Genomic Determinants of Triglyceride and Cholesterol Distribution into Lipoprotein Fractions in the Rat

The plasma profile of major lipoprotein classes and its subdivision into particular fractions plays a crucial role in the pathogenesis of atherosclerosis and is a major predictor of coronary artery disease. Our aim was to identify genomic determinants of triglyceride and cholesterol distribution into lipoprotein fractions and lipoprotein particle sizes in the recombinant inbred rat set PXO, in which alleles of two rat models of the metabolic syndrome (SHR and PD inbred strains) segregate together with those from Brown Norway rat strain. Adult male rats of 15 PXO strains (n = 8–13/strain) and two progenitor strains SHR-Lx (n = 13) and BXH2/Cub (n = 18) were subjected to one-week of high-sucrose diet feeding. We performed association analyses of triglyceride (TG) and cholesterol (C) concentrations in 20 lipoprotein fractions and the size of major classes of lipoprotein particles utilizing 704 polymorphic microsatellite markers, the genome-wide significance was validated by 2,000 permutations per trait. Subsequent in silico focusing of the identified quantitative trait loci was completed using a map of over 20,000 single nucleotide polymorphisms. In most of the phenotypes we identified substantial gradient among the strains (e.g. VLDL-TG from 5.6 to 66.7 mg/dl). We have identified 14 loci (encompassing 1 to 65 genes) on rat chromosomes 3, 4, 7, 8, 11 and 12 showing suggestive or significant association to one or more of the studied traits. PXO strains carrying the SHR allele displayed significantly higher values of the linked traits except for LDL-TG and adiposity index. Cholesterol concentrations in large, medium and very small LDL particles were significantly associated to a haplotype block spanning part of a single gene, low density lipoprotein receptor-related protein 1B (Lrp1b). Using genome-wide association we have identified new genetic determinants of triglyceride and cholesterol distribution into lipoprotein fractions in the recombinant inbred panel of rat model strains.


Introduction
Utilization of total concentrations of cholesterol (C) and triacylglycerols (TG) together with cholesterol levels in high and low density lipoprotein particles (HDL and LDL, respectively) has been a mainstay of cardiovascular risk assessment [1].However, recent evidence somewhat undermined the prominent position of these relatively coarse measures [2][3][4].New data are emerging from studies that combine advanced phenotyping techniques allowing mean particle diameter measurement [5,6], assessment of C and TG concentration down to particular lipoprotein subfractions [7][8][9] or comprehensive lipidomic approaches [10] together with genetic, epigenetic [11] as well as genome-wide analyses.Several recent studies identified genetic polymorphisms contributing to variation in lipoprotein diameters in distinct geoethnic groups [6,12] or patterns of lipoprotein subfraction distribution [13][14][15][16].As the TG and C content across the lipoprotein spectrum is substantially influenced by both genetic factors and environmental effects including diet [17][18][19] and medication [20][21][22], mammalian models have been successfully used for reduction of complexity in the detailed analyses of their genetic determinants [23,24].Already in 1996, an association study of the major lipoprotein subfractions was performed in 30 recombinant inbred HXB BXH strains derived from spontaneously hypertensive (SHR) and congenic Brown Norway (BN-Lx) rat strains [25].The animals were fed a diet with 5% olive oil and 2% cholesterol for four weeks and the subsequent association analysis using 534 microsatellite markers identified loci on rat chromosomes 4, 19 and 20 linked to HDL2-C and phospholipids [25].One of the strains from the original HXB/BXH set, the BXH2 strain, served as a progenitor of the PXO recombinant inbred strain panel together with chromosome 8 congenic strain SHR.Lx [26,27].Diets with high content of sucrose are well established to induce dyslipidemia (and insulin resistance) as recently reviewed by Gibson et al. [28].We and others have established particular differences in the sensitivity to high-sucrose diet among the three strains contributing to genomic pool of both HXB/BXH and PXO recombinant inbred strains.While the spontaneously hypertensive rat SHR/OlaIpcv [29] and the polydactylous rat PD/Cub [30,31] are quite sensitive both to hyperlipidemic and diabetogenic effects of HSD, the Brown Norway (BN) strain is less responsive in this respect [32,33].Furthermore, all the PXO strains carry the chromosome 8 region of PD/Cub origin that was shown to be responsive to HSD-induced dyslipidemia on both SHR and BN genomic backgrounds [33,34].In the current study, we performed genome-wide association of lipoprotein particle size, detailed distribution of C and TG into 20 distinct lipoprotein fractions and morphometric measures in the whole PXO set followed by in silico prioritization of candidate variants using dense map of over 20,000 single nucleotide polymorphisms and available whole genome sequences of SHR and BN rat strains.

Materials and Methods
All experiments were performed in agreement with the Animal Protection Law of the Czech Republic (311/1997) which is in compliance with the European Community Council recommendations for the use of laboratory animals 86/609/ECC and were approved by the Ethical Committee of the First Faculty of Medicine of the Charles University.

Experimental protocol
Animals were held under temperature and humidity controlled conditions on 12 h/12 h light-dark cycle.At all times, the animals had free access to food and water.At the age of 4 months, male rats of 15 PXO strains (n = 8-13/strain) and the progenitor strains SHR-Lx (n = 13) and BXH2 (n = 18) were fed a high-sucrose diet (HSD, proteins (19.6 cal%), fat (10.4 cal%), carbohydrates (sucrose, 70 cal%) and balanced levels of micronutrients) for 1 week.Blood samples were drawn after overnight fasting from the tail vein.Using high performance liquid chromatography (HPLC), triglyceride (TG) and cholesterol (C) concentrations in 20 lipoprotein fractions and the size of major classes of lipoprotein particles were determined as described previously [9,22].Then the rats were sacrificed and the weights of heart, liver, kidneys, adrenals, epididymal and retroperitoneal fat pads were determined.The detailed results of all measurements are provided in Tables S1-S6.

Genomic characterization
The rat genomic DNA was isolated from the tail incision samples using a modified phenol extraction method.Polymorphic microsatellite loci were amplified by PCR using conditions optimized for each marker.The PCR products were separated on polyacrylamide (7-10%) or agarose (2-4%) gels, stained by ethidium bromide and visualized using Syngene's G-Box.We scanned over 1000 microsatellite markers for polymorphisms between the two progenitor strains.Subsequently, we selected markers polymorphic between the progenitor strains for genotyping all 15 PXO recombinant inbred strains.Total of 704 microsatellite markers were used for association analysis.Once the QTLs were identified, we determined in silico the extent of the haplotype blocks using the map of .20,000single nucleotide polymorphisms, which has been recently made available by STAR consortium (http://www.snp-star.eu/)[35].

Statistical analysis
All statistical analyses were performed using STATISTICA 10 CZ.Unpaired Student's t-test was used for comparison of metabolic and morphometric traits of the two progenitor strains.When comparing morphometric and biochemical variables in the whole PXO set, one-way ANOVA was used with STRAIN as major factor followed by post-hoc Tukey's honest significance difference test for comparison of the specific pairs of variables.Null hypothesis was rejected whenever p,0.05.

The morphometric profile of progenitors and the PXO set
The two progenitor strains showed significant differences in both morphometric and overall lipid profiles (Table 1).SHR-Lx rats were 69% heavier, displaying comparable amounts of visceral fat, however, more than twice as much of retroperitoneal fat per 100 g body weight compared to HXB2.Within the PXO recombinant inbred panel, body weight was distributed between the values of progenitors, while the fat deposit relative weights displayed a transgressive variation (Figure 1) as did other organ weights (Table S1).

The lipid profile of progenitors and the PXO set
Concentration of total C was slightly higher in SHR-Lx compared to BXH2, while total TG did not differ between the progenitors (Table 1).However, the detailed assessment of the C and TG distribution into 20 lipoprotein fractions revealed a rather unfavorable profile in SHR-Lx rats (Tables S5 and S6).Compared to BXH2, SHR-Lx had increased levels of TG and C in VLDL fraction and most significant cholesterol enrichment in the very small LDL particles (Figure 2).Except for very large and very small HDL, the HDL-C content was decreased in SHR-Lx.When assessing the lipid profile of the whole PXO panel, we observed a substantial gradient in C and TG concentrations within the major lipoprotein classes (Figure 3, Tables S2 and S3), particularly in VLDL (13-fold and 12-fold gradient across the PXO strains for C and TG, respectively).Combined with limited intra-strain variance of these measures this suggested actual segregation of alleles affecting the distribution of TG and C, making the studied traits amenable to genetic dissection in the PXO panel.

The genome-wide association
We scanned over 1000 microsatellite markers for polymorphisms between SHR-Lx and BXH2 and subsequently we genotyped 704 polymorphic markers in all PXO recombinant inbred strains, evenly covering autosomes and X-chromosome (except for chromosomes 2 and 13 that are of SHR origin in both parental strains).Blocks of adjacent markers with identical segregation patterns were collapsed into the final strain distribution pattern (SDP).The complete SDP was used for association analysis of morphometric and lipid-related traits.Using interval mapping, we have identified nine loci showing suggestive or significant association to 13 studied traits (3 morphometric and 10 lipid-related loci, summarized in Table 2).Except for association signals for adiposity index and LDL-TG on chromosomes 3 and 12, PXO strains carrying the SHR-Lx (i.e.SHR) allele displayed significantly higher values of the linked traits, e.g.LDL-C (21.260.4 vs. 12.560.4mg/dl in PXO strains with SHR allele vs. BXH2 (BN) allele in the D3Rat50-D3Got19 block (Table 2).

In silico analyses
In order to narrow down the genomic extent of the identified QTLs, we have retrieved SNP genotypes of progenitors and PXO strains for the identified associated regions using the SNPlotyper tool (http://snplotyper.mcw.edu/,Table 2) [35].In some cases, we were able to significantly reduce the extent of the particular haplotype block (e.g. for VLDL-C QTL on chromosome 4 from 8.7 Mb to just 1.1 Mb containing only 4 genes, i.e. by 86%), in other cases the information provided by microsatellites was almost identical to that by SNPs (e.g. the LDL-C QTL on chromosome 3 was shortened by 350 kb, i.e. by 8%).In the next step, we utilized the Variant Visualizer resource provided by Rat Genome Database (http://rgd.mcw.edu/rgdweb/front/select.html) to identify highly conserved variants between SHR and BN genome sequences [38] among the genes captured within the narrowed QTLs as potential candidates for further targeted studies.These genes represent primary candidates for the identified QTLs.At the same time, we have searched for the presence of copy number variants distinct in SHR an BN genomes within the QTLs [39].
The strongest association signal was detected for cholesterol content in total LDL and very small LDL particles on rat chromosome 3.Among 21 protein-coding genes in the associated segment, seven showed non-synonymous variants between the SHR and BN alleles: ring finger and CCCH-type domains 2 (Rc3h2), G protein-coupled receptor 21 (Gpr21), RAB GTPase activating protein 1 (Rabgap1, there was another variant within 59-untranslated region of this gene), crumbs homolog 2-Drosophila (Crb2), similar to cis-Golgi matrix protein GM130 (LOC690458), Golgin subfamily A member 1 (Golga1) and 40 s ribosomal protein S17-like (LOC100362666).Four intronic variations were found within spermatid perinuclear RNA binding protein (Strbp).
A distinct locus on chromosome 3 was linked to cholesterol concentration in large, medium and very small LDL particles.The whole genomic block spanning about 2 Mbp overlaps with a single gene, low density lipoprotein receptor-related protein 1B (Lrp1b).While no mutation putatively changing an amino-acid was found when comparing the sequences of the two progenitors, there were .20 intronic differences spread across the associated region as well as ten and five distinct variants within 59-and 39-untranslated regions (UTR) of this gene, respectively.Five genes within the third locus on chromosome 3 linked to cholesterol and triacylglycerol concentrations in large LDL particles and cholesterol in medium LDL have distinct non-synonymous variants in SHR: hypothetical protein LOC680314, tetratricopeptid repeat, ankyrin repeat and coiled-coil containing 1 (Tanc1, additional 3 differences were found in 59-UTR), bromodomain adjacent to zinc finger domain, 2B (Baz2b, within 59-UTR of this gene there were 9 other polymorphisms), lymphocyte antigen 75 (Ly75) and phospholipase A2 receptor 1 (Pla2r1).
The quantitative trait locus identified on chromosome 4 shows association to overall concentration of cholesterol in VLDL as well as in medium and large VLDL particles.Spanning just over 1 Mbp, it contains two genes with predicted aminoacid changes in SHR, the protein tyrosine phosphatase, receptor-type, Z polypeptide 1 (Ptprz1) and aminoadipate -semialdehyde synthase (Aass).More than 10 intronic variants were found in the Ca++-dependent secretion activator 2 (Cadps2) gene.
Chromosome 7 was a site of two association hits; the one associated with morphometric traits for liver and kidney weight encompasses 5 genes with differences between SHR and BN strains [frizzled family receptor 6 (Fzd6, 1 synonymous and 2 nonsynonymous variants in SHR), dendrocyte expressed seven transmembrane protein (Dcstamp, 1 nonsynonymous variant and 4 SNPs in introns), regulating synaptic membrane exocytosis 2 (Rims2, 9 SNPs in introns), non-synonymous variants were found also in dihydropyrimidinase (Dpys) and zinc finger protein, FOG family member 2 (Zfpm2)].This region actually overlaps with previously identified salt-loaded kidney mass in Dahl salt-sensitive rats [40].The genes containing non-synonymous variants distinguishing the SHR and BN alleles within the second locus on chromosome 7 linked to size of HDL cholesterol particles were transmembrane protein 71 (Tmem71), thymoglobulin (Tg) and WNT1 inducible signaling pathway protein 1 (Wisp1).

Discussion
This study represents one of the first scans of genomic determinants of detailed distribution of cholesterol and triacylglycerols into lipoprotein fractions in the rat.The PXO recombinant inbred set represents a genetically fixed system with segregating alleles of SHR, i.e. hypertensive model of metabolic syndrome; limited portion of chromosome 8 of polydactylous rat PD/Cub, a model of metabolic syndrome without hypertension [44] and normotensive and normolipidemic Brown Norway (BN/Cub) strain [27].The studies of PXO and the original HXB BXH recombinant inbred strain panels have repeatedly provided crucial insights into the genomic architecture of human complex cardiometabolic traits [26,45,46] and has served as a platform for the ''proof-of-concept'' type of studies, e.g. for genetical genomics [47].
Majority of significant associations in our study were found for lipid concentrations in subclasses of LDL lipoproteins, three of them mapped to distinct regions of rat chromosome 3.So far, two rat QTLs for total cholesterol levels were reported in the overlapping regions, one in a metabolic syndrome model -the Wistar Ottawa Karlsburg W (WOKW) rat [48] [49] and the second in an intercross of Dahl S x R rats transgenic for human cholesteryl ester transfer protein (CETP) [50].A single gene, orphan G protein coupled receptor Gpr21 stands out as possible candidate for the most significant observed association of cholesterol in total and very small LDL, although no direct connection to lipid metabolism is currently known.The Gpr21 gene is widely expressed [51] and has been shown to possess  Summary of quantitative trait loci identified in the PXO recombinant inbred rat panel.The letter ''f'' followed by a number (e.g.f5) corresponds to a particular fraction of lipoproteins as depicted in critical function in coordinating macrophage proinflammatory activity in the context of obesity-induced insulin resistance [52].
Gpr21-knockout mice were shown to have improved insulin sensitivity and lower triacylglycerol content in liver when fed highfat diet [52].Altogether, we have identified three significant associations for cholesterol concentration in very small LDL particles.It has been repeatedly shown that small, dense LDL are clearly connected with metabolic syndrome, type 2 diabetes and fatty liver [53] [54] While the unfavorable lipidemic profile is greatly influenced by environment, particularly lifestyle, possible identification of sensitizing genetic variants may eventually allow early identification of individuals at enhanced risk of coronary artery disease and related conditions.We have also compared the complete set of genes present in the regions with identified association to any of the followed traits with the repository of results coming from human genome-wide association studies (http://www.genome.gov/gwastudies/) [37].We did not identify any directly corresponding associations (see Methods) between human GWAS and the putative variants in SHR genome.This is not surprising given the rather subtle nature in risk increase conferred by most SNPs identified in this type of human studies.On the other hand, variants in several genes including LRP1B and DGKG, associated with cholesterol concentrations in very small LDL particles in our study, were associated with body mass index and weight human studies [55,56].LRP1B belongs to the LDL receptor family and its mutations are often found in human cancers [57].Furthermore, polymorphisms in the LRP1B gene were found to be associated with insulin resistance [58] and preeclampsia [59] in human studies.Also, Lrp1b was identified as significant member of genome-wide signatures of abdominal fatness in chicken [60].The importance of identification of very narrow QTL encompassing only part of single gene (Lrp1b) is enhanced by in silico data suggesting presence of five non-synonymous variants distinct in SHR vs. BN within the QTL.Since the expression of LRP1B was found to be relatively low compared to the homologous LRP1 (and due to its localization mainly in brain, skeletal muscle and thyroid gland) it was suggested that LRP1b does not play a major role in the clearing of circulating lipoproteins from plasma, on the other hand, its capability of binding apoE-containing ligands has been demonstrated [61].Detailed study of this large gene is needed to validate causality of the ascertained variants in SHR.
The third region on rat chromosome 3 identified in the current study affecting cholesterol and triacylglycerol concentrations in LDL harbors three genes with non-synonymous mutations in SHR with indirect supportive evidence from other studies.The variants in BAZB2 gene have been shown as prime candidates for sudden cardiac death in GWAS meta-analysis performed in individuals with European ancestry [62] while LY75 was identified as a prime expression SNP for type 2 diabetes [63] and significant changes in LY75 expression due to aberrant methylation was found in human dilated cardiomyopathy [64].The Pla2r1 functions as a receptor of the enzyme secretory phospholipase A2, which in turn catalyzes the hydrolysis of phospholipids.The PLA2R1 gene has been proposed as a candidate for coronary artery disease in subjects with normal levels of total cholesterol [65] and very recently the Pla2r-deficient mice were shown to be susceptible to cardiac rupture after myocardial infarction [66].
While two regions associated in our study to levels of cholesterol in VLDL and very small LDL (chromosomes 4 and 11, respectively) were not yet linked to cholesterol in other rat studies, there are several reports showing their association to triacylglycerol concentrations in WOKW [67], SHR [68] [69] and Dahl salt-sensitive rat models [40].One of the limitations of the current study is the incomplete coverage of all autosomes, as chromosomes 2 and 13 are of SHR origin in both parental strains.Therefore, possibly existing polymorphisms with influence on lipid traits residing at the two chromosomes could not be identified by our association analysis.To that end, we have previously derived dedicated chromosome 2 congenic strain and reported the effect of SHR alleles on TG and C distribution into lipoprotein fractions [70].Another limitation is represented by the fact that all measurements were performed under high-sucrose diet-feeding and therefore the identified associations may be related only to this particular nutritional condition (i.e. the ''nutrigenetic'' layer of the dynamic genetic architecture [30] of lipid profile).In summary, we have combined detailed phenotyping of a unique rat strain panel together with classical and in silico approaches to identify novel candidates for genetic determinants of cholesterol and triacylglycerol distribution into lipoprotein subfractions.The process of identification of candidate genes and positional cloning of causal variants is substantially streamlined thanks to the availability of comprehensive genotype data and full genomic sequences of the spontaneously hypertensive and Brown Norway rat strains.

Figure 1 .
Figure 1.SNP (single nucleotide polymorphism) focusing column denotes the position of the QTL on the chromosome (Chr.)according to the Rat Genome V3.4 Assembly.% -percentage of trait variance explained by QTL, C -cholesterol, TGtriacylglycerols; LOD -likelihood of odds score (bold for genome-wide significance, italics for suggestive association); VLDL -very low-density lipoprotein; LDL -low density lipoprotein; HDL -high-density lipoprotein; wt -weight; adiposity index -calculated as epididymal (visceral) fat weight per 100 g of body weight; *…region contains known CNV distinct between SHR and BN-Lx [39].The total number of genes within the QTL region is shown.Mean values 6 SEM are shown for PXO strains carrying the SHR-Lx (SHR/SHR) allele vs. those carrying the BXH2/Cub allele (BN/BN) for the given trait and locus.doi:10.1371/journal.pone.0109983.t002

Table 1 .
Phenotypic comparison of BXH2 and SHR-Lx progenitor strains.Basic phenotypic profile of BXH2 vs. SHR-Lx rats after one week of high-sucrose diet feeding with p-values (significant p values in bold, non-significant in italics) for unpaired Student's t-test.Values are shown as mean 6 S.E.M.; b.wt.…body weight; EFP…epididymal fat pad; RFP…retroperitoneal fat pad; VLDL -very low-density lipoprotein, LDL -low density lipoprotein, HDL -high-density lipoprotein. doi:10.1371/journal.pone.0109983.t001

Table 2 .
Suggestive and significant associations in the PXO recombinant inbred set.