Biocatalysis for industry, medicine and the circular economy: general discussion

First published on 27th August 2024

Michal Lihan opened a general discussion of the paper by Per-Olof Syrén: This is a very fascinating proof-of-concept study illustrating how PETases could be employed to tackle microplastic particles in human blood (https://doi.org/10.1039/d4fd00014e). This brings with it a number of complications, such as recognition of these bacterial enzymes by the human immune system. Because of this, have you given thought to the approaches that could be used to engineer these PETases to evade the immune system?

Per-Olof Syrén responded: Thank you for this comment. This is an important question, we have considered PEGylation and protein design for this purpose.

Abdulrahman Alogaidi asked: have you considered increasing the ability of the enzyme to do both reactions (the flat chain and the twisted chain), and would that reduce the activity of the enzyme?

Per-Olof Syrén replied: This is an interesting question. I think our S238A PETase variant, that is trans selective, constitutes a nice complement to existing gauche-selective enzymes. As our microwave pre-treatment strategy renders PET enriched in trans, an additional enzyme would not be needed in this case.

Thomas King commented: Hi Per, cool work. You mention that the concentration of microplastics in blood has been shown to be approximately 1.6 μg mL⁻¹, but the lowest concentration that you have tested is 250 μg mL⁻¹, about 150 times more concentrated. Given that enzyme kinetics are related to binding of substrate and K_M, are you concerned about a loss of activity in real-world blood samples compared to your experiments?

Per-Olof Syrén responded: Thank you for the nice comment. This is a very relevant question. As we have observed that the enzymatic activity seems to scale with substrate concentration, we are confident that the enzyme would still be active at low substrate concentrations (i.e., lower than the 0.25 mg mL⁻¹ we tried herein, a concentration that is easily detectable by ordinary HPLC analysis). The activity in plasma was however not tried and it is possible that components in that matrix could influence enzymatic activity.

Lu Shin Wong asked: Can you comment on the size distribution of the polymers as the PET particles are degraded? Does the size distribution move around and/or widen; and do you eventually get monomers or (small) oligomers?

Per-Olof Syrén answered: Yes, the PDI of the polymer increases upon enzymatic treatment and eventually monomers are released. We actually performed an in-depth study of this in another paper.¹

1 P.-O. Syrén, et al., ChemSusChem, 2023, 16, e202300742.

E. Neil G. Marsh asked: The proposed application of a PETase as a therapeutic agent to degrade microplastics in the body is very intriguing. But PETases are not processive, so the enzyme would generate a large number of intermediate molecules on the way to the monomers. Some of these might be much more toxic than the starting microplastic. Can you comment on this possibility?

Per-Olof Syrén responded: This is a very important question. Whether or not microplastic accumulated in human serum, tissue and organs is toxic, is not fully understood. It is known that TPA is significantly less toxic than MHET and its oligomers, from which it is ultimately generated by hydrolysis. Hence it is important to implement an enzyme that generates TPA, such as our S238A variant that seems proficient in this respect.

Donald Hilvert said: Your TEM image suggests that your PETase variant binds the crystalline form of the polymer. How does this happen? The model of the enzyme complex suggests that the enzyme binds a single strand of the polymer in a deep cleft. Is the strand extracted from the crystal? Are more accessible free strands displayed on the surface of the crystal? Can you comment on these issues?

Per-Olof Syrén answered: This is a very interesting point. The polymer would contain both amorphous and crystalline domains. Our imaging studies suggest that the enzymes can “line up” along a crystalline part, exactly how this happens we don't know, and it is something we would like to study further as I find it highly interesting.

Donald Hilvert added: It's amazing that your variant can degrade crystalline PET at 30 °C, as opposed to 50 to 70 °C. Do you understand why?

Per-Olof Syrén answered: Thank you for this comment. I think matching enzyme and substrate conformation is key. Further studies on the fundamental underlying mechanism are needed.

Marc Van der Kamp remarked: Do you think the mutation S238A, leading to trans-specificity for PET substrates, may be transferable to other PETases?

You note that this mutation leads to a change in ‘wobbling’ tryptophan side-chain orientation, allowing the substrate to bind. Do you think that there is also a change in π–π stacking with the (trans) substrate involved? This is potentially relevant to catalytic efficiency/specificity – we recently found that the π–π stacking between the ‘wobbling’ tryptophan and the substrate changes during the enzyme-catalyzed reaction cycle (from T-stacking to parallel, see ref. 1). This behaviour may be different in your trans-specific S238A variant.

1 A. Jäckering, M. Van der Kamp, B. Strodel and K. Zinovjev, Influence of Wobbling Tryptophan and Mutations on PET Degradation Explored by QM/MM Free Energy Calculations, bioRxiv, 2024, preprint, DOI: 10.1101/2024.04.30.591886v1.

Per-Olof Syrén responded: Thanks for these very good questions. We have observed that S238A affects W185 in IsPETase by changing its rotamer orientation, allowing extended trans chain conformers to pass. It is difficult to predict whether the responding loop mutation will lead to the exact same effect (i.e. causing a flat to semi-perpendicular flip of the tryptophan ring) in the active site of other PETases. The π-stacking interactions will change depending on the tryptophan ring orientation and whether the polymer chain binds in a flat or twisted conformation.

Amulyasai Bakshi communicated: During the discussion, it was mentioned that the system is not suitable for therapeutics currently. However, the vision of the project is to use it in therapeutics. PET has been pre-treated under microwaves for degradation. How is it possible to pre-treat the PET that is in the blood when we use it in therapy? If the enzyme is being evolved for degradation under pre-treated substrates, would the same evolution help in non-pretreated samples?

Per-Olof Syrén communicated in reply: Thank you for this question. Pre-treatment was used here with the purpose to make bottle-derived PET microplastic-like. As our enzyme variant is active against this material, and in addition also on the second microplastic-like PET-derived substrate that was tested, we are confident that real microplastic PET can also be degraded. Accordingly, in a biopharmaceutical application no pre-treatment is needed.

Alexander McKenzie remarked: Since your S238A PETase mutant shows increased activity towards crystalline trans-PET, have you tried applying it in combination with WT/other PETase mutants to heterologous PET samples to see if this more effectively degrades mixed crystalline/amorphous PET?

Per-Olof Syrén replied: We are working according to a one-enzyme for all concept and as our microwave pre-treatment strategy yields trans PET, our trans-selective variant is solely capable of converting pre-treated PET into monomers.

Bruce R. Lichtenstein suggested: Perhaps you can comment more on toxicity: you've shown that your IsPETase variant is not cytotoxic, but what about the products of the reaction? I suspect TPA would get metabolised quite effectively in the liver and be eliminated either metabolically or excreted, but ethylene glycol is incredibly toxic. Are you thinking about biomedical applications of your enzyme? Have you considered or are you worried about ethylene glycol being released in sufficient amounts to overwhelm the body's ability to deal with it?

Per-Olof Syrén replied: Yes, we have thought about this. Ethylene glycol (EG) can be metabolized at levels below approximately 125 mg kg⁻¹ (see references in our paper). Due to the low concentration of microplastics currently found in blood we speculate that EG levels due to enzymatic hydrolysis would be lower than this value.

Agata Raczyńska commented: It is probably not exactly known what happens to microplastics that are consumed or inhaled; they might not circulate in the bloodstream but instead accumulate in cells.

Per-Olof Syrén answered: Yes, we do not yet fully understand the mechanisms associated with microplastics in humans. I think it is important to degrade the microplastics during circulation before it accumulates in organs.

Lu Shin Wong remarked: Following on from Bruce R. Lichtenstein's comment, since ethylene glycol is toxic, perhaps what is needed is a “less good” enzyme that ceases hydrolysis at BHET or MHET?

Per-Olof Syrén responded: This is a very important point. EG should not be toxic up to approximately 125 mg kg⁻¹ (see references in our paper). Due to the low concentration of microplastics currently found in blood we speculate that EG levels due to release by hydrolysis would not reach this threshold.

Artur Góra asked: When you perform modelling, you are working with a single chain, however crystalline fibers are not single chain. Have you considered modelling interactions of protein with the fiber? Is it possible that the enzymes can accommodate in its binding groove fibers? How these could be arranged?

Per-Olof Syrén replied: This is a very relevant question. Due to what is practically possible, we did not consider models taking an extended polymer chain into account. Instead we used oligomers spanning the active site as a representation of the plastic chain.

Artur Góra added: When I am thinking about differences in length and arrangement of chains I think that binding can be crucial for catalysis. It can be the rate limiting step, and thus we cannot improve decomposition without improvement of the binding affinity to spatial arrangement of chains especially for surfaces/fibers with lower flexibility.

Per-Olof Syrén answered: I agree with this comment. I think binding in the right conformation is crucial and for that reason, having an enzyme variant with an active site pre-organized to bind certain (inflexible) conformers with high affinity constitutes an advantage.

Bruce R. Lichtenstein said: To follow up on the prior comment: we know that for PET degrading enzymes that binding is not rate limiting; the K_m/K_d are in the low nanomolar range, and from prior studies it appears that hydrolysis is rate limiting.

Per-Olof Syrén replied: Thank you for this nice discussion point on a very important topic. My opinion is that binding in the right conformation (i.e. productive binding) can be rate limiting.

Ben Orton asked: With regards to distal mutations, are you exploring allosteric interactions to optimise access (and potentially reduce pre-treatment requirements) to crystallographic PET?

Per-Olof Syrén replied: This is an interesting question. We have not yet investigated allosteric pathways but rather focused on affecting the ground-state conformational landscape of the enzyme.

Nicholas Turner enquired: Regarding the delivery for PETase in vivo: we haven't discussed this today. Do you know the challenges for this? Do you need to make it protease resistant as well as encapsulate it. Any other challenges?

Per-Olof Syrén answered: This is a very relevant question. We hypothesize that delivery through encapsulation by erythrocytes could by one path forward.

Nicholas Turner added: In biopharma, do you know of any other enzyme based therapies that have gone through this process that you can use as a guidance for this?

Per-Olof Syrén responded: Yes, enzyme-based replacement therapies for lysosomal diseases have been developed (e.g. idursulfase).

Lu Shin Wong remarked: In regards to improving the pharmacokinetic properties (and perhaps reduction of immunogenicity) of administered PETase, perhaps PEGylation might be an “easy” modification. This approach is used in several approved therapeutic proteins.

Per-Olof Syrén answered: Thank you for this comment. Yes that could be one option.

Meilan Huang opened a general discussion of the paper by Louis Y. P. Luk: Is it known how the OaAEP1-C247A single mutation would affect the interdomain interface stability? Would the possible influence on interface stability account for the reduced activity you observed compared to the free enzyme (https://doi.org/10.1039/d4fd00002a)? How would the C247A mutation affect the catalytic activity? Would the mutation communicate with the rest of the catalytic pocket and hence affect the enzyme's function?

Louis Y. P. Luk replied: Many thanks for your question. C247A locates in the substrate binding pocket. Mutation to a smaller residue, such as alanine, enhances ligation turnover.

Meilan Huang asked: In addition to the C247A variant, did you look at the effect of mutating C247A into similar amino acids (such as C247S) and would it improve the activity while such variant is compartmentalized?

Louis Y. P. Luk responded: Interesting question! Cys247S has been mutated into Gly, Ala, Val, Ser, Thr, Met, Leu, and Ile, as reported by Yang et al.¹ Mutation to Gly and Ala gave the best turnover. The C247S mutant exhibits similar activity to that of the wild-type enzyme. Bulkier residues (Thr, Met, Leu and Ile) all resulted in poorer turnover when compared to WT.

1 R. Yang, et al., J. Am. Chem. Soc., 2017, 139(15), 5351.

Donald Hilvert queried: Do you have a sense of why the fusion protein is inactive? Is it properly folded on the capsid surface? Is it oriented incorrectly?

Louis Y. P. Luk answered: Thank you for the question. It remains unclear why the PAL fusion protein is less active. Plausible explanations include misfolding of PAL and an inaccessible active site. Isolating the PAL for further analysis (CD and additional kinetic characterisations) will be useful. Furthermore, use of different linkers between cpAL and PAL can be explored.

Donald Hilvert suggested: Can you test this? Could you proteolytically cleave at the linker connecting PAL to the capsid? You could then isolate the free enzyme and test whether it is active in vitro. This might help you assess whether folding is a problem.

Louis Y. P. Luk answered: Thank you for your suggestion. Yes, we should've also cleaved the PAL from the patchwork cage and examined its activity. The enzyme may not be active for various reasons, e.g. orientation of the active site and misfolding due to lack of a cap domain. Cleaving the PAL for further characterization will help pinpoint the issue.

Joelle N. Pelletier said: Taking a page from the expression of microbial transglutaminase, which is also toxic if expressed in its active form, have you considered extracellular expression? This could be full secretion, or membrane-associated (you mentioned the possibility of phage display, I believe). This is not a simple question!

Louis Y. P. Luk responded: Thank you for this question. Indeed, it is very important for my research development. In parallel, we will try the “split” approach reported in my group. It is also worth testing various secretion pathways for variant screening. Other methods such as yeast display can be considered, but there may be scenarios where PAL hydrolyses itself from the cell surface.

Lu Shin Wong asked: Can you elaborate on the benefits of encapsulation? In your talk you seem to mention an increase in stability of the encapsulated protein, but in your case the desired outcome is a reduction of toxicity from the encapsulated protein? I suppose it could be both?

Louis Y. P. Luk responded: That's an interesting question. In nature, encapsulation has various purposes: (1) compartmentalizing toxic gene products or chemicals, e.g., reactive peptidases; (2) enabling regulation by substrate sorting; (3) establishing a unique environment for designated reactions, e.g. a low pH environment for enzyme activation; (4) capturing volatile or reactive intermediates for reaction. PALs are almost always found to be compartmentalized (to my knowledge). Its activity could be cytotoxic, modifying cytoplasmic proteins randomly. Verifying PAL's cytotoxicity is technically challenging, as production of core PAL in E. coli cytoplasm often results in inclusion bodies (soluble form is low yielding). Moreover, only specific peptide substrates (pro-albumin and secondary peptide metabolites) are subjected to modifications by PALs in plant cells. Furthermore, the vacuole has a lower pH (5.0–5.5), which is more optimal for PAL's activity. Bringing these observations together, PAL compartmentalization appears to fit in purposes (1), (2) and (3).

Sabine Flitsch said: You show that you do not get a sorting effect with substrates having differently charged linkers. What is your explanation for this observation?

Louis Y. P. Luk answered: This is a good question. A protein container, entirely composed of AaLS-13 subunit, has a super-negatively charged lumen, enabling selective encapsulation of positively-charged cargoes.^1,2 In turn, the substrate-sorting capacity of a patchwork cage, composed of both AaLS-13 and circularly permutated LS (cpLS), has not been explored.³ Our result demonstrates a lack of selectivity between the neutral and positively-charged peptide substrates. This suggests that the lumen of the patchwork cage does not have sufficient negative charges for substrate-sorting.

1 E. Sasaki, D. Böhringer, M. van de Waterbeemd et al., Nat. Commun., 2017, 8, 14663.

2 R. Zschoche and D. Hilvert, J. Am. Chem. Soc., 2015, 137, 16121.

3 D. Hilvert et al., J. Am. Chem. Soc., 2018, 140, 558.

Nicholas Turner remarked: You mentioned the formation of a cyclic asparagine imide. Is that a byproduct or the peptide cleavage mechanism?

Louis Y. P. Luk responded: This is a very interesting question. Cyclisation of Asn and Asp has been discussed in different aspects of PAL catalysis. When (thio)depsipeptide was used as a label for irreversible PAL reaction, a notably higher concentration was needed.¹ This is most likely caused by aspartimide formation (Fig. 1).²

Within the active site of various PAL homologues, one of the Asp residues located in the active site (Asp174 in OaAEP1-C247A) was proposed to be rearranged in the form of succinimide.^2–5 In our group, mass spectrometry analysis of OaAEP1-C247A revealed loss of a water molecule in its molecular weight, suggesting formation of the succinimide.

1 J. P. Tam et al., Angew. Chem., Int. Ed., 2015, 54, 15694.

2 T. M. Simon Tang and Louis Y. P. Luk, Org. Biomol. Chem., 2021, 19, 5048.

3 J. S. Mylne et al., eLife, 2018, 7, e32955.

4 E. Dall and H. Brandstetter, Proc. Natl. Acad. Sci. U. S. A., 2013, 110, 10940.

5 E. Dall et al., Angew. Chem., Int. Ed. Engl., 2015, 54, 2917.

Nicholas Turner added: When it does the normal cleavage, does it involve the asparagine side chain? Is there less hydrolysis with aspartic acid?

Louis Y. P. Luk replied: Based on our experience, PAL can recognize its own aspartate residues leading to self-cleavage. Cleavage, instead of ligation, takes place when there is a lack of suitable nucleophile. To my knowledge, there is no report where an internal Asp residue was used for ligation by PAL. It is worthwhile to investigate.

Nicholas Turner remarked: What does it do with aspartic acid? Do you get more hydrolysis and less conjugation?

Louis Y. P. Luk responded: Many thanks for the question. This is an interesting point. To my knowledge, no one has systematically compare the ligation and hydrolysis ratio for Asp. PAL is most often used for Asn ligation. It is worthwhile to investigate.

Nicholas Turner said: I am trying to understand if the side chain is involved in the cleavage mechanism?

Louis Y. P. Luk replied: For the PAL OaAEP1-C247A (PDB 5H0I), His174 and Cys217 were proposed to be important for catalysis through proton shuffling and formation of a covalent enzyme–peptide intermediate (Fig. 2).¹

1 T. M. Simon Tang and Louis Y. P. Luk, Org. Biomol. Chem., 2021, 19, 5048.

Nicholas Turner further queried: How does it recognise asparagine? How is it able to recognise it and then cleave at that bond?

Louis Y. P. Luk responded: This is a very interesting question. Based on comparative sequence analysis, it was proposed that the P1 pocket, being responsible for recognizing an internal Asn within a peptide, is composed of six highly conserved residues: Arg72, His73, Asp174, Ser245, Glu215, Asp267.¹ Based on the crystal structure PDB 5H0I, Ser245 is most closely located to the Asn residue and thus likely to be involved in the interaction. Arg72 and His73 likely form part of the oxyanion hole for activation of the peptide bond (while other Asp174, Glu215 and Asp267 are in proximity and appear to be in a hydrogen bond network). Please note these are only speculations, because the PAL was solved in its inactive form with a cap domain bound.

His175 and Cys217 are essential for catalysis. Previous work has illustrated that the C217S mutation abolishes PAL's activity.² Another interesting note is that Asp174 was suggested to be rearranged in the form of aspartimide upon acid activation of PAL, but its role in catalysis has not been solved.

1 J. Lescar and J. P. Tam et al., ACS Catal., 2020, 10(15), 8825.

2 R. Yang, et al., J. Am. Chem. Soc., 2017, 139(15), 5351.

Nicholas Turner asked: Does the incoming nucleophile have to be a peptide?

Louis Y. P. Luk answered: Thank you for your question. PAL can accept nucleophiles other than peptides. These include cysteamine, ethylene diamine, ethanolamine, and 3-azido-1-propanamine, benzylamine, p-anisidine and D-alanine.¹ Nevertheless, very high concentrations are typically required.

1 D. J. Craik and T. Durek et al., J. Am. Chem. Soc., 2021, 143(46), 19498.

Nicholas Turner added: Could it be an amino sugar or an amino nucleotide?

Louis Y. P. Luk replied: The nucleophile can be an amine other than amino acid. Various types of amines, including glucosamine and a derivative of cyclodextrin, can be accepted as nucleophiles for the PAL reaction, as reported in ref. 1. To my knowledge, an amino nucleotide has not been tested but is likely feasible, considering PALs relaxed substrate scope at the P1′′ position.

1 D. J. Craik and T. Durek et al., J. Am. Chem. Soc., 2021, 143(46), 19498.

Sabine Flitsch asked: Have you tried nucleophiles other than amines? Thiols for example?

Louis Y. P. Luk responded: That's a very interesting point. PAL has been used for ligation of a glycine thioester, which enables chemical ligation.¹ It has also been tested for ligation of different amines.² To my knowledge, no one has reported testing of different thiols. One possible technical challenge is that the Asn thioester may rearrange intramolecularly to form aspartimide.³ Nevertheless, this idea is worth pursuing, such that “activated” carbonyl motifs can be generated via biocatalysis. For example, “latent” thiols (e.g. bis(2-sulfanylethyl)amine) or alcohols can be tested as nucleophiles for PAL addition.

1 J. P. Tam and C.-F. Liu et al., Chem. Commun., 2015, 51, 17289.

2 D. J. Craik and T. Durek et al., J. Am. Chem. Soc., 2021, 143(46), 19498.

3 T. M. Simon Tang and Louis Y. P. Luk, Org. Biomol. Chem., 2021, 19, 5048.

Sabine Flitsch opened a general discussion of the paper by Lu Shin Wong: Have you observed any stereoselectivity in your enzymatic formation of polymers (https://doi.org/10.1039/d4fd00003j)?

Lu Shin Wong answered: We have not investigated this aspect yet. However, this is something that we already have in mind (it was in our recent EPSRC grant). Producing polysiloxanes with controlled tacticity would be very interesting since they are very difficult to produce by other means. Generating molecules with chirality at the silicon centre is generally difficult since they are more susceptible to epimerisation (compared to the carbon analogue). In this respect, biocatalysis may offer an advantage since they can be carried out under milder conditions.

Bruce R. Lichtenstein commented: Have you tested the substituent effects on the silicon's reactivity? It would appear that you might have quite a lot of flexibility in adjusting the electrophilicity of the silicon as well as potentially the nucleophilicity of a free silanol?

Lu Shin Wong responded: We have not tested any other monomers other than the ones described in the paper. We have so far focused only on those monomers since they would give polymers that are relevant for technological applications. Altering the substituents around the silicon would certainly have large effects on the reactivity of the monomers. For example, considering the silyl protecting groups, the dimethylphenyl silyl ethers are more labile compared to the t-butyldimethyl silyl ethers and a contributing factor is the aromatic phenyl substituent.

Of course there are also other ways to manipulate the product formation, rather than engineering the enzyme, reaction engineering – altering the reaction conditions – is an alternative.

Louis Y. P. Luk asked: In terms of the mechanism, is it necessary to have hydrolysis? Is it possible for there to be an enzyme substrate that is covalently linked?

Lu Shin Wong responded: Yes, it is possible. The analogy here is with the cathepsin L, which are homologues to the silicateins. The cathepsin is a cysteine protease, with a classical “ping-pong” mechanism where the key intermediate in the catalytic cycle is an acyl-enzyme covalent intermediate. This intermediate is then attacked by the incoming nucleophile to release the substrate. In the silicateins, it has been hypothesised – but not proven – that the catalytic cycle would pass through an analogous silyl-enzyme covalent intermediate. If this intermediate was attacked by a silanol, then water would not be necessary.

Fraser Armstrong commented: With this unusual biocatalytic system, it should be possible to design to produce, in a highly controlled manner, unusual molecules with Si–O linkages.

Lu Shin Wong replied: Yes, in principle this should be possible. We have shown with our earlier work (with silyl ethers possessing a single Si–O bond) that the enzyme is relatively unselective and will accept a variety of silyl groups. It would be a good starting point for the design of biocatalysts and/or processes for the synthesis of siloxanes that have so far not been synthesised, or can only be produced with great difficulty. Of particular interest would be the synthesis of molecules with chirality at the silicon atom - as mentioned in my response to Sabine Flitsch's comment.

Thomas King remarked: Hi Lu Shin, nice work. You mentioned that the methylphenylsilane was difficult to polymerise, perhaps due to stability issues but perhaps also due to steric demands. Have you tried copolymerisation with the dimethylsilane, to reduce the steric clash?

Lu Shin Wong answered: Thanks for the question. We have not tried mixtures of monomers, but that's a good idea.

E. Neil G. Marsh remarked: The enzyme-catalyzed polymerization of silanes seems unusual – although longer polymers are produced, they are much more polydisperse. Normally polymerase enzymes produce products with low polydispersity. How do you think the enzyme works? Does it activate the monomer, which then polymerizes non-enzymatically, or does it catalyze the addition of one monomer to another? One way to address this question might be to examine how the distribution of products changes as a function of enzyme to substrate monomer concentration.

Lu Shin Wong responded: Thanks E. Neil G. Marsh for the comment. My feeling is that the enzyme catalyses the hydrolysis of the monomer to a free silanol, which then polymerises non-enzymatically. The rather wide size distribution (Fig. 1 in our paper) is consistent with your comment.

We have not carried out an experiment to specifically address this issue, but I entirely agree with your suggestion. We need a more detailed analysis of how product distribution varies with the substrate:enzyme ratio.

E. Neil G. Marsh added: Perhaps you could follow the enzyme activity directly if the silane monomer had a leaving group that could be monitored spectroscopically? This could provide a handle on the unusual kinetics of polymerization.

Lu Shin Wong replied: Yes, I agree. We had originally considered using a bis(4-nitrophenoxy)dimethylsilane substrate, since hydrolysis would release the nitrophenoxide chromophore. We had used the same approach in our previous work with silyl ethers.^1,2

However, these bis-nitrophenoxy substrates are extremely unstable. We did try to synthesise them but were unsuccessful. This is consistent with the fact that there are no examples in the literature of such substrates being used.

(We did find a couple of papers claiming their synthesis from the 1960–70's but we were unable to reproduce the reported results.)

1 S. Y. Tabatabaei Dakhili, S. A. Caslin, A. S. Faponle, P. Quayle, S. P. de Visser and L. S. Wong, Proc. Natl. Acad. Sci., 2017, 114, E5285–E5291, DOI: 10.1073/pnas.1613320114.

2 Y. Lu, C. S. Egedeuzu, P. G. Taylor and L. S. Wong, Biomolecules, 2024, 14, 492, DOI: 10.3390/biom14040492.

Donald Hilvert queried: You run the polymerization reactions in toluene. How much water is present?

Lu Shin Wong answered: Yes, water is present. We have not quantified it precisely, but there is residual water in the lyophilised enzyme preparation. I presume at least a catalytic amount of water is needed to hydrolyse one of the alkoxy groups to initiate polymerisation.

Donald Hilvert added: The building block apparently needs to be hydrolyzed to start polymerization. What is the rate of the reaction?

Lu Shin Wong responded: Yes, we presume so. We have not investigated in detail the rate of reaction. This is somewhat difficult to do in the current context. Since the reactions are being carried out in toluene, we are taking aliquots of reaction mixture for analysis of fixed time points (i.e. not in situ monitoring).

Nicholas Turner asked: What about if you look at polysiloxanes as substrates?

Lu Shin Wong replied: Thanks Nicholas Turner for the question. Yes, we have started to investigate this aspect, but I do not yet have sufficient data to report. It would be relevant for the recycling of waste silicones. We do observe a gradual lowering of MW upon prolonged reaction (see Fig. 5 in our paper), but we do not yet have details of the product distribution.

Nicholas Turner asked: Do you have access to well defined polysiloxanes?

Lu Shin Wong answered: No, these are not commercially available at the size range of interest. Standards are available for larger sizes, but these are still heterogeneous (though with a known molecular weight).

Per-Olof Syrén said: How do you plan to achieve more control of polymerization reaction?

Lu Shin Wong replied: At this stage we are only just carrying out preliminary scoping reactions, with an initial aim of maximising the molecular weight. If the polymerisation is not “processive” then our other option would be to investigate the reaction parameters. The actual numerical values of the PDI (Table 1 in our paper) are probably a reflection that the polymers are quite short, perhaps if we were able to get larger polymers we could indirectly change the PDI.

Per-Olof Syrén asked: Have you thought about learning from radical polymer chemistry, e.g. ATRP, to tune the polymerization reaction (PDI, molecular weight)?

Lu Shin Wong replied: Thanks Per-Olof Syrén for the very interesting idea. This is related to other comments by Bruce R. Lichtenstein and E. Neil G. Marsh about whether the enzyme was “processive”. If so, then one could envisage a living polymerisation with good control of molecular weight, PDI. If the current case was analogous to a living polymerisation, one could even envisage the synthesis of block co-polymers.

Friedrich Johannes Ehinger remarked: How were the control reactions, in which you see a certain degree of polymerization despite the absence of enzyme, performed? Is the matrix, i.e. lyophilisate components, the same as for the active enzyme reaction? And have you maybe even performed control reactions containing (heat-)inactivated enzyme?

Lu Shin Wong replied: The negative controls were carried out only in the presence of the lyophilisation additives (buffer salts, crown ether) but without the enzyme. Further details are given in the methodology section of our paper. We did not investigate controls involving heat-denatured enzyme in this case. However, in our previous work on the hydrolysis of small molecule silyl ethers,¹ we did try heat denatured enzymes and they gave much lower activity.

1 E. I. Sparkes, C. S. Egedeuzu, B. Lias, R. Sung, S. A. Caslin, S. Y. Tabatabaei Dakhili, P. G. Taylor, P. Quayle and L. S. Wong, Catalysts, 2021, 11, 879, DOI: 10.3390/catal11080879.

Bruce R. Lichtenstein commented: Returning to the previously mentioned RAFT or living polymerisation idea. If we push the analogy further, those processes have a separate initiation and catalytic equilibrium growth step. Have you considered perhaps introducing a means to initiate the reaction separately from the enzyme, perhaps creating a sort of ‘capped’ end? Perhaps through introduction of some fluoride salt?

Lu Shin Wong replied: Yes, if the enzymes were “processive” (i.e. adding one monomer in a stepwise fashion) then we could envisage a living polymerisation. We have not tried this, but it is certainly worth investigating, and might also help us discover whether the enzyme is in fact catalysing the hydrolysis or the condensation step (or both). The addition of a “cap” with just one Si–O bond would be a way to do this, e.g., trimethylsilanol or trimethylmethoxysilane.

Artur Góra addressed Per-Olof Syrén and Lu Shin Wong: Regarding polymer degradation. I think we need to verify the approaches which we use, for example we are trying to transfer definition of enzymatic activity on homogeneous substrates. It will not work in most cases. Binding to different surfaces or spatial arrangements can be rate limiting for some polymers rather than hydrolysis. We have a truly different situation with substrate – instead of small molecules, for which conformational changes are fast, we have long chains, fibers, crystal or amorphous surfaces where conformational sampling of the hydrolysable bond can be restricted. I think that our methodology needs to be different and reflect such changes.

Per-Olof Syrén responded: This is a very relevant comment. In my opinion, the field would benefit from standardization in terms of reporting data for enzymatic plastic hydrolysis. Still, solubility issues make analytics challenging in retrieving high-quality datasets.

Lu Shin Wong responded: Thank you for your comment. Yes, I agree. We do not really know the overall rate-limiting step. We are at the boundary of homogeneous and heterogeneous catalysis. The models we use for interfacing small molecules to proteins would not work well in our situation for the reasons you stated. However, I am not sure if the methods used by researchers modelling conventional heterogeneous catalysts would be suitable either. They are typically modelling a relatively small number of surface atoms by DFT (with some predetermined boundary conditions), and in the gas phase. For the current scenario of proteins binding to polymers on a surface, we would need to model many atoms, and in the presence of solvent molecules – I think this would be computationally very expensive. Perhaps some coarse-grain modelling could be done, but the results may not be sufficiently accurate to give useful insights. I would welcome further discussion and comments on how to address this problem by our computational colleagues.

Artur Góra addressed Per-Olof Syrén and Lu Shin Wong: To develop methods for polymer degradation we need to combine approaches from homogenic catalysis and heterogeneous catalysis, we need to bridge the different fields of expertise to provide a more consistent description for example of catalytic efficiency. In studying polymer degradation, linking efficiency with substrate consumption or product production can be too harsh an approximation. In fact we need to calculate both and also validate all products. This will be especially important for polymers which contain various building blocks. Without knowledge of composition and distribution of products (monomers, dimers, heteromers, fragments) the level of polymer degradation is not quantified well.

Per-Olof Syrén answered: I agree with this comment. We have used NMR for such in-depth studies allowing for detailed monitoring of constituents (polymer, oligomers, monomers) simultaneously using different solvents (see ref. 1).

1 P.-O. Syrén, et al., ChemSusChem, 2023, 16, e202300742.

Lu Shin Wong responded: Yes, I agree. We need more detailed experimental analysis of the product distribution and their rates for formation/consumption. We are trying to work on this but analysing and quantifying small oligomers is challenging (10–20′ mers in our case). Analytical methods for small molecules (e.g. NMR) do not work well due to the number of repeating units; but polymer analysis methods are not very suitable either. For example, GPC is not very accurate since we are at the edge of the size range that can be accurately quantified.

Artur Góra added: We are building a computational approach for fast simulation of binding between the enzyme and surface. I believe we need a fast method to access the preferences of a particular surface. As a first step we are trying to validate our approach on some known examples with proteins which bind specifically to native polymer surfaces, it is crucial to know experimentally how the protein binds, which surface it attaches to, what is the geometry of the surface. We don't have access to such experimental high quality data. Perhaps we could benefit from solid state catalysis and how they analyse their problems. Then most likely we will need to explore how important defects are for initiation of the hydrolysis of solid surfaces, I think this is still a completely unexplored field.

Per-Olof Syrén replied: Thank you for this interesting reflection, I agree that we do not fully understand the fundamental mechanism underpinning degradation at the interface between enzyme and polymer.

Lu Shin Wong replied: Yes, this is really a frontier area. Convenient computational approaches would be very welcome. We would also ideally like to get more experimental data to synergistically support such efforts. However, this is challenging and we are really only at the very early stages.

Bruce R. Lichtenstein said: In terms of rate limiting steps: it depends upon where the hydrolysis is occurring and what enzyme/polymer we are considering. For PET, there is growing evidence that much of the productive hydrolysis occurs in solution, with the enzymes taking soluble chunks of the polymer and most bond cleavages occurring away from the plastic surface, although this depends upon the condition and enzyme. This is obviously very difficult to model, because although we have inverse Michaelis–Menten (and related Langmuir) models that work well enough for PET, it does not appear that we have a good framework for modeling such biphasic (bio)-chemical catalysis. Are we looking at something for which we have no models?

Per-Olof Syrén answered: I agree with this nice discussion point. I think further studies are needed to fully understand the mechanism underpinning enzymatic plastic depolymerization.

Lu Shin Wong added: Yes, I think so. The classical Michaelis–Menten model would not be applicable in the current scenario since we are not dealing with interfacing a protein and a small molecule. I am not sure about the inverse Michaelis–Menten model, since this also has a lot of assumptions that are implicit in the model. We could certainly carry out experiments and calculate some values, but it is not clear to me what such values actually represent, and I am not comfortable with how they could be used to draw any meaningful insight. Nevertheless, yes, this is something we can try.

Sabine Flitsch asked: How do you evaluate solid substrates. Do you use kinetic models? How do you deal with substrate concentrations. How do you consider K_m for example?

Per-Olof Syrén responded: Thank you for this very relevant question. We focused on specific yield as we think this is an important parameter to evaluate process performance. We did consider relative apparent binding affinity to plastic by microscopy using labelled protein.

Louis Y. P. Luk asked: How do you compare with other PETase reports in literature without universal kinetic parameters?

Per-Olof Syrén answered: This is a very relevant question. I advocate for the use of parameters from chemical engineering such as specific yield. In this way, different reports and datasets with respect to efficiency of depolymerization can be compared.

Bruce R. Lichtenstein continued: There is an emphasis on absolute quantitation of the reaction kinetics, and we do have limited models, like the inverse Michaelis–Menten approaches, but are such metrics actually valuable? We are developing these enzymes for industrial processes constrained and defined by the technoeconomic analyses (TEA) and the life cycle assessments (LCA), and the constrictions place certain practical requirements on how we evaluate the enzymes and what we define as better or worse. Faster kinetics does not necessarily mean more product produced overall, but which on balance is better for an industrial process is defined by outside market parameters and costs. Although there is a great deal of heterogeneity in the field, there is a growing consensus on how to evaluate enzymatic activity: HPLC (and spectroscopic) methods have supplanted inaccurate weight loss parameters for measuring product release, there is a growing tendency of using a single universal source of standardised film, etc. These need to be more systematically applied to polymers in general, but the framework for good practice is there.

Per-Olof Syrén answered: This is an important discussion-point. In my opinion, implementing parameters from chemical process engineering such as specific yield, is one way to achieve standardisation of reporting of data for enzymatic depolymerization. Analytics and associated solubility issues constitute challenges when evaluating depolymerization, as PET and its constituent monomers and oligomers often show low solubility, requiring different solvents.

Sabine Flitsch asked: What underlying physical parameters are you trying to improve with rational redesign of enzymes for polymer degradation? Is there a difference to methods used for soluble substrates?

Per-Olof Syrén replied: This is a very interesting fundamental question. We are aiming to generate a match between enzyme ground state and plastic chain conformation to increase the depolymerization rate. One difference here is accessibility which constitutes one challenge limiting plastic degradation rate by enzymes.

Sumire Honda opened a general discussion of the paper by Stefan Lutz: Could you comment on where mutations of your best variant T7-68 mainly lie (https://doi.org/10.1039/d4fd00023d)? Were most of them located in the active site or in another region of the enzyme? Why was T7 RNA polymerase used as the scaffold for engineering? Would you have considered first identifying a homologue with higher stability or activity?

Stefan Lutz answered: Beneficial mutations are typically found throughout the entire enzyme. Given our focus on the polymerase's ability to more efficiently process capping reagents, our library design in some rounds probed for amino acid substitutions closer to the active site. As far as selecting the most suitable starting point for this (or really any) enzyme engineering project, some level of detectable activity is desirable in order to quantify improvements of variants over the starting enzyme. That said, there are ample examples among engineered enzymes at Codexis and in the literature where evolution resulted in substantial gains in both activity and stability.

Peter Stockinger remarked: As far as I remember from the literature, these RNA polymerases are highly dynamic. Conformational changes upon substrate binding and during catalysis play a critical role (e.g. in byproduct formation). Did you address the conformational dynamics during your engineering campaign?

Stefan Lutz answered: The literature certainly discusses the importance of protein dynamics and conformational changes as part of RNA polymerase function, in particular during the transition from initiation to extension. While probing regions throughout the enzyme structure for favorable amino acid changes, one is likely to impact areas responsible for conformational changes as well. More specifically, our work on the RNA polymerase is likely to have included such regions, though they were not specifically targeted, nor did we study the impact of favorable substitutions on specific properties of the enzyme, but instead focus on the overall mRNA synthesis performance.

Lubomír Rulíšek asked: Do the dsRNA by-products interact with the cGAS-STING pathway in the immune system in the same way as dsRNA does?

Stefan Lutz answered: According to our reading of the literature related to the immunogenic effects of these dsRNA byproducts, the primary mechanism involves pattern recognition receptors, endosomal-bound toll-like receptors, and other proteins. Please see the discussion of specifics in our published manuscript, including relevant references to other sources in the literature.

Nicholas Turner commented: It seems that there will be need to develop enzymes that are broad with respect to the template, capping agent, nucleotides etc. Is this a different paradigm to other biocatalyses? Ideally we would like just one engineered enzyme that can be universal. How have you thought about this in terms of screening and protein engineering strategies to achieve enzymes that can achieve broad specificities in terms of substrates?

Stefan Lutz replied: In traditional enzyme engineering, the focus has been on developing biocatalysts for one specific transformation. For the application described in our manuscript, the manufacturing of mRNA, and really the broader use of enzymes in molecular biology and in the life sciences specifically, we need to consider a new engineering challenge – the inherent biases of enzymes towards specific substrates and sequences. To be more precise, these biases are small and might go unnoticed in day-to-day use but negatively impact applications such as next-generation DNA sequencing, DNA synthesis, and RNA manufacturing. In these use-cases, enzymes such as ligases and polymerases must ideally be agnostic to the particular nucleic acid substrate (i.e. sequence composition). For example, a polymerase interacting with 3 nucleotide positions on the primer sequence plus the incoming nucleoside triphosphate must ideally be processing (3 + 1)⁴ or 256 different combinations of substrates. To evolve a biocatalyst for maximum substrate promiscuity/minimal sequence bias, presents a new challenge for enzyme engineering. At Codexis, we have successfully tackled this challenge by consistently evaluating our libraries of enzyme variants against a diverse set of possible substrates to identify variants with reduced substrate bias.

Nicholas Turner remarked: Now you have 35 substrates, whereas previously you had 1. Do you now have to compensate down on the library size?

Stefan Lutz responded: We typically evaluate 1000–2000 variants of our libraries, a size compatible with assessing a larger number of substrates.

Donald Hilvert commented: Post-transcriptional modification or RNA is important in biology. Are you also investigating enzymes that can promote such reactions, for example site-specific methylate mRNA?

Stefan Lutz replied: Yes – the use of native and engineered enzymes for introducing co and post-transcriptional modifications in RNA sequences is an exciting area for leveraging the power of biocatalysts, in particular their regioselectivity and mild reaction conditions.

Donald Hilvert said: Can you say how you assay for such reactions?

Stefan Lutz responded: The work described in our paper highlights a very effective and successful strategy to screen under process-relevant conditions while maintaining throughput. The search for an improved RNA polymerase for mRNA manufacturing is most meaningful when all variants are being evaluated on the actual synthesis of an mRNA strand. To facilitate the efficient analysis of enzymes that successfully cap their nucleic acid sequence, we considered a self-cleaving ribozyme resulting in liberation of a small 5′-fragment that could be characterized using rapid LC-MS methods. A similar fragmentation approach could be envisioned for other modifications.

Dominic J. Campopiano asked: Because of the mechanism you described, does that mean they will not work in flow at all? Because you are going to have this templated substrate? Otherwise, the enzyme won't work?

Stefan Lutz responded: It is my understanding that the traditional in vitro transcription reaction for mRNA manufacturing is run as a batch process. That said, there are alternate processes that rely on immobilized templates that enable continuous production with the added potential benefit of simplifying down-stream processing (eliminates the need for DNA template hydrolysis) and the opportunity to recycle template. Either way, the high processivity of the RNA polymerase should ensure compatibility with both operational modes.

Dominic J. Campopiano added: So, at scale, is that a typical vessel?

Stefan Lutz responded: I can't speak to the typical process scale but in batch, the in vitro transcription reaction places few requirements on infrastructure beyond temperature control and mixing.

Dominic J. Campopiano commented: Cellulose isolation? The isolation technology afterwards?

Stefan Lutz replied: Down-stream processing of in vitro transcription reaction at scale-up/manufacturing scale typically involves chromatography and filtration. While ion exchange and ion-pair reverse phase chromatography methods are routinely used, a number of other more specialized separation strategies can be deployed. Cellulose chromatography was recently reported to remove double-stranded RNA from mRNA preparations with high specificity and reasonable recovery of the mRNA product. Given the extra unit operation to remove the dsRNA side product and associated losses in mRNA yield, manufacturing strategies that eliminate or minimize such side products in the first place certainly seem more desirable.

Anthony P. Green commented: How do you assess polymerase fidelity and how does fidelity change during evolution?

Stefan Lutz replied: Polymerase fidelity of variants was assessed by a direct sequencing approach as outlined in our manuscript. Enzymes displaying inferior performance were de-selected in subsequent rounds of evolution.

Anthony P. Green furthered: Is it technically challenging to assess fidelity when you make long sequences containing chemical modifications?

Stefan Lutz answered: Assessing polymerase fidelity as part of a directed evolution campaign is technically not trivial but necessary to ensure that final variant meets a key performance indicator required for mRNA manufacturing. To confirm the incorporation of chemical modifications in longer nucleic acid sequences, autocatalytic fragmentation as demonstrated for successful capping of mRNA in our screen with the help of an encoded ribozyme, proved highly advantageous for faster sample analysis.

Abdulrahman Alogaidi asked: With respect to the safety of this technology when applied as a vaccine, would the mRNA, modified with an unnatural cap that protects from degradation, pose a burden to the cell as it does not degrade as fast as the natural one?

Stefan Lutz responded: I am not aware of any such effects. To the best of my knowledge, these capped mRNAs might prolong the asset’s half-life in the cell but do not delay degradation any more than naturally capped transcripts.

Isabelle Bruton opened a general discussion of the paper by Daniel Dourado: The integration of biocatalysis into organic synthesis can be very valuable especially for the pharmaceutical industry. How did you develop your process engineering (https://doi.org/10.1039/d4fd00011k)? Did you find that the enzyme caused problems in synthesis later downstream? Once biosynthesis was done, did you have to do any specific work-up? And how did the work-up compare to a typical chemical work-up?

Daniel Dourado responded: Thank you for your question. After the biocatalytic production of intermediate (13R-17S)-5, the compound was crystallized and isolated with a Hep/IPA (142 mL, 5:1, 1 vol) mixture achieving a yield of 80%. Moreover, Celite filtration removes the bulk of the protein residue and that work-up is virtually identical to a typical chemical work-up (i.e. extractions, aqueous washes, solvent swaps, crystallisation). There were no issues related to the enzyme further downstream.

Isabelle Bruton addressed Gareth Brown and Daniel Dourado: Did you find selectivity had to be directed by any of the groups rather than by enzyme engineering? Or was your synthesis already established and the enzyme was simply slotted in, i.e., was the synthetic route designed with the enzymatic transformation in mind (and did this result in groups being in place to drive selectivity) or did the enzyme simply replace a step in an already established route?

Daniel Dourado answered: Thank you for your question. Synthetic routes to produce the APIs ENG and LNG are known and have been used in industry. However, the cost associated with these routes is high. For instance, ENG production cost is ∼200000 dollars per kg. ENG and LNG can be made via late-stage optical resolution of a racemate or via early intermediate stage asymmetric synthesis. We focused on the latter, which can be assumed to be the most cost-effective approach. There is a literature precedent for asymmetric synthesis involving the production of a key stereocentre (13R-17S)-5. We introduced a CRED enzyme to catalyse this step, converting compound 4 into the above mentioned (13R-17S)-5 intermediate. This step was followed by the further diastereoselective synthetic transformations we developed. In the end the cost associated with the developed biocatalytic route to produce these valuable APIs decreased substantially, allowing its use in developing countries.

Rhiannon E. H. Jones asked: It was during early scale up work that you realised solubility and solvent tolerance were key properties for a successful enzyme variant. With this new information, did you consider going back to screen the full panel again, this time looking for solvent tolerance?

Daniel Dourado responded: Thank you very much for your question. We started by screening 297 CRED enzymes from Almac's selectAZymeTM panel. The best 11 enzymes showed, under the initial reaction conditions, high conversions and selectivity towards the wanted product enantiomer. All these 11 enzymes were submitted to initial single parameter optimization trials (co-factor, temperature, pH, and solvent tolerance). CRED from Bacillus wiedmannii (CRED-BW) was the best enzyme showing complete conversion at 20 wt% enzyme loading. Further enzyme loading studies with CRED-BW revealed that ∼2 wt% of this enzyme was sufficient for reaction completion. It was decided to move forward with this enzyme, to rational engineering it to increase stability so its lifetime in the presence of organic solvent would increase – its only major drawback. Eventually, we could have gone back and tested the initial 297 panel for organic solvent stability and possibly we would have found one that is more tolerant to organic solvent. However, the timelines for enzyme engineering, process development, fermentation development and reaction development were very strict. For instance, we had 1 month to in silico rational designed CRED_BW, and 2 months more to perform the in vitro screenings. Since, we had already identified an enzyme with 100% conversion and very high selectivity (ee value 99), which we knew we could engineer to improve the stability issue, it was decided to move forward with it. This was the best decision since in the end we met the deadlines. One of two contraceptives is at the moment in manufacturing with the biocatalytic route we developed, and the second will follow shortly. If we had delayed the entire development to find a better initial enzyme, this would most likely not be possible.

Peter Stockinger commented: Following-up on the previous question – I had a similar question in mind before it was asked. Looking back, do you think it would have been more sensible to choose a template with greater (solvent) stability for engineering, and then adapt the binding site to achieve optimal substrate fit?

Daniel Dourado responded: Thank you for your question. Eventually, we could have gone back and tested the initial 297 panel for organic solvent stability and possibly we would have found one enzyme that is more tolerant to organic solvent. In the previous response we answered why it was decided not to follow that strategy and how we moved forward. If we just focus on the rational engineering perspective of the problem, even if we had considerably more time to engineer the enzyme, we would still prefer to start with an enzyme that shows high specificity, high selectivity and a low organic stability than the other way around. Typically improving an enzyme’s stability is easier and more cost-effective than improving specificity and selectivity. Also, some CREDs on the panel also produced unwanted diol by-products. If we mutate the active centre we might inadvertently increase the production of these diols.

Friedrich Johannes Ehinger addressed Gareth Brown and Daniel Dourado: Initial rounds of enzyme screening were performed under low substrate loading and only later, process-relevant conditions were applied. Could you comment on this approach?

Daniel Dourado replied: Thank you for your question. The main goal of the initial screenings in the 96 well plate was to identify which enzyme variants were more stable in the presence of co-solvent and under high temperatures. Enzyme loading was kept low to prevent the reaction reaching complete conversion, thus facilitating a meaningful comparative readout. Similar to screening, initial single parameter optimization trials were conducted at low substrate concentration to avoid any impeding inhibitory or physical effects. Once the best variants were identified and having assessed the performance and true parameter preference in low substrate titre conditions, we were in a position to start increasing the reaction scale and explore the potential for increasing substrate concentration, so as to establish a volumetrically productive and cost-efficient process towards delivery of (13R,17S)-5.

Gareth Brown added: In Almac we typically carry out initial enzyme screening reactions under dilute conditions with high enzyme loading to avoid issues with product/substrate inhibition and also to avoid necessity for strict pH control (in this instance production of gluconic acid via the glucose/GDH co-factor recycle system will cause a drop in pH). Once we have established that an enzyme is suitable (in terms of selectivity) for a desired transformation, we then move into screening with defined enzyme loading to allow meaningful comparisons of activity between potential ‘hit’ enzymes. Having identified enzyme(s) with the desired selectivity and activity under dilute conditions, we next establish temperature, pH, cosolvent, cofactor preferences etc. of the enzymes in question before moving progressively towards larger scale reactions with pH control (where necessary) and more process relevant substrate titre conditions.

Tara C. Lurshay addressed Daniel Dourado and Gareth Brown: When choosing cosolvents, is there a standard selection of these that you screen, and do you take into account how green these are when you are moving towards a biochemical process?

Gareth Brown responded: For enzyme screening we typically begin with DMSO as a cosolvent if we are screening a substrate with limited water solubility. DMSO as 5–10% v/v of an aqueous mixture is usually fairly well tolerated by a broad swathe of our enzymes. However, it is not always the best choice, sometimes we require a water-immiscible cosolvent to act as a reservoir for substrate or product to mitigate against enzyme inhibition, in this case MtBE or toluene, for example could be better choices. For a typical cosolvent screen we would cover a range of water-miscible and water immiscible solvents (DMSO, MtBE, toluene, 2-MeTHF, MeOH, EtOH, IPA, acetone etc.) but the list can be extended substantially depending on the needs of the particular project (substrate/product solubility etc.). We would tend to stick to greener solvents as far as possible and would avoid chlorinated solvents (such as DCM), DMF, dioxane etc.

Francesca Valetti said: Is it relevant to have a regeneration system for NADPH in your set-up or, cost wise, it is the same to consume NADPH and re-add it once it is oxidized?

Daniel Dourado responded: Thank you for your question. Yes, having a NADPH regeneration system is the best time and cost-effective approach. We use our own glucose dehydrogenase (GDH) enzyme to recycle NADPH.

Nicholas Turner asked: Is methionine oxidation a common problem? Is this something you routinely do to mutate these residues?

Daniel Dourado answered: Thank you for your question. Yes, we have more examples of enzyme engineering projects where taking out methionines and introducing new strong intermolecular interactions in its place, was a successful approach. Nevertheless, it does not mean we always mutate these residues. In this present CRED rational engineering project, the specificity and selectivity were not an issue. The main problem was the poor stability of the WT enzyme in organic solvent. So, the rational engineering was entirely focused on increasing the structural stability in these conditions. As described in our paper, we followed different rational engineering approaches, including mutating methionine residues, not only on the surface of the enzyme but also in the core of it. We designed 93 mutants, which could be efficiently tested in a single 96-well plate. We have worked on other enzyme engineering projects where selectivity and specificity needed to be improved and so mutating methionines was not a priority.^1–8 Nevertheless, we always follow a multi-approach strategy to increase our success rate and be less exposed to other variables like expression and folding issues.

1 L. J. Bannon, D. F. A. R. Dourado, D. A. Lipscomb, P. B. A. McIntyre, S. Mix, T. S. Moody, D. Quinn, P. D. Jones and A. P. S. Narula, WO Pat., WO2022051761A2, 2021.

2 T. S. Moody, I. R. Miskelly and D. J. Quinn, WO Pat., WO2023245039A1, 2018.

3 H. Namanja-Magliano, J. L. Bettiol, S. Rupard, J. Velasquez, D. Gonzales, E. Barnard, D. Quinn, T. Moody and D. Dourado, WO2023044183A1, 2023.

4 L. D. Humphreys et al., Org. Process Res. Dev., 2022, 26, 849.

5 T. S. Moody and D. Castagnolo et al., ACS Catal., 2023, 13, 4742.

6 A. T. P. Carvalho et al., Org. Process Res. Dev., 2022, 26, 2351.

7 M. Huang et al., ACS Catal., 2016, 6, 7749.

8 J. S. Carey and T. S. Moody et al., Org. Process Res. Dev., 2024, 28, 729.

Dominic J. Campopiano addressed Daniel Dourado and Gareth Brown: Regarding the pathway flowchart: so, you said you might use other biocatalysts in this pathway. If you go from 10 to 11, are you oxidising effectively at the same place that you are reducing. I don't know what opener oxidation is, but can you use your enzyme there? Or screen again there to oxidise at that position?

Daniel Dourado responded: Thank you for your question. You could use a CRED for this step, however the Oppenauer oxidation is the most efficient and cost-effective approach to convert 10 to 11.

Dominic J. Campopiano addressed Daniel Dourado and Gareth Brown: Regarding funding – Bill and Melinda Gates Foundation? Did you approach them? How did that happen?

Daniel Dourado answered: Thank you for your question. The Bill and Melinda Gates Foundation did fund this project. It started with an initial approach by our Associate Director Stefan Mix, the corresponding author of this publication.

Alexander McKenzie opened a general discussion of the paper by Bruce R. Lichtenstein: Your study tested the effects of PETase fusion partner stability on activity in vitro; have you carried out any in vivo experiments to test this (https://doi.org/10.1039/d4fd00067f)? An unstable protein might be more severely effected in terms of expression or in vivo activity.

Bruce R. Lichtenstein responded: For PETases in vivo activity is not easily defined – these are proteins that cleave their natural substrates outside of the bacterial/yeast cell. We did find that the fusion partners may actually express slightly better than the wildtype proteins, but it is difficult to pin down the causes of this to the protein itself: there are other reasons this may be outside of the protein stability itself, and the variation was not more than a typical expression-to-expression variance.

Agata Raczyńska said: Did you investigate the impact of the fusion on enzyme flexibility? PETases have flexible binding sites that can accommodate large ligands, but more thermostable variants often lose this flexibility, which can reduce their activity. However, your method appears to improve thermostability without making the structure more rigid.

Bruce R. Lichtenstein replied: We did not look at protein flexibility, but we do suspect that some of the small changes in activity and thermal melting properties that we are seeing are a consequence of variations in the protein flexibility. It should be noted that the fusions are quite far away from the active site (on the opposite side of the protein) so these effects are likely small. I would not necessarily characterise our results as enhancing the thermal stability of the protein(s), the variations are small (equivalent to the melting temperature change expected for a single point mutation). I think what the work highlights is that for very small fusion partners, especially those that reversibly fold, changes in stability in the protein of interest may be very subtle.

Daniel Dourado commented: Did you try different sizes for the linkers?

Bruce R. Lichtenstein answered: We did. As discussed in the manuscript briefly, we originally tried the full length SpyCatcher003 sequence but found through experiments and modelling that this interfered with both expression and activity, likely owing to the potential of the longer linker in the full length construct to wrap around the PETases and sit within the groove of the active site. Switching to the ΔN1 variant alleviated these issues.

Daniel Dourado remarked: SpyTag and SpyCatcher form a covalent bond, right?

Bruce R. Lichtenstein replied: Yes. This is critical for this work, we wanted an irreversible change in the melting temperature of the fusion partner, so either an exceptionally high affinity (like monomeric streptavidin to biotin) or covalent linkage.

Dominic J. Campopiano said: In terms of the AlphaFold model displayed. So that is what AlphaFold does with PolyGly?

Bruce R. Lichtenstein answered: It is indeed. I have not yet tried this since releasing the AlphaFold3 model, but these are obviously just a snapshot view of what AF2 thinks a PolyGly linker would look like.

E. Neil G. Marsh said: There are many examples in the literature where people genetically join two or more enzymes together with the idea that the proximity of active sites will make a cascade reaction proceed more efficiently by setting up an assembly line. Sometimes higher activities/conversions are reported using this approach. It intuitively sounds like an attractive approach and it is a misconception that is widely prevalent. A rigorous analysis of the problem – see e.g. the work of the Hess group at Columbia¹ – shows there is very little improvement to be gained by this approach. On the other hand, making fusion proteins to enhance enzyme stability does seem like a promising strategy.

1 O. Idan and H. Hess, ACS Nano, 2013, 7, 8658.

Bruce R. Lichtenstein replied: I agree. This work was conducted in light of the fact that, as expected from the Hess group and others, most enzymatic fusions with PETases show little benefit with one highlighted exception. Our construction of this research was to establish an approach and investigate the effect of the thermodynamics of the fusion partner on the stability and activity of PET degrading enzymes, hoping to shed light on whether a priori assumptions concerning the stability of these fusion partners were correct: that is, that the fusion partners have to be folded to be able to witness an improvement in stability/activity of the enzyme and that if they are unfolded their effects would be limited or interfere. The results of our study indicate that what dominates in these considerations, at least using small reversibly folding fusion partners, is the intrinsic stability and dynamics of the enzyme itself (i.e. IsPETase is generally intolerant to fusions, while more thermally stable PETases are not particularly bothered by being associated with unfolded (or folded) protein partners).

I do want to point out more generally that the work indicating the limited utility of enzyme fusions as a tool to enhance flux or accelerate reaction cascades has not been shown to apply to constrained or contained systems. Examples of metabolic processes where intermediates are diffusion limited to a small volume because of larger, encapsulated assemblies, clearly benefit from fusion-like architectures. Similarly at the surface of plastics, or at other solid substrates, activities of enzyme fusions may be enhanced because of restricted diffusion of both released soluble products and the enzymes themselves.

Louis Y. P. Luk remarked: Does the melting temperature of the fusion construct increase?

Bruce R. Lichtenstein answered: There are several constructs with varying effects, but in general the effective melting temperature does not change much (sometimes apparently increasing, sometimes apparently decreasing). The only exception to this are the fusions to IsPETase for which the effective melting temperature decreases substantially regardless of the thermal stability of the fusion partner. I would say there is no defined pattern for this property: fully denatured SpyCatcher constructs evidently stabilise LCC(ICCG) but destabilise TfCut1, while the opposite is true for the thermostable SpyCatcher:SpyTag complexes.

Nicholas Turner opened a general discussion of the Concluding Remarks paper by Uwe T. Bornscheuer: What about biosynthetic enzymes – has anyone covered this? Enzymes from secondary metabolism also provide a rich source of new chemistry. These communities are somewhat separate (biocatalysis and biosynthesis). Could you comment on this?

Uwe T. Bornscheuer replied: This is indeed a rich source of novel enzymes, especially if they can be “repurposed” into other synthetic routes outside of their natural function/pathway. This was not covered during this meeting, I assume simply because there was no time (https://doi.org/10.1039/d4fd00127c). In the metabolic engineering field, many interesting and useful examples are indeed covered/published. I believe that many scientists active in biocatalysis are still aware of novel and synthetically useful enzymes discovered in that context.

Joelle N. Pelletier remarked: One purpose of biocat is understanding enzymes, and developing theory of enzyme improvement. A further key goal is having engineered enzymes applied in the real world. Application of ‘advanced’ biocat in industry is fairly recent: the 2010 report of engineering for sitagliptin production by Merck/Codexis was a forerunner. Companies present at this discussion are biocat friendly, yet there is still a barrier to a wider adoption of biocatalysis and enzyme engineering: so much of the chemical production could potentially benefit from biocat is not yet going there. Teaching and the classroom are one of the ways to eventually improve adoption.

Joelle N. Pelletier said: That was a terrific summing up of the discussion. In thinking of landmark advancements, I think that cascade reactions are worth noting – they began being reported near the time the Nobel prize was awarded (50%!) to Frances Arnold.

Uwe T. Bornscheuer responded: I am happy to learn that you liked my “Concluding Remarks”. Indeed, cascade reactions are also very important. Many examples have been published long before the Nobel Prize in 2018, we have for instance reviewed this here.¹ Also the (highly relevant) combination of chemo- and biocatalysis was reviewed by us in 2018 already.²

1 J. Muschiol, C. Peters, N. Oberleitner, M. D. Mihovilovic, U. T. Bornscheuer and F. Rudroff, Cascade catalysis – strategies and challenges en route to preparative synthetic biology, Chem. Commun., 2015, 51, 5798–5811.

2 F. Rudroff, M. D. Mihovilovic, H. Gröger, R. Snajdrova, H. Iding and U. T. Bornscheuer, Opportunities and challenges for combining chemo- and biocatalysis, Nat. Catal., 2018, 1, 12–22.

Stefan Lutz commented: To raise awareness and the visibility of the promise of biocatalysis, is the current focus on engaging with pharmaceutical organizations too narrow? Are we missing an opportunity for greater impact by not actively reaching out to regulators, politicians and environmental agencies as highly complementary major drivers for adoption of new technologies?

Uwe T. Bornscheuer answered: I think this is already taking place. There are several (industrialized) examples of products made using enzymes outside of the pharmaceutical industry with acrylamide (using nitrilhydratases as the best example) but also for emollient esters for cosmetics as produced by Evonik, Germany, for decades. Another emerging area is plastic recycling with PET hydrolysis by engineered enzymes.

As for “reaching out to regulators, politicians and environmental agencies”, this already takes place (but could and should be pushed more), i.e. for the many bioeconomy initiatives in various countries to help the transition from classical chemistry to new methods based on biocatalysis, biotechnology and related fields.

Conflicts of interest

Stefan Lutz declares the following competing financial interest(s): his work has been included in patent application(s) and he is an employee and shareholder of Codexis Inc. There are no other conflicts to declare.

This journal is © The Royal Society of Chemistry 2024