Diagnosis and Prognosis of aGVHD By Metabolic Biomarkers after Allogeneic Hematopoietic Stem Cell Transplantation

Introduction

Acute graft-versus-host disease (aGVHD) significantly impacts the survival and prognosis of AML patients undergoing allogeneic hematopoietic stem cell transplantation (allo-HSCT), making accurate diagnosis crucial. Currently, aGVHD diagnosis relies mainly on clinical assessment and invasive histopathological examinations, with protein-related markers such as Elafin, TIM3, ST2, and REG3α aiding in diagnosis. Compared to proteomics, metabolomics can more directly reflect the immediate physiological and pathological states, providing clearer insights into biological phenomena. Although metabolic biomarkers like fatty acids, lipids, and carnitines have been used for aGVHD diagnosis and prognosis, the metabolic changes during aGVHD development remain unexplored. This study utilizes metabolomics technology to explore the dynamic metabolic changes during aGVHD progression and assess their correlation with clinical outcomes, aiming to provide a scientific basis for non-invasive diagnosis and prognosis of aGVHD in allo-HSCT patients.

Methods

This study included 70 AML patients who underwent allo-HSCT at Hebei Yanda Lu Dao Pei Hospital, with a total of 110 plasma samples, including 33 aGVHD samples, 22 non-aGVHD (control) samples, 12 pre-aGVHD and 43 post-aGVHD samples. Patients who received secondary or tertiary transplantations were excluded. The diagnosis and grading of aGVHD were based on the clinical characteristics and the modified Glucksberg method. The last follow-up date was December 27th, 2023. Non-targeted metabolomics detection of samples was conducted using a liquid chromatography-tandem mass spectrometry system. Data processing and statistical analysis were performed using Progenesis QI, Simca 14.1, SPSS 27.0, R 4.1.2, and the fuzzy C-means clustering algorithm.

Results

Compared with non-aGVHD patients, there were significant changes in 36 metabolites in plasma of aGVHD patients, including 17 metabolites with a significant increase and 19 metabolites with a significant decrease (VIP>1.0 and P < 0.05). KEGG pathway analysis revealed three significantly affected pathways: arginine biosynthesis, glycine, serine, and threonine metabolism, and glycerophospholipid metabolism. Cluster analysis showed that the 36 metabolites were clustered into 6 types, presenting regular changes. Pathway enrichment, cluster analysis and ROC analysis showed that PC(P-16:0/18:1), PC(o-16:1/18:0), PC(16:0/16:0), PE(P-18:0/20:5), PC(20:4/20:4), and PC(18:0/20:3) were candidate metabolites, with AUC values of 0.831 (95%CI, 0.724-0.940), 0.789 (95%CI, 0.670-0.909), 0.764 (95%CI, 0.640-0.889), 0.764 (95%CI, 0.628-0.900), 0.782 (95%CI, 0.658-0.909), and 0.758 (95%CI, 0.627-0.891), respectively. The first three metabolites were upregulated from control to pre-aGVHD and aGVHD stages and dropped in the post-aGVHD period, whereas the latter three were downregulated from control to pre-aGVHD and aGVHD stages but significantly rebounded in the post-aGVHD period. Aadditionally, the AUC value of the 6-marker panel was 0.960 (95%CI, 0.914-1.000), indicating that the panel of 6 metabolites has higher diagnostic accuracy in distinguishing between aGVHD and non-aGVHD patients.

Survival curves and multivariate COX regression show that the aGVHD group had a worse survival rate (60.61%, P=0.0097) compared to the non-aGVHD, and PC(20:4/20:4), PC(P-16:0/18:1), PC(16:0/16:0), and PC(o-16:1/18:0) were independent risk factors for survival (P=0.0029, 0.0290, 0.0061, and 0.0294). Additionally, the nomogram model based on the levels of these six metabolites can predict the survival outcome of allo-HSCT patients within four years.

Conclusion

The six glycerophospholipid metabolites show significant metabolic differences at various stages of aGVHD and have the potential to serve as biomarkers for the aGVHD diagnosis in AML patients. The COX regression model based on these metabolites can predict the survival of allo-HSCT patients within four years. Future validation with a more extensive and multicentric clinical sample set is needed.