Recurrent WWTR1 S89W mutations and Hippo pathway deregulation in clear cell carcinomas of the cervix

Clear cell carcinoma (CCC) of the cervix (cCCC) is a rare and aggressive type of human papillomavirus (HPV)‐negative cervical cancer with limited effective treatment options for recurrent or metastatic disease. Historically, CCCs of the lower genital tract were associated with in utero diethylstilbestrol exposure; however, the genetic landscape of sporadic cCCCs remains unknown. Here we sought to define the molecular underpinning of cCCCs. Using a combination of whole‐exome, targeted capture, and RNA‐sequencing, we identified pathogenic genetic alterations in the Hippo signaling pathway in 50% (10/20) of cCCCs, including recurrent WWTR1 S89W somatic mutations in 40% (4/10) of the cases harboring mutations in the Hippo pathway. Irrespective of the presence or absence of Hippo pathway genetic alterations, however, all primary cCCCs analyzed in this study (n = 20) harbored features of Hippo pathway deregulation at the transcriptomic and protein levels. In vitro functional analysis revealed that expression of the WWTR1 S89W mutation leads to reduced binding of TAZ to 14‐3‐3, promoting constitutive nuclear translocation of TAZ and Hippo pathway repression. WWTR1 S89W expression was found to lead to acquisition of oncogenic behavior, including increased proliferation, migration, and colony formation in vitro as well as increased tumorigenesis in vivo, which could be reversed by targeted inhibition of the TAZ/YAP1 complex with verteporfin. Finally, xenografts expressing WWTR1 S89W displayed a shift in tumor phenotype, becoming more infiltrative as well as less differentiated, and were found to be composed of cells with conspicuous cytoplasmic clearing as compared to controls. Our results demonstrate that Hippo pathway alterations are likely drivers of cCCCs and likely contribute to the clear cell phenotype. Therapies targeting this pathway may constitute a new class of treatment for these rare, aggressive tumors. © 2022 The Pathological Society of Great Britain and Ireland.


Introduction
Clear cell carcinoma (CCC) of the cervix (cCCC) is a rare subtype of endocervical adenocarcinoma, accounting for 3% of all cervical adenocarcinomas [1]. In contrast to the most prevalent type of cervical cancer, squamous cell carcinoma (SCC), cCCC is not related to human papillomavirus (HPV) infection [1,2]. Historically, CCCs of the lower genital tract were associated with in utero exposure to diethylstilbestrol (DES) [3,4]; however, today most CCCs are sporadic without a known etiology [2]. Although early-stage cCCCs have a good overall prognosis following radical surgical resection, cCCCs have a greater propensity to recur and metastasize, and advanced-stage disease is often incurable and associated with high mortality [5,6]. Current treatment options for recurrent or metastatic disease are limited, and novel therapeutic approaches for cCCCs are needed.
The etiology and genomic features of non-HPV-related cervical cancers, including cCCCs, remain active areas of investigation. Boyd et al [7] performed a molecular analysis of 24 vaginal and cervical CCCs and found microsatellite instability in 50% and 100% of non-DES and DES-exposed cases, respectively. As part of The Cancer Genome Atlas (TCGA) project, common-type HPV-associated SCCs and adenocarcinomas were found to display an enrichment in APOBEC signatures, recurrent PIK3CA mutations (26%), 3q copy number gains (66%), and amplification of the Hippo signaling pathway component YAP1 (16%) [8]. Based on our previous genomic studies of rare epithelial malignancies that have resulted in the identification of novel genotypicphenotypic correlations and convergent phenotypes [9][10][11][12][13], we posited that if cCCCs were driven by a highly recurrent if not pathognomonic genetic alteration these could potentially serve as therapeutic targets. Here we sought to define the molecular underpinning of cCCCs using a combination of targeted, whole-exome, and RNA-sequencing. Our analyses revealed recurrent alterations affecting the Hippo pathway in cCCCs, including WWTR1 S89W mutations. Functional in vitro and in vivo studies further demonstrate that WWTR1 S89W mutations are not only oncogenic and associated with phenotypic changes in these tumors but may also constitute the basis for new targeted therapeutics for patients with this disease.

Samples
Following approval by the Institutional Review Boards (IRBs) of the authors' institutions, unstained tissue sections from formalin-fixed paraffin-embedded (FFPE) cCCCs were retrieved from Weill Cornell Medical Center (New York, NY, USA) and Memorial Sloan Kettering Cancer Center (MSK; New York, NY, USA). Patient consents were obtained following the respective IRB protocols approved by the authors' institutions, and samples were anonymized before analysis. All tumors were centrally reviewed by two pathologists (KJP and CM), and a total of 20 tumors were classified as cCCCs and included in this study. The histopathologic criteria are detailed in the Supplementary materials and methods.

Western blotting, immunohistochemistry, and immunofluorescence
Protein extraction, standard western blotting, immunohistochemical, and immunofluorescence analyses were performed as described previously [11,29,30], see also Supplementary materials and methods.
Proliferation, scratch wound healing, and colony formation assay Cells were seeded, then proliferation, scratch wound healing, and colony formation assays were performed as previously described [10][11][12] (see also Supplementary materials and methods).

Xenograft studies
NOD Scid Gamma (NSG) mice (Jackson Laboratory, Bar Harbor, ME, USA) were used for in vivo studies and were cared for in accordance with guidelines approved by MSK's Institutional Animal Care and Use Committee and Research Animal Resource Center. Six 8-week-old female mice were injected subcutaneously with 7.5 million HaCaT cells with Matrigel (Corning, Oneonta, NY, USA) at a 1:1 ratio. Once tumors reached an average volume of 100 mm 3 , mice were randomized to receive either verteporfin 100 mg/kg or vehicle control intraperitoneally every other day for 1 month. Mice were observed daily throughout the treatment period for signs of morbidity/mortality. Tumors were measured twice weekly using calipers, and volume was calculated (length Â width 2 Â 0.52). Body weight was assessed twice weekly. After 4 weeks of treatment, tumor samples were collected for histological and immunohistochemical analysis. All tumors were reviewed by two pathologists (JSR-F and KJP).

Statistical analyses
Statistical analysis was performed using Prism 7 (GraphPad Software, San Diego, CA, USA). Student's two-tailed t-test was employed for comparison of means in parametric data. Heteroscedasticity was assessed for each comparison, and homoscedastic or heteroscedastic t-tests were employed as appropriate, as described previously [10,11]. A p value<0.05 was considered significant.

Clinical and histologic features of cCCCs
Following central pathology review, 20 cCCCs were included in this study (supplementary material, Table S1). Histologically, these tumors displayed the characteristic histologic features of cCCCs, consisting of tubulocystic, papillary, and solid structures composed of cells with clear or eosinophilic cytoplasm and 'hobnail' cells with moderate-to-marked atypia and infrequent mitotic figures ( Figure 1A). Patient median age at diagnosis was 50.5 years (range, . At the time of diagnosis, all stages, following the 2018 FIGO staging system for cervical cancer [32], were Genomics of cervical clear cell carcinomas 637   Table S1).
cCCCs harbor recurrent genetic alterations in the Hippo pathway To define their genomic landscape, 20 tumor-normal cCCC pairs were subjected to WES (n = 5) or targeted MSK-IMPACT sequencing due to limited quantity/ quality of available DNA (n = 15). cCCCs were found to have a relatively low mutation burden. with a median of 0.94 somatic mutations/Mb (range, 0.43-1.47) on WES, and of 2.63 somatic mutations/Mb (range, 0-9.5) by MSK-IMPACT (supplementary material, Table S2). In addition to ARID1A mutations (4/20, 20%), a known genomic feature of other Mullerian CCCs [33], this analysis revealed the presence of recurrent TERT (3/20, 15%), NFE2L2 (3/20, 15%), and WWTR1 mutations (4/20, 20%; Figure 1B). The WWTR1 mutations affected the S89W residue, which maps to the TEAD binding domain of TAZ, an effector of the Hippo signaling cascade ( Figure 1C). A pathway analysis based on the presence of pathogenic somatic mutations, amplifications, and homozygous deletions further revealed a significant enrichment for alterations in curated Hippo pathway-related genes in these tumors (p < 0.01, EASE Score, Fisher's exact test; Figure 1D) [31]. Of the 20 cCCCs analyzed, 10 (50%) were found to harbor pathogenic alterations affecting components of the Hippo signaling pathway, including WWTR1 (5/10, 50%), YAP1 (2/10, 20%), TEAD3 (1/10, 10%), TEAD4 (1/10, 10%), and LATS2 (1/10, 10%), which occurred in a mutually exclusive pattern. Specifically, we found four hotspot mutations and one amplification in WWTR1, one likely a pathogenic S61W missense mutation [34,35] and one deletion in YAP1, one homozygous deletion in TEAD3, one amplification in TEAD4, and one likely a pathogenic K784N missense mutation in LATS2 ( Figure 1B and supplementary material, Table S2). Both the WWTR1 S89W and the YAP1 S61W mutations were found to be clonal (supplementary material, Figure S1) and to affect residues in the TEAD binding domains of their respective proteins, TAZ and YAP1, which contain the 14-3-3 binding motifs [36] ( Figure 1C). We subjected five cCCCs lacking mutations, amplifications, or homozygous deletions affecting Hippo pathway-related genes to RNAsequencing, but no fusion transcripts affecting Hippo pathway genes were identified (supplementary material, Table S3). Importantly, irrespective of the presence or absence of genetic alterations in the Hippo pathway, RT-qPCR analysis provided transcriptomic evidence of deregulation of the Hippo pathway in all cCCCs analyzed (supplementary material, Figure S2), suggesting the presence of epigenetic or genetic alterations not detected by the targeted sequencing approaches employed. Furthermore, using RNA-sequencing, cCCCs displayed a significant increase in the expression of WWTR1 and TAZ/YAP transcriptional targets (15/26; p < 0.05; supplementary material, Figure S2) [36], providing further evidence of alteration of the Hippo pathway in cCCCs.

WWTR1 S89W expression results in the acquisition of oncogenic properties
There is burgeoning evidence to demonstrate that the Hippo pathway plays important roles in the pathogenesis of different types of cancer [37], with recurrent amplification of YAP1 and WWTR1 being reported in SCCs of the head and neck, esophagus, lung, cervix [31], and ovarian cancers [38]. Somatic hotspot mutations have not been previously described as a recurrent mechanism of the Hippo pathway alteration in cancer. Based on a study that experimentally assessed the effect of 19 WWTR1 nonsynonymous mutations in MCF10A nonmalignant breast epithelial cells, which identified the S89W mutation as functionally activating [36], we hypothesized that the expression of WWTR1 S89W would be associated with oncogenic properties. To test this, we performed functional in vitro studies by expressing WWTR1 wildtype (WT), WWTR1 S89W mutation and empty vector as control (supplementary material, Figure S3). Given the scarcity of commercially available HPV-negative nonmalignant cervical cell lines, we selected primary nontumorigenic differentiation-competent HaCaT keratinocytes as a model system, which has been employed by others for the study of cervical tumorigenesis [39][40][41]. As a second model, and to confirm the results obtained in HaCaT cells, we employed HEK-293 cells, which are widely used in cell biology/ cancer research studies given their reliable growth and propensity for transfection [42]. In both HaCaT and HEK-293 cells, WWTR1 S89W expression resulted in a significant increase in proliferation and colony formation as compared to cells expressing empty vector control (control) or WWTR1 WT (Figure 2A,B). A scratch assay analysis revealed increased migration of HaCaT and HEK-293 cells upon stable expression of WWTR1 S89W when compared to cells expressing WWTR1 WT or control ( Figure 2C). As expected, based on observations of WWTR1 amplification/overexpression [37], WWTR1 WT-expressing cells showed increased proliferation, colony formation, and migration as compared to controls; however, cells harboring WWTR1 S89W consistently exhibited a significantly stronger oncogenic phenotype (Figure 2A-C).
Expression of WWTR1 S89W results in constitutive deregulation of the Hippo signaling pathway TAZ has been reported to play pivotal roles in regulation of the Hippo signaling pathway [43,44]. Upon nuclear translocation, along with YAP, TAZ acts as a cotranscriptional activator of the pathway, thereby promoting tissue growth and cell viability [37]. We investigated whether WWTR1 S89W expression would result in alteration of this pathway in our cell models. We found that both HaCaT and HEK-293 cells expressing WWTR1 S89W displayed Genomics of cervical clear cell carcinomas 639 increased expression levels of key Hippo pathway target genes, including AXL, ITGB2, LMNB2, and CYR61 [44,45], when compared to the respective control and WWTR1 WT cells ( Figure 3A). This finding was further supported by quantitative western blot analyses, which revealed increased levels of AXL, ITGB2, LMNB2, and CYR61 proteins upon WWTR1 S89W expression in both cell models ( Figure 3B). Again, the expression of WWTR1 WT was also associated with increased AXL, ITGB2, LMNB2, and CYR61 gene expression; however, at significantly lower levels than in those expressing the WWTR1 S89W mutation ( Figure 3A). Our in vitro findings suggest that the oncogenic behavior of cCCCs harboring the WWTR1 S89W mutation can, at least in part, be explained by the deregulation of the Hippo pathway, corroborating the observation that this tumor suppressor pathway is consistently repressed in human cCCCs (supplementary material, Figure S2).

WWTR1 S89W mutation stabilizes TAZ protein deregulating the Hippo pathway
We investigated the molecular mechanisms underlying the deregulation of the Hippo pathway in cells expressing the WWTR1 S89W mutation. TAZ activity is regulated through phosphorylation by LATS2, promoting proteolytic degradation or retention in the cytoplasm due to sequestering by the 14-3-3 protein [46,47]. To define whether TAZ stability is affected by the presence of the S89W mutation, we treated HaCaT and HEK-293 cells with the protein synthesis inhibitor cycloheximide (CHX) and observed that in cells expressing WWTR1 S89W, TAZ had a longer half-life than in those cells expressing WWTR1 WT in both HaCaT and HEK-293 cell lines ( Figure 3C). Additionally, we hypothesized that WWTR1 S89W cells would exhibit constitutive Hippo pathway repression, and thereby be functionally activating, with expression of Hippo pathway targets All experiments were performed in technical triplicate in at least three independent replicates, and statistical analyses were performed comparing each condition to Control. Mean AE SD. n.s., not significant; *p < 0.05, **p < 0.01, ***p < 0.001; two-tailed unpaired t-test.  Figure 3D). Moreover, immunoprecipitation Figure 3 Legend on next page.

Genomics of cervical clear cell carcinomas 641
of TAZ in HEK-293 cells indicated that in WWTR1 S89W cells, the TAZ protein did not bind efficiently to 14-3-3, demonstrating that the S89W mutation reduces 14-3-3 cytoplasmic sequestration of TAZ, leading to TAZ nuclear translocation [46] ( Figure 3E). Finally, we performed a luciferase reporter assay with a specific Hippo/TEAD reporter containing eight TEAD binding sites [48] in HEK-293 cells stably expressing control, WWTR1 WT, and WWTR1 S89W. Increased activation of the Hippo/TEAD reporter was observed in cells expressing WWTR1 WT and WWTR1 S89W; however, Hippo/TEAD activation was significantly enhanced in WWTR1 S89W-expressing cells ( Figure 3F). Taken together, these findings demonstrate that the WWTR1 S89W mutation increases TAZ stability and promotes nuclear localization, thereby resulting in downstream constitutive repression of the Hippo signaling pathway.

Oncogenic and phenotypic properties can be reversed by inhibition of the TAZ/YAP/TEAD transcriptional complex
Verteporfin (Vysudine) targets the Hippo pathway by repressing the interaction between TAZ, YAP1, and TEAD, and is FDA-approved for photodynamic treatment of vascular disorders of the eye [49]. Currently, verteporfin is being studied in clinical trials for the treatment of different tumor types, including cutaneous metastasis of breast and pancreatic cancer [50]. Given the deregulation of the Hippo pathway upon WWTR1 S89W expression in our in vitro models, we assessed whether inhibition of the TAZ/YAP1/TEAD transcriptional complex would result in reversal of the observed oncogenic phenotype induced by WWTR1 S89W expression. Upon treatment of HaCaT and HEK-293 cells with verteporfin, we observed significant inhibition of proliferation and colony formation in cells expressing WWTR1 S89W ( Figure 4A,B and supplementary material, Figure S4). Verteporfin treatment also led to a significant decrease in migration rates of HaCaT and HEK-293 cells expressing WWTR1 S89W, which was not observed in controls or cells expressing WWTR1 WT ( Figure 4C).

Impact of the WWTR1 S89W mutation on tumorigenesis
We next sought to define the impact of WWTR1 S89W on tumorigenicity. HaCat cells expressing control, WWTR1 WT, and WWTR1 S89W were injected subcutaneously into NSG mice. Consistent with our observations made in vitro, HaCaT cells expressing WWTR1 S89W grew faster and formed significantly larger tumors than cells expressing WWTR1 WT or control ( Figure 5A). In addition, treatment of mice with verteporfin for 33 days inhibited the acquired oncogenic potential of WWTR1 S89W expressing cells to control levels ( Figure 5A), indicating that the Hippo pathway deregulation is responsible for the increased tumor growth. Consistent with the in vitro observations and the notion that WWTR1 WT or WWTR1 S89W-expressing xenografts are dependent on the Hippo pathway, verteporfin treatment resulted in necrosis only in the xenografts expressing WWTR1 WT or WWTR1 S89W (supplementary material, Figure S5). Importantly, mice treated with verteporfin did not show detectable clinical side effects or significant loss of body weight (supplementary material, Figure S5), indicating that verteporfin treatment is likely safe [51,52]. Histologic review of the xenografts revealed that, while control HaCaT cells resulted in pure SCCs with squamous pearls and overt keratinization ( Figure 5B), xenografts expressing WWTR1 WT and, in particular, WWTR1 S89W, displayed a phenotype shift, with tumors becoming more infiltrative and less differentiated, being composed of infiltrating ragged gland-like structures, nests, and cords of tumor cells, with marked atypia. A systematic histologic analysis of the xenografts revealed that the degree of differentiation decreased in a stepwise manner according to the expression of empty vector (EV), WWRT1 WT to WWTR1 S89W, whereas immunohistochemical analysis of the Ki67 labeling indices demonstrated that the proliferation levels increased in a stepwise manner from EV, WWRT1 WT to WWTR1 S89W (Ki67 labeling index 33, 60, and 89, respectively; Figure 5B). Consistent with these observations, the expression and distribution of p40, a protein expressed in cells of the squamous lineage apart . Quantification (bottom) of TAZ intensity/ cell in a nucleus/cytoplasm ratio. (E) Immunoprecipitation assay with TAZ antibody of HEK-293 cells stably expressing Control, WWTR1 WT, or WWTR1 S89W. Western blot was performed using anti-TAZ and anti-pan 14-3-3 antibodies, and tubulin as loading control. (F) Hippo pathway luciferase reporter assay of HEK-293 cells stably expressing Control, WWTR1 WT, and WWTR1 S89W. SV40-Renilla was used to normalize transfection efficiency. All experiments were performed in technical triplicate in at least three independent replicates and statistical analysis was performed comparing each condition to Control (A,B,F), or to WWTR1 WT (C,D). Mean AE SD; n.s., not significant; *p < 0.05, **p < 0.01, ***p < 0.001; two-tailed unpaired t-test.

642
SH Kim, T Basili, H Dopeso, A Da Cruz Paula et al from terminally differentiated squamous cells, was higher in WWTR1 WT and WWTR1 S89W xenografts than in models expressing EV (p40-positive cells: EV 58%, WWTR1 WT 77%, WWTR1 S89W 92%). These findings are further supported by immunohistochemical analysis of vimentin expression (which is Figure 4 Legend on next page.

Genomics of cervical clear cell carcinomas 643
detected in poorly differentiated SCCs and in clear cell carcinomas (Figure 6), but either absent or only focally present in well-differentiated SCCs), with an antibody that reacts with the human but not mouse vimentin, revealed a higher level of vimentin expression in xenografts expressing WT WWTR1 and WWTR1 S89W ( Figure 5B). Immunohistochemical analysis further revealed that, while TAZ expression was either absent or primarily cytoplasmic in the control and WWTR1 WT-expressing xenografts, WWTR1 S89W tumors showed increased nuclear TAZ expression ( Figure 5B). Given that we observed WWTR1 S89W expression lead to nuclear TAZ expression in both the in vitro and in vivo settings, we performed TAZ immunohistochemical analysis of primary cCCCs (n = 13). Consistent with our in vitro and in vivo findings, strong nuclear TAZ expression was present in the human primary cCCCs (Figure 6).
Taken together, our results not only implicate the Hippo pathway in the pathogenesis of human cCCCs and tumor phenotype changes, but also support the contention that pharmacologic inhibition of the Hippo pathway may constitute a potential novel targeted therapeutic approach for cCCCs.

Discussion
cCCCs, while curable in the early stage, are associated with poor prognosis in the advanced setting [6]. Treatment options for metastatic or recurrent cCCC remain limited in part due to a lack of understanding of its genetic basis and pathogenesis. Here we demonstrate that cCCCs harbor recurrent alterations in the Hippo signaling pathway, with WWTR1 being the most commonly affected gene. At variance with other cancer types, including renal clear cell carcinomas, in which WWTR1 amplification has been documented [36], in cCCCs we detected the presence of a recurrent WWTR1 S89W somatic mutation. In cases lacking this mutation, other somatic genetic alterations affecting the Hippo pathway, including those affecting YAP1, TEAD3/4, and LATS2, and/or activation of downstream targets were identified, consistent with the notion that CCCs may constitute a convergent phenotype, which is dependent on deregulation of this signaling pathway.
The Hippo pathway, critically important in the regulation of organ size and tissue homeostasis, is recurrently altered in certain cancers including mesothelioma, glioma, SCCs of the head and neck, and cervix [31,36]. Upregulation and/or overexpression of YAP1 and TAZ, the transcriptional co-activators of the pathway, contribute to tumor progression in multiple cancer types, including breast, lung, liver, colon, and ovarian cancer [52,53]. The canonical Hippo pathway comprises a serine/threonine kinase cascade, in which MST1/2 interacts with Salvador homolog 1 (SAV1) and phosphorylates LATS1/2, thereby activating them. Activated LATS1/2 in turn phosphorylates conserved serine residues on YAP1/TAZ, leading to cytoplasmic retention of YAP1/ TAZ via 14-3-3 binding and degradation of YAP1/ TAZ [47]. Recent studies have highlighted the oncogenic effects of mutating the LATS2 phosphorylation site of TAZ, S89, resulting in constitutive activation of TAZ (TAZ-S89A), leading to increased cellular proliferation in multiple cell models [54][55][56]. Here we demonstrate that the S89W missense mutation in WWTR1 increases TAZ stability and nuclear localization, resulting in constitutive repression, and thereby functional activation, of the Hippo pathway. As expected, we observed increased proliferation, colony formation, and migration in our cell models in cells expressing WWTR1 S89W. Furthermore, we observed increased tumorigenesis in xenograft models injected with WWTR1 S89W cells, which could be reversed by verteporfin, a potent Hippo pathway transcriptional complex inhibitor, both in vitro and in vivo. Expression of WWTR1 S89W was not only found to be oncogenic but may also result in phenotypic changes, including the clearing of the cytoplasm and increased expression of markers of dedifferentiation, corroborating the notion that constitutive repression of the Hippo pathway plays a role in the pathogenesis of this aggressive form of HPV-negative cervical cancer.
In addition to the recurrent S89W mutation in WWTR1, we report on the S61W mutation in YAP1. The downstream effects of the S61W YAP1 mutation, akin to those of the WWTR1 S89W mutation described here, have been reported to include reduced phosphorylation, and subsequent protein stabilization and increased nuclear localization of YAP1, thereby resulting in an increase in TAZ/YAP1/TEAD transcriptional activity [34,35]. Interestingly, one cCCC studied here contained a YAP1 deletion, which would presumably have the opposite effect; however, there are reports that YAP inversely regulates TAZ protein abundance [57,58]. VP, Verteporfin. All experiments were performed in technical triplicate in at least three independent replicates and statistical analysis was performed comparing each condition to its corresponding untreated pair. Mean AE SD; n.s., not significant; *p < 0.05, **p < 0.01, ***p < 0.001; two-tailed unpaired t-test.

644
SH Kim, T Basili, H Dopeso, A Da Cruz Paula et al

Genomics of cervical clear cell carcinomas 645
While previous studies have reported cervical cancers to have alterations in Hippo pathway genes, specifically YAP1 and WWTR1 amplification, the majority of these findings were reported irrespective of histologic subtype or in the more common type of HPV-driven SCCs [8,59,60]. The AACR Project GENIE reported the WWTR1 S89W mutation to be present in 0.02% of all cancers, with the majority being cancers of gynecologic origin, including two cCCCs [61]. In the TCGA pancancer study [62], the WWTR1 S89W mutation was only detected in a single case (1/10967 samples, 0.009%), namely, a FIGO grade 1 endometrioid endometrial cancer. Given that expression of WWTR1 S89W in HaCaT cells resulted in conspicuous cytoplasmic clearing in vivo, we sought to define whether this mutation would constitute a feature of not only cCCCs but also of other Mullerian carcinomas with a clear-cell phenotype. Reanalysis of the MSK-IMPACT sequencing data (v6/7) of ovarian and endometrial cancers revealed the presence of a WWTR1 S89W mutation in 1/1790 (0.06%) ovarian cancer, which was of clear-cell histology, and WWTR1 S89W or S89L mutations in 5/2,093 (0.24%) endometrial carcinomas, two of which were CCCs (supplementary material, Figure S6). None of the remaining CCCs of the ovary (n = 141) or endometrium (n = 78) harbored WWTR1 S89 mutations. These data suggest that mutations affecting codon S89 of WWTR1 may be causally linked to a clear-cell phenotype not only in cCCCs, but also in a subset of noncervical Mullerian cancers. Of note, a subset of ovarian CCCs is known to express high levels of YAP at the protein level, correlating with poor prognosis [63,64]. Aside from a single case with a YAP1 missense mutation, we did not observe YAP1 amplification in the 141 ovarian CCCs analyzed here.
This study has several limitations. Given the scarcity of nonmalignant HPV-negative cervical cancer cell lines, the functional studies were performed in immortalized human keratinocytes (HaCaT) and in HEK-293 cells, whose exact cell type is still a matter of contention [42]. HaCaT WWTR1 S89W xenografts were found to display phenotypic features distinct from the control cells, including clear cytoplasm, supporting the notion that these cells may be an adequate disease model. Although a likely pathogenic genetic alteration affecting the Hippo pathway was detected in only 50% (10/20) of the cases, features of Hippo pathway deregulation, including Hippo downstream target and/or nuclear TAZ protein expression, were found in all cCCCs analyzed here, supporting the importance of this pathway in the pathogenesis of these aggressive cancers. Further studies are warranted to determine alternative mechanisms that lead to increased TAZ/YAP1 complex activity. Given that this study was based on WES, MSK-IMPACT, and RNA-sequencing, we cannot rule out genetic alterations affecting noncoding regions of Hippo pathway genes or translocations resulting in the deregulation of this pathway in cases not subjected to RNA-sequencing or epigenetic regulation of this pathway. Whole-genome sequencing and epigenetic profiling of cCCCs lacking the canonical Hippo pathway alterations reported here are warranted. Finally, given

646
SH Kim, T Basili, H Dopeso, A Da Cruz Paula et al the small number of patients in our cohort, and the multiinstitutional nature of the study, correlations with mutation status and patient outcomes could not be performed. Larger studies are required to elucidate the impact of Hippo pathway alterations on clinical outcomes of cCCC patients. Despite these limitations, here we report on the observation that Hippo pathway alterations, particularly the somatic WWTR1 S89W mutation, likely constitute drivers of cCCCs, and may also be causative of the clear-cell phenotype in a subset of cCCCs and, potentially, of Mullerian cancers. Our findings also offer a new potential therapeutic target for patients with these aggressive forms of HPV-negative cervical cancers, with FDA-approved inhibitors of the Hippo pathway available and currently under investigation in clinical trials for patients with other cancer types.

SUPPLEMENTARY MATERIAL ONLINE
Supplementary materials and methods  Table S1. Clinicopathologic characteristics of patients with clear cell carcinoma of the cervix included in this study Table S2. Non-synonymous somatic mutations identified in clear cell carcinomas of the cervix subjected to whole-exome or targeted massively parallel sequencing Table S3. Fusion candidate genes and/or readthroughs identified by RNA-sequencing analysis of clear cell carcinomas of the cervix Genomics of cervical clear cell carcinomas 649