Introduction
The concept of sustainability undoubtedly occupies a prominent position in the vast scientific area, demonstrated by the impressive number of scientific papers that mention this topic in 2024 (24212 vs 4118 in 2014, source: Web of Science). Since the late 90s, the 12 guiding principles coined by Anastas and Warner have been pointing the road towards a more acceptable way to design manufacturing processes in any chemical field (1).
Whereas the optimum relies in preventing waste generation and maximizing the atom economy concept coined by Barry Trost (2), the use of catalyst moves towards this direction, demonstrated by the high number of scientific works done by the community over the year. Accordingly, it is not surprising that, from the vast realm of catalyst toolbox available to the (bio)chemist, the use of enzyme as powerful tools for exploiting the desired chemical reactions occupies a privileged position.
The unique combination of complex machines designed by Nature with the recent improvement driven by the direct evolution to “educate” enzymes to perform reaction not in their scope fosters the field to unprecedent level. Moreover, with the impressive improvement of the biological tools to engineer the enzyme allowed the consistent reduction of the random-mutation protocols and the relevant cost for the consumables needed to perform the design experiments. As a reference, today the price for plasmid production for research purposes goes lower than 100 USD (for 4 ug) and needs less than 1 week to have it ready in hand. With regards to timeframes for mutation screening, a single aminoacid random mutation screening on shake flask can be completed in less than 2 weeks and the recent application of artificial intelligence platforms promises to consistently decrease the number of experiments required to reach the desired target (3). There is vast literature about the topic that interested readers can access, such as the joint Academic and Industrial review coordinated by Prof. Bornscheuer (4). However, despite the massive improvements observed in these years in the field, the application of enzymes as catalyst for the manufacturing of goods of industrial relevance is still not heavily applied. Despite few notable examples, such as the application of PenG acylase to produce 6-aminopenicillanic acid (6-APA) and of invertase to produce syrup at tons scale, the use of enzyme at production scale remains underused due to their intrinsic problems. Among all, the limited reaction space (principally involving hydrolysis and oxidation), the high cost for development and manufacturing of the selected mutant and their stability in the reaction condition. While the scientific community is brilliantly demonstrating the ability to engineer enzymes to perform reactions that are outside their natural scope, enlarging the chemical space available, all these applications still are applied at lab scale, where the enzyme design and production are finalized to screen its ability to exploit the required reaction with the available substrates. On the other side, when the same enzyme has to move at industrial scale, “allosteric” parameters must be considered, such as their stability and catalytic activities at high substrate concentration or when immobilized to a solid support. Furthermore, their expression in the host must be robust and reproducible, in order to guarantee a sustainable and economically viable source of the desired enzyme. When coming to industrial scale, the enzymatic catalysis must compete with the existing classical chemical approaches, equalizing or surpassing their productivities.
1st Case study: biocatalitic approaches in therapeutic peptide
It is thus not surprising that the application of biocatalysis in the pharmaceutical contest become effective with complex targets such as peptides or oligonucleotides, where selective reactivities are needed. On the other side, the recent years are showing an increase of drug candidates belonging to these families approved by FDA (5), with the concurrent needs of industrially transferable biocatalytic approaches for their manufacturing. An exemplary case can be considered the Chemo-Enzymatic Peptide Synthesis (CEPS) approach currently applied in our laboratory. The technology, originally developed by Enzypep BV in the Netherland, is focused on the use of engineered Subtilisin (a serine protease from Bacillus amyloliquefaciens) to selectively ligate peptide fragments of the interested therapeutic target, improving the overall purity of the produced crude peptide (6). In fact, it is well known that the peptide length is among the most critical parameters for the complexity of that manufacturing process, whereas longer sequences have higher probability of incurring into the classical impurities of the Solid Phase Peptide Synthesis (SPPS) (7). The probability exponentially increases with length and in case of “difficult” sequences, which shows high tendency to aggregate or mask the free amino group that must react. Splitting the desired peptide sequence into two (or more) fragments leads to intermediates with higher purity and lower production effort. Furthermore, their assembly can be parallelized, speeding up the manufacturing timeframe and increasing productivity.
This approach is particularly powerful in the current trend of the therapeutic peptides, which is focused on longer sequences (FDA classifies peptides as aminoacid sequences with not more than 40 residues) with unnatural aminoacid added to enhance the physiochemical properties. As a reference example, Tirzepatide from the company Ely Lilly (MonjuaroTM, ZepboundTM) is a 39 aminoacid long Glucagon-like peptide (GLP-1) agonist showing two amino-isobutyrric acid residues (Aib) and a lipidated fatty side chain crucial for the elongation of its bioavailability (weekly injection).
Due to its length and complexity, the Ely Lilly manufacturing strategy adopted splits the molecule into 4 fragments, which are independently synthesized and then sequentially coupled (8). However, the coupling of each fragment through classical chemical approach (typically a carbodiimide and an additive) lead to the epimerization of the adjacent stereocenter though the well-known oxazolone mechanism. Epimers may become cumbersome to purify due to the high similarity with the parent compound, so the coupling site must be carefully selected and the coupling condition finely tuned in order to minimize this side reaction. Furthermore, the chemical coupling requires the use of protected fragments which may be poorly soluble even in polar aprotic solvent like dimethylformamide (DMF).
On the other hand, using an enzyme that allows the selective coupling of unprotected fragments in aqueous solution constitutes a breakthrough in terms of process quality and sustainability. From the combination of initial works done by Wells on the active site (particularly the mutation on Cys221 and Pro225) and Brian on the enzyme stability (removing the Ca2+ dependent domain), the first generations of mutants were created, which showed a reversed activities respect to the parent enzyme (hydrolase) with appreciable synthesis/hydrolysis ratios (S/H). An acyl donor, such as the carboxamide methyl (OCam) ester or the more stable HMBA (hydroxymethylbenzoic) ester, is required to form a thioester complex with the enzyme active site which is subsequentially attacked by the amine nucleophile (peptide acyl acceptor). The coupling is performed under aqueous condition at room temperature with a pH in the range of 7.5-8.5 and stoichiometry can be tuned according to the cost and availability of the fragments selected. The coupling activity of the peptiligase is conserved into six recognition pockets: four on the acyl donor side (S1-S4) and two that recognize the acyl acceptor (S1’ and S2’). Using structure-guided site-evaluation libraries, the scope of peptiligase could be broadened or modulated to selectively and efficiently recognize the selected fragments. The design of the optimal mutant follows a well-established protocol, starting from the “synthesis to hydrolysis” approach, where the catalytic Cys221 of a model peptiligase is retro-mutated to serine to restore its hydrolytic activities. The desired peptide target is then subjected to this enzyme to evaluate its preferred hydrolytic site, which will be also the most preferred coupling site when the cysteine is restored to position 221. After that, a combination of rational design followed by random mutation of selected residues and their expression in bacillus subtilis generates a set of mutants that are screened with a peptide model mimicking the desired sequence, with the final target to select the best candidate (see Figure 1). This sequence of activities, called “DSP cycle”, is repeated until the achievement of the desired performance in terms of efficiency and selectivity. Following this approach, tailored enzymes were developed for the synthesis of several therapeutic peptides, demonstrating the broadness of the approach. The list of the produced peptides ranges from dasiglucagon, glepaglutide, elsiglutide, teriparatide, salmon calcitonin, bivaluridin to the blockbuster GLP-1 agonist liraglutide and Semaglutide (9). The latter products deserve a special mention due to the extraordinary market attractiveness achieved after their approval for body weight management. Liraglutide and Semaglutide show high similarities in the primary structure, both composed by 31 residues and by the presence of a lipidated fatty side chain bound to the lysine in position 20. Consequently, the “synthesis to hydrolysis” approach provided the same coupling site, between Ser11 and Ser12, splitting the molecules into two fragments, namely fragment 1-11 ester and fragment 12-31. However, semaglutide shows also the presence of an Aib residue in position 2,mstrategically placed to reduce the activity of the Dipeptidyl peptidase-4 (DDPIV) and elongate the bioavailability. This difference greatly impacts the development of the tailored peptiligase, since the presence of the Aib function hampers its recognition by the enzyme, avoiding the possible double coupling from two fragment 1-11. On the other side, Liraglutide suffers from the potential unselective coupling, forcing the team to an extensive set of “DSP cycles” (more than 15 generations) to achieve the desired mutant able to obtain good selectivity (Fragment 1-11/Fragment 1-11 coupling <0.1%) with reasonable coupling Yield (80-85 % range). The processes metrics are outlined in Figure 2 and were performed at Kg scale using crude fragments, manufactured by SPPS followed by cleavage from the solid support and not subjected to further purification. CEPS process mass intensities (PMI) can be reduced to 50% with respect to full SPPS approaches thanks to the higher fragment purity/yield and lower downstream effort, leading to a consequent reduction of manufacturing costs. Additional design of the enzymes and their immobilization in solid supports to explore the flow application are currently on-going to reduce the process PMI. Parallelly, the optimization of the Bacillus Subtilis host was required to ensure robust production of the required peptiligases. The host was engineered with 10+ gene deletions to erase sporulation and allow an extracellular excretion of the desired enzyme at multiple grams per liter, easing its purification and isolation.
2nd Case study: biocatalitic approaches in small molecules
The application of the CEPS protocol summarizes the critical aspect that a biocatalytic process must achieve to move to the industrial level: overcoming technical difficulties, improving sustainability and ensuring cost-effectiveness. When more established products are in scope or cheap reagent/catalyst are involved, the development of a competitive biocatalytic route becomes more complex. An example from our laboratories is related to lactulose, a probiotic molecule used as osmotic laxative and as detoxifying agent in pathological frame (e.g. chronic kidney disease). It is presented as 67% syrup or in crystal and it is currently manufactured through a chemical approach based on the Lobry de Bruyn-Van Ekenstein isomerization of lactose catalyzed by bases. According to the manufacturing route selected, the lactose-lactulose isomerization ranges from 12% to >90%. However, when soda (NaOH) is used as the base for the isomerization, the conversion is usually low (<20%) and the purity profile generated is rather complex, which leads to the use of extensive downstream processing to isolate the final product. Lastly, the process foresees the heating of the solution to reflux for 10 minutes, followed by cooling of the solution, leading to an energy expenditure to rapidly heat and cool the lactose solution.
In an effort to improve the reaction performance and product quality profile, we investigated the aid of enzymes to perform this isomerization. It is known from literature that the Cellobiose-2 epimerase (CsCE) enzyme family is able to isomerize lactose to two main isomers, namely lactulose and epilactose (see Figure 3).
However, the rate of isomerization toward epilactose is kinetically faster respect to the lactulose one, leading to the production of the unwanted isomer to content above the permitted specification (not more than 8% according to US Pharmacopoeia).
The 1st part of the work was focused on the screening for the best CsCE variant among those available in terms of kinetics and lactose-epilactose conversion (50°C in PIPES buffer at pH 7.5, lactose conc.: 50 g/L, 2.5% w/w enzyme). From the screening, the 5xCsCE mutant (10) showed the fastest kinetics (about 8 hours to reach the reaction equilibria of 62% (lactulose) – 24% (lactose) – 13% (epilactose). This was partially in accordance with literature except for the formed epilactose, which was found out of specification.
Accordingly, a rational design was done on the crystal structure of a related CsCE available in the Protin Data Bank (PDB), finding that the active site contains 3 catalytic histidine, each of those proximal to a specific functionality of lactose. Interestingly, the histidine in position 247 (in 5xCsCE) is oriented towards the hydrogen potentially involved into the isomerization to epilactose. Thus, despite its reported critical role in the formation of the cis-enediol intermediate required for both isomerization and epimerization (11), a random mutation was performed on this residue. The generated 19 new mutants were tested in the previously mentioned reaction condition, showing to still retain the epimerization activity with an averagely lower tendency to form unwanted epilactose. For example, the H247E mutant showed an excellent result, with a % ratio of 64% (lactulose) – 28% (lactose) – 6% (epilactose), in the same reaction time of the wild type 5xCsCE (12). Furthermore, the overall purity of the obtained lactulose resulted higher vs the chemical approach, since all the unwanted side reactions driven by sodium hydroxide were ruled out.
These results, albeit encouraging, still lead to a product that contains a % of residual lactose out of specification (content not more than 10% according to the EU Pharmacopoeia) plus the buffer salts (if enzyme is immobilized), which requires the use of a downstream protocol to remove them. Furthermore, the reaction condition requires anyhow the use of heating to ensure the full dissolution of lactose, which must be considered in the total production cost. Lastly, the costs to produce the enzyme are still far more expensive when compared to the sodium hydroxide used for the chemical approach (currently at 0.50 Eur/Kg) and forces the enzyme to be immobilized with high recyclability and robustness. When all these topics are taken into account, it is clear that the biocatalytic route requires more efforts to reach the same level of the chemical approach.
Conclusion
The two examples reported are representative of the direction where biocatalysis needs to move. The approaches behind enzyme design should not only be focused on the explication of the desired reaction in aqueous systems but must consider manufacturing efficiency. Biocatalytic process robustness, broadness of application, increase of productivity with the consequent reduction fixed cost impact (e.g. utilities, plant occupancy) are only a few of the parameters to be taken into account during their development, with the final target to introduce them as a routinary option in the chemist toolbox.
References and notes
- Anastas PT, Warner JC. Green Chemistry: Theory and Practice, Oxford University Press, New York, 1998
2. Trost BM. Science 1991, 254, 1471
3. Callaway E. Nature 2024, 634, 525
4. Buller R, Lutz S, Kazlauskas RJ, Snajdrova R, Moore JC, Bornscheuer U.T. Science 2023, 382,899
5. De la Torre, BG, Albericio F. Molecules 2025, 30, 482
6. a. Nuijens T, Toplak A, Schmidt M, Ricci A, Cabri W. Front. Chem. 2019, 7, 829. b. Toplak A.; Ricci A, Cabri W. Sustainability in Tides Chemistry Green Approaches to Oligonucleotides and Oligopeptides Synthesis; RSC London, 2024, London, chapter 6, pp 109-132.
7. Wolf AJ, Ricci A, in Sustainability in Tides Chemistry Green Approaches to Oligonucleotides and Oligopeptides Synthesis, RSC London, 2024, chapter 8, pp. 169-193
8. Frederick MO et al. Org. Proc. Res. Dev. 2021, 25, 1628-1636.
9. Nuijens T, Mateman J, De Visser R, Cabri W, Wallraven K, Wijker E, Ricci A, Toplak A. US2024043897 (2024)
10. Shen Q, Zhang Y, Yang R, Pan S, Dong J, Fan Y, Han L. Food Chem. 2016, 207,60
11. Zhang Y, Zhanga H, Zheng Q. Phys. Chem. Chem. Phys. 2017, 19, 31731
12. Toplak A, Wijma HJ, Cabri W, Ricci A, Janssen DB WO2024208687 (2024)
Figure 1. Overview of the protocol for the design of the optimal peptiligase specific for the interested target.
Figure 2. Process performances of the CEPS approach for Liraglutide and Semaglutide. a Process Mass Intensity (PMI) does not take into consideration the reagents and solvents used to produce protected aminoacids, SPPS reagents and enzymes; b Comparative PMI coming from a full SPPS approach protocol.
Figure 3. Schematic representation of the cellobiose 2-epimerase (CE) catalytic mechanism. Crystal structure of CsCE with a focus into the active site, highlighting the spatial arrangement of the three catalytic histidine (PBD code: 7D5G).
