Novel transcriptional signatures associated with antitumor activity in vidutolimod (vidu)-treated patients (pts) with anti-PD-1-refractory melanoma and non-small cell lung cancer (NSCLC).

Abstract Background: Vidu is a first-in-class CpG-A TLR9 agonist in a virus-like particle that activates plasmacytoid dendritic cells (pDC), thus bridging innate and adaptive immunity. Intratumoral (IT) vidu alone or in combination with intravenous (IV) anti-PD-(L)1 has shown evidence of antitumor activity in pts with anti-PD-(L)1-refractory melanoma or NSCLC. As biomarkers associated with anti-PD-(L)1 response were not predictive of vidu activity, we sought to identify novel transcriptional signatures. Methods: RNA-Seq was performed on baseline biopsies from pts with anti-PD-(L)1-refractory melanoma (NCT02680184; RECIST v1.1 responders [R; n=20]; nonresponders [NR; n=78; 30 pts with stable disease + 48 pts with progressive disease [PD]]) or NSCLC (NCT03438318; N=11) treated with vidu ± anti-PD-(L)1. IFNg18 signature was used to characterize melanomas as non-T cell-inflamed (non-Tinfl), T cell-inflamed (Tinfl), or intermediate. Gene Set Enrichment Analysis (GSEA) using MSigDB and other signatures (>19,000) was performed on non-Tinfl melanoma in R vs NR (false discovery rate [FDR] <0.25). Deconvolution of immune cells was performed using TIMER2.0. Prediction models were generated using QLattice. Signatures were characterized using both publicly available bulk and single-cell (sc) RNA-Seq datasets of pDC subsets or PD-1/CTLA-4 blockade response datasets. Results: GSEA on non-Tinfl melanoma biopsies revealed 2 gene signatures (COPII vesicle and Golgi targeting) most strongly associated with R (FDR <0.16). Leading edge analysis of these signatures identified 35 common core genes (CC) that strongly differentiated RECIST 1.1 R vs PD. As independent validation, CC enrichment was also significantly associated with R in intermediate melanoma (p=0.009) and with tumor shrinkage in NSCLC (p=0.027). CC was significantly associated with R to vidu single-agent or combination treatment, but not with clinical baseline prognostic factors or IFNg18. In Tinfl melanoma, CC was not associated with R, but myeloid signatures were significantly associated with NR. A model based on CC and transcription factor ELF2 predicted R in melanoma (AUC 0.93 [95% CI 0.82-1.00]). In public datasets, CC was not associated with R to PD-1/CTLA-4 blockade, but CC was highly expressed in the type I interferon-secreting subset of pDCs and some myeloid cells. CC expression was prevalent in most tumor types and was independent of IFNg18 in TCGA. Conclusion: In pts with anti-PD-1-refractory melanoma or NSCLC, transcriptional signatures of COPII vesicle and Golgi targeting (functionally related to TLR9 activation) were associated with antitumor activity of IT vidu ± IV anti-PD-(L)1. Ongoing bulk and sc RNA-Seq analyses may clarify the underlying biology and the signature’s potential role as a predictive biomarker for clinical response to vidu. Citation Format: Hong Liu, Luping Zhao, Ping Zheng, Riyue Bao, Jason J. Luke, Marcelo V. Negrao, Shakoora A. Sabree, George J. Weiner, Sujatha Kumar, Dmitri Bobilev, James E. Wooldridge, Arthur M. Krieg. Novel transcriptional signatures associated with antitumor activity in vidutolimod (vidu)-treated patients (pts) with anti-PD-1-refractory melanoma and non-small cell lung cancer (NSCLC) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr LB107.