The deep penetrating nevus (DPN) is a variant of benign melanocytic nevus with clinical and histologic features mimicking vertical growth phase, nodular malignant melanoma (NMM). Because fatal misdiagnosis such as NMM occurs in 29% to 40% of the DPN, molecular differentiation markers are highly desirable. Beyond the clinical demand for precise diagnosis and diagnosis-adapted, preventive therapeutic strategies, the DPN represents a valuable natural model for melanocytic invasion without metastatic potential that per se deserves further investigations. In the present study, at first, we used a genome-wide, microarray-based approach to systematically prescreen for possible molecular markers differentially expressed between selected cases of typical DPN (n = 4) and metastatic NMM controls (n = 4). Gene expression profiling was done on Affymetrix Human X3P microarrays. Of the 47,000 genes spotted, we identified a list of 227 transcripts, which remained significantly regulated at a false discovery rate of 5%. Subsequently, we verified the expression of a subset of the most interesting transcripts in a larger immunohistochemical series (DPN, n = 17; NMM, n = 16). Of these transcripts, three were selected for immunohistochemical confirmation: tissue inhibitor of metalloproteinase-2, tumor protein D52, and ataxia telangiectasia-mutated gene (ATM). Additional criteria for selection from the list of 227 significantly regulated transcripts were grouping into functional Ingenuity networks and a known melanoma- or cancer-relevant function. Following these criteria, we detected a highly significant up-regulation of ATM transcription in NMM, which was also mirrored by ATM protein up-regulation. In contrast to the other markers, ATM particularly might serve as a suitable diagnostic and reliable discriminator of DPN/NMM because ATM immunoreactivity also showed a reliable staining consistency within all samples of both entities. (Cancer Epidemiol Biomarkers Prev 2007;16(11):2486–90)

In 1989, Seab et al. (1) reported a series of invasive, but nonmetastatic, pigmented melanocytic tumors and coined the term deep penetrating nevus (DPN). The DPN is a variant of benign melanocytic nevus with clinical and histologic features that may be alarming and can be mistaken for vertical growth phase, nodular malignant melanoma (NMM; refs. 1-4). In some cases, only histologic features together with clinical follow-up information can confirm that a lesion was truly benign. Due to its wedge-shaped growth into the dermis and subcutis (4) and further pseudomalignant features, such as absence of melanocytic maturation and lesional asymmetry (4, 5), misdiagnosis such as NMM is reported to occur in 29% to 40% of the cases. As a consequence, such patients would be excessively overtreated, causing a tremendous reduction of quality of life due to extensive follow-up examinations and possible side effects due to adjuvant therapy regimens (1, 4, 5). The opposite (i.e., taking advanced melanomas as deep penetrating nevi) may also occur, certainly with a more fatal consequence for the patient because a potentially curative therapy or life-prolonging follow-up would be withheld. To unravel this unsolved dilemma of daily pathologic routine, new predictive biomarkers for early differentiation of DPN and NMM are highly needed. Beyond the clinical demand for precise diagnosis and diagnosis-adapted, preventive therapeutic strategies, the DPN represents a valuable natural model for melanocytic invasion without metastatic potential that particularly deserves further investigations.

In a former study on DPN, Mehregan et al. (6) reported a subtle increase in the expression of proliferating cell nuclear antigen in NMM compared with DPN. However, this possible marker has never been considered for routine diagnostics, perhaps because it has not been confirmed in larger studies. Extending the quest for significant discriminators, in a recent approach, we focused on an arbitrary selection of common melanoma-associated candidate markers such as retinoblastoma protein, MIB-1/Ki-67, integrin β3, different types of matrix metalloproteinases, and dipeptidyl peptidase IV (CD26). By this, only dipeptidyl peptidase IV, a marker affecting both proliferation and invasion of malignant melanocytic tumors, was discriminative due to a consistent lack in NMMs (7).

In the present study, it is our primary objective to further improve the molecular diagnostics of DPN preferentially by increasing our understanding of its paradoxical biology. At first, we used a genome-wide, microarray-based approach to systematically prescreen for possible molecular markers differentially expressed between selected cases of typical DPN and metastatic NMM controls. Subsequently, we verified the differential expression of the preselected subset of the most interesting markers in a large immunohistochemical series to find a new immunohistochemical discriminator for daily practice. NMMs were chosen as controls because (a) they represent the major diagnostic problem and (b) we hoped to learn more about the mechanisms leading to the contrary metastatic potential of DPN and NMM. Because our former data on dipeptidyl peptidase IV suggest that markers discriminating between DPN and metastatic melanoma might also be directly involved in melanoma progression, our secondary objective was to link the newly found markers to already known melanoma pathways on the basis of published data.

Due to the diagnostic dilemma of DPN illustrated above, cryoconserved material is usually not available for research purposes. Therefore, paraffin blocks from 17 DPNs and 16 NMMs were collected at our department over a 10-year period. To ensure that the intended microarray analysis reveals valid data that reliably reflect differential gene expression profiles of true DPN and NMM, from all collected paraffin blocks, the five most typical DPNs (according to the clinical course and histopathology; ref. 1) and five matched cases (according to tumor thickness and the respective microdissected region, e.g., invasive part or tumor center) of typical NMM were preselected. To overcome the problem of formalin-induced RNA degeneration, the Paradise Reagent System (Arcturus), which has been particularly established for extraction and amplification of RNA from formalin-fixed paraffin-embedded material, was applied to the preselected five DPN and five NMM samples. RNA quality was checked by quantitative reverse transcription-PCR using two β-actin primer sets detecting two different mRNA fragments, one close to the 3′ poly(A) tail between bp 1,650 and 1,717 and one between bp 1,355 and 1,472. The ratio between both PCR products determines the degree of mRNA degradation. According to the manufacturer's recommendations, only samples with a ratio of (1,650-1,717)/(1,355-1,472) < 50 were considered for further evaluation. As a consequence, one DPN and one NMM were sorted out due to high RNA degradation. Amplification, probe labeling, and microarray hybridization of the remaining four DPN and four NMM were done on GeneChip Human X3P Arrays (Affymetrix) according to the manufacturer's recommendations. Normalization of the raw intensity data was carried out using the variance-stabilizing procedure of Huber et al. (8). Probes within probe sets were summarized by fitting the additive model of Irizarry et al. (9) using a median-polish procedure. Differentially expressed genes have been identified by a regularized, paired t test using the method implemented in the Limma software of Smyth (10). The false discovery rate was computed from the gene-wise P values according to Benjamini and Hochberg (11, 12). Of the 61,359 probe sets spotted on the microarrays representing ∼47,000 genes, we identified a list of 227 transcripts that were differentially expressed between both entities with a false discovery rate of 0.05 (193 up-regulated and 34 down-regulated transcripts in NMM; Supplementary Table).

To discover possibly hidden connections and pathways among the regulated genes that could have been overlooked by pure examination of the gene cluster and to reveal functional links to known melanoma-associated genes and pathways, a network analysis of the prefiltered genes was done as previously described (13) using the Ingenuity Pathways Analysis online application5

(Ingenuity Systems). By this, Ingenuity's Global Analysis Summary filter found functional groups on cellular growth and proliferation (32 genes), cell death (33 genes), cellular development (14 genes), and cell cycle (16 genes). Applying the Ingenuity Network filter, seven functional networks could be constituted with significant Ingenuity scores (13). Particularly, networks 1, 4, and 6 attracted our attention because they could be merged to one single network displaying links to many known melanoma-relevant signaling pathways such as Akt/phosphatase and tensin homologue, retinoblastoma/cyclin-dependent kinase, and tumor necrosis factor signaling (Fig. 1; ref. 14).

Figure 1.

Ingenuity Pathways Analysis of microarray data. Depicted are the results of GeneChip Human X3P arrays hybridized with probes from RNAs extracted from formalin-fixed paraffin-embedded DPN (n = 4) and advanced vertical growth malignant melanomas (NMM; n = 4). After stringent data filtering (paired t test plus Benjamini-Hochberg test), of 47,000 genes spotted, 227 transcripts remained significantly regulated with a false discovery rate q < 0.05 and P < 0.0002. Ingenuity network analysis discovers functional connections among regulated genes with known relevance to melanoma biology (e.g., TIMP-2 and ATM) or to other cancer types such as breast cancer (e.g., TPD52). Red symbols, up-regulation in NMM (positive t values); green symbols, down-regulation (negative t values); white symbols, no regulation detected. A, activation; B, binding; T, transcription; E, expression.

Figure 1.

Ingenuity Pathways Analysis of microarray data. Depicted are the results of GeneChip Human X3P arrays hybridized with probes from RNAs extracted from formalin-fixed paraffin-embedded DPN (n = 4) and advanced vertical growth malignant melanomas (NMM; n = 4). After stringent data filtering (paired t test plus Benjamini-Hochberg test), of 47,000 genes spotted, 227 transcripts remained significantly regulated with a false discovery rate q < 0.05 and P < 0.0002. Ingenuity network analysis discovers functional connections among regulated genes with known relevance to melanoma biology (e.g., TIMP-2 and ATM) or to other cancer types such as breast cancer (e.g., TPD52). Red symbols, up-regulation in NMM (positive t values); green symbols, down-regulation (negative t values); white symbols, no regulation detected. A, activation; B, binding; T, transcription; E, expression.

Close modal

Of the 227 regulated transcripts at a false discovery rate of 0.05, three markers were selected for further immunohistochemical confirmation: tissue inhibitor of metalloproteinase-2 (TIMP-2), tumor protein D52 (TPD52), and ataxia telangiectasia-mutated gene (ATM). Criteria for further marker selection were (a) grouping into a functional Ingenuity network and (b) a known melanoma-relevant or, at least, cancer-relevant function. Immunohistochemistry was done with 17 DPNs and 16 NMMs according to standard protocols. As primary antibodies, anti–TIMP-2 DM318P (Acris Antibodies), anti-TPD52 d1C5, and anti-ATM ATX08 ab2354 (Abcam) were used. To reduce bias, immunoreactivity was assessed by two independent investigators (A.R. and T.V.) in a blinded fashion using uniform criteria to maintain the reproducibility of the method. For each sample, three representative fields of vision were evaluated at ×400 magnification in a tumor region with maximum and a region with minimum staining signals as recently published (7, 15, 16) and as explained in Supplementary data. For assessment of relative protein expression levels, the nonparametric Mann-Whitney U test was applied (SPSS 13.0 software package, SPSS, Inc.).

The transcriptional up-regulation of TPD52 in NMMs was also mirrored on protein level with P = 0.04. TPD52 has been suggested to be involved in calcium-mediated signal transduction and cell proliferation in early breast cancer progression (17, 18). A relation of TPD52 to melanoma has never been reported thus far, but a possible influence on melanocytic invasion can be envisioned due to its considerable immunoreactivity in the deep dermal portions of both tumor entities analyzed here (Supplementary Fig. S1).

In accordance with former studies on matrix metalloproteinases in pigmented tumors (19), TIMP-2 staining was positive in all samples as well as in benign DPNs. Both entities showed a similar enhancement of TIMP-2 immunoreactivity at the invasive rims and subepidermal portions. Regarding the percentage of stained tumor cells per section, a trend toward a higher TIMP-2 expression in DPN reflects the microarray data. However, TIMP-2 seems to be less suitable for discrimination of DPN and NMM than the aforementioned TPD52.

Finally, in agreement with a recent study showing a progressive increase of ATM in melanocytic tumors (20), we detected within the list of 227 differently regulated genes an up-regulation of ATM transcription in NMM, which was also mirrored by a significant ATM protein up-regulation in NMM versus DPN (U test, P = 0.011). The ATM protein is a member of the phosphatidylinositol-3 kinase family of proteins that respond to UV-induced DNA damage by activating key substrates involved in DNA repair and/or cell cycle control (21, 22). In contrast to all other markers analyzed in this study, ATM particularly might serve as a suitable melanoma discriminator because ATM immunoreactivity also showed a reliable staining consistency within all samples of both entities (Fig. 2). Moreover, in a recent publication, a possible connection between abnormal ATM functioning and an altered dipeptidyl peptidase IV–mediated immune response was discussed (23), a finding that additionally links the current data to our previous observations on dipeptidyl peptidase IV in DPN (7).

Figure 2.

Comparative immunostaining of ATM. Immunoreactivity was semiquantitatively assessed with regard to staining quantity and subcellular staining intensity in 12 DPNs and 11 NMMs. Staining quantity and intensity were summarized to a single expression score (ES) as recently published (7, 15, 16) and as explained in Supplementary data. Due to the strong intratumoral staining heterogeneity observed in all DPN and NMM samples, ES was always recorded for a maximum staining region (invasive tumor portion) and a minimum staining region (tumor core) per sample. To get more information about the ATM expression in relation to the complete tumor, ESmax and ESmin were proportionally summarized as EStotal. Afterward, ESmax, ESmin, and EStotal of each sample were grouped into one of five ES categories (ES 0, ES 1-50, ES 51-100, ES 101-150, and ES >150) and displayed as percentage of all samples of one entity. By this, we found a statistically significant trend to higher scores in NMMs (U test, P < 0.05).

Figure 2.

Comparative immunostaining of ATM. Immunoreactivity was semiquantitatively assessed with regard to staining quantity and subcellular staining intensity in 12 DPNs and 11 NMMs. Staining quantity and intensity were summarized to a single expression score (ES) as recently published (7, 15, 16) and as explained in Supplementary data. Due to the strong intratumoral staining heterogeneity observed in all DPN and NMM samples, ES was always recorded for a maximum staining region (invasive tumor portion) and a minimum staining region (tumor core) per sample. To get more information about the ATM expression in relation to the complete tumor, ESmax and ESmin were proportionally summarized as EStotal. Afterward, ESmax, ESmin, and EStotal of each sample were grouped into one of five ES categories (ES 0, ES 1-50, ES 51-100, ES 101-150, and ES >150) and displayed as percentage of all samples of one entity. By this, we found a statistically significant trend to higher scores in NMMs (U test, P < 0.05).

Close modal

Taken together, this short report particularly suggests ATM as a diagnostic marker for a more objective discrimination of DPN and NMM with subsequent diagnosis-adapted therapy, and that ATM may direct further molecular studies in melanoma biology (e.g., targeting the ATM pathway). In addition, our study highlights DPN as a valuable natural model for melanocytic invasion without metastatic potential that deserves future investigations.

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.

Note: Supplementary data for this article are available at Cancer Epidemiology, Biomarkers & Prevention Online (http://cebp.aacrjournals.org/).

1
Seab JA, Jr., Graham JH, Helwig EB. Deep penetrating nevus.
Am J Surg Pathol
1989
;
13
:
39
–44.
2
Barnhill RL, Mihm MC, Jr., Magro CM. Plexiform spindle cell naevus: a distinctive variant of plexiform melanocytic naevus.
Histopathology
1991
;
18
:
243
–7.
3
Barnhill RL, Barnhill MA, Berwick M, et al. The histologic spectrum of pigmented spindle cell nevus: a review of 120 cases with emphasis on atypical variants.
Hum Pathol
1991
;
22
:
52
–8.
4
Mehregan DA, Mehregan AH. Deep penetrating nevus.
Arch Dermatol
1993
;
129
:
328
–31.
5
Robson A, Morley-Quante M, Hempel H, et al. Deep penetrating naevus: clinicopathological study of 31 cases with further delineation of histological features allowing distinction from other pigmented benign melanocytic lesions and melanoma.
Histopathology
2003
;
43
:
529
–37.
6
Mehregan DR, Mehregan DA, Mehregan AH. Proliferating cell nuclear antigen staining in deep-penetrating nevi.
J Am Acad Dermatol
1995
;
33
:
685
–7.
7
Roesch A, Wittschier S, Becker B, et al. Loss of dipeptidyl peptidase IV immunostaining discriminates malignant melanomas from deep penetrating nevi.
Mod Pathol
2006
;
19
:
1378
–85.
8
Huber W, Heydebreck A, Sueltmann H, Poustka A, Vingron M. Variance stabilization applied to microarray data calibration and to the quantification of differential expression.
Bioinformatics
2002
;
18
:
96
–104.
9
Irizarry RA, Bolstad BM, Collin F, Cope LM, Hobbs B, Speed TP. Summaries of Affymetrix GeneChip probe level data.
Nucleic Acids Res
2003
;
31
:
15
.
10
Smyth GK. Limma: linear models for microarray data. In: Bioinformatics and computational biology solutions using R and bioconductor. Gentleman R, Carey V, Dudoit S, Irizarry R, Huber W, editors. New York: Springer; 2005. p. 397–420.
11
Scheid S, Spang R. Twilight; a Bioconductor package for estimating the local false discovery rate.
Bioinformatics
2005
;
21
:
2921
–2.
12
Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing.
J Royal Stat Soc, Ser B
1995
;
57
:
289
–300.
13
Raponi M, Belly RT, Karp JE, et al. Microarray analysis reveals genetic pathways modulated by tipifarnib in acute myeloid leukemia.
BMC Cancer
2004
;
4
:
56
.
14
Miller AJ, Mihm MC, Jr. Melanoma.
N Engl J Med
2006
;
355
:
51
–65.
15
Roesch A, Becker B, Meyer S, et al. Retinoblastoma-binding protein 2-homolog 1: a retinoblastoma-binding protein down-regulated in malignant melanomas.
Mod Pathol
2005
;
18
:
1249
–57.
16
Roesch A, Becker B, Meyer S, et al. Overexpression and hyperphosphorylation of retinoblastoma protein in the progression of malignant melanoma.
Mod Pathol
2005
;
18
:
565
–72.
17
Byrne JA, Tomasetto C, Garnier JM, et al. A screening method to identify genes commonly overexpressed in carcinomas and the identification of a novel complementary DNA sequence.
Cancer Res
1995
;
55
:
2896
–903.
18
Byrne JA, Nourse CR, Basset P, et al. Identification of homo- and heteromeric interactions between members of the breast carcinoma-associated D52 protein family using the yeast two-hybrid system.
Oncogene
1998
;
16
:
873
–81.
19
Hofmann UB, Westphal JR, Van Kraats AA, et al. Expression of integrin α(v)β(3) correlates with activation of membrane-type matrix metalloproteinase-1 (MT1-MMP) and matrix metalloproteinase-2 (MMP-2) in human melanoma cells in vitro and in vivo.
Int J Cancer
2000
;
87
:
12
–9.
20
Smith AP, Weeraratna AT, Spears JR, et al. SAGE identification and fluorescence imaging analysis of genes and transcripts in melanomas and precursor lesions.
Cancer Biol Ther
2004
;
3
:
104
–9.
21
Kurz EU, Lees-Miller SP. DNA damage-induced activation of ATM and ATM-dependent signaling pathways.
DNA Repair (Amst)
2004
;
3
:
889
–900.
22
Lavin MF, Birrell G, Chen P, et al. ATM signaling and genomic stability in response to DNA damage.
Mutat Res
2005
;
569
:
123
–32.
23
Garcia-Perez MA, Allende LM, Corell A, et al. Novel mutations and defective protein kinase C activation of T-lymphocytes in ataxia telangiectasia.
Clin Exp Immunol
2001
;
123
:
472
–80.