Crystal Structure of the Cysteine-rich Secretory Protein Stecrisp Reveals That the Cysteine-rich Domain Has a K+ Channel Inhibitor-like Fold*

Stecrisp from Trimeresurus stejnegeri snake venom belongs to a family of cysteine-rich secretory proteins (CRISP) that have various functions related to sperm-egg fusion, innate host defense, and the blockage of ion channels. Here we present the crystal structure of stecrisp refined to 1.6-Å resolution. It shows that stecrisp contains three regions, namely a PR-1 (pathogenesis-related proteins of group1) domain, a hinge, and a cysteine-rich domain (CRD). A conformation of solvent-exposed and -conserved residues (His60, Glu75, Glu96, and His115) in the PR-1 domain similar to that of their counterparts in homologous structures suggests they may share some molecular mechanism. Three flexible loops of hypervariable sequence surrounding the possible substrate binding site in the PR-1 domain show an evident difference in homologous structures, implying that a great diversity of species- and substrate-specific interactions may be involved in recognition and catalysis. The hinge is fixed by two crossed disulfide bonds formed by four of ten characteristic cysteines in the carboxyl-terminal region and is important for stabilizing the N-terminal PR-1 domain. Spatially separated from the PR-1 domain, CRD possesses a similar fold with two K+ channel inhibitors (Bgk and Shk). Several candidates for the possible functional sites of ion channel blocking are located in a solvent-exposed loop in the CRD. The structure of stecrisp will provide a prototypic architecture for a structural and functional exploration of the diverse members of the CRISP family.

The cysteine-rich secretory protein (CRISP) 1 family is a large group of secreted proteins that function in many vertebrates and is characterized by 10 of 16 conserved cysteine residues residing in the carboxyl-terminal portion. This evolutionarily highly conserved family was originally described in the rodent male reproductive tract (1)(2)(3) and, more recently, in all mammals studied such as mouse, rat, equine, and human. Most mammalian CRISP members have been identified in three classes, CRISP-1, CRISP-2, and CRISP-3. CRISP-1 (DE/AEG, ARP) is secreted in an androgen-dependent manner by the proximal epididymis, associates with the sperm surface during maturation, and is latterly involved in gamete fusion (4). It is localized on the dorsal region of the sperm head and migrates to the equatorial segment during capacitation (5). CRISP-1 has been demonstrated to bind to the egg surface (6), and the presence of an excess of exogenous CRISP-1 can actually block sperm-egg fusion, possibly by inhibiting capacitation (7). CRISP-2 (TPX-1) is expressed and secreted from spermatocytes and mediates enhanced binding of spermatocytes to Sertoli cells (8). The distribution of CRISP-3 is wider than that of the other two classes and includes saliva, the pancreas, murine pre-B cells, human neutrophils, the thymus, the colon, and the ovary, indicating a possible role of CRISP-3 in innate host defense (9). It has been suggested as a potential biomarker for prostate cancer (10). Several CRISP members expressed in the embryos of Xenopus laevis and chicken have been reported (11,12).
CRISPs are also found in the snake venom of many species and comprise a new group of snake venom proteins (13). Some of these proteins have been shown to alter a variety of ion channels. PsTx and pseudecin act by blocking olfactory and retinal cyclic nucleotide-gated channel currents (14). Ablomin, latisemin, ophanin, triflin, and piscivorin can inhibit the contraction of smooth muscle induced by a potassium ion with a high concentration (13,15). Helothermine, from lizard venom, has been shown to block voltage-gated calcium channels, potassium channels, and ryanodine receptors (16,17).
Distantly related proteins such as the plant pathogenesisrelated group 1 (PR-1) proteins, insect allergens, and human Golgi-associated plant pathogenesis-related protein GAPR-1 exhibit sequence homologies to the N-terminal domain of CRISPs but do not have the carboxyl-terminal cysteine-rich region (23)(24)(25) (Fig. 1). Such proteins are grouped into a PR-1 superfamily. Structures of these proteins (P14a, Ves v 5, and GAPR-1) have been reported, revealing a unique ␣-␤-␣ sandwich fold of the PR-1 domain (26 -28). A cleft with two con-served histidines and two glutamates was noticed and suggested as a possible active site (29). Recently, one cone snail PR-1 protein, Tex31, was found with an endopeptidase activity (30), implying further the possibility of the PR-1 domain as an enzymatic module.
To better understand the biological function of the CRISPs, we have determined the first structure of the CRISP family, stecrisp, from Trimeresurus stejnegeri snake venom by x-ray crystallography. It revealed a separate chimera structure of CRISP. The carboxyl-terminal cysteine-rich region folds into a hinge and a cysteine-rich domain (CRD). Several loops around the well ordered ␣-␤-␣ sandwich core in the PR-1 domain show a great structural difference with those in P14a, GAPR-1, and Ves v 5. CRD is linked by a hinge to the N-terminal domain and possesses a similar fold with two K ϩ channel-blocking toxins.

MATERIALS AND METHODS
Crystallization and Data Collection-Stecrisp was purified from T. stejnegeri venom, and a native data set was collected and processed as described previously (31). For a derivative, crystals with a typical size of 0.3 ϫ 0.05 ϫ 0.05 mm 3 were transferred to 2 l of former reservoir solution (100 mM Tris-HCl, pH 7.8, 44% polyethylene glycol 4000, and 20 mM sodium acetate) to equilibrate with 60% polyethylene glycol 4000. Three days later, a new reservoir solution that contained 100 mM Tris-HCl, pH 7.8, 60% polyethylene glycol 4000, 6 mM UO 2 acetate, and 20 mM sodium acetate was added in to soak for 20 days. This derivative data set was collected at the beamline 3W1A of the Beijing Synchrotron Radiation Facility at 100 K and processed using Automar 1.3 (Mar Research GmbH).
Phasing and Model Refinement-Because an initial attempt to solve the structure by molecular replacement using the known models of PR-1 domains did not give an apparent solution, the structure of stecrisp was determined by SIRAS (single isomorphous replacement with anomalous signal) using a uranium derivative. The heavy atom parameters were first determined by SHELXD (32) after extracting Bijvoet differences by XPREP (Bruker Analytical X-ray Systems) and then refined, and the phases were calculated using the program SHARP (33). Seven uranium sites were located on anomalous Patterson maps or difference Fourier maps. The resulted phases were extended to 1.6 Å and improved with solvent flattening and histogram matching using DM and SOLOMON (34). A model with 210 of 221 amino acids was auto-built using ARP/wARP (35) and refined at a 1.6-Å resolution using CNS and O by manual model correction (36,37). Diagrams were computed using PyMOL (www.pymol.org) and GRASP (38). The crystal belongs to the orthorhombic space group I222 with one molecule per asymmetric unit. The final refined model, with an R-factor of 19.2% and an R free of 21.9%, contains all 221 residues, and the atomic coordinates have been deposited in the Protein Data Bank with accession code 1RC9 (see Table I).
Size Exclusion Chromatography-Samples of the purified stecrisp were run on a size exclusion Superdex 75 column (16/800; Amersham Biosciences) with a fast protein liquid chromatography system (Waters 650E). The column was equilibrated with running buffer (50 mM Tris-HCl, pH 8, and 150 mM NaCl) and calibrated using a series of molecular weight marker proteins for gel filtration chromatography according to the LMW calibration kit (Amersham Biosciences).
Proteolysis Activity Assay-Proteolysis activity assays were performed according to those of Tex31. A typical reaction mixture consisted of 100 mM buffer, 100 M pNA substrate, and 1 mM cation (or without cation) with a final volume of 500 l. 5 l of 1 mg/ml stecrisp was added to the premixed reaction mixture, and the time course absorbance was measured at 405 nm on a UV-visible spectrophotometer (600 s per reaction; Jasco V-550) at both 20 and 37°C in matrix mode combinations of these variables. The buffers, which had been tried for identification of possible proteolytic activity, included sodium acetate-acetic acid (pH 4.5), sodium citrate-HCl (pH5.5), MES-NaOH (pH 6.5), bis-Tris (pH 7. . Proteolytic assays toward Ac-KLEKRG-NH 2 (synthesized) were performed in a mixture of 100 mM NH 4 HCO 3 , 1 mg/ml peptide, and 0.1 mg/ml stecrisp with or without the cations mentioned above at 37°C overnight. The reaction mixtures were analyzed by an analytical reversed phase high performance liquid chromatography C18 column (16 ϫ 150 mm; Waters). Proteolytic assays toward protein ␤-casein (C6905; SIGMA; further purified by reversed phase high performance liquid chromatography C4 column) were performed in a similar manner, except that the reaction products were analyzed by SDS-PAGE.

RESULTS
Overall Structure-The structure of stecrisp comprises two well defined domains named the PR-1 domain and cysteinerich domain. The CRD consists of the last 39 amino acids and is connected with the PR-1 domain by a well folded hinge region (Fig. 2, A and B). The hinge clings to the side of the PR-1 domain opposite to the possible active site. CRD is positioned on the flank of PR-1 domain. The 16 strictly conserved cysteines of CRISP family form eight paired disulfide-bridges with three in the PR-1 domain, three in the CRD, and two in the hinge (Fig. 1).  2C). Most of the hydrogen bonds and hydrophobic interactions conserved among these proteins also exist in stecrisp. These interactions are crucial for the stable space arrangement of secondary structure elements and the formation of a bipartite hydrophobic core associated with the twolayer interfaces of the ␣-␤-␣ sandwich fold. An extended Nterminal helix and six segments (30 -38, 55-58, 62-72, 105-

PR-1 Domain of CRISP Has Special Features-The
Values in parentheses show the last resolution shell in Ångstroms. c For the single wavelength anomalous diffraction (SAD) data set, Friedel pairs were counted as independent.
d Calculated from 10% of the reflections selected in thin shells and omitted from refinement. e Wilson plot B-factor in parentheses.
113, and 153-159) show differences with P14a, GAPR-1, and Ves v 5. The outspread loop 30 -38 between ␣2 and ␤1 provides a scaffold for interaction with the hinge region and is attributed to an insertion of cis-bonded Pro 30 as a Pro kink, which is strictly conserved in the carboxyl-terminal motif-containing members, including CRISPs and GliPR-like proteins (Fig. 5).
The segment 62-72 is a longer insertion of ten amino acids (compared with p14a) and consists of one short ␣-helix (␣4) and two antiparallel ␤-strands, ␤2 and ␤3. This segment is stabilized by a CRISP-unique disulfide between Cys 73 and Cys 148 at the end of ␤8 ( Although all PR-1 domains described to date possess a number of conserved amino acid residues (His 60 , Glu 75 , Glu 96 , and His 115 in stecrisp), the exact roles that they may play are still unknown. The corresponding residues are present in the CRISP family, except that His 60 and Glu 75 are not strictly conserved. A similar Gln substitutes for Glu 75 in some CRISPs, and His 60 is changed to Glu in CRISP-1s of human and equine. In stecrisp, these residues are also in close spatial proximity and are located in the largest surface cleft, a candidate for the substrate binding site, surrounded by segments 62-72, 105-113, and 153-159, with ␤4 as the floor (Fig. 2D). In the crystal structure of stecrisp, this cleft is free from crystal packing and is entirely exposed to the solvent tunnel. A stable water molecule (B-factor of 16.36 Å 2 ) is hydrogen-bonded with imidazole rings of both histidines. Because of this hydrogen bonding linkage, the side chain of His 60 has a defined electron density map rather than multiple conformations as observed in GAPR-1. Superposition of the four residues (we name them as a tetrad) shows they are presented with quite similar conformations among the solved crystal structures (stecrisp, Ves v 5, and one of the dual conformations in GAPR-1; Fig. 3C), although their orientations vary in the solvent models of P14a. Relative to the anti conformation of His 60 and Glu 75 , His 115 and Glu 96 adopt a syn hydrogen bond, which is more familiar in triads of proteases than the anti dyad. The dipole of core helix a5 may help to enhance the interaction between Glu 96 and His 115 . The conformational similarity of these solvent-exposed residues is clearly due to the conservation of vicinal residues that constrain the tetrad by hydrogen bonds (Trp 52 and Ser 61 ), stacking interaction (Trp 93 ), and the dipole of the core helix (␣6; Fig. 3A).
Negative Results of the Proteolytic Assays-Because Tex31, composed of a homologous PR-1 domain (30% identity with that of stecrisp) and a different carboxyl-terminal tail with CRISPs, has been proved to be an endopeptidase, the presence of other substrate-specific proteases in PR-1 superfamily was suggested (30). An initial attempt to identify the proteolytic ability of stecrisp toward peptide Ac-KLEKRG-NH 2 (the substrate of Tex31; 37°C overnight) showed no apparent cleavage, whereupon we performed a series of more comprehensive assays by using a protein substrate (␤-casein) and chromogenic peptides that encompassed different properties of P1 site (details are shown under "Materials and Methods"). No proteolytic activity of stecrisp could be identified in these assays.
Stecrisp Is Present as a Monomer in Solution-Recently, it has been shown that a small fraction of a PR-1 protein, GAPR-1, is present in dimer form in solution and that mutations of the four conserved residues make a significantly increased dimer population. Based on the crystal packing interactions, a likely dimerization surface is predicted. A serine residue (Ser 71 ), which is found to be close to two of the tetrad residues (His 65 and Glu 54 ), is predicted to be a nucleophile. These three residues across the dimer interface may form a catalytic triad like that of trypsin (28). Similar experiments were performed using gel filtration to detect whether stecrisp might exist as a dimer in solution. However, we found that stecrisp was present as a monomer without any detectable dimer population that could be identified from the chromatograms (Fig. 4A). Another CRISP protein we purified also gave a similar result. 2 After superimposing stecrisp to the dimer model of GAPR-1, we found that additional carboxyl-terminal regions were oriented in opposite directions and would not occlude the possible substrate binding site in the superimposed model. Ser 71 in GAPR-1 overlaps with Pro 81 in stecrisp (Fig. 4B). These residues share a similar position at the tip of a loop between ␤4 and ␣5. A nearby Ser 80 is a candidate to correspond with Ser 71 in GAPR-1. The location of Ser 80 in stecrisp is slightly inner at the end of ␤4, making it relatively far from the His 60 -Glu 75 dyad of the other dimer-related molecule. Even after rotating the hydroxyl of Ser 80 toward the imidazole ring of His 60 , the nearest distance between the O of Ser 80 and the N⑀2 of His 60 is 6.3 Å. Although most CRISPs have serine or threonine residues that can be aligned with Pro 81 in stecrisp and may be close to the His-Glu dyad in the dimer model, we find that there are also some exceptions (Fig. 4C). Remarkably, despite containing the conserved tetrad mentioned above, CRISP1 in rat (AEG_RAT; Swiss-Prot accession number P12020) does not have any possible nucleophile residue in this region. On the other hand, a glutamate substitutes for His 60 in CRISP-1s in human and equine (Swiss-Prot accession numbers P54107 and Q8MHX2); however, both of these proteins have serine residues corresponding to Pro 81 in stecrisp. Without the critical histidine, the proposed triad will not also form in them. Interestingly, CRISP1 in mouse (Swiss-Prot accession number Q03401) contains all of the five conserved residues. This inconsistency argues against the universality of this triad. Thus, we speculate that these residues (especially the tetrad) located at the possible active site may act by another unknown mechanism. We still do not rule out the possibility that stecrisp may assemble to form a similar triad by a substrate-inducing dimerization or that CRISP-1s are mutated, non-enzymatic ramifications in the CRISP family.
It was suggested that another relatively conserved Ser 80 in Tex31, located in a loop, might get into close proximity of the highly conserved His 130 and Glu 115 (corresponding to His 115 and Glu 96 in stecrisp) and form a catalytic triad (30). Structurebased sequence alignment shows that the corresponding serine residue is not present in many PR-1 domains (Fig. 1). Some nearby serines, which were previously aligned by the authors, should participate in a conserved hydrogen-bonding interaction with a glutamate of the tetrad (such as Ser 61 -Glu 65 in stecrisp and Ser 55 -Glu 65 in GAPR-1; Fig. 3A). This hydrogen bond is maintained by a similar interaction of Asp 99 -Gln 111 as observed in Ves v 5 (Fig. 3B) when the glutamate is changed into glutamine in some PR-1 domains, including Tex31. Consequently, Ser 62 in stecrisp and Thr 100 in Ves v 5 may give better correspondence to the proposed Ser 80 in Tex31 (Fig. 1). Because of the constraint of conserved hydrogen bonding network com-posed of Ser 61 , Arg 66 , Cys 73 , Glu 75 , and Gln 146 (Fig. 3A), the side chain of the corresponding Ser 62 in stecrisp is positioned toward the solvent channel, far away from His 115 and Glu 96 (ϳ11 Å). Consistent with the constrained state, the relative low B-factors (19.28 -21.98 Å 2 ) of its backbone suggest that Ser 62 is not flexible enough to move the hydroxyl close to His 115 and Glu 96 . Therefore, a possible serine protease-like triad as suggested in Tex31 is not likely to be present in stecrisp. Similar hydrogen bond interactions are found in the structure of Ves v 5 (Fig. 3B). It would constrain the corresponding serine or threonine candidate far away from the His-Glu dyad. We do not exclude the possibility that Tex31 may utilize an un-conserved residue for catalytic purposes, as it is the only enzymatic PR-1 member identified to date. However, if a hydrogen bond between Asp 79 and Gln 94 , similar to that of Asp 99 -Gln 111 in Ves v 5, is present in Tex31 (Fig. 3B), the Ser 80 of Tex31 may also be positioned away from its His-Glu dyad.
Structure of the Hinge Region-It is interesting to find that the carboxyl-terminal portion of stecrisp is separated into two parts and that both are quite structurally condensed, with a strict constraint of local disulfide bridges (Fig. 2B). Four of ten carboxyl-terminal cysteines reside in the hinge region and form two crossed disulfides (167-174 and 170 -179), making the region a well defined left-handed knot with a short ␤-hairpin near the turn and two ends oriented in almost a right angle. Attributed to an insertion of cis Pro 30 kink, the outspread loop 30 -38 provides a scaffold to support the hinge region by hydrogen-bonding linkages mainly between the main chains. A small hydrophobic core composed of the side chains of conserved residues (Leu 37 , Thr 180 , Asn 181 , and Ile 220 ), the backbone of loop 180 -182, and the disulfide bridge Cys170 -179, further enhances the interaction between the PR-1 domain and the hinge (Fig. 5). Two conserved residues in the hinge, Pro 171 and Leu 178 , may be important for this special fold conformation and hydrophobic interaction between ␣2 and ␤10. Interestingly, the consensus of these residues in some other homologous proteins (such as GliPR, allurin, MAK248, and LGL1) with additional carboxyl-terminal motifs other than CRD suggests that they also contain the similar hinge fold. The hinge is presumably evolved to fix the carboxyl-terminal hydrophobic residues (Pro 160 -Tyr 161 ) of the PR-1 domain, which cover the critical hydrophobic core of ␣2-␣5-␣6, from being direct pulled away by the additional carboxyl-terminal patch. This structural segmentation is consistent with the genetic evidence that the last domain (CRD) is encoded by a separate exon (39).
Cysteine-rich Domain Has a Similar Fold with Two Ion Channel Blockers-The CRD (182-221), as a character of the CRISP family, folds into a compact motif with a hydrophobic core composed of several conserved apolar residues (Fig. 6). It contains three short ␣-helixes (␣7-␣9) maintained by two cross-disulfide bridges (192-210 and 201-214) and an eightamino acid loop tied up by the other disulfide bridge (183-216). The majority of the CRD is separated spatially from the PR-1 domain except for some contacts near the hinge. Leu 37 , Cys 170 , Cys 179 , Asn 181 , and Ile 220 make hydrophobic interactions at the interface of the two domains. Tyr 122 and Lys 123 form hydrogen bond and salt bridge linkages with the carboxyl terminus of the protein (Fig. 5). Apparently, these interactions do not seem to be enough to immobilize CRD. Orientation of this rigid domain is likely to swing in solvent and may differ with that in the crystal structure.
Searched by Dali server (40), the structure of CRD shows apparent similarity to that of Bgk and Shk (2.0-Å r.m.s.d. for 31 aligned C␣ and 3-Å r.m.s.d. for 30 C␣, respectively, Fig. 6), despite the low sequence identity (13% with Bgk and 17% with Shk in the aligned sequences). Bgk and Shk are 37-and 35residue peptides, respectively, from the sea anemone acting as voltage-sensitive K ϩ channel-blockers (41,42). They also contain six conserved cysteines and share a similar disulfide link-age mode to CRD (Cys 1 -Cys 6 , Cys 2 -Cys 4 , and Cys 3 -Cys 5 ). An absence of four consecutive residues in the amino acid sequence of Shk, as compared with Bgk and CRD of stecrisp (residues 12-15 using Bgk numbering), leads to a different organization of the N-terminal part of the toxin to that of Bgk and CRD. Thus, the structure of CRD is more similar to that of Bgk.

Great Variance of Functions May Be Present in PR-1 Domains-
We have shown that the characteristic carboxyl-terminal cysteine-rich region in stecrisp forms two compact and discrete motifs appended at the flank of the N-terminal PR-1 domain. Significant similarity between stecrisp and other PR-1 proteins reconfirms that they share a conserved structural domain. A PR-1 member, Tex31, has recently been identified as an endopeptidase. Although the affirmative enzymatic mechanism in Tex31 is still unknown, superposition of the conserved and solvent-exposed residues (His 60 , Glu 75 , Glu 96 , and His 115 ) in the known structures of PR-1 domains indicates that the highly conserved residues may share similar roles involved in catalysis and that CRISPs may also be enzymatic proteins (43).
In further experiments, no proteolytic activity of stecrisp has been unambiguously identified toward a diversity of chromogenic peptides and protein substrates. Previous work also reported similar negative results in identifying the enzymatic activities of some other PR-1 proteins (28,44). By alignment, three segments (62-72, 105-113, and 153-159) surrounding the largest surface cleft are hypervariable in CRISPs and other PR-1 proteins and show apparent structural difference between the superposed structures (as indicated in Figs. 1 and 2, C and D), implying great difference of the possible substrate binding site. Species-specific and substrate-specific interactions may be involved in recognition and catalysis between superfamily members. That may explain these negative results of enzymatic activity assays. As it has been frequently observed that superfamily members that share structural and functional features can catalyze a number of different reactions and that homologous proteins may evolve to have different functions and possess different functional sites (45)(46)(47)(48), it seems that there is an inability to readily transfer this overall reaction of Tex31 to other homologues.
Cysteine-rich Domain May Act as a Protein-Protein Interaction Module-As some CRISP members from venom have been documented to block some types of ion channels and have more distinct high molecule masses than most known blockers (13)(14)(15)(16)49), it is intriguing to find that CRD shares a similar fold with two K ϩ channel-blocking toxins, Bgk and Shk (33,34). A SXC (six-cysteine) domain analogous with these two toxins has been reported to be widespread in Caenorhabditis elegans and other nematodes (50). By alignment, most CRISPs show no obvious sequence homogeneity with them except for the six FIG. 6. Structural feature of CRD. Left, superposition of CRD (orange) with Bgk (white; Protein Data Bank code 1BGK). Side chains of three disulfide bridges are indicated in stick representation. Right, sequence alignment of Bgk, Shk, and CRDs of stecrisp, PsTx (GenBank TM number AY072695), pseudecin (AY072696), and helothermine (HLTX) (U13619). Cysteine residues are highlighted in orange, and other conserved residues in CRD are highlighted in blue. Two distinct residues between PsTx and pseudecin, comprising a possible interaction site for cyclic nucleotide-gated channels, are underlined in red, and the corresponding position is indicated by a red line in the figure. Single letter amino acid abbreviations are used with position numbers. cysteines (Fig. 6). The highest sequence identity is 25%, present between Shk and a CRISP member from Xenopus laevis as noticed previously (11). In addition, despite similar disulfide linkage mode, the strictly conserved pattern CXXXCXC with the last three of six cysteines of CRD discriminates itself from the characteristic CXXXCXXC pattern in the SXC domain and these two toxins (Fig. 6).
Given the relative detachment of CRD with its nitrogen moiety and the unique functions of CRISPs compared with those of other PR-1 homologues, the structural similarity makes it imaginable to attribute the carboxyl-terminal motif to those blockings, although we cannot exclude the possibility that other regions, including the PR-1 domain and the hinge, may provide additional interactions. A functional dyad composed of a lysine and a hydrophobic residue has been found to be conserved in Bgk and other potassium channel blockers in the sea anemone and the scorpion (51), whereas no similar dyad can be easily identified in modeled CRDs of several potassium channel interacting CRISPs. However, in two highly homologous CRISPs that target cyclic nucleotide-gated ion channels with very different affinities, two adjacent basic residues (Lys 175 and Arg 176 ) were suggested to be the interacting site with the channels (14,15). These residues just reside in the N-terminal loop of CRD (Fig. 6). Presumably, this coincidence may be attributed to the highly solvent accessible character of the exposed loop, which is favorable for forming a potential active site. It is important to note that in other mini-proteins/ domains, surface loops could also play a role as the "bait region" or "functional site" (52). In some snake venom CRISPs, two other residues, Phe 189 and Glu 186 , which are also in this loop, have been suggested to be the possible functional residues for blocking the smooth muscle contraction (13). Interaction sites with other channels may be also located in this domain. In fact, many peptide ion channel blockers with different targets are homologous, share same folds even with some protease inhibitors (53)(54)(55), or possess dual functions at the same time (56). Apparently, by repartitioning different amino acid side chains, these compact and stable folds are capable of providing excellent scaffolds for specific binding interactions.
Furthermore, the analogous SXC motif is known as an evolutionarily mobile module by either existing as SXC-only genes or flanking with multiple partner sequences of nonenzymatic and enzymatic domains, and it is suggested to be involved in a protein-protein interaction possibly specific to extracellular matrices or to act as a signaling ligand. Interestingly, several CRISPs (XCRISP and CRISP-1s) have also been reported to associate with the membranes of different cells such as hatching gland cells and spermatozoa (5,11). Deletion of CRD in XCRISP induces significant acceleration of the hatching process, and the carboxyl-terminal six amino acids tail is proved to be essential for XCRISP-mediated functions. Recently, human CRISP-3 was found to bind R1B-glycoprotein (A1BG), a member of the immunoglobulin superfamily in plasma (57). By alignment, most residues of CRD, other than a few structuresustaining ones, show great sequence variance among CRISP members, implying the diversity of possible interacting partners. Additionally, by extending in the direction of the putative substrate binding cleft, CRDs probably provide additional binding sites for the PR-1 domain. Thus, as a characteristic portion of CRISP family, CRD is likely to be an independent regulative module, interacting with different cofactors, receptors, or specific substrates and governing diverse functions of CRISP members along with the possible enzymatic activity of the nitrogen moiety.
Overall, structural detachment implies that there may be a functional separation between two domains in CRISPs. High homogeneity in the CRISP family ensures that other CRISPs will possess very similar three-dimensional structures like stecrisp, and it makes the structure of stecrisp a prototype for identifying possible functional sites (11,13), understanding molecular effects in different processed forms (6,7,58), and guiding further biochemical experiments.