WO1997005275A1

WO1997005275A1 - Method of predicting bone density

Info

Publication number: WO1997005275A1
Application number: PCT/AU1996/000474
Authority: WO
Inventors: Nigel Alexander Morrison; John Allan Eisman
Original assignee: Garvan Institute Of Medical Research
Priority date: 1995-07-27
Filing date: 1996-07-29
Publication date: 1997-02-13
Also published as: AUPN445095A0

Abstract

A method of assessing an individual's predisposition to low or high bone density, development of high or low bone turnover and/or responsiveness to therapy, comprising analysing allelic variation in relation to a retinoic acid receptor gene (RAR) of the individual. In a preferred embodiment of the invention, the method comprises analysing for allelic variation in the RAR-a gene through restriction fragment length polymorphism (RFLP) using Pst I endonuclease digestion.

Description

METHOD OF PREDICTING BONE DENSITY

The present invention relates to a method of identifying allelic variations in trans-acting regulators as a means of identifying individuals at risk of suffering from an adverse pathophysiological condition. The method of the present invention is particularly useful in assessing allelic variations in a retinoic acid receptor (RAR) gene(s) and thereby predicting predisposition to low or high bone density. In particular, the invention relates to the use of diagnostic kits for the discrimination of individuals with different set points of physiological character associated with the RAR regulatory 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 or high risk of osteoporosis.

Morrison et al. (1992) showed that variation in trans-acting (transcriptional) regulators (TAR) can be linked to different set points of physiological traits that are controlled by the regulator in question. It follows that the existence of useful genetic variation in a trans-acting regulator gene can be detected by testing for significant differences in the mean values of target gene products. This concept was demonstated for the relationship between genetic variation in the VDR receptor gene and the physiological set points of serum osteocalcin, a protein product regulated at the gene level directly by an association of the vitamin D receptor protein and a response element in the osteocalcin gene promoter. Since trans-acting regulators control the activity of large numbers of genes, whose regulation is subordinate to the regulation of the TAR, any genetic variation in the TAR gene which leads to functionally different TAR gene variants, can produce a coordinated change in the overall regulation of tl e cohort of downstream target genes. Such a downstream effect can result in a change in the set point of a gross physiological variable, such as bone density, or any other physiological variable which, conceivably, is regulated downstream of the TAR. According to this concept, genetic variation in the VDR gene was also related to different set points of the bone density trait and subsequently related to osteoporosis risk.

This concept suggests that genetic variation that has physiological significance can be identified by a comparison of the level of a physiological target gene regulated by the TAR in question. Evidence exists that the osteocalcin gene is also regulated by retinoic acid (Morrison et al Science 1989) through promoter elements in the osteocalcin promoter. RAR-a (Petkovich et al. Nature 1987) and other RAR's, is an important conduit of developmental signals relating to gene expression and regulation of cellular differentiation in many gene and cellular systems.

RAR-a and RXRs (retinoid X receptors) are derived from a family of related genes with similar function. These proteins form part of a complex web of cross-regulatory actions with other TAR, both of the steroid receptor family [VDR, T3R, ER and orphan receptors] as well as TAR of unrelated types such as the Jun/Fos family leucine zipper regulators. RAR-a and related genes are therefore central regulators of diverse physiological processes. Variation in any member of the RAR-a family, if resulting in functional changes, should result in the alteration of subsequent physiological processes and be reflected in changed set points. If such variation could be identified it may serve as a useful genetic prognostic.

RAR-a is also an important co-factor in VDR gene responses, acting to regulate gene expression by the formation of heterodimers, although VDR and RAR-a can act as homodimers and can form complexes with other members of this and other TAR families as mentioned above. Regulation of the osteocalcin gene by RA metabolites has been described by several authors, including Morrison, and direct association of RAR-a and its related protein X to the osteocalcin gene promoter has been described. Regulation of osteocalcin by RAR-a therefore provides the opportunity to detect functional genetic variation in the gene for RAR-a, and subsequently, members of the RAR-RXR family.

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. Prediction of those at risk of osteoporosis may reduce the incidence of this disease by enabling early attention to those in greatest need. The aim of the present inventors has, therefore, been to define genetic tests that are capable of identifying those individuals at risk of osteoporosis and those protected from the disease.

To this end, the present inventors have now found that there is a relationship between functionally different RAR-a alleles and bone density and that this relationship is significant, indicating that allelic variation in RAR-a can be used as a genetic test for establishing the bone density trait and for examining the interaction of RAR-a variants and VDR variants in regulating physiological variables.

Accordingly, the present invention consists in a method of assessing an individual's predisposition to low or high bone density, development of high or low bone turnover and/or responsiveness to therapy comprising analysing allelic variation in relation to a retinoic acid receptor (RAR) gene of the individual.

Preferably, the RAR gene is the RAR-a gene. In a preferred embodiment of the present invention, the analysis comprises restriction fragment length polymorphism (RFLP) using endonuclease digestion.

In a further preferred embodiment of the present invention, a segment of the retinoic acid receptor (RAR) gene is amplified using polymerase chain reaction prior to endonuclease digestion. In yet a further preferred embodiment of the present invention, the endonuclease is selected from the group of endonucleases which cleave at the nucleotide sequence CTGCAG. Examples of such endonucleases are Bbi I, Noc I, Pma I, Pst I, Xma II and Xor I. Most preferably, the endonuclease is Pst I.

In yet still a further preferred embodiment of the present invention, the method further involves analysing allelic variation in relation to the vitamin D receptor gene and/or the estrogen receptor gene of the individual. Further information regarding the analysis of these genes may be found in International Patent Application No.s PCT/AU93/00394, PCT/AU95/00452 and PCT/AU96/00017. The disclosure of these applications is included herein by cross reference.

As indicated above, it is presently preferred that RAR-a genotypes are detected with the endonuclease Pst I. However, as it will be appreciated by those skilled in the art, other restriction enzymes and other detection systems can be used to detect the same genetic variants, and that other genetic variants can have similar information content and be detected by other means. The Pst I RFLP may be in linkage disequilibrium with other sequence alteration, both known and unknown, that mediate this effect.

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

1. through the use of Southern hybridisation blots to detect the RAR- a restriction fragment length polymorphism(s) using a cDNA fragment(s) as a probe;

2. through the use of polymerase chain reaction (PCR) and subsequent digestion by restriction endonuclease(s) as described below to detect polymorphic site(s) in the RAR-a gene and the subsequent analysis of the genotype(s); and

3. through the use of RAR-a genotypes, detected by any means, in combination with other genes, particularly the VDR gene and the estrogen receptor (ESR) gene, in determining risk of low bone density, differing rates of loss of bone density and subsequent osteoporosis risk; and 4. through the use of RAR-a genotypes, detected by any means, in combination with other genes, particularly the VDR gene and the (ESR) gene, in determining response to therapy directed at reducing osteoporosis risk.

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 accompanying figures.

Brief Description of the Figures

Figure 1 provides a graphical representation of the RAR-a genotype effect across "all females". The result of a Students t-test for the extreme genotypes was p = 0.0272. Simple regression gave p = 0.0405.

Figure 2 provides a graphical representation of the RAR-a genotype effect across "postmenopausal females". When postmenopausal female subjects were analysed alone a significant effect of RAR-a genotype was observed. The result of a Students t-test for the extreme genotypes was p = 0.053. Simple regression gave p = 0.0471.

Figure 3 provides a graphical representation of the RAR-a genotype effect across "premenopausal females". This figure shows the result when premenopausal females were analysed. The result of a Students t-test for the extreme genotypes was p = 0.0116. Simple regression gave p = 0.0265.

Figure 4 provides a graphical representation of the relationship between RAR-a alleles and LS BMD when VDR genotype is controlled. This figure shows that the RAR-a genotype effect persists when the VDR genotype effect is controlled by taking only VDR heterozygotes into consideration. Figure 5 provides a graphical representation of the relationship between RAR-a alleles and serum osteocalcin when the VDR genotype is constrained to heterozygotes.

Fisgure 6 provides a plot of LS BMD from "all female" subjects. This figure shows a scatter plot of lumbar spine bone densities of normal female subjects according to combined VDR and RAR-a genotypes showing an almost linear relationship between BMD and genetic factors.

Figure 7 provides a graph showing a plot of mean ± SEM of LS BMD within each combined genotype group.

EXAMPLE

MATERIALS AND METHODS.

1. Subjects.

A sample of normal subjects was genotyped by standard Southern blot techniques for variants in the RAR-a gene by RFLP. These subjects had previously been genotyped for VDR variation and represented well characterised normal humans. The study population was comprised of monozygotic (MZ) and dizygotic (DZ) twins, representing a total of 252 people with a mean age of 42 years. There were 24 male pairs and 103 female pairs composed of 78 premenopausal and 25 postmenopausal pairs. The total number genotyped for RAR-a alleles consisted of 31 males, 47 postmenopausal females and 132 premenopausal females. All subjects were genotyped for VDR alleles except two MZ pairs (one male and one female). These twins, who were selected only on the basis of being twins, comprised 71 MZ and 55 DZ twin pairs, including 7 male MZ pairs and 6 male DZ pairs. They were aged between 17 and 70 years; MZ 45 ± 13 yrs and DZ 44 ± 11 yrs, mean ± SD. All female twin pairs were concordant for menopausal status and, if postmenopausal, for years since menopause. Bone density was measured at the lumbar spine and proximal femur with a Lunar DP3 dual-photon absorptiometer (LUNAR Corporation, Madison, WI) as previously described (Pocock NA et al., 1987) or Lunar DEXA dual energy x-ray absorptiometry (Pocock NA et al., 1988).

Genetic effects on bone density are reported to be strong at the lumbar spine region of the vertebral column, so this region was selected for analysis of the relationship between gene variants and bone density.

The population distribution of lower spine bone mineral density (LS BMD) was as follows:

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

Pre 152 1.199 0.126 0.01

Post 48 1.066 0.17 0.025

Male 48 1.235 0.155 0.022

2. DNA Analysis.

It will 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 Southern blot with probes to identify polymorphic DNA sequences. 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 (lOmM Tris HCl, pH 7.4, physiological saline and 0.5% w/v sodium dodecyl sulphate). Lysate was treated with proteinase K (Applied Biosciences, Palo Alto, USA) at 50μg/ml for 2 hours at 65°C. 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.

The RAR-a gene probe was a 1.6kb Eco Rl fragment from a cDNA clone of the human RAR-a (Arveiler et al. 1988). Southern blotting was according to methods well known in the art and hybridisation of radioactive probe was under standard conditions at 65°C, and high stringency washes to 0.2% SSC and 0.1% SDS at 65°C prior to autoradiographic exposure. The RFLP was a two allele Pst I polymorphism with bands at 3.0kb in the absence of the Pst I restriction site and two bands (2.6 and 0.4 kb) in the presence of the Pst I restriction site. Invariant bands are also present at high stringency at 5.7, 3.8, 1.5, 1.4, 0.9, 0.8 and 0.3 kb.

3. Statistical analysis.

It has been previously described that the effect of VDR alleles on bone density was not dependent on zygosity, permitting standard statistical analysis such as analysis of variance (ANOVA), Students t-test, and various regression models to be used on a massed twin population. Such a twin population is a representative model of the population as a whole. ANOVA analysis (using Statview from Abacus Inc. Berkeley USA) was used to detect different mean values of parameters in genotypic groupings under Fisher's protected least significant difference test (P.L.S.D.). Student's t-test was used to examine pairwise comparisons. Regression was used to examine genotype effects in conjunction with other anthropomorphic variables known to affect bone density.

4. Allelic frequency.

The allelic frequency was found to be: Presence of Pst I site: 0.81

Absence of Pst I site (AA): 0.19

ANALYSIS AND RESULTS. 1. THE EFFECT OF PST I RESTRICTION FRAGMENT LENGTH POLYMORPHISMS ON HUMAN BONE MINERAL DENSITY AND SERUM OSTEOCALCIN LEVELS.

I.a Identification of significant effects of RAR-a alleles detected in female by comparisons of bone density at the lumbar spine.

Method: Bone density was compared across RAR-a genotype groups in all females, and separately in pre and postmenopausal groups. As bone density at the spine changes with the menopause, a significant result in both pre and postmenopausal groups confirms the gene effect.

l.a(i). All females.

Using ANOVA analysis. Analysis of Variance Table

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

Between groups 2 0.111 0.056 2.518

Within groups 170 3.756 0.022 p = 0.0836

Total 172 3.867

Model II estimate of between component variance = .017

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

AA 23 1.226 0.193 0.040

AB 67 1.165 0.148 0.018

BB 83 1.147 0.135 0.015

Comparison: Mean Diff.: Fisher PLSD:

AA vs. AB 0.062 0.071

AA vs. BB 0.079 0.069*

AB vs. BB 0.017 0.048 * Significant at 95% confidence limit.

l.a(ii). Premenopausal females. Analysis of Variance Table Source: DF: Sum Squares: Mean Square: F-test:

Between groups 2 0.077 0.039 2.507

Within groups 130 2.007 0.015 p = 0.0855

Total 132 2.085

Model II estimate of between component variance = .012

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

AA 14 1.268 0.11 .029

AB 51 1.194 0.134 .019

BB 68 1.187 0.119 .014

Comparison: Mean Diff.: Fisher PLSD

AA vs. AB .073 .074

AA vs. BB .081 .072*

AB vs. BB .008 .046

* Significant at 95% confidence limit. l.a(iii). Post menopausal females. Analysis of Variance Table

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

Between groups 2 .13 .065 2.284

Within groups 43 1.225 .028 p = .1141

Total 45 1.355

Model II estimate of between component variance = .018 Group: Count: Mean: Std . Dev.: Std . Error:

AA 10 1.163 .257 .081 AB 18 1.063 .155 .036 BB 18 1.021 .114 .027

Comparison: Mean Diff.: Fisher PLSD

AA vs. AB 0.100 0.134

AA vs. BB 0.142 0.134*

AB vs. BB 0.042 0.113

''Significant at 95% confidence limit.

CONCLUSION:

In each analysis of the entire group or sub-grouping of pre and postmenopausal females, the same gene effect was seen, i.e. a high or low bone density was associated with RAR-a genotype.

Graphical representations of the results are provided at Figures 1-3.

l.b Identification of significant effects of RAR-a by controlling for the strong variable of genetic variation in the vitamin D receptor gene.

Vitamin D receptor (VDR) genotype has a strong effect on bone density (Morrison et al Nature 1994). To examine the possibility that the RAR-a genotype effect was the result of fortuitous distribution of VDR gene alleles within the RAR-a genotype categorisations, the effect of RAR-a alleles was examined by first controlling for VDR gene variation by selecting VDR gene heterozygotes (the most abundant VDR genotype group) which have intermediate BMD, and then examining the effect of RAR-a on BMD within this group. Method: In this analysis the genotype of the vitamin D receptor was controlled by selecting those subjects heterozygous for the Bsm-1 RFLP (Bb), minimising the effect of VDR gene variation. When this was done, the relationship between RAR-a receptor gene alleles and the parameters of lumbar spine bone density and serum osteocalcin was examined in females by ANOVA.

l.b(i) Result of analysis of lumbar spine bone density: Analysis of Variance Table

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

Between groups 2 .108 .054 3.35

Within groups 85 1.364 .016 0.0398

Total 87 1.472

Model II estimate of between component variance = 0.019

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

AA 13 1.25 0.13 0.036

AB 34 1.154 0.132 0.023

BB 41 1.15 0.121 0.019

Comparison: Mean Diff.: Fisher PLSD:

AA vs. AB 0.096 0.082*

AA vs. BB 0.100 0.080*

AB vs. BB 0.004 0.058

* significant at 95% confidence.

The difference between the means of genotype groupings can be used as a estimate of the allele's power in regulating the trait of bone density. The difference in the means was 0.100, which can be compared against the total variance of the total study group and of the VDR heterozygote group used in this comparison.

Total group:

Xi: LS

Mean: Std. Dev. : Std. Error: Variance: Coef. Var.: Count:

1.167 .149 .011 .022 12.744 200

Minimum: Maximum: Range: Sum: Sum of Sαr.:# Missinε:

.74 1.58 .84 233.42 276.827 6

VDR Bsm-1 allele heterozygotes used to control for the VDR gene effect:

Xi LS

Mean: Std. Dev.: Std. Error: Variance: Coef. Var. : Count:

1.174 .131 .013 .017 11.156 98

Minimum: Maximum: Ranee: Sum: Sum of Sqr.: # Missinε:

.87 1.47 .6 115.091 136.828 4

The mean ± standard deviation of the heterozygote population is 1.174±0.131. The mean difference of the RAR-a allele effect is 0.100 or 76% of the SD of the heterozygote controlled population.

l.b(ii) Result for serum osteocalcin. Analysis of Variance Table

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

Between groups 2 124.744 62.372 2.985

Within groups 73 1525.575 20.898 p=0.0568

Total 75 1650.319

Model II estimate of between component variance = 20.737 Group: Count: Mean: Std. Dev.: Std. Error:

AA 11 3.596 2.682 0.809

AB 32 7.5 5.094 0.901

BB 33 6.491 4.504 0.784

Comparison: Mean Diff.: Fisher PLSD:

AA vs. AB -3.904 3.185*

AA vs. BB -2.895 3.172

AB vs. BB 1.009 2.261

significant at 95% confidence.

CONCLUSION:

These results show that there is a significant effect of RAR-a receptor genotype on both lumbar spine bone density and on serum osteocalcin which is revealed when the genetic variation in VDR gene alleles is controlled by limiting analysis to heterozygotes. Controlling for the VDR genetic variation permits the detection of a significant heterozygote effect on LS BMD.

Graphical representations of the results are provided at Figures 4 and 5.

l.c The effect of RAR-a genotypes on serum osteocalcin was not detectable without controlling of VDR genotype. Analysis of Variance Table

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

Between groups 2 11.472 5.736 0.178

Within groups 156 5016.686 32.158 p=0.8368

Total 158 5028.159

Model II estimate of between component variance = -13.211 Group: Count: Mean: Std. Dev.: Std. Error:

AA 21 6.593 8.083 1.764

AB 67 7.44 5.701 0.696

BB 71 7.23 4.728 0.561

No significant differences.

I.d Comparisons ofthe magnitude ofthe RAR-a gene effect on the population distribution of LS BMD.

Within the RAR-a genotypic groupings mean values from each genotype were:

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

AA 24 1.224 0.189 0.038 AB 69 1.16 0.15 0.018 BB 86 1.152 0.136 0.015

Comparison: Mean DifT. Fisher PLSD:

AA vs. AB 0.064 0.070* AA vs. BB 0.072 0.068* AB vs. BB 0.008 0.048 significant at 95% confidence.

The mean differences were 0.064 and 0.072 between significantly difference genotypes.

When the entire female study group is considered (mean±SD of 1.167±0.149) then the difference is about 50% of a standard deviation.

In a larger Sydney population, the mean of young normal females was 1.2±0.144 (n=70). Therefore difference in means attributable to RAR-a is approximately 0.5 SD of the young normal population.

l.d(i) As a comparison between RAR-a and VDR alleles.

Female subjects split according to VDR allele genotypes. Group: Count: Mean: Std. Dev.: Std. Error:

BB 42 1.063 0.141 0.022

Bb 98 1.174 0.131 0.013 bb 58 1.235 0.141 0.019

In the same population, VDR alleles give extreme means of 1.063 and 1.235 with a total mean difference of 0.172, exceeding one standard deviation of the young normal value.

CONCLUSION:

The effect attributable to RAR-a alleles is approximately one half that of VDR alleles.

2. MULT-VARIATE REGRESSION MODEL.

2. a Analysis of RAR-a genotypes and bone density.

The genotypes of RAR-a were tested with the two strong variables years post menopause (YPM) and weight (Kg) in multiple regression against BMD in the female test population. The result was that RAR-a genotypes are significantly related to the bone mass trait. Regression Table:

Count: 179

R: 0.48

R-squared: 0.23

Adj. R-squared: 0.217

RMS Residual: 0.133

Analysis of Variance Table

Source DF: Sum Squares: Mean Square: F-test: REGRESSION 3 0.926 0.309 17.465

RESIDUAL 175 3.092 0.018 p=0.0001

TOTAL 178 4.018

Beta Coefficient Table

INTERCEPT 1.067

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

RAR-a -0.042 0.014 -0.195 2.873 0.0046

KG 0.004 0.001 0.254 3.823 0.0002

YPM -0.011 0.002 -0.384 5.672 0.0001

Intercept Confidence Intervals and Partial F Table

Variable: 95% Lower: 95% Upper: 90% Lower: 90% Upper: Partial F:

RAR-a -0.07 -0.013 -0.065 -0.018 8.251

KG 0.002 0.005 0.002 0.005 14.616

YPM -0.014 -0.007 -0.014 -0.008 32.175

CONCLUSION:

Different RAR-a alleles are associated with significantly different BMD traits.

2.b On the relative power of variables as factors in assessment of lumbar spine bone density.

Incorporating RAR-a alleles with other variables known to be involved in the bone density trait.

Method. Several variables are assessed in a multivariate model using stepwise regression against the bone density variable (lumbar spine): Summary Information

F to Enter 4

F to Remove 3.996

Number of Steps 4

Variables Entered 4

Variables Forced 0...0

FIRST STEP

R: 0.43

R-squared: 0.185

Adj. R-squared: 0.179

RMS Residual: 0.134

Analysis of Variance Table

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

REGRESSION 1 .509 .509 28.2

RESIDUAL 124 2.238 p=0.018

TOTAL 125 2.747

LAST STEP

R: 0.601

R-squared: 0.361

Adj. R-squared 0.34

RMS Residual: 0.12

Analysis of Variance Table

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

REGRESSION 4 .992 .248 17.107 RESIDUAL 121 1.755 p=0.01

TOTAL 125 2.747

Variables in i Equation

INTERCEPT 0.954

Variable: Coefficient: Std. Err.: Std. Coeff.: F to Remove

KG 0.003 0.001 0.192 6.848

RAR-a -0.045 0.016 -0.218 7.793

YPM -0.01 0.002 -0.391 25.277

VDR 0.08 0.015 0.4 9.625

Variables Not in Equation

Variable: Par. Corr: F to Enter:

25D 0.168 3.476

CM 0.07 0.593 ln(l+x) of Oc -0.157 3.051

Age -0.149 2.735

The equation for 125 subjects (pre and postmenopausal females) is therefore: LS BMD = 0.954 + 0.003 KG -0.01 YPM -0.045 RAR-a + 0.08 VDR.

CONCLUSION:

Genetic markers VDR and RAR-a are independent predictors of bone density, and the strength of the RAR-a system is about one half that of the VDR system.

2.c Genetic factor RAR-a alone in a multi variate model of LS BMD. Method: RAR-a alleles, age, height, weight and years post menopause were incorporated in the regression. Multiple Regression 1ft :LS 5 X variables

Count: R: R-squared: Adj. R-squaredRMS Residual:

179 .489 1 .239 .217 .133

No Residual Statistics Computed Note: 27 cases deleted with missing values.

Multiple Regression Y1 :LS 5 X

Beta Coefficient Table

Variable Coefficient Std Err Std Coeff t-Value Probability

INTERCEPT 943

RAR ALLELE - 042 015 - 199 2 906 0041

CM 001 002 052 721 472

KG 003 001 245 3 433 0007

Age - 001 1 974 1 654 974 3313

YPM - 008 003 - 303 3 035 0028

Multiple Regression Y1:LS 5 X variables

Confidence Intervals and Partial F Table

Vaπable 95% Lower 95% Upper 90% Lower 90% Upper Partial F

INTERCEPT

RAR ALLELE - 071 - 014 - 066 - 018 8 446

CM - 002 004 - 001 004 519

KG 001 005 002 005 11 789

Age -3 898 3 895 -3 266 3 263 949

YPM - 014 - 003 - 013 - 004 9 213

CONCLUSION: RAR-a allele variation is independent of other known predictors of LS

BMD, is highly significantly associated with the BMD trait and is of comparable strength to the other variables (standardised coefficient of ^■ 0.199).

3. COMBINAΗON OF PST I RFLP'S AND VDR ALLELES.

3. a Genetic factors alone can explain a substantial part ofthe variance in bone density.

Method: VDR and RAR-a alleles were cross coded to give a 9 genotype system as follows:

Contingency table analysis of RAR-a and VDR alleles in female subjects (VDR in columns and RAR-a in row). Coding was arranged to create an ascending series of combined genotypes.

3.b Simple regression estimate ofthe component ofthe variance explained by combined RAR-a and VDR alleles.

Method: a simple linear regression with combined genotypes as the only variable was performed against LS BMD, with the result that r²=0.17 (n=117) meaning that this genetic factor alone explains 17% of the population variance in BMD in this sample.

3.c ANOVA analysis was used to determine which genotypes have significantly different bone densities. In this analysis the criterion of acceptance of significant differences is under Fisher's protected least significant difference test (PLSD) which is protected by the p= 0.0001 for the overall effect ANOVA. Analysis of Variance Table

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

Between groups 8 .821 0.103 5.66

Within groups 168 3.045 0.018 p = .0001

Total 176 3.865

Model II estimate of between component variance = .011

Mean LS BMD in each genotype designation.

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

A 23 1.085 0.118 0.025

B 11 1.030 0.181 0.054

C 5 1.036 0.195 0.087

D 42 1.143 0.127 0.020

E 33 1.163 0.124 0.022

F 13 1.250 0.130 0.036

G 21 1.222 0.148 0.032

H 23 1.215 0.105 0.022

I 6 1.323 0.209 0.085 Differences in means were enhanced using a two gene allelic system.

Comparison Mean Diff. Fisher PLSD

1. A vs. B 0.055 0.097

2. A vs. C 0.049 0.131

3. A vs. D -0.058 0.069

4. Avs.E -0.078 0.072*

5. A vs. F -0.1Θ5 0.092*

6. A vs. G -0.137 0.080*

7. Avs.H -0.130 0.078*

8. A vs. I -0.238 0.122*

9. Bvs. C -0.006 0.143

10. Bvs.D -0.114 0.090*

11. Bvs.E -0.133 0.093*

12. Bvs. F -0.221 0.109*

13. Bvs. G -0.192 0.099*

14. Bvs.H -0.185 0.097*

15. Bvs. I -0.294 0.135*

16. Cvs.D -0.107 0.126

17. Cvs. E -0.127 0.128

18. Cvs.F -0.214 0.140*

19. Cvs.G -0.186 0.132*

20. Cvs.H -0.179 0.131*

21. Cvs. I -0.287 0.161*

22. Dvs.E -0.019 0.062

23. Dvs.F -0.107 0.084*

24. Dvs. G -0.079 0.071*

25. Dvs.H -0.072 0.069*

26. Dvs. I -0.180 0.116*

27. Evs.F -0.088 0.087*

28. Evs. G -0.059 0.074

29. Evs.H -0.052 0.072

30. Evs. I -0.161 0.118*

31. Fvs. G 0.028 0.094

32. Fvs.H 0.035 0.092

33. Fvs. I -0.073 0.131

34. Gvs.H 0.007 0.080

35. Gvs.I -0.102 0.123

36. Hvs.I -0.109 0.122

*: significant at 95% confidence limit under Fisher's protected least significance difference test (P.L.S.D.) 3.d Genotypic differences in relation to population characteristics of bone density.

The greatest differences in in comparisons of means between genotypes were:

Comparison. Difference in means PLSD xSD Young normal

30. E vs. I -0.161 0.118* 1.12

5. A vs. F -0.165 0.092* 1.15

20. C vs. H -0.179 0.131* 1.24

26. D vs. I -0.180 0.116* 1.25

14. B vs. H -0.185 0.097* 1.29

19. C vs. G -0.186 0.132* 1.29

13. B vs. G -0.192 0.099* 1.33

18. C vs. F -0.214 0.140* 1.49

12. B vs. F -0.221 0.109* 1.54

8. A vs. I -0.238 0.122* 1.65

21. C vs. I -0.287 0.161* 1.99

15. B vs. I -0.294 0.135 * 2.04

In a larger Sydney population the mean of young normal females was 1.2 ± 0.144 (n=70). The differences in the means of the combined genotypes are greater than 2.0 standard deviations of the mean of young normal.

3.e Multiple regression of this combined genetic coding with other factors as above was used to establish the amount ofthe variance in bone density explained by these factors. Multiple regression table.

Count: R: R-squared: Adj. R-squared: RMS Residual:

177 0.59 0.348 0.333 0.121 Analysis of Variance Table

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

REGRESSION 4 1.345 0.336 22.943

RESIDUAL 172 2.521 0.015 p = .0001

TOTAL 176 3.865

Beta Coefficient Table

Variable: Coefficient: Std. Err.: Std. Coeff.: t-Value: Probal

INTERCEPT 1.015

Combined 0.027 0.004 0.415 6.499 0.0001

KG 0.002 0.001 0.177 2.808 0.0056

Age -0.003 0.001 -0.218 2.308 0.0222

YPM -0.005 1.974 1.654 2.090 0.0381

Confidence Intervals and Partial F Table

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

INTERCEPT

Combined 0.019 0.035 0.02 0.034 42.24

KG 0.001 0.004 0.001 0.004 7.887

Age -0.005 -3.999E-4 -.005 -0.001 5.328

YPM -3.902 3.892 -3.27 3.26 4.366

CONCLUSION:

34.8% of the variance in the LS BMD trait was explained by RAR-a and VDR combined alleles in conjunction with age, weight and years post menopause. The results presented above show the following:-

1. Genetic variation in the RAR-a gene in normal Caucasian volunteer subjects is associated with differences in the bone density trait.

2. The alleles of the RAR-a gene detected by RFLP can be used as markers defining functionally different alleles of the RAR-a gene, which are either themselves involved or are linked to genetic changes which are responsible for, functionally significant variation in physiological parameters of clinical, medical and prognostic use, reflected in the significant differences in the bone density and serum osteocalcin traits described herein.

3. Further genetic variation in the RAR-a gene, analysed in this way will further subdefine the genetic role that RAR-a has in regulating the physiological processes that control the bone density and serum osteocalcin traits. 4. Variation in RAR-a can be usefully combined with variation in VDR and/or ESR gene alleles to permit enhanced genetic prognostic for the trait of bone density, exceeding the present state of the art as demonstrated by VDR or ESR alleles used alone described by Morrison et al Nature (1994) and in International Patent Application No.s PCT/AU93/00394, PCT/AU95/00452 and PCT/AU96/00017.

5. It is highly likely that other genes in the RAR-a family can be similarly analysed for potential interaction with RAR-a and VDR alleles in describing genetic tests capable of detecting differences in the bone density trait or any other physiological variables regulated by the VDR or RAR-a endocrine systems or RAR-a related gene systems.

6. Due to the intimate molecular cross-talk between RAR-a and the VDR and the other protein members of the RAR-a system, it is likely that genetically different alleles of RAR-a and the VDR will have a functional impact on other apparently unrelated endocrine and physiological systems. For instance it is now known that the VDR can participate with other nuclear transactivators in regulation of target genes. RAR-a has the same property. The cross-regulatory participants are vitamin D, thyroid, retinoid, dioxin, peroxysomal proliferator regulator receptor molecules. It is therefore likely that genetic variation in RAR-a or VDR will result in regulatory disturbances in these other endocrine/metabolite/xenobiotic pathways. 7. Functionally different genetic variants of RAR-a that exist in the population at high frequency will have utility in genetic testing for response to RAR-a based compounds and therapeutics and will have utility in predicting the severity of diseases where RAR-a has been implicated by inference of therapy or direct evidence of RAR-a involvement has been confirmed.

In summary, the present inventors have developed a method of predicting genetic risk of having,

(a) low bone density, and

(b) high serum osteocalcin levels, involving the analysis of allelic variation in a RAR gene(s), particularly the RAR-a gene. The method permits the definition of groups of people with differing risks of subsequent bone disease, in particular postmenopausal osteoporosis.

In addition, the present inventors have shown that RAR-a genotypes can be used to decipher relationships between important physiological variables regulated by VDR dependent processes as a result of the fact that permissive genotypes of RAR-a permit mathematical description of the relationship between clinical humoral variables, which on general examination do not have strong relationships. A working example is the relationship between serum 1,25 dihydroxyvitamin D3 and serum osteocalcin which have a relationship governed by RAR-a genotype. 4 1 j RAR-a gene alleles have utility in predicting the diversity of response to therapy based on the vitamin D endocrine system.

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. Morrison NA et al.. 1994. Nature. 367: 284-287.(1994).

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

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

4. Pocock NA et al. J Bone Min Res 3: 601-604 (1988).

5. Morrison NA et al. Science 246: 1158-1161 (1989).

6. Petkovich M et al. Nature 330: 444-450 (1987). 7. Maniatis T et al. Molecular Cloning: a laboratory manual. Cold

Spring Harbour Laboratory, Cold Spring, New York (1982). 8. Arveiler B et al Nucleic Acids Res 16: 6252 (1988).

Claims

CLAIMS:-

1. A method of assessing an individual's predisposition to low or high bone density, development of high or low bone turnover and/or responsiveness to therapy, comprising analysing allelic variation in relation to a retinoic acid receptor gene (RAR) of the individual.

2. A method according to claim 1, wherein the analysis comprises restriction fragment length polymorphism (RFLP) using endonuclease digestion.

3. A method according to claim 2, wherein the analysis involves the amplification of a segment of the RAR gene using polymerase chain reaction prior to endonuclease digestion.

4. A method according to any one of the preceding claims, wherein the retinoic acid receptor gene is the RAR-a gene.

5. A method according to claim 4 wherein the endonuclease is selected from the group of endonucleases which cleave at the nucleotide sequence CTGCAG.

6. A method according to claim 5 wherein the endonuclease is Pst I.

7. A method according to any one of the preceding claims wherein the method further involves analysing allelic variation in relation to the vitamin

D receptor gene and/or the estrogen receptor gene of the individual.