High dose-rate brachytherapy of localized prostate cancer converts tumors from cold to hot

Abstract

Background: Prostate cancer (PCa) has a profoundly immunosuppressive microenvironment and is commonly immune excluded with few infiltrative lymphocytes and low levels of immune activation. High-dose radiation has been demonstrated to stimulate the immune system in various human solid tumors. We hypothesized that localized radiation therapy, in the form of high dose-rate brachytherapy (HDRBT), would overcome immune suppression in PCa.

Methods: To investigate whether HDRBT altered prostate immune context, we analyzed preradiation versus postradiation human tissue from a cohort of 24 patients with localized PCa that received HDRBT as primary treatment (RadBank cohort). We performed Nanostring immune gene expression profiling, digital spatial profiling, and high-throughput immune cell multiplex immunohistochemistry analysis. We also resolved tumor and nontumor zones in spatial and bioinformatic analyses to explore the immunological response.

Results: Nanostring immune profiling revealed numerous immune checkpoint molecules (eg, B7-H3, CTLA4, PDL1, and PDL2) and TGFβ levels were increased in response to HDRBT. We used a published 16-gene tumor inflammation signature (TIS) to divide tumors into distinct immune activation states (high:hot, intermediate and low:cold) and showed that most localized PCa are cold tumors pre-HDRBT. Crucially, HDRBT converted 80% of these 'cold'-phenotype tumors into an 'intermediate' or 'hot' class. We used digital spatial profiling to show these HDRBT-induced changes in prostate TIS scores were derived from the nontumor regions. Furthermore, these changes in TIS were also associated with pervasive changes in immune cell density and spatial relationships-in particular, between T cell subsets and antigen presenting cells. We identified an increased density of CD4⁺ FOXP3⁺ T cells, CD68⁺ macrophages and CD68⁺ CD11c⁺ dendritic cells in response to HDRBT. The only subset change specific to tumor zones was PDL1^- macrophages. While these immune responses were heterogeneous, HDRBT induced significant changes in immune cell associations, including a gained T cell and HMWCK⁺ PDL1⁺ interaction in tumor zones.

Conclusion: In conclusion, we showed HDRBT converted "cold" prostate tumors into more immunologically activated "hot" tissues, with accompanying spatially organized immune infiltrates and signaling changes. Understanding and potentially harnessing these changes will have widespread implications for the future treatment of localized PCa, including rational use of combination radio-immunotherapy.

Keywords: computational biology; gene expression profiling; prostatic neoplasms; radiotherapy.

Figure 1

Overview of experimental plan. (A) Twenty-four patients with localized prostate adenocarcinoma were treated with a single 10 Gy dose of HDRBT with image-guided biopsies taken immediately prior to and 2 weeks post-HDRBT. Following formalin fixation and sectioning, samples were processed for (i) histopathological examination and tumor zone categorization, (ii) Nanostring nCounter Pan-cancer immune gene panel RNA profiling, (iii) digital spatial profiling, (iv) multiplex IHC, and (v) computation spatial analysis. (B) Schematic showing tissue zones categorized by histopathology and incorporated into subsequent immune density analyses. DC, dendritic cells; HMWCK, high molecular weight cytokeratin; HDRBT, high dose-rate brachytherapy; mIHC, multiplex immunohistochemistry; Mφ, macrophage; NT, nontumor; T, tumor; Treg, regulatory T cells.

Figure 2

HDRBT induced a 16-gene TIS in localized PCa. (A) Heatmap of normalized expression levels of 16 genes in TIS and categorization by k-means clustering into three groups: (i) Cluster 1, high TIS, (ii) Cluster 2, intermediate TIS, and (iii) Cluster 3, low TIS. White and black boxes indicate either pre-HDRBT or post-HDRBT tissue, respectively. Colored circles indicate pre-HDRBT samples and their TIS category change post-HDRBT. (B) Averaged and ranked z-scores for the 16 genes in TIS indicated three categories. (C) Proportion of pre-HDRBT or post-HDRBT tissues in each of the three TIS categories identified in (A), the p value was calculated from χ² test. (D, E) Dot plots of (D) mean TIS expression and (E) TGFB1 mRNA levels in patient-matched pre-HDRBT or post-HDRBT-treated PCa tissue. Wilcoxon matched pair test. *P<0.05, **p<0.01, ***, p<0.001, ****p<0.0001. (F) Box-and-whisker plots of expression levels of immune checkpoint molecules in pre-HDRBT and post-HDRBT tissues from all patients in cohort. TLK2 and TRIM39 are provided as invariant controls. Significance was assessed using a Wilcoxon matched pair test. *P<0.05, **p<0.01, ***, p<0.001, ****p<0.0001. † represents RadBank-V1. HDRBT, high dose-rate brachytherapy; PCa, prostate cancer; TIS, tumor inflammatory signature.

Figure 3

Digital spatial gene and protein expression profiling reveals cellular and immune checkpoint drivers of high and intermediate TIS. Heatmaps of normalized Nanostring TIS gene z-scores in two patients using (A) bulk tissue from pre-HDRBT and post-HDRBT tissues or (B) post-HDRBT ROIs (n=24) assessed using DSP human IO RNA panel. Tissue sites were determined using tumor histopathology (tumor and nontumor) and immune cell IHC staining for T cells (CD3) and Mφ (CD68) (12 ROIs per patient). High magnification examples of highest and lowest TIS-expressing ROIs in post-HDRBT tissue from patient (C) RB019 and (D) RB023. (E) Mean Nanostring DSP-derived TIS scores by tumor category in two patient samples (tumor zones: n=7/12(RB019) and n=5/12(RB023)). Pearson correlation of DSP IO proteins in (F) cell profiling, (G) immune checkpoint, or (H) activation protein classes with DSP RNA TIS score. DSP, digital spatial profiling; HDRBT, high dose-rate brachytherapy; ROIs, regions of interest; TIS, tumor inflammatory signature.

Figure 4

HDRBT significantly increased the density of CD4⁺FOXP3⁺ T cells and antigen presenting cells. Multiplex IHC was performed on PCa biopsies pre-HDRBT and post-HDRBT (n=24 patients). (A) Cumulative barplots of mean immune cell density either pre-HDRBT or post-HDRBT. P values calculated from two-tailed student’s t-test with * indicating p<0.05. (B) Immune subset density calculated from total number of identified immune phenotypes (see table 2) per square millimeter pre-HDRBT or post-HDRBT in each of the two designated tissue zones: (i) nontumor or (ii) tumor-containing. Statistical significance was calculated using a nonparametric Wilcoxon signed-rank test. P values indicated were appropriate. *P<0.05, **p<0.01, ***p<0.001. HDRBT, high dose-rate brachytherapy; IHC, immunohistochemistry; PCa, prostate cancer.

Figure 5

Immune cell density relationships correlate with TIS signature. (A) Computed differences in Pearson correlation (Δr) for either (i) nontumor and (ii) tumor following HDRBT. Color indicates direction of change, that is, red=gained association, blue=lost association. Only significant associations are shown. (B) Bubblechart of correlation (Pearson r) between immune cell densities in different tissue zones (total, tumor, or nontumor) and TIS signature. Significant correlations indicated. (C) Scatterplots of highly significant immune cell subset correlations between tumor and nontumor tissue zones, and (D) highly significant cross-subset correlations in total tissue, with relative TIS level and HDRBT radiation-status indicated. (T: tumor zone, NT: nontumor zone, Tot: all tissue zones). Pearson correlation r values and corresponding p values indicated. HDRBT, high dose-rate brachytherapy; TIS, tumor inflammatory signature.

Figure 6

Immune cell spatial interactions change following HDRBT and uniquely associate with TIS signature. (A) Box-and-whisker plots of median cell-cell distance in seven significantly altered immune cell subset pairs in four major tissue zones: (i) pre-HDRBT: nontumor zone, (ii) post-HDRBT: nontumor zone, (iii) pre-HDRBT: tumor zone, and (iv) post-HDRBT: tumor zone. Significance between group is calculated using uncorrected Wilcoxon test. n=17 patients. (B) Bubblechart plot of significant target-cell and neighbor-cell median distance correlations with TIS signature in four tissue groups. Positive correlations indicate a higher median cell-cell distance association with higher TIS. Only significant (p<0.05) correlations are shown. (C) Examples of highly significant median distance to TIS correlations with Pearson r calculations for each of four categories of tissue. *P<0.05, *p<0.01, ***p<0.001. HDRBT, high dose-rate brachytherapy; TIS, tumor inflammatory signature.