for the coretrieval, 9.1% for the control, and
8.2% for the no-reactivation groups (Fig. 2F).
Coactivity of specific subsets of neurons triggers the association of memories (12, 25). To
test whether ensemble coactivity during RCS is
critical for the CTA-AFC interaction, we used
optogenetic manipulation to suppress the AFC
ensemble activity in the basolateral amygdala
during RCS. We applied to c-fos–tetracycline
(tet)–controlled transactivator (t TA) transgenic
mice (26) a tet-OFF lentiviral vector encoding
enhanced archaerhodopsin-T 3.0 fused to enhanced yellow fluorescent protein (Arch T-EYFP)
(Fig. 3, A to C) (4). Arch T-EYFP labeled the AFC
training ensemble in a doxycycline (Dox)– and
AFC training–dependent manner (fig. S5). We
then examined, ex vivo and in vivo, whether optical stimulation suppressed the activity of the
neuronal ensemble that had been labeled during
the AFC training. In ex vivo patch-clamp recordings with yellow-light illumination, Arch T-EYFP-expressing, but not EYFP-expressing, cells elicited
a robust and sustained hyperpolarizing current
that effectively blocked action potentials evoked
by somatic current injections (Fig. 3, D and E,
and fig. S6, A and B). Spontaneous activity was
suppressed with continuous optical stimulation
for 5 min, with the firing properties subsequently
recovering within several minutes (fig. S6, C
and D). In vivo, optical silencing of the AFC
training ensemble (Arch T laser ON group) significantly reduced the percentage of c-Fos+ cells
and tone-induced freezing compared with those
in the Arch T laser OFF (Arch T OFF) and EYFP
laser ON groups (EYFP ON) (Fig. 3, F to H). The
magnitude of freezing correlated well with the
number of c-Fos+ cells (Fig. 3I).
We then applied tet-tagged optical silencing
to mice subjected to RCS (Fig. 3J). Arch T-EYFP–
or EYFP-tagged mice, whose basolateral amygdala
cells were labeled during AFC training, received
continuous light illumination during each of three
RCSs. Optical silencing during RCS had no effect on the retrieval of the original CTA and AFC
memories (Fig. 3, K and L). However, the percentage of saccharin-induced freezing during test
1 was significantly attenuated in the Arch T ON
group as compared with the other groups (Fig.
3M). In catFISH imaging analysis conducted
24 hours after RCS (Fig. 3J), the number of Arc–
or H1a–single-positive cells was similar in each
group. However, the overlapping ensemble ratio
in the Arch T ON group was significantly reduced
as compared with that in the other groups (Fig. 3,
N and O). Therefore, artificial disruption of the
AFC training ensemble activity during RCS led
to a decrease in the proportion of the overlapping
neuronal ensemble and impaired the saccharin-induced freezing behavior.
Our findings raise a critical question: Does the
neuronal activity of the overlapping ensemble
in the basolateral amygdala mediate the cross-
modal association that leads to saccharin-induced
freezing? We developed a c-fos t TA-tet tag (26)
and lentivirus-based genetic targeting system in
combination with a tamoxifen-inducible Cre-loxP
recombination system (27) that enabled us to
specifically target the overlapping ensemble
(Fig. 4A). In a first step, the AFC retrieval en-
semble was labeled with CreERT2 under OFF-Dox
conditions. In a second step, DIO-Arch T-EYFP un-
der the control of the enhanced synaptic activity–
responsive element (E-SARE) promoter (28) was
transcribed in the cells of the CTA retrieval en-
semble. Last, Arch T-EYFP protein was translated
specifically in those cells that were activated
twice, once during the AFC retrieval and again
during the CTA retrieval—that is, in the cells of
the overlapping ensemble.
The time-dependent change in the intracellular
localization of CreERT2 was examined to deter-
mine the appropriate interval between the AFC
and CTA tests (Fig. 4, B and C). Five hours be-
fore the AFC test under OFF-Dox conditions,
4-hydroxytamoxifen (4-OHT) was administered.
CreERT2 signals initially localized in the nucleus
are translocated to the cytoplasm (27). This sig-
nal was detected exclusively in the cytoplasm
24 hours after the AFC retrieval test. Thus, Cre
recombination hardly occurs at 24 hours or
later, indicating that the interval between the
AFC and CTA tests should be at least 24 hours.
Our immunohistochemical analysis revealed an
increase in the number of Arch T-EYFP–positive
cells in the coretrieval group, as compared with
the control group or the group that did not
receive test 2 [test 2(−)] (13.5 ± 1.5%, 8.0 ± 1.2%,
and 4.2 ± 1.0%, respectively) (Fig. 4, D to F).
Thus, the increase depended on the RCS and
the CTA retrieval at test 2. Moreover, the in-
crease was consistent with that observed for the
catFISH results (Fig. 2E), indicating that the
overlapping population of cells was specifically
labeled with Arch T-EYFP. We found no signif-
icant differences among groups for the number
of H2B-mCherry+ cells (Fig. 4F), which repre-
sent the AFC retrieval ensemble, thus showing
an equivalent efficiency in Cre-mediated recom-
bination. The number of Arch T+, H2B-mCherry+
(double-positive) cells in the coretrieval group
was also significantly higher than in the control
and test 2(−) groups (Fig. 4G).
Last, we investigated a causal relationship
between the neuronal activity of the overlapping
ensemble during RCS and the association be-
tween the CTA and AFC memories (Fig. 4, H
to K). Animals labeled with Arch T-EYFP in the
overlapping ensemble in the basolateral amyg-
dala were divided into four groups that received
the same treatments throughout the experiments,
except during test 3 (Fig. 4H). Two groups were
subjected to the CTA retrieval test with or without
optical silencing (CTA-paired or CTA-unpaired,
respectively). The other two groups were sub-
jected to the AFC retrieval test with or without
optical silencing (AFC-paired or AFC-unpaired,
respectively). Optical silencing of the overlapping
ensemble suppressed saccharin-induced freezing
in test 3 (Fig. 4I, CTA-paired group). This group
of animals showed saccharin-induced freezing
comparable with that of the control (CTA-unpaired)
group in test 2 and test 5 without optical illumi-
nation. Original memories for either CTA (Fig. 4J)
or AFC (Fig. 4K) were intact even under light
illumination in test 3. When compared with the
respective CTA- or AFC-unpaired group, the
CTA- and AFC-paired groups showed compara-
ble aversion indexes (Fig. 4J, test 2 and test 5,
and fig. S7E), tone-induced freezing (fig. S7, A,
B, D, and F), and saccharin-induced freezing
(fig. S7C, test 2 and test 5) under no light
Our study reveals that repetitive coretrieval
reorganizes two aversive memory traces to generate an intersectional neuronal ensemble, neurons shared by both prestored CTA and AFC
memories. The overlapping ensemble is responsible for memory association but is dispensable
for the retrieval of original memories. An increase
in overlapping ensembles has been observed
when memories are associated in a variety of
learning paradigms (4, 12–14, 24, 25). Moreover,
memories for events occurring close together in
time generate an interaction that accompanies
an overlap in the cell ensembles of each memory (13, 14). Memory trace is not always allocated at the primary active neurons and could
be potentially reallocated to the different subsets of neuronal ensembles (7, 9). The increase
in the overlap between CTA and AFC ensembles
suggests that the overlapping ensemble is newly
formed by the RCS. Thus, after the RCS, CTA
and AFC memory traces are reallocated and reorganized to form a memory linkage. Generating
an overlapping neuronal ensemble is a general
mechanism underlying a linkage between memories during both acquisition and retrieval. Our
finding of an overlapping ensemble that is responsible for memory association but dispensable
for retrieval of original memories may provide a
way to dissociate daily memories that relate in
some way to the circumstances of a trauma (29)
from a traumatic event. This could help prevent
flashback in individuals with posttraumatic
REFERENCES AND NOTES
1. D. O. Hebb, The Organization of Behavior: A Neuropsychological
Approach (John Wiley and Sons, 1949).
2. J. R. Anderson, G. H. Bower, Human Associative Memory (Wiley,
3. T. Rogerson et al., Nat. Rev. Neurosci. 15, 157–169
4. M. Nomoto et al., Nat. Commun. 7, 12319 (2016).
5. T. Takeuchi et al., Nature 537, 357–362 (2016).
6. X. Liu et al., Nature 484, 381–385 (2012).
7. S. A. Josselyn, S. Köhler, P. W. Frankland, Nat. Rev. Neurosci.
16, 521–534 (2015).
8. S. Tonegawa, X. Liu, S. Ramirez, R. Redondo, Neuron 87,
9. A. Holtmaat, P. Caroni, Nat. Neurosci. 19, 1553–1562
10. A. R. Garner et al., Science 335, 1513–1516
11. S. Ramirez et al., Science 341, 387–391 (2013).
12. N. Ohkawa et al., Cell Rep. 11, 261–269 (2015).
13. D. J. Cai et al., Nature 534, 115–118 (2016).
14. A. J. Rashid et al., Science 353, 383–387 (2016).
15. Y. Zhou et al., Nat. Neurosci. 12, 1438–1443
16. K. Hashikawa et al., J. Neurosci. 33, 4958–4963
17. T. Inui, C. Inui-Yamamoto, Y. Yoshioka, I. Ohzawa, T. Shimura,
Neurobiol. Learn. Mem. 106, 210–220 (2013).
18. J. E. LeDoux, Annu. Rev. Neurosci. 23, 155–184 (2000).
19. C. Herry et al., Nature 454, 600–606 (2008).
402 27 JANUARY 2017 • VOL 355 ISSUE 6323 sciencemag.org SCIENCE
RESEARCH | REPORTS