To identify differentially expressed genes within the HG2D congenic donor region that potentially underlie body weight and obesity QTL, we hybridized whole brain RNA from non-recombinant HG2DF2 mice (mice inheriting intact cast or b6 congenic haplotypes) from each of the three F2 genotypes (b6/b6, b6/cast and cast/cast) 5 to Agilent whole genome mouse microarrays. Comparison of normalized expression values identified 62 genes as differentially expressed at a false discovery rate (FDR) of <0.30 using an additive effects model (Additional File 1). Fifty-five of the genes (88.7%) were located within the HG2D congenic region (from approximately 75 to 180 Mbp).

As confirmation of the microarray results, we next examined the genetic basis of expression for eleven genes (Cd44, AI451465, Fmn, Scg5, 2310032D16Rik, 3300001M20Rik, Pak7, Actr5, D930001I22Rik, Sdccag33l and Rab22a) identified as differentially expressed from the microarray study (Additional File 1). These genes were selected for confirmation because they were located near the peaks of previously identified HG2DF2 body weight and obesity QTL (Table 1) 5. Gene expression was measured in whole brain RNA from 45 recombinant HG2DF2 mice using quantitative PCR. Of the eleven genes, significant eQTL were identified for five (AI451465, Scg5, 2310032D16Rik, 3300001M20Rik and Actr5) (Table 1).

Based on its close proximity to Wg5 5, biological function and magnitude of differential expression, Scg5 was selected as the highest priority candidate gene. From the microarray analysis, Scg5 was the eighth most significantly differentially expressed gene, with an FDR = 0.10, and its expression was up-regulated by approximately 3.0 fold in non-recombinant cast/cast HG2DF2 mice (Additional File 1). qPCR analysis in HG2DF2 mice indicated the expression difference was due to a Scg5 eQTL located at 55.5 cM (approximately 115 Mbp) mapping with a maximum LOD score of 16.4 (Figure 1). The eQTL was primarily additive (a = -0.68 and d = -0.14), with cast alleles increasing expression. The eQTL explained 74.1% of the variation in Scg5 expression. The nearest marker, D2Mit207 (located at 9.7 cM or 111.8 Mbp), was less than 2 Mbp from the physical location of Scg5 (113.6 Mbp), strongly suggesting the variation controlling the difference in expression was local in nature (eQTL have traditionally been referred to as cis if the variation affecting expression lies in the differentially expressed gene as is the case for Scg5, however, true trans eQTL may also show a similar behavior. Therefore, we have adopted the use of the more general local and distant nomenclature recently proposed by 14). Importantly, previously identified QTL for body weight at 9 weeks (Wg5) 5 and body mass index in HG2DF2 mice (located at 54 cM and 54.7 cM, respectively) were coincident with the Scg5 eQTL (Figure 1).

To determine the effect of genetic background on transcriptional regulation of Scg5, we performed QTL analysis with expression data from three different crosses. Two were analyzed using the GeneNetwork resource for system genetics 1516. The GeneNetwork contains microarray data on whole brain gene expression in B6 × DBA (BXD) recombinant inbred (RI) strains 17 and hippocampus gene expression in B6 × BALB (CXB) RI strains. QTL analysis for the levels of Scg5 expression revealed strong local eQTL in both crosses (Figure 2). Scg5 expression was also found to be under the control of a strong local eQTL using whole brain expression data from a recently described B6-ApoE^-/- × C3H-ApoE^-/- F2 intercross (BXH-ApoE^-/- F2) (Figure 2) 18. In all three crosses B6 alleles decreased expression. Additionally, the expression of Scg5 was correlated with body weight in BXD and CXB mice, but not in the BXH-ApoE^-/- F2 cross (Table 2). Thus, genetic variation near the Scg5 locus accounts for the majority of variation in mRNA levels in multiple genetic backgrounds, and importantly the low-expressing allele is unique to B6.

Since Scg5 is regulated by a strong local eQTL in HG2DF2 mice (Figure 1) we also expected it to be differentially expressed in wild type B62D-3 F2 mice. B62D-3 mice are wild-type with respect to the hg deletion and are congenic for cast alleles from 102–115 on chromosome 2 in the region of Scg5 8. To confirm differential expression we measured mRNA levels in mice of all three B62D-3 F2 genotypes (b6/b6, b6/cast and cast/cast). As expected, Scg5 was up-regulated in cast/cast mice by approximately 2.5 fold (Figure 3). Additionally, if differential expression of Scg5 is causal for Wg5 (in HG2DF2 mice) and Wg2b (in B62D-3 mice), then its expression must correlate with growth traits and indeed its expression was significantly negatively correlated (P < 0.01) with body weight and obesity in both crosses (Table 2).

If an interaction exists between Wg5 and hg (as has been hypothesized in 47), it may be mediated via alterations in Scg5 expression. To determine if expression is altered by hg, Scg5 was quantified in HG and B6 mice at 3, 4.5 and 9 weeks of age. At all three time points the relative expression of Scg5 in whole brain was increased in HG mice (Additional File 2). A suggestive (P < 0.10) 25% increase at 3 weeks was followed by approximately 40% increases at 4.5 (P < 0.05) and 9 weeks (P = 0.06).

The hg deletion encompasses three genes; Socs2, Raidd and Plexin C1 19. The extreme growth rate and mature body size of HG mice (homozygous for the hg deletion) is primarily (if not entirely) due to the absence of Socs2 6. In Socs2^-/- mice increased growth is dependent on Stat5b 20. The DNA sequence required for Stat5b binding has been well characterized 21.Stat5b is the primary transcription factor responsible for mediating Gh induced gene expression changes 22 and a modest increase in its activity has been identified in Socs2^-/- mice 20. To identify putative Stat5b binding sites which may be used by Gh (via Stat5b) to regulate the increase of Scg5 expression in HG mice, we screened 72.683 Kbp of Scg5 genomic sequence for the presence of the Stat5b consensus transcription factor binding site TTCYNRGAA (Y = T or C and R = A or G) 21. The sequence included all introns and exons and 10 kbp of 5' and 3' genomic sequence. In total six consensus sites were identified (Additional File 2). Interestingly, two tandem sites in the Scg5 promoter were located between -56 and -39 (relative to the transcription start site; Additional File 2). The probability that these sites occur by chance is 4¹⁸, suggesting they are functional and may mediate the increased expression of Scg5 in HG mice.

It is likely that most polymorphisms giving rise to local (truly cis) eQTL will reside within promoter or intronic based regulatory elements 23. With this in mind, we analyzed the genomic sequence of Scg5 to identify putative polymorphisms that may be responsible for the strong local eQTL. This is important since it could lead to the discovery of novel modes of Scg5 transcriptional regulation. Based on the eQTL results from multiple crosses, we hypothesized that two haplotypes should exist in the genomic region containing the eQTL, one unique to B6 (low expressers) the other shared between CAST, BALB, DBA and C3H (high expressers). To identify potential regions, haplotype blocks were identified by downloading all SNPs within Scg5 from the Mouse Phenome Database 2425. A total of 191 SNPs were identified between 113.571079 and 113.633403 Mbp on chromosome 2 (Additional File 3). This 62.324 Kbp region encompassed the entire Scg5 coding sequence, promoter, over 5 Kbp downstream and the 3' end of Arhgap11a, which is located 2.4 kbp upstream of the first Scg5 exon. As predicted, two haplotype blocks were identified using the Haploview software package 26. Block 1 extended from the beginning of the queried region at 113.571079 Mbp to 113.581978 Mbp (10.899 Kbp) and block 2 spanned from 113.582743 Mbp to the end of the interval at 113.633403 Mbp (50.660 Kbp) (Additional File 3). Block 1 did not fit the necessary criteria of being unique to B6 (B6 and CAST were very similar across this block), however, block 2 consisted of three predominant haplotypes, one unique to B6, one unique to CAST and the other was highly similar among C3H, DBA and BALB. Although CAST differed from all other strains across the whole block, it was much more similar to C3H, DBA and BALB in the region corresponding to exon 2 and the first two introns. The haplotype similarities and differences across both blocks are shown in cladograms in Additional File 4. Therefore, since the region of haplotype block 2 contains a haplotype unique to B6 and similar among the other strains it is likely the genetic variant underlying the eQTL is located in this region.

None of the SNPs included in the above analysis were located within the proximal promoter. To comprehensively screen this region, we sequenced 2500 bps upstream of the transcription start site (as defined by RefSeq NM_009162) in B6, HG, CAST, BALB, DBA and C3H mice. Ten SNPs were identified across all six strains (Table 3). In addition, an 11 bp deletion between -432 and -422 (del-432) was identified in DBA, C3H and BALB. This deletion was previously identified in C3H mice by Schmidt et al. 27. Consistent with the above haplotype analysis, BALB, DBA and C3H shared an identical haplotype across the region. CAST was polymorphic at all 10 of the SNPs (but not del-432) relative to B6 and the other strains were polymorphic at SNPs -1373, -794 and -659. Interestingly, CAST mice did not possess del-432, however, position -432 was polymorphic between B6 and CAST mice. Therefore, since these four SNPs are unique to B6, the low Scg5 expressing strain, it is probable that one (or more) may give rise to the Scg5 eQTL

To characterize the molecular consequences of Scg5 up-regulation, the levels of 7B2 protein, Pcsk2 gene expression, PCSK2 activity, α-MSH and ACTH levels were quantitated in B62D-3 F2 mice. The levels of pituitary (pit) 7B2 mirrored the differences in whole brain Scg5 expression in mice of differing B62D-3 genotype. b6/b6 mice had the lowest levels with 1.25 ± 0.04 pmol/pit, b6/cast mice followed with 1.34 ± 0.04 pmol/pit while cast/cast mice had the highest levels of 1.55 ± 0.04 pmol/pit (Figure 4). The difference between b6/b6 and cast/cast mice was highly significant (P < 0.01). In addition, the activity of PCSK2 also increased coincident with Scg5 mRNA abundance and 7B2 protein levels (Figure 4). The difference in activity between b6/b6 and cast/cast genotypes was 27%. All pairwise genotype comparisons were highly significant (P < 0.0001). The distal border of B62D-3 ends at 115 Mbp and Pcsk2 is located on chromosome 2 at 143 Mbp. Therefore, the difference in PCSK2 activity could not result from CAST and B6 allelic differences at the Pcsk2 locus. To rule out distant-regulation of Pcsk2 in B62D-3 mice we measured Pcsk2 mRNA levels. No difference in expression was observed due to B62D-3 genotype (Figure 4). α-MSH and ACTH are secreted peptides, both products of convertase-mediated proopiomelanocortin processing in the pituitary. Although PCSK2 activity was increased, no differences were seen in the tissue levels of either peptide (Figure 4).

Phenotypic and gene expression data were generated from two inbred strains (B6 and HG 32), a subcongenic strain (B62D-3 8) and two congenic-derived intercrosses (HG2DF2 5 and B62D-3 F2 8). In addition, Scg5 sequence data were collected from four inbred strains C3H/HeJ (C3H), DBA/2J (DBA), BALB/cByJ (BALB) and CAST and Scg5 expression data were obtained from three additional crosses not produced in our laboratory, the BXD RI, CXB RI and BXH-ApoE-/- F2 1618. All animal protocols were managed according to the guidelines of the American Association for Accreditation of Laboratory Animal Care (AAALAC).

Total RNA was isolated using TRIzol reagent (Invitrogen) from homogenized whole brain samples. Whole brain total RNA from four male non-recombinant (hg/hg, hg/cast or cast/cast across the entire congenic interval) mice was used for microarray analysis. Two samples from each genotype were labeled with each dye (Cy3 or Cy5) using the Agilent Low RNA Input Fluorescent Linear Amplification Kit (Agilent Technologies). Samples were hybridized on the Mouse Whole Genome Oligo Microarray (Agilent Technologies) using an incomplete loop design. Using two arrays, two hg/hg samples (one labeled with Cy3 and the other Cy5) were hybridized with two reciprocally labeled cast/cast samples (i.e. hg/hg Cy3 was hybridized along with cast/cast Cy5). Four additional arrays were used for the other two genotype combinations (hg/hg vs. hg/cast and cast/cast vs. hg/cast). TIFF images were obtained using the Agilent DNA Microarray Scanner BA (Agilent Technologies) and intensities were measured with the accompanying Feature Extraction Software. Intensities were normalized by linear and LOWESS regressions and background corrected by subtracting the trend of minimum intensities along the surface of the slide. Data were then analyzed with the maanova package 333435 from the Bioconductor suite for R 36. Genes were tested both for additive and dominance effects by setting the appropriate contrasts in the mixed model y_ijkl = Dye_i + Array_j + Genotype_k + Error_ijkl where Dye and Genotype are fixed and Array and Error are random effects. P-values were corrected for multiple comparisons using an FDR transformation 37.

Duplicate qPCR reactions were carried out with 25 ng of cDNA using ABI Gene Expression Assays (Applied Biosystems) for each gene on the Applied Biosystems 7500 Fast Real-Time PCR System (Applied Biosystems). The ABI Gene Expression Assays were as follows: AI451465, Mm00505280_m1; Fmn, Mm00439021_m1; Scg5, Mm00486077_m1; Cd44, Mm01277160_m1; 2310032D16Rik, Mm00512140_m1; 3300001M20Rik, Mm00558305_m1; Pak7, Mm00556184_m1; Actr5, Mm00615134_m1; D930001I22Rik, Mm00557752_m1; Sdccag33l, Mm01248119_m1, Rab22a, Mm00508287_m1; and Pcsk2 Mm00500981_m1. Sdha (succinate dehydrogenase complex, subunit A, flavoprotein (fp)) and Gus (beta glucuronidase) were used as endogenous control genes. They are not located in the congenic region (Sdha is on MMU13 at 70.4 Mbp and Gus is on MMU5 at 129.2 Mbp) and both are common, stably expressed control genes. The control corrected expression of each target gene was determined as ΔCt = Ct (target gene) - Ct ((Sdha Ct + Gus Ct)/2). These measurements (ΔCt) were then subjected to interval mapping using the "scanone" function of R/qtl 38. Mouse age, litter size and dam's parity were used as additive covariates. The "fitqtl" function was used to estimate genetic effects and percent variance explained.

qPCR reactions for Scg5 and Pcsk2 were performed in B62D-3 mice as described above. The expression of each target gene was determined using the expression, 2^-ΔCt, where ΔCt = Ct (target gene) - Ct (control gene). These values were analyzed with SAS using a linear model that included the fixed effects of genotype 39. The expression of Scg5 was also measured in whole brain from five HG and B6 mice at 3 WK, 4.5 WK and 9 WK. qPCR was carried out as described above and Sdha normalized expression levels (2^-ΔCt) were calculated for each sample. These values were analyzed with SAS using a linear model that included the fixed effects of strain, age and strain by age interaction 39. The mean values for HG at each age were then scaled to B6 expression levels.

PCR amplicons (Additional File 5) covering approximately 2500 bp upstream of the Scg5 transcriptional start site were sequenced from the B6, HG, CAST, BALB, DBA, and C3H strains. PCR amplified fragments were gel purified using the Qiagen QIAquick gel extraction kit and sequenced at the High-Throughput Genomics Unit at the University of Washington.

All of the following extraction steps were performed at 4°C. Pituitaries were briefly sonicated in 100 mM sodium acetate, pH 5.0, and 1% Triton ×-100 in the presence of a protease inhibitor cocktail composed of 1 μM pepstatin, 1 μM trans-epoxysuccinic acid (E-64), and 1 mM phenylmethanesulfonyl fluoride (PMSF). The extracts were centrifuged for 2 min at 15,000 × g. The supernatants were used for PC2 enzyme assays. The assay for PC2 was carried out in triplicate in 96 well polypropylene microtiter plates using 10 μl of each sample in a total volume of 50 μl containing 200 μM fluorogenic substrate, pyr-Glu-Arg-Thr-Lys-Arg-methylcoumarinamide (MCA) as a substrate and 100 mM sodium acetate buffer (pH 5.0) containing 5 mM CaCl₂, 0.5% Triton ×-100 in the presence of a protease inhibitor cocktail above with the addition of 0.14 mM tosyllysyl chloromethyl ketone (TLCK) 40. The activity was also separately measured in the presence of 1 μM 7B2 CT peptide, a specific inhibitor of PC2 40. The fluorescent product MCA was measured with a Fluoroscan Ascent plate fluorometer. The amount of released product was calculated by reference to the fluorescence of the free MCA standard and is given in fluorescence units (FU) per minute, in which one FU corresponds to 5.33 pmol of MCA. These values were analyzed with SAS using a linear model that included the fixed effects of genotype 39.

Pituitaries were homogenized by sonication in 250 μl of ice-cold 0.1 N HCl. The samples were stored frozen, thawed, and centrifuged for 15 min at 13,000 rpm (17,383 × g) at 4°C. The supernatant was lyophilized and resuspended in 0.5 ml of radioimmunoassay (RIA) buffer (100 mM sodium phosphate, pH 7.4, containing 0.1% heat-treated BSA, 50 mM NaCl, and 0.1% sodium azide). All samples were stored frozen at -70°C until use. For 7B2 assays, 50 μl of pituitary samples were subjected to assay in duplicate. The polyclonal antiserum against 7B2 (LSU13BF), directed against residues 23–39 of 7B2 30, was used to detect 7B2. ¹²⁵I-labeled 7B2 was prepared by the chloramine-T method originally described by Hunter and Greenwood 41 RIAs were carried out according to protocols described previously 42. Samples were incubated with 10,000 cpm of iodinated peptide and the appropriate dilution of rabbit antiserum in a final volume of 300 μl at 4°C overnight. To separate the antibody-bound labeled peptide from the unbound labeled peptide, 1 ml of 25% polyethylene glycol and 100 μl of 7.5% carrier bovine γ-globulin (in PBS) were added. The samples were vortexed, kept on ice for 30 min, and then centrifuged for 20 min at 3,000 × g at 4°C using a Sorvall RT6000B refrigerated centrifuge. The supernatant was aspirated, and the radioactivity in the pellets was determined. These values were analyzed with SAS using a linear model that included the fixed effects of genotype 39.

10 μl of a 1/200 dilution pituitary sample in RIA buffer were subjected to assay in duplicate. The polyclonal anti-α-MSH antiserum was commercially purchased from Chemicon (Temecula, CA). α-MSH assays were performed as described above. ACTH assays were carried out using the two-site Nichols human ACTH_1–39 assay kit (Nichols Institute, San Juan Capistrano, CA). The ¹²⁵I-ACTH antibody used in this kit is directed towards both N-terminal and C-terminal regions of intact ACTH molecule and does not recognize ACTH cleavage products. These values were analyzed with SAS using a linear model that included the fixed effects of genotype 39.