Directed evolution of an efficient and thermostable PET depolymerase

The recent discovery of IsPETase, a hydrolytic enzyme that can deconstruct poly(ethylene terephthalate) (PET), has sparked great interest in biocatalytic approaches to recycle plastics. Realization of commercial use will require the development of robust engineered enzymes that meet the demands of industrial processes. Although rationally engineered PETases have been described, enzymes that have been experimentally optimized via directed evolution have not previously been reported. Here, we describe an automated, high-throughput directed evolution platform for engineering polymer degrading enzymes. Applying catalytic activity at elevated temperatures as a primary selection pressure, a thermostable IsPETase variant (HotPETase, Tm = 82.5 °C) was engineered that can operate at the glass transition temperature of PET. HotPETase can depolymerize semicrystalline PET more rapidly than previously reported PETases and can selectively deconstruct the PET component of a laminated multimaterial. Structural analysis of HotPETase reveals interesting features that have emerged to improve thermotolerance and catalytic performance. Our study establishes laboratory evolution as a platform for engineering useful plastic degrading enzymes. Enzymes for poly(ethylene terephthalate) (PET) deconstruction are of interest for plastics recycling, but reports on their directed evolution are missing. Now, an automated, high-throughput directed evolution platform is described, affording HotPETase that effectively achieves depolymerization above the glass transition temperature of PET.

1 million PET bottles being produced every minute 3 . Although mechanical recycling methods are available for PET, recycling rates remain low due to difficulties in collecting and sorting mixed postconsumer waste streams 4,5 , and declining polymer properties after repeated processing cycles 6 . In light of these challenges, depolymerization of PET into its component monomers has attracted interest as a means of circularizing the PET life cycle 7,8 . This can be achieved using chemical recycling techniques, including solvolysis methods such as hydrolysis and glycolysis 9,10 . More recently, enzymatic depolymerizations have emerged as a potentially attractive alternative 7,8 . Techno-economic analysis and life-cycle assessments predict that biocatalysis can offer a cost-effective and energy efficient approach to PET recycling. Furthermore, enzymatic recycling could also facilitate selective depolymerizations of complex mixed feedstock waste streams that are particularly challenging to recycle effectively.
For enzymatic PET recycling to be feasible, suitable biocatalysts must first be discovered and then engineered to tailor their properties for target applications. Unfortunately, while microorganisms are extremely well-equipped to deconstruct biological polymers such as proteins, DNA and carbohydrates, they are generally not well-adapted to achieve efficient depolymerization of synthetic polymers 11 . Nevertheless, some cutinases have been shown to have promiscuous PET degradation abilities [12][13][14][15] . These enzymes typically display poor activity towards PET materials with high crystallinities, akin to those commonly found in postconsumer waste. To function effectively, even engineered cutinases require the extensive preprocessing of PET substrates to amorphize the material 16 , a process that compromises the economic and environmental sustainability of biocatalytic plastic recycling approaches 17 .
The recent discovery of an organism, Ideonella sakaiensis, with the ability to use PET as a carbon source 18 , revealed a naturally evolved, PET-hydrolysing enzyme (IsPETase WT ) that has an enhanced ability to depolymerize more crystalline forms of PET 18,19 . There are interesting structural differences between IsPETase and homologous cutinases, which are thought to be linked to this improved activity 19 , including a conformationally flexible Trp185 that has been proposed to aid polymer binding 20,21 . The unique catalytic properties of IsPETase make it an attractive candidate as a biocatalyst for PET recycling. Unfortunately, the wild-type enzyme suffers from low thermostability 18 , meaning that biotransformations must be run at ambient temperatures far below the glass transition temperature (T g ) of PET (T g of approximately 60-70 °C), which compromises polymer deconstruction rates 22,23 .
In an effort to address these limitations, improvements in PETase stability have been achieved using a variety of rational engineering approaches 19,[24][25][26][27] . In contrast, experimental optimization of IsPETase using directed evolution, which typically offers a more comprehensive approach to enzyme engineering 28,29 , remains under-explored, probably due to the lack of suitable protocols for monitoring the deconstruction of insoluble plastics with sufficient throughput 30 .
Here we establish an automated, high-throughput directed evolution platform for engineering plastic deconstructing enzymes and showcase its use by engineering a thermostable variant of IsPETase that can operate at the glass transition temperature of PET. This engineered biocatalyst can efficiently depolymerize semicrystalline PET and can selectively deconstruct real-world laminated packaging materials. supernatant was monitored by ultra performance liquid chromatography (UPLC); these two products result from the partial and complete hydrolysis of the PET backbone, respectively, and are the known major products of IsPETase-mediated depolymerizations 18,31 . Using the UPLC method developed here, the MHET and TPA produced by a single degradation reaction can be analysed in under 2 minutes ( Supplementary Fig. 2). Using our integrated, automated system, over 2,000 enzyme variants can be assessed for plastic deconstruction activity in around 2 days.
The evolutionary strategy comprised sequential rounds of saturation mutagenesis, using degenerate NNK codons to individually randomize between 24-30 residue positions per cycle. In total, 106 of the 264 residues present in IsPETase were targeted for mutation throughout evolution. Residues were selected for randomization on the basis of a number of considerations, including their identification by online protein stability-enhancing tools, visual inspection of the protein crystal structure or previous reports of their involvement in substrate binding or thermostability (Supplementary Table  1). In each round of evolution, around the top 3% of hits were assessed as purified enzymes. Beneficial mutations found were then combined by DNA shuffling, and the resulting variants assayed as purified proteins to identify the most active sequence, which was then used as a template for the next round of evolution. Between   Fig. 1 | Workflow for the directed evolution of a PET depolymerase. Depiction of the laboratory evolution workflow for a single well in a 96-deep-well plate. Enzymatic PEt depolymerizations mainly produce MHEt, tPA and ethylene glycol (EG). the example UPLC trace demonstrates the MHEt and tPA produced following a 5 h PEt degradation reaction of semicrystalline PEt powder (cryPEt), with absorbance on the y axis in milli-absorbance units (mAu) and retention time on the x axis in minutes (min). Reactions were carried out at 70 °C, with both the best variant following evolution, HotPEtase (pink), and the starting protein IsPEtase tS (yellow) using 0.4% cryPEt substrate loading (4 g l −1 ) and 0.29 mg g −1 enzyme loading (0.04 μM). the crystal structure shows the 21 amino acid positions mutated from IsPEtase Wt : three positions mutated in the starting protein IsPEtase tS (yellow spheres), 16 installed through evolution (pink spheres) and a rationally installed disulfide bridge (black spheres). the catalytic triad and W185 are shown with a ball and stick representation in blue and grey, respectively. rounds 2 and 3, an additional disulfide bridge (N233C, S282C) was rationally generated in the protein, following reports that the inclusion of this structural feature increased protein stability in homologous, promiscuous PET-degrading cutinases 16,32 , leading to a 5.5 °C increase in T m (Extended Data Fig. 1 and Supplementary Fig. 3).
To simultaneously improve both thermostability and activity, the evolutionary pressures applied were gradually changed across rounds by raising both the reaction temperature and extending the reaction time. For rounds 1-4, the primary focus was on improved catalysis at elevated temperatures. Cell lysates were pre-incubated at sequentially higher temperatures (from 55-75 °C) for 1 h, before conducting PET depolymerization reactions for 3 h. The reaction temperature also increased from 55-70 °C during these rounds (Supplementary Table 1). Once a satisfactory level of thermostability was achieved, additional selection pressures of catalyst longevity and activity on more crystalline material were added. To this end, reactions in rounds 5 and 6 were conducted at 70 °C, with reaction times of 5 and 7 h, respectively, initially using amoPET as the substrate. The top 3% of clones identified during rounds 5 and 6 were then screened as purified enzymes against a commercially available semicrystalline PET powder (cryPET, 29.8% crystallinity, sourced from Goodfellow) that has a crystallinity level more reminiscent of material dominant in postconsumer waste streams 33,34 .
The most thermostable and active variant to emerge following six rounds of evolution, HotPETase, contains 21 mutations compared to IsPETase WT : three from the starting protein template IsPETase TS , two from the rational insertion of an additional disulfide bridge and a further 16 found through directed evolution ( Fig. 1 and Extended Data Fig. 1). HotPETase has a melting temperature of 82.5 °C, the highest T m recorded so far of an active IsPETase derivative. This elevated thermostability means that the enzyme can be incubated before reaction at 75 °C for 90 min with only a 6% loss of activity over 24 h (Supplementary Fig. 4). Enzyme pre-incubation at 80 °C for 90 min resulted in a more substantial 35% reduction in activity. Assessment of variants along the evolutionary trajectory demonstrated that evolution led to progressive improvements in both thermostability and activity in cryPET deconstruction assays performed at 60 °C ( Fig. 2a and Supplementary Fig. 3). While IsPETase WT and IsPETase TS have minimal activity at 60 °C, HotPETase operated well under these conditions. Biochemical characterization. We next determined the activity of HotPETase across a range of temperatures by monitoring the release of MHET and TPA over time (Fig. 2b). For comparison, analogous experiments were performed using the starting template, IsPETase TS , and the engineered thermostable cutinase LCC ICCG (ref. 16 ). Comparisons between LCC ICCG and IsPETase variants were carried out under the optimal buffer conditions for each individual protein 16,18 , using cryPET powder as the substrate ( Supplementary  Fig. 5). At 40 °C, slightly improved initial reaction rates were achieved by HotPETase versus IsPETase TS (Fig. 2b), demonstrating that the evolution of thermostability has not compromised activity at ambient temperatures. While the activity of IsPETase TS was severely compromised at higher temperatures, the rate of PET hydrolysis by HotPETase is substantially improved by operating at temperatures approaching the reported T g of PET in aqueous solutions (around 60-65 °C) 14 . At 65 °C, each mole of HotPETase releases 2.7 × 10 4 M of monomers in 1 hour, a time-course over which reaction progression is linear. At the same temperature LCC ICCG produced 5.7 × 10 3 moles of monomer product in the same time frame, highlighting the superior catalytic activity of this engineered IsPETase. For both HotPETase and LCC ICCG , the reaction rates were slightly reduced at 70 versus 65 °C.
Comparison of reactions with HotPETase and IsPETase TS at 40 °C show that evolution has afforded a more robust catalyst with increased longevity (Fig. 2c and Extended Data Fig. 2). For IsPETase TS , soluble product formation essentially ceases after 8 h. In contrast, for reactions with HotPETase, monomeric products continue to accumulate for more than 48 h. Consistent with previous studies 26 , the reaction profile is non-linear with faster initial phase for roughly 8 h, followed by a slower phase from 8-48 h. Similar, but more pronounced, non-linear reaction profiles are observed at elevated temperatures (from 60-70 °C, Extended Data Fig. 2). The time-course of reactions with HotPETase at 65 °C demonstrate that product accumulation rises rapidly for the first of 3 h of reaction (1.51 mM of MHET + TPA), but then slows substantially after this time, producing 1.61 mM of soluble monomers over 48 h (Fig. 2d). As a result, while PET depolymerization with HotPETase is substantially faster at 65 °C, the extent of depolymerization at longer time frames is greater at 40 °C (Fig. 2c,d).
Product accumulation over time is also non-linear for LCC ICCG in reactions at 65 °C, with 0.68 and 1.78 mM of monomers produced over 5 and 48 h, respectively (Extended Data Fig. 3). It is interesting to note that HotPETase operating at 40 °C deconstructs cry-PET more efficiently than LCC ICCG at 65 °C, both with respect to initial rate and extent of depolymerization over 48 h. HotPETase also depolymerizes amoPET discs (used for library screening) more effectively than LCC ICCG across a range of temperatures from 40 to 65 °C (Supplementary Fig. 6). At 70 °C, although HotPETase produces more soluble monomers than LCC ICCG over 3 h, at this temperature over 24 h, LCC ICCG is a more effective depolymerase of amoPET due to its enhanced longevity.
To understand the origins of the non-linear reaction profiles of HotPETase, particularly at elevated temperatures, we conducted experiments to supply additional enzyme or substrate once reaction progression had ceased. Addition of fresh HotPETase, following cryPET depolymerization for 24 h at 60 °C, leads to similar product accumulation versus time trends as observed at the outset of the reaction (Extended Data Fig. 4). In contrast, addition of fresh PET substrate does not give rise to any additional soluble products. These observations suggest that reactions stall due to catalyst deactivation, not as a result of inhibition by soluble released products or exhaustion of available plastic substrate. It is interesting to note that during evolution, IsPETase libraries were analysed over time frames ranging from 3 to 7 h, meaning that limited selection pressure was applied to catalyst longevity at elevated temperatures. We anticipate that adapting selection pressures during future rounds of evolution will lead to improved variants capable of operating efficiently at elevated temperatures for more extended periods.
To further explore the use of HotPETase, we next attempted to deconstruct commercial-grade PET materials. HotPETase can depolymerize milled bottle-grade PET (bgPET, 41.9% crystallinity, full material characterization can be found in Extended Data Table 1 and Supplementary Figs. 7 and 8), albeit with a reduced conversion compared to that observed with cryPET powder (Fig. 3) (9.7 and 2.8% with cryPET and bgPET, respectively). To showcase the selectivity achievable with biocatalytic depolymerizations, HotPETase was used to deconstruct a common laminated packaging tray lid composed of PET and polyethylene (PE) (1.6% crystallinity, thickness of 325 μm PET and 40 μm PE, Extended Data Table 1). This PET/PE laminate is challenging to recycle mechanically, and indeed is considered a pollutant in commercial recycling streams. The HotPETase enzyme is adept at selectively deconstructing the PET portion of this material. In this instance, the extent of depolymerization after 24 h is substantially improved at 60 versus 40 °C (9.2 versus 2.9 mM of soluble monomer products released, corresponding to a degree of depolymerization of 48.1 and 15.3%, respectively, Fig. 3a and Extended Data Fig. 5). Scanning electron microscopy (SEM) reveals significant pitting of the PET surface, whereas the PE surface appears unchanged, compared to control reactions run in the absence of enzyme ( Fig. 3b and Supplementary Fig. 9). The patterns of PET surface erosion differ in samples depolymerized at 40 versus . For all reactions presented in this figure, IsPEtase and its derivatives were assayed in the library screening buffer: pH 9.2, 50 mM Gly-OH with 4% BugBuster; LCC ICCG was assayed in its reported optimal operating buffer: pH 8, 100 mM K-Pi 16 . Error bars represent the s.d. of triplicate measurements, each replicate measurement is represented with a black circle.
60 °C, with defined pits observed at 40 °C compared with a more rugged surface at the higher temperature ( Supplementary Fig. 9). These differences could plausibly arise due to different rates and extents of polymer deconstruction at the two temperatures, or due to increased chain mobility at 60 versus 40 °C.
To improve the rate and extent of PET depolymerization achievable with HotPETase, we next optimized several reaction parameters including pH, reaction buffer, substrate loading and enzyme loading (Supplementary Figs. [10][11][12]. Under optimal conditions using 3.62 mg g −1 HotPETase enzyme loading (0.5 μM) and cryPET as the substrate (0.4% cryPET substrate loading (4 g l −1 ), 20 mg total), 6.07 mM of soluble monomer products were formed (MHET:TPA ratio of 1:0.29) within 5 h at 60 °C, corresponding to a degree of depolymerization of 31% (Extended Data Fig. 6). Differential scanning calorimetry (DSC) analysis of samples before and after depolymerization show an overall increase in crystallinity from 29.8 to 41.7%, suggesting that HotPETase preferentially degrades the amorphous PET domains (Supplementary Fig. 13a and Extended Data Table 2). Size-exclusion chromatography (SEC) analysis shows no substantive change in the molecular weight and dispersity of the remaining PET ( Supplementary Fig. 13b), which may indicate that the enzyme operates in an exo-cleavage fashion, depolymerizing individual polymer chains fully before chain transfer to a new macromolecule, thus retaining the original chain lengths in the bulk of the sample. Applying the optimized reaction conditions for cryPET depolymerization to alternative PET materials (bgPET and PET/PE laminate film) fails to enhance the rate or extent of depolymerization at 60 °C (Extended Data Fig. 5), suggesting that optimal process conditions are highly dependent on the characteristics of the material undergoing deconstruction.

Structural analysis.
To gain insights into the origins of HotPETase thermostability and its improved activity, the crystal structure of the enzyme was solved and refined to a resolution of 2.2 Å for comparison to the starting variant IsPETase TS . The structures of HotPETase (Protein Data Bank (PDB) 7QVH) and IsPETase TS (PDB 6IJ6) superimpose well, with a root-mean-square-deviation of 1.18 Å (Extended Data Fig. 7a). In HotPETase, the disulfide bridge between the Cys233 and Cys282 pair is formed as intended, with an S-S interatomic distance of 2.03 Å (Extended Data Fig. 7b). The P181V mutation results in an additional hydrogen bond between Val181 and Leu199 leading to better packing of the central β-sheet region compared to IsPETase TS (Extended Data Fig. 7c and Supplementary Fig. 14). Analysis of the surface charge distributions of HotPETase and IsPETase TS reveals substantial changes, including in the putative polymer binding cleft ( Supplementary Fig. 15). Ensemble refinements of IsPETase TS and HotPETase demonstrates that regions Ala183 to Asn190 and Cys203 to Leu216 have substantially decreased flexibility in the evolved enzyme ( Supplementary  Fig. 16).
To understand how HotPETase interacts with PET oligomers, we performed in silico docking using distance restraints to the Ser160 catalytic nucleophile and the backbone amides of the oxyanion hole (Tyr87 and Met161). The lowest energy docking pose is shown in Fig. 4a, with the PET oligomer (2-hydroxyethyl-(monohydroxyet hyl terephthalate) 4 , 4PET) occupying a shallow, extended binding cleft. The 'wobbling' tryptophan, Trp185, a feature that is thought to aid substrate binding and catalysis in the wild-type enzyme 20,35 , is present as a single conformer in apo-HotPETase and is suitably positioned to accommodate the docked 4PET in a productive pose for catalysis (Fig. 4b). Extensive remodelling of the loop region connecting β7-α5, including introduction of a bulky Tyr214, leads to a new π-stacking interaction with Trp185 that restricts its conformational freedom (Fig. 4b and Extended Data Fig. 7d). A hydrogen bonding network involving Trp185, Tyr214 and the terminal hydroxyl group of 4PET also contributes to the stabilization of the docked oligomer within the binding cleft. To explore the functional significance of the altered environment around Trp185 in HotPETase, residues installed in the β7-α5 connecting loop during evolution were reverted back to the amino acids present in the wild-type enzyme (HotPETase K212N, E213S, Y214S (HotPETase LR )). These modifications led to a substantial 7.5 °C reduction in T m and compromised catalytic performance at elevated temperatures (Extended Data Fig. 8). Catalytic activity at low temperatures is minimally affected, suggesting that in the heavily engineered HotPETase, the fixed conformation of Trp185 is not detrimental to catalysis. Combined, these results indicate that a flexible Trp185 is not a prerequisite for efficient PET deconstruction.

Conclusions
The catalytic performances of PETases have previously been improved through rational engineering using computational methods, providing an important basis towards the development of commercially viable PET depolymerases. However, the engineering of industrial biocatalysts is most commonly achieved through directed evolution. The notable lack of PETases engineered using laboratory evolution probably reflects the challenges of developing suitable high-throughput, quantitative methods for analysing the catalytic deconstruction of insoluble polymers. Here, we have developed an automated directed evolution platform for engineering plastic deconstructing enzymes and showcase its use through the development of an evolved thermostable PETase (HotPETase, T m = 82.5 °C), that can operate at the glass transition temperature of PET and depolymerizes semicrystalline PET more rapidly than previously reported PETases. HotPETase is able to deconstruct commercial bottle-grade PET and can selectively deconstruct PET in a PET/ PE laminated packaging material, highlighting the potential benefits of enzymatic depolymerizations for real-world samples with minimal pretreatment or processing. Structural characterization of HotPETase highlights formation of the intended Cys233-Cys282 disulfide bridge and improved packing of the central β-sheet region, which probably aids thermostability, alongside the presence of a single well-defined conformer of Trp185, indicating that flexibility of this tryptophan is not a prerequisite for effective catalysis. To maximize the use of our platform moving forward, it will be important to interface our evolution methods with alternative strategies for augmenting biocatalyst function, including computationally guided engineering 36 , introduction of polymer binding domains 37 and the development of multienzyme complexes 38 . Likewise, combining and optimizing biocatalytic deconstructions with enzymatic monomer upcycling methods will be an important avenue for exploration 39,40 .
In all cases, detailed techno-economic and life-cycle analysis will play a crucial role in assessing commercial viability, as well as defining target parameters for future biocatalyst engineering 7,8 .
In the future, we anticipate that by adapting the selection pressures of our directed evolution workflows, we will be able to engineer a suite of useful biocatalysts with complementary functions and improved activities under process-relevant conditions. For example, we can extend catalyst stability and lifetime by increasing reaction times and temperatures, optimize biocatalysts to act on alternative plastic substrates or enhance enzyme specificities in order that they operate on single polymer components from mixed plastic waste streams. In doing so, our laboratory evolution platform will contribute to a biocatalytic recycling strategy to recover value from plastic waste.

Methods
Gene construction. The genes encoding IsPETase TS (IsPETase S121E, D186H, R280A, signal sequence removed as by Son et al. 24 ) and LCC ICCG (LCC F243I, D238C, S283C, Y127G, signal sequence removed as by Tournier et al. 16 ) were commercially synthesized by Integrated DNA Technologies as gBlock fragments with codon optimization for expression in Escherichia coli cells. The IsPETase TS gene was cloned into the NdeI (5′ end) and XhoI (3′ end) sites of a pBbE8K vector modified to contain a C-terminal hexa-histidine tag coding sequence following the XhoI restriction site 41 , to form pBbE8K_IsPETase TS . The gene encoding LCC ICCG was cloned into the NdeI (5′ end) and XhoI (3′ end) sites of pET-22b vector (Novagen) leading to fusion to a C-terminal hexa-histadine tag coding sequence, to form pET-22b_LCC ICCG . Nucleotide sequences and expressed amino acid sequences of the genes used and plasmid maps of the vector constructs are provided in Supplementary Figs. 17-19.
Library construction. Rounds 1-6: iterative saturation mutagenesis. In each round, 24-30 residues were selected and individually randomized using cassette mutagenesis. Positions were chosen for mutation on the basis of a range of factors, detailed in Supplementary Table 1. For residue identification via the Protein One Stop Repair Shop webserver 42 , IsPETase WT was used as the input protein (PDB 5XJH), with all constraints fixed to the default settings; positions identified more than twice by the software were selected for mutation. For residue identification via the B-fitter software 43 , IsPETase WT was again used as the input protein (PDB 5XJH); the 15 top positions ranked by highest B-factor were selected for mutation. DNA libraries at chosen residue positions were constructed via standard overlap-extension PCR, using degenerate primer pairs (containing an NNK codon at the position to be mutated) and pBbE8K_IsPETase TS as the template for round 1, with the most active clone discovered at the end of each directed evolution cycle serving as the template for subsequent rounds. Primer sequences are provided in Supplementary Table 2.
Shuffling by overlap-extension PCR. After each round of evolution, beneficial diversity was combined by a process of DNA shuffling. Fragments were generated by overlap-extension PCR using designed primers that encoded for either an identified beneficial mutation or the parental amino acid. Using these primers, up to six short fragments were created, DpnI digested, PCR-purified and mixed in appropriate combinations in overlap-extension PCRs. The resulting genes contained all possible combinations of mutations (from two to five mutations per gene) and were subsequently cloned into the pBbE8K vector as described previously.
Variant gene construction. HotPETase K212N, E213S, Y214S (HotPETase LR ), was created via overlap-extension PCR with HotPETase as the template protein and primers designed to encode the wild-type residues at positions 212-214. Primer sequences are provided in Supplementary Table 3. The resulting gene was cloned into the pBbE8K vector as described previously.
Protein production for library screening. For all protein expression and screening of libraries, transfer and aliquoting steps were performed using a Hamilton liquid-handling robot. pBbE8K_IsPETase libraries were expressed in chemically competent Origami 2 E. coli cells. Single colonies from a fresh transformation were used to inoculate 180 µl of Luria-Bertani (LB) media supplemented with 25 µg ml −1 kanamycin (to maintain the pBbE8K_PETase plasmid) and 2.5 µg ml −1 tetracycline (to maintain the glutathione reductase (gor) gene-containing plasmid present in Origami 2 cells), in 96-deep-well plates. Each plate contained six positive controls consisting of clones of the parent template, and two negative controls consisting of clones containing pBbE8K_RFP (red fluorescent protein). Plates were incubated overnight at 30 °C, 80% humidity in a shaking incubator (950 r.p.m.). Expression cultures were then prepared by inoculating 460 µl of 2YT media containing 25 µg ml −1 kanamycin and 2.5 µg ml −1 tetracycline with 40 µl of overnight culture in deep-well plates. The inoculated plates were incubated at 30 °C, 80% humidity in a shaking incubator (950 r.p.m.). When an optical density at 600 nm (OD 600 ) of 1 was reached, protein production was initiated by the addition of l-arabinose to a final concentration 10 mM and plates incubated for a further 20 h at 19 °C, 80% humidity in a shaking incubator (950 r.p.m.). Cells were collected by centrifugation at 2,900g for 10 min and the resulting pellets resuspended in a lysis mix consisting of 50 µl of BugBuster Protein Extraction reagent containing 10 µg ml −1 DNase I. Cell lysis was initiated by incubation for 30 min at 30 °C, with 80% humidity in a shaking incubator (950 r.p.m.) and the lysate produced diluted with 300 µl of reaction buffer (pH 9.2, 50 mM glycine-OH (Gly-OH)). Insoluble cell debris was removed via centrifugation for 10 min at 2,900g to produce a clear cell lysate.
Production of purified proteins. IsPETase and its derivatives were expressed in chemically competent Origami 2 E. coli. Single colonies of freshly transformed cells were cultured for 18 h at 30 °C in 5 ml of LB medium supplemented with 25 µg ml −1 kanamycin and 2.5 µg ml −1 tetracycline. 1 ml of the resulting culture was used to inoculate 50 ml of 2YT medium containing 25 µg ml −1 kanamycin and 2.5 µg ml −1 tetracycline. Cultures were grown at 35 °C, 180 r.p.m. to an OD 600 of 1. Protein production was initiated by the addition of l-arabinose (final concentration of 10 mM) and cultures then grown at 19 °C for 20 h. The E. coli cells were gathered by centrifugation at 3,220g for 10 min and resuspended in lysis buffer (pH 7.5, 50 mM Tris-HCl, 10 mM imidazole, 300 mM NaCl, 10 µg ml −1 DNase I). Cells were disrupted by sonication and the resulting lysate clarified by centrifugation (13,500g for 15 min). The soluble fraction was subjected to affinity chromatography via application to Ni-NTA agarose (Qiagen). After washing off unbound proteins with the lysis buffer supplemented with 10 mM imidazole, bound proteins were eluted with elution buffer (pH 7.5, 50 mM Tris-HCl, 300 mM imidazole, 300 mM NaCl). Proteins were desalted by application to 10DG desalting columns (Bio-Rad) and eluted in storage buffer (pH 7.5, 50 mM Tris-HCl, 150 mM NaCl). For the cutinase, LCC ICCG , the gene was expressed in chemically competent E. coli BL21 (DE3). Single colonies of freshly transformed cells were cultured for 18 h at 30 °C in 5 ml of LB medium supplemented with 25 µg ml −1 ampicillin. 1 ml of the resulting culture was used to inoculate 50 ml of auto-inducible 2YT medium containing 25 µg ml −1 ampicillin. Cultures were grown at 35 °C, 180 r.p.m., to an OD 600 of 1, and then cooled to 19 °C, for 20 h. Protein purification then proceeded as detailed for IsPETase, with protein concentrations determined using an extinction coefficient of 37,150 M −1 cm −1 . HotPETase exhibits a high-level of cytosolic protein expression (roughly 110 mg l −1 ); LCC ICCG has a lower level of protein expression (roughly 20 mg l −1 ) (Supplementary Fig. 20).
Library screening using amorphous PET film (amoPET). The clarified cell lysate was incubated in foil-sealed plates for 30 min to 1 h at 55-80 °C (pre-incubation step) and subjected to centrifugation at 2,900g for 10 min to remove any insoluble protein precipitate formed. To initiate the PET degradation reaction, 60-220 µl of clarified cell lysate was transferred to a 96-deep-well plate containing reaction buffer (pH 9.2, 50 mM Gly-OH) and a single 6 mm amoPET disc cut from a sheet in each well, to make a final reaction volume of 220-400 μl. Lysate volume was varied across rounds to avoid overloading the UPLC column by keeping peak areas below 2,000 mAu and to limit evaporation at higher reaction temperatures and extended reaction times. Plates were then foil-sealed and incubated for 3-7 h at 55-70 °C, after which reactions were terminated by the addition of an equal volume of a cold methanol and 12.5 mM trifluoracetic acid solution. Following reaction quenching, plates were foil-sealed and incubated for 30 min at 30 °C, 80% humidity in a shaking incubator, 950 r.p.m. and insoluble protein precipitate removed by centrifugation for 10 min at 2,900g. A UPLC analysis sample was then prepared by transferring 100 µl of the resulting reaction supernatant into a fresh 96-well microtitre plate and the plate foil-sealed. The most active clones of each round were then subjected to a second screening round, where each clone was represented as a triplicate. All expression and screening protocols were as described above, apart from overnight culture preparation, where LB media was instead inoculated with 20 µl of a glycerol stock of the original overnight cultures from the library screening round. Details for the temperatures and lengths of the pre-incubation steps, the lysate volumes added to reactions, and the temperatures and lengths of the reaction incubations for each round of directed evolution are provided in Supplementary Table 1.
Purified protein screening using amorphous PET film (amoPET). AmoPET film assays with purified proteins were conducted as follows: a foil-sealed 96-deep-well plate containing the reaction buffer (library screening buffer, pH 9.2, 50 mM Gly-OH, 4% BugBuster, for IsPETase and its derivatives, pH 8, 100 mM K-Pi, for LCC ICCG , as reported in Tournier et al. 16 ), with a single 6 mm amoPET disc in each well, was incubated for 1 h at the reaction temperature (40-70 °C) to equilibrate all reaction components to the reaction temperature (equilibration step). For directed evolution hit retesting and beneficial diversity shuffling, purified proteins were incubated in foil-sealed plates for 30 min to 1 h at 55-80 °C before reaction set up (pre-incubation step, full details in Supplementary Table 1) and subjected to centrifugation at 2,900g for 10 min to remove any insoluble protein precipitate formed. The reaction was initiated by adding the purified enzymes to the prepared 96-deep-well plate containing reaction buffer and amoPET discs (final reaction conditions: 0.04 μM enzyme, 400 μl total volume). Protein variants were arrayed across the 96-deep-well plate in triplicate. Plates were foil-sealed and incubated for up to 24 h at the desired temperature, after which reactions were terminated by the addition of an equal volume of a cold methanol and 12.5 mM trifluoracetic acid solution. Following reaction quenching, samples were incubated for 30 min at 30 °C, 80% humidity in a shaking incubator, 950 r.p.m. and insoluble protein precipitate removed by centrifugation for 10 min at 2,900g. A UPLC analysis sample was then prepared by transferring 100 µl of the resulting reaction supernatant into a fresh 96-well microtitre plate, and the plate foil-sealed.
Purified protein screening using crystalline PET powder (cryPET) and alternative PET substrates. Crystalline PET powder assays were conducted as follows: a 12 ml lidded glass vial containing 5 ml of reaction buffer (library screening buffer, pH 9.2, 50 mM Gly-OH, 4% BugBuster, for IsPETase and its derivatives or pH 8, 100 mM K-Pi for LCC ICCG ), with 20 mg crystalline PET powder (cryPET) was incubated for 1 h at the reaction temperature (40-70 °C) to equilibrate all reaction components to the reaction temperature (equilibration step). The reaction was initiated by adding the purified protein (0.04 μM final concentration) coupled with incubation at the desired temperature under agitation at 180 r.p.m. Samples were taken at multiple times points, quenched and prepared for UPLC analysis as detailed previously. The percentage depolymerization of plastic was calculated using the mass of TPA and MHET produced, using the concentrations of each compound as determined by UPLC. For assays under optimized conditions, reactions were carried out as above, using a final enzyme concentration of 0.5 μM in pH 9.7, 50 mM Gly-OH buffer, 4% BugBuster.
For bottle-grade PET assays, bottle-grade PET pellets (bgPET, RamaPET N1) were micronized using a RETSCH PM100 Planetary Ball Mill at 500 r.p.m. for 30 min to form a powder. For PET/PE composite packaging tray lid assays, PET/ PE composite packaging lids, thickness of 325 μM PET and 40 μM PE, were cut into 6-mm discs. For both materials, 20 mg of the resulting prepared substrates were used in reactions, with reaction conditions as described above. Full characterizations of the PET substrates described are detailed in Extended Data Table 1.
Chromatographic analysis of reactions. UPLC analysis was carried out on a 1290 Infinity II Agilent LC system including an autosampler with the ultraviolet detector set to 260 nm, using a Kinetex XB-C18 100 Å, 5 µm, 50 × 2.1 mm, LC Column with a stepped, isocratic solvent ratio method. Mobile phase A was water containing 0.1% formic acid and mobile phase B was acetonitrile with a fixed flow rate of 1.1 ml min −1 . Either 1 or 4 μl of sample was injected for library screening reactions or time-resolved purified protein assays, respectively. Following sample injection, the mobile phase was set to 13% buffer B for 52 s to separate TPA and MHET, stepped up to 95% buffer B for 33 s to separate larger reaction products and contaminants, and then stepped back down to 13% buffer B for column re-equilibration until a total run time of 1.8 min. Peaks were assigned by comparison to chemical standards prepared from commercial TPA and in-house synthesized MHET, and the peak areas integrated using Agilent OpenLab software. Using this method, TPA is eluted at roughly 0.4 min, MHET at around 0.6 min and small amounts of bis(2-hydroxyethyl) terephthalate (BHET) and longer oligomers at around 1-1.2 min (Supplementary Fig. 2). TPA and MHET concentrations were calculated by preparation of standard curves (Supplementary Fig. 21).
Characterization of PET substrates pre-and post-degradation. Polymer crystallinity was determined using DSC, using 4 mg of material. DSC data were obtained from using a DSC 2500 TA instrument. Samples were run in triplicate, in series, over a −50 to 300 °C temperature range under a nitrogen atmosphere at a heating rate of ±10 °C min −1 in a 40 μl aluminium crucible. The number and weight average molecular weights (M n and M w ) of polymer chains were determined by SEC. Samples (4 mg) were dissolved in hexafluoro-2-propanol (120 μl) at room temperature. Once dissolved, HPLC-grade chloroform (1,880 μl) was added to form a uniform, colourless solution that was filtered through a 0.24 μm polytetrafluoroethylene filter. SEC analysis was conducted on a system composed of an Agilent 1260 Infinity II LC system equipped with an Agilent guard column (PLGel 5 μm, 50 × 7.5 mm) and two Agilent Mixed-C columns (PLGel 5 μm, 300 × 7.5 mm). The mobile phase used was HPLC-grade CHCl 3 at 35 °C at flow rate of 1.0 ml min −1 . SEC samples were calibrated against linear polystyrene standards (162-2.4 × 10 5 g mol −1 ).

SEM analysis of enzymatic depolymerizations of PET/PE composite packaging.
A section of a PET/PE packaging lid (710 mg) was fully submerged in reaction buffer (pH 9.2, 50 mM Gly-OH, 4% BugBuster, 50 ml total) in a glass bottle, and HotPETase (0.04 μM final concentration) added to initiate the reaction. Reactions were incubated at either 40 or 60 °C with agitation at 120 r.p.m. The PET/PE packaging lid portion was washed and a fresh buffer and enzyme solution added each day over the course of 6 days. The control reactions were run in an identical manner, but with no enzyme added. The percentage depolymerization of the PET portion of each lid section was estimated from the release of MHET and TPA monomers, determined by UPLC analysis of the reaction supernatant taken each day, assuming an estimated 12.6 g l −1 PET substrate loading. The extent of depolymerization was further confirmed by weight loss analysis of samples before and after biotransformations. Samples were analysed by SEM as follows: polymer samples were sputter coated with Au/Pd (thickness 5 nm) to prevent charging during SEM imaging and were observed using secondary electron imaging in a Tescan SC Mira, FEG-SEM with an accelerating voltage of 5 kV and probe current of approximately 2 nA.
Protein melting temperature (T m ) analysis. The melting temperatures (T m ) of IsPETase and its variants were determined using differential scanning fluorimetry. For each protein, a 50 μl sample of 5 μM protein was prepared in buffer (pH 9.2, 50 mM Gly-OH) with a final concentration of 10X SYPRO Orange dye stock solution (Sigma-Aldrich) in an optically clear, lidded PCR tube (Bio-rad). Differential scanning fluorimetry melt-curve experiments were conducted using a Bio-rad CFX Connect 96 Real-Time PCR system set on the fluorescence resonance energy transfer channel to use the 450/490 excitation and 560/580 emission filters. The temperature was increased from 25 to 95 °C with an increment of 0.3 °C s −1 . Each protein's T m was determined from a mean value for the peak of the first derivative of the melt curve from three replicate measurements.
Structure determination of HotPETase. Protein crystallization of HotPETase was achieved by sitting drop vapour diffusion of 20 nl of 6 mg ml −1 protein mixed with an equal volume of reservoir solution and incubated at 20 °C. Crystals were observed after 72 h incubation with a reservoir solution comprising 0.85 M sodium citrate tribasic dehydrate, 0.1 M Tris, pH 8.0 and 0.1 M sodium chloride (LMB screen HT96 H7 Molecular Dimensions). Before data collection, crystals were cryogentically protected with the addition of 20% PEG 200 to the mother liquor and plunge cooled in liquid nitrogen. All data were collected at Diamond Light Source. Data reduction was performed with Dials and the structure solved by molecular replacement using a search model derived from IsPETase WT structure PDB 5XJH. Iterative rounds of model building and refinement were performed in COOT and Phenix using phenix.refine and phenix.ensemble_ refinement. Validation with MOLPROBITY and PDBREDO were incorporated into the iterative rebuild and refinement process. Data collection and refinement statistics are shown in Supplementary Table 4. The HotPETase coordinates and structure factors have been deposited in the PDB under accession number 7QVH. The 4PET docking simulations were performed in ICM-Pro and resulted in a number of potential docked conformations. BHET was first docked into the active site using distance restraints to Ser160 and the backbone amides of the oxyanion hole to guide the docking towards catalytically plausible conformations. The position of the docked BHET was subsequently used as a template restraint for the larger 4PET docking. The top ranking docked pose of 4PET had an ICM VLS score of −31.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Coordinates and structure factors have been deposited in the PDB under accession number 7QVH. Data supporting the findings of this study are available within the paper and its Supplementary Information, or are available from the authors upon reasonable request.