Impact of a multi-strain L. crispatus-based vaginal synbiotic on the vaginal microbiome: a randomized placebo-controlled trial

Abstract

A clinical trial of a multi-strain vaginal synbiotic (NCT05659745, registered 12/19/2022 at clinicaltrials.gov) led to an optimal vaginal microbiome dominated by L. crispatus (CST I). The synbiotic led to a significant increase in L. crispatus compared to placebo (p < 0.05), and conversion to CST I was significantly higher with the vaginal synbiotic than with placebo (90 vs 11%; p < 0.002). Mechanistically, the synbiotic reduced Gardnerella vaginalis and Candida, clinically important microbes.

Introduction

Compared to an optimal vaginal microbiome dominated by Lactobacillus crispatus (Community State Type, CST I)^1,2, vaginal dysbiosis is linked to several gynecological and obstetric conditions, including genital infections³ and pregnancy complications⁴. Biologically, vaginal dysbiosis is associated with increased vaginal inflammation and increased abundance of a diverse set of anaerobic microbes associated with degradation of the protective mucus barrier⁵.

Microbiome-based interventions using exogenous lactobacilli to help restore an optimal vaginal microbiome have demonstrated preliminary clinical benefit in a number of clinical conditions^6–9. However, the efficacy rates of probiotics for vaginal health are heterogeneous¹⁰ and there is a paucity of data demonstrating that microbiome interventions with live beneficial organisms lead to an optimal vaginal microbiome. This is likely due to several reasons: (1) probiotics often include species not native to the human vagina (e.g., L. acidophilus, a gastrointestinal microbe) which have not been shown to effectively colonize the vaginal microenvironment, (2) probiotics often are administered orally, which anatomically limits direct vaginal colonization¹¹, (3) probiotics rarely include additional ingredients to support colonization of introduced strains, despite preclinical data indicating vaginal strains can utilize glycogen breakdown products for growth¹², and (4) few studies have rigorously assessed the correct delivery technology (oral/vaginal)¹³.

Previous studies using a single strain vaginally administered Lactobacillus crispatus product demonstrated that clinical efficacy in the reduction of recurrent urinary tract infections and bacterial vaginosis¹⁴ was associated with high levels of colonization after antibiotic pretreatment. We hypothesized that the inclusion of (1) multiple strains better representative of a native vaginal ecology would enable effective colonization without the need for antibiotic pretreatment and (2) additional ingredients supporting the growth of L. crispatus in the vaginal environment optimize colonization. Thus, we sought to develop a novel vaginal synbiotic to optimize colonization of a multi-strain consortium of vaginally-derived L. crispatus strains.

Specifically, we identified, isolated, and characterized three strains of L. crispatus, which were associated with a stable L. crispatus dominant microbiome in a deeply sampled longitudinal cohort¹⁵ and which together represent an average coverage of 70.2% of the L. crispatus pangenome based on analysis of optimally dominated and stable microbiomes using the VIRGO framework¹⁶. These strains were shown preclinically to inhibit multiple strains of key microbes associated with vaginal dysbiosis– Gardnerella and Candida– using a parallel streak assay (Fig. 2e). This antimicrobial activity was enhanced when assembled in a consortium. We combined these strains with a nutrient complex designed to support the growth of L. crispatus via provision of growth substrates (maltose, glutamine) and optimization of the vaginal environment (including calcium L-lactate and cysteine) and tested alternate formulations in the pilot clinical study reported here.

Fig. 2. Colonization of L. crispatus and modulation of biomarkers of dysbiosis.

a Mean relative abundance of L. crispatus in the multi-strain synbiotic vaginal tablet (VS-01) and placebo arms (n = 19) among participants not in CST I at baseline (<50% relative abundance of L. crispatus, as defined in Methods). * indicates p < 0.05 difference between the multi-strain synbiotic and placebo; b The proportion of participants who converted to CST I at each timepoint is shown for Part A and Part B among those with a baseline dysbiotic/non-optimal microbiome (n = 39), along with associated p values; c VS-01 significantly decreased the mean abundance of Candida spp. (C. albicans, C. tropicalis, and C. parapsilosis) compared to placebo from baseline to D21 (p < 0.05); d VS-01 significantly decreased the abundance of Gardnerella vaginalis, a key dysbiotic/non-optimal bacterium, from baseline to D21 (p < 0.05); e Each strain in VS-01 and the VS-01 consortium inhibited the growth of Gardnerella spp. in vitro. The fraction of inhibition is shown across all strains, with 1 representing complete inhibition and 0 representing no inhibition; f VS-01 significantly decreased sialidase genes from baseline to D21 among participants with baseline detectable sialidase genes (p < 0.05); g VS-01 significantly decreased the concentration of IL-1ɑ from baseline to D21 (p < 0.01).

We conducted a randomized, placebo-controlled clinical trial (n = 70; Fig. 1) with Part A comparing a multi-strain synbiotic vaginal tablet formulation (VS-01) to a placebo, and Part B comparing three formulations (synbiotic vaginal capsule and synbiotic oral capsule, each containing the same three-strain consortium, and an over-the-counter oral supplement containing L. crispatus, Lacticaseibacillus rhamnosus, Lactobacillus gasseri, and Lactobacillus jensenii). The vaginal tablet was formulated for slow release of synbiotic ingredients, while the vaginal capsule was formulated for rapid release (Methods). Participants were randomized to each arm for the 1-month intervention period immediately following menses and assessed through Day 21 (D21) (Fig. 1) with post-dosing follow-up at Day 35 (D35) and Day 51 (D51). Key outcomes included the relative abundance of L. crispatus, taxonomic/functional microbiome profiles as assessed by metagenomic sequencing, the relative abundance of specific organisms and functional pathways associated with vaginal dysbiosis, and cytokine and chemokine profiles using a custom 33-plex Luminex panel. Safety and tolerability was assessed in all participants.

Fig. 1. Study design.

Part A of the trial assessed a multi-strain L. crispatus vaginal synbiotic tablet compared to a placebo, and Part B of the trial assessed several formulations (vaginal multi-strain L. crispatus synbiotic capsule, oral multi-strain L. crispatus synbiotic capsule, over-the-counter oral supplement). Participants collected baseline samples (D0) 5–10 days prior to the onset of menses (M). Twenty-four hours after the conclusion of menstrual bleeding, the intervention period was initiated (as shown in the schematic), and samples were collected prior to the next menses.

The baseline distribution of vaginal microbiome community state types (CSTs) was consistent with prior literature on microbiome profiles present in a population of healthy women¹⁵. The proportion of individuals with a L. crispatus-dominated CST I ranged from 25 to 44% at baseline (Supplemental Table 1) and did not significantly differ between arms. We assessed colonization of L. crispatus in individuals not in CST I at baseline.

Results indicate that L. crispatus successfully colonized individuals who received the multi-strain synbiotic vaginal tablet, and that administration of the tablet drove conversion to an optimal vaginal microbiome. Specifically, the multi-strain synbiotic vaginal tablet arm showed a significantly higher relative abundance of L. crispatus compared to placebo at D21 (p < 0.05) among participants with baseline dysbiosis/non-optimal microbiome (Fig. 2a). Consistent with the increase in L. crispatus relative abundance, the multi-strain synbiotic vaginal tablet arm demonstrated a 90% conversion to an optimal CST I microbiome (defined as L. crispatus >50%) compared to 11% in the placebo (p < 0.002) among participants with baseline dysbiosis/non-optimal microbiome (Fig. 2b).

Further, the summed relative abundance of strains in the multi-strain synbiotic increased significantly over baseline during the dosing regimen at D7 (p < 0.05), D14 (p < 0.001), and D21 (p < 0.01) in the multi-strain synbiotic vaginal tablet arm. The presence of all three strains both during and after active dosing (Supplemental Table 1) is consistent with the hypothesis that a multi-strain consortium facilitates effective colonization in the vaginal environment, though not definitive in the absence of a clinical trial directly comparing single and multi-strain formulations.

Notably, conversion to a CST I microbiome persisted after the end of active dosing for individuals who received the multi-strain synbiotic vaginal tablet. Both at 14 days post-dosing (D35) and at 30 days post-dosing (D51), 66% of individuals who had converted to a CST I microbiome at D21 remained in CST I, and overall CST I conversion rates at both timepoints were 54.6%. Post-dosing conversion rates were statistically significant relative to placebo at both timepoints (D35 and D51, p < 0.02). These findings suggest the possibility of sustained conversion to CST I, although additional studies are needed to identify the optimal dosing regime to drive long-term persistence of CST I conversion.

Additionally, the multi-strain synbiotic vaginal tablet demonstrated statistically significant, superior conversion to an optimal L. crispatus-dominated microbiome compared to alternate formulations with the same strains (multi-strain synbiotic vaginal capsule and multi-strain synbiotic oral capsule), among participants with baseline dysbiosis/non-optimal microbiome (Fig. 2b). The over-the-counter oral supplement did not lead to colonization.

The multi-strain synbiotic vaginal tablet reduced levels of microbes associated with vaginal dysbiosis, which contributed to four key mechanistic findings (Fig. 2c–g). First, the mean relative abundance of Candida spp. was significantly reduced in the multi-strain synbiotic vaginal tablet arm (236-fold) compared to placebo (p < 0.05) at D21 among participants with detectable Candida at baseline (Fig. 2c). Second, the relative abundance of Gardnerella vaginalis was significantly decreased in the multi-strain synbiotic vaginal tablet arm (p < 0.05) from baseline to D21 (Fig. 2d), but was not significantly decreased in the Placebo arm. This result aligned with preclinical data demonstrating that the strains comprising the multi-strain synbiotic inhibited the growth of multiple species of Gardnerella, including G. vaginalis, using a parallel streak assay (Fig. 2e). Third, the abundance of mucin-degrading sialidase genes was significantly decreased in the multi-strain synbiotic vaginal tablet arm (p < 0.05) from baseline to D21, suggesting the potential for improvement in the protective mucus barrier (Fig. 2f). The abundance of these genes was not significantly decreased in the Placebo arm. Fourth, levels of the pro-inflammatory cytokine IL-1ɑ were significantly decreased in the multi-strain synbiotic vaginal tablet arm from baseline to D21 (p < 0.01; Fig. 2g), but were not significantly decreased in the Placebo arm. Overall, there were no reported serious adverse events in any trial arm, and all formulations were well-tolerated, with no participants lost to follow-up.

This is the first randomized clinical study to demonstrate that a vaginally applied multi-strain L. crispatus vaginal synbiotic leads to an optimal CST I vaginal microbiome, a biomarker associated with a reduction in microbiome-mediated women’s health conditions. Three strengths of this study are: (1) it is the first randomized clinical study of a vaginally administered multi-strain L. crispatus synbiotic, (2) the direct comparison of oral and alternate vaginal formulations with the same strains, and (3) the use of novel bioinformatic approaches coupled with deep metagenomic sequencing (up to 100 MM reads) for outcome evaluation.

A multi-strain synbiotic vaginal tablet formulated for slow-release demonstrated superior colonization of L. crispatus and conversion to an optimal CST I microbiome compared to both placebo and a fast-release vaginal capsule. This advantage may result from the tablet’s mucoadhesive properties and slow dissolution due to hydroxypropyl methylcellulose (HPMC), which is absent in the capsule. Mucoadhesive formulations are widely used in the delivery of vaginally administered agents¹⁷. However, differences in supporting ingredients between the tablet and capsule may have also influenced colonization. Notably, the synbiotic vaginal tablet showed superior efficacy in the conversion of the vaginal microbiome to a L. crispatus-dominated state compared to oral administration of the same synbiotic formulation and a commercially available oral supplement product.

Biologically, the multi-strain L. crispatus synbiotic reduced the relative abundance of microbes associated with vaginal dysbiosis and the abundance of Candida, the most common source of vaginal yeast infections. We demonstrated conversion to CST I and colonization of product strains without antibiotic pretreatment, suggesting that administering the synbiotic immediately post-menses may improve the colonization of L. crispatus. Limitations of this study include asymptomatic and otherwise healthy volunteer participants and translational biomarker-based endpoints.

In conclusion, the multi-strain L. crispatus synbiotic effectively establishes an optimal and protective vaginal microenvironment. This synbiotic warrants further exploration with randomized controlled trials, including clinical endpoints.

Methods

Ethics approval

This study was registered at clinicaltrials.gov as NCT05659745 on 12/19/2022, adhered to ethical guidelines and received IRB approval (Argus Independent Review Board). Research was performed in accordance with the Declaration of Helsinki.

Participant recruitment and sample collection

This decentralized trial enrolled healthy women of reproductive age (40% non-White, 60% White) between 18 and 54 years (Demographics: Supplemental Table 2) across the United States. The trial started in November 2022 and was completed in April 2023. Participants were prescreened with the following inclusion criteria: (1) history of regular menses every 21–35 days for 6 months prior to enrollment, (2) agreed to not use specified intravaginal products during the study product use, (3) willing to use an intravaginal suppository or take a dietary supplement during the study, (4) willing to self-collect vaginal swabs, (5) willing and able to provide informed consent in English. Participants were excluded if they: (1) were pregnant or planned to become pregnant during the study, (2) had two or more amenorrheic months in the past 6 months prior to enrollment, (3) were allergic to any components of the supporting or excipient ingredients in the capsules, tablets, placebo, or over-the-counter product, (4) lived in the state of New York.

At baseline, participants were randomized to a 1:1 allocation ratio for the study. The randomization sequence was generated using a computer-generated random number list and prepared by the contract research organization prior to the study. All participants and outcome assessors were blinded to treatment allocation throughout the entire study. Overall, packaging was similar, containing an equal number of intervention products. Each intervention was paired with a specific set of participant administration instructions (e.g., oral vs vaginal administration in Part B). As this study was exploratory in nature and there was an absence of prior data on effect size, a sample size of convenience was adopted based on discussion with subject matter experts.

For Part A, 34 participants were randomized with 17 participants in the multi-strain synbiotic vaginal tablet arm and 17 participants in the placebo arm. All participants were included in the analysis with 0 lost to follow-up. For Part B, 36 participants were randomized with 14 in the multi-strain synbiotic vaginal capsule arm, ten participants in the multi-strain synbiotic oral capsule arm, and 12 in the OTC oral capsule arm. All participants were included in the analysis with 0 lost to follow-up.

Vaginal swabs were collected for metagenomic and cytokine/chemokine analysis at baseline and D1, D4, D7, D14, D21, D35, and D51. Participants were instructed to take swabs prior to use of the vaginal synbiotic tablet. Genomic samples were collected using the OMNIgene Vaginal kits (DNA Genotek, Ontario, Canada), preprocessed with proteinase K, aliquoted, and stored at −80 °C before processing for DNA extraction (ZymoBIOMICS-96 MagBead DNA Kit, Zymo Research, Irvine, CA), quantification, and shotgun metagenomic DNA sequencing using an Illumina NovaSeq 6000 (minimum depth 45 M read pairs per sample, 2 × 150 bp configuration).

Study intervention

The multi-strain synbiotic tablet (850 mg) comprises live probiotic strains (L. crispatus LUCA103, L. crispatus LUCA011, L. crispatus LUCA009 in a 1:1:1 proportion), a nutrient complex (maltose, calcium lactate, glutamine, magnesium citrate, and cystine), and other ingredients (plant-derived cellulose, magnesium stearate, and hydroxypropyl methylcellulose [HPMC]). The multi-strain vaginal capsule (300 mg) comprises live probiotic strains (L. crispatus LUCA103, L. crispatus LUCA011, L. crispatus LUCA009), a nutrient complex (maltose, calcium lactate), and other ingredients (magnesium stearate, plant-derived cellulose, non-animal-derived outer capsule coating).

Metagenomic analysis

For bacterial taxonomic analysis, we developed a vaginal microbiome reference database using a combination of genomes from bacterial isolates and metagenome-associated genomes (MAGs) assembled from public and internal datasets. All quality-controlled samples were aligned to the vaginal microbiome database with XTree (0.92i)¹. For analysis of the synbiotic strains, we set a conservative detection threshold based on genomic coverage and demonstrated high sensitivity and specificity using a custom in silico approach. Detailed benchmarking demonstrated the ability of our method to detect each of the three VS-01 strains against a background of closely related L. crispatus isolate genomes (Supplemental Fig. 2).

For fungal taxonomic analysis, a fungal XTree genome database was created using fungal genomes from NCBI RefSeq (R220) with taxonomy assigned using t2gg¹⁸. Metagenomic samples were profiled using this database as described above.

Community state types (CSTs) were assigned to each sample using a rule-based decision tree considering the relative abundance of specific Lactobacillus species or species sets. Samples were first split into those with a dominance (≥50% relative abundance) of a CST-relevant Lactobacillus or Lactobacillus species set (in the case of CST II and CST V) versus those with no dominant CST-relevant lactobacilli. Lactobacillus-dominant samples were sorted into CST I, II, III, or V based on the maximally-abundant species or species set. Non-Lactobacillus- dominant samples with ≥50% combined relative abundance of species from genera such as Atopobium, Enterococcus, and Gardnerella were assigned to CST IV. Samples with no dominance of any relevant species or species set were considered “unclassified”. The full distribution of CSTs across all arms, all timepoints is shown in Supplemental Fig. 1.

For identification and quantification of sialidase genes, the quality-controlled metagenomic reads were assembled using megahit v1.2.9 3, gene inference was performed on the resulting contigs, and the KEGG V107 HMM database¹⁹, was used to annotate each gene in each sample, and gene abundance was computed based on read mapping and normalization to contig coverage.

Cytokines

Vaginal swab samples were self-collected by participants and placed into 1 mL of a buffer that affords protein stability up to 7 days at room temperature, shipped at room temperature and stored at −80 °C until use. Levels of a custom panel of 33 cytokines²⁰ were measured in the clarified swab extracts using a multiplexed immunoassay kit (MILLIPLEX®, MilliporeSigma Corp., Burlington, MA) read out on a Bio-Plex 200 readout system (Bio-Rad, Hercules, California) utilizing Luminex® xMAP fluorescent bead-based technology (Luminex Corporation, Austin, TX).

Statistical analysis

Analysis of changes in L. crispatus relative abundance were performed using linear mixed effects models on logit-transformed data. CST conversion rates were statistically compared using Fisher’s exact test. Analyses of Candida, Gardnerella vaginalis, and sialidase gene abundance were performed on individuals in Part A with metagenomic data (baseline and D21) that met quality thresholds. All fungal species labeled with genus-level NCBI taxonomy “Candida” were pooled using the mean of their log-transformed relative abundances. The data was filtered to those participants who had detectable Candida at baseline (n = 8); the difference between the multi-strain synbiotic tablet and placebo arms at D21 was assessed with a one-sided two-sample t-test. Analysis of G. vaginalis was performed by pooling relative abundance data from all species annotated as synonyms of Gardnerella vaginalis in the taxonomy database GTDB R214²¹. A paired Wilcoxon test was performed to test the difference between baseline and D21. Analysis of sialidase genes was performed in those individuals who had detectable sialidases at baseline (n = 10). A one-tailed paired t-test was run to test the difference in log₁₀-normalized sialidase abundances between baseline and D21 in the multi-strain synbiotic tablet arm.

Cytokine analysis was performed on those individuals in Part A (Fig. 1) with samples that met quality thresholds (n = 15). Linear mixed effects models were applied with the focal cytokine as the response variable. All analyses were performed using R Version 4.2.2.

In vitro Gardnerella inhibition by L. crispatus

L. crispatus LUCA103, L. crispatus LUCA011, and L. crispatus LUCA009 were grown anaerobically at 37 °C for 24 h in De Man, Rogosa, and Sharpe (MRS) broth, and OD600 adjusted to 0.4. For the consortium, equal volumes of OD600 = 0.4 cultures were mixed to create a 1:1:1 ratio of all three strains. The cultures were parallel streaked onto Tryptic Soy Broth (TSB) agar plates supplemented with 5% horse serum using a sterile swab and incubated anaerobically at 37 °C for 24 h. In addition, all target pathogens were individually grown in TSB supplemented with 5% horse serum under anaerobic conditions at 37 °C for 24 h. To test for antimicrobial ability of L. crispatus strains, 10 μL of the 24 h cultures of 11 Gardnerella strains were pipetted between the parallel L. crispatus streaks. Specifically, the pathogens tested were: G. vaginalis, G. pioti, G. leopoldi, G. swidsinkii, and seven unclassified Gardnerella spp. Plates containing both L. crispatus and Gardnerella spp. spots were incubated for a further 24–48 h anaerobically at 37 °C. Antimicrobial activity was scored using the following five-scale system: 0, 25, 50, 75, and 100% inhibition. Results are averages of biological duplicates.

Author contributions

J.R. and D.G. conceptualized the study. S.S. and D.G. performed the experiments and analyzed data. S.S. wrote the initial draft of the manuscript, assisted by J.R., D.G. and D.M.H. All authors edited the manuscript and approved the final version for submission.

Data availability

Requests for raw sequencing data may be directed to Dr Jacques Ravel, jacques@luca.bio.

Competing interests

The authors declare no competing non-financial interests but the following competing financial interests: D.G. and S.S. are employees/shareholders at Seed Health. J.R. is co-founder of LUCA Biologics, a biotechnology company focusing on translating microbiome research into live biotherapeutic drugs for women’s health. D.M.H. serves as an advisor/consultant at Seed Health.

Supplementary information

The online version contains supplementary material available at 10.1038/s41522-025-00788-6.