Gene arrays

- Genomics, proteomics and bioinformatics of human heart failure

Gene array technology for analysis of changes in gene expression in failing human hearts is now well advanced. Initial sequencing of the human genome suggested it contains an estimated 32,000 genes (Lander et al., 2001; Venter et al., 2001) and more recently this figure has been revised downwards (Collins et al., 2003).

In principle, since the sequence of each gene is known, it can be precisely probed using gene arrays that contain either robotically-arrayed full length (or nearly full length) cDNA cloned probes, or using shorter (approximately 25 base pair) oligonucleotides that are synthesized on glass. The assumption that in practice each probe accurately quantifies the expression of the corresponding gene may be overly simplistic. For eukaryotes, the correlation between the quantity of mRNA and the quantity of the corresponding protein can be poor (Le Naour et al., 2001). Despite these concerns, gene arrays remain the most extensive source of reliable information on the molecular status of an organ like the human heart.

Gene arrays start with intact RNA

Oligonucleotide and cDNA arrays

Two DNA microarray technologies are currently in use for the study of human disease. The first is the Hu133 ‘AffyChip’, an oligonucleotide array manufactured by Affymetrix that contains 22,000 transcripts representing the major transcripts of the human genome. The second, technology is the spotted cDNA arrays. These are both commercially available or can be manufactured in-house (Barrans et al., 2002; Hwang et al., 2002). One example of the latter is the 12,000-element ‘CardioChip’ (Barrans et al., 2001). The CardioChip consists of unique cDNA clones of heart libraries that have been partially sequenced (i.e. ESTs) and annotated. This particular array is customized for the study of heart failure. It contains known cardiovascular-related genes as well as a broad selection of novel genes and ESTs derived from our laboratory's heart cDNA library. The layout consists of 2761 known genes (25.5% of the array), 3406 matched ESTs (31.4%), 4489 novel ESTs (41.4%) and 192 bacterial controls (1.8%). A representative CardioChip (Figure 1) compares Cy3 (red) labelled non-diseased and Cy5 (green) labelled failing heart samples.

Gene clustering

The major advantage of DNA microarray technology lies in its ability to profile and compare thousands of genes simultaneously between several mRNA populations. However, the utility of the DNA microarray extends beyond the concept that it is a novel tool for large-scale transcript profiling and identifying differences in expression between single genes.

It is the wealth of data generated from a series of transcript profiling experiments that makes this technique so powerful. Groups of data can be analyzed to determine the level of ‘relatedness’ between genes or samples in multiple dimensions. Basically, using a set of expression fingerprints or profiles, similarities and differences in gene expression are capitalized on in order to group or cluster different mRNA populations or genes into discrete, related sets or bins. This is extremely powerful as the clusters of co-regulated genes often belong to the same biological pathway, or even to the same protein complex. Thus the clusters of mRNA populations defined by their ‘expression fingerprint’ provide a means of defining differences between samples that would not otherwise be possible.

Gene clustering between mRNA populations can be viewed from two perspectives. Clustering over a time-course can identify genes with a similar function because the expression of functionally related genes tends to be regulated, and thus expressed, in a similar manner (e.g. genes involved in G1 of the yeast mitotic cell-cycle). This is an important concept, as even genes with no known function can now be placed into a biological pathway.

Alternatively, clustering genes between several mRNA populations from a single time point can identify a vast number of associations between the expression of a cluster of genes and a biological phenotype such as marker or sentinel genes. Thus, through clustering the complete sets of gene expression data generated from samples, several gene clusters will be identified and examined to determine the specific relationship with each particular mRNA population. Furthermore, these results can be extended to define different etiologies of cardiomyopathies.

A focus on different causes of human heart failure

Over the past 4 years, we have examined a limited number of explanted heart samples representing hypertrophic and dilated cardiomyopathies (Barrans et al., 2002; Hwang et al., 2002) and suggested that 200–300 genes may be involved in the processes of developing heart failure. Through cluster analysis of gene profiles, donor (n = 6), hypertrophic (n = 3) and dilated (n = 6)samples can be clustered distinctively.

Over the past decade or so, we have accumulated a collection of over 250 explanted hearts from patients undergoing heart transplantations. Based on our available samples we have selected: (1) non-familial idiopathic dilated cardiomyopathy; (2) doxorubicin-poisoned myocardium; (3) peripartum cardiomyopathy; (4) hypertrophic cardiomyopathy; (5) failure due to known genetic causes (including Eisenmenger, Ebstein, and Holt-Oram syndromes, Beckers muscular dystrophy) and (6) ischemic heart disease particularly the diffuse type. Multiple samples have been harvested from left and right ventricles and atria as well as other tissues such as coronary arteries and aorta.

This large collection will help us to define the biological pathways and clusters of genes related to the etiology of heart failure. For example, we can now select multiple examples of idiopathic dilated cardiomyopathy that are essentially free of coronary vascular disease. Importantly, we now have nearly 60 examples of non-diseased, donor hearts that can act as ‘controls’ for the failing hearts.

AffyChip or CardioChip: which is better?

We are aware that a major question currently concerns which of the two microarray technologies is superior? Actually, we feel that neither is superior and that both can have ideal or equivalent situations for use. Thus one or the other should be selected based on an appropriate context. Of course, there are inherent advantages and disadvantages to both technologies. These are summarized as follows:

We feel that both technologies should be used concertedly. They can complement each other and allow us to take the best approach to identify the genes/pathways involved in investigations into the molecular basis of heart failure (Table 1).