Computational design of high-activity PET hydrolase using ADD
To improve the specific interaction of enzyme to the PET ligand and accelerate PET depolymerization, a computational strategy (Affinity Analysis of Dynamic Docking strategy, ADD) was devised. Based on the 3D structure of the proteins, MD simulations can provide more-dynamic conformation information regarding the enzyme-ligand interactions by considering the time-dependent mobility30. Relative to conventional static single-protein conformation analysis, dynamic conformations provide more-realistic and accurate representation of the enzyme-substrate binding state, and thus more-promising hotspots can be discovered. The hotspot residues were determined by the duration of binding to the substrate within the dynamic protein conformations generated by MD simulation trajectories (Fig. 1). We clustered 100 frames of dynamic conformation information extracted from the 20 ns’ MD simulation trajectories into 5 representative conformations for virtual saturation mutations (Supplementary Fig. 1), and assessed the average affinity energy of the variants to the PET substrate by molecular docking (Fig. 1). The average affinity energy reveals the specific interaction of PET ligand to the substrate binding region of the enzyme, rather than the non-specific surface adsorption, which is more critical for the catalytic activity of PET hydrolase.
To obtain the best PET-degrading activity, we selected the LCC variant ICCG (F243I/D238C/S283C/Y127G)13 as the starting enzyme for variant design using ADD (Fig. 1). Briefly, the complex structure of ICCG (PDB ID: 6THT)13 and the 3PET substrate was obtained by molecular docking. 100 frames of dynamic conformation information extracted from the MD simulation trajectories of the ICCG-3PET complex was clustered into 5 representative conformations, and used for accounting the amino acid residues that interact with the 3PET substrate. The amino acid residues with a duration ratio >20% in the dynamic conformations were selected as the hotspots for subsequent variant design (Supplementary Table 1). Virtual saturation mutations were performed on the predicted hotspots in 5 frames of the representative conformations from the 20 ns’ MD simulation trajectories. The affinity energy of the variants to 3PET was assessed by molecular docking, and variants with reduced affinity energy were collected as candidates for activity validation because they were more amenable to binding to the PET substrate. Considering the importance of active centers and evolutionarily conserved residues, a rational screen was performed to avoid designing variants in these residues. Finally, the remaining variants with potentially enhanced PET degradation activity were further verified by experiments.
Iterative engineering of ICCG
The main hydrolysis products of PET (including bis(2-hydroxyethyl) terephthalate (BHET), mono(2-hydroxyethyl) terephthalate (MHET) and TPA) are all benzene-containing compounds, which have a characteristic absorbance at 240 nm (Supplementary Fig. 2)31. Therefore, we designed a high-throughput screening workflow based on the characteristic absorbance of benzene-containing compounds for rapidly identifying variants with high PET-degrading activity (Supplementary Fig. 3).
In dynamic conformation analysis, we identified 33 hotspots with enhanced interactions with the PET ligand (Supplementary Table 1). After virtual saturation mutations and affinity energy analysis, 33 potential variants with reduced affinity energy from six hotspots were selected for activity validation (Supplementary Table 2). High-throughput screening identified 12 variants with an activity >75% of ICCG (Fig. 2a). Those variants were purified to characterize the activity with amorphous PET powder. In comparison to ICCG, 8 variants showed improved activity (Fig. 2b). The PET-degrading activity of the best variant H218Y increased by 27% compared with ICCG at 72 °C. The mutation of H218 to serine and F222 to isoleucine has been previously reported to increase PET-degrading activity at the reaction temperature below 60 °C, but the activity decreased at 72 °C22.
In the second round of engineering, the H218Y variant was selected as the starting enzyme and the number of hotspots was reduced to 27 (Supplementary Table 1). We predicted 44 variants from 17 residues to have lower affinity energy than the H218Y variant, and we assessed the activity of these variants by the high-throughput screening workflow (Supplementary Table 3). There were 10 variants with an absorbance >75% of ICCG, derived from the S100, S101, N246, S247 and N248 hotspots. Among them, three double-point mutations (H218Y/S247E, H218Y/N248D and H218Y/N248Q) exhibited increased absorbance compared with the H218Y variant (Fig. 2a). More importantly, the purified enzyme of H218Y/N248D variant showed a 23% higher PET-degrading activity than H218Y variant (Fig. 2b).
Therefore, the third round of engineering was based on the H218Y/N248D double-point variant. A total of 27 residues were determined as hotspots (Supplementary Table 1). To increase the diversity, we introduced certain hotspots from the conserved domains (W104, H164, A213, H242 and I243) into the variant design. 72 variants from 25 residues were predicted to have lower affinity energy than H218Y/N248D (Supplementary Table 4). There were 34 variants with an absorbance >75% of the ICCG, accounting for 47.2% of all the tested variants (Fig. 2a). Among them, 17 variants with an absorbance higher than the H218Y/N248D double-point variant were selected for protein purification and PET-degrading activity measurement (Fig. 2b). The activity of all variants in the third round was higher than that of ICCG, and two variants (H218Y/N248D/S247A and H218Y/N248D/I234Q) were more active than the best variant H218Y/N248D from the second round.
A time-course activity analysis was performed with the three most active variants (H218Y/N248D/S247A (LCC-A3), H218Y/N248D/I243Q (LCC-A3-2) and H218Y/N248D (LCC-A2)) with amorphous PET powder (Fig. 2c). During the 10 h reaction period, the PET hydrolysis rate of these variants was significantly faster than that of ICCG. LCC-A2, LCC-A3 and LCC-A3-2 released 26.0, 25.8 and 24.0 mM products, respectively (Fig. 2c). As the control, ICCG only produced 18.5 mM products (Fig. 2c), highlighting the superior catalytic activity of the designed variants. The measured products included TPA, BHET and MHET, of which the main component was the terminal degradation product TPA (Supplementary Fig. 4).
Biochemical characterization of the selected variants
Next, we evaluated the PET-degrading activity of LCC-A2 and LCC-A3 across a range of pH and temperatures with amorphous PET power. In comparison with ICCG, the two variants were stable and exhibited high catalytic activity across the pH range of 7.5 to 8.5. The optimum pH of the variants (9.0 for LCC-A2 and 8.5 for LCC-A3) was similar to those previously reported for ICCG13,32, indicating that the mutations did not alter the pH properties of LCC (Fig. 3a). The Tm of LCC-A2 and LCC-A3 were 95.25 °C and 94.75 °C, which were 1.11 °C and 0.61 °C higher than that of ICCG, respectively (Supplementary Fig. 5), demonstrating that the engineering of affinity to PET did not compromise the thermal stability. The PET-degrading activity of the variants and ICCG increased as the increase in reaction temperature from 72 °C to 78 °C, whereas the activity began to decrease as the temperature continued to increase (Fig. 3b). At 78 °C, the relative activity of LCC-A2 and LCC-A3 increased by 61.7% and 64.8%, respectively, compared to ICCG.
Because the PET substrate is insoluble, enzyme can only bind to the substrate exposed on the surface, called the effective substrate concentration, which is less than the total substrate concentration33. The conventional Michaelis-Menten (convMM) requires an excess of substrate relative to the enzyme. The undefined molar concentration of substrate makes it difficult to assess the convMM validity. In comparison, the modified inverse Michaelis-Menten with excess enzyme (invMM) is more suitable for the interfacial enzyme LCC19. In invMM, invKm represents the enzyme concentration required to reach half-saturation under conditions of enzyme excess, kcat (invVmax/S*0) represents the catalytic rate when all attack sites on the PET surface are covered with enzyme. The invKm values for LCC-A2 and LCC-A3 were 1.61 and 1.52 µM, whereas ICCG was 3.44 µM (Fig. 3c and Supplementary Table 5), indicating that the affinity of the two variants to the PET substrate was significantly higher than that of ICCG, and LCC-A3 exhibited the strongest affinity. This result is consistent with the predicted affinity energy of the variants (Supplementary Table 3 and 4). Regarding the catalytic rate, the kcat values for LCC-A2 and LCC-A3 were 539.9 and 525.7 nmol g−1 s−1, whereas ICCG was 426.3 nmol g−1 s−1. These results suggest that the H218Y/N248D double-point mutation provided the fastest catalytic rate and the addition of the third S247A mutation, despite increasing the affinity to PET, had no contribution to the catalytic rate.
We also compared the activity of these two variants with the recently reported efficient PET hydrolase Fast-PETase15. At 50 °C, LCC-A2 and LCC-A3 released 2.58 and 2.44 mM of products (the sum of TPA, BHET and MHET), respectively (Supplementary Fig. 6). Even though the products released from LCC-A2 and LCC-A3 was 110.0% and 98.6% higher than that of ICCG, it was still slightly lower than the 3.47 mM of Fast-PETase (Supplementary Fig. 6). These suggest that the substrate binding pocket of Fast-PETase may be more suitable for the binding of PET under this condition. Another PETase variant, HotPETase, was reported to produce 1.51 mM of product in 3 h, even using a semicrystalline PET substrate (cryPET, 29.8% crystallinity). However, after 2 h of reaction at 72 °C, LCC-A2 and LCC-A3 released 16.76 and 16.81 mM of depolymerized monomers, 4.83- and 4.84-fold more than Fast-PETase at 50 °C, respectively (Supplementary Fig. 6). These indicate that high reaction temperature is more favourable for the bio-depolymerization of PET, which is the reason for extensive work on improving the thermal stability of IsPETase, such as ThermoPETase34, HotPETase35 and DepoPETase36.
To further explore the depolymerization of PET, amorphous PET films were used as the substrate. LCC-A2 and LCC-A3 can efficiently degrade PET films, producing 53.4 mM and 51.0 mM of hydrolysis products in 8 h, 2.6- and 2.4-fold more than ICCG (Fig. 3d). This result is comparable to the recently reported promising PET hydrolase PHL-7, which had high hydrolysis activity for PET films9. However, nearly half of the hydrolysis products from PHL-7 were the incompletely degraded intermediate MHET, whereas almost all of the products from LCC-A2 and LCC-A3 were the terminal product TPA (Supplementary Fig. 7). Moreover, the variants of IsPETase, Fast-PETase and DepoPETase36, could only release 14.07 mM and 6.94 mM of hydrolysis products at a higher enzyme consumption (~11.1 mgenzyme gPET-1) after 24 h reaction at 50 °C, indicating that the LCC variants are promising for the degradation of PET films.
Structural re-analysis of the LCC variants
Considering the invKm values of our designed variants were reduced compared to ICCG, we re-analyzed the specific ligand interactions of the variants to PET through MD simulations and molecular docking. The affinity energy between ICCG and 3PET was −21.39 kcal mol−1, whereas the two variants LCC-A2 and LCC-A3 were −34.31 and −39.35 kcal mol−1, respectively (Fig. 4a). The hydrogen bonds were widely distributed and served an important role in enzyme-substrate binding. We found that the hydrogen bond number between ICCG and 3PET was 2.33, whereas increased to 3.76 and 4.79 in the LCC-A2 and LCC-A3 variants (Fig. 4b). The interaction energy of each amino acid residue to the 3PET substrate was calculated and clustered. Only eight amino acid residues from clusters 4, 6, 8 and 9 in ICCG exhibited prolonged interactions with 3PET (Fig. 4c and Supplementary Table 6). Conversely, for LCC-A2 and LCC-A3, the number increased to 12 (from clusters 1, 2, 4 and 7) and 21 (from clusters 1, 3, 5, 6, 7, 9 and 10), respectively (Fig. 4c, Supplementary Table 7 and 8). Meanwhile, the binding capacity of these amino acid residues in the variants was also higher than that of ICCG (Fig. 4c). These results were consistent with the invKm data. LCC-A3 exhibited the smallest invKm value, followed by LCC-A2, and ICCG was the highest (Supplementary Table 5), indicating that the specific interactions of the variants to PET ligand was enhanced.
It has been reported that IsPETase binding to 4PET substrate through four substrate binding regions37,38. Structural analysis showed that ICCG exhibited similar substrate binding and can be divided into two PET binding regions by a narrow channel formed by S247 and I243 (Fig. 5a, Supplementary Fig. 8). Molecular docking indicated that region 1 can accommodate the 3PET substrate. However, the narrow channel of ICCG restricted the passing of PET chains and interaction with the PET binding region 2, making it difficult to accommodate longer PET chains. We found that the substitution of N248D, S247A and I243Q can broaden the narrow channel between the two PET binding regions, resulting in the channel widths of LCC-A2, LCC-A3 and LCC-A3-2 increased from 5.9 Å to 8.5 Å, 8.9 Å and 9.6 Å, respectively (Fig. 5c, Supplementary Fig. 9 and 10b). Therefore, 4PET could cross the channel and dock to the PET binding region 2 in LCC-A2, LCC-A3 and LCC-A3-2 variants (Fig. 5b, Supplementary Fig. 9 and 10a). Similarly, Austin et al. also found that the diameter of the substrate-binding pocket affected the activity of various PET-degrading enzymes39. We hypothesize that the widened substrate channel is conducive for the LCC variants to efficiently bind longer PET chain, enhancing its substrate adaptation and the specific interactions of the variants to PET ligand. This also partly explains why LCC-A2 and LCC-A3 have lower invKm values.
We also clustered and analyzed the PET binding capacity of the key amino acid residues around the PET binding region (Supplementary Fig. 11 and Supplementary Table 9). The binding capacity of the three mutated amino acid residues H218Y, S247A and N248D to 3PET was improved (Fig. 5d, Supplementary Fig. 10c and Supplementary Table 9). The amino acid residues near the active center (Y95, F125, S165, W190, V212 and H242) also exhibited increased substrate binding capacity, which were all grouped into clusters 2 and 9 (Fig. 5d, Supplementary Fig. 11 and Supplementary Table 9). Furthermore, we found a magic amino acid residue site: H218 located at the edge of the substrate-binding pocket. After mutating H218 to amino acids with benzene ring in the side chains (F, W and Y), the PET-degrading activities of the variants all increased to varying degrees (Fig. 2b). This is consistent with previous reports that mutation of amino acid residues at the edge of the PET binding pocket to aromatic amino acid residues contributed to improving PET-degrading activity21. We hypothesize that mutation of H218 to a benzene ring-bearing amino acid resulted in the formation of an F-type pi-pi bond with the benzene ring of W190, and further corresponded to the benzene ring of W190 forming a T-type pi-pi bond with the benzene ring in the 3PET substrate, which enhanced the binding ability of 3PET to the variant (Fig. 5e). Interestingly, Chen et al. reported that H218 and W190 were key amino acid residues that affect the binding of LCC to PET, and the type of H218 site affected the side chain swing of W19022. Compared with ICCG, we suggest that both the widening of the narrow substrate channel and the improved ligand interactions near the active center were beneficial for the efficient binding and hydrolysis of PET by the LCC variants.
Bio-depolymerization of post-consumer PET bottles.
To demonstrate scale-up bio-depolymerization of PET waste, we evaluated LCC-A2 and LCC-A3 in bioreactor conditions using the Pc-PET bottles (Fig. 6a). The collected Pc-PET bottles were first crushed and amorphized to increase the exchange surface of the enzyme using regular pretreatment (see Methods section). It has been reported that the optimum depolymerization of PET in high concentration (200 g kg-1) can be achieved using 3 mg of ICCG per gram of PET at 72°C13. Therefore, we first placed an initial concentration of 200 g kg−1 pretreated Pc-PET and 3 mg of enzyme per gram of Pc-PET into the bioreactor to react in 0.1 M phosphate buffer (pH 8.5) at 72 °C.
Consistent with previous study35, the reaction curve was non-linear, with a faster initial phase of approximately 4 h followed by a slower phase of 4-8 h (Fig. 6b and Supplementary Fig. 12). In particular, the maximum specific space-time-yield of LCC-A2 reached 110.75 gTPAeq. l−1 h−1 genzyme−1 at the initial 0.4 h, corresponding to a maximum productivity of 66.45 gTPAeq. l−1 h−1. (Fig. 6c and Supplementary Fig. 13). Although LCC-A3 had a lower affinity energy to PET than LCC-A2, it did not improve the catalytic rate in scale-up bio-depolymerization of PET (Supplementary Fig. 12). The maximum specific space-time-yield of LCC-A3 was 93.25 gTPAeq. l−1 h−1 genzyme−1 and the maximum productivity was 55.95 gTPAeq. l−1 h−1. It suggests that the PET-degrading ability of the enzyme can be improved by appropriately enhancing the affinity to PET substrate19. The best variant LCC-A2 only required 5.8 h to degrade >90% of Pc-PET (Fig. 6b and Table 1), whereas the previously reported best variant ICCG required 9.3 h under the same conditions13. More importantly, over 99% of the products in LCC-A2 and LCC-A3 were the terminal degradation products TPA and EG, whereas ICCG contained nearly 20% of the incompletely degraded intermediates BHET and MHET (Supplementary Fig. 14), indicating that the variants not only degrade PET more rapidly, but also more completely.
To further improve the degradation efficiency, we investigated the bio-depolymerization of Pc-PET using the best variant LCC-A2 at higher temperatures. The results showed that the rate of PET depolymerization continuously increased as an increase in reaction temperature from 72°C to 78°C, whereas the catalytic rate began to decrease as the temperature continued to increase (Supplementary Fig. 15). In particular, the time required to degrade >90% of Pc-PET reduced to 3.3 h at 78 °C, which was 2.5 h less than that at 72 °C (Fig. 6b), and 6 h less than the most efficient LCC variant ICCG reported in the literature13. Degrading 80% of Pc-PET only required 2.5 h (Table 1), which was attractive for the rapid bio-depolymerization of PET on a large scale. The maximum specific space-time-yield reached 150.50 gTPAeq. l−1 h−1 genzyme−1, increasing 114.7% over the maximum rate of ICCG (70.1 gTPAeq. l−1 h−1 genzyme−1) (Fig. 6c and Supplementary Fig. 16).
It has been reported that rapid recrystallization of PET will happen at 75 °C, with an increase in crystallinity from 15% to 37.5% within 6 h13, which was detrimental to the bio-depolymerization of PET40. However, LCC-A2 variant only required 2.5 h to degrade 80% of Pc-PET at 78 °C (Table 1), and the PET substrate may not substantially recrystallize in such a short timescale13. We hypothesize that LCC-A2 can efficiently degrade PET in the interval before it recrystallizes, thus causing complete degradation of PET. In addition, the Tm of LCC-A2 was 1.11 °C higher than that of ICCG, which was an excellent feature to maintain high activity at increased reaction temperature (Fig S5). Regardless, the catalytic performance of the LCC-A2 variant outperformed all of the currently reported PET hydrolases, which can greatly facilitate the bio-depolymerization of PET waste.
Table 1: Time for LCC-A2 variant to degrade post-consumer PET bottles.
The depolymerization rate of Pc-PET
|
Time required for depolymerization (h)
|
ICCG-72 °C
|
LCC-A2-72 °C
|
LCC-A2-78 °C
|
|
50%
|
2.5
|
1.7
|
1.2
|
|
80%
|
6.0
|
3.8
|
2.5
|
|
90%
|
9.0
|
5.8
|
3.3
|
|