ARID1A Mutations and PI3K/AKT Pathway Alterations in Endometriosis and Endometriosis-Associated Ovarian Carcinomas

Endometriosis is a common gynecological disease affecting 6%–10% of women of reproductive age and is characterized by the presence of endometrial-like tissue in localizations outside of the uterine cavity as, e.g., endometriotic ovarian cysts. Mainly, two epithelial ovarian carcinoma subtypes, the ovarian clear cell carcinomas (OCCC) and the endometrioid ovarian carcinomas (EnOC), have been molecularly and epidemiologically linked to endometriosis. Mutations in the gene encoding the AT-rich interacting domain containing protein 1A (ARID1A) have been found to occur in high frequency in OCCC and EnOC. The majority of these mutations lead to a loss of expression of the ARID1A protein, which is a subunit of the SWI/SNF chromatin remodeling complex and considered as a bona fide tumor suppressor. ARID1A mutations frequently co-occur with mutations, leading to an activation of the phosphatidylinositol 3-kinase (PI3K)/AKT pathway, such as mutations in PIK3CA encoding the catalytic subunit, p110α, of PI3K. In combination with recent functional observations, these findings strongly suggest cooperating mechanisms between the two pathways. The occurrence of ARID1A mutations and alterations in the PI3K/AKT pathway in endometriosis and endometriosis-associated ovarian carcinomas, as well as the possible functional and clinical implications are discussed in this review.


Introduction
The identification of recurrent somatic mutations in endometriosis-associated ovarian cancer, in particular, AT-rich interacting domain containing protein 1A (ARID1A) mutations [1,2], provided the first molecular evidence of a direct pathogenic link between endometriosis and certain subtypes of ovarian carcinomas [3,4].
Endometriosis is a common gynecological inflammatory disease that affects at least 6%-10% of women of reproductive age and is characterized by the presence of endometrial-like tissue outside of the uterine cavity [5,6]. The prevalence of this disease is higher in women with abdominal pain, infertility or both, where it rises to an incidence rate of 35%-50% [7]. The ectopic endometrial-like tissue is most often found in the pelvic peritoneum, the ovaries and/or the rectovaginal septum, but can also involve uncommon localizations, such as the diaphragm, pleura, pericardium or, even, brain [7,8]. The exact pathogenesis of endometriosis has as yet not been fully elucidated. One of the most widely accepted theories is the retrograde menstruation of fragments of menstrual endometrium through the fallopian tubes that may explain the more frequent cases of peritoneal and ovarian endometriosis [5]. The typical clinical symptoms of the disease consist of dysmenorrhea, dyspareunia and infertility. Current treatment options are the surgical removal of endometriotic implants and hormonal suppressant drugs [8]. The overall risk of ovarian cancer associated with endometriosis can be regarded as generally very low [9], but is still increased compared to women not presenting with endometriosis [8].
Clinical suggestion for a causal relationship between endometriosis and ovarian cancer is not novel, and the possible pathogenic link had already been described at the beginning of the 20 th century by Sampson [10]. However, larger epidemiological studies showing evidence for a causal relationship between endometriosis and ovarian cancer were lacking until the end of the 20 th century [11].
In a large registry study among 20,686 women in Sweden who had been hospitalized for endometriosis, a significantly increased risk for ovarian cancer (standardized incidence ratio (SIR) 1.9, 95% confidence interval (CI) 1.3-2.8) was found after a mean of 11.4 years of follow-up. The risk of ovarian cancer was increased 2.5-fold in women with a follow-up of more than 10 years [12]. A larger Swedish register study involving 64,492 women with endometriosis confirmed an elevated risk of ovarian cancer in these patients with an SIR of 1.43 (95% CI 1. 19-1.71). The risk of ovarian cancer was considerably higher in patients with early diagnosed and long-standing endometriosis (SIR 2.01 and 2.23, respectively) [13]. Several other studies have also described an increased risk of ovarian cancer in women with endometriosis [14][15][16][17][18][19][20][21][22][23][24]. A large epidemiological pooled analysis of 13 case-control studies, including 13,226 controls and 7911 women with invasive ovarian cancer, investigated the frequency of self-reported endometriosis and found significantly increased risks for ovarian cancer in women with history of endometriosis with an odds ratio of 1.46 (95% CI 1.31-1.63, p < 0.0001) after stratifying for age, ethnic origin and adjustment for the duration of oral contraceptive use and parity. Importantly, a significant association of the histological subtypes of clear-cell (OR 3.05, 95% CI 2.43-3.84, p < 0.0001), endometrioid (OR 2.04, 95% CI 1.67-2.48, p < 0.0001), and low-grade serous ovarian carcinomas (OR 2.11, 95% CI 1.39-3.20, p < 0.0001) was found with endometriosis in this study. In contrast, there was no association of endometriosis with mucinous and high-grade serous cancer, as well as borderline tumors [25].
Ovarian cancer is the most lethal gynecological neoplasm and a very heterogeneous disease [26]. Based on extensive morphologic, immunohistochemical and molecular genetic analyses of different epithelial ovarian cancer subtypes, Kurman and Shih proposed that epithelial ovarian carcinomas can be divided in two groups: Type I ovarian carcinomas comprise clear cell carcinomas (OCCC), low-grade endometrioid carcinomas (EnOC), mucinous carcinomas and low-grade serous carcinomas. They tend to be rather slow-growing, low-grade neoplasms and often develop in a stepwise manner from visible precursor lesions. On a molecular level, they often share mutations in different genes, such as KRAS, BRAF, PTEN, PIK3CA, ARID1A, ERBB2, CTNNB1 and PPP2R1A. Type II ovarian carcinomas mainly cover high-grade serous carcinomas and are more frequent, as they represent approximately 75% of all ovarian carcinomas. These tumors are characterized by aggressive behavior, nearly ubiquitous presence of TP53 mutations and a high level of genetic instability, in contrast to type I tumors, which rarely show TP53 mutations [27][28][29][30][31].
There is growing evidence that epithelial ovarian carcinomas, unlike ovarian germline tumors, may often find their origin in non-ovarian tissue [32,33]. This is a shift of paradigms, since until recently, the largely accepted theory was that ovarian epithelial tumors arise from the single cell layer lining the ovarian surface, usually referred to as surface epithelium [34,35]. Hence, numerous molecular studies indicate that the precursor of serous ovarian carcinomas may be localized in the epithelium of the fallopian tubes in the form of a precursor lesion, called serous tubal intraepithelial carcinoma [27,32,[36][37][38][39][40][41][42]. Epidemiologic data are supporting this theory, since it has been reported that the risk of ovarian cancer decreases after tubal ligation or excision [43][44][45][46].
Two distinct histological subtypes of ovarian carcinomas, the OCCC and EnOC, have been directly associated with endometriosis through observational epidemiological studies [14,24,25,[47][48][49]. Atypical endometriosis has long been proposed as the histological precursor lesion of OCCC and EnOC [50,51]. A direct pathogenic link of endometriosis with OCCC and EnOC has been evidenced by the important study of Wiegand et al. demonstrating common truncating mutations and loss of protein expression of the ARID1A tumor suppressor gene in OCCC and contiguous atypical endometriosis [1]. Whole-exome sequencing performed independently by Jones et al. in eight OCCC confirmed frequent mutations of ARID1A, as well as PIK3CA, PP2R1A and KRAS. Validation of these results in 42 OCCC (the eight tumors of the discovery cohort and an additional 34 OCCC samples as a validation cohort) by Sanger sequencing reported the frequencies of these mutations at 57%, 40%, 7.1% and 4.7%, respectively [2].
OCCC is the second most common epithelial ovarian cancer subtype, and its prevalence has been shown to differ significantly between geographic regions [52]. The highest prevalence is described in Asian countries, especially in Japan, where it accounts for 15%-25% of epithelial ovarian cancer [53,54]. The prevalence in Europe and North America is significantly lower, where it accounts for 1%-13% of ovarian epithelial tumors [55][56][57][58]. Interestingly, the prevalence of endometriosis has also been reported to be higher in Asian women in some studies [59]. OCCC are reported to occur at an earlier age than serous ovarian cancer, with a median age at diagnosis of 55 years compared to 64 years [58]. Although low-stage OCCC have a relatively good prognosis, high-stage OCCC have a poorer prognosis than stage-matched high-grade serous ovarian carcinomas and are often characterized by resistance to standard carboplatin paclitaxel chemotherapy [53,54,[60][61][62][63][64]. Interestingly, OCCC are associated with a 2.5-fold higher incidence of clinically significant venous thromboembolism than in women with other histological types of epithelial ovarian cancer [65].
The prevalence of EnOC is estimated to be somewhat lower than in OCCC and accounts for 7%-13% of epithelial ovarian cancer [56,66,67]. There is an association of endometrioid ovarian carcinomas with uterine endometrial carcinomas in 15%-20% of cases [68][69][70][71][72]. Similarly to OCCC, EnOC are often diagnosed as early-stage disease in younger women (median age 47 years) in comparison with serous ovarian cancer and, often, become manifest with pelvic pain, a palpable abdominal mass, abnormal vaginal bleeding and/or newly developed or increased dysmenorrhea and dyspareunia [48,73].
Finally, seromucinous borderline tumors have also been associated with endometriosis. They are a rare subtype of mucinous borderline tumors and show a distinct non-gastrointestinal-type pattern [67]. These tumors coexist with endometriosis in 30%-70% and are likely to originate from endometriotic cysts [67,[74][75][76].
This review focuses on common genetic alterations in endometriosis and the endometriosis-associated OCCC and EnOC, with special emphasis on ARID1A mutations and alterations in the PI3K/AKT pathway and potential cooperative mechanisms between these pathways.

Background
The ARID1A gene encodes the AT-rich interacting domain containing protein 1A (ARID1A), also known inter alia as BAF250a or p270, which is part of a family of 15 proteins in humans that all contain a characteristic 100-amino acid DNA-binding ARID domain that binds in a sequence non-specific manner to DNA [77]. The ARID1 subfamily is a member of seven ARID subfamilies based on degree of homology of the ARID domain and the similarity between the highly variable non-ARID domain structures [78]. ARID1A and ARID1B are the only ARID1 subfamily members and are two mutually exclusive subunits of the SWI/SNF chromatin remodeling complex [79]. It is through this complex that ARID1A probably exerts its role as a tumor suppressor [80]. ARID1A is located at 1p36.11 [81] and encodes a large protein of approximately 250 kD that is expressed primarily (maybe exclusively) in the nucleus [77]. Its expression is cell-cycle dependent and is higher in the G 0 /G 1 -phase compared to the S-and G 2 /M-phases [82]. ARID1A encodes two protein isoforms (2285 and 2086 amino acids), but there is no actual knowledge about a functional difference of the two isoforms. It has been described that ARID1A is post-translationally modified through lysine acetylation and serine/threonine phosphorylation, which may potentially regulate protein expression or protein-protein interactions [77,83]. Although its role as a part of the SWI/SNF complex is the best studied interaction of the ARID1A protein, several other protein-protein interactions, such as interactions with p53 [84], with SMAD3 [84], with hormonal receptors, such as the glucocorticoid receptor [85], and others have been described. Due to its various interactions, the understanding about the functional effects of ARID1A mutations remains quite poor and may vary depending on different cell types. Other components of the ATP-dependent SWI/SNF chromatin remodeling complex, such as SMARCB1, have been found to be mutated in a wide variety of cancers [77]. The frequency of mutations of the different components of the SWI/SNF complex typically differs among different cancers [86].

ARID1A Mutations in Endometriosis-Associated Ovarian Carcinomas
A high frequency of ARID1A mutations has been detected in endometriosis-associated ovarian carcinomas. Studies undertaken by Wiegand et al. [1] and by Jones et al. [2] have reported ARID1A mutations in 46%-57% of OCCC and in 30% of EnOC. In the study of Wiegand et al., validated RNA-sequencing results of 19 OCCC samples (18 solid OCCC tumor samples and one OCCC cell line (TOV21G)) reported three somatic nonsense mutations, two somatic insertion/deletion mutations, one somatic missense mutation (found simultaneously in a sample containing an insertion mutation) and one gene rearrangement of ARID1A with the neighboring gene, ZDHHC18, with the fusion ends mapping to a homozygous deletion involving most of ARID1A. These data were verified in a mutation-validation cohort of 210 samples of ovarian carcinomas, including 101 OCCC, 33 EnOC and 76 high-grade serous ovarian carcinomas, as well as a second OCCC cell line (ES2) (cf. Table 1). All the 65 truncating ARID1A mutations that were found in 47 OCCC and 8 EnOC samples were somatic.
ARID1A is a large gene containing 20 exons, and the mutations found by Wiegand et al. were distributed evenly across the whole gene, with just a few recurrent mutations between different tumors. Most of the detected mutations were truncating (nonsense or frameshift) and correlated strongly with a loss of ARID1A protein expression in the immunohistochemistry (IHC). A total of 27 (73%) of the 37 OCCC and five (50%) of 10 EnOC samples with ARID1A mutations showed loss of protein expression in IHC, whereas four (11%) of 36 OCCC and two (9%) of 23 EnOC samples without ARID1A mutations were negative for ARID1A protein expression. Since data from exon resequencing and RNA sequencing presented excellent correlation, there was no suggestion for a relevant epigenetic silencing of ARID1A [1].
It is interesting that only approximately 30% of the OCCC showed homozygous mutations of ARID1A, whereas 73% of heterozygous mutated tumors showed a loss of protein expression without loss of heterozygosity. Similar observations in other cancer types, as well as results of in vitro studies, therefore, suggest a haploinsufficient tumor suppressor role for ARID1A [1,77].
Jones et al. described 32 mutations in ARID1A that were also distributed throughout the coding region of ARID1A and were all predicted to truncate the protein (nine nonsense and 23 insertion/deletion mutations). Both ARID1A alleles were affected through loss of heterozygosity or through biallelic mutations in 10 of the 24 tumors harboring ARID1A mutations [2].
The frequency of loss of ARID1A protein expression in OCCC and EnOC was verified in multiple studies [99][100][101][102][103][104] and found to be consistent with the initial observations [1,2]. Due to the high frequency of their occurrence, mutations of ARID1A are regarded to be one of the major genetic alterations in endometriosis-associated OCCC and EnOC [67].
An overview of the studies that analyzed ARID1A expression in ovarian cancer by mutational analysis and/or immunohistochemistry is given in Table 1.

Loss of ARID1A Expression in Endometriosis
Mutations of ARID1A have been demonstrated in atypical endometriosis that, in contrast to the adjacent OCCC tissue, was negative for HNF-1β and retained estrogen receptor expression. This indicates that ARID1A mutations are an early event in the pathogenesis of endometriosis-associated ovarian carcinomas. In contrast to tumor-adjacent atypical endometriosis, no mutations or loss of ARID1A expression were found in the distal non-atypical endometriotic tissue of the same patients [1].  This observation sustains the theory of ARID1A being a tumor suppressor in which loss of expression occurs in cell clones that are undergoing a process of precancerous alteration. However, it remains controversial at which stage of pathogenesis ARID1A mutations occur in endometriosis, i.e., if they are limited to atypical endometriosis or if they already occur in a low-frequent manner in non-atypical endometriosis or at the early transition stage from non-atypical to atypical endometriosis.
To date, ARID1A sequencing studies are lacking in non-carcinoma-related endometriosis, probably due to the fact that the occurrence of ARID1A mutations is expected to be low in endometriosis (considering that the relative risk to developing ovarian cancer during a lifetime is approximately 1.5%) and that ARID1A sequencing studies are technically challenging, due to the large size of the gene and the random distribution of the mutations along the gene. Nevertheless, immunohistochemical data from different studies indicated that loss of ARID1A expression is also observable in rare cases of non-atypical endometriosis, especially in endometriotic cysts of the ovary, also referred to as endometriomas [101,104,106]. An overview of the studies that investigated ARID1A expression in endometriosis with or without relation to ovarian carcinomas is given in Table 2.
Despite the good correlation between the immunohistochemical negative ARID1A expression and its mutations (cf. Section 2.4.), it is not definitely clarified if these observations are the result of ARID1A mutations or epigenetic regulation. Sequencing analyses and further functional studies are warranted to elucidate the exact time point of the occurrence of ARID1A mutations and their role in (atypical) endometriosis.

Correlation between ARID1A Mutations and Loss of ARID1A Expression in Immunohistochemistry
In addition to Wiegand et al. [1], other groups found a strong correlation between ARID1A mutations and loss of ARID1A protein expression in different tumor types. Guan et al. [88], in addition to a large IHC analysis of 995 tumor samples of various localizations, sequenced ARID1A in a total of 93 tumor samples and found 10 (40%) of 25 uterine endometrioid carcinomas mutated (all of them insertion/deletion or nonsense mutations), whereas none (0%) of 12 uterine serous carcinomas and none (0%) of 56 ovarian serous and mucinous carcinomas revealed somatic mutations in ARID1A. They correlated the ARID1A status and IHC in the 25 uterine endometrioid carcinomas and in 51 ovarian serous carcinomas and found a significant correlation of ARID1A expression with its mutational status (p = 0.0014). Interestingly, they found an immunohistochemical pattern of clonal loss in ARID1A mutated carcinomas, which was not found in ARID1A wild-type carcinomas. This clonal loss of ARID1A protein expression was observed in four cases, which were classified as immunohistochemically positive, but genetically harbored ARID1A mutations. When combining these cases with completely negative ones, they found an even stronger correlation of mutational status and IHC (p < 0.0001) [88]. In a smaller study performed in OCCC, there was a concordance of ARID1A immunohistochemistry in 91% of 12 OCCC cases with known ARID1A mutational status with a sensitivity of 100% and a specificity of 66% [99]. In gastric adenocarcinomas, Zang et al. found a reduced or absent ARID1A protein expression in 75% (6/8) of ARID1A mutated samples and a strong positivity for ARID1A protein expression in 100% (11/11) of the samples with wild-type ARID1A sequence [92]. Wang et al. sequenced ARID1A in a total of 109 gastric cancers. They found 32 samples with ARID1A mutations, which, as discovered in OCCC and EnOC, were truncating in the majority of cases (85%). Seventy-five percent (24/32) of gastric cancers with ARID1A mutations immunohistochemically exhibited a loss or substantially lower ARID1A protein expression compared to cancers with the wild-type gene (p < 0.001) [91]. Taken together, these results suggest that an immunohistochemical loss of ARID1A expression correlates well, although not perfectly, with truncating ARID1A mutations, which justifies its use as a surrogate marker for the underlying gene mutations [87,101,104,[107][108][109].

Introduction
Activation of the phosphatidylinositol 3-kinase (PI3K)/AKT pathway supports multiple mechanisms responsible for cancer progression, including proliferation, inhibition of apoptosis, cell adhesion and transformation. PI3K activation, e.g., by activating mutations of PIK3CA encoding the catalytic subunit, p110alpha, leads to an activation of AKT, a serine-threonine kinase, which is present in three different isoforms (AKT1-3) in human cancer and leads to increased cellular growth and survival of cancer cells [110]. The mammalian target of rapamycin complex 1 (mTORC1) is one of the major effectors downstream of AKT and is central in controlling cell growth and proliferation. mTORC1 was discovered through its inhibition by the drug, rapamycin [111,112]. Activation of the PI3K/AKT pathway is mainly effected by the activation of receptor tyrosine kinases and by somatic mutations in specific components of the signaling pathway. These mainly include loss of the tumor suppressor, PTEN, activating mutations of p110alpha (PIK3CA) and, less frequently, of the three isoforms of AKT1-3. In addition, amplifications of AKT1 and AKT2, as well as of PIK3CA have been described in some cancers, but seem to play a subordinate role compared to the other described mechanisms [110].

PI3K/AKT Pathway Activation in Endometriosis
Several studies have reported PI3K/AKT pathway activation in endometriosis [113][114][115][116][117][118][119][120]. It has been shown that the PI3K/AKT pathway regulates FOXO1 protein levels, a member of the forkhead-box O family and the decidua-specific gene IGF binding protein-1 (IGFBP-1), which are both involved in the decidualization of endometrial cells. Levels of phospho-AKT (Ser473) were consistently higher in endometriotic stromal cells, and overactivation of PI3K/AKT led to reduced decidualization in primary endometriotic stromal cells issuing from endometriomas [113]. Reduced decidualization and IGFBP-1 secretion have also been observed in primary endometriotic cells from other localizations [121] and in eutopic endometrial stromal cells from women with endometriosis [122,123]. Decidualization of the endometrium is a remodeling event that is physiologically occurring in response to progesterone in the secretory phase of the menstrual cycle in order to prepare the endometrium for the implantation of the embryo [124]. It is known that progestins and cAMP decrease phosphorylated AKT levels and increase nuclear FOXO1 levels in eutopic endometrial stromal cells [125,126]. The response to decidual stimuli by medroxyprogesterone acetate and dibutyryl cAMP was dramatically lower in ectopic endometriotic stromal cells. Interestingly, both inhibition of PI3K and AKT led to increasing nuclear FOXO1 and IGFBP1 levels in response to treatment with medroxyprogesterone acetate and dibutyryl cAMP, supporting evidence that the increased PI3K/AKT pathway is involved in the reduced decidual response in endometriosis [113]. This observation is further interesting, since it may indicate that the PI3K/AKT pathway is involved in processes supporting the effects of progesterone resistance, a well-described characteristic of endometriosis [127]. Therefore, small molecule inhibitors may be preclinically investigated as a therapeutic option, especially in overcoming progesterone resistance in endometriosis [113].

PI3K/AKT Pathway Alterations in OCCC and EnOC
In contrast to high-grade serous ovarian carcinomas, where activation of the PI3K/AKT pathway through mutation of PIK3CA, AKT or inactivating mutations of PTEN is rather rare (< 5%), it is a clearly more frequent event in OCCC and EnOC [128]. Activating mutations in PIK3CA encoding p110α, the catalytic subunit of PI3K, have been described to occur in 33%-40% of OCCC [2,129]. Activation of the PI3K/AKT pathway by loss of PTEN expression has been found in 40% of OCCC [130]. Finally, AKT2 amplification was observed in 14% of OCCC [131]. It is still unclear if aberrations in the PI3K/AKT pathway are critical drivers of cancer growth and, therefore, a possible therapeutic target in ovarian cancers [128]. Nevertheless, preclinical and clinical phase-I studies have suggested that inhibition of this pathway may help to overcome resistance to chemotherapy in ovarian cancer [132], which is a common problem in OCCC and, therefore, would be of specially great interest in this tumor type [52].

PIK3CA Mutations in Endometriosis-Associated Ovarian Cancer and Endometriosis
PIK3CA mutations, as a common mechanism of PI3K/AKT pathway activation in OCCC [133], will be discussed here in detail, since they are frequently associated with ARID1A mutations (cf. Section 4.2.).
Activating mutations in the PIK3CA gene, encoding the p110α subunit of PI3K, have been observed at a frequency of 33%-40% in OCCC [2,129]. Kuo et al. [129] first reported the high frequency of activating PIK3CA mutations in a large cohort of 97 OCCC, including 18 affinity-purified tumor cells from fresh specimen, 69 samples of microdissected paraffin-embedded tumors and 10 OCCC cell lines. They described an overall frequency of 33% PIK3CA mutations and of 46% in the 28 affinity-purified OCCCs and OCCC cell lines. The majority of the PIK3CA mutations were confined to exons 9 and 20, leading to an activation of p110α kinase. This was confirmed by immunohistochemistry, demonstrating an intense diffuse phosphorylated AKT immunoreactivity in all of the 18 specimens with PIK3CA mutations. This was also observed in PIK3CA wild-type tumors in 34 (85%) of 40 cases, indicating that other mechanisms are contributing to AKT phosphorylation in a large proportion of OCCCs with wild-type PIK3CA [129].
Yamamoto et al. [134] sequenced exon 9 and 20 of PIK3CA in 23 OCCC samples and found 10 tumors (43%) with activating mutations (H1047R in all cases). Interestingly, they observed the same mutations not only in nine adjacent atypical endometrioses of the 10 cases (90%), but they found non-atypical endometriotic tissue in six of the 10 samples (60%) with the same H1047R mutation, suggesting that these mutations occur very early in the tumorigenesis of OCCC [134].
An overview of studies that investigated PIK3CA mutations in OCCC and EnOC, as well as in endometriosis, is given in Tables 3 and 4, respectively.    No PIK3CA mutations detected in this collective [141]

Targeting the PI3K/AKT-Pathway in OCCC and EnOC
PI3K/AKT/mTOR signaling is frequently altered in EnOC and OCCC, but it is unclear whether these genetic changes are critical drivers for these carcinomas. It is therefore questionable whether these molecules are susceptible to targeted inhibition and suitable for response prediction. Several PI3K/AKT/mTOR inhibitors were explored as single agents or in combination in clinical trials of different tumor types. In ovarian and endometrial cancers, significant single-agent activity with PI3K/AKT inhibitors is rarely observed [142,143]. However, it is known that PI3K/AKT activation contributes to a reduced response to chemotherapy in ovarian cancer [144] and that modulation of the PI3K/AKT/mTOR pathway is suitable for overcoming resistance to chemotherapy [132]. Thus, Mabuchi et al. demonstrated increased sensitivity of cisplatin-resistant OCCC cell lines to the mTOR inhibitor everolimus, compared to the cisplatin-sensitive parental OCCC cell lines [145].
To date, reliable biomarkers for the prediction of response to therapy in ovarian cancer are not in clinical use [146]. In the past, several studies have explored the effects of PI3K/AKT/mTOR inhibitors on human cancer cells. In a study of several OCCC cells, PIK3CA mutations did not predict the sensitivity to PI3K/AKT/mTOR inhibitors [139]. In contrast, activated AKT was a predictive marker of drug sensitivity in ovarian cancer cells, which were treated with RAD001 (everolimus) to inhibit the mTOR pathway [147].
In a recent study, seven patients with advanced ovarian carcinomas harboring a PI3KCA mutation were enrolled onto clinical trials that included a PI3K/AKT/mTOR inhibitor. Prior to that, all patients had experienced treatment failure with standard therapies. Two patients (2/7, 29%) responded to this targeted therapy, which was combined with a cytotoxic drug. Other gynecologic and breast carcinomas in this study cohort also demonstrated a higher response rate than patients without PIK3CA mutations. Despite the small sample size, the authors conclude that PIK3CA mutation screening is helpful in the use of PI3K/AKT/mTOR pathway inhibition. However, single-agent use seems to not be sufficient to induce a response, because PIK3CA mutations often coexist with other concurrent molecular aberrations. Interestingly, both responders of the ovarian cancer subset had a simultaneous MAPK pathway mutation (one each in KRAS and BRAF) [142].
It is known that PI3K/AKT/mTOR signaling is a complex process, which interacts with the RAS/RAF/MEK/ERK pathway. In this context, it was observed that PIK3CA mutations may predict the response to PI3K/AKT/mTOR inhibitors, whereas concomitant mutations in the MAPK pathway (KRAS, NRAS, BRAF) may mediate resistance [148,149]. Therefore, clinical trials investigate several strategies, like dual targeting of PI3K/AKT/mTOR and RAF/MEK/ERK pathways, as well as the combination of multi-drugs instead of single-agents [128,150].

Functional Studies about the Loss of ARID1A Expression In Vitro and In Vivo
Cancer genome sequencing studies have created a new perspective about disordered chromatin organization as a feature of cancer, showing that mutations of epigenetic regulators are occurring frequently in a wide variety of different human cancers. Although the subject of intensive study, the exact function and role of ARID1A as a tumor suppressor remains far from being elucidated. In vitro studies have suggested different roles for ARID1A to exert its tumor suppressive action, which are mainly through proliferation, differentiation and apoptosis [77]. Knockdown of ARID1A led to increased proliferation of normal ovarian surface epithelial cells [84] and disrupted differentiation of certain cell types (e.g., osteoblasts) [151]. In Jurkat leukemia cells, Fas-mediated cell death was inhibited after ARID1A knockdown [152]. ARID1A, in contrast to the mutually exclusive ARID1B subunit of the SWI/SNF complex, inhibited cell cycle arrest in murine preosteoblasts [153,154]. Although these findings all constitute promising new perspectives in understanding the function of ARID1A as a tumor suppressor, the functional consequences of ARID1A mutations are probably more vast, since ARID1A regulates hundreds of different genes through the SWI/SNF chromatin remodeling complex [80]. Furthermore, it is likely that ARID1A mutations have divergent effects, depending on different cell and tumor types in which they are present, probably also depending on the mutational landscape in different cancer types. As a result, functional studies of ARID1A present a substantial scientific challenge [77].

Evidence for Cooperative Mechanisms between ARID1A and the PI3K/AKT Pathway
Various studies are suggesting cooperating mechanisms in relation to ARID1A mutations. Yamamoto et al. described a frequent co-occurrence between activating PIK3CA mutations and loss of ARID1A expression in OCCC, demonstrating that 46% of ARID1A deficient tumors were harboring PIK3CA mutations versus 17% of the ARID1A expressing tumors [102]. In endometrioid endometrial cancer, a higher frequency of PI3K/AKT-pathway alterations (PTEN loss or PIK3CA activating mutations) was found in tumors with a loss of ARID1A expression, and the number of tumors that showed no alteration in the PI3K/AKT-pathway was 4.6-fold higher in tumors with preserved ARID1A expression (p = 0.042) [107]. An association between PIK3CA and ARID1A mutations has also been reported in gastric cancer [92].
It has been suggested in many studies that loss of ARID1A is usually associated with TP53 wild-type tumors [84,91,92,155], and one study showed evidence for a direct protein-protein interaction between ARID1A and p53 [84].
A very interesting study conducted by Liang et al. [156] not only identified ARID1A as a potential driver gene in endometrial cancer, but also demonstrated that ARID1A mutations frequently co-occur with mutations, leading to activation in the PI3K/AKT pathway. An important observation was that siRNA knockdown of ARID1A in endometrial cancer cell lines per se led to an increased phosphorylation of AKT, indicating a regulation of the PI3K/AKT pathway activity by ARID1A [156].
A recent study in SMARCB1, a potent tumor suppressor with loss of expression, especially in certain rhabdoid tumor types and which is a core subunit of the SWI/SNF complex, showed persistent activation of AKT in SMARCB1-deficient tumor cells, contributing to survival and proliferation. Inhibition of AKT was sufficient to inhibit development of SMARCB1-deficient xenograft-tumors and to inhibit proliferation of SMARCB1-deficient cells in vitro [157].
Remarkable results have been presented at the 104th Annual Meeting of the American Association for Cancer Research by Guan et al. [158]: The authors tested the possibility of molecular dependency of ARID1A and the PI3K/PTEN pathway in an ARID1A knockout mice and ARID1A/PTEN double knockout mice model. After conditional depletion of either ARID1A (n = 10) or simultaneously ARID1A and PTEN (n = 12), the mice with knock-out of only ARID1A did not develop histological alterations, whereas 40% of the mice with double knock-out of ARID1A and PTEN developed poorly differentiated ovarian tumors disseminating in the peritoneal cavity and ascites. In the other 60%, the authors found hyperplasia of the ovarian surface epithelium. This shows that ARID1A inactivation in itself is not sufficient to initiate tumor development and requires a second hit, possibly consistent with an alteration in the PI3K/PTEN/AKT pathway to lead to carcinogenic transformation [158].
These observations are exciting for at least two important reasons. Firstly, it may explain the observation of ARID1A expression in non-atypical endometriosis made by several groups [101,104,106]; loss of ARID1A expression may be an early molecular event in these cases that might increase the overall risk of developing endometriosis-associated cancer, but in itself is not sufficient to initiate cancerogenesis. Secondly, it confirms that there is a close cooperative mechanism between ARID1A mutations and PI3K/AKT pathway alterations that might be of importance for early tumor detection, as well as in a therapeutic context. Interdependency on PI3K/AKT activation of ARID1A mutated tumor clones might be a process that is targetable by small-molecule inhibitors of the PI3K/AKT/mTOR pathway.

Possible Clinical Implications of ARID1A Mutations
The question of clinical implication of ARID1A mutations is not yet thoroughly answered. Whilst there is fast growing knowledge about the distribution of mutations of ARID1A and other members of the SWI/SNF complex in various cancers, functional and clinic-pathological data remain quite sparse.
Many studies have investigated ARID1A mutations as prognostic markers in a multitude of different cancers, such as OCCC, as well as breast, gastric and bladder cancer [91,93,97,100,102,155,[159][160][161], but to date, none have demonstrated a consistent prognostic significance of ARID1A mutational or expressional status. This may be partially due to the lack of larger prospective studies [77]. On the other hand, it has also to be answered if an early loss of ARID1A expression and/or PIK3CA mutation represents an increased probability for developing OCCC or EnOC in endometriosis [3,101,104,106]. Last, but not least, it will be of major interest to determine whether ARID1A inactivation may be therapeutically exploited by targeting downstream and potentially reversible epigenetic gene expression targets altered by remodeler mutations, such as, e.g., oncogenic proteins cyclin D1 and MYC or alterations in the Hedgehog pathway signaling [77,133,162].
An interesting synthetic-lethality therapy principle involving the SWI/SNF catalytic subunit, BRG1/SMARCA4, frequently deficient in non-small-cell lung carcinomas by BRM-ATPase inhibitors targeting another SWI/SNF subunit, has been proposed in a recent study [163]. This demonstrates that although the effects of SWI/SNF subunit mutations are very complex and still poorly understood, their respective downstream effectors are potentially targetable by compounds, opening a wide horizon of mechanisms that may be influenced for therapeutic purposes.

Conclusions
Although still at the beginning, the current genomic characterizations of ARID1A mutations and functional investigations of the loss of ARID1A expression in vitro and in animal studies, combined with analyses of presumed cooperating pathways as, e.g., the PI3K/AKT/mTOR pathway through direct activation and mutations of PTEN or PIK3CA, opens new perspectives for potential therapeutic approaches. This is of great interest in OCCC, since this frequently endometriosis-associated ovarian carcinoma subtype is characterized by resistance to conventional chemotherapy regimens and, therefore, has a poor prognosis in advanced stages. Mutations and consecutive loss of ARID1A expression, as well as activating mutations in the PI3K/AKT pathway (through loss of PTEN expression or PIK3CA activation) are highly frequent in OCCC and have to be further assessed in potential therapeutic strategies. The current intensified research activity in this field promises improved understanding and relevant progress in clinically significant aspects in the near future.