Variability of the response of human vaginal Lactobacillus crispatus to 17β-estradiol

We previously showed that the physiological concentration of 17β-estradiol in the vaginal environment is sufficient to affect the membrane dynamics and adhesion phenotype of the Lactobacillus crispatus strain CIP104459. However, L. crispatus is a heterogeneous species. Here, we investigated the effect of 17β-estradiol on the recently isolated L. crispatus vaginal strain V4, related to a cluster distant from CIP104459 and at the limit of being a different subspecies. Grown in the same medium, the two strains expressed a highly similar pool of proteins. However, in contrast to CIP104459, L. crispatus V4 showed high aggregation potential and 17β-estradiol promoted this phenotype. This effect was associated with large changes in cell-surface polarity and Lewis acid/base properties. In addition, we observed no effect on the membrane dynamics, contrary to CIP104459. These results can be explained by differences in the properties and organization of the S layer between the two strains. However, as for CIP104459, 17β-estradiol increased biosurfactant production of L. crispatus V4 and their adhesion to vaginal cells. This suggests that 17β-estradiol agonists would be valuable tools to favor a stable re-implantation of L. crispatus in the vaginal mucosa.

Quantification of proteins expressed by L crispatus V4 in the absence or presence of 17β-estradiol showed very limited differences. There were no differences in protein expression > twofold, considered to be the limit of significance, between L. crispatus V4 exposed or not to 10 -8 M of 17β-estradiol and only one protein, elongation factor G (MHJPMHAL_00289), showed an increase > twofold after exposure to 10 -10 M of 17β-estradiol (+ 2.079).
17β-estradiol affects the 1L. crispatus V4 aggregation phenotype. Preliminary studies showed that 10 -6 to 10 -10 M of 17β-estradiol had no effect on the growth kinetics of L. crispatus V4 (Suppl. Figure 1A). Lactobacillus can spontaneously form aggregates and L. crispatus is among the species with the highest potential 13 . The aggregation potential of L. crispatus V4 visibly increased after exposure to 17β-estradiol (Fig. 3A). We quantified the aggregation phenotype using the sedimentation technique 14 (Fig. 3B), which showed an inverted dose-related effect for 17β-estradiol was observed. 17β-estradiol at 10 -6 M had no detectable effect, whereas we observed a non-significant increase of the aggregation phenotype with bacteria exposed to 10 -8 M 17β-estradiol. 17β-estradiol at 10 -10 M led to significant stimulation of L. crispatus V4 aggregation, reaching + 114.4 ± 6% of the control (p < 0.05). We further investigated the aggregation phenotype of L. crispatus V4 by flow cytometry analysis (Fig. 3C). The mean size of detected structures was plotted on the horizontal axis (forward scattering fraction of the light, FSC) and the surface heterogeneity (granularity) on the vertical axis (side scatter light, SSC) 15 . Exposure of the bacteria to 10 -6 M 17β-estradiol resulted in a significant increase in mean particles size (+ 20.76 ± 0.5% of the control; p < 0.01). The signal was divided between two peaks, the smaller one likely corresponding to isolated bacteria. In parallel, the surface heterogeneity of detected particles was higher than that of the control (p < 0.05). We observed the same tendency for bacteria exposed to 10 -8 M 17β-estradiol and the FSC signal was divided between two peaks, as before, whereas the surface heterogeneity reverted to a unique group, but the differences were not statistically significant. Bacteria treated with 10 -10 M 17β-estradiol at showed Heatmap and dendrogram of average nucleotide identity (ANI) illustrating the phylogenetic relationships based on the percentage identity [high (red) to low (blue/grey)] shared between the whole sequence of Lactobacillus crispatus V4 (blue boxes and arrows) and 122 other L. crispatus isolates for which the genome sequences were accessible in the NCBI RefSeq database (accessed on 3rd November, 2020), including that of Lactobacillus crispatus CIP104459 from Clabaut et al. 12 (yellow boxes and arrows). The phylogenic tree was rooted using the genome sequence of an unrelated bacterial species, presently Leuconostoc mesenteroides subspecies mesenteroides, ATCC8293. www.nature.com/scientificreports/ the dose-related effect of 17β-estradiol was inverted in most cases and 10 -10 17β-estradiol had the greatest effect on the surface polarity of L. crispatus V4. Calculation of the Lewis acid/base ratio using the two solvent couples hexadecane/chloroform and decane/ethyl acetate showed the evolution of bacterial surface hydrophobicity to be associated with major changes in the electron donor/acceptor character. Indeed, although the surface of the control bacteria was basic, 17β-estradiol exposure made it highly acidic (p < 0.01) (Fig. 4B).
17β-estradiol has no effect on the membrane fluidity of L. crispatus V4. We further investigated the potential effects of 17β-estradiol on the surface of L. crispatus V4 by studying the impact of the steroid on membrane integrity by fluorescence anisotropy. Membrane anisotropy can be measured after the incorporation of a fluorescent probe within the cell membrane and reflects its degree of organization. In a first set of experiments, we assessed this parameter in L. crispatus V4 exposed to 17β-estradiol (at 10 -6 , 10 -8 or 10 -10 M) for 18 h, from the onset of the experiment to the beginning of the stationary growth phase (Fig. 5A). We observed minor variations, but the overall organization of the bacterial membrane remained essentially unchanged. The bacterial membrane has been shown to evolve during the growth of planktonic cultures of Staphylococcus aureus, Listeria monocytogenes, and Pseudomonas aeruginosa, in particular, in terms of fatty-acids composition, which may lead to variations in membrane fluidity 17 . As 17β-estradiol is an amphiphilic molecule and can integrate into phospholipid membranes, it may affect membrane integrity 18 . Thus, we studied the effect of 17β-estradiol on the membrane fluidity of L. crispatus V4 after 6, 12, or 24 h growth before exposure to 17β-estradiol and the fluorescent probe DPH to label membranes [19][20] . The evolution of fluorescence anisotropy was subsequently monitored over 3 h. The membrane fluidity of bacteria collected during late exponential growth phase (6 h of growth) evolved during the 3 h of measurements from a mean "r" anisotropy index of 143 ± 4 to 179 ± 1 (Fig. 5B). However, 17β-estradiol had no effect on the changes in fluorescence anisotropy, and thus membrane integrity, at any concentration studied. Similar data were obtained for 12 or 24 h (early to late stationary growth phase) L. crispatus V4 (Fig. 5C,D). Thus, membrane stiffness increased over the 3 h of monitoring period but there was no effect of 17β-estradiol.
17β-estradiol alters the mean size distribution of L. crispatus V4 cells. Scanning electron microscopy (SEM) observations of L. crispatus V4 showed 17β-estradiol to have a limited impact on the general morphology of the bacteria (Fig. 6A). We then measured the mean size of the bacteria over 60 microorganisms under each condition and classified them from small (< 1 µm) to large (> 3 µm) to quantify the potential effect of 17β-estradiol. The percentage of medium to large size microorganisms remained unchanged, whereas the percentage of small size lactobacilli significantly increased after treatment with 10 -8 and 10 -10 M 17β-estradiol (+ 7.25 ± 0.06% and + 5.76 ± 0.043%, respectively, p < 0.001) (Fig. 6B).

17β-estradiol increases adhesion to vaginal mucosa epithelial cells and promotes biosurfactant production by L. crispatus V4.
We investigated the effect of 17β-estradiol on the adhesion of L.
crispatus V4 to vaginal mucosa epithelial cells using the human vaginal VK2/E6E7 cell line. Exposure of the bacteria to 10 -6 and 10 -10 M 17β-estradiol significantly increased their adhesion to VK2/E6E7 cells (+ 33.22 ± 12.46 and + 40.09 ± 14.75%, respectively) (Fig. 7A). Bacteria exposed to 10 -8 M 17β-estradiol showed the same tendency, but the difference was not statistically significant. The adhesive properties of bacteria depend on the composition of their envelope, as well as that of secreted molecules, such as biosurfactants, which can adsorb to or insert into the bacterial membranes and modify their surface 21 . We investigated the effect of 17β-estradiol on biosurfactant production by growing L. crispatus V4 colonies on MRS-agar with or without the steroid and then delicately scraping them off and extracting the obtained biomass in water. The solution, containing potential biosurfactants, was subsequently tested by the   www.nature.com/scientificreports/ pendant drop method and the mean surface tension calculated from analysis of the curvature of the drops. We observed a tendency towards a decrease of the surface tension but 17β-estradiol showed no dose response (Fig. 7B). The effect of 17β-estradiol on the surface tension of the solution extracted from the L. crispatus V4 cultures was statistically significant at a concentration of 10 -8 M.
17β-estradiol has a minor effect on L. crispatus V4 biofilm formation. We studied biofilm formation by L. crispatus V4 by the crystal violet staining assay and confocal microscopy. No effect of 17β-estradiol on the biofilm formation activity of L. crispatus V4 was observed using the crystal violet staining assay (Fig. 8A).
In order to investigate a potential effect of 17β-estradiol on the biofilm architecture, L. crispatus V4 biofilms were visualized by confocal laser scanning microscopy. Biofilm development in flat glass-bottom microtitration plates required growing the bacteria in a "simulating genital tract secretion" (SGTS) medium 22 , as previously described 8 . Under such conditions, L. crispatus V4 was capable to form a thin and heterogeneous biofilm (Fig. 8B). Treatment of the bacteria with 17β-estradiol had no visible effect on the biofilm structure. Biofilm images were then analyzed using COMSTAT2 software. This analysis revealed no significant variation in biomass (Fig. 8C). However, 10 -6 M 17β-estradiol induced a minor (+ 17.54 ± 9.85%) but significant increase in mean biofilm thickness (p < 0.05) (Fig. 8D). No variations were observed at lower concentrations. The roughness coefficient of the biofilm also appeared to increase after exposure of the bacteria to 10 -8 M 17β-estradiol, but the veracity of such an increase is still uncertain because of the heterogeneity of the biofilm (Fig. 8E).

Discussion
Given the effect of 17β-estradiol on the adhesion of L. crispatus CIP104459 to vaginal mucosa cells 8 and the need to develop new solutions to re-equilibrate the vaginal microbiota, it was necessary to verify whether this property is shared by different strains. Generally, the vaginal microbiota is composed of a single and almost clonal population of Lactobacillus and L. crispatus CIP104459 may be an exception. In addition, this strain was isolated in 1955 and is, thus, old, whereas bacteria, including L. crispatus, can rapidly evolve in response to host 23 and environmental 24 factors, especially by rearrangements within their genomic structure.
Here, we took advantage of the recent collection and draft genome sequencing of the vaginal strain L. crispatus V4, isolated in 2018 from a vaginal swab of a young healthy non-menopausal woman 11 . Comparison of the core genome of L. crispatus V4 to that of 122 L. crispatus strains show L. crispatus V4 and CIP104459 to be related to    www.nature.com/scientificreports/ distinct and separated clusters and L. crispatus V4 appears to be a more typical vaginal strain. The core genomes of L. crispatus V4 and CIP104459 only show 97.42% average nucleotide identity and a 78.70% genome-to-genome distance. The two strains are therefore close to the limit of 96.8%, at which point they should be considered as different sub-species or even species 25 . In addition, alignment of the two whole genome sequences showed the core genome to be limited to 77.96% of the total genome and a high number the ORFs encode proteins specific to one or the other bacterial strains. There are large genomic variations in L. crispatus for genes encoding for glycosylation and glycogen degradation enzymes, which are essential for interaction with the vaginal mucosa 9 . Thus, we examined potential differences between our two strains. Principal genes involved in glycogen metabolism, including glgX, encoding for type 1pullulanase, are highly conserved between the two strains. Given the genomic distance between the two strains, we were surprised to observe that they expressed a highly similar number of proteins under our experimental conditions i.e., when bacteria were grown in MRS medium. However, the protein pools expressed by L. crispatus V4 and CIP104459 were notably different. Indeed, V4 appeared to express a high number of proteins also produced by CIP104459 (81.1%), as well as a series of strain-specific proteins. In addition, the expression of common proteins was generally higher in V4 than CIP104459, suggesting a higher metabolic activity for V4. However, none of the typical glycogen-degradation enzymes essential for interaction with the vaginal mucosa was detected in the proteome of either strain using our method. Medium composition, pH, and temperature strongly affect enzymes expression of L. crispatus 26 . MRS medium used in the present study and classically used for Lactobacillus culture contains yeast and meat extracts but probably not enough glycogen to induce expression of specific degradation enzymes. Proteomic analysis also showed that 17β-estradiol has only a very minor effect on protein expression by L. crispatus V4. This result is coherent with the absence of an effect of 17β-estradiol on the growth kinetics of this strain, as previously observed with L. crispatus CIP104459 8 . Indeed, human hormones and neurohormones rarely affect the growth kinetics of human microbiota bacteria 27 . This is also coherent with clinical studies, which showed that the composition of the vaginal microbiota of women treated with estradiol-containing contraceptives is not significantly altered 28 .  www.nature.com/scientificreports/ All lactobacilli, and particularly L. crispatus, can spontaneously form aggregates 13 . In contrast to strain CIP104459, which shows only a limited potential for auto-aggregation 8 , L. crispatus V4 was capable to form dense aggregates, a phenotype that was increased by 17β-estradiol. The aggregation potential essentially depends on the outer surface properties of the bacteria, particularly S-layer composition 16 . Most lactobacilli have an S-layer, whereas the genes and protein sequences appear to be strain-specific 29 . This very likely explains the differences in strain behaviors in terms of biological activity, including aggregation and adhesion. Interestingly, flow cytometry studies showed the effect of 17β-estradiol to correlate with increased bacterial surface heterogeneity, i.e., presumably with S-layer re-arrangement. These observations are coherent with the results of the MATS studies, which showed that 17β-estradiol induced significant changes in the surface polarity of L. crispatus V4. The affinity of L. crispatus V4 for all solvents, particularly decane and hexadecane, was markedly increased by exposure to 17β-estradiol, showing globally higher surface hydrophobicity of the bacterium. In parallel, the Lewis acid/base properties of L. crispatus V4 evolved from basic to strongly acidic. S-layer proteins are generally hydrophobic 30 and this result suggests reinforcement of the bacterial crystalline envelope. Such behavior of L. crispatus V4 is entirely different from that observed with the strain CIP104459, in which 17β-estradiol has no, or a very limited, impact on aggregation and surface polarity 8 . However, it should be noted that, in sharp contrast to V4, CIP104459 shows low affinity to solvents and its surface is much more polar, suggesting that the composition of its S-layer is different or that it is incomplete or absent. The impact of 17β-estradiol on bacterial membrane fluidity was also very different. L. crispatus V4 showed only minor changes after exposure to 17β-estradiol, whereas L. crispatus CIP104459 reacts to the steroid by a significant decrease in the anisotropy index, indicating membrane alterations 8 . As an amphiphilic molecule, 17β-estradiol can integrate into membranes and destabilize their structure 18 . This phenomenon has been particularly observed in Gram-negative bacteria, such as Helicobacter pylori 18 and Pseudomonas aeruginosa 31 . In this regard, the absence of an effect of 17β-estradiol on the membrane integrity of L. crispatus V4 can be explained by the higher protection conferred by a complete outer S-layer envelope. Nevertheless, the lack of an effect on membrane fluidity of 17β-estradiol, administered during the entire growth phase or at various periods during growth, reinforces the hypothesis that, as previously suggested 8 , 17β-estradiol likely acts on the physiology of L. crispatus through a signal transductiondependent mechanism and not by non-specific membrane integration, at least at the concentrations studied. L. crispatus V4 exposed to 17β-estradiol showed, as noted for CIP104459, a decrease in the mean cell size. We also observed an increase of L. crispatus V4 adhesion to VK2/E6E7 human vaginal cells. In addition to aggregation, the S-layer plays a central role in bacterial morphology and adhesion to various targets [32][33] . However, S-layer composition was not the unique parameter affected by 17β-estradiol. As revealed by measurement of the surface tension of aqueous extracts of L. crispatus V4 colonies grown in the presence of 17β-estradiol, surface tension slightly decreased, suggesting an increase in biosurfactant production. Unexpectedly, the effect of 17β-estradiol on cell adhesion and biosurfactant activity was not no dose-related and there were no significant differences between values obtained after exposure to the hormone. This was also true for L. crispatus V4 aggregation, which showed an inverted dose-relationship for 17β-estradiol. One potential explanation is that the surfactant activity of biosurfactants is not directly related to their concentration. Indeed, their properties and interactions with the environment are different when under or above the critical micellar concentration 34 . Added to the fact that 17β-estradiol must be at the appropriate concentration to favor biosurfactant release and aggregation or adhesion to the target cells, as suggested for L. crispatus CIP104459 8 , the relationship between bacterial phenotype and dose may not be linear. As previously observed, the impact of 17β-estradiol on L. crispatus V4 biofilm formation was marginal. However, the vaginal epithelium is covered by cervicovaginal mucus 35 and Lactobacillus has no need to synthesize extracellular matrix to implant. Pathogens involved in bacterial vaginosis develop in biofilms 36 . Conversely, the installation and persistence of L. crispatus in mucosa appears to depend essentially on its aggregation phenotype 37 and, as demonstrated herein, 17β-estradiol stimulates this process.
These results are of importance, as they show that, although the L. crispatus V4 and CIP104459 strains belong to different genetic clusters and have different surface properties, the effect of 17β-estradiol on adhesion to vaginal mucosa cells is preserved in both. Thus, 17β-estradiol may have a similar effect on most of L. crispatus strains by promoting adhesion to vaginal cells or adsorption in the vaginal mucosa, without modifying the population dynamics. These observations are coherent with clinical studies showing that the vaginal microflora tends to re-equilibrate in favor of L. crispatus and other normal flora Lactobacillus for women under treatment with 17β-estradiol, such as administration during in vitro fertilization protocols 38 . Give the risk inherent to the carcinogenic activity of estradiol 39 , 17β-estradiol is not the ideal candidate drug for the development of a product to re-equilibrate the vaginal microbiota, but there are many other phytoestrogens that may show more specific activity on Lactobacillus.

Methods
Bacterial strain and culture conditions. Vaginal Lactobacillus crispatus V4 was obtained from a swab of a non-menopausal woman 11 . The sample was collected by the CRO Bio-EC (Longjumeau, France) and in agreement with the French and EU ethical guidelines (ARS Biomedical Research Agreement N°2012-12-010, Bioethics Agreement DC-2008-542). This strain was characterized by total proteome identification using a Bruker Autoflex III MALDI-TOF/TOF mass spectrometer coupled to the algorithmic software Biotyper. The draft genome sequence of L. crispatus V4 was determined using an Illumina MiSeq platform and deposited in the DDBJ/ENA/GenBank under accession number SRLG00000000 11 . The bacterium was grown anaerobically in De Man, Rogosa and Sharpe (MRS) medium (VWR, Fontenay-sous-Bois, France) at 37 °C under static conditions. Pre-cultures were prepared anaerobically in the same medium for 48 h to reach the stationary growth phase. Bacterial density of the suspension was determined by measuring the OD 600nm using a ThermoSpectronics  Sample preparation and protein digestion were performed as previously described 43 . For each extract, 25 μg of proteins was loaded onto a 7% polyacrylamide gel (Acrylamide/Bis-Acrylamide 30% [29:1], Sigma-Aldrich) and migration performed for a short period (90 min at 10-20 mA/gel). After Coomassie blue staining, the revealed protein bands were excised and first immersed in reductive buffer (5 mM DTT, Sigma-Aldrich) and then an alkylated buffer (20 mM iodoacetamide, Sigma-Aldrich). After washing, the gel bands were digested with 1 μg of trypsin (Promega, Madison, WI) overnight at 37 °C. Then, several steps of peptide extraction were performed using acetonitrile (Fisher, Hampton, NH) and the peptides dried and stored at − 20 °C.
All experiments were performed using an LTQ-Orbitrap Elite mass spectrometer coupled to an Easy nLC II system (both from Thermo Scientific). Samples were injected onto an enrichment column (Acclaim PepMap100, Thermo Scientific). The separation was achieved with an analytical column needle (NTCC-360/100-5-153, Nik-kyoTechnos). The mobile phase consisted of H 2 O/0.1% fluorhydric acid (FA) (buffer A) and acetonitrile(CAN)/ 0.1% FA (buffer B). Tryptic peptides were eluted at a flow rate of 300 nL/min using a three-step linear gradient: from 2 to 40% B over 76 min, from 40 to 80% B over 4 min, and 100% B for 10 min. The mass spectrometer was operated in positive ionization mode with the capillary voltage set to 1.7 kV and the source temperature 275 °C. The samples were analyzed using CID (collision-induced dissociation) method. The first scan (MS spectra) was recorded in the Orbitrap analyzer (R = 60,000) with a mass range of m/z 400-1800. Then, the 20 most intense ions were selected for MS2 experiments. Singly charged species were excluded from the MS2 experiments. Dynamic exclusion of already fragmented precursor ions was applied for 30 s, with a repeat count of 2, a repeat duration of 30 s, and an exclusion mass width of ± 5 ppm. The precursor isolation width was 2 m/z. Fragmentation occurred in the linear ion trap analyzer with a normalized collision energy of 35. All measurements in the Orbitrap analyzer were performed with on-the-fly internal recalibration (lock mass) at an m/z of 445.12002 (polydimethylcyclosiloxane).
Raw data files were processed using Proteome Discoverer 1.4 software (Thermo Scientific). Peak lists were searched using MASCOT search software (Matrix Science) against the Uniprot database. Database searches were performed using the following parameters: two missed trypsin cleavage sites allowed, variable modifications: carbamidomethylation on cysteine and oxidation on methionine. The parent-ion and daughter-ion tolerances were 5 ppm and 0.35 Da, respectively. The false discovery rate (FDR) threshold for identifications was set to 1% (for proteins and peptides).
For protein quantification, a label-free experiment was performed as previously described by Kentache et al. 44  www.nature.com/scientificreports/ other samples within the experiment were aligned. After alignment and normalization, statistical analysis was performed for one-way ANOVA calculations. For quantitation, peptide features with p-and q-values < 0.05 and a power > 0.8, were retained. MS/MS spectra from selected peptides were exported for peptide identification by Mascot (Matrix Science, version 2.2.04). Database searches were performed using the following parameters: one missed trypsin cleavage site allowed, variable modifications: carbamidomethylation of cysteine and oxidation of methionine. Mass tolerances for precursor and fragment ions were set at 5 ppm and 0.35 Da, respectively. FDRs were calculated using a decoy-fusion approach in Mascot (version 2.2.04). Identified peptide-spectrum matches with a − 10logP value of ≥ 13 were kept. Mascot search results were imported into Progenesis. For each growth condition, the total cumulative abundance of the protein was calculated by summing the abundances of peptides. Proteins identified with < 2 peptides were discarded. Analyses were performed in triplicate on independently prepared samples.
Bacterial aggregation studies. L. crispatus V4 grown in microplates showed the formation of dense aggregates. The aggregation potential of L. crispatus V4 was quantified by sedimentation 14 . Bacteria were grown in MRS at 37 °C for 18 h under anoxic static conditions with or without 17β-estradiol and harvested by centrifugation (7500×g, 10 min, 20 °C). After washing twice in phosphate buffered saline (PBS, Lonza, Thermo Fisher Scientific, Waltham, Massachusetts, USA) and resuspension in 10 mL of the same medium, the bacterial suspension was vortexed for 1.5 min. The OD 600nm of the suspension was measured 30 min later using a Thermo Fisher Scientific spectrophotometer (Waltham, Massachusetts, USA). The percentage of auto-aggregation was calculated as: where OD 0min is the initial OD 600nm at T = 0 and OD 30min is the final OD 600nm after 30 min. The aggregation potential and morphology of L. crispatus V4 were further studied by flow cytometry using a CytoFlex S flow cytometer (Beckman Coulter Life science, Indianapolis, USA) and CytExpert software. Control and 17β-estradiol treated 18-h bacterial cultures were harvested by centrifugation (7500×g, 10 min, 20 °C) and resuspended in PBS. Bacteria were immediately aliquoted, distributed in 96 wells microplates (Thermo Fisher Scientific, Waltham, Massachusetts, USA) and maintained under static conditions for 30 min before flow cytometry analysis. A minimum of 10,000 events at OD 488 ± 4 nm (SSC channel) and OD 525 ± 20 nm (FSC channel) was recorded at a flow rate of 10 µL min −1 in each condition. Data were analyzed using Cytexpert software. The aggregates correspond to the fraction of events appearing in the Q1-UR quarter of the graph (red zone). Particles size is given by the FSC-A (horizontal) axis. Isolated bacteria appear in the Q1-UL and Q1-LL zones. Granulometry (surface heterogeneity) is plotted in the SSC-A (vertical) axis. Increased granulometry appears in the Q1-UL and Q1-UR areas of the graph. Measures were realized in triplicate.
Investigation of bacterial ultra-surface properties. L. crispatus V4 ultra-surface properties were investigated by determining the surface polarity and Lewis acid-base characteristics using the microbial adhesion to solvents (MATS) technique 45 . Bacteria grown for 18 h, with or without 17β-estradiol, were collected as already described. Traces of culture medium were removed by rinsing twice in PBS. Four solvents were used chloroform, hexadecane, decane, and ethyl acetate. For each test, 1.2 mL of bacterial suspension at OD 400nm = 0.8 was mixed with 0.2 mL solvent. After incubation for 15 min and separation of the two phases, the OD 400nm of the aqueous phase was measured. Bacteria divide into the two compartments based on their surface affinity for water or organic solvents. The affinity was calculated using the equation: where [AO] is the OD 400nm of the aqueous phase without solvent and [A] the OD 400nm of the aqueous phase after exposure to the solvent. All experiments were conducted at least in triplicate.
Evaluation of the degree of bacterial membrane organization. Bacterial membrane fluidity, and therefore the degree of organization, was investigated by measurement of the fluorescence anisotropy, as previously described 19 . L. crispatus V4 was grown with or without 17β-estradiol for 18 h and harvested by centrifugation (7500×g, 10 min), washed twice in 10 mM MgSO 4 and stained with the fluorescent probe 1,6-diphenyl-1,3,5-hexatriene (DPH) (Sigma-Aldrich, Saint-Quentin-Fallavier, France) at 4 mM in tetrahydrofuran using a probe to sample ratio of 1/1000. The probe was allowed to incorporate into the membranes by incubating for 30 min in the dark at 37 °C. Then fluorescence polarization was continuously measured using a temperaturecontrolled Spark 20 M multimode microplate reader. The excitation and emission wavelengths of the probe were 365 and 425 nm, respectively. Each measurement was performed in triplicate. The membrane anisotropy index (r value) was calculated according to Lakowicz 46 . Data were analyzed using SparkControl software 2.1 (Tecan Group Ltd., Männedorf, Switzerland). Increases in anisotropy values indicate a decrease in membrane fluidity and vice versa. In a second series of experiments, bacteria grown for 6, 18 or 24 h in the absence of treatment were collected and subsequently exposed to 17β-estradiol or control medium in parallel with incorporation of the fluorescent probe. The anisotropy index was then continuously monitored over 3 h.
Scanning electron microscopy. The morphology of L. crispatus V4 was studied by scanning electron microscopy (SEM) using a TENEO VolumeScope microscope (FEI, Hillsboro, OR, USA). Bacteria grown for 18 h with or without 17β-estradiol were harvested by centrifugation (7500×g, 10 min) and fixed by immersion in 1 mL of 2.5% glutaraldehyde in 1 M PBS, pH 7.1, for 1 h. Samples were prepared by hexamethyldisilazane www.nature.com/scientificreports/ treatment and coating in a LEICA EM ACE600 sputter coater (Wetzlar, Germany) with a 25 nm thick layer of platinum alloy, as previously described 47 . The SEM was operated at 10 kV.

Bacterial adhesion to vaginal cells.
Adhesion of L. crispatus V4 was studied using the VK2/E6E7 mucosa vaginal cells line (ATCC CRL-2616). These cells were originally collected from the vaginal mucosal of a healthy pre-menopausal women. They were grown and propagated using keratinocyte serum-free medium (KSFM) (Thermo Fisher Scientific, Waltham, Massachusetts, USA), supplemented with 0.05 mg/mL bovine pituitary extract, 0.1 ng/mL human recombinant EGF, and 44.1 mg/mL calcium chloride, as recommended by the provider. VK2/E6E7 cells (80% confluence) were exposed to L. crispatus V4 at a multiplicity of index (MOI) of 100 bacteria/cell. Prior to infection, bacteria grown with or without 17β-estradiol for 18 h were collected by centrifugation, rinsed twice with PBS to remove any trace of the steroid, and resuspended in KSFM without antibiotics. After 1 h of interaction with the bacteria, the medium was removed and the VK2/E6E7 cells rinsed carefully with KSFM without antibiotics to withdraw all planktonic bacteria. Then, the vaginal cells were lysed by the addition of 0.1% triton X-100 in PBS. Preliminary studies showed that the cultivability of L. crispatus V4 was not affected by such treatment. The bacterial solution was then diluted in MRS medium and plated on MRS agar Petri dishes. The number of adherent bacteria was determined by the direct counting of L. crispatus colonies grown after 48 h at 37 °C under anoxic conditions. Investigation of biosurfactant production. L. crispatus V4 biosurfactant production was studied using the pendant drop method. As described by Meylheuc et al. 48 , the bacterial mat of L. crispatus V4 grown on normal or 17β-estradiol supplemented MRS-Agar Petri dishes was scraped off and resuspended in 15 mL Volvic water (selected for its neutral surface tension value). The solution was homogenized by vortexing for 3 min and centrifuged twice for 30 min (10,000×g, 4 °C) to remove all debris. The biosurfactant containing supernatant was collected and stored at 4 °C. Measurements were performed using a DSA30 tensiometer temperature controlled drop shape analyser (Kruss, Hamburg, Germany). The surface tension was calculated from the drop shape curvature analysis based on the Young-Laplace equation using the tensiometer drop shape analysis software 49 . Surface tension values were merged using the OD 600nm of the bacterial suspension collected by scrapping of the Petri dishes to account for the variations due to the biomass collected.
Biofilm formation study. The impact of 17β-estradiol on biofilm formation of L. crispatus V4 was evaluated by the crystal violet staining assay and confocal laser scanning microscopy. Crystal violet studies were performed following a procedure adapted from O'Toole 50 . After pre-culture for 48 h in MRS medium, the bacterial solution was adjusted to an OD 600nm = 0.1 and supplemented with 17β-estradiol to reach a final concentration of 10 -6 , 10 -8 , or 10 -10 M. The same volume (1 mL) was distributed into each well of flat glass bottom 24-well polystyrene plates (Falcon ® , Durham, USA). Plates were incubated at 37 °C without agitation for 48 h in a Whitley A85 anaerobic Workstation. After incubation, non-adhered bacteria were removed by aspiration of the medium and rinsing with physiological water (0.9% NaCl). The crystal violet solution (0.1% w/v in sterile pure 18.2 MΩ water) was distributed in microwells and allowed to stain the biofilm for 10 min. Then, the excess of dye was removed by washing with pure 18.2 MΩ water. Absolute ethanol was added to dissolve the crystal violet adsorbed in the biofilm matrix and onto bacteria and the OD 595nm of the solution measured using an automated Te-Cool plate reader (Tecan Group Ltd., Männedorf, Switzerland).
L. crispatus V4 is unable to form biofilms on glass bottom microplates in MRS medium. To visualize the biofilm structure by confocal microscopy, L. crispatus V4 was grown in a special medium for vaginal microflora lactobacilli, designated as "simulating genital tract secretion" (SGTS) medium 20 . 17β-estradiol, or an equivalent volume of ethanol in water, was added from the onset of the culture to the end of the experiment. The absence of effect of 17β-estradiol on L. crispatus V4 growth in SGTS medium was verified in preliminary studies (Suppl. Figure 1B). For confocal microscopy, pre-cultures were performed in normal MRS medium. At the end, bacteria were collected by centrifugation (7500×g, 10 min), re-suspended in normal (control) or 17β-estradiol supplemented SGTS medium at an OD 600nm = 0.1 and distributed (1 mL aliquots) into 24-well glass flat bottom microplates (Sensoplate, Greiner Bio-One, Germany). Biofilm was allowed to develop for 48 h under static conditions in an anoxic chamber (DW scientific, Bingley, UK). Nonadherent bacteria were removed by rinsing twice with PW and the biofilms stained with SYTO 9 Green Fluorescent Nucleic Acid Stain (Thermofisher, Waltham, Massachusetts, USA). Biofilms were visualized under an LSM 710 inverted confocal laser scanning microscope (Zeiss, Marly-le-Roi, France) equipped the Zen 2009 software package (version 12.0.1.362). The average biofilm thickness (μm), mean biomass volume (μm 3 /μm 2 ), and roughness coefficient were calculated over a minimum of 30 observations for each condition using COMSTAT2 software. All experiments were repeated at least of three times. Statistical analysis. The statistical significance of experimental values was evaluated using the Prism GraphPad online tool (https:// www. graph pad. com/ quick calcs/ ttest1/). Data were analyzed using unpaired (two sample) two-tailed t tests to calculate p values.
Ethics approval and consent to participate. Not applicable, this study did not involve a clinical trial. Consent for publication. All authors have read and approved the manuscript. LMSM accepts to be responsibility for the publication fees.