In order to investigate SARS-CoV-2 mutations and their impact on immune evasion and infectivity, we developed a Deep Mutational Scanning (DMS) platform utilizing an inverted infection assay to measure spike expression, ACE2 affinity, and viral infectivity in human cells. Surprisingly, our analysis reveals that spike protein expression, rather than ACE2 affinity, is the primary factor affecting viral infectivity and correlated with SARS-CoV-2 evolution. Notably, within the N-terminal domain (NTD), spike expression and infectivity-enhancing mutations are concentrated in flexible loops. We also observed that Omicron variants BA.1 and BA.2 exhibit immune evasion through receptor binding domain (RBD) mutations, although these mutations reduce structural stability. Interestingly, the NTD has evolved to increase stability, compensating for the RBD instability and resulting in heightened overall infectivity. Our findings, available in SpikeScanDB, emphasize the importance of spike expression levels and compensatory mutations in both the NTD and RBD domains for shaping Omicron variant infectivity.