Sleep and memory consolidation are linked by RNA processing genes in the Drosophila mushroom body

Reviewer #2 (Public review):

Prior work by the Sehgal group has shown that a small group of neurons in the fly brain (anterior posterior (ap) α'β' mushroom body neurons (MBNs)) promote sleep and sleep-dependent appetitive memory specifically under fed conditions (Chouhan et al., (2021) Nature). Here, Li, Chouhan et al. combine cell-specific transcriptomics with measurements of sleep and memory to identify molecular processes underlying this phenomenon. They define transcriptional changes in ap α'β' MBNs and suggest a role for two genes downregulated following memory induction (Polr1F and Regnase-1) in regulating sleep and memory.

The transcriptional analyses in this manuscript are impressive. The authors have now included additional experiments that define acute and developmental roles for Polr1F and Regnase-1 respectively in regulating sleep. They have also provided additional data to strengthen their conclusion that Polr1F knockdown in α'β' mushroom body neurons enhances sleep.

The resubmitted work represents a convincing investigation of two novel sleep-regulatory proteins that may also play important roles in memory formation.

The authors have comprehensively addressed my comments, which I very much appreciate. I congratulate them on this excellent work.

Reviewer #3 (Public review):

Previous work (Chouhan et al., 2022) from the Sehgal group investigated the relationship between sleep and long-term memory formation by dissecting the role of mushroom body intrinsic neurons, extrinsic neurons, and output neurons during sleep-dependent and sleep-independent memory consolidation. In this manuscript, Li et al., profiled transcriptome in the anterior-posterior (ap) α'/β' neurons and identified genes that are differentially expressed after training in fed condition, which supports sleep-dependent memory formation. By knocking down candidate genes systematically, the authors identified Polr1F and Regnase-1 as two important hits that play potential roles in sleep and memory formation. What is the function of sleep and how to create a memory are two long-standing questions in science. The present study used a new approach to identify novel components that may link sleep and memory consolidation in a specific type of neuron. Importantly, these components implicated that RNA processing may play a role in these processes.

While I am enthusiastic about the innovative approach employed to identify RNA processing genes involved in sleep regulation and memory consolidation, I feel that the data presented in the manuscript is insufficient to support the claim that these two genes establish a definitive link between sleep and memory consolidation. First, the developmental role of Regnase-1 in reducing sleep remains unclear because knocking down Regnase-1 using the GeneSwitch system produced neither acute nor chronic sleep loss phenotype. In the revised manuscript, the author used the Gal80ts to restrict the knockdown of Regnase-1 in adult animals and concluded that Regnase-1 RNAi appears to affect sleep through development. Conducting overexpression experiments of Regnase-1 would lend some credibility to the phenotypes, however, this is not pursued in the revised manuscript. Second, while constitutive Regnase-1 knockdown produced robust phenotypes for both sleep-dependent and sleep-independent memory, it also led to a severe short-term memory phenotype. This raises the possibility that flies with constitutive Regnase-1 knockdown are poor learners, thereby having little memory to consolidate. The defect in learning could be simply caused by chronic sleep loss before training. Thus, this set of results does not substantiate a strong link between sleep and long-term memory consolidation. Lastly, the discussion on the sequential function of training, sleep, and RNA processing on memory consolidation appears speculative based on the present data.

Reviewer #4 (Public review):

Summary:

Li and Chouhan et al. follow up on a previous publication describing the role of anterior-posterior (ap) and medial (m) ɑ′/β′ Kenyon cells in mediating sleep-dependent and sleep-independent memory consolidation, respectively, based on feeding state in Drosophila melanogaster. The authors sequenced bulk RNA of ap ɑ′/β′ Kenyon cells 1h after flies were either trained-fed, trained-starved or untrained-fed and find a small number of genes (59) differentially expressed (3 upregulated, 56 downregulated) between trained-fed and trained-starved conditions. Many of these genes encode proteins involved in the regulation of gene expression. The authors then screened these differentially expressed genes for sleep phenotypes by expressing RNAi hairpins constitutively in ap ɑ′/β′ Kenyon cells and measuring sleep patterns. Two hits were selected for further analysis: Polr1F, which promoted sleep, and Regnase-1, which reduced sleep. The pan-neuronal expression of Polr1F and Regnase-1 RNAi constructs was then temporally restricted to adult flies using the GeneSwitch system. Polr1F sleep phenotypes were still observed, while Regnase-1 sleep phenotypes were not, indicating developmental defects. Appetitive memory was then assessed in flies with constitutive knockdown of Polr1F and Regnase-1 in ap ɑ′/β′ Kenyon cells. Polr1F knockdown did not affect sleep-dependent or sleep-independent memory, while Regnase-1 knockdown disrupted sleep-dependent memory, sleep-independent memory, as well as learning. Polr1F knockdown increased pre-ribosomal RNA transcripts in the brain, as measured by qPCR, in line with its predicted role as part of the RNA polymerase I complex. A puromycin incorporation assay to fluorescently label newly synthesized proteins also indicated higher levels of bulk translation upon Polr1F knockdown. Regnase-1 knockdown did not lead to observable changes in measurements of bulk translation.

Strengths:

The proposed involvement of RNA processing genes in regulating sleep and memory processes is interesting, and relatively unexplored. The methods are satisfactory.

Weaknesses:

The main weakness of the paper is in the overinterpretation of their results, particularly relating to the proposed link between sleep and memory consolidation, as stated in the title. Constitutive Polr1F knockdown in ap ɑ′/β′ Kenyon cells had no effect on appetitive long-term memory, while constitutive Regnase-1 knockdown affected both learning and memory. Since the effects of constitutive Regnase-1 knockdown on sleep could be attributed to developmental defects, it is quite plausible that these same developmental defects are what drive the observed learning and memory phenotypes. In this case, an alternative explanation of the authors' findings is that constitutive Regnase-1 knockdown disrupts the entire functioning of ap ɑ′/β′ Kenyon cells, and as a consequence behaviors involving these neurons (i.e. learning, memory and sleep) are disrupted. It will be important to provide further evidence of the function of RNA processing genes in memory in order to substantiate the memory link proposed by the authors.

Author response:

The following is the authors’ response to the original reviews.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Below, I will list the points that should be addressed by the authors:

(1) Line 139: The authors conclude that the lack of a phenotype induced by knockdown of Polr1F is due to reduced baseline sleep because of the leakiness of the Genswitch system. However, it is not clear why the argument of the SybGS being leaky should not apply to all experiments done with this tool. The authors should comment on that aspect. Furthermore, this claim is testable since it should be detectable against genetic controls. An alternative explanation to the proposed scenario is that the Polr1F sleep phenotype observed in the constitutive knockdown experiment is based on developmental defects. The authors should provide additional evidence to explain the discrepancy.

We appreciate the reviewer’s insightful feedback. We assume the reviewer is referring to Regnase-1 RNAi (and not Polr1F) as Regnase-1 RNAi flies exhibit reduced sleep before dusk, potentially hindering further detection of sleep reduction. The leaky sleep reduction was based upon comparison with genetic controls in that experiment. Nevertheless, to discern whether our observations stem from developmental effects, we conducted adult-specific knockdowns of both Polr1F and Regnase-1 using the TARGET system. We generated the R35B12-Gal4:TubGal80ts line and crossed it with the UAS-Polr1FRNAi and UAS-Regnase-1RNAi lines. We confirmed that Polr1F RNAi promotes sleep when knocked down in adults (Figure 3 - supplemental figure 1). Conversely, Regnase-1 showed no effect on sleep in the adult stage, which is consistent with our nSyb-GS experiments, and suggests, as noted by the reviewer, that the Regnase-1 RNAi sleep effect is likely developmental (Figure 3 – supplemental figure 3).

(2) Line 170: Regnase1 knockdown affects all memory types, including short-term and long-term memory. The authors conclude that these genes are involved in consolidation. However, besides consolidation, it has been shown that α′β′ KCs are involved in short-term appetitive memory retrieval. Thus, an equally possible explanation is that the knockdown impairs the neuronal function per se, which would lead to a defect in all behaviors related to α′β′ KCs, rather than a specific role for consolidation. The authors have to provide additional evidence to substantiate their claim.

The exact role of Regnase-1 in the α′β′ KCs remains unclear. We acknowledge the reviewer’s concern and have amended our conclusion to include this potential explanation suggested by the reviewer.

(3) Line 87-88: For the protocol used, it was reported that GFPnls cannot be used for FACS sorting. The authors might want to comment/clarify that aspect. https://star-protocols.cell.com/protocols/1669.

For our RNA-seq experiments, we conducted single cell isolation by FACS sorting cells, instead of nuclei, labeled with GFP.nls. The protocol mentioned that GFP.nls is not effective for single nuclear RNA-seq as it is not specific for nuclei, but for our cell sorting purposes that did not matter.

(4) Line 131: The authors should report the concentration of RU486.

Sorry, this is now in methods.

(5) Line 155: Is that really 42 hours? This might be a typo. If not, it would be good to justify the prolonged re-starvation period.

Flies fed after training form sleep-dependent memories but did not show robust long-term memory after 30 h of restarvation. As starvation is a requisite for appetitive memory retrieval (Krashes and Waddell 2008), the low memory scores after 30 h could be due to inadequate starvation. Therefore, we starved flies for 42h, which is similar to the sleep-independent memory paradigm in which flies are starved for 18 h before training and then tested 24 h after training; this protocol resulted in robust long-term memory performance. These flies were fine and able to make choices in a T-maze after 42 h starvation.

(6) I will be listing mistakes/unclear points in the figures. However, all figures should be checked very carefully for clarity.

Thanks for these valuable comments. We have gone over the figures carefully and fixed any issues we found.

(7) Figure 1C: It is not entirely clear to me how this heatmap was created and what the values mean.

The 59 differentially expressed genes (DEGs) were selected based on DESeq2 described in the methods. For the heatmap, Transcripts per million (TPM) of these 59 DEGs were log-transformed and then scaled row-wise and plotted with IDEP v0.95 (http://bioinformatics.sdstate.edu/idep95/).

(8) Figures 2A and 2B: The units might be missing. For Supplementary Figure 2, it is not clear what the different groups are without looking at the main figure.

Fixed.

(9) Figure 3: The panel arrangement is confusing. Furthermore, the "B)" is cut. The same issue is present in the Supplementary Figure.

Sorry! We rearranged the panels, and fixed the issue in both figures.

(10) Figure 5B: It is not clear what the scale bar means.

Now indicated

(11) Line 119: The citation "Marygold et al n.d."?

Fixed

(12) Line 620: I'm not sure that the rate and localization of nascent peptide synthesis are measured.

Great point. We used the puromycin assay to estimate significant changes in translation. However, we did not measure the absolute translational rate or the localization of newly synthesized proteins. We rephrased this in the updated manuscript.

(13) Line 627, the authors should give the NA of the objective, further the authors should double-check the information they provide on the resolution.

Fixed, it was 20X.

(14) Line 629 "Fuji" is unclear, it might refer to the Fiji software, and in that case, it should be listed in the used software. Further, the authors have to check on the information they provide on the intensity, e.g. is that GFP fluorescence?

Yes, it was Fiji and GFP. The manuscript has been updated accordingly.

(15) Line 634, It is stated that two concentrations of CX-5461 are used, however, as far as I can see only data for the 0.2 mM.

We apologize for the confusion. Data are indeed only shown for 0.2 mM. We also tested 0.4 mM and 0.6 mM under fed conditions once and 0.1 mM under starved conditions twice. Since all effects were not significant, we only presented the complete 0.2 mM results in the supplementary figure.

(16) Line 352 "Marygold et al nd" is probably a glitch in the citation?

It’s a citation tool issue and has been fixed.

(17) The authors use apostrophe rather than a prime in describing the α "prime" β "prime" KCs

We have corrected this.

Reviewer #2 (Recommendations For The Authors):

The authors have generated an interesting study that promises to advance the understanding of how context-dependent changes in sleep and memory are executed at the molecular level. The manuscript is well-written and the statistical analyses appear robust. Major and minor comments are detailed below.

Overall, I would suggest that the authors try to obtain additional evidence that Pol1rF modulates sleep and test the effect of acute adult-stage knockdown of Polr1F and Regnase-1 specifically in ap α'β' MBNs rather than pan-neuronally.

Major comments

(1) In Figures 2 and 3 and associated supplemental figures, the authors first test for a role for Polr1F and Regnase-1 specifically in ap α'β' MBNs (Fig. 2), then test for an acute role for these proteins via pan-neuronal drug inducible expression (Fig. 3). Because the former manipulation is cell-specific and the latter is pan-neuronal, it is hard for the reader to draw conclusions pertaining to ap α'β' MBNs from the second dataset. Perhaps Regnase-1 indeed acutely regulates sleep in ap α'β' MBNs, but that effect is masked by counteracting roles in other neurons? Conversely, it remains possible that Polr1F and Regnase-1 act during development in ap α'β' MBNs to modulate sleep. Indeed, since silencing the output of ap α'β' MBNs using temperature-sensitive shibire does not alter baseline sleep (Chouhan et al., (2021) Nature), the notion that Regnase-1 could act acutely in ap α'β' MBNs to reduce baseline sleep is somewhat surprising.

The authors could address this by using a method such as TARGET (temperature-sensitive GAL80) to acutely reduce Polr1F and Regnase-1 expression specifically in ap α'β' MBNs and test how this impacts sleep.

Thanks for the very helpful suggestions. We have done the suggested experiments and discuss them above in response to Reviewer 1. They are included in the manuscript as Figure 3 – supplemental figure 1 and figure 3 – supplemental figure 3.

(2) Figure 4 presents data examining whether Polr1F and Regnase-1 knockdown suppresses training-induced increases in sleep. For the untrained flies, based on the data in Fig. 2C, E I expected that Polr1F knockdown flies would exhibit more sleep than their respective controls (Fig. 4E), but this was not the case. These data suggest that more evidence may be warranted to strengthen the link between Polr1F (and potentially Regnase-1) knockdown and sleep. Could the authors use independent RNAi constructs or cell-specific CRISPR (all available from current stock centres) to validate their current results? Related to this, it would be useful to know whether the authors outcrossed any of their transgenic reagents into a defined genetic background.

The untrained flies in figure 4E are not equivalent to flies tested for Polr1F effects on sleep in figure 2C. In Figure 4E, flies were starved for 18 h and then exposed to sucrose without an odor at ZT6. Following sucrose exposure, flies were moved to sucrose locomotor tubes, and sleep was assessed only in the ZT8-12 interval. Sleep was not significantly different between untrained R35B12>Polr1FRNAi and Polr1FRNAi/+ flies, and while it was higher in R35B12>Polr1FRNAi than in R35B12/+ untrained flies, the data overall indicate that Polr1F downregulation has no impact on sleep under these conditions and at this time. Similarly, in fully satiated settings (Figure 2C), we found no difference in sleep during the ZT8-12 period between R35B12>Polr1FRNAi flies and genetic controls. We did not outcross our transgenic lines but have now tested another available Polr1F RNAi (VDRC: v103392) (Figure 3 – supplemental figure 1). As shown in the figure, adult-specific knockdown of Polr1F by this RNAi line promoted sleep, as did the initial RNAi line.

(3) Could the authors provide additional evidence that Polr1F knockdown in ap α'β' MBNs does not enhance sleep by reducing movement? A separate assay such as climbing would be beneficial. Alternatively, examining peak activity levels at dawn/dusk from the 12L: 12D DAM data.

We checked the peak activity per minute per day for adult specific knockdown of PorlF1 and Regnase-1 (data shown in Figure 3 – supplemental figure 4). The results show that Polr1F knockdown in ap α'β' MBNs does not enhance sleep by reducing movement.

(4) In terms of validating their proposed model, over-expressing of Polr1F during appetitive training might be predicted to suppress training-induced sleep increases and potentially long-term memory. Do the authors have any evidence for this?

We were unable to find any Pol1rF overexpression line. However, we obtained the Regnase-1 over-expression line from Dr. Ryuya Fukunaga’s lab and found that Regnase-1 OE does not affect sleep (Figure 4 – supplemental figure 1).

Minor comments

(1) Abstract: can the authors please define 'ap' as anterior posterior?

Fixed.

(2) Figure 2 Supplemental 1: can the authors please denote the genotypes each color refers to in?

Fixed.

(3) In Figure 3 Supplemental 1, the authors state that acute Regnase-1 knockdown did not reduce sleep, but sleep during the night period does appear to be reduced (panel A). Was this quantified?

We quantified this, and it was not significant.

(4) Discussion, line 234: the heading of this section is 'Polr1F regulates ribosome RNA synthesis and memory' but the data presented in Figure 4 suggests that Polr1F does not affect memory. Can the authors clarify this?

We made an adjustment to the title and acknowledge that at the present time we cannot say Polr1F affects memory.

(5) Methods, Key Resource Table: can the authors please identify which fly lines were used for Polr1F and Regnase-1 knockdown experiments?

Fixed. Fly line BDSC64553 was used for Polr1F RNAi except in Figure 3 – supplemental figure 1 and 4, where VDRC 103392 was used. VDRC 27330 was used for Regnase-1 knockdown experiments.

Reviewer #3 (Recommendations For The Authors):

(1) Figure 1B: This plot is currently labelled as PCA of DEGs, which I believe is inaccurate, as such a plot is a quality control that examines the overall clustering of samples by using all read counts (not just the DEGs). In addition, the color key value of this Figure 1B is not provided.

Thank you for the insightful suggestion. The reviewer’s comment here that typically PCA plots are used for overall clustering of RNA-seq samples is indeed valid. We've acknowledged that our samples, due to their high similarity in cell populations and mild treatments, do not exhibit clear separation when we use all genes. However, we show a PathwayPCA plot of all DEGs. We aim to highlight that RNA processing pathways enriched among the DEGs account for much of the separation of the groups.

(2) A reviewer token is not provided to examine the sequencing data set.

The RNA-seq data has been submitted to the Sequence Read Archive (SRA) with NCBI BioProject accession number PRJNA1132369. The reviewer token is https://dataview.ncbi.nlm.nih.gov/object/PRJNA1132369?reviewer=cvqkddp8rjuebsjefk0f19556r.

(3) In the discussion, the author pointed out that many of the 59 DEGs have implicated functions in RNA processing. To strengthen the statement, it would be beneficial to conduct the Gene Ontology analysis to test whether the DEGs are enriched for RNA processing-related GO terms.

We have included the GO analysis results in Figure1 and another GO analysis of all DEGs in Figure 1 – supplemental figure 1.

(4) Figure 4E presents an intriguing finding because it shows that the untrained R35B12>Polr1FRNAi flies exhibit reduced sleep (instead of increased sleep) when compared to untrained Polr1/+ control flies.

Please see above response to reviewer #2 question2.

(5) For the memory assay method, the identity of odor A and odor B is not provided.

We used 4-methylcyclohexanol and 3-octanol; this information has been added into the methods section.

(6) Female flies were used for the sleep assay. However, it is not clear whether only female flies were used for the memory assay.

Mixed sexes are used for memory assays because a huge number of files is needed for these experiments. We added this information in the methods.

(7) It is important to provide olfactory acuity data on control and experimental animals to rule out that the learning/memory phenotype is caused by defects in sensing the odor used for training and testing.

Since Polr1F RNAi flies perform well, odor acuity is not an issue. Regnase1RNAi affects both short-term and long-term memories, but this seems to be a developmental issue, so we did not do the odor acuity experiments here.

(8) Line 20: "ap alpha'/beta'" neurons should be spelled as "anterior posterior (ap) alpha'/beta' neurons", as this is the first time that this anatomical name appears in this manuscript.

Fixed.

(9) Figure 2C and 2D labelling: R35B12>control; UAS control should be changed to R35B12/+ control; UAS-RNAi/+ control.

Fixed.

(10) Line 155: it is unclear why the flies were re-starved for 42hr before testing. Is this a different protocol from the 30hr re-starvation that was used by Chouhan et al., 2021?

We have explained the rationale above. The starvation period was increased to get better memory scores.

(11) Line 160: it is stated that knocking down Polr1F did not affect memory, which is consistent with Polr1f levels typically decreasing during memory consolidation. Is there a reference demonstrating that Polr1f levels typically decrease during memory consolidation?

It’s from our RNA-seq dataset from Figure1C. The level of Polr1F decreased in fed trained flies compared with other control flies.

(12) Genotype labeling in Figure 4F is inconsistent with the rest of the manuscript.

Fixed.