Mapping and Congenic Dissection of Genetic Loci Contributing to Hyperglycemia and Dyslipidemia in Mice

Background Patients with dyslipidemia have an increased risk of developing type 2 diabetes, and diabetic patients often have dyslipidemia. Potential genetic connections of fasting plasma glucose with plasma lipid profile were evaluated using hyperlipidemic mice. Methods 225 male F2 mice were generated from BALB/cJ (BALB) and SM/J(SM) Apoe-deficient (Apoe−/−) mice and fed a Western diet for 5 weeks. Fasting plasma glucose and lipid levels of F2 mice were measured before and after 5 weeks of Western diet and quantitative trait locus (QTL) analysis was performed using data collected from these two time points. 144 SNP(single nucleotide polymorphism) markers across the entire genome were typed. Results One major QTL (logarithm of odds ratio (LOD): 6.46) peaked at 12.7 cM on chromosome 9,Bglu16, and 3 suggestive QTLs on chromosomes 15, 18 and X were identified for fasting glucose, and over 10 loci identified for lipid traits. Bglu16 was adjacent to a major QTL, Hdlq17, for high-density lipoprotein (HDL) cholesterol (LOD: 6.31, peak: 19.1 cM). A congenic strain with a donor chromosomal region harboring Bglu16 and Hdlq17 on the Apoe−/− background showed elevations in plasma glucose and HDL levels. Fasting glucose levels were significantly correlated with non-HDL cholesterol and triglyceride levels, especially on the Western diet, but only marginally correlated with HDL levels in F2 mice. Conclusions We have demonstrated a correlative relationship between fasting glucose and plasma lipids in a segregating F2 population under hyperlipidemic conditions, and this correlation is partially due to genetic linkage between the two disorders.


Results
One major QTL (logarithm of odds ratio (LOD): 6.46) peaked at 12.7 cM on chromosome 9, Bglu16, and 3 suggestive QTLs on chromosomes 15, 18 and X were identified for fasting glucose, and over 10 loci identified for lipid traits. Bglu16 was adjacent to a major QTL, Hdlq17, for high-density lipoprotein (HDL) cholesterol (LOD: 6.31, peak: 19.1 cM). A congenic strain with a donor chromosomal region harboring Bglu16 and Hdlq17 on the Apoe −/− background showed elevations in plasma glucose and HDL levels. Fasting glucose levels were significantly correlated with non-HDL cholesterol and triglyceride levels, especially on the Western diet, but only marginally correlated with HDL levels in F 2 mice.

Conclusions
We have demonstrated a correlative relationship between fasting glucose and plasma lipids in a segregating F 2 population under hyperlipidemic conditions, and this correlation is partially due to genetic linkage between the two disorders.

Introduction
Dyslipidemia, characterized by elevations in plasma triglyceride and LDL cholesterol levels and reductions in HDL cholesterol levels, frequently occurs with hyperglycemia as part of the metabolic syndrome, which also includes abdominal obesity, insulin resistance, and hypertension [1]. Although the nature for the close association between dyslipidemia and hyperglycemia is not well understood, pleiotropic effects of genetic mutants or variants affecting both traits appear to play a role. Indeed, a few rare genetic mutations involving ABCA1 [2], LIPE [3], LPL [4], or LRP6 [5]cause both dyslipidemia and hyperglycemia. Genome-wide association studies (GWAS) have identified a number of common variants associated with variations in plasma lipids [6] [7] and fasting plasma glucose [8][9] [10]. Over a dozen of them are associated with both traits at the genome-wide significance level (http://www.genome.gov/GWAStudies/). Unexpectedly, half of them, including CETP, MLXIPL, PLTP, GCKR, APOB, APOE-C1-C2, CYP7A1, and TIMD4, have exhibited opposite allelic effect on plasma lipid and glucose levels [11], a finding that is incontrary to the positive correlations observed in the clinical situation. Furthermore, it is quite challenging to establish causality between a common variant and a complex trait in humans due to small gene effect, complex genetic structure, and environmental influences.
One approach to the problems encountered in human genetic studies is to use inbred strains of mice differing in glucose and lipid profile. Apolipoprotein E-deficient (Apoe −/− ) mice develop spontaneous dyslipidemia on a low fat chow diet, with elevated non-HDL cholesterol levels and reduced HDL levels [12] [13]. Feeding a high fat diet aggravates dyslipidemia. Moreover, these mice develop all phases of atherosclerotic lesions seen in humans [14][15] [16][17] [18]. We have found that Apoe −/− mice with the C57BL/6J, C3H/HeH, SM/J (SM) or SWR/J genetic background develop significant hyperglycemia when fed a Western diet but become resistant when transferred on to the BALB/cJ (BALB) background [19][20] [16]. In the present study, we performed quantitative trait locus (QTL) analysis using a male cohort derived from BALB-Apoe −/− and SM-Apoe −/− mice to find potential genetic connections between plasma glucose and lipid traits.

Methods
All procedures were carried out in accordance with current National Institutes of Health guidelines and approved by the University of Virginia Animal Care and Use Committee (Assurance #A3245-01, Animal Protocol #3109).

Mice
BALB and SM Apoe -/mice were created using the classic congenic breeding strategy, as described [16]. BALB-Apoe -/mice were crossed with SM-Apoe -/mice to generate F 1 s, which were intercrossed by brother-sister mating to generate a cohort of F 2 mice. Mice were weaned at 3 weeks of age onto a rodent chow diet. At 8 weeks of age, male F 2 mice were started on a Western diet containing 21% fat, 34.1% sucrose, 0.15% cholesterol, and 19.5% casein by weight (Harlan Laboratories, TD 88137) and maintained on the diet for 5 weeks.

Measurements of plasma glucose and lipid levels
Mice were bled twice: once before the start of the Western diet and once at the time of euthanasia. Mice were fasted overnight before blood was drawn from the retro-orbital venous plexuswith the animals under isoflurane anesthesia. Plasma glucose was measured with a Sigma glucose (HK) assay kit, as reported [21]. Total cholesterol, HDL cholesterol, and triglyceride were measured using Thermo DMA (Louisville, CO) assay kits [13]. Non-HDL cholesterol was calculated as the difference between total and HDL cholesterol.

Genotyping
Genomic DNA was isolated from the tails of mice by using the phenol/chloroform extraction and ethanol precipitation method. F 2 mice were genotyped by the Jackson Laboratory Genotyping Services using mouse strain-specific SNP arrays. DNA samples from the two parental strains and their F 1 s served as controls. 144 SNPs and 225 F 2 mice were included for QTL analysis.

Studies with congenic mice
Construction of a congenic strain in which a chromosome 9 segment from 5 to 61 cM was transferred from C3H/HeJ-Apoe -/mice onto the C57BL/6J-Apoe -/background was previously reported [22]. Male congenic and C57BL/6J-Apoe -/control mice were started with the Western diet at 6 weeks of age and maintained on the diet for 12 weeks. Blood samples were collected from overnight fasted mice before and after 12 weeks of Western diet.

Statistical analysis
QTL analysis was performed using the standard analysis softwared J/qtl and Map Manager QTX as we previously reported [19][23] [24]. One thousand permutations of trait values were run to define the genome-wide LOD (logarithm of odds) score threshold for significant or suggestive linkage of each trait. Loci that exceeded the 95th percentile of the permutation distribution were considered significant (P<0.05) and those exceeding the 37th percentile were suggestive (P<0.63). Student's unpaired t test was used to determine statistical significance between congenic and control mice in trait values.

Trait value distributions
Values of fasting plasma glucose and triglyceride levels of F 2 mice on both chow and Western diets and of HDL and non-HDL cholesterol levels on the chow diet are normally or approximately normally distributed (Fig 1).Values of Ln (natural logarithm)-transformed HDL and non-HDL cholesterol levels on the Western diet approach the normal distribution. These data were analyzed using J/qtl software to search for QTLs affecting the traits. Loci with a genomewide suggestive or significant P value are presented in Table 1.

Fasting glucose levels
A genome-wide scan for main effect QTL revealed a highly significant QTL in the proximal region of Chr9 for fasting glucose levels when mice were fed the chow diet (12.7cM, LOD:6.463) (Fig 2 and Table 1) (original genotype and phenotype data used for QTL analysis are provided in Table A in S1 text). This locus is overlapping in position with Bglu16, recently mapped in female F 2 mice derived from BALB and SM Apoe -/mice. Two suggestive loci, located on Chr15 and Chr18, for fasting glucose were also detected when the cross were on the chow diet. The Chr15 locus replicates Bglu8 and the Chr18 locus replicates Bglu10, initially mapped in a NZB/B1NJ x NZW/LacJ intercross [25]. When F 2 mice were fed the Western diet, a suggestive locus for fasting glucose was detected in the distal region of ChrX (68.38 cM, LOD: 3.069). This locus was novel. Inheritance of BALB alleles conferred an increased glucose level for the Chr9 and Chr15 QTLs while inheritance of SM alleles conferred increased glucose levels for Chr18 and ChrX QTLs ( Table 2).
For triglyceride, 3 suggestive QTLs, located on Chr1 and ChrX, were detected when mice were fed the Western diet (Fig 6). The QTL on ChrX had a LOD score of 3.25 and peaked at 66.4 cM. This QTL was replicated on the chow diet and thus named Tgq35. The 2 suggestive QTLs on Chr1, peaked at 41.8 and 80.1 cM, replicated Tgq9 and Tglq1, respectively [34].

Correlations between plasma glucose and lipid levels
Correlations of fasting plasma glucose levels with plasma levels of HDL, non-HDL cholesterol or triglyceride were evaluated in the F 2 population fed either chow or Western diet (Fig 7). Significant correlations of fasting glucose with non-HDL cholesterol and triglyceride were observed when mice were fed either chow (R 2 = 0.1498 and P = 5.3E-4 for non-HDL; R 2 = 0.1193 and P = 6.56E-7 for triglyceride) or Western diet (R 2 = 0.3899 and P = 4.48E-25 for non-HDL; R 2 = 5782 and P = 2.61E-43 for triglyceride). F 2 mice with higher non-HDL cholesterol or triglyceride levels also had higher fasting glucose levels, especially on the Western diet. In contrast, HDL cholesterol levels were only marginally correlated with fasting glucose levels on either chow (R 2 = 0.0724 and P = 6.3E-6) or Western diet (R 2 = 0.0199 and P = 0.035).

Discussion
BALB Apoe -/mice have much higher HDL and lower non-HDL cholesterol levels and are more resistant to development of type 2 diabetes compared to SMApoe -/mice [16] [20]. In this study, we performed QTL analysis using a male F 2 cohort derived from the two Apoe -/mouse strains to investigate genetic connections between glucose and lipid-related traits. One major QTL for fasting glucose, Bglu16, is immediately adjacent but not coincident with a major QTL for HDL cholesterol, Hdlq17, on proximal chromosome 9. The presence of these two QTLs was confirmed with a congenic strain. Moreover, strong correlations of fasting glucose with non-HDL and triglyceride levels were observed in F 2 mice when fed the Western diet.
In this study, one significant QTL and three suggestive QTLs have been identified to influence glucose homeostasis under fasting conditions. The significant QTL on proximal chromosome 9 is coincident with Bglu16, recently mapped in a female intercross between BALB and SM Apoe -/mice. A suggestive QTL for fasting glucose has also been mapped to this position in a C57BL/6 x BALB Apoe -/intercross [21]. For all three intercrosses, inheritance of BALB alleles at the locus contributed to increased fasting glucose levels. A locus for glucose-stimulated insulin secretion has been mapped to this position in a C57BL/6J x C3H/HeJ intercross [36].The suggestive QTL on chromosome 15 is close to Bglu8, mapped in a NZB x NZW intercross [25]. Linkages close to this locus have been detected in two other crosses involving BALB mice but inheritance of BALC alleles was associated with reduced fasting glucose levels [21] [37]. The QTL on chromosome 18 is coincident with Bglu10, mapped in a NZB x NZW intercross [25]. All QTLs identified for plasma lipids confirm those mapped in previous studies except for the one on X chromosome for triglyceride that is new and named Tgq30. This QTL occurred with a suggestive LOD score under both chow and Western diet feeding conditions. It is considered appropriate to name a suggestive QTL if it has been repeatedly observed [38]. The major QTL for HDL is coincident with Hdlq17, mapped in a C57BL/6 x 129S1/SvImJ female intercross [26]. This QTL has been replicated in multiple crosses, including B6 x 129, B6 x CAST/EiJ, B6-Apoe -/x C3H-Apoe -/-, and B6-Apoe -/x BALB-Apoe -/intercrosses [33][39] [40] [41] [30] [31]. We conducted haplotype analysis for this QTL and narrowed candidates down to two dozen genes (Table C in S1 Text). These candidate genes contain one or more non-synonymous SNPs in the coding regions or SNPs in the upstream regulatory region that are shared by the high allele strains but are different from the low allele strains at the QTL. Among them, Ubash3b, Phldb1, Sorl1, Sik3, and Apoa1 have been shown to be associated with variations in total, HDL cholesterol or triglyceride levels in humans (http://www.ebi.ac.uk/gwas/home). Linkage close to this locus has also been detected in a female intercross derived from BALB and SM Apoe -/mice but BALB alleles were associated with to reduced HDL levels [42]. The opposite allelic effect on HDL in the male vs. female crosses suggests that two or more genes in this region contributed to the trait. As BALB and C3H/HeJ strains share essentially identical haplotype blocks for the chromosomal region harboring Bglu16 and Hdlq17 and as QTLs for fasting glucose and HDL have been mapped to this region in crosses derived from C3H/HeJ mice [19][35], we used a congenic strain carrying the C3H/HeJ chromosome 9 donor alleles to confirm the presence of the two QTLs. However, as the congenic strain carries a chromosomal segment much longer than the confidence interval of Bglu16 and Hdlq17, other QTLs in the congenic region might also contribute to the QTL effects observed in the congenic mice.
We have observed positive correlations of fasting glucose levels with non-HDL cholesterol and triglyceride levels in the F2 population under either feeding condition. The correlations were extremely high when mice developed significant dyslipidemia on the Western diet. Similar findings have been observed in other crosses derived from Apoe -/mouse strains [21] and humans [11] [43]. Emerging human studies have also shown associations of non-HDL cholesterol and ApoB with incident type 2 diabetes [44], [45], [46]. Despite the strong correlations, only one suggestive QTL for fasting glucose was coincident with one QTL for triglyceride on chromosome X and there was no coincident QTL between glucose and non-HDL traits. We observed a slight but positive correlation between HDL cholesterol and fasting glucose levels in male F 2 mice on either chow or Western diet. This finding is in contrast with the negative correlation between the two traits in a female F 2 cohort derived from the same parental strains [42]. Prospective human studies have shown that HDL is inversely correlated with the risk of type 2 diabetes [47] [48]. HDL can increase insulin secretion from β-cells, improve insulin sensitivity of the target tissues, and accelerate glucose uptake by muscle due to its diverse functions, including cholesterol efflux and reverse cholesterol transport, anti-oxidation, antiinflammation and activation of the AMP-activated protein kinase [49] [50]. The positive correlation between HDL and fasting glucose observed in this study may suggest that HDL has lost its anti-diabetic function. Indeed, under pathological conditions such as the acute phase response and chronic inflammatory diseases, HDL undergoes qualitative changes in both components and structure and can lose protective function [51]. The observed correlation between HDL and fasting glucose could also be derived from the genetic effect of two closely linked QTLs with each affecting one trait, like Bglu16 and Hdlq17. The reasons for the discrepancy between male and female F 2 mice in the correlations are unknown. Multiple factors could contribute: First, female mice were fed the western diet for 12 weeks starting at 6 weeks of age while males were fed the diet for 5 weeks starting at 8 weeks of age. Second, male F 2 s had higher glucose levels (chow: 110 vs 99, Western: 191 vs 147 mg/dl) than their female counterparts, suggesting that males are more susceptible to diet-induced type 2 diabetes. Finally, sex differences in metabolic traits have been observed in humans and mice [52] [53].
Dyslipidemia and hyperglycemia are integral components of metabolic syndrome, a group of risk factors that increase risk for cardiovascular disease and type 2 diabetes. We have identified multiple loci contributing to dyslipidemia and hyperglycemia from a male F 2 cohort. One major QTL for fasting glucose, Bglu16, is adjacent to Hdlq17, a QTL for HDL on chromosome 9. The strong correlations of fasting glucose with non-HDL cholesterol and triglyceride support the hypothesis that dyslipidemia plays a causative role in the development of type 2 diabetes [54]. Supporting Information S1 Text. Supporting tables: genotypic and phenotypic data used for quantitative trait locus (QTL) analysis, characterization of congenic strains, and haplotype analysis. (XLSX)