Control of protein synthesis and memory by GluN3A-NMDA receptors through inhibition of GIT1/mTORC1 assembly

Essential revisions:

Reviewer #1:

1. The presentation of the RNAseq data is confusing. It is clear over-all that the effects of GluN3A are not at the level of transcription; however, the labels indicating various messages in Figure 1D are not effective because it is not clear which dots are indicated. The authors should eliminate all labels except those for Arc and c-Fos, perhaps adding arrows.

We have remodeled Figure 1D as suggested.

Figure 1D should be moved to one of the two figure 1 supplements.

2. In contrast, the data in Figure 2 supplement 1 are important for the overall conclusions. This figure could be added to the text as an additional figure.

We agree with the reviewer and figure 2 supplement 1 is now part of the main manuscript.

There are not limitations on the number of figures for an eLife paper.

3. Figure 4 supplement 1 should be added to Figure 4 as a quantitation of Figure 4 F.

Done as suggested.

6. The limitations of interpretation of the i.p injection of rapamycin into grin3a knockout mice should be clarified. Because mTORC1 is important for many fundamental cell processes, the effect on memory could be non-specific. Thus, its effect on memory cannot be isolated to synaptic effects of rapamycin.

We have moderated our claim as suggested. The original sentence: “The reversal by rapamycin strongly suggests” has been changed to “The reversal of rapamycin is consistent with the notion that the enhanced readiness of the mTORC1 translational machinery in GluN3A-deficient mice expands the range for consolidation of memory traces”. We have additionally expanded the Discussion to mention potential non-synaptic effects of rapamycin.

The Discussion contains some interesting ideas, but is difficult to read and needs clarification and reorganization. Several compound sentences should be divided or clarified.

We thank the reviewer for his careful reading and suggestions. We have included the changes indicated in points 6 to 20 to the Discussion section.

Reviewer #3:

General

22. One can say mRNA translation or protein synthesis but not translation. I believe that careful and articulate writing would help to improve the manuscript.

We agree with the reviewer and have corrected the instances where the incorrect term “protein translation” was used.

23. If they want to convey the message it's a protein synthesis dependent Long Term Memory we are talking about, I would emphasize the Fear Conditioning (1 hr versus 24 hr or more). The Conditioned Taste Aversion helps to show it's a border line effect that can be detected in weak but not strong paradigms.

We agree with the reviewer. In the behavioural section we now state that we used the two paradigms to assess associative memory formation, and emphasize how the fear conditioning allows to distinguish between unchanged short-term (1 h) and enhanced long-term (24h, 48h, 7 d) memory formation.

24. Injecting rapamycin i.p. can do many things, it is not obvious to connect the genetic manipulation to mTOR dependent effect but their data clearly points at this direction.

We have moderated our claim as suggested. The sentence that stated: “The reversal by rapamycin strongly suggests” has been changed to “The reversal by rapamycin is consistent with the notion that the enhanced readiness of memory traces”. We have additionally expanded the Discussion to address potential non-synaptic effects of rapamycin.

Introduction.

25. It is believed- "Brains are made of complex neuronal networks" (the brain is made up of more than neurons) and "memories are encoded by modifying the synaptic connections between them" (There may also be non-synaptic modifications that contribute to memory storage).

We have rewritten our beginning sentence. It is widely assumed that memory involves synaptic modifications, but we agree with the reviewer that non-synaptic modifications such as changes in neuronal excitability or metabolic states likely contribute to memory storage. Our purpose was to focus the Introduction in the subject of our study and we did not mean to exclude other mechanisms.

26. "Encoding" requires de novo mRNA and protein synthesis in response to neuronal activity and sensory experience" (Do you mean consolidation? You may have short term memory.)

We have changed the term “encoding” for “lasting encoding”.

27. "but eIF2α affects general mRNA translation and evidence for a role in local translation is lacking (Sharma et al., 2020; Shrestha et al., 2020b)." As a matter of fact, there is high correlation between proteome of eIf2alpha and learning (its not general as the authors claim). Please look at Sharma et al., 2020.

We have re-written this sentence. The current view is that elF2α phosphorylation, by blocking the action of the initiation factor elF2B, reduces overall levels of protein synthesis. Yet we agree with the reviewer that it is unclear whether elF2α de- phosphorylation might drive transcript-specific translation in settings such as learning (for instance as a function of the abundance of constitutively available or activity-induced mRNAs). The reviewer is also correct that Sharma et al., in their excellent 2020 study attempted to address this issue by mapping the translatome and proteome of CA1 hippocampus in mice with reduced elF2α phosphorylation (see also Eacker et al., BioxRiv https://doi.org/10.1101/169425). Both studies identified different sets of translational changes that resembled those induced by learning, suggesting that the phosphor-status of elF2α might be a significant contributor following a memory experience.

28. Long and not clear sentence- "Negative regulation is mediated by inhibition of the assembly of mTORC1 complexes that contain the postsynaptic adaptor GIT1 (G protein-coupled receptor kinase-interacting protein) and Raptor, are located at or near synaptic sites, and couple mTORC1 kinase activity to synaptic stimulation. It would be better to break the three clauses up into sentences.

The sentence has been re-written as suggested.

Results

29. Figure 1e representative, MG 132 seems to induce expression in Zif268 and arc as measured by WB?

The reviewer is correct. MG132 potentiates the BDNF induction of Arc and Zif268 by inhibiting the proteosomal degradation that contributes to terminate the IEG (compare BDNF alone vs BDNF + MG132) (see Rao and Finkbeiner, Nature Neuroscience 9,887, 2006; or Mabb et al., Neuron 82, 1299, 2014).

30. Figure 6. "These results showed that GluN3A deletion facilitates spatial learning and memory without the unwanted effects associated to other modulators of translation." I do not understand what these sentence mean and what these results mean? there is no learning curve for the WT? if you do repeated measure after the reversal there might be a difference in the curve, what does it mean? There is no difference on latency after 7 days but statistical difference in the probe test? What does it mean?

Two-trial per day Morris water maze experiments as the one in Figure 6 (now Figure 7; another example in females can be found in the Figure 7 – supplement figure 1B) often do not show a clear learning curve and latency averages are noisy when compared to the classical 4 trails per day (Figure 7 —figure supplements 1A). However, mice eventually learn the platform position as evident in the probe test conducted at day 7.

The sentence summarizing the first set of behavioral testing meant to emphasize differences between the behavioral outcome of GluN3A deletion with the observed after other manipulations of translation. For instance, GluN3A deletion leads to enhanced spatial memory in the Morris water maze without alterations in memory flexibility in the same test or signs of perseveration in the alternation test. By contrast: (i) reducing phosphorylation of elF2α in lateral amygdala enhances memory strength but reduces behavioral flexibility (Strestha et al., Nature Neuroscience 2020; Trinh et al., Cell Reports 2012); (ii) genetic removal of the translation modulator 4EBP impairs T-maze spontaneous alternation (Banko et al., Neurobiology of learning and memory 2007); (iii) enhancing mTOR signaling via brain disruption of FKBP12 enhances contextual fear memory but compromises behavioural flexibility in the Morris water maze and other tasks (Hoeffer et al., Neuron 2008). Our current plan is to complete the behavioural characterization of Grin3a KO mice and see if other domains of memory of cognition are compromised.

31. "We then assessed long-term memory formation". 24 hrs. between trial in the Morris Water Maze (MWM) do not consider Long Term Memory?

We agree with the reviewer that 24 hours is considered long-term memory. However, Morris water maze testing involves cumulative training with several trials per day and is not normally used to assess long-term memory. To avoid misunderstanding, we have changed the sentence to: “We then assessed associative memory formation”.

32. Rapamycin erased the weak Conditioned Taste Aversion (CTA) memory in Grin3a-/- mice (Figure 6H), supporting the notion that disinhibited mTOR signaling causes the cognitive enhancement. Would it be the case if anisomycin or other Protein Synthesis Inhibitors will not have an effect?

We did not perform this experiment, but we agree that if would support our claim and we will include in our next round of experiments as Grin3a KO mice become available.

33. Also Injecting an inhibitor i.p. (Rapamycin) in CTA is an issue since you inject twice and it may affect behavior in different ways.

We agree, and have moderated our claims as suggested (see response to point 30).

34. As for the extinction, what is the p value for day1 Fear Extinction? This is also a memory test and if there are differences, it should be measured.

For the results reported in the manuscript, we used a rmANOVA using mean freezing during each FE session (Fes) as within-subject factor and group (Grin3a -/- Wild-type) as between-subject factor. Here, we found an effect of fear extinction sessions, meaning that animals were successfully extinguishing fear throughout time (F(3,75)=84.035, p<0.001), but without differences between groups (tx*Fes) (F(3,75)=1.402, p=0.254). A between-subject (F(1,25)=4.402, p=0.046) showed that Grin3a -/- animals had overall lower freezing rates compared to Wild-type animals.

As the interaction tx*Fes was not significant, we proceeded to analyze freezing in each FE session individually. Freezing levels were not different among treatments in FE1 (t(25)=0.760, p=0.455) or FE2 (t(25)=1.948, p=0.063), but Grin3a -/- mice had significantly less freezing compared to Wild-type in FE3 (t(25)-2.127, p=0.043) and FE4 (t(25)=2.108, p=0.045). The information on FE day 1 requested by the reviewer has now been included in the figure legend.

Given your observation, we considered important to carry out additional analysis to explore if there were memory differences within each FE session. This would inform if the differences were arising from FE session performance or memory consolidation throughout the days. We did this by averaging freezing during 5 CS presentations for each FE session (CS 1 to 5, CS 6 to 10, CS 11 to 15) and used them as within-subject factors in rmANOVA analysis. Within-session freezing levels were similar among groups in FE1 (F(2,50)=1.218, p=0.304), FE2 (F(2,50)=0.120, p=0.887), FE3 (F(2,50)=0.645, p0.529) and (FE4 (2,50)=0.988,p=0.376), meaning that animals extinguished fear at similar rates in each FE session independently of their genotype.

Discussion

Physiological relevance, Do you think GluN3A is a removable constraint on memory in the adult brain? Or is it a way to inhibit plasticity?

1) As we mention in the Discussion, GluN3A is developmentally regulated but also retained in specific brain areas and subsets of synapses. Furthermore, neurons are endowed with specific mechanisms for local, activity-dependant removal from synapses. These properties make GluN3A a suitable candidate for a removable memory constraint.

2) We have considered the alternative possibility, that GluN3A synapses would be a conduit for short-term memories while GluN3A-lacking synapses would encode long-lasting memories. Discerning between these and other possibilities will require further work.

"for spine and memory consolidation" – there is no spine consolidation, what do you mean?

We have changed the wording of this sentence.