IDDF2023-ABS-0129 Disentangling the complex interplay of pathological images, tumor microbes and tumor microenvironment and elucidation of the intertumor heterogeneity

Background

Paramount of studies on tumors has revealed that cancer development is associated with many aspects, including pathological images, intratumoral microbes, and tumor microenvironment (TME), underscoring the influence of intratumoral microbes on TME, which could be reflected in pathological images. However, there is a lack of investigation into the underlying biological links among these three factors from a pan-cancer perspective. In this study, we attempted to disentangle the association among these three factors for 13 cancer types, including digestive tract tumors.

Methods

We collected a cohort of 6,095 samples from The Cancer Genome Atlas (TCGA) covering 13 cancer types, each with matching histology slides, tissue microbiome samples, and host transcriptome samples (IDDF2023-ABS-0129 Figure 1. Study cohorts and tumor microbiome-host associations for 13 cancer types). Association of pathological images, tumor microbes, and TME was conducted, followed by multi-feature clustering to fill the knowledge gaps of intra- and intertumor heterogeneity.

Results

We have categorized the 13 types of cancers into three intertumor types: T1, T2 and T3 (IDDF2023-ABS-0129 Figure 2. Three types of cancer and their type-specific microbial signatures). T1 included low-grade glioma (LGG), esophageal cancer (ESCA) and pancreatic adenocarcinoma (PAAD). T1 tumor tissues had the lowest abundance of Actinobacteria, and there was no significant interaction between tumor microbes and the host. T2 included seven types of cancer, including lung adenocarcinoma (LUAD) and liver hepatocellular carcinoma (LIHC). Proteobacteria and Bacteroidetes were overabundant in T2 tumor tissues, accompanied by significant interactions between microbes and host genes, but no histological image differences were observed. Colorectal cancer (CRC), stomach adenocarcinoma (STAD), and lung squamous cell carcinoma (LUSC) belonged to T3. Firmicutes and Actinobacteria were enriched in T3 tumors. In T3, microbial and host interactions led to profound variation in the TME in tissues, which were reflected in histological images. Specifically, in CRC, Acinetobacter and Campylobacter were significantly positively associated with stromal cell components, while increased stromal scores were associated with poor prognosis.

Conclusions

For the first time, we proposed a combination of pathological images, tumor microbes, and TME for tumor typing, providing new insights for intra- and intertumor heterogeneity.