Clinicopathologic and Molecular Characterization of High-Risk Human Papillomavirus–Positive Carcinomas of the Urinary Tract

Institutional review board approval was obtained for this retrospective study. We identified 12 urethral and urinary bladder cases in which HPV association was confirmed by in situ hybridization for hrHPV in our surgical case files. In brief, we reviewed the pathology reports of all penile, urinary bladder, ureter, and renal pelvis cases in which “HPV” was written in the final diagnosis or diagnosis comment from 2000–2020. Cases with a suspicion of direct extension from the anogenital tract, including penile skin, foreskin or glans mucosa, vulva, cervix, and anus, were excluded. Of the 15 candidate cases, hrHPV in situ hybridization (hrHPV-ISH) was already performed on the case or subsequent case from the same location and was positive in 9 (Supplemental Table 1, see the supplemental digital content containing 3 tables and 2 figures at https://meridian.allenpress.com/aplm in the January 2025 table of contents). For the remaining cases, we performed hrHPV-ISH as part of this study (see below), which resulted in 3 additional positive cases (total 12). The original diagnoses for these cases included urothelial carcinoma, carcinoma with urothelial and squamous features, and SCC (Supplemental Table 1).

For comparison, tumors with predominantly squamous features were preferentially selected owing to HPV⁺ carcinomas typically showing squamous features. We conducted another database search among all of the bladder and urethral cases accessioned between years 2000 and 2020, using the search term squamous written in the final diagnosis. This search identified 200 cases, of which 11 had molecular sequencing data already available. hrHPV-ISH was performed as part of this study and was negative in all 11 cases. These cases were originally diagnosed as SCC, carcinoma with extensive squamous differentiation, and urothelial carcinoma with extensive squamous differentiation (Supplemental Table 1). Relevant clinical history was obtained from the electronic medical record.

All available slides were reviewed. Basaloid features and squamous differentiation were documented as percentage involvement of total tumor area for each case. Basaloid features were defined as nests and ribbons of hyperchromatic tumor cells with high nuclear-to-cytoplasmic ratio, frequent mitotic figures, and necrosis.²⁹  Only moderate to well-differentiated areas of squamous differentiation were quantified and were defined by the presence of intercellular bridges, intracellular keratinization, or the formation of squamous pearls. Representative formalin-fixed paraffin-embedded (FFPE) tumor sections of either the primary resection, biopsy, or a recurrence, based on tissue availability, were selected to perform clinically validated IHC and comprehensive molecular testing.

Immunohistochemical staining for uroplakin II (ready to use [RTU]; BC21, Biocare Medical, Pacheco, California), GATA3 (RTU; L50-823, Ventana Medical Systems, Tucson, Arizona), cytokeratin (CK) 7 (1:100; OV-TL12/30, Dako, Carpinteria, California), CK5/6 (1:200; D5-16B4, Millipore, Billerica, Massachusetts), CK903 (RTU; 34BE12, Enzo Diagnostics, Farmingdale, New York), p63 (RTU; 4A4, Ventana Medical), p16 (RTU; E6H4, Ventana Medical), p53 (1:100; DO-7, Biocare Medical) was performed by using tissue microarrays containing triplicate 1.5-mL punches of tumor from FFPE tissue or on whole slide sections, if not already performed during clinical evaluation. Immunohistochemical staining for uroplakin II, GATA3, CK7, CK5/6, CK903, and p63 was scored as strong and diffuse (++), patchy and weak (+/−), and negative (−). Aberrant p16 staining was defined as blocklike strong nuclear staining.³⁰  p53 wild-type pattern was defined as heterogeneous distribution of variable-intensity nuclear staining. Aberrant p53 expression included overexpression (diffuse strong) and null pattern, as previously described.³¹ 

For all cases, in situ hybridization for hrHPV E6/E7 mRNA was performed to confirm presence/absence of hrHPV, using the RNAscope 2.5 VS Probe-HPV-HR18 (Advanced Cell Diagnostics Inc, Hayward, California) according to the manufacturer’s instructions and 4-μm FFPE tissue sections. The HPV genotypes included in this assay were HPV 16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73, and 82. Positive staining was identified as punctate nuclear and/or cytoplasmic dotlike staining. The slides were classified as positive or negative.

Genomic DNA was extracted from punch biopsies or macrodissected unstained sections, targeting tumor-only from FFPE tissue. Sequencing libraries were prepared from genomic DNA, and target enrichment was performed by hybrid capture using a custom oligonucleotide library (Roche Nimblegen, Madison, Wisconsin). Capture-based next-generation sequencing was performed by using a clinically validated DNA sequencing assay that targets the coding regions of 529 cancer-related genes and select introns from 47 genes with a total sequencing footprint of 2.9 Mb (gene list https://genomics.ucsf.edu/content/ucsf-500-cancer-gene-panel-test-ucsf500-uc500, version 3; UCSF Clinical Cancer Genomic Laboratory, San Francisco, California). Sequencing of captured libraries was performed on a NovaSeq 6000 (Illumina, San Diego, California) as previously described.³²  Molecular analysis was performed blinded to the component being analyzed and the HPV status. Somatic single-nucleotide variants and indels were visualized and verified with the Integrative Genomics Viewer. Genome-wide copy number analysis based on on-target and off-target reads was performed by using CNVkit and Nexus Copy Number (Biodiscovery, Hawthorne, California). Molecular alterations were manually classified as pathogenic or likely pathogenic on the basis of information from the following databases: cBioPortal for Cancer Genomics, ClinVar, and PubMed. Only pathogenic or likely pathogenic molecular alterations are reported and included in the analysis. Tumor mutation burden was calculated from the number of mutations in the coding regions of the genes in the UCSF500 panel, counting single-nucleotide variants and small indels and excluding known germline variants. Microsatellite instability (MSI) was analyzed by MSIsensor v0.2. Samples with 30% or more MSI detection were deemed MSI-high; 20% to less than 30%, indeterminate; and less than 20%, microsatellite stable.

Categorical and quantitative variables were evaluated by using 2-tailed Fisher exact test and 2-tailed Student t test, respectively. Results were considered statistically significant if P < .05.

We identified 12 cases of HPV⁺ UTC: 9 primary resections (75%) and 3 transurethral resections/biopsies (25%), which were treated with primary chemotherapy and/or radiation therapy (Table 1). The cases came from 5 female and 7 male patients with an average age of 66 years at initial diagnosis/primary resection. The primary sites at presentation were penile urethra (7 [58%]), female urethra (2 [17%]), and urinary bladder (3 [25%]); one of the urinary bladder cases also involved the female urethra but was classified as urinary bladder, as the mass was centered in the bladder (case 12). Notably, nearly half the patients (5 of 12, 42%) had lymph node metastases at presentation, including 2 who presented with a bulky groin mass and small primary tumor (cases 5 and 6). Two cases had a history of long-standing instrumentation (1 for urethral stricture disease, case 7; and 1 for neurogenic bladder, case 9). Additionally, 1 patient had concomitant B-cell lymphoma in the same groin lymph node with metastatic HPV⁺ UTC at initial diagnosis, and the same chemotherapy regimen was used to treat both tumors (case 5).

On hematoxylin-eosin–stained sections, the HPV⁺ tumors frequently and extensively showed characteristic basaloid morphology, which was present in all cases and comprised on average 70% (range, 5%–100%) of total tumor area (Figure 1, A and B; Supplemental Table 1). Moderate to well-differentiated squamous areas were seen focally in 8 cases (67%), with total area of squamous differentiation in each tumor averaging 8% (range, 0%–30%) (Figure 1, C). The remaining tumor areas were considered “poorly differentiated,” characterized by sheets and nests of cells with high-grade nuclei and moderate amounts of eosinophilic cytoplasm but without definite keratinization, making it challenging to classify as either urothelial or squamous (Figure 1, D). Two tumors showed noninvasive flat lesions, mimicking conventional urothelial carcinoma in situ (Figure 2, A), and 8 cases showed a noninvasive papillary component, mimicking conventional papillary urothelial carcinoma (Figure 2, B). Noninvasive lesions were also hrHPV positive by ISH (Figure 2, C and D).

We evaluated a panel of IHC markers commonly used for squamous and urothelial differentiation (Table 2): uroplakin II, CK7, GATA3, p63, CK903, and CK5/6. p63, CK903, and CK5/6 were strongly expressed in all cases with stains available. GATA3 was weakly expressed in 8 of 12 (67%) and CK7 was strongly expressed in 4 of 11 (36%) cases. Uroplakin II was weak and rarely seen, appearing in only 1 case (8%).

The most frequent pathogenic variants in HPV⁺ UTC were seen in KMT2C (42%), PIK3CA (42%), and KMT2D (25%), followed by TGFBR2 (17%) genes (Figure 3). One tumor (8%) showed a TERT promoter mutation (TERTp). None of the tumors in the HPV⁺ group had any alteration in TP53 or CDKN2A genes. Copy number analysis showed 1 tumor (8%) with deep deletion in RB1 gene, 1 (8%) with MYCL, and 1 (8%) with EGFR focal amplifications. Tumor mutational burden was variable, averaging 11.7 mutations/megabase (range, 3.4–23.4). All HPV⁺ tumors were microsatellite stable.

Follow-up data were available for 11 patients (Table 1). During follow-up (median, 39 months; range, 9–115 months), 6 of 11 (55%) required treatment for recurrence. Ultimately, 8 patients (73%) had no evidence of disease, 2 (18%) were alive with disease, and 1 patient (9%) died of disease at the end of the follow-up period (Table 1).

We evaluated 11 squamous predominant HPV⁻ UTCs to identify contrasting features that might aid in the detection of HPV⁺ UTCs. Detailed clinical, morphologic, immunohistochemical, and molecular features of these are presented in Supplemental Tables 1 through 3 and Supplemental Figures 1 and 2.

Notably, HPV⁺ UTCs showed more characteristic basaloid features than HPV⁻ squamous predominant UTCs (70% versus 15% of total tumor area, P < .001) and significantly fewer moderate to well-differentiated squamous areas (8% versus 67% of total tumor area, P < .001). On the other hand, HPV⁻ UTCs showed significant immunohistochemical overlap with the HPV⁺ UTCs: diffuse positive CK5/6, CK903, and p63 staining was seen in all the cases. GATA3 and CK7 was variably expressed in 7 (64%) and 8 (73%), respectively. Uroplakin II expression was the only marker to show significant difference, with 54% positivity in HPV⁻ UTCs and 8% positivity in HPV⁺ UTCs (P = .02).

Finally, in contrast to the HPV⁺ UTCs, our HPV⁻ UTCs showed mutational profiles similar to that more commonly reported in urothelial and squamous carcinomas of the lower urinary tract, with most frequently altered genes including TERTp (54%), TP53 (45%), PIK3CA (45%), CDKN2A (36%), RB1 (27%), KMT2D (27%), and KDM6A (27%) (Supplemental Figure 2).^26,33,34  The degree of tumor mutational burden was similar to that of the HPV⁺ group (average, 13.4 mt/bp; range, 6.9–27.5 mt/bp; P = .56). Like the HPV⁺ group, all HPV⁻ cases were microsatellite stable.

Since p53, and indirectly, p16 proteins are the main targets of HPV oncoproteins E6 and E7, IHC was performed to determine whether their expression is altered in HPV⁺ UTCs and whether they can be useful tools for diagnosis. All the HPV⁺ tumors showed aberrant, blocklike p16 expression by IHC (Figure 4, A). Three HPV⁻ tumors (27%) also showed aberrant p16 staining (Figure 4, B), all of which on molecular analysis had a loss-of-function mutation in the RB1 gene (Figure 4, C). Overall, the sensitivity and the specificity of p16 IHC were 100% and 73%, respectively, for detection of HPV⁺ UTCs.

All HPV⁺ tumors had a wild-type p53 expression pattern, consistent with their molecular profile showing lack of TP53 mutations. Seven HPV⁻ tumors (64%) showed aberrant p53 staining: 4 of 7 as overexpression (Figure 4, D) and 3 of 7 as the null pattern (Figure 4, E). All 4 cases with overexpression had a corresponding missense mutation in the TP53 gene on molecular analysis, whereas 1 of 3 cases with the null pattern contained a loss-of-function mutation in TP53 gene (Figure 4, C). The sensitivity and the specificity of p53 IHC for detection of TP53 mutation were 100% and 88%, respectively.

HPV-associated UTCs are uncommon, with limited studies to date. In this study, we found that HPV⁺ UTCs show clinical, morphologic, and molecular characteristics that are similar to what has been reported in HPV⁺ SCCs of other sites and are also distinct from their HPV⁻ UTC counterparts. At presentation, it is well known that HPV⁺ oropharyngeal SCCs frequently present with lymph node metastases in which the primary tumor can be small and hardly detectable by clinical examination alone.³⁵  Indeed, 2 of our cases presented with bulky lymphadenopathy with a small primary site in the urethra.

On morphology, it is also well known that HPV-associated SCCs of varying anatomic locations commonly share characteristic basaloid morphology.^12,36  In our series of 12 HPV⁺ UTCs, all showed similar overtly basaloid components (average 70% of total tumor area), albeit in 3 cases, the basaloid component was only focal (5%–10%). Notably, the remaining tumor areas were typically “poorly differentiated,” with areas of more obvious squamous differentiation only focal to absent (Supplemental Table 1). Review of the morphology for this study highlighted that these poorly differentiated areas could be interpreted as poorly differentiated SCC or poorly differentiated urothelial carcinoma. Likewise, the basaloid components of an HPV⁺ UTC can also mimic conventional high-grade urothelial carcinoma, particularly if there is papillary configuration of the surface tumor, which can resemble an in situ papillary urothelial carcinoma component. Indeed, a subset of our cases was initially called urothelial carcinoma by unsuspecting pathologists. In our own clinical practice, we would recommend testing (1) in any urethral tumor (given that about 30% of urethral carcinomas are HPV⁺ in the literature);²⁸  (2) in any bulky urinary bladder tumor in a female patient to also exclude possibility of gynecologic tract origin; and (3) in the presence of basaloid appearance with focal or absent squamous features. Typically, we would sign these cases out as “HPV⁺ carcinomas with squamous features” with a descriptive comment to ensure origin from other sites has been excluded. While they are morphologically, immunohistochemically, and molecularly similar to HPV⁺ SCC in other sites, given the unique location in the urothelial tract in which the background epithelium is urothelial rather than squamous, and the morphologic overlap between urothelial carcinomas with squamous differentiation and SCCs, we defer official naming to a consensus of urologic pathologists. Finally, while GATA3, p63, CK903, CK5/6, CK7, and uroplakin II may be used to support a urothelial/squamous origin of a poorly differentiated carcinoma, they would not be helpful in distinguishing HPV⁺ from HPV⁻ cases.

HPV⁺ UTCs and HPV⁻ UTCs showed distinct molecular profiles. While there were some commonly overlapping, recurrently altered genes including PIK3CA and KMT2D, the HPV⁺ tumors rarely showed TERTp, TP53, or CDKN2A alterations, which were more common in HPV⁻ UTCs. These findings in our HPV⁻ UTCs are similar to what has been reported for urothelial and non–HPV-associated SCCs of the urinary bladder and in other sites.^20,26  The absence or rarity of alterations in CDKN2A, TP53, and TERTp in HPV⁺ UTCs relative to HPV⁻ UTCs is likely due to HPV pathogenesis and the impact of the E6 and E7 oncoproteins.

First, the HPV E7 oncoprotein binds to the phosphorylated form of pRb and accelerates its degradation, which causes the release of transcription factor E2F and cell cycle progression.³⁷  p16 is a product of CDKN2A, which functions as a tumor suppressor protein, and CDKN2A is commonly mutated in cancers including urothelial carcinomas.²⁶  p16 inhibits the cyclin-dependent kinases CDK4 and CDK6, thereby preventing phosphorylation of pRb and progression to S phase of the cell cycle. When HPV E7 oncoprotein causes degradation of pRb and progression of cell cycle, this in turn increases p16 expression as a feedback mechanism.³⁷  Increased p16 expression is easily detected on IHC and is often used as a surrogate marker for HPV infection. However, expression of p16 is not only dependent on the function of HPV E7 protein, as mutations in RB1 gene can also have a similar effect as E7 protein, with decreased pRB expression and increased p16 expression. Indeed, in 27% of our HPV⁻ cases, p16 expression was found to be aberrant and all had an associated loss-of-function mutation in the RB1 gene. Inactivating mutations and deep deletions in the RB1 gene are reported in 10% and 15% of urinary bladder urothelial carcinomas, respectively.²⁶  Therefore, use of p16 IHC should be limited to screening for HPV and needs to be followed up with hrHPV in situ hybridization. RB1 loss was also seen in one of the HPV⁺ UTCs in our cases, which is surprising because RB1 loss would seem redundant in HPV⁺ tumors owing to the function of HPV E7 protein on pRB protein; however, similar findings have also been reported in HPV⁺ head and neck SCC.²⁰  In our HPV⁺ case with RB1 loss, the loss was due to deep deletion, whereas in the HPV⁻ cases, the RB1 loss was due to point mutations. HPV DNA integration into the host genome is frequently associated with chromosomal structural changes including deletions.^38,39  It would be interesting to learn if the deep deletion in RB1 gene is related to HPV insertion into the genome, though this was not possible in our assay.

Second, the HPV E6 oncoprotein induces degradation of the tumor suppressor protein p53, leading to resistance to cell death.⁴⁰  IHC for p53 was variable in all our HPV⁺ cases, suggesting that not all p53 is degraded. Aberrant p53 expression was seen in 4 HPV⁻ cases as overexpression pattern, all of which had associated missense mutations in TP53 gene. The HPV-associated carcinomas of anogenital and oropharyngeal regions rarely have TP53 mutations, likely because the p53 function is already altered by HPV oncoprotein E6. Similarly, no TP53 mutations were seen in our cohort of HPV⁺ UTCs.

Third, TERTp mutations are one of the most common molecular alterations in urothelial and squamous cell carcinomas of the urinary tract.^26,34  A similar finding to ours was previously reported for penile and head and neck SCC: TERTp mutations were frequent in non–HPV-associated SCC and significantly less frequent in the HPV-associated tumors.^41,42  Interestingly, one of the many functions of the HPV E6 oncoproteins is to promote telomerase activity by multiple mechanisms involving the regulation of transcriptional, epigenetic, and posttranscriptional processes of the TERT gene.^43,44  Furthermore, HPV DNA has been shown to integrate near the TERT locus, with viral enhancers activating TERTp in cervical cancer.^45,46  Therefore, we hypothesize that HPV in our HPV⁺ UTC cases may increase telomerase activation via multiple mechanisms, resulting in cellular immortalization without the need for TERTp mutations. TERT expression studies would be necessary to confirm this, which are beyond the scope of this study.

Limitations to our study include insufficient number of cases and an insufficiently robust control group required to determine the effect of HPV on clinical outcomes in HPV⁺ UTCs. In the oropharynx, HPV⁺ oropharyngeal SCC usually presents at a lower pathologic T stage though with more common lymph node involvement.^47,48  Treatment response and outcomes are better in HPV⁺ oropharyngeal SCC, and studies are underway to determine if less aggressive treatment strategies can be applied for these patients.^15,17  Regardless of their mutational burden, HPV-associated cancers are also more immunogenic owing to expression of viral antigens and therefore good candidates for immunotherapy.¹⁶  Unlike in oropharyngeal and anogenital sites where HPV⁺ SCCs are the most common malignancy, HPV⁺ UTC is rare, making comparative studies difficult. The clinical features and the outcome of the HPV⁺ UTCs are not clear in the current literature.^49–52  In our study, the HPV⁺ cases had more frequent lymph node involvement at presentation and better apparent disease outcomes than our HPV⁻ cases; however, it would be presumptuous to draw this comparison owing to differences in primary location of our HPV⁺ and HPV⁻ UTCs, different treatments received, and also the potentially heterogeneous nature of the HPV⁻ group due to the inability to distinguish with certainty a pure SCC from a urothelial carcinoma with extensive squamous differentiation in which the conventional urothelial component was not sampled. Additionally, there is significant selection bias in the HPV⁻ UTC cases to worse outcomes, as sequencing was performed at the request of treating oncologists in most cases. Nevertheless, molecularly, conventional urothelial carcinoma and SCCs of the urothelial tract share common alterations in TP53, CDKN2A, and TERT genes in the literature^26,34,53  and in our HPV⁻ UTC cohort, which were rarely seen in our HPV⁺ UTCs. The parallels seen in our study to the genetics of HPV-associated tumors in oropharyngeal and anogenital sites further support the molecular findings presented here.^20,41  Finally, the molecular profile of urethral tumors has been largely unknown before this study and therefore our study provides new molecular data on these cases.

In conclusion, HPV⁺ UTCs have a morphologic, immunohistochemical, and molecular presentation similar to HPV⁺ SCCs of other sites, though they may be difficult to distinguish from conventional urothelial carcinoma owing to overlapping morphologic and immunohistochemical features. Given their rarity in the urothelial tract, more studies are needed to better define these tumors’ behavior and treatment strategies.

The authors would like to thank Jolene Guerin for technical support in construction of the tissue microarrays and Tom Rocereto for support in retrieval of slides and tissue blocks.