No support for linkage to the bipolar regions on chromosomes 4p, 18p, or 18q in 43 schizophrenia pedigrees

To the Editor:

Following the distinction proposed by Kraepelin [1919], who built on the work of Morel [1860], Hecker [1871] and Kahlbaum [1863], bipolar affective disorder (BPAD) and schizophrenia are generally thought of as separate disorders. Modern epidemiological studies support this view since these disorders generally do not aggregate in the same families [Kendler et al., 1993; Maier et al., 1993]. An alternative view, originally put forward by Griesinger in 1861, is that schizophrenia and affective psychoses may be different expressions of the same disorder [Crow, 1986; Griesinger, 1861, as referenced by Maier et al., 1993]. In support of this view, cross prevalence studies have demonstrated a significantly higher rate of unipolar affective illness amongst the relatives of schizophrenia probands, compared with that observed amongst the relatives of control probands [Kendler et al., 1993; Maier et al., 1993; Taylor et al., 1993]. Furthermore, commonality in symptomatology (with both schizophrenic and bipolar patients experiencing Schneiderian first rank symptoms), in illness course (deterioration in some severe bipolar cases is more typical of the pattern seen in schizophrenia), and in effective treatments (neuroleptics, lithium) raise the possibility of overlapping causative factors, both genetic and nongenetic.

Patients with schizoaffective disorder exhibit both schizophrenic and affective symptoms in varying patterns over time, and in describing this group Kendell and Brockington [1980] raised four possible explanations: “that most are really schizophrenic illnesses, that most are really affective illnesses, that they are a mixture of schizophrenic and affective illnesses, and that they constitute a third independent type of psychosis” [Kendell and Brockington, 1980, p326]. Genetic studies have forced the need for pragmatic distinctions to be made within this group of patients for inclusion/exclusion in either schizophrenia or BPAD linkage studies. For example, RDC [Endicott and Spitzer, 1978] “schizoaffective, mainly schizophrenic” cases have been included in schizophrenia linkage studies (including the present author's study), while RDC “schizoaffective, mainly affective” cases have been included in BPAD linkage studies [Gershon et al., 1988]. Taken together, these factors suggest that, whether classified as separate disorders or as a continuum, overlap exists between affective and schizophrenic illnesses, and that the existence of this clinical and familial overlap raises the possibility of overlapping aetiologies, and perhaps shared susceptibility genes.

Blackwood and co-workers [Blackwood et al., 1996] reported a peak lod score of 4.1 coincident with D4S394 (α=0.35) on chromosome 4p16.1 in a cohort of 12 Scottish BPAD pedigrees. More recent genome screen results provide additional support for a bipolar predisposition gene in this region, particularly within Caucasian populations [Detera-Wadleigh et al., 1997; Ewald et al., 1998; McInnis, 1997; Nothen, 1997; Philibert et al., 1997]. As with most psychiatric genetics findings, there are also negative reports of linkage to bipolar disorder in this region [Raeymaekers, 1997; Rice, 1997; Schofield, 1997].

There is one report (an abstract) of a family with cases of schizophrenia and schizoaffective disorder that gave a positive linkage score (lod 1.96) to marker D4S403, near DRD5, although analysis in an additional 23 pedigrees collected by the same group failed to provide supportive evidence for this finding. [Asherson et al., 1998].

On chromosome 18 there are two distinct regions of interest for affective psychoses. Berrettini and co-workers [Berrettini et al., 1994, 1997, 1998] reported a suggestive finding in the analysis of five chromosome 18 pericentromeric marker loci (APM, p < 0.0001) as part of a partial genome screen in 22 bipolar pedigrees. Some support for this finding comes from positive linkage reports for bipolar disorder [Stine et al., 1995; Tsiouris et al., 1996] and schizophrenia [Schwab et al., 1998] data sets, although there are more negative reports for both BPAD and schizophrenia [Craddock, 1997; Kaneva, 1997; Murphy, 1997]. It is also noteworthy that the follow-up analysis of Stine et al. [1995] in 30 additional pedigrees was unable to replicate the original positive finding [McMahon et al., 1997].

McInnes [McInnes et al., 1996] analysed over 470 autosomal and X-linked markers in two BPAD pedigrees from a restricted gene pool in Costa Rica, and although positive for chromosome 18p, the peak score occurred within the chromosomal region 18q22-q23, consistent with the initial investigation of the same two pedigrees [Freimer et al., 1996]. Moreover, the analysis of other bipolar cohorts has demonstrated additional support for an 18 q locus [De bruyn et al., 1996; Detera-Wadleigh et al., 1997; McMahon et al., 1997], although not all studies agree [reviewed by Van Broeckhoven and Verheyen, 1999].

Notwithstanding conflicting results, the positive linkage results in some of the data sets suggest that BPAD predisposition genes and perhaps schizophrenia genes may lie within regions on 4p, 18p, and 18q. In the current study we have extended our 10 cM genome screen [Levinson et al., 1998] in 43 U.S./Australian pedigrees with multiple cases of schizophrenia to specifically investigate regions on chromosomes 4p, 18p, and 18q with previous evidence of linkage to affective psychoses. In this sample, the testable hypothesis, as presented in Levinson et al. [1998], is that a gene of major to moderate effect (i.e., λ_sibs of 3–5) for schizophrenia is present in one of these bipolar susceptibility regions.

The 43 pedigrees, with structures as published in Levinson et al. [Levinson et al., 1998], were recruited as previously described [Mowry et al., 1995] in Australia (18), Philadelphia (14), Iowa (8), and New York (3) with each containing a chronic schizophrenia proband with either two available first- or second-degree relatives with any schizophrenia-related disorder (as defined below), or a sibling with schizophrenia. Pedigrees were excluded if bilineal involvement (within two degrees) was suspected based on family history. Two diagnosticians at each site and one from another site arrived at a consensus Diagnostic and Statistical Manual of Mental Disorders (third edition, revised; DSM-III R) diagnosis [American Psychiatric Association, 1987] after reviewing all available information from the modified SADS [Endicott and Spitzer, 1978] or CASH [Andreasen et al., 1992] diagnostic interviews (including inventories of schizotypal and paranoid personality traits) carried out with each proband and other available relatives, as well as from medical records.

The 269 genotyped subjects included 126 cases of psychotic disorders known to aggregate in families of schizophrenia probands [Kendler et al., 1993; Levinson and Mowry, 1991; Maier et al., 1993], including diagnoses of schizophrenia (n=103), chronic schizoaffective disorder (n=4), atypical (nonaffective) psychosis (n=16), delusional disorder (n=2), and schizophreniform disorder (with comorbid paranoid personality disorder) (n=1). To avoid inclusion of psychotic mood disorders, schizoaffective cases with clear melancholic or euphoric mood and associated symptoms suggestive of primary mood disorders were not considered affected, nor were nine cases of schizotypal and/or paranoid personality disorders without psychosis because, although these disorders coaggregate with schizophrenia in families, the associated relative risk is substantially lower than that for psychotic disorders [Webb and Levinson, 1993].

Genotyping was carried out using two methods:

1. Three hundred ten fluoresceinated autosomal and X-chromosome microsatellite markers drawn from the CHLC/Weber 6.0 map (http://genetics.mfldclin.edu/scrset/scrset.6) were typed by Research Genetics, Inc., using PE/ABI 377 sequencers and GENESCAN/GENOTYPER software (PE/ABI), as part of the previously published search for schizophrenia susceptibility genes [Levinson et al., 1998].

2. Additional markers were typed by radiolabeled analysis as previously described [Dann et al., 1997; Mowry et al., 1995;] and then integrated into the CHLC/Weber 6.0 maps used for our genomewide analysis [Levinson et al., 1998] using marker orders and distances derived from Southhampton LBD maps [Collins et al., 1996] These included chromosome 4p markers D4S394, D4S403, and D4S419, chromosome 18p markers, D18S78, D18S53, and GOLF, and chromosome 18q markers D18S61, D18S58, and D18S70. Markers D4S394 and D18S58 were chosen since they represented the peak markers found in the originally reported BPAD studies for 4p Blackwood et al., 1996] and 18q [Freimer et al., 1996] whereas D18S53 and GOLF were the peak markers, by different analysis methods, in the study of Schwab et al. [Schwab et al., 1998] in 65 schizophrenia pedigrees. Marker D4S403 had been typed by both the fluorescent and radioactive methods in our data set to yield a 98.2% genotypic concordance between the two methods.

Multilocus linkage analysis using the model-free nonparametric linkage (NPL) test offered by GENEHUNTER 1.1 [Kruglyak et al., 1996] was carried out as previously described [Levinson et al., 1998], to avoid multiple testing in this relatively small sample.

The results of these integrated analyses for chromosomes 4 and 18, including the intermarker distances used, multilocus NPL scores, and locus-specific information contents are presented in Table I. These data provide no support for the hypothesis that a schizophrenia susceptibility gene of major to moderate effect within our data set maps to one of these bipolar susceptibility regions on 4p, 18p, and 18q. When our results are considered in the context of those from other researchers, two explanations arise: either the initially positive scores within these regions are the result of type one linkage errors (false positive scores); or a gene(s) of small effect for BPAD, schizophrenia, or both, maps to one or more of these regions. Given issues of power and population stratification in individual data sets, identifying such genes of small effect remains a challenge for the field.