Impact of sustained TGFβ receptor inhibition on chromatin accessibility and gene expression in cultured human endometrial MSC

Endometrial mesenchymal stem cells (eMSC) drive the extraordinary regenerative capacity of the human endometrium. Clinical application of eMSC for therapeutic purposes is hampered by spontaneous differentiation and cellular senescence upon large-scale expansion in vitro. A83-01, a selective transforming growth factor-β receptor (TGFβ-R) inhibitor, promotes pharmacological expansion of eMSC in culture by blocking differentiation and senescence, but the underlying mechanisms are incompletely understood. In this study, we combined RNA-seq and ATAC-seq to study the impact of sustained TGFβ-R inhibition on gene expression and chromatin architecture of eMSC. Treatment of primary eMSC with A83-01 for 5 weeks resulted in differential expression of 1,463 genes. Gene ontology analysis showed enrichment of genes implicated in cell growth whereas extracellular matrix genes and genes involved in cell fate commitment were downregulated. ATAC-seq analysis demonstrated that sustained TGFβ-R inhibition results in opening and closure of 3,555 and 2,412 chromatin loci, respectively. Motif analysis revealed marked enrichment of retinoic acid receptor (RAR) binding sites, which was paralleled by the induction of RARB, encoding retinoic acid receptor beta (RARβ). Selective RARβ inhibition attenuated proliferation and clonogenicity of A83-01 treated eMSC. Taken together, our study provides new insights into the gene networks and genome-wide chromatin changes that underpin maintenance of an undifferentiated phenotype of eMSC in prolonged culture. Significance statement Cycling human endometrium is a rich source of adult stem/progenitor cells that could be exploited for clinical purposes. Small molecules, such as A83-01, that modulate cell identity may open new avenues to maintain the functional properties of eMSC upon expansion in culture. By integrating complementary genome-wide profiling techniques, we mapped the dynamic changes in chromatin landscape and gene expression in response to prolonged A83-01 treatment of eMSC. Our findings provide new insights into the mechanisms of action of TGFβ-R inhibition that may lead to the development of more targeted pharmacological approaches for MSC expansion.


Introduction
The human endometrium is a highly dynamic tissue that generates 4-10 mm of mucosa in each menstrual cycle 1 . This extraordinary regenerative capacity is mediated by resident epithelial progenitors and mesenchymal stem/stromal cells (MSC) 2, 3 . Cultured endometrial MSC (eMSC) are clonogenic, highly proliferative, multipotent, and express the International Society for Cellular Therapies (ISCT) surface markers. 4 Human eMSC can be purified as CD140b + CD146 + cells 5 , or by means of the surface marker SUSD2 (Sushi Domain-containing 2, formerly W5C5). 6 While eMSC also express Stro-1, this marker does not enrich for clonogenic cells. 7 eMSC further express MSCA-1 8 and foremost reside in the perivasculature of spiral arterioles and venules. 9-11 A major advantage of the endometrium over other sources of adult MSC, such as bone marrow or adipose tissue, is its accessibility. Endometrial sampling is a routine office-based procedure that does not require anesthesia. 9 Preclinical animal studies showed that eMSC are a promising cell source for treatment of gynecological disorders, including pelvic organ prolapse. 12,13 For example, eMSC seeded or bio-printed onto meshes with biomechanical properties matching the human vagina (e.g. non-degradable polyamide/gelatin composite meshes), 14, 15 16 , or on degradable nanofibers, 17 promote angiogenesis, collagen deposition, and cellular infiltration into biomaterials when transplanted into rodent or ovine models. They also elicit an early inflammatory response, characterized first by influx of M1 macrophages, which then switch to the M2 wound healing phenotype. 18,19 In vitro, eMSC seeded on polyamide/gelatin meshes differentiate into smooth muscle cells and fibroblasts; cell types that are important for restoring vaginal structure and function. 20 Hence, the endometrium is a rich source of MSC for autologous and allogenic cellbased therapies, including pelvic organ prolapse, urinary incontinence, and regeneration of scarred endometrium in women with Asherman's syndrome. 2,21,22 5 Clinical application of eMSC requires expansion of cells in culture. 12,23 As is the case for MSC from other sources, 24 eMSC cultured over several passages differentiate spontaneously and are subjected to replicative stress caused by telomere shortening. 23,25,26 Consequently, the cells lose their proliferative capacity as well as the ability to reconstitute tissue in vivo. 24 To overcome this impediment to the clinical translation, we and others have explored the use of small molecules to maintain the in vivo properties of eMSC and MSC in prolonged culture. 23,25,27 We recently reported that A83-01, a selective inhibitor of TGF-β type I receptor (TGFβ-R) ALK4, 5 and 7 kinase, increases proliferation and inhibits apoptosis and senescence of cultured eMSC, thereby safeguarding their functional properties in vitro 23 and prolonging their survival in vivo. 28 Further, A83-01-treated eMSC exhibit increased angiogenic activity and express a proangiogenic, antifibrotic, and immunomodulatory gene profile. 23,25 The mechanisms underlying pharmacological expansion of cultured eMSC in response to sustained TGFβ-R inhibition are incompletely understood. Adult stem/progenitor cells have more open chromatin than their differentiated progeny, but less than pluripotent stem cells.
Further dynamic chromatin changes underpin subsequent differentiation into mesodermal lineages. 29 Based on these observations, we hypothesized that sustained TGFβ-R inhibition leads to divergence in the chromatin landscape of cultured eMSC and activation of transcriptional regulatory circuitries that maintain the cells in a more naïve or undifferentiated state. To test this hypothesis, eMSC cultured with or without A83-01 were subjected to integrated RNA-sequencing (RNA-seq) and Assay for Transposase Accessible Chromatin through sequencing (ATAC-seq).

Endometrial biopsies
The study was approved by the Monash Health and Monash University Human Research Ethics committees. Endometrial biopsies were obtained from seven pre-menopausal women, without endometrial pathologies, following written informed consent and according to The Declaration of Helsinki (2000) guidelines. Participant information is kept confidential and samples deidentified prior to use. None of the participants received hormonal treatment within three months prior to the biopsy.

eMSC isolation and culture
Endometrial biopsies were processed and single-cell suspensions of eMSC obtained as described previously. 6 Briefly, finely minced endometrial tissue was enzymatically and  The cells   7 were washed, re-suspended in separation buffer 500 µL and applied to a Miltenyi column (Miltenyi Biotech) in a magnetic field. Columns were washed three times with buffer.
Magnetically labelled SUSD2 + eMSC were eluted with buffer and cell number was determined using Koya glasstic slides (KOVA International, USA). eMSC were cultured in DMEM/F12 medium containing 10% FCS, 1% primocin and 2mM glutamine (Invitrogen), supplemented with 10 ng/ml basic fibroblast growth factor (bFGF) (Peprotech, USA) and scaled-down to an in-house DMEM/F12 serum free medium (SFM) over 48 h within the first passage and incubated at 37°C in 5% CO2/5% O2/90% N2, as described previously 30 , with 1 µM TGFβ-R inhibitor, A83-01 (Tocris Bioscience, USA) or vehicle control (0.01% DMSO). Cells were seeded at 5,000 cells/cm 2 with medium changed every 48 h and passaged on days 15, 22, 29 and 36 into fibronectin-coated (10 µg/ml; BD, USA) culture flasks as described previously 23 . At each passage, cells were counted using Kova glasstic slides and cumulative cell population (total cell number) calculated by multiplying total number of cells yielded at the current passage by total number of cells yielded at the previous passage and then dividing by the number of cells seeded at the current passage, as described previously. 31 At passage 4 (i.e. 36 days in culture), untreated and A83-01-treated eMSC cultures were subjected to functional assays, and RNA-and ATAC-seq.
Colony-forming unit-fibroblast (CFU-F) assay eMSC were cultured with or without A83-01 for 2 passages (29 days) and then seeded at 50 cells/cm 2 to assess cloning efficiency as described previously, 23 with minor modifications.
Briefly, eMSC were seeded on fibronectin-coated 10 mm culture dishes and cultured in SFM, supplemented with bFGF and EGF (both 10 ng/mL; Invitrogen), in the presence or absence of A83-01 (1 μM) in SFM at 37°C in 5% CO25% O2/90% N2 for 2 weeks, with weekly medium changes. Cultures were then fixed in 10% formalin and stained with hematoxylin (Amber Scientific, USA), washed and blued in Scott's tap water. Colonies were counted and colony efficiency determined by dividing total number of colonies by number of cells seeded and multiplied by 100.

RNA extraction, RNA quality control and RNA libraries
Total RNA was extracted from untreated and A83-01-treated eMSC using RNeasy Mini kit (Qiagen, Germany) according to the manufacturer's instructions, with some variations. Briefly, eMSC were trypsinised using TryLE TM (Life Technologies), resuspended in DMEM/F12 containing 5% Albumax and then centrifuged at 300 g for 5 min. Cell were lysed with RNeasy Lysis Buffer, genomic DNA contamination removed with RNase-free DNase (Qiagen) and

ATAC-seq data and motif analyses
Sequenced paired-end reads were aligned to the University of California Santa Cruz (UCSC) human genome 19 (hg19) assembly using Bowtie2-2.2.6 32 and Samtools-1.2.0 33 and peak calling performed using MACS-2.1.0. HTSeq-0.6.1 34 to count the reads overlapping the peaks and differential expression analysis of sequencing data 2 (DESeq2) was used to determine opening and closing regions of the chromatin. 35 Fastq, metadata spreadsheet and table of counts have been deposited in the National Centre for Biotechnology Information Gene Expression Omnibus/sequence Read Archive with GEO accession number GSE146067. Differential open chromatin regions were mapped to cis-regulatory elements of their proximal genes using ENCODE DNaseI hypersensitivity data. 40 Physical interaction and distance no greater than 10 kb were used as criteria to assess association between ATAC-seq peak and proximal gene regulatory element. De novo short sequence motif analysis using Hypergeometric Optimization of Motif EnRichment (HOMER) v.4.8 was performed on 3,555 opening and 2,412 closing ATAC-seq peaks to determine enrichment and depletion of TF short sequence binding motifs in the differential ATAC-seq peaks. 41

RARb inhibition experiments
Passage 3 cultured eMSC from 4 biological samples were treated with 1 µM A83-01 or 0.01% DMSO (vehicle control) for 7 days, then trypsinized and seeded in triplicate at 1000 cells/well (3.125x10 3 cells/cm 2 ) into fibronectin-coated wells of two 96 well plates per biological replicate to assess proliferation on day 0 (D0) and day 3 (D3) in the presence or absence of 10 µM RARβ antagonist LE135 (Tocris Bioscience) using the cell viability MTS assay. After seeding, cells were allowed to adhere for 1 h before addition of 100 μl medium containing A83-01+LE135 or A83-01+DMSO. After 2 h, the MTS reagent (20µl, CellTiter 96® AQueous One Solution Cell Proliferation Assay, Promega, USA) was added to wells for the D0 timepoint and incubated for 2.25 h in the dark, then absorbance read at 490 nm on a spectrophotometer (Spectramax i3, Molecular Devices, USA). The medium was changed on the D3 plate at 24 h and the viability assay completed 72 h after seeding, as described above. Parallel cultures were subjected to CFU-F assays. Data were corrected for background readings (medium only) for each plate and normalized to D0 A83-01 for each sample and reported as fold-change.

Statistical Analysis
Statistical analyses were performed with GraphPad Prism 8. Technical replicates were inspected for outliers, which were removed from the analysis using Grubbs test. Normality of the data was determined with Shapiro-Wilk normality test. Individual data points and mean ± standard error of the mean (SEM) are shown when appropriate. Statistical significance was 12 determined using two-way analysis of variance (ANOVA) or ratio paired t-tests with p<0.05 considered statistically significant. For RNA-seq data analysis, statistical significance was assessed using the Benjamini-Hochberg procedure to control false discovery rate. Changes in gene expression were deemed statistically significant if the adjusted p-value (q-value) was less than 0.05.

Phenotypic characterization of sustained A83-01-treated eMSC
To investigate if eMSC can be expanded more efficiently when maintained under sustained TGFβ-R inhibition from culture initiation, cells isolated from 3 individual biopsies were seeded at 5000 cells/cm 2 in SFM supplemented with either A83-01 or vehicle (DMSO). The medium was refreshed every 48 h and cumulative cell population calculated at each passage (culture days 15, 22, 29 and 36). As shown in Figure 1A, treatment with A83-01 from culture initiation progressively conferred a proliferation advantage. After 36 days in culture, at least one order of magnitude more cells were produced in the A83-01 medium when compared to control medium (Fig. 1A). Next, we performed colony-forming unit-fibroblast (CFU-F) assays on eMSC first cultured with or without A83-01 for 2 passages (29 days) and then seeded at a low density (50 cells/cm2) to allow colony formation for a further 14 days. In keeping with our previous study, 23 exposure of cultured eMSC to A83-01 increased the CFU-F activity of primary cultures between ~ 4-10 fold (p = 0.0146; Fig. 1B). We then chose the 36-day timepoint (Fig. 1A) to analyze A83-01-treated cells for the expression of phenotypic eMSC markers CD140b, SUSD2 and CD90 by flow cytometry. The abundance of CD90 + cells was 99.1±0.5% (n=3) for the control and did not change upon TGFβ-R blockade (99.7±0.2% (n=3) (Fig. 1C). By contrast, A83-01 treatment consistently increased the abundance of CD140b + cells (p=0.031) and SUSD2 + cells (p=0.078) (Fig. 1C), although not statistically significant for SUSD2. Mean fluorescence intensity (MFI) was also calculated to evaluate the abundance of different cell surface molecules/cell. The MFI for CD90 did not show a consistent change upon A83-01 treatment, in keeping with previous observations that this ISCT minimal criteria MSC marker 42 shows little change under varied culture conditions. 23,30 By contrast, the MFI of 14 CD140b and SUSD2 (Fig. 1D) increased in response to A83-01 treatment in each culture, although statistical significance was not reached (p=0.0715 and 0.0897, respectively).

Transcriptional profiling of sustained A83-01-treated eMSC
To explore how cultured eMSC are maintained in a more naïve state upon sustained TGFβ-R inhibition, 3 independent cultures treated with or without A83-01 for 36 days were subjected to RNA-seq. Principal component analysis revealed that the greatest variation in gene expression is accounted for by intrinsic differences between primary cultures. The effect of A83-01 treatment was apparent in principal component 2 (PC2), which accounted for 28% of the variance in gene expression ( Fig. 2A). Following Benjamini-Hochberg correction for multiple testing, 1,463 genes were differentially expressed upon A83-01 treatment (Fig. 2B (Fig. S1). Many of the ECM genes repressed by A83-01 are very highly expressed in untreated cells with levels ranging from 342 to 14,586 TPM (Fig. S1). These data suggest that the effect of TGFβ-R blockade on eMSC is mediated, at least in part, by limiting ECM synthesis and deposition in prolonged culture. As shown in Figure S2, several angiogenic, anti-inflammatory, immunomodulatory, antifibrotic and anti-apoptotic genes were significantly upregulated in A83-01-treated cells, in keeping with our previous report. 43 Further annotation of differentially expressed genes using the KEGG Pathway database underscored the functional differences between A83-01-treated and untreated cultures (Fig. 2D). Notable pathways enriched in eMSC in response sustained TGFβ-R inhibition included the 'cyclic guanosine monophosphate (cGMP)-protein kinase G (PKG) signaling pathway', which is implicated in nitric oxidemediated cardioprotection during acute ischemic preconditioning, 44 , cell growth, and inhibition of apoptosis, 45 and 'cytokine-cytokine receptor interaction', in keeping with the innate paracrine function of MSC. 17,46 Conversely, A83-01 repressed genes were enriched in 'pathways in cancer', 'focal adhesion', and 'ECM-receptor interaction'. Notably, the 'PI3K-Akt signaling pathway' was common to both up-and down-regulated genes. This multifaceted pathway controls key cellular processes by phosphorylating substrates involved in apoptosis, protein synthesis, metabolism, and cell cycle.

Chromatin changes induced by sustained A83-01 treatment of eMSC
Dynamic changes in chromatin structure and epigenetic code drive gene expression and ultimately define cell identity. 29 To map the global changes in the genomic architecture of cultured eMSC in response to sustained TGFβ-R inhibition, 3 independent eMSC cultures treated with or without A83-01 for 36 days were subjected to ATAC-seq, which profiles chromatin accessible regions as a sequencing depth readout. 39 Based on q ≤ 0.05, DESeq identified 5,967 differential ATAC-seq peaks upon A83-01 treatment, 60% of which involved opening of genomic regions and 40% closing of specific loci. Out of 5,967 peaks, 31% and 29% of the opening and closing ATAC-seq peaks, respectively, fell within -10 to +1 kilobases (kb) around transcriptional start sites (TSSs). RARB (coding retinoic acid receptor beta, RARb), TGFBR3 (transforming growth factor beta receptor 3) and SUSD2 exemplify genes that showed increased chromatin accessibility at and upstream of their proximal promoters upon A83-01 treatment (Fig. 3A). Cross-referencing with RNA-seq data showed a significant increase in RARB, TGFBR3 and SUSD2 transcript levels in response to A83-01 treatment (q=1.1´10 -33 , q=1.2´10 -22 , and q=2.7´10 -5 , respectively). Conversely, CADM1 (cell adhesion molecule 1), COL1A1 (collagen type I alpha 1 chain), and WNT5A are examples of genes repressed in response to A83-01 treatment (q=1.9´10 -46 , q=1.4 ´10 -9 , and q=5.0´10 -30 , respectively). As shown in Figure 3B, downregulation of CADM1 and COL1A1 is associated with closure of their proximal promoters whereas silencing of WNT5A coincides with closure of a distal enhancer.
The gain or loss of ATAC-seq peaks upon A83-01 treatment indicates that altered transcription factor (TF) binding drives differential gene expression. However, both activating and repressive TFs can potentially bind at different regulatory sites, rendering it challenging to confidently predict gene expression from dynamic chromatin changes at specific loci alone. 39 Nevertheless, analysis of 200 genes associated with the most induced or repressed ATAC-seq peaks (within 10 kb of TSSs) revealed a strong association with increased or decreased expression, respectively, upon A83-01 treatment (p=1.0´10 -6 ) (Fig. 4). Next, we interrogated the ATAC-seq data to gain insight into the cis-regulatory landscape that underpins the transcriptional responses of eMSC to A83-01 treatment. De novo binding motif enrichment analysis was performed using HOMER on 3,555 opening and 2,412 closing ATAC-seq peaks.
This annotation yielded 19 significantly overrepresented motifs in opening peaks (Fig. S3), and 17 overrepresented motifs in closing peaks (Fig. S4). Next, we matched the motifs against canonical TFs that are differentially expressed in cultured eMSC treated with and without A83-01 (Fig. 5A). CCAAT/enhancer binding protein beta and delta (CEBPB/CEBPD), RARB, RAR-related orphan receptor alpha (RORA), and nuclear receptor subfamily 4 group A member 1 (NR4A1, also known as NUR77) were amongst the most plausible differentially expressed genes of TFs that can bind the enriched motifs in opening ATAC-seq peaks with high affinity (Fig. 5A, B). Conversely, reduced expression of transcription factor 21 (TCF21), TGFB induced factor homeobox 2 (TGIF2), and nuclear transcription factor Y subunit alpha (NFYA) paralleled the loss of their corresponding binding sites in closing loci (Fig. 5A, B).
However, addition of LE135 to A83-01 treated cultures indicated that RARβ inhibition partially reverses the proliferation advantage conferred by TGFβ-R inhibition (Fig. 6D). In addition, LE135 reduced the clonogenicity of A83-01 treated eMSC (Fig. 6E). We reported previously that short-term treatment with A83-01 in SFM (7 days) following extensive culture of eMSC in serum medium (6 passages) increases proliferation and enhances clonogenicity. 23 Here we report that sustained TGFβ-R inhibition from culture initiation similarly promotes eMSC clonogenicity and proliferation with significant differences observed by passage 3. Prolonged culture under these conditions also increased the percentage of SUSD2-and CD140b-expressing cells and the density of these surface markers on the A83-01-treated cells, but not the representative ISCT marker, CD90, indicating that the eMSC phenotype was retained and spontaneous differentiation attenuated. Our data for these parameters also highlighted intrinsic differences in A83-01 responsiveness between eMSC cultures established from different donors, a well-known feature of primary MSC cultures. 2 Recently it was shown that body mass index of patients, but not age, correlate inversely with the abundance of clonal SUSD2 + eMSC. 49 These observations raise the possibility that clinical variables, such as obesity, impact on the in vitro expandability of eMSC.

This study demonstrates that eMSC cultured under continuous
The divergent expression of a large number of genes in A83-01-treated cultures suggests either profound transcriptional reprogramming or -perhaps more likely -attenuation of differentiation cues imposed by in vitro culture conditions. For example, a striking but not entirely unexpected observation is that A83-01 represses multiple highly expressed genes involved in ECM deposition and collagen metabolism, including 8 collagen subunit genes, Interestingly, NR4A1 inhibits TGFβ signaling in the nucleus by promoting the assembly of a repressor complex that binds to the promoters of TGFβ target genes. 68 Because of its role as an endogenous TGFβ inhibitor, small-molecule NR4A1 agonists are under development for the treatment of fibrotic disorders. 68 In summary, by integrating genome-wide expression and DNA accessibility profiling techniques, this study has advanced our understanding of the cis-regulatory DNA landscape and gene networks that safeguards eMSC against spontaneous differentiation and loss of function in prolonged cultures. Analyses of these two large data sets revealed novel pharmacological targets that could be exploited to accelerate clinical translation of autologous eMSC therapies for a variety of reproductive disorders. Furthermore, the data sets constitute a robust resource to interrogate fundamental molecular questions pertaining to human endometrial biology.       Y-axis shows TPMs; *indicates q < 0.05, ** q < 0.01 and *** q < 0.001.