This page has been archived and is no longer updated
Integrated transcriptional profiling and linkage analysis for identification of genes underlying disease.
Author: Hubner, N.
Keywords
Keywords for this Article
Add keywords to your Content
Save
|
Cancel
Share
|
Cancel
Revoke
|
Cancel
Rate & Certify
Rate Me...
Rate Me
!
Comment
Save
|
Cancel
Flag Inappropriate
The Content is
Objectionable
Explicit
Offensive
Inaccurate
Comment
Flag Content
|
Cancel
Delete Content
Reason
Delete
|
Cancel
Close
Full Screen
"Integrated transcriptional profiling and linkage analysis for identification of genes underlying disease Norbert Hubner 1 , Caroline A Wallace 2,7 , Heike Zimdahl 1,7 , Enrico Petretto 2,7 , Herbert Schulz 1 , Fiona Maciver 2 , Michael Mueller 2 , Oliver Hummel 1 , Jan Monti 1 , Vaclav Zidek 3 , Alena Musilova 3 , Vladimir Kren 3,4 , Helen Causton 2 , Laurence Game 2 , Gabriele Born 1 , Sabine Schmidt 1 , Anita Mu�ller 1 , Stuart A Cook 2 , Theodore W Kurtz 5 , John Whittaker 6 , Michal Pravenec 3,4 & Timothy J Aitman 2 Integration of genome-wide expression profiling with linkage analysis is a new approach to identifying genes underlying complex traits. We applied this approach to the regulation of gene expression in the BXH/HXB panel of rat recombinant inbred strains, one of the largest available rodent recombinant inbred panels and a leading resource for genetic analysis of the highly prevalent metabolic syndrome. In two tissues important to the pathogenesis of the metabolic syndrome, we mapped cis-andtrans- regulatory control elements for expression of thousands of genes across the genome. Many of the most highly linked expression quantitative trait loci are regulated in cis, are inherited essentially as monogenic traits and are good candidate genes for previously mapped physiological quantitative trait loci in the rat. By comparative mapping we generated a data set of 73 candidate genes for hypertension that merit testing in human populations. Mining of this publicly available data set is expected to lead to new insights into the genes and regulatory pathways underlying the extensive range of metabolic and cardiovascular disease phenotypes that segregate in these recombinant inbred strains. Determining the molecular basis of natural phenotypic variation, including interindividual susceptibility to common diseases, is a central challenge of postgenome genetics. The availability of genome sequences and genome-scale technologies has enabled new strategies for identifying genes underlying complex phenotypes and has greatly accelerated progress in this field. The high heritability of variation in gene expression 1 has suggested that identification of the genetic determinants of gene expression may give insights into the molecular basis of complex traits. One justifica- tion for studying the genetics of gene expression is that transcript abundance may act as an intermediate phenotype between genomic DNA sequence variation and more complex whole-body phenotypes. Mapping of gene expression levels as quantitative trait loci (called eQTLs) has been undertaken in yeast 2,3 and more recently in mam- mals 4,5 and has shown that the approach is feasible. As a tool for studying disease phenotypes, however, aside from a single study in F 2 mice 4 , the approach has not been extensively tested. For the past 50 years, the rat has been a leading model for the study of common, complex human diseases 6,7 . The availability of the rat genome sequence 8 has made feasible studies of gene expression at the level of the genome alongside well-characterized rat phenotypes. The spontaneously hypertensive rat (SHR) is a widely studied model of human hypertension and also has many features of the metabolic syndrome 9?13 . In the early 1980s, the SHR strain was crossed with the normotensive Brown Norway (BN) strain to generate the BXH/HXB panel of recombinant inbred (RI) strains 14,15 . Rodent RI panels are powerful and permanent resources for genetic mapping that offer the opportunity to accumulate genetic and physiological data over time 16 . A further advantage of RI strains, as with chromosome substitution strains 17,18 , is the ability to study genetically identical biological replicates, which increases trait heritability by reducing environmental variance 19 . Here we applied integrated gene expression profiling and linkage analysis to the regulation of gene expression in fat and kidney tissue in the BXH/HXB panel of rat RI strains. We found that these RI strains are a suitable genetic system for large-scale identification of posi- tional candidates and regulatory pathways for previously mapped physiological QTLs (called pQTLs). By comparative mapping, we compiled a data set of candidate genes for investigating the molecular basis of human hypertension. By identifying hundreds of robustly mapped cis-andtrans-acting eQTLs in a model system with large numbers of existing pQTLs, we generated a unique Published online 13 February 2005; doi:10.1038/ng1522 1 Max-Delbru�ck-Center for Molecular Medicine, Berlin-Buch 13125, Germany. 2 MRC Clinical Sciences Centre, Faculty of Medicine, Imperial College, London W12 0NN, UK. 3 Institute of Physiology, Czech Academy of Sciences and Centre for Applied Genomics, 142 20 Prague 4, Czech Republic. 4 Institute of Biology and Medical Genetics, Charles University, 120 00 Prague 2, Czech Republic. 5 University of California, San Francisco, California 94143-0134, USA. 6 Department of Epidemiology and Public Health, Imperial College, London W2 1PG, UK. 7 These authors contributed equally to this work. Correspondence should be addressed to M.P. (pravenec@biomed.cas.cz) or T.J.A. (t.aitman@csc.mrc.ac.uk). NATURE GENETICS VOLUME 37 [ NUMBER 3 [ MARCH 2005 243 ARTICLES � 2005 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics and accessible resource to test the hypothesis that genetic variation in gene expression has a key role in the molecular evolution of complex physiological and pathophysiological phenotypes. RESULTS Linkage analysis of expression profiles in RI strains We analyzed genome-wide expression data of 15,923 transcripts collected from fat and kidney of 30 RI strains and the SHR and BN progenitor strains. To assess variability within RI strains and to generate a robust set of data, we analyzed gene expression in fat and kidney from four independent rats from each RI strain and from four or five rats from each progenitor strain. Including parental progenitor strains, we carried out 259 individual array hybridizations. We first compared gene expression profiles generated from fat and kidney tissue from the parental strains (SHR and BN). Of the 15,923 probe sets present on the array, 2,046 and 1,553 detected differential expression (P o 0.05) in fat and kidney, respectively, between SHR and BN parental strains. We next carried out genome-wide linkage analysis for the expression data generated in the RI strains. Because genetic regulation of gene expression may be detected for genes that are not differentially expressed between the parental strains, we carried out the linkage analysis for expression profiles from all 15,923 probe sets without filtering. We generated likelihood ratio statistic (LRS) values and established empirical genome-wide significance by permutation testing. At genome-wide significance of P � 0.05, we detected 3,520 and 4,530 linkages in fat and kidney, respectively. To account for multiple testing of 15,923 expression phenotypes, we evaluated the false discovery rate (FDR) at several levels of significance. Although the FDR is relatively high at P � 0.05, the expected number of true linkages is 2,644 in fat and 2,917 in kidney (Supplementary Table 1 online). At a more stringent level of P � 10 C03 , the FDR is B4% in both tissues, corresponding to 509 and 761 expected true positive linkages in fat and kidney, respectively. An independent test for linkage between marker and transcript using the Wilcoxon-Mann-Whitney test showed that 65?68% of these linkages at P � 10 C03 are common to the two analyses (Supplementary Table 2 online). To validate differential expression detected by microarray and to determine whether interstrain differences in gene expression could be due to DNA sequence variants affecting probe binding, we measured mRNA abundance in parental and RI strains by quantitative RT-PCR and sequenced the target region of a subset of cis-regulated transcripts. We found significant sequence differences that could affect probe binding for only 1 of the 15 transcripts that we examined: this transcript, in the major histocompatibility complex (MHC), showed significant sequence variation between SHR and BN that accounted for the apparent differential expression detected on the microarray (data not shown). We observed strong concordance between the microarray and RT-PCR data for differential expression between the parental strains (Supplementary Fig. 1 online). In addition, we con- firmed eQTL linkages for nine transcripts showing strong genome- wide linkage (P o 10 C03 ; Supplementary Table 3 online), including those showing small relative changes between genotypic classes. Characterization of cis-andtrans-acting eQTLs Linkages of individual expression phenotypes to multiple tightly linked markers give rise to artificially inflated numbers of eQTLs. We removed this redundancy using a custom algorithm (Supple- mentary Fig. 2 online), delineating a data set of 2,118 and 2,490 nonredundant eQTLs at genome-wide significance level of P o 0.05 in fat and kidney, respectively (Ta bl e 1). We examined which of these eQTLs were regulated in cis or in trans by defining a cis-acting eQTL as having a linkage peak within 10 Mbp of the physical location of the probe set (Supplementary Fig. 3 online). The proportion of eQTLs regulated in cis or in trans varied substantially in accordance with the genome-wide significance of the eQTL linkage. At a genome- wide threshold of P o 0.05, 60?65% of the eQTLs were regulated in trans in both tissues. At a higher significance level (P r 10 C04 ), however, 85?100% of eQTLs were regulated in cis (Ta bl e 1). These results may reflect the large gene effects of some cis-acting eQTLs and the probable oligo- or polygenic trans-acting influences on gene expression. A small proportion (B15%) of the eQTLs detected independently in kidney and fat were common to both tissues. These 311 eQTLs (Ta bl e 1) can be considered replicated linkages and probably reflect common regulatory mechanisms that are shared between the two tissues. A very high proportion of these shared linkages are cis-acting (70% at P o 0.05). This suggests that the preponderance of trans- acting eQTLs observed in the distinct fat and kidney data sets belong to tissue-specific networks for control of gene regulation. Some of the most significantly linked cis-acting eQTLs detected in the RI strains also showed marked differential expression in the parental strains (Fig. 1 and Ta bl e 2). Although a number of these eQTLs were shared between kidney and fat (e.g., Cd36 and Ilf3), others showed tissue-specific differences in regulation of gene expression (Sah,alsocalledSa, and the transcribed sequence 1388437_at). Many of these cis-acting eQTLs showed similar distributions of gene expression in the parental strains and in the RI strains when separated according to marker genotype at the linkage peak (Fig. 1). This is suggestive of a monogenic, or near monogenic, cis-acting regulation of Table 1 Number of eQTLs detected in fat and kidney data sets at different genome-wide thresholds of significance Fat Kidney Shared a Threshold Cis Trans Total Unknown b Total Cis Trans Total Unknown b Total Cis Trans Total Unknown b Total P o 0.05 622 1,211 1,833 285 2,118 800 1,251 2,051 439 2,490 222 44 266 45 311 P r 0.01 392 268 660 94 754 514 295 809 140 949 138 18 156 27 183 P r 10 ?3 189 46 235 36 271 270 56 326 58 384 75 9 84 13 97 P r 10 ?4 97 18 115 22 137 129 10 139 29 168 45 3 48 7 55 P r 10 ?5 49 4 53 8 61 62 4 66 6 72 23 0 23 2 25 P r 10 ?6 18 0 18 2 20 28 3 31 3 34 5 0 5 3 8 a For each threshold of significance, overlapping eQTLs between fat and kidney data sets were identified using the following criteria: (i) the eQTLs are linked to the same probe set and (ii) the eQTLs have the same set of genetic markers identifying the locus or they have both the same genetic marker and P-value threshold at the peak of linkage. b These eQTLs could not be defined as cis or trans because the physical location of either the probe set or the peak marker is unknown. 244 VOLUME 37 [ NUMBER 3 [ MARCH 2005 NATURE GENETICS ARTICLES � 2005 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics gene expression for these genes and for others with linkage at similar levels of significance (Ta bl e 2). Nevertheless, 675 of all linked transcripts showed linkage to two or more loci, emphasizing the generally complex nature of regulation of gene expression. SNP detection and SNP frequency in eQTL genes To identify DNA sequence variants that may represent candidate quantitative trait nucleotides underlying eQTLs, we generated sequence data for seven of the most statistically significant cis-acting eQTL genes (Pik3c3, Myh9, Kif1c, Aox1, Ascl3, Dgat2 and Gnpat). Although variations in all parts of the gene can affect transcript levels, we focused on variations in the promoter or cDNA that might directly influence transcriptional activity or mRNA stability. We identified sequence variants in the exons or upstream regulatory regions in six of the seven genes. As a first step towards assessing the functional importance of these sequence variants, we determined the 10 12 13 5 6 7 a Fat KidneyTranscript: 1367689_a_at; Marker: Cd36 9 10 12 13 5 6 7 8 9 b Fat KidneyTranscript: 1386901_at; Marker: Cd36 6 7 8 5 6 7 i Fat KidneyTranscript: 1369655_at; Marker: D18Rat103 5 7 9 11 4 6 8 j Fat KidneyTranscript: 1371776_at; Marker: D2Rat94 4 5 6 7 4 5 6 e Fat KidneyTranscript: 1377212_at; Marker: D12Rat56 6 7 8 9 5 6 7 f Fat KidneyTranscript: 1377244_at; Marker: D12Cebrp97s4 4 5 5 6 7 8 9 c Fat KidneyTranscript: 1383303_at; Marker: Sah 5 6 7 8 9 5 6 7 8 g Fat KidneyTranscript: 1387366_at; Marker: D8Rat68 6 7 8 9 5 6 d Fat KidneyTranscript: 1387791_at; Marker: D10Utr2 8 9 10 6 7 8 k Fat KidneyTranscript: 1388437_at; Marker: D7Cebr77s1 6 7 8 9 8 9 10 h Fat KidneyTranscript: 1388617_at; Marker: D17Rat144 5 6 7 8 5 6 7 l Fat KidneyTranscript: 1390364_at; Marker: D11Mit2 All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN All RI SHR BN RI SHR RI BN Expression Expression Expression Expression Expression Expression Expression Expression Expression Expression Expression Expression 111 11 Figure 1 Expression values of parental and RI strains for 12 transcripts in kidney and fat. The first column in each plot shows expression levels for all 30 RI strains. The second and third columns show expression levels for the replicates from each parental strain (SHR and BN, respectively). The fourth and fifth columns show expression values for RI strains by SHR and BN marker genotype, respectively. RMA-normalized expression values are shown on the y axis. Detailed properties for all genes depicted are given in Table 2. These genes, all with cis-acting eQTLs at genome-wide significance level of P r 5 C2 10 C04 in either kidney or fat, were selected according to the following criteria: (i) four genes previously reported to show cis-acting differential expression or association with an SHR phenotype (a?d); (ii) four genes with highly significant cis-acting eQTLs in both tissues (e?h); (iii) two genes with highly significant eQTLs and with biological relevance in kidney tissue (i?j); (iv) two genes with highly significant eQTLs in fat tissue (k?l). Gene names and linkage statistics are given in Table 2. NATURE GENETICS VOLUME 37 [ NUMBER 3 [ MARCH 2005 245 ARTICLES � 2005 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics allele status for all sequence variants in three additional inbred strains by resequencing and compared the results to the reference rat genome sequence. Only one of these genes, Pik3c3, had a distribution of allelic variants (one promoter SNP and one silent substitution in exon 19) that was common to all of the studied hypertensive strains (SHR/Ola, SHR/Mdc and SHRSP/Mdc) and distinct from that of the normoten- sive strains (BN.Lx/Cub, BN/SsNHsd/Mcwi and WKY/Mdc). Further sequencing of Pik3c3 from genomic DNA detected 19 SNPs distrib- uted throughout the gene on two distinct haplotypes (Supplementary Fig. 4 online). Pik3c3 showed allelic segregation with hypertension across strains, is one of the most statistically significant (P o 10 C06 ) cis-acting eQTL genes in our data set and resides within the blood pressure QTL BP46 (pQTL ID 107) and in a chromosome 18 SHR congenic strain (SHR.BN-D18Rat32/D18Rat12;pQTLID105;Supplementary Ta bl e 4 online). Pik3c3 was upregulated by a factor of 1.4, as shown by quantitative RT-PCR, in kidneys from the SHR strain compared with those from both the BN and the chromosome 18 congenic strains (both Po0.05), confirming cis-regulated control of Pik3c3 expression in the SHR strain. Kidneys transplanted from congenic SHR.BN- D18Rat32/D18Rat12 rats into SHR rats led to a highly significant drop in blood pressure (by 12 mmHg) compared with kidneys transplanted from SHR rats into other SHR rats (data not shown). These data indicate that the congenic kidney is sufficient to mediate blood pressure changes and support the idea that Pik3c3 is a good candidate for involvement with hypertension in the SHR strain. The finding of SNPs in six of the seven cis-acting eQTL genes that we studied led us to investigate the SNP frequency at the level of the genome compared with the frequency in eQTL genes. We first inspected the number of SNPs identified in a previously published data set of rat cDNAs 20 . For the 8,986 genes on the RAE230A array with Ensembl IDs, SNPs were detected between the stroke-prone SHR (SHRSP) and the BN reference sequences in 2,092 genes (23.3%), which we considered the observed SNP rate across the genome in this SNP data set. We then determined the number of genes in this data set that contained SNPs in which we detected cis-andtrans-acting eQTLs and found significant enrichment for SNPs in the cis-regulated eQTL genes compared with either the trans-regulated eQTL genes or the observed rate across the genome (Ta bl e 3). Relationship of eQTLs to pQTLs We addressed how the eQTL data can be used to identify candidate genes underlying SHR phenotypes, many of which have been analyzed by QTL mapping in experimental crosses over the past 15 years. Cis- acting eQTLs are good candidates for these pQTLs because they show strain-specific differences in gene expression that are under the control of DNA sequence variants in or close to the gene itself. To illustrate the overlap between individual cis-acting eQTLs and previously mapped SHR pQTLs, we plotted the locations of the most stringently mapped (P r 10 C04 ) cis-acting eQTLs against the chromosomal locations of known SHR pQTLs (Fig. 2). This generated a data set of genes that have cis-acting sequence variants that mediate interstrain differences in gene transcription or mRNA stability (Fig. 2; details of the pQTLs and lists of the cis-acting eQTLs with P r 10 C04 (FDR r 1% in both tissues) are given in Supplementary Ta bl es 4 and 5 online). The sequence variants may also underlie Table 2 Linkage statistics and biological properties of cis-regulated genes Transcript Gene symbol Gene name Transcript physical position (Mb) Genetic marker at peak of linkage Marker physical position (Mb) P value* kidney P value* fat 1367689_a_at Cd36 Cd36 antigen 13.51 Cd36 13.51 3.0 C2 10 C06 1.0 C2 10 C06 1386901_at Cd36 Cd36 antigen 13.51 Cd36 13.51 o10 C06 1.0 C2 10 C06 1383303_at Sah SAH (Sah) gene, complete cds 178.22 Sah 178.17 1.0 C2 10 C05 40.05 1387791_at Ace Angiotensin 1 converting enzyme 1 95.39 D10Utr2 95.58 2.1 C2 10 C04 40.05 1377212_at ? Transcribed sequences 3.57 D12Rat56 3.02 1.0 C2 10 C06 1.0 C2 10 C06 1377244_at ? Similar to zinc finger protein 95 (LOC304275), mRNA 9.73 D12Cebrp97s4 5.55 1.5 C2 10 C05 1.0 C2 10 C06 1387366_at Ilf3 Interleukin enhancer binding factor 3 20.5 D8Rat68 19.46 o10 C06 o10 C06 1388617_at C0 Similar to RIKEN cDNA 2010012D11 (LOC361239), mRNA 37.15 D17Rat144 37.97 3.0 C2 10 C06 3.0 C2 10 C06 1369655_at Pik3c3 Phosphatidylinositol 3-kinase 22.56 D18Rat103 20.16 o10 C06 0.025 1371776_at Pik3r1 Phosphatidylinositol 3-kinase, regulatory subunit, polypeptide 1 32.52 D2Rat94 31.1 1.4 C2 10 C04 8.6 C2 10 C04 1388437_at C0 Transcribed sequences 115.17 D7Cebr77s1 115.04 40.05 o10 C06 1390364_at C0 Transcribed sequences 32.78 D11Mit2 30.89 40.05 o10 C06 *P value from eQTL Reaper linkage analysis in RI strains. Table 3 Percentage of transcripts with sequence variants eQTL significance Total number of eQTLs Total number of eQTLs containing SNPs Total number of cis-acting eQTLs Number of cis-acting eQTLs containing SNPs Total number of trans-acting eQTLs Number of trans-acting eQTLs containing SNPs P o 0.05 1,510 429 (28.4%) a 548 204 (37.2%) a,b 1,096 272 (24.8%) b P o 0.001 238 87 (36.6%) a 199 80 (40.2%) a,b 50 11 (22.0%) b a Significant difference (P o 0.001 by w 2 test) from overall proportion of genes with SNPs across the genome (23.3%). b Significant difference (P o 0.001 by w 2 test) between cis- and trans-acting eQTL genes. The observed percentage of transcripts with sequence variants is significantly different (P o 0.001) from the expected percentage across the genome (23.3%) due to increased sequence variability in cis-regulated transcripts. The sum of cis-andtrans-regulated transcripts is more than the total number because some transcripts show both cis and trans regulation. 246 VOLUME 37 [ NUMBER 3 [ MARCH 2005 NATURE GENETICS ARTICLES � 2005 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics (patho)physiological phenotypes mapped as pQTLs in crosses between SHR, BN and other strains (Fig. 2). These cis-regulated eQTL genes merit testing in genetic and functional assays as positional candidates for colocalizing SHR pQTLs and potentially for related human QTLs. Other cis-acting eQTLs may be considered ?orphan? eQTLs, candidate genes for pQTLs that have yet to be mapped. Trans-acting eQTLs represent loci that influence expression of genes or transcripts remote from the eQTL itself. Coincidental mapping of trans-acting eQTLs for multiple transcripts to the same chromosomal location, as observed on chromosomes 3 and 17 (Fig. 3), may represent a shared regulatory transcriptional control mechanism by a single gene at the eQTL. The locations of trans- acting eQTLs in relation to known SHR pQTLs (Fig. 3), together with the locations of cis-acting eQTLs (Fig. 2), may point to genes and regulatory pathways underlying individual SHR pQTLs. Comparative analysis of blood pressure QTLs To explore the applicability of the detected fat and kidney cis-acting eQTLs to human hypertension, we formed a data set of 255 cis-acting eQTL genes with FDR r 5% that were contained within rat pQTLs for blood pressure and left ventricle and cardiac mass previously mapped in the SHR strain (Supplementary Table 4 online). We used the Ensembl Ensmart database to map Affymetrix probe set identifiers of each cis-acting eQTL to rat Ensembl genes and then to identify the putative human orthologs. We determined the physical location of each human ortholog on the human genome sequence and compared 0 12 4 50 100 150 200 250 0 50 100 150 0 50 100 150 5 50 0 100 9 50 0 100 10 50 0 11 50 0 12 0 100 6 50 0 100 7 50 0 100 8 50 0 100 150 0 50 100 150 200 250 1 2 3 4 5 6 7 8 9 10 11 12 13 14 18 29 30 31 32 33 34 38 39 40 41 43 44 45 46 47 48 49 50 51 52 53 35 36 37 19 20 21 22 23 24 25 26 27 15 16 17 54 55 56 57 58 59 60 61 76 13 50 0 100 14 50 0 100 92 17 50 0 100 18 50 0 19 50 0 20 50 0 104 105 106 107 108 109 110 112 113 114 115 116 117 101 102 103 93 94 95 77 78 79 80 81 82 83 96 15 50 0 16 50 0 100 97 98 99 85 86 87 88 89 90 91 62 63 64 65 65 67 68 70 71 72 73 75 Figure 2 Locations of cis-acting eQTLs and previously mapped SHR pQTLs. Chromosomes are shown in blue. The arrowheads on the left of each chromosome (yellow, fat; red, kidney; green, shared) represent the location of the probe set for each cis-acting eQTL with P r 10 C04 . Previously identified pQTLs in the SHR strains and in the RI strains are shown on the right of each chromosome. Gray boxes represent pQTLs for which both flanking markers are mapped. White boxes represent pQTLs for which incomplete flanking marker information was available, in which case the flanking marker(s) are estimatedto be 10 Mbp from the linkage peak. The numbers on the pQTL bars correspond to pQTL information reported in Supplementary Table 4 online. Some pQTLs in Supplementary Table 4 online do not appear in this diagram owing to a lack of physical mapping information for the genetic markers that define the pQTL. Gene names, relative changes in parental strains and public database references for probe sets are given in Supplementary Table 5 online. Scale in Mbp. NATURE GENETICS VOLUME 37 [ NUMBER 3 [ MARCH 2005 247 ARTICLES � 2005 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics these locations with the locations of previously mapped human blood pressure QTLs (Supplementary Table 6 online). This yielded a set of 73 human orthologs of cis-regulated rat eQTL genes that are contained within human blood pressure QTLs and are candidate genes for human hypertension (Ta bl e 4). DISCUSSION Identification of genes that underlie complex (polygenic) traits remains a challenge despite the availability of the genome sequences of humans and other species 21,22 . We previously combined linkage analysis with microarray-based expression profiling to identify a defective gene, Cd36, that underlies complex disease phenotypes in the SHR strain 12,23 . Other investigators have successfully used this approach in the study of both monogenic and complex traits 24?26 .But the relative paucity of successful studies using this strategy, and the nature of the Cd36 genomic deletion 27 that led to detection of dysregulated Cd36 expression on the microarray, raised questions about the generality of this approach. Here we applied global gene expression profiling and linkage analysis to study the regulation of gene expression in fat and kidney 1234 5678 910112 13 14 15 16 17 18 19 20 0 50 100 2 1 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 54 55 56 57 58 59 60 61 76 77 78 92 93 94 95 100 101 102 103 104 105 106 107 108 109 110 112 113 114 115 116 117 96 97 98 99 79 80 81 82 83 85 86 87 88 89 90 91 62 63 64 65 66 67 68 70 71 72 73 75 18 29 30 31 32 33 34 35 36 37 38 39 40 41 43 44 45 46 47 48 49 50 51 52 53 19 20 21 22 23 24 25 26 27 150 200 250 0 50 100 150 0 50 100 150 0 50 100 0 50 100 0 50 100 0 50 0 50 0 50 0 50 0 50 100 0 50 100 0 50 0 50 100 0 50 0 0 50 100 0 50 100 0 50 100 150 0 50 100 150 200 250 Figure 3 Location of trans-acting eQTLs and previously mapped SHR pQTLs. pQTLs are shown as in Figure 2. The arrowheads on the left of each chromosome (colors as in Fig. 2) represent the locations of the markers at the linkage peak for each trans-regulated eQTL with P r 10 C02 . This threshold of significance was chosen because trans-acting eQTLs are expected to have smaller effects than cis-acting eQTLs 4 . Gene names, relative changes in parental strains and public database references for probe sets are given in Supplementary Table 8 online. Scale in Mbp. 248 VOLUME 37 [ NUMBER 3 [ MARCH 2005 NATURE GENETICS ARTICLES � 2005 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics Table 4 Comparative analysis of rat and human blood pressure and blood pressure?related QTLs Rat tissue Rat eQTL probe set Rat gene Rat chr. Overlapping rat BP QTLs (ID) Putative human ortholog Human chr. Overlapping human BP QTLs a Shared 1387812_at Pace4 1 BpQTLcluster1 (1) PCSK6 15 BP8_H, BP_H_K1 b 1376780_at C0 1 BpQTLcluster1 (1), Lvm2 (3) C6orf119 15 BP32_H, BP_H_K1 b 1377614_at C0 1 Bp26 (8), Bp (Huang et al) (9) NP_775889 16 BP33_H, BP27_H 1389264_at ? 7 Lvm7 (Huang et al) (67) NP_620152 22 BP55_H, BP38_H 1371725_at C0 7 Lvm7 (Huang et al) (67) MYH9 22 BP55_H, BP38_H 1373838_at Fut4 8 Bp35 (71) FUT4 11 BP31_H 1390185_at Dcps 8 Bp35 (71), Bp22 (72) DCPS 11 BP22_H 1371442_at Hyou1 8 Bp35 (71), Bp62 (73) HYOU1 11 BP31_H, BP22_H 1380286_x_at C0 8 Bp35 (71) Q9C0D2 11 BP31_H 1398460_at C0 8 Bp35 (71) Q9C0D2 11 BP31_H 1377245_a_at C0 8 Bp35 (71) Q9C0D2 11 BP31_H 1375988_at C0 8 Bp35 (71), Bp22 (72) NP_659451 11 BP22_H 1382105_at Gnb5 8 Bp35 (71), Bp62 (73) GNB5 15 BP32_H 1370176_at Als2cr3 9 Bp53 (76) ALS2CR3 2 BP23_H, BP46_H 1374196_at Lancl1 9 Bp53 (76), Bp108 (78) LANCL1 2BP46_H 1387376_at Aox1 9 Bp53 (76) AOX1 2 BP23_H, BP46_H 1371803_at Gm2a 10 BpQTLcluster9 (79) GM2A 5BP21_H 1398309_at Pigl 10 BpQTLcluster9 (79) PIGL 17 BP16_H, BP15_H, BP34_H 1370930_at Kif1c 10 BpQTLcluster9 (79), Bp45 (82) KIF1C 17 BP15_H, BP34_H 1368285_at Shbg 10 BpQTLcluster9 (79) SHBG 17 BP15_H, BP34_H 1373034_at C0 11 BpQTLcluster10 (85) WRB 21 BP12_H, BP_H_K_2 b Kidney 1375983_at C0 1 Cm7 (12), Bp42 (13), Map1 (14), Km1 (15) MAMDC2 9 BP_H_K_3 b 1371615_at Dgat2 1 BpQTLcluster1 (1), Lvm2 (3), Bp44 (6), Bp26 (8) DGAT2 11 BP31_H 1383303_at Sah 1 Bp26 (8), Bp (Huang et al) (9) SAH 16 BP27_H 1388567_at C0 1 Bp26 (8), Bp (Huang et al) (9) THUMPD1 16 BP27_H 1389300_at C0 1 Bp26 (8), Bp (Huang et al) (9) C0 16 BP27_H 1386985_at Gstm1 2 Bp105 (21), BpQTLcluster3 (23), Bp101 (24), Bp16 (26), Bp63 (27) GSTM1 1BP45_H 1372782_a_at Ampd2 2 Bp105 (21), BpQTLcluster3 (23), Bp101 (24), Bp16 (26), Bp63 (27) AMPD2 1BP45_H 1387067_at C0 2 Bp105 (21), BpQTLcluster3 (23), Bp101 (24), Bp19 (25) NP_958800 1 BP5_H 1389142_at C0 3Cm17(35) SQRDL 15 BP32_H 1388178_at Ncoa3 3 Bp20 (37) NCOA3 20 BP29_H, BP_H_K_4 b 1373204_at C0 4 Bp79 (44), Bp21 (46), BpQTLcluster5 (47) NP_060957 7 BP48_H 1389475_at Smo 4 Bp79 (44), Bp21 (46) SMO 7BP48_H 1372980_at C0 4 Bp79 (44), Bp21 (46) NP_848657 7 BP48_H 1370401_at Ly6h 7 Lvm7 (Huang et al) (67) LY6H 8BP1_H 1367917_at Cyp2d26 7 Lvm7 (Huang et al) (67) CYP2D7P1 22 BP55_H, BP38_H 1370249_at Bzrp 7 Lvm7 (Huang et al) (67) BZRP 22 BP55_H, BP38_H 1376501_at C0 7 Lvm7 (Huang et al) (67) ARHGAP8 22 BP55_H 1372897_at Plod2 8 Bp35 (71) PLOD2 3BP24_H 1374933_at Mcam 8 Bp35 (71), Bp62 (73) MCAM 11 BP31_H, BP22_H 1369665_a_at Il18 8 Bp35 (71), Bp62 (73) IL18 11 BP31_H 1376921_at C0 8 Bp35 (71), Bp22 (72) NP_060017 11 BP22_H 1377457_a_at C0 8 Bp35 (71), Bp22 (72) SORL1 11 BP31_H, BP22_H 1390710_x_at C0 8 Bp35 (71), Bp22 (72) SORL1 11 BP31_H, BP22_H 1376110_at C0 8 Bp35 (71), Bp62 (73) RPP25 15 BP32_H 1373887_at Sf3b1 9 Bp53 (76) SF3B1 2 BP23_H, BP46_H 1376285_at C0 9 Bp53 (76), Bp113 (77) GULP1 2 BP23_H, BP37_H, BP46_H 1390063_at C0 10 BpQTLcluster9 (79) MFAP3 5BP21_H 1368574_at Adra1b 10 BpQTLcluster9 (79) ADRA1B 5 BP_H_K_5 b 1373665_at C0 10 BpQTLcluster9 (79) NP_079375 17 BP15_H, BP34_H 1371645_at C0 10 BpQTLcluster9 (79), Bp45 (82) SDF2 17 BP16_H, BP34_H 1377353_a_at C0 10 BpQTLcluster9 (79), Bp45 (82) TNFSF12 17 BP15_H, BP34_H 1367989_at Slc2a4 10 BpQTLcluster9 (79), Bp45 (82) SLC2A4 17 BP15_H, BP34_H 1372064_at C0 10 BpQTLcluster9 (79), Bp45 (82) CXCL16 17 BP15_H, BP34_H 1372397_at C0 13 Bp80 (95) RPL34P1 1BP5_H 1376259_at PRKCQ 17 Lvm6 (103) PRKCQ 10 BP52_H 1388656_at C0 18 Bp46 (107) UBE2D2 5BP21_H Table 4 continued on following page NATURE GENETICS VOLUME 37 [ NUMBER 3 [ MARCH 2005 249 ARTICLES � 2005 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics tissue in the BXH/HXB panel of rat RI strains. We used this panel to map the genetic determinants of gene expression in the SHR strain for 15,923 genes in two of the key tissues in the pathophysiology of the metabolic syndrome. After assessing genome-wide significance, removing redundancy and accounting for multiple testing using FDR, we found more than 1,000 eQTLs in each tissue, of which several hundred were common to both kidney and fat. These eQTLs represent a large source of candidate genes for the scores of pQTLs that have been mapped in the SHR strain. As previously observed 2?5,28 , a powerful feature of this experimental design is the ability to discriminate between cis-andtrans-acting influences on gene expression. In keeping with previous results 4 ,we found that, at relatively low levels of significance (P o 0.05), a high proportion of detected eQTLs, B65%, result from trans-acting regulators of gene expression. At higher levels of significance (P r 10 C04 ), 85?100% of eQTLs arise from cis-acting regulation and have larger, mainly monogenic effects on gene expression. This suggests that, owing to sequence variation in the gene itself, cis-acting regulation is more easily detected than variation in genes with secondary effects on transcription of other genes and underscores the more complex nature of trans regulation. Investigation of sequence variation in eQTL genes showed that the SNP frequency was much higher in genes with cis-acting eQTLs than in genes with trans-acting eQTLs or than the rate observed across the genome (Ta bl e 3). This observation has a number of possible explanations. First, because the SHR and SHRSP strains are closely related, and both are genetically distant from the BN strain, the increase in polymorphisms detected in cis-regulated genes could reflect identification of causative nucleotide variants that underlie cis-acting control of gene expression. Second, cis-regulated genes may lie in SNP-rich chromosomal regions, and SNPs in these genes could therefore be considered markers of chromosomal regions of phyloge- netic diversity between SHR-related strains and the BN strain. If this is the case, a haplotype map of the rat could point to chromosomal regions containing a high density of cis-regulated genes. Cis-acting eQTLs are of particular interest as positional candidate genes for pQTLs (Fig. 2 and Supplementary Tables 4 and 5 online). Among many other cis-acting eQTLs, our data showed cis-acting con- trol of gene expression for Cd36 and Sah,inwhichcis-acting control of gene expression or significant intragene sequence variation has been documented 23,29 . Cd36 is represented twice on the microarray by non- overlapping probe sets. Our data identified strong cis-acting eQTLs for both probe sets (P E 10 C06 ) in fat and in kidney (Ta b l e 2 and Supplementary Table 5 online). One of the probe sets (1386901_at) is derived from sequence in the 3� untranslated region that is deleted from the SHR genome 23,27 , and the second (1367689_a_ at) is derived from sequence in the Cd36 coding region and uses oligonucleotides that do not differ in sequence between the SHR and BN strains 27 .This confirms that we can detect the chromosomal deletion found pre- viously with cDNA microarrays 23 and also shows, using probe set 1367689_a_at, that the cis-acting eQTL for Cd36 cannot simply be attributed to strain differences in probe affinity for Cd36 mRNA. For Sah, our data showed highly significant differential expression in kidney (P � 10 C09 ) between SHR and BN parental strains and strong cis linkage in the RI strains (P � 10 C05 ; Fig. 1c and Ta ble 2 and Supplementary Table 5 online). The genotype-dependent bimodal distribution of Sah expression in the RI strains (Fig. 1c)indicates essentially monogenic cis regulation, as previously reported 29 . Although Sah is now known not to be a primary determinant of hypertension in the SHR strain 30,31 , the detection of this cis-acting eQTL in kidney but not in fat demonstrates the ability of this system to identify tissue-specific differences in gene expression, as shown also for many other genes in our data set. In this study, we investigated seven of the most statistically sig- nificant cis-regulated eQTL genes from kidney as positional candidates for involvement in hypertension. Nucleotide variations in one of these genes, Pik3c3, formed a haplotype that associated with hypertension in a small range of hypertensive and normotensive strains (Supplemen- tary Fig. 4 online). Pik3c3 resides in both an SHR blood pressure QTL and an SHR congenic segment that carries a hypertension gene on this region of chromosome 18 (refs. 32,33). Additional physiological and biochemical data (Supplementary Note online) support the idea that Pik3c3 is a candidate for involvement in hypertension in the SHR strain. Existing functional data should similarly encourage investiga- tion of other cis-acting eQTL genes (Ta bl e 2 and Supplementary Ta b le 5 online), such as Pik3r1, the regulatory subunit of PI3K, and Ace, which has been extensively investigated as a candidate for involvement in cardiovascular phenotypes 34,35 . Table 4 Continued Rat tissue Rat eQTL probe set Rat gene Rat chr. Overlapping rat BP QTLs (ID) Putative human ortholog Human chr. Overlapping human BP QTLs a Fat 1375516_at C0 1 BpQTLcluster1 (1), Lvm2 (3), Bp44 (6), Bp26 (8) NDUFC2 11 BP31_H 1389300_at C0 1 Bp26 (8), Bp (Huang et al) (9) C0 16 BP27_H 1369866_at LOC56825 2 Bp105 (21), BpQTLcluster3 (23), Bp101 (24), Bp16 (26), Bp63 (27) C0 1BP45_H 1373243_at C0 2 Bp105 (21), BpQTLcluster3 (23), Bp101 (24), Bp19 (25) PMVK 1BP5_H 1398960_at C0 5 Cm (Hamet et al) (89), Bp142 (91) COPS5 8BP49_H 1370377_at Cyp2d26 7 Lvm7 (Huang et al) (67) CYP2D7P1 22 BP55_H, BP38_H 1375940_a_at C0 7 Lvm7 (Huang et al) (67) NP_078957 22 BP55_H, BP38_H 1389229_at C0 8 Bp35 (71) ACPL2 3BP24_H 1390206_at C0 8 Bp35 (71), Bp22 (72) NP_060017 11 BP22_H 1390827_at P34 8 Bp35 (71), Bp62 (73) NP_078942 15 BP32_H 1372500_at Tmod2 8 Bp35 (71), Bp62 (73) TMOD2 15 BP32_H 1371951_at Fhl2 9 Bp53 (76), Bp113 (77) FHL2 2BP46_H 1372646_at C0 9 Bp53 (76), Bp113 (77) NP_115787 2 BP46_H 1373417_at C0 10 BpQTLcluster9 (79), Bp45 (82) NP_689979 17 BP15_H, BP34_H 1368304_at Fmo3 13 Bp80 (95) FMO3 1BP5_H 1371732_at C0 13 Bp80 (95) DPT 1BP5_H a Full details of rat and human genes and pQTLs are given in Supplementary Table 6 online. b BP_H_Kx are newly designated QTLs 50 . BP, blood pressure. 250 VOLUME 37 [ NUMBER 3 [ MARCH 2005 NATURE GENETICS ARTICLES � 2005 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics Eleven of the detected cis-acting eQTLs are located in or close to the MHC on chromosome 20 (Fig. 2). Because we observed cDNA sequence variation between the SHR and BN strains in the single MHC gene that we tested, which accounted for apparent differential expression between strains, several of the apparent cis-acting eQTLs in the MHC probably have a similar basis. Because we found no significant sequence variation in an additional 15 non-MHC genes, however, we do not believe that this accounts for more than a very small proportion of all cis-acting eQTLs. The detection in this study of known cis-acting regulation of gene expression raises the question of whether this approach is of value in identifying the genes that underlie pQTLs. The SHR defect in Cd36 has been conclusively shown in complementation studies to result in pathophysiological phenotypes 36 , whereas upregulation in SHR Sah gene expression, detected here and elsewhere, is not now believed to contribute to the hypertensive phenotype 30,31 . Several other examples now exist in which a causal relationship has definitively been shown between cis-acting eQTLs and functional or physiological pheno- types 2,4,26 . The available data therefore suggest that the combined expression and linkage approach may be useful for pQTL gene identification, particularly when applied on a genome scale. The trans-acting eQTLs that we mapped represent transcripts whose abundance is regulated by loci remote from the genomic locus of each of these genes. Yvert et al. analyzed in detail a set of trans-acting eQTLs and defined clusters of significantly coregulated genes with functional effects on yeast biology 3 . We found several large groups of genes, up to 43 in a single group (Fig. 3), with colocalizing trans-acting eQTLs, suggestive of coregulation by a common gene in the eQTL. Some of these groups of trans-acting eQTLs overlap with pQTLs, suggesting that they may play a part in mediating the development of SHR phenotypes. Functional and in silico analyses of the genes in these groups may advance understanding of the regulatory pathways that underlie these phenotypes. To show how mining of our data set may be applied to the study of human disease phenotypes, we identified a set of 73 robustly mapped cis-regulated rat eQTL genes with FDR o 5% that lie in SHR blood pressure?related pQTLs and whose human orthologs reside in QTLs for human hypertension (Ta bl e 4 and Supplementary Table 6 online). These genes are good candidates for underlying human hypertension, although for many there is little functional information. Some of the genes for which functional information does exist show circumstantial association with hypertension. For example, glutathione S-transferase mu-type 1 (Gstm1) has a strong cis linkage (P o 10 C05 )inkidney tissue, is differentially expressed by a factor of B2.4 fold between the parental strains and lies in a cluster of blood pressure QTLs on rat chromosome 2. Gstm1 has also been implicated previously in hyper- tension in genetic studies in the SHRSP strain 37 and in humans 38 , suggesting that it may have a role in hypertensive rat strains other than SHR and in humans. Like Gstm1, other genes in this data set play a part in cellular resistance to oxidative stress or exhibit different biological characteristics that make them plausible positional candi- dates for human hypertension (Supplementary Note online). Previous studies taking the combined expression and linkage approach used simple eukaryotes, transformed cell lines from healthy individuals or segregating populations that are no longer available for additional phenotyping 2?5 . Here we applied the same strategy to study, in two different mammalian tissues, regulation of gene expres- sion in the BXH/HXB panel of rat RI strains. RI strains have a number of advantages for this type of study, including the ability to make measurements in multiple genetically identical animals from the same strain to increase trait heritability and the ability to accumulate new phenotypes over time. In addition, the continued breeding of these strains, which are publicly available on a collaborative basis (from M.P. and V.K.), the public availability of the expression data set in this study and of the linkage maps generated in these strains 39 give value to these results outside the context of the immediate findings presented here. Alternative study designs in consomic or congenic strains could provide similar long-term resources and should complement the approach presented here 16?18 . In our study, we mapped genetic determinants of gene expression in a single strain combination, SHR C2 BN. But the coincidental mapping of physiological phenotypes in several crosses 23,40 , taken together with the data for Pik3c3 showing association between allelic variants and hypertension across strains, suggests that the eQTLs identified in this study will probably be relevant to SHR phenotypes mapped under a variety of environmental conditions or in strain combinations other than SHR C2 BN. The value of our results is enhanced by the extremely detailed studies (over a 30-year period) of hypertension, other components of the metabolic syndrome and further unrelated SHR pheno- types 9,12,15,23,36,41?43 . Many of these phenotypes have been subjected to genetic analysis in the BXH/HXB RI panel or are amenable to such analysis. Any phenotype that segregates in the SHR C2 BN strain combination can potentially be analyzed by this approach, either using the RI strain expression data set from this study or by generating new expression data sets in other tissues or studying additional transcripts in this RI panel. METHODS Strains and tissue. We produced a set of RI strains by inbreeding between members of the F 2 generation resulting from the cross of the two highly inbred strains: BN (BN.Lx/Cub) and SHR (SHR/Ola) 14 . We used 30 RI strains (BXH and HXB) at F 60 . We housed rats in an air-conditioned animal facility and allowed them free access to standard laboratory chow and water. All experi- ments were done in agreement with the Animal Protection Law of the Czech Republic (311/1997) and were approved by the Ethics Committee of the Institute of Physiology, Czech Academy of Sciences, Prague. We killed rats at 6 weeks of age. We collected tissues from four unfasted males of each RI strain and from four or five rats from each parental strain between 9:00am and 10:00am, froze them in liquid nitrogen and stored them at C080 1C. Preparation of labeled cRNA and hybridization. We extracted total RNA from retroperitoneal fat pads 12 and from whole kidney from four or five rats of each strain using Trizol reagent (Invitrogen) and purified it using an RNeasy Mini kit (Qiagen) in accordance with the manufacturer?s protocol. We synthesized double-stranded cDNA from total RNA without pooling samples, synthesized biotinylated cRNA from cDNA using the MEGAscript T7 kit (Ambion) and nucleotide analogs (Perkin Elmer) in the kidney samples and using Bioarray High Yield RNA Transcript Labelling Kit (Enzo Diagnostics) in fat. We hybridized 15 mg of the fragmented cRNA samples to rat expression Affymetrix RAE 230A GeneChips arrays in accordance with the Affymetrix protocol. Analysis of expression data. We computed gene expression summary values for Affymetrix GeneChip data using the robust multichip average (RMA) algorithm 44 , which uses background adjustment, quantile normalization and summarization. Statistical comparison of expression data in the parental strains was by student?s t-test (two-tailed) without correction for multiple testing. To determine relative changes, we back-transformed raw RMA output values to the raw intensity scale (anti-log). Validation of microarray gene expression data. We used quantitative real-time PCR (TaqMan) to compare mRNA levels of 16 transcripts in kidneys that show a range of relative changes in expression between the SHR and BN progenitors. To validate the linkage data, we also measured mRNA levels by quantitative real-time PCR for nine cis-regulated transcripts across all RI strains. We reverse- transcribed DNA-free total RNA (2 g) with oligo(dT) primers (Gibco-BRL), Superscript II reverse transcriptase (Gibco-BRL) and dNTP (Boehringer NATURE GENETICS VOLUME 37 [ NUMBER 3 [ MARCH 2005 251 ARTICLES � 2005 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics Mannheim) in 40 ml of reaction buffer (Gibco-BRL). We designed primers and probes using Primer Express 1.0 (Applied Biosystems). TaqMan analysis used an Applied Biosystems 7700 system (Perkin Elmer). We normalized expression levels to 18S rRNA expression by using the 2 C0DDCT method. Map construction. We constructed a linkage map of 1,011 autosomal markers for all chromosomes using MAPMAKER/Exp. 3.0 (ref. 45) using two- and four-point linkage analysis and published marker genotypes 39,46,47 . We cor- rected the map based on known physical positions of markers and optimized it by 19 iterative steps using multipoint linkage analysis. Map details are given in Supplementary Table 7 online. We retrieved physical map positions of genetic markers from Ensembl or by alignment of available marker sequences to the rat genome using blastn. Markers that could not be mapped using blastn but that were located between physically anchored markers were placed on the physical map by interpolation. Eleven percent of all genetic markers could not be placed on the physical map and were used only for the linkage analysis. Mapping of eQTLs. We derived mean expression values from the four replicates for each RI strain and each tissue after application of the Nalimov outlier test at P o 10 C03 . We carried out genome-wide linkage analysis for each of the 15,923 expression traits and the 1,011 genetic markers using the eQTL Reaper program (K.F. Manly; University of Tennessee Health Science Center, Memphis, Tennessee), which generates an LRS as a measure of the significance of a possible eQTL. eQTL Reaper establishes genome-wide significance by permutation test, estimating an empirical genome-wide probability for obser- ving a given LRS score by chance 48 . For each probe set, permutations are carried out until an LRS greater than that for the real data is observed or until 1,000,000 permutations have been done, conditional on at least 1,000 permuta- tions being done. The permutation procedures for the eQTL Reaper program correct, for individual expression phenotypes, for multiple testing across genetic markers to give a genome-wide corrected P value. Calculation of FDR. The eQTL Reaper genome-wide P value accounts for multiple testing across genetic markers but not for multiple testing across the 15,923 expression measurements. We estimated the number of falsely discov- ered linkages at a given genome-wide significance level by calculating the q value 49 , defined as the minimum positive FDR for a fixed significance threshold (Supplementary Table 1 online). Definition of cis-andtrans-acting eQTLs. Cis-acting regulatory variants are polymorphisms located at or near a gene that influence mRNA levels of that gene. We defined cis-acting eQTLs as eQTLs that map within 10 Mbp of the physical location of the probe set on the genomic sequence (20 Mbp total window size; Supplementary Fig. 3 online). Other eQTLs were defined as acting in trans. We obtained physical locations of probe sets from Affymetrix or Ensembl. SNP detection in eQTL genes. To identify DNA sequence variants that could underlie eQTLs, we generated sequence data for the cDNA sequence and 2?5 kb upstream of exon 1 for seven of the most statistically significant cis-regulated eQTL genes from the kidney data set. We carried out direct sequencing from PCR-amplified cDNA or genomic DNA using primers designed by Primer 3.0 from Ensembl annotations. We purified PCR products by shrimp alkaline phosphatase (Promega) and exonuclease I (Promega) treatment and sequenced them directly on an ABI 3730 Sequencer (Applied Biosystems). To determine allele status in three additional inbred strains, we resequenced SNPs in promoters, exons and exon-intron boundaries on genomic DNA from WKY/ Mdc, SHR/Mdc and SHRSP/Mdc strains. URLs. ArrayExpress is available at http://www.ebi.ac.uk/arrayexpress/. Accession numbers. dbSNP: unique SNP identifiers, ss35032354?ss35032394. ArrayExpress: scanned microarray data, E-AFMX-7. Note: Supplementary information is available on the Nature Genetics website. ACKNOWLEDGMENTS We thank H. Banks, N. Cooley, F. Rahman, M. Gerhardt, H. Kistel, S. Blachut and R. Sarwar for technical assistance; K. Manly for providing the eQTL Reaper software; and Affymetrix for donation of microarrays. We acknowledge funding to T.J.A. from the MRC Clinical Sciences Centre, from the British Heart Foundation and from a Wellcome Trust Cardiovascular Functional Genomics initiative; to N.H. from the German Ministry for Science and Education (National Genome Research Network); to M.P. and to V.K. from the Grant Agency of the Czech Republic; to M.P. and T.J.A. from the Wellcome Trust Collaborative Research Initiative Grant; to T.W.K. from the US National Institutes of Health; and to T.W.K. and M.P. from a Fogarty International Research Collaboration Award. M. Pravenec is an International Research Scholar of the Howard Hughes Medical Institute. COMPETING INTERESTS STATEMENT The authors declare that they have no competing financial interests. Received 29 November 2004; accepted 26 January 2005 Published online at http://www.nature.com/naturegenetics/ 1. Cheung, V.G. et al. Natural variation in human gene expression assessed in lympho- blastoid cells. Nat. Genet. 33, 422?425 (2003). 2. Brem, R.B., Yvert, G., Clinton, R. & Kruglyak, L. Genetic dissection of transcriptional regulation in budding yeast. Science 296, 752?755 (2002). 3. Yvert, G. et al. Trans-acting regulatory variation in Saccharomyces cerevisiae and the role of transcription factors. Nat. Genet. 35, 57?64 (2003). 4. Schadt, E.E. et al. Genetics of gene expression surveyed in maize, mouse and man. Nature 422, 297?302 (2003). 5. Morley, M. et al. Genetic analysis of genome-wide variation in human gene expression. Nature 430, 743?747 (2004). 6. Jacob, H.J. & Kwitek, A.E. Rat genetics: attaching physiology and pharmacology to the genome. Nat. Rev. Genet. 3, 33?42 (2002). 7. Wallace, C.A. & Aitman, T.J. The rat comes clean. Nat. Genet. 36, 441?442 (2004). 8. Gibbs, R.A. et al. Genome sequence of the Brown Norway rat yields insights into mammalian evolution. Nature 428, 493?521 (2004). 9. Okamoto, K. Spontaneous Hypertension: Its Pathogenesis and Complications (Springer, Berlin, 1972). 10. Rao, R.H. Insulin resistance in spontaneously hypertensive rats: difference in inter- pretation based on insulin infusion rate or on plasma insulin in glucose clamp studies. Diabetes 42, 1364?1371 (1993). 11. Hulman, S., Falkner, B. & Freyvogel, N. Insulin resistance in the conscious sponta- neously hypertensive rat: euglycemic hyperinsulinemic clamp study. Metabolism 42, 14?18 (1993). 12. Aitman, T.J. et al. Quantitative trait loci for cellular defects in glucose and fatty acid metabolism in hypertensive rats. Nat. Genet. 16, 197?201 (1997). 13. Grundy, S.M., Brewer, H.B. Jr., Cleeman, J.I., Smith, S.C. Jr. & Lenfant, C. Definition of metabolic syndrome: Report of the National Heart, Lung, and Blood Institute/ American Heart Association conference on scientific issues related to definition. Circulation 109, 433?438 (2004). 14. Pravenec, M., Klir, P., Kren, V., Zicha, J. & Kunes, J. An analysis of spontaneous hypertension in spontaneously hypertensive rats by means of new recombinant inbred strains. J. Hypertens. 7, 217?221 (1989). 15. Pravenec, M. et al. Genetic analysis of metabolic syndrome in the spontaneously hypertensive rat. Physiol. Res. 53 (Suppl 1), 15?23 (2004). 16. Silver, L.M. Mouse Genetics: Concepts and Applications (Oxford University Press, New York, 1995). 17. Cowley, A.W. Jr., Roman, R.J. & Jacob, H.J. Application of chromosomal substitution techniques in gene?function discovery. J. Physiol. 554, 46?55 (2004). 18. Singer, J.B. et al. Genetic dissection of complex traits with chromosome substitution strains of mice. Science 304, 445?448 (2004). 19. Belknap, J.K., Mitchell, S.R., O?Toole, L.A., Helms, M.L. & Crabbe, J.C. Type I and type II error rates for quantitative trait loci (QTL) mapping studies using recombinant inbred mouse strains. Behav. Genet. 26, 149?160 (1996). 20. Zimdahl, H. et al. A SNP map of the rat genome generated from cDNA sequences. Science 303, 807 (2004). 21. Glazier, A.M., Nadeau, J.H. & Aitman, T.J. Finding genes that underlie complex traits. Science 298, 2345?2349 (2002). 22. Abiola, O. et al. The nature and identification of quantitative trait loci: a community?s view. Nat. Rev. Genet. 4, 911?916 (2003). 23. Aitman, T.J. et al. Identification of Cd36(Fat) as an insulin-resistance gene causing defective fatty acid and glucose metabolism in hypertensive rats. Nat. Genet. 21, 76?83 (1999). 24. Berge, K.E. et al. Accumulation of dietary cholesterol in sitosterolemia caused by mutations in adjacent ABC transporters. Science 290, 1771?1775 (2000). 25. Lawn, R.M. et al. The Tangier disease gene product ABC1 controls the cellular apolipoprotein?mediated lipid removal pathway. J. Clin. Invest. 104,R25?R31 (1999). 26. Karp, C.L. et al. Identification of complement factor 5 as a susceptibility locus for experimental allergic asthma. Nat. Immunol. 1, 221?226 (2000). 27. Glazier, A.M., Scott, J. & Aitman, T.J. Molecular basis of the Cd36 chromosomal deletion underlying SHR defects in insulin action and fatty acid metabolism. Mamm. Genome 13, 108?113 (2002). 28. Jansen, R.C. & Nap, J.-P. Genetical Genomics: the added value from segregation. Trends Genet. 17, 388?391 (2001). 252 VOLUME 37 [ NUMBER 3 [ MARCH 2005 NATURE GENETICS ARTICLES � 2005 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics 29. Frantz, S.A. et al. Successful isolation of a rat chromosome 1 blood pressure quantitative trait locus in reciprocal congenic strains. Hypertension 32, 639?646 (1998). 30. Hubner, N., Lee, Y.A., Lindpaintner, K., Ganten, D. & Kreutz, R. Congenic substitution mapping excludes Sa as a candidate gene locus for a blood pressure quantitative trait locus on rat chromosome 1. Hypertension 34, 643?648 (1999). 31. Frantz, S., Clemitson, J.R., Bihoreau, M.T., Gauguier, D. & Samani, N.J. Genetic dissection of region around the Sa gene on rat chromosome 1: evidence for multiple loci affecting blood pressure. Hypertension 38, 216?221 (2001). 32. Kovacs, P., Voigt, B. & Kloting, I. Novel quantitative trait loci for blood pressure and related traits on rat chromosomes 1, 10, and 18. Biochem. Biophys. Res. Commun. 235, 343?348 (1997). 33. Pravenec, M., Zidek, V., Kren, V., St Lezin, E. & Kurtz, T.W. Genetic isolation of a quantitative trait locus on chromosome 18 associated with blood pressure and salt sensitivity in the SHR. Am.J.Hypertens.14, 82A (2001). 34. Cambien, F. et al. Deletion polymorphism in the gene for angiotensin?converting enzyme is a potent risk factor for myocardial infarction. Nature 359, 641?644 (1992). 35. Lindpaintner, K. et al. A prospective evaluation of an angiotensin?converting?enzyme gene polymorphism and the risk of ischemic heart disease. N. Engl. J. Med. 332, 706?711 (1995). 36. Pravenec, M. et al. Transgenic rescue of defective Cd36 ameliorates insulin resistance in spontaneously hypertensive rats. Nat. Genet. 27, 156?158 (2001). 37. McBride, M.W. et al. Microarray analysis of rat chromosome 2 congenic strains. Hypertension 41, 847?853 (2003). 38. Rice, T. et al. Genome-wide linkage analysis of systolic and diastolic blood pressure: the Quebec Family Study. Circulation 102, 1956?1963 (2000). 39. Jirout, M. et al. A new framework marker-based linkage map and SDPs for the rat HXB/BXH strain set. Mamm. Genome 14, 537?546 (2003). 40. Kato, N. et al. Complete genome searches for quantitative trait loci controlling blood pressure and related traits in four segregating populations derived from Dahl hyper- tensive rats. Mamm. Genome 10, 259?265 (1999). 41. Pravenec, M. et al. Mapping of quantitative trait loci for blood pressure and cardiac mass in the rat by genome scanning of recombinant inbred strains. J. Clin. Invest. 96, 1973?1978 (1995). 42. Bottger, A. et al. Quantitative trait loci influencing cholesterol and phospholipid phenotypes map to chromosomes that contain genes regulating blood pressure in the spontaneously hypertensive rat. J. Clin. Invest. 98, 856?862 (1996). 43. Printz, M.P., Jirout, M., Jaworski, R., Alemayehu, A. & Kren, V. Genetic Models in Applied Physiology. HXB/BXH rat recombinant inbred strain platform: a newly enhanced tool for cardiovascular, behavioral, and developmental genetics and geno- mics. J. Appl. Physiol. 94, 2510?2522 (2003). 44. Irizarry, R.A. et al. Exploration, normalization, and summaries of high density oligo- nucleotide array probe level data. Biostatistics 4, 249?264 (2003). 45. Lander, E.S. et al. MAPMAKER?an interactive computer package for constructing primary genetic linkage maps of experimental and natural populations. Genomics 1, 174?181 (1987). 46. Pravenec, M. et al. A genetic linkage map of the rat derived from recombinant inbred strains. Mamm. Genome 7, 117?127 (1996). 47. Pravenec, M. et al. HXB/Ipcv and BXH/Cub recombinant inbred strains of the rat: strain distribution patterns of 632 alleles. Folia Biol. (Praha) 45, 203?215 (1999). 48. Churchill, G.A. & Doerge, R.W. Empirical threshold values for quantitative trait mapping. Genetics 138, 963?971 (1994). 49. Storey, J.D. A direct approach to false discovery rates. J. R. Stat Soc (B) 64, 479?498 (2002). 50. Krushkal, J. et al. Genome-wide linkage analyses of systolic blood pressure using highly discordant siblings. Circulation 99, 1407?1410 (1999). NATURE GENETICS VOLUME 37 [ NUMBER 3 [ MARCH 2005 253 ARTICLES � 2005 Nature Pub lishing Gr oup http://www .nature .com/natureg enetics "
Add Content to Group
|
Bookmark
|
Keywords
|
Flag Inappropriate
share
Close
Digg
Facebook
MySpace
Google+
Comments
Close
Please Post Your Comment
*
The Comment you have entered exceeds the maximum length.
Submit
|
Cancel
*
Required
Comments
Please Post Your Comment
No comments yet.
Save Note
Note
View
Public
Private
Friends & Groups
Friends
Groups
Save
|
Cancel
|
Delete
Please provide your notes.
Next
|
Prev
|
Close
|
Edit
|
Delete
Genetics
Gene Inheritance and Transmission
Gene Expression and Regulation
Nucleic Acid Structure and Function
Chromosomes and Cytogenetics
Evolutionary Genetics
Population and Quantitative Genetics
Genomics
Genes and Disease
Genetics and Society
Cell Biology
Cell Origins and Metabolism
Proteins and Gene Expression
Subcellular Compartments
Cell Communication
Cell Cycle and Cell Division
Scientific Communication
Career Planning
Loading ...
Scitable Chat
Register
|
Sign In
Visual Browse
Close
Comments
CloseComments
Please Post Your Comment