Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Recessive genetic mode of an ADH4 variant in substance dependence in African-Americans: A model of utility of the HWD test

Recessive genetic mode of an ADH4 variant in substance dependence in African-Americans: A model... Background: In our previous studies, we reported positive associations between seven ADH4 polymorphisms and substance dependence [i.e., alcohol dependence (AD) and/or drug dependence (DD)] in European-Americans (EAs). In the present study, we address the relationship between ADH4 variation and substance dependence in an African-American (AA) population, and report evidence that supports an association between a different ADH4 polymorphism (rs2226896) and these phenotypes in AAs. Methods: Two family-based association study methods, i.e., TDT and FBAT, were applied to test the relationship between ADH4 variation and substance dependence in Sample 3 (112 small nuclear families) and in Sample 4 (632 pedigrees), respectively. A population-based case-control association study method was also applied to test this relationship in 1303 unrelated subjects, with and without controlling for admixture effects. Finally, a Hardy-Weinberg Disequilibrium (HWD) test was applied to examine the association in the case-only sample, infer the genetic disease models, and distinguish the disease and non-disease factors contributing to HWD. Results: The marker examined was found to be in significant HWD in AA alcoholics (p = 0.0071) and drug dependent subjects (p = 0.0341), but in Hardy-Weinberg Equilibrium (HWE) in all other subgroups. Other association methods failed to detect any association between this variation and phenotypes. The best-fit genetic disease model for this marker is a recessive genetic model. Conclusion: ADH4 variation might play a role in risk for substance dependence in AAs, potentially via a recessive mechanism. Under certain conditions, the HWD test could be a more powerful association method than conventional family-based and population-based case-control association analyses, for which, the present study provides an extreme example. Page 1 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 We also demonstrated that ADH4 genotypes predispose Background The rate of alcohol metabolism influences drinking to AD and DD consistent with a recessive mode of inher- behavior and affects risk for alcohol dependence (AD). itance in EAs. These associations remained after control- The alcohol dehydrogenases (ADH) and the aldehyde ling for admixture effects and were confirmed by a dehydrogenases (ALDH) are the two major categories of family-based association study. However, in AAs, these alcohol-metabolizing enzymes in the liver. ADH converts seven ADH4 markers were in HWE in both cases and alcohol to acetaldehyde (which is toxic) and then ALDH controls; no association between alleles, genotypes, hap- converts acetaldehyde into acetate. Acetate is then oxi- lotypes, or diplotypes of these markers and AD and DD dized into CO and H O via the tricarboxylic acid cycle. were found in this population, even after controlling for 2 2 The development of AD is related to an individual's level admixture effects. of ethanol consumption, which is influenced by ADH and ALDH activity. The human ADH4 enzyme (i.e., π subunit) Although we did not find a significant association is an important member among these alcohol-metaboliz- between these seven ADH4 variants and AD or DD in AAs ing enzymes. It mainly contributes to liver ADH activity, in our previous studies, in view of the power variance of and at intoxicating levels of alcohol it may account for as association methods and the population-specificity of much as 40% of the total ethanol oxidation rate [1]. Thus, associations, we did not exclude the possibility that these physiologically significant variation in ADH4 activity associations might be significant if adequate statistical could clearly contribute to variability in risk for AD. power were available, or that other markers might be asso- ciated with these phenotypes in this population [6,7]. The The human ADH4 enzyme is encoded by the ADH4 gene, present study aimed to identify an association between a which maps to 4q22 within the ADH gene cluster. We specific polymorphism, rs2226896 (for convenience and focused on the ADH4 gene in this study because of the to distinguish it from the seven polymorphisms on which functional importance of its protein product, remarkable we reported previously, we refer to this variant as SNP8) for its activity at high ethanol concentrations that might and AD, in AAs. This polymorphism has a rare allele and be of particular relevance in the context of AD risk [2]; and a rare homozygote which could result in its being inde- because previous studies have demonstrated that ADH4 is pendent of the flanking haplotype block and could an important risk gene for AD and AD-related traits in increase the relative risk between genotypes so that it European-Americans (EAs) and African-Americans (AAs). could be a more powerful marker for some specific asso- Guindalini et al. [3] reported that the ADH4 promoter var- ciation analysis methods (e.g., the HWD test) to detect iants -75A/C (rs1800759), which can significantly alter marker-phenotype association [9]. To test the population- the expression of the ADH4 enzyme [4], and -159A/G, specificity of this association, we also tested for it in EAs, were significantly associated with AD in EAs and AAs in a the most common population in the US. Brazilian population. Edenberg et al. [5] reported that six- teen ADH4 polymorphisms (including rs2226896) were As different association methods have different advan- associated with AD in an independent Collaborative tages, disadvantages, and power associated with them, in Study on the Genetics of Alcoholism (COGA) sample of the present study, we applied several methods and com- pedigrees. We have previously reported associations pared their power. Family-based association study meth- between seven ADH4 polymorphisms (including -75A/C) ods are immune to population stratification and and AD and drug dependence (DD) in EAs in the US pop- admixture effects, and (as generally applied) are not ulation [6,7]. We also reported that three ADH4 polymor- informed by the affection status of parents. This method phisms (including -75A/C) were associated with has been used as a valid confirmatory method for a popu- personality traits in AD and DD subjects [8]. lation-based association study in our previous study [6] and was applied in the present study as well. However, In our previous studies [6,7], we genotyped seven mark- because the sample size of families is usually limited, and ers in some of the subjects included in the present study. families usually provide around two-thirds of the power These seven markers span the full length of ADH4 and provided by unrelated case-control samples of similar size are in one haplotype block both in EAs and AAs. The [10], we also performed a case-control association study. seven ADH4 markers showed deviation from Hardy- AAs are an admixed population, and EAs are admixed as Weinberg Equilibrium (HWE) (called Hardy-Weinberg well, although much less so [11-14]. Admixture effects Disequilibrium, HWD) in EA substance dependent sub- may result in spurious observed association between gene jects (including patients with AD and DD), but were in and disease. A population-based case-control association HWE in EA healthy controls. Significant differences in study is vulnerable to these effects. Therefore, the degree genotype and diplotype, but not in any allele or haplo- of admixture was also measured in the case-control sam- type, frequency distributions for all seven ADH4 markers ple, and admixture effects on the case-control association were found between cases and controls in EAs (adjusted analysis were controlled by a structured association (SA) global p = 0.0070, 0.0004 for AD and DD, respectively). method [15]. Page 2 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 Our previous study suggested that a case-only HWD test Methods was more powerful (with lower p values) than a case-con- 1. Subjects The clinical samples are listed in Tables 1 and 2, including trol association analysis under a recessive model [7]. The present study provided an additional application for this Sample 1 (907 unrelated case-control subjects), Sample 3 powerful association method. Meanwhile, we used an (112 small nuclear families (SNFs); each family includes HWD test to infer the genetic disease model, via a set of parents and 1 to 2 offspring) [both samples were used in novel software programs developed by Wittke-Thompson previous studies by Luo et al [6,7]], Sample 2 (271 newly et al. [9], which incorporate the HWD information from recruited unrelated AA cases and 125 unrelated AA con- both cases and controls. HWE is subject to a number of trols), and Sample 4 (1613 related subjects from 632 ped- assumptions regarding genotyping error, selection, igrees; each pedigree has 1 to 6 affected siblings with or genetic drift, inbreeding, gene flow, and mutation, and to without parents) [Sample 4 was previously described by other factors such as population stratification, admixture, Gelernter et al [18,19]]. Samples 1 and 2 included a total nonrandom patterns of missing data, and nonrandom of 1303 unrelated subjects (757 males and 546 females), allele dropout with increasing age. HWD may be caused with 391 "genetic" EA cases, 310 "genetic" EA healthy by the violation of any one of these assumptions; thus, we controls (European ancestry proportion > 0.5), 429 also distinguished the underlying causes for HWD by the "genetic" AA cases, and 173 "genetic" AA controls (African above methodology via application of the recently-devel- ancestry proportion > 0.5; see Methods). Newly-recruited oped programs and by reviewing almost all possibility for subjects (Sample 2) were diagnosed using the Semi-Struc- the causes of HWD in details. tured Assessment for Drug Dependence and Alcoholism (SSADDA) [18,20]. The healthy control subjects (in Sam- Drug dependence (DD), which mainly includes cocaine ples 1 and 2) were screened to exclude major Axis I mental dependence and opioid dependence in our sample, is one disorders, including alcohol or drug use disorders, psy- of the most common phenotypes comorbid with AD [16]. chotic disorders (including schizophrenia or schizophre- DD has many features in common with AD, including nia-like disorders), mood disorders, and major anxiety symptomatology, neuropsychological impairment, disorders. Sample 4 included pedigrees having affected hypothesized pathogenetic mechanisms, and response to probands with substance dependence, previously used for specific treatments (reviewed by Luo et al [7]). DD has genome-wide linkage studies [18,19]. Diagnosis was also been reported to share some susceptibility genes with made according to DSM-III-R or DSM-IV criteria [21,22]. AD (reviewed by Luo et al [6,7,17]). Our previous study demonstrated that DD and AD share ADH4 as a suscepti- Samples 1, 2 and 3 were recruited at the University of bility gene in EAs [6,7]. Thus, in this study, we investi- Connecticut Health Center, the VA Connecticut Health- gated the association not only in AD, but also in DD, to care System-West Haven Campus, or the Medical Univer- test the phenotypic specificity of the observed associa- sity of South Carolina. Sample 4 was recruited at four tions. sites: University of Connecticut Health Center (Farming- Table 1: Sample size for the unrelated sample and the related small nuclear families. Sample sizes %Male Age (years) EA AA Hispanic Others Sample 1 Healthy controls 39.9 28.2 ± 9.1 310 48 Cases 73.4 39.4 ± 9.2 391 158 AD 76.1 40.3 ± 9.2 326 101 DD 68.2 37.1 ± 8.1 204 145 Sample 2 Healthy controls 28.0 33.5 ± 12.6 125 Cases 64.9 40.0 ± 9.3 271 AD 65.1 40.0 ± 9.3 269 DD 64.8 42.0 ± 8.3 71 Sample 3 SNF size 92 12 6 2 Parents 49.8 65.9 ± 8.6 183 24 11 3 %Unaffected 45.5 66.8 ± 8.6 79.8 37.5 63.6 100.0 Offspring 54.8 39.2 ± 11.1 96 13 5 1 AD 58.9 37.3 ± 8.2 77 9 3 1 DD 55.2 38.0 ± 7.3 77 13 5 1 EA, European-American; AA, African-American; AD, alcohol dependence; DD, drug dependence; SNF, small nuclear families. Samples 1 and 2 are unrelated case-control samples; Sample 3 is related SNF sample. Page 3 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 Table 2: Structure and sizes of pedigrees (Sample 4) 125 unrelated AA controls (Sample 2), and the related pedigree subjects (Sample 4) were not included in this Total EAs AAs Others analysis because the other flanking seven ADH4 markers have not been genotyped in these subjects. Subjects 1613 766 833 14 %Male 46.3 51.4 41.2 71.4 2) Transmission disequilibrium test (TDT): Family-based Age (year) 39.4 ± 8.9 37.8 ± 10.2 40.8 ± 7.2 43.7 ± 5.7 Pedigrees 632 311 316 5 association analysis was performed by comparing "cases" With 1 AD 242 115 126 1 (i.e., transmitted allele and genotypes) and artificial "con- With 2 AD 166 87 78 1 trols" (i.e., untransmitted allele and genotypes) in Sample With 3 AD 11 5 6 0 3 by the program TDTPHASE [24]. All subjects in Sample With 4 AD 52 30 3, including EA, AA, Hispanic, and others, and the affected With ≥ 5 AD 10 10 and unaffected parents, were combined in the TDT analy- With 1 DD 57 25 32 0 sis. This analysis was also performed separately within EAs With 2 DD 460 243 216 1 With 3 DD 90 35 55 0 and AAs. With 4 DD 16 7 9 0 With ≥ 5 DD 30 30 3) Family-Based Association Tests (FBAT): The structure of Sample 4 is much more complicated than Sample 3, thus, EA, AA, AD, DD, same as Table 1. "With n AD", a pedigree having n we used the program FBAT [25] to test the gene-AD and offspring with AD; as analogy. gene-DD associations under the general, additive, domi- nant, and recessive genetic models respectively. We have ton, CT), Yale University School of Medicine (New Haven, already genotyped 419 short tandem repeats (STRs) in CT), McLean Hospital, Harvard Medical School (Belmont, this sample in the initial whole genome scan linkage stud- MA), and Medical University of South Carolina (Charles- ies [18,19], and inferred the ancestry proportions in each ton, SC). All subjects gave informed consent before partic- individual. The whole sample can thus be separated into ipating in the study, which was approved by the "genetic" EAs (European ancestry proportion > 0.5) and Institutional Review Board at the respective institutions. "genetic" AAs (African ancestry proportion > 0.5), accord- ing to the ancestry proportions in probands. FBAT was 2. Genotyping performed in the whole sample including EAs, AAs, His- A marker (rs2226896; SNP8; at Chr04: 100460117) at the panics, and others. This analysis was also performed sep- putative 5' regulatory region, close to the functional vari- arately within "genetic" EAs and AAs, because the ant -75A/C (at Chr04: 100458289; 1.8 Kb to 3' of SNP8) population-specificity of the gene-disease linkage and the at the promoter and other disease-related polymorphisms disease-risk sites has ever been demonstrated in this sam- at this region (e.g., -159G/A at Chr04: 100458373, 1.7 Kb ple [18,19]. to 3' of SNP8 [3]; rs1984362 at Chr04: 100463753, 3.6 Kb to 5' of SNP8 [6,7]) was genotyped in all subjects using 4) Case-control comparisons for allele and genotype fre- the TaqMan technique ("assay-on-demand"). These eight quencies: Allele and genotype frequencies for SNP8 ADH4 SNPs were all of those that were available from among EAs and AAs in Samples 1 and 2 are shown in public sources when we started genotyping. SNP8 was not additional file 1. Associations between alleles, genotypes reported together with other seven SNPs in our previous and phenotypes were tested by comparing the allele and studies [6,7] because it was not associated with any phe- genotype frequency distributions between cases and con- notype in EAs. The detailed genotyping procedure is trols with the exact tests in the program PowerMarker described elsewhere [7]. All genotyping was performed in [26]. duplicate and compared to ensure validity of the data. Mismatched genotypes, if any, were discarded. Thirty- 5) Structured association (SA) analysis: EAs and AAs were eight unlinked ancestry-informative markers (AIMs) were taken as admixed populations with different degrees of also genotyped in unrelated subjects to estimate the admixture. The extent of admixture (i.e., average ancestry ancestry proportions for each subject (detailed by Luo et proportions) in unrelated subjects was estimated using al [6]). the program STRUCTURE [15] by analyzing 38 AIMs (in combined EA and AA subjects) [6,27,28]. Admixture 3. Statistical analysis effects on case-control association analysis can be control- 1) Linkage disequilibrium (LD) analysis: Pairwise LD led using the program STRAT [29] by conditioning the between this marker and the other seven ADH4 markers analysis on the ancestry proportions, i.e., structured asso- studied previously was analyzed separately for EAs and ciation (SA) analysis (separately for EAs and AAs). Fur- AAs in Sample 1 via the program Haploview [23]. The thermore, associations between this ADH4 variation and additional newly recruited 271 AA unrelated cases and phenotypes were also analyzed by a regression method Page 4 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 (in combined EAs and AAs), with ancestry proportions, p-value. If this p-value is less than 0.05, the genetic disease age, and sex serving as covariates. model can be rejected as a poor fit to the observed data. The best-fit genetic model would correspond to a p-value 6) Hardy-Weinberg equilibrium (HWE) test and genetic which is larger than 0.05 and is the largest among all disease model inference: An HWE test can be a valid asso- genetic models. If all of the genetic models are poorly fit- ciation method, and can be used to infer genetic disease ted, then alternative explanations for the HWD, including models. HWE was tested within populations, and sepa- chance, genotyping error, and/or violations of the requi- rately for cases and controls, using the goodness-of-fit χ test implemented in the program PowerMarker [26]. If site assumptions of HWE, must be considered. there was one cell with expected genotype count less than The direction of HWD statistics (Δ, F and J values, detailed 5, HWE was also tested by the exact test implemented in the program PowerMarker [26] and confirmed by the definitions follow) in cases is also helpful to judge the genetic models. Under the dominant and the additive exact test implemented in the online program FINETTI models, the values of Δ, F and J are negative, because [30], and the exact tests implemented in the program HWD is caused by an excess of observed heterozygote TFPGA [31] which applies conventional Monte Carlo and Markov Chain Monte Carlo (MCMC) methods. (Aa) and homozygote (AA) frequencies over the expected frequencies. Under the recessive model, the values of Δ, F and J are positive, because the HWD is caused by excess of To distinguish the causes of HWD, which include gene- observed homozygote (aa) frequency over the expected disease association and non gene-disease association fac- frequency, and in contrast, observed heterozygote (Aa) tors such as genetic drift, migration, mutation, non-ran- frequency is less than expected. These statistics were calcu- dom mating, etc., and then to infer the genetic disease lated based on the following formula: Assuming that fo', model for the marker in HWD, we used the goodness-of- fo", fo"', and fe', fe", fe"' denote the observed and expected fit χ test defined by Wittke-Thompson et al. [9], which frequencies of susceptibility homozygote, heterozygote and non-susceptibility homozygote in cases, respectively, incorporated the HWE information both from cases and then Δ = fo'''-fe''', F = 2Δ/fe'' and J = Δ/fe"' [34-36]. controls (here denoted WT goodness-of-fit χ test) [9]. The O − E O − E () j j Results 2 () i i test statistic c=+ (i, j = 1, 2 ∑∑ E E 1. SNP8 was in LD with the seven flanking markers in EAs i j ij (D' > 0.9 in each case), but provided information inde- or 3), where O and E are the observed and expected num- i i pendent of these markers in AAs (D' < 0.10 in each case) th bers of i genotype of this marker in patients, respectively; (Figure 1). TDT in Sample 3 and FBAT in Sample 4 th O and E are the observed and expected numbers of j gen- showed no significant association between the ADH4 j j SNP8 and AD or DD, whether analysis is conducted by otype of this marker in controls, respectively. Assuming combining or separating the different ethnicities (all p > different baseline penetrance of disease (from 0 to 1) for 0.05) (Sample 3: see additional file 2; Sample 4: data not non-susceptibility homozygotes, different heterozygote shown). The genotype frequency distributions in Sample relative risk, and different susceptibility homozygote rela- 4 were shown in Table 3, not for the association analysis, tive risk that determine different E and E given different i j but for providing a replication of the rare genotype fre- genetic models, and setting the lifetime population prev- quencies to those in the unrelated Samples 1 and 2 (see additional file 1). alence of alcohol dependence and drug dependence at 5.5% and 3.8% [32], respectively, we can obtain a differ- 2. SNP8 was not associated with phenotypes via case-con- 2 2 ent set of χ values. Minimizing this χ statistic over the trol comparisons either before or after controlling for entire parameter space with the appropriate constraints admixture effects. Measured admixture degrees were low on the parameters approximates a maximum-likelihood both in EAs and AAs. estimate procedure. We can then obtain the minimal (χ ) 2 No significant difference in allele or genotype frequency value, which is approximately distributed as a χ with 1 df distribution for this marker was found between unrelated for a general model and 2 df for restricted models (i.e., cases and controls in either EAs or AAs (all p > 0.05; addi- dominant, recessive, additive, and multiplicative models) tional file 1). After controlling for admixture effects by the [33]. Simulating 173 AA controls and 370 AA AD subjects SA method, these negative associations were essentially or 216 AA DD subjects for 1000 replications to obtain unchanged (all p > 0.05). Regression analysis, which takes 1000 minimal χ values and comparing these values to into account the effects of age and sex, also showed no sig- nificant association. the original minimal (χ ) value, we can derive an empiric Page 5 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 Table 3: Genotype frequency distributions of SNP8 in pedigrees (Sample 4) European-American African-American AD DD Unaffected AD DD Unaffected Nf N f NfNf N f Nf A/A 271 0.903 541 0.905 85 0.895 293 0.945 626 0.963 127 0.992 A/G 26 0.087 53 0.089 8 0.084 16 0.052 24 0.037 1 0.008 G/G 3 0.010 4 0.007 2 0.021 1 0.003 0 0.000 0 0.000 A 568 0.947 1135 0.949 178 0.937 602 0.971 1276 0.982 255 0.996 G 32 0.053 61 0.051 12 0.063 18 0.029 24 0.018 1 0.004 AD, alcohol dependence; DD, drug dependence. Two main ancestries, i.e., European and African, were with insufficient power might not detect a true genotype- detected in our sample. In the 701 "genetic" EA subjects phenotype association (in fact, evaluated by exact test as a (European ancestry proportion > 0.5), the total estimated case-control genotype frequency distribution comparison, weight of African ancestry proportions is 11.3; the admix- this comparison shows p = 0.391 and 0.225 for AD vs. ture degree of "genetic" EAs in this sample is 1.6% (= controls and DD vs. controls, respectively). Thus, in the 11.3/701). In the 602 "genetic" AA subjects (African present study, several association methods with differing ancestry proportion > 0.5), the total estimated weight of power and that take different views of the data were European ancestry proportions is 23.7; the admixture applied and compared. The most powerful HWD test sug- degree of "genetic" AAs is 3.9% (= 23.7/602). gests that the ADH4 variant (rs2226896) might play an important role in risk for substance dependence (includ- 3. SNP8 was in significant HWD in both AA AD and DD ing AD and DD) in AAs, probably via a recessive genetic subjects (with an excess of observed homozygotes over mechanism. The association between this variant and expected homozygotes): using the exact tests: p = 0.0071 phenotype is population-specific, that is, it appears in for AD and p = 0.0341 for DD, respectively. These HWDs AAs, but not in EAs. This association herein first discov- remained significant or suggestive after correction for ered in AAs is a complementary finding to a previous set multiple testing (α was set at 0.025). However, SNP8 was of genotype-phenotype relations we described for other in HWE in EA AD and DD subjects (p > 0.05) (additional markers at this locus in EAs [6,7]; based on this result, we file 1). This marker was in HWE in controls, both EA and can now provisionally conclude that ADH4 affects SD risk AA (p > 0.10). The WT goodness-of-fit χ test showed that in both EAs and AAs, but different variants are important the best-fit genetic disease model for this marker was a in the different populations. It would be of great interest recessive model (in AA AD subjects: χ = 0.621, df = 2, p = to study this variant in other populations, e.g., Asians, to 0.733; in AA DD subjects: χ = 1.596, df = 2, p = 0.450; see further characterize the population specificity we report additional file 1). In AA AD subjects for this marker, Δ = here. This variant is independent of the other seven poly- +0.008, F = +0.182 and J = +0.009; in AAs with drug morphisms that were reported previously to have no asso- dependence, Δ = +0.008, F = +0.204 and J = +0.010. ciation with substance dependence in AAs [6,7]. ADH4 gene variation is thought to influence the risk for Discussion In African-Americans (AAs), the SNP8 G/G genotype was AD by modulating ethanol metabolism. However, we find never observed in any of the control subjects in Sample 1 that it is associated with DD too. This is reasonable (n = 48), the newly-recruited unrelated controls in Sample because DD has many features in common with AD 2 (n = 125), or the additional related unaffected pedigree which are reviewed above and because the development subjects (n = 128; see Table 3) (i.e., in total there were 0/ of AD and DD might have some similar pathophysiologi- 301 observations); this genotype was, however, observed cal mechanisms. ADH4 enzyme (π ADH) catalyzes syn- in unrelated cases (e.g. 4/370, for AD). We cannot com- thesis from substrates (which include, e.g., pletely exclude the possibility that the difference between norepinephrine aldehydes, including 3,4-dihydroxyman- cases and controls is attributable to sampling bias and the delaldehyde (DHMAL) and 4-hydroxy-3-methoxymande- findings are false positive, but we conjecture that it is laldehyde (HMMAL)) to create the intermediary glycols of more likely that this genotype is related to phenotype in norepinephrine metabolism, including 3,4-dihydroxy- AAs, although there is not enough information in this phenylglycol (DHPG) and 4-hydroxy-3-methoxyphenylg- observation alone to support such a conclusion. Because lycol (HMPG), respectively. This catalysis is considerably this genotype is so rare (1%), some association methods more efficient via this isozyme than for any of the class I Page 6 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 Figure 1 Pairwise LD between the candidate marker SNP8 and other seven flanking markers in unrelated African-American subjects Pairwise LD between the candidate marker SNP8 and other seven flanking markers in unrelated African- American subjects. [D' = 1.00 in the blank squares; the numbers inside the red squares are D' × 100%; the blue squares rep- 2 2 resent low r values; the white squares represent low D' and low r values. SNPs 1–7 were reported previously in Luo et al. (2006) and span from 3' to 5'.]. isozymes (α, β and γ ADHs); and class III ADH (χ ADH) aldehydes. To block the turnover of norepinephrine alde- does not have any detectable catalytic activity towards hydes, perhaps one might self-administer ethanol to com- these substrates at all [37]. Increased π ADH activity – e.g., pete DHPG and HMPG, because ethanol is an external through genetic variation, such as, potentially, the SNP8 competitor for internal DHPG and HMPG on π ADH [37]. G/G genotype – could lead to increasing levels of DHPG This mechanism could lead to AD. Cocaine, which par- and HMPG, and a very high turnover of norepinephrine tially functions as a norepinephrine re-uptake inhibitor, Page 7 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 can activate the noradrenergic system [38]. Plasma epine- For this particular marker, the allele frequency of the rare phrine and norepinephrine concentrations were signifi- allele is higher for EAs than for AAs (0.077 vs. 0.046, for cantly increased in response to cocaine injection [38]. the control subjects). EAs are therefore expected to have a Intravenous opioids stimulate norepinephrine and acetyl- substantially higher frequency of rare homozygotes than choline release in cerebrospinal fluid [39]. Therefore, self- AAs – 0.006 vs. 0.002, i.e. about three times as many. administration of drugs (cocaine and opiates) could ele- Therefore, we specifically considered possible European vate norepinephrine aldehydes too, which may lead to admixture in the four homozygous AA patients. We found DD. that the European ancestry proportions in these four AA subjects were less than 0.72%, indicating these observa- A family-based association study is immune from popula- tions of the rare homozygote are unlikely to be related to tion stratification effects. Thus, in the present study, sub- the genomewide European admixture in these AA sub- jects with different ethnicities, including EA, AA, jects. Hispanic, and others, and the affected and unaffected par- ents, were combined in the analysis, to increase the statis- HWD at SNP loci in the case sample could be an indicator tical power. Allowing for the possible population- of gene-phenotype association [7,9,35-42]. Cases are specificity of association, we also performed this analysis ascertained due to their "affected" status, so disease sus- separately within EAs and AAs. However, the family-based ceptibility genotypes or alleles should be present at high association studies revealed no significant association rates in the case sample, which might violate HWE. Fur- between this ADH4 variation and substance dependence ther, because cases are not randomly sampled from the (both in Samples 3 and 4). This is likely due to the limited general population where there is random mating and 2N statistical power, given the small sample size (Sample 3: alleles among N subjects are independent, HWE of dis- 115 affected offspring and 221 parents), in the context of ease-related marker loci in cases could be violated, and 2N the fact that the SNP8 variant has a rare minor allele (Sam- alleles could become dependent. Only when the marker ple 3: frequency ≤ 0.049 in AAs and ≤ 0.077 in EAs; Sam- has no LD with the disease locus, i.e., the marker genotype ple 4: frequency ≤ 0.020 in AAs and ≤ 0.053 in EAs;). frequency distributions are independent of the diagnosis, Additionally, only heterozygous parents yield TDT infor- can the case group and the control group have the same mation, which further limits the power for the family- genotype frequency distributions, with both in HWE. based association studies. Therefore, the HWD of SNP8 in AD and DD among AAs in the present study suggests an association between SNP8 In the present study, our case-control sample (820 cases; and both AD and DD. Usually, susceptibility loci are in 483 controls) has approximately five times the power of HWD in cases, but in HWE in controls [7], as observed for the family Sample 3 [10]. However, neither allelewise nor SNP8. This is because a much greater sample size is genotypewise case-control comparisons showed any sig- needed to detect HWD in controls than in cases [9]. If the nificant association between ADH4 variation and sub- predisposing effect of the disease susceptibility allele is stance dependence. The case-control design is strong enough and the sample size for controls is large theoretically vulnerable to population stratification that enough, this locus could also be in HWD in controls, but could result in false negative findings. We therefore used with an excess of the protective genotype, the opposite of the structured association (SA) method [15] to exclude the situation for cases [9]. SNP8 does not, apparently, population stratification and admixture effects on associ- have a strong enough effect on risk to distort HWE in con- ations. The results did not change substantively after con- trols; alternatively, the size of the control sample is not trolling for population stratification and admixture large enough to detect HWD in that sample. effects, i.e., both were negative. Similarly, despite taking into account the potential confounding effects of age and Additionally, substance dependence significantly sex via regression analysis, no association between this increases mortality [43-45], leading to age cohort-related variation and phenotypes was detected. Additionally, the dropout of the disease-associated genotypes or alleles low detected admixture degrees in EAs (1.6%) and AAs from the population (i.e., natural selection). Selection by (3.9%) (which may have appeared especially low, partic- mortality may violate an assumption for HWE and cause ularly in the AAs, because of lack of inclusion of an ances- altered distribution of genotype frequencies (i.e., HWD) tral African population) suggest that admixture effects [46]. This dropout makes the risk genotype or allele rarer, should not have substantially affected the analysis in this but the risk genotype or allele is still more common in study. It is possible that the negative findings from the cases than in controls, consistent with what we observed case-control association might, like the TDT analysis and for SNP8, and providing additional evidence that SNP8 the FBAT analysis, result from insufficient power. might be a disease-associated locus. Page 8 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 The magnitude of the HWD test statistic varies with the recessive allele (i.e., the disease-risk allele) can be distance between the marker loci and the disease locus; "masked" by the dominant allele (i.e., the non-disease- that is, deviations from HWE are greatest at trait suscepti- risk allele), which yields negative results in case-control bility loci, and can also be detected for benign polymor- frequency comparisons [under HWD, the two alleles are phisms that are in LD with the susceptibility locus dependent and affect each other]. From the formula of 2 2 [35,40]. In the present study, SNP8 is in HWD in cases, goodness-of-fit χ test for HWD, the χ value (HWD statis- which suggests that the risk locus for substance depend- tic) is proportional to the sample size (N) and the squared ence might be located in the SNP8-containing haplotype difference between observed genotype frequency and the block or be SNP8 itself. expected frequency (Δ ), and inversely proportional to the expected genotype frequencies. Thus, even when the Δ is The association evident from the HWD test was not small, one rare genotype frequency could generate a high detected by case-control frequency comparison. This is HWD statistic. If the expected count of this rare genotype because the HWD test as an association method is, some- is less than 5, we use an exact test; the exact p value is usu- times, much more powerful than case-control compari- ally consistent with that from the goodness-of-fit χ test, son [7], for which, the present study provides an extreme as seen in additional file 1. This is why HWD test is espe- example. One reason is that, from a statistical perspective, cially sensitive to a marker with one rare genotype. the HWD test in cases has one degree of freedom (df = 1), rather than df = 2 for case-control genotype frequency HWD might not be more powerful than LD method in comparisons. Another reason for greater power of the detecting gene-disease association when a trait-associated HWD test in the present study is the age difference marker acts via a multiplicative mode of inheritance, between cases and controls, from an epidemiological because HWD test would have very little power under this viewpoint, as discussed in our previous study [7]. In this disease model [35,40,41,47-49,51]. However, SNP8 study, the average age for controls was 29.7 years, about unlikely acts via a multiplicative model in the present 10 years younger than that for cases, 39.6 years. Many study (p ≤ 0.007, see additional file 1). A new method, the healthy controls, although presently unaffected, have not weighted average (WA) statistic test, has been reported to completely passed through the age of risk to manifest AD be even more powerful than the HWD test to detect asso- or DD. The healthy controls have a probability (≈ lifetime ciation between disease susceptibility and marker loci prevalence of disease; less than the cumulative prevalence under many genetic inheritance models, including the by the subject's age) to develop disease at some point in recessive, additive and multiplicative models [49]. How- the future, and this probability increases with the residual ever, application of this method is beyond the scope of prevalence, so that the case-control association design the present study. may be less powerful than a case-only study using the HWD test. That is, some associations that can be discov- The HWD test can not only detect gene-phenotype associ- ered by the use of a case-only study might not be detected ation, but can also reflect a genetic disease model, because using a case-control design. Meanwhile, because the drop- the direction of HWD statistics (Δ, F and J) varies with the out of disease-related genotypes or alleles increases with genetic model [9,35,36]. In the present study, the Δ, F and the age of cases (due to increasing mortality), an HWD J for SNP8 are positive, suggesting that SNP8 appears to test that reflects the dropout could be more sensitive to follow a recessive genetic disease model. detect this disease-related locus, especially when cases are much older than controls. We also identified the genetic disease model for SNP8 with the best fit to the genotypic proportions observed in The HWD signal of a marker locus decays more rapidly patients and controls using the Mathematica Notebooks with distance from a causative locus than the LD signal written by Wittke-Thompson et al. [9]. Consistent with [35]. The closer a marker is to the causative locus, the the above inference, the "best-fit" model for SNP8 is a greater the excess of power for the HWD test over the LD recessive model. This model-fitting method can not only test. Nielsen et al. [40] demonstrated that the HWD identify the genetic model, but can also tell us that other method was more powerful than the LD method under explanations for the observed HWD, including chance, certain conditions (recessive and additive models), which genotyping error, and/or violations of the requisite was also supported by many other studies [7,9,35,47-49]. assumptions of HWE, are less likely, if one "best-fit" Kocsis et al. [50] demonstrated that even in the absence of model can be identified [9]. However, it should also be significant differences in genotype frequency distribution noted that just because an observed HWD is consistent between cases and controls, associations can be detected with a "best-fit" genetic model does not completely guar- by HWD, as observed for SNP8 in present study. This is antee that errors, missing data patterns, or violations of particularly true for a trait-associated marker that acts via HWE assumptions do not generate or contribute to the a recessive mode of inheritance, because the effect of the observed HWD. Actually, HWD can be attributable to a Page 9 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 combination of factors [9]. But the following analyses fur- sentative of the general population [57], which supports ther support the interpretation that non-disease factors the interpretation that HWD in AA cases is probably not underlying HWD are less likely to be important in due to a sample size issue; however, considering the small explaining our data. We note, though, that it might take a number of homozygote observations that are critical in very large sample to fully support this conclusion. driving the finding, we cannot exclude this possibility. Second, inbreeding can cause HWD. Inbreeding is a type Signal intensity, background noise, and clustering proper- of positive assortative mating which is non-random. Most ties all play a role in the ability to assign genotypes cor- populations are geographically divided, and mating is rectly, and in determining the types of errors that occur local, so inbreeding could be common, but to varying [52]. Genotyping error is one of the greatest concerns for extents. During inbreeding, individuals are more likely to causing spurious HWD observation. First, DNA contami- mate with relatives than with non-related individuals. nation can result in the lack of one homozygote in the One common consequence of inbreeding is that the PCR product, which leads to a deficiency of observed number of heterozygotes decreases and the number of homozygotes [52]. This runs counter to our data and thus homozygotes increases [58], which leads to HWD. is not a possible explanation for the present study. Sec- Another common consequence of inbreeding is that the ond, incomplete digestion of PCR product (relevant only expression of deleterious recessive alleles in the popula- when using the RFLP technique, not the TaqMan tech- tion increases, which reduces average fitness and increases nique) or poor amplification of one of the alleles will lead mortality ("inbreeding depression") [59], which, as to heterozygous genotypes being read as homozygous described above, can also contribute to HWD. However, genotypes [53]. This kind of allele dropout can lead to an our case and control samples are ascertained as unrelated, excess of apparent homozygous genotype observations, and we have no evidence for the existence of overlapping which does fit our data and thus needs to be considered. generations, making inbreeding unlikely in the present Also, when genotypes are read, heterozygote genotypes study. Third, gene flow may result in HWD. Gene flow is could, theoretically, be more ambiguous, and therefore the result of migration. Immigrants carrying new alleles more likely to be scored as "missing," than homozygote into the population may change the genotype frequency genotypes. To detect possible genotyping error, for family distribution of that population with resulting HWD in data, we assessed the data for Mendelian consistencies by that generation. Contrary to selection and genetic drift, the program PEDCHECK [54], with no non-Mendelian gene flow eventually homogenizes allele frequencies inconsistencies detected. For all subjects, including family among populations. Although gene flow occurs in most and case-control subjects, we also replicated the geno- populations, its contribution to major shifts in allele fre- types (the most accurate way to estimate genotyping error quencies is usually negligible. The AA population has rates), so that all genotypes were matched. Missing geno- been in the US for an average of about five generations type data rate was not significantly different between cases and we do not have evidence of major immigration for the and controls. Additionally, controls were tested for HWE current AA generation, so gene flow resulting in HWD in and did not show the same direction of HWD statistics as our AA sample is unlikely. Fourth, mutation may increase cases. Together, the evidence suggests that genotyping the genetic variability due to genetic drift and might cause error as an explanation for the observed HWD is improb- HWD. But because change in allele frequencies induced able. by mutation is so small from one generation to the next, we can safely ignore mutation as a factor in HWD. Unless Violation of one of the other HWE assumptions (besides mutation rates are abnormally high, for which we have no selection of alleles by disease) can also cause HWD. First, evidence in the present data, the change in allele frequen- genetic drift can cause HWD. Genetic drift is the effect of cies is believed to be virtually nil. In conclusion, violation finite population size [55], such that the smaller the pop- of one of the above HWE assumptions causing HWD is ulation, the more noticeable the effects of drift. All popu- unlikely. However, the caution that these results are lations are finite and all genetic variation is subject to driven primarily by a small group of subjects, and that our genetic drift. In a finite population, allele frequencies fluc- conclusions would be different if just a few of them were tuate by chance randomly and the fluctuation leads to omitted or somehow changed diagnosis, bears repeating; deviations from HWE (in this context, this is "sampling this reliance on a small number of subjects requires us to error") [56]. If a population is small enough, the effects of be very tentative in our conclusions. drift may overwhelm the other forces described below, even selection. Our AA case sample is large enough such In addition to the factors that have been discussed, other that genetic drift at the disease susceptibility locus and the factors can also cause HWD. For example, population marker locus can reasonably be ignored. Additionally, in stratification and admixture can cause HWD, as demon- our AA control sample, which is smaller than that of cases, strated by Luo et al. [6]. However, we have demonstrated HWE was not violated. Further, our AA samples are repre- that the admixture degrees in our EA and AA samples are Page 10 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 relatively low, suggesting that this factor can be ignored as Limitations the sources of HWD in our case samples. In addition, the The limitations of the present study have been discussed cases and the controls are drawn from the same popula- above. In summary, in the present study, because the risk tions, but the controls are in HWE for SNP8, reducing the genotype is rare, some analytic approaches including the possibility that HWD in cases results from an effect of goodness-of-fit test might not have sufficient power in admixture. Nonrandom patterns of missing data may also detecting associations; how conservative the correction generate a relatively consistent pattern of HWD (e.g., dis- for multiple testing should be remains disputed; and the proportionate missing data in heterozygotes may lead to function of the risk marker remains unclear. Allowing for a consistent pattern of HWD, with an excess of homozy- these limitations, the possibility that the findings are false gotes, as observed for SNP8). This possibility is common positive cannot be completely excluded. Future replica- and inevitable and thus should not be ignored. However, tion studies with stricter design and improved sampling our cases and controls were genotyped using same geno- methods, and increased study power are warranted. typing systems, reducing the possibility that non-random patterns of missing data cause HWD only in cases but not Competing interests The authors declare that they have no competing interests. in controls. Further, if the cases or controls are old enough, the marker can be in HWD because a specific mortality-related allele drops out due to death associated Authors' contributions with advancing age [51]. However, the ages of our subjects XL and LZ designed and coordinated the study. XL and LZ range from 17 to 78 years, making this explanation drafted the manuscript. LZ and SW participated in data unlikely. An unrecognized polymorphism in primer management and statistical analysis. RFA participated in sequences used in PCR may also lead to HWD, with an sample collection and provided critical comments. JG and excess of homozygotes, as observed for SNP8, particularly HRK participated in the supervision of the project. All when the primer polymorphism is in LD with the tested authors read and approved the final manuscript. marker [9]. Finally, genomic duplications or deletions (a copy deletion could lead to hemizygosity) can also lead to Additional material HWD [9]. We believe that these explanations are not appropriate to explain our data in the present study, but Additional file 1 these factors can be excluded only through extensive Genotype and allele frequencies of SNP8 and p-values for genetic model sequence analysis. fitting. Click here for file In conclusion, the presence of HWD for SNP8 suggests [http://www.biomedcentral.com/content/supplementary/1744- 9081-4-42-S1.doc] that this polymorphism might be a risk locus for sub- stance dependence in AAs, although our direct evidence Additional file 2 for this conclusion is weak and the false positivity cannot Transmitted and non-transmitted allele frequencies in small nuclear fam- completely excluded. SNP8 is located at the putative 5' ilies. regulatory region of ADH4. It might indirectly modulate Click here for file risk for disease via LD with an unknown nearby func- [http://www.biomedcentral.com/content/supplementary/1744- tional variant, e.g., in ABI and HapMap database, it is 2.5 9081-4-42-S2.doc] kb far from and in LD with rs7434491 which could signif- icantly alter the secondary structure of ADH4 mRNA (IDT SciTools: http://www.idtdna.com/SciTools/Sci Tools.aspx); it might also alter the transcription initiation Acknowledgements Ann Marie Lacobelle and Greg Kay provided excellent technical assistance. site or the capacity of transcription factors to bind to the Thomas Hudak, Ellen Koucky, and Michelle Slivinsky assisted with patient DNA sequence, and consequently, directly affect tran- interviews. Dr. Frances Jurnak, University of California Irvine, provided a scription levels; it might result in mRNA instability, helpful suggestion regarding understanding ADH4 effects on norepine- altered translational efficiency, or even different protein phrine metabolism. This work was supported in part by National Institute expression levels in different tissues. Considering our on Drug Abuse (NIDA) grants R01-DA12849, R01-DA12690, and K24- sample size limitations, we believe that replication of DA15105, National Institute on Alcohol Abuse and Alcoholism (NIAAA) these results is critical. Nevertheless, given our findings, grants R01-AA11330, R01-AA016015, P50-AA12870, P50-AA03510, and we believe that it would be productive to study the effect K24-AA13736, National Center for Research Resources (NCRR) grant M01-RR06192, Alcoholic Beverage Medical Research Foundation (ABMRF) of this variation directly on protein expression, in order to grant award R06932 (X Luo); and by the U.S. Department of Veterans provide convergent validation of the findings reported Affairs (the VA Medical Research Program, VA Alcohol Research Center, here and to elucidate the specific mechanism underlying and the VA Connecticut-Massachusetts Mental Illness Research, Education the association of SNP8 at ADH4 to both AD and DD. and Clinical Center [MIRECC], and the VA Research Enhancement Award Program [REAP] research center). Page 11 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 25. Horvath S, Xu X, Laird NM: The family based association test References method: strategies for studying general genotype–pheno- 1. Ditlow CC, et al.: Physical and enzymatic properties of a class type associations. Eur J Hum Genet 2001, 9(4):301-6. II alcohol dehydrogenase isozyme of human liver: pi-ADH. 26. Liu K: PowerMarker: new genetic data analysis software. Biochemistry 1984, 23(26):6363-8. 2007. Free program distributed by the author over the internet from 2. Li TK, et al.: Isolation of pi-alcohol dehydrogenase of human http://www.powermarker.net liver: is it a determinant of alcoholism? Proc Natl Acad Sci USA 27. Stein MB, Schork NJ, Gelernter J: A polymorphism of the beta1- 1977, 74(10):4378-81. adrenergic receptor is associated with low extraversion. Biol 3. Guindalini C, et al.: Association of genetic variants in alcohol Psychiatry 2004, 56(4):217-24. dehydrogenase 4 with alcohol dependence in Brazilian 28. Yang BZ, et al.: Practical population group assignment with patients. Am J Psychiatry 2005, 162(5):1005-7. selected informative markers: characteristics and properties 4. Edenberg HJ, Jerome RE, Li M: Polymorphism of the human alco- of Bayesian clustering via STRUCTURE. Genet Epidemiol 2005, hol dehydrogenase 4 (ADH4) promoter affects gene expres- 28(4):302-12. sion. Pharmacogenetics 1999, 9(1):25-30. 29. Pritchard JK, et al.: Association mapping in structured popula- 5. Edenberg HJ, et al.: Association of alcohol dehydrogenase genes tions. Am J Hum Genet 2000, 67(1):170-81. with alcohol dependence: a comprehensive analysis. Hum Mol 30. Wienker T: FINETTI: Tests for deviation from Hardy-Wein- Genet 2006, 15(9):1539-49. berg equilibrium and tests for association in case-control 6. Luo X, et al.: ADH4 gene variation is associated with alcohol studies. 2002. Free program distributed by the author over the and drug dependence: results from family controlled and internet from http://ihg.gsf.de/cgi-bin/hw/hwa1.pl population-structured association studies. Pharmacogenet 31. Miller M: TFPGA: Tools For Population Genetic Analyses. Genomics 2005, 15(11):755-68. 2006. Free program distributed by the author over the internet from 7. Luo X, et al.: ADH4 gene variation is associated with alcohol http://www.marksgeneticsoftware.net dependence and drug dependence in European Americans: 32. Demyttenaere K, et al.: Prevalence, severity, and unmet need results from HWD tests and case-control association stud- for treatment of mental disorders in the World Health ies. Neuropsychopharmacology 2006, 31(5):1085-95. Organization World Mental Health Surveys. Jama 2004, 8. Luo X, et al.: Personality traits of agreeableness and extraver- 291(21):2581-90. sion are associated with ADH4 variation. Biol Psychiatry 2007, 33. Cramer H: Mathematical methods of statistics. Princeton, NJ: 61(5):599-608. Princeton University Press; 1946. 9. Wittke-Thompson JK, Pluzhnikov A, Cox NJ: Rational inferences 34. Weir B: Disequilibrium, in In: Genetic data analysis II: meth- about departures from Hardy-Weinberg equilibrium. Am J ods for discrete population genetic data. Sinaur Associates: Hum Genet 2005, 76(6):967-86. Sunderland, MA; 1996:91-139. 10. McGinnis R, Shifman S, Darvasi A: Power and efficiency of the 35. Feder JN, et al.: A novel MHC class I-like gene is mutated in TDT and case-control design for association scans. Behav patients with hereditary haemochromatosis. Nat Genet 1996, Genet 2002, 32(2):135-44. 13(4):399-408. 11. Parra EJ, et al.: Estimating African American admixture pro- 36. Jiang R, et al.: Fine-scale mapping using Hardy-Weinberg dise- portions by use of population-specific alleles. Am J Hum Genet quilibrium. Ann Hum Genet 2001, 65(Pt 2):207-19. 1998, 63(6):1839-51. 37. Mardh G, et al.: Human class II (pi) alcohol dehydrogenase has 12. Hoggart CJ, et al.: Control of confounding of genetic associa- a redox-specific function in norepinephrine metabolism. Proc tions in stratified populations. Am J Hum Genet 2003, Natl Acad Sci USA 1986, 83(23):8908-12. 72(6):1492-1504. 38. Sofuoglu M, et al.: Intravenous cocaine increases plasma epine- 13. Shriver MD, et al.: Skin pigmentation, biogeographical ancestry phrine and norepinephrine in humans. Pharmacol Biochem Behav and admixture mapping. Hum Genet 2003, 112(3):387-99. 2001, 68(3):455-9. 14. Collins-Schramm HE, et al.: Mexican American ancestry-inform- 39. Bouaziz H, et al.: Intravenous opioids stimulate norepinephrine ative markers: examination of population structure and and acetylcholine release in spinal cord dorsal horn. System- marker characteristics in European Americans, Mexican atic studies in sheep and an observation in a human. Anesthe- Americans, Amerindians and Asians. Hum Genet 2004, siology 1996, 84(1):143-54. 114(3):263-71. 40. Nielsen DM, Ehm MG, Weir BS: Detecting marker-disease asso- 15. Pritchard JK, Stephens M, Donnelly P: Inference of population ciation by testing for Hardy-Weinberg disequilibrium at a structure using multilocus genotype data. Genetics 2000, marker locus. Am J Hum Genet 1998, 63(5):1531-40. 155(2):945-59. 41. Lee WC: Searching for disease-susceptibility loci by testing 16. Gossop M: Multiple substance use and multiple dependencies, for Hardy-Weinberg disequilibrium in a gene bank of in Dual Diagnosis and Psychiatric Treatment. Edited by: Kran- affected individuals. Am J Epidemiol 2003, 158(5):397-400. zler H. Marcel Dekker: New York; 2004:129-156. 42. Hao K, et al.: Power estimation of multiple SNP association 17. Luo X, et al.: CHRM2 gene predisposes to alcohol dependence, test of case-control study and application. Genet Epidemiol drug dependence and affective disorders: results from an 2004, 26(1):22-30. extended case-control structured association study. Hum Mol 43. Shai D, Aptekar L: Factors in mortality by drug dependence Genet 2005, 14(16):2421-34. among Puerto Ricans in New York City. Am J Drug Alcohol Abuse 18. Gelernter J, et al.: Genomewide linkage scan for cocaine 1990, 16(1–2):97-107. dependence and related traits: significant linkages for a 44. Poser W, Poser S, Eva-Condemarin P: Mortality in patients with cocaine-related trait and cocaine-induced paranoia. Am J Med dependence on prescription drugs. Drug Alcohol Depend 1992, Genet B Neuropsychiatr Genet 2005, 136B(1):45-52. 30(1):49-57. 19. Gelernter J, et al.: Genomewide linkage scan for opioid depend- 45. Dawson DA: Alcohol consumption, alcohol dependence, and ence and related traits. Am J Hum Genet 2006, 78(5):759-69. all-cause mortality. Alcohol Clin Exp Res 2000, 24(1):72-81. 20. Pierucci-Lagha A, et al.: Diagnostic reliability of the Semi-struc- 46. Rundle AT, Atkin J, Sudell B: Serum and tissue proteins in tuber- tured Assessment for Drug Dependence and Alcoholism ous sclerosis. I. Serum and red-cell polymorphic systems. (SSADDA). Drug Alcohol Depend 2005, 80(3):303-12. Humangenetik 1975, 27(1):15-22. 21. American Psychiatric Association: Diagnostic and statistical man- 47. Deng HW, Chen WM, Recker RR: QTL fine mapping by measur- ual of mental disorders. revised ed. third edition. Washington, ing and testing for Hardy-Weinberg and linkage disequilib- DC: American Psychiatric Press; 1987. rium at a series of linked marker loci in extreme samples of 22. American Psychiatric Association: Diagnostic and statistical man- populations. Am J Hum Genet 2000, 66(3):1027-45. ual of mental disorders. fourth edition. Washington, DC: Ameri- 48. Song K, Elston RC: Tests for a disease-susceptibility locus can Psychiatric Press; 1994. allowing for an inbreeding coefficient (F). Genetica 2003, 23. Barrett JC, et al.: Haploview: analysis and visualization of LD 119(3):269-81. and haplotype maps. Bioinformatics 2005, 21(2):263-5. 49. Song K, Elston RC: A powerful method of combining measures 24. Dudbridge F: Pedigree disequilibrium tests for multilocus hap- of association and Hardy-Weinberg disequilibrium for fine- lotypes. Genet Epidemiol 2003, 25(2):115-21. mapping in case-control studies. Stat Med 2006, 25(1):105-26. Page 12 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 50. Kocsis I, et al.: Examination of Hardy-Weinberg equilibrium in papers of Kidney International: an underused tool. Kidney Int 2004, 65(5):1956-8. 51. Weinberg CR, Morris RW: Invited commentary: Testing for Hardy-Weinberg disequilibrium using a genome single- nucleotide polymorphism scan based on cases only. Am J Epi- demiol 2003, 158(5):401-3. discussion 404-5 52. Leal SM: Detection of genotyping errors and pseudo-SNPs via deviations from Hardy-Weinberg equilibrium. Genet Epidemiol 2005, 29(3):204-14. 53. Geller F, Ziegler A: Detection rates for genotyping errors in SNPs using the trio design. Hum Hered 2002, 54(3):111-7. 54. O'Connell JR, Weeks DE: PedCheck: a program for identifica- tion of genotype incompatibilities in linkage analysis. Am J Hum Genet 1998, 63(1):259-66. 55. Hedrick PW: Recombination and directional selection. Nature 1983, 302(5910):727. 56. Barton NH, Turelli M: Effects of genetic drift on variance com- ponents under a general model of epistasis. Evolution 2004, 58(10):2111-32. 57. Luo X, et al.: Response to Dr. Kopke's comments on haplo- types at the OPRM1 locus. Am J Med Genet B Neuropsychiatr Genet 2005, 135B(1):102. 58. Soltis D: Isozymes in plant biology. Dioscorides Press: Portland, Oregon; 1989:130. 59. Charlesworth B, Charlesworth D: The genetic basis of inbreed- ing depression. Genet Res 1999, 74(3):329-40. Publish with Bio Med Central and every scientist can read your work free of charge "BioMed Central will be the most significant development for disseminating the results of biomedical researc h in our lifetime." Sir Paul Nurse, Cancer Research UK Your research papers will be: available free of charge to the entire biomedical community peer reviewed and published immediately upon acceptance cited in PubMed and archived on PubMed Central yours — you keep the copyright BioMedcentral Submit your manuscript here: http://www.biomedcentral.com/info/publishing_adv.asp Page 13 of 13 (page number not for citation purposes) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Behavioral and Brain Functions Springer Journals

Recessive genetic mode of an ADH4 variant in substance dependence in African-Americans: A model of utility of the HWD test

Download PDF
13 pages

Loading next page...
 
/lp/springer-journals/recessive-genetic-mode-of-an-adh4-variant-in-substance-dependence-in-SOPPDdcArG

References (66)

Publisher
Springer Journals
Copyright
Copyright © 2008 by Luo et al; licensee BioMed Central Ltd.
Subject
Biomedicine; Neurosciences; Neurology; Behavioral Therapy; Psychiatry
eISSN
1744-9081
DOI
10.1186/1744-9081-4-42
pmid
18801187
Publisher site
See Article on Publisher Site

Abstract

Background: In our previous studies, we reported positive associations between seven ADH4 polymorphisms and substance dependence [i.e., alcohol dependence (AD) and/or drug dependence (DD)] in European-Americans (EAs). In the present study, we address the relationship between ADH4 variation and substance dependence in an African-American (AA) population, and report evidence that supports an association between a different ADH4 polymorphism (rs2226896) and these phenotypes in AAs. Methods: Two family-based association study methods, i.e., TDT and FBAT, were applied to test the relationship between ADH4 variation and substance dependence in Sample 3 (112 small nuclear families) and in Sample 4 (632 pedigrees), respectively. A population-based case-control association study method was also applied to test this relationship in 1303 unrelated subjects, with and without controlling for admixture effects. Finally, a Hardy-Weinberg Disequilibrium (HWD) test was applied to examine the association in the case-only sample, infer the genetic disease models, and distinguish the disease and non-disease factors contributing to HWD. Results: The marker examined was found to be in significant HWD in AA alcoholics (p = 0.0071) and drug dependent subjects (p = 0.0341), but in Hardy-Weinberg Equilibrium (HWE) in all other subgroups. Other association methods failed to detect any association between this variation and phenotypes. The best-fit genetic disease model for this marker is a recessive genetic model. Conclusion: ADH4 variation might play a role in risk for substance dependence in AAs, potentially via a recessive mechanism. Under certain conditions, the HWD test could be a more powerful association method than conventional family-based and population-based case-control association analyses, for which, the present study provides an extreme example. Page 1 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 We also demonstrated that ADH4 genotypes predispose Background The rate of alcohol metabolism influences drinking to AD and DD consistent with a recessive mode of inher- behavior and affects risk for alcohol dependence (AD). itance in EAs. These associations remained after control- The alcohol dehydrogenases (ADH) and the aldehyde ling for admixture effects and were confirmed by a dehydrogenases (ALDH) are the two major categories of family-based association study. However, in AAs, these alcohol-metabolizing enzymes in the liver. ADH converts seven ADH4 markers were in HWE in both cases and alcohol to acetaldehyde (which is toxic) and then ALDH controls; no association between alleles, genotypes, hap- converts acetaldehyde into acetate. Acetate is then oxi- lotypes, or diplotypes of these markers and AD and DD dized into CO and H O via the tricarboxylic acid cycle. were found in this population, even after controlling for 2 2 The development of AD is related to an individual's level admixture effects. of ethanol consumption, which is influenced by ADH and ALDH activity. The human ADH4 enzyme (i.e., π subunit) Although we did not find a significant association is an important member among these alcohol-metaboliz- between these seven ADH4 variants and AD or DD in AAs ing enzymes. It mainly contributes to liver ADH activity, in our previous studies, in view of the power variance of and at intoxicating levels of alcohol it may account for as association methods and the population-specificity of much as 40% of the total ethanol oxidation rate [1]. Thus, associations, we did not exclude the possibility that these physiologically significant variation in ADH4 activity associations might be significant if adequate statistical could clearly contribute to variability in risk for AD. power were available, or that other markers might be asso- ciated with these phenotypes in this population [6,7]. The The human ADH4 enzyme is encoded by the ADH4 gene, present study aimed to identify an association between a which maps to 4q22 within the ADH gene cluster. We specific polymorphism, rs2226896 (for convenience and focused on the ADH4 gene in this study because of the to distinguish it from the seven polymorphisms on which functional importance of its protein product, remarkable we reported previously, we refer to this variant as SNP8) for its activity at high ethanol concentrations that might and AD, in AAs. This polymorphism has a rare allele and be of particular relevance in the context of AD risk [2]; and a rare homozygote which could result in its being inde- because previous studies have demonstrated that ADH4 is pendent of the flanking haplotype block and could an important risk gene for AD and AD-related traits in increase the relative risk between genotypes so that it European-Americans (EAs) and African-Americans (AAs). could be a more powerful marker for some specific asso- Guindalini et al. [3] reported that the ADH4 promoter var- ciation analysis methods (e.g., the HWD test) to detect iants -75A/C (rs1800759), which can significantly alter marker-phenotype association [9]. To test the population- the expression of the ADH4 enzyme [4], and -159A/G, specificity of this association, we also tested for it in EAs, were significantly associated with AD in EAs and AAs in a the most common population in the US. Brazilian population. Edenberg et al. [5] reported that six- teen ADH4 polymorphisms (including rs2226896) were As different association methods have different advan- associated with AD in an independent Collaborative tages, disadvantages, and power associated with them, in Study on the Genetics of Alcoholism (COGA) sample of the present study, we applied several methods and com- pedigrees. We have previously reported associations pared their power. Family-based association study meth- between seven ADH4 polymorphisms (including -75A/C) ods are immune to population stratification and and AD and drug dependence (DD) in EAs in the US pop- admixture effects, and (as generally applied) are not ulation [6,7]. We also reported that three ADH4 polymor- informed by the affection status of parents. This method phisms (including -75A/C) were associated with has been used as a valid confirmatory method for a popu- personality traits in AD and DD subjects [8]. lation-based association study in our previous study [6] and was applied in the present study as well. However, In our previous studies [6,7], we genotyped seven mark- because the sample size of families is usually limited, and ers in some of the subjects included in the present study. families usually provide around two-thirds of the power These seven markers span the full length of ADH4 and provided by unrelated case-control samples of similar size are in one haplotype block both in EAs and AAs. The [10], we also performed a case-control association study. seven ADH4 markers showed deviation from Hardy- AAs are an admixed population, and EAs are admixed as Weinberg Equilibrium (HWE) (called Hardy-Weinberg well, although much less so [11-14]. Admixture effects Disequilibrium, HWD) in EA substance dependent sub- may result in spurious observed association between gene jects (including patients with AD and DD), but were in and disease. A population-based case-control association HWE in EA healthy controls. Significant differences in study is vulnerable to these effects. Therefore, the degree genotype and diplotype, but not in any allele or haplo- of admixture was also measured in the case-control sam- type, frequency distributions for all seven ADH4 markers ple, and admixture effects on the case-control association were found between cases and controls in EAs (adjusted analysis were controlled by a structured association (SA) global p = 0.0070, 0.0004 for AD and DD, respectively). method [15]. Page 2 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 Our previous study suggested that a case-only HWD test Methods was more powerful (with lower p values) than a case-con- 1. Subjects The clinical samples are listed in Tables 1 and 2, including trol association analysis under a recessive model [7]. The present study provided an additional application for this Sample 1 (907 unrelated case-control subjects), Sample 3 powerful association method. Meanwhile, we used an (112 small nuclear families (SNFs); each family includes HWD test to infer the genetic disease model, via a set of parents and 1 to 2 offspring) [both samples were used in novel software programs developed by Wittke-Thompson previous studies by Luo et al [6,7]], Sample 2 (271 newly et al. [9], which incorporate the HWD information from recruited unrelated AA cases and 125 unrelated AA con- both cases and controls. HWE is subject to a number of trols), and Sample 4 (1613 related subjects from 632 ped- assumptions regarding genotyping error, selection, igrees; each pedigree has 1 to 6 affected siblings with or genetic drift, inbreeding, gene flow, and mutation, and to without parents) [Sample 4 was previously described by other factors such as population stratification, admixture, Gelernter et al [18,19]]. Samples 1 and 2 included a total nonrandom patterns of missing data, and nonrandom of 1303 unrelated subjects (757 males and 546 females), allele dropout with increasing age. HWD may be caused with 391 "genetic" EA cases, 310 "genetic" EA healthy by the violation of any one of these assumptions; thus, we controls (European ancestry proportion > 0.5), 429 also distinguished the underlying causes for HWD by the "genetic" AA cases, and 173 "genetic" AA controls (African above methodology via application of the recently-devel- ancestry proportion > 0.5; see Methods). Newly-recruited oped programs and by reviewing almost all possibility for subjects (Sample 2) were diagnosed using the Semi-Struc- the causes of HWD in details. tured Assessment for Drug Dependence and Alcoholism (SSADDA) [18,20]. The healthy control subjects (in Sam- Drug dependence (DD), which mainly includes cocaine ples 1 and 2) were screened to exclude major Axis I mental dependence and opioid dependence in our sample, is one disorders, including alcohol or drug use disorders, psy- of the most common phenotypes comorbid with AD [16]. chotic disorders (including schizophrenia or schizophre- DD has many features in common with AD, including nia-like disorders), mood disorders, and major anxiety symptomatology, neuropsychological impairment, disorders. Sample 4 included pedigrees having affected hypothesized pathogenetic mechanisms, and response to probands with substance dependence, previously used for specific treatments (reviewed by Luo et al [7]). DD has genome-wide linkage studies [18,19]. Diagnosis was also been reported to share some susceptibility genes with made according to DSM-III-R or DSM-IV criteria [21,22]. AD (reviewed by Luo et al [6,7,17]). Our previous study demonstrated that DD and AD share ADH4 as a suscepti- Samples 1, 2 and 3 were recruited at the University of bility gene in EAs [6,7]. Thus, in this study, we investi- Connecticut Health Center, the VA Connecticut Health- gated the association not only in AD, but also in DD, to care System-West Haven Campus, or the Medical Univer- test the phenotypic specificity of the observed associa- sity of South Carolina. Sample 4 was recruited at four tions. sites: University of Connecticut Health Center (Farming- Table 1: Sample size for the unrelated sample and the related small nuclear families. Sample sizes %Male Age (years) EA AA Hispanic Others Sample 1 Healthy controls 39.9 28.2 ± 9.1 310 48 Cases 73.4 39.4 ± 9.2 391 158 AD 76.1 40.3 ± 9.2 326 101 DD 68.2 37.1 ± 8.1 204 145 Sample 2 Healthy controls 28.0 33.5 ± 12.6 125 Cases 64.9 40.0 ± 9.3 271 AD 65.1 40.0 ± 9.3 269 DD 64.8 42.0 ± 8.3 71 Sample 3 SNF size 92 12 6 2 Parents 49.8 65.9 ± 8.6 183 24 11 3 %Unaffected 45.5 66.8 ± 8.6 79.8 37.5 63.6 100.0 Offspring 54.8 39.2 ± 11.1 96 13 5 1 AD 58.9 37.3 ± 8.2 77 9 3 1 DD 55.2 38.0 ± 7.3 77 13 5 1 EA, European-American; AA, African-American; AD, alcohol dependence; DD, drug dependence; SNF, small nuclear families. Samples 1 and 2 are unrelated case-control samples; Sample 3 is related SNF sample. Page 3 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 Table 2: Structure and sizes of pedigrees (Sample 4) 125 unrelated AA controls (Sample 2), and the related pedigree subjects (Sample 4) were not included in this Total EAs AAs Others analysis because the other flanking seven ADH4 markers have not been genotyped in these subjects. Subjects 1613 766 833 14 %Male 46.3 51.4 41.2 71.4 2) Transmission disequilibrium test (TDT): Family-based Age (year) 39.4 ± 8.9 37.8 ± 10.2 40.8 ± 7.2 43.7 ± 5.7 Pedigrees 632 311 316 5 association analysis was performed by comparing "cases" With 1 AD 242 115 126 1 (i.e., transmitted allele and genotypes) and artificial "con- With 2 AD 166 87 78 1 trols" (i.e., untransmitted allele and genotypes) in Sample With 3 AD 11 5 6 0 3 by the program TDTPHASE [24]. All subjects in Sample With 4 AD 52 30 3, including EA, AA, Hispanic, and others, and the affected With ≥ 5 AD 10 10 and unaffected parents, were combined in the TDT analy- With 1 DD 57 25 32 0 sis. This analysis was also performed separately within EAs With 2 DD 460 243 216 1 With 3 DD 90 35 55 0 and AAs. With 4 DD 16 7 9 0 With ≥ 5 DD 30 30 3) Family-Based Association Tests (FBAT): The structure of Sample 4 is much more complicated than Sample 3, thus, EA, AA, AD, DD, same as Table 1. "With n AD", a pedigree having n we used the program FBAT [25] to test the gene-AD and offspring with AD; as analogy. gene-DD associations under the general, additive, domi- nant, and recessive genetic models respectively. We have ton, CT), Yale University School of Medicine (New Haven, already genotyped 419 short tandem repeats (STRs) in CT), McLean Hospital, Harvard Medical School (Belmont, this sample in the initial whole genome scan linkage stud- MA), and Medical University of South Carolina (Charles- ies [18,19], and inferred the ancestry proportions in each ton, SC). All subjects gave informed consent before partic- individual. The whole sample can thus be separated into ipating in the study, which was approved by the "genetic" EAs (European ancestry proportion > 0.5) and Institutional Review Board at the respective institutions. "genetic" AAs (African ancestry proportion > 0.5), accord- ing to the ancestry proportions in probands. FBAT was 2. Genotyping performed in the whole sample including EAs, AAs, His- A marker (rs2226896; SNP8; at Chr04: 100460117) at the panics, and others. This analysis was also performed sep- putative 5' regulatory region, close to the functional vari- arately within "genetic" EAs and AAs, because the ant -75A/C (at Chr04: 100458289; 1.8 Kb to 3' of SNP8) population-specificity of the gene-disease linkage and the at the promoter and other disease-related polymorphisms disease-risk sites has ever been demonstrated in this sam- at this region (e.g., -159G/A at Chr04: 100458373, 1.7 Kb ple [18,19]. to 3' of SNP8 [3]; rs1984362 at Chr04: 100463753, 3.6 Kb to 5' of SNP8 [6,7]) was genotyped in all subjects using 4) Case-control comparisons for allele and genotype fre- the TaqMan technique ("assay-on-demand"). These eight quencies: Allele and genotype frequencies for SNP8 ADH4 SNPs were all of those that were available from among EAs and AAs in Samples 1 and 2 are shown in public sources when we started genotyping. SNP8 was not additional file 1. Associations between alleles, genotypes reported together with other seven SNPs in our previous and phenotypes were tested by comparing the allele and studies [6,7] because it was not associated with any phe- genotype frequency distributions between cases and con- notype in EAs. The detailed genotyping procedure is trols with the exact tests in the program PowerMarker described elsewhere [7]. All genotyping was performed in [26]. duplicate and compared to ensure validity of the data. Mismatched genotypes, if any, were discarded. Thirty- 5) Structured association (SA) analysis: EAs and AAs were eight unlinked ancestry-informative markers (AIMs) were taken as admixed populations with different degrees of also genotyped in unrelated subjects to estimate the admixture. The extent of admixture (i.e., average ancestry ancestry proportions for each subject (detailed by Luo et proportions) in unrelated subjects was estimated using al [6]). the program STRUCTURE [15] by analyzing 38 AIMs (in combined EA and AA subjects) [6,27,28]. Admixture 3. Statistical analysis effects on case-control association analysis can be control- 1) Linkage disequilibrium (LD) analysis: Pairwise LD led using the program STRAT [29] by conditioning the between this marker and the other seven ADH4 markers analysis on the ancestry proportions, i.e., structured asso- studied previously was analyzed separately for EAs and ciation (SA) analysis (separately for EAs and AAs). Fur- AAs in Sample 1 via the program Haploview [23]. The thermore, associations between this ADH4 variation and additional newly recruited 271 AA unrelated cases and phenotypes were also analyzed by a regression method Page 4 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 (in combined EAs and AAs), with ancestry proportions, p-value. If this p-value is less than 0.05, the genetic disease age, and sex serving as covariates. model can be rejected as a poor fit to the observed data. The best-fit genetic model would correspond to a p-value 6) Hardy-Weinberg equilibrium (HWE) test and genetic which is larger than 0.05 and is the largest among all disease model inference: An HWE test can be a valid asso- genetic models. If all of the genetic models are poorly fit- ciation method, and can be used to infer genetic disease ted, then alternative explanations for the HWD, including models. HWE was tested within populations, and sepa- chance, genotyping error, and/or violations of the requi- rately for cases and controls, using the goodness-of-fit χ test implemented in the program PowerMarker [26]. If site assumptions of HWE, must be considered. there was one cell with expected genotype count less than The direction of HWD statistics (Δ, F and J values, detailed 5, HWE was also tested by the exact test implemented in the program PowerMarker [26] and confirmed by the definitions follow) in cases is also helpful to judge the genetic models. Under the dominant and the additive exact test implemented in the online program FINETTI models, the values of Δ, F and J are negative, because [30], and the exact tests implemented in the program HWD is caused by an excess of observed heterozygote TFPGA [31] which applies conventional Monte Carlo and Markov Chain Monte Carlo (MCMC) methods. (Aa) and homozygote (AA) frequencies over the expected frequencies. Under the recessive model, the values of Δ, F and J are positive, because the HWD is caused by excess of To distinguish the causes of HWD, which include gene- observed homozygote (aa) frequency over the expected disease association and non gene-disease association fac- frequency, and in contrast, observed heterozygote (Aa) tors such as genetic drift, migration, mutation, non-ran- frequency is less than expected. These statistics were calcu- dom mating, etc., and then to infer the genetic disease lated based on the following formula: Assuming that fo', model for the marker in HWD, we used the goodness-of- fo", fo"', and fe', fe", fe"' denote the observed and expected fit χ test defined by Wittke-Thompson et al. [9], which frequencies of susceptibility homozygote, heterozygote and non-susceptibility homozygote in cases, respectively, incorporated the HWE information both from cases and then Δ = fo'''-fe''', F = 2Δ/fe'' and J = Δ/fe"' [34-36]. controls (here denoted WT goodness-of-fit χ test) [9]. The O − E O − E () j j Results 2 () i i test statistic c=+ (i, j = 1, 2 ∑∑ E E 1. SNP8 was in LD with the seven flanking markers in EAs i j ij (D' > 0.9 in each case), but provided information inde- or 3), where O and E are the observed and expected num- i i pendent of these markers in AAs (D' < 0.10 in each case) th bers of i genotype of this marker in patients, respectively; (Figure 1). TDT in Sample 3 and FBAT in Sample 4 th O and E are the observed and expected numbers of j gen- showed no significant association between the ADH4 j j SNP8 and AD or DD, whether analysis is conducted by otype of this marker in controls, respectively. Assuming combining or separating the different ethnicities (all p > different baseline penetrance of disease (from 0 to 1) for 0.05) (Sample 3: see additional file 2; Sample 4: data not non-susceptibility homozygotes, different heterozygote shown). The genotype frequency distributions in Sample relative risk, and different susceptibility homozygote rela- 4 were shown in Table 3, not for the association analysis, tive risk that determine different E and E given different i j but for providing a replication of the rare genotype fre- genetic models, and setting the lifetime population prev- quencies to those in the unrelated Samples 1 and 2 (see additional file 1). alence of alcohol dependence and drug dependence at 5.5% and 3.8% [32], respectively, we can obtain a differ- 2. SNP8 was not associated with phenotypes via case-con- 2 2 ent set of χ values. Minimizing this χ statistic over the trol comparisons either before or after controlling for entire parameter space with the appropriate constraints admixture effects. Measured admixture degrees were low on the parameters approximates a maximum-likelihood both in EAs and AAs. estimate procedure. We can then obtain the minimal (χ ) 2 No significant difference in allele or genotype frequency value, which is approximately distributed as a χ with 1 df distribution for this marker was found between unrelated for a general model and 2 df for restricted models (i.e., cases and controls in either EAs or AAs (all p > 0.05; addi- dominant, recessive, additive, and multiplicative models) tional file 1). After controlling for admixture effects by the [33]. Simulating 173 AA controls and 370 AA AD subjects SA method, these negative associations were essentially or 216 AA DD subjects for 1000 replications to obtain unchanged (all p > 0.05). Regression analysis, which takes 1000 minimal χ values and comparing these values to into account the effects of age and sex, also showed no sig- nificant association. the original minimal (χ ) value, we can derive an empiric Page 5 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 Table 3: Genotype frequency distributions of SNP8 in pedigrees (Sample 4) European-American African-American AD DD Unaffected AD DD Unaffected Nf N f NfNf N f Nf A/A 271 0.903 541 0.905 85 0.895 293 0.945 626 0.963 127 0.992 A/G 26 0.087 53 0.089 8 0.084 16 0.052 24 0.037 1 0.008 G/G 3 0.010 4 0.007 2 0.021 1 0.003 0 0.000 0 0.000 A 568 0.947 1135 0.949 178 0.937 602 0.971 1276 0.982 255 0.996 G 32 0.053 61 0.051 12 0.063 18 0.029 24 0.018 1 0.004 AD, alcohol dependence; DD, drug dependence. Two main ancestries, i.e., European and African, were with insufficient power might not detect a true genotype- detected in our sample. In the 701 "genetic" EA subjects phenotype association (in fact, evaluated by exact test as a (European ancestry proportion > 0.5), the total estimated case-control genotype frequency distribution comparison, weight of African ancestry proportions is 11.3; the admix- this comparison shows p = 0.391 and 0.225 for AD vs. ture degree of "genetic" EAs in this sample is 1.6% (= controls and DD vs. controls, respectively). Thus, in the 11.3/701). In the 602 "genetic" AA subjects (African present study, several association methods with differing ancestry proportion > 0.5), the total estimated weight of power and that take different views of the data were European ancestry proportions is 23.7; the admixture applied and compared. The most powerful HWD test sug- degree of "genetic" AAs is 3.9% (= 23.7/602). gests that the ADH4 variant (rs2226896) might play an important role in risk for substance dependence (includ- 3. SNP8 was in significant HWD in both AA AD and DD ing AD and DD) in AAs, probably via a recessive genetic subjects (with an excess of observed homozygotes over mechanism. The association between this variant and expected homozygotes): using the exact tests: p = 0.0071 phenotype is population-specific, that is, it appears in for AD and p = 0.0341 for DD, respectively. These HWDs AAs, but not in EAs. This association herein first discov- remained significant or suggestive after correction for ered in AAs is a complementary finding to a previous set multiple testing (α was set at 0.025). However, SNP8 was of genotype-phenotype relations we described for other in HWE in EA AD and DD subjects (p > 0.05) (additional markers at this locus in EAs [6,7]; based on this result, we file 1). This marker was in HWE in controls, both EA and can now provisionally conclude that ADH4 affects SD risk AA (p > 0.10). The WT goodness-of-fit χ test showed that in both EAs and AAs, but different variants are important the best-fit genetic disease model for this marker was a in the different populations. It would be of great interest recessive model (in AA AD subjects: χ = 0.621, df = 2, p = to study this variant in other populations, e.g., Asians, to 0.733; in AA DD subjects: χ = 1.596, df = 2, p = 0.450; see further characterize the population specificity we report additional file 1). In AA AD subjects for this marker, Δ = here. This variant is independent of the other seven poly- +0.008, F = +0.182 and J = +0.009; in AAs with drug morphisms that were reported previously to have no asso- dependence, Δ = +0.008, F = +0.204 and J = +0.010. ciation with substance dependence in AAs [6,7]. ADH4 gene variation is thought to influence the risk for Discussion In African-Americans (AAs), the SNP8 G/G genotype was AD by modulating ethanol metabolism. However, we find never observed in any of the control subjects in Sample 1 that it is associated with DD too. This is reasonable (n = 48), the newly-recruited unrelated controls in Sample because DD has many features in common with AD 2 (n = 125), or the additional related unaffected pedigree which are reviewed above and because the development subjects (n = 128; see Table 3) (i.e., in total there were 0/ of AD and DD might have some similar pathophysiologi- 301 observations); this genotype was, however, observed cal mechanisms. ADH4 enzyme (π ADH) catalyzes syn- in unrelated cases (e.g. 4/370, for AD). We cannot com- thesis from substrates (which include, e.g., pletely exclude the possibility that the difference between norepinephrine aldehydes, including 3,4-dihydroxyman- cases and controls is attributable to sampling bias and the delaldehyde (DHMAL) and 4-hydroxy-3-methoxymande- findings are false positive, but we conjecture that it is laldehyde (HMMAL)) to create the intermediary glycols of more likely that this genotype is related to phenotype in norepinephrine metabolism, including 3,4-dihydroxy- AAs, although there is not enough information in this phenylglycol (DHPG) and 4-hydroxy-3-methoxyphenylg- observation alone to support such a conclusion. Because lycol (HMPG), respectively. This catalysis is considerably this genotype is so rare (1%), some association methods more efficient via this isozyme than for any of the class I Page 6 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 Figure 1 Pairwise LD between the candidate marker SNP8 and other seven flanking markers in unrelated African-American subjects Pairwise LD between the candidate marker SNP8 and other seven flanking markers in unrelated African- American subjects. [D' = 1.00 in the blank squares; the numbers inside the red squares are D' × 100%; the blue squares rep- 2 2 resent low r values; the white squares represent low D' and low r values. SNPs 1–7 were reported previously in Luo et al. (2006) and span from 3' to 5'.]. isozymes (α, β and γ ADHs); and class III ADH (χ ADH) aldehydes. To block the turnover of norepinephrine alde- does not have any detectable catalytic activity towards hydes, perhaps one might self-administer ethanol to com- these substrates at all [37]. Increased π ADH activity – e.g., pete DHPG and HMPG, because ethanol is an external through genetic variation, such as, potentially, the SNP8 competitor for internal DHPG and HMPG on π ADH [37]. G/G genotype – could lead to increasing levels of DHPG This mechanism could lead to AD. Cocaine, which par- and HMPG, and a very high turnover of norepinephrine tially functions as a norepinephrine re-uptake inhibitor, Page 7 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 can activate the noradrenergic system [38]. Plasma epine- For this particular marker, the allele frequency of the rare phrine and norepinephrine concentrations were signifi- allele is higher for EAs than for AAs (0.077 vs. 0.046, for cantly increased in response to cocaine injection [38]. the control subjects). EAs are therefore expected to have a Intravenous opioids stimulate norepinephrine and acetyl- substantially higher frequency of rare homozygotes than choline release in cerebrospinal fluid [39]. Therefore, self- AAs – 0.006 vs. 0.002, i.e. about three times as many. administration of drugs (cocaine and opiates) could ele- Therefore, we specifically considered possible European vate norepinephrine aldehydes too, which may lead to admixture in the four homozygous AA patients. We found DD. that the European ancestry proportions in these four AA subjects were less than 0.72%, indicating these observa- A family-based association study is immune from popula- tions of the rare homozygote are unlikely to be related to tion stratification effects. Thus, in the present study, sub- the genomewide European admixture in these AA sub- jects with different ethnicities, including EA, AA, jects. Hispanic, and others, and the affected and unaffected par- ents, were combined in the analysis, to increase the statis- HWD at SNP loci in the case sample could be an indicator tical power. Allowing for the possible population- of gene-phenotype association [7,9,35-42]. Cases are specificity of association, we also performed this analysis ascertained due to their "affected" status, so disease sus- separately within EAs and AAs. However, the family-based ceptibility genotypes or alleles should be present at high association studies revealed no significant association rates in the case sample, which might violate HWE. Fur- between this ADH4 variation and substance dependence ther, because cases are not randomly sampled from the (both in Samples 3 and 4). This is likely due to the limited general population where there is random mating and 2N statistical power, given the small sample size (Sample 3: alleles among N subjects are independent, HWE of dis- 115 affected offspring and 221 parents), in the context of ease-related marker loci in cases could be violated, and 2N the fact that the SNP8 variant has a rare minor allele (Sam- alleles could become dependent. Only when the marker ple 3: frequency ≤ 0.049 in AAs and ≤ 0.077 in EAs; Sam- has no LD with the disease locus, i.e., the marker genotype ple 4: frequency ≤ 0.020 in AAs and ≤ 0.053 in EAs;). frequency distributions are independent of the diagnosis, Additionally, only heterozygous parents yield TDT infor- can the case group and the control group have the same mation, which further limits the power for the family- genotype frequency distributions, with both in HWE. based association studies. Therefore, the HWD of SNP8 in AD and DD among AAs in the present study suggests an association between SNP8 In the present study, our case-control sample (820 cases; and both AD and DD. Usually, susceptibility loci are in 483 controls) has approximately five times the power of HWD in cases, but in HWE in controls [7], as observed for the family Sample 3 [10]. However, neither allelewise nor SNP8. This is because a much greater sample size is genotypewise case-control comparisons showed any sig- needed to detect HWD in controls than in cases [9]. If the nificant association between ADH4 variation and sub- predisposing effect of the disease susceptibility allele is stance dependence. The case-control design is strong enough and the sample size for controls is large theoretically vulnerable to population stratification that enough, this locus could also be in HWD in controls, but could result in false negative findings. We therefore used with an excess of the protective genotype, the opposite of the structured association (SA) method [15] to exclude the situation for cases [9]. SNP8 does not, apparently, population stratification and admixture effects on associ- have a strong enough effect on risk to distort HWE in con- ations. The results did not change substantively after con- trols; alternatively, the size of the control sample is not trolling for population stratification and admixture large enough to detect HWD in that sample. effects, i.e., both were negative. Similarly, despite taking into account the potential confounding effects of age and Additionally, substance dependence significantly sex via regression analysis, no association between this increases mortality [43-45], leading to age cohort-related variation and phenotypes was detected. Additionally, the dropout of the disease-associated genotypes or alleles low detected admixture degrees in EAs (1.6%) and AAs from the population (i.e., natural selection). Selection by (3.9%) (which may have appeared especially low, partic- mortality may violate an assumption for HWE and cause ularly in the AAs, because of lack of inclusion of an ances- altered distribution of genotype frequencies (i.e., HWD) tral African population) suggest that admixture effects [46]. This dropout makes the risk genotype or allele rarer, should not have substantially affected the analysis in this but the risk genotype or allele is still more common in study. It is possible that the negative findings from the cases than in controls, consistent with what we observed case-control association might, like the TDT analysis and for SNP8, and providing additional evidence that SNP8 the FBAT analysis, result from insufficient power. might be a disease-associated locus. Page 8 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 The magnitude of the HWD test statistic varies with the recessive allele (i.e., the disease-risk allele) can be distance between the marker loci and the disease locus; "masked" by the dominant allele (i.e., the non-disease- that is, deviations from HWE are greatest at trait suscepti- risk allele), which yields negative results in case-control bility loci, and can also be detected for benign polymor- frequency comparisons [under HWD, the two alleles are phisms that are in LD with the susceptibility locus dependent and affect each other]. From the formula of 2 2 [35,40]. In the present study, SNP8 is in HWD in cases, goodness-of-fit χ test for HWD, the χ value (HWD statis- which suggests that the risk locus for substance depend- tic) is proportional to the sample size (N) and the squared ence might be located in the SNP8-containing haplotype difference between observed genotype frequency and the block or be SNP8 itself. expected frequency (Δ ), and inversely proportional to the expected genotype frequencies. Thus, even when the Δ is The association evident from the HWD test was not small, one rare genotype frequency could generate a high detected by case-control frequency comparison. This is HWD statistic. If the expected count of this rare genotype because the HWD test as an association method is, some- is less than 5, we use an exact test; the exact p value is usu- times, much more powerful than case-control compari- ally consistent with that from the goodness-of-fit χ test, son [7], for which, the present study provides an extreme as seen in additional file 1. This is why HWD test is espe- example. One reason is that, from a statistical perspective, cially sensitive to a marker with one rare genotype. the HWD test in cases has one degree of freedom (df = 1), rather than df = 2 for case-control genotype frequency HWD might not be more powerful than LD method in comparisons. Another reason for greater power of the detecting gene-disease association when a trait-associated HWD test in the present study is the age difference marker acts via a multiplicative mode of inheritance, between cases and controls, from an epidemiological because HWD test would have very little power under this viewpoint, as discussed in our previous study [7]. In this disease model [35,40,41,47-49,51]. However, SNP8 study, the average age for controls was 29.7 years, about unlikely acts via a multiplicative model in the present 10 years younger than that for cases, 39.6 years. Many study (p ≤ 0.007, see additional file 1). A new method, the healthy controls, although presently unaffected, have not weighted average (WA) statistic test, has been reported to completely passed through the age of risk to manifest AD be even more powerful than the HWD test to detect asso- or DD. The healthy controls have a probability (≈ lifetime ciation between disease susceptibility and marker loci prevalence of disease; less than the cumulative prevalence under many genetic inheritance models, including the by the subject's age) to develop disease at some point in recessive, additive and multiplicative models [49]. How- the future, and this probability increases with the residual ever, application of this method is beyond the scope of prevalence, so that the case-control association design the present study. may be less powerful than a case-only study using the HWD test. That is, some associations that can be discov- The HWD test can not only detect gene-phenotype associ- ered by the use of a case-only study might not be detected ation, but can also reflect a genetic disease model, because using a case-control design. Meanwhile, because the drop- the direction of HWD statistics (Δ, F and J) varies with the out of disease-related genotypes or alleles increases with genetic model [9,35,36]. In the present study, the Δ, F and the age of cases (due to increasing mortality), an HWD J for SNP8 are positive, suggesting that SNP8 appears to test that reflects the dropout could be more sensitive to follow a recessive genetic disease model. detect this disease-related locus, especially when cases are much older than controls. We also identified the genetic disease model for SNP8 with the best fit to the genotypic proportions observed in The HWD signal of a marker locus decays more rapidly patients and controls using the Mathematica Notebooks with distance from a causative locus than the LD signal written by Wittke-Thompson et al. [9]. Consistent with [35]. The closer a marker is to the causative locus, the the above inference, the "best-fit" model for SNP8 is a greater the excess of power for the HWD test over the LD recessive model. This model-fitting method can not only test. Nielsen et al. [40] demonstrated that the HWD identify the genetic model, but can also tell us that other method was more powerful than the LD method under explanations for the observed HWD, including chance, certain conditions (recessive and additive models), which genotyping error, and/or violations of the requisite was also supported by many other studies [7,9,35,47-49]. assumptions of HWE, are less likely, if one "best-fit" Kocsis et al. [50] demonstrated that even in the absence of model can be identified [9]. However, it should also be significant differences in genotype frequency distribution noted that just because an observed HWD is consistent between cases and controls, associations can be detected with a "best-fit" genetic model does not completely guar- by HWD, as observed for SNP8 in present study. This is antee that errors, missing data patterns, or violations of particularly true for a trait-associated marker that acts via HWE assumptions do not generate or contribute to the a recessive mode of inheritance, because the effect of the observed HWD. Actually, HWD can be attributable to a Page 9 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 combination of factors [9]. But the following analyses fur- sentative of the general population [57], which supports ther support the interpretation that non-disease factors the interpretation that HWD in AA cases is probably not underlying HWD are less likely to be important in due to a sample size issue; however, considering the small explaining our data. We note, though, that it might take a number of homozygote observations that are critical in very large sample to fully support this conclusion. driving the finding, we cannot exclude this possibility. Second, inbreeding can cause HWD. Inbreeding is a type Signal intensity, background noise, and clustering proper- of positive assortative mating which is non-random. Most ties all play a role in the ability to assign genotypes cor- populations are geographically divided, and mating is rectly, and in determining the types of errors that occur local, so inbreeding could be common, but to varying [52]. Genotyping error is one of the greatest concerns for extents. During inbreeding, individuals are more likely to causing spurious HWD observation. First, DNA contami- mate with relatives than with non-related individuals. nation can result in the lack of one homozygote in the One common consequence of inbreeding is that the PCR product, which leads to a deficiency of observed number of heterozygotes decreases and the number of homozygotes [52]. This runs counter to our data and thus homozygotes increases [58], which leads to HWD. is not a possible explanation for the present study. Sec- Another common consequence of inbreeding is that the ond, incomplete digestion of PCR product (relevant only expression of deleterious recessive alleles in the popula- when using the RFLP technique, not the TaqMan tech- tion increases, which reduces average fitness and increases nique) or poor amplification of one of the alleles will lead mortality ("inbreeding depression") [59], which, as to heterozygous genotypes being read as homozygous described above, can also contribute to HWD. However, genotypes [53]. This kind of allele dropout can lead to an our case and control samples are ascertained as unrelated, excess of apparent homozygous genotype observations, and we have no evidence for the existence of overlapping which does fit our data and thus needs to be considered. generations, making inbreeding unlikely in the present Also, when genotypes are read, heterozygote genotypes study. Third, gene flow may result in HWD. Gene flow is could, theoretically, be more ambiguous, and therefore the result of migration. Immigrants carrying new alleles more likely to be scored as "missing," than homozygote into the population may change the genotype frequency genotypes. To detect possible genotyping error, for family distribution of that population with resulting HWD in data, we assessed the data for Mendelian consistencies by that generation. Contrary to selection and genetic drift, the program PEDCHECK [54], with no non-Mendelian gene flow eventually homogenizes allele frequencies inconsistencies detected. For all subjects, including family among populations. Although gene flow occurs in most and case-control subjects, we also replicated the geno- populations, its contribution to major shifts in allele fre- types (the most accurate way to estimate genotyping error quencies is usually negligible. The AA population has rates), so that all genotypes were matched. Missing geno- been in the US for an average of about five generations type data rate was not significantly different between cases and we do not have evidence of major immigration for the and controls. Additionally, controls were tested for HWE current AA generation, so gene flow resulting in HWD in and did not show the same direction of HWD statistics as our AA sample is unlikely. Fourth, mutation may increase cases. Together, the evidence suggests that genotyping the genetic variability due to genetic drift and might cause error as an explanation for the observed HWD is improb- HWD. But because change in allele frequencies induced able. by mutation is so small from one generation to the next, we can safely ignore mutation as a factor in HWD. Unless Violation of one of the other HWE assumptions (besides mutation rates are abnormally high, for which we have no selection of alleles by disease) can also cause HWD. First, evidence in the present data, the change in allele frequen- genetic drift can cause HWD. Genetic drift is the effect of cies is believed to be virtually nil. In conclusion, violation finite population size [55], such that the smaller the pop- of one of the above HWE assumptions causing HWD is ulation, the more noticeable the effects of drift. All popu- unlikely. However, the caution that these results are lations are finite and all genetic variation is subject to driven primarily by a small group of subjects, and that our genetic drift. In a finite population, allele frequencies fluc- conclusions would be different if just a few of them were tuate by chance randomly and the fluctuation leads to omitted or somehow changed diagnosis, bears repeating; deviations from HWE (in this context, this is "sampling this reliance on a small number of subjects requires us to error") [56]. If a population is small enough, the effects of be very tentative in our conclusions. drift may overwhelm the other forces described below, even selection. Our AA case sample is large enough such In addition to the factors that have been discussed, other that genetic drift at the disease susceptibility locus and the factors can also cause HWD. For example, population marker locus can reasonably be ignored. Additionally, in stratification and admixture can cause HWD, as demon- our AA control sample, which is smaller than that of cases, strated by Luo et al. [6]. However, we have demonstrated HWE was not violated. Further, our AA samples are repre- that the admixture degrees in our EA and AA samples are Page 10 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 relatively low, suggesting that this factor can be ignored as Limitations the sources of HWD in our case samples. In addition, the The limitations of the present study have been discussed cases and the controls are drawn from the same popula- above. In summary, in the present study, because the risk tions, but the controls are in HWE for SNP8, reducing the genotype is rare, some analytic approaches including the possibility that HWD in cases results from an effect of goodness-of-fit test might not have sufficient power in admixture. Nonrandom patterns of missing data may also detecting associations; how conservative the correction generate a relatively consistent pattern of HWD (e.g., dis- for multiple testing should be remains disputed; and the proportionate missing data in heterozygotes may lead to function of the risk marker remains unclear. Allowing for a consistent pattern of HWD, with an excess of homozy- these limitations, the possibility that the findings are false gotes, as observed for SNP8). This possibility is common positive cannot be completely excluded. Future replica- and inevitable and thus should not be ignored. However, tion studies with stricter design and improved sampling our cases and controls were genotyped using same geno- methods, and increased study power are warranted. typing systems, reducing the possibility that non-random patterns of missing data cause HWD only in cases but not Competing interests The authors declare that they have no competing interests. in controls. Further, if the cases or controls are old enough, the marker can be in HWD because a specific mortality-related allele drops out due to death associated Authors' contributions with advancing age [51]. However, the ages of our subjects XL and LZ designed and coordinated the study. XL and LZ range from 17 to 78 years, making this explanation drafted the manuscript. LZ and SW participated in data unlikely. An unrecognized polymorphism in primer management and statistical analysis. RFA participated in sequences used in PCR may also lead to HWD, with an sample collection and provided critical comments. JG and excess of homozygotes, as observed for SNP8, particularly HRK participated in the supervision of the project. All when the primer polymorphism is in LD with the tested authors read and approved the final manuscript. marker [9]. Finally, genomic duplications or deletions (a copy deletion could lead to hemizygosity) can also lead to Additional material HWD [9]. We believe that these explanations are not appropriate to explain our data in the present study, but Additional file 1 these factors can be excluded only through extensive Genotype and allele frequencies of SNP8 and p-values for genetic model sequence analysis. fitting. Click here for file In conclusion, the presence of HWD for SNP8 suggests [http://www.biomedcentral.com/content/supplementary/1744- 9081-4-42-S1.doc] that this polymorphism might be a risk locus for sub- stance dependence in AAs, although our direct evidence Additional file 2 for this conclusion is weak and the false positivity cannot Transmitted and non-transmitted allele frequencies in small nuclear fam- completely excluded. SNP8 is located at the putative 5' ilies. regulatory region of ADH4. It might indirectly modulate Click here for file risk for disease via LD with an unknown nearby func- [http://www.biomedcentral.com/content/supplementary/1744- tional variant, e.g., in ABI and HapMap database, it is 2.5 9081-4-42-S2.doc] kb far from and in LD with rs7434491 which could signif- icantly alter the secondary structure of ADH4 mRNA (IDT SciTools: http://www.idtdna.com/SciTools/Sci Tools.aspx); it might also alter the transcription initiation Acknowledgements Ann Marie Lacobelle and Greg Kay provided excellent technical assistance. site or the capacity of transcription factors to bind to the Thomas Hudak, Ellen Koucky, and Michelle Slivinsky assisted with patient DNA sequence, and consequently, directly affect tran- interviews. Dr. Frances Jurnak, University of California Irvine, provided a scription levels; it might result in mRNA instability, helpful suggestion regarding understanding ADH4 effects on norepine- altered translational efficiency, or even different protein phrine metabolism. This work was supported in part by National Institute expression levels in different tissues. Considering our on Drug Abuse (NIDA) grants R01-DA12849, R01-DA12690, and K24- sample size limitations, we believe that replication of DA15105, National Institute on Alcohol Abuse and Alcoholism (NIAAA) these results is critical. Nevertheless, given our findings, grants R01-AA11330, R01-AA016015, P50-AA12870, P50-AA03510, and we believe that it would be productive to study the effect K24-AA13736, National Center for Research Resources (NCRR) grant M01-RR06192, Alcoholic Beverage Medical Research Foundation (ABMRF) of this variation directly on protein expression, in order to grant award R06932 (X Luo); and by the U.S. Department of Veterans provide convergent validation of the findings reported Affairs (the VA Medical Research Program, VA Alcohol Research Center, here and to elucidate the specific mechanism underlying and the VA Connecticut-Massachusetts Mental Illness Research, Education the association of SNP8 at ADH4 to both AD and DD. and Clinical Center [MIRECC], and the VA Research Enhancement Award Program [REAP] research center). Page 11 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 25. Horvath S, Xu X, Laird NM: The family based association test References method: strategies for studying general genotype–pheno- 1. Ditlow CC, et al.: Physical and enzymatic properties of a class type associations. Eur J Hum Genet 2001, 9(4):301-6. II alcohol dehydrogenase isozyme of human liver: pi-ADH. 26. Liu K: PowerMarker: new genetic data analysis software. Biochemistry 1984, 23(26):6363-8. 2007. Free program distributed by the author over the internet from 2. Li TK, et al.: Isolation of pi-alcohol dehydrogenase of human http://www.powermarker.net liver: is it a determinant of alcoholism? Proc Natl Acad Sci USA 27. Stein MB, Schork NJ, Gelernter J: A polymorphism of the beta1- 1977, 74(10):4378-81. adrenergic receptor is associated with low extraversion. Biol 3. Guindalini C, et al.: Association of genetic variants in alcohol Psychiatry 2004, 56(4):217-24. dehydrogenase 4 with alcohol dependence in Brazilian 28. Yang BZ, et al.: Practical population group assignment with patients. Am J Psychiatry 2005, 162(5):1005-7. selected informative markers: characteristics and properties 4. Edenberg HJ, Jerome RE, Li M: Polymorphism of the human alco- of Bayesian clustering via STRUCTURE. Genet Epidemiol 2005, hol dehydrogenase 4 (ADH4) promoter affects gene expres- 28(4):302-12. sion. Pharmacogenetics 1999, 9(1):25-30. 29. Pritchard JK, et al.: Association mapping in structured popula- 5. Edenberg HJ, et al.: Association of alcohol dehydrogenase genes tions. Am J Hum Genet 2000, 67(1):170-81. with alcohol dependence: a comprehensive analysis. Hum Mol 30. Wienker T: FINETTI: Tests for deviation from Hardy-Wein- Genet 2006, 15(9):1539-49. berg equilibrium and tests for association in case-control 6. Luo X, et al.: ADH4 gene variation is associated with alcohol studies. 2002. Free program distributed by the author over the and drug dependence: results from family controlled and internet from http://ihg.gsf.de/cgi-bin/hw/hwa1.pl population-structured association studies. Pharmacogenet 31. Miller M: TFPGA: Tools For Population Genetic Analyses. Genomics 2005, 15(11):755-68. 2006. Free program distributed by the author over the internet from 7. Luo X, et al.: ADH4 gene variation is associated with alcohol http://www.marksgeneticsoftware.net dependence and drug dependence in European Americans: 32. Demyttenaere K, et al.: Prevalence, severity, and unmet need results from HWD tests and case-control association stud- for treatment of mental disorders in the World Health ies. Neuropsychopharmacology 2006, 31(5):1085-95. Organization World Mental Health Surveys. Jama 2004, 8. Luo X, et al.: Personality traits of agreeableness and extraver- 291(21):2581-90. sion are associated with ADH4 variation. Biol Psychiatry 2007, 33. Cramer H: Mathematical methods of statistics. Princeton, NJ: 61(5):599-608. Princeton University Press; 1946. 9. Wittke-Thompson JK, Pluzhnikov A, Cox NJ: Rational inferences 34. Weir B: Disequilibrium, in In: Genetic data analysis II: meth- about departures from Hardy-Weinberg equilibrium. Am J ods for discrete population genetic data. Sinaur Associates: Hum Genet 2005, 76(6):967-86. Sunderland, MA; 1996:91-139. 10. McGinnis R, Shifman S, Darvasi A: Power and efficiency of the 35. Feder JN, et al.: A novel MHC class I-like gene is mutated in TDT and case-control design for association scans. Behav patients with hereditary haemochromatosis. Nat Genet 1996, Genet 2002, 32(2):135-44. 13(4):399-408. 11. Parra EJ, et al.: Estimating African American admixture pro- 36. Jiang R, et al.: Fine-scale mapping using Hardy-Weinberg dise- portions by use of population-specific alleles. Am J Hum Genet quilibrium. Ann Hum Genet 2001, 65(Pt 2):207-19. 1998, 63(6):1839-51. 37. Mardh G, et al.: Human class II (pi) alcohol dehydrogenase has 12. Hoggart CJ, et al.: Control of confounding of genetic associa- a redox-specific function in norepinephrine metabolism. Proc tions in stratified populations. Am J Hum Genet 2003, Natl Acad Sci USA 1986, 83(23):8908-12. 72(6):1492-1504. 38. Sofuoglu M, et al.: Intravenous cocaine increases plasma epine- 13. Shriver MD, et al.: Skin pigmentation, biogeographical ancestry phrine and norepinephrine in humans. Pharmacol Biochem Behav and admixture mapping. Hum Genet 2003, 112(3):387-99. 2001, 68(3):455-9. 14. Collins-Schramm HE, et al.: Mexican American ancestry-inform- 39. Bouaziz H, et al.: Intravenous opioids stimulate norepinephrine ative markers: examination of population structure and and acetylcholine release in spinal cord dorsal horn. System- marker characteristics in European Americans, Mexican atic studies in sheep and an observation in a human. Anesthe- Americans, Amerindians and Asians. Hum Genet 2004, siology 1996, 84(1):143-54. 114(3):263-71. 40. Nielsen DM, Ehm MG, Weir BS: Detecting marker-disease asso- 15. Pritchard JK, Stephens M, Donnelly P: Inference of population ciation by testing for Hardy-Weinberg disequilibrium at a structure using multilocus genotype data. Genetics 2000, marker locus. Am J Hum Genet 1998, 63(5):1531-40. 155(2):945-59. 41. Lee WC: Searching for disease-susceptibility loci by testing 16. Gossop M: Multiple substance use and multiple dependencies, for Hardy-Weinberg disequilibrium in a gene bank of in Dual Diagnosis and Psychiatric Treatment. Edited by: Kran- affected individuals. Am J Epidemiol 2003, 158(5):397-400. zler H. Marcel Dekker: New York; 2004:129-156. 42. Hao K, et al.: Power estimation of multiple SNP association 17. Luo X, et al.: CHRM2 gene predisposes to alcohol dependence, test of case-control study and application. Genet Epidemiol drug dependence and affective disorders: results from an 2004, 26(1):22-30. extended case-control structured association study. Hum Mol 43. Shai D, Aptekar L: Factors in mortality by drug dependence Genet 2005, 14(16):2421-34. among Puerto Ricans in New York City. Am J Drug Alcohol Abuse 18. Gelernter J, et al.: Genomewide linkage scan for cocaine 1990, 16(1–2):97-107. dependence and related traits: significant linkages for a 44. Poser W, Poser S, Eva-Condemarin P: Mortality in patients with cocaine-related trait and cocaine-induced paranoia. Am J Med dependence on prescription drugs. Drug Alcohol Depend 1992, Genet B Neuropsychiatr Genet 2005, 136B(1):45-52. 30(1):49-57. 19. Gelernter J, et al.: Genomewide linkage scan for opioid depend- 45. Dawson DA: Alcohol consumption, alcohol dependence, and ence and related traits. Am J Hum Genet 2006, 78(5):759-69. all-cause mortality. Alcohol Clin Exp Res 2000, 24(1):72-81. 20. Pierucci-Lagha A, et al.: Diagnostic reliability of the Semi-struc- 46. Rundle AT, Atkin J, Sudell B: Serum and tissue proteins in tuber- tured Assessment for Drug Dependence and Alcoholism ous sclerosis. I. Serum and red-cell polymorphic systems. (SSADDA). Drug Alcohol Depend 2005, 80(3):303-12. Humangenetik 1975, 27(1):15-22. 21. American Psychiatric Association: Diagnostic and statistical man- 47. Deng HW, Chen WM, Recker RR: QTL fine mapping by measur- ual of mental disorders. revised ed. third edition. Washington, ing and testing for Hardy-Weinberg and linkage disequilib- DC: American Psychiatric Press; 1987. rium at a series of linked marker loci in extreme samples of 22. American Psychiatric Association: Diagnostic and statistical man- populations. Am J Hum Genet 2000, 66(3):1027-45. ual of mental disorders. fourth edition. Washington, DC: Ameri- 48. Song K, Elston RC: Tests for a disease-susceptibility locus can Psychiatric Press; 1994. allowing for an inbreeding coefficient (F). Genetica 2003, 23. Barrett JC, et al.: Haploview: analysis and visualization of LD 119(3):269-81. and haplotype maps. Bioinformatics 2005, 21(2):263-5. 49. Song K, Elston RC: A powerful method of combining measures 24. Dudbridge F: Pedigree disequilibrium tests for multilocus hap- of association and Hardy-Weinberg disequilibrium for fine- lotypes. Genet Epidemiol 2003, 25(2):115-21. mapping in case-control studies. Stat Med 2006, 25(1):105-26. Page 12 of 13 (page number not for citation purposes) Behavioral and Brain Functions 2008, 4:42 http://www.behavioralandbrainfunctions.com/content/4/1/42 50. Kocsis I, et al.: Examination of Hardy-Weinberg equilibrium in papers of Kidney International: an underused tool. Kidney Int 2004, 65(5):1956-8. 51. Weinberg CR, Morris RW: Invited commentary: Testing for Hardy-Weinberg disequilibrium using a genome single- nucleotide polymorphism scan based on cases only. Am J Epi- demiol 2003, 158(5):401-3. discussion 404-5 52. Leal SM: Detection of genotyping errors and pseudo-SNPs via deviations from Hardy-Weinberg equilibrium. Genet Epidemiol 2005, 29(3):204-14. 53. Geller F, Ziegler A: Detection rates for genotyping errors in SNPs using the trio design. Hum Hered 2002, 54(3):111-7. 54. O'Connell JR, Weeks DE: PedCheck: a program for identifica- tion of genotype incompatibilities in linkage analysis. Am J Hum Genet 1998, 63(1):259-66. 55. Hedrick PW: Recombination and directional selection. Nature 1983, 302(5910):727. 56. Barton NH, Turelli M: Effects of genetic drift on variance com- ponents under a general model of epistasis. Evolution 2004, 58(10):2111-32. 57. Luo X, et al.: Response to Dr. Kopke's comments on haplo- types at the OPRM1 locus. Am J Med Genet B Neuropsychiatr Genet 2005, 135B(1):102. 58. Soltis D: Isozymes in plant biology. Dioscorides Press: Portland, Oregon; 1989:130. 59. Charlesworth B, Charlesworth D: The genetic basis of inbreed- ing depression. Genet Res 1999, 74(3):329-40. Publish with Bio Med Central and every scientist can read your work free of charge "BioMed Central will be the most significant development for disseminating the results of biomedical researc h in our lifetime." Sir Paul Nurse, Cancer Research UK Your research papers will be: available free of charge to the entire biomedical community peer reviewed and published immediately upon acceptance cited in PubMed and archived on PubMed Central yours — you keep the copyright BioMedcentral Submit your manuscript here: http://www.biomedcentral.com/info/publishing_adv.asp Page 13 of 13 (page number not for citation purposes)

Journal

Behavioral and Brain Functions – Springer Journals

Published: Sep 18, 2008

There are no references for this article.