Application of the uridine auxotrophic host and synthetic nucleosides for a rapid selection of hydrolases from metagenomic libraries

Summary A high‐throughput method (≥ 106 of clones can be analysed on a single agar plate) for the selection of ester‐hydrolysing enzymes was developed based on the uridine auxotrophy of Escherichia coli strain DH10B ΔpyrFEC and the acylated derivatives 2′,3′,5′‐O‐tri‐acetyluridine and 2′,3′,5′‐O‐tri‐hexanoyluridine as the sole source of uridine. The proposed approach permits the selection of hydrolases belonging to different families and active towards different substrates. Moreover, the ester group of the substrate used for the selection, at least partly, determined the specificity of the selected enzymes.


Introduction
Discovery of novel enzymes and their engineering is the basis of the tremendously fast-developing industrial biotechnology. Enzyme-based organic synthesis constantly needs innovative, diverse and robust biocatalysts (Bornscheuer, 2018). Ester hydrolases (EC 3.1), commonly known as esterases or lipases, represent a diverse group of hydrolases catalysing the cleavage and formation of ester bonds (Jaeger and Eggert, 2002;Dimitriou et al., 2017;Bornscheuer, 2002). Hydrolases have a growing number of applications in biotechnology, particularly in fine chemicals industries, because of their enantioselectivity and regioselectivity. Catalytic promiscuity, no requirement for cofactors as well as stability and activity in organic solvents also stimulate interest in the aforementioned enzymes (Jaeger and Eggert, 2002;Bornscheuer and Kazlauskas, 2004;Rauwerdink and Kazlauskas, 2015;Mart ınez-Mart ınez et al., 2018).
The hunt for novel enzymes, including esterases, is usually based on one of the following strategies: (i) construction of metagenomic libraries, and subsequent functional screening Ferrer et al., 2015;Taupp et al., 2011;Peña-Garc ıa et al., 2016;Ferrer et al., 2016;Steele et al., 2009), (ii) (meta) genome mining based on the homology analysis, chemical synthesis of the target genes and evaluation of the catalytic properties of the recombinant proteins (Mart ınez-Mart ınez et al., 2018), (iii) directed evolution, including a high-throughput scale (HTS) microfluidics screening of the combinatorial libraries of randomly generated enzyme variants (Colin et al., 2015;Bunzel et al., 2018), (iv) rational in silico design followed by an experimental verification of selected variants Packer and Liu, 2015;Bornscheuer et al., 1999;Fern andez-Alvaro et al., 2011). Novel powerful techniques notwithstanding, the functional metagenome screening remains an interesting and promising option, since it offers a possibility of discovering enzymes with unique properties and novel scaffolds applicable for further evolution in vitro. However, the number of tested clones and positive hits strongly depends on the screening method applied, as well as on the substrate used and can vary from several thousands to billions with a success rate ranging from 1:11 to 1:193 200, in the case of ester hydrolases (Peña-Garc ıa et al., 2016;Ferrer et al., 2016).
At present, a functional screening of esterases is usually performed either by employing chromogenic substances (e.g. p-nitrophenyl esters) or by using tributyrinsupplemented agar plates Peña-Garc ıa et al., 2016). In addition, other esters, various fluorogenic substrates or enzyme cascades have been applied for the identification of ester hydrolases-producing microorganisms (Peña-Garc ıa et al., 2016;Rossum et al., 2013). Although both low-throughput and robotic-based HTS systems are being developed, the exploitation of the genotype-phenotype linkage or of the genetic selection is not very common (Reetz et al., 2008). Mutations in the genes of host cell metabolism can be used for discovery of specific enzymes (Forney et al., 1989), their analysis (Delauney et al., 1993) and for the design of antibiotics-free (Dong et al., 2010) or catalytic antibodies-based selection systems (Smiley and Benkovic, 1994). Since only a few of such systems have been developed for the selection of ester hydrolases, the hydrolysis reaction, which liberates glycerol as a carbon source, has been applied for the generation of active esterase mutants (Bornscheuer et al., 1998). The appropriate aspartate esters and the Escherichia coli strain, in which both pathways leading to the synthesis of aspartate are blocked, have been used to select mutants of Bacillus subtilis lipase A (Boersma et al., 2008). Also, a mixture of isosteric (R)-and (S)-enantiomers has been used as a substrate in the bond-breaking reaction that generates a growth-promoting energy source and a growth-inhibiting compound (Reetz and R€ uggeberg, 2002). However, a wider application of such system is slightly restricted, since the starting substrates must be non-toxic to the host organism. In the case of esterasebased generation of energy and/or carbon source, high concentrations of the substrates have to be applied to support a satisfactory growth of the positive clones (Acevedo-Rocha et al., 2014).
Recently, it has been shown that the uracil auxotrophic strain of E. coli can be successfully used for an identification of the specific genes encoding the catabolism of the modified uracil base(Au cynait _ e et al., 2018). Here, we present a method of functional screening of hydrolases based on uridine auxothrophy and the appropriate uridine derivatives that serve as substrates for the ester-hydrolysing enzymes. The proposed method combines the best features of the genotype-phenotype linkage and the flexibility in the chosen substrate hence allowing an efficient selection of esterases from metagenomics libraries. Compared with the known methods developed for screening of esterases/lipases, the proposed selection system has many advantages including a rapidness, a high throughput (millions of clones can be analysed by using a single agar plate) and possibility to apply different substrates for identification of the enzymes with the desired catalytic properties.

Results and discussion
To develop the selection method, 2 0 ,3 0 ,5 0 -tri-O-acetyluridineand 2 0 ,3 0 ,5 0 -tri-O-hexanoyluridine ( Fig. 1)hereafter indicated respectively as compound 1 or 2 throughout the text were chosen as the sole source of uridine, supporting growth only of those recombinant clones (Fig. S1), which encode ester hydrolases that complement the uridine auxotrophy of the E. coli DH10B DpyrFEC::Km (Au cynait _ e et al., 2018) strain by hydrolysis of compound 1 or 2 to uridine (Fig. S1). Due to location of the pyr genes in different parts of the chromosomes of many bacteria, the rationale of using E. coli mutant harbouring triple mutations in the genes encoding the pathway of uridine biosynthesis was to lower a level of the false-positive hits. Regarding the application of two structurally different substrates used for the selection, it was supposed that the uridine derivatives with a varied acyl length could predetermine the properties of the selected hydrolases, and the enzymes with a different preference to acyl size would be identified. The principle of the selection method is shown in Fig. 2.
The metagenomic DNA isolated from different soil samples was partially digested with several restriction endonucleases, and the fragmented DNA was used for the construction of metagenomics libraries. In total, 19 libraries (Table S1) were tested. Clones exhibiting acylesterase activity were selected on MD medium containing 100 lg ml À1 ampicillin and 40 lg ml À1 kanamycin, as well as 20 lg ml À1 of the compound 1 or 2 (Fig. 1). In total, 87 positive clones were selected. All clones were re-streaked on MD medium without uridine or uridine derivative. Four of the selected hits capable of forming colonies on MD medium without the compound 1 or 2 were considered as false positives (approximately 5%) and were omitted from the further analysis. The plasmid DNA from the remaining clones was isolated, and the fragments obtained after restriction digestion were analysed by sequencing to exclude the redundancy. Hence, 30 clones were chosen for further analysis.
Bioinformatics analysis showed that the selected clones exhibiting esterase activity contained ORFs with medium (31%) to high (100%) sequence identity to proteins found in the databases ( Table 1). Ten of the closest homologs were annotated as hypothetical proteins.
The phylogenetic analysis of the selected hydrolases showed that the enzymes represent very diverse groups of proteins (Fig. 3). Most of the hits were representatives of ABhydrolase (a/b hydrolase) superfamily (SSF53474) (19 hits) and SGNH hydrolases (SSF52266) (4 hits) followed with proteins belonging to b-lactamases (SSF56601) (2 hits), a/b hydrolases/galactose-binding domain-like (2 hits), glycosyl hydrolases (SSF51445) (1 hit), N-acyltransferase superfamily (SSF55729) (1 hit) and one DUF998 family protein (Fig. 3). All identified proteins of the ABhydrolases group had a conserved Gly-x-Ser-x-Gly catalytic motif (Bornscheuer, 2002) and were distributed among different families: a/b hydrolase-1, a/b hydrolase-3 and peptidase S9 (Fig. S2A). Two esterases MO4B and EN1H consisted of two domains -ABhydrolase and galactose-binding-like domain. The common function of these domains is to bind to specific ligands, such as cell-surface-attached carbohydrate substrates. Both hydrolases had the conserved motif of serine proteases (Delauney et al., 1993), and the consensus sequence surrounding the active-site serine had been  The principle for the selection of ester-hydrolysing enzymes. Metagenomic DNA isolated from the environmental samples is fragmented, inserted into an appropriate vector and used to transform the competent cells of E. coli DH10B DpyrFEC::Km. A mineral medium containing 2 0 ,3 0 ,5 0 -tri-O-acetyluridine or 2 0 ,3 0 ,5 0 -tri-O-hexanoyluridine as the sole source of uridine is used to select the clones exhibiting acylesterase activity (left column). The positive hits complement the uridine auxotrophy of the E. coli DH10B DpyrFEC::Km strain by hydrolysis of the substrate (compound 1 or compound 2) to uridine allowing the colony formation. A standard tributyrin agar plate method is represented on the right. identified as G-X-S-Y-X-G ( Fig S2B). Based on the phylogenetic and BLAST analyses, the hits 33T3, BD9, PLA1 and 36T1 belonged to SGNH hydrolase superfamily. The catalytic Ser-His-Asp (Glu) triad (Polg ar, 2005) was determined in the amino acid sequences of esterases 33T3, BD9, PLA1, but no such motif was found in 36T1 ( Fig. S2C). Two esterases 12T and SVG1 were similar to b-lactamases and had the conserved S-X-X-S (Wagner et al., 2009) and LLXHXXG motifs (Ranjan et al., 2005) of Esterase VIII, but two other highly conserved b-lactamase motifs (Y-A-N) and (K-T/S-G) (Joris et al., 1988) were not found (Fig. S2D). Signal peptide sequence analysis showed the presence of signal peptides in 17 out of 30 selected esterases (Table S3). Various ABhydrolases, SGNH hydrolases and b-lactamases were previously isolated from the metagenomic libraries using tributyrin or other method of functional screening (Popovic et al., 2017). However, in addition to the novel variants of known groups of hydrolases, the method presented here also allowed the selection of novel scaffolds, such as C233, 45T3 and 1315H. The protein C233 belonged to PF00933 (Glyco_hydro_3) family of glycoside hydrolases (Bourne and Henrissat, 2001) with unknown esterolytic activity. The hydrolase 45T3 belonged to the family PF00583 and was similar to a ribosomal-protein-alanine N-acetyltransferase (Yoshikawa et al., 1987) Hu et al., 2010Jones and O'Connor, 2011). The most interesting hit 1315H encoded a protein with a DUF998 domain. The protein sequence analysis using SMART (http://smart.e mbl-heidelberg.de/) revealed that the enzyme 1315H may be a transmembrane protein without any predictable function.
Thus, a functional selection of metagenomic libraries revealed enzymes from very diverse protein families, including several hits that, based on bioinformatics, could not be annotated as hydrolases active towards esters.
To confirm that the hits encoded the enzymes with esterolytic activity, the selected genes were PCR-amplified, and the resulting fragments were ligated into pET21a or pLATE31 expression vectors. E. coli strain BL21 (DE3) was transformed with the recombinant plasmids and used for the expression of the recombinant proteins. In total, 27 recombinant proteins were purified by Ni-NTA chromatography (Fig. S3), and 23 of them showed purity higher than 90% (Table 2). Due to hydrophobic nature, the protein encoded by the clone 1315H was not purified, and an insoluble fraction of the cells was used for the determination of activity. The proteins encoded by clones RIEB and 4H1T were not purified due to a poor expression. The purity of the protein MO4B was not analysed because the concentration of the purified enzyme was very low due its poor solubility. During overexpression, sufficient amounts of five proteins 24T1, 24T3, 33T1, SVG1 and SVG3 were released into the extracellular space, and the enzymes were purified from the medium, most likely, without the signal peptides.
The hydrolytic activity of the purified proteins was analysed with various p-nitrophenyl (pNP) esters: acetate, butyrate, valerate, decanoate, palmitate and stearate. All hydrolases were active towards the shortchain esters pNP-acetate and pNP-butyrate, most of them used pNP-valerate, but only a few of them hydrolyzed pNP-decanoate (Table 2). Neither pNPpalmitate or stearate was recognized as the substrate (data not show).
The esterase C233 exhibited activity towards shortchain pNP esters but did not hydrolyze the established substrates of glycoside hydrolases, such as 4-nitrophenyl derivatives of a-L-arabinofuranoside, a-and b-L-arabinopyranoside, a-and b-D-xylopyranoside or b-Dglucopyranoside.
The recombinant enzymes selected using compoud 1 were able to hydrolyze short-chain esters of pNP with a high efficiency, but displayed a very weak activity against pNP-decanoate. Only five enzymes from the 22 selected hydrolyzed pNP-decanoate, although three of them (24T5, 21T1 and 33T1) exhibited a very weak activity against this ester. Six enzymes MO101T, SVG3, GRU1, PLA1, 3T and 33T3 exhibited a strong preference for pNP-acetate as the substrate ( Table 2).
The enzymes selected on compound 2 demonstrated the activity towards the longer-chain esters. These esterases hydrolyzed pNP-decanoate with a high efficiency (Table 2), however the highest specific activity in this group of hydrolases was towards pNP-butyrate (EN3H, K3H2, CAP3H) or pNP-valerate (EN1H, 1315H) ( Table 2). These results indicated that the bulkiness of the ester group of the substrate used for the selection, at least partly, determined the specificity of the selected enzymes. These findings may be useful for the development of a more effective direct selection of the enzymes with the desired properties.
Further analysis of the substrate specificity of the selected enzymes showed that approximately half of the tested esterases could accept the bulky peracetylated carbohydrates, and thirteen esterases could hydrolyze tributyrin (Fig. 4) that was confirmed by the cultivation of those clones on tributyrin agar (Fig. S4). Such results definitely showed that only the fraction of hits would be screened using a standard tributyrin agar approach.
To test if the standard selection on agar plates with tributyrin would result in clones exhibiting activity towards compoud 1 or compound 2, several metagenomics libraries were assessed, and two hits (Tb7_1T and Tb10_7T) forming the halos indicative of the hydrolysis were screened. Tb7_1T (MH423281) encoded the group b-lactamases protein, and Tb10_7T (MH423280) was most similar to ABhydrolases. Neither of the two clones was capable of growing on compound 1 or 2 as a uridine source. Tb7-1T was not further analysed because of its insolubility. The purified protein Tb10_7 was not only active towards pNP esters (Table 2), but also hydrolyzed 3 0 -O-acetylated nucleosides in vitro  . 4). Moreover, different activity profiles of identified esterases towards monoacylated nucleosides harbouring short, aromatic and bulky aliphatic groups (Fig. 4) suggest that a more selective system may be designed and applied, for example, for a substrate-guided isolation and tailoring of the desired mutants. To determine enantioselectivity of the esterases, R/S-1-phenylethyl esters with varied length of the ester groups were synthesized and applied for analysis (Fig. 4). The enzymes 24T1 and 30T1 were enantioselective towards R-1-phenylethyl acetate. Four (30T2, SVG1, 1315H, Tb10-7T) and two (SVG1, Tb10_7T) esterases preferred S-1-phenylethyl hexanoate and S-1-phenylethyl benzoate respectively. No esterase was enantioselective towards R-1-phenylethyl hexanoate or benzoate. Furthermore, to analyse the promiscuity of the selected esterases, several amides were tested as substrates. Consequently, three esterases were found to hydrolyze acetanilide, and 18 were active towards nitrocefin (Fig. 4).
In conclusion, a combination of uridine esters (compound 1 and 2) and the uridine auxotroph mutant Dpyr-FEC of E. coli allows for a functional selection of novel ester hydrolases from metagenomic libraries. Evidently, the selection method presented here is highly complementary to traditional approaches and is applicable for allowing the discovery of novel esterases with different structural and catalytic characteristics. Compared with the known methods, the proposed selection system has many advantages: (i) it is a HTS method that allows for a rapid (1-4 days) processing of large (meta)genome libraries (≥ 10 6 of clones can be analysed on a single agar plate) with low number of false positives, ii) depending on the selection media used, it permits the functional selection of hydrolases belonging to different protein families, (iii) it is flexible since, by using different substrates, for example, esters of the fatty acids of different lengths or bulky carboxylic acids, it allows the identification of enzymes with desired properties, (iv) it would be suitable for the selection of regioselective esterases if the appropriate substrate was used, (v) it allows the identification of novel proteins that previously have not been known as the ester-hydrolysing enzymes or cannot be screened using, for example, a tributyrin method, (vi) the appropriate uridine auxotrophs are available for a wide spectra of microbial hosts, including extremophiles and yeasts or can be easily constructed, for example, using 5-fluororotic acid selection (Boeke et al., 1987;Kondo et al., 1991) and/or recombineering (Xu et al., 2015;Aparicio et al., 2018) hence offering a possibility of a functional screening of enzymes under distinct conditions. Definitely, the desired host cannot be capable to hydrolyze the chosen uridine ester. Further applications

Environmental samples, DNA extraction and construction of the metagenomics libraries
Metagenomic libraries were constructed from soil and sediment samples using a pUC19 vector (Table S1). For DNA isolation directly from soil or sediment 'ZR Soil Microbe DNA MidiPrep TM ' was used. The total DNA was partially digested with restriction endonuclease selected from the list: PstI, HindIII or BamHI. The different restriction endonuclease were used to obtain a larger diversity of the DNA fragments which were inserted into the pUC19 vector and used to transform E. coli DH5alpha competent cells by electroporation as described (Stanislauskien _ e et al., 2018). To analyse the number of clones in the library, quality of the library (a ratio of white/blue colonies) and the average insert length, an aliquot of the transformed cells was diluted and spread on LB agar (10 g l À1 tryptone, 5 g l À1 yeast extract, 10 g l À1 NaCl and 15 g l À1 agar) supplemented with the appropriate antibiotics (40 lg ml À1 kanamycin, 100 lg ml À1 ampicillin), 1 mM IPTG and 1 mM X-gal. Ten to twenty of individual white colonies-forming clones were randomly chosen for plasmid DNA isolation by Thermo Scientific TM GeneJET Plasmid Miniprep Kit (Thermo Fisher Scientific) and analysis of the length of the insert. The remaining undiluted mixture of the cells was spread on LB agar (10 ll of the bacterial suspension per one Petri dish (92 mm)) supplemented with 40 lg ml À1 kanamycin and 100 lg ml À1 ampicillin. After cultivation for 14-16 h, a biomass was scraped from agar, and total plasmid DNA was isolated by Thermo Scientific TM GeneJET Plasmid Midiprep Kit (Thermo Fisher Scientific). The obtained DNA mixture (a metagenomics library) was stored at -20°C and used for a further transformation of the E. coli DH10B DpyrFEC::Km cells.
Screening of esterases by using the tributyrinsupplemented agar plates. LB agar medium supplemented with 1% tributyrin was used to screen for lipolytic/esterolytic activity. The clones showing a halo around the individual colonies, which indicated hydrolysis of the tributyrin, were selected on the emulsified tributyrin medium after 1-2 days of growth at 37°C. (Popovic et al., 2017) Peña-Garc ıa et al., 2016Ranjan et al., 2005) DNA sequencing and gene annotation Nucleotide sequences were determined at Macrogen Europe (Netherlands) and using the following sequencing primers: M13F-pUC (5 0 -GTTTTCCCAGTCACGAC-3 0 ), M13R-pUC (5 0 -CAGGAAACAGCTATGAC-3 0 ), T7 Promoter (5 0 -TAATACGACTCACTATAGGG-3 0 ), T7 terminator (5 0 -TAATACGACTCACTATAGGG-3 0 ) or LIC Reverse Sequencing primer, 24-mer (5 0 -GAGCGGATAA-CAATTTCACACAGG-3 0 ). Some individual clones contained more than one ORF in each DNA fragment. ORFs were analysed by using the Unipro UGENE program, and homology search was conducted using the Blast server (http://www.ncbi.nlm.nih.gov/BLAST). For further analysis the ORFs encoding putative hydrolases were chosen. When homology search did not predict a hydrolase, the deletion analysis to obtain the truncated variants of the plasmid and the functional reselection on the appropriate substrate was carried out. To confirm that the hits encoded the enzymes with an esterolytic activity, the selected genes were PCR-amplified, and the resulting fragments were ligated into pET21a or pLATE31 expression vectors. Phylogenetic analysis was conducted using the Maximum Likelihood Tree routine of MEGA 7 software. (Kumar et al., 2016;Thompson et al., 1994) The alignment was performed using CLUSTALW in MEGA 7.

Expression vectors and PCR primers
The genes of the selected enzymes were amplified with Phusion DNA polymerase using primers (Table S2). Metagenomic esterase-encoding genes of 12T, 24T5, 45T3, 33T1, and 3T clones were ligated into pET21a, and all other genes were ligated into pLATE31. E. coli cells transformed with recombinant plasmids were cultivated on LB agar supplemented with 100 lg ml ampicillin.

Overexpression and purification of esterases
The recombinant proteins were overexpressed in E. coli strain BL21 (DE3). E. coli cells were grown in 20 ml BHI (Brain-Heart-Infusion Broth) medium containing ampicillin (100 mg ml) at 37°C with aeration. Protein expression was induced by adding 0.5 lM IPTG at 0.6-1 OD 600 , and cells were grown for a further 4-18 h at 30°C. Cells expressing 12T and 3T were grown at 23°C after induction to increase protein solubility. Wet cell biomass from 20 ml culture broth were suspended in 5-10 ml of buffer A (50 mM potassium phosphate, pH 7.5) and disrupted by sonication for 2.5 min. A lysate was cleared by centrifugation at 15 000 9 g for 4 min. Cleared lysate was applied to 0.2 ml HisPur TM Ni-NTA spin column (equilibrated with the buffer A). The column was washed with the buffer A, and the proteins were eluted with buffer A containing 300 mM imidazole. The active fractions were combined and dialyzed against the buffer B (50 mM potassium phosphate, pH 7.5), at 25°C. All the purification procedures were performed at room temperature.

Proteins concentration and purity determination
The concentration of protein was determined using Pierce TM Coomassie Plus (Bradford) Assay Reagent by Standard Microplate Protocol. Proteins were analysed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE, 14% separating and 4.0% stacking) according to Laemmli. Gels were developed in Coomassie Brilliant Blue G-250 dye, scanned in 16 bit format and quantified by GelAnalyser program. Each sample contained 2 lg of total protein. Quantities of impurities and target proteins were estimated using calibration curve generated from known amounts of BSA: 0.125, 0.25 and 0.5 lg per band. The purity of analysed protein was calculated as the ratio between quantity of target protein and quantity of all proteins.

Enzymatic activity of esterases
Hydrolysis of pNp esters. The activity of esterase were assayed by incubating the enzyme with 1 mM pNPsubstrate (from 10 mM stock in DMSO) in 50 mM potassium phosphate, pH 7.5, buffer at a 37°C for 10 min in 100 ll reaction volume. 5-300 ng of protein, depending on the enzyme specificity for the substrate, were present into the reaction mixture. The absorbtion of the reaction mixture at 405 nm was measured against enzyme-free blank to compensate for the substrate auto-hydrolysis (Hern andez-Garc ıa et al., 2017;Beisson et al., 2000). One unit is defined as the amount of enzyme that catalyses the formation of 1 lmol of 4-nitrophenol (molar extinction coefficient e = 12.3 M À1 cm À1 ) per minute. Enzyme activity was tested against different pNP-acyl esters [acetate, butyrate, valerate, decanoate, palmitate and stearate].
Hydrolysis of nitrocefin. Nitrocefin is a chromogenic cephalosporin substrate routinely used to detect the presence of beta-lactamase enzymes (Petersen et al., 2001;Ohlhoff et al., 2015;Chow et al., 2013). Once hydrolyzed, the degraded nitrocefin compound rapidly changes colour from yellow to red. A hydrolytic activity was assayed in 50 mM potassium phosphate buffer, pH 7.0, containimg 1 mM nitrocefin at 37°C for 2 h. Total reaction volume was 50 ll. The change of the colour was evaluated visually.
Hydrolysis products test by thin-layer chromatography (TLC) method. A hydrolytic activity was assayed in reaction mixture containing 45 mM potassium phosphate buffer, pH 7.5, 1 ll enzyme (0.1-4.6 lg/reaction) and 10 mM substrate: b-D-glucose pentaacetate, b-Dgalactose pentaacetate (from 100 mM stock in acetone), 3 0 -O-benzoyl-2 0 deoxyuridine (10 mg ml À1 stock in DMF), 3 0 -O-levulinyl-N 4 -benzoyl-2 0 deocytidine and 5 0 -Olevulinyl-N 4 -benzoyl-2 0 -deocytidine (from 100 mM stock in DMSO). The total reaction mix volume was 20 ll. Reaction mixture was incubated at 30°C temperature up to 3 h. Thin-layer chromatography (TLC) was conducted on the Merck silica gel 60F254 plates, using the dichloromethane and methanol (9:1) mixture of solvents. b-D-glucose pentaacetate and b-D-galactose pentaacetate were visualized by anisaldehide stain (50 ml ethanol, 1.9 ml of concentrated sulfuric acid, 0.54 ml of glacial acetic acid and 0.14 ml of p-anisaldehyde). The plate was developed by heating on a hot plate. Synthetic nucleosides were exposing to UV light. Table S1. Metagenomic libraries used in this work. Table S2. Primers used for amplification of genes of selected hydrolases. Table S3. Predicted signal peptide sequences identified in the selected esterases by bioinformatics. Fig. S1. Selected clones on MD medium, MD+ compound 1, MD+compound 2 and MD+uridine after 2 days of incubation at 37°C. Fig. S2. The alignment was performed by ClustalW software. Fig. S3. Analysis of the purified esterases by SDS-PAGE. Fig. S4. LB medium with tributyrin plates after 2 days of incubation at 37°C. Scheme S1. Synthesis of optically active esters (3-8).