Results

Speed congenic strain development
Two speed congenic panels were created, the first comprehensively dissected MMU2 while the second isolated QTL on MMU1, 5, 8, 9, 11 and 17 (Table 1). The MMU2 panel consisted of eight congenic strains developed by introgressing four overlapping donor regions onto both B6 and HG genetic backgrounds (Table 1). Single donor regions bred onto an HG background were created to isolate the remaining QTL on MMU1, 5, 8, 9, 11 and 17. Implementation of a speed congenic approach using marker-assisted breeding with 79 genome-wide microsatellite markers accelerated production of all strains [5,6] (Additional File 1 and Figure 1). Strain abbreviations and genomic region isolated by each strain are listed in Table 1. In addition to both congenic panels, two control strains homozygous B6 (B6.CASTC; B6C) or HG (HG.CASTC; HGC) for all genome-wide markers genotyped were developed from the same cross (see Methods) and served as the basis for strain comparisons. After stabilizing each congenic, 12 of 14 were phenotypically characterized for growth and adiposity. The HG2D and HG5 strains were not characterized due to reproductive problems. The recombinant end points for all strains were refined using microsatellite markers flanking each donor region (Additional File 2).

MMU2 speed congenic strain characterization
Due to the overlapping nature of the MMU2 congenics, five distinct chromosomal regions (Regions I–V) were queried for the presence of QTL (Figure 2). As a separate analysis we also tested for interactions between each donor region and the two distinct genetic backgrounds. The following sections describe the phenotypes for each B6.CAST and HG.CAST speed congenic strain as well as the results of the interaction analysis.

B6.CAST MMU2 speed congenic strains
The B62D strain exhibited the largest (P < 0.0001) decreases in body weight of any B6.CAST congenic (Additional File 3). Both sexes had reductions in weight at 2 (2WK), 3 (3WK), 6 (6WK) and 9 (9WK) weeks of age compared to control B6C mice (Figure 3). Despite large decreases in body weight, no differences in growth rates (G26, weight gain from 2 to 6 weeks and G29, weight gain from 2 to 9 weeks) were observed (Additional File 3). Therefore, the B62D unique region (Region V) harbors either an early-growth QTL or maternal genotype effect which produces a distinct decrease in body weight prior to 2WK. These effects are evident in the growth curves for both sexes of each strain (Figure 3). Additionally, B62D mice displayed significantly shorter tail lengths despite no difference in naso-anal length (NA) (Table 2).
In contrast, B62M mice displayed small decreases in body weights and growth (male 9WK was significant, female 6WK and 9WK were suggestive at P < 0.05) and significant decreases in NA (male P = 0.0152) and tail length (Additional File 3 and Table 2). The decrease in NA in B62M mice is surprising since the 2M donor region is entirely nested within the 2D donor region and B62D mice showed no difference in NA (Figure 2). These data indicate the presence of three QTL with the 2D unique region. Two of which are disjoined by the B62M and B62D strains and additively decrease body weight and tail length, but not NA. The third QTL, located within the B62D unique region (Region V), increases NA canceling the effects of the B62M QTL (Region IV) decreasing NA.
B62PM females had decreases in all growth-related traits except 2WK and 3WK, while no differences were seen in males (Additional File 3). Similarly, only B62P females and not males showed significant decreases in 9WK, G26 and G29 (Additional File 3), indicating these two strains share a female specific growth QTL (Region II).
In the original genome scan body fat percentage as determined by chemical compositional analysis was not linked to markers on MMU2 [24]. However, since numerous obesity QTL have been found on MMU2 [2], some of which were discovered in B6 by CAST crosses [9,25], we chose to measure dissected fat pad weights as a more sensitive measure of adiposity.
Table 2 lists the weights of gonadal (GFP), femoral (FFP), mesenteric (MFP) and retroperitoneal (RFP) fat pads, along with total fat pad weight (sum of the four fat pads), adiposity index (AI = TF/weight at sacrifice (WSAC)) and body mass index (BMI = WSAC/NA2 *100) for each MMU2 congenic strain. Similar to large differences in body weight the B62D strain displayed highly significant decreases in adiposity. B62D was the only strain where both sexes had significant decreases in TF, AI and BMI. Interestingly, no difference in adiposity was seen in B62M mice. This discordance indicates the leanness promoting QTL is located in the B62D unique region (Region V). Only minor differences in various fat pad weights were observed for the B62P and B62PM strains (Table 2).

HG.CAST MMU2 speed congenic panel
HG2P and HG2PM mice of both sexes displayed significant decreases in post-weaning body weights and growth rates (Additional Table 3). In addition, the HG2M strain displayed highly significant decreases in length, similar to B62M mice (Table 2). These results indicate QTL within Region II shared between the HG2P and HG2PM strains decrease body size, weight gain and NA (only in females), while a QTL in the HG2M region significantly decreases tail length.
Similar to growth traits a general decrease in GFP, MFP, RFP, TF and AI were seen in HG2P and HG2PM mice (Region II) (Table 2). These differences in adiposity are likely pleiotropic effects of the growth QTL shared between these two strains. Surprisingly, HG2M mice displayed striking sex differences in adiposity. Males exhibited an increase in AI (P = 0.0256) with a decrease in BMI (P = 0.0462) and the exact opposite was seen in females (AI, P < 0.0001; BMI, P = 0.0425) (Table 2). Therefore, the obesity QTL in the HG2M unique region (Region IV) is profoundly impacted by the presence of hg.

Confirmation of QTL-hg interactions
An important feature of our experimental design was the ability to test for interactions between QTL in each MMU2 donor region and genotype at the HG locus, since identical donor regions were introgressed on two genetic backgrounds, B6 (+/+) and HG (hg/hg). Significant interactions between donor region and HG genotype were viewed as strong evidence that hg modifier QTL reside within that unique genomic region.
In the original linkage analysis the mode of gene action and peak location of Wg2 were dependent on the presence of hg [24] (Table 1). In the present study, only homozygous congenic mice were characterized, thus, the overdominant effects of Wg2 were not tested. However, since its original peak location differed dependent on background we hypothesized that Wg2 represents a set of linked QTL between 74.9 and 181.8 Mbp within the 2M and 2D donor regions (Region III–V), some of which interact with hg. Following this logic we expected one or both of these strains would exhibit donor region by HG genotype interactions. Unfortunately, we were unable to characterize HG2D mice. However, as noted above, strong sex-specific effects were seen in HG2M mice. Significant 2M donor region by HG genotype by sex three-way interactions for AI (P = 0.0004; Figure 4) and TF were identified (Table 3). The basis of these interactions was a decrease in HG2M female and an increase in HG2M male adiposity, with no differences in fat accumulation across B62M sexes. Significant 2PM donor region by HG genotype two-way and 2PM donor region by HG genotype by sex three-way interactions were also seen for TF and AI (Table 3). In addition, significant 2P donor region × HG genotype interactions were observed for all traits listed, although some traits did reached significance at the critical P < 0.0071 (Table 3). The basis for each interaction was a decrease in phenotype in HG2P compared to HGC and no difference between B62P and B6C. Together these data confirm that MMU2 QTL modify the effects of the hg deletion.

Speed congenic strains for QTL on MMU1, 8, 9, 11 and 17

HG1
Q1Ucd1 affecting G26 was the least significant of any QTL identified in the original intercross and only reached a suggestive level of significance (LOD = 2.46, P < 0.05) [26]. In accordance, HG1 mice had small differences in all growth and length traits. NA in both sexes and 2WK, 3WK, 6WK and tail length in females was significant at a critical P < 0.005 (Table 4). The only difference in adiposity was a significant decrease in male MFP (Table 5). The successful capture and confirmation of the small effect Q1Ucd1-w26 QTL adds support to pursuing QTL identified at suggestive levels of statistical confidence from genome scans.

HG8
Markers on MMU8 were linked with G29 in the original intercross [24]. This QTL was small, only accounting for 4.3% of the phenotypic variance. Interestingly, in HG8 mice no differences in growth were detected; however, they were significantly leaner than controls (Table 5). In males a general decrease in fat pad mass, TF and AI was detected; however, only MFP was significant at a critical P < 0.005. Male AI was nearly significant at a nominal P = 0.0082. Larger decreases in fat mass were seen in HG8 females, with significant decreases detected for GFP, MFP and AI. Additionally, RFP and TF were nearly significant with nominal P values of 0.0208 and 0.0153, respectively. HG8 mice represent a lean congenic model in which CAST alleles protect against fat accumulation.

HG9
CAST alleles in the proximal region of MMU9 were previously found to be linked with an increase in body fat percentage and a decrease in femur length [24]. The effects appeared to be the result of two linked QTL, Carfhg2 at 10 cM and Feml2 at 20 cM, instead of one pleiotropically impacting both traits (Table 1). In support of this, the increase in fat was dependent on hg; however, differences in femur length were not [24].
In confirmation of Carfhg2 all measures of adiposity in HG9 mice were increased relative to control mice; except MFP in both sexes and BMI in males (Table 5). In males, GFP, RFP, FFP and AI were increased by 29.0%, 51.2%, 42.9% and 30.0%, respectively. In females these increases were much larger. The same measurements were increased by 68.8%, 99.4%, 69.7% and 57.3%, relative to HGC females. A 2.9% and 3.6% reduction in male and female NA, respectively, was also detected. This was expected since NA was previously found to be an excellent indicator of femur length (R2 = 0.88) in HG mice [27] and as noted above this region contains the Feml2 QTL [24]. These data confirm the effects of Carfhg2 on adiposity and Feml2 on length and indicate the isolated effects of Carfhg2 are more significant than originally observed.

HG11
In the original intercross CAST alleles near 50 cM on MMU11 were linked with a reduction in G29, carcass ash (ASH) and carcass protein (PROT) [24]. Therefore, we measured each carcass component using chemical compositional analysis in the HGC, HG11 and HG17 strains. Originally, Carp2 was dependent on the presence of hg and no significant sex effects were detected [24]. In contrast to those results, HG11 mice displayed significant differences in growth and obesity between the sexes. In general, males demonstrated a 5% increase and females a 5% decrease in all growth-related traits (Table 4). Most of these differences, however, were not significant at a critical P < 0.005, although almost all were at a nominal P < 0.05. Following the same trend, males displayed an increase in tail length and a small suggestive increase in NA (P = 0.06), while females displayed significant decreases in NA. In support of the sex-specific QTL effects strain by sex interactions (P < 0.04) were seen for the following growth traits: 3WK, 6WK, 9WK, G26, G29 and NA (data not shown). The basis of each interaction was an increase in the males and a decrease in females.
In our analysis of the HGC, HG11 and HG17 strains, concordant results were observed measuring adiposity using chemical lipid extraction and weighing fat pad mass. The phenotypic correlation between carcass fat (FAT) as a percent of the empty carcass weight (ECW; weight of total body (including organs) minus the head and gastrointestinal tract) (%FAT) determined by chemical analysis and AI, using data from all three lines (HGC, HG11 and HG17), was 0.92. These data suggest that individual fat pad dissection is an excellent proxy for measuring whole body adiposity and provides a much more sensitive technique to measure fat accumulation in specific body regions.
HG11 females displayed slight increases in all fat pads (except MFP) and AI, although, only the 16.5% increase in AI was significant (Table 5). In contrast, HG11 males did not show a difference in fat mass leading to a significant strain by sex interaction (P = 0.008) for AI (data not shown). Identical to the results obtained using fat pad weights; female HG11 carcasses displayed an increase in %FAT (Table 6).
HG11 male carcasses had higher levels carcass ash (ASH) as a percent of ECW (%ASH) and female carcasses displayed lower levels of H2O and ASH (Table 6). Strains by sex interactions (P < 0.005) were identified for H2O, %H2O, %FAT, ASH and PROT (data not shown). If the aggregate phenotype in HG11 is due to a single locus, its function may be to disrupt energy partitioning by decreasing the deposition of lean tissue and increasing lipid accumulation in females, while having the opposite action in males. However, it may also be due to distinct sex-specific QTL with opposing action.

HG17
In the original intercross CAST alleles at MMU17 markers were associated with decreases in femur length, ASH and PROT [24] and similar results were seen in HG17 mice. Interestingly, both sexes were heavier at 2WK (males, P = 0.0045; females, P = 0.0342), but lighter at 9WK, leading to substantially lower G26 and G29 (Table 4). In general growth differences ranged between 5% and 15% lower in congenic mice.
Both sexes had significant reductions in length traits relative to control mice. NA was reduced by 2.7% (P = 0.0186) and 3.4% in males and females, respectively (Table 4). In addition, tail was reduced by approximately 6% in both sexes (Table 4).
In general, HG17 mice displayed increases in all fat pads (except MFP), TF and AI (Table 5). However, none of these differences were significant at critical P < 0.005. In addition, HG17 male carcasses possessed lower levels of H2O and ASH, while female carcasses contained lower levels of % H2O, ASH, %ASH and %PROT (Table 6).

MMU2, 9, 11 and 17 hg modifier candidate gene sequencing
The HG phenotype is due to the deletion of Socs2, therefore we reasoned that QTL on MMU2, 9, 11 and 17 interacting with hg possibly represent variation within genes participating in various aspects of Gh function. To select candidate genes for sequencing we identified genes known to be or potentially involved in Gh signaling, are responsive to Gh or that propagate downstream Gh functions. Forty-four hg modifier candidate genes were identified from primary literature, reviews and book chapters and coordinated using the GenMAPP (Gene MicroArray Pathway Profiler) pathway building software (Figure 5). The coding region of each gene was sequenced from the CAST strain and compared to the publicly available B6 sequence to identify polymorphisms (Additional File 4). A total of 94.492 kbp was sequenced (75.378 kbp CDS and 19.114 Kbp 5' and 3' untranslated region (UTR)), representing 25,083 amino acids. Comparison with the public B6 assembly (May 2004 University of California, Santa Cruz (UCSC [28]) mm5 genome assembly, National Center for Biotechnology Information (NCBI) Build 33) identified 307 polymorphisms between CAST and B6. Of these, 295 were single nucleotide polymorphisms (SNP) and 12 were insertions or deletions in CAST (Additional File 4). All 12 insertions or deletions were located within 5' and 3' UTRs. Fifty-six nonsynonomous SNP (nsSNP) were identified in 14 different genes (Additional File 4). PolyPhen [29] and SIFT [30] are software programs designed to identify nsSNP which potentially alter protein function by evaluating evolutionary conservation at specific amino acid residues using a multiple sequence alignment of protein sequences homologous to the query. When applied to our data set PolyPhen, SIFT or both programs predicted that 15 of the 56 nsSNP in 9 different genes would possibly alter protein function (Table 7).