Cigarette smoking and long-term incidence of colorectal cancer subclassified by tumor-infiltrating myeloid cells: A prospective cohort study.

Evidence indicates that cigarette smoking affects anti-tumor immunity. We tested a hypothesis that the association of smoking with long-term colorectal cancer (CRC) incidence might differ by myeloid cell infiltrates in tumors, which may play roles in cancer immune evasion. We utilized the prospective cohort incident-tumor biobank method on two longitudinal prospective cohort studies, namely the Nurses' Health Study and the Health Professionals Follow-up Study. To identify myeloid cells in the tumor microenvironment, we applied multiplexed immunofluorescence (for ARG1, CD14, CD15, CD33, HLA-DR, and KRT) combined with digital image analysis and machine learning. We examined the association of smoking with long-term incidence of CRC subclassified by myeloid cell densities using the multivariable inverse-probability-weighted duplication-method Cox regression model. During the follow-up of 131,143 cohort participants (3,647,434 person-years), we documented 3092 incident CRC cases, including 854 cases with available data on tissue myeloid cells. The association of pack-years smoked with long-term CRC incidence differed by intraepithelial CD14+HLA-DR- cell densities (Pheterogeneity<0.001) but not by CD15+ cell densities (Pheterogeneity = 0.32). Smoking pack-years were strongly associated with the incidence of tumors with high CD14+HLA-DR- cell densities (multivariable HR for ≥40 vs. 0 pack-years, 1.86; 95% CI, 1.35-2.57; Ptrend<0.001) but not with that of tumors with intermediate or low CD14+HLA-DR- cell densities (Ptrend≥0.1). These differential associations generally persisted in non-microsatellite instability-high tumors. In conclusion, the association of smoking with long-term CRC incidence was stronger for tumors with more abundant CD14+HLA-DR- immature monocytes, highlighting the interplay of smoking and immature monocytic myeloid cells in colorectal tumorigenesis.