Osteoporosis is a common disorder of the skeleton characterized by low bone mineral density (BMD) and structural deterioration of skeletal tissue, leading to an increased risk of fragility fractures. BMD, which can be measured by dual energy X-ray absorptiometry (DXA), is currently the best clinical predictor of future osteoporotic fracture risk.1, 2 However, BMD is a complex trait that is controlled by the interactions of many environmental factors with multiple genetic determinants, each with individually modest effects.3 Although recent reports show promise in identifying some of the genetic influences on BMD and bone strength in humans,4 this has proven to be a difficult undertaking in patient groups because of the heterogeneity of human populations. One approach to gain insights to help unravel this problem has been to exploit genetically tractable animal model systems to identify candidate genes for more focused human investigation.5–7 Although no animal model can duplicate all aspects of human osteoporosis, the characterization of individual genetic influences on specific traits, such as BMD, can be useful for subsequent study of their potential contribution to disease susceptibility in human patients.
Quantitative trait locus (QTL) mapping is a powerful method for identifying genomic regions that harbor genes (quantitative trait loci [QTLs]) involved in shaping complex phenotypes, such as BMD.6, 8 QTL analysis typically employs genetically heterogeneous populations derived from two or more highly inbred progenitor strains. Several investigators have used genome-wide linkage scans to search for QTLs associated with BMD in mice (reviewed in Klein5 and Xiong and colleagues6, 8), and several QTLs have been mapped to similar locations in the mouse genome in studies involving different murine strains,6, 9 thus lending support to the validity of this experimental approach.
We previously applied QTL analysis to a large population of male and female F2 mice derived from a cross between C57BL/6 (B6) and DBA/2 (D2) strains, and reported the identification of five genomic regions on chromosomes (Chr) 1, 2, 4, 7, and 11 that were linked to acquisition of whole-body BMD.10, 11 Here we expand these studies to examine femoral BMD, and find that it also is a polygenic trait in mice, which shares some QTLs with whole-body BMD but has others that appear distinct. Further analysis of the Chr 7 QTL after the generation of reciprocal congenic strains has allowed us to establish a functional relationship of this QTL with femoral bone density and strength between B6 and D2 founder mice, and to demonstrate the cell autonomous nature of the QTL both on osteoblast differentiation of mesenchymal stem cell progenitors in culture and on ex vivo bone growth. Additional studies suggest that Ltbp4, which encodes a binding protein for latent transforming growth factor β1,12, 13 is a strong candidate gene underlying the Chr 7 BMD locus.
Materials and Methods
All mice used in these experiments were bred under identical conditions at the Portland VA Veterinary Medical Unit from stock obtained originally from the Jackson Laboratory (Bar Harbor, ME, USA). Mice used for breeding were maintained for no more than three generations. After weaning, mice were housed in groups of two to five animals per cage, and were provided with rodent chow (Diet 5001; PMI Feeds, Inc., St. Louis, MO, USA) and water ad libitum, and were maintained in a 12-hour light-dark cycle at 21 ± 2°C.
B6D2 F1 mice were bred from Jackson Laboratory parental lines (C57Bl/6 (B6) females and DBA/2J (D2) males; B6D2), and intercrossed to generate a total of 1116 B6D2 F2 mice (589 female and 527 male). In order to attain significance (at p ≤ 5 × 10−5) of a given QTL at a minimal heritability of 4% and a power of 0.9, the minimal sample size was estimated to be ∼1000 (see Dupuis and Siegmund,14 Lander and Botstein,15 and Rabbee and colleagues16). The congenic strains used for these experiments were created in our animal colony over the past 5 years. Reciprocal congenic mice were generated with a 105-Mb region of chromosome 7 of the D2 genome introgressed onto a B6 background (B6.D2.Ch7) and vice versa (D2.B6.Ch7) by breeding mice heterozygous for markers flanking the chromosome 7 femoral BMD QTL (D7Mit340 and D7Mit1000) with the appropriate progenitor strain for 10 generations. During creation of these strains, femoral BMD was assessed in heterozygous congenic mice at generations N4 and N7 to insure capture of phenotypic effects. The congenic interval in both reciprocal mouse strains spanned a region of Chr 7 from 4.6 to 110.9 Mb, with the marker mostly tightly linked to femoral BMD residing at 27.5 Mb (D7Mit114). All mice were studied at 4 months of age when acquisition of adult bone mass is complete.17 Mice were euthanized by CO2 inhalation, and weighed to the nearest 0.1 g. The spleen and left femur were harvested immediately from each mouse. Spleens were frozen in liquid N2 and stored at −80°C for subsequent genomic DNA extraction. The left femora were wrapped in sterile gauze soaked in phosphate buffered saline, and stored at −20°C until subsequent analyses. All animal procedures were approved by the VA Institutional Animal Care and Use Committee, and were performed in accordance with National Institute of Health guidelines for the care and use of animals in research.
Murine skeletal phenotyping
Whole femoral bone mineral measurements were determined by peripheral dual energy X-ray absorptiometry (pDXA; PIXImus; GE-Lunar, Madison, WI, USA). Routine calibration was performed daily with a defined standard (phantom). Cortical femoral shaft bone geometry was examined with a desktop X-ray microtomographic scanner (SkyScan Model 1074; Skyscan, Aartselaar, Belgium). Images were analyzed with Optimas software (version 6.2; Media Cybernetics, Silver Spring, MD, USA). To determine femoral structural properties, the left femur was tested to failure by three-point bending on a high-resolution materials test apparatus (Model 4442; Instron Corp., Canton, MA, USA). Load and displacement data were recorded and failure load was determined using system software.
Genotyping by PCR
Genomic DNA was isolated from individual mouse spleens as described.10, 18 Mice were genotyped with microsatellite (simple sequence length) polymorphic markers chosen from the Mouse Genome Database (http://www.nih.gov/science/models/mouse/resources/mgd.html). PCR primers were purchased from Research Genetics (Hunstville, AL, USA), and DNA amplifications were performed with a Perkin-Elmer 9700 thermocycler (Perkin-Elmer Cetus, Branchburg, NJ, USA) using a 25-µL reaction volume (∼150 ng of genomic DNA, 264 nM of forward and reverse primers, 0.2 mM dNTPs, 1 U Taq polymerase, and 2.5 µL of GeneAmp 10× PCR buffer; all from Perkin-Elmer Cetus). Reactions included two initial denaturation steps of 5 minutes each at 95°C and 80°C, followed by 40 cycles consisting of 30 seconds each of 94°C, 53°C, and 72°C, and a final extension step for 10 minutes at 72°C. Products were detected after separation on 4% agarose gels and visualization under long-wave UV light after staining with ethidium bromide.
The following primer pairs used to amplify mouse genomic DNA for Ltbp4 were synthesized at the OHSU DNA Services Core: top strand: 5′-cagagggttttcgggagat-3′, bottom strand: 5′-cctgggtcgcacgcacaag-3′. DNA sequencing was performed by the OHSU DNA Services Core.
Reagents for cell-based studies
Fetal calf serum (FCS), alpha minimal essential medium (αMEM), Dulbecco's modified Eagle's medium (DMEM), phosphate-buffered saline (PBS), trypsin/EDTA, TRIzol Reagent, and the Superscript III first-strand cDNA synthesis kit were purchased from Invitrogen (Carlsbad, CA, USA). The BCA protein assay kit was from Pierce Biotechnologies (Rockford, IL, USA). Protease inhibitor and 4-nitro blue tetrazolium (NBT)/5-bromo-4-chloro-3′-indolyphosphate p-toluidine salt (BCIP) tablets were from Roche Applied Sciences (Indianapolis, IN, USA). Alizarin red, ascorbic acid, β-glycerol phosphate, type 1 collagenase, and sodium orthovanadate were purchased from Sigma-Aldrich (St. Louis, MO, USA). Okadaic acid was from Alexis Biochemicals (San Diego, CA, USA). Porcine TGF-β1 was purchased from R&D systems (Minneapolis, MN, USA). Immobilon-FL was from Millipore Corporation (Billerico, MA, USA). Rat Bone morphogenetic protein 2 (BMP2) was produced as described.19 AquaBlock EIA/WIB solution was from East Coast Biologicals (North Berwick, ME, USA). Primary antibodies were from the following suppliers: anti-phospho-Smad-5, anti–protein kinase B 2 (anti-Akt2), and anti-Runx2 (Cell Signaling Technology, Beverly, MA, USA); anti-phospho-Smad3, anti-Smad 2,3, and anti-Akt1 from Abcam (Cambridge, MA, USA); anti-α-tubulin from Sigma-Aldrich (St. Louis, MO, USA); anti-Dlx5 and anti-Smad5 from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Polyclonal anti-LTBP4 has been described20 and was a gift from Dr. Lynn Sakai. Secondary antibodies (goat-anti-rabbit immunoglobulin G [IgG]-IR800, goat anti-mouse IgG-IR680) were from Rockland Immunochemical (Gilbertsville, PA, USA). Other chemicals were reagent grade and were purchased from commercial suppliers.
Mouse bone marrow stromal cell and calvarial osteoblast cultures
Mouse bone marrow stromal cells (MSCs) were isolated from 12- to 14-week-old adult male B6 and B6.D2.Ch7 mice, as described,21 and were maintained in αMEM plus 10% heat-inactivated FCS at 37°C in humidified air with 5% CO2. Calvarial osteoblast-enriched progenitors were isolated from frontal and parietal bones of 1- to 3-day-old mice. Dissected bones underwent five successive digestions with Type I collagenase, with the last three fractions being used for culture, as described.22 To initiate osteoblast differentiation, confluent MSCs and calvarial osteoblast precursors were incubated in osteogenic media (OM: αMEM plus 10% FCS, 50 µg/mL ascorbic acid, and 10 mM β-glycerol phosphate) plus 200 ng/mL rat BMP2. OM with BMP2 was replaced every 48 hours for up to 12 days. For studying short-term signaling, confluent MSCs differentiated in OM with BMP2 for 3 days, were serum-starved for 18 hours, and TGF-β1 (1 ng/mL) was added in serum-free αMEM for the times indicated in the figure legends. RNA and whole-cell protein lysates were extracted and analyzed as described below.
Mouse metatarsal bone culture
Metatarsal bones were isolated from newborn B6 and B6.D2.Ch7 mice as described,19, 23 and were cultured in DMEM with 0.5% bovine serum albumin, 50 µg/mL ascorbic acid, 1 mM β-glycerol phosphate, and 100 µg/mL penicillin-streptomycin solution at 37°C in humidified air with 5% CO2 for up to 10 days. Images were captured with a Nikon DXL1200 camera (Melville, NY, USA) attached to a Leica MZ FLIII microscope (Wetzlar, Germany) at days 0, 4, 7, and 10. RNA and protein lysates were extracted on day 10 as described,23 and were analyzed as stated below.
Analysis of gene expression by RT-PCR
Whole-cell RNA (2 µg) from metatarsals, from adult mouse calvariae and the diaphysis of long bones (femur and tibia), and from MSC cultures was reverse-transcribed with the Superscript III first-strand synthesis kit using oligonucleotide (dT) primers in a final volume of 20 µL, followed by PCR with 1 µL of cDNA per reaction and previously published primer pairs for mouse Akt1, Akt2, Id1, JunB, Dlx5, Runx2, Smad7, osterix, osteocalcin, and S17.19, 21, 23 Other primers are as follows:
Mouse CTGF: sense strand, 5′-CCACCCCAAACCAGTCATAA-3′; anti-sense strand, 5′-TGCTGTGCAGGTGATAAAGC-3′
Mouse Type I collagen: sense strand, 5′-GACCCATTGACCTGAACCG-3′; anti-sense strand, 5′-TTCTTCTGGTCCTCGTGGTCTC-3′
BMP2: sense strand, 5′-TGGAAGTGGCCCATTTAGA G-3′; anti-sense strand, 5′-TGACGCTTTTCTCGTTTGTG-3′
BMP4: sense strand, 5′-TGATACCTGAGACCGGGAAG-3′; anti-sense strand, 5′-CTGCTCTTCCTCCTCCTCCT-3′
BMP7: sense strand, 5′-TACGTCAGCTTCCGAGACCT-3′; anti-sense strand, 5′-GGTGGCGTTCATGTAGGAGT-3′
BMP receptor Ia: sense strand, 5′-ATGCAAGGATTCACCGAAAG-3′; anti-sense strand, 5′-AACAACAGGGGGCAGTGTAG-3′
BMP receptor Ib: sense strand, 5′-GTACTGCAGGGCCACAATTT-3′; anti-sense strand, 5′-TCTTCCAGGCTCACGTGACT-3′
BMP receptor II: sense strand, 5′-GGGAGCACGTGTTATGGTCT-3′; anti-sense strand, 5′-CAGAAACTGATGCCAAAGCA-3′
noggin: sense strand, 5′-TGTGGTCACAGACCTTCTGC-3′; anti-sense strand, 5′-GTGAGGTGCACAGACTTGGA-3′
gremlin: sense strand, 5′-TGGAGAGGAGGTGCTTGAGT-3′; anti-sense strand, 5′-GTGAGGTGCACAGACTTGGA-3′
TGF-β1: sense strand, 5′-TTGCTTCAGCTCCACAGAGA-3′; anti-sense strand, 5′-TGGTTGTAGAGGGCAAGGAC-3′
Cycle numbers were within the linear range for each primer pair, and ranged from 22 to 30. Images were captured after agarose gel electrophoresis by visualization under long-wave UV light after staining with ethidium bromide, and were quantified by densitometry using ImageJ software (NIH; http://rsbweb.nih.gov/ij). Quantitative real-time PCR (qRT-PCR) was performed using a BioRad Chromo4 Real-Time PCR detection system (Hercules, CA, USA). Primer sets are as above for TGF-β1; others are as follows:
LTBP4: sense strand, 5′-CTATGGTGAAGCCTGGGGTA-3′; anti-sense strand, 5′-AGGGGTCGTAGG GTAGCACT-3′
S17: sense strand, 5′-TACACGCGTCTGGGTAATGA-3′; anti-sense strand, 5′-TAGGGCTGAGACCTCAGGAA-3′
Runx2: sense strand, 5′-CAGACCAGCAGCACTCCATA-3′; anti-sense strand, 5′-CAGCGTCAACACCATCATTC-3′.
Results were calculated using the 2(–ΔΔC(T)) method,24 with values normalized to levels of S17.
Protein extraction and immunoblotting
Whole-cell protein lysates were prepared as described.19, 21, 23 Protein samples (20 µg/lane) were resolved using SDS-PAGE, and after transfer to Immobilon-FL membranes, immunoblots were performed as described.19, 21, 23 using primary antibodies as follows: anti-Akt1, anti-Akt2, anti-phospho-Smad-3, anti-phospho-Smad5, anti-Runx2, anti-Dlx5, anti-Smad-2/3, and anti-LTBP4 at dilutions of 1:1000; anti-Smad5 at 1:2000; and anti-α-tubulin at 1:5000. Conjugated secondary antibodies were used at 1:5000 dilutions. Results were visualized and images captured using the LiCoR Odyssey Infrared Imaging System and version 3.0 analysis software (LiCoR, Lincoln, NE, USA), and quantified using ImageJ software.
Alkaline phosphatase and Alizarin red staining
Cells were stained for alkaline phosphatase enzymatic activity after fixation with 70% ethanol for 10 minutes, incubation with 500 µL of NBT/BCIP substrate solution for 20 minutes at 20°C, and washing with distilled water, as described.19, 21, 23 Images were captured using the LiCoR Odyssey. Alkaline phosphatase activity was quantified in cell lysates as described, using triplicate sample measurements.19, 21, 23 For detection of mineralization, cells were fixed in 70% ethanol for 10 minutes, and stained with 2% Alizarin red solution (pH 4.1–4.5) for 1 minute at 20°C, and images were captured with a Canon flat-bed scanner.19, 21, 23
Data analysis for animal studies
All F2 progeny were tested for correlations of BMD data with segregation of 115 PCR-based microsatellite markers distributed on the 19 mouse autosomes and the X chromosome. The B6D2 F2 genotyping data were analyzed by Map Manager QT (beta version 28; http://www.mapmanager.org) to determine the position of the peak likelihood for linkage (LOD score) and one LOD support interval.25 For the chromosome X linkage analysis, the backcross option in Map Manager QT was used to deal with the fact that female F2 mice can be B6B6 or B6D2 for markers on the X, but males can be only B6Y or D2Y. A gender difference was ascribed when the LOD score for a phenotype–genotype correlation differed by three or more LOD units between the two genders, which is approximately equivalent to p < 0.001. Individual p values for evaluations of each congenic strain versus the appropriate background strain were determined using Student's t test (one-tailed, because the direction of effect was already known). StatView statistical software for the Macintosh was used to perform all other statistical analyses.
The criteria for statistical significance for an F2 population were based on recommendations of Lander and Kruglyak26–28 and the permutation test,29 as implemented in the Map Manager QT program. The former presumes a near-infinite number of mice and markers in a full-genome search (p = 5 × 10−5 for significance), and the latter estimates an appropriate significance threshold based on a finite number of markers and the number of mice actually evaluated in our experiment (p = 2.6 × 10−4 for significance). Both are intended to establish thresholds that yield a 5% chance of even one false positive appearing anywhere in the genome. The difference between the two criteria relates to the fact that our search did not fully cover the mouse genome, and that we used finite numbers of both mice and markers. The permutation test has the advantage of not being sensitive to departures from a normal trait distribution, and is specific to the finite nature of the experiments reported here.
Data analysis for cell-based studies
Results from cell-based experiments are presented as mean ± SD. Statistical significance was calculated using unpaired Student's t test. Data are listed in each figure legend, and were considered statistically significant when p < 0.05.
Femoral BMD is a polygenic trait in laboratory mice
We scored DNA from 1116 F2 progeny obtained from B6D2 F1 intercross matings (589 female and 527 male) for genetic analysis of femoral BMD. Male B6D2 F2 mice weighed more than females (34.0 ± 0.2 versus 26.8 ± 0.1 g, p = 4.0 × 10−139) and had greater femoral BMD (58.3 ± 0.2 versus 54.7 ± 02 mg/cm2, p = 1.3 × 10−41). A relationship between body weight and femoral BMD was observed in both male and female F2 mice (males: r2 = 0.11, p = 1.5 × 10−14; females: r2 = 0.14, p = 1.9 × 10−21), but mathematically varying the body weight had little impact on the correlation. Thus, for simplicity we used unadjusted weights to correct individual F2 BMD values (WC-femoral BMD). In males and females, the distribution of WC-femoral BMD was continuous, inferring polygenic control of this skeletal trait in both genders (Fig. 1). Tests of skewness and kurtosis showed slight but significant departures from normality, and to address this concern we used the permutation test to establish significance thresholds, which are not sensitive to departures from a normal distribution.29 This modification had no impact on the significance of any of the QTLs reported here.
To identify the putative location of relevant WC-femoral BMD QTLs, we performed a whole-genome scan of the DNA of the 1116 B6D2 F2 male and female mice, using 115 selected microsatellite markers distributed across the 19 autosomes and the X chromosome at an average spacing of 14 cM. Interval mapping of the data identified four loci in which associations between genotype and WC-femoral BMD were present both in male and female mice. In addition, we found seven loci with gender specificity (Supplemental Table 1). In each instance, a large difference was seen in mean BMD values between mice homozygous for the B6 allele versus the D2 allele, with heterozygous mice being intermediate (no dominance). Loci affecting femoral BMD in both genders mapped to Chr 7 (nearest marker: D7Mit114), 8 (D8Mit113), 9 (D9Mit182), and 15 (D15Mit63), those with male specificity to Chr 6 (D6Mit55) and X (DXMit144), and those with female specificity to Chr 2 (D2Mit166), 4 (D4Mit48), 13 (D13Mit193), 14 (D14Mit142), and 19 (D19Mit16) (Supplemental Table 1). Of the 11 QTLs identified, a B6 allele contributed to the increase in femoral BMD at five loci (Chr 6, 9, 13, 15, and X), and a D2 allele at six loci (Chr 2, 4, 7, 8, 14, and 19).
We determined LOD scores for the identified WC-femoral BMD QTLs by analyzing combined male and female data for loci on Chr 7, 8, and 15, and using gender-separate data sets for loci on Chr 2, 4, 6, 9, 13, 14, 19, and X. As seen in Fig. 2, which presents results of an unconstrained model for QTL effects, in which no a priori assumptions were made regarding mode of inheritance, all 11 femoral BMD QTLs reached levels of statistical significance to support linkage (LOD ≥ 4.328). Although statistical guidelines have not been established yet for defining gender divergence of inherited traits, results with the loci on Chr 2, 4, 6, 9, 13, 14, 19, and X (LOD difference > 3) clearly support male- and female-specific linkage for some QTLs (Fig. 2, Supplemental Table 1).
Generation of congenic mouse strains
Identifying a gene underlying a QTL requires refining its position at a much higher chromosomal resolution that does initial detection. Congenic strains, in which a chromosomal region from a donor strain is introduced into a recipient of contrasting phenotype, offer a means to more precisely localize a QTL by constraining the critical chromosomal interval. We generated strains congenic for the femoral BMD QTL on Chr 7 by crossing the genomic interval harboring the QTL from donor D2 mice into B6 recipients (B6.D2.Ch7) and vice versa (D2.B6.Ch7) through a DNA marker-assisted breeding strategy involving 10 backcrosses with the respective background strains. The result is a strain that has a Chr 7 chromosomal segment from one progenitor strain superimposed onto the genetic background of the other progenitor. The congenic interval in both reciprocal mouse strains spanned a region of Chr 7 from 4.6 to 110.9 Mb, with the marker mostly tightly linked to femoral BMD residing at 27.5 Mb (D7Mit114).
Skeletal phenotyping of congenic mouse strains
Identifying a gene underlying a QTL requires refining its position at a much higher chromosomal resolution that does initial detection, and can be facilitated by generating congenic strains, in which a chromosomal region from a donor strain is introduced into a recipient of contrasting phenotype. Physiological analysis of the reciprocal congenic mice and their B6 and D2 background strains revealed no changes in overall viability, body weight, or femoral length (Table 1). Introgression of D2 Chr 7 alleles onto a B6 background resulted in a 7% to 10% rise in femoral BMD (B6.D2.Ch7 congenic mice, p < 0.05, Table 1), whereas reciprocal crosses led to an 8% to 12% decline in femoral BMD (D2.B6.Ch7 mice, p < 0.001, Table 1). Consistent with these changes in femoral BMD were increases in both femoral mid-shaft cortical thickness and resistance to fracture (failure load) in mice bearing the D2 allele, and comparable decreases in mice inheriting the B6 allele (Table 1).
Table 1. Analysis of the Reciprocal Congenic Mice and Their B6 and D2 Background Strains
Background (n = 9)
Congenic (n = 9)
Background (n = 14)
Congenic (n = 14)
Chr 7 Allele
Body weight, g
22.5 ± 0.4
22.8 ± 0.3
29.0 ± 0.5
28.8 ± 0.5
15.7 ± 0.1
15.8 ± 0.1
15.9 ± 0.1
15.9 ± 0.1
52.0 ± 0.9
55.8 ± 0.6
54.6 ± 1.1
60.2 ± 1.3
0.162 ± 0.003
0.180 ± 0.001
0.163 ± 0.003
0.180 ± 0.001
Failure load, N
17.7 ± 0.9
21.1 ± 0.7
19.1 ± 0.5
22.3 ± 0.5
Background (n = 10)
Congenic (n = 14)
Background (n = 17)
Congenic (n = 19)
n.s., not significant; BMD, bone mineral density; Ct.Th, cortical thickness.
Chr 7 Allele
Body weight, g
24.9 ± 0.8
22.3 ± 0.9
29.7 ± 0.5
29.0 ± 0.5
15.4 ± 0.1
15.0 ± 0.2
15.3 ± 0.1
15.2 ± 0.1
56.2 ± 0.9
52.0 ± 0.9
59.5 ± 0.6
53.0 ± 0.3
0.207 ± 0.002
0.196 ± 0.001
0.212 ± 0.003
0.196 ± 0.002
Failure load, N
19.2 ± 0.3
17.5 ± 0.6
19.9 ± 0.5
18.6 ± 0.3
Enhanced metatarsal bone growth in B6.D2.Ch7 congenic mice
Short-term culture of neonatal rodent metatarsal bones has been used to study bone growth and endochondral ossification,19, 23, 30, 31 and we have found that these bones increased in length by >35% in serum-free medium over a 10-day culture period, but that growth and mineralization were blocked by inhibition of the phosphatidylinositol 3-kinase (PI3-kinase)–Akt pathway.23 Evaluation of 10 metatarsals from newborn B6 and 10 metatarsals from B6.D2.Ch7 congenic mice revealed similar average length and width at the time of isolation (postnatal day 3), but after 4 days in culture in serum-free medium, the B6.D2.Ch7 metatarsals were significantly longer than the B6 background strain, with differences in growth rate and overall length being sustained for up to 10 days (Fig. 3A, B).
BMPs are secreted during endochondral ossification by osteoblasts and hypertrophic chondrocytes32, 33 and BMP-mediated signaling can drive osteogenesis in vivo and in vitro.34, 35 We tested the hypothesis that BMP actions were enhanced in B6.D2.Ch7 mice versus the B6 background strain by first examining metatarsal RNA for gene expression of components of BMP-activated signaling pathways. We found that transcripts for both Type I and Type II BMP receptors were found at equivalent levels in both mouse strains, as was mRNA for the ligand, BMP2, and the inhibitors, noggin and gremlin (Fig. 3C). In contrast, steady-state concentrations of BMP4 and BMP7 mRNAs were increased in metatarsals from B6.D2.Ch7 mice by 1.7-fold and 1.4-fold, respectively, and based on the extent of Smad 5 phosphorylation (increased by 1.8-fold), we infer that BMP-stimulated signaling was more active in B6.D2.Ch7 metatarsals than in B6 background mice (Fig. 3C, D). In addition, steady-state levels of mRNAs induced by BMP-stimulated Smads, including Id1, Smad7, and JunB, were up to 1.5 times higher in metatarsals from B6.D2.Ch7 congenic mice than from the B6 strain, as were transcripts exclusively expressed by differentiating osteoblasts (Type I collagen), or expressed in both hypertrophic chondrocytes and differentiating osteoblasts (Runx2) (Fig. 3E). Based on these results, the increased rate and extent of longitudinal growth of cultured metatarsals of B6.D2.Ch7 congenic mice compared with the B6 progenitor strain correlates with the elevated femoral BMD found in the former mice in vivo. Moreover, these observations suggest that the high BMD trait, which tracks with the Chr 7 D2 allele, is a consequence of a biologic difference intrinsic to skeletal tissue, and that it does not depend on responses to in vivo mechanical loading or to alterations in levels of systemic regulators of bone mass.
Enhanced osteogenic differentiation of bone marrow stromal cells from B6.D2.Ch7 congenic mice
We next tested the hypothesis that the Chr 7 QTL caused a cell autonomous increase in osteogenesis by evaluating osteoblast differentiation of femoral and tibial bone MSCs in primary culture from B6.D2.Ch7 congenic mice and the B6 progenitor strain. Incubation in OM plus BMP2 caused equivalent induction of Smad5 phosphorylation and expression of the mesenchymal cell transcription factor Dlx5, but led to twofold increased accumulation of the critical osteoblast transcription factor, Runx2 in cells isolated from B6.D2.Ch7 bone marrow (Fig. 4B). The higher levels of Runx2 were accompanied by more extensive osteogenesis, as measured by greater increases in the mineralization of extracellular matrix (Fig. 4C) and bone-specific alkaline phosphatase activity (1.7-fold at day 12 by quantitative assay, Fig. 4D). In contrast, the rate and extent of differentiation of calvarial osteoblast progenitors was identical in cells isolated from newborn B6 and B6.D2.Ch7 congenic mice (Fig. 4E). Thus, the increased femoral BMD and bone strength seen in B6.D2.Ch7 congenic mice correlates with enhanced osteogenic differentiation of osteoblast precursors obtained from the marrow of long bones but not from neonatal calvariae.
Identifying candidate genes for the Chr 7 QTL
As noted above, the congenic interval in reciprocal B6.D2.Ch7 and D2.B6.Ch7 mouse strains spanned a region from 4.5 to 110.9 Mb, with the marker mostly tightly linked to femoral BMD residing at 27.5 Mb (D7Mit114), which is near the locations of the genes for Akt2 (28.4 Mb) and Ltbp4 (28.1 Mb), as well as ∼150 other genes (± 2.5 Mb), including Tgfβ1 (26.5 Mb). The three Akts found in mammals (Akt1, 2, 3) comprise a highly conserved family of signaling enzymes that function downstream of class Ia PI3-kinases in pathways activated by many growth factors and hormones.36, 37 Recent studies from our laboratory have shown that targeted knockdown or knockout of Akt2 in osteoblast precursors inhibited osteogenic differentiation, whereas inhibition of Akt1 had a minimal effect.21 However, analysis of Akt2 gene expression from neonatal metatarsals and from bone marrow stromal cells showed no differences in steady-state mRNA levels between B6.D2.Ch7 and B6 mouse strains (Fig. 5A, C), and Akt2 protein abundance was identical in marrow stromal cells from each genetic background (Fig. 5B). Similarly, mRNA levels for TGF-β1 were identical in marrow stromal cells and in neonatal metatarsal bones between both mouse strains (Fig. 5A, C).
Ltbp4 is one of four mammalian Ltbp genes, which encode proteins that are part of the large latent complex for the growth factor, TGF-β.12 This multi-protein complex maintains TGF-β in an inactive form and directs it to the extracellular matrix.12, 13 Of the four LTBP proteins, LTBP4 uniquely binds only TGF-β1,12, 13 and natural null mutations in humans have been found to cause a multiorgan system developmental deficiency syndrome that involves the gastrointestinal and genitourinary tracts, and musculoskeletal and dermal tissue systems.38 However, there appeared to be no difference in expression of LTBP4 mRNA or protein between B6.D2.Ch7 and B6 marrow stromal cells (Fig. 5A, B), although a twofold increase in LTBP4 transcript abundance was detected in B6.D2.Ch7 metatarsals by qRT-PCR (Fig. 5C).
Gene expression also was evaluated in RNA isolated from adult calvariae and the diaphysis of femora and tibiae from B6 and B6.D2.Ch7 mouse strains. No differences were observed in steady-state levels of transcripts for Akt1, Akt2, or LTBP4 between B6 and B6.D2.Ch7 mice, or in several bone-enriched mRNAs, including Runx2, Type I collagen, and osteocalcin (Fig. 5D), although the abundance of Akt2 and LTBP4 mRNAs was higher in femora and tibias than in calvariae (Fig. 5D). In both tissues, levels of TGF-β1 mRNA were reduced ∼10-fold in B6.D2.Ch7 compared with control B6 mice (Fig. 5D).
DBA/2J and 129T2/SvEmsJ mouse strains have been found to encode polymorphic variants of Ltbp4 in which there is an in-frame 36-nucleotide insertion/deletion within exon 12 of the gene that adds or removes 12 codons, and thus changes the length of the proline-rich domain of the LTBP4 protein by 12 amino acids.39 Sequencing of genomic DNA showed that D2 mice encoded the deleted allele, B6 mice the encoded the inserted allele (Fig. 5E), and analysis of mRNA from bone marrow stromal cells from B6.D2.Ch7 congenic and B6 background strain mice revealed that the congenics exclusively expressed Ltbp4 transcripts for the deleted allele, whereas B6 mice expressed only mRNA with the insertion (Fig. 5F).
Reduced TGF-β1–activated signaling in bone marrow stromal cells from B6.D2.Ch7 congenic mice
We next assessed the hypothesis that the alteration in the LTBP4 protein isoform expressed in B6.D2.Ch7 congenic mice could lead to changes in TGF-β1–mediated signaling. Bone marrow stromal cells from both B6.D2.Ch7 congenic and B6 mice were incubated for 3 days in OM plus BMP2 to initiate osteogenic differentiation. Then, after serum starvation for 18 hours, these cultures were incubated with TGF-β1 or BMP2 (Fig. 6). The acute signaling response to TGF-β1 was substantially attenuated in B6.D2.Ch7 osteoblast precursors compared with cells from B6 mice, as indicated by a more transient duration of phosphorylation of Smad3 after growth factor treatment (Fig. 6B; quantified values of pSmad3/Smad2,3 were equal at 0 and 30 minutes after stimulation but were twofold higher in B6 at 120 and 150 minutes). Longer-term signaling responses also were diminished in B6.D2.Ch7 osteoblast precursors, as indicated by consistently smaller increases in expression of TGF-β and Smad-activated genes, including CTGF, Type I collagen, and JunB (Fig. 6C; values in B6.D2.Ch7 osteoblast precursors were 50% to 75% of B6). In contrast, acute signaling responses to BMP2 were comparable in osteoblast precursors from both mouse strains (Fig. 6D).
We have mapped a QTL for femoral BMD to chromosome 7 using reciprocal congenic strains derived from B6 and D2 backgrounds. The high BMD trait, that tracks with the D2 chromosome, exerts equivalent effects in male and female mice, and is also manifested by enhanced osteogenic differentiation of long-bone MSCs but not calvarial preosteoblasts in primary culture, and by increased ex vivo metatarsal bone growth. An insertion/deletion DNA polymorphism in Ltbp4 exon 12 that causes the in-frame removal of 12 codons in the D2-derived gene maps within 0.6 Mb of the marker most tightly linked to the QTL. LTBP4 is expressed in differentiating MSC-derived osteoblasts, and in long bones but minimally in calvariae of adult mice, and reduced responsiveness to TGF-β1 is observed in MSCs of mice homozygous for the D2 chromosome. Remarkably, TGF-β1 gene expression is also decreased in diaphyseal RNA from femora and tibias of congenic mice with the high BMD trait, and its diminished abundance may contribute additionally to the phenotype. Taken together, our results identify a potential genetic and biochemical relationship between decreased TGF-β1–mediated signaling and enhanced femoral BMD that is regulated by a variant or variants residing within the Chr 7 introgressed chromosomal region.
Many QTLs for BMD but few genes
BMD is a quantitative trait that is controlled by combinatorial interactions among multiple genetic and environmental factors, with the genetic components being responsible for most of the variability.3 Over the past decade numerous investigative groups have attempted to map genetic loci regulating BMD by identifying QTLs arising after interbreeding different inbred mouse strains (reviewed in Xiong and colleagues6, 8), and through more limited genome-wide association studies (GWAS) in human populations.4 Recent compilations of QTLs have counted more than 100 loci potentially contributing to BMD in mice,6, 9 and these map to all 19 autosomes and to the X chromosome,9 with over one-half being associated with femoral BMD and the others with either vertebral or whole-body BMD.9 In addition, analyses by comparative genomics have found that most of the human GWAS loci linked to BMD are located within the statistical confidence interval of the orthologous mouse QTL.9 However, despite these advances, very few BMD genes have been definitively characterized to date,9, 18, 40 demonstrating that much work needs to be done to develop effective strategies for identifying the specific genetic regulators of BMD, and for defining their modes of action. In this context, the recent characterization by DNA sequencing of the genomes of 17 strains of inbred laboratory mice41 will help establish the genetic basis of phenotypes, including those contributing to QTLs for BMD.
TGF-β, LTBP4, BMPs, and bone
The three TGF-β isoforms found in mammals are highly similar proteins encoded by paralogous genes.13 All three proteins are found in bone, with TGF-β1 being the most abundant species.13 TGF-β proteins exert complex influences on bone biology and physiology.13, 32, 42 They have been shown to increase bone formation in vitro through stimulation of osteoprogenitor proliferation, thus expanding the number of osteoblast precursors,43, 44 but appear to inhibit osteoblast differentiation and mineralization.45, 46 TGF-β also has been found to play a role in osteoclast–osteoblast coupling and thus in bone remodeling by facilitating recruitment of osteoblast progenitors to sites of bone resorption,47 and by stimulating production by osteoblasts of receptor activator of NF-κB ligand (RANKL),48, 49 a critical secreted factor that promotes osteoclast differentiation and function.50, 51 Accordingly, in mouse models, increased production of TGF-β in bone has led to osteoporosis and fragility,52, 53 and conversely, reduction in TGF-β–mediated signaling has caused enhanced bone mass and quality.54, 55 Our finding of reduced TGF-β1 gene expression in long bones from D2 congenic mice suggests one possible mechanism underlying the observed improvement in bone mass and strength in these mice compared to background B6 control mice.
However, we also observed diminished TGF-β1–mediated signaling in MSC-derived osteoblast progenitors from the congenic D2 mice. The decrease in the activity of exogenous TGF-β1 correlates with the presence of a LTBP4 protein that is missing 12 amino acids in its proline-rich domain. LTBP4, a multidomain secreted protein, interacts via its COOH-terminal cysteine-rich region with the NH2-terminal part of the TGF-β1 precursor, termed the latency-associated peptide,12, 13 to form the large latent complex that directs TGF-β1 to the extracellular matrix while maintaining it in an inactive form and shielding it from its signaling receptors.12, 13 The proline-rich segment in the NH2-terminal third of LTBP4 has been thought to be a target for proteases acting within the extracellular matrix that play a role in release of latent TGF-β1 during the steps that lead to its bioavailability for signaling.12, 13 Homozygosity for the same D2-derived deletion allele of mouse Ltbp4 that we have identified here has been associated with worsening muscle function and increased pathology in a mouse model of muscular dystrophy.39 That study found enhanced TGF-β1–mediated signaling in fibroblasts from affected mice carrying the deletion allele of Ltbp4,39 which is the opposite of what we show here in MSCs, in which the LTBP4 protein lacking 12 residues in its proline-rich domain was associated with a reduction in TGF-β1-stimulated Smad phosphorylation and a decrease in TGF-β1–activated gene expression (Fig. 6). We cannot explain the discrepancy between the two observations, except to note that the differences may reflect cell-type–specific or tissue-type–specific modifiers that may be distinct for muscle or bone.
The enhanced growth of neonatal metatarsal bones in short-term primary culture that we see in D2-congenic mice with increased BMD was also associated with higher expression of BMP4 and BMP7 mRNAs, and increased BMP-mediated Smad5 phosphorylation and BMP-activated gene expression (Fig. 4). Our results parallel observations using lung fibroblasts from mice engineered to lack LTBP4,56 in which enhanced BMP-mediated signaling was linked to upregulation of BMP4 mRNA and diminished expression of the BMP inhibitor, gremlin.56 The pathway by which loss or modification of LTBP4, which does not bind BMPs, regulates BMP gene expression and actions, presumably reflects altered crosstalk from decreased activity of TGF-β1, although the precise molecular mechanisms have not been elucidated.
Implications and limitations
As demonstrated here, congenic strains of mice are excellent tools for confirming the existence of an individual QTL, defining allele effects on phenotypes, and biological studies of QTL effects on cellular functions. However, hundreds of genes reside within the Chr 7 introgressed region (including Tgfb1 and Ltbp4). Future breeding strategies (eg, generation of subcongenic mice carrying small segments from the D2 Chr7 QTL region) will be necessary to refine the position of this QTL enough to identify the specific molecular basis (or bases) of its effect. Nevertheless, our studies here identify an association between altered growth factor signaling and inherited differences in bone mass and strength. Both TGF-β and BMPs play complex roles in bone development and physiology,32, 42 with TGF-β primarily being linked to reduced osteogenesis and bone mass and strength,13, 42 and BMPs being associated with enhanced bone development and function.42, 57 The identification of altered bone TGFβ1 mRNA expression and an insertion/deletion polymorphism in exon 12 of Ltbp4 in this unique mouse model provides an opportunity to examine the cellular and molecular basis of interactions between signaling pathways regulated by these growth factors in osteoblasts, osteoclasts, and their progenitors, as well as to investigate the biochemical basis by which LTBP variants influence the bioavailability and functions of TGF-β1 in bone. Interestingly, recent population studies have identified strong genetic associations in or around the genes for LTBP258 and LTBP359 with hip BMD variation and fracture. Further study also will be needed to confirm that Ltbp4 represents a BMD-modifying gene in mice and to discern if its human orthologue also modulates bone development or function.
All authors state that they have no conflicts of interest.
These studies were supported in part by National Institutes of Health grants R01 DK42748 (to PR), P50 AA10760 (to JKB), R01 AR44659 (to RFK), and the VA Medical Research Service (to RFK). We thank Dr. Lynn Sakai for polyclonal antibodies to LTBP4.
Authors' roles: AM performed, analyzed, and interpreted the in vitro data and wrote the manuscript. EAL, under the direction of RFK, carried out the congenic strain development, murine skeletal phenotyping, and data analysis. ASC, under the direction of RFK, performed the DNA sequencing and data analysis. JKB contributed to congenic strain development, data analysis, and manuscript preparation. PR directed the in vitro studies, interpreted results, and wrote the manuscript. RFK conceived of the study, designed the murine experiments, analyzed and interpreted the data, and wrote the manuscript.