Haplotype Analysis in Multiple Crosses to Identify a QTL Gene

Most common human diseases, such as atherosclerosis, diabetes, and obesity, are complex traits determined by many genetic and environmental factors. The genetic factors are usually studied in animal models, most commonly mice, and frequently through a process known as quantitative trait locus (QTL) analysis, which has the advantage of finding novel key genes in a metabolic pathway. To date, more than 1800 mouse QTLs have been found (Mouse Genome Informatics, http://www.informatics.jax.org); however, identifying the genes underlying these QTLs has been an extremely challenging task (Nadeau and Frankel 2000; Korstanje and Paigen 2002).

The level of plasma high-density lipoprotein cholesterol (HDL), although not a disease, is also a complex trait. It has been intensely studied because, in humans, it is inversely correlated with the risks of coronary artery disease, and therapies that raise HDL levels may significantly reduce these risks (Boden and Pearson 2000). QTL analysis has identified many genomic regions that regulate HDL levels, both in mice and in humans—to date, 37 mouse and 29 human HDL QTLs have been identified (update of review by Wang and Paigen 2002). One of the mouse HDL QTLs, Hdlq5 (Wang et al. 2003), on distal chromosome 1 (cM 92), has been repeatedly identified in nine of those crosses (update of review by Wang and Paigen 2002). An obvious candidate for the Hdlq5 gene was Apoa2 (cM 92.6), because its encoded protein, apolipoprotein A-II (APOA2), helps maintain plasma HDL levels in mice (for review, see Blanco-Vaca et al. 2001).

To determine whether Apoa2 was the Hdlq5 gene, we took the approach of haplotype analysis. Recent single nucleotide polymorphism (SNP) maps indicate that the genome of common inbred mouse strains is defined by 1-2 Mb haplotype blocks (Wade et al. 2002; Wiltshire et al. 2003), which can be used to narrow a QTL, because its underlying gene should be in subregions where the parental strains have different haplotypes (Park et al. 2003; Manenti et al. 2004). We extended this haplotype analysis further by using it to identify a QTL gene. In doing so, we performed a sequence analysis of the coding region of Apoa2 in 46 genetically diverse mouse strains, and a haplotype analysis of strains both in the nine crosses that detected Hdlq5 and in the 12 crosses that failed to detect Hdlq5. We found that not only did the haplotype analysis support Apoa2 as the Hdlq5 gene, but it also suggested that the Apoa2 mutation is responsible for the Hdlq5 phenotype. Thus, haplotype analysis in multiple crosses can be used to support a QTL gene.

RESULTS

Inbred Mouse Strains Have Five APOA2 Protein Types

Polymorphisms of mouse Apoa2 coding sequence have been reported before; a total of 15 SNPs cause eight amino acid changes in APOA2 (Doolittle et al. 1990; Higuchi et al. 1991; Purcell-Huynh et al. 1995; Suto et al. 1999; Kitagawa et al. 2003). For our purpose of testing Apoa2 as the candidate gene for Hdlq5, we sequenced all four Apoa2 exons in 46 genetically diverse and widely used inbred mouse strains, including 43 whose plasma HDL concentrations were known (Mouse Phenome Database, http://www.jax.org/phenome) and three that were parents in crosses to map HDL QTLs (CASA, MRL, and NZO). We found that the four exons of Apoa2 had a total of 16 SNPs, resulting in nine amino acid changes and producing five APOA2 protein variants, which we named type “a” (APOA2^a) to “e” (APOA2^e; see Table 1 and Supplemental Fig. 1). The “b” strains were distinct with regard to two features: Their APOA2 sequence had an Ala⁶¹-to-Val⁶¹ substitution, and none of them were wild-derived strains (all were inbred laboratory mice). The other APOA2 types each had at least one wild-derived strain: “a” type contains M. m. castaneus, M. m. musculus, and M. m. molossinus (a mixture of the previous two); “c” type contains M. m. musculus; “d” type contains M. m. domesticus; and “e” type contains M. spretus.

Ala⁶¹ Is Conserved in Eight Mammalian Species

By comparing the APOA2 peptide sequences of eight mammal species, we found that APOA2 Ala⁶¹ was conserved in mice (except those having APOA2^b), humans, chimpanzees, monkeys, horses, cattle, pigs, and rats (Fig. 1), suggesting that Val⁶¹ is a mutation in mice with APOA2^b.

Hdlq5 Was Only Detected in Crosses Whose Parental Strains Differed at APOA2 Amino Acid 61

Of the 21 different mouse crosses used to identify HDL QTLs, Hdlq5 was found only in the nine where one parent carried APOA2^b and the second parent carried other APOA2 types (Table 2). Because mice with APOA2^b have a unique Val⁶¹, our results clearly showed that Hdlq5 was found only when the two parental strains differed at amino acid 61, that is, one should carry Ala⁶¹ and the other should carry Val⁶¹. Hdlq5 may not be detected even though the two parental strains have different APOA2 types; for example, Hdlq5 was not found when C57BL/6 (having APOA2^a) and SPRET (having APOA2^e) were crossed (Table 2).

Haplotype analysis of the strains carrying APOA2^b excluded the possibility that an unknown gene in linkage disequilibrium with Apoa2 was underlying Hdlq5 in these nine crosses: The APOA2^b strains do not share a common haplotype block near Apoa2 (Table 3). The nearby SNPs reduced the region the strains shared to one containing only two genes, Apoa2 and Fcer1g (Fc receptor, IgE, high affinity I, γ polypeptide). There is no evidence that Fcer1g is involved in lipoprotein metabolism.

Mice with APOA2^b Had Higher Plasma HDL Concentrations Than Did Those With Either APOA2^a or APOA2^c

To determine whether any of the APOA2 variants were associated with variant plasma HDL levels, we analyzed the plasma HDL concentrations of the 43 inbred mouse strains from the Mouse Phenome Database. In male mice, HDL levels of strains having APOA2^b (103 ± 6 mg/dL, mean ± SEM, 12 strains) were significantly higher than those of strains having either APOA2^a (67 ± 3 mg/dL, n = 24) or APOA2^c (62 ± 4 mg/dL, n = 4) (P = 9 × 10^-7 and P = 0.002, respectively; Fig. 2A). Similar results were found in female mice: HDL levels of strains having APOA2^b (81 ± 8 mg/dL, n = 12) were significantly higher than those of strains having either APOA2^a (52 ± 2 mg/dL, n = 24) or APOA2^c (46 ± 2 mg/dL, n = 4) (P = 4 × 10^-5 and P = 0.02, respectively; Fig. 2B).

The distinguishing feature of the APOA2^b is the presence of the amino acid substitution of alanine to valine at position 61. HDL levels of strains having Val⁶¹ were significantly higher than those having Ala⁶¹ in both male and female mice (103 ± 6 vs. 69 ± 3 mg/dL, P = 3 × 10^-6, and 81 ± 8 vs. 53 ± 3 mg/dL, P = 6 × 10^-5, respectively; Fig. 2C,D).

DISCUSSION

Identifying the gene underlying a QTL is a challenging task. Whereas considerable success has been achieved in identifying genes responsible for Mendelian traits—more than 1400 genes for them have been found (Page et al. 2003), fewer than 50 have been identified for polygenic or quantitative traits (Glazier et al. 2002; Korstanje and Paigen 2002).

Although there is no “gold standard” for positively identifying a QTL gene, the Complex Trait Consortium recently suggested that an identified gene should meet more than one of the following eight criteria (Biola et al. 2003): (1) polymorphisms in either its coding or regulatory regions have been found; its function has been (2) linked to the quantitative trait being analyzed, (3) tested in vitro, (4) tested in transgenic animals, (5) tested in knock-in animals, (6) assessed in deficiency-complementation tests, or (7) tested by mutational analysis; and (8) it has an homologous QTL for the same phenotype in another species. The most conclusive evidence comes from knock-in studies by replacing one allele with another and testing for function. We herein report using haplotype analysis in multiple crosses as a way of testing polymorphisms (criterion 1) to provide genetic evidence that a gene underlies a QTL.

By using this genetic approach, we provided two strong lines of evidence that Apoa2 underlies Hdlq5. First, haplotypes of Apoa2 were associated with plasma HDL concentrations: Mice with APOA2^b had significantly higher plasma HDL levels than those with either APOA2^a or APOA2^c. Second, among the 21 crosses used to map HDL QTL, Hdlq5 was only detected in the nine where the parental strains had different amino acid at position 61 in APOA2 protein. We excluded the possibility that a gene in linkage disequilibrium with Apoa2 was underlying Hdlq5 because Fcer1g, the only gene that was in linkage disequilibrium with Apoa2, has no known functions in lipoprotein metabolism.

Haplotype analysis enabled us not only to identify Apoa2 as the gene for Hdlq5 but also to pinpoint the Ala⁶¹-to-Val⁶¹ substitution as the possible causal change for this QTL. First, mice with Val⁶¹ had significantly higher HDL levels than did those with Ala⁶¹. Second, Hdlq5 was only found in the crosses in which one parental strain had Val⁶¹ and the other had Ala⁶¹ in APOA2. Third, in the nine crosses that detected Hdlq5, the strain with Val⁶¹ always had the high allele of this QTL. This method may not be limited to finding a QTL gene with amino acid differences among the parental strains; it may also be used to find the key SNPs in the regulatory elements of a candidate gene.

Although Apoa2 is mostly likely the gene underlying Hdlq5, and Ala⁶¹-to-Val⁶¹ substitution in APOA2 is most probably the causal polymorphism, other possibilities exist. First, polymorphisms in the promoter region of Apoa2 may lead to mRNA difference and thereby difference in plasma HDL levels. The following results, however, suggest that there is no correlation between Apoa2 mRNA and plasma HDL levels and that Apoa2 promoter polymorphisms are unlikely to affect plasma HDL levels. In the B6xNZB cross in which Hdlq5 was found, B6 mice (having APOA2^a) have higher levels of Apoa2 mRNA but lower levels of plasma HDL levels, compared with NZB mice (having APOA2^b) (P < 0.05) (Wang et al. 2003). In the B6xC3H cross in which Hdlq5 was found, B6 mice had lower levels of both Apoa2 mRNA and HDL levels, compared with C3H mice (having APOA2^b; P < 0.05; X. Wang and B. Paigen, unpubl.). On the other hand, B6 mice have similar levels of Apoa2 mRNA but lower levels of plasma HDL levels, compared with BALB/c mice (Doolittle et al. 1990). Second, a polymorphic enhancer in the same haplotype as Apoa2 coding sequence may regulate another gene, and this gene could regulate HDL levels. Although unlikely, we cannot exclude this possibility.

It has been reported that compared with C57BL/6 mice (having APOA2^a), BALB/c mice (having APOA2^b) have similar liver Apoa2 mRNA levels, but higher APOA2 protein synthesis rate in hepatocytes (Doolittle et al. 1990). This suggests that a change in messenger sequence, presumably the nucleotide change that causes Ala⁶¹-to-Val⁶¹ substitution, increases the efficiency of Apoa2 messenger translation, leading to more rapid production of APOA2 protein and therefore larger HDL particle and higher plasma HDL concentrations. We found no sequence changes in the 3′- or 5′-UTR regions between C57BL/6 and BALB/c mice (Supplemental Fig. 1).

Our findings are important for future HDL research for two reasons. First, to avoid Hdlq5, whose strength may mask other HDL QTLs (and hence their discovery), future HDL QTL mapping efforts in mice should cross parental strains with the same amino acid 61 in APOA2. Second, a human QTL for plasma HDL levels was identified in the APOA2 region (Elbein and Hasstedt 2002). To determine whether APOA2 regulates human HDL concentrations, APOA2 SNPs should be analyzed in human populations.

Obviously, the analysis we conducted will be more powerful when the same QTL is detected in multiple crosses. Now that about 1500 QTLs, many of which were detected in multiple mouse crosses, have been mapped in the mouse genome (http://informatics.jax.org) and a wide variety of phenotypic data from many commonly used and genetically diverse mouse strains are easily accessible in the Mouse Phenome Database (hppt://www.jax.org/phenome), many QTL genes can be discovered quickly by analyzing their SNPs and haplotypes in multiple crosses, as exemplified in this study.

METHODS

Obtaining and Genotyping Genomic DNA

We obtained genomic DNA from Mouse DNA Resource at the Jackson Laboratory (http://www.jax.org/dnares/index.html). PCR genotyping of the four microsatellite markers (Wrn3, Wrn1, Wrs1, and Wrs3) in the Apoa2 region was carried out for 35 cycles under the following conditions: 94°C, 30 sec; 55°C, 30 sec; 72°C, 1 min. Polymorphisms were detected by electrophoresing the PCR products on 4% Nusieve 3:1 agarose gels in 1× Tris-borate-EDTA running buffer for 2 h at 190 volts. Gels were then stained with ethidium bromide and photographed under ultraviolet light.

Sequencing Apoa2

To define Apoa2 haplotypes, we sequenced the four exons of mouse Apoa2. We amplified each of the four exons with PCR with the same protocol as shown above except using an annealing temperature of 59°C instead of 55°C, and checked the product sizes on 4% Nusieve 3:1 agarose gels. The PCR products were sequenced using Big Dye Terminator Cycle Sequencing Chemistry and the ABI 3700 Sequence Detection System.

Retrieving Data

We retrieved plasma HDL concentrations of the 43 inbred mouse strains from Mouse Phenome Database. Mice had been fed LabDiet 5K52 (6% fat) and were fasted for four hours before their blood was sampled. After precipitating non-HDL cholesterol with polyethylene glycol (20% PEG 8000 in 0.2 M glycine), plasma HDL concentrations were measured with Beckman Coulter Synchron CX5 chemistry analyzer.

APOA2 protein sequences of human, chimpanzee, monkey, cattle, pig, and rat were obtained from GenBank, and D. Puppione (UCLA, Los Angeles, CA) kindly provided horse APOA2 sequence.

Analyzing Statistical Difference

Student's t-test was used to compare the plasma HDL concentrations among the different groups.