Ancestry-related differences in gene expression: findings may enhance understanding of health disparities between populations

`Analysis of variation in gene expression between human populations will enhance our understanding of the contribution of genetics to health disparities.'

The unveiling of the first human genome draft sequences almost 7 years ago [1,2] marked the begining of the postgenomic era. However, because of the existence of huge genetic variations, such as those in the form of SNPs [3] and copy-number variations (CNVs) [4], there is still much to be discovered. In recent years, functional studies have revealed that natural genetic variation plays a role in complex human traits, such as risks of common diseases [5], for example, asthma, stroke, heart attack, diabetes and cancer, and variability in drug response [6-8]. The importance of genetic variation in the human genome has increasingly attracted researchers' efforts in the past few years, leading to its selection as the Breakthrough of the Year in 2007 by Science [9].

To explore the complexity of the human genome, parallel efforts to the Human Genome Project have included the International Hap-Map Project [10,11]. This project has been invaluable in understanding the complexity of human genetic variation not only among individuals of the same ancestry [12], but also between different human populations [13-16]. The goal of the HapMap Project is to develop a haplotype map of the human genome and to describe the common patterns of DNA sequence variation using 270 lymphoblastoid cell lines (LCLs) derived from individuals of African (YRI: 30 parents-child trios from the Yoruba people in Ibadan, Nigeria), Asian (CHB: 45 unrelated Han Chinese from Beijing, China; JPT: 45 unrelated Japanese from Tokyo, Japan) and European ancestry (CEU: 30 parents-child trios from the Centre d'Etude du Polymorphisme Humain samples collected from UT, USA). The newly released Phase II component of the HapMap genotypic data contains over 3.1 million human SNPs [3] that reveal common haplotype patterns and differences in genetic variation among these four geographically distant populations. As Hap-Map LCL samples are commercially available, researchers can obtain relevant phenotypes (e.g., drug response and gene expression) on those cell lines and perform population-based whole-genome associations with the publicly available genotypic data to identify responsible genetic determinants or SNP markers linked to the causal elements [17].

Among various phenotypes, gene expression acts as an intermediate phenotype situated between variation in DNA sequence and other more complex cellular, tissue, organ or whole-body phenotypes. Quantitative variation in gene-expression level (e.g., mRNA transcript abundance) as a complex trait is heritable and has been mapped to the human genome as expression quantitative trait loci (eQTLs), which represent genomic regions for the genetic control of gene expression [18-20].

`Quantitative variation in gene-expression level as a complex trait is heritable and has been mapped to the human genome as expression quantitative trait loci…'

Significant gene-expression variation has previously been observed within the same human population [18-20]. More recently, using the LCL samples from the four HapMap populations, the ancestry-related or population differences in gene expression has been described by investigators utilizing various whole-genome expression microarrays and statistical approaches [17]. Comparing the unrelated 60 CEU samples and 82 Asian samples (41 CHB and 41 JPT), Spielman et al. showed that common genetic variants account for differences in gene expression among ethnic groups [14]. They found that approximately 25% of genes analyzed (out of approximately 4200 deemed expressed in LCLs) using the Affymetrix (CA, USA) Human Focus array were differentially expressed between the European-derived and Asian-derived populations, while there were few differences between the two Asian populations [14]. Using the same microarray platform, Storey et al. found that approximately 17% of genes (out of ~5200 deemed expressed in LCLs) are differentially expressed between eight CEU samples and eight YRI samples [15]. They also found that genes differentially expressed between the CEU and YRI samples were strongly enriched in inflammatory pathways even after a strict Bonferroni correction for multiple-hypothesis tests. Included in this set of genes were several cytokines and chemokine receptors that have been implicated in numerous cardiovascular, infectious and immune-related diseases [15]. Using the Affymetrix Human Exon array [21] (containing ~1.4 million exon-level known and predicted probesets), we recently performed a comprehensive comparison of the expression of approximately 9200 transcript clusters (gene-level), which were deemed reliably expressed out of approximately 18,000 transcript clusters with RefSeq-supported [22] annotatation, between 87 CEU samples and 89 YRI samples [13]. Approximately 4.5% (383) of the tested genes were found to be differentially expressed between the CEU and YRI populations. While no differential genes appeared to be over-represented on certain chromosomes, biological processes including ribosomal biogenesis, tRNA processing and antimicrobial humoral response were found to be enriched in these differential genes, suggesting their possible roles in contributing to the population differences at a higher level than that of mRNA expression and in response to environmental information [13]. Impressively, our finding of the enrichment of immune response-related genes (e.g., CCR7 and CXCR3) was in agreement with the Storey et al. smaller-scale analysis [15]. Our findings also suggest that the frequencies of common genetic variants contribute to a substantial fraction of gene-expression variation between human populations. Previous studies have focused on cis-acting elements [14,15], but our results suggest that distant or trans-acting elements can also contribute to the population differences in gene expression. Using the Illumina (CA, USA) whole-genome expression array, Stranger et al. tested the population differences in gene expression among all the avaliable HapMap populations (CEU, YRI and CHB/JPT), with the major aim to dissect genetic regulators of gene expression [16]. They estimated that the fraction of genes with significant gene-expression variation between any two populations is between 17 and 29%. Their results also support an abundance of cis-regulatory variation in the human genome with the existence of limited trans effects [16].

`…gene expression differs among human populations and common genetic variants account for this variation in cis and in trans mode.'

Collectively, these studies demonstrate that gene expression differs among human populations and common genetic variants account for this variation in cis and in trans mode [13-16]; however, some differences exist in these findings. The most apparent discrepancy is the proportion of differential genes between two populations, ranging from 4.5 to 29% [13-16]. This is primarily due to differences in sample size, statistical approaches, level of cutoff for significance, data preprocessing/normalization procedures and microarray platforms utilized for the studies. The two most comprehensive studies [13,16] used different statistical approaches to accommodate the trio structures of the Hap-Map CEU and YRI samples. Our analysis incorporated both a permutation-based approach and a general linear model with Toeplitz structure for modeling parents-child covariance, while Stranger et al. defined a cutoff based on the assumption that parents and CEU or YRI children represent the null distribution of differences in median expression values [13,16]. The other two studies [14,15] used unrelated samples and therefore did not have to consider relatedness. There were also other technical differences in the analysis. First, we made an attempt to address the potential confounding factor of SNPs in probes [23], while other studies did not take this potential problem into account in their analyses. Second, although, generally speaking, the microarray data are consistent with other experimental quantitative techniques and across microarray platforms [24,25], the Affymetrix GeneChips® and Illumina BeadArrays used in these studies have differences in probe-set design, gene coverage and probe-set replication. Furthermore, some limitations or confounding factors (both measurable or unmeasurable) could influence the accuracy of these results [26]. A re-analysis by Akey et al. of the data from one expression study [14] suggested that a large proportion of the differential genes between CEU and CHB/JPT samples could be due to systematic and uncorrectable bias [27]. This confounding factor is probably due to the age (in culture) of the CEU samples [28] and time at which analysis was performed (batch effect) relative to the other recently established CHB/JPT and YRI samples. To address the issue of age in culture, we tried a reduced quantitative transmission disequilibrium test [29] model to evaluate the contribution of genetic and nongenetic factors for the differential genes associated with SNPs [13]. The majority of our differential genes were not explained by the population identity alone, suggesting the substantial contribution of genetic factors.

`…ancestry is a significant determinant of both susceptibility to some diseases and response to therapeutic treatments.'

Gene expression can affect susceptibility to diseases and drug response. A widely appreciated example is the multidrug resistance exhibited by tumor cells through overexpression of MDR genes [30]. Although disease risk and drug response are likely to be contributed by the interplay between both genetic and nongenetic factors, gene-expression variation plays a significant role, probably through upregulating or downregulating genes within physiological pathways. For example, Caucasians and African-Americans differ significantly in EGFR expression in prostate cancer [31], suggesting potential differences in response to EGFR inhibitors between these two populations. Therefore, analysis of variation in gene expression between human populations will enhance our understanding of the contribution of genetics to health disparities.

Clinically, studies have shown that ancestry is a significant determinant of both susceptibility to some diseases and response to therapeutic treatments. For example, African-American, Hispanic, Asian and Native American women have a lower incidence of breast cancer but higher mortality compared with non-Hispanic white women [32]. Differences in response to anticancer agents docetaxel and carboplatin have been observed between Asian and Caucasian advanced non-small-cell lung cancer patients [33]. Findings of population differences in expression are critical for us to investigate these health disparities between human populations. Interestingly, the enrichment of immune response-related genes in both ours [13] and a previous smaller study [15] suggests that individuals of African and European ancestry may have different susceptibility to infectious diseases. In fact, it has been observed that African-American adults may be more susceptible to infection by certain microbes such as Porphyromonas gingivalis, which causes periodontitis [34].

Obviously, the recent progress demonstrated the usefulness of the HapMap resource, which is comprised of comprehensive genotypic and gene-expression data, in identifying genetic regulators for gene-expression variation among populations. However, as the current HapMap samples include only four populations (three if combining the two Asian populations), we are still far from understanding the differences in gene expression involving other major human populations (African-Americans, Mexicans and Pacific Islanders, to name a few). To extend these kinds of studies to other populations would present more challenging problems; for example, the African-American population is believed to be an admxiture of African, European and Native American descent, thus a relatively clear cut of ancestry may not be as evident as the current HapMap populations. Another limitation of the current studies is that only mRNA-level expression has been compared between populations; a natural question is does this equate to gene- or pathway-level differences in protein expression? Progress in areas such as proteomics and metabolomics will facilitate future studies of population differences in gene expression. Therefore, current success in evaluating global differences in expression is just the beginning of our understanding of how differences in genetic variation and expression variation explain differences in disease susceptibility or drug response.

Biography