Modelling the impact of decidual senescence on embryo implantation in human endometrial assembloids

  1. Thomas M Rawlings
  2. Komal Makwana
  3. Deborah M Taylor
  4. Matteo A Molè
  5. Katherine J Fishwick
  6. Maria Tryfonos
  7. Joshua Odendaal
  8. Amelia Hawkes
  9. Magdalena Zernicka-Goetz
  10. Geraldine M Hartshorne
  11. Jan J Brosens  Is a corresponding author
  12. Emma S Lucas
  1. Division of Biomedical Sciences, Warwick Medical School, University of Warwick, United Kingdom
  2. Centre for Early Life, Warwick Medical School, University of Warwick, United Kingdom
  3. Centre for Reproductive Medicine, University Hospitals Coventry and Warwickshire NHS Trust, United Kingdom
  4. Department of Physiology, Development and Neuroscience, University of Cambridge, United Kingdom
  5. Tommy’s National Centre for Miscarriage Research, University Hospitals Coventry & Warwickshire NHS Trust, United Kingdom
  6. Synthetic Mouse and Human Embryology Group, California Institute of Technology (Caltech), Division of Biology and Biological Engineering, United Kingdom

Abstract

Decidual remodelling of midluteal endometrium leads to a short implantation window after which the uterine mucosa either breaks down or is transformed into a robust matrix that accommodates the placenta throughout pregnancy. To gain insights into the underlying mechanisms, we established and characterized endometrial assembloids, consisting of gland-like organoids and primary stromal cells. Single-cell transcriptomics revealed that decidualized assembloids closely resemble midluteal endometrium, harbouring differentiated and senescent subpopulations in both glands and stroma. We show that acute senescence in glandular epithelium drives secretion of multiple canonical implantation factors, whereas in the stroma it calibrates the emergence of anti-inflammatory decidual cells and pro-inflammatory senescent decidual cells. Pharmacological inhibition of stress responses in pre-decidual cells accelerated decidualization by eliminating the emergence of senescent decidual cells. In co-culture experiments, accelerated decidualization resulted in entrapment of collapsed human blastocysts in a robust, static decidual matrix. By contrast, the presence of senescent decidual cells created a dynamic implantation environment, enabling embryo expansion and attachment, although their persistence led to gradual disintegration of assembloids. Our findings suggest that decidual senescence controls endometrial fate decisions at implantation and highlight how endometrial assembloids may accelerate the discovery of new treatments to prevent reproductive failure.

eLife digest

At the beginning of a human pregnancy, the embryo implants into the uterus lining, known as the endometrium. At this point, the endometrium transforms into a new tissue that helps the placenta to form. Problems in this transformation process are linked to pregnancy disorders, many of which can lead to implantation failure (the embryo fails to invade the endometrium altogether) or recurrent miscarriages (the embryo implants successfully, but the interface between the placenta and the endometrium subsequently breaks down).

Studying the implantation of human embryos directly is difficult due to ethical and technical barriers, and animals do not perfectly mimic the human process, making it challenging to determine the causes of pregnancy disorders. However, it is likely that a form of cellular arrest called senescence, in which cells stop dividing but remain metabolically active, plays a role. Indeed, excessive senescence in the cells that make up the endometrium is associated with recurrent miscarriage, while a lack of senescence is associated with implantation failure.

To study this process, Rawlings et al. developed a new laboratory model of the human endometrium by assembling two of the main cell types found in the tissue into a three-dimensional structure. When treated with hormones, these ‘assembloids’ successfully mimic the activity of genes in the cells of the endometrium during implantation. Rawlings et al. then exposed the assembloids to the drug dasatinib, which targets and eliminates senescent cells. This experiment showed that assembloids become very robust and static when devoid of senescent cells.

Rawlings et al. then studied the interaction between embryos and assembloids using time-lapse imaging. In the absence of dasatinib treatment, cells in the assembloid migrated towards the embryo as it expanded, a process required for implantation. However, when senescent cells were eliminated using dasatinib, this movement of cells towards the embryo stopped, and the embryo failed to expand, in a situation that mimicks implantation failure.

The assembloid model of the endometrium may help scientists to study endometrial defects in the lab and test potential treatments. Further work will include other endometrial cell types in the assembloids, and could help increase the reliability of the model. However, any drug treatments identified using this model will need further research into their safety and effectiveness before they can be offered to patients.

Introduction

Upon embryo implantation, the cycling human endometrium transforms into the decidua of pregnancy to accommodate the placenta (Gellersen and Brosens, 2014). Transition between these physiological endometrial states requires intensive tissue remodelling, a process termed decidualization. Notwithstanding that decidualization in early pregnancy cannot be studied directly, a spectrum of prevalent reproductive disorders is attributed to perturbations in this process, including recurrent implantation failure and recurrent pregnancy loss (Dimitriadis et al., 2020; Macklon, 2017; Zhou et al., 2019). By contrast, the sequence of events that renders the endometrium receptive to embryo implantation has been investigated extensively, starting with obligatory oestrogen-dependent tissue growth following menstrual repair. As a consequence of rapid proliferation of stromal fibroblasts and glandular epithelial cells (EpCs), which peaks in the upper third of the functional layer (Ferenczy et al., 1979), endometrial volume and thickness increases multifold prior to ovulation (Raine-Fenning et al., 2004; Dallenbach-Hellweg, 1981). After the postovulatory rise in progesterone levels, proliferation of EpCs first decreases and then ceases altogether in concert with the onset of apocrine glandular secretions, heralding the start of the midluteal window of implantation (Dallenbach-Hellweg, 1981). Concurrently, uterine natural killer (uNK) cells accumulate and endometrial stromal cells (EnSCs) start decidualizing in a process that can be described as ‘inflammatory programming’ (Brighton et al., 2017; Chavan et al., 2021; Erkenbrack et al., 2018; Salker et al., 2012). Morphological decidual cells, characterized by abundant cytoplasm and enlarged nuclei, emerge upon closure of the 4-day implantation window, meaning that the endometrium has become refractory to embryo implantation (Gellersen and Brosens, 2014). In pregnancy, decidual cells form a robust, tolerogenic matrix in which invading trophoblast cells cooperate with local immune cells to form a haemochorial placenta (Aplin et al., 2020; Vento-Tormo et al., 2018). In non-conception cycles, however, falling progesterone levels and influx of neutrophils lead to breakdown of the superficial endometrial layer and menstrual shedding (Jabbour et al., 2006).

Recently, we highlighted the importance of cellular senescence in endometrial remodelling during the midluteal implantation window (Brighton et al., 2017; Lucas et al., 2020; Kong et al., 2021). Senescence denotes a cellular stress response triggered by replicative exhaustion or other stressors that cause macromolecular damage (Muñoz-Espín and Serrano, 2014). Activation of tumour suppressor pathways and upregulation of cyclin-dependent kinase inhibitors p16INK4a (encoded by CDKN2A) and p21CIP1 (CDKN1A) lead to permanent cell cycle arrest, induction of survival genes, and production of a bioactive secretome, referred to as the senescence-associated secretory phenotype (SASP). The composition of the SASP is tissue-specific but typically includes proinflammatory and immunomodulatory cytokines, chemokines, growth factors, and extracellular matrix (ECM) proteins and proteases (Birch and Gil, 2020). Acute senescence, characterized by transient SASP production and rapid immune-mediated clearance of senescent cells, is widely implicated in processes involving physiological tissue remodelling, including during embryo development, placenta formation, and wound healing (Muñoz-Espín and Serrano, 2014; Van, 2014). By contrast, persisting senescent cells cause chronic inflammation or ‘inflammaging’ (Birch and Gil, 2020), a pathological state that underpins ageing and age-related disorders. We demonstrated that inflammatory reprogramming of EnSC burdened by replication stress leads to the emergence of acute senescent cells during the implantation window (Brighton et al., 2017; Lucas et al., 2020; Kong et al., 2021). Upon successful implantation and continuous progesterone signalling, decidual cells co-opt uNK cells to eliminate their senescent counterparts through granule exocytosis (Brighton et al., 2017; Lucas et al., 2020; Kong et al., 2021). Clearance of senescent decidual cells likely necessitates recruitment of bone marrow-derived decidual precursor cells, which confer tissue plasticity for rapid decidual expansion in early pregnancy (Diniz-da-Costa et al., 2021). Importantly, lack of clonogenic decidual precursor cells and a pro-senescent decidual response are linked to recurrent pregnancy loss (Lucas et al., 2016; Lucas et al., 2020; Tewary et al., 2020).

Based on these insights, we hypothesized that acute senescence is integral to successful implantation by creating conditions for anchorage of the conceptus in an otherwise tightly adherent decidual matrix. To test this hypothesis, we developed an ‘assembloid’ model, consisting of endometrial gland-like organoids and primary EnSC, which recapitulates the complexity in cell states and gene expression of the midluteal implantation window, improving resemblance to endometrial tissue in comparison with existing co-culture models (Cheung et al., 2021; Rawlings, 2021). We used this model to establish co-cultures with human blastocysts and demonstrate that aspects of different pathological states associated with implantation failure and miscarriage can be recapitulated in endometrial assembloids by modulating decidual senescence.

Results

Establishment of endometrial assembloids

Organoids consisting of gland-like structures are established by culturing endometrial EpCs seeded in Matrigel in a chemically defined medium containing growth factors and signal transduction pathway modulators (Supplementary file 1: Table 1; Turco et al., 2017; Boretto et al., 2017). Gland-like organoids grown in this medium, termed expansion medium, are genetically stable, easily passaged, and can be maintained in long-term cultures (Boretto et al., 2017; Turco et al., 2017). Oestradiol (E2) promotes proliferation of gland-like organoids and cooperates with NOTCH signalling to activate ciliogenesis in a subpopulation of EpC (Haider et al., 2019). Further, treatment with a progestin (e.g. medroxyprogesterone acetate [MPA]) and a cyclic AMP analogue (e.g. 8-bromo-cAMP) induces secretory transformation of gland-like organoids in parallel with expression of luteal-phase marker genes (Turco et al., 2017; Boretto et al., 2017).

We modified the gland-like organoid model to incorporate EnSC. To this end, midluteal endometrial biopsies (Supplementary file 1: Table 2) were digested and gland-like organoids established from isolated EpC (Figure 1A). In parallel, purified EnSC were propagated in standard monolayer cultures. At passage 2, single-cell suspensions of EnSC were combined with organoid EpC, seeded in hydrogel, and cultured in expansion medium supplemented with E2 (Figure 1A). The hydrogel matrix comprised 97% type I and 3% type III collagens, which are both present in midluteal endometrium (Oefner et al., 2015; Aplin et al., 1988; Aplin and Jones, 1989; Iwahashi et al., 1996), and has a predicted in-use elastic modulus (Pa) of comparable magnitude to non-pregnant endometrium (Abbas et al., 2019; Bagley, 2019). As shown in Figure 1B, gland formation was unperturbed by the presence of EnSC and assembloids resembled the architecture of native endometrium more closely than organoids. Further, decidualization of assembloids with 8-bromo-cAMP and MPA for 4 days (Figure 1C) resulted in robust secretion of decidual prolactin (PRL) and C-X-C motif chemokine ligand 14 (CXCL14) (Figure 1D). Immunofluorescence microscopy provided further evidence that decidualizing assembloids mimic luteal phase endometrium, exemplified by laminin deposition by decidualizing EnSC, induction of osteopontin (SPP1) and accumulation of glycodelin (encoded by PAEP) in the lumen of secretory glands, and downregulation of the progesterone receptor (PGR) in both stromal and glandular compartments (Figure 1E).

Establishment of endometrial assembloids.

(A) Schematic for establishing endometrial assembloids. (B) Structural appearance of hematoxylin and eosin stained secretory endometrium, E-cadherin labelled gland-like organoids, and E-cadherin and vimentin stained endometrial assembloids. Scale bar = 50 µm. (C) Schematic summary of experimental design. (D) Secreted levels of PRL and CXCL14 were measured by ELISA in spent medium at the indicated timepoints. Data points are coloured to indicate secretion in assembloids established from different endometrial biopsies (n = 3). (E) Representative immunofluorescence labelling of laminin and vimentin, progesterone receptor (PR), glycodelin, and osteopontin (OPN) in undifferentiated (day 0, top panels) and decidualized (day 4; bottom panels) assembloids. Nuclei were counterstained with DAPI. Scale bar = 50 µm. ELISA data in (B) are available in Figure 1—source data 1.

Figure 1—source data 1

Secretion of PRL and CXCL14 by endometrial assembloids.

Secretion of PRL and CXCL 14 (pg/ml) was measured by ELISA in spent medium from assembloids. Supports Figure 1D.

https://cdn.elifesciences.org/articles/69603/elife-69603-fig1-data1-v2.xlsx

We reasoned that once established assembloids may no longer require exogenous growth factors and pathway modulators for differentiation because of the presence of EnSC. To test this hypothesis, parallel gland-like organoids and assembloids were established from three endometrial biopsies and decidualized with E2, 8-bromo-cAMP, and MPA for 4 days in either expansion medium, base medium (Supplementary file 1: Table 1), or base medium with each exogeneous factor added back individually. Induction of PAEP and SPP1 was used to monitor the glandular differentiation response. As shown in Figure 2, differentiation of gland-like organoids in base medium markedly blunted the induction of PAEP and SPP1 when compared to expansion medium. Add-back of individual factors did not restore the glandular response, with the exception of N-acetyl-L-cysteine (NAC). Addition of NAC at low concentration (1.25 mM) to base medium resulted in a robust glandular response in assembloids. Thus, in subsequent experiments assembloids were grown in expansion medium supplemented with E2 and then decidualized in minimal differentiation medium (MDM), consisting of base medium containing NAC, E2, 8-bromo-cAMP, and MPA.

Characterization of a minimal differentiation medium for endometrial assembloids.

Parallel gland-like organoids (red) and assembloids (blue) were established from three endometrial biopsies and decidualized with 8-bromo-cAMP and MPA for 4 days in either expansion medium (ExM), base medium (BM), or BM with each exogeneous factor added back individually (+). Induction of PAEP and SPP1 was used to monitor the glandular differentiation. The grey bar indicates the composition of the minimal differentiation medium selected for further use (BM supplemented with NAC, E2, cAMP, and MPA). Data are presented as fold-change relative to expression levels in undifferentiated organoids or assembloids cultured in ExM+ E2. Bars present minimal, maximal, and median fold-change. * and ** indicate p<0.05 and p<0.01 obtained by Friedman’s test for matched samples. Relative expression values for biological replicates are available in Figure 2—source data 1.

Figure 2—source data 1

RTqPCR data associated with the minimal differentiation medium (MDM) experiments.

Parallel epithelial gland organoids and assembloids were established from endometrial biopsies and decidualized with 8-bromo-cAMP and MPA for 4 days in either expansion medium (ExM), base medium (BM), or BM with each exogeneous factor added back individually (red).

https://cdn.elifesciences.org/articles/69603/elife-69603-fig2-data1-v2.xlsx

Cellular complexity of decidualizing assembloids mimics midluteal endometrium

We hypothesized that, depending on the level of replicative stress (Lucas et al., 2020; Abbas et al., 2019), individual EpC and EnSC adopt distinct cellular states upon decidualization of endometrial assembloids. Based on previous time-course experiments in 2D cultures, we further speculated that divergence of cells into distinct subpopulations would be apparent by day 4 of differentiation (Brosens et al., 1999; Lucas et al., 2020). To test this hypothesis, we performed single-cell RNA sequencing (scRNA-seq) on undifferentiated assembloids grown for 4 days in expansion medium and assembloids decidualized in MDM for four additional days (Figure 3A). Eleven distinct cell clusters were identified by Shared Nearest Neighbour (SNN) and Uniform Manifold Approximation and Projection (UMAP) analysis, segregating broadly into epithelial and stromal populations within the UMAP-1 dimension and into undifferentiated and differentiated subpopulations within the UMAP-2 dimension (Figure 3B). Each cell cluster was annotated based on expression of curated marker genes, which were cross-referenced with a publicly available data set (GEO: GSE4888) to determine their relative expression across the menstrual cycle in vivo (Talbi et al., 2006).

Figure 3 with 3 supplements see all
Decidualizing assembloids mimic midluteal endometrium.

(A) Schematic overview of experimental design. ExM: expansion medium; MDM: minimal differentiation medium. (B) Uniform Manifold Approximation and Projection (UMAP) visualizing epithelial and stromal subsets (EpS and SS, respectively) identified by single-cell transcriptomic analysis of undifferentiated and decidualized assembloids. A transitional population (TP) consisting of cells expressing epithelial and stromal markers is also shown. Dotted lines indicate the separation of EpS and SS in UMAP_1 and of undifferentiated and differentiated subpopulations in UMAP_2. Dotted circles indicate ciliated (EpS3) and TP, which did not fit these broad segregations. (C) Composite heatmaps showing relative expression (Z-scores) of epithelial marker genes across the menstrual cycle in vivo and in undifferentiated and decidualized assembloids. Highlighted in green are genes that mark the midluteal window of implantation (Díaz-Gimeno et al., 2011), whereas genes encoding secreted proteins are indicated by * (Uhlén et al., 2015). See also Figure 3—figure supplement 1. (D) Dot plots showing GO terms related to biological processes enriched in different epithelial populations in decidualizing assembloids. The dot size represents the number of genes in each GO term and the colour indicates FDR-corrected p-value. (E) Composite heatmaps showing relative expression (Z-scores) of stromal marker genes across the menstrual cycle in vivo and in undifferentiated and decidualized assembloids. Highlighted in green are genes that mark the midluteal window of implantation (Díaz-Gimeno et al., 2011), whereas genes encoding secreted proteins are indicated by * (Uhlén et al., 2015). (F) Dot plots showing GO terms related to biological processes enriched in different stromal subpopulations in decidualizing assembloids. See also Figure 3—figure supplements 1 and 2 and 3. Complete epithelial subpopulation marker lists can be found in Figure 3—source data 1. GO analysis outputs can be found in . Complete stromal subpopulation marker lists can be found in .

Figure 3—source data 1

Epithelial subpopulation markers.

Population markers were generated in Seurat v3 using FindMarkers on specified comparisons. Supports Figure 3C.

https://cdn.elifesciences.org/articles/69603/elife-69603-fig3-data1-v2.xlsx
Figure 3—source data 2

GO analysis of differentiated subpopulations.

https://cdn.elifesciences.org/articles/69603/elife-69603-fig3-data2-v2.xlsx
Figure 3—source data 3

Stromal sub-population markers.

Population markers were generated in Seurat v3 using FindMarkers on specified comparisons. Supports Figure 3E.

https://cdn.elifesciences.org/articles/69603/elife-69603-fig3-data3-v2.xlsx

We identified five unambiguous EpC subsets. The glandular component of undifferentiated assembloids harboured actively dividing EpC (EpS1; n = 198) as well as EpC-expressing marker genes of E2-responsive proliferative phase endometrium (EpS2; n = 692), including PGR and CPM (Figure 3C). EpS3 (n = 29) consisted of ciliated EpC, expressing an abundance of genes involved in cilium assembly and organization, including DNAI1 and TUBA4B (Figure 3C). Ciliated cells are the only glandular subpopulation present in both undifferentiated and decidualized assembloids. In vivo, EpS3 marker genes transiently peak during the early-luteal phase (Figure 3C). Decidualization of endometrial assembloids led to the emergence of two distinct EpC subsets, EpS4 (n = 434) and EpS5 (n = 208). Both clusters expressed canonical endometrial ‘receptivity genes’ (annotated in green in Figure 3C), that is, genes used in a clinical test to aid the timing of embryo transfer to the window of implantation in IVF patients (Díaz-Gimeno et al., 2011). In agreement, induction of EpS4 and EpS5 marker genes in vivo coincides with the transition from early- to midluteal phase. However, while expression of EpS4 marker genes, including SOD2, MAOA, and PTGS1, generally peaks during the midluteal window of implantation, EpS5 genes tend to persist or peak during the late-luteal phase (Figure 3C). Additional mining of the data revealed that transition from EpS4 to EpS5 coincides with induction of p16INK4a and p21CIP1 in parallel with upregulation of 56 genes encoding secretory factors (Figure 3—figure supplement 1). Notably, several canonical implantation factors secreted by this subpopulation are also well-characterized SASP components, including dipeptidyl peptidase 4 (DPP4; Kim et al., 2017), growth differentiation factor 15 (GDF15; Basisty et al., 2020), and insulin-like growth factor binding protein 3 (IGFBP3; Elzi et al., 2012). Thus, EpS5 consists of senescent EpC producing an implantation-specific SASP.

Decidualized endometrial assembloids also harboured a sizable population of ambiguous cells expressing both epithelial and stromal genes (Figure 3C and Figure 3—figure supplement 2). A hallmark of this subset, termed ‘transitional population’ (TP; n = 472), is the induction of long non-coding RNAs involved in mesenchymal-epithelial and epithelial-mesenchymal transition (MET/EMT), such as NEAT1 (nuclear paraspeckle assembly transcript 1) and KCNQ1OT1 (KCNQ1 opposite strand/antisense transcript 1) (Bian et al., 2019; Chen et al., 2021). GO analysis showed that both EpS5 and the transitional population comprised secretory cells involved in ECM organization (Figure 3D). However, while EpS5 genes are implicated in neutrophil activation (a hallmark of premenstrual endometrium), genes expressed by the transitional population are uniquely enriched in GO terms such as ‘wound healing’, ‘regulation of stem cell proliferation’, ‘blood coagulation’, and ‘blood vessel development’ (Figure 3D), which points towards a putative role in tissue repair and regeneration.

The stromal fraction of undifferentiated assembloids consisted of actively dividing EnSC (stromal subpopulation 1 [SS1]; n = 434) and E2-responsive EnSC (SS2; n = 874) expressing proliferative phase marker genes, such as PGR, MMP11, and CRABP2 (Figure 3E). As anticipated, decidualization of assembloids for 4 days led to a preponderance of pre-decidual cells (SS3; n = 495) as well as emerging decidual cells (SS4; n = 87) and senescent decidual cells (SS5; n = 118) (Figure 3E). Each of these subpopulations expressed marker genes identified previously by scRNA-seq reconstruction of the decidual pathway in standard primary EnSC cultures (Lucas et al., 2020). Pre-decidual cells in SS3 express HAND2, a key decidual transcription factor (Marinić et al., 2021), as well as previously identified genes encoding secreted factors, including VEGFA (vascular endothelial growth factor A), CRISPLD2 (a progesterone-dependent anti-inflammatory response gene coding cysteine-rich secretory protein LCCL domain containing 2), IL15 (interleukin 15), and TIMP3 (TIMP metallopeptidase inhibitor 3) (Lucas et al., 2020). Novel candidate pre-decidual genes were also identified, such as DDIT4 (DNA damage-inducible transcript 4), encoding a stress response protein intimately involved in autophagy, stemness, and antioxidative defences (Ho et al., 2020; Miller et al., 2020). Decidual cells (SS4) and senescent decidual cells (SS5) express SCARA5 and DIO2, respectively (Figure 3E), two stroma-specific marker genes identified by scRNA-seq analysis of mid- and late-luteal endometrial biopsies (Lucas et al., 2020). SS3 and SS4 genes mapped to the early- and midluteal phase of the cycle, whereas SS5 genes peak in the late-luteal phase, that is, prior to menstrual breakdown. Notably, the transcriptomic profiles of SS3 and SS5 are enriched in GO terms such as ‘Wound healing’, ‘Response to hypoxia’, and ‘Inflammatory response’, suggesting that both clusters comprise stressed cells (Figure 3F). However, the nature of the cellular stress response differs between these populations with only senescent decidual cells (SS5) expressing genes enriched in categories such as ‘Embryo implantation’, ‘Cellular senescence’, ‘Aging’, and ‘Leukocyte activation’. By contrast, few notable categories were selectively enriched in decidual cells (e.g. ‘Mesenchymal cell differentiation’), rendering the lack of GO terms that pertain to stress, inflammation, or wound healing perhaps the most striking observation. In keeping with the GO analysis, senescent decidual cells (SS5) express a multitude of SASP-related genes (Figure 3—figure supplement 3), including matrix metallopeptidases (e.g. MMP3, 7, 9, 10, 11, and 14), insulin-like growth factor binding proteins (e.g. IGFBP1, 3, 6, and 7), growth factors (e.g. AREG, FGF2, FGF7, HGF, and VEGFA) and growth factor receptors (PDGFRA and PDGFRB), cytokines (e.g. LIF, IL6, IL1A, and IL11), chemokines (e.g. CXCL8 and CXCL1), and members of the TGF-β superfamily of proteins (e.g. GDF15, INHBA, and BMP2). By contrast, decidual cells are characterized by expression of a unique network of secretory genes, some encoding ECM proteins (e.g. COL1A1, COL3A1, and LAMA4) and other known decidual markers (e.g. PRL, PROK1, and WNT4) as well as factors involved in uNK cell chemotaxis and activation (e.g. CCL2, CXCL14, and IL15) (Figure 3—figure supplement 3).

Taken together, single-cell analysis of undifferentiated and decidualized assembloids revealed a surprising level of cellular complexity. Each epithelial and stromal subpopulation appears functionally distinct and maps to a specific phase of the menstrual cycle. Transition between cellular states is predicated on changes in cell cycle status, ranging from actively dividing cells in proliferating assembloids to the emergence upon differentiation of highly secretory senescent epithelial and decidual subpopulations, resembling premenstrual endometrium. However, the dominant subpopulations on day 4 of decidualization are EpS4 and SS3, which map to the midluteal implantation window in vivo.

Receptor-ligand interactions in decidualizing assembloids

We used CellPhoneDB, a publicly available online repository of highly curated receptor-ligand interactions, to explore putative interactions between subpopulations in decidualizing assembloids. This computational tool also takes into account the subunit architecture of both ligands and receptors in heteromeric complexes (Efremova et al., 2020; Vento-Tormo et al., 2018). The number of predicted interactions is depicted in Figure 4A, showing a conspicuous lack of crosstalk between the transitional population and any other populations. Conversely, the most abundant interactions centre around the secretory subpopulations, EpS5 and SS5.

Putative receptor-ligand interactions in decidualizing assembloids.

(A) Heatmap showing the total number of cell-cell interactions predicted by CellPhoneDB between different subpopulations in decidualizing assembloids. (B) Dot plots of representative ligand-receptor interactions between stromal subsets (SS) and epithelial subsets (EpS) (upper panel) and EpS and SS (lower panel) in decidualizing assembloids. Circle size and colour indicate p-value and the means of the average expression value of the interacting molecules, respectively. Shaded boxes were used to group putative interactions by level of selectivity. (C) Dot plot of representative ligand-receptor and receptor-ligand interactions between stromal subpopulations in decidualizing assembloids. Direct and indirect tyrosine kinase interactions are indicated by red and blue labels, respectively. Complete tables of predicted ligand-receptor interactions can be found in Figure 4—source data 1.

Figure 4—source data 1

CellPhoneDB prediction of cell-cell interactions.

Ligand-receptor interactions were predicted using CellPhoneDB on counts data from Seurat v3.

https://cdn.elifesciences.org/articles/69603/elife-69603-fig4-data1-v2.xlsx

A total of 270 significantly enriched (non-integrin) receptor-ligand interactions (FDR-corrected p<0.05) were identified between epithelial and stromal subsets in decidualizing assembloids (Figure 4—source data 1), a representative selection of which are shown in Figure 4B. Within the multitude of predicted complex interactions, three broad categories can be discerned. First, there are non-selective interactions involving ligands produced by all subpopulations in one compartment acting on receptors expressed by all subsets in the other compartment. Second, there are semi-selective stromal-epithelial interactions involving three or four subpopulations across both compartments. For example, binding of WNT5A secreted by all decidual stromal subsets to FZD3 (frizzled class receptor 3) expressed on all EpCs represents a non-selective receptor-ligand interaction, whereas binding of WNT5A or WNT4 to FZD6 is a predicted semi-selective interaction, involving all stromal subsets (SS3-5) and EpS4 but not EpS5 (Figure 4B). While FZD3 activates the canonical β-catenin pathway, FZD6 functions as a negative regulator of this signalling cascade (Corda and Sala, 2017). Finally, we identified only three highly selective receptor-ligand interactions (Figure 4B), two of which involved secretion of decidual ligands, prolactin (PRL) and C-X-C motif chemokine ligand 12 (CXCL12), acting on their cognate receptors expressed on receptive EpC (EpS4). CXCL12-dependent activation of C-X-C motif chemokine receptor 4 (CXCR4) has been shown to promote motility of EpC (Zheng et al., 2020), whereas PRL is a lactogenic hormone that stimulates glandular secretion in early pregnancy (Burton et al., 2020).

In contrast to stromal-epithelial communication, non-selective interactions are predicted to be rare between decidual subsets. Instead, communication appears governed largely by a combinatorial network of receptor-ligand interactions (Figure 4C). For example, colony stimulating factor 3 (CSF3) and vascular endothelial growth factor A (VEGFA) produced by senescent decidual cells (SS5) are predicted to impact selectively on pre-decidual cells (SS3), whereas secretion of inhibin A (INHBA) may engage both pre-decidual and decidual cells (SS4). Other interactions are predicted to govern crosstalk between SS3 and SS4, such as modulation of the WNT pathway in response to binding of R-spondin 3 (RSPO3) to leucine-rich repeat-containing G protein-coupled receptor 4 (LGR4). A striking observation is the overrepresentation of receptor tyrosine kinases implicated in SS3 and SS5 signal transduction as well as the involvement of receptors that signal through downstream cytoplasmic tyrosine kinases, including CSF3 receptor (CSF3R) and CD44 (Figure 4C; Corey et al., 1998; van der Voort et al., 1999).

Tyrosine kinase-dependent stress responses determine the fate of decidual cells

The CellPhoneDB analysis inferred that epithelial-stromal crosstalk in assembloids is robust, buffered by numerous non-selective interactions, whereas decidual subsets are reliant on selective receptor-ligand interactions and activation of distinct signal transduction pathways. For example, the predicted tyrosine kinase dependency of pre-decidual (SS3) and senescent decidual cells (SS5) raised the possibility that these subpopulations can be targeted by tyrosine kinase inhibitors, such as dasatinib (Brighton et al., 2017; Zhu et al., 2015), a second-generation, broad-spectrum ATP-competitive protein tyrosine kinase inhibitor (Aguilera and Tsimberidou, 2009; Li et al., 2010). To test this supposition, we generated single-cell transcriptomic profiles of assembloids decidualized for 4 days in the presence of dasatinib (Figure 5A). We found that decidualization in the presence of dasatinib had a dramatic impact on stromal subpopulations, virtually eliminating senescent decidual cells (SS5, n = 7) and increasing the abundance of decidual cells ninefold (SS4, n = 882; Figure 5B). Apart from a modest reduction in pre-decidual cells (SS3), dasatinib also impacted markedly on transitional cells, reducing their numbers by 76%. By contrast, the effect on epithelial populations was confined to a modest reduction in senescent EpC (EpS5) (Figure 5B). Further, relatively few genes were perturbed significantly (FDR-corrected p<0.05) upon dasatinib treatment in epithelial populations (Figure 5C). In the stroma, dasatinib triggered a conspicuous transcriptional response in pre-decidual (SS3) and transitional cells, whereas gene expression in decidual cells (SS4) and the few remaining senescent decidual cells (SS5) was largely unaffected (Figure 5C). In transitional cells, dasatinib simultaneously upregulated genes encoding canonical mesenchymal markers (e.g. SNAI2, TWIST2, ZEB1, COL1A1, and FBN1; Owusu-Akyaw et al., 2019) and decidual factors (e.g. SCARA5, FOXO1, GADD45A, IL15, CXCL14, and SGK1; Gellersen and Brosens, 2014), suggesting that MET accounts for the emergence of this population upon decidualization (Figure 5—source data 1). In pre-decidual cells, dasatinib inhibited the expression of a network of genes enriched in GO categories such as ‘Response to wounding’ (FDR-corrected p=3.5 × 10–5), ‘Response to stress’ (FDR-corrected p=3.8 × 10–5), and ‘Response to oxidative stress’ (FDR-corrected p=1.3 × 10–4), indicative of a blunted stress response. To substantiate this finding, we measured the secreted levels of CXCL8 (IL-8), a potent inflammatory mediator implicated in autocrine/paracrine propagation of cellular senescence (Acosta et al., 2008; Kuilman et al., 2008), in assembloids decidualized with or without dasatinib. CXCL14, IL-15, and TIMP3 levels were also measured to monitor the decidual response. As shown in Figure 5D, dasatinib completely abrogated the release of CXCL8 by pre-decidual cells while markedly enhancing subsequent secretion of CXCL14, IL-15, and TIMP3, which are involved in effecting immune clearance of senescent decidual cells (Brighton et al., 2017; Lucas et al., 2020; Kong et al., 2021). Together, these observations not only support the CellPhoneDB predictions but also indicate that the amplitude of the cellular stress response during the pre-decidual phase determines the subsequent decidual trajectory, with low levels accelerating differentiation and high levels promoting cellular senescence and MET.

Tyrosine kinase-dependent stress responses determine the fate of decidual cells.

(A) Schematic overview of experimental design. ExM: expansion medium; MDM: minimal differentiation medium. (B) Uniform Manifold Approximation and Projection (UMAP) visualization (left panel) and relative proportions (right panel) of subpopulations in endometrial assembloid decidualized in the presence or absence of dasatinib. (C) Number of differentially expressed genes (DEGs) in each subpopulation in response to dasatinib pre-treatment. (D) Secreted levels of CXCL8 and decidual cell factors in spent medium from assembloids treated with or without dasatinib. Secreted levels in individual assembloids established from four different endometrial assembloids decidualized with or without dasatinib are shown by dotted and solid lines, respectively. Full lists of DEGs and associated GO analysis can be found in Figure 5—source data 1 and Figure 5—source data 2, respectively. Data used in (D) are available in Figure 5—source data 3.

Figure 5—source data 1

Differentially expressed genes for day 4 populations treated with and without dasatinib.

Specified pairwise comparisons were generated in Seurat v3 using FindMarkers. Supports Figure 5C.

https://cdn.elifesciences.org/articles/69603/elife-69603-fig5-data1-v2.xlsx
Figure 5—source data 2

GO analysis for day 4 populations treated with and without dasatinib.

https://cdn.elifesciences.org/articles/69603/elife-69603-fig5-data2-v2.xlsx
Figure 5—source data 3

ELISA data.

Secreted levels of key senescent (CXCL8) and decidual stromal cell markers (CXCL14, IL-15, TIMP3) (pg/ml) were examined by ELISA in spent medium from assembloids treated with or without dasatinib. Supports Figure 5D.

https://cdn.elifesciences.org/articles/69603/elife-69603-fig5-data3-v2.xlsx

Modelling the impact of decidual subpopulations on human embryos

We postulated that decidual invasion by human embryos that have breached the luminal endometrial epithelium depends on an acute cellular senescence and transient SASP production, rich in growth factors and proteases. Conversely, we reasoned that lack of senescent decidual cells or unconstrained SASP should simulate pathological implantation environments associated with implantation failure and early pregnancy loss, respectively. To test this hypothesis, we constructed a simple implantation model by embedding human embryos in endometrial assembloids. To this end, assembloids were first decidualized for 96 hr in the presence or absence of dasatinib, washed and cultured in embryo medium, consisting of MDM with added supplements (Figure 6A and Supplementary file 1: Table 1). Day 5 human blastocysts were placed into small pockets created in the decidualized assembloids (Figure 6B), one embryo per assembloid, and individual co-cultures imaged using time-lapse microscopy over 72 hr. Co-cultured blastocysts (n = 5) expanded markedly when placed in decidualized assembloids that were not pre-treated with dasatinib (Figure 6C and D). Time-lapse microscopy revealed intense cellular movement in the stromal compartment as well as evidence that interaction between migratory decidual cells and polar trophectoderm promotes adherence and early invasion of the embryo (SI Video 1 and Figure 6—figure supplement 1). Retrieval and processing of one attached embryo demonstrated proliferating polar trophectoderm and expression of OCT4 and GATA6 in the epiblast and hypoblast, respectively (Figure 6E). A major limitation of this implantation model is that persistence of senescent decidual cells also causes gradual disintegration of the assembloids (Figure 6—figure supplement 2). By contrast, pre-treatment with dasatinib, which accelerates decidualization and all but eliminates decidual senescence, resulted in much more robust assembloids. However, all embedded blastocysts (n = 5) failed to expand in this model (Figure 6C and D). Further, movement of the decidual matrix was greatly reduced and directed migration or attachment of decidual cells to the blastocyst was not observed (SI Video 2). Secreted levels of human chorionic gonadotropin (hCG) did not differ between co-cultures (Figure 6E), suggesting that all embryos remained viable over the 72 hr observation period. Thus, while our experimental design precluded modelling of physiological embryo implantation, aspects of different pathological endometrial states underlying reproductive failure, that is, implantation failure and miscarriage, were recapitulated in assembloids.

Figure 6 with 2 supplements see all
Impact of decidual senescence in assembloids on co-cultured human blastocysts.

(A) Diagram showing experimental design. ExM: expansion medium; MDM: minimal differentiation medium; EM: embryo medium. (B) Schematic drawing of co-culture method. (C) Representative time-lapse images of blastocysts embedded in assembloids following decidualization for 96 hr in the absence (upper panels) or presence (lower panels) of dasatinib. Scale bar = 100 µm. See also Figure 6—figure supplement 1. (D) Embryo diameters (µm) measured over 72 hr when embedded in decidualizing assembloids pre-treated with or without dasatinib. (E) OCT4 and GATA6 immunofluorescence marking the epiblast and hypoblast, respectively, in a blastocyst attached by proliferating polar trophectoderm (arrowhead) to decidual assembloids. Scale bar = 50 µM. (F) Secreted levels of human chorionic gonadotropin (hCG) in blastocyst-endometrial assembloid co-cultures. Individual embryo diameter measurements for biological replicates in (D) are available in Figure 6—source data 1. Individual ELISA data used in (F) are available in Figure 6—source data 2.

Figure 6—source data 1

Embryo expansion measurement.

Embryo diameters were measured over 72 hr co-culture with assembloids previously differentiated with or without dasatinib. Images were captured using time-lapse imaging. Supports Figure 6D.

https://cdn.elifesciences.org/articles/69603/elife-69603-fig6-data1-v2.xlsx
Figure 6—source data 2

Embryo human chorionic gonadotropin (hCG) secretion.

Secreted levels of hCG (pg/ml) by embryos in co-culture with assembloids previously differentiated with or without dasatinib were examined by ELISA in spent medium. Supports Figure 6F.

https://cdn.elifesciences.org/articles/69603/elife-69603-fig6-data2-v2.xlsx
Video 1
Time-lapse microscopy of a human blastocyst embedded in a decidualizing assembloid.

Representative video of a human blastocyst embedded in an assembloid, as imaged by time-lapse microscopy over 72 hr with images captured every 60 min.

Video 2
Time-lapse microscopy of a human blastocyst embedded in a decidualizing assembloid pre-treated with dasatinib.

Representative video of a human blastocyst embedded in an assembloid which had been pre-treated with dasatinib, as imaged by time-lapse microscopy over 72 hr with images captured every 60 min.

Discussion

Here we report on the development of endometrial assembloids, consisting of gland-like organoids surrounded by a matrix rich in primary EnSC, as novel model to parse the cellular dynamics that govern embryo implantation in cycling human endometrium. While assembloids complement and advance other recently described endometrial organoid models (Boretto et al., 2017; Cheung et al., 2021; Fitzgerald et al., 2019; Luddi et al., 2020; Turco et al., 2017), they still lack the cellular complexity of native endometrium, including uNK cells, macrophages, and vascular cells. Nevertheless, we demonstrated that aspects of pathological implantation events can be recapitulated in assembloids, rendering them useful as novel models to study mechanisms of reproductive failure and evaluate potential therapeutic interventions.

Single-cell analysis of differentiating endometrial assembloids indicates that the sequence of events leading up to the implantation window, and beyond, requires divergence of both glandular EpC and EnSC into differentiated and senescent subpopulations, a process likely determined by the level of replication stress incurred by individual cells in the preceding proliferative phase (Brighton et al., 2017). Importantly, we demonstrate that acute senescence in glandular EpC (EpS5) underpins production of an implantation-specific SASP, comprising canonical implantation factors and growth factors, such as amphiregulin (AREG) and epiregulin (EREG), implicated in transforming cytotrophoblasts into extravillous trophoblasts (Cui et al., 2020; Yu et al., 2019). On the other hand, the transcriptome profile of differentiated EpC (EpS4) revealed a pivotal role for this subpopulation in prostaglandin and glycodelin synthesis. Prostaglandins, and specifically PGE2, are indispensable for implantation (Ruan et al., 2012), whereas glycodelin is an abundantly secreted, multifaceted glycoprotein involved in blastocyst attachment, trophoblast differentiation, and immune modulation in early pregnancy (Lee et al., 2016). Further, differentiated EpC highly express SLC2A1, encoding the major glucose transporter GLUT1. Glucose is required for glycogen synthesis, an essential component of glandular secretions that nourishes the conceptus prior to the onset of placental perfusion around 10 weeks of pregnancy (Burton et al., 2020). The fate and function of senescent EpC in pregnancy are unknown. Arguably, localized secretion of proteinases by senescent EpC may promote breakdown of the surrounding basement membrane, thereby facilitating endoglandular trophoblast invasion and access to histotrophic nutrition in early gestation (Huppertz, 2019; Moser et al., 2010). In non-conception cycles, the abundance of p16INK4-positive glandular EpC rises markedly during the late-luteal phase (Brighton et al., 2017), indicating that senescent EpC are progesterone-independent and likely responsible for glandular breakdown in the superficial endometrial layer at menstruation.

Decidual transformation of EnSC in assembloids unfolded largely as anticipated from previous studies, that is, starting with an acute pre-decidual stress response and leading to the emergence of both decidual and senescent decidual subpopulations (Brighton et al., 2017; Lucas et al., 2020; Kong et al., 2021). Like their epithelial counterparts, senescent decidual cells have a conspicuous secretory phenotype. We identified 56 and 72 genes encoding secreted factors upregulated in senescent epithelial and decidual subpopulations, respectively. However, only 15 genes were shared, indicating that the SASP generated in both cellular compartments is distinct. As glandular secretions drain into the uterine cavity, the embryonic microenvironment is therefore predicted to change abruptly upon breaching of the luminal epithelium. Recent comparative metabolomics of apical and basolateral endometrial gland-like organoid secretomes also supports the prediction of an asymmetrical profile of glandular secretions in the pre- and post-implantation microenvironments (Simintiras et al., 2021).

Based on computational predictions of ligand-receptor interactions, we demonstrated that the decidual response in assembloids can be targeted pharmacologically with only modest impact on glandular function and, by extension, the preimplantation embryo milieu. Specifically, dasatinib, a tyrosine kinase inhibitor, was highly effective in blunting the pre-decidual stress response, leading to a dramatic expansion of anti-inflammatory decidual cells and near-total elimination of senescent decidual cells. Dasatinib also inhibited the emergence of TP and shifted the transcriptional profile of the remaining transitional cells towards a decidual phenotype. An analogous population of ambiguous cells expressing both epithelial and mesenchymal marker genes was recently identified in midluteal endometrium by scRNA-seq analysis (Lucas et al., 2020). Further, based on CellPhoneDB and GO analyses, transitional cells are predicted to be highly autonomous and involved in tissue regeneration, in line with experimental evidence that MET drives re-epithelization of the endometrium following menstruation and parturition (Owusu-Akyaw et al., 2019; Patterson et al., 2013). Thus, the level of endogenous cellular stress generated by the endometrium during the window of implantation calibrates the subsequent decidual trajectory, either promoting the formation of a robust decidual matrix or facilitating tissue breakdown and repair. Further, an in-built feature of both trajectories is self-enforcement as decidual cells recruit and activate uNK cells to eliminate their senescent counterparts (Brighton et al., 2017; Lucas et al., 2020; Kong et al., 2021), whereas senescent decidual cells induce secondary senescence in neighbouring decidual (Brighton et al., 2017; Ozaki et al., 2017) and, plausibly, uNK cells (Rajagopalan and Long, 2012).

Clinically, recurrent pregnancy loss, defined as multiple miscarriages, is associated with loss of endometrial clonogenicity (Lucas et al., 2016; Diniz-da-Costa et al., 2021), uNK cell deficiency and excessive decidual senescence (Lucas et al., 2020; Tewary et al., 2020), and rapid conceptions (also referred to as ‘superfertility’) (Dimitriadis et al., 2020; Ticconi et al., 2020). Conversely, lack of a proliferative gene signature in midluteal endometrium and premature expression of decidual PRL have been linked to recurrent implantation failure (Berkhout et al., 2020b; Koler et al., 2009; Koot et al., 2016), a pathological condition defined by a failure to achieve a pregnancy following transfer of one or more high-quality embryos in multiple IVF cycles (Polanski et al., 2014). We reasoned that these aberrant implantation environments can be recapitulated in assembloids by manipulating the level of decidual senescence. In line with these predictions, the presence of senescent decidual cells created a permissive environment in which migratory decidual cells interacted with expanding blastocysts, although continuous SASP production also promoted breakdown of the assembloids. Conversely, in the absence of senescent decidual cells, non-expanding embryos became entrapped in a robust but stagnant decidual matrix. These observations are in keeping with previous studies demonstrating that implantation of human embryos depends critically on the invasive and migratory capacities of decidual cells (Berkhout et al., 2020a; Gellersen et al., 2010; Grewal et al., 2008; Weimar et al., 2012). Our co-culture experiments also highlighted the shortcomings of assembloids as an implantation model, including the lack of a surface epithelium to create distinct pre- and post-implantation microenvironments and the absence of key cellular constituents, such as innate immune cells.

In summary, parsing the mechanisms that control implantation has been hampered by the overwhelming complexity of factors involved in endometrial receptivity. Our single-cell analysis of decidualizing assembloids suggests that this complexity reflects the reliance of the human endometrium on rapid E2-dependent proliferation and replicative exhaustion to generate both differentiated and senescent epithelial and stromal subpopulations in response to the postovulatory rise in progesterone. We demonstrate that senescent cells in both cellular compartments produce distinct bioactive secretomes, which plausibly prime pre-implantation embryos for interaction with the luminal epithelium and then stimulate encapsulation by underlying decidual stromal cells. Based on our co-culture observations, we predict that a blunted pre-decidual stress response causes implantation failure because of a lack of senescence-induced tissue remodelling and accelerated decidualization. Conversely, a heightened stress response leading to excessive decidual senescence may render embryo implantation effortless, albeit in a decidual matrix destined for breakdown and repair. Finally, we demonstrated that pre-decidual stress responses can be modulated pharmacologically, highlighting the potential of endometrial assembloids as a versatile system to evaluate new or repurposed drugs aimed at preventing reproductive failure.

Materials and methods

Ethical approvals, endometrial samples, and human blastocysts

Request a detailed protocol

Endometrial biopsies were obtained from women attending the Implantation Research Clinic, University Hospitals Coventry and Warwickshire National Health Service Trust. Written informed consent was obtained in accordance with the Declaration of Helsinki 2000. The study was approved by the NHS National Research Ethics Committee of Hammersmith and Queen Charlotte’s Hospital NHS Trust (1997/5065) and Tommy’s Reproductive Health Biobank (Project TSR19-002E, REC Reference: 18/WA/0356). Timed endometrial biopsies were obtained 6–11 days after the post-ovulatory LH surge using a Wallach Endocell Endometrial Cell Sampler. Patient demographics for the samples used in each experiment are detailed in Supplementary file 1: Table 2.

The use of vitrified human blastocysts was carried out under a Human Fertilisation and Embryology Authority research licence (HFEA: R0155) with local National Health Service Research Ethics Committee approval (04/Q2802/26). Spare blastocysts were donated to research following informed consent by couples who had completed their fertility treatment at the Centre for Reproductive Medicine, University Hospitals Coventry and Warwickshire National Health Service Trust. Briefly, women underwent ovarian stimulation and oocytes were collected by transvaginal ultrasound-guided aspiration and inseminated with prepared sperm (day 0). All oocytes examined 16–18 hr after insemination and classified as normally fertilized were incubated under oil in 20–25 µl drops of culture media (ORIGIO Sequential Cleav and Blast media, CooperSurgical, Denmark) at 5% O2, 6% CO2, 89% N2 at 37°C. Following culture to day 5 of development, the embryo(s) with the highest quality was selected for transfer, whereas surplus embryos considered top-quality blastocysts were cryopreserved on day 5 or 6 by vitrification using Kitazato vitrification media (Dibimed, Spain) and stored in liquid nitrogen. Prior to their use in the co-culture, vitrified blastocysts were warmed using the Kitazato vitrification warming media (Dibimed, Spain) and underwent zona pellucida removal using a Saturn 5 Laser (CooperSurgical). Blastocysts were then incubated for 1 hr under oil in 20 µl drops of culture media (ORIGIO Sequential Blast media, CooperSurgical) at 5% O2, 6% CO2, 89% N2 at 37°C and allowed to re-expand.

Processing of endometrial biopsies and primary EnSC cultures

Request a detailed protocol

Unless otherwise stated, reagents were obtained from Life Technologies (Paisley, UK). Cell cultures were incubated at 37°C, 5% CO2 in a humidified incubator. Centrifugation and incubation steps were performed at room temperature unless stated otherwise. Fresh endometrial biopsies were processed as described previously (Barros et al., 2016). Briefly, tissue was finely minced for 5 min using a scalpel blade. Minced tissue was then digested enzymatically with 0.5 mg/ml collagenase I (Sigma-Aldrich, Gillingham, UK) and 0.1 mg/ml deoxyribonuclease (DNase) type I (Lorne Laboratories, Reading, UK) in 5 ml phenol red-free Dulbecco’s Modified Eagle Medium (DMEM)/F12 for 1 hr at 37°C, with regular vigorous shaking. Dissociated cells were washed with growth medium (DMEM/F12 containing 10% dextran-coated charcoal stripped FBS [DCC-FBS], 1% penicillin-streptomycin, 2 mM L-glutamine, 1 nM E2 [Sigma-Aldrich] and 2 mg/ml insulin [Sigma-Aldrich]). Samples were passed through a 40 µm cell sieve. EnSCs were collected from the flowthrough, while epithelial clumps remained in the sieve and were collected by backwashing into a 50 ml Falcon tube. Samples were resuspended in growth medium and centrifuged at 400× g for 5 min. EnSC pellets were resuspended in 10 ml growth medium and plated in tissue culture flasks. To isolate EnSC from other (non-adherent) cells collected in the flowthrough, medium was refreshed after 24 hr. Thereafter, medium was refreshed every 48 hr. Sub-confluent monolayers were passaged using 0.25% Trypsin-EDTA and split at a ratio of 1:3.

Endometrial gland-like organoid culture

Request a detailed protocol

Endometrial gland-like organoids were established as described previously (Turco et al., 2017), with adaptations. Freshly isolated endometrial gland fragments were resuspended in 500 µl phenol red-free DMEM/F12 medium in a microcentrifuge tube and centrifuged at 600× g for 5 min. The medium was aspirated and ice-cold, growth factor-reduced Matrigel (Corning Life Sciences B.V., Amsterdam, Netherlands) was added at a ratio of 1:20 (cell pellet: Matrigel). Samples mixed in Matrigel were kept on ice until plating at which point the suspension was aliquoted in 20 µl volumes to a 48-well plate, one drop per well, and allowed to cure for 15 min. Expansion medium supplemented with E2 (Supplementary file 1: Table 1; Turco et al., 2017) was then added and samples cultured for up to 7 days. For passaging, Matrigel droplets containing gland-like organoids were collected into microcentrifuge tubes and centrifuged at 600× g for 6 min at 4°C. Samples were resuspended in ice-cold, phenol red-free DMEM/F12 and subjected to manual pipetting to disrupt the organoids. Suspensions were centrifuged again, resuspended in ice-cold additive-free DMEM/F12, and then subjected to further manual pipetting. Suspensions were centrifuged again and either resuspended in Matrigel and plated as described above for continued expansion or used to establish assembloid cultures.

Establishment of assembloid cultures

Request a detailed protocol

At passage 2, EnSC and gland-like organoid pellets were mixed at a ratio of 1:1 (v/v) and ice-cold PureCol EZ Gel (Sigma-Aldrich) added at a ratio of 1:20 (cell pellet: hydrogel). Samples were kept on ice until plating. The suspension was aliquoted in 20 µl volumes using ice-cold pipette tips into a 48-well plate, one droplet per well, and allowed to cure in the cell culture incubator for 45 min. Expansion medium supplemented with 10 nM E2 was overlaid and the medium was refreshed every 48 hr. For decidualization experiments, assembloid cultures were grown in expansion medium supplemented with E2 for 4 days to allow for growth and expansion. Assembloids were then either harvested or decidualized using different media as tabulated in Supplementary file 1: Table 1 for a further 4 days. Again, the medium was refreshed every 48 hr and spent medium stored for further analysis. For tyrosine kinase inhibition, MDM was supplemented with 250 nM dasatinib (Cell Signaling Technology, Leiden, NL).

Fluorescence microscopy

Request a detailed protocol

For fluorescent microscopy, assembloids were removed from culture wells and transferred into tubes for fixation. Samples were washed in PBS and fixed in 10% neutral buffered formalin in the tube for 15 min, then washed three times with PBS, and stored for use. Samples were dehydrated in increasing concentrations of ethanol (70% then 90% for 1 hr each, followed by 100% for 90 min), then incubated in xylene for 1 hr. After paraffin wax embedding, 5 µm sections were cut and mounted, then incubated overnight at 60°C. Slides were then stored at 4°C until further processing. De-paraffinization and rehydration were performed through xylene, 100% isopropanol, 70% isopropanol, and distilled water incubations. Following antigen retrieval, permeabilization was performed where appropriate by incubation with 0.1% Triton X-100 for 30 min. Slides were then washed, blocked, and incubated in primary antibodies overnight at 4°C. Antibody details are presented in Supplementary file 1: Table 3. After washing three times, slides were incubated with secondary antibodies for 2 hr, then washed as before and mounted in ProLong Gold Antifade Reagent with DAPI (Cell Signaling Technology). Slides were visualized using the EVOS Auto system, with imaging parameters maintained throughout image acquisition. Images were merged in ImageJ and any adjustments to brightness or contrast were applied equally within comparisons.

Real-time quantitative polymerase chain reaction

Request a detailed protocol

After removal of spent medium, gland-like organoid cultures were washed in PBS and harvested in 200 µl Cell Recovery Solution (Corning). Gel droplets were transferred to nuclease-free microcentrifuge tubes and placed at 4°C for 30 min. Samples were then washed in PBS, centrifuged at 600× g for 6 min twice, and snap frozen as cell pellets. Assembloid cultures were washed with PBS and then recovered by directly scraping the samples into nuclease-free microcentrifuge tubes. Samples were centrifuged at 600× g for 6 min. The cellular pellet was resuspended in 500 µl of 500 µg/ml collagenase I diluted in additive-free DMEM/F12 and incubated at 37°C for 10 min with regular manual shaking. Samples were washed twice in PBS, with centrifugation at 600× g for 6 min, then cell pellets were snap frozen. RNA extraction was performed using the RNeasy Micro Kit (QIAGEN, Manchester, UK) according to the manufacturer’s instructions. RNA concentration and purity were determined using a NanoDrop ND-1000. All RNA samples were stored at –80°C until use. Reverse transcription was performed using the QuantiTect Reverse Transcription (RT) Kit according to the manufacturer’s protocol (QIAGEN). Input RNA was determined by the sample with lowest concentration within each experiment. Genes of interest were amplified using PrecisionPlus SYBR Green Mastermix (PrimerDesign, Southampton, UK). Amplification was performed in 10 µl reactions containing 5 μl PrecisionPlus 2× master mix, 300 nM each of forward and reverse primer, nuclease-free water, and 1 µl of cDNA or water control. Amplification was performed for 40 cycles on an Applied Biosystems QuantStudio 5 Real-Time PCR System (qPCR). Data were analysed using the Pffafl method (Pfaffl, 2001) and L19 was used as a reference gene. Primer sequences were as follows: L19 forward: 5′-GCG GAA GGG TAC AGC CAA T-3′, L19 reverse: 5′-GCA GCC GGC GCA AA-3′, PAEP forward: 5′-GAG CAT GAT GTG CCA GTA CC-3′, PAEP reverse: 5′-CCT GAA AGC CCT GAT GAA TCC-3′, SPP1 forward: 5′-TGC AGC CTT CTC AGC CAA A-3′, SPP1 reverse: 5′-GGA GGC AAA AGC AAA TCA CTG-3′.

Enzyme-linked immunosorbent assay

Request a detailed protocol

Spent medium was collected every two days during a 4-day decidual time course, with or without dasatinib treatment. Duoset solid-phase sandwich enzyme-linked immunosorbent assay (ELISA) kits (Bio-Techne, Abingdon, UK) were used for the detection of PRL (DY682), TIMP3 (DY973), IL-8 (DY208), IL-15 (DY247), CXCL14 (DY866), and HCG (DY9034). Assays were performed according to the manufacturer’s instructions. Absorbance at 450 nm was measured on a PheraStar microplate reader (BMG LABTECH Ltd, Aylesbury, UK), with background subtraction from absorbance measured at 540 nm. Protein concentration was obtained using a four-parameter logistic regression analysis and interpolation from the curve. As medium was collected at different timepoints in a time-course culture, secreted levels were not normalized to total cell or protein contents.

Single-cell capture, library preparation, and sequencing

Request a detailed protocol

Assembloids were dissociated to single cells by incubation of gel droplets with 0.5 mg/ml collagenase I for 10 min in a 37°C water bath for 10 min with regular vigorous shaking. Samples were washed with additive-free DMEM/F12 phenol-free medium and incubated with 5× TrypLE Select diluted in additive-free DMEM/F12 phenol-free medium for 5 min in a 37°C water bath. Cell clumps were disrupted by manual pipetting, then suspended in 0.1% bovine serum albumin (BSA) in PBS and passed through a 35 µm cell sieve. Droplet generation was performed using a Nadia Instrument (Dolomite Bio, Cambridge, UK) according to the manufacturer’s guidelines and using reagents as described by Macosko et al., 2015 and the scRNAseq v1.8 protocol (Dolomite Bio). Pooled beads were processed as described previously (Lucas et al., 2020) and sequenced using a NextSeq 500 with high-output 75-cycle cartridge (Illumina, Cambridge, UK) by the University of Warwick Genomics Facility.

Bioinformatics analysis

Request a detailed protocol

Initial single-cell RNAseq data processing was performed using Drop-Seq_tools-2.3.0 (DropseqAlignmentCookbook_v2Sept2018, http://mccarrolllab.com/dropseq) and as described previously (Lucas et al., 2020). To select high-quality data for analysis, cells were included when at least 200 genes were detected, while genes were included if they were detected in at least three cells. Cells with more than 5000 genes were excluded from the analysis as were cells with more than 5% mitochondrial gene transcripts to minimize doublets and low quality (broken or damaged) cells, respectively. The Seurat v3 standard workflow (Stuart et al., 2019) was used to integrate datasets from biological replicates. Clustering and nearest-neighbour analysis was performed on the full integrated dataset using principal components 1:15 and a resolution of 0.6. The ‘subset’ function was applied for interrogation of specific experimental conditions and timepoints. Gene Ontology (GO) analysis was performed on differentially expressed genes from specified ‘FindMarkers’ comparisons in Seurat v3 using the Gene Ontology Consortium database (Ashburner, 2000; THE GENE ONTOLOGY, 2019; THE GENE ONTOLOGY, 2019; Mi et al., 2013). Dot plots of significantly enriched GO terms (FDR-adjusted p<0.05) were generated in RStudio (version 1.2.5042). CellPhoneDB was used to predict enriched receptor-ligand interactions between subpopulations in decidualizing assembloids (Efremova et al., 2020; Vento-Tormo et al., 2018). Significance was set at p<0.05. Annotated tyrosine kinase interactions were curated manually.

Co-culture of human blastocyst and endometrial assembloids

Request a detailed protocol

Prior to co-culture, decidualized assembloids were washed in PBS and medium was replaced with embryo medium (Supplementary file 1: Table 1). Assembloids were lightly punctured with a needle to create a small pocket, to enable one re-expanded day 5 human blastocyst to be co-cultured per assembloid. The plate was transferred to a pre-warmed and gassed (humidified 5% CO2 in air) environment chamber placed on an automated X-Y stage (EVOS FL Auto Imaging System with onstage incubator) for time-lapse imaging. Brightfield images were captured every 60 min over 72 hr. Captured images were converted into videos using ImageJ.

For fixation, assembloid co-cultures were removed from culture wells and transferred into tubes. Samples were washed in PBS and fixed in 10% neutral buffered formalin in the tube for 15 min, then washed three times with PBS. Assembloids were permeabilized for 30 min in PBS containing 0.3% Triton X-100 and 0.1 M glycine for 30 min at room temperature. Samples were incubated overnight at 4°C in primary antibodies diluted in PBS containing 10% FBS, 2% BSA, and 0.1% Tween-20. Samples were then washed in PBS (0.1% Tween-20) and incubated for 2 hr at room temperature protected from light in fluorescently conjugated Alexa Fluor secondary antibodies 1:500 (ThermoFisher Scientific) and DAPI (D3571, ThermoFisher Scientific, dilution 1/500), diluted in PBS containing 10% FBS, 2% BSA, and 0.1% Tween-20. Samples were imaged on a Leica SP8 confocal microscope using a ×25 water objective, with a 0.6 µm z-step and 2× line averaging.

Statistical analysis

Request a detailed protocol

Data were analysed using GraphPad Prism. Pairwise comparison of non-parametric data was performed using Mann–Whitney U test. For paired, non-parametric significance testing between multiple groups, the Friedman test, and Dunn’s multiple comparisons post hoc test were performed. Only values of p<0.05 were considered statistically significant.

Data availability

Single cell RNAseq data presented in this paper are openly available as a Gene Expression Omnibus DataSet (https://www.ncbi.nlm.nih.gov/gds) under accession number GSE168405. Other source data are presented in the Source Data tables as indicated in the corresponding Figure legends.

The following data sets were generated
    1. Brosens JJ
    2. Lucas ES
    3. Rawlings TM
    (2021) NCBI Gene Expression Omnibus
    ID GSE168405. Single-cell RNA Sequencing of Endometrial Assembloid Cultures.
The following previously published data sets were used

References

  1. Book
    1. Aplin JD
    2. Jones CJP
    (1989) Extracellular matrix in endometrium and decidua
    In: Klopper A, Beaconsfield R, editors. Placenta as a Model and a Source. Springer. pp. 23–25.
    https://doi.org/10.1007/978-1-4613-0823-2
  2. Software
    1. Bagley B
    (2019) Advanced Biomatrix
    Collagen Gelation Kinetics and Shear Modulus.
  3. Software
    1. Rawlings TM
    (2021)
    Organoids to model the endometrium: Implantation and beyond
    Reproduction and Fertility.

Decision letter

  1. Thomas E Spencer
    Reviewing Editor; Fred Hutchinson Cancer Research Center, United States
  2. Jonathan A Cooper
    Senior Editor; Fred Hutchinson Cancer Research Center, United States
  3. Thomas E Spencer
    Reviewer
  4. Gunter Wagner
    Reviewer; Yale University, United States
  5. Hugo Vankelecom
    Reviewer
  6. Ji-Yong Julie Kim
    Reviewer

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Acceptance summary:

This paper elegantly combines single cell transcriptomic and novel three-dimensional culture models of the human endometrium to reveal how cell senescence impacts the ability of the uterus to prepare and support embryo implantation. This work provides important insights into the etiology of pregnancy failure in women and may accelerate the discovery of new treatments to improve reproductive health.

Decision letter after peer review:

Thank you for submitting your article "Modelling the impact of decidual senescence on embryo implantation in human endometrial assembloids" for consideration by eLife. Your article has been reviewed by 4 peer reviewers, including Thomas E Spencer as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Jonathan Cooper as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Gunter Wagner (Reviewer #2); Hugo Vankelecom (Reviewer #3).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) The story is not always easy to follow if not completely submerged in this field of endometrial receptivity, decidualization, importance of senescence, of NK cells, etc. Authors should at certain points be more focused and should less diverge from the subject (certainly in the Discussion) which would better retain the attention of the reader to the central message of the study. Also, the Abstract could be simplified in this regard. Moreover, the paper will highly benefit from a graphical abstract of summarizing figure, since messages conveyed are not easy for the broader reader.

2) The epithelial organoid cultures are referred to as glandular organoids throughout the manuscript; however, the cultures may contain a mixed population of glandular and luminal epithelial cells. It would be beneficial to clarify the proportion of glandular and luminal epithelial cells within the culture. This can be done using established markers of endometrial glandular epithelial cells such as FOXA2.

3) The authors should consider testing their minimal differentiation medium to determine whether it's sufficient to induce a stromal cell response, rather than only testing the secretory response of the endometrial epithelium. If these tests have been done, the authors should consider mentioning them at the place in the text where the epithelial secretion response is discussed.

Reviewer #1 (Recommendations for the authors):

1. As the manuscript introduces a new co-culture model of endometrial stroma and epithelial cells, the authors must describe the relative proportion of cells used to establish co-culture. Specifically, a 1:1 (v/v) ratio of different cell types is ill-defined and will make replication difficult as pellet size can depend on the amount of organoid disruption. Please clarify to enhance reproducibility.

2. The epithelial organoid cultures are referred to as glandular organoids throughout the manuscript; however, the cultures may contain a mixed population of glandular and luminal epithelial cells. It would be beneficial to clarify the proportion of glandular and luminal epithelial cells within the culture. This can be done using established markers of endometrial glandular epithelial cells such as FOXA2.

3. Although the authors have introduced a new model that likely will advance our understanding of implantation and pregnancy loss in the future, it is unclear if it provides an advancement in the study of uterine stromal cell decidualization. An unaddressed question from this study is the impact of epithelial cells on stromal cell decidualization. In the future, it will be necessary to directly compare this model with 3D stromal cell culture in the absence of epithelial cells.

4. In Figure 6, the authors should provide images in the z-plane to show definitive attachment. Likewise, it is advisable to define firm attachment in line 349. As this is only observed as an n=1, the location of the embryo may be a fixation artifact. Additionally, it would not be expected for the trophectoderm to attach to glandular organoids on the side of the basal lamina. Thus, the authors should address the polarity of the organoids should be explored.

5. Line 259 refers to EpC5 when it should be EpS5.

Reviewer #2 (Recommendations for the authors):

P 7: I would be interested in your thinking about the role of NAC in the base medium. There is a role for NOX4 and ROS in decidualization, so should NAC not be inhibiting decidualization?

Line 157: the role of the level of stress is not really addressed in this paper as I see it, i.e. there is no comparison between more or less replication. May be this statement is not helpful here.

Also it seems not clear how your experimental model deals with differences between luteal phase changes and those caused by the embryo. Do you expect cAMP signaling be part of the spontaneous decidualization in the cycle or dependent on embryo attachment?

Line 237: PDGFRA and -B are receptors not growth factors.

Line 276: EpC4 should be EpS4.

281 to 282: isn't that always the case?

Random thought: Given what you write about spread of senescence to neighboring cells, I am wondering whether menstruation results from the withdrawal of DSC directed recruitment of uNK???

Reviewer #3 (Recommendations for the authors):

1) The story is not always easy to follow if not completely submerged in this field of endometrial receptivity, decidualization, importance of senescence, of NK cells, etc. Authors should at certain points be more focused and should less diverge from the subject (certainly in the Discussion) which would better retain the attention of the reader to the central message of the study. Also, the Abstract could be simplified in this regard. Moreover, the paper will highly benefit from a graphical abstract of summarizing figure, since messages conveyed are not easy for the broader reader.

2) Although the term assembloids has mostly been used for constructs combining organoids from different tissues or multiple cell types, it sounds okay and can be used here.

3) The use of a single, very broad tyrosine kinase (TK) inhibitor seems a bit crude as perturbation approach. Not all TKs act in similar processes. Do the authors have an idea about more specific TK pathways that may be involved, starting from their data? It could be interesting to just touch on one of them to zoom in on more specific molecular mechanisms. Or will authors elaborate on this in follow-up studies? Moreover, the effect of dasatinib could also be briefly confirmed with another broad TK inhibitor (although experiments with a more specific one would be rather preferred to fine-tune the focus). In lines 424-426, authors claim a role for the "level of endogenous cellular stress …" from the impact of dasatinib; however, TK signaling does not only underlie celluar stress. Please explain this better.

4) Authors describe the epithelial compartment in the organoids/assembloids as glandular. Do they have indications from their scRNA-seq data that may point to presence of some luminal epithelium? As far as the reviewer remembers, other scRNA-seq studies may have shown this (Fitzgerald et al., PNAS 2019; Cohrane group; Turco group in BioRxiv). Please discuss.

5) The study should briefly discuss similarities and differences with a recent paper of the Kessler group (Cell Rep 2021) also having designed an epithelial-stromal co-culture model.

Specific comments on the different parts:

Impact statement:

– " … reveals novel mechanisms of reproductive failure …" should be somewhat toned down since not extensively supported yet, e.g. to "… may lead to new insights into …" or " … will help to unravel/decipher …".

Abstract:

– line 37-38: not understandable for the broader reader.

– line 40: "… mesenchymal-epithelial transition, processes involved in endometrial breakdown and regeneration": is this sufficiently accepted among endometrium researchers to include it in the Abstract? I would suggest to remove this here.

– Line 46-47: is rather hypothetical for an abstract.

Introduction:

– Line 67: "phenotypic decidual cells"? Please explain or rephrase.

– Lines 107-109: "… demonstrate that different pathological states can be recapitulated …": this conclusion sounds too far, since authors use only one approach (i.e. the broad TK inhibitor), which cannot be immediately extrapolated to multiple ("different") pathological states. This generality is indeed not supported yet by experiments; some more proof-of-principle experiments would then be needed (e.g. even Crispr/Cas could be considered). This conclusion is repeated in lines 361-362 and should also there be moderated.

Results:

– Line 111, 153: "simple": why is this adjective used? As compared to?

– Line 113: "progenitor cells": it is not known whether organoids are formed by "progenitors" (neither is there conclusive evidence that they exist); this should be removed.

– Line 122: also Boretto et al. 2017 should be mentioned here.

– Line 127: how was the 1:1 ratio defined or controlled since organoids were not dissociated into single cells (see Methods line 539)? This is important technical information for a new model.

– Line 137: "… mimic luteal phase endometrium": please, show some stainings of primary tissue to support this.

– Line 162-1645: authors compare day 4 assembloids (undifferentiated), thus being in a proliferative phase, with assembloids in decidualized phase (meaning 4 days longer in culture in non-proliferative conditions). I guess the authors took this into account for DEG/GO analysis, by extracting the impact of cell cycle genes? Please explain.

– Lines 172-173: "actively dividing EpC" and "EpC with marker genes of the E2-responsive proliferative endometrium": both are proliferative, so what is their exact difference then? What are the in vivo counterparts of both subclusters? Same question applies to the stromal subclusters SS1 and SS2 (lines 206-207).

– Line 180: canonical endometrial receptivity genes: please indicate them in the figure, or mention a few here in the text.

– Lines 193 …: regarding the transitional population (TP): what may be the in vivo correlate? What could be the in vivo function of the TP? The TP appears not to be found in the in vivo endometrium (see Figure 3C). Line 257: there is a lack of crosstalk of TP with other cells, so is the TP an in vitro population not present in vivo? Should be discussed.

– Line 218: "… novel pre-decidual genes were also identified" … moderate to "novel candidate pre-decidual genes …".

– Lines 247…: "… transition between cellular states is predicated …". This could be further supported by applying pseudotime analysis (using Monocle, RNAvelocity, …). Did the authors consider or perform this?

– Line 299: dasatinib needs somewhat more introduction/explanation here. How does it work? What does it exactly do?

– Line 315: TP develop by MET: what are the predicted source cells then? No relevant information available from pseudotime analysis?

– Line 328-331: this conclusion is not clear from the data; please rephrase and/or include in graphical abstract.

– Line 342: please immediately specify that exactly one embryo is added per assembloid drop/well.

– Line 358: all embryo cocultures were found to secrete hCG; thus the 'receptive' co-cultures (without dasatinib) do not show more features of correct interaction ('pregnancy') than the dasatinib-treated ones? Please discuss.

– Some typo's: line 259: EpS5; line 276: EpS4.

Discussion:

– Lines 364-373: This part is not really to the point, and could be removed to better retain the focus.

– Line 423: "… MET drives re-epithelialization …". how sure is this in the field? Has this recently been unequivocally underpinned? If not, please moderate.

Methods:

– Line 526-527: if I understand well, authors use 95% Matrigel/5% cells in medium (1:20), which is high and completely different from other studies using lower % of Matrigel; please confirm or explain.

– ELISAs: how were secretions in the medium normalized to the number of organoids/cells present in the assembloid drop? Were they normalized to extracted cellular protein b-actin? Or in other words, how was the number of organoids/cells in the structures standardized among wells/plates/independent experiments for secreted protein measurements?

Figures:

– Figure 2: PAEP and SPP1 in ExM/E2/cAMP/MPA are indicated as non-significantly different; is this true or correct?

– Figure 3 C and E: in vivo: please add reference of these data in the legend.

– Figure 6D: right figure is not explained.

– Figure 6: "… attached by proliferating polar trophectoderm": is "proliferating" supported by experimental evidence? Please show then.

Reviewer #4 (Recommendations for the authors):

Suggestions for improved or additional experiments

– The authors should consider testing their minimal differentiation medium to determine whether it's sufficient to induce a stromal cell response, rather than only testing the secretory response of the endometrial epithelium. If these tests have been done, the authors should consider mentioning them at the place in the text where the epithelial secretion response is discussed.

– Line 166 and 167 mentions the dotted line in figure 3B. Please add an explanation for the dotted line circles.

– Line 172-173: Please provide more clarity on the difference between actively dividing epithelial cells and epithelial cells expressing markers of proliferating endometrium.

– Line 218: Were these novel pre-decidual genes verified in human pre-decidual samples? Can the authors confirm that they are not artifacts of their culture system?

– Please clarify how ligands and receptors were chosen to be entered into the CellPhoneDB program.

– Line 260: Please clarify the CellPhoneDB data. It appears that only the RNA expression data is used for this experiment, so can it be assumed that the two proteins are both interacting, even if they are both expressed? The CellPhoneDB data should be validated.

– Line 316: Why does the fact that dasatinib upregulates mesenchymal genes and decidual factors in transitional cells show that MET causes the transitional cells to emerge during decidualization? More evidence is needed for this claim.

– Line 361: Please clarify which different pathological endometrial states were recapitulated.

– Is there data demonstrating that implantation failure is associated with insufficient senescence?

– The "assembloids" are new but there are renditions of the 3D models using epithelial cells and stromal cells that are published and could be cited for comparison purposes. Given the different way the stromal cells were cocultured (in hydrogel) in the assembloid, the morphology is unique and the transcriptomics are new.

– Single cell RNA-seq of organoids have been performed by Spencer's group although they did not have the cocultured stromal cells. In that sense, the scRNA-seq of epithelial/stromal assembloids have revealed new data and it would be interesting to determine how epithelial transcriptomics changed in the presence of stromal cells.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "Modelling the impact of decidual senescence on embryo implantation in human endometrial assembloids" for consideration by eLife. Your article has been reviewed by 4 peer reviewers, including Thomas E Spencer as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Jonathan Cooper as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Gunter Wagner (Reviewer #2); Hugo Vankelecom (Reviewer #3); Ji-Yong Julie Kim (Reviewer #4).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

Please address the suggestions from the reviewers and submit a final version of the manuscript.

Reviewer #1 (Recommendations for the authors):

The authors have satisfactorily addressed the majority of the comments and suggestions from the initial review. They are encouraged to submit a final manuscript that incorporates the remaining suggestions of the reviewers.

Reviewer #2 (Recommendations for the authors):

Thank you for the revision of the paper, and congratulations to that exciting paper. I am happy with the revision, with a few minor suggestions:

Line 95: please clarify what you mean by "turnover" of senescent cells. Do you mean removal? Since turnover seems to mean elimination and replacement. How does turnover recruit decidual precursor cells? Or do you mean secretions from senescent cells recruit mesenchymal precursor cells?

The Diniz-da-Costa reference is incomplete.

Line 380ff: not sure whether it is appropriate to say that SC analysis, which is a descriptive tool, gives evidence for causality (requires divergence…).

Reviewer #3 (Recommendations for the authors):

The authors responded appropriately to the suggestions and remarks, and very well clarified the points of confusion. I have no further comments.

Reviewer #4 (Recommendations for the authors):

The authors have addressed most of Reviewer 4's concerns. However, it remains important that other epithelial and stromal organoid systems are cited in this study, as also suggested by Reviewer 3.

Overall, this is a lovely study.

https://doi.org/10.7554/eLife.69603.sa1

Author response

Essential revisions:

1) The story is not always easy to follow if not completely submerged in this field of endometrial receptivity, decidualization, importance of senescence, of NK cells, etc. Authors should at certain points be more focused and should less diverge from the subject (certainly in the Discussion) which would better retain the attention of the reader to the central message of the study. Also, the Abstract could be simplified in this regard. Moreover, the paper will highly benefit from a graphical abstract of summarizing figure, since messages conveyed are not easy for the broader reader.

The current implantation paradigm in humans is informed by animal models, foremost by studies in mice. Consequently, the implantation process is often reduced and equated to breaching of the luminal epithelium by an embryo. In mice, a polytocous rodent, the barrier function of the luminal epithelial plays a critical role in synchronising implantation of multiple embryos, which can transiently arrest development (diapause) in the uterine cavity while awaiting a maternal nidation signal (estrogen). Implantation in humans, on the other hand, normally involves a single embryo, often harbouring complex chromosomal errors and lacking the ability to transiently arrest development while awaiting a maternal implantation signal (which does not appear to exist). Hence, whether the barrier function of the luminal endometrium epithelium in mice is maintained or degraded in humans is questionable and, by extension, so is the prevailing implantation paradigm. More importantly, the major hurdle at implantation in humans, and other menstruating primates, is not synchronised implantation of multiple blastocysts, but transformation of cycling endometrium into the decidua of pregnancy robust enough to accommodate a deeply invading placenta throughout gestation. This process inevitably requires intense tissue remodelling and the underlying mechanisms that control endometrial fate decision at implantation are explored in this study. Hence, we make no apologies for describing the major cellular players in this process. In fact, all the players involved in endometrial remodelling (acute senescent cells, differentiated cells, innate immune cells, and chronic senescent cells) will be instantaneously recognisable to anyone working in the field of tissue remodelling, whether in the context of development, wound healing, ageing or even cancer. Nevertheless, we appreciate this request for more clarity. Hence, we have produced a Graphical Abstract, simplified the Abstract, and amended the Discussion.

2) The epithelial organoid cultures are referred to as glandular organoids throughout the manuscript; however, the cultures may contain a mixed population of glandular and luminal epithelial cells. It would be beneficial to clarify the proportion of glandular and luminal epithelial cells within the culture. This can be done using established markers of endometrial glandular epithelial cells such as FOXA2.

The use of the word ‘glandular’ was intended to indicate the in vitro characteristics of the tubular, gland-like organisation of epithelial cells in assembloids, rather than the identity of the progenitor cells, but we see how this can be misconstrued. We note that the Turco group used the same terminology in their original report (PMID: 28394884), as well as describing the establishment of the organoid cultures from ‘retained glandular elements … backwashed from the sieve membranes’. Nevertheless, we have changed the term ‘glandular’ organoids to ‘gland-like’ organoids to avoid confusion.

We note the reviewer’s suggestion to explore the proportions of glandular and luminal cells and attempted to do so by examining the expression of putative maker genes in the different populations of undifferentiated and decidualized assembloids. However, constitutive expression of commonly described markers is not apparent either in whole tissue (see GEO Dataset GDS2053: Endometrium through the Menstrual Cycle, and PMID: 16306079) nor in our assembloid cultures (Author response image 1A), and markers described as being specific to luminal epithelium are expressed and temporally regulated at RNA level in laser capture micro-dissected glands in vivo (Author response image 1B).

Author response image 1
(A) Expression of FOXA2, a glandular epithelial marker (green font), and putative luminal epithelial marker genes (black font) in assembloid subpopulations.

Dot size indicates the proportion of cells expressing the marker, while colour indicates the level of expression. Note that none of the cell populations express a compelling luminal epithelial marker gene signature. (B) Expression of the same markers in laser-capture micro-dissected endometrial glands in vivo, obtained during the early- and mid-luteal phase (LH+5 and LH+8, respectively).

Furthermore, putative protein markers identified in tissue sections (e.g. PMID: 28394884 and Garcia-Alonso, bioRxiv https://doi.org/10.1101/2021.01.02.425073) are not well-represented at the RNA level, perhaps due to the dropout phenomenon observed in single-cell datasets (see PMID: 32127540). Further, in the Garcia-Alonso preprint article, label transfer from in vivo single-cell data onto the gland-like organoid dataset reveals very few luminal epithelial cells to our eyes, and expression of luminal epithelial markers (e.g. LGR5, SOX9,WNT7Ahigh) appears lost upon hormonal differentiation of organoids. It is also notable that the overlap in epithelial gene expression between gland fragments and organoids reported by Turco et al., (PMID: 28394884) was fairly low (30-37%), suggesting that cells do undergo a phenotypic change in vitro, although the full gene lists are not available alongside that paper.

3) The authors should consider testing their minimal differentiation medium to determine whether it's sufficient to induce a stromal cell response, rather than only testing the secretory response of the endometrial epithelium. If these tests have been done, the authors should consider mentioning them at the place in the text where the epithelial secretion response is discussed.

We have now tested the decidualization response of primary endometrial stromal cells in standard 2D cultures to the minimal differentiation medium supplemented with E2, cAMP and MPA (MDM+E2+cAMP+MPA) in comparison to our standard differentiation medium for 2D cultures (2% DCC-DMEM+cAMP+MPA). Further, the impact of assembloid expansion medium supplemented with E2 (ExM+E2) during the growth phase was compared to that of our standard growth medium (10% DCC-DMEM) for primary cultures. Secreted levels of IL-8 (released by pre-decidual cells without corresponding induction at mRNA level) and CXCL14 and TIMP3 (induced upon decidualization) were monitored in two biologically independent cultures. Unfortunately, access to endometrial biopsies is currently very limited because of the impact of the pandemic on our research clinic. As shown in Author response image 2, kinetics of decidual secretions were identical under all treatment conditions, characterised by a rapid but transient rise in IL-8 secretion on day 2 of decidualization and followed by accelerated rise in CXCL14 and TIMP3 secretion by day 4. Further, when subjected to the same differentiation medium (2% DCC-DMEM+cAMP+MPA), primary endometrial stromal culture grown first in ExM+E2 tended to secrete lower IL-8 levels on day 2 of decidualization but higher CXCL14 and TIMP3 levels by day 4 when compared to cultures first grown in 10% DCC-DMEM. These observations reinforce our assertion that the amplitude of initial pre-decidual stress response is determined during the preceding proliferative phase and that a blunted pre-decidual stress response accelerates decidualization and vice versa. We have reported similar observations previously (PMID: 31965050).

Author response image 2
Comparison of secreted levels of IL-8, CXCL14 and TIMP3 in two independent primary endometrial stromal cells maintained in standard 2D cultures in response to standard growth medium (10% DCC-DMEM+E2) or assembloid expansion medium (ExM+E2) and standard differentiation medium (2% DCC-DMEM+cAMP+MPA) or assembloid differentiation medium (MDM+E2+cAMP+MPA), as indicated.

Reviewer #1 (Recommendations for the authors):

1. As the manuscript introduces a new co-culture model of endometrial stroma and epithelial cells, the authors must describe the relative proportion of cells used to establish co-culture. Specifically, a 1:1 (v/v) ratio of different cell types is ill-defined and will make replication difficult as pellet size can depend on the amount of organoid disruption. Please clarify to enhance reproducibility.

Thank you for raising this issue. The stated 1:1 (v/v) ratio upon mixing of the different cell types equates to 5×104 stromal cells and epithelial organoid fragments passaged at a ratio of 1:2 per assembloid culture. As mixing involved passaged organoid fragments, we cannot elaborate precisely on the number of individual epithelial cells. Nevertheless, this protocol is robust. The procedure has been clarified in the methods section.

2. The epithelial organoid cultures are referred to as glandular organoids throughout the manuscript; however, the cultures may contain a mixed population of glandular and luminal epithelial cells. It would be beneficial to clarify the proportion of glandular and luminal epithelial cells within the culture. This can be done using established markers of endometrial glandular epithelial cells such as FOXA2.

We refer the Reviewer to our previous response (Essential Revisions, point 2).

3. Although the authors have introduced a new model that likely will advance our understanding of implantation and pregnancy loss in the future, it is unclear if it provides an advancement in the study of uterine stromal cell decidualization. An unaddressed question from this study is the impact of epithelial cells on stromal cell decidualization. In the future, it will be necessary to directly compare this model with 3D stromal cell culture in the absence of epithelial cells.

Thank you – we appreciate this point. In fact, as part of a different project, we are assessing if stromal cells from control subjects can attenuate epithelial defects in gland-like organoids of women with recurrent missed miscarriage. As implied by the Reviewer, endometrial assembloids are indeed a powerful tool to dissect how defects in one cellular compartment influence differentiation response in the other compartment.

4. In Figure 6, the authors should provide images in the z-plane to show definitive attachment. Likewise, it is advisable to define firm attachment in line 349. As this is only observed as an n=1, the location of the embryo may be a fixation artifact. Additionally, it would not be expected for the trophectoderm to attach to glandular organoids on the side of the basal lamina. Thus, the authors should address the polarity of the organoids should be explored.

The time-lapse imaging shows decidualizing stromal cells extending toward the polar trophectoderm of the blastocyst and making contact by 46-50 hours in co-culture. Please see Figure 6—figure supplement 1, where stromal cell migration is demarcated by a dotted white line. Migrating cells continue to surround the embryo at this site and then draw the embryo toward the assembloid matrix. We are confident in the attachment of this embryo since it remained associated with the assembloid through transfer from the culture well into fixation and washing solutions, postage and staining processes while unattached embryos were lost during processing. However, we have removed the word ‘firm’ from the text to avoid potential overstatement. Future work will involve more extensive imaging and the use of additional labels to confirm implantation and outgrowth. Interestingly, endoglandular trophoblast invasion has been reported in human pregnancies (PMID: 13362122, PMID: 26493408), including at very early post-implantation. So, while we agree that initial embryo-endometrium interaction in vivo would not be expected to involve the basal lamina of glands, this may well occur shortly after implantation.

5. Line 259 refers to EpC5 when it should be EpS5.

Thank you. We have corrected this error.

Reviewer #2 (Recommendations for the authors):

P 7: I would be interested in your thinking about the role of NAC in the base medium. There is a role for NOX4 and ROS in decidualization, so should NAC not be inhibiting decidualization?

We agree with the Reviewer that the mechanisms accounting for the beneficial action of NAC in our model warrant further exploration. However, the concentration of NAC used in the minimal differentiation medium is an order of magnitude lower than needed for ROS clearance (MDM contains 1.25 mM NAC, 10-15 mM NAC) is used elsewhere (PMID: 17343919, PMID: 19498006, PMID: 31885807, PMID: 33585475). Therefore, we speculate that low-dose NAC may not constrain decidualization-associated ROS production in assembloids, yet confer some protection against oxidative cell death and therefore beneficial upon differentiation. Additional roles for NAC include serving as a reserve of amino acid cysteine and maintaining redox homeostasis, which may also contribute to its beneficial effects in minimal differentiation medium (PMID: 18671159, PMID: 20602078). We have now stated in the text that NAC was used at low concentration.

Line 157: the role of the level of stress is not really addressed in this paper as I see it, i.e. there is no comparison between more or less replication. May be this statement is not helpful here.

In line with previously reported observations (PMID: 31965050), the dastinib experiments reinforces our assertion that the level of stress in pre-decidual cells (for example, measured by the secreted levels of IL-8), determines the kinetics and quality of the subsequent decidual response.

Also it seems not clear how your experimental model deals with differences between luteal phase changes and those caused by the embryo. Do you expect cAMP signaling be part of the spontaneous decidualization in the cycle or dependent on embryo attachment?

Endometrial cAMP levels rise sharply upon transition from the proliferative to secretory phase in non-conception cycles. We do not have information on whether embryonic cues stimulate cAMP/PKA signalling in endometrial stromal cells. In a different project, we exhaustively investigated if hCG increases cAMP levels in primary stromal cells but found no evidence for this, contrary to some claims in the literature.

Line 237: PDGFRA and -B are receptors not growth factors.

Thank you. We have amended the text.

Line 276: EpC4 should be EpS4.

Thank you. We have corrected this error.

281 to 282: isn't that always the case?

Apologies but we are unsure what the Reviewer is referring to.

Random thought: Given what you write about spread of senescence to neighboring cells, I am wondering whether menstruation results from the withdrawal of DSC directed recruitment of uNK???

Absolutely! We believe indeed that this is indeed the case. In the absence of an implanting embryo, and in response to falling progesterone (which disables the uNK-decidual cell partnership/cooperation), unopposed bystander senescence driven by mounting SASP is anticipated to cause sterile inflammation in the superficial layer, triggering a concatenation of events that ultimately results in menstrual breakdown. Across the menstrual cycle, expression of canonical senescence-associated genes generally rises sharply during the late-secretory phase. It is also plausible that transient senescence-associated inflammation in the superficial layer prior to menstruation ‘primes’ endometrial progenitors residing in the basal layer to promote tissue regeneration in the next cycle. The biological basis for this assertion is that transient SASP induces de-differentiation of cells to a more progenitor-like state and ‘lock-in’ of stem cells, thereby accelerating tissue regeneration upon resolution of SASP. Thus, it is conceivable, if not likely, that endometrial repair mechanisms become poised to be activated prior to menstruation. Notably, persistent, chronic SASP causes stem cell depletion and reduces regenerative capacity of tissues, a process widely believed to drive ageing and age-related disorders.

Reviewer #3 (Recommendations for the authors):

1) The story is not always easy to follow if not completely submerged in this field of endometrial receptivity, decidualization, importance of senescence, of NK cells, etc. Authors should at certain points be more focused and should less diverge from the subject (certainly in the Discussion) which would better retain the attention of the reader to the central message of the study. Also, the Abstract could be simplified in this regard. Moreover, the paper will highly benefit from a graphical abstract of summarizing figure, since messages conveyed are not easy for the broader reader.

We refer the Reviewer to our previous response (Essential Revision 1).

2) Although the term assembloids has mostly been used for constructs combining organoids from different tissues or multiple cell types, it sounds okay and can be used here.

Thank you.

3) The use of a single, very broad tyrosine kinase (TK) inhibitor seems a bit crude as perturbation approach. Not all TKs act in similar processes. Do the authors have an idea about more specific TK pathways that may be involved, starting from their data? It could be interesting to just touch on one of them to zoom in on more specific molecular mechanisms. Or will authors elaborate on this in follow-up studies? Moreover, the effect of dasatinib could also be briefly confirmed with another broad TK inhibitor (although experiments with a more specific one would be rather preferred to fine-tune the focus). In lines 424-426, authors claim a role for the "level of endogenous cellular stress …" from the impact of dasatinib; however, TK signaling does not only underlie celluar stress. Please explain this better.

We agree that the broad action of dasatinib limits mechanistic understanding of senescent cell clearance in our assembloid cultures. Our data show enrichment of SRC, ABL2 and EPHB1 expression in the senescent stromal cells (SS5), suggesting that these are the key cell survival pathways that could be targeted by dasatinib. By contrast, the transitional population (TP) shows enriched expression of EPHB2 and EPHB4, suggesting a different mode of action for dasatinib in these cells. The molecular mechanisms for senescence clearance using more specific inhibitors is currently under investigation.

Author response image 3

4) Authors describe the epithelial compartment in the organoids/assembloids as glandular. Do they have indications from their scRNA-seq data that may point to presence of some luminal epithelium? As far as the reviewer remembers, other scRNA-seq studies may have shown this (Fitzgerald et al., PNAS 2019; Cohrane group; Turco group in BioRxiv). Please discuss.

We refer the Reviewer to our previous response (Essential Revisions, point 2).

5) The study should briefly discuss similarities and differences with a recent paper of the Kessler group (Cell Rep 2021) also having designed an epithelial-stromal co-culture model.

While there are parallels between the models, these authors used iPSC-derived fibroblasts to establish their 3D culture but were unable both in this and previous study (PMID: 32937244) to develop the model using primary stromal cells. The reasons for this are not entirely clear but may relate to the use of the basement membrane matrix, Matrigel, in their model. Indeed, clustering of their stromal lineage is very restricted to the periphery of the organoids, while our stromal cells form a matrix throughout the gel droplet. Thus, the approach taken by Cheung et al., i.e. using iPSC-derived stromal cells, has the advantage of being able to model mechanisms of lineage commitment. However, our model successfully incorporates primary endometrial stromal cells in a collagen matrix to model the functional layer of the endometrium. This offers a more physiological approach to study implantation processes in patient-specific assembloids, investigate endometrial dyshomeostasis, and evaluate therapeutic intervention. Additionally, our characterisation of a minimal differentiation medium further advances a more physiological model, enabling intrinsic cell-cell communications to regulate growth and differentiation.

Specific comments on the different parts:

Impact statement:

– " … reveals novel mechanisms of reproductive failure …" should be somewhat toned down since not extensively supported yet, e.g. to "… may lead to new insights into …" or " … will help to unravel/decipher …".

Thank you – done.

Abstract:

– line 37-38: not understandable for the broader reader.

Thank you, amended.

– line 40: "… mesenchymal-epithelial transition, processes involved in endometrial breakdown and regeneration": is this sufficiently accepted among endometrium researchers to include it in the Abstract? I would suggest to remove this here.

– Line 46-47: is rather hypothetical for an abstract.

Thank you. It is but we have nevertheless simplified the abstract.

Introduction:

– Line 67: "phenotypic decidual cells"? Please explain or rephrase.

Thank you, done.

– Lines 107-109: "… demonstrate that different pathological states can be recapitulated …": this conclusion sounds too far, since authors use only one approach (i.e. the broad TK inhibitor), which cannot be immediately extrapolated to multiple ("different") pathological states. This generality is indeed not supported yet by experiments; some more proof-of-principle experiments would then be needed (e.g. even Crispr/Cas could be considered). This conclusion is repeated in lines 361-362 and should also there be moderated.

We are confident of these statements as the findings from this study complement and reinforce our previous observations using patient samples and primary cultures (and indeed ongoing studies). However, we appreciate the cautioning against overstated claims. Hence, we have rephrased this claim as ‘ …that aspects of different pathological states…’.

Results:

– Line 111, 153: "simple": why is this adjective used? As compared to?

Thanks, we removed this inappropriate adjective throughout.

– Line 113: "progenitor cells": it is not known whether organoids are formed by "progenitors" (neither is there conclusive evidence that they exist); this should be removed.

Thanks, done.

– Line 122: also Boretto et al. 2017 should be mentioned here.

Thank you, we have included this reference.

– Line 127: how was the 1:1 ratio defined or controlled since organoids were not dissociated into single cells (see Methods line 539)? This is important technical information for a new model.

Thank you. As mentioned in response to Reviewer 1 (Comment 1), the 1:1 (v/v) ratio of the different cell type pellets equates to 5×104 stromal cells and epithelial organoid fragments passaged at a ratio of 1:2 per assembloid culture. The reference to the ratio in line 123 has been removed, and the procedure has been clarified within the methods section.

– Line 137: "… mimic luteal phase endometrium": please, show some stainings of primary tissue to support this.

The spatiotemporal expression of these markers in secretory endometrium is well-described in the literature, e.g. PMID: 9022601, PMID: 9021377 (Laminin); PMID: 11591413, PMID: 29420252 (Osteopontin); PMID: 25695723, PMID: 22215622 (Glycodelin) and PMID: 22215622, PMID: 9620842 (PGR).

– Line 162-1645: authors compare day 4 assembloids (undifferentiated), thus being in a proliferative phase, with assembloids in decidualized phase (meaning 4 days longer in culture in non-proliferative conditions). I guess the authors took this into account for DEG/GO analysis, by extracting the impact of cell cycle genes? Please explain.

We appreciate the reviewer’s comment on our experimental design. As outlined in our response to a similar comment from Reviewer 2, our experimental design mimics the temporal transition from proliferative to secretory phase endometrium. Cell cycle phase was determined in our scRNAseq data according to the Seurat standard workflow. However, we made a conscious decision not to perform regression of cell cycle gene expression due to studying a model of differentiation wherein a switch in cell cycle state from proliferative (S, G2/M) to a predominantly non-dividing (G1/0) status is expected.

– Lines 172-173: "actively dividing EpC" and "EpC with marker genes of the E2-responsive proliferative endometrium": both are proliferative, so what is their exact difference then?

We apologise for the lack of clarity in our expression here. The reviewer is right that both populations are proliferative. EpS1 represents a highly proliferative population of cells enriched in expression cell cycle markers, with GO terms relating to DNA replication, centromere complex assembly and chromosome organisation. By contrast, EpS2 represents cells are enriched in markers of the proliferative phase endometrium with GO terms including ‘reproductive structure development’, ‘epithelial cell development’ and ‘response to hormone’.

What are the in vivo counterparts of both subclusters? Same question applies to the stromal subclusters SS1 and SS2 (lines 206-207)

.

Recently we reported the characterisation of a population of highly proliferative mesenchymal cells (hPMC) in mid-luteal endometrial samples (PMID: 33764639). As our assembloids were established from luteal-phase biopsies, the SS1 population seen in our present data may be the in vitro representation of this mesenchymal population, i.e. clonogenic cells which retain a partially conserved signature of bone marrow-derived decidual progenitors. The SS2 population, by contrast, represents the resident stromal cells of the proliferative phase endometrium. Markers for endometrial epithelial progenitors are less well described. We hypothesise that the cells in EpS1 may represent a clonogenic population of epithelial cells, which have been shown to give rise to organoids when cultured as single cells (PMID: 28394884, PMID: 28442471) but this remains to be confirmed. Whether this population differs in patients with reproductive dysfunction (e.g. recurrent miscarriages) is also of interest to future studies.

– Line 180: canonical endometrial receptivity genes: please indicate them in the figure, or mention a few here in the text.

In both panels C and E of Figure 3, these genes are highlighted in green font, as indicated in the legend. We have now added the citations to the legend as well, per the request below.

– Lines 193 …: regarding the transitional population (TP): what may be the in vivo correlate? What could be the in vivo function of the TP? The TP appears not to be found in the in vivo endometrium (see Figure 3C). Line 257: there is a lack of crosstalk of TP with other cells, so is the TP an in vitro population not present in vivo? Should be discussed.

Based on single-cell analysis of fresh biopsies (i.e. processed immediately after collection), we reported the presence of an analogous, ambiguous population of cells co-expressing epithelial and stromal marker genes in midluteal endometrium (PMID: 29227245). As outlined in our response to similar questions raised by Reviewer 2, it is plausible that this transitional population engages in turnover of luminal epithelium during implantation and there is robust evidence that MET drives re-epithelisation following the onset of menstruation (see additional comments below).

– Line 218: "… novel pre-decidual genes were also identified" … moderate to "novel candidate pre-decidual genes …".

We have amended the statement, but like to note that both DDIT4 (PMID: 29447340) and P4HA2 (PMID: 25781565) are upregulated in vitro prior to the emergence of mature decidual cell markers, and in the case of DDIT4 is critical for decidualization. Our exploration of publicly available microarray data (GEO dataset GDS2052 and manuscript Figure 3C) confirms their expression in early luteal phase endometrium, prior to decidualisation. Examination of these markers in the context of decidualization is otherwise lacking, but they appear to be exciting future targets.

– Lines 247…: "… transition between cellular states is predicated …". This could be further supported by applying pseudotime analysis (using Monocle, RNAvelocity, …). Did the authors consider or perform this?

Thank you, we did consider this suggestion. We made several attempts at pseudotime analysis of our dataset using different approaches but found that it did not perform well in the presence of only two timepoints (Day 0 and Day 4). We do consistently observe in such approaches that the (progesterone-resistant) senescent cells link closely with undifferentiated populations, concordant with shared pathways functioning to prepare for tissue renewal after menstruation.

– Line 299: dasatinib needs somewhat more introduction/explanation here. How does it work? What does it exactly do?

Thank you, we have amended the text to introduce the main target pathways of dasatinib.

– Line 315: TP develop by MET: what are the predicted source cells then? No relevant information available from pseudotime analysis?

As mentioned above, pseudotime analysis was not informative when applied to this dataset. However, based on the reduction in TP numbers and corresponding increase in decidual cells after dasatinib treatment, we propose that these cells arise after divergence of pre-decidual cells (SS3) at differentiation. The apparent dependence of TP cells on tyrosine kinase signalling agrees with developmental models of mesenchymal-to-epithelial transition (PMID: 13678588). Mesenchymal-to-epithelial transition is a well-described phenomenon in tissue differentiation and repair processes, and fundamental to endometrial biology (see PMID: 30407544, PMID: 23216285). The use of single-cell analysis, combined with our dasatinib experiments, enabled us to serendipitously identify this population.

– Line 328-331: this conclusion is not clear from the data; please rephrase and/or include in graphical abstract.

A graphical abstract is now included.

– Line 342: please immediately specify that exactly one embryo is added per assembloid drop/well.

Thank you, we have amended the text.

– Line 358: all embryo cocultures were found to secrete hCG; thus the 'receptive' co-cultures (without dasatinib) do not show more features of correct interaction ('pregnancy') than the dasatinib-treated ones? Please discuss.

The comparable levels of hCG secretion suggest equivalent blastocyst quality across the co-cultures, supporting our assertion that imbalance in decidual subpopulations is responsible for entrapment and collapse of embryos in dasatinib-treated assembloids (manuscript Figure 6D). Note, however, that the media was not changed during embryo co-culture, so we do not have temporal data on hCG expression for these embryos and the data presented are merely snapshots.

– Some typo's: line 259: EpS5; line 276: EpS4.

Thank you, we have corrected these errors.

Discussion:

– Lines 364-373: This part is not really to the point, and could be removed to better retain the focus.

The point we intended to convey is that estrogen-dependent proliferation of epithelial and stromal cells is spatially controlled in the endometrium, leading to accumulation of replicative damaged cells underneath the luminal epithelium, i.e. the site of embryo implantation and tissue remodelling. However, we have amended the Discussion as requested.

– Line 423: "… MET drives re-epithelialization …". how sure is this in the field? Has this recently been unequivocally underpinned? If not, please moderate.

At menstruation, concurrent piecemeal repair of the luminal epithelium takes place adjacent to shedding functionalis (PMID: 19252193). Emerging cells exhibit features of migratory capacity characteristic to fibroblasts (i.e. intracellular microtubular systems and pseudopodial projections), but basement membrane formation and intercellular desmosomes in line with epithelial function (PMID: 2064209). These luminal EpC closely relate to the underlying EnSC, therefore mesenchymal to epithelial transition (MET) is the most likely mechanism to explain the rapid repair process (PMID: 30407544) since tissue shedding largely precludes the likelihood that other LE progenitors are present.

Methods:

– Line 526-527: if I understand well, authors use 95% Matrigel/5% cells in medium (1:20), which is high and completely different from other studies using lower % of Matrigel; please confirm or explain.

Our protocol uses a Cells-to-Matrigel ratio of 1:20 (v/v) as described by Turco and colleagues (PMID: 28394884) in their method for the establishment of gland-like organoids.

– ELISAs: how were secretions in the medium normalized to the number of organoids/cells present in the assembloid drop? Were they normalized to extracted cellular protein b-actin? Or in other words, how was the number of organoids/cells in the structures standardized among wells/plates/independent experiments for secreted protein measurements?

We controlled the seeding density of assembloid cultures as described above and in the manuscript. Cells were harvested entirely for staining or single-cell sequencing experiments, so we were unable to extract cellular protein from these cohorts. ELISAs were performed in secretions from samples which were grown in parallel (i.e. not sequential) experiments so densities could be assessed objectively. Statistical analysis took pairing into account.

Figures:

– Figure 2: PAEP and SPP1 in ExM/E2/cAMP/MPA are indicated as non-significantly different; is this true or correct?

We agree that the lack of statistical significance was unexpected. We have reviewed the statistical analysis (Friedman test and Dunn’s multiple comparisons post-hoc test, significant accepted at P<0.05) for these data and the results presented are correct. This is likely due to the spread of the data (note the logarithmic axis), despite consistent induction of SPP1 in all 3 cultures. We used primary cells in these experiments, thus patient variability is the probable source of variance – however induction was seen across all patients, independent of phenotypic differences, indicating a robust response within the model. Culture in the NAC+ media was the only condition which appeared to support induction of both markers (regardless of the statistical outcome) and thus was selected for further experiments.

– Figure 3 C and E: in vivo: please add reference of these data in the legend.

Thank you, now added.

– Figure 6D: right figure is not explained.

Figure 6D shows the changes in diameters of embryos (µm) over 72 hours of time-lapse imaging, as determined in ImageJ. We have amended the legend to clarify this.

– Figure 6: "… attached by proliferating polar trophectoderm": is "proliferating" supported by experimental evidence? Please show then.

The polar trophectoderm of this embryo became multi-layered, hence our interpretation that proliferation is/has taken place. We agree that this requires further confirmation and will be accounted for in future immunolabelling experiments.

Reviewer #4 (Recommendations for the authors):Suggestions for improved or additional experiments

– The authors should consider testing their minimal differentiation medium to determine whether it's sufficient to induce a stromal cell response, rather than only testing the secretory response of the endometrial epithelium. If these tests have been done, the authors should consider mentioning them at the place in the text where the epithelial secretion response is discussed.

See response above (Essential Revisions, point 3).

– Line 166 and 167 mentions the dotted line in figure 3B. Please add an explanation for the dotted line circles.

Thank you. The dotted circles were intended to highlight ciliated and TP populations. These populations were not segregated by the UMAP co-ordinates referenced, which we note broadly segregated the populations into stroma/epithelium and Day0/Day4. We have amended the Figure legend to make this clear.

– Line 172-173: Please provide more clarity on the difference between actively dividing epithelial cells and epithelial cells expressing markers of proliferating endometrium.

We apologise for the lack of clarity. EpS1 represents a highly proliferative epithelial population, enriched in expression cell cycle genes and functional GO terms relating to DNA replication, centromere complex assembly and chromosome organisation. By contrast, EpS2 represents cells enriched for proliferative phase endometrial genes with GO terms including ‘reproductive structure development’, ‘epithelial cell development’ and ‘response to hormone’.

– Line 218: Were these novel pre-decidual genes verified in human pre-decidual samples? Can the authors confirm that they are not artifacts of their culture system?

We thank the Reviewer for this query. Actually, upon closer scrutiny of the literature, both DDIT4 (PMID: 29447340) and P4HA2 (PMID: 25781565) have already been implicated in the decidualization process, although not as yet mapped to pre-decidual cells. Mining of publicly available microarray data (GEO dataset GDS2052 and manuscript Figure 3E) confirmed their expression in early luteal phase endometrium.

– Please clarify how ligands and receptors were chosen to be entered into the CellPhoneDB program.

CellPhoneDB is an online, publicly available repository of ligands, receptors and their interactions (https://github.com/Teichlab/cellphonedb), which integrates various existing datasets and new manually reviewed information. In order to use this computational tool, expression counts and cell metadata were extracted from our single-cell data for decidualizing cells (i.e. those populations present in D4 cultures) according to pipelines provided by the CellPhoneDB vignette. The CellphoneDB package then derives enriched receptor–ligand interactions between two cell types based on expression of a receptor by one cell type and a ligand by another cell type (as described here: https://www.cellphonedb.org/explore-sc-rna-seq). We chose to exclude integrin-interactions from our analysis and focus on cell-cell interactions rather than cell-ECM, but future investigation of these is certainly of interest to the progression of the model.

– Line 260: Please clarify the CellPhoneDB data. It appears that only the RNA expression data is used for this experiment, so can it be assumed that the two proteins are both interacting, even if they are both expressed? The CellPhoneDB data should be validated.

CellphoneDB is a curated repository of ligands and receptors, and their interactions, which is used to predict biologically relevant interacting ligand-receptor partners computationally from single-cell transcriptomics (scRNAseq) data. During the analysis pipeline, CellphoneDB considers the absolute expression levels of ligands and receptors within each cell type (using counts data) and applies empirical shuffling to calculate which ligand–receptor pairs display significant cell-type specificity. This predicts molecular interactions between cell populations via specific protein complexes and generates potential cell–cell communication networks. The dasatinib experiments were based on CellphoneDB predictions and in, that sense, constitute a validation experiment.

– Line 316: Why does the fact that dasatinib upregulates mesenchymal genes and decidual factors in transitional cells show that MET causes the transitional cells to emerge during decidualization? More evidence is needed for this claim.

We agree with the Reviewer. We amended the text and now state that the data only suggest that these cells are of stromal origins. This conjecture is supported indirectly by studies highlighted the critical role of MET in endometrial repair.

– Line 361: Please clarify which different pathological endometrial states were recapitulated.

Thank you, we have elaborated this statement.

– Is there data demonstrating that implantation failure is associated with insufficient senescence?

Yes, direct evidence has come from Cornelis Lambalk and colleagues who demonstrated in a prospective cohort study premature expression of the decidualization marker PRL is associated with repeated implantation failure (RIF) (PMID: 31389284). This study was based on immunohistochemistry. At mRNA level, PRL is not a particularly useful marker to analyse the decidual response in whole endometrial samples. We recently reported much more specific marker genes selectively expressed in decidualising stromal cells (e.g. SCARA5) and senescent decidual cells (e.g. DIO2) in luteal endometrium. We can share with the Reviewer that an ongoing analysis of almost 800 endometrial biopsies indicates that RIF is associated with increased frequency of cycles characterised by excessive decidualisation (high SCARA5, low DIO2). Conversely, RPL is associated with excessive senescence (high DIO2, low SCARA5) and the frequency of cycles with excessive senescence increases stepwise with the number of previous miscarriages, and thus the recurrence risk of miscarriage.

– The "assembloids" are new but there are renditions of the 3D models using epithelial cells and stromal cells that are published and could be cited for comparison purposes. Given the different way the stromal cells were cocultured (in hydrogel) in the assembloid, the morphology is unique and the transcriptomics are new.

Thank you. We are pleased to inform the Reviewer that we now have a comprehensive review of endometrial co-culture models under review.

– Single cell RNA-seq of organoids have been performed by Spencer's group although they did not have the cocultured stromal cells. In that sense, the scRNA-seq of epithelial/stromal assembloids have revealed new data and it would be interesting to determine how epithelial transcriptomics changed in the presence of stromal cells.

We agree that comparison between the transcriptomes of epithelial organoid cultures and epithelial/stromal assembloids would be of interest both to ourselves and others. Unfortunately, the single-cell data for the previously published results are not publicly available. Differences in the culture conditions (matrix, media etc) may also confound comparisons between datasets generated in different labs. As mentioned in our response to Reviewer 1, we are in the process of assessing if stromal cells from control subjects can attenuate epithelial defects in gland-like organoids of women with recurrent missed miscarriage.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

Reviewer #4 (Recommendations for the authors):

The authors have addressed most of Reviewer 4's concerns. However, it remains important that other epithelial and stromal organoid systems are cited in this study, as also suggested by Reviewer 3.

Other key papers are now cited in both the introduction and discussion.

Overall, this is a lovely study.

Thank you very much.

https://doi.org/10.7554/eLife.69603.sa2

Article and author information

Author details

  1. Thomas M Rawlings

    1. Division of Biomedical Sciences, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    2. Centre for Early Life, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    Contribution
    Formal analysis, Investigation, Methodology, Project administration, Visualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
  2. Komal Makwana

    1. Division of Biomedical Sciences, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    2. Centre for Early Life, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    Contribution
    Formal analysis, Investigation, Writing – review and editing
    Competing interests
    No competing interests declared
  3. Deborah M Taylor

    1. Division of Biomedical Sciences, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    2. Centre for Early Life, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    3. Centre for Reproductive Medicine, University Hospitals Coventry and Warwickshire NHS Trust, Coventry, United Kingdom
    Contribution
    Conceptualization, Writing – review and editing
    Competing interests
    No competing interests declared
  4. Matteo A Molè

    Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom
    Contribution
    Investigation, Writing – review and editing
    Competing interests
    No competing interests declared
  5. Katherine J Fishwick

    Division of Biomedical Sciences, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    Contribution
    Investigation, Writing – review and editing
    Competing interests
    No competing interests declared
  6. Maria Tryfonos

    1. Division of Biomedical Sciences, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    2. Centre for Early Life, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    Contribution
    Investigation, Writing – review and editing
    Competing interests
    No competing interests declared
  7. Joshua Odendaal

    1. Division of Biomedical Sciences, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    2. Tommy’s National Centre for Miscarriage Research, University Hospitals Coventry & Warwickshire NHS Trust, Coventry, United Kingdom
    Contribution
    Conceptualization, Writing – review and editing
    Competing interests
    No competing interests declared
  8. Amelia Hawkes

    1. Division of Biomedical Sciences, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    2. Tommy’s National Centre for Miscarriage Research, University Hospitals Coventry & Warwickshire NHS Trust, Coventry, United Kingdom
    Contribution
    Conceptualization, Writing – review and editing
    Competing interests
    No competing interests declared
  9. Magdalena Zernicka-Goetz

    1. Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge, United Kingdom
    2. Synthetic Mouse and Human Embryology Group, California Institute of Technology (Caltech), Division of Biology and Biological Engineering, Pasadena, United Kingdom
    Contribution
    Funding acquisition, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7004-2471
  10. Geraldine M Hartshorne

    1. Division of Biomedical Sciences, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    2. Centre for Early Life, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    3. Centre for Reproductive Medicine, University Hospitals Coventry and Warwickshire NHS Trust, Coventry, United Kingdom
    Contribution
    Conceptualization, Writing – review and editing
    Competing interests
    No competing interests declared
  11. Jan J Brosens

    1. Division of Biomedical Sciences, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    2. Centre for Early Life, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    3. Tommy’s National Centre for Miscarriage Research, University Hospitals Coventry & Warwickshire NHS Trust, Coventry, United Kingdom
    Contribution
    Resources, Formal analysis, Supervision, Project administration, Conceptualization, Funding acquisition, Visualization, Writing – original draft, Writing – review and editing
    For correspondence
    J.J.Brosens@warwick.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0116-9329
  12. Emma S Lucas

    1. Division of Biomedical Sciences, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    2. Centre for Early Life, Warwick Medical School, University of Warwick, Coventry, United Kingdom
    Contribution
    Formal analysis, Investigation, Methodology, Project administration, Funding acquisition, Visualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8571-8921

Funding

Wellcome Trust (212233/Z/18/Z)

  • Jan Joris Brosens

MRC Doctoral Training Partnership (MR/N014294/1)

  • Thomas M Rawlings

Warwick-Wellcome Trust Translational Partnership

  • Thomas M Rawlings

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We are grateful to the women and couples who participated in this research. We are indebted to Dr. Siobhan Quenby and all the staff in the Centre for Reproductive Medicine and Biomedical Research Unit, University Hospitals Coventry and Warwickshire National Health Service Trust, for facilitating sample collection. This work was supported by a Wellcome Trust Investigator Award to JJB (212233/Z/18/Z). TMR was supported by the MRC Doctoral Training Partnership (MR/N014294/1) and a fellowship from Warwick-Wellcome Trust Translational Partnership initiative.

Ethics

Human subjects: Endometrial biopsies were obtained from women attending the Implantation Research Clinic, University Hospitals Coventry and Warwickshire National Health Service Trust. Written informed consent was obtained in accordance with the Declaration of Helsinki 2000. The study was approved by the NHS National Research Ethics Committee of Hammersmith and Queen Charlotte's Hospital NHS Trust (1997/5065) and Tommy's Reproductive Health Biobank (Project TSR19-002E, REC Reference: 18/WA/0356). The use of vitrified human blastocysts was carried out under a Human Fertilisation and Embryology Authority research licence (HFEA: R0155) with local National Health Service Research Ethics Committee approval (04/Q2802/26). Spare blastocysts were donated to research following informed consent by couples who had completed their fertility treatment at the Centre for Reproductive Medicine, University Hospitals Coventry and Warwickshire National Health Service Trust.

Senior Editor

  1. Jonathan A Cooper, Fred Hutchinson Cancer Research Center, United States

Reviewing Editor

  1. Thomas E Spencer, Fred Hutchinson Cancer Research Center, United States

Reviewers

  1. Thomas E Spencer
  2. Gunter Wagner, Yale University, United States
  3. Hugo Vankelecom
  4. Ji-Yong Julie Kim

Publication history

  1. Preprint posted: March 2, 2021 (view preprint)
  2. Received: April 20, 2021
  3. Accepted: September 3, 2021
  4. Accepted Manuscript published: September 6, 2021 (version 1)
  5. Version of Record published: October 18, 2021 (version 2)

Copyright

© 2021, Rawlings et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,808
    Page views
  • 530
    Downloads
  • 24
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Thomas M Rawlings
  2. Komal Makwana
  3. Deborah M Taylor
  4. Matteo A Molè
  5. Katherine J Fishwick
  6. Maria Tryfonos
  7. Joshua Odendaal
  8. Amelia Hawkes
  9. Magdalena Zernicka-Goetz
  10. Geraldine M Hartshorne
  11. Jan J Brosens
  12. Emma S Lucas
(2021)
Modelling the impact of decidual senescence on embryo implantation in human endometrial assembloids
eLife 10:e69603.
https://doi.org/10.7554/eLife.69603

Further reading

    1. Cell Biology
    2. Physics of Living Systems
    Danielle Holz et al.
    Research Article Updated

    Single molecule imaging has shown that part of actin disassembles within a few seconds after incorporation into the dendritic filament network in lamellipodia, suggestive of frequent destabilization near barbed ends. To investigate the mechanisms behind network remodeling, we created a stochastic model with polymerization, depolymerization, branching, capping, uncapping, severing, oligomer diffusion, annealing, and debranching. We find that filament severing, enhanced near barbed ends, can explain the single molecule actin lifetime distribution, if oligomer fragments reanneal to free ends with rate constants comparable to in vitro measurements. The same mechanism leads to actin networks consistent with measured filament, end, and branch concentrations. These networks undergo structural remodeling, leading to longer filaments away from the leading edge, at the +/-35° orientation pattern. Imaging of actin speckle lifetimes at sub-second resolution verifies frequent disassembly of newly-assembled actin. We thus propose a unified mechanism that fits a diverse set of basic lamellipodia phenomenology.

    1. Cell Biology
    2. Developmental Biology
    Anna Keppner et al.
    Research Article Updated

    Spermatogenesis is a highly specialized differentiation process driven by a dynamic gene expression program and ending with the production of mature spermatozoa. Whereas hundreds of genes are known to be essential for male germline proliferation and differentiation, the contribution of several genes remains uncharacterized. The predominant expression of the latest globin family member, androglobin (Adgb), in mammalian testis tissue prompted us to assess its physiological function in spermatogenesis. Adgb knockout mice display male infertility, reduced testis weight, impaired maturation of elongating spermatids, abnormal sperm shape, and ultrastructural defects in microtubule and mitochondrial organization. Epididymal sperm from Adgb knockout animals display multiple flagellar malformations including coiled, bifid or shortened flagella, and erratic acrosomal development. Following immunoprecipitation and mass spectrometry, we could identify septin 10 (Sept10) as interactor of Adgb. The Sept10-Adgb interaction was confirmed both in vivo using testis lysates and in vitro by reciprocal co-immunoprecipitation experiments. Furthermore, the absence of Adgb leads to mislocalization of Sept10 in sperm, indicating defective manchette and sperm annulus formation. Finally, in vitro data suggest that Adgb contributes to Sept10 proteolysis in a calmodulin-dependent manner. Collectively, our results provide evidence that Adgb is essential for murine spermatogenesis and further suggest that Adgb is required for sperm head shaping via the manchette and proper flagellum formation.