Comprehensive Molecular, Pathological, and Clinical Characterization of Clear Cell Adenocarcinoma of the Urinary Tract.

We identified 35 patients with primary utCCA, of whom 30 (86%) were female. Only one patient had a remote history of clear cell carcinoma of the ovary, 12 years prior, and developed clear cell adenocarcinoma in the distal ureter that was histologically different from the ovarian tumor as it also contained a urothelial carcinoma component and was therefore considered as primary CCA at this site. Median age at diagnosis was 63 years (IQR 46-68). The most common presenting symptoms were urinary urgency, hesitancy, urge incontinence, dysuria, and hematuria. Patients details are available in Supplementary Table 3 . All patients had invasive disease (≥pT1) at diagnosis. Radical curative-intent surgery was performed on 25 patients. Lymph node metastasis was present in 9 of 15 patients (60%) whose surgery included pelvic lymph node dissection. No patients had clinical evidence of distant metastasis at diagnosis. Most patients were treated with multimodality therapy including surgery, intra- and/or post-operative radiation, and/or systemic therapy. OS and DFS data were available for 28 and 27 patients, respectively. Cancer recurrence occurred in 14 patients, and 14 patients died during a median follow-up of 50.5 months (range: 1-150 months). Median OS was 65 months (95%CI 53 - not reached), 5-year OS was 55% (95% CI, 37%-81%), and median DFS was 42 months (95% CI 23 - not reached) ( Supplementary Figure 1 ). Twenty-nine tumors arose in the urethra, four in the bladder and one each in the distal ureter and renal pelvis. In all five male patients, the tumors arose in the urethra (3 prostatic, 1 membranous/bulbar, and 1 penile urethra). Eighteen tumors were associated with a urethral diverticulum, all were in female patients (a representative such example is shown in Supplementary Figure 2 ). The spectrum of histologic features was variable, but all tumors included a combination of tubulocystic, tubular/glandular patterns and the majority contained areas of papillary or solid growth and myxoid or hyalinized stroma ( Figure 1 ).Uncommon histologic findings include yolk sac tumor-like morphology in two tumors and the presence of sarcomatoid features and squamous differentiation, in one tumor each ( Figure 1 ). As expected for these tumors, most cases (91%, 32/35) expressed PAX-8 ( Figure 1J ), and HNF-1β (22/22, 100%, including all three cases that were negative for PAX-8) ( Figure 1L ). GATA-3 expression was negative in 24/33 (73%) of tumors, whereas the remaining 9/33 (27%) expressed GATA-3 only focally or in rare tumor cells. Expression of the mismatch repair proteins MLH1, PMS2, MSH2 and MSH6 was retained (normal pattern) in all cases tested (n=27). The diverticular spaces adjacent to the invasive CCA were lined by columnar pseudostratified epithelium with variable inflammatory infiltrate in both the underlying stroma and within the epithelium. In some case, the epithelial lining was denuded and the underlying stroma was fibrotic with associated inflammation. Intestinal metaplasia was identified in the diverticular lining in only 3 cases, but endometriosis or benign Müllerian remnants were identified in any case. Importantly, in 15 tumors, the diverticular spaces show focal to prominent areas in which the epithelial lining was composed of cells with atypical nuclei characterized by nuclear pleomorphism and hyperchromasia with occasional hobnail morphology, and clear or eosinophilic cytoplasm, similar to those of the adjacent invasive tumor ( Figure 2 ). While these features may superficially resemble the morphology seen in nephrogenic adenoma, the association with invasive carcinoma, the atypical/dysplastic nuclei, lack of prominent basement membrane and lack of associated granulation tissue appearance argue against this consideration (see further discussion below). Nineteen tumors underwent whole exome sequencing (WES) and 21 underwent MSK-IMPACT targeted sequencing. This includes 18 tumors with both WES and MSK-IMPACT, 1 tumor with WES only and 3 tumors with MSK-IMPACT only) ( Figure 3A , Supplementary Table 3 ). The median tumor mutational burden (TMB) was 3.1 mutations/megabase (IQR 2.7–5.1). ARID1A and TP53 (29%) were the most frequently mutated genes, with pathogenic/likely pathogenic mutations identified in six tumors for each (29%) and co-occurrence of both mutations in one tumor. TERT promoter mutations were detected in two tumors (see below) while no FGFR3 mutations were identified. Notable recurrent copy number alterations included gains or amplifications of chromosome arm 8q in 12 tumors that were associated with or without whole genome duplication in 6 tumors each ( Figure 3B ). Focal amplifications were present in ERBB2 (n=5), KRAS (n=3), EGFR (n=2) and CD274 (which encodes PD-L1, n=1), and deletion of MGA , ATM, ARID1A , SMARCA4, KMT2A, CREBBP, SETD2, TP53 and CDKN2A ( Figure 3A ). Given the recent tumor agnostic accelerated FDA approval of the HER2-targeted antibody drug conjugate trastuzumab deruxtecan (TDx-T) for adult patients with unresectable or metastatic HER2-positive (IHC3+) solid tumors who have received prior systemic treatment and have no satisfactory alternative treatment options 46 , 47 , we performed HER2 IHC on 12 tumors. Two tumors harbored ERBB2 amplification and had IHC scores of 3+ and 1+ (one example is illustrated in Supplementary Figure 3 ). The remaining 10 tumors were wildtype for ERBB2 and had IHC score of 2+ (n-=2), 1+ (n=5) and 0 (n=3) ( Supplementary Table 3 ). Mutational signature analysis revealed that the predominant single base pair substitutions were those associated with aging (SBS1) in 13/18 tumors, and APOBEC (SBS2 and/or SBS13) in 8/18 tumors. Notably, SBS24, which has been associated with aflatoxin exposure, was identified in 11 tumors ( Figure 3C ), an unusual finding in tumors arising in the urinary tract 43 . TERT promoter mutations were observed in two tumors (C228T and C105T), one of which arose in the ureter with a co-existing component of nested subtype of urothelial carcinoma ( Supplementary Figure 4 ). Among ARID1A- mutated tumors, 3/6 had complete loss of ARID1A protein expression and one tumor had partial loss ( Figure 4A - 4D ). All tumors with wild-type ARID1A retained ARID1A expression by IHC. Interestingly, one case showed loss of ARID1A expression both in the invasive CCA and in an adjacent region of clear cell dysplasia ( Figure 2E - 2H ), whereas the adjacent normal appearing urethral lining retained ARID1A expression. IHC for p53 showed mutant/abnormal expression in 3 of 4 tumors with TP53 mutations ( Figure 4E - 4H ) that were tested. Mutations in SMARCA4 were identified in two tumors, one of which harbored a nonsense mutation (p.Q847*, c.2539C>T), and had BRG1 (encoded by SMARCA4 ) loss of expression by IHC in both the invasive CCA and adjacent diverticular clear cell dysplasia ( Supplementary Figure 5 ). There were no specific histopathologic features related to any of the identified genomic alterations. Germline analysis of over 76 genes with known hereditary cancer predisposition association, for which all coding regions were sequenced, was available for 9 patients, none of whom harbored any pathogenic germline mutations, including DNA mismatch repair genes MLH1, MSH2, MSH6 and PMS2 . To gain insight into the molecular pathogenesis of utCCA, we compared the somatic mutational profiles of 22 CCA of the urinary tract with those of 140 primary ovarian CCA, 72 primary uterine CCA and 477 primary invasive UC, not otherwise specified (NOS) tumors analyzed using the MSK-IMPACT assay as part of a pan-cancer, institutional scale, prospective tumor sequencing initiative ( Figure 5A ). ARID1A was the most commonly mutated gene in CCA arising in the urinary tract (32%) or ovary (64%), and was also mutated at high frequency in UC, NOS tumors (26%). Mutations in other chromatin modifying genes and cell cycle regulators, including KDM6A , KMT2D , CREBBP, CDKN2A , and RB1, were significantly more common in UC, NOS tumors compared to CCA arising at all three primary sites (all q<0.05). TERT promoter mutations were also significantly more common in UC, NOS compared to CCAs of the urinary and gynecologic tracts, whereas FGFR3 mutations were exclusively present in UC, NOS and were not detected in any of the CCA tumors of any site (q<0.001 for both). TMB was significantly higher for UC, NOS tumors (median 9.7 mut/mB, IQR 5.6-15.8) compared to CCAs arising in the urinary tract, ovary (median 4.4 mut/mB, IQR 3-6.1), or uterus (median 4.1 mut/mB, IQR 2.6-6.6) (p<0.01). To determine a potential pathogenic process of invasive utCCA, we performed WES of multiple micro-/macro-dissected regions of dysplasia and invasive cancer from two illustrative cases. The first (patient 31) had CCA arise within a urethral diverticulum, from which the following areas were micro-dissected and analyzed separately by WES: 1) normal-appearing urethral (urothelial) mucosa, 2) diverticular lining, which contained reactive urothelium, 3) diverticular lining with intestinal metaplasia, 4) diverticular lining with atypical/dysplastic clear cells, and 5) two separate regions of the invasive CCA ( Figure 6 ). Notably, there were no mutations shared among all six samples, and no mutations were shared between the invasive CCA and the normal appearing urethral mucosa, the diverticular urothelial lining, or the adjacent diverticular intestinal metaplasia. However, there were 242 shared somatic mutations between the atypical/dysplastic clear cells lining the diverticulum and the invasive CCA, in addition to 255 and 749 private mutations in each, respectively. The results suggest that these dysplastic clear cells lining diverticular spaces represent precursor lesions to the invasive CCA tumors. Although we did not identify driver mutations in genes that were mutated at relatively higher frequency in this cohort, we do highlight two of the shared mutations that occurred in known driver genes in AGO2 and ASXL2 . The specific missense mutations, however, were designated as variants of unknown significance based on the currently available evidence using OncoKB knowledgebase. 36 Further supporting the above observation is the identification of clear cell dysplasia in 15 of the 25 tumors in which the radical resection performed allowed for such evaluation, including two cases in which loss of ARID1A and BRG1 (SMARCA4) expression was similar to that of corresponding invasive tumor (also see Figure 2 and Supplementary Figure 5 ). The second case (patient 13, Supplementary Figure 6 ) had an invasive CCA of the bladder neck adjacent to two distinct regions of marked urothelial atypia. Similar phylogenic analysis as above of the dissected and individually sequenced regions showed no clonal relationship between the invasive CCA and juxtaposed areas of marked urothelial atypia, arguing against a urothelial precursor for this CCA of the urinary tract.

The study was approved by the Memorial Sloan Kettering Cancer Center (MSK) Institutional Review Board (MSK IRB Protocols #12-245, 06-107). All primary utCCA between 1994 to 2023 were identified and available hematoxylin and eosin-stained slides and the accompanying immunohistochemistry (IHC) were reviewed by two genitourinary pathologists (HA, RR) to implement the most recent WHO classification. 14 Patients’ demographic and clinical data were manually extracted from medical records. Cases were excluded if the diagnosis of clear cell adenocarcinoma of the urinary tract could not be confirmed as primary in the urinary tract (i.e., patients with history of clear cell carcinoma of the female genital tract with inability to rule out metastasis/direct extension from the gynecologic tract). Pathologic TNM staging was implemented according to the American Joint Committee on Cancer (AJCC) TNM staging system for the urethra, urinary bladder, renal pelvis and ureter, depending on tumor location. 31 Disease-free survival (DFS) was estimated by Kaplan-Meier method and calculated as time from the most definitive treatment to recurrence of disease, death, or last follow up. Overall survival (OS) was calculated as time to death from any cause. All tumors were stained for PAX8 and HNF-1β, two markers expected to be positive in CCA 19 , 32 , and GATA3 (GATA binding protein 3), which is one of the commonly used markers of urothelial differentiation 33 . Due to association of endometrial clear cell carcinoma with microsatellite instability/Lynch syndrome in a subset of cases 5 , the four most commonly used markers to assess for DNA mismatch repair proteins (MLH1, PMS2, MSH2 and MSH6) were evaluated in 27 tumors with available tissue. IHC for p53, ARID1A, HER2, BRG1 and BRM was performed for select cases based on the corresponding genomic findings. All stains were performed using DAB detection Kit for the development of the stain (correspondent to the staining platform). Detailed information on used antibodies is provided in Supplementary Table 2 . Formalin-fixed, paraffin-embedded (FFPE) tissue was the source of tumor-DNA and patient-matched blood was collected as a source of germline DNA. For 21 patients, DNA sequencing was performed using MSK-IMPACT (Memorial Sloan Kettering Integrated Mutation Profiling of Actionable Cancer Targets), which can detect somatic mutations, copy number alterations, and structural variants in up to 505 cancer-associated genes 34 , 35 . Alterations were classified as oncogenic or likely oncogenic using the OncoKB knowledgebase 36 . Whole exome sequencing (WES) was performed from newly extracted DNA or through re-capture of existing MSK-IMPACT sequencing libraries using the xGen Exome Research Panel v2.0 (Integrated DNA Technologies, Inc., IDT), as previously reported. 35 , 37 - 39 Briefly, for existing, post-capture libraries that corresponded to clinical MSK-IMPACT sequencing, 74–500 ng of remaining barcoded library was captured by hybridization using either the SureSelectXT Human All Exon V4 (Agilent 5190-4632) or xGen Exome Research Panel v1.0 (Integrated DNA Technologies, Inc, IDT) according to the manufacturer’s protocol. PCR amplification of the post-capture libraries was carried out for 8 cycles. Samples were run on NovaSeq 6000 SBS v1 Kit and an S2 flow cell (Illumina) with median WES target coverage of 151X (tumor) and 86X (normal). WES data were analyzed using the TEMPO pipeline (Time-Efficient Mutational Profiling in Oncology; https://github.com/mskcc/tempo ). Briefly, reads were aligned to the human genome (hg19) using BWA MEM, followed by post-processing using the Picard Mark Duplicates tool and GATK 40 . Somatic mutations (point mutations and small insertions and deletions) were identified using MuTect2 41 and Strelka2 42 . Mutational signatures were inferred from single-nucleotide mutations for all sequenced samples with five or more such mutations. The fraction of mutations attributable to each of the known mutational signatures was determined using a basin-hopping algorithm ( https://github.com/mskcc/mutation-signatures with cosmic v3.0). 43 Signatures with a known common source of somatic hypermutation were considered together. Variants of germline and clonal hematopoietic origin are filtered in bioinformatics analysis. The FACETS (Fraction and Allele-Specific Copy Number Estimates from Tumor Sequencing) algorithm and the FACETS-suite package ( https://github.com/mskcc/facets-suite ) were used to generate purity-corrected fraction of genome altered estimates and assess whole-genome duplication (WGD). Tumors were considered to have undergone WGD if at least 50% of their autosomal genome had a major copy number of 2 or more. To gain further insights into the molecular pathogenesis of utCCA, we compared the somatic mutational profiles of utCCA with primary CCA of ovary and uterus and primary invasive UC, not otherwise specified (NOS) that were analyzed using the MSK-IMPACT assay as part of a pan-cancer, institutional scale, prospective tumor sequencing initiative. Two-sided Fisher’s exact test was used for comparison of genomic alteration frequencies among these tumor categories. Continuous variables were compared using a Wilcoxon test. Multiple testing correction was performed using the Benjamini-Hochberg method ( q -value cutoff of 0.1). All analyses were performed using R v4.1.1. When available, we reviewed results of germline variants in ≥76 genes known to have association with hereditary cancer predisposition, including those related to DNA mismatch repair, for which all coding regions were sequenced in both germline normal and tumor tissue. 44 Laser capture microdissection (LCM) was performed on two cases associated with urethral diverticula and additionally contained cystic areas lined by either cuboidal cells with clear cell features, or intestinal metaplasia. Using LCM technique, the different areas of the tumor and associated lesions were separately microdissected and profiled. Twenty sections were cut from each FFPE block with the corresponding areas of interest, stained with H&E and used for LCM. In a third case, macrodissection was performed of two spatially distinct areas of the urothelial lining exhibiting urothelial atypia that was suspicious but not sufficient for the diagnosis of urothelial carcinoma in situ. All micro- and macro-dissected tissue was subjected to WES along with the invasive utCCA from the same bladder. Mutation phylogeny between multiple tumors/lesions from the same patients was inferred using the union of somatic mutations called in any of the matching samples and performed with R package ape 45 .

Clear cell adenocarcinoma of the urinary tract is a molecularly heterogeneous disease with variable mutational profiles and mutational signatures. The somatic mutational landscape of primary utCCA is more like CCA of the female genital tract than to urothelial carcinoma. Our data support a mechanism of disease pathogenesis in which utCCA likely arise from regions of dysplasia, possibly resulting from chronic inflammation in the setting of urinary stasis, and not through tans-differentiation from regions of intestinal metaplasia or divergent differentiation of a precursor urothelial carcinoma.

In this study, we performed a comprehensive histopathological, clinical, and genomic profiling analysis of 35 CCA of the urinary tract treated at a single tertiary care referral center. Consistent with prior studies, we found that utCCA is more common in women and has a propensity to arise in the female urethra, in association with urethral diverticula. Genomic analysis revealed that the somatic mutational profile of utCCA more closely resembles clear cell adenocarcinomas arising in the gynecological tract (ovarian, uterine, cervical primary sites) than invasive urothelial carcinomas. Since the majority of utCCA occur in women, a Müllerian origin has been hypothesized, but other origins such as mesonephric remnant, divergent differentiation from urothelial carcinoma, or progression from nephrogenic adenoma/metaplasia have also been considered 2 , 3 , 6 - 8 , 16 - 19 . In our cohort, no endometriosis or Müllerian or mesonephric remnants were identified in any cases, arguing against the origin from a Müllerian or mesonephric remnant. The strong association with urethral diverticula suggests that chronic inflammation – resulting from urinary stasis – is likely a contributing factor leading to dysplastic changes and malignant transformation of utCCA 27 , 28 , 48 . In this study, intestinal metaplasia in the diverticular lining was identified in 3 cases. As highlighted in one illustrative case ( Figure 6 ), phylogenic analysis did not identify a clonal relationship between the invasive utCCA and these adjacent regions of intestinal metaplasia. Furthermore, the genomic profile of utCCA, characterized by low TMB (3.1 mut/mB), high rates of ARID1A loss-of-function mutations, and a lack of mutations in TERT promoter and other chromatin modifying gene mutations, is distinct from that of UC and argues against divergent differentiation from a classic UC precursor. As an origin from a urothelial carcinoma precursor has been suggested in prior publications, 4 , 16 , 49 such an association has not been confirmed in the more recent literature, making it difficult to determine the true prevalence of this phenomenon. Nonetheless, our study suggests that this association is rare as we observed only one case of mixed urothelial and CCA components that harbored both TERT promoter and ARID1A mutations, among others. Unfortunately, we did not have sufficient material to assess the clonal relatedness of the utCCA and urothelial carcinoma regions of this case. Importantly, we did identify a potential non-invasive precursor lesion in some tumors characterized by atypical or dysplastic clear cells within the lining of diverticular compartments adjacent to the invasive disease. In the case for which we were able to analyze the dysplastic cells separately from the nearby invasive utCCA, the dysplastic cells shared morphologic features and somatic mutations with the adjacent invasive utCCA. These findings support a potential mechanism of disease pathogenesis in which invasive utCCA develops from dysplastic changes affecting the diverticular lining rather than from intestinal metaplasia or a precursor urothelial carcinoma, supporting a distinct pathogenetic pathway from those of classic UC. Conversely, this process does not apply to all tumors as our study showed, particularly those not associated with diverticula , or CCA developing in extravesical locations such as ureter or renal pelvis as documented in two of the tumors in this series as well as in a recently published report of CCA of the renal pelvis. 50 Despite the morphologic similarities between urinary and gynecologic tract CCAs, notable genomic differences exist between these groups. Ovarian CCA had significantly higher rates of ARID1A and PIK3CA mutations and endometrial CCA had more frequent TP53 and PIK3CA mutations than utCCA, but with similar rates of ARID1A mutation. Moreover, both ovarian and endometrial CCA are often associated with DNA mismatch repair deficiency 5 , a finding not identified in any of the urinary tract CCA in our study. Despite representing the largest and most comprehensive molecular analysis of utCCA, the study is limited by its retrospective design which resulted in incomplete clinical or molecular data for some patients. The heterogeneity of treatment approaches and follow up may also limit meaningful clinical comparisons between utCCA and urothelial carcinomas. Thus, conclusions about the optimal management of this disease are limited and will require larger and prospective cohorts.

Clear cell adenocarcinoma of the urinary tract (utCCA) is a rare malignant neoplasm that typically arises in the urethra of adult female patients and histologically resembles CCA of the female genital tract 1 - 5 . utCCA often arises within a urethral diverticulum or the periurethral glands 1 , 6 , 7 . Less frequently, utCCA has been described in men in the bladder or prostate 6 - 10 . The histological pattern of CCA of the urinary bladder was first reported by Dow and Young in 1968 who categorized these tumors as “mesonephric adenocarcinomas” 2 , 11 . Young and Scully later introduced the term clear cell adenocarcinoma in 1985 based on their resemblance to ovarian tumors with the same name 12 , 13 . Since then, only about 250 cases of utCCA have been reported in the English literature ( Table 1 , Supplementary Table 1 ), underscoring the rarity of this entity. In the current World Health Organization (WHO) classification of urinary and male genital tract tumors, utCCA is classified as a Müllerian-type tumor 14 , 15 , but its exact histogenesis or molecular pathogenesis remains largely unknown. A Müllerian cell of origin (from endometriosis or benign Müllerian remnants) has been postulated 2 , 6 , 8 , as well as the possibility of divergent differentiation from urothelial carcinoma 2 , 4 , 7 , 16 , or progression from nephrogenic adenoma 17 - 19 . There is significant morphological and immunohistochemical overlap between utCCA and those of CCA of gynecologic origin and nephrogenic adenoma, which may make the diagnosis of utCCA challenging, especially in small biopsy specimens or in female patients. Once the diagnosis of CCA is suspected, immunohistochemical stains can help to confirm the microscopic impression, particularly the expression of paired box gene 8 (PAX-8) and hepatocyte nuclear factor-1 beta (HNF-1β). PAX8 is a transcription factor with important roles in embryonic development of many organs such as thyroid, kidney, mullerian system and thymus, and is typically expressed in epithelial neoplasms arising from these organs. 19 , 20 HNF-1β is a transcription factor that is vital for embryonic development whose early expression is noted in the kidney, liver, bile ducts, thymus, genital tract, pancreas, lung, and gut, and has roles in glucose metabolism and antiapoptosis. 21 , 22 By immunohistochemistry, it is highly expressed in clear cell adenocarcinoma (87%) but it is not specific to this entity as it is also expressed in other ovarian and gynecologic malignancies. 21 , 23 - 25 Insights into the somatic landscape of driver mutations has been limited to a few small case series 26 - 28 . Most notably, putative oncogenic alterations in tumor suppressor ARID1A , a SWI/SNF family member also commonly mutated in ovarian and endometrial CCA, have been identified in utCCA 5 , 29 , 30 . A subset of endometrial CCA is associated with DNA mismatch repair deficiency (28718916), but no association of utCCA with any genetic predisposition has been reported. Due to its rarity, the natural history of utCCA is poorly understood as most of our knowledge is based on case reports or small series ( Table 1 , Supplementary Table 1 ), and as a result, the clinical management of this disease is not well defined. 7 The aim of this study was to define the molecular features of primary utCCA and their association with clinical, demographic, and cancer specific outcomes.