Glutamatergic basolateral amygdala to anterior insular cortex circuitry maintains rewarding contextual memory

Findings have shown that anterior insular cortex (aIC) lesions disrupt the maintenance of drug addiction, while imaging studies suggest that connections between amygdala and aIC participate in drug-seeking. However, the role of the BLA → aIC pathway in rewarding contextual memory has not been assessed. Using a cre-recombinase under the tyrosine hydroxylase (TH+) promoter mouse model to induce a real-time conditioned place preference (rtCPP), we show that photoactivation of TH+ neurons induced electrophysiological responses in VTA neurons, dopamine release and neuronal modulation in the aIC. Conversely, memory retrieval induced a strong release of glutamate, dopamine, and norepinephrine in the aIC. Only intra-aIC blockade of the glutamatergic N-methyl-D-aspartate receptor accelerated rtCPP extinction. Finally, photoinhibition of glutamatergic BLA → aIC pathway produced disinhibition of local circuits in the aIC, accelerating rtCPP extinction and impairing reinstatement. Thus, activity of the glutamatergic projection from the BLA to the aIC is critical for maintenance of rewarding contextual memory.

While I believe this manuscript has many strengths, it also suffers from several weaknesses that need to be addressed before being considered for publication.
1. I would recommend that the authors restructure their Introduction to better frame their study. Currently, the introduction describes the role of the AIC and BLA-AIC projects in the context of addiction. None of the experiments, however, involve any drug administration or manipulation. I agree that it these experiments will indeed inform our understanding of the neurobiology underlying drug addiction, but I do not think that two paragraph's worth of detail is necessary. 2. Given the methods and results, the authors should introduce memory mechanisms and how the AIC and BLA-AIC projections may be involved. 3. The authors should provide a stronger rationale for the approach they chose when designing the experiment. For example, why did the authors choose to use VTA stimulation rather than a drug stimulus to promote CPP? 4. I would recommend that the authors review the paper from Bernard Balleine's lab that focuses on BLA interactions with AIC (e.g., Parkes and Balleine, 2013)-it might be useful in the Introduction or Discussion. 5. There is no need to discuss the results in such detail in the Introduction. 6. The experimental group labeling is extremely confusing in the text and in the figures. Instead of ChR2-, for example, it would be easier to understand that this is the control vector if the authors used eYFP. Similarly, in Figure 1d, ChR2/extinction and ChR2/reinstatement imply that one group is stimulated during extinction and the other during reinstatement, which is not the case. Please use more intuitive group names for ease of reading. 7. In the Methods section, it is difficult to understand which details pertain to which manipulation. Perhaps it would be helpful to separate out the Methods by "experiment" or use subsections. For example, in the photoinhibition section, it would be easier to read and understand if the authors had separate subsections for 1. The ChR2 and eNpHR3.0 injections + optode implant vs. 2. eNpHR3.0 + optode. It could be framed either as separate experiments or just have subheadings; in either case, it will allow readers to follow the methods more clearly. 8. What is the timing/durations of the behavioral sessions? In particular, are the 3 conditioning sessions on the same day or do they occur across 3 days? What is the interval between the last conditioning session and the post-conditioning session? What is the interval between the postconditioning session and the first extinction session? Do all extinction sessions occur on the same day or do they occur across 5 days? These are important details as they can have a significant impact on the interpretation of the results. 9. I was slightly confused about the experimental design for experiments reflected in Figure 3. It was not clear whether VTA stimulation occurred during a CPP session (causing CPP) or that VTA stimulation occurred alone (i.e., not in the CPP training context). The authors loosely equate VTA stimulation with conditioning, but it is not clear whether conditioning is actually occurring-if VTA stimulation occurs during CPP training, I would agree that this is conditioning. If it is just VTA stimulation alone, I would argue that no conditioning is occurring and that the authors need to adjust the wording. I would also argue that, if the authors are trying to claim that these neurotransmitters are being released/involved during reconsolidation, you need to actually condition the animals (VTA stimulation in the CPP session). 10. I think that the title is misleading: the glutamatergic BLA-AIC circuit experiment was only one small part of the study. It mostly focused on the INS and the neurochemistry within. I would recommend changing the title to more broadly capture the results in the manuscript. 11. I do not think that "contextual reward" is the best way to describe what is being modeled by CPP. The association between context and VTA stimulation (which presumably is reward) is a more accurate description. 12. Is it possible for the authors to dissociate whether the effects of the various manipulations are affecting memory processes, as the authors suggest, or the reinforcing properties of the context imbued by VTA stimulation? 13. Did the authors sample neurotranmisster release in the AIC during the first extinction session? 14. There are several typos throughout the manuscript. For example, a. Mistake in caption for Figure 5-the figure shows A, B and C, but the caption describes A, C and D. b. On page 9, in the NMDA infusion section, the authors begin the paragraph by stating that they are going to characterize the role of dopamine and noreadrenaline in the AIC, but later in that same paragraph, they describe a pharmacological approach targeting the glutamatergic system. Did the authors mean glutamate in the beginning of the paragraph rather than dopamine and noreadrenaline? Could the authors clarify this discrepancy? c. I do not think "labialize" is actually a word (or at least word related to memory). Please re-word (pg. 13).
Reviewer #3 (Remarks to the Author): This paper by Gil-Lievana et al. investigates the role of the BLA-aIC projection in the maintenance of rewarding contextual memories. The authors employ optogenetic stimulation of the TH+ neurons of the VTA to induce real-time conditioned place preference (rtCPP) and to establish reward-based contextual memories in mice.
The paper is subdivided into five major sections: 1) the authors first show how optogenetic stimulation of the VTA TH+ neurons induces robust rewarding context memories. 2) An assessment of how photoactivation of the VTA TH+ neurons modulates neuronal responses in the anterior insular cortex (aIC). 3) An assessment of glutamate, noradrenalin and dopamine release during rtCPP memory retrieval. 4) Intra-aIC infusions of dopamine, noradrenalin and glutamate receptor antagonists after rtCPP retrieval and assessment of extinction. 5) photoinhibition of BLA-aIC pathway during rtCPP extinction.
Overall, my impression is that the story would have the potential to be a valid contribution yet at this point it lacks crucially in clarity and straightforwardness. It is hard to follow the logic of the experimental line-up. It seems as if titration experiments, control experiments and important results are intermingled, making it hard to understand the take home messages. Yet, the manuscript addresses an important question, employs many different and adequate techniques to tackle this question, and contains some important results. Therefore, my overall recommendation would be a thorough revision mainly of the way the paper is presented and the data are lined up.
Major issues: -Lines 156 -167 and Fig. 2b (left) are titration experiments aimed at testing the VTA stimulation protocol -this comes after Fig. 1 where a stimulation frequency of 20 Hz has already been chosen and shown to trigger rtCPP -this does not make much sense to me. I'd suggest to put this titration into the supplement and mention it before the Fig1 rtCPP results. The same inconsistency is the summary sentence in line 174-176, where the conclusion is that 20 Hz was used -but the behavior is already shown before this in  Fig. 2b right and 2c right seem to show that VTA stimulation does not evoke much change in aIC firing. Since it seems that the authors report both, stimulation activated and inhibited neurons it would make sense to show these responses separately and quantitatively (similar to Fig. 6?). What was the threshold to calls a neuron activated / inhibited? Even though the quantifications in d, e, f show some modulation these are not shown with error bars or statistical measures. Also, it is not clear what the threshold for 'modulation' was. I did not find a threshold or measure for categorizing 'modulation' in the methods either. Neurons in VTA (Fig. 2b lower left) seem to react to photostimulation already before the laser onset (or is this a plotting issue?).
Lines 189 and following -The paragraph starts stating: 'we decided to pharmacologically characterize the role of dopamine and noradrenaline receptors in the aIC on retrieval and extinction' …. (lines 191-193). Then, they write 'we thus administered bilateral injections of AP5' (line 200). This is totally inconsistent. Why now AP5 while the whole Fig. 3 showed post-conditioning increases of not only glutamate but also NA and DA and they themselves state they want to test roles of NA and DA? -While it is nice to see that intra-aIC injections of AP5 post-conditioning accelerates extinction -these results come totally out of the blue -why then bother and go through all the pain of microdialysis in Fig. 3? -only in line 206 the authors then resolve this issue (partially and inadequately) mentioning in brief that SCH23390 or propranolol had no effect. The authors have to keep an order in presenting the results that remains penetrable for the reader. They also do not mention in the main text at all what SCH23390 is, neither do they explain what propranolol is -nor why they used it. Only the informed reader knows what these drugs do -it should not be assumed that the common reader of this journal know what SCH23390 does! This paragraph belongs -properly explained BEFORE the AP5 and CNQX results (at least in the current format). I would even suggest to present all the results side-by-side in Fig 4 (include data from S1 b and c -only left panels). I believe that this would be more consistent since data in Fig. 3 are presented also for NA, DA and glutamate -so it would become easier to follow for the reader.
-The second problem with the AP5 experiment is that the authors themselves cite papers that already showed that AP5 accelerates extinction -what is the novelty then here? Why is the AP5 result in this paper novel and important? What are the main differences to the earlier studies?
BLA-aIC photoinhibition -In the last section of the paper the authors then jump to the BLA-aIC pathway. Why? While it is true that there were earlier studies it is a hard shift of direction and question.
-When exactly were the photoinhibitions done -during the post-conditioning or when done after postconditioning. Which parameters?
-It seems very odd that even after reinstatement to a similar level the BLA-aIC inhibited group is again strongly facilitated in extinction, could this be due to opsin-cross-talk?
-An important control would be to demonstrate whethet BLA-aIC inhibition itself has a valence in the rtCPP task -I do not find the opsin-cross-talk control convincing. At least according to excitation spectra of the opsins they should activate CHR2 expressed in VTA terminals with the green light and also inhibit BLA terminals with the blue light. It doesn't help that I couldn't find the exact wavelength utilized but maybe I oversaw the mentioning? -Implantation sites and viral spreads have to be shown for all individuals included in the analyses. -the methods section should be extended with more detail throughout to really understand all parameters of stimulation and the timing of different experiments etc.

Reviewer 1
Comment: The manuscript by Gil-Lievana et al. examines the role of the anterior insular cortex (AIC) and, particularly, the projections from the basolateral amygdala (BLA) to the AIC in the maintenance/extinction of a conditioned place preference (CPP). To induce the CPP, the authors used photostimulation of the VTA to produce the CPP. They found that blockade of the NMDA receptors in the AIC accelerated the extinction and reduced the ability of a reinstatement procedure to restore the CPP. Similarly, they found that photoinhibition of BLA axons in the AIC produced essentially the same results. The authors place these findings in the context of reconsolidation. Overall, the findings are rather interesting, and the studies appear to be well done. Considering the increased interest in the insular cortex, the manuscript is also rather timely, providing the field with new ideas about the interaction between the amygdala and insular cortex with regard to reward learning and extinction. The statistical analyses and methods are well described, and I believe that a research could replicate the work. The manuscript is generally well written, though I have a few specific concerns detailed below.

1.
To fully place the findings in a reconsolidation context, the authors would need to do additional studies, such as giving the manipulation without the CPP test. Moreover, there appears to be a problem is comparing the NMDA receptor blockade to the inhibition of BLA inputs to the AIC. For the NMDA receptor blockade, AP5 was administered into the AIC after the post-conditioning test and was found to be effective for enhancing extinction. In contrast, inhibition of BLA inputs to the AIC was only effective when performed during the test and not after. If the idea is that the AP5 was blocking glutamatergic signaling that included glutamate released from the BLA terminals, these findings are not consistent with that. Indeed, one would expect both manipulations to work equally effectively. The authors should address this discrepancy.

Response:
We want to thank the reviewer's comments for improving our manuscript. We observed that the post-retrieval blockade of the glutamatergic NMDA receptors with AP5 accelerates the extinction of rtCPP. However, inhibition of BLA terminals to the aIC postretrieval did not accelerate rtCPP extinction. Interestingly, when the BLA terminals in the aIC were inhibited during retrieval, the post-conditioning rtCPP scores were not affected, while 24 hours later, the mice that received the photoinhibition showed an accelerated extinction rate. It is possible that the glutamate release in the aIC necessary for the maintenance of the rtCPP memory not only comes from the BLA, but other structures could contributes with the glutamatergic activity in the aIC after the retrieval. Although the similarities of the cellular mechanisms involved in the photoinhibition of glutamatergic terminals and the glutamatergic receptor antagonists remain to be determined, the difference observed between optogenetic inhibition of the BlA terminals in the aIC and the pharmacological inhibition of the NMDA receptors in the aIC could be also explained due to that optogenetic inhibition of BLA terminals in the aIC during retrieval unleashed similar molecular events as the blockade of the AP5 receptor after the retrieval. In this regard, we are currently evaluating whether photoinhibition of the BLA-aIC terminals blocks the release of glutamate that affects the cellular events necessary to reconsolidate (update) context reward associations, in a similar way than we have reported with AP5 (please see, Osorio-Gomez et al., 2017) (please see, page 11, lines 256-266).

2.
My own view is that this calls into question whether these are truly "reconsolidation" results. One possibility is that BLA inputs to the AIC may be important for "maintaining" the memory (as the title suggests) and that disrupting this enables extinction to forward more quickly. This is quite different, however, than demonstrating reconsolidation. As it is, I would suggest that the authors re-frame their findings and only acknowledge that reconsolidation is a *possible* explanation of the results.

Response:
We agree with the reviewer, the reconsolidation phenomenon is one possibility. We hypothesized that the optogenetic inhibition of the BLA projections in the aIC impairs the reconsolidation process of the rewarding contextual memory, but we can not rule out that alternative mechanisms involved in the maintenance of the rewarding memory, could be disrupted and be reflected in an accelerated extinction rate (please see, page 11, lines 264-268).

3.
The authors appear to have used both males and females but, unless I missed it, do not comment on whether any sex differences (trends or otherwise) were observed. They should do so.

Response:
No difference was observed between mice's sex in any of the different phases analyzed in the rtCPP or in the neurotransmitter profile, which is reflected in the small standard error of the mean in the different groups. Figure 4 has panels b and c reversed from the text.

Response:
We apologize for this oversight, it is now fixed.

Reviewer 2
In this study, the authors focused on the role of the anterior insular cortex (AIC) and its input from the basolateral amygdala (BLA) in mediating the association between contextual information and reward. The authors used various techniques to show that stimulation of tyrosine hydroxylase (TH)-expressing neurons in the ventral tegmental area (VTA) is sufficient to induce conditioned place preference (CPP) and the retrieval of such memories require neurotransmission in the AIC. Further, intra-AIC microinjections of the glutamate receptor antagonist NMDA facilitate extinction of the CPP memory, as does inhibition of BLA projections to the AIC. While I believe this manuscript has many strengths, it also suffers from several weaknesses that need to be addressed before being considered for publication.
1. I would recommend that the authors restructure their Introduction to better frame their study. Currently, the introduction describes the role of the AIC and BLA-AIC projects in the context of addiction. None of the experiments, however, involve any drug administration or manipulation. I agree that it these experiments will indeed inform our understanding of the neurobiology underlying drug addiction, but I do not think that two paragraph's worth of detail is necessary. Fig. 2b right and 2c right seem to show that VTA stimulation does not evoke much change in aIC firing. Since it seems that the authors report both, stimulation activated and inhibited neurons it would make sense to show these responses separately and quantitatively (similar to Fig. 6?). What was the threshold to calls a neuron activated / inhibited? Even though the quantifications in d, e, f show some modulation these are not shown with error bars or statistical measures. Also, it is not clear what the threshold for 'modulation' was. I did not find a threshold or measure for categorizing 'modulation' in the methods either. Neurons in VTA (Fig.  2b lower left) seem to react to photostimulation already before the laser onset (or is this a plotting issue?).

3.
Response: Thank you for your thoughtful comments. As suggested by the Reviewer, we have modified the previous Figure 2 now Figure 1 with the same style than Figure 6. Figure 1 now had a heatmap with the neuronal responses split as a function of increase or decrease responses (please see, page 25). Moreover, we also include a new figure, the supplementary Figure S1 (please see, page 31), where it is plotted the percentage of inhibited or activated neurons for each stimulation frequency and brain region. This figure also displays the analysis of spike-laser coherence.
Regarding your question about the threshold, we apologize for this missing information. We use a Kruskal-Wallis test to determine a modulation, we have now written in Methods the following: "In the case of electrophysiological recordings, the firing rate was compared using a nonparametric Kruskal-Wallis test followed by a Tukey-Kramer post hoc. Neurons which firing rates, during the laser "on" period was significantly different (p<0.05), at any laser frequency, relative to the activity in the control trial was considered as modulated. The difference between firing rates during laser stimulation vs. control trials determined the sign of the modulation (i.e., increased or decreased)." (please see, page 18-19, lines 437-442) Regarding your question: "Even though the quantifications in d, e, f show some modulation these are not shown with error bars or statistical measures." Note that there are no error bars, because they are the percentage of total neurons recorded (for this information see now Supplementary Figure 1 on page 31).
Finally, relative to your question: "Neurons in VTA (Fig. 2b lower left) seem to react to photostimulation already before the laser onset (or is this a plotting issue?)." Thank you for noticing this point. There is no activation preceding laser stimulation the reason for this is that for visualization purposes, the responses were smoothed, to minimize this issue, we have now used a smaller bin size and less smoothing.

4.
The paragraph starts stating: 'we decided to pharmacologically characterize the role of dopamine and noradrenaline receptors in the aIC on retrieval and extinction' …. (lines 191-193). Then, they write 'we thus administered bilateral injections of AP5' (line 200). This is totally inconsistent. Why now AP5 while the whole Fig. 3 showed post-conditioning increases of not only glutamate but also NA and DA and they themselves state they want to test roles of NA and DA? -While it is nice to see that intra-aIC injections of AP5 post-conditioning accelerates extinction