IC neurons activated by memory retrieval do not depend on neural activity during learning
We have previously shown that CREB levels determine which the IC neurons that proceed to encode a given conditioned taste memory [28]. However, the cellular mechanism of memory allocation in IC is unclear. We as well as other researchers showed that excitable neurons are preferentially involved in the process of memory formation in the amygdala, and these neurons are reactivated during memory retrieval [33, 37]. In the previous study, hM3Dq, a Gq protein coupled receptor- was used to manipulate neuronal activity in the amygdala neurons [37]. The hM3Dq showed no constitutive activity in increasing neuronal activity [39]. A synthetic ligand, clozapine-N-oxide (CNO), but not an endogenous ligands for muscarinic receptors can selectively bind to and activate the hM3Dq receptors [40]. The activation of amygdala neurons expressing hM3Dq during auditory fear conditioning determined the neurons that are involved in fear memory retrieval [37]. In this study, hM3Dq was expressed in a subset of IC neurons to induce neuronal competition; we also tested whether the activation of IC neurons by hM3Dq enhances the process of memory allocation as in the amygdala (Fig. 1a). To express hM3Dq in a subset of IC neurons, we utilized the Cre-dependent hM3Dq expression system: both hSyn-DIO-hM3Dq-mCherry and diluted CaMKIIa-Cre (109 gc/mL) adeno-associated viruses (AAVs) were mixed and infused bilaterally into the IC according to the brain atlas. Immunohistochemical studies with an antibody against mCherry detected the expression of this viral gene in a region about 1 mm away from the injection sites in the IC. These mice were then subjected to CTA (Fig. 1a). To increase the neuronal activity in a subset of IC neurons, 2 mg/kg CNO was systemically administrated 30 min before CS and US presentations. A memory retrieval test was implemented 1 day following the conditioning in the absence of CNO.
First, we analyzed whether hM3Dq-positive (hM3Dq+) neurons are preferentially activated by conditioning with CNO administration. To visualize these neurons, we used the c-fos protein. The expression of c-fos is rapidly and transiently induced by neuronal activity with the elevation of intracellular Ca2+ concentration. The expression of c-fos protein in hM3Dq + neurons was significantly higher in the CNO group than in the saline group (Fig. 1b and c; c-fos+/hM3Dq + neurons (%), N = 3 in each group: saline group, 3.0% ± 1.7% CNO group, 66.1% ± 4.9%; p = 0.0003). The hM3Dq was expressed in a subset of neurons around the injection site (hM3Dq + neurons/DAPI (%), N = 3 in each group: saline group, 1.5% ± 0.3%; CNO group, 2.4% ± 0.2%). However, CNO administration did not affect to the total number of c-fos positive (c-fos+) neurons in IC (Fig. 1d; c-fos+/DAPI (%), N = 3 in each group: saline group, 5.9% ± 1.6%; CNO group, 6.4% ± 0.9%; p = 0.82) and in BLA (Supplementary Fig. 2A; c-fos+/DAPI (%), N = 3 in each group: saline group, 8.0% ± 0.7%; CNO group, 7.6% ± 1.4%; p = 0.80). This result indicates that neuronal activation is preferentially induced in hM3Dq-positive neurons by CTA conditioning with CNO administration. Subsequently, we tested whether highly excitable neurons during conditioning are preferentially reactivated by CTA memory retrieval. The hM3Dq was bilaterally expressed in a subset of IC neurons (hM3Dq + neurons/DAPI (%): N = 7 in each group; saline group, 3.5% ± 0.4%; CNO group, 3.8% ± 0.5%). After the memory retrieval test, we imaged and analyzed the colocalization of c-fos protein in hM3Dq + neurons. Contrary to our prediction, the probability of c-fos expression induced by memory retrieval in hM3Dq + neurons was not different between the CNO and saline administrated groups, though the probability tended to be increased in the CNO group (Fig. 1e and f; c-fos+ /hM3Dq + neurons (%): N = 7 in each group; saline group, 9.4% ± 3.6%; CNO group, 16.4% ± 5.6%; p = 0.32). The activation of a small population of IC neurons did not affect the total number of c-fos-positive neurons after memory retrieval (Fig. 1g; c-fos+/DAPI (%):N = 7 in each group; saline group, 4.5% ± 0.6%; CNO group, 4.7% ± 0.9%; p = 0.88), the quantity of saccharine solution consumed during conditioning (Fig. 1h; drinking amount (g): N = 13 in each group; saline group, 1.6 g ± 0.1 g; CNO group, 1.4 g ± 0.1 g; p = 0.21), and aversion index (Fig. 1i; aversion index(%): N = 13 in each group; saline group, 64.6% ± 6.2%; CNO group, 65.4% ± 7.4%; p = 0.94). The aversion index in our methods is relatively weaker than other studies. Therefore, we also showed drinking amounts during retrieval test, in which mice avoided drinking saccharine solution (Supplementary Fig. 1A). These data suggest that an increase in the neuronal activity induced by hM3Dq activation with learning only in a subset of IC neurons cannot determine the neurons that are preferentially activated by CTA memory retrieval.
Activation of BLA neurons involved in a CTA memory retrieval depends on neuronal activity during learning
Previous studies showed that fear memory is preferentially recruited into highly excitable neurons in the amygdala [33, 37]. As in the case of the previous study, we used hM3Dq to manipulate neuronal activity. Similar to the IC-related experiment, hM3Dq was bilaterally expressed in a subset of BLA neurons (Fig. 2a). First, we investigated whether hM3Dq + BLA neurons were preferentially activated by conditioning with CNO administration. The c-fos expression in hM3Dq + neurons was significantly higher in the CNO group than in the saline group (Fig. 2b and c; c-fos+/hM3Dq + neurons (%): saline group, 5.6% ± 1.2%, N = 4; CNO group, 46.8% ± 3.4%, N = 5; p = 0.00002). CNO administration had no effect on the total number of c-fos + neurons in either the BLA (Fig. 2d; c-fos+/DAPI (%): saline group, 9.4% ± 0.9%; CNO group, 9.4% ± 0.8%; p = 0.98) or in the IC (Supplementary Fig. 2B; c-fos+/DAPI (%): saline group, 6.6% ± 0.8%, N = 4; CNO group, 5.6% ± 0.9%. N = 5; p = 0.42). This result indicates that neuronal activation is preferentially induced in hM3Dq + BLA neurons by CTA conditioning with CNO administration. Next, we tested whether the recruitment of taste memory is also regulated by neuronal activity in the BLA. Again, hM3Dq was bilaterally expressed in a subset of BLA neurons (hM3Dq + neurons/DAPI (%): saline group, 2.1% ± 0.5%, N = 7; CNO group, 2.9% ± 0.4, N = 8), and the activity of hM3Dq + neurons was increased by the administration of CNO during conditioning (Fig. 2a). Memory retrieval test was implemented 1 day after learning, and brain samples were harvested 90 min after the test. Subsequently, we analyzed the colocalization of c-fos protein in hM3Dq + neurons. Consistent with memory allocation in fear learning, the probability of c-fos expression in hM3Dq + neurons was significantly higher in the CNO group than in the saline group (Fig. 2e and f; c-fos+/hM3Dq + neurons (%): saline group, 5.3% ± 4.1%, N = 7; CNO group, 30.1% ± 6.4%, N = 8; p = 0.008). However, the total number of c-fos + neurons was not different between the saline and CNO groups (Fig. 2g; c-fos+ /DAPI (%): saline group, 3.3% ± 0.8%, N = 7; CNO group, 4.4% ± 0.9%, N = 8; p = 0.41). For the behavioral performance, increased neuronal activity in a subset of BLA neurons did not affect to the quantity of saccharine solution consumed during conditioning (Fig. 2h; drinking amount (g), N = 9 in each group: saline group, 1.5 g ± 0.1 g; CNO group, 1.4 g ± 0.1 g; p = 0.56) and aversion index (Fig. 2i; aversion index (%), N = 9 in each group: saline group, 67.6% ± 7.9%; CNO group, 77.6% ± 6.0%; p = 0.33). We confirmed that mice avoided drinking saccharine solution in the retrieval test (Supplementary Fig. 1B). These data suggest that, during learning, an increase in the neuronal activity by hM3Dq in a subset of BLA neurons can determine the neurons that are preferentially activated by CTA memory retrieval. In the amygdala, the cellular mechanism of memory allocation is shared by the fear and taste memory formation.
Functional connectivity between the insular cortex and the basolateral amygdala is changed by the formation of stronger conditioned taste memory
Our results suggest that memory allocation in the IC could be regulated by different cellular mechanisms associated with the BLA. The IC and BLA are reciprocally connected, and these interconnections are important for the retention of CTA memory [17]. To know the interaction between IC and BLA, we analyzed the ratio (%) of c-fos + neurons in three brain areas, IC, BLA, and prelimbic cortex (PrL) and the correlations of the ratios between two different brain areas during stronger and weaker US- associated learning. A previous study has demonstrated the involvement of the PrL in CTA acquisition [41]. The novel saccharine taste (CS) was associated with different intensities of aversive experience depending on the administration of 150 mM lithium chloride (LiCl; Stronger US), 50 mM LiCl (Weaker US), or 150 mM NaCl (no US) in the stronger, weaker, or no US-associated learning groups, respectively. Memory retrieval test was implemented 1 day after conditioning (Fig. 3a). Aversion index was significantly higher in the stronger-US associated learning group than in the weaker and no US-associated learning groups, and aversion index of the weaker US group was significantly higher than that of the no US group (Fig. 3b; aversion index: Stronger US group, 74.7% ± 2.7%, N = 14; Weaker US group, 41.0% ± 4.7%, N = 14; no US group, 13.7% ± 2.3%, N = 13; one-way analysis of variance (ANOVA); F(2,38) = 77.9, p = 3.6 × 10− 14; Tukey-Kramer test, Stronger vs. Weaker US group, p = 7.1 × 10− 8; Stronger vs. No US group, p = 1.4 × 10− 13; Weaker vs. No US group, p = 6.8 × 10− 6). These results suggest that the association between saccharine taste and aversive experience developed in both the stronger and weaker US-associated learning groups, and the strength of the association was dependent on the concentration of LiCl.
To test whether the strength of CS-US association is related to the number of neurons activated by memory retrieval, we analyzed the number of c-fos + neurons in the IC, BLA, and PrL, after the retrieval test. In the three brain areas, the number of c-fos-positive neurons showed no differences between the stronger and weaker US-associated learning groups (Fig. 3c-f; c-fos+ /DAPI (%), N = 5 in each group: [D]; IC, stronger-learning group, 6.8% ± 0.9%; weaker-learning group, 5.3% ± 1.1%; p = 0.31; [E]; BLA, stronger-learning group, 8.2% ± 1.4%; weaker-learning group, 9.4% ± 0.9%; p = 0.50; [F]; PrL, stronger-learning group, 6.6% ± 0.7%; weaker-learning group, 7.6% ± 2.0%; p = 0.65). These results suggest that the strength of CTA memory recall does not depend on how many neurons are activated in these brain areas. Subsequently, we analyzed the correlation of c-fos + neurons between brain areas. The correlation coefficients for the number of c-fos-positive neurons between the IC and BLA were 0.85 and 0.83 in stronger and weaker US-associated learning groups, respectively (Fig. 3g). However, such as strong correlation was not observed between the BLA-PrL (Fig. 3h; stronger-learning, r = − 0.13; weaker-learning, r = 0.67) and IC-PrL (Fig. 3i; stronger-learning, r = 0.01; weaker-learning, r = 0.38). These results suggest that IC and BLA neurons could be coactivated during memory recall. Subsequently, we analyzed the differences in the number of c-fos + IC neurons correlating to that of c-fos + BLA neurons between the stronger and weaker US-associated learning groups. The number of c-fos-positive neurons in IC correlating to that in BLA was significantly higher in the stronger US-associated learning group than in the weaker US-associated learning group (Fig. 3g, p = 0.04). These results suggest that the interaction of cell assembly between the IC and BLA is correlated with the strength of memory retrieval.
BLA neurons projecting on the IC are preferentially activated by CTA memory retrieval
To test which neural circuit is related to CTA memory retrieval, we used the retrograde neuronal tracer cholera toxin B subunit (CTB) along with c-fos protein imaging. When Alexa 555 labeled CTB was locally infused into the BLA or IC (Supplementary Fig. 3A), and CTB-positive (CTB+) neurons were observed in the IC or BLA, respectively (Supplementary Fig. 3A and B). We analyzed c-fos expression on CTB+ and CTB- excitatory neurons following memory retrieval. CaMKIIa protein was used as a marker of excitatory neurons, and the probability of c-fos expression on CTB+ excitatory neurons was compared with that on CTB- excitatory neurons. There was no difference between numbers of c-fos+/CTB+ and c-fos+/CTB- cells in IC, when CTB was infused into the BLA (Supplementary Fig. 3B and C; N = 6, c-fos+/CTB+ neurons (%), 4.4% ± 1.6%; c-fos+/CTB- neurons (%), 5.3% ± 1.0%; p = 0.67). On the contrary, the probability of c-fos expression on CTB+ BLA neurons was significantly higher than that on CTB- neurons (Supplementary Fig. 3B and D; N = 6, c-fos+/CTB + neurons (%), 9.0% ± 1.2%; c-fos+/CTB- neurons (%), 4.3% ± 0.6%; p = 0.005), when CTB was infused into the IC. We confirmed that CTA memory formation was not impaired by the infusion of CTB into BLA (Supplementary Fig. 3E; drinking amount (g), N = 6, saccharine solution, 0.3 g ± 0.1 g; tap water, 1.2 g ± 0.1 g; p = 0.00008) and IC (Supplementary Fig. 3F; drinking amount (g), N = 6, saccharine solution, 0.5 g ± 0.1 g; tap water, 1.0 g ± 0.2 g; p = 0.03). These results indicate that neuronal populations in the BLA which projecting to the IC would be preferentially activated by CTA memory retrieval in comparison to other populations. However, additional experiments such as the analysis of no retrieval group is important to further show that BLA neurons projecting to IC are preferentially incorporated into CTA memory trace. Additionally, further investigations are warranted to show the contribution of BLA and IC projecting circuits, their functional significance, and contribution of other circuits to memory formation.
IC neurons reciprocally connected with the BLA are preferentially activated by CTA memory retrieval
To examine further a neural circuit related to CTA memory retrieval, we used the AAV-mediated anterograde transsynaptic tagging [42] combined with CTB and c-fos imaging. In this experiment, Alexa 647 labeled CTB and hSyn-Cre AAV1 were locally infused into the BLA, and the Cre-dependent YFP AAV was locally infused into the IC (Fig. 4a). CTB-, YFP- and double-positive cells were observed in IC (Fig. 4a, N = 5, CTB+ cells/DAPI (%), 3.6% ± 1.0%; YFP+ cells/DAPI (%), 3.7% ± 0.9%; double-positive cells/DAPI (%), 0.5% ± 0.2%). CTA was performed 3 weeks after the surgery, and mice were perfused for brain sampling after the memory retrieval test (Fig. 4b; drinking amount (g), N = 5, saccharine solution, 0.2 g ± 0.02 g; tap water, 1.5 g ± 0.2 g; p = 0.00005). Subsequently, we analyzed the probability of c-fos expression on IC neurons that have a different neural connection with BLA neurons. Interestingly, the probability of c-fos expression in double-positive neurons was significantly higher than that in solely CTB+ and YFP+ cells (Fig. 4c and d, N = 5, c-fos+/CTB+ ∩ YFP+ [reciprocal] (%), 45.1% ± 8.3%; c-fos+/CTB- ∩ YFP+ [IC-from-BLA] (%), 14.0% ± 2.7%; c-fos+/CTB+ ∩ YFP- [IC-to-BLA] (%), 9.0% ± 2.7%; F(2, 12) = 13.8, p = 0.0008; Tukey-Kramer test, [reciprocal] vs. [IC-from-BLA], p = 0.003; [reciprocal] vs. [IC-to-BLA], p = 0.001; [IC-from-BLA] vs. [IC-to-BLA], p = 0.78). Collectively, these results suggest that IC neurons reciprocally connected with the BLA are preferentially activated by CTA memory retrieval.
Co-activation of the IC and BLA determines the IC neurons that are activated by CTA memory retrieval
Our c-fos imaging experiments suggest that IC neurons activated by CTA memory retrieval could be regulated by the interaction between the IC and BLA. Subsequently, we tested whether activating both the IC and BLA neurons during conditioning changes the IC neurons that are activated by memory retrieval. hM3Dq was expressed in a subset of IC and BLA neurons (hM3Dq + neurons/DAPI (%): IC, saline group, 2.3% ± 0.3%, N = 5; CNO group, 2.2% ± 0.3%, N = 6; BLA, saline group, 1.2% ± 0.3%, N = 5; CNO group, 1.8% ± 0.1%, N = 6), and neuronal activity in the hM3Dq neurons were induced by the administration of CNO during conditioning (Fig. 5a). Memory retrieval test was implemented 1 day after conditioning, and mice were perfused for brain sampling, 90 min following the retrieval test. For the CTA memory formation, increased neuronal activity in a subset of IC and BLA neurons did not affect the quantity of saccharine solution consumed during conditioning (Fig. 5b; drinking amount (g): saline group, 1.5 g ± 0.1 g, N = 16; CNO group, 1.7 g ± 0.1 g, N = 17; p = 0.15) and aversion index (Fig. 5c; aversion index (%): saline group, 66.4% ± 5.3%, N = 16; CNO group, 75.1% ± 5.4%, N = 17; p = 0.26). We confirmed that mice avoided drinking saccharine solution in the retrieval test (Supplementary Fig. 1C). In the imaging analysis, the total number of c-fos + neurons was not different between the saline and CNO groups (Fig. 5e; c-fos+ /DAPI (%): IC, saline group, 5.1% ± 0.8%, N = 5; CNO group, 4.3% ± 0.7%, N = 6, p = 0.50; BLA, saline group, 3.7% ± 0.7%, N = 5; CNO group, 4.1% ± 0.5%, N = 6; p = 0.63). Subsequently, we analyzed the colocalization of c-fos protein in hM3Dq + neurons. Interestingly, unlike the individually increased neuronal activity in the IC using hM3Dq alone, the probability of c-fos expression in hM3Dq + neurons in IC was significantly higher in the CNO group than in the saline group (Fig. 5d and f; c-fos+/hM3Dq + neurons (%): saline group, 2.6% ± 0.8%, N = 5; CNO group, 26.5% ± 6.3%, N = 6; p = 0.007). In the BLA, consistent with the above results, the corresponding probability was significantly higher in the CNO group than in the saline group (Fig. 5d and f; c-fos+/hM3Dq + neurons (%): saline group, 1.0% ± 1.0%, N = 5; CNO group, 26.9% ± 8.6%, N = 6; p = 0.02). To test whether CNO administration affects c-fos expression in IC and BLA neurons, mCherry was expressed in a subset of IC and BLA neurons and analyzed for c-fos expression. CNO was administrated during conditioning. Memory retrieval test was implemented 1 day after conditioning, and mice were perfused for brain sampling, 90 min following the retrieval test. For the CTA memory formation, CNO administration did not affect the quantity of saccharine solution consumed during conditioning (Supplementary Fig. 4A; drinking amount (g): N = 6 in each group: saline group, 1.6 g ± 0.1 g; CNO group, 1.5 g ± 0.1 g; p = 0.28) and aversion index (Supplementary Fig. 4B; aversion index (%): N = 6 in each group: saline group, 83.2% ± 5.4%; CNO group, 80.2% ± 4.9%; p = 0.70). In imaging analysis, the total number of c-fos + neurons was not different between the saline and CNO groups (Supplementary Fig. 4C; c-fos+ /DAPI (%): N = 5 in each group: IC, saline group, 4.5% ± 0.5%; CNO group, 4.7% ± 1.4%, p = 0.89; BLA, saline group, 5.6% ± 0.7%; CNO group, 4.6% ± 0.8%, p = 0.36). The probability of c-fos expression in hM3Dq + neurons showed no differences between the IC and BLA (Supplementary Fig. 4D; c-fos+ / mCherry+ neurons (%): N = 5 in each group: IC, saline group, 6.2% ± 0.7%; CNO group, 5.8% ± 0.9%, p = 0.75; BLA, saline group, 7.3% ± 1.1%; CNO group, 6.0% ± 1.2%, p = 0.46). These results suggest that the coactivation of the IC and BLA during conditioning determines the IC neurons that are activated by CTA memory retrieval.