NOTCH3 signalling controls human trophoblast stem cell expansion and differentiation Free

Accurate development of the human placenta and its different trophoblast cell types is key to physiological embryonic development and a successful pregnancy outcome. Originating from the trophectoderm, the functional units of this unique extraembryonic organ, the placental villi, evolve within 3 weeks post-implantation (Hamilton and Boyd, 1960; Hemberger et al., 2020; James et al., 2012; Knöfler et al., 2019; Turco and Moffett, 2019). The villous structures rapidly expand by branching morphogenesis, generating a surface of 12-14 m² towards the end of gestation, and fulfil all essential tasks of the mature placenta, such as fetal nutrition, gas exchange, hormonal adaption of the maternal endocrine system and immunological acceptance of the conceptus (Burton and Fowden, 2015; Erlebacher, 2013; Evain-Brion and Malassine, 2003; Napso et al., 2018). Distinct epithelial trophoblast subtypes that already emerge at the early stages of placental development execute these duties. Extravillous trophoblasts (EVTs) develop in specialized villi, the anchoring villi that attach to the maternal decidua, i.e. the endometrium of the pregnant uterus. Upon differentiation, EVTs detach from proliferative cell columns, which harbour NOTCH1/ITGA2⁺ EVT progenitors (Haider et al., 2016; Lee et al., 2018), and invade the decidual stroma and glands, ensuring allorecognition and histiotrophic nutrition of the fetus during the early phases of gestation (Burton et al., 2020; Moffett and Shreeve, 2023). EVTs also play a crucial role in controlling adapted blood flow to the placenta by altering uterine vessel function during pregnancy (Burton et al., 2010; Pijnenborg et al., 1983, 1980). At early stages of development, endovascular EVTs migrate into the maternal spiral arteries and plug the vessels, thereby preventing early onset oxygenation and damage of the placenta. However, during establishment of the fetal-maternal connection, these EVTs remodel the maternal arteries, thereby enlarging their diameter (Pijnenborg et al., 2006). This process allows the precise regulation of blood flow to the developing placenta at later stages of pregnancy when the embryo is fed by hemotrophic nutrition. In contrast to EVTs that interact with uterine cells (Pollheimer et al., 2018), multinucleated syncytiotrophoblasts (STBs), the second differentiated trophoblast cell type of the placenta, reside in placental floating villi and are surrounded by maternal blood. They develop by cell fusion of underlying cytotrophoblasts (CTBs), which represent the proliferative progenitor cell pool of the bi-layered trophoblast epithelium. STBs represent the barrier between maternal and fetal circulation, and fulfil a plethora of tasks, including secretion of pregnancy hormones, transport of nutrients and waste products, as well as oxygen delivery to the growing fetus (Lager and Powell, 2012; Maltepe and Fisher, 2015; Renaud and Jeyarajah, 2022). Defects in the placentation process have been associated with the great obstetrical syndromes, including spontaneous abortion, preterm labour, fetal growth restriction (FGR) and pre-eclampsia (PE) (Brosens et al., 2011). Failures in physiological remodelling of the spiral arteries have been found in these disorders, provoking malperfusion and subsequent oxidative stress of the placenta (Khong et al., 1986; Pijnenborg et al., 1991; Romero et al., 2011). Abnormal development, expansion and differentiation of trophoblasts could represent underlying causes, as CTBs of FGR and PE tissues show alterations in proliferation, STB formation or EVT differentiation (Farah et al., 2020; Lim et al., 1997; Redline and Patterson, 1995; Zhou et al., 1997).

Although our knowledge on placental defects in gestational diseases is largely based on histopathological examinations, adequate trophoblast systems allowing in-depth investigations of the underlying molecular mechanisms have been poorly developed. In particular, ethical constraints, the limited access to early placental material and the lack of self-renewing trophoblast stem cell models in the past have been a hindrance. However, with the recent establishment of two-dimensional (2D) human trophoblast stem cells (TSCs) and 3D trophoblast organoids (TB-ORGs) considerable progress has been made (Haider et al., 2018; Okae et al., 2018; Turco et al., 2018). Using these models, developmental signalling pathways, mediated through epidermal growth factor (EGF), Wingless (WNT)-dependent T-cell factors (TCFs) and transforming growth factor β (TGFβ)-activated SMAD3, have been delineated as crucial regulators of TSC renewal and/or differentiation (Haider et al., 2022, 2018; Okae et al., 2018; Turco et al., 2018). The HIPPO downstream factors TEAD4, YAP and TAZ are also crucial for expansion of TSCs and TB-ORGs, and abnormal expression of these factors was noticed in different pregnancy complications (Meinhardt et al., 2020; Ray et al., 2022; Saha et al., 2020).

NOTCH represents another signalling cascade that was shown to play a major role in human placentation and trophoblast development (Dietrich et al., 2022; Haider et al., 2017). In the canonical pathway, a membrane-bound ligand of the Serrate-like (JAG1 and JAG2) or Delta-like family (DLL1, DLL3 or DLL4) interacts with one of the four NOTCH receptors (Kopan and Ilagan, 2009). Subsequently, two sequential proteolytic cleavage steps result in the generation of the transcriptionally active domain of NOTCH (Fortini, 2009; Siebel and Lendahl, 2017; Wolfe and Kopan, 2004). In the first cleavage step, ADAM protease produces the NOTCH extracellular truncation (NEXT). NEXT is then chopped by γ-secretase at the membrane, or in the cytoplasm after endocytosis, thereby generating the NOTCH intracellular domain (NOTCH-ICD). The latter translocates into the nucleus with the help of importin α proteins and converts the repressor protein recombination signal-binding protein for immunoglobulin kappa J (RBPJκ) into a transcriptional activator by recruiting mastermind-like (MAML) proteins and other co-activators (Huenniger et al., 2010; Kitagawa, 2016; McElhinny et al., 2008; Siebel and Lendahl, 2017). Repressors of the hairy/enhancer of split (HES) and Hey (HES-related with YRPW motif) family members are well known targets genes of NOTCH signalling and control diverse developmental processes (Borggrefe and Oswald, 2009; Siebel and Lendahl, 2017; Weber et al., 2014).

NOTCH regulates trophoblast invasion and the expression of its receptors and ligands dynamically changes during EVT differentiation (Haider et al., 2014; Hunkapiller et al., 2011). NOTCH1 is specifically expressed in the proximal cell column of anchoring villi and has been delineated as a key regulator of EVT development, whereas NOTCH2 has been shown to control EVT migration and spiral artery remodelling (Haider et al., 2016; Hunkapiller et al., 2011; Plessl et al., 2015). Although first insights into the role of NOTCH in EVT differentiation have been unravelled, expression patterns of NOTCH components in early floating villi and their roles in trophoblast stem cell maintenance and STB formation have not been elucidated. Using primary CTBs, TSCs and TB-ORGs, we herein demonstrate that canonical NOTCH3 signalling plays a crucial role in trophoblast self-renewal. NOTCH3 is specifically expressed in CTB progenitors and NOTCH3-ICD, which binds to its co-activator MAML1, promotes expression of cell cycle and trophoblast stemness genes. Consequently, NOTCH3-ICD fosters expansion of 2D TSCs and 3D TB-ORGs, and inhibits differentiation into multinuclear STBs.

Tissue and cellular distribution of NOTCH receptors, their membrane-bound ligands as well as MAMLs were analysed in isolated first trimester trophoblasts and early placental tissues (Fig. 1). To obtain cell pools enriched with villous CTBs and STBs, respectively, a novel two-step protease digestion protocol was developed. After mechanical dissection and removal of villous tips from single placental tissues, the residual material was digested twice, allowing sequential isolation of STBs and CTBs (Fig. S1A). Real-time qPCR analyses for makers of trophoblast stemness and cell fusion indicated high purity of the isolated trophoblast subtypes (Fig. S1B). Using these samples, abundant mRNA levels of NOTCH3, MAML1, MAML2, MAML3, JAG1 and DLL1, as well as low levels of JAG2 and DLL4, were detected in CTBs and STBs, whereas NOTCH4 and DLL3 were absent (Fig. 1A,B). Although mRNA levels of the aforementioned genes were present in both villous trophoblast subtypes, protein expression of NOTCH3, MAMLs and ligands was highly restricted. NOTCH3, MAML1 and MAML3, JAG1 and DLL1 were exclusively detected in isolated CTBs, expressing the stemness markers TEAD4 and YAP1, whereas MAML2 was present in ENDOU/GDF15⁺ STBs (Fig. 1C,D). The 90 kDa NOTCH3-NEXT domain was the predominant signal obtained in the western blot analyses, suggesting canonical activation of the pathway. Immunofluorescence of 6th and 12th week placental tissues also indicated specific expression of NOTCH3 in TEAD4⁺ CTBs (Fig. 1E). NOTCH3 immunofluorescence signals were mainly detected in the cytoplasm of CTBs; however, a small fraction (below 1%) of the progenitors showed nuclear NOTCH3-ICD. MAML1 and MAML3 also localized to CTB nuclei, whereas MAML2 was mainly detected in STB nuclei (Fig. 1E). MAML1 was more abundant in 6th week placentae than in 12th week placentae, and could be detected in both Ki67-positive and Ki67-negative cells (Fig. S2A). NOTCH1 and NOTCH2 were absent from both epithelial trophoblast layers and could only be observed in stromal cells of the villous core (Fig. S2B). Expression in self-renewing TB-ORGs, established from first trimester CTB progenitors, mimicked the patterns monitored in early placental tissues, but with subtle differences (Fig. S2C). MAML1 localized to the outer CTB layer(s) and MAML2 was mainly detected in the centre of 3D TB-ORGs, where syndecan (SDC1)⁺ STBs are formed. However, MAML3 was uniformly expressed. Likewise, STBs in proximity to the TEAD4⁺ CTB layer maintained NOTCH3 expression and only the inner core was negative (Fig. S2C).

Self-renewing and fused TSCs showed a similar expression pattern of NOTCH receptors and MAML co-activators to purified CTBs and STBs, respectively (Fig. 2A). However, these cells expressed low levels of DLL1 and NOTCH1 (Fig. 2A,B). In accordance with the distribution in primary trophoblast subtypes, NOTCH1 and NOTCH3, MAML1 and MAML3, as well as JAG1, decreased during cAMP-mediated TSC fusion, whereas MAML2 increased (Fig. 2C-E). Besides the NEXT domain, NOTCH3-ICD was also detectable in protein lysates, which might indicate a high activity of the pathway in proliferating TSCs (Haider et al., 2016; Rand et al., 2000). Accordingly, immunofluorescence in self-renewing TSCs not only suggested membrane and cytoplasmic localization of NOTCH3, but also a higher proportion of NOTCH3⁺ nuclei compared with primary tissues (Fig. 2E). MAML1 and MAML3 also localized to nuclei of expanding TSCs, while NOTCH1 was only detected in a small subset of these cells. Immunoprecipitation using cell lysates prepared from TSCs revealed binding of NOTCH3-ICD to its co-activator MAML1 (Fig. 2F).

To specifically inactivate canonical NOTCH3 signalling in self-renewing TSCs, progenitor cells isolated from three different first trimester placentae were transfected with a doxycycline (Dox)-inducible dominant-negative (DN) MAML1 construct. In this plasmid, the basic domain (BD) of MAML1, which is required for binding to NOTCH-ICD, was fused to eGFP (Fig. 3A). Thus, truncated DN-MAML1 lacks protein sequences downstream of BD that are required for recruitment of other transcriptional co-activators, such p300/CBP, YAP/TAZ or β-catenin (McElhinny et al., 2008; Zema et al., 2020). Three different TSC clones were obtained that expressed fluorescent DN-MAML1 and DN-MAML1 transcripts after induction with Dox (Fig. S3A). As increased cell fusion was observed in these cultures, we performed kinetic analyses to determine the optimal time span for Dox treatment. Transcript levels of trophoblast progenitor markers were lowest after 6 days of Dox addition, whereas STB markers were highest (Fig. S3B). This condition was selected for all subsequent experiments using DN-MAML1-expressing TSC lines. Bulk RNA-seq and bioinformatics analyses of the three clones revealed 1182 upregulated and 637 downregulated genes after Dox induction (Fig. 3B). Well-known STB markers, such as genes encoding chorionic gonadotrophin β (CGB3, CGB5, CGB7 and CGB8), the syncytins (ERVW-1 and ERVFRD-1), SDC1, leptin (LEP) and growth and differentiation factor 15 (GDF15) were among the upregulated genes. The full list of regulated genes has been deposited in GEO under accession number GSE233140. In contrast, genes for trophoblast stemness (TEAD4, MYC and LGALS1) and proliferation/mitosis (CCNA2, CCNB1, CCNB2, CCND1, CCNE2 and CDK1) were significantly decreased. Although the three TSC lines differed in the principal component analysis (PCA) (Fig. S3C), relative up- or downregulation of these marker genes was consistent in each of the cultures (Fig. 3C).

To study possible side effects of the Dox supplementation, we also profiled non-transfected TSCs after 6 days of incubation with the antibiotic. Treated and untreated TSCs were highly similar in the PCA and none of aforementioned trophoblast progenitor markers and STB-specific genes was significantly changed (Fig. S3C,D). Only two of the 21 genes, significantly upregulated by Dox, were also increased upon DN-MAML1 expression, whereas none of the eight Dox-downregulated transcripts was among the DN-MAM1-suppressed genes (Fig. S3E). DN-MAML1 induction also increased transcription factors that were previously defined as a network associated with STB differentiation (Chen et al., 2022) in the three TSC clones (Fig. 3D).

To verify bulk RNA-seq data, trophoblast subtype-specific markers were analysed in the three different TSC lines (Fig. 4). Upon DN-MAML1 expression, mRNA and/or protein levels of trophoblast stemness (TEAD4, YAP1 and MYC) and proliferation/mitosis-associated genes (cyclin A, cyclin B and cyclin D) were diminished, whereas levels of STB-specific genes (GDF15, ENDOU and CGβ) were increased (Fig. 4A,B). DN-MAML1 expression also downregulated endogenous levels of NOTCH3-NEXT and NOTCH3-ICD in the different TSC clones. To assess the overall effect of DN-MAML1 on NOTCH activity, the TSC lines were transfected with a canonical NOTCH reporter in the absence or presence of Dox (Fig. 4C). Expression of the inhibitor significantly decreased reporter activity in the three different clones. Moreover, further analyses of differentially expressed genes (DEGs) using the g:Profiler tool revealed that expression of DN-MAML1 was associated with biological processes of STBs, such as response to hormones, transport and cell differentiation (Fig. S4A). In addition, target genes harbouring binding motifs for transcription factors (AP-2α, SP1 and SP3), which have previously been shown to control CGB5 expression during trophoblast cell fusion (Knöfler et al., 2004), were upregulated. In contrast, cell division, mitosis and DNA replication, as well as genes containing cognate sequences for stemness/proliferation-associated transcription factors (HES-7, HEY-1, E2F and MYC) were significantly downregulated (Fig. S4B).

Next, we investigated the biological effects of DN-MAML on TSC expansion and STB formation. Dox-induced expression of the inhibitory protein provoked downregulation of proliferation, measured by EdU labelling, whereas cell fusion, analysed by the formation of the SDC1⁺ area, increased in the three TSC lines (Fig. 5A,B). Notably, the three clones differed in proliferation and particularly in their capacity for STB formation upon DN-MAML1 expression (Fig. 5A,B). Furthermore, chemical blockage of NOTCH signalling with the γ-secretase inhibitor DBZ decreased EdU labelling in freshly prepared TSCs, diminished expression of cyclin A, cyclin B and markers of trophoblast stemness (TEAD4 and YAP1), but increased expression of STB-specific factors (GDF15 and CGβ) (Fig. 5C,D). In agreement, inhibition of the pathway with L-685458 also impaired growth of self-renewing TB-ORGs and downregulated CTB-specific proteins, whereas expression of STB markers was elevated (Fig. 5E,F). Similar results were obtained when using DBZ in the 3D system (Fig. S5A,B).

To assess the role of transcriptionally active NOTCH3 in trophoblast self-renewal, trophoblast progenitors were transiently transfected with a plasmid encoding NOTCH3-ICD (Fig. 6). Primary CTBs and TSCs overexpressing NOTCH3-ICD showed elevation of markers for trophoblast stemness (TEAD4 and YAP1) and proliferation/mitosis (cyclin A, cyclin B and cyclin D), whereas CGβ was downregulated (Fig. 6A,D). Accordingly, NOTCH3-ICD increased the number of EdU⁺ cells in TSC cultures, whereas the multinuclear SDC1⁺ area was diminished (Fig. 6B,C). Moreover, NOTCH3-ICD upregulated endogenous NOTCH3-NEXT signals in the two cell types (Fig. 6A,D). Compared with controls, the active domain also increased expression of the full-length receptor in primary CTBs and TSCs (Fig. S6A,B). In agreement with that, Dox-induction of DN-MAML1 in the TSC lines decreased the faint signals of the 270 kDa NOTCH3 protein (Fig. S6C). To gain first insights into the putative role of NOTCH3 in EVT differentiation, its expression pattern was also analysed in first trimester anchoring villi (Fig. S7A). NOTCH3 was present in EVT progenitors of the proximal cell column co-expressing MAML1 and MAML3, and, like the co-activators, decreased upon EVT differentiation in the distal region. Overexpression of NOTCH3-ICD in primary CTBs reduced expression of EVT markers, i.e. HLA-G, diamine oxidase (DAO) and fibronectin (FN1), suggesting that NOTCH3 might control EVT progenitor expansion at the expense of differentiation (Fig. S7B).

Recent omics approaches have been increasing our knowledge about human placental and trophoblast development. For example, single cell analyses of primary tissues, TSCs and TB-ORGs has provided previously unreported insights into underlying mechanisms, the diversity of trophoblasts, and their putative interactions with other placental and decidual cell types (Arutyunyan et al., 2023; Liu et al., 2018; Shannon et al., 2022; Vento-Tormo et al., 2018). Chromatin immunoprecipitation and sequencing has allowed the establishment of a landscape of TSC enhancers that harbour binding sites for pivotal transcriptions factor of trophoblast development (Frost et al., 2023; Kim et al., 2023). Moreover, integrated bioinformatics of diverse datasets have suggested signalling molecules and networks that likely play major roles in trophoblast self-renewal (Chen et al., 2022; Cox and Naismith, 2022). Although functional analyses that underpin the molecular roles of most of these genes are missing, various transcriptional regulators have been studied in detail. For example, HIF, ASCL2, GCM1 and TCF4 have been identified as crucial regulators of EVT differentiation (Jeyarajah et al., 2022; Meinhardt et al., 2014; Varberg et al., 2021; Wakeland et al., 2017), whereas factors such as TEAD4, GATA2, MSX2 and TFAPC promote TSC self-renewal (Chen et al., 2022; Hornbachner et al., 2021; Kim et al., 2023; Saha et al., 2020). Transcriptional co-activators, i.e. enhancer-associated p300, and the HIPPO factors TAZ and YAP, are also crucial for trophoblast stemness, with YAP actively repressing genes associated with cell fusion (Kim et al., 2023; Meinhardt et al., 2020; Ray et al., 2022).

In the present paper, we show that another transcriptional regulator, the NOTCH3-ICD, plays a fundamental role in TSC expansion and differentiation. NOTCH3 is the only NOTCH receptor expressed by villous CTB progenitors of first trimester placental tissues and self-renewing CTBs in TB-ORGs, whereas TSCs additionally display low levels of NOTCH1 (Figs 1 and 2). NOTCH1 was not uniformly expressed and appeared in small cell clusters of expanding TSCs. Notably, placental NOTCH1 expression has previously been identified as a marker of proliferative cell column trophoblasts representing the EVT progenitor pool of anchoring villi (Haider et al., 2016). NOTCH1-ICD has been delineated as a crucial regulator of EVT progenitor formation and survival, which represses the villous CTB progenitor phenotype by downregulating the self-renewal markers TEAD4 and p63, and activates proximal cell column-specific genes (Haider et al., 2016). However, NOTCH1 is absent from the bi-layered trophoblast epithelium of first trimester placental villi and can be detected only in the underlying stroma (Fig. S2B), as previously shown (Haider et al., 2016). Hence, upregulation of NOTCH1 could be beneficial for the viability of TSCs upon cultivation in 2D. On the other hand, passaging of TSCs in 2D could allow the expansion of some of the EVT progenitors, whereas 3D TB-ORG formation might suppress their abundance. Indeed, TB-ORGs established from the first established TSC lines showed higher NOTCH1 expression than TB-ORGs isolated from freshly prepared material from patients (Sheridan et al., 2021).

Akin to EVT differentiation (Haider et al., 2014; Hunkapiller et al., 2011), NOTCH receptors, ligands and MAMLs are dynamically regulated during STB formation. NOTCH3, JAG1, DLL1, MAML1 and MAML3 are specific CTB progenitor markers of early placental tissues, suggesting that these factors could be associated with trophoblast self-renewal (Fig. 1). Protein levels of these factors were absent from STBs and decreased during cAMP mediated cell fusion of TSCs (Fig. 2). Downregulation of NOTCH3 was also observed during in vitro EVT differentiation of primary CTBs upon cultivation on fibronectin, reinforcing the idea that the particular receptor could be crucial for trophoblast proliferation (Haider et al., 2014). In contrast to MAML1 and MAML3, MAML2 was upregulated during cell fusion. In the absence of any NOTCH receptor, MAML2 likely fulfils a differential role in STBs. Indeed, MAML proteins affect nuclear functions of other developmental regulators in different tissues (Zema et al., 2020). Notably, activation of these factors might predominantly occur when NOTCH is inactive, suggesting that NOTCH-ICDs are the preferred binding partners of MAMLs (McElhinny et al., 2008; Shen et al., 2006). MAML1 could be the predominant co-factor of NOTCH3-ICD, as MAML3 has been shown to more efficiently bind and co-activate NOTCH4-ICD than other NOTCH-ICDs (McElhinny et al., 2008; Wu et al., 2002). Whereas protein expression of NOTCH3, JAG1 and DLL1 was strongly regulated during cell fusion in situ and in vitro, their transcript levels showed modest changes. Indeed, NOTCH receptors, ICDs and ligands are subject to different post-translational modifications, intracellular trafficking routes and degradation pathways to fine-tune the signal output (Kopan and Ilagan, 2009; Siebel and Lendahl, 2017). Hence, NOTCH signalling in trophoblasts may not require strict regulation at the mRNA level.

In situ, NOTCH3 was detected at the membrane and in the cytoplasm, suggesting basal activation of the canonical pathway by ADAM cleavage and internalization of the NOTCH3-NEXT fragment, as occurs in other cellular systems (Siebel and Lendahl, 2017). Accordingly, NOTCH3-NEXT was abundant in protein lysates of primary CTBs (Fig. 1). The transcriptional co-activator NOTCH3-ICD, produced by γ-secretase in the subsequent proteolytic step, was noticed in a small percentage of CTB progenitors of early placental tissues, but was hardly detectable in cellular extracts. However, NOTCH3-ICD was present in lysates of TSCs and a higher proportion of these cells showed nuclear staining compared with primary CTBs (Fig. 2). We assume that variances in the proliferation rates of in situ CTB progenitors versus in vitro-cultivated TSCs could provide an explanation for the observed differences. TSCs, expanding in the presence of EGF, the WNT activator CHIR99021 and the TGF-β inhibitor A8301, likely exhibit elevated growth rates. This might also explain the subtle alterations of NOTCH3 regulation in TB-ORGs, undergoing self-renewal in the presence of the same factors. In agreement with the staining pattern in early placentae, NOTCH3 specifically localized to CTB progenitors of TB-ORGs (Fig. S2C). However, in contrast to its restricted in situ expression, NOTCH3 downregulation was delayed during spontaneous STB formation in this system. Nonetheless, detection of NOTCH3-ICD in situ, which has a half-life of only ∼45 min in the presence of MAML (Fryer et al., 2004; Hein et al., 2015), and the functional assays discussed below strongly suggest a role for canonical NOTCH3 signalling in self-renewal of CTB progenitors. The process of NOTCH3 activation in these cells and whether NOTCH ligands are truly involved remains subject to future investigations. Classically, activation of the NOTCH pathway requires juxtaposed cells, expressing receptor and ligand, respectively, whereas cis-expression is thought to elicit inhibitory signals (D'Souza et al., 2008; Siebel and Lendahl, 2017). Yet trans-expression could not be observed in early placental tissues as NOTCH receptors and ligands colocalize in the different trophoblast subtypes (Haider et al., 2014; Hunkapiller et al., 2011). However, among all NOTCH receptors, NOTCH3 is most easily cleaved (Choy et al., 2017). NOTCH3 shows basal signalling activity in other cells and its activation can occur in the absence of a ligand (Xu et al., 2015). Moreover, cis-expression with JAG1, which is also the major ligand of TSCs and primary CTBs (Figs 1 and 2), has been shown to promote NOTCH3-dependent gene expression (Pelullo et al., 2014). Overall, these data suggest that the mechanisms of NOTCH3 activation are highly variable and context dependent.

To exert their biological roles, ICDs of the different NOTCH receptors displace transcriptional repressors from RBPJk complexes and recruit co-activators such as p300, PCAF, MAML and others (Borggrefe and Oswald, 2009; Siebel and Lendahl, 2017). Immunoprecipitation experiments showed that the NOTCH3-ICD binds MAML1 in self-renewing TSCs (Fig. 2), suggesting that functional trimeric protein complexes between RBPJk, NOTCH3-ICD and MAML1 are formed. Indeed, individually, RBPJk and NOTCH-ICD cannot bind MAML as single proteins, only when a complex is built between the two (Nam et al., 2003). This provided the basis for the development of the pan-NOTCH inhibitor DN-MAML, which allowed the investigation of the role of canonical NOTCH signalling in many different biological contexts (Siebel and Lendahl, 2017). As DN-MAML lacks TAD1 and TAD2, which are required for binding of other regulators, it is thought to act as a specific inhibitor of the NOTCH pathway (McElhinny et al., 2008; Zema et al., 2020). Here, we designed a truncated DN-MAML1 protein, harbouring the sequence of the BD fused to GFP, and used a Dox-dependent plasmid for inducible expression in TSCs (Fig. 3). G418-seletion yielded three stable DN-MAML1-expressing TSC lines that were subjected to genome-wide expression analyses. The three clones varied in the PCA in the absence or presence of Dox, which might be explained by distinct origins of the transfected trophoblast cells. Indeed, different clusters of CTB progenitors have been described in placental tissues, primary cultures and TB-ORGs (Arutyunyan et al., 2023; Liu et al., 2018; Shannon et al., 2022; Vento-Tormo et al., 2018). However, upon Dox-induced DN-MAML1 expression, markers characteristic for STBs were upregulated in all three TSC lines, suggesting that inhibition of basal NOTCH3 activity, as indicated by downregulation of a canonical NOTCH reporter, promotes cell fusion (Figs 3 and 4). DN-MAML1 elevated activators of syncytialisation (ERVW-1 and ERVFRD-1), a cluster of genes (CGA, CGB3, CGB5, CGB7 and CGB8) encoding the pregnancy hormone CG, other STB-secreted factors (LEP and GDF15), as well as a set of DNA-binding proteins associated with cell fusion (Fig. 3D). HDAC5, which controls the activity of the master regulator of trophoblast cell fusion, GCM1, was also upregulated (Chang et al., 2013; Jeyarajah et al., 2022). Moreover, genes harbouring binding motifs for the transcriptional regulators of CGB5 were enriched (Fig. S4A). In contrast, DN-MAML1 suppressed trophoblast stemness genes (TEAD4, YAP1, MYC and LGALS1), biological processes such as cell cycle progression and mitosis, and their regulators (cyclins A and B, CCND1, CCNE2 and CDK1), as well as HES-7/HEY-1 target genes (Figs 3 and 4; Fig. S4B). Accordingly, Dox-dependent expression of the NOTCH inhibitor decreased TSC expansion by around half, while cell fusion was increased (Fig. 5A,B). Similarly, blockage of NOTCH activity by inhibiting γ-secretase diminished proliferation, and markers of trophoblast stemness/proliferation in TSCs, but upregulated STB-specific proteins (Fig. 5C,D). In agreement, the latter was confirmed in NOTCH-inhibited TB-ORGs (Fig. 5E,F; Fig. S5), suggesting that the pathway is also crucial for 3D TSC expansion.

Some of the genes downregulated by DN-MAML1 have been identified as direct targets of NOTCH/NOTCH3-ICD such as CCND1, CCNE2 and MYC, which is also bound by NOTCH1-ICD in EVT progenitors (Borggrefe and Oswald, 2009; Haider et al., 2016; Man et al., 2012; Siebel and Lendahl, 2017; Zender et al., 2016). Interestingly, NOTCH3 also maintains its own expression in TSCs (Fig. 4B), as shown in other cells (Liu et al., 2009; Weng et al., 2006), suggesting that an autocrine NOTCH3 loop governs CTB proliferation. Whether genomic sequences of other DEGs identified in this study are bound by NOTCH3-ICD requires further analyses. Furthermore, we speculate that DN-MAML1-induced STB formation and upregulation of fusogenic genes, such as the syncytins, could be largely a consequence of the downregulation of proliferation markers and exit from the cell cycle. Indeed, progression into the G0 phase of the cycle and G0-restricted expression of syncytin 2 was shown to be necessary for developing functional STBs (Lu et al., 2017).

To exclude any side effects of DN-MAML1 on other signalling pathways, TSC and CTB progenitors were also transfected with NOTCH3-ICD. Overexpression of the active domain increased TCS expansion, cyclins and trophoblast stemness genes, whereas cell fusion and STB-specific gene expression were suppressed (Fig. 6). Noteworthy, transient expression of NOTCH3-ICD also provoked upregulation of the full-length NOTCH3 receptor and its NEXT domain (Fig. 6. and Fig. S6), again suggesting that the basal activity of the pathway is maintained in an autocrine fashion. Moreover, NOTCH3 and MAML1/3 were also enriched in EVT progenitors of anchoring villi and expression of NOTCH3-ICD diminished EVT marker expression in differentiating primary CTBs (Fig. S7). Hence, these first data could suggest an additional role for NOTCH3 in EVT progenitor expansion that warrants further investigations. In summary, the results of the NOTCH3-ICD experiments are in line with observations on the DN-MAML1-expressing TSC lines, reinforcing the idea that NOTCH3 promotes trophoblast expansion at the expense of differentiation.

In conclusion, the present data suggest that NOTCH3 is a crucial regulator of early placental development (Fig. 7). Villous CTB progenitors and TSCs specifically produce NOTCH3 and its cleavage product NOTCH3-ICD, which sustains NOTCH3 expression and, consequently, the basal activity of the pathway. NOTCH3-ICD, when binding to the co-activator MAML1, maintains expression of cell cycle and stemness genes that control trophoblast self-renewal. As a result, the default differentiation pathway of floating placental villi, i.e. formation of STBs by cell fusion, is suppressed.

First-trimester placental tissues (6th to 12th week of gestation) were obtained from elective pregnancy terminations performed by the Gynmed Clinic, Vienna, Austria. Use of tissues and experimental procedures were approved by the ethical committee of the Medical University of Vienna (084/2023) which required written informed consent from donating women. If not stated otherwise, cell isolations were performed from single placentae.

Placental tissues (6th-12th week of gestation) were washed with Mg²⁺/Ca²⁺-free Hanks balanced salt solution (HBSS, Gibco) and villus trees were mechanically dissected. All villous tips were then thoroughly cropped and removed to avoid contamination with cell columns. Two digestion steps were performed using 5 ml digestion solution/ml tissue for 10 and 15 min, respectively, in a 37°C water bath with careful inversion every 2-3 min. The digestion solution was composed of HBSS containing 0.25% trypsin (Gibco) and 1.25 mg/ml DNAse I (Sigma). Each digestion reaction was stopped by using 10% FBS (vol/vol, Sigma), and the remaining tissue was collected and further processed. The cells resulting from both digestions were treated separately: the first digestion solution was enriched for STBs and the second for CTB progenitor cells. Contaminating red blood cells (if present) were removed upon incubation with erythrocyte lysis buffer (155 mM NH₄Cl, 10 mM KHCO₃ and 0.1 mM EDTA at pH 7.3) for 5 min at room temperature. Afterwards, both cell populations were washed with 1×phosphate-buffered saline (PBS; PanBiotech) and stored as cell pellets for subsequent real-time qPCR and Western blotting analyses. In addition, tissue was put aside before and after cropping of the villous tips, as well as after each digestion step, and then fixed and embedded in paraffin wax. Using a microtome, 1.5 µm sections of these tissues were prepared and stained with Haematoxylin and Eosin (H&E) using standard procedures.

Trophoblast stem cells (TSCs) were isolated by three consecutive enzymatic digestion steps followed by density gradient centrifugation according to protocols mentioned elsewhere (Haider et al., 2018). Briefly, placental tissues (6th-8th week of gestation) were washed with HBSS, villi were cut and further chopped into small pieces (1-3 mm). Three digestion steps were performed as described above. Whereas the first digestion was discarded, the second and third digestions, mainly containing the trophoblast progenitor cells, were pooled and purified using a Percoll gradient [10-70% (vol/vol)]. Cells were collected between 35% and 50% of Percoll layers, while contaminating red blood cells (if present) were removed by incubation with erythrocyte lysis buffer. Afterwards, cells were washed with HBSS and either seeded onto fibronectin-coated culture dishes using culture medium that promotes trophoblast stemness (TSC medium) or processed for organoid formation as described below. TSC medium contained DMEM/F12 (Gibco) supplemented with 1×B-27 (Gibco), 1×Insulin-Transferrin-Selenium-Ethanolamine (ITS-X; Gibco), 1 µM A83-01 (Tocris), 50 ng/ml recombinant human epidermal growth factor (rhEGF; R&D Systems), 2 µM CHIR99021 (Tocris) and 5 µM Y27632 (Santa Cruz). Cells were passaged with 80-90% confluence at a split ratio of 1:2-1:4 using TrypLE (Gibco) for 8 min at 37°C. The trypsinization reaction was stopped with 10% FBS (vol/vol) and cells were pelleted by centrifugation for 5 min at 450 g and then seeded again or subjected to cryopreservation. For cryopreservation, harvested TSCs were re-suspended in Cellbanker2 (Zenoaq) and frozen at −80°C according to the manufacturer's instructions. Molecular characterization and differentiation of TSCs as described below was performed at P3-P6.

TSCs were grown to ∼60-70% confluence in TSC medium before induction of syncytialization. STB formation was performed as previously published (Okae et al., 2018), with minor modifications. TSCs were fused after removal of A8301, rhEGF and CHIR99021 from the TSC medium by adding 2 mM forskolin. Medium was replaced after 3 days and the cells were harvested using TrypLE at day 5 or fixed for immunofluorescence staining. To remove possible contamination with TSCs that have spontaneously differentiated into HLA-G⁺ EVTs, harvested cells were re-suspended in MACS buffer (MACS Miltenyi Biotec) followed by magnetic cell separation of HLA-G-negative cells using HLA-G-PE antibodies (PE1P-292-C100, Exbio) and anti-PE MicroBeads (MACS Miltenyi Biotec), as instructed by the manufacturer. As fused TSCs are large and would become stuck in the MACS column, negative selection of forskolin-treated cells was performed by using an EasySep cell separation magnet (Stem Cell Technologies). Cell pellets were stored at −80°C for later use.

Trophoblast organoids (TB-ORGs) were established and cultivated as previously published (Haider et al., 2018) with minor modifications. Either freshly isolated trophoblast progenitor cells or cultured TSCs (P3-P6) were used for formation of TB-ORGs. Briefly, cells were washed in ice-cold advanced DMEM (Gibco) and re-suspended in ice-cold TB-ORG medium containing advanced DMEM supplemented with 1×B-27, 1×ITS-X, 10 mM HEPES (Gibco), 2 mM glutamine (Gibco), 1 µM A8391, 100 ng/ml rhEGF and 3 µM CHIR99021. Growth-factor reduced Matrigel (GFR-M; Corning) was added to reach a final concentration of 60%. Drops containing 10⁵ cells per drop in 40 µl TB-ORG medium/60% GFR-M solution were placed centrally onto each well of a 24-well plate. After 1 min of incubation at 37°C, the plates were flipped upside down to ensure equal spreading of the solidifying domes. After 20 min, the plates were turned again and 500 µl pre-warmed TB-ORG medium was added to each well. The medium was changed every 3-5 days and organoids were allowed to form for 7 days at P0 before passaging at a split ratio of 1:4. For passaging, cell recovery solution (Corning) was used according to the manufacturer's guidelines. Organoids were overlaid with ice-cold cell recovery solution and were incubated for 40 min at 4°C to depolymerize the Matrigel. After a washing step using advanced DMEM and centrifugation for 5 min at 450 g, organoids were re-suspended in TB-ORG medium/60% GFR-M. Molecular characterization was performed at P1-P2. In addition, inhibition of NOTCH signalling in P1 organoids was achieved by adding either 50 µM [(2R,4R,5S)-2-benzyl-5-(Boc-amino)-4-hydroxy-6-phenyl-hexanoyl]-Leu-Phe-NH₂ trifluoroacetate salt (L-685,458; Bachem) or 10 µM Deshydroxy LY-411575 (DBZ; Sigma-Aldrich) for 7 days. Treated organoids were fixed and subjected to immunofluorescence staining or were harvested as cell pellets for western blot analyses.

To generate the inducible DN-MAML1 expression construct, a truncated version of MAML1, fused at its C terminus to eGFP (synthesized by GenScript), was cloned into piggyBac-Tre-Dest-rtTA-HSV-neo plasmid. Truncated MAML1 contained amino acids 13-74, encoding only the basic domain needed for interaction with NOTCH-ICD. For genetic modification, three different TSC isolations (at P3-P5) derived from different first trimester placentae were used. 1 µg piggyBac plasmid together with 1 µg transposase plasmid was transfected into 70-80% confluent TSCs using DNAfectin Plus (abm). After transfection, TSCs were selected with 250 µg/ml G418 (Gibco). After about 2 weeks, GFP⁺ single clones were picked and further maintained in TSC medium in the presence of 125 µg/ml G418. DN-MAML1-GFP expression was induced with 1 µg/ml doxycycline (Sigma) for 24 h for up to 6 days. To generate TSCs constitutively overexpressing NOTCH3 ICD, three different TSC isolations (at P3-P5) were transfected with hNICD3(3xFLAG)-pCDF1-MCS2-EF1-copGFP (Addgene 40640; Zhao et al., 2012) or pcDNA3.1(-) as a control plasmid using DNAfectin Plus. The transfection rate was determined by counting GFP-positive cells. After 48 h of overexpression, cells were subjected to downstream analyses.

Pooled first trimester placental tissue (6th-8th week of gestation) was used for isolation of primary CTBs. Purification was performed as described above using three digestion steps followed by Percoll density gradient centrifugation. Before seeding onto fibronectin-coated dishes, cells were transfected with plasmids encoding hNICD3(3xFLAG)-pCDF1-MCS2-EF1-copGFP (Addgene 40640) or pcDNA3.1(-) as a control using the AMAXA SG Cell line kit (4D-Nucleofector program EO-100; Lonza). Cells were further cultivated in DMEM/F12 supplemented with 10% FBS (Sigma) and 0.05 mg/ml gentamycin for 24 h before harvesting and subjecting to downstream analyses.

TB-ORGs were fixed in 4% formaldehyde solution and embedded in paraffin wax as described elsewhere (Haider et al., 2018), while placental tissues were fixed in 7.5% formaldehyde and further processed as described previously (Haider et al., 2016). Briefly, de-paraffinized and re-hydrated serial sections were subjected to antigen retrieval using 1×PT module buffer 1 (ThermoFisher Scientific) for 36 min at 93°C using a KOS microwave Histostation (Milestone). Slides were then treated with blocking solution [Tris-buffered saline (pH 7.6) and 0.1% Tween (TBST) supplemented with 5% normal goat serum] for 1 h at room temperature and subsequently incubated with primary antibodies (listed in Table S1) diluted in TBST/5% normal goat serum overnight at 4°C. The following day, sections were washed three times with TBST and then incubated with appropriate secondary antibodies (listed in Table S1) [2 µg/ml in TBST/1% bovine serum albumin (BSA)] and 1 µg/ml DAPI (Roche) for 1 h at room temperature. Stained sections were analysed by fluorescence microscopy (Olympus BX50) and digitally photographed (CellP software; Olympus). Images were further processed using Adobe Photoshop CC 2023 including the automated merge function. For calculating the percentage of Ki67-positive cells in TB-ORGs, the total number of DAPI-positive nuclei and Ki67-stained nuclei were counted manually using Adobe Photoshop CC 2023.

TSCs were fixed with 4% paraformaldehyde for 10 min at room temperature. Subsequently, cells were washed with PBS, treated with 0.1% Triton X-100/PBS for 5 min and then washed again. After a 30 min blocking step using 0.05% Fish Skin Gelatine/PBS, cells were incubated with primary antibodies (listed in Table S1) diluted in 0.05% Fish Skin Gelatine/PBS overnight at 4°C. The following day, cells were washed with PBS and were incubated with the appropriate secondary antibody (listed in Table S1) (2 µg/ml in 0.05% Fish Skin Gelatine/PBS) and 0.5 µg/ml DAPI for 1 h at room temperature. Stained cells were analysed by fluorescence microscopy and digital pictures were taken with the EVOS FL Cell Imaging System (Life technologies). Images were further processed using Adobe Photoshop CC (2023) including the automated merge function. For quantification of cell fusion, TSC cultures were stained for syndecan 1 (SDC-1) and with DAPI. ImageJ 1.52p (Wayne Rasband, National Institutes of Health, USA) was used for semi-automated quantification of SDC1⁺ areas. Stained areas were thus encircled manually and the measured area was calculated in mm² by using the length of the incorporated scale bar to set the correct distance in pixels. The SDC1-positive area was normalized to the total number of nuclei, determined by DAPI staining.

DN-MAML1 transfected TSCs were cultured for 6 days with and without Dox stimulation before the addition of 10 µM 5-ethynyl-2′-deoxyuridine (EdU) (EdU-Click 488; BaseClick) overnight. In addition, TSCs were cultured for 5 days with and without 10 µM DBZ for pan-NOTCH inhibition before the addition of EdU overnight. Moreover, NOTCH3-ICD or pcDNA3.1(-) transfected TSCs were cultured for 48 h before incubation with 10 µM EdU for 5 h. Subsequently, TSCs were fixed and EdU was detected according to the manufacturer's instructions. In addition, nuclei were stained with 0.5 µg/ml DAPI. Cells were analysed by fluorescence microscopy and digital pictures were taken with the EVOS FL Cell Imaging System (Life Technologies). Images were processed and the percentage of proliferating cells was calculated by manually counting the numbers of total nuclei and EdU-positive nuclei using Adobe Photoshop CC (2023).

DN-MAML1 transfected TSCs were cultured for 5 days with and without Dox stimulation, and subsequently co-transfected with 0.6 μg/ml of a luciferase reporter containing 4 RBPjκ (recombination signal binding protein for Ig kappa J) binding sites (mutant or wild-type plasmids) and 0.2 μg/ml pCMV-β-galactosidase (CMV-βGal; normalization control) using DNAfectin Plus, as recently shown (Haider et al., 2016). Proteins were harvested after an additional 24 h of cultivation with and without Dox stimulation. Luciferase activity was determined on a luminometer (CLARIOstar Plus, BMG Labtech) using a luciferase assay system (Promega). Activity of β-galactosidase was quantitated on a photometer by measuring the conversion of the chromogenic substrate chlorophenol red-β-d-galactopyranoside (Roche) at 570 nm. For each sample, luciferase and β-Gal assays were performed in duplicate and mean values were calculated. To correct for variations in transfection efficiency luciferase activities were normalized to β-Gal values.

RNA isolation (PeqGold Trifst; PeqLab) and reverse transcription (RevertAid H Minus Reverse Transcriptase; Thermo Fisher Scientific) were performed according to the manufacturer's instructions. For qPCR, the following TaqMan Gene Expression Assays (ABI) were used: CCNA2 (Hs00996788_m1), CCND1 (Hs00277039_m1), CGB3 (Hs00361224_gH), DLL1 (Hs00194509_m1), DLL3 (Hs01085096_m1), DLL4 (Hs00184092_m1), ENDOU (Hs00195731_m1), GDF15 (Hs00171132_m1), JAG1 (Hs01070032_m1), JAG2 (Hs00171432_m1), MAML1 (Hs00207373_m1), MAML2 (Hs00418423_m1), MAML3 (Hs07289055_g1), MYC (Hs00153408_m1), NOTCH1 (Hs01062014_m1), NOTCH3 (Hs01128541_m1), NOTCH4 (Hs00965889_m1), TEAD4 (Hs01125032_m1), TP63 (Hs00978340_m1) and YAP1 (Hs00902712_g1). Signals (ΔCt) were normalized to TATA-box binding protein (TBP, 4333769F). To determine the amount of DN-MAML1 expression in TSC clones cultured in the absence or presence of doxycycline, SYBR Green dye-based qPCR detection was performed. The primers used for DN-MAML1 verification were: 5′-AGCACATGGTGAGCAAGG-3′ and 5′-GCAGATGAACTTCAGGGTCAG-3′. BrightGreen Express 2X qPCR MasterMix – Low ROX (abm) was used, together with a final primer concentration of 250 nM. Cycling settings were applied according to the manufacturer's instructions. Signals (ΔCt) were normalized to GAPDH expression (GAPDHfw70 5′-CCACCCATGGCAAATCC-3′ and GAPDHrev70 5′-GATGGGATTTGCATTGATGACA-3′; Sigma). The amplification efficiencies were determined by serial dilution and calculated as E=10^−1/m×100, where E is the amplification efficiency and m is the slope of the dilution curve. The Pfaffl method was used for calculation of the gene expression ratio (Pfaffl, 2001).

Whole-cell lysates were prepared using standard protocols as previously described (Haider et al., 2022; Meinhardt et al., 2020). Protein extracts were separated on SDS/PAA gels, transferred onto Hybond-P PVDF membranes (GE Healthcare) and incubated with primary antibodies (listed in Table S1) overnight at 4°C. On the following day, membranes were washed and incubated for 1 h with appropriate HRP-conjugated secondary antibodies. Signals were developed using WesternBright Chemiluminescence Substrat Quantum (Biozym) and visualized with a ChemiDoc Imaging System (BioRad).

Whole-cell lysates were prepared using cell lysis buffer and sonication as described in the manufacturer's protocols (Cell Signaling). Immunoprecipitations were performed using NOTCH3 or MAML1 antibodies, as well as appropriate rabbit IgG controls (listed in Table S1) according to manufacturer's instructions (Cell Signaling, 73778). Briefly, protein lysates were pre-cleared and then incubated with primary antibodies or IgG controls overnight at 4°C. Immunocomplexes formed between proteins and antibodies in lysates were incubated with protein A-coupled magnetic beads to form pellets. Pellets were washed with lysis buffer and then re-suspended in 3×SDS sample buffer and subjected to western blotting analyses for the detection of co-immunoprecipitated proteins.

For bulk RNA-seq, total RNA was prepared by using an AllPrep DNA/RNA/miRNA Universal Kit (Qiagen). Sequencing libraries were prepared at the Core Facility Genomics of the Medical University of Vienna, using the NEBNext Poly(A) mRNA Magnetic Isolation Module and the NEBNext UltraTM II Directional RNA Library Prep Kit for Illumina according to the manufacturer's instructions (New England Biolabs). Libraries were QC-checked on a Bioanalyzer 2100 (Agilent) using a High Sensitivity DNA Kit for correct insert size and quantified using Qubit dsDNA HS Assay (Invitrogen). Pooled libraries were sequenced on two flowcells of a NextSeq500 instrument (Illumina) in 1×75 bp single-end sequencing mode. On average, 33 million reads per sample were generated. FASTQ files were generated using the Illumina bcl2fastq command line tool (v2.19.1.403) including trimming of the sequencing adapters. Read quality was assessed by FASTQC.

The DESeq2 package was used to analyse the count matrix, with the model being constructed based on conditions and donor identity (Love et al., 2014). We applied variance stabilizing transformation for the PCA plot, followed by plotting of the key components PC1 and PC2. The log fold change shrinkage of the results was executed using approximate posterior estimation for GLM coefficients (Zhu et al., 2019). Differentially expressed genes were identified based on an adjusted P-value of less than 0.05 and an absolute log2 fold change of greater than 0.5. The identified differentially expressed genes were used to create the Venn diagrams. Heatmaps were produced using the pheatmap package, with genes selected manually. The EnhancedVolcano package was used to generate volcano plots and key transcripts were labelled (Blighe and Lewis, 2023). The entire analysis was conducted using R version 4.0.4 (https://www.r-project.org/). For generating Manhattan plots, the g:Profiler tool for enrichment analysis of DEGs was used (URL: https://biit.cs.ut.ee/gprofiler/gost). The tool was employed with an ordered query based on the results of the DESeq2 analysis. Specifically, genes were sorted by log2FoldChange and filtered to include only those with an adjusted P-value of less than 0.05. The complete output containing all results of the enrichment analyses are provided in Table S2 (genes upregulated at day 6 versus day 0 in the presence of DN-MAML1) and Table S3 (genes downregulated at day 6 versus day 0 in the presence of DN-MAML1).

All statistical analyses were performed using GraphPad Prism 9.5. A D'Agostino-Pearson normality test was performed to test Gaussian distribution, and equality of variances was examined with a Bartlett's test.

The authors thank C. Fiala (Gynmed Clinic, Vienna, Austria) for providing placental material and for gathering written informed consent of patients. We are grateful to the Core Facilities of the Medical University of Vienna, a member of VLSI, for performing RNA-seq. We thank P. Latos, Medical University of Vienna, for providing the piggyBac-Tre-Dest-rtTA-HSV-neo plasmid. We are grateful to Hubert Schwelberger, Medical University Innsbruck, for providing the DAO antibody.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202152.reviewer-comments.pdf