Abstracts J Immunother : first published as 10.1136/jitc-2020-SITC2020.0058 on 9 November 2020. Downloaded from

58 EXCLUDING TREG AND INTEGRATING 59 INTEGRATING DEEP PROTEOMICS PROFILING WITH CD8 AND CD4 EFFECTOR NEOEPITOPE CONTENT SURVIVAL ANALYSIS TO IDENTIFY NOVEL BIOMARKERS IMPROVES PROGNOSTIC BIOMARKER TOOL IN OF RESPONSE TO PD-1 BLOCKADE IN NSCLC PATIENTS BLADDER CANCER 1Kamil Sklodowski*, 1Vito Dozio, 2Silvia Lopez-Lastra, 1Andrés Lanzós, 1Kristina Beeler, 1Guilhem Richard*, 2Gary Steinberg, 3Tzintzuni Garcia, 4Matthew Ardito, 4William Martin, 2Emanuela Romano. 1Biognosys AG, Schlieren, Switzerland; 2Institut Curie, Paris, France 1Gad Berdugo, 1Michael Princiotta, 2Arjun Balar, 4Anne de Groot, 3Randy Sweis. 1EpiVax Oncology, Inc., Providence, RI, USA; 2NYU Langone Health, New York, NY, USA; 3University Background Immune checkpoint inhibitors have improved clin- of Chicago, Chicago, IL, USA; 4EpiVax, Inc., Providence, RI, USA ical responses and overall survival for patients with non-small cell lung cancer (NSCLC). However, the response is not equal Background Improvement of current prognosis biomarkers and known NSCLC biomarkers are not sufficient in predicting will enhance our ability to identify cancer patients at higher therapy outcome. Deep proteomic analysis of NSCLC patient‘s risk of recurrence and will further advance the personaliza- plasma treated with anti-PD-1-blockade using a state-of-the-art tion of patient monitoring and treatment. We hypothesized data independent acquisition mass spectrometry (DIA-MS) is a that the presence of a alone is not sufficient to powerful and unbiased way of identifying signatures generate an immunogenic neoepitope, but that significant dif- associated with disease stage or response to treatment. How- ferences must exist between the Human Leukocyte ever, to unravel these associations large-scale omics data (HLA)- and/or Receptor (TCR)-interfaces of the neoe- should be analyzed with respect to available clinical informa- pitope and its non-mutated form, or with other self-epitopes, tion. To achieve this goal, we have used an approach previ- in order to be recognized as non-self by the . ously applied by Uhlen et al., 20171 for transcriptomic As such, cancer patient clinical outcomes may be better datasets. In this approach survival data is used to set the most understood by neoepitope analyses that integrate these optimal thresholds for candidate biomarkers. considerations. Methods 125 plasma samples were analyzed by capillary flow Methods We analyzed large scale (n=412) bladder cancer liquid chromatography coupled to DIA-MS. Data were genomic data from The Cancer Genome Atlas (TCGA) using extracted with latest SpectronautTM and were quanti- Ancer, an automated machine-learning-based pipeline we fied. Each recorded protein intensity was used as a threshold designed for neoantigen screening and design. Ancer for two groups of samples for which Kaplan-Meier estimates shares components with other commercial-grade screening were generated using ‘survival’2 package in R. Benjamini- platforms used routinely in assessments of Hochberg correction was applied and p-values with corre- biologics and infectious disease , such as the EpiMa- sponding intensity cut-offs were extracted to generate panels trix algorithm for HLA-I and HLA-II neoepitope identifica- of potential biomarkers. copyright. tion, and the JanusMatrix algorithm for tolerated, Results 125 plasma samples (in total 75 baseline and 50 after tolerogenic, and cross-reactive T cell identification. 8-weeks treatment) from advanced NSCLC patients treated Evaluation of patient survival with Ancer was compared to with an anti-PD-1 inhibitor following at least 1 prior line of other analyses employing tumor mutational burden (TMB) treatment were analyzed. 727 unique proteins were quantified or neoepitopes identified with the commonly used across all samples. Data analysis was performed separately for NetMHCpan-4.0 and NetMHCIIpan-3.1 T cell epitope pre- each line of treatment and treatment status resulting in more diction tools. than 100’000 p-values. For each group, panels of proteins Results We stratified bladder patients based on their Ancer with best performance in separating progression free survivals HLA-I and HLA-II neoepitope burdens and observed signifi- were defined at FDR of 0.10, giving 64 unique proteins cantly prolonged disease free and overall survival in patients which were mapped to acute phase response, platelet degranu- whose tumor contained both high numbers of HLA-I and lation and complement activation. Several of these proteins http://jitc.bmj.com/ HLA-II neoepitopes compared to other individuals. Stratifica- were listed in the Early Detection Research Network database tions performed with Ancer were superior to comparative of the National Cancer Institute, and one of them – LYPD3, analyses performed with TMB or with NetMHCpan and was a potential therapeutic target in a preclinical study for NetMHCIIpan. In addition, we showed that Ancer’s precise NSCL treatment.3 Selected proteins were then used to cluster filtering and characterization of mutated epitopes contributed patients into cohorts that showed association with the to this increased association with survival, as showcased by response to therapy. on September 25, 2021 by guest. Protected gradual improvements in survival analyses performed after Conclusions Deep proteomic profiling of plasma samples using each of its filtering step. Multivariate survival analyses indi- DIA-MS in conjunction with clinical outcome enables a holis- cated that Ancer neoepitope content remained a significant tic and stringent analysis of potential circulating biomarkers. factor in patient overall survival even when adjusted for TMB, Such analysis generates functional insights into the plasma pro- and other clinical covariates such as age at diagnosis and dis- teome that enable deeper understanding and comprehensive ease stage, unlike analyses involving NetMHCpan and integration of clinical data with proteomics markers at differ- NetMHCIIpan neoepitopes. ent disease stages and treatment phases. Conclusions Our analysis suggests that enhanced presence of CD8, CD4 T cell epitopes, and limited inclusion of Treg epit- REFERENCES opes in the tumor genome plays a key role in cancer survival. 1. Uhlen M, Zhang C, Lee S, Sjöstedt E, Fagerberg L, Bidkhori G, Benfeitas R, Arif Ancer scoring provides a predictive method for predicting M, Liu Z, Edfors F, Sanli K, von Feilitzen K, Oksvold P, Lundberg E, Hober S, Nils- son P, Mattsson J, Schwenk J. patient outcomes, by defining the number of true neoepitopes 2. Therneau TM, Grambsch PM. Modeling Survival Data: Extending the Cox Model. and by identifying Treg epitopes that would interfere with T Springer. 2000, New York, ISBN 0-387-98784-3. cell-based immune activation and response to the tumor. 3. Willuda J, Linden L, Lerchen H, Kopitz C, Stelte-Ludwig B, Pena C, Lange C, Gol- fier S, Kneip C, Carrigan P E, Mclean K, Schuhmacher J, von Ahsen O, Müller J, http://dx.doi.org/10.1136/jitc-2020-SITC2020.0058 Dittmer F, Beier R, El Sheikh S, Tebbe J, Leder G, Apeler H, Jautelat R, Ziegelba- uer K, Kreft B, Preclinical Antitumor Efficacy of BAY 1129980-a Novel Auristatin-

J Immunother Cancer 2020;8(Suppl 3):A1–A559 A37