“Expression Linked Polymorphism Database (EXPOLDB): A resource for linking genome wide expression with cis modulators of transcription in the Human Genome ”

The availability of the human genome sequence and the parallel accumulation of microarray data offer new opportunities to examine the role of repetitive sequences as modulators of gene expression underlying the variation in expression. Therefore, in order to obtain insight into gene regulation in humans, a necessary pre-requisite step is to link gene expression data from microarray experiments with the distribution of cis modulators of transcription such as the dinucleotide (TG/CA)_n repeats. At present there is no database linking genome wide expression with the distribution of (TG/CA)_n and other simple repeats.

EXPOLDB contains information on (i) transcript levels and their variation from oligonucleotide microarray (HG U95A v2) data, (ii) distribution of (TG/CA)_n repeats and their polymorphic status in more than 5000 human genes, (iii) a plethora of functional information such as biochemical roles of the gene products, tissue specific expression, and associated hyperlinks and (iv) levels and variation in expression of known human housekeeping genes across unrelated individuals and monozygotic twins. In addition, examples of well studied genes regulated by (TG/CA)_n repeats are included for reference. EXPOLDB can be queried through a user-friendly interface and allows navigation to obtain additional functional information through combinatorial queries. Further information can be obtained by browsing HTML links to the publicly available databases inclusing Gene Cards, UCSC Golden Path, Entrez Gene, BodyMap, GXD, ENSEMBL, HuGE Index and PubMed.

EXPOLDB incorporates an online tool 'SimRep' which can be used to examine the distribution of simple repeats and patterns in EXPOLDB genes or in a given nucleotide sequence. Users can search either for a dinucleotide repeat by selecting it from the pull down menu or for a specific microsatellite repeat of their choice by entering the pattern to be searched. Patterns can be specified using both the standard four base symbols A, T, G and C. Other symbols recommended by IUPAC can also be used. The minimum length for scoring a repeat can be specified in the field 'Enter Cut-Off'. SimRep reports the length and location of the repeats or patterns in the given sequence in the form of a table. Repeats are reported in both strands if the option 'All' is selected. The positions of repeats either in forward strand (+) or in reverse strand (-) are reported with respect to the forward strand only as per the convention followed by genome sequence annotation groups. In the case of palindromic dinucleotide repeats such as GC, AT only one strand is reported.

As a resource aimed to provide both genotype and phenotype data by including information on gene expression, variability and presence of cis regulatory elements, EXPOLDB is likely to have wider usability and can serve as a useful resource to researchers assessing natural variation in gene expression and those interested in examining the role of (TG/CA)_n repeats in gene expression. A few case studies describing the implementation of EXPOLDB are highlighted below.

Tetrapodic layout of EXPOLDB: The 4 domains of layout are shown in bold face type. Note that all attributes of a gene are singularly linked to its official HGNC symbol serving as the primary key.

The backend data was prepared in MS Access 2000 (Microsoft Corporation, Inc. USA). Server site scripting was prepared using ASP (Active Server Pages, version 3.0), PHP (PHP: Hypertext Preprocessor, version 5.0) and Perl (Practical Extraction Report Language, version 5.8.1). The client site scripting was prepared using JavaScript and HTML (Hyper Text Markup Language, version 4.0). Internet Information Server (IIS) version 6.0 was used as web server.

Natural variation in gene expression between healthy human individuals has been largely unexplored. The variation in gene expression is an outcome of the complex inter-play of genetic polymorphisms (acting in cis or in trans), physiological variations (such as time of day, gender) and environmental factors. In order to understand the genetic basis of variation in gene expression between normal human individuals, we need to obtain genome-wide expression data from various populations. We examined the gene expression profiles in 13 normal human individuals including five pairs of monozygotic twins and three unrelated individuals in blood leukocytes measured using HG U95Av2 oligonucleotide microarrays consisting probes for ~10, 000 genes.. A total of 5,407 genes were found expressed in blood leukocytes. Of these, a total of 2,888 genes were found to be differentially expressed in pairwise comparisons between unrelated individuals and 212 genes in monozygotic twins. Information on mean expression and variability (CV) for human housekeeping genes that had present call (P) in any of the used arrays is also provided in EXPOLDB. This database is likely to be a useful resource for those that are interested in studying natural variations in humans.
The raw data from the array experiments has been submitted to Gene expression Omnibus (GEO; www.ncbi.nlm.nih.gov/geo) under the Series ID: GSE928, and the following accession numbers: GSM14477, GSM14478, GSM14479, GSM14480, GSM14481, GSM14482, GSM14483, GSM14485, GSM20645, GSM29053, GSM29054, GSM29055, GSM29056, GSM29057 and GSM29058.

The mean expression of each gene was computed from the log10 transformed 'signal' values with P calls across all 13 arrays. All genes considered for mean expression had present (P) call in arrays. Replicate probe sets if present, were clustered together and averaged. A total of 5,407 genes were found expressed in blood leukocytes using these criteria. For these genes, the coefficient of variation (CV) was computed as SD/Mean where SD is the standard deviation of the log10 transformed 'signal' values across the 13 arrays.
Two measures have been used to assess variation in gene expression. The metric ‘signal log ratio’ is used to query differentially expressed genes. The metric ‘coefficient of variation (CV)’ is used to query the variability in expression levels across individuals. It is important to note the difference between the two metrics. First, the computation of ‘signal log ratio’ was carried out according to the Affymetrix procedure whereas the CV was carried out according to Hsiao et al. 2001. Second, the two metrics do not signify the same biological phenomena. For example, signal log ratio is measured in pair wise comparisons whereas CV is measured across all arrays. Although high signal log ratio may relate to high CVs, the exact relationship has not been established since, to our knowledge only HuGE Index reports these values in addition to EXPOLDB. Thus, users are advised to exercise care while using the two metrics and the use of a particular metric will depend on the nature of the study.
To enable distinction between the two query strategies and to avoid confusion, we have provided the 'signal log ratio' selection option only with ‘Differentially Expressed Genes’ query page. Similarly, the CV option is available only with ‘Expression in Blood’ page. Searches for a set of differentially expressed genes can be carried out by either defining a range of signal log ratio or selecting from a given list of specified range. The variability in the expression of genes (CV) can be queried through the page entitled “Expression in blood” by either selecting a specified CV range from the list box or specifying a range of CV or the entire range of CV.

Examining the role of (TG/CA)_n repeats as modulators of gene expression

About 50% of the human genome consists of repetitive elements comprising of simple repeats, short interspersed nucleotide elements (SINEs), medium reiteration, long terminal repeats and long interspersed nucleotide elements (LINES). Among the dinucleotide repeats, (TG/CA)_n is the most frequent in the human genome and many of these repeats exhibit length polymorphism. This property has been extensively used in the construction of genetic maps. (TG/CA)_n repeats which have alternating purine/pyrimidine sequences have a propensity to undergo conformational transition on methylation under physiological conditions.

(TG/CA)_n repeats can influence transcription of a gene in cis (variations within the gene) due to ‘incidence’ or ‘secondary elongation’ in these repeats (either within or in close proximity to the gene). The functional roles of (TG/CA)_n repeats are beginning to emerge. Since the early observation on the modulation of transcription by (TG)_n tracts by Hamada et al (1984), experimental evidences on the role of (TG/CA)_n repeats in the regulation of gene expression have been steadily accumulating. The up-regulation or down-regulation of transcription by (TG/CA)_n repeats has been reported for the following genes: rat alpha-lactalbumin (Meera et al., 1989), rat prolactin (Naylor et al. 1990), Acetyl-CoA carboxylase ACC (Tae et al. 1994), matrix metalloprotease MMP-9 (Shimajiri et al. 1999), gamma interferon IFN-gamma (Pravica et al. 1999), epidermal growth factor receptor EGFR (Gebhardt et al. 1999), salt sensitivity HSD11B2 (Agarwal et al. 2000), and tilipia Prolactin1 (Streelman et al. 2002). Expression differences in these genes due to polymorphic (TG/CA)_n repeats varies from one gene to another, exhibits a wide range, and in some cases is reported to be as high as 20 folds. While the effects induced by microsatellite variation may be complicated by the presence of other transcriptional regulatory elements in the proximity of a given gene, these observations underscore the importance of the (TG/CA)_n repeats and their polymorphisms in gene regulation. A list of more examples from literature illustrating the role of these repeats as modulators of gene expression is given below.

Recently, several reports describing the association of polymorphism in (TG/CA)_n repeats with genetic diseases have appeared such as in the coronary heart disease (eNOS, endothelial nitric oxide synthase; Laule et al. 2003), in diabetic retinopathy (ALR2, aldose reductase; Kumaramanickavel et al. 2003), in asthma (IFN-gamma, gamma interferon; Nagarkatti et al. 2002) and in breast cancer (IGF-I, insulin-like growth factor-I;Yu et al. 2001).

Earlier we have shown in rat alpha-lactalbumin (Meera and Brahmachari et al., 1989) that the presence of (TG/CA)_n repeats correlates with lower gene expression levels. Recently, we examined the distribution of the (TG/CA)_n repeats in the human genome and also the correlation of incidence of the repeats with gene expression levels of housekeeping genes measured using oligonucleotide microarrays. The number of short intragenic (TG/CA)_n repeats was significantly higher than the number of long repeats, the proportion of genes with (TG/CA)_n repeats (n ≥ 12 units) had lower mean expression levels compared to those without these repeats, the genes belonging to the functional class of ‘signaling and communication’ had a positive association with repeats in contrast to the genes belonging to the ‘information’ class that were negatively associated with repeats (Sharma et al. 2003). Taken together, these observations underscore the importance of investigating natural variation in transcript levels at the genomic scale with respect to the distribution of (TG/CA)_n repeats both within the genes and in their close proximity.

Case studies to examine the role of (TG/CA)_n repeats as modulators of gene expression
using EXPOLDB