Neoantigen vaccines Flashcards
Define neoantigen
tumour specific antigen generated by mutations in tumour cells, which are not expressed within normal tissue
- The presence of neoantigens is one of the key differences between normal and tumour cells
Difference between classsical antigens and neoantigens
Self-antigens expressed primarily on tumour cells (e.g. oncofetal or cancer-testes antigen)
o Usually gene products from early development or from specific tissue
Self-tolerance (our T cells have been made tolerant) and easily evaded
Not many results have been disappointing – a small number of vaccines with limited effectiveness
o Neoantigens
Mutant proteins unique to the tumour
New promising approach
• Examples are mutated normal protein (will focus on these), but also viral antigens being expressed (e.g. in cervical cancer) and also ‘dark antigens’ – come from genomic dark matter e.g. the portion of the genome that is not normally expressed as protein, genes usually silenced in healthy but activated in cancer cells
What has recently allowed us to study neoantigens between?
- Together with bioinformatics, NGS has enabled mapping of all mutations in cancer and the prediction of MHC molecule-binding neoepitopes (epitope = part of antigen that binds to receptor)
What are 3 possible ways of targeting neoantigens?
- Synthetic neoantigen vaccines
- ICIs
- T cells specific to the neoantigen
o Castle et al 2012
First direct evidence of cancer exome-based approach used to identify neoantigens that could be targetted by T cells cells (this was only one of the groups, another also independently contributed)
Used NGS of exome to identify ~1000 mutations present in cells from a murine model of melanoma
Immunised melanoma mice with 50 synthetic long peptides encoded by some of these mutated epitopes.
1/3 of these peptides were found to be immunogenic
Results showed that neoantigen peptide vaccines targeting MU30 and MUT44, two mutated antigens, had significant preventative and therapeutic effects in the mouse tumour models
• They don’t seem to have any stats on their in vivo data for example, they don’t show a stat sig difference between untreated mice and the two vaccinated groups in terms of decrease in tumour growth and survival
Carreno et al 2015
In-human support of Castle 2012
• Trialled 3 patients with metastatic melanoma
• Given autologous DCs loaded ex vivo with synthetic peptides representing individual mutations of each patient predicted to bind to frequent class I haplotype HLA-A2
• Vaccine-induced CD8+ T cell responses with specificity for the immunogen were detected in several peptides (9 of 21 peptides)
• Cav – recognition of autologous melanoma cells was not assessed, so the relevance of the vaccine response was not clear
What 3 features does a vaccine in clinical practice need?
- Target tumour specific antigens that induce potent T cell responses. Multiple needed to avoid immune evasion
- Platform need a suitable approach – either provide and concentrated antigen source or material that can be used by the body to generate antigens (e.g. mRNA)
- Delivery combine with adjuvant & ICI
When were neoantigen vaccines first properly shown to be successful? By whatt groups? Which cancer type?
2017 - two small trials of pts with melanoma
o Sahin et al 2017
o Ott et al 2017
o Limitations of both – neither had a control arm, small cohorts
o Sahin et al 2017
Same group that made Pfizer vaccine
1-3 months of neoepitope discovery and vaccine manufacturing
8 neoepitope RNA vaccinations over 43 days - previously shown that RNA molecules can be taken up by APCs
13 patients given it
8 of these had no tumour development on follow-up (12-23 months)
Looked at number of metastatic events before and after the vaccine was given – metastatic event plateaued
Immunosurveillance analysis of PBMCs in pts showed the RNA vaccine could enhance the neoantigen-specific T cell response and induce new T cell response
They’re now conducting phase 2 trials and we hope they will be published soon
o Ott et al 2017
6 people who had undergone surgery to remove a tumour
Sequenced DNA and healthy cells and then used algorithm to predict which neoantigen would bind well to MHC
Each participant vaccinated with long peptides, representing up to 20 neoantigens
4/6 no tumour recurrence, 2/6 had tumours but these regressed after PD1 inhibition
None of these patients had measurable disease at the time of vaccine initiation this prevents direct evaluation of in situ vaccine induced tumour efficacy
Discuss technical limitations of methods used to if neoantigens in studies such as Sahin and Ott 2017?
Technical limitations!
- The technological pipeline used to identify neoantigens in these studies varies substantially and further optimisation is likely possible
- Also, erroneous mutation detection
o Standard techniques work well for SNPs – but other mutation types might be relevant such as gene fusions, indels and epigenetic, transcriptional, translational and post-translation aberrations may generate neoepitopes
o These are difficult to assess with current technologies
- Process of preparing a vaccine from a sample usually takes 3-5 months limits the clinical applications – Sahin et al (2017) took 1-3 months
What do we currently use to find mutations with likelihood of immunogenicity? What is an example of a model used to predict this and what is it?
(1) expression of the gene in the tumour and (2) an estimation of MHC binding
NetMHC is an example
• Artificial neural networks trained on binding data from different MHCs much better for MHC I than MHC II
o Liu et al (2019)
found that only 14% of predicted neoantigens (via computational analysis) induced IFN gamma response, 12% of antigens not robustly expressed induced a response
o Has been suggested that the stability of peptide-MHC complex is a better predictor for immunogenicity rather than MHC binding alone – current prediction algorithms for stability cover only a few MHC alleles and rely on small data sets
Aside from technical limitations of methods used to identify neoantigens what else is there?
Also limitations to do with the tumour itself
e.g. tumour heterogeneity, the dynamic nature of the genome, targetting of driver mutations
How may tumour heterogeneity be a limiting factor?
o Typically, analyses rely on small biopsy from a single tumour may not represent the full clonal spectrum
o Targeting one clone may cause outgrowth of antigen-negative clones
- Tran et al (2015)
suggests only high-TMB tumours will qualify
o Whole exome seq in 10 pts with metastatic gastrointestinal cancer
o Performed TCR-Vbeta deep seq on the tumours
o Profiling of individual cancer pts revealed spontaneous immune responses against only a small fraction of their mutations (<1%), casting doubt on the immunogenicity of mutations per se
o This lead to the presumption that only tumours with a high mutational burden, and accordingly a higher diversity of cancer-reactive T cells may qualify for neoantigen-based immunotherapy
o These results differed from their in vitro work
Generate TIL culture from each pts and used a TMG (tandem minigene) approach. Strong of minigenes transcribed to RNA in vitro, transfected into autologous APCs and APCs co-cultured with TILs
They found 9/10 pts elicit T cell responses against at least one somatic mutation expressed by their tumours
This could be due to poor clonal expansion, survival and infiltration of the TME, also tumour heterogeneity
o Laumont et al (2018)
Developed a new proteomics approach for analysing non-coding regions of the genome
In two murine cancer cell lines and seven primary human tumours, they identified a total of 50 tumour-specific antigens, about 90% of which derived from non-coding regions that would have been missed by standard exome-based seq approaches.
They further found that one of these TSA was shared between two diferrent tumours v promising
The finding that non-coding regions could serve as a source of TSAs – more targetable antigens across tumours
Criticisms of this paper
• Their approach was not compatible with computation of false discovery rates – all the TSAs had to be validated by MS
• They only detected the most abundant TSAs – likely underestimate extent of sharing between pts
- Abelin et al (2019)
Improving MHCII binding epitope prediction
o It remains difficult to predict the antigens that will be presented by HLA-II (due to inaccurate peptide-binding prediction and unsolved complexities of HLA-II pathway)
o They created a scalable monoallelic HLA ligandome profiling workflow called MAPTAP (mono-allelic purification with tagged allele constructs)
o Outperformed NetMHC in predicting presented peptides from HLA-II ligandome experiments
o Hu, 2021 (Ott lab)
, single cell sequencing to provide insight into composition of induced T cell responses at a clonal level
Clinical outcome and circulating immune response of 8 patients with surgically resected melanoma after median 4 years treatment with NeoVax – follow up of earlier study (peptide vaccine)
Used PBMC samples from patients to identify that T cell responses against most immunising peptides could sustained up to 4.5 years after vaccination
Demonstrated importance of CD4 responses – 2 patients had no detectable CD8 response despite sustained response to vaccine
Ex vivo these were found to exhibit a memory phenotype + distinct functionalities – used transcriptomic signatures of individual cells
• Looked in 3 patients
• Isolated neoantigen reactive T cells ex vivo from PBMCs
• Characterised individual transcriptional status
• Pooled all these cells, identified four clusters (each composed of cells from all three patients) - evaluated significant genes to identify (1) naïve-like, (2) cytotoxic-like, (3) activation induced cell death like, and (4) memory like.
• Assessed the presence of these over time and found that most T cells had memory-like phenotype (pooling to do with gene expression)
However, would have been better to see functional assays of these ‘clusters’ – this is just RNA pattern
Didn’t compare how non-responders T cell phenotype differed from responders
Also found evidence of non-vaccine neoantigen specific T cells, suggestion of epitope spreading – non-vaccine antigen-directed T cell responses were detected in two patients
However see several recurrences - neoantigen vaccination alone may not be sufficient – 5/8 patients had recurrence
Despite this, data clearly demonstrate peptide vaccines can elicit memory T cell responses that can be sustained over several years
Ott lab (2020)
Vaccine plus nivolumab in 82 patients with metastatic melanoma/ NSCLC/ bladder cancer
No treatment related serious adverse events injection-site reactions and influenza-like illness)
De novo antigen specific T cell responses to vaccinating peptides, 42% and 24% of the vaccinating peptides, respectively, triggered CD4+ and CD8+ T cell responses
Again prominent CD4+ T cell induction
Also reported epitope spreading - correlated with improvements in progression free survival
• Suggests that neoantigen-induced T cells are not only capable of trafficking to the tumour but are also potentially able to kill tumour cells, leading to the release of additional neoantigens that become targets for additional T cells
• Could be used make broad immune response
Study specific problems
• Clinical benefit appeared similar to anti PD1 alone - need a RCT to find additive benefit and need to identify biomarkers
• Identify neoantigen specific T responses in those that responded and those that didn’t to guide future neoantigen prediction methods
Concluding remarks
- Stratified treatment has been considered a synonym with personalised medicine (e.g. think of examples in SCZ or NDD)
o Stratified therapy excludes a whole group of pts who do not carry a biomarker
o True pt-specific therapy should be achievable by neoantigen vaccination - Questions
o What is optimal clinical setting? – if there is already a tumour how much more will adding tumour-associated peptides provide?
o Are they effective in pts with low TMB? - Success of Pfiezer mRNA COVID-19 vaccine – could we apply this?
First direct evidence of cancer exome-based approach used to identify neoantigens that could be targetted by T cells cells (this was only one of the groups, another also independently contributed)
Used NGS of exome to identify ~1000 mutations present in cells from a murine model of melanoma
Immunised melanoma mice with 50 synthetic long peptides encoded by some of these mutated epitopes.
1/3 of these peptides were found to be immunogenic
Results showed that neoantigen peptide vaccines targeting MU30 and MUT44, two mutated antigens, had significant preventative and therapeutic effects in the mouse tumour models
• They don’t seem to have any stats on their in vivo data for example, they don’t show a stat sig difference between untreated mice and the two vaccinated groups in terms of decrease in tumour growth and survival
o Castle et al 2012
In-human support of Castle 2012
• Trialled 3 patients with metastatic melanoma
• Given autologous DCs loaded ex vivo with synthetic peptides representing individual mutations of each patient predicted to bind to frequent class I haplotype HLA-A2
• Vaccine-induced CD8+ T cell responses with specificity for the immunogen were detected in several peptides (9 of 21 peptides)
• Cav – recognition of autologous melanoma cells was not assessed, so the relevance of the vaccine response was not clear
Carreno et al 2015
Same group that made Pfizer vaccine
1-3 months of neoepitope discovery and vaccine manufacturing
8 neoepitope RNA vaccinations over 43 days - previously shown that RNA molecules can be taken up by APCs
13 patients given it
8 of these had no tumour development on follow-up (12-23 months)
Looked at number of metastatic events before and after the vaccine was given – metastatic event plateaued
Immunosurveillance analysis of PBMCs in pts showed the RNA vaccine could enhance the neoantigen-specific T cell response and induce new T cell response
They’re now conducting phase 2 trials and we hope they will be published soon
o Sahin et al 2017
6 people who had undergone surgery to remove a tumour
Sequenced DNA and healthy cells and then used algorithm to predict which neoantigen would bind well to MHC
Each participant vaccinated with long peptides, representing up to 20 neoantigens
4/6 no tumour recurrence, 2/6 had tumours but these regressed after PD1 inhibition
None of these patients had measurable disease at the time of vaccine initiation this prevents direct evaluation of in situ vaccine induced tumour efficacy
o Ott et al 2017
found that only 14% of predicted neoantigens (via computational analysis) induced IFN gamma response, 12% of antigens not robustly expressed induced a response
o Has been suggested that the stability of peptide-MHC complex is a better predictor for immunogenicity rather than MHC binding alone – current prediction algorithms for stability cover only a few MHC alleles and rely on small data sets
o Liu et al (2019)
suggests only high-TMB tumours will qualify
o Whole exome seq in 10 pts with metastatic gastrointestinal cancer
o Performed TCR-Vbeta deep seq on the tumours
o Profiling of individual cancer pts revealed spontaneous immune responses against only a small fraction of their mutations (<1%), casting doubt on the immunogenicity of mutations per se
o This lead to the presumption that only tumours with a high mutational burden, and accordingly a higher diversity of cancer-reactive T cells may qualify for neoantigen-based immunotherapy
o These results differed from their in vitro work
Generate TIL culture from each pts and used a TMG (tandem minigene) approach. Strong of minigenes transcribed to RNA in vitro, transfected into autologous APCs and APCs co-cultured with TILs
They found 9/10 pts elicit T cell responses against at least one somatic mutation expressed by their tumours
This could be due to poor clonal expansion, survival and infiltration of the TME, also tumour heterogeneity
- Tran et al (2015)
Developed a new proteomics approach for analysing non-coding regions of the genome
In two murine cancer cell lines and seven primary human tumours, they identified a total of 50 tumour-specific antigens, about 90% of which derived from non-coding regions that would have been missed by standard exome-based seq approaches.
They further found that one of these TSA was shared between two diferrent tumours v promising
The finding that non-coding regions could serve as a source of TSAs – more targetable antigens across tumours
Criticisms of this paper
• Their approach was not compatible with computation of false discovery rates – all the TSAs had to be validated by MS
• They only detected the most abundant TSAs – likely underestimate extent of sharing between pts
o Laumont et al (2018)
Improving MHCII binding epitope prediction
o It remains difficult to predict the antigens that will be presented by HLA-II (due to inaccurate peptide-binding prediction and unsolved complexities of HLA-II pathway)
o They created a scalable monoallelic HLA ligandome profiling workflow called MAPTAP (mono-allelic purification with tagged allele constructs)
o Outperformed NetMHC in predicting presented peptides from HLA-II ligandome experiments
- Abelin et al (2019)
, single cell sequencing to provide insight into composition of induced T cell responses at a clonal level
Clinical outcome and circulating immune response of 8 patients with surgically resected melanoma after median 4 years treatment with NeoVax – follow up of earlier study (peptide vaccine)
Used PBMC samples from patients to identify that T cell responses against most immunising peptides could sustained up to 4.5 years after vaccination
Demonstrated importance of CD4 responses – 2 patients had no detectable CD8 response despite sustained response to vaccine
Ex vivo these were found to exhibit a memory phenotype + distinct functionalities – used transcriptomic signatures of individual cells
• Looked in 3 patients
• Isolated neoantigen reactive T cells ex vivo from PBMCs
• Characterised individual transcriptional status
• Pooled all these cells, identified four clusters (each composed of cells from all three patients) - evaluated significant genes to identify (1) naïve-like, (2) cytotoxic-like, (3) activation induced cell death like, and (4) memory like.
• Assessed the presence of these over time and found that most T cells had memory-like phenotype (pooling to do with gene expression)
However, would have been better to see functional assays of these ‘clusters’ – this is just RNA pattern
Didn’t compare how non-responders T cell phenotype differed from responders
Also found evidence of non-vaccine neoantigen specific T cells, suggestion of epitope spreading – non-vaccine antigen-directed T cell responses were detected in two patients
However see several recurrences - neoantigen vaccination alone may not be sufficient – 5/8 patients had recurrence
Despite this, data clearly demonstrate peptide vaccines can elicit memory T cell responses that can be sustained over several years
o Hu, 2021 (Ott lab)
Vaccine plus nivolumab in 82 patients with metastatic melanoma/ NSCLC/ bladder cancer
No treatment related serious adverse events injection-site reactions and influenza-like illness)
De novo antigen specific T cell responses to vaccinating peptides, 42% and 24% of the vaccinating peptides, respectively, triggered CD4+ and CD8+ T cell responses
Again prominent CD4+ T cell induction
Also reported epitope spreading - correlated with improvements in progression free survival
• Suggests that neoantigen-induced T cells are not only capable of trafficking to the tumour but are also potentially able to kill tumour cells, leading to the release of additional neoantigens that become targets for additional T cells
• Could be used make broad immune response
Study specific problems
• Clinical benefit appeared similar to anti PD1 alone - need a RCT to find additive benefit and need to identify biomarkers
• Identify neoantigen specific T responses in those that responded and those that didn’t to guide future neoantigen prediction methods
Ott lab (2020)