Notch1 mediates uterine stromal differentiation and is critical for complete decidualization in the mouse

Abstract

Uterine receptivity implies a dialogue between the hormonally primed maternal endometrium and the free-floating blastocyst. Endometrial stromal cells proliferate, avert apoptosis, and undergo decidualization in preparation for implantation; however, the molecular mechanisms that underlie differentiation into the decidual phenotype remain largely undefined. The Notch family of transmembrane receptors transduce extracellular signals responsible for cell survival, cell-to-cell communication, and differentiation, all fundamental processes for decidualization and pregnancy. Using a murine artificial decidualization model, pharmacological inhibition of Notch signaling by γ-secretase inhibition resulted in a significantly decreased deciduoma. Furthermore, a progesterone receptor (PR)-Cre Notch1 bigenic (Notch1^d/d) confirmed a Notch1-dependent hypomorphic decidual phenotype. Microarray and pathway analysis, following Notch1 ablation, demonstrated significantly altered signaling repertoire. Concomitantly, hierarchical clustering demonstrated Notch1-dependent differences in gene expression. Uteri deprived of Notch1 signaling demonstrated decreased cellular proliferation; namely, reduced proliferation-specific antigen, Ki67, altered p21, cdk6, and cyclinD activity and an increased apoptotic-profile, cleaved caspase-3, Bad, and attenuated Bcl2. The results demonstrate that the preimplantation uterus relies on Notch signaling to inhibit apoptosis of stromal fibroblasts and regulate cell cycle progression, which together promotes successful decidualization. In summary, Notch1 signaling modulates multiple signaling mechanisms crucial for decidualization and these studies provide additional perspectives to the coordination of multiple signaling modalities required during decidualization.— Afshar, Y., Jeong, J.-W., Roqueiro, D., DeMayo, F., Lydon, J., Radtke, F., Radnor, R., Miele, L., Fazleabas, A. Notch1 mediates uterine stromal differentiation and is critical for complete decidualization in the mouse.

A large body of experimental evidence has established the role of the ovarian hormones, estrogen and progesterone, in modulating blastocyst implantation in the maternal uterus. Uterine receptivity involves the coordination of multiple cellular and molecular events triggered by the presence of embryonic cues. These events allow embryonic intrusion into the luminal epithelium and the underlying stromal compartment that is essential for sustaining a pregnancy. Together, these early signals promote the differentiation of the endometrial stroma into a dense cellular matrix known as the decidua (1, 2). In an attempt to regulate trophoblast invasion precisely, the decidua obstructs the movement of the trophoblast by forming a physical barrier to cell penetration and generating a local cytokine environment that promotes trophoblast attachment, in a process known as decidualization (3). Decidualization requires a cellular differentiation of elongated spindle-shaped stromal fibroblast rounded decidual phenotype. As a result, the stromal cells acquire new functions that are essential for coordinated trophoblast invasion and placental formation. In rodents, the decidualization can be recapitulated without a specific embryonic signal; it can be induced by mechanical stimulation of the uterus sensitized by an appropriate hormonal milieu. The resulting response has been termed deciduoma, which represents the differentiation of the maternal uterine tissue in the absence of a fetus. This decidual environment entails the prevention of cellular apoptosis and the differentiation of the stromal fibroblasts to a metabolically active decidual cell. This decidual response is essential for coordinated trophoblast invasion and placental growth and formation.

Signaling and transformation during preimplantation and decidualization are strikingly similar to epithelial-mesenchymal interactions during embryogenesis, development, and tumorogenesis and involve multiple evolutionarily conserved developmental and regulatory pathways, such as Notch, among others (4). The family of Notch receptors regulates cell differentiation, cell cycle progression, and ultimately cell death in a multitude of cellular systems and niches (5). Significantly, all of these cellular processes are essential for successful decidualization. Here, we present a novel mechanistic role for Notch1 during the process of decidualization.

Notch proteins are ligand-dependent transmembrane receptors that transduce extracellular signals responsible for cell fate and differentiation throughout development and across species (6–9). Notch1 is a heterodimeric, 300-kDa type 1 transmembrane receptor, which mediates signaling induced by cell-to-cell contact. The extracellular and transmembrane Notch1 subunits are noncovalently held together by calcium-dependent interactions. Notch is initially synthesized as a single polypeptide precursor in the endoplasmic reticulum, and then it is cleaved into a bipartite protein by a furin-like convertase in the trans-Golgi (10, 11). Transmembrane ligand binding at the extracellular domain results in dissociation of the extracellular subunit from the transmembrane subunit, followed by 2 sequential cleavages of Notch. An ADAM (a disintegrin and metalloproteinase) protease catalyzes the first cleavage at the plasma membrane generating a short-lived intermediate called NEXT (notch extracellular truncation), while γ-secretase is responsible for the second cleavage (12). The final protein cleavage step occurs between glycine 1743 and valine 1744 (13, 14), located within the Notch transmembrane domain (15), and is essential for Notch function in vivo (16). Proteolytic cleavage of the NEXT Notch1 intermediate (Notch1-NEXT) by γ-secretase releases an active ∼100-kDa intracellular (Notch1-IC) peptide, which translocates to the nucleus and activates gene transcription (10, 17). Release of Notch1-IC activates transcription by binding to ubiquitous Notch transcription factor CSL [CBF1/Su(H)/Lag2] and recruiting coactivators that are essential for transcription (18).

The well-established role of Notch1 as a regulator of cell fate in various cell types led us to hypothesize that this protein might play a critical role in the differentiation of stromal fibroblasts into decidual cells, which precedes successful implantation. Although Notch receptors, ligands, and downstream effectors form a complex signaling pathway that plays multiple roles in a variety of malignancies, the physiological role of Notch in endometrial cell differentiation, as well as embryo implantation, has never been studied, although the ability of Notch to regulate proliferation, apoptosis, and differentiation is central to this process. The data presented here demonstrate a major physiological role for Notch1 in endometrial stromal cell differentiation and suggest that, in the uterus, Notch1 regulates decidualization by preventing stromal fibroblast apoptosis and promoting changes in gene expression and cytoskeleton reorganization associated with decidualization.

MATERIALS AND METHODS

Animals and tissue preparation

Mice were maintained in the designated animal care facility at the Baylor College of Medicine according to the institutional guidelines for the care and use of laboratory animals. To investigate the function of Notch1 during embryo implantation, a loss-of-function approach was utilized using genetically engineered mice. A conditional Cre-LoxP-knockout strategy was implemented. PR-Cre mice expressing Cre under the control of progesterone receptor (PR) promoter were used previously to ablate “floxed” genes in the uterus (19–21). PR-Cre mice were crossed with those harboring the floxed Notch1 gene (Notch1^f/f) provided by F.R. (22), creating the bigenic PR^Cre/+Notch1^f/f (Notch1^d/d) mice, in which the Notch1 gene is deleted in cells expressing PR. The ablation of the Notch1 gene in the uterine tissue of Notch1^d/d mice was confirmed when uterine sections obtained from these mice failed to show any Notch1 protein in the stromal cells surrounding the implanting embryo.

Hormonally induced decidual reaction

Mice (6 wk old) were ovariectomized. After 2 wk, ovariectomized mice were treated with 3 daily injections of estradiol (E2; 100 ng/mouse/d). After 2 d of rest, mice were then treated with 3 daily injections of methoxyprogesterone acetate (MPA; 1.0 mg/mouse/d) and E2 (6.7 ng/mouse/d) by subcutaneous (s.c.) injection. At 6 h after final injection, unilateral scratching on the antimesometrial lumen induced a decidual reaction. Mice were then given daily s.c. injections of MPA (1.0 mg/mouse/d) and E2 (6.7 ng/mouse/d) for 1–5 d after stimulation that followed the induction of the uterine decidual response. Mice were anesthetized with 2,2-tribromoethyl alcohol (Avertin; Sigma-Aldrich, St. Louis, MO, USA) and sacrificed by cervical dislocation to collect the uteri. Uterine tissues were weighed and then either flash-frozen and stored at −80°C until RNA isolation or fixed in 4% paraformaldehyde and paraffin-embedded for histological analysis.

γ-Secretase inhibitor (GSI) rodent chow

Wild-type C57Bl/6J mice were fed a GSI, LY-411575, via a rodent chow. LY-411575 was formulated to deliver 5 mg/kg/d based on average consumption rates for a 2-mo period, to inhibit Notch1 activity. This dosing regimen was sufficient to achieve desired Notch ablation without adversely affecting the animal's health (23).

Microarray analysis

Microarray analysis was performed by Affymetrix murine genome 430 2.0 mouse oligonucleotide arrays (Affymetrix, Santa Clara, CA, USA), as described previously (24, 25). The RNA was pooled from the uteri of 3 mice per genotype and treatment. All RNA samples were analyzed with a Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA), before microarray hybridization. The gene expression data were normalized and background corrected with the robust multichip average (RMA) method in R/Bioconductor (26). Nonspecific filtering was performed by obtaining the overall variability of each probe set across all arrays. Probe sets with low variability were discarded safely, since they do not add value in the inference of differential expression of their target genes (27). This type of filtering is called nonspecific due to the fact that no phenotype information is used in the filtering process.

To find expressed probe sets differentially, the limma package in R/Bioconductor was utilized (28). Limma uses linear models to analyze a designed experiment and to determine differential expression. It is particularly well suited for experiments with a small number of arrays, as in our case (29). To determine the linear coefficients of the model, a design matrix was created that included a coefficient for the Notch1^d/d (knockout) vs. the Notch1^f/f (wild-type) difference. Due to the small number of arrays, a moderated t statistic was used. The obtained P value for each probe set was then corrected with the Benjamini-Hochberg algorithm for multiple hypothesis testing (30). This method controls the expected false discovery rate (FDR), and probe sets with an adjusted value of P < 0.05 are considered differentially expressed. We then attempted to map all differentially expressed probe sets to gene symbols according to the Affymetrix 430 2.0 mouse genome chip annotation. The probe sets for which no gene symbol was found were discarded.

The rationale for using statistical tests to determine differentially expressed probe sets is based on the fact that fold change, as the sole criterion for differential expression, does not provide an estimation of the significance of the variations in gene expression. The normalization of the expression data, the use of a moderated t test, and the posterior correction for multiple hypothesis testing are robust methods to obtain a list of differentially expressed genes based on the significance.

Finally, 3 types of analysis were performed. First was pathway and Gene Ontology (GO) term enrichment with DAVID bioinformatics (31, 32). For this analysis, the list of differentially expressed genes was used to find enriched pathways and GO terms, classified according to GO function. Second, ingenuity pathways analysis (IPA) was also fed to the list of differentially expressed genes to detect enriched pathways. Third, gene set enrichment analysis (GSEA; refs. 33, 34) was used to find whether the two biological states present in our experiment have a statistically significant enrichment with any of the sets stored in the Molecular Signatures Database (MSigB; ref. 34). For these analyses, we did not restrict ourselves to the list of differentially expressed genes. Instead, the entire list of probe sets with their expression data (after normalization and nonspecific filtering) was used.

RNA extraction and real-time quantitative PCR (qPCR)

Total RNA was extracted from mouse uteri using TrIzol reagent (Invitrogen, Carlsbad, CA, USA). RT-PCR was performed using iScript (Bio-Rad, Hercules, CA, USA) according to the manufacturers' instructions. Real-time qPCR primers were designed using Primer Express open-source software (Applied Biosystems, Carlsbad, CA, USA). Primers were designed for Notch1, Bcl2, Bad, Tpr53, CyclinD2, p21, Hes5, Hey1, Bmp2, CyclinD3, and 18S. All real-time PCR was done with RNA samples from 3 to 5 mice. All mRNA quantities were normalized against 18S gene expression. Gene expression levels were measured by real-time RT-PCR SYBR Green analysis using the ABI Prism 7700 Sequence Detector System (Applied Biosystems) according to the manufacturer's instructions. Amplification conditions included holding for 10 min at 95°C, 40 thermal cycles of denaturing for 15 s at 95°C, and annealing/extending for 1 min at 60°C. A SYBR Green dissociation step was added to the end of the PCR cycle. Relative fold induction or expression levels were calculated using the comparative threshold cycle method for separate tube amplification.

Immunohistochemistry

For immunostaining uterine tissue was fixed overnight in 4% paraformaldehyde, and the processed tissues were embedded in paraffin. Uterine sections from paraffin-embedded tissue were cut at 5 μm and mounted on slides. Sections were deparaffinized, preincubated with 10% normal rabbit or goat serum in PBS (pH 7.5), and then incubated with Notch1 C-20 antibody (1:100; Santa Cruz Biotechnology, Santa Cruz, CA, USA), Notch1 Val-1744 (1:50; Cell Signaling, Danvers, MA, USA), Notch4 (1:100; Santa Cruz Biotechnology), and Ki-67 (1:100; BD, Franklin Lakes, NJ, USA). Controls consisted of nonimmune IgG at the same dilution. Following overnight incubation, sections were washed in PBS and incubated with biotinylated secondary antibody (Vector Laboratories, Burlingame, CA, USA) for 1 h at room temperature. Immunoreactivity was detected by using the Vectastain Elite ABC kit (Vector Laboratories), and immunoreactivity was visualized as brown staining.

Sections were reviewed in a blind procedure and scored quantitatively using positive pixel counts and the percentage of staining of each compartment, according to the intensity of DAB chromagen deposition, normalized to nuclear staining. Pictures of 5 fields/slide (0.09 mm² each) were taken using a Nuance multispectral imaging system (CRi, Hopkinton, MA, USA) at ×400 magnification and multispectral acquisition software. The images were processed by Nuance 1.6.8 image processing software to measure the spectral absorbance curve of each of the stains and then were unmixed. The percentage of positive staining was then quantified and expressed as the percentage of positive pixels to total pixels of the analyzed area. Comparison of 2 means was made with Student's t test. Comparison of ≥3 means was made by 1-way ANOVA. Values of P < 0.05 were considered statistically significant.

RESULTS

Notch1 protein expression profile during normal pregnancy

Notch1 is spatially and temporally expressed in decidualizing stromal cells of the mouse uterus (Fig. 1). Notch1 protein is evident in uterine stromal cells both during normal pregnancy and in mice undergoing artificial decidualization and increases during gestation. Protein levels are minimal in undifferentiated stromal cells during the preimplantation period (Fig. 1A, B). However, with the initiation of implantation, a marked induction in Notch1 protein was observed in the stromal cells (Fig. 1C). As pregnancy progressed, Notch1 immunoreactivity was reduced in the primary decidual zone but remained strong in the secondary decidual zone (Fig. 1D, E). The relationship between Notch expression and decidual progression implies an important role for Notch1 during the decidualization process.

Figure 1.

Notch1 protein expression profile during normal pregnancy. A, B) Notch1 protein levels are low in undifferentiated stromal cells during the preimplantation period, at d 1.5 (A) and 4.5 (B). C) However, with the initiation of implantation, a marked induction in Notch1 protein expression, which was predominantly in the nucleus, was observed in stromal cells of the primary decidual zone (C). D, E). As pregnancy progressed, Notch1 protein expression was decreased in the primary decidual zone (D), but the decidual cells in the secondary decidual zone continued to stain intensely (E). F) Negative control from d 6 of pregnancy that was treated with nonimmune IgG at the same concentration as the C-20 Notch1 antibody. Immunohistochemical images are at ×40.

Decidualization defect in the absence of Notch1: pharmacological evidence

All Notch receptors are proteolytically cleaved and activated by the multisubunit enzyme, γ-secretase, which produces the Notch transcriptionally active forms. As such, cleavage of Notch family members can be inhibited by a GSI, LY-411575, and levels of active Notch1 protein can subsequently be measured. This treatment avoids the functional redundancy by other Notch receptors (35). Wild-type C57/B6 mice were fed a chow containing LY-411575 formulated to deliver 5 mg/kg/d, based on average consumption rates, for 2 mo. This dosing regimen does not adversely affect the animal's health (36). Notch receptor ablation was verified by immunohistochemistry of Notch1, which was reduced markedly in the LY-411575-fed mice (Fig. 2A). In the same series of experiments, Notch4, a transcriptional target of Notch1, was also decreased (data not shown). Consistent with the data in the bigenic Notch1 mice, Notch1^d/d mice (described below) that were fed a LY-411575 chow demonstrated a significant reduction in decidual wet weight following artificial decidualization (Fig. 2B, C and ref. 37), compared to controls (P=0.035, n=9; Fig. 2).

Figure 2.

Decidualization defect with pharmacological inhibition of Notch1. A) Uterine section from wild-type (WT) vs. GSI-fed mice demonstrated an absence of Notch1 immunoreactivity for the cleaved Notch1-IC, using an antibody Notch1 Val1744, that specifically recognizes Notch1-IC. B) Gross morphology of the deciduoma in WT and GSI-fed mouse uteri, after decidual stimulus, shows a decidualization defect in the GSI mice. Right uterine horn was stimulated; left horn was maintained as an internal control. C) Uterine wet weight ratio of the decidual horn to the control horn demonstrated a statistically significant decrease in uterine weight in the absence of Notch1 *P < 0.035; n = 9.

Generation of uterine-specific Notch1-knockout mice

To bypass the embryonic lethality (16, 38) associated with Notch1 deficiency in transgenic knockouts and to assess whether Notch1 specifically plays a functional role in uterine physiology, a loss-of-function approach was undertaken to elucidate the role of Notch1 during decidualization. The floxed Notch1^flox/flox (Notch1^f/f) mouse, in which the exon coding for the signal peptide is floxed (39), was crossed with a mouse in which the Cre recombinase was inserted into the PR gene, PR^Cre/+ (40). On activation of the Cre recombinase, the LoxP sites in the floxed Notch1^f/f at exon 1 recombine and excise Notch1. This resultant bigenic line is PR^Cre/+Notch1^f/f (Notch1^d/d) and demonstrated high-levels of recombination throughout the uterus (Fig. 3).

Figure 3.

Loss of Notch1 expression in uterus of conditional-knockout mice, PR^Cre/+Notch1^flox/flox (Notch1^d/d). A) Mice carrying a Cre recombinase gene controlled by a PR promoter were crossed with mice carrying a floxed allele of the Notch1 signal peptide (exon1, Notch1^f/f). The resultant bigenic, Notch1^d/d, inactivates Notch1 in PR-positive tissues. B) Ovariectomized mice were subject to artificial decidualization. At d 5 following the decidual reaction, uteri were snap-frozen, and RNA was extracted and assessed by RT-PCR for Notch1. Notch1 (exon1) transcript is absent in Notch1^d/d uteri. C) Protein lysate was extracted from snap-frozen d-5 uterine horns immunobloted with Notch1-FL and Notch1-TM. Western blotting demonstrated attenuated and efficient protein reduction of Notch1 in Notch1^d/d uteri. Likewise, immunoblotting for Notch1-IC demonstrated decreased transcriptionally competent Notch1 in Notch1^d/d uteri. D) Immunohistochemical staining of Notch1 shows efficient protein ablation in uterine stromal cells from Notch1^d/d mouse uteri obtained at d 4 following the decidual stimulus (view: ×100).

To confirm that Notch1 expression is absent in the bigenic Notch1^d/d mouse model, uteri were snap-frozen, and RNA was extracted and assessed by RT-PCR for Notch1 on d 5 following the decidual stimulus. Notch1 (exon1) transcript is reduced markedly in Notch1^d/d uteri (Fig. 3B). Concomitantly, protein lysate was extracted from snap-frozen uterine horns and immunobloted for both Notch1 full-length (Notch1-FL) and Notch1 transmembrane (Notch1-TM). Western blotting demonstrated significantly reduced protein levels of Notch1 in Notch1^d/d uteri. Likewise, immunoblotting for Notch1-IC demonstrated a marked decrease in transcriptionally competent Notch1 in Notch1^d/d uteri (Fig. 3C). To further confirm the experimental paradigm, immunohistochemical staining of Notch1 verified efficient protein ablation in uterine stromal cells from Notch1^d/d mouse uterine sections as compared to Notch1^f/f controls following artificial decidualization on d 4 (Fig. 3D).

Decidualization defect in the absence of Notch1: genetic evidence

To investigate the physiological role of Notch1 during decidualization and uterine receptivity, the bigenic Notch1^d/d mice were subjected to the well-established Finn and Martin in vivo artificial decidualization protocol (37). In all mice, the left uterine horn was mechanically stimulated; the right horn was maintained as an internal control. Notch1^d/d mice showed significant decidualization failure, as quantified by uterine wet weight and gross morphology, as early as 3 d following mechanical decidualization stimuli (Fig. 4B). Notably, the unstimulated Notch1^d/d uteri were morphologically similar to the Notch1^f/f uterine horns, implying that the absence of Notch1 does not affect early postnatal uterine development.

Figure 4.

Decidualization defect in PR^Cre/+Notch1^flox/flox (Notch1^d/d) mice. A) Gross morphological evidence of the significant defect of a decidual response in Notch1d/d. In both groups, the left uterine horn was stimulated; the right horn was left untreated and served as an internal control. B) Quantification of uterine wet weight ratio of the decidual horn to the control horn in Notch1^d/d vs. Notch1^f/f uteri demonstrates a significant decrease in uterine weight in the absence of Notch1 that is apparent at 3 d after decidual reaction. **P < 0.01; ***P < 0.001.

To confirm the attenuated decidual response, both Bmp2 and Wnt4, which are conserved genes that mediate progesterone-induced stromal decidualization in mice, were decreased during decidualization in Notch1^d/d (Fig. 5A, B). These genes are also essential for fertility in mice (41). To determine the effect of the ablation of Notch1 signaling on female fertility, control, Notch1^f/f, and Notch1^d/d mice were bred to wild-type male mice. Both control and Notch1^d/d mice exhibited normal fecundity until the sixth liter (Supplemental Table S1). In the Notch1^d/d mice, the number of pups in the first litter was significantly decreased. However, in the subsequent 5 litters, all parameters (number of liters, number of pups, number of liters per mating pair, and average pups per liter) were similar between Notch1^d/d and Notch1^f/f. These results demonstrate that Notch1 is not essential for female fertility in rodents. It is possible that the Notch1 ablation in the uterus is compensated for by other Notch receptors; similar to the steroid receptor coactivator (19), in which ablation of both receptors is essential for infertility.

Figure 5.

Diminished expression of decidualization-specific genes associated with Notch ablation Notch1-dependent decidualization defects parallel diminished transcription of two decidualization-specific genes, Bmp2 (A) and Wnt4 (B), in the Notch1^d/d uteri. Data represent mean values in tissues pooled from 3 animals/group.

Microarray: pathway analysis and hierarchical clustering

To identify new genes and pathways that are dynamically regulated by Notch1 signaling during decidualization, we performed sequential DNA microarray analyses (42) of Notch1^f/f (n=3) and Notch1^d/d (n=3) deciduoma using the Affymetrix murine genome 430 2.0 mouse oligonucleotide arrays, as described previously (24, 25).

A combined list of differentially expressed probe sets was used in all class comparisons (Notch1^f/f control vs. Notch1^d/d), following artificial decidualization. Microarray analysis was performed using the limma package in R/Bioconductor (28). A probe set was considered differentially expressed if its P value, adjusted for multiple hypothesis testing, was <0.05. The probe sets that passed this criterion were mapped to gene symbols, and we finally obtained a total of 1133 differentially expressed genes. Of this group, 730 genes were significantly down-regulated, and 403 genes were up-regulated. All gene expression data, cluster groups, and functional categories were placed in the microarray database web server on the National Center for Biotechnology Information Gene Expression Omnibus (GEO; accession number GSE31244).

Pathway analysis using DAVID bioinformatics (31, 32) provided a means to quantitatively present pathways affected in the microarray, namely apoptosis, cell growth, cell structure, and developmentally significant pathways. Table 1 demonstrates a complied list of GO terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways of interest to cellular processes of decidualization. The list of all differentially expressed genes was used to enrich GO terms in 3 ontologies: Biological Process, Molecular Function, and Cellular Compartment. The enriched GO terms were divided into different groups of interest, as listed below. For each group, we took the genes enriched by the GO terms in the group. The expression level of these genes was clustered (hierarchical clustering) across the 6 samples. A heat map for each group (and its subclusters) was obtained to show the variation in expression levels (Fig. 6 and Supplemental Table S2). Pathway analysis provided clues to the etiology of Notch1-dependent decidualization failure, such as altered cell cycle checkpoint control, cellular proliferation, and differential apoptotic profiles, which are discussed further below.

Table 1.

Enriched GO terms

Function	Term	P
Apoptosis	Cell death	0.0016
	Death	0.0025
	Negative regulation of apoptosis	0.016
	Antiapoptosis	0.018
	Programmed cell death	0.019
	Negative regulation of programmed cell death	0.02
	Regulation of apoptosis	0.029
	Regulation of cell death	0.038
Cell growth	Regulation of growth	0.00023
	Regulation of cell proliferation	0.0048
	Regulation of cell growth	0.024
	Positive regulation of cell growth	0.05
	Regulation of mitotic metaphase/anaphase transition	0.05
Cell structure	Cytoskeletal protein binding	3.3E-09
	Actin binding	2.3E-08
	Extracellular matrix structural constituent	2.9E-07
	Extracellular matrix part	0.000023
	Cytoskeleton-dependent intracellular transport	0.000039
	Microtubule binding	0.00011
	Extracellular matrix	0.00022
	In utero embryonic development	0.0002
	Actin cytoskeleton organization	0.0068
	Actin filament bundle	0.036
	Negative regulation of Wnt receptor signaling pathway	0.047
Miscellaneous	Response to hormone stimulus	0.03
	Regulation of mesenchymal cell proliferation	0.031
	Negative regulation of cell differentiation	0.033
Development	Placenta development	0.017
	Reproductive structure development	0.021
	Blastocyst formation	0.027
	Reproductive developmental process	0.042

Enriched GO terms in 3 different ontologies (Biological Process, Molecular Function, and Cellular Compartment) are listed. The enriched GO terms were divided into groups of interest and expression levels of significance (P value, Bonferroni, Benjamini, and FDR). Enriched genes were clustered (hierachical clustering) across all samples for each group of selected genes.

Figure 6.

Heat maps and dendrograms. Hierarchical clustering was applied to the expression values of the cell growth- and apoptosis-related genes in each heat map. Both Notch1^d/d and Notch1^f/f were tested for biological enrichment using DAVID bioinformatics. The expression level of these genes was clustered (hierarchical clustering) across the 6 samples. A heat map for each group (and its subclusters) was obtained to show the variation in expression levels. Dendrograms at left mark the intercluster distances between the different clusters. Genes with similar expression patterns across samples cluster together.

Clustering the genes based on their expression pattern provides more insight into how the genes are regulated. The heat map summarizes the expression status (down-regulated in green and up-regulated in red) of tightly related genes (see Fig. 6). These genes belong to handpicked GO terms of interest. The clustering of these genes showed different expression profiles among them, and this may indicate different mechanisms of regulation in the absence of Notch1.

Hierarchical clustering was applied to the expression values of the genes in each heat map with a dendrogram presented to the left (Fig. 6). The dendrogram marks the intercluster distance between the different clusters. The clusters are evident by traversing the clusters from top to bottom. Genes with similar expression pattern across samples cluster together. For example, in the apoptosis heat map (Fig. 6), the expression patterns of genes Foxo1 and TGFβ2 are more closely related than the expression patterns of genes: Gata6 and Wwox. The 4 genes are down-regulated in the bigenic Notch1^d/d mice, but their expression levels varied.

Notch1 is critical for stromal cell proliferation

Cell cycle regulation is binary to cell fate and as such plays an important role in the uterus during hormonal stimulation (43) and the natural reproductive cycle (44). In anticipation of decidualization, the uterine stroma undergoes a progesterone-mediated increase in cell proliferation (37, 45, 46). Significantly, hierarchical clustering and pathway analysis demonstrated that the G1/S cell cycle checkpoint is down regulated in the Notch1^d/d mice (Fig. 6, left panel, and Supplemental Fig. S1).

To confirm the alteration of cell cycle regulatory pathways in the bigenic mice, we investigated a number of cell cycle genes that were regulated differentially in the microarray. Reduced cellular proliferation in Notch1^d/d uteri is evident, as assessed by immunoreactivity to the proliferation-specific antigen Ki-67 (Fig. 7A). This defect in proliferation provides direct in vivo evidence that the Notch pathway is required for cellular proliferation. Furthermore, The G1/S cyclin, cyclinD2, which must be induced in decidualizing stroma (47) is decreased in Notch1^d/d uteri (Fig. 7B), implying that defective decidualization parallels altered cell cycle in the absence of Notch1. In addition, it has been shown that the cyclin-dependent kinase inhibitor p21, which regulates the G₁ phase of the cell cycle, is a downstream target of Notch1 (48, 49). Accordingly, p21 mRNA is decreased without functional Notch1 (Fig. 7). Control of mammalian cell proliferation occurs mainly through G1 of the cell cycle. CyclinD2 forms a complex with and functions as a regulatory subunit of Cdk6, whose activity is required for cell cycle G1/S transition. Cdk6 mRNA is also down-regulated in the absence of Notch1 signaling. Likewise, the promiscuous cell cycle regulator and transcription factor Trp53 was also significantly reduced in our mice that demonstrate a decidual defect, in line with decreased cellular proliferation (1.75±0.51 fold reduction, data not shown).

Figure 7.

Reduced stromal cell proliferation in the absence of Notch1 signaling. A) Markedly attenuated stromal cell proliferation in Notch1^d/d mice, as evident by reduced immunohistochemical positivity to proliferation-specific antigen Ki67, at both d 0 and 4 following artificial decidualization. B) Schematic model of the G1/S transition complex. Real-time RT-PCR, presented as fold change, demonstrated down-regulation of cyclinD2, which correlates with the cessation of proliferation. Transcript levels of the regulator of cell cycle G1 progression, p21, the cyclin-dependent kinase inhibitor 1A, is diminished in Notch1^d/d uteri, implying a failure to undergo growth arrest and differentiate into the decidual phenotype. In addition, the transcript of the cyclin-dependent kinase, cdk6, is also reduced with Notch1 abrogation.

Inhibition of Notch1 induced apoptosis through multiple mechanisms

A coordinated process of programmed cell death is involved in both embryonic development and tissue homeostasis in the adult uterus (50) and is an important regulator of uterine function. Multiplicities of neoplastic and transformed cells depend on the Notch pathway for survival (18, 51). Aversion of uterine stromal cell apoptosis is necessary for successful decidualization. We investigated whether a survival-related gene signature is altered in the absence of Notch1 signaling. Following microarray analysis on Notch1^f/f vs. Notch1^d/d, an apoptotic relevance network for Notch1 was predicted, using DAVID bioinformatics and IPA (Supplemental Fig. S2), and multiple molecules were identified as putative Notch1-related gene targets, such as Bad, Bcl2, Bcl2l11, Gzmb, Irf6, and Prf1, which were modulated in the absence of Notch1 signaling.

Heat-map analysis (Fig. 6B) and pathway mapping (Table 1 and Supplemental Fig. S2) demonstrated a relationship between Notch1 and a number of genes in the apoptotic pathway. These data suggest that Notch may work upstream of survival factors and that the prosurvival cascade of Notch1-dependent proteins is altered in Notch1^d/d decidualizing uteri. These studies demonstrate that uterine stromal fibroblasts in Notch1^d/d mice are undergoing apoptosis, in lieu of differentiation and successful decidualization. Concomitantly, the executor of apoptosis, caspase-3, is increased in mice that have decidualization failure (Fig. 8A, B). In addition, Notch1-dependent decidualization failure was associated with a transcriptional decrease in the antiapoptotic gene Bcl2 (Fig. 8C), which is expressed in preparation for uterine implantation (52) and decreases with decidualization (53), and a corresponding increase in the proapoptotic gene Bad (Fig. 8D).

Figure 8.

Functional Notch1 is necessary to avert stromal apoptosis. A) Stromal fibroblasts undergo apoptosis, rather than decidualization, in the absence of Notch1 signaling after artificial decidualization, as demonstrated by Western blot for the executer of apoptosis, caspase-3. B) Quantification of the caspase-3-positive cells (arrowhead: positive caspase-3) from 3 independent experiments that confirm increased caspase-3 activity in Notch1^d/d mice. This finding implies that cells are undergoing apoptosis in the absence of Notch1 signaling. C) Bcl2 serves an antiapoptotic function. Bcl2 expression is markedly decreased in uteri from Notch1^d/d mice that are undergoing apoptosis. Bcl2 transcript levels are 2.4-fold (P=0.0023) lower in the Notch1^d/d uteri. D) In contrast, the proapoptotic gene, Bad, is 1.6-fold higher (P=0.0058) in Notch1^d/d uteri that are undergoing apoptosis.

DISCUSSION

The uterus is a dynamic physiological system in which cellular proliferation, terminal differentiation, and apoptosis occur in a cell-specific manner, regulated by the reproductive cycle or pregnancy. Embryo implantation requires a modified uterine milieu that prevents apoptosis and promotes differentiation of the stromal fibroblast to a specialized secretory cell, called a decidual cell. Elucidating the role of Notch1 contributes to a better understanding of the poorly understood processes that underlie the molecular mechanisms of decidualization, which are essential for the establishment and maintenance of pregnancy in species with a hemochorial placenta (namely, rodents and primates). In addition, this study adds a novel perspective and role to the remarkable diversity in Notch signaling by utilizing a mouse model with a significant hypomorphic phenotype, which adds insight to the Notch1 signaling mechanisms that participate in the formation of the decidua.

Decidualization involves cellular and molecular events that result from tightly regulated interactions among various cell types, multiple signaling mechanisms, growth factors, and cues from the extracellular matrix. This complexity limits conclusions that can be derived from in vitro systems that lack a physiological context. For this reason, we utilized a novel in vivo system to elucidate the mechanisms that underlie decidualization and the transient decidual state, as related to Notch1 signaling in the uterus. Our data identified that Notch1 plays a significant role during decidualization, where the absence of Notch1 inhibits a complete deciduoma in the mouse uterus, as a result of attenuated cell proliferation and cell death. Concomitant with the deceased deciduoma, the decidualization-specific genes Bmp2 and Wnt4 were decreased in the absence of Notch, 1 together with Foxo1 has been shown to be a cAMP-inducible gene in human stromal cells that regulates transcriptional competency during decidualization (54).

Microarray analysis revealed that many decidualization-specific pathways were affected in the absence of Notch signaling. GSEA (33, 34) was used to find whether the biological states present in our experiment have a statistically significant enrichment with any of the sets stored in the MSigB (34). For this analysis, we did not restrict ourselves to the list of differentially expressed genes. Instead, our entire list of probe sets, with their expression data (after normalization and nonspecific filtering), was used.

DAVID bioinformatic analysis revealed that Notch crosstalks with the multiple established decidualization pathways; namely, transforming growth factor-β (TGFβ), bone morphogenic protein (Bmp), and Wnt signaling pathways, which are all signaling pathways necessary for decidualization (55–58). Both Bmp2 and Wnt4, which regulate decidualization and are essential for fertility in mice (41), were significantly down-regulated in the deciduoma of the Notch1^d/d mouse. Previous microarray analysis has demonstrated that ablation of Bmp2 leads to alterations of specific regulator pathways, including the Wnt signaling pathway, PR signaling, and Cox-2 induction, which our microarray correlated to Notch signaling (59). When cell-to-cell contact is made, the initiation of Notch signaling allows TGFβ and Bmp signaling to cooperate with Notch and induce Notch-responsive genes that are elucidated here. In addition, these responses are mediated by progesterone, and our current human data suggest that progesterone plays a crucial role in activating Notch1 (unpublished results).

Significantly, clusters of uterine natural killer (uNK) cell-specific genes, i.e., granzymes, are differentially expressed in our microarray data. Both Notch and Wnt signaling pathways have been independently shown to regulate hematopoietic cell fate decisions between uNK and T-cell phenotypes (60). uNK cells play an important role in decidualization in the murine uterus (61), and our microarray data suggest that Notch1 may modulate uNK signaling and function. Therefore, additional studies need to be conducted to investigate the role of Notch1 signaling in uNK cell function in the context of decidualization and embryo implantation.

Not only is Notch1 signaling essential for the formation of a deciduoma, but Notch-specific genes are responsible for modulating the decidual processes. A variety of mechanisms have been implicated in the Notch prosurvival pathway. The absence of Notch signaling was concomitant with increased apoptosis. Notch1 has been demonstrated to have a dose- and time-dependent effect on the levels of apoptotic inhibitor Bclx (62) and cell cycle regulators p21, Trp53, and p27, (63, 64) and has been shown to induce Bcl2 (65), which are genes affected in our hypomorphic phenotype. Our studies also demonstrated that the ablation of Notch1 results in a marked down-regulation of p21 and cell cycle regulation. It has previously been shown that p21 is a downstream target of Notch1 and can negatively regulate Wnt4 expression (66). In addition, p21 has been shown to bind to E2F, a key transcription factor important for regulating numerous genes critical for cell proliferation, namely Wnt4 (67). Wnt4 has been demonstrated to be a Bmp2-specific downstream target during decidualization. Both Bmp2 and Wnt4 are essential for successful decidualization, and their expression is down-regulated in the similar experimental paradigm that we have utilized (41). As such, we propose that these studies add clarity to the complexity of decidual signaling.

In mice, Notch1 signaling may be necessary for ER- and PR-induced uterine function. Notch, ER, and PR signaling have also been implicated in a handful of cancers, which, similarly to decidualization, involve a multiplicity of cellular signaling pathways and processes (67–69). Rizzo et al. (67) recently demonstrated that in breast cancer cells, a bidirectional crosstalk occurs between ER and Notch1. Estrogen causes accumulation of Notch at the membrane but decreases Notch activation in the absence of progesterone. In turn, Notch1 activates ER-dependent transcription even in the absence of estrogen by physical interaction between the Notch transcriptional complex and ERα on chromatin (70). Our microarray data are consistent with a model in which Notch significantly contributes to ER signaling in the murine endometrium (Supplemental Fig. S3). This finding suggests that the Notch-ER crosstalk identified in breast cancer cells may have a physiological counterpart in the endometrium during implantation. Furthermore, our studies also demonstrate novel hormonal regulation of Notch in a physiological context and suggest that crosstalk between ovarian steroids and Notch signaling may be involved in a variety of physiological processes that require cell fate decisions.

Differentiation to a decidual cell involves dynamic changes in cell function and morphology governed by cytoskeletal reorganization. The microfilaments actin and myosin are the major contractile components in the cytoskeleton. To support the notion that the cytoskeleton is significant during decidualization, our laboratory has demonstrated previously that the cytoskeletal protein, alpha-smooth muscle actin (αSMA) is tightly regulated during decidualization (71–74). Furthermore, cytoskeletal integrity plays a critical role in preventing apoptosis (75–77). Interestingly, Notch has been reported to directly regulate the expression of αSMA by activating a cis element consensus sequence, which is required for Notch-mediated αSMA induction (78, 79). Our bigenic mice harboring a Notch1 deletion demonstrated that actin cytoskeletal integrity is greatly affected and is associated with increased apoptosis seen in these mice (Supplemental Table S2). Seemingly, the regulation of cytoskeletal dynamics is not only critical for the inhibition of apoptosis but also decidualization and that Notch1 signaling in the uterus is essential for this process.

The preimplantation uterus undergoes extensive proliferation and differentiation that results in a decidual phenotype. If implantation is successful, the endometrial stromal compartment forms the decidua, a morphologically and functionally distinct gestation-specific tissue, representing the maternal side of the fetomaternal interface. Although estrogen and progesterone have long been believed to be essential for developing an appropriate endometrial environment for blastocyst implantation, it is now evident that other signals and growth factors further modulate the endometrium. As demonstrated in this study, the expression of Notch1 in stromal cells is intimately associated with decidualization and mediates multiple known signaling pathways crucial for the decidualization process. We demonstrate here that the steroid-primed uterus relies on Notch signaling for cell cycle progression and to rescue stromal fibroblasts from undergoing apoptosis and promote successful decidualization. In addition, array analysis demonstrated that Notch signaling modulates multiple other signaling already established in decidualization and adds perspective to the coordination of signaling modalities that are required for this process (Fig. 9).

Figure 9.

Working model: Notch1 regulates cell fate in endometrial stromal fibroblasts and is essential for successful decidualization and the aversion of apoptosis, which are critical for maintaining endometrial integrity for the successful establishment of pregnancy.