Introduction
Heart failure is an epidemiologically important disease, with 
30% mortality at 1 year after diagnosis1, 2, 3. Most affected individuals have impaired systolic function with a reduced ejection fraction; their prognosis is particularly poor3. Heart failure is thought to result from complex interactions between genetic susceptibility and lifestyle and environmental factors4. Hypertension is one of the major risk factors5, 6. Starting from high blood pressure, the pathophysiological cardiac remodeling cascade proceeds to left ventricular hypertrophy (LVH) as a primarily adaptive process, which then proceeds to left ventricular dilatation, decreased systolic function and cardiac arrhythmias in some individuals, whereas others retain stable systolic function without clinical signs of heart failure. Elucidation of genetic factors has been limited by heterogeneity and by gene-environment interactions, as well as by the confounding presence of hypertension that serves as the driving force but is not sufficient as an intermediary phenotype. Familial predispositions contribute to heart failure in humans7. We used the spontaneously hypertensive heart failure (SHHF) rat, an inbred, genetically homogenous rat model that mirrors human hypertension-associated heart failure8. SHHF rats not only develop heart failure late in life after high blood pressure and left ventricular hypertrophy have developed9, but also show many of the associated transcriptional and metabolic features of the human disease10, 11, 12. The model's genetic propensity is underscored by the fact that a closely related strain, the stroke-prone spontaneously hypertensive (SHRSP) rat, does not develop heart failure despite similarly elevated blood pressure.
We conducted a segregation analysis in F2 hybrids bred from SHHF and SHRSP, thus removing blood pressure and LVH as confounding variables. To test the overall blood pressure– and cardiac hypertrophy–related impact on the development of heart failure, we generated a second F2 cross-bred from SHHF and the normotensive Wistar-Kyoto (WKY) control rat. We integrated linkage analysis of hemodynamic phenotypes with cardiac genome-wide expression profiling in the F2 population, a strategy others and we have successfully used earlier to elucidate complex traits13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23. We identified the gene encoding soluble epoxide hydrolase, Ephx2, as a heart failure susceptibility gene. Finally, we validated our findings using Ephx2 knockout mice and expression studies in human heart failure.
Results
Determination of cardiac phenotypes
We catheterized the left ventricle to precisely measure aortic and left ventricular end-diastolic pressures (LVEDP) and left ventricular end-diastolic volumes (LVEDV), as well as to construct pressure-volume loops. We found similar blood pressure and heart weight:body weight ratios, indicating LVH, in both SHHF and SHRSP compared to WKY rats. SHHF rats showed substantial left ventricular dilatation and reduced left ventricular ejection fraction, consistent with previous observations24 (Supplementary Fig. 1 online). Both SHHF and SHRSP rats had elevated LVEDP, with higher values in SHHF rats (Supplementary Fig. 1).
Blood pressure variance and LVH variance were similar or smaller in the F2(SHHF 
SHRSP) population than among progenitor strains, showing that blood pressure and LVH did not segregate in this cross. In contrast, the ejection fraction showed a large variance, ranging from 14% to 82% with a near normal distribution, consistent with the independent segregation of multiple alleles. In addition, we obtained quantitative hemodynamic parameters for LVEDP, LVEDV and left ventricular contractility (
pressure/time (mm Hg/s), dP/dtmax) (Supplementary Table 1 online). In contrast to F2(SHHF SHRSP) rats, F2(SHHF WKY) rats showed decreased blood pressure and reduced LVH (Supplementary Table 1).
The F2(SHHF WKY) data show that the overall lower average blood pressure was paralleled, on average, by a higher ejection fraction, decreased LVEDP, lower LVEDV, improved contractility and decreased B-type natriuretic peptide (Supplementary Table 1). These observations support our contention that genetic factors contributing to the complex heart failure phenotype should be assessed in an experimental cross that equalizes the influences of high blood pressures and LVH as permissive heart failure factors, thus removing them as confounders. The F2(SHHF SHRSP) population that we characterized fulfills both requirements.
Heart failure phenotypes map to rat chromosome 15
We genotyped all F2 rats as described in the Supplementary Note online. In F2(SHHF SHRSP), a locus on rat chromosome 15 centered at marker D15Rat10 showed statistically highly significant linkage (lod score 4.3) to left ventricular ejection fraction, the principal clinical parameter used to characterize the failing heart (Fig. 1). At the same locus, we further demonstrated statistically significant linkage for cardiac contractility and highly suggestive linkage to LVEDV. The linkage peak on chromosome 15 for LVEDV, however, was some 20 cM away (Fig. 1). Nonetheless, the SHHF allele was invariably associated with cardiac dysfunction. Grouping F2 rats according to zygosity at D15Rat10 did not show any genotype association with blood pressure or LVH, as determined by analysis of variance (data not shown). This finding indicated that linkage to this locus was independent of blood pressure. Our data indicated that this quantitative trait locus (QTL) strongly affected heart failure in an additive mode, accounting for 15.1% of the phenotypic variance observed for the ejection fraction. As we observed clustering of three distinct clinical traits (ejection fraction, LVEDV and contractility) at D15Rat10 with varying degrees of significance, we speculated that a composite quantitative trait variable could extract further genetic information from this locus. Using principal component analysis, we next derived a composite phenotype for each F2(SHHF SHRSP) rat from three traits (ejection fraction, LVEDV and dP/dtmax) and then subjected this composite trait to linkage analysis. This analysis resulted in a statistically more significant linkage at D15Rat10 (lod 4.8; Fig. 1), suggesting a complex but directional interplay of these phenotypes in the predisposition of heart failure at this locus, as well as indicating the limitations of individual, reductionist parameters as surrogate measurements for a complex phenotype in general. In contrast to the above, we found no evidence for linkage of ejection fraction in the F2(SHHF WKY) population anywhere in the genome (Supplementary Note and Supplementary Fig. 2 online).
Figure 1: Colocalization of cardiac phenotype QTL on chromosome 15.
Multipoint lod plots in F2(SHHF SHRSP) for ejection fraction, dP/dtmax, LVEDV. Lod scores are plotted against genetic map distances in cM. Gray marks on x axis represent marker positions. Genome-wide corrected significance levels (P = 0.05) were determined by 1,000 permutations. The dashed horizontal line indicates genome-wide significance for ejection fraction linkage; the thick and thin solid red bar indicates 1- and 2-lod-unit drop confidence interval, respectively. Significance levels for the other traits reported are slightly different and are not depicted for clarity (genome-wide significance at lod 3.2 and 3.5 for dP/dtmax and LVEDV, respectively). The composite trait variable constructed from the three traits notably increased the significance of the findings.
Transcript profiling in parental strains and F2 populations
Within the region of linkage to heart failure on chromosome 15, we identified only two transcripts that were differentially expressed with a false discovery rate (FDR) of <5% between the parental SHRSP and SHHF strains and showed a cis-regulated expression QTL (eQTL) at an FDR <5% in the F2(SHHF SHRSP) population, consistent with the close relatedness of SHHF and SHRSP rats used in this experiment (Supplementary Note). Of the two cis-regulated transcripts, Ephx2 showed the strongest genetic evidence for significant allele-specific regulation in the F2(SHHF SHRSP) population. The linkage peak (genome-wide corrected P < 10-
6; FDR < 10-
4; Supplementary Fig. 3 online) for the Ephx2 eQTL was at D15Rat10, the same marker that also defined the peak of the heart failure QTL on chromosome 15 in this cross. D15Rat10 is the nearest marker to Ephx2 and is located within 2.7 Mb, based on the rat genome reference sequence version 3.4. This finding suggested that Ephx2 transcript levels are regulated in cis at the Ephx2 gene itself and coincide with the peak of linkage to the heart failure QTL. The second cis-acting eQTL gene in this region was Mmp14. Even though it was located within the 95% confidence interval encompassing the heart failure QTL, Mmp14 was located at a distance of more than 10 Mb from the heart failure QTL peak marker D15Rat10. Strongest eQTL linkage of Mmp14 was observed with marker D15Rat83 (genome-wide corrected P < 10-
3; FDR < 0.05). Based on statistical significance and proximity to the heart failure peak marker, we thus prioritized Ephx2 as candidate for further investigations.
Sequence analysis of Ephx2
We next aimed to elucidate regulatory sequence variations underlying the observed Ephx2 cis-regulated eQTL and to screen for coding sequence differences possibly altering protein function. Sequencing 5,000 bp upstream of the first exon, we found three SNPs and a 2-nt deletion in the putative Ephx2 promoter in SHHF as compared to SHRSP and WKY rats (Fig. 2). Genomic sequencing of the entire coding exon sequence, including adjacent splice sites, revealed two synonymous and three nonsynonymous SNPs (Fig. 2).
Figure 2: Polymorphisms in Ephx2 regulatory and coding region.
Four polymorphisms were identified in the putative promoter region, of which a 2-nt insertion in SHRSP and WKY was directly located in an AP-1 binding site (arrow). Five SNPs were detected in Ephx2 exons, three of which led to an amino acid change (marked in parentheses). We found no variation in the 3' UTR. Exons are pictured as black boxes. In all instances, SHRSP and WKY shared the same allele and haplotype at Ephx2. A variant allele status was observed for SHHF.
Full size image (27 KB)Ephx2 is a heart failure susceptibility gene
We established that Ephx2 was genetically linked to the peak marker of linkage for heart failure D15Rat10 by linkage analysis (Supplementary Note). The allelic status at Ephx2 accounted for 64% of the expression differences observed, with highest expression values associated to the SHHF genotype (Fig. 3a,b).
Figure 3: Ephx2 expression and sEH activity in heart tissue.
(a) Ephx2 expression measured using microarrays in parental strains (n = 5 per group) and F2 rats according to their genotype at D15Rat10: SHRSP/SHRSP (SP/SP), n = 29; SHHF/SHRSP (HF/SP), n = 76; SHHF/SHHF (HF/HF), n = 32. (b) RT-PCR to confirm microarray expression results; parental strains (n = 5 per group) and F2 rats according to their genotype at D15Rat10 (SHRSP/SHRSP n = 10; SHHF/SHRSP n = 10; SHHF/SHHF n = 10). (c) Representative blot of sEH protein in F0 and F2 rats. Membranes were stripped and rehybridized with an antibody to 
-tubulin as a loading control. Quantification of the membrane for F2 rats is shown next to blots. (d) sEH hydrolase activity per milligram of heart tissue in parental (n = 5) and F2 rats (n = 10 per genotype). Enzymatic activity was determined by measuring the hydrolysis of 14,15-EET, which is the preferred substrate of sEH and a prominent metabolite among regioisomeric EET present in rat heart tissue, to 14,15-dihydroxyeicosatrienoic acids (14,15-DHET). (e) 14,15-EET per gram of heart tissue in parental (n = 5) and F2 rats (n = 10 per genotype). Data are given as mean 
s.d. (*P < 0.05, **P < 0.01; ***P < 0.001).
We measured Ephx2 abundance by protein blot analysis. Protein amounts in parental rats and F2 rats grouped according to zygosity at D15Rat10 mirrored the differences observed at the transcript level (Fig. 3c). Ephx2 encodes the soluble epoxide hydrolase (sEH), which hydrolyzes cardioprotective epoxyeicosatrienoic acids (EETs)25. To test the functional relevance—that is, effect on the substrate—of the allele-specific upregulation of Ephx2, we measured enzymatic hydrolase activity of Ephx2 in cardiac tissue of the parental and F2 rats grouped by zygosity at the peak marker D15Rat10. Our results demonstrated that the SHHF allele–specific upregulation of the transcript and protein also segregated with an augmented cardiac sEH activity (Fig. 3d). Finally, we directly measured the total amount of 14,15-EETs in the heart of parental and F2 rats, again grouped according to genotype. This analysis showed an effect of genotype on the total amount of cardioprotective EETs, the amount being significantly reduced when Ephx2 was transcribed from the SHHF allele (Fig. 3e). Taken together, the results suggest that genetic variation in cis leads to upregulation of the Ephx2 transcript, elevated protein, increased hydrolysis of EETs and a decrease of cardioprotective EETs in the heart mediated by the Epxh2SHHF allele. The colocalization of this pathophysiological cascade triggered by the Ephx2 eQTL with the mapping of a heart failure QTL strongly suggests that Ephx2 contributes to the observed heart failure QTL.
This interpretation of the data is corroborated by the fact that, in all instances, the Ephx2SHHF allele was associated not only with upregulation of the Ephx2 transcript and protein expression, but also with heart failure in the F2 population. Using the 10% upper and lower values of the tails of the ejection fraction distribution from F2(SHHF SHRSP) rats, we observed a marked correlation between Ephx2 transcript levels and ejection fraction (correlation coefficient r2 = -
0.219, P < 0.01); rats with low ejection fraction had higher Ephx2 transcript levels, whereas rats with high ejection fraction had lower transcript levels. No correlation was detected when the entire dataset was analyzed (r2 = -
0.048, P not significant).
Ephx2 promoter deletion creates an AP-1 binding site in SHHF
To test whether the allelic promoter variants influence Ephx2 gene expression, we performed luciferase reporter assays and compared the variant promoters among SHHF, SHRSP and WKY rats. We found a strong increase in promoter activity with the SHHF compared to the SHRSP allele (Fig. 4a). The findings are consistent with the observed cis-regulated eQTL in F2(SHHF SHRSP), in which the SHHF allele was the allele that showed higher expression (Fig. 3a,b).
Figure 4: Promoter polymorphisms of Ephx2 influence promoter activity.
(a) Schematics of reporter constructs used for transfection of H9c2 cells and, below, luciferase expression of each construct and empty pGL3-Basic vector in H9c2 cells. Each value represents mean s.d. of at least three experiments; ***P < 0.001. RLU, relative light units. (b) Top, sequence alignment of SHHF, SHRSP and WKY Ephx2 promoter sequences containing the AP-1 binding sequence. Computational analysis predicts that, in SHHF, the AP-1 consensus sequence is completely restored, suggesting that it was intact ancestrally, whereas, in SHRSP and WKY, a 2-bp insertion disrupts AP-1 binding. Bottom, an electrophoretic mobility shift assay with biotin-labeled probes. Lanes 1 and 4, probes only; lanes 2, 3, 5 and 6, probes and AP-1 recombinant protein; lanes 3 and 6, probes, AP-1 recombinant protein and unlabeled competitor.
We also confirmed the effect of allele-specific regulatory variation using allele-specific real-time PCR of cDNA from heart tissue of ten F1(SHHF SHRSP) rats. This experiment confirmed that the expression of Ephx2 was regulated in cis and that the Ephx2SHHF allele-specific transcript levels were significantly elevated compared to the Ephx2SHRSP allele (P = 8 10-
12; Supplementary Fig. 4 online).
Using computational predictions, we identified a consensus AP-1 (activator protein 1) transcription factor binding site in the Ephx2 promoter that exactly covered the 2-nt deletion in the SHHF strain (Fig. 4b). To investigate whether the mutated AP-1 binding site affected AP-1 binding in vitro, we performed electrophoretic mobility-shift assays; these showed specific AP-1 binding to the SHHF promoter, whereas AP-1 binding in SHRSP was abolished (Fig. 4b).
SNPs in the coding region do not alter hydrolase activity
Sequencing detected three nonsynonymous SNPs that could potentially affect Ephx2 enzymatic activity (Fig. 2). Conceivably, an allele-specific altered protein could confound the results attributed to allele-dependent expression differences. We demonstrated by in vitro expression of recombinant proteins that both SHHF and SHRSP Ephx2 proteins were equally active, which excludes a substantial effect of polymorphisms on the enzyme activity (Supplementary Fig. 5 online). Thus, the observed allele-dependent differences in the hydrolysis of 14,15-EET (Fig. 3d,e) were due to allele-specific expression differences of Ephx2.
Ephx2 pathway shows genotype-dependent coregulation
Because the transcriptional regulation of pathways related to Ephx2 function is of major interest, we next focused on transcripts of the arachidonic acid metabolic pathway that includes several important enzymatic reactions in the generation of epoxides (for example, 14,15-EETs) from arachidonic acid and the enzymes that regulate arachidonic acid synthesis or beta-oxidation, according to the Kyoto Encyclopedia of Genes and Genomes. We ranked the genes from the arachidonic acid metabolic pathway according to their association with the heart failure locus on chromosome 15. None of these genes had a genome-wide significant eQTL at this locus when correcting for all transcripts tested. Nevertheless, we observe a concerted differential expression—in trans to Ephx2—of the top-ranking genes, with individual P values <0.1. Applying this relaxed filtering resulted in a set of genes containing six members of the CYP superfamily and three members of the PLA2 superfamily that will be referred to as the Ephx2 pathway (Fig. 5). To quantify the small but consistent effect, we computed a score for the association of gene expression at the level of the whole pathway with a genetic marker, which was applied as a genome-scan over all genetic markers used in F2(SHHF SHRSP). We found that the maximum association of the Ephx2-pathway transcripts occurred at the Ephx2 locus and that the pathway association score had genome-wide significance (P = 2.4 10-
5).
Figure 5: Identification of regulatory trans eQTLs mapping to chromosome 15.
A simplified overview of the Ephx2 pathway is depicted. Each box is headed by an enzyme classification number describing the reaction that is catalyzed. All genes analyzed with the corresponding enzymatic function are listed in the boxes below. For each gene, a color code indicates the genotype-specific average expression values according to zygosity—SHHF/SHHF (HF/HF), SHHF/SHRSP (HF/SP) and SHRSP/SHRSP (SP/SP), from left to right—at the Ephx2 locus. The expression matrix has been centered and scaled to 
= 1 before visualization. Orange, overexpression; white, baseline; cyan, underexpression. CYP genes are subdivided into two columns, with the gene products of those on the left producing predominantly EETs, which are converted to dihydroxyeicosatrienoic acids (DHETs) by sEH, and the right producing predominantly hydroxyeicosatetraenoic acids (HETEs). The Ephx2 pathway has its genome-wide pathway linkage peak at the Ephx2 locus with P < 10-
6.
Ephx2 gene ablation protects from cardiac dysfunction
To corroborate our findings that Ephx2 is involved in the pathogenesis of hypertension-induced heart failure, we next studied Ephx2 knockout mice. We hypothesized that Ephx2 knockout mice could represent an extreme opposite compared to the detrimental effects associated with elevated Ephx2 expression observed in F2(SHHF SHRSP).
At baseline, Ephx2 null mice and C57BL/6ByJ wild-type controls had normal blood pressure and normal left ventricular function (Fig. 6a,b) without signs of LVH. To investigate cardiac function under pressure overload conditions similar to those we observed in SHHF rats, we used prolonged (3 weeks) chronic intraperitoneal infusion of the potent vasoactive peptide angiotensin II (Ang II) at high dosages. Because the Ephx2 promoter harbors an Ang II-inducible AP-1 binding site that is also predicted in mouse, this experimental strategy was expected to amplify Ephx2-specific pathophysiological responses to pressure overload in wild-type but not in Ephx2 knockout mice.
Figure 6: Characterization of Ephx2- /- mice by telemetry and echocardiography.
![Figure 6 : Characterization of Ephx2|[minus]|/|[minus]| mice by telemetry and echocardiography.](/ng/journal/v40/n5/thumbs/ng.129-F6.jpg)
(a) Blood pressure measurements (mean arterial pressure, MAP) in conscious mice (values given as mean s.d.; n = 4 per group). (b) Ejection fraction, measured by echocardiography (mean s.d.), in Ephx2-
/-
(n = 17) and wild-type (n = 15) mice before and after Ang II treatment; **P < 0.01. (c) Examples of M-mode echocardiography in Ephx2-
/-
and wild-type mice after Ang II treatment. Blue vertical lines, left ventricular end-systolic diameter; orange vertical lines, left ventricular end-diastolic diameter.
Telemetric blood pressure recordings showed a similar temporal blood pressure increase in both groups (Fig. 6a). As expected from our rat experiments, in which the SHHF allele was associated with high Epxh2 expression and reduced ejection fraction, we found a substantially decreased ejection fraction after 3 weeks of Ang II infusion in wild-type mice compared to knockout mice (Fig. 6b,c).
Cardiac arrhythmias are common features and a principal cause of death in the failing human heart26. We therefore tested whether Ephx2 gene ablation protected against Ang II-induced arrhythmias. The ventricular refractory period was noticeably prolonged in Ephx2 knockout mice compared to wild-type mice (Fig. 7a), supporting a prolonged—that is, beneficial—cardiac repolarization in Ephx2 knockout mice. Notably, there were spontaneous supraventricular and premature ventricular contractions in wild-type mice that were not observed in Ephx2 knockout mice (Fig. 7b). We then applied programmed electrical stimulations and induced supraventricular tachycardia and atrial fibrillation, as well as short episodes of non-sustained ventricular tachycardia, in all wild-type mice, whereas Ephx2 knockout mice had stable sinus rhythm even under isoproterenol challenge (Fig. 7c). These findings were corroborated using an independent pressure overload model using thoracic aortic banding (Supplementary Note and Supplementary Fig. 6 online).
Figure 7: Electrophysiological characterization of Ephx2- /- and wild-type mice after Ang II treatment (n = 4 per group).
![Figure 7 : Electrophysiological characterization of Ephx2|[minus]|/|[minus]| and wild-type mice after Ang II treatment (n = 4 per group).](/ng/journal/v40/n5/thumbs/ng.129-F7.jpg)
(a,b) Marked prolongation of ventricular refractoriness (a) and spontaneous premature ventricular contractions (b,*) were seen in all wild-type mice but could not be observed in any Ephx2 null mice (values given as mean (central line) s.d.; wide bar, minimum and maximum values). Recordings from a single representative mouse from each group are given. The lower trace for ECG recordings in a wild-type mouse represents a magnification of the boxed area from the upper trace. (c) Representative recordings from in vivo electrophysiological studies. Programmed stimulation (stim) in the right ventricle of wild-type mice induced non-sustained atrial fibrillation, indicated by the high frequency signals in the right atrium (RA) and the variability in R–R intervals in the surface ECG (upper panel). Stimulation also induced ventricular tachycardia in wild-type mice. High rate pacing in the right ventricle (RV) is followed by non-sustained fast ventricular tachycardia, eventually recovering into normal sinus rhythm (middle panel). Programmed stimulations could not induce arrhythmias in Ephx2-
/-
mice (lower panel).
EPHX2 expression is decreased in humans with heart failure
If an imbalance of EET concentrations is critical to the pathogenesis of heart failure, we would expect transcriptional regulation of Ephx2 to maintain high EET levels to be a general defense mechanism, irrespective of the trigger eliciting heart failure. To test this hypothesis, we determined cardiac EPHX2 expression in biopsies from subjects with ischemic heart failure and from controls. Cardiac EPHX2 expression was reduced by 61.5% (P = 0.025) in subjects with heart failure, compared to that in controls (Fig. 8). This finding supports transcriptional downregulation of EPHX2 as a potentially beneficial adaptive mechanism in human heart failure. This hypothesis is supported by the fact that wild-type mice that developed a reduced ejection fraction under two independent models of pressure overload conditions (chronic Ang II infusion and aortic banding) showed a significant downregulation of Ephx2 (Supplementary Note and Supplementary Fig. 7 online).
Figure 8: EPHX2 mRNA expression is downregulated in individuals with heart failure.
Ejection fraction in individuals with heart failure was severely reduced (n = 6 per group; ejection fraction 29% 5% in affected individuals versus >50% in controls), although systolic and diastolic blood pressures were not different between groups. EPHX2 mRNA expression was determined using quantitative TaqMan RT-PCR (values given as mean s.d.; n = 6 per group); *P = 0.025.
Discussion
In humans, hypertension associated with sustained cardiac hypertrophy represents one of the most common causes of heart failure27, 28. Cardiomyocyte hypertrophy is a primary adaptive response to diminish wall stress induced by mechanical overload. However, the resultant LVH is clinically associated with increased morbidity and mortality2. So far, the genetic and molecular checkpoints gating the transition from hemodynamic overload to maladaptive cardiac remodeling and heart failure are poorly understood29, 30. Recent data suggest a substantial contribution of familial factors that could account for as much as 70% increased risk for heart failure, after adjustments for other risk factors including LVH7.
We present evidence that genetic variation in Ephx2 facilitates progression from hypertension and cardiac hypertrophy to heart failure in a rat model of human hypertension and heart failure. We first generated experimental F2 intercrosses of SHHF with SHRSP and WKY as reference strains with different set points regarding blood pressure and LVH. Hemodynamic characteristics of parental rat strains were in line with already established echocardiographic and hemodynamic data for rats of strains SHHF24 and SHR31, a close relative of the SHRSP. We documented marked cardiac hypertrophy in SHHF rats, together with decreased left ventricular ejection fraction, indicating the expected alteration of systolic function. By contrast, SHRSP rats retain systolic function, as documented by others31, 32, 33, 34 and as we have demonstrated here. Thus, we used SHRSP, a strain that is genetically closely related to SHHF, as a reference model. However, rats of our candidate strain, SHHF, clearly demonstrated decreased ejection fraction and other heart failure features also common to humans.
Linkage analysis to left ventricular hemodynamic parameters of cardiac function in experimental SHHF crosses with WKY and SHRSP revealed significant segregation of ejection fraction with a QTL on rat chromosome 15 harboring Ephx2. The fact that significant linkage of the Ephx2 locus to left ventricular ejection fraction was detected in the F2(SHHF SHRSP) cross but not in the F2(SHHF WKY) cross underscores the pathophysiological importance of hypertension and LVH as permissive factors in deteriorating cardiac function. In human heart failure, a decreased ejection fraction and clinical symptoms of heart failure are preceded by structural and hemodynamic abnormalities, such as increased LVEDV and reduced contractility35. We identified these phenotypes in F2 populations; both traits significantly segregated with the Ephx2 locus when both crosses were analyzed jointly (data not shown), suggesting a common regulation of hemodynamic responses to pressure overload by the chromosome 15 QTL.
Our genome-wide linkage analysis revealed two more significant QTLs, on chromosome 8 and chromosome 4, for ejection fraction and associated hemodynamic traits in the F2(SHHF SHRSP) population (J.F., J.M., S.P., H.S., M.H. et al., unpublished data). Further characterization of these loci, together with the eQTL data obtained, may generate additional insights into the genetic factors contributing to heart failure in this model.
We performed linkage analysis of cardiac gene transcription as intermediate phenotypes and identified Ephx2 as a prime candidate gene within the heart failure chromosome 15 QTL due to its highly significant cis regulation and its expression pattern across parental strains. Ephx2 encodes an enzyme responsible for the conversion of EETs to their corresponding diols36, thus diminishing their biological activity37, 38, 39. We were aided by the recent development of pharmacological Ephx2 inhibitors. These experiments showed that EETs elicit a number of cardioprotective actions including vasodilation, anti-inflammatory effects, endothelial cell proliferation, antiaggregatory effects and inhibition of vascular smooth muscle cell migration25. In addition, recent studies have shown an increased availability of cardioprotective EETs in Ephx2 knockout mice with associated benefits on left ventricular function due to reduced infarction size after ischemic injury40. Similar to our own observations, baseline cardiac anatomy and contractile function in this study were not different in Ephx2 knockout and wild-type mice, suggesting that detrimental effects of environmental stress to the heart in general (ischemia or pressure overload) can be counteracted by a decreased action of Ephx2 together with the maintenance of high EET levels.
The Ephx2 locus accounted for almost two-thirds of the Ephx2 expression variability in a large proportion in the F2(SHHF SHRSP) population. However, the association of the Ephx2 locus with ejection fraction as a far more complex phenotype was smaller in magnitude. Overall, the Ephx2 locus accounted for about 15% of the phenotypic variance for ejection fraction. We demonstrated that Ephx2 transcript levels significantly correlate with the tails of the distribution for ejection fraction, providing independent evidence that the allelic modulation of Ephx2 transcript levels plays a causal role in the development of heart failure. Rats with high transcript levels had lower ejection fractions, whereas rats with lower transcript levels had higher ejection fractions. It is important to note that allelic variants of Ephx2 appear to impact heart failure as a complex disease through perturbations of trans-regulated gene networks, and that transcriptional phenotypes are again intermediate to the clinically defined disease. We showed that transcriptional regulation not only of Ephx2 but also of several other genes in the entire Ephx2 pathway are affected by the chromosome 15 QTL in our model. We applied pathway linkage scores to assess the statistical significance of this genetic effect and found a genome-wide significant association of expression on a pathway level with the maximum effect at the Ephx2 locus.
To uncover DNA variants that can account for the observed alterations of Ephx2 expression and function, we performed comparative sequence analysis of Ephx2 in heart failure and reference strains. We detected a regulatory insertion/deletion polymorphism within the AP-1 binding site and showed enhanced promoter activity in SHHF. However, our findings do not preclude that other regulatory variants contribute to the allele-specific differences in Ephx2 expression besides this regulatory insertion/deletion polymorphism in AP-1. Since AP-1 activation has also been demonstrated in human heart failure41, and Ang II serves to increase Ephx2 expression through AP-1 activation42, it seems possible that EPHX2 promoter variants affecting AP-1 binding may also constitute a contributory mechanism for heart failure susceptibility in humans. It would be interesting to test variation at EPHX2 for association with human hypertensive heart failure.
Ephx2 was established as a primary heart failure candidate gene by combined evidence from linkage analyses, global F2 transcription profiling, sequence analysis and functional data on protein abundance and Ephx2 enzymatic activity. Additional experimental support for the specific relevance of Ephx2 in the genesis of hypertension-induced heart failure was generated by experiments in mice with targeted disruption of Ephx2. In keeping with previous studies in two independently-derived colonies of Ephx2 knockout mice, including the one used in our experiment43, we found no differences in blood pressure and ejection fraction between knockout and wild-type mice at baseline. However, after exposing both strains to pressure-overload conditions, Ephx2 knockout mice showed only about half of the ejection fraction decline seen in wild-type mice, although blood pressure increases to high and prolonged dosages of Ang II were similar in both strains. This situation resembles the SHHF model, which develops heart failure while its equally hypertensive cousin, the SHRSP, does not. Furthermore, hypertensive Ephx2 knockout mice are protected from cardiac arrhythmias. This is in agreement with findings using pharmacological inhibition of Ephx2 (ref. 44) and may be explained by previous observations suggesting that EETs are modulators of ion channel function in cardiomyocytes (for review, see ref. 25).
We are the first to present evidence that Ephx2 is a heart failure susceptibility gene. Ephx2 allelic variability leads to quantitative differences in transcript levels that in turn quantitatively regulate the metabolism of cardioprotective epoxyeicosatrienoic acids. The implications of our findings are important as Ephx2 has already found attention as a pharmacological target for cardiovascular disease44. In light of elevated Ephx2 expression levels associated with the development of heart failure in rats, and protection from heart failure in Ephx2-deficient mice, we interpret EPHX2 downregulation in human failing hearts as an adaptive transcriptional process aimed at maintaining high EET levels. EETs are known to exert cardioprotective actions45. Conversely, individuals with genetically determined, persistently high cardiac EPHX2 expression, together with arterial hypertension and LVH, probably represent a high-risk group for disease progression to heart failure. The role we have shown for Ephx2 in the complex initiation of the disease in rats and mice, as well as its regulation in human heart failure, suggest a potential avenue for developing new heart failure treatments.
Methods
Experimental crosses and mice.
SHHF/Bbb, WKY/Bbb and SHRSP/Bbb rats were from our colonies at the Max-Delbrück Center for Molecular Medicine and were maintained under strict inbreeding. SHHF breeders (SHHF/Mcc) were initially obtained from the original colonies of S.A. McCune (Department of Integrative Physiology, University of Colorado at Boulder). SHHF were mated with SHRSP and WKY, respectively, to generate F1 progenies. F1 rats for each cross were brother-sister mated, resulting in 166 F2(SHHF SHRSP) (87 male and 79 female) and 189 F2(SHHF WKY) (93 male and 96 female) rats. All rats were investigated at 14 months of age. Ephx2-
/-
mice were previously described43 and were further backcrossed for nine generations onto C57BL/6ByJ before used in our studies. Heterozygote F9 mice were crossed to obtain homozygote null and wild-type littermates used in this study. All mouse experiments were performed starting at 16 weeks of age; only male mice were used. All animal experiments were conducted according to the guidelines of the American Physiological Society and with approval of the animal protection authorities of the state of Berlin (Landesamt Für Gesundheit und Soziales).
Hemodynamic measurements.
Rats were intubated and ventilated under isoflurane anesthesia. A 2-French conductance catheter (SPR 838 Aria, Millar Instruments) was inserted into the left ventricle through the right carotid artery. Left ventricular contractility was obtained from the ventricular pressure curves. Ventricular volumes were determined as described46.
Telemetry.
Telemetric blood pressure measurements were performed as previously described47. The data from the TA11PA-C20 device were sampled every 5 min for 10 s continuously day and night with a sampling rate of 1,000 Hz. Mice were allowed to recover for three days before baseline recording started. SBP, DBP and HR were recorded using the DATAQUEST software.
Echocardiography.
Echocardiography was performed before and after 3 weeks of Ang II treatment as previously described48. Briefly, mice were anesthetized with 2% isoflurane and kept warm on a heated platform. Temperature and ECG were continuously monitored. Cardiac function and morphology were assessed by echocardiography with a VisualSonics Vevo 770 High-Resolution Imaging System with the use of a high-resolution (40 MHz) transducer.
Surface ECG and in vivo electrophysiology studies.
Each mouse was slightly anaesthetized with 1.6% isoflurane and surface Electrocardiograms (ECG) were obtained using adhesive electrodes on each limb after 3 weeks of Ang II treatment. ECG standard intervals were measured in six standard leads. In vivo electrophysiological studies (EPS) with programmed electrical stimulations (PES) were performed at baseline and after isoproterenol treatment (intraperitoneal injection of 0.02mg isoproterenol/g body weight). Briefly, an octapolar 2-French electrode catheter (CIBer mouse cath, NuMed Inc.) was placed via the right jugular vein into the right atrium and ventricle under electrogram guidance. Electrical stimuli were delivered at twice diastolic threshold with a pulse duration of 1.0 ms. PES used up to three extra stimuli after 10 basal stimuli (cycle length 90/100/120 ms). Additionally, short episodes of burst pacing (10 s) with cycle lengths down to 60 ms were applied. Only reproducible arrhythmias of three beats or more were considered. All electrograms were digitally stored for off-line analysis.
Marker selection and genotyping.
SNPs were identified by the STAR consortium and are publicly available at http://www.ensembl.org. SNP genotyping in F2 crosses was carried out using the GoldenGate genotyping platform (Illumina Inc., San Diego). Microsatellite markers were genotyped using fluorescence-based semi-automated genotyping on an ABI3730 sequencer. SNP and microsatellite genotyping results are reported in Supplementary Table 2 online. Statistical methods for genetic mapping and linkage analysis are reported in Supplementary Methods online.
Preparation of labeled cDNA and hybridizations to microarrays.
We pulverized whole frozen hearts and stored aliquots at - 80 °C for RNA and - 20 °C for protein experiments. Total RNA was extracted using Trizol reagent (Invitrogen) and purified using RNeasy Mini kit (Qiagen) in accordance with the manufacturer's protocol. Double-stranded cDNA was synthesized from total RNA using the One-Cycle cDNA Synthesis Kit (Affymetrix) and cRNA biotinylated from cDNA using the IVT Labeling Kit (Affymetrix). We hybridized the fragmented cRNA samples to Affymetrix Rat Genome 230 2.0 arrays in accordance with the Affymetrix protocol. Statistical methods for eQTL analysis are reported in Supplementary Methods.
Real-time RT-PCR.
We reverse-transcribed total RNA with oligo(dT) primers (Gibco-BRL) and Superscript II reverse transcriptase (Invitrogen). We designed primers and probes using Primer Express 2.0 (Applied Biosystems) and placed them in non-polymorphic regions of the rat Ephx2 gene (Supplementary Table 3 online). For human TaqMan probes, Applied Biosystems assay (hs00157403_m1) was used. TaqMan analysis was performed using an Applied Biosystems 7900HT system. We normalized expression levels to GAPDH RNA expression by using the 2-
CT method.
Analysis of Ephx2 hydrolase activity.
Hydrolase activities were determined by using 10 
g of heart-extracted proteins (10,000g supernatant) dissolved in 100 mM potassium phosphate buffer, pH 7.2. Proteins were incubated with 10 M [1-14C]14,15-EET for 5 min at 37 °C. Adding ethyl acetate terminated the reaction. After extraction, products were evaporated under nitrogen and dissolved in ethanol. Metabolites were resolved by reverse-phase HPLC (Shimadzu LC 10 Avp). Retention times for 14,15-EET/-DHET were 23.4 and 12.7 min, respectively. Methods for protein blot analyses and production of recombinant proteins are reported in Supplementary Methods.
Promoter activity assays.
The - 2079 to +41 bp Ephx2 promoter was cloned into a pGL3-Basic vector (Promega) containing the firefly luciferase gene as a reporter encompassing either the SHHF or the SHRSP promoters, and integrity was confirmed after cloning by sequence analysis. Since SHRSP and WKY share common haplotypes, we cloned only the promoter variants from one of the strains (SHRSP) and tested the variants against the SHHF promoter. For transient transfections, H9c2 (ATCC) cells were cotransfected with pRL-SV40 plasmid (Promega) as a normalizing control. All transfections were carried out in triplicates. Firefly and Renilla luciferase activity was analyzed with the Dual-Luciferase Reporter Assay System according to manufacturer's description (Promega).
Electrophoretic mobility shift assay.
Binding of AP-1 to the Ephx2 promoter in vitro was carried out as previously described49. Briefly, synthetic double-stranded and 3' biotin-labeled oligonucleotides and recombinant protein (Promega) were incubated at room temperature for 20 min using the LightShift Chemiluminescent EMSA Kit (Pierce). The binding reaction contained 1 binding buffer, 0.01 g/l poly(dI:dC), 0.05% NP-40, 50 mM KCl, 0.5 mM MgCl2 and 0.5 mM DTT. Two-hundred-fold molar excesses of unlabeled oligonucleotides were used as competitors. Reaction mixtures were separated with 5% polyacrylamide gels. Probes were transferred to a positively charged nylon membrane (Ambion), UV-cross-linked and detected by stabilized Streptavidin-Horseradish Peroxidase Conjugate (Pierce).
Quantification of EETs in the heart.
Heart tissue was homogenized using a biopulverizer (Biospec Products). 40 mg aliquots were subjected to alkaline hydrolysis, followed by a solid phase extraction. Quantification of EETs in the heart homogenates was done by the Lipidomix GmbH, Berlin, using HPLC (Agilent 1200) with tandem mass spectroscopy Triple Quad MS/MS (Agilent 6410). Detection limit for EETs was 0.2 ng per gram of tissue.
Human myocardial biopsies for mRNA expression analysis.
Full thickness myocardial biopsies were harvested at the time of cardiac surgery with the approval of the Hammersmith Hospital Research Ethics Committee. Heart function was characterized by trans-thoracic echocardiography. Individuals with coronary artery disease but having normal ventricular function were used as controls. Individuals with heart failure were defined as an ejection fraction of <45% secondary to ischemic heart disease. Written informed consent was obtained from all subjects.
URLs.
SNPs used are available at http://www.ensembl.org and microsatellite markers at http://rgd.mcw.edu/.
Accession codes.
Scanned microarray data have been submitted to ArrayExpress with accession code E-TABM-418. Ephx2 DNA variants have been submitted to dbSNP: ss86236892, ss86236893, ss86236894, ss86236895, ss86236896, ss86236897, ss86236898, ss86236899 and ss86236900.
Note: Supplementary information is available on the Nature Genetics website.
Author contributions
N.H. and J.M. developed the project. N.H. directed the project. J.F., S.P., C.G., C.S., H.M., G.P., K.S., S.A.C. and W.-H.S. performed genetic and biochemical analysis. J.M., A.H., R.F., A.S., V.G., R.D. and F.C.L. performed and analyzed physiological experiments. M.H., H.S., O.H., M.V. and K.R. carried out statistical analysis. B.D.H., S.M.W., K.L. and S.A.C. contributed materials. N.H. wrote the paper with J.M., J.F. and F.C.L.