Roles of Deletion of Arid1a, a Tumor Suppressor, in Mouse Ovarian Tumorigenesis

Abstract

The chromatin remodeling gene, ARID1A, has been implied as a tumor suppressor, and its somatic inactivating mutations occur in a wide variety of human cancers, most frequently in ovarian and uterine endometrioid and ovarian clear cell carcinomas. Tumors with ARID1A mutations also frequently harbor PTEN or PIK3CA mutations, suggesting their collaboration in tumorigenesis. Here, we used a conditional knockout mouse model in which Arid1a and Pten were deleted either individually or in combination in the mouse ovarian surface epithelium. After 6 months, 59.1% of mice with Arid1a and Pten double knockout developed ovarian endometrioid or undifferentiated carcinoma, whereas the remaining mice showed hyperplasia of ovarian surface epithelium. In contrast, 52 mice with homozygous or heterozygous deletion in either Arid1a or Pten did not develop ovarian lesions. These results demonstrate that inactivation of Arid1a alone is insufficient for tumor initiation but it requires additional genetic alteration(s) such as Pten deletion to drive tumorigenesis.

ARID1A is one of the most commonly mutated chromatin remodeling genes in human cancer, especially in endometrium-related neoplasms, including ovarian clear cell carcinoma, ovarian endometrioid carcinoma, and uterine endometrioid carcinoma (1–5). The great majority of ARID1A mutations are frameshift or nonsense mutations that occur throughout the coding region, suggesting that ARID1A functions as a tumor suppressor. Indeed, restoring expression of wild-type ARID1A in ovarian cancer cells harboring deleterious ARID1A mutations suppresses cellular proliferation and growth of tumor xenografts in mice (6). Conversely, silencing ARID1A expression in nontransformed epithelial cells enhances cellular proliferation and tumorigenicity (6–9). It has been demonstrated that loss of ARID1A expression, presumably due to inactivating mutations, occurs in the precursor stage of ovarian clear cell and ovarian endometrioid carcinomas (2,10,11), and loss of ARID1A immunoreactivity correlates with tumor progression in endometrial carcinomas (12). These studies indicate that ARID1A functions as a tumor suppressor and may participate in both tumor initiation and progression of endometrium-related neoplasms. However, it remains unclear whether ARID1A inactivation per se is sufficient to initiate tumor development in vivo or whether tumor initiation requires molecular collaboration with aberration in other cancer pathway(s). Here, we sought to determine the role of Arid1a in tumor development using Arid1a conditional knockout mice.

We first tested whether deletion of Arid1a alone in mouse ovarian surface epithelium (MOSE) was sufficient to initiate tumor development. All of the animal procedures were approved by the Johns Hopkins University Animal Care Committee. Detailed information on the mouse strains is given in the Supplementary Methods (available online). We performed intra-ovarian bursa injection of AdCre virus on left ovaries in 29 Arid1a ^flox/flox mice (see Supplementary Methods, available online). The right ovaries were not injected and were used as controls. Twelve months after AdCre administration, no gross or microscopic lesions were observed in either AdCre-injected or control ovaries (Table 1), suggesting that additional molecular changes are required to collaborate with Arid1a inactivation in transforming MOSE. It has been noted that ARID1A mutations frequently co-occur with either PIK3CA mutations or PTEN mutations in human tumor samples (5,13,14); therefore, in this study, we tested whether Arid1a deletion in collaboration with abnormalities in the PTEN/PI3K/AKT pathway facilitated ovarian tumorigenesis. We used conditional Pten knockout mice to test the above hypothesis because 84% of uterine endometrioid carcinomas with ARID1A mutations contain PTEN mutation [based on The Cancer Genome Atlas data (5)] and 49% of ovarian endometrioid carcinomas with ARID1A mutations contain PTEN mutation (14), Pten knockout mice are commercially available (15), and Pten deletion in combination with Apc or Kras mutations induces ovarian cancers after intrabursal administration of AdCre virus (16,17).

Table 1.

Summary of tumor phenotypes of ovary in mice with different genotypes

Genotype	Time after AdCre injection	No. of mice with			Total No. of mice	P*
		Tumor	Hyperplasia	No lesion
Arid1a flox/flox; Pten flox/flox	≥6 Months	13	9	0	22	NA
Arid1a +/flox; Pten flox/flox	12 Months	0	0	5	5	<.001
Arid1a flox/flox; Pten +/flox	12 Months	0	0	5	5	<.001
Arid1a +/flox; Pten +/flox	12 Months	0	0	13	13	<.001
Arid1a flox/flox; Pten +/+	12 Months	0	0	29	29	<.001

* Statistical analyses were performed using two-tailed Fisher exact test to compare the frequency of ovarian tumor formation after AdCre administration between genotype Arid1a ^flox/flox;Pten ^flox/flox and other genotypes (Arid1a ^+/flox;Pten ^flox/flox, Arid1a ^flox/flox;Pten ^+/flox, Arid1a ^+/flox;Pten ^+/flox, and Arid1a ^flox/flox;Pten ^+/+). NA = not applicable.

Arid1a ^flox/flox mice were crossed with Pten ^flox/flox mice to generate Arid1a and Pten conditional double knockout mice (see Supplementary Methods, available online). The littermates with double knockout were identifed by genotyping polymerase chain reaction (see Supplementary Methods and Supplementary Figure 1A, available online). We observed MOSE hyperplasia as early as 2 months after AdCre adminstration in the ovaries treated with AdCre virus (Figure 1A; Supplementary Table 1, available online). Six months after AdCre injection, six of 13 (46.2%) Arid1a ^flox/flox ;Pten ^flox/flox mice developed ovarian tumors, and the remaining seven mice developed MOSE hyperplasia (Figure 1A; Supplementary Table 1, available online). Eight to 9 months after knockout of both genes, seven of nine (77.8%) mice developed ovarian tumors, and the remaining two developed MOSE hyperplasia (Figure 1A; Supplementary Table 1, available online). In this study, we did not observe transformation of potential stem cells in the ovarian hilum (data not shown). In contrast, none of the littermates without double knockout developed ovarian neoplasms or hyperplasia (Table 1). In all, by 9 months after AdCre injection, 13 of 22 (59.1%) Arid1a/Pten double knockout mice had developed ovarian tumors, whereas none of the 52 mice without double knockout had developed ovarian lesions (P < .001, two-tailed Fisher exact test; P < .05 was considered statistically significant) (Table 1). These results are consistent with previous reports demonstrating that deletion of Pten alone in MOSE is insufficient to generate neoplasms and at least two molecular genetic alterations are needed for tumorigenesis (16–20).

Figure 1.

Ovarian lesions in Arid1a ^flox/flox;Pten ^flox/flox mice. A) Schematic of the phenotypes of all 30 Arid1a and Pten conditional double knockout mice. B) Gross and histologic features of an endometrioid carcinoma (left panels) and an undifferentiated carcinoma (right panel). Scale bar = 1cm. C) Immunostaining of markers in a representative undifferentiated carcinoma. Scale bar = 250 μm. D) Immunostaining of markers in a representative endometrioid carcinoma. Scale bar = 250 μm. CA = carcinoma; OSE = ovarian surface epithelium; SM = squamous metaplasia.

Histologically, five and eight of 13 ovarian neoplasms were endometrioid and undifferentiated types, respectively (Figure 1, A–D). Prominent squamous differentiation, which occurs in up to 50% of human ovarian endometrioid carcinoma (21), was detected in all mouse endometrioid carcinomas (Figure 1B). Of note, two endometrioid carcinomas also contained undifferentiated carcinoma, and morphological transition from endometrioid to undifferentiated carcinoma was evident (Supplementary Figure 2, available online). All pure endometrioid carcinomas were confined to ovaries, whereas the eight undifferentiated carcinomas and the two endometrioid carcinomas containing undifferentiated carcinoma had disseminated within the peritoneal cavity. In humans, ovarian and uterine undifferentiated carcinomas (22–25) are frequently associated with and clonally derived from low-grade endometrioid carcinoma (26). Unlike endometrioid carcinoma, undifferentiated carcinoma takes a highly aggressive clinical course. Hence, our mouse model appears biologically relevant and may, in fact, recapitulate late ovarian tumor progression.

Immunohistochemically, tumor cells were positive for the epithelial cell marker cytokeratin-8 and phospho-AKT (S473) but negative for ARID1A and PTEN (Figure 2, C and D). Western blot analysis confirmed the immunostaining result (Supplementary Figure 1B, available online). Detailed immunohistochemistry and western blot methods can be found in the Supplementary Methods (available online). For those 16 mice that did not develop ovarian tumors, we observed MOSE hyperplasia by hematoxylin and eosin staining (Figure 1A), which showed the same staining pattern as tumors (Supplementary Figure 3, available online).

Gene expression profiling analysis (see Supplementary Methods, available online) of the mouse Arid1a ^flox/flox; Pten ^flox/flox tumors were performed by referencing to a set of 128 discriminant genes (Supplementary Table 2, available online) selected from the dataset of human ovarian serous, endometrioid, clear cell, and mucinous carcinomas (16). We found that mouse tumors had the overall highest mean probability to be human endometrioid carcinoma as compared with other ovarian tumor subtypes. Specifically, four of the five representative mouse tumors were most likely to be endometrioid type, whereas one was serous type (Supplementary Table 3, available online).

ARID1A mutations are frequently obse rved not only in endometrioid carcinoma but also in clear cell carcinoma; however, we did not observe definitive morphological features characteristic of clear cell carcinoma (Figure 1; Supplementary Figure 2, available online). This might be expected because we codeleted Pten rather than expressing activating Pik3ca together with Arid1a deletion in MOSE. ARID1A and PIK3CA mutations frequently co-occur in human ovarian clear cell carcinomas (1,2), whereas ARID1A and PTEN mutations frequently co-occur in human endometrioid carcinomas (5,14). Moreover, PTEN has been reported to cooperate with p53 in a lipid phosphatase-independent manner (27). Thus, this AKT-independent mechanism may facilitate the development of endometrioid rather than clear cell carcinomas.

Although this study demonstrates new functional roles of Arid1a in mouse ovarian cancer initiation, several questions remain to be answered. For example, it would be of great interest to determine the effects of Arid1a deletion with other molecular genetic alterations that are common in human endometrioid carcinomas such as Pik3ca and Ctnnb1 activating mutations in mouse ovaries. Future studies should also focus on using this mouse model to determine the molecular contributors that propel the tumor progression from endometrioid carcinoma to undifferentiated carcinoma.

In summary, our results show that Arid1a deletion collaborates with Pten deletion in driving tumorigenesis of MOSE, whereas Arid1a deletion alone is insufficient to do so. Our results provide biological significance of molecular genetic changes involving ARID1A and PTEN that have been identified in human ovarian endometrioid carcinomas. This mouse model is potentially useful for future molecular studies of ARID1A in cancer and for preclinical studies of molecular targeted therapy for tumors with ARID1A mutations.

Funding

This study was supported by the National Institutes of Health/National Cancer Institute grants CA129080 and CA165807, National Institutes of Health/National Heart, Lung, and Blood Institute HL109054, and by the Ovarian Cancer Research Fund and the Tell Every Amazing Lady, Louisa M. McGregor Ovarian Cancer Foundation (POE/JHU/01.12).

The authors declare no conflict of interest. The funders did not have any involvement in the design of the study; the collection, analysis and interpretation of the data; the writing of the article; or the decision to submit the article for publication.