WO1996003524A1 - Diagnostic method - Google Patents

Abstract

The present invention provides a method of assessing an individual's predisposition to low or high bone density and/or risk of fracture comprising assessing the vitamin D receptor genotype of the individual by haplotype analysis. Preferably the genotype of the individual is assessed by haplotype analysis by assessing the genotype for the polymorphic restriction endonuclease sites for Bsm-1, Apa-1 and Taq-1 in the combinations of Bsm-1 and Apa-1, or Taq-1 and Apa-1, or Bsm-1 and Taq-1, or Bsm-1, Apa-1 and Taq-1.

Description

DIAGNOSTIC METHOD The present invention relates to the use of vitamin D receptor (VDR) gene locus variants assessed by haplotype analysis and their use in methods of assessing genetic resistance or sensitivity to differences in bone mineral density, bone turnover, osteopenia and osteoporosis and differences in physiological characteristics related to the vitamin D endocrine system.

In particular, the invention relates to the use of diagnostic kits for the discrimination of subjects with different set points of physiological characters associated with the vitamin D endocrine system and to the discrimination of bone density traits associated with health and disease. The invention also relates to the identification of individuals at low risk of osteoporotic bone disease. The use of VDR gene haplotypes has utility defined below, beyond the current state of the art. The present invention provides a method of assessing a number of genetically variant sites in the VDR gene, which when combined to form haplotypes results in the further discrimination of genetic subtypes with differring traits.

Osteoporosis is a debilitating bone disease that affects a high proportion of women and a lesser number of men. Due to the considerable public health problem associated with osteoporosis efforts have been focussed at identifying diagnostic and predictive markers associated with the disease. Since osteoporosis has a significant genetic component, the possiblity exists for genetic prediction of susceptibity and an understanding of the underlying pathophysiology of disease. Prediction of those at risk of osteoporosis may reduce the incidence of this disease by focussing early attention at those in greatest need. Our aim was to define a genetic test that is capable of identifying those individuals at risk of osteoporosis and those protected from disease. Osteoporosis is defined generally as low bone density associated with fracture however for genetic analysis, a bone density two standard deviations below young normal is a more useful criterion of osteoporosis since subsequent fracture is dependent on a traumatic event. Several humoral markers, notably osteocalcin, have been widely used as indicators of bone turnover. Genetic effects on osteocalcin levels have been reported in twin studies with a correlation between high osteocalcin levels and lower bone mineral density (Kelly et al., 1989).

Furthermore, this genetic effect is explained in part by allelic variation in the vitamin D receptor gene which influences osteocalcin serum levels (Morrison et al . 1992) and bone mineral density (Morrison et al. 1994). In this study, haplotypes of the vitamin D receptor gene are combined with serum osteocalcin in an analysis of age related changes in bone mineral density inferred from a cross sectional population study. The use of haplotypes defines extreme high and low bone density groups, and combined with osteocalcin data reveals a sub group essentially resistant to osteoporosis. The different cross sectional rates of bone loss may provide a genetic basis for explaining "fast" and "slow" bone losers (Christiansen) . The present invention consists in a method of assessing an individuals predisposition to low or high bone density and/or risk of fracture comprising assessing the vitamin D receptor genotype of the individual by haplotype analysis. In a preferred embodiment of the present invention the genotype is assessed by haplotype analysis by assessing the genotype for the polymorphic restriction endonuclease sites for Bsm-1, Apa-1 and Taq-1 in the combinations of Bsm-1 and Apa-1, or Taq-1 and Apa-1, or Bsm-1, Apa-1 and Taq-1. In a further preferred embodiment the assessment includes the polymorphic restriction endonuclease sites Sph-1 and/or the poly adenosine sequence microsatellite in the 3 prime untranslated region of exon 9 of the vitamin D receptor gene.

In yet a further preferred embodiment the assessment further includes measuring the individuals serum osteocalcin level.

The present invention can be practised in a number of ways in particular:

1. The use of Southern hybridisation blots to detect the Vitamin D receptor restriction fragment length polymoprhism using a cDNA fragment as a probe.

2. The use of polymerase chain reaction (PCR) and subsequent digestion by restriction endonucleases as described below to detect polymorphic sites and the subsequent analysis of the genotypes by combining Bsm-1 and Apa-1 or Apa-1 and Taq-1 RFLP data points and most preferably by the combination of all three RFLP genotypes. 3. Utilising a combination of serum osteocalcin and VDR haplotype derived genotype data to discriminate the relationship between bone turnover and bone density in genetic subgroups. 4. Provision of genomic cosmid clones containing DNA spanning the entire Vitamin D gene locus which can be used as probes to identify further genetic variants, which, if in strong linkage disequilibrium with the markers herein described will provide similar discriminatory genetic power sufficient to detect the same haplotypes. In order that the nature of the present invention may be more clearly understood preferred forms thereof will now be described with reference to the following examples and Figures in which:- Figure 1. Osteocalcin according to genotype and menopause. Genotype was defined as haplotypes derived from using Bsm-1, Apa-1 and Taq-1 RFLPs. Mean serum osteocalcin ± SEM is shown from the different genotypes as the pre and post menopausal values, open circles and closed circles, respectively.

Figure 2. Lumbar spine bone density as a function of VDR genotype in premenopausal females. The Mean ± SEM of lumbar spine BMD is plotted according to genotype as described by haplotypes see Figure 1. Significant differences are shown and the p values shown are derived from Student's t-test. The difference in mean BMD between homozygotes 1,1 and 2,2 exceeds one SD of the young normal population (0.145 gm/cm*-*).

Figure 3. The relationship between osteocalcin and bone density at the lumbar spine (A) (LS BMD = 1.299 ± 0.017 OC+, r² = 0.218) and femoral neck (B) (FN BMD =

2

0.954 ± 0.01 OC, r = 0.118) in postmenopausal females of genotype haplotype 1,1.

Figure 4. Menopause related change of lumbar spine bone density in major genotypic grouping defined by haplotypes of the vitamin D receptor gene.

A:- LS BMD 1.132 - 0.012 YPM, r² = 0.283, Intercept with

-2.0 SD value = 18.3 YPM B:- LS BMD 1.114 - 0.005 YPM, r² = 0.099, Intercept with -2.5 SD value = 40.4 YPM; mean menopause age is 48 years

C:- LS BMD 1.085 - 0.005 YPM, r² = 0.051, Intercept with -2.0 SD value = 34.6 YPM DD::-- LLSS BBMMDD 11..113355 -- 00..001111 YYPPMM,, r² = 0.311, Intercept with -2.0 SD value = 20.2 YPM EE::-- LLSS BBMMDD 11..111188 -- 00..001100 YYPPMM,, r² = 0.051, Intercept with -2.0 SD value = 20.6 YPM Figure 5. Significantly lower serum calcium in postmenopausal females of the low bone density genotypes and higher serum calcium in those genotypes associated with higher bone densities.

Figure 6. Illustrates the genetic effect on serum calcium, identified using a single RFLP (Bsm-1) (open circles premenopausal, closed circles postmenopausal).

Figure 7. Shows graphically the data presented in Table 20 illustrating haplotype derived genotypes and serum calcium in post menopausal women.

Figure 8. Shows the profound difference in bone density observed in male subjects across genotype using Taq-1.

Figure 9. Shows the analysis of bone density with VDR haplotypes in males .

Study 1. Methods.

Detection of Vitamin D receptor haplotypes.

Extraction of DNA from blood for genetic studies will be appreciated by those skilled in the art. It will also be appreciated by those skilled in the art that sequence differences in genes can be detected by numerous methods including polymerase chain reaction (PCR), Southern blot, ligase chain reaction, allele specific hybridisation and by hybridisation in solution. The examples below use PCR with subsequent restriction endonuclease digestion to identify polymorphic DNA sequences .

DNA analysis. Southern blot, PCR (polymerase chain reaction) and RFLP analysis using endonuclease digestion. Blood was collected into heparin treated tubes and leukocytes separated by sedimentation through physiological saline solution in a clinical centrifuge. Purified leukocytes were lysed in leukocyte lysis buffer (10 mM Tris-HCl, pH7.4, physiological saline and 0.5% w/v sodium dodecyl sulphate) . Lysate was treated with proteinase K (Applied Biosciences, Palo Alto USA) at 50 μg/ml for 2 hour at 65 Celcius. DNA was extracted by repetitive phenol chloroform solvent extraction as described in Maniatis et al . (1982) and ethanol precipitated. DNA was redissolved in TE buffer ( lOmM Tris-HCl, ImM EDTA, pH 8.0) and quantitated by absorbance at 260 nM. Other methods of DNA preparation (Kawasaki, 1990) are compatible with the PCR procedure.

The Bsm-1 and Apa-1 RFLPs (Morrison et al . , 1992) are in intron 8 while the Taq-1 RFLP is a synonymous isoleucine codon change (ATT to ATC, isoleucine codons) in the VDR coding sequence in exon 9 (Morrison et al . , 1994). To facilitate sequencing the region of interest, a 2.4 kilobasepair (kb) PCR fragment was generated spanning exons 7 to exon 9, in a 50μl reaction using a Pharmacia GeneAtaq thermocycler. This fragment originated in the end of intron 6 and terminated in exon 9, spanning intervening exons and introns and contains all the amino acid coding potential of exons 7 to 9. Some subjects were genotyped using the 2.4 kb or an alternative 2.1 kb fragment spanning the RFLPs, prior to the use of shorter fragments . Primer pairs derived from the sequence were tested for utility in fast capillary PCR using a Corbett Research FTS-1 thermocycler.

It will be appreciated by those skilled in the art that any oligonucleotide primer pair derived from the sequenced region or derived from immediately adjacent regions could be used for detection by amplification by PCR. Although a number of primer pairs were tested, four oligonucleotide primers were selected arbitrarily as standard primers to amplify regions of the VDR gene for these studies. Detection of the Bsm-1 site was facilitated by amplifying a region spanning the site, with one primer originating in exon 7 (primer 1: 5'-CAACCAAGACTACAAGTACCGCGTCAGTGA-3' ) and the other in intron 8 (primer 2: 5 '-AACCAGCGGAAGAGGTCAAGGG-3' ) producing a 825 basepair (bp) fragment. Detection of Apa- 1 and Taq-1 sites was facilitated using a single amplification with one primer in intron 8 (primer 3: 5'-CAGAGCATGGACAGGGAGCAAG-3' ) and the other in exon 9 (primer 4: 5 '-GCAACTCCTCATGGGCTGAGGTCTCA-3 ' ) producing a 740 basepair fragment. PCR reactions were 20μl containing 200ng genomic DNA, 20pmol of each primer, 200μM dNTPs, 50mM KC1, lOmM Tris (pH8.3), 1.5mM MgCl2 and 1U Tth DNA polymerase (Thermus thermophilus. Toyobo, Osaka, Japan) . Each sample was subjected to 30 amplification cycles as follows:

Step 1 - 3 min at 94°C, 1 min at 62°C, 2 min at 72°C;

Steps 2 to 6 - 20 sec at 94°C, 20 sec at 62°C, 1 min at

72°C;

Steps 7 to 30 - 5 sec at 94°C, 5 sec at 62°C, 30 sec at 72°C.

Total duration of amplification was 30 minutes. If product is insufficient this protocol can be extended another 10 cycles. Amplification regimes should be optimised for any particular thermal cycling device. Increased extension times may be necessary for standard tube machines and for 96 well microplate format machines.

A 10 μl aliquot of each PCR product was digested with 5 units of endonuclease Bsm-1 at 65 C (New England Biolabs, MA, USA), Apa-1 at 37°C or Taq-1 (Promega Co. Australia) at 65°C for 1 hour. A clone of an unrelated gene was used as an internal control for both Bsm-1 and Apa-1 digestion. For Taq-1 digestion, this plasmid control was not used since an invariant Taq-1 site in the PCR product itself was used as an internal control. The plasmid used was pOSCATl (Morrison et al . , 1989), however any other unrelated plasmid with a Bsm-1 or Apa-1 sites can be substituted for pOSCATl.

The digested PCR products were separated on 1.2% (Bsm-1 and Apa-1), or 2.0% (Taq-1) agarose gels containing 0.5μg/ml ethidium bromide, 0.09M Tris-borate and 0.002M

EDTA, pH 8.3 for 1 hr at 100V. EcoRI digested SPP1 marker (Bresatec Limited, Adelaide, Australia) was used as the size standard for all agarose gels. The Bsm-1 site is 650 bp into the 825 bp fragement, giving 175 bp and 650 bp fragments, while the absence of the site allele remains at 825 bp. In homozygous absence of the Taq-1 site, Taq-1 digestion of the 740 bp PCR product yields bands of 245 bp and 495 bp. Homozygous presence of the Taq-1 site yields fragments of 205 bp, 245 bp and 290 bp. Heterozygotes for the Taq-1 RFLP ( Tt ) exhibit fragments of 490 bp, 290 bp, 245 bp and 205 bp. Apa-1 digestion of the 740 bp PCR product yields fragments of 220 bp and 520 bp for the presence of the site allele and in the case of the absence of the site allele the fragment is not digested. Primer 1 and primer 4 can be combined in a standard slow PCR reaction (using a Pharmacia GeneAtaq cycler) to yield a 2.1 kb fragment suitable for genotyping all three RFLPs from the same fragment. However the smaller reaction volume and shorter time of fast capillary PCR proved more time effective for genotyping.

Due to the sequence of the relevant sites several other restriction enzymes can be used to detect these polymorphisms. Arbitrary laboratory designations for the RFLPs are displayed in this paper: a capital is used to signify the absence of the site and lower case signifies the presence of the site. Sequences flanking polymorphic sites are as follows. Bsm-1 site sequence from an invariant adjacent Stu-1 site; b allele AGGCCTGCG-CATTCCC, B allele underlined G is an A. This sequence change can be detected with Aos-1, Fsp-1, Mst-1, Fdi-2, Hinp-1, Hha-1 and their isoschizomers . Sequence at the polymorphic Apa- 1 site ending in an adjacent invariant Pvu-2 site is; a allele GAGGfiGCCCAGCTG, in the A allele the underlined G is a T. The presence of the G can be detected by Ban-2, Aoc- 2, Pss-1, Pal-1, Hae-3, Cfr-3I, Asu-1, Sau-961, Eco-01091, Dra-2, and isoschizomers . The presence of the T creates a polymorphism for Ban-1, and its isoschizomers. The sequence of the Taq-1 polymorphism spanning invariant Hha- 1 to Hae-3 sites is: T allele GCGCTGATTGAGGCC, in the t allele the underlined T is a C. This polymorphism can be also detected by Mbo-1, Sau-3A, Dpn-1 and their isoshizomers.

Genotypes were judged after electrophoresis by band sizes. Apa-1 and Taq-1 genotypes were assessed in different reactions of the same PCR product. Since Taq-1 can cut at 37°C it is possible to judge Apa-1 and Taq-1 genotypes by double digest of the product, although it was technically easier to digest separately. VDR gene haplotypes were given arbitrary numerical laboratory designations based on frequency. Statistics. Coassociation of genotype was tested using Chi-square and contingency table analysis. The Statview+graphics system by Abacus Inc. (Berkeley, California, USA) was run on a Macintosh IlSi. Results. Haplotype distributions in Caucasian subjects.

Subjects were recruited from the Sydney metropolitan area by advertisement. The total study population was 940 individuals some of whom were genotyped by Southern blot and were previously described in a study of the relationships between vitamin D receptor genotype and osteocalcin (Morrison et al . 1992) and bone density (Morrison et al . 1994). A total of 716 genetically unique and unrelated individuals were typed for three RFLPs using polymerase chain reaction (PCR) in this study. The frequencies of Bsm-1, Apa-1 and Taq-1 RFLPs are shown in Table 1 and are similar to that reported previously (Morrison et al. , 1992). Association between RFLPs was significant according to Chi-square (Table 2).

Table 1. Genotype and RFLP frequencies in the test population (n = 716).

Genotype Cou Percent Frequency

Bsm-1

BB 117 16.3 0.404 (-)

Bb 345 48.2 0.596 (+) bb 254 35.5

Apa-1

AA 185 25.8 0.510 (-)

Aa 361 50.4 0.490 (+) aa 170 23.7

Taq-1

TT 267 37.3 0.607 (-)

Tt 335 46.8 0.393 (+) tt 114 15.9

Note: the + and - in the frequency refers to the presence and absence of the site alleles respectively.

Table 2. RFLP genotypes are highly correlated. Contingency table analysis of RFLP genotypes detected by Bsm-1, Apa-1 and Taq-1. Bsm-1 versus Taq-1.

RFLP BB Bb bb Total tt 111 3 0 114

Tt 5 325 5 335

TT 1 17 249 267

Totals 117 345 254 716

Percent of each cell

RFLP BB Bb bb Taq-1 tt 15.5 0.4 0 15.9

Tt 0.7 45.4 0.7 46.8 T 0.1 2.3 34.8 37.3

Bsm-1 16.3 48.1 35.5 100

Total Chi square = 1259, p< 0.0001.

Taq-1 versus Apa-1.

RFLP tt Tt TT Total

AA 108 58 19 185

Aa 3 269 89 361 aa 3 8 159 170

Total 114 335 267 716

Percent of each ce 11.

Rflp tt Tt TT Apa-1

AA 15.1 8.1 2.7 25.8

Aa 0.4 37.6 12.4 50.4 aa 0.4 1.1 22.2 23.7

Taq-1 15.9 46.8 37.3 100

Total Chi square = 611.43, p < 0.0001 Rate of recombination and co-association of RFLP markers is non-random. Of 185 individuals homozygous for AA RFLPs, representing 370 chromosomes, 277 carried the absence of the site Bsm-1 ( B) allele while 93 carried the Bsm-1 presence of the site(j ) allele, representing coassociation of A and B RFLP alleles of 75%. The same rate of association was not observed when the Apa-1 a allele was considered. Of 170 aa genotypes (340 individual chromosomes) only 17 chromosomes did not carry the Bsm-1 b allele, representing 5.0% apparent recombination or 95.0% coassociation of b with a. On the other hand, 254 bb individuals were identified (508 chromosomes) and within this group 78% carried the a allele while 113 chromosomes, or 22.0% carried the A allele. Therefore the genotype of aa is found coassociated with bb at a greater rate than the reciprocal comparison of genotype BB coexisting with AA. These data indicate that the various markers exist at uneven frequencies of coassociation, a fact which is explained by haplotype distribution.

Taq-1 RFLPs were in high disequilibrium with Bsm-1 RFLPs giving coassociation of genotypes of about 98%. The relationship was such that presence of the Bsm-1 site was correlated with absence of the Taq-1 site. When Bsm-1 and Taq-1 were considered on a per chromosome basis (1432 in number) there were only 15 chromosomes representing recombination between Bsm-1 and Taq-1 or a 1.0% recombination frequency. In comparison, on a per chromosome basis there were 196 potential recombinants between Bsm-1 and Apa-1, or 13.7% recombination frequency. Taq-1 RFLPs correlated with Apa-1 RFLPs in the same manner as descibed for Bsm-1 RFLPs: on a per chromosome basis there were 202 potential recombinants, giving a frequency of 14.1%. Further discrimination of genotype is possible using combined RFLP markers. If three dimorphic RFLPs are used, there are 2*3 or 8 possible haplotypes and subsequently 64 possible combinations of haplotypes, if the RFLPs are in equilibrium. Genotypes identified by RFLP can often appear the same but can be derived from different haplotype combinations. As a result there are 27 possible genotypic combinations of three dimorphic RFLPs (3*--).

In any given population it is usual to observe only a restricted subset of genotypes. When Bsm-1, Apa-1 and Taq-1 RFLPs were combined, 20 of the possible 27 genotypes had 6 or fewer people observed and nine of these genotypes had zero representatives. A total of 31 (out of 716) subjects were spread over these 20 infrequent genotypes. In contrast, a total of 686 people or 96% of the population were in the seven remaining frequent genotypes (Table 3) .

Table 3. Genotype distribution using three RFLPs detected by Bsm-1, Apa-1 and Taq-1 in 716 Caucasians

Taq-1 Genotype

Bsm-Apa TT Tt tt n

BBAA 0 3 105 108

BBAa 0 2 3 5

BBaa 1 0 3 4

BbAA 3 55 3 61

BbAa 12 263 0 275

Bbaa 2 7 0 9 bbAA 16 0 0 16 bbAa 77 4 0 81 bbaa 156 1 0 157

Totals 267 334 114 716

Note: Bsm-l/Apa-1 genotypes are listed in the first column with the last column giving the number of each Bsm-l/Apa-1 genotype observed. The last row shows the number of Taq-1 genotypes observed. These data demonstrate that certain haplotype combinations of the vitamin D receptor gene are not observed in a large population of Caucasians.

The coassociation of RFLP markers provides the basis for defining frequent haplotypes of the vitamin D receptor gene. In the absence of family data, haplotypes can be inferred from the most frequent homozygous genotypes (Table 4) .

Table 4. Four most frequent haplotypes and frequencies in the test population as homozygote and heterozygote combinations.

Haplotypes derived from homozygotes

RFLP array n percent

Haplotype 1 b a T 156 21.8

Haplotype 2 B A t 105 14.7

Haplotype 3 b A T 16 2.2

Haplotype 4 B a t 3 0.4

Total homozygotes 280 39

Heterozygotes number percent

1,2 (or 3,4) 263 37.2

1,3 77 10.7

2,3 55 7.7

1,4 7 1.0

2,4 3 0.4

Total heterozygotes 405 56.5

Total genotypes explained by four frequent haplotypes: 685/716 or 95.7%. Rare haplotypes observed :as homozygotes :per chromosome.

Haplotype 5 B a T 1 4/1432

Haplotype 6 b A t 0 3/1432

Haplotype 7 B A T 0 3/1432

Haplotype 8 b a t 0 1/1432

Bsm-Apa n Percent Haplotype

BBAA 109 15.1 2,2

BBAa 5 2,4

BBaa 4 4,4

BbAA 61 8.4 2,3

BbAa 276 38.3 1,2

Bbaa 9 1,3 bbAA 16 2.2 3,3 bbAa 81 11.3 1,3 bbaa 159 22.8 1,1

Totals 720 98.1

Four different homozygotes were observed at greater than one subject per 716 unique Caucasian individuals. This defines four haplotypes of the VDR gene which can explain 95.7% of the population. Using three frequent haplotypes explains 79.9% of the population. The heterozygote BbAaTt (37.1%) is comprised mostly 1,2 heterozygotes. Heterozygotes of 3,4 haplotypes give the same genotypic pattern but would only exist at 1.8% of the population, as judged by the Hardy Weinberg binomial distribution. This means that only 13 of the 266 BbAaTt genotype subjects can be expected to be haplotype 3,4 heterozygotes rather than haplotype 1,2 heterozygotes.

One other rare haplotypes was seen as a homozygote (one individual) and several others were observed as heterozygotes (Table 3). These genotypes are a minor proportion of the Caucasian sample, and may be the result of meiotic crossover, although it is possible they relate to minor ethnic admixture in Caucasians.

The order of markers is Bsm-1, Apa-1 and Taq-1 with the markers in close proximity ( lkb and 400bp between the RFLPs respectively). Despite this, Bsm-1 and Apa-1 have about 80% coassociation of genotype while Bsm-1 and Taq-1 (flanking the Apa-1 site) have a high 98% coassociation of genotype and low recombination rate (<2%). Of 236 chromosomes of BB homozygotes, only 13 were recombinants between the Bsm-1 B marker and Apa-1 A marker (5.5%). Of 512 chromosomes from bb homozygotes, there were 93 recombinants (18.1%) between the b allele and the a allele of the Apa-1 RFLP. If genetic crossover between Bsm-1 and Apa-1 was producing the recombinants, one would expect equal numbers and greater recombination with the flanking Taq-1 marker.

Haplotype analysis of the vitamin D receptor gene revealed only four frequent haplotypes (see Figure 1 A) of a possible 8 using three RFLPs (Morrison et al . 1994a).

Of 27 possible genotypes only seven were needed to explain 96% of the population (Morrison et al. 1994a). The relationship between these genotypes defined by haplotypes and bone density traits were examined. Study 2. The use of VDR gene haplotypes to identify subjects with genetically influenced high or low bone density. Methods.

Subjects. Only females are used in this study. Subjects were recruited from the Sydney Metropolitan area by advertisement and from a twin registry. DNA was taken from all subjects. All were given a comprehensive physical examination and those with conditions or treatments known to influence bone were excluded. Of the 815 female volunteers or twins, some of whom have been described previously (Pocock et al. , 1990, Kelly et al. , 1991, Morrison et al. , 1992, 1994), not all subjects had all tests. Bone mineral density was measured on 600. Of a total of 600 (269 pre and 331 postmenopausal) women with bone density measurements, 563 were typed for all three RFLPs, comprising 262 premenopausal and 301 postmenopausal. The total number of women from whom we had assayed osteocalcin, regardless of genotyping, was 349, comprising 175 premenopausal and 174 post menopausal. Of subjects with bone mineral density and osteocalcin data, a total of 323 were typed for all three RFLPs, comprising 169 premenopausal and 152 postmenopausal women. A number of postmenopausal women could not accurately recall the timing of the menopause and these were excluded from analysis involving menopausal years. Osteocalcin assay. Serum was separated after clot and stored at -70 C prior to assay. The osteocalcin concentration in serum was estimated by radioimmunoassay using an in house polyclonal antisera raised against purified ovine osteocalcin. This assay has a normal range of 2 to 18 ng/ml. Serum samples were assayed once only for osteocalcin in the majority of cases. Where repeat osteocalcin data were available, the mean value was taken. Statistics. Analysis of variance (ANOVA) , multiple regression and stepwise multiple regression were used to compare parametric variables . Results pertaining to osteocalcin are presented based on In (1 + osteocalcin) osteocalcin values as previously described and on raw osteocalcin values. Natural logarithm transformed, In (1 + osteocalcin), in both parametric and non-parametric analysis (Kruskal-Wallis) gave similar results.

Significance values presented are for the overall effect in the ANOVA analysis and as the pairwise comparisons of means under Fishers' protected least significance difference test (PLSD) . Students' t-test was used to compare the pairwise means of individual groups. Pearsons correlation was used to meaure the relationship of variables. The Statview+graphics system by Abacus Inc. (Berkeley, California, USA) was run on a Macintosh IlSi. Both twin and non twin populations were available and these were combined to cover all subjects recruited.

In several cases we included data on a singleton of a twin pair (due to dropout of the co-twin) and there were 34 females recruited from male/female twin pairs. Of females there were 251 from the twin group (including 65 MZ pairs) and 534 others.

Several statistical approaches are possible to permit analysis of such a population. We previously showed that the effect of VDR alleles on bone density is not dependent or related to zygosity, suggesting that lumping all subjects is valid. The populations (MZ twin, DZ twin or non-twin) were not significantly different as assessed by two factor ANOVA, when menopause was included. Statistical treatments that define different populations usually rely on a comparison of variances, which in the case of bone density were not different between twin and non-twin groups. A second approach is to include singletons of twin pairs as surrogate individuals. This approach was not taken as the choice of a singleton produces a large number of possible combinations. The simplest approach is to lump all subjects and ignore zygosity, which we have done. However we have taken the precaution of repeating the analysis on the non-twin population, and although not quoted, the conclusions and significant relationships described herein hold when the non-twin group is analysed alone. Bone density in relation to a fracture threshold.

Population effects on bone density can be analysed in relation to a critical value of bone density at which fracture risk is heightened. We chose a value of two standard deviations below the mean of young normal females (30 to 45 years, n = 70) derived from this population (1.2 ± 0.144 and 0.95 ± 0.125 gm/cm² for LS and FN BMD respectivley) . These values, 0.912 and 0.70 gm/cm², for the lumbar spine and femoral neck respectively, are almost identical to the mean BMD of fracture patients observed in an Australian epidemiology study as published previously (see Morrison et al 1994) and corresponds to similar values from the Mayo Clinic, Rochester, Minnesota, USA. An alternative value is -2.5 SD below the mean of young normal (WHO 1994) giving values of 0.84 and 0.64 gm/cm2. Our analysis was based on the estimate of how many years after the menopause are required for the mean bone density of the group to decrease to the same value as the fracture threshold.

Serum calcium. Serum calcium was measured by standard hospital automated blood chemistry procedures. Results.

Menopause effect on osteocalcin in relation to genotype. We previously described an association between osteocalcin in serum and vitamin D receptor genotype using 90 individuals (Morrison et al. PNAS 1992). In the new subjects described in this study (n = 267 females), the relationship between RFLP marker and osteocalcin was investigated in the females to confirm the previously described association in a different population. Osteocalcin levels in serum were significantly different according to genotype for both the Bsm-1 and Taq-1 RFLPs, with ANOVA p values for the overall effect of 0.0088 and 0.0059 (p=0.022 on untransformed osteocalcin, n=265 for both), for Bsm-1 and Taq-1 respectively. Apa-1 genotype was not significantly associated with the effect. Since these results replicated our original findings, all subjects were included in subsequent analysis using single RFLPS (see Table 4) and haplotypes (see Table 6).

Table 4. Mean osteocalcin values of different vitamin D receptor genotypes assessed in females.

Bsm-1, ANOVA: F test = 8.2 p = 0.0003 Bsm-1 n Mean SD SEM

BB 59 11.7 6.5 0.8

Bb 165 9.9 7.7 0.6 bb 114 7.5 5.1 0.5

Students t-tests BB vs Bb p = 0.101 BB vs bb p < 0.0001 Bb vs bb p = 0.0053

Apa-1, ANOVA: F test = 2.0 p = 0.13 Apa-1 n Mean SD SEM

AA 83 10.1 6.5 0.7

Aa 161 9.2 7.3 0.6 aa 78 8.0 5.7 0.6

Students t-tests AA vs Aa p = 0.326 AA vs aa p = 0.028 Aa vs aa p = 0.203

Taq-1, ANOVA: F test = 9.4 p<0.0001 Taq-1: n: Mean: SD SEM

TT 118 7.5 5.2 0.5

Tt 164 9.8 7.5 0.6 tt 56 12.1 6.8 0.9

Students t-tests

TT vs Tt p = 0.0065 TT vs tt p < 0.0001

Tt vs tt p = 0.038_ Osteocalcin, VDR genotype and the menopause.

Osteocalcin levels are reported to rise at the menopause. The menopause effect was examined in relation to VDR genotype by two factor ANOVA. Osteocalcin values were significantly different for both genotype and menopausal status as follows: Bsm-1 or Taq-1, p= 0.0001; menopausal status, p=0.0001 (n=337), while Apa-1 (n=321) genotype had a weaker effect on osteocalcin (p= 0.0373) (see Table 5 for mean osteocalcins of Bsm-1 and Taq-1 genotype and menopause groups).

In each case osteocalcin values were significantly different according to menopausal status. There was no evidence of interaction between menopausal status and genotype on osteocalcin. There were 175 premenopausal and 174 postmenopausal subjects that had serum osteocalcin data. The effect of menopause on osteocalcin (premenopausal 7.7 ± 5.3 ; postmenopausal 11.1 ± 7.7, ± SD; p <0.0001), indicated a rise of about 44% over the menopause with a large degree of value overlap. We found similar changes in osteocalcin values across the menopause between the various genotypes. Considering Bsm-1 genotypes, the increases in osteocalcin over the menopause were virtually identical in each genotypic group ( BB, 52%; Bb, 40% and bb, 52%, not shown). In premenopausal subjects, Bsm-1 genotypes had mean osteocalcin values as follows: BB, 9.3+5.0 ng/ml (n= 35); Bb, 8.3+5.7 ng/ml (n=87); bb, 5.9+4.2 ng/ml (n=62). The total magnitude of the genetic effect in both premenopausal and post menopausal groups was essentially identical when expressed as the ratio of the mean osteocalcin of BB and bb genotypes (postmenopausal, BB/bb = 14.3/9.1 or 1.57: premenopausal, BB/bb = 9.5/6.0 or 1.58). The Taq-1 genotype gave similar results. The magnitude of the genetic effect (57.5% from bb to BB) was comparable to the magnitude of the total menopause effect (44%) . The mean premenopausal osteocalcin of the genotype bb (6.0 ± 4.3 SD) , rises after the menopause to a value (9.1 ± 5.4 SD) similar to the premenopausal value for the genotype BB (9.5 ± 5.2 SD) . These data suggest that menopause and genotype effects on osteocalcin serum concentrations are independent of each other and of similar strength.

Table 5. Menopause and genotype effects on osteocalcin serum concentrations.

Premenopausal females. F-test: 5.332, p = 0.0057 Group: n: Mean: SD: SEM

BB 32 9.5 5.2 0.9

Bb 83 8.2 5.8 0.6 bb 57 6.0 4.3 0.6

Students' t-test

BB vs bb; p = 0.0010

BB vs Bb; p = 0.2856 Bb vs bb; p = 0.0141

Postmenopausal female. F-test:4.516, p = 0.0123 Group: n: Mean: SD.: SEM:

BB 27 14.3 7.0 1.3

Bb 82 11.5 9.0 1.0 bb 57 9.1 5.4 0.7

Students' t-test

BB vs bb; p = 0.0003 Bb vs bb; p = 0.14

Bb vs bb; p = 0.0716

Results using Taq-1 genotypes.

Premenopausal females. F-test: 6.8, p = 0.0014 Group: n: Mean: SD: SEM tt 31 9.8 6.1 1.1

Tt 82 8.2 5.4 0.6 T 59 5.8 4.3 0.6

Students' t-test tt vs TT; p = 0.0006 tt vs Tt; p = 0.19

Tt vs TT; p = 0.0052 Postmenopausal female. F-test:5.13, p = 0.0069 Group: n: Mean: SD. : SEM: tt 25 15 . 0 6 . 7 1 . 3

Tt 82 11 . 2 9 . 0 1 . 0

TT 59 9 . 3 5 . 5 0 . 7

Students' t-test tt vs TT; p = 0.0001 tt vs Tt; p = 0.055

Tt vs TT: p = 0.13

VDR haplotypes and serum osteocalcin. Haplotype analysis showed significant relationships between minor genotypes and serum osteocalcin levels (Table 4). Haplotype 2,2 homozygotes (BBAAtt, n=49) and 1,1 homozygotes (bbaaTT, n=69) had mean osteocalcin values of 11.7+6.8 and 7.5 +5.0 ng/ml, respectively. The difference between these extremes of the genotypes (56%) was comparable to that seen with Bsm-1 alone. Other genotypes with mean osteocalcin values significantly different from those of the 2,2 genotype were the 1, 3 heterozygotes (n=32, mean osteocalcin 6.7+5.2 ng/ml) and 2,3 heterozygotes (n=21, mean osteocalcin 6.7+4.5 ng/ml), which were not significantly different from and virtually identical to those of the 1,1 homozygotes. Therefore, the less frequent haplotype 3 is associated with a low osteocalcin phenotype in the presence of both haplotypes 1 and 2. It would be expected from this that haplotype 3,3 homozygotes would have a low osteocalcin phenotype. Only seven of these homozygotes were observed and while the mean osteocalcin of this genotypic grouping was 8.8+7.6 ng/ml it was not significantly different from the other genotypes. The 1,2 heterozygotes (n=121) had mean osteocalcin (9.9 ± 7.8 ng/ml, +SD) which was significantly different from that of 1,3 and 1,1 genotypes but was not significantly different from the osteocalcin levels of the 2,2 homozygote.

Using two rather than three RFLPs, either Bsm-1/Apa- 1 or Apa-l/Taq-1 combinations, the conclusions did not change.

In Figure 1 the relationship between genotype and menopause is shown for the most frequent haplotype genotypes. Menopause and genotype were again significant predictors of serum osteocalcin in two factor ANOVA analysis (menopause F score, 18.148, p = 0.0001; genotype F score 4.291, p = 0.0009) (see Table 6).

Table 6. Statistical analysis of mean osteocalcin values in different genotypic groupings using a two RFLP , nine genotype system with either Bsm-1 and Apa-1 or Apa-1 and Taq-1.

Bsm-1 Apa-1 genotypes

Genotype n Mean SD SEM p vs BBAa p vs BbAa

BBAA 5 511 1 111..66 6 6..77 0 0..99

BBAa 4 4 1 100..77 7 7..00 3 3..55

BBaa 1 1 9 9..33

BbAA 2 233 7 7..44 5 5..22 1 1..11 **

BbAa 1 12255 9 9..88 7 7..77 0 0..77

Bbaa 6 6 1 122..88 9 9..66 3 3..99 bbAA 7 7 8 8..88 7 7..66 2 2..99 bbAa 3 344 7 7..11 5 5..11 0 0..99 ** * bbaa 7 700 7 7..44 4 4..99 0 0..66 ** *

significant at confidence level: *, 99%; **:99.9% .

Repeat analysis using Apa-1 and Taq-1.

Genotype n Mean SD SEM p vs AAtt p vs AaTt

AAtt 49 11.7 6.8 0.9

AATt 2 13.3 9.4 6.7

AATT 1 9.3

Aatt 25 7.2 4.5 0.9 **

AaTt 123 9.9 7.7 0.7

AaTT 6 12.2 10.1 4.1 aatt 9 10.1 7.7 2.6 aaTt 36 6.7 5.0 0.98 ** ★ aaTT 70 7.5 4.9 0.6 ** *

significant at confidence level: *, 99%; **:99.9% Haplotype and the menopause. The menopause effect was examined in relation to haplotypes (Table 7 and 8). The extremes of the genotype effect were again comparable to the menopause effect, such that the osteocalcin level in the high osteocalcin genotype, the haplotype 2,2 homozygotes, (BBAAtt) was 9.3 ng/ml, prior to the menopause was comparable to that of low osteocalcin genotypes after the menopause (BbAATt, 9.4 ng/ml; bbaaTT, 9.1 ng/ml) .

Table 7. Mean osteocalcin values according to menopause and genotype using a four haplotype, nine genotype system with either Bsm-1 and Apa-1 or Apa-1 and Taq-1.

Premenopausal Postmenopausal

Genotype n: Mean SD SEM n: Mean: SD SEM Δ%

BBAA 29 9.5 5.4 1.0 22 14.4 7.3 1.5 51

BbAA* * 14 5.4 3.7 1.0 9 10.6 5.8 1.9 96

BbAa 64 8.4 5.4 0.7 61 11.2 9.4 1.2 33 bbAb* 17 6.5 5.4 1.3 17 7.7 4.9 1.2 18 bbbb* * 36 6.0 3.8 0.6 34 8.9 5.6 1.0 48

* significantly different from BBAA genotype only.

Premenopausal Postmenopausal

Genotype n: Mean SD SEM n: Mean: SD SEM Δ%

AAtt 28 9.3 5.4 1.0 21 14.8 7.2 1.6 59

AATt** 16 6.4 4.4 1.1 9 8.7 4.5 1.5 36

AaTt 62 8.6 5.4 0.7 61 11.2 9.4 1.2 30

AbTT* 19 6.1 5.2 1.2 17 7.5 4.8 1.2 23 aaTT* * 36 6.0 3.8 0.6 34 9.0 5.5 1.0 50

* significantly different from AAtt genotype ! only

Table 7B. Serum osteocalcin levels in the six frequent genotypic groupings and significance of differences, disregarding the menopause.

Genotype: n: Mean SD SEM

BBAAtt 49 11.7 6.8 1.0

BbAATt 21 6.7 4.5 1.0

BbAaTTl 4 5.0 3.2 1.6

BbAaTt 121 9.9 7.8 0.7 bbAATT 7 8.8 7.6 2.9 bbAaTT 32 6.7 5.2 1.0 bbaaTT 69 7.5 5.0 0.6 Significant differences between genotypes. Genotype BbAATt bbAaTT bbaaTT BBAAtt 0.0032 0.0013 0.0002 BbAaTt 0.0715 0.0442 0.0195

Notes: serum osteocalcin levels between genotypic grouping were assessed by pair-wise Students' t-tests. p value for the overall effect by ANOVA was 0.0007. 1 not included in ANOVA analysis due to low numbers and uneven menopause distribution.

Table 8. Serum osteocalcin levels in the six most frequent genotypic groupings using three RFLPs and according to the menopause.

Premenopausal Pos1tmenopausal

Genotype n Mean SD SEM n Mean SD SEM Signif

BBAAtt 28 9.3 5.4 1.0 21 14.8 7.2 1.6 O.0038

BbAATt 14 5.4 3.7 1.0 7 9.4 5.0 1.9 0.052

BbAaTt 61 8.6 5.5 0.7 60 11.3 9.5 1.2 0.0578 bbAATT 4 3.6 4.1 2.1 3 15.7 4.7 2.7 0.0155 bbAaTT 16 6.4 5.5 1.4 16 7.5 5.0 1.2 0.5484 bbaaTT 36 6.0 3.8 0.6 33 9.1 5.6 1.0 0.0084

Sig: refers to the p value of Students' t-test between pre and postmenopausal groups.

Bone mineral density and haplotypes of the VDR gene. Bone density in premenopausal subjects was not significantly related to age. This enables a simple ANOVA analysis of the effect of genotype on bone density to be used to distinguish differences in genotypic groupings. The five most frequent genotypes were analysed, corresponding to haplotype combinations: 1,1; 1,2; 2,2; 1,3; 2,3 (Figure 2 and Table 9 below) .

Table 9. Haplotype analysis of bone mineral density in premenopausal subjects.

LS BMD ANOVA in premenopausal subjects.

Group: Count; Mean: Std. Dev, Std. Errors

2,2 48 1.116 0.127 0.018

2,3 19 1.185 0.122 0.028

1,2 104 1.173 0.151 0.015

3,3 4 1.311 0.1 0.05

1,3 25 1.189 0.16 0.032

1,1 50 1.275 0.115 0.016

Significant differences. Comparison: Diff, Fisher PLSD

2,2 vs . 1,2 -0.057* .048

2,2 vs. 2,3 -0.195* .142

2,2 vs . 1,3 -0.074* .067

2,2 vs . 1,1 -0.16* .055

2,3 vs. 1,1 -0.091* .073

1,2 vs. 1,1 -0.102* .047

1,3 vs . 1,1 -0.086* .067

* Signifies significant difference,

Insufficient numbers of 3,3 homozygotes were present and this genotype was excluded. The analysis demonstrates highly significant differences in the mean lumbar spine BMD between the different genotypes. The most important result is that the total genetic effect increases when using combined RFLP markers, as judged by the absolute difference in the means of the extremes (that is genotype 1,1 compared with 2,2). The difference in the means 0.159 g /cm², which is greater than one standard deviation (0.14 gm/cιrι2) for young normals of this population. This value is larger than that obtained using Bsm-1 (0.125 gm/cm ) or Taq-1 (0.138 gm/cm²) RFLPs alone. The low bone density genotype (BB or tt) is not overly affected by using extra RFLP markers or haplotypes. The increased genetic effect is realised by Apa-1 RFLPs splitting the high BMD group into two groups (haplotypes 1,1 and 1,3) one of which is revealed as an extra high BMD group (mean BMD of 1.275 gm/cm²). This group is haplotype 1,1, defined as genotype bbaaTT, but also detectable as genotype bbaa or aaTT, due to the high co-association of Bsm-1 and Taq-1 RFLPs.

The relationship between bone mineral density and osteocalcin. In simple regression in all female subjects osteocalcin was significantly related to bone mineral density at the lumbar spine and the femoral neck (p = 0.0001 for both sites against osteocalcin, Table 10). Although this association was significant, such that subjects with higher serum osteocalcin tended to have lower bone mineral density, the amount of the variance in bone mineral density explained by this variable was low

(r² value of 0.068 for lumbar spine and 0.044 for femoral neck) . This supports previous twin studies that related higher osteocalcin values with lower bone density (Kelly et al) .

Table 10. Regression analysis of the relationship between osteocalcin and bone mineral density at the lumbar spine (A) and the femoral neck (B) .

LUMBAR SPINE n:346

R = 0.261

R² = 0.068

Adjusted.R² = 0.065

RMS Residual = 0.173

Analysis of Variance Table

Source DF: Sum Mean Square: F-test: Squares:

REGRESSION 1 0.754 0.754 25.048 RESIDUAL 344 10.35 0.03 p = 0.0001 TOTAL 345 11.104

Beta Coefficient Table

Std. Err.: Std. Coeff.: t-Value: Probability!

INTERCEPT = 1.263

SLOPE -0.066 -0.066 - 0.013 0.261 - 5.005 p = 0.0001

Confidence Intervals Table

Variable: 95% Lower: 95% Upper: 90% Lower: 90% Upper:

MEAN (X,Y) 1.103 1.14 1.106 1.137 SLOPE -0.093 -0.04 -0.088 -0.045 B.

Regression results for osteocalcin and femoral neck BMD, n: 347

R:0.209

R²:0.044

Adj. R²:0.041

RMS Residual:0.141

Analysis of Variance Table Source DF: Sum Squares: Mean Square: F-test:

REGRESSION 1 0.31 0.31 15.801 RESIDUAL 345 6.84 0.02 p = 0.0001 TOTAL 346 7.16

Beta Coefficient Table

Coefficient: SE Std. Coeff t-Value: Probability:

INTERCEPT - 0.901 SLOPE -0.004 0.001 -0.209 3.975 0.0001

Confidence Intervals Table Variable: 95% Lower: 95% Upper: 90% Lower: 90% Upper:

MEAN (X,Y) 0.845 0.875 0.847 0.872 SLOPE -0.007 -0.002 -0.006 -0.003

Beta Coefficient Table

Final equations:

Lumbar spine BMD = 1.263 - 0.066 osteocalcin,

Femoral neck BMD = 0.95 - 0.043 osteocalcin. Using a multi-variate model. To address the effect of menopause on the relationship between osteocalcin and BMD, menopausal status (coded as zero and one) was included in multiple regression analysis, with the result that both variables were significantly associated with bone mineral density (regressed against lumbar spine BMD: p = 0.0001 and 0.003 for menopausal status and osteocalcin respectively, r² = 0.302). No interaction between menopausal status and osteocalcin effect on lumbar spine BMD was found, demonstrating that these factors are independent variables. Bone mineral density is affected by anthropomorphic factors such as height and weight and life variables such as age and years post menopause (YPM) . These variables were included with osteocalcin in multiple regression analysis against BMD at the lumbar spine and the femoral neck. Since BMD at the lumbar spine is not associated with age in premenopausal females (p = 0.99) whereas it is in postmenopausal females (p = 0.0001), age was included with YPM with premenopausal females being coded as zero YPM. Surprisingly, osteocalcin was the strongest predictor of lumbar spine BMD in this analysis (p = 0.0001) and was also a significant predictor of femoral neck BMD (p = 0.0014). All variables were significantly associated with FN BMD, while height was the only variable not significantly associated with lumbar spine BMD. Results are presented in Table 11 below.

Table 11. Multiple regression analysis of variables associated with bone mineral density. n: 338

R:0.536

R-squared:0.288

Adj. R-squared:0.277

RMS Residual:0.151

Analysis of Variance Table

Source DF: Sum Squares: Mean Square: F-test:

REGRESSION 5 3.039 .608 26 .806

RESIDUAL 332 7.529 .023 P = 0.0001

TOTAL 337 10.568

Beta Coefficient ' Table

Variable: Coefficient: Std. Std. t-Value: Prob:

Err. : Coeff. :

INTERCEPT 0.776

Ht 0.002 0, ,001 0.081 1.596 0.1114

Kg 0.003 0, .001 0.171 3.507 0.0005

OC -0.005 0, .001 -0.203 4.337 0.0001

YPM -0.005 1, .967 1.65 2.957 0.0033

AGE -0.003 0, .001 -0.191 2.536 0.0117

Confidence Intervals and Partial . F Table

95% 95% i ?0% 90% Partial F

Ht -0.001 0.005 -7.3 x lO-⁵ 0.004 2.548

Kg 0.001 0.005 o. .002 0.005 12.298

OC -0.008 -0.003 -0 .007 -0.003 18.808

YPM -3.875 3.866 3.25 3.241 8.746

AGE -0.005 -0.001 -0 .004 -0.001 6.431 Osteocalcin and genotype in regression analysis.

Vitamin D receptor genotype is significantly associated with both osteocalcin and bone mineral density as described above. Osteocalcin is inversely related to bone mineral density, and genotypes with high osteocalcin have lower bone density. As a result of these relationships it is possible that the correlation of osteocalcin with bone mineral density results from entirely from the genetic effect. This was tested by including genotype and serum osteocalcin in multiple regression against bone mineral density, incorporating age, heighty and weight variables. Surprisingly, osteocalcin was a strong predictor of bone mineral density in the presence of vitamin D receptor genotype (n=335) in the multiple regression (p = 0.0004 and 0.0015 for osteocalcin and Bsm-1 respectively). The same result held for Taq-1 (p = 0.0004 and p = 0.003 for osteocalcin and Taq-1 respectively) while Apa-1 again gave a weaker effect (p = 0.0004 and p = 0.055 for osteocalcin and Apa-1 respectively, n=318). Using haplotype based genotypes, osteocalcin and genotype were independent predictors of lumbar spine bone density (p = 0.0016 for osteocalcin and p = 0.0039 for genotype, n=318). Table 12 below shows the results of using single RFLP markers and Table 13 shows the results of using haplotype relationships.

Table 12. Multiple regression analysis of lumbar spine BMD against vitamin D receptor RFLPs and other variables using single RFLP genotypes.

Bsm-1. Intercept = 0.583

Variable Std. Coeff. Partial F p value

Bsm-1 0.16 20.0 0.0001

Height 0.09 5.3 0.022

Kg 0.17 19.2 0.0001

AGE 1.648 13.0 0.0003

YPM -0.243 16.6 0.0001

Apa-1 Intercept = 0.615

Variable Std. Coeff. Partial F p value

Apa-1 0.133 13.3 0.0003

Height 0.09 4.5 0.0347

Kg 0.17 19.3 0.0001

AGE 1.65 8.4 0.0039

YPM -0.29 22.6 0.0001

Taq-1 Intercept = 0.598

Variable Std. Coeff. Partial F p value

Taq -0.16 20.3 0.0001

Height 0.09 5.2 0.0235

Kg 0.16 18.3 0.0001

AGE 1.65 13.0 0.0003

YPM -0.24 16.3 0.0001 Table 13. Multiple regression analysis of lumbar spine BMD with combined RFLP systems using either Bsm-1 or Taq- 1 with Apa-1. n = 540 R= 0.56 R² = 0.31 Intercept = 0.625 Lumbar spine BMD =

Variable Coefficients Stand.Coeff F score p value

Bsm-l/Apa-1 0.011 0.16 18.7 0.0001

Weight 0.002 0.09 4.6 0.0329

Height 0.003 0.17 19.8 0.0001

Age -0.003 1.65 9.2 0.0026

YPM -0.006 0.28 21.8 0.0001

Results for Apa- 1, Taq-1 genotypes. n = 540

R = 0.557

R² = 0.31

Intercept = 0.628

Beta Coefficient Table

Coefficients Stand.Coeff . F score p value

Apa-l/Taq-1 0.011 0.16 18.7 0.0001

Weight 0.002 0.09 4.6 0.033

Height 0.003 0.17 19.8 0.0001

Age -0.002 1.65 6.62 0.0025

YPM -0.006 -0.28 24.3 0.0001 Analysis of the relationship between serum osetocalcin, VDR genotype and bone density in premenopausal and post menopausal women analysed separately. The results suggest that serum osteocalcin is a predictor of bone density, despite the presence of a genotype variable that is strongly correlated with osteocalcin. The anthropomorphic, genotype and osteocalcin variables were examined for their effect on BMD separately in pre and post- menopausal groups. The approach taken was both multiple and stepwise regression. In premenopausal females (n = 171) only genotype (Bsm-1, Taq-1, or haplotypes, with p values of 0.0001 for each genotype mode assessed separately) and weight (p = 0.0051) were significant with respect to lumbar spine BMD. Osteocalcin was therefore not a significant predictor of bone density in premenopausal females (not displayed in a table) .

When femoral neck BMD was considered the significant variables were genotype, weight and age with p values of 0.0014 (Taq-1), 0.0001 and 0.0037, respectively. P values were similar using the other genotype formats. Apa-1 was not a significant variable for bone density at either the lumbar spine or the femoral neck. Serum osteocalcin was not a significant variable in the equation, by multiple or stepwise regression, even when genotype was removed (n = 264). Of the variables considered, only genotype and weight were significant in multiple regression against femoral neck BMD.

When postmenopausal women were analysed separately the result was different. Stepwise regression incorporating genotype, osteocalcin, height, weight, age and YPM was repeated in the postmenopausal subjects (n=157, see Table 14 below). This analysis gave the unexpected result that osteocalcin was the strongest predictor, followed by weight and height, while genotype, age and YPM were not included in the equation. Incorporating all variables in multiple regression gave the following rank order of variable strength: osteocalcin (p = 0.0081) and weight (p = 0.0177), then non-significant variables genotype (Bsm-1, p = 0.37), age (p = 0.97) and YPM (p = 0.55).

Table 14. Stepwise regression results in postmenopausal females. Only height, weight and osteocalcin are significant predictors of lumbar spine BMD. Variables in Equation

Variable: Coefficient: Std. Std. F/Remove:

Err. : Coeff. :

INTERCEPT = 0 .038

Osteocalc -0.005 0.0002 -0.219 8.7 in

Height 0.005 0.002 0.188 5.7

Kg 0.004 0.001 0.217 7.6

Variables Not in Equation

Variable •• Par.Corr: F to Enter ••

BSM 0.076 0.872

AGE -0.076 0.882

YPM -0.099 1.509

In postmenopausal women, removal of osteocalcin from the equation (n=293) had the effect of making other variables significant: weight (p = 0.0002), Bsm-1 (p = 0.043) and YPM (p = 0.0005) with an R-squared of 0.186.

When lumbar spine bone density in postmenopausal females was analysed in multiple regression against weight, height, age, YPM and genotype (Bsm-1) only three variables were significant (n=262 subjects): weight (p = 0.0007), YPM (p = 0.001) and Bsm-1 (p = 0.023). In the smaller group (n=126) for whom we had osteocalcin data, Bsm-1 was still significant (p=0.0372) while weight and YPM were no longer significant. When osteocalcin data was included in the regression (n=126), the genotype effect was weakened (Bsm-1, p = 0.08) while osteocalcin was highly significant (p = 0.0082).

In summary, in the total subject population multiple regression of variables against lumbar spine bone mineral density showed an unexpected ability of osteocalcin and genotype to coexist as variables in the equation. In premenopausal females genotype eliminated osteocalcin from the equation while in postmenopausal women osetocalcin eliminated genotype from the equation. Relationship between osteocalcin and bone density in different genetic subtypes. In multiple regression analysis, osteocalcin was independent of genotype, despite the fact that osteocalcin is associated with genotype. This suggested a difference in the behaviour of osteocalcin in different genetic groups. This was analysed by examining the correlation between osteocalcin and BMD within the genotype (Bsm-1) groups and in menopausal categories (using a single RFLP criterion, Table 15 below) .

Table 15. Simple correlation coefficients between osteocalcin and lumbar spine BMD within genotypic groupings .

Bsm-1 only: all subjects

Genotype n Covariance Correlation R-squared

BB 59 -0.139 -0.123 0.015

Bb 165 -0.298 -0.219 0.148 bb 112 -0.352 -0.375 0.140 Bsm-1 only: premenopausal subjects

Genotype n Covariance Correlation R-squared

BB 32 -0.040 -0.060 0.004 Bb 83 -0.044 -0.059 0.003 bb 57 -0.007 -0.013 0.0001

Bsm-1 only: postmenopausal subjects

Genotype n Covariance Correlation R-squared

BB 27 -0.066 -0.051 0.003 Bb 82 -0.254 -0.167 0.028 bb 55 -0.365 -0.401 0.161

Haplotype systems: Combined Bsm-1 Apa-1 groups.

Genotype n Covariance Correlation R-squared

BBAA 51 -0.059 -0.051 0.003

BBAa 4 -1.048 -0.972 0.944

BBaa 1

BbAA 23 -0.197 -0.253 0.064

BbAa 125 -0.24 -0.176 0.031

Bbaa 6 -0.692 -0.381 0.145 bbAA 7 -0.89 -0.555 0.308 bbAa 32 -0.16 -0.163 0.027 bbaa 70 -.404 -0.467 0.218

Total 319 -.326 -.266 .071

The strongest correlation of osteocalcin to bone mineral density occurs within postmenopausal subjects of genotype bb , characterised by low osteocalcin and high BMD. In this group the correlation coefficient was -0.401 compared to 0.071 for the entire population. This indicates that even within the genetic group with high BMD, variation of BMD within that group is strongly associated with osteocalcin. An intermediate effect was seen with the heterozygote Bb (r = -0.167), while in the high osteocalcin, low BMD group no correlation was apparent.

Again the genetic effect was superior with haplotype based genotypes. The effect was increased using haplotype based genotypes, notably for haplotype 1,1 homozygotes (the extreme high bone mineral density genotype) where the correlation improved from -0.401 (Bsm-1 bb group only) to -0.467 while the r improved from 0.140 to 0.218.

These data demonstrate that the relationship between osetocalcin and bone density is much stronger in certain genotypes than others and shows that the information content of osteocalcin determinations in a clinical setting can be dramatically enhanced with VDR haplotype allele genotyping. Regression analysis was used to determine the significance of these relationships (Table 16 below) .

Table 16. Regression of osteocalcin against lumbar spine BMD according to genotype and menopause.

A. Postmenopausa .1

Genotype n R R2 F score p value

BB 27 0.051 0.003 0.07 0.8

Bb 82 0.167 0.028 2.3 0.13 bb 55 0.401 0.161 10.2 0.0024

Total 164 0.203 0.041 6,98 0.0091

B. All subjects •

Genotype n R R F score p value

BB 59 0.123 0.015 0.9 0.4

Bb 165 0.219 0.048 8.3 0.0046 bb 112 0.375 0.140 17.9 0.0001 C. Regression of osteocalcin and bone density in all female subjects regardless of genotype.

Subject n R R² score p value

Pre 175 0.093 0.009 1.5 0.22

Post 172 0.197 0.039 6.88 0.0096

Total 347 0.266 0.071 26.4 0.0001

Testing simple regression of osteocalcin against LS BMD in each haplotype based genotype (data presented in Table 17 below) , the only significant regression was obtained with the extreme high BMD group (1,1 homozygotes) in the postmenopausal females (p = 0.001). Multiple regression within this group (r² = 0.405) gave osteocalcin as the strongest variable (p = 0.0023, standardised coefficient of -0.318) with age the only other significant variable. A similar result was obtained for the femoral neck BMD but with a weaker effect of osteocalcin (p =

0.0512). Osteocalcin was not a significant predictor of bone density at either the lumbar spine or femoral neck, in the other major genotypes, 2,2 and 1,2. In stepwise regression of these variables in the minor genotypes osteocalcin was not a significant predictor of bone density at either the lumbar spine or the femoral neck .

Table 17. Regression of multiple variables against BMD in homozygote haplotype 1,1 and 2,2; postmenopausal subjects .

Haplotype 1,1 Homozygotes multiple regression p values

LS result: n = 68 R² = 0.405 F score = 8.45 p = 0.0001 FN result: n = 68 R² = 0.543 F score = 14.7 p = 0.0001

Variable LS p value FN p value

OC 0.0023 0.0512

Height 0.7369 0.3836

Kg 0.3062 0.0014

AGE 0.0382 0.0027

YPM 0.2049 0.0404

Haplotype 2,2 homozygotes multiple regression.

LS result: n = 51 R² = 0.327 F score = 4.38 p = 0.0025 FN result: n = 51 R² = 0.457 F score = 7.584 p = 0.0001

Variable LS p value FN p value

OC 0.98 0.21

Height 0.68 0.0112

Kg 0.85 0.60

AGE 0.64 0.56

YPM 0.0475 0.072

The strong correlation between osteocalcin and BMD in the high BMD genotype (homozygote haplotype 1, 1) (see Fig.3) is not simply the result of a coordinated increase in osteocalcin with YPM coupled with a decrease in BMD. Although BMD in this genotype does decrease with YPM (r² = 0.102) there was no relationship between YPM and osteocalcin (r² = 0.002). Rather the relationship between osteocalcin and BMD was not dependent on age or YPM and osteocalcin was the only variable which was significant in multiple regression against LS BMD in this group. Intermediate effects are observed in the minor genotypes (heterozygote 2,3 and 1,3), judged by the correlation coefficients described above in Table 13.

Age and menopause related changes in bone mineral density as a function of genotype and the fracture threshold. Genotype effects are particularly strong in premenopausal females, where age related changes in bone density are minimal. Age related changes in lumbar spine BMD were not apparent in premenopausal females, regardless of genotype when assessed by regression. Premenopausal and postmenopausal women were analysed separately to eliminate possible age related cohort effects.

A prevalent concept in osetoporosis research is the concept of a fracture threshold, which is a critical value of bone density at which heightened risk of disease is apparent. Postmenopausal women were analysed for changes in bone density in relation to the fracture threshold (descibed above in the analysis section) using YPM as well as age. Discriminating genetic groups using Bsm-1 RFLPs alone, the results are similar to those reported previously (Morrison et al . Nature 1994) with intercepts with the fracture threshold for the genotypes of: 18.7, 32.8 and 33.7 YPM for the lumbar spine and 17.9, 24.2 and 23.7 YPM for the femoral neck for genotypes BB, Bb and bb, respectively. In this analysis, there is a 15 year difference in the timing when, on average, the genotypic groupings reach a critical value of lumbar spine bone density. When analysed in this way, there is little difference between bb homozygotes and the Bb heterozygotes. The most striking difference between this analysis and that done previously (Morrison et al . 1994), where premenopausal subjects were included, is that the slope of the regression relationship is different between the different genotypes. The BB genotype, which is associated with low bone density in premenopausal subjects, has an age related decline in bone density almost twice that of the of bb genotype which is associated with high bone density. The heterozygote Bb has a slope similar to that of bb genotype. These data raise the suggestion that a difference in the rate of change of bone density after the menopause may exist in these genotypes . Since the Bb and bb genotypes have similar rates, these data demonstrate that genetic variants of the VDR can be used to define categories of postmenopausal bone loss: "fast-losers" and "slow-losers". The difference in the slope was similar for the femoral neck (data presented below in Table 18).

Again the use of haplotype information was superior to a single RFLP. Using haplotypes resulted in further discrimination of the genetic effect, with an increase in the magnitude of the genetic effect and the difference in the fracture threshold intercept between genotypes (data presented below in Table 18). For haplotype 1,1 homozygotes, the result was 40.4 YPM before the mean genotype LS BMD reached the critical fracture threshold. Since the mean menopausal age of all subjects was 48.4 years (this was not different between genotypes), this value translates to an age of 89.6 years before a person of this genotype will, on average, reach this critical fracture threshold. In contrast, the intercept value for haplotype 2,2 homozygotes was 18.3 years. The extremes of the genotypic groupings therefore translate to an average 22.1 year difference in reaching the critical fracture threshold. The 1,2 heterozygotes, had an intercept value of 34.6 years, giving an 16.3 year difference with the low bone density haplotype 2,2 homozygote group (see Fig. 4). Although the relationships between YPM and lumbar spine BMD were significant for all three genotypes (1,1, 1,2 and 2,2; p = 0.01, 0.02 and 0.0002 respectively) the relationship was strongest in the 2,2 genotype which is associated with low bone density. In this group the slope of the relationship (1.2% per annum, r² = 0.283,) was approximately twice that of the other two genotypes (both 0.5% per annum) with 28.3% of the variance in postmenopausal bone density in this group explained by YPM. On the other hand, in the genotypes 1,1 and 1,2 , the relationship between BMD and YPM had lesser slopes. In these two genotypes osteocalcin is a much stronger variable in relationship to postmenopausal bone density than YPM. Essentially the same result was obtained by examining the non-twin population alone.

Table 18. Regression analysis of bone density versus years post menopause in different genotypic groups.

Lumbar spine

Genotype Regression equation R square Intercept

BB BMD = 1.099 - 0.010 YPM r2 = 0.238 18.7 Bb BMD = 1.076 - 0.005 YPM r2 = 0.059 32.8 bb BMD = 1.114 - 0.006 YPM r2 = 0.136 33.7

Femoral neck.

Genotype Regression equation R square Intercept

BB BMD = 0.897 -0.011 YPM r2 = 0.384 17.9 Bb BMD = 0.845 -0.006 YPM r2 = 0.164 24.2 bb BMD = 0.866 -0.007 YPM r2 = 0.272 23.7 Apa-1 Taq-1 combined genotypes. Lumbar spine

Genotype Regression equation R square Intercept 2,2 BMD = 1.132 -0.012 YPM r2 = 0.283 17.3 1,2 BMD = 1.085 -0.005 TPM r2 = 0.051 34.6 1,1 BMD = 1.114 -0.005 YPM r2 = 0.099 40.4

Femoral neck

Genotype Regression equation R square Intercept

2,2 BMD = 0.908 - 0.012 YPM r2 = 0.426 17.3 1,2 BMD = 0.854 - 0.006 YPM r² = 0.159 25.6 1,1 BMD = 0.864 - 0.007 YPM r2 = 0.264 23.4

Changes in bone density in the near menopause years in different genotypic groupings. It is usually considered that menopause related bone loss is most rapid closer to the menopause. The difference between genotypes in the rate of change of bone density in the postmenopausal years was even greater when the analysis was restricted to less than 20 years after the menopause. Haplotype 2,2 homozygotes (low bone density group) had an intercept with the fracture threshold of 14 years and an apparent rate of decline of 2.0% per annum while for haplotype 1,1 homozygotes the intercept was 50 years and the rate 0.4% per annum and heterozygote 1,2 intercept was >50 years and slope was 0.05%. Indeed in the 1,1 homozygote and 1,2 heterozygote the LS BMD was not significantly related to YPM and slopes not significantly different from zero.

Postmenopausal BMD was not related to osteocalcin in the 2,2 genotype while it had strong relationships with 1,1 homozygotes and 1,2 heterozygotes. For these genotypes, associated with higher bone density, postmenopausal BMD is more related strongly to osteocalcin than YPM. In comparing these extremes of genotype, only osteocalcin was significantly related to postmenopausal LS BMD in the 1,1 genotype group and only YPM was significant in the 2,2 genotype group, when these variables were assessed by multiple regression including height and weight. In this near postmenopausal (<20 YPM) group, osteocalcin was still strongly associated with LS BMD (r² = 0.34) within the haplotype 1,1 homozygotes. These data demonstrate the utility of combining osetocalcin determinations with genotype classification in increasing the information content of these clinical parameters.

As had been seen in premenopausal BMDs, the minor genotypes had intermediate phenotypes . The fracture intercepts of 2,3 and 1,3 heterozygotes were 20.2 and 20.6 YPM, respectively making these genotypes comparable to the 2,2 homozygotes.

These conclusions were drawn from the total study population which were a mixture of twin and non-twin groups. The same overall conclusion on intercept values and differences in menopause related changes in bone density were derived from an analysis were the subjects were constrained to those who were not twins, thereby representing a different population. The intercepts were 40, 30 and 18.5 YPM for haplotype 1,1 homozygotes (n=52), haplotype 1,2 heterozygotes (n=76) and haplotype 2,2 homozygotes (n=30) respectively with a greater difference in apparent slope of decline of bone density in the postmenopausal years (1.4% versus 0.05% per annum) between the extremes of genotype.

When the femoral neck fracture threshold (0.7 gm/cm²) was considered, the relationship was similar, with intercepts as follows: 2,2 homozygotes, intercept 17.9 YPM; 1,2 heterozygotes, intercept 26 YPM; 1,1 homozygotes, intercept 24 YPM. Study 3 .

The effect of VDR genotype on serum calcium levels.

Method.

Normal female subjects had serum taken after overnight fast and serum calcium was measured by standard hospital chemistry according to methods readily available in clinical pathology laboratories. Results.

Serum calcium levels are different across genotypes. Serum calcium levels were mesured on a subset of subjects and correlated with vitamin D receptor genetics (see Fig. 5). In premenopausal subjects genetic effects were not obvious (see Fig. 6). In postmenopausal females a highly significant effect of genotype on serum calcium was observed. In the subset of individuals in which serum calcium was measured the serum calcium result mirrored the findings with bone density, with low bone density groups having lower serum calcium (see Table 19 below) . The high bone density genotypes had higher serum calcium. A similar trend which was not statistically significant was observed when ionised calcium was examined.

Again using haplotypes resulted in a greater difference between the mean value of the variable in question when comparing the extremes of genotype indicating a greater ability to discriminate an effect with haplotype information (Table 20 and Fig. 7). The difference between the means of the extremes went from 0.058 mM Ca in serum to 0.075 mM Ca. Considering that the mean + Standard deviation of all postmenopausal subjects is 2.3391.092, the difference in serum calcium observed here between the extremes of genotype is equal to an increase from 63% to 76.5% of one standard deviation of the test population. Table 19. Demonstration of significantly different serum calcium levels in different genotypes; postmenopausal females using Bsm-1 genotype marker.

Analysis of Variance Table

Source: DF: Sum Squares: Mean Square: F-test:

Between groups 2.087 0.044 5.309 Within groups 143 1.177 0.008 p = 0.006

Total 145 1.265

Model II estimate of between component variance = .018

Group: Count: Mean: Std. Dev. : Std. Error:

BB 23 2.313 0.101 0.021

Bb 71 2.324 0.081 0.010 bb 52 2.372 0.098 0.014

Comparison Mean Diff Fisher PLSD.

BB vs. Bb -0.010 0.043 BB vs. bb* -0.058 0.045 Bb vs. bb** -0.048 0.033

* p = 0.0132, **p = 0.0035

Table 20. Different serum calcium levels in different genotypes as determined by haplotypes: comparing the 5 most frequent haplotype derived genotypes .

Analysis of Variance Table

Source: DF: Sum Squares: Mean Square: F-test:

Between groups 4 0.104 0.026 3.343

Within groups 129 1.002 0.008 p=0.0122

Total 133 1.106 Model II estimate of between component variance = .005

Group: Count: Mean: Std. Dev, Std. Error:

2,2 18 2.307 0. .090 0.021

2,3 15 2.322 0, .054 0.014

1,2 53 2.325 0, .088 0.012

1,3 10 2.326 0, .071 0.023

1,1 38 2.382 0. ,101 0.016

Comparison: Mean Diff. : Fisher PLSD

2,2 vs. 2,3 -0.015 0.061

2,2 vs. 1,2 -0.018 0.048

2,2 vs . 1,3 -0.019 0.069

2,2 vs . 1,1*** -0.075 0.050

2,3 vs. 1,2 -0.003 0.051

2,3 vs. 1,3 -0.004 0.071

2,3 vs. 1,1* -0.060 0.053

1,2 vs . 1,3 -0.001 0.060

1,2 vs. 1,1** -0.057 0.037

1,3 vs. 1,1 -0.056 0.062

*** p = 0.0099, ** p =0.0055, * p = 0.0347

Study 4.

VDR haplotypes have utility in assessing the bone density of male subjects.

Method.

A total of 226 males were genotyped for haplotypes

(Taq-1 or Bsm-1 and Apa-1). Subjects were males from the

Sydney region (Table 21A) who were unrelated to each other and in the second instance all males genotyped (including twins) from Sydney and Dubbo City, NSW, Australia were combined for analysis (Table 22).

Results. Results are presented in Table 21A and 2IB below and in Figures 8 and 9.

Table 21. Significant effect of VDR genotype on bone density in males.

A. Using a single RFLP (Bsm-1).

Analysis of Variance Table

Source: DF: Sum Squares: Mean Square: F-test:

Between groups 2 .158 .079 3.17

Within groups 78 1.946 .025 p = .0475

Total 80 2.104

Model II estimate of between component variance = 0.027

Group: Count: Mean: Std. Dev.: Std. Error:

BB 14 1.089 0.148 0.040

Bb 46 1.177 0.160 0.024 bb 21 1.226 0.159 0.035

Comparison: Mean Diff.: Fisher PLSD

BB vs . Bb -0.088 0.096

BB vs . bb* -0.137 0.109

Bb vs . bb -0.049 0.083 p = 0.04

B. Using haplotype analysis: three most frequent genotypes only with mean differences between genotypes

Group: Count: Mean: Std. Dev. : Std. Error:

2,2 14 1.089 0.148 0.040

1,2 38 1.177 0.164 0.027

1,1 9 1.231 0.197 0.066 Comparison: Mean Diff . : Fisher PLSD

2,2 vs. 1,2 -0.089 0.104

2,2 vs. 1,1* -0.142 0.142

1,2 vs. 1,1 -0.053 0.123

* p = 0.05.

Even with a smaller group of males typed using haplotype analysis, a significant result is observed and the difference between the means again is larger with haplotype based genotypes, 0.142, compared with 0.137 when Bsm-1 alone is used. In a larger group of males genotyped (n=226) there was a highly significant relationship between bone density and genotype, such that a difference between means of extreme genotypes was 0.159 using a single RFLP (Taq-1 or Bsm-1) with a very strong heterozygote effect (Table 22).

Table 22. VDR genotype affects bone density in the total male study group.

A. Using a single RFLP (Taq-1) (n=208) .

Analysis of Variance Table

Source: DF: Sum Squares: Mean Square: F-test:

Between groups 2 0.683 0.341 9.519 Within groups 206 7.388 0.036 p = 0.00 Total 208 8.071

Model II estimate of between component variance = 0.153 Group: Count: Mean: Std. Dev.: Std. Error:

TT 62 1.278 0.203 0.026

Tt 99 1.213 0.184 0.018 tt 48 1.119 0.182 0.026 Comparison: Mean Diff.: Fisher PLSD:

TT vs . Tt 0.065 0.060

TT vs . tt 0.159 0.072

Tt vs. tt 0.094 0.066

All groups were significantly different from each other.

B. Using haplotype analysis (n=204). Analysis of Variance Table

Source: DF: Sum Squares: Mean Square: F-test:

Between groups 4 0.693 0.173 4.853

Within groups 200 7.138 0.036 p =0.00

Total 204 7.83

Model II estimate of between component variance = 0.034

Genotype n Mean SD SEM

2,2 49 1.124 0.182 0.026

2,3 17 1.242 0.162 0.039

1,2 78 1.207 0.190 0.021

1,3 20 1.284 0.167 0.037

1,1 41 1.280 0.213 0.033

Comparison: Mean Diff.: Fisher PLSD:

2,2 vs . 2,3* -0.119 0.105

2,2 vs . 1,2* -0.084 0.068

2,2 vs . 1,3* -0.160 0.099

2,2 vs . 1,1* -0.157 0.079

2,3 vs . 1,2 0.035 0.100

2,3 vs . 1,3 -0.041 0.123

2,3 vs . 1,1 -0.038 0.107

1,2 vs . 1,3 -0.076 0.093

1,2 vs . 1,1* -0.073 0.072

1,3 vs . 1,1 0.003 0.102

* Highly significant difference p<0.001

While the mean difference between genotypes is not increased in males by using haplotypes (0.157, 2,2 compared to 1,1) the use of haplotpe analysis demonstrates a difference between homozygote 2,2 and heterozygote 2,3 (mean difference 0.16). A combination of haplotype and weight explained 19.1% of the variance in lumbar spine bone density and 8.9% of the variance in femoral neck BMD. Haplotype data alone explained 8.1% and 5% of the variance in BMD for the lumbar spine and femoral neck, respectively.

Study 5. The genotype of the VDR receptor gene predicts the risk of bone fracture. Method. The study plan was to conduct a prosepective epidemiological study of incident bone fracture events in the city of Dubbo New South Wales Australia. This study is described in Nguyen et al in the British Medical Journal Volume 307:1111-1115, 1993. The utility of vitamin D receptor in predicting fracture was determined by genotyping a random sample of 269 females with mean age 70+7 years, selected from the Dubbo study. Bone density was measured at the femoral neck at recruitment into the study and the statistics of logistic regression were used to discriminate the relationship to subsequent fracture incidence. Fracture was assessed by radiology. Genotype was assessed using Taq-1 and Apa-1 RFLP sites. Results.

In this sample, 88 subjects had suffered fracture since recruitment (5 year period) into the prospective study. Females with the genotype tt had higher incidence (p=0.03) of fracture at any bone site (42%) than either Tt (33%) or TT genotypes (30%). The observed genetic effect was more pronounced after controlling by logistic regression for the distribution of bone densities in genotypic groups and the contribution of bone density to fracture risk. Of those in the lowest quartile of femoral neck bone density (<0.70 gm/cm²) a significantly higher proportion (p<0.001) of the tt fractured (69% of the genotype numbers) than of the heteozygote (Tt, 54%) or the the homozygote (TT, 52%). Significant genotypic differences exist in the odds ratio of probability of fracture at a bone density of one standard deviation decrease in femoral neck bone density. The overall odds ratio was 1.9 for fracture at this level of decrease, while odds ratios were different according to genotype with TT, Tt and tt genotypes having values of 1.40, 2.59 and 3.72 respectively with significance at p=0.03. These data clearly show that fracture risk is related to genotype with particular genotypes having lower relative risk of fracture. Discussion.

Osteocalcin is a serum marker derived from osteoblasts that is widely used as a marker of bone turnover (see Morrison et al 1992). Different mean serum levels of osteocalcin were previously reported in different genetic groups defined by simple RFLP markers in the vitamin D receptor gene, leading to the conclusion that the RFLPs are markers for functionally different vitamin D receptor gene alleles. The information supporting the utility of the invention has been increased and the present invention demonstrated to be an improvement on the state of the art. This has been achieved by investigating the relationships between osteocalcin, age and menopause related changes in bone density using analysis based on combined RFLPs and haplotypes. Strong linkage disequilibrium between the RFLPs provides the basis for haplotype analysis, which showed three major haplotypes in Caucasians. A genetic analysis of premenopausal bone density using haplotypes revealed that the genetic effect is stronger than previously described, defining a genetic group with extremely high BMD. In addition, genotypes were observed with intermediate phenotype (1,3 and 2,3 heterozygotes) . Haplotype analysis explains the relationship between the three RFLPs, since there are only three major haplotype forms seen in Caucasians (baT, BAt and bAT) .

The invention comprises, in a particular preferred embodiment, the use of certain oligonucleotide primers to amplify by PCR regions of the VDR gene containing polymorphic restriction endonuclease sites. The genotypes thus derived are combined to increase the genetic information content. In another preferred embodiment the genetic information is combined with serum osteocalcin data to provide enhanced predictiveness of bone density within certain genetic groups.

In this invention we also describe a preferred embodiment in which 2.4kb PCR fragment spanning from exon 7 to midway in exon 9 and covering these polymorphic sites such that all three RFLPs cannot be detected in a single fragment (Morrison et al 1994a) . Osteocalcin was originally used as the trait to discriminate functionally different VDR alleles. This conclusion was verified using new subjects, confirming the existence of a VDR genotypes with traits of high osteocalcin or low osteocalcin. Menopause effects on osteocalcin were detected in all genotypic groupings, with similar rises in osteocalcin after the menopause. Remarkably, the menopausal rise in osteocalcin seen in the genotypes characterised by low osteocalcin (haplotype 1,1 homozygotes), merely increased osteocalcin to the premenopausal values seen in the high osteocalcin haplotype 2,2 group. The high osteocalcin groups had lower bone density in both premenopausal and postmenopausal years. The menopause decline in bone density, which is the primary cause of postmenopausal osteoporosis, was comparable in magnitude to the genetic effect. The high BMD group declined after the menopause to a mean BMD which was similar to the mean BMD of the (high osteocalcin) low BMD group in premenopausal years. The inverse relationship between osteocalcin and bone mass was also seen. Across all subjects, osteocalcin was negatively associated with BMD. When haplotype 2,2 subjects were considered there was no relationship between osteocalcin and BMD. In haplotype 1,1 homozygotes, postmenopausal BMD was strongly related to osteocalcin and not to YPM.

Therefore the invention permits the detection of different genetic groups in which bone turnover markers are differentially related to BMD. These unexpected results show that: 1, even within a high BMD grouping, bone turnover state is associated with bone density and 2, in the converse group, haplotype 2,2, (that has high osteocalcin and low BMD), there was no relationship between osteocalcin and BMD in postmenopausal females, rather LS BMD declined strongly with years after menopause. Intermediate effects were observed in heterozygotes. Thus the invention has utility in descriminating the signifigance of osetocalcin and presumably other bone turnover markers .

Despite the fact that 1,1 homozygotes have higher mean bone density, some of these individuals at the low end of the genotype range will be at risk of osteoporosis. The correlation between osteocalcin and bone density in the 1,1 homozygote permits a method of discriminating that risk. Osteoporotic risk is related to the postmenopausal decline in bone density and premenopausal BMD. Longitudinal studies have shown that the premenopausal BMD of an individual is a strong correlate of subsequent postmenopausal BMD, up to 15 years after menopause. The present invention showed that the genetic effect is larger than previously realised, exceeding one standard deviation of the young normal range. Since genotype contributes to a substantial difference in premenopausal BMD it would be expected to contribute to postmenopausal BMD. The present invention detects a difference in the inferred rate of change of BMD at the LS and FN ranging from 0.5% per annum (genotypes 1,1 and *1,2) up to 1.2% (genotype 2,2). When this analysis was focussed within 20 YPM the difference was even greater, with 2.0% per annum for genotype 2,2 subjects and low rates and non¬ significant relationships (slopes not significantly different from zero) for the heterozygote 1,2 and homozygote 1,1. Although this relies on extrapolating rates of change from cross sectional data, the slopes are significantly different and the invention provides a means of detecting two patient groups which can be characterised as "fast losers" and "slow losers". These data predict that "fast losers" would constitute about 15 % of the population. The difference in the rate of change of BMD, coupled with the difference in premenopausal BMD, results in a large difference in the actual menopause related regressions. This translated to an approximate 23 year difference in the projected time required for the different genotypes to reach a critical value of bone density. We chose a fracture threshold of -2 SD of the young normal population which is reasonably close to that accepted in the field.

The present invention detects a genetic group (haplotype 1,1 homozygotes) who, if female, will on average reach the fracture threshold at about 89 years of age. This result represents an effective genetic resistance to osteoporosis, since at least 50% will still be above the threshold at this age and bone fracture may not be their only health concern. Furthermore a preferred embodiment of the present invention, in which genotype data is combined with serum osteocalcin data, detects those individuals who have a propensity to different levels of bone density within this high bone density group. In this group, which comprises 22% of our population, serum osteocalcin is related to LS BMD in the postmenopausal years, so an embodiment of the present invention utilising a combination of serum osetocalin and genotype detects a smaller subgroup, referred to as "genetic resistance to osteoporosis", that has very little prospect of reaching the fracture threshold.

The present invention also detects the status of haplotype 1,2 heterozygotes as having a projected time of reaching the fracture threshold of 34.6 YPM (around 83 years), as a result of partial genetic dominance of haplotype 1 on the rate of postmenopausal bone loss.

The present invention detects a group most at risk of osteoporosis (the haplotype 2,2 genotype), who have lower premenopausal BMD and faster bone loss in the postmenopausal years. This group reaches the fracture threshold at about 66 years of age, only 18 years after the menopause with many subjects being below the threshold earlier. The relationship between fracture and genotype is clearly defined.

The present invention provides a means of predicting different serum calcium set point levels in postmenopausal females and defines a link between serum calcium, and therefore presumably intestinal calcium uptake, such that low bone density genotypes are characterised by low serum calcium. Therefore the invention further relates to the potential for selection of subjects for calcium therapy.

It will be appreciated by persons skilled in the art that numerous variations and/or modifications may be made to the invention as shown in the specific embodiments without departing from the spirit or scope of the invention as broadly described. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive.

References.

1. Riggs BL & Melton LJ III. N Engl J Med 314: 1676- 1686 (1986).

2. Nguyen TV, PN Sambrook, PJ Kelly, SR Lord, J Freund, JA Eisman. Brit Med J 1993 (in press)

3. Smith DM et al . J Clin Invest 52: 2800-2808 (1973) .

4. Moller M et al. Calcif Tissue Int 25: 197-201 (1978) . 5. Pocock NA et al. J Clin Invest 80: 706-710 (1987).

6. Dequeker J et al. Bone 8: 207-209 (1987).

7. Seeman E et al . N Engl J Med 320: 554-558 (1986).

8. Kelly PJ et al . J Clin Endocrinol Metab 72: 808-814 (1991) . 9. Kelly PJ et al. Osteoporosis Int. 1:1-5 (1990).

10. Morrison NA, Yeoman R, Kelly PJ, Eisman JA. Proc Natl Acad Sci (USA) 89: 6665-6669 (1992).

11. Pocock NA et al . Med J Aust 146: 293-297 (1987).

12. Pocock NA et al . J Bone Min Res 3: 601-604 (1988). 13. Pocock NA et al . J Clin Invest 78: 618-621 (1986).

14. Pocock NA et al. J Bone Min Res 4: 441-448 (1989).

15. Haseman JK & Elston RC. Behav Genet 1: 11-19 (1970) and 2: 3-19 (1972) .

16. Morrison NA et al. Science 246: 1158-1161 (1989). 17. Demay MB, Roth DA, Kronenberg HM. J Biol Chem 264:

2279-2282 (1989).

18. Kerner SA, Scott RA, Pike JW. Proc Natl Acad Sci (USA) 86: 4455-4459 (1989)

19. Eisman JA. In: Physiology and Pharmacology of Bone Eds: Mundy GR, Martin TJ. Handbook of

Experimental Pharmacology 107: 333-375 (1993).

20. Suda T, Takahashi N & Martin TJ. Endocrine Rev 13:66-80 (1992).

21. Bell NH et al . J Clin Invest 76:470-473 (1985) . 22. Arveiler B, Petkovich M, Mandel JL, Chambon P.

Nucleic Acids Res 16: 6252 (1988).

Claims

CLAIMS : -

1. A method of assessing an individuals predisposition to low or high bone density and/or risk of fracture comprising assessing the vitamin D receptor genotype of the individual by haplotype analysis .

2. A method as claimed in claim 1 in which the genotype of the individual is assessed by haplotype analysis by assessing the genotype for the polymorphic restriction endonuclease sites for Bsm-1, Apa-1 and Taq-1 in the combinations of Bsm-1 and Apa-1, or Taq-1 and Apa-1, or Bsm-1 and Taq-1, or Bsm-1, Apa-1 and Taq-1.

3. A method as claimed in claimed in claim 2 in which the assessment includes the polymorphic restriction endonuclease sites Sph-1 and/or the poly adenosine sequence microsatellite in the 3 prime untranslated region of exon 9 of the vitamin D receptor gene.

4. A method as claimed in any one of claims 1 to 3 in which a segment of the vitamin D receptor is amplified using polymerase chain reaction prior to haplotype analysis.

5. A method as claimed in any one of claims 1 to 4 in which the assessment further includes measuring the individuals serum osteocalcin level.