Generation of Oxidoreductases with Dual Alcohol Dehydrogenase and Amine Dehydrogenase Activity

Abstract The l‐lysine‐ϵ‐dehydrogenase (LysEDH) from Geobacillus stearothermophilus naturally catalyzes the oxidative deamination of the ϵ‐amino group of l‐lysine. We previously engineered this enzyme to create amine dehydrogenase (AmDH) variants that possess a new hydrophobic cavity in their active site such that aromatic ketones can bind and be converted into α‐chiral amines with excellent enantioselectivity. We also recently observed that LysEDH was capable of reducing aromatic aldehydes into primary alcohols. Herein, we harnessed the promiscuous alcohol dehydrogenase (ADH) activity of LysEDH to create new variants that exhibited enhanced catalytic activity for the reduction of substituted benzaldehydes and arylaliphatic aldehydes to primary alcohols. Notably, these novel engineered dehydrogenases also catalyzed the reductive amination of a variety of aldehydes and ketones with excellent enantioselectivity, thus exhibiting a dual AmDH/ADH activity. We envisioned that the catalytic bi‐functionality of these enzymes could be applied for the direct conversion of alcohols into amines. As a proof‐of‐principle, we performed an unprecedented one‐pot “hydrogen‐borrowing” cascade to convert benzyl alcohol to benzylamine using a single enzyme. Conducting the same biocatalytic cascade in the presence of cofactor recycling enzymes (i.e., NADH‐oxidase and formate dehydrogenase) increased the reaction yields. In summary, this work provides the first examples of enzymes showing “alcohol aminase” activity.


Introduction
The concept of enzyme promiscuity referstoa ne nzyme's ability to catalyzem echanistically distinct reactions (i.e.,c atalytic promiscuity) or convert structurally diverse substrates by following the same mechanism (i.e.,s ubstrate promiscuity). [1]The promiscuous catalytic behavior of enzymes has been harnessed for chemical synthesis and has served for the evolution of variants possessing novel catalytic activities and/or enhanced substrate scope. [2]In some cases, the enzyme's catalytic or substrate promiscuity can be simply tuned by changing the reactionc onditions (i.e.,c ondition promiscuity), ac lassical example of which is the catalytic activity of hydrolases in non-aqueous media wherein (trans)esterification and amidation reactions of carboxylic acids ande sters are enabled. [3]lcohol dehydrogenases (ADHs, EC 1.1.1.X) from the nicotinamide adenine dinucleotide-o rn icotinamide adenine dinucleotide phosphate-dependent [NAD(P)] category catalyzer eversible interconversion between alcohols and carbonyl compounds. [4]4b, d, e] For instance, ADH-catalyzed oxidation of secondary alcohols can be exploited for the kinetic resolution or even deracemization of racemic alcoholm ixtures. [5]6c] Amine dehydrogenases (AmDHs,E C1 .4.1.X) catalyzet he reversible reductive amination of carbonyl compounds at the sole expense of ammonia and NAD(P)H, the latter of which is appliedi nacatalytic amount and recycled with established methods. [14]14a, 15] Our group has discovered the catalytic promiscuity of AmDHs for the synthesis of secondary or tertiary amines, and other groups have also recently investigated this property. [16]15k] The best variant possessedaF 173A single mutation, which created an ew hydrophobic cavity in the active site wherein aromatic ketones can be accommodated and converted into a-chiral amines with excellent enantioselectivity.N otably,w er ecently observed that the wild type LysEDH is also capable of producing primary alcohols starting from aromatica ldehydes.The capability of an oxidoreductase to reduce both C=Oa nd C=Nb onds was rarely observedu ntil to date.2s] They later discovered thatashort chain dehydrogenase/reductase (SDR)-namely the noroxomaritidine reductase from Narcissus pseudonarcissus (NR)-could reduce C=C( i.e.,o fenones), C=Oa nd C=Nb onds, whereas another SDR from Zephyranthes treatie also exhibited dual C=Oa nd C=Na ctivity. [17]Furthermore, aS DR from Methylobacterium sp.77 that only exhibits ketoreductase activity was recently engineered to gain imine reductase activity by introducing four mutations in its active site to resemble some of the structural features of the NR reductase. [18]Notably,i na ll of these cases, the ketoreductase and the imine reductasea ctivities were strictly substrate dependent;f urthermore, the SDR enzymes were active toward preformed (cyclic) imines but reductive amination between ac arbonyl compound and an amine donor was not reported.Conversely, other groupsh ave independently reported that few imine reductases (IReds)a nd reductivea minases (RedAms) possess promiscuousk etoreductasea ctivity on very specific substrates such as tri-, di-andm ono-fluorinated acetophenones at the terminalcarbon position. [19]n the presentw ork, we studied the catalytic promiscuity and exploited the high evolvability of LysEDH to create new variants that possess enhanced ADHa ctivity or even both AmDH activity (i.e.,f or the reductivea mination of carbonyl compounds with free ammonia) and ADH activity.T he best variant exhibiting dual ADH-AmDH activity was harnessed to accomplish the first example of one-enzyme hydrogen-borrowing amination of benzylic alcohol.

Compoundselection
We conducted this study with ap anel of aldehydes and ketones as depicted in Figure 1.Group Ac omprises substituted benzaldehydes and arylaliphatica ldehydes, whereas Group B comprises aromatic,a rylaliphatica nd aliphatic ketones.W e tested aldehydes (20 mm)o rk etones( 10 mm)f or reduction to the corresponding alcohols and reductive amination to the correspondingp rimary amines, the latter in the presence of 2 m ammonium/ammonia species (1b-21 b,S cheme 1).
However,b iocatalytic transformationso fa ldehydes can sometimes be complicatedd ue to their volatility andl imited extractability from an aqueous medium.Therefore, we initially checked if the selected aldehydes (Figure 1a nd Supporting Information Figure S1, 1b-9b and 22 b-26 b), the corresponding primary alcohols (1a-9a and 22 a-26 a), and terminal amines (1c-9c and 22 c-26 c)w ere not excessively volatile during incubation for 24 ha nd could then be efficiently extractedf rom the aqueous buffer (see Supporting Information section 3f or details).Aldehydes that could be recovered with analytical yields between 66 %a nd 90 %u poni ncubation in an aqueous buffer (see Supporting Information, Table S1 for details) were considered in this study (Figure 1, Group A).However,a lthought hey were recovered in 79 %a nd 88 %a nalytical yields, respectively (Supporting Information, Table S1), aldehydes 25 b and 26 b were not included in this study because the related alcohols (25 a, 26 a)a nd amines (25 c, 26 c)t hat are attainable from the reduction of aldehydes were recoveredi no nly an 8% yield or not recovered at all following incubation for 24 h.In contrast, all of the selected ketones (Figure 1, Group B) could be included in this study because we did not observe any significant volatility or extractabilityi ssues.

Reductive amination of aldehydes
The reductive amination of aldehydes wasp erformed using wild-typeL ysEDH,L E-AmDH-v1 [15k] -the latter of whiche xhibited high activity toward the reductive amination of benzaldehyde-and four additional LE-AmDH variants that were originally designed to reductively aminate bulky-bulky ketones.These four new variants were generated by mutating the amino residues of LysEDH (i.e.,Y 238A and/or T240A)t hat are involved in the interaction between the enzyme's active site and the a-amino group of the natural substrate l-lysine.We combined these new mutations with either the F173S or F173A mutation,t he latter of which was already present in the LE-AmDH first generationv ariants; [15k] this resulted in LE-AmDH-v22 (LysEDH Y238A/F173A), LE-AmDH-v24 (Y238A/ F173S), LE-AmDH-v25 (Y238A/T240A/F173A), and LE-AmDH-v27 (Y238A/T240A/F173S).Detailed information regarding primers for mutations and biocatalysts preparation is reported in Supporting Information section 4. Notably, LE-AmDH-v22a nd v25 possess the F173A mutation as the first-generation variant LE-AmDH-v1;t herefore, these new variants could possibly retain the amine dehydrogenase activity of the parentL E-AmDH-v1 toward aromatic substrates.Our previous study also demonstrated that the LE-AmDH enzyme family catalyzesa genuiner eductive amination by sourcingt he carbonylc ompound and ammonia/ammonium from the reactionm edium.15k] In this context, Nestl's group showedt hat spontaneous imine formation from 1b in aqueous medium is negligible (or does not occur at all) at pH levels below 8. [20] Furthermore, LE-AmDH-v1 exhibited higha ctivity for the reductivea minationo fa cetophenone (11 b)w ith NH 3 /NH 4 + at pH levels rangingf rom 7t o9 .5. [15k] Under such conditions, the spontaneous formation of the imine of 11 b was never observed in an aqueous buffer. [20]These observations confirmt hat LE-AmDHs catalyzet he reactionb etween a carbonyl compound and free NH 3 /NH 4 + along with ap ossible reduction of any pre-formed aldimine in solution at more basic pH values.
Finally,t he F173S mutation was introduced into LE-AmDH-v24 and -v27 in order to investigate the effects of ah ydrophilic, non-bulky residue in this position.In all of the reactions, two samples were included as negative controls( NC1 and NC2, respectively).NC1o nly contained the recombinantf ormate dehydrogenase from Candida boidinii (Cb-FDH) and NAD + to verify that neither amine nor alcohol product formation is obtained as the result of apossible promiscuouscatalytic activity of the cofactor-recycling enzyme.N C2 contained none of the enzymes, and only NAD + and the aldehydes ubstrate were added to the buffer.T able 1s ummarizes the results obtained for the reductive amination reactions of selected aldehydes 1b-9b (20 mm;s ee Experimental Section and Supporting Information Section 5.1 for detailso nb iocatalytic reactions and quantitative analytical determination).The reactions were run at 30 8Ci na na mmonium formate buffer (2 m, pH 8.2-9) in the presence of NAD + (1 mm), Cb-FDH (16 mm) and LE-AmDH enzyme( 45 mm).It is evidentt hat all of the variants possessing the F173A mutation (v1, v22 and v25) performed better in the reductive amination of aldehydes than those possessing the F173S mutation (v24 and v27).In particular,L E-AmDH-v22 generally performed slightly better than LE-AmDH-v1 and -v25, as it yielded the highest amine formation for the reductiono fs ubstituted benzaldehydes 3b-7b (11-> 99 %).The reductive amination of benzaldehyde( 1b)a nd para-fluorobenzaldehyde (2b)e ssentially proceeded equally well with LE-AmDH-v1, -v22 and -v25.Compound 1b was the most converted substrate (> 99 %y ield), whereas 2b was aminated in 89-91 %y ield by the three variants.However,n one of the variants could aminate either phenylacetaldehyde (8b)o r 3-phenylpropanal (9b).As mentioned above, the LE-AmDH variants containing the F173S mutation (v24 and v27) exhibited al ower capability of aminating aldehydes, although LE-AmDH-v24 generally yielded slightly highera mine formation for the reduction of 1b, 3b and 4b (11-80 %).Notably,w ed id not observe any amine formation in any of the negative control reactions (NC1 andN C2), thereby provingt hat only the LE-AmDHscatalyze the reductivea mination of aldehydes.
Table 1.a] %A nalyticalyields of amine [b] and (alcohol) [c] Sub.W Tv 1v22 v24 v25 v27 Surprisingly,w ealso observedt hat the wild-type LysEDH and both variants containingt he F173S mutation produced significant amounts of alcoholb y-products (4-33 %) in a number of biocatalytic reductionsi na mmonium buffers.In particular, the alcoholy ield was greater than the amine formation for the reductionso fb enzaldehyde and derivatives 1b-3b and 3-phenylpropanal( 9b)c atalyzedb yL ysEDH.The alcohols were also the main products of the reductions of 3b and 9b catalyzed by LE-AmDH-v27.Despite the presence of ammonium species in the reactionm edium, 3-phenylpropanol (9a) was the only product of the reduction of 9b catalyzed by LysEDH,L E-AmDH-v22,-v24 and -v27, whereas LE-AmDH-v1 was inactive toward 9b.L E-AmDH-v27 produced the highest alcohol yield (18-33 %) in an ammoniumb uffer for all the substrates that exhibited this behavior (1b-3b, 5b, 9b).When incubated with 1b and 2b,w ild-type LysEDH produced equal amountso fa lcohol as LE-AmDH-v27( 23 %a nd 28 %a lcohol yields, respectively).Conversely,w ith the exceptiono fL E-AmDH-v22 with 9b (9 %a lcohol yield), none of the reactions catalyzed by LE-AmDHs variant possessing the F173A mutation (v1, v22 and v25) resulted in any alcohol formation.M oreover, alcohol formation was never observed in the negative control reactions (NC1 and NC2), thereby proving that the introduction of F173S mutations omehow promoted promiscuousa lcohol formation.T his phenomenon was furtherc onfirmed in biocatalytic reactions in which the reductions were run in an ammonia-freee nvironment (described later), thus precluding any amine product formation.
The stereoselective outcome of the reductive amination was measured for all cases in which sufficient conversion was achieved.Notably,t here was alwaysa tl east one LE-AmDH variant that could produce the amine products with > 99 % ee (R).In general,t he stereoselectivity of the reaction was always perfect except for LE-AmDH-v1 with 13 b and 14 b and LE-AmDH-v22 with 13 b.

Investigation on the promiscuous reduction of aldehydes to alcohols
As mentioned above,t he reductive amination of aldehydes catalyzed by wild-type LysEDH andv ariants LE-AmDH-v24a nd -v27 led to alcohol product formation although the reaction mediumc ontained ammonia/ammonium ions.In particular, LE-AmDH-v27 produced the highest amount of alcohol among all of the tested variants.T herefore, we envisioned that promiscuous alcoholf ormationc atalyzed by LE-AmDH-27 could be enhanced in an ammonia-free environment, in which the imine intermediate for the reductive amination cannotb eg enerated.Therefore, we investigated the reduction of the test substrate benzaldehyde (1b,2 0mm)t ob enzyla lcohol (1a)c atalyzed by LE-AmDH-v27 (45 mm).The optimum of pH for this biocatalytic transformation was initially investigated using the Britton-Robinson universal bufferi np Hl evels rangingf rom 6.5-9.0.R esults showedt hat the highest velocity (57 mm min À1 )w as obtained at pH 7( Figure2;s ee Supporting Information section 5.2, Ta ble S3 for details).15k] However,F igure 2 shows that the apparent rate for the reduction of 1b to 1a decreaseda pproximately 4-fold at pH 9-9.5 compared with the apparent rates obtaineda tp H7.
In the next step, the progress of the reduction of 1b to 1a was monitored using five different types of buffers (HEPES, Tris, MOPS, KPi, NaPi)a tp H7 and 100 mm concentration (see Supporting Information section5.3and Figure S3).These results showed that LE-AmDH-v27 has no apparent preference for ac ertain type of buffer because the reaction progress curves essentially overlapped and the reactions alwaysresulted in !98.6 %c onversion after 24 h( corresponding to ! 18.8 mm of formed 1a).Therefore, we decided to continue our study using the KPi buffer.
Using the optimal reactionc onditions for the reduction of 1b to 1a (KPi buffer 100 mm,p H7;L E-AmDH variant 45 mm, Cb-FDH1 6mm,N AD + 1mm,H COONa 100 mm), we investigated the promiscuousa lcohold ehydrogenasea ctivity of the LE-AmDH variants.Table 3r eportst he results for the reduction of aromatic and arylaliphatica ldehydes( Scheme 1, group A) to the correspondingp rimary alcohols using the six selected variants (see Experimental Sectiona nd Supporting Information section 5.4 for details).a] %A nalyticalyieldofa lcohol [b] Sub.
WT  version).C onversely,o nly LE-AmDH-v27 could reduce orthofluorobenzaldehyde (4b)t ot he corresponding alcohol with a moderate yield (40 %);t he other variants exhibited poor activity (max 11 %c onversion).Furthermore, none of the enzymes could convert ortho-methylbenzaldehyde (7b).These results are different from the data obtainedf or the reductive amination reactions in which LE-AmDH-v1,-v22 and -v25 could aminate 4b with 98-> 99 %c onversionsa nd 7b with 87-92 % conversions( Ta ble 1).Therefore, the low or lack of catalytic activity for the reduction of 4b and 7b to the corresponding alcohols 4a and 7a by the LE-AmDH variants cannotbea ttributed to particulars teric (i.e.,f or 7b)o re lectronic (i.e.,f or 4b)e ffects.The different reactivity is ue to the varying distances and orientations between the prochiral carbon atom of the ligand (i.e.,k etone substrate for ADH reactiono ri minium intermediate for AmDH reaction) and the departing hydride of the NADH coenzymei nt he enzyme's active site.For instance, analysiso ft he X-ray structures of a l-phenylalanine amino acid dehydrogenasef rom Rhodococcus sp.demonstrated that this critical distance changes from more than 5 (i.e.,u nproductive binding)t o~3-3.5 (i.e.,p roductiveb inding) when the ketone ligand is converted into its imine/iminium intermediate during the catalytic cycle in the enzyme's active site. [21]16a] In anothers tudy, Grogan's and Turner's groupsc rystallizedareductive aminase (AtRedAm) with ak etone substrate and NADPH coenzyme in the active site, and a4 .5 distance was observed between the ketone's prochiral carbon and the hydride of NADPH, thereby precluding any substrate reduction. [22]ther authors have also attempted to explain the different carbonyl reductase and imine reductasea ctivitieso fI Reds and RedAmsb ased on the determination of Gibbs energy barriers for the hydride transfer. [19]Although these calculations correlated with the experimentally observed reduction of 2,2,2trifuoroacetophenonet ot he related alcohol, compared with other non-converted ketones,a ninspection of the X-ray crystal structurer evealed that the two fluorine atoms of the substrate directly interactw ith one hydroxy group of the ribose of NADP;t his interaction pushed the ketone substrate toward NADP in such am anner that ac loser distance between the prochiral ketones'c arbon atom and the departingh ydride of NADPH was again attained. [23]Our independentc alculations on carbonyl reduction (pH 7, 100 mm KPi)and reductive amination (pH 9, 2 m NH 3 /NH 4 + )u nder the experimentallya ttainedr eaction conditions show negligible differences of the reaction's Gibbs energy (i.e.,f or benzaldehyde:r eduction DrG' = À45 kJ mol À1 ;r eductivea mination DrG' = À49 kJ mol À1 ;f or acetophenone:r eduction DrG' = À45 kJ mol À1 ;r eductive amination DrG' = À42 kJ mol À1 ;s ee Supporting Information section6 for details).In general, the aforementioned distance between the carbonyl's carbon atom and NAD(P)H's hydride appearst o be ac ritical factor,a lthougho ther parameters can determine carbonyl vs. imine reductase activities, such as as uitable cofactor domain, protond onors adjusted at suitable pK a for efficient imine protonation, flanking residues for pK a adjustment, and a negative electrostatic potentiali nt he substrate-binding site. [24]n the case of the reduction of phenylacetaldehyde (8b), only LE-AmDH-v27 was capable of producing the corresponding alcohol 8a,a lbeit in 4% analytical yield.Av ery similar result was obtained for the reductivea mination of 8b (Table 1), for which none of variants was active.Finally,3 -phenylpropanal (9b)-which possesses one carbon more on its aliphatic chain compared with 8b-was apparently converted into the alcohol 9a in good analytical yields (67-68 %).However,a61 %y ield was also obtainedi nn egative control reactions that were devoid of LE-AmDH variant but included Cb-FDH (NC1).This result indicates that Cb-FDH, whichi su sed for NADH-recycling, is mainly or even solely responsible for this transformation.I nf act, negative control reactions, which were also devoid of Cb-FDH( NC2), gave no detectable conversion.Interestingly,T able 3s hows that Cb-FDH also exhibited al ow level of activity for the reduction of 1b-4b and 6b.H owever, in these cases, the yields obtainedw itht he LE-AmDH variants were highert han one or two orders of magnitude, thus confirming promiscuous activity.I ns ummary,T able 3s hows that LE-AmDH-v27 (possessing the F173S mutation) gave the highest conversion for the reduction of aldehydes 2b (80 %), 4b (40 %), 5b (> 99 %) and 6b (> 99 %) and wast he only enzyme capable of converting 8b.W ild-typeL ysEDH andL E-AmDH-v24 (also possessing the F173S mutation) were, respectively the best performing enzymes forr eduction of 1b (82 %) and 3b (94 %) to the related alcohols.Conversely,t his trend was reversed in the reductive amination reaction, for which LE-AmDH-v1, -v22 and -v25 gave the highest analytical yields of amine products 1c-7c (Table1).Therefore, it appears that the F173A mutation favors AmDH catalytic activity, whereas the F173S mutation favors ADH catalytic activity.Finally,comparing the outcomes of the reactions catalyzed by LE-AmDH-v24a nd -v27, it is also evident that an additional alanine mutation in position2 40 (T240A) further enhances the alcohol dehydrogenasereaction.
Notably,n one of the LE-AmDH variants could reduce any of the ketones 10 b-12 b, 14 b and 16 b-21 b (Group B, Figure 1) to the corresponding secondary alcohols.Only ketones 13 b and 15 b were reduced albeit with 1% conversion (see Supporting Information section 7f or details on chromatographic separations).Therefore, the stereoselective outcomeo ft he reaction could not be determinedd ue to the excessively low conversion.T he lack of ADH activity might be due to the impossibility of binding the ketone substrates in the enzyme's active site with the correct distance and position for NADH hydride delivery.I nc ontrast, such ap ositioning is more probable for aldehydes than ketones because the former possesso nly one carbon chain, thereby increasingt he probability of generating one or more productive binding modes.
All of the alcohol yields reported in Table 3w ere obtained using 20 mm of aldehyde substrate.Therefore, we determined the influence of substrate concentration on the reaction conversions.B enzaldehyde (1b)w as selected as the test substrate at concentrations ranging from 10 to 100 mm,w hereas wildtype LysEDH (45 mm)w as selected as the promiscuousA DH enzyme because it gave the highestc onversion in the reduction of 1b to 1a.T he reactions were run again at 30 8Cf or 24 hi naKPi buffer (pH 7, 100 mm)s upplementedw ith HCOONa (100 mm), NAD + (1 mm)a nd Cb-FDH( 16 mm).The conversion of 1b to 1a was 90 %a t1 00 mm substrate concentration,w hich resulted in a7 6% analytical yield (Figure 3).As mentioned, the apparent loss of mass balance is due to the volatility of 1b.N otably, a1 0-fold increase in 1b concentration did not significantly affect the reactiony ield, which ranged from 95 %( at 10 mm of 1b)t o76% (at 100 mm of 1b).
Michaelis-Menten kinetic parameters were determined in our previous study for the reductive amination of 1b to 1c catalyzed by LE-AmDH-v1 (k app = 43.615k] However,M ichaelis-Menten kinetics for the reduction of 1b to 1a could not be performed in this study because the alcohol 1a possesses high extinction coefficient at the same range of wavelengths that is needed for the quantitative monitoring of NADH depletion.However, para-fluorobenzaldehyde (2b)t urned out to be the suitable substrate for our kinetic study.B ecauseL E-AmDH-v1 was the enzyme that afforded high conversions for both the reduction of 2b to 2a (64 %, Ta ble 3) and the reductive amination of 2b to 2c (90 %, Ta ble1), it was an excellent case for comparing carbonyl reduction with reductivea minationa ctivities.Herein, Michaelis-Menten kinetics conducted at 60 8Cs how that LE-AmDH-v1 acts preferentially as AmDH toward 1b (for reductive amination: k app /K Mapp 6082 m À1 min À1 ;f or reduction to alcohol: k app /K Mapp 272 m À1 min À1 ;s ee Supporting Information section 8 and Ta ble S10f or details).In fact, k app value is greatlyi nf avor of the reductive amination over the reduction to alcohol (22.0 AE 0.9 min À1 vs. 0.15 AE 0.01 min À1 ).In contrast, K Mapp value is ~4-fold better for the reduction to alcohol than for the reductive amination (0.55 AE 0.01 mm vs. 3.62 AE 0.46 mm).These data shows that the relative levels of AmDH vs. ADH activity cannotb ea ssessed based on only specific activity data (as reportedi nS upporting Information, Ta ble S12), but K Mapp has also an important contribution.
One-enzyme, dual-activity( ADH-AmDH)f or alcohol amination As the LE-AmDH variants exhibited both native AmDH activity and promiscuousA DH activity,w ee nvisioned that this dual-activity could be harnessed for the amination of primary alcohols via ao ne-enzyme oxidative-reductive cascade.Such ar edox self-sufficient process is often referred as "hydrogen-borrowing" or,m ore properly," hydride-borrowing" because the hydride abstracted from the first oxidation of the alcohol substrate to the ketone intermediate is delivered backi nt he second reductivea mination step.This process was first reported by our group through the combination of an ADH with an AmDH. [25]Therefore, herein, we studied whether as ingle dehydrogenase enzyme could enable the same two-step redox process.However,t he optimal reaction conditions for the oxidation of an alcohol to ac arbonyl compound are different from the optimal conditions for reverser eduction.I np articular, basic pH values appear to favor the oxidationr eaction, whereas pH close to or at neutrality favors the reduction reaction. [26]herefore, we performed as tudy on the oxidationo fb enzylic alcohol( 1a)t ob enzaldehyde (1b)i nd ifferent buffers, at different pH levels, and using LE-AmDH-v27 due to its high catalytic activity.F igure 4d epictst he progress of the analytical yield over time for the oxidation reaction (see Supporting Information section 5.5.and Ta ble S5 for details) in which aw aterformingn icotinamide adenined inucleotide oxidase (NOx) was used as NAD + -recyclinge nzyme. [27]igure4 shows that higher pH values favor the alcohol oxidation reactions, which is in agreement with the literature.In fact, the highesta nalytical yield of 66 %a fter 24 hr eaction time was obtained in Tris-HCl (pH 9, 100 mm).However,the analyticaly ield was only 7% for the oxidation performed for 24 h, at pH 9a nd in a2m ammoniumf ormate buffer.W ei nfer that the lower aldehydef ormation observed in the ammonium  formate buffer compared with the other buffers might be due to anegative effect on the NOx cofactor-recycling enzyme.
The oxidation of 1a to 1b was tested with all of the variants in the best performing buffer,n amely Tris-HClb uffer (pH 9, 100 mm). Figure 5s hows that all of the LE-AmDH variants were capable of producing 1b in moderate to high yields (44-86 %; see Supporting Information section 5.5 and Table S6 for details).However,t he negative control reactions in whicho nly NOx was present as the enzyme( NC 1) afforded 1b in a3 1% yield;t herefore, the oxidation of 1a to 1b (Figure 5) is partly due to an alcohol oxidase activity of NOx.Nevertheless,t he highest yield for the oxidationr eactionw as achieved using LE-AmDH-v27 as the biocatalyst (86 %), which possesses the beneficial F173S substitution for ADH activity along with the additional T240A and Y238A mutations.However,t he 66 %y ield given by wild-type LysEDH is significantly highert han that obtained from the negative controlr eaction (NC1).Considering both the oxidation (Figure 5) and the reduction (Table 3) experiments, LE-AmDH-v27 was the variant that exhibited the overall highest alcohol dehydrogenase activity.In fact,L E-AmDH-v27 was also the variant possessing the highest specific activity forc arbonyl reduction to alcohol when assayed for 2b as substrate (see Supporting Information section 9a nd Ta ble S12f or details).
In this context, the LE-AmDH variant assumes ab i-functional role, simultaneously acting as ADH and AmDH.T he catalytic amount of NAD + /NADHc ofactorw as internally regenerated due to the inherent redox-neutralityo ft he process.However, the previous resultsd emonstrated that the most active LE-AmDH variants for the oxidation of 1a to 1b have am odest activity for the reductive aminationo f1b to 1c (and vice versa).A dditionally,t he formedb enzaldehyde( 1b)i ntermediate can undergo both reductive amination to 1c as well as reductionb ack to 1a.W ith the aim of finding ac ompromise between the alcohol oxidation and ketone reductive amination steps, we performed an umber of experiments (see Supporting Information section5.6).Under optimized conditions, the hydride-borrowing alcohol aminationo f1a (10 mm)w as conductedw ith all of the LE-AmDH variants (90 mm)i nT ris-HCl (100 mm,p H9)a t1m NH 4 OH and for 48 h. Figure 6( cascade 1) shows that a4 -5 %m aximum yield of 1c was obtained using LE-AmDH-v1,-v22 and-v25, whereas conversion with LE-AmDH-v1 was just above the detection limit.Varying the NAD + concentration (1, 5a nd 10 mm)a sw ell as the substrate  concentration (10, 50 and 100 mm,r espectively)w hile keeping ac onstant NAD + vs. substrate molar ratio (1:10) resulted in increasedp roduction of 1c (0.4, 2.1 and 2.5 mm,respectively).
Finally,t he bi-functional dehydrogenases weret estedi na two-step amination of 1a to 1c in as imilarm anner as described in our previous publication (Scheme2b). [13]However, in this case, although the oxidative and the reductives teps were still performed in one-pot,t hey were separated in time because both steps require NAD as hydride transfer agent (Figure 6, cascade 2).In practice, the reactionw as initiated by adding NAD + (1 mm), NOx (10 mm), LE-AmDH variant (90 mm) and 1a (20 mm)i nT ris-HCl (0.5 mL, pH 9, 100 mm,N H 4 OH 1 m) at 30 8Cf or 24 h.During this time, 1a wasc onverted into 1b by the LE-AmDH variant while NAD + was recycled by NOx at the expense of dioxygen.A fter the oxidative step, other 0.5 mL of the reactionb uffer containing HCOONa( 200 mm)a nd Cb-FDH (16 mm)w ere added sot hat the same LE-AmDH variant could catalyzethe reductiveamination of 1b to 1c;NADH was recycled by Cb-FDHa tt he expense of HCOONa.Therefore, the final concentrations in this second step were LE-AmDH variant (45 mm), Cb-FDH( 8mm)a nd HCOONa (100 mm).Under these reactionc onditions, LE-AmDH-v1 produced the highest analytical yield of 1c (17 %), followed by LE-AmDH-v25, -v22 and -v24 with 13 %, 12 %a nd 5%,r espectively.F urthermore, the analytical yield of 1c could be increased up to 34 %a nd 32 %u sing LE-AmDH-v1 and -v25, respectively,b ya pplying a slight modification of the procedure.In practice, the oxidative step of the cascade wasp erformed for 24 hi nT ris-HCl buffer (0.5 mL, pH 9, 100 mm)a sp reviously (1a 20 mm,L E-AmDH variant 90 mm)b ut in the absence of ammonia.Then, the ammonia solution (0.5 mL, 1 m,p H9)c ontaining HCOONa (200 mm) and Cb-FDH (16 mm)was added to initiate the reductive amination step, which was runf or additional 24 h.Therefore, the final concentrationsi nt his second step were againL E-AmDH variant (45 mm), Cb-FDH( 8mm)a nd HCOONa (100 mm).This increaseo fthey ield for 1c is in agreement with the data reported in Figure 4( purple line) for the oxidation of 1a to 1b, which was indeed impeded by the presence of ammonia in solution.A sp reviously described, we attribute this negative behaviort ot he poor stabilityo fN Ox at high concentration of ammonia/ammonium species.
Figure 6a lso shows that neither wild-type LysEDH nor LE-AmDH-v27 could produce any detectablea mount of amine 1c;h owever,b oth enzymesc ould oxidize 1a to 1b in an efficient manner,a nd LE-AmDH-v27w as in fact the best variant for this transformation (Figure 5).Therefore, wild-type LysEDH and LE-AmDH-v27m ust be incapable of converting 1a into 1c due to al imitation in the reductive amination step.Notably, LysEDH and LE-AmDH-v27w ere the only two enzymes that produced alcohol 1a as the by-product (23 %and 22 %, respectively) along with the amine 1c (17 %a nd 52 %, respectively) in the reductive amination experimentsi na na mmonium formate buffer (Table 1).Therefore, we concludet hat LysEDH and LE-AmDH-v27 cannot convert 1a into 1c in the alcohol amination cascadesb ecause the generated intermediate 1b is preferentially reduced back to 1a rather than aminated to 1c.I nc ontrast, the other LE-AmDH variants( v1, v22, v24 and v25) fully behaved as amine dehydrogenases when the aldehyde 1b was reducedi nt he ammonium buffer,t hereby yielding 1c as the sole product (Table 1, 80-> 99 %).In summary,t his chemoselectivedual-activity of LE-AmDH-v1,-v22, -v24 and -v25 in an ammonium buffer (i.e.,A DH activity for the oxidation of 1a to 1b and AmDH activity for the reduction of 1b to 1c)e nables this unprecedented one-dehydrogenase alcohola mination.
Furthermore, all of the LE-AmDH variants (butn ot the wildtype enzyme) could perform the reductive amination of some ketone substrates and yield the amine product with excellent enantiomeric excess (> 99 %) in the large majority of cases.L E-AmDHsv 1, v22 and v25 were again the best aminating variants, whichc ould convert structurally diverse ketones such as acetophenone, 1-indanone, propiophenone, 1-tetralone, 4chromanone and other aliphatic ketones (11 b-15 b,18 b-21 b)  into enantiopure amines with maximum yields of 76-99 %.However,n one of the enzymes could practically reduce the same ketone substrates to any of the relateds econdary alcohols.This observation suggestst hat in contrast to the possible productiveb inding of ketimines as "AmDH-type intermediates", ketones as "ADH-type substrates" cannot bind in any reactive conformation in the enzyme's active site.Nonetheless, both aldehydes as "ADH-type substrates" and aldimines as "AmDH-type intermediates" were convertede qually well, which is probably because aldehydes possess ah igher conformationalf lexibility than ketones when they are bound in the enzyme's active site.Therefore, at least ap roductiveb inding mode was attained for benzaldehydes reductions to alcohols.
In principle, the promiscuousADH activity of the LE-AmDH variants could be extended to ketones by applying further enzyme engineering.
Finally,A DH activity was also tested with all of the LE-AmDH variants for the oxidation of benzylic alcohol to benzaldehyde in the presenceo ra bsence of ammonia/ammonium species.LE-AmDH-v27 turned out to be the best variant (86 %y ield).Therefore, this unprecedentedd ual ADH/AmDH activity was appliedfor the first example of one-enzyme hydride-borrowing alcohol amination, which yielded5%o fb enzylamine product using LE-AmDH-v1.Notably,L E-AmDH-v1 and -v25 could catalyze benzyl alcohol amination with 34 %a nd 32 %y ields of benzylamine, respectively,b ys eparating the oxidative and the reductives teps in time.To the best of our knowledge,L E-AmDH-v1, -v22 and -v25 represent the first examples of oxidoreductases that have been appliedf or the one-enzyme conversion of an alcoholi ntoa na mine, thereby exhibiting an "alcohol aminase" activity.

Experimental Section
For general information, material, enzymes preparation, details on biocatalytic reactions, analytics and chromatograms, see the Supporting Information.

General procedure for the reductive amination of aldehydes
The biocatalytic reaction was carried out in an ammonium formate buffer (1 mL, 2 m,p H8.5) by adding NAD + (1 mm), Cb-FDH (16 mm), LE-AmDH (45 mm)a nd aldehyde (20 mm)i nc onsecutive order.T he reaction was incubated at 30 8Ci na no rbital shaker (170 rpm) for 24 h.Next, the reaction was acidified with formic acid (20 mL, until pH < 4) and the organic compounds were extracted with EtOAc (2 500 mLE tOAc, containing internal standard).The aqueous layer was basified with KOH (300 mL, until pH > 12) and the organic compounds were extracted (2 500 mLE tOAc, containing internal standard).The acidic and basic extracts were dried with MgSO 4 and analyzed separately with GC-FID.For details, see Supporting Information, section 5.

General procedure for the reduction of ketones and aldehydes to alcohols
The biocatalytic reaction was carried out in ap otassium phosphate buffer (1 mL, 100 mm,p H7)s upplemented with sodium formate (100 mm)b ya dding NAD + (1 mm), Cb-FDH (16 mm), LE-AmDH (45 mm)a nd aldehyde or ketone (20 mm)i nc onsecutive order.T he reaction was incubated at 30 8Ci na no rbital shaker (170 rpm) for 24 h.Then the reaction was extracted with EtOAc (2 500 mL, containing internal standard).The combined organic phase was dried with MgSO 4 and analyzed by GC-FID.For details, see Supporting Information, section 5.

One-enzyme conversion of benzylic alcohol (1 a) to benzylamine (1 c)
The biocatalytic reaction was carried out in aT ris-HCl buffer (1 mL, 100 mm)s upplemented with NH 4 OH (1 m)a tafinal pH value of 9 and by adding NAD + (1 mm), LE-AmDH (90 mm)a nd 1a (10 mm).The reaction was incubated at 30 8Ci na no rbital shaker (170 rpm) for 48 h.Then the reaction was basified with KOH (200 mL, 10 m) and extracted with EtOAc (2 500 mL, containing internal standard).The combined organic phase was dried with MgSO 4 and analyzed by GC-FID.For details, see Supporting Information, section 5.

Figure 1 .
Figure 1.List of compounds used in this study.Groups Aand Bd epict the aldehydes and ketones,respectively, that were testedf or both reductive amination and reduction to alcohols.