The adoptive transfer of lymphocytes genetically engineered to express tumor-specific antigen receptors is a potent strategy to treat cancer patients. T lymphocyte subsets, such as naïve or central memory T cells, selected in vitro prior to genetic engineering have been extensively investigated in preclinical mouse models, where they demonstrated improved therapeutic efficacy. However, so far, this is challenging to realize in the clinical setting, since good manufacturing practices (GMP) procedures for complex cell sorting and genetic manipulation are limited. To be able to directly compare the immunological attributes and therapeutic efficacy of naïve (T_N) and central memory (T_CM) CD8⁺ T cells, we investigated clinical-scale procedures for their parallel selection and in vitro manipulation. We also evaluated currently available GMP-grade reagents for stimulation of T cell subsets, including a new type of anti-CD3/anti-CD28 nanomatrix. An optimized protocol was established for the isolation of both CD8⁺ T_N cells (CD4⁻CD62L⁺CD45RA⁺) and CD8⁺ T_CM (CD4⁻CD62L⁺CD45RA⁻) from a single patient. The highly enriched T cell subsets can be efficiently transduced and expanded to large cell numbers, sufficient for clinical applications and equivalent to or better than current cell and gene therapy approaches with unselected lymphocyte populations. The GMP protocols for selection of T_N and T_CM we reported here will be the basis for clinical trials analyzing safety, in vivo persistence and clinical efficacy in cancer patients and will help to generate a more reliable and efficacious cellular product.