Human vagina chip culture
Microfluidic two-channel co-culture organ chip devices were obtained from Emulate Inc. (Boston, MA). The apical channel (1 mm wide × 1 mm high) and basal channel (1 mm wide × 0.2 mm high) are separated by the porous membrane (7 μm diameter pores) along their length (16.7 mm). For activation, both channels were filled with 0.5 mg/mL ER1 solution in ER2 buffer (Emulate Inc.) and placed under UV light for 20 min followed by washing with ER2 buffer and phosphate-buffered saline (PBS). This chemical activation improves the surface hydrophilicity of the PDMS membrane and increases the efficiency of anchoring of extracellular matrix (ECM) proteins to the chip membrane. Before cell seeding, the apical channel was incubated with collagen IV (30 μg/mL) from human placenta (Sigma, cat. no. C7521) and collagen I (200 μg/mL) from rat tail (Corning, cat. no. 354236) in DMEM (ThermoFisher, cat. no. 12320-032) at 37 °C with 5% CO2 for 2–3 h. The basal channel was incubated with collagen I (200 μg/mL) from rat tail (Corning, cat. No. 354236) and poly-L-lysine (15 μg/mL) (ScienCell Research Laboratories, cat. no. 2301) in DMEM (ThermoFisher, cat. no. 12320-032) at 37 °C with 5% CO2 for 2–3 h.
Primary human vaginal epithelial cells (Lifeline Cell Technology, cat. no. FC-0083; donors 05328 and 04033) were expanded in 75-cm2 tissue-culture flasks using vaginal epithelium growth medium (Lifeline Cell Technology, cat. no. LL-0068) to 60-70% confluency. Primary human uterine fibroblasts (ScienCell Research Laboratories, cat. no. 7040) were expanded in 75-cm2 tissue-culture flasks coated with poly-L-lysine (15 μg/mL, ScienCell Research Laboratories, cat. no. 2301) using fibroblast growth medium (ScienCell Research Laboratories, cat. no. 7040) to 60–70% confluency. In the current configuration, we utilized commercially available uterine fibroblasts due to limited availability of the primary human vaginal fibroblasts.
To create the human vagina chip, fibroblasts (1 × 106 cells/mL) were seeded first in the basal channel by inverting the chip for 1 h in human fibroblast growth medium. Chips were inverted again, and human vaginal epithelial cells (3 × 106 cells/mL) were seeded in the apical channel for 4 h in human vaginal growth medium. The chips were incubated at 37 °C with 5% CO2 overnight under static aerobic conditions. The basal channel was continuously perfused with fibroblast growth medium using the Zoe culture module (Emulate) at a volumetric flow rate of 40 μL/h. The apical channel was intermittently perfused with vaginal epithelium growth medium by changing the flow rate in the apical channel from 0 to 40 μL/h for 4 h per day by using the Zoe culture module to mimic episodic flow. Various organ models developed using commercial Emulate chips utilize 30 to 60 μL/h flow rate and 40 μL/h was selected empirically based on the effective epithelial differentiation we observed in the human vagina chip. After 5–6 days, the apical medium was replaced with Hank’s Balanced Salt Solution (HBSS; ThermoFisher, cat. no. 14025092) and the basal medium was replaced with in-house differentiation medium (see below) for 8 days following same intermittent and continuous perfusion regime, respectively. The HBSS was further replaced with customized HBSS Low Buffer/+Glucose (HBSS (LB/+G)) for 2 days followed by 3 days with microbial co-culture as described below.
Customized HBSS (LB/+G) medium is composed of 1.26 mM calcium chloride (Sigma, cat. no. 499609), 0.49 mM magnesium chloride hexahydrate (Sigma, cat. no. M2393), 0.41 mM magnesium sulfate heptahydrate (Sigma, cat. no. M2773), 5.33 mM potassium chloride (Sigma, cat. no. P5405), 0.44 mM potassium phosphate monobasic (Sigma, cat. no. P5655), 137.93 mM sodium chloride (Sigma, cat. no. S5886), and 5.56 mM D-glucose (Sigma, cat. no. G7021).
In-house differentiation medium is composed of DMEM (ThermoFisher, cat. no. 12320-032), Ham’s F12 (ThermoFisher, cat. no. 11765-054), 4 mM L-glutamine (ThermoFisher, cat. no. 25030081), 1 μM hydrocortisone (ThermoFisher, cat. no. H0396), 1× Insulin-Transferrin-Ethanolamine-Selenium (ITES; Lonza, cat. no. 17-839Z), 20 nM triiodothyronine (Sigma, cat. no. T6397), 100 μM O-phosphorylethanolamine (Sigma, cat. no. P0503), 180 μM adenine (Sigma, cat. no. T6397), 3.2 mM calcium chloride (Sigma, cat. no. 499609), 2% heat inactivated fetal bovine serum (FBS; ThermoFisher, cat. no. A3840001), 1% penicillin-streptomycin (ThermoFisher, cat. no. 15070063), and 4 nM β-Estradiol (Sigma, cat. no. E2257). This medium is adapted from a previously published medium composition used for engineered human skin [24].
Immunofluorescence microscopy
The vagina chips were fixed with 4% paraformaldehyde (Alfa Aesar, stock no. J61899) for 30 min at room temperature and washed with phosphate buffered saline (PBS). The channels were filled with 2% agarose (Lonza, cat. no. 50302) and the whole chip was embedded in O.C.T. compound (Fisher Scientific, cat. no. 23-730-571) and stored at – 80 °C until sectioning. Chips were cryosectioned at a thickness of 50 μm on a cryostat (Leica CM3050 S). The cryosections were then permeabilized using 0.1% Triton-X (Sigma-Aldrich, SKU no. X100) in PBS, blocked with 5% goat serum (Life Technologies, cat no. 16210072) in 0.01% Triton-X in PBS for 1 h at room temperature, and then incubated at 4 °C overnight with primary antibodies against CK13 (Abcam, cat. no. ab92551 at 1:200 dilution), CK14 (Abcam, cat. no. ab119695 at 1:200 dilution), E-cadherin (Abcam, cat. no. ab40772 at 1:200 dilution), ZO-1 (Abcam, cat. no. ab276131 at 1:40 dilution), Involucrin (Abcam, cat. no. ab68 at 1:200 dilution), DSG1 (Abcam, cat. no. ab12077 at 1:400 dilution), and DSG3 (Abcam, cat. no. ab231309 at 1:400 dilution). The sections were washed 3 times with PBS, and then incubated with secondary antibody (Abcam, cat. no. ab150077) at a dilution of 1:500 for 1 h at room temperature. Some sections also were incubated with directly labeled fluorescent with Alexa Fluor® antibodies against CK5 (Abcam, cat. no. ab193894) or CK15 (Abcam, cat. no. ab194065) or phalloidin (Invitrogen, cat. no. A22287) in the dark at 4 °C. Vagina chip sections were stained with as-received Eosin Y solution (Abcam, cat. no. ab246824), which fluoresces under blue-green excitation, for 30 s at room temperature to obtain pseudo-H&E staining. All stained sections were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; Invitrogen, cat. no. D1306) at a concentration of 1 μg/mL for 15 min at room temperature and mounted using ProLong Glass Antifade Mountant (ThermoFisher, cat. no. P36980). Images were acquired with an inverted laser-scanning confocal microscope (Leica SP5 X MP DMI-6000) and processed using ImageJ/Fiji. Pseudo H&E images were processed using ImageJ/Fiji and MATLAB (Mathworks) using a previously published method [25].
Barrier permeability
Cascade blue (Invitrogen, cat. no. OC3239) was added to apical channel media at a concentration of 50 μg/mL. Effluent from the apical and basal channels were collected and measured for fluorescence intensity at an excitation wavelength of 380 nm and an emission wavelength of 420 nm using a multi-mode plate reader (BioTek NEO). The apparent permeability (Papp) was calculated using the equation as previously reported [26]: Papp = (V
r * Cr)/A * t * (C
d-out * V
d + C
r * V
r)/V
d + V
r)), where V
r is the volume of the receiving channel, V
d is the volume of the dosing channel, A is the area of the co-culture membrane, t is the total time of effluent flow, C
r is the measured concentration of Cascade Blue in the receiving channel effluent, and C
d-out is the measured concentration of Cascade Blue in the dosing channel effluent.
RT-qPCR
Total RNA was extracted from vaginal epithelial cells from pre-differentiated (day 5 of expansion) and differentiated (day 10 of differentiation; exposed to 0.4 nM and 4 nM of β-estradiol for 10 days) vagina chips using QIAzol lysis reagent (Qiagen, cat. no. 79306). Complimentary DNA was synthesized using a SuperScript VILO MasterMix (Invitrogen, cat. no. 11755-500). The cellular gene-expression levels were determined using RT–qPCR, according to the TaqMan fast advanced master mix (ThermoFisher, cat. no. 4444964) with 20 μL of a reaction mixture containing gene-specific primers (ThermoFisher) for estrogen receptor (ESR1, Hs01046816), progesterone receptor (PGR, Hs01556702), phosphoenolpyruvate carboxykinase 1 (PCK1, Hs00159918), claudin 17 (CLDN17, Hs01043467), glucagon receptor (GCGR, Hs00164710), keratin15 (KRT15, Hs00951967), and zonula occludens-1 (ZO-1, Hs01551871). The expression levels of the target genes were normalized to GAPDH (Hs04420632).
Computational simulations
Using COMSOL Multiphysics 5.5 (COMSOL, Inc.) a two-dimensional model of two-channel microfluidic device was developed. The co-culture window was used to model the oxygen gradient with 80 μm epithelium and 50 μm stroma in the apical and basal channel, respectively. The apical and basal PDMS blocks are 3.5 mm and 1 mm thick respectively, and the PDMS membrane is 50 μm thick. The 2D oxygen distribution was simulated by coupling laminar flow with dilute species transport. The oxygen-saturated medium was fed through the inlet at the flow rate of 40 μl/h and goes out of the outlet after being partially consumed by the cells via aerobic respiration. Oxygen consumption by the epithelium and stroma was simulated using Michaelis-Menten-type kinetics. Navier-Stokes equations for incompressible flow were used to simulate fluid flow, and Fick’s second law was used to simulate oxygen transport through the PDMS, culture medium, epithelium, and stroma. Steady-state and time-dependent simulations were performed at 37 °C and with 145 mmHg atmospheric pO2 to simulate the conditions in the cell culture incubator. The entire vagina chip contained 145 mmHg atmospheric pO2 at t = 0 min and the time-dependent model was simulated for 200 min of continuous flow.
Isolation and selection of L. crispatus strains (C0006A1, OC1, OC2, and OC3)
As recently reported, vaginal microbiota dominated by Lactobacillus spp. comprise of multiple strains of the same species [27]. Consequently to mimic the ecology of these optimal vaginal microbiota, three L. crispatus multi-strain consortia were reconstructed that contain L. crispatus isolates cultivated from women with stable L. crispatus dominated microbiota who participated in the UMB-HMP study [28]. One optimal consortium (OC1) contained four L. crispatus strains (C0175A1, C0124A1, C0112A1, and C0059A1), while two of the optimal consortia (OC2 and OC3) contained three L. crispatus strains (OC2: C0175A1, C0124A1, and C0059A1 and OC3: C0175A1, C0112A1, and C0006A1); C0006A1 contains a single strain that is also found within OC3 consortium.
Isolation and selection of Gardnerella strains and consortia (BVC1 and BVC2)
In non-optimal vaginal microbiota, Gardnerella species are typically found as dominant bacteria [7,8,9] accompanied by other frequent taxa such as Prevotella species and Atopobium species [2]. To mimic the ecology of non-optimal vaginal microbiota, two dysbiotic consortia (BVC1 and BVC2) were reconstructed from isolates cultivated from women with asymptomatic BV. The first contained complex consortia of taxa (BVC1: G. vaginalis E2, G. vaginalis E4, P. bivia BHK8, and A. vaginae) and second contained two strains of G. vaginalis (BVC2: G. vaginalis E2 and E4). Recent studies have highlighted genomic diversity among Gardnerella spp. and the co-existence of multiple strains and species within an individual. The Gardnerella isolates used in this study were selected because they represent distinct genomic groups (C0011E2 and C0011E4), exhibit phenotypic diversity in vitro, and were co-resident, meaning that they were co-isolated from a single participant in the UMB-HMP study [28]. P. bivia (strain 0795_578_1_1_BHK8) and A. vaginae (strain 0795_578_1_1_BHK4) are prevalent species in Lactobacillus-deficient vaginal microbiota. The two strains used in this study were co-resident, isolated from a single participant in the Females Rising Through Education Support and Health study [29].
Construction of the multi-strain L. crispatus consortia and inoculation in the vagina chip
Each unique L. crispatus strain was grown overnight at 37 °C in De Man, Rogosa, and Sharpe (MRS) broth (Fisher Scientific, cat. no. 288210) under complete anaerobic conditions (83% N2, 10% CO2, 7% H2) in an anaerobic chamber. Subcultures were made from overnight cultures and once mid-logarithmic phase was reached, aliquot stocks were made and frozen at -− 80 °C with 16% sterile glycerol (MP Biomedicals, cat. no. 76019-966). To enumerate colony forming units (CFU)/mL in stocks, a single aliquot was thawed and spread plated on MRS agar (Hardy, cat. no. G117) under anaerobic conditions. Colonies were counted after 48 h of incubation at 37 °C and CFU/mL was calculated for stocks of each strain.
To generate consortia inoculum, required volumes from stocks of each strain were calculated in order to create equal L. crispatus strain cell density per 1 mL of inoculum. Cells were washed, spun, and resuspended in 1 mL of HBSS (LB/+G) and kept on ice. The apical channel of the vagina chip was inoculated with prepared L. crispatus consortia (~ 105 CFU in 35 μL) on day 11 of differentiation and cultured for 72 h. The chips were incubated statically at 37 °C and 5% CO2 for first 20 h of culture before starting the flow using the Zoe culture module. The basal channel was continuously perfused with in-house differentiation medium and apical channel was perfused for 4 h per day with customized HBSS (LB/+G) medium at a volumetric flow rate of 40 μL/h. Non-adherent bacterial CFU were quantified by measuring their presence in chip effluents (160 μL) collected at 24-, 48-, and 72-h post-inoculation and adherent bacteria were measured within epithelial tissue digests at 72 h.
Culture of a non-optimal Gardnerella vaginalis containing consortium in the vagina chip
Two G. vaginalis strains (G. vaginalis E2 and G. vaginalis E4) and two other anaerobic bacteria found in non-optimal microbiota of patients with BV (P. bivia BHK8 and A. vaginae) were grown individually in peptone, yeast, and tryptone (with hemin and vitamin K1) broth at 37 °C under complete anaerobic conditions (83% N2, 10% CO2, 7% H2) in an anaerobic chamber. Subcultures were made from overnight cultures and once mid-logarithmic phase was reached, aliquot stocks were made and frozen at – 80 °C with 16% sterile glycerol (MP Biomedicals, cat. no. 76019-966). To enumerate CFU/mL in stocks, a single aliquot was thawed, serial diluted, and spread plated on Brucella blood agar (with hemin and vitamin K1) (Hardy, cat. no. W23) under anaerobic conditions. Colonies were counted after 72 h of incubation at 37 °C and CFU/mL was calculated for stocks of each strain.
We then generated two consortia: one containing two G. vaginalis strains along with P. bivia BHK8 and A. vaginae species (BV Consortium 1, BVC1) and the other containing only G. vaginalis E2 and G. vaginalis E4 (BVC2). To generate these consortia, required volumes from stocks of each of the four bacterial strains were calculated in order to create equal strain cell density per 1 mL of inoculum. Cells were washed, spun, and resuspended in 1 mL of HBSS (LB/+G) and kept on ice. The apical channel was inoculated with the prepared BVC1 or BVC2 consortia (~ 106 CFU in 35 μL) and then chips were incubated statically at 37 °C and 5% CO2 for 20 h before starting the flow using the Zoe culture module. The basal channel was continuously perfused with in-house differentiation medium and apical channel was perfused for 4 h per day with customized HBSS (LB/+G) medium at a volumetric flow rate of 40 μL/h.
Bacterial enumeration from vagina chip co-culture
To enumerate all cultivable bacteria in the effluents, effluent samples (50 μL out of total 160 μL collected after 4 h of flow at 40 μL/h) were collected at 24, 48, and 72 h, diluted with glycerol to a final concentration of 16%, and frozen at – 80 °C. L. crispatus samples were spread plated on MRS agar under complete anaerobic conditions. After 48 h of incubation, colonies were counted, and CFU/mL was calculated for each sample. Effluent samples from the vagina chips containing BVC2 and BVC1 consortia were plated on Brucella blood agar (with hemin and vitamin K1) (Hardy, cat. no. W23) at 37 °C under complete anaerobic conditions. After 72 h of incubation, colonies were counted, and CFU/mL was calculated for each sample. To enumerate all cultivable bacteria adhered in the vagina chip, the whole cell layer was digested for 3 h with digestion solution containing 1 mg/mL of collagenase IV (Gibco, cat. no. 17104019) in TrypLE (ThermoFisher, cat. no. 12605010). Cell layer digest was then diluted with glycerol to a final concentration of 16% and frozen at – 80 °C. Digestion samples were processed in the same way as effluent samples and CFU/mL was calculated for each chip digest. CFU/chip was calculated from CFU/mL by taking into account the volume collected for effluent (at 24, 48, and 72 h) and digest (at 72 h). Percent adherence of OC1, OC2, OC3, and C0006A1 in vagina chip was calculated as the number of viable bacteria recovered from the chip epithelium digest at 72-h post-inoculation compared to the T0 inoculum.
Strain ratio analysis
DNA was extracted using the Qiagen AllPrep PowerViral DNA/RNA extraction kit (Qiagen; Hilden, Germany; Cat. 28000-50) from a 200-μL aliquot of the L. crispatus consortia inocula and from 200 μL of vaginal epithelial tissue digests after 72 h of co-culture with each of the L. crispatus consortia. Four co-culture replicates were performed for each L. crispatus consortia. Following DNA extraction Illumina shotgun sequence libraries were prepared using the Kapa HyperPrep kit according to manufacturer specifications (Roche; Basen, Switzerland). Libraries were sequenced on an Illumina NovaSeq S4 flow cell (Illumina; San Diego, CA) yielding on average 45 million (range 37.6–67.6 million) pairs of 150 bp reads. Human reads were identified and removed using BMtagger [30]. Sequence datasets contained on average, 97.4% human reads (range 96.3–98.3%). No human reads were identified in the inocula. Ribosomal RNA sequence reads were removed using sortmeRNA [31] (version 2.1) and the remaining reads were subjected to quality filtering and trimming using fastp [32] (version: 0.21, sliding window size: 4 bp, minimum average q-score: 20). The relative abundance of each L. crispatus strain in the samples was estimated by mapping the sequence reads to a database containing strain-specific marker genes. To build this database, single-copy genes uniquely present in the genomes of individual L. crispatus strains used in the experiments were identified using OrthoMCL [33] (all-versus-all BLAST e value threshold 10−5, 70% percent identity, 70% overlap). Reads were mapped to the marker gene database using Bowtie2 [34] and per gene coverage was estimated using SAMtools [35]. L. crispatus strain composition was determined using the median coverage of each strain’s marker genes relative to sum of median coverage values for all strains in the consortia.
Lactate and pH
For lactate analysis, samples from apical effluent were collected at every 24-h timepoint of the experiment and briefly equilibrated under anaerobic conditions (83% N2, 10% CO2, 7% H2) in an anaerobic chamber. Cells in each sample were pelleted and supernatant was collected then stored at 4 °C. D-lactate and L-lactate concentrations were measured separately in each sample using BioAssay Systems EnzyChrom Lactate Assay Kits (cat. no. EDLC-100 and ECLC-100 respectively) according to the manufacturer’s protocol. During effluent collection, pH was measured using pH paper (Micro Essential, Hydrion 325) for all chips.
Cytokines and chemokines analysis
Samples (100 μL) of the apical effluents from vagina chips were collected and analyzed for a panel of cytokines and chemokines, including TNF-α, INF-y, IL-1α, IL-1β, IL-10, IL-8, IL-6, MIP-1α, MIP-1β, IP-10, TGF-β, and RANTES using custom ProcartaPlex assay kits (ThermoFisher Scientific). The analyte concentrations were determined using a Luminex 100/200 Flexmap3D instrument coupled with the Luminex XPONENT software.
Statistical analysis
All of the results presented are from at least three independent experiments and all of the data points shown indicate the mean ± standard deviation (s.d.) from n > 3 organ chips unless otherwise mentioned. Tests for statistically significant differences between groups were performed using one-way ANOVA followed by Tukey multiple comparison, statistical analyses were performed using GraphPad Prism 9.0.2.