Immunogenicity and Efficacy of Personalized Adjuvant mRNA Cancer Vaccines

The field of mRNA vaccination for cancer and infectious agents concentrates much preclinical and clinical attention, and results on its CD8⁺ and CD4⁺ T-cell immunogenicity against cancer antigens in the clinic are of much interest, such as those reported in this issue of Cancer Discovery by Gainor and colleagues (1). Emerging evidence on mRNA vaccines in oncology speaks of the value of preventing relapse in adjuvant settings upon combination with checkpoint inhibitors, but the clinical efficacy of the approach remains to be seen to successfully treat established or spread solid malignancies.

SARS-CoV-2 mRNA prophylactic vaccines have been instrumental in controlling the COVID-19 pandemic and are being actively tested in the clinic for other infectious diseases. Critically, mRNAs encoding antigens need to be nanoformulated in lipid nanoparticles to achieve intracellular penetration and reach the host ribosomal machinery to express the antigens. Cationic ionizable lipids that allow endosomal escape are critical components of those mRNA vaccines.

Therapeutic cancer vaccines have been very disappointing in terms of antitumor efficacy over the years, despite their ability to induce measurable T-cell responses, in many instances, to shared tumor antigens (2). These antigens, also termed tumor-associated antigens, are aberrantly expressed or overexpressed by tumor cells, but it is often contended that high-affinity T-cell receptors recognizing these “self” antigens are not in the repertoire because of thymic negative selection. For decades, cancer cells were known to harbor exclusive protein sequences as a result of mutational and/or gene-rearrangement processes. However, such individual neoantigens were out of reach until next-generation sequencing and biotechnology tools became available to identify the sequences and synthesize/formulate peptides or mRNAs encoding such neoantigens. Bioinformatics tools have proprietarily improved versions that have become a trick of the trade and a trade secret in the individualized cancer vaccination field, along with the optimized composition of lipid nanoparticles to encapsulate mRNA as the crucial vehicles (Fig. 1).

Whether a mutation encodes an immunogenic antigen is a difficult question. Prediction is based on trained algorithms that consider the affinity of the sequence for the binding groove of MHC antigen-presenting molecules, mRNA expression in the tumor, and other features. However, the best bioinformatics methods allow only a 5% to 30% successful prediction rate, and perhaps the most immunodominant antigens are overlooked and not predicted.

Neoantigen pioneering work in patients with melanoma with long peptide vaccines (3) and mRNAs encoding a string of neoantigen sequences (4) have already reported immunogenicity and signs suggesting efficacy in a few cases of patients with metastatic melanoma. Preclinical and clinical research have shown that vaccine combinations with PD(L)1 checkpoint inhibitors are synergistic.

Established solid tumors are challenging for vaccination because many elements in the tissue microenvironment curtail the efficacy of T-cell antitumor immune responses. However, treating minimal residual disease in postsurgical adjuvant settings is conceivably the best-case scenario to attain efficacy. Weber and colleagues (5) reported a phase II clinical trial in which a favorable comparison was made between the mRNA-4157 individualized vaccine plus pembrolizumab versus pembrolizumab alone in the setting of resected cutaneous melanoma. Moderna’s mRNA-4157 vaccines are repeatedly injected intramuscularly. The results indicated the benefit of the combination, with a hazard ratio of 0.561 (0.309–1.017) and P = 0.053 according to a stratified log-rank test. These data have received a breakthrough designation by the FDA and triggered an ongoing phase III clinical trial (NCT05193377). Interestingly, the putative benefit includes both patients with relatively high and low tumor mutation burdens and seems to be very clear in the small number of patients with residual circulating tumor DNA following surgery.

In parallel, an alliance between BioNTech and Genentech is developing a similar approach in which the vaccine is given intravenously to preferentially reach dendritic cells in the spleen and other secondary lymphoid organs (6). The route of administration is of importance. In the case of mRNA-4157, presumably skeletal muscle fibers are the main cell type expressing antigens, and therefore, cross-presentation/cross-priming by cDC1 dendritic cells will be the most prominent antigen presentation pathway. Upon intravenous delivery of autogene cevumeran, the direct presentation pathway by mRNA-transduced dendritic cells probably dominates (Fig. 1). Both vaccines share the approach of identifying multiple epitopes and formulations in lipids to be effective.

Vaccines need an antigen component, but they also need elements that alert the innate immune system, such as microbial moieties, which are referred to as immune adjuvants. RNAs have intrinsic adjuvant activity as agonists of toll-like receptor 7/8. However, in the case of mRNA-4157, such activity is precluded by substituting uridine with N1-methyl-pseudouridine (1, 5). In contrast, in the case of BioNTech’s autogene cevumeran vaccine, such toll-like receptor 7/8 agonist activity is preserved (6). Notably, lipid nanoparticles themselves exert some level of adjuvant effect, resulting in the release of IL1β (7).

As a consequence of its reduced proinflammatory activity, mRNA-4157 is very well tolerated upon intramuscular injections of 1 mg of mRNA (5). However, much lower doses of the autogene cevumeran vaccine elicit circulating levels of cytokines and a certain degree of systemic inflammatory signs at relatively low tolerable doses in the range of 25 μg. Autogene cevumeran vaccines also induced measurable T-cell responses against approximately 20% of the predicted encoded antigens (6). For autogene cevumeran, clinical results have been reported in the adjuvant setting of pancreatic cancer surgery in 16 patients treated with a vaccine combined with atezolizumab. Very interestingly, those eight patients who showed measurable T-cell responses to the vaccine tended not to relapse, whereas those eight patients who failed to respond, relapsed for the most part (6). These data should be interpreted with caution not only because of the limited patient numbers but also because the results could reflect the overall immunocompetence of the subjects. However, specific immune correlates strongly suggest otherwise.

Both mRNA vaccines seem to result in durable CD4⁺ and CD8⁺ T-cell responses measurable by IFNγ enzyme-linked immunospot (ELISPOT) assays, which can be boosted in the range of 10 to 1,000 spots per 10⁶ lymphocytes in peripheral blood (1, 6). The fact that mRNA vaccines induce T cells with antitumor antigen reactivity in most patients is exciting. However, its clinical efficacy could be restricted to adjuvant/minimal residual disease settings, since its clinical activity in terms of the overall response rate against metastatic disease is weak for the time being (Burris and colleagues, ASCO 2019. Abstract).

The most plausible explanations for the divergence of results in preventing relapse or treating established disease are likely the immune suppression mechanisms of the tumor tissue microenvironment and the lack of proper homing of induced T cells to the tumor. In addition, the levels of T cells in vivo as elicited by the neoantigen vaccines might still be too low to allow for meaningful efficacy against bulky or disseminated disease. Numbers might indeed matter, but it is illustrative that only an 18% overall response rate was achieved in a series of patients with HPV⁺ head and neck cancer treated with an lymphocytic choriomeningitis virus (LCMV)-based E6/E7 vaccine that amazingly attains circulating levels >6,000 IFNγ-spots per 10⁶ lymphocytes (8).

Another interesting point raised by Gainor and colleagues is that most T-cell responses are de novo rather than the result of the expansion of preexisting clones. These findings lead to the concept of tumor-antigen immune ignorance (9), as previously noted by Beatriz Carreño’s group, who vaccinated patients with melanoma with neoantigen epitopes pulsed on autologous dendritic cells (10). Immunologic ignorance refers to the peaceful coexistence of antigens and potentially responsive lymphocytes without evidence for immunization or tolerance induction. The mechanisms behind this puzzling specific immune state are unknown but may imply low levels of antigen expression and a lack of cross-presentation/cross-priming by cDC1 dendritic cells.

Tumor heterogeneity, the loss of antigens, or MHC-loss variants are also likely substantial obstacles for efficacy. Hence, as mentioned before, minimal residual disease seems to be the scenario of choice for these strategies at the current point in time.

The field of personalized antigen–mRNA vaccines holds promise and should focus on advancing the following points:

• a)

  Improving epitope prediction, including immunopeptidomics from resected tumor specimens, to identify the truly presented antigens.

• b)

  Focus on hotspot mutations in genes encoding cancer-related proteins (e.g., KRAS and Tp53) or viral oncogene antigens that are shared by at least a subset of patients. Such antigens offer a number of advantages including feasibility of collections of off-the-shelf vaccines to be used according to the mutations and human leukocyte antigen alleles of each patient. The fact that losing an oncogene-encoded antigen could be detrimental to malignant phenotype of cancer cells is also advantageous.

• c)

  Combinations with agents that promote efficient lymphocyte performance and tumor tissue entry beyond PD(L)1 checkpoint inhibitors.

• d)

  Combinations of neoantigen vaccines with adoptive T-cell therapies.

mRNA is endowed with many advantages as a vaccine platform. The good news for cancer therapeutics is that neoantigens can be immunogenic to a certain degree, resulting in sustained immune responses upon repeated boosting. We still need breakthrough advancements and to meet the manufacturing challenges, but these results concerning immunogenicity are solid foundations on which to build.

M.F. Sanmamed reports grants and personal fees from Roche and Bristol Myers Squibb, and personal fees from MSD and Numab outside the submitted work. I. Melero reports personal fees from Bristol Myers Squibb, F-star, Pierre Fabre, HotSpot, Highlight Therapeutics, Bright Peak, Pioneers, and BioNTech and grants and personal fees from AstraZeneca, Roche, Genmab, PharmaMar, and Boehringer Ingelheim outside the submitted work. No disclosures were reported by the other authors.

We were recently deeply saddened by the news that Dr. Jeffrey Weber passed away from pancreatic cancer. He was a great physician-scientist whose pioneering contributions to immunotherapy will last, including those on mRNA-based cancer vaccines, and whose life will continue inspiring us. We acknowledge funding from ERC Advanced Grant RIPECROP (TLSaRNA 101149181), Gobierno de Navarra Proyecto ARNMUNE (Ref: 0011–1411-2023), and Fundación Fero. A. Tejeira, F. Aranda, and A. Martinez-Riaño are acknowledged for scientific discussions.