1. family and parenting

© 2001 Oxford University Press
Human Molecular Genetics, 2001, Vol. 10, No. 26 3037–3048
Genome-wide scan in a nationwide study sample of
schizophrenia families in Finland reveals susceptibility
loci on chromosomes 2q and 5q
Tiina Paunio1,2,*, Jesper Ekelund1,2, Teppo Varilo1,3, Alex Parker4, Iiris Hovatta1,
Joni A. Turunen1,3, Kate Rinard4, Alessandro Foti4, Joseph D. Terwilliger5,6,
Hannu Juvonen2, Jaana Suvisaari2, Ritva Arajärvi2, Jaana Suokas2, Timo Partonen2,
Jouko Lönnqvist2, Joanne Meyer4 and Leena Peltonen1,3,7
1Department
of Molecular Medicine and 2Mental Health and Alcohol Research, National Public Health Institute,
Helsinki, Finland, 3Department of Medical Genetics, University of Helsinki, Finland, 4Millennium Pharmaceuticals,
Inc., Cambridge, MA, USA, 5Department of Psychiatry and Columbia Genome Center, Columbia University,
6Division of Medical Genetics, New York State Psychiatric Institute, New York, NY, USA and 7Department of Human
Genetics, UCLA, Los Angeles, CA, USA
Received August 22, 2001; Revised and Accepted October 23, 2001
We have previously carried out two genome-wide
scans in samples of Finns ascertained for schizophrenia from national epidemiological registers.
Here, we report data from a third genome scan in a
nationwide Finnish schizophrenia study sample of
238 pedigrees with 591 affected individuals. Of the
238 pedigrees, 53 originated from a small internal
isolate (IS) on the eastern border of Finland with a
well established genealogical history and a small
number of founders, who settled in the community
300 years ago. The total study sample of over 1200
individuals were genotyped, using 315 markers. In
addition to the previously identified chromosome 1
locus, two new loci were identified on chromosomes
2q and 5q. The highest LOD scores were found in the
IS families with marker D2S427 (Zmax = 4.43) and in the
families originating from the late settlement region
with marker D5S414 (Zmax = 3.56). In addition to 1q, 2q
and 5q, some evidence for linkage emerged at 4q, 9q
and Xp, the regions also suggested by our previous
genome scans, whereas, in the nationwide study
sample, the region at 7q failed to show further evidence
of linkage. The chromosome 5q finding is of particular
interest, since several other studies have also shown
evidence for linkage in the vicinity of this locus.
INTRODUCTION
Schizophrenia is assumed to be an etiologically heterogeneous
psychiatric disorder characterized by periodic psychotic symptoms, and by problems in reality testing. The disease affects
∼1% of people throughout the world (1), although a number of
recent studies have suggested lower prevalence rates (2).
Moreover, a higher prevalence, potentially affected by genetic
drift, has been observed in certain populations (3).
In addition to environmental factors, such as infection during
the second trimester of pregnancy (4), obstetric complications
(5) and the early rearing environment (6,7), multiple lines of
evidence imply that genetic factors contribute to the pathogenesis. Relatives of schizophrenia probands are at increased
risk of developing schizophrenia or other schizophrenia
spectrum disorders (8), as are adoptees with schizophrenic
biological parents (9). Perhaps the most convincing evidence
for genetic factors influencing schizophrenia comes from twin
studies. In a Finnish study, the proband-wise concordance was
shown to be around five times higher (46%) among monozygotic than among dizygotic twins (9%) (10), and similar
findings have been made in studies of other populations (11).
On the basis of population-wide health care registers, the
lifetime prevalence of schizophrenia in Finland has been
estimated to be 1.2%. Interestingly, the prevalence was found
to be higher in the rural eastern and northeastern parts of the
country than in the more densely inhabited urban population in
southern Finland, one of the areas with particularly high agecorrected lifetime risks being found in an isolated municipality
in northeastern Finland [internal isolate (IS)] (3.2%) (3).
Genetic analyses of schizophrenia families in this isolate
provided some evidence for a putative predisposing locus on
chromosome 1q (3). Another study of a population-wide
collection of affected sib pairs in Finland showed some
evidence of two loci influencing the risk of schizophrenia on
chromosomes 7q and 1q (12).
Despite the efforts of several groups worldwide, no major
locus for schizophrenia has yet been identified and even
replication of previously implicated loci has been rare (13).
This may be due to the etiological complexity of the disorder
*To whom correspondence should be addressed at: Department of Molecular Medicine, National Public Health Institute, Biomedicum, PL104, 00251 Helsinki,
Finland. Tel: +358 9 47448751; Fax: +358 9 47448480; Email: [email protected]
3038 Human Molecular Genetics, 2001, Vol. 10, No. 26
Table 1. Study sample
Nuclear
families
Extended
pedigrees
IS
90
53
AF
192
185
Total (Com) 282
238
Genotyped
individuals
Affected individuals
LC1
LC2
LC3
LC4
389
96
+21
+6
+13
876
340
+54
+40
+21
1265 (1241a) 436
+75
+46
+34
Overlap with 30
Hovatta et al.
(3)
18 (16a)
137 (29a)
37
+8
+2
+8
Overlap with 116
Ekelund et al.
(12)
115 (51a)
447 (98a)
212
+29
+14
+8
aNumbers of individuals who were genotyped genome-wide. There was an
overlap of three families in Hovatta et al. (3) and Ekelund et al. (12), including
three individuals in the genome-wide stage of scan.
or to genetic, environmental and sociocultural variation among
the populations studied, as well as to inconsistent diagnostic
schemes and samples of insufficient sizes. Here, a populationwide study sample of more than 1200 individuals from 238
Finnish pedigrees including multiple cases of schizophrenia
were genotyped, using a genome-wide collection of microsatellite markers. In addition to linkage findings on chromosome 1q, the data provide evidence for two additional
schizophrenia loci on chromosomes 2q and 5q.
RESULTS
Genome-wide scan (stages 1 and 2)
Our nationwide schizophrenia study sample of more than 1200
individuals included two categories of families: those originating from in an isolated municipality in northeastern Finland
with a particularly high age-corrected lifetime risk for schizophrenia (IS) and those from elsewhere in Finland [all-Finland
sample (AF)] (Table 1 and Fig. 1). These samples partially
overlapped with those analyzed in two previous studies (3,12).
However, at the genome-wide stage of the search, the overlap
for the individual samples was only 8% for the IS sample and
11% for the AF sample (Table 1).
In the initial analyses with a marker set of 315 markers
equally spaced across all 22 autosomes and the X chromosome, the IS and AF study samples were treated separately:
two-point linkage analyses were performed in each subsample,
using both liability classes 1 and 2 (LC1 and LC2). In order to
evaluate the statistical significance of these findings, which is
complicated by the multiple testing, we performed simulations
using estimated allele frequencies, marker maps and existing
pedigree structures. The markers giving LOD scores ≥1 were
then chosen for a linkage analysis now including the broader
liability classes 3 and 4 (LC3 and LC4) in IS, AF and LC1–LC4
in the combined sample (Com).
Isolate sample. In the initial (stage 1) pairwise analyses of the
53 IS pedigrees, the best evidence for linkage was obtained
with the marker D2S427 on chromosome 2q37. Using the
recessive model of inheritance in nuclear families without
genealogical links and for the most restricted diagnostic class
Figure 1. Stratification of the study sample according to genealogical data.
When collecting the study sample, families were categorized into two classes:
those in which at least one of the parents originated from the internal isolate
(IS) and those in which both parents originated from elsewhere in Finland
(AF). The AF sample could be further divided by choosing families with all
known ancestors from the late-settlement region of Finland (AF/LS). The
population in this region arose from a second wave of inhabitation that
occurred in Finland in the 16th century.
(LC1), the LOD score (Zmax) reached for this marker was 4.43.
When the full pedigree information was incorporated in the
analyses, Zmax of 2.94 was obtained. For markers on chromosomes 1 (D1S1728), 3 (D3S1311), 5 (GATA81C06) and 18
(D18S877), we obtained Zmax values ≥2. Altogether 37 autosomal markers and two of the X-chromosomal markers
produced LOD scores with Zmax ≥ 1 in the stage 1 analysis of
the IS families (Fig. 2A). These 39 markers were then chosen
for two-point linkage analysis, also using the broader diagnostic
classes LC3 and LC4 (stage 2 of the analyses). Typically, with
the exception of a few markers, evidence for linkage remained
the same or decreased, and none of the markers resulted in a
LOD score ≥3. LOD scores ≥2 were obtained for markers on
chromosomes 1, 3, 5 and 19, specifically D1S1728, D3S2460,
D5S395, GATA81C06 and D19S418. Table 2 summarizes the
results of the analyses of stages 1 and 2.
All-Finland sample. In the two-point linkage analysis of the
185 pedigrees from the nationwide sample, excluding the IS
pedigrees, 27 of the markers of the basic scan set gave a Zmax ≥ 1
using the disease status of LC1 or LC2. The strongest evidence
Human Molecular Genetics, 2001, Vol. 10, No. 26 3039
Table 2. Results of the two-point linkage analysis of the genome-wide scan
Location (cM)
Pedigree typea
Mode of inheritance
Zmax in Com
2
Extended pedigrees
Dominant
2.68
1
Nuclear families
Recessive
3.30
3
Nuclear families
Recessive
0.89
1
Extended pedigrees
Recessive
0.36
2.33
4
Nuclear families
Dominant
1.41
2.36
4
Nuclear families
Dominant
0.44
Nuclear families
Dominant
0.01
Extended pedigrees
Recessive
0.73
Nuclear families
Dominant
3.55
Nuclear families
Recessive
1.60
Nuclear families
Dominant
1.07
Marker
IS
D1S1728
1
97.17
2.40
IS
D2S427
2
224.70
4.43
IS
D3S2460
3
127.52
2.17
IS
D3S1311
3
217.76
2.79
IS
D5S395
5
52.55
IS
GATA81C06
5
57.50
IS
D18S877
18
54.40
2.37
2
IS
D19S418
19
82.72
2.04
4
AF
D5S820
5
159.77
3.16
1
AF
D12S85
12
54.92
2.11
3
AF/LS
D5S1480
5
143.40
2.05
1
aRefers
Chromosome
Zmax
LC
Material
to the pedigrees used in linkage analyses producing the Zmax (Materials and Methods).
for linkage, a LOD score of 3.16, was obtained with the marker
D5S820, using the diagnostic class of LC1 and the dominant
inheritance model in the nuclear families. When the complete
pedigree information was used, the evidence for linkage was
slightly lower (Zmax = 2.86). No other marker provided a LOD
score ≥2.0 in the first stage of the analysis (Fig. 2B and
Table 2). In the stage 2 analyses with broader diagnostic
classes (LC3 and LC4), only one of the markers, D12S85, gave
a LOD score ≥2 (2.11) (Table 2). Thus, the strongest evidence
for linkage was obtained with the marker D5S820 on the long
arm of chromosome 5.
Families from the late settlement region (AF/LS). The AF
sample could be stratified genealogically by grouping the
families with ancestors from the late settlement region of
Finland (AF/LS) (Fig. 1). A total of 118 families originating
from the northern and eastern parts of the country were
analyzed with the 27 markers that had produced a LOD score
≥1 in the AF sample. In an analysis using all four LCs, only
four markers (D1S431, D4S2394, D5S1480 and D12S1294)
provided increased evidence for linkage as compared with the
AF sample, and only the marker D5S1480 showed a Zmax ≥ 2
(Zmax = 2.05 in LC1) (Table 2).
Combined study sample (Com). When all the study families
were pooled together, the total study sample consisted of 238
pedigrees and more than 1200 genotyped individuals. This
sample was analyzed with the 62 markers that had given a Zmax
≥ 1 in at least one analysis in either the IS or the AF sample,
and all four diagnostic classes were used. The best evidence of
linkage was obtained for chromosome 5q, in which a Zmax of
3.55 was obtained with the marker D5S820 in LC1 and a
recessive inheritance model in the nuclear families. When LC1
was used with full pedigree information, the marker D2S427
also yielded a LOD score ≥3 (3.30) (Fig. 2C and Table 2).
Simulation. Approximate solutions to the genome-wide
significance of point-wise LOD scores have been derived on
the assumption that meiotic information is complete (13).
However, in reality, the situation is much more complicated by
the presence of pedigrees of varying size and structure, often
with a significant amount of missing genotype data. The
proposed criteria of Lander and Kruglyak (13) are not entirely
appropriate, and only a simulation based on the real markers
and data structures used in a given study can provide realistic
inferential guidance (14), especially when multiple analytical
techniques have been used for data analysis.
In this experiment, we estimated the distribution of the LOD
scores maximized over all models within the AF and IS data
sets separately and then maximized over both data sets. The
resulting distribution functions are shown in Figure 3A. Figure
3B presents the density functions for the same LOD scores
maximized over the models. For a range of false positive rates,
the equivalent LOD score value is indicated in Table 3, first for
an average model within a given data set, then maximized over
all models in each data set, and finally, maximized over both
models and data sets. The highest LOD score observed in this
study has a genome-wide significance of ∼0.01, exceeding the
0.05 cut-off recommended by Lander and Kruglyak (13). In
other words, only one in 100 genome scans conducted on these
data sets with the analysis models applied would be expected
to lead to such a significant LOD score if there were no linkage
to any marker in the genome scan. The data also shows that the
penalty for performing such a large number of analyses and
maximizing over two independent data sets was only ∼0.8
LOD score units. This could be subtracted from any of the
observed LOD scores, to get a statistic which is somewhat
more comparable to traditional single-model LOD scores.
After doing this, our best LOD score is still comparable in
significance to a single-analysis LOD score of roughly 3.6.
Multipoint analysis of chromosomes 2q and 5q
According to the pairwise linkage analyses of stages 1 and 2 in
the genome-wide search, two chromosomal regions appeared
most interesting: chromosomes 2q and 5q. We conducted
multipoint non-parametric analyses on these chromosomal
regions using Simwalk2 software. We used the LC and study
3040 Human Molecular Genetics, 2001, Vol. 10, No. 26
Figure 2. Pairwise linkage analysis of the IS families (A) and the AF families (B) with the genome-widely distributed markers in LC1 and LC2 (stage 1 of the
analyses). (C) The Com sample was analyzed with 62 markers that had given a Zmax ≥ 1 in stage 1, using all four diagnostic classes in the analyses.
sample combination that had shown the best evidence for
linkage in the two-point analyses in the corresponding region.
Chromosome 2q. Multipoint analyses were performed on
chromosome 2q in the IS and AF samples as well as in the Com
sample. The best evidence for sharing of a marker allele was
obtained for the marker D2S427 in the pedigrees of the Com
sample with statistic A (P = 0.0032). Analyses of other data
sets also yielded the best evidence for allele sharing at D2S427
(Table 4 and Fig. 4A).
Chromosome 5q. The Simwalk2 analyses were first conducted
on chromosome 5q, using markers of the basic scan set in the
Com, AF and AF/LS samples. In all of them, the best evidence
Human Molecular Genetics, 2001, Vol. 10, No. 26 3041
Figure 3. Distribution (A) and density (B) functions estimated from the simulation study based on the MLE assuming that all the models had the same numbers
of equivalent analyses in each data set (IS and AF); LOD scores maximized over models for each sample separately (IS-MOM and AF-MOM) and maximized
over the two samples together (ALL). A shift of the order of magnitude of 1 LOD score can be observed in the simulated values when multiple models are used
in the analyses.
for marker allele sharing was obtained with statistic B for the
marker D5S1480, located 13 cM centromeric of D5S820, the
statistically most significant result for the AF/LS study sample
(P = 0.0016) (Fig. 4B). Encouraged by these results, especially
in view of previously conducted studies of schizophrenia
families from other populations (15,16), we pursued fine
mapping of chromosome 5q. A total of 30 additional markers
in chromosome 5q were genotyped; 29 of these covered a
distance of 14 cM with an average intermarker distance of
0.4 cM.
In pairwise linkage analysis of the IS families, none of the
markers on chromosome 5q gave a LOD score ≥2, whereas, in
the AF sample, the best two-point LOD score value was
obtained, as earlier described, with the marker D5S820
(Zmax = 3.16). However, in the genealogically stratified AF/LS
sample, analyses with the marker D5S414 from the dense map,
using LC4 and a recessive inheritance model in pedigrees,
resulted in a Zmax = 3.56 (Table 4). D5S414 is located 4 cM
centromeric of the marker D5S1480, which had given the best
LOD score (2.05) in stage 2 analyses of the AF/LS families.
3042 Human Molecular Genetics, 2001, Vol. 10, No. 26
Table 3. Critical LOD scores for given levels of genome-wide significance
Genome-wide P-value
One model, one data set
Maximized over models (one data set)
Maximized over data sets
and models
IS
AF
IS
AF
0.10
2.56
2.54
3.18
2.99
3.38
0.05
2.85
2.83
3.48
3.28
3.68
0.01
3.52
3.50
4.16
3.95
4.36
0.001
4.48
4.45
5.11
4.91
5.31
Table 4. Comparison of the results of two-point linkage analyses and Simwalk2 non-parametric multipoint analyses
Chromosome Sample
2q
IS
5q
aAs
Model
Two-point LOD score analysis
Simwalk2
LC
Type of families
analyzed
Mode of
inheritance
Marker
Zmax
P-valuea
Marker
Statistic b
P-value
1
Nuclear families
Recessive
D2S427
4.43
0.0000032
D2S427
A
0.013
AF
1
Nuclear families
Dominant
D2S434
1.63
0.0031
D2S427
A
0.013
Com
1
Pedigrees
Recessive
D2S427
3.3
0.000049
D2S427
A
0.0032
IS
1
Nuclear families
Recessive
D5S498
1.32
0.0069
D5S476
A
0.26
AF
1
Nuclear families
Dominant
D5S820
3.16
0.000069
D5S210
A
0.0079
AF/LS
4
Pedigrees
Recessive
D5S414
3.56
0.000026
FGF1
A
0.00019
Com
1
Nuclear families
Recessive
D5S820
3.55
0.000027
FGF1
A
0.0079
can be calculated directly from the two-point LOD score under homogeneity.
A measures the number of founder alleles contributing to the alleles of the affected individuals.
bStatistic
Finally, multipoint analyses were conducted, using 25
markers over an interval of 70 cM on chromosome 5q with the
IS, AF, AF/LS and Com study samples. In every case, the best
evidence for shared marker alleles in those affected was
obtained between markers D5S414 and D5S1480. Just as in the
two-point LOD score analyses, the most significant result
was obtained for the families from the late settlement region
at marker FGF1, located 2.7 cM centromeric to D5S1480
(P = 0.00019, statistic A). The results of these analyses are
summarized in Table 4 and Figure 4C.
DISCUSSION
Studying schizophrenia in the Finnish population: genetic
and environmental aspects
Genetic studies of schizophrenia in study samples collected
from different populations have given inconsistent results.
Almost every autosome has a putative locus for schizophrenia
and replication has been rare. This is most likely due to the
complexity of the disease phenotype and its inheritance, but
also to a host of other factors, such as diagnostic uncertainties,
selection bias and variability in the methods of data collection
(17). It is possible and even probable that many of the reported
findings are false-positives, since the significance levels
observed are typically within the range of magnitudes expected
from genome scan in the absence of a real signal. Furthermore,
a uniform distribution of schizophrenia loci across the genome
is consistent with what might be expected by chance alone.
Studies of large, statistically powerful samples are likely to
produce more compelling results than those of more typical
small-scale studies (17).
Here, we present the results of the largest single study
published so far, in which schizophrenia susceptibility loci
have been studied in a genome-wide scan. Additionally and
perhaps more importantly, all the families participating in this
study originate from the same culturally, historically and
genetically isolated population of Finland. They were ascertained according to the same study scheme, their genealogical
background was carefully evaluated and consensus diagnoses
were assessed after elaborate examination of life time diagnoses by the same team of experienced psychiatrists.
Isolated populations established by a limited number of
founders have proven useful for mapping genes of rare monogenic disorders, such as those of the Finnish disease heritage
(18,19). In genetic studies of complex traits, such as schizophrenia in primis, with multiple genes forming the genetic
background, the advantage of genetic isolates is probably less
distinct (20). However, families with multiple affected individuals
arising in a population with a restricted number of founders
still offer an advantage as compared with study material from
more heterogeneous populations. Further, if environmental
Human Molecular Genetics, 2001, Vol. 10, No. 26 3043
Figure 4. Simwalk2 non-parametric analysis of chromosomes 2q (A) and 5q with the markers of the basic scan set (B) and with the dense map (C). In (A) and (B),
the genetic distances used were based on information from the Marshfield meiotic map. In (C), data from the Human Genome Browser (http://genome.cse.ucsc.edu/
index.html) were applied, presuming the equivalence of 1 × 106 bp to 1 cM in the analyses.
factors play an important role in the disease pathogenesis,
homogeneity of the environmental components is a distinct
advantage for genetic studies. The Finnish population has a
relatively uniform culture, educational system, religion and
language, and, in subpopulations, social and environmental
similarities extend even further beyond the national homogeneity. The family structures can be determined reliably and
common ancestors identified in subisolates, using complete
population registers, church records and archived information
dating from 1634 (3). The high quality of the health care
3044 Human Molecular Genetics, 2001, Vol. 10, No. 26
Table 5. Comparison of chromosomal regions identified in Finnish schizophrenia families
Material
Chr 1q
Chr 4q
Chr 7q
Marker
Zmax
Marker
Zmax
Marker
Zmax
Marker
Zmax
Marker
Zmax
IS
None
>1
D4S2361
1.3
None
>1
None
>1
DXS1214
1.85
AF
D1S1595
1.51
D4S2394
1.04
None
>1
D9S1825
1.28
None
>1
IS without families
in Hovatta et al. (3)
D1S1609
1.8
None
>1
None
>1
None
>1
DXS7108
1.17
AF without families
in Ekelund et al. (12)
None
>1
D4S2394
1.8
None
>1
D9S934
1.16
None
>1
system and uniform education of the doctors in only five
medical schools make the registers very reliable, so that most
of the serious ascertainment biases can be avoided. Finally,
there is relatively good compliance of the population toward
medical genetic research in Finland, so that collection of
samples for a population-wide study is feasible.
Even in genetic isolates, the issue of locus heterogeneity in
complex traits is complex. During the past years, this matter
has typically been tackled by studying linkage disequilibrium
(LD) in selected chromosomal regions on the assumption that,
in inbred populations, increased LD between adjacent markers
should be identifiable. This feature has facilitated the LDbased gene mapping and cloning of numerous rare recessive
diseases (19). The intervals of LD around common alleles have
been less striking than around rare alleles, but evidence exists
for wider LD intervals in general chromosomes of Finns as
compared with those of more mixed populations (21–24).
Furthermore, a recent study demonstrates, even within
expanding populations, the existence of subisolates with
dramatically increased LD can exist (25).
Replication of earlier findings in the Finnish family
material
We have carried out two earlier genome scans in Finnish
schizophrenia families (3,12). In the first one, using only 20
families from the internal subisolate, the best evidence for
linkage was obtained for the markers on chromosome 1q with
a Zmax of 3.8 and a potential 6 cM haplotype in some families.
Other chromosomal regions with markers showing some
evidence for linkage included 4q, 9q and Xp (3). In the second
study, which was performed using 134 sib-pair families from
the whole of Finland, some evidence emerged for linkage on
chromosome 1q (Zmax = 2.62), whereas markers on 7q showed
the best evidence for linkage (Zmax = 3.18) (12).
In the present study, we found some evidence for linkage
with markers on chromosomes 1q, 4q and 9q, LOD scores ≥1
being obtained in the initial (stage 1) analyses in either IS or
AF families (Table 5). However, no evidence was obtained for
linkage with the markers of the basic scan set on chromosome
7q. Since this sample overlaps with those in the two previously
conducted studies, we also performed pairwise linkage
analyses with markers on the above-mentioned chromosomal
regions, excluding all previously analyzed families. In these
analyses also, evidence for linkage on 1q, 4q, 9q and Xp
remained, with LOD scores ≥1 obtained for the markers in
those regions (Table 5).
Chr 9q
Chr Xp
Among the chromosomal regions linked to schizophrenia in
Finland, the long arm of chromosome 1 (1q31–1q42) emerges
as particularly promising. Not only have there been the abovementioned findings in Finnish families, but also a linkage
finding in families with bipolar disorder and the observation of
a translocation in a Scottish pedigree with schizophrenia spectrum conditions (3,12,26,27). Consequently, we performed
fine mapping of this wide chromosomal region and the results
are described elsewhere (28).
Chromosomes 2q and 5q as newly implicated putative loci
for schizophrenia susceptibility genes in Finnish families
The present study revealed two interesting loci on chromosomes 2q and 5q, potentially containing genes predisposing to
schizophrenia.
For chromosome 2q, the best evidence of linkage was
obtained in the IS families with a Zmax of 4.43 for marker
D2S427. Non-parametric multipoint analysis with statistic A
of Simwalk2 software gave P = 0.013. When the Com study
sample was analyzed, a maximum LOD score of 3.3 was
obtained in pairwise linkage analyses and the multipoint
analysis using statistic A of Simwalk2 showed some evidence
for allele sharing with P = 0.0032. Findings of linkage to 2q
have been reported by other investigators (29,30), but the
markers giving the statistically most important signals for
linkage to schizophrenia were located >100 cM centromeric
from D2S427.
For chromosome 5q, the best evidence of linkage was
obtained with marker D5S820 in the initial (stage 1) analyses
of the AF study sample. With analysis of the Com study
sample, the maximum LOD score increased to 3.55. When
markers of the dense map were included, the best two-point
LOD score value (3.56) was obtained for the AF/LS families
with the marker D5S414, located 15–20 cM centromeric to the
marker D5S820, but only 4 cM from the marker D5S1480,
which had given the best evidence for linkage in the initial
multipoint analyses. The most significant evidence for allele
sharing in the Simwalk2 analyses with the markers of the
denser map was obtained, using statistic A, for the AF/LS
families at marker FGF1, located in the vicinity of D5S414,
with P = 0.00019. Thus, the chromosomal interval between
markers D5S479 and D5S636, covering 17 cM, remains a
potentially interesting region for predisposing genes in Finnish
schizophrenia families.
The first finding of linkage of schizophrenia to 5q was
published more than a decade ago by Sherrington et al. (31) in
Human Molecular Genetics, 2001, Vol. 10, No. 26 3045
Icelandic and British families, but the report was retracted with
updated information (32). However, a recently conducted
genome-wide scan with 13 pedigrees from Iceland and the UK,
using schizophrenia spectrum diagnosis, gave a three-point
LOD score of 3.6 at D5S422 (33). This marker is located
∼20 cM telomeric from D5S414. In a study of 265 Irish pedigrees
with schizophrenia, a multipoint heterogeneity LOD score of
3.04 was found with marker D5S804, located 8 cM centromeric to marker D5S414, under a narrow phenotypic definition
and with a recessive genetic model (16). In German and Israeli
families, additional support was obtained at marker D5S399
located only 1.1 cM from D5S414, with a Zmax of 1.8 from
multipoint analysis (15). In a pedigree with eight cases of
schizophrenia from Palau, Micronesia, a pairwise LOD score
of 2.66 was obtained at the marker D51480 (34). Finally, in a
reanalysis, using Simwalk2, of previously published genomescreen data from a Costa Rican pedigree with bipolar disorder,
a new locus on chromosome 5q was suggested, with best
evidence for allele sharing at D5S436 (P = 0.04 with allele
sharing statistic D of Simwalk2) (35). Thus, multiple lines of
evidence now imply the possible involvement of 5q as a region
containing genes for schizophrenia or psychosis susceptibility
(Fig. 5), although a recent multicenter study failed to add to the
existing evidence for linkage to 5q (36). In addition to the
above-mentioned findings in other linkage studies, a de novo
interstitial deletion of paternal origin, extending from band q22
to q23.2, has been published. The deleted regions contained the
marker D5S804, but not the marker D5S399 (37).
In conclusion, as implied by numerous other studies, schizophrenia, even in an isolated population, appears to be a genetically complex trait. Four of the five chromosomal loci
previously identified in Finnish schizophrenia families (3,12)
also remain interesting in the genome scan of the extensive
study sample described here. The major finding is the evidence
obtained for linkage of schizophrenia or schizophrenia spectrum conditions to two novel loci on chromosomes 2 and 5.
The implication of the chromosome 5q finding is particularly
interesting, since at least five other independent studies have
shown evidence for linkage in the same region.
MATERIALS AND METHODS
Collection of the study sample
By using nationwide registers of hospitalization for schizophrenia (hospital discharge registers), disability pensions and
free medication (Social Insurance Institution), we identified
30 339 individuals born between 1940 and 1969 who had a
diagnosis of schizophrenia between 1969 and 1991. The data
were combined, using each individual’s unique identification
code with that from the National Population Register to construct
core families. This ascertainment strategy, procedures for
acquiring permission to access the registers and issues concerning
informed consent have been addressed previously (3,38).
Each of the probands was contacted by his or her own
clinician. The probands were asked for written informed
consent and permission to contact their first-degree family
members. Collection of blood samples was carried out as
recommended in the Helsinki Declaration.
Figure 5. Multiple lines of evidence imply the possible involvement of
chromosome 5q as a region containing susceptibility loci for schizophrenia or
other type of psychosis. Positive findings in linkage analyses have been
obtained from various populations, namely, from Iceland and Britain (33),
Ireland (16), Germany and Israel (15), Palau, Micronesia (34) and Costa Rica
(35). The intervals between the markers giving the best signal in each study
are indicated in units based on sequence data from the Human Genome
Project (left) and on the Marshfield genetic map (right).
Assessment of diagnoses
All available inpatient and outpatient records were collected
from the probands and from those of their relatives who had
had any psychiatric diagnosis, according to the registers.
Registration of schizophrenia has been shown, in several
studies, to be highly reliable in Finland (10,39–41). Consensus
diagnosis was made by two psychiatrists blind to family structure to give a best-estimate lifetime diagnosis according to the
criteria of the Diagnostic and Statistical Manual of Mental
Disorders, 4th edition (DSM-IV). According to DSM-IV, the
typical symptoms of schizophrenia include psychotic
symptoms (hallucinations or delusions, disorganized speech,
disorganized or catatonic behavior) and negative symptoms,
such as poverty of speech. The Operational Criteria (OPCRIT)
checklist (42) was also completed for the affected individuals.
In the event of disagreement, diagnosis was based on the
3046 Human Molecular Genetics, 2001, Vol. 10, No. 26
opinion of a third reviewer. Using this diagnostic approach,
agreement between different psychiatrists on lifetime diagnoses has been shown to be good (3,12).
Affected individuals were divided into four LCs according to
the consensus diagnosis. The narrowest was LC1, consisting
only of individuals with schizophrenia. LC2 also included
individuals with schizoaffective disorder, LC3 added those
with schizophrenia spectrum conditions (schizoid, schizotypal
and paranoid personality disorder, schizophreniform, delusional and brief psychotic disorder, as well as psychosis NUD)
and LC4 was the broadest, including all individuals with
severe major affective disorders. Of the 1265 individuals
participating in the study, 511 (41%) were affected according
to LC2 and 80 (6%) more were affected according to LC4
(Table 1). The affected individuals included showed a slight
over-representation (59%) of males.
Families participating in the study
We categorized families into two classes: those in which at
least one of the parents of the proband originated from the IS
and those in which both parents came from elsewhere in
Finland (AF). The AF sample was further stratified genealogically by identifying families originating from the latesettlement region of Finland (AF/LS). The population of this
region arose from a second wave of inhabitation that occurred
in Finland in the 16th century (24). At that time, a small population in South-Savo inhabited the northern and eastern parts of
the country, including the isolate (Fig. 1). The names, dates
and places of birth of the ancestors were traced via local church
registers going as far back as 1850. For earlier periods, microfilm and microfiche copies of all local church records, available in the National Archives of Finland, were used for
genealogical studies.
Altogether 1265 individuals from 238 pedigrees participated
in the study; 53 pedigrees originated from the IS. The largest
pedigree (a2) in the IS sample consisted of 151 individuals,
including 19 sibships with DNA available from 88 individuals,
31 of whom were affected. Of the 185 pedigrees in the AF
sample, 118 originated from the late settlement region of
Finland (AF/LS) with all known ancestors born in that region.
In 27 families, all known ancestors originated from the early
settlement region in the southern part of Finland, and in the
remaining 40 families, a varying proportion of the ancestors
came from the early and late settlement regions, respectively.
Table 1 gives the numbers of nuclear families, pedigrees,
and affected individuals per pedigree that participated in the
study. In the majority (88%) of the pedigrees, DNA was available from at least two affected individuals and, in 75 pedigrees
(32%), DNA was available from three or more affected individuals. The present study sample partially overlaps with those
analyzed in two previous studies (3,12). Approximately onehalf of the individuals (46%) had been included in the previous
analyses. However, only a fraction of these individuals were
included in the initial, genome-wide stage of the scan in the
earlier studies, so that the overlap in the genome-wide search is
8% for the IS sample and 11% for the AF sample (Table 1).
Genotyping
A genome-wide scan was carried out, using a marker set of 315
markers spaced at ∼10–20 cM intervals on average across all
the human chromosomes (analyzed in stages 1 and 2). The
markers, di-, tri- and tetranucleotide repeats, were selected
from the CHLC-6 set and supplemental markers were added
from the Généthon map. Altogether 30 markers were additionally mapped on chromosome 5q, 29 of them covering a
distance of 14 cM. The maps we used incorporated information
from the Marshfield meiotic map and Stanford University G3
RH data. In addition, for estimation of the marker order and
intermarker distances on chromosome 5, we used the Human
Genome Browser (http://genome.cse.ucsc.edu/index.html;
draft assembly issued on January 9, 2001) which is based on
sequence data from the Human Genome Project. In the dense
map region, the genetic distance was estimated from the
physical distance by presuming the equivalence of 1 × 106 bp
to 1 cM.
Polymerase chain reactions (PCR) were set up with 20 ng
genomic DNA, and PCR cycling consisted of denaturation at
95°C for 5 min, followed by 30 cycles at 95°C for 30 s, five
cycles at 5°C for 30 s, and at 72°C for 60 s, and concluded with
extension at 72°C for 10 min at the end. The gels were run on
an Applied Biosystems (ABI) 377 DNA sequencer, using ABI
Prims* 377 data collection software. The data were analyzed
with the ABI Prims* GeneScan* 2.0.2 with Genotyper 1.1.1.
Statistical analyses
Two-point LOD score analysis. The analyses were done using
two affecteds-only models, one for dominant and the other for
recessive gene inheritance, each assuming an absence of
phenocopies and a very rare disease allele for the reasons
outlined by Göring and Terwilliger (43), who demonstrated that
these models lead to LOD score tests conceptually equivalent to
model-free analysis, but with superior statistical properties
under the null hypothesis.
The analyses were done with the MLINK program of the
linkage package (44), using the ANALYZE package to streamline the analysis (45). Marker locus allele frequencies were
estimated by gene counting (46), applying the DOWNFREQ
program on the conservative assumption that all the individuals in a pedigree are a random sample of the population, an
approximation that is mandated by the paucity of founders
available for genotyping and the complexity of some of the
pedigree and population structures (47). Because the overwhelming majority of families were small nuclear pedigrees,
there was no power to detect locus heterogeneity or epistatic
interactions, owing to insufficient degrees of freedom in the
data. For this reason, such analyses were not done systematically (48).
The analyses were done according to the following strategy:
at stage 1, we used LC1 and LC2 as criteria for the disease
status and analyzed the IS and AF samples separately. At stage
2, markers that had given LOD scores of at least 1 were
included, so that the criteria of LC3 and LC4 were applied in
the analyses of the IS and AF samples. In addition, the AF/LS
sample and the Com sample were analyzed from these selected
markers in all the diagnostic models, LC1–LC4.
Simulation. To evaluate the genome-wide statistical significance of the highest LOD score obtained in this experiment,
we performed a simulation study with our data set. Genotypes
were simulated for all the markers used in the scan, the marker
Human Molecular Genetics, 2001, Vol. 10, No. 26 3047
allele frequencies and intermarker distances being fixed at the
estimates assumed from our analysis. At each marker locus,
genotypes were assigned only to those individuals who were
genotyped for that marker in the actual study. One such simulation was performed for each of the 22 autosomes independently, the maximum LOD score (Zmax) over that chromosome
being stored separately for the AF and IS data sets for each of
the 100 replicates. At the end, an array of maximum LOD
scores were available and one replicate of each chromosome,
model, diagnosis and data set, respectively, was randomly
selected to represent a simulated whole genome scan. The
LOD score was maximized over the 22 autosomes in order to
get a genome-wide first-order statistic separately for each
model, diagnosis and data set and, subsequently, over all the
models and diagnoses. Finally, the distribution of such
genomewide first-order statistics was estimated from the
resulting empirical distributions.
It was not computationally feasible to simulate and analyze
the thousands of replicates of the entire genome scan in the two
data sets (IS and AF) needed to accurately estimate the small
P-values we are concerned with making inference about.
Therefore, we performed a maximum-likelihood estimation
(MLE) of the equivalent number of independent tests
performed in these scans. The simulated and bootstrapped
histogram of the Zmax values was fitted to the density function
of the first order statistic of N independent χ2 random variables,
on the assumption that a given two-point LOD score is distributed
according to 2Zln(10) ∼ 0.5χ2(1) pointwise. When the equivalent number of independent tests were considered to be equal
for all models, the estimated numbers were 353 for IS and 338
for AF. This difference is due to the fact that, in the AF sample,
the pedigrees are generally smaller and the autocorrelation of
the LOD score between linked markers was therefore greater
than in the larger pedigrees with many more ungenotyped
family members characteristic of the IS sample. As expected,
the differences between the models were statistically significant when whole pedigree analyses were compared with
nuclear families only (P < 0.001), most strikingly in the isolate
sample (P < 0.00001).
Additionally, there was concern about the effect of maximizing
the LOD score on eight models in each of the samples separately.
The ‘equivalent number of independent tests’ estimated for the
LOD scores maximized over the models were 1008 in the AF
sample and 1654 in the IS sample, the difference being statistically
significant (P < 10–10). In the AF sample, the eight models
applied were equivalent stochastically to three independent
analyses, whereas in the isolate sample, they were equivalent
to ∼4.7 of the independent models analyzed. The difference is
largely due to the many more degrees of freedom in the larger
pedigrees of the IS sample (14), making the effect of the
different models on the predicted genotypes more variable than
in the simpler family structures of the AF sample. Since the
two study samples were independent, the effect of maximizing
the LOD scores for both studies with eight models in each is
roughly equivalent to 2662 independent χ2 tests. Based on the
equivalent number of independent χ2 tests, we determined the
appropriate critical values for inference shown in Table 3 for
this experiment.
Multipoint analysis. For the regions showing the strongest
evidence of linkage in the two-point linkage analysis,
multipoint non-parametric analysis was performed, using
SIMWALK 2.60 (49). This program slides an imaginary trait
locus across the marker map. It calculates several statistics at
each position of the map by using Markov chain Monte Carlo
methods to sample patterns from the complete distribution of
underlying inheritance, in proportion to their likelihood, which
is calculated from the genotype data observed. In this study,
two of the statistics, A and B, were followed systematically.
Statistic A measures the total number of different founder
alleles contributing to the alleles of the affected individuals,
whereas statistic B measures the maximum number of alleles
among the affected individuals identical by descent from any
founder. In the analyses we chose the LC and family type that
had yielded the strongest evidence for linkage in the two-point
analyses of the markers in the studied region.
ACKNOWLEDGEMENTS
We would like to thank Ms M.Schreck, and Messrs S.Maruti,
T.Perheentupa, P.Haimi and A.Tanskanen for their part in the
computational issues of the paper. The authors also want to
warmly thank the participating patients and their families as
well as all the individuals who participated in collecting the
samples. The work was supported by Millennium Pharmaceuticals,
Inc.
REFERENCES
1. Schultz, S. and Andreasen, N. (1999) Schizophrenia. Lancet, 353,
1425–1430.
2. Kendler, K., Gallagher, T., Abelson, J. and Kessler, R. (1996) Lifetime
prevalence, demographic risk factors, and diagnostic validity of nonaffective
psychosis as assessed in a US community sample. Arch. Gen. Psychiatry,
53, 1022–1031.
3. Hovatta, I., Varilo, T., Suvisaari, J., Terwilliger, J.D., Ollikainen, V.,
Arajärvi, R., Juvonen, H., Kokko-Sahin, M.L., Väisänen, L., Mannila, H.
et al. (1999) A genomewide screen for schizophrenia genes in an isolated
Finnish subpopulation, suggesting multiple susceptibility loci.
Am. J. Hum. Genet., 65, 1114–1124.
4. Mednick, S.A., Machon, R.A., Huttunen, M.O. and Bonnett, D. (1988)
Adult schizophrenia following prenatal exposure to an influenza
epidemic. Arch. Gen. Psychiatry, 45, 189–192.
5. Sacker, A., Done, D., Crow, T. and Golding, J. (1995) Antecedents of
schizophrenia and affective illness. Obstetric complications.
Br. J. Psychiatry, 166, 734–741.
6. Alanen, Y. (1990) Need-adapted treatment of schizophrenia and other
psychoses: notes on the theoretical background and practical issues.
Psychiatr. Fennica, 21, 31–43.
7. Cannon, T. and Mednick, S. (1993) The schizophrenia high-risk project in
Copenhagen: three decades of progress. Acta Psychiatr. Scand., 370
(suppl.), 33–47.
8. Kendler, K. and Diehl, S. (1993) The genetics of schizophrenia: a current,
genetic–epidemiologic perspective. Schizophr. Bull., 19, 2.
9. Tienari, P., Wynne, L.C., Moring, J., Laksy, K., Nieminen, P., Sorri, A.,
Lahti, I., Wahlberg, K.E., Naarala, M., Kurki-Suonio, K. et al. (2000)
Finnish adoptive family study: sample selection and adoptee DSM-III-R
diagnoses. Acta Psychiatr. Scand., 101, 433–443.
10. Cannon, T.D., Kaprio, J., Lonnqvist, J., Huttunen, M. and Koskenvuo, M.
(1998) The genetic epidemiology of schizophrenia in a Finnish twin
cohort. A population-based modeling study. Arch. Gen. Psychiatry, 55,
67–74.
11. Cardno, A.G., Marshall, E.J., Coid, B., Macdonald, A.M.,
Ribchester, T.R., Davies, N.J., Venturi, P., Jones, L.A., Lewis, S.W.,
Sham, P.C. et al. (1999) Heritability estimates for psychotic disorders: the
Maudsley twin psychosis series. Arch. Gen. Psychiatry, 56, 162–168.
12. Ekelund, J., Lichtermann, D., Hovatta, I., Ellonen, P., Suvisaari, J.,
Terwilliger, J.D., Juvonen, H., Varilo, T., Arajärvi, R., Kokko-Sahin, M.L.
et al. (2000) Genome-wide scan for schizophrenia in the Finnish
3048 Human Molecular Genetics, 2001, Vol. 10, No. 26
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
population: evidence for a locus on chromosome 7q22. Hum. Mol. Genet.,
12, 1049–1057.
Lander, E. and Kruglyak, L. (1995) Genetic dissection of complex traits:
guidelines for interpreting and reporting linkage results. Nat. Genet., 11,
241–247.
Sawcer, S., Jones, H.B., Judge, D., Visser, F., Compston, A.,
Goodfellow, P.N. and Clayton, D. (1997) Empirical genomewide
significance levels established by whole genome simulations.
Genet. Epidemiol., 14, 223–229.
Schwab, S.G., Eckstein, G.N., Hallmayer, J., Lerer, B., Albus, M.,
Borrmann, M., Lichtermann, D., Ertl, M.A., Maier, W. and
Wildenauer, D.B. (1997) Evidence suggestive of a locus on chromosome
5q31 contributing to susceptibility for schizophrenia in German and
Israeli families by multipoint affected sib-pair linkage analysis.
Mol. Psychiatry, 2, 156–160.
Straub, R.E., MacLean, C.J., O’Neill, F.A., Walsh, D. and Kendler, K.S.
(1997) Support for a possible schizophrenia vulnerability locus in region
5q22–31 in Irish families. Mol. Psychiatry, 2, 148–155.
Baron, M. (2001) Genetics of schizophrenia and the new millennium:
progress and pitfalls. Am. J. Hum. Genet., 68, 299–312.
Norio, R., Nevanlinna, H.R. and Perheentupa, J. (1973) Hereditary
diseases in Finland; rare flora in rare soul. Ann. Clin. Res., 5, 109–141.
Peltonen, L., Jalanko, A. and Varilo, T. (1999) Molecular genetics of the
Finnish disease heritage. Hum. Mol. Genet., 8, 1913–1923.
Peltonen, L., Palotie, A. and Lange, K. (2000) Use of population isolates
for mapping complex traits. Nat. Rev. Genet., 1, 182–190.
Eaves, I.A., Merriman, T.R., Barber, R.A., Nutland, S., Tuomilehto-Wolf, E.,
Tuomilehto, J., Cucca, F. and Todd, J.A. (2000) The genetically isolated
populations of Finland and Sardinia may not be a panacea for linkage
disequilibrium mapping of common disease genes. Nat. Genet., 25, 320–323.
Taillon-Miller, P., Bauer-Sardina, I., Saccone, N.L., Putzel, J.,
Laitinen, T., Cao, A., Kere, J., Pilia, G., Rice, J.P. and Kwok, P.Y. (2000)
Juxtaposed regions of extensive and minimal linkage disequilibrium in
human Xq25 and Xq28. Nat. Genet., 25, 324–328.
Boehnke, M. (2000) A look at linkage disequilibrium. Nat. Genet., 25,
246–247.
Varilo, T., Laan, M., Hovatta, I., Wiebe, V., Terwilliger, J. and
Peltonen, L. (2000) Linkage disequilibrium and the demographic history
of isolated populations: Finland and a young subpopulation of Kuusamo.
Eur. J. Hum. Genet., 8, 604–612.
Zavattari, P., Deidda, E., Whalen, M., Lampis, R., Mulargia, A.,
Loddo, M., Eaves, I., Mastio, G., Todd, J.A. and Cucca, F. (2000) Major
factors influencing linkage disequilibrium by analysis of different
chromosome regions in distinct populations: demography, chromosome
recombination frequency and selection. Hum. Mol. Genet., 9, 2947–2957.
Detera-Wadleigh, S.D., Badner, J.A., Berrettini, W.H., Yoshikawa, T.,
Goldin, L.R., Turner, G., Rollins, D.Y., Moses, T., Sanders, A.R.,
Karkera, J.D. et al. (1999) A high-density genome scan detects evidence
for a bipolar-disorder susceptibility locus on 13q32 and other potential
loci on 1q32 and 18p11.2. Proc. Natl Acad. Sci. USA, 96, 5604–5609.
Millar, J.K., Wilson-Annan, J.C., Anderson, S., Christie, S., Taylor, M.S.,
Semple, C.A., Devon, R.S., Clair, D.M., Muir, W.J., Blackwood, D.H. et al.
(2000) Disruption of two novel genes by a translocation co-segregating with
schizophrenia. Hum. Mol. Genet., 22, 1415–1423.
Ekelund, J., Hovatta, I., Parker, A., Paunio, T., Varilo, T., Martin, R.,
Suhonen, J., Ellonen, P., Chan, G., Sinsheimer, J.S. et al. (2001)
Chromosome 1 loci in Finnish Schizophrenia Families. Hum. Mol. Genet.,
10, 1611–1617.
Faraone, S.V., Matise, T., Svrakic, D., Pepple, J., Malaspina, D.,
Suarez, B., Hampe, C., Zambuto, C.T., Schmitt, K., Meyer, J. et al. (1998)
Genome scan of European–American schizophrenia pedigrees: results of
the NIMH Genetics Initiative and Millennium Consortium. Am. J. Med.
Genet., 81, 290–295.
Levinson, D.F., Mahtani, M.M., Nancarrow, D.J., Brown, D.M., Kruglyak, L.,
Kirby, A., Hayward, N.K., Crowe, R.R., Andreasen, N.C., Black, D.W. et al.
(1998) Genome scan of schizophrenia. Am. J. Psychiatry, 155, 741–750.
Sherrington, R., Brynjolfsson, J., Petursson, H., Potter, M., Dudleston, D.,
Barraclough, B., Wasmuth, J., Dobbs, M. and Gurling, H. (1988)
Localization of a susceptibility locus for schizophrenia on chromosome 5.
Nature, 336, 164–167.
32. Kalsi, G., Mankoo, B., Curtis, D., Sherrington, R., Melmer, G.,
Brynjolfsson, J., Sigmundsson, T., Read, T., Murphy, P., Petursson, H.
et al. (1999) New DNA markers with increased informativeness show
diminished support for a chromosome 5q11–13 schizophrenia susceptibility
locus and exclude linkage in two new cohorts of British and Icelandic
families. Ann. Hum. Genet., 63, 235–247.
33. Gurling, H.M., Kalsi, G., Brynjolfson, J., Sigmundsson, T., Sherrington, R.,
Mankoo, B.S., Read, T., Murphy, P., Blaveri, E., McQuillin, A. et al.
(2001) Genomewide genetic linkage analysis confirms the presence of
susceptibility loci for schizophrenia, on chromosomes 1q32.2, 5q33.2, and
8p21–22 and provides support for linkage to schizophrenia, on chromosomes
11q23.3–24 and 20q12.1–11.23. Am. J. Hum. Genet., 68, 661–673.
34. Byerley, W., Tiobech, S., Blakis, A., Zuo, J., Zhao, M., Hoff, M.,
Bennet, P., Caleb, O. and Myles-Worsly, M. (1999) Evidence for a 5q31
schizophrenia locus in a large multipex kindred from Palau, Micronesia.
Mol. Psychiatry, 4, 4.
35. Garner, C., McInnes, L.A., Service, S.K., Spesny, M., Fournier, E.,
Leon, P. and Freimer, N.B. (2001) Linkage analysis of a complex
pedigree with severe bipolar disorder, using a Markov chain Monte Carlo
method. Am. J. Hum. Genet., 68, 1061–1064.
36. Levinson, D.F., Holmans, P., Straub, R.E., Owen, M.J., Wildenauer, D.B.,
Gejman, P.V., Pulver, A.E., Laurent, C., Kendler, K.S., Walsh, D. et al.
(2000) Multicenter linkage study of schizophrenia candidate regions on
chromosomes 5q, 6q, 10p, and 13q: schizophrenia linkage collaborative
group III. Am. J. Hum. Genet., 67, 652–663.
37. Bennett, R.L., Karayiorgou, M., Sobin, C.A., Norwood, T.H. and
Kay, M.A. (1997) Identification of an interstitial deletion in an adult
female with schizophrenia, mental retardation, and dysmorphic features:
further support for a putative schizophrenia-susceptibility locus at
5q21–23.1. Am. J. Hum. Genet., 61, 1450–1454.
38. Suvisaari, J.M., Haukka, J.K., Tanskanen, A.J. and Lonnqvist, J.K. (1999)
Decline in the incidence of schizophrenia in Finnish cohorts born from
1954 to 1965. Arch. Gen. Psychiatry, 56, 733–740.
39. Pakaslahti, A. (1987) On the diagnosis of schizophrenic psychoses in
clinical practice. Psychiatr. Fennica, 18, 63–72.
40. Isohanni, M., Makikyro, T., Moring, J., Rasanen, P., Hakko, H.,
Partanen, U., Koiranen, M. and Jones, P. (1997) A comparison of clinical
and research DSM-III-R diagnoses of schizophrenia in a Finnish national
birth cohort. Clinical and research diagnoses of schizophrenia.
Soc. Psychiatry Psychiatr. Epidemiol., 32, 303–308.
41. Mäkikyrö, T., Isohanni, M., Moring, J., Hakko, H., Hovatta, I. and
Lönnqvist, J. (1998) Accuracy of register-based schizophrenia diagnoses
in a genetic study. Eur. Psychiatry, 13, 57–62.
42. McGuffin, P., Farmer, A. and Harvey, I. (1991) A polydiagnostic
application of operational criteria in studies of psychotic illness.
Development and reliability of the OPCRIT system [news]. Arch. Gen.
Psychiatry, 48, 764–770.
43. Göring, H.H. and Terwilliger, J.D. (2000) Linkage analysis in the
presence of errors IV: joint pseudomarker analysis of linkage and/or
linkage disequilibrium on a mixture of pedigrees and singletons when the
mode of inheritance cannot be accurately specified. Am. J. Hum. Genet.,
66, 1310–1327.
44. Lathrop, G.M., Lalouel, J.M., Julier, C. and Ott, J. (1985) Multilocus
linkage analysis in humans: detection of linkage and estimation of
recombination. Am. J. Hum. Genet., 37, 482–498.
45. Terwilliger, J.D. and Göring, H.H. (2000) Gene mapping in the 20th and
21st centuries: statistical methods, data analysis, and experimental design.
Hum. Biol., 72, 63–132.
46. Smith, C. (1957) Counting methods in genetical statistics. Ann. Hum.
Genet., 21, 254–276.
47. Göring, H.H. and Terwilliger, J.D. (2000) Linkage analysis in the
presence of errors III: marker loci and their map as nuisance parameters.
Am. J. Hum. Genet., 66, 1298–1309.
48. Terwilliger, J.D. (2000) A likelihood-based extended admixture model of
oligogenic inheritance in ‘model-based’ and ‘model-free’ analysis.
Eur. J. Hum. Genet., 8, 399–406.
49. Sobel, E. and Lange, K. (1996) Descent graphs in pedigree analysis:
applications to haplotyping, location scores, and marker-sharing statistics.
Am. J. Hum. Genet., 58, 1323–1337.