The Changing Landscape of Therapeutic Cancer Vaccines—Novel Platforms and Neoantigen Identification

Cancer immunotherapy is defined as the manipulation of the immune system to recognize and destroy cancer cells. Among approved immunotherapeutic agents, therapeutic cancer vaccines have the advantage of eliciting specific immune responses to tumor antigens. Accordingly, choice of target antigen is of utmost importance when considering vaccine design (1). Tumor-associated antigens (TAA) are self-antigens abnormally expressed by tumor cells. As a result of central and peripheral tolerance mechanisms, the bank of high-affinity T cells for TAAs may be insufficient to elicit an immune response. Cancer vaccines using TAAs must, therefore, be potent enough to “break” these tolerance mechanisms (2). In contrast, tumor-specific antigens (TSA), some of which are neoantigens, are tumor and often patient specific, arising from nonsynonymous mutations, genetic alterations, or virally introduced genetic information in cancer cells. TSAs recognized by high-affinity T cells are, therefore, less likely to be subject to central tolerance and induce autoimmunity (1, 3). Figure 1 provides a summary of TAAs and TSAs in terms of specificity, central tolerance, and prevalence.

Platforms for cancer vaccines are categorized as cellular, viral vector, or molecular (peptide, DNA, or RNA; ref. 1). Cellular vaccines are developed using autologous patient-derived tumor cells or allogeneic tumor cell line–derived cells (4). Dendritic cells (DC) are used to develop cellular cancer vaccines due to their role as consumers, processors, and presenters of tumor antigens. Genetically modified oncolytic viral vaccines are designed to replicate within and eradicate tumor cells (5). Beyond their oncolytic mechanisms, viral vector vaccines also promote tumor-directed immune responses by delivering tumor antigens via more conventional T-cell priming mechanisms (3). MHC proteins present peptides on the cell surface for recognition by T cells (6). Peptide-based cancer vaccines are designed through understanding of peptide–MHC and T-cell receptor/peptide–MHC interactions. Short peptides (typically nine amino acid residues in length) bind directly to MHC molecules, potentially inducing tolerance, and are subject to degradation (7). Longer (typically 30-mer) peptides may be more immunogenic as they are internalized by antigen-presenting cells (APC) and processed for MHC presentation, inducing memory CD4⁺ and CD8⁺ T-cell immune responses (7). DNA vaccines are closed circular DNA plasmids (naked DNA) encoding TAAs and immunomodulatory molecules, aimed at inducing tumor-specific responses (8). Advantages include simplicity, ease of manufacture, and safety; however, naked DNA vaccines have limited efficacy as a result of low transfection rates into target tumor cells. Similarly, mRNA vaccines are synthesized in vitro to encode antigen(s) and express proteins following internalization that stimulate an immune response. mRNA vaccines can deliver a high number of antigens and costimulatory signals, with no risk of infection or insertional mutagenesis, and manufacturing is rapid and inexpensive; however, they are limited by instability and inefficient delivery (8).

Two major advances in the field of therapeutic vaccines, therefore, have been novel platforms and characterization of TSAs. This review focuses on these two essential elements for successful immunotherapy.

Challenges facing T-cell–based cancer immunotherapy include low immunogenicity, as a result of aging or immune cell exhaustion (9, 10) after multiple previous treatment lines; high disease burden; and the immunosuppressive tumor microenvironment (TME), whereby potent immunosuppressive mechanisms evolve throughout cancer progression, enabling cancer cells to escape immune attack (11). Objective responses can be limited to specific subsets of patients with particular genetic mutations, molecular profiles, or recruitment of tumor-infiltrative T cells (12).

Tumor immunogenicity depends on antigenicity and the TME (13). Antigenicity is determined by immune cell infiltration (inflammation) and high mutational burden (genomic instability). High mutational burden in the absence of inflammation can lead to increased antigens with mechanisms for preventing immune cells from infiltrating the TME, as seen in small-cell lung cancers (14). Inflammation without high mutational burden is present in cancers such as renal cell carcinoma, hepatocellular carcinoma, triple-negative breast cancer, gastric cancer, and, to an extent, head and neck cancers (14).

Mechanisms of primary escape (nonresponse to cancer immunotherapy) are thought to depend on underlying drivers of the associated tumor (14). Tumors in sites such as the lymph nodes, lungs, and skin, with a relatively high presence of immune cells, exogenous DNA-damaging insults, or oncolytic viral infections, may be promising sites for anticancer immunity. Conversely, sites such as the bone, intraperitoneal cavity, or blood–brain barrier may be more challenging targets as a result of the high concentration of cytokines, myeloid-derived suppressor cells, and unique stromal interactions indicative of immune-excluded tumors (where CD8⁺ T cells accumulate, but cannot infiltrate; ref. 14). Mechanisms of secondary escape (cancer progression despite previous clinical response), or acquired resistance, can develop from genetic changes in antigen presentation machinery or target antigen loss. Acquired resistance to anti-programmed death (PD)-1 agents in patients with melanoma is associated with loss-of-function mutations in genes encoding IFN receptor–associated JAK1 or JAK2 (15). Immune pressure shapes intratumor genetic heterogeneity, favoring clonal restriction and dominance, and can have important implications for designing therapeutic strategies (16). Loss of antigenicity leads to weak immune response, allowing tumor cells to develop immune evasion mechanisms (13, 17). Thus, different approaches to cancer immunotherapy may be required in these varying TMEs.

Several strategies have been investigated to overcome immunosuppressive mechanisms of the TME and counteract tumor escape, including improving antigen selection, refining immunotherapy delivery platforms, and combination therapies (1). Chemotherapeutic agents, immunomodulatory molecules, checkpoint inhibitors (CPI), and radiation, together with cancer vaccines, may induce neoantigens and work synergistically to target the TME (18–20).

Table 1 presents an overview of current cell-based vaccine strategies under investigation. Examples of cancer vaccines using whole-tumor cells include GVAX (4), which has shown promising activity in several pancreatic cancer trials (21–24) and in hormone-refractory prostate cancer (25, 26). Vigil, an autologous tumor cell vaccine, is also currently being evaluated in phase I and II studies of patients with advanced-stage ovarian cancer, with prolonged relapse-free survival compared with placebo observed in a recent interim analysis of the phase II study (27, 28). Other cell-based vaccine studies have been discontinued as a result of futility (29–31).

Sipuleucel-T (PROVENGE), targeting prostatic acid phosphatase, is approved for the treatment of asymptomatic or minimally symptomatic metastatic castration-resistant prostate cancer. However, despite positive efficacy and safety data, since its approval, barriers to administration of sipuleucel-T and approval of competing cancer therapies have hampered its widespread adoption (32). Several other vaccines derived from ex vivo DCs are being investigated, for example, against melanoma antigen, MART-1 (33). In a phase I trial, a vaccine using autologous monocyte-derived DCs pulsed with oxidized autologous whole-tumor lysate significantly prolonged survival in patients with recurrent ovarian cancer (34). In addition, a vaccine using yeast cell wall particles (YCWP) to load autologous tumor lysate into autologous DCs is being studied for melanoma and for solid tumors (35–38); in a phase II trial, the YCWP vaccine resulted in prolonged disease-free survival in patients with resected melanoma, with a disease-free interval of >3 months, compared with those who received unloaded YCWP (36, 37). A further phase II study of ilixadencel, an off-the-shelf, cell-based immune primer, in combination with sunitinib, pre- and post-nephrectomy, showed greater rates of complete and objective response, but similar progression-free survival, compared with sunitinib monotherapy in patients with newly diagnosed metastatic renal cell carcinoma (39).

Table 2 provides an overview of current virus-based vaccine strategies. The first FDA-approved oncolytic virus for cancer treatment was talimogene laherparepvec (T-VEC; ref. 40). T-VEC relies on direct intratumoral injection, which overcomes dilution and neutralization in blood, to induce cell lysis and promote antitumor immune responses in distant lesions (40–43). A phase II trial of T-VEC in combination with ipilimumab, first or second line, demonstrated a significantly higher objective response rate (ORR) compared with ipilimumab alone in patients with pancreatic ductal adenocarcinoma, with no additional safety concerns (42). The phase III OPTiM study also demonstrated improved progression-free survival, ORR, and overall survival (OS) with T-VEC compared with GM-CSF, particularly in previously untreated patients (41, 43).

A heterologous prime-boost strategy has more recently been used to educate T cells and achieve a robust immune response, where a tumor antigen is delivered with one virus vector first, followed by a boost with the same tumor antigen delivered by a different viral vector or vector type. PROSTVAC-VF/Tricom, using a vaccinia virus encoding prostate-specific antigen (PSA) for priming, followed by subsequent booster doses of a fowlpox virus encoding PSA, demonstrated OS benefit in prostate cancer (44). However, a more recent phase III trial of PROSTVAC in castration-resistant prostate cancer was discontinued as a result of futility (45).

Viruses have also been engineered to simultaneously express targeted antigens and immunomodulatory molecules to disrupt the TME. TG4010 contains the modified vaccinia virus (MVA)-expressing tumor antigen, MUC-1, and immunostimulatory cytokine, IL2 (46, 47). TroVax is an MVA-expressing oncofetal antigen 5T4 (MVA-5T4; ref. 48). MG1 is a version of the oncolytic Maraba virus engineered with added transgene capacity for targeted expression of TAAs and immunomodulatory agents (49) being evaluated in non–small cell lung cancer (NSCLC; ref. 50) and human papilloma virus (HPV)-positive tumors (51).

More recently, several fusion-enhanced oncolytic immunotherapies based on herpes simplex virus (HSV-1; RP1, RP2, and RP3) were engineered to express gibbon-ape leukemia virus envelope proteins (52). In addition, MEDI5395, an attenuated Newcastle disease virus (NDV) genetically modified to express GM-CSF, entered phase I clinical trials for intravenous administration late in 2019. In murine models, intravenous delivery of NDV leads to long-lasting tumor-selective replication, transgene expression, and TME transformation (53). Finally, a B cell/monocyte-based vaccine, BVAC-C, transfected with recombinant HPV 16/18 E6/E7 showed efficacy in activating virus-specific T cells in a phase I study of patients with recurrent cervical cancer. A phase II study of BVAC-C in patients with cervical cancer is underway (54), as is a phase I study of BVAC-B, transfected with recombinant HER2/neu, in patients with gastric cancer (55).

Nanoparticle-based cancer vaccines and adjuvants have been used to target cancers through modification of surface properties and/or composition to prolong bioavailability, protect antigens from degradation, and control antigen release (56). Nanoparticles tested include polymeric nanoparticles, liposomes, micelles, carbon nanotubes, mesoporous silica nanoparticles, gold nanoparticles, and virus nanoparticles, which have been assessed in cancer types such as melanoma, NSCLC, breast, prostate, and cervical (56). However, further studies are needed to address concerns of poor reproducibility with uniform size and shape, aggregation, instability, and rapid clearance before widespread clinical use (56). To date, only one nanoparticle vaccine, tecemotide (L-BLP25), an MUC1 antigen-specific vaccine, has reached clinical trial. In a phase III trial of tecemotide compared with placebo for stage III NSCLC no difference in OS was found (57). Similarly, a phase II trial in early breast cancer demonstrated a good safety profile, but showed no significant difference in residual cancer burden or pathologic complete response with tecemotide compared with standard of care (58).

Synthetic long peptide (SLP) immunotherapeutics have been developed, consisting of highly immunogenic long peptides designed to avoid central tolerance mechanisms by efficiently delivering antigens to DCs, inducing CD4⁺ and CD8⁺ T-cell responses (59). In a phase II trial, an SLP vaccine, ISA101, combined with the anti-PD-1 immune checkpoint antibody, nivolumab, was found to be well-tolerated in patients with HPV-16–positive cancer (n = 24), with additive effects observed relative to nivolumab monotherapy (60). In addition, a phase I/II study of ISA101 in combination with standard-of-care chemotherapy in patients with HPV-16–positive cervical cancer (n = 77) demonstrated a longer OS in patients who expressed a stronger-than-median vaccine-induced HPV-16–specific T-cell response (61). A phase II study of ISA101 with cemiplimab for oropharyngeal cancer is underway (NCT03669718).

SVN53-67/M57-KLH (SurVaxM) is a vaccine containing an SLP mimic designed to stimulate an immune response targeting survivin, a TSA which is highly expressed in glioblastomas, among other cancers types (62, 63). In a phase II trial (NCT024455557), patients with newly diagnosed glioblastoma who received SurVaxM in the adjuvant setting demonstrated a significantly longer 12-month OS of 93.4% from diagnosis, compared with 65% survival from historic studies (64, 65). Interestingly, SurVaxM is now being investigated in a phase I study of patients with survivin-positive neuroendocrine tumors (NCT03879694; ref. 66).

A novel technology platform, T-win, was developed to allow identification, design, and validation of immune modulatory peptide-based vaccine candidates targeting the TME (67). T-win vaccines engage and activate a subset of naturally occurring proinflammatory T cells specific for immune inhibitory molecules, for example, indoleamine 2,3 dehydrogenase (IDO), PD-ligand 1 (L1), PD-L2, arginase, or CCL22 (68, 69). These T cells were initially termed “antiregulatory T cells” (“anti-Tregs”) for their specificity against cells with immunoregulatory functions. These autoreactive T cells, found in high frequencies in patients with cancer, recognize and kill both tumor cells and normal immune cells that express their cognate targets (70, 71), as well as assist in expansion of effector T cells against viral and tumor antigens in vitro (70, 72). Importantly, both CD4⁺ and CD8⁺ T cells also contribute to the immunoregulatory function of anti-Tregs through the secretion of proinflammatory cytokines (73, 74). Thus, these T cells assist the adaptive immune response either through their involvement in the direct elimination of the immunosuppressive cells or through the secretion of proinflammatory cytokines (75, 76). In preclinical models, T-win vaccination has led to an antitumor response and synergizes with anti-PD-1 antibody treatment (77). In mice treated with a T-win vaccine against IDO, a substantial reduction of IDO⁺ immune suppressor cells in the TME was observed, accompanied by an increased expansion of infiltrating tumor-specific T cells (77). Taken together, evidence suggests that T-win vaccination can lead to the expansion of T cells that counteract and modulate the immune suppressive environment within TME, allowing for efficient antitumor responses to take place. Because T-win vaccines aim to expand intrinsic/preexisting T cells in patients with cancer, T-win vaccines do not need to “break tolerance” in the same way as cancer vaccines targeting TAAs (67).

The major challenge of the T-win technology is to activate the most potent anti-Treg immune response without inducing autoimmunity and toxicity. However, circulation of a measurable number of such specific T cells in patients with cancer has been described without autoimmunity; thus, the risk of potential long-term toxicity because of vaccine-induced autoimmune mechanisms appears to be minimal, illustrated in murine in vivo studies and clinical trials to date (67). Table 3 summarizes the T-win technology compounds currently in clinical trials.

With the availability of next-generation sequencing, personalized neoantigen-based immunotherapies are emerging. Sequence data from a patient's tumor biopsy are analyzed to predict which mutations will generate tumor-specific neoantigens likely to be presented by MHC molecules on the tumor cell surface in that patient. Most efforts focus on identifying antigen sequences that generate epitopes fitting the groove of a patient's MHC-I molecules. Although personalized cancer vaccines have shown encouraging results (78, 79), neoepitope prediction algorithms return a large number of “candidates,” of which very few trigger genuine antitumor responses (80). To eliminate the tumor, it is likely to be necessary to target clonal or truncal neoantigens present in every cancer cell. Targeting only subclonal or branch mutations, present in a subset of cells, will not eliminate the tumor and can cause resistance to therapy (81). Interestingly, some have proposed that neoantigens may have inhibitory properties that enable tumors to evade immune detection. A few recently emerging personalized neoantigen vaccines and technologies are summarized here and in Table 4.

EDGE is an artificial intelligence platform used to investigate sequence data from tumor biopsies and identify tumor-specific neoantigens (82). GRANITE-001 is a personalized cancer vaccine based on individual patients' predicted neoantigens, targeting a cassette of 20 patient- and tumor-specific neoantigens identified by EDGE. An ongoing phase I/II clinical study is evaluating GRANITE-001 in combination with CPIs for solid tumor treatment (83). Similarly, SLATE is an immunotherapy directed at the top 20 tumor-specific neoantigens shared by a subset of patients, identified by EDGE. For these patients, an “off-the-shelf” therapy that works across multiple tumor types may be appropriate. An ongoing phase I study is evaluating SLATE in combination with CPIs for solid tumor treatment (84). Both GRANITE-001 and SLATE use a priming adenoviral vector and a self-amplifying RNA vector to deliver the neoantigen cassette in a repeated boost sequence.

ATLAS is another technology platform that uses patient's T-cell immune response machinery to identify optimal patient- and tumor-specific neoantigens (85). By including neoantigens to which patients have had preconfirmed responses in vitro, personalized cancer vaccines are created that the patients' immune systems are already primed for. GEN-009 is being investigated in a phase I/IIa trial for multiple tumor types (GEN-009-101), with positive initial results (86, 87), and GEN-011 is in preclinical development.

A RECON Bioinformatics Engine for prediction and identification of therapeutically relevant neoantigen targets (88) was used to investigate cancer vaccines targeting both patient- and tumor-specific and shared neoantigens (present on the same tumor type in multiple patients). NEO-PV-01, a personalized neoantigen vaccine, custom designed on the basis of unique mutational fingerprints of individual patients, is under investigation in multiple phase Ib clinical trials. NEO-SV-01, an “off-the-shelf” multivalent neoantigen vaccine for treatment of a genetically defined subset of hormone receptor–positive breast cancer, is in preclinical development (88).

Several candidate mRNA-based cancer vaccines are being evaluated in phase I trials, based on a “FixVac” platform (fixed combination of shared cancer antigens; ref. 89). These include BNT111 in metastatic melanoma (90), BNT113 in HPV-positive head and neck cancers, and BNT114 in triple-negative breast cancer. Another mRNA-based cancer vaccine candidate, RO7198457 (BNT122), based on an individualized Neoantigen Specific Immunotherapy (iNeST) platform, is being investigated in combination with pembrolizumab for melanoma (phase II), alone and with atezolizumab for solid tumors (phase I; refs. 91, 92), and with atezolizumab for NSCLC (phase II).

Additional personalized mRNA-based cancer vaccines in phase I testing include (93): mRNA-4157 alone or combined with pembrolizumab in solid tumors (KEYNOTE-603; ref, 56) and NCI-4650 (study now terminated). A vaccine encoding the four most common KRAS mutations, mRNA-5671, is also in phase I testing for patients with KRAS-mutant NSCLC, colorectal cancer, or pancreatic adenocarcinoma (93).

Response to immunotherapy often correlates with high tumor mutation load (94) and consequent higher numbers of predicted neoantigens. Researchers at La Jolla Institute for Immunology (La Jolla, CA) and University of California San Diego (San Diego, CA) are working to identify clinically relevant neoantigens in malignancies of moderate or low mutational burden, for example, head and neck squamous cell carcinoma (HNSCC). Validated neoantigens will be further analyzed in HNSCC tumor models (95). They also plan to explore the role of T-cell exhaustion in mouse and human HNSCC, with a view to being able to counteract this and reinvigorate T cells.

Hilf and colleagues are investigating more effective immunotherapies for low mutational load tumors, by integrating highly individualized vaccinations with unmutated antigens and tumor neoepitopes (96). A phase I trial is investigating novel patient-tailored vaccines, APVAC1 (“off the shelf” glioblastoma-associated peptides) and APVAC2 (de novo synthesized patient-specific glioblastoma-associated tumor-mutated peptides), in glioblastoma (96).

It is an exciting time in the field of therapeutic cancer vaccines, with promising developments in both existing strategies for cancer vaccines and in several new cancer vaccine platforms, antigen targets, and methods to identify them. More research is required before the ultimate goal of personalized cancer therapies can be achieved, but there are currently a wealth of ongoing and upcoming trials in therapeutic cancer vaccine that are expected to lend credence to the value of these strategies. In the move toward personalized cancer immunotherapy, panels of genomic and proteomic biomarkers predictive for response following molecular profiling of tumor and host cells using next-generation sequencing, are expected to further aid decision making and improve outcomes (97).

Overall, cancer vaccines could be the next preferred combination partner for long-term cancer treatment, providing a platform that is easily combined with existing therapies, with minimal toxicity and a good safety profile established in vaccines studied to date.

K.J. Harrington reports personal fees from Amgen, Arch Oncology, ISA Therapeutics, Merck-Serono, Oncolys, and Pfizer, and grants and personal fees from AstraZeneca, Boehringer-Ingelheim, MSD, and Replimune during the conduct of the study (all of the aforementioned were paid to institution). M.-B. Zocca reports personal fees from IO Biotech (CEO and shareholder) and Valo Therapeutics (member, board of directors) outside the submitted work. E. Ehrnrooth reports other from IO Biotech (employment) outside the submitted work. E.E.W. Cohen reports personal fees from BioNTech and Pact Pharma outside the submitted work. No disclosures were reported by the other author.

The authors thank Jane Blackburn and Jacqueline Harte, of Watermeadow Medical, an Ashfield company, part of UDG Healthcare plc (funded by IO Biotech ApS), for medical writing and editing assistance. The authors would also like to thank Ayako Wakatsuki Pedersen, employee of IO Biotech ApS, for assisting the authors in addressing peer reviewer comments in relation to antiregulatory T cells. This work was supported by IO Biotech ApS.