Recurrent Vulvovaginal Candidiasis; a dynamic interkingdom biofilm disease of Candida and Lactobacillus

Vulvovaginal Candidiasis (VVC) is the most prevalent Candida infection in humans affecting 75% of women at least once throughout their lifetime. In its debilitating recurrent form, RVVC is estimated to affect 140 million women annually. Despite this strikingly high prevalence, treatment options for RVVC remain limited with many women experiencing failed clinical treatment with frontline azoles. Further, the cause of onset and recurrence of disease is largely unknown with few studies identifying potential mechanisms of failed treatment. This study aimed to assess a panel of clinical samples from healthy women and those with RVVC to investigate the influence of Candida, vaginal microbiome and antagonism between Candida and Lactobacillus on disease pathology. 16S rRNA sequencing characterised disease by a reduction in specific health-associated Lactobacillus such as L. crispatus, coupled with an increase in L. iners. In vitro analysis showed Candida albicans clinical isolates are capable of heterogeneous biofilm formation and show the presence of hyphae and C. albicans aggregates in vaginal lavage. Additionally, the ability of Lactobacillus to inhibit C. albicans biofilm formation and biofilm-related gene expression was demonstrated. Using RNA sequencing technology, we were able to exploit a possible mechanism by which L. crispatus may aim to re-establish a healthy vaginal environment through amino-acid acquisition from C. albicans. This study suggests RVVC is not entirely due to an arbitrary switch in C. albicans from commensal to pathogen and understanding interactions between the yeast and vaginal Lactobacillus species may be more crucial to elucidating the cause of RVVC and developing appropriate therapies.


Importance
Fungal infections are becoming increasingly recognised as a substantial health 2 burden on the global population. Over 1 billion people are estimated to suffer from 3 fungal infections each year, resulting in over 1.5 million deaths [1]. These infections 4 are commonly associated with mucosal sites such as the vagina, gut and oral cavity. 5 Vaginitis is estimated to account for up to 7% of all visits to gynaecologists and for 6 up to 10 million general practitioner (GP) appointments annually [2]. VVC is not a 7 reportable disease and is often self-treated using over-the-counter antifungal agents, 8 subsequently the exact prevalence and distribution is impossible to determine. VVC 9 is reported as the second most common cause of vaginitis and it is estimated that 10 75% of women will suffer from VVC in their childbearing years with up to 140 million 11 of these women developing recurrent VVC (RVVC) annually, defined as ≥4 cases  The biofilm-forming yeast Candida albicans is reported as the predominant pathogen 19 responsible for up to 90% of VVC infections [4]. C. albicans is commonly isolated 20 from the vagina as an asymptomatic commensal with a carriage rate of up to 33% in 21 pre-menopausal women [7]. Other non-albicans Candida (NAC) species, including 22 C. glabrata, C. parapsilosis, C. dublinensis and C. krusei account for 10-20% of VVC 23 infections and are associated with complicated VVC, with more severe symptoms 24 and higher recurrence rates [8]. Isolated species in NAC VVC infections include C. 25 glabrata, C. krusei, C. parapsilosis, C. tropicalis and C. dubliniensis, with C. glabrata 1 being the most common [9]. Candida species, predominantly C. albicans, form thick 2 biofilms which dramatically increase fungal tolerance to drugs commonly used in the shown by Beyer and colleagues where the MAP kinase CgHog1 of C. glabrata was 5 upregulated in response to clinically relevant concentrations of lactic acid [22].

23
CVL supernatants were recovered by centrifugation for assessment of IL-8 levels by 24 Enzyme-Linked Immunosorbent Assay (ELISA) (Invitrogen, Paisley, UK), following 25 manufacturer's instructions. CVL was diluted 1:2 and absorbances were measured 1 using a spectrophotometer at 450nm and 570nm.       Antagonism of C. albicans and Lactobacillus in co-culture 14 The ability of the following 7 Lactobacillus strains to inhibit C. albicans SC5314   quantification of total C. albicans DNA per biofilm. All samples were incubated in the 24 dark for 10 min then placed on ice and exposed to a 650W halogen light for 5 min.     Figure 2). Bacterial load was found to be comparable at 25 ~8×10 7 CFE/mL between the two disease states (Figure 1c). Finally, quantities of 1 bacterial to fungal DNA were found to be lower in RVVC, confirming that patients 2 with disease have a higher fungal burden (Figure 1d).  and 3b). Interestingly, this reduction was coupled with an increase in L. iners from 21 just 19% in health to 40% in RVVC. Additionally, when predicted using random forest 22 plots, levels of L. iners were suggested to be the most distinct between health and 23 RVVC (Figure 3c).

24
To investigate the influence of bacterial taxa further, microbial populations were 1 observed with respect to patient metadata (Figure 4). In patients with culturable 2 Candida compared with those who were culture-negative, a reduction in 3 Lactobacillus species including L. crispatus from 44% to 29%, coupled with an 4 increase in L. iners from 23% to 35% was observed, similar to the RVVC profile 5 (Figure 4a). Further, L. iners was predicted to be the second-most likely organism to 6 define presence or absence of Candida using random forest analysis to identify   Table   9 3). In this study, we aimed to utilise whole-genome transcriptional sequencing of C.  Given the antagonism observed between the two organisms, we then aimed to 10 investigate the in vitro probiotic potential of L. crispatus against C. albicans infection 11 in a complex biofilm model (Figure 9). After 2 consecutive days of probiotic 12 treatment, a slight reduction in total and live C. albicans composition within the 13 biofilm is seen, however this was not significant (P = 0.55, P = 0.16, respectively) 14 ( Figure 9a and 9b). Following a 4-day treatment regimen with L. crispatus, total C.   Table 1). An increased pH in 15 RVVC, although not diagnostic, is seen in practice, likely due to a loss of  Table 1). It is important to note that these clinical 21 samples are from women regularly attending sexual health clinics, and those who 22 have suffered from RVVC for longer may be more likely to be receiving antifungal 23 treatment. This should be taken in to account when interpreting Candida CFE/mL 24 data observed in this study as untreated RVVC may present differently. Unlike vaginal infections such as BV, the microbial communities present during VVC 2 have been shown to be similar with those present in health at the phylum and genus-3 level [43,44]. Our study confirms these findings where we report a Lactobacillus-4 dominated population with vaginal anaerobes including Gardnerella, Prevotella and 5 Atopobium, with no significant differences in diversity or composition between the 6 two cohorts at the genus-level (Figure 2). This suggests that the functional capacity  Figure 3). We observed a reduction in specific Lactobacillus 14 species, including those associated with maintaining health due to their ability to 15 produce L-lactic acid and H2O2, such as L. crispatus and L. jensenii (Fig 4). This 16 reduction coupled with an increase of L. iners has been shown previously and is 17 thought to be indicative of vaginal dysbiosis [43]. We investigated this phenomenon 18 further with respect to individual patient metadata (Fig 5). We report that this loss of 19 health-associated Lactobacillus species and increase of L. iners is also seen 20 specifically in patients with positive Candida cultures and those who have suffered 21 from RVVC for >6 months. These findings may be important for future studies 22 investigating RVVC. It is hypothesised that changes within Candida allow for it to 23 switch from asymptomatic commensal to pathogenic yeast. It may now be more 24 important to study RVVC with specific focus on the microbes present during RVVC, 25 specifically Lactobacillus, and interkingdom interactions which may influence this 1 pathogenic switch in Candida.

3
Consistent with epidemiological reports, we found >70% of isolates from our cohort 4 were C. albicans with the remainder comprised primarily of C. glabrata (Figure 6).   Finally, we investigated the potential probiotic effect of L. crispatus against C. 24 albicans within a complex biofilm model (Figure 9). Although studies have assessed

Author Contributions
EM, LS, RK, CD and SW participated in study design and experimental procedures. EM, LS, RK and CD were responsible for reparation of the manuscript. RM and RT were responsible for clinical sample collection and collection of patient metadata. CW participated in study design and contributed to the manuscript. GR conceived the study, participated in study design and was responsible for producing the final manuscript. All authors have read and approved the final manuscript.

Competing interests' statement
The authors declare no conflict of interest.

Correspondence
Address correspondence to Gordon Ramage, gordon.ramage@glasgow.ac.uk. To asses fungal load, patient lavage was incubated on Candida Chromogenic agar plates and colonies counted after 48h (a). For calculation of CFE/mL, the ITS region of Candida was amplified using genus-specific primers (b). Levels of bacteria DNA were also assessed molecularly by amplification of the 16S rRNA region (c); Correlation between bacterial and fungal burden was also observed (d). Data represents the mean ± SD (*, P < .0.05, **, P < 0.01); Statistical significance was calculated using unpaired t-tests with Welch's correction as data did not share equal standard deviations.  lavage from a patient with a HBF isolate. White arrows represent pseudo-hyphal/hyphal formation and red arrows depict cell aggregates. Data represents the mean ± SD (****, P < 0.0001), statistical analysis was performed using unpaired t-tests. To observe inhibitory effects of Lactobacillus against C. albicans biofilm formation, C.
albicans and a panel of Lactobacillus species were co-cultured together in THB/RPMI (1:1) media in 5% CO2 either for 24h (a) or C. albicans was grown for 4h prior to addition of Lactobacillus species for 20h (b). C. albicans biofilm-associated gene expression was measured in the presence L. rhamnosus, which is associated with 'health', and L. iners, which is hypothesised to indicate dysbiosis. The mean log fold-change relative to single species C. albicans biofilms is shown (c). Data represents the mean + SD. Heatmap displaying the top 50 significantly differentially expressed genes in C. albicans between single and dual species 24h biofilms (P < 0.05). crispatus (c). Nodes are coloured by significance. All GO terms have an adjusted P value < 0.05. Networks were created using ClueGO. The potential probiotic properties of L. crispatus against C. albicans were assessed in a 11species biofilm model treated twice daily with L. crispatus. Live/Dead qPCR allowed for quantification of percentage composition of total (a) and live (b) C. albicans DNA within the biofilm. Average fold change in the C. albicans percentage composition from the untreated 11-species biofilm is also shown (c). Data represents mean + SD. Statistical analysis was performed using paired t-tests comparing raw CFE values (*, P < 0.05).
36 Tables   Table 1: Up-regulated genes in 24h dual-species biofilm associated with amino acid biosynthesis and/or breakdown