Immunohistochemical and Molecular Features of Melanomas Exhibiting Intratumor and Intertumor Histomorphologic Heterogeneity

Melanoma is a heterogeneous neoplasm at the histomorphologic, immunophenotypic, and molecular levels. Melanoma with extreme histomorphologic heterogeneity can pose a diagnostic challenge in which the diagnosis may predominantly rely on its immunophenotypic profile. However, tumor survival and response to therapy are linked to tumor genetic heterogeneity rather than tumor morphology. Therefore, understating the molecular characteristics of such melanomas become indispensable. In this study, DNA was extracted from 11 morphologically distinct regions in eight formalin-fixed, paraffin-embedded melanomas. In each region, mutations in 50 cancer-related genes were tested using next-generation sequencing (NGS). A tumor was considered genetically heterogeneous if at least one non-overlapping mutation was identified either between the histologically distinct regions of the same tumor (intratumor heterogeneity) or among the histologically distinct regions of the paired primary and metastatic tumors within the same patient (intertumor heterogeneity). Our results revealed that genetic heterogeneity existed in all tumors as non-overlapping mutations were detected in every tested tumor (n = 5, 100%; intratumor: n = 2, 40%; intertumor: n = 3, 60%). Conversely, overlapping mutations were also detected in all the tested regions (n = 11, 100%). Melanomas exhibiting histomorphologic heterogeneity are often associated with genetic heterogeneity, which might contribute to tumor survival and poor response to therapy.


Introduction
Melanoma is one of the most heterogeneous neoplasms at both the histophenotypic and molecular levels. Melanomas often exhibit histophenotypic and molecular variations, both within the same tumor (intratumor) and between different tumors from the same patient (intertumor) [1]. The degree of histophenotypic variation can range from mild and focal changes to complete non-overlapping features of distinct histophenotypes, referred to as histophenotypic plasticity. This variation in histology and immunohistochemical expression of melanoma poses a diagnostic challenge that can alter disease stage and therapy.
Although both histophenotypic plasticity and tumor heterogeneity of melanoma are well recognized, their relationship is not well explored. Therefore, we interrogated the molecular characteristics of a series of cutaneous melanomas exhibiting extreme intertumor and intratumor histomorphologic heterogeneity using region-specific next-generation sequencing (NGS) in an attempt to unmask the underlying tumor heterogeneity and to search for cancer-associated driver mutations that may play a helpful diagnostic or therapeutic role and may predict disease outcome.

Clinicopathological Characteristics of Patients and Tumors Evaluated by NGS
The demographic, clinical characteristics, and disease outcome of the five patients are summarized in Table 1. Only one patient had a known history of therapy prior to sample collection (patient 4). Two patients (patients 1 and 2) had only a primary melanoma evaluated, and three patients (patients 3-5) had both a primary melanoma and a metastatic melanoma evaluated. All the primary melanomas were treated with wide local excision and sentinel lymphadenectomy. Complications included microsatellitosis (1.8 mm in largest dimension) identified in one patient (patient 3, 20%), regional nodal metastasis identified in three patients (patients 3-5, 60%), and distant metastases that developed during follow-up after initial surgical treatment in two patients (patients 1 and 4).
Histologically, all five primary melanomas exhibited a vertical growth phase that extended to the reticular dermis (Clark level IV; patients 2-4, 60%) and to the subcutaneous fat (Clark level V; patients 1 and 5, 40%), with a mean Breslow thickness of 5.14 mm (range: 1.2-12 mm) and mean mitotic rate of five mitoses per mm 2 (range: 2-14 mitoses per mm 2 ); however, no regression or brisk tumor-infiltrating lymphocytes was detected. Immunohistochemical studies revealed discordant patterns in expression of at least one melanocytic marker in four cases (patients 1-3 and 5, 80%), among whom a complete loss of at least one melanocytic marker was observed in two patients (patients 2 and 5, 50%). Alternatively, concordant pattern in expression of all the tested melanocytic markers was present in only one case (patient 4, 20%) ( Table 2).  In primary melanomas, the tumor of patient 1 had nevoid and spindled regions ( Figure 1, patient 1, A and B, respectively). S100 and SOX10 exhibited diffuse expression in both regions, whereas pan-melanocytic cocktail exhibited diffuse expression in the nevoid region and focal expression in the spindled region. HMB-45 antigen and Melan-A were positive only in the nevoid region ( Figure 1, patient 1, A1-A5 and B1-B5, respectively). NRAS, TP53, and KDR mutations were detected in both regions, whereas APC mutation was detected only in the nevoid region. The tumor of patient 2 had superficial epithelioid and deep spindled regions ( Figure 1, patient 2, A and B, respectively). S100 and SOX10 were expressed diffusely in both regions, whereas pan-melanocytic cocktail, HMB-45 antigen, and Melan-A were expressed only in the epithelioid region ( Figure 1, patient 2, A1-A5 and B1-B5, respectively). The same driver mutations of TP53 and CDKN2A were detected in both regions, whereas STK11 mutation was detected only in the epithelioid region.     In paired (primary and metastatic) melanomas, the primary tumor of patient 3 exhibited epithelioid morphology ( Figure 2, patient 3, A) and was diffusely positive for S100, SOX10, pan-melanocytic cocktail, HMB-45 antigen, and Melan-A ( Figure 2, patient 3, A1-A5), whereas the metastatic lesion exhibited rhabdoid morphology ( Figure 2, patient 3, B) with patchy expression of S100, SOX10, and pan-melanocytic cocktail, focal expression of HMB-45 antigen, and negative expression of Melan-A ( Figure 2, patient 3, B1-B5). Molecular analysis revealed identical mutations of BRAF, TP53, and ATM in the primary and metastatic tumors, whereas RAC1 mutation was identified only in the metastatic lesion.
In patient 4, the primary tumor exhibited epithelioid morphology, and the metastatic lesion exhibited foamy/balloon-like morphology ( Figure 2, patient 4, A and B, respectively). S100, SOX10, pan-melanocytic cocktail, and Melan-A were diffusely expressed in both lesions, in addition to the positive expression of HMB-45 antigen in the primary lesion ( Figure 2, patient 4, A1-A5 and B1-B3).
NGS revealed mutation in HNFA1 that was detected only in the primary lesion, and a mutation in APC was detected only in the metastatic lesion. In addition, the primary and metastatic lesions harbored identical KIT mutations.

Distribution of the Detected Driver Mutations
The detected mutations are summarized in Figure 3. In the 11 histologically distinct regions of eight tumors retrieved from five patients, mutations were detected in 12 cancer-associated genes. Mutations in five of these genes (42%), namely, APC, STK11, RAC1, HNFA1, and CTNNB1, were detected in only one of the histologically distinct regions within the same patient (non-overlapping). Mutations in APC and STK11 were each non-overlapping in two patients ( Table 2). Mutations in the other seven cancer-associated genes (58%) were present in each histologically distinct region within the same patient (overlapping). Of the detected genes, the most commonly mutated gene was TP53, the mutations of which were detected in each histologically distinct region in four patients. Quantification of allele frequency revealed that CTNNB1 (patient 5, epithelioid region of primary lesion) had the highest allele frequency (32.97%) among the non-overlapping genes, whereas CDKN2A (patient 2, primary tumor, epithelioid and spindled regions) had the highest allele frequency (83.09%) among the overlapping genes. In each case in which a BRAF, NRAS, or KIT mutation was identified, the mutation was present in all the histologically distinct regions. Alternatively, case 2 had no BRAF, NRAS, or KIT mutations.

Discussion
In this study, we analyzed mutation in 50 cancer-associated genes in melanomas exhibiting intertumor and intratumor histophenotypic variation. Our results showed that, in each patient, at least one non-overlapping mutation was detected in only one of the morphologically distinct regions, whereas one or more identical mutations were present in all of the morphologically distinct regions. Our findings indicate that histomorphologically heterogenic melanomas are often molecularly heterogenic, too. The latter may point toward a more aggressive disease and higher rates of treatment failure. Alternatively, the detection of overlapping mutations demonstrated in our current study may help pathologists accurately diagnose primary melanoma in cases with unusual histophenotypic presentations which may lead to a misdiagnosis of dedifferentiated or collision tumor, and may be useful in the diagnosis of histomorphologically heterogeneous metastases mimicking carcinoma or unclassified sarcoma [12]. Also, loss or gain of genetic alterations present only in metastatic lesions may reveal therapeutically actionable targets that can significantly impact the clinical management of patients with melanoma refractory to conventional therapy [13][14][15][16][17][18].
Melanoma can exhibit a variety of morphologic patterns, including rhabdoid, lipoblastic, cartilaginous, adenocarcinomatous, balloon cell, signet-ring cell, and small cell, which are often associated with several immunophenotypic alterations. However, their molecular characteristics are not well explored. The degree of the tumor molecular heterogeneity that is associated with its histologic heterogeneity is still unknown. Tumor genetic heterogeneity can occur at the level of the epigenome, genome, or transcriptome (proteome).
Although tumors originate from monoclonal cancer stem cells, as tumors evolve, tumor cells adopt new mutations and give rise to new subclones, which result in both spatial and temporal heterogeneity [19][20][21][22]. In melanoma, variable rates of mutations in cancer-related genes have been detected, ranging from 0.1 mutations per Mb to 100.0 mutations per Mb [23]. In our present study, we found that, in 80% of cases (n = 4) in which a BRAF, NRAS, or KIT mutation was identified, the mutation was present in all histologically distinct regions. These findings support the idea that the driver mutations are more conserved than passenger mutations in metastatic lesions. Most of the mutations other than BRAF, NRAS, and KIT mutations that we detected in metastatic lesions seemed to be associated with promotion of disease progression. These findings are similar to those of a prior study by Harbst et al. [24], who examined eight melanomas for genetic heterogeneity and found that 12% of the detected cancer-associated mutations were heterogeneously distributed, whereas BRAF and NRAS mutations, when detected, were homogeneously distributed among all the sampled regions.
In our current study, the quantification of mutant allele frequency (MAF) revealed that CDKN2A (patient 2, primary tumor, epithelioid and spindled regions) had the highest allele frequency (f = 83.09), suggesting loss of heterozygosity (LOH) of this gene. These findings are supportive of those of several recent studies, leading to the conclusion that MAF is a result of tumor heterogeneity in melanoma [25][26][27][28]. These studies examined the MAF of BRAF mutation in melanoma and revealed that the rates ranged from 0% to 97%.
Our study demonstrates that, in cases of melanoma metastasis, there may be at least one new mutation in the metastatic tumor in addition to the mutations detected in the primary melanoma. Our findings are similar to previous findings of tumor heterogeneity at the genetic level, which can alter the treatment response. Sakaizawa et al. [29] reported intertumor heterogeneity in BRAF mutation in circulating melanoma cells. Similarly, Lin et al. studied the MAF of BRAF mutation in three paired cases of primary and metastatic melanomas and found that two of the three cases had a higher MAF of BRAF mutation in the metastatic site than in the matched primary tumor [30]. Prior studies have also detected heterogeneity in members of signaling pathways in metastatic melanoma. Casula et al. [31] observed CDKN2A mutation in 11 of 16 metastatic tumors but not in paired primary melanomas whereas Goswami et al. [32] observed heterogeneity in KRAS and KIT mutations between paired cases of primary and metastatic melanomas.
In melanoma, tumor genetic heterogeneity remains one of the most common reasons for resistance to therapy [33], in which certain mutations may lead to resistance to targeted therapy. For instance, in our current study, Rac1 P29S mutation was identified in the metastatic melanoma of patient 3, which would carry a potential risk for resistance to Raf-therapy [34]. Similarly, variable response to immune checkpoint blockade was observed in cases with high tumor heterogeneity [35]. Additionally, high tumor heterogeneity may result in multiple tumor subpopulations that harbor heterogeneous mutations in cancer-related genes, which promote tumor survival and disease progression [36].
To detect tumor heterogeneity, although various methods can be used, NGS-based analysis is considered one of the most helpful tools for detecting mutations that explain tumor heterogeneity due to its high sensitivity that can detect mutations that occur in a small percentage of tumor cells [37]. In addition to being precise and accurate, NGS-based methods are relatively cost-effective compared to other methods, such as real-time PCR and Sanger sequencing [38]. Although our analysis was limited to the 50 genes interrogated by the NGS panel, a more extended panel or the use of whole-genome sequencing may result in lower rates of heterogeneity or false homogeneity.

Materials and Methods
Formalin-fixed, paraffin-embedded tissue samples of melanomas exhibiting intertumor or intratumor histologic heterogeneity were obtained from the archived materials of the Department of Pathology at the University of Texas MD Anderson Cancer Center over a two-year period. Sample collection and study protocols were approved by the MD Anderson Cancer Center Institutional Review Board. Intratumor histologic heterogeneity was defined as tumor with more than one distinct histologic pattern but maintaining well-formed boundaries, whereas intertumor histologic heterogeneity was defined as paired samples of primary and metastatic melanomas with completely distinct and non-overlapping mutually exclusive histologic patterns. A cutoff of 1.0 mm Breslow thinness and 1.0 mm in diameter of tumor deposit were used as criteria for sampling adequacy of primary and metastatic melanomas, respectively. To limit the sampling error using laser capture microdissection (LCM), melanomas exhibiting intermingled or overlapped histomorphologies in primary and metastatic lesions were excluded. For unequivocal histologic and phenotypic interpretation, tumors with heavy melanin pigmentation were also excluded from the study. After application of these selection and exclusion criteria, primary tumors from 2 patients and paired primary and metastatic tumors from 3 patients yielded 8 tumors and 11 histomorphologically distinct regions were identified. Whenever feasible, the immunohistochemical profile of the retrieved cases was also obtained to permit comparison of melanoma marker expression among the histologically distinct regions. Additional immunohistochemical staining for melanocytic markers, including S100 protein, SOX10, HMB-45 antigen, Melan-A, and pan-melanocytic cocktail, were performed in our Clinical Laboratory Improvement Amendments (CLIA)-certified laboratory on all the samples with sufficient tissue to complete the immunohistochemical whenever needed. Tissue blocks were sectioned at 5 µm thickness, placed on uncharged slides, and stained with hematoxylin and eosin. After analysis of the histomorphology, tumor tissue was laser microdissected for NGS analysis.

Laser Capture Microdissection and DNA Extraction
From each of the 11 histomorphologically distinct regions, an area in which the neoplastic melanocytes compromise at least 90% of the cells, was obtained by LCM. Approximately 4 × 10 µm 3 of tissue per sample was captured after adjustment of the laser amplitude, pulse duration, and number of hits. DNA was isolated by treating the tissue samples with 100 µL of 0.04% proteinase K at 42 • C overnight followed by heat-denaturing at 95 • C for 10 min. The genomic DNA extracted from each microdissected area was subjected to NGS-based analysis for the detection of somatic mutations in the coding sequence of 50 genes.
We determined the effective lower limit of detection of this assay (analytical sensitivity) for single nucleotide variations to be in the range of 5% (1 mutant allele in the background of 19 wild-type alleles) to 10% (1 mutant allele in the background of 9 wild-type alleles) by taking into consideration the depth of coverage at a given base and the ability to confirm low-level mutations using independent conventional platforms. The NGS-based study was targeted to detect mutations in cancer-related genes and genes related to signaling pathways that can affect tumor cell viability, promote disease progression, and predict response to therapy.

Conclusions
Our present study confirmed that melanomas exhibiting morphologic heterogeneity are often molecularly heterogeneous neoplasm. In each patient, one or more identical driver mutations were present in each of the morphologically distinct regions, and at least one genomic alteration in a gene not usually associated with melanoma was present in only one of the morphologically distinct regions. These findings support performing comprehensive molecular profiling beyond the major mutations of melanoma (BRAF, NRAS, and KIT) to detect rare cancer driver genes. Currently, NGS is the most reliable and cost-effective method for evaluating tumor heterogeneity. Genetic heterogeneity may reveal therapeutically actionable targets that can significantly impact the clinical management of patients with melanoma refractory to conventional therapy.
Author Contributions: H.A.M.: conceptualization, data curation and configuration, investigation, methodology, result analysis, writing-original draft preparation, and writing-reviewing, revising, and editing manuscript. S.K.C.A.: conceptualization and data curation. D.P.: conceptualization, data curation, writing and editing draft abstract, and poster presentation of preliminary findings. C.A.T.-C.: conceptualization, data curation, investigation of original material, and methodology. P.N.: conceptualization, data curation, investigation of original material, methodology, and writing-reviewing and editing manuscript. M.T.T.: conceptualization, data curation, investigation of original material, and methodology. J.L.C.: conceptualization, data curation, investigation of original material, and methodology. D.I.: conceptualization, data curation, investigation of original material, and methodology. D.Y.D.: data curation, software and scientific material analysis, and molecular diagnostic project administration. R.L.: conceptualization, data curation, investigation of original material, methodology, and molecular diagnostic project administration and analysis. V.G.P.: conceptualization, data curation, investigation of original material, and methodology. L.Y.B.: conceptualization, data curation, investigation of original material, methodology, molecular diagnostic project administration and analysis, and writing-reviewing and editing manuscript. P.P.A.: conceptualization, funding acquisition, archived material search and retrieval, investigation of original material, data collection and curation, methodology, project administration, supervision, mentoring, and writing-editing and revising manuscript.