Type 2 diabetes (T2D) and heart failure (HF) are both growing in prevalence and concern with T2D patients at 2-fold increased risk of developing HF, regardless of other cardiovascular diseases. Protein biomarkers are routinely used as prognostic and diagnostic tools for disease however, in HF they lack specificity and sensitivity. Furthermore, they often vary in concentration within different ethnicities, ages, genders, or diseases. A main limitation of plasma for biomarker discovery is the high dynamic range limiting the detection of low abundant proteins. Therefore, emerging methods have been utilised to enrich low abundant proteins from plasma. The aim of this study is to identify protein biomarkers for the detection of HF in a T2D cohort based on subclinical myocardial changes.
Extracellular vesicles (EVs) were isolated through precipitation with acidic ammonium acetate and lysed. The unbound fraction was added to lipid removal agent (LRA) to bind for an hour. The following steps were done to both EV and LRA methods separately. Proteins were reduced and alkylated with DTT and IAA, and digested overnight with trypsin. Salts and impurities are removed through solid phase extraction and 250 ng was loaded onto the mass spectrometer for analysis. The sample was desalted and chromatographically focused by a Symmetry C18 Trap column before separation by an Acquity UPLC M-Class HSS T3 column. Separation occurred over 125 minutes using aqueous mobile phase A (0.1 % FA) and organic mobile phase B (80 % ACN, 0.1 % FA). Initial conditions were 97 % A this was held for 4 minutes followed by a linear gradient to 60 % A at 97 minutes. Residual proteins were washed off with 10 % A for 5 minutes then an isocratic hold at 97 % A for 17 minutes. The column was directly coupled to a Waters Synapt G2-S mass spectrometer (MS) by an electrospray ionisation source. The MS was operated in HDMSe positive ion mode with an ion mobility velocity of 326 m/s and wave height of 40 V. Low CID energy of 15 V was applied across the transfer ion guide. High CID energy used a ramped look up table. Argon was used as the CID gas. Data was acquired using MassLynx4.1 and data processed in Progenesis QI. Statistical analysis was performed using RStudio, where identified proteins were filtered for redundancy and low counts. Batch effects were removed using surrogate variables before the proteins were investigated for significant correlations to multiple models indicative of HF. These models were comprised of clinical imaging and exercise variables including, global longitudinal strain (GLS), average EE (EE), extracellular volume (ECV), and peak VO2 max (VO2). Further statistical analysis was undertaken using weighted correlation network analysis to identify modules of proteins associated with disease.
A multivariate model was created to interrogate the relationship between expression levels of proteins and HF variables. Initial processing in Progenesis QI resulted in 3086 and 2247 proteins being identified in the EV and LRA datasets, respectively. After filtering, 2685 and 1874 proteins remained in the EV and LRA datasets, respectively. Each dataset was separately correlated against each variable and then underwent weighted correlation network analysis to identify significant proteins associated with each variable. The two datasets were then combined after statistical analysis. There were 238, 432, 530, and 360 proteins that were significantly associated with variable GLS, EE, ECV, and VO2, respectively. Comparing the lists identified 93 proteins that were common between all four variables of interest and these underwent protein-protein interaction (PPI) analysis to further reduce the number of significant proteins. Based on the PPI analysis proteins were filtered based on the number of interactions they shared. Any protein with less than 3 interactions was removed from the candidate list. Finally, the remaining proteins were filtered based on their original confidence score and the number of unique peptides identified, given by Progenesis QI. This resulted in a final list of 27 proteins to be verified and validated.
The identification of 27 candidate protein biomarkers is the first step in developing a fully validated biomarker panel for use in a clinical setting. These proteins represent potential changes that occur within an individual’s protein expression levels reflecting their state of subclinical HF. To be confident in the identification of these proteins they first need to be verified using targeted MS. This would be done by identifying the most prototypic peptides for each protein and developing a single reaction monitoring assay. Any protein that was a false positive would be identified at this stage and removed before commencing validation. The assay developed for the identification of the peptides will undergo a bioanalytical method validation and then the peptides will undergo a clinical validation to assay their specificity and sensitivity as biomarkers of HF in T2D. Not only do these protein identify novel markers for HF in T2D but they provide insights into the disease pathophysiology. Further work could be undertaken to elucidate the specific mechanisms that are altered in T2D patients that increases their risk and ultimately causes them to develop HF.