Vascular andhttp://ves.sagepub.com/ Endovascular Surgery

Systems Biology of Human Atherosclerosis Joseph Shalhoub, Markus B. Sikkel, Kerry J. Davies, Panagiotis A. Vorkas, Elizabeth J. Want and Alun H. Davies VASC ENDOVASCULAR SURG 2014 48: 5 originally published online 7 November 2013 DOI: 10.1177/1538574413510628 The online version of this article can be found at: http://ves.sagepub.com/content/48/1/5

Published by: http://www.sagepublications.com

Additional services and information for Vascular and Endovascular Surgery can be found at: Email Alerts: http://ves.sagepub.com/cgi/alerts Subscriptions: http://ves.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav Citations: http://ves.sagepub.com/content/48/1/5.refs.html

>> Version of Record - Dec 17, 2013 OnlineFirst Version of Record - Nov 7, 2013 What is This?

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

Basic Science Review Article

Systems Biology of Human Atherosclerosis

Vascular and Endovascular Surgery 2014, Vol 48(1) 5-17 ª The Author(s) 2013 Reprints and permission: sagepub.com/journalsPermissions.nav DOI: 10.1177/1538574413510628 ves.sagepub.com

Joseph Shalhoub1, Markus B. Sikkel2, Kerry J. Davies1, Panagiotis A. Vorkas3, Elizabeth J. Want3, and Alun H. Davies1

Abstract Systems biology describes a holistic and integrative approach to understand physiology and pathology. The ‘‘omic’’ disciplines include genomics, transcriptomics, proteomics, and metabolic profiling (metabonomics and metabolomics). By adopting a stance, which is opposing (yet complimentary) to conventional research techniques, systems biology offers an overview by assessing the ‘‘net’’ biological effect imposed by a disease or nondisease state. There are a number of different organizational levels to be understood, from DNA to protein, metabolites, cells, organs and organisms, even beyond this to an organism’s context. Systems biology relies on the existence of ‘‘nodes’’ and ‘‘edges.’’ Nodes are the constituent part of the system being studied (eg, proteins in the proteome), while the edges are the way these constituents interact. In future, it will be increasingly important to collaborate, collating data from multiple studies to improve data sets, making them freely available and undertaking integrative analyses. Keywords systems biology, atherosclerosis, genomics, transcriptomics, proteomics, metabolomics, metabonomics Atherosclerosis is a multistage entity characterized by the interaction between multiple biological processes. These center upon inflammation and also include matrix degradation, angiogenesis, intraplaque hemorrhage, oxidative stress, lipid metabolism, apoptosis, and autophagy.1 This is by no means an exhaustive list but offers insight into the complexity of the biology of atherosclerosis development, progression, destabilization, and the complications thereof. The reason why some atheroscelorotic plaques arise, rupture, and embolize, while others do not, has been an important area of clinical, imaging, and laboratory research. Effective atherosclerosis research has the facility to elucidate mechanisms of atherosclerosis destabilization, uncover targets for functional imaging for risk stratification, discover biomarkers or biomarker signatures, and contribute to the development of targeted plaque stabilizing therapies. Reflecting its share of the global burden of morbidity and mortality, atherosclerosis research has been increasing steadily over the past 2 decades (Figure 1). This was particularly the case since the late 1990s with the formal acknowledgment that atherosclerosis is an active inflammatory process driven by inflammation.3 Systems biology is a term which describes a holistic and integrative approach to the understanding of physiology and pathology. By adopting a stance which is opposing (yet complimentary) to conventional research techniques, it offers an overview by assessing the ‘‘net’’ biological effect imposed by a disease or nondisease state.

to understand fundamental laws at 1 level does not automatically give us the ability to extrapolate those laws and to reconstruct the universe based on those laws.4 In other words, understanding what happens at 1 organizational level does not give us the ability to understand what will happen at higher organizational levels—this is termed the principle of emergent complexity.5 In the modern study of systems biology, there are a number of different organizational levels to be understood; from the arbitrary starting point of DNA to proteins, metabolites (Figure 2), cells, organs and organisms, and even beyond this to the organism’s context (its physical environment and interaction with other organisms). The important aspect of seeing an organism at these different levels, as systems biologists seek to do, is the recognition that with each step change in scale, properties emerge that could not be predicted from analysis of the individual components at lower scales. To take a real-world example visible to all of us, it is impossible to predict the complex patterns formed by a flock of birds in flight by studying the individual animal or even 10 individual animals.

The Philosophy of Systems Biology

Corresponding Author: Joseph Shalhoub, Academic Section of Vascular Surgery, 4th Floor, Charing Cross Hospital, Fulham Palace Road, London W6 8RF, United Kingdom. Email: [email protected]

The limitations of reductionist science have been appreciated in the physical sciences for longer than in biology. The ability

1 Department of Surgery & Cancer, Academic Section of Vascular Surgery, Imperial College London, United Kingdom 2 Myocardial Function Section, National Heart & Lung Institute, Imperial College London, United Kingdom 3 Department of Surgery & Cancer, Computational & Systems Medicine, Imperial College London, United Kingdom

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

6

Vascular and Endovascular Surgery 48(1) system being studied (eg, proteins in the proteome) while edges are the way these constituents interact. Despite the widespread viewpoint that systems biology is somehow synonymous with high-throughput detection techniques such as genomics and proteomics, this is an oversimplification. Although such disciplines form an important part of systems biology, they only give answers about the nodes in a system rather than the edges. Systems biologists are often more interested in the edges in a system than the nodes.9 In fact, systems biology can deal with conceptual nodes and edges and not just those that are physically present. As an example of a conceptual systems framework, autonomous oscillations of the cell’s mitotic cycle require similar concepts to understand as the rhythmic contraction of a cardiomyocyte (both being autonomous oscillators with external inputs). Beginning with complex biological phenomena and providing a simpler and more abstract framework that explains why these events occur is just as important in systems biology as high-throughput screening methodologies.6

Figure 1. The rise of atherosclerosis research. One of the numerous reasons for the increase in such publications is thought to be the use of traditional laboratory techniques by focal research groups working in isolation. This results in the duplication of process and, often, data and publications based upon smaller numbers of samples than would be possible through collaboration.2 Such an approach is, however, essential particularly for (but not exclusively in) the validation of findings and is often driven by a very narrow experimental hypothesis. Traditional laboratory methodologies are reductionist in philosophy, understanding the complex processes underlying atherosclerosis by deconstructing them into constituent components or pathways.

Returning to biological concepts, if biologists limited themselves to the study 2 interacting proteins in a system they would never observe oscillation, which requires 3 proteins connected in a negative feedback loop.6 This concept can also be seen with respect to our lack of understanding of the complexity of human life. A satisfying explanation of our complexity is impossible with reductionist theorization. To say that the mysteries of life were unraveled with the discovery of DNA, and the subsequent flow of information encoded therein to produce proteins is overstating the case. There are a disturbingly small number of coding genes (*25 000) which cannot approximate the number of functions in an organism.7 The viewpoint that DNA can directly lead to an understanding of an organism ignores interactions between gene products, their subcellular compartmentalization, posttranslational modification, and degradation, not to mention the complexity that comes from each subsequent organizational level. In addition, the interspecies similarity of genomes, about 96% between humans and chimpanzees,8 is another compelling argument against the ‘‘constructionist’’ viewpoint that organisms are merely a result of the correct combination of gene transcription events. Systems biology seeks to gain an insight into what constitutes the remainder of the complexity of the function of an organism, both in health and in disease. An important concept in systems biology, regardless of the level at which it is being studied, is the existence of ‘‘nodes’’ and ‘‘edges’’ (Figure 3). Nodes are the constituent parts of the

Systems Biology Versus Reductionist Science Much of modern biology has been shaped by attempts in the last century or so to reduce biological phenomena to the behavior of molecules.10 Although this has been remarkably successful, it has become clear that a discrete biological function can only rarely be ascribed to a single molecule in the sense that, for example, hemoglobin is a molecule that transports gas molecules in the bloodstream. In the rare situations in which this is the case, reductionist methodologies can provide an excellent understanding of why abnormalities of these molecules can cause disease. To take an example familiar to the physician managing atherosclerosis, familial hypercholesterolemia (FH) is one such disorder. However, even with this example of an apparent triumph for reductionist biology, the single abnormal gene cannot explain the complexity of the resultant disease phenotype with a wide variability in the severity of clinical symptoms even in patients with the same mutation.11 Systems biologists would look to the network of molecules participating in cholesterol uptake and atherosclerosis to explain such variability. For example in patients with milder manifestations of FH, other molecules in the ‘‘module’’ of molecules whose function is to take up low-density lipoprotein (LDL) from the bloodstream (eg, LDL receptor [LDLR] adaptor protein 1, LDLR adaptor protein 2, and clathrin triskelion) may alter their activity to provide some form of compensation for the LDLR part of the module. This sort of compensation can only be seen from a ‘‘systems-biology’’ viewpoint, which places molecules in their network context. Of course, systems biology stands to gain a lot from reductionism and in some ways is dependent on it—truly understanding a system requires knowledge of the constituent parts.6 On the other hand, most diseases do not result from a singlemolecular change like this and are instead the result of alterations in more than one component of a complex biological module. The abnormality causing the disease may not be overt at lower organizational levels but may only become evident at

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

Shalhoub et al

7

Figure 2. The synthetic ‘‘workflow’’ of a cell. Simplification of a single pathway in a single cell of a particular cell type. In reality, there are multiple interacting pathways in multiple cells of different types, with microenvironmental conditions asserting alterations at multiple levels on multiple pathways. The biological entities within the workflow may be found within different ‘‘compartments’’ of atherosclerotic environment: the cell nucleus, cytoplasm, extracellular matrix, and some components are released into the bloodstream.

higher hierarchical levels. With reference to the pathophysiology of heart failure, this has been termed ‘‘failing complexity’’ (rather than failure of an organ, cell, molecule, transcript, or gene) by De Keulenaer and Brutsaert.12 The same can be said for atherosclerosis. Clearly, there are multiple organizational levels at which atherosclerosis can be studied, and each of these can result in significant insights into the disease process and potential therapy. However, the disease itself cannot be seen at any of these individual levels as the result of failure but instead is contingent upon the failure of complexity that results from the integration of lower levels into higher levels (Figure 3). That all these levels are important is indicated by recent data pertaining to obesity. At the level of the genome, there are predictably important highrisk alleles that can increase the risk of obesity by up to 67% among homozygous persons.13 At the other end of the network scale, the social network can provide just as important an influence on the risk of obesity than the genetic one.14 Christakis and Fowler studied the social networks of participants in the Framingham Heart Study and observed that if a friend becomes obese, the chances of following suit are increased by 171%.15 One of the central concepts in systems biology is that networks, rather than linear pathways, underlie biological processes. This began to be understood when classic metabolic pathways were represented in network diagrams constituting

nodes and edges.16 This type of interconnection is seen at all organizational levels (Figure 3). More importantly, the type of organization in all these networks seems to follow the same mathematical rules—exhibiting an architecture described as ‘‘scale-free.’’ Such scale-free networks are so-called because this mathematical architecture holds no matter how large or small the particular network is (much as a fractal pattern looks similar however ‘‘zoomed-in’’ the view of it17). As described earlier, this architecture results in a network in which most nodes have few links (edges) connected to them but there are a few highly connected nodes (hubs). As the system grows over time, new links to the highly connected nodes are favored (‘‘the rich get richer’’). Such networks are inherently robust, since there are multiple different pathways to get from one node to another. In a typical scale-free network, up to 80% of the links can be randomly destroyed before there is a catastrophic failure of complexity. A challenge is constructing detailed biological networks at each level from gene to organism (and beyond, and then to connect the levels by integrating orthogonal data sets.16 Another point of difference between reductionist and systems biology methodology is the frequent simplification of inputs and outputs in reductionist experimentation. This is often done out of necessity, since it is impossible to apply the entirety of the biological complexity that results in a disease like atherosclerosis to the model organism or cell line being

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

8

Vascular and Endovascular Surgery 48(1)

Figure 3. The various hierarchical levels of organization within an organism and how their study can be applied to atherosclerosis. The reductionist approaches at each level are represented on the right. The systems biology approach can function at any of these levels and even at multiple levels to provide an understanding of how a network of biological dysfunction, that is, a failure of complexity, rather than the failure of a single component of the system, causes disease. The organization of nodes and edges within the network at each level is similar, represented by a few hub nodes (red) with multiple connections and the remaining nodes (green) having fewer connections. These are termed ‘‘scale-free networks.’’ ACE indicates angiotensin converting enzyme; LDLR, low-density lipoprotein receptor; MI, myocardial infarction; SOD, superoxide dismutase; TGF, transforming growth factor; TLR, toll-like receptor.

studied. Hence, we are left with surrogate, oversimplified inputs applied to model systems—for example estrogen in place of gender, leptin and adiponectin for obesity, and hyperglycemia and hyperinsulinemia for diabetes.12 Even the genetic knockout animal is an oversimplification, since in most diseases gene products are not entirely eradicated but instead perturbed in quantity or function with compensatory adjustments in the levels of other molecules in the network. Hence, we can have an idea of what the function of a gene product is within a knockout model but to make sense of what this means in the real world, this result needs to be placed in the context of the entire network. Outputs are also often simplified by imposition of an arbitrary binary cutoff in a disease (eg, ‘‘significant’’

carotid stenosis being deemed to be >50%). Systems biology approaches aim to identify the continuum of eventualities resulting from the continuum of possible initiating factors that occur in the real world. For example, Ramsey and colleagues explored the macrophage transcriptional network mediated by Toll-like receptors (TLRs) and modeled time course expression data using whole-genome expression array analysis of bone marrow macrophages from various strains of mice with 6 different TLR agonists at multiple time points. In all, 95 combinations of strain, stimuli, and elapsed times were analyzed to give a broad picture of the dynamic transcriptional program of the TLR network.18 Such an approach would not be possible with, for example, a TLR knockout animal alone.

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

Shalhoub et al

9 between high- and low-risk individuals.19 The majority of atherosclerotic biomarkers that have been identified participate in pathways associated with the disease, namely, inflammation and cholesterol biosynthesis.20 Such biomarkers, therefore, may not add significantly to the value afforded by risk stratification on the basis of conventional clinical risk factors. Additionally, and relevant when considering creating a ‘‘score’’ based on multiple markers, a large number of correlated biomarkers is substantially less informative than a small number of uncorrelated biomarkers.21

The Omic Disciplines

Figure 4. A ‘‘top-down’’ approach to biology. Systems biology employs a top-down approach which is broad and integrative, looking at ‘‘net’’ effects in a given biological system in the context of health or disease. Systems biology, therefore, is often described as ‘‘nonhypothesis driven.’’ Conversely, traditional laboratory techniques tend to adopt a ‘‘bottom-up’’ approach which (driven by a focused hypothesis) investigates a single DNA, RNA, protein, or metabolite entity, then works back toward physiology or pathology.

This includes (most commonly) the status of DNA, transcription, translation, and metabolism, although other disciplines focus specifically on lipids (lipidomics) and carbohydrates (glycomics). The investigation in the omic disciplines may be hypothesis driven or nonhypothesis driven. The former are studies on a small to intermediate scale, and the latter a large-scale examination of the genome, transcriptome, proteome, or metabolome requiring complex (often multivariate) statistical analysis which can be ‘‘hypothesis generating.’’ There is a growing utility of these techniques in atherosclerosis research.

Systems Biology in Practice Having emphasized the ideals of systems biology, it is important to stress that the current state of the art does not match this ideology. Much of the early work within the discipline of systems biology has focused on the identification of nodes: separate genes in the genome, transcripts in the transcriptome, proteins in the proteome, metabolites in the metabolome, and so on. Progress would have been almost impossibly slow without the paradigm shift in methodologies used to identify these nodes. They are collectively termed the ‘‘omic’’ disciplines and allow qualitative, and often also semiquantitative or quantitative, understanding of the components of the biological network. Such methodologies and their findings are the focus of the remainder of this article and form a way of defining what is happening in an organism without focusing on a specific pathway—that is a ‘‘top-down’’ approach (Figure 4). Such approaches need integration with ‘‘bottom-up’’ reductionist approaches in order to fully comprehend how the edges connecting the nodes function and thus how the entire network operates both in health and in disease.

Biomarkers and Biomarker ‘‘Signatures’’ A biomarker is a substance or characteristic that is objectively measured and evaluated and is used as an indicator of a biological state. These biological states include normal biological processes, disease states, or responses to therapy. A major acknowledged limitation of circulating biomarkers is that their concentration range is often broad. Therefore, even if a difference exists in a particular disease state, the breadth of range of the biological analyte will mean there is often overlap in levels

Genomics Genome wide association studies (GWAS) are a systems biology approach for the identification of genetic risk factors for susceptibility to a particular condition. There are a number of ways that DNA can vary from individual to individual, some of which may be biologically and/or clinically relevant. Of these variations, the most commonly studied is the single nucleotide polymorphism (SNP) which is where a single nucleotide in a gene’s DNA sequence is replaced with another nucleotide, for example, -CC-G-T-A-C- and -C-C-G-A-A-C-. Genetic variations can be investigated en mass with high-throughput technology to determine the effect of genotype on a particular phenotype—the so-called next generation sequencing. Non-GWAS studies have previously linked high-carotid artery intima–media thickness (IMT, a surrogate measure of subclinical atherosclerosis and strong predictor of future ischemic stroke) with variations in genes (identified by polymerase chain reaction, PCR) related to matrix degradation (matrix metalloproteinase 3), inflammation (interleukin 6), blood pressure homeostasis (angiotensin converting enzyme [ACE]), and lipid metabolism (hepatic lipase, apolipoprotein E [APOE], cholesteryl ester transfer protein, paraoxonase 1, and methylenetetrahydrofolate reductase [MTHFR]).22 Of these, the contribution of ACE, APOE, and MTHFR appears robust in their association with subsequent stroke and have been confirmed by meta-analysis.22 In a study of 928 patients and 602 control individuals, 100 genetic polymorphisms in 47 suspected susceptibility genes for ischemic cerebrovascular disease were investigated. Genotyping was undertaken using intermediate throughput multiplex

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

10

Vascular and Endovascular Surgery 48(1)

PCR and linear immobilized probe array assays. Three polymorphisms were found to be associated with ischemic cerebrovascular disease and were located within the lipoprotein lipase, angiotensinogen, and guanine nucleotide-binding protein b-3 genes.23 Furthermore, comparing patients with carotid stenosis and those without, factor VII, APOE and 2 renin polymorphisms were differentially frequent.23 Lanktree et al tested the association between subclinical atherosclerosis, measured by IMT, and ethnicity using a targeted cardiovascular SNP microarray consisting of approximately 50 000 SNPs found within 2100 genes.24 They found the effects of genetic variants to be small. Genotypes from many SNPs were nominally associated with total cholesterol correlated with IMT. More recently, a collaborative undertaking across 92 centers worldwide sought to identify genetic variations associated with hypertension, a key risk factor in the development and progression of atherosclerosis and cardiovascular disease. A total of 34 433 European patients from 17 GWAS studies were genotyped to identify candidate polymorphisms which were then validated in 2 further patient populations, 71 225 individuals of European descent and 12 889 individuals of Indian Asian ancestry.25 Association between systolic and diastolic blood pressure was found with variants near 8 genes: CYP17A1, CYPIA2, FGF5, SH2B3, MTHFR, c10orf107, ZNF652, and PLCD3.25

Transcriptomics Transcriptomics (epigenomics, gene expression, or transcription profiling) studies RNA and identifies gene activation or transcription. As a patient’s genotype is the same regardless of which somatic cell is looked at, genomics is patient specific. On the other hand, gene transcription varies within individuals from tissue to tissue and under different circumstances, such that they will have a unique genotype but a number of transcriptomic profiles. A further challenge with transcriptomics is that, while the genome is relatively stable, and DNA is relatively robust, RNA is less so and subject to degradation. RNA quality is looked at prior to a transcriptomic study and this is achieved by calculation of the RNA integrity number (RIN) from a number of parameters obtained by electrophoresis of the RNA. The RIN may be artificially lowered in circumstances where degraded RNA exists, for example, in the mural thrombus of an aneurysm or—in the context of atherosclerosis—the necrotic core of a plaque. This is one reason why transcriptome-wide analysis of human atherosclerotic tissue has not been readily performed. Following RNA extraction from samples, double-stranded complementary DNA (cDNA) is synthesized. This is a stable copy of the RNA which can be used for gene profiling. The cDNA is labeled, hybridized, and quantified producing a gene expression pattern.26 Microarrays allow the comparison of global expression changes in thousands of genes. Results can be validated using PCR.27-31 These techniques have led to the

identification of unique subsets of genes associated with specific diseases and disease processes.29 There have been several studies validating the technique of gene expression profiling in normal and atherosclerotic tissue.27,28,30-33 Seo et al generated gene expression data from human aortas with varying degrees of atherosclerosis to create genomic phenotypes which may allow prediction of disease states from any given sample of aorta.34 Microarray analysis has also demonstrated that 97% of the genes are unaffected when atherosclerotic samples were taken during surgery compared to postmortem samples. Differentially expressed genes were shown to be involved in basal cell metabolism and hypoxia at messenger RNA (mRNA) level, but these changes were not reflected in protein expression.35 Further to the assessment of extent of atherosclerosis (ie, atherosclerotic ‘‘burden’’) is examination of the transcriptional profile of plaque instability. Papaspyridonos et al examined total RNA from stable and unstable areas from 3 human carotid plaques, identifying the differential expression of 170 genes between these areas of which 4 varied by greater than 5-fold: retinoic acid receptor responder, junctional adhesion molecule, hemoglobin scavenger receptor CD163, and ganglioside activator protein GM2A.36 One way to circumvent the challenges inherent to transcriptomics of human atherosclerosis tissue is by profiling peripheral blood. This has the further utility of direct biomarker discovery. An example of this was a study that used arrays to expression profile peripheral blood from 10 patients with carotid stenosis and 10 matched controls. Of the 14 000 transcripts represented on the arrays, 82 genes were found to be differentially expressed between these 2 groups.37 Of the 82 differentially expressed genes,14 were selected as candidate genes and subsequently confirmed by PCR on separate validation groups of 40 patients with carotid stenosis and 40 controls. These genes were largely involved in immune activities and oxygen transport.37 Table 1 summarizes studies pertaining to transcriptomics in human atherosclerosis.

Proteomics The proteome is defined as all proteins present in a cell, tissue, or organism.39 Proteomics, the analysis of a proteome, can detect proteins that are associated with a disease by measuring levels in control and disease states.40 Protein identification by mass spectrometry (MS) allows for the analysis of hundreds to thousands of proteins at a time, including posttranslational modifications.41 Proteomic studies have been carried out on both human plaque tissue and on plasma samples from individuals with atherosclerosis41 (Table 2). Recent reviews have examined the proteomics of atherosclerosis,48,49 including 2 relevant studies where proteomic analysis was undertaken on patients with carotid atherosclerosis. Martin-Ventura et al looked at 25 carotid endarterectomy specimens, comparing these to 36 control endarteries.50 The end artery segments were used to condition protein-free medium, which was subsequently separated by 2-dimensional

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

11

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

Carotid stenosis versus healthy individuals Peripheral artery disease as classified by the American Heart Association

Plaque instability

Advanced atherosclerotic plaques. Surgery versus autopsy samples

3 cohorts of 5 mice with varying disease susceptibility and diet compared to human data sets to identify shared genes

Human aorta samples with varying degrees of atherosclerosis

Comparison of coronary plaques in stable and unstable angina

Differential gene expression in human coronary artery disease. Atherosclerosis versus normal

Human arterial samples from varying locations from aorta to lower limb

Comparison of human stable and ruptured plaques. American Heart Association classification

Rossi et al37

Papaspyridonos et al36

Sluimer et al35

Tabibiazar et al32

Seo et al34

Randi et al28

Archacki et al27

Hiltunen et al30

Faber et al38

20 Human arterial plaques from varying locations (10 stable, 10 ruptured)

12 arterial samples obtained after amputation of limb or from organ donors

9 severe and 6 nonatherosclerotic human coronary arteries from explanted hearts

26 carotid segments taken at autopsy from 11 donors. 11 samples from live donors Entire aortas from euthanized mice. Once data sets were identified, they were compared with those of 40 coronary artery samples from 17 transplant patients Cadaveric whole aortas. Severity assessed as mild proximally with increasing severity distally 8 coronary plaques from live patients

40 human carotid stenosis, 40 healthy controls 30 whole femoral arteries; 14 advanced, 11 intermediate, 5 normal from live individuals 3 human carotid plaques

Sample Source

3000 clones upregulated and 2000 downregulated in ruptured plaques

75 differentially expressed genes previously unknown to be associated with atherosclerosis

Established gene profiling as reliable tool for analysis of atherosclerosis. 56 genes differentially expressed

Identified gene expression ‘‘signatures’’; a set of 208 genes, relating to disease severity Examination of genes using data sets for inflammation, adhesion and hemostasis

Established ‘‘classifier gene set’’ which can distinguish specimens with atherosclerotic disease state

97.2% genes have similar expression levels in advanced lesions

170 genes differentially expressed

366 genes differentially regulated in intermediate and 447 in advanced

82 genes differentially expressed

Results

Multiple, including apoE, osteopontin, OLR1. Cell cycle regulation and inflammatory response Lymphocytes adhesion molecule (MadCAM) in all plaques. Anticoagulant protein S and TF increased, and COX-1, IL-7, MCP-1 and MCP-2 decreased in unstable plaques Altered gene expression of inflammation, cell necrosis/apoptosis, altered cell migration, matrix degradation genes Subgroup for cell signaling and proliferation selected for further analysis; receptors for signaling in activated macrophages, angiogenic and vasculoprotective SSH6 and Perilipin (SSH1 and 11) expressed in ruptured plaques in the cytoplasm of cells surrounding cholesterol clefts and in foam cells

Immune/inflammatory genes enriched especially Toll-like receptor signaling, NK-mediated cytotoxicity Retinoic acid receptor responder, junctional adhesion molecule, hemoglobin scavenger receptor CD163, ganglioside activator protein GM2A Differentially expressed genes involved in basal cell metabolism and hypoxia driven pathways Gene subset was found to be equivalent in humans when grading human coronary disease

Immune activities and oxygen transport

Genes of Interest

Abbreviations: ApoE, apolipoprotein E; COX, cyclooxygenase, IL, interleukin; MCP, monocyte chemotactic protein; OLR, oxidized low-density lipoprotein receptor; SSH, suppression subtractive hybridization; TF, tissue factor; NK, natural killer.

Fu et al31

Methodology

Study Reference

Table 1. Summary of Transcriptomic Studies in Human Atherosclerosis.

12

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

Combination of proteomics, metabolomics, and immunomics of human atherosclerotic plaquederived microparticles

Proteomics, metabolomics and immunomics on human atherosclerotic plaques Proteomics of human coronary atherosclerotic plaques

Proteomics of coronary atherosclerosis

NMR mass spectrometry of serum

Mayr et al46

Martinez-Pinna et al41

You et al40

Brindle et al47

36 patients with 3 vessel disease on coronary angiography and 30 individuals with angiographically normal coronary arteries

10 diseased and 7 normal coronary arteries from explanted hearts

35 archival coronary vessels from 28 explanted hearts. Artery and plaque analyzed separately

Human carotid atherosclerotic plaques from live patients

19 histologically stable and 29 histologically unstable (thin or fissured fibrous cap, foamy or necrotic core, with hemorrhage, ulceration, or thrombosis) plaques 26 carotid atherosclerotic plaques

Plasma from 16 patients with stable atherosclerosis and 28 normal individuals

Discovery cohort of 20 patients with 3 vessel coronary disease and 20 with normal vessels on coronary angiography, followed by 2 larger validation cohorts of 67 and 197 patients with varying levels of coronary artery disease and a fourth cohort of 20 individuals with carotid atherosclerosis

Plasma from 35 patients with carotid stenosis, carotid plaques from 6 symptomatic and 6 asymptomatic patients, 3 radial artery controls and a femoral plaque

Sample Source

Application of supervised multivariate statistics allowed >90% of patients with 3 vessel coronary disease to be distinguished from patients with angiographically normal coronary arteries, with a specificity of >90%

Increased expression of ferritin light chain

Immunoglobulins are present within microparticles derived from plaque macrophages. Distinct expression of 806 proteins. Selected 4 subgroups for further analysis; extracellular, lipid binding, inflammatory, and apoptotic-cell

Extracted proteins were separated by 2DE revealing 57 distinct spots representing 33 different proteins which were subjected to MALDI-TOF mass spectrometry Immunoglobulins present within the microparticles were derived from plaque macrophages. Plaque antibodies were different to circulating antibodies in the plasma as they recognize carbohydrate antigens including blood group antigen A

Development of atherosclerosis directly perturbed fatty acid metabolism

The level of a 14.7 kDa protein was elevated; identified as lysozyme. Arterial plasma lysozyme levels in 197 patients with varying degrees of coronary artery disease, using a cutoff value of 1.5 mg/mL, was able to distinguish patients with 1 or more occluded coronary arteries, with 86% sensitivity and 93% specificity

150 lipid species from 9 different classes of which 24 were detected in endarterectomies only

Results

May contribute to CAD by modulating oxidation of lipids within the vessel wall through generation of reactive oxygen species Accurate and noninvasive separation of individuals 3 vessel coronary disease from controls

Anti-A antibodies (IgG, IgA, IgE), anticarbohydrate antibodies, and antibodies to oxidized LDL detected in plaques Periostin (cell migration), PEDF (expressed when cells are exposed to oxidized LDL), MFG-E8 (phagocytosis ligand), annexin-1 (expressed by macrophages of foam cell phenotypes, linked to apoptosis and phagocytosis)

Immunoglobulins are present within microparticles derived from plaque macrophages, that the portfolio of plaque antibodies is different from circulating antibodies in plasma, and that anticarbohydrate antibodies are retained in human atherosclerotic lesions

Unstable plaques showed reduced abundance of the protective SOD-3, GST, HSP27, HSP20, annexin A10, and r-GDI

Palmitate was confirmed as a phenotypic biomarker for the clinical diagnosis of atherosclerosis

Raised plasma lysozyme levels may be a useful biomarker of atherosclerotic cardiovascular disease and response to therapy. Additional studies to investigate this are warranted

Polyunsaturated cholesteryl esters with longchain fatty acids and certain sphingomyelin species showed the greatest relative enrichment in plaques compared to plasma and formed part of a lipid signature for vulnerable and stable plaque areas in a systems-wide network analysis

Key Findings

Abbreviations: GDI, guanosine 50 -diphosphate dissociation inhibitor; GST, glutathione S-transferase; HSP, heat shock protein; Ig, immunoglobulin; LDL, low-density lipoprotein; MALDI-TOF, matrix-assisted laser desorption/ionization-time of flight; MFG-E8, milk fat globule epidermal growth factor 8; NMR, nuclear magnetic resonance; PEDF, pigment epithelium-derived factor; SELDI-TOF, surface-enhanced laser desorption/ionization-time-of-flight; SOD, superoxide dismutase; 2DE, 2-dimensional electrophoresis.

Bagnato et al39

Proteomic analysis of human carotid atherosclerotic plaques

Lepedda et al45

SELDI-TOF mass spectrometry of plasma

Abdul-Salam et al43

Plasma metabolomics for biomarkers of atherosclerosis

Lipidomic analysis of lipids in tissue sections and extracts and plasma

Stegemann et al42

Chen et al44

Methodology

Study Reference

Table 2. Summary of Proteomic, Lipidomic, and Metabolite Profiling Studies in Human Atherosclerosis.

Shalhoub et al

13

electrophoresis (2DE) and subjected to matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) MS. The main finding of this study was that heat shock protein 27 secretion was significantly less in carotid atherosclerosis and was barely detectable in the secretome of complicated plaques (determined histologically by the presence of rupture, intraplaque hemorrhage, and a high proportion of inflammatory cells). You et al took a proteomic approach to coronary atherosclerosis in 10 diseased and 7 normal individuals.40 The 2DE plates were directly visualized to compare diseased and normal specimens. Protein sequencing using MS was performed on a protein spot that was shown to have higher levels of expression in the diseased sample. Proteomics identified that ferritin light chain was increased by 1.9-fold in coronary artery disease. The mRNA expression level of ferritin light chain was investigated using PCR and was increased to a lesser extent than protein, highlighting the discrepancy often seen between mRNA and protein differences of the same analyte. Proteomic and lipidomic studies in human atherosclerosis are summarized in Table 2. Early proteomic studies in atherosclerosis failed to identify inflammatory cytokines that are involved in plaque inflammation raising the possibility that the extraction technique was poor at identifying components that existed at low levels.39 Extraction of protein in samples and enriching for lowabundant analytes (depleting potentially ‘‘masking’’ high abundant entities such as albumin and hemoglobin) has been previously undertaken effectively in proteomic analysis of aortic aneurysm tissue.51 This approach is potentially pertinent to atherosclerosis research, where the level of complexity is high, as individual components can be analyzed separately and low-level components can be unmasked.

Metabolic Profiling Metabolic profiling (metabonomics and metabolomics) aims to measure the global dynamic metabolic response of living systems to biological stimuli or genetic modification.52,53 Metabolic profiling seeks an analytical description of complex biological samples and aims to characterize and quantify (by relative intensity or absolute quantification) small molecules in a sample. The metabolic alterations due to biological interventions or disease processes may be modeled, as the metabolic profile of an organism is analyzed and compared to the normal/ physiological conditions.53 A metabolic profiling assessment in the context of atherosclerosis is possible as is the analysis of subgroups thereof (eg, plaque vs normal tissue, or stable vs unstable atherosclerosis). Systems biology has a disadvantage as gene expression, protein expression, and metabolism operate on different timescales, making it difficult to find causal linkages. However, compared to other omic disciples, metabolic profiling is able to deliver the most ‘‘up-to-date’’ view of a system. The detected molecules are the final products of biological processes, thus a more accurate view of the system’s phenotype is generated. In addition, lifestyle and environmental factors are not incorporated within genomic data, making disease

phenotypes hard to interpret. On the contrary, metabonomics monitors the global outcome of all influencing factors,52 while individual metabolites can be compared between diseased and nondiseased states to produce diagnostic biomarkers.52 In addition to serving as disease biomarkers, metabolites may have previously unanticipated roles such as regulatory signals with hormone-like functions.41 Tools for Metabolic Profiling Applications. In metabolic profiling, samples are analyzed by either nuclear magnetic resonance (NMR) spectroscopy, usually as proton-NMR (detecting hydrogen atoms in molecules) or MS, which is generally applied after separation of the metabolic components using gas chromatography54 or liquid chromatography,55 and less frequently capillary electrophoresis.56 An advantage of NMR is its nondestructive and closed-tube format, while methods such as magic-angle spinning NMR can provide spectra of intact tissues.57,58 However, its main drawback is the reduced sensitivity when compared to MS techniques which, by employing highly efficient multichannel plates, orthogonal time-of-flight tubes, and improved analog-to-digital converters, has resulted in greater sensitivity and resolution and faster scanning rates. There are 2 contemporaneous reviews on the subject of metabolic profiling in atherosclerosis.41,59 It is clear from these that, to date, there has been limited application of these techniques in the field. The Choice of Substrate for Metabolic Profiling. The systematic collection and storage of tissue and biofluids are necessary to support these disciplines. This can be achieved through the establishment of collaborative biobanks or through linking systems biology programs to the infrastructure of clinical studies or trials.2 Such collaborations are often broad and extend across institutions, industries, and countries. Untargeted screening of biological samples, such as biofluids, cells, or tissues is usually performed as a first stage toward obtaining biomarker or metabolic pathway information relating to disease. This is performed in order to obtain a global overview on the low-molecular weight compounds (typically those less than 1 kDa) due to changes in genetic variation, disease, environmental, or nutritional status. It has been shown that metabolic profiling of biofluids that closely interact with the tissue of interest delivers clinically and biologically significant results.60 Urine is a particularly attractive substrate for biomarker discovery; collected in a noninvasive fashion and in ample quantity. Conversely, urine can be affected substantially by diet and fluid intake. However, in order to elucidate mechanisms of the pathophysiology of the disease, it is important to be able to relate structural and functional changes at local tissue levels to an organism’s response at a systems level. With these approaches, the significant parameter of spatial distribution of the biomolecules through the tissue, which may include different cell types, is disregarded. Thus, new screening techniques for untargeted determination of biomolecules in a manner that incorporates the spatial distribution have been developed, such as imaging

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

14

Vascular and Endovascular Surgery 48(1)

MS.61-64 The spatial distribution of biomolecules is known to be intrinsically related to their functions65,66 and so revealing metabolite or protein distribution patterns in tissues can provide an understanding of the physiological functioning of specific biomolecules and enhance our knowledge of disease states.

Addressing Heterogeneity Within Biological Specimens— MS Imaging Applications of imaging techniques that can deliver biological profiles related to organ or tissue topography have recently been demonstrated.61 The MS imaging (MSI) techniques offer a powerful tool for in situ protein and metabolite screening. Of the several techniques proposed for MSI, MALDI-MSI appears to be a dominant presence in recent years. Combining the chemical specificity and parallel detection of MS with microscopic imaging capabilities, MALDI-MSI offers a sensitive alternative to chromatography-based approaches for protein, peptide, and small molecule profiling.62 When combined with multivariate statistics, MALDI-MSI can lead to improved biomarker discovery and disease diagnosis67; images can be obtained simultaneously from all detected analytes in a sample at both the tissue and single-cell levels, and this information can subsequently be correlated with physical features of the tissue.68 Grey and colleagues have shown that MSI can be applied to chick heart tissue sections acquired from both fixed and paraffin-embedded samples to produce molecular images that can be related to high-quality histological tissue sections.69 Matrix-assisted laser desorption/ionization-mass spectrometry is also growing in use for lipid profiling.62 Recent significant improvements in ionization sensitivity, sample preparation, spatial resolution, and data acquisition speed for MALDIMSI of biological samples have made this approach viable for larger scale studies.62 A relatively recent technique for MSI, namely desorption electrospray ionization MS (DESI-MS),70 also comes with advantages making it a competitor to MALDI-MS. The DESI is applied under ambient conditions, where charged nebulized solvents are sprayed at a surface (sample) causing the ejection of small secondary droplets containing dissolved, ionized analytes. Unlike MALDI, there is no necessity for the ionization of assisting material. Its only disadvantage is the reduced spatial resolution compared to MALDI. Manicke et al imaged lipids in human atheroma by DESI-MS, identifying abundant cholesterol esters.71 Adding a charged labeling reagent to the solvent enabled imaging and subsequent mapping of the distribution of the lipid-rich regions within the samples. Imaging revealed areas of lipoprotein accumulation but the lipid composition was not constant within these regions raising the question of whether these may influence plaque stability.

Application of Multivariate Statistics to Omic Techniques Data acquired from omic technologies are highly complex and require robust statistical methods for handling and analysis.72

These data tend to be more short and wide (ie, containing a large number of variables relative to sample number) rather than long and lean (ie, high sample number with limited variables). Common multivariate data analysis (MVDA) methods can be divided into unsupervised and supervised approaches.73 Unsupervised methods, where the algorithm is not given information about different groups or classes (ie, input data without identifying samples), include principal components analysis (PCA)74 and hierarchical cluster analysis (HCA).75 The central idea behind PCA is to reduce the dimensionality of a data set consisting of a large number of interrelated variables, while retaining as much of the variation present in the data set as possible.76 This is achieved through the transformation of data into new sets of variables which are uncorrelated—the principal components (PCs). In other words, the first PC would be the vector that explains the greatest variation in the data. The second PC would be the vector explaining the second greatest variation, but always orthogonal to the first, thus uncorrelated. In the same way, more PCs can be calculated.72 Data are often described in relation to two or three PCs. The PCA is useful for data dimensionality reduction, investigation of clustering tendencies, outlier detection, and data structure visualization. However, PCA gives a simplified representation of the information contained in the data matrix and generally cannot make use of additional metadata. It can however provide discriminating features between groups, thus potential biomarkers, with reduced risk of statistical artifacts. The HCA organizes information about variables in a data matrix, forming ‘‘clusters,’’ where the degree of association is strong between samples within the same cluster and weak between those in different clusters. The HCA may reveal previously hidden associations and structures in data and can be represented as a tree, or dendrogram, where each step in the clustering process is illustrated by a branch of the tree. Hierarchical cluster analysis is a commonly used statistical methodology in transcriptomics and can be applied in metabonomics to provide unsupervised clustering of groups. Commonly used supervised methods include discriminant analysis (DA), for example, projection to latent structures (PLSs)77 and orthogonal PLSs (O-PLSs).78 Contrasting with unsupervised methods, here a classification system is given group information, that is data (commonly referred to as the X matrix) together with sample information commonly referred to as the Y matrix, used to build a model and estimate necessary parameters. In metabolic profiling studies, PLS-DA or O-PLS-DA would often follow unsupervised methods such as PCA, where PLS-DA is performed in order to enhance the separation between groups of observations, often by rotating PCA components to achieve maximum separation between classes and to understand which variables are responsible for separating the classes. The O-PLS-DA is similar to PLS-DA but has improved predictive capability as uncorrelated variables can be filtered out via the orthogonal component. Supervised methods are generally followed by validation testing, for example, permutation testing. The rationale behind this is to deliberately mislabel the samples into a different group. A

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

Shalhoub et al

15

valid model should have its predictive value (also referred to as Q2Y) lower than the real model (preferably for all recalculations), as this mislabeling is randomly permuted a number of times. Other statistical tests employed include cross-validated analysis of variance79 and receiver–operating characteristic (ROC) curve analysis.80 When undertaking MVDA, there are some relevant additional general points one may wish to consider. The importance of detailed clinical characterization—including demographic, comorbidities, and pharmacotherapeutic parameters—for incorporation into multivariate statistical models cannot be overstated. Cases can be used as their own controls where this is possible in the context of the study design and ethical constraints. It needs to be remembered that statistical significance may not always be reflected in a biological or clinical significance. Bioinformatics can be used for ontology, network, and pathway analyses, allowing data to progress from a series of analytes to categorize these into biological themes, processes, networks, and pathways.

Integration of Findings From Different Omic Techniques and at Different Organizational Levels Integration of metabonomic data with that from other techniques in molecular biology, such as transcriptomics and proteomics, is feasible.55 It is also important to be able to integrate findings at different organizational levels with each other in order to assess the network context of such differences. Here the fact that there is information flow from DNA to RNA to protein to metabolite is important since it means that DNA can serve as a ‘‘causal anchor.’’16 As an example, it is often found in a system that there are perturbations at multiple levels of the network. Hence there may be genetic loci that are associated with changes in the quantity of a protein and are also associated with alterations in the quantity of a metabolite present. This suggests 3 possibilities if DNA is viewed as a causal anchor: 1. 2. 3.

The loci control protein expression which in turn effects metabolite production, the loci effect metabolite levels directly which secondarily perturb protein expression, or the loci independently control levels of protein and metabolite.

Which of these is the case can be predicted by causal modeling via mathematical conditioning on transcript levels using partial correlation coefficients. A detailed description is beyond the scope of this article, but the approach is exemplified by Schadt and co-workers81 who developed a statistical procedure to distinguish where variation is most important and thus where the causal root of complex traits lie. The application of this procedure successfully identified a number of genes involved in obesity.82

Conclusion Use of the -omic disciplines, to date, has resulted in advancement of our understanding of the biology of human cardiovascular disease. In the future, it will be increasingly important to collaborate and collate data from multiple studies in order to improve data sets and to make them freely available,33 as well as looking across genomic, transcriptomic, proteomic and metabolic profiling, and undertaking analyses that consider and integrate these. The ultimate goal is the development of insight into potential preventative, diagnostic, and therapeutic strategies in human atherosclerosis.32 Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: MBS is supported by the Wellcome Trust –Grant number 092852. PAV would like to thank the Royal Society of Chemistry for funding his PhD studentship. EJW would like to acknowledge Waters Corporation for funding.

References 1. Narula J, Strauss HW. The popcorn plaques. Nat Med. 2007; 13(5):532-534. 2. Shalhoub J, Davies KJ, Hasan N, Thapar A, Sharma P, Davies AH. The utility of collaborative biobanks for cardiovascular research. Angiology. 2012;63(5):367-377. 3. Ross R. Atherosclerosis–an inflammatory disease. N Engl J Med. 1999;340(2):115-126. 4. Anderson PW. More is different. Science. 1972;177(4047): 393-396. 5. Prigogine I, Stengers I. La nouvelle alliance. Paris: Editions Gallimard; 1979. 6. Ferrell JE Jr. Q&A: Systems biology. J Biol. 2009;8(1):2. 7. Kirschner MW. The meaning of systems biology. Cell. 2005; 121(4):503-504. 8. Varki A, Altheide TK. Comparing the human and chimpanzee genomes: Searching for needles in a haystack. Genome Res. 2005;15(12):1746-1758. 9. Lander AD. The edges of understanding. BMC Biol. 2010;8:40. 10. Hartwell LH, Hopfield JJ, Leibler S, Murray AW. From molecular to modular cell biology. Nature. 1999;402(6761 suppl):C47-C52. 11. Soutar AK, Naoumova RP. Mechanisms of disease: Genetic causes of familial hypercholesterolemia. Nat Clin Pract Cardiovasc Med. 2007;4(4):214-225. 12. De Keulenaer GW, Brutsaert DL. Systolic and diastolic heart failure are overlapping phenotypes within the heart failure spectrum. Circulation. 2011;123(18):1996-2004. 13. Frayling TM, Timpson NJ, Weedon MN, et al. A common variant in the fto gene is associated with body mass index and predisposes to childhood and adult obesity. Science. 2007; 316(5826):889-894.

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

16

Vascular and Endovascular Surgery 48(1)

14. Barabasi AL. Network medicine–from obesity to the ‘‘diseasome’’. N Engl J Med. 2007;357(4):404-407. 15. Christakis NA, Fowler JH. The spread of obesity in a large social network over 32 years. N Engl J Med. 2007;357(4):370-379. 16. Lusis AJ, Weiss JN. Cardiovascular networks: Systems-based approaches to cardiovascular disease. Circulation. 2010;121(1): 157-170. 17. Strogatz SH. Exploring complex networks. Nature. 2001; 410(6825):268-276. 18. Ramsey SA, Klemm SL, Zak DE, et al. Uncovering a macrophage transcriptional program by integrating evidence from motif scanning and expression dynamics. PLoS Comput Biol. 2008;4(3): e1000021. 19. Ware JH. The limitations of risk factors as prognostic tools. N Engl J Med. 2006;355(25):2615-2617. 20. Gerszten RE, Wang TJ. The search for new cardiovascular biomarkers. Nature. 2008;451(7181):949-952. 21. Pepe MS, Thompson ML. Combining diagnostic test results to increase accuracy. Biostatistics. 2000;1(2):123-140. 22. Humphries SE, Morgan L. Genetic risk factors for stroke and carotid atherosclerosis: insights into pathophysiology from candidate gene approaches. Lancet Neurol. 2004;3(4):227-235. 23. Kostulas K, Brophy VH, Moraitis K, et al. Genetic profile of ischemic cerebrovascular disease and carotid stenosis. Acta Neurol Scand. 2008;118(3):146-152. 24. Lanktree MB, Hegele RA, Yusuf S, Anand SS. Multi-ethnic genetic association study of carotid intima-media thickness using a targeted cardiovascular snp microarray. Stroke. 2009;40(10): 3173-3179. 25. Newton-Cheh C, Johnson T, Gateva V, et al. Genome-wide association study identifies eight loci associated with blood pressure. Nat Genet. 2009;41(6):666-676. 26. Archacki SR, Wang QK. Microarray analysis of cardiovascular diseases. Methods Mol Med. 2006;129:1-13. 27. Archacki SR, Angheloiu G, Tian XL, et al. Identification of new genes differentially expressed in coronary artery disease by expression profiling. Physiol Genomics. 2003;15(1):65-74. 28. Randi AM, Biguzzi E, Falciani F, et al. Identification of differentially expressed genes in coronary atherosclerotic plaques from patients with stable or unstable angina by cdna array analysis. J Thromb Haemost. 2003;1(4):829-835. 29. Archacki S, Wang Q. Expression profiling of cardiovascular disease. Hum Genomics. 2004;1(5):355-370. 30. Hiltunen MO, Tuomisto TT, Niemi M, et al. Changes in gene expression in atherosclerotic plaques analyzed using DNA array. Atherosclerosis. 2002;165(1):23-32. 31. Fu S, Zhao H, Shi J, et al. Peripheral arterial occlusive disease: global gene expression analyses suggest a major role for immune and inflammatory responses. BMC Genomics. 2008;9:369. 32. Tabibiazar R, Wagner RA, Ashley EA, et al. Signature patterns of gene expression in mouse atherosclerosis and their correlation to human coronary disease. Physiol Genomics. 2005;22(2):213-226. 33. Bijnens AP, Lutgens E, Ayoubi T, Kuiper J, Horrevoets AJ, Daemen MJ. Genome-wide expression studies of atherosclerosis: critical issues in methodology, analysis, interpretation of transcriptomics data. Arterioscler Thromb Vasc Biol. 2006;26(6):1226-1235.

34. Seo D, Wang T, Dressman H, et al. Gene expression phenotypes of atherosclerosis. Arterioscler Thromb Vasc Biol. 2004;24(10): 1922-1927. 35. Sluimer JC, Kisters N, Cleutjens KB, et al. Dead or alive: gene expression profiles of advanced atherosclerotic plaques from autopsy and surgery. Physiol Genomics. 2007;30(3):335-341. 36. Papaspyridonos M, Smith A, Burnand KG, et al. Novel candidate genes in unstable areas of human atherosclerotic plaques. Arterioscler Thromb Vasc Biol. 2006;26(8):1837-1844. 37. Rossi L, Lapini I, Magi A, et al. Carotid artery disease: Novel pathophysiological mechanisms identified by gene-expression profiling of peripheral blood. Eur J Vasc Endovasc Surg. 2010; 40(5):549-558. 38. Faber BC, Cleutjens KB, Niessen RL, et al. Identification of genes potentially involved in rupture of human atherosclerotic plaques. Circ Res. 2001;89(6):547-554. 39. Bagnato C, Thumar J, Mayya V, et al. Proteomics analysis of human coronary atherosclerotic plaque: a feasibility study of direct tissue proteomics by liquid chromatography and tandem mass spectrometry. Mol Cell Proteomics. 2007;6(6):1088-1102. 40. You SA, Archacki SR, Angheloiu G, et al. Proteomic approach to coronary atherosclerosis shows ferritin light chain as a significant marker: evidence consistent with iron hypothesis in atherosclerosis. Physiol Genomics. 2003;13(1):25-30. 41. Martinez-Pinna R, Barbas C, Blanco-Colio LM, et al. Proteomic and metabolomic profiles in atherothrombotic vascular disease. Curr Atheroscler Rep. 2010;12(3):202-208. 42. Stegemann C, Drozdov I, Shalhoub J, et al. Comparative lipidomics profiling of human atherosclerotic plaques. Circ Cardiovasc Genet. 2011;4(3):232-242. 43. Abdul-Salam VB, Ramrakha P, Krishnan U, et al. Identification and assessment of plasma lysozyme as a putative biomarker of atherosclerosis. Arterioscler Thromb Vasc Biol. 2010;30(5): 1027-1033. 44. Chen X, Liu L, Palacios G, et al. Plasma metabolomics reveals biomarkers of the atherosclerosis. J Sep Sci. 2010;33(17-18): 2776-2783. 45. Lepedda AJ, Cigliano A, Cherchi GM, et al. A proteomic approach to differentiate histologically classified stable and unstable plaques from human carotid arteries. Atherosclerosis. 2009;203(1):112-118. 46. Mayr M, Grainger D, Mayr U, et al. Proteomics, metabolomics, and immunomics on microparticles derived from human atherosclerotic plaques. Circ Cardiovasc Genet. 2009;2(4): 379-388. 47. Brindle JT, Antti H, Holmes E, et al. Rapid and noninvasive diagnosis of the presence and severity of coronary heart disease using 1h-nmr-based metabonomics. Nat Med. 2002;8(12): 1439-1444. 48. Didangelos A, Simper D, Monaco C, Mayr M. Proteomics of acute coronary syndromes. Curr Atheroscler Rep. 2009;11(3):188-195. 49. Didangelos A, Stegemann C, Mayr M. The -omics era: proteomics and lipidomics in vascular research. Atherosclerosis. 2011; 221(1):12-17. 50. Martin-Ventura JL, Duran MC, Blanco-Colio LM, et al. Identification by a differential proteomic approach of heat shock protein

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

Shalhoub et al

51.

52. 53.

54.

55.

56.

57.

58.

59. 60. 61.

62. 63.

64.

65.

17

27 as a potential marker of atherosclerosis. Circulation. 2004; 110(15):2216-2219. Didangelos A, Yin X, Mandal K, Baumert M, Jahangiri M, Mayr M. Proteomics characterization of extracellular space components in the human aorta. Mol Cell Proteomics. 2010;9(9):2048-2062. Nicholson JK, Lindon JC. Systems biology: metabonomics. Nature. 2008;455(7216):1054-1056. Nicholson JK, Lindon JC, Holmes E. ‘Metabonomics’: understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological nmr spectroscopic data. Xenobiotica. 1999;29(11):1181-1189. Chan EC, Pasikanti KK, Nicholson JK. Global urinary metabolic profiling procedures using gas chromatography-mass spectrometry. Nat Protoc. 2011;6(10):1483-1499. Lindon JC, Holmes E, Nicholson JK. Metabonomics techniques and applications to pharmaceutical research & development. Pharm Res. 2006;23(6):1075-1088. Hirai MY, Klein M, Fujikawa Y, et al. Elucidation of gene-togene and metabolite-to-gene networks in arabidopsis by integration of metabolomics and transcriptomics. J Biol Chem. 2005; 280(27):25590-25595. Beckonert O, Keun HC, Ebbels TM, et al. Metabolic profiling, metabolomic and metabonomic procedures for nmr spectroscopy of urine, plasma, serum and tissue extracts. Nat Protoc. 2007; 2(11):2692-2703. Beckonert O, Coen M, Keun HC, et al. High-resolution magicangle-spinning nmr spectroscopy for metabolic profiling of intact tissues. Nat Protoc. 2010;5(6):1019-1032. Goonewardena SN, Prevette LE, Desai AA. Metabolomics and atherosclerosis. Curr Atheroscler Rep. 2010;12(4):267-272. Schmidt C. Metabolomics takes its place as latest up-and-coming ‘‘omic’’ science. J Natl Cancer Inst. 2004;96(10):732-734. Bunch J, Clench MR, Richards DS. Determination of pharmaceutical compounds in skin by imaging matrix-assisted laser desorption/ionisation mass spectrometry. Rapid Commun Mass Spectrom. 2004;18(24):3051-3060. McDonnell LA, Heeren RM. Imaging mass spectrometry. Mass Spectrom Rev. 2007;26(4):606-643. Trim PJ, Djidja MC, Atkinson SJ, et al. Introduction of a 20 khz nd: Yvo4 laser into a hybrid quadrupole time-of-flight mass spectrometer for maldi-ms imaging. Anal Bioanal Chem. 2010; 397(8):3409-3419. Trim PJ, Henson CM, Avery JL, et al. Matrix-assisted laser desorption/ionization-ion mobility separation-mass spectrometry imaging of vinblastine in whole body tissue sections. Anal Chem. 2008;80(22):8628-8634. Simpson JC, Wellenreuther R, Poustka A, Pepperkok R, Wiemann S. Systematic subcellular localization of novel proteins identified by large-scale cdna sequencing. EMBO Rep. 2000;1(3): 287-292.

66. Kumar A, Agarwal S, Heyman JA, et al. Subcellular localization of the yeast proteome. Genes Dev. 2002;16(6):707-719. 67. Djidja MC, Claude E, Snel MF, et al. Novel molecular tumour classification using maldi-mass spectrometry imaging of tissue micro-array. Anal Bioanal Chem. 2010;397(2):587-601. 68. Stoeckli M, Chaurand P, Hallahan DE, Caprioli RM. Imaging mass spectrometry: a new technology for the analysis of protein expression in mammalian tissues. Nat Med. 2001;7(4): 493-496. 69. Grey AC, Gelasco AK, Section J, Moreno-Rodriguez RA, Krug EL, Schey KL. Molecular morphology of the chick heart visualized by maldi imaging mass spectrometry. Anat Rec (Hoboken). 2010;293(5):821-828. 70. Takats Z, Wiseman JM, Gologan B, Cooks RG. Mass spectrometry sampling under ambient conditions with desorption electrospray ionization. Science. 2004;306(5695):471-473. 71. Manicke NE, Nefliu M, Wu C, et al. Imaging of lipids in atheroma by desorption electrospray ionization mass spectrometry. Anal Chem. 2009;81(21):8702-8707. 72. Trygg J, Holmes E, Lundstedt T. Chemometrics in metabonomics. J Proteome Res. 2007;6(2):469-479. 73. van der Greef J, Smilde AK. Symbiosis of chemometrics and metabolomics: past, present, and future. J Chemometrics. 2005; 19(5-7):376-386. 74. Yeung KY, Ruzzo WL. Principal component analysis for clustering gene expression data. Bioinformatics. 2001;17(9):763-774. 75. Scholz M, Selbig J. Visualization and analysis of molecular data. Methods Mol Biol. 2007;358:87-104. 76. Jolliffe IT. Principal component analysis. New York: Springer; 2002. 77. Lutz U, Lutz RW, Lutz WK. Metabolic profiling of glucuronides in human urine by lc-ms/ms and partial least-squares discriminant analysis for classification and prediction of gender. Anal Chem. 2006;78(13):4564-4571. 78. Bylesjo M, Rantalainen M, Cloarec O, Nicholson JK, Holmes E, Trygg J.Opls discriminant analysis: Combining the strengths of pls-da and simca classification. J Chemometrics. 2006;20(8-10): 341-351. 79. Eriksson L. Cv–anova for significance testing of pls and opls1 models. J Chemometrics. 2008;22(11-12):594-600. 80. Blaise BJ, Shintu L, Elena Bnd, Emsley L, Dumas M-E, Toulhoat P. Statistical recoupling prior to significance testing in nuclear magnetic resonance based metabonomics. Anal Chem. 2009; 81(15):6242-6251. 81. Schadt EE, Lamb J, Yang X, et al. An integrative genomics approach to infer causal associations between gene expression and disease. Nat Genet. 2005;37(7):710-717. 82. Yang X, Deignan JL, Qi H, et al. Validation of candidate causal genes for obesity that affect shared metabolic pathways and networks. Nat Genet. 2009;41(4):415-423.

Downloaded from ves.sagepub.com at GEORGE MASON UNIV on June 16, 2014

Systems biology of human atherosclerosis.

Systems biology describes a holistic and integrative approach to understand physiology and pathology. The "omic" disciplines include genomics, transcr...
470KB Sizes 0 Downloads 0 Views