W hile memory T cells are maintained by continuous turnover, it is not clear how human regulatory CD4⁺CD45RO⁺CD25^hi Foxp3⁺ T lymphocyte populations persist throughout life. We therefore used deuterium labeling of cycling cells in vivo to determine whether these cells could be replenished by proliferation. We found that CD4⁺CD45RO⁺Foxp3⁺CD25^hi T lymphocytes were highly proliferative, with a doubling time of 8 days, compared with memory CD4⁺CD45RO⁺Foxp3^–CD25^– (24 days) or naive CD4⁺CD45RA⁺Foxp3^–CD25^– populations (199 days). However, the regulatory population was susceptible to apoptosis and had critically short telomeres and low telomerase activity. It was therefore unlikely to be self regenerating. These data are consistent with continuous production from another population source. We found extremely close TCR clonal homology between regulatory and memory CD4⁺ T cells. Furthermore, antigen-related expansions within certain TCR Vβ families were associated with parallel numerical increases of CD4⁺CD45RO⁺CD25^hiFoxp3⁺ Tregs with the same Vβ usage. It is therefore unlikely that all human CD4⁺CD25⁺Foxp3⁺ Tregs are generated as a separate functional lineage in the thymus. Instead, our data suggest that a proportion of this regulatory population is generated from rapidly dividing, highly differentiated memory CD4⁺ T cells; this has considerable implications for the therapeutic manipulation of these cells in vivo.