Abstracts J Immunother Cancer: first published as 10.1136/jitc-2020-SITC2020.0078 on 10 December 2020. Downloaded from

resistance to single-agent PD-1 inhibitors and is suggestive of potential treatment strategies. Validation is warranted in future clinical trials (NCT03412773). Acknowledgements This study was sponsored by BeiGene, Ltd. Writing and editorial assistance was provided by Regina Swit- zer, PhD, and Elizabeth Hermans, PhD (OPEN Health Medi- cal Communications, Chicago. IL), and funded by the study sponsor. Trial Registration NCT02407990, NCT04068519

http://dx.doi.org/10.1136/jitc-2020-SITC2020.0077

Abstract 77 Figure 1 Median difference in signatures before and after sorafenib exposure 78 T-CELL, MHC I, AND TUMOR INTRINSIC GENE SIGNATURES PREDICT CLINICAL BENEFIT AND sorafenib treatment. Tislelizumab, an anti-PD-1 monoclonal RESISTANCE TO TISLELIZUMAB MONOTHERAPY IN antibody, has demonstrated single-agent antitumor activity in PRETREATED PD-+ UROTHELIAL CARCINOMA patients with advanced, previously treated HCC in two early 1Dingwei Ye*, 2Aiping Zhou, 3Qing Zou, 4Hanzhong Li, 5Cheng Fu, 6Hailong Hu, phase studies (NCT02407990, NCT04068519). Association of 7Jian Huang, 8Wei Shen, 8Yun Zhang, 8Xiaopeng Ma, 8Pei Zhang, 8Ruiqi Huang, biomarkers, including PD-L1 expression and 8Xiusong Qiu, 8Lilin Zhang, 9Feng Bi. 1Fudan University Shanghai Cancer Center, Shanghai, profiles (GEP), with response and resistance to tislelizumab China; 2Chinese Academy of Medical Sciences, Beijing, China; 3Jiangsu Cancer Hospital, were explored. Nanjing, China; 4Peking Union Medical College Hospital, Beijing, China; 5Liaoning Cancer Methods PD-L1 expression was evaluated on tumor cells (TC) Hospital and Institute, Shenyang, China; 6Second Hospital Tianjin Medical University, Tianjin, 7 8 using the VENTANA PD-L1 (SP263) assay in baseline tumor China; Sun Yat-sen University, Guangzhou, China; BeiGene (Beijing) Co., Ltd., Beijing, 9 samples collected before or after sorafenib treatment. GEP China; West China Hospital, Sichuan University, Changdu, China were assessed using the 1392-gene HTG GEP EdgeSeq panel. Background Tislelizumab, an anti-PD-1 monoclonal antibody, Signature scores were calculated using the Gene Set Variation demonstrated efficacy as monotherapy in patients with previ- Analysis package with publicly available gene signatures (GS). ously treated PD-L1+ urothelial carcinoma (UC) during a Wilcoxon -sum test was used to analyze differential gene phase 2 study (NCT04004221). Here, gene expression profiles signatures (DEG); GS association with PFS and OS was eval- copyright. correlating with response and resistance to tislelizumab treat- uated using Cox proportional hazards models. ment are reported. Results Single-agent tislelizumab demonstrated antitumor activ- Methods Gene expression profiling (GEP) was conducted on ity in advanced, previously treated HCC (ORR=13%; CB [PR baseline tumor samples from 100 Chinese patients with UC +SD >6 months]=31%, median PFS=3.3 months; median enrolled in the phase 2 tislelizumab study using a 1,392-gene OS=13.3 months). PD-L1+ (TC1%) prevalence and GEP panel by HTG EdgeSeq. Gene Signature (GS) scores were cal- showed different patterns in samples collected before and after culated using the Gene Set Variation Analysis package. Differ- sorafenib exposure (figure 1). While non-exposed samples ential gene expression (DEG) analysis was performed between (n=16) were enriched for immune suppressive signatures, sor- responders and non-responders using the Wilcoxon test; sur- afenib-exposed samples (n=41) showed higher PD-L1+ preva- vival was evaluated using the Cox proportional hazards model lence (53.7% vs 25%; P=0.08) and immune-cell activation and the odds of tumor response for subgroup analysis was signatures along with co-inhibition molecules. In sorafenib- http://jitc.bmj.com/ estimated by logistic regression. exposed samples, PD-L1 expression was positively correlated Results Of patients with available confirmed response results with CB (P=0.0027) and a trend of longer PFS (HR=0.56, (n=85), DEG analysis found that responders had significantly 95% CI:0.28–1.13). ORR was higher in PD-L1+ than PD-L1 higher T-cell GS (CD3D, CD3E, CD3G, CD6, SH2D1A, sorafenib-exposed samples (23.8% vs 0%; P=0.049). DEG TRAT1) (P=0.04) and MHC I GS (HLA-A, TAP1) (P=0.05), analysis in sorafenib-exposed samples demonstrated that NK- respectively. Using median GS scores as a cutoff, improvement mediated cytotoxicity GS was positively correlated with CB –

in overall survival (OS) was observed in T-cell high versus T- on September 30, 2021 by guest. Protected (P=0.03), as well as a trend of longer PFS (HR=0.43, 95% cell–low groups (P=0.01) and a trend of longer OS was seen CI:0.17–1.06). Across the different analyses, no correlation between MHC I–high versus MHC I–low groups. Patients in with OS was observed. Patients considered non-responders T-cell and MHC I–double-high subgroups showed further (NRs) were found clustered into three distinct GEP subgroups (NR1, NR2, NR3). Compared with responders, NR1 had enhanced angiogenesis signatures (P=0.01), including TEK, Abstract 78 Table 1 T-cell and MHC I gene signatures associated KDR, HGF, and EGR1. Despite high inflamed tumor signa- with clinical efficacy of tislelizumab in patients with UC tures, NR2 had increased expression of T-cell inhibition GS scores (P=0.01), including CD274, CTLA4, TIGIT, and CD96. The NR3 subgroup showed higher cell-cycle GS scores compared with responders (P=0.05), including E2F7, FOXA1, and FANCD2. Conclusions Prior sorafenib exposure appears to be associated with increased PD-L1 expression and tumor microenviron- ment-related GS, as well as response and PFS from tislelizu- mab in advanced HCC patients. Elevated angiogenesis, immune exhaustion, and cell-cycle GS levels may indicate

A48 J Immunother Cancer 2020;8(Suppl 3):A1–A559 Abstracts J Immunother Cancer: first published as 10.1136/jitc-2020-SITC2020.0078 on 10 December 2020. Downloaded from improvement in clinical efficacy (40% objective response rate Abstract 79 Table 1 Tumor-immune signatures associated with [ORR], 5.26 month median progression-free survival [PFS], clinical efficacy of tislelizumab in patients with ESCC and 15.2 month median OS) than other subgroups (table 1). In addition to immune-related in the microenvironment, DEG analysis also revealed that tumor-related genes were highly expressed in non-responders, such as intrinsic genes related to angiogenesis (VEGFA [P=0.07], KDR [P=0.07]), the mTOR pathway (MTOR [P=0.015]), and DNA damage repair (REV3L [P=0.007]). MTOR and REV3L were associ- ated with shorter PFS (P=0.02; P=0.003) and OS (P=0.03; P =0.008), respectively. Conclusions By using GEP, T-cell and MHC I GS were identi- fied as potentially predictive biomarkers of response to tisleli- zumab monotherapy in PD-L1+ UC in this retrospective analysis. By combining these two GS scores, patients with (as defined by median cutoff), the prediction of clinical efficacy optimal efficacy responses could be identified. Conversely, was further improved (table 1). In addition to Treg scores, non- high expression of tumor intrinsic genes related to angiogene- responders (NR) to tislelizumab monotherapy could be further sis and the mTOR pathway may indicate resistance and sug- clustered into four subgroups (NR1, NR2, NR3, and NR4), gest potential future drug combinations for these patients. each exhibiting distinct resistance signatures. Despite a high level Both findings warrant further validation in a phase 3 study of immune infiltration, NR1 expressed a higher exhaustion sig- (NCT03967977). nature (CD96, CTLA4, TIGIT, HAVCR2, etc.) versus responders Acknowledgements Editorial assistance was provided by Ste- (P=0.01). Both NR2 and NR3 demonstrated a trend of phan Lindsey, PhD, and Elizabeth Hermans, PhD (OPEN enhanced cell-cycle signatures versus responders (P=0.07 and Health Medical Communications, Chicago. IL), and funded by P=0.08, respectively), accompanied by a lower NK signature the study sponsor. (KIR2DS4, KIR.panL, CD56) in NR2 and a lack of immune Trial Registration CTR20170071 infiltration in NR3. In the NR4 subgroup, a trend toward higher TH17 (P<0.01) and IL-17F signatures (Log2FC=0.56, http://dx.doi.org/10.1136/jitc-2020-SITC2020.0078 P=0.10) versus responders was observed. GEP-evaluable patients (n=12) receiving tislelizumab in combination with chemotherapy

had an objective response rate of 58% (n=7), with a different copyright. gene signature pattern than those observed in patients receiving 79 TUMOR-IMMUNE SIGNATURES ASSOCIATED WITH monotherapy. Responders to combination therapy showed RESPONSE OR RESISTANCE TO TISLELIZUMAB (ANTI-PD- higher DNA repair expression versus NR (P=0.07), while 1) IN ESOPHAGEAL SQUAMOUS CELL CARCINOMA angiogenesis signatures were significantly higher in NR vs res- (ESCC) ponders (P=0.01). Consistent with this, NR exhibited higher 1 2 3 4 5 6 Jianming Xu, Nong Xu, Yuxian Bai, Chia-Chi Lin, Michael Millward, Jingwen Shi, expression of VEGFC at a single gene level (Log2FC=2.46, 6Yun Zhang, 6Xiaopeng Ma, 6Zhirong Shen, 6Ruiqi Huang, 6Wei Huang, 7Lin Shen*. 1The P <0.01). 2 Fifth Medical Center, General Hospit, BeiJing, China; The First Affiliated Hospital, College o, Conclusions While higher TLR signaling was associated with 3 4 Hangzhou, China; Harbin Medical University Cancer Hospita, Harbin, China; National clinical benefit of tislelizumab monotherapy, elevated Treg, Taiwan University Hospital, Taipei, Taiwan, Province of China; 5Linear Clinical Research exhaustion, cell cycle, and TH17 signatures may indicate Limited, Western Australia, Australia; 6BeiGene (Beijing) Co., Ltd., Beijing, China; 7Peking

University Cancer Hospital, Beijing, China resistance. Signatures predictive for combination therapy may http://jitc.bmj.com/ vary. Both immune- and tumor-related features may be consid- Background Tislelizumab, an anti-PD-1 monoclonal antibody, ered for validation in phase 3 studies (NCT03430843, showed promising clinical outcomes for patients with ESCC. NCT03783442). Here, the tumor and immune microenvironment is investigated Acknowledgements Editorial assistance was provided by Ste- using gene expression profiles (GEP) and gene signatures asso- phan Lindsey, PhD, and Elizabeth Hermans, PhD (OPEN ciated with clinical efficacy in patients with ESCC receiving Health Medical Communications, Chicago. IL), and funded by tislelizumab either as monotherapy (NCT02407990, the study sponsor. on September 30, 2021 by guest. Protected NCT04068519) or in combination with chemotherapy (5-fluo- Trial Registration NCT02407990, NCT04068519, rouracil plus cisplatin; NCT03469557). NCT03469557 Methods Baseline tumor samples were subjected to GEP using http://dx.doi.org/10.1136/jitc-2020-SITC2020.0079 a 1392-gene HTG EdgeSeq panel. Signature scores were calcu- lated using the Gene Set Variation Analysis package with pub- licly available gene signatures. Differential gene signature (DEG) analysis was performed between responders and non- 80 EVALUATION OF THE TRUSIGHT ONCOLOGY 500 ASSAY responders using the Wilcoxon rank-sum test. Associations FOR ROUTINE CLINICAL TESTING OF TUMOR between gene signatures and survival were evaluated using the MUTATIONAL BURDEN (TMB) AND CLINICAL UTILITY Cox proportional hazards model. FOR PREDICTING RESPONSE TO PEMBROLIZUMAB Results In GEP-evaluable patients receiving monotherapy Bo Wei*, John Kang, Miho Kibukawa, Gladys Arreaza, Maureen Maguire, Lei Chen, (n=43), DEG analysis showed toll-like (TLR) signature Ping Qiu, Lixin Lang, Deepti Aurora-Garg, Razvan Cristescu, Diane Levitan. Merck and Co., scores,drivenbyTLR8,TLR6,TIRAP,andTLR4,wereposi- Inc., Kenilworth, NJ, USA tively correlated with response and survival, while Treg scores, driven by FOXP3, EBI3, TNFRSF18, and BATF, showed a neg- Background Various biomarkers have been investigated for ative correlation. After combining TLR-high and Treg-low scores their ability to identify patients more likely to respond to

J Immunother Cancer 2020;8(Suppl 3):A1–A559 A49