NEW YORK – A pair of studies published Wednesday in Cell Reports Medicine have identified proteins linked to type 2 diabetes and metabolic syndromes.
The studies, one led by researchers at Sweden's Karolinska Institute and another by researchers at China's Westlake University, used large-scale proteomic datasets to identify potential protein biomarkers as well as proteins causally linked to these conditions.
In the Karolinska-led study, the scientists analyzed genetic and proteomic data collected as part of the UK Biobank Pharma Proteomics Project and by Decode Genetics along with data from the DIAGRAM (Diabetes Genetics Replication and Meta-analysis) consortium and the FinnGen human genetics study to identify 47 plasma proteins with a possible causal link to type 2 diabetes as well as the role of 17 proteins in diabetic complications.
In the UK Biobank dataset, 1,463 plasma proteins were measured in 54,306 subjects using Olink's Explore platform, while in the Decode dataset, 4,907 proteins were measured in 35,559 individuals using SomaLogic's SomaScan platform. Proteogenomic analyses of these two datasets identified 1,161 protein quantitative trait loci (pQTLs) — genomic loci linked to variations in protein expression — in the UK Biobank set, 1,423 in the Decode set, and 509 in common across the two sets.
They then used this pQTL data with the DIAGRAM consortium and FinnGen datasets to conduct Mendelian randomization and colocalization analyses, through which they identified proteins linked to diabetes as well as complications including ketoacidosis, retinopathy, stroke, and coronary artery disease. The researchers noted that four of the diabetes-linked proteins — HLA-DRA, AGER, HSPA1A, and HSPA1B — appeared associated with most of the complications studied and that this "may suggest the roles of inflammation and oxidative stress in diabetic progression."
They added that their findings indicate "that AGER inhibitors or antagonists may be a promising therapeutic target for diabetic complications."
In the Westlake-led study, researchers used data-independent acquisition mass spec to identify blood-based protein markers than could help identify individuals at risk of developing metabolic syndrome.
To collect the data for their model, the scientists measured roughly 400 proteins in 7,890 serum samples collected longitudinally from 3,840 participants over the course of 10 years as part of the Guangzhou Nutrition and Health Study (GNHS).
They first split the 7,890 samples into a 4,794-sample discovery cohort and a 3,094-sample validation cohort. They then ran them on a Sciex TripleTOF 5600 mass spectrometer using the SWATH DIA workflow with 20-minute LC gradients and the DIA-NN software package, quantifying the levels of 438 proteins.
They then took a subset of this discovery cohort consisting of 267 individuals without metabolic syndrome at baseline but who developed it within the 10-year follow-up period and 588 individuals who did not develop metabolic syndrome during the follow-up period and used machine learning to build an 11-protein model for predicting the likelihood of developing metabolic syndrome over 10 years.
Testing the model in the discovery cohort, they found it was able to predict development of the condition with an area under the curve (AUC) of 0.78. Testing the model in a 242-cohort drawn from the validation samples, they found it performed with an AUC of 0.77.
The authors noted that their findings also provided insight into proteins and mechanisms underlying metabolic syndrome, demonstrating, for instance, the significant role of apolipoproteins in both the development and onset of the condition. They also identified several proteins, including the immune-related proteins A1BG and ATRN, that had not previously been reported as linked to metabolic syndrome.
They also suggested that, given the breadth of clinical and phenotypic data collected on the GNHS cohort, the proteomic data generated by the study could be useful for exploring a variety of other conditions.