Full text of DOI:10.1359/jbmr.2002.17.8.1527

JOURNAL OF BONE AND MINERAL RESEARCH
Volume 17, Number 8, 2002
© 2002 American Society for Bone and Mineral Research
Alleles of RUNX2/CBFA1 Gene Are Associated With
Differences in Bone Mineral Density and Risk of Fracture
TANYA VAUGHAN,1 JULIE A. PASCO,2 MARK A. KOTOWICZ,2 GEOFF C. NICHOLSON,2
and NIGEL A. MORRISON1
ABSTRACT
The aim of this study was to determine if DNA polymorphism within runt-related gene 2 (RUNX2)/core
binding factor A1 (CBFA1) is related to bone mineral density (BMD). RUNX2 contains a glutamine-alanine
repeat where mutations causing cleidocranial dysplasia (CCD) have been observed. Two common variants
were detected within the alanine repeat: an 18-bp deletion and a synonymous alanine codon polymorphism
with alleles GCA and GCG (noted as A and G alleles, respectively). In addition, rare mutations that may be
related to low BMD were observed within the glutamine repeat. In 495 randomly selected women of the
Geelong Osteoporosis Study (GOS), the A allele was associated with higher BMD at all sites tested. The effect
was maximal at the ultradistal (UD) radius (p ؍ 0.001). In a separate fracture study, the A allele was
significantly protective against Colles’ fracture in elderly women but not spine and hip fracture. The A allele
was associated with increased BMD and was protective against a common form of osteoporotic fracture,
suggesting that RUNX2 variants may be related to genetic effects on BMD and osteoporosis. (J Bone Miner Res
2002;17:1527–1534)
Key words: bone mineral density, RUNX2, fracture, alleles, osteoporosis
analysis(7–10) tests the involvement of a locus in controlling
BMD directly by ANOVA or association using available
INTRODUCTION
ONE MINERAL density (BMD) is determined by a com-
bination of environmental and genetic factors.(1) Be-
polymorphisms and appropriate patient cohorts. As human
genome data become more comprehensive and dense single
nucleotide polymorphism (SNP) maps become available,
candidate gene analysis may become the preferred means of
identifying BMD-related genes.
This study addresses an essential candidate gene that has
not been investigated previously in relationship to BMD and
osteoporotic fracture. The osteoblast-specific transcription
factor known officially as Runx2 (protein) or RUNX2 (gene)
and variously as osteoblast-specific factor (Osf-2), poly-
omavirus enhancer binding protein 2␣A (PEBP2␣A), and
core binding factor ␣ (Cbfa1) is a “runt” domain transcrip-
tion factor(11) that is essential for osteoblast differentia-
tion.(12) RUNX2 gene knockouts result in animals with no
B
cause BMD is a strong correlate of osteoporotic fracture,
genes that influence BMD also may contribute to fracture
risk. Because up to 40% of postmenopausal women suffer
from at least one osteoporotic fracture, identifying any
genetic relationship with fracture may be important for a
large number of patients.(2) Whole genome analyses in
animal models of BMD(3,4) use mapping of extreme traits.
Genome scans in human studies(5,6) have not identified
major gene loci that control BMD, rather the loci identified
are of moderate to low impact on the trait. Candidate gene
The authors have no conflict of interest.
1Genomics Research Center, School of Health Science, Gold Coast Campus Griffith University, Queensland, Australia.
2Clinical and Biomedical Health Sciences–Barwon Health, The University of Melbourne, The Geelong Hospital, Geelong, Australia.
1527

1528
VAUGHAN ET AL.
osteoblasts and, consequently, unmineralized skeletons be- spine, 1.6% for the femoral neck, 2.1% for Ward’s triangle,
cause of a failure in osteoblast differentiation.(13,14) In hu- 1.6% for the trochanter, 0.4% for the whole body, 2.1% for
mans, mutations in RUNX2 cause cleidocranial dysplasia the UD forearm, and 1.1% for the midforearm sites. Bone
(CCD), a disorder characterized by hypoplasia or aplasia of mineral apparent density (BMAD) was calculated for the
the clavicles, short stature, supernumerary teeth, patent fon- spine (L2–L4) as bone mineral content (BMC)/(area)1.5 ac-
tanelles, and other changes in skeletal patterning and cording to Carter et al.(25)
growth.(15) Heterozygous RUNX2 knockout mice have ab-
normalities that are characteristic of CCD, confirming that
Population sample
CCD derives from haplo-insufficiency for RUNX2.(13,14)
A random sample of 495 subjects of the GOS study, who
were free of diseases and drugs known to influence BMD,
was examined to determine if different genotypes were
significantly related to BMD. In a population sample, BMD
follows a normal distribution and is influenced by anthro-
pometric variables such as age, height, and weight as well as
lifestyle factors. Analysis of covariance (ANCOVA) can be
used to determine the significance of a genotype effect
while simultaneously adjusting for the effect of typical
covariates such as age, height, and weight. In a cross-
sectional population sample from a single ethnic group, the
population mean for BMD is determined by the sum of
effects of common allelic variants and environment. In the
standard model, an allele confers an effect (␣) on a trait. The
variance attributable to the locus is given by 2pq␣2, where
p and q are the frequencies of alternative alleles. Therefore,
high-effect alleles (large ␣) of low frequency are detectable
because the mean trait value of carriers is significantly
different from the population mean. The population sample
provides an estimate of ␣, obtained through an ANCOVA
model that incorporates other covariates that influence BMD.
For the population sample, genotype data were obtained on 495
randomly selected subjects, representing one-third of the total
study. This number was based on power calculations for an allele
of 0.1 frequency acting in a dominant or additive manner with a
moderate additive genetic effect (an ␣ of 0.25 SD).
RUNX2 is expressed in differentiated osteoblasts and trans-
activates genes involved in the deposition of bone matrix,
such as type 1 collagen, osteocalcin, and osteopontin, fur-
ther indicating a role for RUNX2 in mature osteoblasts.(16,17)
The RUNX2 gene resides on 6p21 over ϳ250 kb and pro-
duces different isoforms that may have differing func-
tions.(18) Overexpression of a dominant negative form of
RUNX2 in mature osteoblasts of transgenic mice leads to
less active osteoblasts, reduced bone formation, and short
stature.(16)
Preceding the runt domain, RUNX2 has a polyglutamine
polyalanine (polyQ/polyA) repeat that acts as one of three
transactivation domains in the protein.(19) Expansion and
deletion of RUNX2 repeats may result in phenotypes other
than CCD. One CCD family had a deletion in this area,(15)
and an in-frame expansion of the polyalanine repeat oc-
curred in a CCD family with brachydactyly and minor
craniofacial features.(20) BMD has not been examined rou-
tinely in CCD, although a patient with fracture and low
BMD was reported.(20) The central role of RUNX2 in de-
termining osteoblast differentiation makes RUNX2 a prime
candidate gene for regulating adult bone density.
We hypothesized that genetic variation in RUNX2 may re-
sult in differences in skeletal development and bone density.
To test these hypotheses, we determined the extent of natural
deletion or expansion in polyQ/polyA region in a normal
population sample and a cohort of fracture cases. Alleles of
RUNX2 were significantly related to BMD and fracture.
Fracture cases
Fracture cases were ascertained from sequential fracture
cases at radiology practices that service the Barwon Statistical
Division. DNA was taken from consenting female fracture
patients over 35 years of age drawn from a population catch-
ment of 218,000 inhabitants observed for a 2-year period, as
previously described.(23) Approximately 800 patients with
fractures of all types donated DNA. There was no overlap
between the fracture cases and the 495 subjects of the random
GOS subsample. To define the fracture type more closely, only
hip, spine, and Colles’ fracture were accepted into the geno-
typing study. An appropriate control population for the fracture
study was all postmenopausal females from the GOS random
sample, because these individuals did not sustain fracture dur-
ing the study period.
MATERIALS AND METHODS
Study subjects
The Geelong Osteoporosis Study (GOS), based in the
Barwon Statistical Division of Victoria (Australia), was
used as a source of subjects with BMD data. The GOS study
has been described previously.(21–24) DNA extraction was
performed using standard procedures. GOS comprised 1494
white female volunteers with BMD measurements and
DNA. The Barwon Health Research and Ethics Advisory
Committee approved research protocols.
Bone density measurements
DXA was performed, using a GE Lunar DPX-L densi-
tometer, and analyzed using GE Lunar DPX-L software
version 1.31(GE Lunar, Madison, WI, USA). BMD was
measured at the spine (L2–L4) in the posterior-anterior
Detection of alleles in RUNX2 polyQ/polyA
repeat sequence
Primers were, forward, 5Ј-AGCCTGCAGCCCGGCAA-
projection, proximal femur (femoral neck, Ward’s triangle, AATGAGC-3Ј, and, reverse, 5Ј-GGGTGGTCGGCGAT-
and trochanter), whole body, ultradistal (UD), and midfore- GATCTCCACCATG-3Ј, (Fig. 1). The buffer contained 3.5
arm sites. In vivo short-term precision was 0.6% for the mM of MgCl2, 20 mM of Tris-HCl, pH 8.4, 50 mM of KCl,

RUNX2 VARIANTS AND BMD
1529
FIG. 1. Position of RUNX2 variants. The line depicts the genomic organization of RUNX2, with exons (E) denoted by vertical bars. The
glutamine-alanine repeat (Q/A) and the runt DNA binding domain are indicated. The segment of exon 1 in the figure begins at codon 18, nucleotide
52 of RUNX2. The positions of the PCR and sequencing primers are indicated by arrows. The underlined region denotes the position of the
glutamine insertion (30Q and 24Q) and deletion (15Q and 16Q) variants. The position of the 11Ala deletion previously identified(15) is indicated
by the rectangle. The G/A SNP is denoted in bold.
FIG. 2. This figure shows the resolution of
deletion and SNP genotypes using the heterodu-
plex assay. Track 1 is DNA standards with mo-
lecular sizes indicated on the left side of the
figure. Tracks 2–4 are amplified DNAs of indi-
viduals of different genotypes indicating the lack
of resolution when the wild-type 17 Ala frag-
ment is electrophoresed. Track 5 is amplified
DNA from an individual homozygous for the 11
Ala allele. Tracks 6–8 show the heteroduplex
assay. To each of the DNAs shown in tracks
2–4, an equal amount of the homozygous 11 Ala
allele DNA (track 5) was added and then the
mixture underwent a cycle of denaturation and
renaturation to form heteroduplexes. The upper
retarded bands are the heteroduplexes, clearly
showing resolution of homozygous G genotype
(track 6), heterozygous G/A (track 7), and ho-
mozygous A genotype (track 8).
10% DMSO, and 0.5 mM of deoxyribonucleotides triphos- the reaction mix denatured at 94°C for 5 minutes followed by
phates (dNTPs), with 0.5 U Taq DNA polymerase (Gibco a return to 65°C for 5 minutes. Heteroduplexes are resolved
BRL, Sydney, Australia). The amplification conditions were then on 7.5% PAGE. SNP heterozygote genotype was con-
95°C for 3 minutes, 65°C for 1 minute, and 74°C for 1 firmed by a second procedure using ultrathin 80-␮m flat non-
minute (cycle 1); 95°C for 20 s, 65°C for 30 s, and 74°C for denaturing PAGE gel combined with a laser detector in a
30 s (cycles 2–6); 95°C for 5 s, 65°C for 5 s, and 74°C for Corbett GS2000 device (Corbett Research, Mortlake, Sydney,
15 s (cycles 7–35).
Australia) using ethidium bromide and real-time laser detec-
Heteroduplex analysis by PAGE (Fig. 2) was used to deter- tion. Using this device, the 1-bp mismatched heteroduplex in
mine the SNP genotypes using a detector strand of DNA that an SNP heterozygote has an apparent mobility shift of 9 bps.
was prepared by large-scale polymerase chain reaction (PCR) All genotypes were obtained in duplicate on two different
amplification of the homozygous 18-bp deletion (11Ala) allele. occasions.
When heteroduplexed with a 232-bp fragment (17Ala), the
detector strand forms a heteroduplex bubble that overlaps the
site of the SNP. The mobility of the heteroduplex is retarded in
the presence of the A allele, relative to the G allele. To perform
Sequencing of alleles in RUNX2 polyQ/polyA
repeat sequence
this assay, patient DNA was first PCR-amplified, and then an
The PCR fragment was sequenced using both the forward
equal mass of detector strand was added to the test DNA and and the reverse direction using the aforementioned PCR

1530
VAUGHAN ET AL.
TABLE 1. GENOTYPE DISTRIBUTION, ANTHROPOMETRIC DATA, AND BONE DENSITY PARAMETERS AT DIFFERENT SKELETAL
SITES ACCORDING TO RUNX2 GENE POLYMORPHISM IN ALL SUBJECTS
All
G/G
G/A
A/A
G/11Ala
A/11Ala
11Ala/11Ala
Parameter
(n ϭ 492)
(n ϭ 361)
(n ϭ 77)
(n ϭ 6)
(n ϭ 40)
(n ϭ 5)
(n ϭ 3)
Age (years)
Weight (kg)
Height (cm)
BMD
52 Ϯ 20
67 Ϯ 14
161 Ϯ 7
52 Ϯ 20
67 Ϯ 15
161 Ϯ 7
55 Ϯ 20
66 Ϯ 13
160 Ϯ 7
52 Ϯ 20
66 Ϯ 10
162 Ϯ 8
51 Ϯ 20
70 Ϯ 12
162 Ϯ 7
53 Ϯ 23
67 Ϯ 3
162 Ϯ 3
41 Ϯ 13
65 Ϯ 5
168 Ϯ 8
Spine (L2–L4)
Femoral neck
Ward’s
Trochanter
Whole body
UD forearm
Midforearm
BMAD
1.15 Ϯ 0.19
0.90 Ϯ 0.17
0.78 Ϯ 0.20
0.79 Ϯ 0.15
1.11 Ϯ 0.12
0.30 Ϯ 0.07
0.65 Ϯ 0.10
1.14 Ϯ 0.19
0.90 Ϯ 0.17
0.78 Ϯ 0.20
0.78 Ϯ 0.15
1.11 Ϯ 0.12
0.30 Ϯ 0.07
0.65 Ϯ 0.10
1.17 Ϯ 0.21
0.91 Ϯ 0.16
0.78 Ϯ 0.20
0.80 Ϯ 0.15
1.12 Ϯ 0.12
0.31 Ϯ 0.07
0.65 Ϯ 0.10
1.27 Ϯ 0.15
1.03 Ϯ 0.21
0.89 Ϯ 0.25
0.90 Ϯ 0.18
1.17 Ϯ 0.09
0.34 Ϯ 0.06
0.71 Ϯ 0.08
1.17 Ϯ 0.20
0.91 Ϯ 0.19
0.78 Ϯ 0.20
0.79 Ϯ 0.14
1.11 Ϯ 0.11
0.30 Ϯ 0.06
0.65 Ϯ 0.10
1.19 Ϯ 0.17
0.97 Ϯ 0.18
0.89 Ϯ 0.26
0.85 Ϯ 0.13
1.17 Ϯ 0.06
0.34 Ϯ 0.04
0.71 Ϯ 0.04
1.12 Ϯ 0.08
0.91 Ϯ 0.10
0.78 Ϯ 0.10
0.73 Ϯ 0.13
1.11 Ϯ 0.05
0.33 Ϯ 0.15
0.72 Ϯ 0.05
Spine (L2–L4)
0.18 Ϯ 0.03
0.18 Ϯ 0.03
0.18 Ϯ 0.03
0.19 Ϯ 0.02
0.18 Ϯ 0.03
0.19 Ϯ 0.03
0.17 Ϯ 0.01
Data are mean Ϯ SD. Three individuals with glutamine deletions are not represented in these data.
primers. Terminator dye ready reaction mix was used ac- ferences in heteroduplex analysis, permitting a method of
cording to the manufacturer’s instructions (Applied Biosys- genotyping. This polymorphism exists in codon 66 of the
tems, Foster City, CA, USA) and the sequencing reaction RUNX2 transcript. A number of individuals were noted that
was followed by ethanol precipitation. The polyA/polyQ had rare deletions or expansions. On sequence, these proved
repeat was sequenced in a total of 30 chromosomes. The to be glutamine repeat variants. No other alleles of the
previously identified alanine deletion(15) was sequenced in 6 alanine repeat were observed.
chromosomes. Various mutants identified by unique mobil-
ity patterns were sequenced according to the aforemen-
tioned protocol.
Relationship of RUNX2 alleles to bone density
The relationship between genotype and BMD parameters
was assessed in subjects of the random GOS sample (n ϭ
495). Of the original 495 subjects, 3 individuals were ex-
cluded who had glutamine repeat mutations, leaving 492
individuals shown in Table 1. The following were the allele
frequencies estimated from the random sample: G allele,
0.85 Ϯ 0.01; A allele, 0.10 Ϯ 0.01; and 11Ala alanine
deletion allele, 0.05 Ϯ 0.01. Because the G and A alleles
exist within the region that is deleted in the 11Ala allele, it
is not possible to construct haplotypes; these three variants
are best considered allelic. Genotype distributions (Table 1)
were consistent with the Hardy-Weinberg equilibrium.
Mean anthropomorphic and BMD variables are presented
in Table 1 for subjects in the random sample. There were no
significant differences in age, weight, and height in relation
Statistical analysis
ANOVA was used to compare subject anthropomorphic
characteristics, BMD, and BMAD in different RUNX2 ge-
notype groups. ANCOVA was used to adjust for the covari-
ates age, weight, and height for BMD at the spine; age and
weight for BMD at all other skeletal sites; and age and
weight for BMAD at the spine. Student’s t-test was used in
2
simple comparisons of two groups. ␹ was used to deter-
mine association between control and fracture cases.
RESULTS
Sequence of RUNX2 exon 1
A 232-bp fragment of the polyQ/polyA repeat region of to genotypes. Examination of BMD mean values suggested
exon 1 was amplified by PCR and sequenced (Fig. 1). A that heterozygotes for the A allele (n ϭ 77) had higher BMD
previously identified common polymorphic deletion of 18 than G/G homozygotes (n ϭ 361). Although only six indi-
bp in the alanine repeat was sequenced.(15) Because the viduals were A/A homozygotes, these subjects had mean
deletion alters the alanine repeat, the two alleles were BMD levels higher, again, than that of G/A heterozygotes,
termed 17Ala and 11Ala, with 17Ala being the most com- suggestive of allele dosage effect. Five A/11Ala deletion
mon form. PCR amplification of 11Ala/17Ala heterozygotes heterozygotes were observed. These subjects had signifi-
result in heteroduplex structures, resolvable on PAGE (Fig. cantly higher BMD at a number of sites compared with G/G
2). DNA sequence analysis revealed a G to A transition, homozygotes.
resulting in a synonymous codon polymorphism (GCG to
Because BMD is influenced by age and weight, the effect
GCA with alleles denoted as G and A, respectively) at the of the A allele was examined using ANCOVA to adjust for
eighth alanine of the 17 alanine repeat in exon 1 of RUNX2 age and weight covariates. Carriers of the A allele (A/A
(Fig. 1). This polymorphism resulted in electrophoretic dif- homozygotes, G/A heterozygotes and A/11Ala heterozy-

RUNX2 VARIANTS AND BMD
1531
TABLE 2. COMPARISON OF ADJUSTED MEAN BMD ACCORDING TO CARRIER STATUS FOR THE SNP A ALLELE
A positivea
(n ϭ 88)
A negativea
(n ϭ 401)
Percent
variance
SD
difference
BMD parameter
p Value
Spine (L2–L4)
Femoral neck
Ward’s triangle
Trochanter
Whole body
UD forearm
1.18 Ϯ 0.017
0.94 Ϯ 0.013
0.82 Ϯ 0.015
0.83 Ϯ 0.013
1.16 Ϯ 0.009
0.33 Ϯ 0.006
0.70 Ϯ 0.007
0.18 Ϯ 0.003
1.13 Ϯ 0.008
0.90 Ϯ 0.006
0.77 Ϯ 0.007
0.79 Ϯ 0.009
1.13 Ϯ 0.005
0.32 Ϯ 0.004
0.68 Ϯ 0.005
0.18 Ϯ 0.001
0.007
0.003
0.006
0.003
0.001
0.001
0.004
0.007
1.5
1.8
1.5
1.9
2.1
2.4
1.7
1.5
0.32
0.35
0.33
0.36
0.38
0.41
0.34
0.32
Midforearm
BMAD spine (L2–L4)
Data are presented as adjusted means Ϯ SE, p value related to the gene effect, percent of the variance in BMD explained by A allele
status, and difference in terms of the adjusted SD at various skeletal sites with respect to the presence or absence of the A allele. ANCOVA
used age and weight as covariates for all sites except spine BMD and BMAD where age, weight, and height were covariates.
a A positive refers to genotypes A/A, A/G, and A/11Ala. A negative refers to genotype G/G and G/11Ala.
gotes) were compared with A allele noncarriers (G/G ho- over the same 2-year interval in which the fracture cases
mozygotes and G/11Ala heterozygotes). A allele carriers were collected and who were from the same geographic
had significantly higher BMD at all skeletal sites compared region. Four individuals from the fracture cohort were het-
with A allele noncarriers (Table 2). When expressed as a erozygous for mutations in the glutamine repeat: an inser-
proportion of the SD of the adjusted BMD at each site, the tion of 21 bp (30Q) in a hip fracture case, deletions of 21 bp
difference in BMD ranged from 0.32 SD at the spine (p ϭ (16Q) in a hip and a spine fracture case, and in a third hip
0.007) to 0.41 SD at the UD forearm (p ϭ 0.001; Table 2). fracture case insertion of 3 bp (24Q). The two 16Q mutants
Accounting for age and weight covariates, the effect of A were not allelic, because one was from an A allele–
allele status accounted for 1.8% of the variance in BMD containing chromosome and the other from a G allele chro-
averaged across all skeletal sites (ranging from 1.5% at mosome. Four glutamine mutants were observed in 278
Ward’s triangle to 2.4% at the UD forearm) and 2.1% of the fracture cases compared with none in the 224 postmeno-
variance for whole body BMD.
pausal women from the GOS (p ϭ 0.13, Fisher’s exact test).
Confining the comparison to hip fracture versus control
gives a value of p ϭ 0.011 (Fisher’s exact test). Three
glutamine repeat variants observed in the GOS random
sample were premenopausal (these were 24Q/23Q, 16Q/
23Q, and 15Q/23Q heterozygotes). The clinical character-
istics of the seven subjects with glutamine repeat variants
were interesting; the three premenopausal women with glu-
tamine variants had BMD values below the mean of young
normal women at all sites examined and the four fracture
cases also had low BMD at all sites (data not shown).
Effect of deletion genotypes on BMD parameters
In contrast to the effect of the A allele, the presence or
absence of the 11Ala deletion allele was not significantly
related to BMD at any site using ANCOVA (data not
shown). In 495 subjects of the random sample, three ho-
mozygotes of the 11Ala deletion allele and 45 heterozygotes
for the 11Ala deletion allele were observed. Mean height,
weight, and age values were not significantly related to the
presence of the RUNX2 11Ala allele. SNP G/G homozy-
gotes were compared against genotype G/11Ala (11 alanine
deletion heterozygotes with a G allele) to isolate the effect
of the deletion allele without considering the effect of the A
allele. No significant difference was observed. There was a
trend toward lower anterior-posterior spine BMD and
BMAD in the three 11Ala/11Ala homozygote premeno-
pausal women (p ϭ 0.1 and p ϭ 0.03, respectively) when
compared with all other premenopausal women.
Genotype distributions in fracture cases
The fracture group mean age was 70 Ϯ 12 (SD) years.
The postmenopausal (nonfracture) women from the GOS
random population selection had a mean age of 72 Ϯ 10
(SD) years and did not have any fractures during the study
period. Spine and hip fracture groups were not significantly
different from the control group (Table 3). Allele counts in
the Colles’ fracture group (numbers: G allele, 182; A allele,
9; 11Ala allele, 15) were significantly different from the
Unusual alleles observed in the fracture cases
Patients with hip (n ϭ 67), Colles’ (n ϭ 103), and spine postmenopausal group (numbers: G allele, 382; A allele, 49;
(n ϭ 108) fracture were genotyped (Table 3). The appro- 11Ala allele, 17) with p ϭ 0.005. This effect is generated by
priate nonfracture control group was postmenopausal a decrease in the A allele in the fracture group. To consider
women from the GOS random sample because fracture and the 11Ala deletion alone, A and G alleles were pooled. The
bone-related disease had been part of the exclusion criteria. 11Ala allele showed a trend for an increase in the Colles’
This represented 224 women of similar mean age to the fracture group (p ϭ 0.07). When 11Ala alleles were pooled
fracture cases and who had not suffered incident fracture with G alleles, carriers of the A allele were associated with

1532
VAUGHAN ET AL.
TABLE 3. GENOTYPES AND FRACTURE SITE
Genotype counts
Site
G/G
G/A
A/A
G/11Ala
A/11Ala
Uniquea
Total n
Spine
Hip
Colles’
Control
76
46
79
20
13
9
0
0
0
3
11
5
15
15
0
0
0
2
16Q
30Q, 24Q, 16Q
108
67
103
224
163
41
Allele counts and (frequency)
Site
G allele
A allele
11Ala allele
Total n
p Value*
Spine
Hip
Colles’
Control
183 (0.86)
110 (0.86)
182 (0.88)
382 (0.85)
20 (0.09)
13 (0.10)
9 (0.04)
11 (0.05)
5 (0.04)
15 (0.07)
17 (0.04)
214
128
206
448
0.6
0.9
0.005
49 (0.11)
Counts of genotypes and alleles according to fracture type were compared with postmenopausal women from the GOS.
a Unique allele given as glutamine variants; all were heterozygotes with the 23Q wild-type allele. ␹2 for the overall distribution, p ϭ
0.16.
2
*p Value considering each fracture class independently against the control group. ␹ overall, p ϭ 0.092; unique individuals were
excluded from allele count analysis.
a reduced frequency of Colles’ fracture (2.7-fold; 95% CI of were observed also: 30Q, 24Q, 16Q, and 15Q compared
the odds ratio [OR], 1.3–5.5). These data provide support with the normal 23Q. All three premenopausal glutamine
for the relationship between the A allele and high BMD
(potentially reducing fracture risk), particularly at the fore-
arm sites where the association with BMD was strongest.
mutation carriers had BMD lower than the mean of young
normal women and the four other elderly glutamine re-
peat mutation carriers had osteoporotic fractures and low
BMD (data not shown). These data suggest that glu-
tamine mutations may be correlated with reduced BMD
and fracture. Seven glutamine repeat mutations were
observed in 773 genotyped subjects (495 subjects from
the GOS sample and 278 fracture cases), representing
0.9% of the population. The frequency of gross variation
of the glutamine repeat region suggests that it is a fre-
quent enough event to have a potential impact when
considered in large populations.
The strongest effect of the A allele was observed at the
UD forearm, where carriers of the A allele had BMD 0.41
SD higher than noncarriers. The contribution of the A allele
to the phenotypic variance of BMD derived from ANCOVA
models that incorporated age and weight as covariates was
2.1% of the variance in whole body BMD in the entire
group and 2.4% for the UD forearm. A recent whole ge-
nome scan(5) for forearm BMD did not show linkage to
6p21, where RUNX2 maps, perhaps because of inadequate
sample size to detect an allele of this effect size; however,
a genome scan of 595 sister pairs showed a significant
linkage (LOD score, 2.15) on 6p21.(6)
DISCUSSION
BMD is influenced by both genetic and environmental
factors.(1,26) Notable examples of candidate genes that
have shown positive associations with adult bone density
include the vitamin D receptor, interleukin-1 receptor,
type 1 collagen, and transforming growth factor (TGF)
␤.(1,26) These genes have widespread expression patterns,
suggesting that genes involved in determining BMD may
not be derived exclusively from bone. However, genetic
variation in bone-specific genes has a greater likelihood
of impacting on bone rather than in other tissue systems.
Of genes recently examined for relationships with
BMD,(26) only osteocalcin represents a bone-specific
gene. A master regulator gene such as RUNX2 that de-
termines the differentiation fate of bone cells is a prime
candidate gene for regulating adult BMD. Our hypothesis
was that variations within the polyQ/polyA repeat region
may be a determinant of BMD. In the course of this
study, we identified an SNP allele that was statistically
related to BMD. A common 18-bp deletion polymor-
phism of the alanine repeat was observed in this white
population. Our data match a previous observation of this
deletion, which we term the 11Ala allele in comparison
with the 17Ala wild-type allele.(15) The SNP allele that
we identified is on codon 66 of the RUNX2 transcript and
The A allele was significantly associated with decreased
incidence of Colles’ fracture. The lack of a relationship with
hip and spinal fracture may relate to the lower impact of the
A allele on BMD at these sites. BMD at UD forearm showed
the strongest relationship to the A allele. Therefore, it is
perhaps not surprising that Colles’ fracture was associated
with A allele status. The common alanine deletion polymor-
appears to be different from that reported by Quack et phism (11Ala allele) was not related to any clinical param-
al.,(20) which is on codon 69. Glutamine repeat mutations eters. Because of the allele frequency, the statistical test was

RUNX2 VARIANTS AND BMD
1533
6. Koller DL, Econs MJ, Morin PA, Christian JC, Hui SL,
Parry P, Curran ME, Rodriguez LA, Conneally PM, Joslyn
G, Peacock M, Johnston CC, Foroud T 2000 Genome
screen for QTLs contributing to normal variation in bone
mineral density and osteoporosis. J Clin Endocrinol Metab
85:3116–3120.
7. Uitterlinden AG, Burger H, Quiju H, Fang Y, McGuigan FE,
Grant SF, Hofman A, van Lleuwen JP, Pols HA, Ralston SH
1998 Relation of alleles of the collagen type 1␣1 gene to bone
density and risk of osteoporotic fractures in postmenopausal
women. N Engl J Med 338:1016–1021.
8. Langdahl BL, Lokke E, Carstens M, Stenkjaer LL, Eriksen EF
2000 Osteoporotic fractures are associated with an 86-base
pair repeat polymorphism in the interleukin-1-receptor antag-
onist gene but not with polymorphisms in the interleukin-1beta
gene. J Bone Miner Res 15:402–414.
9. Raymond MH, Schutte BC, Torner JC, Burns TL, Willing MC
1999 Osteocalcin: Genetic and physical mapping of the human
gene BGLAP and its potential role in postmenopausal osteo-
porosis. Genomics 60:210–217.
10. Bertoldo F, D’Agruma L, Furlan F, Colapietro F, Lorenzi MT,
Maiorano N, Iolascon A, Zelante L, Locascio V, Gasparini P
2000 Transforming growth factor-beta1 gene polymorphism,
bone turnover, and bone mass in Italian postmenopausal
women. J Bone Miner Res 15:634–639.
11. Werner MH, Shigesada K, Ito Y 1999 Runt domains take the
lead in hematopoiesis and osteogenesis. Nat Med 5:1356–
1357.
12. Lee B, Thirunavukkarasu K, Zhou L, Pastore L, Baldini A,
Hecht J, Geoffroy V, Ducy P, Karsenty G 1997 Missense
mutation abolishing DNA binding of the osteoblast-specific
transcription factor OSF2/CBFA1 in cleidocranial dysplasia.
Nat Genet 16:307–310.
13. Komori T, Yagi H, Nomura S, Yamaguchi A, Sasaki K,
Deguchi K, Shimizu Y, Bronson R, Gao Y, Inada M, Sato M,
Okamoto R, Kitamura Y, Yoshiki S, Kishimoto T 1997 Tar-
geted disruption of Cbfa1 results in a complete lack of bone
formation owing to maturational arrest of osteoblast. Cell
89:755–764.
14. Otto F, Thornell AP, Crompton T, Denzel A, Gilmour KC,
Rosewell IR, Stamp GW, Beddington RS, Mundlos S, Olsen
BR, Selby PB, Owen MJ 1997 Cbfa1, a candidate gene for
cleidocranial dysplasia syndrome, is essential for osteoblast
differentiation and bone development. Cell 89:765–771.
15. Mundlos S, Otto F, Mundlos C, Mulliken JB, Aylsworth AS,
Albright S, Lindhout D, Cole WG, Henn W, Knoll JH, Owen
MJ, Mertelsmann R, Zabel BU, Olsen BR 1997 Mutations
involving the transcription factor CBFA1 cause cleidocranial
dysplasia. Cell 89:773–779.
16. Ducy P, Starbuck M, Priemel M, Shen J, Pinero G, Geoffroy
V, Amling M, Karsenty G 1999 A Cbfa1-dependent genetic
pathway controls bone formation beyond embryonic develop-
ment. Genes Dev 13:1025–1036.
17. Sato M, Morii E, Komori T, Kawahata H, Sugimoto M, Terai
K, Shimizu H, Yasui T, Ogihara H, Yasui N, Ochi T, Kitamura
Y, Ito Y, Nomura S 1998 Transcriptional regulation of os-
teopontin gene in vivo by PEBP2alphaA/CBFA1 and ETS1 in
the skeletal tissues. Oncogene 17:1517–1525.
18. Harada H, Tagashira S, Fujiwara M, Ogawa S, Katsumata T,
Yamaguchi A, Komori T, Nakatsuka M 1999 Cbfa1 isoforms
exert functional differences in osteoblast differentiation. J Biol
Chem 274:6972–6978.
restricted to a comparison of the heterozygote 11Ala/17Ala
with the 17Ala/17Ala genotype. This test is only applicable
to additive or dominant effects. Three individuals of ho-
mozygous 11Ala/11Ala genotype were observed and in
these subjects there was a trend toward lower BMD that was
significant only when spine BMAD was considered (p ϭ
0.03). Consequently, we cannot rule out a recessive effect of
the alanine deletion polymorphism on BMD.
In conclusion, variation in RUNX2 may be a contributory
determinate of adult BMD. We detected common polymor-
phism and rare mutations within the polyQ/polyA repeat
region of exon 1, with rare glutamine repeat variants ac-
counting for 0.9% of subjects. The alanine deletion poly-
morphism was not related to BMD whereas the A allele was
related to significantly greater BMD at all measured sites
and is associated with approximately threefold protection
against Colles’ fracture. Because the A allele does not alter
the amino acid sequence of the Runx2 protein, the most
plausible explanation of the relationship between the A
allele and BMD, is that the A allele is in linkage disequi-
librium with some other unidentified functional alteration in
the RUNX2 gene. Glutamine repeat mutations were found in
fracture cases but not in age-matched normal controls, sug-
gesting that glutamine mutation in Runx2 may be related to
fracture in some cases. Further analysis in other populations
is required. Because RUNX2 is a master regulator of osteo-
blast function, these data lend hope to manipulating osteo-
blast cell function as a means of preventing osteoporotic
fracture.
ACKNOWLEDGMENTS
We thank Rebecca Wunsch for expert genotyping analy-
sis and the staff of the GOS for their assistance in recruiting
subjects. This research was supported by grants from Grif-
fith University, the Victorian Health Promotion Foundation,
and the Geelong Region Medical Research Foundation.
Tanya Vaughan is the recipient of a scholarship from the
Australian Research Council.
REFERENCES
1. Ralston S 1999 The genetics of osteoporosis. Bone 25:85–86.
2. Cummings SR, Kelsey JL, Nevitt MC, O’Dowd KJ 1985
Epidemiology of osteoporosis and osteoporotic fractures. Epi-
demiol Rev 7:178–208.
3. Benes H, Weinstein RS, Zheng W, Thaden JJ, Jilka RL,
Manolagas SC, Shmookler Reis RJ 2000 Chromosomal
mapping of osteopenia-associated quantitative trait loci us-
ing closely related mouse strains. J Bone Miner Res 15:
626–633.
4. Shimizu M, Higuchi K, Bennett B, Xia C, Tsuboyama T, Kasai
S, Chiba T, Fujisawa H, Kogishi K, Kitado H, Kimoto M,
Takeda N, Matsushita M, Okumura H, Serikawa T, Nakamura
T, Johnson TE, Hosokawa M 1999 Identification of peak bone
mass QTL in a spontaneously osteoporotic mouse strain.
Mamm Genome 10:81–87.
19. Thirunavukkarasu K, Mahajan M, McLarren KW, Stifani S,
Karsenty G 1998 Two domains unique to osteoblast-specific
transcription factor Osf2/Cbfa1 contribute to its transactivation
function and its inability to heterodimerize with Cbf-beta. Mol
Cell Biol 18:4197–4208.
5. Niu T, Chen C, Cordell H, Yang J, Wang B, Wang Z, Fang Z,
Schork N, Rosen CJ, Xu X 1999 A genome-wide scan for loci
linked to forearm bone mineral density. Hum Genet 104:226– 20. Quack I, Vonderstrass B, Stock M, Aylsworth A, Becker A,
233. Brueton L, Lee P, Majewski F, Mulliken J, Suri M, Zenker M,

1534
VAUGHAN ET AL.
Mundlos S, Otto F 1999 Mutation analysis of core binding 25. Carter DR, Bouxsein ML, Marcus R 1992 New approaches for
factor A1 in patients with cleidocranial dysplasia. Am J Hum
interpreting projected bone densitometry data. J Bone Miner
Genet 65:1268–1278.
Res 7:137–145.
21. Pasco JA, Henry MJ, Gaudry T, Nicholson GC, Kotowicz MA 26. Audi L, Garcia-Ramirez M, Carrascosa A 1999 Genetic de-
1999 Identification of incident fractures: The Geelong Osteo-
porosis Study. Aust N Z J Med 29:203–206.
terminants of bone mass. Horm Res 51:105–123.
22. Henry MJ, Pasco JA, Nicholson GC, Seeman E, Kotowicz MA
2000 Prevalence of osteoporosis in Australian women:
Geelong osteoporosis study. J Clin Densitom 3:261–268.
23. Sanders KM, Pasco JA, Ugoni AM, Nicholson GC, Seeman E,
Martin TJ, Skoric B, Panahi S, Kotowicz MA 1998 The exclusion
of high trauma fractures may underestimate the prevalence of
bone fragility fractures in the community: The Geelong Osteo-
porosis Study. J Bone Miner Res 13:1337–1342.
Address reprint requests to:
Nigel A. Morrison, Ph.D.
Genomics Research Centre
School of Health Science
Gold Coast Campus Griffith University
Queensland 9726, Australia
24. Sanders KM, Seeman E, Ugoni AM, Pasco JA, Martin TJ,
Skoric B, Nicholson GC, Kotowicz MA 1999 Age- and
gender-specific rate of fractures in Australia: A population- Received in original form August 17, 2001; in revised form Feb-
based study. Osteoporos Int 10:240–247. ruary 15, 2002; accepted March 8, 2002.