Bringing Biocatalysis Into the Deuteration Toolbox

Bringing biocatalysis into the deuteration toolbox

J. S. Rowbotham,1 O. Lenz,2 H. A. Reeve1* and K. A. Vincent1*

1Department of Chemistry, University of Oxford, Inorganic Chemistry Laboratory, South Parks Road, Oxford, OX1 3QR, United Kingdom

2Department of Chemistry, Technische Universität Berlin, Strasse des 17. Juni 135, 10623 Berlin, Germany

*Correspondence to: [email protected], [email protected]

Chemicals labelled with the heavy hydrogen isotope deuterium (2H) have long been used in chemical and biochemical mechanistic studies, spectroscopy, and as analytical tracers.1 More recently, demonstration of selectively deuterated drug candidates that exhibit advantageous pharmacological traits has spurred innovations in metal-catalysed

2H insertion at targeted sites,2–5 but asymmetric deuteration remains a key challenge.6–8

Here we demonstrate an easy-to-implement biocatalytic deuteration strategy, achieving

2 high chemo-, enantio- and isotopic selectivity, requiring only H2O (D2O) and unlabelled dihydrogen under ambient conditions. The vast library of enzymes established for

NADH-dependent C=O, C=C, and C=N bond reductions9 have yet to appear in the toolbox of commonly employed 2H-labelling techniques due to requirements for suitable deuterated reducing equivalents.10 By facilitating transfer of deuterium atoms from

2 + H2O solvent to NAD , with H2 gas as a clean reductant, we open up biocatalysis for asymmetric reductive deuteration as part of a synthetic pathway or in late stage functionalisation. We demonstrate enantioselective deuteration via ketone and alkene reductions and reductive amination, as well as exquisite chemo-control for deuteration of compounds with multiple unsaturated sites.

Deuteration has long been used to alter the properties of molecules across a broad range of disciplines, from physical and life sciences, to forensics and environmental monitoring.1,11–13

Indeed, the presence of deuterium can have implications for molecular properties far beyond the augmented mass, including the alteration of spin-characteristics (particularly valuable in

NMR spectroscopy), vibrational states, optical properties, neutron-scattering and even solubility.1 One of the most significant impacts of deuteration is manifested by the so-called deuterium kinetic-isotope effect (DKIE), where cleavage of a C-2H bond can be many times slower than for the C-1H counterpart.13 There has been extensive use of DKIEs to provide insight into the mechanisms of chemical and biological processes.1,11,14 There has been recent interest in exploiting potential therapeutic benefits of the DKIE to alter the pharmacokinetic properties of drug candidates and their associated metabolic profiles.13,15–18 Here, deuterium atoms serve to stabilise compounds against physiological degradation routes, such as racemisation and oxidation, leading to lower dose requirements and potentially fewer toxic side products.13,17,19

The interest in deutero-drugs has increased the demand for catalysts that are capable of

“precision-deuteration”,17 installing deuterium atoms at well-controlled molecular positions.

Hydrogen isotope exchange (HIE) or halogen – deuterium exchange catalysts allow for post-

2 2 2,17,20,21 synthetic deuteration by treating the compound in H2O or under H2 gas.

Developments in these catalysts have greatly improved the regioselectivity of deuteration, with notable examples highlighted in Figure 1A. Significantly, however, the enantioselective installation of deuterium is not well established. Whilst a limited number of catalysts have been demonstrated for asymmetric HIE,6,7 these post-synthetic methodologies risk deteriorating the enantiomeric purity of the compound in question.17 In this work, we demonstrate that NADH-dependent biocatalysis, together with an appropriate cofactor

2 recycling system, can be harnessed for reductive asymmetric deuteration with H2O as deuterium source and H2 as a clean reductant (Figure 1B).

A) State of the art approach to 2H-incorporation Ref.

(2)

(3)

(4)

(5)

B) Asymmetric reductive deuteration (this work) Ref.

this work

Figure 1: Methods for introduction of deuterium into organic molecules and drug candidates. (A) State of the art approaches for carbon-hydrogen or carbon-halogen hydrogen 2 2 isotope exchange; these methods use either H2 or H2O as the deuterium source. (B) Overview 2 of the biocatalytic reductive deuteration strategy demonstrated in this work, using H2O as the deuterium source and H2 as a clean reductant.

In the field of asymmetric synthesis, biocatalysis has profoundly expanded the toolbox available to organic chemists, with enzymatic transformations offering near-perfect selectivity and mild operating conditions.9 Large scale enzyme evolution has dramatically expanded the scope for biocatalysis, with Sitagliptin offering a notable case study in

3 biocatalytic pharmaceutical synthesis.22 However, despite the growing importance of NADH- dependent reductase enzymes in chiral fine chemical synthesis,23 significant hurdles have impeded their use in deuteration chemistry due to the requirement for deuterated, reduced cofactor, [4-2H]-NAD(P)H which must be continually regenerated in situ (Figure 2A).

Laboratory-scale demonstrations have relied on a super-stoichiometric supply of a sacrificial deuterated reductant, [2H]-ethanol, [2H]-isopropanol, [2H]-glucose, or [2H]-formate, for example, with the appropriate dehydrogenase enzyme (Figure 2B).10,24 Such compounds are labour-intensive to prepare, or expensive.

In order to overcome these problems, we sought to develop a catalyst that can recycle [4-2H]-

+ 2 NADH from NAD using (unlabelled) H2 gas as a clean reductant and H2O as the deuterium source. We have previously demonstrated H2-driven NADH recycling using hydrogenase and

NAD+ reductase (both produced in-house, see SI Section S2.1) co-immobilised on a carbon

25,26 support. This heterogeneous biocatalyst system couples H2 oxidation by a nickel-iron hydrogenase to the reduction of NAD+ to NADH by a flavin-containing NAD+ reductase

26 + moiety via electron transfer through the carbon. Since the H2 oxidation and NAD reduction

2 catalytic sites are spatially separated, we hypothesised that operation of this system in H2O

+ instead of H2O, would enable the NAD reductase to transfer deuterium ions from solution to

NAD+, thereby forming [4-2H]-NADH (Figure 2C). This suggested the possibility of a new route to recycling the deuterated cofactor, suitable for application in the biocatalysed synthesis of selectively deuterated chemicals, in which the heavy atom label is provided

2 solely by the solvent, H2O.

A) reagent 2H-labelled product

alkene alkane imine NADH amine carbonyl dependent alcohol enzyme

Cofactor turnover

Regenerating deuterated [4-2H]-NADH NAD+ cofactor

2 2 B) Current methods for regenerating [4- H]-NADH C) H2-driven regeneration of [4- H]-NADH (this work)

NAD+ [4-2H]-NADH NAD+ [4-2H]-NADH

2H O Sacrificial Carbon waste 2 2 [ H]-reductant NAD+ product Heterogeneous biocatalyst dependent dehydrogenase H 2 2H+ + NAD+ [4-2H]-NADH [2H]-glucose gluconolactone 2H+ 2 [ H]-formate CO2 [2H]-isopropanol acetone 2e- Carbon [2H]-ethanol acetaldehyde Hydrogenase NAD+ reductase

Reductant: Deuterium source: Reductant: Deuterium source:

2 2 Sacrificial [ H]-reductant H2 H2O

Figure 2: Biocatalytic reductive deuteration via recycling of the deuterium labelled cofactor, [4-2H]-NADH. (A) Biocatalytic reductive deuteration requires a supply of [4-2H]-NADH. (B) Conventional [4-2H]-NADH recycling strategies are driven by a sacrificial deuterated reductant. (C) Schematic representation of the heterogeneous biocatalytic system demonstrated 2 2 in this work for H2-driven generation of the labelled cofactor [4- H]-NADH, with H2O supplying the deuterium atoms. For complete structures of the cofactors, see SI Figure S1 and for details of heterogeneous biocatalyst, see SI Figure S2.

To test our hypothesis that the heterogeneous biocatalyst system would form deuterated

+ 2 NADH, we supplied it with 4 mM NAD in H2O under an atmosphere of H2 gas (see SI

Section S3.1 for details). Subsequent analysis by HPLC and 1H NMR (SI Figure S3), 5 confirmed that only the biologically-relevant [4]-NADH was generated (at 97% conversion), with the anticipated incorporation of 2H at the [4]-position on the nicotinamide ring ([4-2H]-

NADH).26

Following this successful demonstration, we then sought to use H2-dependent biocatalysis to recycle [4-2H]-NADH to drive a range of NADH-dependent reductases. The selected reductases were all from commercial enzyme suppliers, and were chosen to cover three of the most prominent classes of reductive biotransformations: carbonyl-reduction, carbonyl reductive amination, and alkene-reduction. Through operating the heterogeneous biocatalyst

2 2 system with the relevant reductase in H2O, under mild reaction conditions (p H 8.0, 1 bar

o H2, 20 C), we were able to demonstrate reactions to form a diverse array of molecules bearing deuterium in carefully-controlled positions (Figure 3, full reaction schemes and product characterisation in SI Section S3). A steady flow of H2 across the reaction headspace was found to be sufficient to drive the reactions, but conventional balloons and sealed pressure vessels were also shown to be suitable alternative setups. In all cases, the chemo-, stereo- and regio-selectivity known for the enzymatic reactions in natural abundance H2O was fully retained upon translation to the deuteration conditions.

Figure 3: Asymmetric biocatalytic reductive deuteration to generate a library of chiral 2 products via H2-driven recycling of [4- H]-NADH. Typical reaction conditions: 0.5 mL, 5 mM + 2 2 2 2 reagent, 0.5 mM NAD , H2O (98 %), Tris- HCl (p H 8.0, 100 mM), 1-5 vol.% H6-DMSO. 16 o hours, 20 C, 1 bar H2, shaking at 500 rpm. Catalysts: Excess NADH-dependent reductase, 400 mg carbon (160 pmol hydrogenase, 260 pmol NAD+ reductase). For reductive amination reactions: 25 mM NH4Cl. Extensive experimental and analytical detail is provided in SI Section S3.

2 The reductive deuteration of a suite of reagents, via H2-driven recycling of [4- H]-NADH, is achieved by combining the heterogeneous biocatalytic cofactor recycling system with an appropriate reductase, either co-immobilised or in solution. The products in Figure 3 demonstrate the versatility of this novel approach. For carbonyl reduction, for example, it was possible to generate either [1S-2H]- or [1R-2H]-1-phenylethanol (1a and 1b) from acetophenone, with ≥ 93 % isotopic selectivity and ≥ 99 % enantioselectivity, by choosing an

(S)- or (R)-selective alcohol dehydrogenase (keto reductase), respectively. Pyruvate could be converted to [2L-2H]-lactate (2) by the action of L-lactate dehydrogenase, or to [2L-2H]- alanine (3) by the action of L-alanine dehydrogenase in the presence of an amine source

+ (NH4 ). Reductive amination was also carried out on phenylpyruvate using L-phenylalanine

2 dehydrogenase; prior exchange of the β-protons with H2O yields a triply deuterated product:

2 2 [2- H2, 3L- H]-phenylalanine (4).

Asymmetric alkene reduction was demonstrated by the use of an ‘ene’ reductase on (5R)- and

(5S)-carvone reagents, yielding trans-(2R, 5R)- and cis-(2R, 5S)-dideuterocarvone (5a and

5b), respectively. Here, the enzyme generates (R)-stereochemistry at the prochiral 2-position of the two parent carvones, with equally high de in each case. 1H NMR spectroscopic analysis of product 5a (see SI Section S3.6.3), demonstrates that the deuterium atom at the 3- position of the dideuterocarvone is also added asymmetrically, giving (R)-stereochemistry at this position, in accordance with the established mechanism of ene reductases in reducing double bonds in an anti-manner.27,28 The ability to introduce isotopically-induced chirality by selective biocatalytic deuteration may have further important implications in drug design and generation of deuterated chemicals for analytical applications.29

Finally, we demonstrate the potential for chemo-selective deuteration of another molecule containing both alkene and carbonyl functionalities, αβ-unsaturated cinnamaldehyde.

Reduction of cinnamaldehyde by means of conventional heterogeneous catalysis can suffer

8 from poor chemoselectivity, yielding a mixture of alcohol and alkane products.30 However, by utilising an ene-reductase and alcohol-dehydrogenase in isolation or sequentially, the singly-deuterated cinnamyl alcohol (6a), doubly-deuterated hydrocinnamaldehyde (6b), or multiply-deuterated 3-phenyl-1-propanol (6c) can all be prepared.

In summary, we have demonstrated a versatile approach to asymmetric reductive deuteration that exploits the high chemo-, regio-, and stereo-selectivity of NADH-dependent reductases to add deuterium at specific positions on organic molecules. By using a site-separated heterogeneous biocatalyst, a simple reductant (H2) can be coupled with a cheap and abundant

2 2 source of deuterium ( H2O) to provide a widely applicable [4- H]-NADH regeneration system. Importantly, this system removes the requirement to supply specialist [2H]-labelled sacrificial carbon-based reductants in order to use enzymes in deuteration reactions. Instead, the simple reaction conditions employed here are competitive with the most advanced HIE catalysts. The high %2H incorporation achieved across diverse functional group space indicates the possibility that this approach could be used to generate libraries of well-defined

[2H]-labelled chiral building blocks for drug synthesis and analytical applications. The syntheses for a number of pharmaceutical molecules already incorporate asymmetric biocatalytic reductions which could be extended for asymmetric deuteration via the method we demonstrate here. Furthermore, the approach we have developed would lend itself to late stage functionalisation by reductive introduction of chirality and a deuterium label simultaneously to a complex molecule without compromising product purity. We anticipate that this approach will enable NADH-dependent reductases to be implemented as a powerful complementary addition to the growing toolbox of deuterium-incorporation techniques.

Acknowledgments

The research of K.A.V., H.A.R., and J.S.R. is supported by Engineering and Physical Sciences Research Council (EPSRC) IB Catalyst award EP/N013514/1. In addition, J.S.R. is 9 grateful to Linacre College, Oxford for the support of an EPA Cephalosporin Junior Research Fellowship. O.L thanks the Cluster of Excellence “Unifying Concepts in Catalysis” (EXC 314‐2, Deutsche Forschungsgemeinschaft) for support. Ailun Huang and Thomas Lonsdale are acknowledged for laboratory assistance in initial experiments. We thank Janina Preissler (TU Berlin) and Dr Miguel Ramirez (Univ. Oxford) for assistance with the isolation and purification of the NAD+ reductase and E. Nomerotskaia (Univ. Oxford) for preparing E. coli hydrogenase 1. Dr Beatriz Dominguez (Johnson Matthey) is gratefully acknowledged for providing (S)-ADH, (R)-ADH and ene reductase enzymes. Dr James Wickens (Univ. Oxford) is thanked for helping to perform analysis by GC-MS.

Author contributions

J.S.R., O.L., H.A.R. and K.A.V. all helped to design experiments; J.S.R performed experiments; H.A.R and J.S.R. analysed data; all authors contributed to writing the paper.

Competing interests

A patent application detailing some of this research has been filed through Oxford University Innovation (Feb 2018).

Additional information

All data are provided in the main text or the supplementary information (SI).

References

1. Atzrodt, J., Derdau, V., Kerr, W. J. & Reid, M. Deuterium- and Tritium-Labelled Compounds: Applications in the Life Sciences. Angew. Chemie Int. Ed. 57, 1758–1784 (2018).

2. Nilsson, G. N. & Kerr, W. J. The development and use of novel iridium complexes as catalysts for ortho-directed hydrogen isotope exchange reactions. J. Label. Compd. Radiopharm. 53, 662–667 (2010).

3. Pony Yu, R., Hesk, D., Rivera, N., Pelczer, I. & Chirik, P. J. Iron-catalysed tritiation of pharmaceuticals. Nature 529, 195–199 (2016).

4. Loh, Y. Y. et al. Photoredox-catalyzed deuteration and tritiation of pharmaceutical compounds. Science (80-. ). 358, 1182–1187 (2017).

5. Liu, C. et al. Controllable deuteration of halogenated compounds by photocatalytic D2O splitting. Nat. Commun. 9, 80 (2018).

6. Palmer, W. N. & Chirik, P. J. Cobalt-Catalyzed Stereoretentive Hydrogen Isotope Exchange of C(sp3)–H Bonds. ACS Catal. 7, 5674–5678 (2017).

7. Hale, L. V. A. & Szymczak, N. K. Stereoretentive Deuteration of α-Chiral Amines

with D2O. J. Am. Chem. Soc. 138, 13489–13492 (2016).

8. Taglang, C. et al. Enantiospecific C H Activation Using Ruthenium Nanocatalysts. Angew. Chemie Int. Ed. 54, 10474–10477 (2015).

9. Turner, N. J. & Humphreys, L. Biocatalysis in Organic Synthesis: The Retrosynthesis Approach. (RSC, 2018).

10. Wong, C. H. & Whitesides, G. M. Enzyme-catalyzed organic synthesis: regeneration of deuterated nicotinamide cofactors for use in large-scale enzymatic synthesis of deuterated substances. J. Am. Chem. Soc. 105, 5012–5014 (1983).

11. Yang, J. Deuterium: Discovery and Applications in Organic Chemistry. Deuterium: Discovery and Applications in Organic Chemistry 1–116 (2016). doi:10.1016/C2016- 0-00652-6

12. Hanson, J. R. in The Organic Chemistry of Isotopic Labelling (ed. Hanson, J. R.) 40– 61 (RSC, 2011).

13. Gant, T. G. Using Deuterium in Drug Discovery: Leaving the Label in the Drug. J. Med. Chem. 57, 3595–3611 (2014).

14. Pudney, C. R., Hay, S. & Scrutton, N. S. Practical Aspects on the Use of Kinetic Isotope Effects as Probes of Flavoprotein Enzyme Mechanisms. Methods Mol. Biol. 1146, 161–175 (2014).

15. Liu, J. F. et al. A Decade of Deuteration in Medicinal Chemistry. Annu. Rep. Med. Chem. 50, 519–542 (2017).

16. Foster, A. B. Deuterium isotope effects in studies of drug metabolism. Trends Pharmacol. Sci. 5, 524–527 (1984).

17. Pirali, T., Serafini, M., Cargnin, S. & Genazzani, A. A. Applications of Deuterium in

Medicinal Chemistry. J. Med. Chem. acs.jmedchem.8b01808 (2019). doi:10.1021/acs.jmedchem.8b01808

18. Schmidt, C. First deuterated drug approved. Nat. Biotechnol. 35, 493–494 (2017).

19. Davey, S. Relief from racemization. Nat. Chem. 7, 368–368 (2015).

20. Atzrodt, J., Derdau, V., Kerr, W. J. & Reid, M. C−H Functionalisation for Hydrogen Isotope Exchange. Angew. Chemie Int. Ed. 57, 3022–3047 (2018).

21. Atzrodt, J., Derdau, V., Fey, T. & Zimmermann, J. The Renaissance of H/D Exchange. Angew. Chemie Int. Ed. 46, 7744–7765 (2007).

22. Savile, C. K. et al. Biocatalytic Asymmetric Synthesis of Chiral Amines from Ketones Applied to Sitagliptin Manufacture. Science (80-. ). 329, 305–309 (2010).

23. Prier, C. K. & Kosjek, B. Recent preparative applications of redox enzymes. Curr. Opin. Chem. Biol. 49, 105–112 (2019).

24. Yahashiri, A., Sen, A. & Kohen, A. Microscale synthesis and kinetic isotope effect analysis of (4R)-[Ad-14C,4-2H] NADPH and (4R)-[Ad-3H,4-2H] NADPH. J. Label. Compd. Radiopharm. 52, 463–466 (2009).

25. Reeve, H. A., Lauterbach, L., Ash, P. A., Lenz, O. & Vincent, K. A. A modular system for regeneration of NAD cofactors using graphite particles modified with hydrogenase and diaphorase moieties. Chem. Commun. 48, 1589–1591 (2012).

26. Reeve, H. A., Lauterbach, L., Lenz, O. & Vincent, K. A. Enzyme-Modified Particles for Selective Biocatalytic Hydrogenation by Hydrogen-Driven NADH Recycling. ChemCatChem 7, 3480–3487 (2015).

27. Toogood, H. S. & Scrutton, N. S. New developments in ‘ene’-reductase catalysed biological hydrogenations. Curr. Opin. Chem. Biol. 19, 107–115 (2014).

28. Toogood, H. S., Gardiner, J. M. & Scrutton, N. S. Biocatalytic Reductions and Chemical Versatility of the Old Yellow Enzyme Family of Flavoprotein Oxidoreductases. ChemCatChem 2, 892–914 (2010).

29. Sato, I., Omiya, D., Saito, T. & Soai, K. Highly Enantioselective Synthesis Induced by Chiral Primary Alcohols Due to Deuterium Substitution. J. Am. Chem. Soc. 122, 11739–11740 (2000).

30. Durndell, L. J. et al. Selectivity control in Pt-catalyzed cinnamaldehyde

12 hydrogenation. Sci. Rep. 5, 9425 (2015).

Supplementary Information Bringing biocatalysis into the deuteration toolbox J. S. Rowbotham,1 O. Lenz,2 H. A. Reeve1* and K. A. Vincent1*

1 Department of Chemistry, University of Oxford, Inorganic Chemistry Laboratory, South Parks Road, Oxford, OX1 3QR, United Kingdom

2 Department of Chemistry, Technische Universität Berlin, Strasse des 17. Juni 135, 10623 Berlin, Germany

*Correspondence to: [email protected], [email protected]

Table of contents

S1. List of abbreviations ...... 3 S2. Materials and methods ...... 4 S2.1 Solvents, reagents and catalysts...... 4 S2.2 General procedures for conducting and analysing 2H-labelling reactions ...... 5 S3. Experimental protocols and results ...... 8 S3.1 Conversion and characterisation of NAD+ to [4S-2H]-NADH ...... 8 S3.2 Conversion of AcPh to [1S-2H]-1-PhEtOH (1a) or [1R-2H]-1-PhEtOH (1b) ...... 10 S3.2.1 Experimental conditions ...... 10 S3.2.2 Analytical procedures ...... 10 S3.2.3 Results ...... 11 S3.3 Conversion of pyruvate to [2L-2H]-lactate (2) ...... 13 S3.3.1 Experimental conditions ...... 13 S3.3.2 Analytical procedures ...... 13 S3.3.3 Results ...... 15 S3.4 Conversion of pyruvate to [2L-2H]-alanine (3) ...... 17 S3.4.1 Experimental conditions ...... 17 S3.4.2 Analytical procedures ...... 17 S3.4.3 Results ...... 19 2 2 S3.5 Conversion of phenylpyruvate to [2- H2, 3L- H]-phenylalanine (4) ...... 21 S3.5.1 Experimental conditions ...... 21

S3.5.2 Analytical procedures ...... 21 S3.5.3 Results ...... 23 2 S3.6 Conversion of carvones to [2,3- H2]-dihydrocarvones (5a, 5b) ...... 25 S3.6.1 Experimental procedures ...... 25 S3.6.2 Analytical procedures ...... 26 2 S3.6.3 Results A: [2, 3- H2]-(2R, 5R)-dihydrocarvone (5a) ...... 27 2 S3.6.4 Results B: [2, 3- H2]-(2R, 5S)-dihydrocarvone (5b) ...... 30 S3.7 Selective biocatalytic deuteration of cinnamaldehyde (6a, 6b, 6c) ...... 33 S3.7.1 Experimental conditions ...... 33 S3.7.2 Analytical procedures ...... 33 S3.7.3 Results A: [1-2H]-cinnamyl alcohol (6a) ...... 35 2 S3.7.4 Results B: [2, 3- H2]-hydrocinnamaldehyde (6b) ...... 37 2 S3.7.5 Results C: [1, 2, 3- H3.5]-3-phenyl-1-propanol (6c)...... 39 S4. References ...... 42

S1. List of abbreviations

NAD+ Nicotinamide adenine dinucleotide (oxidised form)

[4-2H]-NAD+ Nicotinamide adenine dinucleotide (oxidised form), deuterated at 4-position of nicotinamide ring

NADH Nicotinamide adenine dinucleotide (reduced form)

[4S-2H]-NADH Nicotinamide adenine dinucleotide (reduced form), singly deuterated at 4-position of nicotinamide ring. (S)-form.

2 [4- H2]-NADH Nicotinamide adenine dinucleotide (reduced form), doubly deuterated at 4-position of nicotinamide ring.

A B

NAD+ R

NADH

Figure S 1: Structure of the nicotinamide cofactors (A) in full (oxidised only) and (B) abbreviated (oxidised and reduced).

S2. Materials and methods S2.1 Solvents, reagents and catalysts

General reagents: General reagents and buffer salts (Sigma Aldrich), NAD+ and NADH (Prozomix), and carbon black particles (Black Pearls 2000, BP2000, Cabot Corporation), were all used as received without further purification. All non-deuterated solutions were 2 prepared with MilliQ water (Millipore, 18 MΩcm), and deuterated solutions with H2O (99.98 %, Sigma Aldrich). All solvents were deoxygenated by sparging with dry N2 for 60 minutes prior to use. All other reagents and solvents were purchased from Sigma Aldrich and used as received, unless specified otherwise.

2 2 Deuterated buffers: ( H5)-Tris- HCl was prepared by dissolving the required amount of 2 Trizma® base in H2O and then evaporating to dryness. After repeating twice more, the pD (p2H) of the Tris solution was adjusted to 8.0 by the addition of small aliquots of 2HCl (3.0 2 M) prior to deoxygenation by sparging with dry N2. Solutions of fully deuterated ( H11)-Tris- 2HCl were also used for analytical purposes, and were prepared in a similar manner. 1H NMR 2 spectroscopy indicated that the final % H2O of deuterated buffer solutions was not below 99 1 mol.% (unless specifically diluted with H2O).

Enzymes: The hydrogenase (Escherichia coli hydrogenase 1, molecular weight 100 kDa),1 and NAD+ reductase (I64A variant of the soluble hydrogenase from Ralstonia eutropha, in which the inherent hydrogenase moiety is inactivated, molecular weight 170 kDa)2 were isolated and purified following published protocols. Commercial samples of alcohol dehydrogenases (R)-ADH (ADH101) and (S)-ADH (ADH105), and an ene-reductase (ENE101) were obtained from Johnson Matthey Catalysis and Chiral Technologies. L- phenylalanine dehydrogenase (from Sporosarcina sp.), L-alanine dehydrogenase (recombinant from E. coli) and L-lactate dehydrogenase (from rabbit muscle) were all purchased from Sigma Aldrich. All commercial enzymes were received from the vendors in their lyophilised forms and used without further purification.

Preparation of the heterogeneous biocatalyst (bio system) for H2-driven cofactor recycling: All catalysts were prepared in a glovebox (Glove Box Technology) under a protective N2 atmosphere (O2 < 0.1 ppm), and in deoxygenated Tris-HCl buffer (100 mM, pH 8.0). Catalysts were not originally prepared in deuterated conditions but, when required, were exchanged into the required solution by centrifugation (10, 000 × g, 5 mins), removing the 2 2 supernatant, and re-suspending in an equal volume of ( H5)-Tris- HCl (100 mM, pD 8.0).

Heterogeneous biocatalytic system (bio system) x + H2O x H2 2H [4- H]-NADH NAD+ xH+

2e- Particle Hydrogenase NAD+ reductase

Figure S 2: Schematic representation of the heterogeneous biocatalyst system (bio system) for H2-driven cofactor recycling, where X = 1 or 2. The hydrogenase has a nickel-iron catalytic site, and the NAD+ reductase has a flavin mononucleotide active site, and both enzymes possess a relay chain of iron-sulfur clusters to allow rapid electron transfer between the catalytic site and the surface of the protein.

The heterogeneous catalyst system was prepared by sonication of a 20 mg/mL suspension of BP2000 carbon black in Tris-HCl buffer for 5 × 15 minutes (with agitation of the solution in between). Equal volumes of hydrogenase (4 mg/mL) and NAD+ reductase (1.4 mg/mL) were then pre-mixed together, and added immediately to an aliquot of sonicated carbon. After standing at 4 ˚C for 60 minutes, the solid catalyst was then removed from the preparatory solution by centrifugation (10, 000 × g, 5 mins) before being re-suspended in deuterated or non-deuterated buffer as required. Enzyme loadings: hydrogenase loading of 39 pmoles per 100 µg of carbon, NAD+ reductase loading of 65 pmoles per 100 µg of carbon. The disappearance of the brown hue of the starting enzymatic solution, and the near-complete drop in detectable solution enzyme activity (as determined by standard benzyl viologen assays)3 were taken as signs that virtually all of the enzymes were immobilized. If an additional enzyme (namely an NADH-dependent oxidoreductase) was to be co- immobilised on the particles, this was done after the particles had already been coated with hydrogenase and NAD+ reductase.

S2.2 General procedures for conducting and analysing 2H-labelling reactions

Typical reaction conditions: All reactions were set up in a glovebox under a protective N2 atmosphere (O2 < 0.1 ppm) and were conducted on a 500 μL scale in sealed 0.5 mL micro- centrifuge tubes (Eppendorf) punctured with five holes in the lid (Ø 1.0 mm). The hydrogenase and NAD+ reductase modified carbon particles (heterogeneous biocatayltic system) were added at a loading of 100 - 400 μg carbon mL-1. Reaction solutions were presaturated with H2 gas and the cofactor dissolved prior to the addition of the catalyst, to ensure that both hydrogenase and NAD+ reductase enzymes were exposed to their respective substrates simultaneously. In order to help dissolve organic compounds 2 ( H6)-dimethylsulfoxide was included in the reactions as required (1-5 vol.%). The punctured tubes were then transferred to a specially adapted shaker plate, which allowed for a steady flow of H2 across the headspace. Alternatively, if a pressurized H2-atomsphere

S5 was required, a tinyclave steel pressure vessel (Buchi) was used, and rocked back and forth at 15 rpm whilst the reactions took place. In the case of volatile substrates, the H2-saturated reaction solution was held in a sealed tube with a balloon to maintain the atmosphere. Following the reaction, solid catalyst was removed by two rounds of centrifugation (10, 000 × g, 5 mins), and the solution was analysed by the methods described below. Conversions were calculated based upon the molar ratio of product(s) to substrate. For each reaction, a control experiment in the absence of catalyst was prepared to ensure there was no background loss of reactant. Overview of techniques for analyzing 2H-labelling reactions: A suite of analytical techniques was employed in order to ascertain the amount of desired deuterated product that been formed in each reaction.  The extent of conversion was calculated by one or more of the following: 1H NMR spectroscopy, GC-FID, HPLC, UV-vis spectroscopy.  For chiral products, enantiomeric excess (ee) was determined by either chiral HPLC or chiral GC-FID.  The %2H incorporation was determined from 1H NMR spectroscopic analysis (from the loss of signal intensity and collapse of coupling associated with incorporation of a 2H unit) and/or GC-MS (from the changes in m/z value associated with installation of the heavy 2H unit).

The specific details of such arguments are explained for each individual case in SM Section S.3.1 (for cofactors) and SM Section S.3.2-S.3.7 (for all other reactions). The experimental procedures required to perform each of the analytical techniques were as follows: a. Procedure for analyzing reactions using UV-vis spectroscopy Reactions which resulted in changes in the relative concentration of NAD+/NADH could be followed be UV-vis spectroscopy. Firstly, solid particulates were removed from the solution 1 by centrifugation (as described above) and the sample diluted in MilliQ ( H2O) water so that the cofactors were in the range 0.1 – 0.2 mM. A background spectrum of pure MilliQ was subtracted from that acquired for the sample. Measurements were made in a quartz cuvette (path length 1 cm, Hellma) on a Cary 60 UV/Vis spectrophotometer (Agilent). b. General procedure for analysing reactions by 1H NMR spectroscopy Following removal of the catalyst particles by centrifugation, 450 μL of the sample solution was transferred to a Norell® SelectSeries™ 5 mm 400 MHz sample tube. A further 50 μL of 2 H2O was added to make up the volume to 500 μL and, in the case of non-deuterated solutions, for field locking purposes. 1H NMR spectroscopy was carried out on either a Bruker Avance III HD nanobay (400 MHz) or Bruker Avance III (500 MHz), with samples for direct comparison always being run on the same machine. Both machines were equipped with 5 mm z-gradient broadband multinuclear probes. Spectra were acquired according to the parameters in Table S.1. In the first instance, the Bruker proc_1d or proc_1dakps processing algorithms were applied, followed by additional manual re-phasing where necessary. A multipoint baseline correction was also applied across

S6 the entirety of the spectral window and a line broadening corresponding to between 0.3 – 1.0 Hz was applied to each spectrum to improve the S/N ratio. Signals were referenced against an 2 appropriate signal: acetone (δ = 2.22 ppm), ( H5)-dimethylsulfoxide (δ = 2.71), or the peak arising from the Tris buffer (δ = 3.70 ppm at pH 8.0), all of which were referenced originally to 4,4-dimethyl-4-silapentane-1-sulfonic acid (δ = 0.00 ppm). Integration of peaks arising from substrate and product could be used to judge the degree of conversion in cases where intensity was not diminished from 2H-incorporation (such as aromatic protons). Similarly, comparison of the intensity of signals arising from the cofactor, buffer, and/or co-solvent could all be used to verify that a mass balance had been achieved in the conversion of substrate to product.

Table S 1: Relevant 1H NMR spectroscopic parameters Machine Avance III - 400 Avance III - 500 Experiment 1H 1H RF pulse energy (MHz) 400.13 499.9 Temperature (K) 298 ± 2 298 ± 2 Number of scans As required As required Pulse width (μs) 14.0 10.3 Spectral width (Hz) 8000 8000 Acquisition time (s) 4.09 2.04 Relaxation delay (s) 1.00 2.00 c. General procedures for analysing reaction products by Gas Chromatography (GC) GC-FID was carried out on a ThermoScientific Trace 1310. GC-MS was carried out on an Agilent 7890B GC coupled to an Agilent 7200 Accurate Mass Q-ToF MS operating under EI mode. For both machines, injections were carried out by means of a robotic autosampler. The relevant columns and chromatography conditions are stated as required. d. General procedures for analysing reactions by HPLC All HPLC was conducted on a Shimadzu UFLC LC-20AD Prominence liquid chromatograph equipped with a dual wavelength UV-spectrophotometric detector and a robotic autosampler. MilliQ water and HPLC grade solvents were used throughout. The relevant columns and chromatography conditions are stated as required.

S3. Experimental protocols and results S3.1 Conversion and characterisation of NAD+ to [4S-2H]-NADH

Experimental conditions A sample of the heterogeneous biocatalytic system (100 μg), prepared according to SM 2 2 2 Section S2, was added to 500 μL of [ H5]-Tris HCl (100 mM, p H 8.0) containing 4 mM + 1 NAD , which was presaturated with H2 gas. The reaction solution was sealed under 2 bar 1 H2 for 16 hours, whilst rocking at 30 rpm. Following removal of the solid particulates by centrifugation, the reaction was analysed 1H NMR and UV-vis spectroscopies. Characterisation The spectra of nicotinamide cofactors and their deuterated analogues were assigned according to well established literature arguments4,5 using the peak(s) between 2.64 – 2.84 ppm (corresponding to the 4-position of the dihydronicotinamide ring), Figure S 3. The spectrum of the NADH standard shows a multiplet with expected splitting. The spectrum of the cofactor generated shows only a single peak in this region, demonstrating the introduction of a deuterium atom. The ratio of NAD+ to NADH (and therefore extent of conversion) could be determined from the UV-vis spectra by measuring the ratio A260 nm : A340 nm, as described previously.6 Analysis by 1H NMR spectroscopy and UV-spectrophotometry together, indicated complete conversion of the cofactor to generate only [4-2H]-NADH. By comparison of the 1H NMR spectra to the literature, it can be deduced that the [4S-2H]-form of the reduced cofactor is generated (observed by the peak position, and the absence of 3.5 Hz splitting).

1H NMR spectra UV-vis spectra 260 nm 340 nm

17.5 Hz 17.5 Hz

3.5 Hz

300 400 500 Wavelength / nm

Figure S 3: 1H NMR and UV-vis spectra of cofactor standards and the cofactor product generated using 2 1 the heterogeneous biocatalytic system operating in H2O. Only the diagnostic regions of the H NMR spectra 2 2 are shown ( H2O, p H 8.0, 400 MHz, 293 K).

S3.2 Conversion of AcPh to [1S-2H]-1-PhEtOH (1a) or [1R-2H]-1-PhEtOH (1b)

S3.2.1 Experimental conditions 2 2 2 Reaction was conducted in 0.5 mL of ( H5)-Tris- HCl (100 mM, p H 8.0) containing 5 mM 2 + acetophenone (AcPh), 2 vol.% [ H6]-DMSO, 0.5 mM NAD , and 500 μg of (S)- or (R)- alcohol dehydrogenase (ADH) in solution. The mixture was presaturated with H2 gas prior to addition of the heterogeneous bio system at a loading of 400 μg. The reactions were shaken at 500 rpm under a steady flow of H2 for 16 hours. S3.2.2 Analytical procedures Chiral HPLC: The ee of the 1-phenylethanol (1-PhEtOH) product was determined by chiral- phase HPLC, by comparison of the reaction mixture with relevant commercial standards. Reactions containing AcPh/PhEtOH were extracted with a 2 × volume of heptane:2-propanol (99:1 vol/vol), and then centrifuged at 18 800 × g for 5 minutes before being transferred to glass vials for HPLC according to method shown below. The results are depicted in Figure S 4(a). Chiral HPLC for resolution of 1-phenyethanol stereoisomers Column: Chiralpak IA column (15 cm × 4.6 mm, 5 μm particle size) with a 20 × 2.1 mm guard Buffer: heptane : 2-propanol (99:1 vol/vol) Column temperature = 40 ˚C Flow rate = 1 mL/min (isocratic) Injection volume = 10 μL Detection = 210 nm

1H NMR spectroscopy: After removal of the catalyst, the unmodified reaction solutions were analysed by 1H NMR spectroscopy at 500 MHz and 298 K, with the results being shown in Figure S 4(b).

GC-MS: Following analysis by chiral HPLC (see above) the extracted reaction solutions were analysed by GC-MS according to the method shown below. A commercial sample of 1-PhEtOH of natural isotopic abundance was derivatised and analysed in an identical manner for comparison, with the results depicted in Figure S 4(c).

S10

GC-MS method for determining the extent of isotopic labeling Column: DB-1701 (agilent), 30 m length, 0.25 mm diameter, 0.25 μm (film thickness) Carrier: He (CP grade), 100 kPa (constant pressure) Inlet temperature: 250 ˚C Injection conditions: Split (25:1) with split flow 30 mL/min, splitless time 0.7 mins, purge 5 mL/min Detection: Mass Spec, EI (70 eV), source temperature 230 ˚C. Scan range 50 – 650 amu with a scan rate of 5 Hz. Transfer line 300 ˚C. Oven heating profile:

Time (minutes) Temperature 0  5 Hold at 50˚C 5  16.3 Ramp to 275 ˚C at 20 ˚C/min 16.3  18.8 Hold at 275 ˚C for 2.5 mins

S3.2.3 Results The extent of reaction was calculated by 1H NMR spectroscopy, by comparing the integrals of peaks arising from the aromatic protons of AcPh with those of 1-PhEtOH (Figure S 4(b)). Here it was found that the conversion was very high (> 93 %) in both. Chiral-phase HPLC demonstrated that only a single enantiomer had been formed in each case, with the (R)-ADH giving only (R)-1-PhEtOH (RT = 11.9 mins), and (S)-ADH giving only (S)-1-PhEtOH (RT = 12.6 mins), within the limits of detection (see Figure S 4(a)). The site and extent of 2H labelling in the 1-PhEtOH product was determined through a combination of 1H NMR spectroscopy and GC-MS, Figure S 4(b) and (c). In the 1H NMR 2 spectrum, the signal corresponding to the –CH3 protons could be used to detect H labelling on the α-carbon, giving a doublet (δ = 1.47 ppm, J = 6.3 Hz) when a 1H nucleus was attached to the α-carbon, and a broad singlet (δ = 1.46 ppm) when a 2H moiety was present. Similarly, a quartet (δ = 4.91 Hz, J = 6.3 Hz) corresponding to the α-1H signal was present in non- deuterated samples, but absent in those with a 2H-label. Low levels of deuterium exchange 2 on the –CH3 group to give –CH2 H was also evidenced by a small, broad shoulder with δ = 2 1.44 ppm (further H exchange on the –CH3 groups was not detected). Deconvolution of these peaks could be used to determine the relative proportion of each species in solution. In the mass spectrum of the 1-PhEtOH product extracted from the GC-MS trace, m/z values +• between 107 – 109 (corresponding to the [C7H7O ] fragment) could also be used to determine the extent of 2H incorporation at the α-position, Figure S 4(c). Similarly, the signals corresponding to the molecular ion (m/z 120 – 122) could be used to determine deuteration occurring on the methyl carbon. Whilst results arising from both the NMR and MS measurements were generally in good agreement, the latter was found to be more accurate for determining 1H:2H at the α-position, and the former more accurate for 2 determining % H exchange on the –CH3 group. In all experiments reported, the extent of secondary 2H labelling on the β-carbon was less than 4% of the [α -2H]-PhEtOH product.

S11

Figure S 4: Characterisation of products generated using the heterogeneous biocatalytic system with either an (S)- or (R)-alcohol dehydrogenase (ADH). In each case the results are compared to an isotopically unenriched commercial PhEtOH product standard. (a) HPLC-UV spectra for determination of the stereochemistry of PhEtOH produced compared to acetophenone, 1-(S)-PhEtOH, and racemic 1-PhEtOH 1 2 2 standards showing that only a single enantiomer is made in each reaction. (b) H NMR analysis ( H2O, p H 8.0, 500 MHz, 293 K) of the PhEtOH produced shows that high 2H incorporation is achieved. (c) Mass spectra (EI, 70 eV) of the PhEtOH produced shows an increase in molecular weight of 1 for both systems, further demonstrating high 2H incorporation.

S12

S3.3 Conversion of pyruvate to [2L-2H]-lactate (2)

S3.3.1 Experimental conditions 2 2 2 Reactions was conducted in 0.5 mL of ( H11)-Tris- HCl (100 mM, p H 8.0) containing 5 mM sodium pyruvate, 0.5 mM NAD+, and 500 μg of L-lactate dehydrogenase (Merck). The mixture was presaturated with H2 gas prior to addition of the heterogeneous biocatalytic system at a loading of 400 μg. The reactions were shaken at 500 rpm under a steady flow of H2 for 16 hours. S3.3.2 Analytical procedures GC-FID: In order to determine the conversion of pyruvate to lactate, the analytes were first derivatised and then subjected to analysis by GC-FID according to a modified version of the method of Bertrand et al.7 (see Figure S 5(a)(i)). Here, 100 μL of reaction mixture was combined with 23 μL of EtOH and 13 μL of pyridine by means of a vortex mixer. Subsequently, 15 μL of ethyl chloroformate was added, and the reaction allowed to proceed until no further CO2 was evolved. The derivatised pyruvate/lactate was extracted with 100 μL 2 of C HCl3, and transferred to a glass vial for analysis by GC according to the method shown below. The results are depicted in Figure S 5(a)(ii).

GC-FID method for quantification of conversion of pyruvate to lactate Column: DB-1701 (Agilent), 30 m length, 0.25 mm diameter, 0.25 μm film thickness Carrier: He (CP grade), 1.1 mL minute (constant flow) Inlet temperature: 250 ˚C Injection conditions: Splitless with split flow 50 mL/min, splitless time 0.7 mins, purge 5 mL/min Injection volume: 0.5 μL Detection: FID Detector temperature: 200 ˚C Detection gases: H2 (35 mL/min), air (350 mL/min), makeup N2 (40 mL/min) Oven heating profile: Time (minutes) Temperature 0  2 Hold at 45˚C 2  15 Ramp to 250 ˚C at 16 ˚C/min 15  20 Hold at 250 ˚C for 5 mins

Chiral GC-FID: The ee of the lactate product was determined by first derivatising with a chiral alcohol ((2R,3R)-(-)-2,3-butanediol) and then analysing the reaction solution by GC- FID according to a modified version of the method of Bertrand et al.7 (see Figure S 5(b)(i)). Here, 100 μL of reaction mixture was combined with 35 μL of (2R,3R)-(-)-2,3-butanediol and

S13

13 μL of pyridine by means of a vortex mixer. Subsequently, 15 μL of ethyl chloroformate was added, and the reaction allowed to proceed until no further CO2 was evolved. Commercial standards of L- and D-lactate were similarly derivatised for comparison. The 2 derivatised analytes were then extracted with 100 μL of C HCl3, and transferred to a glass vial for analysis by GC. The formation of diastereomeric products allowed for the analysis to be conducted on an achiral GC column according to the method shown below. The results are shown in Figure S 5(b)(ii). GC-FID method for determining enantiomeric excess of lactate Column: DB-1701 (Agilent), 30 m length, 0.25 mm diameter, 0.25 μm film thickness Carrier: He (CP grade), 170 kPa (constant pressure) Inlet temperature: 230 ˚C Injection conditions: Splitless with split flow 50 mL/min, splitless time 0.7 mins, purge 5 mL/min Injection volume: 0.5 μL Detection: FID Detector temperature: 250 ˚C Detection gases: H2 (35 mL/min), air (350 mL/min), makeup N2 (40 mL/min) Oven heating profile: Time (minutes) Temperature 0  5 Hold at 100˚C 5  30 Ramp to 150 ˚C at 2 ˚C/min 30  35 Hold at 150˚C for 5 minutes 35  41.5 Ramp to 280 ˚C at 2 ˚C/min 41.5  46.5 Hold at 280 ˚C for 5 minutes

GC-MS: Following analysis by GC-FID (see above) the EtOH/ethyl chloroformate derivatised reaction solutions were further analysed by GC-MS according to the method shown below. A commercial sample of sodium lactate of natural isotopic abundance was derivatised and analysed in an identical manner for comparison, with the results depicted in Figure S 5(c). GC-MS method for determining the extent of isotopic labeling Column: DB-1701 (agilent), 30 m length, 0.25 mm diameter, 0.25 μm (film thickness) Carrier: He (CP grade), 100 kPa (constant pressure) Inlet temperature: 250 ˚C Injection conditions: Split (10:1) with split flow 12 mL/min, splitless time 0.7 mins, purge 5 mL/min Detection: Mass Spec, EI (70 eV), source temperature 230 ˚C. Scan range 50 – 650 amu with a scan rate of 5 Hz. Transfer line 300 ˚C. Oven heating profile:

Time (minutes) Temperature 0  5 Hold at 50˚C 5  16.3 Ramp to 275 ˚C at 20 ˚C/min 16.3  18.8 Hold at 275 ˚C for 2.5 mins

S14

1H NMR spectroscopy: After removal of the catalyst, the unmodified reaction solutions were analysed by 1H NMR spectroscopy at 500 MHz and 298 K, with the results being shown in Figure S 5(d). S3.3.3 Results The very high conversion of pyruvate to lactate was confirmed by GC-FID analysis of the derivatised reaction solution (see Figure S 5(a)(ii)). Here, the peak corresponding to the pyruvate starting material (detected as ethyl pyruvate, RT = 5.85 minutes) disappeared almost entirely in the bio-hydrogenation reaction compared to a control in which no catalyst was included. The appearance of a peak at RT = 12.55 minutes, corresponding to the derivatised form of lactate, provided confirmation that the reaction had proceeded to 95 % completion. Similarly, derivatisaition of L- and D- lactate with a second chiral alcohol affords diastereomeric products with differing retention times (L = 25.5 mins, R = 25.3 mins, see Figure S 5(b)(ii)). Hence, by comparison, it could be confirmed that the lactate produced in the biohydrogenation was entirely of the L-form, within the limits of detection (> 99 %). The site and extent of 2H labelling on the L-lactate product was determined through a combination of GC-MS and 1H NMR spectroscopy, Figure S 5(c) and (d), respectively. In the 1 2 H NMR spectrum, the signal corresponding to the –CH3 protons could be used to detect H labelling on the α-carbon of L-lactate, forming a doublet (δ = 1.34 ppm, J = 9.2 Hz) when a 1H nucleus was at the α-position, and a broad singlet (δ = 1.34 ppm) when a 2H was present. Similarly, a quartet (δ = 4.13 ppm, J = 8.7 Hz) corresponding to the α-1H signal was present in non-deuterated samples, but absent in those with a 2H-label. Deuterium exchange on the – 2 CH3 group to yield –CH2 H was also evidenced by a smaller, broad shoulder at δ = 1.32 (although it is not possible to determine the extent of complete deuterium exchange of the – CH3 group via this method). Analysis of the ethoxycarbonyl ethylester derivatised lactate by mass spectroscopy gave a series of diagnostic peaks, notably the molecular ions at m/z = +• 2 +• 190.1 [C8H15O5 ] and m/z = 191.1 [ H-C8H14O5 ], which could be used to determine the extent of 2H incorporation at the α-position as > 97 %. The peak at m/z = 192.1 could also be 2 +• used to estimate the extent of formation of [ H2-C8H13O5 ] (Figure S 5(e)).

S15

Figure S 5: Characterisation of the products from the reductive deuteration of pyruvate by using the heterogeneous biocatalytic system and lactate dehydrogenase. The results are compared to isotopically unenriched commercial lactate standards and a control reaction with no bio system present. (a)(i) Reaction scheme showing the derivatization of lactate for analysis by (a)(ii) GC-FID. (b)(i) Reaction scheme showing the diastereotopic derivatization of lactate for analysis by (b)(ii) GC-FID. (b) Mass spectra (EI, 70 eV) and (c) 1H 2 2 NMR analysis ( H2O, p H 8.0, 500 MHz, 293 K) of L-lactate produced in the deuteration reaction compared to an isotopically unenriched standard. (d) Determined distribution of 2H labels across the product L-lactate.

S16

S3.4 Conversion of pyruvate to [2L-2H]-alanine (3)

S3.4.1 Experimental conditions 2 2 2 Reaction was conducted in 0.5 mL of ( H11)-Tris- HCl (100 mM, p H 8.0) containing 5 mM + sodium pyruvate, 25 mM NH4Cl, 0.5 mM NAD , and 500 μg of L-alanine dehydrogenase in solution. The mixture was presaturated with H2 gas prior to addition of the heterogeneous biocatalytic system at a loading of 400 μg. The reactions were shaken at 500 rpm under a steady flow of H2 for 16 hours. S3.4.2 Analytical procedures GC-FID: In order to determine the conversion of pyruvate to alanine, the analytes were first derivatised and then subjected to analysis by GC-FID according to a modified version of the method of Bertrand et al.7 (see Figure S 6(a)). Here, 100 μL of reaction mixture was combined with 23 μL of EtOH and 13 μL of pyridine by means of a vortex mixer. Subsequently, 15 μL of ethyl chloroformate was added, and the reaction allowed to proceed until no further CO2 was evolved. The derivatised pyruvate/alanine was extracted with 100 2 μL of C HCl3, and transferred to a glass vial for analysis by GC according to the method shown below. The results are depicted in Figure S 6(b).

GC-FID method for quantification of conversion of pyruvate to alanine Column: DB-1701 (Agilent), 30 m length, 0.25 mm diameter, 0.25 μm film thickness Carrier: He (CP grade), 1.1 mL minute (constant flow) Inlet temperature: 250 ˚C Injection conditions: Splitless with split flow 50 mL/min, splitless time 0.7 mins, purge 5 mL/min Injection volume: 1.0 μL Detection: FID Detector temperature: 200 ˚C Detection gases: H2 (35 mL/min), air (350 mL/min), makeup N2 (40 mL/min) Oven heating profile: Time (minutes) Temperature 0  2 Hold at 45˚C 2  15 Ramp to 250 ˚C at 16 ˚C/min 15  20 Hold at 250 ˚C for 5 mins

Chiral GC-FID: The same samples from the GC-FID analysis described above were then subjected to analysis on a chiral-phase GC column according to the method described below in order to determine the ee of the alanine product. Similarly derivatised commercial standards of L- and D-alanine standards were also run for comparison, with the results being depicted in Figure S 6(c).

S17

GC-FID method for determining enantiomeric excess of alanine Column: CP-Chirasil-Dex CB (Agilent), 25 m length, 0.25 mm diameter, 0.25 μm (film thickness), fitted with a guard of 10 m undeactivated fused silica of the same diameter Carrier: He (CP grade), 170 kPa (constant pressure) Inlet temperature: 200 ˚C Injection conditions: Splitless with split flow 60 mL/min, splitless time 0.8 mins, purge 5 mL/min. Injection volume = 0.5 μL Detection: FID Detector temperature: 200 ˚C Detection gases: H2 (35 mL/min), air (350 mL/min), makeup N2 (40 mL/min) Oven heating profile: Time (minutes) Temperature 0  5 Hold at 70˚C 5  30 Ramp to 120 ˚C at 2 ˚C/min 30  36 Ramp to 180 ˚C at 10 ˚C/min 36  41 Hold at 180 ˚C for 5 minutes

Time (minutes) Temperature 0  5 Hold at 50˚C 5  16.3 Ramp to 275 ˚C at 20 ˚C/min 16.3  18.8 Hold at 275 ˚C for 2.5 mins

1H NMR spectroscopy: After removal of the catalyst, the unmodified reaction solutions were analysed by 1H NMR spectroscopy at 500 MHz and 298 K, with the results being shown in Figure S 6(e).

S18

S3.4.3 Results The very high conversion of pyruvate to alanine was confirmed by GC-FID analysis of the derivatised reaction solution (see Figure S 6(b)). Here, the peak corresponding to the pyruvate starting material (detected as ethyl pyruvate, RT = 5.95 minutes) disappeared almost entirely in the bio-hydrogenation reaction compared to a control in which no catalyst was included. The appearance of a peak at RT = 10.53 minutes, corresponding to the N- ethoxycarbonyl ethylester derivatised form of alanine, provided confirmation that the reaction had proceeded to 95 % completion. Similarly, derivatisaition of commercial samples of L- and D-alanine and analysis on a chiral GC-column gave rise to two peaks (L = 25.5 minutes, D = 25.3 minutes, see Figure S 6(c)). Hence, by comparison, it could be confirmed that the alanine produced in the biohydrogenation was entirely of the L-form, within the limits of detection (> 99 %). The site and extent of 2H labelling on the L-alanine product was determined through a combination of GC-MS and 1H NMR spectroscopy (see Figure S 6(d) and (e), respectively). 1 In the H NMR spectrum, the signal corresponding to the –CH3 protons could be used to detect 2H labelling on the α-carbon of L-alanine, forming a doublet (δ = 1.47 ppm, J = 7.4 Hz) when a 1H nucleus was at the α-position, and a broad singlet (δ = 1.46 ppm) when a 2H was present. Similarly, a quartet (δ = 3.77 Hz, J = 7.4 Hz) corresponding to the α-1H signal was present in non-deuterated samples, but absent in those with a 2H-label. Integration of the relevant peaks indicated a 2H-incorportation of 97 % on the α-carbon. Single and double 2 2 deuterium exchange on the –CH3 group to yield –CH2 H and CH H2 was also evidenced by smaller, broad shoulders at δ = 1.44 and δ = 1.43 ppm respectively (although it was not 2 possible to quantify the extent of –C H3 formation via this method). Analysis by mass spectroscopy gave rise to a high intensity peak corresponding to the derivatised alanine following cleavage of the N-ethoxycarbonyl group, with m/z = 116.1 (corresponding to +• 2 +• [C5H10NO2 ]) and m/z = 117.1 (corresponding to [ H-C5H9NO2 ]). The peaks at m/z = 116.1 and 117.1 could therefore be used to verify the high %2H incorporation at the α-position discerned previously by NMR spectroscopy. Similarly, the signals with m/z 118 – 120 could 2 also be used to better quantify the extent of H-exchange that had occurred on the –CH3 carbon of the α-deuterated alanine, which was subsequently shown to be minimal (see Figure S 6(f)).

S19

Figure S 6: Characterisation of the products from the reductive deuteration of pyruvate by using the heterogeneous biocatalytic system and L-alanine dehydrogenase. The results are compared to isotopically unenriched commercial alanine standards and a control reaction with no bio system present. (a)(i) Reaction scheme showing the derivatization of alanine for analysis by (b) GC-FID and (c) chiral-GC-FID. (d) Mass 1 2 2 spectra (EI, 70 eV) and (e) H NMR analysis ( H2O, p H 8.0, 500 MHz, 293 K) of L-alanine produced in the deuteration reaction compared to an isotopically unenriched standard. (f) Determined distribution of 2H labels across the product L-alanine.

S20

2 2 S3.5 Conversion of phenylpyruvate to [2- H2, 3L- H]-phenylalanine (4)

S3.5.1 Experimental conditions 2 2 2 Phenylpyruvic acid (5 mM) in 0.5 mL of ( H11)-Tris- HCl (100 mM, p H 8.0) was first allowed to stand for 24 hours in order to exchange labile protons for deuterons, before the + addition of 25 mM NH4Cl, 0.5 mM NAD , and 500 μg of L-phenylalanine dehydrogenase. The mixture was presaturated with H2 gas prior to addition of the Bio-catalyst systems at a loading of 400 μg. The reaction was then shaken at 500 rpm under a steady flow of H2 for 16 hours. S3.5.2 Analytical procedures GC-FID: In order to determine the conversion of phenylpyruvate to alanine, the analytes were first derivatised and then subjected to analysis by GC-FID according to a modified version of the method of Bertrand et al.7 (see Figure S 7(a)(i)). Here, 100 μL of reaction mixture was combined with 23 μL of EtOH and 13 μL of pyridine by means of a vortex mixer. Subsequently, 15 μL of ethyl chloroformate was added, and the reaction allowed to proceed until no further CO2 was evolved. The derivatised pyruvate/alanine was extracted 2 with 100 μL of C HCl3, and transferred to a glass vial for analysis by GC according to the method shown below. The results are depicted in Figure S 7(a)(ii).

GC-FID method for determining conversion of phenylpyruvate to phenylalanine Column: DB-1701 (Agilent), 30 m length, 0.25 mm diameter, 0.25 μm film thickness Carrier: He (CP grade), 170 kPa (constant pressure) Inlet temperature: 230 ˚C Injection conditions: Splitless with split flow 50 mL/min, splitless time 0.7 mins, purge 5 mL/min Injection volume: 0.5 μL Detection: FID Detector temperature: 250 ˚C Detection gases: H2 (35 mL/min), air (350 mL/min), makeup N2 (40 mL/min) Oven heating profile: Time (minutes) Temperature 0  5 Hold at 100˚C 5  30 Ramp to 150 ˚C at 2 ˚C/min 30  35 Hold at 150˚C for 5 minutes 35  41.5 Ramp to 280 ˚C at 2 ˚C/min 41.5  46.5 Hold at 280 ˚C for 5 minutes

S21

GC-MS: The N-ethoxycarbonyl ethylester-derivatised samples from the GC-FID analysis described (see above) were subsequently analysed by GC-MS according to the method below. A commercial sample of L-phenylalanine of natural isotopic abundance was derivatised and analysed in an identical manner for comparison, with the results depicted in Figure S 7(c). GC-MS method for determining the extent of isotopic labeling Column: DB-1701 (agilent), 30 m length, 0.25 mm diameter, 0.25 μm (film thickness) Carrier: He (CP grade), 100 kPa (constant pressure) Inlet temperature: 250 ˚C Injection conditions: Split (10:1) with split flow 12 mL/min, splitless time 0.7 mins, purge 5 mL/min Detection: Mass Spec, EI (70 eV), source temperature 230 ˚C. Scan range 50 – 650 amu with a scan rate of 5 Hz. Transfer line 300 ˚C. Oven heating profile:

Time (minutes) Temperature 0  5 Hold at 50˚C 5  16.3 Ramp to 275 ˚C at 20 ˚C/min 16.3  18.8 Hold at 275 ˚C for 2.5 mins

Chiral GC-FID: Derivatisation of the samples for analysis by chiral GC-FID to establish the enantiomeric excess of the reaction, was carried out according to the procedure described in the previous section, except with nPrOH (30 μL) substituted for EtOH (see Figure S 7(b)(ii). Chiral GC-FID was then carried out according to the following method, with the results shown in Figure S 7(b)(ii).

GC-FID method for determining enantiomeric excess of phenylalanine Column: CP-Chirasil-Dex CB (Agilent), 25 m length, 0.25 mm diameter, 0.25 μm (film thickness), fitted with a guard of 10 m undeactivated fused silica of the same diameter Carrier: He (CP grade), 2 mL/min (constant flow) Inlet temperature: 200 ˚C Injection conditions: Splitless with split flow 80 mL/min, splitless time 0.8 mins, purge 5 mL/min. Injection volume = 0.3 μL Detection: FID Detector temperature: 200 ˚C Detection gases: H2 (35 mL/min), air (350 mL/min), makeup N2 (40 mL/min) Oven heating profile: Time (minutes) Temperature 0  5 Hold at 70˚C 5  115 Ramp to 180 ˚C at 1˚C/min 115  130 Hold at 180 ˚C for 15 minutes

1H NMR spectroscopy: After removal of the catalyst, the unmodified reaction solutions were analysed by 1H NMR spectroscopy at 500 MHz and 298 K, with the results being shown in Figure S 7(d).

S22

S3.5.3 Results The very high conversion of phenylpyruvate to phenylalanine was confirmed by GC-FID analysis of the derivatised reaction solution (see Figure S 7(a)(ii)). Here, the peak corresponding to the derivatised phenylpyruvate starting material (RT = 15.05 minutes) disappeared almost entirely in the bio-hydrogenation reaction compared to a control in which no catalyst was included. The appearance of a peak (RT = 14.94 minutes) corresponding to the N-ethoxycarbonyl ethylester derivatised form of phenylalanine, provided confirmation that the reaction had proceeded to ≥ 99 % completion. Similarly, derivatisation of commercial standards of L- and D-phenylalanine and analysis on a chiral GC-column gave rise to two peaks (L = 91.78 minutes, D = 92.00 minutes, see Figure S 7(b)(ii)). Hence, it could be confirmed that the alanine produced in the biohydrogenation was entirely of the L-form, within the limits of detection. The site and extent of 2H labelling on the L-phenylalanine product was determined through a combination of GC-MS and 1H NMR spectroscopy (see Figure S 7(c) and (d) respectively). In this case, the phenylpyruvate starting material was left in the reaction solution to allow exchange of the labile protons for deuterons in the solvent. Following reduction by the biocatalyst system and L-phenylalanine dehydrogenase, the resulting product was labelled on both α- and β-positions. Accordingly, the signals at δ = 4.01 ppm (dd, 1H, αH, J = 4.1 Hz, J = 6.2 Hz), δ = 3.31 ppm (dd, 1H, βH, J = 4.1 Hz, J = 11.7 Hz), and δ = 3.14 ppm (dd, 1H, βH, J = 6.4 Hz, J = 11.7 Hz) which were all observed in the commercial standard of L- phenylalanine, were found to be absent in the reaction product. Similarly, GC-MS analysis of the ethoxycarbonyl ethylester-derivative of phenylalanine demonstrated the expected +3 mass shift of the molecular from m/z 220.1 to 223.1 (see Figure S 7(c)).

S23

Figure S 7: Characterisation of the products from the reductive deuteration of phenylpyruvate by using the heterogeneous biocatalytic system and L-phenylalanine dehydrogenase. The results are compared to isotopically unenriched commercial phenylalanine standards and a control reaction with no bio system present. (a)(i) Reaction scheme showing the derivatization of phenylalanine for analysis by (a)(ii) GC-FID. (b)(i) Reaction scheme showing the derivatization of phenylalanine for analysis by (b)(ii) chiral-GC-FID. (c) Mass 1 2 2 spectra (EI, 70 eV) and (d) H NMR analysis ( H2O, p H 8.0, 500 MHz, 293 K) of L-phenylalanine produced in the deuteration reaction compared to an isotopically unenriched standard.

S24

2 S3.6 Conversion of carvones to [2,3- H2]-dihydrocarvones (5a, 5b)

S3.6.1 Experimental procedures 2 2 2 Reaction was conducted in 0.5 mL of ( H5)-Tris- HCl (100 mM, p H 8.0) containing 5 mM 2 (5R)-carvone (Reaction A) or (5S)-carvone (Reaction B), 4 vol.% [ H6]-DMSO, 0.5 mM + NAD , and 500 μg of ene reductase (JM, ENE101). The mixture was presaturated with H2 gas prior to addition of the heterogeneous bio-system at a loading of 400 μg. The reactions were shaken at 500 rpm under a steady flow of H2 for 16 hours. Evaporation of the product dihydrocarvones proved to more problematic than in other cases so alternative reaction setups were trialled, including a pressure vessel (Büchi Tinyclave) sealed under 2 bar H2 (50 mL headspace), and a reaction in a 1.5 mL glass GC-vial sealed under H2 balloon pressure. Whilst a commercial standard of mixed of (2R, 5R)- and (2S, 5R)-dihydrocarvone was commercially available, the 5S-epimer was unavailable. Instead, this was synthesised by reduction of (5S)-carvone by Zn metal, according to the method of Leiva de Faria et al.8 Here, Zn metal (625 mg, 9.6 mmol) and KOH (250 mg, 4.5 mmol) were combined in 3.5 mL of MeOH/H2O (95/5 v/v) and the slurry bought to reflux under stirring. (5S)-carvone (500 mg, 3.3 mmol) was dissolved in 1 mL of MeOH/H2O (95/5 v/v) and added dropwise to the Zn-solution over 6 hours. Following the reaction, the solution was filtered through a syringe filter, extracted into pentane (3 × 1 mL) and dried over Na2SO4. The solvent was removed in vacuo to give (5S)-dihydrocarvone (400 mg, 80 %) as a mixture of cis-2,5 and trans-2,5 epimers in the ratio 83:17 (by GC). In order to assist in the assignment of the 1H NMR spectra of the deuterated samples, (2R, 5R)- and (2R, 5S)-dihydrocarvones of natural isotopic abundance were also prepared at high de from their parent (5R)- and (5S)-carvones by reaction with an ene reductase. Here, the relevant carvone (5 mM) was shaken with 500 μg of ene reductase (JM, ENE101) in 500 μL of non-deuterated Tris buffer (100 mM, pH 8.0) containing 2 vol.% DMSO and 10 mM 2 NADH for 6 hours at room temperature. H2O (100 μL) was added to the sample for field locking purposes, and the reaction mixture was analysed immediately by 1H NMR spectroscopy.

S25

S3.6.2 Analytical procedures GC-FID: Following the reactions, an aliquot of the solution was extracted with a 2 × volume 2 of C HCl3, and then centrifuged at 18 800 × g for 5 minutes before being transferred to glass vials for analysis by GC-FID according to the method below. The results for reactions A and B are shown in Figure S 8(a) and Figure S 9(a), respectively.

GC-FID method for identification of products in the reduction of carvones Column: DB-1701 (Agilent), 30 m length, 0.25 mm diameter, 0.25 μm film thickness Carrier: He (CP grade), 0.5 mL minute (constant flow) Inlet temperature: 230 ˚C Injection conditions: Splitless with split flow 60 mL/min, splitless time 0.8 mins, purge 5 mL/min. Injection volume: 0.5 μL Detection: FID Detector temperature: 150 ˚C Detection gases: H2 (35 mL/min), air (350 mL/min), makeup N2 (40 mL/min) Oven heating profile:

Time (minutes) Temperature 0  5 Hold at 50˚C 5  25 Ramp to 250 ˚C at 10 ˚C/min 25  35 Hold at 250 ˚C for 10 mins

GC-MS: Following extraction and analysis by GC-FID (above) the reaction solutions were analysed by GC-MS according to the method below. The results for reactions A and B are shown in Figure S 8(c) and Figure S 9(c), respectively. GC-MS method for determining the extent of isotopic labeling Column: DB-1701 (agilent), 30 m length, 0.25 mm diameter, 0.25 μm (film thickness) Carrier: He (CP grade), 100 kPa (constant pressure) Inlet temperature: 250 ˚C Injection conditions: Split (10:1) with split flow 12 mL/min, splitless time 0.7 mins, purge 5 mL/min Detection: Mass Spec, EI (70 eV), source temperature 230 ˚C. Scan range 50 – 650 amu with a scan rate of 5 Hz. Transfer line 300 ˚C. Oven heating profile:

Time (minutes) Temperature 0  5 Hold at 50˚C 5  16.3 Ramp to 275 ˚C at 20 ˚C/min 16.3  18.8 Hold at 275 ˚C for 2.5 mins

1H NMR spectroscopy: After removal of the catalyst, the unmodified reaction solutions were analysed by 1H NMR spectroscopy at 500 MHz and 298 K, with the results for reactions A and B being shown in Figure S 8(b) and Figure S 9(b), respectively. In order to confirm assignments, 1H-1H correlation spectroscopy (COSY) was also conducted on the samples as required.

S26

2 S3.6.3 Results A: [2, 3- H2]-(2R, 5R)-dihydrocarvone (5a) Analysis by GC-FID confirmed the high conversion of substrate to product, and the high selectivity for the trans-2,5 diastereomer (see Figure S 8(a)). The slightly compromised diastereomeric excess (97 %) is most likely a consequence of the racemisation of the dihydrocarvone product, which occurs slowly in aqueous solutions. Quantitative analysis revealed that there was a considerable loss of product from the reaction mixture when H2 was flowed across the headspace, and so a mass balance was not achieved in this case. Sealing the reaction under 2 bar H2 minimised product losses due to evaporation, but still resulted in recovery of only about 30 % of the theoretical quantity. The site and extent of 2H labelling was demonstrated by 1H NMR spectroscopy (see Figure S 8(b)), the assignment of which correlated well with those of previous authors.8 Changes to the spectrum following the deuteration reaction confirmed both the highly specific 2H incorporation, and high diastereoselectivity that was achieved. Disappearance of the multiplet at δ = 2.61 ppm (labelled b in the diagram) and collapse of the coupling on the doublet at δ = 0.99 ppm with J = 6.4 Hz, were clear evidence of the 2H installation on C2. Similarly, loss of the pseudo-quartet peak at δ = 1.41 ppm (labelled cb in the diagram), with simultaneous loss of coupling on signals representing neighbouring protons (da and db at δ = 1.94 ppm and δ = 1.72 ppm, respectively), provided evidence of a 2H nucleus at C3. The preservation of the peak at δ = 2.18 ppm (labelled ca in the diagram), which collapses from a multiplet to a broad singlet, is evidence that a single 2H nucleus is installed stereoselectively on C3.9 As the 2H on C3 has higher chiral priority than the 1H, the product may be assigned 2 10 2 as [2R, 3R- H2]-(5R)-dihydrocarvone. Hence, the two H nuclei are added in an anti- relationship across C2 and C3, which is in accordance with previous mechanistic studies of similar ene reductase enzymes.9,10 Finally, GC-MS analysis of the reaction solution further confirms the addition of two 2H nuclei across the molecule, with the molecular ion being seen to shift from m/z = 152.1 for +• 2 +• [C10H16O ], to 154.1 for [ H2-C10H14O ], as shown in Figure S 8(c).

S27

S28

Figure S 8: Characterisation of products from the reductive deuteration of (5R)-carvone by using the heterogeneous bio-system with an ene reductase. In each case the results are compared to an isotopically unenriched commercial dihydrocarvone standard, or a sample prepared by supplying an ene reductase with substrate and super-stoichiometric NADH. (a) GC-FID determination of the dihydrocarvone produced 1 2 2 compared to standards and a control reaction (with no bio-system added). (b) H NMR analysis ( H2O, p H 8.0, 500 MHz, 293 K) and (c) Mass spectra (EI, 70 eV) of the dihydrocarvone produced from the reaction compared to isotopically unenriched samples.

S29

2 S3.6.4 Results B: [2, 3- H2]-(2R, 5S)-dihydrocarvone (5b) Analysis by GC-FID yielded results that were similar to those for Reaction A described above. Again, conversion from (5S)-carvone to (2R, 5S)-dihydrocarvone was found to be virtually quantitative (see Figure S 9(a)), though evaporation meant that recovery of the product was quite poor. Conducting the reaction in a sealed vial under a balloon pressure of H2 (as opposed to a dynamic flow of H2 across the headspace) helped to improve the product yield to around 30 % of the theoretical quantity. Unlike reaction A, the cis- (2R, 5S)-isomer if formed, which represents the least stable of the two possible diastereomeric products (with racemisation giving rise to a slightly diminished de of 96 %). 1H NMR spectroscopic analysis of the reaction solution provided insight into the extent and location of 2H incorporation into the product. Here, the dihydrocarvone standard prepared by a conventional Zn reduction was not found to be useful, as it contained mostly the thermodynamically favoured trans- (2R, 5R)-dihydrocarvone isomer (see Figure S 9(a)). Instead, a standard prepared by supplying the ene reductase with (5S)-carvone and a super- stoichiometric quantity of NADH in non-deuterated buffer was found to generate a suitable sample of (2R, 5S)-dihydrocarvone of natural isotopic abundance for comparison. The 1H NMR spectrum of the deuterated product is less well resolved than in Reaction A, but the disappearance of signals at δ = 2.56 ppm and δ = 1.58 ppm (labelled b and cb respectively) along the collapse of coupling on the doublet at δ = 1.05 ppm can be clearly seen (Figure S 9(b)). The convoluted peak at δ = 1.93 ppm retains a relative integration of 3H on going from the standard to the deuterated reaction, confirming that the area of signal ca is not diminished. All of these factors provide evidence for two 2H nuclei being installed across the double bond, although the likely anti-arrangement cannot be confirmed unambiguously. Finally, GC-MS analysis of the reaction solution further indicates the addition of two 2H nuclei across the molecule, with the molecular ion being seen to shift from m/z = 152.1 for +• 2 +• [C10H16O ], to 154.1 for [ H2-C10H14O ], as shown in Figure S 9(c).

S30

S31

Figure S 9: Characterisation of products from the reductive deuteration of (5S)-carvone by using the heterogeneous bio-system with an ene reductase. In each case the results are compared to an isotopically unenriched dihydrocarvone samples, prepared by zinc reduction of the carvone or by supplying the ene reductase with substrate and super-stoichiometric NADH. (a) GC-FID determination of the dihydrocarvone 1 2 produced compared to standards and a control reaction (with no bio-system added). (b) H NMR analysis ( H2O, p2H 8.0, 500 MHz, 293 K) and (c) Mass spectra (EI, 70 eV) of the dihydrocarvone produced from the reaction compared to isotopically unenriched samples.

S32

S3.7 Selective biocatalytic deuteration of cinnamaldehyde (6a, 6b, 6c)

S3.7.1 Experimental conditions 2 2 2 Reactions were conducted in 0.5 mL of ( H11)-Tris- HCl (100 mM, p H 8.0) containing 5 mM cinnamaldehyde, 2 vol.% DMSO, 0.5 mM NAD+, and 500 μg of (A) alcohol dehydrogenase (JM, ADH 105), and (B-C) ene reductase (JM, ENE101). The mixture was presaturated with H2 gas prior to addition of the heterogeneous biocatalytic system at a loading of 400 μg. The reactions were shaken at 500 rpm under a steady flow of H2 for 16 hours. In the case of reaction C, the initial step was allowed to proceed for 24 hours before addition of a further 200 μg of the heterogeneous biocatalytic system and 500 μg of alcohol dehydrogenase (JM, ADH 105). S3.7.2 Analytical procedures GC-FID: Following the reactions, an aliquot of the solution was extracted with a 2 × volume 2 of C HCl3, and then centrifuged at 18 800 × g for 5 minutes before being transferred to glass vials for analysis by GC-FID according to the method below. The results for reactions A, B, and C are shown in Figure S 10(a), Figure S 11(a), and Figure S 12(a), respectively. GC-FID method for identification of products in the reduction of cinnamaldehyde Column: DB-1701 (Agilent), 30 m length, 0.25 mm diameter, 0.25 μm film thickness Carrier: He (CP grade), 0.5 mL minute (constant flow) Inlet temperature: 230 ˚C Injection conditions: Splitless with split flow 60 mL/min, splitless time 0.8 mins, purge 5 mL/min. Injection volume: 0.5 μL Detection: FID Detector temperature: 150 ˚C Detection gases: H2 (35 mL/min), air (350 mL/min), makeup N2 (40 mL/min) Oven heating profile:

Time (minutes) Temperature

S33

0  5 Hold at 50˚C 5  25 Ramp to 250 ˚C at 10 ˚C/min 25  35 Hold at 250 ˚C for 10 mins

GC-MS: Following extraction and analysis by GC-FID (above) the reaction solutions were analysed by GC-MS according to the method below. The results for reactions A, B, and C are shown in Figure S 10(c), Figure S 11(c), and Figure S 12(c), respectively.

Time (minutes) Temperature 0  5 Hold at 50˚C 5  16.3 Ramp to 275 ˚C at 20 ˚C/min 16.3  18.8 Hold at 275 ˚C for 2.5 mins

1H NMR spectroscopy: After removal of the catalyst, the unmodified reaction solutions were analysed by 1H NMR spectroscopy at 500 MHz and 298 K, with the results for reactions A, B, and C being shown in Figure S 10(b), Figure S 11(b), and Figure S 12(b), respectively. In the case of reactions A and C, it was necessary to acquire additional spectra under non- 2 aqueous conditions by extracting the product into an equal volume of C HCl3 and re- analysing it.

S34

S3.7.3 Results A: [1-2H]-cinnamyl alcohol (6a) Analysis by GC-FID, Figure S 10(a), shows a broad peak for the cinnamyl alcohol product (RT = 19.9 mins) and one for unreacted cinnamaldehyde (RT = 19.4 mins), with no additional products being detected (including hydrocinnamalehyde and 3-phenyl-1-propanol, which are common side products in the hydrogenation of cinnamaldehyde).REF Integration of the peak areas compared to a set of standard solutions indicated a conversion of 90 %, which could be confirmed by comparison of clearly resolved peaks in the 1H NMR spectrum of the crude reaction mixture (see Figure S 10(b)). The site and extent of 2H labelling in the cinnamyl alcohol product was determined through 1H NMR spectroscopy, Figure S 10(b). The intensity of the doublet at δ = 4.30 ppm (J = 5.5 Hz) in the unlabelled cinnamyl alcohol standard (corresponding to the α-protons) halves in intensity in the spectrum of the biocatalytic reaction product. Overlapping peaks from the ribose ring of the cofactor make a precise integration of the peak from the α-proton difficult, 2 though extracting the product into C HCl3 and re-acquiring the spectrum enabled the measurement to be conducted unambiguously. Hence, it was confirmed that the α-site had been labelled with a single 2H nucleus. Accordingly, the peak corresponding to the neighbouring C2 proton is also found to change from a doublet of triplets in the unlabelled standard (δ = 6.47 ppm, dt, J = 16.1 Hz, J = 5.8 Hz), to a doublet of doublets in the deuterated product (δ = 6.47 ppm, dd, J = 16.1 Hz, J = 5.8 Hz).

Unfortunately, mass spectroscopic analysis could not be used to unambiguously evaluate the extent of %2H labelling, as such compounds are known to re-arrange upon ionization, causing large differences in the fragmentation pattern of different isotopomers.11 However, the spectra were still acquired, and are shown in Figure S 10(c) for completeness. Here, the +• molecular ion peak in the commercial cinnamyl alcohol standard ([C9H10O ], m/z 134.1) is seen to increase by + 1.0 for the product of the biocatalytic hydrogenation, consistent with the 2 2 +• addition of a H nucleus ([ H1-C9H9O ], m/z 135.1).

S35

Figure S 10: Characterisation of products from the reductive deuteration of cinnamaldehyde by using the heterogeneous bio-system with an alcohol dehydrogenase. In each case the results are compared to an isotopically unenriched commercial cinnamly alcohol standard. (a) GC-FID determination of the cinnamly alcohol produced compared to cinnamaldehyde, cinnamly alcohol standards and a control reaction (with no bio system added) showing that only a single product is made in each reaction with no side product formation. (b) 1 2 2 2 H NMR analysis ( H2O, p H 8.0, 500 MHz, 293 K) of the cinnamly alcohol produced shows that high H incorporation is achieved. (c) Mass spectra (EI, 70 eV) of the cinnamyl alcohol produced compared to an isotopically unenriched standard.

S36

2 S3.7.4 Results B: [2, 3- H2]-hydrocinnamaldehyde (6b) The near quantitative conversion of cinnamaldehyde to hydorcinnamaldehyde could was demonstrated by GC-FID analysis, see Figure S 11(a). Here, the cinnamaldehyde substrate (RT = 19.4 minutes) was absent from the reaction solution, which showed a peak only for hydrocinnamaldehyde (RT = 17.5 mins), without the formation of other possible side products (namely cinnamyl alcohol and/or 3-phenyl-1-propanol). This result was further confirmed by 1H NMR analysis of the crude reaction mixture. (see Figure S 11(b)). The site and extent of 2H labelling in the hydrocinnamaldehyde product could also be determined through 1H NMR spectroscopy, Figure S 11(b), though the interpretation is complicated by the speciation of the molecule under aqueous conditions. As such, each of the aliphatic proton environments on the molecule (denoted a, b, and c in Figure S 11(b) give rise to three separate peaks in the corresponding 1H spectrum, which must be summed together in order to determine the extent of 2H incorporation at each site. Accordingly, signals corresponding to sites a and b lose half of their intensity when the molecule is prepared by biocatalytic deuteration, whilst proton c (and the aromatic protons) remain unchanged. The extent of the collapse in coupling across all of the peaks is also commensurate with the addition of single deuterons at positions a and b. Unfortunately, mass spectroscopic analysis could not be used unambiguously to evaluate the extent of %2H labelling as such compounds are known to re-arrange upon ionization, thereby causing large differences in the fragmentation pattern of different isotopomers.11 However, the spectra were still acquired, and are shown in Figure S 11(c) for completeness. Here, the +• molecular ion peak in the commercial standard of hydrocinnamaldehyde ([C9H10O ], m/z 134.1) is seen to increase by + 2.0 for the product of the biocatalytic hydrogenation, 2 2 +• consistent with the addition of two H nuclei ([ H2-C9H8O ], m/z 136.1).

S37

Figure S 11: Characterisation of products from the reductive deuteration of cinnamaldehyde by using the heterogeneous bio-system with an ene reductase. In each case the results are compared to an isotopically unenriched commercial hydrocinnamaldehyde standard. (a) GC-FID determination of the cinnamly alcohol produced compared to cinnamaldehyde, hydrocinnamaldehyde standards and a control reaction (with no bio- system added) showing that only a single product is made with no side product formation. (b) 1H NMR analysis 2 2 2 ( H2O, p H 8.0, 500 MHz, 293 K) of the hydrocinnamaldehyde produced shows that high H incorporation is achieved. (c) Mass spectra (EI, 70 eV) of the hydrocinnamaldehyde produced compared to the isotopically unenriched standard.

S38

2 S3.7.5 Results C: [1, 2, 3- H3.5]-3-phenyl-1-propanol (6c) The conversion of cinnamaldehyde to hydrocinnamaldehyde by the action of the heterogeneous biocatalytic system and an ene reductase was shown to proceed quantitiatively in SM Section S3.7.3). The second half of Reaction C, in which the hydrocinnamaldehyde was reduced further to 3-phenyl-1-propanol by the introduction of an alcohol dehydrogenase, was shown to proceed to a similarly high degree by GC-FID analysis (see Figure S 12(a)). Here, signals corresponding to the substrate (RT = 19.4 mins) and intermediate (RT = 17.5 mins) compounds were replaced in the chromatogram of the reaction mixture by the product peak (RT = 18.7 mins). Similarly, analysis of the 1H NMR spectrum of the crude reaction mixture indicated that the reaction had proceeded to ≥ 98 % conversion. Recovery of the product was not, however, quantitative, with around 30 % having been lost to evaporation over the course of the reaction. The site and extent of 2H labelling in the 3-phenyl-1-propanol product was determined through 1H NMR spectroscopy, Figure S 12(b). Analysis of the crude reaction mixture is complicated by the coincidence of the diagnostic peaks with those from other sources (such as the co-solvent and buffer, see Figure S 12(b)(i)). As such, the product is better analysed by 2 extracting into C HCl3 and re-acquiring the spectrum (see Figure S 12(b)(ii)). Here the intensity of signals corresponding to protons in position b (δ = 2.72 ppm) and d (δ = 3.68 ppm) are reduced by 50 % following the deuteration reaction (compared to a standard sample of natural isotopic abundance), consistent with the addition of a single deuteron at each of these sites. The signal for the proton at position c (δ = 1.90 ppm) was found to decrease by 75 % however, indicating that 1.5 2H units had been installed at this site. The additional 2H incorporation in this latter case most likely occurred through tautomeric solvent exchange following the reduction of the C=C double bond. Upon deuteration, the clear splitting patterns for peaks b (a pseudo-triplet), c (a pseudo-quintet), and d (a pseudo-quartet), are replaced by broad, poorly defined signals, consistent with the influence of much weaker proton-deuteron coupling. Mass spectroscopy was also used to study the isotopic substitution pattern along the 3- phenyl-1-propanol product (see Figure S 12(c). Here the molecular ion peak was found to be diagnostic, with the m/z 136.1 representing the unfragmented molecule.12 Accordingly, peaks at m/z 137.1, 138.1, and 139.1, in the analysis of the reaction mixture could be attributed to the inclusion of two, three, and four 2H nuclei, respectively, consistent with the various levels of labelling on C2 described above.

S39

S40

Figure S 12: Characterisation of products generated using a two-step, one pot reductive deuteration reaction. Each step used the heterogeneous bio-system with (i) an ene reductase acting on cinnamaldehyde and (ii) and alcohol dehydrogenase acting on the intermediate hydrocinnamaldehyde. In each case the results are compared to an isotopically unenriched commercial 3-phenyl-1-propanol standard. (a) GC-FID determination of the 3-phenyl-1-propanol produced compared to cinnamaldehyde, hydrocinnamaldehyde and 3-phenyl-1-propanol standards and a control reaction (with no bio-system added) showing that only a single product is made in each reaction with no side product formation. (b) 1H NMR 2 2 2 analysis ( H2O, p H 8.0, 500 MHz, 293 K) of the 3-phenyl-1-propanol produced shows that high H incorporation is achieved. (c) Mass spectra (EI, 70 eV) of the 3-phenyl-1-propanol produced compared to the isotopically unenriched standard. (d) Determined distribution of 2H labels.

S41

S4. References 1. Lukey, M. J. et al. How Escherichia coli is equipped to oxidize hydrogen under different redox conditions. J. Biol. Chem. 285, 3928–3938 (2010). 2. Lauterbach, L. & Lenz, O. Catalytic Production of Hydrogen Peroxide and Water by Oxygen-Tolerant [NiFe]-Hydrogenase during H2 Cycling in the Presence of O2. J. Am. Chem. Soc. 135, 17897–17905 (2013). 3. Lauterbach, L., Idris, Z., Vincent, K. A. & Lenz, O. Catalytic Properties of the Isolated Diaphorase Fragment of the NAD+-Reducing [NiFe]-Hydrogenase from Ralstonia eutropha. PLoS One 6, e25939 (2011). 4. Birrell, J. A. & Hirst, J. Investigation of NADH binding, hydride transfer, and NAD+ dissociation during NADH oxidation by mitochondrial complex i using modified nicotinamide nucleotides. Biochemistry 52, 4048–4055 (2013). 5. Mostad, S. B. & Glasfeld, A. Using High Field NMR to Determine Dehydrogenase Stereospecificity with Respect to NADH. J. Chem. Educ. 70, 504–506 (1993). 6. Reeve, H. A., Lauterbach, L., Lenz, O. & Vincent, K. A. Enzyme-Modified Particles for Selective Biocatalytic Hydrogenation by Hydrogen-Driven NADH Recycling. ChemCatChem 7, 3480–3487 (2015). 7. Bertrand, M., Chabin, A., Brack, A. & Westall, F. Separation of amino acid enantiomers VIA chiral derivatization and non-chiral gas chromatography. J. Chromatogr. A 1180, 131–137 (2008). 8. Leiva de Faria, M. et al. Enantiodivergent syntheses of cycloheptenone intermediates for guaiane sesquiterpenes. Tetrahedron: Asymmetry 11, 4093–4103 (2000). 9. Shimoda, K. & Hirata, T. Biotransformation of enones with biocatalysts — two enone reductases from Astasia longa. J. Mol. Catal. B Enzym. 8, 255–264 (2000). 10. Knaus, T., Toogood, H. S. & Scrutton, N. S. in Green Biocatalysis 473–488 (John Wiley & Sons, Inc, 2016). doi:10.1002/9781118828083.ch18 11. Venema, A., Nibbering, N. M. M. & De Boer, T. J. Mass spectrometry of aralkyl compounds with a functional group—IX: Hydrogen scrambling in the molecular ion of hydrocinnamaldehyde. Org. Mass Spectrom. 3, 583–595 (1970). 12. Nibbering, N. M. M. & de Boer, T. J. Mass spectrometry of aralkyl compounds with a functional group—III: Specific exchange between some hydrogen atoms in the molecular ion of γ-phenylpropanol. Tetrahedron 24, 1415–1426 (1968).

S42