Extended Scope and Understanding of Zinc‐Dependent Alcohol Dehydrogenases for Reduction of Cyclic α‐Diketones

Alcohol dehydrogenases (ADH) are important tools for generating chiral α‐hydroxyketones. Previously, only the ADH of Thauera aromatica was known to convert cyclic α‐diketones with appropriate preference. Here, we extend the spectrum of suitable enzymes by three alcohol dehydrogenases from Citrifermentans bemidjiense (CibADH), Deferrisoma camini (DecADH), and Thauera phenylacetica (ThpADH). Of these, DecADH is characterized by very high thermostability; CibADH and ThpADH convert α‐halogenated cyclohexanones with increased activity. Otherwise, however, the substrate spectrum of all four ADHs is highly conserved. Structural considerations led to the conclusion that conversion of diketones requires not only the expansion of the active site into a large binding pocket, but also the circumferential modification of almost all amino acid residues that form the first shell of the binding pocket. The constellation appears to be overall highly specific for the relative positioning of the carbonyl functions and the size of the C‐ring.


Introduction
In chemical synthesis, alcohol dehydrogenases (ADH) are firmly established as catalytic tools. Due to high enantiospecificities and -selectivities, they are mainly used to generate chiral hydroxyl compounds. [1a-c] In this context, some ADHs also exhibit the ability to selectively reduce prochiral diketones to the corresponding hydroxyketones with only one defined chiral center. [2a-d] These compounds are intermediates of numerous active pharmaceutical ingredients [3a-j] and therefore valuable synthetic targets.
Only a few ADHs are capable to generate sterically demanding cyclic hydroxyketones with αor βsubstitutions. [2c,4a-f] However, through a rational screening of metabolic pathways we succeeded in identifying a NADHdependent ADH of the medium chain dehydrogenase/reductase (MDR) superfamily that reduces 1,2-cyclohexanedione to (S)-α-hydroxycyclohexanone with outstanding activity. [5a,b] In this, the enzyme from Thauera aromatica (ThaADH) differs significantly from other known representatives of the MDR-ADH superfamily. The special substrate acceptance goes along with a very specific design of the enzyme's active site. [5b] Most apparent is the formation of a single large binding pocket instead of the usual division into a large and a small binding pocket, [6a,b] due to the replacement of a bulky aromatic amino acid residue. [5a,b] Unfortunately, the detailed study of the substrate acceptance of ThaADH revealed a spectrum very narrowly focused on cyclic α-substituted ketones with a ring size of six C-atoms. The second carbonyl function could only be substituted by halogens (Br, Cl, F); methyl substitutions were not accepted. Cyclic diketones in which the carbonyl functions were not in the α-position and/or which had slightly smaller or larger rings were hardly converted. [5b] With broader synthetic applicability in mind, we here aimed to expand the spectrum of NADH-dependent ADHs with the ability to reduce cyclic diketones. Based on our knowledge of the particular active site structure of ThaADH and our positive experience with the use of the Conserved Domains Database (CDD) as a means for identifying functionally similar ADHs, [7] we chose a sequence-based approach for this purpose. Promising candidates were produced in Escherichia coli and investigated for substrate scope and biochemical features. We applied structural models to explain our observations and better understand acceptance of cyclic diketones by NADH-dependent ADHs of the MDR superfamily.

ADH candidates
From the database search for amino acid sequences with at least 30 % identity to ThaADH, and from a following analysis of the obtained sequences for typical characteristics of NADH-  (Table S1, Supporting information), we identified four promising candidates (Table 1). The highest sequence identity had a candidate from Thauera phenylacetica (ThpADH), followed by a candidate from Citrifermentans bemidjiense (CibADH) and a candidate from Deferrisoma camini (DecADH). The candidate with the lowest sequence identity was from Rhodococcus jostii (RhjADH). Sequence characteristics, predicted monomer sizes (Table S1, Supporting information) and the ability for carbonyl reduction with NADH (see below) placed all four candidates in the NADHdependent MDR-ADH superfamily. [8] Among themselves, the new candidates had sequence identities ranging from 28 % to 73 % (Table S2, Supporting information). In each case, RhjADH had the lowest identity of about 30 %. The three candidates with the higher sequence identities (CibADH, DecADH and ThpADH) were annotated in NCBI as 6-hydroxycyclohex-1-ene-1-carbonyl-CoA dehydrogenases (reference sequences WP_ 012529862.1, WP_025323423.1 and WP_004358383.1, respectively) indicating the same natural task as ThaADH [5a,9] and thus, a close functional relationship. RhjADH was annotated more generally as a zinc-binding dehydrogenase (NCBI reference sequence WP_073362856.1). Notably, all four candidates as well as the ThaADH have an identity of about 30 % with the carbonyl reductase 2 of Candida parapsilosis (CPCR2 ; Table S2, Supporting information), which is a typical NADH-dependent MDR-ADH without the ability to convert cyclic α-diketones. [5a] A first investigation of the catalytic activity of crude preparations of recombinant CibADH, DecADH and ThpADH, respectively, demonstrated NADH-dependent reduction of 1,2cyclohexanedione as a model cyclic α-diketone (Table 2). In contrast, RhjADH did not accept this substrate, but reduced aliphatic 2,3-pentanedione, which is a typical substrate of CPCR2, with good activity (1.1 U/mL). As our former studies indicated that good acceptance of cyclic and aliphatic αsubstituted ketones is almost mutually exclusive, [5a,b] and the observed activity was in accordance with the lower sequential similarity of RhjADH with ThaADH, we assumed that RhjADH was not able to convert cyclic diketones effectively. Consequently, we excluded this enzyme from further study.

Substrate scope
Detailed examination of the reducing activities of purified CibADH, DecADH and ThpADH, respectively, revealed an overall very similar substrate scope of all three enzymes (Figure 1), which also strongly correlated with the substrate acceptance of ThaADH. [5b] Methylated cyclohexanones (2-, 3-and 4-methylcyclohexanone), cyclic diketones whose carbonyl groups were not in the α-position (1,3-and 1,4-cyclohexanedione) or compounds with aromatic substituents, such as acetophenone and benzaldehyde, were not converted. Cyclic ketones and diketones with a ring structure of five or seven carbon atoms were hardly accepted, whereas there was moderate conversion of the cyclic ketone cyclohexanone. Among the linear aliphatic carbonyl compounds, only diketones underwent a small conversion.
Under conditions that suited the requirements of each enzyme best, 1,2-cyclohexanedione was reduced to (S)-2hydroxy-cyclohexanone with an activity of 3.1 U mg À 1 (CibADH), 4 U mg À 1 (DecADH) and 2.5 U mg À 1 (ThpADH). Thus, the stereoselectivity was the same ( Figure S1, Supporting information), but the activity of all three enzymes was overall significantly smaller compared to ThaADH (6.3 U mg À 1 ). [5b] However, the activities of the three enzymes relative to their conversion of 1,2-cyclohexanedione were mostly comparable. The only significant difference occurred with α-halogenated cyclohexanones. CibADH and ThpADH converted 2-chloro-and 2-bromocyclohexanone, respectively, with a distinctly higher activity than 1,2-cyclohexanedione, whereas DecADH showed a lower activity with these substrates. The latter was in agreement with the behavior of ThaADH, which always worked best on 1,2cyclohexanedione. [5b] pH effects CibADH, DecADH and ThpADH showed pH-dependent activity profiles ( Figure S2, Supporting information) similar to ThaADH, [5b] and overall typical for MDR-ADHs. [10a,b,c] Best activities towards reduction were found at pH 6 to pH 7, mostly with potassium phosphate (KPi) buffer. Hence, KPi buffer with pH 6.5 provided the standard system for all experiments.

Temperature effects
Regarding activity dependence on temperature (Figure 2), CibADH and ThpADH behaved were similar to ThaADH [5b] with a maximum activity at 50°C. In contrast, DecADH was increasingly active up to 80°C. The activity at even higher temperatures could not be determined due to limitations of our experimental setup. On the other hand, DecADH had no measurable activity at room temperature, while the activities of CibADH and ThpADH were around 20 %.
In accordance with these observations, DecADH demonstrated a considerably higher stability at elevated temperatures than CibADH and ThpADH, as well as ThaADH (Table 3). The The findings place DecADH among the thermostable MDR-ADHs. In fact, its thermostability is one of the highest ever described for an enzyme in the MDR-ADH family. ADHs from Sulfolobus solfataricus, [10c] Aeropyrum pernix, [11] Flavobacterium frigidmaris KUC-1, [12] Picrophilus torridus, [13] Sulfolobus tokodai [14] and Chloroflexus aurantiacus [7] have activity maxima between 70°C and 95°C. The half-life times at 70°C of the ADHs from F. frigidimaris and S. solfataricus are six minutes and five hours, respectively. The stability of DecADH correlates with the fact that the source organism was isolated from a deep-sea hydrothermal vent, although D. camini itself is classified as only moderately thermophilic. [15] Figure 1. Substrate specificity of CibADH, DecADH and ThpADH for reductions. Activity with 1,2-cyclohexanedione was set as 100 % for each individual enzyme and is equivalent to 3.1 U mg À 1 (CibADH), 4.0 U mg À 1 (DecADH) and 2.5 U mg À 1 (ThpADH), respectively. Error bars indicate the standard deviation of triplicate measurements. A R : relative activity, Me: methyl. Missing structures are given in Table S3 (Supporting information).

Solvent effects
All three enzymes showed partially higher catalytic activities in the presence of selected water-miscible solvents than in pure water ( Figure S3, Supporting information). However, there were significant differences with regard to the specific solvent and its proportion in the reaction medium. At high solvent contents (10 % v/v and 25 % v/v), we observed high residual activities for all enzymes only with glycerol and ethylene glycol, while methanol, ethanol, and isopropanol generally caused the greatest loss of activity. Most sensitive reacted CibADH. In contrast, DecADH showed significantly better residual activities than the other enzymes in the presence of both high concentrations of DMSO and ethylene glycol (25 % v/v each). Also in direct comparison with ThaADH, DecADH showed significantly higher residual activities after incubation in an aqueous solution with 20 % (v/v) methanol, DMSO or ethylene glycol ( Figure 3). However, no differences were observed with respect to all other solvents. Given the previously observed significantly better thermostability of DecADH, this was interesting, as stability against elevated temperatures and against solvents are often related to the same structural features in the literature and are affected by the same mutations. [16a-e] Structural features Superimposition ( Figure S4, Supporting information) of monomeric structural models of CibADH, DecADH and ThpADH, respectively, derived from AlphaFold ( Figure S5, Supporting information) on the crystal structure of ThaADH, [5b] revealed a high degree of conservation in the active site of the four enzymes ( Figure 4). The 15 amino acids forming the first shell of the substratebinding pocket were almost identical (Table 4 and Figure S6, Supporting information); only DecADH deviated in two amino acid residues (Y285 and T307, entries 10 and 11, respectively, in Table 4), but the replacements were conservative (Y for F) or semi-conservative (T for N), respectively.   In contrast, compared with the active site of CPCR2 [17] ( Figure S6, Supporting information), the four enzymes showed only four identical first-shell residues, all of them being involved in proton transfer or zinc binding, i. e. a general feature in the whole superfamily.
The observation strongly implies that not a few selected amino acids in the active site of a typical MDR-ADH determine the ability to accept cyclic α-substituted ketones as substrates, but the almost complete first-shell assortment. This agrees with results from a former study, in which we observed that replacement of single amino acids in the first shell of the active site of CPCR2 hardly influenced acceptance of cyclic ketones. [18] Only one of five tested residues (L119) had an influence at all. Replacement by methionine improved conversion of cyclohexanone and its methyl-substituted derivatives, including 2methylcyclohexanone, but did not extent the substrate scope. Most probably, the effect was due to a slight enlargement of the active site. Slight differences in the size of the active sites of CibADH, DecADH and ThpADH compared with ThaADH were observed in our study here as well. The residues correlating with W302 in ThaADH (e. g. W308 in DecADH) are slightly shifted in all three novel enzymes, resulting in a tiny size reduction. Since already in ThaADH the distance between W302 and the C4 atom of a 1,2-cyclohexanedione molecule docked in the active site is only 3.3 Å, [5b] this further reduction might account for the overall smaller catalytic activity of the novel ADHs compared to ThaADH. In contrast, we could not find structural explanations for the high activity of CibADH and ThpADH on α-halogenated cyclohexanones compared with DecADH and ThaADH.
The two amino acid exchanges detected in the active site of DecADH certainly do not explain the much higher thermostability of this enzyme compared with CibADH, ThpADH and ThaADH. General sequence identity or even an overall structural similarity is also highly unlikely to help much. [16d] According to current knowledge, thermostability is favored by a higher degree of oligomerization, stronger intra-and intermolecular interactions and shorter loops at the molecular surface. [16d] In fact, more compressed structures combined with shorter surface loops have already been reported for thermostable MDR-ADHs. [19a,b] However, superposition of the monomeric structural model of DecADH on the structure of ThaADH ( Figure 5) shows only a limited increase in compactness. Two of the surface loops (83-90 and 145-150) are shortened by three and five amino acids, respectively, compared to the complementary loops of ThaADH (74-84 and 137-147, respectively). At the same time, a loop (14-27) lengthened by eight amino acids occurs. Extended loops at a similar position are also present in other thermostable ADHs ( Figure S7, Supporting information), e. g. from S. solfataricus (PDB 1R37, loop 47-61) [20] and Rhodococcus ruber (PDB 3JV7, loop 47-55). [21] Analysis of the interaction between the subunits of dimeric structural models of DecADH, CibADH, and ThpADH derived from AlphaFold ( Figure S8, Supporting information) and the dimeric structure of ThaADH [5b] using PISA [22] revealed no significant differences in the extent of the interface or the solvation free energy gain which corresponds with hydrophobic interfaces (Table 5). Therefore, the increased thermostability of DecADH does not seem to be due to these properties. In contrast, DecADH basically exhibited significantly more salt bridges than the other enzymes and compared to CibADH and ThaADH also more hydrogen bonds. Consequently, the thermostability of DecADH seems to be indeed due to the stronger interactions between the subunits of the dimeric structure. This  ΔiG: solvation free energy gain upon formation of the interface. The value is calculated as the difference between total solvation energies of isolated and interfacing structures. A negative ΔiG corresponds to hydrophobic interfaces, or positive protein affinity. The value does not include the effect of hydrogen bonds and salt bridges across the interface.
is consistent with our earlier conclusion that already the good thermal stability of ThaADH is due to a higher number of salt bridges at the dimer interface, [5b] and with our successful stabilization of CPCR2 by improving the binding between its subunits. [23] Similar findings have been reported for other thermostable MDR-ADHs. [19a,b,21] Going further, the only slightly better stability of DecADH against solvents compared to ThaADH, despite its significantly higher thermostability, might be due to a lower hydrophobicity of the surface of DecADH, as revealed by a comparison using the Hydrophobic Intensity Patch (hi-patch) tool [24] (Table 6). This agrees with the assumption that both high thermostability and solvent stability of ADH-A from Rhodococcus ruber is attributed to the strong interaction of the subunits in this enzyme combined with the pronounced hydrophobicity of its surface. [21] Conclusions Our sequence-based screening targeting conserved domains within the enzyme structure was again successful in this study. Three further alcohol dehydrogenases of the MDR superfamily, which show a special preference for cyclic diketones were identified and characterized. With DecADH, an enzyme with very high thermostability and thus great potential for synthetictechnical application is among them. CibADH and ThpADH show particularly promising activities for the reduction of αhalogenated cyclohexanones.
The newly identified enzymes also provide interesting insights into the particular structural adaptations associated with the conversion of cyclic, multifunctionalized ketones by MDR-ADH. Conversion of diketones appears to require not only the expansion of the active site into a large binding pocket, but also the circumferential modification of almost all amino acid residues that form the first shell of the binding pocket. The constellation appears to be highly specific for the relative positioning of the carbonyl functions and the size of the C-ring. Thus, the conversion of other and/or differently positioned functional groups and differently sized ring structures most likely requires a completely different design of the substrate binding site. Thus, expansion of the substrate spectrum by replacement of single residues in the active site by directed mutation has little chance of success. For identification of new constellations, classical screening or metabolic pathway screening, which had been the key for the idenfication of ThaADH, are strongly recommended.
Screening. Database screening was performed on the amino acid sequence of ThaADH (NCBI reference sequence O87871.2) using the BLASTp [25] and the Conserved Domain Search [26] of NCBI. Hits were further analysed for medium-chain length, NADH-binding sites and zinc-dependency, the latter using the Pfam-database. [27] Cloning of genes. The cibADH, decADH, rhjADH and thpADH gene was obtained by PCR using genomic DNA from the organisms mentioned above, as a template and attaching overhangs complementary to the vector sequence, respectively. The PCRs for gene extraction as well as for vector amplification was performed with Phusion® high-fidelity DNA polymerase (New England Biolabs Inc., MA, USA) according to the manufacturer's instructions. All primer sequences are listed in the supporting information (Table S4, Supporting information). For the isolation of the decADH gene an initial nested PCR was necessary. ADH genes were introduced in the same expression vector as the thaADH gene [5a] by Gibson assembly [28] according to the standard protocol. Thereby the genes were fused to a sequence coding for a C-terminal Strep-tag II. The 50°C incubation of the assembly was extended to 1 h followed by 15 min at 72°C. Due to the lack of expression, it was necessary to reclone the decADH and rhjADH gene into into a pET28b vector as described above. By this a sequence coding for a N-terminal His-tag was fused to the ADH genes. After vector construction, E. coli TOP10 cells were transformed with the gene containing plasmids. Positive clones were screened by colony PCR and their isolated vector was sequenced by GATC Biotech (Konstanz, Germany).

Expression of genes.
The thaADH gene was already constructed within a modified pET22b expression vector present in E. coli BL21 (DE3). [5a] E. coli BL21 (DE3) cells were transformed with a pET22b or pET28b vector containing the four new ADH genes, respectively. The expression was performed in 400 mL Luria-Bertani medium containing either 35 mg·mL À 1 kanamycin (pET28b vectors) or 200 mg·mL À 1 ampicillin (pET22b vectors) as an antibiotic marker. Medium was inoculated with cells of an overnight culture to a final OD 600 of 0.1 and incubated at 37°C with shaking at 180-220 r.p.m. until the induction of expression. Gene expression was induced with different concentrations of isopropylβ-D-1-thiogalactopyranoside (IPTG), between and OD 600 of 0.7 and 1.0. In addition, 0.43 mmol L À 1 ZnSO 4 was added to the culture and the cultivation temperature was reduced with the same shaking speed as before. The best conditions of ADH synthesis are given in the supporting information (Table S5, Supporting information). The cells were harvested by centrifugation (4°C, 5000 rpm, 30 min) and stored at À 20°C.
Cell disruption. Cell pellet was resuspended in lysis buffer with pH 8.0 containing 100 mmol L À 1 triethanolamine (TEA), 150 mmol L À 1 NaCl, 1 mmol L À 1 phenylmethylsulfonyl fluoride (PMSF) and 50 mg L À 1 DNaseI. A French press was used for cell lysis, utilizing two passages at a pressure of 1200 psi. Cell debris was removed by centrifugation (14000 rpm, 1 h, 4°C). Table 6. Surface analysis of DecADH and ThaADH with the hi-patch tool. [24] ADH Purification of Strep-tagged CibADH, ThpADH and ThaADH. Cellfree crude extract containing the respective Strep-tagged ADH was applied to gravity flow affinity chromatography with a 5 mL Strep-Tactin Superflow high capacity column (IBA GmbH, Goettingen, Germany) according to the manufacturer's instructions. The washing buffer was composed of 100 mmol L À 1 TEA, 150 mmol L À 1 NaCl at pH 7.5, the elution buffer includes additionally 2.5 mmol L À 1 desthiobiotin at the same pH. Enzyme concentrations were determined using a bicinchoninic acid assay kit (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions and with a bovine serum albumin standard. Elution fractions were tested for activity, pooled and analyzed by SDS-PAGE ( Figure S9, Supporting information). Enzyme storage was either at 6°C (ThaADH) or at À 20°C using 40 % glycerol as antifreeze agent (CibADH and ThpADH).
Purification of His-tagged DecADH. The crude extract containing the His-tag-DecADH fusion protein was applied to a immobilized metal affinity chromatography (IMAC) using a 5 mL HisTrap FF Crude column (GE HealthcareLife Sciences, Chicago, IL, USA) in combination with an AKTA FPLC system (GE Healthcare Life Sciences, Chicago, IL, USA). The column was equilibrated with washing buffer (100 mmol L À 1 TEA, 500 mmol L À 1 NaCl at pH 7.5) and crude extract was loaded onto the column. The His-tagged enzyme molecules bound to the column were eluted by increasing the imidazole concentration with a linear gradient up to 300 mmol L À 1 imidazole using increasing portions of elution buffer (100 mmol L À 1 TEA, 500 mmol L À 1 NaCl, 500 mmol L À 1 imidazole at pH 7.5). Fractions containing the biocatalyst were pooled and dialysed against dialysis buffer (100 mmol L À 1 TEA, 150 mmol L À 1 NaCl at pH 7.5). Elution fractions were tested for activity, pooled and analyzed by SDS-PAGE ( Figure S9, Supporting information). The determination of the protein concentration was performed as described for Strep-tagged enzymes. His-tag-DecADH was stored at 6°C.
Determination of ADH activity. Photometric assays were performed for the determination of ADH reductive activity by measuring the consumption of the cofactor NADH using a Cary 60 UV-vis spectrophotometer (Agilent, Waldbronn, Germany). The absorption at 340 nm was therefore detected for 1 min at 40°C in triplicate for calculating the initial reaction rates. The standard reduction assay was performed in 100 mmol L À 1 potassium phosphate buffer at pH 6.5, 25-50 mmol L À 1 substrate (Table S3, Supporting information), 0.25 mmol L À 1 NADH and an appropriate concentration of ADH in a 1 mL scale. For the initial substrate screening with crude extract containing ADH, the decreasing absorption was measured once with 10 mmol L À 1 substrate. Substrate specificity was achieved at 50°C for DecADH and at 40°C for CibADH and ThpADH with 10 mmol L À 1 substrate. For the determination of the thermal and pH-dependent activity profiles of the ADHs the assay was carried out either at different temperatures with 100 mmol L À 1 potassium phosphate buffer at pH 6.5 or at 40°C with variable buffers and pH values. For the investigation of the effect of organic solvents on the ADHs activity, solvents were added to the assay in different concentrations, respectively.

Determination of ADH stability.
To study the thermal stability, the ADHs were incubated in 100 mmol L À 1 TEA, 150 mmol L À 1 NaCl, pH 7.5 at different temperatures for various lengths. Thereafter the photometric activity assay was performed as described above for determining the ADH activity. Half-life times were calculated using the following equation, whereas the slope of the natural logarithm of the enzyme activity as a function of the incubation time represents the inactivation constant.
t1=2 ¼ ln2 k t 1/2 : half-life k: inactivation constant The effect of water-miscible solvents on enzyme stability was verified by the incubation of the biocatalyst in 20 % (v/v) polar organic solvent for periods of time followed by the measurement of the reductive activity as mentioned above.
Determination of stereoselectivity. Product analytics by GC were performed as in our former work.
[5b] Retention times of substrate 1,2-cyclohexanedione and the (R)-and (S)-enantiomer of the corresponding product 2-hydroxycyclohexanone were 14.9 min, 15.3 min and 15.4 min. Generation of homology models. While a crystal structure of ThaADH is available (PDB 7QUY, [5b] monomeric and dimeric structure models, respectively, were created for the other ADHs using AlphaFold. [30a,b]

Supporting Information
Additional references cited within the Supporting Information. [31,32a,b]