The entorhinal cortex modulates trace fear memory formation and neuroplasticity in the lateral amygdala via cholecystokinin

Although the neural circuitry underlying fear memory formation is important in fear-related mental disorders, it is incompletely understood. Here, we utilized trace fear conditioning to study the formation of trace fear memory. We identified the entorhinal cortex (EC) as a critical component of sensory signaling to the amygdala. Moreover, we used the loss of function and rescue experiments to demonstrate that release of the neuropeptide cholecystokinin (CCK) from the EC is required for trace fear memory formation. We discovered that CCK-positive neurons extend from the EC to the lateral nuclei of the amygdala (LA), and inhibition of CCK-dependent signaling in the EC prevented long-term potentiation of sensory signals to the LA and formation of trace fear memory. Altogether, we suggest a model where sensory stimuli trigger the release of CCK from EC neurons, which potentiates sensory signals to the LA, ultimately influencing neural plasticity and trace fear memory formation.


25
Learning to associate environmental cues with subsequent adverse events is an important 26 survival skill. Pavlovian fear conditioning is widely used to study this association and is 27 performed by pairing a neutral stimulus (conditioned stimulus, CS), such as a tone, with a 28 punishing stimulus (unconditioned stimulus, US), such as a shock (Pavlov, 1927). The CS-US 29 pair elicits fear behaviors, including freezing and fleeing, which are often species-specific. 30 Canonical delay fear conditioning is performed by terminating the CS and US at the same time. 31 However, conditioned and unconditioned stimuli do not necessarily occur simultaneously in 32 nature, and the brain has evolved mechanisms to associate temporally distinct events. Trace 33 fear conditioning is used to study these mechanisms by inserting a trace interval between the 34 end of the CS and the beginning of the US. The temporal separation between the CS and the 35 US substantially increases the difficulty of learning as well as the recruitment of brain 36 structures (Crestani et al., 2002;Runyan et al., 2004). Although trace fear conditioning 37 provides essential insight into the neurobiology of learning and memory, many unanswered 38 questions remain. For instance, the detailed neural circuitry underlying the formation of this 39 trace fear memory and the potential modulatory chemicals involved in this process need to be 40 further characterized. 41 Synaptic plasticity is the basis of learning and memory and refers to the ability of neural 42 connections to become stronger or weaker. Long-term potentiation (LTP) is one of the most 43 widely-studied forms of synaptic plasticity. The lateral nucleus of the amygdala (LA) receives 44 multi-modal sensory inputs from the cortex and thalamus and relays them into the central 45 nucleus of the amygdala (CeA), which then innervates the downstream effector structures 46 (Phelps & LeDoux, 2005). LTP is developed in the auditory input pathway that signals to the 47 LA. Auditory-responsive units in the LA fire faster after auditory-cued fear conditioning (Quirk 48 et al., 1995). Optogenetic manipulation of the auditory input terminals in the LA leads to the 49 suppression or recovery of LTP in the LA and can correspondingly suppress or recover 50 conditioned fear responses (Nabavi et al., 2014). Together, these studies demonstrate that 51 synaptic plasticity in the LA is impressively correlated with the formation of fear memory. 52 In addition to the amygdala, other neural regions, including the hippocampus (Bangasser, 2006), 53 anterior cingulate cortex (ACC) (Han et al., 2003), medial prefrontal cortex (mPFC) (Runyan 54 et al., 2004), and entorhinal cortex (EC) (Ryou et al., 2001), take part in trace fear conditioning. 55 The EC is integrated in the spatial and navigation systems of the animal (Fyhn et al., 2004;56 Hafting et al., 2005) and is essential for context-related fear associative memory (Maren & 57 Fanselow, 1997). Moreover, the EC functions as a working memory buffer in the brain to hold 58 information for temporal associations (Fransén, 2005;Schon et al., 2016). Here, a scenario of 59 the dependence on the EC to associate the temporally-separated CS and US is manifested. 60 The neuropeptide cholecystokinin (CCK) is universally accepted as the most abundant 61 neuropeptide in the central nervous system (CNS) (Rehfeld, 1978). CCK is recognized by two 62 receptors in the CNS: CCK A receptor (CCKAR) and CCK B receptor (CCKBR). Previous 63 studies in our laboratory unveiled that CCK and CCKBR enabled neuroplasticity as well as 64 associative memory between two sound stimuli and between visual and auditory stimuli in the 65 auditory cortex (X. Chen  attacks in individuals with a panic disorder as well as in healthy human subjects (Bradwejn, 70 1993). Despite the clear connection between CCK and fear-related disorders, it remains elusive 71 that the involvement of CCK in Pavlovian fear conditioning and the formation of cue-specific 72 fear memory, which is possibly the neural foundation of these disorders. 73 In the present study, we investigated the involvement of CCK-expressing neurons in the EC in 74 trace fear memory formation. We then examined how CCK enabled neuroplasticity in the 75 auditory pathway to the LA by conducting the in vivo recording in the LA. Finally, we studied 76 the contribution of the pathway from the EC to LA on the formation of trace fear memory in a 77 physiological and behavioral context. 78

Loss of CCK results in deficient trace fear memory formation in CCK -/mice 80
The first question we asked here was whether CCK is involved in trace fear memory formation. 81 We studied transgenic CCK -/mice (Cck-CreER, strain #012710, Jackson Laboratory), which 82 lack CCK expression (X. Chen et al., 2019). We subjected CCK -/and wildtype control (WT, 83 C57BL/6) mice to trace fear conditioning using two training protocols: long trace interval and 84 short trace interval training. 85 Trace fear conditioning was performed by collecting baseline readouts on pre-conditioning day, 86 training with the appropriate CS-US pairings on conditioning days, and testing the conditioned 87 fear responses on post-conditioning/testing day. In the long trace protocol, mice sequentially 88 received a 10-second pure tone (as the CS), a 20-second gap (trace interval), and a 0.5-second 89 foot shock (as the US) ( Figure 1a). We calculated the percentage of time frames where mice 90 displayed a freezing response as the measure of fear memory. Freezing percentages were 91 compared before (baseline) and after (post-training) trace fear conditioning as well as before 92 (    (a) Schematic diagram of the fear conditioning paradigm with a long trace interval of 20 s. Gray and light blue shadowed areas indicate the time frames before and after the onset of the CS (Before CS, After CS). CS, conditioned stimulus; US, unconditioned stimulus. (b-c) Freezing percentages before (b) and after (c) the CS. Freezing percentages were recorded at baseline on the pre-conditioning day and post-training on the post-conditioning day. WT, wildtype, N = 14; CCK -/-, CCK-knockout, N = 10. *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant. Statistical significance was determined by two-way RM ANOVA with Bonferroni post-hoc pairwise test. RM ANOVA, repeated measures analysis of variance. (d) Freezing score plot of the two groups of mice during the testing session. Solid lines indicate the mean value, and shadowed areas indicate the SEM. The black bar indicates the presence of the CS from 0 s to 10 s. *P < 0.05; two-sample t-test; SEM, standard error of the mean. (e) Schematic diagram of the fear conditioning paradigm with a short trace interval of 2 s. (f-g) Freezing percentages before (f) and after (g) the CS. WT, N = 11; CCK -/-, N = 14.
(h) Freezing score plot of the two groups of mice during the testing session. The black bar indicates the presence of the CS from 0 s to 3 s. *P < 0.05; two-sample t-test.
showed significantly lower freezing percentages (32.5% ± 6.2%) than WT mice after receiving 100 the CS (61.6% ± 4.6%, pairwise comparison, P < 0.01), indicating poor performance in 101 associating the CS with the US (Figure 1c, Movie S1, S2). This effect was not due to elevated 102 basal freezing levels caused by training in WT animals ( Figure 1b). Instead, we found that 103 CCK -/mice (20.7% ± 3.0%) had slightly higher freezing percentages than WT mice (14.0% ± 104 1.7%) in the absence of the CS (pairwise comparison, P > 0.05). Together, these results suggest 105 that trace fear conditioning results in elevated conditioned freezing percentages in WT mice, 106 which are primarily elicited by the CS, and that loss of CCK impairs the freezing response to 107 the CS. Furthermore, we defined an empirical threshold of moving velocity and converted the 108 moving velocity to a binary freezing score plot, in which value 1 represents active status, and 109 value 0 represents freezing status (see Methods). Using this method, we were able to assess the 110 freezing response of the animal as it occurred during the CS presentation. Again, we found that 111 WT mice obtained higher average freezing scores than CCK -/mice during the presentation of 112 the CS (Figure 1d, *P < 0.05, two-sample t-test). 113 In addition to the long trace interval, we also investigated freezing responses of mice during a 114 short trace fear conditioning paradigm. Mice were presented a 3-second CS followed by a 2-115 second trace interval and a 0.5-second electrical foot shock (Figure 1e). Before training, WT 116 (N = 11) and CCK -/-(N = 14) mice showed similarly low freezing percentages both before 117 ( Figure 1f) and after ( Figure 1g) presentation of the CS (Figure 1g, Figure 1g, Movie S3, S4). Additionally, we observed no significant difference between 125 fear conditioned WT and CCK -/mice prior to the presentation of the CS (Figure 1f, pairwise  126 comparison, WT vs. CCK -/in the post-training session, 12.4% ± 1.4% vs. 16.0% ± 2.4%, P > 127 0.05). Finally, we found significant differences in freezing scores between WT and CCK -/mice 128 when presented the CS (Figure 1h, *P < 0.05, two-sample t-test). 129 We conducted the innate hearing and fear expression examinations to rule out a potential 130 inherent deficit derived from genome editing in CCK -/transgenic mice. To evaluate hearing, 131 we recorded the open-field auditory brainstem response (ABR) in anesthetized animals. We 132 observed five peaks in both WT and CCK -/mice at sound intensities above 50 dB of sound 133 pressure level (dB SPL) (Figure S1b), and we did not observe any remarkable differences 134 between the waveforms. Compared to WT mice, CCK -/mice had better hearing (40.0 ± 1.2 dB 135 in CCK -/mice, N = 15, vs. 47.3 ± 2.1 dB in WT mice, N = 11, two-sample t-test, P < 0.01, 136 Figure S1c). Thus, auditory perception does not account for the deficient trace fear memory 137 formation of CCK -/mice. 138 As fear expression is the behavioral output of fear conditioning, we wondered if CCK -/mice 139 suffered from a deficit in fear expression, which is observed in Klüver-Bucy syndrome and 140 other diseases (Lilly et al., 1983). To test whether the CCK -/mice have a deficit in fear 141 expression, we presented a loud (90 dB SPL) white noise and quantified sound-driven innate 142 freezing. We found no statistical difference between WT (46.1% ± 5.5%, N = 11) and CCK -/-143 mice (46.5% ± 6.6%, N = 14, two-sample t-Test, P > 0.05, Figure S1d), indicating that CCK -/-144 mice can express passive defensive behaviors such as freezing. Thus, the deficient trace fear 145 memory formation of CCK -/is not due to a deficit in fear expression and may be due to a 146 deficit in establishing an association between the CS and the US. 147 In summary, CCK -/mice display deficient trace fear memory formations in both short and long 148 trace models that are not caused by inherent abnormalities in hearing or fear expression.   (g) Time course plot of the normalized AEP slope during LTP. The WT group is indicated in black, and the CCK -/group is indicated in red. Representative traces of the AEP before (dotted line) and after (solid line) TBS are shown in inset panels for both groups. The average normalized slopes 10 min before pairing (-10-0 min, before) and 10 min after pairing (50-60 min, after) in the two groups of mice are shown on the right. ***P < 0.05; two-way RM ANOVA with Bonferroni post-hoc pairwise test; RM ANOVA, repeated measures analysis of variance; NS, not significant. comparison, after vs. before induction, 142.7% ± 17.5% vs. 99.1% ± 2.8%, P = 0.011 < 0.05), 159 whereas CCK -/mice showed no potentiation (pairwise comparison, after vs. before induction, 160 98.0% ± 5.8% vs. 100.6% ± 3.4%, P > 0.05). These results suggest that CCK -/mice have a 161 deficit in neural plasticity in the LA that may contribute to their reduced response to trace fear 162 conditioning. 163

Stimulation of CCKBR rescues the formation of trace fear memory in CCK -/mice 164
Although the translation and release of CCK are disrupted in CCK -/mice, we found that the 165 predominant CCK receptor, CCKBR, was expressed normally in both WT and CCK -/mice 166 ( Figure 2h). Therefore, we hypothesized that exogenous stimulation of CCKBR might rescue 167 trace fear memory deficits in CCK -/mice. CCKBR can be stimulated by several agonists, 168 including CCK octapeptide sulfated (CCK-8s) and CCK tetrapeptide (CCK-4). As CCK-8s is 169 a potent agonist of both CCKAR and CCKBR, we selected CCK-4, which is a preferred 170 CCKBR agonist (Berna et al., 2007). To monitor CCK signaling in vivo, we expressed a G CCKBR with endogenous or exogenous CCK results in increased fluorescence intensity, which 174 we measured by fiber photometry in the LA ( Figure S2a). We first confirmed that 175 intraperitoneal (i.p.) administration of CCK-4 permeated the blood-brain-barrier (BBB) and 176 activated the CCK2.0 sensor. Moreover, we demonstrated that administration of CCK-4 177 evoked a clear and long-term increase in the fluorescent signal ( Figure S2b). Together, these 178 data verify that CCK-4 passes through the BBB and binds with CCKBR in the LA. 179 After validating our model, we conducted short trace fear conditioning in CCK -/mice on two 180 consecutive days just after intraperitoneal administration of CCK-4 or the corresponding 181 vehicle (VEH) (Figure 2i-j). We collected data during the two conditioning days to monitor 182 the learning curve of mice as conditioning progressed. The learning curves were plotted as the 183 freezing percentages of CCK-4-or VEH-treated CCK -/mice during the six training trials 184 (Figure 2k-l). During the first three trials on the first conditioning day and even in the fourth 185 trial on the second conditioning day, we did not observe any statistical differences between the 186 two groups. During the fifth and sixth training trials conducted on the second conditioning day, 187 we found that CCK-4-treated mice had significantly higher freezing levels than VEH-treated 188 mice (Figure 2l, 84.2% ± 8.4% in the CCK-4 group [N = 6] vs. 48.4% ± 11.5% in the VEH 189 group [N = 7] in the fifth trial; 84.4% ± 7.3% in the CCK-4 group vs. 52.9% ± 13.0% in the 190 VEH group in the sixth trial, two-sample t-test, both P < 0.05). In support of this evidence, we 191 did not find a statistical difference between the two groups prior to CS presentation during the 192 fifth or sixth trials (Figure 2k, 53.8% ± 11.5% in the CCK-4 group vs. 52.5% ± 11.8% in the 193 VEH group in the fifth trial; 56.0% ± 10.8% in the CCK-4 group vs. 47.8% ± 11.8% in the 194 VEH group in the sixth trial, two-sample t-test, both P > 0.05). Together, these data suggest 195 that mice in the CCK-4-and VEH-treated groups showed similar baseline freezing levels and 196 that CCK-4 treatment improved trace fear conditioning learning responses in CCK -/mice. 197 We went on to assess the conditioned fear response in CCK-4-and VEH-treated CCK -/mice 198 two days after training in comparison to fear responses at baseline prior to training (Figure 2m-199 n). We found that CCK4-treated mice showed remarkably higher freezing levels than VEH-200 treated mice post-training, whereas no significant difference was detected at baseline ( Figure  201 2n, two-way RM ANOVA, significant interaction, F [1,11] = 6.40, P = 0.028 < 0.05; pairwise 202 comparison, CCK-4 vs. VEH at baseline, 10.4% ± 2.8% vs. 7.0% ± 1.4%, P > 0.05; CCK-4 vs. 203 VEH post-training, 54.3% ± 5.9% vs. 30.4% ± 3.3%, P < 0.05; Movie S5, S6). There was no 204 statistical difference between the two groups before the presentation of the CS (Figure 2m,  205 two-way RM ANOVA, the main effect of drug application [CCK-4 vs. VEH] on freezing 206 percentage was not significant, F [1,11] = 0.15, P = 0.70). Additionally, CCK-4-treated mice 207 had significantly higher freezing scores than VEH-treated mice ( Figure 2o). These results 208 indicate that CCK-4 treatment effectively improved learning response to trace fear conditioning 209 in CCK -/mice. Moreover, this rescue was not an artifact caused by reduced locomotion after 210 drug application and fear conditioning training, as there was no difference between the two 211 groups in the freezing percentage prior to presentation of the CS (Figure 2m). Therefore, the 212 exogenous application of a CCKBR agonist activated endogenous CCKBR and improved the 213 fear memory formation of CCK -/mice after trace fear conditioning. 214 CCK neurons in the EC are critical for the formation of the trace fear memory 215 We next examined the source of endogenous CCK that signals to the LA. We injected a potent 216 retrograde neuronal tracer Cholera Toxin Subunit B (CTB) conjugated to a fluorescent tag 217 Alexa-647 (CTB-647) into the LA and dissected the upstream anatomical brain regions that 218 project to the LA (Figure 3a). In addition to regions that are canonically involved in fear 219 circuitry, including the auditory cortex (AC) and the medial geniculate body (MGB), we found 220 that EC was also densely labeled with retrograde CTB-647, suggesting that the EC is connected 221 with the LA (Figure 3b-e). We next injected a Cre-dependent retrograde AAV (retroAAV-222 hSyn-FLEX-jGcamp7s) into the LA of CCK-ires-Cre (CCK-Cre) mice to label CCK-positive 223 Activation of hM4Di by CLZ induces membrane hyperpolarization, effectively silencing 240 neurons. We verified EC neuron silencing by in vivo electrophysiological recording ( Figure  241 4b-d and Figure S3). We found that a low dose of CLZ (0.  (k-l) Freezing percentages before (k) and after (l) the CS during the testing session in mice treated with CLZ or vehicle (VEH). *P < 0.05; NS, not significant; two-way RM ANOVA with Bonferroni post-hoc pairwise test. (m) Freezing score plot of CLZ-and VEH-treated mice during the testing session. The black bar indicates the presence of the CS from 0 s to 10 s. *P < 0.05; two-sample t-test; SEM, standard error of the mean.
9.4% ± 1.4%, P > 0.05; CLZ vs. VEH post-training, 18.0% ± 3.2% vs. 18.3% ± 3.4%, P > 0.05; 272 Movie S9, S10). These results mirror those observed in CCK -/mice and suggest that trace fear 273 memory formation relies on intact and functional CCK-positive neurons in the EC. were densely labeled with EGFP, whereas mCherry labeling of CCK-projections was 284 dramatically weaker. Quantitative analysis revealed that the projection intensity of the 285 EC CCK+ →LA was 3-fold higher than the EC CCK-→LA (35.6% ± 9.5%). In other words, CCK-286 positive afferents constituted approximately 75% of total afferents from the EC to the LA 287 ( Figure 5e-f). 288  (e-f) Visualization (e) and quantification (f) of EGFP-expressing (CCK+) and mCherryexpressing (CCK-) afferents in the amygdala stemming from the EC. The fluorescent intensity of neuronal projections was normalized to the EGFP+ signal, which was approximately 3-fold stronger than the mCherry+ signal (35.6% ± 9.5%). (g-h) Schematic diagram of Cre-dependent color-switch labeling in the EC-LA pathway. A mixture of AAV-hSyn-DIO-mCherry and AAV-EF1α-FAS-EGFP was injected in the EC. Using this labeling scheme, mCherry is expressed in CCK+ neurons, and EGFP is expressed in CCK-neurons.  Figure  295 5i-j). The higher percentage of double-positive neurons present in this system indicates a 296 higher probability of off-target effects compared to the previous color-switching AAV (8.9% 297 ± 2.7% vs. 2.5% ± 1.1%). Consistent with the previous color-switching AAV, we observed that 298 CCK+ (mCherry+) projections were predominant. Specifically, the intensity of the 299 EC CCK+ →LA was approximately 4-fold higher than the EC CCK-→LA (24.0% ± 5.6%). 300 Altogether, our results suggest that the EC CCK+ →LA is the predominant subpopulation of 301 projections, and that these projections are of functional significance in the EC-LA pathway. 302 CCK-positive neural projections from the EC to the LA enable neural plasticity and 303 modulate trace fear memory formation 304 Finally, we asked whether CCK-positive projections from the EC modulate neural plasticity in 305 the LA. First, we expressed a Cre-dependent high frequency-responsive channelrhodopsin 306 (ChR2) variant E123T (ChETA) under control of the universal EF1α promoter in CCK-Cre 307 mice ( Figure 6a). Then, we implanted optic fibers targeting the LA to illuminate EC CCK+ →LA 308 projections and electrodes to conduct in vivo electrophysiological recording as before ( Figure  309 6b). Post-hoc anatomical analysis confirmed the distribution of ChETA in the EC-LA axon 310 terminals (Figure 6c). These CCK+ projections were innervated with postsynaptic CCKBR 311 (Figure 6d), suggesting that CCK+ projections from the EC effectively activated CCKBR in 312 the LA. Finally, we recorded auditory evoked potential (AEP) and visual evoked potential 313 (VEP) in the LA of anesthetized mice (Figure 6e-g). Although AEP and VEP had similar 314 waveforms, the latency of AEP was much shorter than VEP (Figure 6e-f, peak latency: 38.9 ± 315 3.2 ms for AEP, N = 13, vs. 89.5 ± 3.1 ms for VEP, N = 11, two-sample t-test, P < 0.001). This 316 observation implies that input pathways other than the canonical thalamo-cortico-amygdala 317 and thalamo-amygdala projections regulate the transmission of visual cues. We applied high-318 frequency-laser-stimulation (HFLS, Figure 6h) of the EC-LA axons before the auditory 319 stimulus (AS) to trigger AEP-LTP in the LA. After induction, the AEP slope in the ChETA-320 expressing group (n = 10) increased significantly, whereas the VEP slope did not change 321 (Figure 6i- we injected a non-opsin expressing control AAV (AAV-EF1α-DIO-EYFP, n = 20) and the 325   In the next experiment, we examined the possibility of other neuroactive molecules that are co-330 released with CCK and contribute to HFLS-induced AEP-LTP. We adopted an RNA 331 interference technique that specifically knockdown the CCK expression in the EC. We 332 accomplished this by injecting a Cre-dependent AAV cassette carrying a ChR2 variant 333 (E123T/T159C) and a short hairpin RNA (shRNA) targeting CCK (anti-CCK) or a nonsense 334 sequence (anti-Scramble) into the EC of CCK-Cre mice (Figure 7a-c). The inclusion of laser-335 responsive ChR2 allowed us to induce the above AEP-LTP by specifically stimulating the 336 EC CCK+ →LA pathway. We applied our HFLS pairing protocol in these mice and found that 337 AEP-LTP could not be induced in the anti-CCK group but could successfully induced in the 338 anti-Scramble group (Figure 7d- To dissect the real-time behavioral dependency of trace fear memory formation on the 344 EC CCK+ →LA pathway, we employed optogenetics. We expressed the inhibitory opsin 345 eNpHR3.0 (AAV-EF1α-DIO-eNpHR3.0-mCherry) or GFP control (AAV-hSyn-FLEX-GFP) 346 in the EC of CCK-Cre mice. We also implanted optic fibers targeting the LA in these mice and 347 then subjected the mice to trace fear conditioning (Figure 8a-b). During trace fear conditioning, 348 EC CCK+ →LA were stimulated at a frequency of 5 Hz (i.e., 100 ms illumination + 100 ms 349 interval) by the optic fibers for the duration of the CS and trace interval, as indicated in Figure  350 8a. For these experiments, mice were positioned in a head-fixed setup on a moveable surface, 351  and an electrical tail shock was given as the US. After administration of the US, we most 352 commonly observed flight (running). Interestingly, we found that after a few training trials, 353 some GFP control mice (3/6 animals, data not shown) began running before the US was given, 354 suggesting that GFP mice associate the CS with the US and make predictions in subsequent 355 training trials (Movie S11). In contrast, we observe much fewer conditioned defensive 356 responses in the eNpHR group throughout the training process (1/8 animals and 2/40 observed 357 training trials, data not shown, Movie S12). Additionally, we recorded the freezing percentages 358 in response to the CS before and after head-fixed fear conditioning (Figure 8c-d). We found 359 that mice in the eNpHR group showed impaired freezing percentages post-training compared 360 to mice in the GFP group (Figure 8d, two-way RM ANOVA, significant interaction, F [1,12] 361 = 19.20, P < 0.001; pairwise comparison, GFP vs. eNpHR post-training, 39.0% ± 2.0% vs. 362 12.2% ± 4.8%, P < 0.001; Movie S13, S14). We did not observe any differences between the 363 two groups at baseline (Figure 8d, pairwise comparison, GFP vs. eNpHR at baseline, 12.7% ± 364 3.4% vs. 12.2% ± 4.8%, P > 0.05) or prior to the CS (Figure 8c,   . In our study, we used in vivo recording to measure auditory-evoked 400 field excitatory postsynaptic potential (fEPSP) or AEP. We did not find any apparent 401 abnormalities in AEP (such as amplitude or latency) in CCK -/mice, suggesting that cortical 402 and thalamic auditory inputs to the LA were functional. CCK -/mice did fail to induce AEP-403 LTP in the LA, strongly suggesting a deficiency in neural plasticity. However, we cannot 404 simply assume that AEP-LTP induction is equivalent to trace fear memory. Occasionally, AEP-405 LTP is not sufficient to trigger the expression of fear behaviors. Kim and Cho reported that 406 LTP in the LA was maintained during fear extinction (Kim & Cho, 2017) . Thus, LTP in the 407 LA is necessary but not sufficient for fear memory formation. 408 As the EC has been previously implicated in trace fear memory and behaviors, we manipulated 409 EC function in our present study and investigated the behavioral and signaling outcomes. We 410 found that silencing EC neurons with DREADD hM4Di impaired the formation of trace fear 411 memory, which is consistent with several previous studies. For instance, electrolytic lesion of 412 the EC impairs trace eyeblink conditioning performance in mice (Ryou et al., 2001), and 413 neurotoxic lesions as well as M1 receptor blockade in the EC impair trace fear memory 414 formation but not delay fear memory formation (Esclassan et al., 2009). Although this 415 preferential association with trace fear memory has also been observed in certain areas of the 416 hippocampus (Bangasser, 2006), the EC is a promising regulatory unit, because EC neurons 417 maintain persistent spikes activity in response to stimuli (Egorov et al., 2002;Fransén et al., 418 2006). This sustained neuronal activity is thought to be the neural basis of 'holding' CS 419 information during trace intervals to allow for CS-US association even after long trace intervals 420 (20 seconds in our study). This information 'holding' theory is consistent with neuroimaging 421 reports on working memory in subjects who 'hold' stimuli for specific periods (Nauer et  the CS-US association in a more complicated manner, and this mechanism requires further 431 investigation. We speculate that this mechanism is probably similarly as our previous finding 432 in the sound-sound association (X. Chen et al., 2019) and visuo-auditory association (Z. Zhang 433 et al., 2020), which is neuropeptide-based hetero-synaptic modulation machinery. 434 With cell type-specific tracing systems, we demonstrated that the EC is an upstream brain 435 region that projects CCK-positive afferents to the LA, and these CCK-expressing EC neurons 436 are primarily excitatory (Figure 3). Using anterograde Cre-dependent color switch labeling in 437 the EC, we also found that CCK-expressing neurons were the predominant source of EC-LA 438 projections, implying that CCK is integral to EC-LA connection and communication. Cell type-439 specific chemogenetic inhibition of CCK-expressing neurons in the EC also impaired the 440 formation of trace fear memory. However, we cannot exclude the possibility that CCK may 441 originate in other brain regions and contribute to fear memory formation. 442 We triggered the release of CCK from axon terminals after in vivo HFLS of CCK-expressing 443 fibers in the LA (Hökfelt, 1991). In the presence of this artificially released CCK neuropeptide, 444 we then presented the AS. The AS activates presynaptic axons via the canonical LA fear circuit, 445 which is supported by the known role of the LA in receiving auditory input from both the 446 auditory cortex and the thalamus (Romanski & LeDoux, 1992). In our study, the AS triggered 447 postsynaptic neural firing. Therefore, our HFLS-mediated AEP-LTP induction protocol 448 combines the released CCK with pre-and postsynaptic activation altogether in the LA and this 449 pairing leads to the potentiation of AEP in the LA. 450 In the current study, we successfully excluded the contribution of substances co-released with 451 CCK to the induction of AEP-LTP by applying the in vivo RNA interference to knockdown 452 the expression of Cck in CCK-positive neurons of the EC. We found that knockdown of Cck 453 blocked the induction of AEP-LTP and our in vivo application of shRNA supports the clinical 454 use of shRNA to target mental disorders related to the CCK system. Our results that the 455 inhibition of CCK-positive EC afferents to the LA impaired trace fear memory formation 456 during both the learning and response phases suggest that establishing the CS-US association 457 during trace fear conditioning requires functional CCK-positive EC-LA projections. 458 In conclusion, we found that EC-LA projections modulate neuroplasticity in the LA and 459 therefore contribute to the formation of trace fear memory. The CCK terminals of the EC 460 neurons in the LA release CCK that enable hetero-synaptic neuroplasticity of the auditory 461 pathway to the LA. Our findings add a novel insight into the participation of the neuropeptide 462 CCK in the formation of the trace fear memory. As various mental disorders, including anxiety 463 (Davis, 1992)  in vivo recording, auditory stimuli were delivered via a close-field speaker placed 508 contralaterally to the recording side. The sound intensity that induced 50%-70% of the 509 maximum response was selected. Visual stimuli were generated by a direct current (DC)-driven 510 torch bulb via the analog voltage output of the TDT workstation. Light intensity was roughly 511 quantified as the value of the trigger voltage. For in vivo recording, the light intensity that 512 induced 50%-70% of the maximum response was selected. 513 Auditory brainstem response recording 514 Mice were anesthetized with pentobarbital sodium (80 mg/kg, i.p.) and placed on a clean and 515 warm blanket in a soundproof chamber. A free-field magnetic speaker (MF-1, TDT) was placed 516 10 cm away from the right ear of mice. Recording, reference and ground needle electrodes 517 (Spes Medica, Genova, Italy) were subcutaneously inserted below the forehead, right ear and 518 left ear, respectively. Auditory stimuli (wide spectrum clicks, 0.1 ms) were presented to the 519 mouse with a decreasing level from 80 dB to 20 dB with an interval of 5 dB. For each level of 520 click stimulus, total 512 times of presentation were given at a frequency of 21 Hz. ABR signals 521 were collected via a specialized processor (RZ6, TDT) and digitalized with a bandpass filter 522 from 100 Hz to 5 kHz. Stimuli generation and data processing was performed with software 523 BioSigRZ (TDT). 524

Trace fear conditioning 525
On pre-conditioning day, each mouse was placed into the testing context (acrylic box with 526 white wallpaper measuring 25 cm × 25 cm × 25 cm) for habituation and baseline recording. 527 After 3 min of habituation, a CS (2.7 kHz or 8.2 kHz pure tone, 70 dB SPL, 3 s for the short 528 trace paradigm and 10 s for the long trace paradigm) was given three times within 20 min. 529 On conditioning day, the mouse was placed into the fear conditioning context (acrylic box with 530 brown wallpaper measuring 18 cm wide × 18 cm long × 30 cm high and equipped with foot 531 shock stainless steel grid floor). After 3 min of habituation, a CS-US pairing was given. In the 532 short trace interval paradigm, an US (0.5 mA foot shock, 0.5 s) was given 2 s after a 3-s-long 533 CS. Three trials were given on each training day, and the interval between trials was 10-15 534 min. Totally two training days were given. The mouse was kept in the fear conditioning context 535 for a 10 min consolidation period after the last training trial. In the long trace interval paradigm, 536 an US was given 20 s after a 10-s-long CS. Eight training trials were given each training day, 537 and the interval between trials was 2-3 min. The mouse was kept in the fear conditioning 538 context for a 5 min consolidation period after the last training trial. After training, each animal 539 was kept in a temporary cage and returned to their home cage after all individuals finished 540 training. 541 On post-conditioning day (test day), the mouse was placed into the testing context. After 3 min 542 of habituation, a CS was presented to the animal twice with a 2 min-long interval between 543 stimuli. Two min after the last trial, the animal was transferred to a temporary cage and returned 544 to its home cage after all individuals in its cage finished testing. 545 All contexts were cleaned thoroughly with 75% ethanol after each individual session. All of 546 the above procedures were conducted in a soundproof chamber, and all videos (baseline, 547 training, and testing) were recorded with a webcam (Logitech C270) set in the ceiling of the 548 chamber. Videos were analyzed with a custom program based on an open-source platform 549 (Lopes et al., 2015) (https://bonsai-rx.org). Briefly, the centroid of the animal was extracted 550 from the videos. By comparing the coordinates of the centroid frame by frame, we then 551 calculated the distance moved between two frames. The instant velocity of the animal was 552 calculated by dividing this distance by the time span between two adjacent frames. The freezing 553 percentage was defined as the percentage of frames with an instant velocity lower than the 554 threshold of all frames in an observed time window. We compared the output of this program 555 to results observed by the naked eye. Finally, we selected 0.1 (pixel 2 /s) as the appropriate 556 moving threshold to define freezing. Freezing score was defined as the binary value (0 or 1) of 557 time frame with instant velocity higher (0, 'not freezing') or lower (1, 'freezing') than the 558 threshold. For freezing score plot shown in Figure 1, 2 and 4, freezing scores from all test 559 sessions were averaged per second for data visualization. 560 Electrophysiological recording in the LA and EC 561 Mice were subjected to the surgical procedures describe above. Tracheotomy was conducted 562 to facilitate breathing and to prevent asphyxia caused by tracheal secretions during the 563 experiment. Craniotomy was performed 1.0-2.0 mm posterior and 3.0-4.0 mm lateral to the 564 bregma to target the LA. Dura mater was partially opened using a metal hook made of a 29G 565 syringe needle. Tungsten recording electrodes (0.5-3.0 MΩ, FHC, Bowdoin, ME USA) were 566 slowly inserted into the LA (approximately 3.5 mm from the brain surface). For laser 567 stimulation experiments, another craniotomy was performed at the temporal lobe (1.0-2.0 mm 568 posterior to the bregma) to expose the lateral rhinal vein. One optic fiber (200 µm diameter, 569 0.22 NA, Thorlabs, Newton, NJ, USA) was inserted below the rhinal vein and forwarded till 570 1.0-1.5 mm from the surface. The angle of the optic fiber was approximately 75° from the 571 vertical reference. Responses were recorded and passed to a pre-amplifier (PZ5, TDT) and an 572 acquisition system (RZ5D, TDT). Signals were filtered for field potential or spikes with 573 respective bandwidth ranges of 10-500 Hz and 1-5000 Hz. All recordings were stored using 574 TDT software (OpenEx, TDT). The maximum sound intensity was defined as the intensity that 575 elicited a saturated AEP. The AEP baseline was recorded with 50% of the maximum sound 576 intensity at a 5 s intertrial interval (ITI) for 20 min. interparietal bones and exposed the caudal rhinal vein and the transverse sinus ( Figure S3). 584 Electrodes were inserted approximately 1 mm below the dura mater. 585 All field potential data were extracted and processed in the MATLAB program, and all single 586 unit data were extracted from the TDT data tank to the Offline Sorter (Plexon) for spike sorting. 587 Sorted data were forwarded to the Neuroexplorer (Plexon) for additional processing and 588 visualization. 589

Viral and tracer injection 627
Mice were subjected to the surgical procedures described above. For viral injection into the EC, 628 the following rostral parameters were used: Anterior-Posterior (AP) = 3.25 mm, Medial-Lateral 629 (ML) = 3.80 mm, Dorsal-Ventral (DV) = 3.60 mm from the surface, volume = 100 nL. 630 Similarly, the following caudal parameters were used: AP = 4.25 mm, ML = 3.60 mm, DV = 631 2.60 mm from surface, volume = 200 nL. For injection of tracer or virus into the LA, we used 632 the following parameters: AP = 1.70 mm, ML = 3.40 mm, DV = 3.70 mm from the surface, 633 volume = 200 nL. Craniotomy was performed after skull levelling and partial opening of the 634 dura mater using a syringe needle hook (29G). We used the Nanoliter2000 system (World 635 Precision Instruments [WPI], Sarasota County, FL, USA) for all infusions. Viral or tracer 636 infusions were slowly pumped into brain tissue trough a fine-tip glass pipette filled with silicon 637 oil at a speed of no more than 50 nL/min. After infusion, the pipette was left in the injection 638 site for an extra 5-10 min before slow withdrawal. After withdrawal of the pipette, the scalp 639 was sutured, and a local anesthetic was applied. The animal was returned to its home cage after 640 awaking. For axon stimulation (observation), the virus was expressed for at least 7 weeks, and 641 for cell body stimulation (observation), the virus was expressed for at least 4 weeks. For CTB 642 tracer labeling, we perfused animals after 7 days of viral expression. 643

Optic fiber implantation 644
Mice were subjected to the surgical procedures described above. Craniotomy was performed 645 bilaterally to target the LA using the coordinates described above. Optic fibers (optic cannulae) 646 were gently inserted into the LA (50-100 µm above the target area) and fixed with dental 647 cement (mega PRESS NV + JET X, megadental GmbH, Büdingen, Germany). For head 648 fixation, a long screw was fixed to the skull with dental cement at a 45° angle from the vertical 649 axis. 650