Organoid co-culture model of the human endometrium in a fully synthetic extracellular matrix enables the study of epithelial-stromal crosstalk.

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Linda Griffith ( [email protected] ). Hydrogel reagents and peptides described in the manuscript are commercially available. The reagents to make the synthetic hydrogel are in the process of being commercially licensed with the purpose of providing accessibility of these gels to the community. For immediate availability, a Material Transfer Agreement may be required to provide reagents due to transfer restrictions. Single cell RNA and Bulk RNA sequencing data have been deposited into the database of Genotypes and Phenotypes (dbGAP) and are available under dbGaP accession phs003326.v1.p1. Proteomics data was deposited into PRIDE, and processed data into the Gene Expression Omnibus (GEO). All data and links are being compiled into a central study site, which can be accessed here: “Integrating endometrial proteomic and single cell transcriptomic pipelines reveals distinct menstrual cycle and endometriosis-associated molecular profiles”: https://fairdomhub.org/studies/1139 . Accession numbers are listed in the key resources table . Imaging data have been deposited at Omero, an Open microscopy image repository client-specific repository, and bioassay data are publicly available as of the date of publication and can be found in https://fairdomhub.org/studies/1139 . This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. All participants provided informed consent in accordance with a protocol approved by the Partners Human Research Committee and the Massachusetts Institute of Technology Committee on the Use of Humans as Experimental Subjects (Protocol number IRB-P001994). Endometrial tissue was obtained from pipelle biopsies from reproductive age women (N=18, ages 18–45; Table 1) undergoing laparoscopic surgery for non-malignant gynecologic indications. Study enrollment was limited to pre-menopausal women and excluded patients with an irregular or ambiguous cycle, or a history of hormone use in the 3 months prior to surgery. Cycle phase dating was performed as described elsewhere 4 by histological analysis ( Fig S20 in Data File 1 ). Endometrial epithelial glands and stromal cells were isolated by enzymatic digestion and filter-separated as previously described 125 , resulting in ≥ 95% purity as assessed by positive staining for vimentin and morphological assessment. Briefly, single stromal cells were separated from intact epithelial glands fragments and cultured separately. Stromal cells were expanded in traditional 2D cultures (as described below) and epithelial glands were cultured as organoids using protocols adapted from Turco et al. 11 All cells were passaged a maximum of 5 times for all experimental in vitro assessments. For this study, we only included all female (individuals with a uterus). Limited or no metadata information regarding clinical and ethnic or socioeconomic background was available for some patients. Primary cells were utilized from the donors described and used for in vitro culture and analysis. 15 different endometrial donors were used for isolation and cell culture of those 12 were used to establish co-culture models. Stromal cells were cultured and maintained in phenol red-free DMEM/F12 with 5% charcoal-stripped calf serum, 1 nM 17-β estradiol (E 2 , Sigma Aldrich) and 1× Pen-Strep solution (Sigma Aldrich). Some cell cultures were treated with 500 nM synthetic progesterone medroxyprogesterone acetate (MPA, Sigma Aldrich), or the progesterone antagonist mifepristone (RU-486 10 µM). Primary endometrial epithelial organoids (EEOs) were generated from the primary epithelial fragments (p0) according to established protocols and maintained in endometrial epithelial organoid expansion medium (EEO medium). 10 , 11 EEO media is modified from the protocols established by Turco et al. 11 Briefly, EEO media was defined as the minimal set of components needed to establish organoid cultures expansion and was composed of advanced basal media cocktail 11 (ABM, 1X); rhEGF (50 ng/mL, Corning); rhNoggin (100 ng/mL, Peprotech); rhRspondin-1 (200 ng/mL, R&D systems); rhFGF-10 (50 ng/mL, Peprotech); E 2 (1nM); Nicotinamide (1mM); Insulin-Transferrin-Selenium (ITS, 1%, Invitrogen); N-Acetyl-L-Cysteine (1.25 mM); TGFB pathway inhibitor (A83-01, 500 nM, Peprotech); and Rock inhibitor Y-27632, (ROCKi, 10 µM, Tocris). For seeding, epithelial cells, as fragments or single cell suspensions, were embedded in 70% Matrigel and cultured for 6 days, with media was changed every two days ( Fig S2 in Data File 1 ). For passaging, 6-day old organoids were incubated in Cell Recovery Solution (CRS, ThermoFisher) for 30 mins at 4°C to dissolve Matrigel, then pelleted and digested with Tryp-LE supplemented with 1:100 DNAse (40k units, Sigma) for 15 minutes in a 37°C water bath followed by mechanical dissolution to generate single cells. For expansion purposes, single epithelial cells were seeded in Matrigel at a density of 1,000 cells/µL in 60µL droplets in non-tissue culture treated plates. We refer to organoids generated from single cells, in regular expansion or in experiments, as scEEOs. For co-culture experiments of ESC and scEEOs, ROCKi was omitted from EEO media since intact organoids were utilized to establish the co-cultures. ESCs retained hormone sensitivity and morphological characteristic when cultured in this EEO-based common media (See Results). Due to the presence of progesterone in the N2 and B27 supplements in the advanced basal media a homemade formulation of EEO media termed neutral EEO (nEEO) media that omitted P4 in the supplements was also explored in some experiments. All media changes for the co-cultures occurred every 3 days for up to 15 days unless noted otherwise. Spent conditioned media from these cultures was collected and stored at −80°C for downstream analysis. Sex of cells were all female because this population has a uterus and endometrial tissue. 8-arm poly(ethylene glycol) (PEG) macromers (20 kDa) functionalized with vinyl sulfone (20 kDa PEG-VS) were purchased from JenKem Technology (Beijing, China). All peptides were custom synthesized and purified (>95%) by Boston Open Labs (Cambridge, MA), GenScript (Piscataway, NJ), or CPC Scientific (Sunnyvale, CA). Peptides used in these studies include: a dithiol crosslinking peptide containing a matrix metalloproteinase (MMP)-sensitive substrate (Ac)GCRD-LPRTG-GPQGIWGQ-DRCG(Am) ( CL-LW ); fibronectin (FN)-derived peptide containing both the canonical RGD motif from the 10 th FN type III domain as well as the PHSRN synergy site from the 9 th FN Type III repeat in a branched configuration akin to the biophysical presentation in FN, (Ac)PHSRNGGGK-GGGERCG(Ac) -GGRGDSPY(Am) ( PHSRN-K-RGD ); 56 , 126 , 127 a Collagen-derived peptide, (Ac)GGYGGGPG(GPP) 5 GFOGER(GPP) 5 GPC(Am) ( GFOGER ); 22 , 40 a peptide with affinity for sequestering cell-produced FN, (Ac)KKGCRE-TLQPVYEYMVGV(Am) ( FN-binder ); 128 and a peptide with affinity for sequestering the basement membrane proteins type IV collagen and laminin, (Ac)GCRE-ISAFLGIPFAEPPMGPRRFLPPEPKKP(Am) ( BM-binder ). 129 All peptides were reconstituted in acidic (pH 5.5) Milli-Q water (Millipore). The concentration of free thiols in all peptides was determined using Ellman’s reagent (Sigma Aldrich). Synthetic matrices were assembled using materials at the concentrations indicated throughout the text. Briefly, 8-arm 20 kDa PEG-VS macromers were prepared in a 10 w/w% solution in ultra pure water. Macromers were used at concentrations ranging from 14.4–33.6 mM VS (nominally, 3–7 wt%). Macromers were functionalized (fPEG-VS) via a Michael-type addition reaction with the integrin ( PHSRN-K-RGS, GFOGER ) and matrix-binding ( FN-binder, BM-binder ) by adding 10% of the total solution volume in 10x PBS + 1M HEPES, pH 8.2, vortex-mixing, and reacting at room temperature for 30 minutes. Note that lyophilized peptides containing HCl salt can be very acidic in stock solutions, so the pH of the final reaction mix should be maintained at 7.8 by adjusting the pH of the buffer solution. During functionalization, cells were counted and pelleted at the appropriate number for the target final cell densities (10,000 ESC and 10 EEO per µL of gel solution). Following functionalization, care was taken to ensure that all cell media was removed without disturbing the cell pellet, and pellets were gently resuspended in fPEG-VS. Upon resuspension, cross linker (CL-LW) was added to the mixture at a stoichiometric ratio of 0.45 thiol:VS to initiate gelation. The solution was gently mixed and plated accordingly, typically in 3 µL droplets in non-tissue culture treated 96-well plates unless otherwise specified in the text. Cell-fPEG-VS mixture was incubated for 25 minutes in a humidified incubator at 37°C until gelation was complete. To measure bulk bio-mechanical properties of crosslinked gels, 25 µL of each gel matrix mixture was loaded into a 1 mL syringe that had the tip cut off at the 0.1 mL mark. The matrix was allowed to gel at 37 °C for 25 min and then extruded from the open syringe as a cylindrical disc and moved to a 24 well plate that contained 400 µL of 1X PBS. The plate was incubated for 24 hours in a humidified incubator at 37 °C, 95% air, and 5% CO 2 , to allow equilibrium swelling to occur prior to rheological characterization. This procedure generates hydrogel discs of 1.4 mm in thickness. The discs were sandwiched between an 8 mm sandblasted parallel plate geometry and sandblasted base. The shear modulus was determined by performing small-strain oscillatory shear measurements on an Anton Parr MCR 302 rheometer. The mechanical response was recorded by performing frequency sweep measurements (0.1–10 Hz) at a constant strain (0.05), at 37 °C. The elastic modulus (E) is reported as a measure of bulk matrix mechanical properties derived from the recorded storage modulus (G’) in the linear elastic regime where: 𝐸 = 2𝐺 ′ (1 + 𝜈) with the assumption that 𝜈 = 0.5 for ideal elastic hydrogel materials. Endometrial organoids were established and expanded in Matrigel according to established protocols 11 with some minor modifications. Lumenized intact organoids with a diameter greater than 100 µm were harvested with cold CRS. Matrigel dollops were manually detached with a wide bore pipette tip, making sure to fully scrape the entire well to lift off the entire gel. The dollops were resuspended in cold CRS (5mL) for at least 30 mins on ice to dissolve remaining Matrigel, followed by centrifugation at 500–1000g, which resulted in a clear pellet of organoids. Organoids were broken into single cell suspensions using 1mL of prewarmed TrypLE Express (ThermoFisher) supplemented with DNAase (10µl, 1:100) and continuous mechanical agitation until a single cell suspension of epithelial cells was observed. Single cells were counted using a hemocytometer then resuspended in the functionalized PEG precursor (fPEG-VS) solutions prior to the addition of the crosslinker. In parallel, single cells were resuspended in Matrigel that served as an experimental control during matrix evaluation. In both cases, cells were encapsulated at a density of 500 cells/µL of matrix. Cell-matrix suspensions were seeded as 3 µL droplets on a non-tissue treated 96 well plate and allowed to polymerize for 25 min in a humidified incubator at 37 °C, 95% air, and 5% CO 2 . Complete gelation was examined visually on a test gel droplet prior to the addition of media. Gel droplets were maintained in 100 µL of expansion EEO media supplemented with ROCKi (10 µM). Media was changed every three days. Six-day old intact organoids (~100 µm diameter) were grown and expanded in Matrigel and harvested as described above. To maintain scEEO integrity, organoids were incubated in CRS until Matrigel was fully dissolved. Careful resuspension of intact scEEOs in DMEM/F12 was performed to prevent disrupting scEEO integrity. In parallel, donor matched ESCs were expanded using traditional 2D polystyrene cell culture plates and grown to 80% confluency. ESCs were lifted from 2D culture plates using trypsin (1X) and processed to generate a single cell suspension. Intact scEEOs and ESCs were prepared in parallel to encapsulate both cell types simultaneously and establish a co-culture in the PEG matrix solution. Cells were encapsulated at a density of 10,000 ESC and 10 intact scEEOs per µL of matrix. This epithelial-stromal ratio (approximately 1:1) was determined experimentally based on the cell ratios used to generate established endometrial xenograft models (sub-renal capsule) described elsewhere 130 , 131 and validated using whole-mount 3D image analysis of endometrial tissue we performed (see Results). Embedded cells in fPEG-vs were seeded in 3 µL droplets and allowed to polymerize as described above in non-tissue culture treated 96-well plates. Co-culture was maintained in EEO media without Y-27632, a common media recipe utilized in other co-culture studies 61 , unless otherwise noted. Media was changed every three days and maintained for up to 15 days. It was noted that some basal components of the organoid medium contained P4 (N2 and B27) which may impact hormone responsive behavior, so a P4-reduced ‘neutral’ formulation of this organoid medium (nEEO medium) that completely lacked P4 was tested in some co-cultures. Both media formulations were sufficient to maintain support cell culture and induced hormone-mediated responses ( Fig S19 in Data File 1 ); however, as expected, co-cultures maintained in nEEO media conferred a more robust response to progestin. Matching monocultures of ESCs-only or scEEOs-only were established in parallel for some experiments. Cultures were imaged daily. To quantify organoid diameter and total organoid formation efficiency, single epithelial cells were encapsulated in fPEG-VS hydrogels or Matrigel and cultured over 14 days. scEEO formation efficiency was calculated as the percentage of organoids with a clear lumen with a diameter (>100 um) relative to Matrigel. 4X brightfield (BF) z-stack images of 10-day old endometrial organoids were captured using an EVOS M500 microscope (Invitrogen). The images were processed in Fiji using the time lapse Gaussian-based stacker focuser plugin to generate maximum intensity projections. Images were manually analyzed in Fiji to obtain a diameter and organoid count distribution for each condition. Daily images of co-cultures were manually curated to obtain relative growth rates of the organoids in single culture (N=3) or co-culture (N=8) by measuring scEEO diameter and quantifying the daily fold change relative to day 1 of culture ( Videos S4 – S16 in Data File 4 ). Co-culture cell viability and cell death was measured using the LIVE/DEAD ™ Cell Imaging Kit (488/570) (Thermofisher Scientific) according to the manufacturers protocol. Briefly, at day-15 prior to fixation, cells were incubated the reagents provided by the kit and incubated at 37 °C, 95% air, and 5% CO 2 for 20 minutes. Images were captured using Keyence BZ-X800 Fluorescence Microscope. Cell proliferation was assessed by EdU incorporation using the Click‐iT Plus EdU imaging kit with Alexa Fluor 594. All co-culture and monoculture experiments were treated with EdU (5-ethynyl-2′-deoxyuridine, 20 μM) for a 24-hour incubation period at 37°C, 95% air, 5% CO 2 prior to fixation at day 15 of culture. Click-iT reaction cocktail was prepared as described by the manufacturer (ThermoFisher Scientific). Cells were incubated for 1 hr at RT protected from light and counter stained with DAPI (1 mg/mL, 1:2000). Images were captured using a ZEISS confocal Laser Scanning Microscope (LSM 880) and analyzed by 3D image processing was performed using the Surface and Spots function in Imaris (Bitmap) software (9.6.0). At least three individual organoids were analyzed per donor, and the proliferative index was measured as the number of EdU + positive cells per organoid surface area (mm 2 ). ESC proliferative index was calculated as the EdU + cells relative to the total number of DAPI stained nuclei. ESCs and intact scEEOs cultured in Matrigel or synthetic hydrogels were used for molecular analysis. After 14 days of culture, intact scEEOs were released from Matrigel and the synthetic ECM, using Cell Recovery Solution, or recombinant Sortase, respectively. Intact scEEO-only cultures, ESC-only cultures, or co-cultures were pelleted, resuspended in TRIzol reagent (ThermoFisher Scientific), and then stored at −80°C until processing. RNA was extracted using the Directzol RNA Mini-Prep kit (Zymo Research) per the manufacturer’s protocols with the inclusion of an on-column DNase step using the PureLink DNase Set (ThermoFisher Scientific). At least eight replicates (3 µl droplets) were pooled together to obtain sufficient RNA. cDNA was synthesized from ~1 µg of total RNA using the High-Capacity RNA-to-cDNA Kit (ThermoFisher Scientific, 4387406) per manufacturer’s protocols. TaqMan Fast Advanced Master Mix (ThermoFisher Scientific) was used in congruence with the cell-specific probes for qPCR. Gene expression was determined using the StepOnePlus real-time PCR system (Applied Biosystems) and calculated using the ΔΔCt method using GraphPad Prism. Gene expression was first normalized using the housekeeping GAPDH gene in each sample, then the relative fold change relative to experimental control. High Throughput 3’ Digital Gene Expression (3’DGE) RNAseq was performed on the isolated RNA using NovaSeq flowcells at the BioMicroCenter (MIT) and processed using the BMC/BCC 1.8 pipeline (cocultures n=7; ESC monocultures, n=6). Spent media was stored at −80°C until analyzed by ELISA for prolactin (R&D Systems). Protocols provided by the manufacturer were adapted to allow ELISAs to be performed in a 384-well plate (ThermoFisher) to minimize medium (sample) volume needed. 26 Multiplexed Luminex assays were performed to measure cytokines in 48- or 72-hour undiluted conditioned medium throughout the co-culture period using the magnetic Human Luminex XL Cytokine 45-PLEX pre-mixed kit (R&D systems). Protocols provided by the manufacturer were adapted to allow the assay to be performed in a 384-well plate to avoid introducing batch variability. Ten-point standard curves with culture medium and assay buffer blanks were included for quantification. Conversion from mean fluorescence intensity (MFI) values to absolute concentration was performed by interpolating 5-parameter logistical fits of standard curves for each analyte. Following the same downstream analysis protocol, secreted matrix metalloproteinases (MMP) concentrations were measured using the magnetic Human MMP Magnetic Luminex Performance Assay (R&D Systems). This assay required dilutions of cell culture supernatants in fresh culture medium at either 2x or 200x to detect the reported analytes within a quantifiable range. Hydrogel samples were fixed in 4% paraformaldehyde (Electron Microscopy Sciences) for 30 mins, permeabilized in 0.2% TritonX-100 in PBS (Sigma) for at least 30 mins at room temperature, and then blocked overnight with blocking buffer (1% BSA (Sigma), 5% normal donkey serum (Electron Microscopy Sciences) in PBS) at 4°C. Hydrogels were incubated with primary antibodies overnight while rocking at 4°C, rinsed with blocking buffer 3 times (5 min each while rocking at RT), and treated with secondary-conjugated antibodies overnight while rocking at 4°C. Specific antibodies and relevant controls can be found in the Key Resources Table and supplementary table of reagents ( Table S2 in Data File 1 ). Images were captured using a ZEISS confocal Laser Scanning Microscope (LSM 880) or a Keyence BZ-X800 Fluorescence Microscope. Image analysis was performed using Fiji open software and 3D image processing was performed using Imaris (Bitmap) software (9.6.0). Endometrial tissue processing for proteomics, proteomics data collection and analysis is described elsewhere (in preparation) 44 . In this study, the proteomic processing and pipelines were applied to 3 additional donors (N=11). Briefly, fresh endometrial biopsies were collected and processed using 50 µl of RapiGest (Waters Corporation) reconstituted in 50 mM NH 4 HCO 3 at a concentration of 2mg/ml to digest the endometrial biopsy sample for 24 hours at 37°C. Next, urea (ThermoFisher) and dithiothreitol (ThermoFisher) were added to the sample to bring the mixture of a final concentration of 4 mM and 5 mM, respectively. Finally, iodoacetamide (MilliporeSigma) was added to the sample, for a final concentration of 15 mM and kept in the dark, at 37°C, for 30 minutes. Samples were frozen at −80°C until all patient samples had been collected and processed. Samples were lyophilized overnight and then reconstituted in NH 4 HCO 3 . Samples were digested with Trypsin Gold, mass spectrometry grade (Promega), using a ratio of 1 µg trypsin to 50 µg sample protein as quantified using a Pierce BCA Protein Assay Kit (ThermoFisher). Samples were digested overnight at room temperature using a rotor for mixing and then placed in a speed vacuum to reduce sample volume. Dried down digested peptide samples were labeled with TMT10plex kit (Pierce). After labeling, the samples were mixed and purified with peptide desalting spin columns (Pierce). 1/5 of the sample was used for one LC/MS/MS analysis (Thermo Exploris480 mass spectrometer). The dried peptide mix was reconstituted in a solution of 2% formic acid (FA) for MS analysis. Peptides were loaded with the autosampler directly onto a 50cm EASY-Spray C18 column (ThermoFisher). All MS/MS samples were analyzed using Sequest (ThermoFisher; version IseNode in Proteome Discoverer 2.3.0.523). Sequest was set up to search uniprot_human_reviewed_032120.fasta (version March 21, 2020) assuming the digestion enzyme trypsin. Scaffold Q+ (version Scaffold_5.0.1, Proteome Software Inc.) was used to quantitate Label Based Quantitation peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 10.0% probability to achieve an FDR less than 1.0% by the Percolator posterior error probability calculation. 132 Normalization was performed iteratively (across samples and spectra) on intensities, as described in Statistical Analysis of Relative Labeled Mass Spectrometry Data from Complex Samples Using ANOVA. 133 Medians were used for averaging. The protein list was then compared to the curated human matrisome ( http://matrisome.org/ ) list to look for ECM proteins. Samples were normalized using a mixed aliquot of all patient samples that was analyzed in each LC/MS-MS run. Overlapping ECM proteins found in all 3 LC/MS-MS runs were identified and the Limma package 134 in R was used to remove batch to batch effects. scRNAseq data collection, processing and analysis as described elsewhere (in preparation) 44 and is available (see Data and Code Accessibility). Data was downloaded and Seurat (V3) was used to create dot plots and violin plots of specific genes of interest on scaled data. Aligned RNAseq transcript expected counts were analyzed in R (version 4.0.0) where data was modeled using the edgeR package (version 3.34.0). 135 , 136 Counts were normalized by library size and filtered to exclude genes with less than 20 transcripts across all samples. Principal component analysis was performed using data batch-corrected by donor via the ComBat-seq method of the sva package (version 3.13) in R, and each gene was scaled to have unit variance before modeling. 137 Differential gene expression was calculated by modeling transcript data in a quasi-likelihood negative binomial generalized log-linear model using the edgeR function “glmQLFit”. The use of a generalized log-linear model allowed for differences between hormone treatment groups to be compared while accounting for inter-donor variability. Genes were ranked by fold change in expression between hormone treatment groups to produce a ranked list of genes for gene set enrichment analysis (GSEA) ( http://www.broadinstitute.org/gsea/index.jsp ). Gene ontology (GO) term analysis was performed on ranked gene lists of genes with pvalues <= 0.05. No adjustment was performed at this step due to the low sample number, high variability, and intent to use the RNA sequencing as a hypothesis generation step. The R package topGO (version 2.44.0) was used for GO term identification and the package clusterProfiler (version 4.0.2) was used for visualizations. A curated list of menstrual related genes was assembled from existing literature. 62 , 138 A heatmap of log 2 transformed data was created using the R package pheatmap ( https://CRAN.R-project.org/package=pheatmap ; version 1.0.12) with the “complete” clustering method. Analysis was performed on 7 co-culture experiments comprised of 5 donors, 2 independent technical replicates and results were compared between experiments that were maintained in nEEO media (N=2) or traditional EEO media (N=5). Data are expressed as average ± standard error of the mean (SEM). Statistical analysis was performed using GraphPad Prism 8. Unpaired t-tests assuming unequal variances and two-way ANOVAs with appropriate post-tests were performed. Holm-Sidak's multiple comparison of the mean or two-tailed Mann-Whitney tests when comparing two groups. qPCR data was analyzed using multiple t-tests. Specific sample sizes for each assay is specified in the figure legends or methods. Statistical significance was defined as * p < 0.05, ** p < 0.01, *** p < 0.0001, and ****P < 0.0001.

The ECM composition of the endometrium, along with cell-matrix receptor expression, are regulated by sex steroids throughout the menstrual cycle, 33 , 34 giving rise to dramatic shifts in tissue structure and morphology ( Fig 1A ). The overarching design goal for our synthetic ECM is to define a minimal set of biophysical and biomolecular cues required to support the dynamic phenotypic functions of heterogenous endometrial cell populations, over weeks of culture. As a first step, we performed a targeted proteomic analysis (LC-MS/MS, N=11) of the endometrial ECM-associated proteins (matrisome) across the menstrual cycle to benchmark the ECM composition in the native tissue ( Fig 1B ). Consistent with immunohistochemical reports 34 – 36 , FN abundance was greater in the proliferative phase, while fibrillin, collagen and laminin proteins increased during the secretory phase, corresponding to the differentiation of stromal fibroblasts, epithelial gland maturation, and vascular remodeling in response to progesterone ( Fig 1B ). Our initial studies demonstrating the establishment of endometrial organoids in a synthetic ECM 27 were guided in part by existing histologic analysis of endometrial integrin expression patterns from a subset of the 24 known integrins. 36 – 43 We sought additional insights into cell population-specific integrin expression by probing a single cell RNA-sequencing dataset 44 (scRNAseq, N=6; Fig 1C ; Fig S1 in Data File 1 ) to profile the integrin expression across the endometrial proliferative or secretory phase epithelia and stroma in addition to the immune and vascular cellular compartments. Across all cells in the data set, transcripts for all eighteen α chains and all eight β chains, except integrin β3, were detected (data not shown). Integrin β3, a contested marker of fertility that appears in the secretory phase 45 – 47 , was also absent in another endometrial scRNAseq data set. 48 Consistent with the immunohistochemistry literature 38 , αV was robustly expressed by epithelia and stroma mostly during the proliferative phase. Its primary binding partner β5 was primarily expressed by the stroma across the cycle ( Fig S1A in Data File 1 ), while β6 was sparse except in proliferative phase epithelia ( Fig S1A in Data File 1 ). Unexpectedly, integrin β8, an additional αV partner that has not previously been characterized via immunohistochemistry in the endometrium, was robustly expressed in several epithelial and some stromal subpopulations ( Fig 1C ; Fig S1A in Data File 1 ). Also consistent with the immunohistochemistry data 36 , 39 , 41 – 43 integrin β1 was robustly expressed by both epithelia and stroma across the cycle. Expression of the collagen-binding α1 dimer (at lower levels, especially in proliferative epithelia) along with strong expression of α2 in the epithelia was also observed, while the stroma primarily expressed α5 ( Fig 1C , Fig S1B in Data File 1 ). Finally, laminin-binding α6 and its heterodimer β1 was robustly expressed in all epithelia, while expression in the stroma was higher in the proliferative phase ( Fig S1 in Data File 1 ). We investigated the expression of a subset of these integrins in cultured endometrial cells via qPCR confirming comparable expression of β1, α1, α2 and αv in both stromal and epithelial cells, but significantly greater expression of α5 and α8 in stromal cells ( Fig S1C in Data File 1 ). The minimal synthetic ECM formulation must provide integrin-engaging adhesions cues that support cell viability immediately post-encapsulation upon dissociation from tissue or a prior culture state 49 , 50 , while enabling remodeling and stabilization of cell-produced ECM that provides a fuller spectrum of adhesion cues as the culture progresses. We thus include the collagen-derived peptide GFOGER , recognized by the α1β1 (expressed on stroma and epithelia) and α2β1 integrin heterodimers (expressed on epithelia), 51 – 53 , 54 together with the fibronectin-derived PHSRN-K-RGD peptide 4 , 26 as an RGD ligand for αVβ1 and α5β1 ( Fig 1D ). This ligand incorporates the PHSRN synergy site recognized by α5β1 heterodimer primarily expressed by the stroma and, to a lesser extent, by the secretory epithelium ( Fig 1C ). As cells remodel the local microenvironment, proteolytically degrading the synthetic ECM, they deposit cell-specific ECM, creating an additional dominant source of adhesion cues for a broad spectrum of receptors as the culture progresses. To sequester the cell-produced ECM and enhance its interactions with the synthetic ECM, we incorporated peptides that bind to basement membrane components produced by epithelia (and some decidual fibroblasts) and also to fibronectin produced by stromal cells, as previously described. 26 , 27 , 32 This strategy obviates the need to include laminin-derived peptides. 26 , 27 , 32 To synthesize the synthetic ECM, we first functionalized an 8-arm PEG macromers with a combination of GFOGER and PHSRN-K-RGD (“ MIX” ) in defined ratios (see below), then crosslinked this macromer solution, in the presence of cells, with a protease-degradable crosslinking peptide (“ CL-LW” ) to enable cell-dependent matrix remodeling, migration and proliferation during hormone-driven morphogenesis. 26 , 55 , 56 Using this basic framework, individual synthetic ECM parameters (e.g., matrix stiffness) were systematically and independently varied to mimic aspects of healthy and disease states ( Fig 1D ). Altogether, a tissue-inspired approach was implemented to design a synthetic matrix that would initiate and maintain the culture of EEOs and ESC in a controlled environment that mimics the molecular and biophysical properties of the endometrial ECM. We previously demonstrated that a particular formulation of synthetic ECM designed for enteric organoids also supported endometrial epithelial organoids 27 ; however, the impact of the biophysical properties and peptide composition on organoid phenotype and function was largely unexplored. These first-generation hydrogels 27 incorporated only one adhesion ligand (i.e., GFOGER ) and were substantially stiffer (~2000 Pa, corresponding to 5wt% PEG gels) than Matrigel (~150–443 Pa) 21 , 57 or the native endometrial tissue (~250 Pa) 58 . Thus, to better understand EEO generation in synthetic matrices, we probed the biophysical parameters that more closely mimicked native endometrial stiffness and additionally interrogated the consequences of integrin ligand composition. We first generated a tissue bank of EEOs from 8 endometrial donors with and without endometriotic disorders (Table 1 and Fig S2A in Data File 1 ) and defined a panel of gels with varying elastic moduli ranging from “soft” gels (3wt% PEG, ~300 Pa), comparable to native stiffness of the endometrium, to the “stiffest” gels (7wt% PEG, ~6,000 Pa; Fig 2A ). To test these formulations, we dissociated organoids to single cells, passaged them into synthetic matrix or Matrigel, and followed the emergence into single cell-derived endometrial epithelial organoids (“scEEO”, Fig S2B in Data File 1 ). In agreement with our previous work 27 , employing synthetic hydrogels functionalized with GFOGER was sufficient to support the growth of scEEOs, even in a relatively stiff (5wt% PEG) environment. However, scEEOs generated in stiffer gels developed a crenelated morphology ( Fig 2B ; Fig S4 in Data File 1 ) compared to those cultured in softer conditions (3wt% PEG), and this was consistent across all donors regardless of the donor’s disease state (data not shown). Anticipating the integration of α5β1-expressing stromal cells, we then evaluated the influence of an additional integrin ligand on scEEO growth behavior and morphologies. We compared ECM formulations incorporating nominal concentrations of GFOGER (3mM), PHSRN-K-RGD (3mM), or a MIX of both peptide (1.5 mM each) in the soft (~300 Pa, 3wt% PEG) gel regime. Biomechanical properties were not significantly affected by differences in integrin ligand composition (data not shown). As expected, GFOGER was necessary to generate scEEOs in synthetic matrices ( Fig 2C ; Fig S3 in Data File 1 ), presumably through engagement of α1β1 and α2β1, and although PHSRN-K-RGD conferred minimal scEEO generation alone the incorporation of both adhesion ligands in the MIX formulation robustly promoted the growth of the scEEO despite the lower (1.5 mM vs 3 mM) concentration of GFOGER in the MIX (1.5 mM GFOGER + 1.5 mM PHSRN-K-RGD ) formulation. When visualized by brightfield imaging on sequential days after seeding single cells, organoids appeared to emerge more slowly in PEG matrices (soft, 3wt%) than in Matrigel, yet they retained similar morphologies and diameters by day 9 of culture ( Fig 2D ) suggesting that the 3wt% gels function with comparable efficacy to Matrigel. Having confirmed the requirement for GFOGER , and comparable scEEO emergence in both GFOGER -only and MIX gels, we next compared how matrix stiffnesses impacted scEEO emergence by varying the PEG polymer content from 3wt% (~300 Pa) to 7wt% (~6 kPa) to mimic physiologic 58 and pathologic 59 uterine regimes, respectively ( Fig 2E – F ). The efficiency of scEEO formation, as assessed by the number of organoids at day 14 relative to Matrigel, was slightly inversely correlated with matrix stiffness ( Fig 2E ). Moreover, organoid diameter distributions in stiffer matrices, where organoids were sparser, skewed slightly higher in the 5wt% and 7wt% gels compared to softer 3wt% matrices ( Fig 2F ). Consistently, the organoids cultured in stiffer matrices (5wt% and 7wt%) manifested the meandering, crenelated morphology described above, but not in softer PEG matrices or Matrigel ( Fig S4 in Data File 1 ). These results suggest that EEO clonal establishment is robust across a range of biophysical properties ( Fig 2E – F ) and provide a preliminary indication that tissue stiffness influences epithelial morphology and behavior. Based on these results, we focused further investigations using the more physiologically relevant 3wt%- MIX gels to characterize scEEO architecture, hormone response and cellular heterogeneity in the synthetic ECM. PEG-derived scEEOs exhibited proper polarity based on basolateral laminin (LMN) deposition and apical F-actin accumulation, comparable to those generated in Matrigel ( Fig 2G ). Moreover, scEEOs generated in synthetic matrices retained their epithelial origin and architecture as assessed by EpCAM and F-actin staining ( Fig S5A – B in Data File 1 ), were mitotically active ( Fig S5C in Data File 1 ), and exhibited appropriate response to the sex steroids E 2 (1nM, 14 days) or the synthetic progestin (MPA, 500 nM, 14 days). As demonstrated by immunohistochemical staining of progesterone receptors (PGR), EEOs cultured in both hydrogels express PGR when maintained with E2 ( Fig 2H ). Treatment with progestin reduced nuclear staining of PGR while maintaining estrogen receptor (ESR1) ( Fig S5E in Data File 1 ). Downregulation of PGR RNA transcript, but not ESR1, was also in response to progestin treatment by qPCR ( Fig 2I ). To confirm scEEO maturation, live-imaging and immunostaining revealed distinct populations of motile ciliated and secretory epithelial cells ( Fig S5D in Data File 1 ; Video S1 – S2 in Data File 2 ). These results were consistent across several tissue donors (N=8). Altogether, these results justified the use of soft (3wt%) PEG hydrogels, functionalized with a dual presentation of GFOGER and PHSRN-K-RGD ( MIX ), as a functional synthetic ECM to generate and maintain long-term cultures of endometrial organoids ( Fig 1D ). Our results examining organoids in synthetic matrices demonstrated that both the biophysical and the molecular properties of the hydrogels are important to promote adequate scEEO growth and function; therefore, we set out to evaluate, in parallel, how adhesive cues and matrix stiffness influenced ESC behavior in 3D ( Fig 3A ). Embedded ESCs cultured in the gels were viable in all conditions tested ( GFOGER , PHSRN-K-RGD , or MIX for the 3–7% PEG) and maintained comparable cell numbers across all conditions (data not shown). However, ESCs adopted a more characteristic stellate morphology in softer (3wt%) hydrogel microenvironments compared to the stiffer (5wt%) gels which constrained fibroblast spreading and outgrowth ( Fig 3B ; 7% not shown). Furthermore, in support of previous studies 26 , hydrogels containing PHSRN-K-RGD further provided a modest increase in stromal dispersion and spreading, as assessed by the fluorescence projected area covered on day 7 of culture, compared to matrices containing GFOGER alone where cells retained a more clumped distribution ( Fig 3C ). Although the ESCs can enzymatically remodel the synthetic matrices due to the incorporation of the MMP cleavable crosslinker, PEG gels remained intact at the 15-day time point, whereas cultures in Matrigel or collagen I gels have shrunken and at least partly disintegrated by this time point ( Fig 3D ). This ability to maintain long-term cultures enabled us to assess stromal function in the gels by maintaining the 3D cultures under sex hormone stimulation throughout the 15-day experiment to mimic the proliferative (E 2 1nM, 15-days, “E”), the secretory (E + 500 nM synthetic progestin MPA, “PE”), or menstrual (12 days of PE treatment followed by hormone withdrawal and treatment with 10mM RU-486,72 hrs, “W”) phases of the menstrual cycle. In response, subsets of ESCs underwent decidualization in response to progestin treatment, a differentiation process necessary for the establishment of pregnancy, characterized by the secretion of pro-gestational protein prolactin (PRL) and morphological changes toward an epithelial-like decidual cell. 60 – 62 Although increased matrix stiffness curtailed ESC elongation in the E-containing medium before decidualization, ESC developed characteristic rounded epithelioid morphologies in response to progestin in the MIX gels, in both standard ESC culture media as well as EEO media ( Fig 3E ). Detection of PRL secretion in the spent media confirmed that ESC embedded in 3D gel underwent robust decidualization across all matrix stiffnesses response to progestin treatment ( Fig 3F ), suggesting that stiffness is not a strong influence on this metric of stromal hormone sensitivity. We next investigated the effects of exogenous treatment with interleukin-1β (IL1B), a prototypical inflammatory cytokine implicated in the pathogenesis of endometriosis, 63 on ESC phenotype in the synthetic gels using a dose of 1 ng/mL (57 pM) as this concentration is well above the reported 0.2 ng/mL IC50 for suppressing the decidualization response of primary human stromal cells in vitro. 64 , 65 Compared to control groups, which exhibited robust proliferation at day-15 in the 3wt% MIX hydrogel in both E and PE media with 50–60% of cells incorporating EdU over a 24-hour incubation period ( Fig 3G – H ), IL1B suppressed ESC mitotic activity resulting in ~5% of cells incorporating EdU ( Fig 3G ) and induced an increase in apoptosis ( Fig 3I ). Furthermore, treatment with IL1B also dramatically suppressed PRL production by day 15 even in the presence of the progestin MPA ( Fig 3J ). Brightfield and immunofluorescent examination of stromal morphology in response to the inflammatory cue revealed a drastic change toward an elongated morphology, characterized by a thin architecture ( Fig S6A – B in Data File 1 ) and decreased cell surface area ( Fig S6C in Data File 1 ). This change was attributed to an apparent increase in ECM remodeling capacity characterized by increased cellular motility ( Videos S4 – S13 in Data File 3 ). These results suggest that despite a reduction of mitotic activity, IL1B induces a more activated ESC phenotype in 3D. In line with this result, morphological assessment revealed that ESC cultured in Matrigel also developed the phenotype akin to those exogenously treated with IL1B ( Fig 3K ), highlighting the limitations conferred by Matrigel to study the stroma 28 . Altogether, these results demonstrate that the relatively soft, dual integrin ligand (3wt% MIX ) synthetic matrix formulation is suitable as a “one size fits all” scaffold for culturing both stromal cells and epithelial organoids. In the endometrium, both stromal and epithelial cells express hormone receptors for E 2 and P4, yet proper tissue-level function is mediated by the crosstalk between these cell types. 66 – 69 Parsing this crosstalk in vitro using Matrigel - which contains numerous exogenous growth factors, rapidly degrades, and largely lacks the physiologic ECM molecules for stromal cells – is arguably a fraught endeavor. Thus, we used the synthetic matrices described to develop a co-culture model using defined cell ratios that were shown to approximate those observed in vivo ( Fig S7 in Data File 1 ) to interrogate these dynamic phenomena. Donor matched EEOs and ESCs populations (N=12) were co-cultured in the synthetic ECM (3wt%- MIX ) using a fully defined common media (see Methods ) and maintained for up to 15 days of culture ( Fig 4A ). Immunostaining demonstrated the persistence of morphologically well-defined stromal and epithelial populations ( Fig 4B ; Video S3 in Data File 2 ). All cultures were viable, with minimal cell death, throughout the length of the experiments as demonstrated by live/dead staining ( Fig S12 in Data File 1 ) at day 15. Co-cultures were exposed to hormone stimulation designed to mimic the proliferative (E 2, ‘E’), secretory (E 2 + MPA, ‘PE’), or menstrual (72 hr PE withdrawal + RU-486 (10 µM, ‘W’) phases of the idealized human menstrual cycle ( Fig 4C – D ). To determine whether the PEG-based co-culture model could phenocopy canonical molecular pathways involved in the human menstrual cycle, we characterized the co-cultures (N = 5 donors) by both transcriptomic and functional analysis after hormonal treatment. First, we implemented bulk RNA sequencing (n=7) and demonstrated that after 15 days, co-cultures were transcriptionally distinct from either ESC or EEO monocultures ( Fig S8B , Fig S19 in Data File 1 ). Further analysis using a generalized linear model compared gene expression between ESC monocultures and co-cultures using Benjamini-Hochberg (BH) multiple hypothesis test correcting 70 , resulted in 7,821 genes to be differentially expressed at an adjusted p-value threshold of 1×10 −5 ( Fig S8C in Data File 1 ). These findings are suggestive of a crosstalk between these stromal and epithelial populations when cultured together. Next, we analyzed the response of co-cultures with respect to hormonal stimulation. Further transcriptomic analysis of the co-cultures by hierarchical clustering demonstrated that the expression of endometrium-associated genes segregated samples by hormonal treatment rather than donor variability (N=2 donors), and adequately reproduced gene expression profiles associated with their respective menstrual cycle phases ( Fig S9A in Data File 1 ). Across the menstrual cycle, P4 is known to regulate numerous cellular processes that lead to the preparation of implantation including immune-modulatory, proteolytic MMP activity and cellular differentiation of the epithelial and stromal compartments toward a secretory phenotype. 1 , 71 , 72 To identify the molecular functions, biological processes, and pathways differentially regulated by the experimental hormone treatments, we performed Gene Ontology (GO) analysis and detected key reproductive biological processes associated with progesterone signaling to be significantly (p>0.05) enriched in the progestin-treated secretory and menses groups, including ECM composition and receptor signaling activity ( Fig 4E ). Gene network visualizations of these pathways confirmed that, in response to hormone treatment, the co-cultures reproduced P4-induced expression of specific secretory phase associated genes including PRL , PAEP , and ECM-associated genes and downregulation of CPM , MMP9 , and cytokine-mediated inflammatory genes ( Fig S9B in Data File 1 ). Finally, Gene Set Enrichment Analysis (GSEA) further confirmed recapitulation of canonical menstruation-associated hallmarks ( Fig S10D in Data File 1 ), including significant upregulation of senescence, inflammatory 66 , and hypoxia 73 in the withdrawal treatment group compared to E or PE ( Fig 4F ). To expand on these transcriptomic findings, we also observed expression of matrix metalloproteinases (MMPs), including those primarily produced by either the stromal (e.g., MMP-2) or epithelial populations (e.g., MMP-7, MMP-26). 11 , 74 To corroborate these transcriptomic findings, we also measured temporal changes in matrix metalloproteinases (MMPs) from the spent media ( Fig S14C in Data File 1 ). Together, we co-cultures recapitulated similar temporal expression trends that are observed in vivo 75 , 76 , specifically a reduced MMP-1, -3 and -10 expression in the secretory phase compared to proliferative group and reduced MMP-2 compared to the menses group ( Fig S10C in Data File 1 ). Finally, we experimentally validated the co-culture model using functional assays to assess the morphologic and biochemical changes observed across the human menstrual cycle. First, we examined whether the model captured morphologic features of secretory endometrial glands in response to hormonal stimulation observed in vivo ( Fig 5A ). Progestin treatment caused changes in EEO morphology characterized by thickening of the pseudo-stratified columnar epithelial layer and increased epithelial invaginations ( Fig 5B ; Fig S10A in Data File 1 ). Moreover, epithelial secretory function was demonstrated by positive staining of progestogen-associated endometrial protein (PAEP), a marker of the secretory epithelium, in the lumen of the organoids treated with progestin ( Fig 5C ; Fig S10B in Data File 1 ). Similarly, a robust induction of stromal decidualization in response to progestin also occurred in the co-cultures by day 15 of the experiment resulting in increased secretion of PRL in the secretory groups and reduced in the menses group ( Fig 5D ). Moreover, the characteristic negative regulation of PGR during the secretory phase compared to the proliferative phase observed in vivo 77 , 78 was demonstrated via immunostaining in both epithelial and stromal populations in vitro ( Fig 5E ). In accordance with literature observations 6 , the temporal analysis across the 15 days of treatment showed that co-cultures maintained in baseline E2 conditions showed no significant increase in PRL ( Fig 5F ), but co-cultures treated with the progestin MPA showed a significant increase in PRL production starting 9 days after MPA treatment initiation ( Fig 5F ). Finally, we also observed a significant (p=0.031, N=3) doubling of apoptosis ( Fig 5G – H ; Fig S11 in Data File 1 ) in epithelial glands of the co-cultures in the menses groups (3.06 Casp + /cm 2 ) compared to the secretory groups (1.37 Casp + /cm 2 ) mimicking the increased cell death observed during the induction of menstrual phase in vivo. 79 , 80 Altogether, these results demonstrate the establishment of a co-culture model of the endometrium can recapitulate in vitro several of the hormone-induced phenotypic, morphologic, and biochemical changes associated with the idealized human menstrual cycle and could be used to investigate on reproductive function in mechanistic fashion. The suppression of stromal cell response to progestin in the presence of IL1B ( Figs 3H , J ) motivated us to analyze the molecular and phenotypic responses of epithelia and co-cultured stromal and epithelial cells to IL1B, hypothesizing that outcomes in the latter case would likely be influenced by stromal-epithelial cross talk. Separate proliferative, secretory, and menses treatment groups were co-stimulated with IL1B (1 ng/mL) throughout the 15-day experiment ( Fig 4D ). First, in accord to the results described above ( Fig S6 in Data File 1 ), the stromal population in co-cultures also developed an elongated (inflamed) morphology ( Fig S15 in Data File 1 ). Furthermore, co-cultures exposed to the inflammatory stimulant IL1B showed suppression of a stromal decidualization response. Control co-cultures showed a canonical increase in production of PRL following progestin exposure through the 15 days of treatment ( Fig 5F ); however, PRL production in co-cultures treated with IL1B showed no difference in the presence or absence of progestin by day 15 despite an unexpected spike in PRL production in the immediate 3 days after stimulation with the inflammatory cue which fell to negligible levels for the remainder of the 15 days ( Fig. 5F ). Analysis of the inflamed (IL1B treated) co-cultures by bulk transcriptomic ( Fig S13 in Data File 1 ) and multiplex immunoassay ( Fig S14 in Data File 1 ) analysis confirmed the decreased PRL expression observed at the protein level ( Fig S13B in Data File 1 ). Gene expression analysis in a generalized linear model using 8,423 genes expressed across all samples comparing controls to IL1B stimulated cocultures ( Fig S13A in Data File 1 ) revealed an expected global up-regulation of several additional pro-inflammatory cytokines and chemokines (e.g., IL6, TNF-α, CCL2, CCL8, CCL5 ) and down-regulation of anti-inflammatory cytokines (e.g., IL10 ) in the IL1B-treated groups ( Fig S13B in Data File 1 ). Hallmark pathway enrichment analysis of the unstimulated and IL1B-stimulated co-cultures detected disruptions of progesterone-regulated pathways including the downregulation of ECM structural constituent, and upregulation of inflammatory receptor-ligand signaling processes ( Fig S13C in Data File 1 ). Immunostaining revealed that co-cultures with IL1B, compared to the untreated controls, further increased apoptosis in the menses phase ( Fig 5H ; Fig S11 in Data File 1 ), an observation that was corroborated by a separate assay of cell death ( Fig S12 in Data File 1 ). These findings show that IL1B can disrupt progesterone actions. We confirmed this via immunostaining for PGR expression in the co-cultures which showed reduced PGR signal, even in the E-containing media, ( Fig 6A ) in both the epithelial and stromal populations. Lastly, stimulating the co-cultures with IL1B caused a significant thinning of epithelial morphology causing a shift from the pseudo-columnar architecture in the controls toward a squamous cell structure with thin apical F-actin staining ( Fig 6B ). Whereas progestin treatment significantly increased cell height in the control co-cultures, as characterized by the distance from apical to basolateral side (E vs PE; p=0.0098), in the IL1B-treated groups, epithelial cells were thinner than in controls (PE vs PE+IL1B; p <0.0001) and were relatively unaffected by progestin ( Fig 6C ). This flattening phenomenon closely mimics the de-differentiation events observed in the wound repair cell response of the intestine. 81 The change in morphology of the epithelial cells was accompanied by a slight, but non-significant (p=0.127), increase in spacing between nuclei compared to the round nuclei in the controls at the single cell level ( Fig 6B ; Fig S16A in Data File 1 ), suggesting the flattened cell in the IL1B-treated groups have a slightly increase in projected cell area compared to the controls to compensate for the reduction in cell height while keeping similar volume. An alternate metric of projected cell area, the density of nuclei counted per area of organoid monolayer surface, was comparable in control and IL1B-treated co-cultures ( Fig S16B in Data File 1 ) though analysis of the morphometric changes cell volume is a topic for a future investigation. Altogether, these results highlight the extent by which IL1B can induce the overexpression of additional pro-inflammatory cytokines associated with reproductive diseases 82 – 84 and can potentiate progesterone resistance, a phenomenon that is a characteristic feature of endometriotic disease in vivo. 85 , 86 Chronic inflammation of ectopic lesions is a characteristic feature of endometriosis; wherein endometrial epithelial and stromal cells are found growing outside the uterus. 87 We thus next assessed whether inflammation altered epithelial proliferation in ways that may illuminate mechanisms of disease. Tissue-recombinant murine models have previously demonstrated that estrogen and progesterone mediate epithelial mitogenic processes indirectly through signals from the hormone-responsive stromal fibroblast, 69 , 88 , 89 therefore we investigated whether IL1B exerts its effect via similar mechanisms. To test this, we measured the effects of IL1B treatment on epithelial growth, in epithelial monocultures and in co-cultures ( Fig 6D ; N=4 donors, matched cultures), using both morphological analysis of organoid diameters and measuring DNA synthesis. First, we measured changes in organoid size (outer diameter) across the 15 days of culture as a proxy for growth rates ( Fig S17A in Data File 1 ). As was observed with endometrial stromal cells in monoculture ( Fig 3G – H ) IL1B treatment significantly (PE vs PE+IL1B; p=0.0121) suppressed the growth of the epithelial organoid diameter in monoculture, compared to controls ( Fig S17A in Data File 1 ), thus we hypothesized that IL1B would negatively impact epithelial mitotic processes in the co-cultures. Unexpectedly, when co-cultured with stroma, IL1B treatment caused an observable increase in epithelial organoid diameter compared to untreated controls ( Fig 6E ). Quantification of the organoid diameter across the 15 days of culture demonstrated that IL1B stimulation positively impacts scEEO diameter growth in the co-cultures, but not in the monocultures (Coculture vs monoculture; p=0.0026) resulting in a mean fold change in organoid diameter of 7.3 and 3.9-fold in the co-cultures and monocultures, respectively ( Fig 6F ). As scEEO diameter is only a proxy for changes in cell number, the mitotic activity of these cultures, at day 15, was examined by measuring EdU incorporation by the organoids after 24 hours of incubation ( Fig 6G ). In EEO monocultures, a non-significant (p > 0.05) trend toward reduction of DNA synthesis in scEEOs upon treatment with IL1B (E=1822; PE=2333; E+IL1B=1185; PE+IL1B=1407 EdU + /mm 2 ; N=3) and a surprising trend toward an increase in DNA synthesis in the presence of progestin compared to E-only conditions ( Fig 6H ) was observed. In stark contrast, stimulation with IL1B resulted in a greater than 3-fold increase (PE = 228; PE+IL1B = 978 EdU + /mm 2 ; P= 0.0014; N=4) of scEEO DNA synthesis in co-cultures compared to controls ( Fig 6I ), consistent with the observed increase in organoid diameters in IL1B-treated co-cultures ( Fig 6E ). Further, unlike monocultures, the proliferative index of scEEOs in co-culture decreased in progestin-treated controls compared to E-only controls, as would be expected in the transition from proliferative to secretory phase in vivo 90 . These phenotypic changes in epithelial mitotic activity mirror the increased epithelial proliferative profiles observed in the eutopic and ectopic endometriotic tissues in vivo. 91 , 92 Although the scEEO monocultures appeared to have a greater proliferative index than the co-cultures in some groups, this difference could be attributed to the contribution of the ESC in mediating hormone signaling, as shown in other models. 93 Perhaps because of this increased growth rate, IL1B stimulation in co-cultures resulted in greater organoid collapse by day 15 and could also explain the ~1.5-fold (p=0.012) increase in apoptosis (Casp-3 + staining) observed in the epithelial ( Fig 5G ) populations that resulted in an overall increase in cell death (ethidium homodimer staining, Fig S12 in Data File 1 ). Altogether, these results support a working model by which inflammatory signals and sex hormone signaling in the endometrium are partially mediated indirectly via the stroma ( Fig 6J ).

The endometrium is a marvelous example of regenerative biology wherein sex hormones mediate rapid growth and maturation of the tissue, accompanied by equally dynamic changes in ECM-cell interactions that are both mechanically and molecularly linked to reproductive function. 94 These remarkable regenerative properties also contribute to common debilitating diseases like endometriosis, for which new therapies are desperately needed. Here, we developed a model for dissecting molecular and phenotypic consequences of endometrial epithelial-stromal crosstalk in long term (>15 day) cultures of patient-derived endometrial cells by defining a completely synthetic extracellular matrix hydrogel that is tailored to both replace Matrigel for organoid culture, and simultaneously to support stromal culture. We then used this model to show how the hormone-dependent behaviors of the endometrial epithelium in co-culture with stroma diverge from those in monoculture, observing for example that the pro-inflammatory cue IL1B appears to drive the endometrial co-cultures, but not monocultures, to a state that phenocopies features of diseases like endometriosis. 83 , 87 The synthetic ECM described here overcomes certain limitations of organoid technologies and other co-culture approaches for parsing cell-cell communication in general, especially in the endometrium. Natural matrices like Matrigel and collagen – which have been used either alone for co-culture 28 , 29 , 61 , 95 or combined in creative ways 96 , 97 – include many extraneous signaling molecules that may drown out the signals produced by the cells they support. This limitation is underscored by the inability to design experiments that require longer term stable cultures, as is the case for hormone signaling, or require specific ECM stiffness. The main constituent of this synthetic ECM, PEG, is a blank slate known for its relative lack of interaction with proteins. Thus, growth factors, cytokines, and other molecules produced by each cell type can freely dominate the cell-cell communication networks. By design, the synthetic ECM comprises only a minimal set of biological cues: two integrin ligands, two ECM-binding proteins, and a peptide crosslinker, all produced by standard, well-defined chemical peptide synthesis methods. These minimal cues, which were modified from a previous formulation used for organoids 27 and monolayer cell line cultures 26 , include ligands for a spectrum of integrins expressed differentially by epithelial and stroma (see Results) populations and enable cell-mediated remodeling of the microenvironment as well as the accumulation of cell-produced ECM. By demonstrating that this synthetic matrix is sufficient to generate scEEOs from the original primary glands without the need to first expand in Matrigel ( Fig S18 in Data File 1 ), we provide a hydrogel alternative to alleviate the dependency on Matrigel. Because the synthetic matrix is relatively simple and is chemically defined, it is straightforward to implement reproducibly for the culture of organoids and co-cultures. Reproducibility within and between labs is highly desirable to analyze phenotypes of patient-derived tissue models due to the inherent biological variability in the samples themselves. Finally, although only an initial characterization is provided here, the synthetic ECM arguably offers greater potential to mimic and study certain features of endometrial diseases compared to ECM-free co-culture models 98 , 99 , as its biophysical properties can be tuned systematically to mimic the “soft” regimes of healthy endometrium or the “stiff” regimes of the myometrium 100 or fibrotic tissue ( Fig 2A ). Steroid hormone-driven stromal-epithelial crosstalk in the endometrium has been illustrated in mouse models involving cell type-specific receptor knockouts in the eutopic endometria 71 , 77 and recombinant implantation of tissue fragments in the renal capsule, 69 , 93 but crosstalk in normal and diseased human co-cultures has been more challenging to parse due to the disparate matrix environments preferred by each cell type 101 , 102 , and the relative short-lived nature of cultures in Matrigel or collagen. For example, work by Rawlings and co-workers showed stromal-epithelial crosstalk enabled a simplified medium to support an elegant collagen gel-embedded “assembloid” co-culture model that captured both normal and drug-skewed co-evolution of the stromal population to subpopulations of decidual and acutely senescent fibroblasts 61 . They thus defined a tantalizing model for analysis of embryo implantation if the limitations of the culture longevity, which disintegrate before the entire sequence of implantation steps is realized under some conditions, and the limited access organoid apical interface, can be overcome 61 . Thus, the long-term capacity of co-cultures in our synthetic hydrogel may provide one approach to extending the assembloid model if similar evolution of fibroblasts populations can be demonstrated. In concordance with a different short term (7-day) co-culture model, employing EEOs cultured in Matrigel in a Transwell membrane above a coverslip coated with stroma 103 , we observed the expected suppression of proliferation by E 2 and P4 compared to E 2 in control co-cultures ( Fig 6F ; Fig S17 in Data File 1 ). However, we are also able to observe more pronounced effects at the two-week time point, especially in cultures treated with IL1B ( Fig 6F ; Fig S17 in Data File 1 ) and provide evidence that the inflammatory cytokine IL1B can also modulate epithelial proliferation via the stroma ( Fig 6J ). Indeed, the response of co-cultures to an inflammatory cue at extended time points is one of the most striking findings in this study, as it may shed light on inflammatory disorders such as endometriosis. We chose the cytokine IL1B as a prototypical modulator of both physiological inflammation in the eutopic endometrium 104 and in pathophysiological inflammation in endometriosis lesions 63 , recognizing that our co-culture system configuration is similar to a lesion model in its cyst-like structure of epithelia. IL1B is produced by myeloid cells, endothelia, epithelia, and fibroblasts throughout the body 105 , 106 . Upon binding to the IL1 receptor type 1 (IL1R1) and recruitment of IL1R3, IL1B signals through the NF-kB pathway and through JNK, ERK1/2 and p38 to the AP1 complex to trigger production of additional cytokines, including IL-6, IL-8, and IL-1. 105 , 106 The canonical pathways activated by IL1B are shared by other interleukins, cell stressors, and pathogens, making IL1B a reasonable proxy for other triggers of innate inflammatory signaling 105 . IL1B signaling is regulated at multiple levels, including: competition for binding to IL1R1 by the decoy ligand IL1ra; neutralization by shed receptor sIL1R1; neutralization by binding to the non-signaling receptor IL1R2; and downregulation of IL1R1 by receptor-mediated endocytosis 64 , 105 , 106 . In the human endometrium, expression of IL1B is present across the menstrual cycle and is implicated in the physiology of implantation and maintenance of pregnancy, with conflicting reports on how alterations in expression levels in the endometrium, uterine fluid and systemic circulation are correlated with fertility. 104 , 107 – 114 Furthermore, the presence of IL1B is consistently implicated in the pathophysiology of endometriosis, though its contributions to lesion phenotypes and symptoms remain challenging to dissect in a landscape of large clinical variation in both patient and lesion characteristics, and the complex dynamic interplay between the entire network of regulatory molecules involved in IL1B action 63 – 65 ; nevertheless, inhibition of IL1B is the focus of an ongoing clinical trial in endometriosis patients ( NCT03991520 ). The reported ranges (0.01 – 2 pM) of IL1B concentration in plasma 113 , 115 , 116 and uterine fluid 108 , 117 are far below the reported 100–1000 pM KD values for receptor binding 106 , though ~25 pM values in the saliva of menopausal women are close. 118 In vitro culture models that investigate endometrial cell responses to IL1B typically employ ligand concentrations in the range of 0.1–10 ng/mL (5.7–570 pM), with a reported IC50 for inhibiting decidual responses of human endometrial stromal cells of 0.2 ng/mL (11 pM). 64 , 65 We had previously observed that the concentration of IL1B in the local microenvironment of hydrogel-encapsulated endometrial cells reaches a steady state about 20–25% lower than the external concentration in the culture medium when dosed at 10 ng/mL 55 , suggesting that cells may be internalizing and degrading the ligand, thus we chose a ligand concentration of 1 ng/mL (57 pM) for stimulating gel-encapsulated cells to ensure a chronic tonic stimulus mimicking what might be present in an endometriosis lesion, given the gradients produced in 3D tissues by ligand diffusion and consumption. When the co-culture became exposed to this concentration of IL1B, organoid diameter ( Fig 6F ) and epithelia mitotic activity ( Fig 6I ) are both increased, even in the presence of MPA compared to the E2 condition, whereas the canonical response to progestin resulted in a decrease in epithelial proliferation in the control co-cultures and in monocultures. We speculate that this observation of an IL1B-MPA synergy may have clinical relevance for the observed failure of some endometriosis patients to respond to progestin therapies. Moreover, greater proliferation indices are observed in the eutopic endometrium of endometriotic patients, compared to controls 90 , 92 . Similar to our observations about IL1B inducing profound epithelial morphological thinning along with increased proliferation, in deep infiltrating endometriosis lesions, an “invasive front” characterized by highly flattened epithelia with greater proliferative index than epithelia in a more quiescent core has been described. 119 While more translational validation is warranted to confirm mechanistically similar behaviors, the results are suggestive that the model may capture features that are important clinically. Importantly, the timely and acute induction of inflammatory mediators are also well known to be critical for maintaining normal reproductive physiology during implantation. 104 This may explain why a transient (day 3) spike in PRL secretion was observed in the co-cultures in response to IL1B as acute-phase inflammatory insults that activate the nuclear factor–kappaB (NF-KB) system, a master regulator of cell survival and proliferation in endometriotic disease 120 , have been shown to also regulate PGR expression in vivo 63 , 77 . IL1b can enhance stromal decidualization 121 and blastocyst implantation 35 , 122 during short exposure times, suggesting IL1B may have a biphasic response. Because most assays were performed at end point measurements (day 15) in this study, future mechanistic studies that dissect the timing (acute vs chronic), abundance (low vs high concentration), and synergism with sex hormones (estrogenic vs progestin conditions) will be critical to fully understand the extent by which IL1B expression can positively promote physiologic (e.g., implantation) or induce pathologic processes (e.g., endometriosis, infertility) 86 , 120 , and ultimately will help identify therapeutic approaches for targeting the disease while maintaining reproductive function ( Fig 6J ). Finally, we are just beginning to explore how the properties of the synthetic ECM may enable more complex tissue-level components beyond the organoid-stroma culture described herein. This relatively simple model does not reflect the tissue-like architecture and heterogenous cell densities that makes up the in vivo tissue. Although we focused on primarily on paracrine crosstalk, and not on cell-cell contact as a method for cell communication in this study, we did observe that stromal cells can migrate through the gel to become juxtaposed by the epithelia ( Fig 6B ). Nonetheless, alternative approaches, including spheroid aggregate cultures embedded in the synthetic ECM, may be better suited to directly investigate communication via cell-cell contact. Furthermore, while we did not aim to mimic tissue architecture, we observed that organoids close to the gel-liquid interface sometimes merge with each other and erupt to form mature epithelial organoid-monolayers structures that are more representative the glandular-luminal epithelial tissue-like structures of a true endometrial mucosal barrier ( Fig S21 in Data File 1 ). Thus, we are now defining protocols that foster efficient and reproducible organoid fusion model. Finally, while this model employed two key cellular components of the endometrium, a parallel effort to represent the heterogenous cell populations of the endometrial microenvironment by combining the immune component to the stroma and epithelia, as we have described for pancreatic tumors 32 , is on-going with the goal of incorporating these complex endometrial models to perfusable blood vessels in microfluidic devices, thereby extending our initial findings that we can form microvascular networks in synthetic PEG gels. 123 , 124 We speculate that the co-culture model described here can be productively extended by the research community as a platform to mechanistically parse stromal-epithelial crosstalk in the human endometrium.

There are a few limitations to this study. First, the sample size for scRNAseq and proteomic analysis is not sufficient to make any definitive conclusion about the different phases of the menstrual cycle endometrium. Only subset from datasets were used to help design the model and were not intended to atlas of the endometrium. Secondly, synthetic hydrogel design only includes two peptides designed to engage the integrins expressed by the stromal and epithelial populations. Although this bottom-up approach was sufficient to support the metrics we set to achieve in this study, the role of additional components of the ECM biomaterials remains unexplored. Similarly, as discussed, the endometrial microenvironment is composed of many different cell populations, in this study only the epithelial and stromal populations were examined. Third, most of the analysis of this study were performed at end-point measurements at day-15 to accommodate for study design; thus, future studies can be designed with greater temporal granularity as well analysis of the cell populations with single cell resolution. Fourth, this study is not intended to recreate the intact regenerative capacity and architecture of the native endometrium, but rather a model than recapitulates hormone-mediated features of the menstrual cycle in vitro thereby enabling the study of endometrial cell-communication and ECM-cell interactions. Finally, enrollment did not consider general population demographics including the social-economic status, ethnicity and disease status and included donors undergoing treatment for a type of uterine disorders. While the use of heterogenous populations demonstrated the potential of hydrogel to function on cells obtained from distinct clinical phenotypes and cycle phases, the impact of disease status on cell behavior and patient diagnosis was not examined in this study. Prospective studies with appropriate power analysis will use this hydrogel platform to investigate this in more detail.

The endometrium is the mucosal lining of the uterus in which the establishment and maintenance of pregnancy occurs 1 . It is a highly dynamic tissue that undergoes spatial and temporal changes in response to endocrine signaling by the ovarian sex hormones estradiol (E 2 ) and progesterone (P4). Cyclical changes in these hormones dictate the timing and functional capabilities of the endometrium to support nidation by driving cell-specific morphological and biochemical processes 2 . Histologically, the endometrium is primarily composed of the hormone-responsive epithelium and specialized reticular fibroblasts (stromal cells) embedded in a dynamic extracellular matrix (ECM). 1 , 3 , 4 Endocrine-induced reproductive function is mediated through both direct and indirect mechanisms governed by the crosstalk between these cell populations. 5 – 8 The idealized 28-day human menstrual cycle is characterized by three distinct phases in which the endometrium undergoes E 2 -mediated tissue regeneration (the ‘proliferative phase’) followed by a 14-day P4-dominant phase (the ‘secretory phase’) that governs the differentiation processes necessary for the successful establishment of pregnancy ( Fig 1A ). In the absence of embryo implantation, a sharp withdrawal of hormones triggers a cascade of inflammatory processes that result in the shedding of the endometrial tissue (the ‘menstrual phase’) after which the cycle initiates again. 1 , 4 Despite the fundamental roles E 2 and P4 play in maintaining reproductive functions, a thorough understanding of the cellular mechanisms driving these processes remains elusive due in part to a lack of physiological models that recapitulate the complex multi-cellular human condition. 1 , 9 Experimental models capable of recapitulating epithelial-stromal crosstalk are necessary to understand human reproductive physiology and pathology. Endometrial stromal cells (ESCs) are routinely used for experimentation in reproductive biology due to their ability to be cultured and expanded in vitro , however culturing epithelium has been more challenging. Recently, protocols to expand and culture primary endometrial epithelial cells as endometrial epithelial organoids (EEOs) have transformed the experimental landscape for in vitro analysis of endometrial behavior. 10 , 11 Organoids are self-organizing 3D cell structures that retain many physiologically relevant and functional features of their native tissues of origin. 12 Thus, EEOs provide opportunities to investigate human reproductive function, understand disease pathogenesis 13 , and test therapeutic compounds for clinical applications. 10 , 11 EEOs are typically generated from isolated primary endometrial glands embedded in Matrigel, a complex, murine tumor-derived ECM primarily comprising heterogenous basement membrane (BM) proteins and growth factors (GFs), together with a defined culture medium designed to enrich the expansion of the stem/progenitor compartment. 14 , 15 Cultured in Matrigel, clonally-generated EEOs retain columnar epithelial architecture of the native tissue and form polarized cyst-like structures that are genomically stable, hormone-responsive, and cell-heterogenous. 10 , 11 , 16 However, the reliance on Matrigel to culture organoids presents a critical obstacle for numerous experimental, analytical, and therapeutic applications for dynamic and hormone-sensitive tissues such as the endometrium. 17 , 18 Matrigel’s inherent lot-to-lot variability, poorly-defined molecular composition, limited biophysical properties, and poor suitability for co-culture of epithelia with stromal cells have driven efforts to develop synthetic ECMs that are fully defined, modular, and tunable. 13 , 19 – 27 Although endometrial epithelia and stroma are often co-cultured in naturally-derived ECM hydrogels 28 – 30 , these models are limited by rapid ECM breakdown, which ultimately leads to highly abbreviated experimental conditions and limited control over the biophysical and molecular ECM compositions that can reliably support the morphogenic and functional behavior of multiple cell types. Recently, the development of fully or semi-synthetic ECMs to culture enteric organoids were reported for both murine 24 , 31 and human 18 , 22 , 27 , 31 tissues. We have previously demonstrated that poly(ethylene glycol) (PEG)-based hydrogel systems can be applied to model some aspects of the endometrial mucosal barrier, 26 and developed a completely synthetic hydrogel formulation for the culture, differentiation and passaging of human intestinal enteroids that also supported culture of endometrial epithelial cells 27 . This synthetic ECM also supports the culture of pancreatic tumor organoids with a complex stroma. 32 Despite these advances, a fully-defined synthetic ECM that can support the long-term culture and function of multiple endometrial cell populations, especially under the dynamic conditions of hormone variation in a simulated menstrual cycle, is still an unmet need. Inspired by cell-matrix interactions in native tissue, we designed and synthesized a synthetic ECM that targets cell-specific integrins on epithelia and stroma, supports cell-secreted ECM deposition, and mimics the biophysical properties of the endometrium in health and disease. We show that this bio-labile PEG-based synthetic ECM supports the 3D co-culture of primary human EEOs and ESCs and recapitulates characteristic phenotypic properties of the endometrial responses to sex hormone changes across a simulated 15-day menstrual cycle in vitro . We used this 3D co-culture model to parse stromal-epithelial crosstalk and disease-related phenotypes in response to inflammatory cues.