AZGP GENE SINGLE NUCLEOTIDE POLYMORPHISMS (SNPs)

The present invention relates to SNPs and haplotypes in the AZGPl gene associated with obesity, and methods for determining predisposition of an individual to obesity by the presence or absence of said SNPs and/or haplotypes in the AZGPl gene.

Multifactorial diseases such as obesity are caused by mutations in more than one gene with a large contribution from environmental factors. There has been spectacular success in identifying the genes responsible for Mendelian disorders, whereas finding the susceptibility genes involved in multifactorial diseases has so far been difficult. The evidence suggests that humans inherit a genetic predisposition to gain weight on a high fat diet. It would be useful to identify markers of predisposition of individuals to obesity.

AZGPJ is a Zn-Alpha2-glycoprotein the gene of which is down-regulated in obesity (EP 1548445), and up-regulated in cachexia (Russell and Tisdale, 2005, Brit. J. Cancer 92, 876-881; Russell et al., 2004, Biochem. Biophys. Acta 1636, 59-68; Sanders and Tisdale, 2004, Cancer Lett. 212, 71-81; Bing et al., 2004, Proc. Natl. Acad. Sci USA 101, 2500-2505).

So far, no AZGPl haplotypes have been associated with obesity.

Description of the invention

The problem to be solved by the present invention was to provide markers for the predisposition of individuals to obesity. The problem was solved by the present invention by the identification of SNPs and haplotypes in the AZGPl gene which are associated with obesity. DNA samples obtained from lean and obese subjects were used to identify haplotypes and SNPs in the AZGPl gene. These SNPs and haplotypes were associated with obesity. As it is known from the literature that obesity is associated with insulin resistance, these SNPs may also be linked to insulin resistance. Obese subjects who participated in this study were non-diabetic when the samples were taken. DNA fragments of the AZGPl gene were amplified by PCR and sequenced. Following sequencing, polymorphism analysis was performed using the Polyphred software (University of Washington). Table 1 lists all markers identified in AZGPl .

Table 2 is showing the allele frequency of all polymorphic sites found in AZGPl DNA samples. For haplotype frequency calculations, only SNPs with a minor allele frequency higher than 5% were used. The less frequent markers are not likely to be selected in further association studies and will not contribute substantially to the common haplotypes. Out of the 28 markers presented in Table 2, 15 (in bold) were further included in the haplotype analysis.

Table 3 is providing the haplotype frequencies on the 15 frequent markers of AZGPl. As can be seen in the table some marker couples were completely redundant (equivalence of occurrence of alleles in the different haplotypes): • zaglδ and zagl9

• zagl7, zaglό and the intronic deletion (zag del).

• The cluster zaglό, zagl7 and the intronic deletion (zag del) is nearly redundant with zagl5 and zag35. By looking at Table 6, from zagl7 to zag35, only 3 haplotypes are present: AdelATT, GwtGCC and GwtGTC. The last haplotype is only present in H12 which is a rare haplotype (f=0.005).

Table 1. Characteristics of all markers identified in AZGPl, in DNA samples

Position of the marker in the EMBL accession number ac004977. ² Position of the marker in Seq ID No. 2

Table 2. Allelic frequency of discovered SNPs in AZGPl.

Distance

Marker region (bp) ¹ AIM AII2 N ² f AIM f AII2 zagO6 5'reg - G A 93 0.995 0.005 zagO5 5'reg 110 T C 92 0.549 0.451 zagO4 5'reg 16 T A 92 0.745 0.255 zagO3 5'reg 132 T C 92 0.022 0.978 zagO7 lntroni 1347 T C 93 0.823 0.177 zagO8 lntroni 65 T G 93 0.005 0.995 zagO9 lntroni 2 G A 93 0.027 0.973 zag13 lntroni 1766 G A 93 0.995 0.005 zag12 lntroni 485 T C 91 0.973 0.027 zag10 lntroni 355 T G 91 0.709 0.291 zag14 exon2 1192 T C 89 0.461 0.539 zag23 Intron2 169 G A 91 0.456 0.544 zag22 Intron2 164 T C 91 0.033 0.967 zag18 Intron2 308 C A 92 0.75 0.25 zag19 Intron2 17 G A 93 0.753 0.247 zag20 Intron2 235 T C 93 0.968 0.032 zag21 Intron2 5 T C 93 0.962 0.038 zag17 Intron2 588 G A 92 0.527 0.473 zag del Intron2 30 Wt del 91 0.527 0.473 zag16 Intron2 423 G A 90 0.533 0.467 zag 15 Intron2 1048 T C 91 0.478 0.522 zag24 Intron3 653 C A 92 0.005 0.995 zag25 Intron3 237 G A 93 0.995 0.005 zag26 Intron3 581 T G 92 0.005 0.995 zag35 3'reg 975 T C 93 0.478 0.522 zag34 3'reg 390 T G 93 0.022 0.978

: distance between the current SNP and the previous one ² : number of DNA samples with sequencing data wt: wild type sequence del: sequence in which positions 8077-8083 are deleted

Table 3. Raw haplotype frequency table

8

F is the Frequency. A test of Hardy- Weinberg (H-W) equilibrium was performed for each marker separately. No significant departure from H-W equilibrium was found at the 5% type I error. Haplotype frequency estimation conditions were met (Zhao et al., 2003).

The haplotypic characteristic of AZGPl is commonly observed in other human genes in Caucasians: a set of few common haplotypes (here 5), and a series of rare haplotypes.

The alleles of the markers identified as associated with obesity (zagl5, zagl7 and zag35), were present at the heterozygote state in the L3 and L21 lean subjects (see table 4). The presence of those alleles in the two subjects with the lowest AZGPl gene expression level provide some evidence of the importance of AZGPl in the obesity status. This observation is reinforced by the genomic study performed which shows clearly that the L3 and L21 subjects are close to obese subjects when looking at their entire gene expression profile. Table 4: Characteristics of markers identified in AZGPl (associated with obesity) and AZGPl mRNA expression levels in the lean and obese patients (see EP 1548445).

The p-values obtained from each Fisher's tests are presented in Table 5.

Table 5. Association results between each SNP and the obesity status p-value p-value

SNP (dominant coding) (recessive coding) zagO4 1 0.476 zagO5 0.361 0.361 zagO7 0.198 0.214 zaglO 0.08 0.476 zagl4 0.361 0.361 zagl5 0.311 0.03 zagl7** 0.03 0.311 zagl8 1 zagl9 1 1 zag23 0.361 0.361 zag35 0.311 0.03

*: uninformative coding, as all 21 individuals were in the same category. **: zaglδ and the intronic deletion (zag_del) are not displayed in the table as they are completely redundant with zagl7 (see Table 3).

Three markers were significant: zagl5, zagl7 (which represents zaglό and zag_del) and zag35.

Thus, the cluster of markers zagl5, zagl6, zagl7, zag_del and zag35 from AZGPl is associated with the obese status in samples from the Oestensson cohort (EP 1548445). As these five markers are strongly correlated (see Table 3), it is consistent to see that they provide the same strength of evidence.

Therefore, the present invention provides an isolated nucleic acid comprising SEQ ID No. 2, or a fragment thereof including position 8047, 8077-8083, 8500, 9556 or 12002, except for a single polymorphic change at one of the positions as shown below:

zagl5 at position 9556, wherein the T at this position is replaced by a C

zaglό at position 8500, wherein the A at this position is replaced by a G

zagl7 at position 8047, wherein the A in this position is replaced by a G

zag_del at position 8077-8083, wherein the nucleic acids in these positions are deleted

zag35 at position 12002, wherein the T in this position is replaced by a C.

These polymorphisms are the basis for a method of determining the predisposition of an individual to obesity, comprising the steps of: a) isolating a nucleic acid from a sample that has been removed from the patient and b) detecting the nucleotide present at one or more polymorphic sites within Seq ID No. 2 as listed hereinbefore, wherein the presence of the nucleotide specified at the polymorphic site as listed hereinbefore is indicative of a propensity of a patient to obesity.

The polymorphisms described hereinbefore define several haplotypes in the AZGPl gene that are associated with obesity. Therefore, the present invention also provides an isolated nucleic acid molecule selected from the group consisting of haplotypes 1, wherein each of haplotypes 1-3 comprises SEQ ID No. 2 with the exception that the nucleotides specified in the table 6 below for each haplotype are present at the corresponding position in Seq ID No. 2:

Table 6. Haplotypes for markers of interest

As used herein, the term "del" relates to a sequence derived from Seq ID No. 1, wherein the nucleic acids from 8077 to 8083 in Seq ID No. 2 are deleted from the corresponding position in Seq ID No. 1. The term "wt" relates to a sequence derived from ID No. 2 wherein the nucleic acids from positions 8077 to 8083 are present.

Furthermore, a method for haplotyping the AZGPl gene in an individual is provided comprising the steps of: a) isolating a nucleic acid from a sample that has been

removed from the individual; b) determining the presence of the nucleotides present at positions 8047, 8077-8083, 8500, 9556 and 12002 of the individual's copy of gene AZGPl, wherein the position numbers are determined by comparison to SEQ ID No. 2; c) assigning the individual a particular haplotype by comparison of the nucleotides present at said positions to the nucleotides recited in the haplotypes of the table 6 set forth hereinbefore. Preferably, the presence of at least one of the haplotypes set forth in the table 6 is indicative of the propensity of the individual to obesity.

The expression levels of -5000 different genes in fat biopsies taken from 7 lean and 9 obese were measured by high-density oligonucleotides microarray. This constituted their gene expression profile. A correspondence analysis (Benzecri JP. L'analyse des donnέes. Dunod, Paris; 1979; Greenacre M. Theory and application of Correspondence Analysis. 1984; Academic Press, London; Fellenberg K, Hauser N, Brors B, Neutzner A, Hoheisel JD, and Vingron M. Correspondence analysis applied to microarray data. PNAS 1998:10781-86) was then performed on these gene expression levels. Each data point in Figure 2 represents a projection of the entire gene expression profile of one subject in a three-dimensional space, as determined by correspondence analysis. The distance between subjects reflects the distance between their entire gene expression profiles. All obese subjects - but 016 patient - are located on the right side of the F3 axis while the lean subjects are on the left side of this same axis, but four lean subjects - L3, LI l, L17 and L21 - who appear among the obese subjects.

From the statistical work performed, many differentially expressed genes were found when obese subjects were compared to lean ones. The AZGPl gene, which is among these differentially expressed genes, appears down-regulated in obese subjects compared to lean subjects (see graph 2, fold change= -11.5 with a P-value <5%). The lean subjects having the lowest AZGPl gene expression level (L3 and L21) are also the ones who appear close to the obese subjects in Figure 2. The clinical parameters of those same lean subjects indicate that their percentage of truncal fat is higher than in the lean subjects who exhibit a high level of AZGPl mRNA. L21 has also a very low value of energy expenditure, compared to the values observed for the other lean subjects.

Short description of the Figures:

Figure 1: Markers of interest mapped on the genomic sequence used for SNP discovery in AZGPl. The following sequence is derived from the EMBL accession number ac004977. Markers of interest are highlighted (SNPs and deletion described in the statistical analysis). In this sequence, the deletion of zag_del is present.

Figure 2: Correspondence analysis performed on the entire gene expression profiles of 7 lean and 9 obese subjects, measured with high-density oligonucleotide microarray. Each data point corresponds to the entire gene expression profile of one subject. Lean subjects are depicted by black squares and obese subjects by grey squares. The analysis was performed using the statistical package XlStat 6.0 (Addinsoft; New York, NY).

Figure 3: AZGPl expression profile measured with high-density oligonucleotide microarray (see values in table 4).

Examples:

Example 1: DNA samples

DNA samples used for SNP discovery were from two different origins:

Most of them were purchased directly as DNA samples from the Coriell Institute for Medical Research.

Twenty one of them were prepared at RCMG from whole blood provided by Professor Claes Oestenson (see EP 1548445). All patients were non diabetic at the time when samples were taken. DNA was extracted from 200 μl of the whole blood using a silica gel -based extraction method (MagNA Pure LC DNA Isolation KIT I, Roche Molecular Biochemicals).

Example 2: SNP discovery

The mRNA sequence of AZGPl (NCBI accession number NM_001185) was mapped on the genomic sequence (EMBL accession number ac004977, LocusLink 563) to identify the genomic organization of AZGPl (exons and exons/intron boundaries). Primers were designed to amplify DNA fragments that would cover the whole gene sequence and additionally 1.5 kb upstream AZGPl start codon (ATG) and 1 kb downstream AZGPl stop codon (TAG) (Table 7). These fragments are overlapping each other. Fragments were amplified by PCR using DNA sample from several individuals as a template. The amplification conditions were as following, in a final volume of 20 μl:

• 4 ng DNA • Buffer Ix (see Table 8 for details)

• 50 μM of each dATP, dCTP, dGTP and dTTP

• 0.4 μM of each primer

• 0.4 u of polymerase (see Table 8 for details)

Amplification reactions were prepared in 96-well amplification plates with an aliquoting robot (Tecan biorobot). Parameters for procedures performed by the robot were set to minimize the possibility of cross- contamination. The thermal cycling conditions were as following: 15 minutes at 95°C for DNA polymerase activation, 35 cycles of the following steps: denaturation at 94°C for 1 min, hybridization at the

annealing temperature (Table 8) for 30 s and extension at 72°C for 1 min, and a final extension step at 72°C for 5 min. The amplification reactions were run on an MJ Research PTC-200 DNA Engine. After PCR amplification, fragments were purified using 384 Cleanup Millipore plates on a Tecan biorobot. Double strand DNA sequencing of all fragments was performed using ABI Big Dye terminator chemistry according to the manufacturer's instructions. Primers used for sequencing were the same as the ones used for fragment amplification. Sequencing reactions were performed on an MJ Research PTC-200 DNA Engine and run on an ABI 3730 sequencer. After sequencing, the polymorphism analyses were done using Polyphred software (licensed from University of Washington). Table 3 is listing all markers identified in AZGPl. Position of these markers on AZGPl genomic sequence is also highlighted in Figl.

Table 7. Primers used to amplify and sequence AZGPl.

Position in SEQ ID No. 1

Table 8. Amplification conditions for all fragments

FastStart u er Ix: 50 mM Tris-HCl, 10 mM KCl, 5 mM (NH ₄ ) ₂ SO ₄ , 2 mM MgCl ₂ , pH

8.3 25°C

² Roche buffer Ix: 10 mM Tris-HCl, 50 mM KCl, 1.5 mM MgCl ₂ , pH 8.3 25°C

³ Buffer D Ix: 30 mM Tris-HCl, 7.5 mM (NH ₄ ) ₂ SO ₄ , 3.5 mM MgCl ₂ , pH 8.5 25°C

Example 3: Haplotype frequency estimation method

Haplotype frequencies were estimated using an E-M algorithm as implemented in

Genecounting (Zhao JH, Lissarrague S, Essioux L, Sham PC. GENECOUNTING: haplotype analysis with missing genotypes. Bioinformatics. 2002 Dec, 18(12):1694-5).

This program takes into account individuals with untyped sites, and is thus providing more accurate estimations.

Example 4: Analysis of the deletion findings in 10 obese/11 lean

The genomic sequence of AZGPl was sequenced in 10 obese patients and 11 lean samples from Professor Oestenson's cohort (EP 1548445). All frequent SNPs from Table 3 were present. Association tests between the obese status and the genotypes were carried in the 11 non-redundant frequent SNPs: zagO5, zagO4, zagO7, zaglO, zagl4, zag23, zaglδ, zagl9, zagl7/zagl6/zag del, zagl5 and zag35.

Compared to previous analyses zaglδ from zagl9 could be separated. They were thus treated as two non redundant SNPs.

Two coding schemes were applied:

• A dominant coding where the heterozygotes and the homozygotes for the rarer allele are pooled in one category.

• A recessive coding where the heterozygote and the homozygotes for the most common allele are pooled in one category.

Under each coding scheme, each genotypic variable is a binary variable. For each variable created an exact 2x2 fisher test was performed. The significance threshold taken was 0.05.

Example 5: AZGPl mRNA profiling in Lean and obese subjects

Subcutaneous fat biopsies were obtained from the twenty one subjects coming from the cohort described in EP 1548445. For five subjects (Ll, L5, L12, 02 and O6), it was not possible to perform microarrays with the corresponding biopsies.

A gene expression study was performed using high-density oligonucleotide microarray gene technology provided by Affymetrix (Affymetrix GeneChip ^® Technology; Affymetrix, Inc.; Santa Clara, CA) on the remaining sixteen samples.

Example 5.1: RNA preparation

Total RNA from 500 mg subcutaneous fat tissue was isolated using the TriZol reagent (Life Technologies) and the Fast RNA green (BIOlOl) kit according to the manufacturer's protocols. Total RNA was purified from contaminating DNA using the RNeasy kit (Qiagen).

Example 5.2: Gene expression profiling by high-density oligonucleotide microarray

Syntheses of first and second strand cDNA were performed using the Superscript

Choice Gene Chip Kit (Life Technologies) and reagents from Gibco. Double stranded cDNA, containing an incorporated T7 RNA polymerase binding site, was purified by extraction with a mix of phenol: chloroform: isoamylalcohol (v/v/v. 25/24/1, Life Technologies). The organic and aqueous phases were separated by Phase Lock Gel (Eppendorf) and double stranded cDNA was recovered by precipitation according to the manufacturer's protocol and then resuspended in water. Double stranded cDNA was converted to biotin-labeled cRNA by in vitro transcription (IVT) using a T7 kit (Ambion) and biotin-containing ribonucleotides (Enzo - LOXO GmbH). The IVT-material was purified from unincorporated ribonucleotides using RNeasy spin columns (Qiagen). Following cleanup, single stranded biotin-labeled cRNA was chemically hydrolyzed to smaller fragments in 500 mM calcium acetate, 150 mM magnesium acetate, pH 8.1 for 35 min at 95°C. The reaction was terminated by chilling samples on ice.

One U95-A Affymetrix GeneChip Microarray was hybridized per sample. Each microarray contains 12559 probe sets representing -10,000 genes. All washing, hybridization, detection, and signal amplification steps were performed using a GeneChip Fluidics Station (Affymetrix Inc.; Santa Clara, CA). Fluorescence intensity data was collected from the hybridized GeneArrays using a GeneArray scanner (Affymetrix Inc.; Santa Clara, CA). The raw files containing the fluorescence intensity information were transformed into data files using the MAS 5.0 algorithm (component of GCOS 1.0 software). Only 45% of the genes mapped on the microarray were used in the analysis as

the rest of them were called absent by the MAS 5.0 algorithm. Differentially expressed genes were identified using the Roche Affymetrix Chip Experiment Analysis (RACE-A) software.

Example 6: Genotyping of zagl4, zagl5 and zaglό

Example 6.1: Cohort description

86 impaired glucose tolerant (IGT) obese male patients and 290 normal glucose tolerant (NGT) male controle subjects, with normal BMI (BMI<25 Kg/m ² ), were studied. All were Swedish male patients selected from the Stockholm Diabetes Prevention Program. IGT obese subjects had normal birth weight, normal BMI (<25 Kg/m ² ), and normal plasma glucose levels 2 hours after oral glucose tolerance tests. Concentrations of plasma glucose, plasma insulin, and other clinical characteristics were measured as described in Gu et al., (Single nucleotide polymorphisms in the proximal promoter region of the adiponectin (APMl) gene are associated with type 2 diabetes in Swedish Caucasians, Diabetes 53 Suppl 1: 31-5, 2004). Informed consent was obtained from all subjects, and the study was approved by the local ethics committees. Genomic DNA was extracted from peripheral blood using a Puregene DNA purification kit (Gentra) (Gu et al., supra).

Example 6.2: PCR-dynamic allele-specific hybridization (DASH) assay design and genotyping

A high throughput SNP (Single Nucleotide Polymorphism) scoring technique called dynamic allele-specific hybridization (DASH) (Howell, et al., Dynamic allele- specific hybridisation: a new method for scoring single nucleotide polymorphisms, Nat Biotech 17: 87-88, 1999) was used for SNP genotyping. PCR-DASH assay design and SNP genotyping were performed as described previously (Prince, et al., Robust and accurate single nucleotide polymorphism genotyping by dynamic allele-specific hybridization (DASH): design criteria and assay validation, Genome Res 11: 152-162, 2001).

Example 6.3: Statistical analyses

The aim of the statistical analysis was to confirm the previous results: at the genetic polymorphism zagl5, patients homozygotes TT and heterozygotes CT were at higher risk of being IGT obese when compared to patients homozygotes CC. A 2-by-2 contingency table was formed. The statistical test hypotheses were, using unilateral alternatives hypotheses:

Null hypothesis (HO): pi = ρθ Alternative hypothesis (Hl): pi > p0

The parameters pi and p0 are proportions of patients carrying at least one copy of the T allele at zagl5 among IGT obese patients and controls respectively. The statistical test for proportion comparison was based on the normality of the arcsinus-transformed proportions. Under the null hypothesis, the test follows a normal distribution N(0, 1). An exeat test of proportion was also added (Agresti, Categorical data analysis. New York: Wiley, pp. 59-66, 1990).

The test was performed at the type I error of 5 %. The odd ration (OR) of developing impaired glucose tolerance and obesity associated with the tested genetic characteristics at the SNP zagl5 was computed. The corresponding 95 % confidence intervals were computed using the free statistical software R.

The table below is showing the distribution of each genotype at zagl5 between the two patients groups.

The proportion of TT and CT patients was 0.79 in the obese IGT group compared to 0.7 in the control group. Carrying at least one copy of the T allele increased the odds of being IGT obese by 1.65 (95 % CI: [0.93 ; 2.94]. The null hypothesis of independence between

the genetic model and the obese IGT status was rejected versus a higher proportion of TT/CT patients in the obese IGT group at the 5 % level (z=1.77, p=0.04). Using the exeat proportion test (Agresti, supra), the results were borderline significant (p=0.055).

With this extended cohort coming from the same ethnic origin and prevention study as described in Examples 1 - 5 the genetic association between zagl5 and the obesity impaired glucose tolerance status was confirmed. In view of the complete genetic equivalence between the polymorphism zagl5, zaglδ, zag_del, zagl7 and zag35, the association is also holding true for all polymorphism in this cluster, namely zaglό, zagl7, zag_del and zag35.