Comparative transcriptome analysis and CRISPR/Cas9 gene editing reveal that E4BP4 mediates lithium upregulation of Per2 expression

Bipolar disorder (BPD) is a psychiatric disorder characterized by alternate episodes of mania and depression. Disruption of normal circadian clock and abnormal sleep cycles are common symptoms of BPD patients. Lithium salt is currently an effective clinical therapeutic drug for BPD. Animal and cellular studies have found that lithium salt can upregulate the expression of the clock gene Per2, but the mechanism is unknown. We aim to understand the mechanism underlying the Per2 upregulation by lithium treatment. By taking approaches of both comparative transcriptome analysis and comparative qPCR analysis between human and murine cells, Lumicycle assay, luciferase assay and RT-qPCR assay showed that lithium could significantly upregulate the expression of Per2 in both mouse and human cells, and significantly inhibit the expression of E4bp4, which encodes a transcriptional inhibitor of Per2. After knocking out the cis-element upstream on the Per2 promoter that responds to E4BP4, the upregulation effect on Per2 by lithium disappeared. When E4bp4 gene was knocked out, the upregulation effect on Per2 by lithium salt disappeared. This study has found that lithium upregulates Per2 expression by reducing the expression of transcription factor E4BP4, but the mechanism of lithium salt downregulation of E4BP4 remains to be further studied. Our study provides a new therapeutic target and approaches for treating BPD.

XQ, 0000-0003-2496-2434 Bipolar disorder (BPD) is a psychiatric disorder characterized by alternate episodes of mania and depression. Disruption of normal circadian clock and abnormal sleep cycles are common symptoms of BPD patients. Lithium salt is currently an effective clinical therapeutic drug for BPD. Animal and cellular studies have found that lithium salt can upregulate the expression of the clock gene Per2, but the mechanism is unknown. We aim to understand the mechanism underlying the Per2 upregulation by lithium treatment. By taking approaches of both comparative transcriptome analysis and comparative qPCR analysis between human and murine cells, Lumicycle assay, luciferase assay and RT-qPCR assay showed that lithium could significantly upregulate the expression of Per2 in both mouse and human cells, and significantly inhibit the expression of E4bp4, which encodes a transcriptional inhibitor of Per2. After knocking out the cis-element upstream on the Per2 promoter that responds to E4BP4, the upregulation effect on Per2 by lithium disappeared. When E4bp4 gene was knocked out, the upregulation effect on Per2 by lithium salt disappeared. This study has found that lithium upregulates Per2 expression by reducing the expression of transcription factor E4BP4, but the mechanism of lithium salt downregulation of E4BP4 remains to be further studied. Our study provides a new therapeutic target and approaches for treating BPD. complexes (B/C complexes in the following text) and then activate transcription of Per and Cry genes. After transcription, translation and post-translational modifications, PERs and CRYs enter the nuclei to repress the transcriptional activity of B/C complexes [6,7]. As PERs and CRYs degrade, the repression function fades away and B/C complexes activate transcription again to start a new circadian cycle. The cycle is approximately 24 h. Accessorial feedback loops include Reverb [8] and Ror [9] and their coding products. REV-ERB and ROR regulate the transcription of the Bmal1 gene that encodes the BMAL1 protein by associating with RORE elements in opposite directions. E4BP4 and DBP bind to D-box elements to regulate the transcriptional activity of relative genes [5,10].
Human genetic mutations or polymorphisms in these clock components have been suggested to link to BPD, including BMAL1, CLOCK, PER2, PER3, CRY2, RORα and RORβ genes [11][12][13][14][15][16][17]. Mice that carry a mutation on the Clock gene exhibit human mania-like symptoms which can be attenuated by chronic lithium administration [18]. As the most applied medicine to treat BPD, lithium was reported to affect the phosphorylation of GSK-3β, which in turn could reduce its kinase activity, resulting in a longer circadian period [19,21]. Yin et al. showed that the inhibition of GSK-3β by lithium promotes the degradation of the nuclear receptor Rev-erbα, leading to increased Bmal1 expression [22]. However, both GSK-3β specific inhibitors and siRNA knockdown of GSK-3β caused a short circadian period, in contrast to the effect of lithium [23]. The accessorial clock gene Dbp was revealed to be downregulated with chronic lithium treatment of lymphoblastoid cell lines [24]. Egr1 has been reported to be a molecular target of lithium to enhance Per2 expression [25]. Iitaka and his colleagues originally reported that the peak level of mPer2 mRNA increased in the presence of lithium in a dose-dependent manner [19]. A microarray study has shown that Per2 expression was upregulated in mouse brain with the treatment of lithium carbonate [26]. Recent studies have further demonstrated that lithium induced the expression of Per2 gene [27,28].
Lithium has been widely prescribed for treatments of BPD patients in current guidelines [29], while most mechanistic studies on lithium related to circadian rhythms were carried out in rodent models [19,25,27]. The mechanisms behind Per2 upregulation in the presence of lithium remain unknown. In our study, we found that the presence of cycloheximide (CHX), a translation inhibitor, abolished the upregulation of Per2 by lithium treatment. Therefore, in order to understand the direct target of lithium, we carried out a comparative transcriptome analysis using lithium-treated human cell lines in the absence or presence of CHX. To our knowledge, no comparative studies between human and rodents have been performed to investigate the common circadian gene target of lithium. We hypothesized that lithium may have common targets in cells derived from both human and mouse sources. In this report, we took advantage of comparative approaches, circadian luminescence analysis, luciferase assays and qRT-PCR assays to investigate common targets of lithium in both human and mouse cell lines. Furthermore, we employed CRISPR genomic editing tools to knock-out the E4bp4 gene, which was coordinately regulated together with Per2 in our comparative analysis. Collectively, we conclude that lithium upregulates Per2 expression by reducing the expression of the repressive transcription factor E4BP4.

Sequence alignment
Promoter sequences of the human PER2 and mouse Per2 genes were searched in the Eukaryotic Promoter Database (https://epd.epfl.ch//index.php). In order to find the homology of Per2 promoter in human and mouse, DNAman software (Lynnon, Quebec, Canada) was used for the sequence alignment.

NIH3T3/mPer2::luc cell line
A hygromycin resistance gene cassette (hygroR) was inserted into pGL3-PmPer2-Luc, resulting in the reporter construct pGL3-Hygro-PmPer2-luc. Cultured NIH3T3 cells were transfected with the construct pGL3-Hygro-PmPer2-luc, followed by hygromycin screening at 300 µg ml −1 . After a population of hygromycin-resistant cells was successfully obtained, monoclonal cell lines were screened by the limiting dilution method. Luminescence recording assays were performed with monoclonal cell lines to confirm that mPer2 reporter was stably transfected into NIH3T3 cells. The 2F5 monoclonal cell line was used in this study.
royalsocietypublishing.org/journal/rsob Open Biol. 11: 210140 2.4. Circadian luminescence recording assay 1.5 × 10 5 U2OS PER2-Luc (U2OS-D15) or U2OS BMAL1-Luc (U2OS-C26) cells were inoculated into 35 mm culture plates (Costar, #9102) in triplicate. On the next day, cells in each well were synchronized with 2 h treatment of 100 nM dexamethasone and recorded in a Lumicycle as described previously [31]. Different concentrations of lithium chloride (LiCl) were added at peaks of the luminescence oscillation, using sodium chloride (NaCl) as the vehicle control. Bioluminescence data were analysed with the Lumicycle analysis program (Actimetrics, USA) to obtain circadian parameters such as period and amplitude.
2.5. Dual luciferase assay to assess the activity of lithium chloride NIH3T3 cells were seeded into six-well plates, and 1 µg promoter reporter constructs of pGL3-PhPER2-Luc and pGL3-PmPer2-Luc were separately transfected into the cells using Lipo6000 (Beyotime, Beijing, China), using a Rluc reporter as reference; 24 h post-transfection, 20 mM LiCl or NaCl was added into the media for 8 h, then the levels of firefly luciferase activity were measured sequentially using Dual Luciferase Reporter Assay Kit (Beyotime, China).

Cycloheximide treatments to inhibit protein translation
U2OS-D15 cells were plated in the 60 mm dishes, and then 10 µM CHX (solved in DMSO) was added into each dish at the peaks of the luminescence oscillation, using NaCl as the vehicle control. One hour later, 20 mM LiCl was added in the CHX-treated and DMSO-treated groups, each group using the 20 mM LiCl as the control.

RNA-seq analysis
U2OS-D15 cells were simultaneously inoculated into 60 mm and 35 mm Petri dishes for culture. When the number of cells reached 60% confluency, the cells were synchronized with dexamethasone for 2 h. The cells in the 60 mm dish continued the culture, while the cells in the 35 mm dish were recorded by the Lumicycle in the presence of fluorescein. When the Per2-Luc oscillation in the cells was about to reach its peak, CHX was added to the 60 mm Petri dish for 1 h, then 20 mM LiCl (NaCl as the control) was added for 4 h, and the cells were collected into Trizol reagent. The collected cells were sent to Personal Biotechnology in Shanghai for RNA-seq. Three biological replicates were carried out for this experiment. The cDNA library of each sample was prepared and sequenced by Personal Biotechnology. Using the second-generation sequencing technology, based on the Illumina sequencing platform, double-terminal sequencing of these libraries was carried out. FastQC software is used to analyse the quality control of the preprocessed data. Based on the human reference, genome sequence downloaded from the Ensembl database (Homo_sapiens. GRCh38). The filtered Reads were compared to the reference genome using HISAT2 software. We used HTSeq to compare the Readcount value of each gene, and FPKM was used to standardize the expression. We used DESeq to analyse the difference of gene expression. Those genes with a false discovery rate < 0.1 were considered as differentially expressed between control and exposed groups, and genes with |log 2 (Fold_change)|> ± 0.5 were considered with a relevant biological signal. We use topGO for GO enrichment analysis.
In the analysis, we used the differential genes annotated by GO term to calculate the gene list and the number of genes in each term and find out the GO term, in which the differential genes are significantly enriched compared with the whole genome background, so as to determine the main biological functions of the differential genes.

Site-directed mutagenesis
The B-site and E2-site were, respectively, mutated by amplifying the whole plasmid of the pGL3-PhPER2-Luc using a highfidelity Taq polymerase. A pair of complementary primers 5 0 -GGCCCCGCTGCGGCCAGATCTCCCGCCGCCGGC-3 0 and 5 0 -GCCGGCGGCGGGAGATCTGGCCGCAGCGGGGCC-3 0 were used to mutate the B-site 5 0 -CTTACGTAACC-3 0 . Primers 5 0 -CGGTCACGTTTTCCACGCTAGGAACAGCGGCGA-3 0 and 5 0 -TCGCCGCTGTTCCTAGCGTGGAAAACGTGACCG-3 0 were used to mutate the E2-site 5 0 -TCCACTATGTGAA-CAG-3 0 . After the amplification, we used DpnI (New England Biolabs) to digest the template plasmids. Then, the product was transformed into the DH5α competent cells. To obtain the site-directed mutagenesis plasmids, the plasmids from the transformed E. coli colonies were sent to sequencing for identification.
2.9. Quantitative real-time PCR to assess gene expression U2OS-D15 cells and NIH3T3-Per2 cells in 60 mm dishes were synchronized for 2 h, and then the cells were treated with 20 mM LiCl (NaCl as the control) for 4 h since the peak of circadian oscillation. Total RNA was extracted from cells using the RNeasy kit (Qiagen, #74104), according to the manufacturer's instructions. One microgram of total RNA was reverse-transcribed to cDNA using HiScript III RT SuperMix for qPCR (+gDNA wiper) (Vazyme, #R323-01, Nanjing, China) according to the manufacturer's instructions. SYBR green fluorescence-based qRT-PCR was performed for RNA quantification (#Q511-02, Vazyme, China). The Delta-Delta-Ct method was used for the normalization of qPCR results. β-actin and rplo were used as the internal control to normalize the gene expression level of NIH3T3 and U2OS cells, respectively. The primer sets used in this study to quantify both human and mouse clock gene expressions were listed in the electronic supplementary material, tables S1 and S2.

Immunofluorescence microscopy
NIH3T3 cells were grown as described in above section. A sterile cover glass was put in each well and 3 × 10 5 cells were seeded per well 1 day before transfection. When cells grew up to 70-80% confluency, they were transfected with pcDNA-flag-mE4BP4 using Lipo6000 (Beyotime, China). Twenty-four hour after transfection, cells were fixed in 4% paraformaldehyde. Then, cells were permeabilized and blocked with 0.3% TritonX-100, 2% BSA, 0.05% Tween in PBS for 1 h at room temperature, followed by sequential royalsocietypublishing.org/journal/rsob Open Biol. 11: 210140 washing with PBS and were subjected for incubation using FLAG antibody. The second antibody was a Cy3-conjugated rabbit anti-mouse IgG (#D110172, BBI, Shanghai, China). Cells were embedded in Mounting Medium, Antifading (S2110, Solarbio). Immuno-labelled cells were analysed using immunofluorescence microscopy (TCS SP8, Leica).

Immuno-blotting assay
Cells were harvested in lysis buffer and mixed with reduced loading sample buffer, followed by separation by 10% SDS-PAGE (60 V for 30 min, 120 V for 60-90 min). The gel was then transferred onto nitrocellulose filter (NC) membranes and blocked in 10% skim milk in PBS-Tween (PBST) for 1 h at room temperature. Antibodies for Per2 (PER21-A; Alpha Diagnostic International) and tubulin (KM9007; Tianjin Sungene Biotech) were used. Cy3-conjugated rabbit anti-mouse antibodies (#D110172, BBI) were used as the secondary antibodies. Immunoreactive bands were detected by Typhoon laser scanners (FLA-9500, GE, USA).

CRISPR/Cas9 technology to knock-out mE4bp4 gene in NIH3T3 cells
Oligonucleotides specific for the second exon of mE4bp4 genomic locus was designed using the Optimized CRISPR Design tool (http://crispr.mit.edu/) [32]. The synthesized oligonucleotides (detailed sequence present in the text) were annealed, dephosphorylated and then ligated into the expression vector pX459 (pSpCas9(BB)-2A-Puro was a gift from Feng Zhang, Addgene plasmid # 48139) using BbsI restriction site. NIH3T3 cells were seeded into six-well plates 1 day before transfection at a density of 2 × 10 5 cells per well. Cells were transfected with the construct targeting E4bp4 gene locus using Lipo6000 reagents (#C0529, Beyotime, China) according to the manufacturer's instruction. Twenty-four hours after transfection, cells were screened with 1 µg ml −1 puromycin. Monoclonal cell lines were screened by the limiting dilution method. The genomic information of the monoclonal cell lines were further confirmed by DNA sequencing.

Statistical analysis
The data are presented as the mean ± standard deviation (s.d.). All experiments were repeated at least three times. The Statistical Package for the Social Sciences (IBM Corporation, Armonk, NY, USA) was used to assess significant differences. Data were analysed using a two-tailed Student's t-test. p < 0.05 was considered statistically significant.

Lithium treatment upregulates the luminescence that is under the control of PER2 promoter
Human osteosarcoma U2OS cells were engineered to express firefly luciferase reporters under control of BMAL1 promoter (U2OS-C26) and PER2 promoter (U2OS-D15), respectively, as described above. In order to exhibit the effect of lithium on the expression of luminescent reporters, LiCl was added during the second or third cycle instead of at the beginning of the experiments, using NaCl as the control. Furthermore, since the rising phase of the luminescence may interfere with the upregulation effect, lithium was treated at the peak of luminescence, as determined by monitoring the luminescent reporters (figure 1a,b). As presented in figure 1, lithium treatments acutely enhanced the luminescence in the D15 cells, but not in the C26 cells. Like control traces, adding lithium did not affect the trend of luminescence in C26 cells (figure 1a). However, for D15 cells, adding lithium suddenly raised the luminescence trace while the control trace was still in its falling trend (figure 1b). Different concentrations of lithium (10, 20 and 40 mM) were used for the experiments along with corresponding sodium controls (figure 1c,d). The effect of lithium was measured by relative amplitudes that were defined as comparing the peak levels (labelled with black triangles) of luminescence after and before the treatments (figure 1a,b). We observed that the relative amplitude of the luminescence under the control of Per2 promoter was upregulated in a dose-dependent manner (figure 1f ), but not for these C26 cells (figure 1e). Also, the overall baseline of the luminescence was high under the treatments of lithium in the D15 cells (figure 1b). We further tested the consequence of treating D15 cells with lithium at the trough of luminescence, and we observed higher-luminescence peaks after the treatments, which is also dependent on the lithium concentration (electronic supplementary material, figure S1). These data indicate that lithium specifically enhances the activity of Per2 promoter, but not that of Bmal1 promoter. The basal level of luminescence under the control of Per2 promoter is not affected when the D15 cells were either treated with NaCl or PBS (electronic supplementary material, figure S2). Thus, we chose NaCl as a control in the following experiments.

Inhibition of protein translation blocked the upregulation of PER2 by lithium
Multiple signalling pathways are capable of activating circadian clock gene expressions both in vitro and in vivo [33,34]. We wished to determine whether signalling pathways are involved in activating PER2 expression under lithium treatments. If any signalling pathway that can directly target a specific response element on the promoter of PER2 was activated by lithium, inhibition of translation would not prevent the upregulation of PER2 expression. DMSO or 10 µM protein synthesis inhibitor CHX was added to the human U2OS-D15 cells 1 h before the lithium treatment. Then, 20 mM NaCl or 20 mM LiCl was employed. In spite of being treated with sodium or lithium, PER2 expressions were accumulating over time in the presence of CHX, more than the DMSO controls ( figure 2a). PER2 transcripts showed a significant difference within 1 h when protein synthesis was blocked by CHX, leading to accumulated RNA. Some other clock or clock-regulating genes were also quantified under this condition, such as DEC1, DEC2 and EGR1 genes. Transcripts of these genes accumulated as well in the presence of CHX (electronic supplementary material, figure S3). When cells were incubated with DMSO, 20 mM LiCl apparently upregulated PER2 expression at time points 2 h and 4 h (figure 2b). In this experiment, time 0 indicated the peak of the PER2 expression, since the luminescence was monitored under the lumicycle detector. When the translation inhibitor CHX was applied, transcripts of PER2 were quickly royalsocietypublishing.org/journal/rsob Open Biol. 11: 210140 accumulating for both sodium and lithium groups, and there was no difference of PER2 level between two groups at any time point (figure 2c). Thus, lithium treatment cannot directly activate signalling pathways that could target cis-regulating elements on the PER2 promoter region to upregulate its expression. Transcription factors have to be involved in the phenomenon of PER2 upregulation by lithium. Our observations indicated that there is a factor which was directly affected by lithium transcriptionally and its translation product can regulate PER2 expression. Therefore, we conducted comparative transcriptome analysis of the U2OS-D15 cells in the absence or presence of CHX under the treatment of LiCl.

Comparative transcriptome analysis indicates that
E4BP4 may be the direct target of lithium U2OS-D15 cells were treated with CHX for 1 h before adding LiCl or NaCl, which was described in the Material and methods section. Four hours after LiCl (NaCl as the control) treatment, cells were collected and the transcriptomes under both treatments were analysed by RNA-seq. Based on the Illumina sequencing platform, the transcriptome sequencing of the samples was carried out. The average raw data of each sample was 6G, and the range of original reading times was 6356378700-7367250600. Sequencing readings that contain low quality, joint contamination and high content of unknown base (N) readings were removed. The gene differentiation analysis was carried out by DESeq, and the screening condition was that the gene expression was upregulated by 150% and downregulated by 75%, and the significant p-value was less than 0.1. Based on the analysis of differential expression of genes in the non-CHX group (LiCl versus NaCl) and the CHX group (LiCl versus NaCl), 2298 differentially expressed genes (753 upregulated and 1545 downregulated) and 292 differentially expressed genes (138 upregulated and 154 downregulated) were detected, respectively (figure 3a). GO enrichment analysis of differentially expressed genes was carried out by topGO, and the top 30 GO terms with the most significant enrichment were selected for functional analysis (figure 3b). We found that the differentially expressed genes are mainly related to metabolic regulation, transcriptional regulation and biosynthesis. In addition, we analysed the top 30 genes of upregulated and downregulated genes in the non-CHX and CHX groups of samples (figure 3c). In the non-CHX group, PER2 was upregulated about 2.5 change folds after LiCl treatment, while in the CHX group, the upregulation of PER2 disappeared after LiCl treatment. Through the analysis of the common changed 141 genes (41 upregulated, 100 downregulated) between the non-CHX and CHX groups (figure 3a), in the common downregulation set, we found that the E4BP4 was downregulated about 0.65 change folds in both groups. We demonstrated the heatmap of expression values for the 14 most important biological clock genes (figure 3d). Among these genes, PER2 was upregulated only in the non-CHX group, while E4BP4 was downregulated in both groups. These results indicate that LiCl can upregulate the expression of PER2 in cells, but when CHX is used to block the translation pathway, the upregulation of PER2 expression disappears. However, the presence of CHX did not affect the downregulation of E4BP4 by lithium. Most mammalian circadian clock genes were discovered during the past few decades, and the clock runs rhythmically under the control of transcriptional and translational feedback loops. PER genes, including three homologous members, stand in the main path of this system [5]. Positive transcriptional factors BMAL1 and CLOCK form heterodimers and then activate the gene transcription of PER genes through associating with the E-box that locates upstream to their promoter regions, including the PER2 gene. After transcription and translation, PER proteins form complexes with key clock components CRY1/CRY2, therefore inhibiting the activation of BMAL1/CLOCK complexes. Other regulatory clock components were discovered, including several pairs that play opposite roles during transcription, such as REV-ERBs and RORs, DBP and E4BP4, and the DEC family.
Since lithium can enhance the luminescence expression that is under the control of an exogenous PER2 promoter in U2OS cells (figure 1b), we aimed to verify that the endogenous PER2 was indeed upregulated by the treatment of lithium besides using the RNA-seq technique. Furthermore, these main clock genes and the clock-related genes that play regulatory functions in transcription were examined under the lithium treatment situation in U2OS cells.
U2OS-D15 cells were treated with 20 mM LiCl at the luminescence expression peak time, which can be determined by monitoring the luminescence. We measured expression levels of BMAL1, CLOCK, PER1/2/3, CRY1/2, REV-ERBα/β, RORα, DEC1/2, DBP and E4BP4 genes at 0 and 4 h after the lithium treatment. Compared to the sodium controls, expression levels of PER1, PER2 and CRY1 were upregulated, while those of BMAL1, REV-ERBα, DEC2, DBP and E4BP4 were downregulated in U2OS cells (figure 4). For PER2 gene, the expression level increased to 1.5-fold compare to its sodium control (figure 4d). The remaining tested genes, CLOCK, PER3, CRY2, REV-ERBβ, RORα and DEC1, were not affected. The upregulation of endogenous PER2 confirmed that lithium can activate its expression through regulating its promoter region.
Then, we treated NIH3T3-2F5 cells with 20 mM LiCl, using NaCl as controls. The genes that are homologous to those being tested with U2OS cells were measured at the same time points within NIH3T3-2F5 cells. In the case of using mouse-derived NIH3T3 cells lines, different gene expression patterns were observed. Compared to the sodium controls, expression levels of Clock and Per2 were upregulated, while only E4bp4 was downregulated in NIH3T3 cells ( figure 5). For the genes    royalsocietypublishing.org/journal/rsob Open Biol. 11: 210140 (figures 4 and 5). Meanwhile, we tested whether lithium can upregulate Per2 and downregulate E4bp4 in a neuronal cell line, U373 cells. Consistently, we found that Per2 was upregulated with lithium treatments (electronic supplementary material, figure S5a), while E4bp4 was downregulated with lithium treatments in U373 cells (electronic supplementary material, figure S5b). Next, we tested E4BP4 expression under the treatments of LiCl or NaCl in the presence of CHX in U2OS cells. A significant decrease of E4BP4 expression was observed at time point 2 h (electronic supplementary material, figure S6b). When the translation inhibitor CHX was applied, E4bp4 was also accumulating in the first 2 h in the presence of CHX (electronic supplementary material, figure S6a), but its level was significantly lower at 4 h under the lithium treatment (electronic supplementary material, figure S6c). In our observation, even without de novo synthesis of new transcription factors, E4bp4 was downregulated to a similar level as in the absence of CHX (figure 4; electronic supplementary material, figure S6). For other genes being tested, Dec1, Dec2 and Egr1 also decreased to some extent when treated with lithium in the presence of CHX (electronic supplementary material, figure S7).
As a transcription factor, E4BP4 is known to repress Per2 expression when it binds to separate DNA elements upstream the Per2 promoter region. Two E4BP4-binding sites were reported locating on the Per2 promoter region, B-site and E2-site, respectively [36]. According to the promoter sequences in Eukaryotic Promoter Database (https://epd. epfl.ch//index.php), we constructed a firefly luciferase reporter under the control of mPer2 promoter region from −798 to +331 sites relative to the transcription start site, and another reporter under the control of hPer2 promoter region from −500 to +196 sites. These cis-elements on both the mouse Per2 and the human PER2 promoter regions were identified and labelled, as presented in figure 6a. The DNA sequences from both promoter regions are highly conserved, especially in these cis-element regions (figure 6b). We transiently transfected NIH3T3 cells with both reporter constructs, and then 20 mM LiCl or NaCl was added 24 h after transfection. Both firefly luminescence reporters driven by Per2 promoters were strongly induced when the cells were treated with 20 mM lithium (figure 6c). The induction levels of both reporters are equivalent to each other, less than twofold (figure 6c). This effect by lithium on transiently expressed reporters is also consistent with the effects observed both from stably expressed reporter cell lines and from endogenous Per2 expression experiments.

The effect of lithium on the activation of Per2
expression was mediated by E2-site which is an E4BP4-binding cis-element The amplitude of Per2 expression can be suppressed by E4BP4 [36,37]. We hypothesized that the upregulation effect of lithium is mediated by downregulated E4BP4 based on our comparative analysis between human and mouse cells. N-terminal FLAG-tagged E4BP4 was expressed in NIH3T3 cells and immunofluorescence staining experiments confirmed that E4BP4 localizes to the nuclei (figure 7a), where a transcription factor can function. Then we over-expressed E4BP4 together with transiently expressed mouse Per2 luciferase reporter in NIH3T3 cells. As the amount of E4BP4 plasmid constructs increased, the luciferase activity was repressed in a dose-dependent manner (figure 7b). Among these cis-elements on the Per2 promoter region, B-site and E2-site are where E4BP4 functions as a transcriptional repressor. We mutated these two sites separately on the mouse Per2 luciferase reporter construct into the sequences as shown in figure 7c. From the results of transient transfection experiments, the mutation of E2-site greatly enhanced luciferase activity up to eightfold higher, which is most likely to be due to a higher luciferase expression level (figure 7d). In comparison, the mutation of B-site slightly enhanced the royalsocietypublishing.org/journal/rsob Open Biol. 11: 210140 luciferase activity by 1.9-fold (figure 7d ). We further examined the lithium effect on these two mutant reporters. Like our WT control that lithium can enhance the luciferase activity, lithium-treated cells showed 1.8-fold higher luciferase activity with the B-site mutant reporter; by contrast, the upregulation effect by lithium was withdrawn when the E2site mutant reporter was used in the assay (figure 7e). These results supported our hypothesis that E4BP4 mediates lithium upregulation of Per2 expression.

Lithium cannot upregulate
Per2 expression in a cell line in which E4bp4 gene was knocked out using CRISPR/Cas9 genomic editing tools Mouse E4bp4 gene has two exons on the chromosome and a 20 nt guide sequence targeting exon 2 was designed according to the CRISPR Design Tool [32], as presented in figure 8. The Cas9 nuclease normally produces a double-strand break at the −3 position relative to the 5 0 TGG PAM site. Mouse NIH3T3 cells were transfected with the CRISPR construct that targets exon 2 on the E4bp4 locus. Stably transfected cells were screened afterwards and one single-colony cell was named E4bp4-B12 cell strain. Using genomic DNA extracted from E4bp4-B12, flanking regions around the target PAM site were PCR amplified and cloned into bacterial vectors and followed with DNA sequencing. Sequencing results from over 15 clones gave us only one result that −4 position to the PAM site was changed from CG to T (figure 8b,c). The consistent results indicate that both endogenous alleles were simultaneously mutated into the same sequence. This mutation caused codon shift in the open reading frame of E4BP4, resulting in a STOP codon shortly after the modification site (figure 8d). Since the target site is at one-third of the coding sequence, we concluded that even a truncated protein was made, it would not function normally. Next, we tested the effect of lithium on this single-colony cell E4bp4-B12. Twenty millimolar LiCl was treated to cultured cells and the expression levels of Per2 were quantified by qRT-PCR. The results showed that Per2 was not induced in this background (figure 8e), while Per2 was greatly upregulated in the wild-type cells.
We further examined the Per2 protein level within this cell strain when the cells were treated with lithium, compared to its sodium control. No significant difference could be detected by the western blotting results (figure 8f ). royalsocietypublishing.org/journal/rsob Open Biol. 11: 210140

Discussion
By taking comparative transcriptome analysis between CHX and non-CHX conditions, we found that among the 141 common direct regulated genes under the treatment of LiCl, E4bp4 might be the responding transcription factor that was involved in the Per2 upregulation by lithium. Next, we show from a comparative study that lithium treatment leads to induced expression of Per2 in both human and mouse cells. Fifteen circadian clock and clock-related genes were examined in both cell lines and E4bp4 gene expression was also influenced by lithium treatment. E4BP4 functions as a transcription factor that represses many gene expressions, including the Per2 gene [36,37]. Conserved cisregulatory elements were identified by aligning the promoter region of both human and mouse Per2 locus. We further employed luciferase reporter systems and CRISPR/Cas9 genomic editing technology to demonstrate that the Per2 upregulation by lithium is mediated through the E4BP4 transcription factor.
Previous studies have reported that clock component Dbp is sensitive to lithium treatments, resulting in a decreased expression [24]. Another study indicates that both Per2 and Egr1 (early growth response 1) are targets of lithium, while Per2 lies downstream in the ERK/Egr1 pathway [25]. In our study, we observed downregulation of DBP (figure 4m) and upregulation of EGR1 (electronic supplementary material, figure S8) in U2OS cells, which is consistent with previous observations that were carried out using patient-derived lymphoblastoid cells [24] and human neuroblastoma cells [25]. However, their expression in NIH3T3 cells (derived from mouse) was not affected in this study (figure 5m; electronic supplementary material, S8). These data indicate there may be multiple targets and multiple pathways involved in the lithium-induced effect.
Since various pathways may activate Per2 expression directly or indirectly under the lithium treatment situation, we applied protein synthesis inhibitor CHX to U2OS cells before adding lithium. Inhibition of translation prevented the upregulation of Per2 by lithium (figure 2c) and did not prevent downregulation of E4bp4 (electronic supplementary material, figure S6). These results were consistent with our transcriptome RNA-seq results. Interestingly, three other genes being tested, Egr1, Dec1 and Dec2, were also downregulated in the presence of CHX (electronic supplementary material, figure S7). Thus, lithium does not directly activate Per2 promoter. Therefore, multiple cis-regulating elements that respond to different transcription factors may be downstream to the lithium treatment. Since most of the clock genes have E-box in their promoter regions and lithium was unable to induce these circadian genes' expression in the current comparative study, the E-box element was excluded from the factors that could mediate the lithium effect. Egr1 expression was not affected in mouse NIH3T3 cells and was downregulated in human U2OS cells in the presence of CHX, indicating different pathways that lead to downstream Per2 upregulation in human and mouse cells.
Lithium not only activated Per2 expression transiently and rapidly (figure 1b), but also maintained a high expression of Per2 for relatively long time, which can be observed from the overall elevated baseline of Per2 promoter-driven luciferase activity (figure 1b). In this circumstance, the period lengthening effect by lithium which was described and discussed previously [21,28] may be caused by both inhibiting GSK-3β and continuously activating Per2 transcription. As a mood-stabilizing medicine, lithium was considered to treat BPD patients partially through its effect on the circadian system [38]. It is therefore interesting that we identified both Per2 and E4bp4 conservatively being influenced by lithium (figures 4 and 5; electronic supplementary material, S5). As a central component in the circadian system, perturbed Per2 expression level would affect the whole  Figure 8. E4bp4 mediates the effect of lithium that upregulates Per2 expression. (a) Schematic of the mouse E4bp4 locus that is composed of two exons. The gRNA target region was designed to a TGG site in exon 2, whose detailed sequences were presented. (b) Sanger sequencing data of the wild-type sequence of the target site. The TGG site is highlighted in red. (c) The gRNA target region mutated from CG to T at the -4 position to the TGG site, which caused a frame shift mutation. (d ) The mutation caused a codon shift in the E4bp4 locus, resulting in an early stop codon in the translating product. (e) LiCl did not induce Per2 expression in E4bp4 knock-out cell strain. Experiments were carried out with triplicates. ( f ) Per2 protein level was unaffected in the E4bp4 KO cells when treated with lithium, compared to the sodium control. Tubulin was used as a loading control for the western blotting data.
circadian clock in various aspects, including period change, amplitude and phase shift. E4bp4 is widely expressed in peripheral tissues and regulates downstream output genes through the D-box element [10,39]. Perturbed E4bp4 expression may further affect the amplitude of the circadian rhythm, since E4bp4 functions as a repressive transcription factor. Recently, E4bp4 and Dec2 were reported to form a heterodimer to repress Per2 expression through binding to the E2 element [37]. In this study, both E4bp4 and Dec2 were downregulated in human cells, but only E4bp4 was downregulated in mouse cells in the presence of lithium. What is more, we found that Dec2 was repressed both in the presence and the absence of CHX when human U2OS cells were used. We also find the E2 site rather than the B site is the preferred response element for E4bp4. These results support our findings that E4bp4 mediates the upregulation of Per2 expression by lithium, possibly with partial help from Dec2. Thus, our findings suggest lithium can affect both the amplitude and the phase of circadian rhythms, which are key circadian clock symptoms being disrupted in the BPDs. Both SCN and peripheral tissues show robust circadian oscillations, which involve the functions of Per2 and E4bp4 genes. The present study demonstrated that lithium upregulates Per2 expression in human and murine fibroblast cells and neuronal U373 cells (figures 4 and 5; electronic supplementary material, S5). SCN belongs to the central neuron system. Therefore, lithium may affect the expressions of Per2 in SCN. However, a solid conclusion requires further investigation. Revealing the targets of lithium will be an important step towards understanding the relationship between the circadian system and the BPD. Our translation inhibition assay with CHX indicates that E4bp4 might be directly regulated by lithium treatments. Yet how E4bp4 gene was downregulated in the lithium treatments requires further study. As the smallest ion, lithium may compete with other metal ions to inhibit some key enzymes in signalling pathways or key enzymes in transcription processes, such as RNA polymerase II. Furthermore, Per2 upregulation by lithium might also be due to a smaller degradation rate of Per2 transcript. In our future plan, we will test whether the treatment of lithium would affect the degradation rate of Per2 with the application of a transcription inhibitor. Another important task will be to evaluate this lithium-E4bp4-Per2 axis in intact animals or in cells from BPD patients.
Data accessibility. The data are provided in the electronic supplementary material [40].