High-Plex Predictive Marker Discovery for Melanoma Immunotherapy-Treated Patients Using Digital Spatial Profiling

Abstract

Purpose: Protein expression in formalin-fixed, paraffin-embedded tissue is routinely measured by IHC or quantitative fluorescence (QIF) on a handful of markers on a single section. Digital spatial profiling (DSP) allows spatially informed simultaneous assessment of multiple biomarkers. Here we demonstrate the DSP technology using a 44-plex antibody cocktail to find protein expression that could potentially be used to predict response to immune therapy in melanoma.Experimental Design: The NanoString GeoMx DSP technology is compared with automated QIF (AQUA) for immune marker compartment-specific measurement and prognostic value in non-small cell lung cancer (NSCLC). Then we use this tool to search for novel predictive markers in a cohort of 60 patients with immunotherapy-treated melanoma on a tissue microarray using a 44-plex immune marker panel measured in three compartments (macrophage, leukocyte, and melanocyte) generating 132 quantitative variables.

Results: The spatially informed variable assessment by DSP validates by both regression and variable prognostication compared with QIF for stromal CD3, CD4, CD8, CD20, and PD-L1 in NSCLC. From the 132 variables, 11 and 15 immune markers were associated with prolonged progression-free survival (PFS) and overall survival (OS). Notably, we find PD-L1 expression in CD68-positive cells (macrophages) and not in tumor cells was a predictive marker for PFS, OS, and response.

Conclusions: DSP technology shows high concordance with QIF and validates based on both regression and outcome assessment. Using the high-plex capacity, we found a series of expression patterns associated with outcome, including that the expression of PD-L1 in macrophages is associated with response.

Figure 1:. NanoString DSP to QIF comparison in Yale NSCLC cohort A.

Regression of NanoString DSP counts to QIF scores for (A) CD3 in stroma (B) CD4 in stroma (C) CD20 in stroma (D) CD8 in stroma (E) PD-L1 in tumor (F) PD-L1 in stroma (G) CD3 in tumor (H) CD4 in tumor and (I) CD8 in tumor.

Figure 2:. Prognostic value of NanoString DSP and QIF in Yale NSCLC cohort B.

Kaplan Meier 5-year survival curves of NSCLC patients in Yale cohort B, stratified by median tumor and stroma CD3 expression measured by QIF and NanoString DSP. (A) CD3 in tumor by QIF, p=0.0019 HR:0.41, 95% CI: 0.24–0.72 (B) CD3 in stroma by QIF, p=0.036 HR: 0.55, 95% CI: 0.32–0.96 (C) CD3 in tumor by NanoString DSP, p=0.034 HR: 0.54, 95% CI: 0.30–0.95 (D) CD3 in stroma by NanoString DSP, p=0.26 HR: 0.73 95% CI: 0.42–1.27. Survival analysis by log-rank (Mantel-Cox) test.

Figure 3:. Representative images of NanoString DSP compartment selection in melanoma cohort C.

(A) H&E image of a representative TMA spot (B) Low resolution immunofluorescence image of the markers that define the selected compartments. Melanocytes are in green, CD68+ is in blue and CD45+ is in purple. Selection of (C) CD68+ compartment, (D) CD45+ compartment, (E) tumor compartment and (F) remaining DNA+ compartment.

Figure 4:. Candidate predictive markers in immunotherapy treated melanoma cohort C by NanoString DSP.

Kaplan Meier 5-year survival and progression free survival curves of immunotherapy treated melanoma patients in Yale cohort C. (A) OS by CD8 counts in CD68+ compartment, p=0.0119 (B) PFS by CD8 counts in CD68 compartment, p=0.0082 (C) Response to ICIs by CD8 counts in CD68+ compartment, p=0.014 (D) OS by PD-L1 in CD68+ compartment, p=0.0032 (E) PFS by PD-L1 in CD68+ compartment, p=0.0072 (F) Response to ICIs by PD-L1 in CD68+ compartment, p=0.011 (G) OS by PD-L1 in tumor compartment, p=0.072 (E) PFS by PD-L1 in tumor compartment, p=0.054 (F) Response to ICIs by PD-L1 in tumor compartment. Survival analysis by log-rank (Mantel-Cox) test. Two-tailed Mann-Whitney U test, bars represent means with standard deviation. *Denotes statistical significance p<0.05.