Transcriptome and proteome dynamics of cervical remodeling in the mouse during pregnancy

Abstract

During gestation, the female reproductive tract must maintain pregnancy while concurrently preparing for parturition. Here, we explore the transitions in gene expression and protein turnover (fractional synthesis rates [FSR]) by which the cervix implements a transition from rigid to compliant. Shifts in gene transcription to achieve immune tolerance and alter epithelial cell programs begin in early pregnancy. Subsequently, in mid-to-late pregnancy transcriptional programs emerge that promote structural reorganization of the extracellular matrix (ECM). Stable isotope labeling revealed a striking slowdown of overall FSRs across the proteome on gestation day 6 that reverses in mid-to-late pregnancy. An exception was soluble fibrillar collagens and proteins of collagen assembly, which exhibit high turnover in nonpregnant cervix compared with other tissues and FSRs that continue throughout pregnancy. This finding provides a mechanism to explain how cross-linked collagen is replaced by newly synthesized, less cross-linked collagens, which allows increased tissue compliance during parturition. The rapid transition requires a reservoir of newly synthesized, less cross-linked collagens, which is assured by the high FSR of soluble collagens in the cervix. These findings suggest a previously unrecognized form of “metabolic flexibility” for ECM in the cervix that underlies rapid transformation in compliance to allow parturition.

Flux proteomics identify a steady reservoir of newly synthesized fibrillar collagens in the pregnant cervix, which in conjunction with pregnancy-specific transcriptional programs ensure increased cervical compliance during the process of birth.

Graphical Abstract

Graphical Abstract.

Introduction

Over the course of pregnancy, the cervix undergoes a remodeling process that is unprecedented in most adult tissues. The period of competence, which spans the majority of pregnancy, is referred to as the cervical softening phase of remodeling, whereas the period of maximal compliance is referred to as the cervical ripening and dilation phase.

The cervix epithelial and stromal compartments play distinct roles in the remodeling process. During pregnancy, the epithelia undergo a marked proliferation, provide a physical and immunological barrier against pathogens, and synthesize enzymes that modulate the local steroid hormone environment [1–3]. The stroma comprises multiple cell types that include fibroblasts and smooth muscle cells as well as a complex extracellular matrix (ECM; [4]). Changes in composition and structure of the ECM drive the transition of the nonpregnant cervix from closed and rigid to one that becomes soft yet competent during pregnancy and at term capable of large deformations. Studies in the mouse continue to expand our understanding as to how changes in processing, assembly, and organization of the major structural protein, collagen, is modulated starting in early pregnancy to allow for the gradual decline in mechanical strength of the cervix [5–8]. In addition, the demonstrated structural reorganization of elastic fibers likely contributes to the ability of the cervix to undergo large deformations without permanent tissue damage [9, 10]. Modulation of collagen and elastic fiber structure and function is dependent on the composition of nonstructural ECM proteins such as proteoglycans, hyaluronan, and non-fibrillar collagens as well as post-translational modifications to collagen and elastic fibers (e.g., cross-link density and hydroxylation; [11, 12]).

The process of cervical remodeling is exquisitely regulated by the steroid hormones, progesterone and estrogen, and include the transition from competence to compliance by regulation of cervical ECM structure and function [9]. Cervical softening, which encompasses the majority of pregnancy, is achieved in a progesterone dominant phase, whereas ripening/dilation at the end pregnancy is achieved in an estrogen dominant environment. A marked decline in genes encoding lysyl oxidase (Lox) and lysyl hydroxylase (Plod2) by gestation day 8 result in a decline in formation of the stronger lysylpyridinoline (LP) and hydroxylysylpyridinoline (HP) intermolecular cross-links between collagen fibrils and rise in immature cross-links by gestation day 12. The replacement of collagen fibers containing mature by those with immature cross-links corresponds to the decline in mechanical strength that begins by gestation day 12 [5, 13]. This understanding supports a further subdivision of the softening period into an early period prior to a significant decline in stiffness (up to gestation day 11) and late softening period (gestation days 12–17; [10, 14]). In early softening cervical tissue stiffness is similar to or greater than that of the nonpregnant (NP) cervix, whereas in late softening the cervix undergoes a gradual decline in stiffness [7, 10, 15]. In late pregnancy during cervical ripening/dilation, a shift in the steroid environment to one of estrogen dominance results from a decline in progesterone action in humans and mice. This phase begins on gestation day 18 in the mouse with birth on day 19. Key events in this phase are an increase in hyaluronan synthesis and a marked change in elastic fiber ultrastructure [9, 16]. Although prior studies highlight key regulatory steps of ECM remodeling that begin in the early softening period, our current understanding of the phase-specific sequence of molecular events remains incomplete. In particular, the transcriptional and protein synthetic events from early to late softening that allow for protection of the upper reproductive tract and preparation for parturition remain uncertain.

Previous gene expression microarray studies in humans and mice provide partial insights into transcriptional events altered during cervical ripening/dilation and postpartum repair, but they fail to provide insights into the regulatory events that allow for cervical softening and that are likely to be required for the subsequent transition to ripening/dilation [17–21].

Furthermore, a gap in knowledge remains in how transcriptional changes connect with protein synthesis and breakdown rates and how these can provide a more complete biological picture. Steady state collagen gene expression and protein content have been reported during pregnancy, but rates of newly synthesized proteins have not been studied previously [15, 22, 23]. Further, ECM degradation pathways that maintain ECM homeostasis throughout pregnancy have not been identified. Prior studies reporting proinflammatory induced activation of ECM-targeting proteases during cervical ripening would not account for the replacement of highly cross-linked collagen with collagen that has a reduced cross-link density during the softening phase. The role of collagen-degrading matrix metalloproteases (MMP)-in late pregnancy during cervical ripening remains unresolved. A study of mechanical assessment of collagenase treated cervix did not support collagen degradation by MMP, whereas other studies suggest elevations in collagenase activity at term [24–27].

To address these gaps in knowledge and understand the basis of misregulation that contributes to premature birth, which globally impacts 15 million children annually and in the USA 10.3% of pregnancies [28, 29], we focus here on proteome dynamics and transcriptional regulation in the mouse cervix during pregnancy. We interrogated concurrently in the cervix, the dynamic changes in proteome-wide protein fluxes, by use of stable isotope labeling with tandem mass spectrometric analysis as well as changes in transcriptional patterns using RNA sequencing (RNASeq). These measurements include turnover rates of ECM proteins. We report several phase-specific molecular alterations, with coordinated shifts in transcriptional patterns and proteome-wide changes in protein synthesis rates. These findings uncover a high turnover of fibrillar collagens in both the nonpregnant and pregnant cervix, which serves to facilitate the replacement of more cross-linked collagens by less cross-linked collagen species as a mechanism for increasing tissue compliance in late cervical ripening.

Results

Transcriptional changes

To elucidate the transcriptional pathways that guide early cervical softening (gestation day 6), mid-to-late cervical softening (gestation days 12 and 15, respectively) and cervical ripening and dilation (gestation day 18) in mice, we utilized RNASeq. The selected gestational time points are distinct in the ratio of circulating progesterone to estrogen and cervical tissue mechanics and thus likely to best capture the spectrum of transcriptional changes over the course of pregnancy [14, 30–32]. Table 1 provides a schematic to describe the study comparisons and highlight key features that distinguish these unique stages of remodeling. As a reference point for early cervical softening, we generated an RNAseq dataset from the nonpregnant cervix at metestrus (NP). To define the gene signatures that are established in early remodeling, we compared gene expression patterns between NP and gestation day 6. To understand the gene expression changes that transition from early softening to later stages of softening, we compared expression patterns on gestation day 6 with days 12 and 15 (hereafter d12_15). Because of similarities in gene expression patterns (Supplementary Figure 1), steroid hormone profiles, and mechanical properties the days 12 and 15 gestational time points were assessed together and are referenced in this study as mid-to-late softening [14, 30, 31]. Finally, to understand the transcriptional signatures that change during cervical ripening/dilation, we compared gene expression patterns between gestation days 12 and 15 with gestation day 18. As depicted in the heatmap in Figure 1A, gene expression patterns are altered as early as gestation day 6 and patterns between each phase are distinct. The Venn diagram in Figure 1B indicates upregulated and downregulated genes for each comparison. The greatest increase in gene expression occurred between early and late softening (d6 vs day 12_15) and the greatest number of downregulated genes was observed in early softening (NP vs d6). A greater number of up- and downregulated genes were common between the NP vs d6 group and the d6 vs d12_15 group as compared with the d12_15 vs d18 group.

Table 1.

Distinct patterns of mechanical stiffness and circulating progesterone to estrogen ratio (P:E ratio) during early, mid, late softening, and ripening/dilation phases of cervical remodeling.

Figure 1.

RNAseq analysis of the cervix from nonpregnant and pregnant mice demonstrate dynamic changes in pathway specific transcript patterns over the course of pregnancy. Panel A: Heatmap comparing relative changes in expression between NP and gestation day 6 in the first column, gestation day 6 versus days 12 and 15 in the second column and gestation days 12 and 15 versus day 18 in the last column. Panel B: Venn diagram identify the number of common and distinct up- and downregulated genes between comparison groups. Panel C: GO pathway analysis of the most down- and upregulated biological processes between nonpregnant and early softening ([NP vs D6]; top panels), early softening and mid-to-late softening ([D6 vs D12_15]; middle panels), and mid-to-late softening and ripening ([D12_15 vs D18; lower panels). Orange bars depict the statistically significant biological processes as indicated by the log10 P-values. Blue bars indicate the percentage of genes in the process that are differentially regulated in the dataset. Panel D: Transcriptional patterns of genes associated with synthesis, processing, assembly, organization of fibrillar collagen, and cell-ECM communication. The absence of FPKM values for Col1a1 on d12 is an unexplained observation of the RNASeq dataset. Panel E: Transcriptional patterns of MMPs that breakdown fibrillar collagen and other ECM components.

Pathways that inform the succession of changes required for the maintenance of cervical competence and simultaneous cervical remodeling, were identified by gene ontology (GO) analysis. Figure 1C lists the most enriched biological processes identified for downregulated or upregulated pathways. Genes in the upregulated biological processes in early softening included steroid receptors (androgen receptor (Ar) and estrogen receptor alpha (Esr1)), cell proliferation markers (cyclin D 1 (Ccnd1), cyclin D2 (Ccnd2), and kit ligand (Kitl)) transcription factors (Sry (sex determining region)-box 9 (Sox9), T-box transcription factor 2 (Tbx2), and Forkhead box A2 (Foxa2)), and antiviral innate immunity (Interferon-induced protein with tetratricopeptide repeats 1 (Ifiti1), Interferon-induced protein with tetratricopeptide repeats 2 (Ifit2), Interferon-induced protein with tetratricopeptide repeats 3 (Ifit3), and 2′-5′-oligoadenylate synthase 2 (Oas2)). Downregulated pathways/genes highlight a marked suppression of inflammatory responses (chemokine (C-X-C motif) ligand 1 (Cxcl1), chemokine (C-X-C motif) ligand 5 (Cxcl5), NF-kappa-B inhibitor delta (Nfkbid), interleukin 1 beta (IL1b), and protein S100a8 (S100a8)) and loss of epithelial cell structural components associated with keratinocytes (e.g., small proline-rich protein genes (Sprr genes), transglutaminase 1, K polypeptide, transglutaminase 3, K polypeptide (Tgm1/3), involucrin (Ivl), and loricrin (Lor)). The transcriptional complexity from early softening to mid-to-late softening is further increased by upregulation of pathways/genes associated with ECM reorganization (e.g., collagen type XIV alpha 1 chain (Col14a1), a disintegrin-like and metallopeptidase (reprolysin type) with thrombospondin type 1 motif (Adamts14), collagen type III alpha 1 chain (Col3a1), and periostin (Postn)), the interaction of cells with the ECM (e.g., integrin alpha 1 (Itga1), integrin subunit beta 3 (Itgb3), and tenascin XB (Tnxb)), and proteolysis (e.g., a disintegrin-like and metallopeptidase (reprolysin type) with thrombospondin type 1 motif 15, procollagen C-endopeptidase enhancer (Adamts15, Pcolce), and a disintegrin-like and metallopeptidase (reprolysin type) with thrombospondin type 1 motif 9 (Adamts9)). Downregulated pathways from early to late softening indicate further suppression of structural proteins important for keratinocyte differentiation (e.g., keratin 6A (Krt6a) and keratin 16 (Krt16)) and a decline in the defense response to virus (e.g., Ifit1 and Oas1b). The transition from mid/late cervical softening to cervical ripening/dilation is accompanied by an upregulation of pathways that promote epithelia keratinization (e.g., Krt6a, Tgm1, and small proline rich protein 1A (Sprr1a)). Biological pathways associated with collagen fibril organization, proteolysis, and cell adhesion are downregulated (e.g., fibromodulin (Fmod), collagen type XIV alpha 1 chain (Col14a1), and carboxypeptidase M (Cpm)). A complete list of genes in each GO pathway is provided in Supplementary Tables 1–3.

The unbiased assessment of GO pathways regulated in the cervix at distinct time points in gestation extends current understanding of cervical remodeling in that ECM reorganization, particularly in the mid-to-late softening phase, is a key process. We sought to further leverage the current RNAseq datasets to discern the pattern of expression of specific genes critical to the synthesis, processing, assembly, and turnover of fibrillar collagen. Figure 1D displays a heatmap that presents relative Fragments Per Kilobase of transcript per Million mapped reads (FPKM) values of genes encoding fibrillar collagens, processing enzymes, collagen fibril assembly and packing, and collagen organization as well as cell-ECM communication from NP, and gestation days 6, 12, 15, and 18. Along with the previously reported increase in transcription of fibrillar collagens I and III [5], fibrillar collagen V subunits (Col5a1 and Col5a2) were also increased in mid-to-late cervical softening. Transcripts encoding factors necessary for collagen fibril processing, assembly and packing are induced through pregnancy. Examples of previously described factors identified in the dataset include downregulation of Lox and upregulation of the proteoglycans important in collagen assembly; decorin, biglycan, and fibromodulin [5, 9, 33]. Newly identified factors include upregulation of Adamts2 that encodes a protease required for N-terminal processing of newly synthesized collagen and an increase in transcripts that regulate collagen assembly such as the proteoglycan lumican and the Fibril Associated Collagen with Interrupted Triple helices (FACIT collagens), collagen type XII alpha 1 chain (Col12a1), and collagen type XIV alpha 1 chain (Col14a1) [34, 35]. Interestingly, the composition of matricellular proteins is uniquely changed in pregnancy. The noted decline in thrombospondin 2 and tenascin C is consistent with previous findings [5]. Novel to the current dataset, we identified an increase in dermatopontin (Dpt), secreted protein acidic and cysteine rich (Sparc), and periostin (Postn). Collectively, these data identify new components of the ECM that may contribute to cervical function and provide further demonstration of the transcriptional basis of dynamic changes in the processing and composition of the cervical ECM that impacts mechanical function of the cervix through pregnancy.

Previous reports demonstrate a reduction in HP and LP collagen cross-links that begin on gestation day 12 yet collagen content remains constant through pregnancy [14, 15]. This finding raises the question as to how turnover of fibrillar collagen is achieved in pregnancy that allows both an increase in tissue compliance and simultaneously maintains collagen homeostasis. In view of the robust induction of transcripts that promote collagen structural reorganization in the cervix during pregnancy, we hypothesized that collagen turnover rates (replacement of old by new collagen molecules) may be greater in pregnancy relative to the NP cervix. The assessment of MMP gene expression patterns in RNAseq datasets at time points in gestation relative to the NP cervix identifies transcripts encoding MMPs (Figure 1E) but cannot provide evidence to support MMP activity nor importantly a direct assessment of collagen turnover in pregnancy. We sought a direct approach to evaluate the kinetics of collagen turnover as well as general turnover of proteins across the proteome.

Protein synthesis across the proteome, including ECM proteins

We use a dynamic proteomics approach that quantifies protein fluxes by stable isotope labeling and LC/MS–MS-based analysis. The method is based on mass isotopomer distribution analysis (MIDA) of tryptic peptides from a proteome after in vivo metabolic labeling with heavy water (²H₂O) as previously described [36]. Cervical tissue was collected from NP and gestation days 6, 12, 15, and 18, in each case after 3 days of heavy water treatment as depicted in Supplementary Figure 2. Cervical tissue proteins were extracted into sodium dodecyl sulfate (SDS)-soluble, a guanidine hydrochloride-soluble, and a guanidine-insoluble fractions. Physical separation of tissue fractions allows different classes of proteins to be measured by liquid chromatography with tandem mass spectrometry (LC/MS–MS), particularly for ECM proteins [36–39]. Fractional synthesis rates (FSRs) were calculated for proteins identified in each fraction.

The protein replacement levels of SDS-soluble proteins are indicated in the heatmap in Figure 2A. Interestingly, there was a clear global slowdown of protein turnover rates on gestation day 6. In particular, chromatin associated proteins, an indirect marker of cell proliferation, were enriched in the soluble fraction and showed coordinated downregulation of synthesis rates in the day 6 cervix.

Figure 2.

Heat maps of protein replacement rate levels. Panel A: Clustering of proteomics data from SDS soluble samples based on their FSRs (expressed as a percentage). On the right side, chromatin associated proteins are identified. Panel B: Clustering of proteomics data from guanidine soluble samples based on their FSRs (expressed as a percentage). On the right side, ECM proteins identified and clustered by category. Panel C: Clustering of proteomics data based on FSRs of proteins that are insoluble in guanidine. On the right side are ECM proteins identified and clustered based on category.

ECM proteins were identified in the guanidine-soluble and insoluble fractions. The former comprises soluble ECM proteins (e.g., newly synthesized proteins or degraded fragments), including less cross-linked collagen, whereas the latter contains more highly cross-linked, mature collagens [36, 40]. The replacement rates of guanidine-soluble proteins are illustrated in the heatmap in Figure 2B with a subset of ECM proteins clustered by category shown on the right. Major components of this fraction were ECM proteins including collagens, proteoglycans, and microfibrillar proteins. In addition, actin subunits as well as the epithelial cell specific intermediate filament keratin proteins were also enriched. Compared with the pattern of overall replacement rates in the NP cervix, there was a striking decline in general protein replacement rates on gestation day 6 that reversed in the mid-to-late softening period and tended to fall again in the ripening phase at day 18. In contrast to this general pattern, however, fibrillar collagens and proteins involved in the more soluble collagen category maintained high replacement rates throughout pregnancy.

Assessment of the insoluble collagen fraction (Figure 2C) revealed lower protein replacement rates on gestation day 6 as compared with other time points in pregnancy and the nonpregnant state. Replacement rates for fibrillar and nonfibrillar collagens were lower in the insoluble fraction at day 6.

As dynamic protein turnover is influenced by both the rates of protein synthesis and degradation, we sought to determine the correlation between gene expression and FSRs for proteins identified in each fraction. The statistical tested used was a binomial analysis of their increased or decreased relative expressions compared with the NP control (Supplementary Figure 3). In the SDS-soluble fraction, the strongest direct correlation between protein FSR and cognate gene expression was at gestation day 15 (2-tailed P: 0.03). In contrast, gene expression follows most strongly the FSR expression in the guanidine-soluble set of proteins. Binomial distribution analysis of gene expression to their corresponding protein FSR (Supplementary Figure 3) showed a consistent correlation for all time points (2-tailed P: day 6 = 0.02; day 12 = 0.02; day 15 = 0.01; and day 18 = 0.04). When comparing gene expression to the insoluble FSR of proteins identified, there was no correlation observed at any time point (Supplementary Figure 3). Collectively this indicates that the gene expression was most directly related to directing soluble collagen synthesis rates. Table 2 illustrates comparisons between gene expression and FSR in both the guanidine soluble and insoluble fractions for fibrillar and non-fibrillar collagens and other ECM proteins required for collagen fiber assembly. The correlation between gene expression and FSRs was strongest in the guanidine-soluble group at gestation day 15 for the fibrillar collagens being compared.

Table 2.

Gene expression and protein FSR comparisons between gestational time points and nonpregnancy. Values are represented as the log2 of the fold change (FC). Negative values represent decrease expression compared with nonpregnant control. Positive values represent increased expression compared with nonpregnant control.

Interestingly, fibrillar collagen turnover in the guanidine soluble pool in the nonpregnant cervix is quite rapid (65% replaced over 3 days) compared with other tissues in the mouse that we have measured with the stable isotope labeling approach (Table 2) and [36]. This relatively high turnover rate of cervix fibrillar collagens in the soluble fraction continues throughout pregnancy. Fibrillar collagen turnover in the insoluble collagen pool is much slower (10–20% new over 3 days). It should be noted that the great majority of total collagen in most tissues is present in the insoluble, more highly cross-linked pool [36, 41, 42]. A high turnover of collagen in both NP and pregnancy requires upregulation of synthesis and degradation. Using a fluorescently tagged collagen binding peptide (CHP) that recognizes unfolded or damaged collagen as a proxy for collagen degradation [43], we observed a robust and similar level of staining in cervical sections from NP and gestation days 6, 12, 15, and 18 (Supplementary Figure 4). Thus, the observed pattern of collagen degradation is consistent with the high rates of collagen turnover in NP and pregnancy.

The protein replacement rates of individual collagen proteins expressed as a fraction replaced over 3 days are shown for fibrillar and nonfibrillar collagens in Figure 3A and B. The FSR of newly synthesized fibrillar collagens was higher in the guanidine-soluble fraction than in the insoluble-fraction, consistent with the presence of older, more mature matrix components including more highly cross-linked collagens in the insoluble fraction. The FSR for fibrillar collagens (collagen I, III, and V) showed no significant differences between the NP and pregnancy time points. The replacement rates for soluble nonfibrillar collagens (Figure 3B) were lower than fibrillar collagens in most cases with the exception of COL15A1 and the replacement rates for nonfibrillar collagens in both the soluble and insoluble fractions transiently but consistently fell at day 6 of pregnancy, with a statistically significant decline for basement membrane collagen six subunits.

Figure 3.

Protein replacement rates for collagens and ECM proteins. Panel A: Protein replacement rates expressed as a % in the guanidine and insoluble fractions for fibrillar collagens: Col1a1, Col1a2, Col3a1, and Col5a1. Panel B: Protein replacement rates expressed as a % in the guanidine and insoluble fractions for nonfibrillar collagens. * indicates statistical significance relative to other gestational time points in the same fraction. Panel C: Hydroxyproline Kinetics. (Left) Abundance analysis of insoluble OHP. (Middle) Fraction of newly synthesized OHP present in protein extracts from insoluble fraction of mouse cervices during 6, 12, 15, or 18 days of pregnancy compared with NP. (Right) Comparison between % newly synthesized OP in the insoluble fraction to the % of hydroxyproline (HP) crosslinks in cervical tissue as reported in (4). Values represent fold change of the mean ± SEM to NP (n = 5–10 cervices/time point) from two separate experiments with Dunnett multiple comparison test between NP and pregnant groups at each time point (*P < 0.03, **P < 0.01, and ****P < 0.0001). Panel D: Protein replacement rates expressed as a % in the guanidine and insoluble fraction for ECM proteins involved in assembly and organization of fibrillar collagen. Biglycan: Bgn, Decorin: Dcn, Lumican: Lum, Collagen 12a1: Col12a1, and Dermatopontin: Dpt.

Our group and others have previously demonstrated that cervical collagen content remains constant between nonpregnant and pregnancy on gestation days 6, 12, 15, and 18 [14, 15]. Quantification of collagen content in the insoluble fraction by the measure of hydroxyproline (OHP) content (Figure 3C) demonstrates no significant change in collagen content between NP and gestation days 6, 12, and 15, whereas a significant increase in collagen content on gestation day 18 was observed. Because the fractional replacement of insoluble collagens in the cervix was very low, the LC/MS–MS analytic approach for measuring individual collagen turnover was at the lower end of sensitivity for label enrichment here. Accordingly, we also measured label incorporation into total OHP in the insoluble fraction by gas chromatography–mass spectrometry (GC/MS), as this focused approach can be more sensitive to low levels of isotopic enrichment [44–46], although it does not reveal dynamics of individual collagen protein species. Hydroxyproline content did not change but the fraction of newly synthesized OHP in the insoluble fraction was significantly increased from gestation day 12 to day 18 (Figure 3C). Comparison with our previously published data on HP/triple helix content (a mature collagen cross-link) in mouse cervix during pregnancy [5] revealed a clear inverse correlation between cross-link density and its synthesis rate during the course of pregnancy. This finding is consistent with replacement of highly cross-linked collagens by newer, less highly cross-linked collagen species as a mechanism of increased tissue compliance in cervical ripening.

The rapid incorporation of newly synthesized fibrillar collagens into the insoluble fraction requires assembly of higher order collagen structures. Proteins with critical roles in assembly of fibrillar collagens were identified in the guanidine soluble or insoluble fractions (Figure 3D). These included the small leucine rich proteoglycans (SLRPs) decorin, biglycan and lumican, the FACIT collagen, COL12A1, and the matrix protein dermatopontin, DPT [47, 48]. Most of these proteins exhibited relatively high replacement rates in the range of 40–80% over 3 days, in both the nonpregnant and pregnant cervix. The replacement rate for lumican declined significantly at all pregnancy time points as compared with the nonpregnant cervix, whereas DPT increased significantly on gestation days 12 and 15.

Several ECM proteins that have not been previously described in the cervix were identified in the proteomics dataset such as the SLRP, osteoglycin (OGN), as well as proteins with roles in tissue healing (e.g., POSTN), cell-matrix interactions (annexin A2 (ANXA2)), components of the basement membrane (e.g., HSPG2, perlacan, proline arginine-rich end leucine-rich repeat (PRELP), and laminin subunti beta 1 (LAMB1)), elastic fibers/microfibrils (e.g., fibulin (FBN) and elastin microfibril interfacer 1 (EMILIN1); Supplementary Figure 5). Several groups of non-ECM proteins exhibited reduced protein replacement rates in the cervix during pregnancy as compared with the nonpregnant cervix. One example is the keratin intermediate filament proteins made by epithelial cells (Figure 4A and B). Both type I and type II keratins had reduced replacement rates during pregnancy as compared with the nonpregnant cervix. The type I keratins (KRT13, KRT15, and KRT19) and type II keratin (KRT6A) were significantly reduced at all gestational time points evaluated. The changes in keratin protein dynamics correlates well with the change in keratin transcription noted in the RNAseq data as depicted in the heatmap (Figure 4C) and in the changes in epithelial functions during pregnancy as suggested by biological pathway analysis (Figure 1).

Figure 4.

Protein replacement rates expressed as a % in the guanidine and insoluble fraction for keratin proteins. Panel A: Type I keratins. Panel B: Type II Keratins. Panel C: Heat map comparing absolute level of expression for keratin genes in RNAseq libraries from nonpregnant and pregnant gestation days 6, 12, 15, and 18.

Discussion

We leveraged bulk RNAseq transcriptome analysis and broad proteome flux analysis after heavy water labeling of proteins to elucidate transcriptional dynamics and protein replacement rates over the course of cervical remodeling during pregnancy in the mouse. Previous molecular and tissue mechanical studies have led to our current understanding that cervical remodeling starts early in pregnancy with a series of events that first ensure a balance between softening and competence followed by transition to maximal tissue compliance. However, our understanding of pathways (particularly in early pregnancy) that may drive this fine-tuned series of molecular events has remained incomplete. We identified here transcriptional pathways that are put into place in early softening (gestation day 6), mid-to-late softening (gestation days 12 and 15), and finally during cervical ripening (gestation day18). In addition, we utilized here for the first time, stable isotope labeling and proteome-wide flux analysis to assess FSRs of proteins, with a focus on ECM proteins. Each of these time points has a distinct tissue strength and steroid hormone environment (Table 1). This unique dataset has uncovered molecular signatures that define each phase of cervical remodeling.

Several insights were gained by the collective assessment of transcriptome and flux proteome datasets. Gene expression dictated most strongly the fractional synthesis of soluble collagens. Most dramatic in the early softening period is the marked transcriptional suppression of roughly 40% of genes as compared with the NP cervix and resulting global slowdown of protein synthesis rates. The noted decline in chromatin associated proteins identified in the soluble protein fraction at gestation day 6 suggests one mechanism by which rapid transcriptional suppression is achieved. The shift in transcriptional dynamics on gestation day 6 drives phenotypic changes in the epithelial cell population and suppression of chemokine driven immune signaling pathways. The downregulation of proinflammatory pathways is counterbalanced by the increase in transcriptional pathways that promote antiviral defense in response to interferon alpha and gamma, and the induction of epithelial specific genes that promote mucosal immunity. Collectively, these observations suggest a shift during pregnancy in the types of immunoprotective pathways driven by the immune and epithelial cell subpopulations.

The sample preparation for our proteome flux studies was designed to explore the various physical fractions of tissue proteins and how they may differ in flux rates during pregnancy. The SDS fraction reveals the behavior of soluble cellular proteins. Day 6 of pregnancy exhibited the greatest slow-down in FSRs across the global proteome, whereas days 12 and 15 reversed this change to prepregnancy replacement rates. When interrogating the dataset, it was interesting that histones almost universally reflected this pattern, perhaps indicating parallel changes in turnover of cells in the cervix. The guanidine-soluble fraction, on the other hand, which is enriched in ECM proteins including the more soluble collagen species [36, 49] revealed soluble collagen kinetic signatures for both fibrillar and nonfibrillar classes. Globally, within the proteome, we saw a similar pattern as in the SDS fraction, with coordinated slowing on gestation d6 that returns to basal levels in later pregnancy, but fibrillar collagens did not follow this pattern. Fibrillar collagens exhibited a much faster turnover rate that remained stable at all stages of pregnancy, whereas nonfibrillar collagens had a markedly slower turnover rate and resembled the general proteome pattern more closely. The guanidine-insoluble fraction represents the more insoluble proteins in a tissue. Globally, the general proteome pattern was repeated (early slowing with return to basal levels later) but fibrillar collagens had a much more reduced rate of replacement as compared with the soluble fibrillar collagens. In addition, nonfibrillar collagens in the insoluble fraction showed a more robust reflection of the global pattern in the proteome in that almost no new nonfibrillar collagen was synthesized at day 6 while a substantial increase in its turnover rate was observed on days 12–18.

Perhaps most striking, the fraction of newly synthesized OHP in the insoluble fraction was significantly increased from gestation day 12 to day 18 although total OHP content did not change for days 12 and 15 consistent with previous findings in cervical tissue [5, 14]. Because tissue OHP isotope labeling can be more sensitively detected by GC/MS compared with LC/MS–MS of individual collagen species and OHP content provides an accurate measure of tissue collagen content, we applied this approach for measuring turnover of the insoluble fraction as we have previously described [44–46]. The fraction of newly synthesized collagen in the insoluble fraction was plotted in comparison to prior measures of HP cross-links in cervical tissue [36, 44, 50]. The striking increase in OHP turnover (newly synthesized collagen) inversely correlated with the decline in HP cross-link density over time during pregnancy depicted in Figure 3C and [5]. This finding is consistent with the model that replacement of highly cross-linked collagens by newer collagen with reduced cross-links represents an underlying mechanism for increasing tissue compliance in cervical ripening.

Indeed, the rapid turnover of fibrillar collagens in the soluble fraction of nonpregnant cervical tissue is striking (t ½ of 2.1 days). Estimates of collagen half-life are derived from FSR values, with tissue comparisons are provided in Table 3. The magnitude of the turnover rate in the NP cervix is remarkably rapid when compared with turnover rates of fibrillar collagen in other adult tissues whose half-lives range between 13.9 and 267.1 days. These cervical protein turnover rates may be compared conceptually with substrate cycling in control [51, 52], of intermediary metabolism, whereby a tissue remains poised to adapt rapidly to physiologic changes by maintaining a high basal flux rate through both directions of a pathway. Here, the high rates of collagen synthesis and breakdown allow rapid replacement of old, highly cross-linked collagen molecules by newer, less cross-linked species as parturition approaches [51, 52]. Through pregnancy, the increase in cervical tissue compliance occurs rapidly over the course of a few days and appears to involve replacement of cross-linked insoluble collagen species by less cross-linked species. This replacement at constant collagen content requires that a reservoir of newly synthesized, less cross-linked collagens be available, which in turn is assured by the continuously high synthesis rate of soluble collagens in the cervix. This may represent a previously unrecognized form of “metabolic flexibility” for ECM in a tissue like the cervix that needs to undergo rapid transformation in compliance, in this case to allow parturition.

Table 3.

Mice (strain C57Bl/6J [cervix, liver, kidney, dermis, and lung] C57Bl/10 [ventricle, diaphragm, and old gastroc]; and C57Bl/N [for young gastroc]) were given ²H₂O ad libitum* for 2–30 days: 3 in [cervix]; 5 in [kidney]; 7 in [liver and lung]; 2–5 in [young gastroc]; 14 in [ventricle, diaphragm, and gastroc]; 30 in [liver, lung, kidney, and dermis] OHP measurement prior to terminal necropsy. Synthesis rates of OHP were assessed by gas chromatography–mass spectrometry. Synthesis rates of collagen, type 1, alpha 1 (Colα1) were measured by liquid chromatography–mass spectrometry. ND: not determined.

		t ½ (days)
Tissue	Age (weeks)	Soluble	Insoluble	OHP
Cervix	12–24	2.1	10.5	8.8
Liver	12	38	240.2	58.3
Kidney	25	13.9	171.5	83.7
Dermis	25	“ND”	“ND”	93.2
Lung*	25	41.6	78.4	104.8
Left ventricle	44	48.3	300.1	108
Diaphragm	44	32.7	198.7	110.1
Gastrocnemius	8	31.8	59.4	“ND”
Gastrocnemius	44	267.1	391.8	306.7

Integration of transcriptomic and proteomic flux studies also provide new evidence to illustrate dynamic changes in epithelial cell populations as illustrated by the marked change in FPKM values for the epithelial intermediate filament keratin transcripts and the decline in protein replacement rates for several keratin proteins (Figure 4). The pattern of changes in keratin transcription varied to include downregulation of keratins at all gestational time points (e.g., Krt1), downregulation in mid-to-late softening with a rebound in expression during ripening (e.g., Krt6a, Krt6b, Krt10, and Krt16), and gained expression in pregnancy (e.g., Krt8 and Krt18). Integration of keratin patterns with other epithelial cell markers reveals a pattern in which pathways were downregulated in early, mid or late softening yet began a gradual reversal during cervical ripening, by both proteome flux and gene expression analyses. This includes a modest reversal of the downregulation of genes encoding structural proteins characteristic of keratinocytes. One specific epithelial cell population that appeared to increase in the softening period was epithelial goblet cells. These cells secrete mucins and antimicrobial peptides to protect the underlying epithelium and are identified in the human and mouse cervix [53, 54]. In the colon, goblet cells are critical for maintenance of the colonic barrier and mucosal immunity and goblet cell malfunction is a key driver of inflammatory bowel disease [55]. Translating these findings to the cervical epithelia, we suggest in early pregnancy the implementation of transcriptional programs that ensure antiviral defense and the establishment of mucosal immunity and barrier function properties. These protective mechanisms set the stage to prevent ascending infection, an established risk factor for premature birth [56, 57]. Functional interrogation of the unique pregnancy-specific epithelial populations will provide insight into mechanisms by which loss of cervical epithelial mucosal barrier integrity may predispose to ascending infection mediated preterm birth in women.

The induction of protective measures in early softening is followed in mid-to-late softening with activation of a transcriptional program to allow progressive ECM remodeling is suggested by GO assessments. Both RNAseq and proteomic dataset identified ECM targets such as FACIT collagens, matricellular proteins (POSTN and DPT), and proteoglycans (OGN) that have not been previously identified as contributors to ECM function in cervical remodeling. Taken together, the RNAseq and proteomic datasets identify expression of ECM components that promote collagen fibril assembly, as well as promote changes in structure and function of the ECM in pregnancy. The collagen peptide binding studies serve as an indirect measure of collagen degradation and support the continual and robust turnover of collagens in the nonpregnant and pregnant cervix. In addition, this opens the door to future studies that can expand on the identification of noninflammatory induced intracellular/extracellular collagen degradation pathways that allow for rapid turnover and homeostasis [58, 59]. Indeed, novel pathways of collagen remodeling may allow for the dynamic turnover of ECM in the nonpregnant cervix and during pregnancy.

In summary, these data collectively characterize the continuous process of cervical remodeling that begins in early pregnancy and encompasses both transcription and protein turnover events, with the dual purpose to protect the cervix for pregnancy maintenance and to prepare for parturition. In particular, the continuously high fibrillar collagen synthesis rates in the nonpregnant and pregnant cervix that we report here allow rapid replacement of highly cross-linked collagens by newer, less highly cross-linked collagens in late cervical ripening. We propose that this protein turnover of metabolic flexibility may be a key mechanism to ensure the rapidly increase compliance that allows for birth. Integration of mRNA and protein datasets along with consideration of progressive changes from early, mid-to-late softening, and cervical ripening has expanded our understanding of the diverse and changing pathways that must be balanced to achieve on time and sufficient remodeling of the cervix at birth. This understanding may, ultimately provide insights into how perturbations in these pathways compromise cervical functions and contributes to preterm birth.

Experimental procedures

Mice

A breeding colony of mice strain C57BL6/129sv was maintained at University of Texas Southwestern Medical Center (UTSW) and used in described studies. For timed pregnancies, breeding pairs were set up in the morning for 6 h. The presence of a vaginal plug at the end of the 6 h period was considered day 0 of pregnancy. The birth of pups generally occurred in the early morning on day 19. All animal procedures were performed in accordance with the standards of humane animal care following the NIH Guide for the Care and Use of Laboratory Animals. The research protocols were reviewed and approved by the Institutional Animal Care and Use Committee at the University of Texas Southwestern Medical Center.

RNASeq and data analysis

Total RNA was isolated using RNeasy Micro Kit from Qiagen (Hilden, Germany). RNA Quantity was determined using a Nanodrop and quality was determined using Agilent Tapestation. A total of 10 cDNA libraries (duplicate of five groups) were prepared from NP, gestation days 6, 12, 15, and day 18 cervix samples. The NP cervices were collected at metestrus. RNA (3 μg) from three different samples was pooled for 9 μg of starting RNA in each library. Library preparation was carried out as described previously [60]. Single end sequencing was carried out on an Illumina sequencer to a depth of 400 million reads. The raw files were subjected to QC analyses using the FastQC tool. The reads were mapped to the mouse reference genome (mm10) using the spliced read aligner TopHat version 2.0.12 [61]. Default parameters were applied for Transcriptome assembly using cufflinks version 2.2.1 [62]. Cuffmerge and Cuffdiff were used for generating distinct transcripts and calling the differentially regulated transcripts. The significantly (P < 0.05) regulated genes upon different conditions were used.

In vivo heavy water labeling for proteomics

Five groups of mice (nonpregnant [n = 8], gestation days 3 [n = 7], 6 [n = 7], 12 [n = 10], 15 [n = 7], and 18 [n = 6]) were administered deuterium-labeled drinking water (²H₂O) as described previously [39] and cervical tissue was collected after 3 days at which time the mice were NP, or at gestation days 6, 12, 15, or 18 (Supplementary Figure 2). Briefly, mice were given a bolus injection of 35 ml/kg 100% ²H₂O in two divided doses at a 30-min interval followed by administration of 8% ²H₂O in drinking water until tissue collection 3 days later. This labeling protocol maintains body water enrichment constant at ~5% [44, 63, 64]. The cervix and plasma were collected after 3 days of ²H₂O exposure in all animals.

Cervix tissue preparation for proteomics

Mouse cervix (n = 5–10 cervices/time point) was fractionated as previously described [36, 65]. In summary, whole cervix was resuspended in 0.08% SDS (20:1 volume:weight) and vortexed at low speed for 16 h to extract SDS-soluble proteins. The supernatant was transferred to a low-binding microcentrifuge tube (Corning, Tewksbury, MA), spun at 16 000 × g for 30 min at RT to clear any insoluble proteins, and stored at −80°C until further analysis. Remaining tissues were rinsed with ddH₂O and resuspended in 4 M guanidine-HCl, 50 mM sodium acetate, pH 5.8, at a 10:1 volume:mass ratio; and vortexed at high speed for 48 h. The supernatant was then removed, cleared of insoluble proteins, and stored as described above. Residual guanidine-insoluble tissue was rinsed three times with ddH₂O and stored at −80°C. All extraction buffers and rinses were spiked with protease inhibitor mixture at manufacturer-recommended concentrations (Millipore, Billerica, MA). Soluble proteins from extracted fractions were precipitated by overnight incubation at −20°C in ethanol (5:1 ethanol:extraction buffer) followed by centrifugation at 16 000 × g for 45 min. Pelleted proteins were rinsed twice with 90% ethanol, allowed to air dry, and resuspended in 8 M urea prior to trypsin digestion. Insoluble proteins were digested directly from residual guanidine-insoluble tissue fragments.

LC/MS preparation

Protein concentrations in extractable fractions were determined via NanoDrop A280 measurement (Thermo, Waltham, MA). Up to 80 μg of each fractionated protein sample was denatured using ProteaseMax surfactant (0.1%; Promega, Madison, WI) and 4 M urea in 25 mM ammonium bicarbonate (pH 8). Proteins were reduced with TCEP (5 mM) for 20 min at room temperature with vortexing and then incubated with iodoacetamide (10 mM) in the dark for 20 min to chemically modify reduced cysteines. Proteins were then digested with trypsin (Promega) at 37°C overnight using a 1:25 trypsin:protein mass ratio. Guanidine-insoluble protein fractions were processed in an identical manner using a volume of trypsin sufficient for 80 μg of protein. The following day, formic acid was added to a total concentration of 5%, and samples were centrifuged at 14 000 × g for 30 min. The supernatant was transferred to a fresh tube, desalted with a C18 spec tip (Varian, Palo Alto, CA), dried via vacuum centrifugation, and resuspended in 0.1% formic acid/3% acetonitrile prior to LC/MS analysis.

Determination of the fractional synthesis of proteins (FSR) was done as previously described [36, 66, 67]. The isotopic labeling pattern of newly synthesized protein molecules was first determined, and then the isotopic incorporation in the mixture of molecules was compared in order to establish the dilution by unlabeled molecules. The peptide elemental composition and curve-fit parameters were calculated to determine peptide isotope enrichment in newly synthesized proteins during the period of heavy water exposure. This was based on precursor body water enrichment and the number of amino acid C–H positions per peptide actively incorporating hydrogen and deuterium from body water. FSRs were calculated as rate constants for proteins with fractional synthesis values (also known as f-values) by fitting to a mono-exponential rise-to-plateau curve [FSR = −ln(1 − f)/t]. The half-life (t_1/2) was calculated using the rate constant (k) calculated from curve fitting fractional synthesis values (f) vs. time (t) according to a single pool, monoexponential model: f = 1 − e^(−kt), so that k = [−ln(1 − f)]/t, where t_1/2 = ln(2)/k.

Plasma ²H₂O measurement

²H₂O enrichment of body water for each animal (n = 5–10 cervices/time point) was quantified using 100 μL of plasma as previously described [36]. Briefly, body water was evaporated from plasma via overnight incubation at 80°C. Samples were mixed in 10 M NaOH and acetone and underwent a second overnight incubation. This material was extracted in hexane and dried with Na₂SO₄ prior to gas chromatographic/mass spectrometric analysis alongside a standard curve of samples prepared at known ²H₂O concentrations.

Collagen hybridization peptide staining

Cervical OCT-embedded 8-micron frozen tissue sections from two different animals per time point were used. The OCT compound was removed by warming the slides to room temperature and rinsing three times with PBS for 5 min. The sections were blocked with 5% goat serum diluted in PBS for 20 min at room temperature. The trimeric F-CHP (20 uM, 3-Helix Cat. No. FLU300) was heated at 80°C for 5 min to mediate dissociation into single strands followed by incubation on ice for 30 s prior to placement atop the sections. For the positive control slide, the sections were denatured using 80°C hot PBS for 5 min. For the negative control slide, the sections were incubated with the unheated (trimeric) CHP diluted in PBS. The slides were incubated in a humidity chamber at 4°C overnight. The slides were washed in 100 mL of PBS at room temperature for 3 cycles of 5 min. Sections were mounted with ProLong Gold-containing 40,6-diamidino-2-phenylindole (Life Technologies, Carlsbad, CA) to counterstain the nuclei. The cervical sections were imaged using the Zeiss LSM880 LSM confocal microscope at 20X or 40X objectives. Fluorescence from 5-FAM/FITC was used to detect CHP binding to denatured collagen.

Statistical analysis

Heatmap datasets were generated by the average of the FSR for each group (NP, d6, d12, d15, and d18). FSR values were then compared between NP and each pregnancy time point to quantify the number of proteins whose FSR was either increased (faster) or decreased (slower). A binomial distribution comparison was calculated for the upper-tail of the distribution by a two-tailed probability test for significance (P < 0.05). Bar graphs were generated as the mean of the FSR for each group (n = 5–10). One-way analysis of variance with post-hoc correction for multiple comparisons was performed to determine which group protein FSR means differed compared with NP. Differences were considered significant at corrected P < 0.05. Bars display standard errors.

Data availability

The RNAseq datasets were submitted to GEO with the series number GSE164059.

Conflicts of interest

The authors declare that they have no conflicts of interest with the contents of the article.