To evaluate the effects of sleep loss, we compared three sleep-related conditions for differentially expressed genes (DEGs) from samples of the frontal cortex (FC, the cortical region showing the highest power of SWS-SWA) of male C57BL/6J (WT) mice (Figure 1A) including: (1) control sleep (CS; mice free to sleep, ad libitum; n = 6), (2) sleep deprived for 6 hr (SD; n = 7), and (3) sleep deprived for 4 hr followed by recovery sleep for 2 hr (RS; n = 7). All samples were collected 6 hr after lights on (ZT6) to control for circadian influence and were processed for RNA-seq analysis (see Materials and methods).

Figure 1 with 3 supplements see all

Download asset Open asset

The CS samples expressed, on average, ~13,000 genes with counts per million reads (CPM) ≥1.0 in the FC. Comparing the SD samples to the CS samples, we observed 6,248 DEGs with roughly half the number of genes having decreased expression (FDR <= 0.05; |log₂ (fold change)|>=0.3; Figure 1B; Supplementary file 1), confirming an earlier report of massive transcriptome changes associated with SD (Diessler et al., 2018). Similarly, the RS cohort showed 3827 DEGs compared to CS, again with approximately half the number of genes showing decreased expression (Figure 1C; Supplementary file 1).

In contrast, only 81 genes showed significantly altered expression comparing the SD condition to RS (Supplementary file 1), suggesting that for the majority of SD-sensitive genes, RS is more than simply a direct recovery of altered gene expression back to CS levels. Rather, more than half of the differential expression initiated during SD is continued during RS (Figure 1D), significantly extending the SD-induced change.

Because SD may be considered stressful (Mongrain et al., 2010), we compared classic glucocorticoid-mediated response genes to the increased DEGs observed in response to SD (Figure 1—figure supplement 1, Supplementary file 2). No significant increase in glucocorticoid (GC) intracellular response genes was found in the upregulated SD DEGs (p-value=0.513 using Fisher’s exact test), suggesting a more SD-specific response than other GC-mediated stress responses.

This bulk transcriptomic analysis includes mRNA from all cell types found in frontal cortical tissue. Many of the genes expressed in these tissues are, nonetheless, cell-type specific, allowing potential correlation between DEGs specific for a sleep condition and specific cell types. Using previously published single-cell RNA-seq from the adult mouse cortex (Hrvatin et al., 2018), we identified cell-type-specific gene expression patterns across sleep conditions (Figure 1E). Interestingly, the downregulated DEGs shared by SD and RS were enriched for genes associated with both excitatory and inhibitory neurons, whereas the upregulated DEGs shared by SD and RS were enriched for non-neuronal genes that are down in the Mef2c-cKO^Emx1 mice (Harrington et al., 2016 ; Figure 1E).

We also examined MEF2C target gene expression in response to SD or RS. We found that the downregulated DEGs shared by SD and RS were enriched for neuronal genes affected by Mef2c loss of function as previously observed in a conditional knockout of pan-neuronal Mef2c, Mef2c-cKO^Emx1 (Harrington et al., 2016). Whereas the upregulated DEGs in response to SD and RS were enriched for Mef2c non-neuronal genes (Figure 1E). Thus, the DEGs altered by SD and RS in the cortex significantly overlap those altered by the loss of cortical MEF2C function.

We next functionally characterized the sleep-loss induced transcriptional changes shared by SD and RS states by carrying out a gene ontology (GO) analysis. Overall, there are distinct GO categories for the upregulated and downregulated genes (Figure 1F,G; Supplementary file 2). The upregulated categories include mitochondrial components and signaling pathways of oxidative phosphorylation and anaplerosis (Figure 1—figure supplement 2), indicative of energy mobilization, as well as anabolic protein metabolism that includes ribosomal components together with transcriptional, translational and related processes. These results are consistent with and extend those from a previous study using microarrays (Cirelli et al., 2004) that lead to a conclusion, ‘… that sleep, far from being a quiescent state of global inactivity, may actively favor specific cellular functions’. Recently, similar findings were reported in a RNA-seq analysis from C57BL/6J whole cortex and a ‘gentle handling’ method of SD (Hor et al., 2019).

Several studies provide converging evidence for a general decrease in cortical excitatory synaptic strength in association with sleep (de Vivo et al., 2017; Liu et al., 2010; Maret et al., 2011) with respect to protein expression in synaptosomes (Diering et al., 2017) and spontaneous miniature excitatory post-synaptic currents and frequency (Liu et al., 2010; Vyazovskiy et al., 2008). Consistent with this, we observed that the downregulated genes observed in SD and RS are enriched for GO categories reflecting synaptic biological processes and cellular components. These genes are also enriched for genes involved in diseases of cognition associated with abnormal synaptic function, including schizophrenia, autism, intellectual disability and neurodegenerative disorders (Figure 1G, Supplementary file 3).

Some of the SD upregulated DEGs have a notable role as both transcription factors and/or first responders to neural activity and sensory stimulation. Rapid, primary response genes (rPRG) have been identified in cortical neuronal cultures as responders to both brief or 6 hr exposures to a depolarizing concentration of potassium in the presence of a translational blocker (Tyssowski et al., 2018). We find that DEGs for CS compared to SD and/or RS conditions are significantly enriched for these rPRGs (Figure 1E). Furthermore, upregulated transcription factors were identified in visual cortex that responded to a short exposure to light after being in constant darkness (Hrvatin et al., 2018). We observed that 50% of these were upregulated DEGs for CS to SD (listed in Supplementary file 1). Notably, these upregulated, transcription factors in the visual cortex overlapped with DEGs from the FC, and showed sustained, increased expression even after 6 hr of SD-induced activation.

It was recently shown that the influence of circadian rhythms on gene transcript expression in synaptosomes was relatively unaffected by SD, although SD did prevent translation (Noya et al., 2019). Nevertheless, many core clock genes are known to affect sleep homeostasis and the expression of some core clock genes, in particular, Per2, are altered by SD, as previously reviewed (Franken, 2013). To control for circadian-mediated changes in gene expression, we assessed DEGs at the same circadian time, ZT6, changing only the 6 hr, immediate sleep history. In contrast to DEGs in synaptosomes, we observed in FC tissue that SD from ZT = 0 to 6 hr resulted in a number of core clock DEGs (compared to CS; Figure 1—figure supplement 3), including downregulation of Arntl, Clock and Npas2 and downregulation of a set of genes regulating circadian entrainment (MGI; GO:0009649). Taken together, these data are consistent with an SD-induced attenuation and destabilization of circadian-mediated gene expression and confirm recent observations using an RNA-seq assessment in time series together with SD (Hor et al., 2019).

To further uncover the organizational architecture and prioritize the extensive SD-mediated transcriptome, we used weighted gene co-expression network analysis (WGCNA), with expressed genes from all three conditions (CS, SD, and RS) to generate a network of 59 modules. Many of the modules (53%) were enriched for either up or downregulated DEGs from CS to SD conditions (Figure 2A). Most of those enriched for upregulation from CS to SD (n = 11), were also enriched for upregulation from CS to RS (n = 10), and similarly, most modules enriched for downregulation from CS to SD (n = 18) were also enriched for downregulation from CS to RS (n = 13), showing an organization primarily driven by DEGs, that is consistent between SD and RS conditions. The GO of these shared modules (Supplementary file 4) is also consistent with those seen using only the lists of DEGs.

Figure 2 with 1 supplement see all

Download asset Open asset

The advantage of WGCNA is that it allows for the prioritization of gene modules that may be central drivers of biological networks independent of arbitrary cutoffs of p-values and fold changes for individual genes. WGCNA identifies these modules containing genes with highly correlated expression that can be further assessed for significant expression patterns related to features such as the sleep conditions of CS, SD and RS via determination of module eigengenes.

Notably, putative target genes of MEF2C, were enriched in a surprising, 15 of 59 modules generated from the three sleep conditions employed in this study. Using a Monte Carlo model to assess the likelihood of such a result gave a probability of p<0.001 from a sample of 1000 permutations of genes randomly assigned to the experimentally defined number of modules and module size. It is consistent with a widespread influence of MEF2C on the transcriptome as it varies from CS to SD to RS. Furthermore, most of the enriched modules had sleep-condition-related eigengenes indicating that sleep condition was the principal component factor driving their coordinated expression.

We identified two modules (M04 and M10) whose module eigengenes were more positively correlated with CS or ad lib sleep or, conversely, more negatively correlated with SD and RS samples (Figure 2B and Figure 2—figure supplement 1), indicating that the primary component driving their pattern of coexpression is CS versus SD and RS conditions. These two modules were of interest due to significant enrichment of several major categories of genes (Figure 2A): (1) DEGs in neuronal Mef2c conditional KO cortex (Mef2c-cKO^Emx1; Harrington et al., 2016), (2) excitatory and inhibitory neuronal genes, (3) genes downregulated in SD and RS, and (4) genes involved in synaptic function, epilepsy and cognitive pathologies including autism (Figure 2D). Both of these modules have hub genes encoding receptors, ionic channels or cell adhesion, and identified as autism and epilepsy risk factors (for complete GO of M04 and M10 and other modules, see Supplementary file 4). These characteristics are generally consistent with modulation by the MEF2C transcription factor and suggest the modulation by MEF2 is closely associated with sleep conditions. Also, we observe significant enrichment for Mef2c downstream genes (Harrington et al., 2016) in 15 out of 59 modules, which is significantly greater than randomly assigning the genes to simulated 59 modules of the same size (Monte Carlo p-value: 0.001, N: 1000 iterations).

An association of MEF2C target genes was observed by comparing the CS to SD DEGs; we found enrichment for DEGs in Mef2c-cKO^Emx1 compared to WT cortex involved in regulating synaptic strength (Figure 1E,G, Supplementary file 3). Additionally, modules of correlated gene expression across sleep conditions showed a similar enrichment of putative MEF2C targets, pattern of expression and GO as the DEGs across sleep conditions (Figure 2A, Supplementary file 5).

To better understand the role of MEF2C across sleep conditions, we examined a conditional knockout mouse using a Cre-loxP system. We crossed a Mef2c^f/f mouse line (Arnold et al., 2007; Zang et al., 2013) with a transgenic Camk2a-Cre line and examined these Mef2c-cKO^Camk2a-Cre mice (see Materials and methods for details) for their response to sleep loss. Mef2c^f/f mice (see Materials and methods for comparisons to WT DEGs in response to SD) had 767 DEGs within 12,403 total expressed genes from CS (n=4) to SD (n=3); whereas, strikingly, the Mef2c-cKO^Camk2a-Cre mice had only 36 DEGs within 12,620 total expressed genes in CS (n=3) to SD (n=3; Figure 3A, B, C). Both the Mef2c^f/f and the Mef2c-cKO^Camk2a-Cre genotypes expressed similar numbers of FC genes as the WT, using a CPM ≥1.0, but the responses of Mef2c-cKO^Camk2a-Cre to sleep conditions were attenuated to ~7% of the Mef2c^f/f responses, indicating the essential role of Mef2c in the transcriptome response to sleep loss.

Figure 3

Download asset Open asset

Although the observed attenuation of Mef2c^f/f mouse DEGs from CS to SD induced by the conditional loss of Mef2c is striking, it is of note that the number of CS to SD DEGs comparing WT to Mef2c^f/f also shows some attenuation. We further investigated the differences between WT and Mef2c^f/f DEGs by down-sampling and permutation of the WT samples so that the numbers of observations between the two cohorts was comparable (CS (n = 4) and SD (n = 3)). We observed increased variability in gene expression and such increased variability can lead to decreased number of observed DEGs (Supplementary file 9). However, the decrease due solely to increased variability does not appear of sufficient magnitude to fully account for the difference between WT and Mef2c^f/f mice.

A potential contributing factor could result from an effect of the LoxP insertions on MEF2C expression. However, an assessment of the specific genomic expression showed only minor variability in exon expression between WT, Mef2c^f/f and Mef2c-cKO^Camk2a-Cre samples with only the expected specific reduction of the floxed exon in Mef2c-cKO^Camk2a-Cre mice (Figure 4).

Figure 4

Download asset Open asset

Another potential source of difference is from small differences in gene expression due to genetic background. The Mef2c^f/f line employed in this study has been employed in several other studies of its role in synaptic function (Barbosa et al., 2008; Harrington et al., 2016; Rajkovich et al., 2017), but this does not rule out differences in transcriptional expression due to genetic background. While the Mef2c^f/f mice were backcrossed to the same genetic background of the WT mice for five generations, there could still be variability in gene expression due to incomplete congenicity and/or epigenetic variability. Interestingly, an examination of the overlapping DEGs comparing CS to SD from WT and Mef2c^f/f mice used in this study (Supplementary file 10, Tab ‘Overlap’), shows a gene ontology represented by genes controlling synaptic function (Supplementary file 10, Tab ‘Gene Ontology), indicating that sleep-condition regulation of synaptic genes is maintained between the two genotypes. Furthermore, the synaptic and EEG phenotypes of these genotypes across sleep conditions are indistinguishable as shown below (Figure 6—figure supplement 1 and Figure 5).

Figure 5 with 1 supplement see all

Download asset Open asset

MEF2 transcriptional activity is increased upon activity-dependent dephosphorylation of S396 (Flavell et al., 2006). Expression of a constitutively-active form of MEFC (MEF2C-VP16) is sufficient to reduce cortical excitatory synapses (Harrington et al., 2016), mimicking reduction of excitatory synaptic strength previously reported in response to recovery sleep (Liu et al., 2010). This led to our speculation that sleep loss might decrease MEF2C phosphorylation at S396. Using a phosphorylation site-specific antibody (Flavell et al., 2006), we examined the MEF2C phosphorylation status in FC under three sleep conditions (i.e. CS, SD and RS). We found that compared to CS, SD produced a significant reduction of MEF2C P-S396 (Figure 3D), consistent with an increase in MEF2C activity during SD, but without altering total MEF2C protein levels (Figure 3D). In contrast, RS did not significantly alter MEF2C P-S396 levels (Figure 3D), suggesting that SD triggers signaling events to transiently activate MEF2C-dependent transcription and promote the weakening and/or elimination of excitatory synapses.

As noted above (Figure 1E), CS to SD and RS downregulated DEGs were enriched for genes that were downregulated in an earlier study comparing WT to Mef2c-cKO^Emx1 in control conditions (Harrington et al., 2016). Similarly, the upregulated genes from CS to SD and RS are enriched for up-regulated WT to Mef2c-cKO^Camk2a-Cre genes. In other words, the DEGs of Mef2c-cKO^Camk2a-Cre (compared to WT) in control conditions are similar to the DEGs of CS to SD in WT mice. However, in the Mef2c-cKO^Camk2a-Cre used in our study, there is little response to SD (Figure 3B,C), rather the DEGs of the WT sleep loss response are now differentially expressed in CS conditions in Mef2c-cKO^Camk2a-Cre mice, potentially, a compensatory response, no longer responsive to sleep loss.

Based on our transcriptome analyses, SD and RS induced many DEGs affecting synaptic strength and synaptic cellular components that required the MEF2C transcription factor (Figures 1, 2 and 3). MEF2C can regulate forebrain excitatory synapse strength of pyramidal neurons by decreasing synaptic density (Harrington et al., 2016) and frequency (Flavell et al., 2006). Down-scaling of cortical synaptic strength is associated with sleep (de Vivo et al., 2017; Diering et al., 2017). Previous observations indicate that acute loss of sleep is associated with increased miniature excitatory post-synaptic (mEPSC) frequency and amplitude and that recovery sleep reverses these changes in cortical neurons recorded from C57BL/6 mouse FC (Liu et al., 2010). However, neither the involvement of changes in probability of release nor changes in active excitatory synapse number and the role of Mef2c have been previously considered.

To directly examine sleep-associated changes in synaptic strength, the same conditions as employed for the transcription analyses (i.e. CS, SD, RS) were applied to Mef2c^f/f male mice. Acute ex vivo slices of FC were prepared from these mice at ZT = 6 hr, for whole cell voltage-clamp recordings from layer 2–3 pyramidal neurons. Because sleep-loss is associated with an increased extracellular concentration of adenosine throughout the forebrain (Porkka-Heiskanen et al., 2000; Porkka-Heiskanen et al., 1997) which decreases glutamatergic mEPSC frequency and amplitude (Brambilla et al., 2005; Scanziani et al., 1992), all recordings were made in the presence of adenosine receptor 1 and 2 blockade (cyclopentyl-theophyline, 1 μM).

As expected, we observed an increase of mEPSC frequency and amplitude after SD and the recovery to CS-levels in RS condition in the WT and Mef2c^f/f mice (Figure 6A–F; Figure 6—figure supplement 1; Supplementary file 6).

Figure 6 with 1 supplement see all

Download asset Open asset

To determine if the SD-induced increase in mEPSC frequency was caused by an increased presynaptic release probability or an increase in synapse number, we examined the paired-pulse (P₂/P₁) ratio from CS to SD and RS (Figure 7A,B).

Figure 7

Download asset Open asset

The P₂/P₁ ratio tended toward an increase from CS to SD condition (this is consistent with a decreased probability of release), suggesting that the SD-associated increase in mEPSC frequency resulted from an increased number of active synapses. However, for CS to RS, the observed significant increase in P₂ to P₁ ratio and the normalization of mEPSC frequency after RS is due to either a decrease in release probability, a decrease in active synapse number or both.

In Mef2c-cKO^Camk2a-Cre mice, mEPSC frequency and amplitude was greater than in Mef2c^f/f in CS condition. Consistent with the lack of change in the Mef2c-cKO^Camk2a-Cre transcriptome, a significant change in frequency and mEPSC amplitude in response to SD and RS sleep conditions was absent (Figure 6G–L, Supplementary file 7). Finally, as expected from the absence of SD or RS induced effects on mEPSCs, the Mef2c-cKO^Camk2a-Cre mice had no change in P₂/P₁ ratio across sleep conditions (Figure 7C,D). These observations support a necessary role for Mef2c in SD and RS associated changes in synaptic strength for FC layer 2/3 excitatory neurotransmission.

Sleep need, inferred from the duration of time in waking without sleep (Borbély, 1982; Franken et al., 2001), is directly correlated with the ensuing SWS-SWA power calculated from surface EEG recorded from control Mef2c^f/f and Mef2c-cKO^Camk2a-Cre mice. In WT and Mef2c^f/f mice, SWS-SWA is greatest at the start of the sleep phase and resolves over the sleep phase course as the mice spend time in SWS (ZT 0–12 hr; Figure 5A,C,E). In contrast, Mef2c-cKO^Camk2a-Cre mice express, for their averaged SWS-SWA episodes, higher SWA (Figure 5E) that fails to normally resolve over the course of the sleep phase (Figure 5B,C). Another indicator of increased sleep need is increased consolidation of SWS as suggested by a rightward shift in the cumulative histogram of SWS episode duration going from CS to SD conditions (Bjorness et al., 2016), which is occluded in Mef2c-cKO^Camk2a-Cre mice (Figure 5—figure supplement 1A).

With respect to arousal-state duration (waking vs SWS vs REM), Mef2c-cKO^Camk2a-Cre mice had no apparent significant differences and there were minimal differences in the power distribution across the frequency spectrum (Figure 5—figure supplement 1B,C). In response to SD, the Mef2c-cKO^Camk2a-Cre mice show an attenuated or absent rebound of SWS-SWA (Figure 5F,G). These observations suggest that a loss of MEF2C function both increases the need for sleep and attenuates its resolution during SWS.

Since MEF2C Haploinsufficiency Syndrome (MCHS) in humans is associated with sleep-disruption (Harrington et al., 2020; Paciorkowski et al., 2013), we examined constitutive gene dosage on SWS-SWA power of a global Mef2c heterozygous gene-deletion mouse model (Mef2c^+/-) of MCHS under CS and SD conditions (Figure 8).

Figure 8 with 2 supplements see all

Download asset Open asset

Under CS conditions, SWS-SWA power resolved across the sleep phase, unlike the Mef2c-cKO^Camk2a-Cre mice, indicative of a resolution of sleep need. Similarly, Mef2c^+/- mice showed resolution of sleep need over an average SWS episode, again, unlike Mef2c-cKO^Camk2a-Cre mice; however, the rate of resolution (as indicated by the rate of decay of SWA in an average SWS episode) was slowed. On the other hand, SD did not induce the expected rebound SWS-SWA in Mef2c^+/- mice similar to Mef2c-cKO^Camk2a-Cre mice. Another similarity of Mef2c^+/- mice with the Mef2c-cKO^Camk2a-Cre mice is the apparent occlusion of the rightward shift in the cumulative histogram of SWS episode duration going from CS to SD conditions (Figure 8—figure supplement 1), associated with increased sleep need in WT mice (Bjorness et al., 2016). The Mef2c^+/- mice are already shifted to the right compared to Mef2c^+/+ mice in both CS and SD conditions. Furthermore, the Mef2c^+/- mice had an abnormal relative increase in waking SWA power (in the high delta/low theta range of 4–5 Hz; Figure 8—figure supplement 1).

By comparing differential gene expression under CS conditions in each of the Mef2c models, we found a significant overlap of genes changing in the Mef2c^+/- and Mef2c-cKO^Camk2a-Cre compared to their respective controls (142 overlapping genes; p-value=5.28e-08, Fisher’s exact test (Figure 8—figure supplement 2, Supplementary file 8)). The identified overlapping gene set can provide clues into the molecular mechanisms underlying a requisite Mef2c sufficiency. Moreover, the non-overlapping genes can reflect either developmental effects or non-neuronal effects present in the haploinsufficiency model, but absent in the Mef2c-cKO^Camk2a-Cre mice. Taken together, these findings suggest constitutive loss of one Mef2c allele is sufficient to alter the normal response to sleep loss, but the loss appears to be at least partially compensated when spontaneous sleep remains undisturbed. More acute, complete loss of Mef2c function as with the Mef2c-cKO^Camk2a-Cre mouse, even further disrupts the homeostatic sleep loss response.

All mice (Mus musculus) had a C57BL/6J background, were male and of 8–19 weeks of age at time of data collection. Mice were on a 12 hr light-dark cycle with access to food and water ad libitum. For RNA-seq experiments, we employed three genotypes: wild type (WT); a knock-in of loxP sites flanking exon 11 of the Mef2c gene, the second coding exon for the DNA-binding site (Arnold et al., 2007)(Mef2c^f/f), a control for the Mef2c-cKO^Camk2a-Cre; and a CamKII-promoter (Tsien et al., 1996), CRE-mediated recombination of the Mef2c^f/f loxP sites (founders obtained from Jackson Labs Jackson labs; B6.Cg-Tg(Camk2a-Cre)T29-1Stl/J), resulting in a region and cell-type-specific conditional knockout (Mef2c-cKO^Camk2a-Cre [Arnold et al., 2007]). Primers and primer products are shown along with a map of the Mef2c gene (Figure 9). Initially, we used a primer pair, Mef2C.2 and Mef2C.3 that gave a product of 650 BP for WT, 750 BP for Mef2c^f/f and no product for a CRE-mediated recombined allele to distinguish homozygous and heterozygous floxed knockins from WT mice.

Figure 9

Download asset Open asset

In order to confirm CRE-mediated recombination in the brain, we used another primer pair, Mef2C KO F1 and Mef2C KO R2 (Figure 9) that gave products of 2250 BP for non-recombined Mef2c^f/f and 1000 BP for recombined Mef2c-cKO^Camk2a-Cre alleles. All brain samples of Mef2c-cKO^Camk2a-Cre mice expressed both products as expected.

We also determined that some of the offspring of male Mef2c^f/wt;Camk2a-Cre mice were heterozygote for a single recombined allele, showing Mef2c^f/- in tail samples, consistent with CRE-mediated recombination in spermatocytes (Luo et al., 2020). Thus, gene-dosage-dependent developmental effects of Mef2c loss of function in all cell types may contribute to the transcriptome of Mef2c-cKO^Camk2a-Cre mice in addition to the homozygous conditional loss of Mef2c function in the brain. Nevertheless, a heat map of differentially expressed genes (DEGs) from Mef2c-cKO^Camk2a-Cre mice showed similar clustering of DEGs from samples in CS conditions with, and without (second column Figure 10) recombination in the tail.

Figure 10

Download asset Open asset

For in vivo electrophysiology (EEG/EMG analysis for sleep/waking state determination), in addition to the three strains described above, we also included constitutive heterozygous Mef2c knockouts (Mef2c^+/-) and their littermate controls (Mef2c^+/+); generation of these strains has been described previously (Harrington et al., 2020).

All experimental procedures were approved by the University of Texas Southwestern IACUC or VA North Texas Health Care System IACUC.

All mice underwent an accommodation protocol of 3 weeks duration, in 12/12 light/dark schedule, in plexiglass, false-bottomed cages, suspended over a treadmill (see Video 1) with ad libitum water and mouse chow as previously described (Bjorness et al., 2016). The treadmill (speed of treadmill = 3 cm/sec) was turned on daily for 10 min to accommodate the mice to the slow movement (mice are able to eat and drink with it running) and to clean it. For RNA-seq, 36 hr prior to the experimental day (CS, SD or RS) a dark/dark schedule was instituted. The use of the treadmill prevented transition to sleep, did not increase motor activity (due to the slow speed) and provided a mild, familiar, non-varying, arousing stimuli that reduced the experimental variability associated with more traditional ‘gentle handling’ techniques. Animals were randomly assigned to CS, SD or RS groups.

Video 1

Download asset

Total RNA was extracted from all the samples using TRIzol reagent (Thermo Fisher Scientific, Cat. No. 15596026) and was purified with RNeasy mini kit (Qiagen, Cat No. 74104) according to manufacturer’s instructions. All the samples were then randomized in order to reduce the batch effect and submitted to McDermott Sequencing Core at University of Texas Southwestern Medical Center for library preparation and sequencing. Libraries were prepared using TrueSeq Stranded mRNA Library Prep (Illumina, Cat. No. 20020596) as per manufacturer’s instructions by McDermott Sequencing Core. The quality and concentration of the libraries was checked on Agilent Bioanalyzer High Sensitivity DNA chip (Agilent, Cat. No. 5067–1504). Samples were pooled and sequenced by McDermott Sequencing Core using Illumina’s NextSeq500 to yield paired-end (150 bp long) strand-specific reads.

De-multiplexed raw reads were received from McDermott Sequencing Core. Raw reads were first filtered for phred quality and adapters using FASTQC (FastQC, Babraham Bioinformatics, URL: https://www.bioinformatics.babraham.ac.uk/projects/fastqc) and Trimmomatic (Bolger et al., 2014). Filtered reads were then aligned to the reference mouse genome mm10 (https://genome.ucsc.edu) using STAR (Dobin et al., 2013) aligner. Uniquely mapped reads were used to obtain the gene counts using HTSeq package (Anders et al., 2015), and the read counts were normalized using the CPM (counts per million) method implemented in the edgeR package (McCarthy et al., 2012; Robinson et al., 2010). For further analysis, we performed a sample-specific CPM filtering, considering genes with CPM values of 1.0 or greater in all replicates in one or more of the sleep conditions. DESeq2 (Anders and Huber, 2010; Love et al., 2014) was used to detect the DEGs across sleep states. We applied a filter of an adjusted p-value of ≤0.05 and absolute log fold change of ≥0.3 to identify DEGs. Significant DEGs are visualized using volcano plots. UpSetR (Conway et al., 2017) R package was used to identify shared and unique sets of DEGs across sleep states.

GO analysis of the significant, sleep-states shared and unique DEGs was performed using ToppGene (https://toppgene.cchmc.org) (Chen et al., 2009) and GO terms were reduced using REVIGO (Supek et al., 2011). GO categories were considered significant if they contained at least three genes and if they had a Benjamini–Hochberg (B–H)-corrected p-value of ≤0.05.

The mouse visual cortex single cell RNA-seq dataset was downloaded from NCBI Gene Expression Omnibus (accession no. GSE102827) and processed using Seurat (Butler et al., 2018). Retaining cluster annotations from the original source (Hrvatin et al., 2018), genes enriched in each cell-type/cluster were identified using the FindAllMarkers function of Seurat (Butler et al., 2018) with default parameters. The list of MEF2C target genes was obtained from Harrington et al., 2016. The list of rapid primary response genes was obtained from Tyssowski et al., 2018. The list of differentially expressed genes from Mef2c^+/- compared to Mef2c^+/+ at CS was obtained from Harrington et al., 2020. Enrichment analyses for shared and unique list of DEGs across sleep states against (i) cell types from mouse visual cortex single-cell datasets, (ii) MEF2C targets and (iii) rPRGs were performed using a Fisher’s exact test. Also, enrichment analysis for differentially expressed genes across Mef2c^f/f and Mef2c-cKO^Camk2a-Cre under CS and differentially expressed genes from Mef2c^+/- compared to Mef2c^+/+ at CS was performed using a Fisher’s exact test.

Weighted gene co-expression network analysis (WGCNA) was performed on 19 total RNA-seq samples, including six control sleep (CS) samples, six sleep deprived (SD) and seven recovery sleep (RS) samples. The R package for WGCNA (Langfelder and Horvath, 2008) was used to build a gene co-expression network using filtered CPM data (CPM >= 1 across all replicates of one or more of each of the conditions). A signed network was constructed using the blockwiseModules function of the WGCNA R package. A value of 26 was chosen as Beta with highest scale-free R square (R² = 0.796). For other parameters, we used corType = pearson, maxBlockSize = 15000, reassignThreshold = 1 × 10⁻⁵, mergeCutHeight = 0.05, deepSplit = 4, detectCutHeight = 0.999 and minModuleSize = 50. Visualizations of network plots were created using Cytoscape v3.4.0 (Shannon et al., 2003), representing the top 250 edges based on ranked weights.

List of genes per WGCNA module were enriched against (i) DEGs per sleep state, (ii) cell type genes identified in mouse cortex (Hrvatin et al., 2018), (iii) MEF2C targets (Harrington et al., 2016), and (iv) module eigengene correlation to each sleep state using supertest function of SuperExactTest R package. The heat map is generated using ggplot2 functions geom_tile. The modules are ordered based on module eigengene correlation to CS sleep state starting from highest to lowest correlation.

DESeq2 (Anders and Huber, 2010; Love et al., 2014) differential expression analysis for down-sampled WT CS (n = 4) and SD (n = 3) was performed by first randomly down-sampling the WT CS and SD sample sizes to match the sample sizes of Mef2c^f/f. Down-sampling was performed for 100 permutations and number of DEGs were reported for all 100 permutations.

Differential exon usage for exons across Mef2c was accessed using DEXSeq (Anders et al., 2012) indicating only minor variability in exon expression between WT, Mef2c^f/f and Mef2c-cKO^Camk2a-Cre samples. Analysis also shows the specific reduction of floxed exon (Exon 11) in Mef2c-cKO^Camk2a-Cre mice (Figure 4). As expected, this reduction was only partial due to the selective expression of Camk2a-Cre that limited CRE-mediated recombination to CAMK-expressing cells.

Frozen brain tissue from the frontal cortex of mice were immediately homogenized by sonication after addition of 600 mL mRIPA buffer with inhibitors (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% iGEPAL CA-630, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM sodium fluoride, 1 µM cyclosporine A, 1 mM sodium orthovanadate, 50 nM okadaic acid, 1 tablet of cOmplete, EDTA-free Protease Inhibitor Cocktail (Sigma/Roche 11836170001) in deionized water for 10 mL total volume of solution). Samples were then centrifuged at max speed and supernatant was aliquoted. Protein concentration was quantified via Bio-Rad DC Protein Assay (Bio-Rad 5000112). Protein lysates (1.5 mg) were diluted in mRIPA buffer with inhibitors to a total volume of 500 mL. A volume of 2 µL of mouse monoclonal anti-MEF2C antibody (Novus NBP2-00493) was added to each diluted protein lysate and incubated for 2 hr at 4°C. Samples were then added to 75 µL washed protein G Plus/A agarose bead slurry (Millpore IP05) and incubated for 1 hr at 4°C. The samples were then washed three times with mRIPA buffer with inhibitors and eluted with 2X sample buffer + 10% 2-mercaptoethanol.

Immunoprecitation (IP), pre-IP, and post-IP samples were loaded into 7.5% Mini-PROTEAN TGX Stain-Free gels (Bio-Rad 4568026). Gel electrophoresis was run at 250 V for 25 min in 1X Tris-gycine SDS buffer. Proteins were then transferred onto PVDF membranes using the Bio-Rad Trans-Blot Turbo System on the high-MW setting. Membranes were blocked in 1:1 Odyssey blocking buffer/1X PBS for 1 hr. Membranes were incubated overnight with primary antibodies: rabbit anti-phospho-MEF2 (1:200, lab purified) or rabbit anti-MEF2C (1:2500, Abcam ab197070). After washing, Licor secondary antibodies were incubated on the membranes (1:20,000, Licor IRDye 800CW goat anti-rabbit 926–32211). After washing, blots were imaged on the Licor Odyssey CLx system. The pre-IP sample membranes were then incubated with rabbit anti-beta-actin antibody (1:1000, Cell Signaling 4970S) overnight. After washing, the membrane was incubated with the Licor goat anti-rabbit secondary and imaged.

Licor image files were imported into Licor ImageStudio software and protein bands were quantified. We calculated a ratio for phospho-MEF2C signal to the total MEF2C signal for each sample. The average ratio for the control samples was used as a normalization factor for each sample. The ratio for each sample were then divided by the normalization factor to generate normalized signal values for each sample. Similar analysis was used to compare total MEF2C signal to beta-actin signal. One-way ANOVA statistical analyses were performed in Graphpad Prism software.

Brain slices 300 µm thick, cut 15°-angled in rostral direction to keep perpendicular blade penetration in the anterior cingulate area, were prepared from 8- to 12-week-old control (Mef2c fl/fl: Cre-) and Mef2c-cKO^Camk2a-Cre mice at the same Circadian time ZT6 immediately after subjecting to one of the three 6 hr sleep/wake protocols (CS, SD, RS). The dissection was done in ice-cold solution containing (mM): NaCl 84, KCl 3, NaH₂PO₄ 1.25, CaCl₂ 0.5, MgSO₄ 7, NaHCO₃ 26, glucose 20, sucrose 70, ascorbic acid 1.3, kynurenic acid 2 (brought to pH 7.3 by continuous saturation with carbogen gas containing 95% of O₂ and 5% of CO₂). After dissection, slices were transferred to the carbogen gas-saturated artificial cerebrospinal fluid (ACSF) solution containing (mM): NaCl 125, KCl 5, NaH₂PO₄ 1.25, CaCl₂ 2, MgCl₂ 1.3, NaHCO₃ 25, glucose 12 (pH 7.3) at 32° C, and were used for experiments after 1 hr equilibration.

Slices were placed in 400 µL recording chamber and continuously perfused at 2 mL/min with heated (33° C) carbogen gas-saturated ACSF containing 1 µm of adenosine receptor 1 blocker 8-cyclopentyl theophylline (CPT) during entire experiment. Solutions containing additionally 100 µm of picrotoxin (PTX) or 1 µl tetrodotoxin (TTX) were washed-in to block GABA receptor synaptic currents during paired-pulse stimulation experiments, and Na⁺ voltage-dependent channel-generated spontaneous presynaptic spikes during recordings of miniature excitatory post-synaptic currents (mEPSC), respectively.

Pyramidal neurons in cortical layers 2/3 at the anterior cingulate cortex (ACC) area were visually identified with upright microscope equipped with infra-red light sensitive camera, and voltage-clamped at −70 mV holding potential (Cl^- equilibrium potential in our solutions) with patch pipette containing (mM): K-methanesulfonate 130, NaCl 5, MgCl₂ 2, Hepes 10, EGTA 1, CaCl₂ 0.4 (free Ca²⁺ ~80 nM), MgATP 2, Na₂GTP 0.2 (pH ~ 7.3 adjusted with NaOH). Multiclamp 700A amplifier, Digidata 1440A and pClamp 10.2 software (all Molecular Devices, CA) were used for patch-clamp recordings both in voltage-clamp and current clamp modes. The identity of each pyramidal neuron was confirmed by recording a specific pattern of trains of action potentials in current-clamp mode, clearly different from fast-spiking activity of interneurons in response to a series of depolarizing current steps.

CaCl₂ was removed, and 0.25 mM of QX 314, a Na⁺ voltage-dependent channel blocker was added to the patch pipette solution to suppress generation of postsynaptic Na⁺ currents during recordings of evoked excitatory post-synaptic current (eEPSC) in paired-pulse stimulation experiments. In these experiments, the identity of pyramidal neurons was verified immediately after breaking the membrane patch before suppression of action potentials. Stimulation of axons forming synapses at apical dendrites of L2/3 pyramidal cells during paired pulse experiments was done by placing a bipolar stimulation electrode in the L1 area at a distance ~ 100–200 µm from the recorded neuron axial projection to this area. Brief 0.4 ms biphasic stimulation current pulses of 30 to 60 µA delivered with Biphasic Stimulus Isolator BSI-950 (Dagan, MN) were enough to produce minimal-amplitude reproduceable eEPSCs in 95% of experiments.

Functional parameters of miniature excitatory post-synaptic currents (mEPSCs) were calculated from 2 to 4 min continuous recordings following steady-state after application of 1 µl TTX. Recordings were high-pass filtered at 10 Hz and low-pass filtered at 500 Hz. Individual mEPSC were detected by the event template detection algorithm embedded into the pClamp 10 software. A combination of five mEPSCs templates varying from 6 to 25 ms durations was used, each template recognition window being ‘trained’ from ~ 100 real events. No minimal threshold for mEPSC amplitude was applied for medium and longer events to avoid potential exclusion of low-amplitude attenuated mEPSCs originating from distal dendrites. The same template recognition set of parameters was applied for the analysis of data obtained from all cells included in this study. The microscopic parameter values of individual events (peak amplitude, area, instant frequency) were calculated by the pClamp event analysis algorithm and transferred to the Microsoft Excel for secondary off-line general statistics analysis. Standard box and whiskers graphs for each analyzed mEPSC parameter averaged for all cells within a particular experimental group were plotted using GraphPad Prism software, which represents 25% to 75% interval as vertical boxes, the whole range of data points as whiskers, the line inside boxes as a median, and a '+' sign inside boxes indicates the mean value. One-way ANOVA analysis with Tukey’s multiple comparisons test was used to define significant differences within three experimental groups (CS, SD, RS for each genotype) for each reported parameter.

Cumulative histograms for each measured mEPSC parameter of a particular experiment were generated from a continuous 2 to 4 min recording and comprised values falling within a mean ± 3 SD interval (99.7% of events). Within a particular experimental group (CS, SD, RS for each genotype), histograms from individual cells were binned at equal 0.01 steps and then averaged for each binning point over the entire Y-scale range (from 0 to 1). Averaged cumulative histograms were plotted using GraphPad Prism and included mean ± SEM (X-axis error bars) for each binning point. Ordinary one-way ANOVA analysis with Tukey’s multiple comparisons test was used to define significant difference between resultant averaged histograms within each genotype group.

Paired-pulse eEPSC analysis. For each paired-pulse experiment, peak amplitudes of coupled evoked EPSCs (eEPSCs) were obtained after averaging traces of 5 to 10 repetitive paired stimulation sequences separated by 20 s relaxation time. P2/P1 amplitude ratio was calculated for three interpulse intervals: 20, 50 and 100 ms. Standard box and whiskers graphs for each P2/P1 ratio averaged for all cells within a particular experimental group were plotted for each of three interpulse intervals using GraphPad Prism software. Ordinary one-way ANOVA analysis with Tukey’s multiple comparisons test was used to define significant differences within three experimental groups (CS, SD, RS for each genotype) for each measured P2/P1 ratio.

Animals (Mef2c-cKO^Camk2a-Cre n = 7, Mef2c^f/fn = 11, C57BL/6 n = 6, Mef2c^+/- n = 6, Mef2c^+/+ n = 6) were implanted with EEG and EMG electrodes using standard procedures (Bjorness et al., 2016). Briefly, mice were anesthetized with isoflurane and placed into a stereotaxic apparatus after which the hair is sheared and the scalp cleaned and incised. Four small holes were drilled for the placement of bilateral skull screw style electrodes over the frontal cortex (AP +1.7, ML +/- 1.77), parietal cortex (AP −1.7, ML +2.0), and occipital cortex (AP −5.5, ML −1.5). Paddle style EMG electrodes (Plastics One) were placed under the dorsal nuchal muscle. Electrode pins were gathered into a six pin pedestal (Plastics One) which was cemented to the skull using dental cement. The wound was closed using absorbable sutures and coated with triple antibiotic ointment. Mice received buprenorphine for analgesia and were given 14 days to recover from surgical procedures prior to experimental procedures.

Mice were placed into individual cages suspended above the belt of a custom treadmill apparatus with food and water available ad libitum. Mice were given 2 weeks to acclimate to the new environment with daily short duration exposure to the treadmill belt moving after which mice were connected to an amplifier system via a tether attached to the pedestal with an additional week of acclimation to the tether. A balance bar was set between the tether and commutator to allow for unrestricted movement. Following acclimation, EEG and EMG signals were acquired using a 15LT amplifier system (Natus Neurology) for 2 days under baseline (undisturbed) conditions followed by 2 days of chronic, partial sleep deprivation (Bjorness et al., 2016). Sleep deprivation consisted of eight cycles of 4 hr TM 'ON' (belt moving) followed by 2 hr TM 'OFF' (belt fixed). State assignment was determined via a custom Matlab (Mathworks) autosorter program using standard criteria (Bjorness et al., 2016) for waking, SWS, and REM after which files were manually checked to ensure correct assignment of state and flag epochs featuring artifact. Epochs were scored in 10 s bins with artifact-flagged epochs excluded from power spectral analysis. Episodes began with 30 consecutive seconds of one state and ended with 30 consecutive seconds of a different state. Power spectrum values were calculated in a 2 s window with a 1 s overlap and a Hamming window using the mean squared spectrum function in Matlab. For statistical analyses, Mef2c-cKO^Camk2a-Cre was compared to Mef2c^f/f and C57BL/6, while Mef2c^+/- was compared to Mef2c^+/+.

SWA (0.5–4.5 Hz) was normalized to the 24 hr average SWA and the percentage change between the first and last consolidated 1 hr sleep period was calculated as previously described (Nelson et al., 2013). The baseline day with the clearest consolidated sleep periods early and late in the light phase were used for statistical comparison via a one-way ANOVA with Sidak’s multiple comparison test.

For a full description of the decay analysis see Bjorness et al., 2016. Briefly, SWS-SWA power was normalized to W SWA power after which SWS episodes of at least 5 min in duration were collected and gathered into an excel file with one episode per column such that the normalized SWS-SWA power of the first epoch was in the first row of each column, the normalized SWS-SWA power of the second epoch was in the second row of each column, and so forth. Based on the 5 min duration criteria, all columns had at least 30 rows; the longest four episodes were truncated to leave five episodes at the longest time point. One Mef2c-cKO^Camk2a-Cre, one Mef2c^f/f, and one Mef2c^+/+ mouse per group was excluded on the basis of insufficient episodes of at least 12 min in duration (Bjorness et al., 2016). The natural log was taken after which episodes were averaged within animal to create a single averaged SWS episode per mouse followed by averaging across animals within group to result in a single episode per group. Next, the group average episode was transformed by taking the exponential. A fragment of the average was created (excluding the initial ascending phase and periods of high variability toward the end) and fit using a single phase decay (GraphPad Prism) in order to determine the time constant of decay (τ). Based on previously described exclusion criteria (absolute sum of squares, Bjorness et al., 2016), the averaged group episode for Mef2c-cKO^Camk2a-Cre mice could not be fit. Thus, the time constant of decay determined for the Mef2c^f/f group was used to determine the plateau and R² value for the Mef2c-cKO^Camk2a-Cre group.

Mice underwent 8 cycles of 4 hr SD (TM ‘ON’) followed by 2 hr RS (TM ‘OFF’) starting at ZT0. SWS-SWA power was averaged in 2 hr bins. Rebound SWS-SWA was determined by the percent change in SWS-SWA from baseline conditions to TM ‘OFF’ conditions using time-matched circadian bins (i.e. ZT 4–6, ZT 10–12, ZT 16–18, ZT 22–24). A one-way ANOVA was used to compare average rebound across the 8 RS periods and a One sample T test was used to compare average values to a theoretical mean of 0 (GraphPad Prism).

Time in each sleep/waking state (waking, SWS, REM) under baseline conditions was averaged in 2 hr bins and calculated as the percent time of the total period. Averaged percent time in state for each state was compared between groups using a one-way ANOVA (GraphPad, Prism).

SWS episodes under baseline and SD (recovery periods) conditions were binned by duration using the histogram function in excel (Microsoft); the first bin was 90 s with successive bins increasing by 40 s. The number of episodes in each bin was divided by the total number of episodes for the condition and multiplied by 100 to get a percentage of the total episodes represented by each bin. Next, a cumulative percentage was calculated by adding the percentage of successive bins. A two-way ANOVA (condition, bin) was used to compare cumulative histogram episode durations across groups and conditions (baseline and SD).

Under baseline conditions, individual 10 s epochs were separated by state (waking, SWS, REM) after which spectral power in each 1 Hz bin up to 50 Hz was normalized by the total power for that state (waking epoch normalized to waking total power, etc). Fraction power was then averaged across epochs within each 1 Hz bin to get an average spectral distribution from 1 to 50 Hz by state. Two-way ANOVA (group, frequency) was used to compare spectral distributions by state across groups.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank Mr. To Thai, Ms. Lilian Zhan and Dr. Busra Goksu for technical assistance and Dr. Stefano Berto for discussions about genomic analyses. Funding: This study was funded by NIH grant R01 MH080297 and WPI program for IIIS to RWG; NIH grants RO1 MH102603, DC014702 and the James S McDonnell Foundation 21^st Century Science Initiative in Understanding Human Cognition – Scholar Award (220020467) to GK; NIH grants R01 AG045795, R01 NS 106657 to JST; JST is an Investigator in the Howard Hughes Medical Institute; NIH grant RO1 MH111464 to CWC and F30 HD098893 to CMB. The contents do not represent the views of the U.S. Department of Veterans Affairs or the United States Government.

Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to approved institutional animal care and use committee (IACUC) protocols of the UT Southwestern Medical Center and the North Texas VA Health Care System. All surgery was performed under isoflurane anesthesia, and every effort was made to minimize suffering.

1. Received: April 27, 2020
2. Accepted: August 26, 2020
3. Accepted Manuscript published: August 27, 2020 (version 1)
4. Version of Record published: September 14, 2020 (version 2)

© 2020, Bjorness et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.