A PRRX1 Signature Identifies TIM-3 and VISTA as Potential Immune Checkpoint Targets in a Subgroup of Microsatellite Stable Colorectal Cancer Liver Metastases

Abstract

Disease recurrence and drug resistance are major challenges in the clinical management of patients with colorectal cancer liver metastases (CLM), and because tumors are generally microsatellite stable (MSS), responses to immune therapies are poor. The mesenchymal phenotype is overrepresented in treatment-resistant cancers and is associated with an immunosuppressed microenvironment. The aim of this work was to molecularly identify and characterize a mesenchymal subgroup of MSS CLM to identify novel therapeutic approaches. We here generated a mesenchymal gene expression signature by analysis of resection specimens from 38 patients with CLM using ranked expression level of the epithelial-to-mesenchymal transition-related transcription factor PRRX1. Downstream pathway analysis based on the resulting gene signature was performed and independent, publicly available datasets were used to validate the findings. A subgroup comprising 16% of the analyzed CLM samples were classified as mesenchymal, or belonging to the PRRX1 ^high group. Analysis of the PRRX1 signature genes revealed a distinct immunosuppressive phenotype with high expression of immune checkpoints HAVCR2/TIM-3 and VISTA, in addition to the M2 macrophage marker CD163. The findings were convincingly validated in datasets from three external CLM cohorts. Upregulation of immune checkpoints HAVCR2/TIM-3 and VISTA in the PRRX1 ^high subgroup is a novel finding, and suggests immune evasion beyond the PD-1/PD-L1 axis, which may contribute to poor response to PD-1/PD-L1-directed immune therapy in MSS colorectal cancer. Importantly, these checkpoints represent potential novel opportunities for immune-based therapy approaches in a subset of MSS CLM.

Significance: CLM is an important cause of colorectal cancer mortality where the majority of patients have yet to benefit from immunotherapies. In this study of gene expression profiling analyses, we uncovered novel immune checkpoint targets in a subgroup of patients with MSS CLMs harboring a mesenchymal phenotype.

FIGURE 1

PRRX1 gene and signature expression. A, Analysis of EMT-related transcription factor gene expression range in the COMET CLM cohort. Underlying bar displays variance per gene. B,PRRX1 gene expression in quartile-based selected groups. The samples were categorized into three groups according to levels of PRRX1 expression: Q1 (low): 0%–25% quartile, Q2+Q3 (inter-medium): 25%–75% quartile, and Q4 (high): 75%–100% quartile. Significant differences in pairwise comparisons where Q4 was set as reference (t test). C, Heatmap and hierarchical clustering of the CLM samples based on the 405 PRRX1 DEGs. Red-blue scale reflects log₂ expression (range, 4.3–16.3). The rows above the heatmap depict main clusters (blue and red as shown on the top “Cluster” row) and subgroups (blue, light-blue, orange-red, and red as shown on the second “PRRX1_subgroups” row) defined by dendrogram subbranches. Seven samples fall into a distinct PRRX1^high cluster (red box).

FIGURE 2

Immunologic features associated with the PRRX1 signature. A, “Cell proliferation of T lymphocytes” was a significant biological function to be associated with the PRRX1 signature genes. The function was predicted to be inhibited in the PRRX1^high subgroup. The blue lines indicate candidate upregulated genes associated with inhibition of proliferation of T cells. B, Gene expression of HAVCR2/TIM-3 across PRRX1 subgroups. Statistical significance was assessed by ANOVA in a multiple group comparison. C, Correlation matrix between the PRRX1 signature score and expression of key immune checkpoint genes and M2 (CD163) and M1 (NOS2) macrophage markers. D, Boxplots of immune cell abundances. Estimated abundance of immune cells based on gene expression data using the TIMER application. Abundance of B cells, DCs, macrophages, neutrophils, CD4⁺ T cells, and CD8⁺ T cells in CLM samples estimated according to PRRX1 subgroup. Statistical significance was assessed by ANOVA in a multi-group comparison.

FIGURE 3

Venn diagram of signature genes. Overlapping sets of genes between the PRRX1 signature and three public EMT core signatures based on meta-analyses [References: Broad Institute (9, 10); see Supplementary Data S5]. The diagram was made using the online tool Venny (http://bioinfogp.cnb.csic.es/tools/venny/index.html).

FIGURE 4

Boxplots of protein quantification. A, Protein expression of PRRX1 signature genes. Of the eight genes with protein data, PRRX1 subgroup specific expression is significant for five proteins, all showing a higher expression in the PRRX1^high subgroup. B, Protein quantification by RPPA of EMT- (top) and immune-related proteins (bottom). Statistical significance was assessed by ANOVA in a multi-group comparison.