Amino acid sequence and homology modeling of obtustatin, a novel non-RGD-containing short disintegrin isolated from the venom of Vipera lebetina obtusa

Abstract

Disintegrins represent a group of cysteine-rich peptides occurring in Crotalidae and Viperidae snake venoms, and are potent antagonists of several integrin receptors. A novel disintegrin, obtustatin, was isolated from the venom of the Vipera lebetina obtusa viper, and represents the first potent and selective inhibitor of the binding of integrin α₁β₁ to collagen IV. The primary structure of obtustatin contains 41 amino acids and is the shortest disintegrin described to date. Obtustatin shares the pattern of cysteines of other short disintegrins. However, in contrast to known short disintegrins, the integrin-binding loop of obtustatin is two residues shorter and does not express the classical RGD sequence. Using synthetic peptides, a KTS motif was identified as the integrin-binding sequence. A three-dimensional model of obtustatin, built by homology-modeling structure calculations using different templates and alignments, strongly indicates that the novel KTS motif may reside at the tip of a flexible loop.

Crotalid and viperid venoms contain a large number of pharmacologically active proteins. Disintegrins represent a family of low-molecular-mass, cysteine-rich peptides derived by proteolysis of the PII class of mosaic snake venom metalloproteases (SVMP; Kini and Evans 1992^; Jia et al. 1996). Disintegrins have been divided into four different groups according to their polypeptide length and number of disulfide bonds (McLane et al. 1998). The first group includes short disintegrins composed of 49–51 residues and 4 disulfide bonds (Smith et al. 2002). The second group is formed by medium-size disintegrins, which contain ∼70 amino acids and 6 disulfide bonds (McLane et al. 1998). The majority of the more than 30 different disintegrins characterized to date belong to this group. The third group is formed by the long disintegrins bitistatin, an 84-residue polypeptide cross-linked by 7 disulfide bonds (Calvete et al. 1997), and salmosin 3 (Park et al. 1998). Unlike short, medium, and long disintegrins, which are single-chain molecules, the fourth group of disintegrins is composed of dimeric disintegrins (Calvete et al. 2002). Each subunit contains ∼67 residues including 10 cysteines involved in the formation of 4 intrachain disulfide bonds and 2 interchain cystine linkages (Calvete et al. 2000).

The integrin inhibitory activity of disintegrins depends on the primary structure of the inhibitory loop, whose active conformation is critically determined by the appropriate pairing of cysteine residues (McLane et al. 1996). NMR studies of several short (echistatin) and medium (kistrin, flavoridin, albolabrin) disintegrins revealed that the active tripeptide is located at the apex of a mobile loop protruding 14–17 Å from the protein core (Adler et al. 1991; Saudek et al. 1991; Senn and Klaus 1993; Smith et al. 1996). In most single-chain disintegrins, the active sequence is RGD (McLane et al. 1998), the exceptions being barbourin and ussuristatin 2, two medium disintegrins possessing KGD as the active sequence (Scarborough et al. 1991; Oshikawa and Terada 1999), and atrolysin E, which displays an MVD motif in its inhibitory loop (Hite et al. 1994). RGD-containing disintegrins show different binding affinity and selectivity toward integrins that recognize the RGD sequence in their ligands, that is, α_IIbβ₃, α_vβ₃, and α₅β₁. KGD-containing barbourin inhibits the α_IIbβ₃ integrin with a high degree of selectivity. The integrin specificity profile of atrolysin E/D is unknown, although because of its inhibition of ADP- and collagen-stimulated platelet aggregation, α_IIbβ₃ may be one target receptor (Shimokawa et al. 1998). Dimeric disintegrins exhibit more variability in their active-site sequences. EC3, a heterodimeric disintegrin from Echis carinatus venom, contains VGD and MLD sequences and is a selective and potent inhibitor of integrins α₄β₁, α₄β₇, and α₉β₁ (Marcinkiewicz et al. 1999a, 2000a). EMF10, a heterodimeric disintegrin isolated from the venom of Eristocophis macmahoni, is a potent and selective inhibitor of integrin α₅β₁, and this activity has been mapped to both the MGDW sequence (B-subunit) and the RGDN motif of the A-subunit (Marcinkiewicz et al. 1999b). Furthermore, the presence of the WGD motif in disintegrin CC8 from Cerastes cerastes venom increases its inhibitory effect toward α_IIbβ₃, α_vβ₃, and α₅β₁ integrins (Calvete et al. 2002).

Here, we describe a novel short disintegrin, obtustatin, isolated from the venom of the viper Vipera lebetina obtusa. It represents the smallest reported disintegrin and contains an integrin-recognition loop that is two residues shorter than that of other disintegrins, and whose sequence (WKTSLTSHY) strongly departs from the structure of the RGD-motif-containing loops of all known short disintegrins. Using a set of synthetic peptides representing the entire loop with single alanine mutations, the integrin recognition was narrowed down to the tripeptide KTS. This novel motif endows the disintegrin with selective α₁β₁-integrin inhibitory activity. A molecular model for obtustatin has been built.

Results and Discussion

Obtustatin was purified to homogeneity from the venom of Vipera lebetina obtusa using two steps of reverse-phase HPLC. Screening against a panel of integrins revealed that obtustatin is a very potent (IC₅₀ = 2 nM) and selective inhibitor of the binding of integrin α₁β₁ to collagen IV (C. Marcinkiewicz, P.H. Weinreb, J.J. Calvete, D.G. Kisiel, S.A. Monsa, G.P. Tuszynski, and R.R. Lobb, in prep.). MALDI-TOF mass spectrometry yielded an isotope-averaged molecular mass (M + H⁺) of 4394.2 D for native obtustatin. Incubation of obtustatin with 4-vinylpyridine under denaturing, nonreducing conditions did not change its molecular mass. However, upon reduction and pyridylethylation, the molecular mass of EP-obtustatin was 5243.3 D. These results indicated that obtustatin contained 8 cysteine residues involved in the formation of 4 disulfide bonds. The amino acid sequence of obtustatin (Fig. 1A ▶) was established using automated Edman degradation of the reduced and pyridylethylated protein and of peptides obtained after degradation of EP-obtustatin with endoproteinase Lys-C. Obtustatin contains 41 amino acids and is therefore the shortest disintegrin reported to date. The calculated isotope-averaged molecular mass of fully oxidized obtustatin (4394.2 D) accurately matches the experimentally determined mass. The primary structure of obtustatin displays many of the strongly conserved features of known disintegrins, including the pattern of cysteines of short disintegrins and several highly conserved residues distributed throughout the whole amino acid sequence (Fig. 1A ▶). Noteworthy, the highest divergence from other disintegrin sequences involves the integrin-binding loop encompassing residues 20–28. In all reported short disintegrin sequences, the integrin-binding loop contains 11 residues, including an RGD motif at the 4th–6th position (Smith et al. 2002 and references therein). The length of this loop is also absolutely conserved in long, medium, and dimeric disintegrins, and despite sequence variance, all these molecules express a conserved aspartate at position 6 of their integrin-binding loops. The recent crystal structure of the extracellular segment of integrin α_vβ₃ in complex with an RGD ligand (Xiong et al. 2002), showing that the ligand Asp carboxylate oxygens protrude into a βA-domain cleft and participate in an extensive network of polar interactions, provided a basis for understanding the involvement of the conserved aspartate in the interaction of integrins with larger RGD-containing ligands. Similarly, the crystal structure of the complex between the integrin α₂β₁ I-domain and a synthetic collagen triple-helical peptide (Emsley et al. 2000) showed the key role of a glutamate residue for ligand interaction. However, neither the length of the integrin-binding loop nor the presence of an acidic residue is conserved in obtustatin (Fig. 1A ▶). The structural requirements of obtustatin’s α₁β₁ inhibitory activity was investigated by means of synthetic peptides covering the entire integrin-binding loop of the disintegrin (¹⁹CWKTSLTSHYC²⁹) and using a set of peptides containing a single alanine mutation at every amino acid position (Table 1). Substitutions of residues ²¹KTS²³ reduced the inhibitory activity of the corresponding peptide on the adhesion of K562 cells transfected with the α₁β₁ integrin to collagen IV. Threonine 22 appears to represent the most critical residue for blocking the α₁β₁-integrin–collagen IV interaction.

Figure 1.

(A) Sequence alignment of obtustatin and the template proteins (echistatin, kistrin, and flavoridin) used for homology-modeling calculations. A bar chart shows the degree of sequence similarity among the four proteins. The numbers at the right of the sequences indicate the number of amino acids of each protein. Stars indicate conserved residues; a dot or a colon indicates positions with partially conserved residues. (B) Front and rear views (top and bottom, respectively) of the superpositions of the best energy models obtained using kistrin (red), flavoridin (green), and echistatin (blue) as templates. Wider ribbon sections indicate larger dispersion in the bundle of calculated structures. A lighter color has been chosen for the flexible loop containing the integrin recognition site. (C) Ribbon diagram of the best energy homology model of all the calculated structures. The ²¹KTS²³ sequence is displayed in the red and wider ribbon. The disulfide-bonded cysteines are numbered and in yellow.

Table 1.

Inhibition activity on the adhesion of α1-transfected K562 to immobilized collagen IV of synthetic peptides representing the native sequence and single-alanine-mutated sequences of the integrin-binding loop of obtustatin

Sequence of peptide	IC50 (μM)
CWKTSLTSHYC	586
CAKTSLTSHYC	612
CWATSLTSHYC	1185
CWKASLTSHYC	>2000
CWKTALTSHYC	916
CWKTSATSHYC	698
CWKTSLASHYC	603
CWKTSLTAHYC	632
CWKTSLTSAYC	658
CWKTSLTSTAC	669

Residues substituted for alanine are in bold and underlined. The data present the mean of three different experiments.

To gain insight into the molecular architecture of obtustatin, we have worked out a molecular model for this novel disintegrin. Figure 1A ▶ shows the sequence alignment between obtustatin and the different templates used for modeling. Conserved residues are mainly those involved in disulfide bonds. The set of models calculated with the disulfide pattern of echistatin gave better energy values and structural quality than those calculated with the alternative disulfide pattern II. However, the differences in energy were not large enough to determine uniquely the disulfide pattern of obtustatin from the results of the homology-modeling calculations. Nevertheless, in general, lower values of van der Waals energy of the models calculated with disulfide pattern I (echistatin) were associated with a better agreement between the rest of distance restraints. Moreover, disulfide pattern II has not been experimentally demonstrated in any of the known disintegrins. All the models selected fulfilled the criteria implemented in the program PROCHECK 3.5 (Laskowski et al. 1993), including Ramachandran plot, planarity of peptide bonds, tetrahedral distortions on C_α atoms and nonbonded interactions, as required for medium quality structures. Thus, all model residues reside in the allowed areas of the Ramachandran plot, although only ∼70% reside in the most favored areas. This outcome reflects the little or none regular secondary structure of the structure template used in the calculation of the obtustatin models.

Figure 1B ▶ displays views of the superposition of the best energy-refined structures of obtustatin obtained by homology-modeling calculations using the different templates. The homology model with the best values of energy and best conformational quality as calculated with PROCHECK 3.5 is shown in Figure 1C ▶. As expected from the high mobility reported in NMR studies of disintegrins, the largest conformational differences between obtustatin models obtained using different templates are linked to the loop formed by residues 19–29, whereas the part of the molecule containing the 4 disulfide bonds displayed the largest structural conservation regardless of the template used for the calculation. As for other disintegrins, the structure of obtustatin can be described as a bundle of loops with no regular structure protruding from a rigid body cross-linked by disulfide bonds. However, in contrast to other disintegrins, whose integrin-binding loop joins two short antiparallel β-strands, the mobile integrin-binding loop of obtustatin is two residues shorter, and no secondary structure could be modeled. Obtustatin residues W20 and Y28 appear to act as hinges for the overall movement of the loop. Initial NMR data of obtustatin showing a double number of signals in the aromatic region, indicating two conformations for W20 and Y28 (M. Paz Moreno-Murciano, D. Monleón, J.J. Calvete, B. Celda, and C. Marcinkiewicz, in prep.), support this hypothesis.

The tip of the flexible integrin-binding loop harbors the amino acid sequence KTS, which represents a novel integrin-binding motif, further suggesting a similar functional role as other known integrin inhibitory sequences. Moreover, in all the calculated models, the C-terminal tail of obtustatin lays in the vicinity of the integrin-binding loop. This feature has been observed in the NMR structures of other disintegrins and appears to be of functional significance in echistatin for recognition of α_IIbβ₃ and α_vβ₃ integrins and expression of ligand-induced binding sites (Marcinkiewicz et al. 1997). However, the question of whether obtustatin uses an equivalent binding site on integrin α₁β₁ as the XXD-containing disintegrins on their integrin targets deserves further investigation.

Materials and methods

Isolation and primary structure determination of obtustatin

Obtustatin was purified from the venom of Vipera lebetina obtusa (Latoxan) using two steps of reverse-phase HPLC as described (Marcinkiewicz et al. 1999a,b, 2000a). The yield was 12 mg of purified obtustatin per gram of crude venom. The purity was assessed by SDS-PAGE and MALDI-TOF mass spectrometry (using an Applied Biosystems DE-Pro spectrometer and α-hydroxycinammic acid saturated in 0.5% TFA in 40% acetonitrile as matrix). For structure determination, purified obtustatin was reduced and alkylated with 4-vinylpyridine as described (Marcinkiewicz et al. 2000a). S-Pyridylethylated (PE) obtustatin was subjected to N-terminal sequencing (using an Applied Biosystems 477A instrument) and mass spectrometry (as above). The primary structure of PE-obtustatin was established from N-terminal sequence analyses of the PE-protein and of overlapping peptides obtained by proteolytic digestion. To this end, 2 mg/mL obtustatin in 100 mM ammonium bicarbonate, pH 8.3, was degraded with endoproteinase Lys-C (Boehringer Mannheim) at 37°C for 18 h using an enzyme: substrate ratio of 1:100 (w/w). Peptides were separated on a 4 × 250-mm C18 (5 μm particle size) Lichrospher RP100 (Merck) column eluting at 1 mL/min with a linear gradient of 0.1% TFA in water (solution A) and acetonitrile (solution B). The amino acid sequence of obtustatin is accessible in the SWISS-PROT database under accession number P83489.

Peptide synthesis and inhibition assays

Peptides were synthesized on a Symphony multiple-peptide synthesizer using Fmoc [N-(9-fluorenyl)methoxycarnonyl] chemistry as described previously (Calvete et al. 2002). Adhesion studies of cultured K562 cells transfected with α₁β₁ integrin and labeled with 5-chloromethyl fluorescein diacetate (CMFDA) to collagen IV were performed as described (Marcinkiewicz et al. 2000b). For inhibition studies, CMFDA-labeled cells (1 × 10⁵/sample) were incubated at 37°C for 30 min with increasing amounts of the peptide before adding them to the wells of a 96-well microtiter plate (Falcon) coated with collagen IV (Chemicon Inc.). After 30 min at 37°C, unbound cells were removed by aspiration, the wells were washed, and bound cells were lysed by the addition of 0.5% Triton X-100. A standard curve was prepared using known concentrations of labeled cells. Plates were read using a Cytofluor 2350 fluorescence plate reader (Millipore) with a 485-nm excitation filter and a 530-nm emission filter. All determinations were carried out in triplicate.

Molecular modeling

An homology model for obtustatin was built using translation of distance restraints from the template structure to the target protein (Sali and Blundell 1993; Li et al. 1997) with minor modifications. The structures of three different disintegrins, flavoridin, kistrin, and echistatin (PDB IDds 1FVL, 1KST, and 2ECH, respectively), were used as templates for the homology-modeling calculations. Only distances between homologous atoms, as defined by the atom type and the sequence alignment, were used for distance restraints translation (Li et al. 1997; Donaire et al. 1998). To generate a preliminary set of models, instead of using a full potential-energy function method for structure calculations, we have used a much faster algorithm based on dihedral angle molecular dynamics. This set of models was refined by simulated annealing of restrained molecular dynamics with CONGEN (Bruccoleri 1993). A density of ∼20 meaningful constraints per residue was chosen to be comparable to the number of NOE distance constraints of high-resolution NMR structures. However, the restraints derived for homologous atoms are more restrictive than the NOE-based distance restraints because the homologous atoms are all heavy atoms. Hence, sets of more convergent models are expected. A cutoff of 15 Å was used for distance restraint selection. Distances were chosen randomly from all the distances shorter than 15 Å between homologous atoms in the template. From each distance, upper and lower bounds were created by adding or subtracting 10% of the exact distance. The final sets of upper and lower bounds were introduced in the molecular dynamics program DYANA for simulated annealing calculations (Güntert et al. 1997). Two different patterns of disulfide bonds were taken into consideration in the calculations because of the close spatial proximity of cysteines 6, 7, 29, and 34 in all the template structures. Disulfide pattern I involves disulfide connections between cysteines 1 and 10, 6 and 29, 7 and 34, and 19 and 36, and represents the pattern present in echistatin (Calvete et al. 1992; Bauer et al. 1993). The alternative disulfide pattern II involves the cysteine pairings 1 and 10, 6 and 34, 7 and 29, and 19 and 36. Disulfide bonds were introduced in DYANA as “ssbond-type” restraints. The standard DYANA simulated annealing protocol starting from random coordinates was used for structure calculation. A total of 100 structures were calculated for each homology model, and the 20 with the lowest target function were selected for CONGEN simulated annealing refinement. Customized CONGEN protocols for simulated annealing, including stages of weight annealing and temperature annealing, were used for structure refinement (Tejero et al 1996).