Development of a Non-invasive Genomic-based Assay to Detect Melanoma William Wachsman1,2, Tara Palmer3, Lory Walls1,2, Sherman Chang3 1Research Service, VA San Diego Healthcare System, San Diego, CA; 2Division of Hematology-Oncology and Moores Cancer Center, UC San Diego; 3DermTech International, San Diego, CA, USA

Introduction Characterization of a 19-Gene Classifier that Identifies Melanoma in Pigmented Lesions Melanoma incidence is increasing in the US population faster than all other cancers. It is already the seventh most TreeNet analysis for class selection and prediction common serious cancer in the United States with a lifetime risk of 1 in 41 in men and 1 in 61 in women. Melanoma accounts for more than 70% of skin cancer deaths, but when detected early it is considered highly curable. In current -Training set: 37 melanomas and 37 nevi • Background subtraction: 100 clinical practice, the detection of melanoma is based upon visual cues, including the “ABCDE” criteria for pigmented Training Set Test Set • Remove gene ≤ 2x difference nevi and results of optical imaging techniques, such as dermoscopy and confocal microscopy. However, only 2 to 10% • t-test with multi-testing correction identified 7,199 differentially of lesions biopsied for suspicion of melanoma are positive for this disease upon histopathologic examination. In expressed genes (p < 0.05; FDR < 0.05) Predicted Class Predicted Class addition, differences in histopathologic review of biopsy specimens by experienced dermatopathologists can result in • TreeNet analyses of differentially expressed genes identified a discordant tissue diagnoses for 10-35% of melanoma cases. Thus, there is a need for a melanoma test that improves 19-gene classifier (sensitivity 100%, specificity 95%) Diagnosis Diagnosis the accuracy, objectivity and ease of detection in clinical practice. -Test set: an independent set of sample data (39 The patented Epidermal Genetic Information Retrieval (EGIR) technology (DermTech International, Inc.) uses a melanomas and 89 nevi) custom adhesive film to non-invasively obtain RNA from stratum corneum (i.e. the upperrmost layer of the skin). • PPV: 39/50 = 78%, NPV: 78/78 = 100% Observed Melanoma Nevi Observed Melanoma Nevi Developed by Morhenn and Rheins (J Am Acad Dermatol 4: 687 [1999]). The EGIR methodology uses four 20 mm • Sensitivity 100%, specificity 88% tapes (pictured below) to noninvasively sample stratum corneum. RNA so harvested has been found to be stable for -The 19-gene classifier also distinguishes melanoma assay, when kept at ambient temperature for at least 72 hours. This makes it an ideal method for use in the clinical from other pigmented lesions and normal skin Melanoma 37 0 Melanoma 39 0 setting. • Basal cell carcinoma (18) • Solar lentigines (12) Previous work showed that EGIR, non-invasive tape stripping of stratum corneum, identified 284 genes that • Non-lesional normal skin (73) Nevi 2 35 Nevi 11 78 differentiated melanoma from atypical nevi and normal skin (p<0.001, false discovery rate q<0.05) . (Wachsman et al. J - Biological functions of the 19-gene classifier Invest Derm 128: S213 [2008]) (see Figure below). Several of the genes were found by Ingenuity Pathways analysis to • 9/19 melanoma/melanocyte associated 78 play a role in melanocyte development and function, as well as, skin development, cellular proliferation, and cancer. Total 39 35 Total 50 • 7 of remaining 10 are cancer associated These findings further demonstrated that the presence of melanoma, directly or indirectly, alters the gene expression profile of stratum corneum.

Study Objective Analysis of False Positive Nevi Identified by the 19-Gene Classifier Translation of the Assay from Microarray to qRT-PCR Platform • The 19-gene classifier identified 11 nevi as melanoma (false positive) Goals The goal of this study is to develop a more objective means for melanoma detection by use of the EGIR technology • Each false positive nevus was serial sectioned and re-reviewed • To transition the 19-gene classifier from the microarray discovery platform to a that non-invasively samples stratum corneum by means of tape stripping. • 10 of 11 remain false positive quantitative RT-PCR platform that will improve clinical utility W • Specimen A (denoted below): Methods • Clark nevus on initial pathological review • Specimens: pre-amplified with WT-Ovation FFPE kit (NuGen, Inc) • Identified as melanoma when analyzed by the 19-gene classifier • qRT-PCR assay for expression of the 19-gene classifier in triplicate • Confirmed by both primary and central pathologists • Data normalized with a reference control (TRIM65) Results • Preliminary results (10 melanomas & 10 nevi) indicate that qRT-PCR recapitulates data obtained using the GeneChip microarray A

Melanoma Atypical nevus Materials, Methods and Clinical Protocol Melanoma Atypical nevus Assay methods • EGIR specimens collected at 9 sites in U.S. • Shipped and stored at -800 C A 32-Gene Classifier Differentiates Solar Lentigo from Lentigo Maligna/Lentigo Maligna Melanoma • MELT (Ambion, Inc.) used to extract RNA - Only used tape demarcated over pigmented lesion - RNA product pooled from 4 tapes • Lentigo maligna/lentigo maligna melanoma (LM/LMM) and solar - Yield and quality assayed • RNA (300 - 500 pg) amplified using WT-Ovation Pico kit (NuGen,Inc.) lentigo are often difficult to distinguish on sun damaged skin • GeneChip assay using U133 plus 2.0 microarray (Affymetrix, Inc.) • Objective: To develop a classifier to differentiate LM/LMM from solar - Data QC using Simpleaffy (Bioconductor) lentigo by EGIR-based genomic assay - GCRMA used to normalize data • Methods Class prediction algorithm • EGIR tape-stripped specimens were expression profiled by GeneChip assay • Microarray data divided into training and test sets • Lentigo maligna (N=5), lentigo maligna melanoma (N=5) and solar lentigo (N=12) • t-test with multi-testing correction to identify differentially expressed genes between melanoma and nevi • GCRMA normalization • Training set: TreeNet® (Salford Systems) used for class selection • Background subtraction: 100 • Test set: TreeNet used for class prediction • t-test between lentigo maligna and solar lentigo (p< 0.01) • Multiple testing correction (False discovery rate (q< 0.05)) Clinical Protocol • 32 differentially expressed genes identified • Inclusion criteria - Subjects 18 or older Conclusion: - Pigmented lesion, suspicious for melanoma that requires biopsy • A 32-gene classifier distinguishes solar lentigo from both lentigo maligna and - Lesion size: 4 mm or greater lentigo maligna melanoma. - If 2 lesions, must have > 4 mm separation • Exclusion criteria Lentigo maligna Solar lentigo - Lesion that is ulcerated, bleeding or weeping - Use of topical moisturizer or sunscreen on sites within 24 h - Allergy to tape or latex • Procedures - Tape stripping of lesion(s) and uninvolved, control skin Conclusions - Demarcate lesion edge on tape - Biopsy lesion as per standard of care • EGIR-harvested stratum corneum RNA can be used for the detection of both in situ and invasive melanoma. - Primary and central dermatopathology review • A 19-gene classifier, developed by microarray analyses of EGIR specimens, discriminates melanoma from Specimens • Training set: 37 melanomas and 37 nevi benign, pigmented nevi. - Superficial spreading melanoma (33):in situ (14) and invasive (19) - Lentigo maligna (3) & lentigo maligna melanoma (1) • Testing of this classifier, shows it to be 100% sensitive and 88% specific for detection of in situ and - Nevi (37): Clark (29) and lentiginous (8) invasive superficial spreading melanoma and lentigo maligna – with a negative predictive value of 100% and • Test set: 39 melanomas and 89 atypical nevi - Superficial spreading melanoma (32): in situ (11) and invasive (21) a positive predictive value of 78%. - Lentigo maligna (2) & lentigo maligna melanoma (4) - Nodular melanoma (1) • Most of the genes in the classifier are known to have a function in melanocyte development, melanoma, or - Nevi (89): Blue nevus (1), Clark (67), congenital (12) and lentiginous (9) cancer biology – suggesting that there is a direct or indirect effect of the malignant melanocyte upon surrounding keratinocytes.

ACKNOWLEDEGMENTS • Further analyses of specimens, deemed to be false positive for melanoma by the 19-gene classifier, is We would like to acknowledge the following clinical sites for the melanoma study: DermTech International • Dr. Tissa Hata, Department of Dermatology, UCSD, La Jolla, CA • Alejandra Ramos needed to determine whether it is more sensitive than standard histopathology. • Dr. Robert Scheinberg, Dermatologist Medical Group of North County, Oceanside, CA • Cheryl Peters • Kamaryn Peters • Dr. Ken Gross, Skin Surgery Medical Group, San Diego, CA • George Schwartz • Dr. Serena Mraz, Solano Dermatology Associates, Vallejo, CA • Preliminary results indicate discovery data, obtained through microarray analyses of EGIR specimens, • Dr. Harold Rabinovitz, Skin and Cancer Associates, Plantation, FL • Dr. Shondra Smith, Dermatology and Advanced Aesthetics, Lake Charles, LA Salford Systems translates well onto a quantitative RT-PCR platform that can be used in the clinical setting. • Dr. Howard Sofen, Dermatology Research Associates, Los Angeles, CA • Dan Steinberg • Dr. James Zalla, Dermatology Associates, Florence, KY • Mikhail Golovnya