Prof Francesco Tedesco, Department of Physiology and Pathology, University of Trieste, via Fleming 22, Trieste, Italy. E-mail: email@example.com
In our study we examined the early complement components in patients with bacterial vaginosis (BV), vulvovaginal candidiasis (VVC) and in healthy controls. The levels of C1q, mannose-binding lectin (MBL) and C3 were measured by ELISA in the cervicovaginal lavage (CVL) from gynaecological patients and controls. No significant differences were observed in the levels of these proteins in the three study groups. Immunofluorescence analysis of the clue cells and Candida hyphae from BV and VVC patients for surface-bound complement components showed the presence of C3, while C1q was undetectable. MBL was revealed on clue cells but not on Candida. Binding of MBL to Candida, grown or cytocentrifuged from the CVL of VVC patients, was found to be pH dependent and occurred between pH 4·5 and pH 5·5. In conclusion, we demonstrated that MBL and C3 present in the vaginal cavity act as recognition molecules for infectious agents that colonize the cervicovaginal mucosa. Our finding that MBL, but not C1q, binds to bacteria and fungi in vagina suggests that the lectin and classical pathways of complement activation may play a different role in immune defence in the female genital tract.
The vagina represents the portal of entry for several infectious agents that colonize the vaginal mucosa under permissive local conditions and cause a chronic and disabling inflammatory process. Vulvovaginal candidiasis (VVC) and bacterial vaginosis (BV) are the most frequent vaginal infections in Western Countries . VVC is caused by Candida spp., while BV is a polymicrobial infection associated with overgrowth of several anaerobes, and more frequently of Gardnerella vaginals, which releases a cytolysin highly destructive for the vaginal epithelium . The vaginal secretion of these patients contains epithelial cells, known as clue cells, which detach from the mucosa and carry surface bound bacteria. These cells represent a diagnostic marker for BV together with an increase in the pH value above 5  in contrast to the lower pH around 4 observed in patient with VVC.
Under normal circumstances, the damaging effect of these infectious agents is controlled by the intervention of the local defense system that interferes with the microbial colonization and prevents invasion of the cervicovaginal mucosa. While the protective function of antibodies to several infectious agents in the cervicovaginal fluid is well established , less clear is the role played by the complement (C) system, a humoral component of the innate immune system actively involved in the host protection against bacterial infections
C recognizes the infectious agents utilizing three different molecules, C1q, MBL and C3, that trigger the classical, lectin and alternative pathways of C activation, respectively . Babula and colleagues have recently reported the presence of MBL in nanogram amounts in the cervicovaginal lavage (CVL) of healthy women and lower levels of vaginal MBL in patients with recurrent VVC patients suggesting that the deficient level of this C component may be a predisposing factor to Candida infection . However, the question as to whether MBL interacts with the Candida spp. colonizing the vaginal mucosa remains unanswered. We have extended this observation evaluating the vaginal level of the three recognition molecules of the C system in control women and in patients with VVC and BV and examined the ability of these C components to bind to C. albicans and to G. vaginalis that cause these infections.
Materials and methods
The study groups comprise 23 healthy women with no clinical signs of cervicovaginal inflammation and with negative culture for pathogens and 47 women with gynaecological symptoms of infections, including leukorrhea, itching, burning, vulvar oedema, and dyspareunia. The Amsel criteria  were followed to diagnose BV in 22 patients, while a diagnosis of VVC was made in the remaining 25 patients, who had a history of more than three episodes of vaginal candidiasis, as ascertained by a positive culture of vaginal smear for C. albicans. The age of the control and the patient groups ranged from 18 to 40 years. Both patients and controls were followed at the Department of Obstetrics and Gynaecology of ‘Burlo Garofolo’ Hospital (Trieste, Italy) and informed consent was obtained according to the guidelines of the Bioethical Committee of IRCCS, Burlo Garofolo, Italy.
C components and antisera
Human purified C1q and C3 and goat antisera to C1q and C3 were purchased from Quidel (San Diego, CA, USA). Human recombinant MBL was prepared as previously reported . Murine monoclonal antibody to MBL (HYB 131–01) was obtained from Antibodyshop (Copenhagen, Denmark). The IgG were purified from the antisera by affinity chromatography on protein G-Sepharose columns (Pharmacia Biotech, Milan Italy) and were biotin-labelled with biotinamidocaproate-n-hydroxysuccinimide ester (Sigma-Aldrich, Milan, Italy) following a previously described procedure . Alkaline phosphatase-labelled and FITC-conjugated rabbit anti-goat IgG were purchased from Sigma and Southern Biotechnology Associates (Birmingham, AL, USA), respectively, and FITC-conjugated rabbit anti-mouse IgG from DakoCytomation (Milan, Italy).
Collection and processing of cervicovaginal lavage
CVL was obtained by instilling 10 ml of sterile saline after insertion of a speculum and the lavage collected with a syringe was centrifuged at 400 × g. The supernatant was filtered through 0·2 µm filters (Corning Inc., NY, USA), lyophilized and, after resuspension to a final volume of 2 ml, was kept in small aliquots at − 20°C until use. The pellet was resuspended in PBS containing 2% bovine serum albumin (BSA) to about 1 × 106 cells/ml and cytocentrifuged for further analysis of the clue cells and C. albicans present in the CVL.
Bacterial and fungal culture
C. albicans and G. vaginalis were collected with cotton swabs from patients with VVC and BV, respectively, and grown under appropriate cultural conditions in selective media, that included BBL™ CHROMagar™ Candida (Difco, Milan, Italy) and Gardnerella Vaginalis selective Agar (Oxoid, Milan, Italy).
DNA was extracted from peripheral whole blood following a standard procedure reported by Sambrook et al.. MBL2 exon I genotyping was performed using a melting temperature assay according to Hladnik and coworkers . As suggested by Garred et al., the three variant alleles were grouped in the same category of allele O because they are all associated with a substantial reduction in the serum level of MBL, while the wild-type allele was called A. Allele frequencies were calculated by direct gene counting and the differences analysed by the χ2 test using 2 × 2 contingency tables.
Measurement of C1q, C3 and MBL in vaginal washings
The level of C1q and C3 in the CVL was measured by ELISA on microtiter plates (Corning) as previously described  using 5 fold concentrated CVL samples. The concentration of the C components was evaluated with reference to a standard curve established with increasing amounts of purified human C1q and C3. An MBL oligomer ELISA kit (Antibodyshop) was used to quantify the level of MBL in the CVL.
Cytochemical and immunofluorescence analysis
The epithelial cells were stained with haematoxylin-eosin (Dip-quick stain Kit DYASET s.r.l., Ferrara Italy) and G. vaginalis with Giemsa staining. Deposition of C components was evaluated on cytocentrifuged vaginal epithelial cells and on C. albicans and G. vaginalis grown from vaginal smears. The Candida suspension was washed twice with PBS by centrifugation at 2000× g for 5 min and the pellet was cytocentrifuged following resuspension at 104 cells/ml. G. vaginalis was isolated from the culture plate, washed with PBS and air dried on a glass slide. Cells and bacteria were fixed with 1·5% paraformaldehyde for 15 min at room temperature (R.T.) and then incubated for 1 h at 37°C either with normal human sera (NHS) diluted 1 : 2 or with the purified C components MBL (10 µg/ml), C1q (5 µg/ml) or C3 (5 µg/ml). After two washings with 10 m m Tris-HCl containing 10 m m Ca2+, the cells were first incubated with goat IgG to C1q or C3 (5 µg/ml) or with monoclonal antibody to MBL (5 µg/ml) for 1 h at R.T. and then with FITC-labelled rabbit IgG anti-goat for C1q and C3 and FITC-labelled IgG goat anti-mouse for MBL both at a final dilution of 1 : 200. All the experiments were performed using 10 m m Tris-HCl buffer containing 1 m NaCl, 10 m m Ca2+, and 0·1% BSA and the cells were observed under the fluorescence microscope (Zeiss 250-CF Jenalumar, Jena, Germany).
Statistical significance was determined using Student's t-test to compare two groups of data. Values of P < 0·01 or less were considered to be statistically significant.
Vaginal concentrations of MBL, C1q and C3 in healthy women and in patients with VVC and BV
The CVL obtained from 25 healthy women, 20 patients with VVC and 19 patients with BV were analysed for the level of the recognition molecules of the lectin, classical, and alternative pathways of C activation. As shown in Fig. 1, all three C components were detected in the CVL of healthy women, even if in different amounts. The level of C3 ranged between 250 and 750 ng in the CVL of more than 90% of the healthy controls, while that of C1q and MBL was approximately 100 fold lower. The concentration of these C components in the CVL of patients suffering from VVC and BV was not significantly different from that of the healthy controls. It is interesting to note that three patients with VVC had a level of MBL substantially higher than that of the other patients. The increased amount of vaginal MBL in these patients was found to be related to a higher level of circulating MBL.
Since MBL was either undetectable or present at very low levels in the CVL of 36% of both healthy and BV women and in 20% of VVC patients, we sought to determine whether this was related to an increased frequency of the mutant MBL2 O allele in our study groups. The results presented in Table 1 show that about 26% of healthy women, 28% of VVC and 19% for BV patients were heterozygous for the mutant O allele, while 8% of VVC patients and 4% of BV patients were found to be homozygous for the same allele. The absence of circulating MBL in the homozygotes and the reduced level in the heterozygotes may explain our inability to detect measurable concentration of MBL in the CVL of a substantial number of both controls and patients.
Table 1. MBL2 allelic, genotype and phenotype frequencies in Italian gynaecological patients and healthy controls.
Phenotype AO + OO
P > 0·001 (ns)
Analysis of clue cells for bound MBL, C1q and C3
We evaluated the role played by the early C components of the three activation pathways in the recognition of the infectious agents associated with BV. To this end, CVL samples obtained from two BV patients were cytocentrifuged and the cells were examined by immunofluorescence for the presence of the C components on their surface. Analysis of the cells stained with haematoxylin-eosin shows the presence of clue cells with the typical granular appearance resulting from the adherence of numerous bacteria among several epithelial cells with a smooth surface (Fig. 2). Both C3 and MBL were detected by immunofluorescence on a few cells in both CVL, though with a different staining intensity. Conversely, C1q was barely visible, despite the fact that it was present in the CVL of both patients in measurable amount. Interestingly, the staining pattern exhibited by the positive cells was not homogeneous, but had a granular appearance, that corresponds to the distribution of bacteria on the surface of the clue cells. All the other epithelial cells present in the same CVL that did not carry surface-bound bacteria were found to be negative. To further prove that the bacteria attached to the clue cells were the target of the C components, we decided to examine G. vaginalis, the bacterial species most frequently isolated from patients with BV, for its ability to bind the early C components. To this end, G. vaginalis were cultured from one of the two CVL and incubated either with NHS, to activate the classical and the alternative pathways, or with MBL for the lectin pathway. The results presented in Fig. 3 essentially confirm the observation made on the clue cells showing that C3 and MBL bind to G. vaginalis. We failed to detect C1q on the clue cells irrespective of whether the cells were exposed to the purified protein or to NHS as a source of C1q.
Detection of C components on Candida in the vaginal cavity
A procedure similar to that used for the clue cells was followed to reveal binding of the early C components to Candida obtained by cytocentrifugation of the CVL from patients with VVC. Figure 4 shows several epithelial cells present in the CVL obtained from one VVC patient and interspersed among these cells a few Candida hyphae that stained for C3 by immunofluorescence. We failed to detect C1q on the surface of Candida confirming similar results obtained with the clue cells and G. vaginalis possibly due to the absence of C-fixing antibodies in the CVL. Our inability to reveal bound MBL was surprising because Candida is well known to fix MBL. A possible explanation for this finding is that the low pH usually observed in the CVL of healthy women and in patients with VVC may cause a structural alteration of the fungal hyphae that may account for their inability to fix C. To test this hypothesis, Candida yeast cells were incubated for 1 h at R.T. with recombinant MBL in Tris-HCl buffer pH 4, and examined for their ability to bind the initiating component of the lectin pathway by immunofluorescence. Under these experimental conditions, we failed to show interaction of MBL with Candida(Fig. 5a). A faint immunofluorescence staining was seen on Candida incubated with MBL in buffer at pH 4·5 and the staining increased when the pH was further raised to 5·5. The finding that MBL binding to Candida grown in culture is markedly inhibited at low pH prompted us to perform a similar experiment on the Candida hyphae present in the CVL. To this end, CVL containing Candida hyphae was cytocentrifuged and incubated with recombinant MBL in buffers at different pH ranging between 4 and 5·5. The results presented in Fig. 5 (panel B) confirmed the data obtained with the Candida yeast cells showing that the interaction of MBL with the fungi is pH dependent.
The presence of C components has been documented in various tissues and extravascular fluids where they contribute to the local protection. We now show that the three recognition molecules of the C system, MBL, C1q and C3, can be detected in the CVL of healthy women and in patients with VVC and BV.
The vaginal levels of C3, C1q and MBL reflect the different concentrations of these components in the circulation suggesting that a large proportion of the C proteins detected in the vaginal fluid derive from the circulation, although a contribution of locally produced C components can not be excluded. The undetectable level of MBL in the CVL of a good proportion of healthy women can be explained by the relatively high incidence of heterozygotes for the mutant O allele (26%) in this study group. This incidence is essentially similar to that reported by Boniotto et al. in a non selected population of the same regional area.
The level of the three recognition molecules did not increase in the CVL of patients with BV and VVC, and, if anything, was somewhat lower then that found in the control group, although the difference was not statistically significant. This is surprising because an increased level of these components is usually associated with a marked inflammatory process, that has not been clearly documented in VVC and BV patients [14,15]. It should be emphasized that Babula et al. were also unable to find increased levels of MBL in the CVL of patients with VVC. The concentration of MBL in these patients was less than half that of the control subjects, due to high frequency of the heterozygotes for the mutant O allele, which was not observed in our VVC patients.
The detection of MBL and C3 on clue cells raises the possibility that these C components help to recognize the infectious agents, which colonize the cervicovaginal mucosa. The granular pattern of fluorescence staining of these cells, which corresponds with the similar distribution of bacteria on their surface, suggests a specific interaction of MBL and C3 with the bacteria adhering to clue cells. This conclusion is also supported by the negative staining of epithelial cells uncoated by bacteria and by the binding of MBL and C3 to G. vaginalis grown from the vaginal discharge of BV patients with clue cells. Our failure to stain clue cells for C1q and to show binding of C1q to cultured G. vaginalis excludes a direct interaction of C1q with these bacteria. This may be due to a limited amount of specific IgM or IgG directed against these bacteria in the CVL of BV patients, as opposed to IgA antibodies . The colocalization of MBL and C3 on the surface of clue cells, in the absence of C1q, suggests that the MBL pathway may contribute to C3 deposition on Gardnerella and that MBL deficiency may increase the susceptibility to this infection in patients.
Binding of C3 to Candida hyphae is in line with the known ability of Candida to activate the alternative pathway, as shown by the ability of C2 deficient serum and EGTA-treated normal serum to induce C activation . The classical pathway, that may be activated by anti-mannan antibodies frequently found in patients with candidiasis, is unlikely to be implicated in the binding of C3 to the hyphae because C1q was not detected on the surface of the fungi. Our data do not exclude that C3 binds, at least in part, to C3 receptors that are particularly expressed on the surface of hyphae [17,18]. The interaction of C3 activation products with C3 receptor is believed to account for approximately 50% of bound C3. This conclusion is based on the finding that C3 bound to C3 receptor on Candidae is not released by treatment with hydroxylamine as are the C3 fragments bound following C activation by either alternative or classical pathway .
Contrary to the results obtained with G. vaginalis, we failed to detect binding of MBL to C. albicans present in the vaginal secretion of patients with VVC. This observation was surprising because Candida is one of the infectious agents shown to strongly bind MBL  due to a substantial amount of mannan present on their surface . A lower concentration of MBL in the CVL of VVC patients can not account for our finding because the MBL level in these patients was not statistically different from that found in BV patients. An intrinsic chemical alteration of Candida may provide an alternative explanation for our failure to detect bound MBL on the surface of hyphae. However, this possibility is ruled out by the observation that binding of MBL to Candida increases substantially by raising the pH from 4, usually found in the vaginal discharge of VVC patients, to pH 5·5. This finding suggests that the interaction of MBL with Candida is pH dependent and probably explains the beneficial effect of local treatment that tends to increase the local pH.
In conclusion, we demonstrated that MBL and C3 present in the vaginal cavity act as recognition molecules for infectious agents that colonize the cervicovaginal mucosa. Our finding that MBL, but not C1q, binds to bacteria and fungi in vagina suggests that the lectin and classical pathways of complement activation may play a different role in immune defence in the female genital tract.
This work was supported by Italian MIUR (Cofin 2003-06220-004 and FIRB 2001 RBAU01C3CJ), the Danish Medical Research Council, the Novo Nordisk Research Foundation and EU concerted action (Contract N° QLG1-CT-2001–01039). We thank Alberto Candiotto (Department of Obstetrics and Gynecology, IRCCS, Burlo Garofolo, University of Trieste, Italy) for his assistance in collecting cervicovaginal lavage from patients and controls.