Targeting the Sugar Code: Serum Glycoproteome Profiling for Early Detection of Advanced Adenoma and Colorectal Cancer

Introduction: Colorectal cancer (CRC) is a leading cause of cancer-related death. Immune responses within the colonic crypt have been found to contribute to CRC development. Abnormal protein glycosylation has been linked to the malignant transformation process. This study leverages stable circulating glycoproteomic markers as a means of identifying advanced adenomas (AAs) and CRC at an early stage. Methods: Prospective (NCT05445570) and biorepository samples (to supplement AA/CRC cases) from multiple sites were included. AA was defined as polyps ≥1 cm, any size polyp with >25% villous features or high-grade dysplasia. Utilizing a glycoproteomic profiling platform that combines liquid-chromatography/mass-spectrometry and artificial-intelligence-powered data processing, we assessed glycopeptide (GP) and non-glycosylated peptide quantification transitions in peripheral blood serum. The samples were split into training, validation and hold-out testing sets. Statistical analyses were performed on normalized data from the optimized assay to develop and validate a classifier to predict probability of AA/CRCs against controls. Primary outcomes for the validation test-set were sensitivity for AA and CRC, as well as AUC for AA/CRC combined. Results: We analyzed 1,356 prospectively collected samples and 681 biorepository samples: Group 1: 545 CRC (27%); Group 2: 383 AAs (19%); Group 3: 154 non-AAs (8%); Group 4: 955 colonoscopy negative controls (47%). We identified 84, 89, and 16 GPs/peptides with statistically significant abundance differences (FDR < 0.001), when comparing CRCs with controls, high-grade dysplastic (HGD) AAs with controls, and AAs to non-AAs respectively. A subset of 24 of these biomarkers were used to generate a Multivariable Classifier Model using the training and validation data sets. When the Classifier was applied to the test set it yielded an area under the receiver-operating characteristic of 0.83 for the detection of AA and CRC. Using a defined cutoff, the Classifier sensitivity for all CRC stages was 80.9% (85.0% stage 1&2); for HGDs 89.6% and for all AAs 43.8% with specificity of 90.4% for controls and 89.6% for non-AAs in the test set (Figure 1). Conclusion: Using glycoproteomic profiling, we detected glycosylation changes associated with the host immune response to AAs with HGD and CRC. The high sensitivity and specificity of this strategy supports the development of a non-invasive blood-based screening test that complements current screening strategies.Figure 1.: Distributions of predicted probabilities in each phenotype, stratified by dataset.