sign in sign up
<<
Home , Chromosome 18, Genetic linkage

Molecular Psychiatry (1997) 2, 200–210  1997 Stockton Press All rights reserved 1359–4184/97 $12.00

PERSPECTIVE Genetic linkage and bipolar affective disorder: progress and pitfalls M Baron

Department of Psychiatry, Columbia University College of Physicians and Surgeons, and Department of Medical , New York State Psychiatric Institute, 722 West 168th Street, New York, NY, USA

The history of linkage studies in bipolar affective disorder is a convoluted affair punctuated by upswings and setbacks, hope and skepticism. Advances in genomics and statistical tech- niques, and the availability of well-characterized clinical samples, have bolstered the search for disease , leading to a new crop of findings. Indeed, recent reports of putative loci on 18, 21, X, 4, 6, 13 and 15 have rekindled a sense of optimism. The new findings are reviewed and scrutinized, with implications for future research. Keywords: molecular genetics; DNA markers; genes; linkage analysis;

Linkage studies of bipolar affective disorder (manic nostic procedures and in statistical techniques appli- depression) have charted an unsteady course marred cable to complex traits.4–15 The availability of dense by fits and starts.1 Findings that appeared at first unas- genomic maps and suitable clinical samples is an sailable could not be replicated or faltered on further added asset. Indeed, the emergence of promising new scrutiny of the evidence. The highs and lows of this findings in the last 2 years may signal the turning of decades-long search were aptly dubbed ‘a manic the tide, though the problems posed for geneticists are depressive history’:2 bouts of optimism alternating far from over. In this article I review and critique the with dashed expectations. This predicament is not new findings, with an eye to future research. unique: in the wake of conflicting results and reversals of earlier findings, schizophrenia was nicknamed ‘a New findings graveyard for molecular geneticists’.3 Other complex disorders, which bear likeness to psychiatric disorders Several chromosomal regions have recently been by way of genetic and phenotypic uncertainties, were implicated in bipolar disorder: 18p16,17; 18q16–18; 21q19– similarly afflicted: some of the initial linkage findings 21;Xq22;4p23; and 6p, 13q and 15q.24 Altogether, eight in diabetes mellitus, Alzheimer’s disease, and breast putative loci are invoked in the latest spate of link- were not compelling or readily interpretable, age reports. although subsequent studies firmed up the evidence. For the most part, the ill-fated findings in bipolar dis- 18 order predate the molecular genetics era. This may be Berrettini et al16 reported evidence of linkage to the partly attributable to the low information content of pericentromeric region of . They con- classical markers and the dearth of polymorphisms that ducted a genomic scan in 22 North-American extended can serve to verify linkage assignments. But a host of pedigrees comprised of 365 individuals. Affected sib- other factors must be considered including the com- pair (ASP) analysis of 11 markers suggested linkage to plex inheritance of the disorder, leading to diminished the D18S21 (P = 0.0007); the affected-pedigree- power and difficulties in discerning true from false- member (APM) method also suggested linkage with positive findings; the involved procedures for data col- multilocus analysis of five pericentromeric markers lection and processing, causing variability among stud- (P Ͻ 0.0001 and P = 0.0007, depending on the weight- ies and proneness to error; selection bias; statistical ing function of frequencies). Lod score analysis artifacts, and misinterpretation of data.1,2,4–14 The did not yield statistically significant evidence of link- growing awareness of potential pitfalls has led to age, though modest lod scores were observed in some methodological advances in ascertainment and diag- pedigrees. The strongest results were obtained with a disease comprised of bipolar I and II, schi- zoaffective, and recurrent major depression. Correspondence: M Baron, Department of Psychiatry, Columbia In an attempt to replicate the Berrettini et al finding, University College of Physicians and Surgeons, and Department Stine et al17 studied 28 North-American nuclear famil- of Medical Genetics, New York State Psychiatric Institute, 722 West 168th Street, New York, NY, USA. E-mail: mb17Ȱcolum- ies consisting of 243 individuals, using 31 markers. bia.edu ASP analysis indicated excess allele sharing for 18p Received 17 October 1996; accepted 18 October 1996 markers, especially at D18S37 (P = 0.0003); addition- Genetic linkage and bipolar affective disorder M Baron 201 ally, excess sharing of paternally, but not maternally, evidence for linkage with either parametric or nonpara- transmitted was observed for three markers on metric (APM and ASP) methods, though suggestive evi- 18q, especially at D18S41 (P = 0.0004). The evidence dence was obtained for some loci. Subdividing the for linkage to loci on both 18p and 18q was strongest pedigrees according to maternal versus non-maternal in the paternal pedigrees (those in which the illness transmission (the approach used by Stine et al17 and appeared to be transmitted from the father’s side of the Gershon et al25 to produce the strongest evidence for pedigree), in particular at D18S41 (P = 0.00002 using linkage) did not alter the pattern of results. ASP analysis, and a lod score of 3.51 at ␪ = 0.0). The positive results were particularly pronounced using an affected phenotype similar to the one used by Berret- Straub et al19 reported a possible susceptibility locus tini et al.16 Stine et al concluded that their results con- at 21q22.3. They performed a preliminary scan firm Berrettini et al’s findings and, in addition, provide in 47 extended pedigrees obtained in the US and Israel support for a parent-of-origin effect. Subsequently, Ger- comprised of 937 genotyped individuals. A lod score shon et al25 reanalyzed the Berrettini et al16 pedigrees of 3.41 was observed at the PFKL locus; the lod score for parent-of-origin effect. Using ASP analysis, linkage was robust to marker allele frequencies, phenocopy to 18p- markers was observed in ‘mixed’ rates, age-dependent penetrance, and changes in affec- paternal-maternal pedigrees, but not the maternal pedi- tion status (sensitivity analysis). Fourteen other mark- grees, especially at D18S32 (P Ͻ 0.00001), D18S56 ers in 21q22.3 were examined in this family with larg- (P = 0.00009) and D18S21 (P = 0.0007). Gershon et al25 ely positive lod score. Five other families also showed concluded that their results are consistent with those positive, though modest lod values with PFKL. The of Stine et al.17 maximum lod score for the entire pedigree series was Finally, Freimer et al18 found evidence for a bipolar 2.8. Extended sib-pair analysis (ESPA) did not yield a locus at 18q23 in two Costa Rican pedigrees obtained statistically significant result, but the single locus and from a genetically isolated population. The evidence, multilocus (PFKL and D21S171) APM method gave sig- obtained after a systematic genome scan had been com- nificant P-values (maximum value: Ͻ10−6). The strong- pleted, was primarily based on the sharing of marker est results were obtained with an ‘intermediate’ pheno- by individuals affected with bipolar I dis- type definition encompassing bipolar I and II, order, the narrowly defined phenotype of the illness. schizoaffective-manic, and recurrent major depression. Altogether, 14 markers spanning the 18q21–23 region Gurling et al20 studied 23 extended pedigrees (17 were studied. Association analyses yielded lod scores English; six Icelandic) comprised of 278 individuals. of 3.7 and 4.06 (both at ␪ = 0.0) at D18S554 and They found a three-point lod score of 1.33, with PFKL D18S70, respectively. Overlap with the putative per- and D21S171. In addition, an overall lod score of 3.58 icentromeric region implicated by Berrettini et al16 was was observed when an oligogenic (two-locus) model excluded; overlap with the location proposed by Stine was used to analyze PFKL/D21S171 and the tyrosine et al17 did not appear likely but could not be excluded. hydroxylase (TH) locus on . (TH itself Other studies failed to turn statistically significant showed a lod score of 1.43 in these families.) ASP linkage to chromosome 18 markers. Maier et al26 stud- analysis was also suggestive of linkage with D21S171 ied six pericentromeric markers in five extended Ger- (P = 0.001). The largest lod scores and highest statisti- man families. Linkage was excluded using both lod cal significance were observed using a narrowly score and APM methods. Pauls et al27 examined four defined phenotype consisting of bipolar I and II. pericentromeric markers in three large Old Order Another set of results supportive of linkage to 21q Amish kindreds obtained from a homogeneous sample. was reported by Detera-Wadleigh et al21 who analyzed They excluded linkage to this region using both lod 22 extended North-American pedigrees consisting of score and ASP analyses. Debruyn et al28 examined 14 365 individuals. Using 18 markers on 21q, single-locus markers spanning the 18p11–18q23 region in two large ASP analysis detected a high proportion of alleles Belgian families. Using lod score analysis, there was shared identical by descent at nine loci (P = 0.049– no significant evidence for linkage; modest lod scores 0.0008). Multilocus analyses revealed locus trios with were observed for some markers in one family. Simi- excess allele sharing in: 1) the distal region between larly, ASP analyses did not yield statistically signifi- D21S270 and D21S171 (P Ͻ 0.01); and 2) a more proxi- cant results save one locus: D18S51 (P = 0.0007), but mal interval spanned by D21S1436 and D21S65 absent multilocus analysis the significance of this (P = 0.03–0.0003). There was, in addition, a cluster of result cannot be evaluated. Coon et al29 studied six markers with positive, though not statistically signifi- North-American pedigrees with 21 markers spanning cant lod scores for the pedigree sample as a whole. The nearly the entire length of chromosome 18. There was positive results were obtained using either a narrowly no statistically significant linkage to any region using defined phenotype (bipolar I and II and schizoaffective either lod score analysis or nonparametric methods disorder) or a broader definition similar to the one (APM and ASP). In the largest sample studied to date, used by Straub et al19 (including recurrent major our group (Rao et al, in preparation) examined 10 mar- depression). kers spanning the pericentromeric region in 53 pedi- There were two negative reports, however. Byerley et grees (39 North-American, 14 Israeli) comprised of al30 studied six large North-American pedigrees using 1013 genotyped individuals. There was no compelling seven markers, including a for PFKL. Genetic linkage and bipolar affective disorder M Baron 202 They did not find evidence of linkage with either lod transmitted to other affected individuals in the pedi- score analysis or the ASP method. Ewald et al31 ana- gree. Linkage to the dopamine D5 receptor (DRD5) lyzed two Danish families with 21q telomeric markers, locus, a potential candidate at 4p16, did not seem including PFKL. They observed small positive lod likely in this pedigree. Other potential candidates on scores, but significant evidence was lacking; they did 4p, the GABA receptor GABRB1 and the region harbor- not perform nonparametric analysis, however. ing the genes for Wolfram Syndrome, were also excluded. The X-chromosome A further 11 families were typed using D4S394 and In a study of an extended Finnish pedigree obtained in all families combined there was evidence of linkage from a genetically isolated population, Pekkarinen et with heterogeneity with a maximum lod score of 4.1 al22 examined 25 markers spanning the Xq24–q28 and an estimated 30–35% of these families linked to region, demonstrating linkage between bipolar disorder this marker. and a distinct chromosomal on Xq24–q27.1. The largest lod score was 3.54 (␪ = 0.0) at the DXS994 Chromosomes 6, 13 and 15 locus. Positive, though more modest, lod values were Ginns et al24 conducted a genome-wide search, with an observed at nearby loci. Analyses with markers outside average spacing of 5–10 cM, in a large Old Order the Xq24–27.1 region resulted in negative lod scores. Amish pedigree complex consisting of 205 individuals. When haplotypes were constructed for Xq24–q27.1 They considered further regions which yielded poss- markers, all the family members with the narrowly ible hints of linkage. Markers at three loci gave P- defined phenotype (bipolar or schizoaffective disorder) values of 0.0001 or smaller using ASP analysis were found to carry the same haplotype, reducing the (SIBPAL program): D6S7 (chromosome 6pter-p24); possibility of false linkage due to undetected recombi- D13S1 (chromosome 13q13), and D15S45 nants. The largest lod scores were obtained under a (chromosome 15q11-qter), all with P = 0.0003. How- stringent phenotypic definition (bipolar I, II and not ever, ASP analysis weighted by a number of meioses otherwise specified, and schizoaffective-manic). gave marginally significant results: maximum P- Because no obligatory recombination events were values = 0.029, 0.0057, and 0.015 for the three markers, observed between the disease and the HPRT locus in respectively. Using lod score analysis, none of the mar- Xq26, the of the gene was sequenced in kers reached a lod score of 3, though positive lod scores one affected female and in one unaffected male, with (range: 0.86–2.47) were observed for some of the mod- no nucleotide variation when the sequences were com- els tested. There was no evidence of linkage disequilib- pared with the published CDNA sequence. rium or specific haplotypes with these markers, nor Additionally, the investigators reanalyzed pre- was there significant evidence for allele sharing ident- viously published data on Xq26–27 markers in bipolar ical-by-descent (IBD) among all possible sib-pairs. The pedigrees,32–37 using uniform diagnostic criteria and results overall were stronger for bipolar I disorder than inheritance models, and found suggestive evidence for for more broadly defined diagnostic categories. linkage in the combined analysis: the maximum poly- locus lod scores for markers DXS51, F9a, and F9b were Critique 2.78, 1.51, and 1.77, respectively. They concluded that the combined data support linkage to the Xq26 inter- The studies reviewed in the preceding section are more val, and that the evidence for linkage diminishes in advanced than earlier attempts by way of diagnostic regions distal to F9. and statistical methods, and the availability of densely mapped highly polymorphic DNA markers. In parti- cular, the use of rigorous standards for data collection Blackwood et al23 studied a large multigeneration and clinical assessment; statistical simulations to Scottish pedigree comprised of 120 individuals. They determine power and to proffer a meaningful interpret- performed a genome-wide search with markers spaced ation of results; parametric and nonparametric at about 40 cM intervals. A positive lod score at methods to address the complex mode of inheritance; D4S230 was followed up by typing an additional 17 the analysis of multiple markers in chromosomal markers mapping to chromosome 4p16. Eleven mark- regions of interest to verify linkage assignments; con- ers gave positive lod scores, with the highest lod score servative definitions of the clinical phenotype to cur- at D4S394: 4.09 at ␪ = 0.0. The positive lod scores were tail phenotypic misclassifications, and reporting of robust to changes in allele frequencies and affection results in the broader context of a genome screen, thus status (sensitivity analysis). The nonparametric allowing a distinction between background noise and extended relative pair analysis showed an increase in findings worth pursuing. Also of note is the apparent allele sharing, but the P-values fell short of suggestive genetic homogeneity in two of the populations,18,24 linkage. Three-point analyses with adjacent markers augmenting the prospects of gave a maximum lod score of 4.8 in the region D4S431– and positional cloning in the event of genuine linkage. D4S403. In all analyses lod scores were greater when On the other hand, uncertainties abound in several the phenotype was defined narrowly (bipolar I and II). areas: statistical significance; robustness of results; Haplotypes involving the seven markers in that region transmission patterns; genetic distances; and repli- were consistent with a haplotype in the founder being cation. Genetic linkage and bipolar affective disorder M Baron 203 Statistical significance reversed. The lod scores reported by Freimer et al18 at To stem ‘a flood of false positive claims’ in linkage D18S554 and D18S70 were statistically significant 13 = studies of complex traits, Lander and Kruglyak pro- (Zmax 3.7 and 4.06, respectively), with more modest posed guidelines for the interpretation of significance results for adjacent markers. levels. The proposed guidelines have taken hold, though opinions vary.38,39 They distinguish between Chromosome 21 The lod score reported by Straub et 19 = significant linkage (statistical evidence expected to al for their large pedigree (Zmax 3.41) was consistent occur 5% of the time in a genome scan) and suggestive with significant linkage; the lod score for the entire ser- = linkage (statistical evidence that would be expected to ies (Zmax 2.8) was suggestive of linkage. The single occur one time at random in a genome scan). In the locus APM results showed significant linkage case of lod score analysis, the two categories would (P Ͻ 10−6) with PFKL using two weighting functions for correspond to lod scores of 3.3 and 1.9, respectively; allele frequencies, and suggestive linkage with a third in the case of sib-pair studies, the corresponding point- function. The multilocus APM results showed sugges- wise significance levels would be P = 0.000022 and tive or significant linkage with two of the weighting 0.00074. For studies involving a mixture of relative functions. The extended sib-pair was not supportive of types, as in APM analysis, a range can be used: linkage, but incomplete genotypic information on par- P = 0.00005–0.00001 and 0.005–0.0001 for significant ents complicates the interpretation of this result. The versus suggestive linkage. How do the recent findings two-locus lod score with PFKL and TH reported by 20 = fare in this regard? Gurling et al (Zmax 3.58) was consistent with sig- nificant linkage. (Lander and Kruglyak’s guidelines are Chromosome 18 The results reported by Berrettini et not explicit with respect to oligogenic scenarios, al16 did not surpass the threshold for significant link- however). Their ASP results with D21S171 (P = 0.001) age. One marker, D18S21, showed suggestive linkage fell short of suggestive linkage. Some of the loci on 21q using ASP analysis, but the authors’ use of SIBPAL, a analyzed by Detera-Wadleigh et al21 met the criterion nonconservative method with no weighting for sibship for suggestive linkage using multilocus ASP analysis size, might have produced inflated significance lev- (P = 0.0003–0.0001). els.12,24,40 Also, other markers in the implicated region fell short of suggestive linkage and the authors did not X-chromosome The highest lod score reported by present a multilocus analysis. The multilocus APM Pekkarinen et al22 for their large Finnish pedigree = results with two of the weighting functions were con- exceeded the significance threshold (Zmax 3.54). The sistent with suggestive linkage in a region proximal to combined analysis of previously published data on = D18S21; the single locus results showed no indication Xq26–27 showed suggestive linkage at F9 (Zmax 2.78); of linkage. When reanalyzed by Gershon et al25 accord- more distal markers were not supportive of linkage. ing to parent-of-origin effect, the ASP results reached statistical significance at D18S32 and were suggestive Chromosome 4 The lod scores for the large pedigree 23 = of linkage at D18S56 and D18S21 in non-maternal (but in the Blackwood et al study (two-point Zmax 4.09; = not maternal pedigrees); however, as with the Berret- multipoint Zmax 4.8) were consistent with significant tini et al16 study, the authors’ use of the SIBPAL pro- linkage in that pedigree, though the extended relative gram and the absence of multilocus ASP analysis puts pair analysis fell short of suggestive linkage. Hetero- limits on the interpretation of results. In Stine et al’s17 geneity testing in the entire series also indicated sig- = study, the ASP results supported suggestive linkage at nificant linkage (Zmax 4.1), though the result was larg- D18S37 in the sample as a whole and at D18S41 in ely due to a high lod score in the single large pedigree. paternally transmitted alleles; nearby markers showed no such evidence, however. In paternal (but not , 13 and 15 There was no evidence of maternal) pedigrees, D18S41 showed significant link- significant linkage. D6S7 yielded suggestive lod scores, age, and D18S64 and D18S38, two adjacent markers, and using SIBPAL for ASP analysis, all three markers showed suggestive linkage. However, the authors did (D6S7, D13S1 and D15S45) produced locus-specific P not present multilocus results for any of their ASP values consistent with suggestive linkage, but these analyses and, moreover, their use of SIBPAL for ASP values evaporated when the ASP analysis was analysis constrains the interpretation of the results (see weighted by the number of meioses. above). Lod scores for the sample as whole were not Although some of the results conformed to ‘signifi- suggestive of linkage, but the paternal pedigrees pro- cant’ linkage, they did not exceed the ‘significance’ duced a significant two-point lod score at D18S41 threshold by much. Consequently, a type 1 error (5% = (Zmax 3.51) and a suggestive score at D18S64 probability) is not discountable. Of course, the results = (Zmax 2.6); the multipoint lod score in these pedigrees showing only ‘suggestive’ linkage have a greater prone- = was also suggestive of linkage (Zmax 3.11). Also, there ness to error. is an apparent inconsistency between the Stine et al17 and Gershon et al25 analyses: in the former, D18S64, Robustness one of the pivotal markers on 18q, appeared linked to The robustness of linkage results can be gauged several the disease in paternal (but not maternal) pedigrees, ways, in particular with respect to multiple testing; whereas in the latter analysis the linkage pattern was phenotypic misclassification or diagnostic error; gen- Genetic linkage and bipolar affective disorder M Baron 204 etic parameters, and updating of pedigrees. All of these the pedigrees.36,44 None of the recent results have been issues can be pivotal to the interpretation of linkage subjected to such a scrutiny, but systematic follow-up results. studies are underway.45,46

Multiple testing Multiple testing can lead to inflation Transmission patterns of lod scores and type 1 errors. With 16 lod score per- Transmission patterns in pedigrees bear upon the link- (four diagnostic classes and four genetic age reports for chromosome 1816,17 and the X-chromo- models) and modest lod scores, the results for chromo- some.22 somes 6, 13 and 15 are particularly vulnerable: Ginns et al24 reported simulations showing only modest mul- Chromosome 18 It appears that linkage to the per- tiple test effects, but previous work advocated a more icentromeric and proximal q regions is restricted to conservative interpretation in similar circum- pedigrees with paternal, or non-maternal trans- stances.41,42 Thus, even a moderate deflation of lod mission.17,25 The investigators’ decision to divide their scores would have downgraded the suggestive linkage sample according to type of transmission stemmed reported in this study. The number of tests conducted from a parent-of-origin effect in the clinical data, in the other studies was too small to be consequential specifically, excess maternal transmission, a pattern in this regard. consistent with genomic imprinting or mitochondrial inheritance. The use of genetic-epidemiological argu- Phenotypic misclassification The impact of pheno- ments to pursue potential leads has its merits, but cau- typic ambiguities on linkage results, as measured by tion is advised for the following reasons: 1) there was sensitivity analysis,43 can be substantial. Sensitivity no prior statistical support for linkage heterogeneity in analysis varies the affection status of each individual the sample as a whole; 2) dividing nuclear families, in the pedigree and computes the impact of these the approach taken by Stine et al,17 according to the changes on the lod score. Large pedigrees are parti- presence or absence of maternal inheritance, does not cularly sensitive to changes in diagnostic status when convey the complete picture because of absent infor- such changes occur in key individuals, but the cumu- mation on previous generations. Indeed, when lative effect of many such errors in smaller families can extended pedigrees were examined in other also be considerable.43 All of the recent studies were samples,25,30 paternal pedigrees, which constituted based on moderate-to-large pedigrees, but only the nearly 40 percent of the Stine et al17 series, nearly van- studies involving chromosome 21 and 4 employed sen- ished in favor of pedigrees with mixed transmission. sitivity analysis, demonstrating the robustness of the This poses difficulties for the interpretation of results original observations.19,23 unless strict guidelines and uniform protocols for the extension and classification of pedigrees are adhered Genetic parameters Penetrance, allel frequencies, and to across studies; 3) as suggested,25 the parent-of-origin phenocopy rates can all impact the outcome of linkage effect may simply be an artifact stemming from differ- analysis. To varying degrees, these parameters were ential reporting of illness in maternal versus paternal examined in all the recent reports, with no substantive branches; differential gene-environment interaction effect on the magnitude of the lod score. However, in according to the sex of the transmitting parent, or a bias many of the pedigrees, the genotypic information was resulting from the small sample size. Indeed, Kato et incomplete due to untyped individuals, especially in al47 did not find a parent-of-origin effect in a series of the early generations (for example, deceased or other- families 10 times as large as either the Berrettini et al16 wise unavailable individuals). The presence of or the Stine et al17 samples. Also, as noted earlier, untyped individuals, even when some of the missing classifying pedigrees according to maternal versus non- information can be deduced from relatives’ , maternal transmission did not produce statistically sig- can bias the lod score results. Also, the APM method nificant linkage in a sample much larger than either of used in two of the studies16,19 is sensitive to misspeci- the studies claiming linkage for this chromosomal fication of marker allele frequencies. Estimating these region.30 frequencies from the data, the approach taken in these studies is a potential countermeasure, but the incom- X-chromosome As discussed elsewhere,48 although plete information on founders’ genotypes can still male-to-male transmission is an exclusion criterion in undermine the accuracy of the results. studies of X-linkage, and offspring of males are gener- ally less informative than those of females, the paucity Updating of pedigrees Follow-up studies to extend of male offspring of affected (and unaffected) males pedigrees and to examine changes in affection status raises the possibility that these subjects were not stud- over time (especially in unaffected individuals who are ied systematically. This could lead to an erroneous still within the risk period) can be crucial to the conclusion of linkage if pedigrees with positive lod interpretation of linkage results. Previously reported scores had unreported instances of male-to-male trans- linkages, including one claimed linkage in the popu- mission. In the pedigree reported by Pekkarinen et al,22 lation with the putative loci on chromosomes 6,13 and there was a clear excess of females, diminishing the 15 (the Older Order Amish studied by Ginns et al24), ability to detect male-to-male transmission. The alter- have been sharply attenuated following an updating of native explanation is the low reproductive rate of affec- Genetic linkage and bipolar affective disorder M Baron 205 ted males because of high mortality; however, there Replication were several unaffected males well into the repro- In conjunction with guidelines for the interpretation of duction period whose offspring were not reported. linkage results in complex traits, Lander and Krug- Also, two affected males had sons who, though judged lyak43 set forth criteria for credible replication. In parti- to be unaffected, were far from completing the risk per- cular, they proposed that the term ‘replication study’ iod. For these reasons, and as acknowledged by the should be reserved for situations in which significant investigators,22 the segregation pattern does not pro- linkage (by their definition, see Statistical significance vide unequivocal evidence of X-linked transmission, above) has already been obtained. Once a significant although X-linkage could still exist alongside other linkage is obtained in an initial study, and because rep- genetic and nongenetic factors that obscure a ‘pure’ lication involves testing an established prior hypoth- transmission pattern. esis, a P value of 0.01 should suffice to declare confir- mation at the 5% level. As noted earlier, replication has been attempted for Genetic distances the putative loci on chromosomes 18 and 21. Strictly In all datasets, the positive results spanned large gen- speaking, because Berrettini et al’s16 results did not etic distances, generally in the 20–50 cM range. This reach the threshold of significant linkage, the Stine et could be ascribed to uncertainties in pinpointing the al17 replication study was not a bona fide confirmation. map location of complex disease genes because of both The reanalysis by Gershon et al,25 with a focus on non- genetic heterogeneity and model misspecification. But maternal pedigrees, did produce significant linkage in in two instances, chromosome 18 and the X-chromo- Berrettini et al’s16 data, but as noted earlier (see Stat- some, there are other issues to consider. istical significance), there are uncertainties about the interpretation of results in the three datasets.16,17,25 The 18 Chromosome 18 The three datasets supportive of Freimer et al study should not be considered in this linkage16–18 implicate disparate chromosomal regions. context because of the aforementioned map inconsist- 16 17 Although Berrettini49 argued that the most simple encies with the Berrettini et al and Stine et al stud- 19 interpretation is that the various reports have identified ies. The Straub et al chromosome 21 finding appears the same locus, the strongest evidence of linkage in the somewhat stronger because some of the initial results Stine et al17 report was obtained at D18S41, 65 cM had the required statistical significance, and both repli- 20,21 away from D18S21, the locus with the most pro- cation studies produced P values consistent with nounced linkage in the Berrettini et al16 study; in fact, confirmation. Other studies yielded negative or 26–29 the two loci are on separate arms of the chromosome ambiguous results with either chromosome 18 or 30,31 altogether (p and q, respectively). The 95% confidence chromosome 21 markers, but most of the samples interval calculated by Stine et al17 covers both arms of were too small to furnish the statistical power required the chromosome, but the distance remains substantial. for replication. Also of interest in this regard is the X- 22 The candidate segment reported by Freimer et al18 in chromosome linkage finding of Pekkarinen et al. 18q does not overlap with either of these regions and, Although the strongest evidence of linkage was in fact, excludes the region implicated by Berrettini et obtained for Xq24–26, there was some overlap with al.16 It thus appears, that the various linkage claims Xq27, a region implicated in earlier linkage 32,34,35 cannot be readily reconciled, raising three possibilities: results. However, these earlier studies did not there are at least two separate loci on chromosome 18 utilize multiple DNA markers for verification and the 50 predisposing to the disease; there is only one such results were not statistically robust. Also, the analysis locus, or, as imputed by Risch and Botstein2 in their of the combined data at Xq27 did not produce signifi- 22 critique, given the map inconsistencies and other cant results and other studies with DNA markers in 33,36 uncertainties in the various studies, there are none. this region yielded negative findings. The positive results have been public knowledge since 1994 (longer, if one considers presentations at X-chromosome The region implicated by Pekkarinen scientific meetings prior to publication). It is possible, et al22 at Xq24–27 is 40–55 cM proximal to Xq28, a therefore, that other groups have made replication region thought to be linked to bipolar disorder in earl- attempts, obtained negative or equivocal results and ier studies using classical markers. As reviewed,50 opted not to publish, in which case the pattern of although subsequent studies with DNA markers have reported ‘replications’ departs from the real-life scen- not confirmed the earlier findings both in independent ario. samples and in a follow-up of one of the original ser- ies,36 the high lod scores obtained in some of the earlier Implications studies remain intriguing and may have to be con- sidered in the overall interpretation of the X-linkage The recent surge in linkage reports may well be the data. Specifically, if new findings point once again to harbinger of a new era in the quest for genes in bipolar Xq28 as a candidate region, a combined analysis of the disorder. The potential replication of some of the find- entire Xq24–28 region in a large dataset may help ings is particularly noteworthy in view of a long his- determine if there is more than one putative disease tory of failed attempts at confirming other claimed locus in this region. linkages. However, given the aforementioned uncer- Genetic linkage and bipolar affective disorder M Baron 206 tainties and pending further inquiry, it wold be prema- large effect, could still be detected by linkage analysis. ture to view any of the reported linkages as definitive. Instead, they advocate genomic association studies. The involved procedures for data collection, pro- Also noteworthy are simulations concerning repli- cessing and analysis, coupled with the ‘soft’ nature of cation. Suarez et al,60 simulating an oligogenic model behavioral , pose another complication with 50% heritability (a scenario commonly proposed which may not be readily tractable.14 For some time to for complex psychiatric traits), estimated that the num- come, researchers will continue to debate the meaning- ber of families required for first detection of linkage fulness and credibility of linkage results and the means with six quantitative trait loci (QTL) may be well over and prospects for translating a linkage finding into gene 100 with a population prevalence (ϰ) of 10% (and over discovery. Some of these issues are briefly discussed, 50 with ϰ = 1%); the number of families required for with special reference to the results reviewed earlier. replication would be far larger (over 700 with ϰ = 10%, The standards advocated earlier in this article, and about 250 with ϰ = 1%). Whatever approach is including the guidelines of Lander and Kruglyak13 for taken, the sample sizes projected by simulations for the interpretation and acceptance of linkage claims, linkage detection and replication are generally much may seem harsh, but in a field with a murky history larger than those used in the positive linkage and repli- they are useful ‘gatekeepers’ that can help guide scien- cation reports described earlier. This raises two possi- tific priorities and ensure that precious resources are bilities: 1) the simulated models are ‘omniscient’, not squandered in vain. Nevertheless, hints of linkage namely, they make oversimplified or otherwise that fail to satisfy stringent criteria need not be ignored; unrealistic assumptions and linkage detection and rep- they should be pursued even though most will prove lication will occur more often than predicted; 2) the to be false. Failure to report or pursue a potential true- models are correct in their predictions and the positive positive may relegate a (true) susceptibility locus to linkage results are either fortuitous or inaccurate. oblivion. This is especially true for genes with small Whatever the accuracy of a particular simulation, it is effects, where a true-positive cannot be readily dis- generally agreed that large datasets will be needed to tinguished from a false-positive due to complex genetic sort out the variety of gene effects that are likely to be mechanisms leading to low statistical power. A case in encountered. Sizeable samples are already in exist- point is insulin-dependent diabetes mellitus, a com- ence, or at various stages of completion, both in single plex trait often used as a genetic paradigm for psychi- centers and as a result of collaborative networks. atric disorders: regions that merely gave a hint of link- Another issue of interest concerns methods of analy- age in the initial genome screen51 showed more sis. As discussed by Greenberg et al,61 the prevailing promise on follow-up and expansion of the data.52,53 In trend to gravitate toward nonparametric methods (ASP a similar vein, failure to replicate does not necessarily and APM) in the analysis of complex traits, a trend evi- refute a hypothesis because of difficulties in detecting dent in some of the aforementioned studies, has its linkage for complex traits when gene effects are weak drawbacks, not the least of which is the risk of overin- or confined to a small fraction of the population. As terpreting a ‘positive’ finding. Apparent misconcep- discussed,13 in such circumstances the initial positive tions concerning the presumed superiority of the non- results may be due to random fluctuations that help parametric approach over parametric methods (ie, lod push weak gene effects above the threshold of score analysis) include: 1) nonparametric methods ‘accepted linkage’, while replication studies will often have high statistical power, compared with lod score miss these effects because of regression to the ‘true’ methods, and are less sensitive to misspecification of value. the genetic model. In fact, it is well established that a A related issue is the power to detect and replicate lod score analysis generally has greater power and that linkage with a given sample size. There have been it is reasonably robust to model misspecification pro- numerous attempts to estimate power under various vided more than one plausible model is tested (eg, genetic models, including oligogenic transmission that dominant and recessive models); 2) nonparametric may have special relevance to bipolar disorder. For methods are ‘model-free’ and therefore are more suit- genes of relatively large effect, the required sample size able for psychiatric disorders whose mode of inherit- is generally within reach, provided heterogeneity is not ance is uncertain. In fact, nonparametric methods are excessive.45,54–58 But genes of lesser effect are more not always truly model-free, and in some instances are problematic. Risch and Merikangas59 estimated, based statistically ‘equivalent’ to parametric analysis. More- on simulations of multiplicative traits, that the number over, with lod score analysis there is the option of of families required to detect genes of modest effect employing several different genetic models (thus with a linkage approach is prohibitively large. In con- covering a range of inheritance patterns) with adequate trast, the required number of affected sib-pairs is power and little danger of missing a true linkage. Such within reach (hundreds), provided the genotypic rela- an option (that is, a range of models) is not available tive ratio (GRR) is of a large enough magnitude for nonparametric analyses; 3) because investigators (GRR = 4); for a GRR of 2 or less, however, the sample using lod score analysis tend to use several models, size is again beyond reach (well over 2000). In the light there is a danger of multiple test effects leading to type of these findings, they are not sanguine about the util- I errors. In fact, if the number of models is limited (as ity of linkage analysis for minor gene effects, though is the case in most studies), there is no such risk. Also, some important loci, especially those of a relatively studies involving nonparametric methods often per- Genetic linkage and bipolar affective disorder M Baron 207 form multiple tests, thus running the same risk; 4) indeed, less likely to uncover minor genes than associ- because the utility of nonparametric methods has been ation studies (and, in some instances, sib-pair demonstrated in several complex traits, such methods analysis). On the other hand, more biological impor- could detect a linkage that a lod score analysis would tance is accorded to genes of relatively large effect. miss. In fact, there are no systematic studies to support Moreover, pedigrees afford ample opportunities for sib- this assertion and such circumstances may be quite pair analysis and for some types of (intrafamilial) rare. The bottom line is that there is no one correct association tests, such as haplotype-relative-risk (HRR) strategy and that complementary approaches must be analysis64 and genomic association studies (using, for considered, with a cautious interpretation of results. example, affected sib-pairs and their parents);59 6) Study design is another issue to contend with. undoubtedly, nuclear families and sib-pairs are rela- Extended multiplex pedigrees, the mainstay of most tively easy to obtain. But the pedigree series reported linkage studies to date, have lost luster in some quar- so far show that when the choice is limited to medium- ters in favor of nuclear families, sib-pairs, and associ- sized pedigrees (as apposed to very large pedigree ation studies in populations.62 Arguments in support structures), the task is not daunting. As with analytic of this approach include: 1) extended pedigrees incur paradigms, a combined approach encompassing differ- more opportunities for introducing ‘extraneous’ genes ent study designs and sampling schemes holds the by way of assortative mating and bilineal transmission; broadest range of scenarios that may underlie this dis- hence, they are more likely to be genetically hetero- order. geneous, leading to reduced power to detect linkage; A total genome scan, the ultimate goal in linkage 2) extended pedigrees represent a particular form of a mapping, is costly and labor-intensive. Potential short- highly familial disease with a ‘dominant-like’ pattern cuts have been proposed to speed up gene localization. to the exclusion of alternate, possibly more representa- For example, Antonarakis65 advocated focussing on tive, genetic mechanisms. This may hamper the gen- GC-rich regions known to harbor disease genes. This eralizability of the results; 3) extended pedigrees are could speed linkage detection rather dramatically. best suited for the analysis of genes of a relatively large However, as discussed,66 where multiple genes are effect and are less likely to detect minor genes; 4) large involved (a likely scenario in complex disorders), pedigrees are hard to come by. To be sure, there are including genes in GC-poor regions, systematic genome counter arguments that carry some weight: 1) most scanning may be unavoidable. Other approaches seek- pedigree studies employ rigorous measures to screen ing to finesse the brute force efforts involved in system- out bilineal transmission. In addition, because hidden atic scanning are genome mismatch scanning (GMS),67 bilineality may still exist, two-trait-locus models repeat expansion detection (RED),68 and mitochondrial allowing for two disease loci within a family can be DNA analysis. GMS seeks to isolate all genomic regions applied to the data.63 In fact, these models can provide identical-by-descent in affected relative pairs. It does substantial linkage information to warrant the not require prior map information and the entire gen- inclusion of bilineality in the study design, thus pro- ome can be scanned in a single hybridization step. RED viding a broader perspective on the complex inherit- examines expansion of trinucleotide repeats as a cause ance of this disorder.63 There are, in addition, methods of genetic disease, especially in disorders displaying to analyze linkage and heterogeneity in extended pedi- ‘anticipation’ (a transmission pattern in which disease grees subdivided into all component nuclear families, severity increases, or age-at-onset decreases, in suc- to account and control for intrafamilial heterogeneity.64 cessive generations), such as bipolar disorder.69 There One should also note that small families and sib-pairs is preliminary evidence that expanded repeats may, are not impervious to heterogeneity: phenocopies may indeed, exist in bipolar disorder,70,71 but more work is be common due to low illness density; 2) extended needed. Mitochondrial disease mutations can explain pedigrees contain more genetic information than the parent-of-origin effects observed in some data on smaller families and have higher statistical power, bipolar disorder,72 whereby maternal transmission especially when heterogeneity is accounted for; 3) the appears to differ from paternal inheritance. The utility ‘dominant, single-gene’ appearance of many extended of these methods for bipolar disorder is far from cer- pedigrees should not be taken to imply that other tain. Specifically, GMS requires further developmental modes of transmission are discounted. Various models, work to be fully applicable to DNA (it has prim- specifically, oligogenic inheritance that involves more arily focussed on the yeast genome). ‘Anticipation’ and than two genes, can be successfully applied to the data; parent-of-origin effects have not been observed consist- 4) large pedigrees obtained from genetic isolates with ently and ascertainment bias is a distinct possi- a small number of founders (as in the studies of Fre- bility.47,73 Another potential shortcut of persistent imer et al,18 Pekkarinen et al,22 and Ginns et al24) may, interest involves chromosomal aberrations and candi- indeed, be special cases by way of mode of inheritance date genes (genes with known neurobiological (likely single-gene) and rarity of the mutations function). But the experience to date suggests that this involved. But they also enhance the prospect of testing may not be a high-yield approach except, perhaps, in for linkage disequilibrium to facilitate the positional genomic regions displaying putative region. Although cloning of putative genes. Such genes, if discovered, the search for some disease genes may be expedited could shed light on the biology of other, more common, with these methods, the rapid advances in automation disease forms; 5) linkage analysis in pedigrees is, of genotyping procedures, and in software for efficient Genetic linkage and bipolar affective disorder M Baron 208 and speedy linkage analysis,74,75 will likely uphold sys- 7 Pauls DL. Behavioral disorders: lessons in linkage. Nature tematic genome scans as a leading strat- Genet 1993; 3: 4–5. egy. 8 Spence MA, Bishop DT, Boehnke M, Elston RC, Falk C, The ultimate proof for a susceptibility locus lies in Hodge SE, Ott J, Rice J, Merikangas K, Kupfer D. Methodo- logical issues in linkage analyses for psychiatric disorders: the identification of the gene and its product. But the secular trends, assortative mating, bilineal pedigrees. Hum road from gene localization to positional cloning will Hered 1993; 43: 166–172. be tortuous and taxing, primarily because of uncer- 9 Baron M. Genetic linkage and mental disorders: an update tainties in pinpointing the map location of complex on analytic methodologies. Biol Psychiatry 1994; 36: 1–4. disease genes. This would necessitate exploration of 10 Cloninger RC. Turning point in the design of linkage stud- relatively large genomic regions. Moreover, the likely ies of schizophrenia. Am J Med Genet 1994; 54: 83–92. oligogenic etiology of bipolar disorder will complicate 11 Plomin R, Owen MJ, McGuffin P. The genetic basis of the identification of the disease gene from a large num- complex human behavior. Science 1994; 264: 1733–1739. ber of ‘candidates’ in the region. On the other hand, the 12 Baron M. Genes and psychosis: old wine in old bottles? effort and related projects are rapidly Acta Psychiat Scand 1995; 92: 81–86. 13 Lander E, Kruglyak L. Genetic dissection of complex changing the landscape for positional cloners. The vast traits: guidelines for interpreting and reporting linkage majority of all human genes are now represented as results. Nature Genet 1995; 11: 241–247. expressed sequence tags (ESTs) sites. Increasingly, pos- 14 Baron M. Genes and manic depression. Psychiat Genet itional cloning will be reduced to the systematic analy- 1996; (in press). sis of candidate genes.76 Multicenter collaborations 15 Owen MJ, Craddock N. Modern molecular genetic may be required to enhance the power of meiotic map- approaches to complex traits: implications for psychiatric ping, to access gene sequence information, and to disorders. Mol Psychiatry 1996; 1: 21–26. employ efficiently the full range of mutational and 16 Berrettini WH, Ferraro TN, Goldin LR, Weeks DE, Detera- population screening strategies. Wadleigh S, Nurnberger JI, Gershon ES. Chromosome 18 DNA markers and manic-depressive illness: evidence for a susceptibility gene. Proc Natl Acad Sci USA 1994; 91: Conclusion 5918–5921. 17 Stine OC, Xu J, Koskela R, McMahon FJ, Gschwend M, The recent outpouring of linkage data on bipolar dis- Friddle C, Clark CD, McInnis MG, Simpson SG, Breschel order attests to the renewed vitality of the field. TS, Vishio E, Riskin K, Feilotter H, Chen E, Shen S, Though its remains to be seen which, if any, of the Folstein S, Meyers DA, Botstein D, Marr TG, DePaulo JR. claimed linkages will lead to gene discovery, the vari- Evidence for linkage of bipolar disorder to chromosome ous studies display greater sophistication than their ill- 18 with a parent-of-origin effect. Am J Hum Genet 1995; fated predecessors. With the full gamut of powerful 57: 1384–1394. molecular and computational techniques being brought 18 Freimer NB, Reus VI, Escamilla MA, McInnis LA, Spesny M, Leon P, Service SK, Smith LB, Silva S, Rojas E, Gal- to bear on suitable clinical samples, and with the com- legos A, Meza L, Fournier E, Baharloo S, Blankenship K, bined resources of academia and industry, the pros- Tyler DJ, Batki S, Vinogradov S, Weissenbach J, Barondes pects for identifying and sequencing genes that predis- SH, Sandkuijl LA. Genetic mapping using haplotype, pose to bipolar disorder are brighter than ever. But association and linkage methods suggests a locus for sev- growing pains are still in evidence and maturity has ere bipolar disorder (BP) at 18q22–q23. Nature Genet yet to be reached, and celebrated. 1996; 12: 436–441. 19 Straub RE, Lehner T, Luo Y, Loth JE, Shao W, Sharpe L, Alexander JR, Das K, Simon R, Fieve RR, Lerer B, Endicott Acknowledgements J, Ott J, Gilliam TC, Baron M. A possible vulnerability locus for bipolar affective disorder on chromosome Supported by NIMH Research Scientist Award 21q22.3. Nature Genet 1994; 8: 291–296. KO5MH00176 and by grants MH42535 and MH43979. 20 Gurling H, Smyth C, Kalsi G, Moloney E, Rifkin L, O’Neal J, Murphy P, Curtis D, Petursson H, Brynjolfsson J. Link- References age findings in bipolar disorder. Nature Genet 1995; 10: 8–9. 1 Turecki G, Rouleau GA, Mari JJ, Morgan K. A systematic 21 Detera-Wadleigh SD, Badner JA, Goldin LR, Bettettini evaluation of linkage studies in bipolar disorder. Acta Psy- WH, Sanders AR, Rollins DY, Turner G, Moses T, Haerian chiat Scand 1996; 93: 317–326. H, Muniec D, Nurnberger JI, Gershon ES. Affected sib-pair 2 Risch N, Botstein D. A manic depressive history. Nature analyses reveal support of prior evidence for a suscepti- Genet 1996; 12: 351–353. bility locus for bipolar disorder, on 21q. Am J Hum Genet 3 Owen MJ. Will schizophrenia become a graveyard for mol- 1996; 58: 1279–1285. ecular geneticists? Psychol Med 1992; 22: 289–293. 22 Pekkarinen P, Terwilliger J, Bredbacka P-E, Lonnqvist J, 4 Risch N. Genetic linkage and complex diseases, with spe- Peltonen L. Evidence of a predisposing locus to bipolar cial reference to psychiatric disorders. Genet Epidemiol disorder on Xq24–q27.1 in an extended Finnish pedigree. 1990; 7: 3–16. Genome Res 1995; 5: 105–115. 5 Baron M. Genetic linkage in mental illness. Nature 1990; 23 Blackwood DHR, He L, Morris SW, McLean A, Whitton 346: 618. C, Thomson M, Walker MT, Woodburn K, Sharp CM, 6 Baron M, Endicott J, Ott J. Genetic linkage in mental ill- Wright AF, Shibasaki Y, St Clair DM, Porteous DJ, Wuir ness: limitations and prospects. Br J Psychiatry 1990; 157: WJ. A locus for bipolar affective disorder on chromosome 645–655. 4p. Nature Genet 1996; 12: 427–430. Genetic linkage and bipolar affective disorder M Baron 209 24 Ginns EI, Ott J, Egeland JA, Allen CR, Fann CSJ, Pauls DL, ted sib-scoring methods to detect HLA-linked disease sus- Weissenbach J, Carulli JP, Falls KM, Keith TP, Paul SM. ceptibility genes. Am J Med Genet 1984; 18: 135–146. A genome-wide search for chromosomal loci linked to 41 Weeks DE, Lehner T, Squires-Wheeler E, Kaufmann C, Ott bipolar affective disorder in the Old Order Amish. Nature J. Measuring the inflation of the lod score due to its maxi- Genet 1996; 12: 431–435. mization over model parameter values in human linkage 25 Gershon ES, Badner JA, Detera-Wadleigh SD, Ferraro TN, analysis. Genet Epidemiol 1990; 7: 237–243. Berrettini WH. Maternal inheritance and chromsome 18 42 Risch N. A note on multiple testing procedures in linkage allele sharing in unilineal bipolar pedigrees. Am J Med analysis. Am J Hum Genet 1991; 48: 1058–1064. Genet 1996; 67: 202–207. 43 Hodge SE, Greenberg DA. Sensitivity of lod scores to 26 Maier W, Hallmayer J, Zill P, Bondy B, Lichtermann D, changes in diagnostic status. Am J Hum Genet 1992; 50: Ackenheil M, Minges J, Wildenauer D. Linkage analysis 1053–1066. between pericentromeric markers on chromosome 18 and 44 Kelsoe JR, Ginns EI, Egeland JA, Gerhard DS, Goldstein bipolar disorder: a replication test. Psychiatry Res 1995; AM, Bale SJ, Pauls DL, Long RT, Kidd KK, Conte G, Hous- 59: 7–15. man DE, Paul SM. Re-evaluation of the linkage relation- 27 Pauls DL, Ott J, Paul SM, Allen CM, Fann CSJ, Carulli JP, ship between chromosome 11p loci and gene for bipolar Falls KM, Bouthillier CA, Gravius TC, Keith TP, Egeland affective disorder in the Old Order Amish. Nature 1989; JA, Ginns EI. Linkage analyses of chromosome 18 markers 342: 238–243. do not identify a major susceptibility locus for bipolar 45 Baron M, Endicott J, Lerer B, Loth JE, Alexander JR, Simon affective disorder in the Old Order Amish. Am J Hum R, Sharpe L, Gibbon M, Hasin D, Lilliston B, Schacht S, Genet 1995; 57: 636–643. Blumenthal R, Alexander J, Verter A, Tubi N, Fieve RR, 28 Debruyn A, Souery D, Mendelbaum K, Mendlewicz J, Van Gilliam TC, Lehner T, Ott J. A pedigree series for mapping Broeckhoven C. Linkage analysis of families with bipolar disease genes in bipolar affective disorder: sampling, illness and chromosome 18 markers. Biol Psychiatry 1996; assessment, and analytic considerations. Psychiat Genet 39: 679–688. 1994; 4: 43–55. 29 Coon H, Hoff M, Holik J, Hadley D, Fang N, Reimherr F, 46 Egeland JA. An epidemiologic and genetic study of affect- Wender P. Analysis of chromosome 18 DNA markers in ive disorders among the Old Order Amish. In: Papolos DF, multiplex pedigrees with manic depression. Biol Psy- Lachman HM (eds). Genetic Studies in Affective Dis- chiatry 1996; 39: 689–696. orders. John Wiley and Sons: New York, 1994, pp 70–90. 30 Byerley W, Holik J, Hoff M, Coon H. Search for a gene 47 Kato T, Winokur G, Coryell W, Keller MB, Endicott J, Rice predisposing to manic-depression on chromosome 21. Am J. Parent-of-origin effect in transmission of bipolar dis- J Med Genet 1995; 60: 231–233. order. Am J Med Genet 1996; 67: 546–550. 31 Ewald H, Eiberg H, Mors E, Flint T, Kruse TA. Linkage 48 Hebebrand J. A critical appraisal of X-linked bipolar ill- study between manic-depressive illness and chromosome ness: evidence for the assumed mode of inheritance is 21. Am J Med Genet 1996; 67: 218–224. lacking. Br J Psychiatry 1992; 160: 7–11. 32 Mendlewicz J, Simon P, Sevy S, Charon F, Brocas H, 49 Berrettini W. Genetic studies of bipolar disorders: new Legrus S, Vassart G. Polymorphic DNA marker on X-chro- and recurrent findings. Mol Psychiatry 1996; 1: 172–173. mosome and manic depression. Lancet 1987; 1: 1230– 50 Baron M. Is there a gene for manic depressive illness on 1232. the long arm of the X-chromosome? In: Gershon ES, Clon- 33 Gejman P, Detera-Wadleigh S, Martinez MM, Berrettini inger CR (eds). Genetic Approaches to Mental Disorders. WH, Goldin LR, Gelernter J, Hsieh W-T, Gershon ES. American Psychiatric Press: Washington, DC, 1994, Manic depressive illness not linked to factor IX region in pp 253–272. an independent series of pedigrees. Genomics 1990; 8: 51 Davis JL, Kawaguchi Y, Bennett ST, Copeman JB, Cordell 648–655. HJ, Pritchard LE, Reed PW, Gough SCL, Jenkins SC, 34 Lucotte GA, Landoulsi A, Berriche S, David F, Babron MC. Palmer SM, Balfour KM, Rowe BR, Farrall M, Barnett AH, Manic depressive illness is linked to factor IX in a French Bain SC, Todd JA. A genome-wide search for human type pedigree. Ann Genet 1992; 35: 93–95. 1 diabetes susceptibility genes. Nature 1994; 371: 130– 35 Jeffries FM, Reiss AL, Brown WT, Meyers DA, Glicksman 136. AC, Bandyopadhyay S. Bipolar spectrum disorder and 52 Hashimoto L, Habita C, Beressi JP, Delepine M, Besse C, : a family study. Biol Psychiatry 1993; Cambon-Thomsen A, Deschamps I, Rotter JI, Djoulah S, 33: 213–216. James MR, Froguel P, Weissenbach J, Lathrop GM, Julier 36 Baron M, Freimer NF, Risch N, Lerer B, Alexander JR, C. Genetic mapping of a susceptibility locus for insulin Straub RE, Asokan S, Das K, Peterson A, Amos J, Endicott dependent diabetes mellitus on chromosome 11q. Nature J, Ott J, Gilliam TC. Diminished support for linkage 1994; 371: 161–164. between manic depressive illness and X-chromosome 53 Copeman JB, Cucca F, Hearne CM, Cornall RJ, Reed PW, markers in three Israeli pedigrees. Nature Genet 1993; 3: Ronningen KS, Undlien DE, Nistico L, Buzzetti R, Tosi R, 49–55. Pociot F, Nerup J, Cornelis F, Barnett AH, Bain SC, Todd 37 Debruyn A, Raeymaekers P, Mendelbaum K, Sandkuijl JA. Linkage disequilibrium mapping of a type 1 diabetes LA, Raes G, Delvenne V, Hirsch D, Staner L, Mendlewicz susceptibility gene (IDDM7) to chromosome 2q31–q33. J, Van Broeckhoven C. Linkage analysis of bipolar illness Nature Genet 1995; 9: 80–85. with X-chromosome DNA markers: a susceptibility locus 54 Lander E, Botstein D. Strategies for studying hetero- in Xq27–28 cannot be excluded. Am J Med Genet 1994; geneous genetic traits in by using a linkage map 54: 411–419. of restriction fragment length polymorphisms. Proc Natl 38 Witte JS, Elston RC, Schork NJ. Genetic dissection of com- Acad Sci USA 1986; 83: 7353–7357. plex traits. Nature Genet 1996; 12: 355–356. 55 Martinez M, Goldin LR. Power of the linkage test for a 39 Curtis D. Genetic dissection of complex traits. Nature heterogeneous disorder due to two independent inherited Genet 1996; 12: 356–357. causes: a simulation study. Genet Epidemiol 1990; 7: 40 Suarez BK, Van Eerdewegh P. A comparison of three affec- 219–230. Genetic linkage and bipolar affective disorder M Baron 210 56 Risch N. Linkage strategies for genetically complex traits. 67 Nelson SF, McCuster JH, Sander MA, Kee Y, Modrich P, II. The power of affected relative pairs. Am J Hum Genet Brown PO. Genome mismatch scanning: a new approach 1990; 46: 229–241. to genetic linkage mapping. Nature Genet 1993; 4: 11–18. 57 Goldin LR. Detection of linkage under heterogeneity: com- 68 Schalling M, Hudson TJ, Buetow KH, Housman DE. Direct parison of the two-locus vs admixture models. Genet Epid- detection of novel expanded trinucleotide repeats in the emiol 1992; 9: 61–66. human genome. Nature Genet 1993; 4: 135–139. 58 Craddock N, Khodel V, Van Eerdewegh P, Reich T. Math- 69 McInnis MG, McMahon FJ, Chase JA, Simpson SG, Ross ematical limits for multilocus models. The genetic trans- CA, DePaulo JR. Anticipation in bipolar affective disorder. mission of bipolar disorder. Am J Hum Genet 1995; 57: Am J Hum Genet 1993; 53: 385–390. 690–702. 70 Lindblad K, Nylander P-O, deBruyn A, Sourey D, Zander 59 Risch N, Merikangas K. The future of genetic studies of C, Engstrom C, Holmgren G, Hudson T, Chotai J, Mendle- complex human diseases. Science 1996; 273: 1516–1517. wicz J, Van Broeckhoven C, Schalling M, Adolfsson R. 60 Suarez BK, Hampe CL, Van Eerdewegh P. Problems of rep- Detection of expanded CAG repeats in bipolar affective licating linkage claims in psychiatry. In: Gershon ES, disorder using the repeat expansion (RED) method. Neuro- Cloninger CR (eds). Genetic Approaches to Mental Dis- biol Dis 1995; 2: 55–62. orders. American Psychiatric Press: Washington, DC, 71 O’Donovan C, Guy C, Craddock N, Murphy KC, Cardno 1994, pp 23–46. AG, Jones LA, Owen MJ, McGuffin P. Schizophrenia and 61 Greenberg DA, Hodge SE, Vieland VJ, Spence MA. Affec- bipolar disorder are associated with expanded CAG/CTG teds—only linkage methods are not a panacea. Am J Hum repeats. Nature Genet 1995; 10: 380–381. Genet 1996; 58: 892–895. 62 Greenberg DA. There is more than one way to collect data 72 McMahon FJ, Stine OC, Meyers DA, Simpson SG, DePaulo for linkage analysis. Arch Gen Psychiatry 1992; 49: 745– JR. Patterns of maternal transmission in bipolar affective 770. disorder. Am J Hum Genet 1995; 56: 1277–1286. 63 Schork NJ, Boehnke M, Terwilliger JD, Ott J. Two-trait- 73 O’Donovan MC, Owen NJ. Dynamic mutations and psy- locus linkage analysis: a powerful strategy for mapping chiatric genetics. Psychol Med 1996; 26: 1–6. complex genetic traits. Am J Hum Genet 1993; 53: 74 O’Connell JR, Weeks DE. The VITESSE algorithm for rapid 1127–1136. exact multilocus linkage analysis via set-rec- 64 Terwilliger JD. A powerful likelihood method for the ording and fuzzy inheritance. Nature Genet 1995; 11: analysis of linkage disequilibrium between trait loci and 402–408. one or more polymorphic marker loci. Am J Hum Genet 75 Kruglyak L, Daly MJ, Reeve-Daly MP, Lander ES. Para- 1995; 56: 777–787. metric and nonparametric linkage analysis: a unified 65 Antonarakis SE. Genome linkage scanning: systematic or multipoint approach. Am J Hum Genet 1996; 58: 1347– intelligent? Nature Genet 1994; 8: 211–212. 1363. 66 Baron M. Searching for complex disease genes: can it be 76 Lander E, Schork NJ. Genetic dissection of complex traits. made any easier? Psychiat Genet 1995; 5: 89–91. Science 1994; 265: 2037–2048.