Human Molecular Genetics, 2000, Vol. 9, No. 19 2895-2908
© 2000 Oxford University Press
Sequence variability and candidate gene analysis in complex disease: association of µ opioid receptor gene variation with substance dependence
1Genome Research, Max-Delbrück-Center for Molecular Medicine, Robert-Rössle-Strasse 10, D-13092 Berlin, 2GenProfile AG, Robert-Rössle-Strasse 10, D-13125 Berlin, Germany, 3Department of Genetics, Yale University School of Medicine, New Haven, CT 06520, USA, 4Center for Neurobiology and Behavior, University of Pennsylvania, Philadelphia, PA 19104, USA and 5Department of Genetics, Harvard Medical School, Boston, MA 02115, USA
Received 14 August 2000; Revised and Accepted 27 September 2000.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
To analyze candidate genes and establish complex genotype–phenotype relationships against a background of high natural genome sequence variability, we have developed approaches to (i) compare candidate gene sequence information in multiple individuals; (ii) predict haplotypes from numerous variants; and (iii) classify haplotypes and identify specific sequence variants, or combinations of variants (pattern), associated with the phenotype. Using the human µ opioid receptor gene (OPRM1) as a model system, we have combined these approaches to test a potential role of OPRM1 in substance (heroin/cocaine) dependence. All known functionally relevant regions of this prime candidate gene were analyzed by multiplex sequence comparison in 250 cases and controls; 43 variants were identified and 52 different haplotypes predicted in the subgroup of 172 African-Americans. These haplotypes were classified by similarity clustering into two functionally related categories, one of which was significantly more frequent in substance-dependent individuals. Common to this category was a characteristic pattern of sequence variants [–1793T
A, –1699Tins, –1320AG, –111CT, +17CT (A6V)], which was associated with substance dependence. This study provides an example of approaches that have been successfully applied to the establishment of complex genotype–phenotype relationships in the presence of abundant DNA sequence variation. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
The analysis of candidate genes is a key step in strategies for disease gene identification. Candidate genes may, on the one hand, be identified based on functional information about the disease and/or on genetic map information obtained by linkage or linkage disequilibrium (1,2). On the other hand, whole genome candidate gene approaches have been proposed as the future in the genetics of complex disease (2–4). In addition, as the human genome reference sequence comes to completion, the analysis of genetic variation is becoming increasingly important (2,5). Those few studies that have systematically compared individual candidate gene sequences, either parts of a gene (6–9) or whole gene sequences (10,11), suggest that genes and the human genome may be much more variable than previously thought. Allelic complexity in candidate genes will be large and such complexity will make the analysis of genotype–phenotype relationships difficult, particularly in the situation of complex disease (12–16). The spectrum of polymorphic profiles may include any variant, or combinations of variants (patterns), that may interact to determine those functional variations that are involved in phenotypic variation. Since it is the entire gene and its encoded protein that act as the units of function which potentially affect a phenotype (and ultimately allow first conclusions on disease mechanisms), we must analyze the entire sequences of the individual genes including their regulatory and critical intronic regions. It is therefore essential in diploid organisms to determine the specific combinations of given gene sequence variants for each of the chromosomes defined here as haplotypes. Given the high level of sequence variation now evident, the number of haplotypes may be unfeasibly high for association studies (6,15). This represents a particular challenge in this approach to candidate gene analysis, leading to the following major questions: how to analyze genotype–phenotype relationships in the situation of abundant sequence variation; how to identify those variants, or combinations of variants, that are of importance for the phenotype, given that the functionally relevant variants represent only a subset of the naturally occurring sequence variation. We have developed approaches to (i) compare candidate gene sequence information in multiple individuals; (ii) predict haplotypes from numerous variants; and (iii) classify haplotypes into functionally related categories to allow identification of those specific sequence variants, or combinations of variants (pattern), associated with the disease phenotype. Using the human µ opioid receptor gene (OPRM1) as a model system, we have combined these approaches to the analysis of genotype–phenotype relationships in complex genetic disease.
The human µ opioid receptor gene (OPRM1) is a prime candidate for substance dependence by both biology and genetic map position. Its gene product is the principal molecular target of morphine, mediating reward, tolerance and dependence (17,18). A marked interindividual variation in euphoric responsiveness to µ opioid receptor (MOR) agonists has been observed in opioid-naïve human subjects (19). Evidence from pharmacological, clinical and animal studies suggests that MORs may also be involved in the reinforcing actions of non-opioid drugs, such as cocaine and alcohol (20,21). Thus, the MOR could be a common component of addiction crossing pharmacological boundaries and may play a role in the mediation of drug reward. This hypothesis is supported by a reward-based animal model of substance dependence, morphine preference in B6 mice which have a genetic predisposition to experience euphoria-producing drugs such as morphine, but also cocaine and alcohol, in an intense and compelling manner (22). This represents on one hand an intriguing parallel to the polysubstance dependence of humans and on the other hand points to a genetic co-determination of polysubstance dependence. Genetic mapping of the morphine self-administration behavior revealed a locus of major effect harboring the OPRM1 gene (22). Thus, analysis of genetic variation in this gene in relation to substance dependence in humans may contribute important clues to the pathophysiology of addiction. In our study, we have focussed on those individuals more likely to have a significant genetic component to etiology of their disease, to increase the power of the sample to detect the relevant gene(s) (23).
In this study, we have—for the case of the OPRM1 gene—systematically analyzed candidate gene variation in all known functionally relevant regions of the gene including 6.7 kb regulatory, exonic and critical intronic sequences in a total of 250 substance-dependent individuals and controls. Critical intronic sequences include splice junctions, branch-point sequences, polypyrimidine tracts and in particular (A/T)GGG repeats (24). We have applied multiplex sequence comparison (MSC) to generate a substantial body of sequence data on multiple individuals for the same gene, equivalent to the analysis of 
1.7 Mb. Evidence for remarkable DNA sequence variation was obtained, as demonstrated by a total of 43 variants and 52 different haplotypes predicted for the subgroup of 172 African-American substance-dependent individuals and controls. Moreover, in our study, we classified both haplotypes and genotypes by similarity clustering, which allowed establishment of associations between (functionally) related groups of haplotypes and genotypes, respectively, and the disease phenotype. By this approach, we were able to extract a characteristic pattern of five sequence variants in the OPRM1 gene, which was observed significantly more frequently in several forms of substance dependence, one of the most devastating complex diseases known.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Analysis of OPRM1 gene sequence variation
High throughput analysis of DNA sequence variation was performed by application of a novel comparative sequencing approach, MSC, which is based on the principle of multiplex DNA sequencing (25). MSC has been designed to screen for DNA sequence differences over multiple candidate DNA segments in many individuals in parallel. This approach, which scales up conventional gel-based sequencing technologies by one order of magnitude, will be described in more detail elsewhere (M.R. Hoehe et al., in preparation). A short description of the principles and the protocols is given in Materials and Methods. This study describes the first application of MSC to the analysis of candidate gene variation on a megabase scale. The OPRM1 gene including its 5' regulatory, exonic and critical intronic regions (24) was dissected into target DNA segments (Fig. 1). For each individual, 15 PCR products covering either coding or complementary strands were pooled and simultaneously sequenced by MSC. Two hundred and fifty individual OPRM1 genes, both strands, were analyzed. Given
6.7 kb per gene, this amounts to 1.7 Mb of ‘comparative’ sequence data. Examples of typical data generated by this approach, demonstrating both heterozygous and homozygous sequence variations in individuals, are shown in Figure 2.
|
|
Regarding nature and distribution of sequence variation in OPRM1 in two major populations, 172 African-Americans and 66 European-Americans, a total of 43 biallelic variants were identified by sequencing (Fig. 3), 40 of which were single nucleotide polymorphisms (SNPs) (93%). Among those, transition substitutions were more prevalent (25 of 40, 62.5%) than transversions (15 of 40, 37.5%). Three insertion/deletion variants (7%) were found. Among these 43 variants, 15 were observed only once. Twenty-four variants were present in the 5' regulatory region (1 every 99 bp), 4 in the 5'-untranslated region (UTR) (1 every 73 bp), 1 in the 3'-UTR (1 every 710 bp), 6 in the coding region (1 every 267 bp), 5 of which clearly affect the encoded protein, and 8 variants in the intronic regions (1 every 261 bp). The critical regions of intron 2 appeared more polymorphic (1 every 110 bp) than those of introns 1 and 3 (1 variant in
1580 bp). Clearly, the relative number of variable sites in non-coding regions (1 every 147 bp) was higher than in coding regions. In the sample of African-Americans, 4 variants, –172, 17, IVS2+31 and IVS2+691, exhibited a frequency of 10% or more in the group of cases (n = 137) and/or controls (n = 35) (Table 1). Of the variants identified in the second pass of sequencing, variants –2358 and –2078 were seen in 5 and 2 of the African-Americans, respectively. Nine variants exceeded a heterozygosity of 10% in the group of cases and/or controls, 4 of these a heterozygosity of 20%, the mean observed heterozygosity for the 43 variants in this ethnic group being <0.065. Among European-Americans, 5 variants (–1748, –1320, –1045, –172 and –38) were present with frequencies between 1 and 10% in the groups of cases and/or controls and 3 variants (118, IVS2+31 and IVS2+691) exceeded a frequency of 10%. Significant ethnic differences were found for variant 17, which was observed in African-Americans only (P < 0.005), and variant 118, which was more frequent in European-Americans (P < 0.05). Variant –1236 was identified in one European-American only. All observed genotype frequencies in all subgroups were in Hardy–Weinberg equilibrium.
|
|
Variants of interest in the 5' regulatory region were those potentially affecting transcription regulatory motifs (24): a T
A exchange in a YY1 transcription factor binding motif at nucleotide position –1793 neighboring a CRE motif and a T insertion in an AP-1 binding site at nucleotide position –1699 (Fig. 3). In the coding region, two amino acid substitutions (A6V at bp 17 and N40D at bp 118) were found in the N-terminus (26–28). Additional substitutions of highly conserved amino acids were observed in the third transmembrane domain, N152D at bp 454, and in the third cytoplasmic loop, R265H at bp 794 and S268P at bp 802, which destroyed a CaM-kinase II-dependent phosphorylation site (29). The N152D mutation was found in two African-American patients, the R265H and the S268P mutations were found in one African-American patient each. The intronic variant IVS2+31GA disrupts one of the (A/T)GGG repeats, which have been found to regulate alternative splicing (30).
Haplotype (genotype) analysis
As the next step, we applied statistical techniques to predict the specific combinations of variants, haplotypes, on each of the two chromosomes of each individual. We estimate that current experimental methods (15,31) to determine haplotypes to be 10 times more costly than the combination of algorithm and diploid (mixed) sequencing applied here. The ethnic groups were analyzed separately; in the following, we concentrate on the notably larger group, the African-Americans. As described the heterozygosity (32) in our sample is not high, the mean observed heterozygosity for the 25 variants included in haplotype analysis being <0.1 (Table 1). We observed 25 of 172 individuals being homozygous at all variable positions and 33 individuals being heterozygous at only one single site, each providing unambiguous information about the haplotypes. The average genotype was heterozygous at three sites. Therefore, a reliable statistical prediction of the genetic haplotypes can be expected. Nevertheless, potentially important results obtained by statistical haplotype prediction will need to be confirmed by molecular genetic techniques. To process numerous variants, the computer program MULTIHAP has been developed by us (see Materials and Methods); MULTIHAP has been specifically designed to predict the most likely haplotype pair for each genotype in a sample. We included in the analysis those 25 variants of the total of 43 that occurred more than once, with the exception of variant 942, a third-base exchange not involving an amino acid exchange (Fig. 3; Table 1). In the combined group of 172 substance-dependent individuals and controls, 52 different haplotypes could be distinguished (Table 2). The five most frequent haplotypes common to both groups were haplotypes no. 43 (39 and 36% in cases and controls, respectively), no. 14, which was completely identical to the reference sequence (12 and 19%), no. 4 (7% in both groups) and nos 24 (4 and 6%) and 7 (4 and 6%). These five haplotypes constituted 66% of the haplotypes in the group of substance-dependent individuals and 74% of haplotypes in the controls. In the smaller European-American sample, 13 different haplotypes were distinguished. The most frequent ones corresponded to those in the African-American sample.
|
Classification of haplotypes (genotypes) and risk pattern identification
Obviously, the number of different haplotypes is unfeasibly large, so that the power is not sufficient to detect an association with any single haplotype. Previous approaches to perform association analysis, many different haplotypes given, were based on evolutionary relationships (33–35). Another approach to cope with the multiplicity of haplotypes could be the classification of haplotypes into functionally related (ideally functionally equivalent) ones based on sequence–structure–function similarity. Once a classification has been derived, the haplotype frequencies of cases and controls in the different classes can be compared. Because it is not known how many different classes may exist (if they exist), a stepwise classification process such as a hierarchical cluster procedure appears suitable. This procedure is based on an algorithm which starts with each haplotype in a separate cluster (step 0) and merges step by step the two most similar clusters until one final cluster is left (for our sample step 21). The results of the cluster analysis are illustrated by a dendrogram. The classification procedure is performed independently of the phenotype. The existence of functionally different classes would be likely, if at least one class included haplotypes from cases significantly more (or less) frequently than haplotypes from controls. In that case, the haplotypes in the different classes are inspected for consensus patterns. The pattern(s) observed more frequently among individuals with disease could be interpreted as susceptibility pattern(s) whereas pattern(s) more frequent among controls could be considered protective.
The results of the hierarchical cluster analysis of OPRM1 haplotypes are presented in Figure 4. This cluster analysis provided the basis for comparing the haplotype frequencies of cases and controls between the clusters, which have been derived at each step of the clustering process; the most significant P-value of 0.017 was reached at step 20, where two clusters were left. Haplotypes in one of these clusters corresponded to a mixture of cases and controls, whereas the other cluster contained with one exception haplotypes of substance-dependent individuals only. Analyzing the haplotype profiles in this second cluster, haplotype numbers 3, 9, 11, 17, 25, 27, 31, 37, 45, 49, 51, 16 and 39 were found combined (Table 2). The first 11 haplotypes featured a unique constellation of five polymorphic sites (pattern) located in the 5' regulatory region and exon 1. The first two sites at nucleotide positions –1793 and –1699 were those affecting sequence motifs potentially critical for transcriptional regulation (24), the third site involved an AG exchange at nucleotide position –1320, the two additional sites were located in exon 1, at nucleotide positions –111 and 17 (A6V). Two more haplotypes were found in this cluster, which contained three of these sites (haplotype nos 16 and 39) and in addition a variation 5' upstream, at nucleotide position –2078, affecting a potential CRE/AP-1 motif (24). This additional variation was observed exclusively in these two individuals. Comparing these haplotypes, one might assume that the second, third and fourth of the polymorphic sites common to all of them may be those related to substance dependence. On the other hand, 92% of the haplotypes in this cluster contain the complete combination of these five sites. This clearly stresses this constellation; the observation that the combination of three of these polymorphic sites in haplotypes 16 and 39 is complemented 5' upstream by a variation at a potentially relevant CRE motif might indicate a second haplotype pattern related to the disease. Since the importance of these polymorphic sites and their combinations is yet unknown, we have conservatively used the class containing the haplotypes with the complete pattern as the basis for association analysis and included haplotypes nos 16 and 39 in a second pass. The remaining cluster, in contrast, did not yield any evidence for a distinctive pattern.
|
Classifying the corresponding genotypes (Table 3) by similarity clustering resulted in the distinction of three categories. The first cluster contained exclusively substance-dependent individuals, who were homozygotes for the described haplotype pattern. The second included—with one exception—substance-dependent individuals, who were heterozygotes, and the third included again a mixture of substance-dependent individuals and controls without any characteristic features. It is interesting to note that genotypes 55 and 56 (Table 3) containing haplotypes 16 and 39 (Table 2) were found included in the second cluster (results of the hierarchical cluster analysis are not shown). These results confirm the haplotype-based analysis and one can assume a dominant mode of action. Importantly, the data suggest that the predicted ‘statistical’ haplotypes may reflect the genetic haplotypes. In fact, subsequent allele-specific PCR experiments utilizing as template a gene segment, which included all five polymorphic sites of this pattern, clearly confirmed this for each carrier of the pattern. Therefore, the genetic haplotypes characterized by the pattern and not the statistically predicted haplotypes were utilized in the final phase of statistical analysis.
|
The frequencies of haplotypes containing this pattern were compared in a final sample of 51 African-American controls and 158 African-American substance-dependent individuals. These included those individuals who had been genotyped by MSC (Table 3) (35 controls, 137 cases) and those who had been recruited in addition (see Materials and Methods: Subjects; 16 controls, 21 cases) and genotyped by means of allele-specific PCR for absence or presence of the pattern. Both groups, the 51 controls and 158 cases, were significantly different (P = 0.002); including haplotypes 16 and 39, a P-value of 0.001 was obtained. Also, the subgroups of cocaine- (n = 125) and opiate- (n = 33) dependent individuals differed significantly from controls (P = 0.004 and P = 0.006, respectively). Including haplotypes 16 and 39 in the subgroup of cocaine-dependent individuals, a P-value of 0.002 was reached. These results were confirmed by association analysis based on genotypes, combining individuals, who were homo- or heterozygous for this pattern. Again, significant differences were obtained in all corresponding tests, including genotypes 55 and 56 in a second pass, with P-values ranging between 0.002 and 0.006. Estimated odds ratios for substance dependence (pooling heterozygous and homozygous genotypes carrying the pattern of five variants) were 9.85 (51 controls versus 158 cases; CI 1.3–74.5) and 10.8 (including genotypes 55 and 56; CI 1.4–81.3) (36). These results suggest that distinct allelic categories may exist, one of which may be related to substance dependence.
A parallel approach to perform association analysis in this situation of haplotype multiplicity was based on established cladistic approaches that estimate the ‘evolutionary relatedness’ of haplotypes. An evolutionary tree was constructed by means of the distance-based neighbor-joining method (37) (see Materials and Methods). One defined region of the cladogram was found associated with a significant phenotypic effect (P = 0.021). Specifically, one clade in this region included precisely those haplotypes which were characterized by the unique combination of five variants described above. Clearly, these results were in excellent agreement with the results obtained by means of the hierarchical cluster analysis.
Accordingly, haplotype calculations and hierarchical cluster analysis were carried out in the European-American sample. The final sample includes 49 controls and 50 cases (n = 99); of these, 38 controls and 23 cases (n = 61) had been genotyped by MSC (data not shown), the additional individuals (11 controls, 27 cases) had been genotyped by allele-specific PCR for absence or presence of the pattern. Haplotype analysis and subsequent haplotype classification were performed. However, no evidence of any haplotype pattern, or any common variant, in this smaller sample was given.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
In our analysis we were forced to address a problem which we expect to be of increasing importance in the future: the enormous sequence variability in human genes, making it extremely difficult to analyze complex genotype–phenotype relationships. Systematic analysis of variation in entire candidate gene sequences results in numerous individually different forms of the gene identified as haplotypes. This requires a logical approach to reduce the multiplicity of haplotypes. A reasonable approach to this problem seems to classify those haplotypes that are similar in function. If a relationship between candidate gene haplotypes and a disease phenotype exists, then significant frequency differences between cases and controls should be observed between the haplotype classes. Consequently, the sequence variants common to such a functionally related class of haplotypes must include those variants that determine (or serve as markers for) differences in gene function. The critical question now arises of how to derive a classification that may reflect functionally different haplotype classes.
The hierarchical cluster analysis used to classify haplotype information is a simple and robust method to describe and analyze haplotype or genotype similarities. Given that different clustering methods may produce different solutions based on the same input data, the question naturally arises of whether the derived clusters are valid. Certain criteria, such as the stability of results or their consistency with the data (38), decide which of two different clustering results may be better. In order to check the stability of our haplotype clustering results, we have in addition applied several combinations of clustering measures (‘Squared Euclidean distance’, ‘Euclidean distance’, ‘Simple matching’) and special methods (‘Between-groups linkage’, ‘Centroid clustering’, ‘Nearest neighbor’). All applications resulted in a similar clustering and, importantly, the same two classes were consistently obtained in the last classification step, which determined here the result. Additionally, applying the non-hierarchical ‘K-Means cluster analysis’ with the initial cluster centers method under the assumption of two groups, exactly the same haplotypes were found combined in two clusters. At present, we are deriving similarity measures for classification procedures such that these measures may reflect properties that determine sequence–structure–function similarities, for instance physicochemical properties.
Another approach to combine information from different haplotypes (e.g. multiple restriction site haplotypes) uses the historical information in haplotypes to construct a cladogram that estimates how the different haplotypes are evolutionarily related (33,39). This allows localization of functional mutational changes in the haplotype network by the identification of phenotypic contrasts between sister clades. The use of this evolutionary tree as a statistical design may become difficult when various forces may have influenced the evolutionary history of a population, such as a high rate of recombination (40), multiple mutations to high susceptibility alleles, rapid population growth, migrations, strong selection, or co-evolutionary pressures like changes in environment. Undoubtedly evolution is an essential force that influences present genome sequence variability and the range of given functional variation. Thus, our approach that relies on sequence–function relations is not incompatible with cladistic approaches. However, reconstruction of the specific evolutionary process is neither feasible in most complex disease studies nor even essential for the analysis of genotype–phenotype relationships. It should be sufficient to focus on the ‘here and now’, the given DNA sequence and its individual variation as it determines structure and function.
Such an approach seems more generally applicable. It may also be more suitable to cope with the various scenarios that may underlie functional variation in complex disease. The spectrum of polymorphic profiles may include single significant mutations as well as mutations in different haplotype frameworks, for instance, or any combinations (patterns) of variants that may cause different or similar pathogenetic constellations. After all, the historically grown concept of ‘single causative mutations’ found to underlie discrete protein defects in Mendelian diseases may be poorly compatible with those graded functional differences active in complex disease. It is unproven and possibly fallacious in the situation of complex disease. It should be added that, in this analysis we have focussed on one candidate gene—a relatively simple level of complexity in view of the general model that both gene–gene and gene–environment interactions play a key role in the genetics of complex disease (13,16,41).
In our study, in the OPRM1 gene, we have identified a combination of variants, which is predominantly composed of a specific constellation of changes in putative transcription regulatory motifs and exists significantly more frequently in African-American substance-dependent individuals likely to have a significant genetic predisposition to their substance dependence. Given that we have strong prior arguments for the OPRM1 candidate gene hypothesis, P-values in the range 0.001–0.006 for both haplotype and genotype analysis appear sufficient to support the hypothesis of an association (42,43). Of course, we recognize that these nominal significance levels for an association are difficult to interpret and cannot be accepted as proof of an etiologic relationship. Additional support for an etiologic relationship must come from examining this hypothesis in new groups of patients and controls, as well as in family-based analysis. Our results suggest that individuals carrying this characteristic pattern may be at increased risk of developing substance dependence under certain additional environmental and genetic circumstances. Both genotypes, the heterozygous and the homozygous ones, characterized by one or two of these putative susceptibility patterns, are apparently associated with substance dependence. It is interesting to note that variation in promoter sequences fits quite well with this (dominant) mode of action, often observed in cases of overexpression or inappropriate expression patterns.
In theory, population stratification would be an alternative explanation for the haplotype–phenotype association described here. Available documentation described in Materials and Methods does not suggest the presence of a different subpopulation among the patients. Additional data, examining candidate genes that are particularly sensitive to population differences such as DRD2 (44,45) and the serotonin transporter gene (46) (B. Wendel et al., in preparation) in these same African-American individuals, show no evidence of such stratification. We therefore conclude that the most likely explanation of our data is a population-specific genetic disposition to substance dependence, which can be observed in 16% of the affected. In a global survey conducted by our group, including eight major populations of different geographic origins, this pattern could be observed in three Biaka-Africans of ten individuals tested. Thus, this observation suggests that this characteristic pattern originated in Africa. Moreover, the conservation of this combination of polymorphisms in relatively old, sub-Saharan populations (47), including the African ancestors of the younger African-American population, might point to a possible functional significance of this gene segment.
What could this association result imply about the genetic mechanism potentially involved in vulnerability to substance dependence? There are two possibilities: this pattern may serve as a marker, which indicates the location of genetic variation conferring risk to the disease and resides in linkage disequilibrium. This may refer to any neighboring genes, remote regulatory elements or any regulatory features within the large introns 1 and 3. No strong evidence for other candidate genes is presently given on either the human or the homologous murine chromosomal region. On the other hand, there is suggestive evidence that this polymorphic pattern, either one or several polymorphisms in combination, may be causative. The fact that the pattern of polymorphisms concentrates on the 5' regulatory region and does not extend further upstream, as demonstrated by additional sequence analysis, may support this hypothesis. Moreover, this pattern reflects a specific constellation of putative transcription regulatory sequence motifs. Given the finding that mice lacking MORs do not experience the reinforcing properties and reward mediated by the MOR agonist morphine and do not develop tolerance and dependence (17), we suggest that this polymorphic pattern may confer regulatory properties that result in an increase of receptor expression on exposure to drugs. Specifically, OPRM1 may be overexpressed as the result of an altered feedback mechanism, or a disturbance of regulatory properties after prolonged agonist exposure; on the other hand, OPRM1 may have altered regulatory properties in specific brain tissues. The existence of individual differences in OPRM1 expression levels in mouse strains and humans has recently been proposed (48).
Importantly, this pattern was found in individuals, who—even though diagnosed with a primary dependence—satisfied criteria for several forms of dependence including cocaine, opiate dependence and alcoholism and were characterized by a potential strong hereditary base for polysubstance dependence. This represents a striking parallel to a reward-based animal model of substance dependence, morphine-preferring B6 mice, which excessively self-administer opioids, cocaine and alcohol and obviously have a genetic co-determination for polysubstance dependence (22). This predisposition genetically maps to a segment of chromosome 10 harboring the murine µ opioid receptor gene (Oprm). Recent findings in this C57B1/6J strain indicate differential agonist regulation of striatal Oprm expression (49). Taken together, this pattern may be related to expressional alterations in central reward structures that may be involved in the genetic co-determination of several forms of substance dependence.
A series of clinical and experimental evidence has suggested an involvement of opioidergic regulatory mechanisms in the disposition to opioid and cocaine dependence and alcoholism (20,21). Interestingly, an upregulation of MOR binding has been detected by PET in cocaine-dependent men and was associated with cocaine craving in living human subjects (50). Moreover, an increase of OPRM1 expression could be compatible with the therapeutic properties of the hMOR antagonist naltrexone, which prevents relapse and craving in alcoholics and exerts differential subjective alcohol responses in subjects with a family history of alcoholism (51,52). As with any newly proposed risk factor, independent tests of our results are necessary. Moreover, evaluation of possible functional implications of the described OPRM1 variations in vitro and in vivo will be necessary to test the hypothesis that OPRM1-related mechanisms may be involved in the pathophysiology of addiction crossing pharmacological boundaries.
To conclude, this study shows the importance of comparing entire candidate gene sequences and future challenges of the genetics of complex disease. The development and application of approaches to the analysis of genotype–phenotype relationships against a background of high natural genome sequence variability can lead to testable genetic and functional hypotheses.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Subjects
Individuals from residential and non-residential addiction treatment programs were included, if they met modified Research Diagnostic Criteria (RDC) (53) and DSM-IV criteria for dependence on opioids or cocaine and whether systematic family history information revealed at least two first degree relatives with a past or present diagnosis of substance dependence. Detailed diagnostic procedures, validation of diagnoses and family histories have been described (44,53). All drug-dependent individuals gave extensive histories of dependence on opioids or cocaine, in which jobs, family, houses and all financial resources were lost in pursuit of drugs. They had serious multi-year drug dependence, with age at onset below 20 years. Although a primary drug dependence was identified (either cocaine or opioids) for each patient, nearly all these patients satisfied diagnostic criteria for dependence or abuse of multiple substances, including cocaine, alcohol and opioids. Normal volunteers were recruited from the same residential area (26). These individuals had an SADS-L (Schedule for Affective Disorders and Schizophrenia—Lifetime Version) interview and a systematic family history was obtained. Individuals with a personal or family history of drug abuse, dependence, alcohol dependence or a major psychiatric disorder (schizophrenia, unipolar or bipolar illness) were excluded from the final control group (26). The volunteers were matched by gender and ethnic background to the substance-dependent group. Recruiting of controls from the same residential area as the substance-dependent group provides some control over different proportions of African and European ancestry among the African-Americans. No attempt was made to match the normal volunteers by age. The mean ages for African-American controls (n = 51) and cases (n = 158) were 35.1 ± 5.5 and 30.1 ± 8.9 years, respectively. Fifty milliliters of EDTA-treated venous blood were obtained for DNA extraction (54). Two hundred and fifty individuals, 172 African-Americans, 66 European-Americans and 12 individuals of other ethnic backgrounds, were analyzed by MSC. The relatively high proportion of African-Americans among the sample reflected the demographics of the Philadelphia neighborhoods served by the programs from which the patients were drawn. On completion of the analysis of the results obtained by MSC, additional cases and controls, who had been recruited according to the same ascertainment and diagnostic criteria, were genotyped by means of allele-specific PCR and included in the final statistical analysis.
MSC
The principle of MSC in brief (M.R. Hoehe et al., in preparation) is as follows. Genes (both strands) are dissected into multiple target DNA segments that are PCR amplified and pooled to perform multiple (to date maximally up to 55) sequencing reactions simultaneously in one reaction tube. Pools of base-specific termination reactions are resolved on a sequencing gel and transferred by direct transfer electrophoresis (DTE) onto a nylon membrane. By sequentially hybridizing the membrane with different probes each highly specific for one of the original target DNA segments and by repeating the hybridization cycle as many times as there are PCR products in the original pool, sequences of all DNA segments can be read. The loading format allows identification of sequence differences between individual target DNA segments immediately by visual inspection of images (skilled pattern analysis) (Fig. 2).
Gene dissection and oligonucleotide design
Dissection of OPRM1, both strands, into target DNA segments is illustrated in Figure 1. PCR primers were designed to allow amplification at 60°C annealing temperature and probes designed to allow hybridization at 42°C. Specific information on the target DNA segments/PCR fragments and the oligonucleotides (PCR and sequencing primers and probes) defining the two plex pools that cover both strands of the gene will be made available on request. The PrimerSelect program (DNASTAR package) was used.
PCR and preparation of single-stranded template
Two sets of 15 PCR products per individual were generated to cover forward and reverse strands (Fig. 1), specifically 5' biotinylated antisense primers were introduced to allow direct solid-phase sequencing of the coding strand. Products were amplified separately in PTC 225 Tetrads (MJ Research, Oldendorf, Germany). The reaction mixtures contained 100 ng of genomic DNA, 10 mM Tris–HCl pH 8.3, 1.5 mM MgCl2, 50 mM KCl, 200 µM of each nucleotide, 30 pmol of each primer, 3 U Taq polymerase and H2O in a final volume of 50 µl each. Polymerase was added during a time interval of 4 min at 88°C, following a denaturation phase of 94°C for 4 min, followed by 30–35 cycles at 94°C (15 s), 60°C (15 s) and 72°C (30 s), or at 94°C (15 s), 60°C (30 s) and 72°C (1 min). All 15 PCR products per strand per individual (10 µl of each equivalent to 200 ng of PCR DNA) were pooled and purified via the 96-well QIAvac manifold system (Qiagen, Hilden, Germany) according to the manufacturer’s protocols. For solid phase-based multiplex sequencing of the coding strand, single-stranded (ss) PCR DNA was isolated by magnetic bead–streptavidin/biotin interaction. Beads had been washed according to recommendations (Dynal, Oslo, Norway), which had been modified to allow simultaneous immobilization of 15 PCR products, utilizing only about a quarter of the amount of beads recommended. Then, the pooled purified PCR products in a final volume of 160 µl were added to equal volumes of pre-washed beads and were incubated for 15 min at room temperature. Melting DNA duplexes and separation of DNA strands were performed according to the manufacturer’s protocol; however, 4- to 5-fold volumes of buffers and solutions were used to accommodate for the 15-fold amount of processed PCR DNA. At the end of the processing, beads bound with ss PCR DNA were resuspended in 16 µl of H2O.
Fifteen plex solid-phase T7/Mn sequencing of PCR DNA
T7 sequencing reactions were performed according to the manufacturer’s protocol (Amersham, Braunschweig, Germany), utilizing, however, the 2-fold amount of reagents compared with the 15-fold amount required if the reactions had have been processed individually, with the exception of the amounts of the 15 different sequencing primers. The specific modifications for 15 plex sequencing will be described in detail elsewhere (M.R. Hoehe et al., in preparation). One microliter of sample was loaded onto the sequencing gel.
Fifteen plex cycle sequencing of PCR DNA
Complementary strand sequencing was performed with 15 plex cycle sequencing utilizing 10 µl, equivalent to 50 ng, of each PCR product, Thermo Sequenase and reaction conditions as described (Amersham), with the exception that the sequencing primer was replaced by a mixture of 15 primers, 0.5 pmol per primer, respectively. Thus, 15 sequencing reactions were carried out with 1x reagents. The major part of the gene was analyzed by 15 plex sequencing; in a second pass of sequencing directed towards the 5' end of regulatory sequence and additional 3' intron 1 sequence (Fig. 1), direct sequencing of these three PCR segments, or an upgraded 18 plex pool, were used.
DTE
Ninety-six reactions (equivalent to the pooled base-specific termination reactions of 24 individuals), 1 µl each, were loaded onto an ultrathin, 0.125 mm, 5% polyacrylamide gel, resolved by size at 3150 V, transferred onto a nylon membrane (32 x 45 cm; neutral 1.2 µm Biodyne A; Pall, Dreieich, Germany) by DTE (55), following a modified speed gradient to guarantee band distances of 0.8 mm. Membranes were cross-linked by UV light for 30 s.
Sequential probing
Oligonucleotides were end-labeled according to the manufacturer’s recommendations (USB, Braunschweig, Germany). After prehybridization at 42°C for at least 15 min, one terminal transferase reaction was diluted into 8 ml of hybridization buffer [7% SDS, 10% polyethylene glycol (PEG), 0.25 M NaCl, 0.051% H3PO4, 82 mM Na2HPO4·2H2O, 10 mM EDTA (free acid), 32 mM NaOH] and incubated at 42°C overnight. The membranes were washed at room temperature in 1% SDS, 0.022% H3PO4, 69 mM NaH2PO4·H2O, 5 mM EDTA (free acid), 32 mM NaOH by use of an automated device (Umweltund Ingenieurtechnik, Dresden, Germany). Membranes were exposed to Phospho-Fluor-Screens up to 24 h at room temperature and scanned (Phospho-Fluor-Imager; Molecular Dynamics, Krefeld, Germany). Radioactive probes were removed with 0.2% SDS and 2 mM EDTA pH 8.3, at least twice for 5–10 min at 65°C.
Allele-specific PCR (AS-PCR)
In order to test the hypothesis that the combination of five variants, which constitute the pattern, specifically variants at nucleotide positions –1793, –1699, –1320, –111 and +17 (A6V), may reflect a genetic haplotype, a series of AS-PCR experiments was performed. The PCR product generated from the first PCR was used as template with two (reference sequence and mutant) reverse primers for a second round of PCR, then these new products were used as templates for a third round with two new (reference sequence and mutant) reverse primers, etc. First, a large PCR fragment of 2470 bp including all five polymorphic sites was generated according to the PCR conditions described earlier as part of the MSC protocol; forward and reverse primers were MOR1X-108F (5'-GGACTTTCATTGTACTGGTAGA-3') and Intron1R (5'-TTACCTGACAATCACATACATGAC-3'). Ten nanograms of this PCR product was then assayed with 20 pmol primers MOR1X-108F and the reverse primers MOR1X+17 (5'-GCTGGCGTTCGTGGGGG/A-3') with either G or A at its 3' end. AS-PCR reactions were performed at 68°C annealing temperature and conditions as described earlier. The PCR products obtained with both primers MOR1X+17 (5'-GCTGGCGTTCGTGGGGG/A-3') were then used separately for AS-PCR reactions utilizing sequentially as reverse primers with variable positions at their 3' ends: –111(G/A), –1320(T/C), –1699+T(G/T), –1793(T/A). Again, at each step, the PCR products were used separately, using as template 1 µl of the PCR product, which had been diluted 5–12 times. For further details regarding sequence-specific PCR reactions see Sander et al. (56). The DNA from individuals carrying the pattern was shown to amplify consistently, when PCR products generated with the mutant reverse primers were used as templates for amplification with the other mutant reverse primers.
Image analysis and genotyping
Computer images were genotyped on PCs by skilled pattern analysis independently by two expert readers, who were blind to diagnoses. Variation data from both strands, each read by the two readers, were compared and scored, if totally consistent. If not, data were checked for transfer or reading errors. In case the genotype could not be consistently resolved, the DNA segment was resequenced. Each genotype entering final analysis was based on a minimum of three consistent scores.
Haplotype analysis
To predict the most likely haplotype pair for each genotype in a given sample, we have developed and applied in this study the program MULTIHAP (http://mahe.bioinf.mdc-berlin.de ) which derives the maximum likelihood estimation of the underlying haplotype frequencies by an EM algorithm, which is similar to that of Excoffier and Slatkin (57) or Hawley and Kidd (58). Besides giving the maximum likelihood estimation of the haplotype frequencies in the population, MULTIHAP has been specifically designed to provide access to the haplotype decomposition of the sample. This forms the basis for an optimal choice of the initial haplotype frequencies as starting points for the EM iterations, which reduces computing time considerably compared with other programs (57,58) and predicts the most likely haplotype pair for each genotype after maximum likelihood estimation. The haplotype frequencies estimated for our sample by application of MULTIHAP were cross-checked against the results obtained by application of the program of Excoffier and Slatkin (57) to this sample; the results were in excellent agreement. Our haplotype program uses similar algorithms for the estimation of haplotype frequencies in a population; it differs from the other haplotype programs in that it predicts the most likely haplotype pair for each genotype in a sample that means that we obtain two haplotypes for each genotype. The prediction of these haplotype pairs is necessary for our subsequent analysis.
Classification of haplotypes (genotypes)
Haplotypes (genotypes) were progressively classified by means of the procedure CLUSTER (method = BAVERAGE, measure = SEUCLID) from the package SPSS for Windows NT, release 7.5. This is a hierarchical cluster analysis procedure, which uses the squared Euclidean distance as measure and the between-groups linkage as the cluster method. The variants were coded with 1 (identical with the reference sequence) or 2 (different from the reference sequence) in a haplotype, with 1 (homozygous for nucleotides/sequences identical with the reference sequence), 2 (heterozygous) or 3 (homozygous for nucleotides/sequences different from the reference sequence) in a genotype, respectively. This procedure is based on an algorithm that starts with each haplotype (independent of phenotype) in a separate cluster and merges step by step the two most similar clusters until one final cluster is left. The results of the cluster analysis are illustrated by a dendrogram.
Construction of an evolutionary tree
Construction was performed by means of the distance-based neighbor-joining algorithm described by Saitou and Nei (37), using the computer program NJ.PAS (http://smiler.lab.nig.ac.jp ). The distance matrix was derived from the unweighted pair-wise Hamming distances between the haplotypes, measuring the number of different variants. The principle of this method is to identify pairs of neighbors that minimize the total branch length at each stage of clustering, starting with a star-like tree. The branch lengths as well as the topology of a parsimonious tree can rapidly be obtained by application of this method. Specifically, this method is not based on a substitution model. Therefore, the NJ method has been often found to be superior to the other established tree reconstruction methods (59). Moreover, this method benefits from its high computational efficiency.
Association analysis and identification of risk patterns
Taking the results of the CLUSTER procedure as a starting point, a series of (2 x ni)-contingency tables were analyzed successively, each table corresponding to exactly one classification step of the i classification steps, each of these defined by an equal distance cluster coefficient; ni symbolizes the number of classes in the i-th classification step. In each table, the haplotype frequencies of the two different phenotypic groups (cases and controls) in the ni classes were compared and a likelihood ratio 
2 statistic was calculated. If necessary, exact tests (60) were performed to calculate significance levels by means of the program package SPSS for Windows NT version 7.5 with the EXACT test procedure. The exact test calculates an accurate significance level for sparse and unbalanced tables. It uses network algorithms based on the methods of Mehta and Patel (61–63). The results of the statistical evaluation at each iterative step are given as P-values. At the stage of maximum separation of phenotypes within the classes, the highest significance level should be reached. If at least two classes differed significantly, the clustered haplotypes were inspected for consensus patterns. Subsequently, the extracted consensus patterns were analyzed for association, using likelihood ratio 2 statistics. In these analytical procedures, corrections for multiple testing seem not to be required, because in the process of step-wise testing related to the iterative clustering steps each step depends on the preceding one as part of the same clustering process; this is no series of independent tests of the same null hypothesis in different data sets. In addition, an extensive computer simulation, which duplicated our experiment, was performed in order to examine whether our result could have been obtained by chance. We generated 10 000 repetitions of our experiment, randomizing the affection status (case or control). Randomization testing is a way of determining whether the null hypothesis is reasonable in this situation. In each of these 10 000 experiments, we have randomly reassigned 70 of the available 344 haplotypes to controls. Then we analyzed the series of (2 x ni)-contingency tables successively, each table corresponding to exactly one of the 20 classification steps in our experiment. In each table, the haplotype frequencies of the two different phenotypic groups in the ni clusters were compared and a likelihood ratio 2 statistic was calculated; if necessary, the exact test was applied. If the exact test in tables with 23 or more clusters was not computable, the asymptotic statistic was calculated.
Then we counted all those experiments of the 10 000 for which the following criteria were met. (i) One or more contingency tables were significant at the 5% level for the exact statistic, or at the 10% level for the asymptotic 2 statistics (if 23 or more clusters were given). The table for the next clustering step (if at least three clusters existed at a step) would not reach a significance level of 30% (because mixing a pattern-defining cluster with any other clusters should lead to a non-significant clustering step). (ii) At least one of the clusters in such a significant table contained more cases than it should in the case of independence, and the likelihood ratio 2 statistic for the consensus pattern from such a cluster reached the 5% level.
The simulation experiments generated at any step any possible cluster with any possible pattern (which could consist of any possible number of positions), without any restriction to the pattern. Two hundred and ninety-four of the 10 000 experiments (2.94%) fulfill the above-described conditions; of these, 105 showed a significant result at step 20, as was observed in our µ opioid receptor gene study. This is equivalent to a probability of 1.05% that our result could have been generated by chance. This corresponds well with the P-value of 0.017 derived by our approach.
Data analysis
Allele and genotype frequencies were calculated by application of the program package SPSS for Windows NT version 7.5. Hardy–Weinberg equilibrium was checked using the 2 test. All clustering methods applied to this data set were performed with the SPSS package.
|
ACKNOWLEDGEMENTS |
|---|
The authors particulary thank W. Schmidt, GenProfile AG, for performing the cladistic analysis and W. Terhalle, GenProfile AG, for performing the simulation study as well as for helpful discussions. The authors are particularly grateful for the valuable contributions of H. Lehrach and R. Reinhard, Max-Planck-Institute for Molecular Genetics, Berlin. The authors are also grateful to M. Vingron, German Cancer Research Institute, Heidelberg, for advice in simulation studies. They also thank J. Ott, Laboratory of Statistical Genetics, Rockefeller University, New York, for helpful discussions. The authors acknowledge V. Höllt for sharing information on transcription regulatory motif sequences. The authors acknowledge the excellent technical assistance of I. Grunewald and P. Heere, as well as valuable help through B. Timmermann, L. Ohl and N. Gscheidel. The authors wish to acknowledge E. Gottheil, N. Wintering, P. DeMaria, S. Weinstein, C. Kaltenbach, C. Berrettini and A. Patkar for their help in patient recruitment. This project was supported by DOE grant DE-FG02-87ER60565 to G.M.C., by German Human Genome grant 01KW9607/0 and an Ernst Schering Research Foundation Award Scholarship to M.R.H., by funding from the European Social Funds (ESF) to K.K. and by NIH grant DA11835 to W.H.B.
|
FOOTNOTES |
|---|
+ To whom correspondence should be addressed at: Genome Research, Max-Delbrück-Center for Molecular Medicine, Robert-Rössle-Strasse 10, D-13092 Berlin, Germany. Tel: +49 30 9489 2167; Fax: +49 30 9489 2166; Email: mhoehe@mdc-berlin.de
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
1 Collins, F.S. (1995) Positional cloning moves from perditional to traditional. Nature Genet., 9, 347–350.[Web of Science][Medline]
2 Collins, F.S., Guyer, M.S. and Chakravarti, A. (1997) Variations on a theme: cataloging human DNA sequence variation. Science, 278, 1580–1581.
3 Risch, N. and Merikangas, K. (1996) The future of genetic studies of complex human diseases. Science, 273, 1516–1517.
4 Lander, E.S. (1996) The new genomics: global views of biology. Science, 274, 536–539.
5 Chakravarti, A. (1998) It’s raining SNPs, hallelujah? Nature Genet., 19, 216–217.[Web of Science][Medline]
6 Nickerson, D.A., Taylor, S.L., Weiss, K.M., Clark, A.G., Hutchinson, T.G., Stengard, J., Salomaa, V., Vartiainen, E., Boerwinkle, E. and Sing, C.F. (1998) DNA sequence diversity in a 9.7 kb region of the human lipoprotein lipase gene. Nature Genet., 19, 233–240.[Web of Science][Medline]
7 Timmermann, B., Mo, R., Luft, F.C., Gerdts, E., Busjahn, A., Omvik, P., Li, G.-H., Schuster, H., Wienker, T.F., Hoehe, M.R. et al. (1998) Beta-2 adrenoceptor genetic variation is associated with genetic predisposition to essential hypertension: the Bergen Blood Pressure Study. Kidney Int., 53, 1455–1460.[Web of Science][Medline]
8 Pang, H., Koda, Y., Soejim, M., Schlaphoff, T., Du-Toit, E.D. and Kimura, H. (1999) Allelic diversity of the human plasma 
-(1,3) fucosyltransferase gene (FUT6). Ann. Hum. Genet., 63, 277–284.[Web of Science][Medline]
9 Fan, F., Liu, Ch., Tavaré, S. and Arnheim, N. (1999) Polymorphisms in the human DNA repair gene XPF. Mutat. Res. Genomics, 406, 115–120.[Web of Science][Medline]
10 Hoehe, M.R., Wendel, B., Köpke, K., Flachmeier, C., Kieffer, B.L., Berrettini, W.H. and Church, G.M. (1998) Genetic variability of the human mu opioid receptor gene and its implication for substance abuse. Am. J. Med. Genet., 81, 507.
11 Rieder, M.J., Taylor, S.L., Clark, A.G. and Nickerson, D.A. (1999) Sequence variation in the human angiotensin converting enzyme. Nature Genet., 22, 59–62.[Web of Science][Medline]
12 Weiss, K.M. (1996) Is there a paradigm shift in genetics? Lessons from the study of human diseases. Mol. Phylogenet. Evol., 5, 259–265.[Web of Science][Medline]
13 Terwilliger, J.D. and Weiss, K.M. (1998) Linkage disequilibrium mapping of complex disease: fantasy or reality? Curr. Opin. Biotechnol., 9, 578–594.
14 Weiss, K.M. (1995) Genetic Variation and Human Disease: Principles and Evolutionary Approaches. Cambridge University Press, Cambridge, UK.
15 Clark, A.G., Weiss, K.M., Nickerson, D.A., Taylor, S.L., Buchanan, A., Stengård, J., Salomaa, V., Vartiainen, E., Perola, M., Boerwinkle, W.H. et al. (1998) Haplotype structure and population genetic inferences from nucleotide-sequence variation in human lipoprotein lipase. Am. J. Hum. Genet., 63, 595–612.[Web of Science][Medline]
16 Willard, H.F. and Davies, K.E. (1998) Genetics of disease. Complex genetics, complex diseases. Curr. Opin. Genet. Dev., 8, 271–273.[Web of Science][Medline]
17 Matthes, H.W., Maldonado, R., Simonin, F., Valverde, O., Slowe, S., Kitchen, I., Befort, K., Dierich, A., Le-Meur, M., Dolle, P. et al. (1996) Loss of morphine-induced analgesia, reward effect and withdrawal symptoms in mice lacking the mu-opioid-receptor gene. Nature, 383, 819–823.[Medline]
18 Kieffer, B.L. (1999) Opioids: first lessons from knockout mice. Trends Phys. Sci., 20, 19–26.
19 Hoehe, M., Duka, T., Doenicke, A. and Matussek, N. (1984) Dose-dependent influence of fentanyl on prolactin, growth hormone, and mood. Neuropeptides, 5, 269–272.[Web of Science][Medline]
20 Kreek, M.J. (1996) Opiates, opioids and addiction. Mol. Psychiatry, 1, 232–254.[Web of Science][Medline]
21 Herz, A. (1997) Endogenous opioid systems and alcohol addiction. Psychopharmacology, 129, 99–111.[Medline]
22 Berrettini, W.H., Ferraro, T.N., Alexander, R.C., Buchberg, A.M. and Vogel, W.H. (1994) Quantitative trait loci mapping of three loci controlling morphine preference using inbred mouse strains. Nature Genet., 7, 54–58.[Web of Science][Medline]
23 Lander, E.S. and Schork, N.J. (1994) Genetic dissection of complex traits. Science, 265, 2037–2048.
24 Wendel, B. and Hoehe, M.R. (1998) The human mu opioid receptor gene: 5' regulatory and intronic sequences. J. Mol. Med., 76, 525–532.[Web of Science][Medline]
25 Church, G.M. and Kieffer-Higgins, S. (1988) Multiplex DNA sequencing. Science, 240, 185–188.
26 Berrettini, W.H., Hoehe, M.R., Ferraro, T.N., DeMaria, P.A. and Gottheil, E. (1997) Human mu opioid receptor gene polymorphisms and vulnerability to substance abuse. Addiction Biol., 2, 303–308.
27 Bergen, A.W., Kokoszka, J., Peterson, R., Long, J.C., Virkkunen, M., Linnoila, M. and Goldman, D. (1997) Mu opioid receptor gene variants: lack of association with alcohol dependence. Mol. Psychiatry, 2, 490–494.[Web of Science][Medline]
28 Bond, C., LaForge, K.S., Tian, M., Melia, D., Zhang, S., Borg, L., Gong, J., Schluger, J., Strong, J.A., Leal, S.M. et al. (1998) Single-nucleotide polymorphism in the human mu opioid receptor gene alters beta-endorphin binding and activity: possible implications for opiate addiction. Proc. Natl Acad. Sci. USA, 95, 9608–9613.
29 Mestek, A., Hurley, J.H., Bye, L.S., Campbell, A.D., Chen, Y., Tian, M., Liu, J., Schulman, H. and Yu, L. (1995) The human mu opioid receptor: modulation of functional desensitization by calcium/calmodulin-dependent protein kinase and protein kinase C. J. Neurosci., 15, 2396–2406.[Abstract]
30 Sirand-Pugnet, P., Durosay, P., Brody, E. and Marie, J. (1995) An intronic (A/U)GGG repeat enhances the splicing of an alternative intron of the chicken beta-tropomyosin pre-mRNA. Nucleic Acids Res., 23, 3501–3507.
31 Yan, H., Papadopoulos, N., Marra, G., Perrera, C., Jiricny, J., Boland, C.R., Lynch, H.T., Chadwick, R.B., de la Chapelle, A., Berg, K. et al. (2000) Conversion of diploidy to haploidy. Nature, 403, 723–724.[Medline]
32 Nei, M. (1987) Molecular Evolutionary Genetics. Columbia University Press, New York, NY.
33 Templeton, A.R., Boerwinkle, E. and Sing, C.F. (1987) A cladistic analysis of phenotypic associations with haplotypes inferred from restriction endonuclease mapping. I. Basic theory and an analysis of alcohol dehydrogenase activity in Drosophila. Genetics, 117, 343–351.
34 Sing, C.F., Havilland, M.B., Zerba, K.E. and Templeton, A.R. (1992) Application of cladistics to the analysis of genotype–phenotype relationships. Eur. J. Epidemiol., 8 (Suppl. 1), 3–9.
35 Valdes, A.M. and Thomson, G. (1997) Detecting disease-predisposing variants: the haplotype method. Am. J. Hum. Genet., 60, 703–716.[Web of Science][Medline]
36 Knapp, M. (1991) Statistische Methoden zur Assoziations- und Linkage-Analyze bei Kernfamilien. [Statistical methods for association and linkage analysis in nuclear families.] Thesis, University of Bonn, Germany.
37 Saitou, N. and Nei, M. (1987) The neighbor-joining method: a new method for reconstructing phylogenetic trees. Mol. Biol. Evol., 4, 406–425.[Abstract]
38 Arabie, P. and Hubert, L.J. (1996) An overview of combinatorial data analysis. In Arabie, P., Hubert, L.J. and De Soete, G. (eds), Clustering and Classification. World Scientific, Singapore, New Jersey, London, Hong Kong, pp. 5–63.
39 Templeton, A.R. (1995) A cladistic analysis of phenotypic associations with haplotypes inferred from restriction endonuclease mapping or DNA sequencing. V. Analysis of case/control sampling designs: Alzheimer’s disease and the apoprotein E locus. Genetics, 140, 403–409.[Abstract]
40 Templeton, A.R. (1996) Cladistic approaches to identifying determinants of variability in multifactorial phenotypes and the evolutionary significance of variation in the human genome. In: Variation in the human genome. Ciba Found. Symp., 197, 259–283.[Medline]
41 Schork, N.J. and Schork, C.M. (1998) Issues and strategies in the genetic analysis of alcoholism and related addictive behaviors. Alcohol, 16, 71–83.[Web of Science][Medline]
42 Kidd, K.K. (1993) Associations of disease with genetic markers: déjà vu all over again. Am. J. Med. Genet., 48, 71–73.[Web of Science][Medline]
43 Carey, G. (1994) Genetic association study in psychiatry: analytical evaluation and a recommendation. Am. J. Med. Genet., 54, 311–317.[Web of Science][Medline]
44 Berrettini, W.H. and Persico, A.M. (1996) Dopamine D2 receptor gene polymorphisms and vulnerability to substance abuse in African Americans. Biol. Psychiatry, 40, 144–147.[Web of Science][Medline]
45 Barr, C.L. and Kidd, K.K. (1993) Population frequencies of the A1 allele at the dopamine D2 receptor locus. Biol. Psychiatry, 34, 204–209.[Web of Science][Medline]
46 Delbrück, S.J.W., Wendel, B., Grunewald, I., Sander, T., Morris-Rosendahl, D., Crocq, M.A., Berrettini, W.H. and Hoehe, M.R. (1997) A novel allelic variant of the human serotonin transporter gene regulatory polymorphism. Cytogenet. Cell Genet., 79, 214–220.[Web of Science][Medline]
47 Tishkoff, S.A., Dietzsch, E., Speed, W., Pakstis, A., Kidd, J.R., Cheung, K., Bonne-Tamir, B., Santachiara-Benerecetti, A.S., Moral, P., Krings, M. et al. (1996) Global patterns of linkage disequilibrium at the CD4 locus and modern human origins. Science, 271, 1380–1387.[Abstract]
48 Uhl, G.R., Sora, I. and Wang, Z. (1999) The µ opiate receptor as candidate gene for pain: polymorphisms, variations in expression, nociception, and opiate responses. Proc. Natl Acad. Sci. USA, 96, 7752–7755.
49 Petruzzi, R., Ferraro, T.N., Kürschner, V.C., Golden, G.T. and Berrettini, W.H. (1997) The effects of repeated morphine exposure on µ opioid receptor number and affinity in C57BL/6J and DBA/2J mice. Life Sci., 61, 2057–2064.[Web of Science][Medline]
50 Zubieta, J.K., Gorelick, D.A., Stauffer, R., Ravert, H.T., Dannals, R.F. and Frost, J.J. (1996) Increased µ opioid receptor binding detected by PET in cocaine-dependent men is associated with cocaine craving. Nature Med., 2, 1225–1229.[Web of Science][Medline]
51 Volpicelli, J.R., Alterman, A.I., Hayashida, M. and O’Brien, C.P. (1992) Naltrexone in the treatment of alcohol dependence. Arch. Gen. Psychiatry, 49, 876–880.
52 King, A.C., Volpicelli, J.R., Frazer, A. and O’Brien, C.P. (1997) Effect of naltrexone on subjective alcohol response in subjects at high and low risk for future alcohol dependence. Psychopharmacology, 129, 15–22.[Medline]
53 Mazure, C. and Gershon, E.S. (1979) Blindness and reliability in lifetime psychiatric diagnosis. Arch. Gen. Psychiatry, 36, 521–525.
54 Lahiri, D.K. and Nurnberger Jr, J.I. (1991) A rapid non-enzymatic method for the preparation of HMW DNA from blood for RFLP studies. Nucleic Acids Res., 19, 5444.
55 Richterich, P. and Church, G.M. (1993) DNA sequencing with direct transfer electrophoresis and nonradioactive detection. Methods Enzymol., 218, 187–222.[Web of Science][Medline]
56 Sander, T., Gscheidl, N., Wendel, B., Samochowiec, J., Smolka, M., Rommelspacher, H., Schmidt, L.G. and Hoehe, M.R. (1998) Human µ-opioid receptor variation and alcohol dependence. Alcohol Clin. Exp. Res., 22, 2108–2110.[Web of Science][Medline]
57 Excoffier, L. and Slatkin, M. (1995) Maximum-likelihood estimation of molecular haplotype frequencies in a diploid population. Mol. Biol. Evol., 12, 921–927.[Abstract]
58 Hawley, M.E. and Kidd, K.K. (1995) HAPLO: a program using the EM algorithm to estimate the frequencies of multi-site haplotypes. J. Hered., 86, 409–411.
59 Hasegawa, M. and Fujiwara, M. (1993) Relative efficiencies of the maximum likelihood, maximum parsimony, and neighbor-joining methods for estimating protein phylogeny. Mol. Phylogenet. Evol., 2, 1–5.[Medline]
60 Agresti, A., Mehta, C.R. and Patel, N.R. (1990) Exact inference for contingency-tables with ordered categories. J. Am. Stat. Assoc., 85, 453–458.[Web of Science]
61 Mehta, C.R. and Patel, N.R. (1983) A network algorithm for performing Fisher’s exact test in r x c contingency tables. J. Am. Stat. Assoc., 78, 427–434.[Web of Science]
62 Mehta, C.R. and Patel, N.R. (1986) A hybrid algorithm for Fisher’s exact test on unordered r x c contingency tables. Commun. Stat., 15, 387–403.
63 Mehta, C.R. and Patel, N.R. (1986) FEXACT: a FORTRAN subroutine for Fisher’s exact test on unordered r x c contingency tables. ACM Transact. Math. Software, 12, 154–161.
64 Wang, J.B., Johnson, P.S., Persico, A.M., Hawkins, A.L., Griffin, C.A. and Uhl, G.R. (1994) Human µ opiate receptor: cDNA and genomic clones, pharmacologic characterization and chromosomal assignment. FEBS Lett., 338, 217–222.[Web of Science][Medline]
65 Auffray, C., Lillie, J.W., Korman, A.J., Boss, J.M., Frechin, N., Guillemot, F., Cooper, J., Mulligan, R.C. and Strominger J.L. (1987) Structure and expression of HLA-DQ alpha and -DX alpha genes: interallelic alternate splicing of the HLA-DQ alpha gene and functional splicing of the HLA-DQ alpha gene using a retroviral vector. Immunogenetics, 26, 63–73.[Web of Science][Medline]
66 Antonarakis, S.E. and the Nomenclature Working Group (1998) Recommendations for a nomenclature system for human gene mutations. Hum. Mutat., 11, 1–3.[Web of Science][Medline]
67 Beutler, E., McKusick, V.A., Motulsky, A.G., Scriver, C.R. and Hutchinson, F. (1996) Mutation nomenclature: nicknames, systematic names and unique indentifiers. Hum. Mutat., 8, 203–206.[Web of Science][Medline]
68 Ad Hoc Committee on Mutation Nomenclature (1996) Update on nomenclature for human gene mutations. Hum. Mutat., 8, 197–202.[Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Ravindranathan, G. Joslyn, M. Robertson, M. A. Schuckit, J. L. Whistler, and R. L. White Functional characterization of human variants of the mu-opioid receptor gene PNAS, June 30, 2009; 106(26): 10811 - 10816. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Zhang, X. Luo, H. R. Kranzler, J. Lappalainen, B.-Z. Yang, E. Krupitsky, E. Zvartau, and J. Gelernter Association between two {micro}-opioid receptor gene (OPRM1) haplotype blocks and drug or alcohol dependence Hum. Mol. Genet., March 15, 2006; 15(6): 807 - 819. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Zhang, D. Wang, A. D. Johnson, A. C. Papp, and W. Sadee Allelic Expression Imbalance of Human mu Opioid Receptor (OPRM1) Caused by Variant A118G J. Biol. Chem., September 23, 2005; 280(38): 32618 - 32624. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R.-S. Wang, L.-Y. Wu, Z.-P. Li, and X.-S. Zhang Haplotype reconstruction from SNP fragments by minimum error correction Bioinformatics, May 15, 2005; 21(10): 2456 - 2462. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. J. Kreek, G. Bart, C. Lilly, K. S. Laforge, and D. A. Nielsen Pharmacogenetics and Human Molecular Genetics of Opiate and Cocaine Addictions and Their Treatments Pharmacol. Rev., March 1, 2005; 57(1): 1 - 26. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. D. Hurley, L. J. Engle, J. T. Davis, A. M. Welsh, and J. E. Landers A simple, bead-based approach for multi-SNP molecular haplotyping Nucleic Acids Res., January 6, 2005; 32(22): e186 - e186. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P.-W. Lee, S. Wu, and Y.-M. Lee Differential Expression of {micro}-Opioid Receptor Gene in CXBK and B6 Mice by Sp1 Mol. Pharmacol., December 1, 2004; 66(6): 1580 - 1584. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. Skarke, A. Kirchhof, G. Geisslinger, and J. Lotsch Comprehensive Mu-Opioid-Receptor Genotyping by Pyrosequencing Clin. Chem., March 1, 2004; 50(3): 640 - 644. [Full Text] [PDF]
|
|
|
|
 |
|
 C. Burgtorf, P. Kepper, M. Hoehe, C. Schmitt, R. Reinhardt, H. Lehrach, and S. Sauer Clone-Based Systematic Haplotyping (CSH): A Procedure for Physical Haplotyping of Whole Genomes Genome Res., December 1, 2003; 13(12): 2717 - 2724. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Kraus, C. Borner, E. Giannini, and V. Hollt The Role of Nuclear Factor {kappa}B in Tumor Necrosis Factor-Regulated Transcription of the Human {micro}-Opioid Receptor Gene Mol. Pharmacol., October 1, 2003; 64(4): 876 - 884. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I. Mai, E. Stormer, M. Goldammer, A. Johne, H. Kruger, K. Budde, and I. Roots MDR1 Haplotypes Do Not Affect the Steady-State Pharmacokinetics of Cyclosporine in Renal Transplant Patients J. Clin. Pharmacol., October 1, 2003; 43(10): 1101 - 1107. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Sauer, H. Lehrach, and R. Reinhardt MALDI mass spectrometry analysis of single nucleotide polymorphisms by photocleavage and charge-tagging Nucleic Acids Res., June 1, 2003; 31(11): e63 - e63. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. D. Mitra, V. L. Butty, J. Shendure, B. R. Williams, D. E. Housman, and G. M. Church Digital genotyping and haplotyping with polymerase colonies PNAS, May 13, 2003; 100(10): 5926 - 5931. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 X. Wang, Z. Fan, J. Huang, S. Su, Q. Yu, J. Zhao, R. Hui, Z. Yao, Y. Shen, B. Qiang, et al. Extensive Association Analysis Between Polymorphisms of PON Gene Cluster With Coronary Heart Disease in Chinese Han Population Arterioscler. Thromb. Vasc. Biol., February 1, 2003; 23(2): 328 - 334. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Wilkie, A. Osei-Lah, B. Chioza, L. Nashef, D. McCormick, P. Asherson, and A. J. Makoff Association of {micro}-opioid receptor subunit gene and idiopathic generalized epilepsy Neurology, September 10, 2002; 59(5): 724 - 728. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. Borner, V. Hollt, and J. Kraus Involvement of Activator Protein-1 in Transcriptional Regulation of the Human {micro}-Opioid Receptor Gene Mol. Pharmacol., April 1, 2002; 61(4): 800 - 805. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Kraus, C. Borner, E. Giannini, K. Hickfang, H. Braun, P. Mayer, M. R. Hoehe, A. Ambrosch, W. Konig, and V. Hollt Regulation of {micro}-Opioid Receptor Gene Transcription by Interleukin-4 and Influence of an Allelic Variation within a STAT6 Transcription Factor Binding Site J. Biol. Chem., November 16, 2001; 276(47): 43901 - 43908. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Befort, D. Filliol, F. M. Decaillot, C. Gaveriaux-Ruff, M. R. Hoehe, and B. L. Kieffer A Single Nucleotide Polymorphic Mutation in the Human {micro}-Opioid Receptor Severely Impairs Receptor Signaling J. Biol. Chem., January 26, 2001; 276(5): 3130 - 3137. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. Wang, J. M. Quillan, K. Winans, J. L. Lucas, and W. Sadee Single Nucleotide Polymorphisms in the Human {micro} Opioid Receptor Gene Alter Basal G Protein Coupling and Calmodulin Binding J. Biol. Chem., September 7, 2001; 276(37): 34624 - 34630. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||