Adoptive cell-transfer therapy (ACT), in which tumour-reactive T cells are activated and expanded in vitro before being returned to the patient, is one of the few methods of cancer immunotherapy that has managed to successfully induce clinical responses against metastatic solid tumours. A particularly attractive aspect of ACT is that specific T cells can be selected for their functional properties before they are transferred into the patient. However, which stage of T-cell differentiation is associated with the successful treatment of tumours in vivo has not been fully investigated, and the current T-cell-selection criteria do not guarantee in vivo efficacy.

Gattinoni et al. analysed the function of CD8+ T cells at different stages of differentiation (naive, early effector, intermediate effector and effector) to determine whether this affected the ability of T cells to mediate tumour regression in a mouse model. Surprisingly, the authors found that, despite having the most effective antitumour effects in vitro, the highly differentiated effector T cells were 100-fold less effective in vivo than T cells that were only at the early effector stage. In fact, the characteristics that are currently used to select T cells for use in the clinic — interferon-γ release and in vitro cytotoxicity — are negatively correlated with in vivo antitumour efficacy.

Microarray analysis revealed that the more differentiated T cells expressed high levels of genes encoding pro-apoptotic molecules, such as BID, BAD and FAS ligand, as well as genes associated with replicative senenscence, indicating that these cells might be less 'fit' in vivo. And the proliferative capacity of the transferred T cells in vivo did, in fact, decrease with the progressive acquisition of in vitro antitumour function.

Analysis of the early-effector T cells identified a subpopulation that expressed high levels of the CD26L marker (CD62Lhigh) and showed superior antitumour efficacy after vaccination with antigen, despite seeming similar to their CD62Llow counterparts. CD62Lhigh cells preferentially home to lymph nodes and analysis showed that this marker targets the transferred T cells to professional antigen-presenting cells (APCs) that are expressing tumour antigen as a result of the vaccination. Loss of CD62L through T-cell differentiation impairs the interaction with APCs and compromises T-cell activation and proliferation in vivo, thereby inhibiting their antitumour activity.

So, early-effector T cells with high levels of lymphoid-homing molecules are the best T-cell populations to use for ACT. However, the necessary step of in vitro T-cell expansion to produce clinically therapeutic cell numbers inevitably causes differentiation and loss of these important cell markers. At present interleukin-2 (IL-2) is used to induce T-cell proliferation, but this also induces differentiation. However, the authors show that IL-15 can uncouple differentiation from proliferation to produce a large T-cell population that is more likely to retain CD62L and is therefore significantly more effective when used in ACT.

These findings will be crucial for the further development of ACT as a clinical treatment, and the authors propose that the current T-cell-selection criteria should be modified to select for less differentiated, more effective T cells.