|Home | About | Journals | Submit | Contact Us | Français|
Dopamine β-hydroxylase (DβH) catalyzes the conversion of dopamine to norepinephrine. DβH enters the plasma after vesicular release from sympathetic neurons and the adrenal medulla. Plasma DβH activity (pDβH) varies widely among individuals, and genetic inheritance regulates that variation. Linkage studies suggested strong linkage of pDβH to ABO on 9q34, and positive evidence for linkage to the complement fixation locus on 19p13.2-13.3. Subsequent association studies strongly supported DBH, which maps adjacent to ABO, as the locus regulating a large proportion of the heritable variation in pDβH. Prior studies have suggested that variation in pDβH, or genetic variants at DβH, associate with differences in expression of psychotic symptoms in patients with schizophrenia and other idiopathic or drug-induced brain disorders, suggesting that DBH might be a genetic modifier of psychotic symptoms. As a first step toward investigating that hypothesis, we performed linkage analysis on pDβH in patients with schizophrenia and their relatives. The results strongly confirm linkage of markers at DBH to pDβH under several models (maximum multipoint LOD score, 6.33), but find no evidence to support linkage anywhere on chromosome 19. Accounting for the contributions to the linkage signal of three SNPs at DBH, rs1611115, rs1611122, and rs6271 reduced but did not eliminate the linkage peak, whereas accounting for all SNPs near DBH eliminated the signal entirely. Analysis of markers genome-wide uncovered positive evidence for linkage between markers at chromosome 20p12 (multi-point LOD = 3.1 at 27.2 cM). The present results provide the first direct evidence for linkage between DBH and pDβH, suggest that rs1611115, rs1611122, rs6271 and additional unidentified variants at or near DBH contribute to the genetic regulation of pDβH, and suggest that a locus near 20p12 also influences pDβH.
Dopamine β-hydroxylase (DβH) catalyses the conversion of dopamine to norepinephrine within synaptic vesicles or chromaffin granules of neurons and neurosecretory cells that release norepinephrine and epinephrine. DβH protein is released, together with other synaptic contents, during sympathetic neuronal activity (Smith et al. 1970), and also appears to be secreted by a constitutive pathway (Oyarce and Fleming 1991). By virtue of its secretion into the extracellular space, DβH enzyme activity can readily be assayed in the plasma or serum. Family and twin studies have demonstrated that plasma DβH (pDβH) activity is a highly heritable trait, with segregation analysis supporting a single major locus responsible for low pDβH activity (Ross et al. 1973; Weinshilboum et al. 1975).
A large number of studies (Cubells and Zabetian 2004) have examined pDβH for associations with disorders such as schizophrenia (MIM 181500), major depression (MIM 608516) and Parkinson's disease (MIM 168100). Unfortunately, this literature is conflicting and difficult to interpret, in large part, because very few such studies have accounted for the genetic influence on the pDβH trait. Despite this shortcoming, one consistent pattern that emerges from studies of pDβH is an association between lower pDβH and psychotic features in several psychiatric disorders. Perhaps the most consistent observation in this regard is that patients who have unipolar major depression with psychotic features exhibit lower pDβH than depressed patients without such symptoms (Cubells et al. 2002; Lykouras et al. 1988; Meltzer et al. 1976; Meyers et al. 1999; Mod et al. 1986; Sapru et al. 1989): five studies have found such a relationship (Cubells et al. 2002; Meltzer et al. 1976; Meyers et al. 1999; Mod et al. 1986), and only one has not (Lykouras et al. 1988). Studies of pDβH in schizophrenia clearly have ruled out a simple association between variation in pDβH and the diagnosis of the disorder, but some studies (Sternberg et al. 1983; van Kammen et al. 1994) have noted differences in symptomatic patterns, premorbid or morbid functioning, or in clinical course, associated with variation in pDβH activity, cerebrospinal fluid DβH activity, or protein levels [which are both highly correlated to pDβH (Castellani et al. 1982; O'Connor et al. 1994)]. The foregoing observations suggest that a detailed understanding of the genetics underlying pDβH might reveal important information about genetic modifiers of psychosis across a variety of psychiatric diagnoses.
Linkage analysis performed with non-DNA markers provided evidence for linkage of pDβH to the ABO blood group locus (MIM 110300) (Elston et al. 1979; Goldin et al. 1982; Wilson et al. 1988), which maps to 9q34. This evidence together with the molecular cloning and mapping of the structural gene DBH (MIM 609312) to 9q34 (Craig et al. 1988; Gelernter et al. 1991; Lamouroux et al. 1987) strongly suggested DBH is a major quantitative trait locus regulating pDβH activity, and this hypothesis was confirmed by reports of associations between molecular markers at DBH and variation in pDβH (Cubells et al. 1998; Wei et al. 1997). Zabetian and colleagues (Zabetian et al. 2001) then resequenced DNA samples from individuals with extreme values of pDβH activity, and identified a single nucleotide polymorphism (SNP), consisting of a cytosine to thymidine transition located 970 bp 5′ to the transcriptional start site of DBH (−970C>T; rs1611115, formerly called −1021C>T) that accounted for 30–50% of the variance in pDβH activity in samples from three human populations of differing geographic ancestry (European Americans, African Americans and Japanese). In all populations examined thus far, the T allele of −970C>T associates in a co-dominant manner with lower pDβH activity (Cubells et al. 2002; Kohnke et al. 2002; Mustapic et al. 2007; Tang et al. 2007). Furthermore, −970C>T accounts for a greater proportion of the variance in pDβH activity than any other SNPs at DBH examined to date (Chen et al. 2010; Zabetian et al. 2003). Transient transfection analysis in PC-12 cells of luciferase reporter constructs driven by human DBH sequence, together with co-transfection and chromatin immunoprecipitation experiments, recently provided strong evidence that −970C>T is a functional polymorphism influencing transcription of DBH (Chen et al. 2010). A non-synonymous SNP located in exon 11 (rs6271) accounts for additional variance in pDβH after allowing for −970C>T statistically (Tang et al. 2006; Zabetian et al. 2001), or by examination of the association of this SNP with pDβH activity in samples chosen for homozygous genotype at −970C>T (Tang et al. 2005).
Some data suggest the presence of at least one additional locus contributing to variation in pDβH. Wilson and colleagues (Wilson et al. 1988), in addition to a strong linkage signal between pDβH activity and ABO, also reported a weaker signal at C3, the complement fixation locus (MIM 120700), which maps to chromosome 19p13.2-13.3. A re-analysis of those data confirmed those results (Province 2000). No linkage analysis to our knowledge has confirmed the linkage between pDβH and a locus on 9q34, nor evaluated the possible linkage near C3 on 19p. Indeed, the linkage studies of the pDβH activity phenotype to date have all used phenotypic markers covering only a small proportion of the genome (~5%).
The current study was undertaken for the following purposes: (1) to confirm linkage and association between pDβH and SNP markers at the DBH locus; (2) to evaluate possible linkage between pDβH and distal chromosome 19p; (3) to identify evidence that any other loci in the genome show linkage to pDβH activity; and (4) to evaluate the contribution of rs1611115 (−970C>T), rs1611122 and rs6271 to the hypothesized linkage at DBH, in preparation for studies of these functional variants as potential modifiers of psychotic symptoms in schizophrenia. We report analysis of linkage between a genome-wide panel of SNP markers (Holmans et al. 2009) and pDβH activity, assayed in archived plasma samples from participants in family-based studies of schizophrenia.
The data were generated from plasma and DNA samples donated by a subset of 123 European American families, multiplex for schizophrenia, comprising 921 individuals (Holmans et al. 2009). A total of 284 individuals from 70 pedigrees had samples of plasma available for assay of pDβH activity. Of the 70 pedigrees, 59 had two generations, 10 had 3 generations, and 1 pedigree had 4 generations.
Samples of acid-citrate anti-coagulated plasma, collected at the time of participant ascertainment, were stored frozen at −80°C until DβH-activity assay. pDβH activity was determined in duplicate 5-μl aliquots of plasma as the rate of conversion of tyramine to octopamine using a method modified from that of Nagatsu et al. (Cubells et al. 1998; Nagatsu and Udenfriend 1972). Octopamine was measured by a column-switching, reverse phase high performance liquid chromatography (HPLC) system, using coulometric electrochemical detection, and synephrine as internal standard. DβH activity is reported as μmols of octopamine formed per minute from a solution of tyramine (0.2 mol L−1) by 1 L of plasma at 37°C (μmoL min−1 L−1). The average (±SD) duration of storage was 10.5 ± 2.0 years.
Genotypes were obtained on 210 SNP markers on chromosome 9 with mean spacing 0.76 cM, and 153 markers on chromosome 19 with mean spacing 0.73 cM. These genotypes were all from the Illumina Version 4 Linkage Panel, except genotypes for rs1611115 and rs6271, which were determined using the Taqman 5′ exonuclease method. The genome-wide analyses employed a total of 5584 markers on 22 autosomal chromosomes passing quality control checks. X-linked markers were not analyzed because familial correlations, calculated by the S.A.G.E. program FCOR, gave no suggestion of X-linkage (supplementary Table 1).
In order to perform model-based linkage analysis, the trait model parameters (the trait genotypic means, a common variance at the trait locus, and the trait locus allele frequencies) were estimated using a diallelic model with the SEGREG program of S.A.G.E. 6.01. In a model-based linkage analysis, it is important to have an appropriate model. Therefore, we explored various modes of inheritance and power transformations of the data values to determine appropriate distributional assumptions for the model used. The Box-Cox transformation is useful in this regard because it comprises a family of power transformations that includes as a special case the logarithmic and square root transformations. Under the assumption of Mendelian transmission, low-value dominant, low-value recessive and additive models were fitted for each of the following four transformations of pDβH activities: (1) no transformation, (2) square root transformation, (3) logarithmic transformation, and (4) Box-Cox transformation with estimated power λ1. Therefore, a total of 12 trait models were produced for the pDβH activity trait. A trait-locus type probability file, which provides individual-specific penetrance functions for linkage analysis, was produced by SEGREG for each of the trait models.
We planned a priori to examine linkage first on chromosome 9, followed by linkage on chromosome 19, because of the prior evidence for linkage on those chromosomes. For the chromosome-specific analyses, two-point and multipoint model-based linkage analyses were, respectively, performed under the trait models produced in SEGREG. Using diallelic trait models produced by SEGREG, two-point linkage was conducted with the LODLINK program, and multipoint linkage was performed with MLOD, these being programs in S.A.G.E. 6.01. After confirming linkage on chromosome 9 but not on chromosome 19, genome-wide linkage analysis was performed.
In addition, model-free linkage was conducted with SIBPAL, with the sibship mean estimated by the best linear unbiased predictor [BLUP; (Wang and Elston 2004)], and the regression method W4 was selected to set the dependent variable as a weighted combination of the squared trait difference and the squared mean-corrected trait sum (Shete et al. 2003). For this analysis, exact multipoint IBD sharing distributions for full siblings were estimated at 2 cM intervals by GENIBD. Both SIBPAL and GENIBD are part of the S.A.G.E. package.
In order to adjust for the linkage signal on chromosome 9 prior to testing for linkage on chromosome 19 and elsewhere, the square root of pDβH activities were regressed on (1) the SNPs rs1611115, rs1611122 and rs6271 in the DBH gene, coded by 14 covariates to represent their additive, dominant and interaction effects, and (2) the 25 SNPs in the 40 cM linkage region on chromosome 9, coded by 40 additive and/or dominant covariates (including the 14 covariates for the 3 SNPs in DBH). The square root of pDβH activities was used for this purpose both because it gave the strongest linkage signal and because our previous study (Zabetian et al. 2001) showed that, for rs1611115 genotypes, a plot of the genotype-specific variance of pDβH activity against mean pDβH activity produced approximately straight lines passing through the origin for three ethnically different populations (see supplementary Fig. 1), suggesting that a square root transformation would be the appropriate transformation to produce a stable variance for the three genotypes (Healy 1968). From the total of 45 SNPs in the 40 cM region, the 25 SNPs individually showing the most significant associations (p value for each covariate less than 0.2 by a likelihood ratio test) were selected for this purpose. The regressions were performed using the ASSOC program in S.A.G.E. 6.01, which allows for familial correlations in the trait. The residuals of the two regression models, which were the differences between the square root of pDβH and the sum of weighted SNP covariate values, were first adjusted to be on a scale comparable to that of the original pDβH activities by adding a constant and then squaring (Wilson et al. 1990); they were then used as new traits. Best-fitting segregation models were fitted to the new traits, to produce a new set of low-value dominant, low-value recessive and additive models under no transformation, a square root transformation, and a logarithmic transformation. Then linkage analyses with these new trait models were performed on chromosome 19 and elsewhere as described above.
Segregation analysis strongly rejected environmental transmission of the trait pDβH under all models (data not shown). Mendelian dominant, recessive and additive transmission models under no transformation, square root transformation, logarithmic transformation, and Box-Cox transformation with the power λ1 estimated were used in both two-point and multipoint linkage analyses. The model producing the best genomewide linkage signal was the low-value dominant model under the square root transformation (Table 1). In addition, there was no significant effect of schizophrenia on pDβH activity when schizophrenia status was included as a covariate of penetrance in the segregation analysis.
The strongest linkage signal was on chromosome 9q (Fig. 1). The highest multipoint linkage LOD score was 6.33 at the position 2.8 cM proximal to the DBH locus. At the three SNPs (rs1611115, rs1611122, rs6271) in the DBH gene, the multipoint LOD score was 6.0. Two-point linkage under this model produced LOD scores at rs1611115, rs1611122 and rs6271 that were, respectively, 2.32, 3.6 and 0.29, each with a recombination fraction of 0. Association analysis of the square root of pDβH, performed using ASSOC within S.A.G.E., showed that SNP rs1611115 associated with pDβH in a co-dominant manner, with means 4.20, 2.90 and 1.06 (μmoL min−1 L−1)0.5, i.e. 17.64, 8.41 and 1.12 μmoL min−1 L−1, respectively, for the CC, CT and TT genotypes (p = 1.14 × 10−18; Table 2).
Linkage analysis was repeated on the new traits formed after adjusting for the linkage signal on chromosome 9, using dominant, recessive, and additive segregation models fitted for the new traits under no transformation, square root, and logarithmic transformations. Using a new dominant model without transformation for the new trait, having adjusted for rs1611115, rs1611122, and rs6271 (Table 1), the best linkage signal was again detected on chromosome 9, but decreased by 47%. The highest multipoint LOD score was 3.37 on chromosome 9q34, 9.5 cM from the DBH gene, with the one-LOD support interval encompassing the DBH gene (Fig. 1). For the new trait obtained on adjusting for the 25 SNPs in the linkage region on chromosome 9, no suggestive linkage was identified at the DBH locus (Fig. 1: the best linkage signal on chromosome 9 was identified under a new additive model using the logarithmic transformation—multipoint LOD score of 1.39 at 6.6 cM from the DBH gene), which suggests that the original linkage signal at DBH was now adjusted out.
Using the new trait models fitted after adjusting for the 25 SNPs, no linkage was found around the C3 locus (19p13) on chromosome 19. Among all these models, a Mendelian dominant model under logarithmic transformation produced the best, but quite weak, multipoint linkage signal on chromosome 19q13 (multipoint LOD = 1.0). The position of that peak was very far from the C3 locus (Supplementary Fig. 2). To confirm the absence of a significant linkage signal on chromosome 19, and to account for the possibility that gametic disequilibrium between DBH and a locus on chromosome 19 explains the lack of a signal after accounting for DBH, we also examined linkage on chromosome 19 without adjusting out DBH. That analysis failed to show any significant linkage across the entire chromosome, the best linkage signals being identified by a Mendelian additive model under logarithm transformation, at positions 0 and 7.7 cM on 19p, with multipoint LOD scores of 1.32 and 1.15, respectively. The positions are 21.6 and 13.9 cM away from the C3 locus (Supplementary Fig. 2).
Model free linkage analysis indicated that the most significant linkage peak for the square root of pDβH activity was at 9q34, 9.4 cM from the DBH gene, with p value = 7.6 × 10−12. After adjusting for the 25 SNPs in the linkage region, the model free linkage signal of the new trait on chromosome 9 decreased to p = 0.01 at 9.4 cM from the DBH gene, and to p = 0.13 at the gene itself (Fig. 2). There was no significant linkage for the new trait on chromosome 19 (Supplementary Fig. 3).
The proportion of the square root pDβH heritability accounted for by the linkage on chromosome 9 was estimated in two ways. First, using the 266 individuals for whom the 25 selected SNPs were available, the estimated heritability of square root pDβH activity was 0.62 ± 0.14. After adjusting for the 25 SNPs in the DBH region, the heritability of the square root of the new trait (comparable to the square root of pDβH in scale) decreased to 0.37 ± 0.18, indicating that the markers at the DBH region accounted for 40% of the genetic component of the square root pDβH activity. Second, the genetic variance of the square root of pDβH due to the DBH gene is equal to the regression coefficient of the proportion of alleles shared IBD at the gene, estimated to be 1.91 for square root pDβH. After adjusting out the linkage signal on chromosome 9, the genetic variance at the DBH gene was estimated to be 0.69 for square root pDβH. Thus the decrease in the genetic variance at the DBH locus is estimated to be 64%. Knowing that the total variance of square root pDβH (based on all 284 individuals for whom pDβH is available) is 3.09, the heritability by this method is estimated to decrease by 39% (consistent with the above method accounting for 40% of the genetic component of pDβH activity).
To provide preliminary data on linkage between loci on other chromosomes and plasma DβH activity, we examined linkage results for the new trait formed by adjusting for the 25 SNPs with markers from the entire genome. The most significant model-based linkage result for the new trait was obtained under the new additive model with a square root transformation (Table 1), and revealed a positive significant peak on chromosome 20p12 (multi-point LOD = 3.1 at 27.2 cM; Fig. 3). No other substantial peaks were observed (Supplementary Fig. 4). The most significant linkage for pDβH other than the one on chromosome 9 was also obtained on chromosome 20p12 (Fig. 3, multipoint LOD = 2.4 at 43.2 cM), under the same segregation model that produced the best linkage signal on the DBH locus (Table 1, dominant model under the square root transformation). Thus adjusting for the effects of the markers on chromosome 9 made the linkage signal on chromosome 20 become somewhat stronger, but its position moved.
The present study, to our knowledge, is the first linkage study of pDβH activity to employ genotypes inferred directly from DNA markers as opposed to protein phenotypes. The results confirm evidence from linkage studies that used protein and blood group markers some 30 years ago, indicating that a major quantitative trait locus regulating pDβH activity resides on distal 9q34 near ABO (Goldin et al. 1982; Province 2000; Wilson et al. 1988), and shows with finer resolution that the linkage peak encompasses the structural gene DBH. Those results are consistent with strong association evidence implicating DBH as a QTL regulating pDβH activity (Chen et al. 2010; Zabetian et al. 2001). However, the present analysis fails to support prior findings suggesting a second QTL on chromosome 19 near C3, which had been reported in one of the earlier linkage studies (Province 2000; Wilson et al. 1988). The prior linkage studies, based on non-DNA markers, covered only ~5% of the genome, and only included one marker, adenosine deaminase, on chromosome 20q. The current analysis, using a genome-wide SNP marker set, revealed an additional positive (but not significant) linkage peak on chromosome 20p12. On the basis of the model-free regression coefficient at the peak signal, the locus-specific variance here is estimated to be 99.6, corresponding to 49.6% of the variance of the original pDβH activity. This estimate, however, is necessarily an overestimate on account of being determined from the point at which the linkage signal is most significant.
The association analysis performed on the markers spanning DBH confirms the well-replicated finding that rs1611115 is strongly associated with pDβH in a co-dominant pattern, in which the T allele associates with lower activity than the C allele. However, the best linkage results were found under a low-value dominant model. The most likely reason for this discrepancy is that the distribution of values of pDβH activity assayed in the present study appears to be somewhat compressed toward the lower end of the expected range based on prior studies (Kohnke et al. 2002; Zabetian et al. 2001). There are two likely explanations for that compression of values. First, the samples assayed for pDβH activity had been stored at −80° C for more than a decade, and it is possible that some loss of activity occurred during storage. Second, the plasma samples assayed in this study were anticoagulated in acid citrate buffer, whereas prior studies have examined either heparin-anti-coagulated plasma (Zabetian et al. 2001) or serum (Weinshilboum 1978). Unpublished experiments performed by R.W.B and J.F.C show that acid citrate diminishes activity by approximately 50% (presumably due to copper chelation by acid citrate). The compression of values due to those, and possibly other factors, is likely to affect higher values more substantially than lower values owing to a “floor effect.” Such asymmetric compression of the distribution of values would, therefore, be expected to exaggerate the influence of the low-activity-associated alleles, leading an additive system to appear low-value dominant. The results were not affected by including the duration of sample storage as a covariate but, since almost all of the samples were stored for more than 10 years, that observation cannot exclude the possibility of a sample-storage effect. Samples of serum or heparin-anticoagulated plasma were not available, so we were unable to evaluate the influence of anti-coagulant directly in this sample set. Regardless of the possible effects of factors such as duration of sample storage or anticoagulants, the strong replication of the 9q34 linkage, in addition to replication of the strong association of rs1611115 with pDβH activity, indicate that the data from this study are useful and reasonably comparable to prior data.
It is important to note that accounting for the SNPs rs1611115, rs1611122 and rs6271 substantially diminished, but did not completely eliminate, the linkage signal at 9q34, resulting in a residual multi-point lod score of 3.37 under a low-value dominant model. However, accounting for all SNPs in the 9q34 linkage region completely eliminated the signal, strongly suggesting that additional variants at or near DBH contribute to regulation of pDβH activity. One evidence-supported candidate for such a variant is rs1989787. Chen and colleagues (Chen et al. 2010) recently showed that SNP to associate with pDβH activity after statistically accounting for rs1611115. In addition, analysis of reporter-gene transcription indicated that rs1989787 appears to alter transcription driven by the DBH promoter (Chen et al. 2010). Prior association studies from our laboratory have also suggested the presence of at least one additional functional variant, in linkage disequilibrium with a SNP within intron 5 of DBH but not with rs161115 or rs6271 (Tang et al. 2006).
The finding of positive (but not genome-wide significant) evidence for linkage between markers on chromosome 20 and variation in pDβH is an entirely new observation. Although the fact that the chromosome-20 model-based linkage peak became stronger after correcting for the linkage at DBH is encouraging, only replication in independent studies will distinguish a true linkage from a chance finding. We note here that the limited sample size in the current study substantially reduced our power to detect linkage in a genome-wide scan, and this is an important limitation of the study. If further research verifies a chromosome-20 linkage to pDβH, it will be interesting to examine how much of the unaccounted-for genetic variance such a linkage explains. It should be noted in passing that the segregation model that best fits the trait data alone is not always the segregation model that, when genetic markers are included for a model-based linkage analysis, gives the strongest linkage signal. We cannot exclude the interesting possibility that the positive linkage signal on chromosome 20 (or the lack of such a signal on chromosome 19) is due to the fact that the current study examines pedigrees ascertained for schizophrenia. However, without additional evidence to support such a hypothesis, we believe it unlikely.
Prior studies of pDβH in psychiatric illness reveal a pattern of association of lower pDβH or CSF levels of DβH with psychosis-related phenotypes, including psychotic features in major depression (Cubells et al. 2002; Lykouras et al. 1988; Meltzer et al. 1976; Meyers et al. 1999; Mod et al. 1986), more florid symptoms of schizophrenia described in the original publications as “reactive” as opposed to “process” schizophrenia (Sternberg et al. 1983; van Kammen et al. 1994), and higher risk for psychosis following treatment with disulfiram (which inhibits DβH) in alcoholics (Major et al. 1979). Furthermore, genotypes and haplotypes at DBH that associate with lower pDβH have also been reported to associate with cocaine-induced paranoia in chronic users of cocaine (Cubells et al. 2000; Kalayasiri et al. 2007), and poor outcome in patients with schizophrenia (Yamamoto et al. 2003). Those findings, taken together, suggest that DBH might influence the phenotypic expression of psychosis in a variety of drug-induced and idiopathic psychotic disorders. While the precise relationship between pDβH and central regulation of DβH activity remains unclear, a single gene encodes the central and peripheral protein, suggesting that molecular mechanisms regulating central and circulating DβH could at least partially overlap. The associations just cited, between variation in pDβH and differences in psychiatric phenomena almost certainly originating in the brain might reflect such common mechanisms. The present results set the stage for more detailed molecular-genetic investigation of that hypothesis in families segregating schizophrenia.
Supplementary Figure 1: The variance of pDβH activity plotted against mean pDβH activity for three populations, obtained from the table published in Zabetian et al (2001). For each population, the blue profile is the genotype-specific variance vs. the mean of pDβH activity at rs1611115, the red straight line connects the estimated mean for the high homozygote and the origin and the vertical lines are the 95% confidence intervals for the observed variances on the assumption of normal distributions.
Supplementary figure 2: Model-based linkage analysis of pDβH activity on chromosome 19. The red solid-line profile is the multi-point linkage for pDβH activity, with Mendelian additive model under logarithm transformation. The blue dotted-line profile is the multi-point linkage for the new trait (after adjusting for 25 SNPs on chromosome 9), with new Mendelian dominant model under logarithm transformation.
Supplementary figure 3: Model-free linkage analysis of pDβH activity on chromosome 19. The red solid-line profile is the linkage for pDβH activity under logarithm transformation. The blue dotted-line profile is the linkage for the new trait (after adjusting for 25 SNPs on chromosome 9), under logarithm transformation.
Supplementary figure 4: Genome-wide model-based linkage of the new trait formed by adjusting for the 25 SNPs, with an additive model under the square-root transformation.
The authors thank Kim Wardlaw for technical assistance. We are grateful to the patients and family members who participated in this study. This work was supported by NIH grants R01 MH 077233 (JFC) and RR 03655 (RCE).
The authors wish to dedicate this manuscript to V. K. Lasseter's memory.
Electronic supplementary material The online version of this article (doi: 10.1007/s00439-011-0989-6) contains supplementary material, which is available to authorized users.
Web resources: S.A.G.E. program: http://darwin.cwru.edu/sage/.
Conflict of interest To the best knowledge of the authors, there are no or any potential conflicts of interest relevant to the content of the current manuscript.
Joseph F. Cubells, Mercer Department of Human Genetics, Emory University School of Medicine, Atlanta 30322, GA, USA; Department of Psychiatry and Behavioral Sciences, Emory University School of Medicine, Atlanta, GA 30322, USA.
Xiangqing Sun, Department of Epidemiology and Biostatistics, Case Western Reserve University School of Medicine, Cleveland, OH 44106, USA.
Wenbiao Li, Mercer Department of Human Genetics, Emory University School of Medicine, Atlanta 30322, GA, USA.
Robert W. Bonsall, Department of Psychiatry and Behavioral Sciences, Emory University School of Medicine, Atlanta, GA 30322, USA.
John A. McGrath, Department of Psychiatry, Johns Hopkins School of Medicine, Baltimore, MD 21231, USA.
Dimitri Avramopoulos, Department of Psychiatry, Johns Hopkins School of Medicine, Baltimore, MD 21231, USA.
Virginia K. Lasseter, Department of Psychiatry, Johns Hopkins School of Medicine, Baltimore, MD 21231, USA.
Paula S. Wolyniec, Department of Psychiatry, Johns Hopkins School of Medicine, Baltimore, MD 21231, USA.
Yi-Lang Tang, Mercer Department of Human Genetics, Emory University School of Medicine, Atlanta 30322, GA, USA.
Kristina Mercer, Mercer Department of Human Genetics, Emory University School of Medicine, Atlanta 30322, GA, USA.
Ann E. Pulver, Department of Psychiatry, Johns Hopkins School of Medicine, Baltimore, MD 21231, USA.
Robert C. Elston, Department of Epidemiology and Biostatistics, Case Western Reserve University School of Medicine, Cleveland, OH 44106, USA.