Despite an apparent effective vaccination, some patients are admitted to the hospital after SARS-CoV-2 infection. The role of adaptive immunity in COVID-19 is growing; nonetheless, differences in the spike-specific immune responses between patients requiring or not hospitalization for SARS-CoV-2 infection remains to be evaluated. In this study, we aim to evaluate the spike-specific immune response in patients with mild-moderate or severeSARS-CoV-2 infection, after breakthrough infection following two doses of BNT162b2 mRNA vaccine. We included three cohorts of 15 cases which received the two BNT162b2 vaccine doses in previous 4 to 7 months: 1) patients with severe COVID-19; 2) patients with mild-moderate COVID-19 and 3) vaccinated individuals with a negative SARS-CoV-2 molecular pharyngeal swab (healthy subjects). Anti-S1 and anti-S2 specific SARS-CoV-2 IgM and IgG titers were measured through a chemiluminescence immunoassay technology. In addition, the frequencies of IFNγ-releasing cells were measured by ELISpot. The spike-specific IFNγ-releasing cells were significantly lower in severe patients (8 0; 26 s.f.c.×106), as compared to mild-moderate patients (135 64; 159 s.f.c.×106; p<0.001) and healthy subjects (103 50; 188 s.f.c.×106; p<0.001). The anti-Spike protein IgG levels were similar among the three cohorts of cases (p = 0.098). All cases had an IgM titer below the analytic sensitivity of the test. The Receiver Operating Curve analysis indicated the rate of spike-specific IFNγ-releasing cells can discriminate correctly severe COVID-19 and mild-moderate patients (AUC: 0.9289; 95%CI: 0.8376-1.000; p< 0.0001), with a diagnostic specificity of 100% for s.f.c. > 81.2 x 106. 2-doses vaccinated patients requiring hospitalization for severe COVID-19 show a cellular-mediated immune response lower than mild-moderate or healthy subjects, despite similar antibody titers.