Despite the successful vaccination against hepatitis B (HBV) and human papillomavirus (HPV), which are known to cause liver and cervical cancer, respectively, the efficacy of cancer vaccines has been problematic.1,2 Unlike these aforementioned vaccines, cancer vaccines are therapeutic, as opposed to prophylactic, and face three major challenges: (1) historical cancer vaccines typically have a low immunogenicity; (2) tumour mutational burden can ‘out-pace’ and evade the adaptive immune response induced by a vaccine; (3) the tumour microenvironment can be immunosuppressive. Despite these hurdles, the advent of checkpoint inhibitors (CPIs) may provide a means by which tumour mutational pathways can be halted, increasing the efficacy of a vaccine-associated adaptive immune response. Activation of both CD4 and CD8 T cells requires the receptor binding of multiple pathways, and combining the stimulants of these receptors with cancer vaccines also has the potential to propagate a clinically meaningful immune response. Furthermore, the low immunogenicity observed in previous cancer vaccine clinical trials may be due to these vaccines using tumour-associated antigens (TAAs), which although are expressed to a high degree on cancer cells, are not tumour-specific and can be found on healthy cells as well. Self-recognising T-cells are predominantly eliminated during development, which can further lower a vaccines immunogenicity. With this knowledge, vaccine research has improved in recent years, utilising more specific antigens, better vectors and efficacious co-stimulants. These advancements, as well as historical limitations, will be discussed in this review. 3-5
The ideal cancer vaccine would present an antigen to the immune system that is expressed exclusively by cancer cells, and is essential to tumour survival. Additionally, this antigen would be highly immunogenic, inducing the clonal expansion of both CD4+ and CD8+ T cells up to, and past, the concentration threshold needed to result in a high degree of efficacy. Achieving all three of these criteria has proven difficult, although advancements in antigen design has resulted in an increase in efficacy in recent years.
Historically, the majority of cancer vaccines have utilised TAAs, which although are overexpressed on cancer cells, are not specific to a tumour, and can be found on the surface of healthy cells as well. This presents several challenges. Firstly, given that these are self-antigens, B cells and T cells that recognise these antigens are largely eliminated from the immune cell reservoir via central and peripheral tolerance. As a result, vaccines that utilise TAAs are stimulating immune cells that are found in very low concentrations. Too low, potentially, to induce a clinically meaningful response. Despite attempts to amplify response via co-stimulators and booster vaccinations, a response to TAA-utilising vaccines is often seen as stimulating an antigen-specific CD8 T cell population to a level of <1% of the total circulating CD8 T cells. In contrast, the YF-Vax yellow fever and Dryvax smallbox vaccines stimulate clonal expansion of CD8 T cells to 12.5% and 40% of total peripheral CD8 T cells, respectively.6-8 Additionally, immune responses to TAAs can result in collateral damage due to their expression on normal cells. Such is the case with CAR-T cell therapy against the CEA antigen for colorectal cancer, which can result in severe colitis as CEA is expressed on healthy intestinal tissue. 9
Tumour-specific antigens (neoantigens)
In contrast to TAAs, antigens that arise from oncogenic mutation are recognised as foreign by the immune system and are highly immunogenic. Although some common neoantigens have been identified between different cancers, the majority of these antigens are unique and specific to an individual’s own tumour. As a result, the production of a cancer vaccine that utilises neoantigens is a personalised, multi-step process; the patients tumour genome is sequenced, mutations are identified and neoantigens are then predicted in silico using machine learning. With these antigens identified, a vaccine containing these predicted neoantigens is then produced and delivered to the patient. Two phase I studies have utilised this approach in melanoma patients with encouraging results. One study of 13 stage III-IV melanoma patients who received a multi-neoantigen RNA-based vaccine found that 8 patients remained recurrence-free for up to 23 months (the maximum follow-up period). The remaining 5 patients experienced relapses shortly after enrolment, although one patient developed an objective clinical response, with another developing a complete response in multiple progressing metastatic lesions when combined with CPIs, and remained relapse-free for 26 months. 10 Ott et al., also demonstrated the efficacy of a personalised neoantigen vaccine in 6 melanoma patients, finding that 4 of these patients were without recurrence at 25 months post-vaccination, and that the remaining 2 who did progress experienced complete tumour regression when treated with anti-PD-1 therapy. 11 These two studies clearly demonstrate a preliminary efficacy with neoantigen cancer vaccines, as well as demonstrating a potential synergistic action with CPIs. However, promising results in early-phase settings have been observed with other cancer vaccine modalities, only to be found ineffective in later phases. Additionally, the development of a personalised, neoantigen vaccine involves a lengthy production process, by which point the patient may have progressed or developed immune evasion mechanisms. It is also a costly process, as an individual’s tumour has to be sequenced in full. Until this process becomes cheaper, this will inevitably price out some patients, and patients of lower-income countries. Finally, the identification and selection of neoantigens relies on mutational burden, which varies between different cancer types. Cancers that have a higher mutational burden have more neoantigens, increasing the chances of an effective therapeutic vaccine. 12
The lack of efficacy from previous cancer vaccine attempts has identified T-cell response as the main effector of therapeutic vaccines. In contrast, prophylactic vaccines aim to induce deep B-cell responses. Current research now focuses on the development of three types of cancer vaccine vector: cellular vaccines, viral vector vaccines and molecular vaccines, which contain either small peptides, DNA or RNA.13-16
Cellular therapies have in fact been used for many years. Perhaps the most well-known, is Bacillus Calmette-Guerin (BCG), which is used to treat non-muscle-invasive bladder cancer. 17 In theory, bacteria can be used as vectors for the delivery of DNA- or RNA-encoded tumour antigens, which are taken up by antigen presenting cells, such as dendritic cells (DCs). 18 However in practice, conversion of early-phase efficacy to later phase trials has not been easy. Such is the case in advanced pancreatic cancer, which saw promising phase II data with a vaccine containing Listeria-expressing mesothelin (CRS-207), only for these results not to be replicated in a larger phase IIb study. 19, 20 Irradiated allogeneic or autologous tumour cell lines have also been used as vaccine vectors. Vaccines such as these benefit from inactive whole tumour cells, which come complete with presentable neoantigens. As a result, the major advantage with this modality is that neoantigens no longer have to be identified and processed. Unfortunately however, these too have seen limited success, offering only moderate efficacy at best in prostate, lung, and pancreatic cancers, as well as melanoma, despite stimulating immune responses. 21-24 Finally, cellular vaccines containing autologous DCs which have been primed to present neoantigens have been studied extensively. 25 Furthermore, the FDA has approved a DC-based cancer vaccine, sipuleucel-T, for use in metastatic castration-resistant prostate cancer (mCRPC). Approval came following the results of the phase III IMPACT trial, which randomised (2:1) 512 mCRPC patients to receive sipuleucel-T or placebo. A modest, but significant improvement in median overall survival (OS) was observed with the vaccine (25.8 months vs. 21.7 months; HR[95%CI]: 0.78[0.61-0.98], P= 0.03). 26 Unfortunately, sipuleucel-T suffers from the timely complexity of production and cost, limiting its widespread use.
Previously, research into peptide vaccines have utilised short peptides. Less than 15 amino acids long, these peptides do not need to be processed by APCs to effectively bind with major histocompatibility complexes (MHCs). However, without APC presentation, direct binding of peptides to MHCs results in tolerogenic signalling and T cell dysfunction. Crucially however, short peptides do not elicit a CD4+ T cell response, which is essential for the full activation and expansion of CD8 cytotoxic T cells. 27, 28 More recently, synthetic long peptides (SLPs) have demonstrated the induction of both CD8 and CD4 T cell responses through the dual presentation of epitopes on both class 1 and class 2 MHCs. 29
Viral vector vaccines
The human immune system is well adapted to react to a viral infection. As a result, viral vector vaccines are able to elicit both an innate and adaptive immune response via the recognition of viral pathogen-associated molecular patterns by APCs. However, this immune response can often neutralise this viral vector, limiting both the efficacy of the initial vaccine, and any repeat vaccination thereafter. To overcome this, a heterologous prime-boost strategy has been investigated, whereby two different viral vectors are used, administered as a primary and booster vaccine. Unfortunately, this approach also appears to have inconclusive efficacy between phase II and III settings. In 2017, the PROSTVAC-VF vaccine, which was comprised of a vaccinia-based primary and six fowlpox-based boosters was found to induce a statistically significant improvement in median overall survival in 125 men with mCRPC (26.2 months vs. 16.3 months; HR[95%CI]: 0.4997[0.3201-0.7801], P= 0.0019). 30 Conversely, the evaluation of this vaccine was stopped early in a phase III setting, after an interim analysis found it was unlikely to improve OS compared to placebo. 31 In light of these results, a vaccine-based immunotherapy regimen (VBIR) has been constructed, incorporating CTLA-4 and PD-1 CPIs to boost T cell priming and prolong activity. 32 This is currently being evaluated in a phase I setting. 33
DNA and RNA Vaccines
Similar the peptide vaccines, DNA and RNA vaccines are structurally simple and are able to induce DC activation without the need for a signal-boosting adjuvant. Furthermore, they do not provoke a strong anti-vector response, allowing for repeated dosing. Despite this, uptake efficiency by immune cells is low, and techniques to improve uptake have been investigated. Delivery via microneedle arrays, gene gun, nanoparticles, and electroporation have all been found to improve transfection. 34 Such vaccines have found particular efficacy in cervical cancer, with a phase IIb trial of 167 women with grade 2/3 cervical intraepithelial neoplasia (CIN) demonstrating significant tumour regression in 49.5% who received the DNA vaccine VGX-3100 (N= 107), compared to 30.6% who received a placebo (N= 42) (P= 0.034). 35 Additionally, RNA vaccines may be more efficacious, as they do not need to pass through the nuclear membrane, unlike DNA vaccines. Although RNA vaccines are more susceptible to degradation via RNases, their half-life can be prolonged by modifying their chemical structure. 36
To evade immunosurveillance, cancers often employ several oncogenic and immunosuppressive pathways at once. In recent years, CPIs have transformed the treatment paradigm for many cancers, and preclinical studies have demonstrated the synergy between CPIs and therapeutic cancer vaccines. 37 More recently, a phase II study offered encouraging data with a CPI-vaccine combination, finding an overall response rate of 33% in 24 patients with incurable HPV-16-positive cancer, when given the long-peptide ISA101 vaccine in combination with nivolumab. 38 Unfortunately, not all CPI-vaccine combinations have been fruitful. In a phase III study by Hodi et al., the addition of the short peptide gp100 vaccine to ipilimumab did not improve survival outcomes in melanoma patients. 39 As previously mentioned, an optimal immune response relies on both activation of CD4+ and CD8+ T cells. Neoantigen presentation via APCs is not enough to elicit a full signalling cascade, and co-stimulators are required to activate CD8 cytotoxic T cells. Normally, these costimulators are received from APCs, particularly DCs, as well as from CD4 T helper cells. With this in mind, co-stimulators can be combined, in theory, with cancer vaccines to further propagate a T cell response. Studies in mice have confirmed this synergy, showing that OX40 antibodies enhanced CD4 and CD8 T cell responses in mice models. 40, 41 Taking this approach further, triple combinations of vaccines, CPIs and costimulators have also been investigated. An adenoviral vaccination in combination with anti-CD40 and anti-CTLA-4 monoclonal antibodies was able to induce complete regression and long-term survival in 30-40% of mice. 42 In another mice study, when combined with a HER2-directed therapeutic vaccine, OX40 agonism and CTLA-4 blockade induced robust CD4 and CD8 T cell responses, as well as extensive tumour destruction and T-cell infiltration. 43 Interleukin-2 (IL-2) is a key co-stimulator of cytotoxic T cells, and has also been investigated as a potential cancer vaccine adjuvant. In an encouraging development, a gp100-like vaccine was found to significantly improve progression-free survival when combined with IL-2, compared to IL-2 alone in 185 patients with stage III or IV melanoma in a phase III trial (2.2 months vs. 1.6 months; P= 0.008), as well as numerically improve median overall survival (25.8 months vs. 11.1 months; P= 0.06). 44
Lastly, chemo/radiotherapy-vaccine combinations have also been investigated. In particular, radiotherapy has been shown to induce immunogenic tumour cell stress, resulting in enhanced T cell response and synergy with cancer vaccines. 45, 46 This is a relatively new area of research that is still being developed. 47 The synergistic relationship between chemotherapy and cancer vaccines is also well documented. 48 Perhaps the most compelling evidence of this relationship comes from a phase IIb study, which found that TG4010, an MUC-1- and IL-2-expressing, modified Ankara viral vaccine, was able to induce a longer progression-free survival and a higher rate of confirmed responses when combined with platinum-based chemotherapy, compared to chemotherapy alone (mPFS: 5.9 months vs. 5.1 months; HR[95%CI]: 0.74[0.55-0.98], P= 0.019).
In closing, although historically unremarkable, emerging research around therapeutic vaccines is encouraging. Learning from previous failures, this area of research now has many antigen and vector combinations to explore, with even greater potential efficacy when combined with other co-stimulants and systemic therapies. However, in the majority of cases, this therapy is far from clinical realisation. In the case of cancer vaccines that utilise neoantigens or whole cells, cost and manufacture time are major hurdles that make this treatment unfeasible for many patients. Furthermore, not all patients respond to immunotherapy, and it is likely that cancer vaccines will not be applicable to everyone. Predictive biomarkers will help to personalise this treatment modality in the future and ultimately, more research is required to establish vaccine designs that can significantly impact survival outcomes in a phase III setting.