New insights on the role of CD8(+)CD57(+) T-cells in cancer.

Incomplete differentiation of CD8+ cytotoxic T-lymphocytes (CTLs) in the tumor microenvironment is associated with cancer progression. We describe a new type of tumor-infiltrating CD8+CD57+ T cell in cancer with hybrid phenotypic and functional properties of both an early effector-memory cell and a terminally-differentiated effector cell. These cells behave as incompletely-differentiated CTLs.

Melanomas are considered to be one of the most immunogenic cancers and are frequently infiltrated with CD8+ and CD4+ T lymphocytes specific for major tumor antigens. CD8+ cytotoxic T lymphocytes (CTL) mediate antigen-specific lysis of tumor cells through cytolytic granule proteins, such as granzymes, granulysin, and perforin. Previously, other investigators have shown a lack of complete CTL differentiation in solid tumors. However, the exact stage of differentiation affected was unclear. We recently reported in Clinical Cancer Research a novel subset of CD8+ tumor-infiltrating lymphocytes (TILs) that co-expressed early effector-memory markers (CD27, CD28) and a marker for end-stage, senescent T cells (CD57).

CD28 and CD27 are markers of early CD8+ effector-memory T (TEM) cells. Based on studies on virus-specific T cells in humans, it was postulated that CD8+ TEM cells differentiate in a linear pathway from CD28+CD27+ (early-differentiated) to CD28-CD27+ (intermediate-differentiated) to CD28-CD27- (late-differentiated). As CD8+ TEM transitions to end-stage effector (TEFF), the loss of CD28 and gain of CD57 is an immunological characteristic of humans and non-human primates, but not of mice. CD57 is a marker on highly differentiated CD8+CD27-CD28- T cells needed to control CMV and other endemic viruses in humans., CD57 was also proposed as a marker of end-stage, senescent CD8+ T cells in HIV patients exhibiting highly cytotoxic potential., However, in HIV progressors, a failure to coordinately downregulate CD27 and upregulate CD57 resulted in an accumulation of an unusual subset of HIV-specific CD8+CD27+CD57+ cells. CD8+CD27+CD57+ T lymphocytes have also been observed in the peripheral blood of melanoma patients after vaccination with gp100 tumor antigen.

We set out to determine the role of CD8+CD57+ T cells in the melanoma. By performing flow cytometry staining on TIL isolated from 44 metastatic tumors, we found that the CD8+CD57+ subset was 16.2 ± 3.5% of the total tumor-infiltrating CD8+ T cells. The vast majority of these CD8+ T cells were also “locked” in the TEM stage and, on average, > 20% of the CD8+CD57+ subset co-expressed CD27 and CD28. These T cells were GBhi but Perflo/-, and they recognized melanoma tumor antigens, MART-1 and gp100 in HLA-A2+ patients. In contrast, only a few (< 5%) of the CD8+CD57+ T cells in the PBMC of the same melanoma patients co-expressed CD27 and CD28, but they were GBhi and Perfhi. We also found a similar population in pleural effusions from metastatic breast cancer. Notably, this TIL subset was different from the Foxp3+ CD8+ early effector TILs, which did not express CD57, as reported earlier.

Whether CD8+CD57+ T cells are really “senescent” has become a controversial issue. We addressed this issue in the CD8+CD27+CD57+ TIL subset and found that they indeed proliferated in response to IL-2. We then purified this subset from IL-2-cultured TILs by cell sorting. Interestingly, we found that the CD27+CD57+ TILs were able to proliferate and produce high levels of IFN-γ upon TCR stimulation, which was inconsistent with CD57 as a general marker for T-cell senescence. We also found that PD-1 expression was higher in the CD27+CD57+ subset compared with the CD27+CD57- subset. Since CD8+ T cells naturally express PD-1 as a result of TCR activation, where it plays a regulatory mechanism to prevent over-reactivity, it is possible that the CD8+CD27+CD57+ subset are more highly activated T cells in the tumor microenvironment.

Building on our observation that CD8+CD57+ TILs were non-senescent, we hypothesized that these T cells were in a transition state and could be induced to differentiate further to cytotoxic end-stage effectors. Indeed, we found that the sorted CD8+CD27+CD57+ TILs, which were Perflo and poorly cytotoxic, could differentiate into a Perfhi, highly cytotoxic CD27-CD57+ or CD27-CD57- subset after TCR stimulation (Fig. 1). IL-2 treatment alone induced a minor fraction of CD27+ precursor cells to become CD27+CD57+, which suggested that IL-2 was sufficient to expand the CD27+CD57+ subset from CD27+ precursors (Fig. 1). TCR-stimulation induced the differentiation of sorted CD8+CD27+CD57- TILs into a Perfhi, CD27-CD57- subset, but the cells differentiating from the CD27+CD57+ subset on average acquire higher perforin levels and killing function (Fig 1). We also found that TGF-β1, produced by many melanomas, arrested the differentiation and cytotoxic activities of both the CD27+CD57- and CD27+CD57+ subsets at the CD27+ stage. Taken together, these findings may help explain why metastatic melanomas are often infiltrated with only early TEM CD8+ T cells that have limited ability to kill tumor cells. Thus, TGF-β1 may be a key suppressor of CTL differentiation in the tumor microenvironment. However, other factors such as PGE2, indoleamine 2,3-diooxygenase (IDO), IL-10, or inhibitory signaling through PD-1, may also play a role.

Figure 1. Differentiation pathway of the tumor-infiltrating CD8+ T cells in metastatic cancer. In situations where CD8+ T cells encounter persistent, chronic antigenic stimulation such as metastatic cancer or uncontrolled chronic viral infections, CD8+ effector-memory T (TEM) cells fail to coordinate downregulation of CD27 with upregulation of an end-stage CTL marker, CD57 and acquire a more cytolytic phenotype. Thus TEM fail to transition from a granzyme B (GB+) perforinlo (Perflo) cells into Perfhi, highly cytotoxic end-stage CTL. This resulted in accumulation of CD8+ T cells at a transitional stage where markers for early TEM (CD27, CD28) are co-expressed with CD57, even though the cells remain Perflo. We also found that TGF-β1, an immunosuppressive cytokine frequently found in the microenvironment of metastatic cancer, could also contribute to the arrested differentiation and accumulation of CD27+CD57- precursor T cells and CD27+CD57+ T cells. We found that PD-1 was expressed more in the CD27+CD57+ T cells, which implied that they may exhibit a higher level of effector activity. When these tumor-infiltrating lymphocytes (TIL) are expanded with IL-2, a minor fraction (~30%) of CD27+CD57- subset differentiated into CD27+CD57+ T cells. After TCR stimulation ex vivo, the CD27+CD28+CD57- subset directly differentiated to become CD27-CD57- T cells. On the other hand, CD27+CD57+ differentiated to become CD27-CD57+, and, in some patients, CD27-CD57-. These phenotypic changes were accompanied by increased perforin expression and acquisition of potent cytotoxicity against tumor cells. We think that CD57 is not a marker for T-cell senescence, but rather marks highly differentiated T cells that are in the process of transitioning into a truly end-stage, effector CTL. Currently it is not known which set of markers define truly senescent, end-stage, highly cytolytic CTL. We also propose that ultimately, both the CD8+CD57+ and CD8+CD57- subsets may ultimately differentiate into a subset of CD8+CD27-CD28-CD57- cells that may be more NK-like, expressing higher levels of CD122, KLRG1, NKG2D and other NK receptors.

Another interesting observation we made was that both the CD8+CD27+CD57+ and CD8+CD27+CD57- T cells seemed to ultimately differentiate into CD27-CD57- (also CD28-) cells subset with high perforin and killing activity. Are these the true end-stage or terminally-differentiated state of CTL that actually may be more NK-like, by expressing CD122, other NK receptors (NKG2D and high KIR levels)? These cells have been described before, but their origin is unknown.

In conclusion, we characterized a novel subset of TILs in metastatic melanoma and breast cancer that seems to be locked in a transitional state between early TEM and fully-differentiated CTL. Our study also underscores the need for a greater understanding of the state of CTL differentiation and activity in patients in relation to tumor control. Overall, there has been an over-emphasis on the role of cytokine release by CD8+ T cells (e.g., IFNγ), which has generally not correlated with clinical efficacy during cancer immunotherapy. A case in point is the recent HIV vaccine trials where effective immunization correlated more with a gain of antigen-specific CTL phenotype and killing function rather than IFN-γ production.