Quantitative Trait Loci for Femoral and Lumbar Vertebral Bone Mineral Density in C57BL/6J and C3H/HeJ Inbred Strains of Mice



Significant differences in vertebral (9%) and femoral (50%) adult bone mineral density (BMD) between the C57BL/6J (B6) and C3H/HeJ (C3H) inbred strains of mice have been subjected to genetic analyses for quantitative trait loci (QTL). Nine hundred eighty-six B6C3F2 females were analyzed to gain insight into the number of genes that regulate peak BMD and their locations. Femurs and lumbar vertebrae were isolated from 4-month-old B6C3F2 females at skeletal maturity and then BMD was determined by peripheral quantitative computed tomography (pQCT). Estimates of BMD heritability were 83% for femurs and 72% for vertebrae. Genomic DNA from F2 progeny was screened for 107 polymerase chain reaction (PCR)-based markers discriminating B6 and C3H alleles on all 19 autosomes. The regression analyses of markers on BMD revealed ten chromosomes (1, 2, 4, 6, 11, 12, 13, 14, 16, and 18) carrying QTLs for femurs and seven chromosomes (1, 4, 7, 9, 11, 14, and 18) carrying QTLs for vertebrae, each with log10 of the odds ratio (LOD) scores of 2.8 or better. The QTLs on chromosomes (Chrs) 2, 6, 12, 13, and 16 were unique to femurs, whereas the QTLs on Chrs 7 and 9 were unique to vertebrae. When the two bone sites had a QTL on the same chromosome, the same marker had the highest, although different, LOD score. A pairwise comparison by analysis of variance (ANOVA) did not reveal significant gene × gene interactions between QTLs for either bone site. BMD variance accounted for by individual QTLs ranged from 1% to 10%. Collectively, the BMD QTLs for femurs accounted for 35.1% and for vertebrae accounted for 23.7% of the F2 population variances in these bones. When mice were homozygous c3/c3 in the QTL region, 8 of the 10 QTLs increased, while the remaining two QTLs on Chrs 6 and 12 decreased, femoral BMD. Similarly, when mice were homozygous c3/c3 in the QTL region for the vertebrae, five of the seven QTLs increased, while two QTLs on Chrs 7 and 9 decreased, BMD. These findings show the genetic complexity of BMD with multiple genes participating in its regulation. Although 5 of the 12 QTLs are considered to be skeleton-wide loci and commonly affect both femurs and vertebrae, each of the bone sites also exhibited unique QTLs. Thus, the BMD phenotype can be partitioned into its genetic components and the effects of these loci on normal bone biology can be determined. Importantly, the BMD QTLs that we have identified are in regions of the mouse genome that have known human homology, and the QTLs will become useful experimental tools for mechanistic and therapeutic analyses of bone regulatory genes.


BONE MINERAL density (BMD), an accurate and precise measure of bone mass, has been identified in several epidemiological studies as being the single most important risk factor for osteoporotic fractures.(1, 2) Hence, research efforts over the past three decades have focused on defining major environmental and hormonal determinants of bone mass and its subsequent loss. Such factors include age, gender, calcium intake during adolescence and senescence, estrogen deficiency in both men and women, glucocorticoid excess, physical activity, and demographic factors.

Recently, it has become apparent from a series of longitudinal studies that BMD at any time is a function of the rate of acquisition of peak bone mass in late adolescence and the subsequent rate of bone loss during adulthood. Although environmental and hormonal factors affect bone turnover, 60-70% of final adult BMD is determined by the process referred to as “peak bone acquisition,”(3, 4) and it is this phase of skeletal accumulation, which occurs over a short time period during puberty in humans and other animals, that is influenced predominantly by genetic factors.(4–6) However, unlike the identification of established environmental risk factors for osteoporosis, progress in delineating specific heritable determinants of BMD has been relatively slow.

Several lines of evidence confirm the thesis that peak BMD is influenced strongly by heritable factors. Estimates of heritability for bone mass in humans have ranged from 40% to 90%, depending on the model. In general, twin studies have indicated that the proportion of variance of BMD accounted for by genetic factors approaches 80%.(4, 5, 7) However, this estimate may be exaggerated by the confounding influence of shared environmental factors that are common in monozygotic twins but more diverse for nontwins.(8) Indeed, sibling studies have yielded somewhat lower estimates of heritability of approximately 50-60%.(4) Moreover, there are certain to be gene by environment interactions and site specificity with respect to BMD measurements, which preclude absolute estimates of the overall genetic influence on bone mass from any population study.(9–11) Notwithstanding these concerns, it is apparent that peak BMD has a major genetic component in humans, and from a public health perspective, the significance of this is immense, because identifying these factors could have huge medical and social implications in terms of targeting susceptible individuals for simple preventive measures.

Animal models are critical for experimentally defining the genetic regulation of bone density. Inbred mice represent an animal model with a short lifespan, rapid generation time, and lower maintenance costs than other mammals. Mice of a given inbred strain represent unlimited numbers of genetically identical “twins” whose genes can be analyzed experimentally and whose environments can be controlled strictly. Equally important, each inbred strain is genetically different from every other inbred strain, making possible planned matings and studies of segregating genes. The mouse genome has become highly defined, especially with respect to protein and molecular polymorphic differences among the various inbred strains. This detailed mapping of chromosomes greatly facilitates rapid location of new genes. In addition, segments of many mouse and human chromosomes have been identified with homologous linked loci.(12) As the human and mouse genome sequences become available, identification and testing of candidate genes will be even more easily accomplished.

In this report, we present the quantitative trait locus (QTL) analyses conducted with B6C3F2 intercross progeny from normal progenitor inbred strains C57BL/6J (B6) and C3H/HeJ (C3H), which differ in BMD of femurs by 50% and of vertebrae by 9%.



The study used two inbred strains of mice1-B6 and C3H—previously shown to differ widely in BMD of femurs, with lesser differences found in tibias, vertebrae, and phalanges.(13) Mice were produced and maintained in our research colony under 14:10 h light/dark cycles. Females were housed in polycarbonate cages (51 in2) in groups of three to five on bedding of sterilized Northern White Pine shavings. Water was acidified with HCl to achieve a pH of 2.8-3.2 (to prevent bacterial growth) and was freely available. The diet used for all mice was autoclaved National Institutes of Health (NIH) 31 (6% fat diet, Ca/P of 1.15:0.85, 19% protein, vitamin, and mineral fortified; Purina Mills International, Richmond, IN, USA) and was freely available. Use of mice in this research project was reviewed and approved by the Institutional Animal Care and Use Committee of The Jackson Laboratory (Bar Harbor, ME, USA).

Progenitor B6, C3H, and their (B6 × C3H)F1 hybrid strain females used in these studies ranged in age from 1 to 12 months. Progeny for genetic analysis of BMD was produced by mating low BMD B6 females to high BMD C3H males and then intercrossing this B6C3F1 hybrid to produce F2 offspring. A total of 1012 F2 females were raised, 986 of which eventually contributed to the genetic analyses. The F2 females were analyzed at 4 months of age when acquisition of adult BMD was achieved. Body weights were recorded at necropsy, and partial carcass preparations were preserved in 95% ethanol as previously described.(13) Kidneys and spleens from each mouse were frozen in liquid N2 and stored at −60°C for later extraction of genomic DNA. Femurs were isolated and their lengths were measured by digital calipers (Stoelting, Wood Dale, IL, USA) before densitometry. The F2 males were not kept because of losses due to aggressive behavior when group housed.

BMD measurements by peripheral quantitative computed tomography

Isolated femora and lumbar vertebrae were assessed using peripheral quantitative computed tomography (pQCT; Stratec XCT 960M, Norland Medical Systems, Ft. Atkinson, WI, USA) as described previously.(13) Briefly, bones were isolated and stored in 95% ethyl alcohol (EtOH) until measured for bone parameters by XCT 960M. Thresholds of 1.300 attenuation units differentiated mouse bone from water, adipose tissue, muscle, and tendon; a threshold of 2.000 differentiated high-density cortical bone from low-density bone. Calibration of the densitometer was done with a set of hydroxyapatite standards (0.050-1.000 mg/mm3) yielding a correlation of 0.997 between standards and pQCT estimation of density. Daily confirmation of that calibration was confirmed using a phantom of known density. Precision of the XCT 960M for repeated measurement of femoral BMD was 1.2%. Isolated femora were scanned at 2-mm intervals over their entire lengths. Total and cortical BMD values were calculated by dividing the total or cortical mineral content by the appropriate bone volume and expressed as milligrams per cubic millimeter. The XCT 960M does not have sufficient resolution to resolve accurately trabecular bone volume; thus, such data are not presented. Femoral periosteal circumferences and cortical thicknesses were calculated at the mid-point of the diaphysis. The CV for progenitor strain femoral parameters at 4 months were (a) 2.8-3.1% for density, (b) 8.4-9.8% for mineral, (c) 7.0-7.4% for volume, (d) 2.1-3.6 for middiaphyseal periosteal circumference, and (e) 3.4-5.0% for middiaphyseal cortical thickness.

Isolated L5 lumbar vertebrae also were evaluated with the XCT 960M, with precision for repeated measurement of vertebral BMD of 1.4%. Vertebrae were scanned at 0.7-mm intervals along their anterior-posterior lengths. The total and cortical BMD values were calculated for entire vertebrae. For the L5 vertebrae, the coefficients of variation for progenitor strain parameters were (a) 3.4-4.0% for density, (b) 5.5-12.4% for mineral, (c) 4.8-15.4% volume, and (d) 10.5-17.2% for cortical mineral.

Genetic analyses

Genomic DNA was prepared by two methods. First, kidney samples were digested with proteinase K and extracted by chloroform/phenol. Second, 20-25 mg of kidney or spleen tissues were heated to 95°C in 0.5 ml of 50 μM NaOH for 10 minutes, and then pH was adjusted to 8.0 with 0.05 mM Tris-HCl. Genotyping of individual mouse DNAs was accomplished by polymerase chain reaction (PCR) using oligonucleotide primer pairs from Research Genetics (Birmingham, AL, USA). These primer pairs amplify simple CA repeated sequences of anonymous genomic DNA that are of different length and via gel electrophoresis can uniquely discriminate between B6 and C3H genomes. Primer pairs identifying simple sequence length polymorphisms between B6 and C3H were selected from more than 6000+ available.(14) Details of standard PCR reaction conditions have been described previously.(15) PCR products from B6, C3H, and (B6 × C3H)F1 hybrids were used as electrophoretic standards in every gel to identify the genotypes of F2 mice (that is, homozygous B6 [b6/b6] or C3H [c3/c3] and heterozygous [b6/c3]). Nine hundred eighty-six F2 progeny provided data for the genetic analyses.

All F2 progeny were tested for correlations of BMD data with segregation of 107 PCR-based simple sequence length polymorphic markers on the 19 autosomes. Four to nine polymorphic DNA markers, spaced at approximately 15 cM intervals from centromere to telomere, were selected for each autosome. The 15-cM genetic distance is easily capable of detecting major loci for bone density in this experimental design, given the large F2 population size. Chromosome (Chr) X alleles were not assessed because reciprocal F1 × F1 matings would be required to yield all possible allelic combinations necessary for genetic evaluation of the BMD phenotype in females.

Statistical analyses

Biological measurements:

Statistical analyses of progenitor and F1 hybrid femoral and vertebral data were performed with StatView 4.5 software from Macintosh (Cary, NC, USA). These data were analyzed first by analysis of variance (ANOVA) to detect major genotype effects. Individual group means were assessed for significant differences by Fisher's Protected least significant difference (LSD) test. Differences were judged statistically significant when p < 0.05.

Genomewide analyses:

The genome scans described in this study were carried out using a software implementation of the pseudomarker algorithm.(16) Software and additional details can be found at http://www.jax.org/research/churchill. The genome scans for QTL main effects produce log10 of the odds ratio (LOD) score curves identical to the standard MapMakerQT package(17) for a univariate normally distributed phenotype. Pairwise genome scans were carried out to search simultaneously for QTL pairs that were associated with the femoral and vertebral BMD traits. These scans allowed us to assess the possibility of gene × gene interactions, in which the combined effects of allelic substitutions at two loci are not equal to the sum of the two individual loci.(18)

Multiple regression:

We used a multiple linear model, based on marker regression, to assess the joint effects of all loci (and interactions) that appeared to be significant in the genomewide scan. Multiple regression models were analyzed using Minitab software (Minitab, Inc., State College, PA, USA).


Inbred and hybrid F1 mice

To define the optimum time at which to conduct a genetic analysis of peak bone density, the progenitor B6 and C3H strains, plus B6C3H-F1 hybrid progeny, were measured for their developmental patterns of femoral and lumbar vertebral BMD. The results presented in Fig. 1A indicated that C3H femoral BMD levels were consistently greater than those of the B6 mice. Both progenitor strains developed maximal adult femoral BMD at 4 months and then generally maintained those levels through 12 months of age. The F1 hybrids achieved their maximum femoral BMD at approximately 8 months, followed by a decline at 12 months of age. BMD levels in the F1 progeny were always higher than those of B6 and were lower than those of C3H from 4 months onward. In Fig. 1B, the lumbar vertebrae of C3H mice reached greater values than observed in B6 at 4, 8, and 12 months. Both progenitors and F1 females showed highest BMD levels at 4 months, followed by a significant decline in BMD levels by 12 months of age. In contrast with femoral BMD, the lumbar vertebral BMD levels were greater in F1 progeny at 2, 4, and 8 months than in either progenitor strain, following a remarkable growth spurt occurring between 1 and 2 months of age. Collectively, these results suggested that peak adult BMD in the femora and lumbar vertebrae were achieved at 4 months of age, making possible genetic analyses of peak BMD data gathered on both bone sites from the same mice.

Figure FIG. 1..

Acquisition of adult peak BMD in B6, C3H, and B6C3F1 (F1) females. Means ± SEM are presented from groups of 7-12 mice that donated femurs and L5 lumbar vertebrae at the indicated ages for pQCT assessment of BMD. (A) Femoral data; (B) vertebral data.

Descriptive statistics for the bones in the 4-month-old progenitor and F1 hybrid mice are found in Tables 1 and 2. After determination of a significant “F” for each main effect, comparisons were made between the progenitor means, and then the F1 hybrid means were compared with each progenitor mean. Inspection of B6 and C3H data in Table 1 shows that the body weights of the progenitors did not differ from each other, whereas the C3H femur lengths were significantly shorter by 1.9% than those of B6 mice. The mineral contents of the B6 progenitor mice were markedly reduced compared with those of C3H mice, whereas femur volumes did not differ between B6 and C3H progenitors. Differences in these components result in the characteristic low femoral density of the B6 strain compared with the C3H strain. At the middiaphyseal region, periosteal circumference was larger in B6 than in C3H mice, whereas cortical thickness and cortical density were less than those values in C3H mice. Body weights, femur lengths, and volumes of the F1 mice were greater than those of the progenitors. For other measures, except periosteal circumference, the F1 mice showed femoral values significantly greater than B6. When F1 data are compared with C3H, measures of bone size (length, volume, and periosteal circumference) were greater than C3H while mineral content, density, and cortical thickness were smaller than these values for C3H mice.

Table Table 1.. Data for Femurs from C3H, B6, and F1 Females Aged 4 Months
original image
Table Table 2.. Data for Entire Lumbar L5 Vertebrae from C3H, B6, and F1 Females Aged 4 Months
original image

Comparisons of the L5 vertebral data are presented in Table 2. Considering the B6 and C3H progenitors first, it can be seen that both the total mineral content and the volume of the L5 lumbar vertebrae of the B6 mice were lower than those of the C3H mice. Consequently, total lumbar vertebral density was significantly lower in B6 than in C3H; however, the percent difference was modest (∼9%) when compared with the percent difference observed for femurs (∼51%). As would be predicted, the cortical mineral content of the B6 vertebrae was significantly less than that of the C3H vertebrae. The F1 hybrid vertebral measurements were intermediate between the B6 and C3H progenitors, with the exception of BMD as discussed previously.

F2 progeny

BMD was measured by pQCT in femurs and lumbar vertebrae for the F2 progeny obtained from B6C3H-F1 intercross matings. nine hundred eighty-six F2 females contributed DNA that could be scored for the genetic analyses of the femoral BMD phenotype. The distributions of femoral and lumbar vertebral BMD data are presented in Fig. 2. The femoral BMD data in Fig. 2A approximate a normal distribution with a grand mean of 0.585 mg/mm3 (minimum = 0.451 mg/mm3; maximum = 0.751 mg/mm3). Heritability of femoral BMD was calculated to be 83%.(19) The position of mean values for the B6 and C3H progenitor strains, as well as the B6C3H-F1 parental mice at 4 months of age, are indicated by arrows at appropriate locations. It is noteworthy that the distribution did not reveal significant numbers of F2 progeny with femoral BMD values less than that of B6 (0.487 ± 0.005 mg/mm3) or greater than that of C3H (0.738 ± 0.006 mg/mm3) progenitor values.

Figure FIG. 2..

Distributions of femoral and vertebral BMD obtained from B6C3F2 progeny. (A) Femoral BMD in female mice at 4 months of age. Positions of mean BMDs for the inbred progenitors and F1 parentals are indicated by arrows, while the bell-shaped line depicts a normal distribution of data. (B) L5 vertebral BMD from the same F2 mice. Location of mean values for the progenitors are markedly different than for the femoral BMD shown in panel A.

In Fig. 2B, the distribution of vertebral BMD from 938 F2 females having intact L5 vertebrae (48 damaged at necropsy or during subsequent isolation) is presented. The positions of mean values for B6 and C3H progenitors as well as for B6C3F1 parental genotypes are indicated by arrows. The overall distribution of vertebral BMD in Fig. 2B was normal in shape, with a grand mean of 0.306 mg/mm3 (minimum = 0.195 mg/mm3; maximum = 0.416 mg/mm3). Heritability of L5 vertebral BMD was calculated to be 72%. In contrast with the femoral data in Fig. 2A, the B6 and C3H progenitor vertebral BMD means were close to each other in the central region of the F2 distribution. Some F2 progeny had vertebral BMD values that were lower than those of B6 progenitor, and there were many more F2 progeny with vertebral BMD greater than that of the C3H progenitor. These transgressive patterns of F2 values suggest that both B6 and C3H strains carry alleles for high and low lumbar vertebral BMD. The mean vertebral BMD for the F1 hybrids was greater, although not significantly so, than for the C3H progenitor.

Genetic analyses

QTLs with LOD scores indicating suggestive (>2.8) or significant (>4.3) linkage for the femoral and vertebral density traits were identified by the whole genome scans.(20) In addition, pairwise genome scans were carried out to detect gene × gene interactions. In Fig. 3, each chromosome is represented on the abscissa, and the percent of the BMD variance accounted for in the F2 population is presented on the ordinate. Dashed lines represent permutation analysis-based thresholds for genomewide significance at levels of p < 0.05 and p < 0.01, according to criteria defined previously by Churchill and Doerge.(21) In Fig. 3A, multiple regression analysis results show that the femoral QTLs on Chrs 1, 2, 4, 6, 11, 12, 13, 14, 16, and 18 collectively accounted for approximately 35.1% of the BMD variance in the F2 population. Similar analyses for vertebral BMD are shown in Fig. 3B where QTLs on Chrs 1, 4, 7, 9, 11, 14, and 18 account for 23.7% of the variance in vertebral density in the F2 population. The QTLs with the greatest effect on both femoral and vertebral variances were those on Chrs 1, 4, and 18. These accounted for 22.34% (femurs) and 16.24% (vertebrae) of the total variances for each skeletal site. The QTLs specifically associated with femurs (Chrs 2, 6, 12, 13, and 16) accounted for 8.36%, while QTLs specifically associated with vertebrae (Chrs 7 and 9) accounted for 4.31% of the F2 population variance by skeletal site.

Figure FIG. 3..

Genomewide scans for correlations of BMD and molecular markers on each of the 19 autosomes. (A) Percentage of femoral BMD variance accounted for by each marker tested on the 19 autosomes. Horizontal lines represent the 1% (upper; nominal p = 8 × 10−4) and 5% (lower; nominal p = 3 × 10−3) permutation thresholds correcting for multiple testing. (B) Percentage of vertebral BMD variance accounted for by each marker tested on the 19 autosomes; details of presentation same as for panel A.

Table 3 gives the Mit marker most highly associated with each BMD QTL, as well as the Mouse Genome Database (MGD) map position for that marker. In addition, the LOD score for each marker and the percent of F2 variance accounted for are presented for each femoral and vertebral QTL detected. Each BMD QTL is named sequentially by Chr number, beginning with Bmd5 on Chr 1.(22)Bmd1-4 were located in the B6CAST-F2 cross reported earlier.(15) A genomewide search for pairwise interaction effects on BMD between markers was carried out but failed to reveal significant interactions when the test statistics were compared with a genomewide permutation threshold.

Table Table 3.. Summary of Genomewide Scans of B6C3F2 Mice for Bone Density QTLs
original image

Interval maps with Mit markers, associated LOD scores, and minimum critical level (p = 0.05, dashed line) for three QTLs accounting for the largest variance in BMD are presented in Fig. 4. These QTLs affected both femoral (solid line) and vertebral (dotted line) BMD, as shown for Chrs 1, 4, and 18. The 95% CIs (femora = black bars; vertebrae = open bars) were estimated from a 1.5 LOD reduction from the peak LOD score generated by MapMakerQT. Overlapping CIs support the possibility that both femora and vertebrae are regulated by the same gene within the QTL region on the respective chromosomes. Two of the remaining 9 QTLs, Chrs 11 and 14, also had overlapping CIs for femoral and vertebral BMD. The interval map for femoral BMD on Chr 1 QTL suggested more than one locus may be present. Accordingly, we carried out additional analyses of F2 QTL models for Chr 1 but could not develop convincing statistical evidence for a second femoral QTL. Resolution of possible tightly linked QTLs will require further decomposition of this region via construction and testing of congenic strains. Finally, interval maps prepared for the remaining chromosomes showed that genetic regions containing BMD QTLs varied in size up to 35 cM of recombination distance.

Figure FIG. 4..

Representative interval maps for three major QTLs (Chrs 1, 4, and 18) affecting both femoral (solid line) and vertebral (dotted line) BMD. Intermarker intervals and chromosome lengths in recombination distances are scaled. Statistical analyses are presented as LOD scores calculated for molecular markers beginning with the centromeric end of each chromosome on the left and extending toward the telomeric end. The 95% CI derived from a change of 1.5× LOD unit is shown by a black bar for femoral BMD (Chr 1 = 12 cM; Chr 4 = 19 cM; Chr 18 = 9 cM) and by an open bar for vertebral BMD (Chr 1 = 22 cM; Chr 4 = 22 cM; Chr 18 = 19 cM). The horizontal dashed line indicates the critical LOD score of 2.8 for each QTL. Locus names, Bmd5, 7, and 16 are assigned to each QTL based on Table 3.

In Figs. 5A and 5B, we present the main effects of C3H (c3) alleles for four selected QTLs on femoral and vertebral BMD. These include the three major effect QTLs on Chr 1, 4, and 18, plus an illustrative site-specific QTL for each bone. Inspection of these figures shows that the c3 alleles for QTLs on Chrs 1, 4, and 18 are additive in action and the presence of one or more c3 alleles significantly increased BMD at both skeletal sites. We also found that c3 alleles on Chrs 6 and 12 significantly decreased femoral BMD, as exemplified by Chr 6 in Fig. 5A. Likewise, c3 alleles on Chrs 7 and 9 significantly decreased vertebral BMD, as exemplified by Chr 7 in Fig. 5B. All remaining QTLs for both skeletal sites had additive effects with c3 alleles increasing BMD (data not presented).

Figure FIG. 5..

The effects of C3H (c3) alleles at three major QTLs (Chrs 1, 4, and 18; Bmd5, 7, and 16) affecting both (A) femoral and (B) vertebral BMD. Lower case letters in each panel indicate significant difference at p < 0.01: a indicates mean differs from that for b6/b6 mice; b indicates mean for c3/c3 mice differs from that for b6/c3 mice. In addition, a representative QTL (Chr 6, Bmd8) for femoral BMD not detected in the vertebral genome scan is presented in panel A. Likewise, a representative QTL (Chr 7, Bmd9) for vertebral BMD not detected in the femoral genomewide scan is presented in panel B.


In this study, we found that heritability for femoral and vertebral BMD was high (83% and 74%, respectively), and that at least 10 QTLs for femoral and 7 QTLs for vertebral BMD had LOD scores greater than 2.8. These loci are distributed widely across the genome, and they accounted for 35% and 23%, respectively, of the variance in BMD among these B6C3F2 female mice. In addition, we have shown that the developmental patterns of peak bone acquisition in both progenitor B6 and C3H, as well as their F1 hybrids, are very similar for both bone sites, and that overall, the maximum differences in BMD values were evident at 4 months of age. Progenitor strain peak femoral BMD was maintained through 12 months of age, whereas vertebral BMD appears to decline after 4 months of age. Thus, we chose 4 months as the time to measure both femoral and vertebral BMD and to perform genetic analyses. The femoral F1 data were intermediate between those of C3H and B6, suggesting that heterozygosity for B6 alleles was suppressive for achievement of the high BMD observed in the C3H progenitor. For the vertebrae, F1 data were different from those of both progenitors with low BMD at one month and higher BMD at adulthood. This suggests that heterozygosity for alleles of bone density genes, rather than homozygosity for either progenitor allele, may result in reduced bone formation, relative to bone resorption, before the first month of age. This delay in bone acquisition was reversed by a remarkable increase in vertebral BMD between 1 and 2 months of age. At 4 and 8 months of age, the F1 BMD was significantly higher than that of the C3H progenitor strain. We also observed F1 values exceeding progenitor strain values for three additional measures: body weight, femoral length, and femoral volume. These phenotypes also have polygenic regulation with individual loci that are dominant or additive in their main effects (Beamer and Donahue, unpublished data, 1998). The hybrid vigor in these phenotypes observed in the F1 progeny simply represents the sum of these main effects when all loci are heterozygous.

Acquisition of a large sample size of one gender was undertaken in this investigation to allow an in-depth look at the genetic regulation of bone density in appendicular and axial bone sites. A single sex was examined for economic reasons as noted in the Material and Methods section. The outcome of the genomewide analyses, judged by percent variance accounted for, clearly revealed three QTLs with major effects on both femoral and vertebral BMD and several QTLs with minor effects that were either unique to one skeletal site or common to both sites. The data for allele effects indicated that C3H and B6 alleles at each locus interacted in an additive fashion. Although we were not surprised by the number of QTLs regulating BMD, the lack of evidence supporting the existence of gene × gene interaction was unanticipated. Thus, we conclude from these findings that loci contributing to BMD are relatively independent.

The total variance for BMD in the B6C3F2 population explained by the QTLs for either femurs or vertebrae was 35% and 23%, respectively. The genomewide analyses (Fig. 3) showed a few additional chromosomes with effects on BMD in which LOD scores were below the minimum critical value established by permutation analyses for statistical significance at p < 0.05.(21) Even if these chromosomes with small effects were considered to harbor putative BMD QTLs, they would offer very little additional contribution to the percent BMD explained in these genetic analyses. Several other sources contributing to the BMD variance are readily discernible. For example, the maternal environment, pre- and postpartum or both, could exert effects via number of pups in a litter, litter number (e.g., first litter vs. fourth litter), or even gender distribution within a litter. Recently, Reifsnyder et al.(23) showed that maternal effects via NZO alleles significantly increased body weight gain in backcross progeny, most likely through elevation of milk lipid levels.

It also is possible that genes regulating variation in other phenotypes could have secondary effects on BMD. In our B6C3F2 females, regression analyses revealed that body weight had a significant effect on both femoral and vertebral BMD (6.3% and 5.7%, respectively). Analyses in progress indicate four body weight QTLs (Chrs 1, 6, 17, and 18) that achieved LOD scores >2.8 (Beamer and Donahue, unpublished observations, 2001). The body weight QTLs on Chr 1, 6, and 18 overlap regions that contain QTL for BMD. However, QTL for the regions are quite large and it can only be speculated as to whether the QTL for both body weight and BMD are the same or different. Another genetically complex phenotype correlated with B6C3F2 mice is serum levels of insulin-like growth factor I (IGF-I), for which we have reported four QTLs (Chrs 6, 10, 11, and 15).(24) Chromosomes 10, 11, and 15 are not shared with BMD, whereas Chr 6 QTLs for serum IGF-I and BMD are shared. The implication is clearly that other phenotypes we are not measuring with X-ray attenuation may have significant roles in peak BMD.(26)

Another source of genetic variability could be Chr X. Our F2 progeny was derived strictly from intercrosses of (B6 × C3H)F1 female with (B6 × C3H)F1 male mice. This design yields the expected genotypic classes of b6/b6, b6/c3, and c3/c3 for autosomal loci, but only b6/b6 and b6/c3, not the c3/c3 genotype, for Chr X loci needed to unambiguously assign phenotype-genotype relationships between Chr X allelic combinations and BMD. From preliminary data analyzing strain distribution patterns for BMD among the 12 BXH recombinant inbred strains,(25, 26) we detected a suggestive association between BMD and the proximal third of Chr X(27) (Beamer et al., unpublished observations, 2001). Therefore, it is very likely that a QTL accounting for significant variance in BMD is located on Chr X. This possibility could be analyzed by intercrossing (B6 × C3H) F1 females with (C3H × B6)F1 males and combining evaluation of both sets of F2 progeny for BMD and Chr X allelic segregation.

Finally, our statistical analyses were capable of evaluating interactions between two genes, but not interaction among three or more genes. If several of the BMD QTLs are within the same biochemical pathway leading to the same cellular change (e.g., mineralization), this multicomponent interaction would remain undetected. Finally, social interactions among individuals sharing the same cage may lead to behavioral consequences that affect acquisition of peak bone mass. Thus, additional experimental work plus new statistical tools are needed to improve our understanding of all genetic factors regulating the BMD variance in this cross.

Defining the exact location of a QTL for BMD or any other trait is limited even in a very large F2 intercross population. Limitations include the number of polymorphic markers available, chromosomal structural features that affect genetic recombination, and random characteristics of meiotic events. Based on our data, 95% CIs for regions containing BMD QTLs ranged from 12 to 35 cM in genetic distance. Any of these regions could contain a single locus or a cluster of linked genes, in which their net effect represents the percent variance accounted for by a given QTL. Because of the large number of genes within our QTLs, it is possible that even regions with overlapping CIs (Fig. 4) may contain more than one BMD regulatory gene. One solution to this complex locus challenge is congenic strains with single BMD QTLs isolated in a common genetic background.(27) Such congenic strains and sublines thereof can be used to confirm the presence of a QTL, narrow the genetic distance containing the putative gene, and decompose the genetic complexity of the region, that is, determine if more than one gene contributes to the difference in BMD caused by distinct parental alleles. Congenic sublines also can be used as bioassays to narrow the QTL regions and to identify candidate genes. For example, if one congenic subline responds to an experimental manipulation, such as ovariectomy, and a second subline for the same QTL does not, then certain candidate genes for the BMD phenotype such as steroid regulatory genes could be included or eliminated for a given QTL region. This would also indicate more than one gene is present within the QTL region.

Although some regions of the genome are more or less gene rich, it is commonly estimated that there are up to 60-70 genes per cM of mouse chromosome recombinational distance. This large number of potential genes argues against speculation about candidate genes, although attractive candidates have been proposed from human association or sib-pair studies. These candidates yield proteins such as (a) calcium sensing receptor,(28) (b) bone osteocalcin,(29) (c) apolipoprotein E,(30) (d) IGF-I,(31) (e) interleukin-1 receptor antagonist,(32) (f) calcitonin receptor,(33) (g) vitamin D receptor,(34) (h) collagen Ia1,(35) (i) interleukin-6,(36) (j) high bone mass,(37) or (k) estrogen and androgen receptors.(38) It is the case that our QTLs for regulation of normal BMD reported here map to other chromosomal regions more than the human genes cited previously, suggesting new loci regulating BMD. Although we are reluctant to offer a selection of candidate genes for any of the BMD loci described in the B6C3F2 cross at this time, it is clear that candidates will become obvious as incipient congenic strains because several of these QTL regions manifest anticipated changes in BMD.(39)

In a previously reported F2 intercross between B6 and CAST/Ei, we found four QTLs for femoral BMD located on Chrs 1, 5, 13, and 15 that exceeded the threshold for statistical significance.(15) In concert with the B6C3F2 data in this report, the proportion of variance explained by the four QTLs was low (13.1%) and the data were statistically devoid of detectable gene by gene interaction. Two of the QTLs—Chrs 5 and 15—were not detected in the B6C3F2 data reported here. Therefore, different crosses will reveal both similar and different BMD QTLs because of differences in alleles carried by particular strains. This is supported by studies from other investigators using a variety of genetic models combined with different measurement methodologies to the study of density. Thus, the osteopenic senescence accelerated mouse (SAM)P6 plus the related normal P2, R1, and AKR inbred strains have been studied with F2 analyses by femoral Cortical Thickness Index(40) and by dual-energy X-ray absorptiometry (DXA) of the spine.(41) The BXD recombinant inbred strain set (26 strains) derived from an intercross between B6 and DBA/2J has been exploited by DXA for analyses of whole body BMD.(42) We have chosen to use F2 intercrosses: one between B6 and CAST(15) and a second between B6 and C3H strains—both assessed by pQCT. Accordingly, insights are emerging about BMD as a complex trait associated with a manageable number of genes amenable to study. Table 4 summarizes the available mouse data on (a) chromosomal locations with genomewide LOD scores suggestive of linkage (i.e., >2.8), (b) measurement techniques, (c) bone sites, and (d) homologous human chromosome locations. Given that CIs were not available for all data sets, it is uncertain how many QTLs or clusters of linked QTLs are represented by multiple markers such as on Chr 7. In general, the closer the markers are to each other, the more likely it is that the same locus is being detected. Collectively, the QTL data found by these groups indicate (a) some of the same loci are being detected by different methods in different crosses, (b) there are some loci with multiple alleles (and thus detectable in different crosses) and some loci with rare allelic differences (and not detectable in every cross), and (c) some loci are site specific.

Table Table 4.. Summary of QTLs for BMD in Mice from Four Different Laboratories
original image

The future of these studies lies in greater refinement of map positions, testing of candidate genes for BMD QTLs, sequencing of BMD genes, and prediction of BMD gene location in the human genome via linkage homology. This will in turn facilitate functional studies to define genes and their roles in bone biology and their potential as therapeutic targets. For example, the QTLs on mouse Chrs 1 and 4 are within regions that share linkage homology with human Chr 1. The importance of this relationship is illustrated by the studies of (a) Devoto et al.,(43) who reported a locus associated with human BMD linked to 1p36; (b) Reed et al.,(44) who described a syndrome of hypercalciuria and low bone mass linked to 1q25; and (c) Koller et al.,(45) who found a significant association in a large sib-pair study with spinal BMD linked to 1q21. Genome sequencing for Homo sapiens is nearly complete, and the B6 mouse sequence will be completed within the next year. Direct comparison of gene sequences from homologous regions will sharpen choice of candidate genes for testing and determining whether, in fact, genes regulating BMD are the same or different between the two species.

In summary, our genetic analyses have shown that (1) BMD is a polygenic trait with a substantial number of genes supporting this bone parameter, (2) the QTL alleles are additive in their actions with respect to BMD, (3) the adult peak bone density in a given strain is the net result of QTLs with both positive and negative effects on BMD, and (4) there appear to be QTLs acting on multiple bone sites, as well as QTLs with site-specific effects. These data show that mouse models with a skeletal phenotype such as congenic strains, induced mutation strains (transgenics, gene knockout(46)), and spontaneous mutant gene-bearing strains(12) are going to be powerful tools for biological investigation of gene effects, as well as for fine mapping and candidate gene testing.


The authors thank Dr. G. Cox and Dr. K. Johnson for critical review of this article. We appreciate the dedicated assistance provided by R. Donahue, V. Haynes, K. Leibwohl, T. Leidy, K. Smith, and C. Wishcamper in the pursuit of this work. The research was supported by grants from the NIH (AR43618 and CA34196) and the U.S. Army (DAMD17-96-1-6306).