View all hidden authors and organizations

Edited by Joseph S. Takahashi of the University of Texas Southwestern Medical Center, approved on October 27, 2021 (review received on May 6, 2021)

Lack of sleep destroys the consolidation of memories that depend on the hippocampus. To understand the cellular basis of this effect, we quantified the RNA associated with the translation of ribosomes in the cytoplasm and cell membranes of different hippocampal neuron populations. Our analysis shows that although lack of sleep (but not learning) changes many ribosomal transcripts in the cytoplasm, learning has a significant impact on the less well-defined transcription profile of membrane-bound ribosomes. We show that post-learning sleep deprivation has minimized learning-driven changes in the cytosolic ribosome. It also changes the transcripts associated with metabolism and biosynthesis in membrane-bound ribosomes in excitatory hippocampal neurons and highly active putative "imprinted" neurons, respectively. Taken together, these findings provide insights into understanding the cellular mechanisms that are altered by learning and the subsequent damage caused by lack of sleep.

The hippocampus is essential for integrating short-lived experiences into lasting memories. Sleep after learning promotes memory consolidation, although the underlying cellular mechanism is largely unknown. We have adopted a fair approach to this problem by using a mouse model of hippocampal-mediated sleep-dependent memory consolidation (contextual fear memory). Because synaptic plasticity is related to changes in neuronal cell membranes (such as receptors) and cytoplasm (such as cytoskeletal elements), we describe how these cell compartments are affected by learning and subsequent sleep or sleep deprivation (SD). Translation ribosomal affinity purification is used to analyze different subcellular compartments (cytoplasm and cell membrane) and different cell populations (whole hippocampus, Camk2a+ neurons, or highly active neurons with phosphorylated ribosomal subunit S6 [pS6+]). Ribosome-associated RNA. We studied how the transcription profile changes with sleep and SD and prior learning (situational fear conditioning; CFC). Although lack of sleep changed many cytoplasmic ribosomal transcripts, CFC hardly changed, and CFC-driven changes were blocked by subsequent SD. In sharp contrast, SD changed a few transcripts on membrane-bound (MB) ribosomes, while learning changed even more (including long non-coding RNA [lncRNA]). The cellular pathways most affected by CFC are involved in structural remodeling. Comparison of MB transcript profiles after CFC between sleep and SD mice showed that the cellular metabolism of Camk2a+ neurons and the changes in protein synthesis in highly active pS6+ (putative “imprints”) neurons, because biological processes were disrupted by SD. These findings provide insights into how learning affects hippocampal neurons and indicate that SD's effects on memory consolidation are specific to cell types and subcellular compartments.

The role of sleep in promoting synaptic plasticity and memory consolidation is an eternal mystery (1). In the past two decades, transcriptomics (2⇓ ⇓ ⇓ –6) and proteomics (7⇓ ⇓ –10) analyses of mammalian brains after sleep and sleep deprivation (SD) have provided information on general brain sleep functions Insights. These works provide a preliminary basis for the hypothesis of synaptic homeostasis (SHY) of sleep function (2), which proposes that synapses are widely "shrinked" during sleep. However, the role of this process in memory consolidation, and its occurrence during sleep after learning, remains a controversial issue (1, 11). The conclusion of the in vivo electrophysiological observation is that the specific activity pattern that occurs during the learning process may be repeated in the subsequent sleep. This mechanism may have instructive significance for memory storage [ie, by selectively reactivating the “imprinted neurons” (12) that participated in previous learning and promoting the synaptic plasticity of these neurons]. In fact, recent research results indicate that in the sensory cortex and other structures, the reactivation of memory imprinted neurons plays a crucial role in consolidating new memories in the first few hours of sleep after learning (12). However, while recent work has demonstrated long-term transcriptional changes in imprinted neurons (13), how sleep-related reactivation of these neurons (and other features of sleep-related brain physiology) affects intracellular synaptic plasticity The access is still unknown. 1, 14).

Recently, transcriptomics and proteomics analysis of synapses or axon organelles have been used to better understand the effects of learning (15) or sleep and wakefulness (10) on synaptic function. However, to date, there has been no experimental work aimed at characterizing cellular changes during sleep-dependent memory consolidation-that is, cellular changes that occur during sleep after learning. Here, we use a complete mouse model of sleep-dependent memory consolidation—contextual fear memory (CFM)—to study this process. CFM can be coded in a learning experiment (Situational Fear Conditioning [CFC]; placed in a new environmental chamber and then given a foot shock). In the next few hours, this contextual shock paired memory was consolidated through the hippocampus-dependent mechanism (16). Crucially, within the first 5 to 6 hours after CFC, SD may disrupt CFM integration (17⇓ ⇓ ⇓ –21). During the same post-CFC interval, the destruction of neuronal activity (16), transcription (22, 23) or translation (24⇓ ⇓ –27) in the dorsal hippocampus will also disrupt consolidation. This suggests that activities and sleep-dependent mechanisms affecting hippocampal biosynthetic pathways are essential for consolidation.

To clarify these mechanisms, we characterized ribosomal-associated messenger RNA (mRNA) from different hippocampal cell populations [including all Camk2a+ excitatory neurons and a subset of highly active hippocampal neurons expressing Ser244/247 phosphorylated S6 (pS6+) Changes (28, 29)] as a function of sleep and SD and previous CFC. By quantifying the mRNA profiles on ribosomes that are differentially located in the cytoplasm and cell membranes, we found that although most of the changes in transcripts on the cytoplasmic ribosomes vary with sleep and wakefulness, most of the membrane-bound (MB) ribosomes The transcript changes as a function of learning. Our research results reveal the subcellular functions of sleep after learning, and propose a cellular mechanism by which sleep can selectively promote memory storage.

To quantify the effects of sleep and learning on hippocampal mRNA translation, we adopted two translation ribosomal affinity purification (TRAP) technologies. First, in order to quantify the ribosome-related mRNA in excitatory neurons, the B6.Cg-Tg(Camk2a-cre)T29-1Stl/J mouse and the B6N.129-Rpl22tm1.1Psam/J mouse line (30, 31) Hybridization. The offspring of this crossover express hemagglutinin (HA)-labeled ribosomal protein 22 (HA-Rpl22) in excitatory (Camk2a+) neurons (Figure 1A, left). Second, in order to quantify the mRNA associated with ribosomes in active hippocampal neurons, we used an antibody against the phosphorylation site (Ser244/247) of the ribosomal protein S6 (pS6) (28) (Figure 1A, left) . Due to high neuronal activity, these sites are phosphorylated in neurons by ERK (32), mTOR-dependent kinases S6K1/2 (33) and CK1 (34), leading to translation initiation (34) and possible selection of transcripts Translation (33). This strategy allows us to compare mRNAs expressed in the entire hippocampus (input) with mRNAs associated with ribosomes in a population of Camk2a+ or highly active (pS6+) neurons from the same hippocampus. To further test how mRNA translation changes as a function of ribosomal subcellular localization, we centrifuged the homogenized hippocampal tissue and collected it from the supernatant (presumed cytosolic) and pellet (presumed membrane-containing) fractions Sample (35). From these two parts, we compared the whole hippocampal (input) transcript with the transcripts isolated from excitatory neurons (Camk2a+) and higher activity (pS6+) neuron populations by TRAP (Figure 1A).

Trap-based analysis of hippocampal cell population and separation of subcellular components. (A, left) Confocal image showing the expression of HA (Camk2a), phosphorylated S6 (pS6) and paralbumin in the CA1 region of the dorsal hippocampus. The highlighted neurons are paralbumin, pS6 and HA. (Scale bar, 100 μm.) (Right) A schematic diagram of a protocol for the isolation of mRNA from subcellular components and different cell populations using TRAP. (B) Camk2a+ (cyan) and pS6+ (orange) TRAP mRNA enrichment values ​​(relative to input) are calculated for activity-dependent (Arc, Cfos, Homer1a), excitatory neurons (Glua1, Vglut1), inhibitory Neurons (Parv, Sst), and glial (Mbp, Gfap) transcripts. *** means the difference in enrichment value between Camk2a+ and pS6+ neuron population P <0.001 (t test, n = 7/group). (C) Encoding secreted (Bdnf), transmembrane (Grin2a, Grin2b), ER (Hspa5) and cytoplasmic (Cfos, Homer1a) proteins. (T test, n = 9/group, *, ** and *** indicate P <0.05, P <0.01 and P <0.001, respectively). (D) PCA plots (variance stable transformation [VST], Deseq2) of RNA-seq data from three cell populations (input, Camk2a+ neurons and pS6+ neurons) and two parts (supernatant and pellet). Data for n = 30 biological replicates (ie, bilateral hippocampus) from 30 mice in the four treatment groups are shown.

We first verified the enrichment of cell type-specific transcripts from Camk2a+ and pS6+ cell populations using qPCR. Consistent with recent data (36), Camk2a+ TRAP produced similar mRNA levels of Arc, Cfos, Homer1a, Glua1, and Vglut1 transcripts relative to Input mRNA, and reduced the expression of interneurons and glial cell markers . Given the hemizygosity of HA-Rpl22, this is consistent with the approximately two-fold enrichment of excitatory markers relative to the input. Compared with Camk2a+ neurons, the mRNA profile of highly active pS6+ neurons showed more activity-driven transcripts (Arc, Cfos, and Homer1a) and interneuron-specific transcripts (Pvalb and Sst), and excitatory nerves The levels of meta-specific mRNAs (Glua1 and Vglut1) are comparable (Figure 1A and B). These data support the explanation that pS6 is present in the highly active subset of Camk2a+ neurons in the hippocampus and selected populations of interneurons (20).

We next used qPCR to initially characterize the subcellular enrichment of mRNA differentially expressed in the supernatant and pellet fraction. A previous report found that isolated granular ribosomes can selectively enrich endoplasmic reticulum (ER) and dendritic-localized transcripts (35). To test whether the transcripts encoding ER and dendritic proteins are different in the supernatant and pellets, we first analyzed Hspa5 (encoding the resident ER chaperone BIP). The content of Hspa5 in the granular fraction was significantly higher than the supernatant fraction of Camk2a+ and pS6+ neurons (Figure 1C; Camk2a+: supernatant, 1.03 × Input; particles, 7.38 × Input; P <0.001, t test; pS6+: upper Clear, 1.17 × input; particles, 10.33 × input; P <0.001). Similarly, compared to the supernatant fraction, Bdnf, Grin2a and Grin2b mRNA (encoding secreted growth factors and glutamatergic receptor subunits) were more enriched on ribosomes isolated from the granule fraction. In contrast, Homer1a [encoding a truncated form of the synaptic scaffolding protein Homer1 whose transcript was originally translated in somatic cells (37, 38)] is more pronounced on the supernatant ribosomes of two neuronal populations. Abundant and immediately early gene transcript Cfos content is also abundant in two parts. These results support the view that the proteins encoded by ribosome-related transcripts observed in the particles and supernatant fractions are enriched in the cell membrane and cytosol, respectively. To further verify this explanation, we used an unbiased method-RNA sequencing (RNA-seq) to further characterize the transcription profiles from different cell populations (Camk2a+, pS6+, input) and subcellular fractions (supernatant, pellet). The biological replicates used for this analysis consisted of biological hippocampus from individual mice (ie, one mouse/sample; a total of n=30 mice across 4 treatment groups as described in the results section below). Principal component analysis (PCA) analysis revealed six discrete clusters of mRNA expression profiles based on RNA-seq, of which PC1 accounted for 83% of the total expression variance (PC2 only contributed 6%). These clusters were separated according to the source of the sample (Figure 1D), and the supernatant and sedimentation parts each formed similar groupings of three different clusters. It is very important that there are three clusters in the data of the supernatant and the particles; these two groups of three clusters show the same correlation in the PC space and are performed according to the cell type (Camk2a+, pS6+ or input) Separate. There are no obvious subpopulations in this PC space (ie, representing the four treatment groups of mice; see the results section below).

In order to characterize the transcripts that are differentially localized in the supernatant and precipitated parts, we next calculated the relative mRNA abundances of the two parts from Camk2a+ neurons, pS6+ neurons, and input (ie, the entire hippocampus) (SI Appendix, Figure S1 A and B and S2A) use Deseq2 (39). Similarly, for this analysis, the hippocampus of each mouse (n = 30 samples from 30 mice under 4 behavioral conditions) constituted a biological replicate. The 2,000 transcripts (based on adjusted P-values) with the largest differences between the parts were then characterized using the cellular component annotation database (40) for annotation, visualization, and integrated discovery (DAVID). The transcripts from the supernatant part of the neuron population (and input) are more abundant, and the function of the encoded protein is located in the cytoplasm and nucleus. The more abundant transcripts in the sediment encode proteins with functions that are localized in the plasma membrane, ER, Golgi apparatus, and synapses (SI appendix, Figure S1 C and D, and S2B). Therefore, for the subsequent analysis, we refer to the supernatant and granule ribosomes as cytoplasmic ribosomes and MB ribosomes, respectively. The Ingenuity Pathway Analysis (IPA) classical pathway (SI appendix, Figure S1 E and F) was used to evaluate signal transduction and metabolic pathways, the components of which are highly represented in the more abundant mRNA on the cytoplasmic ribosome. These include cytosol-localized cellular functional pathways, such as ubiquitination, nucleotide excision repair, hypoxia signaling, and sumoylation pathways. In contrast, transcripts on MB ribosomes are more abundant, rich in synapses (GABAergic receptors, glutamatergic receptors, and endocannabinoid signaling) and endoplasmic reticulum (such as unfolded protein response). ) Components of functional signaling pathways.

In order to further study the subcellular localization of transcripts representing cellular pathways closely related to hippocampal function, we examined the signaling pathways represented by transcripts selectively expressed in the cytoplasm and MB fraction. Because both learning and sleep affect the structure and function of synapses (10, 41⇓ –43), we first pay attention to the signal pathway of synapse occurrence. In Camk2a+ neurons, the more abundant transcripts on MB ribosomes encode secreted proteins (such as Bdnf), transmembrane proteins, including AMPA, NMDA, and ephrin receptors (such as Gria1, Gria2, Gria3, Grin2a, Grin2b, Grin2c, Epha1 And Epha2), and membrane-associated enzymes (Plcγ) (SI appendix, Figure S1G). The more abundant mRNA on the cytosolic ribosome encodes intracellular complexes, including adaptor proteins (Crk, Shc) and kinases (Cdk5, Lmk1, Gsk3b, Mapk1, Mapk2, and P70S6K) in the synaptic pathway. The components of the CREB ​​signal pathway (another pathway regulated by learning and sleep) (18, 44⇓ ⇓ –47) are also highly representative in the transcript. In the Camk2a+ and pS6+ neuron groups, they are respectively Localized in cytoplasmic ribosomes and MB ribosomes (SI appendix, Figure S1 E and F). The mRNA encoding the enzyme that regulates the transcriptional activity of CREB is selectively localized to the ribosome in any compartment (for example, Polr2c, which encodes the RNA polymerase subunit in the cytoplasmic part; Adcy1, which encodes adenosine in the MB part) Acid cyclase). The mRNA encoding G protein-coupled receptors and ion channels (metabotropic glutamate receptor, ER IP3 receptor, calcium channel subunits, AMPA and NMDA receptor subunits) are only located in the MB part, while encoding transcription factors and Those of kinases are mainly located in the cytoplasmic part (SI appendix, Figure S1H). Contrary to Camk2a+ and pS6+ neuron populations, CREB ​​signaling pathway is not present in mRNAs that are differentially located in the input subcellular part (ie the entire hippocampus; SI appendix, Figure S2), indicating that the mRNA encoding CREB ​​signal components The difference in localization is more obvious in neurons than in other hippocampal cell types.

The functional categories represented in the two subcellular fractions in the input RNA follow a pattern roughly similar to that seen in the Camk2a+ and pS6+ neuron populations. However, contrary to the profile from the neuron population, the most representative signal pathway category for mRNAs that are differentially located between the two input parts is the protein ubiquitination pathway (SI Appendix, Figure S2). This finding indicates that in non-neuronal hippocampal cell types (that is, glial cells), there is a more significant subcellular separation of mRNAs encoding components of the ubiquitin pathway.

Because both learning and sleep change hippocampal activity, intracellular signaling and function (18, 19, 48, 49), we next tested how the cytosolic and MB ribosome-related transcripts in different neuron types were previously dependent on hippocampus The impact of sexual memory training tasks. Consistent with previous findings in our laboratory and other laboratories (17⇓ ⇓ ⇓-21), SD after CFC disrupted the CFM integration in the transgenic mice used in this study (SI Appendix, Figure S3). We also tested how these transcription profiles are affected by short (3 hours) subsequent sleep or SD. The 3-hour time window after CFC was selected to correspond to the peak changes in hippocampal network activity induced by CFC post-learning (20, 49, 50), and the time course of protein synthesis required for CFM to consolidate (26, 27). When the light is on (Zeitgeber time [ZT] 0; the beginning of the resting phase), the mice are either left undisturbed in their cage (HC) or undergo a single test CFC. For the next 3 hours, mice in the CFC and HC control groups were either allowed to sleep at will (sleep) or undergo SD in their HC by gentle handling (Figure 2A). Free-sleeping HC + Sleep and CFC + Sleep mice spent 76 ± 2% and 73 ± 4% of their time sleeping in the 3-hour window before euthanasia, respectively. This is consistent with our previous data, showing that about 60% to 75% of the first 6 hours after CFC is used for non-rapid eye movement (non-REM) sleep, and about 3% to 7% is used for REM sleep (19, 49, 50). These operations are followed by RNA isolation and sequencing, as described in the previous results section. Quantify the effects of learning and lack of sleep on mRNA abundance for each cell population (Camk2a+, pS6+ or input) and subcellular parts (cytoplasm or MB; for example, pS6+ MB) in order to preserve the gene level inferences made by the Deseq2 model.

Cytoplasmic ribosomal transcripts are mainly changed by SD, while MB ribosomal transcripts are mainly changed by learning. (A, left) An example of an RNA-seq experiment, showing four treatment groups. The n value represents the number of mice in each group; the bilateral hippocampus of a mouse constitute a biological repeat. When the light is on, the mice will either stay in their HC or undergo a single CFC test. Then, all mice were either allowed to sleep at will or maintained SD for the next 3 hours. (Right) Transcription comparison used to quantify the effects of SD (yellow) includes HC and CFC animals. To quantify the effects of CFC, CFC + sleep (blue) and CFC + SD (red), mice were analyzed separately. After the behavioral manipulation, the cytoplasm and MB parts of different cell groups are separated, as shown in Figure 1. (B) The proportional Venn diagram reflects the number of significantly changed transcripts in each cell population and subcellular fraction (ie Camk2a+/MB), based on the comparison shown in A. Data sets S4, S5, and S6 provide a complete list of transcripts for each comparison. (C) Transcripts changed by SD and CFC in different subcellular components of Camk2a+, pS6+ and Input populations were used to construct Venn diagrams. The complete transcription list is included in the data set S7.

We first assessed the specific effects of SD alone (compare SD and sleep conditions) by combining data sets from naive (HC) and recently trained (CFC) mice (Figure 2A, yellow). Learning and new sensory experiences during SD are usually warnings that recognize brain biochemical changes due to arousal itself (11). Therefore, we pooled samples from the HC and CFC groups for this specific part of the Deseq2 analysis, with the goal of more reliably identifying the most robust transcript changes driven by SD, while ignoring changes that may have occurred mainly as a learning function (11 ). Then, we quantified the learning effects of sleep and SD mice (compare CFC and HC conditions) (Figure 2A, red [SD], blue [sleep]). Comparing the latter two sets of transcript changes is useful to identify those changes that occur during successful CFM consolidation (during sleep) and those that occur in the case of consolidation failure (SD). The Venn diagram (shown in Figure 2B) illustrates the relative number of significant transcript changes resulting from these comparisons. Compared with the effect of learning, SD has a relatively greater impact on cytosolic ribosomal mRNA (Camk2a+: 567 transcripts; pS6+: 913 transcripts; input: 297 transcripts), the latter on cytosolic ribosomal transcription Ben’s influence is extremely mild. In contrast, MB ribosomal mRNA passes through SD (CFC + SD: Camk2a+, 2,396 transcripts; pS6+, 1,908 transcripts; input, 840 transcripts) and sleep group (CFC + Sleep: Camk2a+, 795 transcripts; pS6+, 2,211 transcripts) were significantly changed; input, 208 transcripts). In contrast, relatively few MB ribosomal mRNAs were altered by SD (Camk2a+: 233 transcripts; pS6+: 164 transcripts; input: 95 transcripts). These results indicate that SD and learning have different effects on ribosomal mRNA profiles according to their subcellular localization. SD (as opposed to sleep) seems to have a more pronounced effect in the cytoplasm (ie, significantly affects a large number of transcripts), and learning appears to have a more pronounced effect on MB ribosomes. In addition, transcripts that are significantly affected by sleep or learning have relatively little overlap between different cell populations, as shown in Figure 2C. Together, these data indicate that the effects of learning and sleep on ribosome-related transcripts have a high degree of cell type and subcellular compartment specificity.

In Camk2a+ neurons (SD: 567 transcripts, CFC: 20 transcripts), pS6+ neurons (SD: 913 transcripts, CFC: 43 transcripts), SD (relatively less driven by CFC) Compared with the change) significantly changed more cytosolic ribosome-associated mRNA. , And to a lesser extent, input (whole hippocampus) mRNA (SD: 297 transcripts, CFC: 37 transcripts) (Figure 3 and SI appendix, Figure S4). We have characterized the molecular and cellular pathways that are altered by SD-induced changes in cytosolic ribosomal transcripts. Only the molecular functions most affected by SD (adjusted P value <0.1 based on transcripts annotated with IPA) are overwhelmingly beneficial to Camk2a+ and pS6+ neurons, as well as transcriptional regulation and RNA processing in input (Figure 3B, top). Previous transcriptome analysis showed that mRNA encoding transcriptional regulators after transient SD in the hippocampus (Fos, Elk1, Nr4a1, Creb, and Crem1) (4⇓ –6, 36) and neocortex (Per2, Egr1, and Nr4a1) More abundant (2, 36). Consistent with these findings, SD increased the abundance of multiple mRNAs encoding transcription factors and upstream regulators in all cell populations, including E2f6, Elk1, Erf, Fosl1, Fosl2, Fos, Fosb, Lmo4, Taf12, Xbp1, Atf7, Artn12, Atoh8, Bhlhe40 (Dec1), Crebl2, Crem, Egr2, Nfil3 and Ubp1. The transcripts significantly affected in Camk2a+ and pS6+ neurons overlap with those reported in previous SD experiments, including AMPK, PDGF, ERK/MAPK, IGF1, and components of the ER stress signaling pathway (Figure 3B, bottom) (4 ⇓ –6, 24, 51). However, it has recently been reported that only a small part of the mRNA in the whole hippocampus (5), the ribosome-associated spectrum of hippocampal Camk2a+ neurons (6), or the whole forebrain (10) has an impact with SD in the cytoplasm partially overlaps any cell population (SI Appendix, Figure S5). Only in pS6+ neurons, mRNAs encoding components of the PI3K/AKT and TGF-B signaling pathways were down-regulated in the cytoplasmic part after SD (Figure 3B, bottom), indicating that the activity in these pathways may be higher during sleep.

SD mRNA changes on the cytoplasmic ribosomes encode transcriptional regulators. (A) The proportions and overlapping Venn diagrams of transcripts with significantly changed SD, CFC + Sleep, and CFC + SD in the cytoplasmic part of Camk2a+ neurons, pS6+ neurons, and Input. (B, top) The seven most abundant molecular and cellular functional categories (sorted by Padj value) of transcripts that are individually changed by SD in Camk2a+ neurons, pS6+ neurons, and Input. (Bottom) The 10 most abundant classical pathways of transcripts affected by SD are listed by Padj value (represented by the diameter of the circle), and the z-score reflects the direction of pathway regulation (represented by hue). There is no obvious classical approach in the input part. (C, left) Camk2a+ (top) and pS6+ (bottom) neurons are the 10 most significantly affected transcripts under CFC + Sleep (blue) and CFC + SD (red) conditions, sorted by Padj value. Transcripts that have also changed significantly as a function of the SD alone are highlighted in yellow. (Right) The result of transcription level analysis (56) shows transcripts with altered transcription subtypes in Camk2a+ (top) and pS6+ (bottom) neurons after CFC. Transcription subtypes that have changed significantly as a function of SD are highlighted in yellow. An analysis of the functional categories available in the data set S8.

We next performed upstream regulator analysis to characterize transcript changes, which may be due to SD-related transcriptional regulation. The results of this analysis provide a P value that indicates the importance of a specific common upstream regulator in the regulation of mRNA, and a z-score that indicates the direction of the fold change (ie transcriptional activation or repression) of the regulated mRNA. Taken together, these values ​​predict the activation state of specific gene regulatory complexes during SD (52). Consistent with previous meta-analysis of SD-induced transcripts (53), Creb1 was identified as a transcriptional regulator, and its downstream target was most consistently affected of all cytoplasmic transcripts (SI Appendix, Figure S6A). The Creb1 transcript itself did not increase after SD, although both Creb2 and Crem mRNA were significantly increased, and multiple Creb1 transcription targets (such as Fos, Arc, Fosb, Egr2, Nfil3, Nr4a2, Bag3, and Irs2) Camk2a+ and pS6+ nerves after SD The cytoplasmic score of the metagroup (SI appendix, Figures S6B and S7).

The consolidation of fear memory in the hippocampus within a few hours after CFC depends on sleep (17⇓ –19) and CREB ​​mediated transcription (13, 54). Due to the high activity of Creb1 during SD, we are curious as to what effect SD will have on the abundance of ribosome-related transcripts involved in memory consolidation. As discussed in the previous results section above, CFC rarely changes the mRNA associated with cytosolic ribosomes, regardless of subsequent sleep (Figures 2 and 3). In Camk2a+ neurons, 19 cytosolic ribosome-related transcripts were changed in CFC + Sleep mice (compared to HC + Sleep control), while only 2 transcripts in CFC + SD mice were significantly changed Change (compared to HC + SD control). Among them, most (13/19 from CFC + Sleep, 2/2 from CFC + SD) also increased significantly due to SD alone (Figure 3A). In Camk2a+ neurons, ribosome-associated mRNA whose abundance is significantly affected by SD and CFC include activity-dependent transcripts such as Fosb, Arhgap39 (Vilse), and Errfi1 (Figure 3C, yellow). In contrast, in the input (ie, the entire hippocampus), after CFC + SD (27 transcripts) and CFC + Sleep (14 transcripts), there were slightly more significant changes in cytosolic ribosome-related mRNA. Among them, 6/27 and 9/14 were similarly significantly affected by SD alone (Figure 3A).

These initial data indicate that the changes in cytosolic ribosome-related transcripts after SD alone may mask some of the transcript changes triggered by CFC, that is, the effect of CFC itself on the abundance of transcripts cannot be detected. To better describe how this might affect the neurons most activated by CFC (which can represent CFC "imprinted neurons"), we compared the cytoplasmic transcripts affected by CFC and SD in pS6+ neurons. Although CFC and sleep or SD changed a similar number of transcripts (23 vs. 22), only two transcripts (Fosb and Egr3) were similarly affected in CFC + Sleep and CFC + SD mice (Figure 3A and C) . Among the transcripts altered on the cytosolic ribosome after CFC in pS6+ neurons, several (12/23 affected in free-sleeping mice, 6/22 affected in SD mice) are also individually regulated by SD (Figure 3A), including Fosb, Egr3, Arghap39 (Vilse), Gpr3, Ssh2, Inhba, Rnf19a and Cdkn1a (Figure 3C, yellow). Because SD changes a large number of transcripts involved in RNA splicing/processing (Figure 3B), and splicing isoforms play a key role in synaptic plasticity and memory storage (55), we next evaluated the effect of SD and CFC on differential splicing The effect of mRNA isoforms was analyzed using transcription level analysis (56) (Figure 3C). On the cytoplasmic ribosomes from pS6+ neurons, the activity-dependent splicing isoform of Homer1 scaffold protein (Homer1a) and the activity-dependent splicing isoform of Fosb, encoding a highly stable protein (ΔFosb), followed by CFC After casual use, sleep was significantly increased, but it was not significantly affected by CFC in SD mice (Figure 3C). These isoforms also increase as a function of SD alone, indicating that SD can prevent learning-induced changes in pS6+ neurons through another mechanism (Figure 3C, yellow).

In order to verify and extend these findings and better characterize the persistence of CFC-induced cell changes, we harvested hippocampus from CFC and HC mice after 5 hours of SD or random sleep (ie, later in the learning time point) . Using qPCR, we first quantified the mRNA levels of Fosb and Homer1 splicing isoforms in the entire hippocampus (input), Camk2a+ neurons, and the cytoplasmic part of pS6+ neurons. Similar to what was observed after 3 hours of SD, 5 hours of SD increased the expression of Fosb and its persistent splicing isoform ΔFosb, regardless of the previous CFC (SI appendix, Figure S8A, left). Compared with Input, Fosb and ΔFosb transcripts are relatively enriched on cytosolic ribosomes from Camk2a+ neurons, but both are highly enriched on cytosolic ribosomal pS6+ neurons (SI Appendix, Figure S8A, right ). This finding is consistent with the neural activity that regulates S6 phosphorylation (28) and the abundance of Fosb and ΔFosb transcripts (57, 58). To measure the CFC-driven changes in these transcripts, we normalized Fosb and ΔFosb transcripts in CFC + Sleep or CFC + SD mice to the corresponding HC controls. In pS6+ neurons and Input, relative to the HC control, ΔFosb increased after CFC, regardless of subsequent sleep or SD. In other words, CFC has a significant effect on the abundance of ΔFosb, even after considering the effect of SD alone on the expression of ΔFosb. However, on ribosomes taken from all Camk2a+ neurons, both Fosb and ΔFosb transcripts were increased in CFC + Sleep mice, but this increase was caused by the increase in transcripts caused by SD alone in CFC + SD mice (36) Covered by (SI appendix, Figure 2). S8 A, bottom).

We also used qPCR to quantify the relative expression of Homer1 and its splice variant Homer1a in the cytoplasm of CFC and subsequent 5 hours of sleep or SD. Homer1 itself was moderately affected by 5-hour SD, while its splice variant Homer1a was significantly elevated, consistent with earlier findings (36, 59, 60) (SI appendix, Figure S8B, left). Compared with Input, Homer1 is enriched in both Camk2a+ and pS6+ neurons, while Homer1a is only enriched in pS6+ neurons (consistent with the regulation of neuronal activity) (SI appendix, Figure S8B, right). Similar to ΔFosb, CFC increased Homer1a in all cell populations of mice that allowed sleep at will, but this increase was blocked by SD in Camk2a+ neurons (SI appendix, Figure S8B, bottom).

We quantified additional cytoplasmic localization transcripts of proteins with known functions in hippocampal plasticity and memory to test the effects of CFC and 5 hours of follow-up sleep or SD (SI appendix, Figure S9). These include the increase in transcripts that were increased after SD alone in our Desq2 analysis, and were not affected by CFC (Cfos and Arc), changed only in CFC + SD mice (Atf3), or only in CFC + Sleep mice Change (1700016P03Rik). We also visualized Egr3, which is not affected by SD but is increased by CFC in the CFC + SD and CFC + sleep groups. In Camk2a+ and pS6+ neuron groups, all transcripts except Egr3 were changed by 5-hour SD. In pS6+ neurons, Cfos and long-chain non-coding RNA (lncRNA) transcripts 1700016P03Rik (61, 62) are still significantly increased 5 hours after CFC as a function of learning; SD completely or partially occludes these learning-related Changes (SI appendix, Figure S9). At 5 hours after CFC, no significant CFC-induced changes in these transcripts were detected on the cytosolic ribosomes from Camk2a+ neurons.

Taken together, these data support the hypothesis that CFC-related changes in activity-regulated transcripts in the cytosolic ribosome may be blocked by the subsequent SD (discussion). This effect is most pronounced for highly active (presumed imprinting) hippocampal neurons, and constitutes a reasonable mechanism for SD to disrupt memory consolidation.

Compared with cytosolic ribosomes, there are fewer mRNAs on MB ribosomes that only change as a function of SD—most of the observed changes are driven by SD rather than CFC (Figures 2 and 4A and SI appendix, Figure S10) . The changes in MB ribosomal transcripts caused by SD are also dwarfed by the more changes in MB ribosomal-related transcripts after CFC. These changes are significantly different between Camk2a+ neurons, pS6+ neurons, and Input; therefore, the canonical pathways represented by the transcripts changed by SD in the three populations are also different. Crucially, in Camk2a+ neurons or inputs, SD-induced transcript changes do not significantly enrich the canonical pathways. On MB ribosomes from Input and Camk2a+ cell populations, SD-modified transcripts include components of cell pathways that are significantly affected in SD-regulated transcripts on cytosolic ribosomes (Figure 3B) and in the previous Transcripts affected by SD in hippocampal transcriptome studies (4⇓ –6, 24, 51). These include the components of AMPK (Chrm5, Irs2, Pfkfb3, Ppm1f, Prkab2, Prkag2 and Smarcd2), IGF1 (Elk1 and Rasd1), IL-3 (Crkl and Foxo1), relaxin (Gnaz, Pde4b, Smpdl3a) and neuromodulin (Errfi1) Signal pathway and glucocorticoid receptor signal components in Camk2+ neurons (Elk1, Gtf2e2, Prkab2, Prkag2, Rasd1, Smarcd2, Taf12, Fos, Krt77, Ptgs2 and Tsc22d3), unfolded protein response (Hspa ) And P Camk2a+ neuron and input pressure (Calr and Xbp1) pathways. However, the key point is that only 30 (6%) of the previously reported 511 mRNAs are altered by SD in the entire hippocampus (4) overlap with the mRNAs affected by SD in the MB part of the hippocampus (i.e. input; SI appendix ,image 3). S4). This suggests that even in these common cellular pathways, the individual transcripts of SD changes on MB ribosomes may be very different from those previously reported.

In contrast to the SD-driven changes in Camk2a+ neurons and inputs, the transcripts of SD changes from the MB ribosomes of pS6+ neurons significantly enrich several classical pathways. These include neurotrophin/TRK (Atf4, Bdnf, Fos, Ngf, Plcg1 and Spry2), corticotropin releasing hormone (Crh and Vegfa), ERK5 (Il6st and Rasd1) and EIF2 (Eif2b3, Hspa5, Ptbp1 and Rpl37a) signals Pathway and human embryonic stem cell pluripotency pathway (Bmp2, Inhba and Wnt2). Therefore, the main pathways affected by SD in MB ribosome-related transcripts include receptor signaling pathways, protein synthesis regulation, and ER function.

In contrast to the scarcely expressed CFC-driven transcript changes observed on cytoplasmic ribosomes, most of the transcript changes on MB ribosomes are driven by CFC (Figures 2 and 4A). Crucially, the changes in MB partial transcripts induced by learning are different in all cell populations, depending on whether the CFC is followed by 3 hours of casual sleep or 3 hours of SD. In contrast to the high overlap between SD-driven and CFC-driven transcript changes in the cytosol, on MB ribosomes, SD-affected mRNA and CFC-induced changes showed a significantly smaller overlap ratio (2% to 10%) (Figure 4A)). These data indicate that the translation profile of MB ribosomes is most selectively affected by previous learning, but the specific mRNA associated with MB ribosomes also changes significantly as a function of sleep or SD after learning.

The transcripts altered by CFC on MB ribosomes encode neuronal morphology, intracellular transport and lncRNA regulators. (A) In the MB part from Camk2a+ neuron, pS6+ neuron, and input, SD, CFC + sleep, and CFC + SD significantly changed the ratio of transcripts and the overlapping Venn diagram. (B) The seven most important molecular functions of transcripts changed by CFC + Sleep (top) and CFC + SD (bottom) in Camk2a+ neuron, pS6+ neuron and Input (sorted by Padj value). (C) The 10 transcripts that have the most significant effects on Camk2a+ and pS6+ neurons under the conditions of CFC + sleep (blue) and CFC + SD (red), sorted by Padj value. Transcripts that have also changed significantly as a function of the SD alone are highlighted in yellow. An analysis of the functional categories available in the data set S10.

We first characterized the cellular and molecular functions of MB ribosomal mRNA, which change with CFC and subsequent sleep or SD changes. For Camk2a+ and pS6+ neurons, the most abundant functional categories overlap in CFC + Sleep and CFC + SD mice and represent similar molecular categories—including cytoplasmic organization, cytoskeletal organization, microtubule dynamics, and cell -Cell contact and formation of protrusions. Figure 4B). In contrast, few of these categories have rich input MB scores. There, the most abundant functional categories of transcripts altered by CFC + Sleep include the excretion of sodium and potassium, the formation of cilia, the organization of cell protrusions, the desensitization of phagocytes, and abnormal amounts of phospholipids. The changes in the input mRNA after CFC + SD also enriched the functional categories not represented in the neuron MB component, including cell degeneration, membrane lipid derivative metabolism, sphingolipid metabolism, and ER stress response. Taken together, these data indicate that CFC may alter similar membrane-related cell functions in Camk2a+ and pS6+ neuronal populations, regardless of subsequent sleep or SD. In contrast, CFC may have significant effects on membrane-related functions in other hippocampal cell types (for example, glial), and these effects may vary depending on the sleep state of the animal.

To further characterize the changes in MB-related ribosomal transcripts after learning, we first compared the significantly changed mRNA (based on adjusted P values) in Camk2a+ and pS6+ neurons after CFC under sleep and SD conditions (Figure 4C). In the MB ribosomes of Camk2a+ neurons, CFC + Sleep resulted in an increase in the abundance of mRNA encoding transmembrane receptors (Chrm5 and Htr1a), and significantly reduced the abundance of a variety of lncRNAs, including Kcnq1ot1, Meg3, Mir99ahg, and others. Annotated transcripts (eg Gm37899 and Gm26917. 4C). Many other lncRNAs showed reduced MB ribosome abundance after CFC (including Neat1, Malat1, MiRG, and Ftx). In addition to Mirg and Ftx, these lncRNAs were also significantly reduced after CFC + SD. CFC + SD leads to Lrrc8c (encoding acid-sensing, volume-regulating anion channel) and anti-adhesion extracellular molecules (Sparcl1/Hevin), adhesion molecules (F3/Contactin1), transmembrane receptors (Paqr8) MB ribosomal transcripts The most significant increase of the potassium modulator (Kcng4), ER-bound lipid synthesis molecule (Hacd2) and actin modulator (Fam107a).

Among the highly active (pS6+) neurons, Dynac1h1 was the transcript that changed the most significantly after CFC, and it was significantly reduced in free sleep and SD mice (Figure 4C). Dync1h1 encodes the major retrograde motor protein in eukaryotic cells and supports retrograde transport in axons and dendrites (63). To a lesser extent, the abundance of MB ribosomes from Camk2a+ neurons is also significantly reduced. In any neuron population, Dynac1h1 did not decrease as a function of SD itself, indicating that the decrease in Dynac1h1 was specific to the post-learning condition.

We next constructed a canonical pathway network affected by CFC + Sleep or CFC + SD to visualize the signal transduction and metabolic pathways that change differently under the two conditions. Canonical pathways are represented as hubs and are connected by common transcriptional components. Here, the wheel sizes are weighted by their corresponding P-values ​​and shaded to represent their z-scores (blue represents a decrease in the passage after CFC, and red represents an increase) (Figure 5). A network comparison of MB ribosome-related transcript changes in Camk2a+ neurons revealed that the center of overlap was significant in both CFC + sleep (Figure 5A) and CFC + SD conditions (Figure 5D). These centers represent typical pathways for chondroitin sulfate degradation, unfolded protein response, notch signaling, phagosome maturation, supporting cell connection signaling, and epithelial adhesion signaling. However, the significance value of these pathway centers—similar to the number of altered transcripts in Camk2a+ neurons after CFC—is significantly higher in SD mice than in sleeping mice. For example, the overlap and centrality of Sertoli cell junction signaling, phagosome maturation, and epithelial adhesion junction signaling pathways indicate changes in common transcripts in free sleep and SD mice after CFC. After careful observation, although some tubulin transcripts were reduced after using CFC in the sleep group and SD group (Tuba1a and Tuba1b), a large number of tubulin-encoding mRNAs were reduced only after SD (Tuba4a, Tubb2a, Tubb3, Tubb4a, Tubb4b and Tubg1) (data set S10). Similarly, although the unfolded protein response-related transcripts increased moderately after CFC + Sleep (Calr, Mbtps1, P4hb, Sel1l, and Syvn1), they reacted with unfolded protein after CFC + SD (Amfr, Calr, Canx) Significant increase in related mRNA, Cd82, Cebpz, Dnajc3, Edem1, Eif2ak3, Hsp90b1, Hspa5, Mapk8, Mbtps1, Nfe2l2, Os9, P4hb, Sel1l, Syvn1, Ubxn4 and Xbp1).

The MB ribosomal transcription network affected by CFC changes with subsequent changes in sleep or SD. Classical pathway network analysis of altered transcripts on MB ribosomes from Camk2a+ (A and B) or pS6+ (C and D) neurons after CFC + sleep (A, C) or CFC + SD (B, D) . The center size and color indicate the Padj value and z-score under each condition, respectively, and the connecting lines indicate the genes commonly expressed between the centers. The normative approach is provided in the data set S10.

In Camk2a+ neurons, CFC + SD also changed the expression of MB ribosome-related mRNA related to the metabolic pathway, and was not affected in the CFC + Sleep group (Figure 5A and B). For example, CFC + SD increases the abundance of transcripts related to lipid (triacylglycerol, phosphatidylglycerol, cdp-diacylglycerol) biosynthesis, including encoding 1-acylglycerol-3-phosphate O-acyltransferase (Agpat2, Agpat3 and Agpat4), ELOVL fatty acid elongase mRNA (Elvol1, Elovl2 and Elovl6) and phospholipid phosphatase (Plpp3). CFC + SD also reduced the abundance of transcripts related to glucose metabolism pathways (glycolysis, gluconeogenesis, and TCA cycle), including Aldoa, Adloc, Eno1, Eno2, Gapdh, Gpi1, Pfkl, and Pkm (Data set S10 ). Combined with the results shown in Figure 4, these data indicate that sleep and SD cause different changes in the bioenergetic genes present in MB Camk2a+ ribosomes after learning.

We performed a similar classical pathway network analysis on the transcripts altered on the MB ribosomes of pS6+ neurons after CFC (Figure 5C and D). Many of the same pathways that CFC changes in Camk2a+ neurons (under sleep and SD conditions)—including supporting cell connection signals, epithelial adhesion signals, and phagosome maturation—are also observed in pS6+ neurons, suggesting some overlap. Under CFC + SD conditions, the pathways affected in Camk2a+ neurons include the lipid and carbohydrate pathways affected in pS6+ neurons. Interestingly, under CFC + sleep conditions (where CFM is combined), the abundance of MB ribosome-related transcripts representing protein translation regulation pathways (eIF2, eIF4, and p70S6K regulation and mTOR signaling pathway) increases. Crucially, this change does not exist in CFC + SD mice. In sleep and SD mice, MB ribosomal transcripts with reduced abundance after CFC include eukaryotic initiation factors (Eif3a, Eif3c, Eif3l, Eif4a1, Eif4g1, Eif4g3), mTOR, and Tsc1. However, in mice that allowed sleep after CFC, transcripts related to small ribosomal subunits were elevated in pS6+ neurons, including Rps12, Rps14, Rps17, Rps19, Rps20, Rps21, Rps23, Rps24, Rps26, Rps28 , Rps29 and Rps6. This suggests that after CFC, sleep may promote an increase in overall protein synthesis capacity, which occurs in the most active hippocampal neurons. The fact that these changes occur selectively on MB ribosomes suggests that this increased synthesis capacity may be cell compartment specific.

Our current results indicate that ribosome-related transcripts in the hippocampus not only change as a function of 1) learning and/or 2) sleep and lack of sleep, but also as a function of 3) the cell population being analyzed and 4) subcellular changes happened. The location of the ribosome. We found that the latter aspect (i.e. the location of the cell’s nuclear ribosomes) is the main factor in the effect of observed learning and subsequent sleep or SD on the hippocampal ribosome transcription profile. For a long time, it has been known that neuronal ribosomes are separated by cell compartments and exist in the form of "free-floating" (ie, cytosolic) or MB complexes, which are easily separated by centrifugation (64). These populations are known to participate in the compartmentalized translation of mRNA. The emergence of TRAP gave people an in-depth understanding of the special functions of ribosomes in different cell compartments. It is known that cytoplasmic ribosomes process mRNA encoding proteins that have functions in the cytoplasmic compartment, including transcription factors and kinases. MB ribosomes are usually associated with rough ER and translate mRNA encoding secreted or integral membrane proteins. Available data from non-neural cell types indicate that these two translation environments are biochemically different and can be regulated differently—for example, through cellular pressure (65). In the case of isolating ribosomes from the subcellular compartments of neurons (for example, Purkinje neurons) (35), MB ribosomes have been shown to be enriched in ER-related ribosomes and trees involved in local translation Ribosomes in the protrusion compartment. Our current research results reflect this, proving that the transcription profiles of MB and cytoplasmic ribosomes in hippocampal neurons are very unique (Figure 1 and SI appendix, Figures S1 and S2).

Many forms of hippocampus-dependent memory (in human subjects and animal models) are interrupted by lack of sleep before or after learning (1, 66⇓ –68). In fact, lack of sleep seems to destroy the plasticity mechanism in the hippocampus more than other brain regions (41, 69). The underlying mechanism by which lack of sleep causes these changes and disrupts memory mechanisms remains elusive. Transcriptome analysis of the effect of insufficient sleep alone on the hippocampus showed that SD increased the expression of genes involved in transcriptional activation and down-regulated the expression of genes involved in transcriptional repression, ubiquitination, and translation (4⇓-6). Although we found that the same cellular pathways as in previous studies were affected in the cytoplasm (Figure 3B), the specific transcripts affected by SD in the cytoplasm and MB ribosomal transcripts were compared with those affected by SD in these studies (SI Appendix, Figure S5). Our current data prove that in the entire hippocampus, Camk2a+ neurons, or pS6+ neurons, most of the pure SD-driven transcript changes are present in the cytoplasmic part and the cytoplasmic ribosome (Figure 2B and 3). Although SD-driven mRNA changes also occur on MB ribosomes, the number of these changes is relatively small (Figures 2B and 4). The pathways affected by SD-across neuronal populations and across two subcellular compartments-are related to transcriptional regulation, related to previous findings (4⇓ –6, 36) and AMPK, IL-3, IGF1, and PDGF signaling pathways Unanimous. Crucially, AMPK (70) and IGF1 signals (71) are related to the steady-state sleep response, that is, changes in the sleep structure of the brain that oscillate after SD. Therefore, it is easy to speculate that lack of sleep may lead to subsequent changes in sleep brain dynamics through changes in neuronal signals.

However, there are two unresolved questions: 1) Which SD-related changes in the synthesis or translation of specific transcripts provide a possible mechanism to disrupt hippocampal memory consolidation, and 2) Which cell types in the hippocampus learn after learning Severely affected by SD. This study aims to solve this problem in the context of a form of memory consolidation (CFM) that relies on the hippocampus, which relies heavily on sleep after learning. The work of our laboratory and others has shown that sleep interruption in the first few hours after CFC is sufficient to disrupt CFM consolidation (17, 19, 20). Although some system-level mechanisms that occur during sleep after CFC are related to the consolidation process (19, 20, 49, 50, 72, 73), there is almost nothing about the cellular mechanisms that mediate the effects of sleep (or SD) on sleep Know. CFM integration.

We were surprised that, compared with the large number of cytosolic ribosomal mRNAs affected by SD alone, CFC induced very few transcript changes on the cytosolic ribosomes. However, among those cytoplasmic transcripts that were changed by CFC, almost all 1) were similarly affected under CFC + sleep or CFC + SD conditions, and 2) were similarly changed by SD itself (Figure 3C). This makes sense in view of the fact that many mRNAs on the cytosolic ribosomes are known to be transcribed or translated in an activity-dependent manner after CFC. For example, SD up-regulates cytoplasmic transcripts regulated by Creb1; this effect is evident in hippocampal input, the Camk2a+ neuron population, and the highly active pS6+ neuron population. Importantly, after CFC, SD completely blocked the changes in learning-induced activity-regulated transcripts present on the cytoplasmic ribosomes of Camk2a+ neurons, and partially blocked the similar changes in Input and pS6+ neurons. In other words, if the mouse subsequently experiences SD, because SD alone causes similar changes, the changes in these transcripts attributable to CFC (compared to the HC control) can no longer be detected. Transcript changes showing SD-mediated occlusion include increases in Fosb (and its isoform Δfosb, which encodes a highly stable protein) and Homer1 (and its short isoform Homer1a) (SI Appendix, Figures S8 and S9). These transcript subtypes encode proteins closely related to synaptic plasticity and memory, including CFM (74, 75). Because here we examined the transcription levels in neuronal populations, rather than individual neurons, there are at least two reasonable explanations for the SD-driven occlusion we observed. First, SD and CFC may cause similar intracellular events in a single neuron. This situation is similar to how LTP-driven cellular changes and some forms of learning occlude each other in the hippocampus (76). Since the combination of learning and SD saturates intracellular mechanisms (for example, reaching the limit of transcriptional response), this mechanism may cause occlusion, as we have observed here. Or, the type of occlusion we see after SD may also be consistent with the effect based on neuron populations—that is, although CFC induces learning to activate transcript changes in a subset of neurons, additional (non-imprinted) neurons These same changes have been experienced during SD. This seems to make sense because the overall level of these transcripts increases more with SD alone rather than learning alone. Regardless of the precise mechanism, the time for SD to block these transcript abundance changes (3 to 5 hours after CFC) coincides with the critical window of sleep after CFC, which is essential for CFM consolidation (17, 19). Therefore, this change in cytosolic ribosomal transcription may represent a reasonable mechanism for lack of sleep to cause memory interruption.

This occlusion-based explanation of how sleep deprivation disrupts information storage is consistent with the central idea behind SHY's sleep function hypothesis that sleep is essential for reducing the signal-to-noise ratio in brain circuits (11, 77, 78). SHY proposed that this reduction is ultimately achieved through the reduction of synaptic strength throughout the brain. This hypothesis is supported by biochemical and physiological data, but it is not without controversy (1, 11). Crucially, the transcriptome data provided support for the hypothesis—including transcripts corresponding to the cytosolic ribosomal portion—consistent with our current analysis of the effects of SD on cytosolic ribosomal transcripts. Although it is not clear whether changes in synaptic strength are affected by SD, it is clear that the activity-driven transcription mechanism does change during SD in a way that may interfere with newly encoded information in the hippocampus. However, we found that SD-driven changes in cytosolic ribosomal transcripts represent only a small part of neuronal biology, which changes dynamically as a function of learning and subsequent sleep.

In contrast to the relative lack of transcripts on cytosolic ribosomes that have been altered by previous learning, CFC affects an alarming amount of mRNA on MB ribosomes (Figures 2B and 4A). In general, CFC induces changes in MB ribosome-related transcripts that encode proteins related to neuronal structure remodeling-from cellular pathways involved in cytoskeleton remodeling, intracellular transport, and cell-cell interactions (Figure 4B). Some changes are also very surprising and unexpected—for example, the significant reduction of ribosome-related lncRNA on MB ribosomes in the Camk2a+ neuron population after CFC (Figure 4C). Crucially, the precise transcript and (in some cases) cellular pathways changed significantly after CFC, depending on whether it was sleep or SD after learning (Figures 4 and 5). These differences provide a lot of information about the underlying mechanisms of SD-related CFM interrupts. For example, our current research results show that in the non-neuronal cell types of the hippocampus, CFC induces a unique set of transcript changes-these changes are present in the MB part of the Input, but not in the MB of the neuronal ribosome part. After SD (2⇓ –4), an increase in the abundance of transcripts related to energy metabolism is usually observed, especially those that encode mitochondrial proteins, glucose transporters, and proteins related to glycogen metabolism. Different from previous reports, our data shows that in Camk2a+ neurons, when CFC is followed by SD, cell metabolism/energy pathways may be selectively disrupted, but not when CFC is followed by sleep (Figure 4 and 5). Therefore, SD may disrupt CFM integration by increasing the metabolic demands on the hippocampus.

Among the most active (pS6+) neurons, CFC + Sleep leads to the regulation of many pathways related to the regulation of protein synthesis, including a broad increase in MB ribosomal mRNA encoding the translation machinery itself. When SD is followed by sleep, this change will not be seen. The pS6+ neuron population is likely to be mixed—contains a subset of the most active Camk2a+ neurons (those activated by experience) and some interneurons (20, 79). In fact, S6 phosphorylation may even occur selectively in specific neuronal synapses that are activated by learning (29). Nevertheless, the fact that transcription depends on learning and sleep changes between the pS6+ population and the entire hippocampus (input) or the entire Camk2a+ population indicates that ribosomal functions differ significantly in neuronal functions that depend on state and experience. Or synapse) activation. Therefore, it is easy to speculate that in the neurons and synapses that are most active after memory encoding (putative "imprinted neurons"), long-term changes in protein synthesis on the cell membrane may play a key role in subsequent memory consolidation. Neuropharmacological and biochemical studies have shown that SD's disruption of hippocampal cAMP signaling and protein synthesis may prevent memory consolidation (18, 24, 26, 27, 45). Our current data indicate that CFC may trigger changes in these pathways in specific subgroups of hippocampal neurons, and these changes will subsequently be promoted by sleep after CFC.

Recently, TRAP has been used to characterize compartment-specific ribosomal transcripts that project cortical axons from the amygdala during consolidation of fear memories (15). However, the method used in this study (which is true for most transcriptome and TRAP studies) may mainly report transcript changes related to the cytoplasm rather than MB ribosomes. Here, we show that most of the changes caused by learning itself are expressed on MB ribosomes (Figures 2B and 4)—surprisingly, CFC-induced changes in cytosolic ribosome-related mRNA are rare ( image 3). A recent comparison of hippocampal ribosome-associated and total mRNA abundance has shown that cytosolic and MB ribosome-associated mRNA are significantly regulated in terms of translation efficiency (80). Therefore, understanding the effects of learning and subsequent sleep on structures such as the hippocampus will require further study of their effects on the translation that occurs on the membrane.

An important caveat when considering the results of our current research is that they represent a specific CFC point in time after 3 hours. This time point may be very useful for understanding the CFM consolidation mechanism, because it represents the time of the peak change caused by CFC learning after hippocampal network activity (20, 49, 50) and the combination of the window CFM that requires protein synthesis (26, 27). However, after CFC, the time course of ribosome-related transcript changes in the cytosolic and membrane compartments seems to be different. If this is the case, learning-induced changes in ribosome-related transcripts may appear in the cytoplasmic compartment at an earlier or later time point. Lack of sleep may change the transcripts on the membrane-associated ribosomes at an earlier or later time point, which is also reasonable. Therefore, important research is needed in the future to determine the relative timing of CFC and sleep-induced ribosomal transcription changes in these two cell compartments.

Our current data indicate how common are sleep-dependent memory consolidation mechanisms? For example, in the context of sleep-related information storage, do other brain regions have the same neural mechanisms? It seems that very basic neurobiological mechanisms—such as activity-driven transcriptional regulation and the separation of specific transcripts between the cytosol and MB ribosomes—are similar throughout the brain. In addition, the latest data from our laboratory indicates that the storage of new information in the hippocampus (19, 50) and neocortex (12, 82) requires learning of certain characteristics of post-sleep sleep, such as coordination of network oscillations (14, 81). However, our data also showed that many cell type-specific effects of SD differ between hippocampus and neocortex (20, 36, 69). Future research needs to determine the consistency of these different subcellular responses to learning and subsequent sleep in different cell types and brain regions. However, the integration of various types of memory across species share a common cell matrix (1, 14, 83), among which mRNA translation after learning is an important factor. Changes in neuronal activity patterns and the activation of specific intracellular pathways during sleep after learning have common characteristics, spanning brain structures and species (1, 14). Our current research results have illustrated several sleep-dependent post-learning cellular processes that affect pathways that are critical to learning and memory. Future research will determine whether these processes are the basis of sleep-dependent memory consolidation events in other brain circuits, following different learning styles.

All animal feeding and experimental procedures were approved by the Institutional Animal Care and Use Committee of the University of Michigan (Public Health Service Animal Welfare Assurance Number D16-00072 [A3114-01]). All mice maintained a 12:12-hour light/dark cycle (light on at 8 am) and provided food and water ad libitum. B6.Cg-Tg(Camk2a-cre)T29-1Stl/J mice (Jackson) were crossed with B6N.129-Rpl22tm1.1Psam/J mice (Jackson) to express HA-labeled Rpl22 protein in Camk2a+ neurons. The double-transgenic mice were bred separately for 1 week before the experiment, and were accustomed to handling (5 minutes/day) for 5 days before the experiment. For RNA-seq experiments, mice were randomly assigned to one of four groups: HC + Sleep (n = 8), HC + SD (n = 7), CFC + Sleep (n = 8) or CFC + SD (n = 7). The bilateral hippocampus of each mouse contained a biological replicate for sequencing, and samples were not mixed between mice. Starting from turning on the lights (8 AM), half of the mice received the single CFC test described previously (19, 49, 50). In short, when the light is on (ZT 0), the mouse is placed in a new conditioning room (Med Associates) and allowed 2.5 minutes before passing the grid floor of the room for a 2 second, 0.75 mA foot shock Free time to explore. After a total of 3 minutes in the chamber, the mice returned to their original HC. As a control for the learning effect, the HC control remains in their HC during this period. For the next 3 hours, the HC + SD or CFC + SD mice were treated with gentle treatment (SD; including knocking the cage, disturbing the nest, and stroking with a cotton swab applicator if necessary) to keep awake for all RNA- 5 hours for seq research or all qPCR experiments. HC + Sleep and CFC + Sleep mice were allowed to sleep in their HC at will for the same time interval.

RiboTag TRAP is performed by indirect binding (84) as described previously (31), separating MB from free-floating ribosomes (35). After homogenization and centrifugation, transfer the resulting supernatant (cytoplasmic fraction) to a new tube while resuspending the pellet (MB fraction) in homogenization buffer. Both MB and cytoplasm are divided into input, Camk2a+ and pS6+ parts. To isolate ribosomes from the Camk2a+ population, the fractions were incubated with 1:40 anti-HA antibody (Abcam, ab9110) (85). To isolate ribosomes from highly active (pS6+) neurons, the sections were incubated with 1:25 anti-pS6 244-247 (ThermoFisher 44-923G) (28). The homogenized antibody solution was directly added to Protein G Dynabeads (ThermoFisher, 10009D) for incubation. After washing the bound beads and removing the supernatant, the RNA was eluted by vigorously vortexing the beads in 350 μL RLT (Qiagen, 79216). Purify the eluted RNA using RNeasy Micro kit (Qiagen).

RNA-seq is performed at the core of DNA sequencing at the University of Michigan. The amplified complementary DNA library was prepared using Takara's SMART-seq v4 ultra-low input RNA kit (Takara 634888) and sequenced on Illumina's NovaSEq. 6000 platform. Reads mapped to unique transcripts are counted using featureCounts (86). Using Deseq2 to perform differential expression analysis on all 30 hippocampal samples, the bilateral hippocampus of each mouse constituted a biological replicate (39). To characterize the difference between the effects of SD and CFC, IPA was used to analyze significantly changed transcripts. Gene Ontology (GO) analysis is performed in the functional annotation tools of IPA and DAVID. For comparison of subcellular fractions, 2,000 top cytosolic (Log2 fold change [Log2FC]> 0) and MB (Log2FC <0) differentially expressed transcripts (sorted by adjusted P value) were run through the typical analysis of IPA. . To characterize the differences in common metabolic pathways between the cytosolic and MB fractions, hierarchical clustering was used to visualize the transcripts with the largest expression differences. Since the signal pathways have less overlap between the MB and cytoplasmic parts, they are sorted by enriched P values. Then run these transcripts through DAVID's functional annotation tool, and select the cell composition to describe the cell compartments related to the corresponding protein. The data are plotted in the number of fragments per million, and their correlation value (R) is calculated in the ViDger R package (87).

The complete materials and methods are in the SI appendix, SI materials and methods.

All research data is included in the article and/or supporting information.

We are grateful to the members and PhDs of the SJA laboratory. Natalie Tronson and Ryan Mills provided helpful feedback on this manuscript. This work was funded by NIH (DP2 MH 104119 and RO1 NS118440), SJA's Human Frontier Science Program (N023241-00_RG105), and the Bioinformatics Core of the University of Michigan's Biomedical Research Core Facility.

↵1 Current address: Department of Health Sciences, Mayo Clinic, Rochester, Minnesota 55905.

Author contributions: JD and SJA design research; JD, LW, VK, YW, JM, and SJ conducted research; JD, LW, VK, and SJA analyzed data; JD and SJA wrote this paper.

The author declares no competing interests.

This article is directly contributed by PNAS.

This article contains online support information https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.2108534118/-/DCSupplemental.

This open access article is distributed under the Creative Commons Attribution-Non-Commercial-No Derivative License 4.0 (CC BY-NC-ND).

Thank you for your interest in advertising on PNAS.

Note: We only ask you to provide your email address so that the people you recommend the page to know that you want them to see it and that it is not spam. We do not capture any email addresses.

Feedback privacy/legal

Copyright © 2021 National Academy of Sciences. Online ISSN 1091-6490. PNAS is a partner of CHORUS, COPE, CrossRef, ORCID and Research4Life.