Disruption of NMDAR-dependent burst firing by dopamine neurons provides selective assessment of phasic dopamine-dependent behavior

Abstract

Midbrain dopamine (DA) neurons fire in 2 characteristic modes, tonic and phasic, which are thought to modulate distinct aspects of behavior. However, the inability to selectively disrupt these patterns of activity has hampered the precise definition of the function of these modes of signaling. Here, we addressed the role of phasic DA in learning and other DA-dependent behaviors by attenuating DA neuron burst firing and subsequent DA release, without altering tonic neural activity. Disruption of phasic DA was achieved by selective genetic inactivation of NMDA-type, ionotropic glutamate receptors in DA neurons. Disruption of phasic DA neuron activity impaired the acquisition of numerous conditioned behavioral responses, and dramatically attenuated learning about cues that predicted rewarding and aversive events while leaving many other DA-dependent behaviors unaffected.

Full text

Disruption of NMDAR-dependent burst firing by
dopamine neurons provides selective assessment
of phasic dopamine-dependent behavior
Larry S. Zweifel
a,b
, Jones G. Parker
a,b
, Collin J. Lobb
c
, Aundrea Rainwater
a,b
, Valerie Z. Wall
a,b
, Jonathan P. Fadok
a,b
,
Martin Darvas
a,b
, Min J. Kim
d
, Sheri J. Y. Mizumori
d
, Carlos A. Paladini
c
, Paul E. M. Phillips
e,f
, and Richard D. Palmiter
a,b,1
Departments of dPsychology, ePsychiatry and Behavioral Sciences, fPharmacology, and aBiochemistry and bHoward Hughes Medical Institute, University of
Washington, Seattle, WA 98195; and cDepartment of Biology, University of Texas, San Antonio, TX 78249
This Feature Article is part of a series identified by the Editorial Board as reporting findings of exceptional significance.
Edited by Richard L. Huganir, Johns Hopkins University School of Medicine, Baltimore, MD, and approved February 20, 2009 (received for review
December 31, 2008)
Midbrain dopamine (DA) neurons fire in 2 characteristic modes,
tonic and phasic, which are thought to modulate distinct aspects of
behavior. However, the inability to selectively disrupt these pat-
terns of activity has hampered the precise definition of the func-
tion of these modes of signaling. Here, we addressed the role of
phasic DA in learning and other DA-dependent behaviors by
attenuating DA neuron burst firing and subsequent DA release,
without altering tonic neural activity. Disruption of phasic DA was
achieved by selective genetic inactivation of NMDA-type, iono-
tropic glutamate receptors in DA neurons. Disruption of phasic DA
neuron activity impaired the acquisition of numerous conditioned
behavioral responses, and dramatically attenuated learning about
cues that predicted rewarding and aversive events while leaving
many other DA-dependent behaviors unaffected.
cue-dependent learning 兩mouse behavior 兩electrophysiology 兩
cyclic voltammetry
Dopamine (DA) neurons of the ventral midbrain project to
the dorsal and ventral striatum, as well as to other cortico-
limbic structures such as the hippocampus, amygdala, and pre-
frontal cortex. Differential DA release (tonic or phasic) is
thought to activate distinct signal transduction cascades through
the activation of postsynaptic inhibitory and excitatory G protein
coupled receptors. Phasic DA is proposed to activate excitatory,
low-affinity DA D1-like receptors (Rs) (1, 2) to facilitate
long-term potentiation of excitatory synaptic transmission and
enhance activity of the basal ganglia direct pathway facilitating
appropriate action selection during goal-directed behavior. Con-
versely, tonic DA release is proposed to act on inhibitory,
high-affinity DA D2Rs to facilitate long-term depression of
cortico-striatal synapses and suppress activity of medium spiny
neurons (MSNs) of the basal ganglia indirect pathway (1, 3–5).
Thus, coordinate D1R and D2R activation modulates motor and
cognitive function, and facilitates behavioral flexibility by a
dichotomous control of striatal plasticity (5).
During reinforcement learning shifts in phasic DA neuron
responses from primary rewards, to reward predicting, stimuli
are thought to reflect the acquisition of incentive salience for the
predictive conditioned stimuli (6–10). Coincident DA and glu-
tamate release onto MSNs during conditioned-stimulus re-
sponse learning facilitates long-term potentiation of excitatory
synapses that is thought to underlie reinforcement learning (1, 2,
11). Pharmacological or genetic disruption of D1R signaling
impairs learning in numerous behavioral paradigms (2, 11); thus,
phasic DA acting through D1R is thought to facilitate memory
acquisition by ‘‘stamping-in’’ stimulus-response associations.
Although considerable correlative electrophysiological evi-
dence, as well as pharmacological and genetic evidence, supports
an important role of phasic DA in stamping-in cue-reward
associations, other evidence suggests that DA is not necessary for
learning conditioned-stimulus responses. Mice genetically mod-
ified to be hyperdopaminergic do not learn faster than normal
mice. However, they do demonstrate increased motivation to
work for food reward (12, 13). Also, mice that lack the ability to
synthesize DA (DA-deficient mice) can develop conditioned
reward associations, but lack the motivation to obtain the reward
(14–16). These findings suggest that DA provides an incentive
motivational signal to engage in goal-oriented tasks in response
to learned conditioned stimuli, but is not necessary for learning
conditioned-stimulus associations (17).
Burst firing by DA neurons is mediated, in part, by large
amplitude, slow inactivating excitatory postsynaptic currents
(EPSCs) from NMDARs that allow for the temporal summation
of synaptic inputs (18–20). Iontophoretic administration of
NMDAR antagonists, but not AMPAR-selective antagonists,
attenuates burst firing. Also, NMDAR antagonists attenuate
burst frequency without altering the frequency of nonburst
events (19), suggesting that inactivation of NMDAR signaling in
DA neurons could provide the selectivity necessary to asses the
contribution of phasic DA to DA-dependent behaviors without
producing a complete DA-deficient state.
Results
Genetic Inactivation of NMDAR in DA Neurons Impairs Burst Firing.
Genetic inactivation of the essential NR1 subunit (Grin1)ofthe
NMDAR selectively in neurons expressing the dopamine trans-
porter gene (Slc6a3) is sufficient to inactivate NMDAR currents
in these cells (21, 22). To determine whether burst firing depends
on functional NMDAR signaling, we monitored DA neuron
activity in freely moving control (Slc6a3
⫹/Cre
;Grin1
⫹/lox
) and
knockout (KO, Slc6a3
⫹/Cre
;Grin1
⌬/lox
) mice, chronically im-
planted with recording electrodes in the ventral tegmental
area/substantia nigra pars compacta. Putative DA neurons were
identified by action potential waveform and inhibition by the
D2R autoreceptor, which is present in most, but not all, DA
neurons (23, 24), as described (Fig. S1) (25). In Fig. 1 A
and Bwe show that the wave forms were similar, whereas in
Author contributions: L.S.Z. and R.D.P. designed research; L.S.Z., J.G.P., C.J.L., A.R., V.Z.W.,
J.P.F., and M.D. performed research; M.J.K., S.J.Y.M., C.A.P., P.E.M.P., and R.D.P. contrib-
uted new reagents/analytic tools; L.S.Z., J.G.P., A.R., J.P.F., M.D., and M.J.K. analyzed data;
and L.S.Z. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.
See Commentary on page 7267.
1To whom correspondence should be addressed. E-mail: palmiter@u.washington.edu.
This article contains supporting information online at www.pnas.org/cgi/content/full/
0813415106/DCSupplemental.
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0813415106 PNAS
兩
May 5, 2009
兩
vol. 106
兩
no. 18
兩
7281–7288
NEUROSCIENCE FEATURE ARTICLESEE COMMENTARY

Fig. 1 Cand Dthat quinpirole had similar inhibitory effects in
bothcontrol and KO mice; controls: 81.8 ⫾2.08% inhibition, n⫽
17 cells from 3 mice vs. KO 78.3 ⫾3.81% inhibition, n⫽18 cells
from 4 mice. Phasic activity was defined as bursts of spikes
occurring with an interspike inter val (ISI) of ⱕ80 ms and
terminating with an ISI of ⱖ160 ms (26). NMDAR inactivation
had a significant effect on the pattern of activity reducing the
median frequency of burst events by ⬎6-fold (median burst
sets/s ⫽0.63 Hz control vs. 0.10 Hz, KO, Mann–Whitney Utest
P⬍0.01; see Fig. 1 E–Gand I). The percentage of spikes fired
in bursts (percentage SFB) were similarly reduced (median
percentage SFB ⫽61.4, control vs. 17.3, KO, Mann–Whitney U
test P⬍0.05; see Fig. 1H). We also observed a small reduction
in burst duration (147.5 ⫾18.0 ms, control, vs. 96.2 ⫾17.0 ms,
KO; Student’s ttest P⬍0.05). Total firing rate was reduced in
KO mice, and it correlated with reduced burst set rate (4.86 ⫾
0.61 Hz, control, vs. 2.17 ⫾0.44 Hz, KO; r⫽0.82, Student’s ttest
P⬍0.01; see Fig. 1I). However, the frequency of nonburst spikes
was unaffected (1.66 ⫾0.24 Hz, control, vs. 1.40 ⫾0.35, KO; see
Fig. 1J), indicating that NMDAR inactivation in DA neurons
does not affect tonic activity. Firing rate and bursting activity of
cells that did not fulfill the criteria for DA neurons were similar
between the 2 groups (Fig. S1; average percentage quinpirole
inhibition: 16.5 ⫾7.8; control n⫽13 vs. 9.1 ⫾7.6, KO n⫽11;
median frequency ⫽3.62 Hz, control, vs. 4.04 Hz, KO; median
burst sets/s ⫽0.32 Hz, control, vs. 0.29 Hz, KO; see Fig. 1 K–L).
Burst firing by DA neurons is modulated, in part, by excitatory
(glutamatergic and cholinergic) afferents from the pedunculo-
pontine tegmental nucleus (PPTg), which is thought to relay
cue-related sensory information to these cells (27–29). To con-
firm that burst firing is impaired in KO mice, we assessed stimulus-
evoked burst activity in antidromically-identified DA neurons from
anesthetized control and KO mice (see SI Materials). The success
rate of PPTg-evoked burst firing was higher in control mice than
KO mice (10/19 vs. 5/17 cells). Of those cells in which bursts were
evoked, the percentage of stimulus-evoked bursts and the number
of spikes/burst were reduced in KO mice (35.0 ⫾9.5%, control, vs.
12.1 ⫾3.8%, KO; Mann–Whitney Utest P⬍0.05; median
spikes/burst ⫽3.78, control, vs. 3.00, KO; Mann–Whitney Utest,
P⬍0.05; see Fig. S2). These findings confirm that NMDARs
contribute significantly to burst firing by these cells.
Phasic DA Release Is Impaired in KO Mice. Bursts of DA neuron
activity are thought to facilitate neurotransmitter release, re-
sulting in transient increases in synaptic DA (2). To determine
whether DA release associated with burst firing is altered in KO
mice, we measured PPTg-evoked DA release in the dorsal
striatum using fast-scan cyclic voltammetry (30). Hindbrain
stimulation (0.15 mA at 60 Hz for 1 s) corresponding to the
stereotaxic coordinates for the PPTg reliably evoked DA release
in the dorsal striatum (Fig. 2 Aand B;n⫽9 stimulation electrode
tracts from 6 control, and n⫽5 stimulation tracts from 4 KO mice).
Similar to PPTg-evoked burst firing, the success rate of PPTg-
evoked DA release was twice as high in c ontrol mice compared with
KO mice (n⫽14/24 stimulation sites from 12 control vs. n⫽6/21
stimulation sites from 8 KO mice); only stimulation sites that
evoked release were used in subsequent analysis. Varying PPTg
stimulus intensity and duration had a significant effect on DA
release in control mice that was greatly reduced in KO mice (2-way
repeated measures ANOVA, genotype ⫻stimulus, F
(7, 84)
⫽4.37;
P⬍0.001, and F
(6, 72)
⫽3.26; P⬍0.01, respectively; see Fig. 2 C–F).
To determine whether the releasable pool of DA that can be evoked
by electrical stimulation is altered in KO mice, after PPTg stimu-
lation, we measured DA release evoked by direct stimulation of DA
neuron fibers in the medial forebrain bundle. There was no
significant difference in DA release, with ⬎92% of the stimulation
experiments producing detectable responses in both groups (Fig. 2
G–H;n⫽14/15 stimulation sites, control, vs. n⫽11/12 stimulation
sites, KO). Deficits in PPTg-evoked DA release confirm our
electrophysiology results, and demonstrate that NMDAR inactiva-
tion in DA neurons significantly impairs DA neuron burst firing and
subsequent DA release.
Many DA-Dependent Behaviors Are Unaffected in KO Mice. To assess
whether disruption of phasic DA leads to generalized behavioral
impairment, we performed an extensive analysis of DA-
dependent behaviors (summarized in Table 1). Lack of NMDAR
in DA neurons does not affect 24-hour locomotor activity during
light or dark phase, the locomotor response in a novel environ-
ment, or acute responses to cocaine, amphetamine, morphine, or
D1R agonists (22). Because DA neurons are directly and
indirectly modulated by hormones that regulate feeding behav-
ior (31), we monitored daily ad libitum food consumption and
the latency of calorie-restricted (85% body weight) control and
KO mice to eat freely available food pellets. We did not observe
significant differences between control and KO mice in either
parameter (Fig. S3). Progressive DA deficiency, as observed in
Parkinson’s disease (PD), is associated with impaired motor and
cognitive function (3). To examine motor function, we assessed
the ability of mice to improve their performance on an accel-
erating rotating rod and their latency to escape to a visible
Fig. 1. Burst firing by DA neurons is impaired in KO mice. (Aand B) Waveform
of a DA neuron recorded from a control (A) and KO (B) mouse, and corre-
sponding ISI histogram (10-ms bins). (Cand D) Firing rate histogram (30-s bins)
of the DA neurons in Aand B, from control (C) and KO (D) mouse, demon-
strating sensitivity to D2R agonist, quinpirole (quin), and D2R antagonist,
eticlopride (etic). (Eand F) Burst-rate histogram (30-s bins) of the DA neurons
in Aand B, from control (E) and KO (F) mice. (G) Burst set-rate (burst sets/s) by
DA neurons from control and KO mice. (H) Percentage spikes fired in bursts
(percentage SFB) by DA neurons from control and KO mice. (I) Correlation
between burst set rate and firing rate. (J) Frequency of nonburst spikes. (K)
Firing frequency of non-DA neurons. (L) Burst set-rate of non-DA neurons. (G
and H) Mann–Whitney Utest; *,P⬍0.05; **,P⬍0.01.
7282
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0813415106 Zweifel et al.

platform in a straight-alley, water-escape task, behaviors that are
significantly impaired in DA-deficient mice (32, 33). KO mice
were not significantly impaired in either task (Fig. 3 Aand B).
In addition to modulating sensorimotor function, DA also facili-
tates sustained cortical network activity during working memory
(4), which is impaired in PD (3). We monitored working memory
in control and KO mice in a water-based, T-maze, in which the arms
are bent such that the goal cannot be observed at the choice point.
The mice are presented with a forced choice trial leading to an
escape platform in one arm followed 10 s later by a free choice, in
which the escape platform was located in the opposite arm. KO and
control mice demonstrated equivalent improvement in this task
(Fig. 3C; 2-way repeated measures ANOVA, day: F
(17,324)
⫽11.611;
P⬍0.01). Also, KO mice performed, as well as control mice, in a
novel-object recognition task (Fig. S3).
Altered DA signaling is associated with numerous psychiatric
disorders, including schizophrenia (34). Also, modified behavior
associated with anxiety, sociability, stress, and drug-seeking
behavior are correlated with altered DA neuron activity in mice
(35). To assess anxiety, we monitored the time spent in the open
arm of an elevated, plus maze; however, we did not observe any
difference between KO and control mice (Fig. S3). Likewise,
social interaction did not differ between groups, and the latency
to immobility in a forced-swim test did not differ from controls
(Fig. S3). Alterations of tonic DA signaling in mice are associ-
ated with disruptions of sensor y motor gating in ref lexive startle
paradigms (36). To assess sensory motor gating, we monitored
prepulse inhibition (PPI) of the acoustic startle ref lex; 120-dB
startle pulses were preceded by varying prepulse intensities
above background noise (65 dB), KO mice demonstrated equiv-
alent PPI compared with controls (Fig. S3). Thus, KO mice can
perform many DA-dependent tasks without any apparent im-
pairment. These findings suggest that tonic firing by DA neurons
is sufficient for execution of most behaviors, and that disruption
of phasic DA does not impact performance of these tasks.
Acquisition of Conditioned-Place Preference (CPP) and Learning in a
Water Maze Are Deficient in KO Mice. Drug seeking behavior, as
monitored by acquisition of cocaine CPP, is impaired in these
KO mice during the first 3 days of training (22), but an
association can eventually be formed after 8 context-reward
presentations (21). To assess whether phasic DA facilitates
reinforcement learning for natural rewards, we monitored the
acquisition of food CPP. Food-restricted mice (85% of normal
body weight) were presented with food in 1 of 2 contextually
distinct compartments of a CPP box, and without food in the
other compartment. Pairings of food with context were per-
formed every other day. On intermittent days, mice were tested
Fig. 2. PPTg-evoked DA release is attenuated in KO mice. (A) Schematic
representation of stereotaxic coordinates of stimulating electrode placement
(blue) into caudal (red) and rostral PPTg (red check), representation adapted
from (54). (B) Peak DA oxidation currents from PPTg stimulation at different
depths (mean ⫾SEM). (C) Representative DA oxidation current in response to
increasing stimulus intensity (50 –1,000
␮
Aat60Hzfor1s).(D) Representative
DA oxidation currents in response to decreasing stimulus duration (400
␮
Aat
60 Hz for 83–1,000 ms). (E) Average peak DA oxidation currents in response to
increasing stimulus intensity are reduced in KO compared with control mice
(mean ⫾SEM, 2-way, repeated measures ANOVA; P⬍0.01). (F) Average peak
DA oxidation currents in response to decreasing stimulus duration are reduced
in KO compared with control mice (mean ⫾SEM, 2-way, repeated measures
ANOVA; P⬍0.01). (Gand H) Peak DA oxidation currents after increasing
medial forebrain bundle stimulus intensity or decreasing stimulus duration is
unaltered in KO mice.
Table 1. Summary of behavioral analysis of mice with impaired
phasic DA neuron activity
Food consumption, ad libitum* 7
Latency to eat, free access* 7
Body weight 7
Rotarod 7
Water-escape latency 7
Working memory 7
Novel object recognition* 7
Sociability* 7
Forced-swim test* 7
Elevated plus maze* 7
Prepulse inhibition 7
Locomotor activity, novelty (ref. 22) 7
Locomotor activity, drugs of abuse (refs. 21, 22) 7
Sensitization, cocaine
Acquisition (refs. 21, 22) 7
Withdrawal (ref. 22) 2
Cocaine CPP (refs. 21, 22) 27
Extinction (ref. 21) 7
Reinstatement (ref. 21) 2
Food CPP 2
Cued water maze
Acquisition 2
Recall 7
T-maze 2
FPS 2
Operant conditioning 2
Increase (1,P⬍0.05), decrease (2,P⬍0.05), or no change (7) relative to
control mice. CPP is impaired after 3 days of intermittent cocaine injections
(22), but not after 8 consecutive days (21).
*See SI Methods.
Zweifel et al. PNAS
兩
May 5, 2009
兩
vol. 106
兩
no. 18
兩
7283
NEUROSCIENCE FEATURE ARTICLESEE COMMENTARY

for the development of a preference for the food-paired com-
partment. Preference for the food paired compartment was
significantly impaired in KO mice (n⫽12) relative to controls
(n⫽10) (2-way repeated measures ANOVA, genotype, F
(1, 20)
⫽4.29, P⬍0.05; see Fig. 3D), although food consumption
during the training sessions was equivalent, indicating that the
mice were equally hungry (Fig. S3).
Dopamine signaling has also been demonstrated to modulate
learning in a cue-dependent, Morris water maze, and it is thought
to reflect a disruption of synaptic plasticity within the forebrain
(37, 38). We measured memory acquisition in a modified, Morris
water maze. Mice were given 5 trials per day for 4 days to learn
the location of a hidden platform using cues located within the
maze. KO mice (n⫽8) were significantly slower to learn the
task, as measured by latency to find the hidden platform,
compared with controls (n⫽9) (2-way repeated measures
ANOVA, genotype ⫻day, F
(3,45)
⫽2.88; P⬍0.05; see Fig. 3E);
however, they demonstrated equivalent recall (time spent in
zone where hidden platform was located) once the task was
learned (Fig. 3F). These behavioral analyses suggest selective
impairments in cue-dependent learning.
Phasic DA Neuron Activity Facilitates Learning in T-Maze Tasks. To
further explore whether phasic DA facilitates learning, mice
were trained in an appetitive T-maze task, in which arms were
baited with an accessible food pellet (horizontal stripes, Rew⫹)
in one arm, and an inaccessible food pellet (vertical stripes,
Rew⫺) in the other arm (Fig. 4A), as done previously with
DA-deficient mice (16). Two independent groups of food-
restricted control (n⫽15) and KO mice (n⫽12) were given 10
trials per day for 10 days. Performance of this task (percentage
correct arm choices) was significantly impaired in KO mice
compared with control mice (2-way repeated measures
ANOVA, genotype, F
(1,25)
⫽11.15, P⬍0.01; see Fig. 4B);
however, they eventually made a similar percentage of correct
arm choices in the final 2 days of training. Both KO and control
mice consumed all rewards after a correct arm entry. However,
KO mice appeared slower to make a choice of arms to enter than
control mice. Reduced latency to choice was confirmed in the
second group of mice by quantifying the latencies to choice
(2-way repeated measures ANOVA, genotype, F
(1,13)
⫽6.63, P⬍
0.05; n⫽6 KO, and n⫽9 control; see Fig. 4C).
Although KO mice eventually learned the T-maze task with
repeated training, the reward location did not change. Thus, it
is possible that the mice learned the task in a response-
dependent manner, rather than a cue-dependent manner. To
directly assess the ability of the mice to use the cues to predict
reward availability, mice were trained with cues presented in
pseudorandom order (horizontal stripes, Rew⫹; vertical stripes,
Rew⫺), such that the reward was located in each arm half of the
time for a total of 20 trials per day (Fig. 4D). Maximal time
allotted to make a choice before a forced choice was given was
reduced from 2 to 1 min to facilitate learning. Control mice (n⫽
9) demonstrated significant improvement in the task; however,
KO mice (n⫽8) were significantly impaired relative to controls
(percentage correct choices, 2-way, repeated measures ANOVA,
genotype ⫻day, F
(13,185)
⫽1.81; P⬍0.05; see Fig. 4E). KO mice
were again significantly slower to make a choice relative to
controls (2-way, repeated measures ANOVA, genotype ⫻day,
F
(13,185)
⫽1.86; P⬍0.05; see Fig. 4F).
Phasic DA Is Unnecessary for Motivation to Work for Food Rewards.
Increased latencies to choice in T-maze tasks may reflect deficits
in learning, motivation, or both. To determine whether motiva-
Fig. 3. Selective behavioral impairments in KO mice. (A) Rotarod perfor-
mance during 3 trials per day for 3 consecutive days is not different between
the 2 groups control (n⫽19) and KO mice (n⫽13). (B) Latency to escape to
a visible platform in a straight-alley water-escape task is not different be-
tween control (n⫽13) and KO mice (n⫽14). (C) Performance (percentage
correct choice) in a working-memory task is not impaired in KO mice. (D) CPP
for food is impaired in KO (n⫽12) vs. control (n⫽10) mice (mean ⫾SEM,
2-way, repeated measures ANOVA, Fisher’s LSD; *,P⬍0.05). (E) The acquisi-
tion phase of a cue-dependent Morris water maze is impaired in KO mice
(mean ⫾SEM, 2-way, repeated measures ANOVA, Fisher’s LSD; *,P⬍0.05). (F)
Time spent searching in the area where the hidden platform was located in the
cue-dependent Morris water maze is not different between groups.
Fig. 4. Cue-dependent reward learning is impaired in KO mice. (A) Schematic
representation of T-maze task in which Rew⫹and Rew⫺location did not
change. (B) KO mice were significantly delayed in learning the task (percent-
age correct arm entries) compared with control mice (2-way repeated mea-
sures ANOVA; P⬍0.05). (C) Latency to make a choice is significantly longer in
KO mice compared with controls (2-way repeated measures ANOVA; P⬍0.05).
(D) Schematic representation of T-maze in which Rew⫹and Rew⫺cues were
presented in pseudorandom order. (E) Learning, measured as percentage
correct arm entries was significantly impaired in KO mice compared with
control mice (2-way repeated measures ANOVA; P⬍0.05). (F) Latency to
choice was also significantly impaired in KO mice relative to control mice
(mean ⫾SEM, 2-way, repeated measures ANOVA; P⬍0.05).
7284
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0813415106 Zweifel et al.

tion is impaired in KO mice, we measured their willingness to
work for food in a progressive ratio, instrumental conditioning
task similar to that previously described (39). After 1 week of
pretraining (noncontingent reward pellets delivered coincident
with a lever extension-retraction), instrument al conditioning was
established by using a simple fixed ratio schedule, in which a
lever press delivered a single food pellet (FR1). All KO (n⫽10)
and control (n⫽10) mice reached criterion within 3 days (50
lever presses within 2 h). However, KO mice were significantly
slower to reach criterion on the first day (2-way repeated
measures ANOVA, genotype ⫻day: F
(2,36)
⫽3.66, P⬍0.05;
Fisher’s LSD day 1: P⬍0.05; see Fig. 5A), but not on subsequent
days. Also, KO mice were significantly slower to initiate lever
pressing on the first day (2-way repeated measures ANOVA,
genotype ⫻day: F
(2,36)
⫽3.95, P⬍0.05, Fisher’s LSD: P⬍0.01
day 1; see Fig. 5B), but not on subsequent days (day 2, P⬍0.20
and day 3, P⫽0.10). Assessment of break-point (maximal lever
presses to achieve a single reward pellet) revealed no significant
difference between the 2 groups (Fig. 5C), indicating KO mice
were equally motivated to work for food. When mice were
retested after overnight ad libitum food access to devalue the
food rewards, both groups demonstrated a significant decline in
break-point (2-way repeated measures ANOVA, day: F
(1,18)
⫽
54.01, P⬍0.01; see Fig. 5C). These findings demonstrate that
phasic DA is necessary for cue-dependent reward learning, and
suggest that phasic DA also has a more general role in facilitating
learning about conditioned stimulus-responses (S-R), but not
motivation to work once the S-R is learned.
Phasic DA Neuron Activity Also Facilitates Cue-Dependent Fear Learn-
ing. Some DA neurons are phasically activated by aversive
stimuli, acute stressors, and cues associated with aversive events
(40–42). However, the majority of DA neurons are inhibited by
these stimuli (43). Because DA neurons segregate anatomically,
pharmacologically, and electrophysiologically (23), it is difficult
to generalize their function, which could explain the equivocal
results related to DA neuron activity in response to aversive
stimuli (40–43). To determine whether phasic DA is important
for cue-dependent fear, we assessed learning in a pavlovian
fear-potenitated acoustic startle (FPS) paradigm. Because the
acoustic startle response is ref lexive, it can be examined inde-
pendently of motivation (44). Fear-conditioning was assessed the
day after a conditioning session (10 presentations of a cue that
coterminated with a 0.2-mA footshock) by measuring acoustic
startle responses in the presence or absence of the cue; training
and testing were repeated on subsequent days. After the second
and third conditioning days, control mice (n⫽15) developed
FPS that was absent in KO mice (n⫽16) (2-way, repeated
measures ANOVA, genotype ⫻day, F
(3,87)
⫽5.57, P⬍0.01; see
Fig. 6A). KO mice showed significantly elevated startle re-
sponses, compared with controls, in the absence of the cue on all
test days after a conditioning session (2-way, repeated measures
ANOVA, genotype ⫻day, F
(3,87)
⫽8.52, P⬍0.01) that was the
same as their startle responses in the presence of the cue (Fig.
6B). Thus, KO mice manifested generalized fear responses, but
did not learn to discriminate the cue that predicted the foot-
shock. Control mice also had potentiated responses to acoustic
startle in the absence of the cue after a single training session.
However, this response diminished with further training (Fig.
6B; control: baseline no cue vs. no cue test 1; P⬍0.01; vs. test
2, P⬎0.5; test 3, P⬎0.1), indicating a learned association of the
cue that predicted the footshock. These findings demonstrate
that phasic DA neuron activity is also important for learning
about cues that predict fearful events.
Discussion
Here, we show that NMDARs in DA neurons modulate burst
firing and DA release in postsynaptic brain regions. Remarkably,
the absence of burst firing leaves many DA-dependent behaviors
intact (body weight regulation, working memory, and motor
performance); however, selectively impairs learning in cue-
dependent learning tasks (see Table 1). Some of the behavioral
tasks involve food rewards (CPP, T-maze, instrumental learn-
ing), some involve escape from an unpleasant environment
(Morris water maze), whereas some learning situations are
clearly aversive (FPS paradigm). A unifying interpretation of
these results is that bursts of DA neuron activity in response to
important events provide generalized salience signals that facil-
itate learning associations of environmental cues with these
events, whereas increased tonic DA, such as those measured by
microdialysis, is independent of burst firing, and provides suf-
ficient DA to engage most behaviors.
Fig. 5. Motivation to work for food reward is not impaired in KO mice. (A) Cumulative lever presses in an instrumental task during day 1 is significantly delayed
in KO mice compared with control mice (P⬍0.05). (B) Latency to initiate lever pressing is also significantly delayed on day 1 of preconditioning (mean ⫾SEM,
2-way repeated measures ANOVA, Fisher’s LSD; **,P⬍0.01). (C) Break point in a progressive ratio task is equivalent in KO and control mice (day 1), and is equally
reduced after 24 h ad libitum food access (day 2, 2-way repeated measures ANOVA, Fisher’s LSD; **,P⬍0.01).
Fig. 6. Cued fear is attenuated in KO mice. (A) FPS is significantly attenuated
in KO mice (mean ⫾SEM, 2-way repeated measures ANOVA; P⬍0.01). (B)
Acoustic startle responses in the absence of the cue is significantly elevated in
KO mice compared with controls (mean ⫾SEM, 2-way, repeated measures
ANOVA; P⬍0.01, Tukey’s HSD; **,P⬍0.01; *,P⬍0.05).
Zweifel et al. PNAS
兩
May 5, 2009
兩
vol. 106
兩
no. 18
兩
7285
NEUROSCIENCE FEATURE ARTICLESEE COMMENTARY

Lack of NMDARs in DA neurons not only impairs burst firing,
but also precludes LTP of synaptic AMPARs (21, 22). AMPAR
currents are transiently potentiated in DA neurons after expo-
sure to cocaine, stress, or during learning paradigms (21, 22,
45–47). However, the role of AMPAR LTP in DA neurons is
unclear. Synaptic scaling after NMDAR inactivation in DA
neurons leads to chronically elevated AMPAR currents in KO
mice, similar to levels normally observed after exposure to
cocaine, stress, or during learning (21, 22, 45– 47). Despite the
enhanced level of AMPAR in DA neurons of KO mice, which
one might suspect would enhance firing rate (20), burst firing was
dramatically attenuated in KO mice, and the tonic firing rate was
unaffected. This result is consistent with reports that A MPAR
do not potently modulate burst firing by DA neurons (20), and
observations that numerous behaviors thought to be dependent
on tonic DA signaling [acute locomotor responses to drugs such
as cocaine, amphetamine and morphine (22), rotarod perfor-
mance, working memory, and others; see Table 1] are unaltered
in KO mice. Transient increases in AMPAR currents that have
been demonstrated during conditioned-stimulus reward associ-
ations may provide an important gate for NMDAR-mediated
burst firing by DA neurons. We suggest that spike-timing-
dependent plasticity, in which local dendritic calcium influx
through NMDAR, together with elevated global calcium gen-
erated by NMDAR-dependent burst firing work synergistically
to increase synaptic AMPARs (48). Increased synaptic AMPAR
currents, in turn, facilitate removal of the magnesium block from
the NMDAR; thus, increasing the probability of burst firing.
Previous studies have demonstrated that both hyper and
hypodopaminergic mice can learn various tasks. However, mo-
tivation to engage in the tasks is significantly altered in these
mice, suggesting that DA mediates ‘‘wanting’’ rewards, rather
than learning (17). For example, DA-deficient mice appear to be
unmotivated and will not engage in most tasks (49), whereas
hyperdopaminergic mice perform some tasks more rapidly with
less meandering, suggesting enhanced motivation (13). We ob-
served longer latencies by the KO mice to make choices and
engage in goal-directed behavior in many learning paradigms,
suggesting that they may be less motivated in the absence of burst
firing. Burst-firing increases the probability of neurotransmitter
release (50); thus, bursting by DA neurons likely increases
extracellular DA in target areas that are necessary to engage in
reward-based tasks. Consistent with this idea, elevated synaptic
DA associated with burst firing is proposed to activate D1R (1),
and D1R antagonists decrease the probability of cue-elicited
approach responses in the early stages of S-R training (51). We
observed several cases in which learning was significantly de-
layed in KO mice (water-maze, instrumental conditioning, T-
maze with stationar y cue; also, see Table 1). However, perfor-
mance often became equivalent or nearly equivalent with further
training. Impaired acquisition of some learned behaviors in KO
mice is consistent with a role for phasic DA in facilitating the
acquisition of incentive value for environmental cues that in turn
facilitate engagement of goal-directed behavior (52). However,
with repeated training, responses become habitual, and phasic
DA is no longer necessary. Thus, during early stages of S-R
conditioning, phasic DA facilitates learning. Delays in the
acquisition of the S-R association would in turn manifest as
increased latencies to engage the behavior, making it appear as
if the mice were less motivated. Despite the observation that KO
mice may be slower to make a choice in the appetitive T-maze
tasks, food rewards were always eaten once found. Another
aspect of motivation is the willingness to work for food rewards
(53). We examined this aspect of motivation using the progres-
sive ratio strategy, and found that, although KO mice were
delayed in the acquisition of the S-R association, they were as
motivated as controls once the response was acquired; they
would both press a lever ⬇200 times for a single food pellet. Also,
when the value of the food reward was devalued by prior feeding,
lever pressing declined by a similar amount in both groups.
To generalize, we propose that exposure to a salient event
stimulates excitatory inputs onto DA neurons, and perhaps
reduces inhibitory inputs, which facilitates activation of
NMDARs, allowing calcium inf lux and promoting burst firing
activity. The bursts of activity by DA neurons result in transient
spikes of extracellular DA in synapses within striatum, prefrontal
cortex, amygdala, and/or hippocampus. The transient elevation
in DA concentration would preferentially activate the lower-
affinity D1R, and facilitate LTP in brain regions involved in
spatial memory. Subsequent exposure to cues within the envi-
ronmental context would evoke phasic DA release that would
facilitate engagement of goal-directed behavior. However, with
extended conditioning, these responses would become habitual,
and bursts of DA release would become less important for the
learned response. Conversely, the requirement of phasic DA in
making accurate choices based on discrete cues, such as choosing
to enter 1 of 2 arms of a T-maze to acquire a food reward, or
predict a footshock, does not significantly diminish with repeated
conditioning. Thus, phasic DA remains essential in more com-
plex processes such as 2-choice discrimination.
Materials and Methods
Animals. All behavioral and electrophysiology experiments were approved by
the University of Washington and University of Texas, San Antonio Institu-
tional Animal Care and Use Committees. The KO and control mice were
generated by the breeding scheme described (22). The genetic background of
the mice was almost completely C57BL/6 as a consequence of extensive
backcrossing. Experiments were performed on 8- to 12-week-old male and
female mice, except for electrophysiology, which was conducted in 10- to
12-week-old male mice. During calorie restriction, mice were individually
housed in environmentally enriched cages, and maintained on high-energy
chow (LabDiet 5lJ5) to 85% body weight for a minimum of 1 week.
Electrophysiology. Electrophysiology in freely moving mice was performed by
using HS-16, 4-tetrode microdrives (Neuralynx). Microdrives were implanted
in anesthetized mice by using stereotaxic coordinates for the VTA (3.5 mm A-P,
0.5 mm M-L, and 4.0 mm D-V). Two weeks after surgery, mice were connected
through an HS-16 headstage preamplifier to an ERP27 patch panel, signals
were amplified (200- and 8,000-fold) and filtered (600– 6,000 Hz) by using a
Lynx-8 programmable amplifier, and data were acquired by using Cheetah
acquisition software (Neuralynx). Tetrodes were lowered by ⬇50-
␮
m incre-
ments each day until putative DA neurons were identified by action potential
waveform and sensitivity to quinpirole (Sigma; 0.2 mg mL⫺1i.p.; ⬎70%
inhibition of baseline frequency) and eticlopride (Sigma; 0. mg mL⫺1i.p.;
return to ⬎70% baseline frequency). Baseline DA neuron firing properties
were recorded for 10 min, followed by treatment with confirmation drugs for
10 min each. Tetrode placement was confirmed postmortem by cresyl violet
staining of midbrain sections. Neurons were isolated by cluster analysis using
Offline Sorter software (Plexon). Clustered waveforms were subsequently
analyzed by using MATLAB software (Mathworks). Baseline activity was used
to calculate burst sets (burst onset, ISI of ⱕ80 ms; burst offset, ISI of ⱖ160 ms),
burst set rate (burst sets/s), percentage spikes fired in bursts (burst spike/total
spikes), spikes/burst, burst duration, and firing frequency (total spikes/s). Data
were analyzed by ttest unless normality tests failed, in which case Mann–
Whitney Utests were performed by using Statistica software (Statsoft).
Fast-Scan Cyclic Voltammetry. Fast-scan cyclic voltammetry was performed
using glass-encased, carbon-fiber microelectrodes. A 0.15-mm diameter bipo-
lar stimulating electrode (Plastics One) was used with an analog stimulus
isolator (Model 2200, A-M Systems, Inc.). Stimulation patterns were generated
using Tarheel CV (National Instruments). Mice were anesthetized with 1.5 g/kg
urehane (i.p.) (Sigma), and electrodes were placed based on stereotaxic
alignment. All anterior-posterior (AP) and medial-lateral (ML) coordinates are
reported in millimeter distance from Bregma unless otherwise noted; all
dorsal-ventral (DV) coordinates are millimeter from dura. The carbon-fiber
microelectrode was placed in the dorsal striatum (AP ⫽1.1, ML ⫽1.2, and DV ⫽
⫺2.35). The carbon-fiber was cycled at 60 Hz to allow the electrode to
equilibrate and switched to 10 Hz for data acquisition. The reference elec-
trode was placed AP ⫽4.9 and ML ⫽0.0. For pedunculopontine tegmental
7286
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0813415106 Zweifel et al.

nucleus (PPTg) stimulations, the stimulating electrode was lowered in 0.1-mm
increments at AP ⫽⫺0.68 from lambda and ML ⫽0.7 until dopamine release
was observed. The working electrode was lowered in 0.1-mm increments from
DV ⫽⫺1.5 until dopamine release was observed. The average DV coordinate
for maximal PPTg-stimulated dopamine release was DV ⫽⫺2.69. Following
PPTg stimulation, the stimulating electrode was lowered into the median
forebrain bundle at AP ⫽⫺2.4 and ML ⫽1.1. As described for PPTg stimula-
tions, the electrode was lowered in 0.1-mm increments from DV ⫽⫺3.0 until
dopamine was recorded at the working electrode. The electrode was posi-
tioned where maximum stimulation evoked dopamine was observed, which,
on average, occurred at DV ⫽⫺4.89. With the stimulating electrode in place,
a stimulation-response pattern was obtained by increasing stimulation cur-
rent at 60 Hz and 60 pulses from 50 mA-400 mA and then decreasing the
stimulation duration at 60 Hz and 400 mA from 60 to 5 pulses. For each
stimulation parameter 2 stimulations were conducted, and the current re-
sponse was recorded as the average of the peak dopamine oxidation current
in response to each stimulation. Following surgeries, stimulating electrode
placement was confirmed by cresyl violet staining of hindbrain sections. Data
were analyzed by repeated-measures ANOVA using Statistica software (Statsoft).
Behavioral Testing.
CPP.
Food CPP was performed by using the same procedure
used for cocaine CPP (22).
Water-escape task.
Mice were tested in this task essentially as described (32);
latency to reach the platform was scored.
Rotarod performance.
Mice were testing on an accelerating rotarod as described
(33); latency to fall was scored.
Working memory.
Before discrimination testing in a water based, T-maze, in
which the arms are bent so that the goal cannot be seen at the choice point,
mice were given 5 trials in a water escape task (60 s) in the pool to acclimate
them to swimming. Mice received 6 trials per day, consisting of a sample run,
in which mice were forced to choose one arm by the presence of a door
blocking the entrance to the other arm, according to a pseudorandom se-
quence (equal number of left and right turns per day and with no ⬎2
consecutive turns in the same direction). After completion of the forced
choice, the animal was allowed to rest for 10 s, the door was then removed, the
animal placed at the start position, and a free choice was given with both arms
available and the escape platform located in the alternate arm of the forced
run. Entry into the wrong arm resulted in the mouse being locked in that arm
for 10 s, then allowed to swim to the escape platform and rest for 10 s. The
intertrial interval (ITI) between pairs was 10 min.
Morris water maze.
Water maze performance was assessed by using a small,
3-lobe pool (90-cm diameter) filled to 10 cm with tepid water containing
nonfat dry milk. The hidden platform (weighted white plastic box) was
submerged 1 cm below the pool surface next to 1 of 3 cues. On the first day,
mice were placed on the hidden platform for 30 s, followed by 5 conditioning
trials separated by 10 min. For each trail, mice were placed at new start
location within the pool; 5 conditioning trails were given each day for a total
of 4-consecutive days. On the fifth day, the hidden platform was removed,
mice were placed in the center of the pool, and time spent in the area around
the 3 cues was monitiored for 30 s by using a video acquisition system (Canopus
MediaCruise), data were acquired by using Ethovision software (Noldus In-
formation Technology). Data were analyzed by repeated-measures ANOVA by
using Statistica software (Statsoft).
T-maze.
Performance in cue-dependent T-maze was measured as total correct
arm entries over 20 trials for 13 consecutive days. Mice were allotted 60 s to
leave the start box and make a choice (⬎50% of body across the plane
separating the center chamber from the arm), after which a forced choice to
either the Rew⫹or Rew⫺arm was given and scored as incorrect. Mice were
given 60 s to consume the reward pellet after a correct choice (all reward
pellets were consumed). After reward consumption, mice were immediately
returned to the start box. After an incorrect choice, mice were retained in the
Rew⫺arm for 60 s before being returned to the start box. To assess blocking,
mice were presented with a second set of distinct cues paired with the original
Rew⫹or Rew⫺cue during the last 60 trials, followed by a 20 trial test session,
in which only the second set of cues were presented. Each trial was separated
by an average ITI of 20 s. Data were analyzed by repeated-measures ANOVA
by using Statistica software (Statsoft).
Instrumental conditioning.
Instrumental conditioning was measured in sound-
attenuated, operant chambers (ENV-300; Med Associates). Calorie-restricted
mice received 7 consecutive days of preconditioning, in which 15 food pellets
(20 mg, Bio-Serv) were delivered immediately after a lever-extension for 8 s,
followed by a lever retraction with an average ITI of 90 s. For instrumental
conditioning, sessions began with the simultaneous illumination of the house-
light and extension of both levers. Sessions lasted for2horuntil 50 lever
presses were recorded. Noncontingent food pellets were delivered randomly
(⬇1 per minute) during the first 15 min of each session. At the end of each
session, the houselight and fan were extinguished, and both levers retracted.
Each subject was then placed back in their home-cage. Food pellets left
uneaten in the food hopper were recorded and removed. To assess break
point, a progressive ratio schedule was used by increasing a nonarithmetic
fixed ratio/reinforcement schedule. Data were analyzed by repeated-
measures ANOVA using Statistica software (Statsoft).
FPS and PPI.
FPS and PPI were measured by using sound-attenuated acoustic
startle boxes (San Diego Instruments). A 7-day FPS paradigm was used. For PPI
experiments, mice were given a 10-min habituation period before the test
begun. Throughout the entire test, the background noise level was main-
tained at 65 dB. After the habituation, mice were presented with 5, 40-ms
duration 120 dB, pulse-alone trials to obtain baseline startle responses. Mice
were then presented with 50 trials of either a startle pulse-alone trial, 1 of 3
prepulse trials, or a null trial, in which there was no acoustic stimulus. The ITI
averaged 15 s, (range of 5 to 25 s). A startle trial consisted of a 40 ms, 120-dB
pulse of white noise. The 3 types of prepulse trials consisted of a 20-ms
prepulse of 70-, 75-, or 80-dB intensity (5, 10, and 15 dB above background)
that preceded the 40 ms, 120-dB pulse by 100 ms. Peak amplitude of the startle
response, occurring in the first 65 ms after pulse onset, was used as the
measure of startle response magnitude. For FPS, day 1 (baseline) consisted of
a 5-min habituation period, followed by a series of 20 trials, split evenly
between 2 trial types. The trial types were startle pulse alone, or startle pulse
in the presence of the cue. On startle pulse alone trials, animals were pre-
sented with a 40 ms, 105-dB acoustic pulse. On cue trials, the animals were
presented with a 10 s light cue, which coterminated with a 40 ms, 105-dB
acoustic pulse. These trials were presented in pseudorandom order. The ITI
ranged from 60 to 180 s, (average of 120 s). Throughout the experiment, the
background sound level was maintained at 65 dB. Peak amplitude of the
startle response occurring in the 65 ms after pulse onset was used as
the measure of the acoustic startle response. On days 2, 4, and 6, animals were
placed into the chambers and, after a 10-min habituation period, were given
10 presentations of the cue light, which coterminated with a 0.2 mA, 0.5-s
footshock. The ITI ranged from 60 to 180 s (average of 110 s). Peak responses
occurring during the 500-ms footshock were recorded and averaged for each
animal. The test sessions occurred on days 3, 5, and 7, and were identical to the
baseline session described above. Percentage of FPS was calculated for each
animal. Data were analyzed by repeated-measures ANOVA by using Statistica
software (Statsoft).
ACKNOWLEDGMENTS. We thank Drs. Joe Tsien (Medical College of Georgia,
Augusta, GA) and Xiaoxi Zhuang (University of Chicago, Chicago) for provid-
ing mice with the conditional Grin1lox allele and the Slc6a3Cre allele, respec-
tively. This work was supported by National Institutes of Health Grants
F32DA022829 (to L.S.Z.), MH079276 (to C.A.P.), and MH58755 (to S.J.Y.M.).
1. Goto Y, Grace AA (2005) Dopaminergic modulation of limbic and cortical drive of
nucleus accumbens in goal-directed behavior. Nat Neurosci 8:805–812.
2. Grace AA, et al. (2007) Regulation of firing of dopaminergic neurons and control of
goal-directed behaviors. Trends Neurosci 30:220–227.
3. Frank MJ (2005) Dynamic dopamine modulation in the basal ganglia: A neurocompu-
tational account of cognitive deficits in medicated and nonmedicated Parkinsonism. J
Cogn Neurosci 17:51–72.
4. Wang M, Vijayraghavan S, Goldman-Rakic PS (2004) Selective D2 receptor actions on
the functional circuitry of working memory. Science 303:853–856.
5. Shen W, et al. (2008) Dichotomous dopaminergic control of striatal synaptic plasticity.
Science 321:848– 851.
6. Bayer HM, Lau B, Glimcher PW (2007), Statistics of midbrain dopamine neuron spike
trains in the awake primate. J Neurophysiol 98:1428–1439.
7. Roitman MF, et al. (2004) Dopamine operates as a subsecond modulator of food
seeking. J Neurosci 24:1265–1271.
8. Schultz W, Apicella P, Ljungberg T (1993) Responses of monkey dopamine neurons to
reward and conditioned stimuli during successive steps of learning a delayed response
task. J Neurosci 13:900–913.
9. Schultz W (1998) Predictive reward signal of dopamine neurons. J Neurophysiol
80:1–27.
10. Wise RA (2006) Role of brain dopamine in food reward and reinforcement. Philos Trans
R Soc Lond B Biol Sci 361:1149–1158.
11. Lisman JE. Grace AA (2005) The hippocampal-VTA loop: Controlling the entry of
information into long-term memory. Neuron 46:703–713.
12. Cagniard B, et al. (2006) Mice with chronically elevated dopamine exhibit enhanced motiva-
tion, but not learning, for a food reward. Neuropsychopharmacology 31:1362–1370.
13. Cagniard B, et al. (2006) Dopamine scales performance in the absence of new learning.
Neuron 51:541–547.
14. Hnasko TS, Sotak BN, Palmiter RD (2005) Morphine reward in dopamine-deficient mice.
Nature 438:854– 857.
Zweifel et al. PNAS
兩
May 5, 2009
兩
vol. 106
兩
no. 18
兩
7287
NEUROSCIENCE FEATURE ARTICLESEE COMMENTARY

15. Hnasko TS, Sotak BN, Palmiter RD (2007) Cocaine-conditioned place preference by
dopamine-deficient mice is mediated by serotonin. J Neurosci 27:12484–12488.
16. Robinson S, et al. (2004) Distinguishing whether dopamine regulates liking, wanting,
and/or learning about rewards. Behav Neurosci 119:5–15.
17. Berridge KC (2007) The debate over dopamine’s role in reward: The case for incentive
salience. Psychopharmacology 191:391–431.
18. Komendantov AO, et al. (2004) A modeling study suggests complementary roles for
GABAA and NMDA receptors and the SK channel in regulating the firing pattern in
midbrain dopamine neurons. J Neurophysiol 91:346–357.
19. Overton P, Clark D (1992) Iontophoretically administered drugs acting at the N-methyl-
D-aspartate receptor modulate burst firing in A9 dopamine neurons in the rat. Synapse
10:131–140.
20. Overton PG, Clark D (1997) Burst firing in midbrain dopaminergic neurons. Brain Res
Brain Res Rev 25:312–334.
21. Engblom D, et al. (2008) Glutamate receptors on dopamine neurons control the
persistence of cocaine seeking. Neuron 59:497–508.
22. Zweifel LS, et al. (2008) Role of NMDA receptors in dopamine neurons for plasticity and
addictive behaviors. Neuron 59:486– 496.
23. Lammel S, et al. (2008) Unique properties of mesoprefrontal neurons within a dual
mesocorticolimbic dopamine system. Neuron 57:760–773.
24. Margolis EB, et al. (2006) The ventral tegmental area revisited: Is there an electrophys-
iological marker for dopaminergic neurons? J Physiol 577:907–924.
25. Robinson S, et al. (2004) Firing properties of dopamine neurons in freely moving
dopamine-deficient mice: Effects of dopamine receptor activation and anesthesia.
Proc Natl Acad Sci USA 101:13329–13334.
26. Grace AA, Bunney BS (1984) The control of firing pattern in nigral dopamine neurons:
Burst firing. J Neurosci 4:2877–2890.
27. Pan WX, Hyland BI (2005) Pedunculopontine tegmental nucleus controls conditioned
responses of midbrain dopamine neurons in behaving rats. J Neurosci 25:4725–4732.
28. Floresco SB, et al. Afferent modulation of dopamine neuron firing differentially
regulates tonic and phasic dopamine transmission. Nat Neurosci 6:968–973.
29. Lokwan SJ, et al. (1999) Stimulation of the pedunculopontine tegmental nucleus in the
rat produces burst firing in A9 dopaminergic neurons. Neuroscience 92:245–254.
30. Phillips PE, et al. (2003) Real-time measurements of phasic changes in extracellular
dopamine concentration in freely moving rats by fast-scan cyclic voltammetry. Meth-
ods Mol Med 79:443–464.
31. Palmiter RD (2007) Is dopamine a physiologically relevant mediator of feeding behav-
ior? Trends Neurosci 30:375–381.
32. Denenberg VH, Kim DS, Palmiter RD (2004) The role of dopamine in learning, memory,
and performance of a water escape task. Behav Brain Res 148:73–78.
33. Zhou QY, Palmiter RD (1995) Dopamine-deficient mice are severely hypoactive, adipsic,
and aphagic. Cell 83:1197–1209.
34. Sawa A, Snyder SH (2002) Schizophrenia: Diverse approaches to a complex disease.
Science 296:692–695.
35. Krishnan V, et al. (2007) Molecular adaptations underlying susceptibility and resistance
to social defeat in brain reward regions. Cell 131:391–404.
36. Ralph RJ, et al. (2001) Prepulse inhibition deficits and perseverative motor patterns in
dopamine transporter knock-out mice: Differential effects of D1 and D2 receptor
antagonists. J Neurosci 21:305–313.
37. El-Ghundi M, et al. (1999) Spatial learning deficit in dopamine D(1) receptor knockout
mice. Eur J Pharmacol 383:95–106.
38. Morris RG, et al. (2003) Elements of a neurobiological theory of the hippocampus: The
role of activity-dependent synaptic plasticity in memory. Philos Trans R Soc London B
Biol Sci 358:773–786.
39. Robinson S, et al. (2007) Viral restoration of dopamine signaling to the dorsal striatum
restores instrumental conditioning to dopamine-deficient mice. Psychopharmacology
191:567–578.
40. Mirenowicz J, Schultz W (1996) Preferential activation of midbrain dopamine neurons
by appetitive rather than aversive stimuli. Nature 379:449– 451.
41. Guarraci FA, Kapp BS (1999) An electrophysiological characterization of ventral teg-
mental area dopaminergic neurons during differential pavlovian fear conditioning in
the awake rabbit. Behav Brain Res 99:169–179.
42. Trulson ME, Preussler DW (1984) Dopamine-containing ventral tegmental area neu-
rons in freely moving cats: Activity during the sleep-waking cycle and effects of stress.
Exp Neurol 83:367–377.
43. Ungless MA, Magill PJ, Bolam JP (2004) Uniform inhibition of dopamine neurons in the
ventral tegmental area by aversive stimuli. Science 303:2040–2042.
44. Koch M (1999) The neurobiology of startle. Prog Neurobiol 59:107–128.
45. Chen BT, et al. (2005) Cocaine but not natural reward self-administration nor passive
cocaine infusion produces persistent LTP in the VTA. Neuron 59:288–297.
46. Stuber GD, et al. (2008) Reward-Predictive Cues Enhance Excitatory Synaptic Strength
onto Midbrain Dopamine Neurons. Science 321:1690–1692.
47. Saal D, et al. (2003) Drugs of abuse and stress trigger a common synaptic adaptation in
dopamine neurons. Neuron 37:577–582.
48. Dan Y, Poo MM (2006) Spike timing-dependent plasticity: From synapse to perception.
Physiol Rev 86:1033–1048.
49. Palmiter RD (2008) Dopamine signaling in the dorsal striatum is essential for
motivated behaviors: Lessons from dopamine-deficient mice. Ann NY Acad Sci
1129:35–46.
50. Krahe R, Gabbiani F (2004) Burst firing in sensory systems. Nat Rev Neurosci 5:13–23.
51. Choi WY, Balsam PD, Horvitz JC (2005) Extended habit training reduces dopamine
mediation of appetitive response expression. J Neurosci 25:6729– 6733.
52. Horvitz JC, et al. (2007) A ‘‘good parent’’ function of dopamine: Transient modulation
of learning and performance during early stages of training. Ann NY Acad Sci
1104:270–288.
53. Salamone JD, et al. (2003) Nucleus accumbens dopamine and the regulation of effort
in food-seeking behavior: Implications for studies of natural motivation, psychiatry,
and drug abuse. J Pharmacol Exp Ther 305:1–8.
54. Paxinos G, Franklin K (2001) The Mouse Brain in Stereotaxic Coordinates (Academic,
New York), 2nd Ed.
7288
兩
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0813415106 Zweifel et al.