doi:10.1093/brain/awh496
Brain (2005), 128, 1584–1594
Functional MRI study of diencephalic amnesia in
Wernicke–Korsakoff syndrome
M. Caulo,1,2 J. Van Hecke,2,4 L. Toma,3 A. Ferretti,1,2 A. Tartaro,1,2 C. Colosimo,1,2
G. L. Romani1,2 and A. Uncini3
1
Department of Clinical Sciences and Bio-imaging, University ‘G. d’Annunzio’, 2ITAB Institute for Advanced Biomedical
Technologies, 3Department of Oncology and Neurosciences and Aging Research Center, Ce.S.I. University ‘G. d’Annunzio’
Foundation Chieti-Pescara, Chieti-Pescara, Italy and 4Collaborative Antwerp Psychiatric Research Institute, CAPRI,
University of Antwerp, Antwerp, Belgium
Correspondence to: Prof. Antonino Uncini, Clinica Neurologica, Ospedale ‘SS. Annunziata’, Via dei Vestini,
I-66013, Chieti, Italy
E-mail: [email protected]
Anterograde amnesia in Wernicke–Korsakoff syndrome is associated with diencephalic lesions, mainly in the
anterior thalamic nuclei. Whether diencephalic and temporal lobe amnesias are distinct entities is still not
clear. We investigated episodic memory for faces using functional MRI (fMRI) in eight controls and in a 34-yearold man with Wernicke–Korsakoff syndrome and diencephalic lesions but without medial temporal lobe (MTL)
involvement at MRI. fMRI was performed with a 1.5 tesla unit. Three dual-choice tasks were employed: (i) face
encoding (18 faces were randomly presented three times and subjects were asked to memorize the faces);
(ii) face perception (subjects indicated which of two faces matched a third face); and (iii) face recognition
(subjects indicated which of two faces belonged to the group they had been asked to memorize during encoding). All activation was greater in the right hemisphere. In controls both the encoding and recognition tasks
activated two hippocampal regions (anterior and posterior). The anterior hippocampal region was more
activated during recognition. Activation in the prefrontal cortex was greater during recognition. In the subject
with Wernicke–Korsakoff syndrome, fMRI did not show hippocampal activation during either encoding or
recognition. During recognition, although behavioural data showed defective retrieval, the prefrontal regions
were activated as in controls, except for the ventrolateral prefrontal cortex. fMRI activation of the visual
cortices and the behavioural score on the perception task indicated that the subject with Wernicke–
Korsakoff syndrome perceived the faces, paid attention to the task and demonstrated accurate judgement.
In the subject with Wernicke–Korsakoff syndrome, although the anatomical damage does not involve the MTL,
the hippocampal memory encoding has been lost, possibly as a consequence of the hippocampal–anterior
thalamic axis involvement. Anterograde amnesia could therefore be the expression of damage to an extended
hippocampal system, and the distinction between temporal lobe and diencephalic amnesia has limited value.
In the subject with Wernicke–Korsakoff syndrome, the preserved dorsolateral prefrontal cortex activation
during incorrect recognition suggests that this region is more involved in either the orientation or attention at
retrieval than in retrieval. The lack of activation of the prefrontal ventrolateral cortex confirms the role of this
area in episodic memory formation.
Keywords: Wernicke–Korsakoff syndrome; diencephalic amnesia; fMRI, episodic memory
Abbreviations: AC = anterior commissure; ANOVA = analysis of variance; BOLDc = blood oxygenation level-dependent
contrast; DLPFC = dorsolateral prefrontal cortex; FLAIR = fluid-attenuated inversion recovery; fMRI = functional magnetic
resonance imaging; IFG = inferior frontal gyrus; MMSE = Mini-Mental State Examination; MPFC = medial prefrontal cortex;
MTL = medial temporal lobe; PC = posterior commissure; rCBF = regional cerebral blood flow; ROI = region of interest;
VLPFC = ventrolateral prefrontal cortex; WAIS = Wechsler Adult Intelligence Scale; WK = Wernicke–Korsakoff syndrome
Received September 6, 2004. Revised February 9, 2005. Accepted March 2, 2005. Advance Access publication April 7, 2005
#
The Author (2005). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved. For Permissions, please email: [email protected]
fMRI in Wernicke–Korsakoff syndrome
Introduction
Wernicke encephalopathy is an acute neurological disorder
characterized by ataxia, vestibular dysfunction, a variety of
ocular motility abnormalities, a confusional state and drowsiness, which can progress to a chronic amnesic state called the
Korsakoff syndrome. Both disorders are caused by thiamine
deficiency and are found mainly in chronic alcoholics but also
in cases of persistent vomiting, gastric carcinoma or other
disturbances of the alimentary tract (Victor et al., 1989).
The cardinal feature of the Korsakoff syndrome is severe
and disproportionate anterograde memory loss in an otherwise alert and responsive patient. Patients suffering from this
amnesic syndrome show degeneration of diencephalic
regions: the mammillary bodies and periventricular regions,
especially the anterior thalamic nucleus (Harding et al., 2000).
Anterograde amnesia resulting from diencephalic damage
resembles amnesia following medial temporal lobe (MTL)
damage (Aggleton and Brown, 1999).
Whether these two types of amnesia should be considered
as independent is still a matter of debate. The extensive anatomical connections between the temporal lobe, the mammillary bodies and the thalamus on the one hand and the position
of the diencephalon connecting the MTL and the frontal lobe
on the other suggest a crucial role of the diencephalon in the
network underlying memory formation (Vann and Aggleton,
2004). Both the MTL and the diencephalon could be considered part of a single system for acquiring new declarative
memories. Therefore, both amnesic syndromes could be
variants of a disorder involving the same cerebral network
(Aggleton and Brown, 1999). Alternatively, damage to the
diencephalic structures could result in widespread dysfunction of cortical regions, disrupting memory formation and
retrieval without involvement of the MTL (Mair, 1994).
We used functional MRI (fMRI) to study memory processes for facial episodic recognition to clarify the relationship
between the MTL and diencephalic amnesia in eight healthy
volunteers and in a subject with Wernicke–Korsakoff (WK)
syndrome with diencephalic damage but without MTL
involvement.
Material and methods
Subjects
A 34-year-old, non alcoholic, right-handed man with oesophageal
stenosis after surgery for cancer, persistent vomiting and severe malnutrition was admitted because of gait unsteadiness and confusional
state. Neurological examination performed 3 days later showed
a global confusional state and drowsiness, dysarthria, ophthalmoparesis, nystagmus and ataxia. A clinical diagnosis of Wernicke
encephalopathy was made and thiamine treatment (100 mg/day
intramuscularly) initiated. After 5 days the patient was alert and
attentive but presented retrograde amnesia and severe anterograde
amnesia. Five months after onset, ataxia and ophthalmoparesis
had disappeared, retrograde amnesia shrank, but dense anterograde
amnesia was still present. A neuropsychological examination was
performed 4 weeks and 5 months after onset using the following
tests: the Mini-Mental State Examination (MMSE), the Prose
Brain (2005), 128, 1584–1594
1585
Table 1 Neuropsychological tests and scores 4 weeks
and 5 months after onset
4 weeks
5 months
MMSE (range 0–30)
22
27
Prose Memory test (range 0–28)
1.5
4
*Rey Complex Figure (range 0–36) Immediate 29 Immediate 31
Delayed 3
Delayed 4
Gollin Incomplete Pictures
50
55
Test (range 0–60)
WAIS-R Digit Symbol
79
82
Subtest (range 0–90)
WAIS Information Subtest
17
21
(range 0–29)
Crovitz test (range of mean
0.5
1.5
score for each episode 0–3)
*Copy was correctly performed in both examinations.
Memory test, the Rey Complex Figure, Crovitz test, the Wechsler
Adult Intelligence Scale (WAIS) Information subtest, the Gollin
Incomplete Pictures Test, and the Wechsler Adult Intelligence
Scale—Revised (WAIS-R) Digit Symbol subtest. Results (raw scores)
are reported in Table 1. At 4 weeks the low score obtained on the
MMSE revealed reduced global cognitivity. Poor performances on
the Prose test and the Rey Complex Figure delayed recall test
documented a severe anterograde episodic memory loss. The
good performance observed in the immediate recall on the Rey
Complex Figure test could be explained by the relative preservation
of attention and visuospatial/constructional abilities. Immediate
and non-declarative memory was preserved, as proved by the
scores in the Gollin and Digit Symbol tests. The score obtained in
the WAIS Information subtest demonstrated intact semantic memory storage. A low score on the Crovitz test showed an important
deficit in remote autobiographical memory, also observed during the
clinical interview, which revealed retrograde amnesia spanning
approximately 2 years. At 5 months the global level of cognitive
functions returned within the normal range. We did not find significant changes on the declarative (verbal and non-verbal) memory
test as a whole, although a slight improvement in relative scores was
evident. The scores on semantic and non-declarative tests were
slightly improved. A better performance on the Crovitz test was
concordant with the improvement in retrograde amnesia, which
had declined, as inferred by clinical interview, and spanned approximately 10 months.
Anatomical MRI studies were performed using standard (spinecho, turbo spin-echo and fluid-attenuated inversion recovery;
FLAIR) sequences 7 days, 4 weeks and 5 months after the onset
of symptoms. During the last two sessions, fMRI studies with the
blood oxygenation level-dependent contrast (BOLDc) technique
were performed using a block paradigm to test memory for faces.
Eight healthy right-handed men (30–40 years) were recruited as
controls for the fMRI study. All subjects gave informed consent.
This study had received prior approval from our institutional Ethics
Committee.
Functional MRI
Recordings
BOLDc functional imaging was performed using a Siemens
Magnetom Vision scanner with 1.5 T- and T2*-weighted echoplanar imaging (EPI). Sequences parameters were: repetition
1586
Brain (2005), 128, 1584–1594
time (TR) 2 s, echo time (TE) 60 ms, matrix size 64 · 64, field of view
256 mm, in-plane voxel size 4 · 4 mm, flip angle 90 , slice thickness 5
mm, and no gap. Foam padding was added to the standard head coil
in order to minimize involuntary subject movement.
Functional volumes consisted of 16 transaxial slices parallel to the
anterior commissure–posterior commissure (AC–PC) line and
including the medial and posterior temporal, frontal and occipital
cortexes. A high-resolution structural volume was acquired at the
end of each session with a 3D MPRAGE (magnetization prepared
rapid acquisition gradient echo) sequence (matrix 256 · 256; field
of view 256 mm; slice thickness 1 mm; no gap; in-plane voxel size
1 · 1 mm; flip angle 12 ; TR 9.7 ms; TE 4 ms).
M. Caulo et al.
A
fMRI paradigm
The fMRI study was a block-design study. Each scanning session
consisted of five runs containing five blocks of 24 s. Participants
performed a face-encoding task during the first three runs, a face
discrimination task during the fourth run and a face recognition task
during the last run. For each of the three tasks the subjects were
required to choose between two options (dual-choice), as previously
described by Haxby and colleagues in a PET study (Haxby et al.,
1996). The dual-choice blocks (‘on’ condition) were alternated in an
‘off–on’ fashion, in which the ‘off ’ condition lasted 24 s and consisted of repeating, alternating pressing of the two buttons.
Stimuli for the dual-choice tasks consisted of one upper and two
lower white frames, in which either a greyscale neutral face or a nonsense pattern, according to the task type, was presented. The photos
with a neutral facial expression were obtained with the consent of
volunteers from our institute. During each block, six stimuli were
presented for 4 s each. The stimuli for the face-encoding task were
a nonsense pattern in the upper frame and an alternating face and a
nonsense pattern in the lower frames: the subject was required to
press the button that corresponded to the lower frame which contained the face (Fig. 1A). The stimuli for the face discrimination task
were a master face pattern in the upper frame and two faces in the
lower frames: the subject was required to press the button that
corresponded to the lower frame that contained the matching face
(Fig. 1B). The stimuli for the face recognition task were a nonsense
pattern in the upper frame and two faces in the lower frames: the
subject was required to press the button that corresponded to the
lower frame that contained the face that had already been presented
during the encoding tasks (Fig. 1C). Subjects were asked during the
tasks to always express a choice. For the control task all three frames
contained the nonsense pattern and no perceptual or memory
judgement were required.
Stimuli were projected by means of an LCD projector and two
perpendicular mirrors on a translucent glass plate placed on the back
of the scanner bore. An additional mirror was attached to the head
coil inside the magnet bore. This permitted the subjects to see the
stimuli. The LCD projector was driven by a personal computer
placed in the console room.
The nonsense pattern was obtained by scrambling one face using
a commercial package for picture manipulation (swirl function of
Adobe Photoshop CS).
Behavioural data (correct or incorrect responses) were obtained
by analysing which of the two buttons was activated (left or right
index fingers). These buttons were connected via fibre-optic cables
to the same computer that produced the stimuli. Behavioural data
were recorded using the Psychotoolbox of the Matlab program
(Mathworks, Natick, MA, USA).
B
C
Fig. 1 Examples of faces used as stimuli for the encoding
(A), perception (B) and recognition (C) tasks.
fMRI in Wernicke–Korsakoff syndrome
The subject and the controls received training the day before and
immediately prior to fMRI in order to ensure a correct execution of
the session. Task instructions were repeated orally between runs and
projected on the screen 4 s before the onset of each task block. This
repetitive procedure was necessary because of the increased probability that the subject would otherwise forget the instructions.
fMRI data analysis
Raw data were analysed using Brain Voyager 4.9 software (Brain
Innovation, The Netherlands). Preprocessing of functional scans
included motion correction and removal of linear trends from
voxel time series. A 3D motion correction was performed. Preprocessed functional volumes of each subject were co-registered with the
corresponding structural data set. Since the 2D functional and 3D
structural measurements were acquired in the same session, the
co-registration transformation was determined using the Siemens
slice position parameters of the functional images and the position
parameters of the structural volume. Structural and functional
images were transformed into Talairach space coordinates (Talairach
and Tournoux, 1988). The Talairach transformation was performed
in two steps. In the first step the structural and functional data sets
were rotated in the AC-PC plane. In the second step the extreme
points of the cerebrum were specified, defining 12 subvolumes which
were scaled into Talairach space using a piecewise affine and continuous transformation. Functional volumes were resampled at a
voxel size of 3 · 3 · 3 mm.
Statistical analysis was performed individually and for the control
group using the general linear model (Friston et al., 1995) in order
to reveal cortical regions showing a significantly higher BOLD signal
during the three tasks with respect to the off condition. The boxcar
waveform representing the task and control conditions was convolved with an empirically founded haemodynamic response
function in order to account for the haemodynamic delay (Boynton
et al., 1996).
Individual and group statistical maps were thresholded at
P < 0.0004 at the voxel level and a cluster size of at least four voxels
was required in order to obtain a corrected (for multiple comparisons) significance level: the probability of a false detection for the
entire functional volume was P < 0.05, as estimated by Monte Carlo
simulation (3dFWHM and AlphaSim routines of AFNI package;
Forman et al., 1995; Cox, 1996). Regions of interest (ROI) were
determined for each control and for the WK subject, considering
the mask obtained from the union of voxels activated in each condition separately.
Talairach coordinates of the defined ROI were obtained from the
centroids of clusters of activation. The mean time course of each ROI
was analysed and the relative signal variation between baseline (off
condition) and each task was calculated from the fitted parameters of
the general linear model in order to analyse the effect of the different
task conditions:
BOLD% change ¼ ðb · 100Þ=baseline,
where b represents the estimated amplitude of the variation in the
fMRI signal.
This analysis was restricted to the ROI of activated cortical regions
that have been indicated in the literature as having a major role in
face memory processing: the prefrontal cortex (ventrolateral and
dorsolateral), MTL structures and a visual cortical region (fusiform
gyrus) (Kelley et al., 1988; Clark et al., 1996; Haxby et al., 1996;
Kanwisher et al., 1997). The BOLD signal of the insular and medial
Brain (2005), 128, 1584–1594
1587
prefrontal cortex was also analysed. The regional comparison of
activation for the control group was performed using analysis of
variance (ANOVA) for repeated measures. The dependent variable
of the ANOVA analysis was the BOLD signal relative variation
between the task and off conditions; the factors were the memory
task (encoding, perception, recognition) and the selected ROI. The
Duncan test was used for post hoc comparisons. Activation in the
subject was considered significantly different when it was greater
than 2 SD with respect to controls.
In addition a voxel-wise analysis was performed by concatenating
controls’ and the WK subject’s voxel time courses to look for cortical
regions showing differences between conditions (encoding versus
perception, and recognition versus perception) in the WK patient
but not in controls and vice versa. In this analysis we used a less
stringent threshold (P < 0.001, uncorrected) in order to exclude false
negatives.
Results
Anatomical MRI
Signal changes involving the mammillary bodies, the medial
thalamic region and the fornix, in the absence of any signal
abnormalities in MTL structures, were observed 1 week after
clinical onset (Fig. 2A). The 5-month follow-up study demonstrated almost complete resolution of the lesions and absence
of signal changes in the MTL (Fig. 2B). An expert neuroradiologist made a visual qualitative evaluation of brain atrophy
using morphological MRI sequences. No signs of generalized
or focal MTL atrophy were detected in the 4-week and
5-month MRI studies.
fMRI findings
Controls
Analysis of fMRI data comparing the on and off conditions
revealed that activated areas were bilateral during each task
but with relative right lateralization. The control group map
of cumulative cortical activation for all tasks revealed
enhancement of several cortical regions: a region in the medial
surface of the frontal lobe (medial prefrontal cortex, MPFC)
in a position generally attributed to the supplementary motor
area (Fig. 3); two confluent bilateral regions, one anterior and
one posterior, in the dorsolateral prefrontal cortex (DLPFC)
and a cluster of activation in the low convexity involving
the pars opercularis of the right inferior frontal gyrus (IFG)
(ventrolateral prefrontal cortex, VLPFC) and the insular
cortex (Fig. 3). Given the close position of the insula and
inferior frontal gyrus and of the two regions of activation
of the DLPFC, the group analysis showed confluent activations, whereas analysis of the individual subjects distinguished
two distinct cortical areas. The insular cortex was also activated in the left hemisphere. Regions of cortical activation
were present in the bilateral ventral occipitotemporal cortex
and included the fusiform gyrus (Fig. 3). In this region it was
also necessary to look at individual activations in order to
distinguish the fusiform gyrus from the rest of the extrastriatal
1588
Brain (2005), 128, 1584–1594
M. Caulo et al.
A
B
C
D
Fig. 2 Images from axial and coronal FLAIR sequence of the WK subject. Seven-day MRI study demonstrates signal abnormalities of the
mammillary bodies (A) (stars), the medial aspect of the thalami (B, C) (arrows) and the fornix (B) (arrowheads). Five months later signal
changes have almost completely disappeared (D) (arrows and arrowheads).
cortex. In the MTL, the hippocampus was activated in two
distinct regions, one anterior and one posterior (Fig. 4).
WK subject
Because at the 5-month follow-up the severity of anterograde
amnesia the fMRI activations and the behavioural data were
similar to the study performed at 4 weeks, we merged the data
from the two fMRI examinations.
At the selected threshold, the activations showed marked
right hemisphere predominance. The areas of activation in
the DLPFC, MPFC, insula and ventral occipitotemporal
cortex (including the fusiform gyrus) resembled those
observed in the control group (Fig. 3). The list of activated
areas and corresponding Talairach coordinates is reported
in Table 2.
Activated areas in the MTL and VLPFC were not observed
at the statistical threshold of P < 0.05 (corrected). Even at a
less rigorous threshold (P < 0.001 uncorrected), activated
voxels were not detected in these regions. In order to further
reduce the possibility of false negatives, the fMRI signals for
these cortical areas were derived from ROI determined using
the respective ROI obtained from the group analysis of the
controls. The Talairach normalization can account only
partially for the intersubject anatomical variability. Therefore,
in addition to the use of Talairach coordinates we also visually
checked the overlapping of the group’s ROI with the patient’s
cortical anatomy on the basis of known anatomical landmarks. In order to avoid a methodological bias, the BOLD
response in these regions for each normal subject was also
evaluated considering the group ROI, in addition to the
evaluation performed using individual ROI.
fMRI in Wernicke–Korsakoff syndrome
Fig. 3 Activated areas during the ‘on’ condition in controls
(left column) and the WK subject (right column). Structural and
functional images are shown using the right–left radiological
convention.
ANOVA
The two-way ANOVA (factors: ROI, tasks) in the control
group was focused on eight ROI in the right hemisphere:
anterior and posterior DLPFC, IFG, MPFC, insular cortex,
Brain (2005), 128, 1584–1594
1589
Fig. 4 Activated areas in the hippocampal region in controls
during the ‘on’ condition. Structural and functional images are
shown using the right–left radiological convention. The images are
oriented so that the long axis of the hippocampus is parallel to the
axial section. In A the hippocampus is activated in two distint
regions: anterior (arrows) and posterior (arrowhead). B and C
show coronal slices passing trought the anterior (B) and posterior
hippocampal activations (C). The anterior aspect of the left
hippocampus is also activated during the ‘on’ condition (A).
anterior and posterior hippocampus, and fusiform gyrus. A
statistically significant main effect was observed for the ROI
factor [F(7.49) = 6.22; P < 0.00003] and for the task factor
[F(2.14) = 49.79; P < 0.0000004]. Furthermore, a statistically
1590
Brain (2005), 128, 1584–1594
M. Caulo et al.
Table 2 Cumulative activated areas during all three tasks in controls and the WK subject
Region
Control group
Right middle frontal gyrus (DLPFC)
Right middle frontal gyrus (DLPFC)
Left middle frontal gyrus (DLPFC)
Left middle frontal gyrus (DLPFC)
Right inferior frontal gyrus (VLPFC)
Left inferior frontal gyrus (VLPFC)
Right insula
Left insula
Right hippocampus
Right hippocampus
Left hippocampus
Right superior frontal gyrus (MPFC)
Left superior frontal gyrus (MPFC)
Right fusiform gyrus
Left fusiform gyrus
Right middle temporal gyrus
Left middle temporal gyrus
Right inferior temporal gyrus
Left inferior temporal gyrus
Left lingual gyrus
Right middle occipital gyrus
Left middle occipital gyrus
Right inferior occipital gyrus
WK subject
Talairach coordinates (x, y, z)
Brodmann area
Talairach coordinates (x, y, z)
Brodmann area
47,
47,
44,
45,
45,
48,
33,
39,
15,
24,
18,
5,
4,
29,
33,
32,
37,
40,
48,
3,
41,
45,
26,
9
42,
45,
44,
40,
–
–
28,
34,
–
–
–
5,
4,
30,
36,
32,
39,
39,
40,
2,
38,
46,
27,
9
24,
6,
25,
6,
18,
14,
17,
16,
10,
26,
11,
10,
11,
55,
60,
74,
78,
72,
72,
80,
75,
75,
88,
31
36
34
37
4
3
6
5
10
8
10
44
44
16
13
15
16
1
1
6
7
7
7
9
28
28
28
6, 32
6, 32
39
39
18
29,
5,
27,
7,
31
32
36
31
9
20, 11
15, 11
9,
11,
52,
53,
77,
79,
74,
77,
77,
75,
74,
88,
46
46
16
17
16
12
1
1
11
7
10
6
6, 32
6, 32
39
39
18
Table 3 ANOVA results: comparison of BOLD signal percentage changes during the memory tasks for the different ROI:
probabilities for post hoc test
Encoding vs perception
Encoding vs recognition
Recognition vs perception
Insula
Anterior
DLPFC
Posterior
DLPFC
MPFC
IFG
VLPFC
Anterior
hippocampus
Posterior
hippocampus
Fusiform
gyrus
0.094
0.011*
0.334
0.051
0.000*
0.051
0.016*
0.000*
0.229
0.051
0.000*
0.000*
0.156
0.000*
0.000*
0.835
0.001*
0.001*
0.034*
0.072
0.028*
0.012*
0.000*
0.258
*Significant differences.
significant interaction [F(14.98) = 3.49; P < 0.0001] between
the two factors was observed.
In the prefrontal cortex, the BOLD signal increased during
the three memory tasks (encoding, perception, and recognition) (Fig. 5A). The anterior and posterior DLPFC, the MPFC
and the insular cortex were significantly more activated during
recognition than during encoding. No difference was demonstrated, for the same cortical regions, between recognition and
perception except for the MPFC, which was more activated
during recognition (Fig. 5A and B). Moreover, the anterior
and posterior DLPFC, MPFC and insular cortex showed
no significant difference between encoding and perception,
except for the posterior DLPFC, which was more activated
during perception. The VLPFC activation was significantly
greater during recognition than during both encoding and
perception (Fig. 5B). In the MTL, the ROI in the anterior
hippocampus was significantly more activated during
recognition than during encoding and perception but a difference was not observed between encoding and perception.
The region of activation in the posterior hippocampus showed
a significantly greater BOLD signal during encoding and
recognition than during perception, whereas differences did
not exist between encoding and recognition (Fig. 5C). The
BOLD signal of the fusiform gyrus was significantly greater
during recognition and perception than during encoding.
Statistically significant differences were not observed between
recognition and perception (Fig. 5D). Levels of significance of
the post hoc comparison of the BOLD signal change during the
three different tasks in each ROI are reported in Table 3.
Comparison of cortical activation of the control group
versus the WK subject demonstrated a precise match in
the DLPFC, MPFC, insula and fusiform gyrus. However,
the patient failed to activate the MTL and the VLPFC during
each condition.
The voxel-wise analysis, testing for differences between
conditions in the WK subject compared with controls, did
not show further regions of activation other than those which
entered the ANOVA analysis.
fMRI in Wernicke–Korsakoff syndrome
Brain (2005), 128, 1584–1594
1591
A
B
C
D
Fig. 5 BOLD % signal changes in different cortical regions during task execution in controls and the WK subject. Structural and functional
images are shown using the right–left radiological convention. The y axis shows the BOLD % signal change. The error bars represent
standard deviations. The comparison between the activations of the control group and the WK subject demonstrates a similar BOLD signal
course exept for the hippocampal region and the IFG, where no activation was observed in the WK subject.
1592
Brain (2005), 128, 1584–1594
Behavioural data
Controls
During encoding, the correct responses in the face–nonsense
judgement were virtually 100% in all subjects, indicating
good attention and accurate judgement. Mean correct
responses were 88.6 ± 3.8% (range 83–94%) for perception
and 83.5 ± 4.2% (range 78–89%) for recognition.
WK subject
Cumulative correct responses of the two fMRI studies were
98% for encoding and 88% for perception. These results were
not significantly different from those for controls. Correct
responses for recognition were 54%, which was significantly
lower than the figure for controls.
Discussion
Patients with WK syndrome present persistent anterograde
episodic memory loss with preserved semantic memory, intelligence and learned behaviour (Kopelman, 1995). The severe
anterograde memory loss is similar to the amnesia that is
associated with MTL lesions in humans and experimental
animals (Squire, 1992; Aggleton and Brown, 1999). There
are a number of recent fMRI and PET studies indicating
that MTL is activated during episodic memory encoding
and retrieval. A meta-analysis of 52 PET studies concluded
that the anterior MTL is strongly associated with episodic
encoding, whereas the posterior MTL is associated with retrieval (Lepage et al., 1998). In contrast, Schacter and Wagner
(1999), in their meta-analysis of fMRI studies, indicated
a predominance of posterior MTL activation for encoding.
Data on retrieval were insufficient to draw any conclusion on
rostrocaudal localization. Overall, although there is no general
agreement on regional anatomical differences in activation
between encoding and retrieval, several authors have reported
that MTL activation is observed during both episodic encoding and retrieval (Shacter et al., 1999; Eldridge et al., 2000;
Greicius et al., 2003).
We chose recognition memory for faces because faces are
novel, unfamiliar and unique in their configuration. Therefore, they are more suitable than words for testing episodic
memory (Haxby et al., 1996). Few fMRI studies have investigated facial encoding and recognition. Kelley and colleagues
reported a bilateral MTL (greater on the right) and right dorsal
frontal cortex activation during encoding (Kelley et al., 1998).
Bernard and colleagues tested memory for famous faces using
an event-related fMRI study and demonstrated a functional
distinction between the anterior and posterior aspects of
the hippocampus, the former being involved in successful
episodic encoding and the latter with retrieval of semantic
information (Bernard et al., 2004).
Haxby and colleagues, using PET, found that facial encoding was associated with increased regional cerebral blood flow
(rCBF) in the right MTL and left prefrontal cortex, whereas
facial recognition was associated with increased rCBF in an
M. Caulo et al.
extensive region of the right prefrontal cortex (Haxby et al.,
1996). We employed a modified version of the paradigm
developed by Haxby and colleagues. Since we were dealing
with a patient with memory impairment, we thought that a
shorter study would increase the probability that the subject
would complete it with adequate collaboration. Therefore we
choose a block-design fMRI study, which generally yields
more robust activation compared with an event related
study of the same duration (Friston et al., 1999).
In our fMRI study we found that in controls a right posterior hippocampal region is activated by both facial encoding
and recognition, whereas a more anterior hippocampal region
is greatly activated in recognition. The prefrontal cortex
(greater on the right) was activated during all tasks, especially
during recognition. The discrepancies in the reported studies
may be due to the differences in techniques and paradigms
employed.
The lateralization within MTL and the prefrontal cortex has
been demonstrated to be dependent on the type of material
being remembered (Kelley et al., 1998). The left frontal and
medial temporal network is predominantly activated by
verbal stimuli whereas a contralateral network is activated
by non-verbal stimuli, such as faces. Therefore, we will only
discuss activations of the right hemisphere.
The cortex of the fusiform gyrus has been demonstrated to
be selectively specialized in facial perception (Haxby et al.,
1994; Kanwisher et al., 1997); therefore, we focused our analysis on this region. In our controls and the WK subject, the
bilateral activations of the cortex of the fusiform gyrus were
comparable. In both controls and the WK subject the greater
activation of the fusiform gyrus during the perception and
recognition compared with the encoding task could be due
either to the number of faces employed in each task (one in
encoding, two in recognition, three in perception) or to the
different attention levels (Wojciulik et al., 1998).
In the WK subject fMRI showed no activation of MTL
during encoding, although the behavioural scores indicated
that faces were seen, attention was good and judgement was
accurate. Moreover, the normal activation of the cortex of the
fusiform gyrus excludes haemodynamic impairment of the
cortical–visual network as a possible cause of the lack of
MTL activation. During recognition, the activations observed
in the prefrontal regions were not significantly different from
those in controls, except for the inferior frontal cortex. However behavioural data indicated impaired recognition. These
findings suggest that MTL activation during encoding is crucial for subsequent recognition memory for faces and that
activation of the dorsolateral prefrontal cortex may be more
involved in orientation and/or attention at retrieval than in
retrieval per se (Dobbins et al., 2003). The inferior frontal
cortex is thought to be involved in the elaborative encoding
of information into episodic memory, as well as in the maintenance of retrieved information (Fletcher and Henson, 2001;
Dobbins et al., 2002). Therefore, its lack of activation seems to
be more specifically connected with the defective encoding
and unsuccessful retrieval in WK. Although the poor memory
fMRI in Wernicke–Korsakoff syndrome
performance of the patient and the associated activation
deficits in the MTL and the IFC are consistent with recent
literature (Bunge et al., 2004), we cannot exclude some additional problems of the patient in retrieving and/or perceiving
faces, not detected by our paradigm, that may influence the
differences in BOLD signal between patient and controls.
No activation differences were found between subject and
controls in the MPFC and the insula. As generally acknowledged, the MPFC is involved in motor behaviour (Picard and
Strick, 2001), whereas activation of the insula is found in most
tasks requiring visual attention (Woldorff et al., 2004).
The lack of differences in these regions, which are involved
in the demands of general processing, supports the specificity
of the activation deficits found in regions related to memory
formation in the WK subject.
Neuropathological studies in alcoholic Korsakoff syndrome have identified the dorsal medial thalamic nuclei as
the anatomical site of pathology (Victor and Adams, 1989).
Harding and colleagues, using operational criteria to identify
alcoholics with and without Korsakoff syndrome (Caine et al.,
1997), demonstrated that neuronal loss in the anterior thalamic nuclei is the best predictor of memory loss (Harding
et al., 2000). This supports the view that the hippocampal–
anterior thalamic axis is critical for episodic memory formation (Aggleton and Brown, 1999). Damage to this axis
is responsible for the anterograde amnesia in Korsakoff syndrome, as originally proposed by Delay and Brion (1969).
Pathology and MRI studies support the view that hippocampal structures are not damaged in WK syndrome (Squire et al.,
1990; Harding et al., 1997; Colchester et al., 2001).
The patient we report is not an alcoholic. MRI showed
lesions in the diencephalic structures matching the classical
descriptions of Wernicke encephalopathy (Gallucci et al.,
1990). MRI did not show signal abnormality or cortical atrophy in the MTL structures in either the 4-week or 5-month
studies. There were also no signs of generalized cortical atrophy. Nonetheless, fMRI did not demonstrate hippocampal
activation during encoding, the initial step of episodic memory formation. A PET study in two WK patients documented,
besides the expected hypometabolism in the diencephalic grey
matter, hypometabolism of MTL structures (Reed et al.,
2003). This finding was interpreted as a secondary metabolic
effect of the diencephalic pathology. Several experimental
studies have demonstrated that anterior thalamic damage
disrupts normal hippocampal activity (Jenkins et al., 2002;
Savage et al., 2003).
In conclusion, in the patient we report here, MRI confirms
that Korsakoff syndrome is related to diencephalic lesions and
fMRI indicates that anterograde memory loss is due to a
failure to encode the aspects of the incoming information
at the hippocampal level. These findings support the view
that the distinction between diencephalic and MTL amnesias
is of limited value. The common denominator of anterograde
amnesia is damage to an extended hippocampal system which
seems to work as a unitary network (Aggleton and Brown,
1999).
Brain (2005), 128, 1584–1594
1593
Acknowledgements
This work was supported in part by a grant from the Italian
Ministry of Research to the Center of Excellence on Aging of
the University of Chieti.
References
Aggleton JP, Brown MW. Episodic memory, amnesia, and the hippocampalanterior thalamic axis. Behav Brain Sci 1999; 22: 42589.
Bernard FA, Bullmore ET, Graham KS, Thompson SA, Hodges JR, Fletcher
PC. The hippocampal region is involved in successful recognition of both
remote and recent famous faces. Neuroimage 2004; 22: 170414.
Boynton GM, Engel SA, Glover GH, Heeger DJ. Linear systems analysis
of functional magnetic resonance imaging in human V1. J Neurosci 1996;
16: 420741.
Bunge SA, Burrows B, Wagner AD. Prefrontal and hippocampal contributions to visual associative recognition: interactions between cognitive
control and episodic retrieval. Brain Cogn 2004; 56: 14152.
Caine D, Halliday GM, Kril JJ, Harper CG. Operational criteria for the
classification of chronic alcoholics: identification of Wernicke’s encephalopathy. J Neurol Neurosurg Psychiatry 1997; 62: 5160.
Clark VP, Keii K, Maisog M, Courtney S, Ungerleider LG, Haxby JV. Functional magnetic resonance imaging of human visual cortex during face
matching: a comparison with positron emission tomography. Neuroimage
1996; 4: 115.
Colchester A, Kingsley D, Larserson D, Kendall B, Bello F, Rush C, et al.
Structural MRI volumetric analysis in patients with organic amnesia:
Methods of comparative findings across diagnostic groups. J Neurol
Neurosurg Psychiatry 2001; 71: 1322.
Cox RW. AFNI: software for analysis and visualization of functional magnetic
resonance neuroimages. Comput Biomed Res 1996; 29: 16273.
Delay J, Brion S. Le syndrome de Korsakoff. Paris: Masson; 1969.
Dobbins IG, Foley H, Schacter DL, Wagner AD. Executive control during
episodic retrieval: multiple prefrontal processes subserve source memory.
Neuron 2002; 35: 98996.
Dobbins IG, Rice HJ, Wagner AD, Schacter DL. Memory orientation and
success: separable neurocognitive components underlying episodic recognition. Neuropsychologia 2003; 41: 31833.
Eldridge LL, Knowlton BJ, Furmanski CS, Bookeimer SY, Engel SA. Remembering episodes: a selective role for the hippocampus during retrieval. Nat
Neurosci 2000; 11: 114952.
Fletcher PC, Henson RN. Frontal lobe and human memory insights from
functional neuroimaging. Brain 2001; 124: 84981.
Forman SD, Cohen JD, Fitzgerald M, Eddy WF, Mintun MA, Noll DC.
Improved assessment of significant activation in functional magnetic resonance imaging (fMRI): use of a cluster-size threshold. Magn Reson Med
1995; 33: 63647.
Friston KJ, Holmes AP, Worsley KJ, Poline JP, Frith CD, Frackowiak RSJ.
Statistical parametric maps in functional imaging: a general linear
approach. Hum Brain Mapping 1995; 2: 17381.
Friston KJ, Zarahn E, Josephs O, Henson RN, Dale AN. Stochastic design in
event-related fMRI. Neuroimage 1999; 10: 60719.
Gallucci M, Bozzao A, Splendiani A, Masciocchi C, Passariello R. Wernicke
encephalopathy: MR findings in five patients. Am J Neuroradiol 1990; 11:
88792.
Greicius MD, Krasnow B, Boyett-Anderson JM, Eliez S, Schatzberg AF,
Reiss AL, Menon V. Regional analysis of hippocampal activation during
memory encoding and retrieval: fMRI study. Hippocampus 2003; 13:
16474.
Harding AJ, Wong A, Svoboda M, Kril JJ, Halliday GM. Chronic alcohol
consumption does not cause hippocampal neuron loss in humans.
Hippocampus 1997; 7: 7887.
Harding AJ, Halliday G, Caine D, Kril JJ. Degeneration of anterior thalamic
nuclei differentiates alcoholics with amnesia. Brain 2000; 123: 14154.
Haxby JV, Horwitz B, Ungerleider LG, Maisog JM, Pietrini P, Grady CL.
The functional organization of human extrastriate cortex: a PET-rCBF
1594
Brain (2005), 128, 1584–1594
study of selective attention to faces and location. J Neurosci 1994; 14:
633653.
Haxby JV, Ungerleider LG, Horwitz B, Maisog JM, Rapoport SI, Grady CL.
Face encoding and recognition in the human brain. Proc Natl Acad Sci USA
1996; 93: 9227.
Jenkins TA, Dias R, Amin E, Brown MW, Aggleton JP. Fos imaging reveals
that lesions in the anterior thalamic nuclei produce widespread limbic
hypoactivity in rats. J Neurosci 2002; 22: 52308.
Kanwisher N, McDermott J, Chun MM. The fusiform face area: a module in
human extrastriate cortex specialized for face perception. J Neurosci 1997;
17: 430211.
Kelley WM, Miezin F, McDermott K, Buckner RL, Raichle ME, Cohen NJ,
et al. Hemispheric specialization in human dorsal frontal cortex and medial
temporal lobe for verbal and non verbal memory encoding. Neuron 1988;
20: 92736.
Kopelman MD. The Korsakoff syndrome. Br J Psychiatry 1995; 166: 15473.
Lepage M, Habib R, Tulving E. Hippocampal PET activations of
memory encoding and retrieval: the HIPER model. Hippocampus 1998;
8: 31322.
Mair RG. On the role of thalamic pathology in diencephalic amnesia. Rev
Neurosci 1994; 5: 10540.
Picard N, Strick PL. Imaging the premotor areas. Curr Opin Neurobiol 2001;
11: 66372.
Reed LJ, Lasserson D, Marsden P, Stanhope N, Stevens T, Bello F, et al.
FDG-PET findings in the Wernicke–Korsakoff syndrome. Cortex 2003;
39: 102745.
M. Caulo et al.
Savage LM, Chang O, Gold PE. Diencephalic damage decreases hippocampal
acetylcholine release during spontaneous alternation testing. Learn Mem
2003; 10: 2426.
Schacter DL, Wagner AD. Medial temporal lobe activations in fMRI and PET
studies of episodic encoding and retrieval. Hippocampus 1999; 9: 724.
Schacter DL, Curran T, Reiman EM, Chen K, Bandy DJ, Frost JT. Medial
temporal lobe activation during episodic encoding and retrieval: a PET
study. Hippocampus 1999; 9: 57581.
Squire LR. Memory and the hippocampus: a synthesis from findings with rats,
monkeys, and humans. Psychol Rev 1992; 99: 195231.
Squire LR, Amaral DG, Press GA. Magnetic resonance imaging of the hippocampal formation and mammillary nuclei distinguish medial temporal
lobe and diencephalic amnesia. J Neurosci 1990; 10: 310617.
Talairach J, Tournoux P. Co-planar stereotaxis atlas of the human brain.
New York: Thieme; 1988.
Vann SD, Aggleton JP. The mammillary bodies: two memory systems in one?
Nat Rev Neurosci 2004; 5: 3544.
Victor M, Adams RD, Collins GH. The Wernicke–Korsakoff syndrome and
related neurologic disorders due to alcoholism and malnutrition. 2nd edn.
Philadelphia: FA Davies; 1989.
Wojciulik E, Kanwisher N, Driver J. Covert visual attention modulates facespecific activity in the human fusiform gyrus: fMRI study. J Neurophysiol
1998; 79: 15748.
Woldorff MG, Hazlett CJ, Fichtenholtz HM, Weissman DH, Dale AM,
Song AW. Functional parcellation of attentional control regions of the
brain. J Cogn Neurosci 2004; 16: 14965.