COMMENTARY
‘‘To learn, you must pay attention.’’ Molecular insights
into teachers’ wisdom
Marc G. Carona,1 and R. Mark Wightmanb
aDepartments of Cell Biology, Medicine, and Neurobiology, Duke University Medical Center, Durham, NC 27710; and bDepartment of
Chemistry, University of North Carolina, Chapel Hill, NC 27599
T
he neuromodulator dopamine
(DA) regulates a series of phys-
iological functions in the brain
from the control of locomotion,
emotion, and affect, to mechanisms of
reward and cognition. Neurons that syn-
thesize DA originate mostly from 2 spe-
cialized areas of the basal ganglia, the
substantia nigra and the ventral tegmen-
tal area and project, respectively, to the
dorsal striatum to form the nigrostriatal
pathway and the nucleus accumbens,
amygdala, and hippocampus, as well as
the frontal cortex to form the mesolim-
bic and mesocortical pathways (1). It
has been established electrophysiologi-
cally that dopaminergic neurons charac-
teristically exhibit 2 distinct modes of
firing, a slow or tonic and a burst or
phasic mode (2). Phasic firing of dopa-
minergic neurons has been shown to
occur in response to cues that predict
reward. Indeed, the activity of DA neu-
rons follows the expectations of a pre-
diction error that reflects the difference
between predicted and delivered re-
wards (3). Neurochemical studies have
shown that reward-predicting cues evoke
DA concentration transients in terminal
regions, the predicted outcome of phasic
firing (4). Thus, a large body of evi-
dence supports a central role for phasic
DA release in reward-driven learning.
However, it has been difficult to deter-
mine with certainty the trigger that ini-
tiates this phasic firing. Likewise, the
behavioral roles of phasic and tonic fir-
ing of DA neurons have been unclear.
Answers to these questions are emerging
from a study in this issue of PNAS (5),
in which burst firing and phasic release
of DA are found to be associated only
with cue-conditioned behaviors that pre-
dict reward or aversion. These findings
imply that tonic firing of DA neurons is
sufficient to mediate the large repertoire
of behaviors normally attributed to DA.
NMDA Receptors Trigger Phasic Firing
Attempts to examine the role of DA in
these reward-learning paradigms so far
have used genetic models in which the
dopaminergic tone of the animals was
either increased through knockdown of
the DA transporter or decreased by de-
pletion of DA synthesis. These approaches
revealed that increased dopaminergic
tone in animals does not enhance learn-
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0903306106
ing, despite the fact that motivational
levels for reward are increased (6).
However, mice lacking DA still worked
for a reward but appeared less moti-
vated (7). Thus, more sophisticated
models would be needed to resolve the
issues of triggering firing and associated
behaviors.
Over the past year, 2 different groups
have developed genetically modified
mice that could approach this question
more directly. By the use of very similar
approaches, a deletion of the gene
(Grin1) for the NR1 subunit of the
NMDA receptor was obtained with Cre
A large body of evidence
supports a central role
for phasic DA release in
reward-driven learning.
recombinase expressed either in the lo-
cus of the DA transporter (DAT) or via
a BAC-DAT transgene (8, 9). Both
studies initially demonstrated that, al-
though the strengthening of excitatory
synaptic transmission (LTP) usually as-
sociated with exposure to the psycho-
stimulant cocaine was absent in the
knockout (KO) mice, the initial locomo-
tor response and sensitization to cocaine
were still normal, suggesting that initial
phases of the response to the psycho-
stimulant did not depend on the neuro-
nal plasticity. However, these mice
showed more subtle deficits either in
cocaine withdrawal-induced increases in
sensitization of locomotion or in cue-
dependent reinstatement to the drug
suggesting that a functional NMDA re-
ceptor is necessary for contextual or
cue-induced events associated with the
drug.
The generation of burst firing in DA
neurons is thought to result from
NMDA-mediated signals, because selec-
tive blockade of NMDA rather than
AMPA receptors diminishes the fre-
quency of bursting without affecting
tonic firing (10). Thus, Zweifel et al. (5)
reasoned that genetic inactivation of the
NMDA receptor in DA neurons should
provide a way to selectively attenuate
PNAS 兩 May 5, 2009 兩 vol. 106 兩 no. 18 兩 7267–7268
DA neuron burst firing to assess the
contribution of phasic release of DA to
various behavioral paradigms.
In their feature article, Zweifel et al.
(5) used the DA neuron-NR1–deficient
mice to attempt to answer that specific
question. They found that, in the ab-
sence of a functional NMDA receptor,
generation of phasic firing is markedly
impaired and this effect is selective for
DA neurons. Furthermore, they demon-
strate with fast-scan cyclic voltammetry
that, although DA release in the
NR1DATCre mice can be evoked with
direct stimulations that resemble phasic
firing patterns, release was diminished
relative to wild-type mice when the
stimulation was directed at the glu-
tamergic inputs that synapse onto dopa-
minergic cells in the VTA.
Specific Attentional Deficits in
NR1DATCre-KO Mice
When the KO mice were compared with
controls in a large battery of behavioral
tests, despite their marked reduction in
the burst firing frequency and phasic
DA release, mice lacking a functional
NMDA receptor amazingly displayed a
whole series of normal behaviors com-
monly associated with the function of
DA. NR1DATCre mice showed normal
locomotor activity throughout the light–
dark phases and in response either to
novel environment, psychostimulant,
morphine, or DA receptor agonists. In
addition, food consumption, motor task
performance, working memory, and
cognitive performance all appeared nor-
mal compare with control mice. More-
over, in tests of emotionality, stress,
sociability, and sensory motor gating,
NR1DATCre mice performed equally as
well as control mice, demonstrating that
most DA-dependent tasks do not de-
pend on burst firing and that tonic re-
lease of DA is sufficient to sustain
normal activity.
However, in a series of tests that de-
pend on contextual learning the
NR1DATCre mice displayed deficient per-
Author contributions: M.G.C. and R.M.W. wrote the paper.
The authors declare no conflict of interest.
See companion article on page 7281.
1Towhom correspondence should be addressed. E-mail:
caron002@mc.duke.edu.
formances. Similar to what had been
observed for conditioned place prefer-
ence to cocaine (8), place preference for
a natural reward such as food was im-
paired. In a Morris water maze test
where identifying the location of an es-
cape platform depends on visual cue-
dependent learning, NR1DATCre mice
also showed a significantly longer la-
tency to locate the platform. A subtle
but interesting difference was identified
when the mice were subjected to a T-
maze task baited with food pellets. The
KO mice showed a much slower rate of
making the correct arm choice but after
many trials learned the task if the loca-
tion of the bait remained the same.
However, if the location was changed in
a pseudo-random way, the KO mice
never managed to learn the task. This
observation suggests that, although the
KO mice might be able to learn from
habit, they show impairment in using
predictive cues for learning.
Interestingly, the cue-dependent
learning deficits of the KO mice were
also apparent when the cue predicted an
aversive or fearful event. This finding
was not readily predicted from previous
investigations. Many aversive stimuli
phasically activate some DA neurons but
the majority of these neurons seem to
be inhibited by such stimuli (11). Indeed
in a fear-conditioning paradigm where
mice are exposed to a cue that predicts
1. Haber SN, Fudge JL, McFarland NR (2000) Striatonigros-
triatal pathways in primates form an ascending spiral
from the shell to the dorsolateral striatum. J Neurosci
20:2369 –2382.
2. Grace AA (1991) Phasic versus tonic dopamine release
and the modulation of dopamine system responsivity:
A hypothesis for the etiology of schizophrenia. Neuro-
science 41:1–24.
3. Schultz W (2007) Behavioral dopamine signals. Trends
Neurosci 30:203–210.
4. Sombers LA, Beyene M, Carelli RM, Wightman RM
(2009) Synaptic overflow of dopamine in the nucleus
accumbens arises from neuronal activity in the ventral
tegmental area. J Neurosci 29:1735–1742.
7268 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0903306106
an unpleasant stimulus, NR1DATCre mice
fail to develop elevated startle responses
to an acoustic challenge as control mice
are able to do.
Finally, appropriate performance of a
task requires at least 2 components,
learning the task and motivation to per-
form the task. Although the KO mice
appeared to show longer latencies in
many of the learning paradigms, once a
task was learned they performed nearly
as well as the control mice. This latency
could not be attributed to a lack of mo-
tivation because, as revealed in a pro-
gressive ratio lever pressing for food
paradigm, the KO mice worked as hard
as their controls for the food reward.
These findings suggest that in these ani-
mals it is the acquisition of learning that
is impaired, consistent with a role for
phasic DA release in providing the in-
centive value for the directive cues. This
interpretation is also consistent with the
previous conclusion advanced from ex-
periments performed with hyper- and
hypodopaminergic mice where changes
in tonic DA release were found to more
significantly affect the motivation to
perform a task than the performance of
the task itself.
This study provides an elegant com-
plement to previous studies that had
suggested such a role for burst firing of
DA neurons in coding for the salience
of cues based on pharmacological ap-
5. Zweifel LS, et al. (2009) Disruption of NMDAR-
dependent burst firing by dopamine neurons pro-
vides selective assessment of phasic dopamine-
dependent behavior. Proc Natl Acad Sci USA
106:7281–7288.
6. Cagniard B, et al. (2006) Dopamine scales performance
in the absence of new learning. Neuron 51:541–547.
7. Hnasko TS, Sotak BN, Palmiter RD (2007) Cocaine-
conditioned place preference by dopamine-deficient
mice is mediated by serotonin. J Neurosci 27:12484 –
12488.
8. Zweifel LS, Argilli E, Bonci A, Palmiter RD (2008) Role of
NMDA receptors in dopamine neurons for plasticity
and addictive behaviors. Neuron 59:486 – 496.
proaches (4). The precision of the con-
clusions of this study highlights the
power of the cell-specific genetic ap-
proach used here. However, one should
be mindful that the NR1DATCre mice
used in these studies actually represent a
‘‘genetic sensitized animal’’ because the
mice carried single copies of both Grin1
(NR1) and Slc6a3 (DAT) genes. Thus, it
is possible that the observed behavioral
effects of decreased burst firing could
vary in magnitude in normal animals.
In future experiments it would of in-
terest to explore the proposition that
phasic DA release mainly activates exci-
tatory DA D1 receptors on medium
spiny neurons of the direct pathway,
whereas tonic DA engages inhibitory
DA D2 receptors of the indirect path-
way. Selective genetic inactivation of D1
and D2 receptor genes in D1 and D2
receptor expressing medium spiny neu-
rons coupled with deletion of the NR1
gene in DA neurons would provide a
suitable approach to directly validate
this contention. In vivo voltametric ex-
periments indicate that this should be
the case because release evoked by pha-
sic stimulations of dopaminergic neurons
is much larger than that evoked by tonic
stimulations (12).
ACKNOWLEDGMENTS. The work of M.G.C. and
R.M.W. is supported by National Institutes of
Health Research Grants R01-MH-073853, U01-DA-
022950 (to M.G.C.), and R01-DA 10900 (to
R.M.W.).
9. Engblom D, et al. (2008) Glutamate receptors on dopa-
mine neurons control the persistence of cocaine seek-
ing. Neuron 59:497–508.
10. Overton P, Clark D (1992) Iontophoretically adminis-
tered drugs acting at the N-methyl-D-aspartate recep-
tor modulate burst firing in A9 dopamine neurons in
the rat. Synapse 10:131–140.
11. Ungless MA, Magill PJ, Bolam JP (2004) Uniform
inhibition of dopamine neurons in the ventral teg-
mental area by aversive stimuli. Science 303:2040 –
2042.
12. Wightman RM, Robinson DL (2002) Transient changes
in mesolimbic dopamine and their association with
‘reward.’ J Neurochem 82:721–735.
Caron and Wightman