Pregnancy-associated changes in cervical noncoding RNA

Abstract

Aim:

To identify pregnancy-associated changes in cervical noncoding RNA (ncRNA), including miRNA and long noncoding RNA (lncRNA), and their potential effects on biologic processes.

Materials & methods:

We enrolled 21 pregnant women with term deliveries (≥37 weeks’ gestation) in a prospective cohort and collected cervical swabs before 28 weeks’ gestation. We enrolled 21 nonpregnant controls. We analyzed miRNA, lncRNA and mRNA expression, applying a Bonferroni correction.

Results:

Five miRNA and three lncRNA were significantly differentially (>twofold change) expressed. Putative miRNA targets are enriched in genes mediating organogenesis, glucocorticoid signaling, cell adhesion and ncRNA machinery.

Conclusion:

Differential cervical ncRNA expression occurs in the setting of pregnancy. Gene ontology classification reveals biological pathways through which miRNA may play a biologic role in normal pregnancy physiology.

Regulatory noncoding RNA (ncRNA) are functional molecules that regulate gene expression through transcriptional and translational mechanisms. Both short and long classes of ncRNA can modulate epigenetic processes including DNA methylation, histone modification, chromatin remodeling and gene silencing [1]. miRNA are short ncRNA sequences of less than 30 nucleotides that typically function by binding target mRNA to induce degradation or prevent protein translation. Long noncoding RNA (lncRNA) are greater than 200 nucleotides in length and commonly couple with chromatin proteins to influence transcription. Both miRNA and lncRNA have been implicated in myriad disease states, including tumorigenesis and inflammatory conditions, such as autoimmune disorders, infection and cardiovascular disease [2–6]. ncRNA may function as biomarkers for adverse pregnancy outcomes, including spontaneous abortion, preeclampsia, fetal growth restriction and preterm birth [7–9].

Characterizing pregnancy-associated changes in local expression of regulatory molecules, such as ncRNA in reproductive tissues, is essential; molecules that are dynamic throughout gestation likely are vulnerable to environmental factors and may drive pathologic events in pregnancy. Gaining insight into these processes will enhance understanding of pregnancy complications, including cervical insufficiency and preterm labor. The cervix has been well studied as a clinical predictor of preterm birth [10,11], though the biology underlying normal parturition remains poorly understood. Cervical remodeling is required for vaginal delivery and has been described at a molecular level in animal models [12–18]. Elovitz et al. first described a cervical miRNA signature associated with cervical remodeling and preterm birth [19]. A subsequent study identified an association between expression of select cervical miRNA and length of gestation [20]. Others have implicated specific cervical miRNA in the regulation of progesterone receptor activity and onset of labor [21–23]. Distinct miRNA have been implicated in disruption of the cervical epithelial barrier [24]. Recent work suggests that environmental factors, including cervicovaginal bacterial burden and toxic metal exposure, may alter miRNA expression profiles and confer increased risk of preterm delivery [25,26]. Despite literature exploring the role of cervical miRNA in pregnancy, little is known about the role of the lncRNA class of ncRNA in this capacity.

Data remain limited examining ncRNA expression profiles in the cervix outside of pregnancy and cancer. Given the dynamic nature of the cervix, gaining insight into the mechanisms underlying its response to normal pregnancy is of great clinical relevance and may facilitate early recognition of pathologic processes leading to adverse perinatal outcomes. The objective of this study was to identify pregnancy-associated changes in cervical ncRNA, as well as their potential associations with common functional biologic processes. We specifically sought to identify differences in miRNA and lncRNA expression from cervical samples among pregnant women who went on to have uncomplicated term deliveries compared with matched nonpregnant women. We also examined biological processes through which these miRNA potentially could mediate cervical response to pregnancy by using functional clustering of predicted mRNA targets. We simultaneously measured mRNA expression to confirm predictions from bioinformatics approaches.

Materials & methods

Participant selection

Spontaneous Prematurity and Epigenetics of the Cervix (SPEC) is a prospective pregnancy cohort study at Beth Israel Deaconess Medical Center (BIDMC) in MA, USA [27]. Pregnant women were eligible to enroll if they were at least 18 years old, less than 28 weeks of gestation and receiving routine prenatal care at BIDMC. Nonpregnant women were eligible if they were 18–45 years old and receiving routine gynecology care at BIDMC. After giving written informed consent, participants completed questionnaires that collected relevant data on demographics and medical and obstetric history.

Given that the primary aims of SPEC involve a matched, nested case–control design of spontaneous preterm birth, used samples from pregnant SPEC participants with characteristics that were over-represented in the study population. These women ranged in age from 32 to 39 years and were non-Hispanic white. We enrolled 30 nonpregnant women within ±3 years of age who were of the same race/ethnicity and parity as the selected pregnant participants. There were 21 pregnant and 21 nonpregnant participants with sufficient RNA quantities for miRNA analyses and 22 pregnant and 21 nonpregnant participants with sufficient RNA for mRNA and lncRNA analyses. This study was approved by the Institutional Review Board at BIDMC (Protocol 2013P-000125).

Cervical RNA analysis

Under direct visualization during speculum examination, a provider cleared cervical mucous, then collected two swabs by twirling a Dacron^® swab within the endocervix five-times for each swab. One swab was used for RNA and the other for visual inspection of cell-proportions on a histopathologic slide. Median gestational age at the time of cervical swab collection among pregnant participants was 19.1 weeks (interquartile range [IQR] 18.0–19.4, minimum 6.7, maximum 20.4). Samples were stored in RNAlater (Qiagen, MD, USA), placed in a cooler, and then frozen at –80°C within 4 h of collection. RNA was extracted using the Exiqon miCURY RNA Isolation Kit. Laboratory personnel were blinded with respect to pregnancy status. To assess RNA quality, we analyzed nine samples using the Agilent RNA 6000 Nano Kit following manufacturer’s protocol (Agilent 2100 Bioanalyzer, Agilent Technologies, CA, USA).

TaqMan OpenArray^® technology (Life Technologies, CA, USA) was used to measure the expression profile of 754 miRNAs (miRBase v14). RNA was reverse transcribed and preamplified per protocol using Megaplex™ Reverse Transcription Primers, Human Pool A v2.1, Human Pool B v3.0 and Megaplex™ PreAmp Primers, Human Pool A v2.1 and Human Pool B v3.0. Quantitative polymerase chain reaction was performed on the QuantStudioTM 12K Flex Real-Time PCR System (Life Technologies). Cycle threshold (Ct) values were defined as the PCR cycle at which the amplification curve exceeded the background fluorescence threshold line.

Samples from 22 pregnant and 21 nonpregnant participants, as well as six sets of blinded duplicates, were sent to Arraystar (MD, USA) for analysis of 40,173 lncRNA and 20,730 mRNA. Agilent Array platform was utilized. The sample preparation and microarray hybridization were performed per manufacturer protocol. Samples were amplified and transcribed into fluorescent cRNA along the entire length of the transcripts without 3′ bias utilizing a random priming method (Arraystar Flash RNA Labeling Kit, Arraystar). The labeled cRNA were hybridized onto the Human LncRNA Array v4.0 (8 × 60K, Arraystar). The arrays were scanned by the Agilent Scanner G2505C after slides were washed. Agilent Feature Extraction software (version 11.0.1.1) was used for analysis.

Histopathologic slide assessment for leukocyte burden

Given that cell type can affect RNA expression, a pathologist stained smears of the cervical swab using a modified hematoxylin and eosin staining protocol. We defined leukocyte burden as high (versus low) when leukocytes were so numerous as to obscure the epithelial cells throughout the slide, or in areas of cervical mucus, if there were at least three fields at 10X magnification (10x objective) with greater than two clusters of 15 leukocytes. The majority of leukocytes were neutrophils based on morphology. There were no sperm visualized on the slides.

Data processing

For miRNA, we identified three normalization probes by calculating the standard deviation for the 198 probes that were expressed in at least 50% of the cohort. The three probes that were expressed in the most subjects and had the lowest standard deviations were miR-597, miR-210 and miR-197. We calculated the geometric mean for each of these normalization factors among the participants who had each probe expressed. If a participant did not have one of these three probes expressed, we imputed the study population geometric mean for that probe. We then calculated the geometric mean for these three probes among each participant. To calculate the ΔCT for each participant’s expression of a probe, we subtracted the geometric mean CT value from the CT value for each of the probes they expressed.

To assess the accuracy of the miRNA extraction, the analyses were run twice on five participants. The samples were run in separate batches. After visual inspection and given the high correlations (Pearson correlations ranged from 0.94 to 0.99 between replicates [Supplementary Figure 1]), we averaged the two values for participants who had samples run twice.

For lncRNA and mRNA data processing, quantile normalization and subsequent data processing were performed using GeneSpring GX v12.1 software package (Agilent Technologies). RNA with low intensities in all samples were filtered out. While correlations between blinded technical replicates were high (Pearson correlations ranged from 0.92 to 0.94), as a final step and to maximize signal to noise ratio, we filtered out RNA with more than a twofold difference between technical replicates (Supplementary Figure 1). There were 17,889 lncRNAs and 9600 mRNAs retained for downstream analyses.

Statistical analysis

Our primary analysis was to determine associations between miRNA expression and pregnancy status. We used linear regression to calculate the difference in mean ΔCT between pregnant and nonpregnant participants, applying the Bonferroni correction for multiple comparisons (0.05/195, p < 2.6E4) to identify statistically significant miRNA. We then calculated the fold change in miRNA expression between pregnant and nonpregnant participants for each of these probes using the 2^(-ΔΔCt) [28]. To assess whether histopathological presence of leukocytes may have confounded the associations between pregnancy status and probe expression, we used linear regression to calculate difference in expression between pregnant and nonpregnant participants, adjusting for the presence of leukocytes. We assigned the most common profile (low leukocyte burden) for the one nonpregnant woman without a histopathologic slide available.

We obtained lists of predicted targets of miRNA from publicly available algorithms TargetScan Human Release 7.1 (www.targetscan.org) [29]. Putative targets were queried using Panther Gene Ontology (GO) classification database v13.1 [30–32] available through the Gene Ontology project (www.geneontology.org) to identify enrichment of genes in GO biological processes.

As a secondary analysis focused on lncRNA and mRNA, we also performed linear regression to determine differences in expression of lncRNA and mRNA by pregnancy status and subsequently adjusted for leukocyte burden. We considered lncRNA and mRNA differentially expressed if there were greater than twofold change with Bonferroni significance of p <(0.05/17,889) or 2.8e-6 for lncRNA and p < 0.05/9,600 or 5.2e-6 for mRNA. We also compared our list of mRNA that were differentially expressed to those that were predicted to differ based on our miRNA findings.

Results

There were 21 pregnant and 21 nonpregnant participants with sufficient RNA from cervical swabs for miRNA analysis. Participant demographic characteristics are reported in Table 1. Median age was 30.2 years (interquartile range [IQR] 28.2–33.5) among nonpregnant participants and 30.5 years (IQR 28.6–31.3) among pregnant participants. In both groups, 95.2% of participants were nulliparous. The majority of participants had normal Papanicolaou cytology and negative Human Papilloma Virus testing. Histologic assessment for bacterial vaginosis from cervical specimens revealed a single case of infection in each group. Approximately half of participants from both cohorts had cervical exposure to semen in the week preceding specimen collection.

Table 1. . Participant characteristics among 42 pregnant and nonpregnant, non Hispanic white participants of the Spontaneous Prematurity and Epigenetics of the Cervix study, Boston, MA, USA.

Variables	Nonpregnant n = 21	Pregnant n = 21
Age years (Median [IQR])	30.2 (28.2–33.5)	30.5 (28.6–31.3)
Education
– Less than a college degree	1 (4.8)	2 (9.5)
– College degree	6 (28.6)	8 (38.1)
– Graduate degree	13 (61.9)	11 (52.4)
Household income
– <$75,000	13 (61.9)	3 (14.3)
– ≥$75,000	7 (33.3)	18 (85.7)
Gravidity
– 0	19 (90.5)	0 (0.0)
– 1	0 (0.0)	17 (81.0)
– >1	2 (9.5)	4 (19.5)
Nulliparous	20 (95.2)	20 (95.2)
Most recent Papanicolaou test result
– Normal	17 (81.0)	19 (90.5)
– Atypical cells of uncertain significance	1 (4.8)	0 (0.0)
– Unsatisfactory	1 (4.8)	0 (0.0)
Histopathologic slide leukocyte burden
– Low	16 (76.2)	9 (42.9)
– High	4 (19.0)	12 (57.1)
Sexual intercourse within past week	12 (57.1)	10 (47.6)

Unless specified, data presented as n (column %) and may not add to 100% due to missing data: education n = one pregnant woman, income n = one pregnant woman, Papanicolaou test n = two pregnant and two nonpregnant women; histopathologic slide n = one pregnant woman.

IQR: Interquartile range.

A total of 21 miRNA were differentially expressed in pregnant compared with nonpregnant participants and are reported in Table 2. After accounting for leukocyte burden, five miRNA remained differentially expressed, including miR-21, miR-150, miR-335, miR-126 and miR-101. Notably, expression of all miRNA was significantly downregulated in cervical specimens from the cohort of pregnant participants.

Table 2. . Differential microRNA expression among 21 pregnant and 21 nonpregnant women.

miRNA	Nonpregnant		Pregnant		p-value	Fold change (pregnant/nonpregnant)
	n	Mean ΔCT	n	Mean ΔCT
miR-126	17	2.3 ± 4.2	18	8.7 ± 2.8	0.00000†	85.9
miR-10a	20	0.1 ± 3.5	21	4.4 ± 2.3	0.00004	19.2
miR-218	20	1.5 ± 3.0	18	4.9 ± 1.9	0.00018	11.1
miR-150	19	-3.5 ± 2.1	20	-0.2 ± 2.6	0.00010†	9.7
let-7f	18	0.2 ± 2.1	13	3.4 ± 1.4	0.00004	9.4
miR-21	20	-3.8 ± 2.4	21	-0.6 ± 1.9	0.00002†	9.3
miR-224	17	-0.2 ± 1.7	13	2.8 ± 1.5	0.00002	7.9
miR-101	20	4.4 ± 2.0	19	7.4 ± 1.8	0.00003†	7.5
miR-335	20	2.4 ± 2.0	19	5.2 ± 2.0	0.00011†	6.9
let-7a	14	-3.9 ± 1.4	12	-1.1 ± 1.0	0.00001	6.7
miR-130	20	0.9 ± 2.5	20	3.6 ± 1.2	0.00012	6.6
miR-152	20	1.1 ± 2.4	21	3.7 ± 1.7	0.00018	6.5
let-7e	14	-2.5 ± 1.9	13	0.2 ± 1.0	0.00011	6.4
miR-20a	20	-4.3 ± 2.3	21	-1.9 ± 1.4	0.00016	5.5
miR-148	20	-1.2 ± 2.0	21	1.3 ± 1.7	0.00013	5.5
miR-141	20	-1.3 ± 1.9	20	1.1 ± 1.4	0.00006	5.3
miR-106	20	-3.2 ± 2.1	21	-1.0 ± 1.3	0.00016	4.9
miR-99b	20	-2.1 ± 2.0	21	0.1 ± 1.5	0.00015	4.9
miR-17	20	-3.6 ± 2.0	20	-1.5 ± 1.2	0.00018	4.5
miR-598	17	7.4 ± 1.1	15	9.2 ± 1.2	0.00014	3.4
miR-29a	17	4.8 ± 0.8	12	6.2 ± 0.9	0.00013	2.7

^†Models statistically significant after adjustment for histopathologic evidence of high leukocyte burden.

CT: Cycle threshold.

We next looked at putative targets of the downregulated miRNA to gain insight into the possible biological sequelae of their differential expression. Using publically available online prediction software, a total of 2137 genes were identified, of which 258 were overlapping putative targets of more than one of the upregulated miRNA (Supplementary Table 1). Functional clustering of genes from both composite and overlapping putative targets was then performed using an ontology classification database (Supplementary Tables 2 & 3). A list of the top 30 enriched pathways is reported in Table 3 after exclusion of functionally redundant processes. Among these are genes involved in organogenesis, regulation of glucocorticoid signaling, and cell adhesion and motility, as well expression of ncRNA machinery.

Table 3. . Functional clustering of putative microRNA targets.

Functional pathway	Fold enrichment	p-value	FDR
Regulation of glucocorticoid mediated signaling pathway (GO:1900169)	81.2	8.70E-04	4.71E-02
Beta-catenin destruction complex assembly (GO:1904885)	48.8	9.51E-05	8.24E-03
Regulation of epithelial cell proliferation involved in lung morphogenesis (GO:2000794)	30.5	2.73E-04	1.95E-02
Negative regulation of cytoplasmic translation (GO:2000766)	30.5	2.73E-04	1.94E-02
Mesenchymal–epithelial cell signaling (GO:0060638)	30.5	2.73E-04	1.93E-02
Positive regulation of nuclear-transcribed mRNA poly(A) tail shortening (GO:0060213)	22.2	5.86E-04	3.53E-02
Regulation of nuclear-transcribed mRNA poly(A) tail shortening (GO:0060211)	22.2	5.86E-04	3.52E-02
Positive regulation of nuclear-transcribed mRNA catabolic process (GO:1900153)	20.3	8.88E-05	7.87E-03
miRNA mediated inhibition of translation (GO:0035278)	20.3	7.26E-04	4.09E-02
Regulation of translation, ncRNA-mediated (GO:0045974)	20.3	7.26E-04	4.08E-02
Negative regulation of cardiac muscle cell differentiation (GO:2000726)	18.8	8.85E-04	4.78E-02
Negative regulation of striated muscle cell differentiation (GO:0051154)	10.2	9.26E-04	4.95E-02
Regulation of sodium ion transmembrane transporter activity (GO:2000649)	10.2	4.83E-05	4.83E-03
Phosphatidylinositol-3-phosphate biosynthetic process (GO:0036092)	9.8	5.95E-05	5.66E-03
Negative regulation of muscle cell differentiation (GO:0051148)	9.0	3.47E-04	2.33E-02
Tricuspid valve development (GO:0003175)	8.2	1.89E-03	3.10E-02
Peptidyl-threonine modification (GO:0018210)	7.7	1.79E-05	2.14E-03
Regulation of ventricular cardiac muscle cell membrane depolarization (GO:0060373)	7.0	3.00E-03	4.47E-02
Regulation of metanephric nephron tubule epithelial cell differentiation (GO:0072307)	7.0	3.00E-03	4.47E-02
Regulation of sarcomere organization (GO:0060297)	7.0	3.00E-03	4.46E-02
Parathyroid gland development (GO:0060017)	7.0	3.00E-03	4.46E-02
Positive regulation of cell proliferation involved in heart morphogenesis (GO:2000138)	7.0	3.00E-03	4.45E-02
Semicircular canal morphogenesis (GO:0048752)	7.0	3.00E-03	4.45E-02
miRNA loading onto RISC involved in gene silencing by miRNA (GO:0035280)	7.0	3.00E-03	4.45E-02
Lipid phosphorylation (GO:0046834)	6.5	1.87E-05	2.19E-03
Regulation of cell-matrix adhesion (GO:0001952)	6.4	5.65E-05	5.51E-03
Cardiac muscle cell proliferation (GO:0060038)	6.3	1.15E-04	2.88E-03
Cerebral cortex development (GO:0021987)	6.4	1.99E-05	2.32E-03
Hippocampus development (GO:0021766)	6.3	5.61E-04	3.45E-02
Membrane raft assembly (GO:0001765)	5.9	2.15E-03	3.42E-02

FDR: False discovery rate; GO: Gene ontology; ncRNA: Noncoding RNA.

There were 22 pregnant and 21 nonpregnant participants for whom there were sufficient quantities of RNA for microarray analysis of lncRNA and mRNA. Five of the pregnant participants and four of the nonpregnant participants differed from those in the miRNA analysis. Covariate distributions were similar between the two incompletely overlapping cohorts (Supplementary Table 3). Expression of lncRNA revealed upregulation of 22 lncRNA and downregulation of two lncRNA (Table 4). After adjustment for leukocyte burden, three lncRNA remained differentially expressed, all of which were upregulated among pregnant participants.

Table 4. . Differential long noncoding RNA expression in the cervix among 22 pregnant women and 21 nonpregnant women.

lncRNA	Fold change (pregnant/nonpregnant)	Regulation	p-value
G072782	10.6	up	2.48E-06
RP11-290L1.3	8.4	up	5.67E-07†
uc.329	7.8	up	2.98E-06
TGFA-IT1	7.6	up	2.50E-06
DPY19L2P1	7.4	up	1.40E-06
G015724	6.7	up	1.61E-06
uc.156	6.4	up	1.89E-06
RP11-376M2.2	4.6	up	9.94E-07†
XLOC_003225	4.6	up	1.70E-06
AC006041.1	3.7	up	1.07E-06
LOC374443	3.6	up	2.82E-07
G045267	3.3	down	2.82E-06
G010253	3.0	up	1.24E-06
MYLK-AS1	2.8	up	7.35E-08†
G085423	2.8	down	2.68E-06
G053267	2.8	up	1.53E-06
RP11-403P17.3	2.7	up	9.47E-08
uc.429	2.7	up	2.14E-06
DQ574721	2.4	up	1.83E-06
RP11-213G6.2	2.4	up	1.90E-06
G030628	2.2	up	2.58E-06

^†Models statistically significant after adjustment for leukocyte burden.

lncRNA: Long noncoding RNA.

To extend our analysis and validate miRNA findings, we examined expression of mRNA in pregnant and nonpregnant participants. mRNA transcripts from 20 genes were upregulated, and transcripts from five genes were downregulated (Table 5). After adjustment for leukocyte burden, six mRNA remained differentially expressed, all of which were upregulated among pregnant participants. We then compared putative gene targets from differentially regulated miRNA to mRNA data. Among the six upregulated mRNA, two were putative targets of at least one of the downregulated miRNA and include SRRM4 and GRHL2.

Table 5. . Differential mRNA expression in the cervix among 22 pregnant women and 21 nonpregnant women.

mRNA	Fold change (pregnant/nonpregnant)	Regulation	p-value
CRYGS	5.6	up	1.48E-06
PRR20E	5.4	up	3.09E-06
TMEM254	4.9	up	1.05E-06
DNAH7	4.8	up	3.99E-06
SGPL1	3.4	down	3.44E-06
BUB3	3.3	up	2.61E-06
C9orf135	3.1	up	5.01E-06
AC007952.6	3.0	up	4.66E-08†
NAPA	3.0	down	2.14E-06
EZH1	2.7	up	6.30E-07
WIPF2	2.7	up	3.67E-06
CCDC116	2.6	up	9.83E-07
MFSD12	2.6	down	2.67E-06
ZNF485	2.5	up	3.06E-06
DNASE1L2	2.5	down	4.27E-06
ADGRD1	2.4	up	1.23E-07†
SRRM4	2.4	up	6.03E-07†
OR10W1	2.4	up	2.40E-06
FOXL2NB	2.3	up	6.24E-07
GRHL2	2.3	up	6.32E-07†
SNAP47	2.2	up	2.30E-06
CCDC33	2.2	up	3.33E-06†
PYDC1	2.1	up	5.25E-07†
PTCRA	2.1	up	1.89E-06
ETNK2	2.1	down	4.73E-06

^†Models statistically significant after adjustment for leukocyte burden.

Bold mRNA are putative targets of differentially detected miRNA.

Discussion

This study characterizes a cervical signature of ncRNA associated with normal pregnancy. These data identify a unique subset of downregulated cervical miRNA associated with pregnancy, which are commonly predicted to target functional processes involved in organogenesis, regulation of glucocorticoid signaling, and cell adhesion and motility, as well as expression of ncRNA machinery.

Our finding that all differentially expressed miRNA in pregnancy were downregulated suggests increased expression of downstream targets at the mRNA and protein level. This is consistent with our observation that a greater number of mRNA were upregulated rather than downregulated in response to pregnancy. Global repression of miRNA expression has been reported in cancer models [2,33,34] and may result in part from changes in expression of miRNA processing machinery [35,36]. Interestingly, functional clustering of putative miRNA targets from our study identified enrichment in genes involved in miRNA biogenesis.

With respect to cancer, numerous published studies link differential miRNA expression with cervical carcinoma progression and metastasis [37–39]. Of note, expression patterns generally involve upregulation of these small regulatory molecules [40–44], in contrast to the directionality of our reported findings and general observations that cancer induces global miRNA repression as aforementioned. Individual miRNA have been identified as potentially useful diagnostic markers in cervical cancer, and target validation has provided insight into the biologic sequelae of these expression patterns [37,45]. The majority of the miRNA identified in our study have been implicated in the pathophysiology of cervical cancer, including miR-21, miR-150, miR-126 and miR-101 [37]. This finding may reflect tissue-specific miRNA patterns, supporting the premise that they could be biologically active in pregnancy.

Several of these downregulated miRNA also have been examined in pregnancy and correlated with adverse outcomes, most notably miR-21. Maternal serum and placental miR-21 expression have been linked to increased risk of gestational diabetes [46], fetal macrosomia [47–49], preeclampsia [50–53], fetal hypoxia [54] and intrauterine growth restriction [55,56]. Given tissue-specific expression profiles of miRNA, it is feasible that aberrant expression of a miRNA in maternal serum may correlate with pathologic conditions, while expression of the same miRNA in the cervix may represent normal physiology.

With respect to cervical miRNA expression in pregnancy, only miR-21 among the identified five miRNA has been previously studied. Increased expression of miR-21 in the cervix was associated with longer gestations in a study of women in Mexico City [20]. Our data may seem counterintuitive given these findings; one might expect to observe lower miR-21 expression in nonpregnant participants compared with pregnant participants given the link between increased expression of this miRNA and pregnancy prolongation. One plausible explanation is that the directionality of this association is different in pregnancy. Pregnant participants may have lower miR-21 levels overall, but within that range higher levels are associated with longer gestations. Furthermore, the two datasets do not lend themselves to direct comparison, as they used different platforms. The study of gestational length in Mexico also was demographically and geographically distinct from the present study of non Hispanic women in MA, USA. Of note, none of the miRNA identified in our study were among those linked to preterm birth reported previously [19], supporting the notion that they characterize normal cervical response to pregnancy.

Notably, two cervical miRNA that have been implicated in the regulation of epithelial function in spontaneous preterm birth were not identified in our study. Specifically, Elovitz et al. demonstrated that miR-143 and miR-145 increase cervical epithelial cell permeability, downregulate cell adhesion molecules and initiate cell cycle arrest, thereby leading to epithelial barrier breakdown and cervical remodeling [19,24]. It is plausible that these miRNA did not appear among those differentially detected in cervical specimens of our pregnant versus nonpregnant participants since pathologic cervical epithelial barrier breakdown was not occurring in our cohort. Moreover, initiators of cervical remodeling in spontaneous preterm birth are believed to be different than those mediating normal parturition [57,58].

The observed enrichment in putative miRNA targets involved in cell motility, adhesion and cytoskeletal structure is intuitive and suggests a novel role for these regulatory ncRNA in promoting the cervical remodeling required to maintain a healthy pregnancy. Glucocorticoid signaling emerged in our analysis as another biologic process of interest, though little is known about the pathway in a cervical model outside of cancer. Putative targets that involve regulation of glucocorticoid signaling include genes that are both suppressive and activating in function, indicating that the cervix may continuously fine-tune its response to environmental stress to achieve metabolic balance.

Enrichment in predicted miRNA targets involved in organogenesis, including cardiac, neurological, renal, endocrine, muskoloskeletal and pulmonary development, raises the possibility that cervical specimens contain genetic material from other cell types, including those of fetal origin. Detection of cell-free DNA from placental origin in maternal serum underlies the premise of noninvasive prenatal testing in pregnancy, a technology commonly utilized for screening of autosomal and sex chromosome aneuploidies [59]. Though RNA is less biologically stable than DNA, our data raise the possibility that genetic material arising from the pregnancy may be present in or surrounding maternal tissues such as the cervix. This theory is plausible given the proximity of the vascular cervix to the developing pregnancy, as well as the fact that several studies have demonstrated that genetic material arising from the pregnancy may be detected in cervical specimens [60,61]. Another interpretation of these data is that cervical cells may be responding to cues in the local microenvironment, and that changes occurring in the fetus or placenta may drive cervical physiology or pathology. This is congruent with our explanation of the aforementioned glucocorticoid signaling.

Notably, the directionality of the miRNA expression patterns correlates with SRRM4 and GRHL2 mRNA findings. Specifically, miR-101 is downregulated while putative target SRRM4 is upregulated among pregnant participants. Likewise, miR-150 is downregulated, while its target GRHL2 is upregulated among pregnant participants. These observations support the possibility that these miRNA mediate expression of these genes in the cervix during gestation. SRRM4 is involved in alternative splicing and mRNA binding, while GRHL2 is a transcription factor involved in epithelial differentiation. Both genes are known to be active in embryogenesis. Further research will confirm these expression changes at the protein level, as well as gain mechanistic insight into their role in pregnancy.

Data examining expression of lncRNA in the cervix during pregnancy are sparse. Published literature has largely focused on the role of these regulatory ncRNA in cancer initiation and progression [62–64]. Few studies have examined lncRNA during pregnancy; expression profiles have been limited to maternal serum and placental tissue and, similar to miRNA, highlight a role for lncRNA in potentiating adverse outcomes including preeclampsia, invasive placentation and intrauterine growth restriction [9]. Burris et al. identified a unique signature of nine cervical lncRNA associated with length of gestation [27]. While none of the lncRNA identified in this analysis overlapped with those from the previously published study, this is likely due to use of different platforms, as well as the fact that Burris et al. examined lncRNA patterns in a pregnancy cohort while the current study compares differential lncRNA expression between pregnant and nonpregnant women.

One strength of this study is the incorporation of two classes of regulatory ncRNA, as well as mRNA data, which capture multiple levels of gene regulation and provide a thorough depiction of the cervical response to pregnancy. This study design includes a cohort of healthy women who went on to have term deliveries; such criteria diminish the likelihood that pregnancy-associated complications of preeclampsia, fetal growth restriction, or preterm birth confound interpretation of our findings. These data further account for a potentially heterogeneous population of cells from the cervical swab by correcting for leukocyte burden.

Limitations of our study include the lack of additional RNA to perform technical validation of mRNA and lncRNA using a second platform such as RT-PCR. We did, however, use a PCR-based platform for the miRNA analysis that did not require technical validation. Second, the PCR-based arrays of lncRNA are limited to candidate approaches of approximately 80 lncRNA, such as the Qiagen RT2 platform, which may miss relevant lncRNA [27]. While many studies simply use bioinformatic approaches to identify potential miRNA targets [65], this study combined that approach with measurement of lncRNA and mRNA in the same samples to capture a broader biologic picture. Nonetheless, our use of microarray technology for lncRNA and mRNA warrants replication in other studies.Our study may lack generalizability to other racial/ethnic groups, as it included only non Hispanic white women. While homogeneity offers advantages in limiting the variability of single nucleotide polymorphisms, it cannot provide insight into the cervical physiology among other groups of women. This is particularly pertinent with respect to African American women who have the highest risk of spontaneous preterm birth in the USA. Future, larger studies should focus on these women. Given that changes in the expression of coding and ncRNA do not necessarily correlate with protein expression and function, additional research will focus on validation of identified targets and pathways at the protein level to gain greater insight into the biologic relevance of these data. Further experimental studies to gain mechanistic insight are needed to understand molecular regulation of cervical function.

Conclusion

In conclusion, we identified a unique cervical miRNA and lncRNA expression profile in pregnant women compared with nonpregnant women. If replicated, our findings will inform target-specific, efficient analyses of larger cohorts of women to gain greater insight into the cervical biology of pregnancy.

Future perspective

Future work examining the dysregulation of cervical ncRNA in pregnancy associated with adverse pregnancy outcomes, such as spontaneous preterm birth, is of critical importance. The work presented herein lays the foundation for such studies by describing the normal cervical ncRNA expression profile of the pregnant compared with nonpregnant state. It is plausible that ncRNA identified in this study may be the ones that are mechanistically involved in maintaining adequate cervical function in pregnancy. Future work in human observational studies, as well as mechanistic in vitro and in vivo studies, will be necessary to determine whether dysregulation of these small molecules is involved in the pathogenesis of spontaneous preterm birth. If so, these sequences may be targets of future potential therapeutics, such as antagomirs, to disrupt the cascade of events that lead to premature cervical remodeling and spontaneous preterm birth.

Summary points.

• The purpose of this study was to compare noncoding RNA (ncRNA) expression in the cervix among pregnant and nonpregnant women.

• Understanding the expression of cervical ncRNA in pregnancy may shed light on whether ncRNA are dysregulated in pregnancies complicated by adverse outcomes.

• We identified a unique cervical ncRNA signature, including downregulation of select miRNA and long noncoding RNA, in the pregnant compared with the nonpregnant state.

• All differentially expressed miRNA in pregnancy were downregulated, suggesting increased expression of downstream targets at the mRNA and protein level.

• Differentially detected cervical mRNA were upregulated in response to pregnancy, including putative targets of the identified miRNA.

• Downregulated cervical miRNA in pregnancy are commonly predicted to target functional processes involved in organogenesis, regulation of glucocorticoid signaling, and cell adhesion and motility, as well as expression of ncRNA machinery.

• Future work focused on whether pregnancy-associated ncRNA are dysregulated in pregnancies with adverse outcomes is a research priority.

• If ncRNA are involved in the pathogenesis of spontaneous preterm birth, they may serve as targets for future therapeutics aimed at preterm birth prevention.