Skip Navigation

European Journal of Heart Failure 2002 4(6):687-697; doi:10.1016/S1388-9842(02)00169-1
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Juhasz, O.
Right arrow Articles by Boheler, K. R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Juhasz, O.
Right arrow Articles by Boheler, K. R.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

© 2002 European Society of Cardiology

Analysis of altered genomic expression profiles in the senescent and diseased myocardium using cDNA microarrays

Ondrej Juhasz, Yi Zhu, Rahul Garg, Sergey V. Anisimov and Kenneth R. Boheler*

Laboratory of Cardiovascular Science, Gerontology Research Center, National Institute on Aging, National Institutes of Health 5600 Nathan Shock Drive, Baltimore, MD 21224, USA

* Corresponding author. Tel.: +1-410-558-8095; fax: +1-410-558-8150. E-mail address: bohelerk{at}grc.nia.nih.gov


   Abstract
Top
 Abstract
1. Introduction
 2. What is the...
 3. Towards the development...
 4. Application of DNA...
 5. Data interpretation
 6. Limitations of gene...
 7. Ideal arrays
 8. Conclusions
 References
 
Cardiac function deteriorates with aging or disease. Short term, any changes in heart function may be beneficial, but long term the alterations are often detrimental. At a molecular level, functional adaptations involve quantitative and qualitative changes in gene expression. Analysis of all the RNA transcripts present in a cell's population (transcriptome) offers unprecedented opportunities to map these transitions. Microarrays (chips), capable of evaluating thousands of transcripts in one assay, are ideal for transcriptome analyses. Gene expression profiling provides information about the dynamics of total genome expression in response to environmental changes and may point to candidate genes responsible for the cascade of events that result in disease or are a consequence of aging. The aim of this review is to describe how comparisons of cellular transcriptomes by cDNA array based techniques provide information about the dynamics of total gene expression, and how the results can be applied to the study of cardiovascular disease and aging.

Key Words: Cardiovascular • Heart • Disease • Aging • Microarrays • Chips • Genomics

Received September 28, 2001; Revised February 21, 2002; Accepted May 1, 2002


   1. Introduction
Top
 Abstract
 1. Introduction
2. What is the...
 3. Towards the development...
 4. Application of DNA...
 5. Data interpretation
 6. Limitations of gene...
 7. Ideal arrays
 8. Conclusions
 References
 
Modern medical advances have steadily increased the average human life expectancy. Aging is, however, associated with a higher incidence of disorders like diabetes, Alzheimer's disease, different types of cancer as well as cardiovascular diseases [1–10]. Cardiovascular diseases (atherosclerosis, hypertension, stroke and heart failure), in particular, continue to be the major cause of morbidity and mortality in industrialized societies. Hypertension affects 25% of the adult population of industrialized societies [11], while heart failure is a world-wide public health problem that causes approximately 300 000 deaths annually in the US [12]. Cardiovascular diseases reach epidemic proportions in the very old (≥80 years of age), such that more than 10% of them will be afflicted with some form of cardiovascular disease [9,13].

Molecular genetic studies have identified a number of genes that cause particular cardiovascular diseases. For example, all Mendelian forms of hypertension solved to date are linked to altered net renal salt balance, causing elevated blood pressure [14]. Mutant gene products associated with the sarcomere cause familial hypertrophic cardiomyopathy (FHC). Similarly, several genetic loci are known to be associated with dilated cardiomyopathies, yet only a few of the responsible genes have been identified [15]. Despite their important roles as a cause of disorder, our current understanding of the molecular and cellular events responsible for the development of cardiovascular disease with aging is limited [16,17].

Because gene expression is a function of development, life-style, aging, disease and the genetic makeup of an individual [18], it has been postulated that a complete and simultaneous analysis of gene expression would lead to important insights into the mechanisms responsible for disease. The aim of this review is to describe how gene expression arrays of myocardial transcriptomes (all the RNA transcripts present in a cell population) provide information about the dynamics of total genome expression. Arrays permit a comprehensive analysis of quantitative and qualititative changes in RNA transcript abundance, and the results provide a snapshot of altered patterns of gene expression in response to any genetic or environmental stimulus. Specifically in this review we will describe the use of cDNA arrays, discuss their limitations and indicate how expression profiling can be applied to the study of cardiovascular diseases or aging.


   2. What is the relevance of the Genome Project to disease?
Top
 Abstract
 1. Introduction
 2. What is the...
3. Towards the development...
 4. Application of DNA...
 5. Data interpretation
 6. Limitations of gene...
 7. Ideal arrays
 8. Conclusions
 References
 
The goal of the Genome Project was to sequence the entire genome of several species to understand the function and regulation of genes. In the past few years, we have gone from the first complete genome sequence of a free-living organism to a working draft of the human genome published recently by two groups: the public genome consortium [19] and Celera genomics [20]. Surprisingly, the number of genes (30 000–40 000) predicted from these projects was much lower than previous predictions (45 000–140 000), although recent analyses suggest that the ~30 000 genes may be an underestimation [21]. The Genome Project has thus provided us with clones and sequence data across the genome, with a full catalogue of genes. It thus can be argued that we have now entered the ‘post-genome era’ where it is possible to examine the entire RNA population or transcriptome of cells under various physiological and pathophysiological conditions. Currently two techniques hold the potential to accurately and quickly analyze the cell's transcriptome: Serial Analysis of Gene Expression [17]; and DNA arrays or microarrays.

An array, just as the word describes, is a systematic arrangement of objects. In the case of DNA arrays, the arrangement usually consists of cDNA clones or oligonucleotides spotted at high density on a membrane, specially coated glass slide or other appropriate material [22–24]. Specifically, arrays are tools that when used in hybridization can assay small amounts of DNA or RNA (Fig. 1). DNA arrays can be loosely divided into two groups: genotyping and gene expression arrays. Genotyping arrays typically use short synthetic oligonucleotides to examine the primary structure of DNA probes to confirm the presence of a gene(s) in the genome or identify mutations, genetic variants (SNP) [25], isoforms, or the distribution of gene products. Its use can facilitate the diagnosis of a disease for which a gene mutation has been previously identified. Gene expression arrays are generally composed of either oligonucleotides or cDNA clones, depending on the source. The probe for a gene expression array is usually generated from total RNA (mRNA) and reflects the actual gene expression profile of the cell/tissue examined (Fig. 1). Quantitative differences in expression profiles reflect the cell's transcriptome in response to an environmental or genetic factor. The uses of genomic techniques that can rapidly examine the entire transcriptome portend a new era where gene expression profiles, on a global basis, may be instrumental in determining the molecular basis of disease.


Figure 1
View larger version (29K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 1 Array methodology. Using standard hybridization protocols (top), probes are labeled with radionucleotides, hybridized to an array, exposed to a storage phosphor screen or film and scanned. The signal intensity of an element is a measure of its abundance. To compare the changes in abundance, two arrays must be hybridized with labeled (radioactive or fluorescent) probes of different origin. For competitive hybridizations (bottom), probes are prepared from two mRNA sources and fluorescently labeled, typically with Cy3 and Cy5 labeled dNTPs. The fluorescent probes are mixed in equimolar amounts and hybridized to the array. Once washed, the array is examined for fluorescent signals by scanning confocal microscopy. After image acquisition, fluorescence intensities are corrected because the detection efficiency of the fluorescent labels differs. Comparison of the corrected values (relative intensity of Cy3/Cy5 signal) is a reliable measure of the abundance of specific mRNAs on each element. Because the deposited cDNAs on the array are heterogeneous sequences, a hybridization condition that is optimal for every element is impossible to achieve. Irregularities in probe preparation and hybridization can greatly affect the fidelity of the final data set. Abbreviations: RT: reverse transcriptase labeling reaction; Ctl: control sample; Exp: experimental sample.

 

   3. Towards the development of a cardiac array
Top
 Abstract
 1. Introduction
 2. What is the...
 3. Towards the development...
4. Application of DNA...
 5. Data interpretation
 6. Limitations of gene...
 7. Ideal arrays
 8. Conclusions
 References
 
The first attempt at generating a genome-based resource for the cardiovascular system was performed by the group of Hwang et al. [26]. In 1997, they completed and published a large profile of cDNAs obtained from 13 cardiovascular cDNA libraries, containing a total of 84 904 expressed sequence tags (ESTs—sequences of cloned DNA derived from reverse transcribed mRNA). This number has rapidly increased since that time [27]. The data represent a compendium of cardiovascular genes and an impressive collection of cardiovascular gene expression information. Included in the initial compendium were genes that matched known genes (55%), matched ESTs with sequence similarity to known genes (33%), or did not match any known sequence (12%). Many of the transcripts corresponding to known sequences were nuclear-encoded, mitochondrial, ribosomal, or corresponded to repetitive elements. In one of the first attempts to study the cardiovascular system in the ‘post-genomic era’, Hwang et al. also performed in silico Northern blots (computer based comparisons of EST expression profiles among cDNA libraries) to identify gene products with greater tissue-specificity [26]. They identified 48 gene products with expression levels elevated in cDNA libraries generated from hypertrophied hearts, which were verified and expanded by subsequent analyses [27]. As stated in the original manuscript, the authors ‘demonstrate the untapped potential of genome research for investigating questions related to cardiovascular biology’. They go on and say that this represents ‘a first-generation genome-based resource for molecular cardiovascular medicine’ [26].

An important finding from this early genomic sequencing work and in silico analysis is that the majority of transcripts are not tissue-restricted, but are present to varying degrees in a wide variety of tissues. There are of course exceptions like {alpha}-myosin heavy chain, which is primarily found in heart, but most transcripts are widely distributed. The number of cDNA clones available continues to increase from many sources. As such, it is not necessary to create arrays only from cDNAs obtained from cardiovascular libraries. All available cDNA clones (or appropriate oligonucleotides) can be spotted onto appropriate support substrates and used for screening purposes. While not all of the spots (cDNAs) will hybridize, a large percentage of the DNAs will hybridize, indicating expression in the tissue of study (see Fig. 1). There are, therefore, many sources of clones that can be used for array analyses. These include arrays and microarrays that can be purchased from companies (e.g. Incyte, Affymatrix, among others), which consist of cDNA clones or oligonucleotides that cover a large percentage of the transcripts (known and ESTs) present in public databases. Other sources are publicly available. At the National Institutes on Aging, the Laboratory of Genetics has, for example, prepared an array consisting of cDNA clones obtained from a large number of mouse developmental libraries that represents a unique set of clones particularly relevant to the identification of developmental and signaling pathways [28]. The ability to use arrays to study the expression profiles of the cardiovascular system under a wide variety of conditions is, therefore, readily available. As cDNA sequences are shown to be present in the myocardium, specialty cardiac arrays ultimately will be developed. These may be particularly useful in the study of adult myocardial disease states, but may initially be of limited value for developmental studies where the dynamics of cardiac gene expression are poorly understood.


   4. Application of DNA arrays to the myocardium
Top
 Abstract
 1. Introduction
 2. What is the...
 3. Towards the development...
 4. Application of DNA...
5. Data interpretation
 6. Limitations of gene...
 7. Ideal arrays
 8. Conclusions
 References
 
The ideal gene expression array should contain elements (spotted DNA sequences) that correspond to every gene expressed in a given cell type, tissue or organ. The DNA elements should all be unique in sequence and preferably in size, to minimize cross-hybridization of probes and differences in hybridization conditions (see Section 6). This is currently not possible with cDNA arrays, where the sequence lengths vary considerably between elements. The human and mouse genomes have been fully sequenced, but the presence of isoforms, splicing variants, gene duplications and conserved sequences among disparate transcripts, indicates that not all of the genes and gene transcripts have been identified. The development of ideal comprehensive array, therefore, must await the completion of the genome project when all possible transcripts have been identified, sequenced and analyzed to generate unique DNA sequences of fixed length.

The ideal gene expression array experiments would also utilize labeled probes originating from a single cell type under control conditions and following specific mutational or environmental challenges, and not from a tissue or organ that contains a heterogeneous cell population. Cardiomyocyte cell lines that are stable in culture long-term would, therefore, represent an excellent source of cells, but even though several cardiomyocyte-like cell lines exist, no ‘normal’ cardiomyocyte cell lines have been established [29,30]. Alternatively, primary cardiomyocytes [31,32] could be used, but such cultures invariably contain some contaminating non-myocytic cells and the process of isolation alters gene expression. Additionally, cell cultures are not always consistent between laboratories (differences in culturing conditions or reagents, isolation techniques, etc.); therefore, potential changes in gene expression might be due to genetic or culturing anomalies associated with these cells.

In the absence of appropriate cell cultures, animal models (mouse and rat in particular) represent a viable alternative. Samples taken from ventricles, atria, or entire heart are, however, heterogeneous and will potentially consist of a number of muscle cells (pacemaker, His-Purkinje, atrial, and ventricular cardiomyocytes and smooth muscle cells) and non-muscle cells (endothelial, epithelial, neuronal, fibroblasts, etc.), making identification of the transcript source difficult. Although the cell population is heterogeneous, the ability to manipulate animal models genetically, pharmacologically, or environmentally, under rigorous experimental conditions can, however, lead to the identification of genes with altered expression due to a specific cause. The putative identification of a ‘regulated’ gene may also be a consequence of a number of independent variables (age, gender, housing conditions, diet, circadian rhythms, stress). All data must, therefore, be verified, linked to specific cell types and where appropriate tested according to Koch's postulates.

Examination of human heart samples presents even more challenges. Most of the samples that can be acquired come from: biopsies (fresh, frozen, or fixed); hearts rejected for transplantation but maintained in an artificial environment; or explants of diseased myocardium from patients who may have been under long-term care with multiple pharmacological interventions. Low abundance transcripts may not be detectable in small biopsies and the relative contribution to the transcriptome of specific cells in a heterogeneous cell population is difficult to interpret. The results may, however, have biological implications since disease or aging of a tissue or organ can be understood only in the context of heterogeneous cell population. Such comparisons are, therefore, complicated and require rigorous post-array testing. Ultimately, simpler systems employing animal models or cell culture are necessary to determine how or why the expression of an individual gene may be altered.

Application of gene expression arrays to the cardiovascular system should, therefore, be thought of initially as a screening tool to quantify the expression of gene products in cells under control and experimental (diseased, mutant, senescence) conditions [26,33–36]. Once accomplished, patterns and functional correlates can be formed and potential disease-related genes identified for subsequent analysis. The major goals of early array analysis are: (1) to identify which transcripts are present in sufficient quantities to give reproducible signals on the array; and (2) to compare expression levels between tissues under different experimental conditions, or various disease states. This approach has the advantage of being relatively global but suffers from a lack of immediate conclusive data [37]. The initial experiments have thus been criticized as being ‘fishing expeditions’; however, by performing expression-profiling experiments many times with various samples, expression clusters of co-regulated genes may emerge from the data set, leading to hypotheses of function [37]. Once defined, more specific experiments can be designed to address hypotheses developed from and based on global gene expression profiles.

To illustrate the use of this system, we have performed a series of hybridizations with an Incyte based cDNA microarray [38] to examine transcriptomes of aging rodent myocardium [39]. Using an array with 8734 unique elements and fluorescent (Cy3 and Cy5)-labeled cDNA probes, we first examined the signal reproducibility and compared the results from a replicate hybridization utilizing one RNA sample obtained from 12-month-old rat hearts. Fig. 2a illustrates the linear signal distribution of the fluorescent signals. The relative expression of individually labeled transcripts was nearly identical for both probes (R2=0.99). Failure to maintain linearity indicates that at least one step in the procedure had failed (Fig. 2b).


Figure 2
View larger version (35K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 2 Identification of candidate genes by competitive hybridization assay using Cy3 (Ctl) and Cy5 (Exp) labeled probes. Processed and normalized data are typically presented as scatter plots, permitting a simple evaluation of the probe quality, hybridization and detection efficiencies. It also may indicate gene products that are differentially expressed. Statistically, out of 10 000 spotted cDNAs a total of 500 or 100 false-positives would be expected for a P value of <0.05 or <0.01, respectively—one reason why the initial hybridizations should be used primarily for screening purposes. In this example, each dot on the scatter plot represents the ratio of normalized Cy3/Cy5 intensities of the gene from one array. (a) Results of an aging rodent model hybridization experiment. Two sets of data were generated from one array: raw signal data for each probe (Cy3 and Cy5 fluorescence). The competitive hybridization signals were analyzed with GEM TOOLS 2.4.2 to generate a ratio-based value of Balanced Differentiation Expression (BDE). This value takes into account the intrinsic difference in fluorescence intensities between Cy3 and Cy5 labeled DNA [38]. The signal intensities cover a wide range of values from several hundred to several thousand. A BDE value >1 indicates an mRNA abundance higher in control samples, whereas, a BDE <1 indicates an abundance higher in the experimental group. It has been reported that arrays might detect as little as 1.5-fold changes in expression [66]. An absolute BDE >1.7 may imply a significant change in expression, however the potential significance depends on the relative signal strength. The majority of the expression ratios are centered near the unity line of 1; however, some like ANP clearly show increase in expression with aging. Ratios present in the proximity of the ±2 ratio lanes indicate genes of potential importance for further study. The signal intensities in this experiment having expression ratios more than ±10 are yeast cDNA's used for control purposes. These are expected to have strong fluorescent signal intensities and are used for quality control of labeling (Cy3 and Cy5) and hybridization steps. (b) Example of a different hybridization with poorly labeled Cy3 and Cy5 probes. This is represented by overall low fluorescent signal intensities concentrated between 100 and 1000 on the graph. Secondly, the data are skewed in the direction of the abscissa, which may indicate inefficient Cy5-dNTP probe labeling or incorrect calibration of Cy5 channel. Data such as these should not be used in subsequent analyses.

 
It is important to realize that large scale hybridizations have an inherent probability of generating false positives. For a P-value equal to 0.05, one out of 20 elements would just by chance be statistically significant. From the 8734 elements in a single hybridization, we therefore expected ~437 false positives; however, 7.1% or 624 elements indicated potentially significant differences in expression (fluorescence signal Cy5 to Cy3 ratios of >1.7). In these experiments, 99.0 and 94.5% of the elements passed the minimum hybridization criteria: minimal signal to background ratio ≥2.5 and a minimal fluorescence area ≥40% of the element area; however, 482 of the elements failed to meet the minimum hybridization criteria and could be excluded from further consideration. It is probable that some cDNA clones had been inadequately spotted (no cDNA, cDNA improperly amplified or denatured) or the hybridization signal was too low to be reliable. The remaining 142 elements showed potential significant changes in gene expression, which may be indicative of individual variation between sample labeling or fluorescence.

Since the number of potential false positives in the control data set just described was low, we proceeded to examine the aging Wistar rat myocardium. To further minimize any potential false positives, only fluorescence signals that had at least a twofold difference in expression were accepted as potential and authentic changes in gene expression. Using these criteria, some 158 (1.8%) elements showed age-dependent alterations, most of which occurred at 30 months of age. Of these, 52 were known, 31 showed identity to ESTs with sequence homologies to known genes and 75 were public domain ESTs. In addition to known aging-related gene products, several novel aging-associated gene products were identified (Fig. 3). These results have led to the hypothesis that these transcripts are potential markers of aging in rat myocardium. Our laboratory reported similar findings in two other models examining the effect of the pineal gland hormone melatonin and geroprotective peptides in mouse hearts [40,41]. Follow-up analyses, utilizing quantitative PCR techniques and statistics on a larger set of myocardial and non-myocardial samples with aging and pharmacological intervention are now being used to confirm these initial screening results and determine whether any are aging-dependent. Data such as these illustrate how gene products can be identified as potential candidate genes for further investigation (see Section 5).


Figure 3
View larger version (22K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 3 Graphical representation of gene expression patterns. Plot of the expression ratios (BDE) as a function of aging provides an opportunity to visualize distinct expression features. For illustration purposes, two groups of genes, up- and down-regulated in senescent rodents, are presented but these data should only be considered preliminary. As described under Section 5, we suggest that a two-step methodology is preferable, and that array data not be relied upon exclusively for data interpretations. Most array data, particularly from animal models or human samples, should be used initially for screening purposes and subsequently, the data should be verified by independent techniques or using a smaller focused array that can be assayed until adequate statistical analyses can be employed. In this graph, genes were sorted according the similarity of their expression patterns during aging. Gene names are the standard genetic abbreviations and can be found at http://www.ncbi.nlm.nih.gov/UniGene/. (n=2 for each point).

 
We have also compared the cardiac transcriptome from biopsies obtained from ‘healthy’ (control) and heart failure (idiopathic dilated cardiomyopathy—IDCM) donors [42]. As discussed earlier, analysis of human samples, particularly with disease, is less than ideal. To illustrate this data, we used a cluster analysis with color-coding to show visually relative changes in gene abundance [43]. This system (Fig. 4A) arranges genes according to similarity in the pattern of gene abundance and not on the magnitude of signal strength. The dot product of two normalized vectors (i.e. standard correlation coefficient) is a measure of similarity in the behavior of two or more genes between experiments and conforms well to the intuitive biological notion of ‘coexpression’. Other methods of clustering have also been described [35,36,44], which allow application of more than one data mining and visualization technique to a particular data set to best explore the relationships in the data expression patterns [45,46]. Our human data set were analyzed by Pairwise average-linked cluster analysis. This is a form of hierarchical clustering where the relationship among objects (genes) is represented by a tree. When a graphical output is displayed, it conveys the clustering as well as the underlying expression pattern (Fig. 4A). Like Eisen et al. [43] who found in yeast that such clustering efficiently groups genes of similar function, we found a tendency to group genes with a similar function (ribosomal) in the human data set (Fig. 4B). Thus, patterns seen in cluster analysis of expression microarrays may reflect the status of cellular processes in end-stage IDCM. As in the aging results just reported, results such as these form the basis of testable hypotheses that must be followed up by independent analyses.


Figure 4
View larger version (65K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 4 Cluster analysis of array data. (A) For easy visualization and comparison, each column represents a single experiment and each row the expression profile of the particular gene. The color of the element reflects its relative expression value (red, decreased; green, increased). The color intensity is an indicator of the relative expression value: brighter shades are more differentially expressed. The left portion on the graph is a tree view of the clustering. Single expression profiles are joined to form groups until the last element is included and a hierarchical tree is developed. The length of the branch indicates similarity between the two groups. (B) Pairwise average-linked cluster analysis of human Idiopathic Dilated Cardiomyopathy (IDCM) data. Expression levels of several selected gene products (horizontal rows) are presented from multiple hybridizations with different samples (vertical columns). In this figure, one small cluster indicates how some related gene products (ribosomal) cluster naturally, potentially indicating coordinated regulation.

 

   5. Data interpretation
Top
 Abstract
 1. Introduction
 2. What is the...
 3. Towards the development...
 4. Application of DNA...
 5. Data interpretation
6. Limitations of gene...
 7. Ideal arrays
 8. Conclusions
 References
 
Interpreting the data from a microarray experiment is very challenging. Information generated from cDNA arrays and comparisons between species or with disease states are informatics-intensive [require high-throughput data management, statistical and non-statistical analyses, data mining via UniGene, GenBank (NR, EST) and data dissemination]. A natural choice is to calculate the ratio of measured signal intensities between test and control hybridizations, as done in the aging studies described above (see Figs. 2 and 3). The method is fast but accompanied by a systematic bias. A ratio-based analysis merely provides a list of mRNAs whose abundance is potentially altered by experimental conditions. Without formal significance testing, these lists of data may include a large number of false positives [47]. The number of false positives, however, can be partially alleviated by performing replicate experiments [48,49].

An alternate and preferable approach to ratio analyses is a two-step methodology, where the raw data are first normalized and, then, a statistical based method is used to demonstrate the soundness of an observation. Specifically, arrays should be used initially in a screening assay (as described above). A Poisson distribution or a Z-transformation [49] that adjusts the data in proportion to the standard error is an accepted way of normalizing the array data. These methods are based on the assumption that most of the signals (transcript abundance) are not altered in response to a given condition. The {chi}2 contingency test is not suitable for simultaneous analysis of thousands of data sets. The Bonferroni correction applied to normalized data also lowers the significance threshold and eliminates the majority of false positives, but an inherently rare mRNA with weak signal intensities also will be eliminated [50]. Given a sufficient number of hybridizations, a Student's t-test or ANOVA then can be used to determine whether the means in sample populations are significantly different. From this reduced data set, it should be possible to identify a group of gene targets, which can be hypothesis tested for significance on a larger number of samples. This is best accomplished by the rapid analysis of smaller arrays containing the genes of interest or through the use of independent techniques (SAGE tag counting methods, blot analyses, Quantitative-PCR).

When the sample size is sufficiently large and a two-step process is followed, the data have a high degree of reliability. Such an approach has been successful in identifying genes targeted by specific drugs in yeast [44], associated with hematopoietic differentiation in specific cell lines [45], or implicated in aging rodent and diseased human myocardium [39,42]. Several other applications of the procedure have been discussed elsewhere [51]. There are, however, no perfect methods to analyze data from arrays, but by planning the experiment to minimize any systematic bias and applying suitable statistical methods to the results, arrays can be an excellent way to generate meaningful gene expression profiles on a large scale that can be independently confirmed.


   6. Limitations of gene expression arrays in cardiovascular research
Top
 Abstract
 1. Introduction
 2. What is the...
 3. Towards the development...
 4. Application of DNA...
 5. Data interpretation
 6. Limitations of gene...
7. Ideal arrays
 8. Conclusions
 References
 
Investigation of heart tissue is complicated by factors such as tissue heterogeneity, genetic variability, disease state and pharmacological intervention. To overcome these difficulties, screening of large numbers of unmatched samples or the investigation of specific animal models of cardiac diseases is required. The two-step methodology offers a fast and sensitive approach to identify genes with altered abundance. There are several key factors that will determine the outcome of the expression profiling experiments. First of all, the definition of the control (normal) and experimental (diseased, aged, treated) tissue must be considered. It will be invaluable to profile a range of tissues of both groups to distinguish between expressions that are relevant to the disease process and those reflecting the biological variation of the normal tissue. Second, the limits of sensitivity and measurable expression differences of individual mRNA must be determined. In theory, chips can detect 1 in 250 000 mRNA copies [52,53]. Even though sensitivity may be high, rare transcripts or transcripts present in only small populations of cells may not be detected. Similarly, in the heterogeneous cell population (like heart tissue) a change in abundance may only be detected if the change occurs in the predominant cells of the tissue. Third, variability in array data means that subtle changes in experimental conditions may significantly alter the array results. It is well established that DNA–DNA hybridization efficiencies are strongly affected by probe and target sequence lengths and complexity, and by parameters such as hybridization times, salt content and annealing temperatures [54]. Because of a lack of uniformity, array data are often highly variable, showing in some instances for the same gene product no change, an increase or a decrease in abundance. This makes it difficult for laboratories to compare experimental data. In addition to the lack of standard controls, the use of signal ratios and incompatible reporting of data formats contribute to poor comparability between studies [55]. To overcome inter-laboratory differences, use of standardized procedures (probe preparation, array hybridization, primary data acquisition, data reporting and manipulations) will be required.


   7. Ideal arrays
Top
 Abstract
 1. Introduction
 2. What is the...
 3. Towards the development...
 4. Application of DNA...
 5. Data interpretation
 6. Limitations of gene...
 7. Ideal arrays
8. Conclusions
 References
 
To illustrate how microarray data can lead to the discovery of specific gene function or cellular pathways by gene expression profile matching, we summarize results from a more ideal model system utilizing yeast (S. Cerevisiae). In one set of experiments, Hughes et al. [44] generated and examined expression profiles from 300 diverse mutant yeast lines or after chemical treatments of normal yeast. They found that the expression profiles caused by deletion of the YER044c gene were similar to several reported sterol pathway mutant yeast profiles. The authors surmised that this unknown gene (YER044c) had a potential role in lipid/sterol metabolism. Further independent analyses on the YER044c gene deficient strain, which exhibited reduced growth rate, revealed an unusual lipid content in the cell and synthesis of novel sterols not normally seen in wild type cells. These results agreed with those from other drug studies and helped explain the sensitivity of the mutant strain to certain antifungal agents. The YER044c gene product was, therefore, classified as a sterol pathway enzyme, a finding later confirmed when the human homologue of the YER0044c gene was found to restore its growth rate. Similarly, pathway-specific reporter genes have been identified in yeast. The gene expression pattern of the uncharacterized YER083c mutant line showed a profile similar to the profiles of yeast treated with cell wall affecting agents. Additional properties of the mutant strain (growth and spheroplast lysis rates, sensitivity to cell wall altering agents) ultimately established the YER083c gene as a cell wall marker. Thus, comparing the gene expression profiles of seemingly unrelated mutant genes and test conditions in yeast has allowed a rapid characterization and function of the YER044c and YER083c genes, a task that could not have been achieved easily by conventional screening methods.

Although the heart is a complex organ system and the cardiac transcriptome is not as well characterized as the yeast's transcriptome, there is a considerable amount of published EST and gene information available that may be useful for determining the function of unknown cardiac genes. Liew's group has analyzed cardiac expression profiles during development [56,57], hypertrophy [27] and heart failure [36], and a number of differentially expressed gene clusters have been defined that are indicative of functional responses underlining myocardial infarction [35], cardiac hypertrophy [58], melatonin and geroprotective peptides in heart [40,41]. Their functional significance remains to be determined, but importantly, these data are useful in generating testable hypotheses addressing the molecular basis of coordinated gene expression under physiological heart conditions (normal and diseased) and after treatment with pharmacological agents.

Little data currently exist to explain the molecular genetic basis of cardiovascular disease [14,15], however, the acquisition of array data provides a format for future comparisons. As with yeast, DNA array analyses of cardiovascular tissues from animal models will ultimately provide critical information to address this issue [17]. Ischemic heart disease, heart failure and hypertrophy models have all been established in a variety of animal models (rats, dogs, rabbits, etc.) using surgical techniques. A number of rodent models also spontaneously develop cardiac pathologies [59] or can be manipulated genetically to mimic a disease state [60–63]. Genetic manipulations have successfully generated animals with myocardial hyperplasia, hypertrophy, cardiomyopathy, heart failure and myocarditis. These have become valuable tools for the study of specific pathways and functional abnormalities [59,64,65]. Although the genetic initiation event is known in some transgenic models, the downstream events necessary to elicit some forms of cardiovascular disease remain to be determined. Because expression profiles can be generated in animal models from both cross-sectional and longitudinal study designs, downstream events ultimately must be identified and examined in great detail, with and without pharmacological interventions. We should ultimately even be able to distinguish between changes in gene expression due to aging from those due to progressive disease processes. Finally, animal models will be of great use in developmental studies because of the ethical difficulties of research on humans. Together, these temporal and spatial profiles of the changes in gene expression as a function of the genetic anomaly, or more likely the adaptive response to congestive heart failure with or without drug intervention, will provide critical information relevant to gene function on to the development and treatment of heart failure.


   8. Conclusions
Top
 Abstract
 1. Introduction
 2. What is the...
 3. Towards the development...
 4. Application of DNA...
 5. Data interpretation
 6. Limitations of gene...
 7. Ideal arrays
 8. Conclusions
References
 
Our examples and works from other laboratories demonstrate the potential of arrays in the post-genomic era to contribute to the understanding of cardiovascular diseases. The growing popularity of array screening will lead to construction of specialized arrays to systematically describe disease states, which in turn can be used to examine the role of aging on the disease process. As disease genes are identified, DNA arrays will be used widely to diagnose disease and the potential for disease development. We will increasingly rely on these arrays and databases to organize information about each organ and disease state, and by combining results from many laboratories, we will ultimately be able to view networks from the vantage of the gene.

The availability of complete genome sequences is thus expected to change the way we address biological questions, diagnose and treat disease. The goal of such experiments is to generate new biological hypotheses by characterizing large numbers of gene products using a single high throughput tool (SAGE or arrays). The conventional reductionism approach of studying one gene at a time will be complemented by more global or integrative approaches that consider many genes at once (Fig. 5). This should ultimately facilitate gene and pathway discovery, and assist in the determination of which genes are aberrant or affected in disease states. Once identified, the implications for drug discovery are profound.


Figure 5
View larger version (30K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Fig. 5 Postgenomic cardiac research. Genome wide screens identify genes present in a transcriptome and can be used for downstream analysis (ratio based, clustering). Using different computational tools, patterns of gene expression across many experimental conditions are identified and expression profiles describing disease states can be generated. Genes present in the transcriptome without clear function that match the model profile of a particular disease can be used to construct a new hypothesis. The specific hypothesis can then be tested with smaller specialized arrays or independent techniques. After multiple rounds of computer and experimental data mining, multigene analysis profiles may form the basis of novel treatments as well as highlight the role of new gene products not yet implicated in a disease.

 
Expression profiling is not the endpoint of array data. The mRNA abundance in each element is an indicator of the template availability for protein synthesis—concomitant processes associated with phenotypic change. The comprehensive databases of these changes can facilitate the study of the entire protein complement of the genome (proteome) and the understanding of the process at the protein level. As we pass through the post-genomic era, where transcriptome analyses predominate, can the era of proteomics, metabolomics, etc. be far away?


   References
Top
 Abstract
 1. Introduction
 2. What is the...
 3. Towards the development...
 4. Application of DNA...
 5. Data interpretation
 6. Limitations of gene...
 7. Ideal arrays
 8. Conclusions
 References
 

  1. Yancik R., Ries L.A. Aging and cancer in America. Demographic and epidemiologic perspectives. Hematol Oncol Clin North Am (2000) 14:17–23.[CrossRef][Web of Science][Medline]
  2. Yancik R. Cancer burden in the aged: an epidemiologic and demographic overview. Cancer (1997) 80:1273–1283.[CrossRef][Web of Science][Medline]
  3. Balducci L., Extermann M. Cancer and aging. An evolving panorama. Hematol Oncol Clin North Am (2000) 14:1–16.[CrossRef][Web of Science][Medline]
  4. Beaufrere B., Morio B. Fat and protein redistribution with aging: metabolic considerations. Eur J Clin Nutr (2000) 54:S48–S53.[Web of Science][Medline]
  5. Fossel M. Cell senescence in human aging: a review of the theory. In Vivo (2000) 14:29–34.[Web of Science][Medline]
  6. Ladislas R. Cellular and molecular mechanisms of aging and age related diseases. Pathol Oncol Res (2000) 6:3–9.[Medline]
  7. Wick G., Jansen-Durr P., Berger P., Blasko I., Grubeck-Loebenstein B. Diseases of aging. Vaccine (2000) 18:1567–1583.[CrossRef][Web of Science][Medline]
  8. Mattson M.P., Pedersen W.A., Duan W., Culmsee C., Camandola S. Cellular and molecular mechanisms underlying perturbed energy metabolism and neuronal degeneration in Alzheimer's and Parkinson's diseases. Ann NY Acad Sci (1999) 893:154–175.[CrossRef][Web of Science][Medline]
  9. Schocken D.D. Epidemiology and risk factors for heart failure in the elderly. Clin Geriatr Med (2000) 16:407–418.[CrossRef][Web of Science][Medline]
  10. Marin J., Rodriguez-Martinez M.A. Age-related changes in vascular responses. Exp Gerontol (1999) 34:503–512.[CrossRef][Web of Science][Medline]
  11. Burt V.L., Whelton P., Roccella E.J., et al. Prevalence of hypertension in the US adult population. Results from the Third National Health and Nutrition Examination Survey, 1988–1991. Hypertension (1995) 25:305–313.[Abstract/Free Full Text]
  12. Cohn J.N., Bristow M.R., Chien K.R., et al. Report of the National Heart, Lung and Blood Institute Special Emphasis Panel on Heart Failure Research. Circulation (1997) 95:766–770.[Free Full Text]
  13. Lakatta E.G. Circulatory function in younger and older humans in health. In: Principles of Geriatric Medicine and Gerontology—Hazzard W.R., ed. (1999) New York: McGraw–Hill Companies, Inc. 645–660.
  14. Lifton R.P., Gharavi A.G., Geller D.S. Molecular mechanisms of human hypertension. Cell (2001) 104:545–556.[CrossRef][Web of Science][Medline]
  15. Seidman J.G., Seidman C. The genetic basis for cardiomyopathy: from mutation identification to mechanistic paradigms. Cell (2001) 104:557–567.[CrossRef][Web of Science][Medline]
  16. Hwang J.J., Dzau V.J., Liew C.C. Genomics and the pathophysiology of heart failure. Curr Cardiol Rep (2001) 3:198–207.[Medline]
  17. Anisimov S.V., Lakatta E.G., Boheler K.R. Discovering altered genomic expression patterns in heart: Transcriptome determination by serial analysis of gene expression. Eur J Heart Fail (2001) 3:271–281.[Abstract/Free Full Text]
  18. Finch C.E., Tanzi R.E. Genetics of aging (see comments). Science (1997) 278:407–411.[Abstract/Free Full Text]
  19. Lander E.S., Linton L.M., Birren B., et al. Initial sequencing and analysis of the human genome. Nature (2001) 409:860–921.[CrossRef][Medline]
  20. Venter J.C., Adams M.D., Myers E.W., et al. The sequence of the human genome. Science (2001) 291:1304–1351.[Abstract/Free Full Text]
  21. Hogenesch J.B., Ching K.A., Batalov S., et al. A comparison of the Celera and Ensembl predicted gene sets reveals little overlap in novel genes. Cell (2001) 106:413–415.[CrossRef][Web of Science][Medline]
  22. Lipshutz R.J., Fodor S.P.A., Gineras T.R., Lockhart D.J. High density synthetic oligonucleotide arrays. Nature Genetics (1999) 21:20–24.[CrossRef][Web of Science][Medline]
  23. Cheung V.G., Morley M., Aguilar F., Massimi A., Kucherlapati R., Childs G. Making and reading microarrays. Nature Genetics (1999) 21:15–19.[CrossRef][Web of Science][Medline]
  24. Southern E.M. DNA chips: analysing sequence by hybridization to oligonucleotides on a large scale. Trends Genetics (1996) 12:110–115.[CrossRef][Web of Science][Medline]
  25. Sachidanandam R., Weissman D., Schmidt S.C., et al. A map of human genome sequence variation containing 1.42 million single nucleotide polymorphisms. Nature (2001) 409:928–933.[CrossRef][Medline]
  26. Hwang D.M., Dempsey A.A., Wang R.-X., et al. A genome-based resource for molecular cardiovascular medicine toward a compendium of cardiovascular genes. Circulation (1997) 96:4146–4203.[Abstract/Free Full Text]
  27. Hwang D.M., Dempsey A.A., Lee C.Y., Liew C.C. Identification of differentially expressed genes in cardiac hypertrophy by analysis of expressed sequence tags. Genomics (2000) 66:1–14.[CrossRef][Web of Science][Medline]
  28. Ko M.S., Kitchen J.R., Wang X., et al. Large-scale cDNA analysis reveals phased gene expression patterns during preimplantation mouse development. Development (2000) 127:1737–1749.[Abstract]
  29. Hescheler J., Meyer R., Plant S., Krautwurst D., Rosenthal W., Scultz G. Morphological, biochemical and electrophysiological characterisation of a clonal cell (H9c2) line from rat heart. Circ Res (1991) 69:1476–1486.[Abstract/Free Full Text]
  30. Field L.J. Atrial natriuretic factor-SV40 T antigen transgenes produce tumors and cardiac arrhythmias in mice. Science (1988) 239:1029–1033.[Abstract/Free Full Text]
  31. Simpson D.G., Decker M.L., Clark W.A., Decker R.S. Contractile activity and cell–cell contact regulate myofibrillar organization in cultured cardiac myocytes. J Cell Biol (1993) 123:323–336.[Abstract/Free Full Text]
  32. Iwaki K., Sukhatme V.P., Shubeita H.E., Chien K.R. Alpha- and beta-adrenergic stimulation induces distinct patterns of immediate early gene expression in neonatal rat myocardial cells. J Biol Chem (1990) 265:13809–13817.[Abstract/Free Full Text]
  33. Fung Y.W., Liew C.C. Identification of genes associated with myocardial development. J Mol Cell Cardiol (1996) 28:1241–1249.[CrossRef][Web of Science][Medline]
  34. Sehl P.D., Tai J.T., Hillan K.J., et al. Application of cDNA microarrays in determining molecular phenotype in cardiac growth, development, and response to injury. Circulation (2000) 101:1990–1999.[Abstract/Free Full Text]
  35. Stanton L.W., Garrard L.J., Damm D., et al. Altered patterns of gene expression in response to myocardial infarction. Circ Res (2000) 86:939–945.[Abstract/Free Full Text]
  36. Barrans J.D., Stamatiou D., Liew C. Construction of a human cardiovascular cDNA microarray: portrait of the failing heart. Biochem Biophys Res Commun (2001) 280:964–969.[CrossRef][Web of Science][Medline]
  37. Lockhart D.J., Winzeler E.A. Genomics, gene expression and DNA arrays. Nature (2000) 405:827–836.[CrossRef][Medline]
  38. Yue H., Eastman P.S., Wang B.B., et al. An evaluation of the performance of cDNA microarrays for detecting changes in global mRNA expression. Nucleic Acids Res (2001) 29:E41.[CrossRef][Medline]
  39. Zhu Y., Yang H.-T., Boheler K. Identification of novel transcripts implicated in the progression to a senescent myocardium: results from microarrays. Circulation (2000) 102:II-142.
  40. Anisimov S.V., Boheler K.R., Khavinson V.K., Anisimov V.N. Elucidation of the effects of peptides Vilon and Epitalon on gene expression in mouse heart by microarrays. Bull Exp Biol Med (2002) 133:340–347.
  41. Anisimov S.V., Boheler K.R., Anisimov V.N. Microarray technology in studying the effect of melatonin on gene expression in mouse heart. Dokl Biol Sci (2002) 383:276–281.
  42. Boheler, KR, Garg R, Volkova MV, Morrell C, Margulies K, Lakatta EG. Identification and validation of novel gene products with altered expression in heart failure using microarrays and quantitative PCR. Physiol Genomics 2002.
  43. Eisen M.B., Spellman P.T., Brown P.O., Botstein D. Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci USA (1998) 95:14863–14868.[Abstract/Free Full Text]
  44. Hughes T.R., Marton M.J., Jones A.R., et al. Functional discovery via a compendium of expression profiles. Cell (2000) 102:109–126.[CrossRef][Web of Science][Medline]
  45. Tamayo P., Slonim D., Mesirov J., et al. Interpreting patterns of gene expression with self-organizing maps: methods and application to hematopoietic differentiation. Proc Natl Acad Sci USA (1999) 96:2907–2912.[Abstract/Free Full Text]
  46. Quackenbush J. Computational analysis of microarray data. Nat Rev Genet (2001) 2:418–427.[CrossRef][Web of Science][Medline]
  47. Miller R.A., Galecki A., Shmookler-Reis R.J. Interpretation, design, and analysis of gene array expression experiments. J Gerontol A Biol Sci Med Sci (2001) 56:B52–57.[Web of Science][Medline]
  48. Lee M.L., Kuo F.C., Whitmore G.A., Sklar J. Importance of replication in microarray gene expression studies: statistical methods and evidence from repetitive cDNA hybridizations. Proc Natl Acad Sci USA (2000) 97:9834–9839.[Abstract/Free Full Text]
  49. Anisimov SV, Tarasov KV, Lakatta EG, Boheler KR. Quantitative transcriptome analysis of the adult mouse heart using serial analysis of gene expression. Physiol Genomics 2002: abstract, in press.
  50. Claverie J.M. Computational methods for the identification of differential and coordinated gene expression. Hum Mol Genet (1999) 8:1821–1832.[Abstract/Free Full Text]
  51. Luo Z., Geschwind D.H. Microarray applications in neuroscience. Neurobiol Dis (2001) 8:183–193.[CrossRef][Web of Science][Medline]
  52. Lockhart D.J., Dong H., Byrne M.C., et al. Expression monitoring by hybridization to high-density oligonucleotide arrays. Nat Biotechnol (1996) 14:1675–1680.[CrossRef][Web of Science][Medline]
  53. Kane M.D., Jatkoe T.A., Stumpf C.R., Lu J., Thomas J.D., Madore S.J. Assessment of the sensitivity and specificity of oligonucleotide (50mer) microarrays. Nucleic Acids Res (2000) 28:4552–4557.[Abstract/Free Full Text]
  54. Britten R.J., Davidson E.H. Hybridisation strategy. In: Nucleic Acid Hybridization. A Practical Approach—Hames B.D., Higgins S.J., eds. (1985) Oxford–Washington DC: IRL Press. 3–16.
  55. Watson S.J., Meng F., Thompson R.C., Akil H. The ‘chip’ as a specific genetic tool. Biol Psychiatry (2000) 48:1147–1156.[CrossRef][Web of Science][Medline]
  56. Fung Y.W., Liew C.C. Identification of genes associated with myocardial development. J Mol Cell Cardiol (1996) 28:1241–1249.[CrossRef][Web of Science][Medline]
  57. Rezvani M., Barrans J.D., Dai K.S., Liew C.C. Apoptosis-related genes expressed in cardiovascular development and disease: an EST approach. Cardiovasc Res (2000) 45:621–629.[Abstract/Free Full Text]
  58. Johnatty S.E., Dyck J.R., Michael L.H., Olson E.N., Abdellatif M. Identification of genes regulated during mechanical load-induced cardiac hypertrophy. J Mol Cell Cardiol (2000) 32:805–815.[CrossRef][Web of Science][Medline]
  59. Hasenfuss G. Animal models of human cardiovascular disease, heart failure and hypertrophy. Cardiovasc Res (1998) 39:60–76.[Abstract/Free Full Text]
  60. Chien K.R. Molecular advances in cardiovascular biology. Science (1993) 260:916–917.[Free Full Text]
  61. Gottshall K.R., Becker K.D., Hunter J.J., Chien K.R. A genetic based model of cardiac hypertrophy in MLC-ras mice. J Card Fail (1996) 2:S28–34.[Medline]
  62. Tanaka N., Dalton N., Mao L., et al. Transthoracic echocardiography in models of cardiac disease in the mouse. Circulation (1996) 94:1109–1117.[Abstract/Free Full Text]
  63. Chien K.R. Cardiac muscle diseases in genetically engineered mice: evolution of molecular physiology. Am J Physiol (1995) 269:H755–766.[Web of Science][Medline]
  64. Baker D.L., Hashimoto K., Grupp I.L., et al. Targeted overexpression of the sarcoplasmic reticulum Ca2+-ATPase increases cardiac contractility in transgenic mouse hearts. Circ Res (1998) 83:1205–1214.[Abstract/Free Full Text]
  65. Huang X., Pi Y., Lee K.J., et al. Cardiac troponin I gene knockout: a mouse model of myocardial troponin I deficiency. Circ Res (1999) 84:1–8.[Abstract/Free Full Text]
  66. Mirnics K., Middleton F.A., Lewis D.A., Levitt P. Analysis of complex brain disorders with gene expression microarrays: schizophrenia as a disease of the synapse. Trends Neurosci (2001) 24:479–486.[CrossRef][Web of Science][Medline]

Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?


This article has been cited by other articles:


Home page
 ReproductionHome page
M. Breuiller-Fouche and G. Germain
Gene and protein expression in the myometrium in pregnancy and labor.
Reproduction, May 1, 2006; 131(5): 837 - 850.
[Abstract] [Full Text] [PDF]

Home page
 Cardiovasc ResHome page
M. Volkova, R. Garg, S. Dick, and K. R. Boheler
Aging-associated changes in cardiac gene expression
Cardiovasc Res, May 1, 2005; 66(2): 194 - 204.
[Abstract] [Full Text] [PDF]

Home page
 Journals of Gerontology Series A: Biological Sciences and Medical SciencesHome page
C. C. Deocaris, S. C. Kaul, K. Taira, and R. Wadhwa
Emerging Technologies: Trendy RNA Tools for Aging Research
J. Gerontol. A Biol. Sci. Med. Sci., August 1, 2004; 59(8): B771 - B783.
[Abstract] [Full Text] [PDF]

Home page
 Proc. Natl. Acad. Sci. USAHome page
K. R. Boheler, M. Volkova, C. Morrell, R. Garg, Y. Zhu, K. Margulies, A.-M. Seymour, and E. G. Lakatta
Sex- and age-dependent human transcriptome variability: Implications for chronic heart failure
PNAS, March 4, 2003; 100(5): 2754 - 2759.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Juhasz, O.
Right arrow Articles by Boheler, K. R.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Juhasz, O.
Right arrow Articles by Boheler, K. R.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?