|Home | About | Journals | Submit | Contact Us | Français|
The putative role of (TG/CA)n repeats in the regulation of transcription has recently been reported for several cancer- and disease-related genes, including the genes encoding the epidermal growth factor receptor (EGFR), hydroxysteroid (11-beta) dehydrogenase 2 (HSD11B2) and interferon-gamma (IFNG). These studies indicated a correlation between gene expression levels and the presence or length of (TG/CA)n repeats in their intron 1 regions. A genome-wide search for genes with similar features may provide evidence of whether these dinucleotide repeats represent a class of universal regulators of gene expression, which has recently begun to be investigated as a quantitative complex phenotype. Using a public database of simple repeats, we identified 330 genes containing potentially polymorphic long (TG/CA)n repeats (n ≥ 12) in their intron 1 regions. One known physiological pathway, the calcium signalling pathway, was found to be enriched among the genes containing long repeats. In addition, certain biological processes, such as cation transport, signal transduction and ion transport, were found to be enriched in these genes. Genotyping of the long repeats showed that the majority of these dinucleotide repeats were polymorphic in the HapMap CEU (Caucasians from Utah, USA) samples of northern and western European ancestry. Evidence for a significant association between these repeats and gene expression was not observed in the genes selected based on their expression profiles in the HapMap CEU samples. Our current findings, therefore, do not support a role for these repeats as a class of universal gene expression regulators. A more comprehensive evaluation of the relationship between these repeats and gene expression, potentially in other tissues, may be necessary to illustrate their roles in gene regulation in the future.
Microsatellite sequences that consist of repeating units of two to five base pairs with different lengths and complexities are dispersed along the human genome and exhibit extensive length polymorphism, a feature widely used in genetic mapping [1-3]. Among the dinucleotide repeats, (TG/CA)n (CA-) repeats represent the most common class of microsatellites in vertebrates, including humans . Because of their structure of alternating purine/pyrimidine sequences, CA- repeats have a tendency to form Z-DNA under physiological conditions, suggesting a possible role for such Z elements in chromatin activation or genome rearrangements . Biochemically, this feature of CA-repeats may affect the movement of RNA polymerases, thus modulating gene expression levels . It has also been shown that intronic CA sequences constitute novel and widespread regulatory elements of alternative splicing , which, in turn, adds another layer of complexity in the human genome .
The abundance of intronic CA-repeats and gene expression has been found to be inversely correlated for some highly expressed housekeeping genes, indicating a putative global mechanism of gene expression . This observation raises the question of whether CA-repeats in intronic regions are functional as a gene regulation mechanism for other human genes as well. In other words, does this feature represent a unique class of gene regulators for the whole genome? Interestingly, the relationship between the length polymorphisms of CA-repeats located in the first introns or intron 1 regions (I1Rs) and gene expression has recently been reported in several cancer- or disease-related genes, including the genes encoding the epidermal growth factor receptor (EGFR) , hydroxysteroid (11-beta) dehydrogenase 2 (HSD11B2) and interferon-gamma (IFNG) . These studies indicated a correlation between gene expression level and the polymorphic status of CA-repeats in the I1Rs. While gene expression could be regulated by various genetic or non-genetic factors -- such as the presence of cis- or trans- regulatory elements [13-15], copy number variants (CNVs) and DNA methylation , -- a genome-wide search for genes with similar structure to these genes could potentially provide evidence that this particular feature (I1 repeats) represents a class of universal gene regulation mechanism.
Structurally, CA- repeats can be divided into categories by length, which may reflect their biological properties . For example, short repeats (6 ≤ n < 12 units) have a very low propensity for polymorphisms, while long repeats (n ≥ 12 units) are more likely to be polymorphic and functional [19,20]. More than 90 per cent of the CA-repeats of n ≥ 12 units were found to display length polymorphisms and may act as cis-regulators of transcription . In general, CA- repeats of n ≥ 12 units in highly expressed housekeeping genes show a down-regulatory effect on transcription, which suggests an inverse correlation with the proportion of repeats in introns . Therefore, we sought to explore the relationship between gene expression and relatively long repeats (n ≥ 12 units) located within the I1Rs. In particular, using a public database of small repeats and the RefSeq-supported  genes, we searched for genes with I1 (TG/CA)n repeats in the proximity of their neighbouring exons in the human genome. Since genes within the same biological pathways are potentially under similar levels of evolutionary pressure, our next goal was to evaluate if there were any enriched or over-represented known pathways and biological processes in the genes with this particular feature. Known functional annotations, such as those maintained at the Gene Ontology (GO) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) databases, were searched for the classification of the genes. Statistically significant pathways and biological processes were found to be enriched in the genes with I1 (TG/CA)n repeats. Genotyping of the I1 repeats showed that the majority of these repeats were polymorphic in a panel of human lymphoblastoid cell lines (LCLs) derived from apparently healthy Caucasian individuals of northern and western European ancestry from the International HapMap Project [7,25] (CEU samples; ie Caucasians from Utah, USA). We then tried to evaluate experimentally the relationship between mRNA expression and the polymorphisms of the I1 repeats in these LCLs.
(TG/CA)n (where n ≥ 6) repeats in introns were identified in the human genome using Satellog (http://satellog.bcgsc.ca) , which is a database for the identification and prioritisation of pure repeat unit microsatellite repeats. Satellog runs on v. 34 of the human genome (version hg16, July 2003) and provides the ability to identify repeats based on user-specified characteristics, such as repeat unit, period, length and genomic coordinates. To collect our analysis set of I1 repeats, we mapped the identified intronic repeats from Satellog to the same version of the human genome reference using the exonic positions of the RefSeq-supported genes (curated non-redundant genes) retrieved by PipHelper (http://pipmaker.bx.psu.edu/cgi-bin/piphelper/) . The final analysis dataset comprised I1 (TG/CA)n repeats within the 2.5 kilobase (kb) windows either downstream or upstream of neighbouring exons (Figure (Figure1).1). We limited our analysis to those introns bigger than 2.5 kb.
Distribution of the genes containing I1 (TG/CA)n repeats across the human genome were tested against the null chromosomal distribution of the human genome (version hg16), which has 18,299 RefSeq-supported  genes. Over- or under-represented chromosomes were determined using binomial tests. A false discovery rate (FDR) of 5 per cent after the Benjamini-Hochberg (BH) correction  was used as the cut-off for statistical significance.
We used the Database for Annotation, Visualization and Integrated Discovery (DAVID)[29,30] (http://david.abcc.ncifcrf.gov) to identify enriched GO  (http://www.geneontology.org) or Protein ANalysis THrough Evolutionary Relationships (PANTHER) (http://www.pantherdb.org/pathway/) biological processes, as well as known pathways such as those maintained in the KEGG  (http://www.genome.jp/kegg/), Biocarta (http://www.biocarta.com) and PANTHER  databases among the genes containing long I1 (TG/CA)n repeats. Biological processes or pathways over-represented relative to the whole human genome (version hg16) were identified using Fisher's exact test (five hits or more, FDR < 0.05 after BH correction ).
To evaluate if the I1 (TG/CA)n repeats were polymorphic, we genotyped a list of genes with relatively long (n ≥ 12) repeats in the 59 unrelated HapMap CEU samples. The gene expression levels of these samples previously had been measured using the Affymetrix GeneChip ® Human Exon 1.0ST array (Affymetrix, Santa Clara, CA, USA) and are publicly available through the Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/), accession GSE7851 . Ten genes with the top-ranking variability and associated with enriched pathways or biological processes were included in the genotyping (Table (Table1).1). DNA samples were extracted from LCLs purchased from the Coriell Institute for Medical Research (Camden, NJ, USA). Polymerase chain reaction (PCR), in combination with fluorescently labelled oligonucleotide primers, was used to amplify the I1Rs containing (TG/CA)n repeats. Primers were designed to amplify the repeat region . One primer was labelled with 6-carboxyfluorescein at the 5' end. The primer sequences are listed in Table Table1.1. The PCR reaction mix was prepared with 1 unit of HotStart Taq Polymerase (Qiagen, Valencia, CA, USA), 1.5 mM MgCl2 (Qiagen, Valencia, CA, USA), 250 nM of each primer (IDT, San Jose, CA, USA) and 20 ng of genomic DNA. PCR amplification was performed in a thermal cycler (Bio-Rad, Hercules, CA, USA) with an initial step of 12 minutes at 95°C, followed by 35 cycles of 10 seconds at 95°C, 20 seconds at 56°C and 40 seconds at 72°C. PCR products were separated in the ABI 3700 DNA sequencer (Applied Biosystems, Foster City, CA, USA), and the genotypes were read using GeneMapper® software v4.0 (Applied Biosystems). The Hardy-Weinberg equilibrium (HWE) was tested for the frequencies of each polymorphism. There lative lengths of the repeats were used in further analyses. We set the location of the first peak to appear in the GeneMapper® software as 1. The following peaks were read based on the distance from the first peak. Three random homozygous samples of two genes (the genes encoding the insulin receptor [INSR] and calmin [CLMN]) were sequenced to validate the genotyping.
We quantified the transcriptional expression levels of two genes, INSR and CLMN, by real-time PCR in the 59 unrelated CEU samples. For each cell line, 3 μg of total RNA was used and was reversely transcribed into cDNA with the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). The MxPro - Mx3000P QPCR System (Stratagene, La Jolla, CA, USA) was used to quantify the expression levels of each gene. The reactions were performed in a 96-well plate, in a final volume of 25 μl containing iQ™ SYBR® Green Supermix (Bio-Rad), 500 nM of forward and reverse primers, respectively, and cDNA template corresponding to 4.5 ng of total RNA. The thermal cycling conditions were 10 minutes at 95°C, followed by 40 cycles of 30 seconds at 95°C, 1 minute at 55°C and 30 seconds at 72°C, completing with one cycle of 1 minute at 95°C, 30 seconds at 55°C and 30 seconds at 95°C. A housekeeping gene, the gene encoding FK-506-binding protein 1A (FKBP1A), was included as the internal control. Primer sequences are listed in Table Table11.
Samples were grouped into classes based on their relative repeat genotypes. For example, the samples can be grouped into either 'low category' (with relatively low numbers of repeats '1/1') or 'high category' (with other genotypes such as '2/5' or '1/7'). Other combinations, such as 'samples with allele 1' versus 'other samples' and 'sum of alleles in genotype A' versus 'sum of alleles in genotype B', were also tested. Gene expression levels from realtime PCR were then evaluated for correlation with the repeat genotypes in the 59 unrelated CEU samples. Linear regression was performed using the stat::lm function in the R Statistical Package .
A total of 5,424 (TG/CA)n (where n ≥ 6) repeats in all introns were identified in the human genome (version hg16) using Satellog. By mapping the repeats to the human genome, a total of 1,137 repeats were found to be located in the I1Rs (Table S1 in the supplementary material), corresponding to 839 unique human genes. We grouped these genes according to whether they contained short (six to 11 units) repeats (704 repeats and 536 genes) or long (≥12 units) repeats (433 repeats and 388 genes). We further limited our analysis dataset to the simple repeats within the 2.5 kb windows either downstream of exon 1 or upstream of exon 2, as well as to I1Rs larger than 2.5 kb (Figure (Figure1).1). In total, our final analysis set comprised 371 long (TG/CA)n repeats (n ≥12) (corresponding to 330 genes) in either the 5' (205 repeats and 187 genes) or 3' (194 repeats and 187 genes) 2.5 kb windows (Table (Table22 and Table S2 in the supplementary material). Among these, 44 genes had long repeats in both the 5' and 3' windows.
The chromosomal distribution of the 330 genes containing long I1 (TG/CA)n repeats was compared against the null distribution of the human genome (version hg16). Figure Figure22 shows the chromosomal distribution of these genes. At FDR < 5 per cent, chromosome 11 was found to be underrepresented relative to the whole genome (eight genes; p = 8.3 × 10-4; FDR = 0.02). No other chromosomes were found to be either over-represented or under-represented relative to the human genome reference.
Using the DAVID web application [29,30], four GO  and 17 PANTHER  biological processes were found to be enriched among the 330 gene-containing long I1 (TG/CA)n repeats. The most significant enriched biological process was the PANTHER  term for cation transport (97 genes; p = 6.20 × 10-11; pc = 1.40 × 10-8). One KEGG  pathway -- the calcium signalling pathway (16 genes; p = 2.1 × 10-11, FDR = 0.041) -- was also enriched among these genes. Table Table33 lists these enriched biological processes and pathways (FDR < 0.05).
Ten I1-repeats-containing genes (the genes encoding runt-related transcription factor [RUNX], CD38 molecule [CD38], CLMN, very low-density lipoprotein receptor [VLDLR], calcium channel, voltage-dependent, R type, alpha 1E subunit [CACNA1E], aryl hydrocarbon receptor [AHR], neuregulin 2 [NRG2], par-3 partitioning defective 3 homolog [PARD3], INSR and insulin-like growth factor binding protein 4 [IGFBP4]) with highly variable gene-level expression based on a public microarray expression dataset  in the unrelated CEU samples were genotyped. Table S3 in the supplementary material shows the relative lengths of repeats of the 59 unrelated CEU samples. The data from microsatellite analysis showed that the I1 repeats of these genes were polymorphic. Table Table44 shows the summary of the genotyping results. RUNX and IGFBP4 did not follow the HWE and are therefore not listed in Table Table4.4. All the other genes followed the HWE (Fisher's exact test, p < 0.05). Three PCR products of INSR and CLMN were randomly selected for sequencing validation. The sequencing results were consistent with the genotyping data (data not shown).
The relationship between genotype and gene expression level was evaluated in the remaining eight genes whose repeat genotypes were consistent with the HWE (Table (Table4).4). Real-time PCR was performed to measure their gene expression levels (Table S4 in the supplementary material). When grouping the samples based on repeat length genotypes, among the eight genes we tested, no statistically significant associations were observed using combinations of genotypes, including 'samples with allele 1' versus 'other samples', 'sum of alleles in genotype A' versus 'sum of alleles in genotype B' and 'low category' (with relatively low numbers of repeats '1/1') versus 'other genotypes'. The two most significant relationships were INSR (r = -0.23; p = 0.088) and CLMN (r = -0.19; p = 0.15) for the 'low category' (genotype 1/1) versus 'other genotypes' comparison.
Understanding gene regulation is critical to understanding the underlying mechanisms of health-related phenotypes, such as the risks of common diseases. Except for a few classes of elements, such as the promoter regions , the putative role of most non-coding elements (eg simple tandem repeats) in gene regulation, however, has not been comprehensively investigated. In this work, we explored the relationship between gene expression and a class of dinucleotide repeats in the intron 1 regions -- that is, I1 (TG/CA)n repeats.
We focused on relatively long repeats because they are more likely to be polymorphic , and therefore potentially more likely to have a role in gene regulation. We further limited our analysis set of repeats to the flanking regions of 2.5 kb windows (Figure (Figure1).1). This allowed us to focus on the relationship between gene expression and neighbouring repeats, which also were the targets of a couple of previous studies of EGFR , HSD11B2  and IFNG . Overall, approximately 2 per cent of known human genes have this particular feature based on our criteria. The 330 genes containing long I1 repeats were generally distributed evenly across the human genome, although chromosome 11 was found to be under-represented. This under-representation could be due to the arbitrary cut-offs (ie the 2.5 kb windows and n ≥ 12) we chose to use in the final analysis set. We confirmed that relatively long (TG/CA)n repeats in the I1Rs are universal across the human genome and exist in hundreds of human genes. Furthermore, a couple of biological processes were found to be enriched among these 330 genes using different annotation databases. For example, the GO and PANTHER biological processes related to metal ion transport, as well as the KEGG and PANTHER pathways related to calcium signalling, were found to be enriched, indicating that these repeats may play a role in the function or regulation of certain signalling pathways. Interestingly, the PANTHER biological processes BP00040: mRNA transcription and BP00044: mRNA transcription regulation were enriched among these genes, hinting at a potential relationship between I1 repeats and transcriptional gene expression.
By genotyping the I1 repeats of ten highly variable genes expressed in the unrelated HapMap CEU samples , we demonstrated that the majority of repeats in these LCL samples were polymorphic (Table (Table4).4). Because the functional models of these repeats are not yet known, various classifications of the repeat polymorphisms, including 'low category' (with relatively low numbers of repeats '1/1') versus 'other genotypes' were tested for association with gene expression measured by real-time PCR. No statistically significant associations (p < 0.05) were observed between various repeat genotype classes and gene expression in these LCLs, however. The two top-ranking relationships were identified in INSR and CLMN when comparing the repeat genotypes between 'low category' and 'others'. Overall, these current findings do not appear to support a putative role for I1 repeats as a novel class of universal gene regulation mechanism which contributes to variation in transcript abundance. Because of the limited number of genes we tested experimentally, however, a much more comprehensive investigation might be necessary to confirm finally whether this relationship is a universal mechanism for other genes. Due to the fact that only approximately 50 per cent of genes are believed to be reliably expressed in the LCLs [15,35], and that gene regulation can be tissue specific , it could be interesting to carry out the same investigation in other tissue types, to illustrate the potential role of I1 repeats in gene regulation. Furthermore, since population-level gene expression variation has been observed in [15,35,37,38] humans, a future survey, using different human populations, might be also worthwhile for illustrating the relationship between I1 repeats and gene expression.
This Pharmacogenetics of Anticancer Agents Research (PAAR) Group (http://www.pharmacogenetics.org) study was supported by NIH/NIGMS grant U01GM61393. Data will be deposited into PharmGKB (supported by NIH/NIGMS Pharmacogenetics Research Network and Database grant U01GM61374, http://www.pharmgkb.org/).