Cancer Vaccines: Steering T Cells Down the Right Path to Eradicate Tumors

Although the capability of the immune system to specifically recognize and eliminate cancer cells was recognized more than a century ago, effective strategies, such as adoptive T-cell transfer and immune checkpoint blockade, to consistently direct and enhance tumor-specific immune responses therapeutically in patients have emerged only recently. Adoptive transfer of autologous T cells expanded in vitro from tumor-infiltrating lymphocytes (TIL) and genetically engineered lymphocytes bearing chimeric antigen receptors have demonstrated compelling clinical efficacy in select cancers (1, 2), and targeting inhibitory receptors including PD-1 and CTLA4 with mAbs, a concept termed immune checkpoint blockade (ICB), has demonstrated efficacy in a wide spectrum of solid and hematologic malignancies (3, 4). Approvals of ICB by regulatory agencies have changed the standard of care and improved outcomes for many patients with cancer. The remarkable clinical efficacy of adoptive cell therapy in B-cell malignancies and the broad activity of PD-1/PD-L1 pathway inhibition provide proof-of-concept for cancer immunotherapy in general and the potential of tumor-specific T cells in eradicating tumors. However, with some exceptions (microsatellite-instable tumors, Merkel cell carcinoma, desmoplastic melanoma, cutaneous squamous cell carcinoma, and Hodgkin lymphoma), only up to 30% of patients benefit from the objective responses in most tumor types with proven ICB efficacy, and many malignancies, including highly prevalent ones such as colon and prostate cancers, are largely resistant to this approach.

There is increasing evidence that PD-1 pathway inhibition is effective primarily in tumors with preexistent T-cell infiltration. The observation of high-density CD8⁺ T-cell infiltration at the invasive margins of metastatic melanoma tumors that regressed after PD-1 inhibition, allowing precise prediction of response to anti–PD-1 therapy, provides perhaps the most direct evidence for the importance of tumor T-cell infiltration for effective immunotherapy in humans (5). Additional evidence supporting tumor T-cell infiltration as a prerequisite for the efficacy of ICB is the association of mRNA signatures of IFNγ-related genes and PD-L1 expression on tumor and immune cells with clinical outcomes (6, 7). The marked increase in T-cell numbers and expression of immune-modulatory molecules in early on-treatment tumor samples of patients with advanced melanoma treated with PD-1 inhibition also supports this concept (8). In patients with advanced melanoma, the numbers of Ki67⁺CD8⁺ T cells in the peripheral blood increased after treatment with the PD-1 inhibitor pembrolizumab. These CD8⁺ T cells were CD45RA^loCD27^hi and Eomes^hi and T-bet^lo, indicating an antigen-experienced, exhausted phenotype (9). Ki67⁺PD-1⁺CD8⁺ T cells also increased in the peripheral blood of patients with non–small cell lung cancer who had received PD-1–targeted therapies. Similarly, these peripheral CD8⁺ T cells expressed markers consistent with an antigen-experienced, effector phenotype (HLA-DR⁺, CD38⁺, CD28⁺, CD27⁺, ICOS⁺; ref. 10). These observations indicate that the size and specificity of the T-cell repertoire is shaped by the physiologic interaction between the tumor and the host immune response and is consistent with the mechanism of PD-1 blockade affecting antigen-experienced T cells in the tumor. Mechanisms that can limit the generation of tumor-specific memory effector T cells and their infiltration into a tumor are (i) insufficient T-cell priming; (ii) absence of antigens or dysfunctional antigen presentation leading to immunologic ignorance; (iii) suppressive soluble factors or inhibitory immune cell populations leading to immune tolerance; and (iv) vascular factors, chemokines, or extracellular matrix conditions posing barriers to migration of T cells into the tumor (11). Recent work in mouse models and patients with cancer has highlighted the critical role of dendritic cells (DC) and natural killer (NK) cells in priming and recruiting T cells into the tumor in the context of ICB (12–14).

Therapies that address one, or possibly more, of these rate-limiting steps should have the ability to convert a non–T cell–inflamed “cold” tumor into a T cell–inflamed tumor. It is noteworthy that limiting factors are likely different among individuals, emphasizing the importance of biomarkers that can inform which modality will be most beneficial for an individual patient. Conceptually, several therapeutic modalities can be envisioned given the multitude of mechanisms that can account for the absence of T-cell inflammation. Strategies that induce or enhance innate immune response, thereby providing critical signals to activate specific T-cell immune response (secretion of type I IFNs, activation of DCs leading to augmented antigen presentation and T-cell priming, etc.), have been investigated in preclinical models and are being tested in clinical trials in patients with cancer. Examples include the pharmacologic agonists of Toll-like receptors (TLR) and the STING pathway (15, 16). For instance, intratumoral injection of the TLR9 agonist CMP-001 in combination with PD-1 inhibition was recently shown to mediate objective tumor responses in patients with advanced melanoma who were resistant to PD-1 blockade (17), and the combination of a different TLR9 agonist, SD-101, and low-dose radiation demonstrated clinical efficacy in patients with low-grade B-cell lymphoma (18). In both studies, an increase in CD8⁺ T cells and upregulation of inflammatory genes were seen in post-treatment tumors. Intratumoral administration of oncolytic viruses such as Talimogene laherparepvec (T-VEC) induces both direct tumor cytolysis and an innate immune response, triggering a systemic effect on noninjected metastatic sites (19). In a study of 21 patients with metastatic melanoma combining T-VEC and the PD-1–blocking antibody pembrolizumab, a complete response rate of 33% and overall response rate of 62% suggested a synergistic effect (20); some of these objective responses occurred in patients whose baseline tumors exhibited low CD8⁺ T-cell density and low or absent IFNγ signatures. Increases in the CD8⁺ T-cell density and PD-L1 expression in both injected and uninjected tumors were observed, predominantly in patients who responded to therapy. Blockade of oncogenic pathways such as MAPK, WNT–β-catenin, and PI3K/AKT can reverse defects in T-cell recruitment and priming (21, 22). BRAF and/or combined MEK/BRAF inhibition has been shown to mediate upregulation of melanoma antigens and increased CD8⁺ T-cell infiltration in patients with BRAF^V600-mutant melanoma (23). Preclinical and clinical evidence indicates that a number of other therapeutic approaches including chemotherapy, radiotherapy, agonistic antibodies directed against costimulatory molecules, or the inhibition of VEGF can generate tumor T-cell infiltration (24). Nevertheless, the majority of treatment modalities that mediate anticancer immunity have unintended consequences, such as a decrease in intratumoral T cells, for example, because they are either inherently unspecific (e.g., chemotherapy or radiation) or are directed at targets that are central to many other biological functions (e.g., oncogenic pathways or VEGF).

By exposing a patient with cancer to suitable tumor antigens in the context of T-cell activation signals, a vaccine is designed to amplify preexisting tumor-specific T-cell responses and broaden the T-cell repertoire by inducing de novo responses of either naïve or antigen-experienced T cells. Because of the specificity of the approach, vaccinationis a particularly attractive tool to achieve infiltration of tumors with the “relevant” T-cell populations, that is, T cells that are specific for the tumor and functional, “armed and ready” to destroy the tumor cell. Ideally, a vaccine-induced tumor-directed immune response will be comprised of long-lasting memory T-cell phenotypes mediating durable clinical tumor responses.

The four components of a cancer vaccine are tumor antigens, formulation, immune adjuvants, and delivery vehicles (25), and for decades tumor vaccination was the principal modality in the quest of effective cancer immunotherapy, with much attention on optimizing these components. However, because of the modest clinical efficacy observed in previous large-scale vaccine efforts (26), the overall approach had fallen out of favor, until recently when a new source of tumor antigens became available. The vast majority of previous cancer vaccines have used tumor-associated antigens (TAA), self-antigens which demonstrate tumor-specific or development-specific expression. The development of next-generation sequencing opened up the opportunity to utilize tumor-specific antigens (TSA), such as neoantigens, which arise from somatic cancer mutations. As opposed to TAAs, which also include viral antigens, TSAs are not subject to central tolerance and are therefore more immunogenic.

Neoantigens are generated by cancer mutations and have been implicated as key targets of effective tumor-specific T-cell responses, including the correlation of tumor neoantigen load with tumor cytolytic T-cell infiltration and clinical benefit from immunotherapies, the expansion of neoantigen-specific T cells in patients who derive clinical benefit from immunotherapies, and direct evidence of tumor cell killing by neoantigen-specific T cells in preclinical and human studies (27). The vast majority of cancer mutations are unique to an individual tumor. Coupled with the diversity of HLA molecules, an effective neoantigen-directed therapy will therefore require a personal approach. It is also possible to target multiple neoantigens in each patient, given the observed mutational loads, which can help meet the challenges of clonal heterogeneity and clonal evolution. Technological advances including rapid and cost-effective next-generation sequencing and improved mass spectrometry–based epitope prediction algorithms allow the just-in-time design and production of therapeutic vaccines targeting neoantigens expressed by an individual patient's tumor. Phase I clinical trials in patients with melanoma have already demonstrated that such a personalized vaccination approach is safe and immunogenic in patients with melanoma, with encouraging signs of clinical activity (28–30). Recent studies have demonstrated that tumor mutational burden is not the sole predictive marker of tumor response to ICB and that immune signatures, including those of T cells, DCs, and NK cells, are complementary (5, 31, 32). These data support the investigation of vaccine–ICB combinations in tumors with lower mutational burden. Innovative antigen and adjuvant delivery approaches have the potential to substantially enhance the immunogenicity of cancer vaccines. For example, nanoparticle systems can achieve prolonged and controlled release of antigen and adjuvant by exploiting efficient draining of nanocarriers to lymphoid tissues. In a mouse model, codelivery of peptide and adjuvant in a synthetic high-density lipoprotein (LPS) nanodisc induced more than 30-fold higher frequencies of peptide-specific cytotoxic CD8⁺ T cells compared with the TLR9 agonist combined with Montanide, which had been considered one of the strongest adjuvants currently used in clinical trials (33). Similarly, the delivery of TLRs on polymer scaffolds enhanced DC activation and cytokine production in the draining lymph node and induced up to 100-fold higher numbers of vaccine-specific CD8⁺ T cells (34). For RNA-based vaccines, the challenge of rapid extracellular degradation of RNA can be overcome by formulating the vaccines in a lipid carrier, thus enabling systemic delivery of a vaccine format that has intrinsic adjuvant properties (35).

In an established B16 melanoma model, TEGVAX, an autologous tumor cell vaccine given in combination with GM-CSF and the TLR agonists glucopyranosyl lipid A (TLR4) and resiquimod (TLR7/8), led to significant increases in tumor infiltration with CD4⁺ and CD8⁺ T cells compared with control vehicle, and regression of tumors (36). Notably, when the vaccine was combined with an anti–PD-1 antibody, complete responses of all tumors were seen in 50% of the mice. In the adoptive T-cell transfer Pmel melanoma model, a peptide vaccine directed against gp100 given in combination with an anti-CD40 antibody, the TLR7 agonist imiquimod, and IL2 mediated tumor regression and trafficking of gp100-specific T cells into the tumor (37). In the MC38 colon cancer model, tumors of C57BL/6 mice vaccinated with synthetic long peptides containing MC38-specific neoantigens together with anti-CD40 antibody and poly(I:C), TILs were enriched for CD8⁺ T cells specific for 3 immunogenic neoepitopes (measured by MHC-I dexamer staining), suggesting the generation and migration of vaccine-specific T-cell responses into the tumor (38). The neoantigen peptide vaccine also inhibited the growth of MC38 tumors in both the prophylactic and therapeutic setting. In a separate study, neoantigen-directed RNA vaccines targeting MHC Class II neoepitopes mediated tumor regression and infiltration with both CD8⁺ and CD4⁺ T cells in melanoma, colon carcinoma, and breast cancer models (39). Lymphocytes from tumors that regressed after immunization with the dominant Class II B16 melanoma neoantigen M30 were reactive against B16-M30, demonstrating that neoantigen vaccination led to trafficking of vaccine epitope–specific T cells into the tumor. Separately, the T-cell response induced by vaccination with multiple Class II epitopes spread to an immunodominant Class I epitope that was not contained in the vaccine, indicating that antigen-specific CD4⁺ T cells may promote cross-priming of cytotoxic T-lymphocytes, thus reshaping the repertoire of tumor-specific T-cell responses in the vaccinated host.

Although there is compelling evidence in various mouse models that vaccines can steer cancer-specific T cells into the tumor, evidence in patients has only recently emerged. In one study, in which patients with melanoma were vaccinated with RNA encoding personal neoepitopes, lymphocytes infiltrating a lymph-node metastasis from a patient who had received 4 vaccinations were found to be specific for several vaccine epitopes (29). Reactivity against the neoepitope RETSAT (P546S) was confirmed by MHC multimer staining, and the T-cell receptor (TCR) recognizing RETSAT was characterized by single-cell cloning. Notably, CD8⁺ T cells transduced with the RETSAT TCR demonstrated killing of autologous melanoma cells derived from postvaccination tumors. These data demonstrate that vaccination can effectively drive tumor-specific and functional T cells into a metastatic site in a human patient with melanoma.

We recently demonstrated that personal neoantigen vaccines in patients with glioblastoma induce trafficking of vaccine-specific T cells into intracranial tumors (40). Single-cell RNA sequencing (scRNA-seq) of CD3⁺ T cells present in a relapsed tumor sample revealed cytotoxic CD8⁺ and CD4⁺ T cells and a phenotype of antigen experience and potentially exhaustion. Testing single CD3⁺ T cells from relapsed tumor and circulating T cells reactive to vaccine epitopes revealed that TCR sequences from 4 CD4⁺ T cells and 2 CD8⁺ T cells were identical between tumor and peripheral blood. Specificity for vaccine epitopes was confirmed for a subset of the shared TCRs by cloning and expression into TCR-deficient Jurkat T cells. Notably, although both shared CD4⁺ T cells expressed cytotoxicity markers, one of them also coexpressed the inhibitory receptors PD-1, TIM3, and TIGIT. These data provide definitive evidence that vaccine-induced T cells can traffic into an intracranial tumor in patients with glioblastoma. In another recent study, patients with glioblastoma received two sets of vaccinations (in addition to standard therapy with irradiation and temozolomide): (i) actively personalized vaccines (APVAC1) composed of shared, nonmutated peptides from a premanufactured library that were selected on the basis of mass spectrometry–defined HLA Class I and HLA-DR (Class II) affinity and (ii) preferentially mutated peptides (APVAC2; ref. 41). Approximately 50% of the unmutated APVAC1 peptides induced ex vivo CD8⁺ T-cell responses, whereas >80% of the mutated APVAC2 peptides triggered CD4⁺ T-cell responses. CD4⁺ T cells specific for an APVAC1 unmutated peptide were detected in a post-treatment tumor sample.

As a tool for cancer immunotherapy, vaccines are unique given their ability to focus the host's tumor-specific immune response. By exposing a patient to the relevant (i.e., tumor-specific) antigens in the context of activation signals, vaccines restore a fundamental mechanism that prevents the host from mounting an effective adaptive immune response against cancer cells (despite their “foreignness” given their genetic aberrations). As described, there is now robust evidence that this approach has the potential to activate, amplify, and steer the immune response toward the generation of cancer-specific T cells (Fig. 1, inset). Moreover, recent evidence by us and others, as described above, demonstrates that this approach can drive these tumor antigen–specific T cells into the tumor (Fig. 1). However, there are still many barriers to achieving effective tumor cytolysis. Multiple active suppressive mechanisms in the tumor microenvironment counteract effective tumor-specific immune responses and therefore, particularly for the treatment of advanced metastatic cancers, combinatorial approaches will likely be required. Our own observations of exhausted CD4⁺ T-cell phenotypes and coexpression of three inhibitory receptors on a vaccine-specific CD4⁺ T cell in the tumor of a patient with glioblastoma provide direct evidence that blocking inhibitory pathways may be an essential complementary therapy for successful cancer vaccine approaches.

Indeed, the effectiveness of PD-1/PD-L1–directed monotherapy in subsets of patients with cancer points to this pathway as a dominant suppressive mechanism. Our and others' anecdotal clinical observations in patients with melanoma who experienced disease progression after neoantigen vaccination, followed by rapid-onset complete tumor regressions after treatment with PD-1 inhibitors, are early clinical signals that the combination of vaccine and PD-1 inhibition could be a powerful therapeutic strategy. Clinical trials partnering RNA-based and long-peptide personal neoantigen vaccines with the PD-L1 inhibitors atezolizumab (ClinicalTrial.gov identifier NCT03289962) and nivolumab (ClinicalTrial.gov identifier NCT02897765), respectively, are already ongoing. In the latter of these two studies, a phase Ib clinical trial combining a personal long peptide neoantigen vaccine and nivolumab in patients with metastatic melanoma, non–small cell lung cancer, or urothelial cancer (patients receive nivolumab while vaccines are generated during the first 12 weeks of the trial), objective responses after vaccination were seen in subsets of patients who had not responded to nivolumab alone (42). Circulating immune responses including ex vivo responses directed at the majority of vaccine epitopes were detected in all 16 patients who had completed vaccination and underwent immune analyses. Spreading of circulating T-cell responses to epitopes that were expressed by the tumor but not included in the vaccine, suggesting tumor cell killing (thereby releasing the additional epitopes for recognition by T cells), were detected in 7 of 9 patients tested. Moreover, in a patient with melanoma with durable stable disease after nivolumab and neoantigen vaccine, the postvaccine tumor was specifically enriched for vaccine neoepitope–specific TCR transcripts (not evident prior to nivolumab or vaccination), indicating that the vaccine led to trafficking of vaccine-specific T cells into the tumor in this patient. In future studies, it will be important to further define the specificity, frequencies, and functional states of T cells in tumors after vaccination and whether this information can be linked to clinical benefit from vaccine-based combinatorial therapies.

Although anti–PD-1 or anti–PD-L1 antibodies are an obvious combination partner for a cancer vaccine, recent preclinical data suggest that optimal sequencing of these therapies may be critical for achieving maximal antitumor activity. In the lymphocytic choriomeningitis virus (LCMV) model, abrogation of the PD-1 pathway led to impaired formation of LCMV-specific memory T cells while the addition of PD-1 inhibition to a regimen of combined OX-40 agonistic and a peptide vaccine in the human papillomavirus–associated tumor model TC-1 compromised efficacy and led to apoptosis of tumor-infiltrating vaccine-specific T cells (43, 44). These data indicate that a cancer vaccine may be most effective if administered prior to initiation of PD-1 pathway inhibition. Of note, given the ability of CTLA4 inhibition to enhance T-cell priming in the draining lymph node, there is a rationale to administer vaccine concurrently with a CTLA4 inhibitor (45, 46); PD-1 inhibition could then be given after a series of priming vaccines plus CTLA4 inhibition. These studies also highlight the potential utility of suitable preclinical models to guide the design of clinical studies; a combinatorial vaccine approach could conceivably fail in the clinic simply because of suboptimal dosing and/or scheduling. Given the large number of therapies available for combination with a vaccine and the likely need for comparative studies to definitively prove clinical efficacy, clinical trials must be informed as much as possible by preclinical data and appropriate biomarkers.

Infiltration of tumors with specific and functional memory T-cell responses is critical for durable anticancer immunity. Because of their ability to focus stimulation of tumor-specific T cells, vaccines are arguably the most precise therapeutic approach to achieve effective tumor T-cell infiltration. By targeting neoantigens, delivered in conjunction with enhanced immune adjuvants and partnered with complementary immune therapies that address immune-suppressive circuits in the tumor microenvironment, cancer vaccines may finally fulfill their promise as effective therapy for patients with cancer.

P.A. Ott reports receiving commercial research grants from BMS, Merck, Neon Therapeutics, Celldex, ArmoBiosciences, AstraZeneca/MedImmune, Novartis, Pfizer, CytomX, and Genentech and is a consultant/advisory board member for Merck, BMS, Genentech, Novartis, Pfizer, Neon Therapeutics, Celldex, CytomX, and Array. C.J. Wu is a founder, equity holder, member of the scientific advisory board, and consultant for Neon Therapeutics, Inc.

The authors gratefully acknowledge support from the Francis and Adele Kittredge Family Immuno-Oncology and Melanoma Research Fund (to P.A. Ott), the Faircloth Family Research Fund (to P.A. Ott), the Bender Family Research Fund (to P.A. Ott), the Mathers Family Foundation (to C.J. Wu), and the Blavatnik Family Foundation (to C.J. Wu). C.J. Wu is a Scholar of the Leukemia and Lymphoma Society.