The polydactylous rat strain (PD/Cub) is a highly inbred (F > 90) genetic model of metabolic syndrome. The aim of this study was to analyze the genetic architecture of the metabolic derangements found in the PD/Cub strain and to assess its dynamics in time and in response to diet and medication. We derived a PD/Cub × BN/Cub (Brown Norway) F2 intercross population of 149 male rats and performed metabolic profiling and genotyping and multiple levels of genetic linkage and statistical analyses at five different stages of ontogenesis and after high-sucrose diet feeding and dexamethasone administration challenges. The interval mapping analysis of 83 metabolic and morphometric traits revealed over 50 regions genomewide with significant or suggestive linkage to one or more of the traits in the segregating PD/Cub × BN/Cub population. The multiple interval mapping showed that, in addition to “single” quantitative train loci, there are more than 30 pairs of loci across the whole genome significantly influencing the variation of particular traits in an epistatic fashion. This study represents the first whole genome analysis of metabolic syndrome in the PD/Cub model and reveals several new loci previously not connected to the genetics of insulin resistance and dyslipidemia. In addition, it attempts to present the concept of “dynamic genetic architecture” of metabolic syndrome attributes, evidenced by shifts in the genetic determination of syndrome features during ontogenesis and during adaptation to the dietary and pharmacological influences.
- quantitative trait loci
- insulin resistance
metabolic syndrome is a complex condition arising from an intricate network of interacting genetic and environmental factors. The dynamics of the increase in its worldwide prevalence qualify the syndrome as a major healthcare issue for years to come. The syndrome comprises several clinical features, each representing a complex trait of its kind, i.e., hypertriglyceridemia, hyperinsulinemia, insulin resistance, obesity, and hypertension (11, 36). It is therefore self evident that detailed analysis of the genetic determinants underlying such metabolic clustering is rather complicated in the general human population and only somewhat more feasible in particular circumstances, e.g., in population isolates (32). Although some progress has been made and several studies succeeded in identification of genes potentially involved in the pathogenesis of metabolic syndrome (13, 17, 35, 51), its genetic determination is far from being fully understood.
As in other complex diseases, defined animal models of metabolic syndrome are proving to be important tools for deciphering the causative genes and gene-gene and gene-environment interactions (8). There are numerous rodent strains expressing all or a subset of metabolic syndrome attributes found in humans [e.g., Otsuka Long-Evans Tokushima Fatty rat (16), Goto-Kakizaki rat (10), hereditary hypertriglyceridemic rat (53, 54), and spontaneously hypertensive rat (31, 35)]. The information acquired from the study of diverse strains is not redundant, however, as each model represents a particular combination of alleles leading to the unique form of manifestation of the metabolic syndrome. Each of the strains’ genotypes also provides a singular setting for gene-environment interactions leading to a strain-specific pattern of sensitivity to external stimuli (diet, exercise, stress, medication) and temporal factors (ontogenetic stage, age of disease onset).
Our previous studies and pilot experiments indicate that the polydactylous rat strain (PD/Cub), a genetic model of the metabolic syndrome, is remarkably susceptible to the aggravation of insulin resistance and serum triglyceride concentrations by diets rich in simple carbohydrates (46). Moreover, these predispositions are in striking contrast to the Brown Norway (BN/Cub) strain, which shows relative resistance to such challenge. The observed strain distinctions extend also to the aging process, where dyslipidemia and glucose intolerance tend to progress with time in PD/Cub but not in BN/Cub (L. Kazdova, unpublished observation). Reflecting these findings, we introduced sucrose diet feeding and dexamethasone (a known inducer of insulin resistance) administration in the F2 as tools for nutrigenetic and pharmacogenetic resolution of metabolic syndrome-related traits and designed an experimental protocol to analyze the genetic architecture of the metabolic syndrome in the PD/Cub strain by means of deriving the PD/Cub × BN/Cub F2 intercross population and its subsequent metabolic profiling and genotyping and multiple levels of genetic linkage and statistical analyses at different stages of ontogenesis and under dietary and pharmacological challenges.
Rat strains: progenitors and F2 segregating population.
The polydactylous rat strain (PD/Cub, Rat Genome Database ID no. 728161) is a highly inbred rat strain (F > 90, verified by several total genome scans) kept since 1969 at the Institute of Biology and Medical Genetics, First Faculty of Medicine, Charles University in Prague (IBMG). It carries a mutant allele of Lx gene, which gives rise to an apparent leg malformation phenotype, the polydactyly-luxate syndrome (PLS; Ref. 19). PD/Cub and PLS-carrying recombinant inbred and congenic strains derived from this strain (20) have been exploited as a model of limb development and teratology (2, 21) and hypertriglyceridemia (55), and it was established as a model for metabolic syndrome (46). Recently, the PD/Cub rat was shown to possess a distinct pharmacogenetic profile in response to several ligands of nuclear receptors, e.g., thiazolidinediones (43), fibrates, and retinoids (47).
The Brown Norway (BN/Cub Rat Genome Database ID no. 737899) rat strain was transferred from the United States to IBMG in 1964 and since then bred by brother × sister mating for more than 90 generations. By reciprocal crossing of PD/Cub and BN/Cub progenitors, the F1 hybrids (PD/Cub × BN/Cub and BN/Cub × PD/Cub) were obtained. These were further crossed to generate F2 hybrids. Only male F2 rats were used in the study (n = 149).
All experiments were performed in agreement with the Animal Protection Law of the Czech Republic (311/1997), which is in compliance with European Community Council Recommendations for the use of laboratory animals (86/609/ECC). The experimental protocol was approved by the Ethics Committee of the First Faculty of Medicine of Charles University, Prague.
On the basis of our preliminary observations (44), we established a phenotyping protocol for the F2 male population. At the ages of 2 and 6 mo, the rats were weighed and a blood sample was drawn for determination of fasting glucose and triglyceride (TG) levels. At the age of 10 mo, the rats were subjected to phenotype profiling [weight, oral glucose tolerance test (OGTT), and serum levels of TG, free fatty acids (FFA), insulin, and uric acid (UA)] at baseline after being fed a high-sucrose diet (HSD; 70% calories as sucrose) for 1 wk and after administration of dexamethasone (DEX; Dexamed, Medochemie) in drinking water (0.026 mg/ml) for 3 days (while still fed HSD). At the end of DEX treatment, the animals were killed, and the heart, liver, kidneys, and epididymal fat pads were collected and weighed.
The OGTT was performed after overnight fasting. Blood for glycemia determination (Ascensia Elite Blood Glucose Meter; Bayer HealthCare, Mishawaka, IN) was drawn from the tail vein at intervals of 0, 30, 60, and 120 min after intragastric glucose administration to conscious rats (3 g/kg total body wt, 30% aqueous solution). Blood samples were drawn by puncture of the retroorbital plexus in light halothane anesthesia. The TG and UA concentrations were assessed with Triglycerides Liquicolormono (Human, Wiesbaden, Germany) and UA plus kit (Roche Diagnostics, Mannheim, Germany) enzymatic colorimetric tests, respectively, both on a Roche/Hitachi MODULAR analyzer. Serum FFA concentrations were determined by use of an acyl-CoA oxidase-based colorimetric kit (Roche Diagnostics). Serum insulin concentration was determined using an RIA kit for rat insulin assay (Amersham Pharmacia Biotech, Little Chalfont, UK).
The rat genomic DNA was isolated from tail incision samples using a modified phenol extraction method. Polymorphic microsatellite loci were amplified by PCR, using the ABI GeneAmp 9700 cycler (Applied Biosystems), with conditions optimized for each marker. Sequences of the selected markers were retrieved from public databases (Rat Genome Database, http://rgd.mcw.edu/; Wellcome Trust Centre for Human Genetics, http://www.well.ox.ac.uk/; or Whitehead Institute/MIT Center for Genome Research, http://www.genome.wi.mit.edu/). The PCR products were separated on polyacrylamide (7–10%) or agarose (2–4%) gels, stained by ethidium bromide, and visualized using a Typhoon 8600 (Molecular Dynamics) digitalization system and the ImageQuant analysis software package (Molecular Dynamics). Because this is the first genome-scale analysis of a PD/Cub-based cross, a comprehensive set of polymorphic markers between the two progenitors (PD/Cub × BN/Cub) had to be established. We scanned over 1,000 mostly microsatellite markers, finding 320 of them to be polymorphic between the two progenitor strains. We selected and genotyped 220 markers (Supplemental Table S2), including 34 markers previously established as polymorphic (46), in the whole set of PD/Cub × BN/Cub male F2 progeny to achieve ∼10 cM average spacing between markers distributed across the whole genome.
The normality of distribution of the measured phenotypes was assessed using the Kolmogorov-Smirnov test. When positive, the values were normalized with simple function transformation (log, square root), and the normalized values were used for further statistical analyses. When comparing more than two groups, we used one- or two-way ANOVA with post hoc Tukey’s honest significance difference test for comparison of the specific pairs of variables. Repeatedly assessed phenotypes in the same individuals were analyzed by repeated measures ANOVA with post hoc Tukey’s test for comparison of specific pairs of conditions. Null hypothesis was rejected whenever P < 0.05. All statistical calculations were performed using StatView v.5.0.1. software (SAS Institute).
We performed interval mapping and multiple interval mapping on the dataset of obtained phenotype and genotype information using MapManager v.b20 (26). For interval mapping and multiple interval mapping, the significance threshold was set for each analyzed trait by the permutation test (2,000 permutations in 1-cM steps), and the permutation bootstrap (implemented in MapManager) was used for assessing the confidence intervals.
Metabolic and morphometric profile of the F2 population.
The summary of the metabolic profile of BN/Cub, PD/Cub, and BN/Cub × PD/Cub F2 male rat populations at the age of 10 mo (standard diet), after being fed HSD for 1 wk and after administration of DEX for an additional 3 days, is shown in Table 1. In most of the followed parameters, the F2 rats showed values between those observed in progenitor strains. However, for insulin concentrations, we observed a gradual shift of the distribution of F2 values along with the increasing strength of environmental challenge, i.e., while on standard diet, most of the values were well within the range marked by the progenitors, but a significant proportion of values fell either above the levels found in PD/Cub or below those of BN/Cub after DEX administration. This finding suggests the presence of gene-gene and gene-environment interactions influencing the variation in insulinemia. Sucrose diet administration induced an overall increase in total body weight and TG and FFA concentrations (repeated measures ANOVA, P < 0.001). As expected, DEX impaired all of the assessed metabolic parameters of the F2 rats and increased their total variance, an effect facilitating the search for potential susceptibility loci (Fig. 1). DEX administration induced a selective, substantial (8.7-fold) rise of TGs in the PD/Cub progenitor strain in contrast to the lack of change in BN/Cub and a twofold average increase in F2 rats. The intercorrelation among the followed attributes of metabolic syndrome revealed distinct patterns, changing with diet and DEX administration, the latter inducing the tightest correlation among the parameters (Supplemental Fig. S1; available at the Physiological Genomics web site).1
Interval mapping: single quantitative trait loci.
Interval mapping analysis of the merged phenotyping and genotyping datasets revealed over 50 genomic regions on chromosomes 1, 2, 3, 4, 5, 7, 8, 9, 14, 15, 16 and 17 with significant or suggestive linkage (evaluated by permutation analysis) to one or more of the measured phenotypes in the segregating PD/Cub × BN/Cub population. The highest LOD score overall reached 6.91 for the trait of relative kidney weight (Supplemental Fig. S2) linked to the marker D1Wox22 (within a gene coding for insulin-like growth factor-2; Igf2). All identified quantitative train loci (QTLs) for the analyzed traits are summarized in Table 2, and an extensive overview of their phenotypic relevance is given in Supplemental Table S1. The genomewide distribution of the identified QTLs reveals a pattern with several evident clusters of QTLs present on chromosomes 3, 14, 15, and 17. These may indicate genomic regions potentially harboring genes with a pleiotropic influence on attributes of the metabolic syndrome or simply groupings of several QTLs within a limited location.
Two separate QTL clusters reside on chromosome 3. The first group relates to total body weight, adiposity, and UA concentration, whereas insulin-related traits pertain to the second one (Fig. 2). Moreover, several of the loci are involved in the strongest identified epistatic interactions influencing body weight, adiposity, and insulinemia (Table 3). All QTLs identified on chromosome 14 belong only to glucose tolerance traits (after sucrose and DEX challenges). On the other hand, a common peak linkage for diverse traits including TGs, epididymal (visceral) fat mass, and glucose tolerance was found on chromosome 15 in the vicinity of the marker D15Rat6. Finally, the QTL cluster on chromosome 17p12–14 includes significant linkages for body weights at all time points, kidney and epididymal fat pad weights, and insulin levels after HSD administration. Although the cluster covers a relatively large region of the chromosome, with many known and predicted genes, the result of bootstrap testing suggests that the QTLs lie near the marker D17Rat84.
Multiple interval mapping: epistatic QTLs.
The first analysis showed that there are >60 pairs of loci across the whole genome influencing the variation of particular traits in an epistatic fashion. A considerable number of the identified loci were distinct from the interval mapping group, i.e., such loci would have been missed if only a simple linkage model was employed. For instance, while no single QTLs were identified on chromosome 10, the region D10Rat72–D10Rat161 is present in the strongest interactions involving three glucose tolerance-related traits (Table 3). Because the method considers only loci at the typed markers, usually several “interactions” comprising closely linked loci on two respective chromosomes may reflect the presence of epistasis between two QTLs in the vicinity of the identified markers. The list of the most significant interactions for traits showing the presence of at least one significant epistatic QTL pair is given in Table 3.
Genetic architecture of traits: temporal, nutritional, and pharmacological aspects.
Another way of processing the linkage results is not via positional clustering of the QTLs, but through the “functional” clustering of QTLs linked to a single trait. The design of the protocol allowed us to evaluate the genetic components of metabolic and morphometric parameters in three specific ontogenetic stages and under dietary and pharmacological influences. In this manner, we got several complete sets of single and epistatic pairs of loci influencing the trait variation under specific conditions for each of the analyzed traits. Moreover, by combining these cross-sectional “snapshots,” a much more complex picture of the dynamic “genetic architecture” of the traits emerges. The genetic determinants of body weight in our experimental set may serve as an example of this approach. Altogether, we identified 14 significant and suggestive single QTLs on chromosomes 3, 7, 8, and 17 and 10 significant interactions of loci on multiple chromosomes influencing body weight. When viewed from the temporal perspective, we can observe that the single loci on chromosomes 3 and 17 show significant linkage throughout the study, whereas those on chromosome 7 or 8 and several variable epistatically acting loci on other chromosomes appear only at specific time points (Fig. 3).
Most of the QTL mapping studies in rodent models are performed in a cross-sectional manner; however, current findings suggest that introduction of different environmental challenges and a consecutive sampling protocol enhances the power of detection of loci influencing variation of the studied traits (7, 27, 53). The same is seemingly true for different phases of ontogenesis (12). On the basis of these observations, the set of genes influencing any of the metabolic syndrome features in a particular situation is, to a certain extent, reflecting an “allostatic” (49) process involving both genetic and environmental factors. Indeed, it is not surprising to find that, e.g., distinct yet considerably overlapping sets of loci are linked to glucose tolerance (Table 2 and Supplemental Table S1) or body weight under changing environmental conditions (Fig. 2). We can speculate that the resulting shifting picture is a reflection of the changing importance of different genomic loci for determination of the trait in a given condition (age, dietary, or pharmacological challenge). Several types of QTLs emerge in this way: “constitutive” ones, found irrespectively of the environment, and “facultative” ones that appear only under specific conditions, thus possibly encompassing genes with subtler, yet sometimes crucial importance for the given trait. Of course, sufficient statistical power is required for the detailed dissection of such relations and especially the two- (3-, 4-) way interactions of epistatic loci, meaning at least hundreds of individuals, even in the experimental setting involving crosses of the inbred model strains.
Several of the QTLs reported here are novel, considering the synteny among available rat, mouse, and human studies, e.g., the uric acid locus within the RNO3 QTL cluster or the two separate loci on chromosome 7. One of the latter affects the variation of total body weight. There is an interesting gene within the QTL region coding for complement factor D/adipsin, a serine protease that is secreted by adipocytes into the bloodstream, that was shown to be deficient in several animal models of obesity. A body weight QTL was already reported on rat chromosome 7 in the Goto-Kakizaki rat (5); however, it was in a locus clearly distinct from that reported here. Near the marker D7Rat5, we identified a peak of linkage for the concentrations of FFA after HSD feeding. At the same region, others have found a linkage for serum cholesterol levels in a Wistar-Kyoto (WKY) × stroke-prone spontaneously hypertensive rat (SHRSP) cross (15), although exclusively in female rats.
Some of the QTLs found in the current study are in consent with or enhance previously reported linkages. We have observed two nonoverlapping QTL-containing regions on chromosome 8. Whereas the region between markers D8Rat36 and D8Rat75 showed linkage to several morphometric traits, the second one (D8Rat43–D8Got72) showed significant linkage to TG concentrations in dexamethasone-treated rats. The latter region has been linked to the various aspects of metabolic and hemodynamic derangements in humans and model organisms (18, 22, 41). The confidence interval for the TG QTL lies within the differential segment of PD/Cub origin in the previously established BN.PD-D8Rat39/D8Rat35 (BN-Lx/Cub, Rat Genome Database ID no. 728144) congenic strain. We have compared its metabolic profile to that of the BN/Cub progenitor strain under conditions of standard, high-sucrose (42), and high-fat-high-cholesterol (HFD) diets (unpublished data), showing a significant rise not only in TGs but also in FFA concentrations and aggravation of glucose intolerance in BN-Lx/Cub (42). The only exception was HFD, where there was no difference between the congenic pair, reflecting the susceptibility of BN/Cub to dietary cholesterol load, a phenomenon confirmed recently (45). The potency of this PD/Cub region to elevate TG and FFA levels was also shown by transferring the chromosome 8 segment in question onto the spontaneously hypertensive rat (SHR) genetic background, thus creating the SHR-Lx congenic strain (23). Compared with SHR, the sucrose-fed SHR-Lx congenic strain shows increased levels of TG (41) and FFA concentrations (n = 6/group; SHR FFA = 1.28 ± 0.04 mmol/l, SHR-Lx FFA = 1.58 ± 0.09 mmol/l; P = 0.008). Furthermore, the QTL’s confidence interval overlaps with the region carrying the Lx mutation, causing PLS of the rat, so the loci for metabolic and morphological disturbances are closely linked or even may be identical, as is the case in several syndromes described in humans, e.g., the Bardet-Biedl syndrome (29) or the Smith-Lemli-Opitz syndrome (9). Confirmation of the identified QTLs (especially those with only suggestive linkage), using, e.g., a congenic strain approach (23), is a necessary step in validation of the putative effect of the particular genomic region on trait variance.
The largest clusters of QTLs in this study were identified on chromosome 3. No linkage of insulin, uric acid, FFA, or adiposity was reported for these two genomic regions; however, their importance is further supported by the overlapping findings of linkage of other metabolic syndrome-related traits. So, the region between D3Rat37 and D3Rat257 has been linked to glucose levels after pancreatectomy (30), body weight (14), cardiac mass (7, 27), and heart rate (24), whereas loci within the more telomeric part of RNO3 were reported to influence variation of blood pressure, kidney mass (27), and urine albumin level (48). One of the possible candidate genes present in the region is a paired homeodomain transcription factor Pax6, a key regulator of pancreatic islet hormone gene transcription required for normal islet development (38). The second identified region covers the genomic location of a group of nuclear receptors, hepatocyte nuclear factor-4α (Hnf4α) being the most prominent in terms of insulin-related traits (4). The QTL for relative heart weight on chromosome 2 (LOD = 5.8) overlaps with two previous linkage results for blood pressure (6, 39); the growth hormone receptor gene represents an interesting putative candidate. Chromosome 16 bears a single QTL for insulin concentration at the 120th min of the OGTT in rats after the combined challenge of sucrose diet and dexamethasone. Both peak linkage and the bootstrap test point to the telomeric region of the chromosome (RNO 16p16). A closely linked marker (D16Mit5) was found to be a member of the epistatic QTL pair linked to the dexamethasone-induced change in insulinemia. This region contains at least two genes of possible relevance, coding the voltage-dependent L-type calcium channel-α1D subunit (37) and pancreatic polypeptide receptor Ppr1 (1). A QTL for plasma glucose concentration was reported near D16Mit2 in genetically obese Leprfa/Leprfa F2 WKY13M intercross rats (3) and the Otsuka Long-Evans Tokushima Fatty rat strain (50). This region was also previously shown to be linked to blood pressure regulation in SHR (20) and Dahl salt-sensitive rat (S) × Lewis rat (Lew) cross (6). Both blood pressure linkages were confirmed in congenic strains SHR.BN-D16Mit2/Mit5 (23) and S-Lew RNO16 (28). One of the two constitutive QTLs for body weight was mapped to chromosome 17. Interestingly, the QTL confidence interval encompasses the dopamine receptor-1a gene, which has been shown to be a determinant of heart weight (34) and of the difference between adult and newborn kidney weight (12) in HXB and BXH recombinant inbred strains (developed from a cross between hypertensive SHR/OlaIpcv and normotensive BN-Lx/Cub rat strains). These observations are the only linkage studies reporting weight-related QTL in this particular genomic region and the corresponding regions of murine chromosome 13 and human chromosome 6p24. Finally, the location of Nidd2nsy QTL on mouse chromosome 14 for glucose concentration at the 120th min of the OGTT (52) may actually overlap with our RNO15 QTL cluster.
In summary, we present a comprehensive linkage study of metabolic syndrome attributes in a new genetic model, the PD/Cub rat. Apart from identifying >70 genomic regions with significant or suggestive linkage to particular facets of the syndrome, we demonstrate the dynamic changes of its genetic component during ontogenesis and in response to dietary and pharmacological stimuli.
This work was supported by the following grants: Czech Science Foundation No. 301/04/0248, the Internal Grant Agency of the Ministry of Health of the Czech Republic No. NR/7888, Canadian Institutes of Health Research Nos. MT-14654 and GEI-53958, and the Grant Agency of the Academy of Sciences of the Czech Republic No. B5105401. O. Seda is the recipient of a TACTICS Fellowship.
↵1 The Supplemental Material for this article (Supplemental Figs. S1 and S2, Supplemental Tables S1 and S2) is available online at http://physiolgenomics.physiology.org/cgi/content/full/00230.2004/DC1.
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).
Address for reprint requests and other correspondence: P. Hamet, Centre de recherche, Centre hospitalier de l’Université de Montréal, Laboratoire de Médecine Moléculaire, 3850 rue St-Urbain, Montréal, Québec, Canada H2W 1T7 (E-mail:).
↵* P. Hamet and V. Kren contributed equally to this work.
- Copyright © 2005 the American Physiological Society