Escherichia coli BL21(DE3) was acquired from New England BioLabs (Ipswich, MA, USA) and was cultivated in Luria–Bertani (LB) broth or on LB agar plates at 37°C. Recombinant strains E. coli BL21(DE3) (pET-28b(+)-Ss-IRED_R) and E. coli BL21(DE3) (pET-28b(+)-Ss-IRED_S) were grown in LB broth or LB agar plates, supplemented with the appropriate antibiotic (kanamycin 50 μg/ml). Autoinduction media LB Base including trace elements was used for Ss-IRED_R and Ss-IRED_S protein production. Recombinant strains of E. coli BL21(DE3) coexpressing an IRED and Lb-ADH required a second antibiotic (ampicillin 50 μg/ml) as well anhydrotetracycline (200 μg/ml) for protein production.

Ss-IRED_R (National Center for Biotechnology Information (NCBI) ref. seq. WP_112474446.1) was identified from Streptomyces sp. strain ST1020 (BioSample: SAMN04977949). Ss-IRED_S (NCBI ref. seq. WP_119590418.1) was identified from Streptomyces scabies strain ST129 (BioSample: SAMN07731257).

Biotransformation substrate 1a was kindly provided by Prof. Ignacio Carrera; substrates 1b–1d were generously donated by Prof. Nicholas Turner; substrates 3–5 and a were purchased from Sigma Aldrich (St Louis, MO, USA). All the biotransformation substrates were used without further purification. Racemic amines used as analytical standards were purchased from Sigma Aldrich (St. Louis, MO, USA), while chiral amines were produced from the prochiral imines or ketones through biotransformations with reported biocatalysts: 2a was produced with IRED-K; 2b was produced with IRED-D and K (Velikogne et al., 2018); 2c was produced with Ao-IRED (Aleku et al., 2016) and GF3546-IRED (Mitsukura et al., 2013); and 4a was produced with AspRedAm (Aleku et al., 2017).

Molecular biology reagents were purchased from New England Biolabs (Ipswich, MA, USA) and Thermo Fisher Scientific (Waltham, MA, USA). Expression vector pET-28b(+) was purchased from Novagen (Darmstadt, Germany) and was used for gene expression. Fourteen pET-28a(+) IRED-containing plasmids (IRED-A to IRED-N) and plasmid pASK-IBA5plus–Lb-ADH were very generously donated by Dr. Jörg H. Schrittwieser (Velikogne et al., 2018).

Standard procedures were used for DNA manipulation, and genomic DNA from Streptomyces was extracted according to Sambrook Molecular Cloning: A Laboratory Manual (Sambrook 2001). Plasmid DNA was purified using commercial chromatography kits from Thermo Fisher Scientific (Waltham, MA, USA) and QIAGEN (Hilden, Germany); silica columns from these kits were regenerated and reused following the procedure described by Siddappa (Siddappa et al., 2007), while standard protocols were modified to obtain greater yields of DNA (Pronobis, Deuitch, and Peifer 2016; Wood et al., 2017). Restriction enzymes and thermostable polymerases were used according to the manufacturers’ instructions. PCR amplifications were performed in a GeneAMP PCR system 2400 (PerkinElmer, Waltham, MA, USA) using adequate cycling periods, and DNA samples were routinely analyzed by agarose gel electrophoresis (Sambrook 2001). Nucleic acid concentration and purity were measured using a NanoDrop^® ND-1000 spectrophotometer. Transformations of electrocompetent cells were performed on a BioRad MicroPulser™ Electroporator following manufacturer protocols.

An optimized sequence for E. coli of Ss-IRED_R (provisional ID Bankit 2503810) was synthetized and cloned in pET-28b(+) by GenScript (Piscataway, NJ, USA). The resulting vector was transformed into E. coli BL21(DE3) for protein expression, yielding the recombinant strain E. coli BL21(DE3) (pET-28b(+)–Ss-IRED_R).

Amplification of Ss-IRED_S was performed by PCR with primers IRED NdeI Fw (GCT​CAT​ATG​CCC​GCA​CAC​ATC​GCA​G) and IRED EcoRI Rv (CTA​GAA​TTC​GAC​GAG​GGA​AAC​CCC​GAC). Ss-IRED_S gene was then cloned into pGEM-T Easy vector (Promega, WI, USA), and digestion with NdeI and EcoRI allowed subcloning into pET-28b(+) in frame with the N-terminal His6 tag. The obtained vector was purified by gel electrophoresis using the Freeze ‘N Squeeze kit from Bio-Rad (Hercules, CA, USA) and transformed into E. coli BL21(DE3), yielding the recombinant strain E. coli BL21(DE3) (pET-28b(+)–Ss-IRED_S).

Fresh plates of E. coli BL21(DE3) (pET-28b(+)–Ss-IRED_R) and E. coli BL21(DE3) (pET-28b(+)–Ss-IRED_S) were streaked from frozen stocks, and a single colony was used to inoculate 5 ml of LB-Kan. The culture was incubated in a rotary shaker overnight (150 rpm, 37°C), and 1 ml of this culture was used to inoculate 100 ml of fresh autoinduction media LB Base including trace elements and kanamycin. This culture was incubated in a rotary shaker at 150 rpm for about 2.5 h at 37°C until an OD_{600 nm} ≈ 1; and then it was grown at 150 rpm and 28°C for 48 h. The cells were collected by centrifugation at 4,000 × g and 4°C for 15 min. The pellet was washed three times with 50 ml of Tris buffer (100 mM, pH 7.5). E. coli cells measuring 50 mg (wet weight) were resuspended in 465 μl of the same buffer; glucose and the appropriate substrate were added to a final concentration of 2% and 5 mM, respectively. For RedAm reactions, 50 eq. of propargylamine was used. The reaction mixture was incubated at 28°C and 150 rpm for 48 h. Reactions were quenched by addition of 10 M of NaOH (100 μl). The mixture was extracted with ethyl acetate (750 μl), and the organic layer was separated by centrifugation (5 min, 12,000 × g). The organic phase was dried with Na₂SO₄, and methyl heptadecanoate (2.5 mM) was added as internal standard. Percent conversion and enantiomeric excess were determined as stablished in the Gas Chromatography Analysis section.

Fresh plates of E. coli BL21(DE3) coexpressing an IRED and Lb-ADH were streaked from frozen stocks, and a single colony was used to inoculate 5 ml of LB-Kan + Amp. The culture was incubated in a rotary shaker overnight (150 rpm, 37°C), and 1 ml of this culture was used to inoculate 100 ml of fresh autoinduction media LB Base including trace elements and the appropriate antibiotics. This culture was incubated in a rotary shaker at 150 rpm for about 2.5 h at 37°C until an OD_{600 nm} ≈ 1; ADH production was induced by the addition of anhydrotetracycline. Then, it was grown at 150 rpm and 28°C for 48 h. The cells were collected by centrifugation at 4,000 × g and 4°C for 15 min. The pellet was washed three times with 50 ml of Tris buffer (100 mM, pH 7.5). E. coli cells measuring 50 mg (wet weight) were resuspended in 465 μl of the same buffer; glucose and the appropriate substrate were added to a final concentration of 2% and 5 mM, respectively. The reaction mixture was incubated at 28°C and 150 rpm for 48 h. Reactions were quenched by addition of 10 M of NaOH (100 μl). The mixture was extracted with ethyl acetate (750 μl), and the organic layer was separated by centrifugation (5 min, 12,000 × g). The organic phase was dried with Na₂SO₄, and methyl heptadecanoate (2.5 mM) was added as internal standard. Percent conversion and enantiomeric excess were determined as stablished in the Gas Chromatography Analysis section.

Biotransformation products were analyzed on a GC-2010 Shimadzu (Kyoto, Japan) gas chromatograph equipped with a FID detector. Carrier gas, H₂; makeup gas, N₂. Columns MEGA-SE52 (achiral SP, 30 m × 0.25 mm × 0.25 μm) and MEGA-DEX ASX-1 (chiral SP, 25 m × 0.25 mm × 0.25 μm) from MEGA (Milan, Italy) were used for determination of conversion and enantiomeric excess, respectively. Particular analytical conditions and retention times for each amine can be found in Supporting Information, Supplementary Tables S1, S2.

YASARA’s homology modeling experiments, version 21.8.27, were carried out with five templates and three PSI-BLAST iterations in template search; the remaining parameters were set by default (Krieger and Vriend 2014). The templates were initially prepared by adding bond orders and charges; water molecules were removed, and hydrogens were added accounting for protonation states of all ionizable residues at pH 7.5. IREDs from A. oryzae (PDB 5G6R), A. terreus (PDB 5OJL), Bacillus cereus (PDB 4D3F), and Amycolatopsis orientalis (PDB 5A9T); and RedAm from A. terreus (PDB 6EOD) were used as structural templates for modeling Ss-IRED_R, Ss-IRED_S, IRED-A, B, C, D, F, I, J, K, L, M, and N. To aid alignment correction and loop modeling, a secondary structure prediction for the target sequence was obtained. This was achieved by running PSI-BLAST to create a target sequence profile and feeding it to the PSI-Pred secondary structure prediction algorithm (Jones 1999). A total of 25 independent models were created, and YASARA combined the best parts of the 25 models to obtain a hybrid model. In each homology model using YASARA software, we assigned bond orders and protonation patterns according to pH 7.4 (Krieger et al., 2012).

Refinement by short molecular dynamics simulations was performed to improve structures’ quality: with the use of YASARA’s force field YAMBER3 (Krieger et al., 2004), 500 ps of molecular dynamics simulation was conducted at 298 K and pH 7.4. The simulation was performed in explicit solvent with TIP3P water model, density 0.997. Based on the lowest energy, the final representative structures were considered for further study. Ligands were first prepared using YASARA’s molecular modeling software (Krieger, Koraimann, and Vriend 2002). Solvation of the ligands prior to minimization was carried out in an explicit solvent shell at pH 7.4 and 0.9% NaCl. Each ligand was subjected to energy minimization to refine the bond lengths and bond angles. The minimized conformation with the lowest energy for each ligand was used in the docking experiments. Docking was performed by VINA software (Trott and Olson 2019) using default parameters. However, the number of VINA runs was increased from 10 to 250. Size of the docking box was set to 5 Å around the amino acids in the active site in each structure. Positions with a root mean square deviation (RMSD) lower than 2.5 Å were clustered together, and each cluster was represented by the position with lowest free energy of binding. Every cluster obtained by VINA for each substrate used in imine reduction was analyzed, and parameters such as energy of binding, distance for hydride transfer from NADPH to C═N double bond, and interactions between nitrogen of imine with surrounding possible proton donor amino acids His, Glu, Asp, Ser, Tyr, and Cys were evaluated.

We used BLASTp of NCBI (Altschul et al., 1997) to compare the amino acidic sequences of different R- and S-IREDs previously reported (Mitsukura et al., 2010, 2013; France et al., 2018). Through the alignment of twelve sequences, conserved regions formed by 5 and 13 amino acids were identified, and from these, two were selected as templates for new searches: M1: VWNRT and M2: YxDGAI[ML]AxPx2IG. Sequenced genomes from our collection of native Streptomyces strains were analyzed using Artemis software (Carver et al., 2012), distinguishing nine sequences that presented the abovementioned motifs. These nine potential IREDs were then analyzed using additional criteria reported by Fademrecht et al. (2016) and France et al. (2017). The former characterized the presence of two protein motifs (GLGxMGx5[ATS]x4Gx4[VIL]WNR[TS]x2[KR] and Gx[DE]x[GDA]x[APS]x3 {K}x[ASL]x[LMVIAG]) corresponding respectively to NADPH and substrate binding for IREDs. Additionally, it stablished the importance of amino acids in positions 139 and 194 for determining enzyme stereoselectivity (Fademrecht et al., 2016). Besides, we analyzed the presence of active site residues associated with reductive aminase activity (France et al., 2018). After these analyses, two potential enzymes with complementary stereoselectivity were selected: Ss-IRED_S from S. scabies ST129 and Ss-IRED_R from Streptomyces sp. ST1020. The selected enzymes were cloned in pET-28b(+) and expressed in E. coli BL21(DE3).

IRED activity for the identified enzymes was assessed using a set of substrates 1a–1d (Figure 1) encompassing both monocyclic and bulky prochiral imines. The assays were performed using resting cells under two different conditions that favor NADPH recycling: addition of glucose as co-substrate and coexpression of Lb-ADH. As can be observed in Table 1, the coexpression of ADH for NADPH recycling resulted in the best conversions as it was earlier reported by Velikogne and coworkers (Velikogne et al., 2018). Surprisingly, Ss-IRED_R presented no activity towards these substrates, despite the sequence alignment analysis that suggested IRED activity. On the other hand, Ss-IRED_S exhibited excellent activity towards substrates 1a–1c with superb enantiomeric excess, showing preference for bulky substrates such as those with dihydroisoquinoline and tetrahydro-β-carboline motifs, and a preference for six-membered rings over five-membered rings for monocyclic imines.

To study reductive amination activity, substrates 3–5 were tested using propargylamine (a) as amine donor (Figure 2). For these reactions, we found that recombinant Lb-ADH reduced the ketone; thus, the assays were only performed with glucose as co-substrate for NADPH regeneration. Following that procedure, no ketone reduction was detected. Table 2 illustrates the experimental results obtained. As it can be observed, Ss-IRED_S showed moderate activity for substrate 3 and no activity with the other substrates. In contrast, Ss-IRED_R presented good reductive amination activity with substrates 3 and 5, this last one with excellent enantiomeric excess. Despite many efforts on assaying the reductive amination of phenyl ketone (4), no activity was detected for this substrate.

Imine reduction and reductive amination have been observed with highly diverse sequences, which indicates the importance of analyzing the three-dimensional structure and the characteristics of the active site over that of the linear sequence. In silico analysis was performed for the new enzymes to better understand observed preferences for IRED or RedAm activities, as well as differences in substrate specificity.

Homology models for Ss-IRED_R (Supplementary Figure S1), Ss-IRED_S (Supplementary Figure S2), and 10 IREDs used as controls in complex with NADPH cofactor were constructed using the “closed” form with NADPH of IREDs in PDB because analysis of the (apo-) and (NADPH-bound) subunits within the structure revealed a rotation of 14° (Sharma et al., 2018). Docking studies of substrates 1a–1d and 3–5 and a in the different enzymes’ active site were performed by Autodock-VINA. After clustering of 250 runs of each substrate with each enzyme, the different ligand–protein interactions in the active site were analyzed (Figure 3). To analyze IRED activity, substrates 1a–1d in the different enzymes were selected for docking, and different parameters such as energy of binding, distance for hydride transfer from NADPH to C═N double bond, and distance between nitrogen of imine with surrounding acidic amino acids were evaluated. Adequate distance between the C4 atom of the NADPH and the carbon in the C═N double bond (d_C-C4) is needed for hydride delivery. Additionally, mechanistic studies have suggested the importance of an acidic amino acid at a proper distance to the forming amine to act directly or through a water molecule as a proton donor (Man et al., 2015).

Favorable docking positions were selected as those presenting a d_C–C4 distance minor to 7.5 Å, and a possible proton donor amino acid at a distance minor to 7 Å. Previous reports have indicated Tyr 169 in GF3546-IRED from Streptomyces sp. or Asp 187 in Q1EQE0 IRED from Streptomyces kanamyceticus as candidates for protonating the forming amine (Rodríguez-Mata et al., 2013; Man et al., 2015). The presence of these residues was first analyzed, but the presence of other acidic amino acids at an adequate distance was equally considered to support catalytic activity (Ribeiro et al., 2020). A similar analysis was performed on a group of five IREDs (IREDs*) whose activity was previously reported using the same substrates in order to compare the different parameters (Velikogne et al., 2018). The results are presented in Table 3.

As can be seen in Table 3, the results of this in silico study indicate that no significant difference in energy of binding is detected when considering all the docked positions or only favorable ones, despite evidence of excellent activity associated with these enzymes (columns 1 and 2). Regarding the d_C-C4 distance, it could be observed that selected positions show an average value of 5 Å, as well as those for the control group. Surprisingly, Ss-IRED_R presented the shortest average distance (∼4.5 Å) with all substrates even though no IRED activity was detected with this enzyme. A correlation between d_C–C4 distance and activity cannot be clearly established from these data.

Interestingly, the percentage of docked positions complying with catalytic requirements correlates well with experimental activity. A quantitative analysis of docking positions for substrates 1a and 1b in Ss-IRED_S yields respectively 46% and 73% of favorable positions that meet the catalytic requirements. It is noteworthy that a similar quantitative percentage of positions is observed for the IREDs in control group. Concurrently, for the inactive Ss-IRED_R, only 10% of the positions of substrate 1a and 11% of 1b comply with the expected attributes for catalysis. Additionally, for substrate 1d, IREDs in the control group show an average of 58% of positions presenting an adequate distance to NADPH and the required amino acid for amine protonation. In agreement with these data, Ss-IRED_S and Ss-IRED_R with no detectable activity presented respectively only 0.5% and 31% of positions complying with catalytic requirements. The results for substrate 1c are inconsistent at first sight; however, a deeper analysis allowed us to determine that for the inactive Ss-IRED_R, most docked positions in clusters do not interact with Asp176 but with a different possible proton donor. On the contrary, Ss-IRED_S presents interaction with Asp188 in most positions in the selected clusters, which could explain the difference in activity (Supplementary Table S3).

A similar docking analysis to explain reductive aminase results was investigated; however, no correlation could be established in this case (data not shown). Therefore, we based our assessment of Ss-IRED_S and Ss-IRED_R activity on the work reported by France et al. (2017). The presence of the described critical conserved amino acids in the active site was examined based on an alignment of our enzymes with the best reductive aminases reported by these authors (Supplementary Table S7). We noticed that Ss-IRED_R presents the three amino acids that interact directly with the main functional groups of our substrates via hydrogen bonding, while it also presents the other three conserved amino acids to which no specific function has been assigned. In contrast, Ss-IRED_S only has two of the three conserved amino acids that interact directly with the substrates.