|Home | About | Journals | Submit | Contact Us | Français|
A greater understanding of the parturition process is essential in the prevention of preterm birth, which occurs in 12.7% of infants born in the United States annually. Cervical remodeling is a critical component of this process. Beginning early in pregnancy, remodeling requires cumulative, progressive changes in the cervical extracellular matrix (ECM) that result in reorganization of collagen fibril structure with a gradual loss of tensile strength. In the current study, we undertook a detailed biochemical analysis of factors in the cervix that modulate collagen structure during early mouse pregnancy, including expression of proteins involved in processing of procollagen, assembly of collagen fibrils, cross-link formation, and deposition of collagen in the ECM. Changes in these factors correlated with changes in the types of collagen cross-links formed and packing of collagen fibrils as measured by electron microscopy. Early in pregnancy there is a decline in expression of two matricellular proteins, thrombospondin 2 and tenascin C, as well as a decline in expression of lysyl hydroxylase, which is involved in cross-link formation. These changes are accompanied by a decline in both HP and LP cross-links by gestation Days 12 and 14, respectively, as well as a progressive increase in collagen fibril diameter. In contrast, collagen abundance remains constant over the course of pregnancy. We conclude that early changes in tensile strength during cervical softening result in part from changes in the number and type of collagen cross-links and are associated with a decline in expression of two matricellular proteins thrombospondin 2 and tenascin C.
The composition and structure of the cervical extracellular matrix (ECM) regulates the ability of the cervix to remain closed and firm during pregnancy and to open and distend at the time of parturition. Greater understanding of the molecular changes within the cervical ECM during pregnancy and parturition will enhance our understanding of this critical physiological process as well as identify mechanisms underlying aberrant remodeling that accompany premature birth. The cervical ECM is secreted by the epithelia and fibroblasts within the cervical tissue and is comprised of five major components: fibrillar collagens, elastin fibers, proteoglycans, hyaluronan, and matricellular proteins [1, 2]. Early biochemical modifications result in palpable changes in tissue compliance by the first trimester of pregnancy in women. In mouse studies, quantifiable changes in tissue biomechanics are detectable by mid-pregnancy [2, 3]. This early phase of tissue remodeling, termed cervical softening, is characterized by an increased percentage of soluble collagen and quantifiable changes in collagen fiber microstructure while collagen content remains constant [3, 4]. Molecular processes that account for changes in collagen solubility and microstructure during cervical softening are not well characterized and are the focus of this study.
Fibrillar collagen types I and III are the main structural proteins of the cervix . While the collagen family includes both fibrillar and nonfibrillar collagens, fibrillar collagens are the primary source of the tensile strength of tissue . The load-bearing capacity of a tissue is in part determined by collagen type and abundance, posttranslational processing of collagen, assembly of collagen into fibrils and fibers, deposition of collagen in the ECM, and collagen degradation . Collagen synthesis is complex. Fibrillar collagen is synthesized in the endoplasmic reticulum (ER) in a pro-form that is folded and assembled into a triple helix, initiated at the C terminus, with the aid of chaperone proteins [7, 8]. While still in the ER, lysine residues in the procollagen chain are hydroxylated by the enzyme lysyl hydroxylase. Once the procollagen trimer is secreted into the extracellular space, the N- and C-terminal propeptides are cleaved, and collagen spontaneously aggregates into fibrils, which subsequently assemble into fibers [9, 10]. Assembly of fibrils from mature collagen molecules involves intermolecular cross-links between lysine residues on adjacent collagen molecules. Some lysine residues are hydroxylated in the ER by the enzyme lysyl hydroxylase, and the degree of hydroxylation determines the type of cross-link formed. In turn, the type of cross-link formed determines strength and mechanical stability of the resulting tissue [11, 12]. Collagen cross-linking is catalyzed by lysyl oxidase (LOX) and occurs between hydroxylated or nonhydroxylated lysine residues in collagen, resulting in pyridinoline or nonpyridinoline cross-links, respectively.
The importance of regulated changes in collagen processing during parturition is supported by numerous reports that preterm birth due to cervical insufficiency and preterm premature rupture of membranes are increased in women with inherited defects in collagen and elastin synthesis or assembly (e.g., Ehlers-Danlos and Marfan syndromes) . Recent studies suggest polymorphisms associated with genes important for connective tissue synthesis and metabolism may predispose women to preterm birth due to cervical incompetence and preterm premature rupture of membranes . Previous studies report that activity of the cross-link-forming enzyme, LOX, is reduced in the cervix during mouse pregnancy, and Lox gene expression is regulated in the pregnant mouse cervix [14, 15]. These data suggest a decline in the number of collagen cross-links in the mouse cervix over the course of pregnancy.
In addition to collagen synthesis and cross-link formation, the type of fibrillar collagen can affect ECM composition and strength. Studies in collagen III-deficient mice indicate that collagen III is essential for normal collagen I fibrillogenesis and that changes in the ratio of collagen I:III can alter the mechanical properties of tissue [16, 17]. The ratio of type I-to-III collagen in the human nonpregnant cervix is reported to be 70% and 30%, respectively, but it has not been determined if changes in this ratio contribute to changes in cervical tissue compliance during pregnancy .
ECM architecture and overall tissue strength are determined by a network of interactions among collagens, proteoglycans, and matricellular proteins. Noncollagenous proteins in the cervical ECM, such as proteoglycans and matricellular proteins, can affect matrix organization and consequently tissue strength. Proteoglycans organize the ECM by controlling the size and packing of collagen fibrils . The glycosaminoglycan chains that attach to the protein core can affect the hydration or equilibrium tension of the tissue. Several proteoglycans are known to be expressed in the pregnant cervix, including versican, decorin, biglycan, asporin, and fibromodulin [3, 20–23]. Matricellular proteins, such as tenascins, thrombospondins, and the secreted protein acidic and rich in cysteine (SPARC) protein, are not structural proteins within the ECM, but they modulate the functions of structural proteins such as collagen as well as cell functions through interactions with cell surface receptors, proteases, and growth factors . Targeted gene loss of specific matricellular proteins results in aberrant matrix organization and remodeling [25–27]. In this study, we assess changes in relative abundance of fibril collagens I and III, collagen processing, collagen cross-links, proteoglycans, matricellular proteins, and collagen ultrastructure in order to identify early molecular events leading to increased cervical compliance and collagen solubility required for successful birth.
To assess stage of cycle, vaginal smears were taken and analyzed for characteristic cell structure and immune cell presence specific for each phase of the cycle . Mice were observed through at least one full cycle in order to insure proper cycling prior to tissue collection. In this study the term “NP” refers to data pooled from an equal number of cervices collected in proestrus, estrus, and metestrus. “NP estrus” or “NP metestrus” refers to cervices assessed at a single stage of the estrus cycle.
Females were housed overnight with males and separated the following morning. Vaginal plugs were checked at the time of separation; mice with plugs were considered to be at Day 0 of their pregnancy, with birth occurring early on gestation Day 19. In general, cervices were collected at noon for all time points from gestation Day 8–18, except for Day 18, the day on which cervices were collected between 1800 and 1900 h in order to collect cervices a few hours prior to onset of labor. Postpartum cervices collected after vaginal birth on gestation Day 19 were obtained exactly 2 or 4 h after delivery of the first pup or approximately 10–12 h, 24 h, or 48 h after delivery of the first pup.
Mice used for these studies were of C57B6/129Sv mixed strain. Cervical tissue was dissected from the reproductive tract, and vaginal tissue was carefully removed from cervical tissue. Cervices were flash-frozen in liquid nitrogen immediately following their extraction. All studies were conducted in accordance with the standards of humane animal care described in the National Institutes of Health Guide for the Care and Use of Laboratory Animals, using protocols approved by an institutional animal care and research advisory committee.
Total RNA was extracted from frozen cervical tissue in RNA Stat 60 (Tel-Test Inc., Friendswood, TX). RNA was then treated with DNase (New England Biolabs, lpswich, MA) to remove genomic contaminants. cDNA was synthesized, and RT-PCR was carried out using the SYBR Green detection system (Applied BioSystems, Carlsbad, CA) or using the TaqMan Probes system (Applied BioSystems). All gene expression is relative to gestation Day 18 levels.
Cervical tissue was collected on Days 6, 12, 15, and 18 of gestation and frozen in OCT embedding compound medium (Sakura Finetek USA, Torrance, CA). Five-micrometer cervical sections were cut from tissue blocks. Sections were fixed for 10 min in acetone at −20°C. Tissue was rehydrated in PBS, blocked in 10% normal goat serum (catalog no. 01–6201; Invitrogen, Carlsbad, CA), and incubated with collagen I or collagen III rabbit polyclonal antibodies (codes ab34710 and ab7778; Abcam, Cambridge, UK). Sections were then washed in PBS and incubated with goat α-rabbit immunoglobulin G antibodies coupled to Alexa 488 (product no. A11008; Molecular Probes, Invitrogen). Slides were viewed on a Zeiss LSM510 Meta NLO confocal microscope using an Acroplan 40×/0.8 W objective lens. Fluorescence signal intensity was measured using ImageJ 1.41k software (http://rsbweb.nih.gov/ij/).
Collagen was extracted with 7 M urea (Sigma, St. Louis, MO), 0.1 M sodium phosphate with 1% protease inhibitor (Sigma), overnight at 4°C. The protein concentration was determined by a Bradford protein assay (Pierce; Thermo Scientific, Rockford, IL). Ten micrograms of protein were loaded on a 4%–20% Tris-HCl polyacrylamide gel and electrophoresed at 100 V. After overnight transfer to nitrocellulose membrane and Ponceu S (Sigma) staining to assess equal loading of protein, immunoblotting was performed using rabbit polyclonal anti-mouse collagen I (catalog no. MD20151; MD Biosciences, St. Paul, MN).
Protein used in dot blot analysis was extracted as described for immunoblotting. Two and one-half micrograms of protein were spotted onto a nitrocellulose membrane and probed with rabbit polyclonal anti-collagen I or rabbit polyclonal anti-collagen III primary antibody (codes ab34710 and ab7778; Abcam). Secondary antibody for both Western and dot blots was donkey anti-rabbit horseradish peroxidase (catalog no. 711036152; Jackson ImmunoResearch, Westgrove, PA). Chemiluminescence was visualized using ECL (GE Healthcare, Buchinghamshire, UK). Digital images of the blots were analyzed quantitatively using Multi Gauge software (Fuji Film, Tokyo, Japan).
Frozen cervical tissue was lyophilized, weighed, and then hydrolyzed in 6 M HCl at 100°C for 20–22 h. Collagen cross-links were measured via HPLC as described previously [29, 30]. In brief, cervical samples were lyophilized and then hydrolyzed in 6 M HCl at 100°C for 20–22 h. The samples were then dried and dissolved into 10 μM pyridoxine and 2.40 mM homo-arginine in water. Samples were then diluted 1:4 in 0.5% heptafluorobutyric acid in 10% acetonitrile. Cross-links were analyzed by reversed-phase HPLC and calculated based on internal pyridoxine standards. Hydroxyproline assays were carried out as previously described . Total collagen was calculated using a ratio of 300 hydroxyproline residues per triple helix .
Nonpregnant mice in metestrus or pregnant mice at gestation Days 6, 12, 15, and 18 were perfused with 1% glutaraldehyde, 2% paraformaldehyde fixative in 0.1 M phosphate buffer. Cervical and uterine tissue was removed and fixed in 2.5% glutaraldehyde in 0.1 M cacodylate buffer containing 0.1% ruthenium red overnight at 4°C. Tissue was then cut into 250 μm longitudinal sections using a Vibratome Series 3000 sectioning system (Vibratome Co., St. Louis, MO). Vaginal and uterine tissue was removed from the cervix, and the cervix was rinsed for 30 min with cacodylate buffer. The tissue was postfixed with 1% osmium in 0.1 M cacodylate buffer containing 0.1% ruthenium red for 90 min. Specimens were dehydrated through a graded series of ethanols (50%, 70%, 95%, and 100%) followed by propylene oxide and embedded in Epon (EMbed-812; Electron Microscopy Sciences, Hatfield, PA). Thin sections were stained with uranyl acetate and lead citrate. Sections were viewed by transmission electron microscopy (TEM) with a Tecnai microscope (FEI Company, Hillsboro, OR), and images were captured at 4200× or 20500× magnification with a Morada 11 1-megapixel charge-coupled device camera.
Electron microscope images of transversely sectioned collagen fibrils taken at 20500× magnification were analyzed with Image J 1.41k software (http://rsb.info.nih.gov/ij/). An intensity threshold was interactively determined for optimal segmentation of fibril from background. Segmented images were converted to a binary mask, and erode, open, dilation, and fill holes binary operations were used to separate merged fibrils. Particles were analyzed using the Particle Analysis function of Image J, with parameters set for size, 1000–10000 nm2, and circularity, 0–1. Ellipses were fit to all particles, and the minor angle of the ellipse was taken as the fibril diameter. Ten images from three animals were analyzed for each time point, resulting in n = 8197 fibrils/Day 6, 12958 fibrils/Day 12, 8362 fibrils/Day 15, and 12161 fibrils/Day 18. Due to low contrast ratio of fibrils versus background in NP samples, 516 fibrils were taken from three animals, one image each.
Statistics were performed using Prism software version 5.0b (GraphPad Software, La Jolla, CA). For comparison of QPCR, collagen content, and collagen cross-links data, one-way ANOVA was used, followed by a Dunnet comparison test using NP or Day 8 as the control. Fibril diameter data were analyzed using a one-way ANOVA followed by a Dunn multiple comparison test.
To observe whether changes in the relative amounts of type I and III fibril collagens could account for the progressive decline in tissue compliance through gestation, mRNA levels for the alpha-1 chain of collagen I and collagen III were determined by QPCR. Collagen I alpha-1 mRNA levels in the first half of pregnancy were similar to NP but significantly increased in the latter days of gestation (P < 0.01). Levels declined to NP levels by 24 h postpartum (Fig. 1A). Collagen III alpha-1 mRNA levels were similar to NP throughout gestation. While there was a transient and significant increase in mRNA levels 2 h postpartum, levels quickly declined 24 h after delivery (Fig. 1B). Total cervical collagen protein was estimated by hydroxyproline assay of acid-hydrolyzed tissue. Collagen levels throughout pregnancy were found to be at similar or elevated levels compared to NP estrus cervix (Fig. 1C). Collagen levels were increased significantly compared to NP estrus cervix on gestation Days 5, 8, 16, and 18 and 24 h postpartum when normalized to dry weight (P < 0.0001) (Fig. 1C).
Two methods were used to estimate the relative abundance and ratio of extractable type I and type III fibril collagen. Dot blots were performed on cervical tissue homogenates extracted in 7 M urea and spotted on a nitrocellulose membrane at equal protein loads (Fig. 1D). Conventional Western immunoblotting could not be used for this assay because the antibodies available to distinguish type I from type III do not work on denatured samples. Extractable collagen I increased approximately twofold from NP estrus to early pregnancy (Day 6), and levels remained constant for the remainder of pregnancy (Fig. 1E), while there were no significant differences in the extractable collagen III during gestation compared to NP estrus (Fig. 1F). Immunofluorescent staining for collagen I and III on cervical tissue showed no change in signal intensity throughout gestation. Overall signal intensity for collagen I was greater than collagen III during pregnancy. The collagen I:collagen III ratio was a factor of 2.4 at all time points (Fig. 1, G and H). Taken together, these results indicate that relative to the nonpregnant cervix, collagen levels are constant or elevated during pregnancy and that the relative abundance of fibrillar type I-to-III collagen remains unchanged throughout pregnancy.
Intracellular trafficking and folding of procollagen chains require interaction with ER chaperone proteins, such as heat shock protein 47 (Hsp47, official symbol SERPINH1) and protein disulfide isomerase (PDI), as targeted deletion of these genes results in loss of tissue collagen . Consistent with constant or elevated collagen content through pregnancy (as seen in Fig. 1), we observed that protein expression of SERPINH1 and PDI in pregnant and postpartum cervix appears similar or slightly elevated during pregnancy (Supplemental Fig. S1, available online at www.biolreprod.org).
Once secreted from the cell, C- or N-terminal propeptides are cleaved to form mature collagen. Collagen molecules in which the C-propeptide is not cleaved are unable to form fibrils, while the inability to cleave the N-propeptide results in formation of irregularly shaped fibrils [33, 34]. Both scenarios result in collagen with reduced tensile strength. To examine the possibility that C- or N-terminal propeptide processing may be altered during pregnancy, we determined mRNA levels of enzymes responsible for cleavage of N- and C-propeptides from procollagen. The expression of genes encoding bone morphogenetic protein-1 (Bmp1) and tolloid-like 1 (Tll1), which cleave the C-propeptide from the procollagen molecule, was analyzed via QPCR. mRNA levels for Bmp1 expression were abundant and constant throughout pregnancy, with small but significant elevations observed on gestation Day 18 and during postpartum. There was a trend for reduced Tll1 expression levels during pregnancy, though not all time points achieved significance (Fig. 2, A and B). The procollagen C-endopeptidase enhancer family (PCOLCE1 and −2), reported to enhance the C-proteinase activity of TLL1 and BMP1, was also evaluated, and mRNA levels of Pcolce1 and -2 remained constant throughout gestation compared to NP (data not shown). Despite the decrease in Tll1 expression, immunoblots showed increased abundance of the 30- kDa C-propeptide during pregnancy compared to the NP cervix, suggesting C-terminal processing occurs normally (Fig. 2C).
Three reported ADAMTS proteases (a disintegrin-like and metalloprotease domain with thrombospondin type I motifs), ADAMTS2, ADAMTS3, and ADAMTS14, cleave the N-propeptide from the procollagen molecule. Using standard PCR, Adamts2 and -14 were found to be present in the cervix, while Adamts3 was undetectable (data not shown). QPCR analysis of Adamts2 revealed stable expression throughout pregnancy and postpartum (Fig. 2D). Adamts14 expression increased significantly on Day 10 of gestation and remained elevated through Day 15 (P < 0.01). Levels dropped on Day 18 and remained at NP levels throughout postpartum period (Fig. 2E). While an antibody that recognizes the N-propeptide of mouse collagen is available, we were unable to visualize N-propeptide in our system.
Changes in the type or number of collagen cross-links can affect the tensile strength and solubility of collagen. Hydroxylation of lysine residues by the enzyme lysyl hydroxylase regulates the type of collagen cross-links formed. Expression of three genes that encode lysyl hydroxylase (Plod1, -2, and -3) was measured by QPCR (Fig. 3A). No appreciable change was seen in Plod1 or Plod3 mRNA expression during gestation. In contrast, compared to the NP, Plod2 mRNA levels appeared suppressed throughout gestation and reached significance on gestation Days 9–13. Levels increased twofold at 2 h postpartum and returned to NP levels by 24 h postpartum.
Previous reports of reduced activity of the cross-link-forming enzyme LOX along with our observation that Plod2 is suppressed significantly led us to evaluate the type and degree of lysine cross-links during gestation. Lysylpyridinolines (LP) cross-links derived from one lysine and two hydroxylysines and hydroxylysylpyridinolines (HP) cross-links derived from three hydroxylysines are the stronger pyridinoline cross-links that predominate in rigid connective tissues such as bone, cartilage, and tendon. LP and HP cross-links are relatively low in the more flexible connective tissues such as skin, where cross-links between lysine predominate . Thus, a reduction in LP and HP cross-linking could contribute to increased compliance of the cervix during ripening.
HP and LP cross-links in cervical collagen at time points throughout pregnancy were measured by HPLC. HP cross-links declined significantly from NP estrus levels on Day 12 of gestation and remained low throughout gestation and during postpartum repair (Fig. 3B). LP cross-links also declined by Day 14 of gestation and remained low until Day 18, increasing after birth (Fig. 3C).
The progressive decline in LP and HP cross-links as birth approaches likely contributes toward increased solubility of collagen. Protein was extracted in 7 M urea from NP, gestation Days 6 and 8–18, and 4-, 12-, 24-, and 48-h postpartum cervices, and immunoblotting of equal protein loads were used to detect collagen I. Because cross-linked collagen is not readily soluble, these blots cannot assess changes in total collagen but rather give an estimate of extractable collagen I. Three bands were visible on the Western blot: a broad band at ~250 kDa that we attribute to procollagen dimers and trimers, a mature collagen monomer band at 140 kDa, and an unidentified band at 50 kDa that appears in NP and postpartum samples (Fig. 3D). The 140-kDa mature collagen band is not visible in NP samples. During pregnancy this band increased in density, suggesting that mature collagen is more extractable (soluble) during pregnancy. The band declines in density by 4 h postpartum and is not seen in later postpartum samples. Increased extractability of collagen during pregnancy is consistent with our previous observations of increased collagen solubility as determined by extraction in acetic acid and pepsin and with current data showing a decline in HP and LP collagen cross-links . The 250-kDa band appeared by Day 11 of pregnancy and remained visible throughout gestation but disappeared postpartum. The presence of this procollagen band suggests an increase in the proportion of newly synthesized but unprocessed collagen.
Given the ability of small leucine-rich proteoglycans to modulate collagen structure, the mRNA expression of decorin, biglycan, fibromodulin, asporin, and osteomodulin was quantified by QPCR (Fig. 4). Decorin, biglycan, and osteomodulin mRNA levels showed no significant changes throughout pregnancy (Fig. 4, A–C). Asporin levels remain constant throughout pregnancy, with a transient upregulation 2 h postpartum (Fig. 4D). Fibromodulin levels remained constant throughout gestation and were significantly downregulated postpartum (Fig. 4E).
Expression of mRNAs encoding the matricellular proteins Thbs2, Tnc, and Sparc, were obtained via QPCR (Fig. 5). Thbs2 mRNA expression dropped from NP levels by Day 8 of gestation and remained low until 2 h after birth, returning to NP levels by 24 h postpartum (Fig. 5A). The decline in expression on Days 8, 10, and 11 did not achieve significance. Similar to Thbs2 expression, that of tenascin C mRNA declined significantly on Day 8 of gestation and remained downregulated for the remainder of pregnancy. Tnc mRNA increased sixfold by 2 h postpartum and returned to NP levels by 24 h postpartum (Fig. 5B). In contrast, there was a trend for increased Sparc expression from gestation Day 10 through Day 17, although significance was only achieved at gestation Days 14 and 17. Relative expression returned to NP levels on Day 18 of gestation throughout postpartum (Fig. 5C).
The observed changes in collagen solubility, type, and degree of cross-links as well as matricellular protein composition can affect collagen organization and ultrastructure. Electron micrographs of cervical tissue at Days 6 and 18 were taken at 4200× magnification (Fig. 6). Cellular components and electron-dense components of the ECM appeared to be in close proximity during early pregnancy (Fig. 6A). In late gestation, collagen fibers were more dispersed and not associated with cellular components of the tissue (Fig. 6B), and generally there was a dramatic loss of electron-dense ECM. To evaluate changes in collagen ultrastructure, collagen fibril diameters were measured from TEMs of cervical tissue from NP metestrus and gestation Days 6, 12, 15, and 18 as described in Materials and Methods (Fig. 7A–E). The mean fibril diameter significantly increased from early to late pregnancy along with a shift in the distribution toward a higher frequency of fibrils with a larger diameter. As early as Day 6, there was an increase in fibril diameter of 12.1 nm compared to NP metestrus, with an additional 20-nm increase from Day 6 to Day 18 (Fig. 7, F–J).
This study identifies changes in the cervical ECM early in pregnancy that contribute cumulatively to the progressive decline in tensile strength during the cervical softening phase. Most notable are changes in the degree and types of collagen cross-links and a decline in expression of two matricellular proteins, thrombospondin 2 and tenascin C. The decline in collagen cross-links, as well as decline in expression of THBS2 and TNC, must independently or cumulatively result in an increased solubility of collagen during pregnancy as well as the striking changes in collagen fibril ultrastructure.
Based on these observations, we propose a model in which mature collagen in the cervical ECM is gradually replaced with less cross-linked collagen beginning early in gestation. As the less cross-linked collagen becomes more abundant in the ECM, tissue compliance begins to increase, eventually reaching a threshold of measurable change. This threshold defines the beginning of cervical softening, which, in the mouse, is measurable by gestation Day 12, and tissue stiffness declines progressively thereafter . The decline in LP and HP cross-links correlates with the decline in mRNA expression of Plod2 as well as the reported decline in activity of LOX in the mouse cervix . While the other two lysyl hydroxylase genes (Plod1 and Plod3) were expressed at normal levels in the mouse cervix, they did not compensate for the decline in Plod2 expression. It has been suggested that PLOD2 may have greater specificity for hydroxylation of the telopeptide region of collagen, important in formation of pyridinoline cross-links, while the other lysyl hydroxylases may preferentially hydroxylate lysine residues in the triple helical regions of various collagens . Mutations in the Plod2 gene and a reduction in pyridinoline cross-links have been described in patients with Bruck syndrome, which is characterized by fragile bones, scoliosis, and osteoporosis [36, 37].
In addition to a decline in collagen cross-linking, altered levels of matricellular proteins likely contribute to early changes in tissue compliance as well. In mice, THBS2 appears to regulate cell-matrix adhesion, inhibit angiogenesis, and regulate collagen fibril assembly [25, 38]. Loss of THBS2 results in defective cell adhesion of fibroblasts and increased collagen solubility in skin, and collagen fibril size is increased while tensile strength is decreased. Wound healing is accelerated in these mice, along with increased tissue vascularization . In addition, THBS2 null mice have a reported acceleration of cervical softening without premature birth . Both THBS2 and tenascin C play an important function in promoting cell migration during wound healing after injury and, consistent with this function, are upregulated severalfold in the cervix at the time of birth and postpartum . Future studies are required to understand the mechanisms by which THBS2 and tenascin C contribute to collagen fibril assembly as well as to a potential role in tissue vascularization, which is increased in late pregnancy .
The early, progressive and cumulative changes in the cervical ECM are supported by the observed increase in collagen fibril diameter as measured in electron micrographs. Increased fibril diameter between gestation Days 12, 15, and 18 coincide with biomechanical changes that occur in late pregnancy (Fig. 7) [43, 44]. These changes also correlate with quantifiable changes in cervical collagen fiber morphology, as determined by second harmonic generation microscopy at the same time points in mouse pregnancy . The increasing fibril diameter may result from reduced packing of fibrils due to both the decline in HP and LP cross-links as well as the decline in THBS2 and/or tenascin C expression. The net result is a loss of tensile strength. This is supported by the observation that a reduction in pyridinoline cross-links, or loss of THBS2 in knockout mouse models leads to an increase in collagen fibril diameter as measured by electron micrographs [25, 45].
A number of genes/proteins important in synthesis, trafficking, or processing of collagen were expressed at levels similar to or slightly more elevated than those in nonpregnant cervix (Figs. 2 and and3).3). The resulting continued production of newly synthesized collagen might ensure that collagen is maintained at a constant level, yet it allows for a gradual turnover of well-cross-linked collagen with poorly cross-linked collagen. Consistent with this hypothesis is the observation that Sparc transcripts (Fig. 5B) and protein (data not shown) are elevated during pregnancy and decline to NP levels by gestation Day 18. Sparc expression is frequently associated with tissues in which there is a high rate of collagen turnover, and Sparc is required for appropriate procollagen processing and deposition of collagen in the ECM. Mice deficient in Sparc have reduced collagen content, and collagen is tethered to the cell surface, with reduced fibril aggregation in the ECM [46, 47].
Proteoglycans also regulate and affect collagen fibrillogenesis, as both decorin knockout and the fibromodulin/biglycan double-knockout mice exhibit alterations in collagen ultrastructure in skin [48, 49]. Both the protein core and the GAG chain can influence ECM function . Given the lack of transcriptional regulation of genes encoding proteoglycans in the human cervix  and our studies in the mouse, further investigations are required to evaluate postranslational regulation of the protein core as well as regulation of glycosaminoglycan synthesis, chain length, and degree of sulfation. A role for small proteoglycans, such as decorin, in modulation of cervical collagen structure is also supported by studies in the pregnant rat [51–53].
This work has identified early pregnancy changes in collagen processing and ECM composition that are likely responsible for the initial increase in tissue compliance during cervical softening. In contrast to the accelerated changes that occur during cervical ripening and dilation at the end of pregnancy, these early changes occur in an environment rich in progesterone and relatively low estrogen. Future studies are required to identify steroid and peptide hormones that may regulate collagen cross-link formation and Thbs2 and Tnc expression. These studies not only enhance our understanding of the progressive physiological changes in normal cervical softening, but they also identify specific genes/proteins in which mutations or misregulation may account for clinical complications such as cervical insufficiency or result in premature cervical shortening of the cervix, which is a risk factor for preterm birth [54, 55]. Future studies will address these important questions and provide necessary understanding required for development of therapies to prevent preterm birth, the leading cause of infant death in the first year of life.
We would like to thank Dr. Larry Fisher at the NIH for use of the C-propeptide antibody. We thank Dr. Christopher Gilpin, Dr. Xinran Liu, and the UTSW Molecular and Cellular Imaging Center for help with tissue preparation and visualization for TEM. Finally, we extend our appreciation to Dr. Brenda Timmons for assistance with data analysis.
1Supported by National Institutes of Health grant R01 HD043154.