570P - Subclasses of microsatellite-instability colorectal cancer with unique molecular features and immune cell infiltration patterns

Background

Mismatch repair deficient or microsatellite instable (MSI) tumors are often characterized by high mutational burdens and strong immune infiltration and have shown higher sensitivity to immune checkpoint inhibitors (ICIs) compared to microsatellite stable (MSS) tumors. However, only a portion of the patients with MSI colorectal cancer (CRC) gain survival benefits from ICIs treatment, suggesting unclear heterogeneity in MSI tumors.

Methods

We analyzed the whole-genomes and transcriptomes of 223 ICIs treatment-naïve MSI CRC tumors from U-CAN project. Immune cell infiltration and cancer-relevant pathways were evaluated by CIBERSORT and PROGENy, respectively. An MSI subtyping model was developed based on ResNet50.

Results

Unsupervised classification based on gene expression data divided MSI tumors into two distinct classes (class1 and 2), with MSI class1 characterized by a lower level of chromosomal instability, a higher level of hypoxia, and more activated EGFR, JAK-STAT, TNFa, and NFkB pathways than class2. However, two MSI classes showed no differences in 5-year survival. Further investigation of the immune cell population showed that class1 had considerably higher immune cell infiltration compared to class2, specifically in M0, M1 and M2 macrophages, CD4 and CD8 T-cells, naïve B-cells, resting NK-cells, and neutrophiles. Class2 exhibited significantly higher loss in 10q23.31 (containing PTEN) and reduced expression of PTEN, which may lead to an immunosuppressive tumor microenvironment and partially explain the higher activation in PI3K and VEGF pathways in class2 compared to class1. Importantly, ICIs targets, such as CTLA4, PD-1, and PD-L1 all had higher expressions in calss1. We further developed an MSI subtyping model based on refined pathway features of normal tissue, tumors with MSS, class1 or 2. Ongoing analyses include validation of MSI classification in the TCGA dataset.

Conclusions

We illustrated the heterogeneity in MSI CRCs and revealed two subclasses of MSI with distinct molecular features, which may be informative for immune-based treatments. The MSI subtyping model may further facilitate precision immunotherapy and allow patients to derive clinical benefits.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

Uppsala University Umeå University BGI-Shenzhen.

Funding

The Swedish Cancer Society the Uppsala Cancer Foundation the Guangdong Provincial Key Laboratory of Human Disease Genomics the Swedish Government (CancerUU) the Erling-Persson Foundation.

Disclosure

All authors have declared no conflicts of interest.