Deregulated P4 signaling in the pre-implantation Gata2^d/d uterus

To determine if the reduced PGR expression in Gata2^d/d mice impaired P4 signaling, we assayed the expression of PGR target genes in OVX Gata2^d/d and Gata2^f/f uterus 6 hours following P4 (1 mg, sc) or vehicle (oil) treatment. Gata2^f/f mice showed an increase in the expression of the PGR regulated genes amphiregulin (Areg), cytochrome P450, family 26, subfamily A, polypeptide 1 (Cyp26a1), and Indian Hedgehog (Ihh) (Fig. 2A-C). In contrast, Gata2^d/d mice showed a complete loss of P4-dependent induction of Areg and Cyp26a1 (Fig. 2 A and B) as well as reduced Ihh expression at both baseline and in response to P4 (Fig. 2C). These results indicate that the P4 signaling is attenuated on PGR-dependent marker genes in Gata2^d/d mice.

Open in a separate window

Figure 2

Impaired P4 signaling in Gata2 deficient uterus

OVX mice were treated with vehicle (oil) or 1 mg P4 for 6 hrs. (A-C) mRNA levels of PGR target genes assayed by qRT-PCR. N = 3 for each group. (D-E) Gene expression profiling by microarray assays on denoted treatments and genotypes. (D) Venn diagram of numbers of P4 responsive genes in denoted genotypes. (E) Magnitude of P4 responsiveness on expression of the common P4 responsive genes between denoted genotypes in (D). (F-G) Genome-wide expression profiles of Gata2 dependent genes in the presence of P4. (F) Heatmap of hierarchically clustered profiles. Low expression values are blue and high expression values are yellow. N=3 for each group. (G) Predicted regulators of Gata2 dependent genes by the Ingenuity Pathway Analysis. Error bars represent standard error of the mean. *, p < 0.05; **, p< 0.01; ***, p < 0.001 by student's t-test. See also Figure S2 and Tables S1, S2, S3 and S7.

The global impact of Gata2 deficiency on P4 signaling was assayed by conducting microarray analysis on the uteri of OVX Gata2^f/f and Gata2^d/d mice with P4 treatment. Comparison between P4 vs. vehicle treated groups identified 1861 P4 response genes (Fig. 2D, green box and Table S1) in Gata2^f/f mice, while only 139 genes responded to P4 stimulation in Gata2^d/d mice (Fig 2D, red box and Table S2). Strikingly, P4 failed to regulate 1814 (97%) of P4 response genes in Gata2^d/d mice (Fig. 2D, Venn diagram). Moreover, of the 47 P4 response genes that remained responsive to P4 stimulation in the Gata2^d/d background, many genes exhibit diminished magnitude on change of gene expression in response to P4 (Fig. 2E). We also identified 2461 probes that represent the Gata2 dependent transcription profile in the uterus by comparing expression arrays of Gata2^f/f and Gata2^d/d mice under P4 treatment (Fig. 2F and Table S3). Based on this profile, the Ingenuity Pathway Analysis, IPA, predicted inhibition of P4 signaling (activation z-score −4.05) and activation of P4 antagonist mifepristone (activation z-score 2.09) (Fig. 2G). These results collectively indicate the requirement of Gata2 to sustain PGR expression for P4 signaling.

Suppression of uterine E2 signaling and E2 driven epithelial proliferation is a hallmark of P4-induced preparation for subsequent embryo implantation (Wetendorf and DeMayo, 2014). Compared with Gata2^f/f uteri, Gata2^d/d Day 3.5 pseudopregnancy uteri showed an increased level of phosphorylation at serine 188 on ESR1 as well as increased expression of the target Mucin 1, transmembrane (MUC1) (Fig. S2A) with no significant change in Esr1 mRNA levels (Fig. S2B). Thus, the Gata2^d/d mice exhibited higher E2 signaling activities without elevated ESR1 levels. We also observed altered proliferation of the uterine epithelia in response to heightened E2 signaling after loss of Gata2. Normally epithelial proliferation is high at day 2.5 and diminished at day 3.5 (Fig S2C and D). In contrast, uterine epithelial proliferation at day 3.5 remained as high as at day 2.5 in the Gata2^d/d mice (Fig. S2 C and D). Taken together, these results demonstrate uninhibited E2 signaling and reduced P4 signaling as a result of Gata2 deficiency.

The mechanism by which Gata2 regulates PGR expression was investigated by determining the ability of Gata2 to regulate the expression of PGR at the level of Pgr gene transcription. Pgr mRNA levels was reduced in Gata2^d/d uteri compared to controls (Fig. 3A). ChIP-qPCR analysis further shows enriched GATA2 occupancy at a region that consists of 3 predicted GATA binding motifs upstream and proximal to the Pgr promoter in uteri of OVX Gata2^f/f mice 1 hour post P4 treatment (Fig. 3 B and C). This enrichment was absent in Gata2^d/d mice and in the exon 4 negative control region devoid of GATA binding motifs (Fig. 3 B and C), supporting the finding of GATA2 occupancy at the Pgr locus. Results from luciferase reporter analysis in cultured epithelial cells showed that the Pgr promoter was stimulated with the full length Gata2 construct and with the amino zinc finger deletion (ΔNT) construct, but not with the carboxyl zinc finger deletion (ΔCT) construct (Fig. 3D). This is consistent with a previous finding that the carboxyl zinc finger domain is involved in transcriptional regulation by GATA2 (Tong et al., 2000). The in vitro GATA2ΔCT mutant protein results support the in vivo functional loss of our Gata2 mutant mice in which the carboxyl zinc finger domain is deleted upon cre-mediated excision. Interestingly, deleting the amino zinc finger domain of GATA2 did not affect transactivation activity by GATA2 in the Pgr promoter (Fig. 3D), despite a previous report that both zinc fingers are important for suppressing the PPARγ2 promoter (Tong et al., 2000). Our results demonstrate that GATA2 can enhance Pgr promoter activity with its carboxyl zinc-finger domain. Collectively, we find that Pgr is a direct downstream target of GATA2.

Open in a separate window

Figure 3

Binding and regulation of Pgr by GATA2

(A) Pgr mRNA levels by qRT-PCR in uteri of OVX mice acutely treated with vehicle for 6 hrs. (B) Gata2 binding sites in the Pgr promoter. Red denotes GATA2 binding motifs. PCR primers used for ChIP-qPCR analysis are listed here. (C) ChIP-qPCR analysis of GATA2 occupancy on Pgr proximal promoter in uteri 1 hr after P4 treatment in OVX mice. (D) Luciferase reporter analysis of Pgr promoter activities in response to various GATA2 expression vectors. GATA2, full length GATA2; ΔNT, GATA2 N-terminal zinc-finger deleted mutant; ΔCT, GATA2 C-terminal zinc-finger deleted mutant. Error bars represent standard error of the mean. *, p < 0.05; **, p< 0.01; ***, p < 0.001 by student's t-test.

Significant overlap between GATA2 and PGR binding on uterine P4 response genes

Due to the observed interdependence between PGR and GATA2 expression in uterine P4 signaling, we hypothesized that both transcription factors are required for the regulation of target gene expression. This hypothesis was tested by conducting ChIP-Seq analysis for GATA2 on uterine chromatin isolated from OVX C57BL/6 mice treated with 1 mg P4 sc for 1 hour and immunoprecipitated for GATA2 (Rubel et al., 2012b). Whole uterine analysis resulted in over 22×10⁶ tags mapping to unique locations in the mouse genome. A model-based analysis peak finding algorithm (MACS) (Zhang et al., 2008) was utilized to normalize immunoprecipitated chromatin against uterine input generating a false discovery rate of essentially zero. This high-confidence cutoff produced 46,183 GATA2 binding intervals. Analysis for DNA motifs in the GATA2 cistrome revealed the expected GATA binding motifs as well as PGR binding motifs within the GATA2-occupying intervals (Fig. 4A and B). To investigate whether this PGR binding motif enrichment in GATA2 occupying sites is uterus-specific, we compared GATA2 occupancy patterns and motif enrichment between two distinct tissue types, the uterus and HPC7 hematopoietic precursor cells (Wilson et al., 2010). Among 46183 uterine and 9243 HPC7 cell GATA2 occupying sites, GATA2 exhibited distinct cistrome profiles with 96.7% and 83.3% sites unique to uterus and HPC7 cells, respectively (Fig. S3A). This finding is similar with a previous study in which GATA2 cistromes are significantly different between precursor cells and differentiated mast cells (Calero-Nieto et al., 2014). Motif analyses further revealed that the uterine GATA2 cistrome contained binding motifs enriched for nuclear receptors, such as PGR, glucocorticoid receptor (GR) and androgen receptor (AR) among others (Table S4), while such enrichment was not present in the HPC7 GATA2 cistrome (Table S4). Similarly, motifs of Sex Determining Region Y Box (SOX) factors are also enriched in the uterine, but not the HPC7 cell GATA2 cistrome (Table S4). It is worthy to note that, SOX17 is also expressed in the uterine epithelial cells (Rubel et al., 2012b), where GATA2 is expressed, and functions in the process of embryo implantation (Hirate et al., 2016). These findings revealed a uterine-specific GATA2 occupancy pattern that is associated with the presence of binding motifs of transcription factors, such as PGR, and SOX17 that have been known to regulate uterine function.

Open in a separate window

Figure 4

Binding of GATA2 and PGR at loci of uterine P4 responsive genes

(A, B) Enriched GATA2 and PGR binding motifs at GATA2 binding intervals that were identified by ChIP-Seq in OVX mouse uteri 6 hrs after P4 treatment. (C) Venn diagrams of P4 regulated genes that contain binding sites for PGR or GATA2 within +/− 25 kb of gene boundaries. (D) Occupancy of PGR and GATA2 in putative enhancers for Sox17 and Ihh genes in P4 treated uteri. Blue indicates PGR while red indicates GATA2 occupancy. Brackets mark the regions used for luciferase reporter analysis. (E, F) Luciferase reporter analyses of cotransfection analysis of denoted cis-acting elements in HEC-1A cells. Cotransfection with empty vector, PGR-A, PGR-B or GATA2 expression vectors treated with vehicle or 10⁻⁸ M R5020 for 24 hrs. Experiments were performed in triplicate. Error bars represent standard error of the mean. *, p < 0.05; ***, p < 0.001 by student's t-test. See also Figures S3 and S4 and Table S4.

Since PGR binding motifs are enriched in the GATA2 cistrome, we aligned the GATA2 location data with previous PGR ChIP-Seq performed for mouse uterus (Rubel et al., 2012b). This PGR ChIP-Seq data was also conducted on the uteri of OVX mice treated with 1 mg P4 for 1 hour. Through overlaying the GATA2- and PGR-occupying genes (defined by having at least one ChIP interval within 25 kb of annotated gene boundaries) with P4 responsive genes derived from the microarray results, we found GATA2 occupancy in 1478 (79%) and PGR occupancy in 1013 (54%) P4 response genes (Fig. 4C). Strikingly, GATA2 and PGR occupying sites are present together on 935 (50%) genes regulated by P4 (Fig. 4C). We validated the GATA2 occupancy on 13 genes (15 locations total) that are previously identified to contain PGR binding sites (Rubel et al., 2012b). All validated sites were within known P4 responsive genes and several of these GATA2 occupying locations directly overlapped with those of PGR occupying sites (Fig. S3B). In summary, these findings suggest a major regulatory mechanism of P4 response genes jointly by GATA2 and PGR.

Joint regulation of transcription by GATA2 and PGR

The functionality of the overlap of PGR and GATA2 intervals in the regulation of gene expression was tested by conducting luciferase reporter analysis on putative enhancers containing overlapping PGR and GATA2 binding sites from chromatin locations in two P4 target genes, Sox17 and Ihh, (Fig. 4D) (Lee et al., 2006; Rubel et al., 2012b). The GATA2-PGR intervals were sub-cloned into a luciferase vector containing a minimal promoter and then cotransfected with combinations of vectors expressing PGR-A, PGR-B and GATA2 to evaluate the possible regulatory preferences of these factors on these putative enhancers of P4 target genes. The results showed that luciferase activities under the control of these putative enhancers increased in response to progestin (R5020) treatment in a Pgr isoform specific manner (Fig. 4 E and F). No induction of these reporter genes was observed with cotransfection of the Gata2 alone. However, when Gata2 was cotransfected with the Pgr isoforms, these reporter plasmids showed a significant induction in luciferase activity in the presence of R5020 treatment (Fig. 4 E and F). These results are consistent with our hypothesis that PGR and GATA2 can coordinate the expression of these target genes and with the observation of GATA2 regulating PGR response genes in vivo (Fig. 2C and S3C).

The ChIP-Seq result also identified multiple GATA2 occupying sites in the Pgr locus (Fig. S4A), including the proximal promoter region that contains predicted GATA binding motif and is corresponding to the Pgr promoter fragment in the luciferase construct (Fig. S4B). Interestingly, the Pgr proximal promoter region exhibited a high GATA2 and low PGR occupying profile (Fig. S4B). This occupancy pattern is consistent with the luciferase promoter assay result where GATA2 alone can enhance the Pgr promoter activity (Fig. 3D), which is in contrast with the GATA2/PGR co-occupying enhancers where GATA2 co-activated transcription with PGR (Fig. 4 D-F). Our results demonstrate that GATA2 employs two mechanisms to regulate P4 signaling through directly controlling Pgr gene transcription as well as co-regulating PGR downstream gene expression with PGR.

Progesterone independent function of Gata2 in suppressing p63

The role of Gata2 in hormone independent gene regulation of uterine function was investigated by comparing the expression profiles of Gata2^d/d and Gata2^f/f treated only with vehicle. Microarray analysis revealed 426 genes upregulated and 535 downregulated (Table S5). IPA revealed enrichment in cancer and proliferation of keratinocytes/epidermal cells (Fig. 5A), as evidenced by the Gata2^d/d uteri exhibiting ectopic expression of TRP63 in a layer of cells, morphologically similar to basal cells, underneath the columnar luminal epithelia (Fig. 5D). Since TRP63 is a molecular switch for squamous metaplasia (Koster et al., 2004) associated with in utero exposure to the synthetic E2 diethylstilbestrol (DES) (Franco et al.; Goldberg and Falcone, 1999; Sassoon, 1999), we hypothesized that Gata2 deficiency may increase susceptibility in developing squamous metaplasia in response to E2 signaling. Long-term treatment of OVX mice with E2 resulted in development of focal regions of TRP63-positive cells into squamous cell metaplasia as early as 2 weeks post treatment (Fig. 5F), with Trp63 expression over time (Fig. 5G and S5). An increase in expression of keratin genes including Krt5 and Krt15, in Gata2^f/f mice (Fig. 5 H and I) was also observed. These changes were not observed in the control mice (Fig. 5E). In summary, our results revealed a Gata2 dependent and P4 independent mechanism in suppression of Trp63 expression and a role in squamous metaplasia. This demonstrates that Gata2 is critical for the uterine epithelium to maintain its differentiated state.

Open in a separate window

Figure 5

P4-independent functions of Gata2

(A) Enrichment of functional annotation in differentially expressed genes between Gata2^f/f and Gata2^d/d uteri of OVX, vehicle-treated mice by IPA analysis. (B) mRNA levels of TRP63 in uterus of OVX mice. (C, D) Immunostaining of TRP63 in uteri of OVX mice (control horn of decidualization treatment). Scale bar denotes 100 uM. (E, F) Immunostaining of TRP63 in uteri of OVX mice treated with E2 for 2 wks. (G) qRT-PCR of uterine tissues of OVX mice treated with E2. N = 6 (Gata2^f/f, vehicle, 2 wks), 6 (Gata2^d/d, vehicle, 2 wks), 6 (Gata2^f/f, E2, 2 wks), 6 (Gata2^d/d, E2, 2 wks), 6 (Gata2^f/f, vehicle, 4 wks), 6 (Gata2^d/d, vehicle, 4 wks), 4 (Gata2^f/f, E2, 4 wks), 6 (Gata2^d/d, E2, 4 wks), 6 (Gata2^f/f, vehicle, 2 mo.), 4 (Gata2^d/d, vehicle, 2 mo.), 6 (Gata2^f/f, E2, 2 mo.), 5 (Gata2^d/d, E2, 2 mo.). (H, I) qRT-PCR of uterine tissues of OVX mice treated with E2 for 2 mo. Error bars represent standard error of the mean. *, p < 0.05; **, p< 0.01; ***, p < 0.001 by student's t-test. See also Figure S5 and Table S5.

Microarray analysis

Detailed sample preparation can be found in the supplemental information. All experiments were performed in triplicate with independent pools of RNA. The Partek Genomics Suite 6.6 software (Partek Inc., St. Louis, MO) was utilized to process raw data from CEL files. The Robust Multichip Analysis (RMA) algorithm with quantile for normalization and log2 transformation was applied to generate signal values of all samples. The one-way ANOVA model was used to compare expression profiles from different groups. Differentially expressed genes were defined using the filters of ANOVA unadjusted p value <0.01 and absolute fold change >1.3.

ChIP-Seq and data analysis

GATA2 and Input ChIPs were performed by Active Motif, Inc. (Carlsbad, CA) on mouse uteri treated with P4 and with GATA2 antibody (sc-9008, Santa Cruz) as described in supplemental information. The model-based analysis of ChIP-sequencing (MACS) peak finding algorithm was used to normalize ChIP against Input control (Zhang et al., 2008). Specifically, a P value of 10⁻¹⁰ was used with this software to identify ChIP peaks in this work by comparing them to input. Genes associated with intervals were assessed using two increasingly less stringent requirements; if it was within 10 kb and 25 kb upstream or downstream of a gene, it was counted. Analysis of enriched motifs and CEAS were performed using the Cistrome Analysis Pipeline software (http://cistrome.org/ap/) under default settings (Liu et al., 2011) or using the HOMER (Heinz et al., 2010). For gene functional classifications, the web application tool Database for Annotation, Visualization, and Integrated Discovery (DAVID, http://david.abcc.ncifcrf.gov/) running default settings was used (Huang da et al., 2009).

Signature analysis

The publicly available human recurrent implantation failure (RIF) array dataset GSE58144 was scored for manifestation of the mouse model-derived GATA2 signature, using published methods (Qin et al., 2013). In brief, a gene signature score (t-score) was defined for each human sample as the 2-sided t-statistic comparison of the high GATA2-signature with the low GATA2-signature genes’ expression profile. Where multiple probes in GSE58144 referred to the same gene, the probe with the highest variation was chosen to represent the gene. For each gene, the profiles were centered to the median across samples. Expression pattern was displayed as heatmap using Partek Genomics Suite 6.6 software.

Path analysis on human data

To test the relationships among variables proposed in the research models, structural equation modeling was conducted to indicate the strength of influence among variables by getting an overall fit of model with the data. Goodness-of-fit tests were performed to determine whether the research models should be accepted or rejected. Models were analyzed using Mplus software version 7.11 (Muthen & Muthen, Los Angeles, CA). Based on the research models, the significance of direct and indirect relationships among variables was examined. All observed data were included when fitting the models. There was no missing data in the dataset.

The fit of the models was assessed using several methods including the root mean square error of approximation (RMSEA), along with a 90% confidence interval, the standard root mean square residual (SRMR), the Comparative Fit Index (CFI), and the Tucker-Lewis Fit Index (TLI) to determine the extent to which the relationships existing in the data are consistent with those proposed by the models (Hu and Bentler, 1998, 1999; MacCallum et al., 1996). To interpret the RMSEA, the general rule of thumb is that values <.05 indicate close fit, values between .05 and .10 indicate marginal fit, and values >.10 indicate poor fit (MacCallum et al., 1996). For both CFI and TLI, a value of 1 indicates perfect fit and the values >.9 indicate adequate fit (Hu and Bentler, 1998, 1999). Also, values <.08 for SRMR indicate a very good fit between the model and the data.

Histology

Alkaline Phosphatase Activity Assay was performed on frozen sections. Immunohistochemistry analysis was carried out in paraformaldehyde fixed, paraffin sections with PGR (A0098, Dako) and TP63 (sc-8431, Santa Cruz) antibodies. Detail information is in the supplemental information.

We thank Dr. Paul Labhart of Active Motif for performing the ChIP-Seq and the Genetically Engineered Mouse and the Microarray Cores at BCM. We also thank Mrs. Janet DeMayo for manuscript editing. Funding: This research was supported in part by the Intramural Research Program of the National Institute of Environmental Health Sciences: Project Z1AES103311-01 (FJD). This work was supported by National Institutes of Health grants R01HD042311 (JPL), R01HD057873 (JWJ), R01CA77530 (JPL), U54HD0077495 (FJD) and U54HD055787 (FJD) and U54HD28934 (University of Virginia Center for Research in Reproduction Ligand Assay and Analysis Core) through a cooperative agreement as part of the Specialized Cooperative Centers Program in Reproduction and Infertility Research.