Non-Hebbian plasticity transforms transient experiences into lasting memories

Reviewer #1 (Public Review):

Summary:

The authors' goal here was to explore how a non-hebbian form of plasticity, heterosynaptic LTP, could shape neuronal responses and learning. They used several conceptually and technically innovative approaches to answer this. First, they identified a behavioral paradigm that was a subthreshold training paradigm (stimulation of thalamic inputs with a footshock), which could be 'converted' to memory via homosynaptic LTP (HFS of thalamic inputs). They then found that stimulation of 'cortical' inputs could also convert the subthreshold stimulation to a lasting memory and that this was associated with a change in neuronal response, akin to LTP. Finally, they provided some slice work that demonstrated that stimulation of cortical inputs could stabilize LTP at thalamic inputs.

Strengths:

1. The approach was innovative and asked an important question in the field.

2. The studies are, for the most part, quite rigorous, using a novel dual opsin approach to probe multiple inputs in vivo.

3. The authors explore neural responses both in vivo and ex vivo, as well as leveraging a 'simple' behavior output of freezing.

Weaknesses:

1. There appears to be a flaw in the exploration of cortical inputs. the authors never show that HFS of cortical inputs has no effect in the absence of thalamic stimulation. It appears that there is a citation showing this, but I think it would be important to show this in this study as well.

2. It is somewhat confusing that the authors refer to the cortical input as driving heterosynaptic LTP, but this is not shown until Figure 4J, that after non-associative conditioning (unpaired shock and tone) HFS of the cortex can drive freezing and heterosynaptic LTP of thalamic inputs. Further, the authors are 'surprised' by this outcome, which appears to be what they predict.

3. 'Cortex' as a stimulation site is vague. The authors have coordinates they used, it is unclear why they are not using standard anatomical nomenclature.

4. The authors' repeated use of homoLTP and heteroLTP to define the input that is being stimulated makes it challenging to understand the experimental detail. While I appreciate this is part of the goal, more descriptive words such as 'thalamic' and 'cortical' would make this much easier to understand.

Reviewer #2 (Public Review):

Faress et al. ask the question of how synaptic plasticity (i.e. long-term potentiation, LTP) induced at different time points and different synapses in relationship to learning can transform memories supported by these circuits. The authors adopted an experimental design developed by Nabavi et al, 2014 and used male mice to optogenetically induce a weak fear memory in thalamo-LA circuits by pairing an optical conditioned stimulus (CS) at thalamo-LA synapses with a footshock unconditioned stimulus (US), or subjected the mice to an unpairing of the opto-CS and the footshock US. They then investigated how homosynaptic (thalamo-LA)- or heterosynaptic (cortico-LA) high-frequency stimulation (HFS) -that would induce LTP- delivered at different time points before and after learning can transform the opto-fear memory by using state of the art in vivo dual-wavelength optogenetics. They find that homosynaptic HFS delivered before or after learning transforms weak memories into stronger ones, whereas heterosynaptic HFS can do so when delivered immediately after learning. Both homo- and hetero-HFS delivered after unpairing produce a 24 h fear memory for the opto-CS. Lastly, they show that synaptic potentiation accompanies the strengthening of fear memory induced by hetero HFS in freely moving mice.

The significance of the study lies in showing in vivo that plasticity induced at different times or different synapses than those engaged during learning can modify memory and the synaptic strength in a synaptic pathway related to that memory. While heterosynaptic and timing-dependent effects in synaptic plasticity have been described largely ex vivo on shorter time scales, the discovery of lasting behavioral effects on memory is novel.

A strength of the study is that it uses well-defined and elegant optogenetic manipulations of distinct neural pathways in awake-behaving mice combined with in vivo recordings, which allows the authors to directly manipulate and monitor synaptic strength and memory.

The conclusions of this paper are mostly well supported by the data, but there are some aspects that should be resolved:

1. The experimental design for assessing the effects of applying HFS 24 h after conditioning should be clarified, and it should be re-evaluated which experimental groups can be compared and how. The experimental schemes in Figs. 1 and 3 (and Fig. 4e and extended data 4a,b) show that one group of animals was subjected to retrieval in the test context at 24 h, then received HFS, which was then followed by a second retrieval session. With this design, it remains unclear what the HFS impacts when it is delivered between these two 24 h memory retrieval sessions. It would be nice to see these data parsed out in a clean experimental design for all experiments (in Figs 1, 3, and 4), that means 4 groups with different treatments that are all tested only once at 24 h, and the appropriate statistical tests (ANOVA). This would also avoid repeating data in different panels for different pairwise comparisons (Fig 1, Fig 3, Fig 4, and extended Fig 4).

2. The final experiment (Fig. 5a-c, extended data 5c) combines behavioral assessments with in vivo LFP recordings before and 24 h after hetero-HFS. While this experiment is demanding, it seems a bit underpowered and not well-controlled. It would be critical to know if LFPs change over 24 h in animals in which memory is not altered by HFS, and to see correlations between memory performance and LFP changes, as two animals displayed low freezing levels. Also, the slice experiments (Fig. 5d-f) are not well aligned with the in vivo experiments (juvenile animals, electrical vs. opto stimulation, different HFS protocols, timescale of hours). They would suggest that thalamo-LA potentiation occurs directly after learning+HFS (which could be tested) and is maintained over 24 h.

3. The statistical analyses need to be clarified. All statements should be supported with statistical testing (e.g. extended data 5c, pg 7 stats are missing). The specific tests should be clearly stated throughout. For ANOVAs, the post-hoc tests and their outcomes should be stated. In some cases, 2-way ANOVAs were performed, but it seems there is only one independent variable, calling for one-way ANOVA.

4. There are a number of details in the methods and procedures that need to be elaborated on and clarified for the reader. All of them will be listed in the recommendations to the authors.

Reviewer #3 (Public Review):

Summary:

In this study, Faress and colleagues investigated the differential contributions of Hebbian and non-hebbian plasticity to long-term memory. For this, the authors relied on in vivo optogenetic manipulations of thalamic (Th) and cortical (Ctx) inputs to the lateral amygdala (LA), a circuit whose role in associative memories is well established. The authors first begin by demonstrating that following a weak association protocol (also involving opto stimulation of the Th input) high-frequency stimulation (HFS) of the Th input induces robust conditioned responses (CR) 24 hours later. The authors then use two excitatory opsins to independently manipulate Ctx and Th inputs to the LA. They show that by delivering HFS of the Ctx input, LTP can be observed at Th-LA inputs which is accompanied by long-lasting memory effects.

Strengths:

Overall, the study addresses an important scientific question and could potentially result in a very valuable contribution to the field. The combination of in vivo electrophysiology with optogenetic manipulations of individual input sources to the LA is attractive.

Weaknesses:

While the methods employed in this study are attractive, they are also associated with major weaknesses. In particular, the manuscript lacks convincing validation and sufficient controls. Specific comments are included in the "Recommendations for the authors" section.

Author Response

We would like to thank you and the reviewers for evaluating this manuscript and providing constructive recommendations. Please see our provisional response to the major comments made by the reviewers.

Reviewer #1 (Public Review):

"…the authors never show that HFS of cortical inputs has no effect in the absence of thalamic stimulation. It appears that there is a citation showing this, but I think it would be important to show this in this study as well"

We understand that the reviewer would like us to induce an HFS protocol on cortical input and then test if there is any change in synaptic strength in thalamic input. We agree this is an important experiment which we will do.

Reviewer #2 (Public Review):

“…The experimental schemes in Figs. 1 and 3 (and Fig. 4e and extended data 4a,b) show that one group of animals was subjected to retrieval in the test context at 24 h, then received HFS, which was then followed by a second retrieval session. With this design, it remains unclear what the HFS impacts when it is delivered between these two 24 h memory retrieval sessions."

We understand that the reviewer has raised the concern that the increase in freezing we observed after the HFS protocol (ex. Fig. 1b, the bar labeled as Wth+24hHFSth) could be caused or modulated by the recall prior to the HFS (Fig. 1a, top branch).

If our interpretation of the concern is correct, we think this is unlikely to be the case. The first test, and the following HFS protocol, and the second test, (Fig. 1a, top branch) were all performed in the same chamber. For both the first and the second tests, animals received two 30-second recall trials, separated by 2 minutes (the data presented as the average of the two trials). We did not see a difference in freezing between the first and the second recall trials within each session (data not shown). It was only after the HFS protocol that we observed an increase in freezing.

This shows that in our paradigm the first recall does not impact the next recall in terms of the animals’ freezing levels. It must be noted that in cases where we did not do any testing prior to the HFS protocol, we still observed an increase in freezing after the HFS protocol (ex. Fig. 1a, middle branch and the corresponding data in Fig. 1b, the bar labeled as Wth+HFSth). Also, relevant is the data shown in Fig. 3c. Here, although animals were tested twice (Fig3. a, top branch), there was no increase in freezing in the second test (Fig. 3c, middle panel, Wth+24HFSCtx). That is, in the absence of an effective LTP, there is no significant difference between the two tests.

To further confirm this, in a new group of mice, 24 hours after weak conditioning, we will induce the HFS protocol, followed by testing (that is, no testing prior to the HFS protocol).

“The final experiment (Fig. 5a-c, extended data 5c) combines behavioral assessments with in vivo LFP recordings before and 24 h after hetero-HFS. While this experiment is demanding, it seems a bit underpowered”

We agree with the reviewer that the number of mice used in this experiment is on the lower side. However, this is not unusual for such an experimental configuration. As the reviewer mentioned, this is a demanding experiment for multiple reasons. For example, to confidently demonstrate that our HFS protocol, in addition to long-lasting behavioral changes, produces long-lasting synaptic changes, we must see a significant increase in evoked LFP after the manipulation which is predicted to last at least 24 hours. That is, the change in evoked LFP is not caused by non-related fluctuations, such as movement of the recording probe. For this reason, 3-4 days prior to conditioning, each day we measured evoked LFP. Only those mice that had a stable evoked LFP during this time were used for further conditioning. We will provide exclusion criteria for this experiment in the revised manuscript.

“ It would be critical to know if LFPs change over 24 h in animals in which memory is not altered by HFS,..”

We will perform an experiment where mice undergo a weak conditioning protocol and will record the evoked LFP 1-2 hours following the conditioning protocol, as well as the next day.

“…the slice experiments (Fig. 5d-f) are not well aligned with the in vivo experiments (juvenile animals, electrical vs. opto stimulation, different HFS protocols, timescale of hours).”

Our aim in this part was to demonstrate that the pathways we chose for our study can undergo heteroLTP. For this purpose, we used an already established protocol, which uses electrical stimulation (Fonseca, 2013). For clarification, I have tried to induce optical LTP with a high-frequency stimulation protocol in slices, but I did not succeed. I am not aware of a work that successfully induced optical LTP with a high-frequency protocol.