Crystal structure of human cyclophilin D in complex with its inhibitor, cyclosporin A at 0.96‐Å resolution

Abstract

Cyclophilins belong to a ubiquitous protein family that catalyses the reaction of peptidyl-prolyl cis-trans isomerase (PPIase) on their ligand proteins, and they play a crucial role in protein folding.1-3 Proteins in the cyclophilin family are found in a variety of organisms from prokaryotes to eukaryotes, including humans.4 Five isoforms are known among human cyclophlins.2,5,6 One of these isoforms, cyclophilin D (CypD), is a mitochondria-specific cyclophilin.2 CypD is known to play a pivotal role in mitochondrial permeability transition (MPT), which leads to the loss of mitochondrial membrane potential (ΔΨ), mitochondrial swelling, rupture of the outer membrane, and finally necrotic cell death.7 CypD knock-out mice do not have any detectable phenotypic anomalies, but they resist ischemia/reperfusion-induced injury in the heart8,9 and brain.10 Although the exact target protein molecule for CypD PPIase activity is still unknown, these findings reported to date have suggested CypD as a potential drug target to treat cerebral and myocardial infarctions. An immunosuppressive compound, cyclosporin A (CsA), is known to bind to the proteins of this family and inhibits PPIase activity of cyclophilins and MPT.3, 11 Several crystal structures of cyclophilin including that of CypD have been determined,12-14 but thus far, no structures of CypD complexed with ligands have been determined. It is essential to obtain structural information about CypD-ligand interactions for the development of a CypD-specific inhibitor as an anti-infarction drug. Accordingly, we have determined the ultrahigh resolution crystal structure of CypD in complex with CsA at 0.96-Å resolution. The human CypD expression system starting from Gly2 was constructed as described by Schlatter et al.14 The calculated molecular weight of this CypD recombinant is 17.8 kDa. The gene was cloned into a pET-21a vector (NOVAGEN) using the NdeI and BamHI restriction sites, and a mutation of Lys133 into Ile was introduced on the CypD gene. Expression was conducted in Escherichia coli BL21(DE3) cells. Cell culturing and the induction of CypD expression were also performed as previously described.14 Purification was performed with Hi-trap SP, RESOURCE Q, and Superdex 75 chromatography instead of with the EMD COO−, Q-sepharose, and EMD BioSEC columns used by Schlatter et al.14 The supernatant of the cell homogenate was applied to a Hi-trap SP column equilibrated with buffer A (100 mM Tris-HCl at pH 7.8, 2 mM EDTA, and 2 mM DTT). Bound CypD proteins were eluted with a gradient of0–0.5 M NaCl. The CypD fractions were desalted with dialysis against buffer A and applied onto a ResourceQ column equilibrated with buffer A. CypD did not bind to the column, and the flowthrough fraction was further purified with a Superdex75 column equilibrated with buffer B (50 mM potassium phosphate at pH 7.3, 100 mM NaCl, 2 mM EDTA, and 2 mM DTT). Thirteen milligrams of CypD were harvested from 5 g of E. coli cells. For the crystallization, 25 mg/mL of purified CypD in buffer B was used. CsA, isolated from the fungus Tolypocladium inflatum, was purchased from CALBIOCHEM. Two milligrams of CsA powder were incubated with 100 μL of CypD solution for 5 days in order to obtain the CypD and CsA complex. The nondissolved CsA was removed with a 0.22-μm filter. Rod-shaped crystals of the complex were obtained within a day using 20% polyethylene glycol 3350, 45 mM sodium-citrate buffer at pH 2.9 as a precipitant solution. The crystals were grown to an approximate maximum rod-length of 1.0 mm. The crystals were dipped in a cryoprotectant buffer (35% polyethylene glycol 3350, sodium-citrate buffer at pH 2.9) for ∼10 s and were flash-frozen in a nitrogen gas stream at 100 K. Two sets of diffraction data were collected at the beamline NW12A at Photon Factory, Ibaraki, and at the beamline BL41XU at SPring-8, Hyogo, Japan. The initial dataset was collected with a wavelength of 1.000 Å in Photon Factory. The crystal was exposed for 1.0 s for 1.0° oscillation with a 0.5-mm aluminum attenuator for the low-resolution dataset, and for 3.0 s for 0.5° oscillation without the aluminum attenuator for the high-resolution dataset. The images were integrated with HKL200015 from 50 to 2.2 Å and from 2.5- to 1.06-Å resolution for the low- and high-resolution datasets, respectively. The merged data were phased with the molecular replacement method using Molrep.16,17 The structure of human apo-CypD (PDB ID: 2BIT) was employed as a search model. Model-building and refinement were carried out using COOT18 and REFMAC5,19 respectively. The second dataset was collected at SPring-8 with a 0.1-s exposure for 1.0° oscillation with a 0.3-mm aluminum attenuator using a wavelength of 1.000 Å X-ray for the low-resolution dataset, and with a 0.4-s exposure for 0.5° oscillation without the aluminum attenuator, using a wavelength of 0.85 Å X-ray for the high-resolution dataset. Integration was performed from 200 to 2.2 Å and from 2.5 to 0.96 Å for the low- and high-resolution datasets, respectively. The integrated intensities were scaled and merged into one dataset. Model building was carried out based on the partially refined structure using the initial dataset at Photon Factory. The structure was finally refined to crystallographic R- and Rfree factors of 12.8% and 15.3% by employing anisotropic B factors, respectively. The coordinates and structural factors were deposited with the protein data bank under ID 2Z6W. A K133I mutant of human CypD, instead of the wild-type enzyme, was employed to obtain the present crystal of the CypD-CsA complex, because it was reported that this mutant showed the reasonable solubility in water and the similar activity with the wild-type enzyme, and afforded crystals suitable for the X-ray work.14 CsA shows low solubility for water. CsA is usually soluble in organic solvents such as DMSO and ethanol, which are unsuitable for protein crystallization. We therefore prepared the CypD-CsA complex without the use of organic solvents, and instead utilized the high solubility of CypD in water and its strong affinity to CsA. The addition of a small amount of CsA in ethanol to the aqueous CypD solution resulted in the immediate formation of white precipitants. This precipitation might have been CsA, the solubility of which is very low in water. However, the precipitation disappeared after 3 days of incubation. This result implied that CsA was soluble in the aqueous CypD solution, and that the CypD-CsA complex could be formed by the incubation of a CypD solution with CsA powders. In order to drive all CypD molecules in the solution to form a complex with CsA, excess CsA was added to the aqueous CypD solution and the mixture was incubated for 5 days. As described in the following paragraph, the crystal structure indicated that the CypD-CsA complex was successfully formed according to this protocol, which can be applied for the crystallization of protein–ligand complexes with various water-insoluble ligands. The structure of the CypD-CsA complex was refined at 0.96-Å resolution; the present resolution is the highest among those of known Cyp crystal structures. As the highest resolution of previously known Cyp-complex structures was 1.5 Å,20 the present structure provides additional details regarding the mode of binding between Cyp and CsA. The present crystals belong to the orthorhombic space group P212121 with the unit cell dimensions of a = 40.71 Å, b = 72.97 Å, and c = 112.0 Å. The present crystal form differs from that of the apo-CypD crystals (P41212).14 Two CypD-CsA complexes are found in an asymmetric unit of the present crystal; the Matthews coefficient, Vm, and the solvent content are calculated to be 2.0 Å3/Da and 38.2%, respectively. Multiconformations are assigned for 42 amino acid residues in the structural model, among a total of 328 amino acid residues of the two CypD molecules in an asymmetric unit. The data collection and refinement statistics are summarized in Table I. The structure of human CypD is composed of eightβ-strands, two α-helices, and one 310 helix [Fig. 1(A)], and this structure is similar to that of the other known cyclophilin structures. The RMSDs between the Cα atoms of the present structure and those of CypA-CsA, CypB-CsA, CypC-CsA, and CypE(apo) are 0.50 Å (for 164 atoms), 0.67 Å (159 atoms), 0.68 Å (158 atoms), and 0.68 Å (161 atoms), respectively. (A) Overall structure of CypD in complex with CsA. CypD and CsA are shown as ribbon- and stick-models, respectively. Panels A, B, and D were drawn with PyMOL (http://www.pymol.org). (B) Close-up view of CsA superimposed on its Fo-Fc omit electron density map (blue mesh) contoured at 6.0 σ. (C) Binding geometry of CsA on CypD. Green circles mark the CsA atoms involved in the hydrophobic contact with CypD. The CypD residues in green ellipsoids are involved in the hydrophobic interactions with CsA. The red dotted lines represent hydrogen bonding. This panel was prepared based on a scheme drawn with LIGPLOT.21 Abbreviations of CsA residues: Bmt, (4R)-4[(E)-2-butenyl]-4,N-dimethyl-L-threonine; Aba: L-α-aminobutyric acid, Sar: sarcosine, Mle, N-methyl leucine; Dal, D-alanine; Mva, N-methyl valine. (D) Distribution of the conserved residues among human cyclophilins represented on the surface of the present CypD-CsA structure. Residues conserved in all five known human cyclophilins (CypA, CypB, CypC, CypD, and CypE) are shown in red. CypD residues conserved in 3 of 4, 2 of 4, and 1 of 4 other cyclophilins are shown in orange, yellow, and green, respectively, and the unconserved CypD residues in the other four cyclophilins are shown in blue. CsA is shown as a stick model. The Fo-Fc map phased only with the CypD structure showed additional clear electron densities corresponding to the CsA molecule, confirming that the crystal was composed of the CypD-CsA complex [Fig. 1(B)]. The averaged B-factors of CypD and CsA are 6.7 Å2 (2812 nonhydrogen atoms) and 6.5 Å2 (179 nonhydrogen atoms), respectively. This suggests that the CsA molecule is rigidly bound to the CypD molecule to form the stable CypD-CsA complex. This finding is consistent with the very strong affinity of CsA to CypD (KD= 13.4 nM).14 The CypD structure in the present complex is very similar to that of apo-CypD (RMSD of Cα = 0.48 Å for 164 atoms), suggesting that CypD does not need to undergo a conformational change in order to bind with CsA, in spite of the fact that CsA (1.2 kDa) has one-fifteenth of the molecular weight of CypD (17.8 kDa). The mutated Ile133 was involved in the important intermolecular contacts in the apocrystal.14 In the present crystal composed of two CypD-CsA complexes in an asymmetric unit, one of the two CypD complexes makes no contact with the other complex via Ile133, but Ile133 on the other CypD molecule is directly involved in the intercomplex interaction. Therefore, wild-type CypD is not considered to form the present crystal packing arrangement found in the case of apo-CypD. The binding mode of CsA with CypD is shown on Figure 1(C). One-half of CsA residues (Sar3-Dal8) is exposed to the solvent region and the other half (Bmt1, Aba2, and Mle9-Mva11) is buried in the CypD molecule. Since CsA is a cyclic peptide composed of hydrophobic residues, the molecule interacts with CypD primarily via hydrophobic contacts. Hydrogen bonds are formed only using main-chain N and O atoms of CsA [Fig. 1(C)]. The amino-acid sequence identities of CypD with CypA, CypB, CypC, and CypE are 75%, 61%, 58%, and 67%, respectively. These highly conserved residues are distributed across the molecular surface, as shown in Figure 1(D). All residues in contact with CsA are completely conserved in all human cyclophilins. The PPIase active site is known to be located at the bottom of the CsA binding site. Residues Ser59, Ser81, Arg82, Ile117, Lys148, and Ser149 of CypD are located around the CsA binding site, and they are not well conserved in the other cyclophilins (CypA, CypB, CypC, and CypE). These residues are thought to specifically recognize the ligand proteins. Molecules that bind to these residues could be a leading compound as CypD-specific inhibitors, and they may therefore serve as potential targets for the development of anti-infarction drugs. We appreciate the help of Drs. M. Kawamoto and N. Shimizu of SPring-8 and Drs. N. Igarashi, N. Matsugaki, and Y. Yamada of Photon Factory with the X-ray diffraction experiments.