The crystal structure of V. alginolyticus PepD

3.2-1 Overall structure

The crystal structure of V. alginolyticus PepD was solved by the molecular replacement method using the structure of Xaa-His dipeptidase from Haemophilus somnus 129PT (PDB code 2QYV) as the search model (sequence identity: 50.9%). The 2QYV was solved and deposited with the PDB by the Joint Center of Structure Genomics (JCSG), but was not yet published. The structure of PepD was then refined to a resolution of 3.0 Å with an R factor of 23.1% and an Rfree factor of 27.4% (Apprndix 2).

The overall structure of the PepD monomer was determined to be comprised of a total of 486 residues in two domains: a catalytic domain harboring two zinc ions for catalysis and a lid domain functioning in substrate recognition and protein dimerization (Fig. 3.2A). Analysis of the X-ray absorption measurement and electron density map confirmed the presence and locations of the di-Zn²⁺ ions held captive in the catalytic domain (Fig. 3.3). The high B-factors that were obtained were presumed to reflect the flexible open conformation of the catalytic and lid domains. Upon comparison with PepV and other related di-zinc-dependent M20/M28 family members, PepD was found to share similar structural folds, despite the low sequence similarities that exist among each. PepD and PepV showed root mean square deviations (rmsd) of 4.0 and 4.3 Å for Cα atoms of the catalytic and lid domains, respectively (Appendix 3).

In addition, two asymmetric unit of PepD having dimensions of ~90 × 90 × 95 Å was determined to be two PepD molecules packed together as a dimer (Fig. 3.2B). The apparent dimeric and monomeric characteristics of native and denatured PepD were further supported by evidence from analytical ultracentrifugation, which revealed

42

molecular masses of 100.7 and 51.1 kDa under physiological and denatured conditions, respectively (Appendix 4). The lid domain was found to utilize a hydrogen bonding network between helices from each monomer in order to form the dimer interface. PepD was determined to exist as a dimer, similar to the related di-zinc-dependent enzymes of the M20/M28 family, but different from PepV which uniquely exists as a monomer.

43

Figure 3.2. Overall structure of V. alginolyticus PepD. (A) Stereo view of a subunit of V. alginolyticus PepD. Secondary structure elements are shown in red (α-helices) and blue (β-strands). Gray spheres represent the zinc ions. (B) Ribbon diagram of the PepD dimer. The same color scheme in (A) was used for the left subunit, and the right subunit was distinguished by purple (helices) and green (strands).

Figure 3.3. Determination of the PepD zinc ions. (A) The absorption spectra of anomalous scattering factors for zinc ions in PepD are presented as a function of X-ray energy. (B) The electron density map of PepD zinc ions binding site is presented as part of a composite-omit map contoured at 1.0 σ (blue), and anomalous difference Fourier map contoured at 4.0 σ (red).

3.2-2 The catalytic domain

Comparative analysis of the PepD catalytic domain indicated that it has a fold similar to those of PepV and the related di-zinc-dependent M20/M28 family of enzymes, including CPG2, βAS, mouse CN2, PepT, ApAP and SgAP26, 29, 32, 35, 87, 88. The topology of PepD and PepV is illustrated in Fig. 3.4. The catalytic domain consists of residues 1–186 and 401–490 and has a mixed three-layer α/β/α-sandwich architecture composed

44

of two β-sheet groups and seven α-helices (Fig. 3.5). The large sheet group contains eight strands arranged in the order a-b-f-c-g-j-h-i, in which b is the only antiparallel strand. The small sheet group is composed of four shorter antiparallel strands arranged in the order of d-e-l-k and located on the surface of the catalytic domain. The zinc ions are located at the C-terminal end of the four central parallel strands.

The active site was found to be located within a deep cleft that formed between the lid and the catalytic domain (Fig. 3.2). In the dimer, the two active sites are ~57 Å apart, suggesting that each protomer can function independently. No distinct zinc-bound water molecule was found in our structure analysis; however, a higher electron density peak was observed with the closest zinc-water contact of 2.5 Å. The absence of the zinc-bound water molecule could be a reflection of the relatively limited resolution of the data. The N- and C-termini are both located at the top of the catalytic domain, opposite to the lid domain and the active site.

45

Figure 3.4. Topological diagrams of (A) V. alginolyticus PepD and (B) the lid domain of L. delbrueckii PepV. The secondary structural elements of α-helices and β-strands are represented by ribbons and arrows, respectively. The diagram (A) has been divided into two parts, separated accordingly by the dotted line; the region with gray in the lid domain represents the extra domain.

46

Figure 3.5. Comparisons of catalytic domains from PepD, PepV and CPG2. (A) Superposition of catalytic domains from PepD (blue), PepV (orange), and CPG2 (yellow green) after optimal fit. The zinc ions of PepD are depicted by two gray spheres. (B) Structure-based sequence alignment of the three catalytic domains. The sequence corresponding to the α-helices and β-strands are highlighted in orange and blue, respectively. The conserved metal ion binding residues are highlighted in red. Each of the three catalytic domains has the same di-zinc binding residues, with the exception of Glu²⁰⁰ in CPG2.

47

Biochemical studies of V. alginolyticus PepD revealed its metal-dependent characteristics. Optimal activation of apo-PepD was observed with various divalent metal ions, including Mn²⁺, Co²⁺, Ni²⁺, Cu²⁺, and Cd²⁺. Previous studies have shown that addition of Co²⁺ ions to apo-PepD can sufficiently augment the enzymatic activity by a factor of ~1.4 over that of the wild-type PepD containing Zn²⁺. The fact that the simultaneous presence of the Zn²⁺ ions did not inhibit the Co²⁺-loaded PepD activity indicates that the metal-binding cavity can not only accommodate forced ion loading but can retain functionality. On the other hand, when the Zn²⁺ ions were substituted for Mg²⁺ ions, the enzyme retained ~80% of its optimal activity.

We confirmed the presence and locations of zinc atoms by X-ray absorption and electron density map performed at beamline 13B1 (Fig. 3.3). The di-zinc center was located on the surface of the cleft between the catalytic and lid domains, suggesting it is solvent-accessible. Further analysis of the PepD crystal structure also revealed that several functional residues interact with one another to fix the two zinc ions (Zn1 and Zn2) in place, separated by a distance of 2.8 Å (Fig. 3.6). Zn1 was found to be coordinated by Nε2 from His⁴⁶¹, one of the carboxylate oxygens of Asp¹¹⁹, and by a single putative water molecule bound via hydrogen bonding to the carboxylate group of Glu¹⁴⁹. Zn2 was found to be coordinated by Nε2 of His⁸⁰, the other carboxylate oxygen of Asp¹¹⁹, and by two carboxylate oxygens of Asp¹⁷³. Asp¹¹⁹ appeared to be positioned as a bridging ligand between the two zinc ions. Notably, this residue is connected to an asparagine via a cis peptide bond, identical to the structure that has been observed in many of the other di-zinc-dependent enzymes of the M20/M28 family, including PepV and CPG2 (Appendix 5). This particular peptide bond can break the α-helix at the N-terminal end in order to facilitate closer positioning of the Asp-Asn dipeptide and presumably modulate subsequent enzyme activity.

48

The orientation of the side chain carboxylate group of Glu¹⁵⁰ was noted to differ between the active-site center of PepD and that of the related M20/M28 family metallopeptidases. In PepV, the two carboxylate oxygens of Glu¹⁵⁴ point inward to Zn1 at a distances of 1.9 and 2.6 Å, respectively; in PepD, however, the two carboxylate oxygens of Glu¹⁵⁰ point away from Zn1, increasing the distance between the carboxylate oxygen and Zn1 to 4.5 Å. Thus, the particular role played by Glu¹⁵⁰ for metal ion binding in PepD remains ambiguous.

Figure 3.6. Comparison of the active sites in V. alginolyticus PepD and structural homologs. (A) Local view of the di-zinc center of PepD. The residues involved in coordination of Zn1 and Zn2 (gray spheres) are shown as green sticks. The putative water molecule is depicted by a red dot. Asp¹¹⁹ of PepD serves as a bridging ligand for metal coordination. (B) Local view of the di-zinc center of PepV. The residues involved in metal coordination are shown as cyan sticks, and the phosphinic inhibitor (AspΨ[PO2CH2]AlaOH) is represented by yellow sticks.

3.2-3 The lid domain

The lid domain of PepD consists of 214 residues (187–400) between strands h and i of the β-sheet in the catalytic domain (Fig. 3.4A). The lid domain folds into a central eight stranded antiparallel β-sheet, flanked on one side by four α-helices packed in

49

alternating orientations (Fig. 3.7). The antiparallel β-sheets are arranged in the order of V-II-IV-III and VI-VII-I-VIII, respectively (Fig. 3.4A). Interestingly, the structure of the lid domain of PepD resembles that of PepV, but shares only a portion of its structure with other related dimeric M20/M28 family enzymes, including CPG2, βAS, CN1, CN2, and PepT. The CPG2 dimer exhibits continuous β-sheets across the two monomers to form the dimer interface, whereas the lid domain of PepD formed the dimeric interface through hydrogen bonding between helices. Moreover, PepD formed a unique criss-cross configuration via the interface interaction of the respective lid domains.

Helices 6, 7, and 8 (Fig. 3.4A) were found to participate in the monomer-monomer contacts. Specifically, the carboxylate oxygens of Glu²⁹⁴ and the hydroxyl group of Ser³⁷⁴, as well as the C = O from the amide side chain of Asn³²⁹ and the hydroxyl group of Ser³⁸⁵, are hydrogen-bonded to each other and form the dimeric interface (Fig. 3.8).

Figure 3.7. Comparison of the lid domain structures of PepD, PepV, and CPG2. The three residues that are putatively involved in substrate C-terminal and/or transition state binding of PepD (A), PepV (B) and CP G2 (C) are shown as sticks and labeled. The

“extra” domain regions of PepD and PepV, which are absent in CPG2, are shown in black and blue, respectively. Notably, the Arg²⁸⁸ of CPG2 (C) is located on the opposite side of the monomer lid domain, which is spatially different from that of Arg³⁶⁹ of PepD (A) and Arg³⁵⁰ of PepV (B).

50

Figure 3.8. PepD dimeric interface. Residues involved in the dimeric interface, including Glu²⁹⁴, Asn³²⁹, Ser³⁷⁴, and Ser³⁸⁵ of PepD, are shown as sticks and annotated by labels.

3.2-4 Structure comparison of V. alginolyticus PepD and related di-zinc-dependent M20/M28 family enzymes

To further characterize the structural features of PepD, structures of related M20/M28 metallopeptidase family members were superimposed onto PepD. A close overall similarity was observed between PepD and the uncharacterized PDB code 2QYV protein that had been previously solved by the Joint Center for Structural Genomics (JCSG). However, the PepD shows an open conformation, and 2QYV shows a closed conformation (Appendix 6). Sequence alignment of these two proteins revealed a 50.9%

sequence identity (Appendix 7). The root mean square deviation of structure similarity between PepD and 2QYV for Cα atoms was calculated to be 0.63 and 0.73 Å among the

51

catalytic and lid domains, respectively. Although both proteins share a structurally-conserved active site, two notable regions connecting the catalytic and lid domains (PepD residues 183–187 and 400–403 vs code 2QYV residues 179–183 and 397–400, respectively) showed minor differences between the proteins in loop conformations. In addition, the PepD protein also exhibited limited amino acid yet overall folding similarity to the M20/M28 metallopeptidases, except in the region of the dimer topology.

The catalytic domain of PepD superpositioned well with the single domain structures of the S. griseus SgAP and A. proteolytica ApAP, and to the two domain structures of the Pseudomonas sp. CPG2, S. typhimurium PepT, S. kluyveri βAS, human Acy1, and mouse CN2, as well as to the counterpart of L. delbrueckii PepV. A major structural difference, however, was found to exist between PepD and the related di-zinc-dependent M20/M28 metallopeptidases in the lid domain; PepD consists of an eight-stranded β-sheet and four α-helices, similar to that of PepV, but the enzymes from the di-zinc-dependent M20/M28 family are composed of only one four-stranded antiparallel β-sheet flanked by two α-helices.

Furthermore, part of the lid domain of the PepD structure is completely superimposable to that of the two domain structures of Pseudomonas sp. CPG2, S.

typhimurium PepT, S. kluyveri βAS, human Acy1, and mouse CN2. These proteins are known to form a dimer interface through hydrophobic interactions between helices, as well as through hydrogen bonds between the two β-strands within the lid domain.

Nevertheless, PepD exhibited a different dimeric architecture from that of the compared dimeric proteins in that the two lid domains of the dimeric proteins mediate enzyme dimerization through side-by-side packing of their four-stranded β-sheets to form a contiguous extended eight stranded sheets. In contrast, a crisscross configuration was

52

observed in PepD, wherein the lid domain formed the dimeric interface through hydrogen bonds between two sets of four α-helices (Fig. 3.8). Although the above-mentioned M20/M28 family metallopeptidases all consist of homodimer structures similar to CPG2, no report has appeared in the literature to date that discusses the PepD-like criss-cross dimeric architecture.

The structure of PepD aligns well with the counterpart of PepV, which is a monomer and does not have a known function in subunit dimerization. The lid domain of PepV partially resembles the lid domain of CPG2, but is about two times larger (Fig.

3.7). Moreover, the PepV lid domain extends itself away from the active site of the catalytic domain, folding back over the active site and facilitating the catalytic domain formation of a cavity that is uniquely involved in substrate specificity. Surprisingly, the lid domain of PepD, which is also about twice as large as that in CPG2 and other related dimeric proteins, is able to form a dimer instead of a monomer. Lindner et al. reported that the eight-stranded β-sheets comprising the lid domain of PepV can be divided into two subdomains¹⁰, both of which exhibit the same topology as the lid domain of CPG2

and together can mimic the arrangement of the two lid domains within the CPG2 and PepT dimers. However, dimerization of the subunits in PepD was found to be mediated through hydrogen bonding of the α-helices, and not by side-by-side packing of the β-sheets. Moreover, one additional region encompassing residues 186–203 and 311–400 within the lid domain of PepD was found to be similar to the lid domain of related dimeric proteins, and was renamed as the “extra" domain. Although the function of the extra domain has been hypothesized, the true physiological function of the extra domain of PepD remains unclear.

Currently, the crystal structures of the M20 family of proteins have been reported for two different (open and closed) conformations. When a protein is crystallized in its

53

free form, the catalytic and lid domains are expected to be in an orientation that exposes the active site to bulky water; whereas, when a protein is crystallized in complex with an inhibitor, the closed conformation is expected. In the PepV-inhibitor complex, a fixed

“bridging” water molecule was found to be located between both zinc ions and close to the carboxylate group of the catalytic Glu¹⁵³; this residue corresponds to Glu¹⁴⁹ of PepD and has been proposed as necessary for substrate hydrolysis (Fig. 3.6). Upon binding of the substrate, the water molecule will be positioned between the zinc ions and the carbonyl carbon of the scissile peptide bond. Then, an attacking hydroxyl ion nucleophile is able to be subsequently generated through activation of the water molecule by both the zinc ions and transfer of the resultant proton to the Glu¹⁵³. Proximal to the Glu¹⁵³ of PepV is the conserved metal binding residue, Glu¹⁵⁴, which utilizes its carboxylate oxygen to bind to the zinc ion with a distance of less than 3.0 Å.

The carboxylate oxygen of Glu¹⁵⁴ of PepV is directed toward the Zn1. Nevertheless, our structural analysis of PepD in an open conformation revealed that the carboxylate oxygen of the corresponding Glu¹⁵⁰ residue is directed away from the Zn1 at a distance of 4.5 Å.

It has been suggested that dipeptidases of M20 families can change their conformation from opened to closed during the process of enzymatic catalysis. The conformational change could be achieved by movement of the catalytic and lid domains (Fig. 3.9). Consistent with this notion is the presence of a large clearance between the two domains that would allow a peptide chain to move to the opened active site, as was observed for the PepD structure. We, therefore, speculated that upon substrate binding, the PepD protein may change its metal ions’ coordination and/or its protein conformation; the carboxylate oxygen of Glu¹⁵⁰ would be subsequently swung toward Zn1 and would push the Glu¹⁴⁹-bound water molecule toward Zn2, effectively bridging

54

the water between the two zinc ions. However, the precise molecular interactions between the enzyme active site and the substrate or inhibitor still await final X-ray structure determination.

Figure 3.9. Open and closed conformations of PepD and PepV. (A) The overall structure of PepD in the open conformation is presented. (B) The overall structure of PepV in the closed conformation is presented. The substrate C-terminal binding and/or transition state binding residues, including Arg, His, and Asn, are shown as sticks from left to right in the lid domain; in the catalytic domains, Glu¹⁴⁹ (PepD) and Glu¹⁵³ (PepV) are shown, respectively. The phosphinic AspΨ[PO2CH2]AlaOH inhibitor in PepV is represented by sticks.

The di-zinc binding ligands, His⁸⁰, Asp¹¹⁹, Glu¹⁵⁰, and His⁴⁶¹, were found to be conserved among all of the proteins compared in this study. In contrast, the ligand Asp¹⁷³ was found to be replaced by a Glu residue in CPG2 and hACy1. This finding is consistent with an earlier observation reported by Lindner et al.¹⁰, in which all homologs

55

with proven aminopeptidase or dipeptidase specificity were found to contain an aspartic acid, whereas a glutamic acid residue has been identified in the same position in Acyl1/M20 family members that exhibit either aminoacylase or carboxypeptidase specificity. Moreover, four additional residues were found to be conserved among all of the compared proteins: 1) Asp⁸² is located at two residues downstream from the His⁸⁰ and in the vicinity of the zinc center, and is assumed to clamp the imidazolium ring of His⁸⁰; 2) Glu¹⁴⁹ is a putative general base for enzyme catalysis; and 3) His²¹⁹ and Arg³⁶⁹ are putative substrate C-terminal and/or transition state binding residues. On the other hand, within the PepV, CPG2, and related M20/M28 family metallopeptidases, a cis peptide bond exists between the bridging Asp and the proximal residue. In CPG2, ApAP, PepT, and PepV, this residue is an Asp, which is replaced by Asn in PepD (Asn¹²⁰) and SgAP (Asn⁹⁸). The cis-peptide has been proposed to participate in forcing the bridging carboxylate to conform to the correct geometry required for metal binding.