From the Research Division, Joslin Diabetes Center and Department of Medicine, Harvard Medical School, Boston, Massachusetts
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
There is a worldwide epidemic of obesity and type 2 diabetes (1). This is being driven, at least in part, by the westernization of diet with high-fat and higher caloric intake (2). The impact of these changes in dietary pattern, however, varies dramatically depending on the genetic background of the population exposed, with higher rates of diabetes and obesity in Hispanic, black, and Native-American groups. Defining the background genes that modify this response is central to understanding the pathogenesis of these conditions.
Background genes play a major role in determining the phenotype of mice that are subjected to different diets (3,4), mice with naturally occurring mutations leading to obesity and diabetes (5,6), and mice that have been genetically altered by introduction of transgenes or knockout of endogenous genes (7,8). This complicates both the interpretation of genotype-phenotype relationships and the understanding of what is due to the primary genetic alteration and what is due to background genes or environmental modifiers (7,9).
Several studies have suggested that C57Bl/6 (B6) mice are "obesity prone" as well as "susceptible to insulin resistance and glucose intolerance," whereas other strains, such as C3H/He, 129/Sv, and A/J mice, are resistant to obesity and diabetes (3,4,1013). We have previously shown that mice with a double-heterozygous deletion of the insulin receptor and insulin receptor substrate-1 (IRS-1) become markedly insulin resistant and hyperinsulinemic on the B6 background and that >90% of the mice develop diabetes by 6 months of age, whereas on the background of 129S6/SvEvTac (129), <2% of the double heterozygote mice become diabetic by 6 months, and the mice develop only a mild hyperinsulinemia (14). Using an intercross (B6 x 129)F2 generation, a genome-wide scan of the IR/IRS-1 double heterozygous mice revealed a locus with linkage to hyperinsulinemia on chromosome 14, a locus linked to hyperleptinemia on chromosome 7, and two loci with linkage to hyperglycemia on chromosomes 14 and 12, respectively (15). It is interesting that the donor strain in all four cases was B6. Recently, Colombo et al. (16) showed that mice with lipoatrophic diabetes have less insulin resistance in muscle and more insulin-resistant livers on a B6 background than on an FVB background, possibly as a result of differences in hepatic triglyceride clearance, although the exact loci contributing to this difference have not been mapped.
In the present study, we explored in detail how high-fat dietinduced insulin resistance affects the B6 and 129 strains, as well as an F2 intercross between these strains, and asked whether the genes that control diet-induced insulin resistance are the same as those that control genetic insulin resistance and are dominant-acting genes and whether these genes could be identified using a combination of genetics and genomics.
![]() |
RESEARCH DESIGN AND METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
Phenotype analysis.
Weight, blood glucose, plasma insulin, and plasma leptin were measured at 0, 2, 6, 12, and 18 weeks. Fasted serum free fatty acids were measured at week 0 and week 18. Blood glucose values were determined from whole venous blood using an automatic glucose monitor (Glucometer Elite; Bayer, Mishawaka, IN). Insulin and leptin levels were measured in plasma samples by enzyme-linked immunosorbent assay using mouse insulin and mouse leptin as standards (Crystal Chem, Chicago, IL). Glucose tolerance was performed in the mice after the 18-week period (mice were 7 months old) by intraperitoneal injection of 2 g glucose/kg body wt after an overnight fast. Insulin tolerance tests were performed in random-fed mice by injection of 1 unit insulin/kg body wt.
Indirect calorimetry and activity.
Another set of B6 and 129 mice (n = 4 in each group) were maintained on low- or high-fat diet for 6 weeks. The mice were placed individually in indirect calorimetry chambers (Oxymax OPTO-M3 system; Columbus Instruments, Columbus, OH). After 48 h to allow for adaptation to the metabolic chamber, O2 consumption and CO2 production were measured every 30 min for 48 h and activity was measured as beam break counts. During the first 24 h, the mice had free access to food and water, and during the second 24 h, the mice had access to water only.
Genome-wide scan in 100 F2 intercross mice.
DNA was prepared from tail tips as previously described (15). The mice were genotyped using 114 polymorphic markers covering the 19 autosomal chromosomes and the X chromosome (Research Genetics, Huntington, AL) with a proposed average distance of 20 cM. Genotypes were scored using 4% agarose gels, and this was done blindly. Any ambiguous results were repeated in an independent analysis. Marker linkage maps were generated with our own data and the Celera database and compared with the public maps at the Mouse Genome Database. Reasonable agreement was found with these mouse databases, although the exact position of markers sometimes deviated slightly from the positions identified by genome-wide sequence analysis.
RNA isolation, cRNA preparation, array hybridization, and analysis.
Skeletal muscle, liver, and epididymal fat pads were removed from B6 and 129 mice at the age of 6 months after the mice were maintained on a standard diet that contained 21% of calories from fat, 22% from protein, and 57% from carbohydrates. Total RNA was extracted and purified with RNeasy (Qiagen, Chatsworth, CA), and a total of 25 µg pooled from two to three mice was used for cRNA synthesis (17). A total of 15 µg of cRNA was hybridized on Affymetrix (Santa Clara, CA) murine chips U74Av.2 that contain 12,488 genes or expressed sequence tags with 3045% present on the chips. Three to four chips were used for each strain, and the data were normalized to 1,500 using the GeneChip software MAS 5.0. Genes in Table 2 were obtained using Celera Discovery System (Rockville, MD).
|
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
The fact that the F2 mice show patterns of average weight gain almost identical to those of B6 suggests that B6 mice carry dominant genes in the control of body weight between these strains. In all cases, circulating leptin levels, which generally reflect adipose mass (22), paralleled the changes in body weight (Fig. 2A and B).
|
|
|
|
|
|
Our previous results have suggested a locus on chromosome 12 (D12Mit231) with linkage to hyperglycemia and another on chromosome 7 linked to leptin levels (15). In the present study, there was a QTL on chromosome 12 (D12Mit231) with linkage to leptin at week 2 in mice that were fed a low-fat diet (peak LOD score of 3.2; Fig. 6C). Surprisingly, mice homozygous for the 129 allele exhibited 1.9-fold higher leptin levels than B6 homozygous mice (P = 0.005; Table 1), whereas the previously identified locus on chromosome 7 linked with hyperleptinemia (15) was not statistically defined in these mice during diet-induced insulin resistance.
Five other QTLs were identified in the genome-wide scan (Table 1). These included a hyperleptinemia locus on chromosome 10 (Fig. 6D), and on chromosome 3, we identified two separate loci linked with hyperinsulinemia and hyperleptinemia in mice that were fed a low-fat diet for 12 weeks (Table 1, Fig. 6E). Insulin and leptin levels, respectively, were significantly higher in mice homozygous for the B6 alleles at these markers (P = 0.002). On chromosome 11, a locus was significantly associated with insulin after 6 weeks on high-fat diet (Fig. 6F), and finally on chromosome 15 (Fig. 6G), we identified a locus associated with fasting plasma blood glucose.
Expression analysis of genes located in identified chromosomal areas.
To determine whether any of the annotated genes located in the chromosomal areas with QTLs for different phenotypes (Table 1) were differentially expressed between the B6 and 129 strains, we collected skeletal muscle, liver, and epididymal fat pads from 6-month-old mice that had been maintained on a standard diet containing 21.6% fat, extracted RNA, and performed Affymetrix microarray analysis. All of the genes located within 3 Mb of the marker with the highest LOD score (as shown in Table 1) that were present on the murine chip U74Av.2 are presented in Table 2 in the order of location on the chromosome. The interval of ±3 Mb was arbitrary but was chosen to focus on the differentially expressed genes closest to the LOD peak and probably represents approximately the 50% confidence interval for each QTL. Differential expression between B6 and 129 was observed of more genes than expected in the region of several of the chromosomal markers with significant LOD scores. For example, in the region within 3 Mb of D14Mit52, there were 23 annotated genes represented on the U72Av2 chip. Examining expression of these genes in liver, muscle, and fat demonstrated 11 differences in expression that were significant at the P < 0.05 or greater level. By random chance, one would have expected three or four significant differences at this level out of the 69 measured variables in the region. There were several interesting candidate genes among these 11, including the gene encoding PKC, which showed 2257% upregulation in B6 mice in all examined tissues, with the largest change in adipose tissue reaching statistical significance. Wnt5a is also located under that peak and was expressed at lower levels in all tissues obtained from the B6 strain, with a significant decrease of 78% in skeletal muscle. Other potential candidate genes based on expression differences include nuclear receptor coactivator A; calcium channel, voltage-dependent,
2/
subunit 3; and transketolase, all of which were lower in skeletal muscle of B6 mice as compared with 129. Inter-
trypsin inhibitor, heavy-chain 3, NIMA (never in mitosis gene a)-related expressed kinase 4, and stabilin were all lower in livers of B6 mice compared with 129, whereas transketolase was more highly expressed in this tissue in the B6 strain. The PHD (plant homeo domain) finger protein 7 was significantly higher in skeletal muscle of B6 mice.
A similar enrichment of differentially expressed genes was observed in the region of marker D14Mit192, where there were seven significant changes at the P < 0.05 level of 36 measured defined genes. These included a 2.6-fold higher level of expression of the tyrosine kinase substrate Dok2 in fat of B6 over 129 mice, as well as bone morphogenetic protein 1 and esterase 10. Significant differences in fat of 129 over B6 included erythrocyte protein band 4.9 and protein phosphatase 3, and the latter gene was also upregulated in skeletal muscle, whereas the expression of SH3-domain protein 4 was higher in muscle of B6 than 129. The large number of changes on chromosome 14 between these two strains of mice suggests the action of some global cis-acting regulator of gene expression, such as altered patterns of DNA methylation, chromatic binding, or chromatic structure in these regions between these two strains of mice.
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|
We find that B6 mice are prone to becoming more obese on either high- or low-fat diets as they age compared with 129 mice. Although 129 mice are relatively lean, they are in fact highly susceptible to the content of fat in the diet, with a greater percentage increase in weight and greater deterioration in glucose tolerance than the B6 on a high- versus low-fat diet. Thus, mice of this strain are not obesity resistant, as previously described (4,13), but exhibit a complex interaction among genes, diet, age, and environment. In addition, we find that the higher body weight and tendency to glucose intolerance of the B6 mice are dominantly inherited when these two strains are crossed.
The most obvious potential explanation for differences in body weight would be differences in food intake. Surprising, however, is that 129 mice ate significantly more calories than B6 on the low-fat diet and an equal number of calories on the high-fat diet but still remained leaner than their B6 counterparts. Thus, the B6 mice showed higher feeding efficiency, i.e., greater weight gain per calorie consumed, as compared with 129 mice. This would represent an example of the "thrifty gene" phenotype and represent an advantage during times of food shortage (27).
In the absence of excessive food intake, low energy expenditure would be the major potential risk factor for body weight gain (28). Although B6 mice are actually more active than 129 mice, B6 mice have a lower metabolic rate than the 129 during feeding on both low- and high-fat diets and also during fasting on the high-fat diet. B6 mice also exhibit a smaller change in metabolism in response to a shift from a fasted to a fed state than 129 mice and a smaller change in response to the shift from a low- to a high-fat diet, i.e., reduced diet-induced thermogenesis. Whether the higher basal metabolic rate and higher diet-induced thermogenic response in 129 mice accounts entirely for the diminished weight is hard to predict. It is interesting that studies have shown that diet-induced thermogenesis is higher in lean humans compared with obese individuals (29). Furthermore, the higher metabolic rate in the 129 mouse in the basal state is independent of activity level, because these mice show approximately a 50% lower level of activity than the B6.
Besides a low metabolic rate, a low-fat versus carbohydrate oxidation ratio, i.e., a high respiratory quotient, has been suggested to be a risk factor for body weight gain (23,30,31). Contrary to this dogma, B6 mice exhibit a slightly lower respiratory quotient than the 129 mice.
An increase in body fat causes an increase in circulating leptin levels, which will normally decrease the energy intake and increase energy expenditure (32), yet in the B6 mice, the ratio of leptin to body weight and fat mass was much higher than that in the 129 mice. Thus, the B6 mice develop leptin resistance with age and obesity, mimicking humans with diet-induced obesity (15,33). It is interesting that leptin mRNA on high-fat diet is similar in B6 and 129 mice despite the differences in circulating leptin (manuscript in preparation), suggesting differences in either leptin translation efficiency or protein half-life between strains.
In parallel with the higher body weight, the intraperitoneal glucose tolerance testing showed greater hyperglycemia and little diet-induced change in the B6 mice, whereas 129 mice were glucose tolerant on the low-fat diet but intolerant on the high-fat diet. Likewise, the B6 mice exhibited insulin resistance during the insulin tolerance test on both the low- and high-fat diets. In contrast, 129 mice were relatively insulin sensitive even on a high-fat diet. The development of glucose intolerance in 129 mice on a high-fat diet is in contrast to our previous studies that have shown that 129 mice do not develop glucose intolerance when subjected to genetically induced insulin resistance as a result of heterozygosity of IR and IRS-1 (14). The B6 mice, conversely, become insulin resistant and hyperinsulinemic on either diet and also are more prone to become diabetic when they carry the double heterozygous deletion for IR and IRS-1 (14).
Characterization of the (B6 x 129)F2 intercross suggests that the differences between B6 and 129 mice are primarily determined by dominant-acting genes of the B6 mouse (24). In particular, we found that a locus on chromosome 14 between markers D14Mit55 and D14Mit52 previously identified as a modifier of genetically induced insulin resistance (15) also plays an important role in response to dietary insulin resistance.
Of the potential candidate genes in this area that are represented in the Affymetrix microarray, a disproportionate number of genes exhibited differential expression between B6 and 129 mice. These included several genes that are known to affect metabolism, such as PKC, Wnt-5a, and transketolase. PKC
is one of several serine-threonine protein kinases that has been implicated in action, activity, and intracellular trafficking of the insulin receptor (34) and has been shown to be persistently activated in diabetes (35). Recent studies have also suggested an involvement of PKC
in promotion of apoptosis (36), and it has been shown that increased glucose uptake promotes oxidative stress and PKC
activation in adipocytes of B6 mice (37).
Wnt-5a is a member of a family of paracrine and autocrine factors that regulate cell growth and cell fate (38). Wnt signaling prevents preadipocytes from differentiating into adipocytes (38), and this could fit with the findings of the present study, showing that Wnt expression is upregulated in skeletal muscle, liver, and fat of the lean 129 strain as compared with the obese B6, in some cases by almost fivefold.
The expression level of transketolase was also significantly different between B6 and 129 mice; however, in this case, the expression was significantly higher in skeletal muscle of 129 versus B6, whereas it was significantly lower in liver of 129 versus B6, and the expression was similar in white fat. Transketolase is a ubiquitous enzyme involved in multiple metabolic pathways. Haploinsufficiency of transketolase results in reduced adipose tissue mass (39), whereas the deletion of PKC and Wnt-5a seems to affect ß-cell proliferation and stimulate hematopoietic malignancies, respectively (40,41). Thus, future studies of the knockouts of each of these candidate genes may help to determine their roles in metabolism.
A separate QTL at marker D14Mit192 was identified with linkage to both weight gain and hyperglycemia and inherited by the B6 strain. On the basis of differential expression, Dok2 (also known as p56[Dok2], FRIP, or Dok-R) is an interesting candidate gene located in the area around marker D14Mit192, which is highly upregulated in epididymal fat of B6 mice in comparison with 129. Dok2 is a RasGAP-binding protein suggested to have an important role downstream of growth factor receptors as a potential negative regulator of signal transduction (42). Additional studies will be needed to determine whether this gene or other genes revealed in Table 2 play a role in the development of hyperinsulinemia or insulin resistance. Furthermore, the list of genes under the peaks given in the Table 2 may potentially be the quantitative trait genes; however, several unidentified candidate genes may still be discovered in the areas.
A QTL with linkage to leptin at week 2 was detected around marker D12Mit231 and is inherited from the 129 genome. The locus overlaps the hyperglycemia locus that was identified in the insulin receptor/IRS-1 double-heterozygous mice that interacts significantly with a locus on chromosome 7 in the control of leptin (15). In the New Zealand Obese mouse, this same marker (D12Mit231) has been shown to affect adiposity, BMI, and leptin (43) and is a diabesity locus when interacting with several other loci (44). Loci linked with hyperleptinemia were also discovered on chromosomes 3 and 10 in the present model. Because of the low number of markers in the area of the QTL on chromosome 10 as well as chromosome 15, there is some uncertainty about the significance of those QTLs.
Reed et al. (45) performed an analysis of an F2 intercross of C57BL/6ByJ and 129P3/J and identified QTLs for body weight, length, and adiposity on chromosome 2. We did not identify any significant QTLs on chromosome 2. This discrepancy may be due to small differences of the inbred mouse substrains being used in the two studies (Jackson versus Taconic, 129Ps/J versus 129S6/SvEvTac, and C57Bl/6 versus C57BL/6ByJ), differences in dietary fat intake (4.4 vs. 9% fat) that were used to provoke the insulin-resistance phenotype, and differences in the age at the time of study or time course of study or other unidentified variables.
Combining the mapping data with the microarray expression profile provides an opportunity to establish a potential link between genotype and phenotype and thereby improve the chances of finding new diabetogenes and pathways. If the link between QTL and gene expression can be proved directly, then this would suggest a cis-acting gene, i.e., one in which a DNA variation in the gene itself (most likely the promoter) controls the expression level. As more data on the complete sequence of these candidate genes in both strains of mice become available, this hypothesis can be tested. However, a LOD score can also be due to a trans-acting effect, i.e., a gene at the defined locus acts on other genes to affect their transcription. It has previously been suggested that it is easier to detect effects that are caused by DNA variation acting on the gene itself (46), and this is also suggested by our data as a result of the high number of genes with differential expression at the QTLs (Table 2). For example, around marker D14Mit192, 11 more genes than expected for a P = 0.05 were significantly differentially expressed in B6 versus 129 mice. Furthermore, at the other identified locus on chromosome 14 (D14Mit52), 23 annotated genes were present on the chip, and 11 of those genes were differentially expressed (P < 0.05) in muscle, fat, or liver of B6 as compared with 129, which is
9 more genes than expected by random chance alone. Thus, the QTL and gene expression data support each other and suggest that genes located in the identified loci do play important roles in the outcome of a trait. At this point, however, we cannot determine whether this is the effect of only one or all three candidate genes at the QTL on chromosome 14 (at D14Mit52) or some other gene not yet mapped that is responsible for the hyperinsulinemia/insulin resistance phenotype. Moreover, these multiple changes at individual loci suggest that methylation, chromatin structure, or histone binding may be altered in one strain or the other and consequently affect the gene expression (47,48).
In conclusion, B6 mice are prone to becoming obese, hyperleptinemic, insulin resistant, and glucose intolerant independent of the fat content in food, whereas 129 mice exhibit more modest weight gain and glucose intolerance only on a high-fat diet. This is due to higher caloric efficiency and reduced diet-induced thermogenesis in the B6 mouse, two manifestations of the "thrifty gene." This "thrifty" phenotype with weight gain, hyperinsulinemia, and hyperleptinemia is dominantly inherited when these strains are bred. QTL and gene-array expression data suggest that some of the major genes that control hyperinsulinemia/insulin resistance both as a result of genetic modification and obesity are located on murine chromosome 14 in the region of marker D14Mit52.
![]() |
ACKNOWLEDGMENTS |
---|
We thank Scott Lannon, Laureen Mazzola, Katherine Gartrell, and Pei Lin for technical help and Drs. Andy Norris, Atul Butte, and Isaac Kohane for useful discussions.
![]() |
FOOTNOTES |
---|
Address correspondence and reprint requests to C. Ronald Kahn, MD, Research Division, Joslin Diabetes Center, One Joslin Place, Boston, MA 02215. E-mail: c.ronald.kahn{at}joslin.harvard.edu
Received for publication March 3, 2004 and accepted in revised form August 23, 2004
IRS-1, insulin receptor substrate-1; LOD, logarithm of odds; LRS, likelihood ratio statistic; PKC, protein kinase C; QTL, quantitative trait locus
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() |
---|