VLDL cholesterol associates with higher plasmatic expression of inflammatory proteins and atherosclerotic pathways compared to LDL cholesterol

Abstract

Background and aim: High cholesterol in Low-Density Lipoproteins (LDL-C) is the key target of current pharmacological treatments aimed at reducing atherosclerotic cardiovascular disease (ACVD) risk. Increased cholesterol in Very low-density lipoproteins (“VLDL-C”) is an independent predictor of ACVD. VLDL-C was previously associated with markers of inflammation (for instance C-reactive protein). We now tested the relationship between either VLDL-C or LDL-C with a large spectrum of inflammatory proteins in plasma collected from subjects at different ACVD risks.
Methods: We measured 276 proteins (OlinkTM) in plasma from a primary ACVD risk prevention cohort (“PLIC” in Milan; n=656 (8.2% on statins)) and a secondary ACVD risk prevention cohort (the Second Manifestations of ARTerial disease, “SMART”, the Netherlands, n=630 (50.8% on statins)). Cohorts were divided into three groups for VLDL-C (“Normal” VLDL-C<15 mg/dL, “High” VLDL-C 15-30 mg/dL, “Very high” VLDL-C >30 mg/dL) and LDL-C (“Normal” LDL-C <115 mg/dL, “High” LDL-C 115-155 mg/dL, “Very high” LDL-C>155 mg/dL). The expression (Normalized Protein eXpression, NPX) of each protein was compared among these groups by artificial intelligence. The performance to discriminate subjects with higher VLDL-C or LDL-C was evaluated by comparing the Areas Under the Curve (AUCs) of the Receiver Operating Characteristics curve (ROC) considering proteomics on top of ACVD risk factors (“CVRFs”: age, body mass index, systolic blood pressure, glycemia, therapies), versus the AUC of the ROCs with CVRFs alone.
Results: The number of plasma proteins differentially expressed increased, as a function of higher VLDL-C in PLIC, as the NPXs of 84 were higher in “High” and the NPXs of 136 were higher in “Very high” vs “Normal” VLDL-C respectively. A similar trend was found in SMART, where the NPXs of 30 proteins were higher in “High” and the NPXs of 64 were higher in “Very high” vs “Normal” VLDL-C respectively. 26 proteins were shared between the two populations and recapitulated key atherosclerotic pathways (including chemotaxis of immune cells).
The relationship between LDL-C was less marked; in PLIC, 14 proteins were more expressed in “High” and 33 in “Very high” vs “Normal” LDL-C respectively, while in SMART, the NPXs of 11 proteins were higher in “High” and the NPXs of 36 were higher in “Very high” vs “Normal” LDL-C respectively. Only 4 proteins were shared between high and very high LDL-C in the two populations. Finally, none of the proteins were shared between the groups of “High”/“Very high” VLDL-C and “High”/“Very high” LDL-C in the two cohorts.
Canonical CVRFs alone slightly improved the ability to identify subjects with increased VLDL-C both in PLIC and SMART (AUCs between 0.6 on average), but adding plasma proteomics markedly improved the performance to identify subjects with “High” VLDL-C, in PLIC (AUC=0.767 (0.709-0.837)) and in SMART (AUC=0.781 (0.681-0.873)), and with “Very high” VLDL-C (AUC=0.950 (0.899-0.976) in PLIC, and AUC=0.938 (0.894-0.971) in SMART).
The ROC of plasma proteomics with CVRFs was also superior to the ROC of the CVRFs alone to identify subjects with “High” and “Very high” LDL-C, but, as compared to the ROCs that discriminated subjects with “High” and “Very-high” VLDL-C, the AUCs were attenuated in both cohort (for “High” LDL-C: AUC=0.665 (0.558-0.774) in PLIC and AUC=0.775 (0.704-0.842) in SMART; for “Very high” LDL-C: AUC =0.776 (0.694-0.854) in PLIC and AUC=0.882 (0.825-0.931) in SMART).
Conclusion: High VLDL-C associates with a higher number of differentially expressed plasma proteins versus high LDL-C and none of the proteins were in common. Our data do not underestimate the value of LDL-C in ACVD but reinforce the concept that VLDL-C may also promote different atherosclerotic pathways involved in determining ACVD.