Heterosynaptic Plasticity of the Visuo-auditory Projection Requires Cholecystokinin 1 released from Entorhinal Cortex Afferents 2

33 The entorhinal cortex is involved in establishing enduring visuo-auditory associative memory in 34 the neocortex. Here we explored the mechanisms underlying this synaptic plasticity related to 35 projections from the visual and entorhinal cortices to the auditory cortex, using optogenetics of 36 dual pathways. High-frequency laser stimulation (HFLS) of the visuo-auditory projection did not 37 induce long-term potentiation (LTP). However, after pairing with sound stimulus, the visuo- 38 auditory inputs were potentiated following either infusion of cholecystokinin (CCK) or HFLS of 39 the entorhino-auditory CCK-expressing projection. Combining retrograde tracing and RNAscope 40 in situ hybridization, we show that CCK expression is higher in entorhinal cortex neurons 41 projecting to the auditory cortex than in those originating from the visual cortex. In the presence 42 of CCK, potentiation in the neocortex occurred when the pre-synaptic input arrived 200 ms 43 before post-synaptic firing, even after just five trials of pairing . Behaviorally, inhibition of CCK 44 signaling blocked the generation of associative memory. Our results indicate that neocortical 45 visuo-auditory association is formed through heterosynaptic plasticity, which depends on release 46 of CCK in the neocortex mostly from entorhinal afferents. The results suggest that HFLS of EC → AC CCK + projection rather than VC → AC CaMKII + projection is necessary to induce potentiation of VC → AC inputs, whereby CCK release induced by the former one is an underpinning mechanism. Taken together, our results demonstrate a typical form of heterosynaptic plasticity, in which the potentiation of the VC → AC input is not dependent on HFLS of its own pathway but requires HFLS of the EC → AC projection that presumably triggers CCK release.


Introduction 55
Cross-modal association is crucial for our brain to integrate information from different 56 modalities to provide a useful output. Traditionally, this process is assumed to mainly occur in 57 higher association cortices as evidenced by both anatomical (Cusick et al., 1995;Seltzer et al., 58 1996) and physiological (Fuster et

124
The auditory cortex receives a direct projection from the visual cortex 125 To examine the origin of the visual information underlying the previously observed visual   136 High-frequency stimulation (HFS) is a classical protocol to induce LTP (Bashir et al., 1991; Hz with an ITI of 10 s ( Figure 1D upper). We chose the laser intensity that induced a 50% 151 fEPSP saturation for baseline and the post-HFLS tests and 75% for HFLS. However, no 152 significant LTP was induced in the VC→AC projection by HFLS of this pathway alone ( Figure   153 1D bottom, p = 0.403, n = 10, paired t-test). 155 in the presence of CCK 156 Hebbian theory says that cells that fire together wire together. We next tested if VC→AC inputs 157 can be potentiated after pairing with repetitive AC activation. We used VALS to evoke 158 presynaptic input and noise stimulus to trigger postsynaptic AC firing. Since the latency of 159 fEPSPVALS is approximately 2-2.5 ms, and the firing latency of noise responses in the AC of 160 mice is mostly equal to or longer than 13 ms, we presented the laser stimulus 10 ms after noise. 161 Therefore, we started the presynaptic input just before the postsynaptic firing (Pre/Post Pairing, 162 Figure 1E). Responses to noise at different sound intensities were first tested ( Figure 1F), and we 163 chose the intensity that evoked reliable firing for Pre/Post Pairing. However, even after 80 trials 164 of Pre/Post Pairing, the VC→AC inputs were not potentiated ( Figure 1G, gray, ACSF group). 165 In the previous studies we have shown that CCK has an important role in neocortical  Figures 1G and S1C Collectively, these results provide evidence that CCK together with noise, but not noise alone, 176 enables a visuo-auditory association in the AC via a direct projection from the VC to AC. (G) Normalized fEPSPVALS slopes (bottom) and example traces (upper) before and after Pre/Post Pairing with CCK-8S (red) or ACSF (gray) infusion in the AC. Error bars represent SEM. *** p < 0.001, n = 14 for CCK group, n = 15 for ACSF group, two-way RM ANOVA with Bonferroni post-hoc test. See Table S1 for detailed Statistics. 181 postsynaptic firing in the AC evoked by noise stimuli 182 We have shown that cortical projection neurons in the EC mostly are CCK + and glutamatergic, 183 and that HFS induces CCK release in the auditory cortex (Chen et  with two color activation. We injected AAV9-Ef1α-Flex-Chronos-GFP and AAV9-hSyn-    Table S1 for detailed Statistics. 223 electrical stimulation 224 We also performed similar experiments at the single cell level in vitro. Slices were prepared from 225 CCK-iRES-Cre mice after injection of AAV9-Ef1α-Flex-Chronos-GFP in the EC and of AAV9-226 Syn-ChrimsonR-tdTomato in the VC ( Figure 3A). Pyramidal neurons in the auditory cortex were 227 patched ( Figure 3B), and excitatory postsynaptic currents evoked by VALS (EPSCVALS, Figure   228 3C) and electrical stimulation of the auditory cortex (EPSCESAC, Figure 3D) were recorded.

229
HFLS of the EC→AC CCK + projection ( Figure 3E) was followed by the pairing of VALS and   Figure 3G) Figure 3J). These results, from recording at the synaptic level, provide further evidence 249 for the view that HFLS of EC→AC, CCK + projection is a prerequisite to potentiate the VC 250 inputs to the AC.  Table S1 for detailed Statistics. CCK levels was also higher in the EC compared with the VC ( Figure 4F). These results suggest 268 that after HFLS more CCK is released from EC→AC neurons than from VC→AC neurons, 269 which may, at least, be one explanation why the former but not the latter can produce LTP.  Table S1 for detailed Statistics. 1989), and our results suggest that CCK released from the EC→AC CCK + projection was critical 276 for generating visuo-auditory cortical LTP. We hypothesized that the frequency of the laser used 277 to stimulate the EC→AC CCK + projection was critical for the level of potentiation of the 278 VC→AC input. We therefore varied the frequency of the laser stimulation (80, 40, 10, or 1 Hz).

279
As shown in Figure 5A left, the delay between the termination of repetitive laser stimulation of 280 the CCK + EC→AC projection and presynaptic activation (Delay 1) was set at 10 ms, and the 281 delay between pre and postsynaptic activation (Delay 2) was set at 0 ms. The potentiation level 282 of the VC→AC inputs showed a tendency to increase as the frequency of laser stimulation of the 283 CCK + EC→AC projection increased ( Figure 5A  comparison, p < 0.001, n = 13). If higher than 10 Hz, the VC→AC input was significantly 289 potentiated. However, at 1 Hz no significant potentiation was observed.

334
To that end, we examined the role of CCK in the formation of visuo-auditory associative 335 memory in a fear response test ( Figure 6A). First, a CCKB receptor (CCKBR) antagonist (L-   361 As seen after treatment with a CCK antagonist, we expected that CCK -/mice would show a 362 deficit in the formation of associative memory, and we tested this hypothesis ( Figure 6C). Our 363 previous results demonstrated that 6-9 trials were needed for CCK -/mice to produce a freezing 364 rate of >60% in response to the conditioned AS, whereas only 3 trials were needed for wild-type  is evidence that the tetrapeptide CCK-4 can penetrate the blood-brain barrier (Rehfeld, 2000).

371
The CCK-4 dosage (1ug/kg) was at sub-panic attack level, and animals showed no sign of panic 372 or anxiety after administration ( Figure S5). CCK-4 injection resulted in a significantly higher 373 freezing rate compared to the controls (Videos S13-16; Figure 6D Table S1 for detailed Statistics.

390
In the present study we demonstrate, in the mouse, that a direct input from the VC to the AC can      In summary, we found that a direct projection from the VC to the AC provide an   Figure S2A right, the fEPSP slopes gradually increased with increasing 541 intensity of the 473 nm laser and became saturated at 30 mW/mm 2 (green solid). However, no 542 responses were evoked by the 635-nm laser, even at an intensity of 40 mW/mm 2 (red dash).

543
Conversely, in animal with injection of AAV9-hSyn-ChrimsonR-tdTomato in the VC the fEPSP 544 slopes gradually increased and became saturated (red solid) with the 635 nm laser. However, 545 here 40 mW (green dash) produced fEPSPs, but they were relatively small ( Figure S2B right).   552 At least four weeks after virus injection, acute brain slices were prepared using a protective 553 cutting and recovery method to achieve a higher success rate for patch clamp. Briefly,

608
Associative learning test after CCKBR antagonist application in the AC. 609 After the same anesthesia and surgery as mentioned earlier, a drug infusion cannula was     See Table S1 for detailed Statistics. See Table S1 for detailed Statistics. See Table S1 for detailed Statistics.

Supplementary Figure 4.
Comparison of Cck expression level in neurons in the Ent and VC which project to the VC. For each animal, Cck expression level is normalized to the average Cck expression in projecting neurons in visual cortex. Unpaired t-test, ****p<0.0001, **p<0.01, *p<0.05.