiomers [17]; as a result, for mutation sites previously described, lower stereoselectivity inversions compared to previous work were observed (Table 1).However, these obtained values reflect the synthetic capabilities more realistically.Although it was not possible to completely invert the enantioselectivity of KDPG or KDPGal, it was possible to pin-point I12/V20 (β 1 -sheet) and V154/T161 (β 7 -sheet) as essential residues for stereoselectivity.Variants of these restricted sites allowed not only to reduce stereoselectivity but even inverted it compared to KDPG (ee = -5%) and KDPGal (ee = 10%).This demonstrated that these sites are important for stereoselectivity.
We then transferred these sites onto DERA based on considerations derived from a structural superpositioning of the enzymes.The spatial positions of the stereoselectivity-determining sites should be the same because of their similar geometrical composition and their very same mode of action.Indeed, mutating the found sites caused a significant change in enantioselectivity within the first series of mutations (best ee found for T18S: 20%).Combinations of the identified stereoselectivity-determining sites did not further change the stereoselectivity.DERA has a very different mechanism in realizing its stereoselectivity compared to the homologous pyruvate-dependent aldolase, thus, additional stereoselectivity-determining sites might be involved as seen for the pyruvate-dependent aldolases.Hence, inversion of the stereoselectivity of DERA is likely an especially challenging problem; obviously further stereo-determinants of DERA must be characterized in future studies.However, the presented homologous  grafting approach was shown to serve as a new entry point for difficult tasks for finding enantiocomplementary enzymes.

Fig 2 .
Fig 2. Top view on KDPGal (PDB ID: 2V82)[17].The substrate is highlighted in stick representation.We hypothesize that the blue part of the protein is adapted to fix the nucleophilic carbonyl (here pyruvate), whereas the orange part is adapted to the electrophilic carbonyl (here glyceraldehyde-3-phosphate) and, thus, provides the basis for stereoselectivity.

Fig 3 .Fig 4 .
Fig 3. Geometric analyses of homologous aldolases.(A) -Superposition of DERA (PDB ID: 1JCL, blue) [37], KDPG (PDB ID: 1EUA, orange) [16], and KDPGal (PDB ID:2V82, green) [17] illustrates the structural similarity of these TIM-barrel proteins.(B-D) -Close-up views of the active sites of the three proteins.The active lysine is marked by a triangle (▼).(E) -Superposition of the active sites of KDPG and KDPGal aldolase.Atoms within a radius of 6 Å around the newly formed stereogenic center are indicated as spheres.(F) -Distances from MD simulations between the ε-amino group of the active lysine and all atoms within the 12 Å-Ø-sphere around the newly formed stereogenic center for both the KDPG and KDPGal aldolase, respectively.The mean values over all conformations of the MD simulation are given; bars indicate the standard deviation.doi:10.1371/journal.pone.0156525.g003

Fig 5 .
Fig 5. Model aldol reaction for testing the stereoselectivity of KDPG, KDPGal, and specific mutants.The aldol reaction of propanal (1) and pyruvic acid (2) leads to the enantioselective generation of 3 and, by acidic lactonization, to butyrolactone 4. doi:10.1371/journal.pone.0156525.g005 derivatization and extraction, each series of measurements included a triplicate measurement of DERA wt as a reference.The negative control without enzyme did not show any product formation.A fixed volume of cell free crude extract of each DERA variant was used for the screening reactions whilst the expression level was controlled visually by means of SDS-PAGE (S3-S8 Figs).Only DERA variants T18D, T18C, and T18G had significantly lower expression levels than DERA wt, and for the L20H mutant virtually no soluble protein was expressed (see S1 Fig for details on the mutations).

Fig 7 .
Fig 7. Enantiomeric excesses of the first generation variants of DERA at sites T18, L20, and A203, respectively.The ee was determined by the aldol screening (Fig 6) using chiral HPLC.The error bars represent the standard deviation of triplet measurements.A negative control without enzyme did not show any product formation."*": Results for variants with an activity lower than 2.0% of the DERA wt were omitted.doi:10.1371/journal.pone.0156525.g007

Fig 9 .
Fig 9. Enantiomeric excesses of the second generation of variants of DERA at sites T18 combined with A203G, A203G ΔG204, and A203G ΔG204 ΔG205, respectively.The ee was determined by the aldol screening (Fig 6) using chiral HPLC.The error bars represent the standard deviation of triple measurements.A negative control without enzyme did not show any product formation."*": Results for variants with an activity lower than 2.0% of the DERA wt were omitted.doi:10.1371/journal.pone.0156525.g009 [17]ulated from the enantiomeric excess (ee) value for the formation of compound 7 (Fig 5)using the equation log(E) = log[(-1-ee)/(ee-1)] § Values from Walters et al.[17], deduced from steady-state kinetics with natural substrates in retro-aldol direction.