Pten and Dicer1 loss in the mouse uterus causes poorly differentiated endometrial adenocarcinoma.

Pgr +/+ (control), Pten cKO [ Pten conditional knockout ( Pgr cre/ + ; Pten flox/flox ; Dicer1 +/+ )], and dcKO [double conditional knockout ( Pgr cre/ + ; Pten flox/flox ; Dicer1 flox/flox )] were generated. Pten-Dicer het [ Pten - Dicer heterozygous ( Pgr cre/ + ; Pten flox/flox ; Dicer1 flox/+ )] were studied in parallel, but they were similar to Pten cKO ( Supplementary Figures S3 and Table S2 ). Examination of the reproductive tract of 6-month-old female mice revealed bulky endometrial tumors in both Pten cKO and dcKO mice ( Figure 1A ). Uteri from Pten cKO mice revealed large tortuous uteri with bulky nodules of solid tumors protruding onto the surface, but homogeneous in gross morphology along both uterine horns. Uteri from dcKO mice also contained solid, bulky tumors. Histological analyses revealed adenocarcinoma invading through the myometrium of the uterus and into the adnexa, but without gross metastatic disease outside the female reproductive tract ( Supplementary Figure S4 - S5 ). Poorly-differentiated adenocarcinoma was found in both the uterus and adnexa in 6-month-old dcKO mice ( Supplementary Figure S6 and Supplementary Table S3 ). Due to morbidity, only a small number of mice at this time point were examined ( Supplementary Table S3 ). Based on criteria for humane endpoints, Pten cKO and dcKO mice showed decreased survival compared to Pgr +/+ ( Figure 1B and Supplementary Figure S7 ). Pten cKO had a median survival of 240 days, similar to published results [ 21 ]. The median survival for dcKO mice was 327 days. Many survival mice had necrotic tumors throughout the reproductive tract, making an accurate assessment of tumor histology impossible ( Supplementary Table S3 ). By 12 weeks, dcKO uteri were nearly twice as large as Pgr +/+ but remained significantly smaller than Pten cKO ( Figure 1C ). Uterine morphology, survival studies, and uterine weight suggested distinct phenotypes between Pten cKO and dcKO uteri. Similar to published [ 21 ], a majority of the Pten cKO uteri contained well-differentiated adenocarcinoma, with glands in columnar architecture and nuclear pseudostratification but with retained polarity in many of the nuclei, mild nuclear atypia, and little or no solid pattern ( Figure 2A ). Endometrial adenocarcinoma was defined as confluent growth of the endometrium, with cribriform pattern and/or expansion of the endometrium into the endometrial stroma, or invasion into the surrounding myometrium [ 32 ]. With deletion of Dicer1 , the frequency of poorly-differentiated adenocarcinoma increased significantly (χ 2 = 12.8; P = 0.00035). Most of the dcKO mice exhibited poorly-differentiated adenocarcinoma ( Supplementary Table S3 ), as evidenced by poorly formed malignant glands, solid microscopic architecture, or tubulocystic architecture ( Figure 2A ). Epithelial cells were stained with cytokeratin-8 and myometrium with smooth muscle actin ( Supplementary Figure S8 ). Myometrial invasion was defined as tumor between the two layers of the myometrium. Similar to published results [ 21 ], 85% of tumors from Pten cKO uteri exhibited invasion through the myometrium. All dcKO adenocarcinomas at 12 weeks showed myometrial invasion ( Supplementary Table S3 ). Pgr Cre/ + mice have homologous recombination of floxed alleles within the columnar epithelial cells of the oviduct and the corpus luteum of the ovary when stimulated with gonadotropins [ 33 ]. Older mice of all genotypes ( i.e., survival and 6-month old) exhibited adenocarcinoma of the uterus, oviduct, and ovaries. However, the origin of the tumor, driven by Cre -recombination events in the oviduct and/or ovary versus metastatic disease from the uterus, was unable to be discerned. The oviducts and ovaries of 12-week-old mice were carefully examined for any signs of cancer-associated histological changes. Pten cKO oviducts frequently showed epithelial atypia in the oviducts defined as nuclear stratification (>1 cell thickness epithelium, loss of nuclear polarity) and nuclear atypia in the form of enlarged, rounded nuclei ( Figure 2B ). Adenocarcinoma was infrequently discovered in the oviducts of Pten cKO mice. Adenocarcinoma was found in nearly 50% of the oviducts of dcKO mice. The histology was similar to uterine histology and adenocarcinoma was located on the inside and outside of normal oviducts ( Figure 2B ), consistent with metastatic spread. Pten cKO ovaries showed evidence of both normal follicular development and metastatic disease, as cancer appeared to have invaded at the ovarian hilum ( Figure 2C ). Ovaries of dcKO mice were most frequently normal ( Figure 2C - D ). Examination of the reproductive tracts of dcKO mice showed evidence of direct extension from the uterus to the oviduct with normal, cancer-free, distal fimbria and ovary ( Figure 2D ). Evidence of metastatic disease to the adnexa was observed in approximately 50% of Pten cKO and dcKO mice. Histology was largely similar between the uterus and adnexa in many dcKO mice ( Supplementary Table S3 and Supplementary Figure S9 ). Epithelial cells from dcKO uterine tumors exhibited large, often rounded, and hyperchromatic nuclei and pale-staining to clear cytoplasm ( Figure 2A and Supplementary Figure S4 ). To better describe the poorly-differentiated adenocarcinomas with clear cytoplasm in dcKO mice, immunohistochemical markers of clear-cell adenocarcinoma, Napsin A and hepatocyte nuclear factor 1 homeobox B (HNF1B) [ 34 , 35 ], were evaluated. Poorly-differentiated adenocarcinomas from dcKO mice exhibited malignant epithelial cells with high-frequency, high-intensity cytoplasmic Napsin A staining. Malignant epithelial cells from dcKO mice more frequently stained positive for nuclear HNF1B compared to Pten cKO tumors ( Figure 3 and Supplementary Figure S10 ). To determine the effects of ovarian insufficiency in dcKO mice, both ovaries were removed, and uterine weight and histology were assessed. Gross uterine weight was lower in ovariectomized (ovex) mice compared to intact mice ( Supplementary Figure S11A ). However, the frequency of poorly-differentiated adenocarcinoma was not decreased in dcKO mice ( Supplementary Table S3 ). To examine the effects of long-term progesterone therapy, mice underwent removal of ovaries at six weeks and were randomized to either placebo or progesterone pellets at eight weeks. Treatment lasted 60 days. Treatment with progesterone did not alter the frequency of poorly-differentiated adenocarcinomas in ovex dcKO mice ( Supplementary Table S3 ). Carefully selected women with well-differentiated, early-stage endometrioid adenocarcinoma can be treated with progesterone to preserve fertility options [ 22 ]. Intact mice were randomized at six weeks to either placebo or progesterone pellet for 60 days. Treatment with progesterone did not significantly alter the frequency of poorly-differentiated endometrial adenocarcinomas in intact mice ( Supplementary Table S3 ). Treatment of intact mice with long-term progesterone decreased the rate of metastatic adenocarcinoma to the adnexa, but it was only statistically significantly decreased for Pten cKO mice ( Supplementary Figure S11B and Table S3 ). To create an in vitro human model, CRISPR-Cas9 was used to delete DICER1. Ishikawa cells were chosen because they are PTEN mutant, derived from a well-differentiated adenocarcinoma, and exhibit high expression of E-cadherin (CDH1) consistent with epithelial cell differentiation [ 22 , 36 ]. Western blot confirmed knockdown or knockout of DICER1 ( Figure 4A ). DICER1 deletion in Ishikawa cells led to decreased cellular proliferation and reduced colony formation in soft agar ( Figure 4B - C ), consistent with the smaller-sized uterine tumors in dcKO mice ( Figure 1C ). Examination of CDH1 confirmed high expression at the RNA level by qPCR and the protein level by immunocytochemistry in DICER1 +/+ Ishikawa cells as previously reported [ 36 ]. DICER1 −/− Ishikawa cells exhibited a 4-fold decrease in CDH1 RNA expression ( Figure 4D ) and a significant decline in CDH1 expression by immunocytochemistry ( Figure 4E and Supplementary Figure S12 ), consistent with a more poorly-differentiated phenotype. DICER1 −/− cells exhibited increased expression of the clear cell markers, Napsin A and HNF1B, by immunocytochemistry ( Supplementary Figure S12 ). Poly-A RNA sequencing (RNA-seq) was performed. Principal Component (PC) analysis showed distinct clustering. DICER1 +/+ , DICER1 +/− , and DICER1 −/− differed significantly in PC2, while DICER1 +/+ and DICER1 +/− were most similar in PC1 ( Figure 4F ). Determination of differentially expressed genes between DICER1 +/+ and DICER1 −/− revealed 2444 unique protein-coding genes with log2fold change >ǀ1ǀ significantly dysregulated [false discovery rate (FDR)<0.05, Supplementary Table S4 ], 1154 genes upregulated, and 1290 genes downregulated ( Supplementary Figure S13A ). Poly-A RNA sequencing on mouse uteri was then performed. Three-week uteri were selected because: 1) Pgr Cre/ + -mediated deletion occurs primarily in the luminal and glandular epithelium [ 33 ]; 2) the limited molecular contributions of steroid hormones in pre-adolescent female mice; 3) the ability to explore early molecular contributions; and 4) the similar ratio of cell populations ( i.e., epithelium, stroma, and myometrium) across genotypes. As early as three weeks, Pten cKO and dcKO uteri showed adenocarcinoma ( Supplementary Table S3 ) with activation of phospho-AKT limited to the epithelium ( Supplementary Figure S14 ). PC analysis showed distinct clustering. However, Pten cKO and Pten-Dicer het mRNA profiles were most similar to each other and only varied slightly in PC2. Both dcKO and Pgr +/+ varied significantly in PC1, and dcKO varied in PC2 ( Figure 4G ). Because all experimental mice have a deletion of Pten, transcriptomic profiles were compared to Pten cKO. Determination of differentially expressed genes between dcKO and Pten cKO showed 1635 unique protein-coding genes with log2fold change >ǀ1ǀ significantly dysregulated (FDR<0.01, Supplementary Figure S14B ), 1059 genes upregulated, and 576 genes downregulated ( Supplementary Table S4 ). Gene set enrichment analysis with estrogen-responsive gene sets [ 37 ] showed no significant overlap with mouse dcKO or DICER1 −/− transcriptomic profile ( Supplementary Table S5 ). The genes downregulated in mouse dcKO showed a trend ( P =0.087) towards enrichment in progesterone-responsive gene sets [ 38 ], but no significant enrichment in upregulated progesterone-responsive genes. Both up- and downregulated genes in the DICER1 −/− transcriptomic profile showed significant enrichment in progesterone responsive genes ( Supplementary Table S5 ). Web Gestalt [ 30 ] revealed that the upregulated DICER1 −/− transcriptomic profile was enriched in ephrin-receptor signaling pathway genes ( Supplementary Table S4 ). Studies have shown that high expression of EPH receptor A2 ( EPHA2) in endometrial cancer was associated with high tumor grade, reduced survival, high microvessel density, and high Ki67 expression [ 39 , 40 ]. QPCR of DICER1 −/− cells showed that EPHA2 was 7.6-fold upregulated. Other EPH receptors, including EPH receptor A7 ( EPHA7 ) and EPH receptor B2 ( EPHB2 ), were 8-10-fold upregulated ( Figure 5A ). Further, both the non-canonical WNT signaling molecule, WNT5B , and p21 (RAC1) activated kinase 3 ( PAK3 ) downstream effectors of EPHA2 [ 41 , 42 ], were 9.6- and 9.3-fold upregulated. LIM kinase 2 ( LIMK2 ), an essential molecule in p21 signal transduction and involved in castration-resistant prostate cancer [ 43 ], was similarly upregulated ( Figure 5A ). QPCR on dcKO uteri showed that EPH receptor B1 ( Ephb1 ) was 3.5-fold upregulated, and Pak3 was 5.3-fold upregulated. Cofilin 2 ( Cfl2 ), a potential downstream effector of LIMK2 and PAK3 [ 44 ], was 2.7-fold upregulated ( Figure 5B ). Pathway analysis on the upregulated dcKO uteri transcriptomic profile revealed hedgehog signaling and prostaglandin synthesis genes were upregulated ( Supplementary Table S4 ). QPCR of dcKO uteri showed Indian hedgehog signaling molecule ( Ihh ) 20-fold upregulated and patched 1 ( Ptch1 ) and GLI family zinc finger 1 and 2 ( Gli1 and Gli2 ) were 5-fold upregulated ( Figure 5C ). High CFL2 expression was associated with poor survival in gastric cancer [ 45 ]. Both high CFL2 and EPHA2 have been associated with increased proliferation in cancers [ 39 , 40 , 45 ]. Examination of dcKO uteri showed increased expression of Ki67 compared to Pten cKO uteri ( Figure 5D ). To examine the effects of DICER1 deletion on miRNA expression in endometrial cancer, small RNA sequencing was performed on the same samples as poly-A RNA sequencing. Differentially expressed miRNAs (FDR<10%) in mouse dcKO and DICER1 −/− cells are shown in Supplementary Table S6 . DICER1 −/− cells showed 193, and dcKO uteri revealed 38 mature miRNAs downregulated. There were 22 mature miRNAs downregulated in common. MiRNAs target mRNAs for repression through binding to the 3’ untranslated region of target genes [ 46 ]. Integration of upregulated miRNA-target mRNAs with downregulated miRNAs from DICER1 −/− cells was performed using miRTarBase. MiRTarBase uses curated data from over 10,000 publications to make validated predictions on miRNA-target interactions (MTIs) [ 47 , 48 ]. Out of 1154 upregulated genes in DICER1 −/− cells, 885 were MTIs for the 193 downregulated miRNAs. Supplementary Table S6 lists the 7023 miRNA:mRNA MTIs mapped to 885 unique genes. The genes targeted by miRNAs were involved in the following pathways: microRNAs in cancer, glycosaminoglycan biosynthesis, ephrin-receptor signaling, and transforming growth factor-beta (TGFβ) signaling ( Supplementary Table S6 ). Many TGFβ signaling genes were also dysregulated in dcKO uteri ( Supplementary Table S6 and Figure S15 ). Human DICER1 -mutant TCGA endometrial cancers were enriched in miRNA-target genes from five miRNA families: let-7-5p, miR-17-5p, miR-16-5p, miR-29-3p, and miR-101-3p [ 8 ]. Mature miRNAs from each of these five miRNA families were significantly downregulated in the small RNA sequencing datasets from DICER1 −/− cells ( Supplementary Table S6 ). Out of 1154 upregulated genes in DICER1 −/− cells, 347 unique genes were MTIs for these specific five DICER1 -mutant family members. Pathways enriched included microRNAs in cancer, glycosaminoglycan biosynthesis, ephrin-receptor signaling, renal cell carcinoma, and TGFβ signaling ( Supplementary Table S6 ). Recent work has shown that dysregulation of TGFβ signaling in the mouse uterus results in endometrial hyperproliferation and cancer [ 49 , 50 ]. Members of the TGFβ signaling pathways were targeted by a number of dysregulated miRNAs. Let-7b-5p, miR-16-5p, and miR-26b-5p were mature miRNAs downregulated in both DICER1 −/− cells and dcKO uteri and predicted to target TGFβ signaling genes. By QPCR, DICER1 −/− cells had significant downregulation of let-7b-5p, and let-7b-5p was also 2-fold down-regulated in dcKO uteri. QPCR showed a 71-fold downregulation of miR-16-5p in DICER1 −/− cells, and 2.1-fold downregulation for miR-16-5p in mouse. QPCR showed 939-fold downregulation of miR-26b-5p in human and 1.7-fold downregulation in the dcKO uteri ( Figure 6A - B ). These miRNAs were MTIs for the TGFβ signaling pathway genes: cyclin-dependent kinase 6 ( CDK6 ), bone morphogenetic protein receptor 1B ( BMPR1B ), and transforming growth factor-beta receptor 3 ( TGFBR3 , Supplementary Table S6 ). MiR-26b-5p was also a predicted target for CDK6 , transforming growth factor-beta 1 induced transcript 1 ( TGFB1I1 ), and growth differentiation factor 10 ( GDF10 , Supplementary Table S6 ). QPCR showed higher than 6.5-fold upregulation of CDK6 in human and 2.5-fold upregulation of Cdk6 in the mouse. BMPR1B showed a greater than 14-fold upregulation in DICER1 −/− cells. TGFBR3 showed 6.5-fold upregulation in human, and 4-fold upregulation in dcKO mouse. TGFB1I1 is thought to mediate progesterone resistance in endometriosis [ 51 , 52 ]. TGFB1I1 was 8-fold upregulated in human and 2.9-fold upregulated in mouse. GDF10 was 8.4-fold upregulated in human and 2.3-fold upregulated in mice ( Figure 6C - D ). Other TGFβ signaling genes that are predicted targets of other downregulated miRNAs were upregulated ( Supplementary Figure S15 ). Four publicly available datasets of human endometrial cancer were used to determine the frequency of concurrent mutations in PTEN and DICER1 ( Table 1 ). The TCGA PanCancer dataset ( n =509) contains 75.4% endometrioid, 20.6% serous, and 4% mixed endometrioid/serous [ 53 ]. The MSK-IMPACT dataset ( n =189) contains 40.6% endometrioid, 23.9% serous, 18.8% carcinosarcoma, and 6.6% clear cell [ 54 ]. Available via cBioPortal [ 55 , 56 ] is an additional dataset ( n =16) that contains only uterine clear-cell adenocarcinoma [ 57 ]. An additional dataset ( n =30) not available via cBioPortal also contains only uterine clear cell adenocarcinoma [ 58 ]. These databases showed that mutations in PTEN and DICER1 occur concurrently in up to 11.2% of endometrial cancer samples. Concurrent mutations in PTEN and DICER1 in endometrioid endometrial cancers were frequently grade 3 and disease progression was frequent. Up to 10% of clear cell tumors contained concurrent mutations in PTEN and DICER1 . These tumors were frequently grade 2 without disease progression. Many of the DICER1 mutations (26/65 samples) were known oncogenic changes, including truncating mutations, frame shift deletions, or recurrent oncogenic hotspot mutations ( Supplementary Tables S7 ). Eight tumors were categorized as neither DNA-polymerase-ε ultra-mutated nor microsatellite instability-hypermutated tumors ( Supplementary Tables S7 ). A majority of those contained known DICER1 recurrent oncogenic hotspot mutations.

All animals were bred and kept under standard conditions. Pgr cre/ + ; Dicer1 flox/flox [ 19 ] mice were crossbred to Pgr +/+ ; Pten flox/flox mice [ 21 ] and maintained on a C57BL/6J;129S5/Brd mixed hybrid background. Crossbreeding and genotyping was as described ( Supplementary Methods and Figure S1 ). For survival studies, mice were caged and examined twice weekly. Mice were euthanized at humane endpoints of loss of normal behavior or ambulation, obvious distress, not eating or drinking, loss of 20% of body weight, tumor size greater than 10% of body weight, palpable tumor >1.5 cm in any diameter, ulcerations, roughened hair coat or hunchback, or poor body conditioning score [ 23 ]. At the time points listed, mice were sacrificed; body and uterine weight were recorded. One uterine horn, oviduct, and ovary were snap-frozen, and the other, fixed, and processing, paraffin embedding, and histological and immunohistochemistry staining were as described [ 19 , 24 ]. All histology was interpreted by clinical pathologists (R.E.E. and R.R.B.). FIGO grading was used. A binary system of low-grade (FIGO grade 1 or 2) and high-grade (FIGO grade 3) was used as clinically recommended [ 25 ]. Primary antibodies and conditions are listed ( Supplementary Table S1 ). Immunohistochemistry staining was imaged on a Zeiss Axio Lab.A1 microscope (Oberkochen, Germany) and scored based on the frequency of staining as the percentage of positive staining cells and confirmed with ImageJ using FIJI [ 26 , 27 ]. Immunofluorescence staining was imaged on an EVOS FL Cell Imaging System (Thermo Fisher Scientific, Waltham, MA). Due to the extreme gross and histological differences noted during dissection, investigators were not blinded to genotype. Sample size, statistical analysis, and randomization are described in Supplementary Methods . Mice underwent ovariectomy (ovex) at six weeks. Mice were divided into treatment groups: progesterone pellet (Innovative Research of America, Sarasota, Florida, 25mg/60-day release pellet) or placebo pellet, for 60 days. Ishikawa cells were obtained from the Cytogenetics and Cell Authentication Core at MD Anderson Cancer Center (Houston, TX) and maintained in RPMI1640 (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (Atlanta Biologicals, Minneapolis, MN) and 1% penicillin/streptomycin (Thermo Fisher Scientific). Cell line authentication was confirmed using a short tandem repeat (STR) marker profile (IDEXX BioAnalytics, Westbrook, ME) within six months of experiments and tested for mycoplasma contamination monthly (MycoAlert Plus Mycoplasma Detection Kit, Lonza, Switzerland). DICER1 was deleted in Ishikawa cells using CRISPR-Cas9 ( Supplementary Methods and Figure S2 ). Western blot, proliferation, immunocytochemistry, and colony formation assay are described in Supplementary Methods . Total RNA was extracted from uteri and Ishikawa cells, DNase treated, and assessed for quality [ 19 ]. Details of next-generation sequencing are provided in Supplementary Methods . The mRNA reads were mapped against GRCm38 and GRCh38.p5. Initial mRNA differential expression analysis was carried out using the DESeq2 package (version 1.10.1) in R/Bioconductor (R version 3.2.1) [ 28 ]. MiRNA reads were mapped to miRBase V21. Initial miRNA differential expression analysis was carried out as described [ 29 ]. Dysregulated canonical pathways and gene ontology enrichment analysis were determined from Web Gestalt (WEB-based Gene SeT AnaLysis Toolkit) [ 30 ]. Integrated miRNA:mRNA analyses were carried out using miRTarBase [ 31 ]. Reverse transcription and real-time qPCR were performed on a QuantStudio3 Real-Time PCR System (Thermo Fisher Scientific) and analyzed [ 19 ]. Supplementary Table S1 lists TaqMan assays and Sybr primer sequences. Transcriptomic data have been deposited into the Gene Expression Omnibus (accession pending). Statistical analysis was performed with one of the following: Student’s t -test, Fisher’s exact test, Chi-squared test, multiple t -test, 2-way ANOVA, or log-rank (Mantel-Cox) pairwise comparison with Bonferroni correction. Differences between groups were determined by P <0.05. Data presented are mean and standard error of the mean (SEM) or 95% confidence interval (CI). Statistical analyses were conducted using the InStat package for Prism8 (GraphPad, San Diego, CA).

Although nearly all common cancers have improved cancer survival since the 1970s, endometrial cancer is one of the few cancers with an increased death rate [ 1 ]. Importantly, studies have shown that the increase in incidence is mostly in high-grade or non-endometrioid histologies, including serous and clear-cell adenocarcinomas [ 59 ]. The highest risk factor linked to disease recurrence and death is a high tumor grade [ 5 , 6 ]. New models are needed for high-grade tumors, as these are the tumors leading to the most substantial mortality [ 3 , 4 ]. In the present study, we describe the first mouse model of poorly-differentiated endometrial adenocarcinoma. Because DICER1 mutations are frequently found in ultra-mutated tumors, DICER1 mutations in endometrial adenocarcinoma were thought to be passenger mutations. However, recent genomic profiling of tumors through TCGA PanCancer and MSK-IMPACT showed enrichment of biallelic DICER1 hotspot mutations in endometrial but not other cancers [ 8 ]. Importantly, this analysis removed POLE ultra-mutated samples [ 8 ]. DICER1 mutants play a functional role in endometrial cancers, by affecting miRNA processing, miRNA expression, and miRNA-target gene expression [ 8 , 11 - 15 ]. These studies showed that DICER1 mutations are not passenger mutations in endometrial cancer and suggest that DICER1 may play a unique yet significant role in endometrial cancer. The frequency of concurrent mutations in PTEN and DICER1 is 2-11% ( Table 1 ), but the number of DICER1 loss-of-function and recurrent hotspot mutations comprise nearly 50% of those concurrent mutations, advocating that our mouse model with deletion of Dicer1 , has relevance to human disease. Furthermore, on the molecular level, deregulated miRNA-target genes for let-7-5p and miR-16-5p family members were significantly enriched in published DICER1 -mutant TCGA datasets [ 8 ] and the present study (both dcKO mouse and DICER1 −/− cells; Figure 6 ). Thus, the dcKO mouse model and DICER1 −/− cells represent relevant models. In humans, germline variants in DICER1 are rare and are thought to give rise to the DICER1 Syndrome. Humans with DICER1 Syndrome have an increased risk of certain cancers, but endometrial cancers are not one of those cancers [ 60 ]. However, somatic loss-of-function mutations in DICER1 have been discovered in uterine adenosarcomas and rhabdomyosarcomas [ 61 , 62 ], suggesting that both the mesenchymal and the epithelial component plays a role in development of cancers. Women with Cowden Syndrome are at increased risk of endometrial cancer due to germline PTEN mutations [ 63 ], but the role of DICER1 mutations have not been published in women with Cowden Syndrome. Other mouse models suggest that deletion of Pten and Dicer1 is not sufficient for malignant transformation of all epithelial lineages. In a mouse model of prostate cancer, conditional deletion of Pten and loss of one allele of Dicer1 resulted in a more aggressive phenotype than loss of both alleles of Dicer1 [ 64 ]. Similarly, loss of two alleles of Dicer1 in an oncogenic Kras mouse model of lung cancer was protective against aggressive lung cancer [ 65 ]. Loss of Pten and one allele of Dicer1 in the uterus did not show a significantly different phenotype than Pten cKO alone, while the dcKO showed a poorly-differentiated phenotype. Thus, loss of Pten and Dicer1 leading to endometrial cancer seems to be tissue specific in our mouse model and in human datasets from TCGA and MSK-IMPACT [ 8 ]. The impact of epithelial-stromal communication and specific mutations in smaller subsets of endometrial cells deserve deeper study. Clinically, miR-16-5p and let-7-5p represent important molecular targets for cancer therapy. Treatment with liposomal miR-16 is being tested for mesothelioma [ 66 ] and is frequently considered a tumor suppressor [ 67 ]. Let-7b-5p is a likely tumor suppressor in endometrial cancer [ 68 ]. Also, targets of these miRNA molecules may be very important clinically. For example, CDK6 is a predicted target for multiple robustly downregulated miRNA molecules ( Supplementary Table S6 ). Recent studies suggest that high expression of CDK6 may be a biomarker for poor prognosis endometrial cancers [ 69 ]. Specifically, the high expression of CDK6 with high Ki67 expression was associated with shorter progression-free survival in women with endometrial cancer [ 69 ]. Intact dcKO uterine tumors showed a significantly high frequency of Ki67-staining cells ( Figure 5D ). The use of CDK6 inhibitors, such as palbociclib, has shown promise in vitro and in orthotopic and xenograft models, particularly when tumors are PTEN mutant [ 70 ]. Additionally, high expression of LIMK2 ( Figure 5 ) suggests activation of novel signaling cascades that have not yet been explored in endometrial cancer. Studies indicate that the use of T56-LIMKi, a LIM kinase inhibitor, leads to decreased proliferation and migration in cancer cells [ 44 ]. Future studies will focus on testing these precision therapies in our mouse and cell models. Poorly-differentiation endometrial tumors frequently develop within an atrophic, or postmenopausal endometrium [ 22 ]. The Pgr Cre/ + ; Dicer1 flox/flox uteri represent atrophic endometrium [ 19 ]. Additional deletion of Pten in this background of Dicer1 deletion leads to poorly-differentiated adenocarcinoma. While hyper-proliferation of endometrial epithelium from endogenous or exogenous estradiol with mutation of PTEN is a biologically plausible mechanism for the development of endometrial cancer [ 22 ], little is known about the mechanism of development of endometrial cancer from atrophic endometrium. In most dcKO mice, the tumors seemed to arise from individual areas of the uterus rather than homogenously along the entire uterus. This mechanism is consistent with type II endometrial tumors primarily derived from heterogeneous endometrial polyps in women [ 22 ]. The reason for the heterogeneous nature of the tumor development along the uterine horns is unknown. We suspect that a small portion of cells, potentially a hormone-independent stem cell population, maybe playing a role as decreased expression of DICER1 in human endometrial cancers is associated with increased expression of stem cell markers [ 16 ]. These studies will be the focus of future investigations. Further studies are needed on Dicer1 in the adnexa. High-grade serous ovarian cancer develops in the oviduct of mice with mesenchymal deletion of both Pten and Dicer1 [ 71 ] . Moreover, activation of WNT/β-catenin through deletion of adenomatous polyposis coli ( Apc ) with Pgr Cre/ + showed frequent endometrioid ovarian cancer that began in the epithelial cells of the oviduct [ 72 ]. However, we believe that Pten cKO, Pten-Dicer het, and dcKO mice have metastatic endometrial cancer to the adnexa. There are a number of reasons for this conclusion: 1) the frequency of tumors in the adnexa is of lower penetrance than the uterus; 2) the histology of the adnexal tumors is similar to the uterine tumors; 3) the location of cancer in the adnexa ( i.e., ovarian hilum or adjacent to normal) suggests invasion or direct spread; 4) the large endometrial tumors with myometrial invasion; 5) multinodular location of adenocarcinoma along the oviduct with some areas of normal oviduct; and 6) no evidence of typical spread of oviductal or ovarian cancer such as peritoneal metastasis or ascites. Because Pgr Cre leads to recombination in the oviduct and granulosa cells of the ovary [ 33 , 72 ], it is possible that either the oviduct and/or the ovary are concurrent primary tumors of lower penetrance that exhibit similar histology as the uterine adenocarcinoma. Further, metastatic disease lands in a unique “soil” and may exhibit different histology. Thus, the subtle differences between adnexal and uterine disease may develop from the unique tumor microenvironment. The poorly-differentiated adenocarcinomas of dcKO mice also exhibit features of clear-cell adenocarcinoma. Clinically, clear-cell endometrial adenocarcinomas are challenging to distinguish from high-grade endometrioid endometrial adenocarcinomas with clear-cell changes [ 35 ]. In our mouse model, dcKO uteri show high-intensity staining for Napsin A, which is a sensitive marker for clear-cell carcinoma [ 34 ]. Additionally, dcKO tumors are HNF1B positive, another important marker of clear cells [ 34 , 35 ]. However, neither Napsin A nor HNF1B is specific for endometrial clear cell adenocarcinoma [ 73 , 74 ]. The use of the dcKO model as a model for clear-cell adenocarcinoma would be improved with comparison to transcriptomic studies from clear-cell endometrial adenocarcinoma from women. However, the only existing datasets of clear-cell endometrial cancer profiled five samples, used older microarray technology, and profiled a limited number of genes [ 75 ]. We are encouraged that our gene sets were enriched in let-7 and miR-16 target genes similar to derepressed gene sets from DICER1 -mutant TCGA tumors [ 8 ]. Further studies are needed to transcriptomically interrogate pathologically well-characterized clear-cell endometrial adenocarcinoma from women for comparison to dcKO tumors. In women, poorly-differentiated endometrial adenocarcinomas are associated with worse survival [ 3 - 6 ]. Our dcKO mice had poorly-differentiated endometrial adenocarcinomas that were associated with worse survival than normal control mice, but improved survival compared to Pten cKO mice. This discrepancy is largely based on the humane endpoint for euthanasia being palpable tumor of >1.5 cm. In women who are surgically staged, the size of the uterus is not clinically useful, but higher tumor grade and deeper myometrial invasion portend worse outcomes [ 5 , 6 ]. Importantly, dcKO mice had adenocarcinomas with increased myometrial invasion, a correlate of higher stage in women, in addition to poorly-differentiated tumors. We speculate that if death were an endpoint in our mouse survival studies that dcKO mice would succumb to disease faster than Pten cKO. We speculate that DICER1 loss may result in more indolent, smaller, yet more aggressive tumors. This speculation is supported by published data from women that showed that downregulation of DICER1 expression was associated with increased myometrial invasion, higher FIGO grade tumors, and increase disease recurrence [ 16 - 18 ]. Our data, as presented, show that the dcKO mouse represents poorly-differentiated adenocarcinoma. Mouse models of endometrial cancer that recapitulate human disease represent translational tools for a better understanding of the aggressive disease. We anticipate that this model will be highly relevant, not only to study the molecular characteristics of the rarer forms of human endometrial cancer but also to study the preclinical development of therapeutics.

Endometrial cancer will account for 65,620 new cases and 12,590 deaths in 2020 in the United States [ 1 ]. Well-differentiated [Fédération Internationale de Gynécologie Obstétrique (FIGO), grade 1] endometrioid endometrial adenocarcinoma represents the most common histology and has a 5-year survival for all stages that approaches 83% [ 2 ]. High-risk histologic endometrial adenocarcinoma, such as poorly-differentiated (FIGO grade 3) endometrioid, serous, and clear-cell adenocarcinoma, encompasses 15% of endometrial cancer cases but accounts for nearly 50% of deaths [ 3 , 4 ]. High tumor grade is the most significant risk factor for disease recurrence and subsequent death [ 5 , 6 ]. In endometrial cancer, DICER1 is a prominent cancer-driver gene [ 7 ]. Hotspot biallelic mutations in DICER1 , the endoribonuclease responsible for miRNA genesis, are enriched in endometrial adenocarcinomas over other cancers profiled in both TCGA (The Cancer Genome Atlas) PanCancer and MSK-IMPACT (Memorial Sloan-Kettering Integrated Mutation Profiling and Actionable Cancer Targets) studies [ 8 ]. DICER1 dices the precursor stem-loop miRNA into two single-stranded mature miRNAs [ 9 , 10 ]. Studies have shown that recurrent DICER1 hotspot mutations from ovarian and endometrial cancers resulted in altered miRNA processing [ 8 , 11 - 15 ]. Further, downregulation of DICER1 expression was associated with features of more aggressive endometrial adenocarcinoma, including myometrial invasion, high FIGO grade tumors, and disease recurrence [ 16 - 18 ]. Uterine-specific deletion of Dicer1 using Pgr Cre/ + ; Dicer1 flox/flox mice resulted in a uterine phenotype containing a single layer of luminal epithelium, lack of glandular epithelium, and thin to zero endometrial stroma, consistent with atrophic endometrium from postmenopausal women [ 19 ]. Loss or mutation of the tumor suppressor, phosphatase and tensin homolog ( PTEN ), occurs in more than 80% of endometrial adenocarcinomas [ 20 ]. Mice with conditional deletion of Pten in the uterus ( Pgr Cre/ + ; Pten flox/flox ) have high-penetrance (88.9%) well-differentiated endometrial adenocarcinoma by 90 days [ 21 ]. To study the role of Dicer1 in endometrial cancer, Dicer1 was conditionally deleted in Pgr Cre/ + ; Pten flox/flox mice. Because downregulation of DICER1 expression was clinically associated with high FIGO grade endometrial cancer in women [ 16 ], we hypothesized that deletion of Dicer1 with deletion of Pten in the mouse uterus would result in high FIGO grade endometrial adenocarcinomas. In this study, we present data that underscores the impact of appropriate Dicer1 function in endometrial adenocarcinoma. Importantly, we describe the development of hormone-independent, poorly-differentiated adenocarcinoma with adnexal metastasis that arises from an atrophic endometrium. High-fidelity mouse models of poorly-differentiated endometrial adenocarcinomas have not been described. Further, xenograft models have shown limited success for high-risk histologic endometrial adenocarcinoma types [ 22 ]. Our findings indicate a relationship between appropriate miRNA expression and molecular signaling pathways in the development of poorly-differentiated endometrial adenocarcinoma.