Fig 1.
A: Illustrative representation of the pairwise Ripley’s K function, where the number of cells of interest is identified within a search radius from another cell type, repeating the process over a continuum of radii, for each cell in the study area. B: The Ripley’s K function estimator, where n is the total number of points in the study area, W is the study area, 1{|xi-xj|}≤r} is an indicator worth a value of 1 if points i and j are within distance r, and c(xi,xj;r) represents the applied edge correction (this analysis utilized isotropic edge correction). C: Plot of the naïve K(r) function using a representative slide from our analysis. The black solid line is the observed K(r), and dotted red line is the expected distribution if cells were randomly distributed, assuming a Poisson distribution. D: Normalization of the naïve Ripley’s K function into K(r)/πr2, resulting in an expected distribution of 1 for all values of r. The search radius utilized in this analysis was 25um. In this manuscript, K(25um)/πr2 is abbreviated to “nK(25)”.
Fig 2.
A: Project workflow, in brief. IHC staining for myeloid markers (CD68, CD163, CD206) and PD-L1. Digital pathologic analysis is utilized to convert IHC slides to point pattern plots, which are then utilized to calculate measures of spatial heterogeneity. B: Examples of point pattern plots demonstrating high or low CD68+ cell density (cells/mm2; cut-point = 151.273), and high or low Tumor/CD68+ cell clustering (nK(25), cut-point = 1.30).
Table 1.
Baseline patient characteristics at the time of sample collection.
Fig 3.
A: Scatter plot and Spearman’s correlation of CD68+ cell density (cells/mm2) and Tumor/CD68+ (nK(25)), per ROI (N = 129, r = -0.19, p = 0.046). B: Boxplot diagrams of Ripley’s nK(25) values for Tumor/CD68+, Stroma/CD68+, Tumor/CD163+, Stroma/CD163+, Tumor/CD206+, Stroma/CD206+, Tumor/PDL1+, and Stroma/PDL1+, for all ROIs (N = 129). nK(25) values > 1 indicate cellular clustering, and values <1 indicate cellular dispersion. Asterix in plot title denotes Wilcoxon test p<0.05. C: Multivariable Cox regression for OS, using CD68+ cellular density and Tumor/CD68+ nK(25) as separate covariates, with associated KM estimates for high versus low Tumor/CD68+ nK(25). D: Multivariable Cox regression for OS, stratifying patients who had high CD68+ cellular density and high Tumor/CD68+ nK(25) versus other patients, with associated KM estimates.