Title: Relationship between vaginal microbial dysbiosis, inflammation, and pregnancy outcomes in cervical cerclage. One Sentence Summary

1Results

Retrospective assessment of suture materials’ effects on pregnancy outcomes

A total of 671 women receiving cervical cerclage during pregnancy were identified from five UK university hospitals within a 10-year period. Of these, 327 (49%) received a braided suture material and 344 (51%) received monofilament suture for their cervical cerclage. In women receiving a braided cerclage, higher rates of non-viable births (delivery <24 weeks or intrauterine death) were observed compared to those receiving the monofilament alternative (15% vs 5% respectively; P < 0.0001, Fig 1A). Increased rates of preterm birth (24-37 weeks gestation) were also observed in women receiving braided cerclage (28% braided vs 17% monofilament; P < 0.0001, Fisher’s exact test, Fig 1A). Comparison of available demographics demonstrated that consistent with known clinical practice (21, 22), preference of suture material varied across hospital sites (Table S1). Linear mixed effects modeling excluding hospital location, which is linked to choice of suture material, demonstrated that although history of a previous preterm birth was a significant contributor to non-viable births, braided suture was the primary driver of the observed outcome independent of potential confounders including maternal age, ethnicity, and parity (Fig. S1) but not hospital location. The effect of previous preterm birth was lost when hospital location was included in the model (Table S2). In both analyses, suture material was the major variable influencing the risk of preterm birth independent of maternal age, ethnicity, parity, and history of a previous preterm birth.

Data on gestational age at cerclage insertion and corresponding cervical length were available for women receiving an ultrasound-indicated cerclage (for CL ≤25 mm). Distribution of cervical length was comparable among monofilament and braided groups (P = 0.2; Mann-Whitney, Table S1), and as would be expected, a shorter cervix significantly contributed to the risk of adverse outcome (P = 0.019). A sub-analysis using linear mixed effects regression models demonstrated that suture material remained a significant contributing factor for both preterm birth (P = 0.00002) and non-viable pregnancy (P = 0.006) among ultrasound-indicated cerclages after adjusting for potential confounders including gestational age at insertion (Table S2).
Baseline characteristics of the prospective study subjects

Women who were attending prematurity surveillance clinic after a history of preterm birth and were sonographically identified as having a short cervix (≤25 mm) were prospectively recruited and randomized to receive a cerclage using either braided Mersilene (n=25) or monofilament Ethilon (n=24) suture material. Demographics among suture material groups, including gestation and cervical length at cerclage insertion were comparable (Table 1).

Suture material impact on the vaginal microbiome

The data set consisted of 2,792,842 high quality gene sequences, with a mean sequence read count of 13,825 per sample (range 689 to 1,396,421). Using bacterial genera sequence data, samples were classified according to their vaginal bacterial communities as normal (>90% Lactobacillus spp.), intermediate (30-90% Lactobacillus spp.), or dysbiotic (<30% Lactobacillus spp.) (Fig. 1B and Table S3). Before cerclage insertion, prevalence of intermediate and dysbiotic microbiomes in monofilament and braided patient groups was similar (16.7% v 17.3%, P = 0.7, Fig. 1C). These were higher than the rates of dysbiosis observed in the background low-risk pregnant population at the same gestational age not requiring intervention who had normal pregnancy outcomes (6%, Fisher’s exact, P = 0.03) (6).

Insertion of the braided cerclage caused a dramatic shift towards dysbiosis at 4 weeks after the procedure, which persisted until the final follow up timepoint, at 16 weeks after cerclage insertion (P = 0.008, ANOVA, Fig. 1C, Table S4). When assessed at the species level (Fig. S3, Fig. S4, and Tables S4 and S5), braided cerclage insertion was associated with an increasing proportion of women with community state type (CST) IV, characterized by reduced numbers of Lactobacillus spp. and increased diversity of anaerobic bacteria (6, 25). Before the insertion of braided cerclage, 13% were classified as CST IV, increasing to 45% at 4 weeks and 50% at 16 weeks after cerclage (P = 0.02, Fig. S4A, Table S4). In contrast, microbial disruption was not observed in women receiving monofilament cerclage, who instead demonstrated maintenance of high Lactobacillus spp. abundance (CSTs I, II, III, and V) and stability throughout longitudinal sampling (P = 0.9; Fig. S4A and Table S4). To identify degrees of dysbiosis that were not identified by CST classification, we also undertook alternate species level classification based upon dominance of vaginal bacterial communities by Lactobacillus species associated with stability and health, L. iners dominance associated more frequently with transition to dysbiosis, as well as intermediate and severe dysbiosis (26). Although L. iners has previously been observed as an intermediary towards dysbiosis, there was no significant change in L. iners abundance in association with insertion of a monofilament or braided suture material (Fig. S4B and Table S5).
An in vitro adhesion assay showed that braided suture cultured with the vaginal commensal L. jensenii or the pathobiont E. coli resulted in a 16- and 20-fold greater bacterial load adherence per unit length (cm), respectively, compared to monofilament suture (P = 0.03 and P = 0.0003, respectively, Student’s t-test; Fig. S5).
To identify bacteria specifically associated with braided suture insertion, we performed linear discriminant analysis with effect size (LEfSe)(27) on the 16S rRNA sequence data collected before and 4 weeks after the procedure. Although no differences were identified between patient groups before insertion, braided cerclage resulted in enriched numbers of Gram-negative bacteria at 4 weeks (Fig. 2A and B). This correlated with a 5-fold increase in the number of dysbiotic samples collected after braided cerclage insertion compared to monofilament cerclage (P = 0.04, ANOVA; Fig. 2C). Use of braided cerclage was characterized by increased numbers of bacteria associated with bacterial vaginosis, including species of Prevotella (P = 0.02), Finegoldia (P = 0.02), and Dialister (P = 0.04), and reduced Lactobacillus spp. (Fig. 2A, B and Fig. S6). Targeted qPCR suggested that insertion of braided suture was associated with an increase in mean copy numbers of both A. vaginae (594,326 before cerclage v 5,081,000 after cerclage; P = 0.07) and G. vaginalis (961,805 before cerclage v 10,170,000 after cerclage; P = 0.05; Fig. S7, Table S6). In contrast, no change in the amount of G. vaginalis or A. vaginae was detected after monofilament cerclage. Consistent with these observations, indices of bacterial community richness (Fig. 2D) and alpha-diversity (Fig. 2E) were increased in samples collected after braided suture compared to monofilament, with the greatest differences observed at 16 weeks after cerclage (P = 0.02; ANOVA, Bonferroni multiple comparison).
Inflammatory response to cerclage insertion

Insertion of braided, but not monofilament cerclage, increased the release of inflammatory cytokines into cervico-vaginal fluid, including IL-1β, IL-6, IL-8, TNFα, and MMP-1 (Fig. 3A-F). No change was detected in anti-inflammatory cytokines IL-4 (Fig. 3G) IL-2, or IL-10 (Table S7). We observed a strong association between severe dysbiosis (<30% Lactobacillus spp.) and increased cervico-vaginal fluid concentrations of pro-inflammatory cytokines ICAM-1, IL-1β, IL-6, MMP-1, MCP-1, TNF-α, GM-CSF, and IFN-γ, as well as the anti-inflammatory cytokine IL-10 when compared to women harboring Lactobacillus spp. dominated microbiomes (Fig. 3H).

Impact on cervical vasculature after cerclage

We constructed three-dimensional cervical vascular trees from ultrasound data and assessed the indices of cervical vasculature (Vascularity index, VI) (28) for morphological differences in the cervix according to suture material. The vascularity was not significantly different between the two groups before cerclage insertion (Fig. 4A). Braided cerclage was strongly associated with increased cervical vascularization 4 weeks after the procedure, and this relationship persisted until 16 weeks after insertion (P = 0.0003; ANOVA, Bonferroni multiple comparison, Fig. 4A). A positive correlation of cervical vascularity with both the number of bacterial species (Fig. 4B, R² = 0.09, P = 0.002) and alpha diversity (Fig. 4C, R² = 0.14, P = 0.001) was observed in women receiving a braided cerclage. No relationship between monofilament suture, cervical vascularization, and indices of microbial diversity or richness was observed.

Discussion

Braided suture material is commonly used in preference to monofilament for cervical cerclage (21), because it is assumed that braided suture provides a more secure cerclage that is less likely to slip or tear the cervix (18), however this is not evidence based. Here we show that use of braided suture material is associated with an increased risk of preterm birth and non-viable pregnancy, although this will need to be confirmed in a prospective randomized study. Braided cervical cerclage induces vaginal dysbiosis, increases excretion of inflammatory cytokines into the cervico-vaginal fluid, and induces premature cervical vascular remodeling. In contrast, monofilament suture has minimal impact upon the vaginal microbiome and inflammatory pathways associated with premature onset of parturition. These findings have clinical relevance for cerclage procedures in pregnancy and wider implications for braided suture use in other surgical procedures, particularly in potentially contaminated sites. Based upon an approximated two million cervical cerclages per annum (15), 80% of which are performed using braided suture (21), we estimate that a global shift to monofilament suture use for cervical cerclage would prevent 170,000 preterm births (number needed to treat NNT 9.4; 95% CI 5.9 to 22.6) and 172,000 fetal losses (NNT, 9.3; 95% CI 6.6 to 16.0) per annum world-wide.

Although cervical cerclage is effective in preventing preterm birth in singleton pregnancies with a previous preterm birth (17, 29), our data show that braided cervical cerclage increases vaginal dysbiosis and inflammation, and likely accounts for the doubled risk of puerperal sepsis (19) with no improvement in neonatal outcome (17) after cerclage insertion. Moreover, increasing evidence suggests that the braided cerclage is of no benefit, and may even be detrimental in other groups that are at high risk for preterm birth, such as multiple pregnancies (30) and women with a shortened cervix after excisional treatment for cervical intra-epithelial neoplasia (31-34). The role of suture material in cerclage efficacy has largely been neglected, with studies rarely detailing the suture material used for the procedure (35). To date, only one randomized controlled trial has been conducted to examine the impact of suture material on pregnancy outcomes, but this study was limited to the comparison of two braided materials, and no differences in preterm birth rates were observed (22). Re-evaluation of existing literature on cervical cerclage use in pregnancy for the prevention of preterm birth is required in light of our findings.

Our study is limited by the retrospective nature of the comparisons of pregnancy outcome. However, the cohort size is large, and preterm delivery rates for each suture material were similar at each of the five centers. The size of the prospective study was limited by practicalities and cost of intensive multiple investigations; however, the experimental data provide supporting evidence for the mechanism of poorer outcomes in patients treated with braided suture cerclage. Analysis of the vaginal microbiome was undertaken using three alternate approaches. We primarily used a genera based classification of normal, intermediate and severe dysbiosis, which demonstrated that the principal effect of braided cerclage is to reduce lactobacillus numbers and induce vaginal dysbiosis. We next classified samples at species level into 5 previously described CSTs, however, this analysis was limited to the consideration of only one dysbiotic group. Therefore an alternate species level classification considering two levels of dysbiosis (intermediate and severe) was also undertaken.

The inherent capacity of braided suture material to facilitate bacterial growth has been previously described in other surgical arenas (36, 37); however, our in vivo and in vitro data provide evidence for preferential pathobiont colonization over commensal vaginal species that are important for reproductive health outcomes. Vaginal dysbiosis associated with braided cerclage insertion was characterized by reduced Lactobacillus species and increased diversity and enrichment of bacteria associated with poor pregnancy outcomes, including Peptoniphilus harei (38), species of Bacteriodes (38, 39), Prevotella (40), and Clostridium (41). However, it is recognized that sequencing of specific hypervariable regions of the 16S rRNA gene can result in underrepresentation of key vaginal microbiota such as G. vaginalis, which together with A. vaginae is characteristic of bacterial vaginosis (BV) (42, 43), a condition associated with adverse reproductive health outcomes including pelvic inflammatory disease (44), HIV transmission (45), and preterm birth(46). Using targeted qPCR, we showed that dysbiosis associated with braided cerclage insertion may involve increased abundance of G. vaginalis. The virulence of G. vaginalis is thought to relate to its biofilm-producing capacity and adherence to vaginal epithelial cells (42, 47). It is possible that such characteristics may promote biofilm formation on the surface of cerclage suture material, and this should be examined in future studies. Other bacteria clinically associated with adverse pregnancy outcome include S. agalactiae (Group B streptococcus) and E. coli. However, we did not observe any changes in the amounts of S. agalactiae. E. coli was not detected in our data set, but this may reflect a limitation of the primer set used for 16S rRNA sequencing (48, 49).

Increased bacterial diversity in the vagina corresponded to the induction of a pro-inflammatory cytokine profile in cervico-vaginal fluid involving known mediators of cervical vascular remodeling, including IL-1β, IL-6, IL-8, and TNFα (50-52). Concentrations of MMP-1, a matrix metalloproteinase central to collagenous remodeling preceding preterm birth (53, 54), were also increased after insertion of braided cerclage. Increased cervical vascularity occurs before term parturition (28, 55), and an association between increased cytokine excretion and cervical angiogenesis, vasodilation, and vascular permeability has been previously described (56). Consistent with a role in untimely cervical remodeling preceding preterm birth, increased cervical vascularity was observed as early as 4 weeks after insertion in women receiving a braided cerclage. Our study therefore provides a human model for understanding how pregnant host-vaginal microbial interactions may underpin poor pregnancy outcomes. Cerclage-induced inflammation resulting in premature weakening of the cervix could also provide a mechanism for high rates of intrauterine death in women receiving a braided cerclage, because in-utero exposure of the fetus to elevated concentrations of circulating pro-inflammatory cytokines is known to associate with fetal brain injury (57, 58) and stillbirth (59-61).

In summary, our data provide evidence that cervical cerclage using braided suture associates with increased rates of preterm birth and non-viable pregnancy. Promotion of vaginal bacterial dysbiosis after insertion of braided suture material likely contributes to these adverse pregnancy outcomes through activation of local tissue inflammation and premature cervical remodeling. Because monofilament suture has minimal impact on the host microbiome or inflammation in pregnancy and associates with improved pregnancy outcome, we advocate its use for cervical cerclage. Further clinical trials addressing the impact of cerclage suture material, powered to assess outcomes of preterm birth, neonatal morbidity, and mortality, are therefore urgently required.

Materials and Methods

The study was approved by NHS National Research Ethics Service (NRES) Committee London - City & East (REC 12/LO/2003), and all participants provided written informed consent.

Study design

We initially performed a retrospective data collection to assess outcomes of cervical cerclages in singleton pregnancies considered at risk of preterm birth over a ten-year period between January 2003 and 2013 across five UK hospitals in London, Cambridge, and Birmingham. Cases were identified from operating theatre logs, and all case notes were reviewed where possible. Details regarding cerclage suture insertion, suture material used, outcomes of preterm birth (between 16⁺⁰ and 36⁺⁶ weeks’ gestation), and non-viable birth (still birth or miscarriage >16⁺⁰ weeks’ gestation) were collected. Other metadata collected included maternal age, parity, previous spontaneous preterm birth / midtrimester miscarriage, indication for cerclage (ultrasound indicated or elective), and cervical length at cerclage insertion.

After this, we prospectively recruited pregnant women at risk of preterm birth with sonographic indications for cervical cerclage at preterm surveillance clinics from January 2013 until August 2014 at a single London site (Queen Charlotte’s and Chelsea Hospital). Inclusion criteria were pregnant women with history of spontaneous preterm birth (<37⁺⁰ weeks) and a cervical length (CL) measurement below the 10^th centile (≤25 mm) on transvaginal scan at ≤ 23⁺⁶ weeks’ gestation in the index pregnancy. A normally distributed cervical length range not associated with preterm birth risk is typically 35 mm ± 8.3 mm (mean ± SD) (62).

Exclusion criteria included multiple pregnancy, previous iatrogenic preterm births, HIV positive status, and sexual intercourse or vaginal bleeding in the preceding 48 hours. Eligible women were randomized to either braided Mersilene (n=25) or monofilament Ethilon (n=24) cerclage suture material. The same obstetrician performed the procedure using the MacDonald technique (63). Participants were recruited before cerclage insertion and followed up longitudinally at 4, 8, 12, and 16 weeks after insertion. At each time point, cervico-vaginal fluid was sampled from the posterior fornix under direct visualization using 2 swabs for later 16s rRNA gene sequencing and cytokine analysis: a BBL CultureSwab MaxV Liquid Amies swab (Becton, Dickinson and Company) and a Transwab MW170 with rayon bud type (Medical Wire & Equipment), respectively. Both swabs were immediately placed on ice and snap frozen at -80°C. A transvaginal ultrasound scan was then immediately performed to assess cervical vascularization in the dorsal lithotomy position with an empty bladder, taking care to avoid undue pressure on the cervix.