Background Research on cancer-associated fibroblasts (CAFs) in colorectal cancer (CRC) has advanced, although its clinical effectiveness is hindered by the absence of viable targets for therapy and drug delivery methods.

Methods A realistic CAF risk score (CAFscore) was constructed using three prognostic-related CAF hub genes (CRIP2, FSTL3, and SLC2A3) by univariate Cox and LASSO regression analysis. Redox-responsive micelles were loaded with anti-programmed death-ligand 1 antibodies (aPD-L1). The effectiveness of inhibiting tumor growth was increased by anti-follistatin-like 3 antibodies (aFSTL3) by a method of co-encapsulation to stimulate immunogenic cell death (ICD).

Results The elevated expression of FSTL3 in CAFs was strongly associated with malignant biological function in CRC. FSTL3 expression showed excellent performance in both cohorts utilized for external validation and training. FSTL3 exhibited strong expression in stromal cells of datasets GSE146771 and GSE110009, particularly in CAFs. The micelles successfully targeted CRC and remained in the tumor microenvironment without affecting the bioactivity of aFSTL3. Combining aPD-L1 and aFSTL3 significantly improved the effectiveness against tumors by reducing primary and recurrent CC, increasing cytotoxic T cells, and promoting long-lasting immunological memory in the local immune system.

Conclusions FSTL3 overexpression was linked to poor prognosis, somatic hypermutation, and resistance to drugs. The co-encapsulation method enhances antibody distribution and, when combined with aFSTL3, suppresses vascular mimicry, showing that anti-CAF treatment altered local immunity in CRC.