ORIGINAL ARTICLE

Cell cycle analysis can differentiate thin melanomasfrom dysplastic nevi and reveals accelerated replicationin thick melanomas

Gergo Kiszner & Barnabas Wichmann & Istvan B. Nemeth &

Erika Varga & Nora Meggyeshazi & Ivett Teleki & Peter Balla &

Mate E. Maros & Karoly Penksza & Tibor Krenacs

Received: 12 December 2013 /Accepted: 11 March 2014# Springer-Verlag Berlin Heidelberg 2014

Abstract Cell replication integrates aberrations of cell cy-cle regulation and diverse upstream pathways which all cancontribute to melanoma development and progression. Inthis study, cell cycle regulatory proteins were detected insitu in benign and malignant melanocytic tumors to allowcorrelation of major cell cycle fractions (G1, S-G2, and G2-M) with melanoma evolution. Dysplastic nevi expressedearly cell cycle markers (cyclin D1 and cyclin-dependentkinase 2; Cdk2) significantly more (p<0.05) than commonnevi. Post-G1 phase markers such as cyclin A, geminin,topoisomerase IIα (peaking at S-G2) and aurora kinase B(peaking at G2-M) were expressed in thin (≤1 mm) mela-nomas but not in dysplastic nevi, suggesting that dysplasticmelanocytes engaged in the cell cycle do not completereplication and remain arrested in G1 phase. In malignantmelanomas, the expression of general and post-G1 phasemarkers correlated well with each other implying negligible

cell cycle arrest. Post-G1 phase markers and Ki67 but noneof the early markers cyclin D1, Cdk2 or minichromosomemaintenance protein 6 (Mcm6) were expressed significantlymore often in thick (>1 mm) than in thin melanomas.Marker expression did not differ between metastatic mela-nomas and thick melanomas, with the exception of aurorakinase A of which the expression was higher in metastaticmelanomas. Combined detection of cyclin A (post-G1phase) with Mcm6 (replication licensing) and Ki67 correct-ly classified thin melanomas and dysplastic nevi in 95.9 %of the original samples and in 93.2 % of cross-validatedgrouped cases at 89.5 % sensitivity and 92.6 % specificity.Therefore, cell cycle phase marker detection can indicatemalignancy in early melanocytic lesions and acceleratedcell cycle progression during vertical melanoma growth.

Keywords Cutaneous melanoma . Cell cycle phase analysis .

Mcm6 . Ki67 . Post-G1 phase markers

Introduction

Cutaneous malignant melanoma is the most fatal form of skincancers which shows increasing incidence in white-skinnedpopulations worldwide [1]. Early recognition of melanomas isessential, since even a mole-like thin melanoma may give riseto distant metastasis associated with poor prognosis [2, 3].However, the overlapping clinical and histological featuresand lack of reproducible biomarkers can obscure differentialdiagnosis between dysplastic nevi and thin melanomas indoubtful cases [4, 5].

Deregulation of cell growth (such as MAPK, PI3K/Akt,and Wnt), cell cycle control (loss of p16Ink4a), and apoptosis/pro-survival pathways (such as Apaf1) has been implicated inmelanoma development [6–8]. Benign nevi can already be

Electronic supplementary material The online version of this article(doi:10.1007/s00428-014-1570-1) contains supplementary material,which is available to authorized users.

G. Kiszner :N. Meggyeshazi : I. Teleki : P. Balla :M. E. Maros :T. Krenacs (*)1st Department of Pathology and Experimental Cancer Research andMTA-SE Tumor Progression Research Group, SemmelweisUniversity, Ulloi ut 26, Budapest 1085, Hungarye-mail: krenacst@gmail.com

B. Wichmann2nd Department of Internal Medicine, Semmelweis University,Szentkiralyi utca 46, Budapest 1088, Hungary

I. B. Nemeth : E. VargaDepartment of Dermatology and Allergology, University of Szeged,Koranyi fasor 6, Szeged 6720, Hungary

K. PenkszaInstitute of Botany and Ecophysiology, Department of Botany, SzentIstvan University, Pater Karoly utca 1, Godollo 2100, Hungary

Virchows ArchDOI 10.1007/s00428-014-1570-1

clonal and carry melanoma-predisposing genetic aberrationssuch as an activating BRAFV600E mutation [9]. The redundan-cy in melanoma-related cell growth pathways and the over-lapping genetic deviations further complicate the verificationof malignancy in critical lesions. At the same time, aberrantcell proliferation characterizes malignant phenotype, and thecell cycle regulation machinery integrates all upstream signal-ing pathways that may drive malignancy [8].

Differential expression of major proteins which regulatecell proliferation allows cell cycle phase testing (Fig. 1). DNAis licensed for replication by the binding of minichromosomemaintenance (Mcm) protein complex Mcm2–7 to replicationorigins at early G1 phase [8, 10]. Cell cycle progressionthrough G1-S and G2-M phase checkpoints is mediated bycomplexes of cyclins and cyclin-dependent serine-threoninekinases (Cdks), which are controlled by the cyclin-dependentkinase inhibitors of the CIP/KIP (e.g., p21Waf1 and p27Kip1) orthe INK family (e.g., p16Ink4a) [11]. Checkpoints validate thatall required events in a phase are completed and normal, and ifnot, repair mechanisms or apoptosis will take action [12].Topoisomerase IIα (Top2a) can cleave and ligate DNA duringreplication and contribute to the separation of daughter DNAstrands duringG2 phase [13]. The cell cycle repressor gemininprevents cells from reentering the cycle during S-G2-Mphases [14], while aurora kinases promote G2-M phase tran-sition by regulating mitotic spindle assembly for chromosom-al segregation [15]. Ki67, a general proliferation marker withunspecified role, is expressed throughout G1 to M phases[16, 17].

Aberrant and accelerated cell replication has been linked tomalignancy and aggressive tumor behavior [18] also in mel-anomas [19]. However, most studies have tested only a limitednumber of randomly selected cell cycle markers rather thanusing a comprehensive set of markers, which would allowestimation of major cell cycle fractions in relation tomelanocytic tumor progression [6, 20–23]. Recently, a FISH

method combining MYB (6q23), CCND1 (11q13), RREB1(6p25), and chromosome 6 centromere (CEP6) gene probes[24, 25] and the combined expression profiles of Bim, Brg1,Cul1, and Ing4 proteins [26] has been suggested for diagnos-ing malignancy in critical melanocytic lesions. However, thereproducibility of these methods and the reliability of theirsignal assessment are challenging issues. In contrast, immu-nohistochemical staining of cell cycle regulatory proteins inarchived tissue sections generates clear nuclear signals,allowing determination of the state of the cell cycle in dynam-ic cell populations, which can be used as a primary screeningtest [8].

In this study, we tested cell cycle kinetics in benign andmalignant melanocytic tumors for potential correlations be-tween cell cycle regulation, malignant phenotype, and tumorprogression. Differential expression of cyclin D1 (G1 phase),Cdk2 (G1-S phases), cyclin A, geminin, Top2a (S-G2 phases),and aurora kinases A and B (G2-M phases) served to deter-mine cell cycle fractions. These were correlated with theMcm6 and Ki67 protein positive cell fractions to identify cellpopulations arrested at any stage during cell replication.

Materials and methods

Melanocytic lesions and tissue microarrays

Formaldehyde-fixed, paraffin-embedded samples representingcases of 19 common nevi (10 compound, 8 intradermal, and 1halo), 63 dysplastic nevi (4 junctional, 46 compound, and 13lentiginous) including 36 low-grade and 27 high-grade lesions,63 primary melanomas (23 “thin” of ≤1 mm and 40 “thick”of >1 mm vertical thickness, in situ melanomas wereexcluded), and 22 melanoma metastases (6 lymph nodeand 16 cutaneous), diagnosed between 2003 and 2006and reclassified according to recent TNM criteria by the

Fig. 1 Simplified scheme of thecell cycle and differentialexpression of cell cycle regulatoryproteins tested in this study.Immunohistochemical detectionof these nuclear proteins can beused for assessing major cellcycle fractions such as G1 andpost-G1 (S-G2 and G2-M) phasecells

Virchows Arch

American Joint Committee on Cancer (2009) [27] in theDepartment of Dermatology and Allergology, Universityof Szeged, Hungary, were tested. In critical cases, thediagnosis was based on the consensus between at leasttwo expert dermatopathologists (I.B.N. and E.V.). Themean age of patients was 37.3 (11–64) years for com-mon nevi (6 males and 13 females), 34.3 (12–76) yearsfor dysplastic nevi (38 males and 25 females), 64.4(25–88) years for primary melanomas (31 males and32 females), and 66.5 (43–87) years for metastatic mel-anomas (7 males and 15 females). Clark levels sortedmelanomas into level I 0, II 18, III 29, IV 9, and V 7cases. For thick melanomas, the mean tumor thicknesswas 5.199 mm (1.064–26.524 mm), and the mean mi-totic index was 18.0 (0–71). For thin melanomas, themean thickness was 0.672 mm (0.304–0.988 mm), andthe mean mitotic index was 3.3 (0–11). Patient data werecoded and handled in accordance with the ethical regulationsof the institutional review boards at the Department ofDermatology and Allergology, University of Szeged, and atthe Semmelweis University, Budapest (approval number:KL-37/2006).

Tissue cores of 2-mm diameter were collected into six70-sample tissue microarray (TMA) blocks from repre-sentative areas based on hematoxylin and eosin (H&E)-stained slides using the computer-driven TMA Master(3DHISTECH Kft., Budapest, Hungary) [28, 29]. Smalllesions could be included as a single tissue core. Fromsizable melanomas, duplicate or triplicate cores were takenby systematically selecting samples from the vertical tumorfront, the mid-region, and the uppermost region closest to orincluding the epidermis. Case numbers listed above were usedfor final evaluation after leaving out damaged or nonrepresen-tative samples.

Immunohistochemistry

Four-micrometer-thick sections cut from TMA blocks weremounted on adhesive glass slides (SuperFrost Ultra Plus,Gerhard Menzel GmbH, Braunschweig, Germany) and boiledafter routine dewaxing for antigen retrieval either in a pH 6.0target retrieval buffer (Dako, Glostrup, Denmark) for cyclinD1 or in a pH 9.0 buffer of 0.01 M Tris–0.1 M EDTA for allother antibodies at ~105 °C for 30 min using an electricpressure cooker (Avair Ida, YDB50-90D, Biatlon Kft., Pecs,Hungary). Endogenous peroxidase activity was blocked in a0.5 % hydrogen peroxide methanol solution for 20 min. TheNovolink polymer kit (Leica Novocastra, Newcastle UponTyne, UK) was used for antigen detection as described before[30]. Briefly, the TMA slides were treated in a humiditychamber at room temperature using the protein block for10 min; the primary antibodies including monoclonal anti-mouse Cdk2 (1:300; clone 2B6), cyclin A (1:150; clone 6E6),

Top2a (1:400; clone Ki-S1; all from Thermo Lab Vision,Kalamazoo, MI, USA), geminin (1:150; clone EM6), Mcm6(1:600; clone KAT82; both from Leica Novocastra), Ki67(ready-to-use 1:2; clone MIB-1; Dako), monoclonal rabbitanti-cyclin D1 (1:200; clone SP4; Thermo Lab Vision), aurorakinase A (1:80; clone 1G4; Cell Signaling, Danvers, MA,USA), and aurora kinase B (1:300; clone EP1009Y; AbCamEpitomics, Burlingame, CA, USA) diluted in 1 % bovineserum albumin (BSA) in pH 7.4 Tris-buffered saline (TBS)buffer overnight; the post-primary reagent for 30 min; andfinally with the polymer-peroxidase complex for 30 min. Theslides were washed between all incubation steps for 2×3 minusing TBS buffer containing 0.01 % Tween 20. Enzymeactivity was visualized using a hydrogen peroxide/3-amino-9-ethylcarbazole (AEC) solution at pH 4.5 under microscopiccontrol for 3–5 min, and the slides were counterstained usinghematoxylin.

Scoring and statistical analysis

Immunostained slides were digitalized with Pannoramic Scan(3DHISTECH) and evaluated blindly by three assessors in-cluding a dermatopathologist using the TMA module ofPannoramic Viewer software. Only strong immunostainingof tumor cell nuclei was considered as positive, and thefrequency of positive tumor cells was assessed using fourcategories based on counting with the permanent MarkerCounter option of the digital slide viewer (see Fig. 2c). Finalcutoff values were determined based on interobserver agree-ment and reproducibility with particular attention to the dis-crimination between dysplastic nevi and thin melanomas.Score categories were as follows: 0 (negative) for <1 %, 1(low) for 1–9 %, 2 (medium) for 10–29 %, and 3 (highlabeling index) for ≥30% positive tumor cells in case of cyclinD1, Cdk2, Mcm6, and Ki67 reactions; and 0 for <1 %, 1 for1–4 %, 2 for 5–19 %, and 3 for ≥20 % positive tumor cells forcyclin A, geminin, Top2a, and aurora A and B reactions.Divergent scores were consolidated based on a consensus. Incase of divergent consensus, for scores given on differentsamples of the same case, always the highest score wasconsidered as the one with the highest potential significance.

Two-sided Fischer’s exact test was used to analyze signif-icance of differential protein expression between melanocytictumor progression groups using the SPSS software (IBM,Armonk, NY, US). All possible cutoff values for positivitywere tested for binary results, i.e., scores 0 versus 1–3, 0–1versus 2–3, or 0–2 versus 3 within each group. Significancewas declared at p<0.05. Correlations between cell cycle phasemarker expressions were tested using the Spearman’s rankanalysis with p<0.001 for significance (Statistica, StatSoftInc., Tulsa, OK, US). The sensitivity and specificity of indi-vidual biomarkers were tested using SPSS. Statistical envi-ronment (R 15.2.0) was applied for hierarchical cluster

Virchows Arch

analysis to test the correlations between complex markerexpression profiles in tumor progression groups. Cases whichhad missing marker expression data were left out from thisanalysis. Discriminant analysis was done for predicting diag-nostic group membership using marker combinations and theleave-one-out classification for cross-validation [31, 32].

Results

Cell cycle phase progression in melanocytic tumors

Differential expression of cell cycle phase markers tested insitu through the common nevi, dysplastic nevi, thin (≤1 mm)melanoma, thick melanoma, and metastatic melanoma se-quence is summarized in Table 1. Conclusions were drawnfrom the existing significance between groups by consideringat least one positivity threshold. Common nevi expressed lowlevels of general (Mcm6 and Ki67) and G1 phase (cyclin D1and Cdk2) markers, the latter showing significantly higherexpression (p≤0.000–p=0.014) in dysplastic nevi. In high-

grade dysplastic nevi, the cell cycle protein expression did notdiffer significantly from that of low-grade lesions, with theexception of a slight increase in Cdk2-positive cells.Expression of post-G1 phase markers, including S-G2 (cyclinA, geminin, and Top2a) and G2-M (aurora A or B) phaseregulators, in thinmelanomas showed significant upregulationcompared with nevi, including dysplastic lesions (p≤0.000 forcyclin A, geminin, Top2a, and aurora B). Mcm6 and Ki67expressions were also significantly increased in thin melano-mas compared with dysplastic nevi. In thick melanomas, allcell cycle progression markers showed significant upregula-tion (p≤0.000–p=0.035) compared with thin melanomas,except the early markers Mcm6, cyclin D1, and Cdk2, sug-gesting steady levels of early markers and accelerated post-G1phase. For most markers, with the exception of aurora A andCdk2, no significant differences in expression were foundbetween thick and metastatic melanomas. The frequency dis-tribution of scores within melanocytic tumor subgroups isshown in Fig. S1. Themain features of cell cycle phase proteinexpression through melanocytic tumor progression are sum-marized in Fig. 2.

Fig. 2 Cell cycle regulatory protein expression in dysplastic nevi (a, b),thin (≤1 mm) melanomas (c–f), thick melanomas (g, i, j), and a metastaticmelanoma (h). a, c, g Mcm6 (c demonstrates digital cell counting using

the Marker Counter software). b Cdk2 (positive dysplastic and negativecompound nevus-CN). d, h Ki67. e, i Cyclin A. f Geminin. j Aurorakinase B (inset: late metaphase dividing cell)

Virchows Arch

Table1

Differentialexpressionof

cellcyclephasemarkersin

benign

andmalignant

melanocytictumors

Com

mon

nevus

Com

mon

vs.dysplastic

nevus(p)

Dysplastic

nevus

Dysplastic

nevusvs.

thin

melanom

a(p)

Thinmelanom

a(≤1mm)

Thinvs.thick

melanom

a(p)

Thick

melanom

a(>1mm)

Thick

vs.m

etastatic

melanom

a(p)

Metastatic

melanom

a

Cdk2(%

)

≥11/19

5.3%

0.000

48/60

80%

0.330

21/23

91.3%

1.000

34/39

87.2%

0.471

16/20

80%

≥10

1/19

5.3%

0.002

27/60

45%

0.462

13/23

56.5%

0.587

26/39

66.7%

0.012

6/20

30%

≥30

0/19

0%

0.107

10/60

16.7%

0.224

7/23

30.4%

0.559

9/39

23.1%

0.734

3/20

15%

Cyclin

D1(%

)

≥18/19

42.1%

0.004

46/58

79.3%

1.000

17/22

77.3%

0.086

38/40

95%

0.602

19/21

90.5%

≥10

3/19

15.8%

0.014

29/58

50%

0.805

12/22

54.5%

0.035

33/40

82.5%

0.117

13/21

61.9%

≥30

0/19

0%

0.189

8/58

13.8%

0.032

8/22

36.4%

1.000

15/40

37.5%

0.391

5/21

23.8%

Mcm

6(%

)

≥18/17

47.1%

0.270

36/57

63.2%

0.001

20/20

100%

–39/39

100%

–21/21

100%

≥10

0/17

0%

0.192

7/57

12.2%

0.001

18/20

90%

0.263

38/39

97.4%

1.000

21/21

100%

≥30

0/17

0%

1.000

––

0.001

13/20

65%

0.106

33/39

84.6%

0.404

20/21

95.2%

Ki67(%

)

≥14/19

21.1%

0.449

7/58

12.1%

0.000

17/21

81.0%

0.013

38/38

100%

–22/22

100%

≥10

0/19

0%

–0/58

0%

0.000

12/21

57.1%

0.007

34/38

89.5%

0.643

21/22

95.5%

≥30

0/19

0%

–0/58

0%

0.017

3/21

14.3%

0.214

12/38

31.6%

0.104

12/22

54.5%

Cyclin

A(%

)

≥10/19

0%

0.328

6/63

9.5%

0.000

17/23

73.9%

0.008

39/40

97.5%

1.000

21/21

100%

≥50/19

0%

––

–0.000

6/23

26.1%

0.008

25/40

62.5%

0.160

17/21

81.0%

≥20

0/19

0%

–0/63

0%

–0/23

0%

0.529

2/40

5%

0.329

3/21

14.3%

Top2a(%

)

≥12/18

11.1%

0.332

3/59

5.1%

0.000

10/21

47.6%

0.000

39/40

97.5%

1.000

22/22

100%

≥50/18

0%

–0/59

0%

0.000

7/21

33.3%

0.015

27/40

67.5%

0.136

19/22

86.4%

≥20

0/18

0%

–0/59

0%

0.066

2/21

9.5%

0.470

8/40

20%

0.226

8/22

36.4%

Gem

inin

(%)

≥11/19

5.3%

1.000

4/62

6.5%

0.000

13/23

56.5%

0.000

39/40

97.5%

1.000

22/22

100%

≥50/19

0%

–0/62

0%

0.000

8/23

34.8%

0.001

31/40

77.5%

0.082

21/22

95.5%

≥20

0/19

0%

–0/62

0%

–0/23

0%

0.287

4/40

10%

0.438

4/22

18.2%

AuroraB(%

)

≥12/19

10.5%

0.253

2/58

3.4%

0.000

8/21

38.1%

0.000

33/39

84.6%

0.405

21/22

95.5%

≥50/19

0%

–0/58

0%

0.068

2/21

9.5%

0.000

23/39

59.0%

0.406

16/22

72.7%

≥20

0/19

0%

–0/58

0%

–0/21

0%

0.545

3/39

7.7%

0.060

6/22

27.3%

AuroraA(%

)

≥10/17

0%

–0/61

0%

0.059

2/20

10%

0.000

26/37

70.3%

0.111

18/20

90%

5–20

0/17

0%

–0/61

0%

0.247

1/20

5%

0.076

10/37

27.0%

0.023

12/20

60%

Pvalues

werecalculated

atallavailablepositiv

itycutoffvalues

bythetwo-sidedFisher’sexacttest.Significant

values

ofp<0.05

arein

bold

Virchows Arch

Cluster analysis of melanocytic tumors based on cell cyclephase marker expression

Complex in situ protein expression profiling revealedthat tumor progression from dysplastic nevi to thickmelanomas paralleled the gradual elevation of cell cycleprogression, as shown in heat maps following clusteranalysis (Fig. 3). Benign and malignant cases wereseparated from each other, though subgroups formedsmall clusters mixed within their own (benign or malig-nant) categories. Four cases diagnosed originally as thinmelanoma clustered within the benign groups resulting in amisclassification of potentially adverse clinical significance.However, so far, these patients are disease free after 7 to10-year follow-up.

Dendrograms grouped markers in accordance with theirmain roles and emergence during the cell cycle (Fig. 3). Aclose association was observed between the S-G2 phasemarkers geminin, cyclin A, Top2a, and the G2-M markersaurora B and A, including also the general cell cycle markerKi67. Early cell cycle markers cyclin D1, Cdk2, and Mcm6formed a separate cluster.

Spearman’s rank correlations (ρ) in primary melano-mas were also significant (p<0.001) in these relations,ranging between 0.48 and 0.77 among post-G1 (S-G2-M) phase markers, between 0.44 and 0.65 for Ki67 andpost-G1 phase markers, and between 0.42 and 0.57 forMcm6 and post-G1 proteins or Ki67. A significant associ-ation was also found between cyclin D1 and Cdk2 expression(ρ=0.52) and between cyclin D1 and Top2a expression(ρ=0.45) but not between cyclin D1/Cdk2 and other post-G1 phase markers.

Cell cycle markers can differentiate dysplastic nevi from thinmelanomas

Discriminant analysis was used to determine the predictivevalue of marker combinations for separating dysplastic nevifrom thin melanomas. Mcm6 with Ki67 (at ≥10 % and ≥1 %cutoff, respectively) proved to be the best dual marker com-bination, which correctly classified 95.9 % of the originalsamples and 94.5 % of the cross-validated grouped cases with84.2 % test sensitivity and 100 % specificity. Since theseproteins were also detected in some dysplastic nevi, combin-ing the best post-G1 phase marker cyclin A with Mcm6 andKi67 correctly classified 95.9 % of the original samples and93.2% of the cross-validated grouped cases and improved testsensitivity to 89.5% at 92.6% specificity (Table 2). A detailedlist of test sensitivity and specificity for all markers is sum-marized in Table 3.

Discussion

Dysplastic nevi are a risk factor for and a potential precursorof a melanoma, but the clinical and histological similaritiesand lack of unequivocal biomarkers occasionally obscure thedistinction between dysplastic nevi and transforming earlymalignant lesions [4, 33]. Thus, misdiagnosis of melanomasis one of the most frequent causes of cancer malpractice [34].In this study, in situ cell cycle phase analysis revealed thatpost-G1 phase regulators, emerging in thin melanomas but notyet in dysplastic nevi, can indicate malignancy. The sensitivityof differentiation improved when the S phase promoter cyclinAwas combined with the replication licensing Mcm6 and the

Fig. 3 Heat map of unsupervisedhierarchical cluster analysis basedon complex cell cycle phasemarker testing. Top dendrogramscluster nevi and melanomasseparately, but subgroups aremixedwithin their own (benign ormalignant) categories. Thick(PM>1) and metastaticmelanomas (MetM) are clusteredtogether. Left dendrograms showclose correlations between post-G1 phase markers and Ki67 andbetween early cell cycle markersand Mcm6. CN common nevus,DN dysplastic nevus, PM<1 thinmelanoma (≤1 mm)

Virchows Arch

general proliferation marker Ki67. Significant increase in theexpression of S-G2-M phase regulators in thick (>1 mm)melanomas compared with that in thin melanomas sug-gests that accelerated cell cycle progression in melano-mas peaks when the vertical growth phase is acquired.The lack of major cell cycle changes in metastatic com-pared with thick primary melanomas indicates that high pro-liferation by itself does not determine metastatic melanomainvasion.

Melanoma development and progression can be driven byactivating mutations in growth promoting pathways (BRAFand NRAS), loss of function mutations (CDKN2A), or epige-netic deregulation of the cell cycle checkpoint control, gene

amplification of cell cycle promoters (CDK4 or CCND1), andmicrosatellite instability [7, 35, 36]. These distinct events canall contribute to aberrant cell replication [8], which can be anindicator of malignant phenotype. Elevated replication inmelanomas compared with that in benign nevi has been re-ported by testing Cdk1, Cdk2, and Cdk4 [21, 37]; cyclin D1[20, 38–41], cyclin E [21, 42], cyclin A [20, 43, 44], andcyclin B1 [44, 45]; Mcm2 [46, 47], Mcm3, and Mcm4 [48];geminin [47]; Top2a [49, 50]; aurora kinase A and B [51]; andKi67 [52–55] expressions in melanocytic tumors. However,comprehensive analysis of cell cycle fractions and their rela-tion to melanocytic tumor progression and malignancy has notbeen determined. Furthermore, early comprehensive cell cyclestudies in pigmented skin tumors referring to “very weakreactions” were likely to fail because of the use of inappropri-ate antigen retrieval, primary antibodies, or immunodetectionsystems [21].

Global mRNA expression profiling studies also revealedthat the “proliferation signature” distinguished between mel-anomas and nevi but not between thin/radial melanomas anddysplastic nevi [56] (GEO Series GSE12391). This was prob-ably due to the bias introduced by proliferating basal epider-mal keratinocytes when using whole slides instead of micro-dissected tumor samples. In situ cell cycle protein testing canovercome this sampling defect.

Advanced immunohistochemistry has been a widely avail-able, cost- and tissue-efficient technique which can reliablydetect proteins in situ that regulate the cell cycle. Differentialexpression of these regulators allows assessment of main cellcycle fractions (G1 and post-G1 phase, S-G2 and G2-M) intheir in vivo microenvironment for correlations with tumorprogression [8, 57]. The clear nuclear signals are in line with

Table 2 Classification of atypical nevi and thin (≤1 mm) melanomas bydiscriminant analysis combining Mcm6, Ki67, and cyclin A detection

Marker combination Predicted group membership

Mcm6+Ki67+cyclin A DN PM<1 Total

Original Number DN 53 1 54

PM<1 2 17 25

Percent DN 98.1 % 1.9 % 100 %

PM<1 10.5 % 89.5 % 100 %

Cross-validated Number DN 53 1 54

PM<1 4 15 19

Percent DN 98.1 % 1.9 % 100 %

PM<1 21.1 % 78.9 % 100 %

95.9 % of the original grouped cases correctly classified

93.2 % of the cross-validated grouped cases correctly classified

DN dysplastic nevus, PM<1 thin melanoma

Table 3 Sensitivity and specificity of cell cycle phase markers for separating thin melanomas from dysplastic nevi

Protein Cutoff frequency(%)

Sensitivity(%)

Specificity(%)

PPV(%)

NPV(%)

FNR(%)

FPR(%)

Case number (dysplasticnevi/thin melanomas)

Mcm6+Ki67

≥10 84.2 100 100 94.7 15.8 0 54/19≥1

Mcm6+ Ki67+ cyclin A

≥1 89.5 92.6 81.0 96.2 10.5 7.4 54/19≥1≥1

Mcm6 ≥10 90 87.7 72 96.2 10 12.3 57/20

Ki67 ≥1 81.0 87.9 70.8 92.7 19.0 12.1 58/21

Cyclin A ≥1 73.9 90.5 73.9 90.5 26.1 9.5 63/23

Mcm6 ≥30 65 98.2 92.9 88.9 35 1.8 57/20

Ki67 ≥10 57.1 100 100 86.6 42.9 0 58/21

Geminin ≥1 56.5 93.5 76.5 85.3 43.5 6.5 62/23

Top2a ≥1 47.6 94.9 76.9 83.6 52.4 5.1 59/21

Cutoff values show the minimum percent of positive cells required for positive qualification

PPV positive predictive value, NPV negative predictive value, FNR false negative rate, FPR false positive rate

Virchows Arch

the biological role of cell cycle markers and suited for digitalpathology platforms allowing automated cell counting and therapid and accurate measurement of microscopic tumordimensions [29].

Mcm6 positivity can specify all tumor cells that are li-censed and engaged in the cell cycle [10]. In G1 phase, cellsare under the influence of growth factors for which theybecome refractory from S phase [8]. In melanocytic tumors,the replicative compartment, determined by the growth frac-tion and cell cycle kinetics, reflects deregulated growth pro-moting and cell cycle control pathways and correlates withneoplastic progression and clinical tumor growth [19, 23]. Thegrowth fraction is mostly measured by the Ki67 index.However, both Ki67 and Mcm6 proteins are constitutivelyexpressed from early G1 to late M phases as they also labelcells which start but cannot progress through the replicationcycle, potentially leading to inaccurate results [8, 19].

In dysplastic nevi, positive reactions we detected forMcm6,Ki67, and G1 phase markers cyclin D1 and Cdk2 indicatedcells that were committed to replication [23]. Significantlyhigher expression of cyclin D1 and Cdk2 was noticed indysplastic nevi compared with common nevi but not to mela-nomas. The role of cyclin D1 in melanomas is controversial.Most studies are in line with ours by linking cyclin D1 only tothe initial phases of melanocytic tumor development since itsexpression does not change significantly between dysplasticnevi and melanomas [38]. Expression of cyclin D1 was evenfound reduced in vertical compared with radial growth mela-nomas [20, 39] or metastatic melanomas [40]. In a randomlyselected melanoma cohort, represented by most studies includ-ing ours, the frequent copy gains (in 51%) and losses (in 32%)of the CCND1 gene [58] are likely to neutralize each other’seffect when grouped in protein expression studies. Differingproportions of these cases may lead to controversial data oncyclin D1 expression and melanoma development.

Cell cycle phase markers expressed through S-G2 phasesincluding cyclin A, Top2a, and geminin and those detectedthrough G2-M phases such as aurora kinases A and B canidentify cell fractions which duplicate their DNA and arelikely to complete mitosis [8]. The lack of post-G1 phasemarkers in dysplastic nevi suggests that replication-initiatedcells in these lesions fail to complete the cell cycle and remainarrested in G1 phase. Thus, post-G1 phase markers emergingfirst in thin melanomas can serve as indicators of the malig-nant phenotype. Both in primary and metastatic melanomastested in this study, the Ki67 index showed significant corre-lation with the expression of post-G1 phase markers suggest-ing that, unlike in dysplastic nevi, G1 phase arrest is negligiblein malignant melanomas.

Except for M phase delay in serious mitotic aberrations, thetime cells proceed through S-G2-M phases (cell cycle time) isfairly constant [19]. Thus, post-G1 phase fractions can alsoreflect cell cycle kinetics in melanoma. In our cohort, post-G1

phase markers showed progressive expression from acquiringmalignancy to vertical tumor growth. These data suggest thatin melanocytic lesions, an increasing number of tumor cellspasses through the S phase by replicating their DNA andcompletes mitosis through the dysplastic nevus-thinmelanoma-thick melanoma sequence. This shift to the rightin the cell cycle is consistent with an accelerated progressionof replication when primary melanoma evolved. During pro-gression, melanomas can switch between proliferative andinvasive phenotypes [59]. Future testing of the potential ofinvasive foci, defined as areas with low expression of cellcycle markers, in tumor progression and prognosis may allowfurther functional stratification of melanomas.

The differential power of cell cycle phase markers wassupported by existing significance at least at two scoringthresholds between tumor progression groups, which may beutilized for diagnosing malignancy in critical melanocyticlesions. The relatively high number of Mcm6 (>10 %) andeven a few Ki67 (>1 %)-positive cells raises the probability ofa melanoma, which may be confirmed by few positive cellsfor cyclin A (>1 %) or other post-G1 phase markers. In case ofonly few (>1 %) Mcm6- and Ki67-positive tumors cells, thelesion may equally qualify as a dysplastic nevus or a melano-ma, and detection in >1 % of cyclin A or other post-G1 phasemarkers can support malignancy in doubtful cases. The fourpatients from the early malignant group who we reclassifiedinto the dysplastic nevi group based on cell cycle phasemarker analysis are still disease free for 7–10 years.Retrospective correlations of in situ cell cycle testing withclinical behavior can clarify the frequency of underdiagnosisor overdiagnosis using the presented algorithm.

In conclusion, aberrant cell proliferation resulting fromderegulated cell growth and cell cycle control underlinesmalignant phenotype. In melanocytic tumors, the combineddetection of post-G1 phase and general cell cycle markers mayprovide significant information on malignancy in doubtfulcases, which might prevent false negative diagnosis, andsupports the concept of an association of accelerated cell cycleprogression with vertical melanoma growth.

Acknowledgments The authors are indebted to Irma Korom for super-vising the diagnoses of melanocytic tumors used in this study and to EditParsch for excellent technical support. We also thank Erika Fugg,Krisztina Koraszne Lauf, Diana Papp, and Eva Veszpremi for theirassistance.

Conflict of interest The authors of this manuscript declare no conflictof interest.

References

1. Garbe C, Leiter U (2009) Melanoma epidemiology and trends. ClinDermatol 27:3–9. doi:10.1016/j.clindermatol.2008.09.001

Virchows Arch

2. Leiter U, Meier F, Schittek B, Garbe C (2004) The natural course ofcutaneous melanoma. J Surg Oncol 86:172–178. doi:10.1002/jso.20079

3. Stratigos AJ, Katsambas AD (2009) The value of screening inmelanoma. Clin Dermatol 27:10–25. doi:10.1016/j.clindermatol.2008.09.002

4. Duffy K, Grossman D (2012) The dysplastic nevus: from historicalperspective to management in the modern era: part II. Molecularaspects and clinical management. J Am Acad Dermatol 67(19):e11–e12

5. Rabkin MS (2008) The limited specificity of histological examina-tion in the diagnosis of dysplastic nevi. J Cutan Pathol 35(Suppl 2):20–23. doi:10.1111/j.1600-0560.2008.01131.x

6. Carlson JA, Ross JS, Slominski A, Linette G, Mysliborski J, Hill J,Mihm M Jr (2005) Molecular diagnostics in melanoma. J Am AcadDermatol 52:743–775. doi:10.1016/j.jaad.2004.08.034, quiz775-748

7. Sekulic A, Haluska P Jr, Miller AJ, Genebriera De Lamo J, Ejadi S,Pulido JS, SalomaoDR, Thorland EC,Vile RG, SwansonDL, PockajBA, Laman SD, Pittelkow MR, Markovic SN (2008) Malignantmelanoma in the 21st century: the emerging molecular landscape.Mayo Clin Proc 83:825–846

8. Williams GH, Stoeber K (2012) The cell cycle and cancer. J Pathol226:352–364. doi:10.1002/path.3022

9. Pollock PM, Harper UL, Hansen KS, Yudt LM, Stark M, RobbinsCM, Moses TY, Hostetter G, Wagner U, Kakareka J, Salem G,Pohida T, Heenan P, Duray P, Kallioniemi O, Hayward NK, TrentJM, Meltzer PS (2003) High frequency of BRAF mutations in nevi.Nat Genet 33:19–20. doi:10.1038/ng1054

10. Lei M (2005) The MCM complex: its role in DNA replication andimplications for cancer therapy. Curr Cancer Drug Targets 5:365–380

11. Vermeulen K, Van Bockstaele DR, Berneman ZN (2003) The cellcycle: a review of regulation, deregulation and therapeutic targets incancer. Cell Prolif 36:131–149

12. Sancar A, Lindsey-Boltz LA, Unsal-Kacmaz K, Linn S (2004)Molecular mechanisms of mammalian DNA repair and the DNAdamage checkpoints. Annu Rev Biochem 73:39–85. doi:10.1146/annurev.biochem.73.011303.073723

13. Kellner U, Sehested M, Jensen PB, Gieseler F, Rudolph P (2002)Culprit and victim—DNA topoisomerase. II Lancet Oncol 3:235–243

14. Diffley JF (2004) Regulation of early events in chromosome replica-tion. Curr Biol 14:R778–R786. doi:10.1016/j.cub.2004.09.019

15. Vader G, Lens SM (2008) The aurora kinase family in cell divisionand cancer. BiochimBiophysActa 1786:60–72. doi:10.1016/j.bbcan.2008.07.003

16. Brown DC, Gatter KC (2002) Ki67 protein: the immaculate decep-tion? Histopathology 40:2–11

17. Endl E, Gerdes J (2000) The Ki-67 protein: fascinating forms and anunknown function. Exp Cell Res 257:231–237. doi:10.1006/excr.2000.4888

18. Hanahan D, Weinberg RA (2011) Hallmarks of cancer: the nextgeneration. Cell 144:646–674. doi:10.1016/j.cell.2011.02.013

19. Pierard GE (2012) Cell proliferation in cutaneous malignant melano-ma: relationship with neoplastic progression. ISRN Dermatol 2012:828146. doi:10.5402/2012/828146

20. Alonso SR, Ortiz P, Pollan M, Perez-Gomez B, Sanchez L, AcunaMJ, Pajares R, Martinez-Tello FJ, Hortelano CM, Piris MA,Rodriguez-Peralto JL (2004) Progression in cutaneous malignantmelanoma is associated with distinct expression profiles: a tissuemicroarray-based study. Am J Pathol 164:193–203

21. Georgieva J, Sinha P, Schadendorf D (2001) Expression of cyclinsand cyclin dependent kinases in human benign and malignantmelanocytic lesions. J Clin Pathol 54:229–235

22. Li W, Sanki A, Karim RZ, Thompson JF, Soon Lee C, Zhuang L,McCarthy SW, Scolyer RA (2006) The role of cell cycle regulatory

proteins in the pathogenesis of melanoma. Pathology 38:287–301.doi:10.1080/00313020600817951

23. Quatresooz P, Pierard GE, Pierard-Franchimont C (2009) Molecularpathways supporting the proliferation staging of malignant melano-ma (review). Int J Mol Med 24:295–301

24. Moore MW, Gasparini R (2011) FISH as an effective diagnostic toolfor the management of challengingmelanocytic lesions. Diagn Pathol6:76. doi:10.1186/1746-1596-6-76

25. Vergier B, Prochazkova-Carlotti M, de la Fouchardiere A, Cerroni L,Massi D, De Giorgi V, Bailly C, Wesselmann U, KarlseladzeA, Avril MF, Jouary T, Merlio JP (2011) Fluorescence in situhybridization, a diagnostic aid in ambiguous melanocytic tu-mors: European study of 113 cases. Mod Pathol 24:613–623.doi:10.1038/modpathol.2010.228

26. Zhang G, Li G (2012) Novel multiple markers to distinguish mela-noma from dysplastic nevi. PLoS One 7:e45037. doi:10.1371/journal.pone.0045037

27. Balch CM, Gershenwald JE, Soong SJ, Thompson JF, Atkins MB,Byrd DR, Buzaid AC, Cochran AJ, Coit DG, Ding S, EggermontAM, Flaherty KT, Gimotty PA, Kirkwood JM, McMasters KM,Mihm MC Jr, Morton DL, Ross MI, Sober AJ, Sondak VK (2009)Final version of 2009 AJCC melanoma staging and classification. JClin Oncol : Off J Am Soc Clin Oncol 27:6199–6206. doi:10.1200/jco.2009.23.4799

28. Kononen J, Bubendorf L, Kallioniemi A, Barlund M, Schraml P,Leighton S, Torhorst J, Mihatsch MJ, Sauter G, Kallioniemi OP(1998) Tissue microarrays for high-throughput molecular profilingof tumor specimens. Nat Med 4:844–847

29. Krenacs T, Ficsor L, Varga SV, Angeli V, Molnar B (2010) Digitalmicroscopy for boosting database integration and analysis inTMA studies. Methods Mol Biol 664:163–175. doi:10.1007/978-1-60761-806-5_16

30. Krenacs T, Kiszner G, Stelkovics E, Balla P, Teleki I, Nemeth I, VargaE, Korom I, Barbai T, Plotar V, Timar J, Raso E (2012) CollagenXVII is expressed in malignant but not in benign melanocytic tumorsand it canmediate antibody inducedmelanoma apoptosis. HistochemCell Biol 138:653–667. doi:10.1007/s00418-012-0981-9

31. McLachlan GJ (2005) General introduction. Discriminant analysisand statistical pattern recognition. Wiley, pp. 1–26

32. McLachlan GJ (2005) Likelihood-based approaches to discrimina-tion. Discriminant analysis and statistical pattern recognition. Wiley,pp. 27–51

33. Salopek TG, Kopf AW, Stefanato CM, Vossaert K, Silverman M,Yadav S (2001) Differentiation of atypical moles (dysplastic nevi)from early melanomas by dermoscopy. Dermatol Clin 19:337–345

34. Troxel DB (2003) Pitfalls in the diagnosis of malignant melanoma:findings of a risk management panel study. Am J Surg Pathol 27:1278–1283

35. Hussein MR, Sun M, Tuthill RJ, Roggero E, Monti JA, SudilovskyEC, Wood GS, Sudilovsky O (2001) Comprehensive analysis of 112melanocytic skin lesions demonstrates microsatellite instability inmelanomas and dysplastic nevi, but not in benign nevi. J CutanPathol 28:343–350

36. van den Hurk K, Niessen HE, Veeck J, van den Oord JJ, van SteenselMA, Zur Hausen A, van Engeland M, Winnepenninckx VJ (2012)Genetics and epigenetics of cutaneous malignant melanoma: a con-cert out of tune. Biochim Biophys Acta 1826:89–102. doi:10.1016/j.bbcan.2012.03.011

37. Kuzbicki L, Aladowicz E, Chwirot BW (2006) Cyclin-dependentkinase 2 expression in human melanomas and benign melanocyticskin lesions. Melanoma Res 16:435–444. doi:10.1097/01.cmr.0000232290.61042.ee

38. Alekseenko A, Wojas-Pelc A, Lis GJ, Furgal-Borzych A,Surowka G, Litwin JA (2010) Cyclin D1 and D3 expressionin melanocytic skin lesions. Arch Dermatol Res 302:545–550.doi:10.1007/s00403-010-1054-3

Virchows Arch

39. de Sa BC, Fugimori ML, Ribeiro Kde C, Duprat Neto JP,Neves RI, Landman G (2009) Proteins involved in pRb andp53 pathways are differentially expressed in thin and thicksuperficial spreading melanomas. Melanoma Res 19:135–141.doi:10.1097/CMR.0b013e32831993f3

40. Karim RZ, Li W, Sanki A, Colman MH, Yang YH, ThompsonJF, Scolyer RA (2009) Reduced p16 and increased cyclin D1and pRb expression are correlated with progression in cutane-ous melanocytic tumors. Int J Surg Pathol 17:361–367. doi:10.1177/1066896909336177

41. Ramirez JA, Guitart J, Rao MS, Diaz LK (2005) Cyclin D1 expres-sion in melanocytic lesions of the skin. AnnDiagn Pathol 9:185–188.doi:10.1016/j.anndiagpath.2005.04.018

42. Bales ES, Dietrich C, Bandyopadhyay D, Schwahn DJ, Xu W,Didenko V, Leiss P, Conrad N, Pereira-Smith O, Orengo I,Medrano EE (1999) High levels of expression of p27KIP1 and cyclinE in invasive primary malignant melanomas. J Invest Dermatol 113:1039–1046. doi:10.1046/j.1523-1747.1999.00812.x

43. Florenes VA,MaelandsmoGM, Faye R, Nesland JM, HolmR (2001)Cyclin A expression in superficial spreading malignant melanomascorrelates with clinical outcome. J Pathol 195:530–536. doi:10.1002/path.1007

44. Tran TA, Ross JS, Carlson JA, Mihm MC Jr (1998) Mitotic cyclinsand cyclin-dependent kinases in melanocytic lesions. Hum Pathol 29:1085–1090

45. Stefanaki C, Stefanaki K, Antoniou C, Argyrakos T, Patereli A,Stratigos A, Katsambas A (2007) Cell cycle and apoptosis regulatorsin Spitz nevi: comparison with melanomas and common nevi. J AmAcad Dermatol 56:815–824. doi:10.1016/j.jaad.2006.09.015

46. Boyd AS, Shakhtour B, Shyr Y (2008) Minichromosome mainte-nance protein expression in benign nevi, dysplastic nevi, melanoma,and cutaneous melanoma metastases. J Am Acad Dermatol 58:750–754. doi:10.1016/j.jaad.2007.12.026

47. de Andrade BA, Leon JE, Carlos R, Delgado-AzaneroW,Mosqueda-Taylor A, de Almeida OP (2013) Expression of minichromosomemaintenance 2, Ki-67, and geminin in oral nevi and melanoma. AnnDiagn Pathol 17:32–36. doi:10.1016/j.anndiagpath.2012.05.001

48. Gambichler T, Shtern M, Rotterdam S, Bechara FG, Stucker M,Altmeyer P, Kreuter A (2009) Minichromosome maintenance pro-teins are useful adjuncts to differentiate between benign and malig-nant melanocytic skin lesions. J Am Acad Dermatol 60:808–813.doi:10.1016/j.jaad.2009.01.028

49. Lynch BJ, Komaromy-Hiller G, Bronstein IB, Holden JA (1998)Expression of DNA topoisomerase I, DNA topoisomerase II-

alpha, and p53 in metastatic malignant melanoma. Hum Pathol29:1240–1245

50. Mu XC, Tran TA, Ross JS, Carlson JA (2000) Topoisomerase II-alpha expression in melanocytic nevi and malignant melanoma. JCutan Pathol 27:242–248

51. Wang X, Moschos SJ, Becker D (2010) Functional analysis andmolecular targeting of aurora kinases A and B in advanced melano-ma. Genes Cancer 1:952–963. doi:10.1177/1947601910388936

52. Bergman R, Malkin L, Sabo E, Kerner H (2001) MIB-1monoclonal antibody to determine proliferative activity ofKi-67 antigen as an adjunct to the histopathologic differentialdiagnosis of Spitz nevi. J Am Acad Dermatol 44:500–504.doi:10.1067/mjd.2001.111635

53. Chorny JA, Barr RJ, Kyshtoobayeva A, Jakowatz J, Reed RJ (2003)Ki-67 and p53 expression in minimal deviation melanomas as com-pared with other nevomelanocytic lesions. Mod Pathol 16:525–529.doi:10.1097/01.mp.0000072747.08404.38

54. Kaleem Z, Lind AC, Humphrey PA, Sueper RH, Swanson PE, RitterJH, Wick MR (2000) Concurrent Ki-67 and p53 immunolabeling incutaneous melanocytic neoplasms: an adjunct for recognition of thevertical growth phase in malignant melanomas? Mod Pathol 13:217–222. doi:10.1038/modpathol.3880040

55. Kapur P, Selim MA, Roy LC, Yegappan M, Weinberg AG, HoangMP (2005) Spitz nevi and atypical Spitz nevi/tumors: a histologic andimmunohistochemical analysis. Mod Pathol 18:197–204. doi:10.1038/modpathol.3800281

56. Scatolini M, Grand MM, Grosso E, Venesio T, Pisacane A, BalsamoA, Sirovich R, Risio M, Chiorino G (2010) Altered molecular path-ways in melanocytic lesions International journal of cancer. J Int duCancer 126:1869–1881. doi:10.1002/ijc.24899

57. Loddo M, Kingsbury SR, Rashid M, Proctor I, Holt C, YoungJ, El-Sheikh S, Falzon M, Eward KL, Prevost T, Sainsbury R,Stoeber K, Williams GH (2009) Cell-cycle-phase progressionanalysis identifies unique phenotypes of major prognostic andpredictive significance in breast cancer. Br J Cancer 100:959–970. doi:10.1038/sj.bjc.6604924

58. Vizkeleti L, Ecsedi S, Rakosy Z, Orosz A, Lazar V, Emri G, KoroknaiV, Kiss T, Adany R, Balazs M (2012) The role of CCND1 alterationsduring the progression of cutaneous malignant melanoma. TumourBiol 33:2189–2199. doi:10.1007/s13277-012-0480-6

59. Hoek KS, Eichhoff OM, Schlegel NC, Dobbeling U, Kobert N,Schaerer L, Hemmi S, Dummer R (2008) In vivo switching of humanmelanoma cells between proliferative and invasive states. Cancer Res68:650–656. doi:10.1158/0008-5472.can-07-2491

Virchows Arch