Linkage of bipolar affective disorder on chromosome 8q24 - Nature

Molecular Psychiatry (2004) 9, 191–196 & 2004 Nature Publishing Group All rights reserved 1359-4184/04 $25.00 www.nature.com/mp

ORIGINAL RESEARCH ARTICLE

Linkage of bipolar affective disorder on chromosome 8q24: follow-up and parametric analysis D Avramopoulos1,3, VL Willour1,3, PP Zandi2, Y Huo1, DF MacKinnon1, JB Potash1, JR DePaulo Jr1 and MG McInnis1 1 2

Department of Psychiatry, Johns Hopkins University, School of Medicine, North Wolfe Street, Baltimore, MD 21287, USA; Department of Mental Health, Johns Hopkins University, School of Public Health, N. Broadway, Baltimore, MD 21205, USA Our group first reported a linkage finding for bipolar (BP) disorder on chromosome 8q24 in a study of 50 multiplex pedigrees, with an HLOD score reaching 2.39. Recently, Cichon et al reported an LOD score of 3.62 in the same region using two-point parametric analysis. Subsequently, we published the results of a genome scan for linkage to BP disorder using a sample extended to 65 pedigrees in which chromosome 8q24 provided the best finding, an NPL score of 3.13, approaching the accepted score for suggestive linkage. We have now fine mapped this region of chromosome 8 in our 65 pedigrees by the addition of 19 microsatellite markers reaching a marker density of 0.8 cM and an information content of 0.84. After the addition of the new data, the original NPL score slightly increased to 3.25. Two-point parametric analysis using the model employed by Cichon et al obtained an LOD score of 3.32 for marker D8S256 at h ¼ 0.14 exceeding the proposed threshold for genomewide significance. After adjusting the parameters in accordance with the ‘common disease–common variant’ hypothesis, multipoint parametric analysis resulted in an HLOD of 2.49 (a ¼ 0.78) between D8S529 and D8S256, and defined a 1-LOD interval corresponding to a 2.3 Mb region. No allelic association with the disease was observed for our set of microsatellite markers. Biologically, plausible candidate genes in this region include thyroglobulin, KCNQ3 coding for a voltagegated potassium channel and the gene for brain adenyl-cyclase (ADCY8). Molecular Psychiatry (2004) 9, 191–196. doi:10.1038/sj.mp.4001388 Keywords: bipolar disorder; linkage (genetics); genes; thyroglobulin; potassium channels; adenylate cyclase

Introduction Bipolar (BP) disorder is a severe psychiatric illness characterized by recurrent depressions and manias. It is a major public health problem affecting 1% of the population.1 There is an established but unknown genetic component to BP disorder,2–4 and it is likely that multiple genes contribute to disease susceptibility.5,6 There are now a number of regions (4p, 12q, 13q, 18q, 21q) that have been identified by linkage in independent studies.7 Recently, a susceptibility locus on chromosome 8q24 was supported by two independent genomewide scans of BP disorder. Cichon et al8 reported a peak LOD of 3.62 at D8S514 in 75 BP families using a dominant model in a parametric twopoint analysis, which overlaps with our previously reported finding on 8q24 using our initial 50 multiplex BP pedigrees.9 In that study, the strongest finding was an HLOD of 2.39 (a ¼ 0.42) between D8S256 and Correspondence: D Avramopoulos, Department of Psychiatry, Johns Hopkins University, Meyer Room 4-139, 600 North Wolfe Street, Baltimore, MD 21287, USA. E-mail: [email protected] 3 The first two authors contributed equally to the work. Received 21 March 2003; revised 14 May 2003; accepted 16 May 2003

D8S272, while other positive findings were on chromosomes 4 and 7. We have now extended our sample to 65 families, and in a recent genome scan10 our best finding was an NPL of 3.13 (P ¼ 0.003) at D8S256 on 8q24. We report here the analysis of the fine mapping of this region in these 65 BP pedigrees. An additional 19 microsatellite markers were typed and analyzed using the dominant model described by Cichon et al.8 The linkage evidence in the two-point analysis reaches genomewide significance and the multipoint analysis defines a region of 2.3 Mb likely to contain the susceptibility gene. Three candidate genes on 8q24 are briefly discussed.

Materials and methods The sample and ascertainment criteria have been described elsewhere.9,10 Briefly, our ascertainment criteria required the proband to be a treated BPI patient, from a unilineal pedigree with at least two first-degree relatives affected with BPI, BPII, schizoaffective-manic (SAM) or recurrent unipolar disorder (RUP). All subjects were interviewed using the semistructured Schedule for Affective Disorders and SchizophreniaFLifetime version (SADS-L).11 There were a total of 573 typed subjects for this study. We

Linkage of bipolar disorder to 8q24 D Avramopoulos et al

192

examined a narrow phenotypic model, where the 129 BPI, 97 BPII, and 7 SAM study subjects were considered affected, and a broad phenotypic model where the 69 RUP subjects were also considered affected. In both cases, the remaining relatives were coded as unknown. Genotyping of the original marker-set has been described elsewhere.9,10 The additional markers were selected from the Marshfield database or the DeCode linkage map12 and PCR conditions were retrieved from GDB. Alleles were separated by capillary electrophoresis of fluorescently labeled PCR products on an ABI-377 or an ABI-3700 genetic analyzer, multiplexing 8–10 products at a time. Allele sizes were determined using the Genescan and Genotyper software (Applied Biosystems) and binning was performed using internal software. Mendelian inconsistencies were detected using the Genetic Analysis System (GAS) version 2.0 (Alan Young, Oxford University 1993–95) and UNKNOWN.13 All plates contained two or more DNA samples from the CEPH panel for quality control. Mendelian inconsistencies were checked through review of the original data and when they could not be resolved, the marker data for the family were deleted. Marker HC83REP is a new polymorphic trinucleotide repeat that we identified starting at nucleotide 134 155 624 of the chromosome 8 sequence (UCSC golden path, November 2002 Freeze) within transcript AK023948. It was amplified using standard conditions and primers 50 -HEX-ACCCAGGGCTTCTCAGACTT-30 and 50 -GAAACCTGGCACTCAGCTTC-30 at an annealing temperature of 551C. The Marshfield linkage map (http://research.marshfieldclinic.org/genetics/) was primarily used to determine recombination rates and marker order, while the DeCode map12 and the UCSC (November 2002 Freeze) sequence were used for markers not present or not showing recombination to each other in the Marshfield database. When physical data were used, recombination fraction was approximated based on physical distances and recombination between flanking markers. There are no discrepancies between Marshfield, DeCode, and UCSC regarding marker order in our region of interest. Nonparametric multipoint linkage analyses were performed with the GENEHUNTER-PLUS modification14 of the GENEHUNTER package.15 Parametric analyses were performed using the computer program FASTLINK V4.016 and LODSCORE from the LINKAGE 5.10 program package.13 Like Cichon et al,8 we used a dominant model with a disease allele frequency of 0.009, a phenocopy rate (defined as the disease probability for those not carrying the disease allele) of 0.001 and a penetrance of 0.5. Parametric multipoint analysis was performed using GENEHUNTER. The Cichon et al model provided much lower scores in multipoint than in two-point analyses. As multipoint analysis is much more sensitive to model misspecification,17 we repeated the multipoint analysis using a set of parameters that better reflect the

Molecular Psychiatry

‘common disease–common variant’ hypothesis18 than the Cichon model. We assigned a lower penetrance (0.1) and a higher phenocopy rate (0.01) with a more frequent disease allele (P ¼ 0.06). The model was built based on a disease prevalence of 2%. Family-based association analyses for the detection of excess transmission of specific alleles to affected individuals were performed using the computer program FBAT.19

Results Figure 1 illustrates our nonparametric results, while Table 1 lists our nonparametric and two-point parametric results for both phenotypic models using the parameters of Cichon et al.8 In the parametric analysis (narrow phenotypic model), eight markers exceeded the proposed threshold for suggestive linkage (1.9 for parametric analysis). The LOD score for D8S256 was 3.32, which exceeds the threshold for genomewide significance (3.3 for parametric analysis).20 The results of the nonparametric analysis after the inclusion of the 19 additional markers were consistent with the results in the original marker set: An NPL of 3.25 (P ¼ 0.001), a marginal increase from the previous 3.13 (P ¼ 0.003). Despite the strong two-point results, multipoint parametric analysis using the Cichon et al parameters in GENEHUNTER did not produce equally strong results (HLOD ¼ 1.22 at D8S256). It has been argued that model misspecification may have a serious impact on multipoint analysis, yet little impact on two-point analysis.17 This prompted us to re-examine the Cichon et al8 linkage parameters. A phenocopy rate of 0.001, when the prevalence is 0.01, implies that most cases are due to one gene. With a penetrance of 0.5, a carrier of the disease allele has a disease risk of 0.5 compared to a risk of 0.001 for a noncarrier, suggesting a very high genotype relative risk. This is in contrast to what the ‘common disease–common variant’ hypothesis18 suggests. Using parameters consistent with such a hypothesis as described in our methods, we observed an HLOD of 2.49 (a ¼ 0.78) between markers D8S529 and D8S256. The results of the two-point and multipoint analysis using this model are summarized in Table 2 and Figure 2. In

Figure 1 Chromosome 8–multipoint NPL–multipoint NPL score vs position on the chromosome 8 map (Marshfield). The thicker line corresponds to the narrow and the finer line to the broad phenotypic model. The location of the peak finding of Cichon et al is indicated by an arrow.


Table 1 Nonparametric (GENEHUNTER) and two-point parametric (FASTLINK) linkage analysis results using the Cichon et al model for markers around our region of interest on chromosome 8, for the broad and narrow phenotypic models Narrow phenotypic model (BPI,BPII,SAM) Marker

Position (cM) GENEHUNTER

D8S1782 D8S284 D8S1765 D8S557 D8S1835 D8S558 D8S1740 D8S378 HC83REP D8S529 D8S256 D8S1708 D8S1796 D8S1746 D8S1462 D8S1710 D8S554 D8S1466 D8S1111 D8S1100 D8S274 D8S272 D8S523

142.75 143.82 144.36 145.97 145.97 145.97 147.04 147.04 (147.04) 148.12 148.12 148.12 148.12 149.46 (149.96) 150.8 153.48 (154.71) (155.11) 154.02 154.02 154.02 154.02

Two-point

NPL

(P)

LOD score

1.89 2.20 2.32 2.47 2.56 2.54 3.09 3.00 2.78 3.11 3.25 3.00 2.89 2.56 2.28 2.47 2.86 2.26 2.02 1.87 1.59 2.50 1.65

0.0326 0.0165 0.0124 0.0086 0.0068 0.0073 0.0017 0.0021 0.0039 0.0015 0.0010 0.0022 0.0029 0.0069 0.0136 0.0086 0.0031 0.0144 0.0248 0.0340 0.0584 0.0081 0.0528

1.03 1.50 2.35 2.15 1.04 2.71 2.92 1.00 0.93 1.60 3.32 0.65 2.01 2.11 0.57 0.90 1.67 0.21 0.01 1.29 2.84 1.77 1.55

Maximizing at y=

0.21 0.23 0.19 0.19 0.23 0.15 0.15 0.25 0.25 0.21 0.14 0.25 0.18 0.18 0.23 0.25 0.21 0.31 0.43 0.23 0.13 0.17 0.22

193

Broad phenotypic model (BPI,BPII,SAM,RUP) GENEHUNTER

Two-point

NPL

(P)

LOD score

1.54 1.77 1.67 1.90 1.73 1.62 2.14 2.01 1.87 2.19 2.64 2.39 2.26 1.69 1.29 1.47 2.02 1.38 1.07 0.72 0.73 1.74 0.85

0.0663 0.0432 0.0521 0.0338 0.0465 0.0575 0.0207 0.0269 0.0358 0.0184 0.0067 0.0119 0.0159 0.0501 0.1009 0.0752 0.0266 0.0873 0.1421 0.2310 0.2268 0.0462 0.1943

2.25 2.13 1.47 3.21 0.88 0.43 1.66 0.86 0.88 1.28 1.93 0.45 0.99 0.58 0.40 0.28 1.18 0.06 0.00 0.26 1.40 0.87 1.41

Maximizing at y=

0.14 0.22 0.22 0.17 0.25 0.28 0.20 0.28 0.25 0.22 0.20 0.28 0.25 0.28 0.31 0.31 0.25 0.37 0.50 0.31 0.19 0.22 0.25

Marker positions are in centiMorgans from the Marshfield linkage maps (in parentheses when not actually present in the database). NPL scores, significance levels (GENEHUNTER), two-point LOD scores, and their corresponding recombination fractions are shown. Scores reaching suggestive levels according to Lander and Kruglyak 20 are in bold and the genomewide significant result is underlined.

these analyses, the 1-LOD interval from the peak score of 2.49 is 3.5 cM (2.3 Mb) between markers D8S1835 and D8S1746. Family-based association analysis between all markers on chromosome 8 and the disease did not provide any positive results.

Discussion The present study substantially strengthens the evidence for a susceptibility locus on chromosome 8q24. We have added 19 markers on 8q24 increasing the marker density to 0.8 cM around 8q24. When using parametric analysis and the same model that gave the strongest signal in the study of Cichon et al,8 we detected linkage evidence reaching genomewide significance. Our highest score is with D8S256, a marker located 18 cM (10.7 Mb) telomeric to their peak marker D8S514. Our finding maximizes at W ¼ 0.14 and the finding of Cichon et al maximizes at W ¼ 0.05. These recombination fractions may reflect the effects of intra- and inter-familial heterogeneity and the differences in informativeness of specific families. Precise localization awaits further analysis.

Of the markers reported in Cichon et al,8 we have data for only one locus, D8S284 (see Table 1) for which they report a two-point LOD score of 1.3 (W ¼ 0.15) compared to our 1.5 (W ¼ 0.23). It must be noted that Cichon et al8 used BPI as an affected phenotype, while we include BPII and SAM as affected in our narrow phenotypic model. Coding only BPI patients as affected in our sample results in a three-fold reduction in information followed by the same reduction in scores (LOD ¼ 1.12), without significant changes in localization. One might find the great difference in LOD score in the parametric analysis compared to the original nonparametric analysis surprising. However, if the true mode of inheritance is dominant, the allele sharing assessed by nonparametric linkage is not expected to be as increased as in a recessive model. Additionally, information about the specific alleles that segregate in affected family members is underutilized by allele-sharing methods. This is particularly important with larger pedigrees and dominant diseases. The increased power of two-point parametric linkage is a subject that has long been argued in the literature.21–25 Molecular Psychiatry


194

Table 2 Two-point and multipoint parametric analysis using a model assuming a lower relative risk for a more common allele Narrow phenotypic model (BPI,BPII,SAM)

Broad phenotypic model (BPI,BPII,SAM,RUP)

Marker

Position (cM)

Two-point LOD

Maximizing at y=

HLOD

Alpha

Two-point LOD

Maximizing at y=

HLOD

Alpha

D8S1782 D8S284 D8S1765 D8S557 D8S1835 D8S558 D8S1740 D8S378 HC83REP D8S529 D8S256 D8S1708 D8S1796 D8S1746 D8S1462 D8S1710 D8S554 D8S1466 D8S1111 D8S1100 D8S274 D8S272 D8S523

142.75 143.82 144.36 145.97 145.97 145.97 147.04 147.04 (147.04) 148.12 148.12 148.12 148.12 149.46 (149.96) 150.8 153.48 (154.71) (155.11) 154.02 154.02 154.02 154.02

0.91 1.31 2.38 2.09 0.95 2.32 2.83 0.97 0.80 1.52 3.00 0.65 1.83 1.80 0.34 0.87 1.77 0.27 0.00 0.88 2.17 1.47 1.33

0.09 0.13 0.05 0.06 0.13 0.03 0.00 0.13 0.13 0.08 0.02 0.14 0.05 0.07 0.18 0.14 0.08 0.20 0.50 0.15 0.02 0.04 0.11

0.99 1.30 1.29 1.41 1.48 1.64 2.15 2.10 1.93 2.41 2.39 1.83 1.70 1.34 0.92 1.20 1.52 0.97 0.81 0.72 0.74 1.29 0.75

0.45 0.50 0.49 0.55 0.57 0.60 0.66 0.65 0.62 0.74 0.76 0.64 0.61 0.52 0.44 0.50 0.57 0.47 0.43 0.40 0.41 0.46 0.41

2.26 2.31 1.73 3.20 1.01 0.60 1.79 0.91 0.97 1.38 2.19 0.44 1.06 0.76 0.34 0.41 1.33 0.00 0.11 0.26 1.19 0.94 1.42

0.08 0.16 0.18 0.11 0.20 0.22 0.14 0.24 0.19 0.17 0.14 0.25 0.20 0.22 0.29 0.28 0.19 0.50 0.35 0.28 0.15 0.17 0.19

0.77 0.89 0.78 0.87 0.73 0.75 1.06 0.97 0.97 1.11 1.45 0.98 0.86 0.49 0.28 0.41 0.60 0.33 0.24 0.15 0.23 0.79 0.25

0.25 0.26 0.24 0.26 0.23 0.23 0.28 0.27 0.26 0.29 0.35 0.28 0.26 0.18 0.13 0.16 0.20 0.14 0.11 0.09 0.11 0.17 0.12

The highest multipoint score was 2.49 between markers D8S529 and D8S256 (in bold) for the narrow phenotypic model.

Figure 2 Chromosome 8–multipoint HLOD–multipoint heterogeneity LOD score using our low relative risk model, vs position on the chromosome 8 map (Marshfield). The thicker line corresponds to the narrow and the finer line to the broad phenotypic model. The location of the peak finding of Cichon et al is indicated by an arrow.

As opposed to the two-point, parametric multipoint analysis using the Cichon model failed to give a comparably high LOD score. Two-point analysis is much more robust to model misspecification, as it can account for phenocopies and heterogeneity by allowing for increased recombination. In multipoint analysis, however, model misspecification is more likely to produce lower LOD scores.17 This is especially true for models with low phenocopy rates and disease alleles with low frequencies and high penetrance, because affected individuals not carrying the disease allele will be considered double recombinants and assigned Molecular Psychiatry

extremely low likelihoods. The model of Cichon et al corresponds to a very high genotype relative risk, unlike what is suggested by the ‘common disease– common variant’ hypothesis.18 When we used a model based on this hypothesis, with a frequent allele giving a moderate relative risk and being involved in about half of the disease cases in the population (phenocopies ¼ 12 of prevalence), we observed multipoint results comparable to the two-point results in our narrow phenotypic model. A slight decrease observed in the two-point results (accompanied however by lower thetas, see Table 2) probably suggests that our model is still not perfect and two-point analyses perform better when allowed to account for model misspecifications via recombination. The lack of a positive association finding does not imply the absence of an associated haplotype. Although our high marker density prompted us perform the test for association, the marker spacing still far exceeds the average extent of linkage disequilibrium in the genome26 and a negative finding was the most likely outcome. A study with a high density using the more abundant single nucleotide polymorphisms would be more likely to capture any existing disease-associated haplotypes. Examination of the region in the Refseq database27 showed 11 known genes within the 1-LOD interval. We have identified three genes that are arguably


strong candidates for further study in BP disorder. The gene for thyroglobulin (TG; OMIM: 188450) is located exactly under our peak. The gene KCNQ3 (OMIM: 602232), coding for a voltage-gated potassium channel, is located further towards the centromere at the location of marker D8S558. Just outside the 1-LOD interval in the direction of the finding of Cichon et al is the gene for brain adenyl-cyclase (ADCY8; OMIM: 103070) close to D8S1765. TG is involved in the production, sequestration, and release of thyroid hormones. Thyroid hormones are often useful adjuvants in the treatment of severe refractory BP disorder28 and lithium, one of the most effective treatments for BP disorder, can alter thyroid function and TG structure.29 Additionally, studies have reported increased antithyroid antibodies in BP disorder,30–32 although results are inconsistent.33–35 TG is one of the targets of such antibodies. Interestingly, TG, a gene with over 50 exons, has several coding polymorphisms, 10 of which result in aminoacid changes.36 These are logical starting points for the investigation of this gene in BP disorder. KCNQ3 codes for a voltage-gated potassium channel. KCNQ3 heteromers with other KCNQ family members are responsible for the M-current, which is involved in determining the excitability of neurons and their responsiveness to synaptic input.37 Mutations in KCNQ3 cause epilepsy;38 there is no known association with BP disorder. However, anticonvulsants are a common and successful treatment for BP disorder, and it is possible that subtle abnormalities and/or specific neuroanatomical variations of gene expression could be involved in the susceptibility to BP disorder. Finally ADCY8, although located just outside our region, is a very interesting candidate. It codes for a brain adenyl-cyclase that catalyzes the formation of cyclic AMP (cAMP) from ATP. cAMP is a second messenger involved in the regulation of receptors, enzymes, ion channels, and gene transcription. There are multiple studies suggesting its involvement in BP disorder, both because of alterations in its levels and activity39–42 and because it is a target for lithium and other mood stabilizing agents.43,44 In summary, we present evidence that, taken together with independent findings by other investigators, strongly suggests the presence of a gene for BP disorder on chromosome 8q24. Furthermore, we make an argument for more appropriate models in multipoint linkage of complex disorders that consider the likelihood of the existence of common variants with low penetrance in these diseases. We also discuss three candidate genes, TG, KCNQ3, and ADCY8, that we believe deserve further attention. More work is needed in order to more precisely localize the linked locus and to examine the candidates, which reside in the region. We believe that these results are very promising, warranting intense investigation of 8q24, as there are now two independent investigative teams with evidence of genomewide significance for a susceptibility gene in this region.

195

Acknowledgements This work was supported by grants from the National Institute of Mental Health, the Stanley Medical Research Institute, the National Alliance for Research on Schizophrenia and Depression, the Alex. Brown & Sons Charitable Foundation, and the Charles A. Dana Foundation program ‘The Dana Consortium on the Genetic Basis of Manic Depressive Illness’. We would like to thank Dr Francis J McMahon and Dr Sylvia G Simpson for their participation in the clinical evaluations and Tyra Diggs for providing technical support.

References 1 Goodwin FK, Jamison KR. 1990. Manic Depressive Illness. Oxford University Press: New York. 2 Gershon ES, Hamovit J, Guroff JJ, Nurnberger JI, Goldin LR, Bunney WE. A family study of schizoaffective, bipolar I, bipolar II, unipolar and normal control probands. Arch Gen Psychiatry 1982; 39: 1157–1167. 3 Bertelsen A, Harvald B, Hauge M. A Danish twin study of manic depressive disorders. Br J Psychiatry 1977; 130: 330–351. 4 Mendlewicz J, Rainer JD. Adoption study supporting genetic transmission in manic-depressive illness. Nature 1977; 268: 327– 329. 5 Risch NJ, Botstein D. A manic depressive history. Nat Genet 1996; 12: 351–353. 6 Berrettini WH. Genetics of psychiatric disease. Annu Rev Med 2000; 51: 465–479. 7 Merikangas KR, Chakravarti A, Moldin SO, Araj H, Blangero JC, Burmeister M et al. Future of genetics of mood disorders research. Biol Psychiatry 2002; 52: 457–477. 8 Cichon S, Schumacher J, Muller DJ, Hurter M, Windemuth C, Strauch K et al. A genome screen for genes predisposing to bipolar affective disorder detects a new susceptibility locus on 8q. Hum Mol Genet 2001; 10: 2933–2944. 9 Friddle C, Koskela R, Ranade K, Hebert J, Cargill M, Clark CD et al. Full-genome scan for linkage in 50 families segregating the bipolar affective disease phenotype. Am J Hum Genet 2000; 66: 205–215. 10 McInnis MG, Lan T-H, Willour VL, Zandi PP, McMahon FJ, Simpson SG et al. Genome-wide scan of bipolar disorder in 65 pedigrees: supportive evidence for linkage at 8q24, 18q22, 4q32, 2p13, and 13q14. Mol Psychiatry 2003; 8: 288–298. 11 Endicott J, Spitzer RL. A diagnostic interview: the schedule for affective disorders and schizophrenia. Arch Gen Psychiatry 1978; 35: 837–844. 12 Kong A, Gudbjartsson DF, Sainz J, Jonsdottir GM, Gudjonsson SA, Richardsson B et al. A high-resolution recombination map of the human genome. Nat Genet 2002; 31: 241–247. 13 Lathrop GM, Lalouel JM, Julier C, Ott J. Strategies for multilocus linkage analysis in humans. Proc Natl Acad Sci USA 1984; 81: 3443–3446. 14 Kong A, Cox NJ. Allele-sharing models FLOD scores and accurate linkage tests. Am J Hum Genet 1997; 61: 1179–1188. 15 Kruglyak L, Daly MJ, Reeve-Daly MP, Lander ES. Parametric and nonparametric linkage analysis: a unified multipoint approach. Am J Hum Genet 1996; 58: 1347–1363. 16 Cottingham Jr RW, Idury RM, Schaffer AA. Faster sequential genetic linkage computations. Am J Hum Genet 1993; 53: 252–263. 17 Risch N, Giuffra L. Model misspecification and multipoint linkage analysis. Hum Hered 1992; 42: 77–92. 18 Risch N, Merikangas K. The future of genetic studies of complex human diseases. Science 1996; 273: 1516–1517. 19 Rabinowitz D, Laird N. A unified approach to adjusting association tests for population admixture with arbitrary pedigree structure and arbitrary missing marker information. Hum Hered 2000; 50: 211–223. 20 Lander ES, Kruglyak L. Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results. Nat Genet 1995; 11: 241–247. Molecular Psychiatry


196 21 Goldgar DE. Major strengths and weaknesses of model-free methods. Adv Genet 2001; 42: 241–251. 22 Sham PC, Lin MW, Zhao JH, Curtis D. Power comparison of parametric and nonparametric linkage tests in small pedigrees. Am J Hum Genet 2000; 66: 1661–1668. 23 Goldin LR, Weeks DE. Two-locus models of disease: comparison of likelihood and nonparametric linkage methods. Am J Hum Genet 1993; 53: 908–915. 24 Abreu PC, Greenberg DA, Hodge SE. Direct power comparisons between simple LOD scores and NPL scores for linkage analysis in complex diseases. Am J Hum Genet 1999; 65: 847–857. 25 Durner M, Vieland VJ, Greenberg DA. Further evidence for the increased power of LOD scores compared with nonparametric methods. Am J Hum Genet 1999; 64: 281–289. 26 Gabriel SB, Schaffner SF, Nguyen H, Moore JM, Roy J, Blumenstiel B et al. The structure of haplotype blocks in the human genome. Science 2002; 296: 2225–2229. 27 Maglott DR, Katz KS, Sicotte H, Pruitt KD. NCBI’s LocusLink and RefSeq. Nucleic Acids Res 2000; 28: 126–128. 28 Baumgartner A, Bauer M, Hellweg R. Treatment of intractable nonrapid cycling bipolar affective disorder with high-dose thyroxine: an open clinical trial. Neuropsychopharmacology 1994; 10: 183–189. 29 Lazarus JH. The effects of lithium therapy on thyroid and thyrotropin-releasing hormone. Thyroid 1998; 8: 909–913. 30 Haggerty JJ, Jr., Evans DL, Golden RN, Pedersen CA, Simon JS, Nemeroff CB. The presence of antithyroid antibodies in patients with affective and nonaffective psychiatric disorders. Biol Psychiatry 1990; 27: 51–60. 31 Haggerty JJ, Jr., Silva SG, Marquardt M, Mason GA, Chang HY, Evans DL et al. Prevalence of antithyroid antibodies in mood disorders. Depress Anxiety 1997; 5: 91–96. 32 Wilson R, McKillop JH, Crocket GT, Pearson C, Jenkins C, Burns F et al. The effect of lithium therapy on parameters thought to be involved in the development of autoimmune thyroid disease. Clin Endocrinol (Oxf) 1991; 34: 357–361. 33 Bartalena L, Pellegrini L, Meschi M, Antonangeli L, Bogazzi F, Dell’Osso L et al. Evaluation of thyroid function in patients with

Molecular Psychiatry

34

35 36

37 38

39 40

41

42

43

44

rapid-cycling and non-rapid-cycling bipolar disorder. Psychiatry Res 1990; 34: 13–17. Hornig M, Amsterdam JD, Kamoun M, Goodman DB. Autoantibody disturbances in affective disorders: a function of age and gender? J Affect Disord 1999; 55: 29–37. Rapaport MH. Immune parameters in euthymic bipolar patients and normal volunteers. J Affect Disord 1994; 32: 149–156. van de Graaf SA, Ris-Stalpers C, Pauws E, Mendive FM, Targovnik HM, de Vijlder JJ. Up to date with human thyroglobulin. J Endocrinol 2001; 170: 307–321. Jentsch TJ. Neuronal KCNQ potassium channels: physiology and role in disease. Nat Rev Neurosci 2000; 1: 21–30. Charlier C, Singh NA, Ryan SG, Lewis TB, Reus BE, Leach RJ et al. A pore mutation in a novel KQT-like potassium channel gene in an idiopathic epilepsy family. Nat Genet 1998; 18: 53–55. Abdulla YH, Hamadah K. 30 ,50 cyclic adenosine monophosphate in depression and mania. Lancet 1970; 1: 378–381. Young LT, Li PP, Kish SJ, Siu KP, Kamble A, Hornykiewicz O et al. Cerebral cortex Gs alpha protein levels and forskolin-stimulated cyclic AMP formation are increased in bipolar affective disorder. J Neurochem 1993; 61: 890–898. Rahman S, Li PP, Young LT, Kofman O, Kish SJ, Warsh JJ. Reduced [3H]cyclic AMP binding in postmortem brain from subjects with bipolar affective disorder. J Neurochem 1997; 68: 297–304. Fields A, Li PP, Kish SJ, Warsh JJ. Increased cyclic AMPdependent protein kinase activity in postmortem brain from patients with bipolar affective disorder. J Neurochem 1999; 73: 1704–1710. Mork A. Actions of lithium on the cyclic AMP signalling system in various regions of the brainFpossible relations to its psychotropic actions. A study on the adenylate cyclase in rat cerebral cortex, corpus striatum and hippocampus. Pharmacol Toxicol 1993; 73(Suppl 3): 1–47. Stewart RJ, Chen B, Dowlatshahi D, MacQueen GM, Young LT. Abnormalities in the cAMP signaling pathway in post-mortem brain tissue from the Stanley Neuropathology Consortium. Brain Res Bull 2001; 55: 625–629.