Uterine defects and estradiol‐dependent development of oviductal diverticula in mice lacking the SMAD4 C‐terminal Mad homology 2 domain

In female mammals, the oviduct and uterus are essential sites for female and male gamete transport, fertilization, implantation, and maintenance of a successful pregnancy. To delineate the reproductive function of Mothers against decapentaplegic homolog 4 (Smad4), we specifically inactivated Smad4 in ovarian granulosa cells and, oviduct and uterine mesenchymal cells using the Amhr2‐cre mouse line. Deletion of exon 8 of Smad4 results in the production of an MH2‐truncated SMAD4 protein. These mutant mice are infertile due to the development of oviductal diverticula and defects during the implantation process. The ovaries are fully functional as demonstrated in an ovary transfer experiment. The development of oviductal diverticula occurs shortly after puberty and is dependent on estradiol. The diverticula interfere with sperm migration and embryo transit to the uterus, reducing the number of implantation sites. Analysis of the uterus shows that, even if implantation occurs, decidualization and vascularization are defective resulting in embryo resorption as early as the seventh day of pregnancy. Thus, Smad4 plays an important function in female reproduction by controlling the structural and functional integrity of the oviduct and uterus.


| INTRODUCTION
In 2010, around 48.5 million couples worldwide were infertile. 1 According to the World Health Organization (WHO)'s clinical definition, infertility is 'a disease of the reproductive system defined by the failure to achieve a clinical pregnancy after 12 months or more of regular unprotected sexual intercourse'. 2 In the case of primary infertility, 1.9% of women were unable to have a child. Among women who have at least one child, 10.5% were unable to have another baby (secondary infertility). 1 Infertility in women can have different origins such as ovulatory disorders (24.8%), pelvic adhesions (12.4%), acquired tubal abnormalities (11.2%), bilateral tubal occlusion (10.6%), hyperprolactinemia (6.7%), endometriosis (5.7%), and unexplained (28.6%). 3 Despite the recent progress in reproductive medicine, it is still a long road for couples who attempt to have a baby. Couples with identified or unexplained infertility might experience costly and invasive therapies to optimize the chance of having a baby. 4 To get pregnant, tightly synchronized events must occur from the release of the sperm and oocyte to the implantation and development of the embryo. Once fertilization has occurred in the oviduct, the fertilized egg moves in a specific timing through the oviduct and uterine lumen, where it must interact with the endometrium for implantation. 5 From fertilization to implantation, a cascade of molecular events involving the ovarian hormones and signaling molecules such as cytokines, growth factors, homeobox transcription factors, lipid mediators, and morphogens plays an important role in guiding the embryo to its implantation site. 6,7 Any disorder alongside this female reproductive axis may lead to infertility. 8,9 It is therefore important to understand how the oviduct and uterus develop and function.
The transforming growth factor-β (TGFβ) superfamily including the ligands TGFβ, activins, inhibins, bone morphogenetic proteins (BMPs), growth differentiation factors (GDFs), anti-Müllerian hormone (AMH), and nodal growth differentiation factor (NODAL) is involved in cell proliferation, differentiation, morphogenesis, tissue homeostasis, and regeneration. 10 Ligands activate the pathway through binding to the type I-type II receptor complex, which phosphorylates the receptor-regulated SMAD proteins (R-SMADs: SMAD2/3 for TGFβ, activins and nodal, and SMAD1/5/8 for BMPs, GDFs, and AMH). 11,12 Then, phosphorylated R-SMADs will form a complex with the common SMAD (co-SMAD or SMAD4), which is translocated to the nucleus and acts as a transcriptional regulator complex. 12,13 In the TGFβ and BMP pathways, mutations in one of these factors (ligand, receptor or SMADs) have been described and linked to cancer and genetic disorders. [14][15][16][17] In reproduction, TGFβ superfamily signaling is involved in follicular development, ovulation, oocyte competence, oviduct integrity, uterine development, decidualization, implantation, pregnancy, and embryonic development. [18][19][20][21] While much data are available on the role of the different TGFβ superfamily signaling ligands, receptors, and R-SMADS in reproduction, 18 the function of the common SMAD, SMAD4 is not fully understood.
Smad4-null mutant mice have been generated and died around embryonic day (e) 7.5 before gastrulation. [22][23][24] Therefore, to delineate the function of Smad4 during development and organogenesis, two mouse lines for Cre-recombinase-driven conditional inactivation (cKO) of Smad4 have been generated by floxing either exon 1 (Smad4 RobCA ) 25 or exon 8 (Smad4 tm2.1Cxd ). 26 Although the embryonic phenotype of KOs derived from these two floxed lines, Smad4 RobN (Δexon1) 25 and EIIa +/cre ; Smad4 tm2.1Cxd (Δexon8), 26 resembles that of previously generated Smad4-null mutations, 22,24 it is noteworthy that deletion of exon 8 leads to the generation of a Mad homology 2 (MH2) domain-truncated SMAD4. Using Smad4 RobCA mice 25 and two Cre-expressing lines, SMAD4 has been shown to prevent ovarian granulosa cells from entering premature luteinization 27 and to play a role in LH-induced cumulus expansion, ovulation, and luteinization. 28 To study the function of SMAD4 during the Müllerian duct regression and the reproductive process, we have crossed Smad4 tm2.1Cxd mice (Δexon8) 26 with Amhr2 tm3(cre)Bhr mice. 29,30 The Amhr2 (AMH receptor type 2) promoter allows the expression of the Cre recombinase in the mesenchyme of the Müllerian ducts, fetal and adult ovaries and testes, oviducts, and uteri postnatally. 29,31 Unlike the Amhr2 +/cre ; Smad4 flox/− mice (Δexon1) which did not display any phenotype during Müllerian duct regression or oviduct/uterus function, 27 our Smad4 cKO mice present a partial Müllerian duct retention in male 30 and a female infertility due to defective functions of both the oviduct and uterus. This phenotypic discrepancy could be explained by the presence of an MH2-truncated SMAD4 (p.Ala318Glyfs*4) in our Smad4 cKO mouse model.

| Generation of Smad4 conditional knockout mice
Studies involving animals, including housing and care, method of euthanasia, and experimental protocols were conducted in accordance with the recommendations of the French Accreditation of Laboratory Animal Care. The French Ministries of Agriculture and Research delivered a license to our animal facility (E35-238- 19) and approved the project R-2012-FaPe-01, respectively. Experimental animals were maintained on a C57BL/6J-129/SvEv mixed genetic background. Smad4 tm2.1Cxd or Smad4 f/f mice 26 and Amhr2 tm3(cre)Bhr mice (generously provided by R.R. Behringer (M.D. Anderson Cancer Center, Houston, Texas)) 29 were used to generate the Amhr2 +/cre ; Smad4 f/f mice (Smad4 cKO, noted Smad4 Δ/Δ ). Smad4 f/f littermates were used as controls. Using the primers listed in Table S1, mice were genotyped twice, at weaning and euthanasia, from tail genomic DNA using standard PCR protocols as described. 26,29

| Oviductal dye injection
As previously reported, 32 oviducts of Smad4 f/f and Smad4 Δ/Δ mice were dissected out and a blue dye (0.25% bromophenol blue, 0.25% xylene cyanol, and 15% Ficoll type 400 in deionized water) was injected into the oviductal lumen through either the infundibulum or the uterotubal junction using a 30-gauge needle.

| Estradiol treatment and cyst measurement
Ninety-day release hormone pellets were purchased from Innovative Research of America, Sarasota, FL. For estradiol (E 2 ) treatment, 4 weeks old Smad4 f/f and Smad4 Δ/Δ female mice were ovariectomized. At 7 weeks, these mice received either an E 2 pellet (0.05 mg/pellet) or a control pellet (placebo for E 2 or P-E2) implanted subcutaneously into the neck. Ninety-one days after the pellet implantation, the oviducts were collected, gently uncoiled, and placed in PBS between a microscope slide and a coverslip. Cysts were observed and measured under a stereo microscope Zeiss SteREO Discovery.V8 connected to an AxioCam ICc3 camera. Using Axio Vs40 V 4.8.2.0 software tools (Zeiss), the surface area of flattened cysts was measured and expressed in μm 2 ( Figure S1).

| 2-cell stage embryo recovery
Ten to twelve weeks old Smad4 f/f and Smad4 Δ/Δ female mice were mated with stud C57BL/6NRj male mice. Oviducts were dissected out on day 2 at 3:00 PM (day of vaginal plug is day 1 of pregnancy) and then flushed through the infundibulum using a 30-gauge needle with warmed M2 medium (Sigma). Embryos were collected, characterized for cell number, and counted.

| Induction of decidualization
Artificial decidualization was induced as previously described. 33 Briefly, 9-to 11-week-old mice were ovariectomized. Two weeks after ovariectomy (day 1), mice were treated daily with 100 ng E 2 in sesame oil (days 14 to 16) and then treated with a mixture of 6.7 ng E 2 and 1 mg progesterone (P 4 ) in sesame oil (days 19 to 21). Six hours after the last injection, one uterine horn was traumatized by injection of 50 μL of sesame oil into the lumen. The contralateral horn was not induced and served as a control. Mice were injected with the E 2 /P 4 mixture from day 22 to day 25. Mice were sacrificed on day 26 and the wet weights of their traumatized and control uterine horns were measured. E 2 , P 4 , and sesame oil were purchased from Sigma.

| Vascular permeability assays
Nine to eleven weeks old Smad4 f/f and Smad4 Δ/Δ female mice at gestational day 5 (GD5) were injected intravenously with 100 μL of 1% Chicago Sky Blue 6B (Sigma). After 15 min, mice were euthanized by CO 2 inhalation and the reproductive tract was collected. As previously described, 34 the Chicago Sky Blue 6B dye was extracted from the isolated implantation sites with 70 μL of formamide overnight at 55°C and absorbance was measured at 600 nm. 2.7 | RNA isolation and microarray experiment 2.7.1 | RNA preparation Uteri from 9-11 weeks old female mice at GD6 were collected into RNAlater reagent (Qiagen) prior to RNA extraction. Oviducts were isolated from 8-9 weeks old female mice at pseudoGD1. Respectively, total RNA from oviducts and uteri were isolated using the PicoPure RNA isolation kit (Arcturus) and the RNeasy Plus Mini kit (Qiagen) according to the manufacturer's instructions. Quantity and quality of the total RNA were evaluated using a Nanodrop 8000 (Thermo Fisher Scientific, Wilmington, DE) and a 2100 Bioanalyzer (9.2<RIN>9.9; Agilent, Santa Clara, CA, USA).

| Microarray analysis
Using the Ambion WT Expression kit (Thermo Fisher), 250 ng of total RNA was processed to synthesize cRNA targets. Quality control was performed at each step of the process. After labeling and fragmentation using the GeneChip WT Terminal Labeling and Controls kit (Affymetrix, Santa Clara, CA, USA), 3.75 μg of single strand cDNAs were hybridized with a GeneChip Mouse Gene 2.0 ST Array (Affymetrix) at 45°C for 16 h in a GeneChip Hybridization Oven 645 (Affymetrix). After hybridization, the array chips were stained and washed using a GeneChip Fluidics Station 450 (Affymetrix). Then, image (.DAT) files were preprocessed using the Affymetrix GeneChip Command Console (AGCC) software v.4.0 to generate cell intensity (.CEL) files. The resulting CEL files were processed using the oligo package available performed in R/Bioconductor. 35 Data were then normalized and background corrected using the SCAN. UPC package 36 with the Brainarray custom CDF file for directly mapping Affymetrix probe to Entrez gene identifiers (mogene20st version 21.0.0). 37

| Microarray
The statistical filtration of the genes differentially expressed (DE) between control and wild-type samples was performed using the Annotation, Mapping, Expression and Network (AMEN) suite of tools. 38 Briefly, genes showing a signal higher than a given background cutoff (overall median of the normalized data set) and at least a 1.1-fold change between both experimental conditions were selected. To define a set of differentially expressed genes displaying significant statistical changes across comparisons, the empirical Bayes moderated t-statistics was performed using the limma package (F-value adjusted using the Benjamini & Hochberg (BH) False Discovery Rate approach, p ≤ .05 39 ). The resulting genes were then partitioned into two expression clusters corresponding to transcriptionally repressed and induced genes in the mutant mice.

| Functional analysis
The enrichment analysis module implemented in AMEN was used to identify gene ontology terms (biological processes, molecular functions, and subcellular components) significantly associated with each expression cluster by calculating Fisher's exact probability using the Gaussian hypergeometric function (FDR-adjusted p-value ≤.01 number of genes in a given group associated with a given annotation term ≥5).

| Real-time PCR
For real-time PCR, the first strand cDNA was synthesized using 250 ng of total RNAs and the iScript cDNA Synthesis Kit (Bio-Rad). Then, 5 ng of cDNAs was processed for realtime PCR using the iTaq Universal SYBR Green Supermix kit (Bio-Rad) and the CFX384 Touch Real-Time PCR Detection System (Bio-Rad). The primers (300 nM) used are listed in Table S1. The PCR protocol is as follows: denaturating step at 95°C for 3 min followed by 40 cycles of 95°C for 10 s, 60°C for 30 s. Standard curves were generated using different concentrations of an oviduct or uterine cDNA template. Relative gene expression was normalized using two reference genes (Ribosomal Protein Lateral Stalk Subunit P0 (Rplp0) and Ribosomal protein L13a (Rpl13a) for the oviduct, and Actin Beta (Actb) and Rplp0 for the uterus) with CFX Maestro software (Bio-Rad v2.3).

| Histology, immunohistochemistry (IHC), immunofluorescence (IF), and lectin staining
Reproductive tracts were collected from female mice at 5 weeks, 8-10 weeks, and 20-25 weeks in PBS and visualized under a stereo microscope Zeiss SteREO Discovery. V8 connected to an AxioCam ICc3 camera using the Axio Vs40 V 4.8.2.0 software (Zeiss). All samples were fixed in 4% PFA in PBS at 4°C overnight. Tissues were dehydrated and embedded in paraffin. Five μm sections were hydrated and then used for either hematoxylin/ eosin (Leica Microsystems) staining, IHC or IF as previously described. 30 The primary antibodies used are listed in Table S2. For the fluorescein isothiocyanate (FITC)conjugated lectin (Lycopersicon esculentum (tomato) from Sigma) staining on uterine tissues, the procedure was adapted from a previously described study. 40 Briefly, sections were hydrated before applying an antigen unmasking solution (H3300, Vector Laboratories) for 20 min at 100°C. Section were washed three times with PBS and a blocking solution (3% BSA in PBS) was applied for 60 min. Five μg/ml of a FITC-conjugated lectin was incubated for 2 h at RT. After three washes of PBS, slides were mounted with UltraCruz Hard-set Mounting Medium with DAPI (Santa Cruz Biotechnology).

| Amhr2 +/cre ; Smad4 f/f female mice develop oviductal diverticula
To specifically investigate the function of Smad4 in the female reproductive tract, we have analyzed our previously reported Amhr2 +/cre ; Smad4 f/f mouse model (noted Smad4 cKO or Smad4 Δ/Δ ). 30 To discriminate the expression pattern of the full-length SMAD4 from that of the MH2-truncated SMAD4 (or SMAD4ΔMH2), an immunohistochemistry (IHC) was performed using anti-SMAD4 (directed against the whole protein) and anti-SMAD4 Ct (directed against an epitope located within the MH2 domain) antibodies, respectively. Using the anti-SMAD4 Ct antibody on 3-week-old control mouse (Smad4 f/f ) tissue sections, SMAD4 is detected in all epithelial and mesenchymal cells of the oviduct and uterus ( Figure 1A,D). As expected, in Smad4 Δ/Δ mice, the anti-SMAD4 Ct antibody detects the whole protein only in the epithelial compartment of the oviduct and uterus ( Figure 1B,E). Moreover, the anti-SMAD4 antibody allowed the detection of a protein in all compartments of the oviduct and uterus of the Smad4 Δ/Δ mice ( Figure 1C,F). This result confirms that cre-targeted cells in Smad4 Δ/Δ are of mesenchymal origin and express an MH2-truncated SMAD4.
Anatomical observation of the reproductive tracts reveals no obvious differences between controls and mutants ( Figure 1G,K). Histological analysis of 10-week-old Smad4 Δ/Δ mice in estrus shows that overall the ovary, oviduct, and uterus appear normal compared with Smad4 f/f mice ( Figure 1H-J,L-N). Nevertheless, in mutant mice, one can note the presence of small oviductal diverticula, which are not consistently observed before 10 weeks of age ( Figure 1M).

| Smad4 Δ/Δ female mice are infertile
To assess the reproductive capacity of the Smad4 Δ/Δ mice, we then mated twelve 8-week-old mutant female mice with stud C57BL/6NRj (wild-type or wt) males for a period of six months. Forty-eight wt, Amhr2 +/cre or Smad4 f/f female mice were used as controls. As shown in Figure S2A, Smad4 Δ/Δ female mice were unable to have pups, whereas control females gave birth to 771 pups in 119 litters, with an average litter size of 6.5 pups. The onset of puberty, assessed by the timing of vaginal openings, did not show any significant differences (Smad4 f/f , 31.9 ± 3.8 (SD) days, n = 13; Smad4 Δ/Δ , 32.9 ± 4.3 days, n = 13). Analysis of vaginal smears over a four weeks period revealed no irregularities in the estrous cycle of 8-week-old mutant mice. Moreover, a vaginal plug was regularly observed during matings with wt males. All these data suggest that endogenous estradiol (E 2 ) production was not perturbed in mutants. 41,42 Since Smad4 is inactivated in Smad4 Δ/Δ ovaries, oviducts, and uterus, we performed ovary transfers between wt and transgenic female mice (control Amhr2 +/cre or Smad4 f/f , and mutant Smad4 Δ/Δ ) to identify the organ(s) responsible for this infertility. After mating with stud wt males, we observed that wt females receiving control or mutant ovaries gave birth to offspring from the donor ovaries and that the number of pups per litter was similar between control and mutant ovaries (Amhr2 +/cre , 5.0 ± 2.1 (SD), n = 6; Smad4 Δ/Δ , 5.0 ± 2.7, n = 9; Figure S2B). However, mutant mice with transplanted wt ovaries were infertile despite the fact that copulatory plugs were observed. Meanwhile, control mice with transplanted wt ovaries gave birth to wt pups as expected ( Figure S2B). These observations indicate that mutant ovaries are fully functional, whereas the oviducts and/or uterus are responsible for reproductive failure.
We wondered whether heterozygous Smad4 +/Δ female mice might also have a reproductive defect. Both Smad4 +/f (control) and Smad4 +/Δ female mice are fertile with similar litter size ( Figure S3A). Gross anatomy of reproductive tracts revealed no defects and careful observation of uncoiled oviducts showed no diverticula ( Figure S3B-E). Thus, the presence of a single Smad4 allele is sufficient to maintain the reproductive functions.

| Oviductal diverticula in Smad4 Δ/Δ mice consist of all oviductal cell types
At 8-10 weeks (Figure 2A,B), close examination of mutant oviducts revealed tiny cysts less than 200 μm in diameter ( Figure 2B). After 20 weeks, a few small cysts (<200 μm in diameter) were observed in control oviducts ( Figure 2C), while diverticula are larger (often more than 1 mm in diameter) and numerous in mutants ( Figure 2D). To study the diverticulum ontogeny, we next analyzed the oviducts of control and mutant mice aged 3 to 38 weeks ( Figure 2E-J and Table S3). Under 7 weeks old ( Figure 2E,F and Table S3), no diverticulum was observed in control and mutant mice. At 8 weeks, the first diverticula appear near the utero-tubal junction ( Figure 2H) in 77.8% of mutants but not in controls (Table S3). They then become more numerous and larger along the isthmus portion when the mice are older ( Figure 2H,J-L; average size of 28 089 ± 7473 μm 2 (SEM, n = 21) at 10 weeks and 80 327 ± 11 383 μm 2 (n = 132) at 20 weeks). Importantly, 27.3% of 20-week-old control mice had rare tiny oviductal diverticula (9445 ± 5806 μm 2 , n = 5; Figure 2I,L and Table S3). At 1 year, the tubular structure of mutant oviducts is indistinguishable as diverticula are present all over the isthmus and are gigantic (between 2-and 6 000 000 μm 2 , Figure S4F-J). In control mice, the diverticula also continue to develop but to a lesser extent than in mutants (around 500 000 μm 2 , Figure S4A-E).
The oviduct is formed by a pseudostratified epithelium consisting of ciliated and secretory cells, a thin layer of stroma, and inner circular and outer longitudinal smooth muscle layers. 43 In Smad4 Δ/Δ female mice, the wall of the diverticula appears thinner than that of the oviduct ( Figure 3B-D). As in the control oviduct ( Figure 3A), the epithelial cell marker cytokeratin-8 (KRT8) and smooth muscle cell marker αSMA are expressed in newly developed or well-formed diverticula ( Figure 3B′-D′). Therefore, in oviductal diverticula, both simple epithelium and smooth muscle layers are represented but appear much thinner as if they were stretched out ( Figure 3B ( Figure S5M-O). Ciliated epithelial cells, mainly present in the infundibular and ampulla epithelium, were also observed by immunostaining with acetylated tubulin in the diverticula ( Figure S5P-R). Pax8, a marker of secretory epithelial cells abundantly present in the isthmus, is also expressed in diverticula ( Figure S5S-U). Interestingly, the few diverticula present in one year old Smad4 f/f control mice have the same structure as those in the Smad4 Δ/Δ mice ( Figure S5). Altogether, these results demonstrate that diverticula develop after puberty and contain all cell types of the oviduct.

| Development of oviductal diverticula in Smad4 Δ/Δ mice is dependent on E 2
Because diverticula begin to develop after puberty, we wondered whether E 2 plays a role in triggering this phenotype. Four-week-old Smad4 f/f and Smad4 Δ/Δ mice were ovariectomized before the onset of puberty and diverticulum formation ( Figure 4A). At 7 weeks (at about the time of diverticulum appearance), mice were implanted with a placebo pellet or a 90-day release pellet of 0.05 mg E 2 . At 20 weeks, the oviducts were collected. In three out of five ovariectomized Smad4 Δ/Δ mice receiving a placebo pellet (P-E2), diverticula were absent, while the two others only showed five tiny diverticula (6459 ± 1119 μm 2 (SEM, n = 5), Figure 4C,F). Therefore, the absence of ovaries leads to a dramatic reduction in diverticula development. Mutant mice subjected to continuous exposure to E 2 displayed multiple diverticula with an average size of 26 804 ± 4715 μm 2 (n = 57; Figure 4E,F), which is similar to what was observed in 10-week-old mutants ( Figure 2H,K). One can note that placebo or E 2 exposure did not affect the oviductal morphology in control mice ( Figure 4B,D).

| Well-developed diverticula in
Smad4 Δ/Δ mice perturb the oviductal transport of gametes and embryos leading to a reduced fertility To mimic the path of sperm and embryos, we injected a blue dye solution through the utero-tubal junction or the infundibulum, respectively. In 20-week-old Smad4 f/f oviducts, the dye is dispersed through the entire oviductal lumen highlighting the tortuous path of the oviduct ( Figure 5A). When injected through the infundibulum ( Figure 5B) or the utero-tubal junction ( Figure 5C) of Smad4 Δ/Δ oviducts, the dye followed the lumen and entered the individual cysts as they came on-stream. Therefore, diverticula are directly connected with the oviductal lumen. In 8-week-old mutant oviducts without visible cysts, the dye highlighted the lumen as in controls ( Figure 5A,D).
We next mated Smad4 f/f and Smad4 Δ/Δ female mice with stud wt males and analyzed the embryonic implantation efficiency at gestational day (GD) 6 or 7 in the presence or absence of oviductal cysts. In 10-week-old control and mutant mice, a similar number of implantation sites is visible ( Figure 5E,F), indicating that the presence of small diverticula on the oviducts does not appear to interfere with gamete and embryo transport through the oviductal lumen ( Figure 5G,H). However, at 20 weeks, Smad4 Δ/Δ mice with significantly larger oviductal cysts had fewer implantation sites than Smad4 f/f mice ( Figure 5I-L). As shown in Figure 5M  no implantation site (1.8 ± 0.7 implantation sites (n = 11) versus 9.4 ± 1.1 (n = 5) for the controls). Thus, while control mice had a comparable number of implantation sites over the course of their reproductive lives, Smad4 Δ/Δ mice showed a significant age-dependent reduction in the number of implantation sites. To determine whether the reduced number of implantation sites is linked to a reduced fertilization rate in 20-week-old Smad4 Δ/Δ females, which in turn could be related to a disruption in gamete tubal transport, we collected embryos at GD2 (embryos should be mainly at the 2-cell stage). As shown in Figure 5N, 41.4% of embryos were at the 2-4-cell stage in Smad4 f/f mice while only 4.9% had reached that stage in Smad4 Δ/Δ mice. This observation suggests that contact between sperm and oocytes is more challenging in the presence of diverticula. Indeed, in metestrus, whereas oocytes can be seen in the ampulla of the Smad4 f/f oviduct ( Figure 5O), Smad4 Δ/Δ oviductal cysts are filled with white debris (Figure 5P), which turned out to be accumulated degraded oocytes ( Figure 5Q). Flushing of oviducts at GD2 revealed the presence of 2-cell embryos in control mice, whereas in mutant mice, the majority of oocytes are degraded ( Figure 5R). Overall, these data suggest that the migration of gametes and embryos into the oviductal lumen is compromised by the presence of well-developed diverticula, which then act as traps.
Since the expression of smooth muscle cell markers appears to be affected in diverticulum-developing oviducts in dicer, 32 Smad1/5/4 46 or TGFβ type 1 receptor (Tgfbr1) 47 mutant mice, we performed real-time PCR on Calponin 1 (Cnn1), Desmin (Des) and Myosin heavy chain 11 (Myh11) ( Figure S6). Assessment of the expression of these few markers at the mRNA level was inconclusive, as no significant transcriptional differences between control and mutant were detected. The same observation was also reported in the Yap/Taz cKO. 48 Nevertheless, Smad4 expression was significantly reduced in the mutant isthmi ( Figure S6).

| Infertility in Smad4 Δ/Δ female mice is due to uterine defects
Since embryo implantation occurs in 8-11-week-old Smad4 Δ/Δ mice ( Figure 5F), infertility should be explained by a post-implantation defect. Therefore, we next mated 9-10-week-old Smad4 Δ/Δ females with stud wt males and implantation sites were examined at different GDs ( Figure 6A-D). Sites of resorption were observed as early as GD7 and the phenotype appeared to worsen from GD8. By GD12, the implantation sites were completely resorbed with the uterine horns filled with blood and debris ( Figure 6D). Histological analysis of implantation sites at GD6 revealed that epithelium fusion did not occur in mutant mice as it does in controls ( Figure 6E,F,E′,F′). Moreover, blood cells were present in the lumen suggesting that resorption had begun ( Figure 6F′). Since the Amhr2 promoter-driven cre targets the mesenchymal cells of the endometrium and the smooth muscle cells of the myometrium, we analyzed these compartments using immunohistology. The use of the smooth muscle marker αSMA revealed that the inner circular smooth muscle layer, which had a regular shape in controls, was partially discontinued in mutants ( Figure 6G,H). Observation of the luminal and glandular epithelium using cytokeratin 8 revealed no obvious defect in mutant uterus compared to controls ( Figure 6I,J).

| Decidualization and vascular permeability are reduced in Smad4 Δ/Δ female mice
After the attachment of the blastocyst to the uterine epithelium, the uterine stroma undergoes a process of proliferation and differentiation called decidualization, which is critical for the proper development of the embryo. 49 In 10-week-old control and mutant mice, we analyzed the decidual response of steroid-primed mutant uteri to artificial stimulation ( Figure 7A). A strong decidual response of the stimulated Smad4 f/f uterine horn was observed, whereas the stimulated Smad4 Δ/Δ uterine horn showed a significantly reduced response ( Figure 7B-D).
Once the blastocyst has initiated a contact with the receptive uterus, the decidualization process is associated with an extensive development of the vascular network. 49 Moreover, an increase in vascular permeability around the implanted blastocyst can be observed by injection of a plasma protein-binding macromolecule such as Chicago Sky Blue 6B dye. Therefore, 8-to 10-week-old Smad4 f/f and Smad4 Δ/Δ female mice were injected with blue dye on the morning of GD5 and implantation sites were monitored by the presence of bluing areas ( Figure 7E,F). By spectrophotometric analysis of dye extravasation around the implantation sites ( Figure 7G), vascular permeability was reduced by 27.7% in mutants compared with controls. In addition, Lycopersicon esculentum lectin (LEL), which specifically binds to the carbohydrate N-acetyl-D-glucosamine (GlcNAc) lining the endothelial layer 50 showed that the vascular network in the artificially decidualized uterus was poorly developed in mutants compared with controls ( Figure 7H,I).
To further investigate the role of SMAD4ΔMH2 during the implantation process at the molecular level, gene expression profiling was performed on GD6-uterine samples collected from 8-10 weeks old control and mutant female mice mated with stud wt males. The transcriptomic analysis revealed that 483 genes were repressed and 267 genes were induced in the Smad4 Δ/Δ female mice ( Figure 8A,B and Table S4). Among all these genes, factors involved in decidualization and/ or angiogenesis came to our attention. One group encompassed key factors in uterine stromal differentiation or decidualization such as alkaline phosphatase (Alpl), prolactin-like protein type B (PLP-B or Prl6a1), decidual/trophoblast prolactin-related protein (d/tPRP or Prl8a2), Bone Morphogenetic Protein 2 (Bmp2) and wingless-related MMTV integration site 4 (Wnt4). A second group included angiogenesis-promoting factors or factors involved in vascular remodeling, such as angiopoietin 2 (Angpt2), angiopoietin-4 (Angpt4), Sushi repeat-containing protein X-linked 2 (Srpx2), Endothelial PAS domain-containing protein 1 (Epas1 or Hif2a) and Killer cell lectin-like receptor subfamily G member 1 (Klrg1). Finally, a third group consisted of factors involved in E 2 biosynthesis or retinoid signaling and known to be essential for decidualization and/or angiogenesis such as Cholesterol side-chain cleavage enzyme (Cyp11a1 or P450scc), hydroxy-delta-5-steroid dehydrogenase, 3 beta-and steroid delta-isomerase 6 (Hsd3b6), steroidogenic acute regulatory protein (StAR), cellular retinoic acid binding protein I (Crabp1) and Cyp26a1. As it was observed for the microarray analysis, real-time PCR analysis revealed a decreased gene expression of Alpl, Angpt2, Bmp2, Crabp1, and Wnt4 in the GD6 mutant uteri ( Figure 8C). An increase in gene expression was confirmed for Cyp26a1, Klrg1, Srpx2, and Wnt16, a factor expressed in uterus during the peri-implantation period 51 ( Figure 8C). Altogether, these results show that, in Smad4 Δ/Δ mice, decidualization as well as uterine vascular development are affected and thus compromise the proper implantation and blastocyst development.

| DISCUSSION
Analysis of our Smad4 Δ/Δ mouse model revealed important new reproductive functions for SMAD4. Strikingly, we observed that Smad4 Δ/Δ female mice are infertile due to a reduced decidualization response and an endometrial vascular insufficiency. Moreover, the development of diverticula on the isthmic portion of the oviduct after puberty interferes with sperm and embryo transport, resulting in decreased fertilization and implantation rates ( Figure S7). However, unlike what was observed in the Amhr2 +/cre ; Smad4 f/− mouse model (Δexon1), 27 premature luteinization of ovarian follicles was not observed in our mice. Conversely, no oviduct and/or uterine phenotypes were reported in Pangas' mouse model. Moreover, our Smad4 Δ/Δ male mice displayed partial retention of the Müllerian duct, 30 which was not described in other Smad4 mutant mice to date. Since both mouse models use the same Amhr2-cre mouse line, these phenotypic differences could be related to the generation of an MH2-truncated SMAD4 protein (p.Ala318Glyfs*4) in our mouse model. However, unlike the Amhr2 +/cre ; Smad4 f/− mouse model, 27 the absence of phenotype in ovaries in our mutant mice could be attributed to the mosaic deletion of Smad4 since two floxed alleles are targeted by a cre recombinase under the control of Amhr2 promoter. 52 Interestingly, analyses of numerous mutant mice homozygous for Smad4, generated by different genetic strategies, revealed a role for Smad4 during early embryonic development (Table S5). [22][23][24][25][26]53 These observations suggest that alterations in the MH1 or MH2 domains, or complete inactivation of SMAD4 are sufficient to severely impact SMAD4 functions and result in embryonic lethality. Nevertheless, it has been described that C-terminal truncated SMAD4 protein can exhibit dominant negative properties by interacting with other SMADs, [54][55][56] or are unstable and rapidly degraded. 56,57 Furthermore, these Cterminal truncations do not encompass the whole MH2 domain, which is necessary for interacting with other factors. 58,59 Therefore, in our mouse model expressing a fully MH2-truncated SMAD4, we unveiled novel in vivo functional properties of SMAD4ΔMH2. This result is of great interest since missense or nonsense mutations leading to an inactive SMAD4 protein often occur in the MH2 domain (Table S6). 59,60 Development of diverticula in the oviduct has been well documented in mice carrying disrupted genes such as dicer, 32,61,62 Smad1/5/4, 46 TGFβ type 1 receptor (Tgfbr1) 47 and Yap/Taz. 48 In contrast to these different studies, however, the oviductal diverticula in our Smad4 Δ/Δ mice develop after puberty (around 8 weeks of age) and appear to be E 2 dependent. The oviduct, the site of fertilization and early embryonic development, plays a crucial role in successful implantation. 44,45 Tubal transport of gametes and embryos is important for the initiation of a pregnancy and depends on the proper functioning of three components, ciliary motion, muscular contractility and tubal fluid. 44,45 Prior to the appearance of the first diverticula in Smad4 Δ/Δ oviducts, implantation rates and fertility were not affected, suggesting that the oviduct appears to be fully functional. Histological analyses of the diverticula at different stages of development revealed that all epithelial and smooth muscle cells were present and that the diverticula were directly connected to the oviductal lumen. The continuous development of diverticula was not due to increased cell proliferation since phosphorylated histone H3 expression was not affected ( Figure S8). Moreover, immunostaining of smooth muscle and epithelial cells showed a single layer of flattened cells. Therefore, the growth of the diverticulum causes the existing cells to stretch until they are barely visible. In these 'inflated balloon' like-structures, the epithelial surface area and lumen volume increase such that the epithelial functions of ciliated cells must be impaired. As a result, gametes and embryos that follow the tubal flow are trapped into the cysts where fluid movement is disrupted. Among the pathologies found in infertile women, tubal disease represents 25-35% of the causes. Subtle distal fallopian tube abnormalities, which include fimbrial agglutination, tubal diverticula, tubal accessory ostium, fimbrial phimosis and accessory fallopian tube, are observed in 28.7% of infertile women. 63 Tubal diverticula have been found in about 2% of infertile women, 64,65 and have a histology similar to that of normal fallopian tubes. [65][66][67] When a hysterosalpingography is used to examine infertile patients, tubal diverticula are often misdiagnosed as distal tubal occlusion such as hydrosalpinx. [68][69][70] Laparoscopic evaluation appears to be a more accurate method of differentiating tubal diverticula from various subtle distal fallopian tube abnormalities. 70 Indeed, a recent laparoscopic analysis in a large cohort study revealed that tubal diverticula accounted for 7.5% of infertile women (6.4% unilateral and 1.1% bilateral). 63 Since cases of tubal diverticula are rare, it is difficult to correlate the presence of diverticula with infertility. Nevertheless, some observations suggest a link between tubal diverticula and infertility, endometriosis or ectopic pregnancy. 63,65,67 For example, in a study of 13 patients with tubal diverticula, 11 (84.6%) had endometriosis. 65 To date, laparoscopic treatment of tubal diverticula is the most effective approach, and after surgery, the natural pregnancy rate is 60%. 71 In an attempt to delineate the molecular pathways involved in the development of these oviductal diverticula, qPCR analyses were performed. In contrast to some previous studies showing a change in the expression of genes involved in smooth muscle cell differentiation, 32,46,47 our results do not show differences between controls and mutants, as has also been shown for Yap/Taz cKO. 48 We hypothesize that molecular changes along the oviduct could occur locally, at the site of cyst development, and are therefore minimized when the entire oviduct is exploited. Recent studies have revealed the contribution of smooth muscle in shaping the epithelium folds of several organs including the oviduct. 72,73 Moreover, E 2 does not have a cell proliferative effect on the isthmus but most likely acts by changing the gene expression and cell morphology. 74 Therefore, in our Smad4 Δ/Δ mouse model, the stiffness of the oviductal tube could be locally compromised by the action of E 2 , so that the smooth muscle layer can no longer maintain the structural integrity of the epithelial lining. The epithelium is thus projected outwards to form diverticula and then loses its folded structure. With each estrous cycle, E 2 could act on new areas of the oviduct leading to the development of more and larger cysts. Loss of folds in fallopian tubes has been correlated with lack of pregnancy. 75 After fertilization, the zygotes develop and migrate into the oviductal lumen before reaching the uterus where the implantation process takes place. During the window of receptivity, the closure of the uterine lumen occurs and allows the blastocyst to establish a close contact (defined by three stages: apposition, adhesion and attachment) with the luminal epithelium. 76 Upon blastocyst attachment, the endometrium undergoes morphological and functional changes, and the vascular network is remodeled. 49,76 All these steps are critical for the proper embryo development and maintenance of the pregnancy. In Smad4 Δ/Δ mice at GD6, the implantation process is compromised as the luminal epithelium does not close around the blastocyst. A similar phenomenon has been described in the Smad1/5/4 cKO mouse line 46 but not in the single Smad1, Smad5, or Smad4 cKO mice. 27,77 This observation raises the possibility that SMAD4ΔMH2, in addition to a loss of function of Smad4, acts as a dominant negative by interfering with the Smad1/5 pathway. In rodents, the process of luminal closure is facilitated by progesterone (P 4 )-mediated absorption of uterine luminal fluid through the action of key epithelial Cl − /Na + channel proteins such as cystic fibrosis transmembrane conductance regulator (CFTR) and epithelial Na + channel (ENaC). P 4 induces ENaC expression, but represses CFTR, contributing to the fluid absorption prior to the blastocyst attachment. 78 In our microarray analysis, CFTR or ENaC mRNA expression was not affected but Serine/threonine-protein kinase Sgk1 (Sgk1) mRNA, an activator of ENaC, was downregulated in Smad4 mutant mice (Table S4), suggesting that uterine fluid uptake might be reduced and thus the implantation process compromised. Coinciding with the blastocyst attachment to the luminal epithelium, a dramatic increase in the endometrial vascular permeability is observed through the accumulation of macromolecule-dye complexes at the implantation sites. 79 Interestingly, at GD5, vascular permeability was reduced in Smad4 Δ/Δ mice suggesting that the vascular network necessary for the successful implementation of the gestation is defective. After the attachment step, stromal cells enter an extensive phase of proliferation and differentiation, called decidualization, which is reduced in Smad4 Δ/Δ mice. We then observed that a key factor for the regulation of stromal cell differentiation, such as Bmp2, 80 was decreased in Smad4 Δ/Δ mice. Wnt4, a BMP2-regulated gene involved in endometrial stromal cell survival and differentiation, 81 and the decidual cytokines PLP-B and d/tPRP 82,83 were also downregulated in Smad4 Δ/Δ mice. Remodeling of the maternal uterine vascular network is a prerequisite for successful implantation and pregnancy. 84 In Smad4 Δ/Δ mice, endometrial vascular development is not complete, which may be a consequence of downregulation of angiogenic factors such as Angpt2, Angpt4, and Epas1. In mice, the letrozole inhibition of aromatase in in vitro decidualized stromal cells revealed the importance of estrogen produced by the decidual uterus in promoting decidualization and neovascularization. 85 This inhibition of aromatase leads to the downregulation of decidual and angiogenic factors, such as Alpl, d/tPRP, Bmp2, Epas1, Angpt2, and Angpt4 that are also downregulated in Smad4 Δ/Δ mice. In addition, StAR and P450scc mRNAs, involved in cholesterol transport and cleavage, respectively, are decreased which may contribute to a reduction in uterine E 2 synthesis required for decidualization and angiogenesis. Because SMAD4 and SMAD4ΔMH2, via the MH1 and SAD domains, can act as corepressors of ERα, 86 it can be hypothesized that SMAD4ΔMH2 may compromise the regulation of decidual and angiogenic factors. In addition to the steroidogenesis, the retinoid signaling pathway could also be affected in Smad4 Δ/Δ mice. Indeed, the retinoic acid (RA) transporter Crabp1 mRNA is downregulated while Cyp26a1, an enzyme regulating the cellular level of RA, is upregulated. In human endometrial stromal cells, RA along with E 2 and P 4 play an important role during decidualization 87,88 and are regulators of VEGF secretion. 89,90 Overall, the observation that Smad4 plays an important role in uterine function is consistent with the fact that many members of the TGFβ superfamily signaling pathway are also involved. During the decidualization process, the genes Acvr1 (activin type 1 receptor), 91 Acvr2a, 92 Bmp2, 80 Bmpr1A, 93 Bmpr2, 94 Fst (follistatin), 95 Lefty, 96 Smad1/5, 92 Smad1/5/4, 46 or Smad2/3 97,98 are key factors. During implantation, the genes Acvr1, 91 Acvr2a, 92 Bmpr1A, 93 Bmp7, 99 Tgfb1, 100 Tgfbr1, 101 and Smad1/5 92 play an important role. Nevertheless, it should be noted that inactivation of Smad4 alone does not reveal any uterine function for this factor, since the Smad4 cKO (Δexon1) showed no phenotype in the uterus. From the analysis of Smad1/5/4 cKO mice and our Smad4 cKO, which expresses SMAD4ΔMH2 with potential dominant-negative activity, it appears that SMAD4 functions in close interaction with other SMADs.
In conclusion, we hypothesized that the reproductive dysfunctions observed in our Amhr2 +/cre ; Smad4 f/f mouse model would be attributed to the expression of an MH2truncated SMAD4 protein probably acting as a dominant negative factor in the TGFβ superfamily signaling pathway. SMAD4ΔMH2 disrupts the onset of implantation by interfering with the decidualization process and the set-up of vascular development. In addition, these mice develop estradiol-induced oviductal diverticula, which contribute to further infertility by preventing gamete and embryo transport. Interestingly, the fact that diverticula can develop on control oviducts only after 20 weeks of age suggests that Smad4 may act as a factor delaying the oviduct aging during the female reproductive life span. These results are of great importance in understanding unexplained infertility related in particular to tubal abnormalities that are often not diagnosed. Since mutations affecting the MH2 domain of SMAD4 are common in human, understanding the molecular function of this protein could be of great interest for the clinical research dealing with human somatic cancer.

ACKNOWLEDGMENTS
We would like to thank Richard R. Behringer (M.D. Anderson Cancer Center, Houston, TX) for the initiation of this study in his laboratory and for providing the Amhr2-cre mouse line, and Charles Pineau and Michael Primig for allowing the termination of this work. The TROMA-I monoclonal antibody developed by Dr. Philippe Brulet and Dr. Rolf Kemler (Service de Génétique Cellulaire, Institut Pasteur, Paris, France) was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biology, Iowa City, IA 52242. This work was supported by Inserm, Université Paris-Sud XI, Université de Rennes, EHESP and the "Agence Nationale de la Recherche" [ANR-08-JCJC-0059 to S.P.J., including a one-year fellowship to A.A.B].

DISCLOSURES
The authors declare no competing or financial interests.

DATA AVAILABILITY STATEMENT
The data set has been deposited in the NCBI Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih. gov/geo/) under the accession number GSE221018. 102