The introduction and the unexpected efficacy of checkpoint inhibitors (CPI) and more recently of chimeric antigen receptor T cells (CAR T-cells) in the treatment of malignant diseases boosted the efforts in the development and clinical application of immunotherapeutic approaches. However, the definition of predictive factors associated with clinical responses as well as the identification of underlying mechanisms that promote the therapeutic efficacy remain to be determined. Starting from the first immunotherapeutic trials, it became evident that vaccine-induced tumor-specific T cells or the adoptive transfer of ex vivo–expanded T lymphocytes can recognize and eliminate malignant cells leading to long-lasting remissions in some patients. In addition, a phenomenon called epitope spreading, which was observed in responding patients, seemed to increase the efficiency possibly representing an important predictive factor. This review will focus on experimental and clinical evidence for the induction of epitope spreading and its role in the maintenance of an efficient antitumor immune response in cancer immunotherapy.
Epitope or determinant spreading is a process characterized by the enhancement and diversification of the endogenous T-cell response against antigenic epitopes that are different from and non–cross reactive with originally targeted epitope and can be directed against additional different epitopes derived from a defined protein (intramolecular spreading) or other antigens (intermolecular spreading; ref. 1). Epitope spreading was first described in the experimental autoimmune encephalomyelitis (2). Tissue damage caused by inflammation leads to the release of large quantities of different autoantigens. This inflammation induced autoimmunization by “leaking” autoantigens generates and expands a cascade of autoreactive immune cells. In mouse models, it was shown that the hierarchical order of epitope spreading develops from the most to the least immunogenic epitope (3–6). Epitope spreading is playing an important role in autoimmune diseases and during chronic graft rejection (1–7).
In cancer immunotherapy, it was convincingly shown that tumor-specific T cells that were induced upon vaccination or were adoptively transferred could eliminate malignant cells and epitope spreading might increase their clinical activity (8–15). Vaccine-induced or transferred antigen–specific T cells promote tumor cell lysis upon recognition of the cognate peptide in the context of the MHC-complex. Dying cancer cells or released cell components are engulfed by APCs that leads to the generation of novel MHC class I and II peptides that are transported to the cell surface for the priming of newly generated polyfunctional T cells with diverse specificities (Fig. 1) that migrate into the tumor lesions where they eliminate malignant cells or may encounter the antigens presented by resident APCs. Upon restimulation with the antigen, T cells proliferate and release a cascade of cytokines and chemokines involved in differentiation, activation, and recruitment of additional T- and B-cell populations, myeloid cells, and APCs that determine the nature of the immune responses.
Extracellular antigens are usually processed and presented on the MHC class II molecules that prime CD4+ T-cell responses, while cytosolic proteins are loaded on MHC class I resulting in the activation of CD8+ T lymphocytes. However, more than forty years ago, work by M.J. Bevan showed that exogenous antigens could be presented on MHC class I molecules and induce cytotoxic CD8+ T-cell responses. This novel antigen presentation pathway was called cross-presentation (16, 17).
Cross-presentation as a Central Mechanism of Epitope Spreading
Uptake and cross-presentation of antigens released during tumor cell destruction by T cells represents an important initial step during the induction of epitope spreading in cancer immunotherapy.
Two main pathways of antigen cross-presentation referred to as the cytosolic and vacuolar were described that differ by the intracellular location of antigen processing and subsequent loading on MHC class I molecules (18, 19). The concept of the vacuolar pathway is based on the internalization and cleavage of antigens by proteases in the endocytic compartment, where they undergo lysosomal degradation. The generated peptides are loaded on the MHC class I molecules that are recruited from the endoplasmatic reticulum (ER) or plasma membrane and are then transported to the cell surface. This process occurred independently of the TAP (transporter associated with antigen processing) molecule and the involvement of the ER-Golgi apparatus.
For the cytosolic pathway, exogenous antigens are taken up via phagocytosis, pinocytosis, or receptor-mediated endocytosis and traffic to the endocytic compartment. Several surface receptors including FcγR, Clec9a, the mannose receptor or Dectin-1 were shown to be involved in cross-presentation of extracellular antigens (18, 20).
Some of the engulfed antigens may gain access to the cytosol where novel epitopes are generated by the action of the ubiquitin–proteasome pathway (21). Proteasomes cleave their substrates into long peptides that are further hydrolyzed by peptidases. A fraction of the cleaved peptides is then transported by TAP molecule into the ER where the long peptides are further trimmed by aminopeptidases. Short peptides of the right sequence and length are subsequently loaded on MHC class I molecules and transported via the Golgi apparatus to the plasma membrane (18–20).
The first studies that analyzed the pathways and molecules involved in cross-presentation used macrophages. However, it became clear that dendritic cells are key APC that cross-present antigens due to their unique specialized phagocytic pathways such as the reduced acidification and lower degradation of engulfed antigens (18, 19, 22). DCs consist of ontogenetically and functionally of multiple populations and many of these can cross-present antigens. In mice, cross-presenting DCs are characterized by the expression of CD8a, the chemokine receptor XCR1, and/or CD103 as well as the transcription factor Batf3. In humans, cross-presenting DCs were shown to have the BDCA3+, XCR1+, CD141+ phenotype. However, other DC subtypes such as the inflammatory myeloid-derived DCs can also cross present exogenous antigens. The biological significance of other phagocytic cells with the ability to cross-present is unclear and some level of this activity was observed in monocytes, B cells, neutrophils, osteoclasts, microglia, endothelial cells, and Kupfer cells (18–20, 22, 23).
It was recently shown that PD-1 expression by tumor-associated macrophages negatively correlates with the phagocytic activity and the induction of protective antitumor immunity in vivo. The inhibition of PD-1/PD-L1 interaction increases the uptake of cancer cells, reduces the tumor growth, and prolongs survival of treated mice (24). CD47 mediates an antiphagocytic (“don't eat me”) signal and inhibits phagocytosis of malignant cells by binding to its receptor SIRPα (signal regulatory protein) that is expressed on DCs and macrophages (25). The blockade of CD47 increases the destruction of malignant cells and cross-priming of engulfed antigens and this effect can be increased by combinatorial treatments with PD-1 antibodies (26).
Several other mechanisms have been described that are capable to promote the transfer of antigens or antigenic peptides from cells to DCs such as the formations of GAP junctions between APC and tumor cells or a process called “cross-dressing” characterized by the acquisition of MHC/peptide complexes by DCs from other cells (18, 27). In addition, autophagy may contribute to or mediate cross-presentation of antigens (18, 21). Interestingly, it was previously shown that autophagy plays an important role for the generation and the presentation of MHC class II antigens derived from cytosolic proteins (28).
Epitope Spreading Induced by Therapeutic Cancer Vaccines
One of the first reports that addressed the development of determinant spreading in the context of tumor rejection was reported in a mouse model using ovalbumin (OVA)-transfected tumor cells (EG.7) (29). Rejection of EG.7 cells by OVA-specific MHC class I–restricted T cells was followed by intramolecular (OVA peptides) and intermolecular spreading to other endogenous tumor-derived antigens.
In human studies, the induction of determinant spreading was first analyzed in patients with breast cancer upon vaccination with peptides derived from the HER-2/neu protein. A high frequency of intramolecular epitope spreading was observed in vaccinated patients (30). Antigen spreading has been further reported in vaccination trials in patients with metastatic breast and ovarian carcinomas, malignant melanoma, prostate cancer, lung cancer, pancreatic cancer, and renal cell carcinomas that included both CD8+ and CD4+ T-cell responses (12, 13, 31–34). Interestingly, this effect was usually detected several weeks after the initiation of the treatment in clinically responding patients that generated a detectable immune response to the antigens included in the vaccine formulation. In some of these studies, the magnitude and breadth of the elicited epitope spreading trended toward improved survival (12, 13). The screening for epitope spreading was mainly performed by utilizing intracellular IFNγ staining, ELISpot assays, or tetramer analysis. The specificity of the cross-primed T cells was directed to overexpressed, shared, and cancer-testis antigens such as Her-2/neu, hTERT, MUC1, gp100, tyrosinase, MELAN-A, MAGE-A3 etc. These results were probably biased or limited by focusing on already known and characterized TAAs or antigens found by applying comparative gene expression analyses. Using novel technologies such as next-generation sequencing, RNA-sequencing, proteomics, and ligandome analysis, it would be interesting to evaluate the induction and the specificity of epitope spreading in currently performed vaccine trials.
Vaccine-induced epitope spreading can be mediated by using MHC class I- and/or MHC class II–restricted peptide antigens. In one of the first preclinical tumor models that analyzed the role of CD4 cells in epitope spreading, a vaccine that consisted of a long peptide was utilized. It contained a viral Th epitope for the induction of tumor-specific T cells. This approach resulted in the induction of antigen-specific CD4+ T cells and protective immunity against MHC class II–negative virus-induced cancer cells. Tumor rejection was mediated by CD8+ cytotoxic T cells specific for an unrelated viral peptide distinct from the used helper epitope (35). A recent murine study demonstrated that vaccine-induced cancer neoepitope–specific CD4+ T cells may mediate their antitumor activity by augmentation of CTL responses through epitope spreading indicating that currently developing vaccination strategies that contain neoepitopes derived from mutated antigens can generate polyclonal CD8+ T cells in analogy to previous vaccination trials with peptides from self-antigens (36).
These results support the critical role of cross-presentation for epitope spreading and underline the importance of the cognate help provided by CD4 cells for the induction and maintenance of CD8-directed immune responses. During priming, CD4 T cells derived help signals are delivered to CD8+ CTLs by specific DCs with the ability to cross-present (conventional DC1s). These interactions promote the differentiation of effector and memory CTLs and optimize the expansion and quality of CD8+ T-cell responses (37).
Epitope Spreading Induced by Immune Checkpoint Inhibition
Introduction of mAbs blocking CTLA4 or PD-1/PD-L1 molecules (checkpoint inhibitors, CPI) in the therapy of malignant diseases resulted in a dramatic improvement of overall survival of treated patients and induced long-lasting remissions (38, 39).
The analysis of epitope spreading during treatment with CPI was not addressed in the currently performed clinical trials. However, recruitment and expansion of tumor-specific T cells that were not detectable prior to the therapy in responding patients was reported (40–43) and mechanistically there is experimental evidence that support the induction of epitope spreading by CPI.
The antitumor activity of PD-1 inhibition is thought to be mediated by reinvigoration of tumor-infiltrating T cells. PD-1 is essential for homeostatic regulation of peripheral tolerance and maintenance of the T-cell function within a physiologic range. Upon engagement with its ligands PD-L1 or PD-L2, PD-1 transduces a negative costimulatory signal through the recruitment of tyrosine phosphatase SHP2 and dephosphorylation of the TCR and the costimulatory molecule CD28. In addition, PD-1 signaling may regulate T-cell trafficking and have a direct cell-intrinsic effect on tumor cell function by modulating mTOR signaling (39, 44–48).
Therapeutic application of ipilimumab, a mAb that blocks CTLA-4, results in expanding and broadening of the peripheral TCR repertoire that mediates tumor remissions (39, 49, 50). CTLA-4 attenuates TCR signal strength and reduces the threshold for TCR mediated signaling through competitive inhibition of the binding of B7-1 and B7-2 to CD28. In addition, CTLA-4 is constitutively expressed on Tregs and its blockade may dampen the inhibitory effects (51–55). Thus, upon CTLA-4 blockade even cryptic or subdominant epitopes may generate T-cells with the ability to recognize tumor cells. In addition, CTLA-4 inhibition would increase and boost the activity of high-affinity T cell clones, resulting in a more robust activation of tumor-reactive T cells and reduced immune tolerance.
CPI induced hyperactivation of T cells results in an enhanced lysis of malignant cells and the release and cross presentation of novel tumor antigens that might lead to epitope spreading and diversification of the T cell repertoire as it was observed in melanoma patients (40–43). In a recent report, sequential analyses of tumor tissue obtained from patients with Merkel cell carcinomas during immunotherapy with PD-1/PD-L1 checkpoint inhibitors revealed a more pronounced proliferation and diverse clonal expansion of tumor-infiltrating lymphocytes in responding patients (56).
There is further evidence for possible involvement of epitope spreading in the activity of CPI that is derived from animal studies. In a model of experimental autoimmune myasthenia gravis, treatment of mice with anti-CTLA-4 that were immunized with the immunodominant peptide deduced from the extracellular domain of the acetylcholine receptor resulted in epitope spreading characterized by the diversification of the autoantibody repertoire and T-cell responses. Similarly, in relapsing autoimmune encephalitis, CTLA-4 was shown to downregulate epitope spreading and treatment with anti-B7-1 increased epitope spreading and exacerbated relapses of the disease (57–59).
In the context of tumor rejection, application of PD-1 blocking antibodies promote and broaden antitumor directed CD8+ T-cell responses due to epitope spreading and a selective boost of subdominant clones (60).
The dependence of CTL responses on the CD4 T-cell help for the generation of protective anticancer immunity induced by CPI treatment was recently demonstrated in a murine study using tumor derived MHC class I and II restricted neoantigens. CD4+ T-cell responses induced by the combined application of anti-PD-1 and anti-CTLA-4 antibodies were required for sufficient priming and maturation of CD8+ CTLs. The most effective tumor rejection occurred when malignant cells expressed both MHC-I- and MHC-II–restricted neoepitopes (61).
Taken together, the contribution of epitope spreading to the clinical efficacy of CPI was not precisely addressed in previous clinical trials. However, expansion of tumor-specific T cells and diversification of the TCR repertoire was observed in responding patients. From a mechanistic standpoint and preclinical models, there is emerging evidence for the induction of epitope spreading by CPI.
Epitope Spreading Associated with Adoptive T-cell Therapies
The adoptive transfer of tumor-specific T cells demonstrated clinical efficacy in patients with malignant diseases (10, 15, 62). Epitope spreading seems to contribute to the efficiency and was observed in patients achieving remission of the metastatic lesions as well as in preclinical animal models (10, 15, 63–65).
In a murine model, the further improvement of adoptive T-cell therapies and induction of epitope spreading was analyzed using intratumor injection of tumor-specific T cells genetically engineered to transiently express IL12 that stimulates T cells and APCs. This approach resulted in complete remission (CR) of the injected lesion and of distant tumors that was further improved by concomitant stimulation of T cells with CD137 mAb. This treatment induced epitope spreading and activation of endogenous T cells within the tumor that was promoted by released IL12, CD137 stimulation, and cDC1 (63).
In a previous study, 10 patients with stage IV malignant melanoma received infusions of polyclonal autologous MART1-specific T cells generated in vitro using peptide-pulsed DCs in the presence of IL21 (64). The addition of IL21 during priming promotes expansion of CD28+ CTLs with enhanced persistence abilities. T-cell infusion was followed by ipilimumab to improve their antitumor activity. Two patients achieved CRs, 2 experienced partial remissions (PR), and 3 patients had stable disease (SD) with persisting CTLs for up to 40 weeks. ELISpot analyses using overlapping peptides deduced from melanoma-associated antigens MART1, NY-ESO1, gp100, tyrosinase, and MAGE A3 that bind to all MHC molecules and could be recognized by CD4+ and CD8+ T lymphocytes were used to determine epitope spreading. In contrast to patients with progressive disease, all patients who experienced remissions or SD demonstrated an increased or new activity against analysed epitopes that occurred after 6–12 weeks upon CTL infusion.
To reduce the risk of escape, ex vivo–expanded multitumor-associated antigen-specific T cells that target Wilms tumor gene 1, preferentially expressed antigen of melanoma, and survivin were given to patients with relapsed/refractory solid tumors including high-risk sarcomas and neuroblastomas. Eleven of 15 treated patients had stable disease or better at day 45 and 6 responders remained without progression at a median of 13.9 months after initial infusion of antigen-specific T cells. Antigen spreading was observed in responding patients that developed newly arising specificities to TAAs such as MAGE A3, MAGE A4, SOX-2, and SSX-2 (65).
The role of epitope spreading in CAR-T cell therapies was not addressed in current clinical trials and might be limited by the pretreatment therapy for lymphodepletion. Lymphodepletion was shown to enhance the engraftment and expansion of adoptively transferred CAR T cells by increasing the levels of homeostatic cytokines, such as IL7 and IL15. Cyclophosphamide and fludarabine are currently used in the clinical application of CAR T cells and induce a profound elimination of T cells (66, 67). However, there is some evidence that it might occur as demonstrated in preclinical and first clinical studies (68–70).
Epitope spreading was evaluated in patients with metastatic pancreatic carcinomas using CAR T cells that were transduced with mRNA coding for a CAR recognizing mesothelin. Humoral immune responses were analyzed using a high-throughput serologic analysis to detect the presence of IgG antibodies. In all but 1 treated patient, antibody levels against multiple protein targets were found to increase after 1 or 2 months of treatment demonstrating that mesothelin-specific CAR T cells can mediate a vaccine effect by inducing spreading of antibody responses (69, 70). Interestingly, lymphodepletion was not performed in this trial. In addition, the same group found recently that mesothelin CAR T cells can promote clonal expansion of endogenous T cells recognizing malignant cells, which is, however, lost with cyclophosphamide conditioning (71).
To improve the efficacy of CAR T-cell therapy in the immunosuppressive environment of solid tumors CAR T-cells engineered to secrete single-domain antibody fragments directed against checkpoint molecules were analyzed in a syngeneic, immunocompetent mouse model. It was demonstrated that while the release of ant-PD-1 or anti-CTLA-4 improved the persistence of CAR T cells, the secretion of anti-CD47 promoted the activation of the innate immunity and induced epitope spreading that resulted in an increased anticancer response (72).
The generation of sequential immune responses leading to epitope spreading and to the expansion of newly generated T cells with diverse specificities contributes to the efficacy of immunotherapeutic approaches and remains an important research area in tumor immunology. This knowledge might help to better understand the mechanisms involved in long-lasting responses and identify patients benefiting from the treatment. Immunomodulatory compounds that accelerate T-cell priming and activation like anti-CTLA-4, anti-PD-1/PD-L1, or anti-CD137 and improve the uptake of tumor-derived antigens (anti-CD47, anti-PD1) may enhance efficiency by inducing epitope spreading. In addition, immunologic conditioning of the tumor microenvironment (73) as well as antibodies that inhibit the function of MDSC (anti-CSF1R, anti-CCR2/5, anti-osteoactivin/GPNMB) may further increase the activity of the elicited antitumor immunity, enhance cross-presentation and the generation of polyfunctional T cells (74, 75).
Disclosure of Potential Conflicts of Interest
P. Brossart reports receiving commercial research grants from Bristol-Myers Squibb, speakers bureau honoraria from MSD, Bristol-Myers Squibb, Abbvie, and AstraZeneca, and is a consultant/advisory board member for MSD, Amgen, Roche, AstraZeneca, and Bristol-Myers Squibb. No other potential conflicts of interest were disclosed.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.