Construction of a third NAD+ de novo biosynthesis pathway

Only two de novo biosynthetic routes to nicotinamide adenine dinucleotide (NAD+) have been described, both of which start from a proteinogenic amino acid and are tightly controlled. Here we establish a C3N pathway starting from chorismate in Escherichia coli as a third NAD+ de novo biosynthesis pathway. Significantly, the C3N pathway yielded extremely high cellular concentrations of NAD(H) in E. coli. Its utility in cofactor engineering was demonstrated by introducing the four-gene C3N module to cell factories to achieve higher production of 2,5-dimethylpyrazine and develop an efficient C3N-based whole-cell bioconversion system for preparing chiral amines. The wide distribution and abundance of chorismate in most kingdoms of life implies a general utility of the C3N pathway for modulating cellular levels of NAD(H) in versatile organisms.


Introduction
Nicotinamide adenine dinucleotide (NAD + ) is a universally essential central metabolite consisting of adenosine monophosphate (AMP) linked to nicotinamide mononucleotide (NMN). Since its discovery as the first 'cozymase' over 100 years ago, two de novo biosynthesis pathways for NAD + have been reported (Fig. 1) 1 . Both pathways use a proteinogenic amino acid as the precursor for nicotinamide and converge at a common intermediate quinolinic acid (QA). In plants and most bacteria, the de novo biosynthesis of NAD + starts from L-aspartate (L-Asp), which is converted to QA by L-Asp oxidase (NadB) and quinolinate synthase (NadA) (Pathway I); in mammals, fungi, and some bacteria, nicotinamide is derived from conversion of L-tryptophan (L-Trp) to 3-hydroxyanthranilic acid (3-HAA) by four reactions of the kynurenine pathway (Pathway II). Oxidation of 3-HAA by 3-HAA 3,4-dioxygenase yields 2-amino-3-carboxymuconate semialdehyde (ACMS), which undergoes spontaneous ring closure to form QA [1][2][3] . The QA generated in both pathways is converted to NAD + via a common three-enzyme pathway (catalyzed by enzymes NadC-E): phosphoribosylation of QA to generate nicotinic acid mononucleotide, AMP addition to form nicotinic acid adenine dinucleotide, and a final amidation to afford NAD + . In addition to the two de novo biosynthesis pathways, NAD + is also regenerated by different salvage pathways using varied pyridines (nicotinic acid, nicotinamide, and nicotinamide riboside) as precursors 1,4 . NAD + and its reduced form NADH serve as hydride acceptor and donor, respectively, in thousands of redox reactions in vital metabolic pathways such as glycolysis, citric acid cycle, fatty acid degradation, and oxidative phosphorylation 5  NAD(H) deficiency is also medically significant; for example carcinoid syndrome patients, in whom L-Trp is 4 depleted by excessive production of serotonin, exhibit pellagra-like symptoms caused by NAD(H) deficiency 9 . NAD + has also been recently identified as the substrate for NAD + -consuming enzymes involved in signal transduction pathways that regulate crucial biological processes including DNA repair, transcription, and cell cycle progression [10][11][12] . In some cases, cellular levels of NAD(H) are reduced dramatically by NAD + -consuming enzymes; for example, the hyper-activation of poly(ADP-ribose) polymerase-1 reduced the NAD + pool by about 80% 11 .
Consequently, increasing cellular NAD(H) concentrations can stimulate the NAD + -consuming enzymes and modulate downstream biological processes. For example, increasing NAD(H) levels can activate sirtuins and improve mitochondrial homeostasis and extend the lifespan of different species 13,14 .
Several strategies have been developed to enhance the catalytic efficiencies of NAD(H)-dependent or NAD + -consuming enzymes by increasing cellular NAD(H) levels. These include supplementation with NAD + or its pyridine precursors, limiting NAD + consumption, and reinforcing NAD + salvage or de novo biosynthesis pathways [15][16][17] . However, each strategy has its limitations: supplementation is restricted by poor cellular uptake of NAD + ; reducing NAD + consumption and accelerating its recycling via the salvage pathways can only replenish but cannot expand cellular NAD(H) pools. Considering that the highest cellular NAD(H) levels are limited in large part by NAD + de novo biosynthetic capacity, manipulating these de novo pathways should efficiently expand cellular NAD(H) pools. The most frequently used strategy for increasing cellular NAD(H) levels is by reinforcing the NAD + salvage pathways instead of manipulating de novo biosynthesis. However, the few reported attempts to manipulate NAD + de novo pathways yielded only very limited 18 or nonexistent increases in cellular NAD(H) levels 19 , probably due to the stringent regulation of NAD + de novo biosynthesis at transcriptional 20 , translational 21 , and post-translational levels 22 . In addition, because both NAD + de novo biosynthesis pathways start from a proteinogenic amino acid, their activation may deplete amino acid pools required to efficiently overproduce proteins in engineered cell factories. 5 In this study, we decoupled NAD + de novo biosynthesis and protein synthesis by designing a C3N pathway as the third NAD + de novo biosynthesis pathway, which uses chorismate as the precursor for nicotinamide. The C3N pathway was constructed in E. coli by combining genes from secondary metabolism with the latter steps of pathway I. It effectively circumvents the tight regulatory controls on NAD + de novo biosynthesis and enables extremely high cellular concentrations of NAD(H) in the recombinant E. coli strains. We genetically packaged the C3N pathway as a four-gene C3N module for plug-and-play installation to significantly boost cellular NAD(H) concentrations in E. coli. Its utility in cofactor engineering was demonstrated by improving the bioconversion efficiency of 2,5-dimethylpyrazine (DMP) and by developing a C3N-based whole-cell system for efficient production of chiral amines. 6

Conceptual design of a third NAD + de novo biosynthesis pathway
The conceptual basis for an alternative NAD + de novo biosynthesis pathway arose from the observation that several secondary metabolites contain structures derived from 3-HAA [23][24][25][26][27][28] (Supplementary Fig. 1a), which also occurs in primary metabolism as a key intermediate in NAD + de novo biosynthesis pathway II (Fig. 1). The biosynthetic gene clusters encoding these natural products indicate that chorismate is converted to 3-HAA by three sequential reactions catalyzed by 2-amino-2-deoxyisochorismate (ADIC) synthase, 2,3-dihydro-3-hydroxyanthranilic acid (DHHA) synthase, and DHHA dehydrogenase ( Fig. 1 and Supplementary Fig. 1). Chorismate is an ideal precursor for a new pathway because as a natural branch point for many primary and secondary metabolic processes it is already metabolically promiscuous and abundant in most cells 29,30 . We therefore designed a third de novo biosynthesis pathway to NAD + by combining the chorismate-to-3-HAA pathway with 3-HAA 3,4-dioxygenase (the enzyme converting 3-HAA to QA) and the common three-step process of pathways I and II converting QA to NAD + . This synthetic route was designated as C3N pathway based on its precursor chorismate, the key intermediate 3-HAA and the final product NAD + , and the four enzymes converting chorismate to QA comprised the C3N module.
Stringent regulatory control of NAD + de novo biosynthesis is directed mainly at the first biosynthetic genes like nadB and occurs at the transcriptional, translational, and post-translational levels [20][21][22] . By constructing the C3N pathway using defined promoters, ribosomal binding sites, and enzymes from secondary metabolism, these regulatory controls may theoretically be circumvented. Moreover, the recruitment of chorismate as the precursor for nicotinamide should decouple the C3N pathway from protein synthesis and make it suitable for use in engineered cells that overexpress proteins.
Characterization of Pau20 as a DHHA dehydrogenase 7 Although the ADIC synthase and DHHA synthase enzymes catalyzing the first two steps of the C3N pathway have been well studied 31 , the third enzyme DHHA dehydrogenase has yet to be biochemically characterized ( Supplementary Fig. 1). We chose the putative DHHA dehydrogenase Pau20 from the paulomycin biosynthetic gene cluster for characterization 27 . The pau20 gene was inactivated by gene replacement in Streptomyces paulus NRRL 8115 to construct the S. paulus pau20::aac(3)IV mutant ( Supplementary Fig. 2a), in which the production of paulomycins was totally abolished. In feeding experiments of S. paulus pau20::aac(3)IV, the production of paulomycins could be restored by 3-HAA but not by DHHA, indicating that Pau20 is responsible for the conversion of DHHA to 3-HAA (Fig. 2a). N-His6-tagged Pau20 was then overexpressed in E. coli BL21, purified, and incubated with DHHA and NAD + . Efficient oxidation of DHHA to 3-HAA was observed, verifying Pau20 as a DHHA dehydrogenase ( Fig. 2b and 2c).
Previous studies showed that 3-HAA is prone to oxidation under alkaline conditions and exposure to air 32 . Indeed, when the Pau20 assay was performed at buffers over pH 7.2, spontaneous oxidation of 3-HAA could be observed ( Supplementary Fig. 3a). Therefore, Pau20 characterization was first performed at pH 7.0 and its optimal temperature of 37 o C ( Supplementary Fig. 3b). Steady-state kinetic analysis of Pau20 under these conditions revealed Michaelis-Menten behavior for all substrates and kinetic constants consistent with Pau20 acting as an efficient DHHA dehydrogenase ( Fig. 2d and Supplementary Fig. 3f). Characterization Pau20 in vitro not only supported a useful DHHA dehydrogenase for testing the designed C3N pathway, but for the first time unambiguously verified the chorismate to 3-HAA pathway originally proposed for a number of natural products.

Constructing the C3N pathway in E. coli
The feasibility of the C3N pathway was initially tested by enabling 3-HAA production in E. coli, which lacks the kynurenine pathway and therefore cannot synthesize 3-HAA. A pau20-phzDE cassette consisting of pau20 and two genes encoding the well-characterized ADIC synthase (PhzE) and DHHA synthase (PhzD) from the phenazine 8 biosynthesis pathway of Pseudomonas aeruginosa PAO1 33, 34 was synthesized and inserted into the medium-copy-number plasmid pXB1s downstream of the arabinose-inducible promoter PBAD to construct pXB1s-HAA. This plasmid was introduced into E. coli BW25113 to generate E. coli BW-pXB1s-HAA, which produced 3-HAA at a titer of 9.91 ± 0.14 mg/L (Fig. 3a).
To establish the complete C3N pathway in E. coli, the nbaC gene encoding an efficient 3-HAA 3,4-dioxygenase from the aromatic compound degradation pathway of Pseudomonas fluorescens KU-7 35 was inserted into plasmid pXB1s-HAA to generate pXB1s-QA, in which the four-gene cassette nbaC-pau20-phzDE dictates a C3N module converting chorismate to QA. In addition, to test whether cells could survive with the synthetic C3N pathway as the sole de novo source of NAD + , we constructed E. coli ∆nadAB ( Supplementary Fig.   2b), a null mutant of NAD + de novo biosynthesis, which was unable to grow on M9 plate unless QA was supplemented (Fig. 3b). Indeed, growth could be restored by introducing pXB1s-QA into E. coli ∆nadAB to afford E. coli ∆nadAB-pXB1s-QA, a recombinant strain with the complete C3N pathway. It grew well on M9 plates when the inducer arabinose was added. As a control, E. coli ∆nadAB with empty pXB1s could not grow on M9 plates even when supplemented with arabinose ( Fig. 3c). Cellular NAD(H) levels measured at early stationary phase in liquid M9 medium with 10 mM arabinose were slightly higher in the C3N strain E. coli ∆nadAB-pXB1s-QA (1.18 ± 0.18 mM) than in E. coli BW25113 in M9 medium with (0.85 ± 0.05 mM) or without arabinose (0.89 ± 0.04 mM) (Fig. 3d). These results indicate that the C3N pathway was active in E. coli ∆nadAB-pXB1s-QA and could supply NAD + efficiently for its growth.

Optimizing the C3N pathway in E. coli
To assess the potential of the C3N pathway for improving cellular NAD(H) levels, we cloned the nbaC-pau20-phzDE cassette into the high-copy-number vector pAB1s. Transformation of the resultant plasmid into E. coli ∆nadAB afforded the strain E. coli ∆nadAB-pAB1s-QA with cellular NAD(H) concentrations as high as 4.43 9 ± 0.14 mM. HPLC analysis of the E. coli ∆nadAB-pAB1s-QA metabolic profile revealed that DHHA was accumulated in the fermentation broth ( Supplementary Fig. 3c), implying that NAD(H) levels could be further increased by using more efficient DHHA dehydrogenases. The 9.7-fold increase in cellular NAD(H) concentration generated solely by the C3N pathway compares favorably with previous cofactor engineering efforts. Specifically, cellular NAD(H) levels could be enhanced 7-fold (to 7.03 mM) in E. coli BL21 by reinforcing NAD + salvage via overexpression of the nicotinic acid phosphoribosyltransferase PncB and NAD + synthetase NadE 15 . Other attempts to either limit NAD + consumption or reinforce NAD + de novo pathways achieved no more than 4-fold increases to cellular NAD(H) levels 17,18 .
Interestingly, an upper limit to cellular NAD(H) concentrations was proposed for the E. coli BW25113-derived strain YJE003. By inactivating nadE to block NAD + synthesis and expressing the NAD(H) transporter ntt4 gene to enable NAD + intake, the maximum NAD(H) level of YJE003 was 9.6-fold (8.5 mM) higher than BW25113 36 .
Similarly, our C3N-generated cellular NAD(H) concentration for E. coli ∆nadAB-pAB1s-QA* was 9.7-fold (9.3 mM) higher than BW25113, suggesting that the C3N pathway is optimized for NAD(H) yield and therefore harbors great potential for more general applications that require expanded cellular NAD(H) pools.

Increasing DMP production via the C3N module
We envisioned that the C3N pathway converting chorismate to QA may represent a powerful tool to increase cellular NAD(H) levels. Specifically, we can consider the advantages of a packaged 'C3N module': (i) it can be easily introduced into targeted strains as a four-gene cassette; (ii) the cellular NAD(H) pools can be expanded significantly by the established C3N pathway in E. coli; and (iii) it is compatible with other NAD(H) level increasing strategies (Fig. 4a). We first tested this by introducing pAB1s-QA* into E. coli DMP, a cell factory for producing the food flavor additive 2,5-dimethylpyrazine (DMP), which is also the synthetic precursor of 5-methyl-2-pyrazinecarboxylic acid, a key component of widely used pharmaceuticals like Glipizide and Acipimox 37 . E. coli DMP overexpresses two genes encoding Tdh and SpNox. Tdh is an NAD + -dependent L-threonine-3-dehydrogenase that oxidizes L-threonine to (2S)-2-amino-3-oxobutanoate, which then spontaneously undergoes decarboxylation and dimerization to yield DMP 38 ; SpNox is an NADH oxidase (H2O forming) for NAD + recycling 39 . Compared with E. coli DMP-Con (E. coli DMP with empty pAB1s), the NAD(H) levels of the new strain E. coli C3N-DMP increased 2.7-fold, and DMP production was 3.8-fold higher (Fig. 4b). These results supported the application of the C3N module as a simple and effective tool to increase cellular NAD(H) levels and boost yields in microbial cell factories.

Construction of a C3N-based whole-cell system for preparing chiral amines
Encouraged by the convenience and efficacy of the C3N module towards boosting DMP production, we next sought to develop a C3N-based whole-cell system for preparing chiral amines. Chiral amines are important for 11 synthetic intermediates of pharmaceuticals and other bioactive molecules 40 . Preparative scale amination of alcohols to enantiomerically pure chiral amines has been achieved using a biocatalytic hydrogen-borrowing cascade employing an NAD + -dependent alcohol dehydrogenase (ADH) coupled with an NADH-dependent amine dehydrogenase (AmDH) to enable efficient internal recycling of the nicotinamide coenzyme 40 (Fig. 4c). A recent report described the first whole-cell conversion of alcohols to chiral amines by combining an enantioselective AA-ADH with a chimeric amine dehydrogenase Ch1-AmDH. The bioamination efficiencies in those whole-cell systems were enhanced by increasing cellular NAD(H) levels via the supplementation of NAD + to lyophilized E. coli resting cells (external NAD + cannot diffuse into cells without lyophilization) or by adding glucose to promote NADH recycling. This system could transform the starting alcohol (S)-1a to the chiral amine product (R)-1c (amphetamine) with 46 ± 14% yield (Fig. 4c). Adding the R-selective LBv-ADH enabled conversion of the racemic alcohol substrate rac-1a to product (R)-1c with 21 ± 6% yield 41 .
The critical role of the expensive reagent NAD(H) in this dual-enzyme system suggested it could benefit from C3N-boosted NAD(H) levels, and inspired us to develop an efficient whole-cell system for preparing chiral amines.
Although incorporating the C3N module led to 5.6-fold higher conversion to the final product, further optimization was required to increase the bioamination efficiency of the whole-cell system. We therefore optimized the whole-cell bioconversion procedure for E. coli C3N-ChA1 with rac-1a ( Supplementary Fig. 4a-d) C3N-ChA3 (Fig. 4c and Supplementary Fig.   4e).
Having constructed a set of strains with variable bioamination efficiencies, we sought to assess whether these efficiencies could be linked to C3N-generated NAD(H). Indeed, NAD(H) measurements revealed a clear positive correlation between bioamination efficiency and cellular NAD(H) levels of the resting cells (Fig. 4c). To further verify the effect of the C3N pathway on bioamination, a second control strain E. coli ChA3-Con was constructed by transforming pRSF-TesADH-CalAmDH into BW-Con. Under the optimized conditions the cellular NAD(H) levels of E. coli ChA3-Con (3.60 ± 0.47 mM) were much lower than that of E. coli C3N-ChA3 (7.03 ± 1.57 mM), and the conversion to final amine product was halved (18.7 ± 0.5%, (R)-1c) (Fig. 4c). When the concentration of rac-1a was increased to 20 mM, the conversion to product (R)-1c was much lower for E. coli ChA3-Con (6.7 ± 1.3%) than that for E. coli C3N-ChA3 (48.3 ± 4.1 %, Fig. 4d), further highlighting the important contribution of the C3N pathway.
The C3N-based whole-cell system for chiral amine preparation revealed that the C3N module can be used as a powerful and expedient tool for cofactor engineering in E. coli. Moreover, we believe it has the potential to be used in a versatile range of host organisms because: (i) chorismate, the precursor of the C3N pathway, is a product of the shikimate pathway, which exists widely in bacteria, archaea, fungi, and plants 44 ; (ii) the shared final three steps of NAD + de novo biosynthesis pathways I and II exist in almost all organisms. In most cases, the only requirement for constructing the C3N pathway is to express the C3N module genes in the targeted organisms.

Construction of S. paulus pau20::aac(3)IV
A blue-white screening-based gene inactivation system was used to inactivate gene pau20 49 . The 1.9-kb upstream fragment and the 2.2-kb downstream fragment of pau20 were amplified using primer pairs pau20-s2/pau18R and pau22-R/pau20-R and inserted into the MunI and BlnI sites of pCIMt002 respectively to afford the pau20 23 disruption plasmid, which was then introduced into S. paulus NRRL 8115 via E. coli-Streptomyces conjugation.
Both blue and white exconjugants were obtained on the plate. One of the white exconjugants with apramycin resistance was selected as the desired pau20 gene inactivation mutants S. paulus pau20::aac(3)IV, which was verified by PCR with primer pair pau20-s2/ pau20-R (Supplementary Fig. 2a).

Production of paulomycins
For paulomycin production, 50 μL spores of S. paulus NRRL 8115 or pau20::aac(3)IV were inoculated into GS-7 liquid medium and cultured at 28 o C, 220 rpm for 2 days. The resulting seed culture was inoculated into 50 mL R5α liquid medium at a 2% ratio (v/v). After 4 days fermentation, the broth was harvested by centrifugation, extracted with 50 mL ethyl acetate three times, and concentrated in vacuo. The samples were re-dissolved in 1 mL acetonitrile and subjected to HPLC analysis.
DHHA or 3-HAA was added to the cultures to a final concentration of 2.5 mM two days after seed inoculation.
After culturing at 28 o C, 220 rpm for 2 days, the broth was harvested, extracted, and concentrated as described above. The samples were re-dissolved in 1 mL acetonitrile and analyzed by HPLC.

Expression and purification of the DHHA dehydrogenases
The 0.7-kb pau20 gene was cloned using the S. paulus NRRL 8115 genome as a template with primer pair pauN10ES/pauN10ER, verified by sequencing, and inserted into the NdeI/BamHI sites of pET-28a to generate pET-28a-pau20. A single transformant of E. coli BL21(DE3)/pET-28a-pau20 was inoculated into LB medium with

Enzymatic assays of the DHHA dehydrogenases
The DHHA dehydrogenase activity of Pau20 was tested in a 100 μL mixture containing 50 mM PBS buffer (pH

Construction and evaluation of E. coli BW-pXB1s-HAA
A 2.6-kb DNA fragment containing the 0.6-kb phzD (PhzD Accession: AAC64487) and the 1.9-kb phzE genes (PhzE Accession: AAC64488) was synthesized and the homolog sequences for ligation independent cloning at its both ends were added by a PCR amplification using primer pair pHAAphzDE-F/pHAAphzDE-R. The 0.7-kb DNA fragment containing the pau20 gene was amplified from the S. paulus NRRL 8115 genome using primer pair pHAApau20-F/pHAApau20-R. The two DNA fragments were inserted into the NcoI/EcoRI sites of pXB1s (p15A origin, medium-copy-number) using one-step cloning kit (Vazyme, Nanjing, China) to generate pXB1s-HAA. E.
coli BW-pXB1s-HAA was constructed by introducing pXB1s-HAA into E. coli BW25113. E. coli BW-pXB1s-HAA was cultured in 3 mL LB medium with 50 μg/mL streptomycin at 37 o C overnight. The seed culture was then inoculated into 50 mL M9 medium with 50 μg/mL streptomycin at 1% ratio (v/v). After cultured at 37 o C, 220 rpm for 3 h, 10 mM arabinose was added and the culture was further fermented at 37 o C, 220 rpm for 20 h. The broth was centrifuged to remove the cell pellet and the supernatant was subjected to HPLC analysis directly.  Fig.2b). The 0.6-kb nbaC (NbaC Accession:

Construction of E. coli ∆nadAB-pXB1s-QA, ∆nadAB-pAB1s-QA
WP_006094629) was synthesized and the homolog sequences for ligation independent cloning at its both ends were added by a PCR amplification using primer pair NbaC-F/NbaC-R. The nbaC gene was then inserted into the NcoI site of pXB1s-HAA to generate pXB1s-QA, which was transformed into E. coli ΔnadAB to obtain E. coli ∆nadAB-pXB1s-QA.
Plasmids pAB1s-HAA (ColE1 origin, high-copy-number) was constructed similarly as that of pXB1s-HAA by insertion of the 2.6-kb fragment containing genes phzDE and the 0.7-kb fragment containing gene pau20 into the NcoI/EcoRI sites of pAB1s. The 0.6-kb nbaC gene was then inserted into the NcoI sites of pAB1s-HAA to generate pAB1s-QA, which was transformed into E. coli ΔnadAB to obtain E. coli ∆nadAB-pAB1s-QA.
The 0.7-kb dhbX gene was synthesized and the homolog sequences for ligation independent cloning at its both ends were added by a PCR amplification using primer pair DhbX-F/DhbX-R. It was then inserted into the XbaI site of pAB1s-QA to replace the pau20 gene (excised by XbaI digestion) to generate pAB1s-QA*, which was transformed into E. coli ΔnadAB to obtain E. coli ∆nadAB-pAB1s-QA*.

Growth of E. coli strains on M9 plates
The E. coli strains were cultured in 3 mL LB medium with appropriate antibiotics at 37 o C, 220 rpm overnight and

Measurement of intracellular NAD(H) concentrations
The E. coli strain was cultivated in 3 mL LB medium with appropriate antibiotics at 37 o C, 220 rpm overnight, and the overnight culture was then inoculated into 3 mL or 50 mL M9 medium at 1% ratio (v/v) and

Spectroscopic analysis
HPLC analysis was carried out on a Shimadzu HPLC system (Shimadzu, Kyoto, Japan). The analysis of paulomycin production was performed using an Apollo C18 column (5 μm, 4.6 × 250 mm, Alltech, Deerfield, IL, USA), which was developed with a linear gradient using acetonitrile and water with 0.1% trifluoroacetic acid at a flow rate of 0.8 mL/min. The ratio of acetonitrile was maintained at 5% for 5 min and changed linearly from 5% to 90% over 5-25 min and from 90% to 100% over 25-30 min. The detection wavelength was 320 nm.
The formation of NADH in the DHHA dehydrogenase reactions was monitored by BioTek Synergy H4 Hybrid Reader (BioTek, Vermont, USA) at 340 nm. HPLC analysis of the DHHA dehydrogenase reactions and production of 3-HAA in E. coli was carried out with a ZORBAX SB-Aq StableBond Analytical column (5 μm, 4.6 × 250 mm, Agilent Teologies, Santa Clara, CA, USA). A linear gradient of acetonitrile and water with 0.1% trifluoroacetic acid was used for development of the column at a flow rate of 0.8 mL/min. The ratio of acetonitrile was changed linearly from 1% to 50% over 0-30 min, from 50% to 100% over 30-32 min, and maintained at 100% for 5 min. 31 The detection wavelength was 254 nm or 294 nm.
The bioamination samples were analyzed using an Apollo C18 column (

Chemicals
The racemic alcohols 3a and 5a were chemically synthesized by reduction of the commercially available ketones 3b and 5b; the ketone 1b was chemically synthesized by oxidation of 1a. Enantiomeric standard (S) and (R) amines of 1c-5c were synthesized by stereoselective amination using previously described enzymatic methods 51 .