Neoantigen quality predicts immunoediting in survivors of pancreatic cancer

Cancer immunoediting is a hallmark of cancer 2 that predicts that the immune system will kill more cancer cells to cause less clone growth. Although proven in mice, it's not clear if it occurs naturally in human cancers. Over the course of 10 years, 70 human Pancreatic Cancers evolved. We found that rare long-term survivors of pancreatic cancer who have stronger T cell activity in primary tumours develop less heterogeneous recurrent tumours with fewer immunogenic mutations. We infer that a neoantigen is high-quality by two features. Cancer clone fitness is the aggregate cost of T cells recognizing high-quality neoantigens offset by gains from oncogenic mutations. The model predicts that long-term survivors of pancreatic cancer develop recurrent tumours with fewer high-quality neoantigens. Evidence shows that the human immune system edits neoantigens. We have a model to predict how immune pressure causes cancer cell populations to evolve. Our results show that the immune system is able to suppress cancer.

The immune system in multicellular organisms must eliminate transformed cells as an evolutionary necessity. The immune system must protect the host from cancer, as a consequence of the theory of cancer immunosurveillance. Some of the new genes generated by cancers are called neoantigens. neoantigens can escape T cell central tolerance in the thymus and become an anti-cancer agent. Heterogeneity in cancer cell clones with variable immunogenicity is the result of neoantigens. T cells can induce less immunogenic clones to grow in cancers.

The principle of how human cancers evolve remains uncertain despite the fact that cancer immunoediting has been demonstrated in mice. Longitudinal tracking of large numbers of patients over time is required for definitive evidence. It's difficult to say whether the human immune system naturally edits cancers and whether edited clones can be predicted.

We looked at how 70 Pancreatic ductal adenocarcinomas evolved from 15 patients over the course of 10 years. The immunoediting hypothesis was thought to be an ideal cancer to test. The number of neoantigens in human PDACs is less than in other types of cancer. The ability to distinguish true neoantigen selection from neutral genomic changes over time is maximized by this. Second, T cell infiltrates range from zero to 1,000-fold higher. We can theoretically observe how differential immune selection pressures affect cancer cell clones with the subsets that approximate immune deficient and immuneficient cancers. It is possible to detect how cell-extrinsic immune pressures affect clonal evolution with the help of the oncogenes.

The experimental design includes overall survival and disease-free survival of patients. In g there are omentum, aorta, diaphragm and perirectum, and pericardium. The bars show the values. Two-tailed log-rank tests were used to determine P values.

The data is source data.

We compared the evolution of immune-proficient and immune- deficient human cancers to the evolution of long-term survivors and short-term survivors. We demonstrated that the primary tumours in LTSs have a 12-fold greater number of activated CD8 + T cells 5 that are predicted to target immunogenic neoantigens 5. The largest T cell clones in the current cohort have the same CDR3 sequence as the largest T cell clones in the previous cohort. We hypothesised that the higher immune pressure in LTSs would cause tumours to preferentially lose their clones with neoantigens over time. We compared the evolution of tumours from primary to recurrent. We found that the longer the survival times, the less recurrent it was. More than 75% of LTSs had recurrent tumours that were only metastatic. The tumours recur with their own evolutionary trajectory.

We performed whole-exome sequencing and inferred the clonal structures of matched primary and recurrent tumours to see if differential selection pressure could explain the unique recurrence patterns. We reasoned that the diversity of tumours clones should be limited due to the increased immune selection pressure. We found that the primary tumours in LTSs were more heterogeneous than the recurrent ones. We compared the total number of non-synonymous mutations and predicted MHC-I restricted neoantigens to see if this could be explained by selection pressure. The primary LTS tumours had a similar TMB with a comparable number of neoantigens. Despite the differences, the two tumours had the same number of synonymous mutations and the same number of driver oncogenes. There were fewer co-occurring mutations in oncogenes in recurrent tumours of LTSs than there were in recurrent tumours ofSTSs. Over time 19, LTS recurrent tumours remained largely neutral. There were fewer neoantigens generated by the LTS tumours than by theSTS tumours. The data supports the hypothesis that the immune selection in the tumours edited the clones.

There is a difference in Shannon entropy between recurrent and primary PDACs. The bars show the values. Two-tailed Mann and Whitney U tests were used to determine P values.

The data is source data.

To identify the edited neoantigens, we extended our previous neoantigen quality model, which quantifies the immunogenic features of a neoantigen, to propose that two competing outcomes determine whether a neoantigen is high-quality. To estimate the likelihood of the immune system recognizing a neoantigen, we measure the sequence similarity of the mutant neopeptide. The recognition potential R of p MT is a proxy for the recognition space of the T cell receptor.

The model and experimental approach to estimate cross-reactivity distance is a Neoantigen quality model. The matrix M is defined by the distance from g. The number of substitutions is indicated by circles. The heat maps show the amino acids in p WT. The green line is a regression fit. The dendogram shows the order in which the heat maps are ordered. Two-tailed Pearson correlation and two-sided Kolmogorov tests were used to determine P values.

The data is source data.

The immune system can fail to discriminate p MT from its wild-type (WT) peptide, and therefore tolerate it as self. The immune system must exert greater self discrimination in tumours to overcome the principles of negative T cell selection. The D between p WT and p MT is approximated by two features. The MHC presentation of p WT and p MT is differential. The availability of T cells is estimated. If p WT is not presented to T cells in the middle of the body, it means poor p WT. Cross-reactivity distance C is a new model term that estimates the antigenic distance required for T cells to discriminate between p MT and p WT. The self discrimination D is a proxy for the peptides outside the toleration space. We define neoantigen quality as Q, R, and D, with components that estimate whether a neoantigen can be recognized as non- non- non- non- non- non- non- non- non- non- non- non- non- non- non- non

We used recent findings to model C. T cell cross-reactivity between p MT and p WT could be used to estimate the relative C of different neoantigenic substitutions, as shown by the binding domains of peptide-degenerate TCRs 21, 22 and TCR-degenerate peptides 23. We used a model p WT from human cytomegaloviruses 24 and 22 to determine the strength of theNLVrestricted strong epitope. We varied the NLV peptide by every amino acid at each position to model p MT substitution, and compared how TCRs cross-react between each p MT and its p WT across a 10,000-fold concentration range where p WT changes maximally altered T cell activation. The substitution pattern was either highly, moderately or poorly cross-reactive. There were similar patterns of cross-reactivity between a model HLA-A*02:01-restricted weaker p WT and a melanoma self-antigen. The three p WT -specific TCRs and single-amino-acid-substituted p MT s suggest that the pattern of substitution is defined by C. The cross-reactivity distance between a p WT and its corresponding p MT was quantified. The half maximal effective concentration was chosen because it was consistent with the sigmoidal function. The Hill equation shows how EC 50 determines how areceptor is activated. The model for C that estimates whether a neoantigenic substitution is cross-reactive or tolerant is estimated by the EC 50. There are 6 a, b and 7a. We tested whether C predicted cross-reactive replacements in an LTS. C predicted cross-reactive p WT and p MT in this neopeptide. The antigenic distance for a TCR to cross-react between two pairs was combined. There are two factors that promote cross-re activity: substitution at the termini 27 and within the biochemical families. The definition of self- discrimination between a p WT and its corresponding p MT is now defined by this composite C.

The relative weight between the two terms is set by (w). The parameters of the neoantigen quality model were chosen to maximize the log-rank test score of survival analysis on an independent cohort of 58 patients.

Our model was applied to the PDAC to show how immunoediting depletes neoantigens with higher D in LTS. The antigenic distance is the first thing we divide the frequencies by. The more antigenic distance from self was more important in determining the amount of depleted in both LTS and STS PDACs. The full D model was used to find that neoantigens with both a higher C and D were more deplete in the LTS. The genes in the HLA class-I pathway were not differentially altered, deleted, expressed or localized in the LTSs. Tumours lose high-quality neoantigens.

We put neoantigen quality parameters into a fitness model to see if our model that predicts clonal evolution can identify immunoedited clones. We reconstructed the phylogenies for all the tumours from each patient to give a common structure and track clone frequencies. Positive selection was accounted for due to cumulative mutations in driver oncogenes. This effect is quantified in a minimal model, which counts the number of missense mutations in the same genes. The sum of a negative fitness cost due to immune recognition of high-quality neoantigens and positive fitness gain due to the accumulation is defined by the composite fitness model.

The primary tumour composition and the fitness model are used to predict a recurrent tumours clone composition. The immune fitness cost of recurrent tumours was evaluated with pseudocounts. The green line is a regression fit. The bars show the values. Two-tailed Spearman correlation, two-tailed Pearson correlation and two-tailed Mann correlation were used to determine the P values.

The data is source data.

The free parameters I and P are used to set the fitness components. The model is used to predict the frequencies of clones.

In the primary tumours, the predicted Frequency is (hatx_rmrecalpha ) We looked at how closely the fitness model predicted clonal evolution. We fitted the model parameters for the recurrent tumours in the LTS andSTS cohort.

When we compared our model with a neutral model, we found that our model provided a better fit of the observed evolution of LTS compared to the clones. The partial fitness model that only includes the oncogenicity component showed reduced performance for the LTS tumours. We compared observed and model-fitted clone frequencies between the primary and recurrent tumours. The correlation between therank and theExtended Data Table 1b was 0.65 and 0.28 for the direction of frequencies. The model's better predictions in the tumours are due to the presence of immune selection.

The immune cost is calculated by taking the immune component and dividing it by the total tumour immune cost. The immune fitness cost was lower in recurrent LTS tumours. The immune cost of clones that are new in recurrent tumours but not in primary tumours was considered. New clones with lower immune fitness cost were found in recurrent LTS tumours. The observations suggest that the recurrent tumours had been altered by the immune system.

We confirmed the results by analyzing the available recurrent tumours samples. The specificity of T cell clonal expansion was quantified using the TCR dissimilarity index 18. We found that more T cell clonal expansion in tumours correlated with more highly edited tumours. The results show that neoantigens are immunoedited in PDAC, and that our fitness model captures the pressures of T cells acting on tumours clones.

There are several questions on how the immune system interacts with cancer. Does cancer immunotherapy happen in humans? The theory of cancer immunoediting was developed by studying the effects of cancer on the body, but it is not certain if these principles apply to human cancers. When the immune system recognizes an immunogenic antigen in a primary tumours, it should cause the antigen to be eliminated in the recurrent tumours. Tumours that evolve under stronger immune pressure lose more neoantigens. The proof for immunoediting is found in a low-mutated cancer that is resistant to immunity, and it is noteworthy that we did not address how different cellular compositions and tissue environments may affect editing. The claim that immunoediting is a broadly conserved principle of carcinogenesis is strengthened by this.

Doesimmunoediting manifest as loss of immunogenic antigens, or does cancer also have genetic resistance? We observed the former but not the latter. The magnitude of theselective pressure is what we think governs such phenotypes. The immune pressure in a 32 tumours is lower than in a tumours with a higher immune pressure. When pressure is moderate, the tumours lose their immunity, whereas when pressure is maximal, they gain resistance. Under immune selection, the selection of clonal composition and pressure determine adaptive change. These concepts will be tested further.

Can we quantify how the immune system recognizes changes? We use C to quantify the antigenic distance of mutated peptides in the TCR-recognition space and the qualities that make them immune. We propose that the model we used to identify neoantigens captures common features of the immune system. We anticipate that our model can illuminate the biology of antigens beyond cancer, including T cell cross-reactivity between antigens, as well as therapies that require rational antigen selection.

The immune system's ability to discriminate between changes in single amino acids can be used to predict how cancers will evolve. The function of the immune system is to maintain the integrity of the host genome. Our model captures the mechanisms through which the immune system protects genomic integrity.

Reporting summary

The Nature Research Reporting Summary contains further information on research design.

The European Genome Archive has previously described 19 of the raw data obtained through the medical donation programme. The NCBI Sequence Read Archive has all the other raw data. The data used in this study is available at the ICGC. The data used in this study are from the TCGA-PAAD dataset, which can be found at the NCI Genomic Data Commons.

The code used to build and apply the model can be found at GitHub.

1. V. Shankaran and his colleagues studied the effects of IFN and lymphocytes on cancer. Nature 410 was published in 2001.

   The article is from the Google Scholar.

2. The next generation of cancer. Cell 146, .

   The article is from the Google Scholar.

3. A T-cell-dependent mechanism of cancer immunoediting is revealed in cancer exome analysis. Nature 484, 400, and 404.

   The article is from the Google Scholar.

4. The model predicts the response to checkpoint blockade immunotherapy. Nature 541, 541, 541, 541, 541, 541, 541, 541, 541, 541, 541, 541, 541, 541, 541, 541, 541, 541, 541,

   The article was written by a scholar from the internet search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the

5. Identification of unique neoantigen qualities in long-term survivors of pancreatic cancer. Nature 541, 512.

   The article is from the Google Scholar.

6. F. M. Burnet wrote about the concept of immunological surveillance. There is a prog. Exp. The Tumor Res. 13 was published in 1970.

   The article is from the Google Scholar.

7. Bruce, A. T., Ikeda, H., Old, L. J. and Schreiber, R. D. Cancer immunoediting: from immune evasion to tumor escape. Nat. The immune system. 3, 988 (2001).

   The article is from the Google Scholar.

8. There are neoantigens in cancer immunotherapy. Science 343, 69–74.

   The article is from the Google Scholar.

9. The immune escape from lung cancer is caused by neoantigen. Nature 476, 478, 486, 499, 499, 499, 499, 499, 499, 499, 499, 499, 499, 499, 499, 499, 499, 499, 499,

   The article is from the Google Scholar.

10. In ovarian cancer, there are two types of clonal dynamics. Cell 173, 1755

    The article is from the Google Scholar.

11. Heterogeneous tumor-immune microenvironments in ovarian cancer patients. Cell 170, 928, 928, 928, 928, 928, 928, 928, 928, 928, 928, 928, 928, 928, 928, 928, 928, 928, 928,

    The article was written by a scholar from the internet search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the

12. Balli, D., Rech, A. J., Stanger, B. Z. It was Clin. Cancer Res. 23, 3129

    The article is from the Google Scholar.

13. The sensitivity to the PD-1 blockade in small cell lung cancer is determined by themutational landscape. Science 343, 123,128

    The article is from the Google Scholar.

14. The correlation of response to the blockade in melanoma was studied. Science 350, 207, and211 were published in 2015.

    The article was written by a scholar from the internet search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the

15. During the genetic evolution of the cancer, distant metastasis occurs. Nature 476, 1114, 1117, 2010.

    The article is from the Google Scholar.

16. Ino, Y. and their colleagues studied the immune microenvironment of pancreatic cancer. Brit. J. Cancer had a number of articles.

    The article is from the Google Scholar.

17. The composition and diversity of the Tumor Microbiome influence the outcomes of Pancreatic Cancer. Cell 178, .

    The article is from the Google Scholar.

18. Probing T-cell response by sequence-based modeling. A journal about Comput. There is a Biol. In the year 2021.

    The article is from the Google Scholar.

19. The evolutionary origins of cancer. Cancer discov. 10, 793, 793, 793, 793, 793, 793, 793, 793, 793, 793, 793, 793, 793, 793, 793, 793, 793, 793, 793

    The article is from the Google Scholar.

20. Heteroclitic immunization causes tumor immunity. J. Exp. Med. 186, 1553, and 1561 were published in 1998.

    The article is from the Google Scholar.

21. Predictive features are defined as epitope-specific T cell receptor repertoires. Nature 547, 89–93.

    The article is from the Google Scholar.

22. J. Glanville and his team identified specificity groups in the T cell receptors. Nature 547, 94–98.

    The article is from the Google Scholar.

23. The specificity of T cell recognition is explained. Cell 157, 1073.

    The article is from the Google Scholar.

24. The identification of three HLA-A*0201-restricted cytotoxic T cell epitopes in the cytomegalovirus is related to eight strains of the virus. J. Immunol. 163, 5512 and 5518 were published in 1999.

    There is a CAS and a PubMed and a Google Scholar.

25. The recognition of multiple epitopes in the human melanoma antigen by tumor-infiltrating T lymphocytes is associated with in-vivo tumor regression. J. Immunol. 154, 3962, 3962, 3962, 3962, 3962, 3962, 3962, 3962, 3962, 3962, 3962, 3962, 3962, 3962, 3962, 3962, 3962, 3962, 3962

    There is a CAS and a PubMed and a Google Scholar.

26. Parkhurst and his team improved the CTL's ability to bereactive to melanoma with the help of a modified melanoma antigen. J. Immunol. There were 157, 2539, and 2548 in 1996.

    There is a CAS and a PubMed and a Google Scholar.

27. Predicting cancer neoantigens is an important determinant of themutation position. J. Exp. Med. The year 2020 is 217.

    The article was written by a scholar from the internet search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the

28. Deshwar and A. G. reconstructed subclonal composition and evolution from tumors. There is a genome. 16, 35.

    The article was written by a scholar from the internet search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the

29. Evans, R. A. et al. The JCI Insight was published.

    The article was written by a scholar from the internet search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the search engine, the

30. Barthel, F. P. and others studied the long-term trajectory of diffuse glioma in adults. Nature 576, 112.

    The article is from the Google Scholar.

31. The CD155/TIGIT axis promotes and maintains immune evasion. Cancer Cell 39 will be replaced by a new cell in 2011.

    The article is from the Google Scholar.

32. J. M. Zaretsky and his co-workers studied the causes of resistance to the anti-PD-1 blockade in melanoma. N. Engl. J. Med. 825, 825, 825, 825, 825, 825, 825, 825, 825, 825, 825, 825, 825, 825, 825, 825, 825, 825, 825,

    The article is from the Google Scholar.

Download references

The work was supported by an award from the National Institute of Health. M. is a scholar. The Integrated Genomics Core was funded by the Cancer Center Support Grant.

1. The authors contributed equally.

1. The Departments of Oncological Sciences and Genetics and Genomic Sciences are at the Icahn School of Medicine at Mount Sinai.

   There is a woman named Marta.

2. The Computational Oncology Service is at the Memorial Sloan Kettering Cancer Center.

   Benjamin D. Greenbaum, Jayon Lihm, Yuval Elhanati, and David Hoyos.

3. Memorial Sloan Kettering Cancer Center in New York, NY, USA has a Human Oncology and Pathogenesis Program.

   Luis A. Rojas, Masataka Amisaki, Rebecca Yu, and Abderezak Zebboudj.

4. The Department of Surgery at Memorial Sloan Kettering Cancer Center is known for its Hepatopancreatobiliary Service.

   Kevin Soares, Masataka Amisaki, and Rebecca Yu are all related to the same person.

5. The Ecole Normale Supérieure is a Laboratoire de Physique.

   Barbara Bravi, Remi Monasson,Thierry Mora, Aleksandra M. Walczak, and Simona Cocco.

6. The Department of Mathematics is located at Imperial College London.

   Barbara Bravi is a person.

7. The David M. Rubenstein Center for Pancreatic Cancer Research is located in New York.

   Kevin Soares, Rajya Kappagantula, Alvin Makohon-Moore, Christine Iacobuzio-Donahue, andVinod P.

8. The Center for Cell Engineering is in New York.

   Dobrin and Sadelain are related.

9. The Immunology Program is at the Memorial Sloan Kettering Cancer Center in New York.

   Dobrin and Sadelain are related.

10. The Human Oncology and Pathogenesis Program is at the Memorial Sloan Kettering Cancer Center.

    Christine Iacobuzio-Donahue.

11. The Garvan Institute of Medical Research is located in New South Wales, Australia.

    Anthony Gill and amber johns.

12. New South Wales, Australia is home to the University of Sydney.

    Anthony Gill.

13. Swim Across America and Ludwig Collaborative Laboratory are at the Memorial Sloan Kettering Cancer Center.

    Jedd Wolchok and Taha Merghoub.

14. The systems biology department is at the New York, NY, USA's Weill Cornell Medical College.

    Benjamin D. Greenbaum.

15. The Memorial Sloan Kettering Cancer Center in New York, NY, is home to theParker Institute for Cancer Immunotherapy.

    A man named Vinod P. Balachandran.

The study was conceived by B.D.G. and V.P.B. L.A.R. and Z.M.S. conceived the model. Z.M.S., L.A.R., B.D.G. and V.P.B. came up with the neoantigen quality model. The fitness model was constructed by Z.M.S. and B.D.G. B.B., T. Mora, R.M., A.M.W., and S.C conceived and constructed the TCR dissimilarity index. Z.M.S., L.A.R., K.S., J. Lihm, D.H., R.K., A.M.-M., A.J. T cell transductions were assisted by A.D. and M.S. Z.M.S., L.A.R., E.P., B.D.G. and V.P.B collaborated on the manuscript.

There are letters to Benjamin D. Greenbaum or Vinod P. Balachandran.

L.A.R is an inventor of a patent related to oncolytic viral therapy. L.A.R., Z.M.S. and V.P.B are listed as inventors on a patent application. M., B.D.G. and V.P.B are listed as inventors on a patent application. C.I.-D. has received funding. B.D.G. has received honoraria for speaking engagements from a number of companies and has received research funding from Bristol-Meyers Squibb. Bristol-Myers Squibb and Genentech supported V.P.B. J.W. is a consultant for many companies. J.W. has equity in several companies. T. Merghoub is a co-founder of IMVAQ and holds equity in the company. The other authors do not have competing interests.

Nature thanks Paul Thomas and the other anonymous reviewer for their contribution to the peer review of this work.

T cell receptor CDR3 is a sequence dissimilarity index in the primary and recurrent PDACs. The Restricted Boltzmann Machine model 18 is used to calculate the TCR dissimilarity index. The left panel shows the trend of the P value of the TCR dissimilarity index between the two groups. The blue line shows the mean P value and the circle shows the mean error. The green line is a linear regression fit. Two-tailed Mann-Whitney U test and two-tailed Pearson correlation have a P value.

The data is source data.

Whole-exome sequencing depth and number of synonymous mutations in primary and recurrent PDACs. The percentage of patients with corresponding driver genes is calculated. The bars are horizontal. The P value is by twotailed Mann-Whitney U.

The data is source data.

Tumour clone phylogenies from primary and recurrent PDACs.

There is an experiment to transduce and measure specific T cell receptor (T CR) activation.

The data is source data.

T cell activation to model (bfptextWT) s and single amino acid substitution.

The data is source data.

T cell activation curves can be fitted to model (bfptextWT) s and (bfptextMT) s.

The data is source data.

The cross-reactivity model is based on T cell receptor cross-reactivity to strong and weak. P values by Pearson correlation.

The data is source data.

The number of deletions and copy number neutral loss of Heterozygosity (LOH) changes in the B2M pathway genes. The bars are horizontal. There are horizontal bars on violin plots. Wald's test adjusted for multiple comparison testing.

The data is source data.

The log-likelihood score is based onSupplementary Methods. The amount of evidence of theselective pressures captured by each of the models is estimated by (31). The aggregated log-likelihood scores are shown in the orange bars. The red bars show the aggregated log-likelihood scores. The blue bars show the aggregated log-likelihood scores.

The data is source data.

Supplementary Tables 2 and 3 and Supplementary Methods are legends.

You agree to abide by the Community Guidelines when you submit a comment. If you find something that does not comply with our guidelines, please flag it as inappropriate.