Although much emphasis is given to the use of immune checkpoint inhibitors to restore the functionality of exhausted lymphocytes, very little is known about the fate of cancer cells that escape from the cytotoxic activity of T cells. In a previous issue of Cancer Research, Stein and colleagues investigated the response of cancer cells to CD8+ T cells disarmed of their killing activity. Spared cancer cells acquired stem cell–like features and displayed an enhanced capacity to form tumors and metastasize. These increased tumorigenic properties could represent the other side of the coin of T-cell surveillance seen in wound healing in which recognition of damaged tissue as “self” gives the green light for healing process.
See related article by Stein and colleagues; Cancer Res 79(7):1507–19
The interface between a target cell and CTL is known as an immunologic synapse, and is the location where focal signaling and polarized release of cytotoxic molecules such as perforins and granzymes, cytokines, and FasL occurs. In sites of chronic inflammation such as the tumor microenvironment (TME), T cells become “exhausted” and lose their effector functions. These unresponsive T cells express several inhibitory receptors such as PD-1, CTLA-4, LAG-3, TIGIT, and Tim-3. Under physiologic conditions, these inhibitory pathways, the so-called immune checkpoints, maintain self-tolerance and protect normal tissues from damage caused by the activated immune system. Cancer cells have evolved various mechanisms to elude immune attack including inactivation of CTLs. Reactivation of CTLs is the rationale behind immune checkpoint inhibitors (ICI). However, what happens if CTLs fail to kill cancer cells?
Resistance of cancer cells to CTL cytotoxic activity is known to extend the persistence of the immunologic synapse, resulting in hypersecretion of inflammatory cytokines and chemokines by T cells (1). These inflammatory molecules in turn modify the TME. In a previous issue of Cancer Research, Stein and colleagues investigated another effect of CTL failure to kill cancer cells (2). In their experiments, peptide-loaded human breast cancer MCF-7 cells were challenged with antigen-specific CD8+ T cells deactivated with concanamycin A, which inhibits the release of lytic granules. MCF-7 cells are resistant to FasL due to low levels of its receptor and do not express caspase-3, thus, they are relatively resistant to apoptosis. In this setting, cognate interactions perturbed TNF/NFκB signaling, IFNγ response, IL2/STAT5 signaling, and inflammatory genes in MCF-7 cells. Strikingly, cellular dedifferentiation pathways were stimulated in these cancer cells. The authors showed that nonlytic immune cell interactions increased the expression of pluripotency genes and caused an enrichment of CD44highCD24low cancer stem cells (CSC). Specifically, the authors documented increased NANOG and Oct4 promoter activity that led to increased expression of epithelial-to-mesenchymal (EMT) transcription factors. Of note, while Snai2/Slug and Twist expression was specifically induced by cognate interactions, Snail upregulation was evident in noncognate interactions. The latter finding supports the idea that CD8+ T cells can induce the expression of some EMT factors in cancer cells in the absence of proper TCR engagement. Functional in vivo experiments confirmed that cancer cells challenged by T cells displayed increased tumor-initiating properties and capacity to metastasize to lymph nodes without substantially affecting the phenotypic features of primary tumors including proliferation and expression of the estrogen receptor and progesterone receptor. In other words, this work supports the idea that, if CTLs fail to kill cancer cells, CTLs could eventually contribute to aggressiveness by stimulating surviving cancer cells into a more stem-like phenotype.
The authors speculated that this phenomenon could contribute to hyperprogressive disease (HPD), which is a sudden acceleration of tumor growth observed in a subset of patients during anti-PD-1/PD-L1 immunotherapy (3). In support of this idea, it has been reported that some patients with head and neck and lung cancers treated with ICIs experience HPD, highlighting a need for predictive biomarkers and for understanding the mechanisms underlying HPD. Recently observed antibody–Fc/FcR interactions on macrophages in HPD led to the hypothesis that immunophenotyping these macrophages could be one such biomarker (4). Stein and colleagues suggested that the observed increase in the number of CSCs caused by nonlytic interactions with CD8+ T cells might contribute to metastasis, local, and distant relapse, and resistance to therapy, but it is unlikely this accounts for the extremely rapid tumor growth boost observed in patients with HPD during treatment with anti-PD1 or PD-L1–based ICIs.
This same group previously showed that failure of trastuzumab-mediated antibody-dependent cellular cytotoxicity by natural killer (NK) cells led to the acquisition of CSC-like properties by the surviving cancer cells (5), suggesting that the protumorigenic effect of nonlytic interactions between cancer and immune cells may not be limited to CTLs. Of note, both studies used only cancer cell lines. This study was based on MCF-7 cells, a model of luminal A breast cancer subtype associated with good prognosis. The experimental conditions were optimized to minimize apoptosis to enable the investigation of the surviving cancer cells. As mentioned above, MCF-7 cells are relatively resistant to apoptosis. In this study, CTLs were inactivated with concanamycin A, further preventing cancer cell death. It is worth noting that failure of CTL lytic activity is frequent in cancer. Granzymes and “death ligands” released by CTLs and NKs are known to induce the apoptotic pathway, which is almost always compromised in cancer cells. A plethora of different mechanisms are known to block the expression of proapoptotic proteins (e.g., p53 and caspase loss) or to induce antiapoptotic molecules. This is the case of Bcl2 family members that block mitochondrial (or intrinsic) apoptosis and of members of the inhibitor of apoptosis (IAP) protein family (XIAP, cIAP1, cIAP2, and survivin), which protect cancer cells from apoptosis triggered by CTLs and NKs. From these observations, it is possible to envisage that the administration of proapoptotic compounds together with immunotherapy could increase the efficacy of the treatment and, simultaneously, minimize cancer cell survival after interaction with the immune system. Not surprisingly, recent preclinical in vivo experiments have shown that combination of ICIs with IAP-targeting compounds or Smac mimetics in glioblastoma and multiple myeloma models can induce durable and, in some cases, curable effects (6).
These findings suggest two potential mechanisms underlying ICI treatment failure. First, cancer cells may acquire resistance to CTLs. Second, CTLs in the TME may be intrinsically unable to kill cancer cells. In the work by Stein and colleagues, CTLs were incapacitated with concanamycin A; further studies are required to identify such innately deficient T cells in vivo. Stein and colleagues demonstrated that concanamycin A–treated CD8+ T cells maintained the capacity to be activated (CD137/4-1BB and IFNγ expression), unlike exhausted T cells, which cannot be activated and are unable to kill cancer cells. Therefore, the question is whether exhausted T cells are able to induce the dedifferentiation programs described in this study (2). Another open question is how nonlytic immune cell interactions stimulate the formation of CSCs. Recent data support the notion that PD-L1 itself promotes stemness in cancer cells and stimulates NANOG and Oct4 expression (7). This raises the fascinating possibility that PD1 from exhausted T cells may be able to stimulate PD-L1 “reverse signaling” in cancer cells that would then promote a dedifferentiation program. CSCs have also been shown to express PD-L1 and other immunomodulatory factors, and these cells may contribute to local immunosuppression by encouraging the accumulation of regulatory T cells (Tregs) and myeloid-derived suppressor cells. Therefore, CSCs could actively contribute to the formation of “disarmed” T cells, which in turn would increase the number of CSCs through nonlytic interactions, in a vicious cycle.
A fascinating prospect is that the findings of Stein and colleagues may relate not only to cancer, but also to wound repair in normal tissue. In fact, when nonmalignant tissues are damaged, recruited immune cells could promote tissue regeneration by cells recognized as “self” and hence spared. In this setting, released cytokines might support dedifferentiation gene programs for stem cell production, which would in turn contribute to tissue healing. In this respect, while some work supports a role of Tregs in repairing cutaneous wounds (8), others have shown that CD4+ and CD8+ cells play a role in determining the immune infiltration composition of damaged skin without affecting its repair (9). Therefore, additional models and studies using these models are necessary to dissect this aspect and clarify the role of CD8+ T cells in tissue repair in skin as well as other tissue types. After all, many immune cell populations are already known to be fundamental for tissue remodeling and have been shown to favor the physiologic development of different organs.
In conclusion, the work by Stein and colleagues supports the notion that nonlytic interactions between CD8+ T and other cells could activate a dedifferentiation program and stimulate the production of new stem cells. In normal tissues, this process would be auspicious and promote tissue healing; in malignant tissues, it would favor the accumulation of tumor- and metastasis-initiating CSCs. Consequently, failure of CTLs to induce apoptosis in target cells would result in more aggressive tumors. Notably, apoptotic and stressed cells can also stimulate the growth of bystander cells, both normal and stem-like, through a process called compensatory growth (10). This implies that, whether successful or not in inducing cell apoptosis, CTLs may sometimes be detrimental in cancer.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
D. Lecis is supported by the Italian Ministry of Health (5 × 1000 Funds - 2013). M.P. Colombo is supported by the Italian Ministry of Health (RF-2016-02364121) and Associazione Italiana per la Ricerca sul Cancro (IG 18425 and 12162 Special Program “Innovative Tools for Cancer Risk Assessment and Early Diagnosis,” 5 × 1000).