Is It Possible to Develop Cancer Vaccines to Neoantigens, What Are the Major Challenges, and How Can These Be Overcome?

Abstract

The term “neoantigen,” as applied to molecules newly expressed on tumor cells, has a long history. The groundbreaking discovery of a cancer causing virus in chickens by Rous over 100 years ago, followed by discoveries of other tumor-causing viruses in animals, suggested a viral etiology of human cancers. The search for other oncogenic viruses in the 1960s and 1970s resulted in the discoveries of Epstein–Barr virus (EBV), hepatitis B virus (HBV), and human papilloma virus (HPV), and continues until the present time. Contemporaneously, the budding field of immunology was posing the question can the immune system of animals or humans recognize a tumor that develops from one’s own tissues and what types of antigens would distinguish the tumor from normal cells. Molecules encoded by oncogenic viruses provided the most logical candidates and evidence was quickly gathered for both humoral and cellular recognition of viral antigens, referred to as neoantigens. Often, however, serologic responses to virus-bearing tumors revealed neoantigens unrelated to viral proteins and expressed on multiple tumor types, foreshadowing later findings of multiple changes in other genes in tumor cells creating nonviral neoantigens.

Great Debates

What are the most interesting topics likely to come up over dinner or drinks with your colleagues? Or, more importantly, what are the topics that don't come up because they are a little too controversial? In Immune Memory and Vaccines: Great Debates, Editors Rafi Ahmed and Shane Crotty have put together a collection of articles on such questions, written by thought leaders in these fields, with the freedom to talk about the issues as they see fit. This short, innovative format aims to bring a fresh perspective by encouraging authors to be opinionated, focus on what is most interesting and current, and avoid restating introductory material covered in many other reviews.

The Editors posed 13 interesting questions critical for our understanding of vaccines and immune memory to a broad group of experts in the field. In each case, several different perspectives are provided. Note that while each author knew that there were additional scientists addressing the same question, they did not know who these authors were, which ensured the independence of the opinions and perspectives expressed in each article. Our hope is that readers enjoy these articles and that they trigger many more conversations on these important topics.

Use the right word, not its second cousin.

—Mark Twain

Groundbreaking discoveries in the early 1970s, starting with the identification of the src oncogene, showed that cancer was a genetic disease and that abnormal expression (activation or suppression) of proto-oncogenes could cause malignant transformation. Products of mutated proto-oncogenes became candidate tumor neoantigens (Hellstrom and Hellstrom 1989; Urban and Schreiber 1992). Simultaneous development of molecular techniques and cellular assays to evaluate T-cell responses in vitro identified numerous mutated neoantigens, products of single-point mutations in genes encoding proteins directing various cellular functions newly expressed and unique to each individual mouse and human tumor (Sibille et al. 1990; Lennerz et al. 2005; Sensi and Anichini 2006; Coulie et al. 2014). These data recapitulated results from early work on mouse carcinogen-induced tumors that identified unique mutations generating neoantigens that elicited immunity and protected mice from challenge with the original tumor but no other tumors caused by the same carcinogen (Srivastava 2015).

All of these neoantigens (viral, oncogene-encoded, or randomly mutated) were the predicted targets of cancer immunosurveillance, a function of the immune system that was at that time very much in doubt and for which their identification provided an indisputable support (Dunn et al. 2004). What was not predicted when the search for tumor antigens began was the finding of tumor-specific humoral and cellular immune responses in mice and in cancer patients, which recognized nonmutated cellular proteins as specific antigens on tumor cells and not on normal cells (Sjogren 1967; Finn 1993; Henderson and Finn 1996). Because these were present on multiple tumors rather than being unique to each tumor, they were given a designation “shared tumor antigens” (Ting and Herberman 1974). Because they were tumor-associated versions of proteins expressed by normal cells and thus not theoretically “tumor-specific,” they were also named tumor-associated antigens (TAAs). Over the last two decades, many TAAs have been identified and the nature of tumor-specific changes in normal cellular proteins that created these neoantigens was elucidated. Shared neoantigens belong to one of several categories: oncofetal antigens (expressed in fetal but not adult tissues and reexpressed in tumors), cancer/testis (CT) antigens (present in germ cells that lack major histocompatibility complex [MHC] and not presented to the immune system except on tumor cells), differentiation antigens (specific to differentiated tissues and organs from which the tumor originated), overexpressed antigens and differentially processed antigens (tumor-specific changes in protein glycosylation [Vlad and Finn 2004; Vankemmelbeke et al. 2016], phosphorylation [Mohammed et al. 2008], and citrullination [Brentville et al. 2016]), among others.

In summary, over a period of more than half a century, a long list of neoantigens has been compiled, which regardless of their origin (viral, mutated, nonmutated) share the same characteristics: (1) they are newly and preferentially present on tumor cells; (2) recognized by antibodies and T cells; and (3) elicit spontaneous immunity in cancer patients and tumor-rejection immunity in animal models. It is thus unjustified that the term “neoantigens” has recently been usurped specifically for products of mutated gene segments uncovered by whole exome sequencing or by mass spectrometric analysis of MHC/human leukocyte antigen (HLA)-bound peptides (Bassani-Sternberg et al. 2015; Kalaora et al. 2016). This misuse of terminology hides the nature and immunogenic potential of other neoantigens mentioned above, which have already been shown to be recognized by the immune system, to induce tumor-rejection immunity in animal models, and to have clinical benefit in cancer patients either as targets of immunotherapy or as vaccine antigens (Cheever et al. 2009). To use the right word and not its second cousin, which is especially important in science communication, what are recently being referred to as “neoantigens” should be referred to as “mutated neoantigens” in deference to other members of a large and varied family of tumor neoantigens.

CANCER VACCINES BASED ON MUTATED NEOANTIGENS CAN BE DEVELOPED BUT DO NOT WARRANT SPECIAL ATTENTION

Because it is now relatively easy to sequence genes, a cottage industry has sprung up around identifying and cataloging hundreds to thousands of mutations in cancers, each potentially a candidate neoantigen for an individual tumor/individual patient-specific vaccine, or a target for immunotherapy (Boegel et al. 2014; Yadav et al. 2014). Those who are in the field of tumor immunology trying to better understand tumor immunity and immunosurveillance appreciate the challenges brought about by tumor-specific/patient-specific mutated epitopes (Lutz and Jaffee 2014; Gubin et al. 2015) and their use for immunotherapy or vaccines. The major challenges include (1) how to select from a large number of mutations those few that are made into proteins and processed and presented as antigenic peptides in MHC class I or class II antigens; (2) how to show convincingly that immune responses against mutated neoantigens are superior to nonmutated neoantigens; (3) how to deal with MHC/HLA restriction and further mutations that can generate antigen-escape variants; and (4) how to obtain evidence, other than through single-cell sequencing, that specific mutations are present in all tumor cells, thus not allowing immune escape (Verdegaal et al. 2016).

Some of these challenges are the same for viral and nonmutated neoantigen vaccines but others are not. For example, every tumor cell in a virally induced tumor expresses viral neoantigens required for continued oncogenic transformation and thus each tumor cell is a target of a vaccine based on these neoantigens (McAllister 1965; Javier and Butel 2008). Similarly, many nonmutated neoantigens, but especially those that have oncogenic functions (Cheever et al. 1995; Bright et al. 2014), such as Her-2neu (Disis et al. 1994), MUC1 (Cheever et al. 1995), hTERT (Vonderheide 2002), and p53 (Pedersen et al. 2011), are expressed in all tumor cells and on multiple tumor types, which makes them optimal candidates for vaccines and immunotherapy. Short of sequencing single cells from various sections of primary tumors and metastatic sites, the same cannot be assumed for mutated neoantigens. Furthermore, many nonmutated neoantigens trigger immune responses because of their overexpression in tumors compared to normal cells. Protein abundance and protein turnover are important factors for HLA presentation of antigens (Bassani-Sternberg et al. 2015). Identification of mutated peptides as potential neoantigens currently requires a combination of exome sequencing, messenger RNA (mRNA) microarrays, and epitope prediction algorithms, but also knowing the level of expression of the source protein, which determines whether it can reach the threshold required for its efficient processing and presentation in HLA.

MUTATED VERSUS NONMUTATED TUMOR PEPTIDES AS HLA LIGANDS

Rammensee and colleagues have characterized immunopeptidomes of a number of primary tumors and cells, including leukemias and solid tumors (Walz et al. 2015; Löffler et al. 2016), by eluting and sequencing peptides from purified HLA class I and class II molecules (Berlin et al. 2015). With the current sensitivity of the mass spectrometry methods, 5000 unique nonmutated peptides can be isolated from 1 g of tissue. Hundreds of nonmutated peptides identified from each sample appear to be tumor specific on the grounds that they are not found on any of the numerous normal tissues similarly analyzed. Many of these peptides are immunogenic, as tested by in vitro priming of T cells from healthy donors or by measuring recall T-cell responses from patients. In chronic lymphocytic leukemia (CLL) patients, such T-cell responses correlated with patients’ overall survival (Kowalewski et al. 2015a).

By combining exome sequencing, prediction of mutated peptides as HLA ligands, and their verification by mass spectrometry, a few mutated peptides were also identified; however, only a very small fraction of mutations at the DNA sequence level ended up as peptides in HLA class I. Based on their own mass spectrometry data and the published data on the expression of mutations in proteins and HLA ligands, Rammensee and colleagues (Löffler et al. 2016) estimate that from 1000 nonsynonymous DNA mutations, 10 manifest in mutated proteins and only one as a mutated HLA class I ligand (for illustration, see Fig. 1). Assuming also that the tumor proteome is cross-presented on HLA class II molecules on tumor-associated antigen-presenting cells, the frequency of mutated HLA class II ligands is estimated to be close to the number of mutated proteins, 10/1000. These numbers are roughly in accordance with clinical observations on mutation-specific T cells from patients treated with checkpoint blockade (Schumacher and Schreiber 2015).

Figure 1.

Frequency of mutated neoantigens presented in major histocompatibility complex (MHC) class I and cross-presented in MHC class II. APC, Antigen-presenting cell; HLA, human leukocyte antigen.

In contrast, the number of nonmutated but still highly tumor-specific peptides is much higher with several dozens of such peptides identified in a given tumor tissue, most of them present in one particular tumor and not in most other tumors or normal tissues. One hypothesis is that a mutation anywhere in a protein would lead to changes in its expression and likely expression of downstream proteins in the same functional pathway, thereby influencing their processing and consequently a difference in the tumor immunopeptidome. For example, a mutation in a member of the WNT-signaling pathway (MacDonald et al. 2009) increases levels of cyclin D1, which in turn increases the number of its HLA ligands and leads to their recognition as tumor-specific, nonmutated neoantigens. Peptides from nonmutated cyclin D1 have been shown to be immunogenic and are already being tested in cancer immunotherapy (Walter et al. 2012; Löffler et al. 2016). Another example is cyclin B1, a shared nonmutated tumor antigen that is overexpressed in human tumors where p53 is either deleted or mutated (Kao et al. 2001; Yu et al. 2002). The nonmutated but overexpressed cyclin B1 serves as a tumor-rejection antigen in mouse models (Vella et al. 2009) and elicits antibodies and T cells in cancer patients (Suzuki et al. 2005). Mutation-induced alterations can affect mRNA expression levels, RNA splicing, and other posttranscriptional or posttranslational modifications, resulting in many tumor-specific MHC ligands as nonmutated neoantigens (Kowalewski et al. 2015a,b).

IMPORTANCE OF CHOOSING THE RIGHT NEOANTIGEN(S) FOR EFFECTIVE CANCER VACCINES

A recent explosion in the number of mutated neoantigens and the enthusiasm for their potential as cancer vaccines should not distract from the fact that mutated neoantigen vaccines are not a new concept and have been made and tested before. This includes vaccines based on mutated H-ras (Gjertsen and Gaudernack 1998) and K-ras oncogenes (Carbone et al. 2005), mutated p53 (Carbone et al. 2005; Vermeij et al. 2011), and heat-shock proteins that bind mutated tumor peptides (Srivastava 1993). These vaccines have had as much or as little success as vaccines based on nonmutated overexpressed antigens, CT antigens, or differentiation antigens. The rationale for their development as vaccines was that they were tumor-specific and foreign to the immune system and should induce stronger immune responses than nonmutated neoantigens that might be subject to immune tolerance. This predicted difference between mutated and nonmutated neoantigens did not materialize, nor did the immunogenic superiority of mutated antigens. A good example is p53 where vaccines based on patient- and tumor-specific p53-mutated neoepitopes achieved the same results as vaccines based on nonmutated peptides derived from overexpressed wild-type p53 (Vermeij et al. 2011). It is clear now that the immunosuppressive tumor microenvironment (Palucka and Coussens 2016) can have a much greater effect on vaccine immunogenicity and efficacy than the nature of the vaccine antigen. Given past experience with mutated neoantigens, it is hard to justify the labor- and cost-intensive development of personalized vaccines based on these antigens. It is more than likely that to be more effective in the therapeutic setting, all cancer vaccines will need help from currently available or soon-to-be-developed immunotherapies directed at modulating the tumor microenvironment (Kourie et al. 2016). If these immunomodulators are successful, both nonmutated and mutated neoepitope vaccines will experience a renaissance.

NEOANTIGENS AS PROPHYLACTIC CANCER VACCINES?

In addition to combination therapies designed to enhance efficacy of therapeutic cancer vaccines, some cancer vaccines are already being tested for increased efficacy in individuals without cancer but at an increased risk for cancer. The hypothesis is that in the absence of tumor the immune system is not compromised and effective immunity and immune memory can be elicited to protect from or delay tumor development (Finn 2014). Initial trials are based on several well-known shared tumor antigens that have been extensively studied and thoroughly characterized for their expression on tumors (Finn and Beatty 2016). One of these is MUC1 that is overexpressed in its hypoglycosylated form on all human adenocarcinomas as well as on multiple myeloma and some leukemias and lymphomas. The expectation is that the MUC1 vaccine would elicit or boost strong immune responses and long-term immune memory to prevent cancer development. Finn and colleagues are vaccinating individuals diagnosed with advanced adenomas of the colon, precursors to colon cancer. The vaccine is given postadenoma removal and a booster is administered a year later. Very strong anti-MUC1 immune responses were induced in the initial study that greatly surpassed frequency and intensity of responses seen in cancer patients. Importantly, the vaccine induced immune memory with no evidence of toxicity (Kimura et al. 2013). Vaccine-elicited antibodies had a range of affinities and reacted only with MUC1 on tumors and not on normal tissues (Lohmueller et al. 2016). Inasmuch as the nonmalignant adenomas also overexpress hypoglycosylated MUC1, this vaccine is expected to prevent recurrence of premalignant lesions or their progression to cancer. This is being tested in an ongoing randomized trial.

The absence of cancer as a source of mutated neoantigens would appear to exclude the possibility of developing personalized prophylactic cancer vaccines. Whereas this is the case in the true prevention setting in the complete absence of disease, it does not apply to the setting of premalignant disease (Finn 2003). Most human adenocarcinomas start with readily diagnosed premalignant lesions that can be biopsied and subjected to exome sequencing to identify specific mutations and mutated neoantigens. Efforts are already underway to better understand the molecular events in cancer development from premalignant to malignant disease, with the goal of creating a Pre-Cancer Genome Atlas (PCGA) to parallel efforts on The Cancer Genome Atlas (TCGA) (Campbell et al. 2016). This effort is expected to uncover new targets for cancer prevention (Kensler et al. 2016). Importantly, it would also set the stage for routinely biopsying and sequencing premalignant lesions, which could support development of patient-specific mutated neoantigen vaccines for cancer prevention. The same issues and challenges that apply to the mutated neoantigens of therapeutic vaccines would apply to the prophylactic vaccines, including challenging the wisdom of their development.

CONCLUSION

Sophisticated high-throughput methodologies have been developed for identifying and cataloging genetic mutations in tumors, and these studies as applied to tumor immunity will provide new information about immunosurveillance of genome integrity and its targets (Fritsch et al. 2014; Tran et al. 2015; Gros et al. 2016). Although mutated neoantigens uncovered in this process already have their proponents for immediate clinical application as targets for immunotherapy or antigens in vaccines (Gros et al. 2016; Stronen et al. 2016), the labor-intensive and likely to be a very expensive approach to their identification and vaccine development should not replace or even compete with the development of more broadly applicable, off-the-shelf, cost-effective vaccines based on nonmutated, shared tumor neoantigens.