High-Value Epitope Selection for Mutanome-Directed Immunotherapy Using an Innovative Cancer Neo-Epitope Classification System

Guilhem Richard1, Lenny Moise1,2, Frances Terry1, Bill Martin1, Annie De Groot1,2 1 EpiVax, Inc., Providence, RI, United States; 2 Institute for Immunology and Informatics, University of Rhode Island, Providence, RI, United States Abstract JanusMatrix Discovers Potential Treg Epitopes

T cell epitopes bearing tumor-specific discovered using whole-exomic and transcriptomic Avg Janus Human Score vs +/- Cytokine Release IEDB Positive IEDB Negative JanusMatrix separates the amino acid sequence of T cell 2.5 sequencing of tumor-normal pairs stimulate T cell-mediated processes that lead to tumor regression. p<0.001 p=0.008 A) Peptides documented epitopes into TCR-facing residues (epitope) and HLA 2 as IL-10-positive in Although neo-epitope prediction using computational methods rapidly identifies epitope binding cleft-facing residues (agretope), then compares the candidates in the mutanome, a large proportion of neo-epitopes prove to be non- 1.5 IEDB have significantly TCR face to other putative T cell epitopes. 1 immunogenic. Innovative computational tools validated for infectious disease targets can be applied higher potential for reactivity potential reactivity

- 0.5 to enhance design of personalized cancer immunotherapies by classification of predicted epitopes by Human (HG,

Cross-reactive peptides: Janus Human Average Score 0 above) and Human potential for mounting a tumor-specific response. Tumor-specific epitopes with potential to • Are predicted to bind the same MHC allele. Cross IL-10 IL-4 stimulate naïve T cells and epitopes that may cross-react with -experienced T cells • Share same/similar T cell-facing residues. A Cytokines B Microbiome (HM, below) cross-reactivity raised in infection are valuable targets. Epitopes with potential to activate regulatory T cells trained Avg Janus Microbiome Score vs +/- Cytokine Release IEDB Positive IEDB Negative as measured by on self and commensal are contraindicated. TCR cross-reactivity prediction: 30 p<0.001 p=0.004 JanusMatrix. • Given a protein or peptide, T cell epitopes are 25 20 We developed the JanusMatrix algorithm that parses query sequences into MHC-facing and T cell identified based on MHC contacts (P1, P4, P6, B) Peptides documented 15 as IL-4-positive in IEDB receptor (TCR)-facing sequences and screens sequence databases to identify MHC ligands that P9 for Class II) using EpiMatrix. 10 reactivity potential reactivity

- have significantly lower share TCR faces with host-related proteins. A database of human protein sequences is available • JanusMatrix searches for potentially cross- 5 potential for HG and to identify tumor-specific epitopes that may reduce anti-tumor activity by sequences that 0 activate reactive TCR by screening TCR-facing residues Janus Microbiome Average Score Cross IL-10 IL-4 HM cross-reactivity as regulatory T cells (Tregs) trained in the thymus on self-antigens. Similarly, tumor-specific epitope (P2, P3, P5, P7, P8 for Class II) against a A Cytokines B measured by candidates are screened using databases composed of antigens derived from human commensals or preloaded, EpiMatrix-processed reference Regulatory Inflammatory JanusMatrix. pathogens to identify epitopes that, respectively, may detrimentally or beneficially cross-react with T database. cytokine cytokine cells raised over the course of an individual’s immune history. Neo-epitope candidates are ranked according to MHC binding potential and TCR face homology. iVAX Neo-epitope Selection Pipeline Normal Peptide Mutated Peptide We conducted retrospective analyses of efficacy studies performed in mice vaccine candidate non-binder MHC binder • EpiMatrix identifies neo-epitopes by [Kreiter et al. 2015 Nature 520, 692–696] showing that mutanome-directed vaccines effective at neo-epitope non-self analyzing the binding potential of normal preventing tumor growth contained higher numbers of class I and II MHC neo-epitopes, as identified and mutated peptides. by EpiVax’s iVAX platform, and had lower potential to cross-react with other murine proteins. An • iVAX performs TCR-face analyses to evaluation of mutanomes derived from non-small cell lung cancer patients [Rizvi et al. 2014 TCR face self screen out mutated epitopes that share Science 348, 124-128] confirmed these findings. Improved clinical outcomes were observed in different TCR-face neo-epitope JanusMatrix analysis identical TCR-faces with normal epitopes. patients with mutanomes enriched in class I MHC neo-epitopes with TCR faces distinct from other eliminate TCR-face analysis human-derived epitopes. MHC face • Neo-epitopes highly cross-reactive with other self proteins are filtered out with identical TCR-face eliminate These results highlight the benefits of in silico prediction tools for the selection of neo-epitopes and JanusMatrix. how they can improve the design and efficacy of future cancer vaccines. While retrospective in nature, the suite of tools used for these analyses have been extensively validated in prospective Immunoinformatic Analysis of Cancer Vaccines from Kreiter et al. Nature 2015 vaccine studies for infectious disease. Removal of Treg epitopes, identified by JanusMatrix, has led

to the development of more immunogenic vaccine antigens. EpiVax’s H7N9 influenza Treg epitope- Adapted from Extended Data Figure 3b. Class I Neo-epitope content Class II Neo-epitope content • Pentatopes were constructed from Kreiter et al., Nature 520, 692–696. 2015 depleted vaccine is scheduled for Phase I clinical trial. Direct application of JanusMatrix to the mutated sequences derived from colon 3 2 neo-epitope-driven cancer vaccine discovery and design pipeline will focus candidate carcinoma model CT26 (study ) 2 selection on higher-value sequences than what conventional T cell epitope mapping algorithms performed by Kreiter et al. Nature 520, 1 epitopes iVAX - generate. Neo-epitopes with low Treg activation potential may then be used to support development 692-696. 2015). ( 1

of personalized therapies including vaccination and in vitro expansion of tumor infiltrating Neo - epitope per sequence • Pentatope2 showed greater antitumor - epitope per sequence lymphocytes for adoptive cell transfer. 0 0 Neo activity compared to Pentatope1 in Neo Pentatope 1 Pentatope 2 CT26-M19 Pentatope 1 Pentatope 2 CT26-M19 BALB/c mice inoculated with CT26 Class I Murine cross-reactivity Class II Murine cross-reactivity Mutanome-Directed tumor cells. potential potential • Pentatope2 increased anti-tumor activity 4 4 ) Tumor Biopsy Expressed Mutation Discovery Neo-Antigen Discovery Immunotherapy may be explained by a higher number 3 3

+ iVAX 2 2

CD8 ( Mutanome Tumor of Class I and II neo-epitopes and T cell cell 1 1 reactivity potential reactivity

overall lower potential for cross- - Murine Janus Score reactivity, as identified by iVAX. 0 Murine Janus Score 0 IMMUNOINFORMATICS Pentatope 1 Pentatope 2 CT26-M19 Pentatope 1 Pentatope 2 CT26-M19 Cross

Neo-epitopes Immunoinformatic Analysis of Mutanome Datasets from Rizvi et al. Science 2015 Normal Tumor CD4+ Fig. 3. Rizvi et al. Science 348, 124-128. 2015 T cell APC • Tumors from 34 anti-PD-1 treated patients with non-small cell lung cancer were sampled and sequenced to identify tumor-specific mutations (study performed by Rizvi et al. Science 348, 124-128. 2015). Tumor mutation burden was found to be associated with clinical response (left). • Mutanomes were analyzed using iVAX to identify neo-epitopes and to determine their cross-reactivity potential with other human proteins (below). • Neo-epitope content, as determined by iVAX, was significantly associated with clinical benefit. Association was improved when removing patients not expressing PD-L1 for whom the anti-PD-1 treatment may not be suitable. Acquire Mutanome • Progression-free survival (PFS) of patients expressing some level of PD-L1 was best estimated by iVAX’s neo-epitope content, adjusted for both Human cross- conservation potential (“low Janus”) and PD-L1 expression. A deduction is applied to patients with “weak” PD-L1 expression Scan for HLA Matched Epitopes with EpiMatrix 18 18 [Patient Class I + Class II] 16 16 14 14 Outlier patients do not express PD-L1 12 12 10 10 8 p=0.002 8 Screen for Humanness with JanusMatrix 6 6 PFS (months) PFS PFS (months) PFS ) 4 4 2 2 R² = 0.52 R² = 0.67 0 0 Annotate 0 200 400 600 800 0 200 400 600 800 Neoantigen count (Rizvi et al.) Low Janus neo-epitope count (adjusted for PD-L1) (iVAX) Prevalence/Published Improved association when using iVAX Remove PD-L1-negative patients 18 18 p=6.8e-5 16 16 14 14 Rank and Select Peptide Candidates 12 12 epitope abundance ( iVAX abundance epitope - 10 10 8 8 Neo 6 6 Design String of Beads Construct (months) PFS 4 (months) PFS 4 2 2 using VaxCAD R² = 0.53 R² = 0.59 0 0 NDB: No Durable Benefit 0 500 1000 1500 0 200 400 600 800 DCB: Durable Clinical Benefit Better outcome Neo-epitope count (iVAX) Low Janus neo-epitope count (iVAX) Conclusions

. Sequence analysis of the MHC- and TCR-facing residues of T cell epitopes enables prediction of epitope phenotype. . Epitopes that share a TCR-face with numerous human sequences may activate Tregs. . Sharper definition of neo-antigens by immunoinformatic analysis may improve epitope selection for mutanome-directed cancer immunotherapy. . In silico analyses of mutanomes and cancer vaccines suggest that high neo-epitope density and low cross-reactivity potential may be indicative of improved clinical outcomes. References • Moise L, Gutierrez A, Kibria F, Martin R, Tassone R, Liu R, Terry F, Martin B, De Groot AS. iVAX: An integrated toolkit for the selection and optimization of antigens and the design of epitope-driven vaccines. Hum Vaccin Immunother. 2015;11(9):2312-21. • Moise L, Gutierrez AH, Bailey-Kellogg C, Terry F, Leng Q, Abdel Hady KM, VerBerkmoes NC, Sztein MB, Losikoff PT, Martin WD, Rothman AL, De Groot AS. The two-faced T cell epitope: examining the host-microbe interface with JanusMatrix. Hum Vaccin Immunother. 2013 Jul;9(7):1577-86. • Kreiter S, Vormehr M, van de Roemer N, Diken M, Löwer M, Diekmann J, Boegel S, Schrörs B, Vascotto F, Castle JC, Tadmor AD, Schoenberger SP, Huber C, Türeci Ö, Sahin U. Mutant MHC class II epitopes drive therapeutic immune responses to cancer. Nature. 2015 Apr 30;520(7549):692-6. • Rizvi NA, Hellmann MD, Snyder A, Kvistborg P, Makarov V, Havel JJ, Lee W, Yuan J, Wong P, Ho TS, Miller ML, Rekhtman N, Moreira AL, Ibrahim F, Bruggeman C, Gasmi B, Zappasodi R, Maeda Y, Sander C, Garon EB, Merghoub T, Wolchok JD, Schumacher TN, Chan TA. Mutational landscape determines sensitivity to PD-1 blockade in non–small cell lung cancer. Science. 2015 Apr;348(124):124-8 For questions regarding in silico antigen screening and vaccine design, please contact: Steven Vessella at 401-272-2123, ext. 107; or at [email protected] www.epivax.com