Abnormal activity in reward brain circuits in human narcolepsy with cataplexy

Abstract

Objective
Hypothalamic hypocretins (or orexins) regulate energy metabolism and arousal maintenance. Recent animal research suggests that hypocretins may also influence reward-related behaviors. In humans, the loss of hypocretin-containing neurons results in a major sleep-wake disorder called narcolepsy-cataplexy, which is associated with emotional disturbances. Here, we aim to test whether narcoleptic patients show an abnormal pattern of brain activity during reward processing.Methods
We used functional magnetic resonance imaging in 12 unmedicated patients with narcolepsy-cataplexy to measure the neural responses to expectancy and experience of monetary gains and losses. We statistically compared the patients' data with those obtained in a group of 12 healthy matched controls.Results and InterpretationOur results reveal that activity in the dopaminergic ventral midbrain (ventral tegmental area) was not modulated in narcolepsy-cataplexy patients during high reward expectancy (unlike controls), and that ventral striatum activity was reduced during winning. By contrast, the patients showed abnormal activity increases in the amygdala and in dorsal striatum for positive outcomes. In addition, we found that activity in the nucleus accumbens and the ventral-medial prefrontal cortex correlated with disease duration, suggesting that an alternate neural circuit could be privileged over the years to control affective responses to emotional challenges and compensate for the lack of influence from ventral midbrain regions. Our study offers a detailed picture of the distributed brain network involved during distinct stages of reward processing and shows for the first time, to our knowledge, how this network is affected in hypocretin-deficient narcoleptic patients. ANN NEUROL 2010;67:190–200

Full text

Page 1

1
Abnormal Activity in Reward Brain Circuits in Human Narcolepsy with Cataplexy
Ponz, A. (1,2); Khatami, R. (3); Poryazova , R. (3); Werth , E. (3); Boesiger, P. (4);
Bassetti, C. (3); Schwartz, S. (1,2)

1. Department of Neuroscience, University of Geneva, Switzerland
2. Geneva Neuroscience Center, University of Geneva, Switzerland
3. Neurology Department, University Hospital Zurich, Switzerland
4. Biomedical Engineering, University of Zurich & Swiss Federal Institute of
Technology, Switzerland

Corresponding author:
Sophie Schwartz
Department of Neuroscience, University Medical Center
Michel-Servet 1, 1211 Geneva 4, Switzerland
Phone: +41 22 3795376 Fax: +41 22 379 5402

Running Head: Brain Responses to Reward in Human Narcolepsy


Number of characters in the title: 77
Number of characters in the running head: 45
Number of words in the abstract: 245
Number of words in the body of the manuscript: 3185
Number of figures: 5
Number of color figures: 5
Number of tables: 3
Page 1 of 47
John Wiley & Sons
Annals of Neurology

Page 2

2
ABSTRACT
Objective: Hypothalamic hypocretins (or orexins) regulate energy metabolism and
arousal maintenance. Recent animal research suggests that hypocretins may also
influence reward-related behaviors. In humans, the loss of hypocretin-containing neurons
results in a major sleep-wake disorder called narcolepsy-cataplexy, which is associated
with emotional disturbances. Here, we aim to test whether narcoleptic patients show an
abnormal pattern of brain activity during reward processing.
Methods: We used functional magnetic resonance imaging in 14 unmedicated patients
with narcolepsy-cataplexy to measure the neural responses to expectancy and experience
of monetary gains and losses. We statistically compared the patients’ data to those
obtained in a group of 14 healthy matched controls.
Results and Interpretation: Our results reveal that activity in the dopaminergic ventral
midbrain (ventral tegmental area) was not modulated in narcolepsy-cataplexy patients
during high reward expectancy (unlike controls), and that ventral striatum activity was
reduced during winning. By contrast, the patients showed abnormal activity increases in
the amygdala and in dorsal striatum for positive outcomes. In addition, we found that
activity in the nucleus accumbens and the ventral-medial prefrontal cortex correlated with
disease duration, suggesting that an alternate neural circuit could be privileged over the
years to control affective responses to emotional challenges and compensate for the lack
of influence from ventral midbrain regions. Our study offers a detailed picture of the
distributed brain network involved during distinct stages of reward processing and shows
for the first time, to our knowledge, how this network is affected in hypocretin-deficient
narcoleptic patients.
Page 2 of 47
John Wiley & Sons
Annals of Neurology

Page 3

3

INTRODUCTION
Hypocretins (HCRT1 or orexins2) are produced by specific neurons in the hypothalamus
that have extensive projections throughout the central nervous system.3, 4 Deficient
HCRT transmission was found to be associated with narcolepsy-cataplexy (NC) in
humans, dogs and knock-out mice,5, 6 suggesting a main role for HCRT in sleep/wake
regulation and arousal-maintenance.3, 7 Clinically, human NC is characterized by
excessive daytime sleepiness, fragmented nighttime sleep, sleep onset REM (rapid eye
movement) periods, as well as sudden episodes of postural muscle atonia called
cataplexy.8 Cataplexy attacks are predominantly triggered by emotional experiences,
including the anticipation of reward when playing games.8-10 These behavioral
observations suggest possible interactions between the hypothalamic HCRT system and
reward brain circuits in humans.11, 12

Recent studies in rodents provided evidence for anatomical and functional links between
the HCRT system and the dopamine system, the latter being critically involved in reward
processes and motivated behaviors.13-15 First, hypothalamic HCRT neurons project
densely to reward-associated brain regions, including the nucleus accumbens (NAcc) and
dopaminergic ventral tegmental area (VTA).16 Second, HCRT receptors are expressed at
the surface of VTA dopamine neurons,17, 18 and HCRT administration increases the firing
rate of VTA neurons.19 Third, HCRT is involved in drug seeking behaviors and
associated mesolimbic dopaminergic activity.18, 20-23 These recent findings in rodents
Page 3 of 47
John Wiley & Sons
Annals of Neurology

Page 4

4
suggest that the HCRT system regulates dopamine activity in brain reward circuits and
impacts on the expression of motivated behaviors and addiction.

While human NC is linked to hypothalamic HCRT depletion,5, 6 it remains unknown
whether NC-patients show altered brain reward processing. Scarce behavioral evidence
exists for reduced drug addiction in NC-patients, as these patients are often treated with
addictive amphetamine-like drugs but rarely become addicted to these drugs.8, 24-26 NC-
patients may also present with some psychiatric disturbances.27 In the present study, we
tested whether human NC entails a dysfunction in brain reward circuits. We acquired
whole-brain functional magnetic resonance imaging (fMRI) data in 14 narcoleptic
unmedicated, non-depressed patients with clear-cut cataplexy and 14 healthy matched
controls (Supplementary Table 1). During scanning, the subjects performed an adapted
version of a monetary incentive delay task, which is known to powerfully activate the
mesolimbic and midbrain reward system (Methods, Fig 1).28-30 As suggested by rodent
studies (see above), we hypothesized that HCRT deficiency in NC-patients might lead to
altered reward-related responses in ventral midbrain/VTA regions during reward
expectancy. We also expected that neural responses to reward experience in the
mesolimbic reward system would be affected in NC-patients. Finally, based on our recent
study on humor processing in NC-patients,11 we anticipated abnormal increases in
amygdala activity during positive emotions elicited by our game-like task, i.e., during
winning trials that resulted in high gains. Because they mostly confirm the hypotheses
above, our fMRI results provide the first evidence, to our knowledge, for abnormal brain
Page 4 of 47
John Wiley & Sons
Annals of Neurology

Page 5

5
responses to reward in NC-patients, and thus suggest an implication of HCRT activity in
the regulation of brain reward function in humans.

METHODS
Patients and Controls
Fourteen drug-free narcoleptic patients with clear-cut cataplexy and 14 healthy controls
matched for age, sex, handedness, and body mass index participated to this study
(Supplementary Table 1 and Supplementary Methods). Twelve patients (and their
matched controls) were included in the final statistical analyses. HLA-DQB1*0602 was
positive in all twelve patients. Hypocretin-1 levels (<120 pg/ml) in the cerebrospinal fluid
could be determined in 8 NC patients and was confirmed to be low or undetectable in all
of them.6, 31

Experimental Paradigm
During fMRI scanning, the subjects performed a visuo-motor game-like task in which
they could win (or loose) points if they rapidly pressed on a key while a visual target was
briefly shown (Fig 1A; Supplementary Methods). Each trial started with a preparation
period, during which the subjects saw a cue indicating the potential gain (+1 or +5 points)
or potential loss (-1 or -5 points) associated with that particular trial. After a variable
delay, the visual target that required a rapid key press was briefly presented on the screen.
The trial ended with a feedback display, telling the subjects whether they just won or lost
the trial. This paradigm allowed us to distinguish between a preparation period
Page 5 of 47
John Wiley & Sons
Annals of Neurology

Page 6

6
(presentation of the cue) and an outcome period (presentation of the feedback). In
addition, we could compare trials with high and low incentives (large versus small cues),
as well as successful versus failed trials (Fig 1B). To ensure a fixed proportion of
successful trials (50-60%), a tracking algorithm adjusted the duration of each target
presentation (i.e. task difficulty) based on each subject’s current performance.

Magnetic Resonance Imaging Methods
We acquired functional MRI data on a Philips Intera 3.0 T whole-body system (Philips
Medical Systems, Best, NL) across 4 scanning sessions (229 MRI volumes each)
separated by brief pauses (Supplementary Methods). The functional MRI data were
analyzed using the standard general linear approach in SPM (www.fil.ion.ucl.ac.uk/spm).
The statistical model included 12 main regressors of interest: 4 regressors corresponding
to the presentation of the 4 possible cues during the preparation period, and 8 regressors
corresponding to the presentation of the feedback during the outcome period (winning or
loosing on the 4 different cue-types; Fig 1B). Individual contrast images between
conditions of interest were calculated and entered into random-effects group analyses
using ANOVAs as implemented in SPM.32 Common group effects were assessed using
conjunction analyses to reveal voxels that showed significant activity increase in both the
patient and control populations.33 Direct group comparisons were performed using an
exclusive masking procedure to reveal voxels that showed increased response during an
experimental condition in one population but not in the other population. Using dedicated
connectivity methods, we also assessed changes in the functional coupling between a
main region of interest (ventral midbrain/VTA region) and any other brain region as a
Page 6 of 47
John Wiley & Sons
Annals of Neurology

Page 7

7
function of cue value.34 Finally, we tested for correlations of brain activity with clinical
characteristics in NC-patients (disease duration, cataplexy, and sleep propensity scores)
using regression analyses.

RESULTS
Behavioral Results
Reaction times
We analyzed the reaction times (RT) on successful trials using an analysis of variance
(ANOVA), with Cue Value (1 point, 5 points) and Cue Valence (positive, negative) as
within-subject factors, and Group (NC-patients, controls) as a between-subject factor
(Supplementary Methods). This statistical analysis revealed that RTs during the task did
not differ between the groups (NC-patients: 226.84 ms, SD 31.43; controls: 228.38, SD
23.36, F(1,22) = 0.02, n.s.). RTs were also not influenced by the value of the cue
(F(1,22) = 0.30, n.s.), nor by the valence of the cue (F(1,22) = 0.47, n.s.). There was no
interaction between these factors. RTs also influenced the duration of target presentation
because the online tracking algorithm controlled the success rate on this reaction-time
task by adjusting the target duration during the task. As expected, we found that average
presentation times for the target did not differ between NC-patients and controls (F(1,22)
= 0.92, n.s.).

Successful trials (hits)
Page 7 of 47
John Wiley & Sons
Annals of Neurology

Page 8

8
We analyzed the number of correct responses (hits) using an ANOVA with the same
factors as for the RTs above. There was no main effect of Group on hit rates (patients:
57.29%, SD 11.8; controls: 58.33%, SD 9.96, F(1,22) = 0.77, n.s.), with both groups
achieving better scores for highest incentive trials (1 point: 54.79%, SD 9.72; 5 points:
60.83%, SD 11.21, F(1,22) = 6.16, p < 0.05), as well as for potential gains (gains:
55.31%, SD 10.35; losses: 60.31%, SD 10.88, F(1,22) = 7.87, p < 0.01). There was a
significant interaction of Valence by Group (F(1,22) = 6.79, p < 0.01), due to NC-patients
being less successful on positive than negative trials (positive: 52.5%, SD 9.66; negative:
62.08%, SD 11.97) unlike controls (positive: 58.13%, SD 10.5; negative: 58.54%, SD
9.6).

Failed trials
Failed trials in this task occurred whenever the subject pressed the response key shortly
after the target picture had disappeared from the screen (less than 1300 ms after target
presentation). Controls produced more such errors than NC-patients (main effect of
Group, patients: 40.53%, SD 11.29; controls: 46.35%, SD 8.57, F(1,22) = 11.89, p <
0.01). There was no effect of Cue Value but a main effect of Cue Valence because failed
trials were slightly more frequent with positive than negative cues (positive: 45.91%, SD
9.66; negative: 40.97%, SD 10.60, F(1,22) = 14.64, p < 0.001). However, this valence
effect was not due to a general slowing of the responses after positive cues, as there was
no effect of Valence on RTs. There was no interaction between these factors.

Page 8 of 47
John Wiley & Sons
Annals of Neurology

Page 9

9
Taken together, these behavioral results show that NC-patients could achieve similar
reaction times as controls during our game-like task, which is remarkable given that the
patients had either no chronic treatment or had to withdraw from medication. Moreover,
both populations obtained similar final outcomes, ranging from 21.5 to 24.8 points
(Mann-Whitney, U = 61, p = 0.52). Mostly comparable performance for both groups
ensures that observed differences in brain activity between the groups are not due to
major differences in performance. Importantly, the online tracking algorithm efficiently
controlled the proportion of successful and failed trials, as well as their balanced
distribution across conditions in each group, which guarantees adequate statistical power
for valid inference from the fMRI results.

FMRI Results
Standard SPM Analyses
Preparation period: Brain response to high motivational cues
To identify brain regions more activated during reward expectancy, we compared brain
activity during the presentation of large (+5/-5 points) versus small (+1/-1 point) cue
values (Fig 1B; Supplementary Fig 1). Both NC-patients and controls activated a network
of brain regions involved in the expectation of a reward, including the ventral striatum
(Table 1).29, 30, 35, 36 Both populations also activated the amygdala, the anterior insula, and
the anterior cingulate cortex, consistent with motivational processes enhancing attention
and autonomic reactivity.37, 38 As expected from augmented visual attention and motor
preparation for highest incentives in this visuo-motor task, both groups showed increased
activation in visual and motor regions. When directly comparing the groups, we found
Page 9 of 47
John Wiley & Sons
Annals of Neurology

Page 10

10
increased fMRI signal in controls in a region of the ventral midbrain, compatible with the
VTA (-3x, -24y, -21z; Fig 2; Table 1).13, 14, 39 Activity increases were also observed in the
medial prefrontal cortex and caudate nucleus, which are target regions for dopaminergic
VTA projections and which are involved in reward prediction.40 No such modulation of
activity by Cue Value was observed in these brain regions for NC-patients. When
comparing NC-patients to controls for the same fMRI contrast, we found increased
activation of the insula and inferior orbitofrontal cortex for higher cue values.

Outcome period: Brain response to successful trials
To test for brain activity increases during winning on both positively- and negatively-
cued trials, we compared any successful trials (i.e., gains and no-losses) to failed trials
(i.e., no-gains and losses) at the onset of the feedback display (Fig 1). Both NC-patients
and controls showed increased activity in the dorsal striatum (caudate nucleus), as well as
in a limbic region corresponding to the sublenticular extended amygdala (SLEA; Table
2). Group comparisons revealed increased activity in ventromedial prefrontal cortex
(vmPFC) and NAcc during successful trials in the controls only (Fig 3A). Both these
regions are known to be involved in regulating emotional processes and reward,14, 41-43
and also receive dense dopaminergic projections from the VTA 44. As suggested by
recent animal work,18, 20-22 HCRT depletion may affect dopaminergic modulation within
brain reward networks, which would also be consistent with the observation that the
midbrain/VTA was not responsive to highly motivating cues in NC-patients (see above).
By contrast, NC-patients showed activity increases in the right dorsal striatum (putamen)
and inferior lateral frontal cortex (Fig 3B).
Page 10 of 47
John Wiley & Sons
Annals of Neurology

Page 11

11

Outcome period: Brain response to successful positively-cued trials
NC patients report cataplexy when they experience strong, most usually positive
emotions such as when joking or winning games.9, 10 Our game-like task also elicited
positive emotions, particularly when winning on rewarded trials (i.e. for actual gains on
positively-cued trials). Note that this condition did not trigger cataplexy in the patients
during scanning. To assess changes in regional brain activity during this positive
emotional condition, we compared winning on positively-cued trials to failing on such
trials (Fig 1). In both NC-patients and controls, winning on positively-cued trials
activated the anterior cingulate cortex (ACC) and the vmPFC, encompassing regions
involved in motivation regulation (see Table 3 for a detailed list of activation clusters).41,
45 When directly compared to NC-patients, controls showed increased activity in the
NAcc during actual gains and in the right lateral PFC. Conversely, NC-patients showed
further increases in amygdala/SLEA activity (Fig 4), which is in line with our recent
fMRI study using humorous stimuli.11 They also showed increased activity in the dorsal
striatum (putamen; consistent with the result of the previous contrast reported above).

Functional Connectivity Analysis
To further refine the functional brain network involved in the processing of high
motivational cues, we conducted a functional connectivity analysis (Methods), which
revealed increased functional coupling between midbrain/VTA and left NAcc (medial
part, -3x 12y -12z) selectively during the presentation of large cues in controls. This
Page 11 of 47
John Wiley & Sons
Annals of Neurology

Page 12

12
result is in line with studies indicating that neuronal activity in VTA and nucleus
accumbens may increase in proportion to the magnitude of anticipated reward.13, 39, 40

Regression Analyses
A clinically-relevant question was whether any regional change in fMRI signal could
relate to the patients’ individual clinical characteristics. We found that response to large
cues in the left NAcc (-9x 12y -6z; R2 = 0.82, p < 0.001) and the vmPFC (-3x 42y -12z;
R2 = 0.85, p < 0.001) correlated positively with disease duration (Fig 5). Critically, while
activity in these regions did not reach significance in NC-patients for the contrast of
interest (large versus small cue values), it was significant in controls. This suggests a
recovery of activity in these reward-related regions for patients who had long been
suffering from narcolepsy.

DISCUSSION
Recent animal studies have shown that the HCRT system does not only contribute to
arousal maintenance,3, 7 but is also implicated in the regulation of motivated behaviors
and in rewarding effects of addictive drugs.18, 20, 22, 46 Because NC-patients have low or
undetectable levels of HCRT, we propose that narcolepsy with cataplexy may provide a
valid model from the pathology to test the role of HCRT in the human reward system.
The present fMRI study is the first, to our knowledge, to assess reward-related brain
responses in NC-patients.

Page 12 of 47
John Wiley & Sons
Annals of Neurology

Page 13

13
Abnormal Reward-Related Brain Responses in NC
A major result of our study is the lack of VTA activation during the presentation of high
incentive cues in NC-patients. This observation corroborates our initial hypothesis, based
on the animal literature, that HCRT deficiency in NC-patients would affect activity in
ventral midbrain/VTA regions.13, 14 In addition, functional connectivity analyses in the
control group confirmed increased coupling of activity between the VTA and the NAcc
during the presentation of highly motivating cues, thus substantiating enhanced VTA-
NAcc interplay during large incentive conditions. We also found that brain responses to
reward experience (i.e., during successful trials) in NC-patients were affected in regions
receiving dense HCRT projections and/or those modulated by dopaminergic inputs from
the VTA, such as the NAcc and prefrontal cortex.44 Taken together, these results suggest
that the HCRT system, which is deficient in narcolepsy with cataplexy,5, 6, 31 may
contribute to the regulation of brain reward functions, either via direct HRCT projections
or via a modulation of dopaminergic VTA projections.16, 20-22 It has been proposed that
the HCRT system might regulate reward processing by increasing dopamine output via
the potentiation of glutamatergic/opioid synaptic transmission in the VTA, which would
enhance reinforcing effects of rewards.47-49 Accordingly, a dysfunction across reward-
related regions, as revealed by our fMRI study during game playing, may limit the
reinforcing effects of reward in HCRT-deficient NC-patients, and could explain why
psychostimulant abuse is extremely rare these patients.8, 24-26

Page 13 of 47
John Wiley & Sons
Annals of Neurology

Page 14

14
Atypical Ventral-Dorsal Striatum Dissociation in NC
During successful trials, NC-patients showed enhanced activity in the dorsal striatum
(putamen), which is involved in stimulus-action reward associations and which mediates
affective properties of outcomes in rewarding conditions.36, 50-52 Activity increase in this
region is consistent with NC-patients being particularly responsive to emotionally
positive contexts9, 11 and echoes the recent observation by Chabas et al.53 who reported a
hyperperfusion in this same region during a cataplectic episode. Increased dorsal striatum
activity contrasts with the lack of response in ventral striatum during winning (in
particular the NAcc) in NC-patients. Yet, fMRI signal in the NAcc and the vmPFC
during high incentive cues correlated with the duration of narcolepsy disease in the
patients (Fig 5), suggesting that some adaptive mechanisms in patients more experienced
with the disease could restore neural activity in these regions (in the absence of
associated VTA activation). This result is consistent with the patients’ reports of
increased control over cataplexy attacks with time.54 Thus, HCRT depletion may lead to
an atypical ventral-dorsal striatum dissociation during reward processing, which might
also reflect putative compensatory mechanisms in NC-patients.
Because withdrawal from amphetamine-like drugs may modulate striatal
activity,55, 56 we would like to stress that only 6 out of the 12 NC-patients withdrew from
a chronic treatment (modafinil, n=3; modafinil and fluoxetine, n=1; ephedrine, n=1;
clomipramine, n=1; Supplementary Table 1). Note that modafinil has been found to
modulate many neurotransmitter systems including glutamate, GABA, histamine,
dopamine, and hypocretin activity.57 Abnormalities in striatal activity found at the group
Page 14 of 47
John Wiley & Sons
Annals of Neurology

Page 15

15
level during reward processing in the patient population are therefore unlikely to be due
to withdrawal from amphetamine-like drugs.

Putative Role for the HCRT System in Affective Responses to Reward
In the present study, we used a task in which subjects could win or loose points
specifically because NC-patients report that they may experience cataplexy when
confronted to highly positive or motivating situations, such as when winning games.
When comparing winning to loosing on positively-cued trials, NC-patients showed
increased activity in the dorsal striatum (see previous section), as well as in bilateral
amygdala. Enhanced amygdala activity in the patients during positive emotional signals
replicates our previous fMRI findings using humorous pictures as positive emotional
stimuli.11 We also recently reported that NC-patients failed to exhibit amygdala-
dependent startle potentiation during the presentation of unpleasant stimuli.58 These
observations document abnormal emotional processing for both positive and aversive
signals in NC. The HCRT system may modulate amygdala-mediated emotional responses
either via direct HCRT projections to the amygdala,4, 59 or via projections to the VTA,16
which can in turn increase prefrontal dopamine efflux and attenuate amygdala
response.44, 45 In this context, increased vmPFC activation during high motivating cues in
the patients with a longer disease history could reflect a more efficient control over
affective responses to emotional signals.42, 60

Page 15 of 47
John Wiley & Sons
Annals of Neurology