Overall structure of MslH. To investigate the catalytic mechanism of MslH, we performed crystallographic studies. Recombinant N-terminally His6-tagged MslH was expressed in Escherichia coli and purified for crystallization. The purified recombinant MslH migrated as a single band with a molecular weight of 48.5 kDa on SDS–PAGE, in good agreement with the calculated value of 49.4 kDa (Supplementary Fig. 1). In contrast, a gel-filtration experiment indicated a molecular weight of 410 kDa, suggesting that the recombinant MslH is an octameric enzyme (Supplementary Fig. 2). Meanwhile, to obtain the complex structure of MslH with the substrate MslA, a recombinant N-terminally His6-tagged MslAΔW21, which lacks the C-terminal Trp21, was prepared by co-expression with MslB1 in E. coli, since MslA expressed in E. coli is only soluble when complexed with MslA-MslB116, and used for the co-crystallization experiment with the purified recombinant MslH. The co-crystallization was accomplished by mixing MslH:MslAΔW21-MslB1 in a 1:1.2 mole ratio. As a result, crystals of MslH:apo and the MslH:MslAΔW21 complex were obtained that diffracted to 2.30 Å and 2.29 Å resolutions, respectively (Supplementary Table 1). Diffraction data were indexed in the I422 space group, with one monomer in the asymmetric unit. The MslH:apo crystal structure was initially determined by the sulfur single-wavelength anomalous diffraction (S-SAD) method (Supplementary Fig. 3 and Table 1). Consequently, the MslH:MslAΔW21 crystal structure was solved by the molecular replacement (MR) method, using the MslH:apo structure as the search model (Fig. 2 and Supplementary Table 1). Both structures showed similar conformations (Cα Root Mean Square Deviation (RMSD) = 0.17 Å) (Supplementary Fig. 3b). The monomers formed symmetric dimers with a crystallographic twofold axis, and the resultant dimers formed the biologically active octamer with crystallographic twofold axes (Fig. 2c and Supplementary Fig. 4).
Each monomer is composed of a large domain with a grip-like subdomain (G-domain: Ala141-Pro234) and a cover-like subdomain (C-domain: Ser317-Trp363) (Fig. 2a and Supplementary Figs. 5 and 6a). Of the eleven α-helices and 18 β-sheets observed in the MslH:MslAΔW21 crystal structure, six α-helices and twelve β-sheets comprise the large domain. Among them, six β-sheets (β1, β2, and β15-β18) and six β-sheets (β5, β6, and β11-β14) face each other at the center of the large domain, to form a mixed parallel and antiparallel β-strand structure, surrounded by α-helices (α1–3, α7, α8, and α11) to create the αββα barrel architecture (Fig. 2a). Of the remaining α-helices and β-sheets, two α-helices (α9 and α10) participate in the construction of the C-domain. This domain contains several long loops and partially covers one side of the top of the αββα architecture in the large domain. The G-domain is composed of two α-helices (α5 and α6) and four antiparallel β-sheets (β7-β10), and is attached to almost the opposite site of the C-domain on the top of the αββα architecture. Two α-helices (α5 and α6) of the G-domain on the monomer protrude toward another dimer-forming monomer, and participate in the dimer formation together with part of the outer surface of the αββα core structure (Fig. 2c). In addition, the G-domain is encapsulated by the C-domain of the other monomer via several contacts, such as α9’ of the C-domain in the other monomer with the long loop structure between β6 and β7 of the G-domain (G-loop) (Fig. 2a and 2c). As a result, α6’ from another monomer, together with the four antiparallel β-sheets on the G-domain and part of the C-domain of the monomer, forms a large cleft on the top of the αββα architecture in the core structure (Fig. 3a-b).
Structure-based similarity searches revealed that MslH possesses a calcineurin-like fold, classically categorized as phosphoesterase domains in phosphodiesterases, protein (serine/threonine) phosphatases, and 5’-nucleotidases22–24, as part of its structure. The overall structure of the MslH:MslAΔW21 complex is most similar to that of the YmdB phosphodiesterase in Bacillus subtilis (PDB ID: 4B2O)25, with an RMSD value of 2.26 Å for 71 Cα atoms (corresponding to 16% of the amino acids of MslH), even though their amino acid sequence identity is only 16%. The structural similarity is mainly observed in the αββα architecture (Fig. 2b and Supplementary Figs. 6 and 7). A structural comparison of both proteins revealed that the two β-sheets (β17 and β18) and one α-helix (α11) structures on the C-terminal side were characteristic of MslH, among the αββα architectures. Notable differences between the MslH and YmdB structures were observed in the C-domain and G-domain in MslH (Fig. 2a-b and Supplementary Fig. 7). In these structures, the C-domain in MslH is a steric homolog of the structure consisting of loops and α6 in YmdB, although that of MslH comprises loops with two α-helices, α9 and α10. In contrast, the G-domain in MslH is absent in YmdB.
Active-site architecture of MslH. The MslH:MslAΔW21 crystal structure was obtained through the co-crystallization of MslH with the MslAΔW21-MslB1 complex, as the substrate analog. The electron density in the MslH:MslAΔW21 crystal structure allowed us to locate a cyclic peptide molecule (cMslA Gly12-Phe20), consisting of the Gly12-Phe20 part on the C-terminal side of MslAΔW21 with a disulfide bond between Cys13 and Cys19, in the large cleft (Fig. 3a-b). However, no electron density corresponding to either the MslA leader peptide or MslB1 was identified in the MslH:MslAΔW21 crystal structure. The disappearance of these electron densities suggested that these regions may not be crucial for the substrate recognition by MslH. The cMslA Gly12-Phe20 molecule is accommodated in the cleft with various interactions, such as a σ-π interaction between Phe228 of monomer A and Tyr15 in cMslA Gly12-Phe20 (3.8 Å) and hydrophobic interactions between α6’ of monomer B and MslA Ala16-Val18 (Fig. 3b and Supplementary Fig. 8a). However, the disulfide bond observed between Cys13 and Cys19 in cMslA Gly12-Phe20 is not present in the biosynthesized final product, MS-271, which has disulfide linkages between Cys1-Cys13 and Cys7-Cys19 (Fig. 1). It is hypothesized that during MS-271 maturation, the core peptide of MslA generates disulfide linkages between Cys1-Cys13 and Cys7-Cys19 (Fig. 1). To confirm indirectly whether the cMslA Gly12-Phe20 could be bound to the original substrate-binding site, we performed in vitro assays with and without DTT. As a result, the addition of DTT to the MslH reaction had no effect on the MslH catalytic activity (Supplementary Fig. 9). Furthermore, even MslA variants with serine residues at Cys13 and Cys19 (MslA C13/19S), respectively, were competent substrates in the MslH reaction, which exhibited the same activity as in the case of the intact MslA substrate (Supplementary Fig. 9). These observations suggest that the disulfide linkage between Cys13 and Cys19 on cMslA Gly12-Phe20 is not necessary for the catalytic activity of MslH. Thus, the original substrate could also be accepted in the same cleft observed in the MslH:MslAΔW21 crystal structure, and the catalytic reaction site might be present anywhere on or behind this surface. Interestingly, the structure analysis revealed that the C-terminal carbonyl group of cMslA Gly12-Phe20 is located near a hydrophobic pocket consisting of Gly60-62 and Leu64 on the core domain and Ile319, Tyr345, Leu348, Leu355, and Phe356 on the C-domain in monomer A at the cleft (Fig. 3c and Supplementary Fig. 8a). In addition, an architecture that could possibly construct a six-coordination motif, consisting of four amino acid residues, Glu44, Asn87, His259, and His261 and two water molecules (w1 and w2), was detected below the C-terminal carbonyl group (Fig. 3d and Supplementary Fig. 10). The six-coordination motif is located at the entrance of the central ββ barrel behind the cleft. In the six-coordination, one of the water molecules (w2) hydrogen-bonds with the side-chain of Asp11 (2.4 Å) and is further held in the structure by the steric contraction of the His295 side-chain, which forms a hydrogen-bond with Asp11. The other water (w1) is held in the structure, with Asp91 forming a hydrogen-bond with His88. We also found some remaining electron density, which was obviously different from those of the peptides and water molecules, in the center of the six-coordination system. When another water molecule (w3) was added to the observed electron density at the center of the six-coordination site, the Fo-Fc map remained positive around the added water molecule, indicating the presence of an unidentified metal molecule instead of a water molecule (Supplementary Fig. 11a).
Previous studies predicted that MslH is a non-metal-dependent protein, even though it adopted a calcineurin-like fold structure as mentioned above. Enzymes with calcineurin-like folds, including YmdB, are reportedly metal-dependent enzymes23–25. The comparison of the MslH:MslAΔW21 and YmdB crystal structures indicated that the residues and water molecules involved in the formations of the six-coordination motif and related hydrogen-bond networks from the water molecules in the MslH:MslAΔW21 crystal structure are located at positions very similar to those in the Fe(II) metal-binding-site of YmdB, and consist of Asp8, Glu39, Asn40, Asn67, His150, His175, His177, His68, Asp71, and a water molecule (Supplementary Fig. 10). In particular, the positions of Asn87 and His259 lining the six-coordination motif and His88 and Asp91 forming the hydrogen-bond networks from the water molecules in MslH are nearly identical to those of Asn67 and His150, and His68 and Asp71 of YmdB, respectively (Supplementary Fig. 10c). Although the side-chain of Asp11 in MslH was slightly rotated as compared with that of Asp8 in YmdB, both residues are located in almost the same position. The main-chain of His261 in MslH, corresponding to His175 in YmdB, has a significantly different position from that of His175 in YmdB, due to the substantial conformational change between the loop containing His261 in MslH and the corresponding loop containing His175 in YmdB. Meanwhile, the side-chain of His261 is also in almost the same position as that of His175 in YmdB, to participate in the six-coordination. In contrast, MslH lacked sufficient space to bind the second Fe(II) found in YmdB, presumably due to several conformational changes such as the lack of the amino acid residue at the location occupied by Asn40 in YmdB, the rotational and positional changes of the side-chain of Glu44 corresponding to Glu39 in YmdB, and the rotation of the side-chain of His295 in MslH by 90 degrees, as compared with that of His177 in YmdB. Thus, the MslH:MslAΔW21 crystal structure suggested that MslH possesses a rearranged structural analog of the Fe(II)-binding site in YmdB.
Ca(II)-binding of MslH. To clarify the importance of the six-coordinated motif and the metal-binding in the MslH catalytic mechanism, we performed mutagenesis studies and ICP-MS analyses of purified MslH. We also considered the previous report that YmdB employs His68 and Asp11, corresponding to His88 and Asp91 in MslH, as the catalytic residues. Therefore, the site-directed mutagenesis studies included not only the four amino acid residues (Glu44, Asn87, His259, and His261) in the six-coordination motif, but also the four residues (Asp11, His88, Asp91, and His295) that follow the observed six-coordination motif with the hydrogen-bond networks via the water molecules (Fig. 3d). In the mutagenesis studies, all residues except for His295 were substituted with alanine, while His295 was substituted with asparagine, with the same size as histidine, because of the instability of the His295A variant during the purification and enzyme reaction. The mutagenesis studies revealed that the MslH Glu44A, Asn87A, Asp11A, His88A, Asp91A, and His295N variants lacked epimerase activity and the MslH His259A and His261A mutants had diminished activity, suggesting the crucial roles of the six-coordination motif and the related hydrogen-bonding networks (Fig. 4). We also attempted to crystallize all of the variants to clarify the mutagenesis results from the structural aspect, and successfully obtained a crystal structure of the MslH Asp11A variant (MslH D11A:apo) at 2.2 Å resolution (Supplementary Fig. 12 and Table 1). The MslH D11A:apo crystal structure indicated that the variant retained the same six-coordination motif as that in the MslH:MslAΔW21 crystal structure (Supplementary Fig. 12c-d), whereas the Asp11A substitution led to the flexibility of His295 in the MslH D11A:apo crystal structure, where a 1:1 alternative state was observed for the His295 residue. The main chain carbonyl group of the Pro294 residue in the MslH D11A:apo crystal structure was inverted by 180 degrees, as compared with that of the wild type (Supplementary Fig. 12e). These observations suggested that the Asp11 and His295 residues could play important roles in maintaining the appropriate catalytic state of MslH.
Furthermore, the ICP-MS measurements unveiled the presence of the Ca(II) and Mn(II) ions at ratios of 0.4 mole and 0.3 mole per mole of MslH, respectively (Supplementary Table 2). These results are consistent with previous reports, in which MslH catalyzed epimerase reactions without metal addition16. Presumably, these metal ions were incorporated into MslH during enzyme expression in E. coli and remained tightly bound to MslH. To identify the metal ions integrated into the enzyme, the MslH:MslAΔW21 crystal structure refinements were performed by accommodating Ca(II), Mg(II), Mn(II), and Fe(II) ions in the aforementioned remaining electron density in the six-coordination motif. The Fo-Fc density maps displayed a positive value for Mg(II) and negative values for Mn(II) and Fe(II), but were neither positive nor negative for Ca(II), suggesting that the presence of a Ca(II) ion in MslH was the most reasonable (Supplementary Fig. 11b-e). Since the B-factor values of metal-coordinating amino acid residues and water molecules are typically near the metal’s B-factor value, we also investigated their B-factor values. Consequently, the B-factor value of the Ca(II) ion in the crystal structure was the closest to the average of those of the metal-coordinating residues and water molecules, while that for Mg(II) was below and those for Mn(II) and Fe(II) were above the average (Supplementary Table 3). Taken together, the aforementioned data allowed us to conclude that MslH is a Ca(II)-dependent epimerase.
Catalytic mechanism of MslH. MslH probably catalyzes the epimerization of the C-terminal L-Trp21 residue of MslA in a Ca(II)-dependent manner. However, in the MslH:MslAΔW21 complex structure, the MslA Trp21 residue was absent from the reaction site. Therefore, we attempted co-crystallizations of the wild-type or variants of MslH with full-length MslA to obtain these complex structures, but were unsuccessful. In the absence of a useful structure for further analysis of the MslH catalytic mechanism, we performed docking studies of MslA with the Gly12-Trp21 residues, including the disulfide linkage between Cys13 and Cys19, to the MslH:MslAΔW21 structure. The initial model of MslA Gly12-Trp21 was designed, based on the cMslA Gly12-Phe20 architecture in the MslH:MslAΔW21 structure. As mentioned above, in the cleft of the MslH:MslAΔW21 structure, we observed a hydrophobic pocket near the C-terminal Phe20 residue of MslAΔW21 that could possibly accommodate the MslA Trp21 residue (Fig. 3c). Hence, we selected several amino acid residues contributing to the hydrophobic pocket flexibility, and ran AutoDock Vina docking simulations using the designed MslA Gly12-Trp21 peptide as the initial phase. The docking study provided an MslA Gly12-Trp21 model with peptides bound in the cleft, in almost the same location and orientation as those of MslAΔW21 in the MslH:MslAΔW21 structure, and its C-terminal Trp21 was accommodated in the hydrophobic pocket, as expected (Fig. 5a-b and Supplementary Fig. 8b). In the model structure, the side-chain of Phe20 of the MslA Gly12-Trp21 molecule formed a parallel-displaced π-stacking interaction (2.7 Å) with the indole ring of its own C-terminal Trp21. To accommodate the C-terminal Trp21, the side-chain of Ile319 in the MslH:MslA Gly12-Trp21 model was slightly shifted toward the outside of the cleft to widen the hydrophobic pocket, as compared with that of the wild type. The MslA Gly12-Trp21 model in the docking study and the MslAΔW21 molecule in the MslH:MslAΔW21 crystal structure showed structural differences at Gly12, Tyr15, and Phe20. Among them, in the hydrophobic pocket, the MslH Phe356 residue formed a π-π stack (3.5 Å) with the indole ring of the MslA Trp21. Furthermore, the oxygen atom of the C-terminal carbonyl group of the MslA Gly12-Trp21 model was located close to the water molecule (w1) in the six-coordination motif with the Ca(II) ion (Fig. 5c). Intriguingly, His88 coupled with Asp91 was near the α-proton of L-Trp21 in the MslA Gly12-L-Trp21 model, while His295 coupled with Asp11 was opposite from the α-proton, and these His residues were positioned across the Cα center of the MslA Trp21 residue in a linear manner (Fig. 5c). Similar cases were also found in other epimerases that catalyze reversible epimerization. For examples, the catalytic process of the PLP-independent epimerase, apMurI26, and a Mg(II)-dependent epimerase, AE epimerase27, enable reversible D- and L-tautomerization of an amino acid by positioning the two catalytic Cys residues and the two catalytic Lys residues against the reaction point of the substrate in a linear manner, respectively. It is thus assumed that the His88 and His295 residues observed in MslH could act as the catalytic residues. Furthermore, the observed linear configuration of the His residues coupled with Asp residues would allow MslH to facilitate the reversible epimerization of the L- and D-amino acid orientations of MslA Trp21, which has indeed been seen in the in vitro MslH reaction, as previously reported16.
Based on the comprehensive data described above and previous reports, we propose that MslH employs a unique “acid/base” chemistry in the epimerization reaction to generate MslA D-Trp21, as follows (Fig. 6a): MslH accepts the Gly12-Trp21 portion of the MslA/MslB1 complex substrate within the cleft, and initiates the α-proton abstraction of L-Trp21 in the MslA/MslB1 complex by Asp91-activated His88, to produce an enolate intermediate in a transition state with the Ca(II) ion. Subsequently, a proton transfer from the His295 side-chain, locating on the opposite side, occurs on the α-position of the enolate intermediate to produce MslA D-Trp21 via a tautomerization of the enolate form to the keto form of the intermediate, and the His295 side-chain further abstracts a proton from the Asp11 side-chain. Finally, MslH releases MslA D-Trp21 as the final product, which is subsequently subjected to the tailoring biosynthesis, and recovers the initial state of His88-Asp91 by proton transfer starting from His295 to Asp91 via the Asp11-w2-Ca(II)-w1-His88 network observed in the crystal structure. In contrast, when MslH accepted MslA D-Trp21 as the substrate, the enzyme facilitates the epimerization reaction of MslA D-Trp21 to MslA L-Trp21, by employing Asp11-activated His295 as the base catalyst (Fig. 6b). As a side note, although the initial proton abstraction may occur with His295, His88 most likely acts as the initial catalytic base, based on our docking study (Fig. 5c).