Regulatory T cell (Treg) therapy is a promising approach for transplant rejection and severe autoimmunity. Unfortunately, clinically meaningful Treg numbers can be obtained only upon in vitro culture. Functional stability of human expanded (e)Tregs and induced (i)Tregs has not been thoroughly addressed for all proposed protocols, hindering clinical translation. We undertook a systematic comparison of eTregs and iTregs to recommend the most suitable for clinical implementation, and then tested their effectiveness and feasibility in rheumatoid arthritis (RA). Regardless of the treatment, iTregs acquired suppressive function and FOXP3 expression, but lost them upon secondary restimulation in the absence of differentiation factors, which mimics in vivo reactivation. In contrast, eTregs expanded in the presence of rapamycin (rapa) retained their regulatory properties and FOXP3 demethylation upon restimulation with no stabilizing agent. FOXP3 demethylation predicted Treg functional stability upon secondary TCR engagement. Rapa eTregs suppressed conventional T cell proliferation via both surface (CTLA-4) and secreted (IL-10, TGF-β, and IL-35) mediators, similarly to ex vivo Tregs. Importantly, Treg expansion with rapa from RA patients produced functionally stable Tregs with yields comparable to healthy donors. Moreover, rapa eTregs from RA patients were resistant to suppression reversal by the proinflammatory cytokine TNF-α, and were more efficient in suppressing synovial conventional T cell proliferation compared with their ex vivo counterparts, suggesting that rapa improves both Treg function and stability. In conclusion, our data indicate Treg expansion with rapa as the protocol of choice for clinical application in rheumatological settings, with assessment of FOXP3 demethylation as a necessary quality control step.