3D Cell assemblies for controlled evaluation of response to dual agent delivery nanofibers targeting bacterial vaginosis

Abstract

Inflammatory response to Gardnerella during bacterial vaginosis (BV) is linked to complications including preterm birth and infection susceptibility. Platforms for local antibiotic and probiotic co-delivery have recently emerged to promote user adherence and avoid recurrence. This study designed a novel 3D cell assembly for controlled response evaluation to co-delivery platforms.

Vaginal bilayer model was configured with 3.0 μm pore inserts in 24-well plates, with confluent vaginal epithelial cells (VK2 E6/E7) in apical chamber and peripheral mononuclear cells (PBMCs) in basal chamber. PEO (rapid-dissolving) and PEO:PLGA (sustained-release) nanofibers were loaded with 7.5mg metronidazole and 5×10⁷CFU L.crispatus, and added to apical chamber. After 24-hour Gardnerella treatment followed by 72-hour nanofiber treatment, IL-8 and TNF-α levels were measured.

Probiotic-loaded nanofibers elicited weak inflammatory response compared to Gardnerella or L.crispatus exposure. TNF-α was lower with sustained L.crispatus recovery compared to burst release (without metronidazole). IL-8 was comparable between rapid vs. sustained release. Metronidazole elicited transient Gardnerella decrease while L.crispatus inhibited longer term. Burst release yielded greater Gardnerella inhibition than sustained release, while L.crispatus recovery was relatively unaffected. Nanofibers had no cytotoxicity on VK2 cells or PBMCs.

A 3D cell assembly model was established for controlled response evaluation to dual-agent nanofibers targeting BV, furthering the goal of more refined in vitro new approach methodologies (NAMs) for FRT research.

1. Introduction

Bacterial vaginosis (BV) is a dysbiosis of the lower female reproductive tract (FRT) that affects approximately 29% of women globally [1–3]. In BV, the symbiotic relationship between a multi-layered epithelium and a microbiome predominantly composed of Lactobacillus spp. is disrupted by the overgrowth of Gardnerella vaginalis (Gardnerella), Prevotella spp., Atopobium vaginae, and other facultative anaerobic pathogens [4, 5]. The combination of bacterial species varies with each patient; however, Gardnerella has been observed as the most common, isolated organism of patients with BV, found in 95% of cases [6, 7]. Gardnerella has been characterized as highly virulent due to the production of vaginolysin, a pore-forming cytotoxin, and enhanced adhesion capabilities for dense biofilm formation onto vaginal epithelial cells of the lower female reproductive tract (FRT) [7, 8]. As a result of Gardnerella outcompeting lactobacilli, an increased Gardnerella population raises the pH of the lower FRT to be less acidic (pH>4.5) [4, 5], which promotes pathogen survival. Previous in vitro and in vivo studies resulted in significantly elevated levels of IL-8, TNF-α, IL-10, IL-1β, and NLRP-3 proteins from vaginal epithelial cells and monocytes after exposure to Gardnerella [7–9]. Gardnerella elicits a pro-inflammatory response through the production of short fatty acid chains that enhance toll-like receptors (TLRs) which increase inflammation-associated protein production by FRT cells [7–11]. With biofilm formation, change in vaginal pH, and evoked inflammatory responses, Gardnerella enables other pathogenic species to colonize, further decreasing the commensal bacterial population and promoting adverse health conditions.

Metronidazole and other antibiotics such as clindamycin and tinidazole have been used as treatment for BV. Due to patient non-adherence to frequent and messy application schedules, high recurrence rates and antibiotic resistance can arise. Seeking to avoid this resistance, probiotics such as Lactobacillus have become an alternative treatment strategy [2, 3, 12, 13]. The use of Lactobacillus spp. as a probiotic treatment against endogenous pathogenic bacteria creates a competitive environment through biofilm formation and antimicrobial release [14]. With the highest levels of D-lactic acid production compared to other Lactobacillus species, Lactobacillus crispatus (L. crispatus) has increased in popularity as a probiotic treatment against BV, as D-lactic acid provides a strong microbicidal defense against Gardnerella biofilm formation while also acidifying the FRT [5, 14]. Furthermore, D-lactic acid has been shown to elicit anti-inflammatory responses (quantified by high levels of cytokines IL-1RA and IL-1β) and decrease inflammatory mediators (Il-6, IL-8, RANTES, MIP3-α, and TNF-α) from epithelial cells and monocytes post exposure to Gardnerella [10]. However, vaginal probiotic treatment still requires frequent application schedules.

To promote user adherence and prevent high recurrence rates, alternative intravaginal drug delivery methods have been explored. These methods include vaginally inserted 3D printed scaffolds, meshes, discs, and electrospun nanofibers delivering antibiotics or probiotics [15–20], as summarized in Table 1. Nanofibers offer the capability for soft, pliable, dissolving sustained-release platforms with the goal to promote user adherence. Recent in vitro studies focusing on the efficacy of L. crispatus-loaded polyethylene oxide/poly(lactic-co-glycolic acid) (PEO/PLGA) fibers in prevention of Gardnerella growth yielded promising results by demonstrating significant reduction in Gardnerella viability while retaining viable L. crispatus [16, 19]. Although free lactic acid has been shown to mediate the pro-inflammatory response and elicit an anti-inflammatory reaction, studies have yet to be performed on the effect of lactic acid produced by Lactobacillus spp. in co-culture with vaginal epithelial cells or the use of L. crispatus from biomaterial platforms such as electrospun fibers to counteract the effects of Gardnerella with respect to the FRT inflammatory response.

Table 1.

Overview of intravaginal drug delivery methods

Delivery Platform	Advantages	Challenges	Clinical Translation Potential
Electrospun Fibers	Tailored, individualized treatments Fine tune drug release due to additive manufacturing Controlled drug release High surface area for active ingredient loading Allow combination of treatments in one delivery platform	Scale up challenges Potential solvent toxicity Variation in flexibility and mechanical strength in different formulations	Currently in many preclinical studies Opportunity for single use dosage promoting user treatment adherence Potential for easy application Scale up challenges
3D-Printed Scaffolds	Tailored, individualized treatments Fine tune drug release due to additive manufacturing Controlled drug release High mechanical strength allowing sustained release Allow combination of treatments in one delivery platform	Multi-faceted ink formulations Requires strict conditions and finding optimum printing parameters Risk of swelling may lead to burst effect affecting drug release kinetics Discomfort in placement and use	Currently in pre-clinical studies Opportunity for single use dosage promoting user treatment adherence Difficulty in mass production and shelf-life variability Potential for easy application
Vaginal Rings	Tailored, individualized treatments Fine tune drug release due to additive manufacturing Controlled drug release High mechanical strength allowing sustained release	Discomfort in placement and use Potential for tissue irritation/inflammation Combination of treatments may be challenging	Multiple marketed products available Multiple products in pre-clinical and clinical trials Potential for easy application
Nanoparticles	Variety of fabrication techniques available High customization for targeted release, including combinatorial treatment Increased mucus penetration Variety of API choices (hydrophilic, hydrophobic, proteins, etc.)	Time-consuming fabrication Potential solvent toxicity Uneven distribution Short residence time due to clearance by vaginal fluid	Many current pre-clinical studies with few approved for clinical use Complex manufacturing Scale up challenges No marketed products available
Creams/Gels	Allows easy dispersal into the vaginal tract Low discomfort in application	High leakage probability Poor release and uneven distribution Low user adherence due to messy application and requiring multiple doses Difficult to combine treatments Short residence time due to clearance by vaginal fluid	Multiple products in pre-clinical and clinical trials Multiple marketed products available
Hydrogels	High flexibility High biocompatibility Easily alter volume Non-toxic cross-linking agents Mucoadhesiveness allows increased residence time	High leakage probability Poor release and uneven distribution Low adherence due to messy application and requiring multiple doses Difficult to combine treatments Potential for tissue irritation/inflammation	Currently in pre-clinical studies Marketed products available Manufacturing material selection may be difficult

Evaluation of alternative intravaginal drug delivery platforms such as nanofibers would benefit from new approach methodologies (NAMs) that enable more controlled and efficient preclinical evaluation to complement current in vitro and animal models. Studies using animal models are limited in fully replicating the human in vivo response due to physiological differences [21]. Alternative in vitro models of the vaginal tract have sought to better replicate in vivo processes as well as the inflammatory response [22]. Mahajan et al. [28] created a vagina-on-a-chip to better understand the interaction between L. crispatus, Gardnerella, and vaginal epithelium. Co-culture of epithelial cells, fibroblasts, and bacteria was accomplished along with the examination of pro-inflammatory cytokine response, pH changes to mimic FRT environments (healthy and BV conditions), and lactic acid levels. Although these studies involved cells that can elicit an immune response, they did not include immune cells such as monocytes to further investigate interactions between the epithelium and stromal immune cells. Here, we designed a novel 3D cell assembly for controlled evaluation of vaginal epithelial and immune cell response to dual agent delivery platforms targeting pathological conditions. This system allows modeling bacterial vaginosis and treatment by exposing lower FRT cells to Gardnerella and to dual agent platforms, enabling controlled evaluation of the cells’ inflammatory response in conjunction with the treatment response. This system was used to compare rapid-dissolving vs. sustained-release nanofibers co-delivering probiotics (L. crispatus) and antibiotic (metronidazole).

2. Materials and Methods

2.1. Bacterial cultures

Lactobacillus crispatus was purchased from American Type Culture Collection (ATCC MV-1A-US) and grown using DeMan, Rogosa, and Sharpe (MRS) broth (Sigma, 69966) and agar (Sigma, 69964) supplemented with 0.1% Tween80 (Sigma, P4780). Streptomycin- resistant Gardnerella from the strain background JCP8151B (GenBank JX860320, described in a previous publication [23]) was grown using New York City III (NYC III) broth and agar. The NYC III broth and agar were made using Sterile HEPES (VWR 45000–690), proteose peptone (Millipore Sigma 82450), sodium chloride (VWR, 0241), D-Glucose (Sigma, G7021), agar (FisherSci, BP1423–2, only for NYC III agar plates), yeast extract (Gibco 18180–059), and 10% heat inactivated horse serum (Gibco,16050–122). To allow for selective culture of streptomycin resistant Gardnerella, the NYC III plates and broth were supplemented with 1 mg/mL streptomycin sulfate salt (Sigma, S9137).

Frozen stocks of L. crispatus were streaked on MRS agar plates using a sterile inoculation loop. The plates were then incubated in anaerobic conditions at 37°C using a Coy Anaerobic chamber (Coy Laboratory Products, Grass Lake, Michigan, USA). L. crispatus required a 16-hour(hr) incubation period for colony formation. Once colonies were observed, the bacterial colonies from the plate were collected using an inoculation loop and placed in 10 mL of MRS broth. OD600 was measured by a Genysis 50 UV-Vis Spectrophotometers (Thermofisher, Waltham, USA 840–298000) and an equation was utilized as previously determined relating OD600 and CFU counts of L. crispatus [18]. The theoretical CFU was calculated to be 5×10⁷ CFU/mL. Additionally, the L. crispatus broth inoculum was serially diluted onto a MRS plate, and the CFU was calculated to be 5×10⁷ CFU/mL. L. crispatus inoculum was then added to electrospun fibers or directly into the Transwell^®.

Frozen stocks of Gardnerella were streaked onto NYC III agar plates supplemented with 10% heat inactivated horse serum and streptomycin and incubated at 37°C, 5% CO₂ for 24 hr. Both streaking and incubation were done under strict anaerobic conditions using a Coy Anaerobic Chamber. After incubation, colony formation was seen and was collected. Then, 5 mL of NYC III broth with 10% heat inactivated horse serum and streptomycin was inoculated with colonies collected. Volume of 1 mL of the solution was used to test OD for cell counting. OD600 was measured, and the theoretical CFU was determined to be equivalent to 1×10⁹ CFU/mL. The relationship between OD600 readings of Gardnerella, and the CFU counts were established by correlating OD600 readings of different Gardnerella dilutions to the CFU count by serial diluting on a NYCIII plate. Serial dilutions onto NYC III plates were performed to determine the actual CFU of the initial inoculum to be 1×10⁹/mL.

2.2. Electrospun Nanofiber Synthesis

To fabricate the PEO fiber bioink, 5% (w/w) PEO (Sigma, 182028) was added to 5 mL of MRS broth and incubated at 37°C, 5% CO₂ overnight. PEO bioink was then filtered through a 0.45 μm membrane Pall Acrodisc Syringe Filter (Fisher Scientific, PN4614). For L. crispatus-containing fibers, a calculated volume of MRS broth containing 5×10⁷ CFU L. crispatus/mL MRS was spun down in a centrifuge at 1200 RPM for 15 minutes (min). The pellet containing L. crispatus was obtained and resuspended with 0.5 mL of MRS and then added to the PEO bioink after filtration. The bioink was mixed until homogenous and 3 mL was loaded into a 3 mL syringe for electrospinning. To fabricate the PLGA fiber bioink, 15% (w/w) PLGA (50:50, Polysciences, 23986) was added to 5 mL of 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP) (Sigma, 105228) and incubated at 37°C, 5% CO₂ overnight. The PLGA bioink was then filtered through a 0.45 μm membrane Pall Acrodisc Syringe Filter. The filtered solution was loaded into a 3 mL syringe. For co-loaded L. crispatus/metronidazole-containing fibers, 15 mg of metronidazole (Sigma, M1547) was solubilized in 1 mL of N,N-dimethylformamide (DMF) (Sigma, 227056) and then 500 μL of the metronidazole/DMF solution was added to 2.5 mL of L. crispatus-containing PEO bioink.

With electrospinning PEO fibers characterized by a rapid drug release, one syringe containing 3 mL of PEO bioink was placed in a syringe pump where the syringe needle tip was 15 cm from an 8 mm in diameter mandrel. A positive voltage of 25 kV was applied to the PEO needle, and a syringe flow rate of 0.3 mL/hr was maintained for 10 hr. As the polymer was extruded from the syringe and the voltage was applied, the polymer was collected on the mandrel. This method was the same for blank, L. crispatus-containing, and L. crispatus/metronidazole-containing PEO fibers. For electrospinning PEO:PLGA mesh fibers, a syringe containing 3 mL PEO bioink was placed with the needle tip 15 cm from an 8 mm mandrel. A second syringe containing 1 mL PLGA bioink was placed with the needle tip 18 cm from the 8 mm mandrel in opposing direction of the PEO needle. 25 kV and 18 kV positive voltages were added to the PEO and PLGA needles, respectively. The flow rate was maintained at 0.3 mL/hr on the PEO syringe and 0.1 mL/hr on the PLGA for 10 hours. As the polymers were being electrospun from the syringes with the voltages applied, both polymers were collected onto the mandrel simultaneously, creating a 3:1 PEO:PLGA mesh fiber. This method was the same for PEO:PLGA blank, L. crispatus-containing, and L. crispatus/metronidazole containing fibers, with the metronidazole and L. crispatus loaded in the PEO bioink.

2.3. Nanofiber Characterization

PEO only and PLGA:PEO electrospun fibers (with and without L. crispatus or metronidazole incorporated) have been characterized in our previous studies [16, 18, 24]. Briefly, fiber samples were cut into approximately 3 mm square samples and placed on carbon tape. Samples were then sputter coated with palladium gold alloy (8.2 nm) and imaged using Apreo C LoVac Field Emission scanning electron microscopy (SEM) (Thermo Scientific, Waltham, USA). Differential scanning calorimetry (DSC) and thermogravimetric analysis (TGA) were used to characterize the fiber thermal profiles [24]. Fiber degradation was previously measured in simulated vaginal fluid (SVF), for which fibers were immersed in 1 mL of SVF for 24 hr at 37°C to simulate physiological exposure [24]. Fibers were dried for another 24 hr, and weighed. They were re-immersed in 1 mL SVF, repeating this process until fiber mass remained unchanged for at least 2 days [24]. Recovery of L. crispatus from the fibers was tested in media through serial dilutions while metronidazole release was assessed through UV-vis at an absorbance of 320 nm [24].

2.4. Vaginal Epithelial Cell Cultures

VK2 E6/E7 cell line of immortalized vaginal epithelial cells (ATCC CRL-2616) was maintained in defined keratinocyte serum free medium (KSFM) (FisherSci 10744019) with glutamine and supplemented with defined KSFM Growth Supplement and 1x penicillin/streptomycin for initial culture in a 60 mm culture plate until reaching 80% confluence. KSFM was changed every 48 hr. Cells were then trypsinized using Trypsin-EDTA (FisherSci, 25200–056) and neutralized with 1:1 Ham’s F-12 (FisherSci, 11765–054): Dulbecco’s Modified Eagle Medium (DMEM) (VWR 15–017-CV) solution with 10% fetal bovine serum (FBS) (FisherSci, A52568–01) and 100mg/mL 1x penicillin/streptomycin (VWR, K952). Vaginal epithelial cells were then collected in a 5 mL falcon tube and spun down to collect in pellet. Subsequently, these cells were resuspended in 2 mL of KSFM and 1x penicillin/streptomycin. Then, the cells were distributed into a Transwell^® dual chamber system.

2.5. Primary Peripheral Blood Mononuclear Cells (PBMCs)

First, 1 mL frozen stocks of peripheral blood mononuclear cells (PBMCs) (ATCC PCS-800–011) were thawed and then placed in a 5 mL falcon tube containing 5 mL of Roswell Park Memorial Institute (RPMI) 1640 medium containing 2 mM glutamine (Fisher Scientific 61870–036) supplemented with 10% FBS and 100 units/mL penicillin/streptomycin. The cells were centrifuged at 200xg for 10 min. The supernatant was removed, leaving a pellet of cells in the tube. The cells were then resuspended using 5 mL of RPMI 1640 medium containing 2 mM glutamine supplemented with 10% FBS and 100 units of penicillin/streptomycin. A sample was taken for counting using a 0.1 mm deep hemacytometer (Brightline 480371). The cell concentration was calculated to 1×10³/mL. The PBMCs in the RPMI 1640 medium were centrifuged at 4°C, 300xg for 5 min. The supernatant was aspirated off and the pellet containing the PBMCs was resuspended using 10 mL of RPMI 1640 medium containing 10% FBS, 100 units/mL Penicillin/streptomycin and 2 mM 1-glutamine. Then 100 μL of 1×10³ CFU PBMCs/mL was aliquoted into 48 wells in a 24 well plate. Next, 600 μL of RPMI 1640 was added to the wells and was incubated for 24 hr. After 24 hr the media was removed and 700 μL of RPMI media was added. Media was changed every 48 hr for 10 days. Once the PBMCs reached a concentration of 5×10⁵ PBMCs per well, the media was removed, and Transwell^® inserts containing VK2 cells were added.

2.6. Transwell^® Dual-chamber System

The apical chamber of the 24-well Transwell^® dual system consists of Costar Transwell^® 6.5 mm inserts with a 3.0 μm pore in Costar 24 well plates (Sigma, 3472). Here, the inserts were coated with 40 μL of 200 ng/mL laminin (Sigma-Aldrich L2020) and air dried in a sterile biosafety cabinet for 4 hr. After cells from the primary culture reached 80% confluence, subsequently trypsinized, neutralized, centrifuged, and resuspended with KSFM supplemented with 1x penicillin/streptomycin, cells were counted using a hematocytometer and cell concentration was determined. A volume of 100 μL of 3×10⁴ VK2 cells was added to the apical chamber. An additional 200 μL of KSFM was added to the apical chamber. Then, 1 mL of KSFM was added to the basal chamber of the Transwell^® dual-chamber system. Transwell^® plates were incubated in a 37°C, 5% CO₂ incubator for 10 days. The medium in the basal chamber was changed every other day, and cell layer integrity was determined by measuring transepithelial electrical resistance (TEER) values of each cell layer using an EVOM Manual^™ (World Precision Instruments). After 10 days, cell layers reached 150 Ω*cm² showing an intact and confluent layer, consistent with previous literature measuring vaginal epithelial cell layer integrity on Transwell^® inserts [17]. Values were normalized to a Transwell^® chamber by measuring a well containing an insert without cells and the same medium volume.

After cell layers were shown to be confluent, the media was removed from the apical chamber and the cells were washed with 1xPBS. The media in the 24 well plates containing 5×10⁵ CFU PBMCs/mL was removed and the inserts containing the VK2 cells were added to each of the PBMC wells. Then, 1 mL and 300 μL of KSFM without supplementation of antibiotics were added to the basal and apical chambers, respectively. The plates were incubated for 24 hr at 37°C, 5% CO₂. After 24 hr the media was changed, 100 μL of 1.3×10⁹ CFU/mL of Gardnerella was added to each apical layer of the Transwell^®, and incubated at 37°C, 5% CO₂ for an additional 24 hr. After 24-hr incubation with the Gardnerella, corresponding groups had 15mg electrospun fiber discs or 100 μL of 5×10⁷ free L. crispatus/mL added to the apical chamber. The plate was incubated at 37°C and 5% CO₂ for 24 hr, 48 hr, and 72 hr. After each time point, the supernatant of the apical and basal compartments was removed for cytokine analysis, pH, and bacterial CFU counts. Then, 700 μL and 100 μL of KSFM without antibiotic supplementation was added to the basal and apical layers respectively after each time point. After the 72-hr time point, the Transwell^® inserts were removed and used for MTT cytotoxicity testing of the VK2 cells.

2.7. Detection of Inflammation Markers

IL-8 and TNF-α were chosen as representative key markers of inflammation during BV [25]. After 24-hr infection of Gardnerella, the VK2 and PBMC co-cultures were treated with metronidazole, L. crispatus, or blank fibers. After 24-hr, 48-hr, and 72-hr treatment, the supernatant removed from each timepoint of both the apical and basal chambers of the co-cultures was tested using Invitrogen IL-8 and TNF-α colorimetric ELISA assays. Supernatants of apical and basal chambers from the Transwell^® dual-chamber system were added to 96 well plates and levels of IL-8 and TNF-α were quantified using colorimetric ELISA assays as described by manufacturer’s protocol (Invitrogen; Thermo Fisher Scientific, Inc.). Absorbance was read at 450 nm using a Synergy Microplate Reader, Biotek Instruments Inc. For each cytokine, a standard curve was created to establish the relationship between absorbance at 550nm to concentration of the cytokine. The formula, y = 158.477(x) - 1.4577 (R² = 0.982), was used to convert absorbance to IL-8 concentration (pg). The formula, y = 125.01(x) - 4.981 (R² =0.979) was used to convert absorbance to TNF-α concentration (pg).

2.8. Cytotoxicity Assay

MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide) assay was used to determine cytotoxicity of electrospun fibers on the VK2 cells and PBMCs. VK2 cells were cultured in the apical chamber of 24-well Transwell^® system for 10 days as previously described. After confirmation of cell confluence and integrity using TEER, VK2 cells and PBMCs were incubated with different formulation of electrospun fibers in triplicates for 72 hr. Media only (untreated cells) and 10% DMSO (VWR 67-68-5) were used as negative and positive controls, respectively. Medium containing the solution from the apical chamber was removed and VK2 cells refreshed by 100 μL cell culture medium. Also, medium from the basal chamber was removed and 700 μL of fresh medium was added. 70 μL of MTT reagent (Sigma Aldrich, 475989) was added to basal chamber and incubated at 37°C and 5% CO₂ for 4 hr, followed by adding 385 μL of lysis buffer and plates were incubated overnight. Absorbance was read at 570 nm (SYNERGY Microplate Reader, Biotek Instruments Inc) and normalized to the absorbance of untreated cells.

2.9. Gardnerella and L. crispatus Growth

After VK2 cells and PBMCs were incubated together in the dual co-culture model, Gardnerella was added to certain groups for Gardnerella infection, mimicking BV conditions in the vaginal tract in vitro. Each of the infected groups received 100 μL of 1.3×10⁹ CFU/mL of Gardnerella and were incubated for 24 hours to allow Gardnerella biofilm formation. After 24 hours, cells received either no treatment (Gardnerella control), free L. crispatus treatment, blank PEO treatment, PEO:PLGA blank treatment, PEO L. crispatus-only treatment, PEO:PLGA L. crispatus-only treatment, dual PEO L. crispatus-metronidazole treatment, or dual PEO:PLGA L. crispatus-metronidazole treatment. Supernatants were collected after 24-,48-, and 72-hr treatment time points and were serially diluted onto NYC III agar SRP plates to determine Gardnerella CFU. For each timepoint, the CFU was divided by the initial CFU and multiplied by 100 to obtain percent of Gardnerella growth during each time point to evaluate Gardnerella inhibition.

L. crispatus CFU was also analyzed after 24-,48-, and 72-hrs post treatment. Treatment groups with free L. crispatus were loaded with 5×10⁷ CFU/mL of L. crispatus. PEO and PLGA fibers were loaded with a starting L. crispatus CFU of 5×10⁷. At 24-, 48-, and 72-hr post treatment of free L. crispatus or fibers, the supernatant of each group was used for serial dilutions to determine CFU counts.

2.10. Measurements of pH

KSFM used in the Transwell^® system for both VK2 cells and PBMCs had a standard pH of 7. After 24-, 48-, and 72-hr, the pH of the supernatant from the basal and apical layers were tested and media was refreshed.

2.11. SEM Verification of Intact VK2 Monolayer

Post 96 hours treatment, porous membranes containing VK2 monolayers were removed from the Transwell^® inserts using a sterile razor blade and tweezers. Membranes were then placed on carbon tape with the monolayers facing up. Samples were sputter coated with palladium gold alloy (5mm) and imaged using Apreo C LoVac Field Emission SEM (thermo Scientific, Waltham, USA).

2.12. Statistical Analysis

Statistical analyses were performed in GraphPad Prism version 10.4.1 (GraphPad Software, La Jolla, USA). Statistical tests performed in each experiment are indicated in the figure legend and include one-way ANOVA with Tukey’s multiple comparison test, two-way ANOVA with Tukey’s multiple comparison test, and one-way ANOVA with Šídák’s multiple comparison test. All in vitro experiments were executed with 3–5 replicates per group. All groups infected were from the same Gardnerella and all fibers were manufactured with the same Lactobacillus and fabrication practices.

3. Results

3.1. Study Workflow

The study workflow is illustrated in Figure 1, showing a 24-well plate with 6.5 mm 3.0 μm pore membrane inserts, containing PBMCs on the basal chamber and VK2 cells on the apical chamber with 500uL of KSFM in each well. The 3D cell assembly model was fabricated with VK2 cells in the insert and monocytes in the bottom of the well. The model was then infected with Gardnerella followed by treatment with various fiber treatments and free L. crispatus in the apical chamber. The supernatant and VK2 cells from the model were extracted and used for the various assays.

Figure 1.

3D cell assembly model. Shown are 24-well Transwell^® plate with 6.5 mm 3.0 μm pore membrane insert containing PBMCs on the basal chamber and VK2 cells on the apical chamber with 500uL of KSFM in each well. The assembly was fabricated as a dual layer system with VK2 cells in the insert and monocytes in the bottom of the well. The model was then infected with Gardnerella followed by treatment with fibers and free L. crispatus. The supernatant and VK2 cells were extracted and used for the various assays.

3.2. Nanofiber Characterization

The fiber formulations used in this study have been optimized and characterized in our previous work [16, 18, 24] and the results are briefly summarized here. As observed via SEM, PEO fibers exhibit thin fibrous strands while PEO:PLGA show thin PEO fibers intertwined with the thicker PLGA fibers. The PEO:PLGA 1:3 ratio mesh architecture was verified through fluorescence microscopy [16]. PLGA fibers forming a mesh architecture. PEO and PEO:PLGA fibers with metronidazole show crystals incorporated into the PEO strands [24]. PEO and PEO:PLGA fibers containing L. crispatus, chains of rod-shaped bacteria were visualized within the PEO strands [16, 24]. However, metronidazole and L. crispatus did not appear to exert a significant change on fiber morphology and no beading of the polymers was observed [16, 24]. DSC/TGA provided mass change and heat flow to verify fiber make up and thermal decomposition [24]. As previously observed, PEO-only fibers had 100% degradation within 24 hours, while PEO:PLGA fibers had 75% reduction in mass after 48 hours, but no significant change after 1 week.[16, 24] Initial concentrations of L. crispatus recovered for both fiber types had at least 10⁵ CFU/mL and increased to 10⁸ CFU/mL after 48 hours and 10⁹ CFU/mL after 6 days [24]. Metronidazole release attained 93.95% of theoretical maximum load released in media within 20 minutes [24].

3.3. TNF-α Release by VK2 Cells

Inflammatory cytokine release by VK2 cells and PBMC in the 3D cell assembly model was measured at 24, 48, and 72 hr. post infection with Gardnerella. Cytokine production slowly ramped up during the first 48 hr (Supplementary Figure 1 and Supplementary Figure 2). Post 72-hr treatment of VK2 cells (Figure 2A), the Gardnerella control group had increased to an average of 32.77 pg while the L. crispatus control group had increased to 24.58 pg, which was lower than the Gardnerella control (p<0.0108). The Gardnerella control had a much higher TNF-α concentration compared to all the fiber treatment groups (p<0.0001), except for the uninfected PEO blank that trended lower. The infected PEO blank had lower TNF-α than the uninfected counterpart (p<.0001). Uninfected PEO L. crispatus fiber TNF-α trended lower than when infected, while the uninfected dual PEO L. crispatus-metronidazole group had lower TNF-α concentration compared to the infected group (p<0.0019).

Figure 2.

TNF-α production after 72 hr treatment with free bacteria or electrospun fibers of VK2 cells infected with Gardnerella for 24 hr in coculture with PBMCs. (A) VK2 cells exposed to PEO fibers. (B) VK2 cells exposed to PEO:PLGA fibers. (C) PBMCs exposed to PEO fibers. (D) PBMCs exposed to PEO:PLGA fibers. Statistical significance between experimental groups was determined with one-way ANOVA with Šídák’s multiple comparison test (*p<0.05; **p<0.01; ***p<0.001; **** p<0.0001). ns: not significant. L. cr.: L. crispatus; GV: Gardnerella; MET: metronidazole.

Uninfected PEO:PLGA blank fibers had higher TNF-α release concentration than when Gardnerella was present (p<0.0019) (Figure 2B). In contrast, uninfected PEO:PLGA L. crispatus fibers showed lower TNF-α concentrations compared to their infected counterpart (p<0.0192), while dual-loaded PEO:PLGA L. crispatus-metronidazole fibers exhibited a similar trend (not statistically significant).

The lowest TNF-α release concentration elicited from VK2 cells after 72 hours was for uninfected dual PEO L. crispatus-metronidazole fibers, with an average concentration of 0.769 pg. The second and third lowest concentrations were for uninfected PEO:PLGA L. crispatus fibers (average of 1.71 pg) and uninfected dual PEO:PLGA L. crispatus-metronidazole fibers (average of 2.52 pg).

3.4. TNF-α Release by PBMCs

Post 72-hr treatment, all groups drastically increased or decreased from the previous timepoint. The dual PEO L. crispatus-metronidazole fibers (Figure 2C) and the PEO:PLGA L. crispatus fibers (Figure 2D) had the lowest TNF-α release (average of 0.81pg and 0.89pg, respectively). The Gardnerella control group increased from an average TNF-α release concentration of 5.89 pg at 48 hours to an average of 24.83 pg at 72 hours. A similar result was seen in the L. crispatus control group, with an increase from 5.33 pg to 26.88 pg. The Gardnerella control was higher than all the other groups except for PEO blank, PEO:PLGA blank, and L. crispatus control groups.

Notably, infected dual PEO L. crispatus-metronidazole fibers had higher TNF-α release concentration (11.40pg) compared to uninfected (0.81pg) (p<0.0015). Dual PEO:PLGA L. crispatus-metronidazole fibers followed a similar trend (4.02pg vs. 3.56pg; not statistically significant), while infected PEO:PLGA L. crispatus fibers had average TNF-α concentration of 13.14pg compared to 0.89pg when uninfected (p<0.0015). As noted for the VK2 cells, infected PEO:PLGA blank fibers had significantly lower TNF-α release from PBMCs than uninfected (7.58pg vs. 16.15pg; p<0.0342).

3.5. IL-8 Release by VK2 Cells

Post 72-hr treatment, IL-8 release concentration by VK2 cells for the Gardnerella control had a drastic increase to an average of 46.40pg (Figure 3A). All treatment groups had lower release concentrations than the Gardnerella control (p<0.026). L. crispatus control group was also lower than the Gardnerella control (p<0.0099) but was higher than the infected free L. crispatus (p<0.0023). Uninfected PEO blank and PEO:PLGA blank fibers had higher IL-8 release concentrations than when infected (p<0.0149 and p< 0.0113, respectively) (Figure 3A and Figure 3B). The lowest IL-8 release concentration from VK2 cells after 72 hours of treatment was elicited from uninfected PEO:PLGA L. crispatus fibers (average release concentration of 12.33pg. Figure 3B). PEO L. crispatus fiber and dual PEO L. crispatus-metronidazole fiber groups were closely behind with average IL-8 release concentration of 13.12pg and 13.49pg respectively (Figure 3A).

Figure 3.

IL-8 production after 72 hr treatment with free bacteria or electrospun fibers of VK2 cells infected with Gardnerella for 24 hr in coculture with PBMCs. (A) VK2 cells exposed to PEO fibers. (B) VK2 cells exposed to PEO:PLGA fibers. (C) PBMCs exposed to PEO fibers. (D) PBMCs exposed to PEO:PLGA fibers. Statistical significance between experimental groups was determined with one-way ANOVA with Šídák’s multiple comparison test (*p<0.05; **p<0.01; ***p<0.001; **** p<0.0001). ns: not significant. L. cr.: L. crispatus; GV: Gardnerella; MET: metronidazole.

3.6. IL-8 Release by PBMCs

Post 72-hr treatment for PBMCs, Gardnerella control and L. crispatus control groups elicited the highest IL-8 release concentrations with an average of 36.15pg and 36.89pg, respectively (Figure 3C and Figure 3D). All treatment groups had significantly less IL-8 release concentrations than the Gardnerella control (all p<0.01). Free L. crispatus with Gardnerella infection had a significantly lower IL-8 release concentration than the L. crispatus control and Gardnerella control (p<0.002). Uninfected PEO L. crispatus fibers and dual PEO L. crispatus-metronidazole fibers trended with lower release concentrations compared to their infected counterparts (Figure 3C). Uninfected dual PEO L. crispatus metronidazole had the lowest overall IL-8 release concentration (average of 9.95pg). Lastly, uninfected PEO:PLGA blank fibers had higher release concentrations than when infected (p<0.0047) (Figure 3D).

3.7. Nanofiber Cytotoxicity

Cytotoxic effect of electrospun fibers on VK2 cells and PBMCs was assessed via reduction in cell viability as measured by MTT assay after 72 her incubation (Figure 4). Untreated cells were used as negative control and had a 100% cell viability. A positive control of 10% DMSO had 25% VK2 cell and 22% PBMC viability. All treatment groups with PEO fibers and PEO:PLGA fibers were shown to have similar viability to untreated cells, indicating no cytotoxic effect of the delivery systems, metronidazole, or L. crispatus.

Figure 4.

Cytotoxic effect of electrospun fibers evaluated via MTT viability assay after 72 hr incubation. (A) VK2 cell and (B) PBMC viability. Data represent mean±SD (n=5). Statistical significance between experimental groups was determined with one-way ANOVA with Tukey’s multiple comparison test (****p<0.0001). L. cr.: L. crispatus.

3.8. Gardnerella Growth

The free Gardnerella control remained at 100% or higher of initial culture throughout the entire 72-hour treatment period (Figure 5A). Gardnerella treated with free L. crispatus decreased from 100% to about 60% after 24 hours and then to 47% after 48 hours (p<0.00001); by 72 hours Gardnerella was completely inhibited (Figure 5B). Gardnerella in the presence of PEO blank fibers (Figure 5C) declined by ~20% by 72 hr, while in the presence of PEO:PLGA blank fibers (Figure 5D), growth declined slightly by 48 hr and recovered by 72 hr.

Figure 5.

Evaluation of Gardnerella after 24 hr infection of VK2 cells in coculture with PBMCs. For each timepoint, the CFU is divided by the initial CFU to give a percentage of growth: (A) No treatment. (B) Free L. crispatus. (C) PEO blank fibers. (D) PEO:PLGA blank fibers. (E) PEO L. crispatus-only fibers, (F) PEO:PLGA L. crispatus-only fibers. (G) PEO L. crispatus-metronidazole fibers. (H) PEO:PLGA L. crispatus-metronidazole fibers. Statistical significance between experimental groups was determined with one-way ANOVA with Tukey’s multiple comparison test (*p<0.05; **p<0.01; ***p<0.001; **** p<0.0001). ns: not significant.

For L. crispatus-only treatment groups, Gardnerella growth decreased for PEO fibers to 86% by 24 hours (p< 0.0001) followed by sustained inhibition through 72 hours (Figure 5E). PEO:PLGA treatment groups had a gradual but significant decrease in Gardnerella over 72-hour. Growth was 100% by 24 hours and decreased to 86% by 48 hours (p<0.0437), and to 65% at 72 hours (p<0.0041) (Figure 5F).

Dual L. crispatus-metronidazole treatment groups had a decrease in Gardnerella growth for PEO fibers to 44% by 24 hours (p< 0.0001) followed by an increase (p<0.0045) to 76% at 48 hours. Gardnerella growth decreased (p<0.0013) to 29% by 72 hours (Figure 5G). PEO:PLGA treatment groups had a slight decrease in Gardnerella growth to 77% at 24 hours. increased to 102% growth at 48 hours (p<0.0012) and decreased to 57%.at 72 hours (p<0.0001) (Figure 5H).

3.9. L. crispatus Growth

Post 24-hr treatment, a stable 10⁹ CFU/mL was observed for the L. crispatus control group, free L. crispatus with Gardnerella group, infected PEO L. crispatus fibers, and infected dual PEO L. crispatus-metronidazole fibers (Figure 6). PEO L. crispatus fibers (without or with metronidazole) had 10⁷ and 10⁶ CFU/mL, respectively. Post 48-hr treatment, L. crispatus control, free L. crispatus with Gardnerella infection. infected PEO L. crispatus fibers, and infected PEO L. crispatus-metronidazole fibers, all slightly increased by one log to 10¹⁰ CFU/mL. Uninfected PEO L. crispatus fibers maintained 10⁷ CFU/mL and uninfected PEO L. crispatus-metronidazole fibers increased one log to 10⁸ CFU/mL. Post 72-hour treatment, uninfected PEO L. crispatus stayed at 10⁷ CFU/mL and PEO L. crispatus-metronidazole increased to an average of 10⁹ CFU/mL L. crispatus, while all other groups had >10⁹ CFU/mL.

Figure 6.

Evaluation of L. crispatus when administered to VK2 cells in coculture with PBMCs. For each timepoint, the CFU is divided by the initial CFU to give a percentage of growth: (A) Control (no treatment) and infected for 24 hr with Gardnerella. (B) Recovery from PEO fibers and when infected for 24 hr with Gardnerella. (C) Recovery from PEO:PLGA fibers and when infected for 24 hr with Gardnerella. (D) Recovery from PEO+MET fibers and when infected for 24 hr with Gardnerella. (E) Recovery from PEO:PLGA+MET fibers and when infected for 24 hr with Gardnerella. Statistical significance between experimental groups was determined with two-way ANOVA with Tukey’s multiple comparison test (*p<0.05; **p<0.01; ***p<0.001; **** p<0.0001). ns: not significant.

Post 24-hr treatment of PEO:PLGA groups (Figure 6), infected L. crispatus fibers had >10⁸ CFU/mL while uninfected L. crispatus and L. crispatus-metronidazole fibers had 10³ and ~10⁷ CFU/mL, respectively. Post 48-hr treatment, the former increased by 4 logs and the latter increased by one log. By 72 hours, all groups had >10⁸ CFU/mL.

3.10. Intact VK2 Monolayer Post Infection and Treatment

After infection with Gardnerella and treatment with fibers for 72 hours, cell culture inserts were removed from the 24-well plates and the porous membranes were removed from the inserts containing the VK2 monolayer. The monolayers were then examined by SEM. VK2 cells that received no Gardnerella infection or treatments (fibers, metronidazole, or L. crispatus) showed an intact cell layer of tightly packed squamous epithelial cells with no bacterial cells present (Figure 7A). VK2 cells that received Gardnerella infection, but no treatment with fibers or L. crispatus showed VK2 cells still adhering to the membrane but starting to have gaps within the layer, indicating effect from the infection (Figure 7B). Additionally, there were bacteria on top of the cell layer as well as between the VK2 cells. VK2 cells that received L. crispatus treatment only showed an intact cell layer of tightly packed squamous epithelial cells, with chains of rod-shaped bacteria on top of the epithelial cells (Figure 7C). The bottom of the porous membrane was also examined with SEM after L. crispatus-only treatment (Figure 7D), showing that probiotic cells were visually present, indicating L. crispatus migration through the porous membrane.

Figure 7.

SEM images of the apical side of a 3.0μm porous membrane containing a VK2 monolayer after incubation. (A) Under no infection or treatment; (B) with Gardnerella infection at 24 hours; (C) with L. crispatus treatment at 24 hours; (D) basal side of the 3.0μm porous membrane containing a VK2 monolayer after incubation with L. crispatus treatment showing bacterial infiltration.

3.11. Change in pH

Post 24-hr treatment, all groups had an acidic media pH < 6 (Figure 8). All uninfected L. crispatus groups (free or in fibers) had pH <4, while corresponding infected groups were >4. As expected, infected PEO fibers with L. crispatus (with or without metronidazole) had higher pH than their uninfected counterparts (p < 0.05 and p < 0.0001, respectively). Infected PEO:PLGA fibers with L. crispatus with metronidazole had higher pH than their uninfected counterparts (p < 0.0001 and p < 0.05, respectively).

Figure 8.

Evaluation of pH. The pH of the supernatant was obtained for (A) cells treated with PEO fibers and (B) cells treated with PEO:PLGA fibers. Statistical significance between experimental groups was determined with one-way ANOVA with Šídák’s multiple comparison test (with p-values reported in text). L. cr.: L. crispatus; GV: Gardnerella; MET: metronidazole.

While the lowest pH (~3) for uninfected PEO:PLGA fibers was achieved by the dual loading of L. crispatus and metronidazole, the lowest pH for uninfected PEO fibers occurred when loaded only with L. crispatus, indicating that simultaneous probiotic and antibiotic release decreases probiotic viability. This effect is temporary, however, as shown by PEO L. crispatus fibers having by 48 hours higher pH than dual loaded PEO fibers (p < 0.0001). While the pH of all other fiber groups by 48 hours either declined further or remained stable, by 72 hours, blank PEO and PEO:PLGA fibers (uninfected or infected), infected PEO L. crispatus with metronidazole, and infected PEO:PLGA L. crispatus increased in pH but stayed < 6. In contrast, the infected PEO:PLGA L. crispatus with metronidazole remained < 4, as did the uninfected PEO and PEO:PLGA L. crispatus groups (with or without metronidazole).

4. Discussion

4.1. 3D Cell Assembly Vaginal Model to Study Response to Dual Agent Delivery Platforms

This study designed a novel 3D cell assembly for controlled evaluation of vaginal response to dual agent delivery platforms targeting bacterial vaginosis. 3D Cell assemblies composed of an apical chamber seeded with vaginal epithelial cells (VK2 E6/E7 cells) and a basal chamber seeded with peripheral blood mononuclear cells (PBMCs) was exposed to Gardnerella followed by treatment with nanofibers as a model platform co-loaded with probiotic (L. crispatus) and antibiotic (metronidazole). The system enabled evaluation of the inflammatory response as well as the bacterial microflora restoration. Demonstrating the anti-inflammatory effect of intravaginal probiotic drug delivery methods on Gardnerella-exposed lower FRT cells could provide a better understanding of how probiotics affect endogenous anaerobes associated with BV during this condition as well as the effect they have on the epithelium and immune cells of the lower FRT. Recognizing different levels of cytokine expression and the inflammatory responses during BV in contrast to the healthy state enables evaluation of how efficacious these intravaginal drug delivery vehicles are in restoring a healthy state after BV. Assessing cytokine levels and corresponding inflammatory responses of cells in the FRT is a necessary preliminary indicator before progressing to in vivo studies and clinical trials of alternative multi-agent delivery methods.

To address the persistence of recurring BV in patients, the use of probiotics and alternative probiotic/antibiotic delivery methods has recently been explored. L. crispatus provides competitive benefits against pathogenic bacteria and, being one of the predominant bacteria in healthy vaginal tracts supports its advantageous use as a probiotic to treat BV. Co-spinning L. crispatus in electrospun fibers has shown advantageous competition against Gardnerella and other BV associated bacteria with minimal cytotoxicity to the vaginal epithelium and monocytes [19]. Furthermore, demonstrating decreased inflammatory response when using L. crispatus-containing fibers would support safety and efficacy towards these alternative treatments of BV.

4.2. Evaluation of Inflammatory Response

Previous literature has explored acidic and anti-inflammatory responses to lactic acid treatment on the vaginal micro-environment and immune response [26–28]. However, the inflammatory effect of probiotics that produce lactic acid, L. crispatus specifically, has not been fully examined. The 3D cell assembly model enabled detailed examination of the inflammatory response of VK2 cells and PBMCs after infection of Gardnerella followed by treatment with Lactobacillus and metronidazole. TNF-alpha by VK2 and PBMCs decreased more with sustained L. crispatus recovery (PEO:PLGA fibers) compared to burst delivery (PEO fibers) (p = 0.0059 and p = 0.0017, respectively), but only when metronidazole was not included. In contrast, IL-8 by VK2 and PBMCs seemed impervious to immediate vs. sustained release. Altogether, the results highlight that metronidazole elicits a temporary (initial) Gardnerella decrease while L. crispatus provides longer term inhibition. Short-term release (by PEO fibers) of metronidazole or combination with L. crispatus may yield higher inhibition than with sustained release (by PEO:PLGA fibers).

The results with the 3D cell assembly model furthermore show that blank fibers (PEO or PEO:PLGA) elicit an inflammatory response. This response is attenuated in the presence of Gardnerella and substantially reduced with L. crispatus. Previous work indicates these materials do not cause significant increase in inflammation overall [29, 30] In the model, Gardnerella and L. crispatus on their own also elicited a strong inflammatory response, which was attenuated when both were present. Gardnerella is known to suppress the host immune system via the release of vaginolysin and sialidase [31] and prevention of cytotaxis by monocytes and leukocytes [31]. On the other hand, L. crispatus would not be expected to elicit such an inflammatory response in vivo, due to immunomodulation by the vaginal epithelium. Previous literature suggests, like the gut microbiota, vaginal commensal bacteria can cause inflammatory responses in host cells, but the close interaction between the two cells as well as anti-inflammatory proteins created by the host cells can prevent the commensal bacteria from inducing an inflammatory response. Specifically, CD4⁺ T-regulatory cells present in the vaginal tract maintain the tolerance of commensal bacteria as seen in the gut microbiome, and may allow the blocking of inflammatory cytokines like IL-8 and TNF-α to be produced by vaginal epithelial cells and immune cells due to commensal bacteria through the production of retinoic acid [11, 32]

To further investigate the impact of the fiber formulations on the vaginal mucosa, this study examined cytotoxicity via MTT assay on an intact monolayer of vaginal epithelium as well as monocyte cells that reside beneath the epithelium and help make up the lamina propria, showing no cytotoxic effect of the delivery systems, metronidazole, or L. crispatus. These results are consistent with previous lactate dehydrogenase (LDH) assay measurements performed on vaginal epithelial cells in vitro after exposure to PEO or PEO:PLGA blank, L. crispatus, metronidazole, and dual L. crispatus-metronidazole fibers with no statistically significant elevated levels of LDH compared to the untreated group, indicating no cytotoxicity of fibers on vaginal epithelial cells [24]. PEO fibers were also previously tested on estrogen-treated C57Bl/6J mice in vivo with no overt pathophysiological changes indicated by visualization of an intact epithelium with no polymorphonuclear cell infiltration after exposure to the fibers [18].

4.3. Evaluation of Burst vs. Delayed Delivery

The 3D cell assembly model allowed assessing response to burst vs. delayed release from a dual delivery vehicle. Dual PEO L. crispatus-metronidazole treatment and dual PEO:PLGA L. crispatus-metronidazole treatment exhibited an initial knockdown at 24 hours and then increased at 48 hours with a second knockdown at 72 hours. The first decrease at 24 hours, indicates the rapid release of metronidazole prior to 24 hours knocking down Gardnerella but unable to prevent temporary comeback. One treatment is insufficient to prevent recurrence of Gardnerella, as seen in previous literature [2], where only multiple treatment groups elicited persistent knockdown of Gardnerella. The decrease from 48 hours to 72 hours, indicates the release of L. crispatus from fibers allowing a second knockdown of Gardnerella. The dual PEO L. crispatus-metronidazole treatment had a greater knockdown than the dual PEO:PLGA L. crispatus-metronidazole treatment at 72 hours, potentially due to the PEO fiber having a more rapid instant release of L. crispatus than the slower degrading PEO:PLGA fibers.

4.4. Evaluation of Bacterial Viability and Dynamics

The 3D cell assembly model proved adequate for controlled evaluation of treatment effects. The Gardnerella control remained at 100% or higher growth throughout the timepoints indicating the model was able to sustain Gardnerella and function as an effective BV model, which is further supported by the pH of the Gardnerella control remaining at a pH of ~4.5, consistent with BV conditions in vivo [4]. The PEO blank and PEO:PLGA blank groups also remained at a 100% or higher Gardnerella growth throughout 72 hours of treatment, showing that the PEO and PLGA materials did not affect Gardnerella on their own and indicating any inhibition to be due to metronidazole or L. crispatus from the fibers. PEO:PLGA fibers act as a dual rapid and sustained release system, with PEO being hydrophilic allowing a rapid initial recovery of L. crispatus and PLGA being hydrophobic acting as a sustained source of L. crispatus. Future studies could evaluate PLGA fibers past 72 hours to demonstrate further recovery of L. crispatus and possibly attain a complete knockdown of Gardnerella as seen in the PEO only L. crispatus treatment group. Due to limitations of the proposed model lacking proper dispersion of media across cells to promote healthy growth after 4 days in culture without passage, expanding the testing period past 96 hours post treatment was not performed. A future improvement of the model could include fluid flow to address this limitation.

The proposed model further allowed evaluation of bacterial dynamics. L. crispatus CFU increased in all groups over all time points, demonstrating L. crispatus can proliferate without difficulty in this co-culture model and is not inhibited by the Gardnerella. Dual fibers had low L. crispatus levels after 24 hours of treatment and a slight increase after 48 hours, yet drastically increased after 72 hours, indicating that the initial burst of metronidazole temporarily hinders L. crispatus growth. This observation is consistent with previous work showing initial inhibition of L. crispatus when combined with metronidazole [24]; however, L. crispatus growth after release was not inhibited [33]. The loaded concentration of metronidazole in the fibers was 7.5mg/mL, significantly higher than the documented minimum inhibitory concentration (MIC) for L. crispatus to metronidazole (>128μg/mL) [34] which could account for the initial inhibition. Fiber formulations used in this study were previously fabricated and characterized, with the metronidazole dosage of 7.5mg/mg of material selected for preclinical evaluation in mice so that the theoretical yield of a 15 mg fiber segment yields an approximate dose of 125–256 μg, which has been shown to not hinder L. crispatus growth in vivo [33]. Furthermore, although L. crispatus increased, there was no large knockdown of Gardnerella, indicating co-inhabitation of L. crispatus and Gardnerella within the vaginal FRT model, which could indicate lack of competition between the two species due to abundance of nutrients or too low of an initial L. crispatus inoculum.

Healthy vaginal pH remains acidic at about 2.8–4.2 but becomes more basic during BV infection with a pH typically >4.5, as previously observed [35]. For each timepoint after the initial timepoint assessed in the model, the Gardnerella control remained at a pH of 4.5–5.0, consistent with BV conditions in vitro. The no treatment group and both blank fiber groups remained at pH >5.0 throughout all timepoints, whereas each L. crispatus treatment group remained at pH <4.0 after 24-hr treatment. A drop in pH below 4.2 indicates the treatment groups containing L. crispatus produced lactic acid, consistent with previous findings showing that lactic acid acidifies the vaginal environment [36]. The decrease in pH below 5 after 48 hours and lactic acid production were similar as previously observed for dual release nanofibers [24], confirming L. crispatus bioactivity.

4.5. Advancing NAMS for FRT research

Interest in NAMs complementary to in vitro and animal models to facilitate preclinical evaluation of drug delivery platforms has recently gained momentum. Evaluation of alternative intravaginal drug delivery platforms such as nanofibers would benefit from NAMs enabling detailed study of vaginal response to dual agent delivery by nanofibers targeting pathological conditions. Lack of proper in vitro models of the FRT as well as accurate recapitulation of the FRT mechanisms can hinder the understanding of how the FRT immune system works as well as a closer evaluation of the relationship of FRT host cells with commensal or pathogenic bacteria [28, 37]. Previous studies have relied on the comparison of microbiota of the gut and lungs before applying it to mouse models or 2-D cell cultures of the FRT, but as culture techniques have improved providing new ways of modeling the FRT, further insight into the FRT microbiome has been possible [31]. The 3D cell assembly model proposed here moves towards more refined NAMs for FRT research.

4.6. Study Limitations

The evaluations in this study assessed a limited number of bacterial species and cytokines elicited by nanofiber platforms. Further examination of the FRT inflammatory response and immune system including other cytokines, TLRs, and influence of other molecules present in the FRT such as retinoic acid and T regulatory cells is warranted. Additional or different bacterial species as well as variations in platform types and architectures could be evaluated in high throughput with the goal to enhance the understanding of mechanisms of infection and treatment response in the lower FRT. Future work will examine interactions with mucus and hormones, as well as menstrual cycle fluctuations. The model provides a platform to evaluate polymicrobial biofilm formation and response to treatment. We plan to validate model results with preclinical studies in vivo. Lastly, integration of patient cells and microflora could provide subject-specific evaluation.

5. Conclusion

This study developed a novel 3D cell assembly to grow vaginal epithelial cells in tandem with monocyte cells reflecting the multi-layered FRT. This model was infected with Gardnerella to replicate BV before being treated with electrospun nanofibers containing L. crispatus and metronidazole. Through the permeable insert beneath the vaginal epithelial cell monolayer of the multi-layer model, bacteria, nutrients, and proteins were able to freely pass between the epithelial cells and monocytes. By collecting the supernatant of both apical and basal chambers, cytokines released from both cell types could be measured to provide insight into the inflammatory response when undergoing infection and treatment. Overall, the results support the feasibility of a 3D cell assembly model for controlled pre-clinical evaluation of response to dual-agent nanofibers targeting BV, with the goal to establish more refined NAMs for FRT research.

Supplementary Material

Supplementary material

Highlights.

• Inflammatory response to Gardnerella during BV causes health complications

• Platforms for local antibiotic & probiotic co-delivery seek to avoid BV recurrence

• Designed a novel 3D cell assembly for controlled response to dual-agent nanofibers

• Bilayer co-culture model sustains vaginal epithelial cell interactions with PBMCs

• Results support the goal of advancing NAMs for female reproductive tract research