Diurnal proteome profile of the mouse cerebral cortex: Conditional deletion of the Bmal1 circadian clock gene elevates astrocyte protein levels and cell abundance in the neocortex and hippocampus

Circadian oscillators, defined by cellular 24 h clock gene rhythms, are found throughout the brain. Cerebral cortex‐specific conditional knockout of the clock gene Bmal1 (Bmal1 CKO) leads to depressive‐like behavior, but the molecular link from clock gene to altered behavior is unknown. Further, diurnal proteomic data on the cerebral cortex are currently unavailable. With the aim of determining the diurnal proteome profile and downstream targets of the cortical circadian clock, we here performed a proteomic analysis of the mouse cerebral cortex. Proteomics identified approximately 2700 proteins in both the neocortex and the hippocampus. In the neocortex, 15 proteins were differentially expressed (>2‐fold) between day and night, mainly mitochondrial and neuronal plasticity proteins. Only three hippocampal proteins were differentially expressed, suggesting that daily protein oscillations are more prominent in the neocortex. The number of differentially expressed proteins was reduced in the Bmal1 CKO, suggesting that daily rhythms in the cerebral cortex are primarily driven by local clocks. The proteome of the Bmal1 CKO cerebral cortex was dominated by upregulated proteins expressed in astrocytes, including GFAP (4‐fold) and FABP7 (>20‐fold), in both the neocortex and hippocampus. These findings were confirmed at the transcript level. Cellular analyses of astrocyte components revealed an increased number of GFAP‐positive cells in the Bmal1 CKO cerebral cortex. Further, BMAL1 was found to be expressed in both GFAP‐ and FABP7‐positive astrocytes of control animals. Our data show that Bmal1 is required for proper cellular composition of the cerebral cortex, suggesting that increased cortical astrocyte activity may induce behavioral changes.


| INTRODUCTION
Circadian rhythms, that is, 24 h endogenous biological rhythms, are maintained by a molecular circadian clock that enables living organisms to efficiently synchronize physiological functions to the light regime (Takahashi, 2017). Anatomically, the master clock of mammals resides in neurons of the suprachiasmatic nucleus (SCN), but the molecular clock, as defined by oscillating clock gene expression, is a feature of most if not all cell types (Guilding & Piggins, 2007). In this regard, oscillating circadian clock gene expression has been reported in all major regions of the brain, including neurons of the neocortex (Rath et al., 2013;Rath et al., 2014) and the hippocampus (Honma et al., 1998;Jilg et al., 2010). While rhythmic clock gene expression in the cerebral cortex seems to depend on a circadian input from the SCN (Rath et al., 2013), the functional role of these extrahypothalamic molecular brain oscillators is far less understood. Evidence from studies in humans shows a clear correlation between major depressive disorder and altered clock gene expression in both neocortical and hippocampal areas of the cerebral cortex (Li et al., 2013), suggesting that local changes in clock gene accompanies development of depression; however, the link between brain disorders and circadian dysfunction is largely correlational (Logan & McClung, 2019).
To clarify a possible causal link from disrupted clock gene expression in extrahypothalamic circadian oscillators to altered behavior, we previously generated a mouse strain in which the canonical clock gene Bmal1 (basic helix-loop-helix ARNT like 1, previously known as Arntl) was specifically deleted in the cerebral cortex by use of Cre-LoxP technology (Bering et al., 2018). In this Bmal1 conditional knockout (CKO) model, the Cre-recombinase is driven by the promoter of Emx1 (Guo et al., 2000), which allows us to specifically disrupt the circadian clock in the cerebral cortex, including the neocortex and the hippocampus, while leaving the master clock of the SCN intact (Bering et al., 2018). The Bmal1 CKO showed only minor changes in circadian locomotor rhythmicity, whereas a phenotype with depressive-like behavior and reduced monoamine signaling was evident, suggesting that a circadian malfunction of the clock of the cerebral cortex is a causal component in development of depressive disorder (Bering et al., 2018;Li et al., 2013); however, the molecular connection from the circadian clock of the cerebral cortex to affective behavior still represents a missing link in our understanding of the circadian system of the brain.
Circadian clock gene expression profiles in the cerebral cortex, that is, both the neocortex and the hippocampus, are well documented; also, the daily profiles of gene expression in the mouse neocortex and the hippocampus have been analyzed by microarray and RNA sequencing (Hor et al., 2019;Lananna et al., 2018;Renaud et al., 2015). However, large-scale screening efforts of protein contents of the mouse cerebral cortex so far did not include temporal profiles (Sharma et al., 2015); therefore, general day-night changes in the murine neocortical and hippocampal proteome, which would represent a first-step towards identifying clock-controlled proteins, still represent a gap in our knowledge on circadian biology of the cerebral cortex.
With the purpose to determine daily changes in protein expression in the cerebral cortex and to identify downstream molecular targets of the Bmal1-dependent circadian clock, we here performed proteomic analyses of the neocortex and the hippocampus of normal mice and a Bmal1 CKO model, respectively. The results of these efforts, which would represent an initial link between a disrupted local clock and altered behavior, are presented below.
For proteomic analyses, a total of 26 mice were sacrificed by decapitation; the neocortex and the hippocampus were dissected and immediately frozen on dry ice; animals included Bmal1 CKO mice sacrificed at Zeitgeber time (ZT) 6 (3 male, 5 female, 24-76 weeks of age), Bmal1 CKO mice sacrificed at ZT18 (3 male, 4 female, 20-63 weeks of age) and control mice sacrificed at ZT6 (3 male, 2 female, 18-65 weeks of age) and ZT18 (2 male, 4 female, 18-63 weeks of age), respectively. For quantitative reversetranscription real-time PCR (qRT-PCR), a total of 48 animals were kept in darkness (DD) for 2 days and sacrificed by decapitation in threehour intervals throughout the presumptive day and night; the neocortex and the hippocampus were dissected and immediately frozen on dry ice; animals included 24 Bmal1 CKO mice (11 male, 13 female, 13-25 weeks of age) and 24 control mice (15 male, 9 female, 13-23 weeks of age). For radiochemical in situ hybridization, a total of 38 mice were sacrificed by decapitation in four-hour intervals; the brains were removed and frozen in crushed dry ice; animals included 20 Bmal1 CKO mice (10 male, 10 female, 14-17 weeks of age) and 18 control mice (12 male, 6 female, 13-19 weeks of age). Samples from mice included in the qRT-PCR and in situ hybridization analyses presented here have also been used in a previous publication (Bering et al., 2018). For immunohistochemistry, animals were anesthetized with tribromoethanol (400 mg/kg) at ZT5 and sacrificed by transcardial perfusion with PBS (with heparin) followed by 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7). The brains were removed, postfixed for 48 h in the same fixative, cryoprotected in 25% sucrose in PBS for 48 h and finally frozen in crushed dry ice; animals included four Bmal1 CKO mice (3 male, 1 female, 10-13 weeks of age) and four control mice (1 male, 3 female, 9-22 weeks of age).

| Proteomics
The experimental approach is summarized in Figure 1. Samples were prepared for analysis with the suspension trapping method (Zougman et al., 2014) using S-Trap spin columns (Protifi). Label-free quantification (LFQ) liquid chromatography-tandem mass spectrometry (LC-MS/MS) was performed with the universal methods settings, essentially as previously described (Ludvigsen et al., 2020). In short, 1 μg of each sample was run in triplicate on a Dionex Ultimate™ 3000RSLC system connected to an Orbitrap Fusion Tribrid mass spectrometer with an EasySpray™ ion source (Thermo Scientific). Max-Quant software version 1.6.6.0  was used to perform a search of the raw data files against the reviewed UniProt Mus musculus database (downloaded on September 2, 2019).
The ProteinGroup data file was subsequently analyzed in Perseus software version 1.6.6.0 (Tyanova, Temu, Sinitcyn, et al., 2016). Identification of a protein required at least two unique peptides; quantitative values were log 2 -transformed. Proteins were identified in at least 70% of the samples in each group compared.
Pathway analyses were performed using the online PANTHER Overrepresentation Test (pantherdb.org) in annotation version GO Ontology database DOI: http://doi.org.10.5281/zenodo.6799722 released July 1, 2022. Differentially expressed proteins were tested for overrepresentation of the annotation data sets "GO biological process complete" and "GO molecular function complete." These proteins were tested against two separate reference lists, the full Mus musculus gene database provided in PANTHER, but also the list of proteins detected in the mass spectrometry. Analyses were set to use Fisher's Exact test type corrected for calculated false discovery rate.

| Quantitative RT-PCR
qRT-PCR was performed as previously described (Bering et al., 2017;Bering et al., 2018;Rath et al., 2014). Briefly, RNA was isolated from neocortices and hippocampi by use of TRIzol reagent, DNase treated, F I G U R E 1 Flowchart of experimental approach. Figure was created by use of BioRender. and 500 ng was reverse transcribed by use of Superscript III (Invitrogen).
PCR reactions were run in a Lightcycler 96 (Roche) with 10 μL reactions containing 0.5 μM primers (Table 1), FastStart Essential DNA Green Master (Roche) and 0.2 μL cDNA. Transcript copy numbers were determined by use of internal 10-fold dilutions of pUC57 plasmids containing the sequence to be amplified (GenScript) and normalized against the geometric mean of Actb and Gapdh copy numbers.

| Radiochemical in situ hybridization
Radiochemical in situ hybridization was performed as previously described (Rath & Moller, 2022). In brief, DNA oligo probes targeting two regions of the transcript of interest ( Cryostat sections (12 μm) of unfixed brains mounted on slides were fixed in formaldehyde, acetylated, dehydrated in a graded series of ethanol, and delipidated in chloroform. Sections were incubated with the labeled probe overnight at 37 C, washed in SSC, and exposed to an x-ray film for 3 weeks. Autoradiographs were digitized and optical densities of the neocortex and the dorsal hippocampus were measured in Scion Image and converted to dpm/mg tissue by use of simultaneously exposed 14 C standards (Amersham).

| Immunohistochemistry
Immunohistochemistry was performed on free-floating cryostat sections (40 μm) as previously described (Rath et al., 2014). In brief, sections were blocked in serum and incubated in different combinations of primary antibodies against FABP7, GFAP, and BMAL1 overnight (Table 3). Sections were subsequently incubated in fluorophore-conjugated secondary antibodies (Table 3) or, in case of FABP7, in a biotinylated secondary antibody, followed by ABC Vectastain (Vector Laboratories), tyramide signal amplification (TSA) (Perkin Elmer) and a streptavidin-fluorophore conjugate (Table 3). All sections were finally stained in DAPI before mounting on slides. Sections from at least three animals of each genotype (Bmal1 CKO and controls) were run and analyzed in parallel. Sections were photographed in an LSM980 inverted laser scanning confocal microscope (Zeiss). Adjustments of light and contrast and image overlays were done using ZEN2.3 pro software (Zeiss); images from all animals were processed using identical settings.
The specificity of the antibodies against GFAP and BMAL1 has been previously tested (Rath et al., 2014). However, to test the specificity of the antibody against FABP7, the diluted antibody was preincubated with the immunogen mouse BLBP peptide (Abcam, catalogue number ab32422) at a concentration of 100 μg/mL for 6 days prior to incubation of the sections. Preabsorption completely abolished the immunohistochemical signal in both the neocortex and hippocampus ( Figure S2).
To determine the number of GFAP-positive cells, the dentate gyrus or layer two of the neocortex located above the dentate gyrus were identified in DAPI staining, the filter set of the microscope was switched without changing the position of the slide, and the immunohistochemical signal was photographed. Three brain sections from each mouse were photographed (total of six photographs per mouse). Each photograph corresponded to a tissue volume of 561 μm (horizontal) Â 361 μm (vertical) Â 40 μm (section thickness). GFAP-immunopositive cell bodies in the images were counted blinded. In the same photographs, GFAPsignal intensity was measured in individual cell bodies by use of Scion Image Beta version 4.0.02 (Scion), also blinded.

| Statistics
Quantitative proteome data (LFQ data) were analyzed in Perseus software version 1.6.6.0 by Student's t-test. An uncorrected two-tailed p-value of .01 was considered to represent statistical significance.
T A B L E 1 qRT-PCR primer sequences.

| RESULTS
3.1 | Mitochondrial, cytoskeletal, and neuronal plasticity proteins are differentially expressed between day and night in the mouse cerebral cortex, but rhythmic expression mainly depends on local expression of Bmal1 The proteome of the mouse cerebral cortex has been previously reported (Sharma et al., 2015), but the daily profile is unknown. As a first step towards characterizing proteins potentially regulated by the local circadian clock of the cerebral cortex, we therefore identified the proteome of the neocortex and hippocampus from both control and Bmal1 CKO mice at ZT6 and ZT18 (Tables S1-S4)  In the neocortex of the Bmal1 CKO mouse, four proteins (COX6A1, DHRS1, SLC4A1, and CSTB) were found to be differentially  (Table 4), such as steroid metabolism (DHRS1), ion transport (SLC4A1) and proteinase inhibition (CSTB). Using the criteria detailed above, differential protein expression between day and night was not detected in the hippocampus of the Bmal1 CKO mouse (Figure 2d). In summary, this part of our proteomic analysis suggests that the daynight differences in the proteome of the neocortex are more abundant than those of the hippocampus and that rhythmic protein expression to some extent requires an intact local circadian clock in the brain region itself.
3.2 | Proteins known to be expressed in astrocytes are upregulated in both the neocortex and hippocampus of the Bmal1 CKO mouse To identify proteins differentially expressed between genotypes, proteomic data from samples collected from both the neocortex and hippocampus at ZT18 and ZT6 were compared between the Bmal1 CKO mouse and the control (Tables S5-S8). Again, cut-off was set at >2-fold changes combined with p-values <.01 (Figures 3b and 4, Table 5). Fourteen proteins (FABP7, GFAP, MTHFD1L, TUBB6, F I G U R E 2 Day-night changes in the proteome of the mouse cerebral cortex. Volcano plots of proteomic analyses of differential expression of proteins between night (ZT18) and day (ZT6) in the neocortex of the control mouse (a), the hippocampus of the control mouse (b), the neocortex of the Bmal1 CKO mouse (c), and the hippocampus of the Bmal1 CKO mouse (d). For each identified protein, the negative log 10 of the p-value (Student's t-test) is plotted against the log 2 of the detected fold change calculated as the mean value at ZT18 divided by the mean value at ZT6. Proteins with p < .01 and fold change >2.0 were regarded as being differentially expressed and were included in further analyses (Table 4); shaded areas indicate these cut-off values with proteins that were not included in further analyses. In both the neocortex and the hippocampus and at both time points examined, FABP7, a protein involved in radial glial cell function, and GFAP, a cytoskeletal protein in astrocytes, were found to be highly upregulated in the Bmal1 CKO. GFAP is a well-established astrocyte marker (Middeldorp & Hol, 2011), while FABP7 in addition to oligodendrocyte precursor cells is also expressed in astrocytes of the adult rodent brain (Foerster et al., 2020;Schmid et al., 2006;Sharifi et al., 2011;Sharifi et al., 2013), thus suggesting that astrocytes may be heavily affected by the knockout of the circadian clock. On the other hand, mitochondrial proteins (MTHFD1L and ACAA2) were generally downregulated in the neocortex and hippocampus of the Bmal1 CKO.
In addition, the enzyme ENTPD2, which is involved in extracellular signaling during neurogenesis (Gampe et al., 2015), was upregulated in both the neocortex and the hippocampus. Other proteins were affected by Bmal1 knockout in one brain region only (Figure 4, Table 5).
In the neocortex, the cytoskeletal protein TUBB6, which was identified as being rhythmic in the control mouse with upregulation during nighttime ( Figure 2a, Table 4), was found to be upregulated in the daytime samples of the Bmal1 CKO neocortex (Figure 4a, Table 5) reflecting the loss of daily rhythm in the Bmal1 CKO cerebral cortex. Similarly, GNG4, a G-protein, and DBI, which is involved in lipid metabolism, both exhibited differential diurnal expression in the control mouse with F I G U R E 3 Summary of proteomic data. (a) Day-night changes in the proteome of the neocortex and hippocampus in the control and Bmal1 CKO mouse, respectively. Data from Figure 2 and Table 4 displayed as a Venn diagram and a heatmap. Rhythmic proteins are mainly detected in the neocortex of the control mouse.
(b) Bmal1-dependent changes in the proteome of the neocortex and hippocampus at ZT6 and ZT18, respectively. Data from Figure 4 and Table 5 displayed as a Venn diagram and a heatmap. FABP7 and GFAP are highly upregulated in the Bmal1 CKO at both timepoints in both tissues. Black cells in the heat maps indicate that the protein was not detectable.
upregulation during daytime (Figure 2a; Table 4), but were found to be specifically upregulated in nighttime samples of the Bmal1 CKO neocortex ( Figure 4b, Table 5), again seemingly reflecting a loss of protein rhythms in the Bmal1 CKO cerebral cortex as opposed to a general continuous effect on the proteins themselves. Daytime protein levels were specifically affected for proteins involved in synapse structure (PLXND1 and CAMKV) (Figure 4a, Table 5), whereas an enzyme controlling pH homeostasis (CA4) was upregulated in both day-and nighttime samples of neocortex of the Bmal1 CKO mouse.
In addition to the proteins affected across both brain regions described above (FABP7, GFAP, MTHFD1L, ACAA2, and ENTPD2), three proteins (ANXA4, NOS1, and VIM) were found to be differentially expressed between genotypes specifically in the hippocampus, all of which were affected in daytime-samples only (Figure 4c, T A B L E 4 Differential expression of proteins between day (ZT6) and night (ZT18) in neocortical and hippocampal samples collected from Bmal1 CKO and control mice, respectively.  Table 5). These included functionally diverse proteins involved in exocytosis (ANXA4) and nitric oxide synthesis (NOS1), as well as cytoskeletal components (VIM).

| Transcripts known to be detectable in astrocytes are upregulated in the neocortex and hippocampus of the Bmal1 CKO mouse
The prominent effect of Bmal1 deletion on the proteins FABP7 and GFAP, both of which are expressed in astrocytes (Middeldorp & Hol, 2011;Schmid et al., 2006;Sharifi et al., 2011), as well as previous reports on the connection between Bmal1 and astrocyte biology in other model systems (Lananna et al., 2018;Musiek et al., 2013), prompted us to determine the expression of these proteins in detail at the transcript level both with respect to circadian profiles and genotypical differences. Since DBI is mainly expressed in astrocytes (Tonon et al., 1990;Tonon et al., 2020), and GNG4 is detectable in fetal astrocytes (Bonham et al., 2018), the transcripts of these proteins, which were both identified as being upregulated in the neocortex of the Bmal1 CKO mouse (Figures 3b and 4b; Table 5), were also included in the analyses.
To determine circadian profile of Fabp7, Gfap, Dbi, and Gng4, qRT-PCR was performed on neocortical and hippocampal samples from the Bmal1 CKO and control mice collected in 3 h intervals ( Figure 5). All four transcripts were significantly upregulated in both cortical areas of the Bmal1 CKO (p-values <.01, two-way ANOVA), whereas circadian rhythms were not detected in either genotype (Table 6). Fabp7 has been previously shown to exhibit a diurnal rhythm in the mouse hippocampus (Gerstner et al., 2008), and analyzing the expression profile in control mouse separately confirmed these findings with a just significant circadian variation in the hippocampus (p = .036, one-way ANOVA) ( Figure S4).

F I G U R E 4
Bmal1 dependent changes in the proteome of the mouse cerebral cortex. Volcano plots of proteomic analyses of differential expression of proteins between the Bmal1 CKO and control mouse in the neocortex at ZT6 (a), the neocortex at ZT18 (b), the hippocampus at ZT6 (c), and the hippocampus at ZT18 (d). For each identified protein, the negative log 10 of the p-value (Student's t-test) is plotted against the log 2 of the detected fold change calculated as the mean value in the Bmal1 CKO divided by the mean value in the control. Proteins with p < .01 and fold change >2.0 were regarded as being differentially expressed and were included in further analyses (Table 5); shaded areas indicate these cut-off values. n = 5-8 in each experimental group.
The overall transcript levels of Fabp7 were increased by 18.4-fold and 10.2-fold in the Bmal1 CKO neocortex and hippocampus, respectively, as compared to controls (Table 6). Gfap expression was increased by 3.6-fold in both brain areas, whereas Dbi and Gng4 transcript levels were upregulated by approximately 2.2-and 1.5-fold, respectively (Table 6). An increased abundance of Fabp7 and Gfap transcripts specifically in the cerebral cortex, including both the neocortex and the hippocampus, was also detectable by radiochemical in situ hybridization (Figure 6a T A B L E 5 Differential expression of proteins in between genotypes, that is, Bmal1 CKO versus control, in neocortical and hippocampal samples collected at midnight (ZT18) and midday (ZT6), respectively.   Table 6.  (Bin et al., 2016), microglia (Boeck et al., 2020), and neurons (Mullen et al., 1992) did not reveal differences between genotypes in cell numbers in the neocortex and hippocampus ( Figure S5). These findings are in accord with the our proteomic analyses which did not identify genotype-dependent differences in any of the detected molecular markers (Matson et al., 2022;Russ et al., 2021) of neurons, microglia, or oligodendrocytes, including oligodendrocyte precursor cells (Table S9).
BMAL1 is known to be expressed in neurons of the murine neocortex and hippocampus (Rath et al., 2014); however, FABP7, which is upregulated in the Bmal1 CKO, did not colocalize with the neuronal marker NeuN ( Figure S6). Therefore, to establish if BMAL1 was also present in cortical astrocytes expressing GFAP and FABP7, immunohistochemical staining were performed on sections of control mice (Figure 8). In both the neocortex and the hippocampus, cells expressing both BMAL1, GFAP and FABP7 were identified, suggesting that intrinsic BMAL1 can influence the biology of individual cortical astrocytes. Notably, BMAL1 was undetectable in astrocytes in the neocortex and hippocampus of the Bmal1 CKO mouse ( Figure S7).
T A B L E 6 Summary of quantitative analyses of the expression profiles of Fabp7, Gfap, Dbi, and Gng4 in the neocortex and hippocampus of the Bmal1 CKO and the control mouse determined by qRT-PCR. Difference between genotypes and circadian changes were determined by two-way ANOVA followed by Bonferroni's multiple comparisons post hoc tests. b Gene expression levels were determined by calculating the overall mean of the mean values at each timepoint analyzed. c Fold change between genotypes was determined as the ratio between the overall mean values of the Bmal1 CKO and the control mouse from the same brain area.

| DISCUSSION
The cerebral cortex contains a circadian oscillator as evidenced by rhythmic expression of clock genes in both the neocortex and the hippocampus (Jilg et al., 2010;Rath et al., 2014). We here identified proteins differentially expressed between day and night in the neocortex and hippocampus of the mouse and further determined downstream targets of the cortical circadian oscillator by proteomic analysis of a cerebral cortex-specific Bmal1 clock gene knockout mouse.
In the current study, 15 day-night differentially expressed proteins were identified in the neocortex, while only three were identified in the hippocampus, indicating that the hippocampus is generally less rhythmic than the neocortex, as also seen in analyses of clock gene expression profiles in the mouse, where the amplitude is higher in neocortex as compared to the hippocampus (Bering et al., 2018). In the mouse neocortex, a number of mitochondrial respiratory chain proteins were upregulated during daytime. Enrichment of mitochondrial respiratory chain proteins has also been reported in the rhythmic proteome of the mouse SCN, in which the respiratory chain proteins were mainly upregulated during daytime (Chiang et al., 2014). In the cerebellum, a pattern with upregulation during daytime is also in agreement with our data, although other mitochondrial respiratory chain components were identified as being rhythmic (Plumel et al., 2019). In the liver, proteins of the mitochondrial respiratory chain are also rhythmic, but the rhythms are more complex in that individual components are not uniformly distributed across the 24 h cycle and are not synchronized to the neocortical proteins identified in the current study (Neufeld-Cohen et al., 2016). Previous proteomic day-night analyses of the rat pineal gland also identified mitochondrial proteins, including one mitochondrial respiratory chain protein, as being preferentially expressed during daytime (Moller et al., 2007), but this was not the case in a similar analysis of the mouse retina (Moller et al., 2017).
In the neocortex, we also identified a number of rhythmic neuroplasticity proteins upregulated during daytime. Studies on time of day-dependent neurogenesis and plasticity of the adult brain has mostly focused the hippocampus and its role in learning and memory (Ali & von Gall, 2022;Snider et al., 2018), whereas little is known about the diurnal neuroplasticity processes in the mature neocortex (Maret et al., 2011). In this regard, our data may represent a first step towards identifying the molecular basis for circadian regulation of neocortical plasticity.
Clock gene-encoded proteins were not detected in our proteomic analyses, probably reflecting that expression levels are too low to F I G U R E 8 Colocalization of BMAL1, FABP7, and GFAP in the neocortex and hippocampus of the control mouse. Immunohistochemistry for detection of BMAL1, GFAP, and FABP7 was performed on the same sections. Colocalization of all three proteins were detected in astrocytes of the neocortex (arrows) and the hippocampus (arrow heads). Images of the second layer of the neocortex and the dentate gyrus are displayed. n = 3. Scale bar, 50 μm.
generate a significant protein signal. As described in the introduction, immunohistochemical analyses have identified clock gene-encoded proteins in the cerebral cortex, and large screening efforts at the transcript level also reported rhythmic clock gene expression in the cerebral cortex (Renaud et al., 2015). This apparent discrepancy seems to reflect a limitation of the methodology; thus, even in the pineal gland, a central structure in circadian biology, proteomic analyses in the rat failed to identify clock gene-encoded protein products (Moller et al., 2007), while the corresponding transcripts were detected by microarray analysis (Bailey et al., 2009). Most importantly, the local circadian oscillator of the neocortex seems to play a key-role in governing differential day-night expression of proteins, as evidenced by the marked decline in the number of day-night differentially expressed proteins in the Bmal1 deficient neocortex in the current study.
Apart from the reduced representation of rhythmic proteins, a striking feature of the proteome of the Bmal1 CKO is the upregulation of proteins known to be expressed in astrocytes. FABP7, which is expressed in neural stem cells and radial glia, but seems to persist in astrocytes and oligodendrocytes precursors of the adult brain (Kipp et al., 2011;Schmid et al., 2006;Sharifi et al., 2011), was upregulated by more than tenfold in both the hippocampus and neocortex at both transcript and protein level in the Bmal1 CKO. Upregulation of GFAP, a well-established astrocyte marker (Middeldorp & Hol, 2011), was fourfold, but this protein still represents the second highest level of differential expression between genotypes in both brain regions. Our finding that BMAL1, FABP7, and GFAP are co-expressed in astrocytes of the neocortex and hippocampus of control mice suggests that BMAL1 may act locally in cortical astrocytes to repress activation.
These results support activation of astrocytes as a general feature of CNS model systems with knockout of Bmal1, since a modest upregulation of Fabp7 transcripts in the hypothalamus of global Bmal1 knockout mice was recently reported (Gerstner & Paschos, 2020), whereas the relation between a knockout of BMAL1 and activation of GFAP-positive astrocyte has been previously reported both in vitro and in vivo (Lananna et al., 2018;Musiek et al., 2013).
Despite the apparent connection of GFAP to the circadian clock, GFAP is not rhythmic in the mouse SCN (Moriya et al., 2000) and although day-night differences have been reported in the rat SCN (Gerics et al., 2006;Hajos, 2008), our data establish the non-rhythmic nature Gfap expression in both the neocortex and the hippocampus of the mouse. Despite a rhythm in the hippocampus as reported here and elsewhere (Gerstner et al., 2008), neocortical Fabp7 is also continuously expressed throughout the 24 h circadian cycle in the neocortex. Thus, the clear connection of Gfap and Fabp7 to the circadian clock is not necessarily translated into a down-stream circadian rhythm. Clock-dependent mechanisms involving activation of astrocytes, as reflected by the upregulation of GFAP and FABP7, seem to play a key-role in neurotoxicity and neurodegenerative diseases (Killoy et al., 2020;Musiek & Holtzman, 2016), which may in turn induce the depressive phenotype of the Bmal1 CKO.
Our proteomic analyses expanded the list of Bmal1-regulated astrocytic proteins (Gerstner & Paschos, 2020;Lananna et al., 2018;Musiek et al., 2013) to include GNG4 and DBI (Bonham et al., 2018;Tonon et al., 1990); in addition to the increase caused by of deletion of Bmal1, both of these proteins were found to be upregulated at night in the proteomic analysis of the control neocortex, but did not display a circadian rhythm at the transcript level, suggesting that rhythms are regulated at the posttranscriptional level. ENTPD2, a protein identified as being upregulated in our proteomic analysis of the Bmal1 CKO cerebral cortex, has also been identified in rat astrocytes (Wink et al., 2006), and VIM, another protein upregulated specifically in the hippocampus of the Bmal1 CKO, has also been localized to astrocytes (Schnitzer et al., 1981), which is in accord with our finding of an increased abundance of GFAP-positive astrocytes in both the hippocampus and neocortex of the Bmal1 CKO mouse. On the other hand, previous investigations found that FABP7 is downregulated in the cerebellum of Per1/Per2 global mutant mice (Plumel et al., 2019), whereas Gfap expression seems to be unaffected in the cerebral cortex (Lananna et al., 2018). Therefore, the marked upregulation of astrocyte markers following Bmal1 knockout does not seem to be a general consequence of interfering with the molecular circadian clock.
In the Bmal1 CKO mouse model, we have previously documented depressive-like symptoms (Bering et al., 2018), and now we see a correlation with increased astrocyte activity specifically in the hippocampus and the neocortex. On the contrary, the general picture in humans diagnosed with major depressive disorder seems to be a reduction in astrocyte density in the cerebral cortex (Cobb et al., 2016;O'Leary et al., 2021;Rajkowska & Stockmeier, 2013), but astrocyte hypertrophy has also been reported (Torres-Platas et al., 2011); this points to the possibility that dysregulation rather than absolute levels of cortical astrocyte activity and density may induce a depressive phenotype. Recent data suggest that disruption of cortical astrocyte calcium-signaling induces a depressive-like behavior accompanied by synaptic structure abnormalities in mice (Luo et al., 2023), which may be part of the mechanistic explanation linking astrocytes to depression. Another CKO mouse model, generated by use of a Cre-construct driven by the CamKII-promoter to knock out Bmal1 in all regions of the forebrain, while cellularly limiting the deletion to excitatory neurons, did not exhibit any affective symptoms (Snider et al., 2016). This indicates that the depressive-like behavior observed in the phenotypic characterization of the mouse analyzed in the current study (Bering et al., 2018) is not caused by Bmal1 deletion in these neurons. On the other hand, Emx1 which was driving the Cre used to knock out Bmal1 is known to be expressed in neurons, including excitatory neurons, and astrocyte precursors of the developing cerebral cortex (Gorski et al., 2002). Therefore, we speculate that the cellular distribution of the Bmal1 knockout, including astrocytes, may be a defining feature of the reported differences between the behavioral phenotypes of the two otherwise seemingly similar mouse models (Bering et al., 2018;Snider et al., 2016), again supporting the interpretation that malfunction of cortical Bmal1-positive astrocytes represents a step towards a depressive-like phenotype.