Myometrial progesterone receptor determines a transcription program for uterine remodeling and contractions during pregnancy

Abstract

The uterine myometrium expands and maintains contractile quiescence before parturition. While the steroid hormone progesterone blocks labor, the role of progesterone signaling in myometrial expansion remains elusive. This study investigated the myometrial functions of the progesterone receptor, PGR. Pgr ablation in mouse smooth muscle leads to subfertility, oviductal embryo retention, and impaired myometrial adaptation to pregnancy. While gross morphology between mutant and control uteri are comparable, mutant uteri manifest a decrease of 76.6% oxytocin-stimulated contractility in a pseudopregnant context with a reduced expression of intracellular calcium homeostasis genes including Pde5a and Plcb4. At mid-pregnancy, the mutant myometrium exhibits discontinuous myofibers and disarrayed extracellular matrix at the conceptus site. Transcriptome of the mutant mid-pregnant uterine wall manifests altered muscle and extracellular matrix profiles and resembles that of late-pregnancy control tissues. A survey of PGR occupancy, H3K27ac histone marks, and chromatin looping annotates cis-acting elements that may direct gene expression of mid-pregnancy uteri for uterine remodeling. Further analyses suggest that major muscle and matrix regulators Myocd and Ccn2 and smooth muscle building block genes are PGR direct downstream targets. Cataloging enhancers that are topologically associated with progesterone downstream genes reveals distinctive patterns of transcription factor binding motifs in groups of enhancers and identifies potential regulatory partners of PGR outside its occupying sites. Finally, conserved correlations are found between estimated PGR activities and RNA abundance of downstream muscle and matrix genes in human myometrial tissues. In summary, PGR is pivotal to direct the molecular program for the uterus to remodel and support pregnancy.

Significance Statement.

Parturition disorders, such as preterm birth and laboring dysfunction impose serious health risks on the mother and the child. The myometrium, the muscle compartment of the uterus, is tasked to hold the increasing weight and size of a growing fetus with minimum contractile activities before parturition. To achieve this, the uterus reinforces its structural integrity by remodeling the myometrium. The present study reveals a gene expression program under the control of PGR for this remodeling process in a mouse model. Determination of the epigenomic landscape of pregnant uteri further identified genomic elements pertinent to PGR-dependent transcriptomic regulation. Results of this work demonstrate a novel role of PGR in preparing the myometrium to support pregnancy.

Introduction

Parturition disorders often result in maternal and fetal complications. According to the Centers for Disease Control and Prevention, the 2020 preterm birth rate is 10.1% in the United States with disparities among races (1). Short gestation and low birth weight related disorders are the second highest cause and accounted for 16.5% of infant death in 2019. Laboring dysfunction may also lead to an unplanned cesarean. Emerging evidence indicates an increase of the relative risk of postpartum laparotomy in mothers who received emergency cesarean section compared with planned surgery (2). The costly management of consequences of abnormal parturitions also imposes burden on patients and family members (3, 4).

The uterus undergoes major morphological changes to accommodate pregnancy. In the myometrial compartment, pregnancy stimulates myometrial expansion to meet the functional demand of housing the growing fetus. At the same time, the uterine muscle remains in a noncontractile status to avoid expelling the fetus prematurely and maintains the muscle tone to provide structural integrity of the uterus. At term, the myometrium switches to a contractile role providing mechanical force for parturition. Failure in this process may lead to laboring disorders. While major signaling pathways have been identified for the transition to parturition (5–7), mechanisms behind myometrial remodeling remain elusive.

During pregnancy, rodent studies have revealed that the myometrium undergoes hypertrophy and hyperplasia events during myometrial expansion (8, 9). In the mouse myometrium, cell proliferation peaks at gestation day 2 while the hypertrophic process evidently accelerates, as early as gestation day 7 (8). Previous studies show distinct transcriptomic and epigenomic landscapes between nonpregnant and term pregnant myometrial tissues that underlie their morphological differences (10, 11). Results from the mouse models further shows that the myometrium is epigenetically programed, as early as mid-pregnancy ready for a gene expression pattern that regulates subsequent parturition (12). Based on myometrial gene expression profiles, a vast majority of enriched biological processes that are altered between gestation day 14 and virgin mice are also seen in those enriched between term pregnancy and the nonpregnant state (10, 13). These observations suggest that myometrium simultaneously undergoes hypertrophy and prepares parturition machineries under a dynamic transcriptomic and epigenomic environment. This permits a remodeling process that adapts the uterus structurally and functionally, making it suitable to maintain pregnancy and prepare for parturition.

Progesterone is an ovarian steroid hormone and progesterone signaling is crucial for embryo implantation and uterine contractility (14, 15), while the cognate receptor, progesterone receptor (PGR) is required for female reproductive capacity (16). The PGR-A and PGR-B isoforms are the two major PGR isoforms in the uterus (17). Changes in myometrial progesterone signaling, either through reduction of local ligand availability or via a switch of the PGR-A to PGR-B ratio, determine the status of myometrial contractile activities (13, 15, 18). It is proposed that PGR works with different combinations of transcription factors, likely via PGR isoforms (19), at the prelabor and laboring stages to regulate expression of contractile and inflammatory genes in control of myometrial contractility for parturition (5–7, 12). Nonetheless, the role of PGR in the myometrial remodeling process is unclear.

The present study describes the functional and structural impact of PGR loss on the mouse uterus in a smooth muscle specific PGR knockout model. The in-vivo landscape of uterine transcriptome, PGR cistrome, and epigenome is documented at mid-pregnancy during myometrial expansion. Uterine enhancers that are topologically associated with PGR downstream genes are cataloged, from which a panel of candidate PGR interacting transcription factors are derived to permit investigations on the molecular network program that constructs the uterus structurally and functionally during pregnancy.

Results

Reduced fertility and uterine contractility in myometrial PGR deficient female mice

The mouse Pgr gene was inactivated in smooth muscle cells by crossing the Tg(Myh11-cre,-EGFP)2Mik and Pgr^tm4.1Lyd mice to produce PR^d/d (Myh11^Cre; Pgr^flox/flox) mutant mice. PR^f/f (Pgr^flox/flox) mice were utilized as control. Immunostaining results confirmed the loss of PGR protein in both inner and outer layers of myometrium in mutant uteri, while PGR protein remained detectable in stromal, luminal epithelial, and glandular epithelial cells of the endometrial compartment (Figure S1). In addition, mutant oviducts lost PGR protein expression specifically in the muscularis layer (Figure S1). These findings demonstrated the specificity of inactivating PGR expression in the myometrial compartment of mouse uteri. Furthermore, the adult control and mutant uteri appeared comparable at the histological level. At 2.5 days post-coitum (dpc) before the presence of embryos in the uterus, mutant mice had organized muscle bundles in the outer layer and well-aligned, continuous muscle sheets in the inner layer of the myometrium, like the morphology of the control uteri (Figure S1). Moreover, expression of the smooth muscle marker CNN1 (Calponin) protein was observed in mutant myometrium as that in control (Figure S1). These observations indicate that gross morphology of the mutant uteri was maintained after the loss of myometrial PGR and suggest that uterine development was less likely impacted in the mutant mice, which supports the feasibility of using this mouse model to study pregnancy in the adulthood.

Myometrial PGR loss resulted in subfertility in mutant females. In a 6-month breeding trial, six mutant females produced 13 pups in 3 litters, compared with 126 pups in 22 litters by six controls. Only two out of six mutant females (33.3%) produced offspring and these two mice had significantly lower numbers of total litters over the trial period (Figure 1A). The resulting average litter size in the trial cohort was 5.97 ± 0.74 pups/litter for control mice and 1.25 ± 0.91 for mutant females (Figure 1A). This subfertility phenotype manifested at early pregnancy. At 5.5 dpc, a significant reduction of implantation sites was observed in the mutant uterus (2.78 ± 1.06, N = 9) compared to control (7.67 ± 1.61, N = 6) with p < 0.05 by the two-tailed Mann–Whitney test. Six of the nine mutant females (66.67%) carried implantation sites, while five of the six control uteri (83.33%) had embryo implantations. The lower numbers of implantation sites in mutant females were a result of the embryo transportation defect to the uterus. At 3.5 dpc, statistically comparable numbers of embryos from each individual female were found between control and mutant animals after flushing the oviducts and uterus (Figure 1B), indicating that control and mutant mice had no difference in the ovulation capability. However, only 42.27% ± 20.02% of embryos traveled to the mutant uterus, while all embryos already arrived in the control uteri (Figure 1C). In mutant mice, the remaining 57.73% of embryos were found in the oviducts (Figure 1D). Additionally, embryos were present in the uterus of four out of six (66.67%) mutant females at 3.5 dpc, in contrast to embryo presence in all control mice. Micro-computed tomography further revealed an aberrant spatial structure in all six oviducts from three mutant mice that failed to form the normally seen oviduct looping arrangement that resembles fingers on a fist (Figure 1E). Since successful embryo implantation is permitted only in a narrow window of uterine receptivity and embryos missing this window often fail to develop further (20), subfertility of mutant females is likely a result of the oviductal retention phenotype partially due to oviduct structural defects.

Fig. 1.

Assessment of breeding capacity, oviductal embryo transportation, and oviduct morphology. (A) Summary of 6-month breeding trial data. (B to D) Distribution of embryos in the female reproductive tract at 3.5 dpc. (E) Representative images of volumetric analysis of oviductal morphology and organization with microCT (N = 6 oviducts from three mice for each genotype). The oviducts are highlighted in blue, fragments of the uterus are in yellow, and parts of the ovary are in pink. *P < 0.05, **P < 0.01 by two-tailed Mann–Whitney test. Error bars depict the SEM.

Next, we examined the impact on muscle contractility of the female reproductive tract in relation to PGR loss in the muscle compartment. Results from the wire myograph assay revealed a significant reduction of oxytocin-stimulated contractility in mutant uteri that were isolated at pseudopregnancy day 6 (PPD6), compared to the control (Figure 2A). This observation is consistent with the previous finding that altered progesterone signaling affects uterine contractility (13). Notably, the gross morphology at PPD6 was comparable between control and mutant uteri with a similar structural appearance in the CNN1-stained myometrium (Figure 2B). These findings not only demonstrated the essence of myometrial PGR for uterine contractility, but also implied that PGR may modulate oxytocin signaling effectors rather than control general morphology for contractile regulation in this structural context.

Fig. 2.

Functional, morphological, and molecular assessments of myometrial PGR deficient uteri at PPD6. (A) Contractility of isolated uterine tissues determined by wire myograph. (B) Representative images of CNN1 immunostaining (N = 5 for each genotype). (C) Genome-wide differentially expressed genes (DEGs) between PR^d/d and PR^f/f uteri. The cartoon on the right shows the affected expression of the genes of interest in the regulation of the intracellular calcium levels. OXTR, oxytocin receptor; Gq, G-proteins; PLCβ, phospholipase Cβ; PIP2, phosphatidylinositol 4,5-bisphosphate; IP3, Inositol trisphosphate; ITPR, Inositol 1,4,5-triphosphate receptor; Ca++, calcium; PKG, cGMP-dependent protein kinase; PDE5, Phosphodiesterase-5. (D) qRT-PCR results of the Pde5a, Plcb4, and Cacna1c relative mRNA levels between control and mutant uteri (N = 7 for each group with technical duplicates on individual specimen). *P < 0.05, **P < 0.01, P < 0.001 by two-tailed Mann–Whitney test. Error bars depict the SEM.

The transcriptomic changes in the PPD6 mutant uterus were examined by RNA-Seq to further understand the molecular profile behind the physiological phenotype. Compared to control, mutant mice exhibited 1,743 DEGs with 844 up-regulated and 899 down-regulated using the cutoff criteria of absolute fold change > 1.5 and unadjusted P < 0.05 (Figure 2C, Dataset S1). Ingenuity Pathway Analysis (IPA) identified enrichments of DEGs in annotated pathways pertinent to oxytocin signaling, including the “Oxytocin Signaling Pathway” and “G-Protein Coupled Receptor Signaling” (Dataset S1). Muscle contraction–associated pathways that were annotated under “Nitric Oxide Signaling in the Cardiovascular System,” “Relaxin Signaling,” and the “Dilated Cardiomyopathy Signaling Pathway,” among others, also had overrepresentations of DEGs (Dataset S1). Further examining the DEGs that are associated with oxytocin signaling and muscle contraction revealed reduced mRNA abundance of Plcb4, which encodes the oxytocin receptor downstream effector PLCβ, in the mutant uterus (Figure 2C and D, Dataset S1). In addition, the mutant uterus also had a decreased expression of Cacna1C, Cacna1d, and Cacna1e (Figure 2C and D, Dataset S1) genes that encode voltage-gated calcium channels, which have a role in regulation of myometrial contractions (21). Moreover, examining muscle-associated pathways, including “G-Protein Coupled Receptor Signaling,” “Relaxin Signaling,” “Cellular Effects of Sildenafil (Viagra),” “Nitric Oxide Signaling in the Cardiovascular System,” “Cardiac Hypertrophy Signaling,” and “Cardiac β-adrenergic Signaling,” identified the commonly annotated, phosphodiesterase 5 (PDE5) encoding gene Pde5a (Dataset S1), which had significantly lower mRNA abundance in the mutant than the control uterus (Figure 2C and D, Dataset S1). Pde5a is expressed in the myometrium of pregnant mice (22) and PDE5 inhibitor sildenafil has been reported to reduce intrauterine pressure at labor in a rat model (23). This transcriptomic data supports the ex-vivo functional result of failed oxytocin responses on uterine contractions in mutant mice. In summary, Pgr deficiency in the smooth muscle led to an altered gene expression profile in favor of dampening the oxytocin-dependent muscle contraction. However, the oxytocin receptor gene was unaffected.

Disrupted myometrial remodeling by PGR loss

A subset of embryos successfully proceeded to further development, permitting investigations of the impact of myometrial PGR loss on the uterine remodeling process at the conceptus sites. Since the uterine epigenome and transcriptome are already programed for the need of parturition at mid-pregnancy (10–12), we chose to examine the role of myometrial PGR in uterine remodeling at this stage. Immunostaining of smooth muscle markers ACTA2 and CNN1 shows that, at the conceptus site of the control uterus, the outer smooth muscle area maintained a roundish shape of muscle bundles in cross-sections of the uterus (Figure 3A). Control uteri also had a continuous inner myometrial subcompartment that was consistently layered with multiple sheets of smooth muscle cells stacking closely together (Figure 3A). In contrast, mutant uteri manifested a spindle shape of muscle bundles in the outer part of myometrium (Figure 3A). The mutant inner myometrial region showed discontinuity and uneven distribution of sheets of smooth muscle that were stacked in a relatively loose manner (Figure 3A). Moreover, Mason’s trichrome staining revealed disorganized extracellular matrix (ECM) with loosely arranged collagen fibers in the mutant uterus, rather than a well-aligned matrix that wrapped around muscles as seen in the control uteri (Figure 3A). This data provided a qualitative view of the reduction of smooth muscle mass in mutant uteri. These observations collectively demonstrate that mutant myometrium failed to structurally adapt to pregnancy at the uterine remodeling phase and such a failure led to aberrantly stretched myometrium with disarrayed extracellular matrix.

Fig. 3.

Morphological and transcriptomic assessment of 13.5 dpc uteri. (A) Immunostaining of ACTA2 and CNN1 in brown and hematoxylin counterstaining in blue. Mason’s trichrome staining are presented to the right. (B) Heatmap for relative expression levels of muscle, ECM, and ECM-cell interaction genes identified by RNAseq. Color scale to the right denotes relative expression levels with increase in yellow and reduction in blue. (C) Selected biological processes and pathways from Dataset S2 that are overpresented in the 1,535 PGR downstream genes. (D) Comparison between PGR downstream genes and DEGs that are derived from term versus mid-pregnancy mouse myometrium [National Center for Biotechnology Information (NCBI) accession number: GSE25017].

Transcriptomic profiling of 13.5 dpc uteri agreed with morphological assessments. RNA-Seq data revealed that myometrial PGR loss resulted in 1,535 DEGs between control and mutant uteri at the conceptus sites (Dataset S2). Gene ontology analyses revealed overrepresentations of muscle, extracellular matrix, and cell–matrix interaction genes in these DEGs, in which most of these genes had relatively lower levels in the mutant (Figure 3B and C and Dataset S2). qRT-PCR assays on a panel of smooth muscle markers validated the RNA-Seq findings where the mutant myometrial defects were reflected by lower levels of RNA abundance of Tagln, Acta2, Actg2, Myh11, and Cnn1, as compared with control (Figure S2). Based on the DEG profile, molecular activities of progesterone, predicted by IPA, were decreased in the mutant uteri as expected (Dataset S2). In addition, estimated activities of muscle and matrix regulators, including TGFβ, SRF/MYOCD, and CCN2 (CTGF, Connective Tissue Growth Factor) were also lower in mutant than control uteri (Dataset S2). The significant reduction of RNA abundances and projected activities of muscle regulator Myocd and matrix modulator Ccn2 (Dataset S2 and Figure S2) indicate that both their levels and activities were affected by myometrial PGR and suggest their pivotal role in the impact of myometrial PGR loss on the mutant myometrial phenotype. These findings reveal that myometrial PGR promotes expression of muscle and matrix component genes and their major regulators during uterine remodeling.

Further mining of the mutant transcriptomic profile against publicly available gene expression datasets on the BaseSpace Correlation Engine (Illumina) identified an association with GSE25017, a dataset that compares mid- and term-pregnancy myometrial gene expression profiles in wild type mice (24). This association offers an opportunity to estimate the status of the mutant uterus in relevance to the stage of pregnancy. The BaseSpace Correlation Engine identified 731 mutant DEGs that overlapped with the DEGs between wild type term- and mid-pregnancy myometrium (Figure 3D). Among these genes, 580 were positively correlated and 151 showed an opposite expression pattern between the two datasets (Figure 3D). The pattern of nearly 80% of genes in positive correlations suggests that a molecular profile of the mid-pregnancy mutant uterus is in closer resemblance to late pregnancy of the wild type myometrium. This observation also implies that myometrial PGR is required before term to prevent the uterus from reaching a structurally matured status. In summary, the histological and transcriptomic data supports a crucial role of myometrial PGR in conducting the uterine remodeling process towards term pregnancy.

Genomic and epigenomic landscape of the uterus at mid-pregnancy

Since the myometrial epigenome is prepositioned at mid-pregnancy to prepare gene expression at term (12), we sought to determine the global PGR occupancy pattern in the epigenomic landscape of mid-pregnancy uteri. ChIP-seq assays and chromatin conformation capture by HiC identified 47,535 H3K27ac-positive enhancers, 13,574 PGR occupying sites, and 17,807 topological DNA contacts in 13.5 dpc uteri of wild type animals (Table 1 and Dataset S3). The 13.5 dpc H3K27ac-positive enhancers encompassed 22,607 of 25,633 H3K27ac intervals that were also found previously at mid-pregnancy (12), showing an agreement between two separate studies. Among mid-pregnancy uterine enhancers, nearly 20% (9,386) of 47,535 were not present in other tissues that were previously mapped by the ENCODE project (NCBI accession number: GSE31039) (Figure S3A and B). Further analyzing frequency of uterine enhancers that are also found in other tissues revealed that a great portion of uterine enhancers could also be employed by other tissue types (Figure S3B). Only 12.3% of uterine H3k27ac enhancers were in a 3-kb promoter region, while most of the remaining enhancers were found in the intergenic zone or introns (Figure S3C). As expected, DNA binding motifs of PGR and myometrial regulators AP-1 and STAT5 factors were overrepresented in uterine enhancers among others (Dataset S4). These data not only annotated the genome-wide enhancer distribution patterns in the mid-pregnancy uteri, but also implies additional layers of regulatory mechanisms that may join specific combinations of enhancers beyond the promoter region to define the tissue-specific transcription program.

Table 1.

Annotations of H3K27ac-posiive enhancers with PGR genome occupancy and chromatin conformation patterns in 13.5 dpc mouse uteri.

	Number of intervals
H3K27ac-postitive enhancer	47,535
Super enhancer	3,264
PGR occupancy	13,574
Enhancer with PGR occupancy	10,843 (22.8%)
Super enhancer with PGR occupancy	2,427 (74.4%)
Chromatin loops	17,807
Distal enhancer topologically in contact with the transcription start site (TSS) of active genes	4,232
Enhancer located in the same Chromatin loop with the TSS of active genes	38,117
Distal enhancer topologically in contact with the TSS of PGR downstream genes	648
Enhancer located in the same Chromatin loop with the TSS of PGR downstream genes	14,669
Distal enhancer with PGR occupancy topologically in contact with the TSS of PGR downstream genes	205
Enhancer with PGR occupancy located in the same Chromatin loop with the TSS of PGR downstream genes	3,203

Similar to the enhancers, more than 70% of PGR occupancy was found outside the promoter region (Figure S3D) and 88% of PGR occupying sites are located in the H3K27ac-positive regions (Figure 4A). Notably, PGR-binding was found in 74.4% of uterine super enhancers, compared with only 22.8% enhancers that had PGR occupancy (Table 1). This pattern suggests a preference of PGR in choosing the cis-acting elements that could simultaneously control multiple genes. Overrepresentation of DNA-binding motifs of known myometrial regulators, including AP-1, STAT, and NFkB proteins, can be found in either PGR occupying sites or in the PGR-binding enhancers (Dataset S4). These data suggest that PGR occupancy offers a layer of transcriptomic regulatory mechanism in conjunction with the uterine enhancers.

Fig. 4.

Enhancers with and without PGR occupancy in the same chromatin loops of PGR downstream genes. (A) Overlapped and distinct regions between PGR and H3k27ac intervals. (B) Categorization of Group A (with PGR) and Group B (without PGR) enhancers. Cartoon on the right: blue triangle, enhancers; red triangle, PGR occupying sites; arrows, TSS of PGR downstream genes; green dot, DNA contacts; blue line, DNA loops. (C and D) Genomic loci of Myocd (C) and Ccn2 (D) with PGR occupancy, H3K27ac ChIP signals and DNA loops identified by HiC. (E) Overrepresented transcription factor binding motifs in PGR downstream gene associated enhancers. Motifs with enrichment P value less than 0.05, as denoted in the color scale, in one of the two groups of enhancers are shown. Orders of motifs are arrangement based on hierarchical clustering of the enrichment P values.

Given that the majority of uterine enhancers and PGR occupying sites are outside the promoter region, topologically associated enhancers and active genes were identified based on their relative positions within chromatin loops in mid-pregnancy uteri. There are 4,232 enhancers in direct contact with the TSS of 2,636 active genes via looping (Table 1, Dataset S5). Due to the limitation of tissue Hi-C resolution, we also examined colocalization of enhancers and the TSS of active genes in the same chromatin loops. Based on this criteria, 38,117 enhancers are found to be topologically associated with 8,407 active genes (Table 1, Dataset S5). These data provided a genome-wide map that catalogs the enhancers of high probability in regulation of specific target genes in the context of a remodeling uterus. In summary, results from RNA-seq, ChIP-seq, and HiC assays together annotated the genomic and epigenomic landscape of mid-pregnancy uterine tissues. Our integrative data analysis also infers the transcriptional regulators that may determine the gene expression pattern at the time of uterine remodeling to accommodate fetal growth and to build the uterus for parturition.

Regulation of PGR downstream genes in mid-pregnancy myometrium

These newly annotated interactions among genes and enhancers permit further investigation on the regulon for PGR downstream genes. Among the TSS-contacting enhancers, 648 of them have direct contacts with 301 PGR downstream targets and 205 enhancers have PGR occupancy (Table 1, Dataset S6). On the other hand, there are 14,669 enhancers in the same loops of 1,192 PGR downstream genes (Table 1, Figure 4B, Dataset S6). Examples of PGR downstream genes in association with these enhancers include myometrial remodeling genes Myocd, Col3a1, Col4a1, Col4a2, and Ccn2, as well as the contractile regulator Mylk (Figure 4C and D and Figure S4). The 3,203 enhancers of this cohort that have PGR occupancy are designated as the Group A enhancers (Table 1, Figure 4B, Dataset S6). The remaining 11,466 nonPGR-binding enhancers (Group B) include ones that are with or without other PGR-occupied enhancers in the same loops (Figure 4B). The vast majority (10,723) of Group B enhancers are colocalized with other PGR-occupied enhancers and PGR downstream genes in the same loops. Despite the current resolution of our HiC data was unable to reveal direct contacts of Group A and Group B enhancers with TSS of PGR downstream genes, we reason that these enhancers are more likely to be part of the regulation of PGR downstream genes because they have been brought to a topologically close vicinity of PGR downstream genes by chromatin looping. Therefore, they are chosen for subsequent motif enrichment analyses to identify the candidate regulatory program that modulates gene expression together with PGR.

Group A and Group B enhancers manifest enrichments of common and distinctive transcription factor binding motifs (Figure 4E and Dataset S4). Motifs found in both groups of enhancers include AP-1, ETS, and STAT proteins that are known myometrial regulators (Figure 4E and Dataset S4). Additionally, motifs of major muscle regulators MEF2D and CHOP are also enriched in both enhancer groups (Figure 4E and Dataset S4). Motifs of PGR, JUND, SP1, and HOX proteins are overrepresented only in the Group A enhancers where PGR binds. On the other hand, motifs of muscle growth (SRF, NFAT, and KLF) and inflammatory response (IRF and FOX) factors are statistically enriched only in Group B enhancers (Figure 4E and Dataset S4). The distinct motif enrichment patterns between these two groups of enhancers suggest that the temporal epigenomic landscape enables PGR to enlist transcription regulators in the topological vicinity to increase versatility of PGR-dependent gene expression program. Notably, predicted molecular activities of AP-1 proteins, SRF, and MEF2D are reduced in mutant uteri (Dataset S2), suggesting that PGR likely works with these factors in regulation of PGR downstream genes. In summary, these findings identified candidate PGR partners for transcription regulation beyond PGR occupying regions.

PGR activities in human myometrial tissues

The in-vivo pregnant mouse myometrial PGR gene signature (Dataset S2) enables an estimation of PGR activities in human myometrial tissues. The inferred PGR activity of individual myometrial specimens can be quantitatively presented as a T-score for the manifestation of the mouse PGR gene signature in the human specimen gene expression profile (25–27). As expected, compared with the nonpregnant stage, the term pregnant human myometrial tissues manifest relatively stronger PGR activities, as shown in higher T-scores, reflecting progesterone signaling for myometrium remodeling in adaptation to pregnancy (Figure 5A) (GSE137552) (11). Notably, in three of the four publicly available datasets, the term labor myometrial specimens exhibit significantly lower inferred PGR activities than the tissues from patients at term pregnancy but not in labor (Figure 5B and Figure S5) (28–31). This observation is in line with and supports the functional progesterone withdraw theory for parturition (32).

Fig. 5.

Projection of the mouse myometrial PGR gene signature on human myometrial specimens. (A) GSE137552. (B) The Mittal dataset. T-scores are surrogates for inferred molecular activities of PGR in each individual human specimen based on the manifestation of the mouse PGR gene signature. NP, nonpregnant; TP, term pregnant; TNL, term pregnant nonlabor; TL, term pregnant labor. *P < 0.05 by two-tailed unpaired t-test. Error bars depict the SEM. (C) Associations between inferred myometrial PGR activities and relative mRNA levels of PGR downstream targets and myometrial regulators in the Mittal dataset. Each column presents the gene expression levels of depicted genes in one specimen. Specimens are ordered based on the inferred PGR activities with the highest on the left and the lowest on the right. The relative mRNA levels, across the 39 specimens, are shown in the color code on the side.

The association between myometrial PGR activities and RNA abundance of PGR downstream genes is also examined in human myometrial tissues. Across the 39 human myometrial specimens published by Mittal and colleagues (The Mittal dataset) (29), inferred PGR activities, in T-scores, show correlations with the relative mRNA levels of genes for muscle, extracellular matrix, and extracellular matrix–cell interactions (Figure 5C and Dataset S7), in the same direction as seen in the functional analysis conducted in the mouse model (Figure 3B and Dataset S2). Notably, inferred PGR activities are positively correlated with relative mRNA levels of the muscle gene regulator MYOCD, the extracellular matrix gene modulator CCN2 (CTGF), and contractility repressors PLCL1 and ZEB1 in the human myometrium (Figure 5C and Dataset S7) (6, 13). Moreover, inferred PGR activities are inversely associated with relative mRNA levels of GADD45A and RXFP1 (Figure 5C and Dataset S7). Previous mouse studies reveal that Gadd45a deficiency results in a parturition failure phenotype (33) and Rxfp1 null females manifest a higher rate of stillbirth compared with control animals (34). Collectively, these findings suggest a conserved role of PGR in regulation of parturition-associated and myometrial remodeling genes in the human myometrium.

Discussion

This study reveals a regulatory program for the pregnancy-dependent myometrial remodeling and expansion process under PGR control. The integrative transcriptomic, cistromic, and epigenomic analyses identified candidate cis-acting elements that may partner with PGR to modulate gene expression. With respect to clinical relevance, results from examining human myometrial datasets support the association of PGR activities with expression of downstream genes that are identified in model systems, such as cultured human myometrial cells and genetically engineered mice. The present study focuses on the molecular network that prepares the uterus to term and complements studies that investigate myometrial contractility and the transition to parturition.

Our results show that, while maintaining contractile quiescence, myometrial PGR orchestrates a transcriptional program to supply both intracellular muscle filaments and extracellular matrix in preparation for the mechanical capacity for parturition. Myometrial PGR modulates major pathways, such as SRF/Myocardin and CTGF (Dataset S2),which control expression of genes that encode building blocks for muscle and extracellular matrix, respectively. Myometrial PGR could also affect activities of a key contractile regulator ZEB1 in our mouse model, based on the expression patterns of ZEB1 downstream genes (Dataset S2) (24). Our data supports that myometrial PGR exerts its regulatory function at both upstream regulators and downstream effectors. Taking Myocd and Des as examples, MYOCD is a coregulator of the master smooth muscle transcription factor SRF and Des encodes a muscle-specific intermediate filament protein. Both genes have higher RNA levels in term pregnant myometrium compared with nonpregnancy myometrial tissues (11). RNA abundances of these two genes are positively correlated with estimated myometrial PGR activities (Figure 5C). In model systems, both Myocd and Des RNA levels decrease in oviducts of PGR-null mice (35), while myometrial PGR is required to sustain expression of these two genes in pregnant uteri (Figure 3B, Figure S2, and Dataset S2). Taking together, these data support a promoting role of myometrial PGR on Myocd and Des transcription. Interestingly, progesterone treatment did not lead to an increase of MYOCD and DES RNA abundances in cultured human myometrial cells (36). This discrepancy implicates a necessity of PGR to work with a potential factor that presides only in a physiological context for regulation of MYOCD and DES genes.

Myometrial PGR also regulates multiple component and regulator genes for extracellular matrix in either a direct or indirect manner (Figure 3B, Dataset S2, and Dataset S6). For example, myometrial PGR promotes uterine Col3a1 expression at the conceptus sites likely through direct transcription regulation (Figure 3B, Figure S4, Dataset S2, and Dataset S6). In human, progesterone treatment increases COL3A1 expression of PGR-expressing human myometrial cells (36). Moreover, COL3A1 RNA levels are higher in pregnant than nonpregnant myometrium (11) and are positively correlated with estimated myometrial PGR activities (Figure 5). Notably, mutations in the COL3A1 locus are linked to the Ehlers–Danlos syndrome type IV and one of the complications is uterine rupture during pregnancy (37). These observations indicate that progesterone and myometrial PGR act at the transcriptional level to elevate the production of an important matrix gene Col3a1 in support of pregnancy across species. Another example is CCN2 (CTGF), which stimulates extracellular matrix production by smooth muscle (38, 39). Ccn2-deficient mice exhibited defects in extracellular matrix production and died at the neonatal stage (40). In human, progesterone promotes CCN2 expression in cultured human myometrial cells that express both PGR-A and PGR-B isoforms (36), while presumed PGR activities also positively correlate with CCN2 RNA abundance in human myometrial tissues (Figure 5C). Consistent with these findings, in the mouse model, myometrial PGR is required to promote Ccn2 expression in pregnant mouse uteri amid pregnancy dependent myometrial remodeling (Figure S2 and Table S1). These data together show the regulation of the matrix gene regulator CCN2 by progesterone and PGR. In summary, myometrial PGR supports pregnancy by creating gene expression patterns that permit pregnancy-dependent extracellular matrix remodeling in the uterus.

Preterm birth and apparent parturition abnormalities were not observed during the breeding trials despite the mid-pregnancy mutant uterus exhibited elevated Oxtr levels, a reduction of Plcl2 RNA abundance, and a transcriptomic profile that resembles term pregnancy. In addition, PR^d/d uteri at PPD6 also had a significant reduction on oxytocin-stimulated contractility (Figure 2). OXTR and PLCL2 transduce the oxytocin signaling to downstream effector MYLK to direct muscle contraction. It has been shown that MYLK is primarily expressed in the myometrial compartment (41), and Mylk haploinsufficiency is sufficient to reduce contractility in vascular smooth muscle cells (42). Since Mylk is downstream of Oxtr and Plcl2, the reduction of Mylk levels at mid-pregnancy could partially explain the lack of preterm birth in this model despite that Oxtr and Plcl2 levels may create an environment in favor of muscle contraction. Uterine Mylk expression can be elevated by progesterone in PGR-A, but not in PGR-B background in an ovariectomy model (43). Together with the observation that PGR-A is the dominant isoform in the mouse uterus (13, 44, 45), likely the reduction of Mylk expression in mid-pregnancy PR^d/d uteri is a result of PGR-A loss. On the other hand, the increase of Oxtr expression in the mid-pregnancy PR^d/d uterus may result from the loss of the PGR-B isoform because two PGR-B overexpression mouse models both show a reduction in uterine Oxtr RNA abundance (13, 43). These examples demonstrate that the results of the present study are most likely a summary effect of both PGR-A and PGR-B isoform deficiency. Moreover, previous studies show that, in polytocous species, females with lower numbers of embryos often have a longer gestation period (46, 47). The significantly lower embryo load of the mutant females may also negate the effect of the aberrantly distended uterine wall on parturition timing. Future experimentations using the embryo transfer technique to achieve an embryo load in the mutant uterus similar to that in control mice may remove this potential confounding factor and permit the assessment of the effect of uterine remodeling on parturition timing.

Transcriptional factors that have their DNA binding motifs overrepresented in enhancers with topological associations with PGR downstream genes (Figure 4E) may have a role in concurrent regulation of transcription together with PGR and/or with other factors in the vicinity. For example, both CEBPB and SP1 participate in PGR-dependent regulation of ADAMTS-1, which has no PGR binding motifs in the promoter region (48). CEBPB and SP1 RNA are expressed in human myometrial and pregnant mouse uterine tissues (Dataset S2) (11). Predicted CEBPB and SP1 molecular activities are reduced in myometrial PGR deficient uteri. These data imply a regulatory network of PGR, CEBPB, and SP1, in which these transcription factors work together from topologically associated enhancers, with or without PGR occupancy within, to direct transcription. Another example is SRF, a master regulator for smooth muscle (49). SRF binds DNA via a DNA consensus sequence of CC[AT]₆GG (CArG box) to directly regulate the expression of SRF downstream genes. SRF regulates cell adhesion and muscle contractility through modulating gene expression of extracellular matrix components and cytoskeletal proteins via interactions with myocardin and myocardin-related transcription factors (MRTFs) (50). SRF is also capable to form a complex with ternary complex factors, such as Elk to conduct gene regulation (51). With respect to sex hormone signaling, SRF recruits the androgen receptor (AR) to coregulate androgenic steroid responsive genes in cis-acting elements that lack the AR-binding motif (52). SRF and its coregulators are expressed in human myometrial tissues (11) and the mouse uterus (Dataset S2). Predicted SRF and Myocardin activities are decreased in PR^d/d uteri based on the expression pattern of SRF downstream genes. These observations together suggest a potential interaction between SRF and PGR signaling via the chromatin looping mechanism and warrant future investigations.

The mid-pregnancy myometrial PGR gene signature in this study is derived from the uterine wall at the conceptus site at the time of uterine remodeling and myometrial expansion. The observation that this gene signature manifests a better resemblance to pregnant human myometrial tissues than nonpregnancy specimens (Figure 5A) supports an interpretation that likely this gene signature represents the PGR activity for building the myometrium in preparation of future parturition. Interestingly, this “myometrium-building” activity becomes lower at the transition to laboring (Figure 5B). This change implies that myometrial cells begin to switch their cellular functions from building the myometrium to force generation and post-parturition involution that reversely remodels the uterus back to a nonpregnant size and status.

Loss of PGR in the smooth muscle resulted in embryo retention in oviducts, which contributed in part to the subfertility phenotype. The presence of embryos in the uterus of 66.67% of mutant females at 3.5 dpc and having implantation sites in one-third of mutant females at 5.5 dpc suggests that the uterus of most mutants appeared capable of permitting embryo implantation. In addition, the average 42% of embryos transited to the mutant uterus on time at 3.5 dpc is also in line with the observation that mutant females manifested approximately 36% of embryo implantation sites of those in control mice at 5.5 dpc. These findings support the possibility that the mutant uterus maintained a certain degree of capacity to permit embryo implantation for the number of embryos that arrived on time. Later into pregnancy, a further restriction, such as defective uterine remodeling, may limit the capability of the mutant uterus to support embryo development and result in additional loss of fetuses, leading to the very low number of offspring. Future studies of the oviductal phenotype would determine the extent of oviductal embryo retention in a temporal aspect, which would shed light on whether the delayed embryos have a chance to implant. Notably, the average litter size of mutant females was 21% of that of the control mice, which is considerably lower than the 36% relative embryo carrying load at 5.5 dpc. This further reduction supports the aforementioned view on the importance of uterine remodeling in normal pregnancy. Alternatively, the mutant uterus might have an even lower capacity of permitting implantation of the embryos that arrived on time. Therefore, regardless of the timing of arrival in the uterus, a portion of embryos that did implant failed to advance in development and eventually were resorbed, leading to a further reduction of the number of survivable offspring. Future experimentation by providing exogenous embryos to the uterus via the embryo transfer technique would enable the examination of the mutant uterine embryo implantation capacity and supporting capacity independent of the oviductal defect.

The present study focused on the uterine aspect of the PGR functions in the smooth muscle, despite the oviductal defects likely contributed the largest share to the severe subfertility phenotype. While the current results are qualitative and descriptive, future experimentation with quantifiable indexes on the oviductal topology, high resolution histological assays, functional assessments of the embryo movement in the oviduct, and molecular assays would determine the progesterone and PGR-dependent mechanisms that regulate the oviductal homeostasis.

In summary, the present study determined the role of PGR in the pregnancy part of uterine remodeling and identified the PGR-associated molecular network for myometrial expansion. This new information enables future experimentation on functional analyses of progesterone signaling modifiers that may underlie the variable efficacy of progesterone in blocking the uterine contractions for treatment of premature labor.

Materials and methods

Animals

Myh11Cre mice [B6.Cg-Tg(Myh11-cre,-EGFP)2Mik/J] were acquired from the Jackson Laboratory (stock no: 007,742) (53). Pgr^tm4.1Lyd mice were generated in house and are described elsewhere (54). Mice were maintained in a recurrent photocycle of 12 hours on–off in temperature-controlled rooms within an AAALAC-accredited vivarium at the Baylor College of Medicine and the National Institute of Environmental Health Sciences. Animal handling and procedures were performed following the guidelines detailed in the Guide for the Care and Use of Laboratory Animals [National Research Council (Eighth Edition 2011)]. All procedures were approved by the Institutional Animal Care and Use Committee at Baylor College of Medicine under animal protocol number AN-4203 and by the Institutional Animal Care and Use Committee (IACUC) at the National Institute of Environmental Health and Sciences under protocol number 2015–0012.

Breeding pairs were placed together at 1700 hours and separated by 0700 hours the following morning, which was designated 0.5 day post-coitum (dpc); gestational length was subsequently recorded as dpc.

Micro computer tomography

Uterine horns with the ovary and the intact oviduct attached were dissected out in PBS at room temperature and fixed in 4% paraformaldehyde (Electron Microscope Sciences) overnight at 4°C. To preserve the tissue structure, STABILITY protocol has been implemented (55). For that, the tissues were immersed in the STABILITY buffer: 4% w/v paraformaldehyde (pH 7.2, Electron Microscope Sciences), 4% w/v acrylamide (Bio-Rad), 0.05% w/v bis-acrylamide (Bio-Rad), 0.25% w/v VA044 initiator (Wako Chemicals), and 0.05% w/v Saponin (Sigma) in 1X PBS and incubated overnight at 4°C. The samples were placed in a desiccator to replace the air with nitrogen gas, then incubated in STABILITY buffer at 37°C for 3 hours to initiate the acrylamide-PFA cross-linking. The cross-linking reaction created a hydrogel-tissue mixed structure. Fragments of the hydrogel outside of tissues were carefully removed in a fume hood. The samples were washed in 1X PBS with 0.1% sodium azide (Sigma) at 4°C overnight; then immersed in 0.1 N iodine (Sigma) for 3 days. Samples were mounted in 1% agarose (Fisher) for microCT imaging. The imaging was performed with SKYSCAN 1272 micro-CT scanner (Bruker), and the images were reconstructed by Fledkamp Algorithm for cone-beam CT data by NRecon Reconstruction software (Bruker). Volumetric rendering of the oviduct morphology was performed in Imaris 8.02 (Bitplane) software.

Uterine isometric contractility analysis

At 6.5 dpc of pseudopregnancy, mice were euthanized and the entire uterus was excised. The uterine horns were immediately immersed in oxygenated Krebs physiological saline solution (119 mM NaCl; 4.7 mM KCl; 2.5 mM CaCl2; 1.17 mM MgSO4; 25 mM NaHCO3; 1.18 mM KH2PO4; 0.026 mM EDTA; and 5.5 mM D-glucose). Next, uterine strips (5 × 2 mm) were mounted on a wire myograph (Danish Myograph Technology, Aarhus, Denmark), placed under 1 g of tension in a bath of Krebs solution, and equilibrated for 30 minutes using Labchart software (ADInstruments, Colorado Springs, CO).

After a baseline integral value of unstimulated uterine tone was recorded, the uterine strips were exposed to oxytocin (Sigma-Aldrich, St Louis, MO) in various concentrations as depicted in Figure 2A. The responses were recorded and presented as the integral value over a period of 10 minutes and normalized to each individual strip’s baseline integral value at 1 g tension. The baseline resting uterine tone over the duration of exposure to oxytocin was also recorded.

Statistical analysis for the ex-vivo contractility assays utilized the One-Way ANOVA Calculator for Repeated Measures from Social Science Statistics (https://www.socscistatistics.com/).

Histological analysis

Transverse sections of the mid-uterine horn or the conceptus sites were prepared, processed, and paraffin-embedded, as described (56). Pretreatment of the tissues included deparaffinization in xylene and rehydration through graded ethanol. Heat-induced epitope retrieval was performed using a 10-mM citrate buffer solution, pH 6.0 (Biocare Medical, Concord, CA) in the Decloaker pressure chamber for 5 minutes at 120°C. Endogenous peroxidase was blocked using 3% H₂O₂; after which nonspecific sites were blocked using Vector ImmPRESS RTU 2.5% normal horse serum (Vector Laboratories, Burlingame, CA) for 20 minutes at room temperature. The sections were then incubated with Rabbit monoclonal Anti-Progesterone Receptor antibody (Cell Signaling Technology, Inc., Danvers, MA, Cat# 8757, Lot# 3, 0.062 mg/ml) at a 1:1000 dilution for 1 hour at room temperature. Rabbit IgG Isotype control serum (Calbiochem, Cat# NI01, Lot# 2,659,621, 0.1 mg/ml) at the equivalent dilution of 1:1613 was applied to the negative control for 30 minutes at room temperature. Further, the sections were incubated with Vector ImmPRESS anti-rabbit HRP Polymer (Vector Laboratories, Burlingame, CA) for 30 minutes at room temperature. The antigen–antibody complex was visualized using 3-diaminobenzidine (DAB) chromogen (DakoCytomation, Carpenteria, CA) for 6 minutes at room temperature. For CNN1 staining, proteolytic-induced epitope retrieval was performed using a proteinase K ready-to-use solution (Dakocytomation, Carpinteria, CA) for 2 minutes at room temperature. Endogenous peroxidase was quenched using 3% H₂O₂; after which nonspecific sites were blocked using Rodent Block M (Biocare Medical, Concord, CA, Catalog # RBM961) for 20 minutes at room temperature. The sections were then incubated with rabbit monoclonal anti-Calponin (EP798Y) antibody (abcam, Cambridge, MA, Catalog # ab46794, Lot # GR3196369-3) at a 1:1000 dilution for one hour at room temperature. Negative slides were stained with rabbit monoclonal IgG control serum (abcam, Cambridge, MA, Catalog # ab125938). For detection, the slides were incubated in Rabbit-on-Rodent Polymer (Biocare Medical, Concord, CA, Catalog # RMR622) for 30 minutes at room temperature. The antigen–antibody complex was visualized with 3,3’-diaminobenzidine (DAB) chromogen (Dakocytomation, Carpenteria, CA) for 6 minutes. Finally, the sections were counterstained with hematoxylin, dehydrated through graded ethanol, cleared in xylene, and coverslipped. For ACTA2, endogenous peroxidase was blocked using 3% H₂O_2. Nonspecific sites were blocked using 10% normal goat serum (Jackson Immunoresearch, West Grove, PA) for 20 minutes at room temperature. Next, the sections were incubated utilizing an avidin/biotin blocking kit (Vector Laboratories, Burlingame, CA). The sections were then incubated with rabbit polyclonal anti-Alpha Smooth Muscle Actin antibody (Cat# ab5694, Lot# GR214586-1, abcam, Inc., Cambridge, MA) at a 1:1000 dilution for 60 minutes at room temperature. Normal Rabbit IgG (Calbiochem, Cat# NI01, Lot# 2,659,621) at the equivalent dilution was applied to the negative control. Secondary incubation was done by using a biotinylated goat anti-rabbit IgG antibody (Vector Laboratories, Burlingame, CA) at a dilution of 1:500 for 30 minutes at room temperature. Further, the sections were incubated with a R.T.U. Vectastain Kit Label (Vector Laboratories, Burlingame, CA) for 30 minutes at room temperature. The antigen–antibody complex was visualized using 3-diaminobenzidine (DAB) chromogen (Dako North America, Inc., Carpenteria, CA) for 6 minutes at room temperature.

RNA extraction

The middle section of the uterine horn from PPD6 mice or uterine tissues from the antimesometrial side of implantation sites at 13.5 dpc were isolated, followed by homogenization by the Bead Mill 24 homogenizer (15–340–163, Fisher Scientific, Waltham, MA) in Bead Mill Tubes (15–340–154 Fisher Scientific, Waltham, MA) with 1 mL Trizol (15,596,026, Thermo Fisher Scientific, Waltham, MA). Tissue debris was pelleted and removed by centrifugation at 12,000 × g for 10 minutes at 4°C. After adding 200 uL 1-Bromo-3-chloropropane, samples were manually shaken for 20 seconds followed by incubation at room temperature for 3 minutes. Phase separation was conducted by centrifugation at 12,000 × g for 18 minutes at 4°C. The RNA containing aqueous layer was retained and subsequently mixed with 500 uL 200 proof ethanol. The mixture was then passed through the RNeasy Mini Kit (74,104, Qiagen, Germantown, MD) RNA binding column, followed by washing and elution steps described in the manufacturer’s handbook.

Quantitative RT-PCR

Total RNA was quantified using NanoDrop Spectrophotometer ND 1,000 (NanoDrop Technologies) before RNA (0.5 µg) was reverse transcribed to generate cDNA using the Transcriptor First Strand cDNA Synthesis Kit (04,379,012,001, Roche Life Science, Penzberg, Germany). Quantitative real-time PCR analysis was performed using TaqMan Universal Master Mix II (Life Technologies, Carlsbad, CA), SsoAdvanced Universal SYBR Green Supermix (1,725,270, Bio-Rad, Hercules, CA), or SsoAdvanced Universal Probes Supermix (1,725,280, Bio-Rad, Hercules, CA) on the CFX Connect Real-Time PCR Detection System (1,855,201, Bio-Rad, Hercules, CA) based on the manufacturer’s instruction. Ribosomal RNA (18S) was used as an internal control. The 18S rRNA probe is from ThermoFisher Scientific (4319413E, ThermoFisher Scientific, Waltham, MA). Primer sequences are listed below:

• Pde5a_F: attagaaaggccaccagagacat

• Pde5a_R: aagagcaggactcggtatgg

• Plcb4_F: agtgacaggagcctgctttg

• Plcb4_R: aaatcctgcacagtccccag

• Cacna1c_F: ccaacctcatcctcttcttca

• Cacna1c_R: acatagtctgcattgcctaggat

• Tagln.F: gcccagacaccgaagcta

• Tagln.R: gtaggatggacccttgttgg

• Acta2.F: taacccttcagcgttcagc

• Acta2.R: acatagctggagcagcgtct

• Actg2.F: aggacttctcacacccttgg

• Actg2.R: gtggtctcttcttcacacatgg

• Myh11.F: tggaggccaagattgcac

• Myh11.R: ggccgcctgtttctctct

• Myocd.F: gcaagggcagaaacaggtc

• Myocd.R: atctgagcagttggaatggac

• Cnn1.F: cggcttgtctgctgaagtaa

• Cnn1.R: accccctcaatccactctct

• Ctgf.F: tgacctggaggaaaacattaaga

• Ctgf.R: agccctgtatgtcttcacactg

RNA-seq

The PPD6 RNA libraries were sequenced with a NovaSeq 6000 system [Illumina] with ∼102.4 to 155.4 million, 76-bp paired-end reads for each sample. The raw reads were initially processed by filtering with average quality scores of 20. The reads passing the initial processing were aligned to the mouse reference genome (mm10; Genome Reference Consortium Mouse Build 38 from 2011 December) using HISAT2 version 2.2.1 with “—very-sensitive” option (57). The duplicated reads with the same sequence were removed. Then, the uniquely mapped reads were normalized to the same number of read pairs (45 million) in each sample. Differential gene expression analysis for paired samples was performed using a Bioconductor package edgeR version 3.15 (58). edgeR requires an input of gene expression matrix with read count, which was calculated by counting the total number of paired-end reads mapped to each gene with Python software HTSeq version 0.10.0. Only the genes with at least one sample count larger than 100 were included.

The libraries separately prepared from the extracted 13.5 dpc RNA samples were sequenced with 43.8∼83.8 million, 76-bp paired-end reads per sample on a Nextseq 500 system [Illumina]. The raw reads were filtered by removing low quality reads (average quality scores < 20), followed by removing adapter-carrying reads using cutadapt (v2.7) (59). The reads were further cleaned by decontaminating rRNA reads through mapping to the mouse rRNA sequences using bowtie (v2.2.5) (60). Then the processed reads were normalized to the same number of reads for aligning to mm10 genome using Spliced Transcripts Alignment to a Reference (STAR) software version 2.7.0f (61). The number of fragments (read pairs) that mapped to the gene models were counted using featureCounts (v1.6.4) function available in the software package Subread (62). Differential expression analysis was performed using R package DESeq2 (v1.24.0) (63). DEGs were determined using the combined cutoffs: (1) detected with mean FPKM ≥ 1 in at least one of the conditions; (2) absolute fold change of ≥ 1.3; and (3) unadjusted P-value < 0.05. The RNAseq data has been deposited to NCBI Gene Expression Omnibus repository.

ChIP-seq and chromatin conformation capture

Uteri were collected from 13.5 dpc C57BL/6 J mice. Conceptus, placenta and endometrium were removed surgically. Pools of uteri from three animals were used for ChIP and HiC assays. ChIP library was prepared by Active Motif (Calsbad, CA). HiC library was prepared by Arima Genomics (Calsbad, CA). Library sequencing was performed on NOVAseq 6,000 by the NIEHS Epigenome Core Lab.

The raw ChIP-seq reads (50-bp, single-end) were filtered with average quality scores greater than 20. Then, the reads were mapped to the mouse reference genome (mm10; Genome Reference Consortium Mouse Build 38 from 2011 December) using Bowtie version 1.1.2 (64) with unique mapping and allowing up to 2 mismatches for each read (-m 1 -v 2). The duplicated reads with the same sequence were removed. The ChIP-seq peaks were identified using MACS2 with a cutoff of adjusted P-value less than 0.0001 (65). The H3K27ac ChIP-seq peaks were defined as H3K27ac-positive enhancers. The enhancers within 12.5 Kb were merged. The combined enhancer regions were called super enhancer if they were larger than 15Kb. Each ChIP-seq peak was mapped to nearby gene within 100 Kb of TSS using the “annotatePeaks.pl” function of HOMER (Heinz et al., 2010).

There were 751.6 million, 51-bp paired-end reads generated from HiC library. Those raw reads were mapped to mouse reference genome (mm10; Genome Reference Consortium Mouse Build 38 from 2011 December) using HiCUP version 0.7.1 (66). The uniquely mapped di-tag passing quality filtering with distance larger than 10 Kb were used for downstream analyses. Chromatin loops were identified by Juicer version 1.8.9 with default parameters (67). The A/B compartments were predicted at 25 Kb resolution using “eigenvector” function of Juicer.

Gene ontology and pathway enrichment analyses

DEGs were mined by BaseSpace Correlation Engine (Illumnia, San Diego, CA), IPA (Qiagen, Germantown, MD), and the Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.8 for enrichment of biological processes (68).

Gene signature analysis

The publicly available human gene expression datasets GSE137552 (11), the Mittal dataset (courtesy of Dr. Mesiano for the original data) (29), GSE50599 (28), GSE80172 (31), and GSE202028 (30) were scored for manifestations of the mouse model-derived PGR gene signature. The resulting T-scores serve as quantitative surrogates for the inferred PGR activity of individual specimens in a relative manner among samples within each dataset. The T-score calculation uses the gene expression matrix of the specimens normalized within individual datasets and the mouse PGR gene signature list that is converted to a format of up (positively regulated by PGR) and down (negatively regulated by PGR) based on the 1,535 DEGs in Dataset S2. Within each human dataset, where multiple probes or transcripts referred to the same gene, the ones with the highest variation were taken to represent the gene; and the expression values of each gene are centered to the median across all samples. The SEMIPs R Shiny application (27) was used to calculate the T-scores with the formula T-score = d*TINV(p, df). d equals to 1 if the average human mRNA abundance of the “up” mouse genes is greater than the average human mRNA levels of the “down” mouse genes, otherwise d equals to -1. TINV is the function for the t-distribution. p is the two-tailed t-test P-value of the human mRNA abundance of the “up” mouse genes and the human mRNA levels of the “down” mouse genes assuming equal variance. df is the degrees of freedom as represented by the total number of the human genes in correspondence to mouse signature genes minus 2.

Statistical analysis

Statistical analysis was performed with the R commander package (R-Studio, Inc., Boston, MA) or GraphPad Prism 8 (GraphPad Software, San Diego, CA) unless specified otherwise. Data are presented as the mean ± SEM. P-values less than 0.05 were defined as statistically significant (*P < 0.05; **P < 0.01; ***P < 0.001).

Notes

Competing Interests: The authors declare no competing interest.

Funding

This work is supported by an Intramural Research Program of the National Institute of Environmental Health Sciences, National Institutes of Health (NIH) grants Z1AES103311 (F.J.D.) and Z99-ES999999 (S.P.W.), and an NIH extramural grant R01EB027099 (I.V.L.).

Authors’ contributions

S.P.W. and F.J.D. designed the project and wrote the manuscript; Z.C.Y. and I.V.L. conducted the microCT experiment; M.C.P. performed physiology assays and associated data analysis; S.P.W. and X.L. performed phenotyping and molecular assays; and T.W., L.Z., and S.P.W. conducted bioinformatic analyses.

Data availability

RNA-Seq, ChIP-seq, and HiC data were deposited in the NCBI Gene Expression Omnibus (GEO) under the accession number GSE199104.