Purpose:

Extramammary Paget disease (EMPD) is an uncommon skin malignancy whose genetic alterations are poorly characterized. Previous reports identified mutations in chromatin remodeling genes and PIK3CA. In order to unambiguously determine driver mutations in EMPD, we analyzed 87 EMPD samples using exome sequencing in combination with targeted sequencing.

Experimental Design:

First, we analyzed 37 EMPD samples that were surgically resected using whole-exome sequencing. Based on several in silico analysis, we built a custom capture panel of putative driver genes and analyzed 50 additional formalin-fixed, paraffin-embedded samples using target sequencing. ERBB2 expression was evaluated by HER2 immunohisotochemistry. Select samples were further analyzed by fluorescence in situ hybridization.

Results:

A median of 92 mutations/sample was identified in exome analysis. A union of driver detection algorithms identified ERBB2, ERBB3, KMT2C, TP53, PIK3CA, NUP93, AFDN, and CUX1 as likely driver mutations. Copy-number alteration analysis showed regions spanning CDKN2A as recurrently deleted, and ERBB2 as recurrently amplified. ERBB2, ERBB3, and FGFR1 amplification/mutation showed tendency toward mutual exclusivity. Copy-number alteration load was associated with likelihood to recur. Mutational signatures were dominated by aging and APOBEC activation and lacked evidence of ultraviolet radiation. HER2 IHC/fluorescence in situ analysis validated ERBB2 amplification but was underpowered to detect mutations. Tumor heterogeneity in terms of ERBB2 amplification status was observed in some cases.

Conclusions:

Our comprehensive, unbiased analysis shows EMPD is characterized by alterations involving the PI3K–AKT pathway. EMPD is distinct from other skin cancers in both molecular pathways altered and etiology behind mutagenesis.

Translational Relevance

Extramammary Paget disease (EMPD) is an uncommon skin malignancy whose genetic alterations are poorly characterized. Some EMPD cases show HER2 (ERBB2) amplifications, and sporadic case reports describe efficacy of the HER2 inhibitor trastuzumab to HER2-amplified EMPD. We have analyzed 87 EMPD cases to unambiguously characterize mutational landscape of EMPD. Our results show that ERBB2 amplifications account for a fraction of EMPD cases; rather, ERBB2 mutation/amplification, ERBB3 mutation, and FGFR1 amplification are observed in a mutually exclusive manner, and they are likely to represent disease subsets. We have also found that greater copy-number alteration load is correlated with a greater chance of recurrence. Our data emphasize the need for genetic analysis in treating EMPD; genetic analysis would inform physicians of potential drug targets and help to identify patients at higher risk for recurrence.

Extramammary Paget Disease (EMPD) is a rare skin malignancy that shows a propensity to occur in anogenital/axillary areas of the elderly population (1). The tumor cells are believed to originate from apocrine glands given that EMPD tends to occur in areas rich in apocrine glands (2). Alternatively, Toker cells, a normal constituent of vulvar epithelium, are hypothesized to give rise to EMPD (3). Surgical resection is the standard care for EMPD, but 18% of the cases require systemic therapy due to inoperability or metastasis at some point of the disease, although there is little evidence showing the efficacy of systemic therapy (4). Several reports suggest a role of trastuzumab for those cases showing HER2 overexpression, which has not been validated in clinical trials (5, 6). Other authors who investigated mutations in 10 cancer-related genes in 144 EMPD samples reported frequent RAS pathway mutations, which were found in 19% and 35% of mutations in RAS/RAF and PIK3CA/AKT1 pathway mutations, respectively (7). Two other studies conducted whole-exome sequencing (WES) in 3 and 21 EMPD samples, respectively (8, 9). However, these studies were not large enough to sensitively identify driver mutations nor did they analyze copy-number alterations (CNA). In order to better characterize the driver alterations, which are responsible for the development of EMPD and might potentially be targeted for therapy, we have undertaken an unbiased genomic study, in which 87 tumors were analyzed using combined exome and targeted sequencing.

Patients

For WES, patients with pathologically confirmed EMPD treated at the participating hospitals between 2016 and 2018 were considered for inclusion (Supplementary Table S1). The median follow-up was 3 years (0–9 years). Written informed-consent forms were obtained from all patients. The study was conducted in accordance with the Declaration of Helsinki, and relevant regulatory laws. This study was approved by the institutional review boards of the participating hospitals: Kyoto University Hospital, Kagoshima Medical Center, Kyushu University, Wakayama Medical University, Keio University, Tsukuba University, and Hokkaido University.

For targeted-sequencing experiments, formalin-fixed, paraffin-embedded (FFPE) blocks were retrieved according to the procedure approved by Keio University and Kyoto University. All cases included in the study occurred in the anogenital region.

DNA extraction

Tumor specimens were recovered immediately after surgical resection. Board-certified dermatologists recovered tumor sites using a 4- to 6-mm punch biopsy knife. After excess tissues were trimmed to enrich tumor cells, DNA was extracted using DNeasy Tissue and Blood kit (Qiagen) according to the manufacturer's protocol. Paired normal DNA samples were extracted from patients' blood using the QuickGene Blood kit (Kurabo) or the DNeasy Tissue and Blood kit (Qiagen). DNA was extracted from FFPE samples after microdissection using the GeneRead FFPE kit (Qiagen) according to the manufacturer's protocol.

Sequencing

For WES, DNA libraries were prepared using SureSelect XT (Agilent) and the libraries were subjected to hybrid capture using SureSelect Human All Exon V6 (Agilent) according to the manufacturer's protocol. Prepared library was sequenced with HiSeq 2500 platform (Illumina).

For targeted sequencing, the DNA libraries were generated using SureSelect XT low input (Agilent) and then subjected to hybrid capture using the custom panel according to the manufacturer's protocol. Prepared libraries were sequenced on HiSeq 2500 platform (Illumina). Tumor purity in WES samples was assessed using Sequenza software. Tumor purity of FFPE samples was evaluated histologically.

Mutation calling

Sequenced bcl files were demultiplexed using bcl2fastq (Illumina). Mutation calling was performed using Genomon2 pipeline (https://genomon.readthedocs.io/ja/latest/) as described previously (10). In brief, sequence reads were aligned to the human reference genome (GRCh37) by the Burrows–Wheeler Aligner (version 0.7.10). PCR duplicates were marked with Picard tools (version 1.3.9) and subjected to mutation calling by EBCall (version 2). Mutations called were annotated with ANNOVAR (11). Mutation calls from the exome data were adopted if they met all of the following criteria: the mutation calls were supported by 5%≥ of all the independent reads and ≥4 independent reads in tumor samples, variant reads were observed in 5%< of the germline control, Fisher exact test P values < 0.158, and EBCall P values < 0.001. Tumor-only calling was performed for the targeted-sequencing experiment. Mutation calls were adopted if they met all of the following criteria: mutated genes were included in the target panel, variant frequency >4% and <40%, 3 or more variant reads were observed, mutations were observed in both sense and antisense strands, and EBCall P values < 0.0001. Candidate mutation calls were considered germline SNPs and discarded if the mutations were observed in ≥0.1% in SNP population data sets including ExAC, human genetic variation database, NHLBI exome sequencing project, and the 1000 Genomes Project.

Validation experiments

Validation of mutations detected by WES was performed using PCR-based deep (>500×) sequencing as described previously (12). Briefly, 398 SNVs from 8 samples were randomly selected, and each sequence was amplified from tumor and normal DNAs by multiplex PCR reactions using primers with NotI linkers attached to them. Amplification products were verified by gel electrophoresis, and these products were combined to form tumor and normal pools. Each pool was digested with NotI, and purified with FastGene Gel/PCR Extraction kit (Nippon Genetics). Then, each pool was ligated with T4 DNA ligase for 5 hours, followed by repurification with the aforementioned kit. Each DNA pool was sonicated into ∼200 bp fragments using focused-ultrasonicator E220 (Covaris). DNA libraries were generated using NEBNext Ultra DNA Library Prep kit (New England BioLabs) according to the manufacturer's protocol. These libraries were sequenced on HiSeq 2500 (Illumina) to the average depth of 500×.

Loci that were sequenced to 200× or more in both tumor and normal samples were considered evaluable. Of 343 (86.7%) SNVs successfully amplified using amplicon-based sequencing, 335 (97.7%) were confirmed (Supplementary Fig. S1; Supplementary Table S2).

Two tumors were sequenced in both WES of fresh frozen samples and targeted sequencing of FFPE samples: among 15 mutation calls from FFPE analysis, 13 calls were identical to the ones in fresh sample analysis. Examining the bam files revealed that the remaining two mutations are supported by reads in tumor samples at low variant frequencies (5.1% and 4.0%, respectively). These two mutations were likely detected only in the targeted-sequencing experiment because of its higher depth. We concluded that the two experiments showed good concordance.

Copy-number analysis

Aligned reads were subjected to copy-number analysis using our in-house pipeline, CNACS as previously described (13). All CNA calls were manually inspected. We assessed the sensitivity of CNACS to determine the minimum tumor purity required for copy-number analysis. First, we generated simulated low purity data by mixing tumor cell line WES reads (lung cancer cell line; ATCC CRL-5929) and matching normal WES reads (matching lymphoblast cell line; ATCC CRL-5969) in silico. Then, we ran CNACS analysis on these simulated files. CNA lesions were detectable at 20% tumor purity, but the sensitivity was compromised at 15% or 10% (Supplementary Fig. S2). Based on the benchmark results, we included samples with 20% or higher tumor purity for copy-number analysis, and survival analysis. Sixty-six of 87 samples met the criteria. Log-rank test was used to calculate P values.

CNA calls from exome data were subjected to GISTIC 2.0 analysis, and recurrent CNA lesions were identified with P values < 0.05.

Target panel generation

Among those genes found to be mutated in WES, we selected 126 genes as driver candidates and included in the custom capture panel that was used for targeted-capture sequencing. We used the following criteria for inclusion:

  • Genes mutated in four (10.8%) or more cases in the WES cohort

  • Genes with known hotspots listed in the cancer hotspots database (14)

  • Genes listed in The Cancer Gene Census as of August 7, 2018, and mutated in ≥2 samples (15)

  • Genes showing frameshift/nonsense mutations ≥2 samples

  • Genes in which mutations were clustered within 6 base pairs

  • Genes with a global P < 0.1 in dNdScv analysis (16)

  • Genes with P < 0.02 in MutSigCV (17)

  • Genes with driver Q value < 0.1 in 20/20plus analysis (18)

  • Genes that overlap with GISTIC significant peaks (19)

  • Genes that are reported to be mutated in EMPD and other skin cancer types by previous literature

  • Manually curated genes that are related to the genes of interest (i.e., including ERBB4 and EGFR based on inclusion of ERBB2)

Our target panel covered the entire exonic regions of the selected genes. Our target panel also included 1,250 SNP sites that spanned the entire genome (Supplementary Table S3). These SNP sites were used for copy-number analysis. A full list of genes included in the custom capture panel is presented in Supplementary Table S4. A custom capture panel was built and purchased from Agilent SureDesign.

Bioinformatic analyses

Pathway enrichment analysis was conducted using a bioconductor package, ReactomePA (ver. 1.28.0; ref. 20). Briefly, a list of genes that were mutated in five or more patients were generated and then subjected to enrichment analysis. P values were calculated using the hypergeometric model.

De novo extraction of mutational signatures were performed by running pmsignature (ver. 0.2.1) on exome samples with the number of signatures (k) set to 4 (21). Additionally, deconstructSigs (ver. 1.9.0) was run on each sample against known COSMIC signatures to verify robustness of the pmsignature analysis.

dN/dS analysis was performed using dNdScv (ver. 0.0.1; ref. 16). We performed two analyses: first, the WES data set was analyzed individually in order to select candidates for designing the custom capture panel. After the targeted sequencing was complete, WES and targeted-seq data sets were combined and analyzed, limiting to the genes targeted by the custom panel.

The following software packages were used to conduct corresponding analyses: 2020plus (ver. 1.2.0), GISTIC (2.0), and R (3.6.0; refs. 18, 19). ERBB2 tertiary structure analysis was performed on the MuPit interactive server (22). All analyses were performed using the default parameters unless stated otherwise.

IHC

IHC was performed using a Ventana Benchmark Ultra (Roche Diagnostics) or a BOND-RX (Leica Biosystems) automated immunostainer according to the manufacturer's recommendation. The Ventana PATHWAY anti-HER2 rabbit monoclonal antibody (clone 4B5, 790-2991, Roche) was used as the primary antibody with the ultraView DAB detection kit (760-500, Roche). Anti-phospho-Akt rabbit monoclonal antibody (clone D9E, Cell Signaling Technology) was used with a tyramide signal amplification kit (CSA II K1497, Agilent) and CSA II Rabbit Link (K1501, Agilent). Anti-phospho-Erk1/Erk2 rabbit monoclonal antibody (clone D13.14.4E, Cell Signaling Technology) was used with BOND Polymer Refine Detection (DS9800, Leica). HER2, phospho-AKT, and phosphor-ERK signals were quantified using the scoring systems reported previously (23, 24).

Fluorescent in situ hybridization

HER2 fluorescent in situ hybridization (FISH) was performed on FFPE tissue sections with the Aquarius HER2 Amplification Probe and Tissue Pretreatment kit (LPS001, LPS100, Cytocell) according to the manufacturer's recommendation with minor modifications. Briefly, heat pretreatment and protease digestion procedures were altered to 35 minutes at 95°C and 20 minutes at 37°C, respectively. The probe consists of a red probe for HER2 gene and a green probe for the chromosome 17 centromere. 4′,6-Diamidino-2-phenylindole (DAPI) was used for nuclear counterstain.

FGFR1 FISH was performed using a bacterial artificial chromosome clone (RP11-357D8)-derived DNA probe for FGFR1 and Vysis CEP 8 (D8Z2) SpectrumGreen Probe for the centromere of human chromosome 8 (Abbott Molecular). Pretreatment using VP-2000 Processor (Abbott Molecular) was performed according to the manufacturer's protocol, and hybridization was carried out using the ThermoBrite (Abbott Molecular).

Digital microscopy

The sections were imaged with a BX45 microscope (Olympus) equipped with a DP27 digital camera (Olympus) or a BX63 fluorescence microscope (Olympus) equipped with a DP80 dual-CCD camera (Olympus) and with fluorescent filter sets (Chroma) for red, green, and DAPI fluorescence of the HER2 FISH samples. All images were captured and composited using CellSens Dimension imaging software (Olympus).

Identification of driver genes in EMPD

We analyzed paired tumor/normal DNA from 37 EMPD patients using WES at a mean depth of 120× for tumor and 116× for normal samples. Except for one case who was treated with paclitaxel, all cases did not receive prior systemic therapy. We detected a total of 6,416 mutations with a true-positive rate of 97.7% as confirmed by amplicon sequencing (Supplementary Tables S2 and S5). The median mutation rate in 37 EMPD samples was 92 mutations per exome (range, 2–445), or 2.78 mutations/Mb (range, 0.06–13.5), which roughly corresponded to the mutation rates reported for head and neck cancer (25). None of the tumors showed a hypermutation phenotype. dN/dS analysis identified KMT2C, ERBB2, TP53, PIK3CA, and CUX1 as positively selected genes (Fig. 1). We further interrogated possible driver genes using the 20/20plus criteria for the prediction of driver genes, combining multiple features such as positional clustering of mutations, conservation across species, in silico pathogenicity scores, and protein interaction network connectivity, and identified additional candidates of driver genes, including NIPBL and AFDN. To further interrogate driver genes involved in EMPD, we performed targeted-capture sequencing of 126 candidate genes in 50 FFPE samples at a mean depth of 491× (Supplementary Table S6). dN/dS analysis of the combined data set identified 8 positively selected genes: TP53, ERBB2, CUX1, KMT2C, ERBB3, NUP93, PIK3CA, and AFDN (Supplementary Table S7). All, but NUP93, were tier 1 cancer genes registered in the Cancer Gene Census (15).

CUX1 is a cancer gene that is likely to have dual roles as oncogene and tumor suppressor gene in human (26). In our EMPD cohort, CUX1 was deleted in 19.5% of the cases (17/87). CUX1 was often affected by truncating mutations; 9.2% of the cohort (8/87) showed nonsense mutations or splice-site mutations. The pattern of mutation in our cohort suggests the tumor suppressor role of CUX1 in EMPD.

PIK3CA was mutated in 11 tumors (12.6%). Of 11 tumors, two harbored known hotspot mutations (E542K and E545K; Supplementary Fig. S3). There were three more cases that harbored mutations reported more than 10 times in the COSMIC database (G118D, E453K, and M1043I; refs. 15, 27). These three mutations are experimentally shown to confer gain of function (28–30). Activation of PIK3CA likely plays a role in these EMPD cases.

NUP93 was mutated in eight tumors (9.2%), of which five showed identical nonsense mutations in exon 2 (Q15*) and other two cases had a missense mutation affecting an adjacent amino acid (E14K), which corresponded to a putative hotspot observed in the COSMIC database (Supplementary Fig. S4; ref. 15). The identical NUP93 (E14K) mutation was implicated in metastatic breast cancer, where the mutation was shown to enhance migration of breast cancer cells in vitro (31). Moreover, a recent pan-cancer analysis reported new NUP93 mutations, seen across multiple cancer types albeit at low frequencies (32). NUP93 encodes a nucleoporin that comprises a nuclear pore complex and represses HOXA expression by interacting with chromatin (33). Another report showed NUP93 regulates genes associated with cell migration and is essential for tumor propagation (34). Taken together, NUP93 may represent an underappreciated driver in cancer. Alternatively, NUP93 may be recurrently mutated simply because the sequence context is a preferred substrate for APOBEC3A-mediated mutation (35).

Copy-number analysis shows extensive CNAs in some EMPD cases

CNAs inferred from exome data were subjected to GISTIC analysis, an in silico analysis tool to identify segments recurrently affected by somatic CNAs (19). GISTIC identified a sharp amplification peak on chromosome 17, corresponding to ERBB2 and deletion on chromosome 9, corresponding to CDKN2A (Fig. 2A; Supplementary Tables S8 and S9). Although marginally significant, CDKN2A was ranked high in dN/dS analysis with Q value of 0.08. These data speak to the importance of CDKN2A, a well-known tumor suppressor gene, in EMPD. GISTIC also identified a recurrent deletion peak at 6q27 spanning the AFDN gene. AFDN mutation was also positively selected in dN/dS analysis and 4/4 cases were nonsense mutations. Inactivation of AFDN likely plays a role in a subset of EMPD. GISTIC also detected significant arm-level events. Arm-level amplification was frequent in chromosomes 8q, 16p, and 21q.

CNAs from exome and targeted sequence data are combined in order to further understand the CNA landscape of EMPD (Fig. 2B; Supplementary Fig. S5). Tumors with high CNAs, as defined by tumors harboring >200 Mb lesions in total (near 75 percentile of total CNA load), are more likely to recur (P = 0.002; Fig. 2C).

Three cases showed extensive CNAs, suggesting complex structural variations. In particular, one case showed high MDM2 amplification (Fig. 2D). MDM2 is known to inhibit TP53 activity, and MDM2 amplification is correlated with the increased number of structural variations in melanoma (36). Loss of TP53 activity likely explains extensive CNAs in this case.

Amplifications and/or mutations in ERBB2, ERBB3, and FGFR1 represent distinct EMPD subtypes

Pathway analysis of frequently mutated genes showed an overrepresentation of genes associated with the PI3K–AKT signaling pathway, with the most frequent mutations/amplifications in ERBB2 (29/87, 33.3%), followed by ERBB3 mutations and FGFR1 amplifications (Table 1). Sixty-eight samples had neither ERBB2 amplifications nor mutations. Of these samples, 16 showed ERBB3 mutations. Interestingly, the ERBB3 mutational status showed a trend toward mutual exclusivity with regard to the ERBB2 mutation/amplification status (P = 0.098, Fisher exact test). In addition, of the remaining 52 tumors, three harbored a focal amplification of FGFR1, a tyrosine-protein kinase frequently amplified in lung and breast cancers (37, 38) FGFR1 amplification was confirmed by FISH analysis (Supplementary Fig. S6). Overall, 48 tumors (55.2%) harbored mutations/amplifications in these tyrosine kinase genes, suggesting that alterations in ERBB2, ERBB3, and FGFR1 represented disease subtypes. TP53 was variably affected in these tumors. All three cases with an FGFR1 amplification carried TP53 mutations, whereas 9 of 29, and 2 of 16 cases with mutated/amplified ERBB2 and ERBB3 had deletions/mutations in TP53.

Of 21 (21.6%) ERBB2-mutated cases analyzed, eight showed S310Y substitution, which was identical to the known hotspot registered in the catalog of somatic mutation in cancer (15). The transforming capacity of ERBB2 S310Y was characterized previously (39). Seven mutations clustered in tyrosine kinase domain (Fig. 3A and B). Among them, two cases harbored T862A substitution that was shown to be sensitive to HER2 inhibition by lapatinib in cell line experiments (40). On the other hand, two cases carried T733I mutation, which was shown to confer resistance toward HER2 inhibition by lapatinib (41). Taken together, mutations in tyrosine kinase domains are likely impacting the protein's function in an activating manner.

To further assess the importance of PI3K–AKT pathway activation in EMPD, 10 cases with mutations in ERBB2 or ERBB3 (10/87, 11.5%) were stained for phospho-AKT (pAKT), an activation marker of the PI3K–AKT pathway, and phospho-ERK (pERK), an activation marker of the MAPK pathway. All 10 cases with ERBB2 or ERBB3 mutations showed positive staining for pAKT, confirming PI3K–AKT pathway activity in these cases. Positivity rate for pERK was lower than that for pAKT (pERK: 5/10, pAKT: 10/10), suggesting EMPD favors the PI3K–AKT pathway over the MAPK pathway in cell survival signaling (Supplementary Fig. S7).

IHC of ERBB2 reveals intratumor heterogeneity in some EMPD cases

Given that ERBB2 was the most frequent target of genetic alterations in our EMPD cohort, we analyzed all cases for expression of ERBB2 by IHC. As we expected, ERBB2 amplification status was well correlated with ERBB2 overexpression, supporting the biological relevance of ERBB2 amplification in these cases (P < 0.0001, Mann–Whitney U test; Fig. 4A). However, ERBB2 mutation status did not correlate with IHC.

There are six cases in which IHC staining and ERBB2 amplification status do not correlate (hereafter referred to as “NC tumors”). Upon close inspection, some of NC tumors can be attributed to technical factors such as low tumor contents or low-quality library preparation that compromised the sensitivity of copy-number analysis. Intriguingly, there are NC tumors in which tumor heterogeneity is strongly suspected, as only a part of the section shows strong positivity for ERBB2 (Fig. 4B). We analyzed two cases with HER2 FISH. Indeed, in parts where definite IHC staining was observed, ERBB2 was strongly amplified while in the remaining cells, ERBB2 amplification status remained normal. These findings suggest that ERBB2 amplification was acquired later in the tumor development in some of the NC tumors.

Mutational signature analysis shows contribution of APOBEC activation and aging

Mutational signature analysis enables us to decipher the etiology behind mutations (25). Using nonnegative matrix factorization of mutation counts, we have extracted three signatures from our exome data; they closely resemble COSMIC signatures 1 (cosine similarity: 0.925), 2 (cosine similarity: 0.893), and 13 (cosine similarity: 0.812; Fig. 5A and B; Supplementary Fig. S8). COSMIC signature 1 is associated with aging whereas COSMIC signatures 2 and 13 are known to be related to APOBEC activation.

To our knowledge, this is the largest study of an unbiased analysis of somatic mutations ever performed on EMPD. Our study has clarified a comprehensive list of putative driver mutations, providing new insight into carcinogenesis of EMPD. Our analysis shows that EMPD is different from other skin cancers with respect to both mutational landscape and underlying mutagenic insults that are relevant to its carcinogenesis. Unlike other skin cancers, EMPD carries a relatively low mutational burden probably because EMPD preferentially involve the groin/axillary areas that are minimally exposed to ultraviolet (UV) light. In fact, mutational signature analysis of EMPD showed a paucity of UV-related signatures (COSMIC signature 7), which is predominant in other skin cancers (42–44). These differences might reflect a different cellular origin of EMPD from other skin cancer types (i.e., the apocrine gland, but not keratinocyte or melanocytes).

We identified putative driver mutations in genes encoding receptor tyrosine kinases; converging into PI3K–AKT pathway, lesions involving ERBB2, ERBB3, and FGFR1 may define disease subtypes. Given that more than 50% in our cohort harbor potentially actionable alterations, it is imperative to identify EMPD patients suitable for targeted therapies. Previous reports have suggested utility of HER2 IHC/FISH studies for selecting EMPD cases for trastuzumab treatment (5, 6, 45). Our data show that IHC/FISH studies are unlikely to identify cases with ERBB2 mutations. We propose that clinical sequencing, in addition to HER2 IHC/FISH analyses, should be offered for EMPD patients to correctly identify ERBB2, ERBB3, and FGFR1 alterations. Utility of sequencing studies in clinical setting is further supported by our observation that CNA burden correlated with patients' prognosis in EMPD. Since treatment of EMPD is often complicated by its propensity to recur locally, assessment of CNA burden may aid stratification of patients' care or surveillance after surgeries. A potential caveat is the intratumor heterogeneity observed in some cases; combining sequencing analysis and HER2 IHC/FISH would aid clinicians to identify such cases, which may be less suitable for targeted therapy.

Another clinical implication is that EMPD may respond to PI3K inhibitors. In our cohort, 11 of 87 EMPD cases harbored mutations in PIK3CA, of which five are known activating mutations. Given that PI3K inhibitors have successfully been used for the treatment of PIK3CA-mutated breast cancer patients, there is a strong rationale to consider PI3K inhibitors in these cases (27, 46). Moreover, some of the EMPD cases without PIK3CA mutations may benefit from PI3K inhibition because our genomic analysis and IHC data point to the role of PI3K–AKT pathway activation in EMPD.

The driver mutations detected in our study were considerably different from those reported in a previous study by Zhang and colleagues, even though both used WES (Fig. 6; ref. 9). Although both found KMT2C, TP53, and PIK3CA as potential drivers, Zhang and colleagues did not identify alterations in ERBB2 or ERBB3. Such a difference might have resulted from different experimental conditions between the studies; Zhang and colleagues used whole-genome amplifications that were prone to generate artifacts (47). Alternatively, ethnic or geographical difference may have given rise to the difference, a phenomenon observed in lung cancer or prostate cancer (48).

In summary, our study clarifies the mutational landscape of the largest EMPD cohort by using validated WES and targeted analyses. We have identified driver mutations in EMPD, including potentially actionable ones such as ERBB2, ERBB3, and FGFR1. These data point to future clinical trials assessing the role of targeted therapies for EMPD. Our study provides insights into its distinct origin by highlighting the uniqueness of EMPD in its mutagenesis and mutational landscape.

Y. Ishida reports grants from Japan Society for the Promotion of Science during the conduct of the study. N. Kakiuchi reports grants from Japan Society for the Promotion of Science during the conduct of the study. Y. Inoue reports grants from JSPS KAKENHI during the conduct of the study. H. Irie reports grants from Japan Society for the Promotion of Science during the conduct of the study. S. Matsushita reports grants and personal fees from Ono Pharmaceutical Co. Ltd; personal fees from Bristol-Myers Squibb, Maruho, MSD, Novartis, Merck Serono, and Kyowa Kirin; and grants from Kaken Pharmaceutical Co. Ltd outside the submitted work. K. Takeuchi reports personal fees from Chugai, MSD, Takeda, Janssen, Eizai, Cellgene, and Yakult and grants and personal fees from Kyowa Kirin outside the submitted work. A. Otsuka reports personal fees from Ono, BMS, Novartis, Sanofi, MSD, Torii, and Kyowa Kirin and grants and personal fees from Eisai outside the submitted work. K. Kabashima reports grants from Japan Society for the Promotion of Science during the conduct of the study. No disclosures were reported by the other authors.

Y. Ishida: Investigation, visualization, writing–original draft. N. Kakiuchi: Conceptualization, resources, project administration, writing–review and editing. K. Yoshida: Conceptualization, resources, writing–review and editing. Y. Inoue: Validation and investigation. H. Irie: Investigation. T.R. Kataoka: Resources and investigation. M. Hirata: Resources and investigation. T. Funakoshi: Resources, writing–review and editing. S. Matsushita: Resources, writing–review and editing. H. Hata: Resources and writing–review and editing. H. Uchi: Resources, writing–review and editing. Y. Yamamoto: Resources, writing–review and editing. Y. Fujisawa: Resources, writing–review and editing. T. Fujimura: Resources, writing–review and editing. R. Saiki: Software. K. Takeuchi: Investigation. Y. Shiraishi: Resources. K. Chiba: Resources. H. Tanaka: Resources. A. Otsuka: Conceptualization, resources, supervision, funding acquisition, writing–review and editing. S. Miyano: Resources. K. Kabashima: Supervision, funding acquisition, and project administration. S. Ogawa: Supervision, funding acquisition, project administration, writing–review and editing.

This work is supported by JSPS KAKENHI (JP19J12844 to Y. Ishida, 15H05909 and 19H05656 to S. Ogawa). Satoko Baba conducted FGFR1 FISH experiment. All data have been deposited at European Genome-Phenome Archive (https://ega-archive.org/) under the accession number EGAS00001004746.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Cohen
JM
,
Granter
SR
,
Werchniak
AE
. 
Risk stratification in extramammary Paget disease
.
Clin Exp Dermatol
2015
;
40
:
473
8
.
2.
Mazoujian
G
,
Pinkus
GS
,
Haagensen
DE
. 
Extramammary Paget's disease–evidence for an apocrine origin: an immunoperoxidase study of gross cystic disease fluid protein-15, carcinoembryonic antigen, and keratin proteins
.
Am J Surg Pathol
1984
;
8
:
43
50
.
3.
Willman
JH
,
Golitz
LE
,
Fitzpatrick
JE
. 
Vulvar clear cells of toker: precursors of extramammary Paget's disease
.
Am J Dermatopathol
2005
;
27
:
185
8
.
4.
Hatta
N
,
Yamada
M
,
Hirano
T
,
Fujimoto
A
,
Morita
R
. 
Extramammary Paget's disease: treatment, prognostic factors and outcome in 76 patients
.
Br J Dermatol
2008
;
158
:
313
8
.
5.
Barth
P
,
Al-Saleem
ED
,
Edwards
KW
,
Millis
SZ
,
Wong
YN
,
Geynisman
DM
. 
Metastatic extramammary Paget's disease of scrotum responds completely to single agent trastuzumab in a hemodialysis patient: case report, molecular profiling and brief review of the literature
.
Case Rep Oncol Med
2015
;
2015
:
1
7
.
6.
Karam
A
,
Berek
JS
,
Stenson
A
,
Rao
J
,
Dorigo
O
. 
HER-2/neu targeting for recurrent vulvar Paget's disease A case report and literature review
.
Gynecol Oncol
2008
;
111
:
568
71
.
7.
Kang
Z
,
Xu
F
,
Zhang
Q-a
,
Wu
Z
,
Zhang
X
,
Xu
J
, et al
Oncogenic mutations in extramammary Paget's disease and their clinical relevance
.
Int J Cancer
2013
;
132
:
824
31
.
8.
Kiniwa
Y
,
Yasuda
J
,
Saito
S
,
Saito
R
,
Motoike
IN
,
Danjoh
I
, et al
Identification of genetic alterations in extramammary Paget disease using whole exome analysis
.
J Dermatol Sci
2019
;
94
:
229
35
.
9.
Zhang
G
,
Zhou
S
,
Zhong
W
,
Hong
L
,
Wang
Y
,
Lu
S
, et al
Whole-exome sequencing reveals frequent mutations in chromatin remodeling genes in mammary and extramammary Paget's diseases
.
J Invest Dermatol
2019
;
139
:
789
95
.
10.
Kataoka
K
,
Shiraishi
Y
,
Takeda
Y
,
Sakata
S
,
Matsumoto
M
,
Nagano
S
, et al
Aberrant PD-L1 expression through 3′-UTR disruption in multiple cancers
.
Nature
2016
;
534
:
402
6
.
11.
Wang
K
,
Li
M
,
Hakonarson
H
. 
ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data
.
Nucleic Acids Res
2010
;
38
:
e164
.
12.
Yoshida
K
,
Sanada
M
,
Shiraishi
Y
,
Nowak
D
,
Nagata
Y
,
Yamamoto
R
, et al
Frequent pathway mutations of splicing machinery in myelodysplasia
.
Nature
2011
;
478
:
64
9
.
13.
Yoshizato
T
,
Nannya
Y
,
Atsuta
Y
,
Shiozawa
Y
,
Iijima-Yamashita
Y
,
Yoshida
K
, et al
Genetic abnormalities in myelodysplasia and secondary acute myeloid leukemia: impact on outcome of stem cell transplantation
.
Blood
2017
;
129
:
2347
58
.
14.
Chang
MT
,
Bhattarai
TS
,
Schram
AM
,
Bielski
CM
,
Donoghue
MTA
,
Jonsson
P
, et al
Accelerating discovery of functional mutant alleles in cancer
.
Cancer Discov
2018
;
8
:
174
83
.
15.
Tate
JG
,
Bamford
S
,
Jubb
HC
,
Sondka
Z
,
Beare
DM
,
Bindal
N
, et al
COSMIC: the catalogue of somatic mutations in cancer
.
Nucleic Acids Res
2019
;
47
:
D941
D7
.
16.
Martincorena
I
,
Raine
KM
,
Gerstung
M
,
Dawson
KJ
,
Haase
K
,
Van Loo
P
, et al
Universal patterns of selection in cancer and somatic tissues
.
Cell
2017
;
171
:
1029
41
.
17.
Lawrence
MS
,
Stojanov
P
,
Polak
P
,
Kryukov
GV
,
Cibulskis
K
,
Sivachenko
A
, et al
Mutational heterogeneity in cancer and the search for new cancer-associated genes
.
Nature
2013
;
499
:
214
8
.
18.
Tokheim
CJ
,
Papadopoulos
N
,
Kinzler
KW
,
Vogelstein
B
,
Karchin
R
. 
Evaluating the evaluation of cancer driver genes
.
Proc Natl Acad Sci U S A
2016
;
113
:
14330
5
.
19.
Mermel
CH
,
Schumacher
SE
,
Hill
B
,
Meyerson
ML
,
Beroukhim
R
,
Getz
G
. 
GISTIC2.0 facilitates sensitive and confident localization of the targets of focal somatic copy-number alteration in human cancers
.
Genome Biol
2011
;
12
:
R41
.
20.
Yu
G
,
He
QY
. 
ReactomePA: an R/Bioconductor package for reactome pathway analysis and visualization
.
Mol Biosyst
2016
;
12
:
477
9
.
21.
Marchini
J
,
Shiraishi
Y
,
Tremmel
G
,
Miyano
S
,
Stephens
M
. 
A simple model-based approach to inferring and visualizing cancer mutation signatures
.
PLoS Genet
2015
;
11
:
e1005657
.
22.
Niknafs
N
,
Kim
D
,
Kim
R
,
Diekhans
M
,
Ryan
M
,
Stenson
PD
, et al
MuPIT interactive: webserver for mapping variant positions to annotated, interactive 3D structures
.
Hum Genet
2013
;
132
:
1235
43
.
23.
Hofmann
M
,
Stoss
O
,
Shi
D
,
Buttner
R
,
van de Vijver
M
,
Kim
W
, et al
Assessment of a HER2 scoring system for gastric cancer: results from a validation study
.
Histopathology
2008
;
52
:
797
805
.
24.
David
O
,
Jett
J
,
LeBeau
H
,
Dy
G
,
Hughes
J
,
Friedman
M
, et al
Phospho-Akt overexpression in non–small cell lung cancer confers significant stage-independent survival disadvantage
.
Clin Cancer Res
2004
;
10
:
6865
.
25.
Alexandrov
LB
,
Nik-Zainal
S
,
Wedge
DC
,
Aparicio
SA
,
Behjati
S
,
Biankin
AV
, et al
Signatures of mutational processes in human cancer
.
Nature
2013
;
500
:
415
21
.
26.
Ramdzan
ZM
,
Nepveu
A
. 
CUX1, a haploinsufficient tumour suppressor gene overexpressed in advanced cancers
.
Nat Rev Cancer
2014
;
14
:
673
82
.
27.
Andre
F
,
Ciruelos
E
,
Rubovszky
G
,
Campone
M
,
Loibl
S
,
Rugo
HS
, et al
Alpelisib for PIK3CA-mutated, hormone receptor-positive advanced breast cancer
.
N Engl J Med
2019
;
380
:
1929
40
.
28.
Burke
JE
,
Perisic
O
,
Masson
GR
,
Vadas
O
,
Williams
RL
. 
Oncogenic mutations mimic and enhance dynamic events in the natural activation of phosphoinositide 3-kinase p110alpha (PIK3CA)
.
Proc Natl Acad Sci U S A
2012
;
109
:
15259
64
.
29.
Dogruluk
T
,
Tsang
YH
,
Espitia
M
,
Chen
F
,
Chen
T
,
Chong
Z
, et al
Identification of variant-specific functions of PIK3CA by rapid phenotyping of rare mutations
.
Cancer Res
2015
;
75
:
5341
54
.
30.
Gymnopoulos
M
,
Elsliger
MA
,
Vogt
PK
. 
Rare cancer-specific mutations in PIK3CA show gain of function
.
Proc Natl Acad Sci U S A
2007
;
104
:
5569
74
.
31.
Lee
J-H
,
Zhao
X-M
,
Yoon
I
,
Lee
JY
,
Kwon
NH
,
Wang
Y-Y
, et al
Integrative analysis of mutational and transcriptional profiles reveals driver mutations of metastatic breast cancers
.
Cell Discovery
2016
;
2
:
16025
.
32.
Chang
MT
,
Asthana
S
,
Gao
SP
,
Lee
BH
,
Chapman
JS
,
Kandoth
C
, et al
Identifying recurrent mutations in cancer reveals widespread lineage diversity and mutational specificity
.
Nat Biotechnol
2016
;
34
:
155
63
.
33.
Labade
AS
,
Karmodiya
K
,
Sengupta
K
. 
HOXA repression is mediated by nucleoporin Nup93 assisted by its interactors Nup188 and Nup205
.
Epigenetics Chromatin
2016
;
9
:
54
.
34.
Bersini
S
,
Lytle
NK
,
Schulte
R
,
Huang
L
,
Wahl
GM
,
Hetzer
MW
. 
Nup93 regulates breast tumor growth by modulating cell proliferation and actin cytoskeleton remodeling
.
Life Sci Alliance
2020
;
3
:
e201900623
.
35.
Buisson
R
,
Langenbucher
A
,
Bowen
D
,
Kwan
EE
,
Benes
CH
,
Zou
L
, et al
Passenger hotspot mutations in cancer driven by APOBEC3A and mesoscale genomic features
.
Science
2019
;
364
:
eaaw2872
.
36.
Newell
F
,
Wilmott
JS
,
Johansson
PA
,
Nones
K
,
Addala
V
,
Mukhopadhyay
P
, et al
Whole-genome sequencing of acral melanoma reveals genomic complexity and diversity
.
Nat Commun
2020
;
11
:
5259
.
37.
Weiss
J
,
Sos
ML
,
Seidel
D
,
Peifer
M
,
Zander
T
,
Heuckmann
JM
, et al
Frequent and focal FGFR1 amplification associates with therapeutically tractable FGFR1 dependency in squamous cell lung cancer
.
Sci Transl Med
2010
;
2
:
62ra93
.
38.
Courjal
F
,
Cuny
M
,
SimonyLafontaine
J
,
Louason
G
,
Speiser
P
,
Zeillinger
R
, et al
Mapping of DNA amplifications at 15 chromosomal localizations in 1875 breast tumors: definition of phenotypic groups
.
Cancer Res
1997
;
57
:
4360
7
.
39.
Greulich
H
,
Kaplan
B
,
Mertins
P
,
Chen
TH
,
Tanaka
KE
,
Yun
CH
, et al
Functional analysis of receptor tyrosine kinase mutations in lung cancer identifies oncogenic extracellular domain mutations of ERBB2
.
Proc Natl Acad Sci U S A
2012
;
109
:
14476
81
.
40.
Kancha
RK
,
von Bubnoff
N
,
Bartosch
N
,
Peschel
C
,
Engh
RA
,
Duyster
J
. 
Differential sensitivity of ERBB2 kinase domain mutations towards lapatinib
.
PLoS One
2011
;
6
:
e26760
.
41.
Trowe
T
,
Boukouvala
S
,
Calkins
K
,
Cutler
RE
 Jr
,
Fong
R
,
Funke
R
, et al
EXEL-7647 inhibits mutant forms of ErbB2 associated with lapatinib resistance and neoplastic transformation
.
Clin Cancer Res
2008
;
14
:
2465
75
.
42.
Bonilla
X
,
Parmentier
L
,
King
B
,
Bezrukov
F
,
Kaya
G
,
Zoete
V
, et al
Genomic analysis identifies new drivers and progression pathways in skin basal cell carcinoma
.
Nat Genet
2016
;
48
:
398
406
.
43.
Hayward
NK
,
Wilmott
JS
,
Waddell
N
,
Johansson
PA
,
Field
MA
,
Nones
K
, et al
Whole-genome landscapes of major melanoma subtypes
.
Nature
2017
;
545
:
175
80
.
44.
Inman
GJ
,
Wang
J
,
Nagano
A
,
Alexandrov
LB
,
Purdie
KJ
,
Taylor
RG
, et al
The genomic landscape of cutaneous SCC reveals drivers and a novel azathioprine associated mutational signature
.
Nat Commun
2018
;
9
:
3667
.
45.
Takahagi
S
,
Noda
H
,
Kamegashira
A
,
Madokoro
N
,
Hori
I
,
Shindo
H
, et al
Metastatic extramammary Paget's disease treated with paclitaxel and trastuzumab combination chemotherapy
.
J Dermatol
2009
;
36
:
457
61
.
46.
Verret
B
,
Cortes
J
,
Bachelot
T
,
Andre
F
,
Arnedos
M
. 
Efficacy of PI3K inhibitors in advanced breast cancer
.
Ann Oncol
2019
;
3
0:x12–x20
.
47.
Dong
X
,
Zhang
L
,
Milholland
B
,
Lee
M
,
Maslov
AY
,
Wang
T
, et al
Accurate identification of single-nucleotide variants in whole-genome-amplified single cells
.
Nat Methods
2017
;
14
:
491
3
.
48.
Tan
DSW
,
Mok
TSK
,
Rebbeck
TR
. 
Cancer genomics: diversity and disparity across ethnicity and geography
.
J Clin Oncol
2016
;
34
:
91
101
.