While numerous anti-angiogenic and immune targeting therapies have become standard-of-care treatments for oncology, predictive biomarkers for these agents have been either entirely lacking or challenged by inconsistencies across indications. We have developed and validated the Xerna TME Panel as a novel machine learning-based RNA-sequencing biomarker assay that guides patient selection for tumor microenvironment (TME)-targeted therapies across multiple tumor types. Gene expression data sets from both public sources and clinical practice representing over 5000 samples across 7 different tumor types were analyzed using the Xerna TME Panel. The Xerna TME Panel consists of an artificial neural net that learns complex gene expression interactions between angiogenesis and tumor immune biologies and robustly classifies patient samples into one of four TME biomarker subtypes: Angiogenesis (A), Immune Suppressed (IS), Immune Active (IA), or Immune Desert (ID). The vast majority (>75%) of all samples were assigned a TME class designation with confidence scores in the upper quartile and had nearly bimodal distributions for biomarker-positive versus -negative classifications. When compared to other independent gene signatures, such as those describing angiogenesis/mesenchymal biology, inflammation, and immune suppression, the expression profiles from the Xerna TME subtypes showed enrichment of those biological processes. Each TME subtype represented between ~15-40% of subjects of each tumor type, indicating balanced representation of subgroups within the patient populations. The Xerna TME designations were prognostic across tumor types, with “A” tumors generally associated with the worst survival and “IA” tumors associated with the best survival. The predictive ability of the Xerna TME Panel to enrich for tumor responses to targeted therapies in gastric cancer was also evaluated. In a ramucirumab+paclitaxel clinical cohort, the Xerna TME Panel high Angiogenesis score tumors (A and IS) demonstrated a 48% response rate compared to a 31% for low Angiogenesis score tumors (IA and ID). In an immune checkpoint inhibitor (ICI) cohort, high Immune score tumors (IA and IS) showed a response rate of 34% vs. 5% for low Immune score tumors (A and ID). Within the microsatellite stable patients (MSS), which historically have low response rates to ICIs, the Xerna TME Panel was able to enrich for responses between Immune high vs. Immune low score patients (25% vs. 3%). Currently in use to prospectively enroll patients into a Phase 3 ovarian cancer clinical trial and in development as a companion diagnostic (CDx) assay, the Xerna TME Panel is a robust, pan-cancer biomarker assay capable of characterizing TME dominant biologies to further advance the matching of patients with targeted therapeutics. Citation Format: Seema Iyer, Luka Ausec, Daniel Pointing, Matjaz Zganec, Robert Cvitkovic, Miha Stajdohar, Valerie Chamberlain Santos, Kerry Culm, Mokenge Malafa, Jeeyun Lee, Rafael Rosengarten, Laura Benjamin, Mark T. Uhlik. Xerna࣪ TME Panel: A pan-cancer RNA-based investigational assay designed to predict patient responses to angiogenic and immune targeted therapies [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 1232.

IntroductionMost predictive biomarkers approved for clinical use measure single analytes such as genetic alteration or protein overexpression. We developed and validated a novel biomarker with the aim of achieving broad clinical utility. The Xerna™ TME Panel is a pan-tumor, RNA expression-based classifier, designed to predict response to multiple tumor microenvironment (TME)-targeted therapies, including immunotherapies and anti-angiogenic agents.MethodsThe Panel algorithm is an artificial neural network (ANN) trained with an input signature of 124 genes that was optimized across various solid tumors. From the 298-patient training data, the model learned to discriminate four TME subtypes: Angiogenic (A), Immune Active (IA), Immune Desert (ID), and Immune Suppressed (IS). The final classifier was evaluated in four independent clinical cohorts to test whether TME subtype could predict response to anti-angiogenic agents and immunotherapies across gastric, ovarian, and melanoma datasets.ResultsThe TME subtypes represent stromal phenotypes defined by angiogenesis and immune biological axes. The model yields clear boundaries between biomarker-positive and -negative and showed 1.6-to-7-fold enrichment of clinical benefit for multiple therapeutic hypotheses. The Panel performed better across all criteria compared to a null model for gastric and ovarian anti-angiogenic datasets. It also outperformed PD-L1 combined positive score (>1) in accuracy, specificity, and positive predictive value (PPV), and microsatellite-instability high (MSI-H) in sensitivity and negative predictive value (NPV) for the gastric immunotherapy cohort.DiscussionThe TME Panel’s strong performance on diverse datasets suggests it may be amenable for use as a clinical diagnostic for varied cancer types and therapeutic modalities.

Ethics ApprovalThe study was approved by WCG IRB Ethics Board, approval number 20181863.

Background: Few therapeutic options exist for anti-PD-1 refractory, metastatic melanoma patients, and today’s biomarkers are insufficient to aid in defining who should receive potential combinatorial immunotherapies. Results from a phase Ib, multicenter study (NCT02680184) showed that a combination of vidutolimod and pembrolizumab provided a best overall response of 23.5% in patients with metastatic or unresectable cutaneous melanoma who had received prior anti-PD-1 therapy. The Xerna™ TME Panel consists of an artificial neural net that utilizes the expression of ~100 genes involved in angiogenesis and tumor immune biologies to classify patient samples into one of four tumor microenvironment (TME) biomarker subtypes: Angiogenesis (A), Immune Active (IA), Immune Desert (ID), or Immune Suppressed (IS). We hypothesized that the IS subtype is predictive of vidutolimod + pembrolizumab benefit in this cohort compared to the other TME subtypes (A, IA, and ID). Methods: Total RNASeq was performed on FFPE biopsies collected from a subset of patients prior to therapy (N=38) and 2 weeks post-initiation of therapy (N=10). Gene expression data was analyzed using the Xerna TME Panel algorithm to assign a TME subtype. Correlational analyses between TME subtypes, response to therapy, and other hallmark gene signatures were performed. Results: Overall response rate in the pretreatment cohort available for biomarker analysis was 26%, comparable to the entire vidutolimod/pembrolizumab arm. The cohort had a skewed distribution of TME subtypes with high prevalence of IS (34%) and ID (45%), indicative of immune therapy-refractory biologies. An overall response of 54% was observed in the IS subtype, compared with 12% in the other subtypes combined. Comparison with other “hallmark” gene signatures confirmed enrichment of immune and angiogenesis biologies in the IS subtype, but none of these individual hallmark signatures were found to differentiate between responders and non-responders. The Xerna TME Panel demonstrated superior classification performance across all criteria compared to a baseline classifier, including accuracy (0.76 vs. 0.62), sensitivity (0.70 vs. 0.27) and specificity (0.79 vs. 0.74). Among matched post-treatment samples, 70% revealed a change in TME subtype compared to their pre-treatment status. Three of the post-treatment samples represented changes from an immune-low subtype to an immune-high subtype, including one complete responder with a pre-treatment ID that changed to IA while on therapy, illustrating how the TME Panel may be used to interpret the mechanism of drug response. Conclusions: The Xerna TME Panel shows potential activity as a predictive and pharmacodynamic biomarker for the combination of vidutolimod and pembrolizumab in anti-PD-1 refractory melanoma patients. Citation Format: Mark T. Uhlik, Daniel Pointing, Luka Ausec, Miha Stajdohar, Robert Cvitkovic, Matjaz Zganec, Seema Iyer, Hong Liu, Art Krieg, Laura Benjamin. The XernaTM TME Panel potentially predicts response to a combination of the TLR9 agonist vidutolimod and PD-1 inhibitor pembrolizumab in metastatic melanomas with prior anti-PD-1 treatment [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 3270.