Background

Primary mucosal melanomas (MM), arising from melanocytes in mucosal tissues lining the head and neck, respiratory, gastrointestinal, and urogenital tracts, account for 26% of all melanomas in Asian populations [1]. Atypical symptoms are commonly observed in the early stage because of the occult anatomic locations, and the median age at presentation is 70 years, which is later than that for cutaneous melanomas [2]. Moreover, the clinical presentation of MM is more aggressive and has a poorer prognosis, regardless of the stage at diagnosis, with a 5-year survival rate of 25% owing to the lack of effective targeted therapies [3].

Importantly, MM is markedly different from cutaneous melanoma in terms of biological aspects. Recently, whole-genome and exome sequencing have revealed the molecular landscape and potential oncogenic drivers of MM, which are characterized by greater copy number (CN) variations, structural variations (SVs), and a lower tumor mutation burden (TMB) [4]. Targeted therapy with BRAF/MEK tyrosine kinase inhibitors improves patient survival [5]. However, BRAF mutations are less frequent in MM and are susceptible to resistance, resulting in a lower effectivity of BRAF/MEK inhibitors [6]. Immune checkpoint inhibitors (ICIs) are also beneficial for MM, with a median progression-free survival (PFS) of only 3.9 months [7]. Although BRAF, RAS, KIT, NF1, and SF3B1 have been identified as significantly mutated genes [4], there remains a lack of understanding and identification of oncogenic drivers in MM, probably owing to the rarity of samples and lack of preclinical models.

To further explore the molecular profiles of MM, we performed targeted next-generation sequencing (NGS) of 112 MM patients to explore the clinical significance of genetic characteristics potentially providing molecular targets and individualized treatments and provide a better mechanistic understanding of MM biology across the anatomic sites of MM.

Methods

Study participants

A total of 112 MM samples were collected from patients treated or consulted at the Sun Yat-sen University Cancer Center (SYSUCC, Guangzhou, China) between April 2007 and April 2022. Patients with cutaneous melanoma that had metastasized to the mucosa were excluded. The staging standard that we used followed the newest proposal for all anatomical sites published in April 2022 [8]. For all cases, the diagnoses were reviewed and confirmed by two experienced pathologists (Y. L. and F. W.). Clinical follow-up and treatment information as well as telephone interview data were available from the medical records of inpatients and outpatients (Supplementary Table S1). The study protocol was designed in accordance with the principles of the Declaration of Helsinki and approved by the Research Ethics Committee of the SYSUCC (No. B2016-069-01).

DNA extraction

Formalin-fixed and paraffin-embedded (FFPE) tissue blocks were assessed using hematoxylin-eosin (H&E) staining, and regions containing minimum of 20% tumor cells in unstained sections were selected for microdissection and subsequent experiments. Genomic DNA was extracted from tumors and patient-matched normal tissues or white blood cells using a QIAamp DNA FFPE Tissue Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions [9]. The extracted DNA was then quantified using a Qubit dsDNA BR assay (Life Technologies, USA) [10] according to the manufacturer’s instructions.

Targeted sequencing and data processing

As described previously [11], we used the two targeted sequencing assays: (1) the 295 OncoScreen panel containing whole exons of 287 genes and selected introns of 22 genes (Burning Rock Biotech Ltd., Guangzhou, China) and (2) the 1021-gene panel containing whole exons and selected introns of 288 genes and selected regions of 733 genes (Geneplus-Beijing, Beijing, China). Detailed methods for preparing the DNA, sequencing libraries, and data processing have been previously described [11].

Statistical analysis

TMB comparison among the different subsets of patients was performed using the Wilcoxon rank-sum test. The Kaplan–Meier method was used to estimate overall survival (OS) and PFS, and differences were compared using the log-rank test. Cox proportional hazards regression analyses were used to evaluate the independently predictive factors of each biological and clinical features associated with OS and/or PFS. Statistical significance was defined as a two-tailed P-value of < 0.05. All the statistical analyses were performed using IBM SPSS V.25.0 (Chicago, Armonk, NY, USA).

Results

Patients’ characteristics

In total, 112 patients with MM who underwent tumor sequencing were identified: 86 patients (76.8%) with primary tumors, 17 (15.2%) with metastatic tumors, and 9 (8.0%) with recurrent tumors. A total of 51 patients underwent 295-gene panel sequencing, and 61 underwent 1021-gene panel sequencing for further analysis (Supplementary Table S1). Table 1 summarizes the clinicopathological characteristics of the 112 patients with MM with a median age of 56 years (range 23–82 years); 56.3% were women and 43.8% were men. The anatomic sites of the MM were the head and neck in 33.9% of the patients, genitourinary tract in 19.6%, anorectum in 18.8%, esophagus in 17.0%, uvea in 8.9%, and small bowel in 1.8%. Histologically, 79 cases (70.5%) showed an epithelioid morphology, 8 (7.1%) showed a spindle morphology, and 21 (18.8%) showed a mixed morphology. Forty-six of the 112 (41.1%) patients with clinical stage were unknown due to the unavailability of invasion depth. The median PFS and follow-up times were 7.5 and 23.9 months, respectively. For treatment, 106 patients underwent surgical operations with a gross resection of primary or recurrent lesions (106/112, 94.6%), and 92 patients received various adjuvant treatments, including chemotherapy, radiotherapy, targeted therapy, and/or ICIs (92/112, 82.1%). Importantly, 67.0% (75/112) patients received ICIs and 32.1% (36/112) did not (Table 1).

Table 1 Clinicopathologic characteristics of 112 MM patients
Full size table

TMB in MM

Detailed information on the genetic variation in MM is provided in Supplementary Table S2. The TMB per megabase was relatively low, with a median of 3.1 (range 0–68.2). A significantly lower TMB was observed in primary tumor samples than in recurrent/metastatic tumors (Wilcoxon rank-sum test, P = 0.015, Fig. 1A). There were no significant differences among the primary sites, morphological types, and survival status (Fig. 1B-D).

Fig. 1
figure 1

Box plots showing the differences in TMB among different clinicopathological characteristics. Box plots (A-D) are shown for TMB in sample origins (P = 0.015), tumor site (B), morphology (C) and patients status (D). Significance shown as *** P < 0.001. Abbreviations: TMB, tumor mutational burden

Full size image

Genetic profiling and related pathways of MM

The frequent mutations were sequentially observed in BRAF (19/112, 17%), KIT (17/112, 15%), MYC (17/112, 15%), RAS (16/112, 14%), TP53 (15/112, 13%), NF1 (13/112, 12%), SF3B1 (12/112 11%), TERT (11/112, 10%), GNA11 (8/112, 7%), CBL (7/112, 6%), GNAQ (6/112, 5%) and FBXW7 (5/112, 4%) (Fig. 2). Furthermore, according to the Kyoto Encyclopedia of Genes and Genomes (KEGG) database, the most significantly altered pathway was the MAPK signaling pathway (73%), followed by the ErbB (58%), p53 (55%), Wnt (49%), mTOR (36%), and Notch (23%) signaling pathways (Fig. 2).

Fig. 2
figure 2

The mutational landscape of 112 patients with MM, including point mutation and SVs. From top to bottom: the total tumor mutational burden; the clinicopathological features, such as anatomic site, sex, and sample origins; the landscape of genes mutation (copy number variations, SNVs, indels, fusion gene, 5’UTR); the significantly activated pathway; the potential targeted therapy, each actionable mutation is colored by evidence level: A (NCCN guidelines and FDA guidelines), B (late trials), C (early trials), D (case report). Abbreviations: MM, mucosal melanoma; SVs, structural variants; SNVs, single nucleotide variations; UTR, untranslated regions; NCCN, national comprehensive cancer network; FDA, Food and Drug Administration

Full size image

The frequencies of these common mutations varied according to the origin of MM. Higher percentages of BRAF, RAS, NF1, and CBL mutations were observed in patients with melanoma in the head and neck than in patients with melanoma in the other five sites. Additionally, higher percentages of GNAQ and GNA11 mutations were observed in patients with uveal melanoma, and a higher percentage of SF3B1 mutations in patients with esophageal and uveal melanoma (Fig. 3, Supplementary Figure S1).

Fig. 3
figure 3

Recurring mutated genes (BRAF, KIT, RAS, SF3B1, GNA11, and GNAQ) of MM in different anatomic sites. Abbreviations: MM, mucosal melanoma

Full size image

Additionally, previous studies reported that other genes are frequently mutated in melanoma and other cancer types [4, 12,13,14], including BRCA2 (9/112, 8%), LRP1B (9/112, 8%), MET (8/112, 7%), PRKDC (7/112, 6%), NOTCH4 (7/112, 6%), CCND3 (6/112, 5%), and FGF3/4/19 (6/112, 5%). Germline mutations were observed in two cases (2/112, 2%) with BRCA2 (p.W1692Mfs*3) and ATM (c.331 + 5G > A).

Somatic copy number and structural events

Next, we observed that MM had a large number of chromosomal SVs, including somatic CN changes, fusions, and the telomere length-associated genes TERT and ATRX.

Somatic CN amplification was observed in 143 genes from 85 different chromosomal regions in 73 patients (Supplementary Figure S2). The frequently amplified regions included 6p21 (18/73, 25%), 8q24 (17/73, 23%), 4q12(14, 19%), 8q21 (11/73, 15%), 12q14 (11/73, 15%), 12q15 (10/73, 14%), 11q13 (9/73, 12%), 1q32 (8/73, 11%), and 7q31 (6/73, 8%), along with the recurrently amplified genes MYC (16/73, 22%), KIT (13/73, 28%), DAXX (11/73, 15%), KDR (11/72, 15%), NBN (11/72, 15%), CDK4 (11/72, 15%), MDM2 (10/73, 14%), PDGFRA (9/73, 12%), CCND1 (7/73, 10%), MDM4 (7/73, 10%), CCND3 (6/73, 8%), FGF19 (6/73, 8%), MET (6/73, 8%), FGF3 (5/73, 7%), FGF4 (5/73, 7%), FANCE (5/73, 7%), IRF4 (5/73, 7%), CDK6 (5/73, 7%), BRAF (5/73, 7%), NOTCH1 (4/73, 5%), and NOTCH2 (4/73, 5%). We also found several CN variations and the co-occurrence of KIT, KDR, PDGFRA, FGF3/4/19, FANCE, DAXX, and NOTCH clustered in the same chromosomal segment in one case, whereas CN deletions were detected in 14 cases. For example, CIC and PTEN deletions were more frequent in three and two cases, respectively.

Novel, recurrently fused genes were observed in nine cases (9/112, 8%), among which RNF43-IFLTD1, KCTD15-BRAF, and WDR64-AKT3 had a higher mutational abundance of 25%. The last two fusion genes encode tyrosine kinases, which may activate the MAPK pathway [15].

Finally, highly prevalent TERT mutations were identified in 11 cases (11/112, 10%), all with TERT promoter mutations containing c.146 C > T (4/11, 36%), c.124 C > T (4/11, 36%), c.-124 C > T (2/11, 18%), and c.-58-u68_-58-u66delCCCinsTCT (1/11,9%). Most ATRX variations had single nucleotide alterations or small fragment deletions. Loss of function was only detected in six cases (6/112, 5%). Interestingly, none of the cases had mutations in either TERT or ATRX, indicating they are mutually exclusive in MM, similar to those observed in gliomas [16].

Patients’ survival and clinical response to ICIs

As shown in Table 2; Fig. 4, multivariate and Kaplan–Meier analyses revealed that the anatomic sites of MM in the esophagus and small bowel (PFS: hazard ratio [HR] = 4.78, 95% confidence interval [CI] = 2.42–9.45, P < 0.001; OS: HR = 5.26, 95% CI = 2.51–11.03, P < 0.001) and CBL mutations (PFS: HR = 3.54, 95% CI = 1.46–8.56, P = 0.005; OS: HR = 5.57, 95% CI = 2.04–15.19, P = 0.001) were independently risk factors for survival. In addition, we found that mitotic activity with ≥ 10/mm2 (HR = 2.76, 95% CI = 1.22–6.25, P = 0.015) was an independent risk factor for OS. There were no associations between other factors and PFS or OS (Supplementary Figures S3 and S4).

Table 2 Univariate and multivariate Cox Regression analyses for PFS and OS in MM
Full size table
Fig. 4
figure 4

Kaplan–Meier survival analysis in MM. Forest plot and Kaplan–Meier survival curves for PFS (A-E) and OS (F-J) were performed for the primary tumor site, mitotic activity, CBL mutations, and with/without ICIs treatment. Significance shown as P < 0.05. Abbreviations: MM, mucosal melanoma; PFS, progression-free survival; OS, overall survival; CBL, Casitas B-lineage lymphoma; ICI, immune checkpoint inhibitor

Full size image

The suggested therapy modalities are listed in the bottom panel of Fig. 2, and mTOR, PD1, MEK, CDK4/6, KIT, PARP, VEGF, BRAF inhibitors may have pharmacological effects on targetable treatment [17,18,19,20,21]. In our study, we found that receiving ICIs (HR = 0.39, 95% CI = 0.20–0.75, P = 0.005, Table 2; Fig. 4J) was a favorable factor for OS (Table 2), suggesting that patients with MM may benefit from ICIs treatment.

Discussion

In this retrospective cohort study, we used two commercially customized NGS panels to describe the molecular spectrum of 112 MM patients and explore potential targets for prognosis prediction and further individualized immunotherapy in these patients.

Consistent with previous reports, which revealed that acral melanoma and MM had limited DNA mutational burden [14, 22,23,24], we observed a low TMB in this study. As previously reported [4, 14, 25,26,27], we found that a substantial proportion of MM had alterations in the genes involved in the MAPK pathway, including frequent mutations in BRAF, KIT, RAS, and NF1. It is well known that a BRAF mutation provides a targeted therapy for melanoma with good efficacy [28]. Nevertheless, small bowel melanoma in this study harbored only two in-frame mutations in BRAF (p.T599_V600insT and p.T244_L245delinsS), whereas no V600 mutations were observed. Mutations in other genes, such as KIT, RAS, and NF1, were found in the primary sites of the head and neck, esophagus, anorectum, and genitourinary tract in our cohort. Notably, we found that SF3B1 mutations as recurrent genetic events in MM were more common in esophageal and uveal melanomas, which is similar to that reported in recent studies [29,30,31]. In addition, mutations in GNAQ and GNA11 occurred in the majority of uveal melanomas found in this study, which is consistent with the findings of previous research [29, 32].

A previous whole-genome study showed that the SV load had the feature of complexity and diversity in MM [4]. Not surprisingly, we also found that most patients with MM harbored frequent CN variations and MYC amplification, which exemplified their role in MM pathogenesis and was in line with the results of other investigations [4, 33, 34]. For instance, high MYC expression is associated with aggressive biological behavior in cancers [35]. Additionally, many of the CN amplifications in well-known pathogenic genes (KIT, PDGFRA, MDM2, and MDM4) and cell cycle genes, such as CDK4/6 and CCND1, were found in this study, similar to those reported in previous studies [33, 34]. Bucheit et al. have demonstrated that complete PTEN loss correlates with poor survival in stage III melanoma [36]. In this study, we found PTEN deletion in MM, suggesting a tumorigenic role for progression in MM. The activation of TERT and inactivation of ATRX commonly and exclusively occurred in MM, which has been reported previously, indicating the importance of telomerase activation or alternative telomere lengthening mechanisms in reducing length in MM [4, 14].

A cohort of 466 Chinese patients with melanoma showed that MM was more aggressive and was associated with shorter survival than cutaneous lesions [37]. MM arises from different anatomic sites, and a population-based epidemiological analysis showed that MM from different anatomical sites exhibit different survival outcomes, likely due to the diverse environmental exposures associated with each site [38]. Here, we demonstrated that the anatomic sites of esophagus and small bowel melanomas may be a risk factor for survival, which concurs with the findings of a previous study [39]. However, a cohort of 706 MM patients did not show significant differences in prognosis among the anatomic sites of MM, primarily due to lack of cases of MM in the esophagus and small bowel [40].

We identified CBL mutations in patients with MM; these were risk factors for poor OS and PFS, which are infrequent in published melanoma exome studies but frequent in desmoplastic melanoma [41]. Recurrent CBL mutations occur in myeloid malignancies and have been associated with poor prognosis [42]. Ebert et al. recently identified that the oncogenic function of CBL mutants drive PI3K/AKT signaling and provide a rationale for therapeutic targets in myeloid malignancies [43]. Another study identified that inhibiting CBL mutations can activate the innate immune system to restrain cancer metastasis and improve the sensitivity to immunotherapy [44]. Finally, we found that MM patients receiving ICI treatment experienced a favorable impact on survival, especially OS, which is consistent with the findings of a previous study [45]. Taken together, CBL mutations might be promising targets for MM immunotherapy.

Our study had some limitations. First, the sample size was small owing to the rarity of this disease, especially small bowel melanoma. Second, for molecular profiling in MM, we used targeted NGS, which only targeted genomic regions of partial genes and, therefore, could not reveal novel pathogenic point mutations, rearrangements, or epigenetic changes. Thus, the whole-exome sequencing in a larger cohort is required to comprehensively depict the molecular profile of MM. A better understanding of the molecular mechanisms is required to explore additional avenues for immunotherapy in MM.

Conclusions

Our study demonstrates the molecular landscape of Chinese patients with MM based on targeted sequencing. Our finding on the genetic characteristics of MM among different anatomic sites reveal that CBL mutations in MM are potential targets. Further studies are warranted to elucidate the mechanisms that link CBL mutations to immunotherapy responses, which may provide a rationale for immunotherapy.