Isolation of Vaginal Lactobacilli and Characterization of Anti-Candida Activity

Healthy vaginal microbiota is dominated by Lactobacillus spp., which form a critical line of defence against pathogens, including Candida spp. The present study aims to identify vaginal lactobacilli exerting in vitro activity against Candida spp. and to characterize their antifungal mechanisms of action. Lactobacillus strains were isolated from vaginal swabs of healthy premenopausal women. The isolates were taxonomically identified to species level (L. crispatus B1-BC8, L. gasseri BC9-BC14 and L. vaginalis BC15-BC17) by sequencing the 16S rRNA genes. All strains produced hydrogen peroxide and lactate. Fungistatic and fungicidal activities against C. albicans, C. glabrata, C. krusei, C. tropicalis, C. parapsilosis and C. lusitaniae were evaluated by broth micro-dilution method. The broadest spectrum of activity was observed for L. crispatus BC1, BC4, BC5 and L. vaginalis BC15, demonstrating fungicidal activity against all isolates of C. albicans and C. lusitaniae. Metabolic profiles of lactobacilli supernatants were studied by 1H-NMR analysis. Metabolome was found to be correlated with both taxonomy and activity score. Exclusion, competition and displacement experiments were carried out to investigate the interference exerted by lactobacilli toward the yeast adhesion to HeLa cells. Most Lactobacillus strains significantly reduced C. albicans adhesion through all mechanisms. In particular, L. crispatus BC2, L. gasseri BC10 and L. gasseri BC11 appeared to be the most active strains in reducing pathogen adhesion, as their effects were mediated by both cells and supernatants. Inhibition of histone deacetylases was hypothesised to support the antifungal activity of vaginal lactobacilli. Our results are prerequisites for the development of new therapeutic agents based on probiotics for prophylaxis and adjuvant therapy of Candida infection.


Introduction
The homeostasis of the vaginal ecosystem results from complex interactions and synergies among the host and different microorganisms that colonize the vaginal mucosa [1,2]. Healthy vaginal microbiota is generally dominated by Lactobacillus spp. [3,4]. These bacteria form a critical line of defence against potential pathogens by producing antimicrobial compounds, or through competition for adherence to the vaginal epithelium [5][6][7][8]. For the positive effects of lactobacilli on the health of female genital tract there is an increasing interest for their use in probiotic formulations for the prophylaxis and therapy of several vaginal disturbances [9][10][11].
Vulvovaginal candidiasis (VVC) is a common infection compromising the quality of life of many women. Candida albicans is the most frequent etiologic agent [12]. Information on VVC incidence is incomplete, since the disease is not reportable and data collection is hampered by inaccuracies of diagnosis and the use of non-representative study populations. Thus, the extent to which VVC is a source of population-level morbidity remains uncertain [13]. Although the pathogenesis of VVC remains a controversial issue, it seems that, when the balance of the vaginal microbiota is disrupted, the overgrowth of Candida spp. is facilitated. Individual susceptibility, pregnancy, antibiotic therapy, use of contraceptives and spermicide, frequent sexual intercourse, diabetes and immunosuppression are factors that increase the risk for development of VVC [14][15][16]. Infections by C. albicans are commonly treated with azole antifungal drugs. Since azoles are fungistatic for C. albicans, cells repetitively exposed to these antifungals adapt to the drug pressure and become azole resistant [17]. The high incidence of VVC together with the growing problem of drug resistance highlight the need for the development of new effective agents for the prevention and therapy of this gynecological infection [18].
Establishment of a healthy vaginal microbiota using probiotic Lactobacillus strains might be a supportive and preventive measure against VVC [19]. Lactobacilli are supposed to protect from Candida infection but the mechanisms underlying antifungal activity are still not clearly understood. Although lactobacilli are quite common even in the vaginal epithelium of women with VVC, the composition of Lactobacillus species/strain is different compared to healthy women. In particular, the development of VVC has been associated with the lack of H 2 O 2producing Lactobacillus species [20].
The capacity of lactobacilli to adhere and compete for the adhesion sites on vaginal mucosa can be involved in the inhibition of Candida colonization. The blockage of Candida adherence may be by exclusion, competition for receptors sites and displacement of adhered yeast cells [21][22][23][24]. Several studies have shown that the inhibition of histone deacetylases (HDACs) can affect morphogenesis, attenuate the virulence and reduce the adhesion of Candida spp. to host mucosa [25][26][27], in addition to enhancing the antifungal activity of azole drugs [28]. Butyrate and lactate are known HDAC inhibitors [29,30]. Since these metabolites are produced by several human commensal bacteria, the anti-Candida activity of the vaginal lactobacilli could be mediated by the inhibition of HDACs.
The present study aims to isolate vaginal lactobacilli from healthy women and to evaluate their ability to counteract the infection by Candida spp. In particular, we investigated possible antifungal mechanisms of action, i.e, the production of antimicrobial compounds, the interference with adhesion process and the inhibition of HDACs. We used a metabolomic approach based on 1 H-NMR [31] to find correlations between metabolism of lactobacilli and anti-Candida activity. A major potential application of this study concerns the identification of active Lactobacillus strains to propose as probiotics for prophylaxis and/or adjuvant therapy of VVC.

Isolation of vaginal lactobacilli from healthy women and taxonomic characterization
Fifteen pre-menopausal Caucasian women (aged 18-45 years old), who have no symptoms of vaginal or urinary tract infection, were recruited for the present study. The women were nonmenstruating and not receiving oral or local antimicrobial therapy within the previous 2 weeks. All volunteers provided a written informed consent in accordance with the Ethics Committee of the University of Bologna (52/2014/U/Tess) and the institutional review board approved the study. Mid-vaginal secretions were self-collected by women with E-swabs (Copan, Brescia, Italy) and immediately processed for lactobacilli isolation. The specimens were coded to assure full anonymousness.
Lactobacillus clones were isolated onto de Man, Rogosa and Sharpe (MRS) and Brain-Heart Infusion (BHI) agar plates (Difco, Detroit, MI). Both MRS and BHI agar plates were supplemented with 0.05% L-cysteine. Plates were incubated anaerobically for 24 h at 37°C in anaerobic jars supplemented with Anaerocult C (Merck, Milan, Italy). Colonies with different morphologies yielding variable rods by microscope observation were selected for glycerol stock preparation. To prepare lactobacilli fractions, 18-h MRS/BHI cultures (OD 600 = 0.5) were centrifuged at 5,000 X g for 10 min at 4°C. Supernatants were filtered through a 0.2 μm membrane filter to obtain cell free supernatants (CFS). Cell pellets (CP) were washed in sterile saline.
Genomic DNA was extracted from lactobacilli CP using DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany) following the protocol "Pretreatment for Gram-positive bacteria". The extracted DNA was amplified with Lactobacillus genus-specific primers Lac1 and Lac2 [32]. The positive isolates were taxonomically characterized to the species level by sequencing the 16S ribosomal RNA (rRNA) gene. Briefly, the complete 16S rRNA gene (1.5 kb) was amplified with the universal primers 27F and 1492R [33] and sequenced. The obtained sequences were compared with the sequences available in the Ribosomal Database Project (RDP, http://rdp. cme.msu.edu/) [34] in order to identify the Lactobacillus species.

Determination of hydrogen peroxide
Lactobacillus strains were tested for their ability to produce H 2 O 2 as described by Pendharkar et al. [35] with slight modifications. Isolates were cultured onto MRS agar plate containing 0.25 mg/ml 3,3', 5,5'-tetramethylbenzidine and 0.01 mg/ml of horseradish peroxidase in anaerobic condition for 72 h. Plates were exposed to air and on the basis of the time required for the blue coloration to appear, isolates were scored as low [score 1(>20 min)], medium [score 2 (10-20 min)] and high producing strains [score 3 (<10 min)]. Isolates not producing blue coloration were scored as 0.

H-NMR analysis
One ml of CFS obtained from lactobacilli was added to 160 μl of a D 2 O solution of 3-(trimethylsilyl)-propionic-2,2,3,3-d4 acid sodium salt (TSP) 6.25 mM set to pH 7.0 by means of a 100 mM phosphate buffer. 1 H-NMR spectra were recorded at 298 K with an AVANCE III spectrometer (Bruker, Milan, Italy) operating at a frequency of 600.13 MHz. To avoid the presence of broad signals arising from slowly tumbling molecules, a T 2 filter of 400 echoes, separated by an echo time of 400 μs, was applied. The signals were assigned by comparing their chemical shift and multiplicity with Chenomx software data bank (Chenomx Inc., Canada, ver 8.02). Literature on previous quantitative investigations conducted with the same technique reports a precision error below 2% [36].
The fungistatic activity of Lactobacillus CFS was determined by broth microdilution in accordance with the EUCAST guidelines [37] with slight modifications. Briefly, Candida suspensions were prepared in sterile water from 24-h cultures on SD agar and the turbidity was adjusted to OD 530 = 0.5. RPMI 1640 medium buffered to pH 7.0 with 0.165 M morpholinepropanesulfonic acid buffer and 2% glucose was added to give a yeast suspension of 1-5 × 10 5 CFU/ml. Each well of a flat-bottom microdilution tray was inoculated with 100 μl of yeast suspension, and subsequently filled with 100 μl of Lactobacillus CFS. A growth control well contained 100 μl of sterile MRS medium and 100 μl of the same Candida suspension. The microdilution trays were incubated at 35°C and growth was observed after 24 and 48 h. The results were read considering a prominent decrease in turbidity (at least 50% reduction in growth) relative to the control by measuring the absorbance at 450 nm with Multiskan FC Microplate Photometer (Thermo Fisher Scientific Inc.,Waltham, USA). To determine a fungicidal effect of Lactobacillus CFS, 20 μl of samples from wells exhibiting less than 50% of growth were spotted onto SD agar plates and incubated at 35°C for 24/48 h. Fungicidal activity was defined as a 3 log 10 reduction from the starting inoculum [38]. Following the same methods, fungistatic/fungicidal activities were tested for all the compounds identified by 1 H-NMR as differently expressed among the lactobacilli.
Capability of each Lactobacillus strain to adhere to HeLa cells was evaluated in individual tubes containing sterile coverslips as previously reported [24,39] with slight modifications. One millilitre of HeLa cell suspension, at a concentration of 5 × 10 4 cells/ml, was seeded onto each glass coverslip and incubated in 5% CO 2 atmosphere at 37°C. After 48 h, the cells, grown to approximately 70% confluence, were washed twice with PBS and treated with 100 μl of lactobacilli suspension (5 × 10 8 bacteria/ml). The tubes were then incubated for 1 h at 37°C in 5% CO 2 . Cell monolayers were washed several times in PBS, fixed with May-Grünwald and stained with Giemsa. Results were read at light-microscopy (1000×) and HeLa cells were scored for the presence and number of lactobacilli attached. Each adherence assay was conducted in duplicate and 200 randomly chosen cells were evaluated for lactobacilli adhesion.
To study the interference of vaginal lactobacilli with the adherence of Candida to HeLa cells, C. albicans 1 was chosen as model strain. Yeast culture in BHI broth was incubated at 30°C for 18 h under constant shaking, in order to obtain blastospores at late exponential growth phase [23]. Yeast cells were collected by centrifugation, washed three times and finally suspended in saline solution to the working dilution of 5 × 10 8 yeasts/ml. Three types of assays were performed to study the capacity of Lactobacillus fractions (CP and CFS) to interfere with the adherence of C. albicans to HeLa cells: exclusion, competition and displacement [21]. In the adhesion assays, CP and CFS fractions corresponding to 5 × 10 7 Lactobacillus cells were incubated with 5 × 10 7 Candida cells following the timelines described, as follows. In the exclusion assay, lactobacilli fractions were incubated for 1 h at 37°C on HeLa cells. Afterwards, Candida cells were added and further incubated for 1 h. In the competition assay, lactobacilli fractions and Candida were inoculated simultaneously onto HeLa cells and incubated for 1 h at 37°C. In the displacement assay, Candida cells were inoculated onto HeLa cells for 1 h at 37°C. Successively, lactobacilli fractions were added and further incubated for 1 h. Yeast adhesion to HeLa cells was assessed by microscopy (400×) after Giemsa staining by counting the number of Candida cells attached to 200 randomly chosen cells. Results were expressed as the percentage of C. albicans adherent to each HeLa cell and compared with adhesion without lactobacilli fractions (control value = 100%). Interference experiments were conducted three times with at least three replicates per group.

Histone acetylation profile analysis
Candida albicans 1 was chosen as model strain to study the histone acetylation profile induced by vaginal lactobacilli. Log-phase Candida cells were inoculated at an OD 600 of 0.5 in lactobacilli CFS or MRS broth (negative control) and incubated at 30°C for 6 h. Sodium Butyrate (20 mM) was used as positive control, a culture of Staphylococcus aureus in MRS was used as a representative Gram-positive organism. Histones were extracted from yeast cultures as described by Knapp et al. [40] with slight modifications. A volume equivalent to 20-40 OD units of each culture was collected and subjected to nuclei isolation. Nuclei were washed for 15 minutes on ice in Wash Buffer (10 mM Tris-HCl, pH 8/75 mM NaCl/30 mM Na-Butyrate/0.5% NP-40/1.0 mM PMSF/10 μg/ml each of protease inhibitors). Washes were repeated four times. Histones were acid extracted by incubating nuclei in H 2 SO 4 0.4N for 1 h in ice, then precipitated overnight in acetone at -20°C. Equal amount of histones were loaded to a 15% acrylamide gel and separated by SDS-PAGE, then transferred to a nitrocellulose membrane and probed with anti-acetyl Lysine primary antibody (Merck Millipore, Darmstadt, Germany) and peroxidase-conjugated anti-mouse IgG secondary antibody (GE Healthcare, Milan, Italy). Peroxidase activity was detected by Westar XT system (Cyanagen, Bologna, Italy). Digital images and densitometric analysis were performed by using the GS-800 calibrated densitometer (Bio-Rad Laboratories, Milan, Italy). For each strain, histone acetylation profile was analysed in triplicate.

Statistical analysis
Differences in the metabolome composition were assessed by means of a two-tailed unpaired Wilcoxon test, through the homonym function implemented in R computational software (www.r-project.org). Linear correlations between fungistatic/fungicidal activities and metabolome were assessed by means of ANOVA test. Statistical analyses for the adhesion assays data were performed by using ANOVA test (GraphPad Prism version 5.02 for Windows, GraphPad Software, San Diego California USA, www.graphpad.com). Results were expressed as mean ± Standard Error of the Mean (SEM). Differences were deemed significant for P values < 0.05 or highly significant for P values < 0.01.

Nucleotide sequence accession numbers
The nucleotide sequences of the 16S rRNA genes of the Lactobacillus strains BC1 to BC17 have been deposited in the DDBJ nucleotide sequence database under accession numbers AB976542 to AB976558.

Taxonomy of vaginal lactobacilli and production of antimicrobial compounds
Seventeen Lactobacillus isolates were obtained from vaginal swabs of 15 healthy premenopausal women. All isolates were cultured on MRS supplemented with L-cysteine, except L. gasseri BC14 which was cultured on BHI supplemented with L-cysteine. The Lactobacillus isolates were taxonomically identified to species level by sequencing the 16 rRNA gene: 8 isolates belong to L. crispatus (BC1-BC8), 6 isolates to L. gasseri (BC9-BC14) and 3 isolates to L. vaginalis (BC15-BC17) ( Table 1).
As a first step in the characterization of the antimicrobial properties of the vaginal lactobacilli isolated in this study, we evaluated the pH of the cultural supernatants, production of H 2 O 2 , lactate and butyrate ( Table 1). The pH of the supernatants was in the range 3.7-4.7, showing the ability of all strains to acidify the medium. Lactobacillus isolates were scored for H 2 O 2 production on a scale of 0 to 3. Lactobacillus gasseri BC14 was not tested for H 2 O 2 production since its incapacity of growing on MRS agar plates and it is important to underline that the hydrogen peroxide test cannot be carried out in BHI plates because tetramethylbenzidine and horseradish peroxidase precipitate in the form of crystals in this medium. Hydrogen peroxide was produced by the totality of the strains. The levels of H 2 O 2 production did not seem related to a particular species. The strongest H 2 O 2 -producers (score 3) were L. crispatus BC1, BC2, L. gasseri BC10, BC11, BC12, and L. vaginalis BC17. Lactate and butyrate were measured in CFS of lactobacilli cultures by 1 H-NMR analysis. Lactate was produced by all isolates at concentrations ranging from 1.42 to 47.4 mM. In general, L. vaginalis species was characterized by good production levels, in particular L. vaginalis BC15 was the highest producer strain. Instead, L. crispatus and L. gasseri species showed variable trends of lactate production. Butyrate was produced by 9 of the 17 lactobacilli at concentrations ranging from 1.00 x 10 -2 to 8.33 x 10 -1 mM. Production of butyrate appeared to be negligible in L. gasseri species and variable in the other two species. The highest levels of this metabolite were found in the supernatants of L. crispatus BC6 and L. vaginalis BC15.

Lactobacilli fungistatic and fungicidal activities
The fungistatic and fungicidal activities of CFS of the vaginal lactobacilli were evaluated against 4 clinical isolates of C. albicans and 5 clinical isolates referring to species different from C.  Table). In function of the number of Candida isolates inhibited by lactobacilli supernatants, the fungistatic and fungicidal activities were scored on a scale of 0 to 4 for C. albicans and 0 to 5 for non-C. albicans (Table 3). In general, the strains tested were more active toward C. albicans. No Lactobacillus strains showed activity against C. krusei and C. parapsilosis. Even, L. gasseri BC14 appears to stimulate the growth of C. tropicalis and C. krusei probably through production of metabolites that act as growth factors for these two species of Candida (S2 Table). The strains that showed the broadest spectrum of anti-Candida activity were L. crispatus BC1, BC4, BC5 and L. vaginalis BC15, since they had fungicidal activity against all the isolates of C. albicans and against C. lusitaniae strain. In addition, L. crispatus BC1 and L. vaginalis BC15 exhibited fungistatic activity towards C. tropicalis and C. glabrata, showing the best profile of anti-Candida activity. Lactobacillus crispatus BC4 and BC5 were fungistatic towards only one of C. tropicalis and C. glabrata species. A good spectrum of activity was also shown by L. crispatus BC7 which was fungicidal for all C. albicans isolates and fungistatic for C. tropicalis and C. glabrata. The less active strains were L. gasseri BC10, BC11, BC14 and L. vaginalis BC16. L. gasseri BC10 and BC11 showed no fungistatic/fungicidal activity towards any of Candida isolates. Lactobacillus vaginalis BC16 exerted a fungistatic activity only towards C. albicans 1 while L. gasseri BC14 was fungistatic for C. albicans 1 and C. glabrata. Among the remaining lactobacilli, exhibiting an intermediate profile of antifungal activity, the most interesting were L. crispatus BC3 and BC6 (fungicidal against 3 species of C. albicans and C. lusitaniae), and L. crispatus BC2 (fungicidal against 2 species of C. albicans and C. lusitaniae). In summary, the anti-Candida activity of lactobacilli CFS was strongly associated with L. crispatus sp. because all strains of this species were moderately or highly active against the vaginal pathogen. Conversely, poor anti-yeast activity was exhibited by CFS of L. gasseri spp. Notably, L. vaginalis spp. showed extremely variable profiles of antifungal activity,

Lactobacilli metabolome correlates with taxonomy and fungistatic/ fungicidal activity
We sought a metabolic description of the CFS of the vaginal lactobacilli isolated in the present study. L. gasseri BC14 was not included in the metabolomics analysis because the metabolic profile of BHI supernatant could not be compared with the metabolic profiles of MRS supernatants. We identified 40 molecules mainly belonging to the families of aminoacids, organic acids monosaccharides, ketones and alcohols (S3 Table). A Principal Component Analysis (PCA) was performed on entire set of metabolites identified (Fig 1). In the biplot describing the distribution of Lactobacillus strains in relation to the pool of metabolites, PC1 and PC2 accounted for the 45.4% of the whole variance of the investigated samples (Fig 1A). This multivariate analysis showed two interesting correlations: (i) metabolome versus taxonomy (PC1, expl. var 28%) and (ii) metabolome versus fungistatic/fungicidal activity (PC2, expl. var 17.4%). These correlations were best visualized by means of box blots representing the distribution of Lactobacillus species (Fig 1B) and fungistatic/fungicidal activity scores (Fig 1C) in relation to the metabolome. Metabolic profiles varied to a greater extent according to the taxonomy. In particular, metabolome of L. vaginalis significantly differed from those of L. crispatus and L. gasseri (P < 0.05). The highest metabolic heterogeneity was observed within L. crispatus, as demonstrated by the width of the corresponding boxplot. Even fungistatic and fungicidal activities of the vaginal lactobacilli were related to their metabolome. Strains with different activity scores were clearly separated in the vertical direction: the most active strains occupied the lower positions while the less active strains were placed in the higher areas of the two-dimensional space represented by the biplot. A linear correlation was observed between median metabolic variance on PC2 and antifungal activity scores against both C. albicans (fungistatic, R 2 = 0.86, P = 0.046; fungicidal, R 2 = 0.83, P = 0.02) and C. non-albicans isolates (fungistatic, R 2 = 0.99, P = 0.003). The correlation coefficient related to the fungicidal activity against C. non-albicans was not calculated due to the presence of only two activity scores (0 and 1). We searched by Wilcoxon univariate statistical test the metabolites which varied in relation to antifungal activity. We identified 4 metabolites (butyrate, orotate, pyroglutamate, and isoleucine) whose concentrations significantly increased (P < 0.05) in CSF of active strains. Fungistatic and fungicidal activities of these compounds were evaluated, but no substance was active at the concentration found in the lactobacilli supernatants. No activity was also observed when a mixture of butyrate, orotate, pyroglutamate and isoleucine was tested, suggesting the lack of synergistic effects. However, we cannot exclude a synergistic action of these metabolites in the more complex cultural medium where other bacterial molecules may act as enhancers.
Lactobacilli interference with C. albicans adhesion to HeLa cells The adhesion of vaginal lactobacilli to epithelial tissue represents the first step in the formation of a barrier to prevent undesirable microbial colonization [22][23][24].
Firstly, Lactobacillus strains were examined for their ability to adhere to HeLa cells, a cell line that originated from a human carcinoma of the cervix (Fig 2). Adherence varied greatly among the lactobacilli analysed, in a range between 0.07 ± 0.03 and 17.68 ± 0.78 (mean ± SEM) bacteria/cell. Lactobacillus crispatus BC1, L. crispatus BC3 and L. gasseri BC8 were the most adhesive strains (> 10 bacteria/cell), L. crispatus BC2 and L. vaginalis BC15 showed an  intermediate adhesiveness (2-10 bacteria/cell), whereas the remaining strains adhered at low levels (< 2 bacteria/cell). These data demonstrate that the adhesive properties are strainspecific rather than species-specific, varying considerably between strains of the same species.
To verify the antagonist effect of the vaginal lactobacilli against C. albicans, the influence of CP and their respective CFS on the adhesion capacity of the yeast to HeLa cells was investigated (Fig 3). Three mechanisms of inhibition were examined: exclusion (Fig 3A), competition ( Fig 3B) and displacement (Fig 3C). Ten strains (BC1, BC2, BC5, BC7, BC8, BC9, BC10, BC11, BC12 and BC16) significantly reduced the adhesion of C. albicans through all three mechanisms. The inhibitory effect was exerted in some cases by CP and in other cases by CFS. In particular, the interference by L. crispatus BC2, L. gasseri BC10 and L. gasseri BC11 was mediated by both CP and CFS, suggesting that these strains were the most active in terms of inhibition of the pathogen adhesion. Interestingly, BC2, BC10 and BC11 were not the most adhesive strains, as shown by the Fig 2. Only three strains did not exercise any effect on C. albicans adhesion: L. gasseri BC13, L. gasseri BC14 and L. vaginalis BC17. The remaining strains showed an intermediate behaviour exerting inhibition through one or two mechanisms.
Lactobacilli effects on C. albicans histone acetylation Bothe the yeast histone deacetylase (HDAC) inhibition, and the consequent histone hyperacetylation, represent a novel mechanism by which Candida growth and adhesion to the host epithelium can be reduced [25][26][27]. Thus, we wondered if the fungistatic/fungicidal activity and interference with Candida adhesion exerted by the vaginal lactobacilli could be associated to this mechanism.
Acetylation profiles of H2/H3 and H4 histones of C. albicans 1 were evaluated for all lactobacilli CFS, except for L. gasseri BC14 due to the inability of this strain to grow in MRS. We attributed hyper-acetylating ability to strains that determined acetylation levels of H2/H3 or H4 histones at least equal to those induced by sodium butyrate, used as positive control ( Table 4). The majority of lactobacilli caused histone hyper-acetylation. In detail, histones H2/ H3 were hyper-acetylated by 12 strains and histone H4 by 11 strains. Lactobacillus gasseri BC13 and L. vaginalis BC17 were the only strains that did not cause acetylation of any kind of histone. Notably, these lactobacilli showed no fungicidal activity towards C. albicans 1 nor any  inhibitory effect on the adhesion of the pathogen. These data suggest that the complete lack of inhibition of HDACs could compromise antifungal activity of lactobacilli.

Discussion
The vaginal mucosa is inhabited by both bacteria and fungi, which normally coexist with the host in a tightly regulated manner. Under certain circumstances this ecological balance may break and turn into a pathological state. A decrease in the number of lactobacilli among the vaginal microbiota may be an aid in the transmission of genitourinary pathogens, including Candida spp. While the treatment of VVC by conventional drugs is relatively effective, it has been suggested that women could benefit from restoration of the vaginal communities via supplementation with probiotics [41]. The aim of the present study was to isolate vaginal lactobacilli from healthy women, to characterize them at a molecular level and to evaluate their anti-Candida proprieties, in the perspective to develop successful vaginal probiotics for VVC management, considering that the drug-resistances are at present a major problem for the public health systems [17].
In this context, we isolated strains belonging to L. crispatus, L. gasseri and L. vaginalis species. The isolation of L. crispatus has been strongly associated with a normal vaginal microbiota and absence of vaginal dysbiosis [42]. Longitudinal studies have also shown that the presence of L. crispatus promotes stability of the vaginal microbiota [43]. Despite the high incidence of L. iners in the human vaginal microbiota, as detected by culture-independent molecular studies [4,44], we did not obtain isolates belonging to this species probably because of its stringent nutritional requirements and very low oxygen tolerance [45]. On the other hand, as our goal was to identify health-promoting lactobacilli, L. iners was of little interest given its close correlation with vaginal dysbiosis [46]. In view of potential application of the isolated Lactobacillus strains as vaginal probiotics, we sought to characterize the capacity of these strains to modify the host microenvironment and therefore deliver health benefits. Hydrogen peroxide and lactate are classically associated with the antimicrobial properties of the genus Lactobacillus [5]. All strains produced hydrogen peroxide in agreement with the assumption that the vaginal microbiota of healthy women is dominated by H 2 O 2 -producing lactobacilli [35]. Also lactate was produced by all lactobacilli, while butyrate was produced only by certain strains at concentrations that varied significantly depending on the activity score. Since butyrate is a known HDAC inhibitor [29], we hypothesize that it may enhance the anti-Candida activity of lactobacilli through the mechanism of histone hyperacetylation.
The fungistatic and fungicidal activities of the vaginal lactobacilli were evaluated against C. albicans and C. non-albicans. Compared to previous studies focused on the antifungal activity of lactobacilli [47,48], our work has the additional value of examining Lactobacillus isolates of vaginal source against a broad spectrum of Candida species, including the most represented species responsible for gynaecological infections. Therefore, the results obtained in this work provide important information about the real applicability of vaginal lactobacilli in the prevention and treatment of VVC. The broadest spectrum of activity was observed for L. crispatus BC1, BC4, BC5 and L. vaginalis BC15, exhibiting fungicidal activity against all isolates of C. albicans and C. lusitaniae. Among these strains, L. crispatus BC1 and L. vaginalis BC15 exhibited the best anti-Candida profile covering also C. tropicalis and C. glabrata, albeit in fungistatic mode.
In order to interpret, through a metabolic key, the differences in fungistatic/fungicidal power of the vaginal lactobacilli, we studied by 1 H-NMR their metabolome and we looked for correlations with taxonomy and activity score. The strong correlation between metabolic profile and taxonomy highlighted the inter-specific variability of bacterial metabolism. Metabolic variance was also related to antifungal activity scores, confirming the excellent antifungal profile of the majority of L. crispatus strains and L. vaginalis BC15. These data highlight the potential of metabolomics to measure the taxonomic distance between different Lactobacillus strains and predict their anti-Candida activity. Although metabolomics has been applied to evaluate the impact of probiotics on the host organism [49], to our knowledge this is the first study employing a metabolomic approach to investigate the antimicrobial activity of healthpromoting bacteria, representing a new idea for future researches. Impairment of pathogens adherence to human cells is considered of major importance for the in vitro evaluation of probiotic properties [50]. Most of Lactobacillus strains significantly reduced C. albicans adhesion through every mechanism including exclusion, competition and displacement. In particular, L. crispatus BC2, L. gasseri BC10 and L. gasseri BC11 appeared to be the most active in reducing pathogen adhesion, as their effects were mediated by both cells and supernatants. Interestingly, BC2, BC10 and BC11 were not the most adhesive strains suggesting that the inhibitory effects are not merely due to steric encumbrance and saturation of the adhesion sites, but rather to a reduction of the adherence of the pathogen itself and/or to modifications of the epithelial cells surface. Furthermore, BC2, BC10 and BC11 were not the best performing strains in terms of fungistatic/fungicidal activity. This finding suggest that lactobacilli isolated from healthy vagina can carry out their protective function against Candida infection exploiting one particular strategy (inhibition of growth or adhesion) rather than through the combination of two complementary mechanisms.
Inhibition of HDACs can impair fungal growth and adherence to host cells [25][26][27]29]. For this reason, we investigated whether the antagonism towards C. albicans could be associated with histone hyper-acetylation. We observed that histone hyper-acetylation was a widespread prerogative among lactobacilli of vaginal source. Notably, the only strains that did not cause any kind of histone acetylation (L. gasseri BC13 and L. vaginalis BC17) were the same ones that did not exercise any inhibitory effect on Candida adhesion nor a fungicidal activity, suggesting that inhibition of HDACs could support antifungal activity of vaginal lactobacilli.
Further studies are necessary for a thorough understanding of the antifungal mechanisms of vaginal lactobacilli, i.e. analysis of the antimicrobial activity of the cell-free supernatants to identify specific classes of bioactive molecules, development of a vaginal model with simulated vaginal fluids and in vivo tests with animals. However, the findings from this work have enabled us to achieve two important objectives: (i) identify vaginal lactobacilli active against Candida spp. and (ii) characterize the mechanisms of action underlying antagonism toward pathogen. The application that follows is the combination of strains exerting different modes of action in order to obtain a probiotic blend with enhanced therapeutic properties. In particular, we have identified strains with a good spectrum of fungistatic/fungicidal activity (L. crispatus BC1 and L. vaginalis BC15) that may be associated with strains particularly active in reducing the adhesion of the pathogen (L. crispatus BC2, L. gasseri BC10, L. gasseri BC11). The choice of different species is also an added advantage as it ensures a wider expression of metabolic functions.
Supporting Information S1