The Neo-Anti-Gen Era introduces personalized cancer vaccines
Neo-antigen vaccines- as an evolution from the shared-antigen (tumor associated antigens, TAA and hot-spot antigens) vaccine approaches- hold a 'hot topic' status since 2016 and potentiating throughout last year, moving the field avidly into personalized cancer immunotherapies. Personalized cancer vaccines are therapeutic vaccines custom tailored to target tumor-specific mutations unique to a given patient.
State-of-the-Biotech/Pharma
Just earlier this month (May, 2018) Merck & Co. announced partnering on mRNA-5671, Moderna’s mRNA KRAS cancer vaccine, and other shared antigen (so-called “hotspots,” in tumor driver genes) mRNA cancer vaccines. Moderna and Merck have agreed to jointly advance mRNA-5671 through human clinical studies, including combination studies with Merck’s blockbuster anti-programmed cell death protein 1 (PD-1) antibody Keytruda (pembrolizumab).
The shared antigen approach has been tried before, though unsuccessfully (as reported in January, 2017) by CureVac: in the Phase IIb proof-of-concept study RNActive prostate cancer vaccine CV9104, composed of six protamine-stabilised mRNAs encoding individually antigens, failed to improve overall and progression-free survival in patients with asymptomatic or minimally symptomatic metastatic castrate-resistant prostate cancer vs placebo. Together with the Moderna, and BioNTech AG, which partnered its IVAC platform with Genentech (see below), Curevac is one of three of the main players in the mRNA–based vaccine field and is eying a potential deal with Eli Lilly to develop five mRNA cancer vaccines, and collaborating with Boehringer Ingelheim in non-small cell lung cancer.
In big contrast, already in 2016 BioNTech published the first example of a clinically applicable and systemic mRNA cancer vaccine. They demonstrated that RNA-LPX encoding viral or mutant neo-antigens or endogenous self-antigens induce strong antigen-specific T-cell responses in 3 melanoma patients. With these data adding to evidence that IVAC Mutanome may lead to personalized, broadly applicable cancer vaccines, Genentech has bought into merits of BioNTech’s process and partnered on its mRNA cancer vaccine platform. This was followed last year (July, 2017) by BioNTech's announcement of Phase I trial results of the first-in-human application of a personalized RNA-based vaccine based on patient-specific mutations (IVAC® MUTANOME) demonstrating induction of strong immunogenicity as well as promising anti-tumor activity in high-risk patients with late-stage melanoma (see side bar).
In 2016 Amgen licenced the rights to Advaxis' lead candidate ADXS-NEO (an antigen delivery technology using bacteria -Lm technology- to deliver multiple neoantigens), which was followed by FDA accepting ADXS-NEO IND in March 2017 and the program entering Phase I.
Advaxis plans to file two new IND applications with the FDA, ie. ADXS-HOT leverages Advaxis’ Lm Technology 1. to target tumor driver genes by the developed library of constructs targeting hotspots that could be available to patients following a rapid diagnostic test that does not require sequencing, and 2. in collaboration with SELLAS Life Sciences to use patented WT1 targeted heteroclitic peptide antigen, which is highly expressed in most tumor types.
Aduro Biotech recently (European Neoantigen Summit, Amsterdam, April 2018) presented early observations from its personalized neoantigen-based pLADD therapy Phase I proof-of-concept clinical study, designed to evaluate the safety and tolerability of pLADD immunotherapy in adults with metastatic colorectal cancer that is microsatellite stable (MSS).
The use of personalized neoantigen vaccines in clinical trials was pioneered nearly a decade ago by Wu, Hacohen, and other colleaques and the Dana Farber Cancer Institute (DFCI) continues to be steadily involved in such clinical trials using synthetic long
peptides (see the side bar for their recently published promising clinical findings). Based on the DFCI platform, Neon Therapeutics (co-founded by Wu, Hacohen, Schreiber, Schumacher, Lander, Allison and Fritsch) got the go-ahead for an open-label Phase Ib clinical trial pairing its neoantigen-targeting peptides in NEO-PV-01 (with an alternative name NeoVax) with Bristol-Myers Squibb's checkpoint inhibitor Opdivo (nivolumab) to assess the safety and efficacy in an estimated 90 patients with melanoma, bladder, or lung cancer. The trial also aims to assess whether NEO-PV-01 plus Opdivo will have a greater clinical benefit than Opdivo alone. The trial is currently recruiting participants at sites across the U.S. and is expected to be completed by December 2020.
Additional players in the field are:
- Gritstone Oncology, planning an IND application for mid-2018 for a phase I/II clinical trial of a personalized neoantigen immunotherapy in combination with checkpoint inhibitors in patients with different tumor types in the adjuvant and metastatic settings
- ISA Pharmaceuticals with their preclinical program MyISA® (personalized immunotherapy based on neoantigens)
- Agenus, currently evaluating AutoSynVax (vaccine made by combining synthetic versions of each patient’s own neo-antigens and heat shock protein 70 (HSP70)) in an open-label, single-arm Phase I clinical trial (ClinicalTrials.gov identifier NCT02992977) in about 20 people with very advanced solid tumors, whose cancers are no longer responding to standard treatment
- Vaccibody, announcing approval of clinical trial application for its cancer neoantigen phase I/IIa trial with their individualized neoantigen vaccine VB10.NEO, evaluating the safety, feasibility, and efficacy of VB10.NEO in combination with standard of care checkpoint inhibitor therapy. The clinical trial will enroll patients with locally advanced or metastatic non-small cell lung cancer, melanoma, renal, bladder, and head&neck cancer. A total of 40 patients are planned to be enrolled in the phase I part of the trial.”
Clinical Trials with personalized cancer vaccines
Clinical trials of neoantigen-based personalized cancer vaccines are taking place in multiple tumour types and are using various adjuvants and delivery approaches, including peptide-based vaccines for breast, pancreatic, (paediatric) brain and hepatocellular cancers (ClinicalTrials.gov identifiers: NCT02287428, NCT02427581, NCT02600949 and NCT03068832), poly-epitope-encoding RNA-based or DNA-based vaccines for advanced digestive system neoplasms (Stemirna Therapeutics), breast and pancreatic cancers (ClinicalTrials.gov identifiers: NCT03468244, NCT02316457, NCT02348320 and NCT03122106) and a peptide-loaded DC vaccine for colorectal cancer (ClinicalTrials.gov identifier: NCT01885702). Other ongoing projects include the Glioma Actively Personalized Vaccine Consortium (GAPVAC) and the Cancer Vaccine development for Hepatocellular Carcinoma Consortium (HEPAVAC), which are conducting phase I clinical trials using off-the-shelf peptides matched to highly overexpressed antigens in the tumours of individual patients as well as personalized mutated peptides in glioblastoma (ClinicalTrials.gov identifier: NCT02149225) and hepatocellular cancer (Hu et al. 2017).
Further, clinical trials are ongoing to assess the neoantigen vaccines in combination with
locally administered ipilimumab (NEO-PV-01, for renal cell carcinoma in collaboration with BMS and Oncovir, ClinicalTrials.gov identifier: NCT02950766) and as mentioned above systemic nivolumab (NEO-PV-01, Neon Therapeutics in collaboration with BMS, ClinicalTrials.gov identifier: NCT02897765) as well as atezolizumab (PGV001, for urothelial cancer in collaboration with Genentech, ClinicalTrials.gov identifier: NCT03359239 and RO7198457, in collaboration with BioNTech, ClinicalTrials.gov identifier: NCT03289962). NEO-PV-01 is also being evaluated following front-line rituximab in follicular lymhoma (ClinicalTrials.gov identifier: NCT03361852) as well as in combination with low-dose cyclophosphamide (for CLL in collaboration with Oncovir and Neon Therapeutics, ClinicalTrials.gov identifier: NCT03219450).
Hot-spots vs Private Neo-antigens
Neo-antigens differ from self-antigens (cancer-testis, differentiation-specific or lineage-specific, overexpressed, post-translationally modified) in that they are mutated. As such, they are not subjected to central tolerance due to developmental negative T cell selection (eliminating every T cell recognizing self antigen). Mutated antigens can be divided into mutated neoantigens encompassing random somatic mutations in individual tumor (private) and mutated oncogenes being products of common somatic mutations or gene translocations, as well as post-translational modifications. Klebanoff and Wolchok (2017) elaborate private vs public neoantigens:
Newly created antigens resulting from cancer-specific mutations (SNPs, frame-shift, alternative splicing, fusion proteins) or neoantigens, pose an unprecedented challenge to developing antigen-specific immunotherapies. The human exome is ∼30 megabases in size. Consequently, the chance that any single random somatic mutation will recur in more than one patient is exceedingly small. This fact, combined with the requirement that a mutated protein can only be detected by T cells if it is processed in the proteasome and presented by one of the patient’s complement of HLA molecules, effectively means that most neoantigens are patient specific. Immunotherapies that seek to raise an antigen-specific immune response to such “private” neoantigens must therefore be customized for each individual patient, creating substantial practical and regulatory hurdles.
However, not all somatic mutations occur at random. Mutations that alter protein function to promote oncogenesis, so-called driver mutations, can systematically reappear across patients. Further, these function-altering mutations typically occur in tightly constrained hotspot regions within a protein, with only a single or limited number of amino acid residues substituted that cause altered function. If a peptide containing a hotspot mutation is bound by a relatively common HLA allele, an ideal “public” neoantigen shared across patients would be created. Given the extraordinary precision with which a peptide binds the groove of an HLA molecule, discovering a “public” neoantigen is not straightforward, but definitely not impossible either (Chheda et al. 2018).
Table 1. Shared peptide antigens have been identified which can be used to broadly vaccinate patients of the same cancer type when those patients commonly express that antigen (top). Vaccines incorporating personalized neoantigens cater specifically to that individual patient or tumor, and so are promising targets to activate antitumor immunity (bottom) (Aldous et al., 2017)
Neoantigen Identification and Selection
To help coordinate the efforts (Nielsen et al., Bassani-Sternberg et al., TRON, Oncolmmunity, Personalis, Caprion Biosciences, Immatics Biologics) and accelerate the search for tumor neoantigens the Parker Institute for Cancer Immunotherapy and the Cancer Research Institute (CRI) launched a public-private partnership- the Tumor Neoantigen Selection Alliance (TESLA) in December of 2016. This global bioinformatics collaborative includes scientists from more than 35 of the leading neoantigen research groups in academia, nonprofit and industry. The goal is to find the best algorithms to predict which cancer neoantigens encoded in DNA and RNA can be recognized by and stimulate an immune response.
Samples from cancerous and normal human tissue and blood (with initial focus on focus on advanced melanoma, colorectal cancer and non-small cell lung cancer) will be sequenced. Research teams at participating institutions will be provided the DNA and RNA sequences. Using their own algorithms and methods, participants will use that information to generate a list of predicted neoantigens that can both be recognized by the immune system and cause it to respond. Those predictions will be confirmed in laboratory assays that will test how well the antigens are recognized by the immune system and how much they stimulate T-cells. Each participant will be provided with data to inform and improve their algorithms.
Neo-antigen prioritization is absed on HLA binding strength/stability, presentation by both HLA class I and II, the expression level and clonality, similarity to self and structural change, differential binding, in vivo processing and presentation.
Neoantigen identification (in silico epitope prediction algorithms, MS datasets, large-scale immunogenicity assessment)
Major histocompatibility complex (MHC) molecules capture peptide antigens (MHC class I bind 9-mer peptides-different haplotypes of MHC molecules bind distinct peptides and MHC class II bind 9-18 peptides) for display on the cell surface. These peptide-MHC complexes are recognized by T cells via their T-cell receptors, which results in T-cell recognition being restricted to those peptides that a MHC molecule can present. Interestingly, Pearson et al. (2016) demonstrated that the entire MHC class I–associated peptides (MAPs) repertoire (ie. immunopeptidome) covers only about 10% of the exomic sequences, while 40% of expressed protein-coding genes generate no MAPs.
Computational pipelines have been generated to identify personal candidate
neoantigens in real time. Comprehensive mutational analysis is carried out through whole-exome sequencing (WES) and neoepitopes encoded by somatic mutations
in the tumour are selected that have the highest probability of being presented by the individual’s MHC molecules on the basis of affinity predictions.
Using the current prediction tools, 95% of known CD8+ T cell targets are identified at a specificity of 99%, meaning that for an antigen the epitope will be in the top 1% best predicted peptides in 95% of all cases. [Compared to class I binders, the algorithms for class II give rather high percentage of false predictions and MS elution data are less reliable due to to the relatively higher promiscuity of peptide binding to MHC class II molecules]. There are several different approaches in neo-antigen discovery, the most mainstream one being pipelines/algorithms for neo-epitope prediction from tumor sequencing data. Conceptually, neo-epitope prediction based on next-generation sequencing (NGS) data can be divided into three steps: (a) convert a list of genomic mutations into protein sequences, and retain all possible mutation-containing “neo-peptides” of appropriate lengths; (b) predict binding to the patient-specifc HLA alleles; and (c) assess the immunogenicity of each peptide based on features such as predicted binding, expression level and sequence similarity to unmutated self-proteins.
There are 3 different pan-allele binding prediction approaches: artificial neural network, network model and linear model. NetMHCpan2.4 is currently the most widely used HLA-binding tool.
Eg. MuPeXI, the mutant peptide extractor and informer, is a program to identify tumor-specific peptides and assess their potential to be neo-epitopes. The program input is a file with somatic mutation calls, a list of HLA types, and optionally a gene expression profile. The output is a table with all tumor-specific peptides derived from nucleotide substitutions, insertions, and deletions, along with comprehensive annotation, including HLA binding and similarity to normal peptides. The peptides are sorted according to a priority score which is intended to roughly predict immunogenicity (Bjerregaard et al. 2017).
Table 2. Features of current neo-epitope prediction tools (Bjerregaard et al. 2017).
Current algorithms (see Aldous et al. 2017 for a review of algorithm development for personalized neoantigen identification/prediction) use only the binding affinity (or better affinity ranking - Nielsen and Andreatta, 2016) of putative neoantigens to HLA (see compendium), however predicting the processing (including immuno-proteasome cleavage site), TAP transport and T cell propensity are all important for identification of peptides capable of eliciting a T-cell response. In fact, only 15% - 20% of “predicted” peptide binders are processed or presented, and therefore contribute to the immunopeptidome. Erroneous predictions may be addressed with mass spectrometry (MS).
Mass spectrometry (MS) eluted ligands data sets aid significantly to validating and interpreting the in silico predicted neo-antigens and help guide the algorithm(s) improvements to correct for the effects of the protein abundance and turnover on presentation. [To deconvolute multiple specificities in large-scale peptidome data generated by MS, GibbsCluster 2.0 presents a powerful tool for unsupervised motif discovery because it can simultaneously cluster and align peptide data (Andreatta et al., 2017).]
While the MS information about the eluted ligands capture additional aspects about the peptides compared to binders from the prediction algorithms, MS-based immuno-peptidomics likely favors detection of the abundantly presented neo-antigens.
Bassani-Sternberg et al. (2016) published that direct identification of mutated peptide ligands from primary tumour material by MS is possible and yields true neoepitopes.
Figure 1. Overview of the experimental approach of mutated peptide ligands identification by matching exome sequencing and MS immunopeptidomics. Patient tumour tissue is used for MS analysis and exome sequencing. Mutations are called and matched with MS data. Mutated peptide ligands are then further evaluated for recognition by patient’s autologous and matched allogeneic T cells (Bassani-Sternberg et al. 2016).
Further, PanPro - a machine learning approach trained on MS elution data that disconnects HLA binding prediction from the prediction of natural processing and presentation of potential neoantigens to the cell surface - leads to a significantly improved identification of neoantigen targets compare to other in silico predictions - eg. NetCHOP that have not been trained on MS elution data.
While MHC binding is required for a peptide to be presented to T cells, not all MHC binders are immunogenic. In fact, large proportion of neo-epitopes are non-immunogenic (ie 98%) and self-similarity apart from the lack/improved binding to MHC contributes to T cell immunogenicity. [Only for the peptides that exhibit conserved MHC binding between the mutated and non-mutated peptide, will the self-similarity contribute to the immunogenicity of this peptide.]
The interaction of a TCR to the p:MHC complex (MAP) holds a key, but currently poorly comprehended, component for our understanding of this variation in the immunogenicity of MHC binding peptides. The critical algorithms still missing remain T cell interaction models predicting T cell binding to MAPs. Apart from the extraordinay diversity of the T cell repertoire (estimated to be > 10^15), making the prediction of the antigen specificity of a given T cell from its TCR sequence impracticable, the negative T cell selection imposes additional layer of unpredictability of T cell cross-reactivity and peptide immunogenicity.
Nevertheless, Lanzarotti et al. (2017) demonstrate that identification of the cognate target of a TCR from a set of p:MHC complexes to a high degree is achievable using simple force-field energy terms. They propose a modeling pipeline for TCR:p:MHC structure prediction and with it lay the foundation for future work within prediction of TCR:p:MHC interactions.
Moreover, Calis et al (2013) set out to determine which properties of pMHCs cause differences in immunogenicity, by carefully collecting a large set of immunogenic and non-immunogenic pMHCs, and analysing the difference between these sets. Two important observations were made: First, that positions P4–6 of a presented peptide are more important for immunogenicity. Second, some amino acids, especially those with large and aromatic side chains, seem to be better recognized by T-cells as they associate with immunogenicity. Next, this information was combined into a simple model to predict the immunogenicity of new pMHCs (this model is made available at http://tools.iedb.org/immunogenicity/).
After the past successful elucidation of different steps in the MHC-I presentation pathway, the identification of variables that influence immunogenicity will be an important next step in the investigation of T-cell epitopes.
Neo-antigen Vaccine Development
For an in-depth reading on development of neo-antigen based vaccine approaches refer to the earlier post: Guidelines for the next-generation cancer vaccines.
Figure 2. The typical workflow for neoepitope selection and vaccine manufacture. DNA and RNA are extracted from single-cell suspensions of tumour cells and matched normal tissue cells.
Somatic mutations of tumour cells are discovered by whole-exome sequencing (WES). RNA sequencing (RNA-seq) narrows
the focus to mutations of expressed genes. Clinical HLA typing is carried out on DNA from normal tissue. The potential
antigenicity of neoepitopes identified by WES and RNA-seq is assessed by predicting the affinity of the neoepitopes for
binding to the HLA type of that individual (using NetMHCpan), thereby generating candidate vaccine epitopes. Validated
epitopes are selected for incorporation into the personalized cancer vaccine, which is administered to patients in
combination with an immune adjuvant (Hu et al. 2017).
Figure 3. Strategies to improve personalized neoantigen vaccines for cancer (Hu et al. 2017).
The 2 big challenges to overcome: TIME & COSTS
One of the big challenges to overcome, should this form of personalized treatment prove broadly successful, is the cost in developing these customized vaccines. Current estimates claim a single patient's neoantigen vaccine costs up to $US 60,000 to produce. In tandem with other new drug innovations, some patients could be paying several hundred thousand dollars for these treatments should they reach the market.
The time it takes to produce an individual vaccine is also a concern when considering how this treatment could be rolled out on a mass scale. It took several months to produce the vaccines used in the published clinical studies (see the side bar), but the researchers are confident this time frame could be reduced to six weeks or less. However, this is still a significant amount of time if the process was to be rolled out on a large scale.
