Deciphering the proteomic basis of early vascular aging in hypertensive patients

This editorial refers to ‘Blood pressure, proteomic vascular ageing, and incident cardiovascular disease’, by M. Kou et al., https://doi.org/10.1093/eurjpc/zwaf757. Early vascular aging is a hallmark in today’s medicine, particularly in hypertension.1 This notion, introduced 15 years ago by Nilsson and Laurent,2 covers the fact that for a given exposure to risk factors, target organ damage [hypertension-mediated organ damage (HMOD)] is more severe in some patients [defined as early vascular aging (EVA)] and slower in others [supernormal vascular aging (SUPERNOVA)]. Indeed, EVA and SUPERNOVA are associated with hard outcomes.3 The concept is easy to capture and effective for sorting high-risk from low-risk patients. It has been a great success in terms of communication for cardiovascular risk management. The mechanisms by which certain individuals experience EVA rather than SUPERNOVA remains incompletely understood. In particular, molecular pathways to EVA have been seldom explored. The primary aim of the present study by Kou et al.4 is to use the UK database for leveraging the question. They rely on the notion of ‘vascular age gap’ deriving from heart vessels enriched proteomic analysis and test the relation between this metric of VA and blood pressure categories and incident CVD. The authors used a subset of the UK biobank population in which proteomic data is available along with blood pressure measures (n = 45,387). They selected a subset of the deep proteomic analysis, targeted on heart/vessels proteins panel using the Gene Tissue Expression Atlas (GTEx) for definition. After exclusion of false-positive hits, the vascular signature comprised 21 proteins. They used the approach of Argentieri,5 who selected 204 proteins highly associated with chronological age, of which 24 (21) belonged to heart/vessels categories. The authors then defined a vascular age gap (VAG), corresponding to the subjects presenting with proteomic vascular age different from chronological age. In that sense, VAG is genuinely an index of accelerated vascular aging. The main finding is that high blood pressure is positively associated with the VAG, i.e. that exposure to graduated elevated blood pressure imprints a proteomic signature of EVA. The strength of the association is moderate (0.122 ± 0.048; 0.266 ± 0.049 and 0.494 ± 0.071 years for high-normal blood pressure, Grade 1 and Grade 2 hypertension, respectively), compared to normal BP vascular aging. Differences were highly significant due to the very large population involved. Similar and robust results were obtained after multiple adjustments and using SBP and DBP as continuous variables. Most importantly, the mediation analysis showed that the prediction of incident cardiovascular events was mostly related to a direct effect of SBP and DBP. However, some of the associations remained significantly related to VAG for Grade 2 hypertension only (3–13%). The study brings new elements supporting the notion of EVA, this time based on proteomic signatures. The originality of the study is its large population, the selection of the most meaningful cardiovascular proteins associated with chronological age, and the identification of discordant profiles of vascular aging gap, i.e. those with discordant calculated age relative to chronological age. This study is a convincing proof of concept that exposure to high BP induces changes in the plasma proteome considered characteristic of EVA. At this stage, it does not provide conclusive hints for mechanisms, although several proteins are good candidates. Indeed, despite its massive use of advanced statistical procedures for selecting meaningful proteomic profiles, it is impossible to discriminate physiologically important proteins (although at very low concentrations) from proteins that are structurally meaningful but found only because they are abundant with low turnover (for instance, elastin). Signalling pathways and metabolic cascades were not investigated here; however, this has been explored in other circumstances6 and it might be the next step to take. Furthermore, it is impossible to discriminate what is causally linked to both hypertension occurrence and vascular aging. Indeed, a specific combination of genetic traits may lead to early occurrence of vascular aging and early onset of hypertension in a bidirectional relationship. As an example, large artery stiffness is at the same time a tissular marker of EVA and a risk factor for hypertension,7 and hypertension is the strongest determinant of increased arterial stiffness.8 One of the criticisms of this approach is that it fully relies on the paper by Argentieri, which, despite its major qualities, can be caricaturized as a complicated and expensive way of measuring age, when asking for the date of birth is the gold standard. Argentieri did investigate chronological aging with a quasi-perfect fit (R² = 0.90). The point is somewhat missed since what really matters is to identify proteins associated with accelerated or slow course of aging (understood cross-sectionally or longitudinally). Proteins retained in the present analysis were considered as ‘cardiovascular’ proteins, based on GTEx index. This selection is somewhat arbitrary, since it reflects the criteria of entry in libraries, rather than a true tissue specificity. For instance, tumour necrosis factor receptor superfamily member 11B (TNFRSF11B) emerges as a potentially interesting protein for explaining VAG, yet it is mostly involved in bone metabolism and may only marginally affect the cardiovascular tissues by inhibiting calcification. Conversely, some unselected proteins might be of interest for cardiovascular diseases. For example, apolipoprotein E was not selected because attributed to brain and Alzheimer’s disease rather than cardiovascular notwithstanding its recognized role in atherosclerosis. The protein selection done by the authors was less effective in predicting chronological age (R² = 0.70) than the 204 original proteins from Argentieri (R² = 0.90). It is likely then that a specific set of the 204 proteins can predict the vascular gap, since the populations are identical. This analysis is missing. It could shed some light on the involved mechanisms in vascular aging. Finally, inference from proteins present in the plasma leads to interpretation at tissular level. There is no guarantee that plasma levels are correlated with tissular responses.9 The authors used a gradient boosting machine learning model (LightGBM) for predicting proteomic age. Although referenced and validated, this approach remains a black box, from which no logical inference can be derived. Indeed, while boosting is effective at constructing decision trees or linear regressions in a complex subset of variables, this is done at the cost of intelligibility and interpretability.10 Following the paths of large numbers of decision trees is difficult, and similar datasets may lead to different results. The authors used an approximation for BP values in treated patients by adding 15 and 10 mmHg to systolic and diastolic blood pressures, respectively, to adjust for treatment effects. Even if referenced, this approximation is regrettable, and it would have been better to adjust for treatments in a cleaner way. Indeed, all treatments are not equal relative to EVA, and each therapeutic class induces specific proteomic changes.11 It also implies that a residual risk (corresponding to the risk associated to 15 mmHg) persists in treated hypertensives, irrespective to the control of blood pressure, which is not clearly established. It remains that this is the first study to evaluate the associations between high blood pressure and proteomic vascular aging. It should pave the path to more mechanistic studies, leveraging on the same approach, combined with other omics data (metabolomics, transcriptomics, etc.) on well-phenotyped cohorts, assessing the longitudinal changes in structural markers of vascular aging. This combined approach may offer greater clues for translational applications since proteins and signalling pathways are major drug targets for preventing early vascular aging. Pierre Boutouyrie (Conceptualization [lead], Writing—original draft [lead], Writing—review & editing [lead]), Karen Barchetti (Validation [supporting], Writing—review & editing [supporting]), and Marc Antoine Custaud (Writing—review & editing [equal])