Abstract
Introduction: Individuals diagnosed with colorectal cancer (CRC) frequently confront a grave prognosis and exhibit poor responses to conventional treatment regimens. Immunotherapy, notably modalities centered on natural killer (NK) cells, represents a burgeoning frontier in the management of CRC. This study developed a validated prognostic model using NK-associated long non-coding RNAs (lncRNAs) to predict CRC outcomes. Methods: Integrating single-cell RNA-seq (GSE146771_Smartseq2) and TCGA-COAD/READ bulk transcriptomic data, we identified NK-specific genes and correlated lncRNAs. A multi-step analytical approach-including univariate Cox regression for preliminary screening, LASSO regression to minimize overfitting, and multivariate Cox regression for final model optimization-yielded a robust 16-lncRNA prognostic signature with high predictive accuracy. Results: This model demonstrated robust predictive performance across the training set, validation set, and 76 independent clinical samples. Mechanistic investigations revealed that AC010319.3 is highly expressed in NK cells, where it attenuates NK cell cytotoxicity by suppressing the expression of IFN-γ and granzyme B, thereby promoting the proliferation and invasion of CRC cells. Discussion: This study systematically delineates the regulatory role of NK-associated lncRNAs within the CRC immune microenvironment, offering novel molecular targets and stratification strategies for CRC immunotherapy.