Microbial-driven preterm labour involves crosstalk between the innate and adaptive immune response

Abstract

There has been a surge in studies implicating a role of vaginal microbiota in spontaneous preterm birth (sPTB), but most are associative without mechanistic insight. Here we show a comprehensive approach to understand the causative factors of preterm birth, based on the integration of longitudinal vaginal microbiota and cervicovaginal fluid (CVF) immunophenotype data collected from 133 women at high-risk of sPTB. We show that vaginal depletion of Lactobacillus species and high bacterial diversity leads to increased mannose binding lectin (MBL), IgM, IgG, C3b, C5, IL-8, IL-6 and IL-1β and to increased risk of sPTB. Cervical shortening, which often precedes preterm birth, is associated with Lactobacillus iners and elevated levels of IgM, C3b, C5, C5a and IL-6. These data demonstrate a role for the complement system in microbial-driven sPTB and provide a scientific rationale for the development of live biotherapeutics and complement therapeutics to prevent sPTB.

Fig. 1. Study design.

Women at high risk of preterm birth due to the presence of risk factors were recruited from preterm birth prevention clinics where cervical length ultrasound scans were performed, and cervical cerclage was placed where indicated. Cervicovaginal samples were taken at three timepoints: A (12–16 weeks), B (20–24 weeks) and C (30–34 weeks). Cervicovaginal fluid was analysed for mediators of inflammation (IL-8, IL-6, IL-1β, IL-10, TNF-α, IFN-γ, GM-CSF, IL-5, IL-4, IL-2, IL-18), complement proteins (C3b, C5 and C5a) and mediators of microbial recognition (MBL, IgM, IgG1-4, and IgA). Bacterial DNA was extracted and microbial composition was determined by next generation sequencing. The spontaneous preterm birth rate was 27.82% < 37 weeks and 15.79% < 34 weeks. Of the whole study population, 39.85% of women were identified with cervical shortening (defined as a cervical l length of ≤ 25 mm), compared to 59.46% of women who delivered preterm and 8.93% of women who delivered at term without intervention. Of the cerclages placed, 45.76% were placed due to a short cervix (ultrasound indicated) and 54.24% were placed prior on the basis of clinical history (history indicated). Monofilament was used in 52.54% and braided suture material was used in 45.76% of cases. There was one case where cerclage material type was not known. Figure created with BioRender.com.

Fig. 2. Cervicovaginal pro-inflammatory cytokines increase in mid-pregnancy in women who deliver preterm and early-preterm birth is associated with a predominance of Lactobacillus Iners or diverse vaginal microbial composition compared to late preterm birth.

Longitudinal sampling of cervicovaginal cytokines was performed in 122 women, of whom 34 delivered preterm, 51 delivered at term without intervention, and 37 delivered at term with an intervention. Concentrations of IL-8 (a), IL-6 (b), IL-1β (c), and IL-2 (d) taken at 12–16 weeks and 20–24 weeks are presented. A one-sided Wilcoxon matched pairs signed rank was used for statistical analysis. A heat map shows the distribution of community state types (CST) of studied women, with graphs representing the percentages of samples classed as Lactobacillus dominant (n = 295) and Lactobacillus deplete (n = 90), or classed by vaginal microbial composition; CST I (L. crispatus) (n = 155), CST II (L. gasseri) (n = 48), CST III (L. iners) (n = 100), CST IV (diverse) subdivided into CST IV-A (n = 0), CST IV-B (n = 35), CST IV-C (n = 24), and CST V (L. jensenii) (n = 23), a total of = 385 samples. e Vaginal microbial composition at each timepoint is presented in women who delivered <34 weeks (f), 34–37 weeks (g), at term without intervention (h), and in women who delivered at term following intervention (i), n = 133 women with cross sectional data. PT = preterm, T = term without intervention, TI = term with intervention. Source data are provided as a Source Data file.

Fig. 3. Cervicovaginal microbial composition modulates the local cytokines and leads to a pro-inflammatory response in women who deliver preterm.

Cervicovaginal concentrations of IL-8 (a), IL-6 (b), IL-1β (c), and IL-10 (d) are shown from samples classed as Lactobacillus dominant or deplete, n = 385 samples, n = 133 women. Statistical analysis was performed using a one-sided Mann–Whitney test. Cervicovaginal cytokine concentrations of IL-8 (e), IL-6 (f), IL-1β (g), and IL-10 (h), are shown relating to samples taken from a vaginal composition classed as CST I, II, III, IV-B, IV-C and V, n = 385 samples, n = 133 women. Statistical analysis was performed using the Kruskal–Wallis and Dunn’s multiple comparison’s test. Cervicovaginal concentrations of IL-8 (i), IL-6 (j), IL-1β (k), and IL-10 (l) were compared between women who delivered preterm or at term in those who were classed as being Lactobacillus deplete (n = 22), or abundant in CST III (L. iners) (n = 22) or CST IV (diverse) (n = 15). CST IV was subdivided into CST IV-B (n = 8), CSTIV-C1 (n = 1), CSTIV-C2 (n = 1) and CST IV-C3 (n = 5). Statistical analysis was performed using a one-sided Mann–Whitney test. Data are presented as median values and interquartile ranges (25th and 75th percentiles). Source data are provided as a Source Data file.

Fig. 4. Cervicovaginal concentrations of mediators of microbial recognition differ depending on vaginal microbial composition and birth outcome.

Cervicovaginal concentrations of MBL, IgM, and Ig1-IgG4 were measured and concentrations compared between 12–16 weeks and 20–24 weeks in women who delivered preterm, term without intervention, and at term following intervention a–f, n = 122. Statistical analysis was performed using a one-sided Wilcoxon matched pairs signed rank test. MBL, IgM, and IgG1-IgG4 concentrations were also compared between samples taken from women who were classed as CST I with CST II–V g–l, n = 385 samples from n = 133 women. The Kruskal–Wallis and Dunn’s multiple comparison’s test was used to determine statistical significance. Cervicovaginal concentrations of MBL, IgM, and Ig1-IgG4 were compared between women who delivered preterm or at term in those who were classed as being Lactobacillus deplete (n = 22), or abundant in either VMG 3 (L. iners) (n = 22) or VMG 4 (diverse) (n = 15). CST IV was subdivided into CST IV-B (n = 8), CSTIV-C1 (n = 1), CSTIV-C2 (n = 1) and CST IV-C3 (n = 5) (m–r). Statistical analysis was performed using a one-sided Mann–Whitney test. Data are presented as median values and interquartile ranges (25th and 75th percentiles). Source data are provided as a Source Data file.

Fig. 5. Vaginal microbiota activates complement in women who deliver preterm.

Cervicovaginal concentrations of C3b, C5, and C5a were measured and compared between 12–16 weeks and 20–24 weeks in women who delivered preterm, term without intervention, and at term following intervention (a–c), n = 122. Statistical analysis was performed using a one-sided Wilcoxon matched pairs signed rank test. C3b, C5, and C5a concentrations were also compared between samples taken from women who were classed as CST I with CST II–V (d–f) n = 385 samples from n = 133 women. The Kruskal–Wallis and Dunn’s multiple comparison’s test was used to determine statistical significance. Cervicovaginal concentrations of C3b, C5, and C5a were compared between women who delivered preterm or at term in those who were classed as being Lactobacillus deplete (n = 22), or abundant in either CST III (L. iners) (n = 22) or CST IV (diverse) (n = 15) (g–i). CST IV was subdivided into CST IV-B (n = 8), CSTIV-C1 (n = 1), CSTIV-C2 (n = 1) and CST IV-C3 (n = 5). Statistical analysis was performed using a one-sided Mann–Whitney test. Data are presented as median values and interquartile ranges (25th and 75th percentiles). Source data are provided as a Source Data file.

Fig. 6. Cervicovaginal IgM and complement positively correlate with pro-inflammatory cytokines in women who are lactobacillus deplete and who deliver preterm.

In women who are Lactobacillus deplete and who deliver preterm a positive correlation was seen between cervicovaginal complement and cytokines. Concentrations were log transformed and a one-sided Spearman’s correlation was performed between; MBL and IL-8 (a). IL-6 (b), IL-1β (c); IgM and IL-8 (d). IL-6 (e), IL-1β (f); C3b and IL-8 (g), IL-6 (h), IL-1β (I); C5 and IL-8 (j), IL-6 (k), IL-1β (l). n = 28 samples. Source data are provided as a Source Data file.

Fig. 7. Cervical shortening is associated with activation of the adaptive and innate immune response, especially in the presence of Lactobacillus iners.

Cervicovaginal immune mediators were analysed in women between 12–16 weeks gestation prior to intervention (a–e). Thirteen women had a cervical length (CL) ≤ 25 mm and 96 women had a cervical length >25 mm. Median concentrations of IgM (a), C5 (b), C5a (c), IL-6 (d), and IL-10 (e) are presented, n = 109 women. Statistical analysis was performed using a one-sided Mann–Whitney test. The percentage of women with a short cervix (≤25 mm) or long cervix (>25 mm) with L. iners at 12–16 weeks is presented in f, and the proportion of L. iners sequences is presented in g, n = 111 women. Immune mediators were analysed and compared between women who had L. iners (CST III) who developed cervical shortening (CL ≤ 25 mm) and women who maintained a CL > 25 mm. Mediators of microbial recognition included MBL (h), IgM (i), IgG1 (j), IgG2 (k), IgG3 (l), IgG4 (m), C3b (n), C5 (o), C5a (p), IL-8 (q), IL-6 (r), IL-1β (s), IL-10 (t), n = 64 samples. Statistical analyses were performed using a one-sided Chi square to compare proportions of VMG 3 between women with a CL ≤ 25 mm and >25 mm (f), and a one-sided Mann–Whitney to compare differences between sequence percentages and immune mediators (g–t), (h–t). Data are presented as median values and interquartile ranges (25th and 75th percentiles). Source data are provided as a Source Data file.

Fig. 8. Braided cervical cerclage material is associated with immune activation and higher rates of preterm labour.

Cervicovaginal immune mediators were analysed from women pre- and post-cervical cerclage. The type of material was known for 39 women, 14 women received braided cerclage, and 25 women received monofilament. The fold change concentration of mediators of microbial recognition (a), complement proteins (b), and cytokine concentrations (c) between samples pre-cerclage and post-cerclage are compared between women who had braided cerclage (n = 14) and women who had monofilament cerclage (n = 25). The percentage of women who delivered preterm with a braided cerclage was compared with the percentage who delivered preterm following a monofilament cerclage (d). Statistical analysis was performed using a one-sided Mann–Whitney test and one-sided Fisher’s exact test. Data are presented as mean and standard deviation. Source data are provided as a Source Data file.

Fig. 9. Cerclage using braided material is associated with activation of complement and cytokine production in women who deliver preterm.

Vaginal microbiota composition (a) and cervicovaginal immune mediators (b–i) were analysed in 34 women who delivered preterm between 12–16 weeks and again between 20–24 weeks. Ten women had a cerclage with braided material and 7 women had a cerclage with monofilament material between timepoints, whereas 16 women had no intervention (a). The concentration of MBL (b), IgM (c), C3b (d), C5 (e), C5a (f), IL-8 (g), IL-6 (h), and IL-1β (i) were compared between 12–16 and 20–24 weeks in women who had a braided cerclage, monofilament cerclage and no cerclage. The proportion of women who had an increase in immune mediators were also compared between suture material. Statistical analysis was performed using a one-sided Wilcoxon matched pairs signed rank test and a one-sided Fisher’s exact test, N = 33 women. Source data are provided as a Source Data file.

Fig. 10. Proposed mechanism for microbial-driven preterm birth.

The complement system is an integral component of the immune response and bridges the adaptive and innate response to pathogens. MBL activates the lectin pathway, whereas IgM and IgG1-3 complexes activate the classical pathway (a). Both pathways converge on the central component C3, which is activated via C3 convertase. Upon activation C3 is cleaved to generate C3b and C3a.C3b binds to C3 convertase to form C5 convertase, which leads to release of C5b and C5a.C3a and the more potent C5a lead to increased vascular permeability and chemoattract phagocytes, and C3b opsonises pathogens and aids phagocytosis via C3b receptors. The alternative pathway also activates a positive feed forward amplification loop via C3b deposition on the surface of pathogens. We propose that in women who deliver preterm, CST III (L. iners) and CST IV (diverse species) activate the complement cascade, which leads to a local pro-inflammatory immune milieu. It is likely that the presence of activated neutrophils leads to an increase in the local concentration of IL-8, IL-6, and IL-1β, prostaglandins (PGs) and matrix metalloproteinases (MMPs). If a dysregulated response occurs, this may lead to cervical shortening and preterm labour (b). In women who deliver at term (c), there is a predominance of CST I (L. crispatus). However, if CST III (L. iners) or CST IV (diverse species) is seen, we propose that a more regulated immune response prevents cervical shortening and the triggering of early parturition. Figure created with BioRender.com.