ABSTRACT: Cutaneous melanomas are malignant radio and chemoresistant neoplasias presenting high morbidity and elevated mortality rates. Isolated Limb Perfusion (ILP) with Melphalan (Mel) is used in the treatment of non-resectable locally advanced melanomas of extremity sparing the limb from amputation in 40-50% of the cases. Adding the Tumor Necrosis Factor alpha (Tnfa) improves total complete response to 70-90%. The cellular and molecular mechanisms underlying the changes promoted by Mel and Tnfa are not completely understood. In this study we evaluated the impact of systemic Mel and Tnfa administration on tumor growth, analyzed the morphological changes promoted by each treatment, and searched for early molecular targets of Mel and Tnfa, alone or in combination, in a murine melanoma model. Morphological changes were analyzed by histological analysis, while microarray gene expression followed by quantitative RT-PCR were used to search for early molecular targets. We found that Mel systemic administration accounted for the impairment of tumor-growth (p<0.001) and improvement of survival. Tnfa co-administration augmented necrosis (p<0.024) and decreased mitotic rates (p=0.001). We identified a set of 118 potential molecular markers that might be correlated with the observed biologic response to treatment with Melphalan and Tnfa and which could represent potential therapeutic targets in melanoma. We used amplified RNA (aRNA) from 18 tumor samples (Control – 5 samples; Mel – 4 samples; Tnfa – 4 samples; and Mel+Tnfa – 5 samples). For replica hybridization with dye swap, aRNA from the tumors and from a pool composed of equal amounts of RNA extracted from the K1735M2, M2R, B16F10, and B16F1 melanoma cells were labeled with Alexa555 or Alexa647. Labeled aRNA samples were hybridized against a glass platform containing 16,128 immobilized sense oligonucleotides (Fox Chase Cancer Center, USA) corresponding to murine genes. Slides were pre-hybridized at 42ºC for approximately 6 hours in a solution containing Denhardt’s 5X (Ficoll 400 0.05g/mL, poli-vynil-pyrrolidone 0.05g/mL, bovine serum albumin 0.05g/mL), SSC (sodium saline citrate) 5X, 0.2% SDS (Sigma), and 1% BSA. For hybridization, we used a mix of 3.5µg of each labeled sample aRNA and reference aRNA. The hybridization solution contained Denhardt’s 5X, formamide 25% (Sigma), SSC 5X, SDS 0.1%, salmon sperm DNA 0.1mg/mL (GE Healthcare), Poly-A 0.1mg/mL (GE Healthcare), and Cot-1 0.1mg/mL (GE Healthcare), to a final volume of 100µL. Hybridizations were carried out on a Gene Tac hybridization station (Genomic Solutions) at 42ºC for about 16h. The slides were removed from the hybridization cassettes directly to a recipient containing a washing solution (SSC 2X, SDS 0.1%) at 42ºC, put in a slide rack, transferred to another recipient containing fresh solution, and washed under constant agitation at 42ºC for 5 min. After that, they were washed twice for 5min in a second wash solution (SSC 0.1X, SDS 0.1%) at room temperature and rinsed five times for 1min at room temperature with a SSC 0.1X solution. The slides were dried upside down in a centrifuge at 1500rpm for 5min. The slides were scanned with a confocal laser scanner (ScanArrayTM Express, Perkin-Elmer Life Sciences, USA), and the spot intensities processed with the help of the ScanArray Express program (Packard BioScience), with a 10µm resolution and a PMT of 60% for Alexa 555 and of 70% for Alexa 647. Each slide generated a data set for each channel corresponding to the dyes Alexa 555 and Alexa 647. The slides were scanned with a confocal laser scanner (ScanArrayTM Express, Perkin-Elmer Life Sciences, USA) with a 10µm resolution and a PMT of 60% for Alexa 555 and of 70% for Alexa 647, and data were extracted with ScanArray Express software (Packard BioScience) using the histogram method. Each slide generated a data set for each channel corresponding to the dyes Alexa 555 and Alexa 647. The Locally Weighted Scatter-plot Smoothing method (LOWESS), adjusted for linear and non-linear systematic variations, both of the intensity dependent type, mainly for low intensity spots, was employed for data normalization.