Structural Characterization of PTX3 Disulfide Bond Network and Its Multimeric Status in Cumulus Matrix Organization*

PTX3 is an acute phase glycoprotein that plays key roles in resistance to certain pathogens and in female fertility. PTX3 exerts its functions by interacting with a number of structurally unrelated molecules, a capacity that is likely to rely on its complex multimeric structure stabilized by interchain disulfide bonds. In this study, PAGE analyses performed under both native and denaturing conditions indicated that human recombinant PTX3 is mainly composed of covalently linked octamers. The network of disulfide bonds supporting this octameric assembly was resolved by mass spectrometry and Cys to Ser site-directed mutagenesis. Here we report that cysteine residues at positions 47, 49, and 103 in the N-terminal domain form three symmetric interchain disulfide bonds stabilizing four protein subunits in a tetrameric arrangement. Additional interchain disulfide bonds formed by the C-terminal domain cysteines Cys317 and Cys318 are responsible for linking the PTX3 tetramers into octamers. We also identified three intrachain disulfide bonds within the C-terminal domain that we used as structural constraints to build a new three-dimensional model for this domain. Previously it has been shown that PTX3 is a key component of the cumulus oophorus extracellular matrix, which forms around the oocyte prior to ovulation, because cumuli from PTX3-/- mice show defective matrix organization. Recombinant PTX3 is able to restore the normal phenotype ex vivo in cumuli from PTX3-/- mice. Here we demonstrate that PTX3 Cys to Ser mutants, mainly assembled into tetramers, exhibited wild type rescue activity, whereas a mutant, predominantly composed of dimers, had impaired functionality. These findings indicate that protein oligomerization is essential for PTX3 activity within the cumulus matrix and implicate PTX3 tetramers as the functional molecular units required for cumulus matrix organization and stabilization.

PTX3 is an acute phase glycoprotein that plays key roles in resistance to certain pathogens and in female fertility. PTX3 exerts its functions by interacting with a number of structurally unrelated molecules, a capacity that is likely to rely on its complex multimeric structure stabilized by interchain disulfide bonds. In this study, PAGE analyses performed under both native and denaturing conditions indicated that human recombinant PTX3 is mainly composed of covalently linked octamers. The network of disulfide bonds supporting this octameric assembly was resolved by mass spectrometry and Cys to Ser site-directed mutagenesis. Here we report that cysteine residues at positions 47, 49, and 103 in the N-terminal domain form three symmetric interchain disulfide bonds stabilizing four protein subunits in a tetrameric arrangement. Additional interchain disulfide bonds formed by the C-terminal domain cysteines Cys 317 and Cys 318 are responsible for linking the PTX3 tetramers into octamers. We also identified three intrachain disulfide bonds within the C-terminal domain that we used as structural constraints to build a new three-dimensional model for this domain. Previously it has been shown that PTX3 is a key component of the cumulus oophorus extracellular matrix, which forms around the oocyte prior to ovulation, because cumuli from PTX3 ؊/؊ mice show defective matrix organization. Recombinant PTX3 is able to restore the normal phenotype ex vivo in cumuli from PTX3 ؊/؊ mice. Here we demonstrate that PTX3 Cys to Ser mutants, mainly assembled into tetramers, exhibited wild type rescue activity, whereas a mutant, predominantly composed of dimers, had impaired functionality. These findings indicate that protein oligomerization is essential for PTX3 activity within the cumulus matrix and implicate PTX3 tetramers as the functional molecular units required for cumulus matrix organization and stabilization.
The long pentraxin PTX3 is an acute phase glycoprotein induced in a variety of somatic and natural immunity cells by primary inflammatory stimuli (e.g. Toll-like receptor engagement, interleukin-1␤, tumor necrosis factor-␣, and interleukin-10) (1)(2)(3)(4)(5). Studies performed on Ptx3-null (Ptx3 Ϫ/Ϫ ) mice have shown that this molecule has a number of functions in vivo. PTX3 acts as a soluble pattern recognition receptor with a nonredundant protective role against selected pathogens, mainly the opportunistic fungus Aspergillus fumigatus (6,7). PTX3 is also essential for correct assembly of the viscoelastic hyaluronan (HA) 4 -rich matrix surrounding the oocyte in the preovulatory follicle (namely, the cumulus oophorus). In fact Ptx3 Ϫ/Ϫ mice show severe female subfertility because of defective cumulus organization (8,9). A number of PTX3 ligands have been described so far, suggesting different physiological and/or pathological roles for this protein. PTX3 binds with high affinity to the complement component C1q thus activating the complement system through the classical pathway (10,11). The selective recognition of fibroblast growth factor-2 (FGF-2) by PTX3 inhibits FGF-2 angiogenic activity on endothelial cells (12) and blocks the autocrine and paracrine stimulation exerted by FGF-2 on smooth muscle cells (13). Previous studies on Ptx3 Ϫ/Ϫ mice suggested that PTX3 acts as a central "node" in cumulus matrix organization by establishing multivalent contacts to the HA-binding protein tumor necrosis factor-stimulated gene-6 protein (TSG-6) (9,14,15) and the serum protein inter-␣-inhibitor (16). PTX3 also plays a role in regulating the scavenger activity of macrophages and dendritic cells by direct binding to apoptotic cells (17)(18)(19)(20).
This broad binding capacity is likely to be due to the structural complexity of the protein as compared with the classical short pentraxins, C-reactive protein (CRP) and serum amyloid P component (SAP) (21). Like other members of the long pentraxin subfamily, human PTX3 is composed of a pentraxin-like C-terminal domain (amino acids 179 -381), which is homologous to the entire short pentraxin amino acid sequence, and an unrelated N-terminal region (amino acids 18 -178) (22,23). The human protein has a unique N-glycosylation site at Asn 220 that has been described to be fully occupied by complex type oligosaccharides, mainly fucosylated and sialylated biantennary sugars (24). PTX3 glycosylation status affects the protein interaction with C1q thus modulating PTX3-mediated activation of the complement classical pathway (24). Studies performed with recombinant isolated forms of both PTX3 C-and N-terminal domains provided preliminary information on regions of the molecule involved in ligands recognition. The C1q-binding site appears to be localized in the C-terminal domain as expected from the known interaction of both CRP and SAP with C1q (23,(25)(26)(27)(28). Structural determinants of the interaction with inter-␣-inhibitor and FGF-2 are localized in the N-terminal domain (16,29).
In addition to the multidomain organization, PTX3 shows a complex quaternary structure with subunits assembled into high order oligomers stabilized by disulfide bonds as demonstrated by electrophoretic and chromatographic analyses (23). Emerging evidence suggests that both short and long pentraxins express distinct bioactivities depending on subunit organization. For example, native CRP inhibits platelet activation and prevents platelet capture of neutrophils, whereas monomeric CRP, resulting from loss of the pentameric symmetry of the molecule, displays potent prothrombotic activities (30,31). Moreover disruption of the pentameric structure, as achieved by urea treatment or by site-directed mutagenesis, leads to enhanced CRP binding to C1q and subsequent C1 activation as well as the ability of CRP to interact with complement-regulatory proteins factor H and C4b-binding protein (32). Neuronal pentraxin 2 (NP2), also known as neuronal activity-regulated pentraxin (Narp), and neuronal pentraxin 1 (NP1), two members of the long pentraxin subfamily selectively expressed in brain, are covalently linked by disulfide bonds into highly organized complexes (33,34). The number of protein molecules in these complexes is dynamically dependent upon the activity history of the neurons and the brain developmental stage (35). However, little is known about the effect of protein oligomerization on PTX3 biological activity. It has been shown that the recombinant isolated PTX3 C-terminal domain, expressed as a monomeric species, is not able to bind C1q unless polymerized through chemical cross-linking, thus suggesting a role for the protein multimer in complement recognition possibly dependent on avidity and/or cooperativity mechanisms (23).
Short pentraxins show typical pentameric symmetry with protomer subunits held together by non-covalent interactions (21). PTX3 C-terminal domain three-dimensional structures have been elaborated by homology modeling based on human SAP (Protein Data Bank code 1sac) and CRP (Protein Data Bank code 1b09) crystal structures (22, 24, 36 -38). Most interestingly, the pattern of amino acid residues located at the protomer interface in SAP pentamers is not conserved in PTX3, suggesting that PTX3 quaternary assembly might not fit the characteristic short pentraxin pentameric arrangement (38). In contrast to CRP and SAP, PTX3 oligomers are stabilized by covalent bonds (23). The mature protein contains nine cysteine residues: three are located within the N-terminal region (Cys 47 , Cys 49 , and Cys 103 ), and six are in the C-terminal domain (Cys 179 , Cys 210 , Cys 271 , Cys 317 , Cys 318 , and Cys 357 ). Cys residues at positions 210 and 271 are highly conserved among pentraxins and based on the homology with CRP and SAP are predicted to be engaged in an intrachain disulfide bond (22,23).
In this study we resolved the recombinant PTX3 oligomeric assembly, determining the number of subunits forming the functional protein. Both SDS-PAGE, performed under non-reducing conditions, and native-PAGE, carried out according to the Ferguson plot method, showed PTX3 to form mainly covalent octamers. We then identified the network of disulfide bonds stabilizing PTX3 multimers by exploiting two complementary approaches: mass spectrometry and Cys to Ser sitedirected mutagenesis. Based on data from both techniques a topological map of PTX3 interchain disulfide bonds was drawn where the protein molecule is represented as a covalent octamer composed of two equivalent tetramers. The contribution of disulfide bonds to protein oligomerization was assessed by PAGE analysis under non-denaturing conditions. The new structural constraints were used to build a refined model of the PTX3 C-terminal domain.
A previous study showed that cumuli from Ptx3-deficient mice are unable to organize the glycosaminoglycan HA in a stable matrix and that exogenously added recombinant PTX3 rescues defective cumulus expansion ex vivo (9,16). We report here that the rescue activity of PTX3 mutants, forming tetramers instead of octamers, was comparable to that of the wt protein. PTX3 mutants, lacking the entire set of interchain disulfide bonds and mainly composed of dimers, proved less effective than the wt protein in rescuing defective cumulus expansion. PTX3 oligomerization is thus essential to protein functionality within the cumulus matrix. Moreover the two tetramers composing the native PTX3 octamer appear to act as functional units in cumulus matrix organization. These findings showed that PTX3 multimeric status plays a key role in defining the biological activity of the protein.

EXPERIMENTAL PROCEDURES
Purified Proteins and Antibodies-Human recombinant PTX3 was purified from a CHO 3.5 cell line stably and constitutively expressing the protein as described previously (39). Molecular weight calibrants (thyroglobulin, ferritin, catalase, lactate dehydrogenase, and bovine serum albumin) were purchased from GE Healthcare. Biotin-labeled rabbit anti-human PTX3 polyclonal antibody ␣PTX3pb and rat anti-human PTX3 monoclonal antibody MNB4 were obtained from "Mario Negri" Pharmacological Research Institute, Milan, Italy. Horseradish peroxidase-conjugated streptavidin was from GE Healthcare, and rabbit anti-rat IgG-horseradish peroxidase conjugate was from Dako, Glostrup, Denmark.
Enzymes and Chemicals-Proteomics grade trypsin (from porcine pancreas), ␣-chymotrypsin (N ␣ -p-tosyl-L-lysine chloromethyl ketone-treated from bovine pancreas), endoproteinase Asp-N (from Pseudomonas fragi mutant strain), and endoproteinase Glu-C (from Staphylococcus aureus V8) were purchased from Sigma-Aldrich. DNA Taq polymerase and restriction enzymes applied in molecular cloning were from Takara Bio, Otsu, Japan. Tris acetate 3-8% gradient gels were obtained from Invitrogen. Other electrophoresis reagents and nonfat dry milk were from Bio-Rad. Hybond-C Extra nitrocellulose membranes and ECL Plus Western blotting detection reagents were from GE Healthcare. Phosphate-buffered saline tablets, Tris base, NH 4 HCO 3 , NaCl, Tween 20, dithiothreitol (DTT), iodoacetamide, and ␣-cyano-4-hydroxycinnamic acid were purchased from Sigma-Aldrich. All other reagents and solvents were from Carlo Erba, Rodano, Italy and were of the highest purity available.
Ferguson Plot-Aliquots of purified recombinant PTX3 (5 g/lane) were loaded on homogeneous polyacrylamide gels cast at 5.0, 5.5, 6.0, and 6.5% (w/v) total acrylamide concentration (% T) in the absence of SDS. Runs were performed under non-denaturing conditions in 50 mM Tris-HCl, 200 mM glycine, pH 8.3, at 150-V constant voltage for 1 h, and proteins were detected by Coomassie staining. Thyroglobulin (669 kDa), ferritin (440 kDa), catalase (232 kDa), lactate dehydrogenase (140 kDa), and bovine serum albumin (67 kDa) were used as molecular mass calibrants. Protein mobility was analyzed by Ferguson plot as described previously (40). Briefly the relationship between protein mobility () and sieving matrix concentration (% T) may be written as where 0 is the free solution mobility (zero concentration of sieving matrix) and K r is the retardation coefficient (which is directly related to protein shape and molecular weight). A standard curve (namely the Ferguson plot) was constructed by plotting the logarithm of calibrant relative mobility (log 10 R f ) as a function of total polyacrylamide concentration (% T). Retardation coefficients (K r ) were then computed from the plot slope and graphed versus molecular mass (K r plot). PTX3 samples were processed in the same way as the calibrants, and their molecular mass was finally calculated by K r plot interpolation.
In a separate set of experiments, electrophoresis mobility data for Ferguson plot analysis of purified recombinant PTX3 were derived from Western blots (WBs). Briefly aliquots of protein (20 ng/lane) were loaded on native gels and run under the same experimental conditions as above. Gels were then blotted onto nitrocellulose membranes, revealed by immunodetection (see below), and submitted to Ferguson plot analysis as described above.
SDS-PAGE and WB-Aliquots of purified recombinant PTX3 (either 2 g for Coomassie staining or 20 ng for immunodetection) were resolved on Tris acetate 3-8% gradient or Tris-glycine 9% homogeneous gels under denaturing conditions in the presence or absence of DTT. Following electrophoresis, proteins were either revealed by Coomassie staining or transferred to Hybond-C Extra nitrocellulose membranes for subsequent immunodetection. Blotted membranes were blocked with 5% nonfat dry milk in Tris-buffered saline containing 0.05% (v/v) Tween 20 (blocking buffer) for 2 h at room temperature and then incubated overnight at 4°C with a 1:2,000 MNB4 dilution in blocking buffer. This antibody recognizes a linear epitope mapping in the N-terminal domain of human PTX3 (29). Following an additional incubation with rabbit anti-rat IgG-horseradish peroxidase conjugate, membranes were finally developed with ECL Plus according to the manufacturer's instructions.
Prior to MALDI-MS analysis, peptide mixtures from endoproteinase digestions were desalted by reversed phase chromatography on an Agilent 1200 HPLC system (Agilent Technologies, Santa Clara, CA). Briefly peptide samples were loaded on a Phenomenex Jupiter C 18 column (500 m; 2.0-mm inner diameter ϫ 15-cm length; particle size, 300 Å; Phenomenex, Torrance, CA) equilibrated by a 95% solvent A, 5% solvent B mixture (solvent A: 0.1% (v/v) trifluoroacetic acid; solvent B: 95% (v/v) acetonitrile, 0.1% (v/v) trifluoroacetic acid). Following an isocratic washing step at 5% solvent B for 10 min, a fast linear gradient was applied from 5 to 95% solvent B over 1 min, and peptide mixtures were eluted at 95% solvent B. The flow rate was set to 0.2 ml/min, and peptide elution was monitored as UV absorbance at both 216 and 280 nm. Peptide mixtures, eluted as a single peak, were manually collected and immediately freezedried. Positive ion MALDI-MS analyses of PTX3 peptides were carried out on a Bruker Daltonics Reflex II TOF/TOF MALDI spectrometer (Bruker Daltonics, Billerica, MA) operating in reflectron mode. The MALDI matrix was prepared by dissolving 10 mg of ␣-cyano-4-hydroxycinnamic acid in 1 ml of acetonitrile, 0.2% (v/v) trifluoroacetic acid (70:30, v/v). Typically 1 l of matrix was applied to the metallic sample plate, and 1 l of analyte was then added. Acceleration and reflector voltages were set as follows: target voltage at 20 kV, first grid at 95% of target voltage, and delayed extraction at 200 ns. Raw data were analyzed using computer software provided by the manufacturer and are reported as monoisotopic masses.
LC/ESI-MS/MS-PTX3 peptides from protease digestions were also analyzed on a SpectraSystem HPLC system (Thermo Fisher Scientific, Waltham, MA) equipped with a P6000 pump and a UV6000LP UV-visible detector connected to the ESI probe of an LCQ Duo ion trap mass spectrometer (Thermo-Finnigan, San Jose, CA). Peptide samples were loaded on a Supelco Discovery BIO wide pore C 18 column (500 m; 2.1-mm inner diameter ϫ 15-cm length; particle size, 300 Å; Sigma-Aldrich) equilibrated by a 97% solvent A, 3% solvent B mixture. A linear gradient was then developed from 3 to 58% solvent B over 55 min. Tryptic mixtures were chromatographed on a shallower linear gradient ramping from 3 to 58% solvent B over 80 min. In both cases, the flow rate was set to 0.2 ml/min followed by a splitting ratio of ϳ1:20. Peptide elution was monitored as both UV absorbance at 216 and 280 nm and total ion current. A spray voltage of 4.5 kV, capillary temperature of 280°C, capillary voltage of 39 V, and tube lens offset of 10 V were used. A triple play (full scan, ZoomScan, and CID with 30 -40% collisional energy) data-dependent acquisition method with dynamic exclusion was performed using the Xcalibur software (ThermoFinnigan). Spectra were processed using the base peak ion chromatogram and a seven-point Gaussian smoothing function.
Cell Culture and Protein Expression-CHO cells (BD Biosciences) were transiently transfected with uncleaved plasmid DNA using Lipofectamine reagent (Invitrogen). Following transformation, cells were incubated for 48 h in serum-free medium (Dulbecco's modified Eagle's medium, Sigma-Aldrich) in a humidified incubator gassed with 5% CO 2 , 95% air at 37°C. Cell culture supernatants were then collected by centrifugation at 1,500 rpm for 10 min and stored at Ϫ80°C. The protein concentration in culture supernatants was measured by a sandwich enzyme-linked immunosorbent assay based on the monoclonal antibody MNB4 and the rabbit polyclonal antibody ␣PTX3pb as reported previously (5,24). Supernatant aliquots containing 20 ng of proteins were analyzed by WB under both denaturing and non-denaturing conditions as described above. Protein bands were quantitated by densitometry analysis using Eastman Kodak Co. Gel Logic 100 software. Electrophoretic mobilities from native WB were finally used to draw Ferguson and K r plots.
Aliquots of culture supernatants from CHO cells transfected with C47S/C49S/C103S, C317S/C318S, and C47S/C49S/ C103S/C317S/C318S mutant constructs were submitted to a chromatographic purification. Briefly following a 10-fold concentration step on 100-kDa-cutoff Pellicon-Biomax 100 ultrafiltration membranes (Millipore), supernatant aliquots were loaded onto a 100-ml Q-Sepharose TM Fast Flow (GE Healthcare) packed column (5 ϫ 51 mm). The column was equilibrated previously with 50 mM Tris-HCl, 0.3 M NaCl, pH 7.00, at a linear flow rate of 76 cm/h, and the effluent was monitored by UV detection at 280 nm. Retained material was eluted using a

C47S
Sense a Serine codons are bold. Mutated bases are underlined. b Enzymes used to clone overlapping PCR products into the pSG5hPTX3 backbone.

PTX3 Disulfide Bonds
nonlinear gradient, raising the mobile phase conductivity stepwise to 42 millisiemens/cm to wash the resin and then to 60 millisiemens/cm to elute PTX3. PTX3-containing fractions (eluted at 60 millisiemens/cm) were quantitated by enzymelinked immunosorbent assay as described above and then applied to cumulus cell-oocyte complex (COC) culture and hyaluronan labeling experiments (see below). Structure Prediction-The MacroModel molecular mechanics package (42) was used to generate a PTX3 C-terminal domain three-dimensional model based on the x-ray structure of CRP (Protein Data Bank code 1b09) with which it has 26.8% identity (42.7% similarity). Intrachain disulfide bonds revealed by mass spectrometry and site-directed mutagenesis were integrated into the model, and structure optimization was performed by molecular dynamics (optimized potential for liquid simulation-all atom force field in water, 200 ps at 300 K, using a 300 kJ mol Ϫ1 Å Ϫ2 constraint on ␤-strand and ␣-helical motives). Twenty structures were extracted and minimized using steepest descent and truncated Newton conjugate gradient algorithms up to a convergence criterion of 5 ϫ 10 Ϫ3 kJ mol Ϫ 1 Å Ϫ1 . The best solution in terms of percentage of amino acids in the most favored and allowed regions (Ramachandran plots) was chosen using the utility Procheck (43).
COC Culture and Hyaluronan Labeling-Adult Sv129 wild type and Ptx3-deficient mice (8)  . At the start of the cultures, wild type recombinant PTX3 or PTX3 mutants were added to Ptx3 Ϫ/Ϫ COCs at the concentrations indicated in the text. Medium and cell matrix were collected separately, and the amount of HA in the two compartments was determined as described elsewhere (9).

PTX3 Molecular Weight
Determination-PTX3 is a multimeric glycoprotein composed of identical subunits held together by interprotomer disulfide bonds (23). The number of protomers forming the functional molecule has not yet been clearly defined. To address this point we measured PTX3 molecular weight by PAGE under both native and denaturing conditions. Samples of human recombinant PTX3 expressed in CHO cells were applied onto polyacrylamide continuous gels cast at four different acrylamide concentrations (5.0, 5.5, 6.0, and 6.5% T) and resolved under native conditions. Fig. 1A shows the 5.5% gel as representative of the observed electrophoretic profiles. By Coomassie staining (lane 1) a characteristic pattern was revealed with one major band (a) running between catalase (232 kDa) and ferritin (440 kDa), two minor bands (b and c) migrating between ferritin and thyroglobulin (669 kDa), and an additional signal above thyroglobulin standard (d). An identical profile was observed following gel blotting onto a nitrocellulose filter and PTX3 detection by monoclonal antibody MNB4 (lane 2), thus confirming that the species revealed by Coomassie staining were indeed PTX3 oligomers. Data from native gels were then submitted to Ferguson plot analysis (40). Log 10 relative mobility (log 10 R f ) was plotted versus gel concentration (% T) for both PTX3 and protein standards. Retardation coefficients (K r ) were computed from the slope of the respective plots and graphed versus molecular mass (K r plot). Finally the molecular mass of bands a, b, and c was calculated by K r plot interpolation and shown to be 336, 422, and 514 kDa, respectively (Fig. 1B). In a previous work, we characterized the recombinant human PTX3 glycosidic moiety and showed it to be mainly composed of monofucosylated disialylated biantennary complex type sugars with minor monofucosylated triantennary and tetra-antennary structures exhibiting different degrees of sialylation (24). According to this analysis, the PTX3 glycosidic moiety accounts for a contribution to the protein monomer molecular mass of about 2.5 kDa. As PTX3 polypeptide backbone has an average molecular mass of 40.165 Da (calculated from the coding DNA sequence), we assumed for the protein subunits an average molecular mass of 42.5 kDa. On the basis of this assumption, the major band (a) could be concluded to correspond to protein octamers (336/42.5 kDa, ϳ8), whereas the minor bands (b and c) originated from PTX3 decamers (422/ 42.5 kDa, ϳ10) and dodecamers (514/42.5 kDa, ϳ12), respectively. We estimated for band d a molecular mass of about 700 kDa, which is close to the molecular mass expected for PTX3 16-mers (680 kDa). However, it should be pointed out that this value is only approximate as the corresponding K r value was outside of the K r plot linearity range.
SDS-PAGE analysis, performed on Tris acetate 3-8% gradient gels under non-reducing conditions (Fig. 1C, lane 1), revealed a major band with an apparent molecular mass of 340 kDa, consistent with PTX3 protomers being mainly assembled into octamers and furthermore indicating that protein octamers are likely to be stabilized by disulfide bonds. No signals were detected at higher molecular weight, suggesting that bands b, c, and d from native gels correspond to PTX3 higher order oligomers held together through non-covalent interactions. A minor band was observed at 280 kDa, about 6.6 times the PTX3 protomer mass. This band probably originates from not fully denatured PTX3 octamers and/or heptamers, which retain some secondary structure and, as a consequence, show an aberrant high electrophoretic mobility. After blotting and antibody detection, several additional bands were seen at 210, 172, and 86 kDa that are likely to correspond to PTX3 pentamers, tetramers, and dimers, respectively (Fig. 1C, lane 2). The species with an apparent molecular mass of 63 kDa may correspond to fast migrating protein dimers, which retain some secondary structure, as suggested above for the signal at 280 kDa. Upon reduction, only one band at 43 kDa (close to the expected molecular mass for PTX3 monomers) was observed following either Coomassie staining (Fig. 1C, lane 3) or Western blotting (Fig. 1C, lane 4), indicating that the high molecular weight species detected under non-reducing conditions are indeed PTX3 homo-oligomers and do not contain any other protein.
Mass Spectrometry-As described above, electrophoresis analyses performed under both native and denaturing conditions revealed that recombinant human PTX3 is secreted as a mixture of oligomers the most abundant of which are likely to be covalent octamers. To characterize the network of disulfide bonds stabilizing the protein octameric assembly, PTX3 samples were subjected to proteolysis by incubation with sequence-specific endoproteinases (trypsin, Glu-C, Asp-N, and ␣-chymotrypsin), and the resulting peptide mixtures were analyzed by both MALDI-MS and LC/ESI-MS/MS both in presence and in absence of reducing agents.
Peptides were assigned to the PTX3 amino acid sequence on the basis of measured molecular masses and enzyme specificity. To verify sequence assignments, pseudomolecular cations of relevant peptides were selected as precursor ions and submitted to CID MS/MS experiments performed in the on-line LC/ESI-MS/MS mode. As shown in Fig. 2, assigned peptides covered 94.2% of the mature PTX3 amino acid sequence. PTX3 encodes nine cysteines: three in the N-terminal region and six in the C-terminal pentraxin domain (23). Fragments were detected containing every cysteine apart from Cys 47 probably because peptides containing Cys 47 have low ionization efficiency and/or low solubility under the LC conditions exploited. No peptide was detected covering Asn 220 when endoproteinase incubations were performed on glycosylated PTX3. As stated above, we recently described Asn 220 as a unique N-glycosylation site in PTX3 and showed this site to be fully occupied in the mature secreted protein (24). PTX3 deglycosylation by peptide-N-glycosidase F treatment, in fact, enabled detection of additional peptides spanning Asn 220 thus bringing sequence coverage to 96.2%.
A list of the observed cysteine-containing peptides is reported in Table 2. MS analyses performed on Asp-N digests under both reducing and non-reducing conditions demonstrated Cys 49 to be involved in interprotomer disulfide bonds, i.e. pairs of PTX3 subunits are linked together through Cys 49 -Cys 49 bonds (Table 2 and supplemental Fig. S1). We consistently observed homodimeric peptides spanning Cys 103 , which showed that Cys 103 also forms interprotomer disulfide bonds ( Table 2 and Fig. 3A). From these data it is not possible to establish whether Cys 103 connects PTX3 subunits already cross-linked by Cys 49 or other subunits. Two covalent bonds were described bridging Cys 210 to Cys 271 (Table 2 and Fig. 3) and Cys 179 to Cys 357 (Table 2 and supplemental Fig. S2). These could not be unambiguously sorted into either inter-or intrachain disulfide bonds. However, it has to be noted that Cys 210 and Cys 271 are highly conserved among members of the pentraxins family, and these residues have been described to establish an intrachain disulfide bond essential for pentraxin fold stabilization (22,23). Thus, it is likely that also in PTX3 Cys 210 and Cys 271 form intrasubunit linkages. Moreover Cys 179 and Cys 357 are conserved among the long neuronal pentraxins Narp and NP1 (33,34). Some recent evidence showed that corresponding residues in these proteins do not participate in the formation of multimers (35), strongly suggesting that the disulfide bond Cys 179 -Cys 357 described here is formed within rather than between PTX3 subunits.
MS analysis of trypsin and Glu-C digests provided information on the redox state of cysteines 317 and 318 ( Table 2). As representative of the obtained results, MALDI and CID ESI-MS/MS spectra of PTX3 tryptic mixtures are shown in Fig. 3. MALDI analysis performed under non-reducing conditions revealed a signal at 3574.48 Da (Fig. 3A, upper panel and inset on the right) that was attributed to the homodimeric peptide (315-332)-(315-332) where Cys 317 and Cys 318 were both involved in disulfide bond formation (theoretical monoisotopic molecular mass of 3574.53 Da). Signal assignment was verified by MALDI analysis of DTT-reduced tryptic digests. Following reduction, in fact, the peak at 3574.48 Da disappeared (Fig. 3A, lower panel and inset on the right), whereas a new peak was detected at 1789.79 Da (Fig. 3A, lower panel and inset on the left), matching the theoretical monoisotopic molecular mass of peptide 315-332 in its reduced form (namely 1789.76 Da). We could not establish by MS/MS whether the two bonds were symmetrical (Cys 317 -Cys 317 and Cys 318 -Cys 318 ) or asymmetrical (Cys 317 -Cys 318 and Cys 318 -Cys 317 ) as fragmentation of the dimeric peptide (315-332)-(315-332) or the corresponding species observed in Glu-C digests (see Table 2) preferentially occurred at the disulfide bonds. It is worth noting that no free sulfhydryls were detected in the recombinant PTX3 by 5,5Јdithiobis(nitrobenzoic acid) assay (see below), so the occurrence of (315-332)-(315-332) peptides with one cysteine engaged in a disulfide bond and the other free is very unlikely. Moreover MALDI MS analysis performed under non-reducing conditions showed a signal at 1787.79 Da (Fig. 3A, upper panel and inset on the left). As this value is 2 Da less than the value calculated for peptide 315-332, it likely originates from peptide 315-332 upon formation of an interchain disulfide bond that causes the loss of two protons. Consistent with this view, DTT reduction induced a 2-Da mass shift (1789.79 Da) where the original peak at 1787.79 Da disappeared (Fig. 3A, lower panel and inset on the left). Taken together, these data showed a dual role for Cys 317 and Cys 318 as they are capable of forming both inter-and intrachain disulfide bonds.
Site-directed Mutagenesis-Mass spectrometry identified PTX3 cysteine residues engaged in disulfide bonds. To understand how these bonds are distributed within the subunits forming protein oligomers, PTX3 cDNA sequence was subjected to site-directed mutagenesis. Cysteine codons at amino acid positions 47, 49, 103, 317, and 318 were replaced with serine codons to give both single and multiple Cys to Ser PTX3 mutant constructs combined as follows: C47S, C49S, C103S, C47S/C49S, C47S/C49S/C103S, C317S/C318S, and C47S/ C49S/C103S/C317S/C318S. The cDNA constructs were then used to transiently transfect CHO cells. Mutant proteins, expressed as secretion products into cell culture supernatants, were quantitated by enzyme-linked immunosorbent assay and finally analyzed by Western blot under non-reducing conditions. Supernatants from wt cDNA-and mock-transfected cells were used as positive and negative controls, respectively. wt PTX3 showed bands at 340, 280, 210, 172, 86, and 63 kDa as described for the purified protein (Fig. 4A, upper panel, as compared with Fig. 1C, lane 2), thus demonstrating that PTX3 expression was effective under the applied transfection conditions. Mock-transfected cells did not produce any PTX3 as expected. C47S and C49S mutants exhibited electrophoresis profiles (Fig. 4A, upper panel) very similar to those observed for wt PTX3. However, an additional band at 128 kDa was detected that likely corresponds to PTX3 trimers (128/42.5 kDa, ϳ3). Two more bands were observed at 250 and 195 kDa, the former probably originating from PTX3 hexamers (250/42.5 kDa, ϳ6). Finally the signal at 210 kDa appeared more prominent as com-pared with that described for wt PTX3. The close similarity of the band patterns for C47S and C49S to that of the wt PTX3 indicates that Cys 47 and Cys 49 exert a redundant role in PTX3 covalent oligomerization. Cys 49 was shown by MS to form interchain symmetric disulfide bonds. From the overall analysis of our MS data, we concluded that also Cys 47 establishes interchain symmetric disulfide bridges (see above). Site-directed mutagenesis results shown here are consistent with this and furthermore suggest that Cys 47 -Cy 47 and Cys 49 -Cys 49 disulfide bonds occur within the same pairs of PTX3 subunits. As can be seen from Fig. 4, the double mutant C47S/C49S had greatly reduced ability to form octamers because the band at 340 kDa was barely detectable (Fig. 4A, upper panel). The other bands given by C47S/C49S showed the same electrophoretic mobility as those from single mutants C47S and C49S. The bands at 172 and 128 kDa appeared more intense. One additional signal was faintly detected at 43 kDa, corresponding to PTX3 monomers. These results thus indicated that Cys 47 and Cys 49 are involved in PTX3 multimer formation.
We extended our investigation to Cys 103 and observed that the single mutant C103S gave a band pattern shifted toward lower order oligomers as compared with those from wt PTX3 and C47S/C49S mutants. Similarly to C47S/C49S, C103S exhibited a band at 43 kDa and a low intensity signal at 340 kDa (Fig. 4A, upper panel). Additional bands were found at 260, 240, 200, and 190 kDa, which did not run at the same positions as any of those described above. Moreover the diffuse band at 172 kDa split in two distinct signals at 180 and 160 kDa. The C103S electrophoresis profile indicates that mutants lacking Cys 103 -Cys 103 interchain disulfide bonds form oligomers with electrophoretic properties different from those exhibited by wt PTX3 and C47S/C49S. This evidence is consistent with mass spectrometry data reporting Cys 103 as engaged in symmetric inter-  Table 2). Reported molecular mass values are expressed as monoisotopic mass. Asterisks refer to contaminants. B, non-reduced (ϪDTT) and reduced (ϩDTT) PTX3 tryptic digests were resolved by reversed phase HPLC and revealed by ESI/ion trap MS/MS. Pseudomolecular [MH] ϩ cations at m/z 1787.73 and 1789.82, detected in non-reduced and reduced samples, respectively, were selected as precursor ions and submitted to CID fragmentation with 30% collisional energy. Peaks in both spectra are labeled with y ions based on mass matches to predicted fragments. Ladder diagrams with y ion numbering are shown. a.u., absorbance unit. chain disulfide bonds and confirmed a role for this residue in PTX3 covalent oligomerization.
The triple mutant C47S/C49S/C103S resolved as two major bands, an upper signal at 86 kDa and a lower signal at 43 kDa, representing PTX3 dimers and monomers, respectively (Fig.  4B, upper panel). Although definitely confirming the involvement of the N-terminal cysteines (Cys 47 , Cys 49 , and Cys 103 ) in PTX3 multimerization, these findings demonstrate that disulfide bonds formed by these residues are required to sustain the protein assembly into tetramers. In fact, the triple mutant C47S/C49S/C103S did not exhibit any significant signal above 86 kDa. Given the structural redundancy of Cys 47 and Cys 49 as described above, it follows that Cys 103 -Cys 103 disulfide bonds must join PTX3 subunits complementary to those bound through Cys 47 and Cys 49 . Furthermore the appearance of a dimer band at 86 kDa points to the presence of additional intersubunit disulfide bonds at the level of the PTX3 C-terminal domain. Our MS results showed that cysteine residues at proximal positions 317 and 318 formed both intra-and interdisulfide bonds. Therefore they were further analyzed by mutagenesis where the C317S/C318S double mutant gave two bands on SDS-PAGE at 172 and 86 kDa consistent with PTX3 tetramers and dimers, respectively (Fig. 4B, upper panel). The absence of bands at molecular mass higher than 172 kDa gives compelling evidence that Cys 47 , Cys 49 , and Cys 103 provide the covalent framework for the assembly of PTX3 subunits into tetramers and indicates that the two interchain bonds formed by Cys 317 and Cys 318 link tetramer pairs into octamers. The quintuple mutant C47S/C49S/C103S/C317S/C318S was resolved into a cluster of bands around 43 kDa (Fig. 4B, upper panel), thus allowing the conclusion that this set of cysteines is necessary and sufficient for protein covalent oligomerization.
As expected, a 43-kDa band was detected upon reduction with DTT (Fig. 4, A and B, lower panels). A minor signal at 40 kDa was also observed that may result from oligosaccharide degradation (in particular the partial loss of terminal sialic acid residues) under the experimental conditions applied. Of note, the monomer band intensity appeared comparable among the different mutant species, thus indicating that the quantification by enzyme-linked immunosorbent assay was reliable.
A Model for PTX3 Assembly-A schematic map of the PTX3 interchain disulfide bond network is shown in Fig. 5. As described above, we demonstrated that PTX3 is predominantly organized into covalent octamers, so the protein is here represented as composed of eight subunits. Pairs of subunits are joined together by disulfide bonds involving Cys 47 and Cys 48 . The resulting dimers are then assembled into tetramers through covalent linkages provided by Cys 103 . MS analysis revealed Cys 317 and Cys 318 to form both intra-and interchain disulfide bonds. Site-directed mutagenesis confirmed their involvement in intersubunit bridges. Accordingly in our model both intra-and interchain disulfide bonds were drawn at sites of Cys 317 and Cys 318 . We could not establish whether Cys 317 and Cys 318 are engaged in symmetric or asymmetric disulfide bonds. Nevertheless symmetric bonds are drawn in our model for simplicity. Because both mutants C103S and C47S/C49S exhibited higher oligomers than tetramers (Fig. 4A, upper  panel), we hypothesized that six Cys 317/318 -Cys 317/318 inter-chain and two Cys 317 -Cys 318 intrachain disulfide bonds occur in the molecule. Thus it is possible for PTX3 to form oligomers up to octamers when Cys 103 -Cys 103 , Cys 47 -Cy 47 , and Cys 49 -Cys 49 are missing. However, as Cys 47/49 and Cys 103 mutagenesis greatly reduced the octamer band intensity while increasing the strength of the hexamer, tetramer, and dimer signals, it is likely that the protein assembles into octamers containing not only six but also four or two Cys 317/318 -Cys 317/318 interchain disulfide bonds (see supplemental Fig. S3).
Analysis of PTX3 Mutants by Native PAGE-To further assess the contribution provided by disulfide bonds to protein oligomerization, we submitted PTX3 Cys to Ser mutants to WB analysis under native conditions. Data from gels were then used to graph Ferguson plots as described above. As apparent in Fig.  6, showing the gel at 5.5%, band patterns exhibited by wt PTX3 and C47S/C49S/C103S triple mutant were essentially identical (Fig. 6A, lanes 1 and 2). The major band in C47S/C49S/C103S electrophoretic profile migrated at an apparent molecular mass of 336 kDa, corresponding to PTX3 octamers. This indicates that PTX3 protomers are able to self-assemble into high order oligomers regardless of the interchain disulfide bonds formed by the N-terminal domain cysteine residues.
The double mutant C317S/C318S showed a main signal at 170 kDa, consistent with PTX3 tetramers, and a minor band at 336 kDa (Fig. 6A, lane 3), demonstrating that, despite a clear impairment in PTX3 octameric assembly, the protein retains the ability to form tetramers when the two linkages established by Cys 317 and Cys 318 are missing. Therefore, these disulfide bonds appear to play an essential role in covalent dimerization of PTX3 tetramers to give octamers. The quintuple mutant C47S/C49S/C103S/C317S/C318S migrated on native gels mainly as a dimer (Fig. 6A, lane 4), suggesting that the N-terminal Cys 47 , Cys 49 , and Cys 103 and the C-terminal Cys 317 and Cys 318 provide interdependent contributions to the stabilization of protein oligomers. Moreover the appearance of minor tetramer and octamer bands indicates a residual PTX3 self-assembly ability even in the absence of the entire set of interchain disulfide bonds. Densitometry analysis indicated relative intensities of about 70, 20, and 10% for dimer, tetramer, and octamer bands, respectively, of the quintuple mutant C47S/C49S/C103S/C317S/C318S (data not shown).
Rescue of Defective Cumulus Expansion by PTX3 Mutants-To investigate the effect of PTX3 oligomeric status on the protein biological activity, C47S/C49S/C103S, C317S/C318S, and C47S/C49S/C103S/C317S/C318S mutants were assayed for their ability to rescue normal phenotype in COCs from Ptx3null mice. It has been shown that COCs from Ptx3-null mice, stimulated in vitro for 16 h with hormones, become disorganized as a consequence of a defective HA organization into the viscoelastic matrix holding together cumulus cells and oocyte (8). This phenomenon can be reverted by supplementing Ptx3 Ϫ/Ϫ COC cultures with recombinant PTX3 (9). Recent findings suggest that PTX3 stabilization of cumulus matrix might depend on the protein multimeric arrangement (9,16). Thus, we evaluated Ptx3 Ϫ/Ϫ COC morphology and HA distribution between medium and matrix after 16-h hormone treatment in the presence of increasing concentrations of different PTX3 mutants and wt protein. C47S/C49S/C103S and C317S/ C318S, forming mainly octamers and tetramers, respectively, showed the same potency as the wt PTX3, allowing maximum HA retention within the matrix and COC integrity at the minimum applied dose of 44 nM (Fig. 7, A and B). Conversely the quintuple mutant C47S/C49S/C103S/C317S/C318S, forming mainly dimers, proved to be about 3 times less active because the normal phenotype of Ptx3 Ϫ/Ϫ COCs was rescued only at the dose of 132 nM. This activity tightly correlates with the overall proportion of tetramers and octamers (compared with dimers) in the quintuple mutant (which make up 30% of the total protein; see above). These results suggest that the minimum oligomeric status required for PTX3 to exert its role in COC matrix organization is the tetramer.  Three-dimensional Model of PTX3 C-terminal Domain-We previously built by homology modeling a three-dimensional model of the PTX3 C-terminal domain using as a template the crystallographic structure of CRP (Protein Data Bank code 1b09) (24) (Fig. 8A). This model contains a disulfide bond, involving Cys 210 and Cys 271 , that is highly conserved among pentraxins (21,22). Both mass spectrometry and site-directed mutagenesis data shown here indicated that also Cys 179 and Cys 357 form an intrachain disulfide bond. According to our model, Cys 179 and Cys 357 are in loop regions facing each other that show the flexibility required to bring the two residues close enough to form a disulfide bond, consistent with our experimental evidence. The Cys 179 -Cys 357 disulfide bond was integrated as a new structural constraint in a refined model of the PTX3 C-terminal domain (Fig. 8B). In the new model the Cys 179 -Cys 357 disulfide bond links the N-and C-terminal ends of the PTX3 C-terminal domain, thus limiting the flexibility of these regions. Moreover we report Cys 317 and Cys 318 as engaged in both intra-and interchain disulfide bonds. These two cysteines are localized in an exposed loop, which is likely to represent part of the PTX3 regions at the protomer interface.

DISCUSSION
In this study we have provided new insights on PTX3 oligomeric structure and have begun to address the role that it plays in PTX3 function. The network of PTX3 disulfide bonds was characterized by two complementary approaches: mass spectrometry and site-directed mutagenesis. Mass spectrometry identified cysteine residues involved in disulfide bond formation, whereas PAGE analysis of PTX3 Cys to Ser mutants allowed the determination of how disulfide bonds are distributed within and between the protein subunits.
Previous studies showed that PTX3 is eluted in gel filtration chromatography with an apparent molecular mass of ϳ900 kDa. It was concluded that this corresponds to a 20-mer, whereas native PAGE revealed the protein to predominantly form multimers of ϳ440 kDa thought to correspond to 10-mers (23). It is known that for accurate determination of molecular weight by size exclusion chromatography the protein of interest must have the same relationship between molecular weight and molecular size and shape as the calibration standards (46), which are usually globular. The PTX3 molecule might not have a globular shape. If so, it would show a higher hydrodynamic radius than the one expected for a globular protein with comparable molecular weight, which might account for the molecular mass of ϳ900 kDa measured under size exclusion chromatography conditions. An additional complication might come from PTX3 glycosylation (24) because the determination of the molecular weight for glycoproteins by size exclusion chromatography may not correlate well to the calibration curves plotted for globular non-glycosylated proteins (46). Furthermore it is well established that protein mobility on native gels is a function of its charge, shape, and mass so information on protein molecular weight inferred from single native gels might sometimes be misleading (47). To overcome these limitations we determined the molecular weight of PTX3 by Ferguson plot on native PAGE, which offers the distinctive advantage of providing charge-independent molecular weight measurements (40). We reported a molecular mass of 336 kDa for the major species on native gels; this is consistent with that of a PTX3 octamer. SDS-PAGE analysis confirmed this view because a major band at an apparent molecular mass of 340 kDa was observed on Tris acetate gradient gels. During new investigations on PTX3 quaternary structure, we have recently exploited multiangle laser light scattering to determine the molecular weight of PTX3 under non-denaturing gel filtration conditions. Measurements performed by multiangle laser light scattering were consistent with data from native and denaturing PAGE. 5 Taken together, these results indicate that PTX3 subunits mainly assemble into covalently linked octamers. Although SAP from Limulus polyphemus hemolymph has been described to fold into a doubly stacked octameric ring (48), the octameric assembly here described for human PTX3 appears to be unique among the vertebrates pentraxins (21,22,33).
Western blotting revealed minor bands that can be attributed to lower order oligomers (hexamers, pentamers, tetramers, and dimers). Faint bands were detected on native gels at 422, 514, and 700 kDa (corresponding to PTX3 10-mers, 12-mers, and 16-mers, respectively); thus they may be derived from the non-covalent association of octamers and lower order oligomers, consistent with the tendency of the protein to selfassemble into higher order oligomers (23).
Here we report that the cysteine residues located in the N-terminal domain of PTX3 form disulfide bonds arranging the protein into tetramers. In this regard, Cys 47 -Cys 47 and Cys 49 -Cys 49 bonds stabilize protein dimers, pairs of which are then linked through Cys 103 -Cys 103 bonds. The resulting tetramers are finally assembled into octamers by disulfide bridges involving Cys 317 and Cys 318 (Fig. 5). In principle two Cys 317/318 -Cys 317/318 interchain bonds (Cys 317 -Cys 318 and Cys 318 -Cys 317 or Cys 317 -Cys 317 and Cys 318 -Cys 318 ) between two distinct PTX3 tetramers might be sufficient to organize PTX3 into octamers. However, as octameric species were still observed in the electrophoretic profile of mutants C103S and C47S/C49S and the band intensity of hexamers, tetramers, and dimers increased following both Cys 47/49 and Cys 103 mutagenesis (Fig.  4, upper panel), it is likely that the functional protein forms octamers containing up to six Cys 317/318 -Cys 317/318 interchain disulfide bonds (supplemental Fig. S3). Although we cannot rule out the formation of Cys 317/318 -Cys 317/318 interchain bonds within tetramers, this seems unlikely because these cysteines were described to mediate the covalent association of tetramers into octamers and as such are probably located at the interface between C-terminal domains on distinct tetramers. Cys 317/318 -Cys 317/318 interchain disulfides were described to involve both Cys 317 and Cys 318 residues. Whether these two disulfide bonds are symmetric (Cys 317 -Cys 317 and Cys 318 -Cys 318 ) or asymmetric (Cys 317 -Cys 318 and Cys 318 -Cys 317 ) remains unsolved. Mass spectrometry showed that Cys 317 and Cys 318 form intra-as well as interchain disulfide bonds. Although Cys 317 and Cys 318 are not conserved among pentraxins, two adjacent cysteines were identified at equivalent positions in the C-terminal end of rat CRP (i.e. amino acids 208 and 209) (49). These cysteines were found to be engaged either in inter-or intrachain disulfide bonds analogous to what we described here for PTX3 Cys 317 and Cys 318 (49).
A model of the PTX3 C-terminal domain was constructed by homology modeling using the crystallographic structure of CRP (Fig. 8A). In this model, the Cys 210 -Cys 271 disulfide bond stabilizes two antiparallel ␤-sheets organized in a typical ␤-jelly roll topology, equivalent to the structural role these conserved residues perform in the classical pentraxins CRP and SAP (21,22). Our mass spectrometry data indicated Cys 179 and Cys 357 to form a disulfide bond too. Results from Cys to Ser site-directed mutagenesis allowed us to infer this bond to be of the intrachain type. In the C-terminal domain three-dimensional model, Cys 179 and Cys 357 were found positioned close enough to support a covalent linkage. Experimental data thus provided evidence as to the reliability of the proposed model. We then elaborated a refined structure of the PTX3 C-terminal domain by integrating as structural constraints the newly identified Cys 179 -Cys 357 and Cys 317 -Cys 318 intrachain bonds. The overall domain fold is essentially unaffected by the new bonds (Fig. 8B). The Cys 179 -Cys 357 disulfide links the N-terminal end to the C-terminal end of the pentraxin domain thus limiting the flexibility of these two regions. The Cys 317 -Cys 318 intrachain bond is localized in an exposed loop, which is likely to represent part of the interface between protomer subunits.
We described the disulfide bond network established by the N-terminal cysteines as essential for PTX3 oligomerization. Interestingly the N-terminal domain of other long pentraxins such as the neuronal pentraxins NP1 (33) and NP2 (34) and some collectins, a family of proteins that share with PTX3 the ability to recognize microorganisms and enhance their engulfment by phagocytes (50,51), includes cysteine residues with a similar structural role, although these domains do not share sequence homology.
The N-terminal region of PTX3 is not related to any known protein sequence; nevertheless it has been predicted to be composed of four ␣-helices, which are likely to adopt a coiled coil conformation (52). Coiled coils are also found in the N-terminal domain of NP1 (33) and NP2 (34) as well as in the neck region of collectins (50,51). This structural motif is mainly found in modular proteins and exerts a key function in protein oligomerization through non-covalent interactions (53). In agreement with these structural features, our data from native PAGE analysis of PTX3 Cys to Ser mutants provided evidence that the protein N-terminal domain mainly contains the structural determinants responsible for PTX3 oligomerization. Indeed although C47S/C49S/C103S was resolved under denaturing conditions as a dimer, this mutant migrated on native gels mainly as an octamer, exhibiting a self-assembly ability even when the entire set of N-terminal disulfide bonds was missing. In contrast, the association between the C-terminal domains mostly relied on the Cys 317/318 -Cys 317/318 interchain disulfide bonds to form. C317S/C318S was, indeed, only partially able to assemble into octamers (Fig. 6).
Previous studies support the notion that PTX3 exerts its structural role within the cumulus matrix as a multimeric protein (9,16). A model for the role of PTX3 in cumulus matrix assembly has been proposed where heavy chains, transferred from inter-␣-inhibitor to HA by the catalytic activity of TSG-6, bind distinct protomers of multimeric PTX3, leading to HA cross-linking (16). Here we showed that PTX3 multimeric organization is essential to the protein activity in cumulus matrix. Furthermore we described the tetramer as the lowest PTX3 oligomer still retaining full functionality in cumulus matrix assembly and stabilization. The mutant C317S/C318S, exhibiting mainly a tetrameric assembly under native PAGE conditions, was indeed as effective as the wt protein in restoring normal cumulus expansion ex vivo. Analogous effectiveness on cumuli isolated from Ptx3 Ϫ/Ϫ mice was observed with C47S/ C49S/C103S, which preserved the octameric structure. We could not test the functionality of PTX3 monomers in cumuli as the mutant C47S/C49S/C103S/C317S/C318S lacking the entire network of interchain disulfide bonds predominantly formed dimers. On native gels about 70% of C47S/C49S/ C103S/C317S/C318S was dimer, 20% was tetramer, and the remaining 10% was octamer; these oligomeric states, which result from the non-covalent association between PTX3 monomers, are presumably in equilibrium. The correlation between the reduction in this mutant activity (2-3 times less effective than the wt protein) and the fraction of the dimeric species (about 2 ⁄ 3 of the total) suggest that the residual rescue activity is derived from the tetramers/octamers present in the sample.
The heavy chain-binding site has been localized in the PTX3 N-terminal domain (16). Accordingly a recombinant form of this domain allows normal in vitro expansion of Ptx3 Ϫ/Ϫ COCs, thus mimicking the function of the full-length protein (16). It has been reported that the isolated PTX3 N-terminal domain assembles into disulfide bond-stabilized oligomers (9,16). Here we describe that the N-terminal domains of the full-length PTX3 are composed of two structurally independent tetramers by the three symmetric disulfide bonds Cys 47 -Cys 47 , Cys 49 -Cys 49 , and Cys 103 -Cys 103 .
In previous studies it has been proposed that PTX3 acts as a node in the cumulus matrix for example by establishing direct multivalent interaction with either the hyaladherin TSG-6 (9) or the serum protein inter-␣-inhibitor (16). Here we provide evidence for the nodal activity of PTX3 in that our data indicate that the multimeric organization of PTX3 is essential to its function with the PTX3 tetramer as the functional molecular unit required for cumulus matrix organization. Further studies are in progress to determine whether the degree of oligomerization may influence other biological functions of PTX3, such as complement activation, angiogenesis, and protection against opportunistic pathogens.