Crystal Structure of Prostate Secretory Protein PSP94 Shows an Edge-to-Edge Association of two Monomers to Form a Homodimer

doi:10.1016/j.jmb.2010.02.035 J. Mol. Biol. (2010) 397, 947–956

Available online at www.sciencedirect.com

Crystal Structure of Prostate Secretory Protein PSP94Shows an Edge-to-Edge Association of two Monomersto Form a Homodimer

Ashwani Kumar1, Dhanashree D. Jagtap2,Smita D. Mahale2 and Mukesh Kumar3⁎
1High Pressure PhysicsDivision, Bhabha AtomicResearch Centre, Trombay,Mumbai 400085, India2Division of Structural Biology,National Institute for Researchin Reproductive Health,Jehangir Merwanji Street, Parel,Mumbai 400012, India3Solid State Physics Division,Bhabha Atomic Research Centre,Trombay, Mumbai 400085,India
Received 1 November 2009;received in revised form17 February 2010;accepted 18 February 2010Available online23 February 2010

*Corresponding author. E-mail [email protected] used: PSP94, prosta

94 residues; CRISP, cysteine-rich secrsingle isomorphous replacement wiscattering; PDB, Protein Data Bank; FPEG, polyethylene glycol.

0022-2836/$ - see front matter © 2010 E

Several recent genome-wide association studies have linked the humanMSMB gene, encoding prostate secretory protein of 94 residues (PSP94),with prostate cancer susceptibility. PSP94 is one of the most abundantproteins from prostatic secretions and a primary constituent of humansemen. PSP94 suppresses tumor growth and metastasis, and its expressiongradually decreases during progression of the prostate cancer. It is a rapidlyevolving protein with homologues present in several species with 10conserved cysteine residues. PSP94 homologues show high-affinity bindingwith different proteins from the cysteine-rich secretory protein family, someof which have been shown to be ion channel blockers. Here, we report thecrystal structure of human PSP94 at 2.3 Å resolution. The structure showsthat the amino and the carboxyl ends of the polypeptide chain are held inclose proximity facing each other. A strong hydrogen bond between theseends, which are located respectively on the first and the last β-strands, leadsto formation of an almost straight edge in PSP94 structure. Crystal structureshows that these edges from two PSP94 monomers associate in antiparallelfashion, leading to formation of a dimer. Our studies further show thatdimers dissociate into monomers at acidic pH, possibly through distortionof the straight edge. Further, based on several observations, we proposethat PSP94 binds to cysteine-rich secretory proteins and immunoglobulinG through the same edge, which is involved in the formation of PSP94dimeric interface.

© 2010 Elsevier Ltd. All rights reserved.

Edited by R. Huber
Keywords: microseminoprotein; IgBF; MSMB; PSP94; CRISP
Introduction

The human MSMB gene encoding β-microsemi-noprotein has attracted much attention recentlysince the two genome-wide association studieshave independently identified it to be linked withprostate cancer susceptibility,1,2 which were latelyconfirmed by other studies.3,4 The MSMB gene isexpressed abundantly by prostate epithelial cells

ess:

te secretory protein ofetory protein; SIRAS,th anomalousnI, fibronectin type I;

lsevier Ltd. All rights reserve

and the encoded protein is, therefore, also called theprostate secretory protein of 94 residues (PSP94).During the development of prostate cancer fromearly to late stages, expression of PSP94 progres-sively decreases.5 The loss of PSP94 expressionpossibly contributes to the development of prostatecancer as PSP94 has been reported to suppresstumor growth and metastasis.6

Emanating from the prostatic secretion, PSP94forms an abundant constituent (∼1.0 mg/ml) of thehuman semen. Although exact biological function ofPSP94 remains unknown, several potential roles ofPSP94 have been reported during the last threedecades. Kamada et al. have shown that PSP94 bindsto human immunoglobulin G (IgG) and referred to itas immunoglobulin binding factor.7 It was sug-gested that a high amount of PSP94 present in thehuman semen might be involved in regulation ofimmune response in the female reproductive tract

d.

mailto:[email protected]

http://dx.doi.org/10.1016/j.jmb.2010.02.035

948 PSP94 Crystal Structure

against the allogeneic sperm.7–9 PSP94 has also beenshown to be a motility inhibitor of sperm10 and aninhibitor of sperm acrosome reaction.11 Apart fromreproductive tissues, PSP94 has also been detectedin several nonreproductive tissues, such as therespiratory and gastric tract tissues, and in bothmales and females.12 PSP94 homologues havealso been identified in several other mammals13–15

and nonmammalian species.16–18 There is a largedegree of sequence diversity among these homo-logues, but a striking feature of PSP94 family ofproteins is the presence of 10 highly conservedcysteine residues.Efforts to understand the biological function of

PSP94 led the identification of cysteine-rich secre-tory proteins (CRISPs) as potential natural bindingpartners of PSP94. It has been shown that PSP94binds to human CRISP-3 in seminal plasma19 aswell as PSP94 binding protein (PSPBP or CRISP-9)in blood.20 A recent report has revealed that PSP94from human as well as porcine species form high-affinity complexes even with evolutionary diverseCRISPs present in several snake venoms.21 Thesesnake venom CRISPs have been shown to be ionchannel blockers22 while the mammalian CRISPsare thought to be involved in sperm maturation,gamete fusion, and host defense.23 The fact thatPSP94 from human and porcine species vary intheir amino acid sequences (51% identity) andboth of them bind to CRISPs of diverse specieslike snakes suggests that the interaction betweenthese two families of proteins may be quitegeneral. It is likely that some of the functions ofPSP94 might actually be orchestrated through itsbinding to CRISPs, but how PSP94 binds to avariety of CRISPs, including those from snakevenoms, is not known.PSP94 per se is a nonglycosylated protein

synthesized as a precursor polypeptide of 114residues with a cleavable signal peptide (residues1–20). The mature protein (94 residues) has amolecular mass of 10.7 kDa; however, it migratesas a 16- to 18-kDa band on SDS-PAGE. Mori et al.had speculated earlier that PSP94 is secreted as ahomodimer that does not bind IgG, but thesedimers get activated to monomeric form in femalereproductive tract to interact with IgG.8 However,these aspects have not been investigated furtherand the biologically relevant oligomeric state ofPSP94 remains unknown. Several recent reportshave, however, shown that PSP94 elutes as a singlepeak of ∼21 kDa from gel-filtration columnscorresponding to a possible homodimer.19,21,24

However, no dimers were observed in the NMRstructures of PSP94.The solution structure of PSP94 by NMR

reported earlier by two laboratories25,26 showsthat PSP94 has two distinct domains bridged by adisulfide bond. Although the secondary structureof the individual domains reported by the twolaboratories was similar, the relative orientation ofthe domains was very different (∼90° apart), givingrise to two distinct overall shapes (globular versus

elongated) for this small protein. The discrepancyseems to be due to the different interpretations ofthe 10 specific nuclear Overhauser effects, whichwere probably weak. As there are no crystalstructures known for any member of this familyof proteins, the relative orientation of two domainsof PSP94 could not be confirmed independently.We, therefore, crystallized PSP94 recently,27 andsubsequently, Kumar et al. also crystallized thisprotein under a different crystallization conditionand in a different space group.24 Though the unitcell dimensions along the a- and b-axis reported byKumar et al. are related to ours, it is different alongthe c-axis.24 Here, we determined the crystalstructure of PSP94 by single isomorphous replace-ment with anomalous scattering (SIRAS) methodusing crystals soaked in uranyl nitrate. The overallstructure of PSP94 reported here is similar to theNMR structure reported by Ghasriani et al., but therelative orientation of the two domains is shifted by∼20°.26 The reason for such a shift seems to havefunctional implications and is discussed. Thepresent structure further gives insight into thepossible mode of interaction of PSP94 with CRISPand IgG molecules.

Results

PSP94 was purified from human seminal plasmaand crystallized in tetragonal space group P41212with unit cell dimensions a= b=107.9 Å andc=82.1 Å as described earlier.27 There are fourmolecules per asymmetric unit. PSP94 structure wasdetermined from diffraction data extending to 2.3 Åresolution. Experimental phases were obtained fromSIRAS using crystals soaked in uranyl nitrate.Experimental electron density maps, improved bydensity modification, allowed unambiguous place-ment of most of the residues. The final, refinedatomic model contains 369 out of 376 amino acidresidues in four polypeptide chains of PSP94. Themissing region is a flexible loop (amino acids 10–16)in chain A. The atomic coordinates have beendeposited in the Protein Data Bank (PDB) with theaccession code 3IX0.

Overall structure of PSP94 monomer

PSP94 monomer has a long extended structure,rich in β-sheets (Fig. 1a). There are two distinctdomains: an N-terminal domain from residues 1–52and a C-terminal domain from residues 53–94. Thetwo domains are held together by a disulfide bondbetween Cys37 and Cys73. The N-terminal domainhas four antiparallel β-strands (β1: residues 1–6, β4:residues 30–34, β5: residues 38–42, and β6: residues46–51) arranged in the form of Greek key motif andtwo small antiparallel β-strands (β2: residues 18–21and β3: residues 23–27) forming a flap on top of theGreek key motif. There are three disulfide bonds inthe N-terminal domain. The disulfides Cys2-Cys50between strands β1 and β6 and Cys40-Cys49

Fig. 1. PSP94 monomer and itscomparison with FnI modules. (a)Cartoon representation of PSP94with polypeptide chain in rainbowcolor from blue (N-terminus) to red(C-terminus). Five disulfides areshown in stick representation. (b)Cartoon representation of fibronec-tin module pair (2FnI3FnI from PDBID 2rkz) with polypeptide chain inrainbow color and a bound peptide,making an antiparallel β-strand, inmagenta color. (c) Structure-basedsequence alignment of PSP94 withdifferent fibronectin FnI modules.The PDB files used for structuralsuperposition of different fibronec-tin modules are as follows: 1o9a for1FnI, 3cal for 2FnI and 3FnI, 2rky for4FnI and 5FnI, 1e8b for 6FnI, and3ejh for 8FnI and 9FnI. The com-pletely conserved cysteine residuesmaking disulfides are highlightedin yellow, and the other partiallyconserved residues are shown indifferent colors. The residue num-bering indicated at the top is withrespect to PSP94 sequence.

949PSP94 Crystal Structure

between strands β5 and β6 make the Greek keystructure rigid while the third disulfide Cys18-Cys42 between β2 and β5 orients the flap onto theGreek key motif. The C-terminal domain has twodouble-stranded antiparallel β-sheets. The strand β7(residues 55–58), which is an extension of the N-terminal strand β6, is associated with the C-terminalβ-strand β10 (residues 90–94). Interestingly, β10seems to get extended further by the first strand β1of the N-terminal domain. The other two β-strandsin the C-terminal domain, β8 (residues 64–70) andβ9 (residues 74–79), are longer and separated fromthe first pair. The arrangement of these four β-strands in the C-terminal domain is unusual andgives rise to a unique fold. The lone disulfide Cys64-Cys87 in the C-terminal domain brings rigidity tothe loop structure. One of the remarkable features ofPSP94 monomer is the close proximity of the aminoand carboxyl ends of the polypeptide chain, whichare facing each other with a strong hydrogen bondbetween the main-chain nitrogen atom of the firstresidue and one of the carboxyl oxygen atoms of thelast residue (N–O distances, 2.45–2.66 Å). Thesignificance of this hydrogen bond is discussed inthe next section.All the four molecules in the asymmetric unit are

structurally very similar. The superposition of theirCα atoms shows that the polypeptide chain fromresidues 1–6 and 32–94 in the four molecules followthe same course whereas the polypeptide chain fromresidues 7–31 in one of the molecule (chain D)

follows a slightly different course. The peptidesegment 8–17 contains mostly hydrophilic residuesand is flexible with poor electron density. Thepairwise root-mean-square deviation (RMSD) be-tween the four molecules in the asymmetric unit is0.6–0.8 Å when Cα atoms of the peptide segments 1–6 and 18–94 were used in the superposition by leastsquares.

Structural similarity search

The structural similarity search for PSP94 showsthat the C-terminal domain has no structuralsimilarity with any known proteins, but residues15–52 in the N-terminal domain have a fold similarto the fibronectin type I (FnI) module (Fig. 1b). TheRMSD of these aligned residues between PSP94 anddifferent FnI structures are ∼2.1–3.0 Å. The struc-ture-based sequence alignment further shows thatthe four conserved cysteine residues in the FnIstructures, which make two disulfide bonds andhold the β-strands together, are present at thecorresponding positions even in the PSP94 structure(Fig. 1c). PSP94, however, has two extra cysteineresidues (Cys37 and Cys50) in this region of thesequence, which make additional disulfide bonds.Cys37 makes disulfide bond with Cys73 located inthe C-terminal domain whereas Cys50 makesdisulfide bond with Cys2 located on the first strand.Further, although the first strand, β1, of the Greekkey motif of PSP94 is absent in the native FnI

Fig. 2. PSP94 dimer. (a) Cartoon representation showing the edge-to-edge interaction of two PSP94 monomers withtheir polypeptide chains in blue and orange colors. (b) Topology diagram of dimeric interaction showing extension of β-sheet across the dimer interface (broken vertical line). The β-strands in the N-terminal domain are shown in cyan, andthose in the C-terminal domain are shown in red. (c) Elution profile of monomeric and dimeric PSP94. Purified PSP94(400–500 μg) was subjected to gel-filtration chromatography using a Superdex G-75 column (Hi-Load 16/60, Amersham)at a flow rate of 0.5 ml/min. The elution buffers used were as follows: 50 mM Tris, pH 8.0 (in blue), and 50 mM acetate,pH 4.5 (in red).


structures, comparison with the FnI bound to acognate peptide either from a fibronectin bindingprotein of a pathogenic bacterium28,29 or from the α1chain of type I collagen30 is noteworthy. The boundpeptide, which forms an additional antiparallel β-strand with the FnI module pair, compares well withthe first strand, β1, of PSP94 (Fig. 1b).

PSP94 dimer

The arrangement of four molecules of PSP94 inthe crystallographic asymmetric unit suggests thatthey form two dimers. These two dimers are relatedto each other by 1/2 unit cell translation along thez-axis. Within a dimer, there is a non-crystallo-graphic dyad axis perpendicular to the z-axis thatorients the N-terminal domain of one monomeradjacent to the C-terminal domain of the other

monomer (Fig. 2a). In this arrangement, the β-sheetof the N-terminal Greek key motif in each monomergets extended across the dimer interface by the C-terminal strands β10 and β7 of the adjacentmonomer. This leads to the formation of two six-stranded β-sheets in the PSP94 dimer (Fig. 2b).Because the dimer is formed by edge-to-edgeinteraction of the two monomers, the surface areaburied between them (960 Å2) is relatively less. Thedimeric association is stabilized by a number ofinteractions that include eight interchain backbonehydrogen bonds, a hydrogen bond between Oγ

atom of Ser1 from one of the monomer and theterminal carboxyl oxygen atom (O2) of Ile94 fromthe other monomer, and 66 other interchain con-tacts shorter than 4.0 Å. All these interchaincontacts are listed in Table 1. The dimeric interac-tion further brings the two N- and two C-terminal

Table 1. List of intermolecular contacts at the dimerinterface less than 4.0 Å

Source atoms(from chain A):residue no.(residue name)/atom name

Target atoms(to chain B):residue no

(residue name)/atom name

Distance(Å)

Res no (Res name)/Atom name1(SER)./CA 94(ILE)./O 3.581(SER)./C 94(ILE)./O 3.751(SER)./CB 94(ILE)./O 3.78

94(ILE)./OXT 3.871(SER)./OG 94(ILE)./C 3.97

94(ILE)./O 3.9294(ILE)./OXT 3.291(SER)./N 3.221(SER)./CA 3.131(SER)./CB 2.981(SER)./OG 3.95

2(CYS)./N 94(ILE)./C 3.9594(ILE)./O 2.96

2(CYS)./CA 94(ILE)./O 3.852(CYS)./O 93(ILE)./CG1 3.61

93(ILE)./CB 3.6793(ILE)./CG2 3.5093(ILE)./CA 3.3793(ILE)./C 3.8294(ILE)./N 3.2594(ILE)./O 3.83

2(CYS)./CB 94(ILE)./O 3.863(TYR)./CA 92(TRP)./O 3.323(TYR)./C 92(TRP)./O 3.633(TYR)./CB 91(GLU)./OE2 3.69

92(TRP)./O 3.724(PHE)./N 92(TRP)./O 2.974(PHE)./CA 92(TRP)./O 3.984(PHE)./O 91(GLU)./CA 3.26

91(GLU)./CB 3.9691(GLU)./C 3.6391(GLU)./CG 3.8692(TRP)./N 2.9892(TRP)./O 3.66

4(PHE)./CE1 94(ILE)./CG2 3.635(ILE)./CA 90(SER)./O 3.575(ILE)./CB 90(SER)./O 3.575(ILE)./CG2 90(SER)./O 3.506(PRO)./CD 90(SER)./CA 3.96

90(SER)./C 3.7090(SER)./O 2.99

90(SER)./C 6(PRO)./CD 3.7690(SER)./O 5(ILE)./CA 3.80

5(ILE)./CB 3.995(ILE)./CG2 3.796(PRO)./CG 3.776(PRO)./CD 2.90

91(GLU)./CA 4(PHE)./O 3.2691(GLU)./C 4(PHE)./O 3.6091(GLU)./CB 4(PHE)./O 3.8891(GLU)./CG 5(ILE)./CG2 3.98

4(PHE)./O 3.6492(TRP)./N 4(PHE)./O 2.9492(TRP)./C 4(PHE)./N 3.9692(TRP)./O 3(TYR)./CA 3.39

3(TYR)./C 3.613(TYR)./CB 3.794(PHE)./N 2.884(PHE)./CA 3.894(PHE)./O 3.57

93(ILE)./CA 2(CYS)./O 3.3993(ILE)./C 2(CYS)./O 3.7493(ILE)./CB 2(CYS)./O 3.8593(ILE)./CG1 2(CYS)./O 3.8593(ILE)./CG2 2(CYS)./O 3.7594(ILE)./N 2(CYS)./O 3.0994(ILE)./C 2(CYS)./N 3.81

Table 1 (continued)

Source atoms(from chain A):residue no.(residue name)/atom name

Target atoms(to chain B):residue no

(residue name)/atom name

Distance(Å)

Res no (Res name)/Atom name94(ILE)./O 1(SER)./CA 3.23

1(SER)./C 3.422(CYS)./N 2.712(CYS)./CA 3.722(CYS)./O 3.662(CYS)./CB 3.85

94(ILE)./CG2 2(CYS)./CB 4.004(PHE)./CE2 3.71


ends of the monomers in close proximity to eachother (Figs. 2a and 3a).

pH-driven monomer/dimer transition

To test whether PSP94 dimer is a manifestationof crystal packing or whether it exists even insolution, we performed gel-filtration chromatogra-phy. PSP94 eluted as a single peak correspondingto a molecular mass of ∼21 kDa (dimer) whenelution was carried out with 50 mM Tris at pH 8.0.This was consistent with several independentreports published earlier.19,21,24 However, whenelution was carried out at the pH of crystallizationwith 50 mM sodium acetate, pH 4.5, PSP94 elutedas a single peak corresponding to a molecular massof ∼10 kDa (monomer) (Fig. 2c). The transition ofPSP94 from dimer to monomer seems to occur bysubtle changes in the solution environment. Thefact that we observe a dimer in the crystal seems tosuggest that either the precipitant [45% polyethyl-ene glycol (PEG) 400] in the crystallization bufferor the crystalline environment itself might havecoaxed the monomers to associate in the form ofnative dimers while packing the molecules in thecrystal.

Discussion

Mechanism of monomer/dimer transition

The edge-to-edge interaction of β-sheets fromtwo different polypeptide chains is an importantmode of protein–protein interaction and has beenobserved in several protein structures either be-tween the two identical polypeptide chains result-ing in the formation of homodimers or betweentwo different polypeptide chains as in the hetero-dimer complexes.31,32 The crystal structurereported here shows a rather unique edge-to-edgeinteraction of two PSP94 monomers. This ‘edge-on’interaction is facilitated by the formation of analmost straight edge in PSP94 monomer containingtwo terminal strands, β1 and β10. These twostrands, which terminate in the amino and thecarboxyl ends of the polypeptide chain facing each

Fig. 3. PSP94—dimer to monomer. (a) The stick representation of the terminal residues from two different polypeptidechains (with green and yellow carbons) at the dimeric interface is shown along with their electron density maps (2Fo−Fc)contoured at 1.5σ. Oxygen and nitrogen atoms are colored red and blue, respectively. The hydrogen-bond distances aremarked in angstroms. Superposition of either (b) the C-terminal domain or (c) the N-terminal domain of the NMRmodels(from PDB file 2iz3 and shown in brown) onto the corresponding domain of the crystal structure (four polypeptide chainsin rainbow color) showing a shift in the relative orientation of two domains.


other, are held in place mainly due to a stronghydrogen bond between the main-chain nitrogenatom of the first residue, Ser1, and one of theoxygen atoms (O2) of the terminal carboxyl groupof the last residue, Ile94 (Fig. 3a). The environmentaround the terminal carboxyl group suggests that itis ionized in the present structure. The ionizedterminal carboxyl group seems to be essential forholding the two strands together and providing astraight edge for the dimeric interactions. However,at the acidic pH, close to the pKa (∼4.0) of freecarboxyl group, one of the oxygen atoms (O2) ofthe terminal carboxyl group can get protonated andas the protonated hydroxyl (–OH) oxygen of thecarboxyl group cannot accept a hydrogen bond33

from a potential donor group (NH3+ of Ser1)

nearby, the hydrogen-bonding interaction holdingthe amino and the carboxyl ends of the polypeptidechain will be broken. The terminal residues (Ser1and Ile94) would then move away from each otherto avoid close contacts resulting in distortion of thestraight edge, which would lead to disruption ofthe β-sheet interactions at the dimer interface andsubsequent dissociation of the dimer. Interestingly,in the solution structure of PSP94 by NMR,26 asimilar movement of the terminal residues is seenwhere the main-chain nitrogen atom of Ser1 andthe carboxyl oxygen atom of Ile94 in differentmodels are far apart (6.0–8.9 Å). Thus, the NMRstructure, wherein the samples were prepared inwater at pH 6.0 without any buffer, wouldrepresent the monomeric form of PSP94. In thepresent structure, the crystalline environmentseems to stabilize the dimeric form even at lowerpH. Further analysis shows that the relativeorientation of the two domains of PSP94 in thepresent dimeric structure is slightly different from

that in the NMR structure (PDB ID 2IZ3). When allthe Cα atoms of one of the domains in the NMRstructure are superposed on the correspondingatoms in the present structure, the other domainin the NMR structure appears shifted as comparedto the present structure (Fig. 3b and c) and the shiftis of a pure rotation of ∼20° around an axis locatedbetween the two domains. The domain shift seemsto be needed to avoid close contacts between theamino and carboxyl ends of the polypeptide chainwithout altering the β-sheet structure in theindividual domains. Due to the shift in theorientation of two domains, the strands β1 andβ10 appear twisted in the NMR structure and nolonger form a straight edge that is needed to makethe dimeric interaction.

PSP94–CRISP interaction

PSP94 and CRISP families of proteins are presentin several organisms, and the fact that there is ahigh-affinity binding between them19–21 suggeststhat these interactions must be of considerablephysiological relevance and some of the functionsof PSP94 might actually be mediated through itsinteraction with CRISPs. PSP94 has a partial butstriking structural similarity with FnI modules withconserved disulfides as described in the previoussection. Different proteins are thought to bindfibronectins by adding a β-strand in antiparallelfashion to the existing β-sheet structure of fibro-nectin modules.29 Whether different CRISPs wouldalso bind to the PSP94 family of proteins in asimilar fashion remains a distinct possibility.Interaction of PSP94 with human CRISP-3 hasrecently been studied by NMR.34 It was found thatthe residues in PSP94 that get affected upon

Fig. 4. PSP94 binding partners. (a) A model for interaction of PSP94 (blue) with CRISP-3 (orange). CRISP-3 model wasbuilt by homology modeling based on triflin structure. Right panel shows another view rotated by 180° around a verticalaxis as shown. (b) A model for interaction of PSP94 (blue) with Fab domain of IgG. The Fab model was taken from PDBfile 1igc. The variable and constant domains of the heavy (in orange) and the light (in pink) chains of Fab are labeled. Rightpanel shows another view rotated by 180° around a vertical axis as shown.


complex formation are located on the two terminalβ-strands (β1 and β10) as well as two strands (β6and β7) that are associated with these terminalstrands. Based on this observation, the authorsproposed a model wherein the first β-strand (β1) ofPSP94 binds to an accessible β-strand of CRISP-3,while the last β-strand (β10) of PSP94 binds to adifferent small β-strand in CRISP-3, making twoparallel β-sheet interactions at the PSP94–CRISP-3interface. However, our analysis shows that suchan interaction make a knot between the polypep-tide chains of PSP94 and CRISP-3. In the crystalstructure reported here, we find that the terminalstrands (β1 and β10) of PSP94 are involved in theformation of a dimer with antiparallel associationof β-strands at the protein–protein interface. Sincethese terminal strands get affected upon PSP94–CRISP-3 complex formation, it is possible thatCRISP-3 would replace one of the PSP94 moleculesfrom the dimer and form a PSP94–CRISP-3 hetero-dimer complex. The 1:1 stoichiometry in thePSP94–CRISP complex has also been suggested

recently in the binding study of porcine PSP94 withdifferent snake venom CRISPs.21 We, therefore,attempted to model PSP94–CRISP interaction insuch a way that the terminal β-strands, β1 and β10,of PSP94 interact with a single accessible β-strandof CRISP-3 in an antiparallel fashion (Fig. 4a),similar to the protein–protein interaction observedin the formation of the PSP94 dimer. In this model,a part of the sperm-coating protein domain inCRISP wraps around one side of PSP94 to increasethe intermolecular interaction, which would resultin the formation of a higher affinity PSP94–CRISPcomplex than the formation of the PSP94–PSP94homodimer, consistent with the experimental ob-servation reported earlier.19 Thus, the straight edgeof PSP94 containing two terminal β-strands seemsto be ‘sticky’ as it is involved in binding to differentCRISPs, and when the binding partners are notpresent, two of them associate to form a PSP94dimer. Further, the interaction between the second-ary structural elements of PSP94 and CRISPs, asproposed here, would be tolerant to sequence

Table 2. Data collection and refinement statistics

Crystal Native U derivative

Data collection statisticsa

Space group P41212 P41212Cell dimensionsa, b, c (Å) 107.88,

107.88, 92.13107.90,

107.90, 91.76α, β, γ (°) 90, 90, 90 90, 90, 90Resolution (Å) 50.0–2.3

(2.42–2.30)39.31–2.50(2.64–2.50)

Reflections measured 198,419(28,313)

137,140(19,260)

Unique reflections 24,778 (3528) 19,314 (2767)Completeness (%) 100 (100) 99.9 (99.8)Average I/σ 16.1 (4.1) 13.3 (3.6)Rmerge

b 0.104 (0.460) 0.114 (0.538)Multiplicity 8.0 (8.0) 7.1(7.0)

Refinement statisticsResolution (Å) 39.43–2.30No. of reflections 23,955Completeness (%) 96.87Total non-H atoms 3176Protein 2938Water 238Rcryst/Rfree

c 0.213/0.266Residues in favored regions

of the Ramachandran plot (%)96.7

Residues in allowed regionsof the Ramachandran plot (%)

3.0

RMSD ideal bond length (Å) 0.007RMSD ideal bond angles (°) 1.0

a Data for the highest-resolution shell are given in parentheses.b Rmerge=∑|Ii− ⟨I⟩|/∑Ii, where Ii is the intensity of an

individual reflection and ⟨I⟩ is the mean intensity obtained frommultiple observations of symmetry-related reflections.

c Rcryst =∑||Fobs|−|Fcalc||/∑|Fobs|; 6.2% randomlyomitted reflections were used for Rfree.


variations among members in these two families ofproteins and thus would provide an ingenioussolution in maintaining high affinity betweenproteins from these two families.

PSP94–IgG interaction

PSP94 binds to different human IgG7 and it hasbeen speculated that PSP94 may be involved inregulation of immune response in the femalereproductive tract.8 The binding of PSP94 tohuman IgG remains unaffected even when all thefive disulfides in PSP94 are reduced, suggesting theinvolvement of sequential epitopes of PSP94 in IgGbinding.9 Further, our recent results suggest thatPSP94 binds to Fab as well as Fab2 domains and notto the Fc domain of human IgG (to be published).Interestingly, similar binding characteristics havebeen reported in binding of protein G, a cell surfaceprotein from Streptococcus, with IgG.35 Crystalstructure of protein G with Fab fragment showsthat the outer β-strand in protein G forms anantiparallel interaction with the last β-strand in theconstant heavy-chain domain (CH1) of IgG, leadingto an extension of β-sheet across the bindinginterface. It is likely that PSP94 would also bindto IgG in a similar fashion with its ‘sticky’ edge,containing two terminal β-strands β1 and β10,interacting with the last β-strand in the CH1domain of Fab in an antiparallel fashion (Fig. 4b).Such interaction between the secondary structuralelements of PSP94 and the CH1 domain wouldagain be tolerant to variations among different IgGmolecules.PSP94 has been attracting interest for the last

three decades due to its abundant presence inhuman semen. Several biological roles have beenproposed since then,16 but the exact function ofPSP94 still remains elusive. The overall structure ofPSP94 reported here is similar to the NMR structurereported by Ghasriani et al., but the relativeorientation of the two domains is shifted by∼20°.26 The crystal structure reported here showshow the edges from two PSP94 monomers associateto form a dimer. Further, the present paper showsfor the first time a pH-induced transition of PSP94from dimeric to monomeric form. This assumesbiological significance as PSP94 would function as adimer in the human semen (pH 7.5–8.0), whereasthe dimers may dissociate into ‘active’ monomerswhen it comes to vagina (pH 3.8–5.0). The conceptof inactive dimers and the active monomers, in thecontext of PSP94 binding to IgG,8 was proposedearlier, but the role of vaginal pH in ‘activating’PSP94 had not been investigated there. Further,based on several observations, we propose herethat the interaction of PSP94 with several CRISPsand IgG molecules may also occur through the N-and C-terminal β-strands of PSP94. The presentreport thus provides valuable insight into thestructure of PSP94 as well as its dimeric form andits possible mode of interaction with differentbinding proteins, which, in turn, may provide

useful clues to the biological functions of thisimportant family of proteins.

Materials and Methods

Protein purification, crystallization, and data collection

PSP94 was purified from human seminal plasma usinga well-established protocol.9 Crystals were grown byhanging drop vapor diffusion as reported earlier.27

Briefly, drops were prepared by mixing 2 μl of proteinsolution (10 mg ml−1 in water) with 2 μl reservoirsolution [0.1 M sodium acetate, pH 4.5, 0.2 M lithiumsulfate, and 44–47% (v/v) PEG 400] and 0.4 ml additivesolution (γ-butyrolactone) and equilibrated against 1 mlreservoir solution at 293 K in a sealed well. Thecrystallization buffer, which contained ∼45%(v/v) PEG400, was directly used for flash freezing the crystals inliquid nitrogen without the addition of any furthercryoprotectant solution. Heavy-atom derivatives wereprepared by soaking PSP94 crystals overnight incrystallization buffer containing 1 mM uranyl nitrate[UO2(NO3)2·6H2O]. The soaked crystals were back-soaked in crystallization buffer lacking uranium beforefreezing. X-ray diffraction data on native crystals as wellas uranium derivatives were collected at X06DA (PXIII)beamline of the Swiss Light Source, Paul ScherrerInstitut, Villigen, Switzerland. Diffraction data wereprocessed using MOSFLM36 and SCALA37 or XDS.38


Structure determination

The crystal structure of PSP94 was determined usingSIRAS using the uranium derivative. Table 2 contains datacollection and refinement statistics. The program SOLVEas implemented in the PHENIX suite39 was used to locatesix U atoms, refine their positions, and calculate experi-mental phases. Phase extension, density modification, andautomatic tracing of the polypeptide chain were carriedout in RESOLVE/PHENIX. Almost 80% of the residues infour polypeptide chains were built by RESOLVE. Furthermodel buildings were carried out manually in Coot40 andO41 using the density-modified map produced by RE-SOLVE. Few more residues were added during iterativerounds of refinement in CNS42 followed by rebuilding inO. Finally, the refinement converged to Rwork and Rfree of21.3% and 26.6%, respectively. Illustrations were made inPyMOL (DeLano Scientific).

PDB accession number

The atomic coordinates and structure factors have beendeposited in the PDB with accession number 3IX0.

Acknowledgements

We thank Prof. B. V. V. Prasad (Baylor College ofMedicine, Houston) and Dr. M.V. Hosur forcritically reading the manuscript; Dr. S Kannan(Fuel Chemistry Division, BARC) for providingmany heavy-atom compounds for the soakingexperiments; Dr. K. K. Kannan, Dr. M. Ramanad-ham, Vishal Prashar, Amit Das, and S. C. Bihani formany useful discussions; and S. R. Jadhav for thetechnical help. We are grateful to Dr. Meitian Wangand other staff members of PXIII beamline at SwissLight Source for their help during data collection.We thankfully acknowledge the travel supportprovided by the Department of Science and Tech-nology, Government of India, for our visits to SwissLight Source for data collection.

References

1. Eeles, R. A., Kote-Jarai, Z., Giles, G. G., Olama, A. A.,Guy, M., Jugurnauth, S. K. et al. (2008). Multiple newlyidentified loci associated with prostate cancer suscep-tibility. Nat. Genet. 40, 316–321.

2. Thomas, G., Jacobs, K. B., Yeager, M., Kraft, P.,Wacholder, S., Orr, N. et al. (2008). Multiple lociidentified in a genome-wide association study ofprostate cancer. Nat. Genet. 40, 310–315.

3. Lou, H., Yeager, M., Li, H., Bosquet, J. G., Hayes, R. B.,Orr, N. et al. (2009). Fine mapping and functionalanalysis of a common variant in MSMB on chromo-some 10q11.2 associated with prostate cancer suscep-tibility. Proc. Natl Acad. Sci. USA, 106, 7933–7938.

4. Chang, B. L., Cramer, S. D., Wiklund, F., Isaacs, S. D.,Stevens, V. L., Sun, J. et al. (2009). Fine mappingassociation study and functional analysis implicate aSNP in MSMB at 10q11 as a causal variant for prostatecancer risk. Hum. Mol. Genet. 18, 1368–1375.

5. Beke, L., Nuytten, M., Van Eynde, A., Beullens, M. &Bollen, M. (2007). The gene encoding the prostatictumor suppressor PSP94 is a target for repression bythe Polycomb group protein EZH2. Oncogene, 26,4590–4595.

6. Shukeir, N., Arakelian, A., Kadhim, S., Garde, S. &Rabbani, S. A. (2003). Prostate secretory protein PSP-94 decreases tumor growth and hypercalcemia ofmalignancy in a syngenic in vivo model of prostatecancer. Cancer Res. 63, 2072–2078.

7. Kamada, M., Mori, H., Maeda, N., Yamamoto, S.,Kunimi, K., Takikawa, M. et al. (1998). β-Microsemi-noprotein/prostatic secretory protein is a memberof immunoglobulin binding factor family. Biochim.Biophys. Acta, 1388, 101–110.

8. Mori, H., Kamada, M., Maegawa, M., Yamamoto, S.,Aono, T., Futaki, S. et al. (1998). Enzymatic activationof immunoglobulin binding factor in female repro-ductive tract. Biochem. Biophys. Res. Commun. 246,409–413.

9. Jagtap, D. D., Narahari, A., Swamy, M. J. & Mahale, S.D. (2007). Disulphide bond reduction and S-carbox-amidomethylation of PSP94 affects its conformationbut not the ability to bind immunoglobulin. Biochim.Biophys. Acta, 1774, 723–731.

10. Chao, C. F., Chiou, S. T., Jeng, H. & Chang, W. C.(1996). The porcine sperm motility inhibitor isidentical to beta-microseminoprotein and is a com-petitive inhibitor of Na+,K(+)-ATPase. Biochem.Biophys. Res. Commun. 218, 623–628.

11. Anahi Franchi, N., Avendano, C., Molina, R. I.,Tissera, A. D., Maldonado, C. A., Oehninger, S. et al.(2008). β-Microseminoprotein in human spermatozoaand its potential role in male fertility. Reproduction,136, 157–166.

12. Baijal-Gupta, M., Clarke, M. W., Finkelman, M. A.,McLachlin, C. M. & Han, V. K. (2000). Prostaticsecretory protein (PSP94) expression in human femalereproductive tissues, breast and in endometrial cancercell lines. J. Endocrinol. 165, 425–433.

13. Fernlund, P., Granberg, L. B. & Roepstorff, P. (1994).Amino acid sequence of beta-microseminoproteinfrom porcine seminal plasma. Arch. Biochem. Biophys.309, 70–76.

14. Makinen, M., Valtonen-Andre, C. & Lundwall, A.(1999). New World, but not Old World, monkeyscarry several genes encoding beta-microseminopro-tein. Eur. J. Biochem. 264, 407–414.

15. Xuan, J. W., Kwong, J., Chan, F. L., Ricci, M., Imasato,Y., Sakai, H. et al. (1999). cDNA, genomic cloning, andgene expression analysis of mouse PSP94 (prostatesecretory protein of 94 amino acids).DNA Cell Biol. 18,11–26.

16. Lazure, C., Villemure, M., Gauthier, D., Naude, R. J. &Mbikay, M. (2001). Characterization of ostrich(Struthio camelus) beta-microseminoprotein (MSP):identification of homologous sequences in EST data-bases and analysis of their evolution during specia-tion. Protein Sci. 10, 2207–2218.

17. Aoki, N., Sakiyama, A., Deshimaru, M. & Terada, S.(2007). Identification of novel serum proteins in aJapanese viper: homologs of mammalian PSP94.Biochem. Biophys. Res. Commun. 359, 330–334.

18. Aoki, N., Matsuo, H., Deshimaru, M. & Terada, S.(2008). Accelerated evolution of small serum proteins(SSPs)—the PSP94 family proteins in a Japanese viper.Gene, 426, 7–14.

19. Udby, L., Lundwall, A., Johnsen, A. H., Fernlund, P.,Valtonen-Andre, C., Blom, A. M. et al. (2005). β-


Microseminoprotein binds CRISP-3 in human seminalplasma. Biochem. Biophys. Res. Commun. 333, 555–561.

20. Reeves, J. R., Xuan, J. W., Arfanis, K., Morin, C.,Garde, S. V., Ruiz, M. T. et al. (2005). Identification,purification and characterization of a novel humanblood protein with binding affinity for prostatesecretory protein of 94 amino acids. Biochem. J. 385,105–114.

21. Hansson, K., Kjellberg, M. & Fernlund, P. (2009).Cysteine-rich secretory proteins in snake venomsform high affinity complexes with human and porcinebeta-microseminoproteins. Toxicon, 54, 128–137.

22. Yamazaki, Y. & Morita, T. (2004). Structure andfunction of snake venom cysteine-rich secretoryproteins. Toxicon, 44, 227–231.

23. Gibbs, G. M., Roelants, K. & O'Bryan, M. K. (2008).The CAP superfamily: cysteine-rich secretory pro-teins, antigen 5, and pathogenesis-related 1 proteins—roles in reproduction, cancer, and immune defense.Endocr. Rev. 29, 865–897.

24. Kumar, V., Roske, Y., Singh, N., Heinemann, U.,Singh, T. P. & Yadav, S. (2009). Purification andpreliminary X-ray crystallographic studies of beta-microseminoprotein from human seminal plasma.Acta Crystallogr., Sect. F, 65, 518–521.

25. Wang, I., Lou, Y. C., Wu, K. P., Wu, S. H., Chang,W. C. & Chen, C. (2005). Novel solution structure ofporcine beta-microseminoprotein. J. Mol. Biol. 346,1071–1082.

26. Ghasriani, H., Teilum, K., Johnsson, Y., Fernlund, P. &Drakenberg, T. (2006). Solution structures of humanand porcine beta-microseminoprotein. J. Mol. Biol. 362,502–515.

27. Kumar, M., Jagtap, D. D., Mahale, S. D., Prashar, V.,Kumar, A., Das, A. et al. (2009). Crystallization andpreliminary X-ray diffraction analysis of humanseminal plasma protein PSP94. Acta Crystallogr., Sect.F, 65, 389–391.

28. Bingham, R. J., Rudino-Pinera, E., Meenan, N. A.,Schwarz-Linek, U., Turkenburg, J. P., Hook, M. et al.(2008). Crystal structures of fibronectin-binding sitesfrom Staphylococcus aureus FnBPA in complex withfibronectin domains. Proc. Natl Acad. Sci. USA, 105,12254–12258.

29. Schwarz-Linek, U., Werner, J. M., Pickford, A. R.,Gurusiddappa, S., Kim, J. H., Pilka, E. S. et al. (2003).Pathogenic bacteria attach to human fibronectinthrough a tandem beta-zipper. Nature, 423, 177–181.

30. Erat, M. C., Slatter, D. A., Lowe, E. D., Millard, C. J.,Farndale, R. W., Campbell, I. D. et al. (2009).

Identification and structural analysis of type I collagensites in complex with fibronectin fragments. Proc. NatlAcad. Sci. USA, 106, 4195–4200.

31. Dou, Y., Baisnee, P. F., Pollastri, G., Pecout, Y.,Nowick, J. & Baldi, P. (2004). ICBS: a database ofinteractions between protein chains mediated bybeta-sheet formation. Bioinformatics, 20, 2767–2777.

32. Remaut, H. & Waksman, G. (2006). Protein–proteininteraction through beta-strand addition. TrendsBiochem. Sci. 31, 436–444.

33. Ramanadham, M., Jakkal, V. S. & Chidambaram, R.(1993). Carboxyl group hydrogen bonding in X-rayprotein structures analysed using neutron studies onamino acids. FEBS Lett. 323, 203–206.

34. Ghasriani, H., Fernlund, P., Udby, L. & Drakenberg,T. (2009). A model of the complex between humanbeta-microseminoprotein and CRISP-3 based on NMRdata. Biochem. Biophys. Res. Commun. 378, 235–239.

35. Derrick, J. P. & Wigley, D. B. (1992). Crystal structureof a streptococcal protein G domain bound to an Fabfragment. Nature, 359, 752–754.

36. Leslie, A. G. (2006). The integration of macromolec-ular diffraction data. Acta Crystallogr., Sect. D: Biol.Crystallogr. 62, 48–57.

37. Collaborative Computational Project No. 4. (1994).The CCP4 suite: programs for protein crystallogra-phy. Acta Crystallogr., Sect. D: Biol. Crystallogr. 50,760–763.

38. Kabsch, W. (1993). Automatic processing of rotationdiffraction data from crystals of initially unknownsymmetry and cell constants. J. Appl. Crystallogr. 26,795–800.

39. Adams, P. D., Grosse-Kunstleve, R. W., Hung, L. W.,Ioerger, T. R., McCoy, A. J., Moriarty, N. W. et al.(2002). PHENIX: building new software for automat-ed crystallographic structure determination. ActaCrystallogr., Sect. D: Biol. Crystallogr. 58, 1948–1954.

40. Emsley, P. & Cowtan, K. (2004). Coot: model-buildingtools for molecular graphics. Acta Crystallogr., Sect. D:Biol. Crystallogr. 60, 2126–2132.

41. Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard, M.(1991). Improved methods for building proteinmodels in electron density maps and the location oferrors in these models.Acta Crystallogr., Sect. A: Found.Crystallogr. 47, 110–119.

42. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano,W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998).Crystallography & NMR system: a new software suitefor macromolecular structure determination. ActaCrystallogr., Sect. D: Biol. Crystallogr. 54, 905–921.

Documents

Crystal Structure of Prostate Secretory Protein PSP94 Shows an Edge-to-Edge Association of two Monomers to Form a Homodimer