SALL1 Is an Orchestrator of Uterine Receptivity in Part via Titration of Estrogen Signaling

The generation of mice carrying a floxed allele of Sall1 has been previously described( 62 , 63 ). The Pgr -Cre line( 43 ), obtained from Dr. Francesco DeMayo at the National Institute of Environmental Health Sciences and Dr. John Lydon at Baylor College of Medicine, was crossed with mice homozygous for the Sall1 floxed allele to produce offspring carrying Pgr -Cre and one floxed allele. F1 progeny were backcrossed with Sall1 flox/flox mice to yield males with the genotype Sall1 flox/flox; Pgr -Cre . These males were bred with Sall1 flox/flox females to generate cKO females ( Sall1 fl/fl ; Pgr Cre/+ ) and their littermate controls ( Sall1 fl/fl ; Pgr +/+ ). For breeding study, Sall 1 Pgr females (N=8) and littermate Sall1 Cntrl females (N=6) were bred to wildtype males over a 250-day breeding study and cumulative pups were tracked for each breeding pair. For timed breedings, female mice were considered 0.5 days post-coitum (dpc) pregnant upon observation of vaginal plugs after mating. For estrous phase determination, the vaginal canal was flushed with 20 μL PBS, and vaginal smears were examined microscopically. For blastocyst flush, uteri were excised and cuts were made at each uterine-oviduct junction and directly above the cervix. Blunt needle was inserted a few millimeters into the uterine lumen at each uterine-oviduct junction and 200uL of PBS was used to flush each horn onto a small grid-lined petri dish. Entire surface area of the flushed PBS droplet was examined under >40x magnification and blastocysts tallied. For tail vein injection to assess implantation sites, 100uL of Chicago Blue (Sigma-Aldrich) solution (1% w/v in sterile PBS) was injected into the tail vein of females in the morning of day four of pregnancy. Five minutes after injection, the mice were sacrificed, and the uterine horns were exposed. Intensely stained segments flanked by unstained segments were considered implantation sites and tallied. Artificial decidualization assay was performed as previously described( 64 ). For fulvestrant treatment, pregnant female mice at 2.5dpc received a single subcutaneous injection of fulvestrant (5% DMSO/95% corn oil at 25 or 200 mg/kg B.W.) or vehicle and were sacrificed at 5.5dpc. Artificial decidualization on control and Sall1 cKO females was performed as described previously( 64 ). Progesterone levels were determined by Cayman ELISA using serum samples as previously described( 22 ). All experiments were conducted under strict adherence to the institutional animal care guidelines at Washington University School of Medicine, Missouri. Uterine tissue RNA was extracted using the Qiagen RNeasy Mini Kit following the manufacturer’s instructions. Reverse transcription was performed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems Inc., ABI). Quantitative PCR (qPCR) analyses were conducted on the QuantStudio5 Real-Time PCR System (ABI) using PowerUp ™ SYBR ™ Green Master Mix (ABI) and gene-specific primers. Each experimental condition was tested in three biological replicates. Relative gene expression changes were quantified using the delta-delta Ct method, normalized against the housekeeping gene Rpl7 . Pregnancy timepoints were confirmed by presence of blastocysts upon PBS flush of uterine horns for 3.5dpc or presence of Chicago blue stained implantation sites for 4.5dpc. Uterine luminal epithelium was isolated by cutting horns into ~4mm segments, opening segments longitudinally to form sheets, shaking sheets in 1mL 1% w/v trypsin in calcium and magnesium-free HBSS in 15mL conical tubes for 1.5hr at 4C on a shaking platform, followed by 30min shaking at room temperature. 4mL additional HBSS was added to each tube and tubes were shaken vigorously by hand to encourage LE sheets to detach. Contents of each tube was poured onto a 6cm dish. The large chunks were removed with tweezers and placed in new tube for subsequent stromal processing. The HBSS was collected from the dish by pipetting and spun at 9,000xG for 5 minutes. Supernatant was removed and LE pellets were washed once with full serum hESC growth media to inhibit remaining trypsin, and solutions were spun again at 9,000xG for 5 minutes. Supernatant was removed and remaining pellet was considered isolated LE. To isolate uterine stroma, remaining uterine chunks were incubated in 1mL 0.25% trypsin with 1mg/mL collagenase in HBSS at 37C on a platform shaker at 200rpm for 30 minutes. 3mL additional HBSS was then added to each tube, and 5mL serological pipette was used to pipette up and down at least 25 times to release the stromal cells into solution. This was then again moved to a 6cm petri dish, and large chunks (serous and muscle layers) were removed with tweezers and discarded. Remaining solution was passed through a 100um nylon filter (cell strainer), strainer was rinsed with 1mL full serum hESC growth media, and solution was transferred to tubes and spun at 9,000xG for 5 minutes. Pellets were washed once with 500uL full serum hESC growth media to inhibit remaining trypsin, and solutions were spun again at 9,000xG for 5 minutes. Supernatant was removed and remaining pellet was considered isolated uterine stroma. For RNAseq, RNA was isolated using Qiagen RNeasy Mini Kit following the manufacturer’s instructions. Ribo-depletion method and cDNA library prep for mRNA sequencing was performed by the Washington University Genome Access Technology Center (GTAC) and sequencing performed on a NovaSeq S4 at ~30M reads/sample 2x150. For ChIPseq, freshly isolated 3.5dpc luminal epithelia and stromal cells were washed with cold PBS +Ca and Mg and crosslinked for 30 minutes at room temperature with 2mM Di(N succinimidyl) glutarate (DSG, Proteo Chem C1104) followed by the addition of 1% formaldehyde (Thermo Scientific 28906) and incubated for an additional 10 minutes with rocking at room temperature. Crosslinking was quenched with 125mM glycine for 5 minutes at room temperature. Tissue was washed with PBS + 1mM PMSF and snap frozen and stored at −80°C. Pooled luminal epithelia and stroma was homogenized with a handheld dounce homogenizer in Lysis Buffer 1 (50mM HEPES-KOH pH 7.5, 140mM NaCl, 1mM EDTA, 10% glycerol, 0.5% NP-40, 0.25% Triton-X 100) and incubated for 20 minutes at 4°C with rocking. Samples were pelleted at 2800 rpm at 4°C for 5 minutes and the supernatant discarded. Pellets were resuspended in Lysis Buffer 2 (10mM Tris-HCl pH 8.0, 200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA), briefly homogenized and rocked gently for 10 minutes at room temperature. Nuclei were pelleted at 2800 rpm for 5 minutes at 4°C and resuspended in Lysis Buffer 3 (10 mM Tris-HCl pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.1% Na-Deoxycholate, 0.5% N-lauroylsarcosine). Chromatin was sheared to a size of 100-1000bp using a Covaris ME220. Chromatin was pre-cleared with 20 μl protein A/G magnetic beads (Thermofisher Scientific 88803) for 2 hours rocking at 4°C. Pre-cleared chromatin was incubated with 10 μg primary antibody (Sall1, Abcam 31526 or rabbit IgG, Cell Signaling 2729) overnight with rocking at 4°C. The next morning 25 μl protein A/G magnetic beads were added and incubated for 4 hours at 4°C. Complexes were washed 5 times with cold modified RIPA buffer (50 mM HEPES-KOH pH 7.5, 500 mM LiCl, 1 mM EDTA, 1% NP-40, 0.7% Na-Deoxycholate) and 2 times with TE (10 mM Tris pH 8.0, 1 mM EDTA). Complexes were eluted by incubating at 65°C for 30 minutes in Elution Buffer (50 mM Tris, 10 mM EDTA, 1% SDS). Crosslinks were reversed by adding NaCl (0.2M) and heating at 65°C for 15 hours, 200 μg/mL RNase A was added and incubated for 2 hours at 37°C, and 200 ug/mL Proteinase K added and incubated 2 hours at 63°C. DNA was isolated using the MinElute PCR Purification Kit (Qiagen 28004). 1ng ChIP DNA was used to prepare sequencing libraries using NEBNext Ultra II DNA library prep kit for Illumina (New England Biolabs, E7645) and multiplex oligos (New England Biolabs, E6440). Libraries were sequenced on the NovaSeq 6000 (Illumina) targeting 50 million paired-end reads. 2 independent ChIP-seq libraries were sequenced for Input, Sall1, and IgG for luminal epithelia and stroma. Trimmed fastq files were aligned to the mm10 reference genome using Bowtie2. Aligned reads were extracted and converted to a sorted bam file using SAMtools. Scaled BigWig files and correlation plots were generated using Deeptools. Peaks were called using Macs2, using IgG as the background control. Peak annotation and downstream analysis was performed in R using ChIPseeker. Tissues were fixed in Bouin’s fixative, dehydrated, and embedded at the Developmental Biology Histology Core at Washington University School of Medicine. Eight-micron sections were stained with hematoxylin and eosin (H&E) for histological analysis. Immunofluorescence was performed as previously described using antibodies diluted in a blocking solution (1% BSA, 3% normal goat serum in PBS) at the concentrations listed in File S10 . In situ hybridization was performed on 4% paraformaldehyde (PFA)-fixed, paraffin-embedded 8 μm tissue sections using the RNAscope ® 2.5 HD Assay-RED kit (Advanced Cell Diagnostics, ACD, Newark, CA). Pre-designed gene-specific double-“Z” oligo probes against Hand2 and Bmp2 were purchased from ACD performed according to manufacturer’s instructions. AP activity was determined by staining PFA-fixed paraffin-embedded sections with BCIP/NBT as described previously( 22 ). UM cells were isolated from control and cKO uteri as previously described( 64 ) and allowed to attach to cell culture dishes overnight. UM cells were treated with or without hormone cocktail containing E2, MPA and cAMP (EPC) at previously described concentration( 6 ) to induce decidualization in vitro, fixed in 4% PFA, and subsequently subjected to AP staining( 22 ) to assess decidualization. Immortalized human endometrial stromal cells (hESCs) were cultured as previously described( 6 ). Knockdown of hSall1 in human endometrial stromal cells was performed via reverse transfection using DharmaFECT 4 (DF4) transfection reagent (Dharmacon) according to the manufacturer’s protocol. Briefly, for each well of a 6-well plate, 2 μL DF4 was diluted in 100 μL sterile, phenol red-free Opti-MEM and incubated at room temperature for 5 minutes. Separately, 4 μL of 10 μM siRNA stock solution (Silencer Select Pre-Designed siRNA, ThermoFisher Cat. 4392420) was diluted in 100 μL sterile, phenol red-free Opti-MEM and subsequently combined with the DF4 mixture. The siRNA/DF4 complex was incubated at room temperature for 20 minutes. hESCs were trypsinized, pelleted by centrifugation, resuspended in culture medium, and counted. A total of 25,000 cells were seeded into each well containing the siRNA/DF4 complex and allowed to adhere overnight. The following day, induction cocktail was added to initiate in vitro decidualization. All experimental groups contained at least three biological replicates, if not specified otherwise. The group Sall1 cKO at 4.5dpc for RNA sequencing (both LE and stroma) is N=2. Two-tailed Student’s t-test assuming unequal variance was performed to compare means of the experimental groups. Data are presented as mean ± SD, with raw individual experimental data displayed as dot plots overlay and p-value less than 0.05 was considered statistically significant.

To determine the potential function of Sall1 in female reproductive tract function, quantitative PCR was first used to examine its expression in wildtype adult female mice. Tissue from females in all stages of the estrous cycle were pooled and tested, showing relatively low expression of Sall1 in the vagina and ovaries, with twenty-fold higher expression in the oviducts and uterus ( Figure 1A ). Uterine expression of Sall1 exhibited a wide expression range when pooled from multiple stages of the estrous cycle. To determine whether Sall1 expression fluctuates throughout the estrous cycle and early pregnancy, uterine tissue was analyzed by quantitative PCR for each individual stage, revealing significant variability in Sall1 expression between stages ( Figure 1B ). Its highest expression stages are proestrus and diestrus, as well as increasing levels across early pregnancy progression from 2.5dpc to 4.5dpc. These data suggest that Sall1 expression may be directly or indirectly regulated by female reproductive hormones. The endometrium has three main histological features: the stroma, the luminal epithelium, and the uterine glands. Each feature plays important roles in embryo receptivity and pregnancy. To clarify in which uterine compartments Sall1 may be functioning, immunofluorescence staining was done at all stages of the estrous cycle and during early pregnancy ( Figure 1C – G ). SALL1 protein shows strong nuclear expression in both the luminal and glandular epithelia during proestrus ( Figure 1C , arrowheads ), as well as relatively high stromal expression. Its expression then significantly reduces during estrus and metestrus ( Figure 1D , E ) in all compartments. Diestrus ( Figure 1F ), as well as early pregnancy at 3.5dpc ( Figure 1G ) sees a resurgence of expression with the same pattern as proestrus but with slightly fewer cells exhibiting nuclear localization. The estrous stage-dependent expression and fluctuating subcellular localization pattern of SALL1, with its expression evident across all three main endometrial histological features, suggests SALL1 could be playing important roles in all compartments of the uterus in relation to cycling hormones, justifying further investigation. To examine the consequence of Sall1 loss in the female reproductive tract, mice with conditional deletion of Sall1 ( Sall1 flox) were bred to the Pgr -Cre driver to conditionally delete Sall1 across female reproductive tract tissues. The Pgr -Cre driver is a knock-in of cre-recombinase to the native Pgr locus, which also disrupts Pgr expression( 43 ). For this reason, the Pgr -Cre driver must be used in a heterozygous manner to prevent simultaneous homozygous disruption of Pgr . Mice were bred to be homozygous for Sall1 flox and heterozygous for Pgr -Cre. For the purposes of this manuscript these mice are termed Sall1 Pgr or Sall1 cKO throughout. Littermate controls are homozygous for Sall1 flox but are negative for the Pgr -Cre knock-in allele; these are termed Sall1 Cntrl or simply “control” throughout. To determine the efficiency of Sall1 deletion via Pgr -Cre mediated recombination in the uterus of Sall1 cKO mice, immunofluorescence staining was performed at 3.5dpc ( Figure 1H ) and showed little to no detectable remaining expression of SALL1 protein in the uterus in any of the uterine compartments. Sall1 Pgr females (N=8) and littermate Sall1 Cntrl females (N=6) were bred to wildtype males over a 250-day breeding study and cumulative pups were tracked for each breeding pair. While controls produced the expected number of pups per litter and litter frequency, Sall1 cKO females were entirely sterile, producing no pups and showing no visible signs of pregnancy over the entire study duration ( Figure 2A ). As Sall1 cKO females have Sall1 deleted in multiple female reproductive tract tissues, the next step is to determine at which stage of reproduction the sterility is first evident. Sall1 cKO females and their littermate controls produce comparable numbers of blastocysts, evidenced by their recovery via PBS flush of the uterine lumen at 3.5dpc ( Figure 2B ). Correlating to this finding, the ovaries of Sall1 cKO at 3.5dpc exhibit histologically normal corpora lutea, suggesting normal ovulation ( Figure 2C ) and the blastocysts flushed from Sall1 cKO uteri appear normal in number and morphology with visible inner cell mass under light microscopy ( Figure 2D ). These finding indicate that ovulation in Sall1 cKO is normal, that descent of blastocysts into the uterus is also normal, and the failure of pregnancy in Sall1 cKO is happening after 3.5dpc. This also concludes that the sterility of Sall1 cKO is specific to the uterus and its functions in pregnancy, rather than potential roles of SALL1 in the ovaries and oviducts where Sall1 is also expressed. The next phase of pregnancy is blastocyst attachment to the uterine luminal epithelium (LE), quickly followed by implantation of the blastocyst into the uterine stroma and stromal decidualization. To determine whether the sterility of Sall1 cKO females is circa the implantation phase of pregnancy, Chicago blue tail vein injection was performed on N=7 Sall1 cKO and controls to visualize implantation sites. Chicago blue dye reliably visualizes the increased vascular permeability that occurs at individual implantation sites along the bicornuate mouse uterus at 4.5dpc before any decidual swellings (increases in uterine mass at implantation sites that is secondary to uterine decidualization, resembling small balls) are visible the following day at 5.5dpc. While Chicago blue uptake at implantation sites was clearly visible in controls ( Figure 2E , blue arrows ), with proper embryo spacing and at the expected numbers, Sall1 cKO uteri by contrast showed either no uptake or diffuse uptake of Chicago blue dye ( Figure 2F ), indicating a likely failure of implantation and potentially pathologic vascular permeability. To complement the findings of the Chicago blue test, decidual swelling counts were also taken. Controls (N=8) showed an average of 11 decidual swellings per uterus ( Figure 2G , blue arrows ) at either 5.5 or 6.5dpc, while Sall1 cKO (N=6) showed none ( Figure 2H ). Together these findings indicate that SALL1 is required during the embryo implantation phase of early pregnancy. The implantation phase of pregnancy is a rapidly progressing series of events beginning with blastocyst attachment to the uterine luminal epithelia, followed by uterine stromal layer decidualization and blastocyst invasion into the uterine stroma. Failure of any of these steps results in pregnancy failure. The blastocyst attaches to the uterine epithelium through physical interactions involving integrins binding to ECM molecules, cadherin-mediated adhesion, and L-selectin on trophoblast cells interacting with its ligands on the uterine surface( 3 ). Hence, if blastocysts are not properly attaching to the mouse uterine luminal epithelia, then they can be flushed easily from the uterine lumen directly after the stage where they should already be attached. All blastocysts should be firmly attached by 4.5dpc, and if uterine flush one day later at 5.5dpc produces blastocysts, this indicates that some or all blastocysts are not attaching properly; this is the case in Sall1 cKO ( Figure 3A ). Further investigation via H&E staining showed that only free-floating blastocysts are evident in Sall1 cKO uterine lumen at 4.5dpc ( Figure 3B ). Just prior to blastocyst attachment, MUC1 protein at the apical pole of the luminal epithelia is normally dramatically decreased. Studies suggest this reduction in MUC1 is a prerequisite of blastocyst attachment( 44 , 45 ), and that retained expression prevents embryo adhesion to the uterine LE. This reduction in MUC1 is evident in control uteri at 4.5dpc, while in contrast MUC1 protein persists in Sall1 cKO ( Figure 3C , 3D ) as evidenced by immunofluorescence staining. Additionally, uterine closure is a physiological adaptation in mice during early implantation to ensure that the blastocyst stays in close contact with the uterine lining, promoting stable adhesion, proper positioning, and retention of signaling molecules necessary for successful implantation and pregnancy( 46 ). This uterine closure is evident in controls while the uterine lumen remains open in Sall1 cKO. Expression of LTF also persists in Sall1 cKO at 4.5dpc in the luminal epithelium, whereas it is absent in the LE in controls ( Figure 3E – F ). During early implantation, lactoferrin (LTF) acts as an immunomodulatory glycoprotein that helps maintain uterine receptivity by regulating local immune responses and supporting embryo-maternal interactions. During the pre-implantation period (around 3.5 dpc), lactoferrin is highly expressed in the luminal and glandular epithelium. However, as implantation progresses to around 4.5 dpc, its expression significantly decreases( 47 ). This downregulation is hypothesized to facilitate reduced immune activity, allowing for successful implantation. Sall1 cKO and controls were also tested by immunofluorescence staining to determine whether luminal epithelial cell proliferation ceases as it normally should at 3.5dpc, which is does ( Figure 3G , H ). Both Muc1 ( 48 ) and Ltf ( 49 ) are regulated by estrogen receptor alpha. We therefore tested expression of ESR1 protein. In the luminal epithelium ESR1 expression should be relatively low at 3.5dpc, which is the case in controls ( Figure 3I ) but not in Sall1 cKO ( Figure 3J ). These findings suggest that deregulated estrogen signaling is a likely cause of Sall1 cKO sterility. Because SALL1 is a transcription factor that can act as either an activator or repressor, it is essential to examine global transcriptional deregulation in Sall1 cKO uteri. Quantitative PCR was performed for a panel of known pregnancy health markers at 4.5dpc. While some markers remained normal ( File S1 ), many critical genes are transcriptionally deregulated ( Figure 3K ). This deregulation could be caused from direct SALL1 transcriptional binding, or secondary to and indicative of implantation failure. Transcripts aberrantly upregulated in Sall1 cKO 4.5dpc uteri include Esr1 , Sox17 , Lpar3 , Gata2, and order of magnitude upregulation of Lif and Ltf . Transcripts aberrantly downregulated in Sall1 cKO 4.5dpc uteri include Igfbp1 , Hoxa10 , Bmp2 , Hand2 , Ereg , and Wnt7b . The deregulation of so many critical indicators of pregnancy health in Sall1 cKO implicate SALL1 as a probable upstream player in the progression of the early pregnancy transcriptional cascade. To determine whether deregulation of progesterone signaling may be playing a part in Sall1 cKO sterility, serum progesterone levels were measured by ELISA and determined to not be significantly different from controls ( Figure 4A ). There are not differences in progesterone receptor (PR) staining between controls and Sall1 cKO at 3.5dpc ( Figre 4B , C ). However, PR expression failed to be downregulated in the LE at 5.5dpc, consistent with the non-receptive phenotype in the mutant LE, while other endometrial cell types exhibit downregulated PR ( Figure 4D , E ). Alkaline phosphatase (AP) staining is used to assess stromal cell decidualization at 4.5dpc. AP staining is normally very dark around an implantation site at 4.5dpc, which is true in controls ( Figure 4F ) but it is nearly absent in Sall1 cKO ( Figure 4G ). In these staining experiments, the failure of the uterine lumen to close in Sall1 cKO is again evident ( Figure 4B – G ). Lack of decidualization in the mutant stroma as reflected by AP staining could be due to failure of blastocyst attachment causing a failure of the signaling cascade that triggers decidualization. However, SALL1 protein is expressed in stromal cells as well as LE cells, so it remains a question whether SALL1 also plays a role in decidualization of the uterine stroma independent of blastocyst attachment. To begin to answer this, artificial decidualization assay was performed. When exogenous hormones were provided to ovariectomized control females and trauma is introduced to one uterine horn, it undergoes rapid proliferation and differentiation that mimics the natural decidualization process. Consistently, Sall1 cKO uteri are unable to decidualize in this assay as well ( Figure 4H ). While this does not answer whether there is any defect in the stromal cells independent of the LE stimulus, it does conclude with certainty that even a simulation of blastocyst attachment (the injection) cannot stimulate the Sall1 cKO uterus to decidualize the way that it easily can in controls. Still, this phenotype could be caused by an inability of the LE to signal to the stroma, an inability of the stroma to respond to the signal and begin decidualization, or both. To answer whether SALL1 is required in uterine stromal cells for proper decidualization independent of any LE signaling trigger, primary uterine mesenchymal cells (UM cells) were isolated, cultured in vitro, and subjected to in vitro decidualization assay. UM cells from controls and Sall1 cKO were treated with or without hormone cocktail containing estrodiol (E2), medroxyprogesterone (MPA), and cyclic adenosine monophosphate (cAMP) (EPC) to induce decidualization. Cells were subsequently stained with AP substrates to assess decidualization. While UM cells from controls were able to robustly decidualize in vitro as evidenced by AP staining, UM cells from Sall1 cKO exhibited minimal to no decidual response ( Figure 4I – L ). To further show that UM cells from Sall1 cKO are unable to respond to decidual stimuli, several transcripts that normally are upregulated during decidualization ( Bmp2 , Hand2 , Igfbp1 , and Prl ) were tested by quantitative PCR. In addition to having lower baseline expression of these transcripts in the absence of EPC hormone cocktail, Sall1 cKO UM cells could not properly induce their expression in response to EPC either ( Figure 4M ). In addition to mouse models, in vitro decidualization is also a popular method of showing decidualization defects in cultured human cells. Our group has used this method in the past many times, and we produced a reporter cell line to make the process more streamlined( 6 ). For the regular assay, immortalized human endometrial stromal cells are cultured in vitro and undergo treatment with EPC cocktail. The assay read-out is quantitative PCR of decidual markers like PRL and IGFBP1 . To show some translational relevance of our mouse model, we cultured human endometrial stromal cells and treated them with either scramble siRNA or siRNA against SALL1 , achieving a roughly 70% knockdown, and subsequently subjected them to in vitro decidualization assay. All cells were treated with EPC, as without EPC the read-out transcripts are typically undetectable even in control cells. Cells treated with siRNA against SALL1 exhibit reduced expression of decidual markers PRL and IGFBP1 in the presence of EPC compared to cells treated with scramble siRNA ( Figure 4N ) indicating an inability to properly undergo decidualization reaction and upregulate these transcripts in response to hormone stimulus. To delineate the transcriptional changes secondary to loss of SALL1 in independent uterine compartments, RNA sequencing was performed on control and Sall1 cKO uteri for both isolated luminal epithelia and isolated stroma at both 3.5dpc and 4.5dpc. Transcriptional changes secondary to SALL1 loss are so pronounced that differentially expressed genes (DEGs) segregate first by genotype and then by uterine compartment. This suggests that SALL1 is controlling similar genes in both compartments, and that loss of SALL1 results in a transcriptional overhaul of both compartments ( Figure 5A , 3.5dpc representative heatmap ). To determine which genes are bound directly by SALL1 in early pregnancy, ChIP-sequencing for SALL1 was performed on wildtype 3.5dpc CD1 mice (the same genetic background as our control and cKO mice) with confirmed ovulation via blastocyst flush. Again, experiments were performed on independent uterine compartments. Representative binding tracks ( Figure 5B ) show that SALL1 binds its own promoter, as well as those of Fst and Esr1 in both the LE and stroma. This is consistent with the RNAseq finding that SALL1 loss affects transcripts in similar ways across both compartments. When peaks at gene body and promoter regions are combined to form a list of genes with no duplicates, 3107 genes are bound by SALL1 at 3.5dpc that are common between the LE and stroma, representing an impressive 90% of uterine stroma peaks and 26% of LE peaks. Still, there are 8548 genes bound by SALL1 uniquely in the LE and 338 bound by SALL1 uniquely in the stroma, indicating that while there is extensive overlap, SALL1 still plays unique roles independently in each uterine compartment ( Figure 5C ). As pregnancy progresses from 3.5 to 4.5dpc, more DEGs total are present when comparing Sall1 control uteri to Sall1 cKO uteri, while the percentage of overlap between the DEGs in the LE and stroma remain similar. When comparing DEGs of 3.5dpc to 4.5dpc in the stroma vs the same comparison in the LE, again similar amounts of overlap are observed ( Figure 5D ). Approximately 59% of the DEGs found in the LE at 3.5dpc were also found to be bound by SALL1 at promoters via 3.5dpc ChIP sequencing. This percentage is slightly less, at 55% when comparing DEGs found in the LE at 4.5dpc to SALL1 ChIPseq promoter peaks via 3.5dpc ChIP sequencing, likely due to the difference in timepoint comparison. Less overlap, though still very significant, is found in the LE when comparing DEGs at either timepoint to gene body ChIPseq peaks ( Figure 5E ). In the stroma, as was true in the LE, more DEGs are found at dpc4.5 compared to dpc3.5. About 13% (3.5dpc RNAseq) to 16% (4.5dpc RNAseq) of DEGs in the stroma are also bound by SALL1 at promoters in 3.5dpc ChIPseq. This percentage decreases to 9.7% and 7.3% respectively when looking instead at stromal gene body peaks ( Figure 5F ). All shown is statistically significant compared to random expected overlap with p-value at or less than 1.0E-28. Gene lists with individual p-values from all datasets are available in Files S2 – S9 and raw data is available on GEO (submission pending). To validate RNAseq findings, several DEGs were also tested via quantitative PCR that showed similar patterns in both the LE and stroma, either upregulated in both compartments or downregulated in both compartments in response to SALL1 loss. Quantitative PCR confirmed the RNAseq findings ( Figure 5G ). For genes showing either gene body or promoter peaks in the ChIPseq data, as well as DEGs at both dpc3.5 and dpc4.5 in the RNAseq data, ontology analysis using PANTHER pathways ( pantherdb.org ) was performed independently for the LE (757 genes) and stroma (351 genes). Analysis was also performed for genes that overlap between the LE list and the stroma list (141 genes) ( Table 1 ). Enriched pathways in the LE include the Alzheimer disease pathway (for relevance to reproductive biology, this list includes a large number of Wnt pathway genes), the GnRH receptor pathway, the endothelin signaling pathway, integrin signaling, GABA-B receptor II signaling, angiogenesis, Insulin/IGF/PKB, CCKR signaling, cadherin signaling, PI3 kinase pathway, TGF-beta pathway, EGFR, and PDGF signaling among others. There are fewer enriched pathways in the stromal analysis – only integrin signaling and the GnRH receptor pathway. When all datasets are overlapped, the only pathway enrichment remaining is GnRH receptor pathway, including key players in uterine function like Bmp2 , Fst , and Tgfb2 . The role of Sall1 in uterine transcriptional regulation via direct binding is vast and diverse; this overlap analysis is meant to demonstrate how a large dataset can be analyzed to reveal overarching regulatory themes and should not be misinterpreted to mean that other deregulated genes not fitting nicely into overlap analyses are less functionally relevant. Because expression studies by both quantitative PCR and immunostaining showed persistent estrogen receptor alpha expression at pathologic times during pregnancy progression, as well as persistent expression of the ESR1 targets MUC1 and LTF, and because ChIP sequencing showed binding of SALL1 to the Esr1 promoter, we hypothesized that SALL1 repression of Esr1 in the uterus is required for proper LE differentiation and uterine receptivity. To confirm that Esr1 is downstream of Sall1 in the pregnancy cascade, we treated Sall1 cKO females with the selective estrogen receptor antagonist, fulvestrant. A single subcutaneous injection of either 25mg/kg (low dose) or 200mg/kg fulvestrant (high dose) was given at 2.5dpc and uteri were collected for histology at 5.5dpc. As previously shown, embryos cannot properly attach to the uterine LE in Sall1 cKO and are consistently found floating in the uterine luminal space ( Figure 6A ). However, in Sall1 cKO treated with high dose fulvestrant, some embryos make contact with the LE ( Figure 6B ) and the luminal epithelium adjacent to the attachment site appears on histology to be breaking down in the way that is expected during normal attachment. Additionally, cKO treated with fulvestrant show rescued low levels of MUC1 expression ( Figure 6C – D ). As is expected under treatment with a selective estrogen receptor antagonist, ESR1 protein levels are lower, showing that the treatment is working as expected, with the goal of levels similar to wildtype animals at this stage of pregnancy ( Figure 6E – F ). It is important to note that ESR1 protein expression is lower in some LE cells in the fulvestrant treated group, but not all, suggesting that inhibition of ESR1 activity is not 100% efficient. Quantification of embryos in contact with the LE upon histology shows that ~40% of blastocysts in high dose fulvestrant treated Sall1 cKO were able to attach to the uterine LE, compared to none in the low dose fulvestrant treatment group and untreated Sall1 cKO ( Figure 6G ). Staining for the epithelial marker E-cadherin shows physical contact between these blastocysts and the LE in high dose fulvestrant treated Sall1 cKO. As expected based on decidualization findings, fulvestrant can only rescue attachment, and fertility is still impaired, as evidenced by impaired luminal closure and absence of clear decidual swellings in these mice. Other or additional downstream targets of Sall1 are orchestrating processes downstream of attachment apart from Esr1 . These findings emphasize that Sall1 is playing multiple critical separate roles in early pregnancy, in both the luminal epithelium as well as the stromal uterine compartments.

Transcription factors are responsible for driving morphologically transformative processes across embryonic development, and it follows that they would also drive similar though rarer morphologically transformative processes in adolescent and adult physiology – processes such as puberty, wound healing, response to infection, and pregnancy. Here we have described one transcription factor, SALL1, whose most studied roles are in embryonic development, and shown that it also has significant roles in similarly transformative processes in adult tissues, namely the uterus during pregnancy. Not only this, but its roles are pleiotropic; it is necessary in two independent tissue layers of the uterus during two processes that are absolute prerequisites for healthy pregnancy – embryo attachment and decidualization. The original interest in SALL1 in female reproduction stemmed from its recently identified roles in the development of the lower urogenital system in males. Sall1 is a direct target of androgen receptor in the developing genital tubercle, and its loss gives rise to morphological defects in urethral tubularization partially due to proliferative deficits in the mesenchyme. These proliferative deficits were also seen in female genitalia, despite females having relatively low expression of androgen receptor( 42 ). This suggested a potential role for SALL1 across sexually dimorphic development of the urogenital tract. Other roles for SALL1 in the upper urogenital tract, namely the kidneys, have been previously well-described( 36 , 37 ). The changing expression of Sall1 mRNA and protein across the mouse estrous cycle indicated that it has a transcriptional relationship with cycling hormones, and this observation solidified the need to perform tissue specific knockout experiments. While we hypothesized some level of subfertility in the Sall1 cKO females, the full penetrance of sterility was unexpected and impressive, concluding without doubt that Sall1 is required for pregnancy. The absence of pregnancies in cKO females despite normal ovulation and fertilization led us to consider the uterus as the primary candidate for investigating the requirement of Sall1 . After experiments concluding its pleiotropic roles in embryo attachment and decidualization, the question shifted from physiology to genetics. We understood which processes in pregnancy are affected by SALL1 loss, but what downstream pathways are at play to produce these effects? Sequencing of mRNA and via ChIPseq revealed several key players downstream, two of which may be the keystones of SALL1 uterine function in the separate uterine compartments: estrogen receptor alpha in the luminal epithelium, and follistatin in the stroma. The restrictive implantation window can be viewed as a protective mechanism against pregnancy during times when the body cannot fully support it. If an embryo were to attach during a time when the uterus is not able to support it, the subsequent pregnancy loss would come with all the associated health risks. The pre-receptive phase when the uterus is preparing for but not yet receptive to implantation, and refractory phase when the uterus is no longer accepting blastocysts for implantation, flank this short implantation window of receptive phase when the uterus is optimally prepared( 1 ). Molecular mechanisms are in place to remove these defenses cyclically in response to appropriate stimuli, allowing for pregnancy. During this implantation window, estrogen signaling must be downregulated to allow luminal epithelial differentiation, part of which includes downregulation of MUC1 expression on the apical surface of the uterine luminal epithelium, in order for blastocysts to attach. Our rescue experiments using the selective estrogen antagonist, fulvestrant, show that titrating estrogen receptor levels in the uterine luminal epithelium is a major function of SALL1. When estrogen receptor levels, and MUC1 levels (an ER target) are within a more normal range, many of the competent blastocyts in Sall1 cKO were able to attach to the luminal epithelium. The observation that this rescue was incomplete shows that either estrogen receptor was not titrated to precisely normal levels, or alternatively there are other critical barriers to embryo attachment in Sall1 cKO independent of estrogen receptor. Additionally, even in those uterine regions where embryos successfully attached to the LE in Sall1 cKO with high dose fulvestrant, the uterine stroma failed to decidualize. No decidual swellings or AP staining were detectable, and in most instances luminal closure was still impaired. These findings indicate that additional compensation is required in the molecular network in order to produce a successful pregnancy. The Sall1 cKO phenotype, including the ability of an estrogen antagonist to rescue, is very similar to the phenotype seen in Pgr-cre mediated knockout of another critical transcription factor, NR2F2 (COUP-TFII). Nr2f2 cKO females show infertility due to failure of both embryo attachment and uterine decidualization from disrupted epithelial-stromal signaling. NR2F2, similar to SALL1, was shown across multiple studies to regulate BMP2, HAND2, and estrogen signaling, with fulvestrant (ICI 182780) partially rescuing implantation in the model( 16 , 50 – 52 ). In our RNA-seq datasets at 4.5dpc, Nr2f2 transcript was modestly but significantly downregulated in both the epithelial and stromal compartments, indicating a functional relationship between the two pathways. Decidualization fails completely in Sall1 cKO uteri as assayed during natural pregnancy, artificial decidualization and in vitro uterine stromal decidualization. Genes essential for decidualization such as Bmp2 and Hand2 failed to be induced by E2+P4. Intriguingly, the basal expression levels of these two genes were barely detectable ( Fig. 4 ) suggesting that Sall1 may be required to activate the expression of these two genes during decidualization. Moreover, follistatin ( Fst ), a regulator of Bmp2 signaling, is also downstream of SALL1 in the stroma, with loss of SALL1 resulting in blunted expression of Fst . Loss of FST in the uterus results in severe subfertility, producing only 2% the number of pups expected( 21 ). Partly this is due again to failure of attachment at the level of the LE secondary to hormone stimulus, but similar to Sall1 cKO mice, Fst cKO mice using the same driver also exhibit defects in stromal decidualization. In some models with insufficient Fst in the stroma, where LE attachment does occur but decidualization is deficient, supplementing with exogenous FST can rescue some portions of the decidual response( 22 ). For this reason, we suspect that Fst is at least one of the keystones of uterine function downstream of SALL1 in the stroma. Finally, it is interesting to note that PANTHER analysis revealed altered integrin signaling pathway in the mutant stroma which could contribute to the decidualization phenotype. Integrins play important roles in defining the window of implantation and some studies suggest that they are involved in a reciprocal mechanism to strengthen embryo attachment to the maternal decidua( 53 – 55 ). Progesterone signaling is critical in establishing the window of implantation. It suppresses LE proliferation by upregulating Hand2 in the stroma, which then inhibits expression of fibroblast growth factors (FGFs) that normally stimulate LE proliferation( 16 ). In Sall1 cKO, LE proliferation is properly shut off, suggesting that this pathway is intact despite lower expression of Hand2 . However, differentiation of the LE in Sall1 cKO is abnormal, despite normal cessation of LE proliferation, indicating that this might be a good model to decouple LE differentiation and proliferation. A crucial transcription target of progesterone receptor is SOX17, which is required in the LE for embryo implantation( 56 ). Elevated estrogen signaling was observed in Sox17 mutant LE, however, ESR1 protein was not ectopically expressed. ChIP sequencing of SOX17 in the uterus shows that Sall1 is a transcriptional target of SOX17( 19 ). In 4.5dpc stroma of Sall1 cKO compared to controls, Sox17 is upregulated, suggesting that Sall1 and Sox17 are in a regulatory loop. SALL1 can act as an activator or repressor depending on context( 28 , 29 ). With estrogen receptor in the luminal epithelium, SALL1 is apparently acting as a repressor, titrating its levels to allow for embryo attachment. Prior studies have used a version of SALL1 with a 3 amino acid mutation in its N-terminus, disrupting its interaction with the nucleosome remodeling and deacetylase (NuRD) complex( 37 ). It would be interesting to investigate which portions of Sall1’s pleiotropic uterine functions are due to its NuRD-dependent versus independent (potentially activator) activity. Additionally, being able to individually study Sall1 roles in the LE and stroma using cre-recombinase models that are efficient and specific would be very valuable. A recently generated decidua-specific cre-recombinase mouse model( 57 ) could be used to better distinguish SALL1’s LE versus stromal functions. Luminal epithelial specific cre-recombinase mouse models, such as Ltf-icre( 58 ), could also be useful to investigate SALL1 functions further. For both these potential experiments, the specificity and efficiency of cre recombination would need to be perfect in order to make clear conclusions about a gene like Sall1 with pleiotropic roles. Crossing the Sall1 cKO model to a Muc1 knockout( 59 ) may further elucidate whether persistent MUC1 protein at the apical uterine LE is the primary barrier to blastocyst adhesion and attachment. For clinical translational relevance, it would be helpful to know whether SALL1 heterozygosity is a sensitizer to poorer IVF outcomes, how much its variation in the population contributes to infertility, and whether variation in SALL1 has any relevance to uterine diseases like endometriosis or uterine cancer. A microarray study examining gene expression in the endometrium of patients with implantation failure and recurrent pregnancy loss (RPL) indicated a significant reduction (~50%) of SALL1 expression in implantation failure samples( 60 ). SALL1 expression was similarly reduced by about 40% in RPL samples but did not reach statistical significance. These preliminary data suggest that reduced SALL1 expression may be correlated with implantation failure and/or recurrent pregnancy loss. Further investigations into Sall1 uterine function should focus on further breaking down its individual uterine functions and finding downstream effectors that are subjectable to targeted small molecule modulation. Many master regulators of development retain near ubiquitous expression in the adult. This energy expenditure suggests their function extends into adult tissue maintenance, function, and likely disease. The ability to tease out adult-tissue specific functions of genes whose loss of function are embryonic or perinatal lethal is relatively new, and functions in reproductive biology are often low on the priority list. While life-giving by definition, and despite infertility alone affecting up to 15% of couples( 61 ), reproductive biology and its diseases are considered to be a question of quality of life, and the field receives significantly less funding compared to high-mortality diseases like heart disease, cancer, and diabetes ( https://report.nih.gov/funding/categorical-spending#/ ). This makes hypothesis-forming datasets specific to reproductive biology relatively rare. In the dearth of these datasets, perhaps by looking at readily available datasets detailing regulators of embryonic development, and investigating their potential reproductive roles, we may very well find that there are many more genes like Sall1 with critical reproductive functions well after their developmental roles are complete. In summary, herein it is shown that Sall1, a transcription factor known largely for its roles in kidney development, is also a critical regulator of embryo implantation via its non-negotiable roles in both the uterine epithelia during embryo attachment and uterine stroma during decidualization. Its orchestration of powerful pathways such as estrogen signaling and FST signaling place it upstream in multiple integrated molecular cascades necessary for successful pregnancy. In the clinical realm, this study, like many others before it, reemphasizes the need to identify, address, and treat systemic issues of uterine receptivity, particularly in patients for whom IVF using competent blastocysts results in multiple unproductive IVF cycles.

For women under the age of 35, in the absence of any prior known etiology for either partner, the American Society for Reproductive Medicine currently defines infertility as the inability to conceive a successful pregnancy after at least one year of regular unprotected intercourse. Among the many potential physiological causes of female infertility are defects at the level of embryo implantation( 1 ). These can be caused by failure of embryo adhesion and attachment to the uterine luminal epithelium( 2 ), failure of embryo invasion of the uterine epithelium into the stromal layer( 3 ), or failure of the uterine stroma to differentiate into uterine decidua( 4 – 6 ) to support the pregnancy until the development of the placenta. Because the hormone used to clinically diagnose pregnancy, human chorionic gonadotropin( 7 ), is not detectable until shortly after the implantation stage, infertility at this step of the process presents clinically as rarely if ever producing a positive pregnancy test despite long term attempts to conceive. It is also likely a cause of serial failed in vitro fertilization (IVF) cycles, due to the inability of the uterus to accept a healthy blastocyst despite exogenous administration of appropriate hormones. The financial( 8 ) and timeline burden( 9 ) of multiple failed IVF cycles, often after years of failed natural attempts to conceive, is high, the mental health burden significant( 10 ), and the emotional burden immeasurable. Couples experiencing this subtype of infertility would be spared some part of this burden if the cause of their infertility were identifiable and ideally ultimately treatable. Recent advances in genetics, especially in the form of mouse models, have delineated many previously unappreciated genetic networks regulating the implantation phase of pregnancy( 11 – 22 ). The window of implantation in mice is established by the sequential secretion of ovarian hormones post coitus( 23 – 25 ). Prior to the formation of corpora lutea at 2.5 days post coitus (dpc), the uterus is under the influence of pre-ovulatory estrogen. The production of progesterone from the corpora lutea shifts cell proliferation from the uterine epithelia to the underlying stromal cells. On the morning of 3.5dpc, a nidatory estrogen surge on the progesterone-primed uterus starts the window of implantation. Only during this short time-window (<24 hours), the uterus is receptive to incoming blastocysts. A complex genetic network controls uterine receptivity. Genes encoding critically required proteins during pregnancy, and therefore genes whose dysregulation have consequences to fertility, have been rapidly identified in succession by leaders in the field of reproductive genetics( 26 ). Uncovering the function of such genes allows the development of both clearer diagnostic tests, as well as more functional and patient-specific treatment options. Many of the regulators of pregnancy are also regulators of embryonic development – transcription factors, morphogens, and their receptors act together to create exquisite signaling cascades to coordinate processes that guide towards differentiation. Conditional knockout mouse models make possible the testing of a gene’s function in independent tissues, even when that gene is required for life. In this study, a female-reproductive tissue-specific knockout mouse model of one such vital transcription factor, Sall1 , unveils its critical roles at multiple levels of pregnancy. Sall1 encodes a relatively ubiquitously expressed zinc-finger transcription factor( 27 ) that can act as either a repressor( 28 ) or activator( 29 ) of downstream gene expression, depending on the tissue and process in question. In humans, SALL1 is extensively studied as the gene whose truncation mutations cause Townes-Brocks Syndrome, an autosomal dominant condition often resulting in congenital malformations of multiple tissues including the ears, hands, anus, genitals, and notably for patient survival, the heart and kidneys( 30 – 34 ). In mouse models, knock out of Sall1 across the full body is lethal during the perinatal period, as it results in kidney agenesis( 35 ). Conditional models knocking out Sall1 in specific tissues further clarified its roles across renal development( 36 – 38 ), neurogenesis( 39 , 40 ), and microglia( 41 ). Additionally, recent work from our group clarified its roles in the development of genitalia( 42 ). Because so many roles for SALL1 have been identified across embryonic development, including lower urogenital tract development, and because of the notable overlap between regulators of embryonic development and regulators of early pregnancy, we hypothesized that SALL1 could play roles in the various tissue remodeling processes required for successful embryo implantation.