Abstract 4407: Leveraging spatial profiling combined with bulk RNA sequencing to study patient-specific immunotherapeutic mechanisms

Abstract

Abstract Introduction: Mechanisms of resistance to adoptive T-cell transfer therapy (ACT) in clinical samples have been identified using bulk tumor sequencing, gene expression profiling, and immunohistochemistry. Furthermore, these datatypes are independently used to predict response to immunotherapy. However, this is often done using relative values quantified across clinical cohorts, which is not always feasible in single sample case analyses. Samples and Methods: A patient (INY3) presented with a synovial sarcoma lesion (INY3-P) and had an objective response (OR) to ACT targeting NY-ESO-1. A lung metastasis (INY3-M) progressed 10 months post-treatment, and she had a second OR to ACT plus ipilimumab. Seven months later, the primary lesion recurred (INY3-R). RNAseq was performed on snap-frozen biopsies at each timepoint. Two formalin-fixed, paraffin-embedded slides from each timepoint were assessed by either the Nanostring GeoMx Digital Spatial Profiling (DSP) 60-plex protein or 84-plex RNA panel. Regions-of-interest (ROIs) were selected for those positive for NY-ESO-1, pan-CK, CD45, or combinations of these proteins. We evaluated the correlation of DSP targets across ROIs within each sample, correlation of matched protein/RNA DSP targets across timepoints, the impact of DSP heterogeneity on bulk tumor RNAseq, and DSP patterns in ACT-resistant samples. Results: The Spearman correlation across ROIs ranged from -0.57-0.84 (median -0.14) in DSP protein and -0.59-0.75 (median -0.11) in DSP RNA expression. The variance in correlation was greatest in INY3-M (0.18 protein; 0.11 RNA), compared to INY3-P (0.08 protein; 0.13 RNA) and INY3-R (0.07 protein; 0.02 RNA), suggesting greater differences in ROIs selected in INY3-M. This was confirmed by lowest correlation between DSP RNA expression with bulk RNAseq in INY3-M (0.43-0.76 across ROIs; INY3-P, 0.31-0.61; INY3-R, 0.67-0.79). The median correlation between matched protein/RNA targets (N=31) in each ROI was 0.41 (-0.17-0.90), establishing some differences in RNA and protein expression. This was confirmed by overall lower correlation of DSP protein expression with bulk RNA expression (0.32-0.65 across ROIs; median 0.61). There were three patterns observed in ACT-resistant samples: increased CTNNB1 (INY3-M); lower expression of T cell proteins, as well as B2M, HLA-DRB, CXCL9, and PSMB10 (INY3-M, INY3-R); and increased expression of T cell surface proteins, and antigen presentation machinery (INY3-M, INY3-R). Conclusions: Bulk tumor RNAseq analysis is complicated by tumor heterogeneity, which can be overcome by DSP. Collectively, our results support complementary use of these technologies to better study individual samples, without requiring comparison to clinical cohorts, as well as address the complexities of immunotherapeutic mechanisms, identifying cell type-specific signals that may be lost using bulk RNAseq. Citation Format: Katie M. Campbell, Theodore S. Nowicki, Antoni Ribas. Leveraging spatial profiling combined with bulk RNA sequencing to study patient-specific immunotherapeutic mechanisms [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 4407.