In vitro bacterial vaginosis biofilm community manipulation using endolysin therapy

Bacterial vaginosis (BV) affects approximately 26% of women of childbearing age globally, presenting with 3–5 times increased risk of miscarriage and two-fold risk of pre-term birth. Antibiotics (metronidazole and clindamycin) are typically employed to treat BV; however the success rate is low due to the formation of recalcitrant polymicrobial biofilms. As a novel therapeutic, promising results have been obtained in vitro using Gardnerella endolysins, although to date their efficacy has only been demonstrated against simple biofilm models. In this study, a four-species biofilm was developed consisting of Gardnerella vaginalis, Fannyhessea vaginae, Prevotella bivia and Mobiluncus curtisii. Biofilms were grown in NYC III broth and treated using antibiotics and an anti-Gardnerella endolysin (CCB7.1) for 24 h. Biofilm composition, viability and structure were assessed using colony counts, live/dead qPCR and scanning electron microscopy. All species colonised biofilms to varying degrees, with G. vaginalis being the most abundant. Biofilm composition remained largely unchanged when challenged with escalated concentrations of conventional antibiotics. A Gardnerella-targeted endolysin candidate (CCB7.1) showed efficacy against several Gardnerella species planktonically, and significantly reduced viable G. vaginalis within polymicrobial biofilms at 1 to 4X pMIC (p < 0.05 vs. vehicle control). Collectively, this study highlights the resilience of biofilm-embedded pathogens against the currently used antibiotics and provides a polymicrobial model that allows for more effective pre-clinical screening of BV therapies. The Gardnerella-specific endolysin CCB7.1 demonstrated significant activity against G. vaginalis within polymicrobial biofilms, altering the overall community dynamic and composition.


Introduction
Bacterial vaginosis (BV) affects approximately 26% of women of childbearing age globally [1,2], and has a profound psychological impact on patients with significant consequences for prenatal health. The condition is characterised by compositional alterations in the vaginal microbiome, resulting in a shift from a lactobacilli-dominated microbiome to one composed largely of BV-associated anaerobesincluding Gardnerella spp. (G. vaginalis, G. leopoldii, G. piotii, G. swidsinskii), Fannyhessea vaginae, Prevotella bivia and Mobiluncus curtisii [3][4][5][6]. This microbial shift gives rise to associated symptoms including foul odour, burning during urination, vaginal discharge and itching. Furthermore, there are strong links between BV and risk of miscarriage, pre-term birth and sexually transmitted infections [7].
During the onset of BV, anaerobes such as Gardnerella spp. displace commensal lactobacilli to form biofilms on the vaginal epithelium [8]. This results in an elevated pH and colonisation of a diverse polymicrobial community, contributing to BV onset, progression and treatment failure [9]. Current treatments for BV are extremely limited and most cases are treated using antibiotic therapy (e.g. metronidazole and clindamycin). However, antibiotic regimens have an unacceptably large rate of relapse with >50% of patients suffering from BV recurrence within 6 months [10], believed to be driven, in-part, by the recalcitrance of these microbial biofilms [4]. As such, there is an unmet need to develop novel therapies for BV and evaluate their efficacy against biofilm-embedded bacteria.
One promising avenue in this regard is the use of bacteriophagederived endolysin proteins which are produced during the lytic phase of phage replication, and act to degrade peptidoglycan from within bacterial cells to allow release of progeny [11]. Given this mechanism of action, these enzymes have good promise as therapeutics to be harnessed by clinicians. Previous studies have shown these enzymes are capable of lysing Gram-positive bacteria which lack an outer lipopolysaccharide membrane when added exogenously to culture media and more complex models of disease [12][13][14]. Interestingly, endolysins have also shown promise as antimicrobial agents against Gram-negative bacteria [15], despite their diderm cell wall structure. In addition, the targeted nature of bacteriophages and corresponding endolysin proteins mean they display high specificity for organisms, lending themselves to selective degradation of pathogens within complex microbial ecosystems without disruption of commensal flora [11]. Collectively, these features have resulted in the development of endolysins which target Gardnerella spp. -predominant members of BV-associated biofilms which contain a thin peptidoglycan layer [6,16].
Thus far, several endolysins directed against Gardnerella spp. have demonstrated good activity in vitro. For example, one endolysin (PM-477) was found to be highly selective for Gardnerella spp. [14], capable of eliminating mono-species G. vaginalis biofilms [17] and significantly reducing G. vaginalis load in dual-species biofilm models [18]. Likewise, recent screening of Gardnerella genome sequences has identified a library of endolysin candidates which prevent G. vaginalis biofilm formation and degrade pre-formed mono-and dual-species biofilms [12]. This library comprised 84 endolysin candidates which are diverse at the sequence and structural levels, with 3 different enzymatic active domains and 4 cell wall binding domains, with convergent bioactivities. Of these, several candidates were found to be particularly active against G. vaginalis mono-species biofilms, including CCB7.1 which comprises a C-terminal SH3b cell-wall binding domain and N-terminal LysA-like GH25 enzymatic active domain.
Whilst promising results have been obtained thus far, the screening of these proteins in vitro has been limited to planktonic organisms or simple (mono-, dual-species) biofilm models, which do not truly capture the complexity and diversity of the microenvironment in BV. As such, the current project aimed to build upon previous expansions in this field by developing a multi-species biofilm model that can be used for more robust high-throughput screening of novel compounds in BV treatment. The model consisted of four anaerobes frequently isolated in BV (G. vaginalis, F. vaginae, P. bivia and M. curtisii), with each species previously shown to exhibit monospecies biofilm formation [19]. Notably, both F. vaginae and P. bivia have been reported to influence Gardnerella biofilm-related gene expression in vitro [20], with the presence and persistence of M. curtisii associated with BV recurrence [5]. This biofilm model was initially assessed against current front-line antibiotics (metronidazole, clindamycin) and subsequently against a Gardnerella targeted phage endolysin candidate -CCB7.1 [12]. We sought to characterise the efficacy of this protein against our more complex polymicrobial system, lending further credence to the clinical utilisation of endolysin proteins.
Liquid cultures were prepared by inoculating colonies into New York City III broth supplemented with 10% (v/v) inactivated horse serum (herein termed NYC III), which has been previously shown to be the optimal growth medium biofilm formation of BV-associated anaerobes [19]. Medium was incubated in the anaerobic cabinet 1 day prior to use. All bacterial culture was performed at 37 • C anaerobically within an anaerobic cabinet (Don Whitley Scientific MACS MG-500) with anaerobic gas influx (5% CO 2 , 10% H 2 and 85% N 2 ).

Drug preparation
Endolysin candidates were supplied by CC Biotech Ltd. in buffer 1. Antibiotics were purchased as powders (Sigma-Aldrich, Gillingham, UK) and prepared to stock concentrations of 5000 μg/mL and 10000 μg/mL in 100% ethanol and sterile distilled water for metronidazole and clindamycin, respectively. Stock solutions were freshly prepared for each set of experiments and aliquots stored at − 20 • C between technical replicates. Working concentrations were prepared by dilution in NYC III medium (antibiotics) or buffer 1 where appropriate.

Planktonic MIC testing
Planktonic minimum inhibitory concentrations (pMICs) for antibiotics (clindamycin, metronidazole) and endolysins were performed using the CLSI M11-A8 broth microdilution method in NYC III, as reported previously for G. vaginalis [12,14]. The same experimental protocol was followed for G. vaginalis ATCC 14018, G. vaginalis UG860107, G. swidsinkii CCUG 72429T and G. piotii CCUG 72427T. In brief, bacteria were initially standardised to 1 × 10 8 CFU/mL in NYC III, and then further diluted to 1 × 10 6 CFU/mL in sterile medium (1:100). Antibiotics and endolysins were serially diluted 1:2 in sterile medium. Drugs were initially added at 100 μL to 96-well round bottom microtiter plates at double the required concentration, followed by 100 μL of 1 × 10 6 CFU/mL G. vaginalis. The final bacterial concentration in wells was therefore 5 × 10 5 CFU/mL as is recommended for pMIC testing [23]. Untreated controls (bacteria only) and negative controls (NYC III broth) were included in each assay run. Microtiter plates were incubated for 24 h anaerobically at 37 • C, with pMIC values determined by resuspending pellets and measuring well absorbance at OD 550 using a microtiter plate reader (BMG labtech, Ortenberg, Germany). Results are presented as a percentage inhibition in comparison to untreated controls, with MIC 90 values ≥ 90% inhibition.

Mono-species biofilm formation and treatment
Mono-species G. vaginalis biofilms were prepared by adding 200 μL of 1 × 10 7 CFU/mL in NYC III into flat bottomed 96-well microtiter plates. Plates were incubated for 24 h anaerobically at 37 • C, before medium was removed and 200 μL fresh NYC III was applied for 24 h. After incubations, 200 μL of the desired antibiotic concentration (8-512 μg/mL for metronidazole, 0.0625-4 μg/mL for clindamycin) was applied wells and biofilms were incubated for a further 24-h anaerobically at 37 • C. Each plate contained untreated biofilms and negative controls (NYC III broth) to ensure sterility of culture medium. After treatment, colony forming units (CFU/mL) analysis was performed on mono-species G. vaginalis biofilms. For this, antibiotics were removed and biofilms were gently washed twice with sterile PBS. After washing, 100 μL of PBS was applied to wells and biofilms removed by scraping. Colony counting was performed using the Miles and Misra technique [24], in which serially diluted biofilms were applied in triplicate (20 μL) to Columbia blood agar plates. Plates were incubated for 48 h anaerobically, and CFU/mL estimation was performed by using the average of triplicate values.

Multi-species biofilm formation
A four species BV-associated biofilm model was grown using a 'G. vaginalis primed system', whereby this organism is allowed to colonise for 24 h, followed by addition of remaining bacteria (Fig. 1). This is similar to previously developed biofilm models in which G. vaginalis was initially cultured in isolation, followed by additional organisms [18,25,26]. In brief, G. vaginalis was standardised to 1 × 10 8 CFU/mL, and diluted 1:10 in NYC III broth to 1 × 10 7 CFU/mL. From here, 500 μL was applied to wells of a 24-well microtiter plate containing sterile 13 mm Nunc™ Thermanox™ coverslips (Fisher Scientific, Loughborough, UK). Microtiter plates were incubated for 24 h under anaerobic conditions at 37 • C. After incubation medium was removed and F. vaginae, P. bivia and M. curtisii were standardised separately to 1 × 10 8 CFU/mL. Each species was then combined to create a cocktail containing 1 × 10 7 CFU/mL final concentration of each bacteria. Plates were incubated for a further 24 h under anaerobic conditions at 37 • C. On day 3, treatments were applied as described below.

Polymicrobial biofilm treatment
For polymicrobial biofilms, antibiotics were diluted in NYC III medium to desired concentrations (8 and 64 μg/mL metronidazole, 0.0625 and 0.5 μg/mL clindamycin). From here, 500 μL of each antibiotic was added to biofilms, alongside untreated (medium) controls. A similar protocol was followed for the CCB7.1 endolysin candidate, which was diluted to desired concentrations in buffer 1 to 128, 256 and 512 μg/mL, and added to biofilms alongside buffer 1 controls. For all treatments, plates were incubated for a further 24 h under anaerobic conditions at 37 • C.

DNA extraction and PMAxx treatment
Following multispecies biofilm treatment, coverslips were gently washed three times with sterile PBS and placed into bijou bottles containing 1 mL PBS. To remove bacteria from coverslips, the bijou bottles were sonicated at 35 kHz for 10 min, after which the sample was split into two 500 μL aliquots for live/dead quantification. One of the samples (live cells) was treated with PMAxx dye (Biotium, California, USA) as per manufacturer's instructions, with the other aliquot (total cells) not receiving PMAxx. After addition, treated samples were gently flicked to mix PMAxx dye and incubated in the dark for 10 min at room temperature as previously described for in vitro biofilms [22,27,28]. All samples were then exposed to the PMA-lite device (Biotium) for 15 min. After PMAxx treatment, DNA was extracted from all samples using the DNeasy kit (QIAGEN, Manchester, UK) as per manufacturer's instructions.

Quantitative PCR (qPCR)
All qPCR reactions were performed on an Applied Biosystems ViiA 7 real-time PCR system using 20 μL reactions. Each reaction mixture consisted of 10 μL 2x PowerUp SYBR green mastermix (Fisher Scientific), 7 μL Hyclone molecular grade water, 1 μL forward primer (10 μM), 1 μL reverse primer (10 μM) and 1 μL of DNA. This resulted in final primer concentrations of 0.5 μM per reaction. After preparation, plates were run with the following cycling conditions on the ViiA 7 analyser; 50 • C for 2 min, 95 • C for 2 min, 40 cycles of 95 • C for 3 s followed by 60 • C for 30 s. Primer sequences used throughout this study targeting each species or genus are highlighted in Table 1. For quantification of each organism, a standard curve was prepared using DNA extracted from known concentrations of the target bacteria, with colony forming Fig. 1. Timeline for growth of polymicrobial biofilm formation in vitro, depicting incubation times and conditions. AnO 2 = anaerobic incubation. Biofilms were grown using a 'G. vaginalis primed' system, whereby this organism is allowed to colonise for 24 h, followed by additional BV-associated anaerobes (F. vaginae, M. curtisii, P. bivia). The four species are incubated together for a further 24 h before treatment is applied. Figure created using BioRender. equivalents per mL (CFE/mL) calculated as previously described [29].

Scanning electron microscopy
To assess the ultrastructure of polymicrobial biofilms formed on coverslips, scanning electron microscopy (SEM) was used as previously described [27,31]. Briefly, biofilms were fixed using 2% paraformaldehyde, 2% glutaraldehyde, 0.15% alician blue power and 0.15 M sodium cacodylate. This was followed by counterstaining with uranyl acetate and gradient dehydration in ethanol (30-100%). Samples were then sputter-coated using gold/palladium and visualised using a JEOL JSM-6400 SEM machine at 1000x and 5000x magnification (JEOL Ltd).

Data analysis
Data was analysed and graphs compiled using a combination of Microsoft Excel (version 16.60) and GraphPad PRISM (version 9). Normality tests were performed and appropriate statistical analysis performed. For example, non-parametric data was analysed using Mann-Whitney U test or Kruskall-Wallis with Dunn's post-hoc where appropriate. Statistical significance was defined as p < 0.05 throughout.

Antibiotic treatment of monospecies G. vaginalis biofilms
As mentioned, frontline therapy for BV remains largely limited to two antibiotics. A common pitfall of these treatments is the large discrepancy between bacterial susceptibility in a planktonic and biofilmembedded state. To understand this in the context of BV, we characterised the efficacy of clindamycin and metronidazole against G. vaginalis ATCC 14018. As expected, this strain was susceptible to both antibiotics when assessed planktonically. The pMIC concentrations were 8 μg/mL ( Fig. 2A, resistance breakpoint ≥32 μg/mL) and 0.0625 μg/mL (Fig. 2B, resistance breakpoint ≥8 μg/mL) for metronidazole and clindamycin respectively, which is in line with previous publications [14,32].

Polymicrobial BV biofilm development
Based on current understanding, it is proposed that G. vaginalis may be the primary coloniser of BV associated biofilms [33,34], followed by attachment of additional BV-associated anaerobes. To recapitulate this in vitro, we thus sought to explore a 'G. vaginalis primed'biofilm model, which has been previously used in dual and more recently tri-species biofilm models of BV [20,25]. In this system G. vaginalis was initially incubated on coverslips for 24-h, followed by addition of F. vaginae, M. curtisii and P. bivia for 24-h.
Following biofilm development, the ultrastructure of polymicrobial biofilms was investigated using SEM at 1000x and 5000x magnification (Fig. 3A). As expected, this imaging revealed the formation of a dense, three-dimensional biofilm, characterised by large clusters of interlinked bacteria. Using live/dead qPCR, we confirmed that all input species colonised the resulting biofilm (Fig. 3B), where G. vaginalis was the most abundant (87.6% of total cells, 66.8% of live cells, Fig. 3C), followed by F. vaginae (11.8% of total cells, 32.7% of live cells). In contrast, both P. bivia and M. curtisii were less abundant with roughly 1 × 10 4 CFE/mL live cells, comparable to previously published dual-species models whereby G. vaginalis dominates [20,26]. These data are expected given that a G. vaginalis initiated system was utilised in this study. Nonetheless, recoverable levels of each organism were detectable in polymicrobial biofilms.

Antibiotic treatment of polymicrobial BV biofilms
To investigate how this polymicrobial model responded to antibiotic therapy, both metronidazole and clindamycin were applied at 1x and 8X pMIC of G. vaginalis. These concentrations were selected based on monospecies biofilm experiments -where minimal alterations were observed following 1X pMIC, with more pronounced reductions in viability at 8X pMIC of each antibiotic. Treatment was applied to polymicrobial biofilms for 24-h, and the viability of each species was assessed using live/dead qPCR (Fig. 4).
Following antibiotic treatment, there were no significant alterations to the total number of G. vaginalis cells compared with untreated controls (Fig. 4A). No reduction was also observed when assessing live G. vaginalis cells; this contrasted with mono-species biofilm treatment, where live cell count was significantly impacted by the antibiotics. Despite a subtle reduction in the median live G. vaginalis quantity at 8X pMIC metronidazole, this did not reach statistical significance in comparison to untreated biofilms (adjusted p > 0.05). These findings were consistent for both P. bivia (Fig. 4C) and M. curtisii (Fig. 4D) where no significant alterations in total or live cells were found (adjusted p > 0.05). The only significant difference was a reduction in live F. vaginae at 1X and 8X pMIC metronidazole (adjusted p = 0.01 and 0.0002 respectively), the latter of which also reduced the total F. vaginae load (adjusted p < 0.001, Fig. 4B). Conversely, no significant alterations in total or live F. vaginae were found following clindamycin therapy.

Endolysin treatment of polymicrobial BV biofilms
Data from our polymicrobial model showed no, or limited, alterations in biofilm viability following antibiotic therapy. As such, we sought to evaluate the efficacy of engineered phage endolysins which were developed to target Gardnerella spp., and represent a promising novel therapy in combatting BV-associated biofilm communities.
Although previous studies have documented success using these antibacterial proteins in vitro, we aimed to build upon this data and trial one candidate (CCB7.1) against our more complex system. As expected, this protein was effective against the G. vaginalis ATCC 14018 when assessed planktonically. The pMIC value against this isolate was 128 μg/  (Fig. 5A), corresponding with previously published values [12]. Furthermore, we also observed efficacy against additional Gardnerella isolates such as an additional G. vaginalis strain (64 μg/mL), G. swidsinskii (32 μg/mL) and G. piotii (64 μg/mL) (Fig. 5B).
The endolysin candidate was next applied to pre-formed polymicrobial biofilms, to investigate its efficacy against biofilmembedded G. vaginalis alongside additional BV-associated anaerobes. Three concentrations of endolysin were applied, representing 1X, 2X and 4X pMIC (128, 256 and 512 μg/mL). After addition, we observed a significant reduction in the viable levels of G. vaginalis using live/dead qPCR (adjusted p < 0.05 for all concentrations, Fig. 5C). This represented a 1-2 log 10 reduction in viable G. vaginalis at all tested concentrations, compared with untreated controls. Although reduced, it is noteworthy to mention that G. vaginalis remained viable at ≈1 × 10 6 CFE/mL. Promisingly, the reduction was consistent, and more pronounced than observed for the current standard of care antibiotics against G. vaginalis. Assessment of additional BV-associated anaerobes demonstrated a reduction in live M. curtisii at the highest tested concentration (adjusted p = 0.007), although this species was the least abundant in the model system. Such a reduction may be due to biofilm destabilisation from reductions in G. vaginalis, rather than direct bacteriolytic activity. Endolysin treatment did not impact the total level of any species, which notably included G. vaginalis. Endolysin treatment did not alter total cell counts within the biofilm, although the live bacterial load reduced ≈1-log 10 , which was almost entirely due to G. vaginalis (Fig. 6A and B). From a compositional perspective, G. vaginalis equated to >95% of live cells in untreated controls, which reduced to 78.3%, 79.9% and 42.5% with increasing doses of CCB7.1. Notably, a concomitant increase in the percentage abundance of live F. vaginae (Fig. 6B) was observed in the model alongside the reduction in G. vaginalis abundance. Although F. vaginae did not significantly shift in comparison to untreated controls, its lack of reduction equated to a larger proportion of the live bacterial load when G. vaginalis was reduced.

Discussion
Metronidazole and clindamycin are first-line antibiotic therapies for BV, applied topically or given orally. Whilst these antibiotics provide good coverage of BV-associated anaerobes [35], such therapies are disappointing and appear insufficient for many patients with evidence demonstrating a recurrence rate of >50% [10,[35][36][37]. This is not only hugely debilitating for patients who suffer from recurrent BV, but also highlights a distinct lack of alternative therapies available. One reason believed to drive these high rates of recurrence is the formation and recalcitrance of BV-associated biofilms which are inadequately eradicated by existing antibiotic therapy [4]. It is therefore essential that existing and novel therapeutics targeting BV-associated bacteria are assessed using in vitro models which are more reflective of this biofilm-based clinical presentation.
Accordingly, in this study we first sought to develop a polymicrobial biofilm-based model which is more replicative of clinical dysbiosis, and harness this model to investigate the impact of polymicrobial interactions on antimicrobial tolerance in BV. As expected, mono-species G. vaginalis biofilms were more tolerant of antibiotics (metronidazole, clindamycin) compared with planktonic cultures. Interestingly, our G. vaginalis-containing, four-species model showed almost complete resilience to current antibiotics prescribed for BV (metronidazole, clindamycin), with no statistical change in the composition of G. vaginalis cells even at the highest concentrations tested. Thus, whilst G. vaginalis mono-species biofilms serve as a useful preliminary screening tool, it is clear that this is not the only bacterium involved in BV-associated biofilmsand the presence of additional species such as F. vaginae, P. bivia and M. curtisii may alter biofilm susceptibility to antimicrobial challenge. In this regard, recent work has demonstrated that polymicrobial interactions in vitro facilitate increased resistance to metronidazole and clindamycin [32], supporting these results and providing further evidence for the involvement of recalcitrant polymicrobial biofilms in BV treatment failure.
It is also worth considering the molecular approach used to quantify polymicrobial biofilms in the current study may further increase controls. Each antibiotic was additionally screened against mono-species G. vaginalis biofilms (C), and viability presented as CFU/mL. Graphs display mean ± standard deviation (A,B) or median ± 95% CI (C). Statistics represent Kruskall-Wallis with Dunn's post-hoc, where Asterix above bars indicates a significant difference with the untreated control. Connecting lines represents direct comparison between clindamycin and metronidazole performed using Mann-Whitney tests adjusted using the Holm-Sidak method where *p < 0.05 and **p < 0.01 and ***p < 0.001. MTZ; metronidazole, CLI; clindamycin, LoD; limit of detection. All experiments are n = 3. viability estimates in comparison with colony counting. Specifically, the ability of live/dead qPCR to detect viable but non-culturable cells (VBNCs) and persister cells within biofilms, which would not be detectable using conventional colony counting culture methodology routinely used for mono-species analysis. As such, colony counting more accurately reflects the presence of culturable cells within biofilms, rather than viable cells (including both culturable and non-culturable cells). This has been previously suggested in the context of wound biofilm models comparing colony counting and qPCR [38], and it is likely that this would extend to BV-associated biofilms used in this study.
Following antibiotic treatment, we then sought to test the impact of one novel BV therapy, a bacteriophage-encoded endolysin, upon our model. Given the limited therapeutic options for BV treatment, there has been recent interest in the use of endolysin proteins as a novel alternate therapeutic. These proteins are naturally encoded within bacteriophage genomes, and offer highly selective lysis of target bacteria when recombinantly produced. Previous research has documented promising results following endolysin treatment of biofilm-embedded G. vaginalis [12,14,18]. Herein, we focused on one of these aforementioned endolysin candidates, CCB7.1, which was selected based on its previously elucidated ability to prevent G. vaginalis biofilm formation and disrupt pre-formed G. vaginalis biofilms [12]. Importantly, and as expected, CCB7.1 was previously shown to be ineffective against commensal lactobacilli [12] offering selective therapy of Gardnerella spp. without off-target impacts on beneficial flora.
After first demonstrating that CCB7.1 was effective against additional, and more clinically representative Gardnerella isolates (G. vaginalis, G. swidsinskii, G. piotii), we evaluated the impact of this targeted Gardnerella endolysin against a more complex, BVrepresentative consortium using our polyspecies BV model. Our data demonstrate a significant reduction in the viability of G. vaginalis within this model following treatment with the engineered phage endolysin CCB7.1 at three concentrations (128, 256 and 512 μg/mL), representing a 1-2 log 10 reduction in viability compared with untreated controls.  Thus, whilst the antimicrobial activity of CCB7.1 is encouragingly maintained when G. vaginalis is embedded in a more diverse in vitro biofilm, the magnitude of impact was less than observed in mono-and dual-species biofilm systems [12].
While it was generally expected that the efficacy of endolysin treatment would reduce in a multi-species model compared with monoculture systems, this relative reduction in efficacy compared favourably to the incumbent standard of care antibiotics metronidazole and clindamycin. One reason for reduced endolysin efficacy may result from the formation of a G. vaginalis primed biofilm model used in this study. Specifically, priming biofilms with G. vaginalis will likely result in partial coverage by secondary bacteria (F. vaginae, M. curtisii, P. bivia) which prevent immediate contact between lysins and G. vaginalis. This more accurately mimics the proposed model of polymicrobial biofilm formation in BV where G. vaginalis forms the initial biofilm scaffold [8,25], and it is thus reassuring that CCB7.1 remained capable of reducing G. vaginalis viability when grown in this system.
Taken together, the data obtained in this study underline the importance of using polyspecies models for in vitro evaluation of emerging therapeutic candidates. It is important that more clinically representative data outputs are obtained at earlier stages of preclinical development thereby reducing attrition rates later in the clinical development cycle. Indeed, some researchers have recently started to transition beyond simple BV biofilm models [20,32], and this work, in combination with the model developed in this current study, will form a useful basis for more accurate screening of BV therapies moving forward.
Although the evidence to date suggests that endolysin therapy is promising for BV, such a targeted approach is not devoid of disadvantages and an important consideration moving forward will be the implications of selective knockdown of this organism in vivo. For example, it is known that BV associated biofilms are polymicrobial in nature and knockdown of Gardnerella spp. spp. may result in enrichment of nontarget species. This may be of particular relevance given that F. vaginae was recently reported to display a larger inflammatory burden on a 3D human cervix model when compared with G. vaginalis [39]. This finding may lend itself to the use of endolysin cocktails which can target several BV-associated anaerobes, in a similar fashion to the approach used in bacteriophage therapy [40]. However the development of endolysins against P. bivia and other Gram-negative bacteria represents a continual challenge in the field [41].
A limitation of this study is that only a single method (live/dead qPCR) was used to characterise polymicrobial biofilm viability and composition. Although the propoidium monoazide based assay is widely reported for in vitro polymicrobial biofilms [22,28,[42][43][44], the use of additional techniques may provide a greater insight into the biofilm structure and bacterial organisation following endolysin therapy. In this regard, one technique which could be utilised to further characterise these biofilms is fluorescence in situ hybridisation using peptide nucleic acid probes (PNA FISH). This technique offers the benefit of complimentary imaging and quantitative assessment of biofilms, and PNA FISH probes have been developed and optimised for both G. vaginalis and F. vaginae [20,45,46]. Ensuing research may seek to use a combination of approaches for biofilm characterisation to provide simultaneous assessment of species-specific viability, biofilm structure and composition.
In conclusion, this study outlines the development of a novel polymicrobial in vitro biofilm system which is more replicative of clinical BV. The bioactivity of metronidazole, clindamycin and CCB7.1 were then assessed against this more complex model. Our data demonstrate tolerance of this model against standard of care antibiotic therapy, and a reduction in G. vaginalis viability following endolysin application. Although significant, these viability impacts following endolysin administration were less pronounced than previously observed against mono-species biofilms, which reiterates the importance of using more diverse and clinically relevant biofilms for screening therapies in BV.
Whilst the microbiota in BV extends well beyond the four species used in this study [3], this model does represent an advancement on ubiquitously used mono-species G. vaginalis biofilms. Moving forward, future studies may seek to trial endolysins in cocktails or evaluate synergy with existing antibiotics to provide a larger spectrum against the polymicrobial biofilm model. Additionally, it is important that the collective findings documenting the efficacy of Gardnerella-targeted endolysins are translated into clinically meaningful outputssuch as a reduction in the inflammatory potential of biofilms following endolysin treatment.

Funding
The current study was supported by the National Biofilms Innovation Centre (NBIC) under project number: 04POC21-235 and an Innovation Seed Funding award from Scottish Universities Life Sciences Alliance. [26,32,46,47].

Declaration of competing interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Matthew Cummings and David Corcorcan of CC Biotech Ltd have a patent for the clinical use of the endolysin(s) described in this study.

Data availability
Data will be made available on request.