Plexin B3 is genetically associated with verbal performance and white matter volume in human brain

Abstract

The presence of genetic influences on cognitive performance and brain volume is well established. However, specific genetic determinants of the variance of these quantitative traits are not yet known. Plexins act as receptors for semaphorins and are implicated in axon guidance, which is a key process in brain development. We have previously shown that plexin B3 is a highly potent stimulator of neurite outgrowth, which makes its gene PLXNB3 an intriguing candidate gene for traits related to human brain development and cerebral connectivity. We identified several polymorphisms in PLXNB3 predicting changes of amino acids (V598I, E1156D and V1596E) conserved at the corresponding positions of the orthologs in mouse and chimpanzee. PLXNB3 was genotyped in 303 healthy volunteers and 42 male patients with schizophrenia. Cognitive performance was measured with the vocabulary test (Wortschatztest (WST)), a method to estimate roughly general intelligence (g). Brain morphology was characterized by magnetic resonance imaging. Compared to subjects not carrying the modern, human-specific haplotype A, carriers of A scored higher in vocabulary test (WST) irrespective of diagnosis (P=0.0004). This effect could be observed in three independent groups (healthy males: P=0.048; schizophrenic males: P=0.034 and healthy females: P=0.037). Additionally, the haplotype A was associated with increased volume of brain white matter that in turn correlated with performance in the vocabulary test. These findings suggest that plexin B3 may influence cognitive performance, and the development of white matter in vivo in a way similar to its known stimulating effect on neurite outgrowth in vitro. These novel observations warrant further replication in independent samples.

Full text

ORIGINAL ARTICLE
Plexin B3 is genetically associated with verbal
performance and white matter volume in human brain
D Rujescu
1,4
, EM Meisenzahl
1,4
, S Krejcova
2
, I Giegling
1
, T Zetzsche
1
, M Reiser
3
, CM Born
3
,
H-J Mo
¨
ller
1
, A Veske
2,5
, A Gal
2
and U Finckh
2,6
1
Department of Psychiatry, Ludwig-Maximilians-University, Munich, Germany;
2
Institute of Human Genetics, University
Hospital Hamburg-Eppendorf, Hamburg, Germany and
3
Department of Radiology, Ludwig-Maximilians-University,
Munich, Germany
The presence of genetic influences on cognitive performance and brain volume is well
established. However, specific genetic determinants of the variance of these quantitative traits
are not yet known. Plexins act as receptors for semaphorins and are implicated in axon
guidance, which is a key process in brain development. We have previously shown that plexin
B3 is a highly potent stimulator of neurite outgrowth, which makes its gene PLXNB3 an
intriguing candidate gene for traits related to human brain development and cerebral
connectivity. We identified several polymorphisms in PLXNB3 predicting changes of amino
acids (V598I, E1156D and V1596E) conserved at the corresponding positions of the orthologs
in mouse and chimpanzee. PLXNB3 was genotyped in 303 healthy volunteers and 42 male
patients with schizophrenia. Cognitive performance was measured with the vocabulary test
(Wortschatztest (WST)), a method to estimate roughly general intelligence (g). Brain
morphology was characterized by magnetic resonance imaging. Compared to subjects not
carrying the modern, human-specific haplotype A, carriers of A scored higher in vocabulary
test (WST) irrespective of diagnosis (P = 0.0004). This effect could be observed in three
independent groups (healthy males: P = 0.048; schizophrenic males: P = 0.034 and healthy
females: P = 0.037). Additionally, the haplotype A was associated with increased volume of
brain white matter that in turn correlated with performance in the vocabulary test. These
findings suggest that plexin B3 may influence cognitive performance, and the development of
white matter in vivo in a way similar to its known stimulating effect on neurite outgrowth in
vitro. These novel observations warrant further replication in independent samples.
Molecular Psychiatry (2007) 12, 190–194. doi:10.1038/sj.mp.4001903; published online 10 October 2006
Keywords: plexin B3; adhesion molecule; genetic association study; verbal performance;
cognitive function; brain morphology
Introduction
The presence of genetic influences on cognitive
performance is well established and the various
studies converge on heritability estimates between
0.60 and 0.80.
1,2
Genetic factors also accounted for
most of the individual differences in whole brain
(90%) and white matter (88%) volume.
3
The correla-
tions between brain volume and cognitive perfor-
mance are about 0.40 as demonstrated in 14 studies
consisting of about 700 individuals.
4,5
Furthermore,
cognitive performance is related to the volume of
brain white matter, and this correlation seems to be
due to genetic factors.
6
Establishing that the correla-
tion between brain volumes and cognitive abilities is
mediated by shared genetic factors is only the first
step in unveiling the relation between them. The next
step is to identify specific genes that influence both
brain volume and cognitive abilities. This approach
can also lead to the identification of schizophrenia-
related genes because several brain volume and
neuropsychological alterations have been described
as schizophrenia endophenotypes. The rationale for
their use in schizophrenia gene discovery is that these
endophenotypes are more elementary compared to
clinical phenotypes.
Genes of the plexin family represent intriguing
candidates. Plexins act as transmembrane receptors
for semaphorins and are implicated in axon guidance,
which is a key process in neuronal network forma-
tion.
7
We further characterized a member of this
family, plexin B3 and demonstrated its predominant
neuronal expression in the brain. Expression of
Received 7 September 2005; revised 4 January 2006; accepted 23
January 2006; published online 10 October 2006
Correspondence: PD Dr med D Rujescu, Department of Psychiatry,
Ludwig-Maximilians-University, Nussbaumstr. 7, Munich 80336,
Germany.
E-mail: Dan.Rujescu@med.uni-muenchen.de
4
These authors contributed equally to this work.
5
Current address: Institute of Gene Technology, Tallinn Technical
University, Tallinn, Estonia.
6
Current address: Laboratoriumsmedizin Dortmund, Dortmund,
Germany.
Molecular Psychiatry (2007) 12, 190–194
&
2007 Nature Publishing Group All rights reserved 1359-4184/07
$
30.00
www.nature.com/mp

plexin B3 in the brain was detected throughout mouse
embryonic and postnatal development as well as in
the adult human brain. In addition, we showed that
plexin B3 is a highly potent stimulator of neurite
outgrowth.
8
This prompted us to search for poly-
morphisms in PLXNB3, the human gene for plexin
B3. We identified several single-nucleotide sequence
polymorphisms (SNPs). Three of them, coding for
V598I (exon 9), E1156D (exon 20) and V1596E (exon
28) result in changes of evolutionarily conserved
amino acids of plexin B3. An extended haplotype
analysis was subsequently performed in male subjects
using several SNPs including the mentioned ones.
These polymorphisms cover a genomic interval of
13.271 kb. Genotyping of the SNPs coding for V598I
and E1156D allowed the determination of the most
common PLXNB3 haplotypes A, B and C. Under the
assumption of unique mutation events driving evolu-
tion, human haplotype B may have evolved from
haplotype C, and ‘modern’ haplotype A from haplo-
type B. Based on comparison with published mouse
and chimpanzee sequences, haplotype C is assumed
to represent the archetypic form. In 27% of the
chromosomes analyzed, we found evidence of ances-
tral recombination events within the 13.27 kb region
covered by the SNPs. Therefore, a significant propor-
tion of the haplotypes defined by the SNPs coding for
V598I and E1156D do not extend to neighboring
genes located B19.5 kb upstream (ABCD1)and
B6.4 kb downstream (IDH3G)toPLXNB3. This sup-
ports the view that the SNPs coding for V598I and
E1156D are well suited for haplotype-based associa-
tion studies on PLXNB3. We investigated the associa-
tion between the haplotypes based on the SNPs coding
for V598I and E1156D and cognitive abilities in 345
healthy volunteers and male schizophrenic patients.
Furthermore, the association with brain morphology
was examined in a subset of these individuals.
Materials and methods
Neuropsychological testing and magnetic resonance
imaging analysis of human subjects
Unrelated healthy Caucasians (115 males, 140 females,
mean age 45.4716.0 years, age range 19–79 years),
randomly collected from the general population
of Munich (Germany), as well as a group of 48
healthy, right-handed male volunteers, matched for
age (mean 30.578.9 years) and educational achieve-
ment with a group of 42 male right-handed inpatients
with schizophrenia (mean age 30.279.0 years), com-
pleted the German Wortschatztest ‘WST’.
9
WST was
constructed from the HAWIE-R (German version of
the Wechsler Adult Intelligence Scale, Revised)
10
and
is an economical method to estimate roughly general
intelligence (g). WST consists of 42 lists each
comprising five pseudowords as distractors and one
meaningful, real word to be recognized among the
distractors. The Spearman–Brown coefficient of the
split-half reliability is r = 0.95 and the Cronbach alpha
r = 0.94. This test allows to estimate verbal intelli-
gence quotient. The correlation between the WST
and the vocabulary subtest of the HAWIE-R
10
is
about r = 0.81, and the vocabulary subtest in turn is
correlated (r = 0.73) with g.
11
In addition, all subjects
of the matched subgroup of 48 unaffected controls
and 42 patients with schizophrenia underwent
scanning by magnetic resonance imaging (MRI) as
described previously.
12,13
In brief, MRI images were
obtained (1.5 T Magnetom Vision, Siemens, Germany)
using a coronal T2- and proton density-weighted
Dual-Echo-Sequence (repetition time (TR) 3710 ms/
echo time (TE) 22/90 ms; total acquisition time:
9 min; number of acquisitions: 1; field of view (FOV)
230 mm; matrix 256  256; slice thickness 3 mm)
and a 3D-MPRAGE sequence (TR/TE 11.6/4.9 ms;
total acquisition time: 8 min; number of acquisitions:
1; FOV 230 mm; matrix 512  512; slice thickness
1.5 mm). For further image processing with size
reduction from 16 to 8 bit and transformation to a
uniform matrix of 256  256 on 192 slices of 1.5 mm
slice thickness, the commercial software package
ANALYZE was used (Biomedical Imaging Resource,
Mayo Foundation, Rochester, MN, USA). All data sets
were realigned and resampled three-dimensionally
according to the coordinates of Talairach using the
program BRAINS (Brain Research: Analysis of
Images, Networks and Systems; developed by NC
Andreasen and co-workers).
14–17
Segmentation of the
MRI data sets in gray and white matter tissue classes
and external and internal cerebrospinal fluid (CSF)
compartments was performed by using BRAINS. This
was performed for the total brain volume and for each
cerebral lobe of both sides separately.
Genotyping
Genomic DNA was prepared from 10 ml blood using
the QIAamp DNA Blood Maxi Kit (Quiagen, Hilden,
Germany) following the supplier’s instructions. The
SNPs 1906G > A (V598I) and 3582G > C (E1156D) were
genotyped by restriction fragment analysis of poly-
merase chain reaction amplicons of exons 9 (220 bp)
and 20 (304 bp), respectively. The change 1906G > A
leads to loss of the Tth111I restriction site in exon 9,
with G-allelic bands of 161 and 59 bp after cleavage
by Tth111I, whereas 3582G > C leads to loss of the
EcoR109I site in exon 20, with G-allelic bands of 226
and 78 bp after cleavage by EcoR109I. As PLXNB3 is
located on the X-chromosome, haplotypes may be
determined unambiguously in male subjects. The
haplotypes defined by the two SNPs were defined as
following: A, I598-D1156; B, V598-D1156 and C,
V598-E1156. Based on mouse and chimpanzee
sequences and under maximum parsimonious condi-
tions assumed for the evolution of human haplotypes,
haplotype C was assumed to represent the archetypic
form and haplotype A to represent the most ‘modern’
form.
Statistical analysis
Statistics were performed using the Statistical Pack-
age for Social Sciences (SPSS) 13.0 Software (SPSS
Plexin B3, verbal performance and white matter volume
D Rujescu et al
191
Molecular Psychiatry

Inc., Chicago, IL, USA, 2004). Socio-demographic
covariates of subgroups stratified by genotype were
analyzed by t-test or w
2
-test. Non-conditional linear
regression analysis was first computed in all men
(n = 205) for WST by integrating the factor genotype
(haplotype A vs B or C) controlled for diagnosis, age
and professional achievement (low, middle, high).
The same procedure was then carried out separately
for healthy men, schizophrenic men and healthy
women. MRI data of the 90 subjects were first
analyzed using separate univariate analyses for brain
white matter, gray matter and volume of CSF,
including diagnosis and genotype controlled for age
and intracranial volume. For white matter and CSF
volume, we also performed two separate repeated
multivariate analysis of covariances (MANCOVAs)
assessing the main and interaction effects of the
within-subjects factors brain region (frontal, temporal,
parietal, occipital) and hemisphere (left, right), and
the between-subjects factors diagnosis (patients,
controls) and genotype. The covariates age and intra-
cranial volume were added to the analysis. All
analyses used two-tailed estimation of significance.
The significance level was set at 0.05.
Results
Linear regression analysis entering age, clinical
phenotype (schizophrenia vs healthy), professional
achievement (low, middle, high) and haplotype
(A vs B/C for men and A-carrier vs rest for women)
revealed an association of modern haplotype A with
higher scores in WST (all men (n = 205): P = 0.0004,
r
2
= 0.036; haplotype A (n = 119); mean WST score
(s.d.) 33.66 (4.9); haplotype B/C (n = 86) mean WST
score (s.d.) 31.38 (8.2)). As shown in Table 1
and Figure 1, this association was detected in
three independent subsamples: schizophrenic men:
P = 0.039, r
2
= 0.057; healthy men: P = 0.048, r
2
= 0.016
and healthy women: P = 0.037, r
2
= 0.026 (schizophre-
nic men: haplotype A (n = 26) mean WST score (s.d.)
32.62 (6.5); haplotype B/C (n = 16) mean WST score
(s.d.) 23.31 (11.0); healthy men: haplotype A (n = 93)
mean WST score (s.d.) 33.96 (4.3); haplotype B/C
Table 1 Linear regression analysis of association between verbal performance and PLXNB3 haplotype
Non-standardized coefficient Standardized coefficient TP-value
Beta s.e. Beta
All males (n = 205)
Plexin haplotype
a
2.644 0.737 0.204 3.590 0.0004 (r
2
= 0.036)
Professional status
b
3.993 0.466 0.505 8.566 < 0.0005
Age 0.070 0.025 0.171 2.760 0.006
Diagnosis 3.035 0.962 0.191 3.154 0.002
Sample 1: Healthy males (n = 163)
Plexin haplotype
a
1.396 0.700 0.141 1.994 0.048 (r
2
= 0.016)
Professional status
b
2.978 0.451 0.488 6.601 < 0.0005
Age 0.033 0.023 0.107 1.435 0.153
Sample 2: Schizophrenic males (n = 42)
Plexin haplotype
a
5.021 2.346 0.258 2.140 0.039 (r
2
= 0.057)
Professional status
b
6.200 1.401 0.531 4.427 < 0.0005
Age 0.261 0.118 0.246 2.202 0.034
Sample 3: Healthy females (n = 140)
Plexin haplotype
c
1.588 0.753 0.163 2.108 0.037 (r
2
= 0.026)
Professional status
b
2.561 0.511 0.437 5.016 < 0.0005
Age 0.026 0.022 0.102 1.168 0.245
Dependent variable: Wortschatztest sum score.
a
Haplotype males: A vs B or C.
b
Low, middle, high.
c
Haplotype females: A-carrier vs A-negative.
A B/C AB/C
healthy males schizophrenic males healthy females
WST sum scores
A-pos A-neg
35
34
33
32
35
30
25
20
15
35
34
33
32
31
Figure 1 Association between verbal performance (WST,
sum scores, s.e.) and PLXNB3 haplotypes (A vs B or C;
A-positive vs A-negative) in three independent subsamples
(healthy males: n = 163, P = 0.048, r
2
= 0.016; schizophrenic
males: n = 42, P = 0.039, r
2
= 0.057; healthy females: n = 140,
P = 0.037, r
2
= 0.026).
Plexin B3, verbal performance and white matter volume
D Rujescu et al
192
Molecular Psychiatry

(n = 70) mean WST score (s.d.) 33.23 (6.2); healthy
women: haplotype A-positive (n = 111) mean WST
score (s.d.) 34.26 (4.0); haplotype A-negative (n = 29)
mean WST score (s.d.) 32.59 (3.5)).
The results of the morphometric brain MRI data in
dependence of diagnosis and genotype of 42 male
patients with schizophrenia and 48 control male
subjects are shown in Table 2. Irrespective of the
diagnosis, presence of haplotype A was associated
with higher volume of brain white matter (repeated
measurement MANCOVA F = 4.275, d.f. = 1/85,
P = 0.042; haplotype A (n = 58) mean volume (s.d.)
392.4 (38.2); haplotype B/C (n = 33) mean volume
(s.d.) 371.2 (39.9)). In addition, there was a correlation
of the volume of white matter with performance
in WST (r = 0.29, P = 0.007; controlled for: age,
professional status, diagnosis).
CSF volume was associated with both diagnosis
(higher in patients) and genotype (lower in haplotype
A-carriers); (repeated measurement MANCOVA
F = 5.302, d.f. = 1/85, P = 0.024; haplotype A (n = 58)
mean volume (s.d.) 199.7 (34.6); haplotype B/C
(n = 33) mean volume (s.d.) 214.5 (50.6)).
Discussion
The main finding of this study is that naturally
occurring structural variability of plexin B3, which is
involved in axon guidance and neuronal network
formation, may be related to normal variance of
human-specific cognitive function and brain white
matter volume.
Neurological disorders linked to Xq28, the chromo-
somal band harboring PLXNB3, include a group of
diseases caused by mutations in L1CAM, the gene
coding for the neuronal cell adhesion molecule L1.
This protein has been shown to be involved in axon
guidance and fasciculation, and mutations in this
gene reduce or abolish L1-dependent neurite out-
growth in vitro.
18
We observed that plexin B3 is an
even more potent stimulator of neurite outgrowth
when compared to L1.
8
Therefore, PLXNB3 may be
considered an interesting candidate gene for
the inborn variability in human-specific cognitive
functions and brain volume. Several coding and non-
coding SNPs allowed an evolutionary analysis of
common haplotypes specific for PLXNB3. Two SNPs
underlying major haplotypes of PLXNB3 and used for
haplotyping in this study lead to human-specific
changes of residues otherwise conserved through
evolution. In both schizophrenic and mentally
healthy males, absence of the most modern and
human-specific haplotype A was associated with
reduced performance in WST and reduced volume
of brain white matter that in turn correlated with
reduced performance in WST. As WST gives a rough
estimate for g, these data are in line with recent
findings of a correlation between white matter and g.
6
The association between PLXNB3 haplotype and
white matter volume was independent of the diag-
nosis and was accompanied by an association of the
haplotype with CSF volume in the opposite direction.
As CSF volume was also associated with diagnosis
and there was no interaction between genotype and
diagnosis for CSF volume, we assume a primary
influence of PLXNB3 haplotype on white matter
volume that in turn indirectly influences CSF volume
independently of the previously known association of
schizophrenia with CSF volume.
13
Thus, we were
unable to find evidence for our hypothesis that
genetic variations in plexin B3 might be especially
associated with brain volume changes and neuro-
psychological impairments in schizophrenia, that is
that the association is stronger in the patients in com-
parison to healthy subjects. On the other hand, there
is a trend for a genotype effect on gray matter and the
interaction of diagnosis by genotype on gray matter is
also suggestive though nonsignificant. Thus, it might
well be that our study is underpowered to detect a
schizophrenia-related effect on gray matter. This
important issue should be followed up in replication
studies in larger samples.
It is well worthy to emphasize that the variations in
PLXNB3 were associated with verbal performance in
three independent samples, schizophrenic patients,
healthy males and healthy females. Whereas these
replications add much credibility to the findings, the
associations of the PLXNB3 haplotype with verbal
performance and white matter volume were not
strong. Although this is not unexpected, as both traits
Table 2 Results of the variance analysis for morphometric data (volumes)
Factor
a
CSF White matter Gray matter
d.f. F P-value d.f. F P-value d.f. F P-value
D 1/85 6.527 0.012 1/85 0.001 0.982 1/85 0.026 0.873
G 1/85 5.302 0.024 1/85 4.275 0.042 1/85 2.362 0.128
D  G 1/85 0.966 0.328 1/85 0.050 0.823 1/85 2.699 0.104
Abbreviations: CSF, cerebrospinal fluid; D, diagnosis; G, genotype.
Interactions with (R) region and (H) hemisphere are not presented.
a
 , interaction.
P < 0.05 in bold.
Plexin B3, verbal performance and white matter volume
D Rujescu et al
193
Molecular Psychiatry

are supposed to be under polygenic influence, these
potentially highly important findings warrant further
replication before generalization. Furthermore, WST
was the only cognitive test available from all cohorts.
It remains to be investigated which cognitive dimen-
sion correlates most specifically and thus possibly
stronger with the PLXNB3 haplotype. It should be
emphasized that also these findings warrant further
multiple replications in order to avoid spurious
associations, which are common to genetic associa-
tion studies.
The mechanisms by which the V598I and/or
E1156D missense variations or other linked sequence
variants might actually act on cognitive performance
and white matter volume can only be speculated
about and require further investigation. If the associa-
tion findings can be confirmed in replication studies,
detailed functional analysis of structural variants of
plexin B3 may help to identify the polymorphisms
most important for the association findings.
To our knowledge, this is the first study that
examines the relationship between PLXNB3 haplo-
types and human-specific cognitive abilities and
white matter volume and one of the first reports on
the influence of common genetic variations on both,
brain structure and function. The association
of human-specific higher cognitive functions like
verbal performance and brain white matter volume
with human-specific polymorphisms affecting con-
served amino acids of a protein with a predominant
neuronal expression is intriguing and deserves
further investigation.
References
1 Plomin R, Petrill S. Genetics and intelligence: what’s new?
Intelligence 1997; 24: 53–77.
2 Bouchard TJ. Genetic and environmental influences on adult
intelligence and special mental abilities. Hum Biol 1998; 70:
257–279.
3 Baare WF, Hulshoff Pol HE, Boomsma DI, Posthuma D, de Geus EJ,
Schnack HG et al. Quantitative genetic modeling of variation in
human brain morphology. Cereb Cortex 2001; 11: 816–824.
4 Vernon PA, Wickett JC, Banzana PG, Stelmack RM. In: Sternberg RJ
(ed). Handbook of Intelligence. University Press: Cambridge, 2000,
pp 245–264.
5 Plomin R, Kosslyn SM. Genes, brain and cognition. Nat Neurosci
2001; 4: 1153–1154.
6 Posthuma D, De Geus EJ, Baare WF, Hulshoff Pol HE, Kahn RS,
Boomsma DI. The association between brain volume and intelli-
gence is of genetic origin. Nat Neurosci 2002; 5: 83–84.
7 Tamagnone L, Artigiani S, Chen H, He Z, Ming GI, Song H et al.
Plexins are a large family of receptors for transmembrane,
secreted, and GPI-anchored semaphorins in vertebrates. Cell
1999; 99: 71–80.
8 Hartwig C, Veske A, Krejcova S, Rosenberger G, Finckh U. Plexin
B3 promotes neurite outgrowth, interacts homophilically, and
interacts with Rin. BMC Neurosci 2005; 25:53.
9 Schmidt KH, Metzler P. Wortschatztest (WST). Beltz Test:
Weinheim, Germany, 1992.
10 Tewes U. HAWIE-R Hamburg-Wechsler Intelligenztest fu¨r Erwach-
sene, Revision. Hogrefe: Goettingen, Germany, 1991.
11 Satzger W, Fessmann H, Engel RR. Liefern HAWIE-R, WST und
MWT-B vergleichbare IQ-Werte? Zeitschrift fu¨r Diff und Diag Psy
2002; 23: 159–170.
12 Meisenzahl EM, Rujescu D, Kirner A, Giegling I, Kathmann N,
Leinsinger G et al. Association of an interleukin-1beta genetic
polymorphism with altered brain structure in patients with
schizophrenia. Am J Psychiatry 2001; 158: 1316–1319.
13 Rujescu D, Meisenzahl EM, Giegling I, Kirner A, Leinsinger G,
Hegerl U et al. Methionine homozygosity at codon 129 in the prion
protein is associated with white matter reduction and enlargement
of CSF compartments in healthy volunteers and schizophrenic
patients. Neuroimage 2002; 15: 200–206.
14 Andreasen NC, Cohen G, Harris G, Cizadlo T, Parkkinen J, Rezai K
et al. Image processing for the study of brain structure and
function: problems and programs. J Neuropsychiatr Clin Neurosci
1992; 4: 125–133.
15 Andreasen NC, Cizadlo T, Harris G, Swayze II V, O’Leary DS,
Cohen G et al. Voxel processing techniques for the antemortem
study of neuroanatomy and neuropathology using magnetic
resonance imaging. J Neuropsychiatr Clin Neurosci 1993; 5:
121–130.
16 Harris G, Andreasen NC, Cizadlo T, Bailey JM, Bockholt HJ,
Magnotta VA et al. Improving tissue classification in MRI: a three-
dimensional multispectral discriminant analysis method with
automated training class selection. J Comput Assist Tomogr 1999;
23: 144–154.
17 Magnotta VA, Heckel D, Andreasen NC, Cizadlo T, Corson PW,
Ehrhardt JC et al. Measurement of brain structures with
artificial neural networks: two- and three-dimensional applica-
tions. Radiology 1999; 211: 781–790.
18 Michelson P, Hartwig C, Schachner M, Gal A, Veske A, Finckh U.
Missense mutations in the extracellular domain of the human
neural cell adhesion molecule L1 reduce neurite outgrowth of
murine cerebellar neurons. Hum Mutat 2002; 20: 481–482.
Plexin B3, verbal performance and white matter volume
D Rujescu et al
194
Molecular Psychiatry