[1] 14w Further Evidence of Association of OPRD1 & HTR1D Polymorphisms with Susceptibility to Anorexia Nervosa

[1] 173w Background: A recent study reported strong evidence for the involvement of a region on human chromosome 1 and genetic susceptibility to anorexia nervosa (AN). A more detailed analysis of this region has suggested 2 genes that may account for this susceptibility. These data suggest that polymorphisms in both the serotonin 1D (HTR1D) and opioid delta 1 (OPRD1) receptor genes show a significant association with restricting AN (RAN). Methods: In the current study, we have conducted an independent association study on 226 females meeting DSM-IV criteria for AN and 678 matched volunteers. Results: We genotyped 4 SNPs in HTR1D and 6 SNPs in OPRD1. 3 SNPs were found to be associated with both RAN and binge-purge AN (BPAN) within the gene for OPRD1. We also found evidence of association between 2 polymorphisms within HTR1D and RAN. Conclusions: These data support the hypothesis that polymorphisms within this region form a component of the genetic basis to susceptibility to RAN. However, further work is required to understand the processes that may be mediated by these genes.

[1] 126w Of the 4 SNPs analysed within the gene for HTR1D, 2 SNPs showed statistical evidence of association with DSM-IV characterized anorexia nervosa sub-groups (Table 3). Markers rs674386 and rs856510, which were shown to be in complete LD (Figure 1), were both significantly associated with RANs versus controls when a comparison was made of allele frequencies between the two populations (p ϭ .026 and p ϭ .018, respectively). These results were backed up by the genotypic tests. Marker rs856510 was found to be statistically significantly associated with the restricting anorexia nervosa subtype versus controls when genotype frequencies were compared (p ϭ .038). The same genotypic risk pattern was observed for SNP rs674386, although the significance of the genotypic test was reduced (p ϭ .063) compared to rs856510.

[2] 198w Of the 6 SNPs within the gene for OPRD1, 3 showed some degree of association with DSM-IV characterized anorexia nervosa sub-groups. The strongest associations were observed with SNP rs569356. Both genotype (p ϭ .0011) and allele (p ϭ .0008) frequencies at the marker rs569356 were highly significantly associated with anorexia nervosa case-control status. When subjects with the restricting and binge-purging anorexia nervosa sub-types were independently compared against the control group, both subtypes showed evidence of association with rs569356 genotype and allele frequencies (Table 4). In addition, the comparison of genotype frequencies between the two anorexia nervosa subtypes yielded a significant result (p ϭ .036). Figure 2 shows a more detailed analysis of the association between marker rs569356 and DSM-IV sub-grouped anorexia nervosa. SNP rs521809, which was moderately correlated (r 2 ϭ .48) with SNP rs569356, was also associated with RAN subtype versus controls (genotype test p ϭ .0163). Further evidence of association of the gene for OPRD1 with RAN subtype was demonstrated by SNP rs4654327 (p ϭ .0246 for the genotypic test versus controls, p ϭ .0267 for the allelic test versus controls), which was in very low LD with both SNP rs569356 and SNP rs521809 (Figure 1).

[1] 185w Previous studies have shown that a number of anorexia nervosa patients have disturbances in their central serotonergic system and that in many instances such disturbances can exist beyond successful treatment (Frank et al 2002;Frank et al 2001;Kaye 1997;Knutson et al 1998). Bergen reported the first genetic association of polymorphisms within the region of HTR1D and genetic predisposition to anorexia nervosa (Bergen et al 2003). In the comparison of the two studies only one marker was used in both: this was marker rs674386 designated HTR1D(-1123T Ͼ C) in the Bergen study. Bergen et al failed to show an association with this in the case-control analysis. We also found this to be true but have extended the analysis to look at subtypes and have reported an association between allelic frequencies at this marker. We go on to show a significant association with both genotypic and allelic frequencies in RAN with marker rs856510 (Table 2). Given the previous data of Bergen et al we suggest that these data add support to the hypothesis that the HTR1D gene is involved in susceptibility to anorexia nervosa (Bergen et al 2003).

[2] 149w For a number of years, disturbances in the opioid system have been shown to correlate with alterations in behaviour towards food (Peciña and Berridge 2000;Samarghandian et al 2003). However, until the report by Bergen, there was no evidence for a genetic association between polymorphic variants within the gene encoding OPRD1 and susceptibility to anorexia nervosa (Bergen et al 2003). In the current study we successfully report additional evidence to support the findings of Bergen et al (2003) and report a highly significant association between markers within the gene encoding OPRD1 and genetic predisposition to both anorexia nervosa per se and also to restricting and binge-purging type anorexia nervosa. In particular, the marker rs569356 shows highly significant genotypic (TT vs. CT/CC OR ϭ .54, 95%CI .39 -.75; p ϭ .0011) and allelic (C vs. T OR ϭ 1.67, 95%CI 1.25 -2.23; p ϭ .0008) associations with anorexia nervosa case-control status.

[3] 203w Further characterisation of the anorexia nervosa cases into DSM-IV defined restricting type and binge-purge type anorexia nervosa show that these strong associations are also observed when comparing restricting type patients to controls and to a lesser degree when comparing binge-purge type anorexia ner- vosa to controls. These findings are also supported by Bergen et al (2003), who reported a stronger association with RAN than with BPAN with both the OPRD1 and HTR1D genes. Figure 2 shows a graphical representation of these data. The C,C genotype is rare in all of the populations studied, though the C,T genotype seems to be more common in the disease populations than in the controls and the T,T genotype more common in the controls that in the cases. It is intriguing to note that the highest frequency of the C,T genotype is observed in the restricting type anorexics and the lowest is observed in the binge purge anorexics. The converse is also true for the T,T genotype. Given these data it is compelling to suggest that the genetic variance at this locus not only contributes to the genetic susceptibility to anorexia nervosa per se but may also be involved in defining the DSM-IV subtype of anorexia nervosa present.

[1] 61w Taken together with the data of Bergen et al (2003), these data represent compelling evidence to support the involvement of polymorphisms within gene encoding OPRD1 and genetic susceptibility to anorexia nervosa and particularly suggest a genetic susceptibility to restricting type anorexia nervosa. Alongside this, we are also able to confirm the involvement of polymorphisms within the gene encoding HTR1D and AN.

[2] 108w In their original linkage analysis, Bergen et al (2003) describe a double linkage peak within the distal portion of human chromosome 1. The first of the two peaks encompasses both the genes for HTR1D and OPRD1. Given the current data it is possible that it is one or more polymorphisms distal to the gene encoding HTR1D and closer to the OPRD1 gene that contributes to the linkage peak observed. It remains unclear what genetic polymorphisms are contributing to the second linkage peak b The genotype listed is that which yields the largest chi-squared value when compared against the other 2 genotypes combined (see Methods and Materials for details).

[3] 38w c The term "at risk allele" refers to the allele that occur more frequently in cases than in controls. In the case of the RAN/BPAN comparison, the RAN subjects are deemed 'cases' and BPAN subjects are deemed 'controls'.

[4] 75w observed by Bergen et al (2003), however it is also possible that the double peak does not represent two separate genetic events and that the two linkage peaks represent a region of extended linkage disequilibrium or some form of ancestral haplotype that encompasses polymorphisms close to the OPRD1 locus and polymorphisms towards the distal portion of chromosome 1p. Further linkage and linkage disequilibrium studies will be required to determine the exact nature of this finding.

[5] 302w The exact mechanisms by which these findings impact on the biological basis of anorexia nervosa remain unclear and to date there is no evidence to suggest that any of the polymorphisms examined have a functional consequence on the biological activity of opioid delta receptors or serotonin 1D receptors in the brain. However, recognising the role of opioid systems in hedonic aspects of eating prompts the idea that a functional consequence of an adjustment at this opioid receptor could lead to a reduced desire for food or a marked reduction in the pleasure derived from eating. While previous work has mainly linked the mu and kappa receptors to the control of food related sensations, the genetic information disclosed here should now lead to an examination of the effects of pharmacological or transgenic manipulation of the OPRD1 in animals, in order to determine the effects on motivation to eat and the occurrence of hypophagia. In principle, for anorexic patients, an Genotypic and allelic frequencies and 95% CI for OPRD1 rs569356. A significant difference was found in the genotypic frequencies between cases and controls (p ‫؍‬ .0011). The frequency of the C,T genotype was increased in anorexia nervosa patients (in particular in the restricting group) compared with the controls, this finding is complemented by a decreased frequency of T,T genotype in the anorexia nervosa population compared with the controls. A significant allelic association was also found in the anorexia nervosa population compared with the controls. A significant allelic association was also found in the anorexia nervosa versus control comparison (p ‫؍‬ .0008), with the C allele being more common among anorexia nervosa participants (irrespective of subtype) compared to controls. b The genotype listed is that which yields the largest chi-squared value when compared against the other 2 genotypes combined (see Methods and Materials for details).

[6] 38w c The term "at risk allele" refers to the allele that occur more frequently in cases than in controls. In the case of the RAN/BPAN comparison, the RAN subjects are deemed "cases" and BPAN subjects are deemed "controls."

[7] 97w alteration in specific receptors within the opioid system could lead to significant adjustments in the biological processes underlying the willingness to eat, and could even be responsible for a loss of pleasure from the sensations of eating. In keeping with this proposition is the finding that anorexia nervosa patients show a marked anhedonia and a corresponding low sensitivity to the rewarding effects of food (Davis and Woodside 2002). This anhedonia is likely to be one component in the avoidance of food by anorexics and is, at least, consistent with some modification of the expression of opioid receptors.

[8] 89w This research was carried out while the first author was in receipt of a Medical Research Council Case award at the Institute of Psychological Sciences, University of Leeds. We are grateful to the MRC for providing financial support of this project. We also acknowledge the help and support of the Discovery and Pipeline Genetics, and Translational Medicine and Genetics departments at GlaxoSmithKline for their contributions to this study. Finally we express our gratitude to all the staff and patients at Yorkshire Centre for Eating Disorders that supported this work.

[1] 163w 226 female Caucasian patients, registered at Yorkshire Centre for Eating Disorders (Leeds, UK) between 1998 and 2002, took part in this study. For each patient recruited, 3 female British Caucasian individuals were used as controls (total control population n ϭ 678). These control subjects (initially collected as part of a separate study) were matched with the anorexia nervosa patients based on gender and year of birth. Patients were diagnosed according to DSM-IV criteria for eating disorders by a consultant psychiatrist before being approached to take part in this study. Patients were classified into two groups identified as restricting AN (RAN) or binge-purge AN (BPAN). Based on this analysis of the 226 anorexia nervosa patients, 122 were characterized as having restrictor type anorexia nervosa and 104 were characterized as having binge-purge type anorexia nervosa. After complete description of the study to each subject, written informed consent was obtained. The study was approved by the Leeds United Hospital Trust Ethical Review Committee (submission number 01/085).

[2] 124w All group classifications, based on clinical diagnosis, were confirmed through the administration of the Structured Interview of Anorexia and Bulimia (SIAB) self report questionnaire (Fichter et al 1989). Participants were dichotomized according to DSM-IV subtype classification. The SIAB allows for current diagnosis that is taken as the period covering the previous three months and worst condition in the past. For this study patients had to meet the criteria for current diagnosis (i.e., previous 3 months). We further extended this to ensure that no diagnostic crossover had occurred within the last 12 months. Reports of current age, current weight, lowest and highest weight and age of onset were also recorded. From these data BMIs were calculated (BMI ϭ weight (kg)/height(m) 2 ) see Table 1.

[3] 221w Genomic DNA was prepared from peripheral venous blood samples by Tepnel Life Sciences laboratories (Manchester, UK) using proprietary Nucleon chemistry. Genotyping was carried out using a proprietary bead-based single base chain extension (SBCE) technology. SNPs were identified from public domain databases. Primer design was carried out using a GSK in-house computer program. SNPs were selected based on their physical location within each gene, and the ability to be able to design and validate assays on the selected GSK genotyping platform. Where possible, we aimed for coverage at a density of 1 SNP per 10 kb of genomic DNA (starting 10 kb upstream of the first exon, and extending 10 kb downstream of the last exon). SNP details can be found at http://hgvbase.cgb.ki.se. Primers were purchased from Metabion (D-82152 Planegg-Martinsried, Germany) and were produced in a standard 50plex format. The oligonucleotide primers were re-hydrated with 250l distilled water for 25 nmol per well. In order to validate the primers, commercially available DNA plates produced from cell lines were used (Coriell Cell Repositories, Camden, NJ). Primers were validated using the Single Base Chain Extension (SBCE) method using the GSK designed flow cytometric platform for high throughput SNP analysis (Taylor et al 2001). Genotyping was carried out using the flow cytometric analysis on a Luminex 100 flow cytometer (Taylor et al 2001) (Austin, TX).

[4] 66w Departure from Hardy-Weinberg Equilibrium (HWE) in the control population was assessed for each SNP using a chi-square test. A HWE permutation test was performed if the HWE chisquare p-value was Ͻ .05 and if at least one genotype cell had an expected count Ͻ 5 (Zaykin et al 1995). Polymorphisms with a HWE p-value (chi-square or exact) Ͻ .005 in control subjects were excluded from analysis.

[5] 142w The magnitude of linkage disequilibrium (LD), a measure of the association between alleles at different loci, can be defined by D ϭ p AB Ϫ p A p B , where p A is the allele frequency of allele A at the first locus, p B is the allele frequency of allele B at the second locus, and p AB is the joint frequency of alleles A and B on the same haplotype. In this study, a standardized measure of LD was calculated for each pair of loci: (Weiss and Clark 2002), which is simply the square of the correlation coefficient. The statistical significance of r 2 was assessed using a chi-square test (for pairs of biallelic SNPs, multiplying r 2 by the sample size gives an approximate chi-square distribution with 1 degree of freedom). LD was calculated in the control population.

[6] 44w For each SNP, testing for association between alleles/genotypes and disease status was carried out using the fast Fisher's exact test (FET) procedure. The fast FET computes exact p-values for contingency tables using the network algorithm developed by Mehta and Patel (Mehta and Patel 1983).

[7] 106w For each SNP two parameters are calculated: (1) an odds ratio (95% CI) for the "at risk allele" (the allele that appears more frequently in cases than controls); (2) an odds ratio for the "genotype" (determined by identifying the genotype that has the largest chi-square value when compared against the other 2 genotypes. For example, if a SNP has genotypes AA, Aa and aa, 3 chi-square association tests are performed: (a) AA versus Aaϩaa; (b) Aa versus AAϩaa; and (c) aa versus AAϩAa. If test (a) yields the highest chi-square value, then an odds ratio is calculated for the AA genotype versus the Aaϩaa genotypes combined).

[8] 104w Odds ratios (OR) were constructed for the "at risk allele" and "genotype" according to the formula OR ϭ (n 11 * n 22 )/(n 12 * n 21 ), where n 11 ϭ cases with "at risk allele"/"genotype", n 21 ϭ cases without "at risk allele"/"genotype", n 12 ϭ controls with "at risk allele"/"genotype", n 22 ϭ controls without "at risk allele"/"genotype". In order to avoid division or multiplication by zero, .5 was added to each cell in the contingency table. 95% confidence intervals for the ORs were calculated as follows: lower limit ϭ OR*exp(Ϫz√v), upper limit ϭ OR*exp(z√v), where z ϭ 97.5 th

[9] 102w The current analysis forms part of a much larger study of 42 candidate genes for anorexia nervosa (data to be published elsewhere). In total we tested 176 SNPs for association with anorexia nervosa and its subtypes. Due to large size of the overall study and the number of statistical tests performed, we would expect to see a number of associations, with p Ͻ .05, simply by chance. However, as this current study is essentially a replication of part of that performed by Bergen et al (2003), we are confident that these results represent true findings and are unlikely to be Type-I errors.

[10] 24w Association analysis was performed for each of the described SNPs comparing cases versus controls; RAN versus controls; BPAN versus controls; and, RAN versus BPAN.

[1] 110w A norexia nervosa is a severely debilitating disorder that affects primarily women (Hebebrand et al 1996) and has the highest mortality rate of any of the psychiatric disorders (Sullivan 1995). Occurring predominantly during adolescence (Halmi 1974;Halmi et al 1979) the disease is characterized by a pathological obsession for thinness through the control of eating behaviour. There is a considerable body of evidence to suggest that anorexia nervosa clusters within families (Fairburn et al 1999) and indeed twin studies have suggested a heritability rate of between .5 and .8 (Holland et al 1984;Holland et al 1988) suggesting that there is a strong genetic component involved in the susceptibility to this disease.

[2] 212w The biological basis of anorexia nervosa per se remains poorly understood; however there is a large body of evidence linking neural opioid systems with appetite control. Evidence from animal studies is based on the use of pharmacological agonists and antagonists (Kirkham and Cooper 1988). Human studies using agonists such as butorphenol (Berridge and Robinson 1998) and antagonists such as naloxone (Drewnowski et al 1989;Yeomans et al 1990) have implicated opioids in food craving and possibly bulimia nervosa. Opioids form part of a hedonic system that influences the expression of appetite, and it has been argued that opioid receptors, within reward pathways, may mediate the feeling of pleasure derived from eating food (Van Ree et al 2000). I t follows that a down regulation in opioid circuits, or a hyposensitivity at receptor sites, could lead to a loss of appetite involving either a passive, or active, resistance to eating. Kaye et al report significant reductions in CSF ␤-endorphin concentration in underweight and short-term weight restored anorexia nervosa patients compared with healthy volunteers (Kaye et al 1997). For a detailed review of the opioid system in anorexia nervosa see Kaye 1996. It is a plausible hypothesis that neurotransmitter adjustment at opioid receptor sites could have implications for the aetiology, or maintenance, of anorexia nervosa.

[3] 118w Changes within the serotonergic system have been shown to be involved in causing disturbances in mood and eating behaviour (Brambilla 2001;Enoch et al 1998;Kaye 1997;Kaye et al 2000;Srinvasagam et al 1995). Serotonin has been implicated in the control of appetite for more than 25 years (Blundell 1977) and a series of studies have observed significant alterations in the balance within the serotonergic system in anorexia nervosa patients and have shown that such imbalances can exist even after recovery (Frank et al 2001;O'Dwyer et al 1996;Sakuta et al 2001) Several serotonin subtypes are involved in satiety (Halford et al, 2005) and the 5-HT1B/1D agonist sumatriptan has been shown to reduce food intake in healthy women (Boeler et al 1997).

[4] 81w A recent study has identified a putative susceptibility locus on the short arm of human chromosome 1 (Grice et al 2001). By performing linkage analysis within a population of 192 families with at least two affected relatives Grice et al were unable to show evidence of linkage within the entire population. However, when the authors stratified the population by DSM-IV (American Psychiatric Association 1994) criteria for restricting anorexia nervosa they found a significant linkage with the region on human chromosome 1.

[5] 79w A more detailed analysis of this region by Bergen et al identified two candidate genes within the region (Bergen et al 2003). The genes for OPRD1 and HTR1D co-locate within a 10 cM region on the short arm of human chromosome 1 that is within the linkage peak observed by Grice et al (2001). Bergen et al provided evidence of association between individual SNPs and various haplotypes within both HTR1D and OPRD1 and anorexia nervosa (Bergen et al 2003).

[6] 50w In order to further establish the role of this region of human chromosome 1 in the genetic predisposition to anorexia nervosa, we have performed an independent case-control analysis on 226 patients with anorexia nervosa registered at the Yorkshire Centre for Eating Disorders and a population of 678 age-matched healthy volunteers.

[7] 130w In their case control study Bergen et al performed an analysis on 186 female anorexia nervosa probands and 98 European-American female controls (Bergen et al 2003). As reported this cohort does not have sufficient power to detect genetic association at the 80% level assuming an effect size of 1.6 -2.6 given the frequency of the minor alleles that Bergen observed (Bergen et al 2003). In the current study we report on a population of 226 anorexia nervosa patients and 678 healthy control individuals. This cohort represents the one of the largest case control populations studied to date and given a minor allele frequency of .1 and an expected effect size of between 1.6 and 2.6 this study has power of between 63% and 99% at the p ϭ .05 level.

[8] 79w Extensive in silico mining of our in-house SNP database identified 6 polymorphisms within the gene for OPRD1 and 4 polymorphisms within the gene for HTR1D that, following genotyping on our in-house validation set of Coriell DNA samples, had a minor allele frequency of greater than .10. The SNP identification number, the mapping position on the NCB IV35 map, the region of the associated gene where the polymorphisms map and the observed allele frequency are all shown in Table 2.

[9] 74w Hardy-Weinberg Equilibrium (HWE) was observed for all SNPs in the control population (data not shown). Significant pairwise linkage disequilibrium (LD) was observed for all SNPs in the gene for OPRD1, however the magnitude of this LD varied from weak to moderate between the 6 polymorphisms across the gene (Figure 1). As suggested by Bergen et al (2003) we too identified strong linkage disequilibrium within the gene for HTR1D in our control population (Figure 1).

[10] 153w The genes encoding OPRD1 and HTR1D are co-located within a 6.7 cM/8 Mbp region of human chromosome 1. Bergen et al have previously reported evidence of statistically significant intragenic LD within both the OPRD1 and HTR1D genes (Bergen et al 2003). In the current study, a limited analysis of LD was performed across 6 SNPs within the gene for OPRD1 spanning a distance of over 53kb. This analysis demonstrated that within the control population there is evidence of LD across this entire region. Analysis for LD across the 4 SNPs within the gene for HTR1D that span a distance of over 10kb has also demonstrated that within the control population there is evidence of strong LD across this entire region. The data of Bergen indicated the notion that there is no evidence for intergenic LD between these two genes (Bergen et al 2003) and the current data support these findings (data not shown).

[11] 165w Diagnostic crossover from AN to Bulimia Nervosa (BN) and vice versa as well as crossover from RAN to BPAN and vice versa is a well documented occurrence. It has been reported that between 8% and 62% of individuals with an initial diagnosis of anorexia nervosa go on to develop bulimic symptoms at some point during the course of their illness (Bulik et al 1997;Eckert et al 1995;Eddy et al 2002;Strober et al 1997). This instability of diagnosis introduces a limitation to any study that attempts to characterize groups based on diagnosis. In our study we minimized this through using three diagnostic time periods: the SIAB identified the current diagnosis, by examining clinical records we confirmed that this diagnosis was consistent with the clinical psychiatric diagnosis stretching back over the previous 18 months, the SIAB also makes a diagnosis when the disorder was at its worst. Therefore participants had held their diagnosis for a minimum of 18 months at the time of inclusion into the study.

[12] 120w Although the results in this paper do confirm the findings of Bergen et al (2003), it was not designed as a replication study. It forms part of a much larger investigation where SNPs were identified over 46 genes. SNP identification was carried out to utilize SNPs that were most appropriate for the genotyping platform available. In this case the most appropriate SNPs were not the same SNPS as used by Bergen et al (2003). However as the study selected SNPs across the gene at 10kb intervals, and due to the high level of LD across the region, we are confident that this protocol is sensitive enough to pick up the association in the region without having used the same SNPs.