Background
A short cervix is a risk factor for preterm birth. The molecular drivers of a short cervix remain elusive. Metabolites may function as mediators of pathologic processes.
Objective
We sought to determine if a distinct cervicovaginal metabolomic profile is associated with a short cervix (<25 mm) to unveil the potential mechanisms by which premature cervical remodeling leads to a short cervix.
Study Design
This was a secondary analysis of a completed prospective pregnancy cohort. Cervicovaginal fluid was obtained between 20 and 24 weeks’ gestation. The participants selected for metabolomic profiling were frequency-matched by birth outcome and cervicovaginal microbiota profile. This analysis included 222 participants with cervical length measured. A short cervix was defined as one having length <25 mm, as measured by transvaginal ultrasound. Unpaired t-tests were performed with a Bonferroni correction for multiple comparisons.
Results
There were 27 participants with a short cervix, and 195 with normal cervical length. Of the 637 metabolites detected, 26 differed between those with a short cervix and those with normal cervical lengths; 22 were decreased, of which 21 belonged to the lipid metabolism pathway (all P <.000079). Diethanolamine, erythritol, progesterone, and mannitol or sorbitol were increased in the cases of short cervix. Among participants with Lactobacillus -deficient microbiota, only diethanolamine and mannitol or sorbitol differed between short cervix (n=17) and normal cervical length (n=75), both increased.
Conclusion
A short cervix is associated with decreased cervicovaginal lipid metabolites, particularly sphingolipids. This class of lipids stabilizes cell membranes and protects against environmental exposures. Increased diethanolamine—an immunostimulatory xenobiotic—is associated with a short cervix. These observations begin to identify the potential mechanisms by which modifiable environmental factors may invoke cell damage in the setting of biological vulnerability, thus promoting premature cervical remodeling in spontaneous preterm birth.
Introduction
Lactobacillus -deficient cervicovaginal microbial communities are associated with a short cervix and spontaneous preterm birth (sPTB). A short cervix serves as a sonographic proxy for premature cervical remodeling, which is believed to be a common biological precursor of sPTB. Although cervical length screening is utilized clinically to identify individuals at a high risk of delivering preterm, the mechanisms underlying these observed associations and pathogenic processes remain to be elucidated.
Why was this study conducted?
Lactobacillus -deficient cervicovaginal microbiota are associated with a short cervix, which is a risk factor for preterm birth. However, the molecular drivers of this sonographic finding remain elusive. The characterization of biochemical footprints of microbiota through metabolomics has illuminated host–microbial interactions in other systems. This study sought to determine whether select cervicovaginal metabolites are associated with a short cervix to unveil the potential mechanisms by which premature cervical remodeling leads to a short cervix.
Key findings
A short cervix is associated with decreased cervicovaginal lipid metabolites, particularly sphingolipids, which are implicated in cell membrane stabilization and protection against environmental exposures. A short cervix is also associated with increased cervicovaginal xenobiotics, including the immunostimulatory metabolite diethanolamine.
What does this add to what is known?
These findings identify the potential mechanisms by which modifiable environmental factors may invoke cell damage in the setting of biological vulnerability, thus promoting premature cervical remodeling in preterm birth.
In other organ systems, characterization of the biochemical footprints of microbial ecosystems has illuminated host–microbial interactions. Microbiota exert a large part of their effect through the production of metabolites, which can impact the inflammatory tone, cell signaling, and epithelial barriers. In the obstetrical realm, cervicovaginal epithelial barrier disruption and induction of select immune responses have been implicated in cervical remodeling. Our laboratory has demonstrated that the microbial output from microbes common in Lactobacillus -deficient communities induces epithelial barrier disruption and proinflammatory cytokines in cervicovaginal epithelial cell lines. , Observations from our clinical pregnancy cohorts corroborate these in vitro findings. These collective data provide compelling evidence that microbial output and interactions at the host–microbial interface contribute to the molecular underpinnings of cervical remodeling, leading to a short cervix.
Our understanding of the cervicovaginal metabolome and its role in reproductive health is evolving. Our group and others have demonstrated that select cervicovaginal metabolites and metabolomic profiles are associated with features of vaginal ecosystems, specific host immune response, and sPTB. It remains unknown, however, whether these small molecules are bioactive and whether they function as drivers of cervical remodeling that results in a short cervix on the pathway to sPTB. To begin unraveling these various molecular inputs, we examined the cervicovaginal metabolome in pregnancy with the goal of determining whether select cervicovaginal metabolites were associated with a second-trimester short cervix.
Materials and Methods
Study setting
This was a secondary analysis from a prospective pregnancy cohort study called Motherhood and Microbiome (M&M). A flow diagram of study participants is presented in Figure . For M&M, 2000 pregnant individuals enrolled from December 2013 through February 2017. This study was approved by the institutional review board at the University of Pennsylvania (IRB #818914) on October 23, 2013. The M&M methods have been previously published. In brief, individuals receiving prenatal care at the Hospital of the University of Pennsylvania enrolled after informed consent before 20 weeks’ gestation. The exclusion criteria included major fetal anomalies, HIV seropositive status, history of organ transplant, chronic steroid use, enrollment into the study during a previous pregnancy, or multiple gestations. The participants were followed to delivery. Cases of preterm birth (PTB) were adjudicated by a maternal-fetal medicine physician (M.A.E.) to distinguish sPTB from medically indicated PTB (mPTB). PTB was considered sPTB if a woman presented with either cervical dilation and/or premature rupture of membranes and delivered before 37 weeks of gestation. Cervicovaginal microbiota profiling was done at multiple time points. This secondary analysis includes profiling from samples collected between 20 and 24 weeks’ gestation (second hypervariable regions [V2], mean gestational age 21.7 weeks [standard deviation (SD) 1.4]) for 612 participants. The microbiota were analyzed by 16S ribosomal RNA (rRNA) gene sequencing via amplification of the third and fourth hypervariable regions (V3–V4) regions of the 16S rRNA gene. Microbial communities were classified into community state types (CSTs) as previously reported. Classifications were assigned to each sample using hierarchical clustering with Jensen-Shannon divergence and Ward linkage. CST-I is predominated with L. crispatus , CST-II with L. gasseri , CST-III with L. Iners, and CST V with L. jensenii . CST-IV is defined by a paucity of Lactobacillus species and the presence of a diverse set of strict and facultative anaerobes. The Lactobacillus -dominant communities included CST-I, CST-II, CST-III, and CST V, whereas the Lactobacillus -deficient communities included CST-IV. Metabolomic profiling of V2 cervicovaginal swabs was performed as described below.
The participants were matched by birth outcome (sPTB n=80, mPTB n=40 and term birth ≥38 weeks’ gestation n=153) and cervicovaginal microbiota CST (CST-IV [n=112] vs all other CSTs [n=161]). This present analysis was then restricted to participants who also had a cervical length measurement obtained in the second trimester (mean gestational age, 20.2 weeks [SD, 0.8]) by trained sonographers as part of routine clinical care (n=222). The cervical length was measured by transvaginal ultrasound and dichotomized at 25 mm. A short cervix was defined as one with length <25 mm. Participants with a normal cervical length (ie, ≥25 mm) were used as the reference group in this analysis.
Biospecimen collection
Specimens were self-collected by the participant or collected by a research coordinator during a clinical exam. These included an ESwabs (Copan Diagnostics, Murrieta, CA) stored in 1 mL of Amies Transport Medium and a Dacron swab stored without buffer. All the samples were immediately frozen at −80°C until processing. Microbiome analysis was performed as previously reported and described above.
Nontargeted global metabolite profiling
Sample preparation and analysis were carried out as described previously at Metabolon, Inc. In brief, sample preparation involved protein precipitation and removal with methanol, shaking, and centrifugation. The resulting extracts were profiled on an accurate mass global metabolomics platform consisting of multiple arms differing by chromatography methods and mass spectrometry ionization modes to achieve a broad coverage of compounds differing by physiochemical properties such as mass, charge, chromatographic separation, and ionization behavior. The details of this platform have been described previously. , Metabolites were identified by automated comparison of the ion features in the experimental samples to a reference library of chemical standard entries that included retention time, molecular weight ( m/z ), preferred adducts, and in-source fragments and also the associated mass spectrometry (MS) spectra; they were curated by visual inspection for quality control using software developed at Metabolon. , Metabolomics data have been deposited to the European Molecular Biology Laboratory-European Bioinformatics Institute (EMBL-EBI) MetaboLights database ( https://doi.org/10.1093/nar/gkz1019 , PMID: 31691833 ) with the identifier MTBLS4650. The complete dataset can be accessed here https://www.ebi.ac.uk/metabolights/MTBLS4650 .
Statistical analyses
Bivariate comparisons of maternal characteristics between cases of short cervix and normal cervical length controls were performed using chi-squared (χ 2 ) or student’s t-tests as appropriate. Metabolite analyses were conducted on log 2 -transformed data normalized to the volume available or utilized for extraction. For metabolites with levels below the limit of detection (LOD), the LOD divided by the square root of 2 was assigned. Student t tests were used to compare metabolite abundance between short cervix cases and normal cervical length controls. We performed 2 sensitivity analyses to ensure that our findings were robust. First, we imputed the lowest detectable value for metabolites below the LOD. Second, we restricted analyses to metabolites detected in at least half of participants.
To address the potential that normalization of metabolite abundance to the volume available or utilized for extraction may introduce significant sample variance, we also repeated analyses using a ratiometric mean control. Specifically, 3 metabolites present in all individuals with the lowest SD were identified and used as reference metabolites. A normalization factor was calculated for each participant by taking the geometric mean of the 3 reference metabolites. Each metabolite value for each participant was then divided by their normalization factor and the data were subsequently log 2 transformed. The metabolites were then compared between cases of short cervix and normal cervical length controls, generating a list of significant metabolites as presented in Supplemental Table 1 .
Several secondary analyses were conducted. First, we restricted the analysis to individuals with a Lactobacillus -deficient microbial community. Next, we restricted it to individuals with a short cervix and compared patients with sPTB <34 weeks’ gestation to those with sPTB ≥34 weeks’ gestation. Finally, given that a short cervix prompted clinical use of vaginal progesterone, we compared the metabolic profiles among participants using vaginal progesterone for a short cervix to participants not exposed, excluding the one participant using vaginal progesterone in the absence of a short cervix. Then, among participants using vaginal progesterone for short cervix, we compared the metabolite abundance between those with sPTB <34 weeks compared those with gestations ≥34 weeks.
Fold change was calculated as the difference in log 2 -transformed abundance between groups for each metabolite. Only the metabolites that met the Bonferroni threshold P value of ≤.05/(n metabolites compared) were considered significant. Statistical analyses were done using SAS version 9.4 (SAS Institute Inc, Cary, NC) and R version 4.1.1 (R Core Team, Vienna, Austria).
Results
Demographic characteristics of participants
A total of 222 individuals were included in this study, of whom 27 (12.2%) had a short cervix and 195 (87.8%) had a normal cervical length ( Table 1 ). Characteristics with respect to age, body mass index (BMI), and race or ethnicity were similar between the groups, with most participants self-identifying as non-Hispanic Black. Consistent with our previous work, a Lactobacillus -deficient cervicovaginal microbiota was more prevalent in cases of short cervix. The frequency of vaginal intercourse and douching in the 24 hours preceding swab collection was similar between the groups. The mean difference in time between cervical length screening to swab collection was 1.9 weeks (SD, 1.5) for individuals with a short cervix and 1.5 weeks (SD, 1.6) for those with normal cervical length ( P =.64).
| Characteristic | Normal cervical length (n=195) | Short cervical length (n=27) | P value |
|---|---|---|---|
| mean (SD) | |||
| Age (y) | 28.6 (5.88) | 27.0 (5.18) | .15 |
| Prepregnancy BMI (kg/m 2 ) | 29.5 (8.18) | 30.1 (8.01) | .70 |
| Gestational age at cervical length screening (wk) | 20.2 (0.764) | 19.9 (0.818) | .045 |
| n (column %) | |||
| Obstetrical history | .0055 | ||
| Multiparous with previous sPTB | 26 (13.3) | 9 (33.3) | |
| Multiparous without previous sPTB | 89 (45.6) | 5 (18.5) | |
| Nulliparous | 80 (41.0) | 13 (48.1) | |
| Previous cervical excision surgery | 7 (3.6) | 0 (0) | .68 |
| Race | .66 | ||
| Asian | 5 (2.6) | 1 (3.7) | |
| Black | 140 (71.8) | 22 (81.5) | |
| Other | 1 (0.5) | 0 (0) | |
| White | 49 (25.1) | 4 (14.8) | |
| Hispanic | 10 (5.1) | 1 (3.7) | 1 |
| Lactobacillus -deficient vaginal microbiota | 75 (38.5) | 17 (63.0) | .027 |
| Vaginal progesterone use | 1 (0.5) | 12 (44.4) | <.001 |
| Vaginal douching in previous 24 h | 4 (2.1) | 1 (3.7) | 1 |
| Intercourse in previous 24 h | 43 (22.1) | 4 (14.8) | .54 |
Short cervix and metabolite analyses
A total of 637 metabolites were detected in at least 1 participant, 608 metabolites were detected in at least 10% of the participants, and 530 metabolites were detected in at least 50% of the participants. In the primary analysis, 26 metabolites differed between cases of short cervix and controls with normal cervical length ( Table 2 ). Among differentially detected metabolites, 22 were decreased in abundance among the cases. Notably, all but one of these metabolites (tartronate) belonged to the lipid metabolism pathway (all P <.000079). Consistent with biochemical nomenclature, lipid metabolites are presented when appropriate using the family name (eg, sphingomyelin) followed parenthetically by the common backbone then fatty acid chain length and saturation (eg, d18:1/18:0). Subpathway categories within the lipid pathway included sphingosines, sphingomyelins, dihydrosphingomyelins, sphingolipid synthesis, hexosylceramines, ceramides, dihydroceramines, fatty acid and dicarboxylate, plasmalogen, phosphatidylcholine, and sterols. Notably, 15 of the 22 decreased lipid metabolites are sphingolipids or sphingolipid precursors. The 4 metabolites that were increased in abundance among cases of short cervix include 2 xenobiotics, viz, diethanolamine and erythritol, and progesterone and mannitol or sorbitol. Progesterone elevation in cases was consistent with the clinical practice of prescribing vaginal progesterone for individuals with a short cervix. The results were similar in the 2 sensitivity analyses. In the analysis that imputed the lowest detectable values, only sphingomyelin (d18:1/14:0, d16:1/16:0) no longer differed significantly, whereas the other 25 metabolites remained significant. In the analysis restricted to metabolites present in at least 50% of the participants, only sphingomyelin (d18:1/14:0, d16:1/16:0) and progesterone no longer differed significantly, whereas the other 23 metabolites remained significantly different.
| Super pathway | Subpathway | Biochemical | Fold change | P value a |
|---|---|---|---|---|
| Lipid | Dihydroceramides | N-palmitoyl-sphinganine (d18:0/16:0) | −1.89 | .000000011 |
| Lipid | Ceramides | N-palmitoyl-sphingosine (d18:1/16:0) | −1.86 | .00000011 |
| Lipid | Hexosylceramides (HCER) | glycosyl-N-palmitoyl-sphingosine (d18:1/16:0) | −1.80 | .0000073 |
| Lipid | Sphingosines | heptadecasphingosine (d17:1) | −1.78 | .00000013 |
| Lipid | Sphingosines | sphingosine | −1.77 | .0000015 |
| Lipid | Sphingomyelins | lignoceroyl sphingomyelin (d18:1/24:0) | −1.64 | .000000038 |
| Lipid | Sphingomyelins | stearoyl sphingomyelin (d18:1/18:0) | −1.63 | .00000029 |
| Lipid | Sterol | cholesterol sulfate | −1.63 | .000000012 |
| Lipid | Plasmalogen | 1-(1-enyl-palmitoyl)-2-linoleoyl-GPE (P-16:0/18:2) | −1.59 | .0000045 |
| Lipid | Sphingolipid synthesis | phytosphingosine | −1.58 | .000000098 |
| Lipid | Sphingolipid synthesis | sphinganine | −1.56 | .0000012 |
| Lipid | Sphingomyelins | sphingomyelin (d18:1/20:0, d16:1/22:0) | −1.55 | .0000054 |
| Lipid | Sphingomyelins | behenoyl sphingomyelin (d18:1/22:0) | −1.54 | .000020 |
| Lipid | Dihydrosphingomyelins | palmitoyl dihydrosphingomyelin (d18:0/16:0) | −1.51 | .000000042 |
| Lipid | Sphingomyelins | palmitoyl sphingomyelin (d18:1/16:0) | −1.38 | .00000018 |
| Lipid | Phosphatidylcholine (PC) | 1-stearoyl-2-oleoyl-GPC (18:0/18:1) | −1.37 | .000014 |
| Lipid | Plasmalogen | 1-(1-enyl-palmitoyl)-2-oleoyl-GPE (P-16:0/18:1) | −1.31 | .0000085 |
| Xenobiotics | Food component/plant | tartronate (hydroxymalonate) | −1.21 | .000072 |
| Lipid | Sphingomyelins | sphingomyelin (d18:1/14:0, d16:1/16:0) | −1.18 | .000070 |
| Lipid | Sphingosines | eicosanoylsphingosine (d20:1) | −1.17 | .0000018 |
| Lipid | Fatty Acid, dicarboxylate | undecanedioate (C11-DC) | −0.45 | .0000010 |
| Lipid | Fatty Acid, dicarboxylate | sebacate (C10-DC) | −0.37 | .000011 |
| Carbohydrate | fructose, mannose, and galactose metabolism | mannitol/sorbitol | 0.70 | <.0000000001 |
| Lipid | Progestin steroids | progesterone | 0.83 | .000000058 |
| Xenobiotics | Food component/plant | erythritol | 0.94 | .000057 |
| Xenobiotics | chemical | diethanolamine | 1.08 | .0000000037 |
a All presented P values were significant at the Bonferroni threshold for multiple comparisons of P <7.85E-5 (0.05/637).
Subanalyses were performed excluding all individuals who used progesterone for this indication, to address the possibility that vaginal or intramuscular progesterone used for sPTB risk reduction contributed to the observed associations. Similar results were identified in terms of the top differentially detected metabolites when comparing short cervix cases to normal cervical length controls. Few metabolites remained significant after applying the Bonferonni threshold, presumably because of the small number of short cervix cases (n=8) ( Supplemental Table 2 ).
Analyses among individuals with a Lactobacillus-deficient microbiota
We next restricted the analysis to those individuals colonized by a Lactobacillus -deficient cervicovaginal microbiota, known to be associated with both short cervix and sPTB (CST-IV). This secondary analysis comprised 92 individuals, of whom 17 (18.5%) had a short cervix and 75 (81.5%) had a normal cervical length. The demographic characteristics were largely balanced ( Supplemental Table 3 ). Table 3 shows the top 10 metabolites that differed between the 2 groups; only diethanolamine and mannitol or sorbitol were significantly higher in participants with a short cervix ( P < .000080).
| Super pathway | Subpathway | Biochemical | Fold change | P value |
|---|---|---|---|---|
| Carbohydrate | Fructose, mannose, and galactose metabolism | mannitol or sorbitol | 0.79 | .000000019 a |
| Xenobiotics | Chemical | diethanolamine | 1.31 | .00000051 a |
| Lipid | Fatty acid synthesis | malonate | −0.40 | .00018 |
| Lipid | Progestin steroids | progesterone | 0.71 | .00018 |
| Lipid | Fatty acid, dicarboxylate | undecanedioate (C11-DC) | −0.43 | .00029 |
| Lipid | Sterol | cholesterol sulfate | −1.35 | .00036 |
| Xenobiotics | Food component/plant | erythritol | 1.07 | .00038 |
| Carbohydrate | Pentose metabolism | arabinose | 0.95 | .00054 |
| Lipid | Fatty acid, dicarboxylate | sebacate (C10-DC) | −0.38 | .0011 |
| Lipid | Sphingolipid synthesis | phytosphingosine | −1.39 | .0013 |
a P value was significant at the Bonferroni threshold for multiple comparisons of P <.000080 (0.05/627).
Analyses among individuals with a short cervix
To determine whether metabolites could distinguish the birth outcomes among individuals with a short cervix, we restricted the analysis to the 27 cases of short cervix. There were 13 cases of sPTB <34 weeks and 14 participants with gestations ≥ 34 weeks. The demographic characteristics were largely balanced ( Supplemental Table 4 ). No metabolites differed between the groups ( Supplemental Table 5 ).
Analyses among individuals using vaginal progesterone
Among the 27 cases of short cervix, there were 12 individuals using vaginal progesterone. We examined the metabolomic profiles among these 12 individuals to determine whether any metabolites were associated with progesterone use or efficacy with respect to sPTB risk reduction. Among these 12 individuals, 6 had sPTB <34 weeks and 6 had gestations ≥34 weeks. The demographic characteristics were largely balanced ( Supplemental Table 6 ). Table 4 shows the top 10 metabolites associated with the use of vaginal progesterone; 7 metabolites in this group of patients differed significantly compared with those in participants without vaginal progesterone exposure, including several drugs ( P <.00008). Among the participants who used vaginal progesterone for a short cervix, no metabolites were associated with sPTB <34 weeks ( Supplemental Table 7 ).
| Super pathway | Subpathway | Biochemical | Fold change | P value |
|---|---|---|---|---|
| Lipid | Progestin steroids | progesterone | 1.91 | <.0000000001 a |
| Xenobiotics | Drug: psychoactive | citalopram N-oxide | 0.04 | .000021 a |
| Xenobiotics | Drug: psychoactive | citalopram propionate | 0.04 | .000021 a |
| Xenobiotics | Drug: psychoactive | citalopram or escitalopram | 0.04 | .000021 a |
| Xenobiotics | Drug: psychoactive | desmethylcitalopram | 0.04 | .000021 a |
| Xenobiotics | Drug: metabolic | metformin | 0.04 | .000021 a |
| Xenobiotics | Chemical | diethanolamine | 1.09 | .000055 a |
| Xenobiotics | Food component/plant | tartronate (hydroxymalonate) | −1.66 | .00018 |
| Carbohydrate | fructose, mannose, and galactose metabolism | mannitol/sorbitol | 0.54 | .00023 |
| Xenobiotics | Food component/plant | erythritol | 1.11 | .0011 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
