• Human metagenomic co-occurrence: A.m. rare (4.1%), but strongly co‑associated with GBS (87.1% of A.m.+ samples also GBS+) and higher GBS reads where A.m. is present (749 pregnant swabs)
• In vitro: A.m. + GBS co-aggregate; presence of A.m. increases GBS adherence to human vaginal epithelial cells (VK2/E6E7)
• Transcriptomics (GBS RNA‑seq during hVEC infection): 258 genes uniquely altered in co‑infection vs mono; cell‑envelope, capsule and pilus genes differentially regulated
• In vivo (CD‑1 mice): daily intravaginal A.m. treatment reduces GBS burden by ≈2.1–2.2 log CFU in lavage and tissues by day 5

Human cohort: A.m. — GBS co-occurrence (counts)

Note: paper reports 4.1% (≈31/749) A.m. positive; of those 87.1% were GBS positive (≈27/31) and GBS reads were higher in A.m.+ samples — raw metagenomic read counts not released here but reported in manuscript Human metagenome analysis (preprint)

Key experimental findings (visual first)

1. Co‑aggregation & adherence: GBS + A.m. co-aggregate in PBS and form mixed aggregates visualized by confocal microscopy; A.m. increases GBS adherence to VK2 hVECs (COH1 ≈1.6x; CJB111 ≈2.8x) and GBS increases A.m. adherence (~1.8x) — replication across multiple GBS serotypes and vaginal isolates reported.
2. Transcriptome shifts: RNA‑seq (GBS RNA) after 4h hVEC infection: GBS media control vs GBS+hVEC vs GBS+A.m.+hVEC. Using FDR≤0.05, |FC|≥1.5: 201 genes uniquely changed in mono-infection, 258 uniquely changed in co-infection, 245 shared; ontology shifts show cell-envelope genes up in co-infection but down in mono.
3. Surface factors validated: ΔcpsD (capsule-deficient) fails to show A.m.-dependent aggregation increase (capsule required for aggregation); Δbp-2b (PI‑2b pilus mutant) fails to show A.m.-dependent increase in adherence (pilus required for A.m.-mediated adherence increases).
4. In vivo effect: Daily intravaginal A.m. treatment in estrus-synchronized CD‑1 mice reduces GBS lavage CFU by ~2.2 log at day 5 and similarly reduces GBS in vaginal and cervical tissues by ~2.1–2.2 log; microscopy showed increased GBS aggregation with A.m. treatment.

All above data and exact statistics are reported in the preprint (figures 1–7; methods and sample sizes included) Experimental methods & stats (preprint)

Critical appraisal — strengths & limitations

• Strengths: multi-angle approach (human metagenome reanalysis, in vitro aggregation/adherence, RNA-seq with biological triplicates, targeted mutant validation, and in vivo mouse model) increases internal coherence and mechanistic depth; data deposited (GEO) improves transparency Strengths & data availability (preprint)
• Limitations / blindspots: (a) Human co-occurrence is associative and derived from one cohort (Paris hospitals dataset) with low A.m. prevalence (≈4.1%) — risk of sampling/stochasticity; (b) in vitro aggregation in PBS lacks mucosal/mucin context and host secretions which shape interactions; (c) single A.m. strain (Muc T, ATCC BAA‑835) used — A.m. is genotypically/phenotypically diverse; (d) murine model uses intravaginal dosing every day (non-physiologic dosing regimen relative to rare human vaginal A.m. colonization) — possible ecological perturbation; (e) immunological readouts (host response) are limited — mechanism of GBS reduction in vivo (immune-mediated clearance vs sequestration/aggregate‑loss) remains unresolved Limitations acknowledged (preprint)
• Potential confounders: antibiotic histories of human cohort unknown here, detection thresholds for metagenomic reads (≥5 reads binning) can create false negatives/positives, and mouse estrus synchronization with estradiol alters mucus and immunity, which can change colonization dynamics.

Mechanistic reading & alternative models

Leading mechanistic hypothesis from authors: A.m. physically co‑aggregates with GBS (capsule-dependent) and increases pilus-dependent adherence to epithelial cells (PI‑2b); A.m. also shifts GBS transcriptional state (upregulation of some cps genes and pilus genes, and Srr2 secretion/glycosylation genes) which may change surface properties and biofilm/aggregation. In vivo, persistent A.m. dosing increases aggregation and reduces free GBS CFU, possibly by sequestration or enhanced clearance — authors propose probiotic potential.

Alternative models consistent with data: (1) A.m.-induced GBS transcriptional changes could be indirect via host epithelial signaling (A.m. modifies host mucins or cytokines) rather than direct microbial cross-talk; (2) A.m. sialidase activity might alter GBS capsule sialylation, changing immune recognition and clearance; (3) daily high‑dose A.m. may displace GBS by nutrient/space competition or by altering mucus rheology. Discriminating these requires host immune readouts, capsule composition assays, and testing A.m. mutants lacking sialidase(s).

Where the evidence is weakest (what could falsify the central claims)

• If independent human cohorts do not replicate the A.m.–GBS co-occurrence or show opposite associations, the human correlate is fragile.
• If in vivo experiments with physiologic/low-dose A.m. colonization (not daily bolus) fail to reduce GBS or increase aggregation, the probiotic interpretation weakens.
• If purified A.m. sialidase de-sialylates GBS capsule in vitro and increases GBS susceptibility to phagocytosis, that would support a specific biochemical mechanism; conversely, demonstration that A.m. lacks functional sialidase activity on GBS would disfavor capsule‑targeting mechanisms.

Concrete follow-up experiments (testable & discriminating)

1. Compare daily vs single/physiologic A.m. intravaginal dosing in mice and include heat-killed / pasteurized A.m. controls to separate live‑microbe effects from structural components.
2. Use A.m. sialidase knockout(s) (or chemical inhibition) to test whether capsule sialic acid cleavage mediates GBS aggregation, capsule modulation, or clearance.
3. Perform capsule composition analysis (mass spectrometry of CPS) of GBS after co-incubation with A.m. to detect desialylation or other glycan changes.
4. Measure host innate immune activation (neutrophil influx, complement deposition, cytokine transcripts) in treated vs mock mice to determine whether clearance is immune-mediated.
5. Re-analyze additional human vaginal metagenomes (geographically diverse cohorts) using consistent detection thresholds and include host metadata (antibiotic use, sexual activity, obstetric history) to test generalizability.

Data & methods reproducibility

Raw RNA-seq available at GEO GSE306314 (paper states this). Genome reference used: S. agalactiae COH1 (NZ_HG939456.1). Software: CLC Genomics Workbench v21.0.5, R packages edgeR/EnhancedVolcano for visualization. Sufficient methodological detail provided for replication, but raw read-level metagenomic tables for the human cohort (per-sample read counts) should be made available to fully reproduce co-occurrence statistics.

Primary preprint: Marroquin et al., preprint (2025)

Open BGPT author reviews