Asymmetric bioreduction of β-ketoesters derivatives by Kluyveromyces marxianus : influence of molecular structure on the conversion and enantiomeric excess

This study presents the bioreduction of six β-ketoesters by whole cells of Kluyveromyces marxianus and molecular investigation of a series of 13 β-ketoesters by hologram quantitative structure-activity relationship (HQSAR) in order to relate with conversion and enantiomeric excess of β-stereogenichydroxyesters obtained by the same methodology. Four of these were obtained as (R)-configuration and two (S)-configuration, among them four compounds exhibited >99% enantiomeric excess. The β-ketoesters series LUMO maps showed that the β-carbon of the ketoester scaffold are exposed to undergo nucleophilic attack, suggesting a more favorable β-carbon side to enzymatic reduction based on adopted molecular conformation at the reaction moment. The HQSAR method was performed on the β-ketoesters derivatives separating them into those provided predominantly (R)or (S)-β-hydroxyesters. The HQSAR models for both (R)and (S)-configuration showed high predictive capacity. The HQSAR contribution maps suggest the importance of β-ketoesters scaffold as well as the substituents attached therein to asymmetric reduction, showing a possible influence of the ester group carbonyl position on the molecular conformation in the enzyme catalytic site, exposing a β-carbon side to the bioconversion to (S)and (R)-enantiomers.


INTRODUCTION
Chiral β-hydroxyesters are widely used in the chemical-pharmaceutical industry as intermediates of organic synthesis and some of them have been used for the synthesis of biologically active compounds.For example, the ethyl 3-hydroxybutanoate plays an important role in the synthesis of (+)-decarestrictine L, an inhibitor of cholesterol biosynthesis (Wang et al. 2013), and the methyl 3-hydroxypentanoate is the key chiral building block in the synthesis of (-)-serricornine, the sex pheromone from cigarette SIMONE S.S. OLIVEIRA et al. beetle, Lasioderma serricorne (Pilli and Riatto 1998).In the same way, ethyl 3-hydroxyhexanoate is an important intermediate for the synthesis of (+)-neopeltolide, a potent in vitro antiproliferative agent against the growth of several cancer cell lines and also antifungal activity against Candida albicans (Ramos et al. 2011, Ghosh et al. 2013).And also both enantiomeric forms of methyl 3-aryl-3-hydroxypropionates are important building blocks in the synthesis of several chiral drugs, fine chemicals and pesticides (Borowiecki andBretner 2013, Liu andLiu 2015).
Chiral β-hydroxyesters from β-ketoesters may be obtained by chemical or microbial asymmetric reduction.The chemical asymmetric methods involve metal catalysts that may leave residues in products and should be avoided for pharmaceuticals.Aside from that it generally requires additional steps for protection/deprotection of functional groups and extreme reaction conditions (Hagemann et al. 2005, Ng and Jaenicke 2009, Floris et al. 2009, Sheldon 2016).Alternatively the bioreduction occurs under nontoxic and mild reaction conditions (ambient temperature, atmospheric pressure, and aqueous medium) with little impact on the environmental and avoids the burden of groupprotecting procedures (Milner and Maguire 2012, Oliveira et al. 2013, Regil and Sandoval 2013, Sheldon 2016).
Despite the bioreduction approaches require special precautions, such as sterile conditions to obtain the biomass and a biphasic system or substrate/product reservoir in order to get around the low water solubility of the substrates, this method is attractive because of its advantages and has been widely used for asymmetric reductions (Zeror et al. 2010, Regil and Sandoval 2013, Sheldon 2016, Wachtmeister and Rother 2016).
Asymmetric bioreductions can be performed using both enzymes and whole cells.The use of whole cells has the additional advantage of containing the required cofactors and regenerates them during the reaction (Milner and Maguire 2012, Nakamura et al. 2003, Illanes et al. 2012, Venkataraman and Chadha 2015, Wei et al. 2016).
Several microorganisms have been used to reduce different β-ketoesters and most of them catalyze the reaction to the (S)-configuration (Ramos et al. 2009a).Saccharomyces cerevisiae (baker's yeast) is the most used microorganism to obtain some chiral β-hydroxyesters (Zeror et al. 2010, Fow et al. 2008) due to its ease of handling and commercial availability.However, this method gives predominantly (S)-configuration products, with wide range of enantiomeric excess (Zeror et al. 2010, Dahl and Madsen 1998, Mahmoodi et al. 2006).
The yeast Kluyveromyces marxianus has several characteristics that make it an excellent choice for commercial processes.These includes low pH tolerance, high resistance to furfural and alcohols, a broad range of fermentation temperatures (thermotolerance), and the ability to grow fast and on a wide variety of inexpensive carbon source (Moreno et al. 2012, Chang et al. 2014, Galindo-Leva et al. 2016).Some biotechnological applications have been reported using this yeast, such as food, beverages, enzymes, and fine chemicals production (Foukis et al. 2012, Lane et al. 2011).
In the present work six β-ketoesters (compounds 1-6) were submitted to asymmetric bioreduction by whole cells of K. marxianus.The bioconversion result of these compounds were added to seven others β-ketoesters (compounds 7-13) from the literature to study the structural influence of β-ketoesters substrate on the activity of this yeast (Table I).In an attempt to understand which structural properties are crucial for the enantiomers biosynthesis, the two-dimensional molecular features of the β-ketoesters series were explored using the hologram quantitative structureactivity relationship (HQSAR) (Rodrigues et al. 2002, Magalhães et al. 2013).

MICROORGANISM AND GROWTH CONDITIONS
The yeast Kluyveromyces marxianus belongs to the collection of the Departamento de Engenharia Bioquímica of the Escola de Química at the Universidade Federal do Rio de Janeiro, Brazil.
Cells of K. marxianus were allowed to grow at 30 °C under 150 rpm for 48h in Erlenmeyer flask (250 mL) with 50 mL of sterile growth medium containing 10 g/L glucose, 5 g/L yeast extract, 5 g/L peptone, 1 g/L (NH 4 ) 2 SO 4 , 1 g/L MgSO 4 .7H 2 O and pH 6.5.The grown cells were harvested by centrifugation at 3500 rpm for 10 min.The cells were washed twice with distilled water and used in the bioconversion.

BIOCONVERSION CONDITIONS AND ANALYSIS OF THE PRODUCTS
The bioconversion was carried out in a 250 mL Erlenmeyer flask with 50 mL medium containing 50 g/L glucose and 1 g/L MgCl 2 and pH 6.5 (Ramos et al. 2009a).The K. marxianus cells were inoculated to the bioconversion's medium (5 g/L, dry weight).After 30 min of inoculation under agitation (150 rpm) at 30 ºC, each Erlenmeyer flask received appropriate β-ketoester (0.5%) in aqueous-ethanol.After 24h of incubation at the same conditions, the cells were harvested by centrifugation and the supernatant was extracted with ethyl acetate.The organic phase was dried (anhydrous MgSO 4 ), filtered and concentrated under vacuum.The experiments were performed in triplicate.

MOLECULAR MODELING AND DETERMINATION OF THREE-DIMENSIONAL FRONTIER ORBITAL MAPS
The computations and molecular modeling were performed using SPARTAN´10 software (Wavefunction, Inc., Irvine, CA, USA).The molecular structure of the β-ketoesters series (1-13) were designed and submitted to the calculations simulating the vacuum, in a non-ionized state and without any geometric constraint.Conformational analysis for the selection of the lowest energy conformer was performed by the molecular mechanics method using force field MMFF (Merck Molecular Force Field) (Halgren 1996).Thereafter the conformers were subjected to geometry optimization using the semi-empirical method RM1 (Recife Model 1) (Rocha et al. 2006).The conformer with the lowest energy was subjected to single point calculation by the DFT-B3LYP method based on quantum mechanics, with 6-31G* basis set (Becke 1993, Lee et al. 1988, Kohn et al. 1996).Subsequently, in order to assist the understanding of the stereoelectronics aspects involved in the asymmetric reduction of the β-ketoesters (1-13) by K. marxianus and the favorable points for the chemical reaction, the molecular frontier orbital LUMO (Lowest Unoccupied Molecular Orbital) were evaluated (Oliveira et al. 2013).

HQSAR MODELS
The hologram QSAR method that employs fragment fingerprints as predictive variables of the activity (Rodrigues et al. 2002) were used to investigate the influence of molecular structure of β-ketoesters (1-13) on the conversion and enantiomeric excess of β-hydroxyester produced by K. marxianus activity.The asymmetric reduction of all compounds was expressed as percentage of (S)-and (R)enantiomers, which were converted to logarithmic values before HQSAR models preparation.
The initial set of 1-13 compounds was divided into training and test sets.The training set (11 compounds) represents structurally different compounds with diverse results obtained for the yeast activity.In order to evaluate the predictive capacity of the model the 2 remaining compounds were assigned to the test set.To avoid possible problems during the external validation, the test set containing compounds with low and high asymmetric reduction.
The HQSAR studies were performed using the SYBYL-X 1.2 software package (Tripos International, St. Louis, MO, USA).The structures of β-ketoesters were converted into fragments initially using the default fragment size of 4-7 atoms per fragment.All fragments were allocated in defined molecular hologram lengths (53, 59, 61, 71, 83, 97, 151, 199, 257, 307, 353 and 401 bins) and fragment distinction analysis was performed in terms of atoms, bonds, connectivity, hydrogen atoms and hydrogen bond donor/acceptor atoms.Since these parameters may affect HQSAR models, different combinations of these parameters were considered during the HQSAR runs (Magalhães et al. 2013).Selection of HQSAR models was done on the basis of q 2 values, known as leave-oneout cross-validated r 2 , the most common statistic criteria used for model evaluation.Although the prerequisite for a predictive QSAR model is q 2 >0.5, is fundamental an external test validation (Golbraikh and Tropsha 2002).
Several QSAR models were generated for each distinguishing fragment based on partial least squares (PLS) analysis.All QSAR models were obtained using PLS.Moreover, internal validation was performed by leave-one-out cross-validation and the external validation was performed with the test set compounds, which were not considered for the HQSAR model development (Shinde et al. 2012).
It is noteworthy that among the synthetic method to obtain the hydroxyesters presented here, there are just few reports that use the Kluyveromyces genus (Ramos et al. 2011, Xu et al. 2013), one of which used carbonyl reductase of K. thermotolerans (Xu et al. 2013).The asymmetric bioreduction by whole cells of K. marxianus method was particularly effective in obtaining the (R)-hydroxyesters in high conversions and ee rate (Table I).Other methods described in the literature leading to the (R)-hydroxyesters (1'b, 3'b, 4'b) requires more laborious and expensive approach, that is the complete inhibition of enzymatic activity for (S)-enantiomer using different microorganisms and growth conditions, recombinant microorganisms or enzyme inhibitors generally toxic and volatile (Ramos et al. 2011, Venkataraman and Chadha 2015, Fow et al. 2008, Dahl and Madsen 1998, Dao et al. 1998, Srivastava et al. 2015, Chen et al. 2015, Zhang et al. 2015).
The results demonstrated that yeast K. marxianus (Oliveira et al. 2013, Molinari et al. 1999), also easy to handle and growth like Saccharomyces cerevisiae, led to the (R)-hydroxyesters (1'b-4'b) without any other additive and in a greater extent than that reported for S. cerevisiae (Zeror et al. 2010, Mahmoodi et al. 2006).This is probably due to the different selectivity of the enzymes with specificity for the (R)-configuration in each yeast.On the other side, β-ketoesters 5 and 6 were reduced to their (S)-hydroxyesters (5'a and 6'a) with 44% and >99% ee, respectively (Table I).If on the one hand 5'a has also been reported by different bioreduction methods (Fow et al. 2008, Bisogno et al. 2009, Nakamura et al. 1985, 1989, Patel et al. 1992), on the other there are few and recent reports about 6'a from biocatalysis of 6 acetate (Borowiecki andBretner 2013, Liu andLiu 2015).Because of the higher ee and the simplicity of the synthetic methodology these results showed that the bioreduction method using whole cell of K. marxianus is an attractive alternative to obtain the enantiopure hydroxyesters (2'b-4'b and 6'a).

MAPPING OF MOLECULAR STEREOELECTRONICS CHARACTERISTICS
In order to evaluate on the influence of chemical structures of substrates in the bioreduction using whole cell of K. marxianus 13 different β-ketoesters compounds (1-13) were analyzed, among which the ones presented in this paper (6 compounds) and the other seven from previously published work (Oliveira et al. 2013, Ramos et al. 2013, 2009a, b, Ribeiro et al. 2009, 2014).
Analysis of the LUMO electron density map contribution from 1-13 showed a highlighted LUMO area (blue) at keto group region that stand out its higher tendency to react with nucleophiles leading to β-hydroxyesters (Figure S1 -Supplementary Material).Moreover, based on their minimal local energy conformation, the compounds display different probabilities of nucleophilic attack on both sides of keto group, with exception of compounds 5 and 10 in which both sides of the keto group have been similarly exposed.However, these conformations tend to modify within the enzyme catalytic site according to the local topology, exposing preferably one side of the keto group to the amino acid residues.
Whether there is a single type of enzyme for these substrates (1-13), it would be able to make β-ketoesters conversions to both (S)-and (R)enantiomers.In this case, the conformational flexibility of the β-ketoester scaffold may acquire different molecular arrangements in the catalytic site of enzyme, where the functional groups of this moiety may interact with different amino acids depending on the molecular structure of the derivative.Assuming that, the β-ketoesters derivatives should adjust the molecular conformation to the topology of catalytic site in complementarity with the enzymatic conformational motions.
However, in the hypothesis of the existence of two or more highly selective enzymes capable of catalyzing the asymmetric bioreduction of these β-ketoesters derivatives it would consider that the molecular structure of their substrates influence for the (S)-enantiomer or (R)-enantiomer conversion.In this case, the β-ketoesters derivatives should adjust the molecular conformation to the topology of catalytic site in complementarity with the enzymatic conformational motions.
We assume that the structure of the substituents linked to the β-ketoester scaffold is directly responsible for the steric effects by the influence on the conformation and have influence on the electropositive area at carbon atom of the keto group.The LUMO maps reinforce that bioconversion values are related to the effects of the substituents on the molecule conformations directly affecting the enzymatic asymmetric reduction of β-ketoesters (Oliveira et al. 2013, Dao et al. 1998).

HQSAR MODELS
HQSAR uses an extended form of fingerprint, known as molecular hologram which encodes

TABLE I Bioconversion of β-ketoesters series (1-13) by Kluyveromyces marxianus to correspondent (S)-(1'a-13'a) or (R)-hydroxyesters (1'b-13'b)
. information.This requires values to be selected for the parameters that specify the hologram length, as well as the size and type of fragments to be encoded (Flower 1998).The molecular fragments were generated using the fragment distinction parameters based on atoms, bonds, connections, hydrogen atoms and hydrogen bond donor/acceptor atoms.The HQSAR models were first generated using the default fragment size (4-7 atoms) combined with various fragment types and various hologram lengths.To identify how the fragment size could influence the statistical parameters, different fragment sizes were tested (2-5, 3-6, 4-7, 5-8, 6-9, 7-10, 8-11, and 9-12) on the three best fragment distinction parameters of these models having the highest statistical indexes for (S)-and (R)-β-hydroxyester conversion.Table II shows the highlight HQSAR models.
The best HQSAR model for conversion of these β-ketoesters derivatives to (S)-β-hydroxyester was found to be model 47S, with atoms, connectivity, hydrogen atoms and hydrogen bond donor/acceptor atoms as the fragment distinction parameters, and 5-8 atoms as the fragment size, showing good predictive capacity (q 2 = 0.75), high data fitting (r 2 = 0.97), and low cross-validated standard error (SE cv = 0.66) (Table II).The descriptors contributions to bioconversion by K. marxianus in the model 47S show the most critical parameters.The fragments of 5-8 atoms may be determinant for (S)-enantiomer conversion.
Regarding the best HQSAR model for bioconversion to (R)-β-hydroxyester, 42R was the best model.With connectivity and hydrogen atoms as the fragment distinction parameter, and 8-11 atoms as the fragment size, it shows good predictive capacity (q 2 = 0.86), high data fitting (r 2 = 0.97), and low cross-validated standard error (SE cv = 0.48) (Table II).The descriptors contributions to bioconversion by K. marxianus in the model 42R show that the critical parameters from β-ketoesters to (R)-enantiomer conversion are not the same to (S)-enantiomer conversion.

HQSAR CONTRIBUTION MAPS
The HQSAR contribution map from these models can provide valuable insights about relationship between molecular fragments and bioconversion values in K. marxianus.Hologram QSAR uses molecular holograms and PLS to generate fragment-based structure-activity relationships.According to HQSAR method, it is possible to predict the activity value of a molecule by mapping its structural fragments (Rodrigues et al. 2002).In this sense, HQSAR contribution map analysis may be used as a step on the understanding the individual atomic contributions to the prevalence of (R)-or (S)-enantiomer bioconversion through a color code from red to green in a spectrum.The colors at the green end (yellow, green blue and green) indicate favorable contributions, whereas colors at the red end (orange, red orange and red) indicate unfavorable contributions.The neutral contributions are colored white (Figure 2).
According to the contribution map, the molecular fragment corresponding to the substituents linked to β-ketoester scaffold display important role to (S)-and (R)-conversion.Furthermore the HQSAR contribution maps suggest that β-ketoesters scaffold is important to asymmetric reduction too.It suggests a possible influence from carbonyl of the ester group on the stabilizing the conformation in the enzyme catalytic site for (S)-and (R)-enantiomers bioconversion.This can be seen especially in compounds 2, 3 and 6 which have had their conversions to hydroxyesters influenced by the effect of the alkoxy groups on ester carbonyl in both models.
The β-ketoesters scaffold is the common molecular structure among the present compounds series, indicating the importance of the conformation and the position of the atoms to bioconversion.The substituents attached to β-ketoesters scaffold clearly have effects on the asymmetric reduction.According to both models, hydrogen atoms of the substituents attached to β-ketoesters scaffold should be replaced by others groups with the aim of modifying the asymmetric reduction.A rational attempt would be to modify these hydrogen atoms by bioisosteres monovalent groups such as fluorine, hydroxyl, amino and methyl groups or replace the  III).
Both successful models provide important information about which molecular fragments are directly related to the bioconversion profile in K. marxianus.Moreover, these models may be used to predict the bioconversion of the untested compounds.These data would be useful for application of K. marxianus in bioconversion of any other β-ketoester derivatives.

CONCLUSIONS
In the present study, the wild strain Kluyveromyces marxianus was able to reduce β-ketoesters (1-6) to the corresponding chiral β-hydroxyesters.Four of these were obtained with enantiomeric purity (>99%).Molecular investigation emphasized that the conversion rate and stereoselectivity of the enzymes over the β-ketoesters are related to their different electropositive region observed on LUMO electron density map.Moreover, the β-ketoesters series LUMO maps emphasize the importance of substituent effect and reinforce the influence of the β-ketoesters molecule conformations on enzymatic asymmetric reductions.
Also, the HQSAR contribution maps suggest a possible influence of the ester group carbonyl on the molecular conformation fit in the enzyme catalytic site for (S)-and (R)-enantiomers bioconversion.Thus, it can be suggested that both β-ketoesters scaffold and the substituents therein influence the asymmetric reduction result.Two HQSAR models were established, one for β-ketoesters derivatives conversion to (R)-hydroxyesters and the other to (S)-hydroxyesters, which provide information about the relationship between molecular fragments and K. marxianus bioconversion profile.These models may be used to predict the bioconversion of untested compounds.

TABLE III Experimental and predicted log percentage values of (S)-and (R)-enantiomers formed.
a Test set compounds; b Model 47S; c Model 42R.