Cytotoxic T-cell exhaustion analyses by in situ single-cell immunofluorescence in relation to colorectal cancer patient age, tissue bacteria and mortality

Objectives We hypothesised that cytotoxic T-cell exhaustion status in colorectal cancer (CRC) microenvironment relates to age of diagnosis (eg, early-onset CRC), pathogenic bacteria and clinical outcomes. Design Population-based prospective cohort study using the prospective cohort incident-tumour biobank method (PCIBM). Setting Two US nationwide prospective cohort studies, the Nurses’ Health Study (NHS) and the Health Professionals Follow-Up Study (HPFS). Participants Among 4476 incident CRC cases identified during follow-up, 857 patients with available tumour tissue were included in the present tissue-based analysis. Patients with colon and rectal adenocarcinomas were included. Interventions Not applicable. Main outcome measures In situ single-cell multispectral immunofluorescence combined with computational machine learning for CD3, CD8, CD274 (PD-L1), HAVCR2 (TIM-3), PDCD1 (PD-1) and KRT (keratin) was used to characterise non-exhausted, progenitor-exhausted and exhausted CD3 + CD8 + cells according to PDCD1 (PD-1) and HAVCR2 (TIM-3) expression. Primary measures included densities of CD3 + CD8 + cell exhaustion subsets and their associations with clinicopathological characteristics, age at CRC diagnosis, selected tumour tissue bacteria, CRC-specific mortality and spatial proximity to tumour cells. Results Younger age of CRC diagnosis generally correlated with lower densities of CD3 + CD8 + cells. Multivariable-adjusted ORs (with 95% CI) in age <55 (vs age ≥70) for overall, progenitor-exhausted and exhausted CD3 + CD8 + cell stromal densities (all quartiles) were 0.39 (0.22 to 0.69; P trend =0.0033), 0.40 (0.23 to 0.68; P trend =0.0048) and 0.44 (0.25 to 0.78; P trend =0.0082), respectively. Multivariable-adjusted ORs (with 95% CI) for stromal non-exhausted CD3 + CD8 + cells (quartiles) were 0.81 (0.56 to 1.17) and 0.52 (0.37 to 0.74) (P trend =0.0004) for low-level and high-level Bacteroides fragilis (vs negativity), respectively. Multivariable-adjusted ORs for stromal exhausted CD3 + CD8 + cells (quartiles) were 0.37 (0.19 to 0.72) and 0.60 (0.30 to 1.17) (P trend =0.0081) for low-level and high-level enterotoxigenic Bacteroides fragilis (vs negativity), respectively. Multivariable-adjusted CRC-specific mortality HRs (with 95% CI) in quartile 4 (Q4; vs Q1) of progenitor-exhausted CD3 + CD8 + cells in the intraepithelial region and their proximity to tumour cells were 0.42 (0.26 to 0.67; P trend <0.0001) and 0.50 (0.32 to 0.78; P trend =0.0004), respectively. Conclusion This PCIBM-based study provides evidence that progenitor-exhausted CD3 + CD8 + cell density and their proximity to tumour cells are favourable prognostic biomarkers in CRC. Moreover, lower stromal densities of overall, progenitor-exhausted and exhausted CD3 + CD8 + cells are associated with younger-onset CRC.