Today it is in the pharmaceutical sector that biocatalysis has flourished in particular, driven by the extraordinary vision of companies such as Merck and Co. Inc. (United States), Pfizer Inc. (United States), GlaxoSmithKline plc (United Kingdom) and Codexis Inc. (United States) amongst others, implementing cascade biocatalysis. Here it is not even selectivity which is the main driver, but rather the reduction in the number of reaction steps which becomes possible when synthesizing complex multi-functional small-molecule APIs using enzymatic cascades. Excellent examples such as the synthesis of Islatravir by Merck illustrate the point well (Huffman et al., 2019). A close to ‘ideal’ synthesis has even been reported using analogous technology by Merck for the synthesis of Molnupiravir, an antiviral treatment for COVID-19 (McIntosh et al., 2021). Most significant though, is that the possibilities are not limited to the modification of known pathways, but instead entirely new synthetic routes can be created, even driving thermodynamically unfavorable reactions in reverse (Abu and Woodley, 2015). Both the Islatravir and Molnupiravir examples illustrate running reactions ‘in reverse’, using the subsequent step in the cascade (or side reaction) to pull the equilibrium in the desired direction. Such new-to-nature syntheses offer enormous possibilities (Nazor et al., 2021) and provide an enormous opportunity over fermentation, even using metabolic engineering, where routes need to be based on modifications to existing pathways, and control becomes difficult.

The second catalytic mode for exploiting enzymes is so-called ‘drop-in’ biocatalysis where an otherwise chemical synthetic sequence is 1) initiated by an enzyme step, 2) interrupted by a biocatalytic step, or else 3) concluded with an enzyme step. A common reason for implementing such single steps is the superb selectivity which can be attained, giving the necessary high yield (Heidlindemann et al., 2014; Rudroff et al., 2018). Since the biocatalytic step must be integrated with existing chemical catalytic steps, it is also essential that sufficient productivity (rate) is offered by the enzymatic reaction, otherwise the biocatalytic step becomes the rate-limiting step. Indeed, enzymatic catalysis can deliver high catalytic rates and productivities of 20–100 g/L.h are readily achievable. Some examples of the high productivity of biocatalysis have been collected in an excellent recent review (Wu et al., 2021). These high productivities have also been the driving force, amongst others, behind the implementation of huge volume biocatalytic processes such as the production of high-fructose corn syrup (HFCS) using glucose isomerase in the 1970s (Basso and Serban, 2019). These rates are many-fold those achievable by fermentation, limited to around 10 g/L.h and more usually less than 5 g/L.h, on account of oxygen limitations and requirements for cell growth. Two-stage fermentation and other well-known tricks in fermentation can also deliver higher productivity cell-based processes, but still not near the requirements for an effective drop-in (Burg et al., 2016). However, this does not mean that fermentation will not have a future role in the biological production of chemicals. For longer pathways where the entire synthesis from sustainable feedstock to product can be taken in the cell, based on the modification of existing pathways then fermentation may indeed prove beneficial (Becker and Wittmann, 2015; Lee et al., 2019). Likewise, in an entirely complementary way for low-priced products where a single step may benefit from high selectivity, then biocatalysis will be the choice. An excellent example is the synthesis of derivatives from 5-hydroxymethyl furfural (Dijkman et al., 2014; Carro et al., 2018; Yuan et al., 2020; Birmingham et al., 2021). Some drop-in reactions may even be integrated with the adjacent chemical step in a ‘one-pot’ reactor which can also bring advantages for the yield (Denard et al., 2013; Dumeignil et al., 2018; Gröger and Hummel, 2020).

Figure 1, schematically illustrates some of the opportunities for cascade and drop-in biocatalysis placed on a productivity/product value map. For low value products, drop-in biocatalysis gives an opportunity to add selectivity at high productivity, matching the rate required at such a scale to the other reactions. For intermediate value products, fermentation is both cheap enough but also at a rate appropriate to match the requirements at that scale. Finally, for the high value products the biocatalytic cascade provides the required selectivity. Rate is not as critical here due to the smaller scale of production.

In the examples discussed here and elsewhere, entirely new synthetic routes can be conceived, limited solely by known enzyme activities. Hence, one very interesting scientific challenge for the future is to develop entirely new enzyme activities, catalyzing new-to-nature reactions. This is an enormously exciting field for academic chemists, discovering new enzymes as well as new synthetic pathways. Pioneering work in this direction has already been started, and holds great promise for the future (Kan et al., 2016; Chen and Arnold, 2020). However, availability of enzymes to catalyze all sorts of new reactions is not in itself sufficient. It is also necessary to understand how to link the enzymes together to create new pathways. This is the growing field of biocatalytic retro-synthesis (Turner and O’Reilly, 2013; De Souza et al., 2017; Hönig et al., 2017). Here the approach has been based on the ground-breaking work on retro-synthesis for conventional chemical reactions (Corey, 1991). Nevertheless, there are around 50 enzyme classes, which is far fewer than offered by conventional catalysts. The key again will likely prove to be protein engineering, enabling tuning and modification of enzyme activities against specific targets, as well as finding entirely new activities to expand the number of enzyme classes. As methods for protein engineering use computational power to an ever-increasing extent (Li et al., 2019; Wu et al., 2019; Yang et al., 2019), and this is integrated with high-throughput automation (Dör et al., 2016) together with computer-assisted retro-synthesis (Finnigan et al., 2021) one can envisage great promise for the start of automated route generation, and subsequent screening and evaluation. In this way, entirely new syntheses will be created (either wholly enzyme-based for cascades or partly enzyme–based for drop-in). Matching reaction conditions, kinetics, thermodynamics and (bio)catalyst stability will all prove fascinating challenges.

When considering a single-step biocatalytic reaction other considerations will prove important, especially with a view to drop-in solutions in the chemical industry. While enzymatic synthesis can often deliver the required productivity (volumetric rate or space-time yield, g product/L reactor.h), the challenge here is how much enzyme is required to achieve this productivity (biocatalyst productivity or specific yield, g product/g biocatalyst). This is a dual problem of ensuring adequate specific activity (through protein engineering) as well as operational stability (maintenance of the activity over operating time). Storage stability is also of importance and needs to be considered particularly at larger scale. Methods to stabilize enzymes during operation include protein engineering (Meng et al., 2020) and immobilization (Basso and Serban, 2019; Boudrant et al., 2020; Sheldon et al., 2021). Such technologies continue to develop, but still there is much to understand about the operational stability of enzymes (Polozzi et al., 2007; Bommarius, 2015). Specifically, it is a challenge to understand enzyme stability under industrial conditions, which are far away from conditions investigated in conventional enzymology (e.g., at high concentrations of substrate and/or product, in multi-phase mixtures, and with fluctuating pH values). Here new tools will be required to assess stability in a rapid and more practical way. A particular problem is that those enzymes most attractive for implementation, and therefore highly stable, present a practical problem of measurement within a reasonable timeframe. Approaches using high temperature ramps are one possibility, but other tools will also be required.

A related scientific challenge here, when the enzymes of interest have already been stabilized, is how to make the best use of this activity. Options include operating in continuous flow reactors or via enzyme recycle and reuse. Both options require retention of the enzyme prior to downstream processing, via a second phase, use of membranes or else via immobilization. This will likely be the third area of scientific interest and growth. Here the issue is less about the development of new tools and more about new methodologies for presenting the reaction components to each other. A first target is the interaction of enzymes with new materials for immobilization (Dos Santos et al., 2015). Modification of activity, stability and selectivity also become possible via the correct selection of immobilization support material (Sheldon, 2019). Opportunities are also possible via the use of magnetic support materials for immobilization where there has been considerable progress recently (Zhang et al., 2020; Matveeva and Bronstein, 2021; Zhou et al., 2021). As material science develops further, many more opportunities will be forthcoming to use immobilization to exploit biocatalysis to the fullest possible extent. The cost of immobilization will always need to be considered but is often offset by the benefits for retention of enzyme in continuous reactors or separation and reuse of the enzymes in multiple batches (Thompson et al., 2019). Recently, co-immobilization of multiple enzymes provide entirely new opportunities for managing enzymatic cascades via spatial as well as temporal control. A second target is the reaction medium. In the 1970s and 1980s significant progress was made in the study of enzymes in non-aqueous environments, usually either anhydrous solvents or aqueous-organic liquid-liquid mixtures. The former enables reactions to be carried out which would be thermodynamically impossible in an aqueous environment (such as esterification), and the latter enable reactions with poorly water-soluble inhibitory substrates or products to be run at high overall loading. In the subsequent decades, a range of other non-conventional media have been tested with enzymes,* (Kourist and Gonzalez-Sabin, 2020) including deep eutectic solvents (Pätzold et al., 2019; Tan and Dou, 2020) and ionic liquids (Cantone et al., 2007; Sheldon, 2021). New ionic liquids, based on renewable materials, giving improvements for sustainability are also being developed (Seitkalieva et al., 2021). Of particular interest is that ionic liquids can be tuned to have particular properties and in this way match not only the requirements of the enzyme and reaction, but also the reactor.