NRAS and BRAF Mutations in Melanoma-Associated Nevi and Uninvolved Nevi

According to the prevailing multistep model of melanoma development, oncogenic BRAF or NRAS mutations are crucial initial events in melanoma development. It is not known whether melanocytic nevi that are found in association with a melanoma are more likely to carry BRAF or NRAS mutations than uninvolved nevi. By laser microdissection we were able to selectively dissect and genotype cells either from the nevus or from the melanoma part of 46 melanomas that developed in association with a nevus. In 25 cases we also genotyped a control nevus of the same patients. Available tissue was also immunostained using the BRAFV600E-mutation specific antibody VE1. The BRAFV600E mutation was found in 63.0% of melanomas, 65.2% of associated nevi and 50.0% of control nevi. No significant differences in the distribution of BRAF or NRAS mutations could be found between melanoma and associated nevi or between melanoma associated nevi and control nevi. In concordant cases immunohistochemistry showed a higher expression (intensity of immunohistochemistry) of the mutated BRAFV600E-protein in melanomas compared to their associated nevi. In this series the presence of a BRAF- or NRAS mutation in a nevus was not associated with the risk of malignant transformation. Our findings do not support the current traditional model of stepwise tumor progression.


Introduction
At present, the initial genetic changes in the development of cutaneous melanoma are unclear. Our understanding of the genetic basis of melanoma development and progression is based primarily on the classic multi-step model predicting that the acquisition of oncogenic mutations is a founder event in melanocytic neoplasia. The Clark model of melanoma progression is based on the concept of a sequential accumulation of mutations that is mirrored morphologically by the transformation of a benign melanocytic nevus to a dysplastic nevus and finally to a melanoma [1][2][3][4][5]. At a molecular level it is believed that activation of the mitogenactivated protein kinase (MAPK) signaling pathway as a result of somatic mutations of NRAS or BRAF is a crucial event in this multistep development of melanoma [6][7][8]. These mutations, which occur mutually exclusive [9,10], cause constitutive activation of the serine-threonine kinases in the ERK-MAPK pathway. The role of BRAF-mutations is underlined by advances in the treatment of melanoma with BRAF inhibitors [11][12][13] but the exact role of BRAF in the initiation or progression of melanoma is still unknown. There are conflicting results with regard to the role of BRAF and NRAS mutations in melanomas in their horizontal and vertical growth phase [14][15][16][17][18]. It is also known that BRAF mutations occur at a similar frequency in nevi and in primary and metastatic melanomas [9,[19][20][21][22]. It has been proposed that activating BRAF mutations induce senescence/apoptosis by up-regulating the tumor suppressor IGFBP7, which acts through autocrine/ paracrine pathways to inhibit BRAF-MEK-ERK signaling. Wajapeyee and coworkers suggest that loss of IGFBP7 expression acts as a critical step in melanoma genesis [23]. Decarlo et al. on the other hand found a disparate expression of IGFBP7 in BRAF V600E -positive dysplastic nevi (enhanced in 56% and diminished/absent in 44%) indicating that the behavior of oncogenic BRAF in dysplastic nevi, unlike that in malignant melanoma, does not appear to consistently induce senescence/apoptosis through pathways mediated by IGFBP7 [24][25][26]. Most nevi, including so called "dysplastic nevi", cease proliferation and remain static for decades. If nevi are indeed precursor lesions of melanoma they must acquire genetic alterations to free themselves of growth restraints and become malignant. The fact that oncogenic BRAF mutations are frequent in "dysplastic nevi", congenital nevi, common nevi and especially in growing nevi [20,27,28] has challenged the role of BRAF mutations for the development of melanoma [16] in particular and the model of stepwise tumor progression in general. According to this model the "dysplastic" or "atypical nevus" is the "missing link" between a benign and a malignant melanocytic lesion [4] and should be typified by genetic alterations that differ from "common nevi" and from melanoma. The major inherent problem is the lack of interobserver agreement for the morphology-based diagnosis of "dysplastic nevi" including clinical, dermatoscopic and histopathologic diagnosis [29]. Until now, BRAF or NRAS mutations have been investigated systematically only in nevi that were not associated with melanomas [7,10,20,22,30,31]. Therefore it is unknown if BRAF or NRAS mutations play any role in the progression of a nevus into a melanoma. For a better understanding it is mandatory to study a subset of melanomas that arose in association with a preexisting nevus [32,33]. To gain a more profound understanding in the role of BRAF or NRAS mutations in the development of melanoma from nevi we compared the genotype and BRAF V600E protein expression of melanomas and their associated nevi with control nevi of the same patient.

Immunohistochemical detection of BRAF V600E
Twenty-eight melanomas and 15 control nevi were stained with the BRAF V600E -mutation specific antibody VE1. Immunohistochemistry detected all mutations found by Sanger sequencing (Figure 1 D-G). In 6 (5.1%) of 117 tumor-samples (5 melanomas and 1 associated nevus) immunohistochemistry showed a BRAF V600E mutation not detectable by sequencing. These cases remained discordant after repeated microdissection, amplification and sequencing. While no associated nevus showed strong staining, 10.0% (n=2) of melanomas were strongly positive by immunohistochemistry. The majority of associated nevi (80.0%, n=16) stained weakly as did only 25.0% (n=5) of melanomas. Twenty percent (n=4) of associated nevi and 65.0% (n=13) of melanomas showed an intermediate staining intensity. Paired analysis revealed that melanomas showed a stronger immunohistochemical staining intensity than their associated nevi (Wilcoxon signed-rank test, p=0.002) (

BRAF and NRAS mutations in control nevi, melanomas and associated nevi
The frequency of oncogenic mutations did not differ significantly between melanoma-associated nevi and controlnevi or between melanomas and their associated nevi ( Figures  2 and S4). However, in four pairs a BRAF-wildtype nevus was present next to a BRAF V600E -mutant melanoma; in five cases a BRAF-wildtype melanoma had a preexisting BRAF V600Emutated nevus. BRAF V600E -mutation status was associated with younger age (55.3 years versus 48.2 years, p=0.007). In the present series we found no association between BRAF mutations and gender, invasion thickness or anatomic site. NRAS Q61 mutations were more frequently found in lesions occurring on the lower (18.2%, n=2) and upper extremities (27.8%, n=5) and less common on the trunk (7.8%, n=6). NRAS mutations were not distributed differently between melanomas, associated nevi and control nevi. Mutation status of control nevi and melanoma-associated nevi was concordant in 68.4% (p=0.99), melanomas and preexisting nevi were concordant in 90.9% (p=0.63) (Figure 2). NRAS Q61 -mutation status was not associated with a age, gender or invasion thickness.

BRAF and NRAS mutations of matched melanomas
For 28 cases (26 for NRAS) a melanoma without an associated nevus, could be matched for patient age ± 3 years, gender and anatomic site. Matched melanomas harbored a BRAF V600E mutation in 42.9% (n=12) of cases and NRAS Q61 mutations in 7.7% (n=2). The frequency of V600E or Q61 mutations was not different compared to melanomas with associated nevi (p=0.45 and p=0.63 respectively).
Because nevus type (junctional component present or not) and histomorphologic criteria were strongly correlated we built two different multivariate models to avoid collinearity. One model included nevus type the other histomorphologic criteria. If nevus type was excluded, location on the trunk (OR 6.68, SD:1.19-37.65, p=0.031) was predictive of melanoma

Discussion
The major finding of our study is that in contrast to current belief, neither morphology nor the presence of BRAF or NRAS mutation predicts the chance of malignant transformation of a nevus into a melanoma. The frequencies of oncogenic BRAF mutations were similar in the melanoma part (63.0%) and in the nevus part (65.2%) with a concordance of 80.4%. BRAF or NRAS mutated nevi did not have a higher chance to be associated with melanomas than wild type nevi. In other words, the presence of a BRAF V600 -or NRAS Q61 -mutation within a nevus does not change its risk of transforming into a melanoma. Our results do not support the concept that oncogenic BRAF or NRAS mutations play a major role in the development of melanoma from nevi and do not support the multistep theory of melanoma progression in its current form. Even if the multistep theory is true it is unlikely that the multiple steps of tumor progression correspond to a morphologic spectrum that spans from a benign nevus via a dysplastic nevus to melanoma in situ. It seems that if melanomas develop in a preexisting nevus it is most likely an inconspicuous nevus that shows even less histomorphologic "dysplasia" or "atypia" than a control nevus of the same patient. From a histomorphologic point of view control nevi were even more "atypical" than those nevi associated with melanomas. However, it has to be stated that most control nevi were excised for diagnostic reasons, which would explain why they showed a higher frequency of atypical features histomorphologically. This limitation however does not attenuate the fact that the genotypes of melanoma associated nevi and control nevi were the same. It was discovered recently, that melanomas show heterogeneity in regard to BRAF-genotype [38]. It is therefore possible that the melanoma arose in melanocytes that were not sampled. Our method of genotyping -namely conventional Sanger sequencing -has been shown to have a low sensitivity in detecting BRAF V600E mutations [39]. For this reason we also used immunohistochemistry, a method that has an excellent accuracy in the detection of oncogenic BRAF V600E mutations and has been shown to be particularly advantageous in specimens containing only small tumor parts [40]. Addition of immunohistochemistry data did not change the significance of our results and it is therefore unlikely that our results can be explained by sampling error. Another reason for discordant melanoma-nevus pairs might be that the melanoma might have had its' origin not in the nevus but developed from melanocytes within the epidermis having a different genotype than the nevus. Given the high concordance of BRAF and NRAF mutations in melanoma and their associated nevi we consider this possibility highly unlikely. Finally we cannot rule out completely that epidermal parts of the associated nevus have been overgrown by the melanoma and therefore have not been accessible for analysis.
Evaluating tumors with BRAF V600E -genotype in benign and malignant parts, the melanoma showed a more intense staining with the VE1-antibody than the associated nevus (Figure 1 C,D,F and Figures S2 & S3 H,I). This higher expression of the oncogenic protein might be due to additional genetic or epigenetic events that play a role in oncogene-induced senescence or upregulate the RAS-RAF-pathway. It is interesting that most melanomas arose in the epidermis although most associated nevi were compound. This may be explained by the fact that driver mutations that are detected in addition to BRAF-mutants show a signature of UV-mutagenesis [41].
In our series, melanomas associated with nevi showed a similar frequency of BRAF V600 -and NRAS Q61 -mutations compared to published reports of melanomas of the skin in general [7]. Higher mutation-rates in melanomas of younger individuals are in line with recent findings [41] that V600E mutations might not be related to chronic sun-exposure. The frequent occurrence of V600E mutations in other neoplasms such as thyroid carcinomas [42], ovarian cancer or adenocarcinomas [43] make it unlikely that actinic damage is a major cause of this mutation. A reason for more frequent BRAF V600E -positive melanomas in younger patients might be that in older patients a higher percentage of melanocytic lesions harboring the mutation already have undergone oncogene driven senescence [44]. When using a more sensitive method (immunohistochemistry) to detect the BRAF V600E phenotype the age difference to wildtype-tumors, especially in nevi, became smaller -an effect already shown before by Zalaudek et al. [28]. An explanation might be that, of coexisting subclones in melanocytic tumors [38], those harboring BRAF V600E are more likely to have undergone senescence or regression [44,45] in older patients and are therefore less likely to be detected by methods with lower sensitivity (e.g. Sanger sequencing).
A limitation of our study is that it is difficult to differentiate "severe junctional dysplasia" from melanoma in situ and that this generates a bias towards selection of melanomas that are associated with banal or "moderately dysplastic" nevi. However, according to our experience, the term "severely dysplastic nevi" expresses diagnostic uncertainty and not biologic uncertainty. If the concept of "dysplasia" is based on the impossibility to differentiate "severe junctional dysplasia" of a nevus from "melanoma in situ" it immunizes itself against falsification, making it a problematic concept from a scientific point of view.
In sum we show that the presence of a BRAF-or NRAS mutation, at least in the present series, is not biologically relevant in the development of melanomas that arise in association with a nevus. Other, currently unidentified, genetic or epigenetic changes may play a role in the transformation of nevi to melanomas.

Ethics statement
The study was approved by the local ethics committee of the Medical University of Vienna, Austria. The institutional ethics committee waived the need for written informed consent from those participants from whom only archived biopsy specimens were used in a retrospective manner (Ethics committee protocol 598/2010). Their private information and specimens were coded and not collected through an interaction or intervention with individuals. All other participants gave their written informed consent to participate in the study.

Histologic assessment of melanoma-associated nevi and control nevi
Criteria for the presence of a melanoma-associated nevus were used as published before by Stolz et al. [46]. All nevi were rated for the presence of different postulated morphologic features of "dysplastic" nevi [34] with the following definitions: Bridging: Nests of melanocytes within the epidermis spanning two or more rete ridges. Lentiginous or epitheloid cell proliferation: Melanocytes, arranged as variably sized nests and single cells within the epidermis. Cytologic atypia: Presence of enlarged nuclei, hyperchromatic nuclei and/or prominent eosinophilic nucleoli. Fibroplasia: Increased number of fibroblasts and collagen fibres arranged in a linear fashion around rete ridges beneath the epidermal part of the neoplasm. Fibroplasia around dermal nests of melanocytes was disregarded.

Laser microdissection
For laser-capture-microdissection one or multiple 5µm-thick sections of the specimen were cut onto membrane slides and stained with Haematoxylin and Eosin. The respective tumor areas were cut using the MMI CellCut Plus -Laser Capture Microdissection Microscope (Molecular Machines & Industries, Switzerland). Figure S1.

Immunohistochemical detection of BRAF V600E mutation
Available tumor tissue was stained using the BRAF V600Emutation specific antibody VE1 as published before [47]. VE1 antibody was kindly provided by Prof. Andreas von Deimling (University of Heidelberg). Every slide was evaluated by two observers (PT, HK), blinded to the sequencing result, in correspondence with a standard H&E-stained slide of the same tumor. Positive staining intensity of tumor tissue was rated semiquantitatively ("weak", "intermediate" and "strong", Figure  S5). Conventional Sanger sequencing has been described to be inferior to BRAF V600E -specific immunohistochemistry by VE1 [47] in terms of sensitivity and specificity, therefore, in discordant cases mutation status as detected by immunohistochemistry was used for further analyses (BRAF V600E mutation was detected by IHC in five melanomas and one associated nevus not previously found by Sangersequencing).

Statistical analysis
To compare groups the McNemar-test (paired data), Chisquare (unpaired categorical/nominal data), Fisher's-exact (smaller sample sizes), Wilcoxon signed-rank (paired continuous/nominal data) and T-test (unpaired continuous data; two-tailed p-value) was used. Multivariate comparison of characteristics between melanoma associated nevi and control nevi was performed using a forward conditional logistic regression model. As the presence of a junctional component was highly codependent (data not shown) to all measured histopathologic characteristics, two models were calculated: One excluding a junctional component as a predictor variable, one excluding all histomorphologic features.
All given P-values are two-tailed and a p-value <0.05 was considered significant. All statistical analyses were performed using IBM® SPSS® statistics Version 20.0. Figure S1.

Supporting Information
Tissue as imaged with the laser-capture microdissecting microscope before (A) and after (B) microdissection and the corresponding collected cuts (C). (TIF)