EXPLORING THE COMPLEX BIOLOGY OF THE CAROTID ATHEROSCLEROTIC PLAQUE
M.A. Siemelink
20162300 Marten Siemelink.indd 1 15-03-16 13:22 EXPLORING THE COMPLEX BIOLOGY OF THE CAROTID ATHEROSCLEROTIC PLAQUE © Marten Siemelink, 2016
ISBN: 978-90-393-6532-8
Cover: Max Siemelink and Wendy Schoneveld Design: www.wenziD.nl | Wendy Schoneveld Printed by: BOXPress BV || Proefschriftmaken.nl Published by: Uitgeverij BOXpress, ‘s-Hertogenbosch
This research is supported by the Dutch Technology Foundation STW, which is part of the Netherlands Organisation for Scientific Research (NWO) and partly funded by the Ministry of Economic Affairs (Project 11679)
This research is supported by the European Union FP7 program “BiomarCaRE” (HEALTH-2011-278913).
Financial support by the Dutch Heart Foundation for the publication of this thesis is gratefully acknowledged.
Kindly, the printing of this thesis was financially supported by: Heart and Lung Foundation Utrecht, Chipsoft.
20162300 Marten Siemelink.indd 2 15-03-16 13:22 EXPLORING THE COMPLEX BIOLOGY OF THE CAROTID ATHEROSCLEROTIC PLAQUE
Verkenning van de complexe biologie van de atherosclerotische carotis plaque (met een samenvatting in het Nederlands)
Proefschrift
ter verkrijging van de graad van doctor aan de Universiteit Utrecht op gezag van de rector magnificus, prof.dr. G.J. van der Zwaan, ingevolge het besluit van het college voor promoties in het openbaar te verdedigen op donderdag 12 mei 2016 des middags te 4.15 uur
door
Marten Antoon Siemelink geboren 21 december 1981 te Utrecht
20162300 Marten Siemelink.indd 3 15-03-16 13:22 Promotoren: Prof.dr. G. Pasterkamp Prof.dr. A.B.J. Prakken
Copromotor: Dr. H. el Azzouzi
20162300 Marten Siemelink.indd 4 15-03-16 13:22 Aan mijn ouders
20162300 Marten Siemelink.indd 5 15-03-16 13:22 CONTENTS
CHAPTER 1 General Introduction 9
CHAPTER 2 Taking Risk Prediction to the Next Level. Advances in Biomarker 19 Research for Atherosclerosis Curr Pharm Des. 2013; 19(33):5929-42. Review
CHAPTER 3 Biomarkers of Coronary Artery Disease: The Promise of the 45 Transcriptome Curr Cardiol Rep. 2014 Aug;16(8):513. Review
CHAPTER 4 Coronary Artery Disease and Large Artery Stroke Loci are 63 Associated with Human Atherosclerotic Plaque Characteristics Manuscript in preparation
CHAPTER 5 Common Variants associated with Blood Lipid Levels do not Affect 85 Carotid Plaque Composition Atherosclerosis. 2015 Sep;242(1):351-6
CHAPTER 6 Cardiovascular Risk Loci associate with DNA Methylation in Carotid 99 Plaques Manuscript in preparation
CHAPTER 7 Tobacco Smoking is associated with DNA Methylation in Human 115 Atherosclerotic Plaques Manuscript in preparation
CHAPTER 8 Sex-Specific Differences in DNA Methylation in the Atherosclerotic 131 Carotid Artery Manuscript in preparation
20162300 Marten Siemelink.indd 6 15-03-16 13:23 CHAPTER 9 Systemic Glucocorticoids are associated with Mortality Following 147 Carotid Endarterectomy J Cardiovasc Pharmacol. 2015 Jul 17
CHAPTER 10 Discussion and Future Perspectives 163
CHAPTER 11 Dutch Summary 171
APPENDIX Review Committee 180 Acknowledgements 181 List of Publications 184 Curriculum Vitae 186
20162300 Marten Siemelink.indd 7 15-03-16 13:23 20162300 Marten Siemelink.indd 8 15-03-16 13:23 CHAPTER 1
General Introduction
20162300 Marten Siemelink.indd 9 15-03-16 13:23 10 | CHAPTER 1
Introduction
Cardiovascular disease (CVD) as result of atherosclerosis is a major cause of morbidity and mortality in many societies, despite the enormous research efforts in recent decades directed at prevention and improved treatment. Even more so, its prevalence is steadily rising in developing countries as living standards improve and lifestyles westernize.1–3 Atherosclerotic CVD may occur isolated or simultaneously at several predilection sites within the vasculature, including the carotid arteries, coronary arteries, aortic artery and more peripherally, the iliac and femoral arteries.4 Atherosclerosis of these vessels can result in life threatening health complications including stroke, myocardial infarction, aortic dissection and leg amputation. Cardiovascular disease (CVD) in broad sense of the word, refers to afflictions of the circulatory system, which can be manifold. However, it is often used synonymously with the most prevalent subtype of cardiovascular diseases, i.e. atherosclerotic disease, which is also true for this thesis. The focus of this thesis will primarily be on (atherosclerotic) carotid artery disease, yet concurrent and subsequent manifestations of atherosclerosis at the other vascular territories will also be considered.
Atherosclerosis of the carotid arteries is an insidious process that occurs over decades. Gradually, atherosclerotic plaque accumulates in the carotid artery, in general most prominently near the aortic bifurcation, but it may also extend to the common carotid artery and the internal- and external carotid artery. The process of atherosclerosis had been elegantly described by Weber et al. and by Jackson.5,6 At first, cholesterol and lipid depositions accumulate within the vascular wall, causing a ‘fatty streak’ lesion. This triggers the expression of chemokines in the overlaying vasculatory endothelium, causing the adherence, rolling and extravasation of leukocytes. Mainly monocytes and T-cells invade the plaque in this fashion, probably in an attempt to clean up the vascular lipid deposits. As the clean up fails, monocytes become foam-cells and surrounding cells start dying. This creates a necrotic core filled with atheroma, cell-debris and cholesterol-crystals. In reaction, smooth-muscle cells proliferate which causes the lesion to protrude into the vascular lumen, sometimes even obliterating the lumen completely, blocking blood-flow. At the last stage of atherosclerotic plaque development, the overlying smooth-muscle cells are receding, resulting in a ‘thin-cap atheroma’, with only minimal cells separating the blood flow from the highly thrombogenic necrotic core. Often without heralding symptoms, the unstable plaque erodes or ruptures, releasing the necrotic core into the bloodstream causing local occlusion of the carotid artery or creating an embolus which travels up the carotid arteries towards the brain. In both cases, this may lead to occlusion of cerebral arteries causing complications including transient ischemic attacks, or if the occlusion is more severe, cerebral stroke.
While the exact mechanism of the development of atherosclerotic carotid artery disease still remains to be elucidated, much more insight has been gained in recent years. All evidence indicates a complex disease, to which many risk factors contribute and where many complex biological processes are involved. A number of ‘traditional’ risk factors have been uncovered through epidemiological studies, most notably the Framingham Heart Study.7 These include age, sex, smoking, blood pressure and cholesterol in circulating lipoproteins. Although the traditional risk scores like the Framingham risk score and the
20162300 Marten Siemelink.indd 10 15-03-16 13:23 GENERAL INTRODUCTION | 11
SCORE risk-charts provide patients a 10-year risk of CVD, this is only a very general risk- prediction based on large cohort studies.8,9 Yet, they still form the basis of current clinical 1 risk assessment. That is mainly due to their low cost and easy-of-use for daily practice. While many efforts have been made to improve over these scores, sometimes showing incremental improvement, none have found general acceptance and clinical implementation as of yet. This may be due to the large effort and substantial costs involved in replication and validation of results, as well as reluctance of physicians to adopt new risk assessment methods. Many efforts have been directed at new biomarkers, which may include any patient measurement that provides information on disease risk. In the narrow-sense, many biomarkers are endogenous molecules, obtained through venous puncture, which are assessed for risk prediction. They may be obtained from various biological sources like cells, cell surface markers, proteins, RNA, microRNA, microvesicles, lipids and metabolites. Substantial advances in risk prediction may be expected when genetic risk-profiling is also incorporated, yet when this may be implemented remains to be seen due to increased costs and complexity. Biomarkers have also attracted the interest of large pharmaceutical companies, especially in the form of companion diagnostics.10 These biomarkers are developed to determine if a patient is responsive to a drug and to possibly adjust treatment regimen. Severe carotid artery disease poses a significant risk of further cerebral complications, and is an indication for preventative treatment. The lifestyle changes, medication and sometimes surgery. As all physicians are aware, lifestyle changes are the hardest to achieve. To reduce CVD risk, advice generally includes a healthy diet, management of body weight, regular exercise and to stop smoking, among others. Medication is usually started if an increased risk is predicted. The mainstay of preventative drug treatment for CVD disease is treatment with cholesterol-lowering medication, including statins and sometimes ezetimibe. Further treatment may include anti-hypertensives to reduce blood-pressure and medication to control blood glucose levels in case of diabetes. If patients have symptoms of carotid artery disease, like retinal infarction, transient ischemic attacks, stroke or if patients have severe stenosis of the carotid artery lumen, surgery may be indicated. Surgery comprises of a carotid endarterectomy procedure, whereby the carotid artery is exposed, incised and the vascular stenosis is alleviated by removal of the atherosclerotic stenosis. Alternatively, minimally invasive carotid artery stenting may be performed to ensure patency of the arterial lumen.
The Athero-Express Biobank The carotid endarterectomy procedure provides an excellent opportunity to acquire the disease lesion for further investigation. To leverage this opportunity, the Athero-Express Biobank was started in 2002, including patients that undergo carotid endarterectomy in the University Medical Center Utrecht and the St. Antonius Hospital in Utrecht, the Netherlands.11 Patients that are included are asked to donate the removed carotid plaque lesion as well as blood samples, complete medical questionnaires and agree to medical follow-up for three years to assess clinical outcome. Examination of the atherosclerotic plaque constitution may yield important new insights into the complex biology of cardiovascular disease. In the Athero-Express, carotid artery atherosclerotic plaques obtained during surgery are routinely examined by immunohistochemistry and evaluated by microscope for several histological features, of
20162300 Marten Siemelink.indd 11 15-03-16 13:23 12 | CHAPTER 1
which the most prominent are: smooth-muscle cells, monocytes, microvessels, atheroma, plaque haemorrhage, collagen and calcifications. This allows for the investigation of mechanisms that drive plaque composition in humans, mechanisms which may predispose to eventual plaque rupture or erosion instead of a stable plaque or plaque regression. Furthermore, many proteins of interest have been measured in blood or plaque specimens, and continue to be evaluated for their potential as biomarkers. Finally, for many patients in the Athero-Express, genotyping and methyltyping was performed using microarray technology. Therefore, the Athero-Express is an exceptionally well phenotyped cohort with unique value for the investigation of atherosclerotic cardiovascular disease and carotid artery disease in specific. This has already led to some key publications based on the Athero-Express.
Genetics Research in families and in monozygotic twins has shown that cardiovascular disease has a significant heritable component, with estimates of 32% heritability for stroke death and even 78% heritability for presence of carotid plaque.12,13 With the onset of the genetics revolution, it has become clear that this heritability is due to complex genetic mechanisms that contribute to CVD risk. Part of these heritable factors has been shown to be due to common genetic variation in many single nucleotide polymorphisms. Currently, it is estimated that 12.8% of heritability is explained by approximately twenty thousand nominally associated SNPs, indicating that there is still a significant unexplained heritability.15 Although mechanism and causality has not been proven for most of these variants, they may hold important clues towards uncovering the complex biology of disease. Furthermore, overlap in genetic determinants of both carotid disease and coronary artery disease provides support for substantial common pathophysiology between atherosclerotic CVD at different predilection sites.16 Additionally, there is also substantial genetic overlap between two subtypes of carotid artery disease: large artery atherosclerosis and small vessel disease.15 While for some variants situated in the body of genes the effect is all too clear (e.g. non- synonymous SNPs in exons that truncate the protein), for many variants which are situated in non-coding regions their mechanism of action remains elusive. Oftentimes, variants may be situated in non-coding regions with many genes in close proximity. At other times variants may have an effect over long distances due to enhancer elements. While genome- wide association studies (GWAS) have uncovered variants that predispose to cardiovascular disease, it has also shown that many genetic variants influence predisposing risk factors like circulating lipids17 and blood pressure,18 effects which may persist despite treatment to reduce these risk factors. Ultimately, these variants will also contribute to disease risk and clinical outcome, yet their individual effect may be even smaller. The combined effects of all known genetic variation contributing to disease risk may be captured in a weighted polygenic risk score. In such a score the effects of all identified risk-alleles in an individual are added up to provide an estimate of the total genetic burden contributing to disease. Additionally, these variants may indicate important pathophysiological pathways and mechanisms that qualify as important drug targets.
Epigenetics Equally important as protein function is protein expression. Regulators of protein expression that inherit independent of the genetic sequence include DNA-methylation, histone
20162300 Marten Siemelink.indd 12 15-03-16 13:23 GENERAL INTRODUCTION | 13
modifications and non-coding RNA. These processes, colloquially referred to as ‘epigenetic’, act in close concert to dynamically direct RNA- and protein expression in cells. Although 1 epigenetic mechanisms have been suspected and studied for many years, recent technological advances have really spurred this field into motion. Most interestingly, it has become feasible to measure DNA methylation in an ‘epigenome-wide’ fashion.19 DNA
methylation is an actively controlled process whereby a methyl group (CH3) is covalently bound to cytosine-phosphate-guanine dinucleotide bases (CpGs). The exact mechanisms controlling this are not completely understood. Yet, it has been shown that DNA methylation is highly tissue specific and may be affected by genetic variance as well as environmental factors. Furthermore, there is clear evidence that DNA methylation affects gene transcription. This is dependent on the spatial location of the DNA methylation in relation to the gene.20 DNA methylation in- or near the gene promoter down regulates gene transcription,21,22 whereas there is increasing evidence that DNA methylation in the gene body up regulates gene transcription.23 By these virtues, DNA methylation provides a great resource for the investigation of complex diseases such as atherosclerotic cardiovascular disease. Many efforts in epigenome-wide association studies (EWAS) in the cardiovascular field have been directed at the effects of tobacco smoking on DNA methylation, especially in blood cells.24 Smoking, an important risk factor for cardiovascular disease, was shown to modify gene expression in blood cells, by means of DNA methylation at specific loci.25–27 Additionally, DNA-methylation at some smoking-induced loci was shown to be predictive of clinical outcome.28 Furthermore, DNA methylation at specific CpG-loci has been shown to be affected by other CVD risk factors such as diabetes,29 blood pressure,18 BMI30 and plasma lipids.31 Another risk-factor for CVD is the sex of the patient.32 This may be due to differences in the pathophysiology of CVD between the sexes. Women have a reduced risk of CVD and suffer from a more microvascular type of lesion with increased plaque erosion as opposed to plaque rupture.33 The reasons for which are as of yet unclear. Previous studies by other investigators, in model organisms as well as humans, have shown major differences in DNA methylation between the sexes, which may contribute to a different vascular biology and pathofysiology.34 Differences in DNA methylation between the sexes are especially evident on the sex-chromosomes.35 To ensure equal expression of most X-chromosomal genes between the sexes, dosage compensation is achieved through inactivation of one of the X-chromosomes in women. This process involves DNA- methylation of the inactivated chromosome. In addition to environmental risk factors, epigenetic mechanisms like DNA methylation are also affected by genetic sequence variation. These interactions between the genome and the epigenome are largely unknown. Large-scale efforts have tried to elucidate this, such as the ENCODE-project36,37 and the Epigenome Roadmap,38 which have shown complex mechanisms involving many intermediaries though much remains unexplained. Recent publications have employed microarray data of the genome and epigenome to investigate the effects of common genetic variation on DNA-methylation. These genetic variants are known as methylation quantitative trait loci (meQTL). Some of these meQTLs may only be a reflection of haploblocks without CpG sites, effectively preventing methylation at that locus. Yet, other meQTL associations may presumably indicate important regulatory mechanisms affected by genetic variation. Excitingly, it may also allow for elucidation of gene-gene interactions in-trans, also across chromosomes.39
20162300 Marten Siemelink.indd 13 15-03-16 13:23 14 | CHAPTER 1
To summarize, atherosclerotic carotid artery disease is a common affliction and a major contributing factor to debilitating or fatal cerebrovascular accidents. Major research efforts have been directed at alleviating the burden of carotid artery disease, yet curation and prevention it still remains elusive. The Athero-Express Biobank study offers a unique opportunity to study this disease. With the onset of the (epi)genomics era, powerful new tools have become available, offering new hope to reduce this societal burden and prevent this disease.
Thesis Outline
The aim of this thesis was to explore the complex biology of the human carotid atherosclerotic plaque in order to provide a stepping stone towards improved risk prediction and precision drug development. To this end, we leveraged the unique properties of the Athero-Express Biobank to investigate carotid plaque parameters including histology and DNA methylation in relation to systemic risk factors such as sex, genetic variation, lipid levels, medication and smoking.
In order to advance the prevention of atherosclerotic carotid artery disease, advances in risk prediction are required. This has been an ongoing effort for several decades, which started with the identification of ‘classical’ risk factors and more recently focussed on improved biomarkers of disease. In Chapter 2, a review of the latest international efforts in biomarker research is presented. Additionally, the importance of genetic and epigenetic research for the improvement of risk prediction is highlighted. Changes in genetics and epigenetics finally result in changes to RNA structure and expression. Unaffected by the difficulties encountered in proteomics, investigation of RNA using transcriptomics holds great promise, which is reviewed in Chapter 3. Although current state-of-art research for classical biomarkers may yield improved risk- predictors, it is becoming clear that genetic and epigenetic data may be required to arrive at the age of precision medicine.40,41 Large-scale international efforts have indicated common genetic variation that associates with clinical or subclinical manifestations of atherosclerotic disease. In Chapter 4, we investigate if these variants also affect the composition of the vascular lesion, an important step in understanding the pathophysiological mechanisms that are driven by these variations. While genetic variation may directly associate with the vascular lesion and clinical outcome, it may also affect the severity of risk factors such as blood pressure, inflammation and plasma lipids. Disturbed lipid metabolism is one of the hallmarks of atherosclerotic disease. This is supported by the fact that abnormal concentrations plasma lipids are an important risk factor for carotid artery disease. In Chapter 5, genetic variation affecting the concentration of plasma lipids is used in an effort to elucidate the effect of plasma lipids on carotid plaque composition. In the introduction, the emerging importance of epigenetic research is emphasized. In this thesis, the focus of epigenetic research is on DNA-methylation. DNA-methylation at specific genomic locations may be strongly affected by genetic sequence variation. To investigate if known cardiovascular-risk SNPs also affect local DNA-methylation in the vascular lesion, methylation quantitative trait analysis was performed which is presented in Chapter 6. DNA methylation may provide information on both genetic and environmental effects,
20162300 Marten Siemelink.indd 14 15-03-16 13:23 GENERAL INTRODUCTION | 15
making it a unique resource for risk-prediction. In Chapter 7, these unique properties are utilized to investigate the effect of smoking on the vascular lesion and study a possible 1 heritable susceptibility to smoking. Cardiovascular disease is quite dissimilar between the sexes, and publications by other groups also indicate large differences in autosomal DNA methylation between the sexes in other tissues. Therefore, Chapter 8 investigates if these sex-dependent differences also exist in the atherosclerotic plaque, and if these changes contribute to differences in pathophysiology between men and women. While striving for better risk-prediction and disease prevention by lifestyle alterations, drug treatment is still a mainstay in the attenuation of cardiovascular disease. However, drug treatment may also have unwanted side effects. Due to comorbidities, patients increasingly receive multiple drug regimens adding to drug-drug interactions and side effects. While glucocorticosteroid drugs reduce inflammation, they have many side effects and are suspected to add to atherosclerotic disease. In Chapter 9, the effect of glucocorticosteroid treatment on long-term outcome in carotid endarterectomy patients is examined. Finally, in Chapter 10, we discuss the main findings in the studies presented in this thesis and provide the reader with a perspective on future developments.
20162300 Marten Siemelink.indd 15 15-03-16 13:23 16 | CHAPTER 1
References
1 Mendis S, Puska P, Norrving B. Global Atlas on cardiovascular disease prevention and control. Geneva, 2011. 2 Mathers CD, Loncar D. Projections of global mortality and burden of disease from 2002 to 2030. PLoS Med 2006; 3: e442. 3 Mathers CD, Boerma T, Ma Fat D. Global and regional causes of death. Br Med Bull 2009; 92: 7–32. 4 Imori Y, Akasaka T, Ochiai T, et al. Co-existence of carotid artery disease, renal artery stenosis, and lower extremity peripheral arterial disease in patients with coronary artery disease. Am J Cardiol 2014; 113: 30–5. 5 Weber C, Noels H. Atherosclerosis: current pathogenesis and therapeutic options. Nat Med 2011; 17: 1410–22. 6 Jackson SP. Arterial thrombosis--insidious, unpredictable and deadly. Nat Med 2011; 17: 1423–36. 7 D’Agostino RB, Vasan RS, Pencina MJ, et al. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation 2008; 117: 743–53. 8 Wilson PW, D’Agostino RB, Levy D, Belanger AM, Silbershatz H, Kannel WB. Prediction of coronary heart disease using risk factor categories. Circulation 1998; 97: 1837–47. 9 Conroy RM, Pyörälä K, Fitzgerald AP, et al. Estimation of ten-year risk of fatal cardiovascular disease in Europe: the SCORE project. Eur Heart J 2003; 24: 987–1003. 10 Agarwal A, Snyder G, Ressler D. The current and future state of companion diagnostics. Pharmgeno- mics Pers Med 2015; 8: 99. 11 Verhoeven BAN, Velema E, Schoneveld AH, et al. Athero-express: differential atherosclerotic plaque expression of mRNA and protein in relation to cardiovascular events and patient characteristics. Ra- tionale and design. Eur J Epidemiol 2004; 19: 1127–33. 12 Bak S, Gaist D, Sindrup SH, Skytthe A, Christensen K. Genetic liability in stroke: a long-term follow-up study of Danish twins. Stroke 2002; 33: 769–74. 13 Tarnoki AD, Baracchini C, Tarnoki DL, et al. Evidence for a strong genetic influence on carotid plaque characteristics: an international twin study. Stroke 2012; 43: 3168–72. 14 Fox CS, Polak JF, Chazaro I, et al. Genetic and environmental contributions to atherosclerosis phe- notypes in men and women: heritability of carotid intima-media thickness in the Framingham Heart Study. Stroke 2003; 34: 397–401. 15 Holliday EG, Traylor M, Malik R, et al. Genetic overlap between diagnostic subtypes of ischemic stroke. Stroke 2015; 46: 615–9. 16 Dichgans M, Malik R, König IR, et al. Shared genetic susceptibility to ischemic stroke and coronary artery disease: a genome-wide analysis of common variants. Stroke 2014; 45: 24–36. 17 Lipids G, Consortium G. Discovery and refinement of loci associated with lipid levels. 2013. DOI:10.1038/ng.2797. 18 Kato N, Loh M, Takeuchi F, et al. Trans-ancestry genome-wide association study identifies 12 genetic loci influencing blood pressure and implicates a role for DNA methylation. Nat Genet 2015; 47: 1282– 93. 19 Sandoval J, Heyn H, Moran S, et al. Validation of a DNA methylation microarray for 450,000 CpG sites in the human genome. Epigenetics 2011; 6: 692–702. 20 Varley KE, Gertz J, Bowling KM, et al. Dynamic DNA methylation across diverse human cell lines and tissues. Genome Res 2013; 23: 555–67. 21 Curradi M, Izzo A, Badaracco G, Landsberger N. Molecular mechanisms of gene silencing mediated by DNA methylation. Mol Cell Biol 2002; 22: 3157–73. 22 Ma A-N, Wang H, Guo R, et al. Targeted gene suppression by inducing de novo DNA methylation in the gene promoter. Epigenetics Chromatin 2014; 7: 20. 23 Maunakea AK, Nagarajan RP, Bilenky M, et al. Conserved role of intragenic DNA methylation in regulat- ing alternative promoters. Nature 2010; 466: 253–7. 24 Gao X, Jia M, Zhang Y, Breitling LP, Brenner H. DNA methylation changes of whole blood cells in response to active smoking exposure in adults: a systematic review of DNA methylation studies. Clin Epigenetics 2015; 7: 113. 25 Reynolds LM, Wan M, Ding J, et al. DNA Methylation of the Aryl Hydrocarbon Receptor Repressor Associations With Cigarette Smoking and Subclinical Atherosclerosis. Circ Cardiovasc Genet 2015; 8: 707–16. 26 Steenaard R V, Ligthart S, Stolk L, et al. Tobacco smoking is associated with methylation of genes related to coronary artery disease. Clin Epigenetics 2015; 7: 54. 27 Guida F, Sandanger TM, Castagné R, et al. Dynamics of smoking-induced genome-wide methylation changes with time since smoking cessation. Hum Mol Genet 2015; 24: 2349–59.
20162300 Marten Siemelink.indd 16 15-03-16 13:23 GENERAL INTRODUCTION | 17
28 Zhang Y, Yang R, Burwinkel B, et al. F2RL3 methylation in blood DNA is a strong predictor of mortality. Int J Epidemiol 2014; 43: 1215–25. 29 Soriano-Tárraga C, Jiménez-Conde J, Giralt-Steinhauer E, et al. Epigenome-wide association study 1 identifies TXNIP gene associated with type 2 diabetes mellitus and sustained hyperglycemia. Hum Mol Genet 2015; published online Dec 7. DOI:10.1093/hmg/ddv493. 30 Demerath EW, Guan W, Grove ML, et al. Epigenome-wide association study (EWAS) of BMI, BMI change and waist circumference in African American adults identifies multiple replicated loci. Hum Mol Genet 2015; 24: 4464–79. 31 Pfeiffer L, Wahl S, Pilling LC, et al. DNA Methylation of Lipid-Related Genes Affects Blood Lipid Levels. Circ Cardiovasc Genet 2015; 8: 334–42. 32 Mosca L, Barrett-Connor E, Wenger NK. Sex/gender differences in cardiovascular disease prevention: what a difference a decade makes. Circulation 2011; 124: 2145–54. 33 den Ruijter HM, Haitjema S, Asselbergs FW, Pasterkamp G. Sex matters to the heart: A special issue dedicated to the impact of sex related differences of cardiovascular diseases. Atherosclerosis 2015; 241: 205–7. 34 Singmann P, Shem-Tov D, Wahl S, et al. Characterization of whole-genome autosomal differences of DNA methylation between men and women. Epigenetics Chromatin 2015; 8: 43. 35 Cotton AM, Price EM, Jones MJ, Balaton BP, Kobor MS, Brown CJ. Landscape of DNA methylation on the X chromosome reflects CpG density, functional chromatin state and X-chromosome inactivation. Hum Mol Genet 2015; 24: 1528–39. 36 A user’s guide to the encyclopedia of DNA elements (ENCODE). PLoS Biol 2011; 9: e1001046. 37 Bernstein BE, Birney E, Dunham I, Green ED, Gunter C, Snyder M. An integrated encyclopedia of DNA elements in the human genome. Nature 2012; 489: 57–74. 38 Skipper M, Eccleston A, Gray N, et al. Presenting the epigenome roadmap. Nature 2015; 518: 313. 39 Lemire M, Zaidi SHE, Ban M, et al. Long-range epigenetic regulation is conferred by genetic variation located at thousands of independent loci. Nat Commun 2015; 6: 6326. 40 Khoury MJ, Iademarco MF, Riley WT. Precision Public Health for the Era of Precision Medicine. Am J Prev Med 2015; published online Nov 4. DOI:10.1016/j.amepre.2015.08.031. 41 Hawgood S, Hook-Barnard IG, O’Brien TC, Yamamoto KR. Precision medicine: Beyond the inflection point. Sci Transl Med 2015; 7: 300ps17.
20162300 Marten Siemelink.indd 17 15-03-16 13:23 M.A. Siemelink1 S.W. van der Laan1 L. Timmers1,2 I.E. Hoefer1 G. Pasterkamp1
1 Laboratory of Experimental Cardiology, University Medical Center Utrecht, Utrecht, the Netherlands 2 Department of Cardiology, University Medical Center Utrecht, Utrecht, the Netherlands
20162300 Marten Siemelink.indd 18 15-03-16 13:23 CHAPTER 2
Taking Risk Prediction to the Next Level Advances in Biomarker Research for Atherosclerosis
CURR PHARM DES. 2013; 19(33):5929-42. REVIEW
20162300 Marten Siemelink.indd 19 15-03-16 13:23 20 | CHAPTER 2
Abstract
Advances in risk prediction are necessary to stem the tide of the increasing incidence of global cardiovascular disease. Newly discovered biomarkers are needed for primary and secondary prevention and will undoubtedly play a major role in drug development programs and monitoring of treatment efficacy. The combination of improved –omics technologies and the investigation of relatively untapped sources of biomarkers will likely result in risk algorithms that will add value on top of the traditional risk factors. New sources of biomarkers are being explored with encouraging results. These include microvesicles, microRNAs, circulating cells and atherosclerotic plaques. We will review these sources for their potential for new biomarkers. Furthermore, the major impact of advances in genetics on risk prediction and biomarker development programs will be discussed.
20162300 Marten Siemelink.indd 20 15-03-16 13:23 ADVANCES IN BIOMARKER RESEARCH | 21
Introduction
Preventing the onset and progression of cardiovascular disease is key in the battle against the global epidemic of cardiovascular disease (CVD). Therefore, improved risk prediction and corresponding treatment is paramount. There is an intensified research effort to discover and validate novel biomarkers that facilitate the prediction of clinical manifestations 2 of CVD, in which most attention has been paid to circulating biomarkers.1,2 Serum and plasma are widely available in population and cohort studies allowing fast retrospect validation of potential biomarkers. However, the known circulating markers of inflammation, oxidative stress, endothelial dysfunction, platelet activation have all been tested for potential predictive value for future events,3 but have only shown modest added value for prediction of adverse events on top of the traditional risk factors.4,5 The advent of modern –omics technologies has resulted in a wave of new discoveries of genetic variations, microRNAs, proteins and metabolites that associate with atherosclerotic disease progression. However, these discovery strategies will often fail to expose associations due to lack of sensitivity and inherently yield many false-positive associations. In addition, the function of the proposed target is often unknown, which may increase the chances of pursuing false-positive targets.6 Despite the inherent limitations, these discovery tracks will undoubtedly result in new insights into the pathophysiology of atherosclerotic disease, which is warranted since the decision to pursue a novel biomarker is still based on the available knowledge of the mechanism of action. The utility of promising biomarkers could go beyond the area of diagnosis and prevention. New therapeutic targets are searched for to fight the progression and destabilization of atherosclerotic disease.7,8 Many associations between disease prevalence and explored biomarkers will be the result of reverse causation: the expression is the consequence rather than a causal factor in the disease process.9 However, in other cases the biomarker is causally related with the initiation or progression of atherosclerosis opening up new avenues for drug development programs. Biomarkers can also facilitate the clinical introduction of a compound when treatment efficacy is reflected by changes in biomarker profiles. Especially the latter application is emerging in light of the major investments currently required for the introduction of a new drug into clinical practice. The relevance of the search for biomarkers reflecting treatment efficacy is especially applicable to the domain of atherosclerotic disease: the disease develops over decades and atherosclerosis hard endpoint studies or serial imaging studies demand for significant budgets which can only be afforded by globally operating pharmaceutical companies.
Biomarker discovery: from circulation to the source In physics and biology, the strongest signal is generally obtained closest to the transmitting source. This also applies to biomarkers. The circulating blood is a convenient source for sampling, but it reflects the metabolic and diseased state of any organ in the human body. Unless the biomarker is organ or disease specific, e.g. troponin-T in the diagnosis of myocardial damage, the circulating biomarker will always suffer from a poor signal to noise ratio. This leads to poor predictive values of markers and limited clinical utility. Excretory fluids like urine and saliva are even further from the source, and surpass the scope of this review.
20162300 Marten Siemelink.indd 21 15-03-16 13:23 22 | CHAPTER 2
This noise can be reduced by sampling biomaterial that more specifically reflects the state of a disease. This can be achieved by blood sampling distal of the diseased tissue where biomarker concentrations may be elevated,10 a method that requires an interventional procedure. Also other sources than blood can be considered that more optimally reflect the current disease state. These could be used as a measure of future risk for the development of acute manifestations of cardiovascular disease. In this review we will discuss recent developments in the improvement of risk prediction for CVD, discussing the potential of several biomaterial sources. Furthermore, we discuss the value of the recent developments in genetic research for CVD risk assessment and biomarker validation.
Circulating biomarkers Excellent reviews have been published summarizing the value of known plasma or serum biomarkers that reflect manifest or future development of atherosclerotic disease.11,12 With the exception of lipoprotein profiles, most plasma and serum biomarkers offer limited added value on top of the classical risk factors. Furthermore, many markers that reflect a systemic inflammatory response, such as C-reactive protein, suffer from a lack of specificity and their application for individual risk stratification after adjustment for risk factors is unlikely. The discovery of new potentially strong biomarkers in plasma is complex. Long-distance communication between tissues is critically dependent on circulatory signalling molecules, e.g. classical hormones, cytokines, chemokines and adipokines. These signalling molecules are generally very potent and quite specific, and hence are present in very low concentrations. On the other hand, proteins like albumin and the immunoglobulins are abundantly present in the circulation. However, the low-abundance signalling molecules could be very important in systemic vascular disease like atherosclerosis. Current quantitative proteomic techniques have insufficient sensitivity to reliably identify and quantify these low-abundance signalling substances in plasma. Therefore, several techniques are employed to increase the power of plasma proteomics. To eliminate the obscuring effects of high-abundance proteins, affinity-based protein depletion columns are used. Plasma fractionation using strong-cation exchange can bring the detection limit down to about 2.5 ng/ml. Furthermore, selective enrichment of the sample for a particular sub- proteome, using techniques such as immunoprecipitation or liquid chromatography can also improve precision and increase yields.13,14 It merits careful consideration however, that these enhancing techniques require further processing of the sample and have the potential to influence measurements, reducing reliability and possibly increasing false-positive findings. To complicate matters even further, many of the circulating biomarkers may have opposing or reinforcing effects, forming an intricate signalling network. Measuring just one or several of these circulating elements within a network may not accurately reflect the in vivo effects, and hence may not represent the net biological effect. Thus, it can be hypothesized that assessment of the direction of imbalance, by measuring multiple biomarkers simultaneously in the whole biological cascade is more informative. This review will continue with a summary of research efforts in search for biomarker sources that could more accurately reflect disease progression.
20162300 Marten Siemelink.indd 22 15-03-16 13:23 ADVANCES IN BIOMARKER RESEARCH | 23
Microvesicles Investigation of a specific fraction of the circulation holds promise to reduce signal-to-noise ratio’s and may more accurately represent disease state. Important constituents of the circulation that have received major interest as a source of biomarkers are microvesicles. Microvesicles (MVs) are a heterogeneous group of membrane shed vesicles, including exosomes, membrane particles and apoptotic bodies which can be discriminated by size, 2 antigen expression and behaviour.15 MVs are released by different cell types, including circulating cells and cells present in the vascular wall, into the extracellular space after cell activation or apoptosis.16–18 The release of MVs enables cells to influence processes over distance. Microvesicles can directly interact with ligands and receptors in the surface of target cells. In addition, intercellular communication can be mediated by internalization and the subsequent transfer of surface membrane proteins, (micro)RNAs and bioactive lipids from one cell to another. The expression of MVs is associated with a variety of diseases, such as cancer,19–21 infectious diseases22–25 and cardiovascular disease. Atherosclerotic lesions contain large amounts of MVs, mostly originating from leucocytes.26,27 Several roles of MVs have been identified in the initiation and progression of atherosclerotic disease.28 Microvesicles cause endothelial dysfunction and increased endothelial permeability by impairing endothelial NO bioavailability.29–32 In addition, MVs of different cellular origins increase the synthesis and release of pro-inflammatory cytokines and adhesion molecules by endothelial cells and leucocytes, stimulating chemotaxis of inflammatory cells into the vessel wall.33–37 Also neovessel formation can be stimulated by MVs.38 Finally, highly thrombogenic tissue factor carrying MVs in atherosclerotic plaques and in the circulation contribute to coagulation and thrombosis following plaque rupture.39 On the other hand, MVs can have anti-atherogenic properties, for example the release of anti-inflammatory TGF-beta from macrophages is mediated by neutrophil MVs,40 and lymphocytic MVs inhibit angiogenesis.41 Cells that are activated or apoptotic increasingly release MVs. Therefore circulating levels of MVs can provide clinically relevant information. Both cardiovascular risk factors and overt cardiovascular disease are associated with increases in MVs. Furthermore, recent reports suggest that MVs can predict major adverse cardiovascular disease outcome. Current evidence showing the value of microvesicles as biomarkers for atherosclerosis in humans is summarized in Table 1. A novel concept within biomarker research is the analysis of MV protein composition. This method has already been applied for the discovery of MV related biomarkers for a number of diseases, such as urinary tract, renal and cerebral disease, but not for cardiovascular disease.42 Microvesicles harbour at their surface most of the membrane associated proteins of the cells they originate from and reflect the status of this cell. The protein content and function of the MVs that cells shed, alter with the pathophysiological context in which MVs are released.43,44 Application of the latest proteomics techniques could lead to the discovery of novel MV related biomarkers, and reveal novel mechanistic information regarding the involvement of MVs in disease processes. Thus, MVs constitute an interesting source for biomarker research. The first explorations are promising, but additional studies are required to determine the clinical utility of measuring MV levels for diagnostic and prognostic purposes, since evidence of its discriminative value, predictive value, and reclassification effect is still limited.
20162300 Marten Siemelink.indd 23 15-03-16 13:23 24 | CHAPTER 2
Table 1. Studies associating microvesicles with atherosclerotic disease and risk factors Cohort Outcome Microvesicle type Reference High-risk of CAD Cardiovascular events Endothelial 166 CAD Cardiovascular events Endothelial 167 CAD Stable/unstable angina, MI Endothelial and platelet 168 CAD Acute coronary syndrome Multiple 169 Heart failure Cardiovascular events Endothelial 170 Acute ischemic stroke Severity, outcome Endothelial 43 End-stage renal disease Cardiovascular mortality Endothelial, platelet, erythrocyte 29,171 Type 2 diabetes mellitus CAD Endothelial 172 CAD: Coronary artery disease. MI: Myocardial infarction
MicroRNA Another biomarker source that can be obtained from the circulation is microRNA. MicroRNAs are small pieces of non-coding RNA that regulate gene expression. These microRNAs hybridize with specific sequences in the untranslated regions of target mRNA, thereby inducing the degradation of that mRNA and preventing protein translation. Many microRNAs have been found to be highly conserved among species and can be classified into families on the basis of homology. Furthermore, specific microRNAs may influence the expression of multiple mRNA targets. On the other hand, multiple microRNAs may be involved in regulating the expression of a single mRNA. These interactions form a complicated system of post-transcriptional gene-regulation that is only just starting to be elucidated. MicroRNAs have been shown to be stably present in blood plasma and circulating cells.45 That microRNAs were found in the plasma despite the presence of degrading ribonucleases was surprising.46 Since then, it has been shown that the resistance of MicroRNAs to degradation is due to either packaging in microvesicles or association with proteins or lipoproteins. MicroRNA can be selectively packed in microvesicles and actively secreted to influence expression in targets cells. Zhang et al. identified miR-150 as a microRNA that is abundantly packed into microvesicles and could influence c-Myb protein levels in cultured endothelial cells, and that plasma microvesicles of atherosclerotic patients had increased levels of miR-150.47 Furthermore, Zernecke et al. showed in a mouse study that apoptotic bodies of wild-type mice had an atheroprotective effect in ApoE-/- mice, and that this effect was abrogated when using apoptotic bodies from miR-126 deficient mouse.48 These studies reveal that MicroRNAs can infer a regulatory effect on target tissues, possibly influencing local disease progression. This also indicates they could serve as causal biomarkers. To further investigate the potential of microRNAs as predictive biomarkers, Genome-Wide Association Studies (GWAS) will prove to be instrumental. In a landmark paper Gamazon et al. describe how a genome-wide expression profiling of 118 lymphoblastoid cell-lines from European or African ancestry identified a respective 100 and 114 microRNAs that were associated with the expression of 558 and 663 mRNAs.49 In addition, they found 12
20162300 Marten Siemelink.indd 24 15-03-16 13:23 ADVANCES IN BIOMARKER RESEARCH | 25
SNPs that associated with both microRNA and target mRNA expression levels which supports the view that the use of GWAS to validate microRNA-mRNA interactions could prove an important tool in showing the causality of putative microRNA biomarkers in mRNA expression levels and phenotypic traits. 2
Table 2. MicroRNAs associated with atherosclerotic disease Coronary artery disease Source Reference miR-17, miR-92a, miR-126, miR-145 and miR-155 Plasma 173 miR-340*, miR-451, miR-454, miR-624 Platelets 174 miR-146a/b PBMC 175 miR134, miR-135a, miR-147, miR-198, miR-370 PBMC 176 miR-19a, miR-29a, miR-30e-5p, miR-145, miR-150, miR-155, Whole blood 177 miR-181d, miR-222, miR-342, miR-378, miR-584 miR-92a/b, miR-140-3p, miR-182 Whole blood 178 Myocardial infarction miR-1 Plasma 179 miR-1, miR-133a, miR-208b Plasma 180 miR-1, miR-133a, miR-208a, miR-499 Plasma 181 miR-1, miR-21, miR146a, miR-208a, miR-499 Plasma 182 miR-1, miR-133a, miR-133b, miR-122, miR-375, miR-499-5p Plasma 183 miR-1, miR-133a, miR-208b, miR-499-5p Plasma 184 miR-1, miR-21, miR-133a, miR-423-5p, miR-499-5p Plasma 185 miR-126, miR-133, miR-208a, miR-499 Plasma 186 miR-208b, miR-499 Plasma 187 miR-208b, miR-499 Plasma 188 miR-499 Plasma 189 miR-1 Serum 190 miR-1, miR-133a Serum 191 miR-133, miR-328 Whole blood, plasma 192 miR-30c, miR-145, miR-663b, miR-1291 Whole blood 193 Stroke miR-145 Whole blood 194 miR-210 Whole blood 195 157 micoRNAs Whole blood 196 Peripheral artery disease miR-21, miR-27b, miR-130a, miR-210 Plasma 197 miR: MicroRNA. PBMC: Peripheral blood mononuclear cells.
20162300 Marten Siemelink.indd 25 15-03-16 13:23 26 | CHAPTER 2
Measuring microRNAs in whole blood of CVD patients has yielded several potential biomarker targets that associate with disease. Most studies have investigated microRNAs in whole plasma or serum samples, while microRNAs in other blood compartments could provide another biosource of information. The peripheral blood mononuclear cells (PBMCs) fraction holds several cell types that migrate to atherosclerotic plaques, like macrophages and T-cells. MicroRNAs in these cells can potentially influence disease progression, and could thus be interesting biomarker targets as well. Furthermore, microRNAs in platelets may directly influence platelet function and could induce a prothrombotic phenotype which increases the risk of a cardiac event. MicroRNAs in platelets may thus serve as a marker. Associations between microRNAs and disease are represented in Table 2. The accumulating evidence showing the value of MicroRNAs as a biomarker is encouraging. Furthermore, microRNAs can be reliably measured in the clinic, using current high- throughput PCR-machines. Their potential makes microRNAs an exciting field of research, especially when taking into account that many microRNAs have only recently been discovered.
Circulating cells Numerous studies have proven the involvement of cell populations from the circulation in the inflammatory processes that lead to atherosclerotic plaque development and progression.50–55 Recent advances in analytical equipment have boosted the investigation of circulating cells. Current flow-cytometers are capable of detecting more than ten markers simultaneously, allowing for cell typing and marker expression analysis down to very rare cell populations. Flow-cytometry is currently a standard analytical method for diagnosis and monitoring of haematological disorders, such as leukaemia or HIV. Due to its broad spectrum of applications, it represents a potent tool for the assessment of putative biomarkers on circulating cells. However, an important limitation of flow-cytometry is operational complexity due to working with living cells and the lack of high-throughput. After the initial report on endothelial progenitor cells (EPCs) by Asahara et al. in 1997,56 numerous studies found reduced EPC levels in patients at elevated risk of cardiovascular disease. Generally, patients at high risk for future cardiovascular events, e.g. patients with rheumatoid disorders, kidney disease or established cardiovascular disease, seem to have lower numbers of circulating EPCs compared to healthy subjects.57–64 Importantly, different studies handle different EPC definitions, which merit careful consideration. Monocytes and (T-)lymphocytes are particularly interesting cells for identification of biomarkers because of their causal role in atherosclerosis. Compared to EPCs, their definition and characterization is less controversial and their numbers are much higher, rendering their analysis less susceptible for artefacts. The prevailing opinion is that circulating leukocytes migrate into the vascular wall once having attached to the plaque covering endothelium. However, considering absolute cell counts, half-life in the circulation and the distribution within the vasculature, it is reasonable to assume that the majority of rolling and adhering cells finally detaches and enters the circulation again, carrying information obtained during the interaction with the endothelium. Furthermore, leukocytes are constantly under the influence of the surrounding microenvironment during circulatory transit and hence are affected by systemic conditions such as hyperlipidaemia. These characteristics make circulating leukocytes good candidates for biomarker discovery.
20162300 Marten Siemelink.indd 26 15-03-16 13:23 ADVANCES IN BIOMARKER RESEARCH | 27
Monocytes and T-lymphocytes form heterogeneous populations with deviating biological functions. Currently, at least 3 different monocyte subpopulations have been defined based on their surface expression of CD14 and CD16: Classical monocytes (CD14++CD16-), non- classical “inflammatory” monocytes (CD14+CD16++) and intermediate monocytes (CD14++CD16+).65 The combination of these two markers has recently been shown to be predictive for future cardiovascular events in coronary artery disease patients referred for 2 elective coronary angiography.66 Furthermore, monocyte subpopulations were reported to correlate with the presence of (sub)clinical atherosclerosis in high risk groups.67,68 More importantly, these inflammatory changes are not limited to adult patients but may already occur in early stages of obesity. When compared to lean controls, obese children have higher numbers of CD14++ monocytes with increased expression of the activation marker CD11b.69 This rather simple 2-3 colour panel can be extended with additional markers to improve its value. Especially the analysis of CD143 (ACE) expression on CD14++CD16+ monocytes was shown to improve the prediction of mortality in end-stage renal disease and associate with prevalent cardiovascular disease in dialysis patients.70,71 Like monocytes, T-lymphocytes can be divided into well-defined subpopulations and have gained much attention as carriers of biological information. Both, CD4+ and CD8+ lymphocytes have been associated with the presence and the extent of cardiovascular disease. CD4+CD28- lymphocytes are increased in end-stage renal disease patients with cardiovascular disease.72,73 A major advantage of this cell type is that it is hardly affected by acute coronary events, remaining relatively stable for up to 3 months after an acute MI.74 Regulatory T-cells have been shown to be critical in the maintenance of immune system homeostasis by inhibiting excessive inflammation. Impediment of regulatory T-cell numbers or function has been associated with various auto-immune disorders. Although being involved in atherosclerosis as a protective factor, the number of circulating CD4+CD25hiCD127lo regulatory T-cells do not reflect cardiovascular disease severity or extent.75 This evidence shows that circulating cell (sub)populations and cellular protein expression levels may hold predictive information, and can be reliably measured by flow-cytometry. However, the interest has mainly been on the investigation of a select number of proteins that are hypothesized to determine functional populations. Proteomics on circulating cells may still uncover many new cellular proteins with biomarker potential.
Cell-based functional testing As described above, certain cell populations may be causally involved in atherosclerosis, yet may not correlate with clinical characteristics. Moreover, most biological processes are regulated by a delicate balance of numerous factors. Consequently, subtle differences in this equilibrium may have severe effects. This balance could be studied by subjecting blood cells to a stimulus and evaluate the change in functionality either directly or by measuring surrogate markers of cell activation, e.g. surface marker expression or cytokine release. Platelets and platelet function play a pivotal role in atherothrombosis and platelet inhibitors form the basis of secondary prevention. As certain patients do not adequately react towards platelet inhibitors, many efforts have been made to identify these individuals by testing the response of their platelets to different stimuli. Various platforms for automatic platelet reactivity testing have been developed and introduced into a clinical setting. Independent of their ability to identify responders and non-responders, platelet function has been shown to predict future events.76
20162300 Marten Siemelink.indd 27 15-03-16 13:23 28 | CHAPTER 2
Functional assays are however not limited to platelets but can also be performed on whole blood or isolated blood cells. The innate immune system and its mediators are strongly involved in atherogenesis and plaque progression.77–81 Incubation of whole blood or blood cells with Toll-like receptor (TLR) ligands e.g. lipopolysaccharide, triggers a strong inflammatory response that has been shown to associate with disease severity and clinical presentation.82–84 Furthermore, monocytic TLR expression independently predicts atherosclerosis and cardiovascular events.85,86 This shows that functional assays may be employed to obtain information about responsiveness to a stimulus, and that this information may be predictive of future risk. Yet, such assays are not commonly used in biomarker studies, possibly because of the inability to work with stored samples, leaving much still to be explored in this field.
Cell-based in vitro assays The inherent limitation that a single biomarker is insufficient to represent the complexity of pathophysiological balances in the human biology, has led to the development of multi- marker panels. Alternatively, to circumvent the limitation of complexity and multi-marker testing one could consider indirect in vitro assays. Cell cultures can be used as a sensor to identify changes that occur in complex microenvironments like plasma. A cell-based assay is preferably based on stable human cell-lines, as opposed to ex-vivo cells, to reduce variability and improve standardization. Different stimuli can be exerted on the assay using patient derived biomaterial, e.g. serum, plasma or microvesicles. Measuring the effects of patient plasma on a cell type that is involved in the local pathology of atherosclerosis may thus uncover important signalling routes in the plasma. This type of assay is considered a method of “microenvironment testing” and could be optimised and utilized as a biomarker platform. Moreover, it could hold great value for the monitoring of drug-efficacy and the development of companion diagnostics, when measuring the effect of plasma obtained from individuals who are participating in interventional drug trials. Read-outs can be generated that provide information on the effect of plasma on specific cell functions, such as morphological changes, proliferation, migration, viability and protein expression. Indirect assessment of the effect of the micro environment on biological functions is already widely applicable in the field of thrombosis and haemostasis where thrombin generation and platelet activation tests are commonly used.87,88 Using a cell-line to identify disease specific changes in plasma has been previously shown to be a viable strategy in lipid research. Macrophage cholesterol efflux is regarded as a key mechanism to prevent foam- cell formation and subsequent plaque progression. It is therefore interesting to investigate to what extend cholesterol efflux is influenced by plasma factors. Several studies have shown that the capacity of serum to induce cholesterol efflux predicts cardiovascular disease.89,90 Furthermore, Khera et al showed that the capacity of human serum to induce cholesterol efflux from a murine monocyte cell-line was strongly inversely correlated with carotid intima-media thickness and angiographic coronary artery disease. When adjusted for known risk factors, an increased efflux capacity was associated with a significantly decreased risk of coronary artery disease.91 Furthermore the serum of diabetes type 2 patients, a known cardiovascular risk factor, has been shown to reduce cholesterol efflux capacity compared to controls.92 These studies show that using a cell-line to study the combined effects of known and unknown plasma factors on cholesterol efflux can yield data that are associated with the presence of CVD.
20162300 Marten Siemelink.indd 28 15-03-16 13:23 ADVANCES IN BIOMARKER RESEARCH | 29
Stimulation of THP-1 monocytes with plasma of patients suffering from systemic lupus erythematosus (SLE) has been shown to induce an increase of CD36, an important scavenger receptor involved in monocyte cholesterol uptake.93 This shows, how the use of a cell-based assay to investigate the cardiovascular risks inferred by SLE, can identify the existence of plasma factors that confer this risk. In a similar manner, Verrijn-Stuart et al recently showed that incubation of the human pre-adipocyte SGBS cell-line with plasma of children with type 2 1 diabetes induced increased adipocyte differentiation and a change in cytokine and adipokine levels when compared to plasma of healthy children, evidence that the increased risk of obesity in type 1 diabetes may also be mediated by plasma factors.94 Cell-based assays may thus prove instrumental for biomarker discovery. However, the use of cell-based assays as a direct biomarker in the clinic may be very daunting. This is due to the inherent drawbacks when working with living cells, such as standardization and reproducibility. As a consequence, the value of such assays will probably be confined to a discovery platform for new biomarkers and testing of drug efficacy in clinical trials. When the existence of unknown plasma factors that correlate with disease state is observed, an effort can be made to identify the specific factors that cause the observed effects, so they can be accurately measured using existing clinical laboratory techniques.
Plaques As we zoom in on the source of the acute cardiovascular events that are associated with atherosclerotic disease, we finally end up in the plaque itself. The most appreciated vascular pathological determinants of acute manifestations are encompassed in the definition of the vulnerable plaque: a large lipid-core and a thin overlying fibrotic cap with inflammation. So how predictive is this local plaque composition, when assessed by histological examination? From post-mortem observations we have learned that large lipid-core and cap inflammation are frequently observed phenomena in non-ruptured atherosclerotic lesions throughout the circulation.95 This suggests that the positive predictive value of the determinants of the so-called “vulnerable plaque” (large lipid-rich core, thin-cap with inflammation) for the occurrence of future event is rather limited. The predictive value of the pathological features of the vulnerable plaque has been assessed in a high-risk cohort by making use of techniques to image the composition of the plaque. In that study major adverse cardiovascular events occurring during follow-up were equally attributable to recurrence at the site of culprit lesions and to non-culprit lesions.96 Indeed, most were thin-cap fibroatheromas as determined by gray-scale and radiofrequency intravascular ultrasonography. However, less than 5% of the lesions with a thin-capped fibroatheroma were associated with adverse outcome during follow up. This emphasizes the low positive predictive value of the vulnerable plaque phenotype. Apparently there are also other determinants, which may vary between plaques, and may cause a plaque to destabilize and rupture. Still, for the last two decades the concept of the vulnerable plaque has dominated the research field in atherosclerotic disease. Current concepts exploring the mechanisms that contribute to the initiation and progression of atherosclerotic disease have been based on this vulnerable plaque concept including studies in genetically modified animal models where plaque phenotypes were used as a surrogate for atherosclerotic disease progression and cardiovascular events.
20162300 Marten Siemelink.indd 29 15-03-16 13:23 30 | CHAPTER 2
The concept of vulnerable plaque and its role in the pathogenesis of atherosclerotic disease progression have also inspired the biomarker discovery field. It is evident that most of the proposed circulating biomarkers, such as those reflecting inflammatory or proteolytic activity, represent current concepts of plaque progression and rupture. Cross-sectional pathology studies of atherosclerotic tissue generated ground-breaking hypotheses regarding both the mechanisms of initiation of atherogenesis and of disease progression. However, for prediction studies follow up is required. Atherosclerosis is a systemic disease and plaque composition in different vascular beds display many common features. Therefore, one could hypothesize that local plaque characteristics observed at a particular time-point might reflect the determinants of plaque progression in other territories of the vascular tree. In our group we hypothesized that atherosclerotic plaques obtained during vascular surgery might contain predictive molecular biomarkers for future cardiovascular events in other vascular territories. The Athero-Express is an on-going Biobank cohort study with a longitudinal study design, initiated in 2002. In the Athero-Express study, we have now correlated characteristics of the atherosclerotic plaque, which was removed during vascular surgery, with secondary cardiovascular manifestations during a three-year follow-up in a group of more than 2800 patients. Previously we showed that a stable fibrous plaque composition is an independent predictor of restenosis after carotid endarterectomy.97 This was the first demonstration that the characteristics of an atherosclerotic plaque, obtained at the time of vascular surgery, could provide predictive information about the outcome of the procedure. Next, we searched for plaque characteristics that were predictive for major cardiovascular events in all vascular territories. A proteomic analyses and subsequent validation studies revealed that histological and protein markers detected in local plaques could have strong predictive value for the occurrence of major cardiovascular events independent of traditional risk factors. For instance, studies of atherosclerotic plaques from patients with and without secondary cardiovascular manifestations, revealed that high plaque osteopontin (OPN) levels were associated with a 4 fold increased risk for secondary cardiovascular events.98 Other plaque protein markers that were predictive for events were fatty acid binding protein-4 (FABP4) and matrix metalloprotein-8 (MMP-8).99,100 A feature that was observed in histology slices from a single lesion and that also showed predictive for adverse events was intraplaque haemorrhage. The presence of intraplaque haemorrhage in a single carotid culprit lesion was associated with a twofold increase in risk of secondary manifestations of atherosclerotic disease during follow up;101 an observation that has been supported by similar observations in Magnetic-Resonance Imaging (MRI) studies.102 Unfortunately the protein markers can be measured only after surgical dissection of atherosclerotic tissue. However, non-invasive MRI is able to detect intraplaque haemorrhage and can be applied to study the predictive power of this plaque characteristic in longitudinal studies. Therefore, the observation that one small dissected plaque cross-section reveals information regarding the risk of secondary manifestations at 3 year follow up opens up potential new avenues for research.
Genetics The delivery of the complete sequence of the haploid human genome ushered in an unprecedented era of uncovering genetic variation and characterizing their impact on human disease.103–115 To date more than 16 million single-nucleotide polymorphisms (SNPs),
20162300 Marten Siemelink.indd 30 15-03-16 13:23 ADVANCES IN BIOMARKER RESEARCH | 31
insertions, deletions and other genetic variants are known.108 SNPs facilitate GWAS and help to understand the role of molecular genetic variation in human pathophysiology.116–119 Indeed GWAS have proven paramount in improving our comprehension of genetic risk prediction. GWAS identify genetic loci of interest containing putative biomarkers of disease and efficacy, and novel drug targets. Lastly SNPs can be used to infer causality of a putative biomarker. 2 Essentially all traditional cardiovascular risk factors, intermediate disease phenotypes, and the two most prevalent cardiovascular diseases have been investigated through meta- analyses of GWAS in thousands of individuals. Together these meta-analyses uncovered hundreds of associated SNPs marking previously unrelated loci, while confirming many others.120–148 These studies are summarized in Table 3. Still, some argue that GWAS failed, because none of the analyses fully accounted for the estimated genetic heritability and few of the individual SNPs have clinical impact.149 Weighted polygenic risk scores (PRS) which account for effect size and the number of effect alleles, include many risk modifying SNPs, but fail to substantially add to existing cardiovascular risk models.150 Clinically relevant predictive values could only be reached in silico by including hundreds of SNPs in such PRS.151 Still, PRS are consistently associated with the investigated phenotypes and the success of fitting genetic variations into existing prediction models is only limited by the current statistical methodology. Recently, Bayesian polygenic risk score modelling for various diseases, including myocardial infarction, revealed that potentially thousands of variants await discovery which might substantially add to existing risk models.152 Additionally, a Bayesian approach which accounts for samples size and minor allele frequencies to set the threshold for genome–wide significance, might substantially increase the positive findings in GWAS.153,154 These studies point out that innovative statistical methodology can help uncover many more SNPs associated with CVD and its risk factors, which taken together in PRS might add clinically relevant predictive value to existing risk models.
Genetics in biomarker and drug research Despite the limited clinical impact on predictive modelling, GWAS have confirmed some loci that are of clear clinical importance. For instance, NPC1L1 is a known lipid-lowering drug target and was associated with total cholesterol levels using GWAS.120 Furthermore, statins are among the most widely marketed lipid-lowering drugs, and inhibit hydroxyl-3- methylglutaryl coenzyme A reductase, encoded by HMGCR. The variants associated with this gene were confirmed to be associated with total cholesterol.120 In this way, the confirmation of drug targets by GWAS provides important evidence for the clinical relevance of the drug target. GWAS can also provide important arguments to pursue new drug targets. For example, GWAS showed that genetic variation in proprotein convertase subtilisin/kexin type 9 (PCSK9) is strongly associated with total and LDL-cholesterol levels and mildly with CAD.120,138 This supported the screening of compounds that modify PCSK9 effects. Inhibition of PCSK9 through a monoclonal antibody has additive effect on top of the use of statins in primary hypercholesterolemia and heterozygous familial hypercholesterolemia; it is expected to be marketed within the coming months.155–159 In biomarker research, GWAS have recently proven very influential. Undoubtedly the ideal biomarker is causally influencing the disease while unaffected by any confounders, and
20162300 Marten Siemelink.indd 31 15-03-16 13:23 32 | CHAPTER 2
Table 3. Genome-wide association studies for atherosclerotic disease and risk factors Cohort size Cardiovascular disease Population Loci References range CAD European, South-Asian 7,593-70,595 16 142,198–201 European 19,036-187,667 28 138,202–205 Stroke European 56,075-95,688 3 146,206,207 Risk factor: disease Type 2 diabetes European 71,562-141,454 29 127,128 CKD European 84,740 13 124–126 Hypertension European 91,164-174,347 11 129,130,208,209 Risk factor: quantitative trait Lipid levels 120,121 Total cholesterol European 62,935-100,184 59 LDL European 51,968-95,454 42 HDL European 63,754-99,900 52 Triglycerides European 61,214-96,598 38 Renal function & secretion 124–126 eGFR - creatinine European 90,075 25 eGFR - cystatin European 26,071 4 Blood pressure 129,130,209 208 SBP European 96,425-201,745 25 DBP European 96,406-201,709 26 PP European 105,280-114,758 10 MAP European 73,884-123,059 23 CRP European (82,725 18 210 BMI European 249,796 32 122 Smoking 211 CPD European 72,956-73,853 3 Smoking initiation European 143,023 1 Smoking cessation European 64,924 1 Intermediate Quantitative phenotype Coronary artery calcification European 15,993 14 212 Carotid cIMT European 15,020-41,295 4 148,213 Ankle-brachial index European 31,035-51,708 5 214 Intermediate Binary phenotype Carotid plaque presence European 29,255-29,827 2 212 CAD: Coronary artery disease; including myocardial infarction, percutaneous coronary intervention, coro- nary artery bypass-graft. Stroke: large artery ischemic stroke. Type 2 diabetes: based on diagnosis or medi- cation use. Hypertension: based on diagnosis, a systolic blood pressure > 145 mmHg, or medication use. CKD: chronic kidney disease (KDOQI class≥2). eGFR - creatitine: estimated glomerular filtration rate based on serum creatinine levels, mL/min/1.73m2. eGFR cystatin: estimated glomerular filtration rate based on serum creatinine levels, mL/min/1.73m2. SBP: systolic blood pressure, mmHg. DBP: diastolic blood pres- sure, mmHg. PP: pulse pressure, mmHg. MAP: mean arterial pressure, mmHg. BMI: body-mass index, kg/m2. CRP: C-reactive protein plasma levels, μg/mL. Common cIMT: common carotid intima-media thick- ness. Carotid plaque presence: carotid stenosis >25% (cases) vs. carotid stenosis ≤25% (controls). CPD: number of cigarettes smoked per day. Smoking initiation: ever vs. never been a regular smoker. Smoking cessation: former vs. current smoker. This is not a complete list of GWAS publications; we refer to http:// www.genome.gov/gwasstudies for more GWAS results. European populations include all populations of European descent. South-Asian populations include Japanese and Chinese populations.
20162300 Marten Siemelink.indd 32 15-03-16 13:23 ADVANCES IN BIOMARKER RESEARCH | 33
thereby the ideal drug target candidate. Yet, of many putative biomarkers, it is unknown whether they are causally related to the disease or just an epiphenomenon. Obviously a double-blind randomized controlled trial involving knocking out genes to test causality is not an option in human test subjects. Importantly, GWAS can enable Mendelian Randomization studies to infer causality of putative biomarker in the presence of possible confounders.160–162 As paternal and maternal 2 alleles are randomly assigned when passed to offspring during meiosis, we can assume the population distribution of these alleles is not confounded. The Mendelian Randomization method can be viewed as a “natural” randomized controlled trial, and large well-designed GWAS (N>30,000) provide good estimates of effect crucial for Mendelian Randomization studies. Three recent Mendelian Randomization studies (N>90,000) have disavowed the likelihood of a causal role for C-reactive protein (CRP) and HDL in cardiovascular disease.9,163–165 Zacho et al. genotyped 4 variants in CRP and could reliably estimate the effect on CRP plasma levels.9,163 Moreover, the expected effect of CRP plasma levels on ischemic heart disease (IHD) was observed. In contrast, the observed association between CRP variants and IHD was markedly different from the predicted effects. Another study genotyped 5 SNPs in loci associated with CRP levels, and also failed to show the same proportional effect between CRP genotypes, levels and coronary heart disease.164 Similarly, HDL is commonly viewed as a beneficial lipoprotein particle, however, a study by Voight et al. has given rise to serious doubts on the validity of this view.165 Variants in genetic loci known to modify HDL plasma levels performed as expected, but did not show a unified association with a decrease in MI risk.165
Discussion
As healthcare providers around the globe try to cut back on costs while improving patient care, better information on patient disease risk and disease state is indispensable. Therefore, the importance of biomarkers in cardiovascular disease is likely to increase in the near future. The improvement of sensitive –omics techniques enables the discovery of many new potential biomarkers. This combines well with the investigation of new biosources to open up new hunting grounds to scour for biomarkers. This has already provided some exciting new biomarkers, which will have to be validated in larger cohorts in the coming years. The advent of next-generation genetic research provides the ability to distinguish individual genetic risk factors from acquired risk factors with unprecedented accuracy. Furthermore genetic associations provide a powerful instrument to determine the causality of a biomarker in the pathophysiological processes of complex diseases. This is key to our understanding of the pathophysiology and to determine the value of a biomarker. Data integration from different sources may prove crucial in gaining an improved understanding of the pathophysiological mechanisms of atherosclerotic disease. The interplay between hereditary mechanisms, early-life imprinting, epigenetic mechanisms and current risk factors may prove decisive in the eventual occurrence of an acute event. Getting a grip on that interplay may take us to the next level of risk prediction and eventual risk reduction.
20162300 Marten Siemelink.indd 33 15-03-16 13:23 34 | CHAPTER 2
Finally, these developments will put risk prediction on the much anticipated track towards personal medicine. When all the factors that influence disease pathophysiology are recognized and accurately determined, reliable personal risk prediction comes within reach. This will also offer many opportunities for new drug development, based on increased understanding of the pathophysiology.
20162300 Marten Siemelink.indd 34 15-03-16 13:23 ADVANCES IN BIOMARKER RESEARCH | 35
References
1 Ge Y, Wang TJ. Circulating, imaging, and genetic biomarkers in cardiovascular risk prediction. Trends Cardiovasc Med 2011; 21: 105–12. 2 Pikula A, Beiser AS, DeCarli C, et al. Multiple biomarkers and risk of clinical and subclinical vascular brain injury: the Framingham Offspring Study. Circulation 2012; 125: 2100–7. 3 Wang TJ. Assessing the role of circulating, genetic, and imaging biomarkers in cardiovascular risk 2 prediction. Circulation 2011; 123: 551–65. 4 Wang TJ, Gona P, Larson MG, et al. Multiple biomarkers for the prediction of first major cardiovascular events and death. N Engl J Med 2006; 355: 2631–9. 5 Buckley DI, Fu R, Freeman M, Rogers K, Helfand M. C-reactive protein as a risk factor for coronary heart disease: a systematic review and meta-analyses for the U.S. Preventive Services Task Force. Ann Intern Med 2009; 151: 483–95. 6 Loscalzo J. Association studies in an era of too much information: clinical analysis of new biomarker and genetic data. Circulation 2007; 116: 1866–70. 7 Quillard T, Libby P. Molecular imaging of atherosclerosis for improving diagnostic and therapeutic de- velopment. Circ Res 2012; 111: 231–44. 8 Cole JE, Mitra AT, Monaco C. Treating atherosclerosis: the potential of Toll-like receptors as therapeu- tic targets. Expert Rev Cardiovasc Ther 2010; 8: 1619–35. 9 Schunkert H, Samani NJ. Elevated C-reactive protein in atherosclerosis--chicken or egg? N Engl J Med 2008; 359: 1953–5. 10 Jaumdally R, Varma C, Macfadyen RJ, Lip GYH. Coronary sinus blood sampling: an insight into local cardiac pathophysiology and treatment? Eur Heart J 2007; 28: 929–40. 11 Avgerinos ED, Kadoglou NPE, Moulakakis KG, Giannakopoulos TG, Liapis CD. Current role of biomark- ers in carotid disease: a systematic review. Int J Stroke 2011; 6: 337–45. 12 Battistoni A, Rubattu S, Volpe M. Circulating biomarkers with preventive, diagnostic and prognostic implications in cardiovascular diseases. Int J Cardiol 2012; 157: 160–8. 13 Surinova S, Schiess R, Hüttenhain R, Cerciello F, Wollscheid B, Aebersold R. On the development of plasma protein biomarkers. J Proteome Res 2011; 10: 5–16. 14 Mebazaa A, Vanpoucke G, Thomas G, et al. Unbiased plasma proteomics for novel diagnostic bio- markers in cardiovascular disease: identification of quiescin Q6 as a candidate biomarker of acutely decompensated heart failure. Eur Heart J 2012; 33: 2317–24. 15 Théry C, Ostrowski M, Segura E. Membrane vesicles as conveyors of immune responses. Nat Rev Immunol 2009; 9: 581–93. 16 Mause SF, Weber C. Microparticles: protagonists of a novel communication network for intercellular information exchange. Circ Res 2010; 107: 1047–57. 17 Simons M, Raposo G. Exosomes--vesicular carriers for intercellular communication. Curr Opin Cell Biol 2009; 21: 575–81. 18 Camussi G, Deregibus MC, Bruno S, Cantaluppi V, Biancone L. Exosomes/microvesicles as a mecha- nism of cell-to-cell communication. Kidney Int 2010; 78: 838–48. 19 Hron G, Kollars M, Weber H, et al. Tissue factor-positive microparticles : Cellular origin and association with coagulation activation in patients with colorectal cancer. 2007; : 119–23. 20 Taylor DD, Gercel-Taylor C. MicroRNA signatures of tumor-derived exosomes as diagnostic biomarkers of ovarian cancer. Gynecol Oncol 2008; 110: 13–21. 21 Davila M, Amirkhosravi a, Coll E, et al. Tissue factor-bearing microparticles derived from tumor cells: impact on coagulation activation. J Thromb Haemost 2008; 6: 1517–24. 22 Mack M, Kleinschmidt a, Brühl H, et al. Transfer of the chemokine receptor CCR5 between cells by membrane-derived microparticles: a mechanism for cellular human immunodeficiency virus 1 infec- tion. Nat Med 2000; 6: 769–75. 23 Rozmyslowicz T, Majka M, Kijowski J, et al. Platelet- and megakaryocyte-derived microparticles trans- fer CXCR4 receptor to CXCR4-null cells and make them susceptible to infection by X4-HIV. AIDS 2003; 17: 33–42. 24 Fevrier B, Vilette D, Archer F, et al. Cells release prions in association with exosomes. Proc Natl Acad Sci U S A 2004; 101: 9683–8. 25 Robertson C, Booth S a, Beniac DR, Coulthart MB, Booth TF, McNicol A. Cellular prion protein is re- leased on exosomes from activated platelets. Blood 2006; 107: 3907–11. 26 Mallat Z, Hugel B, Ohan J, Leseche G, Freyssinet J-M, Tedgui a. Shed Membrane Microparticles With Procoagulant Potential in Human Atherosclerotic Plaques : A Role for Apoptosis in Plaque Thrombo- genicity. Circulation 1999; 99: 348–53.
20162300 Marten Siemelink.indd 35 15-03-16 13:23 36 | CHAPTER 2
27 Leroyer AS, Isobe H, Lesèche G, et al. Cellular origins and thrombogenic activity of microparticles isolated from human atherosclerotic plaques. J Am Coll Cardiol 2007; 49: 772–7. 28 Rautou P-E, Vion A-C, Amabile N, et al. Microparticles, vascular function, and atherothrombosis. Circ Res 2011; 109: 593–606. 29 Amabile N, Guérin AP, Leroyer A, et al. Circulating endothelial microparticles are associated with vas- cular dysfunction in patients with end-stage renal failure. J Am Soc Nephrol 2005; 16: 3381–8. 30 Densmore JC, Signorino PR, Ou J, et al. Endothelium-derived microparticles induce endothelial dys- function and acute lung injury. Shock 2006; 26: 464–71. 31 Brodsky S V, Zhang F, Nasjletti A, Goligorsky MS. Endothelium-derived microparticles impair endothe- lial function in vitro. Am J Physiol Heart Circ Physiol 2004; 286: H1910–5. 32 Agouni A, Lagrue-Lak-Hal AH, Ducluzeau PH, et al. Endothelial dysfunction caused by circulating mi- croparticles from patients with metabolic syndrome. Am J Pathol 2008; 173: 1210–9. 33 Barry OP, Pratico D, Lawson J a, FitzGerald G a. Transcellular activation of platelets and endothelial cells by bioactive lipids in platelet microparticles. J Clin Invest 1997; 99: 2118–27. 34 Barry OP, Praticò D, Savani RC, FitzGerald G a. Modulation of monocyte-endothelial cell interactions by platelet microparticles. J Clin Invest 1998; 102: 136–44. 35 Mesri M, Altieri DC. Endothelial cell activation by leukocyte microparticles. J Immunol 1998; 161: 4382–7. 36 Scanu A, Molnarfi N, Brandt KJ, Gruaz L, Dayer J-M, Burger D. Stimulated T cells generate micropar- ticles, which mimic cellular contact activation of human monocytes: differential regulation of pro- and anti-inflammatory cytokine production by high-density lipoproteins. J Leukoc Biol 2008; 83: 921–7. 37 Rautou P-E, Leroyer AS, Ramkhelawon B, et al. Microparticles from human atherosclerotic plaques promote endothelial ICAM-1-dependent monocyte adhesion and transendothelial migration. Circ Res 2011; 108: 335–43. 38 Brill A, Dashevsky O, Rivo J, Gozal Y, Varon D. Platelet-derived microparticles induce angiogenesis and stimulate post-ischemic revascularization. Cardiovasc Res 2005; 67: 30–8. 39 Chou J, Mackman N, Merrill-Skoloff G, Pedersen B, Furie BC, Furie B. Hematopoietic cell-derived mi- croparticle tissue factor contributes to fibrin formation during thrombus propagation. Blood 2004; 104: 3190–7. 40 Gasser O, Schifferli J a. Activated polymorphonuclear neutrophils disseminate anti-inflammatory mi- croparticles by ectocytosis. Blood 2004; 104: 2543–8. 41 Yang C, Mwaikambo BR, Zhu T, et al. Lymphocytic microparticles inhibit angiogenesis by stimulating oxidative stress and negatively regulating VEGF-induced pathways. Am J Physiol Regul Integr Comp Physiol 2008; 294: R467–76. 42 Raimondo F, Morosi L, Chinello C, Magni F, Pitto M. Advances in membranous vesicle and exosome proteomics improving biological understanding and biomarker discovery. Proteomics 2011; 11: 709– 20. 43 Simak J, Gelderman MP, Yu H, Wright V, Baird a E. Circulating endothelial microparticles in acute ischemic stroke: a link to severity, lesion volume and outcome. J Thromb Haemost 2006; 4: 1296–302. 44 Peterson DB, Sander T, Kaul S, et al. Comparative proteomic analysis of PAI-1 and TNF-alpha-derived endothelial microparticles. Proteomics 2008; 8: 2430–46. 45 Mitchell PS, Parkin RK, Kroh EM, et al. Circulating microRNAs as stable blood-based markers for can- cer detection. Proc Natl Acad Sci U S A 2008; 105: 10513–8. 46 Chen X, Ba Y, Ma L, et al. Characterization of microRNAs in serum: a novel class of biomarkers for diagnosis of cancer and other diseases. Cell Res 2008; 18: 997–1006. 47 Zhang Y, Liu D, Chen X, et al. Secreted monocytic miR-150 enhances targeted endothelial cell migra- tion. Mol Cell 2010; 39: 133–44. 48 Zernecke A, Bidzhekov K, Noels H, et al. Delivery of microRNA-126 by apoptotic bodies induces CX- CL12-dependent vascular protection. Sci Signal 2009; 2: ra81. 49 Gamazon ER, Ziliak D, Im HK, et al. Genetic architecture of microRNA expression: implications for the transcriptome and complex traits. Am J Hum Genet 2012; 90: 1046–63. 50 Gosling J, Slaymaker S, Gu L, et al. MCP-1 deficiency reduces susceptibility to atherosclerosis in mice that overexpress human apolipoprotein B. J Clin Invest 1999; 103: 773–8. 51 Nakashima Y, Raines EW, Plump a. S, Breslow JL, Ross R. Upregulation of VCAM-1 and ICAM-1 at Atherosclerosis-Prone Sites on the Endothelium in the ApoE-Deficient Mouse. Arterioscler Thromb Vasc Biol 1998; 18: 842–51. 52 Qiao JH, Tripathi J, Mishra NK, et al. Role of macrophage colony-stimulating factor in atherosclerosis: studies of osteopetrotic mice. Am J Pathol 1997; 150: 1687–99. 53 Emeson EE, Shen ML, Bell CG, Qureshi a. Inhibition of atherosclerosis in CD4 T-cell-ablated and nude (nu/nu) C57BL/6 hyperlipidemic mice. Am J Pathol 1996; 149: 675–85.
20162300 Marten Siemelink.indd 36 15-03-16 13:23 ADVANCES IN BIOMARKER RESEARCH | 37
54 Zhou X, Nicoletti a., Elhage R, Hansson GK. Transfer of CD4+ T Cells Aggravates Atherosclerosis in Immunodeficient Apolipoprotein E Knockout Mice. Circulation 2000; 102: 2919–22. 55 Zhou X, Stemme S, Hansson GK. Short communication. Int J Lang Commun Disord 1996; 149: 359– 66. 56 Asahara T. Isolation of Putative Progenitor Endothelial Cells for Angiogenesis. Science (80- ) 1997; 275: 964–6. 57 Kuroi A, Imanishi T, Suzuki H, et al. Clinical Characteristics of Patients With Kawasaki Disease and Levels of Peripheral Endothelial Progenitor Cells and Blood Monocyte Subpopulations. Circ J 2010; 74: 2 2720–5. 58 Ludwig A, Jochmann N, Kertesz A, et al. Smoking decreases the level of circulating CD34+ progenitor cells in young healthy women--a pilot study. BMC Womens Health 2010; 10: 20. 59 Schmidt-Lucke C, Rössig L, Fichtlscherer S, et al. Reduced number of circulating endothelial progeni- tor cells predicts future cardiovascular events: proof of concept for the clinical importance of endog- enous vascular repair. Circulation 2005; 111: 2981–7. 60 van Oostrom O, Nieuwdorp M, Westerweel PE, et al. Reconstituted HDL increases circulating en- dothelial progenitor cells in patients with type 2 diabetes. Arterioscler Thromb Vasc Biol 2007; 27: 1864–5. 61 Werner N, Kosiol S, Schiegl T, et al. Circulating endothelial progenitor cells and cardiovascular out- comes. N Engl J Med 2005; 353: 999–1007. 62 Westerweel PE, Hoefer IE, Blankestijn PJ, et al. End-stage renal disease causes an imbalance be- tween endothelial and smooth muscle progenitor cells. Am J Physiol Renal Physiol 2007; 292: F1132– 40. 63 Westerweel PE, Luijten RKM a C, Hoefer IE, Koomans H a, Derksen RHWM, Verhaar MC. Haemat- opoietic and endothelial progenitor cells are deficient in quiescent systemic lupus erythematosus. Ann Rheum Dis 2007; 66: 865–70. 64 Westerweel PE, Visseren FLJ, Hajer GR, et al. Endothelial progenitor cell levels in obese men with the metabolic syndrome and the effect of simvastatin monotherapy vs. simvastatin/ezetimibe combina- tion therapy. Eur Heart J 2008; 29: 2808–17. 65 Heine GH, Ortiz A, Massy ZA, et al. Monocyte subpopulations and cardiovascular risk in chronic kidney disease. Nat Rev Nephrol 2012; 8: 362–9. 66 Rogacev KS, Cremers B, Zawada AM, et al. CD14++CD16+ Monocytes Independently Predict Car- diovascular Events: A Cohort Study of 951 Patients Referred for Elective Coronary Angiography. J Am Coll Cardiol 2012; xx. DOI:10.1016/j.jacc.2012.07.019. 67 Barisione C, Garibaldi S, Ghigliotti G, et al. CD14CD16 monocyte subset levels in heart failure patients. Dis Markers 2010; 28: 115–24. 68 Ulrich C, Heine GH, Gerhart MK, Köhler H, Girndt M. Proinflammatory CD14+CD16+ monocytes are associated with subclinical atherosclerosis in renal transplant patients. Am J Transplant 2008; 8: 103– 10. 69 Schipper HS, Nuboer R, Prop S, et al. Systemic inflammation in childhood obesity: circulating inflam- matory mediators and activated CD14(++) monocytes. Diabetologia 2012; 55: 2800–10. 70 Ulrich C, Heine GH, Garcia P, et al. Increased expression of monocytic angiotensin-converting enzyme in dialysis patients with cardiovascular disease. Nephrol Dial Transplant 2006; 21: 1596–602. 71 Ulrich C, Heine GH, Seibert E, Fliser D, Girndt M. Circulating monocyte subpopulations with high ex- pression of angiotensin-converting enzyme predict mortality in patients with end-stage renal disease. Nephrol Dial Transplant 2010; 25: 2265–72. 72 Betjes MGH, de Wit EE a, Weimar W, Litjens NHR. Circulating pro-inflammatory CD4posCD28null T cells are independently associated with cardiovascular disease in ESRD patients. Nephrol Dial Trans- plant 2010; 25: 3640–6. 73 Alber HF, Duftner C, Wanitschek M, et al. Neopterin, CD4+CD28− lymphocytes and the extent and severity of coronary artery disease. Int J Cardiol 2009; 135: 27–35. 74 Dumitriu IE, Araguás ET, Baboonian C, Kaski JC. CD4+ CD28 null T cells in coronary artery disease: when helpers become killers. Cardiovasc Res 2009; 81: 11–9. 75 Ammirati E, Cianflone D, Banfi M, et al. Circulating CD4+CD25hiCD127lo regulatory T-Cell levels do not reflect the extent or severity of carotid and coronary atherosclerosis. Arterioscler Thromb Vasc Biol 2010; 30: 1832–41. 76 Breet NJ, van Werkum JW, Bouman HJ, et al. Comparison of Platelet Function Tests Undergoing Coronary Stent Implantation. JAMA 2010; 303. 77 Edfeldt K, Swedenborg J, Hansson GK, Yan Z. Expression of toll-like receptors in human atheroscle- rotic lesions: a possible pathway for plaque activation. Circulation 2002; 105: 1158–61. 78 Hamann L, Gomma A, Schröder NWJ, et al. A frequent toll-like receptor (TLR)-2 polymorphism is a risk factor for coronary restenosis. J Mol Med (Berl) 2005; 83: 478–85.
20162300 Marten Siemelink.indd 37 15-03-16 13:23 38 | CHAPTER 2
79 Tobias PS, Curtiss LK. Toll-like receptors in atherosclerosis. Biochem Soc Trans 2007; 35: 1453–5. 80 Xu XH, Shah PK, Faure E, et al. Toll-Like Receptor-4 Is Expressed by Macrophages in Murine and Human Lipid-Rich Atherosclerotic Plaques and Upregulated by Oxidized LDL. Circulation 2001; 104: 3103–8. 81 Ishikawa Y, Satoh M, Itoh T, Minami Y, Takahashi Y, Akamura M. Local expression of Toll-like receptor 4 at the site of ruptured plaques in patients with acute myocardial infarction. Clin Sci (Lond) 2008; 115: 133–40. 82 Liuzzo G, Angiolillo DJ, Buffon a., et al. Enhanced Response of Blood Monocytes to In Vitro Lipopoly- saccharide-Challenge in Patients With Recurrent Unstable Angina. Circulation 2001; 103: 2236–41. 83 Methe H, Kim J-O, Kofler S, Weis M, Nabauer M, Koglin J. Expansion of circulating Toll-like receptor 4-positive monocytes in patients with acute coronary syndrome. Circulation 2005; 111: 2654–61. 84 Versteeg D, Hoefer IE, Schoneveld a H, et al. Monocyte toll-like receptor 2 and 4 responses and ex- pression following percutaneous coronary intervention: association with lesion stenosis and fractional flow reserve. Heart 2008; 94: 770–6. 85 Kuwahata S, Fujita S, Orihara K, et al. High expression level of Toll-like receptor 2 on monocytes is an important risk factor for arteriosclerotic disease. Atherosclerosis 2010; 209: 248–54. 86 Lorenzen JM, David S, Richter A, et al. TLR-4+ peripheral blood monocytes and cardiovascular events in patients with chronic kidney disease--a prospective follow-up study. Nephrol Dial Transplant 2011; 26: 1421–4. 87 Hemker HC, Dieri R Al, Smedt E De, Béguin S. Thrombin generation , a function test of the haemo- static- thrombotic system. 2006; : 553–61. 88 Rand ML, Leung R, Packham M a. Platelet function assays. Transfus Apher Sci 2003; 28: 307–17. 89 Pajunen P, Syvänne M, Castro G, Nieminen MS, Taskinen MR. Cholesterol efflux capacity in vitro pre- dicts the severity and extent of coronary artery disease in patients with and without type 2 diabetes. Scand Cardiovasc J 2001; 35: 96–100. 90 Chirinos J a, Zambrano JP, Chakko S, et al. Ability of serum to decrease cellular acylCoA:cholesterol acyl transferase activity predicts cardiovascular outcomes. Circulation 2005; 112: 2446–53. 91 Khera A V, Cuchel M, de la Llera-Moya M, et al. Cholesterol efflux capacity, high-density lipoprotein function, and atherosclerosis. N Engl J Med 2011; 364: 127–35. 92 Zhou H, Shiu SWM, Wong Y, Tan KCB. Impaired serum capacity to induce cholesterol efflux is associ- ated with endothelial dysfunction in type 2 diabetes mellitus. Diab Vasc Dis Res 2009; 6: 238–43. 93 Reiss AB, Wan DW, Anwar K, et al. Enhanced CD36 scavenger receptor expression in THP-1 human monocytes in the presence of lupus plasma: linking autoimmunity and atherosclerosis. Exp Biol Med (Maywood) 2009; 234: 354–60. 94 Verrijn Stuart a a, Schipper HS, Tasdelen I, et al. Altered plasma adipokine levels and in vitro adipocyte differentiation in pediatric type 1 diabetes. J Clin Endocrinol Metab 2012; 97: 463–72. 95 Pasterkamp G, Schoneveld AH, van der Wal AC, et al. Inflammation of the atherosclerotic cap and shoulder of the plaque is a common and locally observed feature in unruptured plaques of femoral and coronary arteries. Arterioscler Thromb Vasc Biol 1999; 19: 54–8. 96 Stone GW, Maehara A, Lansky AJ, et al. A prospective natural-history study of coronary atherosclero- sis. N Engl J Med 2011; 364: 226–35. 97 Hellings WE, Moll FL, De Vries J-PPM, et al. Atherosclerotic plaque composition and occurrence of restenosis after carotid endarterectomy. JAMA 2008; 299: 547–54. 98 de Kleijn DP V, Moll FL, Hellings WE, et al. Local atherosclerotic plaques are a source of prognostic biomarkers for adverse cardiovascular events. Arterioscler Thromb Vasc Biol 2010; 30: 612–9. 99 Peeters W, de Kleijn DP V, Vink A, et al. Adipocyte fatty acid binding protein in atherosclerotic plaques is associated with local vulnerability and is predictive for the occurrence of adverse cardiovascular events. Eur Heart J 2011; 32: 1758–68. 100 Peeters W, Moll FL, Vink A, et al. Collagenase matrix metalloproteinase-8 expressed in atherosclerotic carotid plaques is associated with systemic cardiovascular outcome. Eur Heart J 2011; 32: 2314–25. 101 Hellings WE, Peeters W, Moll FL, et al. Composition of carotid atherosclerotic plaque is associated with cardiovascular outcome: a prognostic study. Circulation 2010; 121: 1941–50. 102 Sun J, Underhill HR, Hippe DS, Xue Y, Yuan C, Hatsukami TS. Sustained acceleration in carotid ath- erosclerotic plaque progression with intraplaque hemorrhage: a long-term time course study. JACC Cardiovasc Imaging 2012; 5: 798–804. 103 Stein LD. Human Genome: End of the beginning. Nature 2004; 431: 915–6. 104 International Human Genome Sequencing Consortium. Initial sequencing and analysis of the human genome. Nature 2001; 409: 860–921. 105 International Human Genome Sequencing Consortium. Finishing the euchromatic sequence of the human genome. Nature 2004; 431: 931–45. 106 Venter JC, Adams MD, Myers EW, et al. The sequence of the human genome. Science (80- ) 2001; 291: 1304–51.
20162300 Marten Siemelink.indd 38 15-03-16 13:23 ADVANCES IN BIOMARKER RESEARCH | 39
107 Wheeler D a, Srinivasan M, Egholm M, et al. The complete genome of an individual by massively paral- lel DNA sequencing. Nature 2008; 452: 872–6. 108 The 1000 Genomes Project Consortium. A map of human genome variation from population-scale sequencing. Nature 2010; 467: 1061–73. 109 The International HapMap 3 Consortium. Integrating common and rare genetic variation in diverse hu- man populations. Nature 2010; 467: 52–8. 110 Frazer K a, Ballinger DG, Cox DR, et al. A second generation human haplotype map of over 3.1 million SNPs. Nature 2007; 449: 851–61. 2 111 The International SNP Map Working Group. A map of human genome sequence variation containing 1.42 million single nucleotide polymorphisms. Nature 2001; 409: 928–33. 112 The International Human Genome Mapping Consortium. A physical map of the human genome. Na- ture 2001; 409: 934–41. 113 The International HapMap Consortium. A haplotype map of the human genome. Nature 2005; 437: 1299–320. 114 Mills RE, Walter K, Stewart C, et al. Mapping copy number variation by population-scale genome se- quencing. Nature 2011; 470: 59–65. 115 Levy S, Sutton G, Ng PC, et al. The Diploid Genome Sequence of an Individual Human. PLoS Biol 2007; 5: e254. 116 Mullen S a, Crompton DE, Carney PW, Helbig I, Berkovic SF. A neurologist’s guide to genome-wide association studies. Neurology 2009; 72: 558–65. 117 Pearson T a, Manolio T a. How to interpret a genome-wide association study. JAMA 2008; 299: 1335– 44. 118 Manolio TA. Genomewide association studies and assessment of risk of disease. N Engl J Med 2010; 363: 2076–7; author reply 2077. 119 McCarthy MI, Abecasis GR, Cardon LR, et al. Genome-wide association studies for complex traits: consensus, uncertainty and challenges. Nat Rev Genet 2008; 9: 356–69. 120 Teslovich TM, Musunuru K, Smith A V, et al. Biological, clinical and population relevance of 95 loci for blood lipids. Nature 2010; 466: 707–13. 121 Asselbergs FW, Guo Y, van Iperen EPA, et al. Large-Scale Gene-Centric Meta-analysis across 32 Stud- ies Identifies Multiple Lipid Loci. Am J Hum Genet 2012; : 1–16. 122 Speliotes EK, Willer CJ, Berndt SI, et al. Association analyses of 249,796 individuals reveal 18 new loci associated with body mass index. Nat Genet 2010; 42: 937–48. 123 Kathiresan S, Willer CJ, Peloso GM, et al. Common variants at 30 loci contribute to polygenic dyslipi- demia. Nat Genet 2009; 41: 56–65. 124 Köttgen A, Glazer NL, Dehghan A, et al. Multiple loci associated with indices of renal function and chronic kidney disease. Nat Genet 2009; 41: 712–7. 125 Köttgen A, Pattaro C, Böger C a, et al. New loci associated with kidney function and chronic kidney disease. Nat Genet 2010; 42: 376–84. 126 Chambers JC, Zhang W, Lord GM, et al. Genetic loci influencing kidney function and chronic kidney disease. Nat Genet 2010; 42: 373–5. 127 Voight BF, Scott LJ, Steinthorsdottir V, et al. Twelve type 2 diabetes susceptibility loci identified through large-scale association analysis. Nat Genet 2010; 42: 579–89. 128 Saxena R, Elbers CC, Guo Y, et al. Large-Scale Gene-Centric Meta-Analysis across 39 studies Identifies Type 2 Diabetes Loci. Am J Hum Genet 2012; : 410–25. 129 Newton-Cheh C, Johnson T, Gateva V, et al. Genome-wide association study identifies eight loci as- sociated with blood pressure. Nat Genet 2009; 41: 666–76. 130 Levy D, Ehret GB, Rice K, et al. Genome-wide association study of blood pressure and hypertension. Nat Genet 2009; 41: 677–87. 131 Ehret GB, Munroe PB, Rice KM, et al. Genetic variants in novel pathways influence blood pressure and cardiovascular disease risk. Nature 2011; 478: 103–9. 132 Hindorff L a, Sethupathy P, Junkins H a, et al. Potential etiologic and functional implications of genome- wide association loci for human diseases and traits. Proc Natl Acad Sci U S A 2009; 106: 9362–7. 133 McPherson R, Pertsemlidis A, Kavaslar N, et al. A common allele on chromosome 9 associated with coronary heart disease. Science 2007; 316: 1488–91. 134 Helgadottir A, Thorleifsson G, Manolescu A, et al. A common variant on chromosome 9p21 affects the risk of myocardial infarction. Science 2007; 316: 1491–3. 135 The Wellcome Trust Case Control Consortium. Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls. Nature 2007; 447: 661–78. 136 Helgadottir A, Thorleifsson G, Magnusson KP, et al. The same sequence variant on 9p21 associates with myocardial infarction, abdominal aortic aneurysm and intracranial aneurysm. Nat Genet 2008; 40: 217–24.
20162300 Marten Siemelink.indd 39 15-03-16 13:23 40 | CHAPTER 2
137 Samani NJ, Raitakari OT, Sipilä K, et al. Coronary artery disease-associated locus on chromosome 9p21 and early markers of atherosclerosis. Arterioscler Thromb Vasc Biol 2008; 28: 1679–83. 138 Schunkert H, König IR, Kathiresan S, et al. Large-scale association analysis identifies 13 new suscep- tibility loci for coronary artery disease. Nat Genet 2011; 43: 333–8. 139 Hiura Y, Fukushima Y, Yuno M, et al. Validation of the association of genetic variants on chromosome 9p21 and 1q41 with myocardial infarction in a Japanese population. Circ J 2008; 72: 1213–7. 140 Shen G-Q, Li L, Rao S, et al. Four SNPs on chromosome 9p21 in a South Korean population implicate a genetic locus that confers high cross-race risk for development of coronary artery disease. Arterioscler Thromb Vasc Biol 2008; 28: 360–5. 141 Assimes TL, Knowles JW, Basu A, et al. Susceptibility locus for clinical and subclinical coronary artery disease at chromosome 9p21 in the multi-ethnic ADVANCE study. Hum Mol Genet 2008; 17: 2320–8. 142 The Coronary Artery Disease (C4D) Genetics Consortium. A genome-wide association study in Eu- ropeans and South Asians identifies five new loci for coronary artery disease. Nat Genet 2011; 43: 339–44. 143 Ikram M, Seshadri S, Bis J, et al. Genomewide association studies of stroke. NEJM 2009; : 1718–28. 144 International Stroke Genetics Consortium, Wellcome Trust Case-Control Consortium 2. Failure to Vali- date Association between 12p13 Variants and Ischemic Stroke. NEJM 2010; 362: 1547–50. 145 Bellenguez C, Bevan S, Gschwendtner A, et al. Genome-wide association study identifies a variant in HDAC9 associated with large vessel ischemic stroke. Nat Genet 2012; 44: 328–33. 146 Traylor M, Farrall M, Holliday EG, et al. Genetic risk factors for ischaemic stroke and its subtypes (the METASTROKE Collaboration): a meta-analysis of genome-wide association studies. Lancet Neurol 2012; : 951–62. 147 Preuss M, König IR, Thompson JR, et al. Design of the Coronary ARtery DIsease Genome-Wide Repli- cation And Meta-Analysis (CARDIoGRAM) Study: A Genome-wide association meta-analysis involving more than 22 000 cases and 60 000 controls. Circ Cardiovasc Genet 2010; 3: 475–83. 148 Bis JC, Kavousi M, Franceschini N, et al. Meta-analysis of genome-wide association studies from the CHARGE consortium identifies common variants associated with carotid intima media thickness and plaque. Nat Genet 2011; 43: 940–7. 149 Manolio TA, Collins FS, Cox NJ, et al. Finding the missing heritability of complex diseases. Nature 2009; 461: 747–53. 150 van der Net JB, Janssens ACJW, Sijbrands EJG, Steyerberg EW. Value of genetic profiling for the prediction of coronary heart disease. Am Heart J 2009; 158: 105–10. 151 Janssens ACJW, Aulchenko YS, Elefante S, Borsboom GJJM, Steyerberg EW, van Duijn CM. Predic- tive testing for complex diseases using multiple genes: fact or fiction? Genet Med 2006; 8: 395–400. 152 Stahl E a, Wegmann D, Trynka G, et al. Bayesian inference analyses of the polygenic architecture of rheumatoid arthritis. Nat Genet 2012; 44: 483–9. 153 Panagiotou O a, Ioannidis JP a. What should the genome-wide significance threshold be? Empirical replication of borderline genetic associations. Int J Epidemiol 2012; 41: 273–86. 154 Wakefield J. Commentary: Genome-wide significance thresholds via Bayes factors. Int J Epidemiol 2012; 41: 286–91. 155 Stein EA, Mellis S, Yancopoulos GD, et al. Effect of a monoclonal antibody to PCSK9 on LDL choles- terol. N Engl J Med 2012; 212: 408–9. 156 Stein E a, Gipe D, Bergeron J, et al. Effect of a monoclonal antibody to PCSK9, REGN727/SAR236553, to reduce low-density lipoprotein cholesterol in patients with heterozygous familial hypercholesterol- aemia on stable statin dose with or without ezetimibe therapy: a phase 2 randomised controlle. Lancet 2012; 380: 29–36. 157 Chan JCY, Piper DE, Cao Q, et al. A proprotein convertase subtilisin/kexin type 9 neutralizing antibody reduces serum cholesterol in mice and nonhuman primates. Proc Natl Acad Sci U S A 2009; 106: 9820–5. 158 McKenney JM, Koren MJ, Kereiakes DJ, Hanotin C, Ferrand A-C, Stein E a. Safety and efficacy of a monoclonal antibody to proprotein convertase subtilisin/kexin type 9 serine protease, SAR236553/ REGN727, in patients with primary hypercholesterolemia receiving ongoing stable atorvastatin thera- py. J Am Coll Cardiol 2012; 59: 2344–53. 159 Stein EA, Mellis S, Yancopoulos GD, et al. Effect of a monoclonal antibody to PCSK9 on LDL choles- terol. NEJM 2012; 212: 408–9. 160 Didelez V, Sheehan N. Mendelian randomization as an instrumental variable approach to causal infer- ence. Stat Methods Med Res 2007; 16: 309–30. 161 Thomas DC, Conti D. Commentary: the concept of ‘Mendelian Randomization’. Int J Epidemiol 2004; 33: 21–5. 162 Davey Smith G, Ebrahim S. ‘Mendelian randomization’: can genetic epidemiology contribute to under- standing environmental determinants of disease? Int J Epidemiol 2003; 32: 1–22.
20162300 Marten Siemelink.indd 40 15-03-16 13:23 ADVANCES IN BIOMARKER RESEARCH | 41
163 Zacho J, Tybjaerg-Hansen A, Jensen JS, Grande P, Sillesen H, Nordestgaard BG. Genetically elevated C-reactive protein and ischemic vascular disease. N Engl J Med 2008; 359: 1897–908. 164 Elliott P, Chambers JC, Zhang W, et al. Genetic Loci associated with C-reactive protein levels and risk of coronary heart disease. JAMA 2009; 302: 37–48. 165 Voight BF, Peloso GM, Orho-Melander M, et al. Plasma HDL cholesterol and risk of myocardial infarc- tion: a mendelian randomisation study. Lancet 2012; 6736: 1–9. 166 Nozaki T, Sugiyama S, Koga H, et al. Significance of a multiple biomarkers strategy including endothe- lial dysfunction to improve risk stratification for cardiovascular events in patients at high risk for coro- 2 nary heart disease. J Am Coll Cardiol 2009; 54: 601–8. 167 Sinning J-M, Losch J, Walenta K, Böhm M, Nickenig G, Werner N. Circulating CD31+/Annexin V+ microparticles correlate with cardiovascular outcomes. Eur Heart J 2011; 32: 2034–41. 168 Bernal-Mizrachi L, Jy W, Jimenez JJ, et al. High levels of circulating endothelial microparticles in pa- tients with acute coronary syndromes. Am Heart J 2003; 145: 962–70. 169 Mallat Z, Benamer H, Hugel B, et al. Elevated Levels of Shed Membrane Microparticles With Pro- coagulant Potential in the Peripheral Circulating Blood of Patients With Acute Coronary Syndromes. Circulation 2000; 101: 841–3. 170 Nozaki T, Sugiyama S, Sugamura K, et al. Prognostic value of endothelial microparticles in patients with heart failure. Eur J Heart Fail 2010; 12: 1223–8. 171 Amabile N, Guérin AP, Tedgui A, Boulanger CM, London GM. Predictive value of circulating endothelial microparticles for cardiovascular mortality in end-stage renal failure: a pilot study. Nephrol Dial Trans- plant 2012; 27: 1873–80. 172 Koga H, Sugiyama S, Kugiyama K, et al. Elevated levels of VE-cadherin-positive endothelial microparti- cles in patients with type 2 diabetes mellitus and coronary artery disease. J Am Coll Cardiol 2005; 45: 1622–30. 173 Fichtlscherer S, De Rosa S, Fox H, et al. Circulating microRNAs in patients with coronary artery dis- ease. Circ Res 2010; 107: 677–84. 174 Sondermeijer BM, Bakker A, Halliani A, et al. Platelets in patients with premature coronary artery dis- ease exhibit upregulation of miRNA340* and miRNA624*. PLoS One 2011; 6: e25946. 175 Takahashi Y, Satoh M, Minami Y, Tabuchi T, Itoh T, Nakamura M. Expression of miR-146a/b is associ- ated with the Toll-like receptor 4 signal in coronary artery disease: effect of renin-angiotensin sys- tem blockade and statins on miRNA-146a/b and Toll-like receptor 4 levels. Clin Sci (Lond) 2010; 119: 395–405. 176 Hoekstra M, van der Lans C a C, Halvorsen B, et al. The peripheral blood mononuclear cell microRNA signature of coronary artery disease. Biochem Biophys Res Commun 2010; 394: 792–7. 177 Weber M, Baker MB, Patel RS, Quyyumi A a, Bao G, Searles CD. MicroRNA Expression Profile in CAD Patients and the Impact of ACEI/ARB. Cardiol Res Pract 2011; 2011: 532915. 178 Taurino C, Miller WH, McBride MW, et al. Gene expression profiling in whole blood of patients with coronary artery disease. Clin Sci (Lond) 2010; 119: 335–43. 179 Ai J, Zhang R, Li Y, et al. Circulating microRNA-1 as a potential novel biomarker for acute myocardial infarction. Biochem Biophys Res Commun 2010; 391: 73–7. 180 Widera C, Gupta SK, Lorenzen JM, et al. Diagnostic and prognostic impact of six circulating microR- NAs in acute coronary syndrome. J Mol Cell Cardiol 2011; 51: 872–5. 181 Wang G-K, Zhu J-Q, Zhang J-T, et al. Circulating microRNA: a novel potential biomarker for early diag- nosis of acute myocardial infarction in humans. Eur Heart J 2010; 31: 659–66. 182 Oerlemans MIFJ, Mosterd A, Dekker MS, et al. Early assessment of acute coronary syndromes in the emergency department: the potential diagnostic value of circulating microRNAs. EMBO Mol Med 2012; 4: 1176–85. 183 D’Alessandra Y, Devanna P, Limana F, et al. Circulating microRNAs are new and sensitive biomarkers of myocardial infarction. Eur Heart J 2010; 31: 2765–73. 184 Gidlöf O, Andersson P, van der Pals J, Götberg M, Erlinge D. Cardiospecific microRNA plasma levels correlate with troponin and cardiac function in patients with ST elevation myocardial infarction, are selectively dependent on renal elimination, and can be detected in urine samples. Cardiology 2011; 118: 217–26. 185 Olivieri F, Antonicelli R, Lorenzi M, et al. Diagnostic potential of circulating miR-499-5p in elderly pa- tients with acute non ST-elevation myocardial infarction. Int J Cardiol 2012; : 1–6. 186 De Rosa S, Fichtlscherer S, Lehmann R, Assmus B, Dimmeler S, Zeiher AM. Transcoronary concentra- tion gradients of circulating microRNAs. Circulation 2011; 124: 1936–44. 187 Corsten MF, Dennert R, Jochems S, et al. Circulating MicroRNA-208b and MicroRNA-499 reflect myo- cardial damage in cardiovascular disease. Circ Cardiovasc Genet 2010; 3: 499–506. 188 Devaux Y, Vausort M, Goretti E, et al. Use of circulating microRNAs to diagnose acute myocardial infarction. Clin Chem 2012; 58: 559–67.
20162300 Marten Siemelink.indd 41 15-03-16 13:23 42 | CHAPTER 2
189 Adachi T, Nakanishi M, Nishimura K, et al. Plasma MicroRNA 499 as a Biomarker of Acute Myocardial Infarction. Clin Chem 2010; 56: 1183–5. 190 Cheng Y, Tan N, Yang J, et al. A translational study of circulating cell-free microRNA-1 in acute myocar- dial infarction. Clin Sci (Lond) 2010; 119: 87–95. 191 Kuwabara Y, Ono K, Horie T, et al. Increased microRNA-1 and microRNA-133a levels in serum of patients with cardiovascular disease indicate myocardial damage. Circ Cardiovasc Genet 2011; 4: 446–54. 192 Wang R, Li N, Zhang Y, Ran Y, Pu J. Circulating MicroRNAs are Promising Novel Biomarkers of Acute Myocardial Infarction. Intern Med 2011; 50: 1789–95. 193 Meder B, Keller A, Vogel B, et al. MicroRNA signatures in total peripheral blood as novel biomarkers for acute myocardial infarction. Basic Res Cardiol 2011; 106: 13–23. 194 Gan CS, Wang CW, Tan KS. Circulatory microRNA-145 expression is increased in cerebral ischemia. Genet Mol Res 2012; 11: 147–52. 195 Zeng L, Liu J, Wang Y, et al. MicroRNA-210 as a novel blood biomarker in acute cerebral ischemia. Front Biosci (Elite Ed) 2011; 3: 1265–72. 196 Tan KS, Armugam A, Sepramaniam S, et al. Expression profile of MicroRNAs in young stroke patients. PLoS One 2009; 4: e7689. 197 Li T, Cao H, Zhuang J, et al. Identification of miR-130a, miR-27b and miR-210 as serum biomarkers for atherosclerosis obliterans. Clin Chim Acta 2011; 412: 66–70. 198 Aoki A, Ozaki K, Sato H, et al. SNPs on chromosome 5p15.3 associated with myocardial infarction in Japanese population. J Hum Genet 2011; 56: 47–51. 199 Takeuchi F, Yokota M, Yamamoto K, et al. Genome-wide association study of coronary artery disease in the Japanese. Eur J Hum Genet 2012; 20: 333–40. 200 Wang F, Xu C-Q, He Q, et al. Genome-wide association identifies a susceptibility locus for coronary artery disease in the Chinese Han population. Nat Genet 2011; 43: 345–9. 201 The IBC 50K CAD Consortium. Large-scale gene-centric analysis identifies novel variants for coronary artery disease. PLoS Genet 2011; 7: e1002260. 202 IL6R Genetics Consortium and Emerging Risk Factors Collaboration. Interleukin-6 receptor pathways in coronary heart disease: a collaborative meta-analysis of 82 studies. Lancet 2012. 203 The Interleukin-6 Receptor Mendelian Randomisation Analysis (IL6R MR) Consortium. The interleu- kin-6 receptor as a target for prevention of coronary heart disease: a mendelian randomisation analy- sis. Lancet 2012; 6736: 6–16. 204 Davies RW, Wells G a, Stewart AFR, et al. A genome-wide association study for coronary artery dis- ease identifies a novel susceptibility locus in the major histocompatibility complex. 2012 DOI:10.1161/ CIRCGENETICS.111.961243. 205 Erdmann J, Grosshennig A, Braund PS, et al. New susceptibility locus for coronary artery disease on chromosome 3q22.3. Nat Genet 2009; 41: 280–2. 206 The International Stroke Genetics Consortium (ISGC) & the Wellcome Trust Case Control Consortium 2 (WTCCC2). Genome-wide association study identifies a variant in HDAC9 associated with large ves- sel ischemic stroke. Nat Genet 2012; 44: 328–33. 207 Holliday EG, Maguire JM, Evans T-J, et al. Common variants at 6p21.1 are associated with large artery atherosclerotic stroke. Nat Genet 2012; 44: 1147–51. 208 Wain L V, Verwoert GC, O’Reilly PF, et al. Genome-wide association study identifies six new loci influ- encing pulse pressure and mean arterial pressure. Nat Genet 2011; 43: 1005–11. 209 The International Consortium for Blood Pressure Genome-Wide Association Studies. Genetic variants in novel pathways influence blood pressure and cardiovascular disease risk. Nature 2011; 478: 103–9. 210 Dehghan A, Dupuis J, Barbalic M, et al. Meta-analysis of genome-wide association studies in >80 000 subjects identifies multiple loci for C-reactive protein levels. Circulation 2011; 123: 731–8. 211 The Tobacco and Genetics Consortium. Genome-wide meta-analyses identify multiple loci associated with smoking behavior. Nat Genet 2010; 42: 441–7. 212 O’Donnell CJ, Kavousi M, Smith A, et al. Genome-wide association study for coronary artery calcifica- tion with follow-up in myocardial infarction. Circulation 2011; 124: 2855–64. 213 Gertow K, Sennblad B, Strawbridge RJ, et al. Identification of the BCAR1-CFDP1-TMEM170A Lo- cus as a Determinant of Carotid Intima-Media Thickness and Coronary Artery Disease Risk. 2012 DOI:10.1161/CIRCGENETICS.112.963660. 214 Murabito JM, White CC, Kavousi M, et al. Association between chromosome 9p21 variants and the ankle-brachial index identified by a meta-analysis of 21 genome-wide association studies. Circ Cardio- vasc Genet 2012; 5: 100–12.
20162300 Marten Siemelink.indd 42 15-03-16 13:23 | 43
20162300 Marten Siemelink.indd 43 15-03-16 13:23 M.A. Siemelink1 T. Zeller2, 3
1 Laboratory of Experimental Cardiology, University Medical Center Utrecht 2 University Heart Center Hamburg, Clinic for General and Interventional Cardiology
20162300 Marten Siemelink.indd 44 15-03-16 13:23 CHAPTER 3
Biomarkers of Coronary Artery Disease: The promise of the Transcriptome
CURR CARDIOL REP. 2014 AUG;16(8):513. REVIEW
20162300 Marten Siemelink.indd 45 15-03-16 13:23 46 | CHAPTER 3
Abstract
The last years have witnessed tremendous technical advances in the field of transcriptomics that enable the simultaneous assessment of nearly all transcripts expressed in a tissue at a given time. These advances harbour the potential to gain a better understanding of the complex biological systems and for the identification and development of novel biomarkers. This article will review the current knowledge of transcriptomics biomarkers in the cardiovascular field and will provide an overview about the promises and challenges of the transcriptomics approach for biomarker identification.
20162300 Marten Siemelink.indd 46 15-03-16 13:23 PROMISE OF THE TRANSCRIPTOME | 47
Introduction
While cardiovascular disease (CVD) has traditionally been considered a disease of the Western society, its global incidence is on the rise and it is currently more prevalent in low- and middle income countries in Asia and Africa.1 To prevent CVD, accurate personal risk-assessment is paramount. The 2012 European Society of Cardiology (ESC) guidelines recommend risk- assessment using the updated SCORE charts based on age, gender, smoking, blood pressure and total cholesterol.2 The recent joint guidelines by the American College of Cardiology and the American Heart Association (ACC/AHA) recommend a model based on the Framingham Risk Score using generally similar parameters.3 However, these current risk prediction models 3 only provide a rough estimate of individual risk. Therefore, great value is posited in the identification and development of new biomarkers for CVD risk prediction. Decades of research have shown that improvement of risk prediction requires comprehensive understanding of the disease mechanism. The tremendous progress achieved in the ‘omics’ field has successfully improved the understanding of CVD pathophysiology by comprehensively interrogating disease states at the molecular level. This molecular phenotyping has become feasible by novel, robust, and fast high-throughput analytic platforms providing novel opportunities for molecular biomarker identification.4 Transcriptomics, the study of ribonucleic acid (RNA) transcripts and their expression patterns at a genome-wide level, is particularly promising for biomarker identification. This article will review current knowledge of transcriptomics biomarkers in the cardiovascular field and provide an overview about the promises and challenges of the transcriptomics approach for biomarker identification.
RNA RNA has long been considered as the messenger molecule between genes and proteins, where RNA is transcribed from DNA to messenger RNA (mRNA) and subsequently translated into protein.5,6 In recent years, non-coding RNA species have been characterized including microRNAs (miRNAs) and long non-coding RNAs (lncRNAs).7,8 miRNAs are endogenous, non-coding small RNAs of about 22 nucleotides regulating gene expression at a post-transcriptional level.9,10 They are involved in a broad range of biological processes and their dysregulation impacts disease development.11 Of great interest is that miRNAs are stable in biological fluids such as blood and urine,12,13 are actively secreted in microparticles and show tissue-specificity, attractive features of potential biomarkers.14 lncRNAs cover RNA molecules over 200 nucleotides and are observed in a wide range of tissues. They exert a broad repertoire of functions and have been linked to differentiation and developmental processes and disease.8,15 Compared to miRNAs, the widespread attention on lncRNAs is a rather recent phenomenon, nonetheless some promising evidence of using lncRNAs as biomarkers exist.16
Technology Platforms Historically, investigation of RNA expression was performed using northern blotting or RT-PCR approaches, at best investigating several RNA targets at once. Since several years, the use of expression microarrays has allowed rapid unbiased screening of nearly the entire transcriptome for discovery of the most promising targets. In microarray-based methods tens of thousands of transcripts are simultaneously analyzed by chemically labelling RNA
20162300 Marten Siemelink.indd 47 15-03-16 13:23 48 | CHAPTER 3
molecules and subsequent hybridization to probes on the microarray. The strength of microarrays lies in the extensive coverage, the high-throughput applicability and the relative inexpensiveness of the microarray approach. However, microarray technology is limited by the amount of RNA required, the limited dynamic range for quantification and can only detect predefined transcripts. Furthermore, questions are raised about the reproducibility and reliability of microarray experiments. Currently, we are at the brink of a new revolution, brought about by the advent of next- generation RNA-sequencing (RNA-seq). Although still prohibitively expensive, advances in RNA-seq will allow for superior scrutiny of the transcriptome, providing absolute quantification of transcripts while including splice variants, non-coding RNA and yet unknown transcripts.17 RNA-seq uses deep-sequencing technologies whereby a population of RNA (e.g. mRNA or miRNA) is converted to a cDNA library which is subsequently sequenced in a high-throughput base-by-base manner to obtain short sequences. The reads, typically 30–400 bp depending on the DNA-sequencing technology used, are used to reconstruct the original RNA-sequence in silico.18 The use of these so called next generation sequencing technology for the analysis of RNA has pioneered work with small regulatory RNAs, possibly because this field has benefited less from microarrays as the usual size of small RNAs is too short to be captured adequately with the limited resolution of microarrays.19 Detailed descriptions of microarray and RNA-seq approaches are out of scope of this work, but many excellent reviews provide a comprehensive overview.19–21 As the technological capabilities for measuring transcript expression have vastly improved, the importance of expression data for the development of new biomarkers has soared. The opportunity for transcriptome-wide screening of biomarkers allows for unbiased investigation of their potential as an individual biomarker for disease.
Transcriptomics-based Biomarkers in Cardiovascular Disease Recent advances in the cardiovascular biomarker field have identified novel and emerging transcriptomics-based biomarkers (Table 1). Here, we highlight examples that have started to emerge into clinical practice.
ST2 (IL-1RL-1, Interleukin 1 receptor-like 1) ST2 represents a promising biomarker identified by a transcriptomics approach. Weinberg and colleagues identified the ST2 gene as upregulated in cardiac myocytes subjected to mechanical strain by microarray analysis.22 Soluble ST2 is a secreted receptor belonging to the IL-1 receptor family that regulates inflammation and immunity.23 The soluble form of the protein can be measured in peripheral blood and a test kit for measurements of soluble ST2 is already commercially available (Critical Diagnostics Presage ST2 Assay). It has been shown that ST2 levels rise above normal in the context of various cardiac diseases24 such as heart failure25 and ischemic heart disease.26 In the Framingham Heart Study, measurements of soluble ST2 showed clear gender differences, an increase with age and increased levels in association with diabetes and hypertension27 and soluble ST2 added prognostic value to standard risk factors.28 Novel findings, however, indicate that genetic factors account for up to 40% of the inter-individual variability of soluble ST2 levels, which must be taken into account in future studies of ST2 as a biomarker.29 ST2 is a clear example how the initial microarray analyses identified a target as cardiac biomarker and led to the development of a suitable assay.
20162300 Marten Siemelink.indd 48 15-03-16 13:23 PROMISE OF THE TRANSCRIPTOME | 49
Table 1. Studies evaluating gene expression data for biomarker identification for coronary artery disease Gene expression (single genes) Study scheme Selected publications sST2 HF, ischemic heart disease 22,25,26 GDF15 acute coronary syndromes, 30,32 angina pectoris, HF FSTL1 ACS, HF 35,37 Gene expression signatures 3 23-gene score (Corus CAD) obstructive CAD 43–45 Circulating microRNAs miR-126, miR-223, miR-197 CHD 53 miR-1, miR-122, miR-133, miR-208a/b, AMI 49,90–94 miR-375, miR-499 miR-21, mir-29a, miR-208a post-MI 95 miR-1, miR-133a, miR-499, Mir-208 CAD 96,97 microRNA signatures 20-miRNA signature AMI, single time point 56 6-miRNA signature AMI, serial time points 98 4-miRNA signature AMI vs Takotsubo Cardiomyopathy 99
Growth Differentiation Factor-15 (GDF-15) GDF-15, a distant member of the TGF-β cytokine superfamily, has been identified by gene expression microarray analyses as being massively upregulated in nitric oxide (NO)-treated cardiomyocytes,30 under oxidative stress, in pressure overloaded left ventricles of mice with aortic stenosis and a mouse model of dilated cardiomyopathy.31 Levels of GDF15 can be measured in serum and plasma and evidence are accumulating that GDF15 is a strong and independent predictor of mortality and disease progression in patients with established disease, such as acute coronary syndromes, angina pectoris, heart failure.32 Moreover, circulating GDF-15 levels are independently related to intermediate cardiovascular phenotypes, including endothelial dysfunction, intima media thickness, plaque burden, and left ventricular hypertrophy and dilatation.33,34 Thus, measurement of GDF-15 may contribute to a refined risk assessment on top of traditional risk factors and biomarkers. The same group that reported on GDF15 as cardiac biomarker identified follistatin-like 1 (FSTL1) as an inducer of GDF15 production and an independent biomarker in acute coronary syndrome by using an expression screen for cDNAs encoding activators of the GDF15 promoter.35 FSTL1 had previously been indicated as a putative biomarker in chronic systolic heart failure36 and has been discussed as a novel therapeutic target for post-myocardial infarction and acute coronary syndrome.37
20162300 Marten Siemelink.indd 49 15-03-16 13:23 50 | CHAPTER 3
Expression Signatures A precise gene expression signature, i.e. a RNA expression pattern, has the promise to diagnose and classify diseases and potentially guide personalized treatment decisions for patients.4 Gene expression signatures have already been shown to accurately predict cardiomyopathy etiology in heart failure38,39 and to be useful in monitoring clinically significant allograft rejection.40,41 These data support ongoing efforts to incorporate biomarkers based on expression profiling to determine prognosis and response to therapy.38,42 In the Personalized Risk Evaluation and Diagnosis in the Coronary Tree (PREDICT) study, a whole blood gene expression score was developed and validated for the assessment of obstructive CAD in non-diabetic patients.43,44 This score is a function of the expression levels of 23 genes grouped into highly correlated terms reflecting biological processes or cell types44 and is associated with the probability of obstructive CAD.45 Subsequently, a multiplex assay for expression levels of the 23 gene transcripts became commercially available (Corus CAD, CardioDx, Palo Alto, CA).45 Multiplex tests are often complex, containing multiple sample processing steps, operators, machines and types of reagents which can affect assay variability. Assessment of the laboratory process variability showed that the Corus CAD intra-batch PCR variability contributed most to the overall variability while the reagent lot contributed most to inter-batch variability. Thomas et al. evaluated the diagnostic accuracy of the gene expression score to determine obstructive CAD in symptomatic patients referred for myocardial perfusion in the multicenter COMPASS study.46 The investigators found that the gene expression score was a significant predictor of obstructive CAD and resulted, at a predefined threshold, in a high sensitivity and high negative predictive value. Although the added value of a transcriptomics profile such as Corus CAD must be rigorously tested against current standard-of-care risk prediction and explored in different populations to define its clinical utility, the Corus CAD assay is extremely promising and one of the best examples of the value of transcriptomics-based biomarkers in the cardiovascular field today.
Circulating microRNAs Changes in the circulating miRNA levels have been associated with cardiovascular disease.47,48 As PCR-based techniques for quantifying circulating miRNAs improved, studies began to explore whether miRNAs could serve as clinical biomarkers, e.g. as biomarkers of the acute coronary syndrome,49,50 acute myocardial infarction,51 heart failure.52 In the Bruneck study, one of the largest studies measuring miRNAs, Zampetaki et al. screened levels of 19 circulating miRNAs by quantitative RT-PCR.53 Three miRNAs formed a signature for myocardial infarction: miR-126, miR-223 and miR-197. Those miRNAs added information to the Framingham Risk Score for the endpoint coronary heart disease and led to better patient stratification to risk categories, indicating the potential value of these miRNAs as biomarkers for cardiovascular risk prediction. However, most published miRNAs studies were small case-control studies and should be interpreted with caution and further work in larger populations is required. Detailed overviews of the current miRNA biomarker literature are given in several publications.9,54,55
20162300 Marten Siemelink.indd 50 15-03-16 13:23 PROMISE OF THE TRANSCRIPTOME | 51
MicroRNA Signatures Similar to specific gene expression signatures, signatures of miRNAs may reflect a given disease state and have potential as biomarker. Meder et al. assessed whole-genome miRNA expression in whole blood samples of patients with acute myocardial infarction (AMI).56 121 miRNAs were identified to be significantly dysregulated in AMI. The predictive power of these miRNAs were evaluated by receiver operator characteristic curves, and area under the curve (AUC) values of up to 0.94 were observed for the most predictive single miRNAs, miR-1291 and miR-663b. Using an algorithm for self-learning pattern recognition, a unique 20-miRNA signature was identified that predicts AMI with higher power and better AUC compared to individual miRNAs, even at stages when troponin T was still negative. These 3 study results implicate that miRNA signatures, derived from peripheral blood, can serve as a valuable biomarker and may improve biomarker-based diagnosis of AMI. However, it needs to be mentioned that the sample size was rather small and larger patient cohorts are needed for validation. In a subsequent miRNA study the same group investigated the kinetics of miRNA dysregulation in serial measurements in AMI patients and confirmed a 6-miRNA signature, including 5 out of the 20 miRNAs identified in the previous study.57 These serial measurements identified distinct miRNA patterns in the very early phase of AMI that resolved within the first days of successful therapy. Significant differences were seen mainly at the two earliest time points, indicating those miRNAs to be early markers of AMI. The authors hypothesize that, although the release of molecules from injured myocardium may be similar for miRNA and proteins, a whole-blood approach may provide further information because it would reflect the disease processes involved in the pathogenesis of rather than solely detecting myocardial necrosis. Clearly, future studies are needed to examine the value of miRNA signatures as potential robust biomarkers; nevertheless, miRNAs and miRNA signatures are emerging promising new players in cardiovascular biomarker research.
Long non-coding RNAs Recently another class of non-coding RNAs, lncRNAs, has aroused interest in cardiovascular function and disease. Growing evidence suggest that lncRNAs are key regulatory molecules at every level of cellular physiology, and their alterations are associated with multiple human diseases and may provide promising new targets for biomarker identification.58,59 Despite the progress made in oncology studies that tested lncRNAs as biomarkers for e.g. breast cancer,60 endometrial carcinoma61 and lung cancer,62 data on lncRNA biomarkers in the cardiovascular field is still poor and further work is essential to improve the overall understanding and value of lncRNAs as biomarkers.
Challenges in Biomarker Development Multiple stages are required for the “pipeline” of transcriptomics biomarker discovery and development. These stages include among others i) discovery of putative biomarkers for the target disease phenotype, ii) (technical) validation of those biomarkers in various disease and population cohorts to characterize biomarker performance and iii) subsequent testing in large prospective clinical trials before translation into clinical routine. In addition, the impact of a new biomarker on clinical outcomes in terms of efficacy and cost effectiveness is a further step that should be taken. Novel technologies have contributed to a massive increase in
20162300 Marten Siemelink.indd 51 15-03-16 13:23 52 | CHAPTER 3
biomarker discovery projects and reports, however, only few have been validated for routine clinical practice.63 Numerous excellent reports are published providing a comprehensive overview of pitfalls and challenges for biomarker discovery and translation.55,63–67 Here, we briefly review the key challenging points (summarized in Figure 1).
Study design An appropriate study design is a foremost requirement for reliable transcriptomics-based biomarker identification, ensuring adequate sample size for analysis and accounting for possible confounders. We recently showed that age, gender, body mass index, inflammatory status, and smoking influence gene expression.68 Likewise, consideration should be given to the influence of cardiovascular risk factors, ethnicity, and medication on gene expression.4 In addition, common gene variants (i.e. single-nucleotide polymorphisms) and epigenetic
Figure 1. Challenges in Transcriptomic Biomarker Development. Figure depicts main steps in Biomarker discovery and development and associated challenges to overcome.
20162300 Marten Siemelink.indd 52 15-03-16 13:23 PROMISE OF THE TRANSCRIPTOME | 53
patterns can influence gene expression.4 To achieve adequate statistical power, large sample sizes, accurate clinical phenotyping and well-characterized populations are mandatory.64,69 Another primary consideration in study design is the choice of tissue or cell type to investigate. Due to the ease of access, circulating blood is often used as surrogate source of diseased tissue. However, it is unclear whether the blood transcriptome is suitable as surrogate for tissues like e.g. heart tissue. One needs to consider that whole blood contains a mixture of cell types whose proportions show inter variability and may alter depending on disease state.70 Animal models and in vitro experiments are still important methods employed for biomarker research. However, the translation of these studies towards clinical application is difficult 3 and could lead to false targets. Comparison of transcriptomics data from ex-vivo monocytes and the in vitro monocytic THP-1 cell-line showed important differences.71 Likewise, recently Seok et al. showed that human inflammatory expression profiles where highly similar between various causes of inflammation, yet very different from mice inflammatory expression profiles.72 This indicates that great care must be taken when translating such results into the clinical setting.
Analytical Considerations and Standardization In contrast to genomic data, a subject’s gene expression data will vary spatially and temporally. To reduce confounding factors influencing gene expression data, such as different sample preparations and differences in the PCR runs, gene expression data have to be normalized. This is a critical issue and a major concern in transcriptomics studies. Especially for circulating miRNA measurements normalization is a “hot topic” in the current discussion, and several normalization approaches are used such as quantile-quantile normalization or spike-in of artificial RNA material.20 However, normalization is currently applied in a non-standardized fashion and application of universal reference material is required. Furthermore, variation caused by pre-analytical and analytical factors can substantially influence gene expression data.4 Schurmann et al. showed that factors such as RNA quality, storage time of blood, and batches of RNA processing and amplification have strong influence on gene expression data.73 Other studies provide evidence for the variability inherent to the PCR process and about batch effects in high-throughput technologies.45,74 In addition, numerous variables have been shown to influence the detection of miRNAs in the pre-analytical phase such as heparin75 and can lead to erroneous results.76 This can be particularly challenging in the clinical setting, as differences in sample collection, sample processing and assay performance in different clinical centers are to be expected. Therefore, to eliminate technical and analytical variability and avoid artefactual data generation, consensus on standard methods for all steps is imperative.
Validation Validation of initial discovery results in independent, large-scale studies are required in the field of biomarker research. Ideally, results of transcriptomics analyses will be validated in multi-center real-world studies, even comprising decentralized processing of RNA and PCR analysis and optimization of (decentralized) clinical laboratory testing procedures.4 After validation of the initial expression results, the putative biomarker must be rigorously tested against the existing standard of care and explored in a wider population to define its clinical utility.
20162300 Marten Siemelink.indd 53 15-03-16 13:23 54 | CHAPTER 3
Another aspect that will become increasingly important is the validation of biomarkers for specific subgroups. It has been common practice for clinical laboratories to use specific reference values for several important subgroups like men and women or children and adults, when evaluating diagnostic markers. However, it is uncommon to determine the predictive value of a biomarker for specific subgroups. This is about to change, as it is clear from the recent recommendations on cardiovascular risk-assessment by the ACC/AHA, stating that race- and sex- specific risk-assessment is highly recommended (3).
Multidisciplinary approaches Getting candidate biomarkers into large-scale validation studies requires the integration of diverse skills. Most biomarker discovery is conducted in labs lacking the resources and multidisciplinary expertise needed.63 Therefore, biomarker discovery should be a component of large research networks, involving industry and experts in distinct fields such as molecular biology, analytical chemistry, bioinformatics, clinical-trial design, epidemiology, statistics, and health-care economics.63 Several collaborative initiatives have emerged in recent years to orchestrate biomarker research efforts (including transcriptomics-based biomarkers). These include, among others, the Innovative Medicines Initiative (IMI) (www. imi.europa.eu/) and the BiomarCaRE Consortium (www.biomarcare.eu), both funded by the European Union.
Transcriptomics, Genomics and Epigenomics The current trend in biomarker research is increasingly focused on the discovery of causal biomarkers indicative of changes in pathophysiological processes that are at the basis of the complex disease and a potential target for drug development. GWAS provide an important tool to reveal causality through the principle of “Mendelian Randomization”. Zacho et al. is a case in point, showing that genetically raised CRP levels did not influence risk of myocardial ischemia.77 Another clear example is the recent landmark paper by Voight et al. which showed that genetic predispositions that raised HDL-cholesterol levels had no influence on disease outcome, as opposed to genetic alterations in LDL-cholesterol levels.78 The method of ‘Mendelian randomization’ is also well-suited to indicate causality of transcriptomics-derived biomarkers. GWAS has found many single nucleotide polymorphisms (SNPs) affecting disease, yet the complex mechanisms through which they exert their effect, is still largely unknown, as many appear in non-coding regions of the genome. SNPs which influence mRNA expression are known as expression Quantitative Trait Loci (eQTL). SNPs associated with complex diseases are more likely to be eQTLs compared to other SNPs and 45% of genes associated with CVD contain eQTLs.79,80 SNPs also influence known risk factors of cardiovascular disease, for example lipoproteins, for which 96 eQTLs have been found in 157 known loci.81,82 This shows that eQTLs may be an important mechanism for cardiovascular risk SNPs, and emphasises the importance of transcriptomics for the interpretation of GWAS results. In addition to genetic biomarkers, epigenetic DNA modifications like DNA methylation and histone modifications could serve as biomarkers of disease. Most interest has recently been directed at DNA-methylation biomarkers, enabled by development of ‘epigenome- wide’ DNA-methylation arrays. To elucidate the tissue specific down-regulation of gene expression by DNA-methylation in a high-throughput fashion, transcriptomics are
20162300 Marten Siemelink.indd 54 15-03-16 13:23 PROMISE OF THE TRANSCRIPTOME | 55
indispensable. In a recent study, Grundberg et al compared DNA-methylation to GWAS and transcriptomics data and found that 28% of methylation quantitative trait loci (meQTLs) are associated with nearby SNPs, and 6% of SNPs played a role in both DNA-methylation and adipose tissue gene expression,83 showing the complex interplay between genetic variants, methylation and expression. In addition, SNPs may also influence the expression of mRNA through interference with non-coding RNA (ncRNA) regulatory activity. For example, Gamazon et al. analyzed the effects of SNP’s on expression (mRNA-eQTL) and microRNA expression (miRNA-eQTL) and showed significant enrichment of miRNA-eQTLs in known mRNA-eQTLs, thereby providing important evidence for specific miRNA-mRNA interactions. Furthermore, many 3 of the found SNPs were associated with traits of complex diseases.84 In an identical fashion, Kumar et al. identified SNPs that influence lincRNA expression, and showed associations of these SNPs with complex diseases.85 This indicates that dysregulation of transcriptome interactions could be an important disease mechanism, and may thus form interesting biomarker targets.
Future Perspectives Despite a tremendous increase of interest in the transcriptome, we are only just scratching the surface of its complexity. To fully elucidate the transcriptome requires robust sample processing as well as advances in technology and analysis methods. Whole transcriptome RNA sequencing is still in its infancy yet new developments seem very promising. Meanwhile, several companies acknowledge the trend for multi-marker diagnostics, and have developed custom expression arrays and multiplex PCR solutions suited for clinical application. Improvements in microfluidics lead to reduced sample volume requirements, smaller machines and laboratory set-ups and will soon culminate in lab-on- chip solutions. Advances in analysis methods require standardization of data normalization and optimal modelling.86 An increasingly important strategy of in silico modelling is the systems biology approach.87 It combines data at various biological levels (e.g. genomic, epigenomic, transcriptomic and proteomic) to identify targets of interest (Figure 2). In addition, it sheds light on the relation of the target biomarker to other markers, paving the way for in silico pathway analysis and enabling the identification of pathological pathways.88 As new biomarkers emerge on the horizon, improved risk prediction will have to be translated in to increased health benefits from therapeutic intervention. This is especially interesting for causal biomarkers, which can themselves act as a target for novel drug development. Furthermore, companion diagnostics indicating individual drug efficacy, will likely take a more prominent role, as we progress towards personalized medicine.
Conclusion
Over the last years, gene expression analyses strongly influenced the area of biomarker identification and development in the cardiovascular field. Several potential biomarkers have been identified including gene expression signatures and non-coding RNAs, and a few have been translated into clinical utility. However, several aspects in the “transcriptomics pipeline” of biomarker development deserve consideration, ranging from appropriate study
20162300 Marten Siemelink.indd 55 15-03-16 13:23 56 | CHAPTER 3
design and material to analytical methods, standardizations, most importantly, and validation. Finally, to reach clinical application of the biomarker, fundamental questions about the clinical potential need to be evaluated as outlined by Morrow and deLemos:89 i) Can the clinician measure the biomarker?, ii) does the biomarker add new information? and iii) does the biomarker help the clinician to manage patients?.
Figure 2. Transcriptomics for Biomarker Discovery. Simplified schematic of relevant transcrip- tome interactions for current biomarker development. Large studies are required to elucidate the complex interactions of the genome and epigenome with the transcriptome and subsequently the proteome. Bullets denote contemporary techniques. eQTL, expression quantitative trait loci; meQTL, methylation quantitative trait loci; mRNA, messenger RNA; ncRNA, non-coding RNA.
20162300 Marten Siemelink.indd 56 15-03-16 13:23 PROMISE OF THE TRANSCRIPTOME | 57
References
1 Mendis S, Puska P, Norrving B. Global Atlas on cardiovascular disease prevention and control. Geneva, 2011. 2 Perk J, De Backer G, Gohlke H, et al. European Guidelines on cardiovascular disease prevention in clinical practice (version 2012). The Fifth Joint Task Force of the European Society of Cardiology and Other Societies on Cardiovascular Disease Prevention in Clinical Practice (constituted by re. Eur Heart J 2012; 33: 1635–701. 3 Goff DC, Lloyd-Jones DM, Bennett G, et al. 2013 ACC/AHA guideline on the assessment of cardiovas- cular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol 2014; 63: 2935–59. 4 Zeller T, Blankenberg S. Blood-based gene expression tests: promises and limitations. Circ Cardiovasc 3 Genet 2013; 6: 139–40. 5 CRICK FH. On protein synthesis. Symp Soc Exp Biol 1958; 12: 138–63. 6 Mattick JS. The genetic signatures of noncoding RNAs. PLoS Genet 2009; 5: e1000459. 7 Eddy SR. Non-coding RNA genes and the modern RNA world. Nat Rev Genet 2001; 2: 919–29. 8 Mercer TR, Dinger ME, Mattick JS. Long non-coding RNAs: insights into functions. Nat Rev Genet 2009; 10: 155–9. 9 Fichtlscherer S, Zeiher AM, Dimmeler S, Sessa WC. Circulating MicroRNAs: Biomarkers or Mediators of Cardiovascular Diseases? Arterioscler Thromb Vasc Biol 2011; 31: 2383–90. 10 Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell 2004; 116: 281–97. 11 Small EM, Olson EN. Pervasive roles of microRNAs in cardiovascular biology. Nature 2011; 469: 336– 42. 12 Mitchell PS, Parkin RK, Kroh EM, et al. Circulating microRNAs as stable blood-based markers for can- cer detection. Proc Natl Acad Sci U S A 2008; 105: 10513–8. 13 Kinet V, Halkein J, Dirkx E, Windt LJ De. Cardiovascular extracellular microRNAs: emerging diagnostic markers and mechanisms of cell-to-cell RNA communication. Front Genet 2013; 4: 214. 14 Saikumar J, Ramachandran K, Vaidya VS. Noninvasive micromarkers. Clin Chem 2014; 60: 1158–73. 15 Ponting CP, Oliver PL, Reik W. Evolution and functions of long noncoding RNAs. Cell 2009; 136: 629–41. 16 Hauptman N, Glavac D. MicroRNAs and long non-coding RNAs: prospects in diagnostics and therapy of cancer. Radiol Oncol 2013; 47: 311–8. 17 Churko JM, Mantalas GL, Snyder MP, Wu JC. Overview of High Throughput Sequencing Technologies to Elucidate Molecular Pathways in Cardiovascular Diseases. Circ Res 2013; 112: 1613–23. 18 Wang Z, Gerstein M, Snyder M. RNA-Seq: a revolutionary tool for transcriptomics. Nat Rev Genet 2009; 10: 57–63. 19 Marguerat S, Bähler J. RNA-seq: from technology to biology. Cell Mol Life Sci 2010; 67: 569–79. 20 Devonshire AS, Sanders R, Wilkes TM, Taylor MS, Foy CA, Huggett JF. Application of next generation qPCR and sequencing platforms to mRNA biomarker analysis. Methods 2013; 59: 89–100. 21 Grant SFA, Hakonarson H. Microarray technology and applications in the arena of genome-wide as- sociation. Clin Chem 2008; 54: 1116–24. 22 Weinberg EO, Shimpo M, De Keulenaer GW, et al. Expression and regulation of ST2, an interleukin-1 receptor family member, in cardiomyocytes and myocardial infarction. Circulation 2002; 106: 2961–6. 23 Weinberg EO. ST2 protein in heart disease: from discovery to mechanisms and prognostic value. Biomark Med 2009; 3: 495–511. 24 Ciccone MM, Cortese F, Gesualdo M, et al. A novel cardiac bio-marker: ST2: a review. Molecules 2013; 18: 15314–28. 25 Weinberg EO, Shimpo M, Hurwitz S, Tominaga S, Rouleau J-L, Lee RT. Identification of serum soluble ST2 receptor as a novel heart failure biomarker. Circulation 2003; 107: 721–6. 26 Manzano-Fernández S, Mueller T, Pascual-Figal D, Truong QA, Januzzi JL. Usefulness of soluble con- centrations of interleukin family member ST2 as predictor of mortality in patients with acutely decom- pensated heart failure relative to left ventricular ejection fraction. Am J Cardiol 2011; 107: 259–67. 27 Coglianese EE, Larson MG, Vasan RS, et al. Distribution and clinical correlates of the interleukin recep- tor family member soluble ST2 in the Framingham Heart Study. Clin Chem 2012; 58: 1673–81. 28 Wang TJ, Wollert KC, Larson MG, et al. Prognostic utility of novel biomarkers of cardiovascular stress: the Framingham Heart Study. Circulation 2012; 126: 1596–604. 29 Ho JE, Chen W-Y, Chen M-H, et al. Common genetic variation at the IL1RL1 locus regulates IL-33/ST2 signaling. J Clin Invest 2013; 123: 4208–18.
20162300 Marten Siemelink.indd 57 15-03-16 13:23 58 | CHAPTER 3
30 Kempf T, Eden M, Strelau J, et al. The transforming growth factor-beta superfamily member growth- differentiation factor-15 protects the heart from ischemia/reperfusion injury. Circ Res 2006; 98: 351– 60. 31 Wollert KC. Growth-differentiation factor-15 in cardiovascular disease: from bench to bedside, and back. Basic Res Cardiol 2007; 102: 412–5. 32 Lindahl B. The Story of Growth Differentiation Factor 15: Another Piece of the Puzzle. Clin Chem 2013; 59: 1550–2. 33 Lind L, Wallentin L, Kempf T, et al. Growth-differentiation factor-15 is an independent marker of car- diovascular dysfunction and disease in the elderly: results from the Prospective Investigation of the Vasculature in Uppsala Seniors (PIVUS) Study. Eur Heart J 2009; 30: 2346–53. 34 Rohatgi A, Patel P, Das SR, et al. Association of growth differentiation factor-15 with coronary athero- sclerosis and mortality in a young, multiethnic population: observations from the Dallas Heart Study. Clin Chem 2012; 58: 172–82. 35 Widera C, Giannitsis E, Kempf T, et al. Identification of Follistatin-Like 1 by Expression Cloning as an Activator of the Growth Differentiation Factor 15 Gene and a Prognostic Biomarker in Acute Coronary Syndrome. Clin Chem 2012; 58: 1233–41. 36 El-Armouche A, Ouchi N, Tanaka K, et al. Follistatin-Like 1 in Chronic Systolic Heart Failure: A Marker of Left Ventricular Remodeling. Circ Hear Fail 2011; 4: 621–7. 37 Ogura Y, Ouchi N, Ohashi K, et al. Therapeutic impact of follistatin-like 1 on myocardial ischemic injury in preclinical models. Circulation 2012; 126: 1728–38. 38 Kittleson MM, Ye SQ, Irizarry RA, et al. Identification of a gene expression profile that differentiates between ischemic and nonischemic cardiomyopathy. Circulation 2004; 110: 3444–51. 39 Kittleson MM. Gene expression analysis of ischemic and nonischemic cardiomyopathy: shared and distinct genes in the development of heart failure. Physiol Genomics 2005; 21: 299–307. 40 Horwitz, PA; Tsai, EJ; Putt, ME; Gilmore, JM; Lepore, JJ; Parmacek, MS; Kao, AC; Desai, SS; Gold- berg, LR; Brozena, SC; Jessup, ML; Epstein, JA; Cappola T. Detection of Cardiac Allograft Rejection and Response to Immunosuppressive Therapy With Peripheral Blood Gene Expression. Circulation 2004; 110: 3815–21. 41 Deng MC, Eisen HJ, Mehra MR, et al. Noninvasive discrimination of rejection in cardiac allograft recipi- ents using gene expression profiling. Am J Transplant 2006; 6: 150–60. 42 Pedrotty DM, Morley MP, Cappola TP. Transcriptomic biomarkers of cardiovascular disease. Prog Car- diovasc Dis; 55: 64–9. 43 Elashoff MR, Wingrove JA, Beineke P, et al. Development of a blood-based gene expression algorithm for assessment of obstructive coronary artery disease in non-diabetic patients. BMC Med Genomics 2011; 4: 26. 44 Rosenberg S. Multicenter Validation of the Diagnostic Accuracy of a Blood-Based Gene Expression Test for Assessing Obstructive Coronary Artery Disease in Nondiabetic Patients. Ann Intern Med 2010; 153: 425. 45 Elashoff MR, Nuttall R, Beineke P, et al. Identification of factors contributing to variability in a blood- based gene expression test. PLoS One 2012; 7: e40068. 46 Thomas GS, Voros S, McPherson JA, et al. A blood-based gene expression test for obstructive coro- nary artery disease tested in symptomatic nondiabetic patients referred for myocardial perfusion imag- ing the COMPASS study. Circ Cardiovasc Genet 2013; 6: 154–62. 47 Tijsen AJ, Pinto YM, Creemers EE. Circulating microRNAs as diagnostic biomarkers for cardiovas- cular diseases. Am J Physiol Heart Circ Physiol 2012; published online Aug 31. DOI:10.1152/ajp- heart.00191.2012. 48 Deddens JC, Colijn JM, Oerlemans MIFJ, et al. Circulating MicroRNAs as Novel Biomarkers for the Early Diagnosis of Acute Coronary Syndrome. J Cardiovasc Transl Res 2013; 6: 884–98. 49 Kuwabara Y, Ono K, Horie T, et al. Increased microRNA-1 and microRNA-133a levels in serum of patients with cardiovascular disease indicate myocardial damage. Circ Cardiovasc Genet 2011; 4: 446–54. 50 Oerlemans MIFJ, Mosterd A, Dekker MS, et al. Early assessment of acute coronary syndromes in the emergency department: the potential diagnostic value of circulating microRNAs. EMBO Mol Med 2012; 4: 1176–85. 51 D’Alessandra Y, Pompilio G, Capogrossi MC. MicroRNAs and myocardial infarction. Curr Opin Cardiol 2012; 27: 228–35. 52 Tijsen AJ, Creemers EE, Moerland PD, et al. MiR423-5p as a circulating biomarker for heart failure. Circ Res 2010; 106: 1035–9. 53 Zampetaki A, Willeit P, Tilling L, et al. Prospective study on circulating MicroRNAs and risk of myocar- dial infarction. J Am Coll Cardiol 2012; 60: 290–9.
20162300 Marten Siemelink.indd 58 15-03-16 13:23 PROMISE OF THE TRANSCRIPTOME | 59
54 Mayr M, Zampetaki A, Willeit P, Willeit J, Kiechl S. MicroRNAs within the continuum of postgenomics biomarker discovery. Arterioscler Thromb Vasc Biol 2013; 33: 206–14. 55 De Guire V, Robitaille R, Tétreault N, et al. Circulating miRNAs as sensitive and specific biomarkers for the diagnosis and monitoring of human diseases: promises and challenges. Clin Biochem 2013; 46: 846–60. 56 Meder B, Keller A, Vogel B, et al. MicroRNA signatures in total peripheral blood as novel biomarkers for acute myocardial infarction. Basic Res Cardiol 2011; 106: 13–23. 57 Vogel B, Keller A, Frese KS, et al. Refining diagnostic microRNA signatures by whole-miRNome kinetic analysis in acute myocardial infarction. Clin Chem 2013; 59: 410–8. 58 Scheuermann JC, Boyer LA. Getting to the heart of the matter: long non-coding RNAs in cardiac de- velopment and disease. EMBO J 2013; 32: 1805–16. 59 Batista PJ, Chang HY. Long noncoding RNAs: cellular address codes in development and disease. Cell 2013; 152: 1298–307. 3 60 Sørensen KP, Thomassen M, Tan Q, et al. Long non-coding RNA HOTAIR is an independent prognostic marker of metastasis in estrogen receptor-positive primary breast cancer. Breast Cancer Res Treat 2013; 142: 529–36. 61 He X, Bao W, Li X, et al. The long non-coding RNA HOTAIR is upregulated in endometrial carcinoma and correlates with poor prognosis. Int J Mol Med 2014; 33: 325–32. 62 Weber D, Johnen G, Casjens S, et al. Evaluation of long noncoding RNA MALAT1 as a candidate blood- based biomarker for the diagnosis of non-small cell lung cancer. BMC Res Notes 2013; 6: 518. 63 Poste G. Bring on the biomarkers. Nature 2011; 469: 156–7. 64 Drucker E, Krapfenbauer K. Pitfalls and limitations in translation from biomarker discovery to clinical utility in predictive and personalised medicine. EPMA J 2013; 4: 7. 65 Tang WHW. Contemporary Challenges in Translating Biomarker Evidence Into Clinical Practice⁎⁎Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American. J Am Coll Cardiol 2010; 55: 2077–9. 66 Perlis RH. Translating biomarkers to clinical practice. Mol Psychiatry 2011; 16: 1076–87. 67 Zampetaki A, Mayr M. Analytical challenges and technical limitations in assessing circulating miRNAs. Thromb Haemost 2012; 108: 592–8. 68 Zeller T, Wild P, Szymczak S, et al. Genetics and beyond--the transcriptome of human monocytes and disease susceptibility. PLoS One 2010; 5: e10693. 69 Filiou MD, Turck CW. Biomarkers of Neurological and Psychiatric Disease. Elsevier, 2011 DOI:10.1016/ B978-0-12-387718-5.00001-8. 70 McHale CM, Zhang L, Thomas R, Smith MT. Analysis of the transcriptome in molecular epidemiology studies. Environ Mol Mutagen 2013; 54: 500–17. 71 Kohro T, Tanaka T, Murakami T, et al. A comparison of differences in the gene expression profiles of phorbol 12-myristate 13-acetate differentiated THP-1 cells and human monocyte-derived macrophage. J Atheroscler Thromb 2004; 11: 88–97. 72 Seok J, Warren HS, Cuenca AG, et al. Genomic responses in mouse models poorly mimic human inflammatory diseases. Proc Natl Acad Sci U S A 2013; 110: 3507–12. 73 Schurmann C, Heim K, Schillert A, et al. Analyzing illumina gene expression microarray data from different tissues: methodological aspects of data analysis in the metaxpress consortium. PLoS One 2012; 7: e50938. 74 Leek JT, Scharpf RB, Bravo HC, et al. Tackling the widespread and critical impact of batch effects in high-throughput data. Nat Rev Genet 2010; 11: 733–9. 75 Boeckel J-N, Thomé CE, Leistner D, Zeiher AM, Fichtlscherer S, Dimmeler S. Heparin selectively af- fects the quantification of microRNAs in human blood samples. Clin Chem 2013; 59: 1125–7. 76 Becker N, Lockwood CM. Pre-analytical variables in miRNA analysis. Clin Biochem 2013; 46: 861–8. 77 Zacho J, Tybjaerg-Hansen A, Jensen JS, Grande P, Sillesen H, Nordestgaard BG. Genetically elevated C-reactive protein and ischemic vascular disease. N Engl J Med 2008; 359: 1897–908. 78 Voight BF, Peloso GM, Orho-Melander M, et al. Plasma HDL cholesterol and risk of myocardial infarc- tion: a mendelian randomisation study. Lancet 2012; 6736: 1–9. 79 Nicolae DL, Gamazon E, Zhang W, Duan S, Dolan ME, Cox NJ. Trait-associated SNPs are more likely to be eQTLs: annotation to enhance discovery from GWAS. PLoS Genet 2010; 6: e1000888. 80 Zhang X, Johnson AD, Hendricks AE, et al. Genetic associations with expression for genes implicated in GWAS studies for atherosclerotic cardiovascular disease and blood phenotypes. Hum Mol Genet 2013; 23: 782–95. 81 Teslovich TM, Musunuru K, Smith A V, et al. Biological, clinical and population relevance of 95 loci for blood lipids. Nature 2010; 466: 707–13. 82 Lipids G, Consortium G. Discovery and refinement of loci associated with lipid levels. 2013. DOI:10.1038/ng.2797.
20162300 Marten Siemelink.indd 59 15-03-16 13:23 60 | CHAPTER 3
83 Grundberg E, Meduri E, Sandling JK, et al. Global analysis of DNA methylation variation in adipose tis- sue from twins reveals links to disease-associated variants in distal regulatory elements. Am J Hum Genet 2013; 93: 876–90. 84 Gamazon ER, Ziliak D, Im HK, et al. Genetic architecture of microRNA expression: implications for the transcriptome and complex traits. Am J Hum Genet 2012; 90: 1046–63. 85 Kumar V, Westra H-J, Karjalainen J, et al. Human disease-associated genetic variation impacts large intergenic non-coding RNA expression. PLoS Genet 2013; 9: e1003201. 86 Shi L, Campbell G, Jones WD, et al. The MicroArray Quality Control (MAQC)-II study of common prac- tices for the development and validation of microarray-based predictive models. Nat Biotechnol 2010; 28: 827–38. 87 Shalhoub J, Sikkel MB, Davies KJ, Vorkas PA, Want EJ, Davies AH. Systems biology of human athero- sclerosis. Vasc Endovascular Surg 2014; 48: 5–17. 88 Diez D, Wheelock AM, Goto S, et al. The use of network analyses for elucidating mechanisms in car- diovascular disease. Mol Biosyst 2010; 6: 289–304. 89 Morrow DA, de Lemos JA. Benchmarks for the assessment of novel cardiovascular biomarkers. Cir- culation 2007; 115: 949–52. 90 Creemers EE, Tijsen AJ, Pinto YM. Circulating microRNAs: novel biomarkers and extracellular com- municators in cardiovascular disease? Circ Res 2012; 110: 483–95. 91 D’Alessandra Y, Devanna P, Limana F, et al. Circulating microRNAs are new and sensitive biomarkers of myocardial infarction. Eur Heart J 2010; 31: 2765–73. 92 Corsten MF, Dennert R, Jochems S, et al. Circulating MicroRNA-208b and MicroRNA-499 reflect myo- cardial damage in cardiovascular disease. Circ Cardiovasc Genet 2010; 3: 499–506. 93 Gidlöf O, Andersson P, van der Pals J, Götberg M, Erlinge D. Cardiospecific microRNA plasma levels correlate with troponin and cardiac function in patients with ST elevation myocardial infarction, are selectively dependent on renal elimination, and can be detected in urine samples. Cardiology 2011; 118: 217–26. 94 Cheng Y, Zhang C. MicroRNA-21 in Cardiovascular Disease. J Cardiovasc Transl Res 2010; 3: 251–5. 95 Zile MR, Mehurg SM, Arroyo JE, Stroud RE, DeSantis SM, Spinale FG. Relationship between the tem- poral profile of plasma microRNA and left ventricular remodeling in patients after myocardial infarction. Circ Cardiovasc Genet 2011; 4: 614–9. 96 Widera C, Gupta SK, Lorenzen JM, et al. Diagnostic and prognostic impact of six circulating microR- NAs in acute coronary syndrome. J Mol Cell Cardiol 2011; 51: 872–5. 97 De Rosa S, Fichtlscherer S, Lehmann R, Assmus B, Dimmeler S, Zeiher AM. Transcoronary concentra- tion gradients of circulating microRNAs. Circulation 2011; 124: 1936–44. 98 Vogel B, Keller A, Frese KS, et al. Multivariate miRNA signatures as biomarkers for non-ischaemic systolic heart failure. Eur Heart J 2013; 34: 2812–22. 99 Jaguszewski M, Osipova J, Ghadri J-R, et al. A signature of circulating microRNAs differentiates takot- subo cardiomyopathy from acute myocardial infarction. Eur Heart J 2013; 35: 999–1006.
20162300 Marten Siemelink.indd 60 15-03-16 13:23 | 61
20162300 Marten Siemelink.indd 61 15-03-16 13:23 S.W. van der Laan1 H. Schunkert7 M.A. Siemelink1 J. Erdmann7 S. Haitjema1 U. Hedin3 H. Foroughi Asl2 G. Paulsson-Berne8 L. Peresic3 J.L.M. Björkegrenn2 M. Mokry1 G.J. de Borst9 J. van Setten1 F.W. Asselbergs10,11,12 R. Malik4 H.M. den Ruijter1 M. Dichgans4 P.I.W. de Bakker13,14* N.J. Samani5,6 G. Pasterkamp1,15*
* Shared last authors
1 Laboratory of Experimental Cardiology, Division Heart & Lungs, University Medical Center Utrecht, Utrecht, the Netherlands 2 Cardiovascular Genomics Group, Division of Vascular Biology, Department of Medical Biochemistry and Biophysics, Karolinska Institutet, Stockholm, Sweden 3 Molecular Medicine and Surgery, Karolinska Instutet, Stockholm, Sweden 4 Institute for Stroke and Dementia Research, Medical Center Ludwig-Maximilians-University Münich, Münich, Germany 5 Department of Cardiovascular Sciences, University of Leicester, BHF Cardiovascular Research Centre, Glenfield Hospital, Leicester, United Kingdom 6 NIHR Leicester Cardiovascular Biomedical Research Unit, Glenfield Hospital, Leicester, United Kingdom 7 Deutsches Herzzentrum München, Klinik an der TU München, Munich Heart Alliance (DZHK), Münich, Germany 8 Unit of Cardiovascular Medicine, Department of Medicine, CMM, Karolinska Institutet, Stockholm, Sweden 9 Department of Surgery, Division of Surgical Specialties, University Medical Center Utrecht, Utrecht, the Netherlands 10 Department of Cardiology, Division Heart & Lungs, University Medical Center Utrecht, Utrecht, the Netherlands 11 Durrer Center for Cardiogenetic Research, ICIN-Netherlands Heart Institute, Utrecht, the Netherlands 12 Institute of Cardiovascular Science, Faculty of Population Health Sciences, University College London, London, United Kingdom 13 Department of Medical Genetics, Center for Molecular Medicine, University Medical Center Utrecht, Utrecht, the Netherlands 14 Department of Epidemiology, Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, the Netherlands 15 Laboratory of Clinical Chemistry and Hematology, Division Laboratories and Pharmacy, University Medical Center Utrecht, Utrecht, the Netherlands
20162300 Marten Siemelink.indd 62 15-03-16 13:23 CHAPTER 4
Coronary Artery Disease and Large Artery Stroke Loci are associated with Human Atherosclerotic Plaque Characteristics
MANUSCRIPT IN PREPARATION
20162300 Marten Siemelink.indd 63 15-03-16 13:23 64 | CHAPTER 4
Abstract
Tens of loci associate with large artery ischemic stroke (LAS) and coronary artery disease (CAD), but deciphering the underlying mechanisms is challenging. Atherosclerosis is cardinal to these diseases and pathohistological studies have identified vulnerable plaque characteristics that associate with clinical outcome. It is unknown to what extent common sequence variation associated with LAS and/or CAD risk relates to such advanced atherosclerotic lesion characteristics. We studied the impact of sequence variation on histological plaque characteristics, tissue-specific gene expression and regulation, using data from three biobanks comprising patients with clinically significant arterial stenosis. We report a ~4-fold enrichment (p = 3.6x10-5, 11 out of 51 variants) of CAD variants associated with plaque characteristics. The CAD risk reducing alleles of rs12539895 and a nearby deletion (chr7:106,901,393 TG > T) at 7q22 associated with less fatty carotid plaques (p < 5.1x10-6), and less circulating LDL levels. Circularized chromosome conformation capture in monocytes revealed many regional genes interacted with rs12539895. Further analyses revealed tissue-specific effects on HBP1, COG5, and GPR22 expression, prioritizing the list of 11 regional genes for future studies. Polygenic score analyses demonstrated only nominal significant overlap between CAD and LAS, and plaque characteristics. Our study supports the view that genetic analyses in deeply phenotyped patient cohorts, aid in prioritizing putative therapeutic targets among clinically relevant loci identified through GWAS.
20162300 Marten Siemelink.indd 64 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE CHARACTERISTICS | 65
Introduction
Atherosclerosis refers to the lifelong process of lesion formation and progression in the inner linings of the arteries1, and is a key contributor to acute clinical manifestations such as myocardial infarction (MI) and large artery ischemic stroke (LAS). Coronary artery disease (CAD), including MI, and LAS are complex diseases with a shared polygenic architecture2,3. To date, meta-analyses of genome-wide association studies (GWAS) have identified 45 loci for CAD and 2 for LAS in populations of European ancestry4,5. Understanding the underlying pathogenetic mechanisms of these loci is of direct interest to drug targeting research programs6, and contributes to fulfilling the promise of precision medicine7. However, the translation from associated variants to potentially causal genes as therapeutic targets remains a major challenge. The process of atherosclerosis has extensively been studied in plaque samples through histological analysis1. The high risk atherosclerotic lesion prone to destabilization and 4 rupture, is characterized by variability in lipid, inflammatory, calcific and thrombotic components1,8,9. Such characteristics have previously been linked to acute manifestations of disease9-12. It is unknown to what extent common genetic variation associated with CAD and/or LAS risk relates to advanced atherosclerotic lesion characteristics. The Athero-Express (AE)13, Stockholm Atherosclerosis Gene Expression (STAGE)14, and Biobank of Karolinska Endarterectomy (BiKE)15 are three independent, deeply phenotyped, biobank studies comprising individuals with clinically significant arterial stenosis. In the AE study the genetic architecture of histologically analyzed plaque specimens and plaque- derived DNA methylation is assessed8. STAGE14 and BiKE15 focus on the genetics of tissue- specific differential gene expression. We investigated CAD and LAS associated loci in the AE using individual variant and polygenic score analyses of plaque characteristics. We report that risk loci associated with plaque characteristics and show evidence of polygenic overlap with disease. One variant (rs12539895, risk allele A) on chromosome 7q22 near COG5, HBP1 and GPR22 was previously associated with reduced CAD risk. In the AE the same allele significantly associated with less fatty carotid lesions. To further increase our understanding of this association, we sought sequence coding and regulatory mechanisms, including carotid plaque DNA methylation and tissue-specific gene expression, in the AE, STAGE and BiKE studies.
Results
Single-variant analysis of CAD and LAS associated loci with plaque characteristics We correlated 47 CAD and LAS associated loci identified through GWAS to commonly assessed plaque characteristics in the Athero-Express; 4 loci identified in a GWAS for bipolar disorder (BD) served as negative controls (these 51 variants are listed in Supplemental Table 3). There was a ~4-fold enrichment for association with plaque characteristics as 11 out of 51 variants reached p < 0.05 (Table 1 and Table 2, binomial p = 3.61x10-5). Interestingly, most single-nucleotide variant (SNV) associations were observed with the amount of fat in atherosclerotic lesions (4 out of 11, p = 1.55x10-3). The strongest association
20162300 Marten Siemelink.indd 65 15-03-16 13:23 66 | CHAPTER 4
was observed for an intronic variant (rs12539895) on chromosome 7q22 in the gene encoding for conserved oligomeric Golgi complex subunit 5 (COG5) that was significantly associated with the presence of fat in plaques (p = 5.09x10-6, Table 1 and Figure 1). The same allele (A) was previously shown to reduce the risk of CAD5. Additionally, in the same region we observed that a deletion at chr7:106,901,393 (TG > T in the intron of COG5) showed the strongest association with intraplaque fat (Figure 1, OR = 0.52, 0.40-0.66 95% confidence interval [CI] per A-allele, p = 2.14x10-7, allele frequency = 0.17). The same deletion was also associated with a reduction in CAD risk (OR = 0.96, 0.93-0.99 CI, p = 0.0028) assessed by CARDIoGRAMplusC4D16. In addition, three other variants were nominally associated with fat in plaques. Including one variant (rs445925) in the ApoE-ApoC1 locus and a variant (rs2954029) in the TRIB1 locus. The third was the CAD associated variant (rs2023938) in the HDAC9 locus (Table 2). In contrast, rs2107595 in the same locus that associates with LAS did not associate with fatty lesions (p = 0.15, r2 = 0.48 with rs2023938). CAD loci were also associated with plaque collagen content, smooth muscle cell percentage, the extent of calcification, and intraplaque hemorrhage, but none associated with macrophage percentages in plaques (Table 1 and Table 2). Both LAS and the BD associated variants did not associate with any of the plaque characteristics.
Table 1. Association of CAD associated SNVs with quantitative plaque phenotypes.
Phenotype Locus SNV Chr BP Alleles EAF β (s.e.m.) P GWAS dir.
Smooth muscle cells HHIPL1 rs2895811 14 100,133,942 C/T 0.445 0.102 (0.040) 0.011 +
Vessel density UBE2Z rs15563 17 47,005,193 G/A 0.540 -0.080 (0.038) 0.034 -
Gene desert (KCNE2) rs9982601 21 35,599,128 T/C 0.120 0.117 (0.059) 0.049 +
Per SNP the chromosomal (Chr) base pair position (BP), the effect and other allele (Alleles), as well as the effect allele frequency (EAF) is given. The effect size (β) ± standard error of the mean (s.e.m.) is relative to the risk allele with its associated p-value (P). GWAS dir. indicates the direction of effect in the CAD GWAS. SMCs, smooth muscle cells. Vessels: intraplaque vessel density. Additional per variant statistics are in Supplemental Table 3.
Gene-based analysis of CAD and LAS associated loci with plaque characteristics Prioritizing genes near GWAS loci can be complicated depending on the underlying genetic architecture. Many causal variants in genes may exist that only show marginal significance in initial analyses of complex traits17, in this case plaque characteristics. A method to assess such gene based associations, is to calculate an empirical p-value per gene by aggregating the associated p-values of all variants in and near these genes while applying permutations17. We mapped 151 genes to the 51 loci and ran gene-based association tests using VEGAS218 on the 7 measured plaque characteristics (supplemental material). Two genes, HMG-box transcription factor 1 (HBP1) and COG5, on the aforementioned chromosome 7q22 significantly associated with fat in lesions (p < 5.0x10-5, Supplemental Table 4). The top SNV in HBP1 was rs10953530 (OR = 1.63, 1.33-2.00 95% CI per A-allele, p = 1.84x10-6) which is in LD (r2 = 0.90) with rs12539895.
20162300 Marten Siemelink.indd 66 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE CHARACTERISTICS | 67
chromosome 7 (q22.3-q31.1) q33 34 35
mQTL, eQTL CAD risk rs80341862 rs12539895 plaque fat content chr7:106,901,393:TG>T
4
CpG Islands K562 CpG Methylation 450K AoSMC
K562 CpG Methylation Bi-Seq AoSMC Leukocytes
K562
DNaseI Hypersensitivity AoSMC
Monocytes CD14+ Layered H3K4Me3 150 _ (often near promoters) 0 _ Layered H3K4Me1 50 _ (often near regulatory elements) 0 _ 100 _ Layered H3K27Ac (often near active regulatory elements) 0 _ 1 _ 4C Monocytes CD14+ 0 _
4.88 _
Conservation (100 vert.) 0 _
-4.5 _ NHGRI GWAS Catalog
Figure 1. Genetic associations in 7q22 with atheroma size. Rs12539895 was previously associ- ated with CAD (purple). The strongest association was for a deletion chr7:106901393 (TG > T, p = 2.14x10-7, see main text, pink). Rs80341862 is a meQTL and eQTL in various tissues (green, see main text). Chromosomal position (1000G, Hg19) on x-axis. The lower panel shows: CpG islands, CpG methylation loci based on 450K array, CpG methylation sites based on bisulfite sequencing, DNaseI hypersensitivity sites, histone methylation sites (H3KMe3, H3K4Me1), histone acetylation sites (H3K27Ac), circularized chromosome conformation capture (4C), conserved in 100 different vertebrate species, NHGRI GWAS catalog variants, and GENCODE genes from UCSC (black arrow indicates direction of transcription). The left y-axis shows the –log10(p-value) of the association with fatty lesions (<10% vs. >10% of plaque area). The right y-axis shows the recombination rate (grey line in the middle panel). The middle panel shows each associated variants colored by the r2 relative to rs12539895.
20162300 Marten Siemelink.indd 67 15-03-16 13:23 68 | CHAPTER 4
One gene, extended synaptotagmin-Like protein 3 (ESYT3) in the 3q22.3 locus (near MRAS), was significantly associated with the amount of collagen in plaques. The top SNV in ESYT3 was rs774007 (OR = 1.56 with collagen in the plaques, 1.26-1.93 95% CI per G-allele, p = 6.52x10-5). However, this variant is not in LD (r2 < 0.1) with the CAD associated variant in this locus (rs9818870), and showed no association with CAD (OR = 1.00, 0.97- 1.04 95% CI per G-allele, p = 0.78 for a proxy rs774008).
Tissue specific sequence variation effects on methylation and expression Given the single-variant and gene-based association presented above, we further explored the 7q22 locus with respect to regional gene expression and regulation. Sequence variation could regulate gene expression by changing tissue-specific DNA methylation19. To uncover cis-acting methylation quantitative trait loci (meQTLs) we associated all genetic variants in 7q22 with the methylation of all CpGs mapped to the same region using fastQTL20. We found 12 variants significantly associated with differential methylation of CpGs in DNA derived from carotid plaques, some of which also associated with CAD risk (p < 1.82x10-7, Table 3). Specifically, a proxy for rs12539895, rs80341862 (LD 2r = 1.0), was significantly associated with a decrease in methylation of a CpG at the 3’ UTR of COG5 and HBP1 (false discovery rate Q = 5.41x10-9 after 1,000 permutations, Table 3). A total of 8 additional variants reached a false-discovery rate (FDR) Q ≤ 0.05 (Supplemental Table 5), but only 1 CpG of the 20 was associated with fat in plaques (Supplemental Table 6). Alternatively, regional genetic variants could regulate gene expression by altering transcription factor binding or splice sites in plaques or other relevant tissues; such variants are known as expression quantitative trait loci (eQTLs). We associated variants in 7q22 with regional gene expression in tissues from the STAGE and BiKE studies whose inclusion criteria are comparable to those of the AE (Supplemental Table 7)14,15. In STAGE we found 4 variants in 3 tissues significantly modulating expression of two different genes (p ≤ 1.82x10-7, Table 4). The effect allele of rs3779495 in the body of COG5 associated with decreased COG5 expression in whole blood (Table 4), also associated with a reduction in CAD risk. In STAGE, 21 variants associated with expression of 10 genes across 6 tissues at a FDR Q ≤ 0.05; most eQTLs were found in liver and whole blood tissue (Supplemental Figure 1 and Supplemental Table 8). The correlated variants rs80341862 and rs12539895 which associated with fatty lesions (Table 2), and a reduction in CAD risk, also associated with a reduction of COG5 expression in atherosclerotic arterial wall tissue, internal mammary artery, and whole blood (Supplemental Table 8). In BiKE we tested 146 regional genotyped variants at 7q22 for association with the expression of 11 regional genes available on the expression array (see material and methods). This revealed three correlated variants that significantly associated with an increase of HBP1 expression in carotid plaques, p = 7.0x10-6; all are proxies for rs12539895 and rs80341862 (LD r2 > 0.90, Supplemental Figure 2 and Supplemental Table 9). The same alleles decrease CAD risk and intraplaque fat content (Supplemental Table 9). Considering the association of rs12539895 with fatty lesions and the central role of the LDL receptor (encoded by LDLR) in lipid uptake, we speculated on whether regional gene expression in carotid plaque is correlated to LDLR expression. Indeed, we find significant associations for PRKAR2B, COG5, DUS4L and CBLL1 with LDLR in plaques from BiKE (Supplemental Table 10).
20162300 Marten Siemelink.indd 68 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE CHARACTERISTICS | 69 - + - + + - - - GWAS dir. GWAS -4 -6 - 3 -4 5.09x10 0.028 8.09x10 5.13x10 0.016 3.54x10 0.034 P 0.014
4 0.626 [0.512-0.765] 0.759 [0.595-0.967] 0.759 [0.646-0.893] 0.654 [0.513-0.833] 0.783 [0.643-0.954] 0.775 [0.653-0.920] 0.673 [0.469-0.968] OR [95% CI] 0.746 [0.590-0.943] dir . indicates the direction of effect in CAD GWAS 0.234 0.135 0.523 0.117 0.333 0.475 0.113 EAF 0.127 A/C C/T C/T T/C G/A T/A A/G Alleles C/T 107,091,849 19,036,775 90,989,109 35,599,128 2,117,945 126,490,972 45,415,640 BP 203,745,885 7 7 10 21 17 8 19 Chr 2 rs12539895 rs2023938 rs11203042 rs9982601 rs2281727 rs2954029 rs445925 SNV rs6725887 7q22 HDAC9 LIPA Gene desert (KCNE2) SMG6 TRIB1 ApoE-ApoC1 Locus WDR12 Fat content , as <10% vs. >10% fat per the total plaque area. IPH intraplaque hemorrhage. Additional variant statistics are in Supplemental Fat content Intraplaque hemorrhage Collagen Per SNP the chromosomal ( Chr ) base pair position BP ), effect and other allele Alleles as well frequency EAF is given. The odds ratio ( OR ) ± 95% confidence interval CI is relative to the risk allele with its associated p-value P ). Calcification Table 2. Association of CAD associated SNVs with semi-quantitative plaque phenotypes. Table Phenotype GWAS. GWAS. 3. Table
20162300 Marten Siemelink.indd 69 15-03-16 13:23 70 | CHAPTER 4 -4 -4 -5 -4 0.960 0.697 0.674 6.60x10 0.293 6.53x10 7.80x10 8.45x10 0.939 0.285 0.803 0.275 P CAD 1.00 (0.98 - 1.02) 1.00 (0.98 - 1.01) 0.99 (0.97 - 1.02) 0.96 (0.94 - 0.98) 1.03 (0.98 - 1.08) 0.96 (0.94 - 0.98) 1.04 (1.02 - 1.06) 0.96 (0.94 - 0.98) 1.00 (0.98 - 1.02) 0.99 (0.97 - 1.01) 1.00 (0.98 - 1.02) 0.99 (0.97 - 1.01) OR (95% C.I.) Body 5’UTR Body 3’UTR Body Body Body Body 5’UTR Body CpG location PRKAR2B PRKAR2B HBP1 COG5 COG5, HBP1 SLC26A4, LOC286002 COG5 CBLL1 CBLL1 SLC26A3 DLD Gene -21 -10 -4 -5 -4 -9 -5 -36 -37 -5 -9 -7 4.67x10 1.02x10 2.12x10 5.41x10 2.96x10 5.18x10 5.48x10 3.09x10 2.46x10 8.24x10 5.63x10 5.10x10 Q ). 16 0.0064 (-0.0008) 0.0062 (-0.0012) -0.0030 (0.0005) -0.0083 (0.0011) 0.0058 (-0.0010) -0.0032 (0.0004) 0.0029 (-0.0004) -0.0103 (0.0009) -0.0074 (0.0014) -0.0094 (0.0017) 0.0232 (-0.0016) -0.0140 (0.0010) β (s.e.m.) -47,491 70,175 58,003 92,099 23,298 -426 41,063 -94 24,951 -359,724 48,510 4,447 Distance cg10691109 cg21250978 cg02696742 cg24556660 cg04276007 cg18560240 cg20501882 AEGS meQTL analysis cg27433759 cg14628708 cg21322654 cg23024343 cg14590325 CPG 0.43 0.10 0.25 0.25 0.27 0.37 0.39 CAF 0.48 0.05 0.27 0.79 0.44 T/G GA/G G/GC G/C G/A T/TGAA A/G Alleles A/AAAG C/T C/T A/G A/T 106,711,492 106,754,716 106,868,150 106,934,731 107,409,810 107,437,230 107,583,241 106,592,532 BP 106,890,156 106,941,324 107,250,260 107,390,163 variant the coded and other allele ( Alleles ) frequency CAF are given. For each associated CpG chromosomal Chr base pair rs3801969 chr7:106754716:I chr7:106868150:D rs80341862 rs41666 chr7:107437230:D rs13222925 chr7:106592532:D Table 3. 7q22 regional association of genetic variation with DNA methylation in plaques. Table Variant rs143162346 rs12112877 rs3779495 rs2023709 Per position ( BP ) as well the effect size β ± standard error of mean s.e.m. with its associated p-value P and false-discovery rate Q is given. The CpG location indicates the position of CpG relative to gene (s). Also indicated for each variant are odds ratio ( OR ) with 95% confidence interval ( 95% C.I. ) and the p-value of association with CAD (from CARDIoGRAMplusC4D
20162300 Marten Siemelink.indd 70 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE CHARACTERISTICS | 71
Functional analysis of chromosomal interactions on 7q22 and eQTL analysis in monocytes The above results show that 7q22 associated with less fatty plaques, reduced methylation in plaques, and differential expression in atherosclerotic tissues, liver, and whole blood cells. Studying the effects of variants on chromosomal structure and spatial organization can help understand the regulation of regional gene expression. Circulating and local monocytes play a central role in atherosclerosis, specifically in scavenging oxidized lipid- particles. We speculated on how and if rs12539895 might interact with regional genes in monocytes (CD14+ cells) and applied circularized chromosome conformation capture (4C) to interrogate regional interactions. This revealed that rs12539895 interacts with almost all genes within 1Mb either side of it (Figure 1 and Supplemental Table 11). These regional interactions of rs12539895 prompted us to investigate genetic effects on regional gene expression in monocytes. We queried data from Zeller et al.21 and found a variant (rs34084719) in perfect LD with rs12539895 that associated with GPR22 expression 4 in monocytes (p = 6.88x10-10). In addition, rs7811034 associated with DLD expression (p = 7.47x10-12) and rs6957510 with DUS4L expression (p = 1.73x10-10) in monocytes; these are uncorrelated to rs12539895.
Cardiovascular risk factors and the 7q22 locus Dominant causal contributing factors for cardiovascular disease risk are changes in circulating lipid proteins, presumably leading to changes in plaque fat content. As the fat and CAD associated alleles correlate with gene expression in relevant tissues for lipid metabolism (liver) and atherosclerosis (carotid plaques, arterial wall tissues, and whole blood), we speculated on whether these variants associated with circulating blood lipid levels too. Indeed, data from the Global Lipids Genetics Consortium22 revealed that the risk-reducing allele of rs12539895 also associated with a decrease circulating LDL and an increase in HDL (Supplemental Table 12). In contrast, the same variant did not associate with blood pressure, BMI or type 2 diabetes (Supplemental Table 12).
Polygenic association of clinically relevant variants with plaque phenotypes Because of the central role of atherosclerosis in CAD and LAS and the presumed polygenic nature of these complex diseases, we further tested the contribution of modestly associated variants, ascertained through GWAS, on atherosclerotic characteristics. We summed genetic variation weighted by the effect on disease risk into quantitative polygenic burden scores23. We made scores using dosage data from AE and the allelic effects estimated through GWAS on CAD and LAS; each score included SNVs selected by using increasingly 23 liberal thresholds of significance (pT, Supplemental Table 2) . We then correlated each of these scores, which capture the per-variant effects on clinical outcome, to the atherosclerotic plaque phenotypes. We found significant associations for the CAD score with the presence of macrophages and smooth muscle cells in atherosclerotic plaques, while the LAS score was associated with intraplaque hemorrhage (Figure 2, p < 0.05).
20162300 Marten Siemelink.indd 71 15-03-16 13:23 72 | CHAPTER 4
Figure 2: Polygenic scores of CAD and LAS associate with different plaque characteristics. Polygenic scores for LAS associate with intraplaque hemorrhage (panel A.), where as polygenic scores for CAD associate with macrophages (panel B.) and smooth muscle cells (panel C.). Scores were constructed based on GWAS summary statistics using increasingly liberal p-value thresholds
(pT, see also supplemental information and Supplemental Table 2). Each pT bin is depicted on the x- axis. The y-axes indicate the r2 of the association; please note the different scales. ** p < 0.005 and * p < 0.05 for association of the score with the plaque characteristic. CAD: coronary artery disease (yellow). LAS: large artery stroke (pink). BD: bipolar disorder (blue).
20162300 Marten Siemelink.indd 72 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE CHARACTERISTICS | 73 eQTL eQTL eQTL meQTL, eQTL, CAD Assoc. -4 0.159 0.208 8.45x10 0.142 P ) indicates which ( Assoc. ) last column ). The 0.99 (0.97 - 1.01) 0.99 (0.97 - 1.01) 1.04 (1.02 - 1.06) 0.99 (0.97 - 1.00) CAD OR (95% C.I.)
16 4 0.093 0.067 0.742 0.063 P 1.16 (0.97 - 1.39) 1.19 (0.99 - 1.42) 0.96 (0.78 - 1.20) 1.18 (0.99 - 1.41) OR (95% C.I.) Fatty lesion Liver SM WB WB Tissue -10 -5 -17 -14 7.08x10 7.84x10 1.57x10 2.85x10 Q -12 -7 -22 -17 2.31x10 5.91x10 1.64x10 P 1.50x10 and CAD (from CARDIoGRAMplusC4D CAD (from lessions and fatty 0.111 (0.019) -0.318 (0.026) 0.144 (0.017) Beta (SE) STAGE eQTL analysis STAGE 0.159 (0.015) C/T A/G C/T Alleles C/T 107582323 107,250,260 107,522,867 BP 107,540,244 DLD COG5 DLD Gene DLD variant the effect and other allele ( Alleles ) frequency EAF are given. For each associated CpG chromosomal Chr base pair rs12672634 rs3779495 associations were found for this variant. rs756382 7q22 regional association of variants with gene expression in 7 tissues in STAGE. 4. 7q22 regional association of variants with gene expression in 7 tissues STAGE. Table Variant Per position ( BP ) as well the effect size β ± standard error of mean s.e.m. with its associated p-value P and false-discovery rate Q is given. The CpG location indicates the position of relative to gene (s). Also indicated for each variant are odds ratio ( OR ) with 95% confidence with p-value of association and the ( 95% C.I. ) interval rs7779923
20162300 Marten Siemelink.indd 73 15-03-16 13:23 74 | CHAPTER 4
Discussion
In this study we associated clinically relevant variants with histological proxies of vulnerable human plaque characteristics; to our knowledge this large-scale study is the first to do so. We report a ~4-fold enrichment for association with plaque characteristics, and most associations were with fatty lesions. One variant, previously associated with a reduction in CAD risk rs12539895 at 7q22, showed the strongest association with less fat in carotid plaques. A further regional analysis revealed the strongest association with fatty plaques for a deletion (chr7:106,901,393 TG > T) in COG5; the same deletion also associated with a decrease in CAD risk. The tissue-specific meQTL and eQTL analyses revealed regulatory and transcriptional effects on COG5, HBP1, and DLD in skeletal muscle, carotid plaque, liver and whole blood tissues. A further investigation into the apparent regulatory and transcriptional role of rs12539895 through circularized chromosome conformation capture revealed many genes interacting in monocytes. And this variant also affected GPR22 expression in monocytes. In line with the current understanding of the role of lipids in cardiovascular disease, rs12539895 also decreases circulating LDL and increases HDL. Interestingly, polygenic scores for CAD associated with intraplaque macrophages and smooth muscle cells, while scores for LAS associated with intraplaque hemorrhage. These results are surprising, given the lack of single variant associations with intraplaque macrophages. The presence of macrophages in advanced plaques is thought to be a complex regulated process1. Our analyses suggest that single variants involved in macrophage infiltration may only have a marginal effect on eventual clinical outcome, or that macrophage infiltration derives as a generic consequence of vascular wall injury. Admittedly, power (due to sample size or a noisy phenotype) in the Athero-Express is limited. However, to our knowledge no other study exists with the size and scope of the Athero-Express to study genetic variation in association with human plaques characteristics. Our study supports the hypothesis that local human advanced plaque characteristics can be influenced by clinically relevant common sequence variation. Of 47 variants that previously associated with CAD or LAS, we detected 11 variants that associated with a plaque characteristic. These observations require confirmation, but the number of positive results is encouraging since phenotyping of atherosclerotic lesions was solely based on histology. Reducing the noise in phenotype measurements through alternative deep- phenotyping methods might be a feasible way to increase power, and likely to unravel more relevant associations24,25. The second strongest associations were with calcification (rs9982601 near KCNE2) and intraplaque hemorrhage (rs11203042 near LIPA). A previous study on coronary calcification found no evidence for association of rs998260126, which was also the case for aortic calcification in another study (upon lookup, p > 0.98)27 . Our results suggest that variation near KCNE2 may have different effects leading to calcification depending on the arterial bed. Inflamed plaques are prone to rupture and give rise to symptoms, and such plaques often show signs of intraplaque hemorrhage28. Indeed, expression analyses confirm that ruptured plaques, identified through high expression of CD16329, express variety of inflammatory genes – one of these also includes LIPA. Mouse models impaired for LIPA show accelerated atherosclerosis and spontaneous thrombosis, thought to derive from an impaired lipid-trafficking in macrophages leading to the accumulation of fat and eventually plaque rupture30.
20162300 Marten Siemelink.indd 74 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE CHARACTERISTICS | 75
Further notable associations were for variants near TRIB1 and ApoE-ApoC1. Genes in both loci have been implicated in lipid metabolism before31,32, and our results suggest this effect extends into carotid plaques. We further explored one variant at 7q22, that associated with lipid content of the carotid lesions in the AE. Strong LD between variants extends between 106.7 and 107.2 Mb in this locus (Figure 1), complicating the identification of causal gene(s). However, results from the meQTL analyses in the AE, the eQTL analyses in BiKE and STAGE, and the lookup in data from Zeller at al.21 converge around only a few genes at 7q22: especially HBP1, COG5, and GPR22. The CAD risk associated variant rs12539895 associates with gene regulation and expression, and less fatty lesions in relevant tissues with directionally consistent effects. Thus, although the current study is unable to point to the causal gene, clearly it prioritized the above genes. HMG-box transcription factor 1 (HBP1) is a transcriptional repressor binding to promoter regions; it plays a role in Wnt1 signaling33 and cancer34,35; and is mostly expressed in the 4 heart. In a rat model of carotid injury therapeutic inhibition of Hbp1 (also known as Hmgb1 in rodents) was associated with a decrease in vascular SMC activation and neointimal formation36. Another study showed that Hbp1 is targeted by miR-15537. Inhibition of miR- 155 attenuated foam cell formation and reduced atherosclerotic plaques in ApoE-/- mice, and Hbp1 knockdown enhanced lipid uptake by macrophages37. Our eQTL analyses in BiKE are the first to confirm these results in humans, as the risk-reducing alleles of rs12539895 (and its proxies), increase Hbp1 expression in carotid plaques (Supplemental Figure 2 and Supplemental Table 8), while decreasing risk of CAD and leading to less fatty lesions. Supportive evidence derives from our meQTL analyses as rs80341862 (which is in LD with rs12539895) decreases methylation – and potentially thereby increases gene expression – near the 3’UTR (the enhancer region) of HBP1 and COG5. COG5 (component of oligomeric golgi complex 5) has not been associated with atherosclerotic disease or traits previously. COG5 is part of an evolutionary conserved protein complex in the Golgi apparatus that is involved in protein and lipid trafficking and sorting38. It plays a major role in glycosylation of proteins38. One report discusses a mutation near the 3’ UTR of COG5 leading to impaired glysolyation of proteins, one of which was ApoC-III39. Another study reported that COG5 depletion in HeLa cells did not significantly affected glycosylation or membrane expression of LDLR40. Strikingly, COG5 knockdown in these cells lead to a significant dilation of Golgi-apparatus cisternae as well as an impaired intracellular trafficking leading to an accumulation of intracellular LDLR40. In this study we confirmed that COG5 expression negatively correlates with LDLR expression in carotid plaques. The CAD risk reducing allele of rs80341862 (a rs12539895 proxy) also associated with decreased COG5 expression in atherosclerotic arterial wall, internal mammary artery tissue and whole blood tissues, as well as decreased methylation at the 3’UTR of COG5 in plaques, and less fatty lesions (Supplemental Table 8). The orphan GPR22 (G-protein-coupled receptor 22) is part of the G-protein coupled receptor 1 family and involved in membrane transport. One study reported on the myocardial expression patterns in human and rodent hearts, and revealed evidence for a protective role in response to hemodynamic stress in GPR22-/- mice. DUS4L (dihydrouridine synthase 4-like) catalyzes synthesis of dihydrouridine which is a modified base found in transfer RNAs. DLD (dihydrolipoamide dehydrogenase) is a moonlighting protein. In a homodimeric form it functions as a dehydrogenase regulating energy metabolism, while as a monomer
20162300 Marten Siemelink.indd 75 15-03-16 13:23 76 | CHAPTER 4
it functions as a protease41. For neither DUS4L or DLD a role in atherosclerosis was ever reported (PubMed search with “[gene name]” and “atherosclerosis” on November 3rd, 2015). Our measure of fatty lesions is an overall measure of plaque fat content; thus we cannot ascertain which fraction of lipids might be involved. Yet, the association of 7q22 with circulating lipids, rather than other cardiovascular risk factors, implicates a role for lipid metabolism that may lead to less fatty carotid plaques and a reduction in CAD risk. This is further substantiated by the regional interaction of rs12539895 with genes in monocytes, and the strong correlation of carotid plaque gene expression with LDLR expression. Our tissue specific gene regulatory and transcription analyses are consistently showing effects of rs12539895 and its proxies on HBP1 and COG5, and GPR22 in both (atherosclerotic) arterial wall tissues and monocytes. As the scavenger cell in atherosclerotic arterial wall tissue, monocytes act to clear lipid particles from plaques; the eQTL effects found in atherosclerotic tissues might well be due to the presence of monocytes. Power may have impeded the identification of CpGs associating with fatty lesions. We were also unable to directly test tissue-specific gene expression to fatty lesions; thus further studies are needed to assess such effects. Clinically relevant genetic variants associated with histological proxies of vulnerable carotid plaque characteristics. Further validation and verification of our findings are needed in future external cohorts and functional studies, prioritizing HBP1, COG5 and GPR22 in tissues relevant to atherosclerosis. Genetic analyses such as ours in deeply phenotyped patient cohorts, could aid in prioritizing therapeutic targets among clinically relevant loci identified through meta-analyses of GWAS.
Methods and Material
Study population To study the impact of the CAD and LAS associated variants on plaque characteristics and plaque-derived DNA methylation, we used data from the Athero-Express Genomics Study (AEGS, n = 1,443), which is part of the Athero-Express Biobank Study (AE, Supplemental Table 4, http://www.atheroexpress.nl)8,42. Expression quantitative trait loci (eQTL) analyses were done using data from the Biobank of Karolinska Endarterectomy (BiKE, n = 127) 15,43- 45 and the Stockholm Atherosclerosis Gene Expression (STAGE, n = 109)14 studies. These studies are comparable in terms of inclusion of patients with significant coronary (STAGE) or carotid (BiKE and AE) stenosis and their study designs were described before. In brief, in the AE blood and plaques are obtained, stored at -80°C and plaque material is routinely used for histological analysis8,9. BiKE includes patients with a comparable clinical presentation and uses plaque and blood material for expression analyses. STAGE comprises well-characterized CAD patients undergoing coronary artery bypass grafting (CABG) surgery in the Karolinska Hospital, Stockholm, Sweden14. During CABG samples from atherosclerotic arterial wall (AAW), internal mammary artery (IMA), liver, subcutaneous fat (SF), skeletal muscle (SM), visceral fat (VF), and fasting whole blood (WB) were taken for DNA and RNA extraction. Informed consent was available for patients. The medical ethical committees of the respective hospitals approved these studies.
20162300 Marten Siemelink.indd 76 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE CHARACTERISTICS | 77
Plaque phenotyping The (immune)histochemical analysis protocols used in the AE have been described in detail before8,9,46. In short, for this study we focused on immunohistological markers in atherosclerotic plaques. We quantitatively scored the number of macrophages (CD68) and smooth muscle cells (SMCs, α-actin), as percentage plaque area. In a subset we counted neutrophils (CD66b)46 and mast cells (α-trypsin)47. Intraplaque vessel density (CD34) was assessed as the average number per 3 hotspots. Intraplaque hemorrhage (IPH) was scored as no/yes. Intraplaque atheroma size was defined as less or more than 10% atheroma per total plaque area. Calcification (hematoxylin & eosin) and collagen (picrosirius red) were binary scored as no/minor vs. moderate/heavy staining.
Genotyping, imputation and gene expression The methods of genotyping and imputation in the AE were described previously42. Briefly, in two experiments 1,858 patients were genotyped using the Affymetrix SNP 5.0 (Athero- 4 Express Genomics Study 1, AEGS1, n = 836) and Affymetrix Axiom CEU arrays (Athero- Express Genomics Study 2, AEGS2, n = 1,022). We applied community standard quality control48,49 to obtain high quality data in AEGS1 (403,789 markers in 571 CEA patients) and AEGS2 (535,983 markers in 868 CEA patients). Autosomal missing genotypes were imputed in both datasets separately based on phased integrated data from 1000G (phase 1, version 3)50 using IMPUTE2 (v2.3.0)51 after pre-phasing with SHAPEIT2 (v2.644)52. The BiKE genotyping and microarray analyses have been previously described15. The microarray dataset is available from Gene Expression Omnibus (accession number GSE21545). Briefly, high-density genotyping was performed by Ilumina 610w-QuadBead SNP-chips and microarray profiling using Affymetrix HG-U133 plus 2.0 Genechip arrays from n = 127 patient carotid plaque samples. For the expression quantitative trait locus (eQTL) analysis all genotyped variants (n = 146) between 106,591,849 and 107,591,849 bp on chromosome 7 (7q22 locus) were tested for association with regional gene expression in plaques. Details of genotyping and expression analysis in STAGE have been previously described14. In short, for imputation, data was filtered on minor allele frequency (MAF < 5%), Hardy- Weinberg Equilibrium (HWE p<1.0x10-6, and a call rate of 100%. IMPUTE2 and SHAPEIT were used for imputation with 1000 Genomes EUR (phase 1, version 3) as the reference53. For eQTL analysis all variants at the 7q22 locus were tested for association with reginal gene expression in all 7 tissues using Matrix eQTL54.
Measurement, quality control and analysis of DNA methylation As part of the Athero-Express Methylation Study, DNA from 509 CEA patients was extracted from 506 plaque samples and 94 blood samples (of which 91 overlap with the 506 plaque samples). We used the Illumina Infinium HumanMethylation450 BeadChip Array according to the manufacturer’s protocol to determine tissue-specific DNA methylation. Upon quality control using the MethylAid R-package55 a total of 443,872 (91.4%) probes remained. The data were normalized and corrected for batch effects using ‘Functional Normalization’56 as implemented in the minfi R-package57. A total of 488 plaque samples and 91 blood samples overlap with the 1,443 genotyped CEA patients. For the methylation quantitative trait locus (meQTL) analysis we tested all variants at 7q22 for association with 160 CpGs in the same region. A total of 3,130 variants were imputed in this region in our data, but 1,721 variants remained after filtering on MAF > 0.005,
20162300 Marten Siemelink.indd 77 15-03-16 13:23 78 | CHAPTER 4
imputation quality > 0.9 and Hardy-Weinberg Equilibrium p-value > 1.0x10-6. For the meQTL analysis we used FastQTL (http://fastqtl.sourceforge.net) and ran 1,000 permutations to derive empirical p-values20. These analyses were corrected for age, sex, the first 2 genetic principal components, the genotyping chip, year of surgery, and the first 2 principal components based on the methylation data. Q-values of the false-discovery rate were calculated using the Storey & Tibsjirani formula and based on the empirical p-values derived after permutation58.
Single-nucleotide variant association study For the individual single-nucleotide variant analyses we focused on the GWAS hits for CAD, LAS, and BD (Supplemental Table 3). Although the two reported variants for LAS (rs2107595 and rs2383207) map to the same genomic regions (HDAC9 and CDKN2BAS1, respectively), they are in modest linkage disequilibrium with the two CAD associated variants that map to the same region (rs2023938, r2 = 0.44, and rs3217922, r2 = 0.37, respectively using SNiPA with 1000G phase 1, version 3 as a reference59). Thus, in total there were 51 independent SNVs analyzed. An additive genetic model was assumed for SNV analyses.
Weighted polygenic scores Weighted polygenic scores for all CEA patients in the combined AEGS1 and AEGS2 dataset (AEGS) were calculated based on the reported estimated effects in the same way as described above60. As a negative control we calculated similar polygenic scores for bipolar disorder (BD). Further details are described in the supplement.
Gene-based association study We used the VEGAS2 software package to analyze the loci in a gene-based manor as described before18. We then ran VEGAS2 on these results to obtain per gene empirical p-values. Further details are described in the supplement.
Statistical analysis For statistical analysis quantitative plaque phenotypes measurements were normalized by Box-Cox transformation; outliers deviating more than ±3 s.d. from the transformed mean were removed. All single variant and polygenic models were analyzed in a linear or logistic regression framework corrected for age, sex, year of surgery, dataset (AEGS1 or AEGS2), and principal components. Nagelkerke’s r2 was used as a metric of the variance explained by the polygenic model. For single variant analyses we considered a variant significantly associated after Bonferroni correction for SNVs (n = 51) and phenotypes (n = 7), thus p = 0.05/(51 x 7) = 1.40x10-4. Likewise, for the gene-based analyses we considered p = 0.05 / (151 genes x 7 phenotypes) = 4.73x10-5 significant. Results from the regional genetic, epigenetic, meQTL, and eQTL analyses were Bonferroni corrected accordingly. For completeness we also considered variants associated with methylation or expression after correction by false-discovery rate, q-value (Q). IBM® SPSS® Statistics version 20 (release 20.0.0, IBM Corp., Armonk, NY, USA, http://www.ibm.com) was used for statistical analyses of baseline characteristics. PLINK v1.9 (beta 3g release March 5th, 2015)61, SNPTEST v2.5.162 and Golden Helix SNP & Variation Suite v8.3.3 MacOSX (Golden Helix, Inc., Bozeman, MT, USA, http://www.goldenhelix.com) were used for quality control and analysis of genetic data in the AE.
20162300 Marten Siemelink.indd 78 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE CHARACTERISTICS | 79
Acknowledgements We would like to thank Freerk van Dijk and Morris Swertz and acknowledge them graciously for imputing our datasets using the “GoNL Impute2” pipeline. Evelyn Velema and Petra Homoet-Van der Kraak are graciously acknowledged for the immunohistochemical stainings.
4
20162300 Marten Siemelink.indd 79 15-03-16 13:23 80 | CHAPTER 4
References
1. Lusis, A. J. Atherosclerosis. Nature 407, 233–241 (2000). 2. Kathiresan, S. & Srivastava, D. Genetics of Human Cardiovascular Disease. Cell 148, 1242–1257 (2012). 3. Dichgans, M. et al. Shared Genetic Susceptibility to Ischemic Stroke and Coronary Artery Disease A Genome-Wide Analysis of Common Variants. Stroke 45, 24–36 (2014). 4. Traylor, M. et al. Genetic risk factors for ischaemic stroke and its subtypes (the METASTROKE col- laboration): a meta-analysis of genome-wide association studies. 11, 951–962 (2012). 5. CARDIoGRAMplusC4D Consortium et al. Large-scale association analysis identifies new risk loci for coronary artery disease. Nature Genetics 45, 25–33 (2013). 6. Kamb, A., Harper, S. & Stefansson, K. Human genetics as a foundation for innovative drug develop- ment. Nature Genetics 31, 975–978 (2013). 7. Pitt, G. S. Cardiovascular precision medicine: hope or hype? European Heart Journal 36, ehv226–1843 (2015). 8. Verhoeven, B. A. N. et al. Athero-express: differential atherosclerotic plaque expression of mRNA and protein in relation to cardiovascular events and patient characteristics. Rationale and design. Eur J Epidemiol 19, 1127–1133 (2003). 9. van Lammeren, G. W. G. et al. Atherosclerotic plaque vulnerability as an explanation for the increased risk of stroke in elderly undergoing carotid artery stenting. Stroke 42, 2550–2555 (2011). 10. Hellings, W. E. et al. Composition of Carotid Atherosclerotic Plaque Is Associated With Cardiovascular Outcome: A Prognostic Study. Circulation 121, 1941–1950 (2010). 11. Virmani, R., Kolodgie, F. D., Burke, A. P., Farb, A. & Schwartz, S. M. Lessons from sudden coronary death: a comprehensive morphological classification scheme for atherosclerotic lesions. ATVB 20, 1262–1275 (2000). 12. Ross, R. Atherosclerosis--an inflammatory disease. N Engl J Med 340, 115–126 (1999). 13. Verhoef, T. I. et al. A Randomized Trial of Genotype-Guided Dosing of Acenocoumarol and Phenprocou- mon. N Engl J Med 369, 2304–2312 (2013). 14. Hägg, S. et al. Multi-organ expression profiling uncovers a gene module in coronary artery disease involving transendothelial migration of leukocytes and LIM domain binding 2: the Stockholm Athero- sclerosis Gene Expression (STAGE) study. PLoS Genet. 5, e1000754 (2009). 15. Folkersen, L. et al. Association of Genetic Risk Variants With Expression of Proximal Genes Identi- fies Novel Susceptibility Genes for Cardiovascular Disease. Circulation: Cardiovascular Genetics 3, 365–373 (2010). 16. CARDIoGRAMplusC4D Consortium. A comprehensive 1000 Genomes-based genome-wide associa- tion meta-analysis of coronary artery disease. Nature Genetics 47, 1121–1130 (2015). 17. Liu, J. Z. et al. A versatile gene-based test for genome-wide association studies. Am. J. Hum. Genet. 87, 139–145 (2010). 18. Mishra, A. & Macgregor, S. VEGAS2: Software for More Flexible Gene-Based Testing. Twin Res Hum Genet 18, 86–91 (2014). 19. Grundberg, E. et al. Mapping cis- and trans-regulatory effects across multiple tissues in twins. Nature Genetics 44, 1084–1089 (2012). 20. Ongen, H., Buil, A., Brown, A., Dermitzakis, E. & Delaneau, O. Fast and efficient QTL mapper for thousands of molecular phenotypes. bioRxiv (2015). 21. Zeller, T. et al. Genetics and beyond--the transcriptome of human monocytes and disease susceptibil- ity. PLoS ONE 5, e10693–e10693 (2009). 22. Global Lipids Genetics Consortium et al. Discovery and refinement of loci associated with lipid levels. Nature Genetics 45, 1274–1283 (2013). 23. The International Schizophrenia Consortium. Common polygenic variation contributes to risk of schizo- phrenia and bipolar disorder. Nature 460, 748–752 (2009). 24. Nelissen, B. G. L., van Herwaarden, J. A., Moll, F. L., van Diest, P. J. & Pasterkamp, G. SlideToolkit: an assistive toolset for the histological quantification of whole slide images. PLoS ONE 9, e110289 (2014). 25. Vrijenhoek, J. E. P. et al. High Reproducibility of Histological Characterization by Whole Virtual Slide Quantification; An Example Using Carotid Plaque Specimens. PLoS ONE 9, e115907 (2014). 26. O’Donnell, C. J. et al. Genome-Wide Association Study for Coronary Artery Calcification With Follow- Up in Myocardial Infarction. Circulation 124, 2855–2864 (2011). 27. van Setten, J. et al. Genome-wide association study of coronary and aortic calcification implicates risk loci for coronary artery disease and myocardial infarction. Atherosclerosis 228, 400–405 (2013). 28. Fuster, V., Moreno, P. R., Fayad, Z. A., Corti, R. & Badimon, J. J. Atherothrombosis and high-risk plaque: part I: evolving concepts. JACC 46, 937–954 (2005).
20162300 Marten Siemelink.indd 80 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE CHARACTERISTICS | 81
29. Puig, O., Yuan, J., Stepaniants, S. & Zieba, R. A gene expression signature that classifies human atherosclerotic plaque by relative inflammation status. Circulation (2011). doi:10.1161/CIRCGENET- ICS.111.960773/-/DC1 30. Hopkins, P. N. Molecular Biology of Atherosclerosis. Physiological Reviews 93, 1317–1542 (2013). 31. Jong, M. C., Hofker, M. H. & Havekes, L. M. Role of ApoCs in lipoprotein metabolism: functional dif- ferences between ApoC1, ApoC2, and ApoC3. ATVB 19, 472–484 (1999). 32. Burkhardt, R. et al. Trib1 is a lipid- and myocardial infarction-associated gene that regulates hepatic lipogenesis and VLDL production in mice. J. Clin. Invest. 120, 4410–4414 (2010). 33. Sampson, E. M. et al. Negative regulation of the Wnt-beta-catenin pathway by the transcriptional repressor HBP1. EMBO J 20, 4500–4511 (2001). 34. Paulson, K. E. et al. Alterations of the HBP1 transcriptional repressor are associated with invasive breast cancer. Cancer Res 67, 6136–6145 (2007). 35. Lemercier, C. et al. Involvement of retinoblastoma protein and HBP1 in histone H1(0) gene expression. Mol Cell Biol 20, 6627–6637 (2000). 36. Chen, J. et al. Inhibition of neointimal hyperplasia in the rat carotid artery injury model by a HMGB1 inhibitor. Atherosclerosis 224, 332–339 (2012). 37. Tian, F.-J. et al. Elevated microRNA-155 promotes foam cell formation by targeting HBP1 in atherogen- esis. Cardiovascular Research 103, 100–110 (2014). 38. Smith, R. D. & Lupashin, V. V. Role of the conserved oligomeric Golgi (COG) complex in protein glyco- 4 sylation. Carbohydr. Res. 343, 2024–2031 (2008). 39. Paesold-Burda, P. et al. Deficiency in COG5 causes a moderate form of congenital disorders of glyco- sylation. Human Molecular Genetics 18, 1–7 (2009). 40. Oka, T. et al. Genetic analysis of the subunit organization and function of the conserved oligomeric golgi (COG) complex: studies of COG5- and COG7-deficient mammalian cells. Journal of Biological Chemistry 280, 32736–32745 (2005). 41. Babady, N. E., Pang, Y.-P., Elpeleg, O. & Isaya, G. Cryptic proteolytic activity of dihydrolipoamide dehy- drogenase. Proceedings of the National Academy of Sciences 104, 6158–6163 (2007). 42. van der Laan, S. W. et al. Variants in ALOX5, ALOX5AP and LTA4H are not associated with atheroscle- rotic plaque phenotypes: The Athero-Express Genomics Study. Atherosclerosis 239, 528–538 (2015). 43. Razuvaev, A. et al. Correlations between clinical variables and gene-expression profiles in carotid plaque instability. Eur J Vasc Endovasc Surg 42, 722–730 (2011). 44. Folkersen, L. et al. Prediction of ischemic events on the basis of transcriptomic and genomic profiling in patients undergoing carotid endarterectomy. Mol. Med. 18, 669–675 (2012). 45. Perisic, L. et al. Profiling of atherosclerotic lesions by gene and tissue microarrays reveals PCSK6 as a novel protease in unstable carotid atherosclerosis. ATVB 33, 2432–2443 (2013). 46. van den Borne, P. et al. Leukotriene b4 levels in human atherosclerotic plaques and abdominal aortic aneurysms. PLoS ONE 9, e86522 (2014). 47. Willems, S. et al. Mast cells in human carotid atherosclerotic plaques are associated with intraplaque microvessel density and the occurrence of future cardiovascular events. European Heart Journal 34, 3699–3706 (2013). 48. Laurie, C. C. et al. Quality control and quality assurance in genotypic data for genome-wide association studies. Genet. Epidemiol. 34, 591–602 (2010). 49. Anderson, C. A. et al. Data quality control in genetic case-control association studies. Nature Protocols 5, 1564–1573 (2010). 50. The International HapMap Consortium. A haplotype map of the human genome. Nature 437, 1299– 1320 (2005). 51. Howie, B., Fuchsberger, C., Stephens, M., Marchini, J. & Abecasis, G. R. Fast and accurate genotype imputation in genome-wide association studies through pre-phasing. Nature Genetics 44, 955–959 (2012). 52. Delaneau, O., Marchini, J. & Zagury, J. F. A linear complexity phasing method for thousands of ge- nomes. Nature methods 9, 179–181 (2012). 53. 1000 Genomes Project Consortium et al. A map of human genome variation from population-scale sequencing. Nature 467, 1061–1073 (2010). 54. Shabalin, A. A. Matrix eQTL: ultra fast eQTL analysis via large matrix operations. Bioinformatics 28, 1353–1358 (2012). 55. van Iterson, M. et al. MethylAid: visual and interactive quality control of large Illumina 450k datasets. Bioinformatics 30, 3435–3437 (2014). 56. Fortin, J.-P. et al. Functional normalization of 450k methylation array data improves replication in large cancer studies. Genome Biol 15, 503 (2014). 57. Aryee, M. J. et al. Minfi: a flexible and comprehensive Bioconductor package for the analysis of In- finium DNA methylation microarrays. Bioinformatics 30, 1363–1369 (2014).
20162300 Marten Siemelink.indd 81 15-03-16 13:23 82 | CHAPTER 4
58. Storey, J. D. & Tibshirani, R. Statistical significance for genomewide studies. Proceedings of the Na- tional Academy of Sciences 100, 9440–9445 (2003). 59. Arnold, M., Raffler, J., Pfeufer, A., Suhre, K. & Kastenmüller, G. SNiPA: an interactive, genetic variant- centered annotation browser. Bioinformatics 1–2 (2014). 60. Psychiatric GWAS Consortium Bipolar Disorder Working Group. Large-scale genome-wide associa- tion analysis of bipolar disorder identifies a new susceptibility locus near ODZ4. Nature Genetics 43, 977–983 (2011). 61. Purcell, S. et al. PLINK: A Tool Set for Whole-Genome Association and Population-Based Linkage Analyses. The American Journal of Human Genetics 81, 559–575 (2007). 62. Marchini, J., Howie, B., Myers, S., McVean, G. & Donnelly, P. A new multipoint method for genome- wide association studies by imputation of genotypes. Nature Genetics 39, 906–913 (2007).
20162300 Marten Siemelink.indd 82 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE CHARACTERISTICS | 83
Supplemental Material
Supplement to: van de Laan, SW et al, Coronary artery disease and large artery stroke loci are associated with human atherosclerotic plaque characteristics. Supplemental material available on publication and on request.
Table of contents Supplemental Text
Supplemental Tables Supplemental Table 1. Sources of summary level GWAS results Supplemental Table 2. Polygenic models. Supplemental Table 3. General statistics of reported GWAS hits for coronary artery disease (CAD), large artery ischemic stroke (LAS), and bipolar 4 disorder (BD) in AEGS. Supplemental Table 4. Plaque phenotype associated genes mapped to GWAS loci. Supplemental Table 5. Additional variants associated with DNA methylation in plaques based on FDR Q ≤ 0.05. Supplemental Table 6. Association of carotid plaque DNA methylation with fatty lesions. Supplemental Table 7. Baseline characteristics of the Athero-Express Genomics Study (AEGS), Biobank of Karolinska Endarterectomy (BiKE), and Stockholm Atherosclerosis Gene Expression (STAGE). Supplemental Table 8. 7q22 regional association of variants with gene expression in 7 tissues in STAGE based on Q ≤ 0.05. Supplemental Table 9. 7q22 regional association of variants with gene expression in carotid plauqes from BiKE. Supplemental Table 10. Correlation of gene expression at 7q22 with LDLR expression in carotid plaques. Supplemental Table 11. Genes interacting with rs12539895 as identified through circularized chromosome conformation capture (4C). Supplemental Table 12. Association of rs12539895 with risk factors of CAD.
Supplemental Figures Supplemental Figure 1. Frequency and QQ-plot of the eQTL analyses in liver (A and B) and whole blood (C and D) tissues from in STAGE. Supplemental Figure 2. Association of rs1548524 with HBP1 expression.
Supplemental References
20162300 Marten Siemelink.indd 83 15-03-16 13:23 M.A. Siemelink1 S.W. van der Laan1 J. van Setten1 J.P.P.M. de Vries2 G.J. de Borst3 F.L. Moll3 H.M. den Ruijter1 F.W. Asselbergs4,5,6 G. Pasterkamp1,7 P.I.W. de Bakker8,9
1 Laboratory of Experimental Cardiology, Division Heart and Lungs, University Medical Center Utrecht, Utrecht, the Netherlands 2 Department of Vascular Surgery, St. Antonius Hospital, Nieuwegein, the Netherlands 3 Department of Vascular Surgery, University Medical Center Utrecht, Utrecht, the Netherlands 4 Department of Cardiology, Division Heart and Lungs, University Medical Center Utrecht, Utrecht, the Netherlands 5 Durrer Center for Cardiogenetic Research, ICIN-Netherlands Heart Institute, Utrecht, the Netherlands 6 Institute of Cardiovascular Science, Faculty of Population Health Sciences, University College London, London, United Kingdom 7 Laboratory of Clinical Chemistry and Hematology, Division Laboratories and Pharmacy, University Medical Center Utrecht, Utrecht, the Netherlands 8 Department of Medical Genetics, Center for Molecular Medicine, University Medical Center Utrecht, Utrecht, the Netherlands 9 Department of Epidemiology, Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, the Netherlands
20162300 Marten Siemelink.indd 84 15-03-16 13:23 CHAPTER 5
Common Variants associated with Blood Lipid Levels do not Affect Carotid Plaque Composition
MANUSCRIPT IN PREPARATION
20162300 Marten Siemelink.indd 85 15-03-16 13:23 86 | CHAPTER 5
Abstract
Introduction Although plasma lipid levels are known to influence the risk of cardiovascular disease (CVD), little is known about their effect on atherosclerotic plaque composition. To date, large-scale genome-wide association studies have identified 157 common single-nucleotide polymorphisms (SNPs) that influence plasma lipid levels, providing a powerful tool to investigate the effect of plasma lipid levels on atherosclerotic plaque composition.
Methods In this study, we included 1443 carotid endarterectomy patients from the Athero-Express Biobank Study with genotype data. Plasma concentrations of high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), total cholesterol (TC) and triglycerides (TG) were determined at the time of endarterectomy. Atherosclerotic plaques, obtained during surgery, were histologically examined. For all patients, we calculated weighted genetic burden scores (GBS) for all lipid traits on the basis of the available genotype data. Plasma lipid levels and GBS were tested for association with 7 histological features using linear and logistic regression models.
Results
All GBS were associated with their respective plasma lipid concentrations (pHDL-C = 2.4 x -14 -6 -8 10 , pLDL-C = 0.003, pTC = 2.1 x 10 , pTG = 3.4 x 10 ). Neither the measured plasma lipids, nor the GBS, were associated with histological features of atherosclerotic plaque composition. In addition, neither the plasma lipids nor the GBS were associated with clinical endpoints within 3 years of follow-up, with the notable exception of a negative association between HDL-C and composite cardiovascular endpoints.
Conclusion This study found no evidence that plasma lipid levels or their genetic determinants influence carotid plaque composition.
20162300 Marten Siemelink.indd 86 15-03-16 13:23 BLOOD LIPID VARIANTS DO NOT AFFECT PLAQUE COMPOSITION | 87
Introduction
There is a wealth of evidence that plasma lipid levels influence the risk of cardiovascular disease (CVD). Most notably, Mendelian randomization studies have contributed unambiguous support for a causal effect of LDL-C and TG on CVD, while the evidence for the causality of HDL-C is perhaps not as clear.1–6 Yet, for all the accumulated evidence on plasma lipids, the biological mechanisms by which they increase CVD risk are still incompletely understood. The traditional paradigm of atherosclerosis has been the accumulation of plasma lipids in the vascular wall contributing to the formation of an unstable plaque, the rupture of which causes a cardiovascular event. Much research has been performed in animal models and in vitro studies, but little is known about changes to human atherosclerotic plaque composition due to changed plasma lipid concentrations. In addition to known determinants of plasma lipid levels such as diet, metabolism and body weight, it is clear that there is a heritable polygenic basis for inter-individual variation in lipid levels.7 A large meta-analysis of genome-wide association studies by the Global Lipids Genetics Consortium (GLGC) identified 157 common single-nucleotide polymorphisms (SNPs) associated with high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), total cholesterol (TC) and triglycerides (TG) blood concentrations.8 5 Although the individual effects of these SNPs on plasma lipid concentrations are modest, the cumulative impact of multiple variants may be substantial. Since these SNPs confer a lifelong effect, they provide a useful tool to investigate the possible effects of plasma lipid levels on atherosclerotic plaque composition. The Athero–Express Biobank Study (AE) has enrolled patients who underwent carotid endarterectomy (CEA) followed by extensive histological analysis of carotid plaque characteristics, genotyping and a 3 year follow-up for clinical endpoints. We investigated the association of plasma lipid concentrations with 7 features of atherosclerotic plaque composition in the Athero-Express. Subsequently, we performed similar associations using weighted genetic burden scores (GBS) to assess the cumulative effect of the known 157 lipid associated SNPs on plaque composition. In addition, we investigated if plasma lipids, or their genetic determinants, were associated with clinical endpoints during follow-up.
Methods
Inclusion The Athero-Express Biobank Study is an ongoing multi-center cohort study including patients that underwent carotid endarterectomy and was previously described in more detail.9,10 In short, blood and atherosclerotic plaques were obtained during surgery and stored at -80°C. Clinical data were obtained through standardized questionnaires, pre- operative admission charts and patient medical files. Medical ethics committees of both hospitals approved the study and all patients included in the study provided informed consent.
Plaque histology Atherosclerotic plaque specimens were collected during carotid endarterectomy, processed and analyzed according to a standardized and previously reported protocol.9 In short, the
20162300 Marten Siemelink.indd 87 15-03-16 13:23 88 | CHAPTER 5
specimen was paraffin-embedded, the culprit lesion was identified and a 5 micron cross- section was sliced, stained and quantified by microscopy. Hematoxylin and eosin (H&E) staining was used for assessment of calcifications and atheroma. Staining with Elastin von Gieson (EVG) was used for plaque hemorrhage and Picrosirius Red for assessment of collagen. Immunohistochemical staining was performed for assessment of macrophages (anti-CD68), microvessels (anti-CD34) and smooth-muscle cells (anti-alpha smooth muscle actin). The presence of calcifications and collagen were classified as scarce/absent or high. Atheroma was classified as below of above 10% of the plaque area. Plaque hemorrhage was classified as present or absent. Plaque microvessels were quantified by number of individual vessels per microscopy field. Plaque smooth-muscle cells and macrophages were quantified as percentage of the microscopy field area. For quantitative measures, multiple random microscopy fields were digitized, quantified and averaged.
Laboratory measurements If routine clinical plasma lipids measurements were performed prior to surgery, that data was used. If available, stored pre-operative plasma samples were used for additional plasma lipids measurements at the clinical chemistry laboratory of the University Medical Center Utrecht. If both routine and additional lipids values were available the values were averaged (correlations routine vs. additional measurements: Spearman’s ρ > 0.94 for all lipid traits).
SNP Genotyping and Imputation Details on genotyping, quality control and imputation have been published previously.11 In brief, DNA was extracted from whole blood, or alternatively from plaque samples, following standardized in-house validated protocols. Genotyping was done using commercially available genotyping chips. The first batch (Athero-Express Genomics Study 1, AEGS1) was genotyped using Affymetrix Genome-Wide Human SNP Array 5.0, the second batch (Athero-Express Genomics Study 2, AEGS2) was genotyped using the Affymetrix Axiom® GW CEU 1 Array (Affymetrix, Santa Clara, CA, USA). We adhered to community standard quality control and assurance (QCA) procedures to clean the genotype data obtained in AEGS1 (N=571) and AEGS2 (N=868).12 We used the 998 phased haplotypes from the Genome of the Netherlands Project release 4 (GoNL4) encompassing 19,763,454 SNPs as the reference panel for imputation.13
Weighted Genetic Burden Scores The Global Lipids Genetics Consortium identified 157 loci associated with circulating lipid levels, and reported their effect sizes on plasma lipids (included in online supplement). We calculated weighted genetic burden scores using the risk alleles from these SNPs and their respective effect sizes. To account for the imputation quality we used the risk allele dosages to construct the GBS. The following formula was used to calculate the GBS for each lipid14–16:
GBSlipid = Σ (βn*Dn)
th Where an individual’s GBSlipid is the sum of the effect size (β) of the risk allele of the n SNP multiplied by the individual’s dosage (D) of the risk allele for the nth SNP.
20162300 Marten Siemelink.indd 88 15-03-16 13:23 BLOOD LIPID VARIANTS DO NOT AFFECT PLAQUE COMPOSITION | 89
Follow-up and clinical outcome Clinical follow-up was performed by contacting patients to fill-in standardized questionnaires at 1, 2 and 3 years after surgery. In case no response to the questionnaire was received the general practitioner was contacted for information. All events were assessed by two members of an outcome assessment committee. Cardiovascular death (CV-death) was defined as death of presumed vascular origin (stroke, myocardial infarction, sudden death, other vascular causes). Composite (cardiovascular) endpoints were defined as the occurrence of any cardiovascular event including CV-death, non-fatal stroke or myocardial infarction, or any vascular intervention not planned at the time of inclusion. Restenosis was determined based on routine clinical assessment of ipsilateral carotid artery stenosis using duplex echocardiography at 1 year post carotid endarterectomy.
Statistics Statistical analysis was performed in RStudio using R (v3.1.2). Regression analysis was performed using an additive linear model for continuous traits and a logistic model for categorical traits correcting for age, sex, hospital of surgery, diabetes or statin use, a dummy-variable representing the genotyping batch (AEGS1 or AEGS2), and the first 10 principal components. Prior to analysis, outliers at more than three standard deviations 5 were removed from the data and traits that were not normally distributed (smooth-muscle cells, macrophages, microvessels) were normalized using Box-Cox transformation. We used PLINK17 (version 1.7) to determine the associations of individual SNPs.
Results
Cohort characteristics Clinical cohort characteristics are summarized in table 1. The Athero-Express cohort is comprised of patients with advanced stages of atherosclerosis, as is evident from a high prevalence of atherosclerotic disease in various vascular territories (cerebrovascular accident 82.3%, coronary artery disease 29.9%, peripheral artery occlusive disease 20.7%). Associations of individual plasma lipids and GBS with risk factors are shown in Supplemental Tables 1 and 2.
GBS associate with circulating lipids To ascertain the validity of the calculated GBS as a means to investigate the relationship between plasma lipids and CVD, we first confirmed the relationship between GBS and measured plasma lipids. All the GBS showed significant associations with their respective -14 -6 plasma lipid concentrations (pHDL-C = 2.4 x 10 , pLDL-C = 0.003, pTC = 2.1 x 10 , pTG = 3.4 x 10-8) and corrected plasma lipids showed an increasing trend with higher GBS for all lipid traits (Figure 1A-D). This replicates previous studies, and confirms the effects of these SNPs on plasma lipids in this cohort of patients with severe atherosclerotic disease. Remarkably, there was only a moderate association between the LDL burden score and measured LDL-C levels, which may be explained by statin-induced favorable LDL-C levels, irrespective of the genetic burden (Supplemental Figure 1). Additionally, we investigated all SNPs individually for their association with each of the plasma lipids (Supplemental Table 3), yet none of the SNPs were significant after Bonferroni correction.
20162300 Marten Siemelink.indd 89 15-03-16 13:23 90 | CHAPTER 5
Table 1. Cohort characteristics Characteristic median (IQR) missing (%) Age, years 69.9 (62.0 - 76.0) 0.0 SBP, mmHg 153 (138 - 170) 16.6 DBP, mmHg 80 (74 - 90) 16.6 BMI, kg/m2 25.9 (24.0 - 28.4) 6.9 GFR, ml/min/1.73m2 72.2 (58.9 - 85.1) 3.5 hsCRP, mg/l 1.97 (0.92 - 4.41) 38.9 HDL-C, mmol/l 1.11 (0.90 - 1.37) 36.2 LDL-C, mmol/l 2.70 (2.02 - 3.39) 39.7 TC, mmol/l 4.63 (3.83 - 5.46) 33.3 TG, mmol/l 1.39 (1.00 - 1.96) 36.5 Smooth-muscle cells, no./field 1.34 (0.50-2.78) 2.9 Macrophages, no./field 0.34 (0.08-1.047) 4.2 Microvessels, no./field 7.0 (4.0-11.3) 10.1
percentage missing (%) Gender, male 67.8 0.0 Smoking 24.3 2.3 Statin 76.3 0.1 Antiplatelets 88.7 0.3 Diabetes 23.0 0.0 Hypertension 85.6 0.1 CAD 29.9 0.1 MI 19.3 1.1 Stroke 33.1 0.0 CVA 82.3 0.0 PAOD 20.7 0.1 Calcification, high 50.2 1.9 Collagen, high 80.3 1.8 Atheroma, >10% 72.5 1.7 Plaque hemorrhage, present 60.4 1.8 Clinical characteristics of all patients at the time of study inclusion. Patient history of CAD, MI, Stroke, CVA and PAOD were scored as percentage of cases prior to inclusion. hsCRP, smooth- muscle cells, macrophages and microvessels are presented as non-transformed data. IQR, in- terquartile range; SBP, systolic blood pressure; DBP, diastolic blood pressure; BMI, body-mass index; eGFR, estimated glomerular filtration rate (MDRD formula); hsCRP, high-sensitive C-reactive protein; CAD, coronary artery disease; MI, myocardial infarction; CVA, cerebrovascular accident; PAOD, peripheral artery occlusive disease.
20162300 Marten Siemelink.indd 90 15-03-16 13:23 BLOOD LIPID VARIANTS DO NOT AFFECT PLAQUE COMPOSITION | 91
5
Figure 1. Relationships between plasma lipids and corresponding genetic burden scores. Combined histograms-boxplots showing the model relationship between genetic burden scores and plasma lipids using linear regression modeling. Histograms show distribution (left axis) of the genetic burden score (x axis) in the cohort population. Boxplots show corrected plasma lipid concentrations (right axis), i.e. model fitted values, within the respective histogram bin. Significance of the (contin- ues) association between genetic burden score and the respective plasma lipid denoted as p-value. A) HDL-C ; B) LDL-C; C) Total TC; D) TG.
Associations with plaque characteristics We proceeded to investigate if measured plasma lipid levels were associated with histological features of atherosclerotic plaque composition. Plaque features were determined quantitatively (number of smooth-muscle cells, macrophages, and microvessels) or semi-quantitatively (for crude levels of collagen, calcification, plaque hemorrhage, and atheroma), and associated with lipid levels. We observed nominally significant associations for HDL-C with fewer plaque macrophages (p = 1.6x10-3) as well as for LDL-C (p = 2.0x10‑2), TC (p = 8.2x10-3) and TG (p = 1.4x10-2) with less plaque calcification (table 2). Similarly, we investigated if GBS of lipids were associated with histological features of atherosclerotic
20162300 Marten Siemelink.indd 91 15-03-16 13:23 92 | CHAPTER 5 0.589 0.159 0.183 0.294 P 0.055 0.334 0.728 P 0.298 0.383 0.942 0.014 P 0.052 0.726 0.678 P 1.15 (0.75 to 1.78) 1.49 (0.94 to 2.39) 1.52 (0.91 to 2.55) 0.77 (0.51 to 1.16) OR (95% CI) 0.53 (0.27) 0.15 (0.16) 0.06 (0.16) beta (SE) 1.1 (0.95 to 1.28) TG-score 1.09 (0.93 to 1.3) 0.99 (0.83 to 1.19) 0.81 (0.7 to 0.93) OR (95% CI) 0.18 (0.09) -0.02 (0.06) 0.02 (0.06) beta (SE) TG 0.348 0.020 0.228 0.254 P 0.939 0.530 0.728 P 0.687 0.135 0.470 0.008 P 0.666 0.371 0.966 P 1.21 (0.86 to 1.7) 1.69 (1.17 to 2.45) 1.35 (0.9 to 2.02) 0.8 (0.57 to 1.11) OR (95% CI) 0.02 (0.22) 0.08 (0.13) 0.04 (0.13) beta (SE) TC-score 1.03 (0.93 to 1.13) 1.11 (0.99 to 1.24) 1.05 (0.94 to 1.19) 0.85 (0.77 to 0.94) OR (95% CI) -0.03 (0.06) -0.03 (0.04) 0. (0.04) beta (SE) TC 0.421 0.026 0.125 0.365 P 0.919 0.174 0.536 P 0.996 0.127 0.144 0.020 P 0.278 0.598 0.206 P 1.18 (0.84 to 1.67) 1.67 (1.15 to 2.44) 1.47 (0.97 to 2.23) 0.83 (0.6 to 1.16) OR (95% CI) 0.02 (0.22) 0.18 (0.13) 0.08 (0.13) LDL-score beta (SE) 1. (0.88 to 1.14) 1.14 (0.99 to 1.32) 1.15 (0.98 to 1.34) 0.84 (0.74 to 0.95) OR (95% CI) -0.09 (0.08) 0.03 (0.05) 0.06 (0.05) beta (SE) LDL-C 0.446 0.799 0.503 0.324 P 0.496 0.667 0.366 P 0.553 0.171 0.351 0.678 P 0.843 0.002 0.097 P 0.85 (0.6 to 1.21) 0.94 (0.64 to 1.38) 0.84 (0.55 to 1.28) 0.82 (0.58 to 1.15) OR (95% CI) -0.15 (0.22) -0.06 (0.13) -0.12 (0.13) HDL-score beta (SE) 0.89 (0.64 to 1.24) 0.74 (0.52 to 1.06) 1.26 (0.84 to 1.89) 1.08 (0.79 to 1.49) OR (95% CI) 0.04 (0.21) -0.39 (0.12) -0.21 (0.13) beta (SE) HDL-C The association of plasma lipids and genetic burden scores with histological features plaque composition. Smooth-muscle cells, macrophages microvessels were scored quantitatively as the number of features per microscopy field. Quantitative histology normalized using Box-Cox were hemorrhage and plaque atheroma collagen, Calcification, modeling. linear regression using GBS and lipids to plasma associated and transformation Plasma lipids and burden scores were associated to increased abundance of the semi-quantitative feature using logistic determined semi-quantitatively. or as odds ratios with 95% confidence interval. regression modeling. Results are given as model beta with standard error, Plaque hemorrhage Atheroma Collagen Calcification Microvessels Macrophages Smooth-muscle cells Genetic Burden Scores Plaque hemorrhage Atheroma Collagen Calcification Microvessels Macrophages Smooth-muscle cells Table 2. Associations of Plasma lipids and their genetic burden scores with plaque composition. Table Lipid Spectra
20162300 Marten Siemelink.indd 92 15-03-16 13:23 BLOOD LIPID VARIANTS DO NOT AFFECT PLAQUE COMPOSITION | 93
plaque composition. Nominally significant associations were observed for both LDL-scores (p = 2.6x10-2) and TC-scores (p = 2.0x10-2) with atheroma size (table 2). After Bonferroni correction however, none of the plasma lipids and none of the GBS were significantly associated with histological plaque features. Since the use of statins was high in this cohort (76.3%) and strongly associated with plasma LDL-C (p=8.9 x 10-20) and plasma TC concentrations (p=2.8 x 10-15) (Supplemental Table 1), we also performed a subgroup analysis in the 344 patients not using statins (Supplemental Table 4). This showed generally similar results, with the exception of stronger negative associations with plaque calcification for the LDL-burden scores (p = 6.5x10-3) and TC-burden scores (p = 7.6x10-3), although these were not significant after Bonferroni correction. Additionally, we investigated all SNPs individually for their association with plaque characteristics, but none reached statistical significance after Bonferroni correction (Supplemental Table 3).
Associations with clinical endpoints Despite the lack of significant associations with plaque features, we investigated whether plasma lipids and their corresponding GBS were associated with the incidence of cardiovascular endpoints within 3-years of follow-up. Increased HDL-C levels were significantly associated with a reduction of composite cardiovascular endpoints, also after 5 correction for multiple testing, whereas none of the GBS were significantly associated (table 3). In addition, we found no significant associations between individual SNPs and clinical endpoints, after correction for multiple testing (Supplemental Table 3).
Discussion
To our knowledge, this study is the first effort to associate plasma lipids and genetic factors influencing them, with histological features of carotid plaques, which are widely considered to be surrogates of cardiovascular disease severity and risk.10,18,19 We show that neither plasma lipids nor genetic burden scores of plasma lipids are associated with carotid atherosclerotic plaque composition. This may seem somewhat surprising in light of the current paradigm of the “vulnerable plaque”. Given the overwhelming evidence of LDL-C as a causal contributor to CVD risk, it is perhaps remarkable that no effects on plaque composition were found. Several points are worth highlighting and may explain the lack of association of plasma lipids or the GBS with plaque features. (i) We performed this study in a patient cohort of limited size, which lacked well-matched healthy subjects as controls as it is unfeasible to obtain carotid plaque samples from healthy subjects. This may have limited biological variance in the observed plaque features. Both cohort size and lack of controls may limit the power to confidently prove a lack of associations, particularly if the magnitude of effect is small. (ii) The patients in this cohort had severe atherosclerotic disease and many of them suffered from comorbidities with effects we may not have been able to fully account for in the analysis. (iii) The majority of patients received lipid altering medications, which may have variable effects on individual lipid level improvement. This is something we cannot correct for in the model, and may explain the relatively weak association of the LDL-burden score with the measured LDL plasma concentrations in these patients, especially when compared to the other lipid fractions. The high use of lipid-lowering
20162300 Marten Siemelink.indd 93 15-03-16 13:23 94 | CHAPTER 5 -2 0.72 0.18 0.76 0.62 0.46 0.94 0.46 0.95 0.25 7.9 x10 p val p val 0.98 (0.88 to 1.09) 1.18 (0.92 to 1.52) 1.04 (0.8 to 1.36) 1.05 (0.86 to 1.29) 1.12 (0.82 to 1.54) 0.99 (0.85 to 1.16) 0.92 (0.76 to 1.1) 1.01 (0.79 to 1.29) 1.18 (0.89 to 1.58) 1.12 (0.99 to 1.27) TG HR (95% CI) TG-score HR (95% CI) -2 -2 -2 -2 0.74 0.56 0.35 5.7 x10 0.53 2.9x10 0.55 5.0x10 0.66 3.6 x10 p val p val 0.98 (0.88 to 1.09) 0.93 (0.71 to 1.2) 0.88 (0.67 to 1.15) 0.71 (0.49 to 1.01) 1.07 (0.87 to 1.31) 0.74 (0.57 to 0.97) 1.11 (0.78 to 1.58) 1.25 (1.04 to 1.52) 1.04 (0.89 to 1.22) 0.86 (0.75 to 0.99) TC HR (95% CI) TC-score HR (95% CI) -2 0.61 0.80 0.52 0.28 0.19 0.17 0.68 0.28 6.0x10 0.63 p val p val 1.03 (0.92 to 1.15) 0.97 (0.75 to 1.26) 0.91 (0.69 to 1.2) 0.81 (0.54 to 1.19) 1.15 (0.94 to 1.41) 0.82 (0.62 to 1.09) 1.08 (0.74 to 1.57) LDL-C 1.15 (0.93 to 1.41) 0.86 (0.74 to 1.01) LDL-score 1.05 (0.89 to 1.23) HR (95% CI) HR (95% CI) -2 -4 0.27 0.46 0.23 0.15 0.37 0.31 0.14 1.1x10 7.0x10 0.81 p val p val 0.94 (0.84 to 1.05) 0.91 (0.71 to 1.17) 0.85 (0.65 to 1.11) 0.77 (0.53 to 1.1) 0.91 (0.74 to 1.12) 0.87 (0.67 to 1.14) 0.75 (0.52 to 1.1) 0.74 (0.63 to 0.86) HDL-C 1.23 (1.02 to 1.48) HDL-score 1.02 (0.88 to 1.19) HR (95% CI) HR (95% CI) Cox proportional hazard model associating plasma lipids and genetic burden scores with clinical endpoints within 3 years. Results are given as ratio per standard deviation increase with 95% confidence intervals. †Restenosis was assessed after 1 year by carotid echography as more than 50% stenosis of the vascular lumen, and was associated using logistic regression modeling, showing odds ratio (95% CI) instead hazard ratio. Composite endpoints Cardiovascular death Myocardial Infarction Cardiovascular death Stroke Stroke Myocardial Infarction Composite endpoints Table 3. Associations of plasma lipids and their genetic burden scores with clinical endpoints Table Restenosis† Restenosis†
20162300 Marten Siemelink.indd 94 15-03-16 13:23 BLOOD LIPID VARIANTS DO NOT AFFECT PLAQUE COMPOSITION | 95
medications may also have improved plaque composition, and may have prevented cardiovascular events in some patients. This could mask the effects of lipids or the GBS, although sub-analysis in non-statin users showed generally similar results. (iv) We have shown that carotid plaque histological examination is prone to some inter-observer bias.20 This is due to a certain level of subjective interpretation by the observer, affecting the precision to determine plaque composition. (v) With regard to plasma lipids, we were only able to include a pre-operative measurement, while plasma lipid concentrations may vary considerably over time. (vi) Also, we suffered from the incomplete availability of plasma lipid concentrations (missingness 33.3% - 39.7%), which may contribute to bias. (vii) For most SNPs previously associated with lipid levels, the mechanism of action is unknown. Conceivably, unknown factors may modify their effect on plasma lipids between individuals. This would also explain the different effect sizes when comparing our study to the GLGC study results. An alternative explanation is that plasma lipid levels may have a negligible effect on carotid plaque composition, and may exert their effects on the risk of CVD through other mechanisms (for example, coronary calcification21). In accordance, a recent meta-analysis showed that statin treatment leads to carotid plaque regression through improvement of inflammation, not lipid levels.22 5 We also investigated if plasma lipids or genetic variants were associated with clinical endpoints within 3 years, but failed to show an association. We also included myocardial infarction as an endpoint, as there is a partially shared etiology23 and patients with carotid artery disease are at increased risk of cardiovascular disease in other vascular territories.24 The notable exception is a negative association between HDL-C levels and the occurrence of composite cardiovascular endpoints, which may represent a non-causal effect (e.g. due to reverse causation or uncorrected confounding), consistent with recent Mendelian randomization studies.2,4,5 The aforementioned limitations as well as the limited follow-up period of 3 years in this actively monitored and treated patient group may have decreased power to show an effect. Replication at present seems challenging, as we are unaware of cohorts that possess similar data with adequate sample size. Future efforts in larger samples will have better power to address the causal role of lipid fractions and the “vulnerable plaque” in cardiovascular disease. If it is true that there is no relation between lipid fractions and plaque features, then that would genuinely question the mechanisms by which plasma lipids contribute to CVD risk, as well as the paradigm of the vulnerable plaque, yet we do not believe our current results are sufficiently robust to arrive at that conclusion. In summary, this study found no evidence that plasma lipid levels or genetic determinants of plasma lipid levels influence carotid plaque composition.
Acknowledgements The authors would like to express their gratitude to Freerk van Dijk and Morris Swertz for imputing the Athero-express datasets using the “GoNL Impute2” pipeline.
20162300 Marten Siemelink.indd 95 15-03-16 13:23 96 | CHAPTER 5
References
1 Sarwar N, Sandhu MS, Ricketts SL, et al. Triglyceride-mediated pathways and coronary disease: col- laborative analysis of 101 studies. Lancet 2010; 375: 1634–9. 2 Voight BF, Peloso GM, Orho-Melander M, et al. Plasma HDL cholesterol and risk of myocardial infarc- tion: a mendelian randomisation study. Lancet 2012; 6736: 1–9. 3 Do R, Willer CJ, Schmidt EM, et al. Common variants associated with plasma triglycerides and risk for coronary artery disease. Nat Genet 2013; 45: 1345–52. 4 Holmes M V, Asselbergs FW, Palmer TM, et al. Mendelian randomization of blood lipids for coronary heart disease. Eur Heart J 2014; published online Jan 27. DOI:10.1093/eurheartj/eht571. 5 Wu Z, Lou Y, Qiu X, et al. Association of cholesteryl ester transfer protein (CETP) gene polymorphism, high density lipoprotein cholesterol and risk of coronary artery disease: a meta-analysis using a Men- delian randomization approach. BMC Med Genet 2014; 15: 118. 6 Burgess S, Freitag DF, Khan H, Gorman DN, Thompson SG. Using multivariable mendelian randomiza- tion to disentangle the causal effects of lipid fractions. PLoS One 2014; 9: e108891. 7 Teslovich TM, Musunuru K, Smith A V, et al. Biological, clinical and population relevance of 95 loci for blood lipids. Nature 2010; 466: 707–13. 8 Willer CJ, Schmidt EM, Sengupta S, et al. Discovery and refinement of loci associated with lipid levels. Nat Genet 2013; 45: 1274–83. 9 Verhoeven BAN, Velema E, Schoneveld AH, et al. Athero-express: differential atherosclerotic plaque expression of mRNA and protein in relation to cardiovascular events and patient characteristics. Ra- tionale and design. Eur J Epidemiol 2004; 19: 1127–33. 10 Hellings WE, Moll FL, De Vries J-PPM, et al. Atherosclerotic plaque composition and occurrence of restenosis after carotid endarterectomy. JAMA 2008; 299: 547–54. 11 van der Laan SW, Asl HF, van den Borne P, et al. Variants in ALOX5, ALOX5AP and LTA4H are not asso- ciated with atherosclerotic plaque phenotypes: The Athero-Express Genomics Study. Atherosclerosis 2015; 239: 528–38. 12 Anderson CA, Pettersson FH, Clarke GM, Cardon LR, Morris AP, Zondervan KT. Data quality control in genetic case-control association studies. Nat Protoc 2010; 5: 1564–73. 13 Whole-genome sequence variation, population structure and demographic history of the Dutch popu- lation. Nat Genet 2014; 46: 818–25. 14 Wray NR, Goddard ME, Visscher PM. Prediction of individual genetic risk to disease from genome- wide association studies. Genome Res 2007; 17: 1520–8. 15 Purcell SM, Wray NR, Stone JL, et al. Common polygenic variation contributes to risk of schizophrenia and bipolar disorder. Nature 2009; 460: 748–52. 16 Ananthakrishnan AN, Huang H, Nguyen DD, Sauk J, Yajnik V, Xavier RJ. Differential effect of genetic burden on disease phenotypes in Crohn’s disease and ulcerative colitis: analysis of a North American cohort. Am J Gastroenterol 2014; 109: 395–400. 17 Purcell S, Neale B, Todd-Brown K, et al. PLINK: a tool set for whole-genome association and popula- tion-based linkage analyses. Am J Hum Genet 2007; 81: 559–75. 18 Hellings WE, Peeters W, Moll FL, et al. Composition of carotid atherosclerotic plaque is associated with cardiovascular outcome: a prognostic study. Circulation 2010; 121: 1941–50. 19 Howard DPJ, van Lammeren GW, Rothwell PM, et al. Symptomatic carotid atherosclerotic disease: correlations between plaque composition and ipsilateral stroke risk. Stroke 2015; 46: 182–9. 20 Hellings WE, Pasterkamp G, Vollebregt A, et al. Intraobserver and interobserver variability and spatial differences in histologic examination of carotid endarterectomy specimens. J Vasc Surg 2007; 46: 1147–54. 21 van Setten J, I gum I, Pechlivanis S, et al. Serum Lipid Levels, Body Mass Index, and Their Role in Coronary Artery Calcification: A Polygenic Analysis. Circ Cardiovasc Genet 2015; 8: 327–33. 22 Ibrahimi P, Jashari F, Bajraktari G, Wester P, Henein MY. Ultrasound assessment of carotid plaque echogenicity response to statin therapy: a systematic review and meta-analysis. Int J Mol Sci 2015; 16: 10734–47. 23 Dichgans M, Malik R, König IR, et al. Shared genetic susceptibility to ischemic stroke and coronary artery disease: a genome-wide analysis of common variants. Stroke 2014; 45: 24–36. 24 van Wijk I, Kappelle LJ, van Gijn J, et al. Long-term survival and vascular event risk after transient ischaemic attack or minor ischaemic stroke: a cohort study. Lancet (London, England) 2005; 365: 2098–104.
20162300 Marten Siemelink.indd 96 15-03-16 13:23 BLOOD LIPID VARIANTS DO NOT AFFECT PLAQUE COMPOSITION | 97
Supplemental Material
Supplement to: Siemelink, MA et al, Common variants associated with blood lipid levels do not affect carotid plaque composition. Supplemental material available online (http:// dx.doi.org/10.1016/j.atherosclerosis.2015.07.041)
Table of Contents Supplemental Tables Supplemental Table 1. Associations of plasma lipids with cardiovascular risk factors. Supplemental Table 2. Associations of genetic burden scores with cardiovascular risk factors. Supplemental Table 3. General information, effect sizes and associations for all individual SNPs included in the study. Supplemental Table 4. Sub-analysis of plasma lipids and their burden scores in patients without statin treatment.
Supplemental Figure Supplemental Figure 1. Histogram showing distribution of plasma LDL-C values. 5
20162300 Marten Siemelink.indd 97 15-03-16 13:23 M.A. Siemelink1 S.W. van der Laan1 S. Haitjema1 H. el Azzouzi2 F.W. Asselbergs3,4,5 H.M. den Ruijter1 B.T. Heijmans6 G. Pasterkamp1,7
1 Laboratory of Experimental Cardiology, University Medical Center Utrecht, Utrecht, The Netherlands 2 Department of Cardiology, University Medical Center Utrecht, Utrecht, the Netherlands 3 Institute of Cardiovascular Science; Faculty of Population Health Sciences; University College London; London, United Kingdom 4 Department of Cardiology; Division Heart and Lungs; University Medical Center Utrecht; Utrecht; the Netherlands 5 Durrer Center for Cardiogenetic Research; ICIN-Netherlands Heart Institute; Utrecht, the Netherlands 6 Department of Molecular Epidemiology, Leiden University Medical Center, Leiden, The Netherlands 7 Reseach Laboratory Clinical Chemistry, University Medical Center Utrecht, Utrecht, The Netherlands
20162300 Marten Siemelink.indd 98 15-03-16 13:23 CHAPTER 6
Cardiovascular Risk Loci associate with DNA Methylation in Carotid Plaques
MANUSCRIPT IN PREPARATION
20162300 Marten Siemelink.indd 99 15-03-16 13:23 100 | CHAPTER 6
Abstract
Introduction Large scale genomic meta-analyses have revealed single nucleotide polymorphisms (SNPs) that associate with the risk of atherosclerotic cardiovascular disease (CVD), including ischemic stroke and coronary artery disease. For most of these variants, the mechanism by which these genetic variants influence the biology of disease is still unknown. Genetic variants can alter DNA-methylation, which may lead to tissue-specific changes in the expression of genes. Such methylation quantitative trait loci (meQTL) may influence atherosclerotic disease progression and may serve as an attractive therapeutic target. We conducted meQTL analyses for the association of SNPs related to CVD traits with DNA- methylation in carotid artery derived atherosclerotic plaques.
Methods For 444 patients that underwent carotid endarterectomy in the Athero-Express Biobank study, both genotyping and measurement of DNA methylation were performed. Genotyping was done using Affymetrix arrays and missing genotypes were imputed using GoNL5 as a reference. DNA methylation was assessed in peroperatively obtained atherosclerotic plaques using the Illumina HM450k array. By means of online literature searches, 106 CVD-associated lead SNPs were identified in 30 GWAS publications. For these SNPs, meQTL analysis was performed with all nearby CpGs.
Results meQTL analyses of 106 CVD related SNPs for association with DNA methylation in carotid plaques showed 23 SNPs with significant associations to DNA methylation at 74 proximal CpG loci. For 9 out of the 23 SNPs, the SNP associated with both expression and CpGs near the same gene (including CELSR2, VAMP5, MRAS, ABO, KIAA1462, HHIPL1 and SNF8).
Conclusion We show that 23 known CVD related SNPs are associated with differential DNA methylation at 74 nearby CpGs in carotid atherosclerotic plaque tissue. Of these 23 SNPs, 9 associated with DNA-methylation and gene-expression at the same gene. This may indicate that these SNPs exert their effects through a change in DNA methylation, and suggests the involvement of these CpGs in CVD-risk.
20162300 Marten Siemelink.indd 100 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE METHYLATION | 101
Introduction
In order to improve therapeutic strategies for atherosclerotic cardiovascular disease (CVD), a better understanding of pathophysiology is required. While atherosclerotic cardiovascular disease occurs at several distinct predilection sites (e.g. carotid bifurcation, coronary arteries, femoral arteries), a substantial fraction of CVD-associated genetic variants associate with atherosclerotic disease across vascular territories, a testimony to the partially shared pathophysiology between atherosclerotic diseases1. Heritability estimates in twin- studies have shown that genetic factors are important contributors to CVD risk.2,3 Several large-scale genome-wide association studies (GWAS) have elucidated common single nucleotide polymorphisms (SNPs) that contribute to this risk, with more likely to be discovered. Collectively, these SNPs explain a sizeable part of the heritability,4,5 and may facilitate the search for improvement of risk prediction.6 However, there is still an unexplained heritability, which may be reduced when epigenetic heritability is included. SNPs may cause a higher susceptibility to atherosclerotic disease through various mechanisms. Aside from well appreciated functional missense mutations, SNPs may affect transcriptional expression, for example through epigenetic mechanisms. Genetic variants that alter DNA-methylation, known as methylation quantitative trait loci (meQTL),7,8 often lead to changes in the expression of genes.9,10 While cis-meQTL associations between SNPs and CpGs are generally within close proximity (within 500kb),11 there are also considerable trans-meQTL associations across chromosomes.12 Interestingly, in addition to the effects of genetic sequence variation on DNA methylation, environmental 6 factors may also affect methylation at the same locus. This suggests that genetic and environmental effects may coalesce at the DNA methylation level to regulate gene expression. Furthermore, in contrast to the genetic sequence, DNA methylation is cell- type and tissue specific. Thus, meQTL analysis in relevant tissues may identify downstream epigenetic regulators of disease risk and may yield important pathophysiological insight. We conducted methylation quantitative trait analyses to investigate the effect of cardiovascular-risk SNPs on DNA-methylation at nearby CpGs in carotid artery atherosclerotic lesions. We present multiple SNPs that strongly associate with DNA-methylation at CpG loci in-cis and with gene expression at the same loci. This may indicate that these SNPs increase cardiovascular risk by affecting local DNA-methylation in the lesion.
Methods
Patient inclusion The Athero-Express Biobank Study (AE, www.atheroexpress.nl) is an ongoing longitudinal Biobank study including patients that undergo arterial endarterectomy in two Dutch referral centers since 2002. A detailed description of the cohort study design has been previously published.13 For the present study, 444 patients were included who underwent carotid endarterectomy (CEA) on the basis of genotyping data and biomaterial availability. Clinical data were extracted from patient medical files and standardized questionnaires. The medical ethics committees of the respective hospitals approved the study and informed consent was obtained from all patients.
20162300 Marten Siemelink.indd 101 15-03-16 13:23 102 | CHAPTER 6
Sample collection Carotid plaque specimens were removed during surgery and processed in the laboratory. Specimens were cut transversely into segments of 5 mm. The culprit lesion was identified, fixed in 4% formaldehyde, embedded in paraffin and processed for histological examination. Remaining segments were snap-frozen and stored at -80°C. Peripheral blood samples were obtained immediately prior to surgery and stored at -80°C.
DNA methylation array DNA was isolated from 506 plaque segments and 94 peripheral blood samples (matched samples for 91 patients) and used for DNA methylation measurements. DNA purity and concentration was determined using the Nanodrop 1000 system (Thermo Scientific, Massachusetts, USA) and equalized at 600ng. DNA samples were randomized over the 96-well plates and bisulphate converted using a cycling protocol and the EZ-96 DNA methylation kit (Zymo Research, Orange County, USA). DNA methylation was measured on the Infinium HumanMethylation450 Beadchip Array (HM450k, Illumina, San Diego, USA) at the Erasmus Medical Center Human Genotyping Facility (Rotterdam, The Netherlands). Processing of the sample and array was performed according to the manufacturer’s protocol.
Quality control of methylation data The quality of the HM450k array data was determined with the MethylAid R-package14 using default settings, controlling for sample dependent and probe dependent parameters. Bisulphate conversion efficiency was determined using dedicated probes on the HM450k. A total of 41,640 probes were excluded, with 443,872 probes (91·4 %) of good quality remaining. We used ‘Functional Normalization’15 (implemented in the minfi R-package)16 with eight control-probe principal components for normalization and correction of batch effects. Principal components of CpG-probes were calculated and used to identify possible mix up of samples by tissue type (i.e. blood or plaque DNA) or sex. Genotype data was correlated to the raw data of the 65 SNPs included on the HM450k array, and samples with poor correlation (R ≤ 0·6) across these 65 SNPs were excluded due to possible mix up.
Epigenome-wide methylation analysis Due to the high number of meQTL associations performed, we chose to use the conservative Bonferroni method to correct for multiple testing. For all other associations, the less conservative false-discovery rate (FDR) was used.17 In both instances, a corrected p value ≤ 0·05 was considered statistically significant. Statistical analyses were performed in R Studio (v0.99.473) using R (v3.2.2). Unless otherwise specified, CpG annotation is derived from the Illumina HM450K annotation file.
Systematic search for cardiovascular disease related SNPs In order to get a comprehensive list of CVD related SNPs discovered by GWAS studies, we performed five queries in the GWAS Catalogue database with search terms “Cardiovascular”, “Carotid”, “Coronary”, “Stroke” and “Peripheral” using standard settings (p-value cut-off = 5x10-8).18 All results were filtered on ‘Reported Trait’, including traits for CVD surrogates and endpoints while excluding studies on subpopulations, risk
20162300 Marten Siemelink.indd 102 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE METHYLATION | 103
factors and interactions. Furthermore we searched pubmed for GWAS on these traits and added recent reports by the CARDIoGRAMplusC4D19 and METASTROKE consortia,20,21 as well as a study on peripheral artery disease,22 which were not yet included in the GWAS catalogue. Sex chromosomal SNPs were excluded from analysis. All resulting autosomal SNPs were queried in SNiPA to retrieve GRCh37 genome positions.23 SNPs that could not be identified in SNiPA were queried in dbSNP to retrieve GRCh37 positions.24 For all resulting SNPs, pairwise LD scores were retrieved from SNAP genome to determine LD structures.25 For SNPs in LD with r2> 0.8, the lead SNP was determined as the SNP in the latest GWAS publication, resulting in 106 lead SNPs, with the other SNPs deemed proxies. The final workflow for the identified SNPs is summarized in Supplemental Table 3.
Genotyping & Imputation DNA was isolated from stored samples and genotyping was performed in two series using commercially available genotyping arrays.26 The first series (Athero-Express Genomics Study 1, AEGS1) was genotyped using Affymetrix Genome-Wide Human SNP Array 5.0, the second (Athero-Express Genomics Study 2, AEGS2) was genotyped using the Affymetrix Axiom® GW CEU 1 Array. Community standard quality control and assurance procedures we used to clean the genotype data obtained in AEGS1 and AEGS2.27 Phased haplotypes from the Genome of the Netherlands (version 5) served as the reference panel for genotype imputation. 6 Methylation Quantitative Trait Analysis Genotyping data was available for 102 of the 106 lead SNPs. Additionally, the CVD related SNPS were filtered on quality measures of our data (minor allele frequency ≥ 0·05; imputation quality ≥ 0·8; Hardy-Weinberg equilibrium p value ≥ 1·0x10-6), resulting in meQTL data for 99 SNPs. A single SNP in the plaque data did not return any results due to missing CpG data. MeQTL analysis was performed in cis for 21,302 CpGs near these SNPs, i.e. within 500kb of the SNP. We used SNPTEST28 (v2.5.0) for the identification of methylation quantitative trait loci in a additive model corrected for covariates age, sex, SNP array type, genotyping principal components 1 and 2, methylation principal components 1 and 2. For the resulting meQTL associations, inflation was calculated for plaque and blood samples separately, and corrected using genomic control (Supplemental Figure 6).
Results
To construct an inclusive list of CVD-risk SNPs across atherosclerotic disease traits, online repositories were surveyed. Online queries resulted in 30 GWAS studies on phenotypic traits of atherosclerotic disease (Supplemental Table 1) which reported 135 SNPs (106 lead SNPs, 29 proxies) on autosomal chromosomes (Supplemental Table 2). Of these SNPs, 91 were associated with coronary artery disease, 13 with myocardial infarction and 1 with coronary artery calcification. Furthermore, 19 SNPs were associated with stroke (including ischemic- and large artery stroke) and 5 with cIMT. Finally, a single SNP was associated with peripheral artery disease. The workflow for these SNPs is summarized in Supplemental Table 3.
20162300 Marten Siemelink.indd 103 15-03-16 13:23 104 | CHAPTER 6 -17 -09 -06 -12 -07 -07 -02 -11 -02 -07 -02 -02 -06 -04 -03 -06 -03 -03 -02 -04 -13 -04 -11 Corr. P Corr. 7.4X10 1.1X10 2.1X10 2.6X10 1.2X10 1.8X10 2.5X10 5.5X10 1.1X10 5.0X10 2.7X10 1.2X10 5.8X10 4.7X10 2.4X10 2.0X10 4.8X10 2.3X10 2.5X10 1.5X10 1.1X10 3.3X10 3.5X10 07 -21 -14 -11 -17 -12 -12 -15 -07 -11 -06 -07 -10 -08 -08 -11 -07 -08 -07 -09 -18 -08 -15 P 2.8X10 4.2X10 7.8X10 9.7X10 4.6X10 6.8X10 9.3X10- 2.1X10 4.3X10 1.9X10 1.0X10 4.5X10 2.2X10 1.8X10 9.0X10 7.7X10 1.8X10 8.8X10 9.2X10 5.8X10 4.1X10 1.3X10 1.3X10 0.07 0.06 0.08 0.08 0.06 0.06 0.07 0.05 0.06 0.04 0.06 0.06 0.06 0.07 0.07 0.05 0.05 0.03 0.06 0.06 0.06 0.07 0.05 S.E. 0.76 -0.53 0.59 -0.77 -0.48 0.44 -0.39 0.46 0.35 -0.28 0.33 0.34 0.46 -0.41 -0.43 0.34 0.30 -0.20 0.33 0.39 -0.57 -0.43 0.41 Beta 109817496 85810744 203926405 138071101 31325085 32375672 106297689 136151359 136151359 30316432 104813866 89920650 111807467 111807467 112123256 110961404 111039873 111040579 100138306 100138306 79140260 2031491 47019901 CpG Pos cg00908766 cg02493740 cg08076091 cg20289491 cg26706521 cg18067840 cg27284331 cg24267699 cg24267699 cg12786570 cg03493300 cg00757033 cg10833066 cg10833066 cg08577424 cg08318510 cg26053697 cg03482866 cg04705318 cg04705318 cg27435867 cg16513277 cg22482690 CpG Dist 1034 755 180520 51021 140889 34319 111763 9156 2809 18690 94770 88309 77141 264957 109762 692 808 9044 4364 7404 1524 95013 31304 SNP Pos 109818530 85809989 203745885 138122122 31184196 32341353 106409452 136142203 136154168 30335122 104719096 90008959 111884608 112072424 112233018 110960712 111040681 111049623 100133942 100145710 79141784 2126504 46988597 Chr 1 2 2 3 6 6 7 9 9 10 10 12 12 12 12 13 13 13 14 14 15 17 17 CELSR2, PSRC1, SORT1 GGCX VAMP8, VAMP5, WDR12 MRAS HCG27, HLA-C BTNL2, C6orf10 PIK3CG ABO ABO KIAA1462 CYP17A1, CNNM2, NT5C2 ATP2B1 SH2B3 SH2B3 ALDH2 COL4A1/A2 COL4A1/A2 COL4A1/A2 HHIPL1 HHIPL1 ADAMTS7 SMG6, SRR UBE2Z,ATP5G1 SNF8, GIP, Nearest genes Table 1. meQTL analysis of CVD SNPs in carotid artery atherosclerotic plaque Table SNP rs646776 rs1561198 rs6725887 rs9818870 rs3869109 rs9268402 rs17398575 rs514659 rs579459 rs2505083 rs12413409 rs2681472 rs3184504 rs11065987 rs10744777 rs4773144 rs11838776 rs9515203 rs2895811 rs10139550 rs7173743 rs216172 rs46522 chromosome; Corrected p values are calculated using the Bonferroni method based on total number of associations performed. Abbreviations: Chr, bonferroni corrected p value. P, Pos, position; Dist, distance in basepairs from CpG to SNP; Corr.
20162300 Marten Siemelink.indd 104 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE METHYLATION | 105
Methylation quantitative trait analysis (meQTL) was performed for all SNPs to assess -17 -09 -06 -12 -07 -07 -02 -11 -02 -07 -02 -02 -06 -04 -03 -06 -03 -03 -02 -04 -13 -04 -11 their association with DNA methylation in carotid atherosclerotic plaque samples. This showed that 23 lead SNPs were associated with a change in carotid plaque DNA Corr. P Corr. 7.4X10 1.1X10 2.1X10 2.6X10 1.2X10 1.8X10 2.5X10 5.5X10 1.1X10 5.0X10 2.7X10 1.2X10 5.8X10 4.7X10 2.4X10 2.0X10 4.8X10 2.3X10 2.5X10 1.5X10 1.1X10 3.3X10 3.5X10 methylation at 74 unique CpG loci in their vicinity. For each SNP, the best associated CpG 07 -21 -14 -11 -17 -12 -12 -15 -07 -11 -06 -07 -10 -08 -08 -11 -07 -08 -07 -09 -18 -08 -15 is presented in table 1. For all SNPs with significant meQTL associations, regional P associations with CpGs are depicted in Supplemental Figure 1.
2.8X10 4.2X10 7.8X10 9.7X10 4.6X10 6.8X10 9.3X10- 2.1X10 4.3X10 1.9X10 1.0X10 4.5X10 2.2X10 1.8X10 9.0X10 7.7X10 1.8X10 8.8X10 9.2X10 5.8X10 4.1X10 1.3X10 1.3X10 SNPs associating to CpGs at- and expression of the same gene, may indicate that the SNP affects gene expression through DNA methylation. For the 23 SNP-to-CpG associations, the relationship of SNPs to CpGs and SNPs to gene expressions (GTEX29 via Haploreg30) 0.07 0.06 0.08 0.08 0.06 0.06 0.07 0.05 0.06 0.04 0.06 0.06 0.06 0.07 0.07 0.05 0.05 0.03 0.06 0.06 0.06 0.07 0.05 S.E. are presented in table 2. This shows that for 9 out of 23, the SNP, CpG and gene expression are corresponding to the same gene (including CELSR2, VAMP5, MRAS, ABO, KIAA1462, HHIPL1 and SNF8) while 2 SNPs associate with DNA-methylation and expression at other 0.76 -0.53 0.59 -0.77 -0.48 0.44 -0.39 0.46 0.35 -0.28 0.33 0.34 0.46 -0.41 -0.43 0.34 0.30 -0.20 0.33 0.39 -0.57 -0.43 0.41 Beta genes (WDR12 and WDR51B). 6 SNPs are associated with DNA-methylation at previously unreported genes or are intergenic (HLA-B, FAM109A, ACAD10 and BRAP), while these SNPs are also implicated in expression at other genes. Finally, 6 SNPs associate with CpGs at the same gene, but do not associate with expression of that gene (BTNL2, CNNM2, 109817496 85810744 203926405 138071101 31325085 32375672 106297689 136151359 136151359 30316432 104813866 89920650 111807467 111807467 112123256 110961404 111039873 111040579 100138306 100138306 79140260 2031491 47019901 CpG Pos COL4A2 and SMG6). The effect of DNA methylation on expression is complex and highly dependent on the location of the CpG in relation to a gene.31–34 To investigate if the genomic region of the identified CpG determines its function, we compared the regions in which the 74 CpGs were situated to the distribution on the microarray (Supplemental Figure 3 and 4), showing cg00908766 cg02493740 cg08076091 cg20289491 cg26706521 cg18067840 cg27284331 cg24267699 cg24267699 cg12786570 cg03493300 cg00757033 cg10833066 cg10833066 cg08577424 cg08318510 cg26053697 cg03482866 cg04705318 cg04705318 cg27435867 cg16513277 cg22482690 CpG enrichment of meQTL CpGs within genes and within 1500 bp of the transcription start 6 site. Since DNA-methylation is tissue specific, we investigated differences in strengths of meQTL effects for the 23 SNPs in carotid plaque samples and in peripheral blood samples, Dist 1034 755 180520 51021 140889 34319 111763 9156 2809 18690 94770 88309 77141 264957 109762 692 808 9044 4364 7404 1524 95013 31304 which are depicted in Figure 1 and shown in detail in Supplemental Table 4. This shows that most meQTL associations are comparable between tissues in direction and strength of effect. However, it also shows several meQTL effects in carotid plaque samples which are undetected in blood samples. SNP Pos 109818530 85809989 203745885 138122122 31184196 32341353 106409452 136142203 136154168 30335122 104719096 90008959 111884608 112072424 112233018 110960712 111040681 111049623 100133942 100145710 79141784 2126504 46988597 Discussion Chr 1 2 2 3 6 6 7 9 9 10 10 12 12 12 12 13 13 13 14 14 15 17 17
While previous epigenome-wide association studies have identified CVD-related CpGs in blood35,36 and vascular tissue,37,38 we used a different approach which offers a unique perspective on the possible epigenetic effects of known cardiovascular risk SNPs. Leveraging the unique strength of the Athero-Express Biobank, we identify which of these SNPs associate with DNA methylation in the atherosclerotic lesion. Furthermore, we find strong support that meQTL SNPs may affect expression of genes by affecting DNA- methylation. Furthermore, our modelling showed generally comparable yet also interesting CELSR2, PSRC1, SORT1 GGCX VAMP8, VAMP5, WDR12 MRAS HCG27, HLA-C BTNL2, C6orf10 PIK3CG ABO ABO KIAA1462 CYP17A1, CNNM2, NT5C2 ATP2B1 SH2B3 SH2B3 ALDH2 COL4A1/A2 COL4A1/A2 COL4A1/A2 HHIPL1 HHIPL1 ADAMTS7 SMG6, SRR UBE2Z,ATP5G1 SNF8, GIP, Nearest genes differences in meQTL effects between atherosclerotic plaque samples and blood samples from the same patients. This may indicate SNPs that have meQTL effects depending on tissue type. It is contrary to findings by Smith et al. whom showed meQTL effects across ancestry, developmental stage and tissues and concluded that peripheral blood may be a reliable correlate to other tissues.39 The main strength of this study is that meQTL effects are uncovered which occur in the
Table 1. meQTL analysis of CVD SNPs in carotid artery atherosclerotic plaque Table SNP rs646776 rs1561198 rs6725887 rs9818870 rs3869109 rs9268402 rs17398575 rs514659 rs579459 rs2505083 rs12413409 rs2681472 rs3184504 rs11065987 rs10744777 rs4773144 rs11838776 rs9515203 rs2895811 rs10139550 rs7173743 rs216172 rs46522 chromosome; Corrected p values are calculated using the Bonferroni method based on total number of associations performed. Abbreviations: Chr, bonferroni corrected p value. P, Pos, position; Dist, distance in basepairs from CpG to SNP; Corr. disease lesion itself and are associated with eQTL effects on gene expression. This may
20162300 Marten Siemelink.indd 105 15-03-16 13:23 106 | CHAPTER 6
have three important implications towards elucidation of cardiovascular pathophysiology and subsequent drug development. First, it identifies which SNPs may exert their effects through epigenetic mechanisms, directing future research efforts and possibly offering opportunities for intervention at the epigenetic level. Second, we show several SNPs that affect DNA-methylation at other genes, which were previously not implicated in cardiovascular disease, identifying them as potential contributors to CVD. Third, it may give an indication which of the SNPs exert their risk inducing effects within the lesion, as opposed to SNPs which influence processes in other tissues (e.g. CVD-risk SNPs affecting metabolism in liver and intestines or affecting coagulation in the circulation). Despite best efforts, this study is subject to several limitations. Even though for the majority of meQTL associations the results in plaque and blood are similar, providing strong support for our study, validation of our results in an independent cohort is warranted. Furthermore, the results from this analysis may actually be an underrepresentation of meQTL effects of the selected SNPs due to several reasons. First, there may be a lack of biological variation
Figure 1. For all CVD-related SNPs, the strength of effect (model beta) is depicted for the meQTL association with DNA-methylation at the strongest CpG-locus. Each point represents a single SNP. Significance of the meQTL association in each tissue is shown by colour (NB. Sample sizes of the tissues are unequal). Abbreviations: NS, non-significant.
20162300 Marten Siemelink.indd 106 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE METHYLATION | 107
of a SNP or CpG in this moderately sized disease cohort, or in these tissues, reducing power. Second, the methylation array does not provide full coverage of genomic DNA methylation loci, possibly missing associations with non-genotyped CpGs. Lastly, we focussed on in-cis meQTL associations within a window of 500kb left and right of the SNP. In general, the strength of meQTL-associations between SNPs and CpGs showed an inverse relationship with the distance between them (Supplemental Figure 2), indicating that many meQTL effects may occur at relatively close distance, as was previously reported.11 However, trans-meQTL effects have also been demonstrated which we did not investigate.40 Irrespective of its limitations, this study puts the identified CpGs in the limelight for further scrutiny, providing important evidence for close interplay between CVD-related genetic variation, DNA methylation and gene expression. Following up on these results, it would be informative to investigate these meQTL-associations and gene-expression in other CVD-related tissues like liver and intestine. Additionally, in-trans meQTL effects of SNPs with CpGs and eQTL effects of CpGs with expression may yield further insights. While the effects of SNPs are generally small, conceivably the effect of methylation on gene expression and thereby on disease risk may be substantial. This may present epigenetic loci as attractive drug targets, which however, first requires proof of causality to CVD risk. This may be investigated by a Mendelian randomization study of these SNPs and their corresponding CpGs to CVD outcome in a larger cohort, as was recently published for blood pressure.40 We showed that 23 known CVD related SNPs are associated with differential DNA 6 methylation at 74 nearby CpGs in carotid atherosclerotic plaque tissue. Of these 23 SNPs, 11 associate with DNA-methylation and gene-expression at the same gene. This may indicate that these SNPs exert their effects through a change in DNA methylation, and suggests the involvement of these CpGs in CVD-risk.
20162300 Marten Siemelink.indd 107 15-03-16 13:23 108 | CHAPTER 6 Yes yes Yes Enh Ocean Island Island North Shelf Ocean Island South Shelf South Shore Ocean South Shore North Shore South Shore Ocean Island Ocean Body TSS1500 TSS1500 Intergenic TSS1500 TSS200 5’UTR TSS1500 Body TSS1500 TSS1500 TSS1500 3’UTR Region Body KIAA1462 ABO ABO Intergenic BTNL2 HLA-B MRAS FAM109A NBEAL1 FAM109A VAMP5 WDR51B CELSR2 Gene CNNM2 cg12786570 cg24267699 cg24267699 cg27284331 cg18067840 cg26706521 cg20289491 cg10833066 cg08076091 cg10833066 cg02493740 cg00757033 cg00908766 CpG ID cg03493300 AA WB;LYM WB;15 other AA AA;CA;TA;TLY;WB;28 AA;CA;TA;TLY;WB;28 other MON AA;TA;WB;3 other AA;TA;WB;3 WB;1 other AA;CA;TA;LYM;12 other AA;CA;TA;LYM;12 WB;1 other AA;TA;LYM;TLY;WB;20 AA;TA;LYM;TLY;WB;20 other TA;WB;1 other TA;WB;1 WB;7 other eQTL tissues LYM;WB;13 other LYM;WB;13 KIAA1462 ABO;GBGT1;SURF4;SURF6 ABO;GBGT1;SLC2A6 PRKAR2B HLA-DQA2;HLA-DRB6;HLA-DQB1;HLA- DRB1;HLA-DRA;HLA-DQB1-AS1;AGPAT1 LOC642073 MRAS;A4GNT;FAIM ALDH2;SH2B3 CARF;ICA1L;NBEAL1;FAM117B ALDH2;ATXN2;SH2B3 GGCX;TMEM150A;VAMP8;VAMP5; GNLY;USP39 ATP2B1;WDR51B PSRC1;CELSR2;SORT1;PSMA5 eQTL genes MARCKSL1P1;ARL3;MIR1307;CYP17A1-AS1; WBP1L;SFXN2;C10orf32;NT5C2 INT INT INT 3’UTR INT NSM INT Region CYP17A1, CNNM2, NT5C2 KIAA1462 ABO ABO PIK3CG BTNL2, C6orf10 HCG27, HLA-C MRAS SH2B3 WDR12 SH2B3 VAMP5, VAMP8, GGCX VAMP8, VAMP5, ATP2B1 ATP2B1 CELSR2, PSRC1, SORT1 Nearest genes Table 2. Annotation of SNPs and associated CpG loci Table rs12413409 rs2505083 rs579459 rs514659 rs17398575 rs9268402 rs3869109 rs9818870 rs11065987 rs6725887 rs3184504 rs1561198 rs2681472 rs646776 SNP ID
20162300 Marten Siemelink.indd 108 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE METHYLATION | 109 Yes Yes Yes North Shelf Ocean Ocean North Shelf North Shelf South Shelf South Shore South Shore Island Body Body Intergenic Body Body Body Body Body TSS1500;Body SNF8 SMG6 Intergenic HHIPL1 HHIPL1 COL4A2 COL4A2 COL4A2 ACAD10;BRAP cg22482690 cg16513277 cg27435867 cg04705318 cg04705318 cg03482866 cg26053697 cg08318510 cg08577424
and annotations for associated CpGs (provided by Illumina). SNPs were shown to associ - 6 30 WB;LYM;10 other WB;LYM;10 AA;TA;LYM;WB;11 other AA;TA;LYM;WB;11 TA;LYM;WB;1 other TA;LYM;WB;1 2 other 2 other TA;WB;7 other TA;WB;7 SNF8;UBE2Z;ATP5G1;TTLL6;CALCOCO2 SRR;SGSM2;TSR1 ADAMTS7;CHRNA5;CTSH HHIPL1 HHIPL1 ALDH2;TMEM116;HECTD4;MAPKAPK5-AS1; C12orf30;ERP29 INT INT INT INT INT INT INT SNF8, GIP, UBE2Z, SNF8, GIP, ATP5G1 SMG6, SRR ADAMTS7 HHIPL1 HHIPL1 COL4A1/A2 COL4A1/A2 COL4A1/A2 ALDH2 Table showing annotations for SNPs including gene expression derived from Haploreg (version 4.1), Table rs46522 rs216172 rs7173743 rs10139550 rs2895811 rs9515203 rs11838776 rs4773144 rs10744777 Abbreviations: ID, identifier; eQTL, expression quantitative trait; Enh, enhancer; INT, intron; 3’UTR, 3’ untranslated ate with gene expression (eQTL genes) in various tissues tissue). Abbreviations: ID, identifier; eQTL, quantitative trait; Enh, enhancer; INT, transformed lymphocytes; WB, whole blood, TSS1500, lymphoblast; MON, monocytes; TLY, tibial artery; LYM, region; NSM, non-synonymous; AA, aortic artery; CA, coronary TA, within the gene. <1500bp of transcription start site, TSS200, <200bp site; Body,
20162300 Marten Siemelink.indd 109 15-03-16 13:23 110 | CHAPTER 6
References
1 Dichgans M, Malik R, König IR, et al. Shared genetic susceptibility to ischemic stroke and coronary artery disease: a genome-wide analysis of common variants. Stroke 2014; 45: 24–36. 2 Zdravkovic S, Wienke A, Pedersen NL, Marenberg ME, Yashin AI, De Faire U. Heritability of death from coronary heart disease: a 36-year follow-up of 20 966 Swedish twins. J Intern Med 2002; 252: 247–54. 3 Flossmann E, Schulz UGR, Rothwell PM. Systematic review of methods and results of studies of the genetic epidemiology of ischemic stroke. Stroke 2004; 35: 212–27. 4 Bevan S, Traylor M, Adib-Samii P, et al. Genetic heritability of ischemic stroke and the contribution of previously reported candidate gene and genomewide associations. Stroke 2012; 43: 3161–7. 5 Traylor M, Rutten-Jacobs LCA, Holliday EG, et al. Differences in Common Genetic Predisposition to Ischemic Stroke by Age and Sex. Stroke 2015; 46: 3042–7. 6 Tada H, Melander O, Louie JZ, et al. Risk prediction by genetic risk scores for coronary heart disease is independent of self-reported family history. Eur Heart J 2015; published online Sept 20. DOI:10.1093/ eurheartj/ehv462. 7 Liu Y, Li X, Aryee MJ, et al. GeMes, clusters of DNA methylation under genetic control, can inform genetic and epigenetic analysis of disease. Am J Hum Genet 2014; 94: 485–95. 8 Pai AA, Pritchard JK, Gilad Y. The Genetic and Mechanistic Basis for Variation in Gene Regulation. PLoS Genet 2015; 11: e1004857. 9 Olsson AH, Volkov P, Bacos K, et al. Genome-wide associations between genetic and epigenetic vari- ation influence mRNA expression and insulin secretion in human pancreatic islets. PLoS Genet 2014; 10: e1004735. 10 Bell JT, Pai AA, Pickrell JK, et al. DNA methylation patterns associate with genetic and gene expres- sion variation in HapMap cell lines. Genome Biol 2011; 12: R10. 11 Grundberg E, Meduri E, Sandling JK, et al. Global analysis of DNA methylation variation in adipose tis- sue from twins reveals links to disease-associated variants in distal regulatory elements. Am J Hum Genet 2013; 93: 876–90. 12 Lemire M, Zaidi SHE, Ban M, et al. Long-range epigenetic regulation is conferred by genetic variation located at thousands of independent loci. Nat Commun 2015; 6: 6326. 13 Verhoeven BAN, Velema E, Schoneveld AH, et al. Athero-express: differential atherosclerotic plaque expression of mRNA and protein in relation to cardiovascular events and patient characteristics. Ra- tionale and design. Eur J Epidemiol 2004; 19: 1127–33. 14 van Iterson M, Tobi EW, Slieker RC, et al. MethylAid: visual and interactive quality control of large Il- lumina 450k datasets. Bioinformatics 2014; 30: 3435–7. 15 Fortin J-P, Labbe A, Lemire M, et al. Functional normalization of 450k methylation array data improves replication in large cancer studies. Genome Biol 2014; 15: 503. 16 Aryee MJ, Jaffe AE, Corrada-Bravo H, et al. Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics 2014; 30: 1363–9. 17 Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B 1995; 57: 289–300. 18 Welter D, MacArthur J, Morales J, et al. The NHGRI GWAS Catalogue, a curated resource of SNP-trait associations. Nucleic Acids Res 2014; 42: D1001–6. 19 A comprehensive 1000 Genomes-based genome-wide association meta-analysis of coronary artery disease. Nat Genet 2015; 47: 1121–30. 20 Kilarski LL, Achterberg S, Devan WJ, et al. Meta-analysis in more than 17,900 cases of ischemic stroke reveals a novel association at 12q24.12. Neurology 2014; 83: 678–85. 21 Traylor M, Mäkelä K-M, Kilarski LL, et al. A novel MMP12 locus is associated with large artery ath- erosclerotic stroke using a genome-wide age-at-onset informed approach. PLoS Genet 2014; 10: e1004469. 22 Matsukura M, Ozaki K, Takahashi A, et al. Genome-Wide Association Study of Peripheral Arterial Dis- ease in a Japanese Population. PLoS One 2015; 10: e0139262. 23 Arnold M, Raffler J, Pfeufer A, Suhre K, Kastenmüller G. SNiPA: an interactive, genetic variant-cen- tered annotation browser. Bioinformatics 2014; 31: 1334–6. 24 Sherry ST, Ward MH, Kholodov M, et al. dbSNP: the NCBI database of genetic variation. Nucleic Acids Res 2001; 29: 308–11. 25 Johnson AD, Handsaker RE, Pulit SL, Nizzari MM, O’Donnell CJ, de Bakker PIW. SNAP: a web-based tool for identification and annotation of proxy SNPs using HapMap. Bioinformatics 2008; 24: 2938–9. 26 van der Laan SW, Asl HF, van den Borne P, et al. Variants in ALOX5, ALOX5AP and LTA4H are not asso- ciated with atherosclerotic plaque phenotypes: The Athero-Express Genomics Study. Atherosclerosis 2015; 239: 528–38.
20162300 Marten Siemelink.indd 110 15-03-16 13:23 CVD LOCI ASSOCIATE WITH PLAQUE METHYLATION | 111
27 Laurie CC, Doheny KF, Mirel DB, et al. Quality control and quality assurance in genotypic data for genome-wide association studies. Genet Epidemiol 2010; 34: 591–602. 28 Marchini J, Howie B, Myers S, McVean G, Donnelly P. A new multipoint method for genome-wide association studies by imputation of genotypes. Nat Genet 2007; 39: 906–13. 29 The Genotype-Tissue Expression (GTEx) project. Nat Genet 2013; 45: 580–5. 30 Ward LD, Kellis M. HaploReg: a resource for exploring chromatin states, conservation, and regulatory motif alterations within sets of genetically linked variants. Nucleic Acids Res 2012; 40: D930–4. 31 Varley KE, Gertz J, Bowling KM, et al. Dynamic DNA methylation across diverse human cell lines and tissues. Genome Res 2013; 23: 555–67. 32 Curradi M, Izzo A, Badaracco G, Landsberger N. Molecular mechanisms of gene silencing mediated by DNA methylation. Mol Cell Biol 2002; 22: 3157–73. 33 Ma A-N, Wang H, Guo R, et al. Targeted gene suppression by inducing de novo DNA methylation in the gene promoter. Epigenetics Chromatin 2014; 7: 20. 34 Maunakea AK, Nagarajan RP, Bilenky M, et al. Conserved role of intragenic DNA methylation in regulat- ing alternative promoters. Nature 2010; 466: 253–7. 35 Breitling LP, Salzmann K, Rothenbacher D, Burwinkel B, Brenner H. Smoking, F2RL3 methylation, and prognosis in stable coronary heart disease. Eur Heart J 2012; 33: 2841–8. 36 Zhang Y, Yang R, Burwinkel B, et al. F2RL3 methylation in blood DNA is a strong predictor of mortality. Int J Epidemiol 2014; 43: 1215–25. 37 Castillo-Díaz SA, Garay-Sevilla ME, Hernández-González MA, Solís-Martínez MO, Zaina S. Extensive demethylation of normally hypermethylated CpG islands occurs in human atherosclerotic arteries. Int J Mol Med 2010; 26: 691–700. 38 Zaina S, Heyn H, Carmona FJ, et al. DNA methylation map of human atherosclerosis. Circ Cardiovasc Genet 2014; 7: 692–700. 39 Smith AK, Kilaru V, Kocak M, et al. Methylation quantitative trait loci (meQTLs) are consistently de- tected across ancestry, developmental stage, and tissue type. BMC Genomics 2014; 15: 145. 40 Kato N, Loh M, Takeuchi F, et al. Trans-ancestry genome-wide association study identifies 12 genetic loci influencing blood pressure and implicates a role for DNA methylation. Nat Genet 2015; 47: 1282– 93. 6
20162300 Marten Siemelink.indd 111 15-03-16 13:23 112 | CHAPTER 6
Supplemental Material
Supplement to: Siemelink, MA et al, Cardiovascular risk loci associate with DNA methylation in carotid plaques. Supplemental material available on publication and on request.
Table of Contents Supplemental Tables Supplemental Table 1. GWAS studies included in the study Supplemental Table 2. CVD-risk SNPs from GWAS studies Supplemental Table 3. Workflow for SNPs Supplemental Table 4. Comparison of meQTL effects in plaque and blood
Supplemental Figures Supplemental Figure 1. Locuszoom plots of all significant meQTL associations Supplemental Figure 2. Scatterplot of meQTL association compared to distance Supplemental Figure 3. Barplot of meQTL distribution over regions Supplemental Figure 4. Barplot of meQTL distribution over islands Supplemental Figure 5. QQplot of plaque and blood meQTL results
Supplemental References
20162300 Marten Siemelink.indd 112 15-03-16 13:23 | 113
20162300 Marten Siemelink.indd 113 15-03-16 13:23 M.A. Siemelink1 S.W. van der Laan1 S. Haitjema1 K.F. Dekkers2 R. Luijk2,3 H. Foroughi Asl4 T. Michoel5 J.L.M. Björkegren4 E. Aavik6 S. Ylä-Herttuala6,7 G.J. de Borst8 F.W. Asselbergs9,10,11 H. el Azzouzi9 H.M. den Ruijter1 B.T. Heijmans2 G. Pasterkamp1,12
1 Laboratory of Experimental Cardiology, University Medical Center Utrecht, Utrecht, the Netherlands 2 Department of Molecular Epidemiology, Leiden University Medical Center, Leiden, the Netherlands 3 Medical Statistics and Bioinformatics, Leiden University Medical Center, Leiden, the Netherlands 4 Cardiovascular Genomics Group, Division of Vascular Biology, Department of Medical Biochemistry and Biophysics, Karolinska Institutet, Stockholm, Sweden 5 Division of Genetics and Genomics, the Roslin Institute, University of Edinburgh, Edinburgh, United Kingdom 6 Department of Biotechnology and Molecular Medicine, A.I. Virtanen Institute, University of Eastern Finland, Kuopio, Finland 7 Science Service Center and Gene Therapy Unit, Kuopio University Hospital, Kuopio, Finland 8 Department of Vascular Surgery, University Medical Center Utrecht, Utrecht, the Netherlands 9 Department of Cardiology, University Medical Center Utrecht, Utrecht, the Netherlands 10 Durrer Center for Cardiogenetic Research, ICIN-Netherlands Heart Institute, Utrecht, the Netherlands 11 Institute of Cardiovascular Science, Faculty of Population Health Sciences, University College London, London, United Kingdom 12 Laboratory of Clinical Chemistry and Hematology, University Medical Center Utrecht, Utrecht, the Netherlands
20162300 Marten Siemelink.indd 114 15-03-16 13:23 CHAPTER 7
Smoking Associated DNA Methylation in Atherosclerotic Carotid Arteries is Predictive of Cardiovascular Risk
MANUSCRIPT IN PREPARATION
20162300 Marten Siemelink.indd 115 15-03-16 13:23 116 | CHAPTER 7
Abstract
Background Tobacco smoking is a major risk factor for atherosclerotic disease and has been associated with DNA methylation patterns in blood cells. However, the effect of smoking on DNA methylation in the vascular lesion is unknown and may prove crucial in understanding the pathophysiology of atherosclerosis. In the current study we performed an epigenome-wide association study (EWAS) to investigate the association between tobacco smoking and DNA methylation in atherosclerotic plaques from patients undergoing carotid endarterectomy.
Methods and Results DNA methylation was assessed by microarray in atherosclerotic plaque samples and peripheral blood samples from 482 patients who underwent carotid endarterectomy. The association of DNA methylation with tobacco smoking at the time of study inclusion was determined. In atherosclerotic plaques, 68 of the 443,872 epigenetic loci were associated with current tobacco smoking. The strongest association with smoking in atherosclerotic plaques was observed at cg25648203 in the gene AHRR, a repressor of the aryl hydrocarbon receptor transcription factor involved in xenobiotic detoxification. Six of the novel CpG loci (near genes PRDM16, GLIS1, SORBS1, NTHL1, and intergenic at 8q24.21) were associated with cardiovascular events after 3 years. Four of the smoking-related CpG loci in atherosclerotic plaques were also found to be associated with genetic variation at nearby SNPs.
Conclusions Tobacco smoking is associated with DNA methylation at 68 loci in atherosclerotic lesions of carotid endarterectomy patients; six of these loci are predictive for the three-year risk of cardiovascular events within this population.
20162300 Marten Siemelink.indd 116 15-03-16 13:23 SMOKING ASSOCIATES WITH PLAQUE METHYLATION | 117
Introduction
Tobacco smoking is a major risk factor for the development of atherosclerosis and subsequent cardiovascular disease (CVD), such as myocardial infarction and stroke. Tobacco smoke contains over 5,000 toxic chemicals which may jointly contribute to CVD risk.1 Smoking activates the immune system, facilitates pro-atherogenic lipid profiles, and induces a prothrombotic state.2,3 Moreover, smoking affects the vascular wall, leading to endothelial dysfunction and atherosclerosis.4 Histological examination of plaques of smokers have shown increased atheroma, decreased fibrous volume,5 more plaque hemorrhage,6 and increased inflammation and tissue destruction.7 All these changes contribute to a plaque composition that is more vulnerable to rupture and more likely to cause cardiovascular events. Yet, a detailed understanding of the pathophysiological mechanisms underlying these changes remains elusive. Such an understanding may help to identify patients at increased risk due to smoking and may contribute to cessation strategies. Of equally importance, it may show common pathophysiological pathways of atherosclerosis, shared by multiple risk factors, which may be important drug targets. The large-scale genetic association studies (GWAS) have proven instrumental in the investigation of many cardiovascular risk factors. However, GWAS results of smoking have mainly identified genetic variants influencing the tendency for smoking.8 Still, smoking has been shown to contribute to CVD risk by modulating the effect of genetic variants on cardiovascular risk factors.9–12 Yet identification of the pathophysiology caused by environmental exposures, such as smoking induced cardiovascular risk, may require other approaches. In contrast to the genetic sequence, epigenetic markers like DNA methylation are commonly affected by environmental factors during life, which can lead to altered gene expression.13 This has been shown in blood cells for several cardiovascular risk factors such as Body- 7 Mass Index (BMI)14 and blood lipid levels.15 Chemicals in tobacco smoke may change gene expression through DNA methylation, either adaptive or pathologic. Such epigenetic changes have indeed been shown in blood cells, in which numerous genetic loci (at adjacent cytosine and guanine bases; CpGs) affected by smoking have been identified by epigenome- wide association studies (EWAS).16–23 Conceivably, the most important insights in vascular pathology may be obtained by scrutinizing DNA methylation in the vascular lesion itself. Yet, to our knowledge, the effect of smoking on DNA-methylation in the vascular lesion has not been studied. In the current study, we performed an epigenome-wide association study of tobacco smoking in carotid atherosclerotic plaques of patients undergoing carotid endarterectomy, confirming known loci as well as reporting novel plaque-specific loci. In addition, we show that common genetic variation associates with methylation and gene expression near some of these loci. Finally, we observed CpGs that associate with cardiovascular events which occurred following carotid surgery. Together our findings point to genetic and epigenetic mechanisms of smoking-induced cardiovascular disease.
20162300 Marten Siemelink.indd 117 15-03-16 13:23 118 | CHAPTER 7
Methods
Patient inclusion and follow-up The Athero-Express Biobank Study (AE) is an ongoing longitudinal biobank study including patients that undergo arterial endarterectomy in two Dutch referral centers since 2002. A detailed description of the cohort study design has been previously published.24 For the present study, patients were included who underwent carotid endarterectomy (CEA) and of which biomaterial and genotyping data were available. Clinical data were extracted from patient medical files and standardized questionnaires. Tobacco smoking was defined as current smoking at the time of admission for CEA and was assessed by questionnaire. Patients included in the AE were assessed for cardiovascular events using questionnaires at one, two, and three years after CEA procedure. Events were validated by reviewing the general practitioner’s patient file. Cardiovascular events included: cerebral infarction, cerebral hemorrhage, myocardial infarction, coronary bypass, coronary angioplasty, peripheral angioplasty, leg amputation, fatal heart failure, fatal aneurysm rupture, sudden death, and other cardiovascular death. The medical ethics committees of the respective hospitals approved the study and informed consent was obtained from all patients.
Sample collection Blood samples were obtained from the radial artery catheter immediately prior to surgery and stored at -80 degrees. Carotid plaque specimens were removed during surgery and processed in the laboratory. Specimens were cut transversely into segments of 5 mm. The culprit lesion was identified, fixed in 4% formaldehyde, embedded in paraffin and processed for histological examination. Plaque histological features were routinely scored through chemical- and immunohistochemical techniques as described in the supplemental methods. Remaining segments were stored at -80 degrees.
Methylation array DNA was isolated from stored plaque segments and stored blood samples of patients and used for DNA methylation measurements. DNA purity and concentration was assessed using the Nanodrop 1000 system (Thermo Scientific, Massachusetts, USA). DNA concentrations were equalized at 600ng, randomized over the 96-well plates and bisulfate converted using a cycling protocol and the EZ-96 DNA methylation kit (Zymo Research, Orange County, USA). Subsequently, DNA methylation was measured on the Infinium HumanMethylation450 Beadchip Array (HM450k, Illumina, San Diego, USA), which was performed at the Erasmus Medical Center Human Genotyping Facility in Rotterdam, The Netherlands. Processing of the sample and array was performed according to the manufacturer’s protocol.
Quality control of methylation data The quality of the HM450k array data was determined with the MethylAid R-package25 using default settings, controlling for sample dependent and probe dependent parameters. Bisulfate conversion efficiency was determined using dedicated probes on the HM450k. A total of 41,640 probes were excluded, with 443,872 probes (91.4 %) of good quality remaining. We used ‘Functional Normalization’26 (implemented in the minfi R-package)27 with eight control-probe principal components for normalization and correction of batch effects. No imputing of data was performed.
20162300 Marten Siemelink.indd 118 15-03-16 13:23 SMOKING ASSOCIATES WITH PLAQUE METHYLATION | 119
Samples with missing phenotype or covariates were excluded. Principal components of CpG-probes were calculated and used to identify possible mix up of samples by tissue type (i.e. blood or plaque DNA) or sex. Where available, genotype data was correlated to the raw data of the 65 SNPs included on the HM450k array, and samples with poor correlation (R ≤ 0.6) across these 65 SNPs were excluded due to possible mix up. After quality control, 475 plaque samples and 91 blood samples obtained from 482 unique patients were remaining. A flow-chart summarizing quality control of samples is presented in Supplemental Figure 1.
Epigenome-wide methylation analysis Epigenome-wide associations were determined using linear regression modeling with the Fast-LMM algorithm.28 Analysis was confined to autosomal chromosomes. Invariable probes (beta-values of all samples within range 0-0.1 or 0.9-1) were removed prior to analysis. Outliers at more than three standard deviations from the mean were removed from each probe. Linear regression modeling was performed with covariates age, sex, hospital of inclusion, and principal components. Resulting p values were corrected for residual inflation using genomic control.29 To control for multiple testing p values were corrected using the Benjamini-Hochberg False-Discovery Rate (FDR), with resulting FDR p values ≤ 0.05 considered statistically significant.30,31 Statistical analyses were performed in R Studio (v0.98.1091) using R (v3.1.3). The relationship between CpGs and nearby genes was described based on the Illumina HM450K annotation file, with CpGs not described as related to a gene denoted as intergenic.
DNA methylation as predictor of clinical events Epigenome-wide significant CpGs found in atherosclerotic plaque samples were tested for association with clinical events within the three years of follow-up. Cox proportional 7 hazard modeling was performed adjusted for age, sex and current smoking behavior. Hazard ratios (HR) were determined per standard deviation of the normalized DNA methylation β-values at a locus. Cox survival plots were drawn by stratifying patients to upper and lower 50-percentiles based on DNA methylation β-values at a locus.
Genotyping & methylation Quantitative Trait Analysis DNA was isolated from stored samples and genotyping was performed in two series using commercially available genotyping arrays.32 The first series (Athero-Express Genomics Study 1, AEGS1) was genotyped using Affymetrix Genome-Wide Human SNP Array 5.0, the second (Athero-Express Genomics Study 2, AEGS2) was genotyped using the Affymetrix Axiom® GW CEU 1 Array. We adhered to community standard quality control and assurance procedures to clean the genotype data obtained in AEGS1 and AEGS2.33 We used phased haplotypes from the 1000 Genomes Project as the reference panel for genotype imputation. Imputed genotypes were available for 454 patients. We used SNPTEST34 (version 2.5) for the identification of methylation quantitative trait loci (meQTL) in a linear regression model corrected for covariates age, sex, SNP array type, genotyping principal components 1 and 2, methylation principal components 1 and 2, and current smoking status. An additive genetic model was used and only high-quality imputed variants (Minor Allele Frequency (MAF) ≥ 0.05; imputation quality ≥ 0.9; Hardy-Weinberg Equilibrium (HWE) p value ≥ 1.0x10-6) in cis, i.e. within 500kb of the CpG, were considered.
20162300 Marten Siemelink.indd 119 15-03-16 13:23 120 | CHAPTER 7
Results
We performed an epigenome-wide association of plaque-derived DNA methylation with smoking in 475 carotid endarterectomy derived plaque samples. Clinical characteristics of the patients included in the study are provided in table 1. The results show that smoking was significantly associated with differential DNA methylation at 68 CpG loci, of which the majority (46 CpGs) showed lower DNA methylation (Figure 1, Supplemental Table 1, Supplemental Figures 2 and 3). 28 of these CpGs have previously been associated with smoking in circulating blood cells. We observed 40 novel CpGs in plaque samples that have not been associated with smoking previously. The most significantly associated CpG (cg25648203, β = -0.29 ± 0.02, FDR-p = 1.1 x 10-22) was in the aryl hydrocarbon receptor repressor (AHRR) gene.
Table 1. Cohort patient characteristics Characteristics Smokers (N = 191) Non smokers (N = 284) missing* median (IQR) missing* median (IQR) p value Age years 0.0% 65 (59 – 71) 0.0% 71 (64 – 76) 7.2x10-9 SBP mmHg 10.5% 150 (137 – 170) 14.1% 155 (140 – 170) 0.713 DBP mmHg 10.5% 80 (75 – 93) 14.1% 82 (73 – 90) 0.870 eGFR ml/min/1.73m2 4.7% 77 (64 – 88) 2.1% 70 (56 – 84) 0.001 BMI kg/m2 6.2% 25.8 (23.8 – 28.7) 4.9% 26.3 (24.4 – 28.4) 0.072 hsCRP mg/l 47.6% 2.6 (1.2 – 7.3) 53.9% 1.8 (1.0 – 3.6) 0.066
missing* N (percentage) missing* N (percentage) p value Sex male 0.0% 128 (67.0) 0.0% 206 (72.5) 0.236 Diabetes 0.0% 42 (22.0) 0.0% 63 (22.2) 0.999 LLDs 0.0% 147 (77.0) 0.0% 216 (76.1) 0.906 Symptoms† 0.0% 87 (45.5) 0.4% 125 (44.0) 0.791 Asymptomatic 28 (14.7) 50 (17.7) TIA 87 (45.5) 125 (44.2) Stroke 45 (23.6) 74 (26.1) Retinal 31 (16.2) 34 (12.0) Patient characteristics at time of inclusion, stratified by smoking status. Patients without data on current smoking were excluded. *Percentage of patients with missing data. †Symptoms at presen- tation, before carotid endarterectomy. Significance shown as p values without FDR adjustment. Ab- breviations: SBP, systolic blood pressure; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate by MDRD-formula; BMI, body-mass index; LLDs, use of lipid-lowering drugs; Retinal, retinal infarction and amaurosis fugax.
20162300 Marten Siemelink.indd 120 15-03-16 13:23 SMOKING ASSOCIATES WITH PLAQUE METHYLATION | 121 (p value) of the association with smok - 10
7 ing. P values have been adjusted for inflation using genomic control. CpGs that were epigenome-wide significant after false-discovery rate correction are shown in red. Figure 1. Manhattan plot of the association DNA methylation in carotid atherosclerotic plaque with tobacco smoking. Each point represents an individual CpG, with the x-axis shows genomic location of each CpG and y-axes observed –log
20162300 Marten Siemelink.indd 121 15-03-16 13:23 122 | CHAPTER 7
To study the possible effect of smoking-induced methylation changes on the carotid atherosclerotic plaque in more detail, we investigated histological features of the plaques. Smoking itself was not directly associated with any of the atherosclerotic plaque features (Supplemental Table 2). Of the 68 smoking associated CpGs, a single CpG in the gene RARA (cg19572487, β = 0.02 ± 0.005, FDR-p = 0.038) was associated with an increase in plaque atheroma.
Clinical events We proceeded to investigate if the smoking-associated CpGs in plaque were associated with clinical manifestations of cardiovascular disease irrespective of smoking status. We performed multivariable Cox-proportional hazards analysis for cardiovascular events during follow-up. For six CpGs, near four genes (PRDM16, GLIS1, SORBS1, NTHL1) or intergenic (at 8q24.21), we found a significant association with cardiovascular events within three years of follow-up (Figure 2, Supplemental Figure 4), after correction for age, sex and smoking. The strongest predictor was cg18767735 near GLIS1 (HR = 0.69 ± 0.08, FDR-p = 4.8x10-4). The direction of effect between smoking and these six CpGs, and the relative risk associated with these CpGs, was consistent with the known increase of cardiovascular risk due to smoking. For example, smoking was associated with an increase in DNA methylation at cg03190891 near SORBS1 (beta = 0.21, SE = 0.04, FDR-p = 2.0x10-3) and increased DNA methylation at this locus was associated with an increased cardiovascular risk (HR = 1.32 ± 0.09, FDR p = 3.6x10-2). Interestingly, none of these CpGs have previously been associated with smoking in blood (Supplemental Table 1), which supports the notion that they may reflect vascular pathophysiological mechanisms that result in cardiovascular events.
Genetic variation We associated DNA methylation at the smoking-associated CpGs in plaque with nearby SNPs35, which identified 126 SNPs (methylation quantitative trait loci; meQTL) that associate with four CpG loci (table 2, Supplemental Figure 5). This suggests that DNA methylation at these smoking-related CpGs may be also be affected by genetic variation. To investigate if these meQTL SNPs also indicate co-regulatory gene-gene interaction, we determined the relationships between SNPs, CpGs, and the expression of the involved genes. We found eight SNPs to be expression quantitative trait loci (eQTL), possibly
Table 2. SNPs associated with DNA methylation in plaque samples CpG Gene meQTL SNP Chr CpG pos SNP pos Allele Beta±SE FDR p value cg09935388 GFI1 rs74767605 1 92947588 92770376 G/A -0.20 ± 0.04 1.0x10-2 cg04135110 AHRR rs7717970 5 346695 181660 G/A -0.56 ± 0.06 1.9x10-9 cg11557553 AHRR rs2288461 5 404996 230980 G/A -0.55 ± 0.12 3.9x10-2 cg09841788 SPTB rs73273524 14 65213478 65193767 T/C -0.30 ± 0.05 6.8x10-5 SNPs (meQTL) that associate with smoking-related CpGs in atherosclerotic plaques, only the most significant SNP is reported. Abbreviations: meQTL, methylation quantitative trait locus; Chr, chro- mosome; CpG pos, base pair position of CpG; SNP pos, base pair position of SNP; Allele, coding and non-coding alleles; SE, standard error; FDR, false discovery rate
20162300 Marten Siemelink.indd 122 15-03-16 13:23 SMOKING ASSOCIATES WITH PLAQUE METHYLATION | 123
Figure 2. Forest plot of the hazard ratios associated with DNA methylation. Smoking associated CpGs that associate with the occurrence of cardiovascular events within three years of carotid en- darterectomy, independent of smoking. Cox proportional hazard modeling was performed corrected for age, sex, hospital of inclusion, and smoking. Cardiovascular events included: cerebral infarction, cerebral bleeding, myocardial infarction, coronary bypass, coronary angioplasty, percutaneous trans- luminal angioplasty, leg amputation, fatal heart failure, fatal aneurysm rupture, sudden death, and cardiovascular death by any other causes. For each CpG, the association with smoking is presented (odds ratio), as well as the association with cardiovascular events (hazard ratios and confidence in- tervals). All measures are shown per standard deviation of DNA methylation at the locus in carotid plaques. *cg27509867 is intergenic at genomic location 8q24.21. Abbreviations: OR, Odds ratio; HR, hazard ratio; CI, confidence interval.
affecting expression of the gene PLEKHG3 (Supplemental Table 3, Supplemental Figure 6) of which the best associated was rs229660 (p = 1.2x10-176).9,36,37 In addition, we found two non-synonymous SNPs, rs4956987 and rs12519352, that may alter the function of gene PLEKHG4B (Supplemental Table 3). Finally, we show positive associations between 7 expression of the genes AHRR and PLEKHG4B as well as between SPTB and PLEKHG3, in multiple CVD related tissues in the STAGE-cohort (Supplemental Table 4). In light of these results, we speculate that PLEKHG3 and PLEKHG4B may be co-regulatory genes that epigenetically regulate SPTB- and AHRR expression (Supplemental Figure 7). To gain further insight into the potential roles of these meQTL SNPs in various tissues relevant for disease, we determined the association of meQTL variants with expression of more distant genes (in trans). This showed eQTL associations for two SNPs (rs140417889 and rs28670180, both also meQTL for cg04135110 near AHRR) with expression of several genes (SCEL, IL20, PRR9, IGSF9, ABHD12B and SDHAP3) in various tissues (Supplemental Table 5).
DNA methylation in Blood Additionally, we performed epigenome-wide associations for smoking with DNA methylation in 91 blood samples (Supplemental Figures 2 and 8). We found 21 significant CpGs in blood, of which 18 have previously been associated with smoking in blood and other tissues (Supplemental Table 6),16–19,21,38–41 confirming the relevance of previously reported loci in patients with severe atherosclerotic disease. The most significant CpG in blood samples (cg05575921, β = -0.39 ± 0.04 FDR-p = 1.1x10-9) was also located in the AHRR gene and correlated to plaque methylation at the most-significant locus cg25648203 (ρ = 0.63, FDR-p = 2.2x10-16).
20162300 Marten Siemelink.indd 123 15-03-16 13:23 124 | CHAPTER 7
Comparing methylation in plaque and blood samples To increase the understanding of tissue specificity of the plaque CpG, we investigated if smoking had similar effects on DNA-methylation in plaque and blood samples (Supplemental Figure 9). This showed that for some CpGs, smoking had either similar effect on DNA methylation in plaque and blood samples, while other CpGs were only affected in plaque samples. To identify blood CpGs that may have potential as a biomarker of smoking-induced plaque methylation, we performed hierarchical clustering based on correlations between the plaque and blood samples of identical patients (n=84). This showed that 21 smoking-related CpGs in blood were associated with DNA-methylation in plaque in two apparent clusters, one correlating positively with 13 plaque CpGs, another correlating negatively with 21 plaque CpGs (Supplemental Figure 10). Interestingly, four of the six CpGs (cg27509867, cg18767735, cg01815912, and cg16650073) that associate with cardiovascular events in plaque were in the first cluster and were strongly correlated to several CpGs in blood, including cg03636183 near F2RL3, a previously reported biomarker for smoking induced mortality.42
Discussion
Presenting data from an epigenome-wide association study of smoking in carotid atherosclerotic plaque samples, we show that smoking is strongly associated with differential DNA methylation in carotid atherosclerotic plaques. A total of 40 novel epigenome-wide significant CpGs in plaque samples were associated with smoking, while we could replicate 28 CpG loci known to associate with smoking in circulating cells. Although results require verification in other studies, they provide strong supportive evidence for an effect of smoking on epigenetic regulation in atherosclerotic vascular tissue. This is strengthened by the partial similarity observed in DNA methylation patterns in blood and plaque. For example, the strongest associations with smoking were observed at CpG loci near AHRR, a regulator of the aryl hydrocarbon receptor (AhR) transcription factor and its pathway, which has been associated with smoking on numerous occasions.16–22,38–41 A lower DNA methylation at the AHRR gene has previously been associated with higher AHRR expression in human lung tissue and a mouse model.19 In previous studies, similar associations between smoking and CpG loci in the AHRR gene were found in other tissues including pulmonary macrophages39 and in cord-blood of neonates born to smoking mothers.43 The AhR transcription factor is a xenobiotic receptor, sensitive to some endogenous ligands as well as many exogenous toxins. These toxins include polycyclic aromatic hydrocarbons and dioxins both of which are important constituents of tobacco smoke44 and lead to upregulation of enzymes involved in the detoxifying metabolism of these substances. Our observations suggest that smoking affects the atherosclerotic vascular lesion at the epigenetic level, which may affect local gene expression levels. Although the concept of transcriptional regulation by DNA methylation has been abundantly shown, the effect of a particular CpG on local gene-expression is complex. Elucidation of the exact effects of these CpGs on gene expression within the atherosclerotic vascular wall may offer important insights into the biological mechanisms by which tobacco smoking confers an increased cardiovascular risk. In that respect, smoking-associated CpGs specific for plaque tissue are of particular interest.
20162300 Marten Siemelink.indd 124 15-03-16 13:23 SMOKING ASSOCIATES WITH PLAQUE METHYLATION | 125
An unfavorable plaque composition, i.e. the vulnerable plaque, has often been associated with clinical events and has sometimes been regarded as a surrogate endpoint for cardiovascular events. In our analysis, we found that DNA methylation at a single CpG near the RARA gene (retinoic acid receptor alpha) was associated with plaque atheroma. The RARA pathway has been shown to interact in the AhR pathway,45,46 suggesting a link between this detoxification pathway and atheroma formation. Most interestingly, we revealed an association of six CpGs with the occurrence of clinical manifestations of atherosclerotic disease, irrespective of smoking behavior. Conceivably, these CpGs may be involved in common pathways of atherosclerotic disease, shared between smoking and other risk factors. Therefore, these CpGs identify genes that may emerge as important candidates for biomarker studies or as potential drug targets. Genetic variation may also affect methylation status of specific genes. Using meQTL analysis, we found strong associations between CpGs and nearby SNPs, showing that some of the smoking-associated CpGs may be affected by genetic variation. Therefore, these SNPs are of particular interest since they may reveal hereditary susceptibility to toxicity in the vascular wall. Not much is known about the biological functions of the PLEKHG3 and PLEKHG4B genes, yet our data supports further research into their relationship with smoking. Most epigenetic smoking studies to date have focused on blood derived DNA-methylation. To gain better insight in the tissue specificity of the methylation results obtained in atherosclerotic plaques and to verify consistency with pre-existing studies, we also did an epigenome-wide association study in blood samples from the same patients. We carefully scrutinized literature investigating blood or other tissues. This showed that only a minority of CpGs found in plaque are also found to be associated with smoking in blood. This emphasizes the importance of investigating DNA methylation in the vascular lesion itself, as well as the need for replication. However, we are unaware of existing cohorts of sufficient size with appropriate data to enable validation of our results. Yet, many plaque-associated 7 CpGs are reported in other publications, and methylation at these CpGs is robustly correlated to blood methylation of the same patients. This strongly supports the validity of our results. Our analyses are based on patients’ current smoking behavior, which will not reflect time- dependent effects of smoking on plaque methylation47, as patients may be light or heavy smokers or may have recently ceased smoking. Although we show strong associations and correct for latent confounders using principal components, it is impossible to exclude some residual confounding. This is complicated by the differences in DNA methylation between cell-types in the sample, on which we elaborate in the supplement. Furthermore, it should be noted that the Athero-Express biobank is a cohort of patients with advanced atherosclerotic disease. Therefore, one should be cautious to make inferences on the contribution of these CpGs to cardiovascular disease risk in other populations, solely based on our results. In summary, we performed an epigenome-wide association study of current smoking in 475 atherosclerotic plaque samples and 91 peripheral blood samples derived from 482 carotid endarterectomy patients. We show that tobacco smoking is associated with DNA methylation at 68 loci in atherosclerotic lesions of carotid endarterectomy patients; six of these loci are predictive for the three-year risk of cardiovascular events within this population.
Acknowledgements The authors would like to express their gratitude to Aisha Gohar for her help with the manuscript.
20162300 Marten Siemelink.indd 125 15-03-16 13:23 126 | CHAPTER 7
References
1 Talhout R, Schulz T, Florek E, van Benthem J, Wester P, Opperhuizen A. Hazardous compounds in tobacco smoke. Int J Environ Res Public Health 2011; 8: 613–28. 2 Messner B, Bernhard D. Smoking and Cardiovascular Disease: Mechanisms of Endothelial Dysfunc- tion and Early Atherogenesis. Arterioscler Thromb Vasc Biol 2014; 34: 509–15. 3 Csordas A, Bernhard D. The biology behind the atherothrombotic effects of cigarette smoke. Nat Rev Cardiol 2013; 10: 219–30. 4 Herder M, Johnsen SH, Arntzen KA, Mathiesen EB. Risk factors for progression of carotid intima- media thickness and total plaque area: a 13-year follow-up study: the Tromsø Study. Stroke 2012; 43: 1818–23. 5 Kumagai S, Amano T, Takashima H, et al. Impact of cigarette smoking on coronary plaque composition. Coron Artery Dis 2015; 26: 60–5. 6 van den Bouwhuijsen QJA, Vernooij MW, Hofman A, Krestin GP, van der Lugt A, Witteman JCM. Determinants of magnetic resonance imaging detected carotid plaque components: the Rotterdam Study. Eur Heart J 2012; 33: 221–9. 7 Kangavari S, Matetzky S, Shah PK, et al. Smoking increases inflammation and metalloproteinase ex- pression in human carotid atherosclerotic plaques. J Cardiovasc Pharmacol Ther 2004; 9: 291–8. 8 The Tobacco and Genetics Consortium. Genome-wide meta-analyses identify multiple loci associated with smoking behavior. Nat Genet 2010; 42: 441–7. 9 Zeller T, Wild P, Szymczak S, et al. Genetics and beyond--the transcriptome of human monocytes and disease susceptibility. PLoS One 2010; 5: e10693. 10 Polfus LM, Smith JA, Shimmin LC, et al. Genome-wide association study of gene by smoking interac- tions in coronary artery calcification. PLoS One 2013; 8: e74642. 11 van Setten J, Isgum I, Smolonska J, et al. Genome-wide association study of coronary and aortic calci- fication implicates risk loci for coronary artery disease and myocardial infarction. Atherosclerosis 2013; 228: 400–5. 12 Hamrefors V, Hedblad B, Hindy G, et al. Smoking modifies the associated increased risk of future cardiovascular disease by genetic variation on chromosome 9p21. PLoS One 2014; 9: e85893. 13 Venolia L, Gartler SM. Comparison of transformation efficiency of human active and inactive X-chromo- somal DNA. Nature 1983; 302: 82–3. 14 Dick KJ, Nelson CP, Tsaprouni L, et al. DNA methylation and body-mass index: a genome-wide analy- sis. Lancet 2014; 383: 1990–8. 15 Pfeiffer L, Wahl S, Pilling LC, et al. DNA Methylation of Lipid-Related Genes Affects Blood Lipid Levels. Circ Cardiovasc Genet 2015; 8: 334–42. 16 Zeilinger S, Kühnel B, Klopp N, et al. Tobacco smoking leads to extensive genome-wide changes in DNA methylation. PLoS One 2013; 8: e63812. 17 Besingi W, Johansson A. Smoke-related DNA methylation changes in the etiology of human disease. Hum Mol Genet 2014; 23: 2290–7. 18 Breitling LP, Yang R, Korn B, Burwinkel B, Brenner H. Tobacco-smoking-related differential DNA meth- ylation: 27K discovery and replication. Am J Hum Genet 2011; 88: 450–7. 19 Shenker NS, Polidoro S, van Veldhoven K, et al. Epigenome-wide association study in the European Prospective Investigation into Cancer and Nutrition (EPIC-Turin) identifies novel genetic loci associated with smoking. Hum Mol Genet 2013; 22: 843–51. 20 Allione A, Marcon F, Fiorito G, et al. Novel epigenetic changes unveiled by monozygotic twins discord- ant for smoking habits. PLoS One 2015; 10: e0128265. 21 Dogan M V, Shields B, Cutrona C, et al. The effect of smoking on DNA methylation of peripheral blood mononuclear cells from African American women. BMC Genomics 2014; 15: 151. 22 Richmond RC, Simpkin AJ, Woodward G, et al. Prenatal exposure to maternal smoking and offspring DNA methylation across the lifecourse: findings from the Avon Longitudinal Study of Parents and Children (ALSPAC). Hum Mol Genet 2015; 24: 2201–17. 23 Joubert BR, Håberg SE, Nilsen RM, et al. 450K epigenome-wide scan identifies differential DNA meth- ylation in newborns related to maternal smoking during pregnancy. Environ Health Perspect 2012; 120: 1425–31. 24 Verhoeven BAN, Velema E, Schoneveld AH, et al. Athero-express: differential atherosclerotic plaque expression of mRNA and protein in relation to cardiovascular events and patient characteristics. Ra- tionale and design. Eur J Epidemiol 2004; 19: 1127–33. 25 van Iterson M, Tobi EW, Slieker RC, et al. MethylAid: visual and interactive quality control of large Il- lumina 450k datasets. Bioinformatics 2014; 30: 3435–7.
20162300 Marten Siemelink.indd 126 15-03-16 13:23 SMOKING ASSOCIATES WITH PLAQUE METHYLATION | 127
26 Fortin J-P, Labbe A, Lemire M, et al. Functional normalization of 450k methylation array data improves replication in large cancer studies. Genome Biol 2014; 15: 503. 27 Aryee MJ, Jaffe AE, Corrada-Bravo H, et al. Minfi: a flexible and comprehensive Bioconductor package for the analysis of Infinium DNA methylation microarrays. Bioinformatics 2014; 30: 1363–9. 28 Lippert C, Listgarten J, Liu Y, Kadie CM, Davidson RI, Heckerman D. FaST linear mixed models for genome-wide association studies. Nat Methods 2011; 8: 833–5. 29 Devlin B, Roeder K, Wasserman L. Genomic control, a new approach to genetic-based association studies. Theor Popul Biol 2001; 60: 155–66. 30 Storey JD, Tibshirani R. Statistical significance for genomewide studies. Proc Natl Acad Sci U S A 2003; 100: 9440–5. 31 Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B 1995; 57: 289–300. 32 van der Laan SW, Asl HF, van den Borne P, et al. Variants in ALOX5, ALOX5AP and LTA4H are not asso- ciated with atherosclerotic plaque phenotypes: The Athero-Express Genomics Study. Atherosclerosis 2015; 239: 528–38. 33 Laurie CC, Doheny KF, Mirel DB, et al. Quality control and quality assurance in genotypic data for genome-wide association studies. Genet Epidemiol 2010; 34: 591–602. 34 Marchini J, Howie B, Myers S, McVean G, Donnelly P. A new multipoint method for genome-wide association studies by imputation of genotypes. Nat Genet 2007; 39: 906–13. 35 Bell JT, Pai AA, Pickrell JK, et al. DNA methylation patterns associate with genetic and gene expres- sion variation in HapMap cell lines. Genome Biol 2011; 12: R10. 36 Grundberg E, Meduri E, Sandling JK, et al. Global analysis of DNA methylation variation in adipose tis- sue from twins reveals links to disease-associated variants in distal regulatory elements. Am J Hum Genet 2013; 93: 876–90. 37 The Genotype-Tissue Expression (GTEx) project. Nat Genet 2013; 45: 580–5. 38 Elliott HR, Tillin T, McArdle WL, et al. Differences in smoking associated DNA methylation patterns in South Asians and Europeans. Clin Epigenetics 2014; 6: 4. 39 Monick MM, Beach SRH, Plume J, et al. Coordinated changes in AHRR methylation in lymphoblasts and pulmonary macrophages from smokers. Am J Med Genet B Neuropsychiatr Genet 2012; 159B: 141–51. 40 Sun Y V, Smith AK, Conneely KN, et al. Epigenomic association analysis identifies smoking-related DNA methylation sites in African Americans. Hum Genet 2013; 132: 1027–37. 41 Wan ES, Qiu W, Baccarelli A, et al. Cigarette smoking behaviors and time since quitting are associated with differential DNA methylation across the human genome. Hum Mol Genet 2012; 21: 3073–82. 7 42 Zhang Y, Yang R, Burwinkel B, et al. F2RL3 methylation in blood DNA is a strong predictor of mortality. Int J Epidemiol 2014; 43: 1215–25. 43 Joubert BR, Håberg SE, Nilsen RM, et al. Research | Children ’ s Health 450K Epigenome-Wide Scan Identifies Differential DNA Methyla in Newborns Related to Maternal Smoking during Pregnancy. Environ Health Perspect 2012; 120: 1425–32. 44 Klingbeil EC, Hew KM, Nygaard UC, Nadeau KC. Polycyclic aromatic hydrocarbons, tobacco smoke, and epigenetic remodeling in asthma. Immunol Res 2014; 58: 369–73. 45 Widerak M, Ghoneim C, Dumontier M-F, Quesne M, Corvol MT, Savouret J-F. The aryl hydrocarbon receptor activates the retinoic acid receptoralpha through SMRT antagonism. Biochimie; 88: 387–97. 46 Ohno M, Ikenaka Y, Ishizuka M. All-trans retinoic acid inhibits the recruitment of ARNT to DNA, result- ing in the decrease of CYP1A1 mRNA expression in HepG2 cells. Biochem Biophys Res Commun 2012; 417: 484–9. 47 Guida F, Sandanger TM, Castagné R, et al. Dynamics of smoking-induced genome-wide methylation changes with time since smoking cessation. Hum Mol Genet 2015; 24: 2349–59.
20162300 Marten Siemelink.indd 127 15-03-16 13:23 128 | CHAPTER 7
Supplemental Material
Supplement to: Siemelink, MA et al, Smoking associated DNA methylation in atherosclerotic carotid arteries is predictive of cardiovascular risk. Supplemental material available on publication and on request.
Table of Content Supplemental Methods
Supplemental Results
Supplemental Tables Supplemental Table 1. Epigenome-wide significant CpGs that associate with smoking in carotid plaque Supplemental Table 2. Association of smoking with carotid plaque histological features Supplemental Table 3. Functional annotation of meQTL SNPs Supplemental Table 4. meQTL Gene-Gene expression associations in STAGE Supplemental Table 5. Association of meQTL SNPs with mRNA expression in STAGE Supplemental Table 6. Epigenome-wide significant CpGs that associate with smoking in whole blood
Supplemental Figures Supplemental Figure 1. Flowchart of samples excluded during quality control Supplemental Figure 2. Quantile-quantile plot of epigenome-wide associations of smoking with DNA methylation in carotid plaques and whole-blood Supplemental Figure 3. Vulcano plot of smoking-association for CpG loci in plaque Supplemental Figure 4. Survival plots for cardiovascular events Supplemental Figure 5. Regional plots of CpG loci with associated genetic variants Supplemental Figure 6. eQTL analysis of rs229660 with PLEKHG3 expression in whole blood Supplemental Figure 7. Schematic view of smoking-associated CpGs with putative epigenetic gene regulation Supplemental Figure 8. Manhattan plot showing the genomic distribution of p values of the association of whole-blood DNA methylation with smoking Supplemental Figure 9. Scatter plot comparing the effect of smoking on DNA-methylation in plaque and blood samples Supplemental Figure 10. Heatmap of hierarchical clustering between plaque and blood methylation Supplemental Figure 11. Quantile-quantile plots of models for inflation correction in plaque
Supplemental References
20162300 Marten Siemelink.indd 128 15-03-16 13:23 | 129
20162300 Marten Siemelink.indd 129 15-03-16 13:23 M.A. Siemelink1* S. Haitjema1* S.W. van der Laan1 R. Luijk2 K.F. Dekkers2 M. Mokry3 G.J. de Borst4 H. el Azzouzi5 B. Heijmans2 G. Pasterkamp1,6 H.M. den Ruijter1#
*Authors contributed equally
1 Laboratory of Experimental Cardiology, University Medical Center Utrecht, The Netherlands 2 Department of Molecular Epidemiology, Leiden University Medical Center, Leiden, The Netherlands 3 Laboratory of Pediatric Gastroenterology, Wilhelmina Children’s Hospital, University Medical Center, The Netherlands 4 Department of Vascular Surgery, University Medical Center Utrecht, The Netherlands 5 Department of Cardiology, University Medical Center Utrecht, The Netherlands 6 Laboratory of Clinical Chemistry and Haematology, University Medical Center Utrecht, The Netherlands
20162300 Marten Siemelink.indd 130 15-03-16 13:23 CHAPTER 8
Sex-Specific Differences in DNA Methylation in the Atherosclerotic Carotid Artery
MANUSCRIPT IN PREPARATION
20162300 Marten Siemelink.indd 131 15-03-16 13:23 132 | CHAPTER 8
Abstract
Background Sex-differences in the aetiology of cardiovascular disease and atherosclerosis have been well defined, yet the extent to which a sex-specific epigenetic signature exists within the atherosclerotic vessel is unknown. We present an epigenome-wide association study (EWAS) into sex-specific differentially methylated regions in carotid atherosclerotic plaques of men and women undergoing carotid endarterectomy. In addition, we investigated whether the resulting loci could be involved in differences in cardiovascular disease pathophysiology.
Methods We performed an epigenome-wide association study in carotid plaque specimens of 488 patients (148 women, 340 men) who underwent carotid endarterectomy. DNA was isolated from the plaque specimens, bisulphate converted and used to interrogate DNA methylation of 443,872 CpG dinucleotides by means of the Illumina Infinium HM450 Beadchip Array. We validated the results in 92 blood samples (61 men). Analysis was confined to autosomal chromosomes. DNA methylation was associated with sex using linear modelling corrected for age. Gene-promoter methylation was determined based on local CpGs. Additionally differential gene methylation was associated to measures of cardiovascular disease severity.
Results We identified 311 differentially methylated CpGs between women and men in atherosclerotic plaques. We found the majority of CpGs to overlap with previously reported sex-differentially methylated CpGs in other tissues and with our data in blood. Analysis of promoter methylation identified 4,568 gene-promoter regions that were differentially methylated between the sexes. We found no indication for confounding through risk factors or atherosclerotic plaque characteristics and no association with measures of disease severity.
Conclusion We present an epigenome-wide association study of sex-differences in DNA methylation in the carotid atherosclerotic plaque. We found 311 differentially methylated CpGs and 4,568 differentially methylated promoter regions. We found no evidence for involvement of these genes in cardiovascular disease mechanisms. Our data confirm that DNA methylation profiles are highly sex-specific, even in the severely diseased atherosclerotic vessel wall. Our results suggest that the search for associations between plaque DNA methylation and cardiovascular disease should be executed in a sex-stratified study design.
20162300 Marten Siemelink.indd 132 15-03-16 13:23 SEX DIFFERENCES IN PLAQUE METHYLATION | 133
Introduction
Large differences have been shown between males and females in the regulation of gene expression.1,2 These differences could account for differences in susceptibility to, and pathophysiology of, diseases. Indeed, such sex-differences have been found for methylation marks in leukocytes.3 Sex-differences in the aetiology of cardiovascular disease and atherosclerosis are well defined. Women tend to form more stable atherosclerotic plaques4 that are more prone to erosion instead of classic plaque rupture.5 Women are also less likely to develop classical features of cardiovascular disease such as myocardial infarction and stroke early in life6 and have a favourable prognosis as compared to men.7 However, these sex-differences may not be fully explained by a difference in traditional cardiovascular risk factors.8 Differences in DNA methylation between the sexes have been found within numerous tissues and across different races. Most of the studies looking at epigenome-wide sex- differences were carried out in healthy individuals.9 Recently, a genome-wide atherosclerosis-specific profile of differentially methylated regions was published.10 This study observed an atherosclerosis-specific methylation signature with genes that play a role in endothelial and smooth muscle functions. Yet, sex-specific analyses were not presented. Our aim was to investigate sex-differences in DNA methylation in the severely diseased atherosclerotic vascular wall. We present an epigenome-wide association study (EWAS) into sex-specific differentially methylated regions in carotid atherosclerotic plaques of men and women. We researched the sex-specifically methylated loci and genes for their involvement in the molecular mechanisms that underlie the sex differences in cardiovascular pathophysiology.
Methods 8 Patient inclusion The Athero-Express is an ongoing Biobank cohort study including patients that undergo carotid or femoral artery endarterectomy at the University Medical Center Utrecht (Utrecht, The Netherlands) or the Sint Antonius Ziekenhuis Nieuwegein (Nieuwegein, The Netherlands). A detailed description of the cohort has previously been published.11 Clinical data are extracted from patient medical files and standardized questionnaires. The medical ethics committees of both hospitals approved of the study and written informed consent was obtained from patients. For this particular study, only patients that underwent carotid artery endarterectomy (CEA) patients were included.
Sample collection Blood samples were obtained from the radial artery catheter immediately prior to surgery. Carotid plaque specimens were removed during surgery and immediately processed in the laboratory. Specimens were cut transversely into segments of 5 mm. The culprit lesion was identified, fixed in 4% formaldehyde, embedded in paraffin and processed for histological examination. Remaining segments were stored at -80 degrees.
20162300 Marten Siemelink.indd 133 15-03-16 13:23 134 | CHAPTER 8
Methylation Array DNA was isolated from 506 stored plaque-specimens and 94 patient-matched blood samples. Patient selection was on the basis of plaque, blood and genotype data availability. To exclude selection bias, selected patients were compared with the whole cohort for changes in general characteristics and no significant differences were observed. Isolated DNA was checked for purity and concentration and was equalized to 600ng DNA prior to bisulphate conversion. Bisulphate converted DNA samples were used to measure DNA- methylation by means of the HM450k Methylation Beadchip Array (Illumina, San Diego, USA). Processing of the sample and array was performed according to the manufacturer’s protocol.
Quality Control The raw data from the array was processed using the ‘MethylAid’ R-package.12 Samples with low-median signal intensity, high background signal, incomplete bisulphate conversion or a low success rate (<95%) were removed. Probes with low beadcount (<3), with high detection p-value or with low success rate (<0.95%) over samples as well as ambiguously mapping probes were removed. Normalization and batch effects were corrected for using ‘Functional Normalisation’ with 8 principal components of control probes. Principal components of CpG probes were calculated and used to identify possible mix up of samples by plotting PCs vs sample type (i.e. blood or plaque DNA) and gender. In addition, where available, genotype data from previous studies (using genotyping arrays) was correlated to the raw data of the 65 SNPs included on the HM450k array, and samples with poor correlation (R < 0.6) across these 65 SNPs were excluded due to possible mix up. Quality control showed 488 plaque samples (96%) and 92 blood samples (98%) of good quality for further analysis. During quality control, 41,640 probes were excluded, with 443,872 probes (91.4 %) of good quality remaining. An overview of the sample quality control is depicted in a flow-chart (Supplemental Figure 1). For the current analyses, only probes on the autosomal chromosomes were included as X-chromosomal DNA methylation is subject to X-chromosomal inactivation and requires a different methodological approach.
Analysis Statistical analyses were carried out in R Studio (v0.98.981) using R (v3.2.2). Fast-LMM13 was used to perform epigenome-wide associations, while applying genomic control to correct for residual confounding. Invariable probes (β range 0-0.1 or 0.9-1) as well as outliers (>3 SD) for each probe were removed prior to epigenome-wide analysis. We refrained from adding methylation principal components to the epigenome-wide association model, since the principal components were correlated to the phenotype (sex). Further association of sex-related CpGs with cardiovascular risk factors and histological parameters was performed by linear regression modelling, using age and sex as covariates.
Athero-Express Genomics Study and methylation Quantitative Trait Analysis Common genetic variation has been examined previously in the Athero-Express, and is described elsewhere.14 In short, patient DNA samples were genotyped in two batches, and imputation was performed using the 1000 Genomes Project phased haplotypes as the reference panel. Imputed genotypes were available for 454 of the total of 509 patients. We used SNPTEST15 (version 2.5) for the identification of methylation quantitative
20162300 Marten Siemelink.indd 134 15-03-16 13:23 SEX DIFFERENCES IN PLAQUE METHYLATION | 135
trait loci (meQTL) in a linear regression model corrected for covariates age, genotyping batch and genotyping principal components 1 and 2. Only high-quality imputed variants (MAF ≥ 0·05; imputation quality ≥ 0·9; HWE p value ≥ 1·0x10-6) within 500kb of the CpG were considered.
Determining DNA methylation of the gene promoter For all autosomal DNA-methylation probes, the distance to the transcription start site (TSS) of nearby genes was determined with custom Perl script using the RefSeq gene annotation (July 2015, GRCh37/hg19). To determine the overall methylation status of the gene promoter, we first discarded outlier-probes that were > 0.2 beta outside of the median across all samples and all probes within a window of 1000 bp upstream and 500 bp downstream of the gene TSS: |x ̃ – y ̃| < 0.2 with x = methylation of a single probe y = methylation of all probes in the window We then calculated gene-promoter methylation for each sample as the median of the included probes for each gene.
Gene-centered analysis We performed an epigenome-wide analysis on median promoter methylation of 14,456 genes for sex differences, which resulted in 4,748 differentially methylated gene promoters. For these genes, annotation using enrichment and pathway analysis was performed by means of the ToppFun suite.16 Significance was determined by an FDR- corrected p-value < 0.05.
Results
Population characteristics DNA-methylation was determined in carotid plaque samples of 488 patients (70% male) and in blood samples of 92 patients (66% male), of which samples of 85 patients were 8 overlapping. Cohort baseline characteristics can be found in Table 1.
Association of CpG methylation with sex An epigenome-wide association study (EWAS) with covariate age was performed on plaque methylation of 340 men and 148 women to identify CpGs that were differentially methylated between the sexes. After applying genomic control to correct for inflation, 311 of 443,872 loci were significantly associated with sex (Figure 1, Figure 2, Supplemental Table 1, Supplemental Figure 3). For both sexes, there were CpG loci that associated with increased DNA methylation, yet male sex was most often associated with increased DNA methylation (Figure 3). The majority of the 311 CpGs showed similarity in sex-specific methylation between tissues (Supplemental Figure 6). The top 20 differentially methylated CpGs are listed in Table 2. Remarkably, all of the top 20 CpGs were previously reported in other tissues and showed overlap with our data in blood (Table 2, Supplemental Figure 4 and 5).
20162300 Marten Siemelink.indd 135 15-03-16 13:23 136 | CHAPTER 8
Table 1. General cohort characteristics Characteristics Male Female p value (N = 340) (N = 148) Age years 68 (61-74) 69 (62-74) 0.809 SBP mmHg 153 (135-170) 155 (140-170) 0.239 DBP mmHg 81 (73-90) 80 (75-90) 0.907 eGFR ml/min/1.73m2 74 (61-86) 68 (55-85) 0.051 BMI kg/m2 26.1 (24.3-28.4) 26.1 (23.8-28.5) 0.519 hsCRP mg/l 1.8 (1.1-4.1) 2.8 (1.5-7.9) 0.017 Diabetes yes 82 (24.1) 28 (18.9) 0.252 Hypertension yes 247 (75.3) 115 (78.8) 0.483 Statins yes 257 (75.5) 112 (75.7) 0.999 Smoking yes 130 (38.5) 64 (44.4) 0.271 Symptoms 0.986 Asymptomatic 56 (16.5) 24 (16.3) TIA 152 (44.7) 65 (44.2) Stroke 89 (26.2) 36 (24.5) Retinal 43 (12.6) 22 (16.0) Patient characteristics at time of inclusion stratified by patient sex, excluding patients without data on sex (N = 15). Continuous variables are shown as medians with interquartile ranges. Categorical variables are shown as number with percentage. Symptoms refer to symptoms at presentation, be- fore carotid endarterectomy. Significance shown as p values without FDR adjustment. Abbreviations: SBP, systolic blood pressure; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate by MDRD-formula; BMI, body-mass index; LLDs, use of lipid-lowering drugs; Retinal, retinal infarc- tion and amaurosis fugax.
Association of sex-differentially methylated CpGs with cardiovascular parameters Conceivably, these 311 CpGs may be confounded by cardiovascular risk factors which may differ between the sexes. However, none of these CpGs were found to associate with cardiovascular risk factors (Supplemental Table 3). A sex-difference in CpGs may also be due to residual confounding due to cell-type heterogeneity in the plaque. Yet we were unable to show an association of the 311 significant CpGs with histological plaque parameters (Supplemental Table 4). Finally, DNA methylation of CpGs may be influenced by sex-independent single nucleotide variants (SNVs), which may argue against tight control of DNA methylation determined by sex. Yet, assessment of this using methylation quantitative trait analysis (meQTL) showed only in a small amount of CpGs (20, 6.4%) an association between CpGs and local SNVs (Supplemental Table 5).
Association of gene promoter methylation with sex The best evidence for an effect of CpGs on gene expression has been found in CpGs in promoter regions of genes. To grasp the biological meaning of the found associations in CpGs, we therefore calculated a composite measure of promoter methylation. We determined median DNA methylation within the promoter for all genes in each individual
20162300 Marten Siemelink.indd 136 15-03-16 13:23 SEX DIFFERENCES IN PLAQUE METHYLATION | 137
patient. Analysis of DNA methylation in gene promoter areas showed relatively minor variability between multiple CpGs in the same promoter (Supplemental Figure 2). We associated promoter methylation of each gene to the patient’s sex, which yielded 4,568 gene promoters that showed sex-specific promoter methylation in atherosclerotic plaque tissue (Supplemental Table 7). We validated the majority of the 4,568 gene promoters in blood (Supplemental Table 7). To investigate whether the differences could be hormone-driven, we annotated all genes with estrogen receptor binding site motifs. The gene promoters of the sex-differentially methylated genes showed no enrichment for such motifs, arguing against estrogen-driven DNA methylation (Supplemental Figure 7).
Association of sex-differentially methylated gene promoters with cardiovascular parameters To further interrogate these genes, we looked for association of their promoter methylation with cardiovascular risk factors and plaque composition, independently of age or sex, which did not show significant associations (Supplemental Table 8, 9). Finally, we annotated the genes with sex-differentially methylated promoters using a variety of enrichment and pathway analysis tools united in the ToppFun suite, which showed no atherosclerosis-specific annotations (Supplemental Table 11).
8
Figure 1. Manhattan-plot of differentially methylated CpGs between the sexes in 488 plaque sam- ples. Vertical axis shows significance of the observed effect, while the chromosomal position is shown on the horizontal axis. Sex-chromosomes where excluded from analysis. Points in red denote significant CpGs.
20162300 Marten Siemelink.indd 137 15-03-16 13:23 138 | CHAPTER 8 Replication PFC;LEU;PAN;WB PFC;LEU;PAN;WB COL;PFC;LEU;PAN;WB PFC;PAN;WB PFC;LEU;PAN;WB PFC;LEU;PAN;WB PAN;WB PFC;LEU;PAN;WB PFC;LEU;PAN;WB PFC;LEU;WB PFC;LEU;PAN;WB PFC;PAN;WB PFC;LEU;WB COL;PFC;LEU;WB PFC;LEU;WB PFC;LEU;PAN;WB PFC;LEU;WB PFC;LEU;WB PFC;LEU;PAN;WB PFC;LEU;WB -35 -35 -32 -30 -27 -24 -22 -21 -19 -19 -19 -19 -17 -17 -17 -16 -16 -16 -16 -15 1,77 1,77 2,21 4,52 2,14 3,33 3,96 1,16 3,40 1,70 3,72 3,09 1,13 2,21 5,49 6,53 3,54 4,83 2,46 5,13 FDR p value Relation to Island Island Open water Island Island North Shore Open water Island Island Open water Open water Open Sea North Shelf North Shore Open water Island Island Island Island Open water Island Relation to gene Body TSS200 5’UTR;1stExon TSS200 Body Body TSS1500 Body Body Body Body 1stExon;5’UTR 1stExon;5’UTR TSS1500 Body 1stExon Related Gene Intergenic KRT77 AMDHD2 CSNK1D Intergenic Intergenic ZNF69 RNASEH2C RFTN1 Intergenic ARID1B FOXN3 TFDP1 RBM39 LOC644649 GPX1 NICN1 TLE1 RFTN1 HAX1 SE 0,008 0,008 0,012 0,013 0,013 0,013 0,013 0,008 0,009 0,010 0,011 0,011 0,012 0,012 0,012 0,013 0,013 0,013 0,011 0,013 beta 0,426 0,421 0,390 0,388 0,378 0,376 0,408 -0,426 -0,378 -0,365 -0,360 -0,360 -0,355 -0,416 -0,376 -0,366 -0,365 -0,361 -0,397 -0,360 Position 2570283 53085038 65487814 89878619 34319899 59789030 80231263 59318136 23489940 11998623 16411667 49395714 49466685 84303915 16428391 243053673 114292172 160865097 157098338 154245232 1 3 1 6 3 3 9 3 1 12 11 14 13 16 20 16 17 11 13 19 Chr cg12691488 CpG cg03691818 cg25294185 cg11643285 cg03618918 cg25568337 cg02325951 cg26355737 cg26921482 cg08906898 cg04946709 cg02758552 cg23814743 cg22345911 cg07852945 cg17238319 cg17232883 cg06710937 cg12177922 cg03608000 Top 20 of autosomal CpGs that differ significantly between women and men 2. Top Table Table showing the top-20 sex-differentially methylated CpGs in carotid plaque tissue. Replication refers to previously reported significantly Table pancreatic islet cells; COL, colon; WB, methylation in other tissues using the HM450k chip. Abbreviations: PFC, prefrontal cortex; LEU, leukocytes; PAN, whole blood from the Athero-Express cohort.
20162300 Marten Siemelink.indd 138 15-03-16 13:23 SEX DIFFERENCES IN PLAQUE METHYLATION | 139
Figure 2. Quantile-Quantile plot comparing observed and expected p-values for the difference in DNA-methylation between the sexes. Each point represents a single CpG. P values have been cor- rected for inflation using genomic control.
Discussion
We present an epigenome-wide association study into sex-specific differentially methylated regions on autosomal chromosomes in vascular tissue of a severely atherosclerotic population. We found many differences between the sexes, from an individual CpG as well as from a gene promoter methylation point of view, most of which we were able to validate 8 in blood samples. It appeared that these sex-differences were neither tissue-specific nor implicated in cardiovascular disease mechanisms. This suggests that sex has a constitutive effect on DNA methylation at these CpGs independent of tissue or cardiovascular disease. Sex-differences in DNA methylation have been reported before in healthy individuals in a wide variety of tissues.9 Interestingly, we replicated many CpGs that have previously been reported to differ between the sexes in prefrontal cortex, peripheral blood cells, intestinal cells and pancreatic islet cells.17–20 Furthermore, we were able to validate most CpGs and promoters in blood of the same patients. We are therefore able to show that the same sex-differences in methylation patterns as previously reported, are also found within severely diseased tissue, namely the atherosclerotic vessel wall. This could indicate the fact that sex-differences are much larger than the effect of the disease on modification of DNA methylation. Indeed, we found no association of DNA methylation at CpGs or gene promoters, with measures of disease severity. Secondly, our DNA methylation measure is one of a heterogeneous mixture of cells in the atherosclerotic plaque. As DNA methylation is strongly correlated to cell type, this could point towards a sex-effect that is even stronger than the effect of cell-type on DNA methylation. This finding is supported by the fact that
20162300 Marten Siemelink.indd 139 15-03-16 13:23 140 | CHAPTER 8
all previous reports of studies that use the same Illumina HM450k methylation array report the same top hits when comparing male and female methylation. Furthermore, we did not find any association between the differentially methylated CpGs or promoters with plaque characteristics and, we found no evidence for a relation between hormone regulation of DNA methylation, as estrogen-specific binding sites were not overrepresented near our significant results. One could speculate about the existence of innate sex-differentially methylated regions. A certain methylation profile, irrespective of tissue type or disease status, may thus in itself be informative as a proxy for sex. Whether the novel sex- associated CpGs that we found, are plaque- or disease specific or merely reflect a gain of statistical power remains to be proven.
Figure 3. Volcano plot Volcano plot showing the relationship between p-value and logarithmic fold-change in DNA-methyl- ation between the sexes. The horizontal axis shows beta value for the effect of male sex on DNA- methylation at a CpG in this gene. The vertical axis shows the FDR-corrected –log10(p value) of the association with sex. Points in red denote significant CpGs. Gene annotation is provided for selected probes, plus signs indicate intergenic probes.
20162300 Marten Siemelink.indd 140 15-03-16 13:23 SEX DIFFERENCES IN PLAQUE METHYLATION | 141
We investigated the possible biological relevance of the differentially methylated CpGs and genes for cardiovascular disease. We found no association of the observed sex- differentially methylated CpGs and promoters between more severely diseased patients (presenting with stroke versus TIA or asymptomatic or absence or presence of contralateral carotid stenosis). Furthermore, enrichment and annotation analysis revealed no specific cardiovascular disease mechanisms of the sex-differentially methylated CpGs and promoters. This is not surprising, given the fact that sex-differences seem to outweigh both the effects of cell type and disease phenotype with respect to DNA methylation. Still, the innate differences could in itself be the starting point for different biological mechanisms that contribute to cardiovascular disease, such as lipid accumulation, inflammatory response or coagulation in female and male cells that reside in the vascular wall. As such, they could still account for differences in cardiovascular disease phenotypes as observed in atherosclerotic plaques. However, these subtle differences cannot be found within the large number of sex- differentially methylated CpGs and promoters. We can only speculate as to the biological effects of the found differences. DNA methylation of CpGs in a gene promoter region is well linked to gene silencing, yet recent research in the field of epigenetics revealed more complex interplay between different regulatory mechanisms for other CpGs sites.21 Unequivocally, differences in regulation of gene expression add to the phenotypic differences between the sexes in mammals.1,2 This study has several limitations. First, due to limited sample size and stringent correction for possible residual confounding using genomic control, we may suffer from a substantial false-negative discovery rate. Thus, there may be many CpGs affected by sex which go undetected in our study. Future studies may solve this through increased statistical power in larger cohorts. Another mechanism that might have led to more significant gene promoters is the fact that we calculated the median of the probes. This leads to exclusion of random noise and outliers. Such a calculation may thus result in less noise in the promoter methylation variable as compared to the raw CpGs. Second, due to the heterogeneous cell types in the plaque we cannot ascertain in which cell types the sex- associated differences arise. Neither can we infer the biological meaning of the related 8 genes in the context of atherosclerotic disease. One of the remaining questions is whether the found methylation differences that are plaque-specific are the cause or the effect of the disease phenotype or the influence of sex. However, we replicated many loci that have been reported previously in different tissues and we found no association with cardiovascular disease risk factors or disease severity measures strengthening the current interpretation of the results. Third, we were unable to confirm the role of methylation in gene expression due to the absence of expression data in the Athero-Express Biobank. It could very well be the case that many of the CpGs we found are affecting the differences between women and men through regulatory mechanisms. Furthermore, we determined promoter methylation of genes based on a non-empirically determined window. This may lead to inaccuracy and as a result, pathway analysis might suffer from a measure of bias. This may have led to an underestimate of the number of gene promoters associated with sex- differences. However, we do not believe the absence of the mechanism of action of the CpGs or promoters makes the differences in itself less true or interesting. Fourth, we chose to exclude the X-chromosome in our study as the imprinted X-chromosome leads to complicated analysis that is not comparable to the autosomal data as it reflects another
20162300 Marten Siemelink.indd 141 15-03-16 13:23 142 | CHAPTER 8
sex-specific biological mechanism. However, we recognize the fact that differential methylation of the (imprinted) X-chromosome can contribute to sex-differences in the atherosclerotic vessel wall. We present the first epigenome-wide association study into sex-differences on autosomal methylation in the atherosclerotic plaque. We found 311 differentially methylated CpGs and 4,568 differentially methylated gene promoters. Our data confirm that DNA methylation profiles are highly sex-specific, in both the severely diseased atherosclerotic vessel wall as well as peripheral blood samples. Our results suggest that the search for associations between plaque DNA methylation and plaque or patient characteristics should take sex into account, preferably by a sex-stratified study design. Further research is needed to investigate the sex-differentially methylated CpGs and genes to determine their role in the molecular mechanisms of sex-differences in (cardiovascular) disease.
20162300 Marten Siemelink.indd 142 15-03-16 13:23 SEX DIFFERENCES IN PLAQUE METHYLATION | 143
References
1 Parsch J, Ellegren H. The evolutionary causes and consequences of sex-biased gene expression. Nat Rev Genet 2013; 14: 83–7. 2 Blekhman R, Marioni JC, Zumbo P, Stephens M, Gilad Y. Sex-specific and lineage-specific alternative splicing in primates. Genome Res 2010; 20: 180–9. 3 Talens RP, Jukema JW, Trompet S, et al. Hypermethylation at loci sensitive to the prenatal environ- ment is associated with increased incidence of myocardial infarction. Int J Epidemiol 2012; 41: 106– 15. 4 Vrijenhoek JEP, Den Ruijter HM, De Borst GJ, et al. Sex Is Associated With the Presence of Athero- sclerotic Plaque Hemorrhage and Modifies the Relation Between Plaque Hemorrhage and Cardiovas- cular Outcome. Stroke 2013; 44: 3318–23. 5 Burke AP, Farb A, Malcom GT, Liang Y, Smialek J, Virmani R. Effect of risk factors on the mechanism of acute thrombosis and sudden coronary death in women. Circulation 1998; 97: 2110–6. 6 Go AS, Mozaffarian D, Roger VL, et al. Heart disease and stroke statistics--2013 update: a report from the American Heart Association. Circulation 2013; 127: e6–245. 7 van der Meer MG, Cramer MJ, van der Graaf Y, Doevendans PA, Nathoe HM. Gender difference in long-term prognosis among patients with cardiovascular disease. Eur J Prev Cardiol 2012; 21: 81–9. 8 Hellings WE, Pasterkamp G, Verhoeven BAN, et al. Gender-associated differences in plaque pheno- type of patients undergoing carotid endarterectomy. J Vasc Surg 2007; 45: 289–96. 9 McCarthy NS, Melton PE, Cadby G, et al. Meta-analysis of human methylation data for evidence of sex-specific autosomal patterns. BMC Genomics 2014; 15: 981. 10 Zaina S, Heyn H, Carmona FJ, et al. DNA methylation map of human atherosclerosis. Circ Cardiovasc Genet 2014; 7: 692–700. 11 Verhoeven BAN, Velema E, Schoneveld AH, et al. Athero-express: differential atherosclerotic plaque expression of mRNA and protein in relation to cardiovascular events and patient characteristics. Ra- tionale and design. Eur J Epidemiol 2004; 19: 1127–33. 12 van Iterson M, Tobi EW, Slieker RC, et al. MethylAid: visual and interactive quality control of large Il- lumina 450k datasets. Bioinformatics 2014; 30: 3435–7. 13 Lippert C, Listgarten J, Liu Y, Kadie CM, Davidson RI, Heckerman D. FaST linear mixed models for genome-wide association studies. Nat Methods 2011; 8: 833–5. 14 van der Laan SW, Asl HF, van den Borne P, et al. Variants in ALOX5, ALOX5AP and LTA4H are not asso- ciated with atherosclerotic plaque phenotypes: The Athero-Express Genomics Study. Atherosclerosis 2015; 239: 528–38. 15 Marchini J, Howie B, Myers S, McVean G, Donnelly P. A new multipoint method for genome-wide association studies by imputation of genotypes. Nat Genet 2007; 39: 906–13. 16 Chen J, Bardes EE, Aronow BJ, Jegga AG. ToppGene Suite for gene list enrichment analysis and candidate gene prioritization. Nucleic Acids Res 2009; 37: W305–11. 17 Kaz AM, Wong C-J, Dzieciatkowski S, Luo Y, Schoen RE, Grady WM. Patterns of DNA methylation in 8 the normal colon vary by anatomical location, gender, and age. Epigenetics 2014; 9: 492–502. 18 Inoshita M, Numata S, Tajima A, et al. Sex differences of leukocytes DNA methylation adjusted for estimated cellular proportions. Biol Sex Differ 2015; 6: 11. 19 Xu H, Wang F, Liu Y, Yu Y, Gelernter J, Zhang H. Sex-biased methylome and transcriptome in human prefrontal cortex. Hum Mol Genet 2014; 23: 1260–70. 20 Hall E, Volkov P, Dayeh T, et al. Sex differences in the genome-wide DNA methylation pattern and im- pact on gene expression, microRNA levels and insulin secretion in human pancreatic islets. Genome Biol 2014; 15: 522. 21 Rothbart SB, Strahl BD. Interpreting the language of histone and DNA modifications. Biochim Biophys Acta 2014; 1839: 627–43.
20162300 Marten Siemelink.indd 143 15-03-16 13:23 144 | CHAPTER 8
Supplemental Material
Supplement to: Siemelink, MA et al, Sex-specific differences in DNA methylation in the atherosclerotic carotid artery. Supplemental material available on publication and on request.
Table of Contents Supplemental Tables Supplemental Table 1. Epigenome-wide analysis in plaque samples Supplemental Table 2. The effect of Sex on DNA methylation in different tissues Supplemental Table 3. Association of CpGs with cardiovascular risk factors Supplemental Table 4. Association of CpGs with carotid plaque histology Supplemental Table 5. meQTL analysis of sex associated CpGs in plaque and blood Supplemental Table 6. Epigenome-wide analysis of gene promoters in plaque samples Supplemental Table 7. Association of genes promoters with cardiovascular risk factors Supplemental Table 8. Association of genes promoters with carotid plaque histology Supplemental Table 9. ToppFun suite analysis
Supplemental Figures Supplemental Figure 1. Flow-chart of quality control Supplemental Figure 2. Variability of DNA methylation in relation to the transcription start site Supplemental Figure 3. Scatterplots comparing DNA methylation in men and women Supplemental Figure 4. Manhattanplot of EWAS on sex differences in blood samples Supplemental Figure 5. Qqplot of EWAS on sex differences in blood samples Supplemental Figure 6. Scatterplots comparing DNA methylation across tissues Supplemental Figure 7. Scatterplot comparing number of estrogen binding motifs and sex-association of gene promoters
20162300 Marten Siemelink.indd 144 15-03-16 13:23 | 145
20162300 Marten Siemelink.indd 145 15-03-16 13:23 M.A. Siemelink1 H.M. Ruijter1 F. van der Valk2 J.P.P.M. de Vries3 G.J. de Borst4 F. Moll4 E.S.G. Stroes2 G. Pasterkamp1,5
1 Utrecht University Medical Center, Laboratory of Experimental Cardiology, Utrecht, the Netherlands 2 Department of Vascular Medicine, Academic Medical Center, Amsterdam, the Netherlands 3 Department of Vascular Surgery, St. Antonius Hospital, Nieuwegein, the Netherlands 4 Department of Vascular Surgery, University Medical Center Utrecht, Utrecht, the Netherlands 5 Reseach Laboratory Clinical Chemistry, University Medical Center Utrecht, Utrecht, the Netherlands
20162300 Marten Siemelink.indd 146 15-03-16 13:23 CHAPTER 9
Systemic Glucocorticoids are associated with Mortality Following Carotid Endarterectomy
MANUSCRIPT IN PREPARATION
20162300 Marten Siemelink.indd 147 15-03-16 13:23 148 | CHAPTER 9
Abstract
Glucocorticoids (GCs) are widely used anti-inflammatory drugs, well known to cause many adverse effects. Still, there is a dearth of data on the long-term cardiovascular effects of GCs in patients with established cardiovascular disease and the effect on atherosclerotic plaque composition. 1894 patients that underwent carotid endarterectomy, of whom 40 patients received systemic GCs, were included in the Athero-Express biobank. Atherosclerotic plaque samples as well as peripheral blood samples were obtained during carotid endarterectomy. Cardiovascular events during 3 years of follow-up were investigated using Cox regression modeling to adjust for possible confounding. Atherosclerotic plaque composition was examined using immunohistochemical staining. Use of GCs at inclusion was associated with markedly increased incidences of ischemic stroke (15.2% vs. 5.9%), composite events (48.5% vs. 26.9%) and cardiovascular death (21.2% vs. 5.7%) as well as an increased risk of cardiovascular death (HR 2.7, 95% CI 1.1 - 6.7) and all- cause death (HR 2.3, 95% CI 1.1-4.8) after 2.6 years of follow-up. None of the histological features of atherosclerotic plaques were significantly different in patients using GCs. Following carotid endarterectomy, the use of systemic glucocorticoids is independently associated with an increased incidence of cardiovascular events and an increased risk for cardiovascular- and all-cause death, but not atherosclerotic plaque composition.
20162300 Marten Siemelink.indd 148 15-03-16 13:23 GLUCOCORTICOID RISK AFTER CAROTID ENDARTERECTOMY | 149
Introduction
Lipid metabolism and inflammation are key determinants driving initiation and development of atherosclerotic cardiovascular disease (CVD). At present, lipid modulating strategies are the cornerstone of CVD management, reaching 33% and 22% risk reductions for primary and secondary prevention of major cardiovascular events, respectively.1,2 Yet, in the context of a substantial residual cardiovascular risk, drug development tracks are currently focusing on the use of anti-inflammatory compounds on top of lipid lowering strategies to further reduce CVD risk.3 Glucocorticoids (GCs), like prednisone and dexamethasone, are well-characterized and potent anti-inflammatory drugs widely prescribed for inflammatory diseases. Nonetheless, hesitations exist in translating GCs’ anti-inflammatory efficacy to CVD patients. The chronicity of atherogenesis warrants long-term systemic GCs use which may be accompanied by significant cardiometabolic side-effects including hypertension, increased insulin resistance, weight gain and dyslipidemia.4,5 The combination of these side-effects and GCs’ non-cardiometabolic complications, for example osteoporosis, cataract and risk of infection, may outweigh GCs’ beneficial anti-inflammatory effects in CVD patients.6 In fact, a large population-study and a case-control study both reported a dose-dependent increased risk of CVD.7,8 This fuels the hypothesis that GCs may have an adverse effect on atherogenesis, eventually resulting in an increased CVD risk. In support, patients who undergo carotid endarterectomy (CEA) due to symptomatic carotid artery stenosis are at an increased risk for stroke within 30 days following CEA when exposed to GCs prior to surgery.9 Nonetheless, solid evidence is lacking on GCs’ effects on the longer term in CEA patients. In the present study, we investigated the cardiovascular risk of GC use in patients that underwent CEA. To investigate the mechanisms by which GCs may affect atherogenesis and cardiovascular risk, we evaluated both systemic cardiometabolic parameters as well as carotid plaque composition in the CEA specimens.
Methods
Patient inclusion and selection This study adheres to the principles outlined in the declaration of Helsinki for human studies 9 and the study protocol was approved by the ethics committees of both participating hospitals. Patients that participated in the study provided written informed consent as reviewed and approved by the ethics committees. The Athero-Express cohort and associated biobank is a large ongoing longitudinal study including patients that underwent carotid endarterectomy due to severe atherosclerotic carotid artery stenosis, of which a detailed description of study design has been previously published.10 We performed analysis on 1894 patients included in the study that underwent carotid endarterectomy at the University Medical Center Utrecht or the Sint Antonius Hospital Nieuwegein (The Netherlands). Clinical patient data was obtained through standardized questionnaires, pre-operative admission charts and patient medical files. The indication for- and use of systemic GCs was determined at time of surgery, based on analysis of the electronic patient file. Medications included were prednisone, prednisolon, beclometasone and budesonide. Patients without data on medication use were excluded.
20162300 Marten Siemelink.indd 149 15-03-16 13:23 150 | CHAPTER 9
In addition, patients using only topical- or inhalation GCs were excluded. The prevalence of comorbidities that are major indications for GC treatment were determined based on the patient medical file. Pulmonary disease and was determined by medication use for chronic obstructive pulmonary disease (COPD) while rheumatoid disease was based on anti-rheumatic medication use. The prevalence of kidney disease was based on the Kidney Disease Outcomes Quality Initiative (KDOQI) classification.
Sample collection Atherosclerotic plaque specimens were collected during carotid endarterectomy procedure, cut in sections and subsequently processed. The plaque section containing the culprit lesion was identified, paraffin-embedded and used for microscopic plaque histology. Adjacent segments to the culprit lesion served for plaque protein isolation.
Carotid Plaque Histology Atherosclerotic carotid plaque samples were obtained during CEA according to a standardized protocol. In short, plaque specimens were cut in 5mm segments, the segment containing the culprit lesion was identified, paraffin-embedded and used for microscopic plaque histology. 10 micron cross-sections of the paraffin-embedded plaque specimens were cut using a microtome, and examined using a microscope. Microscopy-slides were stained with hematoxylin and eosin (H&E) for assessment of calcifications, atheroma and fibrous tissue content. Furthermore, staining with Elastin von Gieson (EVG) was used for plaque hemorrhage and Picrosirius Red for assessment of collagen. Immunohistochemical staining was performed for assessment of macrophages (CD68) and smooth-muscle cells (alpha-actin). The relative abundance of atheromatous and fibrous tissue was used to determine the generalized plaque phenotype. The presence of calcifications, collagen and atheroma were classified as scarce/absent or high. Plaque hemorrhage was classified as present or absent. Plaque microvessels were quantitatively assessed as number of individual vessels per microscopy field. Plaque smooth-muscle cells and macrophages were quantitatively assessed as percentage of the microscopy field area. For quantitative measures, results of multiple random microscopy fields were averaged.
Follow-up Clinical follow-up procedure has been previously described in more detail.10 Patients included in the Athero-Express were assessed using questionnaires at periods of 1, 2 and 3-years post CEA procedure. The occurrence of all endpoints was ascertained by telephone interview of the patient or general practitioner, and by investigation of the electronic patient file.
Statistics Statistical analyses were performed using SPSS 20 (IBM, Armonk, NY, USA). Statistical associations of cohort categorical characteristics and incidence of endpoints were determined using Pearson χ2 test or Fisher’s exact test where appropriate. Continues cohort characteristics were assessed using Mann-Whitney U test. A p-value below 0.05 was considered statistically significant. Cox proportional hazard models were used to analyze cardiovascular outcomes, using three models 1) a univariate model of GC use 2) a multivariate model of GCs, age and gender 3) a multivariate model similar to the second model with addition of Body-Mass Index (BMI), beta-blocker, presenting symptoms and
20162300 Marten Siemelink.indd 150 15-03-16 13:23 GLUCOCORTICOID RISK AFTER CAROTID ENDARTERECTOMY | 151
comorbidities. Survival curves were made using SPSS. The bargraph was made using RStudio v0.98.981 (RStudio, MA, USA) with R Statistics v3.1.1.
Results
Cohort characteristics Of the 1894 patients who underwent carotid Table 1. Indications for glucocorticoids endarterectomy, 40 patients (2.1 %) were Indication N found to use oral GCs. Indications for the use COPD 19 of GCs showed that the majority of patients Rheumatic disease 10 suffered from COPD or rheumatoid disease, with a minority of indications for other causes Kidney transplant 2 (Table 1). General parameters of patients Other 9 either using- or not using GCs, were Indications for glucocorticoid treatment and investigated and showed differences in body- corresponding number of patients. Rheuma- mass index (BMI) and beta-blocker use. toid disease includes patients suffering from rheumatoid arthritis or polymialgia rheumat- Furthermore, there was a significant increase ica. Other less prevalent indications include in the prevalence of comorbidities associated ulcerative colitis, idiopathic thrombocyto- with GC use: COPD, rheumatoid disease and penic purpura, arthritis psoriatica, lupus-like kidney disease. No significant differences in syndrome, prostate cancer, Crohn’s disease. prior history of cardiovascular disease (Table 2) or distribution of presenting symptoms were found (Figure 1).
5 0 L egend p = 0.87 No GCs GCs 4 0 p = 0.21 3 0 9
2 0 p = 0.51
e r c n t a g o f p i s p = 1.00 P 1 0 0
asymptomatic TIA stroke retinal
Figure 1. Symptoms at presentation Bargraph showing the percentages of patients with symptoms at presentation. Asymptomatic, ab- sence of symptoms at presentation (N = 265); TIA, transient ischemic attack (N = 830); stroke, hemorrhagic- or ischemic stroke (N = 495); retinal, retinal infarction or amaurosis fugax (N = 292).
20162300 Marten Siemelink.indd 151 15-03-16 13:23 152 | CHAPTER 9
Table 2. Patient characteristics at time of inclusion Characteristics GCs no GCs GCs no GCs N N P-value Study population 40 1854 N/A Age (years) 40 1852 med (sd.) 71 (8.1) 69 (9.3) 0.23 Gender (male) 40 1854 % 83 68 0.06 Systolic bloodpressure (mmHg) 35 1558 med (sd.) 154 (26) 150 (26) 0.96 Diastolic bloodpressure 35 1557 med (sd.) 80 (12) 80 (13) 0.17 (mmHg) BMI (kg/m2) 38 1736 med (sd.) 24.2 (4.0) 26.0 (4.0) 0.014 Creatinin (umol/L) 39 1789 med (sd.) 103 (46) 89 (47) 0.09 Total cholesterol (mmol/L) 27 1174 med (sd.) 4.9 (0.73) 4.6 (1.2) 0.24 HDL (mmol/L) 25 1111 med (sd.) 1.1 (0.66) 1.1 (2.2) 0.69 LDL (mmol/L) 25 1045 med (sd.) 2.6 (0.41) 2.7 (0.54) 0.52 Triglycerides (mmol/L) 25 1121 med (sd.) 1.8 (1.6) 1.4 (1.0) 0.38 C-reactive protein (mg/L) 19 654 med (sd.) 1.7 (27) 3.0 (15) 0.65 Smoking 38 1782 % 37 24 0.09 Medication Statins 40 1854 % 70 77 0.35 Anti-hypertensiva 40 1854 % 73 76 0.71 Calcium-channel blocker 39 1853 % 23 24 0.86 Diuretic 40 1854 % 40 35 0.62 Beta-blocker 40 1853 % 25 45 0.015 ACE-inhibitor 40 1852 % 38 31 0.49 AT-II receptor antagonist 40 1850 % 20 22 0.85 RAAS medication 40 1852 % 58 51 0.43 Anti-coagulant 40 1854 % 18 12 0.32 Anti-platelet 40 1852 % 95 89 0.31 Insulin 39 1853 % 8 7 1.00 Oral glucose inhibitor 40 1854 % 10 16 0.39 History Diabetes (type I or II) 40 1854 % 18 23 0.57 Hypertension 40 1854 % 75 85 0.11 Coronary artery disease 40 1851 % 33 31 1.00 Myocardial infarction 39 1820 % 21 20 1.00 Coronary intervention 40 1845 % 25 22 0.70 Cerebrovascular accident 40 1854 % 40 33 0.39 CVA/TIA 40 1854 % 93 81 0.10 Peripheral artery disease 40 1851 % 23 21 0.85 Peripheral intervention 40 1847 % 23 20 0.69 Femoral intervention 40 1847 % 13 11 1.00 Pulmonary disease 40 1853 % 60 11 <0.001 Kidney disease 37 1786 % 46 27 0.014 Rheumatoid disease 40 1854 % 33 1 <0.001
20162300 Marten Siemelink.indd 152 15-03-16 13:23 GLUCOCORTICOID RISK AFTER CAROTID ENDARTERECTOMY | 153
Table 2. Continued Characteristics GCs no GCs GCs no GCs N N P-value Presenting Symptoms 40 1842 Asymptomatic 5 260 % 13 14 1.00 TIA 17 813 % 43 44 0.87 Stroke 14 481 % 35 26 0.21 Retinal 4 288 % 10 16 0.51 Comparison of patient characteristics between patient-groups at time of inclusion. All percentages denote percentages within group. Abbreviations: GCs, patients using glucocorticoids; No GCs, patients not using glucocorticoids; N, number; med (sd.), median (standard deviation); %, percentage within group; BMI, body-mass index; HDL, high-density lipoprotein; LDL, low-density lipoprotein; ACE, Angiotensin-convert- ing enzyme; AT-II, Angiotensin-II; RAAS, Renin-angiotensin-aldosterone system; CVA, cerebrovascular ac- cident; TIA, transient ischemic attack; Retinal, retinal infarction or amaurosis fugax.
GCs and cardiovascular risk The chance of restenosis at 1 year post-CEA was determined by duplex-echo of the ipsilateral carotid artery, which showed that patients receiving GC treatment had a lower chance of restenosis compared to those not receiving GCs (9.1% vs. 16.4%). Despite a lower chance of recurrence, analysis of cardiovascular events after 2.6 years of follow-up showed a significant increase of ischemic stroke (15.2% vs. 5.9%, p=0.047), cardiovascular death (21.2% vs. 5.7%, p=0.003) and all-cause death (33.3% vs. 11.4%, p<0.001) as well as composite events (48.5% vs. 26.9%, p=0.009), in patients using GCs versus those who were not reported to use GCs. Furthermore, with the exception of peripheral events, there is a clear trend towards increased incidence of cardiovascular events for all endpoints in patients using GCs (Table 3). Based on Cox proportional hazard modeling, we show that even after correction for confounders (age, gender, beta-blocker, body-mass index, COPD, rheumatoid disease, kidney disease and presenting symptoms) GCs are associated with a significantly increased risk of cardiovascular death (HR 2.7, 95% CI 1.1 to 6.7) and all-cause death (HR 2.3, 95% CI 1.1 to 4.8) within the median 2.6 years of follow-up (Table 4, Figure 2). To reduce possible confounding by indication, we repeated hazard modeling in the subset of 267 patients with the most prevalent 9 indications for GCs: rheumatoid disease and COPD. In this subset, we compared patients using GCs with patients not using GCs. Patients using GCs were found to be at significantly increased risks for all-cause death (HR 2.3, 95% CI 1.0 to 5.5) and composite cardiovascular events (HR 2.1, 95% CI 1.1 to 4.1) while a significant difference in cardiovascular death could not be proven (HR 2.3, 95% CI 0.7-5.5) (Table 4). A higher number of patients were lost to follow-up in the group using GCs (17.5%) compared to those not using GCs (12.9%).
GCs and plaque phenotype To investigate whether the use of GCs was correlated to a more vulnerable plaque phenotype, histological plaque specimens were compared between patient groups by microscopic examination. The generalized plaque phenotypes were classified as either atheromatous, fibrous or atherofibroumatous. In addition, the occurrence of specific
20162300 Marten Siemelink.indd 153 15-03-16 13:23 154 | CHAPTER 9
histological plaque features was determined (atheroma, smooth-muscle cells, macrophages, calcification, collagen, plaque- hemorrhage and microvessels). Neither the generalized phenotype nor any of the specific features associated with the use of GCs (Table 5).
Figure 2. Survival Survival functions for cardiovascular death of patients using glucocorticoids compared to patients not using glucocorticoids. A) Kaplan-Meier curve. B-D) Cox proportional hazard models. B) Model 1: glucocorticoids. C) Model 2: glucocorticoids, age, gender. D) Model 3: glucocortocoids, age, gender, BMI, beta-blocker, comorbidities, presenting symptoms.
20162300 Marten Siemelink.indd 154 15-03-16 13:23 GLUCOCORTICOID RISK AFTER CAROTID ENDARTERECTOMY | 155
Table 3. Incidence of endpoints Endpoints GCs no GCs GCs no GCs N N Units P-value Restenosis 22 1122 percentage 9.1 16.4 0.56 Ischemic stroke 33 1615 percentage 15.2 5.9 0.05 Coronary artery disease 33 1615 percentage 12.1 7.6 0.32 Myocardial infarction 33 1615 percentage 9.1 4.5 0.19 Coronary interventions 33 1615 percentage 9.1 5.0 0.24 Peripheral events 33 1615 percentage 12.1 13.6 1.00 Cardiovascular death 33 1615 percentage 21.2 5.7 0.00 All-cause Death 33 1615 percentage 33.3 11.4 0.00 Composite cardiovascular event 33 1615 percentage 48.5 26.9 0.01 Comparison of the incidence of endpoints between patient-groups, after median follow-up of 2.6 years, with the exception of restenosis which was determined by echo-duplex at 1 year. Abbrevia- tions: GCs, patients using glucocorticoids; No GCs, patients not using glucocorticoids.
Discussion
We present data showing an increased incidence of cardiovascular events as well as an increased risk of cardiovascular- and all-cause death associated with systemic glucocorticoid treatment of patients with severe carotid stenosis. The increase in mortality risk was not reflected by the composition of the atherosclerotic plaque, which was not associated with GC use. The found increased risk could be due to confounding-by-indication, since there is an increased incidence of comorbidities like COPD, rheumatoid- and kidney disease in the GC group. This is compounded by the likelihood that patients receiving GCs have more severe comorbidity. For example, GCs are used in rheumatic patients with active disease to obtain disease control, after which GCs are preferably discontinued in favor of disease-modifying antirheumatic drugs (DMARDs). 9 To adjust for possible confounding by these comorbidities, we included them as covariates in the risk model. In addition, we performed a subanalysis in patients with either COPD or rheumatoid disease, the two main indications for GCs in this cohort. This subanalysis, which suffers in power from a reduction in sample size, still shows increased all-cause mortality as well as a persistent trend, albeit non-significant, for increased risk for all cardiovascular endpoints. However, by virtue of the observational nature of the study, we cannot exclude that there is residual confounding-by-indication. The use of GCs has also been associated with an increase of cardiometabolic risk factors.4,5 We did not observe cardiometabolic differences at time of inclusion, with the notable exception of BMI. To correct for possible confounding by indirect effects of GCs, we included BMI as a covariate in the model which did not significantly alter results. It is interesting that despite the increased risk of cardiovascular events at follow-up due to GCs, we did not observe an association between GC use and carotid plaque composition.
20162300 Marten Siemelink.indd 155 15-03-16 13:23 156 | CHAPTER 9
Table 4. Hazard ratios Outcome in all patients Model 1 (95% CI) Model 2 (95% CI) Model 3 (95% CI) Ischemic stroke 2.8 (1.1 - 6.8) 2.6 (1.0 - 6.4) 2.2 (0.7 - 6.7) Coronary artery disease 1.9 (0.7 - 5.0) 1.7 (0.6 - 4.7) 1.1 (0.4 - 3.2) Myocardial infarction 2.4 (0.7 - 7.5) 2.1 (0.7 - 6.8) 1.4 (0.4 - 5.0) Coronary interventions 2.1 (0.7 - 6.8) 2.1 (0.7 - 6.8) 1.5 (0.4 - 5.2) Peripheral artery events 1.1 (0.4 - 2.9) 1.1 (0.4 - 2.9) 1.4 (0.5 - 4.0) Cardiovascular death 4.5 (2.1 - 9.7) 3.8 (1.7 - 8.2) 2.7 (1.1 - 6.7) All-cause death 3.6 (2.0 - 6.6) 3.0 (1.6 - 5.5) 2.3 (1.1 - 4.8) Composite cardiovascular event 2.1 (1.3 - 3.4) 2.0 (1.2 - 3.3) 1.6 (0.9 - 2.8)
Outcome in subset: COPD and RA patients Model 1 (95% CI) Model 2 (95% CI) Model 3 (95% CI) Ischemic stroke 1.9 (0.6 - 5.6) 1.9 (0.6 - 5.9) 1.8 (0.5 - 7.8) Coronary artery disease 1.2 (0.4 - 3.4) 1.2 (0.4 - 3.6) 1.4 (0.4 - 4.6) Myocardial infarction 1.6 (0.4 - 5.4) 1.7 (0.5 - 5.9) 2.2 (0.5 - 9.0) Coronary interventions 1.4 (0.4 - 4.9) 1.7 (0.4 - 6.0) 2.2 (0.5 - 10) Peripheral artery events 1.3 (0.5 - 3.9) 1.5 (0.5 - 4.3) 2.3 (0.7 - 7.6) Cardiovascular death 2.9 (1.1 - 7.4) 2.8 (1.1 -7.5) 2.3 (0.7 - 6.7) All-cause death 2.7 (1.3 - 5.6) 2.6 (1.2 - 5.3) 2.3 (1.0 - 5.5) Composite cardiovascular event 1.8 (1.0 - 3.1) 1.8 (1.0 - 3.2) 2.1 (1.1 - 4.1) Cox proportional hazard models of cardiovascular outcomes for all patients (N=1854) and for the subset of patients (N=267) with either chronic obstructive pulmonary disease (COPD) or rheuma- toid disease (RA). Hazard ratios with corresponding confidence intervals are depicted for patients using glucocorticoids, compared to patients not using glucocorticoids. Three models of increasing complexity were used to investigate hazards while correcting for possible confounders. Model 1: glucocorticoids. Model 2: glucocorticoids, age, gender. Model 3: glucocorticoids, age, gender, BMI, beta-blockers, comorbidities, presenting symptoms. Abbreviations: CI, confidence interval.
This may be due to limited precision of the histological data, since histological examination of plaque characteristics is subject to interpretation by the observer.11 It could also be that the anti-inflammatory properties of GCs and the presumed pro-atherogenic properties of GCs have opposing effects on the vascular wall, effectively mitigating major changes to the plaque composition. Alternatively, GCs may impart increased risk for cardiovascular events through other mechanisms than alteration of plaque composition, like hypertension or atrial fibrillation.12 The here presented study has several limitations. 1) The main limitation of this cohort study is its observational design that limits conclusions on causality. Although we corrected for known confounders the results may be influenced by unknown confounders we cannot control for. 2) The current study is limited by the number of patients using corticosteroids and the lack of data on dose and duration of GC usage. While the indication for systemic
20162300 Marten Siemelink.indd 156 15-03-16 13:23 GLUCOCORTICOID RISK AFTER CAROTID ENDARTERECTOMY | 157
Table 5. Histology Plaque histology GCs no GCs GCs no GCs N N Units P-value General phenotype 39 1766 Atheromatous 11 454 percentage 28 26 0.85 Fibroatheromatous 13 653 percentage 33 37 0.74 Fibrous 15 659 percentage 39 37 1.00 Features Atheroma 39 1774 percentage 'high' 28 26 0.85 Smooth-muscle cells 35 1636 percentage of area 0.27 0.35 0.89 Macrophages 35 1650 percentage of area 1.22 1.35 0.32 Calcification 39 1771 percentage 'high' 51 49 0.87 Collagen 39 1770 percentage 'high' 80 80 0.84 Plaque Hemorrhage 39 1772 percentage present 67 60 0.41 Plaque Microvessels 32 1494 number per field 10 7.3 0.33 Comparison of plaque phenotypes between patients using glucocorticoids and patients not using glucocorticoids, based on immunohistochemical staining of the endarterectomy specimen. Abbre- viations: GCs, patients using glucocorticoids; No GCs, patients not using glucocorticoids.
GCs in rheumatoid arthritis is often for long-term treatment, GC prescription in COPD patients is often confined to periods of exacerbations of symptoms. The demonstrated association between GCs and CVD risk, despite this data limitation, could indicate that short-duration GC therapy also increases CVD risk or indicate an underestimation of the effect of long-duration GC therapy. 3) In this cohort, several indications for receiving GCs (for example RA and COPD) are risk-factors for CVD. However, details on disease duration and -severity are unavailable. 4) Information on the effect of GCs on cardiac and vascular function is lacking. 5) In addition, most indications for GCs are chronic inflammatory diseases, and chronic inflammation is a known contributor to CVD risk, possibly influencing results. 6) The study suffers from significant loss-to-follow-up. Yet, the percentage of 9 patients which are lost to follow-up are higher in the GC group, and are more likely to represent an underestimation of the GC associated risks. The results of this study are based on epidemiological evidence, and do not provide any information on causality of the found association of glucocorticoids with cardiovascular risk. In contrast to the strong cardiovascular risk increase we report here, studies in patients with chronic inflammatory diseases like rheumatoid arthritis,13–15 polymialgia rheumatic,16,17 systemic lupus erythematosus18 and systemic sclerosis19 for which long-term prescription of GC is common, show equivocal effects of GCs on cardiovascular events. However, in patients with COPD20 or asthma21,22 the use of inhalation GCs seems beneficial. In addition, GCs are close structural homologues of endogenous cortisol and an increased risk of cardiovascular disease has been described for hypercortisolism disorders like Cushing’s syndrome.23–26 So, placing our study in light of previous studies on the effects of glucocorticoids on cardiovascular disease, the evidence is still inconclusive and solid
20162300 Marten Siemelink.indd 157 15-03-16 13:23 158 | CHAPTER 9
evidence on causality and possible mechanism of action is still lacking. While a classical randomized controlled trial would face ethical issues, a genetic ‘Mendelian randomization study’ of cortisol may provide strong evidence of the imposed risks by GCs.
Conclusion
Following carotid endarterectomy, the use of systemic glucocorticoids is independently associated with an increased incidence of cardiovascular events and an increased risk for cardiovascular- and all-cause death, but not atherosclerotic plaque composition.
20162300 Marten Siemelink.indd 158 15-03-16 13:23 GLUCOCORTICOID RISK AFTER CAROTID ENDARTERECTOMY | 159
References
1. Taylor F, Huffman MD, Macedo AF, Moore THM, Burke M, Davey Smith G, Ward K, Ebrahim S. Statins for the primary prevention of cardiovascular disease. Cochrane database Syst Rev. 2013;1:CD004816. 2. Mora S, Wenger NK, Demicco DA, Breazna A, Boekholdt SM, Arsenault BJ, Deedwania P, Kastelein JJP, Waters DD. Determinants of residual risk in secondary prevention patients treated with high- ver- sus low-dose statin therapy: the Treating to New Targets (TNT) study. Circulation. 2012;125:1979–87. 3. Tendera M, Aboyans V, Bartelink M-L, Baumgartner I, Clément D, Collet J-P, Cremonesi A, De Carlo M, Erbel R, Fowkes FGR, Heras M, Kownator S, Minar E, Ostergren J, Poldermans D, Riambau V, Roffi M, Röther J, Sievert H, van Sambeek M, Zeller T. ESC Guidelines on the diagnosis and treatment of peripheral artery diseases: Document covering atherosclerotic disease of extracranial carotid and vertebral, mesenteric, renal, upper and lower extremity arteries: the Task Force on the Diagnosis and Treatm. Eur Heart J. 2011;32:2851–906. 4. Sarnes E, Crofford L, Watson M, Dennis G, Kan H, Bass D. Incidence and US costs of corticosteroid- associated adverse events: a systematic literature review. Clin Ther. 2011;33:1413–32. 5. Czock D, Keller F, Rasche FM, Häussler U. Pharmacokinetics and pharmacodynamics of systemically administered glucocorticoids. Clin Pharmacokinet. 2005;44:61–98. 6. Seguro LPC, Rosario C, Shoenfeld Y. Long-term complications of past glucocorticoid use. Autoimmun Rev. 2013;12:629–32. 7. Varas-Lorenzo C, Rodriguez LAG, Maguire A, Castellsague J, Perez-Gutthann S. Use of oral corticos- teroids and the risk of acute myocardial infarction. Atherosclerosis. 2007;192:376–83. 8. Wei L, MacDonald TM, Walker BR. Taking glucocorticoids by prescription is associated with subse- quent cardiovascular disease. Ann Intern Med. 2004;141:764–70. 9. Gupta PK, Pipinos II, Miller WJ, Gupta H, Shetty S, Johanning JM, Longo GM, Lynch TG. A population- based study of risk factors for stroke after carotid endarterectomy using the ACS NSQIP database. J Surg Res. 2011;167:182–91. 10. Verhoeven BAN, Velema E, Schoneveld AH, de Vries JPPM, de Bruin P, Seldenrijk CA, de Kleijn DP V, Busser E, van der Graaf Y, Moll F, Pasterkamp G. Athero-express: differential atherosclerotic plaque expression of mRNA and protein in relation to cardiovascular events and patient characteristics. Ra- tionale and design. Eur J Epidemiol. 2004;19:1127–33. 11. Hellings WE, Pasterkamp G, Vollebregt A, Seldenrijk CA, De Vries J-PPM, Velema E, De Kleijn DP V, Moll FL. Intraobserver and interobserver variability and spatial differences in histologic examination of carotid endarterectomy specimens. J Vasc Surg. 2007;46:1147–54. 12. Savelieva I, Kakouros N, Kourliouros A, Camm AJ. Upstream therapies for management of atrial fibril- lation: review of clinical evidence and implications for European Society of Cardiology guidelines. Part I: primary prevention. Europace. 2011;13:308–28. 13. Ruyssen-Witrand A, Fautrel B, Saraux A, Le Loët X, Pham T. Cardiovascular risk induced by low-dose corticosteroids in rheumatoid arthritis: a systematic literature review. Joint Bone Spine. 2011;78:23–30. 14. Ajeganova S, Andersson MLE, Frostegård J, Hafström I. Disease factors in early rheumatoid arthritis are associated with differential risks for cardiovascular events and mortality depending on age at on- set: a 10-year observational cohort study. J Rheumatol. 2013;40:1958–66. 15. Del Rincón I, Battafarano DF, Restrepo JF, Erikson JM, Escalante A. Glucocorticoid dose thresholds as- sociated with all-cause and cardiovascular mortality in rheumatoid arthritis. Arthritis Rheumatol (Hobo- ken, NJ). 2014;66:264–72. 9 16. Maradit Kremers H, Reinalda MS, Crowson CS, Davis JM, Hunder GG, Gabriel SE. Glucocorticoids and cardiovascular and cerebrovascular events in polymyalgia rheumatica. Arthritis Rheum. 2007;57:279– 86. 17. Mazzantini M, Torre C, Miccoli M, Baggiani A, Talarico R, Bombardieri S, Di Munno O. Adverse events during longterm low-dose glucocorticoid treatment of polymyalgia rheumatica: a retrospective study. J Rheumatol. 2012;39:552–7. 18. Schoenfeld SR, Kasturi S, Costenbader KH. The epidemiology of atherosclerotic cardiovascular dis- ease among patients with SLE: a systematic review. Semin Arthritis Rheum. 2013;43:77–95. 19. Vettori S, Maresca L, Cuomo G, Abbadessa S, Leonardo G, Valentini G. Clinical and subclinical athero- sclerosis in systemic sclerosis: consequences of previous corticosteroid treatment. Scand J Rheuma- tol. 2010;39:485–9. 20. Ghoorah K, De Soyza A, Kunadian V. Increased cardiovascular risk in patients with chronic obstructive pulmonary disease and the potential mechanisms linking the two conditions: a review. Cardiol Rev. 2013;21:196–202. 21. Iftikhar IH, Imtiaz M, Brett AS, Amrol DJ. Cardiovascular safety of long acting beta agonist-inhaled corticosteroid combination products in adult patients with asthma: a systematic review. Lung. 2014;192:47–54.
20162300 Marten Siemelink.indd 159 15-03-16 13:23 160 | CHAPTER 9
22. Otsuki M, Miyatake a, Fujita K, Hamasaki T, Kasayama S. Reduced carotid atherosclerosis in asthmatic patients treated with inhaled corticosteroids. Eur Respir J. 2010;36:503–8. 23. Tsuiki M, Tanabe A, Takagi S, Naruse M, Takano K. Cardiovascular Risks and Their Long-Term Clinical Outcome in Patients with Subclinical Cushing’s Syndrome. Endocr J. 2008;55:737–745. 24. Fardet L, Petersen I, Nazareth I. Risk of cardiovascular events in people prescribed glucocorticoids with iatrogenic Cushing’s syndrome: cohort study. BMJ Br Med J. 2012;4928:1–13. 25. Di Dalmazi G, Vicennati V, Rinaldi E, Morselli-Labate AM, Giampalma E, Mosconi C, Pagotto U, Pasqua- li R. Progressively increased patterns of subclinical cortisol hypersecretion in adrenal incidentalomas differently predict major metabolic and cardiovascular outcomes: a large cross-sectional study. Eur J Endocrinol. 2012;166:669–77. 26. Neary NM, Booker OJ, Abel BS, Matta JR, Muldoon N, Sinaii N, Pettigrew RI, Nieman LK, Gharib AM. Hypercortisolism is associated with increased coronary arterial atherosclerosis: analysis of noninva- sive coronary angiography using multidetector computerized tomography. J Clin Endocrinol Metab. 2013;98:2045–52.
20162300 Marten Siemelink.indd 160 15-03-16 13:23 | 161
20162300 Marten Siemelink.indd 161 15-03-16 13:23 20162300 Marten Siemelink.indd 162 15-03-16 13:23 CHAPTER 10
Discussion and Future Perspectives
20162300 Marten Siemelink.indd 163 15-03-16 13:23 164 | CHAPTER 10
Discussion and future perspectives
Many societies across the globe suffer from the burden of cardiovascular disease on the health system.1 To lower this burden requires advances in risk prediction and subsequent preventative treatment strategies, for which an improved understanding of complex biology of disease is imperative. Although it may seem that current biomarkers are hard to improve upon, there is still a long road ahead. Current risk prediction only provides a 10-year risk of cardiovascular events, but does not accurately predict individual event occurrence and fails to lower disease burden.2,3 In Chapter 2 we have provided a summary of new efforts in biomarker development across emerging techniques and biological sources. While individual techniques and markers seem promising, integration from different sources is crucial to understand pathophysiology and improve prediction. We speculate that the interplay between genetic, epigenetic and environmental risk factors may prove decisive in event occurrence, thus requiring risk assessment by including information on all of these processes. Furthermore, we conclude that genetic associations provide a powerful instrument to determine the causality of a biomarker in the pathophysiological processes of complex diseases. This is a key step to determine the value of a biomarker as a drug target. RNA is one of the biological sources of biomarkers that have gained much traction in recent years. Assessment of all coding and non-coding RNA in a biological sample, referred to as the ‘transcriptome’, has yielded some promising biomarkers for translation to clinical practice. In Chapter 3 we reviewed these transcriptional biomarkers and associated techniques, evaluating promises, challenges and pitfalls. We describe several RNA- biomarkers under development including expression signatures and non-coding RNA. Furthermore, we describe the importance of an integrative approach with genetic and epigenetic data for the development of transcriptional biomarkers.
Imbued with this knowledge on current biomarker developments, we turn our sights to specific contributors of cardiovascular risk. First we discuss heritable factors of CVD risk. In Chapter 4 known CVD-risk variants were associated with human carotid plaque histological characteristics. We report an approximate 4-fold enrichment of CVD-risk loci for an association with plaque characteristics. Most of these loci were associated with plaque atheroma. Additionally, we show that polygenic risk-scores for CAD associated with the number of macrophages and smooth muscle cells in the plaque, while scores for LAS associated with presence of haemorrhage in the plaque. We continued our scrutiny of the genetics of the carotid plaque in Chapter 5. In this chapter, plasma lipids concentrations and the genetic determinant that influencing these concentrations, are associated with histological features of carotid plaques. In contrast to the current paradigm of the ‘vulnerable plaque’,4 we find no association of plasma lipids or genetic burden scores of plasma lipids with carotid atherosclerotic plaque composition. This is surprising in light of the wealth evidence for LDL-C as a contributor to CVD risk.5,6 However, the lack of associations should be considered carefully due to limited statistical power. Future efforts will have to determine if plasma lipids in the end also affect carotid plaque composition or alternatively, may induce CVD risk through other mechanisms.
20162300 Marten Siemelink.indd 164 15-03-16 13:23 DISCUSSION AND FUTURE PERSPECTIVES | 165
From the genetics that may increase our propensity for vulnerable plaque development, we proceeded to the epigenetics of the carotid plaque. As discussed in the introduction, epigenetic mechanisms are important for the regulation of gene expression. While epigenetics encompasses several mechanisms which act in concert, we focused on DNA methylation. One particular interest in this thesis is how genetic variance may affect DNA methylation, known as ‘methylation quantitative trait loci’ or meQTL.7,8 Chapter 6 describes how we investigated the effects of genetic variation on DNA methylation in the disease lesion. We showed that 23 known CVD-risk SNPs are associated with differential DNA methylation at 74 nearby CpGs in carotid atherosclerotic plaque tissue. Of these 23 SNPs, 9 associate with DNA-methylation and gene-expression at the same gene. This may indicate that these SNPs exert their effects through a change in DNA methylation, and suggests the involvement of associated CpGs in CVD-risk. Furthermore, it implies that several novel CpG loci may be involved in CVD. Finally, this study identifies several meQTL effects which were not found in blood cells and may indicate lesion specific effects. We believe that this study brings the identified CpGs into the limelight for further scrutiny, providing evidence for close interplay between CVD-related genetic variation, DNA methylation and gene expression.
Of particular interest is that DNA methylation can be affected by genetic variation as well as environmental risk factors. One of the most prominent risk factors for the development of CVD is tobacco smoking. In Chapter 7 we engaged in an epigenome-wide association study of smoking in carotid atherosclerotic plaque samples. The study results show that smoking is strongly associated with differential DNA methylation in carotid atherosclerotic plaques. The study replicates 28 CpG loci previously shown in circulating cells to associate with smoking, predominantly in a study by Guida et al.9 This also included several CpGs in the well-known AHRR gene involved in xenobiotic detoxification.10 Notably, it also shows 40 novel epigenome-wide significant CpGs, only associated with smoking in plaque samples. Most excitingly however, is the association of six of these CpGs with the occurrence of clinical manifestations of atherosclerotic disease, irrespective of smoking behaviour. Conceivably, these CpGs may be involved in common pathophysiological pathways between smoking and other risk factors. Finally, meQTL analysis showed four CpGs possibly affected by nearby SNPs, potentially revealing hereditary susceptibility to vascular smoke toxicity.
The epidemiology of CVD shows that men have a higher risk of suffering from CVD, yet there is also a high incidence of CVD in women.11 There is some evidence that women suffer from a more microvascular type of atherosclerosis, more likely to lead to plaque erosion instead of plaque rupture. A first step in bridging the gap may come from DNA 10 methylation studies. To this end, Chapter 8 presents an epigenome-wide association study into sex-specific differentially methylated regions on autosomal chromosomes in the vascular lesion. The results indicate that at least 311 CpGs are differentially methylated between the sexes. The majority of these CpGs have been previously described in similar studies in other tissues.12 Our study confirms these differences in the disease lesion, despite vastly different cell-types compared to the other studies. How- and if this affects the complex biology of atherosclerotic disease is as of yet unclear. Hypothetically, these apparently innate differences may modify physiological processes that contribute to cardiovascular disease.
20162300 Marten Siemelink.indd 165 15-03-16 13:23 166 | CHAPTER 10
To drive down societal CVD burden, requires identification of individuals at risk and subsequent preventative action. Currently, CVD prevention is based on healthy lifestyle changes and pharmaceutical improvement of blood cholesterol using statins. Conceivably, anti-inflammatory drug treatment could also be beneficial to reduce CVD risk. Chronic inflammation is a known risk factor for CVD and glucocorticoid drugs (GCs) have potent anti-inflammatory effects. Yet, GCs are close structural homologues of endogenous cortisol and hypercortisolism disorders such as Cushing’s syndrome increase the risk of CVD.13–16 In Chapter 9 we showed that glucocorticoid use is associated with increased CVD-risk in carotid endarterectomy patients, though we do not find an effect on the carotid plaque composition. Viewing these results in the context of other studies, one has to conclude there is conflicting evidence on the effects of GCs on CVD risk. A Mendelian Randomization (MR) study of cortisol may solve this issue. A MR study uses naturally occurring genetic variation in a population as an instrument, providing a powerful method to investigate causality.20 Ultimately, whether the detriment of glucocorticoid treatment outweighs its benefits due to lowering of inflammation may depend on the treatment indication, as is shown by varying effects in patients with systemic inflammatory diseases.
Following up on the results presented in this thesis, it will be interesting to see if epigenetic blood signatures may serve as an easily accessible source of biomarkers of disease. Although we found some important differences in methylation between peripheral blood and atherosclerotic plaque tissue, blood methylation may still provide crucial information. Indeed, this has already been indicated for DNA methylation at the gene F2RL3, which is associated with smoking as well as mortality.17,18 Additionally, it will be interesting to see if the constitutive differences in DNA-methylation between the sexes fundamentally contribute to the known differences in CVD disease between men and women. This will also require further scrutiny of DNA methylation on the sex chromosomes, which is difficult to investigate with current microarray technology. Furthermore, meQTL associations may develop into an important tool to determine the pathophysiology of CVD risk SNPs, well suited to indicate regulatory effects of SNPs in a genome wide and tissue specific fashion.7,19 As such, it would be informative to investigate the presented meQTL- associations (Chapter 4, 6-8) in other CVD-related tissues like liver and intestine, in order to determine in which tissues CVD-risk SNPs contribute to disease. While the effects of SNPs are generally small, conceivably the effect of methylation on gene expression and thereby on disease risk may be substantial. This may present epigenetic loci as attractive drug targets, given that a causal effect on CVD risk can be proven. This may be investigated by a MR study of comparing SNPs and their corresponding CpGs to CVD-outcome in a larger cohort.
Challenges Despite best efforts, the presented studies in this thesis are subject to several impediments. A number of the encountered challenges stem from limitations due to the nature of the Athero-Express Biobank study. Briefly, they include the observational study design, modest cohort size, possible referral- and selection bias of patients, study withdrawal and lost to follow-up. Furthermore, the assessment of plaque histology parameters inevitably suffers from a lack of accuracy and precision, due to inconsistency
20162300 Marten Siemelink.indd 166 15-03-16 13:23 DISCUSSION AND FUTURE PERSPECTIVES | 167
in culprit lesion identification as well as observer bias.21 Regardless of these challenges, the Athero-Express Biobank remains an unequalled resource for investigation of carotid artery disease.
Validation and replication Large scale discovery studies in genomic and epigenomic research are prone to false positive associations, due to the sheer number of tests performed. This is controlled to some extend by stringent quality control procedures and multiple testing corrections, which we consistently performed according to international consensus. Still, the primary methods to affirm previous findings and reduce false-positive findings are validation and replication of results. Preferably, replication is accomplished by means of the initial technique in a comparable patient cohort of sufficient size. Exactly this requirement of replication is a recurring drawback of the unique properties of the Athero-Express. To the best of our knowledge, there are hardly any cohorts of similar patients of notable size, and none of these cohorts have comparable data including plaque derived phenotypes like histology, plaque proteins and plaque DNA methylation. We did find supporting evidence in previous studies performed in other tissues for associated CpG loci in chapters 7 and 8. Lesser alternatives are replication of findings in cohorts in other vascular territories or in animal- or in-vitro studies, which we did not pursue. Validation of results is performed by measuring the same parameters in the same cohort, by means of another technique. This may show that the findings are not due to technical artefacts or sample handling errors and that the initial results are valid. We did not perform any validation studies; instead we used bioinformatic repositories as additional sources for supportive evidence.
Coverage Importantly, both the genetic and epigenetic research presented here, are based on measurements using microarray technology platforms. Although these arrays offer several advantages which make them suitable for many research purposes, they have important weaknesses for clinical application. Most importantly, one must acknowledge the fact that they only provide data on a small fraction of the genome and epigenome. It has been argued that they provide ample coverage since they cover loci of common variation (genetic) or loci with strong effects (epigenetics) and that missing loci may be accurately imputed. This is a fallacy, as it is a data reduction compared to whole genome (bisulphate) sequencing and imputation quality is poor for rare variants and consequential mutations will be missed altogether. Therefore, it is paramount to realize that current (epi)genome wide studies, such as presented here, may discover genomic regions of importance for disease but do 10 not capture all relevant genomic information.
Future of genomic medicine A solution to the current lack of coverage and resulting data deprivation will be the development of methods that offer better coverage, ultimately towards cost-effective whole-genome sequencing. Currently, next-generation sequencing is still prohibitively expensive for most applications, yet prices are expected to keep falling. However, this will also vastly enlarge the data that will need to be analyzed, creating new challenges. Computing infrastructures will need to be scaled to cope with the copious amounts of data
20162300 Marten Siemelink.indd 167 15-03-16 13:23 168 | CHAPTER 10
and data processing. Additionally, it will aggravate the problem of multiple-testing correction, requiring very large cohorts to attain sufficient power. Despite the aforementioned hurdles which will need to be overcome, whole-genome sequencing is certainly set to radically transform medicine as we know it. Next-generation sequencing is actually a continuing trend of ever faster, cheaper and more reliable technology to measure genetic sequences. As soon as whole-genome sequencing will become affordable, it will surely become a pervasive part of medicine. One does not need to be visionary to envision a future where genetic information adds to personal risk- assessment for all diseases. Finally, unlocking of the genome will contribute significantly to our understanding of cardiovascular disease and will be an important aid for the development of more effective drugs and treatment strategies. Taking these trends into account, the predicted age-of-biology in the 21st century may indeed come true.
Conclusion With this thesis, we embarked on a journey of exploration, aiming to explore the complex biology of the human carotid atherosclerotic plaque. This journey has shown that integration of genotyping, epigenotyping and phenotyping of the atherosclerotic disease lesion may yield meaningful new insights in carotid atherosclerotic disease. This contributes to a genuine understanding of the biology of disease, essential for precise individual risk prediction and disease prevention.
20162300 Marten Siemelink.indd 168 15-03-16 13:23 DISCUSSION AND FUTURE PERSPECTIVES | 169
References
1 Mendis S, Puska P, Norrving B. Global Atlas on cardiovascular disease prevention and control. Geneva, 2011. 2 Perk J, De Backer G, Gohlke H, et al. European Guidelines on cardiovascular disease prevention in clinical practice (version 2012). The Fifth Joint Task Force of the European Society of Cardiology and Other Societies on Cardiovascular Disease Prevention in Clinical Practice (constituted by re. Eur Heart J 2012; 33: 1635–701. 3 Goff DC, Lloyd-Jones DM, Bennett G, et al. 2013 ACC/AHA guideline on the assessment of cardiovas- cular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol 2014; 63: 2935–59. 4 Plasschaert H, Heeneman S, Daemen MJ. Progression in atherosclerosis: histological features and pathophysiology of atherosclerotic lesions. Top Magn Reson Imaging 2009; 20: 227–37. 5 Voight BF, Peloso GM, Orho-Melander M, et al. Plasma HDL cholesterol and risk of myocardial infarc- tion: a mendelian randomisation study. Lancet 2012; 6736: 1–9. 6 Holmes M V, Asselbergs FW, Palmer TM, et al. Mendelian randomization of blood lipids for coronary heart disease. Eur Heart J 2014; published online Jan 27. DOI:10.1093/eurheartj/eht571. 7 Grundberg E, Meduri E, Sandling JK, et al. Global analysis of DNA methylation variation in adipose tis- sue from twins reveals links to disease-associated variants in distal regulatory elements. Am J Hum Genet 2013; 93: 876–90. 8 Liu Y, Li X, Aryee MJ, et al. GeMes, clusters of DNA methylation under genetic control, can inform genetic and epigenetic analysis of disease. Am J Hum Genet 2014; 94: 485–95. 9 Guida F, Sandanger TM, Castagné R, et al. Dynamics of smoking-induced genome-wide methylation changes with time since smoking cessation. Hum Mol Genet 2015; 24: 2349–59. 10 Klingbeil EC, Hew KM, Nygaard UC, Nadeau KC. Polycyclic aromatic hydrocarbons, tobacco smoke, and epigenetic remodeling in asthma. Immunol Res 2014; 58: 369–73. 11 Mosca L, Barrett-Connor E, Wenger NK. Sex/gender differences in cardiovascular disease prevention: what a difference a decade makes. Circulation 2011; 124: 2145–54. 12 McCarthy NS, Melton PE, Cadby G, et al. Meta-analysis of human methylation data for evidence of sex-specific autosomal patterns. BMC Genomics 2014; 15: 981. 13 Tsuiki M, Tanabe A, Takagi S, Naruse M, Takano K. Cardiovascular Risks and Their Long-Term Clinical Outcome in Patients with Subclinical Cushing’s Syndrome. Endocr J 2008; 55: 737–45. 14 Fardet L, Petersen I, Nazareth I. Risk of cardiovascular events in people prescribed glucocorticoids with iatrogenic Cushing’s syndrome: cohort study. BMJ Br Med J 2012; 4928: 1–13. 15 Di Dalmazi G, Vicennati V, Rinaldi E, et al. Progressively increased patterns of subclinical cortisol hyper- secretion in adrenal incidentalomas differently predict major metabolic and cardiovascular outcomes: a large cross-sectional study. Eur J Endocrinol 2012; 166: 669–77. 16 Neary NM, Booker OJ, Abel BS, et al. Hypercortisolism is associated with increased coronary arte- rial atherosclerosis: analysis of noninvasive coronary angiography using multidetector computerized tomography. J Clin Endocrinol Metab 2013; 98: 2045–52. 17 Breitling LP, Salzmann K, Rothenbacher D, Burwinkel B, Brenner H. Smoking, F2RL3 methylation, and prognosis in stable coronary heart disease. Eur Heart J 2012; 33: 2841–8. 18 Zhang Y, Yang R, Burwinkel B, et al. F2RL3 methylation in blood DNA is a strong predictor of mortality. Int J Epidemiol 2014; 43: 1215–25. 19 Lemire M, Zaidi SHE, Ban M, et al. Long-range epigenetic regulation is conferred by genetic variation located at thousands of independent loci. Nat Commun 2015; 6: 6326. 20 Smith GD. Mendelian Randomization for Strengthening Causal Inference in Observational Studies: Application to Gene x Environment Interactions. Perspect Psychol Sci 2010; 5: 527–45. 21 Hellings WE, Pasterkamp G, Vollebregt A, et al. Intraobserver and interobserver variability and spatial 10 differences in histologic examination of carotid endarterectomy specimens. J Vasc Surg 2007; 46: 1147–54.
20162300 Marten Siemelink.indd 169 15-03-16 13:23 20162300 Marten Siemelink.indd 170 15-03-16 13:23 CHAPTER 11
Dutch Summary
20162300 Marten Siemelink.indd 171 15-03-16 13:23 172 | CHAPTER 11
Nederlandse samenvatting (Dutch Summary)
Cardiovasculaire ziekten, beter bekend als hart- en vaatziekten (HVZ), zijn een voorname oorzaak van ziekte en overlijden. Dit ondanks de enorme onderzoeksinspanningen in recente decennia, gericht op verbetering van preventie en behandeling. Atherosclerose, oftewel aderverkalking, is de meest voorkomende vorm van hart- en vaatziekten. Atherosclerose kan zich geïsoleerd of tegelijk voordoen op verschillende voorkeursplaatsen in het lichaam, voornamelijk in de halsslagaderen (arteriae carotis), de kransslagaderen van het hart (arteriae coronariae), de grote lichaamsslagader (aorta) en in de bekken- en beenslagaderen (arteriae iliaca en femoralis).1 Aderverkalking van deze vaten kan leiden tot ernstige gezondheidsklachten zoals een herseninfarct, hartinfarct, het scheuren van de vaatwand in de grote lichaamsslagader (aorta dissectie of -ruptuur) en amputatie van de benen. In deze dissertatie ligt de focus voornamelijk op atherosclerose van de arteria carotis, maar ook andere manifestaties van aderverkalking worden soms besproken. Atherosclerose van de arteria carotis is een sluipend proces dat plaatsvindt over een tijdspanne van vele jaren. Langzaam slibt de slagader dicht door de opbouw van een atherosclerotische plaque (aderverkalking). Dit proces wordt elegant beschreven door Weber en collega’s alsmede door Jackson.2,3 Kort samengevat: eerst worden cholesterol en andere vetten afgezet in de vaatwand, wat leidt tot de aanmaak van signaalstoffen (chemokines) in de bovenliggende cellaag die de vaatwand van binnen bekleedt. Dit leidt vervolgens tot het aanhechten en uittreden van witte bloedcellen (leukocyten) uit de bloedstroom naar de vaatwand. Het zijn vooral monocyten en T-cellen die op deze manier de vaatwand betreden, waarschijnlijk in een poging om de vetten op te ruimen en de lokale onsteking te lijf te gaan. Wanneer het atherosclerotische proces verder vordert en het deze cellen niet lukt om de afzettingen op te ruimen, beginnen deze cellen af te sterven. Hierdoor ontstaat een lokale weefselsterfte (necrose) in de vaatwand, gevuld met vet, overblijfselen van cellen en cholesterolkristallen. Mede als gevolg hiervan beginnen gladde spiercellen in de vaatwand zich te delen, waardoor de diameter en doorgankelijkheid van de slagader afneemt. Uiteindelijk ontstaat er een instabiele situatie, die vaak zonder aankondiging leidt tot het scheuren van de plaque. Wanneer dat gebeurt komt de necrotische inhoud van de plaque in de bloedbaan terecht wat direct leidt tot stolselvorming. Een stolsel kan vervolgens de bloedstroom afsluiten waardoor een herseninfarct kan onstaan.
De Athero-Express Biobank Terwijl de precieze oorzaken en het moleculaire ontstaansmechanisme van atherosclerose nog niet precies bekend zijn, is er de laatste jaren wel meer inzicht verkregen. Alle informatie wijst erop dat atherosclerose een zeer complexe ziekte is waarbij vele risicofactoren en complexe biologische processen een rol spelen. Grote cohortstudies hebben geleid tot de ontdekking van belangrijke risicofactoren voor het ontstaan van atherosclerose, waarvan leeftijd, geslacht, roken, bloeddruk en cholesterol de bekendste zijn. Het bestaan van atherosclerose in de carotis is een belangrijke risicofactor voor verdere complicaties, zoals een herseninfarct. Daarvoor is een preventieve behandeling aangewezen, gebaseerd op een verandering van levensstijl en medicatie. Wanneer de vernauwing van het bloedvat ernstig is of wanneer er al sprake is van symptomen zoals een TIA of infarct, dan is vaak een operatie de aangewezen behandeling. Tijdens deze operatie, de
20162300 Marten Siemelink.indd 172 15-03-16 13:23 DUTCH SUMMARY | 173
“carotisendarteriëctomie”, wordt de halsslagader opengelegd en wordt de aderverkalking aan de binnenzijde van het bloedvat verwijderd, waardoor het bloedvat weer normaal doorgankelijk wordt. Het aangedane weefsel dat uit het bloedvat gehaald wordt, kan worden gebruikt voor wetenschappelijk onderzoek. Om daar optimaal gebruik van te maken bestaat sinds 2002 in het Universitair Medisch Centrum Utrecht en het Sint Antonius Ziekenhuis in Nieuwegein de Athero-Express Biobank. Daarin wordt het verkregen materiaal samen met het bloed van de patiënt opgeslagen. Daarnaast wordt de gezondheid van de patient nog drie jaar te gevolgd. Door de vakgroep Experimentele Cardiologie wordt veel onderzoek verricht op deze Biobank, waaronder microscopische eigenschappen van de aderverkalking alsmede genetische en epigenetische profielen van de desbetreffende patiënten. Dit geeft unieke mogelijkheden voor onderzoek naar atherosclerotische processen in de vaatwand en de onderlinge samenhang in die processen, die uiteindelijk kunnen leiden tot een verbeterde risicoschatting en preventie van atherosclerose.
Genetica en Epigenetica Belangrijke onderwerpen in deze dissertatie zijn de genetica en de epigenetica van atherosclerose. De genetische bijdrage aan atherosclerose van carotislijden wordt hoog geschat. Zo wordt de erfelijke bijdrage aan het overlijden aan een herseninfarct geschat op 32% en wordt de bijdrage aan de aanwezigheid van verkalking van de halsslagader zelfs geschat op 78%.4,5 Door verbeteringen in genetisch onderzoek is duidelijk geworden dat daar zeer veel plaatsen op het genoom bij betrokken zijn. Zo heeft een recente studie laten zien dat slechts 12,8% van de erfelijkheid wordt verklaard door veel voorkomende variatie in 20.000 SNPs (‘single nucleotide polymorphisms’; individuele basenparen) verspreid over het genoom.6 Dit geeft tevens aan dat nog een substantieel deel van de erfelijkheid van atherosclerose van de halsslagader onverklaard is. Onderzoek naar SNPs die bijdragen aan atherosclerose van de halsslagader laat zien dat sommige SNPs ook betrokken zijn bij atherosclerose op andere plaatsen in het lichaam, wat duidt op een deels overeenkomstige ontstaanswijze.7 Om tot een betere risicoschatting te komen, kunnen de effecten van alle bijdragende genetische varianten worden opgeteld in een gewogen score, wat bekend staat als een ‘polygene risicoscore’. Zulke scores zijn momenteel belangrijke onderzoeksinstrumenten, maar hebben nog geen toepassing gevonden in de medische praktijk. Wellicht dat ze daarvoor ook van toegevoegde waarde blijken wanneer de precisie van de scores toeneemt. De coderende sequenties van het genoom beschrijven hoe eiwitten, de bouwstenen van het lichaam, gemaakt moeten worden. De niet-coderende sequenties zijn veelal betrokken bij het reguleren van gen-activatie en daarmee eiwitproductie. Daarnaast zijn er nog andere factoren die de activiteit van een gen bepalen. Sommige van deze regulatoren, waaronder DNA-methylatie, histonmodificaties en niet-coderende RNA’s, erven onhankelijk van de genoomsequentie over en vallen onder de noemer ‘epigenetica’. DNA-methylatie is een
proces waarbij een methyl-groep (CH3) wordt gebonden aan een aanliggend paar van cytosine- en guaninebasen (CpG) op de DNA sequentie. Afhankelijk van de locatie ten 11 opzichte van een gen, kan dit de activatie van het gen positief of negatief beïnvloeden.8 Het is gebleken dat genetische variatie ter plaatse van SNPs, de DNA-methylatie ter plaatse van specifieke CpGs kan beïnvloeden. SNPs die dit effect veroorzaken worden ‘methylation quantitative trait loci’ (meQTL) genoemd. Veel is nog onbekend over dit fenomeen, maar het lijkt er sterk op dat het de regulatie van genen kan beïnvloeden. Wellicht kan het ook
20162300 Marten Siemelink.indd 173 15-03-16 13:23 174 | CHAPTER 11
meer informatie opleveren over de interactie van verschillende genen op het genoom, zelfs als deze op grote afstand van elkaar of zelfs op andere chromosomen liggen.9
Het doel van deze dissertatie is om de complexe biologie die ten grondslag ligt aan aderverkalking van de halsslagader nader te onderzoeken. Daarmee kan op termijn een bijdrage geleverd worden aan betere risicopredictie en het ontwikkelen van beter afgestemde medicatie. Met dat doel voor ogen werden verscheidene parameters gemeten in de Athero-Express Biobank en gerelateerd aan cardiovasculaire risicofactoren, waaronder geslacht, genetische variatie, lipiden (vetten), medicatie en roken.
Om tot betere risicopredictie te komen zijn betere biomarkers voor atherosclerose van groot belang. In hoofdstuk 2 wordt een overzicht gegeven van de laatste internationale ontwikkelingen op dit gebied, waarbij de laatste technieken en bronnen van biomarkers de revue passeren. Dit wordt tevens besproken in het kader van genetisch en epigenetisch onderzoek. We speculeren daarbij dat met name de integratie van informatie uit verschillende bronnen uiteindelijk doorslaggevend zal zijn om tot een betere predictie te komen. Een belangrijke bron voor atherosclerotisch onderzoek en voor de ontwikkeling van biomarkers is het gehele profiel van genen die tot uiting komen, beter bekend als het ‘transcriptoom’. De mogelijkheden die het transcriptoom biedt voor biomarkerontwikkeling en risicopredictie worden besproken in hoofdstuk 3. Vervolgens wordt verder ingegaan op het belang van het genoom in hoofdstuk 4. Daar wordt van SNPs, waarvan bekend is dat ze het risico op atherosclerose beïnvloeden, bekeken of ze ook de samenstelling van de plaque beïnvloeden. Het blijkt dat er een viervoudige verrijking is van deze SNPs in de groep van SNPs die associëren met karakteristieken van de plaque samenstelling. De meeste van deze SNPs waren geassocieerd met vet in de plaque. In hoofstuk 5 wordt er nader ingegaan op hoe lipiden (vetten) in de bloedbaan en de bijbehorende genetische eigenschappen van invloed kunnen zijn op de plaquesamenstelling. In tegenstelling tot het bekende theoretische model van de ontstaanswijze van de plaque, worden er geen aanwijzingen voor dergelijke invloeden gevonden. Dit is verrassend in het licht van de bekende literatuur en kan mogelijk verklaard worden door een beperkte omvang van de beschikbare gegevens. Vanaf hoofdstuk 6 wordt er verder ingegaan op epigenetische processen. Deze processen zijn belangrijk voor de regulatie van genen. In hoofdstuk 6 wordt een studie beschreven waarin van SNPs, waarvan bekend is dat ze het risico op atherosclerose beïnvloeden, wordt bepaald of ze associëren met DNA-methylatie in hun nabijheid. Daarbij worden 23 SNPs gevonden die DNA-methylatie beïnvloeden ter plaatse van 74 CpGs. Van 9 van deze SNPs blijkt dat ze associëren met activatie van een gen alsmede met DNA-methylatie ter plaatse. Dit is een aanwijzing dat deze SNPs mogelijk de genexpressie beïnvloeden doordat ze de DNA-methylatie veranderen. Tevens identificeert deze studie CpGs die voorheen nog niet eerder met atherosclerose in verband zijn gebracht.
Bijzonder interessant is het gegeven dat DNA-methylatie ter plaatse van een CpG beïnvloed kan worden door genetische variatie alsmede door omgevingsfactoren in het lichaam of de buitenwereld. Een van de meest prominente omgevingsfactoren die het risico op atherosclerose verhoogd is het roken van tabak. In hoofdstuk 7 wordt een studie
20162300 Marten Siemelink.indd 174 15-03-16 13:23 DUTCH SUMMARY | 175
beschreven waarin onderzoek wordt gedaan naar DNA-methylatie in het weefsel van de atherosclerotische plaque. Daarbij wordt van een groot aantal CpGs (circa 450.000), verspreid over het genoom, gekeken naar de associatie met roken. Er worden 28 CpGs gevonden die al eerder waren geassocieerd met roken in bloedcellen. Tevens worden 40 voorheen onbekende CpGs gevonden die nog niet eerder beschreven waren in relatie tot roken. Daar zitten ook CpGs bij die in het AHRR gen liggen, dat betrokken is bij de afbraak van giftige stoffen in tabaksrook. Verder werden SNPs aangetoond die meQTL associaties vertonen met enkele van deze CpGs, wat een aanwijzing kan zijn voor een erfelijke invloed op de gevoeligheid voor roken. De epidemiologie toont dat mannen een groter risico hebben op hart- en vaatziekten dan vrouwen. Desondanks is er ook een hoge incidentie van hart- en vaatziekten bij vrouwen.10 Tevens zijn er aanwijzingen dat er bij vrouwen eerder sprake is van een andere opbouw van de plaque, waarbij er meer kleine bloedvaatjes in de plaque ontstaan en de plaque eerder geneigd is uit te slijten dan te barsten. In hoofdstuk 8 wordt onderzocht of DNA- methylatie een bijdrage kan leveren aan de verklaring van deze verschillen. Daarbij worden 311 CpGs gevonden die verschillen tussen de geslachten kunnen verklaren, waarvan het merendeel reeds eerder gevonden werd in andere studies in andere weefsels. Het lijkt er sterk op dat deze verschillen voorkomen in vele of misschien wel álle weefsels. Verder onderzoek zal moeten uitwijzen of deze verschillen ook bijdragen aan atherosclerose.
Om de grote invloed van hart- en vaatziekten op de algemene gezondheid van de bevolking verder terug te dringen zijn betere risicopredictie en behandeling noodzakelijk. Chronische onsteking is een bekende risicofactor voor hart- en vaatziekten. Hoewel het nog niet aangetoond is, kan verondersteld worden dat geneesmiddelen die ontsteking remmen ook het risico op atherosclerose verlagen. Een veel gebruikte en potente groep van ontstekingsremmers is glucocorticoïden (waaronder prednison), waarvan men zich kan afvragen of ze ook het risico op atherosclerose kunnen verlagen. Echter, glucocorticoïden zijn structureel nauw verwant aan het lichaamseigen stress-hormoon cortisol, waarvan bekend is dat het een risico-verhogend effect heeft.11–14 In hoofdstuk 9 wordt het effect van glucocorticoïden onderzocht op patiënten die een carotisendarteriëctomie ondergaan. Het blijkt daarbij dat het gebruik van glucocorticoïden gepaard gaat met een groter aantal complicaties van hart- en vaatziekten binnen een termijn van 3 jaar na de operatie. Of dat direct wordt veroorzaakt door de glucocorticoïden kan in deze studie niet aangetoond worden, daarvoor is verder onderzoek noodzakelijk.
Conclusie Deze dissertatie had ten doel om de complexe biologie van aderverkalking van de halsslagader nader te onderzoeken. Gedurende het onderzoek is gebleken dat de integratie van gegevens uit genotypering, epigenotypering en fenotypering van aderverkalkingsweefsel een belangrijke bijdrage kan leveren aan nieuwe inzichten in het proces van aderverkalking van de halsslagader. Dit draagt bij aan het werkelijk doorgronden van de biologie van 11 aderverkalking, wat essentieel is voor toekomstige individuele risicopredictie en preventie.
20162300 Marten Siemelink.indd 175 15-03-16 13:23 176 | CHAPTER 11
Referenties
1 Imori Y, Akasaka T, Ochiai T, et al. Co-existence of carotid artery disease, renal artery stenosis, and lower extremity peripheral arterial disease in patients with coronary artery disease. Am J Cardiol 2014; 113: 30–5. 2 Weber C, Noels H. Atherosclerosis: current pathogenesis and therapeutic options. Nat Med 2011; 17: 1410–22. 3 Jackson SP. Arterial thrombosis--insidious, unpredictable and deadly. Nat Med 2011; 17: 1423–36. 4 Bak S, Gaist D, Sindrup SH, Skytthe A, Christensen K. Genetic liability in stroke: a long-term follow-up study of Danish twins. Stroke 2002; 33: 769–74. 5 Tarnoki AD, Baracchini C, Tarnoki DL, et al. Evidence for a strong genetic influence on carotid plaque characteristics: an international twin study. Stroke 2012; 43: 3168–72. 6 Holliday EG, Traylor M, Malik R, et al. Genetic overlap between diagnostic subtypes of ischemic stroke. Stroke 2015; 46: 615–9. 7 Dichgans M, Malik R, König IR, et al. Shared genetic susceptibility to ischemic stroke and coronary artery disease: a genome-wide analysis of common variants. Stroke 2014; 45: 24–36. 8 Varley KE, Gertz J, Bowling KM, et al. Dynamic DNA methylation across diverse human cell lines and tissues. Genome Res 2013; 23: 555–67. 9 Lemire M, Zaidi SHE, Ban M, et al. Long-range epigenetic regulation is conferred by genetic variation located at thousands of independent loci. Nat Commun 2015; 6: 6326. 10 Mosca L, Barrett-Connor E, Wenger NK. Sex/gender differences in cardiovascular disease prevention: what a difference a decade makes. Circulation 2011; 124: 2145–54. 11 Tsuiki M, Tanabe A, Takagi S, Naruse M, Takano K. Cardiovascular Risks and Their Long-Term Clinical Outcome in Patients with Subclinical Cushing’s Syndrome. Endocr J 2008; 55: 737–45. 12 Fardet L, Petersen I, Nazareth I. Risk of cardiovascular events in people prescribed glucocorticoids with iatrogenic Cushing’s syndrome: cohort study. BMJ Br Med J 2012; 4928: 1–13. 13 Di Dalmazi G, Vicennati V, Rinaldi E, et al. Progressively increased patterns of subclinical cortisol hyper- secretion in adrenal incidentalomas differently predict major metabolic and cardiovascular outcomes: a large cross-sectional study. Eur J Endocrinol 2012; 166: 669–77. 14 Neary NM, Booker OJ, Abel BS, et al. Hypercortisolism is associated with increased coronary arte- rial atherosclerosis: analysis of noninvasive coronary angiography using multidetector computerized tomography. J Clin Endocrinol Metab 2013; 98: 2045–52.
20162300 Marten Siemelink.indd 176 15-03-16 13:23 | 177
20162300 Marten Siemelink.indd 177 15-03-16 13:23 20162300 Marten Siemelink.indd 178 15-03-16 13:23 APPENDIX
Review Committee Acknowledgements List of Publications Curriculum Vitae
20162300 Marten Siemelink.indd 179 15-03-16 13:23 180 | APPENDIX
Review Committee
Prof. dr. M.C. Verhaar (Chair) Department of Nephrology and Hypertension, Utrecht University Medical Center, Utrecht, the Netherlands
Prof. dr. J. Garssen Division of Pharmacology, Utrecht University, Utrecht, the Netherlands
Prof. dr. F.W. Asselbergs Department of Cardiology, Utrecht University Medical Center, Utrecht, the Netherlands
Dr. B.T. Heijmans Department of Molecular Epidemiology, Leiden University Medical Center, Leiden, the Netherlands
Dr. J.P. van Tintelen Department of Clinical Genetics, Academic Medical Center, Amsterdam, the Netherlands
20162300 Marten Siemelink.indd 180 15-03-16 13:23 APPENDIX | 181
Acknowledgements (Dankwoord)
Aan de totstandkoming van dit proefschrift hebben velen een bijdrage geleverd, waarvoor ik hen dank verschuldigd ben. Bij dezen wil ik dan ook graag eenieder die betrokken is geweest bedanken voor hun inzet en bijdrage. In het bijzonder wil ik mijn bewondering uitspreken voor alle patiënten die tijdens hun eigen ziekteproces onbaatzuchtig hebben deelgenomen aan de Athero-Express studie. Een aantal mensen wil ik hieronder graag specifiek bedanken.
Geachte prof. Pasterkamp, beste Gerard, ik zal nooit vergeten hoe ik voor het eerst bij je op gesprek kwam, nadat Berent had gezegd dat je hoge verwachtingen zou hebben. Vol enthousiasme gaf je meteen een tour van het lab en stelde me aan iedereen voor. Aan het einde bleef ik een beetje verbouwereerd achter en moest ik nog even vragen of ik nu was aangenomen of niet. In de eerste jaren van mijn promotie heb ik veel met je gespard over het ‘high-risk, high-reward’ idee dat we hadden, de ‘cell-based assays’. Vaak werden die discussies door mij bijna filosofisch ingestoken. Gelukkig ben jij veel pragmatischer ingesteld dan ik, waar ik veel van heb geleerd. Helaas hebben we het na twee jaar over een andere boeg moeten gooien, gelukkig heeft dat uiteindelijk goed uitgepakt. Wat ik het meeste aan je waardeer is je open management stijl. Je dwingt niet op autoritaire wijze respect af, je verdient respect omdat je ruimte geeft voor talent. Helaas kreeg je er ook gratis mijn koppige en eigengereide karakter bij, wat zo nu en dan tot goede discussies heeft geleid. Desondanks heb je me de afgelopen jaren alle vrijheid en kansen gegeven om het onderzoek tot een succesvol einde te brengen, waarvoor ik je veel dank verschuldigd ben. Ik heb deze tijd onder jouw hoede ervaren als een prachtige gelegenheid om mezelf op wetenschappelijk niveau alsmede op persoonlijk vlak te ontwikkelen. Ik zal mijn weg vervolgen in de genetica, allicht zullen onze paden in de toekomst nog vaker kruisen.
Geachte prof. Prakken, beste Berent, wat gaat de tijd toch hard. Al in 2005 kwam ik, groen als gras, bij jou aanzetten met de vraag of ik niet twee weken later met een stage kon beginnen. Ik was net toegelaten tot de SUMMA-opleiding en had nog 6 maanden te overbruggen. Nog dezelfde dag stond ik met Alvin op de OK om een thymus op te vangen bij een arteriële-switch operatie van een pasgeborene. In die zomermaanden van 2005 heb ik, met vallen en opstaan, mijn eerste echte stappen in de wetenschap gezet. Wat mij betreft is daar al de basis gelegd voor dit proefschrift. Dat heeft zich later vertaald naar een wetenschapsstage tijdens mijn geneeskunde opleiding en vervolgens naar mijn aanstelling bij Gerard. Ik heb je door de jaren leren kennen als een sympathieke persoonlijkheid en begeesterd wetenschapper. Al die tijd heb je een zekere potentie in mij gezien en deuren voor me geopend, daar ben ik je nog steeds zeer dankbaar voor.
Hamid, je bent echt iemand die de diepte in gaat en je bent altijd bereid om anderen te helpen en adviseren. Dat heb ik zeer aan je gewaardeerd tijdens mijn lab werk. Je hebt wel het een en ander met mij te stellen gehad, maar dat heeft je niet uit het veld geslagen. Bedankt voor je goede begeleiding tijdens mijn lab tijd. A
20162300 Marten Siemelink.indd 181 15-03-16 13:23 182 | APPENDIX
Geachte leden van de beoordelingscommissie en promotiecommissie: prof. dr. Verhaar, prof. dr. Garssen, prof. dr. Asselbergs, dr. Heijmans, dr. van Tintelen. Dank voor uw bereidheid om mijn manuscript te beoordelen en zitting te nemen in de promotiecommissie.
Alvin, mijn eerste echte aanvaring met de wetenschap was als student van jou. Ik was amper begonnen met mijn stage toen ik de boel per ongeluk lelijk verprutste, het was meteen een eerste les in wetenschappelijk vallen-en-opstaan. Daarna volgde velen weken onderzoek aan het thymectomie project, doorspekt van haast filosofische wetenschappelijke discussies, regelmatig onder genot van een speciaalbiertje. Deze tijd heb ik als zeer inspirerend ervaren en het heeft uiteindelijk geleid tot mijn eigen promotie onderzoek.
Hester, wat heeft de experimentele cardiologie er met jouw komst een goudhaantje bij gekregen. Je bent echt een rising star een gaat vast helemaal naar de top, niks glazen plafond. Ik vind het prachtig hoe je onderzoek naar cardiovasculaire ziekten bij vrouwen voor het voetlicht brengt en vond het geweldig om daar mijn steentje aan bij te kunnen dragen.
Bas Heijmans, mede door jouw enthousiaste optreden, zijn we in 2012 echt van start gegaan met de DNA methylatie studie, die uiteindelijk een belangrijk onderdeel is geworden van dit proefschrift. Je bent echt een bioloog in hart en nieren en dat spreek me erg aan, je was soms echt een welkome frisse wind ten opzichte van de medische onderzoekers. Wat vond ik het mooi om in Leiden met je en jouw groep te brainstormen over de basale mechanismen van DNA methylatie. Je hebt daar echt een paar toppers zitten: René en Koen, ontzettend bedankt voor jullie hulp en bijdragen aan de DNA methylatie studie.
Sander, de afgelopen jaren hebben we behoorlijk intensief samengewerkt, waarbij jij de genetica-kar hebt getrokken, en ik de epigenetica-kar. We zaten daarbij soms wel een beetje op ons eigen eilandje, jammer dat het geen tropisch vakantieparadijsje was. Toch heb je me enorm aangestoken met je voorliefde voor genetica, en daarmee heb je tevens aan de wieg gestaan van dit proefschrift. Er is dan ook bijna geen artikel waar je niet op staat, en terecht. Ik vind het dan ook geweldig dat je me als paranimf bij wilt staan. We gaan allebei door in de genetica, die de komende jaren alleen nog maar aan importantie zal toenemen, ik hoop daarbij in de toekomst nog eens in de gelegenheid te zijn met je samen te werken.
Joyce, Vince, Bas, Amir, Crystel, Sander, Saskia, Jessica, Jelte, Ingrid, Aisha, Ellen, Ian, Jonne, Jeroen en natuurlijk Quirina (de vreemde eend in de bijt). Wat was de toren toch leuk met jullie, met echt een goede balans tussen efficiënt werken en een gezellige werkplek. Met prachtige items als de basketball, de stinkende ijskast, de schoonmaker, de plot-wall-of-fame, de toren-app en andere gekkigheden. Daarnaast was er natuurlijk ook de gezelligheid na het werk tijdens de nodige vrijmibo’s op het ledig erf.
Sanne, Ellen, Pleunie en Geert. Ik heb echt een mooie tijd gehad met jullie op de AIO- kamer. Naast dat er altijd serieus gewerkt werd, heb ik ook veel met jullie kunnen lachen. Gezelligheid was nooit het probleem, aangezien we naast het secretariaat zaten en iedereen te pas en te onpas binnen kwam zetten. Sanne en Ellen, met jullie ben ik naar een congres in Barcelona geweest, waar ik mooie herinneringen aan heb. Geert, wat ik niet zal vergeten is de estafette Utrecht-München en hoe fanatiek jij was, wat een bijzonder evenement. Pleunie, er was niets mooier dan even dat koffie momentje ’s ochtends, dat mis ik nog steeds wel eens.
20162300 Marten Siemelink.indd 182 15-03-16 13:23 APPENDIX | 183
Zeliha en Mark, wat waren jullie goede studenten! Toen je bij mij onderzoek kwam doen, Zeliha, was ik zelf nog geen jaar bezig. Amper had ik het onderzoeksplan aan je uitgelegd of je ging er als een speer mee aan de slag, dus jou bezig houden was in den beginne nog niet zo makkelijk. Gelukkig ben je zelf iemand die echt weet van aanpakken. Al snel had je helemaal je draai gevonden op het lab en wist je wie er wat voor je kon betekenen. Je bent daar echt heel zelfstandig mee aan de slag gegaan en hebt echt je eigen project kunnen bedenken en uitvoeren. Hartstikke mooi! Mark, je hebt in je stage bij ons echt laten zien dat je een slimme vent bent. Het project dat je deed, mede begeleid door Zhiyong, was echt wel lastige materie, maar je sloeg je daar zonder veel moeite doorheen. Je hebt een hoop wetenschappelijke ervaring en skills opgedaan. Tijdens je stage heb je je zinnen gezet op de SUMMA-opleiding. Geen gemakkelijke opgave om daar binnen te komen, maar het is je glansrijk gelukt en ondertussen komt je artsenbul al snel dichterbij.
Verder zijn er nog de nodige collega’s van de experimentele cardiologie die niet onvermeld mogen blijven. Imo wat vond ik het altijd mooi als je even nonchalant bij ons op de AIO- kamer kwam buurten. Joost, als ik mijn werk moest presenteren was ik misschien wel het meest beducht voor jouw vragen, echt een scherpe geest. Arjan, Sander, Noortje, Julie, Loes en overige laboranten, ik ben jullie veel dank verschuldigd. Onder jullie begeleiding heb ik veel geleerd in het lab. Daarnaast hebben jullie ook veel werk voor mijn onderzoek verricht. Jullie zijn een belangrijke drijvende kracht achter het laboratorium. Alle overige (ex-)promovendi, collega’s en studenten bedankt voor de leuke tijd samen op het lab, bij congressen en symposia en natuurlijk bij de labuitjes, feestjes en borrels.
Alle collega’s van de Klinische Genetica in het VUmc, dank voor jullie belangstelling en vooral jullie geduld ten tijde van het schrijven van dit proefschrift. Ik vind het verfrissend om de genetica op een klinische manier te benaderen en werk dagelijks met veel plezier met jullie samen.
Vrienden van weleer: Kria, Karianne, Bram, Trix, Isa en alle aanhang en kids, wat zijn we toch een mooi clubje, nog steeds goede vrienden. Wat fijn als we elkaar even niet gezien hebben dat het meteen weer als vanouds is. Kevin, Jules, Gilbert, al jaren goede vrienden, wat hebben we altijd een mooi avonturen samen. Lon en Iwan, bedankt voor jullie goeie vriendschap en de vele gezellige momenten. Robert en Sanne, de soms spaarzame momenten met jullie waardeer ik des te meer. Rikkert, al sinds de kennismaking bij geneeskunde trokken we samen op, als vrienden al veel moois ondernomen, wat goed dat je er nu ook weer bij bent. Kirsten en Jan-Willem, mooi dat we elkaar nog steeds zo nu en dan weer opzoeken, altijd de grootste lol.
Constance en Max, bedankt dat jullie mij door de jaren heen zijn blijven steunen, ook toen het allemaal over een andere boeg ging. Jullie hebben altijd vertrouwen in mij getoond en mij gestimuleerd om mijn doelen te verwezenlijken. Zonder jullie was dit proefschrift er niet geweest. Lieve Isabel, je bent en blijft ontzettend belangrijk in mijn leven, we hebben al zoveel samen meegemaakt. Ik vind het een eer dat je mijn paranimf wilt zijn. Mary, enorm bedankt voor je aanhoudende warme betrokkenheid. Suzanne, wat fijn dat je er bij bent en wat geweldig dat je in Utrecht bent komen wonen. Ik hoop dat er nu meer tijd komt om jullie allemaal weer vaker te zien.
Lieve Judith, wat ben je geweldig enthousiast, lief en zorgzaam. Bedankt voor je onophoudelijke steun en liefde. Ik hoop dat we samen nog vele mooie momenten mogen A beleven.
20162300 Marten Siemelink.indd 183 15-03-16 13:23 184 | APPENDIX
List of Publications
Human validation of genes associated with a murine atherosclerotic phenotype. Pasterkamp G, van der Laan SW, Haitjema S, Foroughi Asl H, Siemelink M, Bezemer T, van Setten J, Dichgans M, Malik R, Schunkert H, Samani N, de Kleijn D, Worrall B, Markus H, Hoefer I, de Jager S, Björkegren J, Michoel T, den Ruijter H, Asselbergs F Accepted in ATVB. 2016
Gene-based meta-analysis of genome-wide association studies implicates new loci involved in obesity. Hägg S, Ganna A, van der Laan SW, Esko T, Pers TH, Locke AE, Berndt SI, Justice AE, Kahali B, Siemelink MA, Pasterkamp G; GIANT Consortium, Strachan DP, Speliotes EK, North KE, Loos RJ, Hirschhorn JN, Pawitan Y, Ingelsson E. Hum Mol Genet. 2015 Sep 16
Systemic glucocorticoïds are associated with mortality following carotid endarterectomy. Siemelink M, den Ruijter H, van der Valk F, de Vries JP, de Borst GJ, Moll F, Stroes E, Pasterkamp G J Cardiovasc Pharmacol. 2015 Oct; 66(4):392-8
Common variants associated with blood lipid levels do not affect carotid plaque composition. Siemelink MA, van der Laan SW, van Setten J, de Vries JP, de Borst GJ, Moll FL, den Ruijter HM, Asselbergs FW, Pasterkamp G, de Bakker PI. Atherosclerosis. 2015 Sep;242(1):351-6
Biomarkers of Coronary Artery Disease: The promise of the transcriptome. Siemelink MA, Zeller T. Curr Cardiol Rep. 2014 Aug;16(8):513.
Differential homeostatic dynamics of human regulatory T-cell subsets following neonatal thymectomy. Schadenberg AW, van den Broek T, Siemelink MA, Algra SO, de Jong PR, Jansen NJ, Prakken BJ, van Wijk F. J Allergy Clin Immunol. 2014 Jan;133(1):277-80
Taking risk prediction to the next level. Advances in biomarker research for atherosclerosis. Siemelink M, van der Laan S, Timmers L, Hoefer I, Pasterkamp G. Curr Pharm Des. 2013; 19(33):5929-42.
20162300 Marten Siemelink.indd 184 15-03-16 13:23 APPENDIX | 185
Articles in preparation
Cardiovascular risk loci associate with DNA methylation in carotid plaques Manuscript in preparation
Tobacco smoking is associated with DNA methylation in human atherosclerotic plaques Manuscript in preparation
Sex-specific differences in DNA methylation in the atherosclerotic carotid artery Manuscript in preparation
Coronary artery disease and large artery stroke loci are associated with human atherosclerotic plaque characteristics Manuscript in preparation
Associations of genetic variation with carotid plaque characteristics: the AE Exome and Genomics Studies Manuscript in preparation
X-chromosomal differences in atherosclerotic plaque methylation Manuscript in preparation
A
20162300 Marten Siemelink.indd 185 15-03-16 13:23 186 | APPENDIX
Curriculum Vitae
Marten Antoon Siemelink was born on the 21st of December 1981. After finishing secondary school in 2000 at the Sint Bonifacius College in Utrecht, he first started studying architecture in Delft where he left after one semester to pursue a more beta-oriented education. He studied biology at the University of Utrecht, where he obtained his bachelor of science in 2005. In the following period he performed an extra-curricular internship in the department of Pediatric Immunology and Pediatric Intensive-Care at the Wilhelmina Children’s Hospital, under supervision of prof. dr. Berent Prakken and dr. Alvin Schadenberg. In the meantime, he passed the entry exams for the Selective Utrecht Medical Masters (SUMMA) program, which he successfully completed in the following four years. After graduation, he worked as a non-training resident in the cardiothoracic surgery department at the OVLG hospital in Amsterdam, until he was offered a research position. In 2011 he started his PhD research in the Experimental Cardiology Laboratory at the Utrecht University Medical Center, working on the Athero-Express Biobank under supervision of prof. dr. Gerard Pasterkamp. During his PhD he worked on many projects, most notably he devised and performed the Athero-Express Methylation study, for which he has set up collaboration with the group of dr. Bas Heijmans at Leiden University Medical Center department of Molecular Epidemiology. The results of the studies described in this thesis, have been published in international scientific journals and were presented at international cardiovascular conferences. Following his work at the UMC Utrecht, he started in the fall of 2015 as a non-training resident in the department of Clinical Genetics at the VU Medical Center in Amsterdam.
20162300 Marten Siemelink.indd 186 15-03-16 13:23