The X-Ray Structure of Carboxypeptidase A Inhibited by a Thiirane Mechanism-Based Inhibitor

Abstract

The three-dimensional X-ray crystal structure of carboxypeptidase A, a zinc-dependent hydrolase, covalently modified by a mechanism-based thiirane inactivator, 2-benzyl-3,4-epithiobutanoic acid, has been solved to 1.38 Å resolution. The interaction of the thiirane moiety of the inhibitor with the active site zinc ion promotes its covalent modification of Glu-270 with the attendant opening of the thiirane ring. The crystal structure determination at high resolution allowed for the clear visualization of the covalent ester bond to the glutamate side chain. The newly generated thiol from the inhibitor binds to the catalytic zinc ion in a monodentate manner, inducing a change in the zinc ion geometry and coordination, while its benzyl group fits into the S1′ specificity pocket of the enzyme. The inhibitor molecule is distorted at the position of the carbon atom that is involved in the ester bond linkage on one side and the zinc coordination on the other. This particular type of thiirane-based metalloprotease inhibitor is for the first time analyzed in complex to the target protease at high resolution and may be used as a general model for zinc-dependent proteases.

Metalloproteases from the M14 family (carboxypeptidases, CPs) are involved in many physiologic and pathological processes such as tissue organogenesis (1–3), acute pancreatitis (4,5), diabetes (6,7), inflammation (8,9), fibrinolysis (10,11), neurologic diseases (12), and cancer (13–16). Circulating forms of CP may be used as prognostic tools for the early detection of cancer and other diseases (17). The M14 family of proteases comprises an extended and complex array of zinc-dependent proteins with wide genomic distribution (18–20). They share with matrix metalloproteases (MMPs; e.g., gelatinases, collagenases, matrilysins, and stromelysins) a similar catalytic mechanism, whereby the catalytic zinc ion predisposes the substrate to turnover with the involvement of a conserved glutamate residue in the active site (21). Notwithstanding the shared features of their respective mechanisms, MMPs are endopeptidases, whereas CPs are exopeptidases that remove the C-terminal residues from peptide substrates.

Most known small-molecule inhibitors of CPs are chelators of the active site zinc ion. These chelators are often non-discriminant, as their function hinges on coordination with metals. The limited selectivity among potent zinc chelators can be exemplified by thiol-dependent CP inhibitors, such as 1 (2-mercaptomethyl-3-guanidino-ethylthiopropanoic acid, Plummer’s inhibitor (22)), which was still found to inhibit other M14 proteases (23). A similar concern has been encountered in the development of MMP inhibitors with subtype selectivity (24). Mechanism-based inactivators (‘suicide substrate’ or ‘k_cat inactivator’) exploit features of the catalytic mechanisms of the targetted enzymes (25). An innocuous agent is converted to the inhibitory species only within the active site of the target enzyme. Compound 2, 2-((4-phenoxyphenylsulfonyl)methyl)thiirane, SB-3CT, is a prototypical example. It was designed as a selective gelatinase (MMP-2 and MMP-9) inhibitor. Its unique features include a biphenyl moiety that would fit the S1′ pocket of gelatinases and a thiirane ring able to revert into a thiol group upon enzymic reaction (26,27). The oxirane counterpart to 2 is three orders of magnitude less potent than the thiirane (28).

Chemotherapeutic compounds that may be used as drugs or imaging agents targeted to M14 proteases are highly sought (18,29), and the complexity among CPs calls for the development of selective compounds. Based on the notion that a covalently bound mechanism-based CP inactivator might be an ideal starting point for such an endeavor, a screening of our collection of thiirane-based MMP ligands (~500 compounds) against representative M14 proteases was performed. Unfortunately, these molecules did not prove efficacious in inhibition of these carboxypeptidases. We hasten to add that thiirane-based inhibitors of MMPs have not been amenable to study of their structures in complex with the MMPs by X-ray analysis, hence structural information is currenly lacking (30). In an effort to elucidate the minimal structural motifs for inhibition of carboxy-peptidases by thiirane-based inhibitors, we turned to two examples from the literature. Compound 3 (2-benzyl-3,4-epithiobutanoic acid), as well as its oxirane version, 4 (2-benzyl)-3,4-epoxybutanoic acid), have been reported as mechanism-based inactivators of carboxypeptidase A (CPA) (31–33). An earlier structural determination of CPA in complex with oxirane 4 (34,35) has been reported, although the atomic coordinates have not been deposited. Kinetic results showed that thiirane 3 and oxirane 4 displayed a similar inhibitory potency as CPA inactivators, contrary to the example of gelatinase inhibition by 2.

We synthesized compound 3 for its study with CPA. Its binding to the active site would allow for the coordination of the thiirane sulfur with the zinc ion, making the thiirane more electrophilic and predisposing it to nucleophilic attack by the active site glutamate. Herein, we report the X-ray structure of the enzyme inhibited by covalent modification with this compound.

Methods and Materials

CPA crystals were grown from a 2 μL:2 μL mixture of enzyme solution (14 mg/mL in 0.02 M Tris, pH 7.5) and precipitant (20% PEG 3350, 0.2 M NH₄Cl, 0.02 M Tris pH 7.0) by the vapor diffusion method. The largest crystals were harvested and transferred to a 2 μL drop containing the reservoir solution plus 10 mM of 3. Soaking lasted for one week before data collection time, at which no crystal cracking or damage was evident. Crystals were harvested, briefly bathed in cryoprotectant buffer (i.e., the reservoir solution with added 30% glycerol), flash frozen in nitrogen stream and diffracted. Some crystals gave excellent quality diffraction images and that of the highest resolution was used for structure determination by the molecular replacement procedure using native CPA (Protein Database: 2ctb) as the model. Programs from the CCP4 suite were used in different stages of structure determination (36). Extra electron density was clearly evident in the region near the catalytic zinc ion and was easily interpreted as the inhibitor molecule. Further, clear continuous electron density unequivocally revealed the covalent linkage between the inhibitor and Glu-270 side chain. The agreement between the model and the experimental observations is excellent, as indicated by the crystallographic and refinement statistics (Table 1). Geometric details of inhibitor–enzyme interactions are listed in Table 2.

Table 1.

Statistics of data collection and refinement for inhibitor-bound CPA

Parameters	Value
Wavelength used during data collection	0.8123 Å
Unit cell constants	a = 42.32 Å, b = 57.50 Å, c = 57.08 Å
	α = 90.0°, β = 99.1°, γ = 90.0°
Resolution range	31.27–1.38 Å
Space group	P21 (1 mol/asymmetric unit)
Number of measured reflections	488 546
Number of unique reflections	54 345
Rmergea (overall/outermost shell)	6.8/37.9%
Completeness & multiplicity (overall/outermost shell)	96.6/77.8% 6.1/5.3
I/σI (overall/outermost shell)	8.8/1.8
Reflections used for refinement (total/test set)	52 590/1102
Crystallographic Rfactorb/Rfreec	14.3/15.7%
Deviation from ideality
r.m.s.d. bond lengths	0.008 Å
r.m.s.d. bond angles	1.14°
Number of protein atoms/total atoms	2438/2930
B-factor statistics (Å2)
Overall B-factor/Wilson plot B-factor	10.7/9.3
Catalytic domain, main/side chain	7.3/8.4
Zn2+ (1 in total/1 mol per monomer)	5.3
Inhibitor atoms (14 in total/1 mol per monomer)	5.8
Glycerol atoms (42 in total/7 mols per monomer)	19.4
Solvent atoms (435 in total)	26.0
Protein geometryd
Ramachandran favored	97.3% (293 of 301 residues)
Ramachandran allowed	99.7% (300 of 301 residues)
Ramachandran outliers	0.3% (1 of 301 residues, Ser-199)
Residues with bad bonds/angles	0.00/0.00%
Rotamer outliers	0.77%

^aR_merge = Σ_hkl Σ_{j = 1} to N |I_hkl−I_hkl (j) |/Σ_hkl Σ_{j = 1} to N I_hkl (j), where N is the redundancy of the data. The outer-most shell is 1.46–1.38 Å.

^bR_factor = Σ_hkl ||F_obs|−|F_calc||/Σ_hkl |F_obs where F_obs and F_calc are the observed and calculated structure factor amplitudes of reflection hkl.

^cR_free = R_factor for a randomly selected 2% subset of reflections that were not used in refinement.

^dAccording to Molprobity (41).

Table 2.

Geometric details inhibitor–CPA interactions

Interaction	Parameter (Å, °)	Comments
Zn/His-69, Zn/His-196, Zn/Glu-72-Zn, Zn/S1	2.10, 2.11, 2.04, 2.33	Zinc coordination sphere
Glu-270 Oε2-C4	1.61	Covalent bond to nucleophile Glu-270
Tyr-248 OH-COOH	2.59	Hydrogen bond phenolic ring
Arg-145 Nη1-COOH	2.89	Salt bridge
Arg-145 Nη2-COOH	2.78
Arg-127 Nη2-COOH	3.35	Hydrogen bond of inhibitor COOH
Asn-144 Nδ1-COOH	2.97
Leu-203 Cδ1-Cphenyl	3.85	Hydrophobic interactions of the benzyl ring (selected)
Ile-243 Cδ1-Cphenyl	3.42
Thr-268 Cγ2-Cphenyl	3.68
C3–S1, C4–C3–S1	1.81, 100.9	Geometry around the sulfur

Results and Discussion

The three-dimensional structure of the complex between CPA, a model zinc-dependent protease, and a thiirane mechanism-based inactivator was solved to 1.38 Å resolution. The CPA crystal structure presented herein is the highest resolution structure for this enzyme, with a nearly 80% data completeness in the highest resolution shell. The structure has been deposited with the PDB with accession code 3i1u. As shown in Figure 1, the existence of continuous electron density between the bound inhibitor and the side chain of the glutamate indicates that Glu-270, corresponding to the conserved active site glutamate of the zinc-dependent proteases, is covalently modified by the inhibitor. The establishment of the covalent bond is accompanied by the opening of the thiirane ring with the attendant coordination of the newly formed thiol group with the catalytic zinc ion in a monodentate manner. This arrangement gives a tetrahedral coordination sphere for the zinc ion, with three ligands from the protein and the thiol from the inhibitor. Because thiiranes are quite stable at physiologic pH, the involvement of the active site zinc ion is essential for the ring opening. The coordinated thiol (likely as a thiolate) is the remnant of the interaction of the thiirane sulfur, before the ring opening. Although inhibitor 3 was prepared as a racemic mixture, the X-ray structure reveals that it is only the 2S,3S-3 configuration that fits into the electron density map, indicating that the enzyme is enantioselective in its interaction with the inhibitor. This configuration corresponds to that of the D-amino acid series, in accordance with what has been observed for oxirane 4 (34,35).

Figure 1.

(A) The structure of the CPA-inhibitor complex. A2F_obs−F_calc electron density map showing the CPA active site with the product of the reaction of thiirane 3 with the enzyme. The electron density map (chicken wire), calculated deleting the Glu-270 side chain and inhibitor coordinates, is contoured at a 1.5σ level. The inhibitor and residues important for binding are shown in capped sticks and are labeled. The catalytic zinc ion is in magenta, while the sulfur atom from the inhibitor is in yellow. Other atoms are colored blue (nitrogen) and red (oxygen). A continuous electron density is clearly seen along the bond linking Glu-270 side chain and the inhibitor. (B) Schematic representation of the CPA-inhibitor complex.

When compared to the structure of the native CPA, some conformational changes occur upon inhibitor binding. The catalytically important Tyr-248 (37) and Arg-145 have experienced motion. Tyr-248 has been observed in two conformational states in the several structures available for CPA. One brings it to a hydrogen bonding distance of the bound peptide substrate (the ‘closed’ position) and the other is away from it (the ‘open’ position) (38,39). In the X-ray structure of the inhibited complex, the Tyr-248 phenol group moves from the surface closer to the active site cleft to make a strong hydrogen bond to the carboxylate group of the inhibitor (the bond distance is 2.59 Å; the ‘closed’ position). This carboxylate group, which corresponds to the terminal carboxylate of the peptide substrate, is held in place by hydrogen bonds to the side chain of Arg-145 and Asn-144 Nδ1. The benzyl ring of the inhibitor is buried in the S1′ pocket, thus interacting with Ile-255. This interaction is believed to impart CPA selectivity toward hydrolytic processing of the C-terminal hydrophobic amino acids. In this pocket, residues Leu-203 and Ile-243 make C_H-π interactions with the benzyl ring of the inhibitor. The presence of this benzyl ring displaces the water molecules present in the native CPA S1′ pocket. A superimposition with the related carboxypeptidae B (CPB) reveals that the benzyl moiety would clash with Asp-255 of CPB, which corresponds to Ile-255 of CPA, hence the origin of the likely selectivity for CPA. The Glu-270 side chain conformation, which is covalently tethered to the inhibitor, is approximately gauche (as measured by the Cα-Cβ-Cγ-Cδ dihedral angle of −74.4°). These observations point to the fact that little changes in key microenvironments in the binding pocket of the enzyme would allow for an exquisite recognition and binding of the ligand, despite it being configurationally dissimilar to the natural peptide substrate.

The zinc environment is perturbed in the complex. The coordination number of the zinc ion is four in the complex with the inhibitor and the resulting geometry is tetrahedral, with three protein ligand atoms at the base and the thiol sulfur of the inhibitor at the apex of a regular tetrahedron. The Zn–N distances are equal (2.10 Å and 2.11 Å for His-69 and His-196, respectively), while the Zn–S distance is 2.33 Å. The two Zn–O distances from the side chain of Glu-72, the third coordinating amino acid, differ by almost 1 Å (2.04 versus 3.00 Å). This observation shows that Glu-72 is a monodentate ligand to the zinc ion. The conformation of the glutamate carboxylate is syn as the O–C–O–Zn dihedral angle is −2.56°. In the native CPA, the Zn coordination number is five because of the bidentate coordination by Glu-72 (the Zn–O distances are 2.13 Å and 2.26 Å). A water molecule is at the apex of the distorted tetrahedron, at a distance of 2.07 Å. Again, the conformation of the glutamate carboxylate is syn (O–C–O–Zn = −1.69°). On the whole, the metal ion environment is perturbed by the inhibitor in the CPA-bound structure.

The inhibitor adopted a trans conformation in binding to CPA (the main C₁–C₂–C₃–S₁ backbone dihedral angle is −162.4°). The C₃–C₄ and C₃–S₁ bond lengths are normal, 1.49 and 1.81 Å, respectively. The distance from C₄ to Glu-270 Oε2, 1.61 Å, is longer than the C–O ester bond distances reported in the other CPA structures (34,35). The C₃–C₄–Oε2 bond angle compares favorably to that seen in CPA inhibited by 2-benzyl-3-iodopropanoic acid, a halogen-based covalent inhibitor of CPA: 114.3° versus 113.5° (40). The bond angles to heavy atoms around C₃ average 108.3°, with a minimum of 100.9° for C₄–C₃–S₁. This angle is below the ideal tetrahedral value of 109.5° indicating a distortion of the tetrahedral geometry around C₃ carbon. This carbon atom resides amidst the ester bond to Glu-270 side chain and the zinc-coordinating sulfur, S₁. Whether this distortion would have an effect over the inhibitor binding to the enzyme, and hence to its inhibitory potency, might only be assessed by analyzing compounds structurally related to the inhibitor analyzed here.

In summary, we observed in atomic detail the binding mode of a mechanism-based thiirane inhibitor for CPA. The bound inhibitor fits in the S1–S1′ subsites of CPA, a model metallopeptidase. The high-resolution complex between a zinc-dependent protease and a thiirane-based inactivator has been obtained for the first time, allowing for a complete analysis of the covalently modified protein. This structure constitutes the foundation for future development of additional reagents targeting metallocarboxypeptidases.