Screening and Characterization of Novel Polyesterases from Environmental Metagenomes with High Hydrolytic Activity against Synthetic Polyesters

Abstract

The continuous growth of global plastics production, including polyesters, has resulted in increasing plastic pollution and subsequent negative environmental impacts. Therefore, enzyme-catalyzed depolymerization of synthetic polyesters as a plastics recycling approach has become a focus of research. In this study, we screened over 200 purified uncharacterized hydrolases from environmental metagenomes and sequenced microbial genomes and identified at least 10 proteins with high hydrolytic activity against synthetic polyesters. These include the metagenomic esterases MGS0156 and GEN0105, which hydrolyzed polylactic acid (PLA), polycaprolactone, as well as bis-(benzoyloxyethyl)-terephthalate. With solid PLA as a substrate, both enzymes produced a mixture of lactic acid monomers, dimers, and higher oligomers as products. The crystal structure of MGS0156 was determined at 1.95 Å resolution and revealed a modified α/β hydrolase fold, with a lid domain and highly hydrophobic active site. Mutational studies of MGS0156 identified the residues critical for hydrolytic activity against both polyester and monoester substrates, with two-times higher polyesterase activity in the MGS0156 L169A mutant protein. Thus, our work identified novel, highly active polyesterases in environmental metagenomes and provided molecular insights into their activity, thereby augmenting our understanding of enzymatic polyester hydrolysis.

Graphical Abstract:

INTRODUCTION

Over the last 50 years, a tremendous increase in production of synthetic polymers and their persistence in the environment resulted in elevated levels of pollution.^1–4 Most petroleum-based plastics (polyethylene, polypropylene, polyethyleneterephthalate) are remarkably stable in the environment, resulting in the accumulation of plastic waste and microplastic particles, which negatively affect marine ecosystems.^4–6 The development of biodegradable plastics represents part of a solution that includes other approaches such as plastic recycling and polymer recovery.^7,8 The most sustainable option for plastic waste treatment appears to be a closed-loop recycling process based on physical, chemical, and biocatalytic depolymerization and the recovery of chemical feedstocks for the synthesis of novel polymers (a circular economy).^9–11 Compared to physical, thermal, and chemical plastics depolymerization, biocatalytic recycling has several advantages, including low energy consumption, mild reaction conditions, and the possibility for stereospecific degradation and enzymatic repolymerization.^9,12

Most conventional polymers (polyethylene, polypropylene, polystyrene, poly(vinyl chloride), and poly(ethylene terephthalate)) exhibit limited or no biodegradability.² The group of biodegradable plastics with commercial relevance is dominated by aliphatic polyesters such as polylactic acid (PLA) and polycaprolactone (PCL), whereas the aromatic–aliphatic copolyester poly(ethylene terephthalate) (PET) is more resistant to microbial or enzymatic attack.^13,14 The complex process of polymer biodegradation in the environment is a combination of various abiotic and biotic factors.^15,16 In nature, different groups of anaerobic (Clostridium and Propionispora) and aerobic bacteria and fungi (including actinomycetes (Amycolatopsis, Lentzea, Pseudonocardia), Bacillaceae (Geobacillus, Brevibacillus, Paenibacillus), Pseudomonas, and Aspergillus)have been found to degrade polyesters.^17,18 Enzymatic hydrolysis of polyesters was demonstrated over 40 years ago using purified lipases and proteases from different fungi and Achromobacter sp., as well as hog liver esterase.^13,19 These enzymes belong to a large group of serine-dependent α/β hydrolases, which use the catalytic triad Ser-His-Asp (or Glu) for the hydrolysis of different monomeric and polymeric substrates. Furthermore, polyester degrading (polyesterase) activity was later demonstrated in cutinases (also serine-dependent α/β hydrolases) from fungal and bacterial plant pathogens, which secrete these enzymes to degrade the plant polyester cutin.²⁰ Cutinases are esterase-like enzymes, hydrolyzing not only cutin, but also water-soluble monoesters and synthetic polyesters, including PCL and PET.²¹

A number of polyester degrading esterases and lipases have been characterized biochemically, including Paenibacillus amylolyticus PlaA, Thermobifida fusca TfH, ABO1197 and ABO1251 from Alcanivorax borkumensis, several clostridial esterases ($\mathrm{C}\mathrm{h}\mathrm{a}\mathrm{t}\stackrel{-}{\mathrm{h}}\mathrm{E}\mathrm{s}\mathrm{t}1$, $\mathrm{C}\mathrm{b}\mathrm{o}\mathrm{t}\stackrel{-}{\mathrm{c}}\mathrm{E}\mathrm{s}\mathrm{t}\mathrm{A}$, $\mathrm{C}\mathrm{b}\mathrm{o}\mathrm{t}\stackrel{-}{\mathrm{c}}\mathrm{E}\mathrm{s}\mathrm{t}\mathrm{B}$), and the metagenomic polyesterases PlaM4, EstB3, and EstC7.^22–27 These enzymes hydrolyzed a broad range of both emulsified and solid polyester substrates, such as PLA, PCL, PET, poly(ethylene adipate), and poly(butylene adipate-co-butylene terephthalate) (PBAT).^27–30 Enzymatic activity of purified polyesterases was also characterized using the soluble monoesters p-nitrophenyl butyrate and p-nitrophenyl acetate, which are common substrates for carboxylesterases.^24,27,30 Moreover, several purified cutinases from bacteria (Thermobifida), fungi (Humicola, Aspegillus, Fusarium), and environmental metagenomes have been shown to hydrolyze synthetic polyesters including PET and polyester polyurethane.^{14,20,31–33} These works also revealed several important factors limiting the biodegradation of synthetic polyesters, including the hydrophobicity, crystallinity, surface topography, and molecular size of synthetic polymers.^2,3,34,35 It has been shown that the hydrophobic polymer surface restricts effective adsorption and activity of polymer-degrading enzymes.^36–38 To improve protein sorption and thereby polymer hydrolysis, the Thermobifida cellulosilytica cutinase $\mathrm{T}\mathrm{h}\stackrel{-}{\mathrm{c}}\mathrm{C}\mathrm{u}\mathrm{t}1$ was fused to noncatalytic substrate-binding modules from the Hypocrea jecorina cellobiohydrolase I or Alcaligenes faecalis polyhydroxyalkanoate depolymerase.³⁶ Both fusion enzymes demonstrated increased adsorption and hydrolytic activity on PET, likely due to enhanced hydrophobic interactions with the substrate.

Crystal structures have been determined for several polyesterases, including several carboxyl esterases (Clostridium hathewayi $\mathrm{C}\mathrm{h}\mathrm{a}\mathrm{t}\stackrel{-}{\mathrm{h}}\mathrm{E}\mathrm{s}\mathrm{t}1$, C. botulinum $\mathrm{C}\mathrm{b}\mathrm{o}\mathrm{t}\stackrel{-}{\mathrm{c}}\mathrm{E}\mathrm{s}\mathrm{t}\mathrm{A}$, and Rhodopseudomonas palustris RPA1511) and cutinases (Thermomyces (formerly Humicola) insolens HiC, Thermobifida alba Est119, and metagenomic LC-cutinase from leaf-branch compost).^{25,30,39–42} Polyesterase structures revealed the classical α/β hydrolase core domain with or without an α-helical lid (or cap) domain covering the active site, with the catalytic Ser residue positioned at the bottom. The polyesterase active site usually represents a wide-open cleft, directly accessible to polymeric substrates as revealed by the structures of RPA1511 and Est119 in complex with polyethylene glycol bound close to the catalytic triad.^41,42 Interestingly, when the active site cleft of the Ideonella sakaiensis PETase (PET-degrading enzyme) was narrowed by site-directed mutagenesis (to make it more cutinase-like), the resulting engineered PETase outperformed the wild-type protein in degradation of both PET and polyethylene-2,5-furandicarboxylate.⁴³ In addition, mutagenesis and protein engineering experiments with the Thermobifida cellulosilytica cutinases $\mathrm{T}\mathrm{h}\stackrel{-}{\mathrm{c}}\mathrm{C}\mathrm{u}\mathrm{t}1$ and $\mathrm{T}\mathrm{h}\stackrel{-}{\mathrm{c}}\mathrm{C}\mathrm{u}\mathrm{t}2$ confirmed an important role of enzyme surface and hydrophobic interactions for polyester hydrolysis.^36,44 In addition, the recent work by Zumstein et al. suggested that polyesterase activity of different carboxylesterases depends on the accessibility of their active sites.⁴⁵

Although recent studies have identified a number of polyester degrading enzymes, the continuously growing global demand for plastics and new polymers has also stimulated interest in novel enzymes and biocatalytic approaches for polymer recycling technologies. The discovery of novel polymer degrading enzymes, engineering of more active enzyme variants, as well as understanding of the molecular mechanisms of these enzymes represent the key challenges for the development of biocatalytic strategies for polymer hydrolysis and synthesis.³ To address these challenges, we have identified over 20 novel polyesterases using enzymatic screening, and biochemically characterized MGS0156 and GEN0105, which showed high hydrolytic activity against a broad range of polyesters (PLA, PCL, polyethylene succinate (PES), poly(butylene succinate-co-adipate) (PBSA), and 3PET). The crystal structure of MGS0156 was solved, revealing an open active site with hydrophobic surface, whereas structure-based mutagenesis studies identified amino acid residues critical for enzymatic activity, with the L169A mutant protein displaying two-times higher polyesterase activity.

MATERIALS AND METHODS

Reagents.

All chemicals and substrates used in this study were of analytical grade unless otherwise stated. Polymeric substrates were purchased from Sigma-Aldrich (St. Louis, MO) except poly(D,l-lactide) PLA2 (M_w 0.2 × 10⁴), PLA70 (M_w 7.0 × 10⁴), as well as poly(l-lactide) PLLA40 (M_w 4.0 × 10⁴), that were obtained from PolySciTech (Akina Inc., West Lafayette, IN). Commercial-grade PLA polymers (Ingeo 4032D, and Ingeo 6400D) were products of NatureWorks LLC (NE, USA), poly(d-lactide) PURASORB PD 24 of Corbion Purac (Amsterdam, The Netherlands), whereas polybutylene succinate (PBS) (Bionolle 1001MD, and Bionolle 1020MD) and polybutylene succinate-co-adipate (PBSA) (Bionolle 3001MD, and Bionolle 3020MD) were purchased from Showa Denko K.K., Japan. The PET model substrate, bis(benzoyloxyethyl) terephthalate (3PET), was synthesized by CanSyn (Toronto, ON, Canada) using a previously reported protocol.⁴⁶ The surfactant Plysurf A210G was obtained from Dai-ichi Kogyo Seiyaku Co. (Tokyo, Japan) and used to emulsify the polymers.

Gene Cloning, Protein Purification, and Mutagenesis.

The coding sequences of selected hydrolase genes were PCR amplified using the genomic DNA of the host organism or the gene constructs from metagenomic libraries as the templates. The PCR products were cloned using a ligation-independent protocol into a modified pET15b (Novagen) expression vector containing an N-terminal 6His tag. The plasmids were transformed into Escherichia coli BL21 (DE3) Codon-Plus strain (Stratagene) as the expression host. Transformants were grown in shake flasks of Terrific Broth (1 L) at 37 °C to the optical density A₆₀₀ ≈ 0.6 followed by an induction with 0.5 mM (final concentration) isopropyl 1-thio-β-D-galactopyranoside (IPTG) and an overnight incubation at 16 °C. The cells were harvested by centrifugation, resuspended in binding buffer (50 mM HEPES pH 7.5, 0.25 M NaCl, 5 mM imidazole and glycerol 5% vol.) and disrupted by sonication followed by another round of centrifugation to remove cell debris. Recombinant proteins were purified to homogeneity (>95%) as described previously,⁴⁷ using metal-chelate affinity chromatography on Ni-NTA Superflow (Ni²⁺-nitrilotriacetate; Qiagen) resin as well as ion exchange chromatography on a Mono Q GL 10/100 column (GE Healthcare) or size exclusion chromatography on a Superdex 200 16/60 column (Amersham Bio- sciences) equilibrated with 10 mM HEPES (pH 7.5), 0.25 M NaCl and 1 mM TCEP [tris(2-carboxyethyl)phosphine] using ÄKTA FPLC (Amersham Biosciences) where necessary.⁴⁸ Site-directed mutagenesis of metagenomics esterases was performed using a QuickChange kit (Stratagene) according to the manufacturer’s protocol. Wild-type MGS0156 and GEN0105 were used as the templates, and mutations were verified via DNA sequencing. The selected residues (Pro167, Leu169, Leu170, Glu172, Cys173, Val174, Ser175, Leu179, Leu197, Arg199, His231, Ser232, Lys233, Ser265, Phe271, Arg277, Glu280, Cys287, Leu296, Leu299, Glu330, Leu335, Phe338, Asp350, Leu352, Val353, Asp372, His373, Met378, Phe380 for MGS0156 and Ser168, Glu262, His292 for GEN0105) were mutated to Ala or Gly (for insoluble Ala mutant proteins). The amino acid numbering is based on the full-length protein. Mutant proteins were overexpressed and purified in the same manner as described for the wild-type proteins. Multiple sequence alignment was conducted by Clustal Omega v1.2.1 through EMBL-EBI server, whereas phylogenetic analysis was performed by MEGA v7.0 using the neighbor-joining method.^49,50

Esterase Assays with Soluble Substrates.

Carboxylesterase activity was measured spectrophotometrically as described previously.⁴⁷ Purified enzymes (0.05–10.0 μg protein/reaction) were assayed against α-naphthyl or p-nitrophenyl (pNP) esters of different saturated fatty acids (C2–C16; 0.25–2.0 mM) as substrates in a reaction mixture containing 50 mM HEPES-K buffer (pH 8.0).⁴⁷ The stock solutions of α-naphthyl (100 mM) and p-nitrophenyl (50 mM) ester substrates were prepared in acetone and isopropanol, respectively. The final concentration of substrates was maintained lower than their water solubility to yield a homogeneous solution. Reaction mixtures (200 μL, in triplicate) were incubated at 30 °C in a 96-well plate format. Enzyme kinetics were determined by substrate saturation curve fitting (nonlinear regression) using GraphPad Prism software (version 7.0 for Mac, GraphPad Software, CA, USA).

Polyester Degradation (Polyesterase) Screens.

Emulsified polyester substrates were prepared in 50 mM Tris-HCl buffer (pH 8.0) as described previously.^41,51 Polyester emulsions were solidified with agarose (1.5%, w/v) and poured into 150 mm cell culture dishes with 20 mm grids to make a uniformly opaque gel. Cylindrical wells (3 mm diameter) were aseptically punched in the assay plate and inoculated accordingly with purified enzyme solutions (20 μL containing 1.5 nmol of protein/well). Sealed assay plates were incubated at 30 °C and monitored for 3 weeks. The presence of polyesterase activity was inferred from the formation of a translucent halo around the wells with purified proteins.^41,51 To compare the size of clear halos formed by different enzymes, image analysis was conducted using the histogram tool on Adobe Photoshop software and pixel counts were plotted on the graph with error bars representing measurements with varying pixel tolerance values.

Analysis of the Reaction Products of Solid PLA Depolymerization.

Purified enzymes (50 μg) and enzyme-free controls were incubated with PLA10 powder (10–12 mg, Resomer R 202 Hpoly(d,l-lactide); M_w 10–18K, melting point 58 °C) in a reaction mixture (1 mL) containing 0.4 M Tris-HCl buffer (pH 8.0) for 18 h at 30 °C with shaking. Supernatant fractions were collected at different time points, passed through centrifugal filters (MWCO 10 kDa), and the produced lactic acid was measured using a lactate dehydrogenase (LDH) assay which enabled the detection of both d- and l-enantiomers of lactic acid with high sensitivity.^41,52 Values for detected lactate in enzyme-free controls were subtracted from the reported results for all enzyme-containing reactions. The l-lactate dehydrogenase (PfLDH) from Plasmodium falciparum⁵³ and the D-lactate dehydrogenase (D-LDH3) from Lactobacillus jensenii⁵⁴ were heterologously expressed in E. coli and affinity purified to near homogeneity as described earlier. Both LDH enzymes were added to the reaction mixture in excess (total 500 μg/mL, 50/50) to maintain the reaction rate in the first order with lactate concentration. For the analysis of oligomeric PLA products in supernatant fractions (passed through 10 kDa filters), the flow-through aliquots (90 μL) were treated for 5 min at 95 °C with 1 M NaOH (final concentration) to convert oligomeric PLA products to lactic acid monomers prior to LDH assay (the data were corrected for the presence of monomeric lactic acid before alkaline treatment). In addition, the filtered supernatant fractions from solid PLA reactions were analyzed using reverse phase liquid chromatography⁵⁵ coupled with mass spectrometry (LC-MS) to identify the water-soluble products of PLA hydrolysis. The platform configuration and methodology were as described previously.⁴¹

Protein Crystallization and Crystal Structure Determination of MGS0156.

Purified MGS0156 (75–421 aa) was crystallized at room temperature using the sitting drop vapor diffusion method by mixing 1 μL of the selenomethionine substituted protein (12 mg/mL) with 1 μL of crystallization solution containing 30% (w/v) PEG 4k, 0.2 M ammonium acetate, 0.1 M sodium citrate (pH 5.6), and 1/70 chymotrypsin. Crystals were harvested using mounted cryo-loops and transferred into the cryo-protectant (Paratone-N) prior to flash-freezing in liquid nitrogen. Data collections were carried out at the beamlines 19-ID of the Structural Biology Center, Advanced Photon Source, Argonne National Laboratory.⁵⁶ The data set was collected from a single crystal to 1.95 Å at the wavelength of 0.9794 Å and processed using the program HKL3000⁵⁷ (Supporting Information (SI) Table S1). The structure of MGS0156 was determined by the Semethionine SAD phasing, density modification, and initial model building as implemented in the PHENIX suite of programs.⁵⁸ The initial models (~90% complete) were further built manually using the program COOT⁵⁹ and refined with PHENIX. Analysis and validation of structures were performed using MOLPROBITY⁶⁰ and COOT validation tools. The final model was refined to R_work/R_free = 0.1532/0.19, and it shows good geometry with no outliers in the Ramachandran plot. Data collection and refinement statistics are summarized in SI Table S1. Surface electrostatic charge analysis was performed using the APBS tool in Pymol on a model generated by the PDB 2PQR server.^61,62 The topology diagram of MGS0156 was generated by HERA program⁶³ through PDBsum server.⁶⁴ The atomic coordinates have been deposited in the Protein Data Bank, with accession code 5D8M.

RESULTS AND DISCUSSION

Screening of Purified Microbial Hydrolases for Polyesterase Activity.

To discover novel polyesterases, 213 randomly selected purified uncharacterized hydrolases from sequenced microbial genomes and environmental metagenomes (SI Table S2, also ref 65) were screened for hydrolytic activity against emulsified PLA10 [poly(Dl-lactide); M_w 10K], PLLA40 [poly(l-lactide); M_w 40K], polycaprolactone PCL10 (M_w 10K), and bis(benzoyloxyethyl) terephthalate (3PET) using agarose-based screens.^24,65 The experimental conditions used in these screens (30 °C, pH 8.0) were within the activity range for most known polyesterases from different organisms (cold-adapted, mesophilic, and thermophilic). These screens revealed the presence of detectable polyesterase activity in 36 proteins, mostly from the α/β hydrolase superfamily (SI Table S3). Most of these proteins were active against poly(d,l-lactide) (22 proteins), followed by 3PET (13 proteins) and PCL (11 proteins), whereas nine proteins exhibited activity toward poly(l-lactide) (PLLA40). At least 10 identified polyesterases exhibited high activity against multiple substrates (e.g., MGS0156 and GEN0105) (SI Table S3). Thus, a significant number of microbial and metagenomic hydrolases exhibit hydrolytic activity against synthetic polyesters.

Since the metagenomic polyesterases MGS0156 (GenBank AKJ87264) and GEN0105 (GenBank AKJ87216) showed high hydrolytic activity against several polyesters (PLA10, PCL10, and 3PET), the present work was focused on the biochemical and structural characterization of these proteins, which were also compared with the recently published metagenomics polyesterases GEN0160 and MGS0084⁶⁵ (Figure 1).

Figure 1.

Polyesterase activity of purified metagenomic carboxylesterases. Agarose-based screen of purified proteins for the presence of polyesterase activity against emulsified PCL10. The presence of polyesterase activity is indicated by the formation of a clear zone around the wells (A) with purified proteins (50 μg of protein/well, 72 h at 30 °C). Agarose (1.5%) plates contained 0.2% emulsified PCL10 in 50 mM Tris-HCl (pH 8.0) buffer. Previously characterized enzymes (metagenomic polyesterases PlaM4,²⁶ GEN0160,⁶⁵ MGS0084,⁶⁵ and porcine liver esterase (PLE)) were used as controls. (B) The bar graph represents the polyesterase activity of tested enzymes as clear halo areas in pixels obtained via image scanning.

Carboxylesterase activity of the selected enzymes was initially identified using tributyrin-based esterase screens of the metagenomic gene libraries from an anaerobic urban waste degrading facility (GEN0105) or paper mill waste degrading microbial community (MGS0156).⁶⁵ The MGS0156 gene encodes a protein comprised of 421 amino acids with a potential N-terminal signal peptide (1–75 aa), whereas the GEN0105 sequence (322 aa) appears to lack an obvious signal peptide (SI Figure S1). Based on sequence analysis, both MGS0156 and GEN0105 belong to serine dependent α/β hydrolases but share low sequence identity to each other (21.1%). Both enzymes represent metagenomic proteins as GEN0105 shares 61% sequence identity with the predicted esterase $\mathrm{B}0\mathrm{L}3\mathrm{I}\stackrel{-}{1}9\mathrm{B}\mathrm{A}\mathrm{C}\mathrm{T}$ from an uncultured bacterium, whereas the closest homologue of MGS0156 ($\mathrm{D}\mathrm{e}\mathrm{s}\mathrm{f}\mathrm{r}\mathrm{D}\mathrm{R}\mathrm{A}\mathrm{F}\stackrel{-}{\mathrm{T}}2296$ from Desulfovibrio fructosivorans) shows 70% sequence identity to this protein (SI Figure S1). When compared to experimentally characterized proteins (Swiss-Prot database), GEN0105 showed the highest similarity to the monoterpene epsilon-lactone hydrolase MlhB from Rhodococcus erythropolis (85% query cover, 35% sequence identity), whereas MGS0156 did not show sequence homology to any characterized hydrolase.⁶⁶ Initially, carboxylesterases and lipases were classified by Arpigny and Jaeger (1999) into eight families based on amino acid sequence homology and biochemical properties, which were later expanded to include novel enzymes (currently over 18 families, some of which containing only one family member).^67–70 Phylogenetic analysis revealed that GEN0105 is associated with esterase family IV, whereas MGS0156, MGS0084, and GEN0160 showed no clustering with known families of lipolytic enzymes, suggesting that these proteins represent new esterase families (Figure 2). Thus, the type II (lipase/cutinase type) polyesterases, including PLA depolymerases, exhibit broad phylogenetic diversity and are associated with esterase families I, III, IV, V as well as with new esterase families.

Figure 2.

Phylogenetic analysis of selected metagenomic polyesterases. Phylogenetic tree of polyesterases showing their relatedness to known esterase families (I –VIII, based on Arpigny and Jaeger, 1999).⁶⁷ The phylogenetic tree was generated by MEGA7 software⁷¹ using the neighbor-joining method. The numbers on the nodes correspond to the percent recovery from 1000 bootstrap resamplings. The evolutionary distances were calculated using the Poisson correction method⁷² and are in the units of the number of amino acid substitutions per site. GenBank accession numbers or Uniprot IDs are shown in parentheses.

Carboxylesterase Activity of MGS0156 and GEN0105 against Soluble Monoester Substrates.

The acyl chain length preferences of purified recombinant MGS0156 (75–421 aa) and GEN0105 were characterized using spectrophotometric assays with α-naphthyl and p-nitrophenyl (pNP) monoesters (Figure 3). For these substrates, MGS0156 was most active against pNP -octanoate and pNP -decanoate (C8–C10 substrates), as well as against pNP-palmitate (C16; Figure 3), which is in line with the lipolytic activity of this protein against olive oil observed in agar-based screens,⁶⁵ indicating that it is a lipase-like enzyme. Compared to MGS0156, the specific activity of GEN0105 was an order of magnitude lower with a preference for shorter substrates (C4 and C5 substrates; Figure 3). With monoester substrates, both enzymes demonstrated saturation kinetics with MGS0156 showing high catalytic efficiencies with low K_m values toward different substrates (k_cat 88.8–1101 s⁻¹, specific activities 170–1700 U/mg) (SI Table S4, Figure 3). Based on our data, both MGS0156 and GEN0105 are among the most active known polyesterases, for which a broad range of specific activities against model monoesters have been reported (4.4–1150 U/mg, k_cat 6–660 s⁻¹).^{26,27,29,30,37,73}

Figure 3.

Esterase activity of metagenomic polyesterases against soluble monoester substrates with different acyl chain length. The reaction mixtures contained 0.5 mM p-nitrophenyl (pNP)- or 1.5 mM α-naphthyl (αN) esters with different chain lengths and 0.01 μg of purified MGS0156 (A) or GEN0105 (B). The white bars show activity against α-naphthyl esters, whereas the gray bars represent activity against pNP-substrates.

Based on temperature profiles of esterase activity, both MGS0156 and GEN0105 are mesophilic carboxylesterases showing maximal activity between 35 and 40 °C and retaining approximately 20% of maximal activity at 5 °C (SI Figure S2). This is similar to mesophilic metagenomic polyesterases MGS0169 and GEN0160, whereas the cold-resistant polyesterase MGS0084 was most active at 20 °C and retained almost 50% of its maximal activity at 5 °C.⁶⁵ While both enzymes showed significant activity at alkaline reactions (pH < 10.5), MGS0156 was more tolerant to high acidity retaining approximately 50% of its optimal activity at pH 2 compared to GEN0105 which was inactive at pH < 4 (SI Figure S2). In addition, MGS0156 and GEN0105 showed similar sensitivity to inhibition by detergents (Triton X-100 and Tween 20), whereas MGS0156 retained higher residual activity (25–75%) in the presence of salts (0.5–2.5 M NaCl or KCl) (SI Figure S2). Thus, with monoester substrates, MGS0156 and GEN0105 exhibit different acyl chain length preferences and salt resistances yet similar sensitivities to temperature and detergents. The biochemical diversity of identified polyesterases presents potential technological advantages for applications in enzymatic degradation of various polyesters under different reaction conditions (temperature, salts, detergents).

Hydrolytic Activity of Metagenomics Polyesterases Against 22 Polyester Substrates.

The polyester substrate ranges of purified MGS0156 and GEN0105 were determined using agarose-based assays with 22 emulsified synthetic polyesters, including PLA, PCL, 3PET as well as copolymers of various composition (Table 1). Since melting temperature (Tm) of the polymers highly affect their enzymatic degradability,⁷⁴ polyester substrates from a range of molecular weight were subjected to enzymatic hydrolysis. Polyesterase activity of these enzymes was compared with the activity of the recently identified metagenomic esterases GEN0160 and MGS0084.⁶⁵ Based on the diameter of clear zones produced in agarose-based screens, the four metagenomic esterases exhibited polyesterase activity against emulsified PCL10, which was higher (GEN0105, GEN0160, and MGS0156) or comparable (MGS0084) to that of the previously identified polyesterase PlaM4 from compost (Figure 1).²⁶ When screened against 22 emulsified polyesters, MGS0156 and GEN0105 degraded 13 and 17 substrates, respectively, including PLA, poly(d,l-lactide-co-glycolide) (PLGA), PCL, PBSA, and 3PET (Table 1). Both enzymes hydrolyzed the majority of the tested PLA polymers, with GEN0105 displaying activity against poly(l-lactide) and neither enzyme displaying activity against poly(d-lactide). Previously, it has been shown that type I (protease) PLA depolymerases are specific toward poly(l-lactide), as opposed to type II (cutinase/lipase) PLA depolymerases, which show preference for poly(d,l-lactide).^17,75 Besides GEN0105, only the cutinase-like type II enzyme CLE from Cryptococcus sp. strain S-2 has been shown to be able to hydrolyze poly(l-lactide).^17,76 PLA substrates with the acid end protected by the addition of an ester group were also hydrolyzed by MGS0156 and GEN0105, suggesting that these polyesterases can exhibit endotype hydrolysis. In contrast, GEN0160 and MGS0084 showed no polyesterase activity against PLA substrates (except for MGS0084 toward PLA2) and 3PET (Table 1). Finally, the four metagenomic esterases showed no hydrolytic activity toward poly(d-lactide), PHB and PBS. Thus, GEN0105 appears to be the most versatile polyesterase from the four tested enzymes, being able to hydrolyze a copolymer of hydroxybutyric acid and hydroxyvaleric acid (PHBV), as well as the commercial polymer Ingeo PLA6400 from NatureWorks (Table 1).

Table 1.

Polyester Substrate Profile of Purified GEN0105, GEN0160, MGS0084, and MGS0156^a

substrate	GEN0105	GEN0160	MGS0084	MGS0156
PLAb (D,L); Mw 2K	+	–	+	+
PLA (D,L); Mw 10K	+	–	–	+
PLA (D,L); Mw 10K, ester terminated	+	–	–	+
PLA (D,L); Mw 18K	+	–	–	+
PLA (D,L); Mw 70K	+	–	–	+
PLA (L); Mw 40K	+	–	–	–
PLA (L); ester terminated	+	–	–	–
PLA (D); Mw 124 K	–	–	–	–
IngeoPLA6400	+	–	–	–
IngeoPLA4032	–	–	–	–
PLGA	+	–	+	+
PHB	–	–	–	–
PHBV	+	–	–	–
PCL; Mw 10K	+	+	+	+
PCL; Mw 45K	+	+	+	+
PCL; Mw 70K	+	+	+	+
Bionolle PBS 1001MD	–	–	–	–
Bionolle PBS 1020MD	–	–	–	–
Bionolle PBSA 3001MD	+	+	+	+
Bionolle PBSA 3020MD	+	+	+	+
PES	+	+	+	+
3PET	+	–	–	+

^aThe presence of hydrolytic activity (+) was inferred from the formation of a clear halo around the agarose wells containing the indicated enzymes. Polyesterase activity against PLA (D,L; M_w 2K) and PCL (M_w 10 K) was also reported in our previous work.⁶⁵.

^bPLA, polylactic acid; PLGA, poly(D,l-lactide-co-glycolide); PHB, poly[(R)-3-hydroxybutyric acid]; PHBV, poly(3-hydroxybutyric acid-co-3-hydroxyvaleric acid); PCL, polycaprolactone; PBS, polybutylene succinate; PBSA, poly(butylene succinate-co-adipate); PES, poly(ethylene succinate); 3PET, bis(benzoyloxyethyl) terephthalate.

Analysis of the Reaction Products of Solid PLA Hydrolysis.

To demonstrate hydrolytic activity of the identified metagenomic polyesterases against solid PLA substrates, purified MGS0156 and GEN0105 were incubated with solid poly(d,l-lactide) (M_w 10 K) powder (12 mg) suspended in 1 mL of 0.4 M Tris-HCl buffer (pH 8.0, equivalent lactate concentrations ~135 mM). At indicated time points (Figure 4), the enzyme and solid PLA particles were removed from the reaction mixture using centrifugal filters (MWCO 10 kDa), and the production of monomeric and oligomeric lactic acid products was analyzed using L- and D-lactate dehydrogenases (as described in Materials and Methods). After 6 h of incubation at 30 °C, MGS0156 hydrolyzed approximately 80% of the solid PLA substrate producing a mixture of oligomeric and monomeric products (Figure 4).

Figure 4.

Production of lactic acid during incubation of solid PLA10 with purified metagenomic polyesterases: wild-type MGS0156 (A), GEN0105 (B), and MGS0156 L169A (C) and cutinase $\mathrm{C}\mathrm{u}\stackrel{-}{\mathrm{t}}2$ from Thermobifida fusca⁷⁷ (D). Monomeric and oligomeric lactic acid products were measured using d- and l-lactate dehydrogenases as described in the Materials and Methods. Results are means ±SD from at least two independent determinations.

The proportion of monomeric lactic acid product increased with longer incubation times resulting in almost full (95%) conversion of solid PLA substrate (monomeric + oligomeric products) after overnight incubation (Figure 4). GEN0105 degraded ~70% of solid PLA after overnight incubation and was able to produce significant amounts of lactic acid within the first 30 min of incubation (Figure 4). Under the same experimental conditions, the well-known polyester-degrading cutinase $\mathrm{C}\mathrm{u}\stackrel{-}{\mathrm{t}}2$ from T. fusca⁷⁷ hydrolyzed 8.5% of solid PLA during the first 6 h of incubation with 30% conversion (monomeric + oligomeric products) after 18 h (Figure 4). Interestingly, these three enzymes showed no significant hydrolytic activity against the oligomeric PLA products (obtained using centrifugal filters). In addition, the formation of significant amounts of oligomeric and monomeric lactate products during incubation of MGS0156 and GEN0105 with solid PLA (Figure 4) suggests that they can catalyze both endo- and exoesterase cleavage of solid PLA.

Liquid chromatography–mass spectrometry (LC–-MS) was used for direct analysis of water-soluble reaction products from solid PLA hydrolysis by MGS0156 and GEN0105 (SI Figure S3). The soluble reaction products were separated using a C18 column and analyzed using mass spectrometry. These analyses revealed that both enzymes produced mixtures of lactic acid monomers and oligomers with different chain lengths (SI Figure S3 and Table S5). In line with the results of LDH-based assays, GEN0105 showed a higher degree of monomeric products compared to lactic acid oligomers suggesting a higher exoesterase activity of this enzyme compared to MGS0156 (Figure 4).

Recently, we have found that the purified polyesterase ABO2449 from Alcanivorax borkumensis required the addition of detergents (e.g., 0.1% Plysurf A210G) for solid PLA hydrolysis, suggesting that detergents can facilitate protein binding to solid PLA.⁴¹ However, in this work detergents (0.1% Plysurf A210G or Triton X-100) significantly reduced hydrolytic activity of MGS0156 against solid PLA, and had no effect on polyesterase activity of GEN0105 (data not shown). With monoester substrates, GEN0105 retained significant catalytic activity in the presence of up to 20% detergent, whereas MGS0156 was much more sensitive to detergents (SI Figure S2). Thus, metagenomic polyesterases show different kinds of responses to detergents.

Crystal Structure and Active Site of MGS0156.

Purified metagenomic esterases (GEN0105, GEN0160, MGS0084, and MGS0156) were submitted for crystallization trials, with only MGS0156 (75–421 aa) producing diffracting crystals (Materials and Methods). The crystal structure of the selenomethionine-substituted MGS0156 was solved at 1.95 Å resolution (SI Table S1), and revealed a protomer with an α/β-hydrolase fold comprised of a slightly twisted central β-sheet with seven parallel β-strands (−5x, −1x, 2x, (1x)₃) and 19 α-helices (Figure 5A and SI Figure S4). The predicted catalytic nucleophile Ser232 is positioned on a short sharp turn (the nucleophilic elbow) between the β4 strand and α8 helix. It is located at the bottom of the MGS0156 active site, which is partially covered by a ring-shaped lid domain formed by seven short α-helices (α4, α10, α11, α14, α15, α16, and α18) connected by flexible loops (Figure 5A).

Figure 5.

Crystal structure of MGS0156. (A) Overall fold of the MGS0156 protomer shown in three views related by a 90° rotation. The protein core β-sheet is shown in cyan with α-helices colored in gray, and the lid domain in magenta. The position of the active site is indicated by the side chain of the catalytic Ser232. (B) Two views of the MGS0156 dimer related by a 90° rotation. The two protomers are colored in cyan and magenta. (C) Two surface presentations of the protein tetramer shown in two views related by 90° rotation. The protomers are shown in different colors, and the active site openings are indicated by arrows.

Analysis of the MGS0156 crystal contacts using the quaternary prediction server PISA suggested that this protein may form tetramers in solution through dimerization of dimers (Figure 5B,C). The tetrameric state of MGS0156 is consistent with the results of size-exclusion chromatography, which revealed a predominance for MGS0156 tetramers (70%), as well as the presence of some octomeric (25%) and monomeric (5%) forms (151 kDa, 296 kDa, and 40 kDa; predicted Mw 39 kDa). The tightly packed MGS0156 dimer is created through multiple interactions between residues located on several α-helices (α1, α2, α10, α13, and α16) and the β1 strand (buried area 4,100 Ų, surface area 24,590 Ų). The two MGS0156 dimers are assembled into a tetramer via interactions between the α11, α15, and α18 helices (surface area 47 980 Ų, buried area 9400 Ų) (Figure 5C). In the MGS0156 tetramer, the four active sites are not adjacent to each other and are separated from the monomer interfaces with the two active site cavities open on the wide sides of the oligomeric assembly (Figure 5C).

A structural homology search of the DALI and PDBeFold databases revealed hundreds of structurally homologous proteins, mostly lipases and carboxylesterases with low overall sequence similarity to MGS0156 (<20% sequence identity). The top structural homologues include the LipA lipases from Pseudomonas aeruginosa (PA2862) (PDB code 1EX9, Z-score 24.3, rmsd 2.5 Å, 17% identity) and Burkholderia cepacia (PDB code 1OIL, Z-score 24.2, rmsd 2.6 Å, 16% identity), as well as the Staphylococcs hyicus lipase Lip (PDB code 2HIH, Z-score 23.2, RMSD 1.89 Å, 13% identity). This Dali search also identified structurally homologous polyesterases from Clostridium botulinum⁷⁸ (PDB code 5AH1, Z-score 22.3, rmsd 2.6 Å, 15% identity) and Pelosinus fermentans⁷⁹ (PDB code 5AH0, Z-score 21.4, rmsd 2.5 Å, 18% identity).

The lid domain of MGS0156 contains many hydrophobic residues creating a hydrophobic surface extending to the catalytic site cavity (SI Figure S5). The lid domain is additionally stabilized by a disulfide bond between the Cys173 and Cys287 (Figure 6). Disulfide bonds are not very common in esterase-type polyester hydrolases, with just a few reports restricted to fungal cutinases (from A. oryzae,⁸⁰ F. solani,⁸¹ and Cryptococcus sp. strain S-2⁸²). However, in cutinases the disulfide bond is involved in the stabilization of the protein core domain.

Figure 6.

Close-up view of the MGS0156 active site. The protein ribbon is colored in gray with amino acid side chains shown as sticks and carbon atoms colored in green. Only the side chains of catalytic triad and residues potentially involved in substrate binding are shown.

The MGS0156 structure revealed two conformations for the catalytic Ser232 side chain, one of which is hydrogen bonded to the Nε2 atom of the catalytic His373 (3.2 Å), whereas the other one is a bit further away (3.9 Å) and appears to be H-bonded to the backbone amide of Lys233 (2.7 Å) (Figure 6). This is similar to the recently reported two conformations for the catalytic Ser130 of the naproxen esterase from Bacillus subtilis, representing the resting and acting states of the active site.⁸³ Like in known α/β hydrolases, the catalytic His373 of MGS0156 is supposed to act as a base, deprotonating the Ser232 side chain to generate a nucleophilic alkoxide group. The MGS0156 structure also indicates that the third member of its catalytic triad is Asp350 (2.8 Å to His373), whereas its oxyanione hole appears to include the main chain NH groups of Lys233 and Leu169 (2.7 and 3.8 Å to Ser232, respectively) (Figure 6).

The composition of the MGS0156 catalytic triad (Ser232, His373, and Asp350) was confirmed using site-directed mutagenesis, demonstrating that alanine replacement of these residues produced catalytically inactive proteins (Figure 7). Like other biochemically characterized carboxylesterases,^25,41,48,84 MGS0156 has a hydrophobic acyl-binding pocket formed by the side chains of Leu169, Phe271, Leu275, Phe278, Leu299, Phe338, and Val353 (Figure 6). The alcohol-binding pocket of the MGS0156 active site is located near the catalytic Ser232 and is also filled mostly with hydrophobic residues, including Leu170, Val174, Ile334, Met378, Phe380, and Ile391 (Figure 6).

Figure 7.

Mutational analysis of MGS0156: hydrolytic activity of purified mutant proteins against mono- and polyester substrates. (A), Agarose-based screen showing polyesterase activity against emulsified PCL10. (B), Monoesterase activity against α-naphthyl acetate (2 mM, 0.02 μg protein/assay, white bars) and polyesterase activity against solid PLA10 (12 mg) measured using LDH assay (50 μg protein/assay, gray bars). Specific polyesterase activities of purified mutant proteins are shown in SI Figure S7.

Since GEN0105 failed to produce diffracting crystals, a structural model of this protein was generated using the Phyre2 server⁸⁵ and was used as a guide to identify its catalytic residues (SI Figure S6). The structural model of GEN0105 revealed a classical α/β hydrolase fold for this protein, with Ser168 as the nucleophilic serine in a conserved GXSXG motif (SI Figure S6). The other two residues of the GEN0105 catalytic triad are His292 (3.1 Å from Ser168) and Glu262 (2.7 Å from His292). The catalytic role of these residues in GEN0105 activity was confirmed using site-directed-mutagenesis (data not shown).

Structure-Based Site-Directed Mutagenesis of MGS0156.

To identify the residues of MGS0156 important for polyesterase activity, 30 active site residues were mutated to Ala or Gly using site-directed mutagenesis. Hydrolytic activities of purified mutant proteins were compared against wild-type protein activity using assays with α-naphthyl acetate, emulsified PCL10, and solid PLA10 as substrates (Figure 7). As expected, these assays revealed a critical role of the MGS0156 catalytic triad (Ser232, His373, and Asp350) for hydrolysis of all tested substrates (Figure 7). These assays also demonstrated the importance of three residues adjacent to the catalytic Ser232 (His231 and Lys233) and His373 (Asp372) (3.7–5.0 Å), which show strong sequence conservation (Figure 7 and SI Figure S1). The side chains of conserved Cys173 and Cys287 form a disulfide bridge stabilizing the protein lid domain, with alanine replacement of these residues reducing the hydrolytic activity of MGS0156 toward all substrates (Figures 6 and 7). In addition, enzymatic activity of MGS0156 against both mono- and polyesters was found to be significantly reduced in the L299G, L335A, and M378G mutant proteins, which are located in the active site cleft, likely contributing to substrate binding (Figures 6 and 7).

Reduced monoesterase activity was also observed in the L169A, L170G, E172G, V174G, S265A, L352G, and F380G mutant proteins (Figure 7). The polyesterase activity of these mutant proteins appeared to be unaffected based on agarose screens with emulsified PCL10 but was reduced (except for L169A and S265A) in LDH-coupled assays with solid PLA10 (Figure 7). These results suggest that the LDH-coupled polyesterase assay is more sensitive than the agarose-based screen. In addition, the LDH-coupled assay with solid PLA10 revealed a greatly diminished polyesterase activity in E330A, L335A, F338G, and V353A mutant proteins, whereas their activity toward α-naphthyl acetate was close to that of the wild-type protein or slightly reduced (Figure 7B). Finally, small negative effects on both polyesterase and monoesterase activities of MGS0156 were observed in the mutant proteins S175G, L179G, L197G, R199G, F271G, R277G, and E280G suggesting that these residues are not essential for substrate binding or activity of this enzyme.

Interestingly, the LDH-based assays with solid PLA10 revealed a two-time increase in polyesterase activity of L169A, whereas its monoesterase activity was reduced to approximately 20% of the wild-type protein (Figure 7). As also shown in Figure 4C, after 3 h of incubation with solid PLA10 the L169A mutant protein demonstrated at least 90% substrate conversion to monomeric and oligomeric products, whereas the wild-type enzyme hydrolyzed only 50% of substrate. This is in line with the LDH-based assays of free lactic acid production from solid PLA, which showed two-times higher activity in the L169A protein compared to wild type MGS0156 (1.1 and 0.5 mmol lactic acid/h/mg protein, respectively). In the MGS0156 active site, the side chain of L169 is located close to the catalytic Ser232 (6.4 Å) and can potentially contribute to substrate binding/coordination (Figure 6). Furthermore, the L169G mutant protein showed lower polyesterase activity against PLA10 and PCL10 compared to L169A both in the LDH- and agarose-based assays (data not shown). Therefore, we propose that hydrophobic interactions with polyester substrates at the position of Leu169 are important for polyesterase activity with the Ala side chain providing a more efficient polyester binding compared to Leu.

Recently, we have determined the crystal structure and identified eight residues critical for PLA hydrolysis by the R. palustris polyesterase RPA1511, which belongs to the esterase family V (Figure 2).⁴¹ However, a structural superposition of this protein with MGS0156 revealed only two apparently homologous residues in MGS0156: Leu296 (Leu212 in RPA1511) and Leu299 (Leu220 in RPA1511). While mutagenesis of Leu299 (to Gly) abolished both polyesterase and monoesterase activities of MGS0156, replacement of Leu296 (by Gly) had no significant effect on both activities (Figure 7). Thus, our results indicate that although the polyesterases from different esterase families have distinct binding modes for polyesters, their active sites contain a high number of hydrophobic residues playing an important role in substrate hydrolysis.

Implications.

This study demonstrated the feasibility and usefulness of medium-throughput enzymatic screening of purified uncharacterized hydrolases for hydrolytic activity against synthetic polyesters. After screening over 200 purified uncharacterized hydrolases from different environmental metagenomes and sequenced genomes for polyesterase activity, we identified 36 active enzymes with 10 proteins showing high hydrolytic activity toward several synthetic polyesters (e.g., MGS0156 and GEN0105). Sequence analysis of identified polyesterases revealed that they exhibit broad phylogenetic diversity and are associated with esterase families I, III, IV, V as well as represent new enzyme families. Detailed biochemical characterization of MGS0156 and GEN0105 showed that they degrade both emulsified and solid PLA with high efficiency, producing lactic acid monomers, dimers and longer oligomers as end products. Assays with soluble monoester substrates demonstrated that these enzymes are active under a broad range of experimental conditions (pH, temperature, salts, detergents), making them useful for various applications. The crystal structure of MGS0156 revealed the classical α/β hydrolase core with a lid domain and highly hydrophobic active site, providing structural insights into substrate binding and activity of polyester degrading enzymes. Mutational studies of MGS0156 identified the residues critical for hydrolytic activity against both polyester and monoester substrates and discovered one mutant protein (L169A) with two times higher (greatly enhanced) polyesterase activity. Our results contribute to the understanding of the polyesterase catalytic mechanism and provide valuable information for the engineering of highly active polyesterases for biocatalytic applications. Owing to their ability to depolymerize a broad range of synthetic polyesters under ambient conditions, the identified metagenomic polyesterases offer great promise for the development of polyester recycling and a more sustainable biobased economy.