Understanding the genetic contribution to phenotype variation of human groups is necessary to elucidate differences in disease predisposition and response to pharmaceutical treatments in different human populations.
We have investigated the genome-wide profile of structural variation on pooled samples from the three populations studied in the HapMap project by comparative genome hybridization (CGH) in different array platforms. We have identified and experimentally validated 33 genomic loci that show significant copy number differences from one population to the other. Interestingly, we found an enrichment of genes related to environment adaptation (immune response, lipid metabolism and extracellular space) within these regions and the study of expression data revealed that more than half of the copy number variants (CNVs) translate into gene-expression differences among populations, suggesting that they could have functional consequences. In addition, the identification of single nucleotide polymorphisms (SNPs) that are in linkage disequilibrium with the copy number alleles allowed us to detect evidences of population differentiation and recent selection at the nucleotide variation level.
Overall, our results provide a comprehensive view of relevant copy number changes that might play a role in phenotypic differences among major human populations, and generate a list of interesting candidates for future studies.
Citation: Armengol L, Villatoro S, González JR, Pantano L, García-Aragonés M, et al. (2009) Identification of Copy Number Variants Defining Genomic Differences among Major Human Groups. PLoS ONE 4(9): e7230. doi:10.1371/journal.pone.0007230
Editor: Marc Bauchet, Max Planck Institute for Evolutionary Anthropology, Germany
Received: January 20, 2009; Accepted: August 20, 2009; Published: September 30, 2009
Copyright: © 2009 Armengol et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. This work was supported by the Generalitat de Catalunya, the Spanish Ministry of Science and Innovation [SAF2008-00357 to X.E., MTM2008-02457 to J.R.G., and BFU2007-60930 to M.C.]; and the European Commission [AnEUploidy project; FP6-2005-LifeSciHealth contract #037627]. M.C. was supported by the Ramon y Cajal Program (Spanish Ministry of Science and Education).
Competing interests: The authors have declared that no competing interests exist.
Although any two humans are 99.9% identical at the nucleotide sequence level , , many phenotypic differences are apparent in individuals within the same and from distinct human populations. Genetic diversity underlying the remaining 0.1% nucleotide differences has been postulated to contribute to phenotypic diversity among humans, and to population-specific susceptibility to disease and variability in the response to pharmacological treatments –. The different prevalence of Mendelian diseases reflects variability in allele frequencies for specific genes and haplotypes, and the relevance of ethnic background in the susceptibility to disease is recognized for several disorders , including cystic fibrosis –, sickle cell anemia –, and deafness , among many others. Similarly, there are differences in the prevalence of common disorders and associated genetic variants in human populations, such as the factor V Leiden (venous thromboembolic disease , ), variants in the CARD15 gene (Crohn's disease , ), the CCR5–Δ32 variant (human immunodeficiency virus (HIV) infection and progression ), and APOE e4 (Alzheimer's disease ). In addition, population-specific genetic variants could also influence the metabolism and response to drugs, exemplified by the population variability in the acetylating activity of the N-acetyltransferase 2 (NAT2) gene . However, despite the progress on the identification of genes for inflammatory bowel disease and neurodegenerative disorders, little is known about the genetic contribution to the prevalence differences of other complex traits among human groups .
The current knowledge of divergence among major populations and subpopulations is based in the study of different sources of genetic variation within and amongst human populations –. The analysis of nucleotide and structural variability on human populations has shown that genetic clusters closely correspond with human groups defined by ethnicity or continental ancestry –. In agreement with human African origins, a higher amount of SNP allelic variability has been found in individuals of African ancestry, while individuals of Asian origin show the smallest amount of nucleotide diversity . Another major conclusion from these studies is that for most nucleotide and small genomic changes (in/dels), intra-population differences among individuals account for the majority of variation, while differences among major population groups represents a minimal fraction , –. However, this variation cannot be neglected.
Despite that the existence of structural variation in the human genome has been known for a while (see  and  for review), we only recently have started to recognize its real scope and wide-spread nature –. This variation is present in different forms, including insertions, deletions and inversions. Due to the relative ease of identification, copy number variants (CNVs) are the most studied and best recognized form of structural variation, and have attracted much attention because of its potential to affect gene expression , .
Most studies aiming at the discovery of CNV regions have made use of genome-wide approaches based on different array-based comparative genomic hybridisation (aCGH) and comparative genomic intensity (aCGI) platforms, and have provided partial evidences of their nature, real size and/or population distribution. The first comprehensive map of CNVs was produced in late 2006  by the analysis of 270 HapMap reference samples using high-density BAC arrays and Affymetrix 500 K EA chips. It allowed the identification of 1,447 CNV regions (~360 Mb; 12% of human genome), showing a 43% concordance between the two platforms. A subsequent analysis of 71 individuals from three different human populations using an equivalent high-density BAC array revealed 315 CNV regions, 40% of them not having been previously reported . A targeted high-density analysis of 1,153 known CNVs (using oligonucleotide-based aCGH at a 1 Kb resolution) on 30 HapMap samples, showed that 76% of CNV regions were at least 50% smaller than previously reported . More recently, using a dense aCGI platform (Affymetrix 6.0), McCarroll and co-workers refined the CNV knowledge on the HapMap samples and identified 3,048 CNV regions, 40% of them not overlapping with previously reported variants . Furthermore, an in-depth study using the Illumina HumanHap550 BeadChip surveyed copy number variation on individuals from 29 different human worldwide populations . Although little experimental validation was performed, their analysis of 396 non-unique autosomal loci revealed that CNV distribution parallels the population structure reported by SNP data. All these studies highlight that, except for those focusing on single or a reduced number of loci –, more detailed analysis of the loci is necessary to determine their potential contribution to normal phenotypic variation and disease susceptibility. In addition, although a big proportion of the human genome has been suggested to be copy-number variable, little is known about the functional consequences of differences in population distribution of CNVs.
A series of studies have already linked rare and common CNVs to complex diseases (such as Alzheimer's disease, HIV and AIDS susceptibility, chronic pancreatitis, autism or schizophrenia) (reviewed in ) and established etiopathogenesis connecting changes in copy number with alteration of expression levels of dosage-sensitive genes. Apart from the direct effect on gene expression, a link between some structural variants and predisposition to disease-causing rearrangements has been proven, especially for genomic disorders –. Thus, population-specific differences of CNV frequency could make a certain human population more prone to suffer certain types of disorders.
In this study we report the identification of CNVs that are present in very different frequency in individuals from the three phase I/II HapMap analysis panels  (YRI - Yoruba in Ibadan, Nigeria; CEU - Utah residents with ancestry from Northern and Western Europe; and CHB+JPT - Han Chinese in Beijing, China and Japanese in Tokyo, Japan), representatives of sub-Saharan Africa, Europe and East Asia, respectively. Most of the genomic variants detected have already been described as CNVs, involve genes, and modify, to a certain extent, their expression levels. Furthermore, we show that these CNVs are sufficient to discriminate population structure at the continental level, and that some of them are embedded in population-specific haplotypes. We finally provide evidences of the effect of natural selection on the variants identified.
Identification of population differences in CNV regions
We have used aCGH ,  on Agilent and BAC-based platforms to identify CNVs with different frequencies in three representative human populations of African (YRI – Yoruba in Ibadan, Nigeria), Asian (JPT – Japanse in Tokyo, Japan, and CHB – Han Chinese in Beijing, China), and European (CEU – Utah residents with ancestry from Northern and Western Europe) ancestry. In order to dilute inter-individual variation and to enrich inter-population differences, we pooled 50 unrelated DNA samples from each of the four human groups studied in the HapMap project , . For each group, we set up independent pools of males and females, and carried out a total of sixteen aCGH hybridization experiments on each platform, by confronting sex-unmatched pools of individuals from the different populations (Figure 1A). As a control, we performed an intra-population male versus female hybridization to discard the probes that called a copy number difference in this situation (see Methods), since we do not expect copy number variability between genders of the same group, other than those affecting the sexual chromosomes. The circular design of the hybridization experiments (Figure 1A) allowed us to determine the most likely population that carries the variant (increase or decrease in copy number) with respect to the other populations, assuming the most parsimonious scenario. In addition, this design also allowed us to minimize the presence of spurious positive signals (Figure 1B).
Figure 1. General Strategy Used in the Analysis of Copy Number Variants (CNVs) with Population-Specific frequency changes.
(A) Pools of genomic DNA of 25 unrelated males and 25 females from the four HapMap populations were prepared and hybridized to the different array platforms. In total, sixteen inter-population CGH experiments were performed in each platform. The circular design of the hybridization pairs (panel A) allowed us to assign the most likely population to carry the variation, assuming the most parsimonious scenario. (B) This combination of hybridizations also allowed us to minimize the calling of CNVs on spurious positive signals. For instance, for a CNV to be called as a YRI-specific loss, the hybridizations of CEU, CHB and JPT male pools were required to show an increased hybridization signal with respect to the YRI female pool (log2 ratio≥0.3, green squares), the hybridization of the YRI male pool versus the rest of pools had to show decreased signals (log2 ratio≤−0.3, red squares), and the rest of combinations should show values around 0 and not above the |0.3| threshold (dark squares).doi:10.1371/journal.pone.0007230.g001
The global analysis of the aCGH data on the three array platforms yielded a total of 111 loci whose copy number state frequencies differed among populations (Table S1), including 59 gains, 43 losses, and 9 complex changes which showed a range of variation that we could not assign to a single population (see Methods). Because of the similarities observed between the JPT and CHB groups, and since they both belong to a population of Asian ancestry, we grouped them and will use the ASN (Asian) term to refer to them. From the copy number changes that could be assigned to a population, 26 were CEU-specific, 23 ASN-specific and 53 YRI-specific (Table 1). The amount of copy number changes found in the YRI population doubles that of any of the other individual populations, and is concordant with previous observations of higher frequency of changes for small, intermediate and large-scale variants and insertion/deletion polymorphisms , . Of the identified regions, 61% (n = 68) overlap with previously reported CNVs, and 32% (n = 35) overlap with segmental duplications.
Table 1. Combined results of Copy Number Variant (CNV) changes in HapMap populations using genome-wide aCGH platforms.doi:10.1371/journal.pone.0007230.t001
Validation of the aCGH findings
To confirm the changes detected with the aCGH platforms, we performed exhaustive validation experiments using Multiplex Ligation-dependent Probe Amplification (MLPA)  on individual DNAs from the HapMap collection. Depending on the length of the variable segment, one or two MLPA probes per region were designed, preferentially within genes (Table S2). We split the predicted CNVs into three different categories depending on the number of consecutive BAC or oligonucleotide probes that supported the altered segment (1, 2 or >2 probes per region, respectively; Table 2). We decided to validate putative CNVs from all three groups, although we anticipated that the lower the number of array probes supporting each segment the higher the likeliness of being a false positive. In total, we analyzed 152 probes in 4 different probe mixes (see Methods) that interrogated 93 of the 111 CNV regions, including 60 CNVs already described in the literature and 33 potentially novel CNVs. Overall, after validation, 47% of the tested loci (n = 44) turned out to be variable, and the majority of them (86%; n = 38) corresponded to already known CNVs, thus only 6 were novel. Out of the 44 variable regions, 33 (75%) were concordant with the results observed in the aCGH experiments, while the remaining 11 did not coincide with the aCGH prediction or did not show significant differences among populations. As expected, the validation rate of the CNVs varied considerably depending on the number of supporting array probes, and was low (33%) for group 1 (CNVs composed of only 1 altered probe) and much higher for groups 2 and 3 (73% and 72%, respectively) (Table 2). On the other hand, our results are consistent with the observations made by others ,  in the sense that the boundaries of many variants reported so far are not precisely defined, because different MLPA probes located within the same theoretical CNV did not behave uniformly (Figure S1).
Table 2. Validation of CNV changes identified in HapMap populations using different aCGH platforms.doi:10.1371/journal.pone.0007230.t002
Although direct comparison is difficult, as CNV region boundaries identified by arrays cannot be well defined and different array platforms use probes located in different regions, we have estimated to what extent our results matched with previous data deposited in the database of genomic variants (DGV - ) by evaluating reported frequencies of overlapping CNVs (Table S3). Of the 33 CNV loci validated as showing populations differences, in 13 of them we could not establish a concordance with published results, either because the CNV was not previously reported, there was not enough population information to compare, or there were inconsistencies in the available data. In 80% of the remaining 20 cases (16/20), the copy number frequencies we obtained matched with the available population data; while in the other 4 cases the information we obtained did not agree with what it had been published beforehand. Thus, most of the CNVs with population differences are independently confirmed by various methods and supported by previously existing evidences.
To test for the existence of population stratification based on CNVs, we took advantage of the MLPA data generated for all analyzed loci to perform an a priori principal component analysis of the copy number information estimated for each individual . We estimated that thirty percent of total variance is explained by the two principal components (PC1 16,62%, and PC2 13,38%) and we observed that using these two axes the majority of individuals consistently cluster according to their population of origin, thus supporting the idea that the three human groups are characterized by a markedly different composition of the identified CNV regions (Figure 2).
Figure 2. Principal component analysis of MLPA data in HapMap populations.
We decomposed the genetic variation within MLPA-estimated copy number into the first two principal axes , plotted each data point, and colored them according to the population of origin. Thirty percent of the total variance is explained by the two first principal components (PC1 16,62%, and PC2 13,38%), and it is evident that the first principal component discriminates variation inherent to the population of African ancestry (YRI), while the second component differentiates the populations of European (CEU) and Asian (ASN) ancestries.doi:10.1371/journal.pone.0007230.g002
Gene content in CNV regions showing population differences
Thirty of the 33 CNVs that we validated to show population differences contain known coding genes, totaling 40 RefSeq genes (Table S1). The three remaining CNV regions are completely devoid of coding sequences, but using the Evolutionary Conservation of Genomes Browser  we identified conserved segments that might have some potential biological function that has not yet been uncovered (data not shown).
To characterize the nature of the genes located within the set of validated CNVs we carried out a GO-term analysis using the Gene Ontology Tree Machine . This analysis revealed a significant enrichment of genes involved in immune response, which is consistent with previous studies –. Other functional categories enriched when compared both to the rest of human genes or the set of genes located within known CNVs (Tables S4 and S5) were protein and lipid metabolism genes, and genes related to the extracellular space. Therefore, the genes identified are mainly related to interaction and adaptation to the environment. In addition, we performed an analysis of the rate of non-synonymous (Ka) to synonymous (Ks) nucleotide changes (Ka/Ks) between humans and chimpanzees to test whether the genes with changes in copy number among human populations are characterized by a different selective regime in the ‘recent’ evolutionary history, which could be consistent with a role in environment response. We found that the average Ka/Ks ratio of our set of genes is not significantly different from what it would be expected by chance compared to the rest of coding genes in the genome. Similar results were obtained using Ka/Ks estimates retrieved from Ensembl (Tables S6 and S7). Therefore, we conclude that the CNV gene set has been subjected to analogous functional constraints than other genes during hominoid evolution.
Several of the genes coinciding with the CNVs detected here attracted our attention because of their biomedical relevance and potential involvement in human adaptation to environmental pressures. For example, the Haptoglobin-related protein (HPR) is found in a CNV region that exhibits a higher copy number in the YRI individuals than in the individuals from the other two populations analyzed, and could be related to adaptation to sleeping sickness disease (see Discussion). Other interesting genes located within regions with considerable copy number differences among populations include several that were previously known to be variable and associated to different biomedical conditions (GSTT1, UGT2B17, PRSS2 and APOBEC), as well as others that, although associated to disease phenotypes, have not been reported to be variable among populations. Among those, there are several promising candidates that might be influencing different population prevalence of diseases like age-related macular degeneration (CFHR3) –, cardiovascular disease (LPA) – and chronic obstructive pulmonary disease (HCK) .
Contribution of the CNV changes identified to gene expression variability
The most intuitive functional consequence of a change in genomic dose is a change in gene-expression levels (i.e. more copies imply higher expression, and vice versa). To check this, we identified a total of 23 RefSeq genes contained within the 44 validated CNV regions (regardless of showing population differentiation) for which expression levels in lymphoblastoid cell lines, derived from the same samples we analyzed, had previously been reported . Of those, 16 transcripts exhibited average expression levels above the threshold of the negative controls and were further considered. First, we wanted to know to what extent global expression level of genes within the variable regions could be attributable to copy number differences. In total, five transcripts showed significant correlation (P<0.0031, after Bonferroni correction) between copy number and expression levels for the three populations combined, which represents a slightly higher proportion (31%) than previously reported  (Table 3). Second, when considering only the CNVs with population-specific frequency changes, 10 out of 14 (71%) transcripts showed significant expression differences among populations (Kruskal-Wallis Test; P<0.0036; after Bonferroni correction; Table 3) and we could assess a positive correlation between expression levels and the copy number in 60% of them (6 out of 10) (Figure S2). One interesting instance is the UGT2B17 gene, whose expression levels have been previously associated with specific cis-acting SNPs , , that we have found to be tightly correlated with the copy number of the gene (Table 3), as previously reported by others , . Thus, it seems reasonable to postulate that the differences in copy number are responsible of controlling expression, and it is possible that the SNPs previously identified are mere tags for the different copy number alleles. Nevertheless, when assessing the relationship between CNVs and gene-expression, we have to keep in mind that we solely used expression data that was collected from a single tissue. In addition, it is important to note that other genes were contained within the identified CNVs, but were not included in our analysis because their expression levels were not reported in the above-mentioned study or because the MLPA probes we used did not coincide exactly with the gene. Given the poor definition of CNV boundaries reported in the currently available databases observed by others ,  and us, we decided not to take into account the genes for which we have not directly inspected copy number.
Table 3. Expression differences in lymphoblastoid cell lines from HapMap samples and correlation with copy number at CNV loci.doi:10.1371/journal.pone.0007230.t003
LD and SNP haplotype structure of population variable CNV regions
Although only a small proportion of human copy number variation can be attributed to population differences , , in this work we specifically sought to enrich for this type of changes. In order to check in more detail the underlying genetic architecture of these regions, we examined the nucleotide variation data from the HapMap project.
Initially, we used the available HapMap genotype information of each individual to search for evidences of LD between the identified CNVs and flanking SNPs. The median of the maximum r2 values for single SNPs located up to 100 kb around the boundaries of each CNV was 0.37, indicating a low overall disequilibrium between SNPs and copy number alleles. Nevertheless, very heterogeneous disequilibrium patterns were observed for individual CNVs. Four loci show very high LD (r2>0.8) with surrounding SNPs, and others present low levels of disequilibrium (Figure 3). This indicates that, as reported previously , –, at least in some cases it might be feasible to define tag-SNPs as proxies for CNV alleles in certain populations that could be helpful in future analyses of those variants (Table S8). Given the complexity of variable regions and the difficulty to genotype within and around them, it is worth noting that the average number of SNPs that could be analyzed (1/2059 bp) was slightly lower than the average density in HapMap (1/1495 bp; according to the total number of polymorphic common-to-all-population SNPs) .
Figure 3. Linkage disequilibrium between HapMap SNPs and CNVs.
The distribution of maximum measured r2 value for the SNPs analyzed for each CNV is plotted in the histogram, which represents the frequency of findings (Y axis) in each range of LD values (X-axis).doi:10.1371/journal.pone.0007230.g003
Due to the population specificity of our set of CNVs, we expected to find some of them embedded in particular haplotypes, and therefore to be able to identify different patterns of LD between surrounding SNPs and copy number alleles in each population. When possible, we performed pairwise comparisons of LD patterns between populations using the Spearman rank correlation value (ρ), and found that they are quite different in most cases (Table S9). Thus, the regions explored here do not seem to follow the general rule of preserved LD patterns across populations that are observed globally in the genome , , , and this is a potential evidence of recent selection. In extreme cases, in which the CNV appears exclusively in a single population, the test could not even be performed. On the other hand, we also identified three cases in which there is a high correlation value. This observation is consistent with a high degree of concordance in the LD patterns observed in the different populations (i.e. CNVs interrogated by probes chr5_150_A, A_14_P133280 and Chr1B_pop_2), and likely indicates that these CNVs appeared as a single event predating the out-of-Africa human migrations.
To examine more carefully the degree of population differentiation at the SNPs that are in highest LD with the identified CNVs (tag-SNPs) we used the FST measure , . FST values range from 0 (no differentiation) to 1 (complete differentiation) and can provide a measure of the action of natural selection on these regions. It is known that negative and balancing selection tends to decrease FST in a locus-specific manner, whereas local positive selection tends to increase FST , although unusual FST values are not a definitive proof of the action of selection. According to the distribution of the available FST data for HapMap SNPs , , we considered values above 0.20 (corresponding to the top 15–20th percentile of the distribution of FST values between each pair of populations) as evidence of population differentiation. Ten of the tag-SNPs had global (i.e. across the three populations) FST values above this threshold. In seven out of the ten cases, the most differentiated population coincided with the one in which the population variable CNV was identified (Table S10). These results strengthen our initial findings with the aCGH and MLPA experiments, and indicate that the CNV differences among populations are often replicated in linked SNPs.
Finally, the analysis of haplotype structure can provide a good estimate of the contribution of positive selection in our regions of interest, which might account for the observed population differences . To test for recent selection events, we used the available Integrated Haplotype Score (iHS) data, which takes into account different measures of haplotype length and composition in each population . We found that seven tag-SNPs showed signals of recent positive selection (|iHS|≥2); and in six of those, the extreme iHS values were corresponded to the population in which we observed the change in CNV frequency (Table S10). Therefore, there are evidences that, at least in some cases, the change in frequency of a CNV in a population was associated to positive selection acting on itself or on a linked nucleotide variant.
Given the unexpected extent of structural variation among individuals, it is very likely that there is an important contribution of such variants to phenotypic differences, including predisposition to common and complex disorders. Although different authors have reported the existence of CNVs in individuals from different geographical locations , , , , experimental validation is lacking for a number of the reported variants. To uncover the main part of such variability, we have performed whole genome array-based comparative genome hybridization of pooled DNA from different populations, and made an important effort to provide validation of the most relevant findings. In addition, we have combined different available sources of information to estimate the functional and evolutionary relevance of these changes.
Given the resolution and coverage of our arrays, we have been able to identify a “relatively” small set of CNVs that show significant differences in population distribution. Despite the stringent criteria for CNV calling, an important number of them could not be validated by MLPA analysis, especially those supported by just one array probe, and several reasons can account for that. First, it is possible that some of the aCGH observations were indeed false positives due to the large number of regions assayed in the arrays (i.e. the larger the number of tests, the larger the type I error). Second, in the case of the variable loci reported by a single probe in the oligonucleotide arrays (Table 2), it was not always possible to design the MLPA probes in the exact same place, and they were designed within one kb around the array probe. As a result, it is possible that the probe ended up located outside of the really variable region. Finally, it is also possible that some of the loci reported had a basal copy number that is too high to be correctly measured by MLPA, that is known to reliably distinguish copy numbers between zero and five copies , , but its sensitivity greatly decreases when higher copy numbers are measured. This is likely the case of the chemokine CCL3L1 gene that appeared as variable in the array but could not be confirmed by MLPA, although we know that it is indeed highly variable, with median values of 3 and 6 copies per diploid genome in non-Africans and Africans, respectively . On the other hand, despite the limitations of matching differently defined CNVs, the comparison of the MLPA validation data with previously reported information has revealed a very good concordance, with 80% of the CNVs we identified agreeing with previously published data. It is a fact that genome-wide platforms have been a great tool for obtaining a global view of copy number variation, but big efforts of validation on individual loci will be required in the future prior to establish the impact of individual CNVs in human health. In this regard, MLPA has proven to be an easy, efficient, reliable and inexpensive method to validate copy number variants in large sets of samples.
CNVs have been suggested to play an important effect on gene-expression levels. In our study, 60% (6 out of 10) of the genes showing inter-population expression differences associate with copy number changes, which reinforce the view that these variants are indeed important modulators of expression. The list of genes whose expression differs among human groups and is correlated to copy number includes, amongst others, genes involved in xenobiotic detoxification (GSTT1), cellular immunity (APOBEC3B and HLA-DRB5) and hormone metabolism (UGT2B17) (Table 3 and Table S1). Although in some cases we have been able to corroborate the intuitive higher dose – higher expression hypothesis, the effect of CNVs on expression can be far more complex. Depending on the modification of the regulatory landscape of specific genes, it has been shown that CNVs can affect distant transcripts in cis or in trans, and even result in a negative correlation between expression and copy number (more copies, less expression) . In addition, the inherent complexity of the regulation of gene expression, with temporal and tissue-specific patterns, makes it difficult to evaluate the global effect of structural variation on expression from the available data coming from only one tissue. In our case, the total number of genes whose expression levels could be evaluated (i.e. because they had levels above the negative control threshold) in relation to copy number was only half of the total number of genes contained within the regions identified.
Despite these limitations, we have observed a number of population-specific copy number variants that affect genes associated to different phenotypes and disorders, and are very interesting candidates for future studies. For instance, we found that YRI have significantly lower copy number of the CFHR3 gene (Complement Factor H-Related gene 3) than CEU and ASN individuals. In addition, we have observed that this variant is well tagged by a SNP (rs16840639; r2 = 0.79) that shows a significant population differentiation and exhibits signatures of recent positive selection. Interestingly, the incidence of age-related macular degeneration (AMD) shows racial differences , , with lower prevalence among individuals of African origin. It is thus tempting to speculate about a connection between an AMD protective role of the deletion –, the population-specific pattern of AMD prevalence, and our findings about copy number population differences. Another attractive candidate is the haptoglobin-related protein gene (HPR), which is found in a CNV that exhibits a higher copy number in Yorubans than in the other two populations analyzed, and the linked SNPs also show high FST values among populations. The HPR is a key component of the serological innate immunity system that protects humans and other primates against Trypanosoma brucei brucei infection –, which causes sleeping sickness disease , . Therefore, it seems likely that the limited geographical distribution of the Trypanosoma brucei's vector, the TseTse fly, which is endemic to Africa , has influenced the elevated frequency of the high copy number alleles in individuals of the African population. Finally, elevated plasma levels of LPA gene product (lipoprotein A or Lp(a)) are associated with increased risk for atherosclerosis and its main manifestations: myocardial infarction, stroke and restenosis , –. In this direction, the increase in copy number of the LPA gene observed in the ASN population is associated with lower gene-expression levels (Table 3), which might explain, at least in part, the reduced risk of cardiovascular disease in individuals of this background.
Genes contained in CNV regions with population copy-number state frequency differences are enriched in functions related to interaction with the environment. For that reason, the different frequencies of these variants could be a sign of recent local adaptation events that have favored certain haplotypes over others in response to specific environmental conditions, related for example to dietary restriction or exposition to pathogens. By looking at the associated nucleotide variation, we have tried to discern if the differences could be attributable to some biological adaptation process driven by natural selection or just to random genetic drift. The diverse LD patterns among populations, the extensive population differentiation (high FST) and the presence of long haplotypes (high absolute values of iHS) point to recent positive selection acting to increase CNV frequency in a particular population. Nevertheless, in other cases, the available evidences suggest that genetic drift is the most likely responsible for the observed differences in CNV population frequency. Therefore, taken as a whole, the data obtained from the tag SNPs points that there is not a general pattern of selection for the different CNVs herein identified.
Our results highlight the importance of CNV differences among populations and, although more detailed studies would be needed to characterize their real functional consequences, provide a validated set of CNVs that might contribute to phenotypic and disease susceptibility differences among human populations. The identification of these variants emphasize the utility of current available platforms to evaluate copy-number differences genome-wide and pinpoint the need to develop specific technologies to reliably genotype CNVs in multiple individuals, since currently available array platforms and other proposed approaches (CNV-SNP tagging) might be hampered by the significant differences in population distribution of some CNVs. The work presented here represents a first step towards characterization of population differences in genome structure and will need to be revisited when denser and more precise tools to study structural variation are available. In addition, it would also be important to evaluate the impact on population-specific variation of balanced rearrangements not detected by aCGH, such as inversions, that, although putatively less frequent, could have a total genome coverage and potential contribution to variation at least as important as that of copy number changes , .
Materials and Methods
aCGH on pooled DNA samples from different populations
We performed array-based Comparative Genome Hybridization (aCGH) on pools of samples from different populations to detect CNVs. To obtain the pools, we used DNA extracted from lymphoblastoid cell lines from the HapMap collection (obtained from the Coriell Institute for Medical Research). The samples consisted of 50 unrelated individuals (25 males and 25 females) from four population groups: Yoruba from Ibadan, Nigeria (YRI), Caucasians of European descent from Utah, USA (CEU), Japanese from Tokyo, Japan (JPT), and Han Chinese from Beijing, China (CHB). In order to complete the 50 unrelated samples from Asian ancestry, we included individuals from HapMap Plate 05. Individual DNA quality was assessed by alkaline gel electrophoresis and the concentration was measured using the PicoGreen® assay (Invitrogen, Life technologies, Carlsbad, CA, US). 500 ng of DNA from each individual were combined to obtain each population-specific pool, which gave enough DNA for the several experiments using different array platforms.
Three different genome-wide platforms were used to perform the CNV detection: an in-house developed BAC-based array consisting of ~32.000 BAC-derived probes from the Human 32 K clone set library (kindly provided by Dr. Joris Veltman), a gene-centered oligonucleotide array of ~44.000 probes (Human Agilent 44 K; Agilent Technologies Inc, Santa Clara, CA, US) and a genome-wide oligonucleotide array of ~185.000 probes (Human Agilent 185 K). The hybridization, washing and scanning of the BAC-arrays was performed as described elsewhere , with minor modifications. Briefly, 400 ng of test and control DNA were labelled by random priming using the BioPrime Array CGH Genomic Labelling System (Invitrogen, Life technologies, Carlsbad, CA, US). The hybridization, washing and scanning of Agilent arrays was performed following manufacturer's instructions with minor modifications. For each array, each population-specific pool was hybridized against all the rest population pools in a sex-unmatched specific manner (i.e. the male pool from a certain population is hybridized with the pools of the opposite sex of the other populations - Figure 1A). In addition, we performed a sex-unmatched, intra-population hybridization to discard clone and oligonucleotide probes behaving abnormally. Arrays were scanned using an Agilent G2565BA MicroArray Scanner System (Agilent Technologies Inc., Santa Clara, CA, US) and the acquired images were analyzed using GenePix Pro 6.0 software (Axon, Molecular Devices, Toronto, Canada) using the irregular feature finding option. Extracted raw data was filtered and Lowess normalized using Bacanal , an in-house web server implementation of the Limma package developed within the Bioconductor project .
Genomic imbalances were determined based on the average log2 of the Cy5/Cy3 ratios of the spotted replicates, and regions were considered as amplified or deleted when probes exceeded the ±0.3 threshold. The design of the hybridizations allowed us to establish a priori combinations of hybridization intensity patterns to assign the most likely population to carry the variation, assuming the most parsimonious scenario (Figure 1B). The combination of hybridizations also allowed us to minimize the calling of CNVs on spurious positive signals. For instance, for a CNV to be called as a YRI-specific loss, the hybridizations of CEU, CHB and JPT male pools were required to show an increased hybridization signal with respect to the YRI female pool (log2 ratio≥0.3, green squares), the hybridization of the YRI male pool versus the rest of pools had to report decreased signals (log2 ratio≤−0.3, red squares), and the rest of combinations should reveal equal intensities (dark squares) (Figure 1B). Any number of consecutive probes that exhibited log2 ratios consistent with a population-specific CNV call after the analysis of the combination of hybridizations on the same platform was considered for further analysis. We discarded probes/clones that exhibited discordant log2 ratios among the different combinations. The high degree of similarity between the patterns obtained in the CHB and JPT samples prompted us to combine the data from both populations and to treat the two populations as a single one named ASN (for Asian). Data obtained from each of the three platforms was considered and analyzed independently.
Detection limits of aCGH
In order to estimate the lower limit of aCGH to detect copy number changes among samples, we set up an experiment in which we diluted DNA from an individual carrying a deletion of known size at 7q11.23 (typical Williams-Beuren Syndrome deletion) into a pool of normal DNA. We determined that we were able to call a copy number change when the alteration was present in between 20% and 30% of the chromosomes that participated in the hybridization (data not shown).
MLPA validation and data analysis
Oligonucleotides for each probe were designed following the recommendations of the proprietary of the MLPA technology (MRC-Holland) using the automated web application called MLPA ProSeek . A total of 133 locus-specific MLPA probes were combined in 5 different mixes to validate the findings. Each probe mix was composed of about 30 probes targeting selected genomic loci (Table S2). In the case of CNV loci reported by a single oligonucleotide probe in the array, the MLPA probe was designed within a window of 1 kb around the original position; otherwise, it was designed within the altered region. Each probe mix contained two extra control probes interrogating regions that are deleted and duplicated in the positive control samples that were included in each experiment, but are invariable in normal individuals. For validation, we analyzed all individual samples included in the pools plus the rest of individuals in HapMap plates 01 (90 CEU), 02 (90 ASN), 03 (90 YRI) and 05 (90 ASN) (Coriell). In addition, control samples were included in each plate: two DNAs with known aberrations (10q duplication, 22q deletion) plus 4 replicates of each of the male pools used in the hybridizations. MLPA was performed as described elsewhere , with minor modifications.
Although DNA concentration was controlled during the preparation of the plates, for data analysis each sample was normalized internally by dividing the height of each individual peak by total sum of peaks heights (dosage quotient). In order to compare results among plates, individual peak dosage quotients of each sample were normalized against the mean dosage quotients of the control samples included in each plate (pooled samples). Thus, a value of 1.0 for a particular probe after normalization corresponds to the population average copy number and does not necessarily denote the usual copy dosage of 2 of a given locus. In order to estimate the copy number state of each locus in each individual, we developed a novel method that will be implemented in the R-package called MLPAstats . In brief, since the MLPA dosage quotient ratios of the individuals at a certain locus are distributed according to a mixture of normal distributions, it is possible to assign to each individual a posterior probability of being located within one of the normal distributions that represent each copy number state. In order to assess if the copy number distribution was significantly different among populations, we performed a Fisher test after constructing contingency tables with copy number states and number of individuals assigned to each state. To determine the population in which a given CNV occurs, its distribution in each population was compared in a pairwise manner also using a Fisher's test (or χ2 test, depending on whether the number of individuals carrying a specific copy number exceeded a certain threshold).
Expression data for genes located within the identified CNVs was obtained from , available at ftp://ftp.sanger.ac.uk/pub/genevar/. We extracted the “globally” normalized information for unrelated individuals for whom we had estimated the CNV status. The existence of population-specific differences in expression levels of probes was determined by comparing the expression data using a Kruskal-Wallis test. In order to test for the relationship between copy number and expression levels we performed an ANOVA by fitting a linear model using the individuals' copy number estimated from MLPA and the expression data. Some of the genes located within CNV regions were not included in the analysis because their expression levels were below the average of negative control genes included in the array (bacterial genes lysA, pheA, thrB and trpF) across the different experiments (6.139), indicating that the hybridization to the array probes is below the level of detection.
Gene ontology analysis
We used the Gene Ontology Tree Machine (GOTM - http://bioinfo.vanderbilt.edu/gotm/) to measure the enrichment in GO categories of the genes located within the CNVs showing population differences in comparison with the rest of genes of the human genome and the subset of genes located within known CNVs. In order to identify GO categories with significantly enriched gene numbers, the GOTM compares the distribution of genes in the user's set in each GO category to those in the reference gene set. GOTM reports only those enrichments that are statistically significant as determined by the hypergeometric test. Further details on the statistical background for the comparison between gene sets are given on the original publication .
Selective constraints of genes located within CNVs
We obtained information on rates of synonymous (Ks) and non-synonymous (Ka) changes between humans and chimpanzees for genes located within the validated CNVs from Ensembl (http://www.ensembl.org/) and Khaitovich et al. . To check for the existence of significant Ka/Ks differences of our gene set, we compared the average and median values with the values obtained from 10000 permutations of the same number of RefSeq genes. The P-value was calculated as the number of times that the Ka/Ks values of the permuted sets equaled or exceeded the observed Ka/Ks divided by the total number of permutations plus one . Whole data set of Ka/Ks values used is provided in Tables S6 and S7.
SNPs and CNVs association
We used HapMart to retrieve the genotypes of all HapMap SNPs (excluding monomorphic ones) in a window of 100 kb at each side of the estimated CNV breakpoints, or around the position of the MLPA probe that we used for the validation of single oligonucleotide probe. The SNPassoc R package  was then used to measure LD by calculating r2 between the set of SNP genotypes and the copy number alleles. LD could only be calculated for biallelic copy number variants. In addition, in some cases, LD value could not be calculated because the CNV was identified exclusively in non-unrelated HapMap individuals or in individuals of Asian ancestry that were used in our analysis but have not been genotyped as part of the HapMap. Pairwise comparison of r2 LD patterns among different populations was done using the Spearman's rank correlation coefficient.
Evidences of natural selection on identified CNVs
For the tag-SNPs showing the highest linkage disequilibrium with the identified copy number variants in the population where the CNV was identified, we obtained the global and pairwise FST values among populations for the Phase II HapMap SNPs from . To detect population differentiation, we assumed a relatively strict threshold of global FST≥0.2, and the comparison of pairwise FST values (i.e. CEU vs YRI, CEU vs ASN, and YRI vs ASN) for a given tag-SNP pointed out which population was more clearly differentiated.
To test for signals of recent positive selection we compared, among populations, the integrated Haplotype Score (iHS) of the SNP that best tagged the variant in the population in which the CNV differences were found. Absolute values of iHS over 2 are generally accepted as signals of recent positive selection . iHS data for HapMap II SNPs was retrieved from the Haplotter website (http://hg-wen.uchicago.edu/selection/index.html).
Details of regions interrogated with the MLPA probes and summary of MLPA results obtained for each probe. Conservation was only explored in regions that were confirmed to be CNVs and did not contain genes. WGTP stands for Whole Genome Tiling Path Array.
(0.16 MB XLS)
Details of MLPA probes used in the study. Genomic positions are referred to the hg18 UCSC genome assembly. MLPA probes are composed of two oligonucleotides that would only render an amplifiable product if properly ligated when bound to the specific DNA sequence. Right oligonucleotide sequences are entirely complementary to the target locus, while left oligonucleotide sequences are formed by the locus-specific sequence plus a stuffer sequence that would allow distinguishing each product in the mix. In the left sequence, stuffer sequence is lowercase and a dash separates it from the specific sequence, that is either upper or lowercase depending on the coincidence with a repeat identified by Repeat Masker. VIC, NED and FAM are the fluorophores that were used to label the specific universal primers combined in each probe mix.
(0.05 MB XLS)
Comparison of CNVs identified in this study and previsouly existing information on the Database of Genomic Variants (DGV). The agreement column states whether our findings agree with previously published results (Agree), do not agree (Disagree), or could not be verified (Unverifiable) either because the CNV herein identified was not previously reported or there were inconsistencies in the description of the CNV in the original paper or in the Database of Genomic Variants.
(0.04 MB XLS)
Gene Ontology Analysis (part I). Gene ontology analysis of genes located within the identified CNVs reveals enrichment in different functional categories when compared to the rest of human genes.
(0.02 MB XLS)
Gene Ontology Analysis (part II). Gene ontology analysis of genes located within the identified CNVs reveals enrichment in different functional categories when compared to genes located within known CNVs.
(0.01 MB XLS)
Comparison of Ka/Ks ratios of genes located within population-specific CNV changes. For the Ensembl dataset, mean and median values are calculated from dN/dS. For the calculations on the Khaitovich dataset, we used the Ka/Ki measure. P-values were calculated as the number of times that the observed mean value exceeded or equaled the value in the permutation analysis out of 10,000 permutations.
(0.02 MB XLS)
Indexes of molecular evolution. All evolutionary measures were published or available elsewhere as described in the Methods section. Values were calculated by comparing human versus chimpanzee sequences. Ka is the number of non-synonymous nucleotide substitutions per non-synonymous site. Ks is the number of synonymous nucleotide substitutions per synonymous site. Ki is the number of substitutions per site in interspersed repeats in a 250-kbp window around the center of each gene (see Khaitovich et. al, 2005). dn/ds: number of non-synonymous substitutions per non-synonymous site divided by the number synonymous substitutions per silent site.
(0.02 MB XLS)
Linkage disequilibrium for SNPs and CNVs. In the table it is reflected the calculated r-squared value (r2) and the SNP showing the maximum LD with the copy number state revealed by the MLPA probes in the first column. Values were obtained for each population independently and considering all populations together. The column SNPs indicates the number of HapMap SNPs retrieved for each analysed region flanking the CNV. For three of the probes it was not possible to measure LD because more than three copy number states were present (**) or because no SNP data was available for individuals showing the CNV (***).
(0.03 MB XLS)
Haplotype structure correlation between populations. LD patterns correlation between populations was measured using the Spearman's rank correlation test. High Rho values indicate correlation and p-values below 0,05 denote significance for the correlation. For three of the probes it was not possible to measure LD because more than three copy number states were present (**) or because no SNP data was available for individuals showing the CNV (***).
(0.02 MB XLS)
Evaluation of selective pressures on SNPs tagging CNVs. SNPs having the highest LD in the population in which the CNV variant was identified are shown in the table. We have recovered FST values calculated after pairwise population comparisons from Barreiro et al, (2008). The Inferred Pop column indicates to which population is attributable the differentiation, according to pairwise comparisons of the SNP FST. We used a strict criterion and assumed values above 0.20 to be indicative of population differentiation. iHS stands for integrated Haplotype Score (Voight et al, 2006) and absolute values higher than 2 indicate recent selection (marked with two asterisks). *: denotes coincidence of the most differentiated population with the population in which de CNV was identified. NA: indicates that we could not find Fst calculations or iHS value for the given SNP. None: indicates that we can not identify which is the differentiated population using the observed values. It is also used to indicate that no population has values above the iHS threshold considered for recent selection. **: indicates an absolute value of iHS above 2.
(0.07 MB XLS)
Evaluation of copy number in large CNVs. MLPA probes located within the same theoretical CNV do not show the same copy number pattern neither, in consequence, the same correlation with gene expression levels.
(0.01 MB PNG)
Correlation between copy number changes and expression levels. Each coloured circle corresponds to the measured copy number state plotted against expression levels for each HapMap individual. Gene name for which expression levels and copy numbers are depicted is shown in the title of each figure. Left panel (A) shows four examples of CNVs identified in which copy number is correlated with gene expression. Right panel (B) exemplifies four cases in which copy number differences exist among individuals from different populations, but it does not translate into gene expression differences.
(5.57 MB TIF)
We would like to thank Luis Barreiro and Lluís Quintana-Murci for kindly providing FST values calculations for the HapMap SNPs. We would also like to thank Ferran Casals for his advice regarding molecular evolution analyses.
Conceived and designed the experiments: LA MC XE. Performed the experiments: LA SV MGA MC. Analyzed the data: LA SV JRG MC. Contributed reagents/materials/analysis tools: LA LP MC. Wrote the paper: LA JRG RR MC XE.
- 1. Lander ES, Linton LM, Birren B, Nusbaum C, Zody MC, et al. (2001) Initial sequencing and analysis of the human genome. Nature 409: 860–921. doi: 10.1038/35057062
- 2. Venter JC, Adams MD, Myers EW, Li PW, Mural RJ, et al. (2001) The sequence of the human genome. Science 291: 1304–1351. doi: 10.1126/science.1058040
- 3. Bamshad M, Wooding S, Salisbury BA, Stephens JC (2004) Deconstructing the relationship between genetics and race. Nat Rev Genet 5: 598–609. doi: 10.1038/nrg1401
- 4. Daar AS, Singer PA (2005) Pharmacogenetics and geographical ancestry: implications for drug development and global health. Nat Rev Genet 6: 241–246. doi: 10.1038/nrg1559
- 5. Burchard EG, Ziv E, Coyle N, Gomez SL, Tang H, et al. (2003) The importance of race and ethnic background in biomedical research and clinical practice. N Engl J Med 348: 1170–1175. doi: 10.1056/NEJMsb025007
- 6. Botstein D, Risch N (2003) Discovering genotypes underlying human phenotypes: past successes for mendelian disease, future approaches for complex disease. Nat Genet 33: Suppl228–237. doi: 10.1038/ng1090
- 7. (1994) (1994) Population variation of common cystic fibrosis mutations. The Cystic Fibrosis Genetic Analysis Consortium. Hum Mutat 4: 167–177. doi: 10.1002/humu.1380040302
- 8. Estivill X, Bancells C, Ramos C (1997) Geographic distribution and regional origin of 272 cystic fibrosis mutations in European populations. The Biomed CF Mutation Analysis Consortium. Hum Mutat 10: 135–154. doi: 10.1002/(SICI)1098-1004(1997)10:2<135::AID-HUMU6>3.0.CO;2-J
- 9. Bobadilla JL, Macek M Jr, Fine JP, Farrell PM (2002) Cystic fibrosis: a worldwide analysis of CFTR mutations–correlation with incidence data and application to screening. Hum Mutat 19: 575–606. doi: 10.1002/humu.10041
- 10. Antonarakis SE, Boehm CD, Serjeant GR, Theisen CE, Dover GJ, et al. (1984) Origin of the beta S-globin gene in blacks: the contribution of recurrent mutation or gene conversion or both. Proc Natl Acad Sci U S A 81: 853–856. doi: 10.1073/pnas.81.3.853
- 11. Kulozik AE, Wainscoat JS, Serjeant GR, Kar BC, Al-Awamy B, et al. (1986) Geographical survey of beta S-globin gene haplotypes: evidence for an independent Asian origin of the sickle-cell mutation. Am J Hum Genet 39: 239–244.
- 12. Lapoumeroulie C, Dunda O, Ducrocq R, Trabuchet G, Mony-Lobe M, et al. (1992) A novel sickle cell mutation of yet another origin in Africa: the Cameroon type. Hum Genet 89: 333–337. doi: 10.1007/bf00220553
- 13. Nagel RL, Fabry ME, Pagnier J, Zohoun I, Wajcman H, et al. (1985) Hematologically and genetically distinct forms of sickle cell anemia in Africa. The Senegal type and the Benin type. N Engl J Med 312: 880–884. doi: 10.1056/NEJM198504043121403
- 14. Pagnier J, Mears JG, Dunda-Belkhodja O, Schaefer-Rego KE, Beldjord C, et al. (1984) Evidence for the multicentric origin of the sickle cell hemoglobin gene in Africa. Proc Natl Acad Sci U S A 81: 1771–1773. doi: 10.1073/pnas.81.6.1771
- 15. Gasparini P, Rabionet R, Barbujani G, Melchionda S, Petersen M, et al. (2000) High carrier frequency of the 35delG deafness mutation in European populations. Genetic Analysis Consortium of GJB2 35delG. Eur J Hum Genet 8: 19–23. doi: 10.1038/sj.ejhg.5200406
- 16. Ridker PM, Miletich JP, Hennekens CH, Buring JE (1997) Ethnic distribution of factor V Leiden in 4047 men and women. Implications for venous thromboembolism screening. Jama 277: 1305–1307. doi: 10.1001/jama.277.16.1305
- 17. Shen MC, Lin JS, Tsay W (1997) High prevalence of antithrombin III, protein C and protein S deficiency, but no factor V Leiden mutation in venous thrombophilic Chinese patients in Taiwan. Thromb Res 87: 377–385. doi: 10.1016/S0049-3848(97)00141-2
- 18. Hugot JP, Chamaillard M, Zouali H, Lesage S, Cezard JP, et al. (2001) Association of NOD2 leucine-rich repeat variants with susceptibility to Crohn's disease. Nature 411: 599–603. doi: 10.1038/35079107
- 19. Yamazaki K, Takazoe M, Tanaka T, Kazumori T, Nakamura Y (2002) Absence of mutation in the NOD2/CARD15 gene among 483 Japanese patients with Crohn's disease. J Hum Genet 47: 469–472. doi: 10.1007/s100380200067
- 20. Stephens JC, Reich DE, Goldstein DB, Shin HD, Smith MW, et al. (1998) Dating the origin of the CCR5-Delta32 AIDS-resistance allele by the coalescence of haplotypes. Am J Hum Genet 62: 1507–1515. doi: 10.1086/301867
- 21. Farrer LA, Cupples LA, Haines JL, Hyman B, Kukull WA, et al. (1997) Effects of age, sex, and ethnicity on the association between apolipoprotein E genotype and Alzheimer disease. A meta-analysis. APOE and Alzheimer Disease Meta Analysis Consortium. Jama 278: 1349–1356. doi: 10.1001/jama.278.16.1349
- 22. Vatsis KP, Martell KJ, Weber WW (1991) Diverse point mutations in the human gene for polymorphic N-acetyltransferase. Proc Natl Acad Sci U S A 88: 6333–6337. doi: 10.1073/pnas.88.14.6333
- 23. Bowcock AM, Kidd JR, Mountain JL, Hebert JM, Carotenuto L, et al. (1991) Drift, admixture, and selection in human evolution: a study with DNA polymorphisms. Proc Natl Acad Sci U S A 88: 839–843. doi: 10.1073/pnas.88.3.839
- 24. Bowcock AM, Ruiz-Linares A, Tomfohrde J, Minch E, Kidd JR, et al. (1994) High resolution of human evolutionary trees with polymorphic microsatellites. Nature 368: 455–457. doi: 10.1038/368455a0
- 25. Ramachandran S, Deshpande O, Roseman CC, Rosenberg NA, Feldman MW, et al. (2005) Support from the relationship of genetic and geographic distance in human populations for a serial founder effect originating in Africa. Proc Natl Acad Sci U S A 102: 15942–15947. doi: 10.1073/pnas.0507611102
- 26. Rosenberg NA, Pritchard JK, Weber JL, Cann HM, Kidd KK, et al. (2002) Genetic structure of human populations. Science 298: 2381–2385. doi: 10.1126/science.1078311
- 27. HapMap C (2005) A haplotype map of the human genome. Nature 437: 1299–1320. doi: 10.1038/nature04226
- 28. Frazer KA, Ballinger DG, Cox DR, Hinds DA, Stuve LL, et al. (2007) A second generation human haplotype map of over 3.1 million SNPs. Nature 449: 851–861. doi: 10.1038/nature06258
- 29. Jakobsson M, Scholz SW, Scheet P, Gibbs JR, VanLiere JM, et al. (2008) Genotype, haplotype and copy-number variation in worldwide human populations. Nature 451: 998–1003. doi: 10.1038/nature06742
- 30. Barbujani G, Magagni A, Minch E, Cavalli-Sforza LL (1997) An apportionment of human DNA diversity. Proc Natl Acad Sci U S A 94: 4516–4519. doi: 10.1073/pnas.94.9.4516
- 31. Watkins WS, Rogers AR, Ostler CT, Wooding S, Bamshad MJ, et al. (2003) Genetic variation among world populations: inferences from 100 Alu insertion polymorphisms. Genome Res 13: 1607–1618. doi: 10.1101/gr.894603
- 32. Beckmann JS, Estivill X, Antonarakis SE (2007) Copy number variants and genetic traits: closer to the resolution of phenotypic to genotypic variability. Nat Rev Genet 8: 639–646. doi: 10.1038/nrg2149
- 33. Hurles ME, Dermitzakis ET, Tyler-Smith C (2008) The functional impact of structural variation in humans. Trends Genet 24: 238–245. doi: 10.1016/j.tig.2008.03.001
- 34. Tuzun E, Sharp AJ, Bailey JA, Kaul R, Morrison VA, et al. (2005) Fine-scale structural variation of the human genome. Nat Genet 37: 727–732. doi: 10.1038/ng1562
- 35. Redon R, Ishikawa S, Fitch KR, Feuk L, Perry GH, et al. (2006) Global variation in copy number in the human genome. Nature 444: 444–454. doi: 10.1038/nature05329
- 36. Kidd JM, Cooper GM, Donahue WF, Hayden HS, Sampas N, et al. (2008) Mapping and sequencing of structural variation from eight human genomes. Nature 453: 56–64. doi: 10.1038/nature06862
- 37. Korbel JO, Urban AE, Affourtit JP, Godwin B, Grubert F, et al. (2007) Paired-end mapping reveals extensive structural variation in the human genome. Science 318: 420–426. doi: 10.1126/science.1149504
- 38. Stranger BE, Forrest MS, Dunning M, Ingle CE, Beazley C, et al. (2007) Relative impact of nucleotide and copy number variation on gene expression phenotypes. Science 315: 848–853. doi: 10.1126/science.1136678
- 39. Stranger BE, Nica AC, Forrest MS, Dimas A, Bird CP, et al. (2007) Population genomics of human gene expression. Nat Genet 39: 1217–1224. doi: 10.1038/ng2142
- 40. de Stahl TD, Sandgren J, Piotrowski A, Nord H, Andersson R, et al. (2008) Profiling of copy number variations (CNVs) in healthy individuals from three ethnic groups using a human genome 32 K BAC-clone-based array. Hum Mutat 29: 398–408. doi: 10.1002/humu.20659
- 41. Perry GH, Ben-Dor A, Tsalenko A, Sampas N, Rodriguez-Revenga L, et al. (2008) The fine-scale and complex architecture of human copy-number variation. Am J Hum Genet 82: 685–695. doi: 10.1016/j.ajhg.2007.12.010
- 42. McCarroll SA, Kuruvilla FG, Korn JM, Cawley S, Nemesh J, et al. (2008) Integrated detection and population-genetic analysis of SNPs and copy number variation. Nat Genet 40: 1166–1174. doi: 10.1038/ng.238
- 43. Armour JA, Palla R, Zeeuwen PL, den Heijer M, Schalkwijk J, et al. (2007) Accurate, high-throughput typing of copy number variation using paralogue ratios from dispersed repeats. Nucleic Acids Res 35: e19. doi: 10.1093/nar/gkl1089
- 44. Kidd JM, Newman TL, Tuzun E, Kaul R, Eichler EE (2007) Population stratification of a common APOBEC gene deletion polymorphism. PLoS Genet 3: e63. doi: 10.1371/journal.pgen.0030063
- 45. White SJ, Vissers LE, Geurts van Kessel A, de Menezes RX, Kalay E, et al. (2007) Variation of CNV distribution in five different ethnic populations. Cytogenet Genome Res 118: 19–30. doi: 10.1159/000106437
- 46. Armengol L, Rabionet R, Estivill X (2008) The emerging role of structural variations in common disorders: initial findings and discovery challenges. Cytogenet Genome Res 123: 108–117. doi: 10.1159/000184698
- 47. Gimelli G, Pujana MA, Patricelli MG, Russo S, Giardino D, et al. (2003) Genomic inversions of human chromosome 15q11–q13 in mothers of Angelman syndrome patients with class II (BP2/3) deletions. Hum Mol Genet 12: 849–858. doi: 10.1093/hmg/ddg101
- 48. Osborne LR, Li M, Pober B, Chitayat D, Bodurtha J, et al. (2001) A 1.5 million-base pair inversion polymorphism in families with Williams-Beuren syndrome. Nat Genet 29: 321–325. doi: 10.1038/ng753
- 49. Stankiewicz P, Lupski JR (2002) Genome architecture, rearrangements and genomic disorders. Trends Genet 18: 74–82. doi: 10.1016/S0168-9525(02)02592-1
- 50. Stankiewicz P, Shaw CJ, Dapper JD, Wakui K, Shaffer LG, et al. (2003) Genome architecture catalyzes nonrecurrent chromosomal rearrangements. Am J Hum Genet 72: 1101–1116. doi: 10.1086/374385
- 51. Pinkel D, Segraves R, Sudar D, Clark S, Poole I, et al. (1998) High resolution analysis of DNA copy number variation using comparative genomic hybridization to microarrays. Nat Genet 20: 207–211. doi: 10.1038/2524
- 52. Solinas-Toldo S, Lampel S, Stilgenbauer S, Nickolenko J, Benner A, et al. (1997) Matrix-based comparative genomic hybridization: biochips to screen for genomic imbalances. Genes Chromosomes Cancer 20: 399–407. doi: 10.1002/(SICI)1098-2264(199712)20:4<399::AID-GCC12>3.0.CO;2-I
- 53. Conrad DF, Andrews TD, Carter NP, Hurles ME, Pritchard JK (2006) A high-resolution survey of deletion polymorphism in the human genome. Nat Genet 38: 75–81. doi: 10.1038/ng1697
- 54. Schouten JP, McElgunn CJ, Waaijer R, Zwijnenburg D, Diepvens F, et al. (2002) Relative quantification of 40 nucleic acid sequences by multiplex ligation-dependent probe amplification. Nucleic Acids Res 30: e57. doi: 10.1093/nar/gnf056
- 55. Iafrate AJ, Feuk L, Rivera MN, Listewnik ML, Donahoe PK, et al. (2004) Detection of large-scale variation in the human genome. Nat Genet 36: 949–951. doi: 10.1038/ng1416
- 56. Price AL, Patterson NJ, Plenge RM, Weinblatt ME, Shadick NA, et al. (2006) Principal components analysis corrects for stratification in genome-wide association studies. Nat Genet 38: 904–909. doi: 10.1038/ng1847
- 57. Ovcharenko I, Nobrega MA, Loots GG, Stubbs L (2004) ECR Browser: a tool for visualizing and accessing data from comparisons of multiple vertebrate genomes. Nucleic Acids Res 32: W280–286. doi: 10.1093/nar/gkh355
- 58. Zhang B, Schmoyer D, Kirov S, Snoddy J (2004) GOTree Machine (GOTM): a web-based platform for interpreting sets of interesting genes using Gene Ontology hierarchies. BMC Bioinformatics 5: 16. doi: 10.1186/1471-2105-5-16
- 59. Cooper GM, Nickerson DA, Eichler EE (2007) Mutational and selective effects on copy-number variants in the human genome. Nat Genet 39: S22–29. doi: 10.1038/ng2054
- 60. Feuk L, Carson AR, Scherer SW (2006) Structural variation in the human genome. Nat Rev Genet 7: 85–97. doi: 10.1038/nrg1767
- 61. Nguyen DQ, Webber C, Ponting CP (2006) Bias of selection on human copy-number variants. PLoS Genet 2: e20. doi: 10.1371/journal.pgen.0020020
- 62. Sharp AJ, Cheng Z, Eichler EE (2006) Structural variation of the human genome. Annu Rev Genomics Hum Genet 7: 407–442. doi: 10.1146/annurev.genom.7.080505.115618
- 63. Edwards AO, Ritter R 3rd, Abel KJ, Manning A, Panhuysen C, et al. (2005) Complement factor H polymorphism and age-related macular degeneration. Science 308: 421–424. doi: 10.1126/science.1110189
- 64. Hageman GS, Hancox LS, Taiber AJ, Gehrs KM, Anderson DH, et al. (2006) Extended haplotypes in the complement factor H (CFH) and CFH-related (CFHR) family of genes protect against age-related macular degeneration: characterization, ethnic distribution and evolutionary implications. Ann Med 38: 592–604. doi: 10.1080/07853890601097030
- 65. Haines JL, Hauser MA, Schmidt S, Scott WK, Olson LM, et al. (2005) Complement factor H variant increases the risk of age-related macular degeneration. Science 308: 419–421. doi: 10.1126/science.1110359
- 66. Hughes AE, Orr N, Esfandiary H, Diaz-Torres M, Goodship T, et al. (2006) A common CFH haplotype, with deletion of CFHR1 and CFHR3, is associated with lower risk of age-related macular degeneration. Nat Genet 38: 1173–1177. doi: 10.1038/ng1890
- 67. Spencer KL, Hauser MA, Olson LM, Schmidt S, Scott WK, et al. (2008) Deletion of CFHR3 and CFHR1 genes in age-related macular degeneration. Hum Mol Genet 17: 971–977. doi: 10.1093/hmg/ddm369
- 68. Brazier L, Tiret L, Luc G, Arveiler D, Ruidavets JB, et al. (1999) Sequence polymorphisms in the apolipoprotein(a) gene and their association with lipoprotein(a) levels and myocardial infarction. The ECTIM Study. Atherosclerosis 144: 323–333. doi: 10.1016/S0021-9150(98)00333-5
- 69. Celik T, Yuksel C, Iyisoy A, Isik E (2008) Lipoprotein(a) and coronary artery disease: Ongoing debate. Nutr Metab Cardiovasc Dis.
- 70. Gaeta G, Cuomo S, Capozzi G, Foglia MC, Barra S, et al. (2007) Lipoprotein(a) levels are increased in healthy young subjects with parental history of premature myocardial infarction. Nutr Metab Cardiovasc Dis.
- 71. Hahmann HW, Schatzer-Klotz D, Bunte T, Becker D, Schieffer HJ (1999) The significance of high levels of lipoprotein (a) compared with established risk factors in premature coronary artery disease: differences between men and women. Atherosclerosis 144: 221–228. doi: 10.1016/S0021-9150(99)00057-X
- 72. Ogorelkova M, Gruber A, Utermann G (1999) Molecular basis of congenital lp(a) deficiency: a frequent apo(a) ‘null’ mutation in caucasians. Hum Mol Genet 8: 2087–2096. doi: 10.1093/hmg/8.11.2087
- 73. Ogorelkova M, Kraft HG, Ehnholm C, Utermann G (2001) Single nucleotide polymorphisms in exons of the apo(a) kringles IV types 6 to 10 domain affect Lp(a) plasma concentrations and have different patterns in Africans and Caucasians. Hum Mol Genet 10: 815–824. doi: 10.1093/hmg/10.8.815
- 74. Zhang X, Mahmudi-Azer S, Connett JE, Anthonisen NR, He JQ, et al. (2007) Association of Hck genetic polymorphisms with gene expression and COPD. Hum Genet 120: 681–690. doi: 10.1007/s00439-006-0253-7
- 75. Spielman RS, Bastone LA, Burdick JT, Morley M, Ewens WJ, et al. (2007) Common genetic variants account for differences in gene expression among ethnic groups. Nat Genet 39: 226–231. doi: 10.1038/ng1955
- 76. McCarroll SA, Hadnott TN, Perry GH, Sabeti PC, Zody MC, et al. (2006) Common deletion polymorphisms in the human genome. Nat Genet 38: 86–92. doi: 10.1038/ng1696
- 77. Cooper GM, Zerr T, Kidd JM, Eichler EE, Nickerson DA (2008) Systematic assessment of copy number variant detection via genome-wide SNP genotyping. Nat Genet 40: 1199–1203. doi: 10.1038/ng.236
- 78. Locke DP, Sharp AJ, McCarroll SA, McGrath SD, Newman TL, et al. (2006) Linkage disequilibrium and heritability of copy-number polymorphisms within duplicated regions of the human genome. Am J Hum Genet 79: 275–290. doi: 10.1086/505653
- 79. Ke X, Hunt S, Tapper W, Lawrence R, Stavrides G, et al. (2004) The impact of SNP density on fine-scale patterns of linkage disequilibrium. Hum Mol Genet 13: 577–588. doi: 10.1093/hmg/ddh060
- 80. Excoffier L, Smouse PE, Quattro JM (1992) Analysis of molecular variance inferred from metric distances among DNA haplotypes: application to human mitochondrial DNA restriction data. Genetics 131: 479–491.
- 81. Weir BS, Hill WG (2002) Estimating F-statistics. Annu Rev Genet 36: 721–750. doi: 10.1146/annurev.genet.36.050802.093940
- 82. Nielsen R (2005) Molecular signatures of natural selection. Annu Rev Genet 39: 197–218. doi: 10.1146/annurev.genet.39.073003.112420
- 83. Weir BS, Cardon LR, Anderson AD, Nielsen DM, Hill WG (2005) Measures of human population structure show heterogeneity among genomic regions. Genome Res 15: 1468–1476. doi: 10.1101/gr.4398405
- 84. Barreiro LB, Laval G, Quach H, Patin E, Quintana-Murci L (2008) Natural selection has driven population differentiation in modern humans. Nat Genet 40: 340–345. doi: 10.1038/ng.78
- 85. Sabeti PC, Schaffner SF, Fry B, Lohmueller J, Varilly P, et al. (2006) Positive natural selection in the human lineage. Science 312: 1614–1620. doi: 10.1126/science.1124309
- 86. Voight BF, Kudaravalli S, Wen X, Pritchard JK (2006) A map of recent positive selection in the human genome. PLoS Biol 4: e72. doi: 10.1371/journal.pbio.0040072
- 87. Gonzalez E, Kulkarni H, Bolivar H, Mangano A, Sanchez R, et al. (2005) The influence of CCL3L1 gene-containing segmental duplications on HIV-1/AIDS susceptibility. Science 307: 1434–1440. doi: 10.1126/science.1101160
- 88. Bressler SB, Munoz B, Solomon SD, West SK (2008) Racial differences in the prevalence of age-related macular degeneration: the Salisbury Eye Evaluation (SEE) Project. Arch Ophthalmol 126: 241–245. doi: 10.1001/archophthalmol.2007.53
- 89. Chang MA, Bressler SB, Munoz B, West SK (2008) Racial differences and other risk factors for incidence and progression of age-related macular degeneration: Salisbury Eye Evaluation (SEE) Project. Invest Ophthalmol Vis Sci 49: 2395–2402. doi: 10.1167/iovs.07-1584
- 90. Drain J, Bishop JR, Hajduk SL (2001) Haptoglobin-related protein mediates trypanosome lytic factor binding to trypanosomes. J Biol Chem 276: 30254–30260. doi: 10.1074/jbc.M010198200
- 91. Vanhollebeke B, De Muylder G, Nielsen MJ, Pays A, Tebabi P, et al. (2008) A haptoglobin-hemoglobin receptor conveys innate immunity to Trypanosoma brucei in humans. Science 320: 677–681. doi: 10.1126/science.1156296
- 92. Vanhollebeke B, Nielsen MJ, Watanabe Y, Truc P, Vanhamme L, et al. (2007) Distinct roles of haptoglobin-related protein and apolipoprotein L-I in trypanolysis by human serum. Proc Natl Acad Sci U S A 104: 4118–4123. doi: 10.1073/pnas.0609902104
- 93. Pays E, Vanhollebeke B, Vanhamme L, Paturiaux-Hanocq F, Nolan DP, et al. (2006) The trypanolytic factor of human serum. Nat Rev Microbiol 4: 477–486. doi: 10.1038/nrmicro1428
- 94. Vanhamme L, Paturiaux-Hanocq F, Poelvoorde P, Nolan DP, Lins L, et al. (2003) Apolipoprotein L-I is the trypanosome lytic factor of human serum. Nature 422: 83–87. doi: 10.1038/nature01461
- 95. Krafsur ES (2003) Tsetse fly population genetics: an indirect approach to dispersal. Trends Parasitol 19: 162–166. doi: 10.1016/S1471-4922(03)00034-5
- 96. Buckley PG, Mantripragada KK, Benetkiewicz M, Tapia-Paez I, Diaz De Stahl T, et al. (2002) A full-coverage, high-resolution human chromosome 22 genomic microarray for clinical and research applications. Hum Mol Genet 11: 3221–3229. doi: 10.1093/hmg/11.25.3221
- 97. Lozano J Unpublished.
- 98. Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, et al. (2004) Bioconductor: open software development for computational biology and bioinformatics. Genome Biol 5: R80. doi: 10.1186/gb-2004-5-10-r80
- 99. Pantano L, Armengol L, Villatoro S, Estivill X (2008) ProSeeK: A web server for MLPA probe design. BMC Genomics 9: 573. doi: 10.1186/1471-2164-9-573
- 100. Gonzalez JR, Carrasco JL, Armengol L, Vilaroto S, Jover L, et al. (2008) Probe-specific mixed-model approach to detect copy number differences using multiplex ligation-dependent probe amplification (MLPA). BMC Bioinformatics 9: 261. doi: 10.1186/1471-2105-9-261
- 101. Khaitovich P, Hellmann I, Enard W, Nowick K, Leinweber M, et al. (2005) Parallel patterns of evolution in the genomes and transcriptomes of humans and chimpanzees. Science 309: 1850–1854. doi: 10.1126/science.1108296
- 102. Good PI (2000) Permutation Tests. A Practical Guide to Resampling Methods for Testing Hypotheses. Berlin-Heidelberg-New York: Springer-Verlag.
- 103. Gonzalez JR, Armengol L, Sole X, Guino E, Mercader JM, et al. (2007) SNPassoc: an R package to perform whole genome association studies. Bioinformatics 23: 644–645. doi: 10.1093/bioinformatics/btm025