We previously constructed a congenic mouse, B6.S-D2Mit194-D2Mit311 (B6.S-2) with 27 Mb of SPRET/Ei donor DNA on distal chromosome 2 in a C57BL/6J background that captured an obesity quantitative trait locus (QTL). Mice homozygous for SPRET/Ei alleles at the donor region had decreased body weight and obesity-related phenotypes (Diament AL, Farahani P, Chiu S, Fisler J, Warden CH. Mamm Genome 15: 452–459, 2004). In this study, we constructed five overlapping subcongenics with smaller SPRET/Ei donor regions to fine map the underlying gene(s). One of the five subcongenic lines derived from the B6.S-2 founding congenic, B6.S-2A, captured the body weight and adiposity phenotypes in a donor region with a maximum size of 7.4 Mb. Homozygous SPRET/Ei donor alleles in both the founding congenic and the derived B6.S-2A subcongenic exhibited significant decreases in body weight, multiple fat pad weights, and adiposity index (total fat pad weight divided by body weight). Interval-specific microarray analysis in four tissues for donor region genes from the founding B6.S-2 congenic identified several differentially expressed genes mapping to the B6.S-2A subcongenic donor region, including prohormone convertase 2 (PC2; gene name: Pcsk2). Quantitative real-time PCR confirmed a modest decrease of PC2 expression in brains of mice homozygous for SPRET/Ei donor alleles. Analysis of the relative levels of mRNA for B6 and SPRET/Ei in heterozygous congenic mice showed differentially higher expression of the C57BL/6J allele over the SPRET/Ei allele, indicating a cis regulation of differential expression. Using subcongenic mapping, we successfully narrowed a body weight and obesity QTL interval and identified PC2 as a positional candidate gene.
- quantitative trait locus mapping
- mouse chromosome 2
- congenic strains
mouse models are useful tools in the genetic dissection of complex traits such as growth and obesity. Monogenic mouse models have led to the discovery of rare human obesity genes; however, common human obesity is multifactorial, resulting from many genes with small to moderate effects as well as gene-gene and gene-environment interactions. Linkage studies have identified over 150 quantitative trait loci (QTLs) influencing body weight, obesity, and obesity-related traits in a variety of mouse crosses (36). These QTLs often cover broad chromosomal regions that contain several hundred genes. Fine mapping strategies must be utilized to positionally clone the causative gene(s).
Congenic strain mapping is one strategy used to confirm and localize QTLs with subsequent isolation of the underlying gene(s). Congenic strains are developed by introgressing a chromosomal region from a donor inbred strain onto the genetic background of a recipient inbred strain through several generations of backcrossing and selection for the desired genomic interval. Subcongenic strains are produced by additional backcrossing to identify recombinant individuals retaining a smaller portion of the donor region. Congenic and subcongenic mapping have been successful in the isolation and fine mapping of QTLs for a wide range of complex traits (4, 5, 18, 40). Recently, subcongenic mapping was used to positionally clone Sorcs1 as the gene underlying a Type 2 diabetes QTL (8) and Sqle as the gene underlying an obesity QTL (44).
BSB mice are a model of spontaneous complex obesity. They are produced by (C57BL/6J × Mus spretus) F1 × C57BL/6J backcross and exhibit a range of body fat percentage from 1% to 50% on a standard chow diet, while parental and F1 mice are lean (15). Linkage analysis identified QTLs influencing body weight and adiposity measures on chromosomes 2, 6, 7, 12, and 15 (9, 45, 46, 48). We previously constructed (9) a congenic strain, B6.S-D2Mit194-D2Mit311, with M. spretus (SPRET/Ei) donor region alleles decreasing body weight, body fat, and weights of individual adipose depots. This donor region spans ∼27 Mb and contains several hundred genes.
Mouse chromosome 2 is rich with body weight, body length, and obesity QTLs. Several independent mouse crosses have mapped QTLs in regions overlapping our congenic donor region (13, 14, 22, 31, 37–39, 43, 47), indicating the importance of this region in the regulation of growth and obesity. Here we describe the construction of five overlapping subcongenic strains to fine map the mouse chromosome 2 obesity QTL, Bsbob. Subcongenic strain mapping revealed the localization of body weight and obesity effects to a 7-Mb region. We also present candidate gene expression data from an interval-specific microarray and investigate prohormone convertase 2 (PC2) as a candidate body weight and obesity gene.
MATERIALS AND METHODS
The B6.S-D2Mit194-D2Mit311 (B6.S-2) congenic strain was generated and described previously (9). It contains SPRET/Ei donor alleles in a B6.129P2-Lipctm1Unc/J background capturing the QTL Bsbob on mouse chromosome 2. Since no interaction was found between the chromosome 2 QTL and Lipc alleles, the background strain was kept as B6.129P2-Lipctm1Unc/J (9). The background strain has a Lipc-null mutation made in 129P2 ES cells and backcrossed to C57BL/6J (B6) for at least 10 generations so that >99.9% of alleles are B6. To generate recombinants, B6.S-2 congenic mice were crossed to B6.129P2-Lipctm1Unc/J mice. Resulting heterozygous mice were intercrossed to generate F2 mice, and recombinant mice were identified and used as founder mice for subcongenic lines. All experiments were performed on male mice generated from an F2 cross of the congenic or each subcongenic line. The three donor region genotypes are designated as BB (B6 background homozygous), BS (B6/SPRET/Ei heterozygous), and SS (SPRET/Ei homozygous).
At weaning, F2 mice were group housed with ad libitum access to water and a standard chow diet (Purina Rodent Chow 5001, Dean's Animal Feed) containing ∼12% energy from fat. To assess the response to high-fat feeding, one group of male B6.S-2 congenic mice were also fed a purified high-fat diet with 60% energy from fat (Research Diets D12492, Research Diets, New Brunswick, NJ). At 16–18 wk, male mice were fasted overnight, anesthetized by inhaled isoflurane, and bled through the retroorbital sinus. Anal to nasal (AN) length was measured, and mice were killed by cervical dislocation. Tissues were taken and flash frozen in liquid nitrogen. Four white adipose depots, femoral (FWAT), epididymal (EWAT), retroperitoneal (RWAT), and mesenteric (MWAT), were dissected, weighed, and flash frozen. The weights of the four adipose depots were used to calculate adiposity index (total adipose depot weight/live body weight × 100). Body fat analysis by chemical extraction was performed on a subset of B6.S-2 mice as previously described (7). In these mice, there was a strong correlation between adiposity index and percent body fat (r = 0.97, P < 0.0001), with no difference in relationship among donor region genotypes. Therefore, we used adiposity index as a surrogate for body fat percentage.
For food intake measurements, one group of B6.S-2 mice were singly housed in wire-bottom cages with standard chow diet at 7 wk of age and acclimated for 1 wk before the start of the measurement period. Food intake was measured at the end of 7 days and averaged to determine daily intake.
All animals were housed and cared for under conditions meeting National Institutes of Health standards as stated in the Guide for the Care and Use of Laboratory Animals and American Association for Accreditation of Laboratory Animal Care accreditation standards. All animal use was conducted according to Institutional Animal Care and Use Committee-approved protocols.
Genomic DNA was isolated from tail clips by overnight digestion in 0.5% NP-40, 0.5% Tween 20, 40 mM tricine-KOH, 150 mM KOAc, 3.5 mM Mg(OAc)2, 7.5 μg/ml BSA, and 100 μg/ml proteinase K. Simple sequence length polymorphism (SSLP) markers polymorphic between C57BL/6J and SPRET/Ei were used to genotype mice by standard polymerase chain reaction (PCR) methods to amplify DNA. PCR products were run on a 2% agarose gel to resolve SSLPs and visualized with ethidium bromide. Table 1 shows the marker names and genomic positions as mapped on Ensembl v39 (www.ensembl.org), which is based on NCBI mouse assembly m36. For regions for which publicly available polymorphic markers were not available, custom SSLP markers were generated with software available at http://danio.mgh.harvard.edu/mouseMarkers/musssr.html. This program searches for SSLPs in the Ensembl/NCBI build 35 assembly and designs primer sequences flanking them. Primers were tested for polymorphisms between C57BL/6J and SPRET/Ei, and these primer sequences are included in the supplemental data for this article.1
Microarray RNA sample preparation and hybridizations.
Total RNA from whole brain, EWAT, gastrocnemius skeletal muscle, and liver from B6.S-2 F2 animals (3 BB and 3 SS genotypes) was isolated with TRIzol reagent (GIBCO/BRL, Grand Island, NY), followed by additional purification with the Qiagen RNeasy kit (Qiagen, Valencia, CA) and the RNA cleanup protocol. Samples were DNase treated (DNA-free, Ambion, Austin, TX), quantified, and run on a 1% denaturing gel to confirm integrity. cRNA was generated with a protocol developed by NimbleGen Systems. Briefly, double-stranded cDNA was synthesized from total RNA (Invitrogen, Carlsbad, CA). After a phenol extraction, cDNA was transcribed into cRNA with the MEGAscript T7 Kit (Ambion) with biotin-labeled CTP and UTP, followed by purification with the RNeasy kit (Qiagen). cRNA was then hybridized to a custom NimbleScreen 12-well array platform. The array was designed to include 24-mer probes specific to exons between D2Mit138 and D2Mit52, with at least 3 probes/exon when the length of the exon allowed for multiple probes, for a total of 3,919 probes/array with 12 identical arrays (wells) per chip. Genes and exons were identified from the UCSC March 2005 mouse genome assembly based on NCBI Build 34 and the Ensembl v31 mouse genome assembly based on NCBI Build 33. The array was custom manufactured by NimbleGen Systems based on its Maskless Array Synthesizer (MAS) technology (33). After hybridization of the samples, the hybridization solutions were removed and the gene chips were installed in a fluidics system for washes and staining with streptavidin-Cy3 conjugate. The arrays were scanned on a 5-μm scanning platform. The averaged images collected from two scanned images were used for analysis. Two 12-well arrays were used, with brain and liver samples on one and adipose and skeletal muscle on the other. The position of each within the 12-well array was randomized by tissue and genotype.
PCR was performed using gene-specific primers to amplify 500- to 700-bp fragments from cDNA of mice homozygous SS for the congenic region. PCR products were gel purified with the QIAEX II Gel Extraction kit (Qiagen). DNA sequencing was performed with an ABI 3730 Capillary Electrophoresis Genetic Analyzer using ABI BigDye Terminator v3.1 Cycle Sequencing chemistry (Applied Biosystems, Foster City, CA). SPRET/Ei protein coding regions of several positional candidates were sequenced and deposited into GenBank (http://www.ncbi.nlm.nih.gov/genbank/). These genes and corresponding accession numbers include Pcsk2 (EF068176), Rbbp9 (EF068177), Sec23b (EF068178), Polr3f (EF068179), Snrpb2 (EF068180), Csrp2bp (EF068181), and Dstn (EF068182). Primer sequences are listed in the supplemental data for this article. Sequencher software (Gene Codes, Ann Arbor, MI) was used to assemble sequences. SIFT (http://blocks.fhcrc.org/sift/SIFT.html) and Polyphen (http://genetics.bwh.harvard.edu/pph) were used to assess putative functional impact of amino acid polymorphisms between B6 and SPRET/Ei sequences. Signal peptide analysis was performed with Signal P 3.0 Server (http://www.cbs.dtu.dk/services/SignalP).
Quantitative real-time PCR.
RNA was isolated as described above from B6.S-2 F2 animals. cDNA was generated from 1 μg of total RNA with Taqman reverse transcription reagents (Applied Biosystems). Quantitative PCR reactions were performed with a SYBR Green I dye detection system and 30 ng of cDNA. Primer pairs were generated with Primer Express software (Applied Biosystems). For genes within the donor region, primer pairs were designed to nonpolymorphic sequences as identified by B6 and SPRET/Ei sequenced DNA. This was done to ensure that polymorphisms did not affect annealing and efficiency of primers between genotypes. The primer pair sequences for 4930529M08Rik, AK045769, Snrpb2, Pcsk2, Polr3f, Ovol2, and Rrbp1 are included in the supplemental data for this article. Two control genes, Gapdh and Gusb, were used to correct for input RNA with the following primers: Gapdh-F: 5′-TTGTCTCCTGCGACTTCAACA-3′, Gapdh-R: 5′-CCCCGGCATCGAAGGT-3′; Gusb-F: 5′-CATGAGAGTGGTGTTGAGGACTA-3′, Gusb-R: 5′-CCCATTCACCCACACAACTG-3′. Samples were run in duplicate, and threshold cycle (Ct) values were averaged. Ct is the PCR cycle in which fluorescent signal associated with the exponential growth exceeds the threshold (10 × noise level). mRNA levels were expressed as fold difference over the lowest value, corrected for the control. This was calculated by first determining the fold difference over the control genes with the equation 2−ΔCt, where ΔCt = Ct(gene) − Ct(control gene). All values were then divided by the group average 2−ΔCt of the BB genotype to set the BB group to a value of 1. We termed this “relative expression.” Gapdh and Gusb Ct values were well correlated with an r > 0.87 in each tissue, so we chose to present data from one control gene, Gapdh.
Pyrosequencing was used to assay for the presence of allelic imbalance of gene expression in mice heterozygous (BS) for the B6.S-2 congenic donor region. Single nucleotide polymorphisms (SNPs) between B6 and SPRET/Ei were identified by sequencing of SPRET/Ei alleles and comparison to publicly available B6 sequence for target genes. cDNA and genomic DNA from mice BS for the congenic region were amplified with gene-specific PCR primers designed with Primer3 software (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi). Primer extension sequencing primers were designed with Pyrosequencing SNP Primer Design Software v1.01 (www.pyrosequencing.com). PCR products were immobilized on a solid phase and prepared for Pyrosequencing according to Dunker et al. (12), sequenced with PyroGold reagents (Biotage), and analyzed with PSQ96 2.1 allele quantification (AQ) software. To assess allelic imbalance, the percentages of B6 and SPRET/Ei alleles detected in the cDNA samples were compared with those detected in genomic DNA. Heterozygous genomic DNA represents a 50:50 mix of B6:SPRET/Ei alleles and corrects for any PCR or Pyrosequencing assay bias. A significant difference between heterozygous cDNA and genomic DNA indicates deviation from 50:50 and preferential expression of one of the two alleles. The PCR primers used were Pcsk2-PCRF: 5′-TGGCAATCATTGAGAATTCATTT-3′ and Pcsk2-PCRR: 5′-ATGGCGTCTGTGTCTTCAAT-3′. Sequencing primers used to assess two SNPs within the PCR product were Pyroseq1: 5′-GACATTCTGTGTGCACAT-3′ and Pyroseq2: 5′-GTGTATATATACCTGTATAA-3′.
Data are presented as least-squares means ± SE. A general linear model was used to assess genotypic effects on phenotypes with genotype, age, and genotype × age as factors with JMP (SAS Institute, Cary, NC). Pairwise comparisons were analyzed by Tukey's post hoc test. A P value <0.05 was considered statistically significant. In Pyrosequencing experiments, a Student's t-test was used to test for pairwise comparisons for RNA vs. DNA. For adipose depot weights, four decimal places were used in calculations; however, only two are presented. Any outliers for the major outcomes of body weight and adiposity index were removed if they met two criteria: 1) the observed values of the phenotypes were outside the range of 3 SD, and 2) the observation had a standardized residual >2 in absolute value. Cook's D statistic was used to determine whether the outliers should be discarded and was done with the PROC REG procedure and SAS 9.1 (SAS Institute).
Before data analysis of microarrays, data preprocessing was performed by quantile-quantile normalization to remove variability among arrays that is unrelated to strain differences in expression. Quantile-quantile normalization is a ranking procedure in which each observation on the chip is ranked by fluorescence level and then converted to the value of a deviation that would be expected from the standard normal distribution based on the observation's rank. This procedure results in data from each chip with a mean of 0 and SD of 1.0. To compare the changes in transcript expression between two B6.S-2 donor region genotypes, homozygous congenic (SS) and homozygous background (BB), we conducted the following statistical evaluation procedure. First, we obtained expression measures at gene level and summarized the gene-level data by averaging the transcript intensity levels from all probes that correspond to each gene. For the primary analysis to discover differentially expressed genes, we used the Student's t-test with multiplicity adjustment by the false discovery rate (FDR) (3) for all pairwise comparisons, each of which generated the distribution of frequent P values for genes. Gene expression changes were called significant when the P value was <0.05 after the FDR control. In the second step, we used a mixture modeling approach (2) that estimates the posterior probabilities that a gene is a true positive (denoted as PTP). The PTP is the Bayesian probability that a gene with a given frequentist P value or smaller is truly different between the two strains being studied. The mixture modeling approach also provides an omnibus test of whether there is any overall effect of genotype on gene expression levels, an estimate of the total number of genes that have their expression levels altered, and a probabilistic model predicting the effects of strain character on gene expression. For brain tissue, the power was not adequate to determine whether the mixture model method would be reliable in estimating the number of genes for which there is a true difference in expression levels, and the PTP that a gene is truly differentially expressed was not calculated. Microarray data were submitted to the Gene Expression Omnibus (GEO) database (www.ncbi.nlm.nih.gov/geo/) under the accession number GSE6115.
Five overlapping subcongenic lines were produced from the original B6.S-D2Mit194-D2Mit311 (abbreviated B6.S-2) congenic strain that collectively captured the full congenic region. Figure 1 outlines the SPRET/Ei donor region introgressed in each subcongenic line. Refinement of the original B6.S-2 congenic borders by querying additional markers resulted in a maximum donor region interval of 26.9 Mb. The derived subcongenic lines range from 6.7 to 19.1 Mb in size. The minimum interval is defined as the most proximal and distal boundary markers with known SPRET/Ei genotypes for a given donor region. The maximum interval is defined as the interval between the two markers just outside the minimum interval with known B6 genotypes for a given donor region. The small interval between minimum and maximum interval markers is of unknown genotype.
Mapping of body weight and obesity traits to proximal region of B6.S-2.
All mice phenotyped were males derived from an F2 cross of heterozygous congenic mice in order to minimize possible environmental and background effects. For subcongenic mapping, only male mice on chow were collected.
We collected and phenotyped a second cohort of B6.S-2 mice, separate from those included in Diament et al. (9), for the duration of this experiment. Similar to previously published data, the original B6.S-2 congenic showed a strong genotype effect for body weight as well as significant genotype effects on individual fat pad weights and adiposity index, with SS mice being significantly leaner than BB mice and BS mice having intermediate values. Of the different adipose depot weights, RWAT and MWAT were reduced the most, with SS mice having a decrease of 27% of RWAT and MWAT compared with BB mice. FWAT and EWAT were reduced 20% and 22%, respectively. Table 2 shows the phenotypes for the B6.S-2 line as well as all subcongenic lines.
The B6.S-2A subcongenic, with a maximum donor region of 7.4 Mb, showed a significant decrease in body weight in BS and SS mice compared with BB mice. Homozygous SS mice also showed a significant decrease in FWAT, EWAT, MWAT, and adiposity index compared with BB mice. Similar to the B6.S-2 founding congenic results, MWAT had the greatest reduction (28%) in SS mice while FWAT, EWAT, and RWAT were reduced 17%, 19%, and 22%, although the difference in RWAT was not significant (P = 0.15). This absence of a genotype effect on RWAT may be due to the very small size of the depot, because any measurement error or variation would greatly affect its value. It is also possible that the original RWAT QTL was due to the sum of multiple small-effect QTLs now separated by subcongenic mapping into individual, undetectable QTLs. Heterozygous BS mice had shorter AN body lengths than homozygous BB mice. Even after adjustment for body length, the genotype effect on body weight, FWAT, EWAT, MWAT, and adiposity index remained significant, with SS mice being consistently leaner than BB mice (data not shown).
The B6.S-2B subcongenic line, with a donor region encompassing the proximal two-thirds portion of the B6.S-2 donor region, showed a similar genotype effect on body weight, with SS mice weighing the least, although the pairwise comparisons did not reach statistical significance. However, for adiposity measures, only EWAT showed a significant effect, with heterozygous BS mice having significantly larger EWAT than homozygous mice. This overdominance pattern was seen in the other fat pads as well as in adiposity index. However, there were no significant genotype effects on these measures. BS mice had significantly longer body lengths than SS mice. Interestingly, the genotype effect on body weight and EWAT disappeared after body length was added as a covariate, suggesting that any body weight and adiposity differences in this subcongenic may be attributable to differences in length. The donor region of this subcongenic line encompasses the entire donor region of the B6.S-2A line. It is possible that the B6.S-2B donor region contains a second gene distal to the B6.S-2A border that modifies the adiposity effects of the gene present in the B6.S-2A line. The fact that this phenotype pattern is not seen in the full B6.S-2 congenic then would be attributed to another locus distal to the B6.S-2B border. A simpler explanation for these results and patterns is that the weight and EWAT genotype effect in the B6.S-2B line is due to the same gene that is present in the B6.S-2A region and results of other phenotypes are nonsignificant because of small effect size and sampling. An analysis within donor genotype of the B6.S-2, B6.S-2A, and B6.S-2B lines showed no significant line effects on fat pad weight or adiposity index of the B6.S-2B line compared with either the B6.S-2 or B6.S-2A line in BB, BS, or SS mice, with the exception of heterozygous B6.S-2 mice having a smaller adiposity index than either B6.S-2A or B6.S-2B.
The B6.S-2C and the B6.S-2E subcongenics showed no genotype effects on any of the measured phenotypes, indicating the absence of a body weight- or adiposity-influencing gene in those regions.
The B6.S-2D subcongenic showed a marginally significant genotype effect (P = 0.0493) on body weight, although pairwise comparisons showed no significant differences between groups and the effect was nonsignificant (P = 0.10) after adjusting for body length. There were no genotype effects on any other phenotype. The B6.S-2D congenic overlaps with both the B6.S-2C and B6.S-2E congenics, neither of which had significant or suggestive genotype effects on body weight. Therefore, this suggestive body weight gene maps to a 2.5-Mb region of nonoverlap.
Identifying positional candidate genes by microarray analysis.
We used a custom-designed microarray containing oligonucleotide probes to genes within the maximum B6.S-2 congenic donor region from D2Mit138 to D2Mit52. The microarray contained at least three 24-mer probes per exon per gene and was designed from publicly available sequences.
Because the B6.S-2A retained a body weight and adiposity phenotype, we focused our interest on differentially expressed genes residing within the B6.S-2A subcongenic donor region. This array included probes for all known protein-coding genes in the B6.S-2A congenic donor region. Comparison with the current Ensembl v39 mouse assembly showed 36 known and predicted protein-coding genes in the minimum donor region and 45 in the maximum donor region. Of the genes in the maximum donor region, 33 were present in the UCSC March 2005 or Ensembl v31 assembly. The B6.S-2A maximum donor region also includes five genes for noncoding RNA, four novel genes identified from Ensembl v39 that have no known homology to other named genes, and eight genes added to GenBank after our experiment was completed. All known and predicted genes identified by the Ensembl v39 assembly within the B6.S-2A maximum donor region are listed in Table 3.
In any one of the four tissues examined only a few genes remained positional candidates. Two genes, 4930529M08Rik in skeletal muscle and a mouse EST clone (accession no. AK045769) in whole brain, passed initial criteria of FDR of P < 0.01. We extended our criteria to those genes passing FDR of P < 0.05, including Pcsk2 and Ovol2 in brain, Snrpb2 in skeletal muscle, Polr3f in adipose tissue, and Rrbp1 in liver. This is summarized in Table 4.
Microarray validation by quantitative real-time PCR.
Using quantitative real-time PCR, we validated those genes in the B6.S-2A donor region identified as differentially expressed by microarray analysis. Because the microarray probe production was based on public sequence derived mostly from the B6 strain, we sequenced a portion of each gene in order to design custom real-time PCR primers to sequences without polymorphisms between B6 and SPRET/Ei. Gene expression validation was performed in the same strain, B6.S-2, as used in the microarray experiment. While there is a possibility that more than one QTL exists in the 27-Mb B6.S-2 donor region as discussed above, this would likely not affect the detection of cis-regulated transcripts underlying any of the possible QTLs within this region. Five of the seven identified genes had significant genotype effects on gene expression (Table 5). Four of the genes, 4930529M08Rik, Snrpb2, Pcsk2, and Ovol2, had genotype effects in the same direction as predicted by the array data.
PC2 as a candidate gene.
We investigated proprotein convertase subtilisin/kexin type 2, also known as prohormone convertase 2 (PC2, gene name Pcsk2), further because of its known function and involvement in several metabolic processes, differential expression in brain, and involvement in the same processing pathways as PC1, a known human and mouse obesity-causing gene. PC2 processes several hormones and proproteins at the COOH-terminal side of dibasic or monobasic amino acid residues including proinsulin, proglucagon, and ACTH and is present in neuroendocrine cells such as the brain, gut, and pancreatic islets. It is in the same protein family as PC1, a gene causing rare Mendelian human obesity (21). A PC1 mutant mouse (N222D) is obese and hyperphagic and is characterized by multiple endocrine effects, while the null mouse is retarded in growth because of a lack of mature growth hormone releasing hormone (28).
Sequencing of SPRET/Ei PC2.
The full-length coding region was sequenced in a homozygous congenic mouse harboring SPRET/Ei alleles at PC2. A total of 18 SNPs were identified between the publicly available B6 coding sequence and SPRET/Ei sequence. Only one SNP resulted in a missense mutation, changing a phenylalanine in B6 to a leucine in SPRET/Ei at amino acid position 14. The potential functional impact of this SNP was tested with both SIFT and Polyphen, two programs that predict functional consequences of coding sequence mutations. The missense mutation was predicted to be tolerated by SIFT and benign by Polyphen and lies within the secretory signal peptide domain of PC2. Analysis of both B6 and SPRET/Ei alleles using the SignalP 3.0 Server prediction program showed that the Phe-to-Leu substitution would not affect location or presence of the signal peptide. While these results are not definitive, they suggest that the Phe-to-Leu polymorphism does not affect functionality of PC2.
Determination of cis/trans regulation of PC2 by Pyrosequencing.
To determine whether the expression difference of PC2 between BB and SS mice was due to cis-acting or trans-acting effects, we performed a Pyrosequencing assay to assess the potential differential expression of B and S alleles in the congenic. Cis-acting effects are those that affect gene expression directly, for example, through polymorphisms in the promoter region or in expressed sequences that influence mRNA stability. Trans-acting effects on gene expression do not depend on sequence variation at regulatory regions of the gene itself, but instead are due to differences in trans-acting factors, such as transcription factors, that interact with the gene of interest to influence its transcript levels. If a gene is responsible for a QTL through differences in its gene expression, this must be due to cis regulation, because trans regulation would implicate another gene. The Pyrosequencing assay assesses the relative contributions of B and S allele expression in BS heterozygous cDNA samples using SNPs between the two strains in a classic F1 cis/trans test (11). The same assay run on DNA from BS mice would represent a theoretical 50-to-50 ratio of B and S alleles. Ratio differences between cDNA and DNA would suggest that one allele was more highly expressed than the other and would indicate a cis effect. Two different SNPs within the same PC2 PCR product were assayed.
For both SNPs, cDNA samples had a significantly larger percentage of B allele expression compared with in DNA (Fig. 2). SNP1 had a %B allele in cDNA of 73.2 ± 0.4 vs. a %B allele in genomic DNA of 70.4 ± 0.5 (P = 0.002). The deviation of genomic DNA from 50% indicates Pyrosequencing assay bias of alleles, and comparison of cDNA to genomic DNA corrects for this. SNP2 had a %B allele in cDNA of 52.7 ± 0.2 vs. 49.7 ± 0.1 in genomic DNA (P < 0.0001). These results show consistently higher expression of the B allele in whole brain cDNA samples of heterozygous mice and, therefore, lower expression of the S allele. The differences, however, are small. This assay also indicates that this imbalance is due to cis regulation rather than trans regulation, meaning there is a polymorphism in the congenic region directly affecting transcript levels of PC2.
Other positional candidate genes.
We sequenced the SPRET/Ei allele of several genes within the B6.S-2A donor region including Csrp2bp, Dstn, Polr3f, Rbbp9, Sec23b, and Snrpb2 and observed no amino acid changes in any of the fully sequenced open reading frames. Of the validated differentially expressed genes, none has published involvement with obesity or energy balance except for PC2.
Food intake is decreased in B6.S-2 congenic mice, but feed efficiency is increased.
We measured food intake in a subset of B6.S-2 congenic mice. There was a significant genotype effect on daily food intake (P = 0.046), with SS mice consuming significantly less than BB mice (3.67 ± 0.09 vs. 4.07 ± 0.12 g) (Fig. 3). This genotype effect was also seen when correcting daily food intake for initial body weight (P = 0.01) or change in body weight during the food intake period (P = 0.01), suggesting an increased feed efficiency. Heterozygous mice had food intake values between those of BB and SS mice.
Effects of B6.S-2 donor region on response to diet with 60% kcal from fat.
Some genes may cause obesity irrespective of diet, while others influence obesity on some diets but not others. To investigate the effects of diet-induced obesity on B6.S-2 mice, F2 mice were placed on a purified high-fat diet consisting of 60% energy from fat. There was a significant genotype effect on body weight, with SS mice weighing the least, consistent with data observed for mice maintained on a low-fat chow diet (Table 6). FWAT and MWAT were also significantly decreased; however, adiposity index was not, suggesting that homozygous congenic SPRET/Ei alleles were able to protect against diet-induced weight gain, partly through decreases in FWAT and MWAT. As seen with mice on chow, MWAT was decreased more than any of the other depots. However, SPRET/Ei alleles were not able to protect against the diet-induced increase in adiposity index, perhaps because effects of this very high-fat diet superseded the genetic effects on adiposity.
In this study, we have used subcongenic mapping to narrow a body weight and obesity QTL captured by the B6.S-D2Mit194-D2Mit52 (B6.S-2) congenic from a 27-Mb region to a 7-Mb donor region in our B6.S-2A subcongenic on mouse chromosome 2 and to identify positional candidate genes.
We used an interval-specific microarray to identify seven genes that were differentially expressed within the congenic donor region of the B6.S-2 line and between homozygous B6 and SPRET/Ei donor genotypes. Pcsk2, or prohormone convertase 2 (PC2), was one of the seven genes shown to be differentially expressed by microarray analysis. PC2 is an enzyme involved in processing neuroendocrine precursors to produce active hormones and neuropeptides including insulin, glucagon, and α-melanocyte-stimulating hormone (α-MSH). PC2-null mice show abnormalities in several endocrine pathways such as a lack of glucagon, severe block in insulin processing, and reduced glucose levels (16, 35). There is evidence that PC2 has epistatic interactions with other genes, because there is no effect on resting glucose levels when the PC2-null allele is transferred from its B6/129 F1 strain background to a B6 background (35). This strongly indicates different interactions of PC2 with B6 and 129 strain alleles of an unknown other gene. PC2-null mice weigh 5–18% less than their wild-type counterparts with no gross anatomic disturbances (16). They lack pituitary and brain α-MSH and demonstrate high levels of ACTH in the pituitary and hypothalamus (24, 32) due to the inability to process α-MSH from ACTH.
In our study, we found a small but statistically significant decrease in brain PC2 expression in homozygous congenic SPRET/Ei mice compared with background controls, which may have effects on whole body energy balance and metabolism. While this decrease may be modest, a recent publication showed the positional cloning of Gpc3 as a causal gene for a growth QTL in mice, with no nonsynonymous amino acid polymorphisms and only an 15% decrease in glypican-3 mRNA (34). This result is strongly supportive of the hypothesis that small changes in gene expression may be physiologically relevant. It is also possible that other time points or tissues may show more marked genotype differences in PC2 expression. For example, our brain samples were taken from mice fasted overnight, and it has been shown that prolonged fasting can decrease PC2 expression in the hypothalamus (41). Future functional studies of the potential effects of reduced PC2 in our B6.S-2 and B6.S-2A mice will test the role of PC2 as an obesity candidate gene.
While we have not identified any other missense mutations in genes sequenced, we identified several genes mapping to the B6.S-2A subcongenic that were differentially expressed between donor region genotypes. Four of these genes (including PC2) were validated by quantitative real-time PCR. 4930529M08Rik is a cDNA clone coding for an unknown hypothetical protein, LOC78774, with increased expression in SS mice. Ovol2, with decreased expression in our SS mice, is a zinc finger transcription factor shown to be involved in developmental signaling pathways such as the positioning of the neuroectoderm/surface ectoderm border in the head and closure of the cranial neural tube (30). Snrpb2 codes for the U2 small nuclear ribonucleoprotein B, and its expression is decreased in our SS mice. While Snrpb2 itself has not been directly associated with obesity in the literature, another small nuclear ribonucleoprotein, Snrpn, lies within the Prader-Willi locus on human chromosome 15. Microdeletions in the Snrpn locus have been implicated as causal in cases of Prader-Willi syndrome, making Snrpb2 a candidate gene by virtue of its functional similarity to Snrpn (6, 19). Our results are consistent with two different models for the underlying cause of body weight and obesity phenotypes in B6.S-2A mice: 1) subtle changes in gene expression of a gene with known involvement in energy balance and metabolism, PC2, are responsible, or 2) changes in expression or presence of sequence variation of a gene or genes with no previously known connection to obesity are the underlying cause. Both outcomes are instructive about the causes of obesity.
Other congenic lines capturing obesity-causing alleles overlapping our congenic and subcongenic donor regions on chromosome 2 have been published. We previously (47) identified an adiposity QTL in the congenic B6.UW-H3b we Pax1un d/Sn with a QTL peak overlapping our B6.S-2A donor region. Recently, Jerez-Timaure et al. (22) fine mapped QTLs for growth and fatness in a M16i.B6-(D2Mit306-D2Mit52) congenic line, using recombinant progeny testing. They found a QTL for body weight, total fat mass, and gonadal fat at D2Mit194 (143.6 Mb), coincident with our B6.S-2A donor region. Additionally, they found a second QTL for gonadal percentage and body fat percentage between D2Mit285 (152.5 Mb) and D2Mit263 (162.0 Mb), which encompasses our second suggestive body weight-critical region. In their model, the very large M16i strain served as the background strain that was selected for rapid weight gain for 27 generations (23). A congenic with B6 alleles in the congenic donor region of the M16i strain promoted leanness. In contrast, B6 alleles in our B6.S-2A congenic promote fatness relative to SPRET/Ei alleles. It is unlikely that our QTL and those described by Jerez-Timaure et al. result from the same alleles of the same genes, yet it is possible they result from polymorphisms in the same quantitative trait gene, with the B6 allele promoting an intermediate phenotype between the very lean SPRET/Ei mice and the very large M16i mice. Alternatively, the quantitative trait genes identified in the two crosses may be completely different. Several candidate genes mapping to the M16i.B6-(D2Mit306-D2Mit52) donor region were identified by gene expression analysis between the congenic strain and its background M16i strain (23). One of these genes, destrin, mapped to our B6.S-2A donor region; however, we found no amino acid polymorphisms or evidence for expression difference.
The donor regions of our congenic and subcongenics are syntenic to human chromosome 20. Hunt et al. (20) and Lee et al. (25) both found QTLs for body mass index in this syntenic region. Norman et al. (32a) identified a QTL for 24-h respiratory quotient (RQ) in a population of Pima Indians on human chromosome 20 in this region. Lembertas et al. (27) mapped a QTL for body fat percentage, mapping slightly more distal on mouse chromosome 2, but overlapping with our SPRET/Ei donor regions. The presence of these QTLs gives support that the human orthologs of genes in this region of mouse chromosome 2 may be relevant to human obesity and energy balance.
The ability to resolve closely linked QTLs is the one major advantage of subcongenic mapping. We have constructed congenics to localize an obesity gene to a 7-Mb region and have also identified a second suggestive body weight QTL mapping to a 2.5-Mb region. The discovery that a congenic actually contains two or more quantitative trait genes is not novel. Several mapping studies with various complex traits have also isolated closely linked QTLs with the use of subcongenics, separating the effects of two or more causal genes (1, 10, 17, 42). Our second body weight QTL is only suggestive, perhaps in part because of smaller numbers of mice studied in the B6.S-2D line or because the causal gene may have a small effect on body weight, not easily detectable in mice on a very low-fat chow diet. Similarly, we were unable to fine map the RWAT effect, perhaps because the small depot size results in larger measurement error or because there exist several small-effect RWAT QTLs in this region that are not detectable in any one subcongenic. There are emerging studies showing the separation of a previously major QTL into two or more smaller effect loci, each with barely detectable or nondetectable effects (26, 29, 49). This phenomenon may make fine mapping of causal genes much more difficult, requiring alternative mapping strategies and/or refined phenotyping to potentially identify a related trait with a larger effect. One could also attempt to exacerbate the phenotype, for instance, by feeding mice a higher- or lower-fat diet in the case of obesity-related traits or by increasing the sample size studied.
Although we did not systematically examine all genes in the B6.S-2A donor region, the production of microcongenics, capturing only a portion of the B6.S-2A region, will allow us to be more exhaustive in our testing of candidate genes. For example, while we did not test recently identified genes not represented on the microarray, mapping by microcongenics will decrease the candidate list substantially, allowing us to individually sequence and perform expression analysis by quantitative PCR on these genes. Microcongenics retaining the body weight and obesity phenotypes will also allow us to confirm our gene expression observations presented in this paper.
In summary, we have successfully constructed subcongenic strains to fine map body weight and obesity QTLs on mouse chromosome 2. We found one highly significant locus mapping to ∼7 Mb. Using an interval-specific microarray, we identified several positional candidate genes including PC2. Expression analysis indicated that the differential expression was due to cis effects of a polymorphism within the congenic donor region. The expression differences were significant, but modest. Functional studies with PC2 in our mice will determine its viability as a candidate obesity gene while other candidate genes identified from the microarray are tested. Further mapping with subsequent candidate gene testing with microcongenics of our 7-Mb region will help identify the causative gene.
This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases Grant DK-69978.
We thank Sue Bennett and Noreene Shibata for their mouse care and husbandry and Jackie Leung for her help with genotyping. We also thank Juan Medrano, Charles Farber, Ricardo Verdugo, and Rodrigo Gularte for their project guidance.
↵1 The online version of this article contains supplemental material.
Address for reprint requests and other correspondence: C. H. Warden, Rowe Program in Genetics, Univ. of California, Davis, CA 95616 (e-mail:)
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).
- Copyright © 2007 the American Physiological Society