Serum glycoproteomic signatures and association with survival in patients with bone and soft tissue sarcoma treated with immune-checkpoint inhibitor therapy.

Abstract

11546 Background: Glycosylation is one of the most ubiquitous and functionally important forms of post-translational modification. The role of differential glycosylation in serum proteins has so far been limited by the technical complexity inherent in generating and interpreting this information. InterVenn has built a novel platform that combines liquid chromatography/mass spectrometry with a proprietary artificial-intelligence-based data processing engine, allowing for highly scalable and reproducible interrogation of glycoproteins with site- and glycan-specificity. Methods: Using this platform, we interrogated 519 glycopeptide (GP) biomarkers derived from 70 serum proteins in pre-treatment samples from a cohort of 103 individuals (56 females, 47 males, age ranging from 18 to 84 years) presenting with one of 20 solid cancer types. All patients were treated with durvalumab and tremelimumab immune checkpoint inhibitor (ICI) therapy. Median follow-up for overall survival (OS) was 11.4 months, with 70 events total observed. OS associations were assessed for individual GPs via Cox regression models and leave-one-out-cross-validation (LOOCV) was employed to generate penalized multivariable prediction scores. Notably, 43 patients had a primary diagnosis of bone and soft tissue sarcoma, and stratified analyses were carried out in this population. Results: We identified 154 biomarkers significantly associated with OS in the full dataset after adjusting for multiple comparisons (FDR < 0.05). Of these, 7 were statistically significant at p < 0.01 in the sarcoma-only subset. LOOCV models built in all cancer types resulted in held-out scores that discriminated those likely to exhibit long-term survival post-ICI therapy from those unlikely to benefit (HR = 4.0, p = 4.91E-08, with 4 GPs included in the final model). Furthermore, LOOCV models including only sarcoma patients demonstrated even stronger predictive attributes (HR = 8.22, p = 2.10E-05, employing 2 glycopeptides). All 9 sarcoma patients with extreme glycosylation signatures for prediction of poor survival displayed quick clinical progression with little benefit from ICI therapy. Relative signal strength and comparative analyses demonstrated strong histotype-specificity inherent in the biomarkers employed for sarcoma vs all cancers. Conclusions: Our results indicate that glycoproteomic liquid biopsy holds potential as a predictive biomarker for identifying sarcoma patients who will derive the greatest benefit from ICI therapy.