Immune checkpoint inhibitors (ICI) have revolutionized the care of patients with advanced melanoma. However, most suffer from immunological side-effects, and half of them experience severe side-effects requiring treatment discontinuation and immunosuppressive drugs. Adverse events that are refractory to corticosteroids, particularly colitis, require the use of infliximab, an anti-TNF antibody. My team previously showed that blocking TNF improves the survival of mice treated with anti-PD-1 in a mouse melanoma model. These preclinical studies allowed my team to initiate two clinical trials: MELANFα, a prospective translational trial designed to assess whether changes in plasma TNF levels can identify advanced melanoma patients whose disease is not progressing on ICI therapy; and TICIMEL, a phase Ib clinical trial designed to assess the safety of concomitant blockade of TNF (infliximab or certolizumab), PD-1 (nivolumab) and CTLA-4 (ipilimumab). My thesis objectives are to evaluate (i) whether TNF is a predictive biomarker of clinical response and a therapeutic target in patients with advanced melanoma treated with ICI; (ii) the impact on the immune response of anti-TNF in combination with ICI in patients and in mouse melanoma models. Although the evolution of plasma TNF does not precisely discriminate patients according to their response, TNF levels increase in the plasma of patients treated with ICI. During treatment, these levels are higher in non-responders than in responders. Single cell RNA sequencing analyses indicate increased TNF signaling in circulating T cells in non-responders compared to responders. TICIMEL has demonstrated the safety of concomitant inhibition of TNF, CTLA-4 and PD-1 to treat patients with advanced melanoma. Inhibition of TNF does not attenuate the response to ICI as we observed activation of the peripheral immune response. The ipilimumab / nivolumab / certolizumab combination is associated with better efficacy signals but more toxicity than the ipilimumab / nivolumab / infliximab combination. In mice, I have confirmed that TNF blockade improves the efficacy of the anti-PD-1 + anti-CTLA-4 combination and increases the survival of B16-OVA melanoma-bearing mice. It also promotes long-term immunological memory in a rechallenge experiment. This increase in efficacy was associated with a reduction of the proportion of regulatory T cells (Treg) infiltrating the tumor and a decrease in the CD8/Treg ratio in the tumors. In the tumor-draining lymph node, I observed that blocking TNF induces increased proportions of CD8 and CD4 T cells expressing Slamf6 and PD-1 compared with the anti-PD-1 + anti-CTLA-4 combination. These cells may correspond to precursor exhausted T cells whose proliferation has a major impact on the efficacy of anti-PD-1. Sequencing of bulk RNA extracted from tumors of mice treated with anti-PD-1 + anti-CTLA-4 ± anti-TNF shows that TNF blockade is associated with signatures of T and B lymphocyte activation and leukocyte-mediated cytotoxicity. Collectively, my thesis results demonstrate that, although TNF does not appear to be a robust predictive biomarker of clinical response in patients with advanced melanoma treated with ICI, plasma levels of TNF and TNF signaling in circulating T cells are increased in non-responders. Moreover, the inhibition of TNF in combination with anti-PD-1 and anti-CTLA-4 is safe in patients and the clinicobiological activity seems to depend on the nature of the anti-TNF. Such a combination does not impair the peripheral immune response and has the potential to improve survival in patients with advanced melanoma, as it seems to be confirmed in a mouse melanoma model.