To better understand the regulatory network shaping the myometrial transcriptome before labor, we analyzed transcriptome and putative enhancers in individual human myometrial specimens. Using RNA-seq, we identified actively expressed RNAs, while ChIP-seq for H3K27ac and H3K4me1 was used to map putative enhancers (Supplementary file 1). Active genes were associated with nearby putative enhancers based on their genomic proximity. Additionally, chromatin looping patterns were mapped using Hi-C to further link active genes and putative enhancers within the same chromatin loops (Supplementary file 1 and Supplementary file 2). The ChIP-seq assay identified an average of 44,238 H3K27ac and 74,325 H3K4me1-positive genomic regions in three term-pregnant, nonlabor myometrial biopsies (Table 1). An RNA-Seq assay revealed an average of 12,157 active genes in these specimens that manifested expression levels of fragments per kilobase of transcript per million mapped reads (FPKM) greater than or equal to 1 (Table 1). A High-throughput Chromosome Conformation Capture (Hi-C) assay further found a total of 27,162 chromatin loops from two of the three myometrial specimens (Table 1, subject 1 and 3). We failed to identify chromatin loops in the second subject’s biopsy due to limited sample availability. Together, these results delineate a map of H3K27ac and H3K4me1-positive signals in the human term pregnant myometrial tissue before the onset of labor, which we use as a resource to investigate the molecular mechanisms that prepare the myometrium for subsequent parturition.

When comparing the present study to previous findings (Dotts et al., 2023), 8563 genomic regions carry common H3K27ac-positive histone marks. However, significant variations in the number and location of H3K27ac-positive intervals are present across the six samples among these two studies (Figure 1—figure supplement 1). Depending on the number of mapped intervals, the 8563 common regions constitute between 20.7% and 71.5% of total H3K27ac-positive intervals across the six myometrial specimens (Figure 1—figure supplement 1). Notably, the fraction of intervals commonly identified in the present study’s three specimens ranges between 56.4% and 70.0% of each of their total H3K27ac-positive regions, while the Dotts dataset has a wider range between 28.0% and 74.0%. With a less stringent criterium, most of the H3K27ac-positive intervals are found in at least two samples, ranging between 75.4% and 98.3% (Figure 1—figure supplement 1). These data together highlight the degree of variation in mapping the epigenome among specimens and datasets.

Since H3K27ac and H3K4me1 marked genome regions are associated with enhancers for gene regulation (Creyghton et al., 2010; Heintzman et al., 2007), we define the regions that have overlapping H3K27ac and H3K4me1 marks as putative myometrial enhancers at the term pregnant nonlabor stage (Supplementary file 3). Among the three specimens, 13,114 putative enhancers are commonly present (Figure 1A). A significant variance in the location of myometrial putative enhancers is also seen among biopsies as evident by the observation that these common putative enhancers make up less than half of the total putative enhancers in each individual specimen (Figure 1A). More than one-third of these 13,114 common putative enhancers are located within a 100-kilobase vicinity of actively high-expressing genes (FPKM≥15), while a much smaller fraction of them (13.3%) are associated with actively low-expressing genes (Figure 1B). Motif enrichment analysis on these common putative enhancers further reveals an overrepresentation of binding motifs for myometrial transcription factors AP-1, STAT5, and NFκB, steroid hormone receptors PGR and NR3C1, and smooth muscle transcription regulators SRF and ELK1 (Figure 1C, Supplementary file 4). These findings collectively suggest that the putative myometrial enhancers bring together smooth muscle and hormonal control programs for the regulation of myometrial gene expression at term pregnancy.

Super enhancers house hormone-dependent gene regulatory programs for female reproductive tract homeostasis (Shin et al., 2016; Hewitt et al., 2020). Across the three human subjects, 540 putative super enhancers are commonly identified in the term pregnant nonlabor myometrial specimens (Figure 2A, Supplementary file 5). More than 40% of the 540 putative super enhancers are located within a 100 kilobase distance to high-expressing genes (FPKM ≥ 15), while only 7.3% of putative myometrial super enhancers are found near low-expressing genes (5>FPKM ≥ 1; Figure 2B). Compared with the regular putative enhancers, the putative myometrial super enhancers are found more frequently near active genes that are expressed at relatively higher levels (Figures 1B and 2B). Notably, 76% of the putative super-enhancers co-localize with known PGR-occupied regions in human myometrial tissue, compared to 20% co-localization observed in regular enhancers. Further examining the myometrial active genes that are associated with putative super enhancers revealed an enrichment of gene functions in cytoskeleton organization, extracellular-receptor interaction, transcription regulation, mRNA stability, and estrogen signaling (Figure 2C, Supplementary file 6). Taken together, these results establish the association among the putative myometrial enhancers, the potential regulatory program within these super enhancers, and the cellular functions of the genes they may control.

We previously demonstrated the positive correlation of PLCL2 and PGR expression in a mouse model and PLCL2’s function on negatively modulating oxytocin-induced myometrial cell contraction (Peavey et al., 2021). However, the mechanism that underlies PGR’s regulation of PLCL2 remains unclear. Taking advantage of the mapped myometrial cis-acting elements, we aimed to identify the cis-acting elements that may contribute to the PLCL2 transcriptional regulation with a special interest in the PGR-related enhancers.

The Hi-C assay identified a chromatin loop with one end at the PLCL2 transcription start site (TSS) and the other end at approximately 90 kilobases upstream of the TSS (Figure 3A). Within this chromatin loop, ChIP-Seq results revealed seven genomic regions that are marked with H3K27ac in all three myometrial specimens (Figure 3A). Six of the seven regions are also co-localized with previously published genome occupancy of transcription regulators curated by the ReMap Atlas (Hammal et al., 2022; Figure 3A). For the purpose of functional screening, we focus on H3K27ac signals instead of using the H3K27ac/H3K4me1 double positive criterium to cast a wider net. Eight guide RNAs (gRNAs) were designed to target these seven candidate regions for CRISPR activation (CRISPRa)-based screening of PLCL2 gene regulation (Figure 3A). The two previously reported gRNAs (PLCL2-7 and PLCL2-8) were able to elevate the endogenous PLCL2 mRNA levels significantly higher than the non-targeting gRNA in immortalized human myometrial cells (hTERT-HM; Figure 3B; Peavey et al., 2021). gRNAs PLCL2-1, PLCL2-3, PLCL2-5, and PLCL2-6 also significantly increased PLCL2 expression above the levels of the non-targeting group (Figure 3B). Moreover, results from the Perturb-Seq assay found that the PLCL2 gene ranks high in the differentially expressed gene list of the PLCL2-5 gRNA expressing cells compared with the non-targeting gRNA expressing cells (Supplementary file 7 and Supplementary file 8), in line with the qRT-PCR assay finding in which the gRNA PLCL2-5 induced PLCL2 expression to the greatest extent among the upstream non-coding genome targeting gRNA (Figure 3B). Moreover, gRNA PLCL2-5 was capable of mediating the increase of the PLCL2 transcript abundance in another uterine mesenchymal cell type T-HESC (Figure 3C). Collectively, these findings support the PLCL2-5-targeted genomic region as a cis-acting element for regulation of the PLCL2 gene.

After determining its cis-acting element role, the PLCL2-5 targeted genomic region was further examined to identify upstream regulators that control myometrial PLCL2 expression. This was achieved by first identifying transcription factor occupancy in this genome region in any tissues/cells that are documented in public databases (ReMAP Atlas; Hammal et al., 2022), followed by filtering for factors that have known mRNA and/or protein expression in hTERT-HM cells and tissues using publicly available transcriptomic and histological data including the Human Protein Atlas (Human Protein Atlas proteinatlas.org; Uhlén et al., 2015) and published RNA-Seq datasets (Stanfield et al., 2019; Wu et al., 2020). The resulting candidate upstream regulators for the myometrial PLCL2 gene are documented in Table 2. Since the PLCL2 mRNA abundance is much lower in the hTERT-HM cells than in the myometrial tissues, these candidate upstream regulators were further grouped into candidate activators and repressors based on an assumption that a candidate activator would be expressed higher in human myometrial tissue, and a candidate repressor being expressed higher instead in the hTERT-HM cells, holding the PLCL2 mRNA levels down. Candidate activators for the PLCL2 gene include known myometrial regulators such as PGR, estrogen receptor alpha (ESR1), and AP-1 (Table 2). Interestingly, the mediator complex protein subunit MED1, cohesion complex members RAD21 and SMC3, and many epigenomic modifiers are among the list of the candidate repressors (Table 2). These findings not only provide candidates for subsequent functional assessment, but also highlight potential pathways for future investigations on the regulation of the contractile gene PLCL2 expression in the myometrium.

We previously demonstrated the regulation of mouse Plcl2 gene by the myometrial PGR (Wu et al., 2022). To further test whether such a regulatory mechanism is also present in humans, the CRISPRa technology was employed to increase the expression of PGR mRNA and proteins in hTERT-HM cells (Figure 4A and B). Given that the PGR640 gRNA led to higher PGR protein production (Figure 4B), this gRNA was utilized to stimulate myometrial PGR expression for subsequent experiments. Indeed, RT-qPCR results showed that PGR overexpression increased PLCL2 mRNA abundance in hTERT-HM cells (Figure 4C), demonstrating a consistent finding with the mouse model.

To further explore this regulatory relationship in the human myometrial tissue, we examined the degree of correlation between the inferred PGR activity and the PLCL2 transcript abundance in myometrial biopsies. Gene expression profiles of myometrial specimens from 13 proliferative-phase, 6 secretory-phase, 14 post-menopausal, and 10 Provera-treated human subjects were determined by the RNA-Seq assay (Supplementary file 9) to take advantage of the wide spectrum of progesterone signaling activities in these samples. Inferred myometrial PGR activities, represented as T-scores (Supplementary file 9), were derived from this normalized gene expression matrix of 43 human myometrial specimens with the previously published mouse myometrial Pgr gene signature (Wu et al., 2022; Li et al., 2021). A Pearson correlation analysis found a positive correlation (r=0.47) between the T-scores and normalized PLCL2 mRNA levels in these 43 specimens (Figure 4D). Together, these in vitro and in vivo findings collectively support that PGR is a PLCL2 activator, likely acting through the 35 kb upstream cis-acting element.

Permission to collect human tissue specimens was prospectively obtained from individuals undergoing hysterectomy or cesarean section for benign clinical indications (H-33461). Gravid myometrial tissue was obtained from the margin of the hysterotomy in women undergoing term cesarean sections (>38 weeks estimated gestational age) without evidence of labor. Non-gravid myometrial tissue was collected from pre-menopausal women undergoing hysterectomy for benign conditions. Specimens from gravid women receiving treatment for pre-eclampsia, eclampsia, pregnancy-related hypertension, or pre-term labor were excluded.

Human immortalized myometrial cells (hTERT-HM) (Condon et al., 2002) cells and human immortalized endometrial cells (T-hESC; ATCC, CRL-4003) were cultured in DMEM/F-12 (Invitrogen, Grand Island, NY, USA; Cat. No. 11320033.) supplemented with 10% Fetal Bovine Serum (FBS, Gibco; Cat. No. 26140079) and antibiotics (10 000 IU/mL penicillin, 10 000 IU/ mL streptomycin; Life Technologies, Grand Island, NY). Cell culture media was filtered using the 0.22 µm Rapid-Flow Sterile Disposable Filter Units (Nalgene). Cells were incubated at 5% CO₂ and 37 °C.

RNA was isolated from cells using TRIzol Reagent (Invitrogen; Cat. No. 15596026) and the RNeasy mini RNA isolation kit (QIAGEN, Hilden, Germany). cDNA was prepared using Moloney Murine Leukemia Virus reverse transcriptase (Thermo Fisher Scientific; Cat. No. 28025013) with Random Hexamers (Invitrogen, Waltham, MA, USA) according to manufacturer protocol. For quantitative analysis of mRNA, SsoAdvanced Universal SYBR Green Supermix (Bio-Rad, Hercules, CA, USA; Cat. No. 1725271) was used according to manufacturer instructions. Relative mRNA expression was determined by standard curve-based method (Larionov et al., 2005). Relative expression of genes of interest was normalized to the 18 S rRNA. Briefly, reaction samples were prepared to a volume of 20 µl with 5 uM of both forward and reverse primers, cDNA, and a final 1 x concentration of the SYBR Green Supermix. The following primers were used: PGR forward primer: TTTAAGAGGGCAATGGAAGG; PGR reverse primer: CGGATTTTATCAACGATGCAG; PLCL2 forward primer: TATGACATGATGATTCAGTCCCTC; PLCL2 reverse primer: TTCCTTGGTGCCTATGCTGT; 18 S forward primer: GTAACCCGTTGAACCCCATT; 18 S reverse primer: CCATCCAATCGGTAGTAGCG. The reaction was heated to 95 °C for 30 s, followed by 39 cycles of denaturation at 95 °C for 15 s and annealing and elongation at 60 °C for 30 s. Temperature cycles were performed on the CFX Connect Real-Time PCR Detection System (Bio-Rad). Experiments were performed in three biological replicates and two technical replicates, where biological replicates describe independent samples, and technical replicates describe repeated measurements on the same biological samples.

Protein was isolated from cells using the RIPA Lysis and Extraction Buffer (Thermo Scientific; Cat. No. 89900) with the following specifications: approximately 300,000 cells were pelleted and lysed with 100 µl of complete Pierce RIPA buffer. Protein was quantified using the BCA Protein Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA) as instructed by the manufacturer, and 40 µg of protein were loaded per lane on a Mini-PROTEAN TGX Precast Protein gel (Bio-Rad; Cat. No. 4568094) with the Precision Plus Protein Dual Color Standards Ladder (Bio-Rad; Cat. No. 1610374). Protein lysates from MCF7 cells were used as positive controls, while protein lysates from HEK293T cells were used as negative controls. Gels were transferred to nitrocellulose membrane using the Turbo-Blot transfer system (Bio-Rad) according to the manufacturer’s instructions. The membrane was blocked with 5% milk (Santa Cruz Biotech, Santa Cruz, CA, USA) in TBST 20 mM Tris, pH 7.4 (Lonza, Morrisville, NC, USA), 140 mM NaCl (Lonza), 1% TWEEN-20 (Sigma). PGR protein was detected using Monoclonal Mouse Anti-Human Progesterone receptor clone PgR 1294 (Dako; GA09061-2). B-actin protein was detected using Rabbit polyclonal Actin Antibody (I-19)-R (Santa Cruz Biotech, Dallas, TX, USA; sc-11616-R; Lot#DO406). Both antibodies were diluted 1:1000 in milk and incubated with blots overnight at 4 °C. Bands were detected using Donkey Anti-Mouse (LI-COR Biosciences, Lincoln, NE, USA; 926–32212; Lot # C60524-15) and Goat anti-rabbit (LI-COR Biosciences; 926–32211; Lot # DOO304-15) diluted 1:20,000 in milk and incubated with blots for 45 min at room temperature. Blots were imaged using Odyssey Fc Imager (LI-COR Biosciences) using the 800 nm channel for 10 min and the 700 nm channel for 30 s (to image ladder).

RNA-Seq libraries were prepared according to the Illumina TruSeq Stranded mRNA protocol Document # 1000000040498 v00. Libraries were sequenced using the Illumina NextSeq 500 and NovaSeq 6000 systems.

The ChIP-Seq assay was performed by the Active Motif service laboratory using snap-freeze human myometrial specimens. The H3K27Ac ChIP reactions were conducted with 10 μg of tissue chromatin and 4 μg of H3K27Ac antibody (Active Motif; Cat. No. 39133). The H3K4me1 ChIP reactions were carried out by using 10 µg of tissue chromatin and 4 µl of H3K4me1 antibody (Active Motif, Cat. No. 39297). Libraries were prepared by a custom Illumina library with the standard Illumina PE adaptors (Short et al., 2008). Libraries were sequenced using the Illumina NextSeq 500 and NovaSeq 6000 systems.

Snap-freeze human myometrial specimens were shipped to the Arima Genomics and the Active Motif service laboratories for preparation of the Hi-C library using the Arima-HiC Kit (Arima Genomics A510008), protocol version A160132 v00. Libraries were sequenced in the NIEHS on the Illumina NovaSeq 6000 platform.

All lentiviruses were packaged in HEK293T/17 cells (ATCC # CRL-11268) according to published method in Current Protocols in Neuroscience (Chen et al., 2019). Briefly, 293T cells were transiently transfected with pMD2G, psPAX2, and a transfer vector containing the desired gene using Lipofectamine 2000. Supernatant was collected 48 hr post-transfection and concentrated by centrifugation at 50,000 × g for 2 hours. Pellets were resuspended in PBS and used for infection. To determine titer, HEK293T/17 cells were infected with lentiviral samples. Five days post-infection, Qiagen DNeasy Blood & Tissue Kits was used to isolate chromosomal DNA from infected cells. All titers were determined by performing droplet digital PCR (ddPCR) to measure the number of lentiviral particles that integrated into the host genome.

hTERT-HM cells were infected with lentiviral gRNA expression vectors with a multiplicity of infection of 4 (MOI = 4) and with dCas9-VPR-mCherry at an MOI = 4. gRNA expression vector and dCas9-VPR-mCherry expression vectors, following lentiviral transduction, confer GFP and mCherry fluorescent markers, respectively.

Cells were examined using a BD FACSAria II cell sorter (Becton Dickinson Biosciences, San Jose, CA) equipped with FACSDiVa software. Initially, a ‘scatter’ gate was set on a forward scatter (FSC-A) versus side scatter (SSC-A) dot plot to isolate the principal population of cells free of debris. Subsequently, cells were consecutively gated on a side scatter height (SSC-H) versus width (SSC-W), then a forward scatter height (FSC-H) versus width (FSC-W) dot plot to isolate single cells. GFP-expressing cells were excited off a 488 nm laser and detected using a 525/50 nm filter. mCherry-expressing cells were excited off a 561 nm laser and detected using a 610/20 nm filter. Double positive cells were collected based on non-transfected controls and used for further cultural/biochemical analysis.

The cells in suspension were counted and examined for viability with trypan blue staining using a TC-20 cell counter (Bio-Rad). Approximately 16,500 live cells at 1×10⁶ cells/ml concentration with 65% or above viability were loaded into the Single Cell Chip to generate single cell emulsion in Chromium Controller with Chromium Single Cell 3’ Library & Gel Bead Kit v3.1 (10x Genomics, Cat. 1000268). Reverse transcription of mRNA and cDNA amplification were carried out following the manufacturer’s instruction (10 x Genomics, Cat. 1000268, Cat. 1000262 with 10 x Genomics protocol CG000316). The amplified cDNA was separated into CRISPR-sgRNA-derived cDNA and transcriptome-derived cDNA. The CRISPR-sgRNA-derived cDNA was used to make NGS sequencing libraries. The transcriptome-derived cDNA was further fragmented to construct NGS libraries. Both libraries were then sequenced together with the molar ratio of 1–4 by the NIEHS Epigenomics and DNA Sequencing Core Laboratory with the parameters recommended in the manufacturer’s instruction.

The raw sequencing FASTQ files generated from both the transcriptome and CRISPR screening libraries were processed together by Cell Ranger software (version 4.0.0, 10 X Genomics). The ‘cellranger count’ pipeline used STAR for aligning the reads to the human reference, GRCh38 ‘refdata-gex-GRCh38-2020-A’ (10 X Genomics), and associated gene expression profile with guide RNA (gRNA) identity by unique barcode in each cell. Seurat software (version 3.6.3) was utilized to perform clustering analysis on the combined dataset (Satija et al., 2015). We applied the SCTransform package to normalize gene expression counts across cells (Hafemeister and Satija, 2019). The cells were clustered based on the number of unique gRNA types detected. The cell populations containing more than one type of gRNA were excluded from further analyses. We applied two approaches to determine the association between gRNAs and target genes:

The raw ChIP-seq reads (51 bp, single-end) were filtered with average quality scores greater than 20. Adapter sequences were trimmed from reads using cutadapt (v1.12). Then the reads were aligned to the human reference genome (hg38) using Bowtie (v1.1.2) (Langmead et al., 2009), requiring unique mapping and allowing up to 2 mismatches per read (-m 1 v 2). Duplicated reads with identical sequences were removed using Picard tools. To visualize the read coverage, BigWig files were generated from the bedgraph files of each sample using bedGraphToBigWig. These bigWig files were displayed as custom tracks on the UCSC genome browser.

The normalization of sequencing depth across all ChIP-seq datasets was achieved by downsampling to 20 million uniquely mapped reads per sample. The initial peaks of each sample were identified using MACS2 with a cutoff of adjusted p-value < 0.0001 (Zhang et al., 2008). The gene associated with each peak was predicted by searching the TSS of nearby genes within a 100 Kb range using HOMER (Heinz et al., 2010). The summary table is available in Supplementary file 1.

Additional H3K27ac ChIP-seq and related input data were obtained from GEO (GSE202027; patients 1, 4, and 7). The same ChIP-seq analysis pipeline described above was applied to those samples. To obtain a comprehensive set of peaks across all H3K27ac ChIP-seq datasets, the peak intervals from each dataset were merged using BedTools.

H3K27ac-positive enhancers were defined as regions of H3K27ac ChIP-seq peaks in each sample. The enhancers within 12.5 Kb were merged by using the bedtools merge function with the parameter ‘-d 12500’. The combined enhancer regions were called super enhancers if they were larger than 15 Kb. The common super enhancers from multiple samples were used for downstream analysis (Wang et al., 2008).

Colocalization of super enhancers and PGR genome occupancy was compared by calling peaks from previously published PGR ChIP-seq data (GSM4081683 and GSM4081684). The percentages of enhancers and super enhancers that manifest PGR occupancy were calculated by overlapping the genomic regions in each category with PGR occupancy regions.

The raw paired-end reads (51 bp, paired-end) were initially processed by applying a filter with an average quality score of 20. Adaptor sequences were subsequently removed from reads using cutadapt version 1.12. The processed reads were then aligned to the human reference genome (hg38) using the STAR aligner version 2.5.2b. The gene expression levels in each sample were determined by counting the total number of paired-end reads mapped to each gene using DESeq2 R package version 1.12.4 (Love et al., 2014). The gene expression count matrix was normalized based on the ratio of total mapped read pairs in each sample to 56.5 million. Only the genes with average normalized count across all samples larger than one were used for further analysis. The Bioconductor package edgeR was applied to the gene expression count matrix to detect differentially expressed genes between groups of interest (Robinson et al., 2010). The false discovery rate (FDR) of differentially expressed genes was estimated using the Benjamini and Hochberg method.

The raw paired-end reads (76 bp, paired-end) were processed by applying a filter with an average quality score of 20. Adapter sequences were removed from reads using cutadapt (v1.12). The processed reads were then aligned to the human reference genome (hg38) using the STAR aligner (v2.5.2b). Fragments Per Kilobase Million (FPKM) was calculated for each gene in each sample by using Cufflinks version 2.0.2.

The raw reads (51 bp, paired-end) generated from the HiC library were mapped to the human reference genome (hg38; Genome Reference Consortium Human GRCh38 from December 2013) using HiCUP version 0.7.1 (Wingett et al., 2015). The uniquely mapped di-tag passing quality filtering with distance larger than 10 Kb was used for downstream analysis. Chromatin loops were identified using the ‘hiccups’ function of Juicer version 1.8.9 with default parameters (Durand et al., 2016). The A/B compartments were predicted at 100 Kb resolution using the ‘eigenvector’ function of Juicer. The Hi-C summary table is available in Supplementary file 1, and the quality control data are provided in Supplementary file 2.

Enriched motifs were identified by HOMER (Hypergeometric Optimization of Motif EnRichment) v4.11 with default background sequences matching the input sequences (Heinz et al., 2010).

The inferred PGR activities were represented by the T-score, which was derived by inputting the mouse myometrial Pgr gene signature, based on the differentially expressed genes between control and myometrial Pgr knockout groups at mid-pregnancy (Wu et al., 2022), into the SEMIPs application (Li et al., 2021). The T-scores were computed using this signature alongside the normalized gene expression counts (FPKM) from 43 human myometrial biopsy specimens.