Evidence of Dysfunction of a Prefrontal-Limbic Network in Schizophrenia: A Magnetic Resonance Imaging and Regional Cerebral Blood Flow Study of Discordant Monozygotic Twins Daniel R. Weinberger, Richard Suddath,
M.D., Karen Faith Berman, M.D., M.D., and E. Fuller Torrey, M.D.
Objective: The authors phrenia the affected twin
previously reported that in monozygotic almost invariably had a smaller anterior
with
imaging
magnetic
resonance
(MRI),
and
invariably
had
twins discordant pes hippocampus,
less regional
for schizomeasured
cerebral
blood
flow
(rCBF) in the dorsolateralprefrontal cortex duringperformance ofthe Wisconsin Card Sorting Test. The present study was an investigation ofthe relationship between hippocampalpathology and prefrontal hypofunction in the same twin pairs. Method: Nine pairs of monozygotic twins discordant for schizophrenia underwent MRI scanning for determination of anterior hippocampalvolume andxenon-inhalation rCBF testingfor determination of prefrontal physiological activation associated with the Wisconsin Card Sorting Test. Results: The differences within twin pairs on the MRI and rCBF measures were strongly and selectively correlated. Specifically, the more an affected twin differed from the unaffected twin in left hippocampal volume, the more they differed in prefrontal physiological activation during the Wisconsin Card Sorting Test. In the affected twins as a group, prefrontal activation was strongly related to both left and right hippocampal volume. These relationships were not found in the group ofunaffected twins. Conclusions: This finding is consistent with the notion that schizophrenia involves pathology ofand dysfunction within a widely distributed neocortical-limbic neural network that has been implicated in, among other activities, the performance ofcognitive tasks requiring working memory. (Am
J
Psychiatry
1992;
149:890-897)
I
is increasingly appreciated that schizophrenia is associated with deficits in various cognitive functions, especially attention, memory, and “executive” functions (1 ). Although the mechanisms underlying these deficits are uncertain, recent speculation has focused on dysfunction of temporal-limbic and prefrontal brain regions as explanations for at least the deficits in memory and executive functions, respectively. Poor perfonmance on prefrontal-type executive tasks, such as the Wisconsin Card Sorting Test, has been directly linked to physiological hypofunction of the prefrontal cortex in patients with schizophrenia (2-7), and subtle anatomical deviations in anteniomedial temporal lobe
Received July 10, 1 991; revision received Nov. 27, 1991; accepted Dec. 31, 1991. From the Clinical Brain Disorders Branch, intramural Research Program, NIMH. Address reprint requests to Dr. Weinberger, NIMH Neuroscience Center, St. Ehizabeths Hospital, 2700 Martin Luther King Jr. Ave., S.E., Washington, DC 20032. Supported
The authors cedures.
890
in part
thank
by NIMH
grant
Mr. Alex lerrazas
MH-41176.
for help with statistical
pro-
structures
implicated
in memory
processing
have
been
interpreted as a pathological correlate of the impairments on memory tests also manifested by these patients (8-1 1). Furthermore, our group has reported data suggesting that these neurobiological findings might be related (12), and we have speculated that the heart of the matter is dysfunctional neocortical-limbic connectivity (12-15). In a recent investigation of sets of monozygotic twins discordant for schizophrenia, we found that the affected twin almost invariably performed worse on memory tests (8) and on the Wisconsin Card Sorting Test (7, 8) than did the unaffected sibling. Moreover, we also found that the affected twin almost invariably had a smaller anterior pes hippocampus ( 1 6) and invariably had diminished relative prefrontal activity as reflected in a measure of regional cerebral blood flow (nCBF) during performance of the Wisconsin Card Sorting Test (7). In this investigation we examined the nelationship between these putative abnormalities in hippocampal anatomy and prefrontal cortical physiology
in this group
of monozygotic
Am
J
twins
Psychiatry
in an effort
149:7,
July
to test
1992
WEINBERGER,
directly
the
limbic
network.
hypothesis
of a dysfunctional
pus junction. The volumes of the pes hippocampus on both the right and left sides were determined. To indude a hypothetically “neutral” MRI measure (i.e., a
neocortical-
measure
that
did
not
discriminate
affected
from
Subjects The study group consisted of nine pains of monozygotic twins discordant for schizophrenia who were recruited as part of a multidimensional study of schizophrenia
in monozygotic
twins.
These
nine
pairs
are
all
of the discordant twin pairs who had undergone both magnetic resonance imaging (MM) and rCBF measurement. Characteristics of these twin pairs were described in detail previously (7, 8, 16). In brief, monozygosity was confirmed by analysis of 19 RBC antigens. Psychiatric diagnosis was established by structured diagnostic interview (SCID parts I and II [17]). The affected memben of each twin pair fulfilled the DSM-III-R criteria for chronic schizophrenia, and the unaffected member had no evidence of major psychopathology and did not fulfill criteria for an axis I psychiatric disorder. Three pains were women, and six pairs were men; their mean age was 32 years (range, 25-44 years). Each twin pair had been discordant for at least 4 years when the study began and had remained discordant for an additional 3year follow-up period. The affected twins were all taking antipsychotic medications at the time of the study. They were in excellent general medical health and had no sign of dehydration or malnutrition. They had been screened for major medical problems and for history of noteworthy
drug
abuse,
head
trauma,
and
other
neuno-
logical problems. The results of rCBF measurements of this group have been reported in detail elsewhere (7). The MM scans of these subjects were part of a study of a larger group of monozygotic twins discordant for schizophrenia,
the MM
and
scans
have
the
anatomical
also
been
measurements
reported
previously
from
measures
have
Regional
Cerebral
Resonance
been
given
Blood
elsewhere
(16).
Flow
The method for determining rCBF by xenon inhalation has also been described in detail elsewhere (2, 4, 7). In brief, the procedure involved a 60-second inhalation of a mixture of xenon-133 gas and air followed by 14 minutes
of “washout,”
during
which
only
room
first
measurement
was
made
during
an eyes-closed
unitless
Initial lapsed
measure
of cortical
slope values from into five regions,
7). The
raw
rCBF
data
gray
matter
32 cortical as previously have
been
perfusion
given
elsewhere
Analysis
Imaging The purpose of this study tionship between anatomical
was to examine the reladifferences in the volume
of the
physiological
unaffected
In the
prior
anatomical
study
(16), the volume of the anterior pes hippocampus was the most robust discriminator of affected from unaffected twin and was selected a priori as the anatomical measure of interest for this investigation. This measure was determined by summing the areas of the pes hipin the
including
J
Psychiatry
first
four
a slice through
149:7,
(7).
(16).
image
pocampus
(18).
regions were coldescribed (2, 4,
a relationship. 1 For each twin pair an anatomical difference was computed by subtracting the pes hippocampal ume in the affected twin from the same measure
system.
rest-
ing condition that served to acclimate the subjects to the testing procedure. The second and third conditions were an automated version of the Wisconsin Card Sorting Test and a simple sensonimotor control task. They were presented in counterbalanced sequence to control for the possibility of an order effect. The control task was designed to serve as a nonspecific sensonimotor baseline activation condition so that rCBF during the Wisconsin Card Sorting Test could be compared with a control value (2). We calculated rCBF as the initial slope, a
The method for making anatomical measurements from MRI scans has been described in detail elsewhere (16). In brief, we used spin-echo coronal sections weighted for T1 relaxation time (repetition time=800 msec, echo time=20 msec); the sections were S-mmthick contiguous slices parallel to the floor of the fourth ventricle and were acquired with the same 1 .S-tesla magnet (General Electric Signa) for each of the subjects. We used an interactive method for outlining structures of interest on digitized images displayed on a computer analysis
air was
inhaled. Regional cerebral blood flow was calculated (by using a biexponential model for gray and white matter compartments) from the radioactivity desatunation curves generated by 32 extracranial gamma ray detectors. Each subject underwent three consecutive rCBF procedures separated by approximately 30 minutes and carried out under three different testing conditions. The
Statistical Magnetic
Am
unaf-
fected twin), we derived cerebral volume by adding the total cerebral areas in the four slices through the antenor hippocampus. The raw data concerning these
METHOD
dala,
El AL.
SUDDATH,
BERMAN,
July
slices
posterior
to the
the amygdala-hippocam-
1992
amyg-
pes hippocampus
and
differences
in prefrontal Sorting Test schizophrenia.
activation during the Wisconsin Card within monozygotic pains discordant for To achieve this goal, we selected two meas-
ures
as those
a priori
most
likely
to test
the presence
.
twin.
If we
is an ideal subject twin in each pair, can
be taken
putative phrenia activation
as a measure
anatomical in the
2. For each
assume
associated
the
unaffected
value volin the twin
with the affected coanatomical difference
of the
pathology
affected
pair
that
for comparison the within-pair
of
relative
extent
associated
with
of the
schizo-
twin.
we computed with
the
the difference Wisconsin
Card
in rCBF Sorting
891
PREFRONTAL-UMBIC
DYSFUNCTION
TABLE 1. Within-Pair Monozygot
Ic
IN SCHIZOPHRENIA
Differences in MRI AnatomiCal for Schizophrenia Within-Pair
Difference
Volume
Twin
(cm3)a
Cerebral
Twin
Hippo-
Hippocampus
Volume Measure
Pair
-0.6 -5.2 -2.2 -8.5 -4.3 11.0 -0.7 4.9 -1.9 in unaffected
campus
1 2
38 38
F F
3
33
M
4 S 6 7 8
31 25 28 44 28
F M M M M
0.35 0.06 0.14 0.16 0.24 0.35 0.36 0.24
M
0.03
0.14 -0.02 0.16 0.20 0.01 0.76 0.29 0.08 0.13
twin was subtracted
from volume
9
27
in affected
First,
each
performance
individual’s
of the
determined.
This
rCBF
Wisconsin
was
activation
Card
accomplished
rCBF data during the control task during the Wisconsin Card Sorting
Sorting
related Test
by subtracting from Test.
during
rCBF
studies
(2, 19).
this
the
differ-
ence in rCBF “activation” for the affected twin was subtracted from the analogous value for the unaffected twin. This within-pair physiological difference can be taken as a measure of the relative extent of the cognition-linked dorsolateral prefrontal cortical hypofunction in the affected twin. The correlation of these within-pair differences in left and night pes hippocampal volumes and within-pair differences in rCBF activation was examined by several procedures, both parametric and nonparametnic. Simplc nonparametnic correlations were performed with the Spearman rho test. Because cerebral perfusion is highly intercorrelated across brain regions and the majority of the variance in regional perfusion is accounted for by the level of global brain perfusion (20), correlations, if found, may not reflect an anatomical nelationship with cortical neuronal activity that is uniquely regional. Alternatively, such correlations may be obscured by the “noise” introduced by global perfusion variance. To account for these possibilities, we performed
a partial
correlation
procedure
to address
the
principal theoretical question of this study, namely, whether and to what extent the MRI data relate to rCBF data (and, by inference, neuronal activity) that are uniquely regional (i.e., not accounted for by whole brain CBF). This approach is analogous to other methods, such as analyses of covaniance (20), that attempt to highlight local brain activity by adjusting regional data for the level of global flow, but the partial correlation approach addresses the strength of the relationship between
procedure the effect
892
variables
rather
involved using of the differences
than
differences
linear regression in whole brain
in means.
Temporal
Parietal 4.0 -4.3 -0.5 -7.6
1 2 3 4 S
23.4 -7.2 1.7 -10.2 4.3
10.7 -5.7 0.1 -7.8 4.0
8.6 -4.8 -0.1 -2.3
6
12.9
10.1
-2.3
7 8 9
19.6 -8.8 4.0
aRgionah
-6.2
8.8 -9.2 -11.5
cerebral
blood
Parieto-
Whole
Occipital
Brain
5.8 -1.8 -9.3 -4.7
8.7
7.7
10.5 -8.8 -9.5
during
10.5 -4.5 -0.7 -8.2 1.9
2.1
2.9
7.9 -4.9 -9.4
flow
Activationa
10.3
11.2 -6.9 -11.4
11.5 -8.2 -8.6 task was subtracted
a control
to
the rCBF data This “subtnac-
Second,
in rCBF
was
tion” method has become widely accepted as the “gold standard” for isolating behavior-specffic regional brain activity
Central
Difference
from rCBF during the Wisconsin Card Sorting Test. The difference value for the affected twin was subtracted from the difference for the unaffected twin. The values shown here are (unitless) initial slope values from 32 cortical regions collapsed into five regions the raw data have been given elsewhere (7).
nwn.
Test.
Prefrontal
Twins Discorda nt for
Test for Monozygotic
Within-Pair
Right
Sex
aVohume
Wisconsin Card Sorting Schizophrenia
in
Left Age (years)
Pair
Measures for
Twins Discordant
The
to remove CBF on the
regional the MRI
differences
and
difference
data.
as a “semipartial
correlating
This
the
residuals
procedure
correlation”
(21).
with
is referred
Because
to
of the nela-
tively small number of subjects, the residual rCBF diffenences and the MM differences were also subjected to a nonparametric Spearman correlational analysis. The statistical procedures were performed on a mainframe computer with Statistical Analysis System programs (SAS Institute, Cary, N.C.).
RESULTS
The twin pains’ age, sex, and in the anatomical measures are within-pair differences in rCBF Wisconsin Card Sorting Test are hippocampal
ences
anatomical
correlated
p0.7,
(rho=0.82,
pO.lS).
pected, rCBF differences were highly correlated between all regions and with mean whole brain CBF (in all cases, nho>0.90, p0.20
0.13
>0.70
-0.02 -0.63 -0.30
>0.90 0.43
0.04 -0.71 -0.29
>0.90 0.40
0.33 -0.60 0.10
>0.30 0.70
-0.19 0.44 -0.35
>0.60 >0.20 >0.30
aff7
tions were found for both left and right hippocampus and residual panietal flow. This correlation suggests that more hippocampal pathology is associated with greater panietal activation. The results of a test for a relationship between within-pair differences in the cerebral volume measure and residual prefrontal rCBF activation were not significant (Spearman nho=0.40, p=O.29; Pearson r=0.26, p=O.SO). Post hoc correlations were performed to address whether the results depended on the presence of illness per se. In the unaffected twins as a group, there were no significant correlations between pes hippocampal volume and rCBF activation during the Wisconsin Card Sorting Test, either without or with partial correlations (p>O.2 in each case). In the affected group, after the partial correlation procedure was performed prefrontal activation correlated strongly with both the left (rho= 0.75, p=O.O2) and right (rho=0.82, p