DNA Microarray Analysis of Genes in Human Atrial Myocardium

Keiji Yamamoto
Jichi Medical School, Tochigi, Japan

Molecular studies in patients are needed to develop novel treatment strategies. DNA microarray analysis is a powerful technology that offers a comprehensive profile of gene expression. Work performed by this group illustrates the powerful benefits of genome informatics.

Study of human right atrium

In 9 patients undergoing cardiac surgery, most with valvular heart disease, this group investigated the transcriptional profile of genes induced in pressure-overload or volume-overload human atria.

The patients were divided into 3 groups based on pressure data and echocardiographic findings: 1) control, 2) pressure-overload (mean right atrial pressure > 7 mm Hg), 3) volume overload (moderate or severe tricuspid regurgitation). The patient characteristics are shown in Figure 1. The expression of 2,139 well-characterized functional genes, such as cell cycle protein, growth factors, and cytokines was investigated.

On DNA microarray analysis, only 4 genes were expressed: cyclin-dependent kinase inhibitor 1A (CDKI1A), MAP kinase phosphatase-1 (MKP-1), glutathione S-transferase theta 1 (GSTT1), and small inducible cytokine A5 (SCYA5) (Figure 2).

In the pressure overload group, CDKI1A and MKP-1 were significantly increased, compared to the control and volume overload groups (12-fold induction versus < 4-fold in control and volume overload for CDKI1A, and for MKP-1 a 25-fold versus < 10-fold induction in control and volume overload). In contrast, the expression of GSTT1 and SCYA5 was the same in each group.

The increased expression of CDKI1A and MKP-1 mRNA in the volume overload group was confirmed by real-time PCR. For this analysis, primers of the 4 expressed genes were designed for real-time PCR. To confirm the preparation expressions of selected genes in the pressure overload group, cDNA from the same RNA used in the microarray analysis was subjected to real-time PCR.

Analysis of MKP-1 and CDKI1A

MKP-1 has dual catalytic activity toward phosphotyrosine- and phosphothreonine-containing proteins, and is known to inactivate extracellular signaling-regulated protein kinases, possibly JNKs and p38-MAP kinases. CDKI1A functions by binding to and inactivating a number of the cyclin and cyclin-dependent kinase complexes, including cyclin-dependent kinase 2. CDKI1A is a regulator of the G1-S cell cycle checkpoint. In the adult heart, the role of the cell cycle, including CDKI1A, is unknown.

Mechanical stress induced MKP-1 and CDKI1A protein expression in cultured neonatal rat cardiomyoctyes in their experiments.

Bueno and colleagues had demonstrated a dramatic attenuation of cardiac hypertrophy in response to aortic banding in alpha-myosin heavy chain cardiac specific MKP-1 expressing transgenic mice. Therefore, these investigators made transgenic mice expressing CDKI1A under the control of an alpha-myosin heavy chain cardiac specific promoter, in collaboration with investigators at Tokyo University. In wild-type mice, abdominal aortic banding induced left ventricular hypertrophy (LVH), while in the CDKI1A transgenic mice the LV was only dilated a little after aortic banding (Figure 3). In the CDKI1A transgenic mice, attenuation of LVH development was also found. These findings suggest the importance of CDKI1A as a counter-balancing regulatory factor in the heart. The precise role of CDKI1A in cardiac hypertrophy is the subject of further study by this group.

Based on the present findings, these investigators speculate that a suppressor of cell cycle, proliferation, or cell hypertrophy may play a critical role in the pathophysiology of pressure overload (Figure 4). 

Molecular mechanisms in AF

Gene expression profiling of human atrial myocardium with AF using DNA microarray analysis was performed by this group.

A comprehensive analysis of the regulatory mechanisms involved in AF had not been performed previously, and molecular research has mainly focused on ion channels and proteins involved in calcium handling.

In 17 fresh cardiac specimens (10 sinus rhythm (SR), 7 AF) from the right atrial appendage, the expression profile of 12,000 human genes was investigated with mRNA from the atria using Human Genome U95A arrays (Figure 5). The left atrial diameter was significantly greater in the AF group than in the SR group (67 mm vs 42.7 mm; p<0.001). However, there were no other differences between the 2 groups.

The DNA microarray analysis identified 33 AF-specific genes that are activated compared to SR (Figure 6). Of these, macrophage migratory inhibitory factor and NF-IL6-beta are involved in inflammation. In contrast, they found 63 SR-specific genes. Sarcoplasmic reticulum Ca²⁺-ATPase 2 is a calcium ion regulatory protein and Connexin 43 is one of the Gap junctions that are clusters of closely packed channels.

NF-IL6-beta mRNA induction in the AF group was confirmed by real-time PCR. A novel and inflammatory mechanism may promote the initiation and persistence of AF, potentially by inducing structural and electrical remodeling of the atria.

Connexin 43 and SERCA2 expression in the AF group was significantly lower than that in the SR group. The abnormalities in inter-cellular calcium ion handling and changes of the atria in chronic AF patients may be involved in the initiation and perpetuation of AF. Changes of Connexin 43 expression may affect conduction velocity and induce and sustain persisting AF.

Summary

Novel and inflammatory mechanisms may be involved in the mechanism of AF.Decreased SERCA2 and connexin 43 mRNA expressions may contribute to the molecular mechanism of AF. Although the roles of many genes including ESTs in the heart remain unknown, the genes screened in this study may provide insights into the initiation or perpetuation of AF and the pathophysiology of atrial remodeling.

Systematic Linkage between the Genomic and Clinical Database of Cardiovascular Disease to Find New Pathophysiology or Effective Treatment

Masafumi Kitakaze
National Cardiovascular Center, Osaka, Japan

This presentation illustrated the use of 4 different strategies to elucidate candidate genes for chronic heart failure (CHF). This group states that these strategies, which use genomics or pathogenomic sciences, combined with molecular biology, gene expression, physiology, and clinical medicine, will identify the genes for certain cardiovascular diseases, and lead to novel therapies.

The 4 strategies used to identify the candidate gene for CHF amongst the 35,000 available genes, were: 1) knowledge from the existing literature, 2) identify common genes that are upregulated or downregulated in the failing myocardia in the studied species, 3) identify heart failure-related genes by linking clinical parameters to gene expression level, and 4) identify genes modulated by drugs known to be cardioprotective, such as ACE inhibitors.

DNA chip analysis was performed in failing myocardium samples obtained from 1) patients with CHF (n=20), 2) canine failing myocardium produced by rapid pacing (n=12), and 3) murine failing myocardium produced by aortic banding (n=60).

Extending literature-based knowledge to pathophysiology of CHF

In the 20 samples of human failing myocardium, of the 10,000 known genes, the expression of about 3% were significantly modulated, that is, about 300 genes were upregulated or downregulated.

The adenosine-related genes among the modulated genes in the failing myocardium were examined, because adenosine is known to be cardioprotective against ischemia/reperfusion injury. The adenosine A2a and A3 receptors are downregulated in the failing myocardium, compared with a commercially-available control myocardium, on DNA microarray analysis.

To prove their hypothesis that the downregulation of the adenosine receptor is linked to the pathophysiology of CHF, they administered the adenosine agonist 2-chloroadenosine in the murine heart failure model produced by thoracic aortic banding (TAC). TAC produced cardiac hypertrophy and 2-chloroadenosine potently inhibited cardiac hypertrophy. The TAC-increased lung edema was strongly attenuated by the adenosine analog.

Therefore, they conclude that the adenosine A2a receptor is downregulated in patients with CHF and the restoration of the adenosine receptor activation improves heart failure in the mouse model.

A difference in the SNP of the A2a receptors was found between the control and patients with dilated cardiomyopathy (DCM). This strengthens the notion that the decreases in the effect of adenosine following a genetic abnormality of the adenosine receptors may be one of the mechanisms that progresses CHF.

To test the effect of adenosine in the clinical setting, they administered the adenosine potentiator dipyradimole in patients with CHF. A 6-month administration of dipyradimole improved NYHA functional class, and increased the ejection fraction (EF) and exercise capacity. These results suggest that augmentation of the effect of adenosine improves CHF.

Identifying commonly-modulated genes

The second strategy was to find commonly-modulated genes in the human, canine, and murine failing myocardium using DNA microarray, and test whether such genes are responsible for heart failure.

In the human heart, the modulated genes were linked to clinical parameters, such as BNP, LVEF, and pulmonary arterial pressure. Of these, approximately 80 genes are only modulated in the human failing myocardium, including heparin-binding epidermal growth factor (HB-EGF). HB-EGF is upregulated in the failing canine myocardium. In the murine model, HB-EGF was closely correlated to changes in DNA expression levels.

HB-EGF is closely related to the pathophysiology of HF, based on experiments by this group. In knockout mice, the deletion of HB-EGF function did not cause systemic blood pressure changes, but provoked enlargement of ventricular diameter and decreased contractile function (Figure 1). Further, the knockout mice, but not the wild type mice, died of cardiovascular (CV) death at about 150 days (Figure 2).

Further investigation also suggested that abnormalities of HB-EGF genes may contribute to the pathophysiology of the human failing heart. Examination of the differences in SNPs related to HB-EGF and ADAM-12, which shares membrane-bound HB-EGF, between control subjects and patients with DCM revealed that the SNPs at exon 3 and exon 5 were found in patients with DCM.

Linkage analysis between gene expression and clinical parameters

The correlation between the change in gene expression and plasma BNP levels, EF or pulmonary arterial (PA) pressure was examined. They found 212 genes tightly and positively linked to PA pressure, and 15 genes were inversely linked to changes in PA pressure.

Next, they investigated if there was a relation between the genes related to drugs used to treat CHF and the genes related to PA pressure. The beta-1 adrenergic receptor, endothelin receptor A, endothelin converting enzyme, and aldosterone receptor are tightly linked to CHF, and this analysis showed these genes are also well-correlated to the PA pressure changes (Figure 3). Therefore, they believe this strategy works for mining for novel genes for heart failure.

They further classified the CHF-related genes according to the cellular function. Genes related to channels, phosphodiesterase, matrix metalloproteinase, and collagen were tightly linked to the PA pressure (Figure 4). This suggests that networks of these gene changes contribute to the pathophysiology of heart failure.

The PA pressure-related genes were examined based on gene onthology, and showed that many unexpected genes are in the groups of striated muscle contraction, nucleobase, nucleoside, and nucleotide, nucleic acid metabolism, and protein modification (Figure 5). These genes are tightly related to PA pressure.

The same genes are modulated when linking was performed to ejection fraction. Among these genes, a second screening was performed to determine the relation to cellular survival or death. They transfected the candidate genes for CHF to the rat cardiomyocyte and examined the amino acid uptake and cell survival in this system. The X2 genes decreased amino acid uptake, and X5 genes are increased in cell survival. This method verified that X3 and X5 genes are crucial for cell survival. Knockout mice with these genes are being developed for further study.

Haplotype analysis revealed that the SNP pattern of 9 genes is significantly different. These genes are tightly related to CHF and testing these genes is worthwhile to determine whether they are related to the pathophysiology of CHF.

Identify genes modulated by drugs

In myocardial samples from failing canine myocardium, of the 20,000 genes examined by DNA microarray, 2,005 genes had active significant signals as assessed by Cy3 or Cy5. Among these genes, 50 genes were upregulated and 25 genes were downregulated—and several of these were potently modulated by the ACE inhibitor temocapril. ACE inhibitors are known to be cardioprotective. Further testing of these genes is worthwhile.

HB-EGF was found to be upregulated by ACE inhibitors and ARBs compared to the failing myocardium without any drugs. These results also suggest that HB-EGF plays a role in the pathophysiology of HF.

Establishment of Clinical Data Management System and Its Practical Application to Genetic Epidemiology in Cardiovascular Medicine

Tsutomu Yamazaki
University of Tokyo, Tokyo, Japan

The trinity of genome information, clinical information, and information technology is indispensable to research in the field of genome informatics in cardiovascular disease (CV). Multivariate analysis using genetic and environmental factors is necessary to investigate the mechanisms of CV disease. The gene-environment interaction also influences the susceptibility of the genesis of CV disease. Thus, a combination approach to genetic and clinical factors has been used by these investigators.

Database development

Comprehensive clinical information from about 3,000 patients admitted to the University of Tokyo has been collected in their clinical database. DNA samples from 1,300 patients who provided signed informed consent are also available for SNP analysis. The genetic study conducted by this group was approved by the Ethics Committee of the University of Tokyo.

More than 500 clinical parameters, such as basic characteristics, diagnosis, risk factors, noninvasive examination, coronary angiography, percutaneous intervention, bypass surgery, cardiac events, complications, and medications, are stored in the clinical database. Analysis of about 50 SNPS known to be implicated in the pathogenesis of CV disease was performed, such as ACE, angiotensinogen, and angiotensin II type 1 receptor.

The Bioluminometric Assay coupled with Modified Primer Extension Reactions (BAMPER) assay was introduced by this group of investigators for analyzing human genetic variations (Figure 1). After PCR, probes are added and pyrophosporic acid is produced, which is converted to ATP, to analyze the luciferase reaction system.

Figure 2 illustrates the chronological sequence of the clinical information in one patient, such as clinical symptoms, hospitalization, complications, noninvasive testing, intervention, and prescriptions. Detailed data on any item can be accessed easily by clicking on any item of interest. Variations in total cholesterol can be investigated by arranging the levels in a side-by-side manner (Figure 3). This analytical method provides for easily conducting epidemiological studies, including prospective and transversal studies.

Medical treatment records of patients at the University of Tokyo are maintained in an electronic database system, which provides the ability to perform chronological reference and prospective and transversal studies and analysis of medical treatment data mainly through data mining.

Data from coronary artery quantitative analysis are included in the database. The QCA data is automatically transferred to the database. Figure 4 illustrates a chronological view of the QCA data in one patient, providing an easy overview of the patient status, including the progression or regression of disease in a vessel.

For a transversal study, the patients in the database were divided into 2 groups based on the presence or absence of diabetes. The ratio of patients who developed restenosis can be easily determined and these findings can be automatically shown in real-time (Figure 5).

Data mining is an exhaustive method that is generally accepted and used in recent years.
An efficient search of a large database can be easily performed using specified criteria.
The table in Figure 6 illustrates the use of data mining. About 81% of the patients with the specified premise had a myocardial infarction (MI). Support means the ratio of patients who have both the premise and conclusion. The odds ratio shows the patients who have the premise and are susceptible to MI, about 4.7 times, compared to the patients who don’t have this premise.

Examples from their database system

Cholesterol esters are transferred via the ABCA1 protein and bind to the lipid-poor Apo-A1 particles. Therefore, a variation in the ABCA1 gene, which regulates the cholesterol influx from the cell, may modify the protein function leading to the alteration of the lipid profiles, such as HDL cholesterol and Apo-A1 levels. So, this group investigated whether a common variant of the ABCA1 gene at the position of Ile823Met, may influence the plasma lipid profiles, including HDL-C, and the susceptibility to coronary artery disease (CAD) in the Japanese population in their database.

In 292 patients untreated for hyperlipidemia, they found significantly higher HDL-C levels in patients with Met/Met (49.4 mg/dl, p<0.05 vs Ile/Ile patients) compared to patients with Ile/Ile (45.7 mg/dl) and Ile/Met (49.8 mg/dl).

The correlation between the Met allele and CAD was then analyzed. Distribution of the Met allele was about 31% and 42% with or without CAD, respectively. This suggests that the Met allele is associated with significantly higher levels of HDL-C, resulting in the inhibition of CAD. On logistic regression analysis, age >65 years, hypertension, diabetes, hyperlipidemia, and ABCA1 gene polymorphism (odds ratio 1.54) were independent risk factors for CAD (Figure 7).

Adiponectin was recently shown to have anti-atherogenic and anti-diabetic effects. Adiponectin, structurally regulated by collagen, has been identified as a novel adipose-specific gene product. The polymorphism at position 276 of adiponectin was analyzed by this group using ELISA. The G allele at position 276, thought to be associated with diabetes or insulin resistance, was conversely and linearly related to plasma adiponectin levels in the Japanese patients in the clinical database.

Next, the severity of CAD and adiponectin polymorphism was evaluated. The G allele at position 276 was significantly associated with the severity of CAD only in obese patients, but not in non-obese patients. However, other genetic polymorphisms implicated in insulin resistance, such as IRS1, IRS2, PPAR-gamma, and the leptin receptor, had no association with the severity of CAD (Figure 8).

Conclusion

The original database system established at this institution supports genetic epidemiologic studies and clinical investigation, including prospective cohort study, clinical trials, and medical research, especially on the adverse reactions of new drugs. This system has great potential to elucidate the susceptibility for CAD and previously unknown information, which should contribute to personalized medicine and novel therapeutics and new drug design for CV disease.         

Pharmacogenomics and Pharamacoinformatics

Toshio Tanaka
Mie University School of Medicine, Tsu, Japan

The goals of pharmacogenomics are to identify gene clusters involved in determining the responsiveness to a given drug, to distinguish responders and nonresponders, and to predict side effects. This presentation focused on drugability and drugable genomics.

The principle of molecular pharmacology is reductionist, whereas the principle of pharmacogenomics is comprehensive and individualized. Studying drugability using molecular pharmacology provides information on the interaction between medicine and the drug target, and some information on disease associated genes. However, there is a large gap between the therapeutic gene cluster and disease-gene clusters.

Studies of cerebrovascular spasm and subarachnoid hemmorhage

Drugability can now be researched using pharmacogenomics. Therapeutic gene clusters, disease gene clusters, and the interaction between these clusters can be exploited to find potential therapeutic actions.

This group constructed an original pharmacogenomic database, in which drug action or disease information can be translated into gene clusters. This enables comparing the relation between therapeutics and disease in the common genomic database. The database comprises 3 dictionaries: gene, chemical, and disease.

To study cerebral vasospasm, this group used their pharmacogenomic database. Cerebral vasospasm can be defined as a delayed onset narrowing of the cerebral artery, which can occur after subarachnoid hemorrhage (SAH). The pathophysiology of cerebral vasospasm remains unknown.

Transcription analysis of the differences in gene expression on H-E staining of the basilar artery in their rat model revealed that encoding for heme oxygenase 1 is highly upregulated by basilar artery by delayed cerebral vasospasm.

Three genes (NOS1, ICAM1, EDM3) are related to both HO1 and cerebrovascular disorders. Thus, they tried to determine the mechanism of action of HO-1 in cerebrovascular disorders. Antisense for HO-1 ODN in vivo inhibited HO-1 induction in basilar artery and in deteriorated cerebral vasospasm. This phenomenon was not observed in the scrambled ODN. This data suggested that HO-1 has a potential protective function in cerebrovascular spasm.

Administration of the potent HO-1 inducer hemin significantly induced HO-1 and significantly improved the diameter of the basilar artery by angiography. This result suggests HO-1 induction is protective of disease.

Other experiments by this group showed that the low molecular weight (LMW) chemical compound AVS selectively induced HO-1 in the basilar artery, only in the presence of SAH. A single administration of AVS in vivo did not induce the basilar artery. In their model, AVS has a potential therapeutic action on cerebrovascular spasm measured by angiography. AVS is being studied in clinical trials, with one report of its efficacy for cerebrovascular spasm in a small number of patients. 

In other work by this group, they found that heat-shock protein (HSP) 72 was greatly induced in SAH. Using their database to check the relationship between HSP72-related genes and cerebrovascular disease, they identified 1 gene related to HSP72 and cerebrovascular disorders. To characterize the function and significance of HSP72 in cerebrovascular vasospasm, they used antisense ODN, which is effective in suppressing HSP72 message induction and inhibiting HSP72 induction at the protein level. Antisense induction was protective of function in this disease.

Oral GGA, another low molecular weight chemical, dose-dependently induced HSP72 in the basilar artery. GGA can induce HSP72 at the messenger level in the presence of SAH. In the rat model, GGA administration significantly improved cerebrovascular spasm, especially day 1 to day 3.