Recent Highlights in ATVB: Systems Biology and Non-Invasive Imaging of Atherosclerosis

Introduction

Atherosclerosis is a systemic disease of the arterial vessel wall. While the mortality due to cardiovascular events is decreasing, the prevalence of atherosclerosis and its comorbidities, and the consequent heath care costs are expected to rise sharply in the near future¹.

Since the precise etiology and pathogenesis of this complex, multi-factorial disease are still not fully understood, the clinical assessment of cardiovascular risk has been traditionally based on population risk factors (https://www.framinghamheartstudy.org²). However, this approach still largely fails to capture the individual's cardiovascular risk: most cardiovascular events occur in patients with 1 or few traditional risk factors, while individuals classified as high-risk may never experience clinical events³.

The past 10 years have seen a significant paradigm shift in our understanding of the mechanisms of atherogenesis. From being considered the mere result of passive lipid accumulation in the vessel wall, atherosclerosis is now classified as an active inflammatory condition^{4, 5}. The presence of abundant, active inflammatory cells is a known hallmark of high-risk, vulnerable atherosclerotic plaques^{4, 5}. Many studies have identified several systemic pro-inflammatory conditions (such as lupus⁶, rheumatoid arthritis^7-9, and primary cardiovascular events themselves¹⁰) as emerging, independent risk factors for atherosclerosis. New evidence suggests that atherosclerosis arises from the complex influence of genetic, environmental, and behavioural variables on systemic and local inflammation through a complex network of molecules, cells and organs.

Thanks to the recent technological advancements of high throughput ‘-omics’, a plethora of these genes, proteins and cells involved in the atherosclerotic cascade have already been identified. However, many steps still need to be taken to fully exploit this information, and improve patients' risk stratification and anti-atherosclerotic therapies. The mutual relationship between genetic and molecular “key drivers”, and their interplay in peripheral blood, atherosclerotic plaques and other organs still need to be established. Furthermore, quantitative methods to non-invasively measure these markers in the vessel wall and other tissues need to be developed and validated before they can be routinely used in the clinical practice.

In this review, we highlight novel work in high-throughput ‘-omics’, systems biology and non-invasive quantitative imaging of atherosclerosis, with specific emphasis on articles recently featured in ATVB. New advancements in these disciplines are discussed separately, as well as in their complementary applications, to showcase how these fields may be successfully integrated to improve cardiovascular risk prediction and patients' stratification in the future clinical practice (Figure 1).

Figure 1.

Schematic representation of the integration between data from high-throughput ‘-omics’ and imaging phenotypes in a systems biology approach, focused on building biological networks from genes and molecules, up to tissues and organs to study disease mechanisms.

Systems Biology and high-throughput ‘-omics’ of atherosclerosis

Systems biology can be broadly defined as the combination of experimental and computational research used to understand complex biological systems¹¹. It involves the integration of data derived from high throughput ‘-omics’ with computational/statistical tools to build comprehensive networks, and predictive physiological models¹² (Figure 1).

In recent years high throughput ‘-omics’ have been intensely applied to the study of atherosclerosis¹³, with the aim to deepen our knowledge of this disease and refine our tools for cardiovascular risk assessment. For example, recent data¹⁴ from the Erasmus Rucphen Family and Rotterdam Study have shown that common genetic variants for total cholesterol and LDL-C are, in combination, significantly associated with subclinical and clinical outcomes of atherosclerosis. Other investigations have suggested association between soluble interleukin (IL) -2 receptor subunit α (sIL-2R α, regulating lymphocytes activation) and cardiovascular disease (CVD), and also uncovered the genetic determinants of its levels¹⁵. These results substantiate our existing knowledge on the impact of long-life, cumulative exposure to modifiable and non-modifiable risk factors on cardiovascular risk^{16, 17}. While previous genome wide association studies (GWAS)¹⁸ have reported only marginal improvements^{19, 20} in risk stratification compared to the Framingham Risk Score^{3, 21, 22}, more recent investigations^{23, 24} found that genetic risk scores (GRS) from validated GWAS significantly improved prediction of cardiovascular events over traditional risk factors, from 4-5%²³ up to 12%²⁴. Extensive bioinformatics analyses²⁵ of loci known to be associated with coronary artery disease (CAD) from GWAS, suggest that sequence variations mainly occur in noncoding regions of the genome and promote CAD risk by either affecting gene expression or by leading to amino acid changes. By extending the list of candidate genes likely linked to CAD, these studies suggest that bioinformatics analysis of GWAS may be beneficial in the context of atherosclerosis as well as other diseases.

By quantifying the expression levels of protein-coding genes, transcriptomic studies ^26,27,28,29 identified several promising biomarkers for cardiovascular disease²⁶ and vulnerable plaques³⁰. The Systems Approach to Biomarker Research (SABRe) study (launched as part of the Framingham Heart Study) recently found that 35 genes and 3 gene clusters (metagenes) were differentially expressed in cases with CHD versus controls, while GUK1 38 and other genes were differentially spliced in the two populations³¹. More recently, transcriptomics has been extended to study non-coding, regulatory gene transcripts, such as microRNAs. Several types of circulating microRNAs have recently been implicated in CVD^{3, 32-45}. Using transcriptomics the prospective Bruneck study found that 3 miRNAs (miR-126, miR-223 and miR-197) originating from platelets improved patients' risk classification for coronary heart disease compared to the Framingham Risk Score⁴⁶. Transcriptomics analyses have also directly implicated several miRNAs in the development of vascular inflammation as mediated by wall sheer stress (WSS)⁴⁷. Numerous flow-sensitive miRNAs (miR-10a, miR-633, miR-21, miR-92a) have already been identified⁴⁸.

Proteomic studies^{49, 50,51-53,54,55,56} collectively identified more than 150 potential single biomarkers of cardiovascular disease^{49, 57, 58}. More recent studies have seen a shift from the analysis of single protein markers to simultaneously quantifying the combination of the levels of several proteins in multi-marker panel analyses. A recent study in 135 myocardial infarction (MI) cases and matched controls⁵⁹ found that both single and multi-marker analysis of plasma proteins were associated with incidence of MI, with multi-markers analysis providing higher discrimination. Similar results were found in a prospective study, where multi-markers analysis in 336 patients was found to be predictive of cardiovascular disease (CVD, p<0.0001), with moderate improvement over traditional risk factors (C-statistic of 0.69 versus 0.73) ⁵⁹.

While these and previous studies offer invaluable insights on the genetic, molecular and metabolic basis of atherosclerosis, they still do not provide a cohesive, integrated approach to the study of this disease. Recent approaches have tried to overcome this obstacle by applying a combination of ‘-omics’. For example a combination of GWAS and transcriptomics is being used to investigate the recently suggested association between the haplogroup 1 of the Y male chromosome and an increased risk of coronary artery disease (CAD)⁶⁰. New studies at the interface between metabolomics (lipidomics^61-63) and proteomics have shed light on the complexity of the human lipoproteome ^{64, 65}, and helped characterizing more than 90 forms of high-density liproproteins (HDL) particles associated with different lipoproteins, with a diverse range of anti-atherosclerotic properties ⁶⁶, beyond the known effects on cholesterol efflux. A combined approach⁶⁷ using adipose tissue transcriptomics, HDL lipidomics, and genotyping found a “shift” towards inflammatory HDL particle types in individuals with low HDL cholesterol, mirrored by an increase in inflammatory markers in adipose tissue and in the peripheral blood.

By taking these approaches a step forward, Shang et al^{68, 69} employ gene subnetworks profiling of the Stockholm Atherosclerosis Gene Expression (STAGE) study to find a candidate gene strongly correlated to leucocyte migration, and assess its association with clinical manifestation of disease (coronary angiography, carotid intima-media thickness by ultrasound). This work offers an example of how a more integrated systems biology approach may be used to better understand the process of atherogenesis, from the genetic determinants to the phenotypic manifestations of systemic and vascular inflammation. As part of the SABRe study mentioned above, Huan et al⁷⁰ also employ an integrated systems biology approach to find differential gene coexpression modules (DMs) in the blood of subjects with coronary heart disease (CHD) and matched controls. By integrating these results with previous GWAS and single nucleotide polymorphisms (SNPs), the authors are able to draw a causal relationship between the DMs and CHD in this cohort. With the further integration of Bayesian networks and protein-protein interaction networks they also identify key drivers (KD), regulatory genes important for the DMs stability and therefore potential targets for novel drugs. This network driven, integrated approach not only identifies genes related to CHD, but also strives to build a network structure that informs on the molecular interactions of genes associated with CHD risk.

Despite the enormous potential of a panomic/systems biology approach to atherosclerosis, several obstacles have to be overcome so that ‘-omics’ can be successfully used in future clinical practice. Firstly, a causal relationship between biomarkers and disease mechanisms has to be solidly established. Dissecting causal effects from confounders can prove very challenging in cross-sectional studies, while prospective, causal studies with cardiovascular events as endpoints are costly and lengthy to perform. Furthermore, performing ‘-omics’ analyses on direct tissue samples may not be always feasible, while peripheral blood analyses may only reflect transient changes in metabolites that do not necessarily inform on the overall disease activity.

Imaging

In recent years medical imaging has made great strides in the evaluation of virtually every organ in the body, including atherosclerotic plaques. Modality specific imaging traits (imaging “phenotypes”) emerge from the combination of tissues structure, physiology and function, and inform on organs physiology and pathology^{71, 72}.

Several imaging modalities have already found widespread use in the clinical practice to evaluate atherosclerotic burden. Carotid intima-media thickness (CIMT) by surface ultrasound (US)^{73, 74} is one such technique, although its clinical usefulness to significantly improve risk prediction over traditional risk factors has recently been questioned^{75, 76}. Recently, CIMT was found to decrease in subjects consuming a Mediterranean diet (MedDiet) supplemented with 30 g/d of mixed nuts, compared with a control, low fat diet, thereby corroborating the results form the Primary Prevention of Cardiovascular Disease with a Mediterranean Diet (PREDIMED) trial⁷⁷. Other than CIMT, surface US can be used to measure other parameters related to plaque vulnerability, such as vascular strain or the extent of plaque microvasculature using non targeted micro bubbles⁷⁸. US with micro bubbles targeted to vascular cell adhesion molecule 1 (VCAM-1) and platelet glycoprotein Ibα was recently validated in genetically modified mice as being able to assess the anti-inflammatory properties of apocynin, before detectable changes in macrophages burden⁷⁹.

More invasive procedures involving intravascular (IVUS) or transesophageal US⁷⁸ can also be performed. For example, transesophageal US was recently employed in mongrel dogs to quantify changes in aortic area and elastic properties from velocity-vector imaging (VVI) with aging⁸⁰. Another study used a combination of multi-vessel IVUS and novel near infra-red spectroscopy (NIRS)^{81, 82} to evaluate features of vulnerability in fibroatheromas of diabetic/hypercholesterolemic pigs⁸³. Longitudinally, IVUS demonstrated a progressive increase in plaque and media areas, with the appearance of necrotic cores and regions of positive vascular remodeling. Compared with histological samples, NIRS positive lesions exhibited features of high-risk fibroatheromas, such as large plaque size, necrotic cores, thin fibrous cap, and abundant presence of inflammatory cells. A more recent study using IVUS demonstrated a greater progression in CAD patients classified as statin hyporesponders, compared to individuals that exhibited LDL-C reductions of more than 15% from baseline⁸⁴.

Coronary calcium score (CAC) evaluated by non contrast enhanced computed tomography (CT)^{85, 86}, is another non-invasive measure of overall atherosclerotic burden, which has been described to predict the risk of future clinical events^{87, 88}. A recent follow-up study of the Multi-Ethnic Study of Atherosclerosis (MESA) trial found that abdominal aortic calcium (AAC) and CAC were predictors of coronary heart disease and cardiovascular events independent of one another, with only AAC being independently associated with cardiovascular mortality, and showing a stronger association than CAC with overall mortality⁸⁹. Recent studies ⁹⁰ suggest that this measure could be complemented by cardiac computed tomography angiography (CCTA)⁹¹ with the use of a iodinated contrast agent to provide additional information on the degree and distribution of coronary plaque stenosis, vessel wall positive remodeling, and plaque composition (such as presence of low-attenuation plaques and spotty calcification), which have been identified as markers of vulnerability⁹².

Other modalities, such as magnetic resonance imaging (MRI), and positron emission tomography (PET) are actively being investigated in both the pre-clinical and clinical research arenas for their potential translation into clinical practice. With its superior soft tissue contrast compared to CT, and the possibility to image large segments of the vasculature with high spatial resolution, non-contrast enhanced MRI has been extensively investigated as a method to characterize atherosclerotic plaques components, such as lipid core, fibrous cap, intraplaque hemorrhage and presence of thrombi^93-95. A recent study⁹⁶ performed on 1016 individuals from the Framingham Heart Study Offspring cohort studied the prevalence and risk factor (RF) correlates for aortic plaque (AP) detected by MRI and CT. The study found that while AP by both imaging modalities is associated with smoking and increasing age, the association with other risk factors differs between calcified plaques detected by CT, and non-calcified lesions detected by MRI. The study postulates that the relative predictive value of AP detected by MRI and CT still needs to be investigated. Combined with the use of non-specific gadolinium (Gd) based contrast agents, MRI has been also use to interrogate plaque physiology and quantify the extent of microvascular permeability^97-99, another important hallmark of plaque vulnerability. Other physiological parameters, such as carotid arterial strain and distensibility calculated from MRI, have been shown to predict the future incidence of cerebral microbleeds (CMBs) in 2512 patients recruited as part of the prospective, population-based Age, Gene/Environment Susceptibility (AGES)-Reykjavik study¹⁰⁰. PET, combined with the anatomical information from CT¹⁰¹ or, more recently, MRI⁸⁸, has been extensively validated to quantify vascular inflammation itself and its changes upon therapeutic intervention, both in humans and animals using the tracer 18F fluorodeoxyglucose (FDG)^{102, 103}. Recently, 18F-FDG PET was used to demonstrate a decrease in vascular inflammation in Ldlr-/- atherosclerotic mice after treatment with melanocortin peptides¹⁰⁴. Other PET tracers are now being investigated, such as sodium fluoride (NaF)¹⁰⁵ targeting micro-calcification, or tracers targeted to specific molecules, such as αvβ3, VCAM-1¹⁰², 68Ga-Fucoidan for P-selectin¹⁰⁶ (abundantly expressed in vulnerable, but not stable plaques), ⁶⁴Cu-FBP8 for thrombus detection and fibrin quantification^{107, and 64}Cu-DOTATATE to selectively quantify plaque macrophages via the somatostatin receptor subtype-2¹⁰⁸.

Among emerging imaging modalities, optical coherence tomography (OCT) is gaining increasing interest because of its ability to provide high-resolution images of tissues microstructure. Recently, OCT was first validated in atherosclerotic rabbits and then used in a prospective study in patients to evaluate vascular healing after the implantation of drug-eluting stents, where it was found to be able to discriminate between immature and mature (healed) neointimal tissue¹⁰⁹. Another study used OCT in 40 patients with mild coronary atherosclerosis to study the composition of coronary segments after stimulus with acetylcholine. The study found that the segments showing presence of macrophages and microchannels (microvasculature), exhibited a more prominent change in the diameter of coronary arteries, indicating higher endothelial dysfunction¹¹⁰. Recently, novel, in vivo multiphoton laser scanning microscopy was used to study plaque microvasculature and confirmed that plaque-associated vasa vasorum exhibit increased permeability, and increased leukocyte adhesion and extravasation¹¹¹.

Imaging as a tool for systems biology

From the account above, it emerges that, through the non-invasive characterization of tissues anatomy and physiology, medical imaging may be an ideal complement to ‘-omics’ technologies for a comprehensive systems biology approach to cardiovascular disease. Several approaches are currently being explored to successfully integrate imaging in this framework.

Although confined to pre-clinical investigations, molecular imaging already reports on specific biological processes (optical imaging, fluorescence imaging), and can even directly quantify gene expression (bioluminescence)¹ or cells (macrophages) development, migration^{112, 113} and presence in tissues throughout the body¹¹⁴. Among translatable modalities, molecular imaging with MRI and PET can also similarly be used for this purpose. Aside from the increasing number of MRI contrast agents being developed to target specific biomarkers⁹³, MRI with ultra small superparamagnetic iron particles (USPIO) has been widely validated as a tool to detect plaque macrophages content in atherosclerotic plaques in both animals¹¹⁵ and patients¹¹⁶. Recent studies in mice have shown the successful integration of proteomics, metabolomics and quantitative and anatomical MR imaging to phenotype transgenic mice in regards to creatinine and phosphocreatinine cardiac metabolism¹¹⁷. In addition to the quantification of plaque local inflammation, the use of 18F FDG PET was recently extended to study the interplay between local and systemic inflammation and to substantiate the existence of a cardiosplenic axis in humans (implicated from animal models¹⁰ in the high incidence of secondary cardiovascular events in patients with previous MI) ¹¹⁸. Similarly, another study has recently demonstrated increased vascular inflammation by 18F FDG PET in patients with psoriasis, independent of cardiovascular risk factors¹¹⁹. These studies shows an example of the integration of 18F FDG PET in a ‘systems physiology’ approach. Similarly, several clinical imaging modalities are currently being investigated to quantify non-invasively the extent (CT, MRI) and metabolic activity (18F FDG PET) of visceral and subcutaneous body fat, regarded as a potential marker and risk predictor of cardiovascular disease¹²⁰.

Some studies have already focused on investigating the genetic and molecular correlates of imaging traits. A recent study identified the genetic variations influencing the effect of smoking on CIMT, thereby exemplifying how the study of gene-environment interactions may explain the interindividual variation in both cardiovascular events and surrogate measures of cardiovascular risk¹²¹. The Genetic Loci and the Burden of Atherosclerotic Lesions (GLOBAL) study¹²² (NCT01738828) brings this concept to a different level by aiming to comprehensively integrate plaque phenotype by cardiovascular imaging, with a panomic approach including genomic, transcriptomic, proteomic, metabolomics and lipidomic in a systems biology framework. The study plans to examine single-omic and multi-omic associations with each imaging phenotype evaluated (CAC and CT angiography) in training and validation datasets.

Final remarks

In this report we have reviewed the most recent advances in ‘-omics’/systems biology and non-invasive medical imaging applied to atherosclerosis and cardiovascular disease, with specific focus on papers recently published in ATVB.

The combination of ‘-omics’ and systems biology is nowadays used more and more frequently to elucidate mechanisms of disease, and is also being investigated as a complement to clinical data to improve patients risk stratification. Examples of these approaches are GWAS^{23, 24}, the study of coding and non-coding RNAs using transcriptomics^{31, 32-45}, as well as tissue and blood proteomics^{49, 50,51-53,54,55,56,57,58,59}. More recent, sophisticated analyses^{68, 69,70} aim to integrate this information in a comprehensive, systems biology approach to build a network structure of the molecular interactions in cardiovascular disease.

Medical imaging is also making tremendous advancements in the diagnosis and characterization of cardiovascular disease. Modalities such as US, CT and MRI already allow for the accurate quantification of plaque burden and lesion characterization. Other approaches, such as PET and MRI with contrast report on relevant physiological parameters, i.e. plaque permeability and inflammation, while molecular imaging techniques can shed light on specific molecular/cellular processes.

While taken separately both ‘-omics’ and medical imaging can already tremendously contribute to our understanding of cardiovascular disease and to our ability to stratify patients' risk, their successful integration may bring additional, significant benefits.

Similarly to what is being recently proposed in oncology^123-129, the first step in integrating these two disciplines will be establishing the association¹³⁰ between “imaging phenotypes” and specific genetic, molecular and cellular signatures in atherosclerotic plaques and other organs involved in atherogenesis. Once these correlations will be robustly established, the use of imaging phenotypes may be extended to function as predictors^{124, 125} of plaques genetic and molecular makeup^123-129 in both the pre-clinical and clinical arenas. In this scenario, the integration of imaging and ‘-omics’ in a systems biology framework may be better positioned to improve risk stratification and assessment of therapeutic response of atherosclerotic patients in the future clinical practice (Figure 1).