The molecular clock protein Bmal1 regulates cell differentiation in mouse embryonic stem cells

mESCs do not show circadian oscillation of clock genes, but these can be induced by in vitro differentiation towards mouse neural stem cells (mNSCs) (Yagita et al, 2010). Reciprocally, circadian oscillations in mNSCs are extinguished when they are artificially reprogrammed to pluripotency (Yagita et al, 2010). We asked whether the onset of circadian oscillations during mESCs differentiation was associated with differences in Bmal1 expression levels. Colonies of mESCs growing in FBS and leukemia inhibitory factor (LIF) (mESCs FBS + LIF or primed serum mESCs) were differentiated into mNSCs that showed the typical fusiform shape morphology and grew as a reticular network (Fig 1A). mNSC cultures showed expected down-regulation of pluripotency genes Nanog, Oct4, and Rex1 coupled to up-regulation of mNSC genes Fabp7, Slc1a3, and Nestin (Fig 1B). Immunofluorescence analysis confirmed that NSC cultures do not express the nuclear pluripotency-associated transcription factor Oct4 and showed homogeneous staining of the NSC protein marker Nestin in their cytoplasms (Fig 1C). Comparison of mRNA expression level in primed serum mESCs and NSCs showed that Bmal1 is expressed at a similar level in both cell types (Fig 1D). Expression of Bmal1 was similar to the transcriptionally active housekeeping gene Hmbs, indicating that Bmal1 is expressed at functional levels in mESCs (Fig 1D). Western blot analysis using anti-Bmal1 antibodies revealed two bands at 71 kD that correspond to previously described phosphorylated and unphosphorylated isoforms of Bmal1 (Yoshitane et al, 2009; Sahar et al, 2010) in mNSCs (Fig 1E). Importantly, primed serum mESCs and mNSCs expressed similar levels of Bmal1 protein (Figs 1E and S1A), suggesting that Bmal1 might be functional in pluripotent cells.

To discard that expression of Bmal1 protein was associated to marginal spontaneous cell differentiation and cell heterogeneity typically found in primed serum mESC cultures (Li & Izpisua Belmonte, 2018), we analysed expression of Bmal1 in naïve pluripotent cells. Primed serum mESCs were transferred to defined media with LIF and inhibitors of MAPK and glycogen synthase kinase 3 (2i + LIF media), to establish cultures of naïve 2i pluripotent cells (Ying et al, 2008). Expression of Bmal1 was similar and slightly increased in naïve 2i mESCs as compared with primed serum mESCs, at the level of both mRNA and protein (Figs 1F and G and S1B), further indicating that Bmal1 is expressed at a functional level in pluripotent cells. To confirm that expression of Bmal1 in pluripotent mESCs is not an artefact of in vitro cell culture conditions, we analysed whether Bmal1 is also transcribed in pluripotent cells in vivo. We interrogated published single-cell messenger RNA high-throughput sequencing (mRNA-seq) datasets of developing preimplantation embryos (Deng et al, 2014) and found a heterogeneous pattern of Bmal1 expression at all stages from four cells to late blastocyst (Fig 1H). Importantly, cells that express a higher level of the pluripotency gene Oct4 also tend to express higher levels of Bmal1 mRNA at the blastocyst stage (Fig 1I), demonstrating that Bmal1 mRNA is expressed in pluripotent cells in vivo.

Last, cell fractionation followed by Western blot experiments showed that distribution of Bmal1 protein in mESCs is similar to its distribution in NSCs (nucleus and cytoplasm) (Figs 1J and S1C), suggesting that the absence of circadian oscillations in ESCs is not a consequence of the exclusion of Bmal1 from the nucleus of pluripotent cells. Taken together, these results solidly demonstrate that albeit mESCs do now display circadian oscillations (Yagita et al, 2010), the molecular clock protein Bmal1 is present in pluripotent cells.

To investigate the role of Bmal1 protein in pluripotent cells, we derived Bmal1−/− mESCs using CRISPR/Cas9 and grew them in the 2i + LIF medium. The amino terminus of Bmal1 was targeted (exon 5) (Fig 2A), and a panel of genetic clonal mESC lines were screened and genotyped. The frequency of Bmal1 mutants in one or both alleles was high (four heterozygous [Bmal1+/−] and four compound heterozygous [Bmal1−/−] mutants of 12 screened clonal colonies), suggesting that Bmal1 is not essential to establish undifferentiated mESC cultures. Of the four Bmal1−/− clonal cell lines, we focused on one in which both alleles contained DNA indel mutations (−22 and +61 nucleotides, respectively) at expected target sites (Fig 2B) and resulted in frame shifts and premature stop codons at the very beginning of Bmal1 open reading frame, impeding the production of peptides that include any of the annotated functional domains included in Bmal1 protein (BHLH, PAS1, PAS2, and TAD). Bmal1−/− mESCs formed typically tight three-dimensional colonies similar to the ones generated by parental wild-type cells (Fig 2C). As expected, expression of Bmal1 mRNA was not detected in Bmal1−/− mESCs compared with wild-type cells (Fig 2D), and Western blot analysis indicated that Bmal1−/− mESCs do not express detectable levels of Bmal1 protein (Fig 2E). Bmal1−/− mESC cultures remain undifferentiated because they retained expression of pluripotency-associated transcription factors Oct4, Nanog, and Rex1 (Fig 2F). Interestingly, we found that Bmal1−/− mESCs express even higher levels of Nanog mRNA and protein than parental wild-type cells (Figs 2F–I and S1D). Taken together, these results demonstrate that Bmal1 protein is not essential for mESCs self-renewal, and that its depletion results in up-regulation of Nanog expression in naïve 2i mESCs.

To further understand the role of Bmal1 in pluripotent cells, we compared mRNA expression in Bmal1−/− and wild-type naïve 2i mESCs by mRNA-seq. 854 gene transcripts were differentially expressed (up- or down-regulated by more than twofold and adjusted P-value < 0.05) in cells that lack Bmal1 as compared with parental wild-type cells (Fig 3A). Unexpectedly, 689 of these transcripts (80.6%) were up-regulated in Bmal1−/− cells, suggesting that, in contrast to the role of Bmal1 in promoting transcription of target genes in adult tissues (Takahashi, 2017), Bmal1 is not acting as a transcriptional activator in pluripotent mESCs. Most of the up-regulated genes in Bmal1−/− mESCs were already transcriptionally active in wild-type cells (Fig 3B), suggesting that Bmal1 negatively modulates transcription of active genes. Functional categories analysis revealed that Bmal1−/− mESCs display mis-regulation of genes encoding for 180 glycoproteins and 59 developmental proteins, suggesting that Bmal1 is involved in the regulation of cell communication and development (Fig 3C). In fitting, gene ontology analysis showed that depletion of Bmal1 results in deregulation of genes involved in cell differentiation and organism development (Fig 3D), and pathway enrichment analysis exposed that Bmal1−/− mESCs display altered regulation of pluripotency-associated signalling pathways including phosphoinositide 3-kinase (PI3K), MAPK, and Wnt (Fig 3E). In agreement with the general trend towards up-regulation of gene transcripts found in Bmal1−/− mESCs, most differentially expressed genes involved in cell differentiation and embryo development were up-regulated in these cells (i.e., 47 of 58 genes involved in development, Fig S2A). In addition, analysis of components of signalling pathways deregulated in Bmal1−/− mESCs confirmed widespread abnormal expression patterns of mRNA transcripts involved in pluripotency signalling and cell differentiation including the MAPK pathway (Figs 3F and S2B). MAPK pathway, also known as extracellular signal-regulated kinases (Erk1/2) pathway, inhibits Nanog expression in mESCs (Wray et al, 2010), and thus, we wondered whether deregulation of this pathway underlie the up-regulation of Nanog protein found in Bmal1−/− mESCs (Fig 2F–I). However, expression of Erk1/2 protein was unaltered in Bmal1−/− mESCs in both primed serum and naïve 2i conditions (Fig S2C), and mRNA expression of Erk1/2 targets (Hamilton & Brickman, 2014) was not affected by depletion of Bmal1 protein (Fig S2D and E), suggesting that up-regulation of Nanog protein is not a consequence of reduced Erk1/2 signalling in Bmal1−/− mESCs.

We reasoned that up-regulation of Nanog protein and mis-regulation of signalling components might hinder the multi-lineage differentiation capacity of Bmal1−/− mESCs. We injected Bmal1−/− and wild-type mESCs into immunodeficient mice and assayed their ability to form teratomas and differentiate into the three germ layers. As expected, imaging analysis of teratomas using haematoxylin and eosin staining confirmed that wild-type mESCs can differentiate into tissues derived from the three germ layers. For example, we could detect skin (ectoderm), glandular tissue (endoderm), and smooth muscle (mesoderm) (Fig 3G). Bmal1−/− teratomas also contained tissues derived from the three germ layers, suggesting that the lack of Bmal1 does not fully block differentiation towards any the of three germ layers in vivo (Fig 3G). To complement this analysis using a quantitative approach, we performed immunohistochemistry using antibodies against glial fibrillary acidic protein (Gfap), cytokeratin, and smooth muscle actin that are well-established markers of ectoderm, endoderm, and mesoderm tissue respectively. Importantly, teratomas formed by Bmal1−/− cells displayed reduced labelling for Gfap, cytokeratin, and smooth muscle actin markers compared with wild-type cells (Fig 3H and I), indicating that although Bmal1−/− cells can produce tissue derived from the three germ layers, their differentiation potential is altered compared with wild-type cells.

We wondered whether reduced differentiation capability of Bmal1−/− mESCs was a consequence of abnormal expression of lineage specifier genes before teratoma formation. Analysis of mRNA expression of a subset of master regulators of ectoderm (Sox1, Wnt7b, Pcdh17, Gsx1, and Pax6), endoderm (Gata4, Gata6, Foxa2, Cxcr4, and Sox7), and mesoderm (Bry, Bmp2, Hand1, Eomes, Mixl1, and Bmp4) lineages showed that Bmal1−/− mESCs retain repression of lineage specifier genes (Fig S2F and G). This suggests that differentiation defects found in Bmal1−/− teratomas are not a consequence of unscheduled transition of Bmal1−/− mESCs into a differentiated state, but rather mis-regulation of lineage specifier gene activation upon induction of cell differentiation.

We compared the dynamics of lineage transition of Bmal1−/− and wild-type mESCs at early time points during cell differentiation using well-established in vitro systems. We induced cell differentiation by plating naïve 2i mESCs in N2B27 media growing as adherent cultures (Ying et al, 2003) and compared down-regulation of pluripotency genes and activation of differentiation markers by RT-qPCR in Bmal1−/− and parental wild-type cells (Fig 4A). As expected, wild-type cells down-regulated pluripotency genes (Oct4 and Nanog; black bars) and induced expression of ectoderm (Sox1, Wnt7b, Pcdh17, Gsx1, and Pax6; blue bars), endoderm (Gata4, Gata6, Bmp2, Cxcr4, and Sox7; green bars), and mesoderm (Bmp2, Hand1, Msx2, Eomes, Bmp4, and Mixl1; red bars) genes after 4 d of differentiation (Figs 4B and C and S3A). Likewise, Bmal1−/− mESCs also down-regulated pluripotency genes and activated ectoderm genes (Figs 4B and C and S3A). However, activation of endoderm and mesoderm markers was severely hindered in differentiating Bmal1−/− cells (grey bars), as compared with wild-type control cells (Figs 4C and S3A).

To confirm that Bmal1 is required for activation of endoderm and mesoderm genes during pluripotent cell differentiation, we compared early time points upon pluripotency exit in primed serum Bmal1−/− and wild-type mESCs growing as adherent cultures (Fig 4D). Bmal1−/− mESCs down-regulated the pluripotency gene Nanog (Fig 4E) and induced expression of ectoderm markers (Sox1, Gsx1, and Pax6) with similar kinetics to wild-type mESCs (Fig 4F). Importantly, primed serum Bmal1−/− mESCs showed hindered activation of endoderm (Gata4, Gata6, Foxa2, Cxcr4, and Sox7) and mesoderm markers (Bmp2, Hand1, and Msx2) (Figs 4F and S3B). Taken together, these results indicate that Bmal1 is required to modulate multi-lineage differentiation in pluripotent mESCs and indicate that its activity favours endoderm and mesoderm specification.

To further examine the altered differentiation capacity of Bmal1−/− mESCs, we analysed their ability to differentiate and produce gastrula-like organoids (gastruloids) in vitro (Beccari et al, 2018). Bmal1−/− and wild-type mESCs were plated in low-attachment conditions to induce the formation of gastruloids, and their morphology and gene expression profile was compared during 5 d of differentiation. Compared with wild-type mESCs, Bmal1−/− cells formed smaller (average diameter on day 5 in wild-type is 371 ± 40 μm compared with Bmal1−/− 160 ± 16 μm) and less dense-looking three-dimensional structures during the course of the experiment (Fig 5A, white arrows). Analysis of mRNA expression revealed that differentiating Bmal1−/− mESCs down-regulated pluripotency genes (Oct4 and Nanog) and induced the activation of ectoderm genes (Sox1, Gsx1, Wnt7, and Pax6) at a similar or even more accused rate than parental control cells (Figs 5B and C and S3C). In contrast, Bmal1−/− differentiating gastruloids showed minor induction of endoderm (Gata4, Gata6, and Foxa2) and mesoderm (Bmp2, Msx2, and Mixl1) genes compared with parental control cells (Figs 5D and E and S3C). Thus, we concluded that Bmal1−/− mESCs consistently fail to activate endoderm and mesoderm genes upon pluripotency exit in all cell differentiation systems tested.

To confirm that the phenotype of Bmal1−/− mESCs is a consequence of Bmal1 protein depletion and is not linked to a CRISPR/Cas9 off-target effect, we derived an independent clonal population of Bmal1−/− mESCs (clone #G10). Genotyping confirmed that these cells are homozygous mutant for a deletion of four nucleotides in exon 5 in both alleles of the Bmal1 gene that result in a frameshift and premature stop codon at the beginning of the open reading frame (Fig S4A). Bmal1−/− (clone #G10) did not express detectable levels of Bmal1 protein (Fig 5F), and similarly to the previously characterized Bmal1−/− mutant (Fig 2), Bmal1−/− (clone #G10) grown in the 2i + LIF medium formed typical tight three-dimensional colonies (Fig S4B) and expressed normal, or slightly increased, levels of the pluripotency-associated transcription factors Nanog, Oct4, and Rex1 (Fig 5G). Importantly, Bmal1−/− (clone #G10) mESCs also displayed higher levels of Nanog protein than wild-type parental controls (Fig 5H) phenocopying the previous Bmal1−/− cells. In addition, analysis of the differentiation potential of Bmal1−/− (clone #G10) further confirmed that the lack of Bmal1 protein results in the formation of smaller gastruloids (Figs 5I and S4C) that can down-regulate pluripotency markers and up-regulate ectoderm genes but have compromised induction of endoderm and mesoderm markers (Figs 5J and S4D). Taken together, these analyses confirm that Bmal1 regulates Nanog expression and the multi-lineage potential of pluripotent cells and suggest that Bmal1 is required for formation of endoderm and mesoderm tissue during embryo development.

JM8 mESCs derived from the mouse strain C57BL6/N were cultured in primed serum conditions as previously described (Landeira et al, 2015); DMEM KO (Gibco) media supplemented with 10% FCS (Gibco), LIF, penicillin/streptomycin (Gibco), L-glutamine (Gibco), and 2-mercaptoethanol (Gibco). Ground-state 2i mESCs were grown in serum-free 2i media as described previously (Ying et al, 2008): N2B27 supplemented with MEK inhibitor PD0325901 (1 μM) (Millipore) and GSK3 inhibitor CHIR99021 (3 μM) (Millipore) in the presence of LIF. Both serum-grown and 2i-grown mESCs were cultured feeder free on 0.1% gelatin-coated dishes at 37°C and 5% CO₂.

Neural stem cell cultures were derived from primed serum JM8 mESCs as previously described (Conti et al, 2005). Ground-state 2i JM8 and Bmal1−/− cells (1 × 10⁵ cells in six well plates) were differentiated as a monolayer as described previously (Ying et al, 2003). JM8 and Bmal1−/− cells growing in serum (12,500 cells/cm²) were differentiated by LIF removal as a monolayer as previously described (Landeira et al, 2010). Primed serum JM8 and Bmal1−/− cells (1 × 10⁴ cells/ml) were induced to form gastrula-like organoids (gastruloids) as described elsewhere (Beccari et al, 2018).

One million cells of each cell line were resuspended in 250 μl of PBS and injected subcutaneously into the flanks of NOD/SCID IL-2Rγ−/− mice in the animal experimentation unit of the University of Granada. 5 wk later, the mice developed tumours that were removed, rinsed with PBS, fixed, and embedded in paraffin at the Biobank of Andalusia Public Health System. Tissue sections were cut and processed for haematoxylin and eosin staining or for immunohistochemistry staining with Anti-Human Smooth Muscle Actin (IR611; Dako), Ms Anti-Human Cytokeratin Clone AE1/AE3 (IR053; Dako), and Rb Anti-Human Glial Fibrillary Acidic Protein (IR 624; Dako). Species-specific secondary antibodies conjugated with peroxidase were used (EnVision Flex/HPR SM802; Dako). Immunohistochemistry staining images were quantified measuring the area that was positive for the signal of the antibody. For each staining, 10 images were quantified using Image J, averaged, and tested statistically using t test.

Single-guide RNA sequences (Oligo 1: 5′-CACCGAACCGGAGAGTAGGTCGGT-3′, Oligo 2: 5′-AAACACCGACCTACTCTCCGGTTC-3′) were designed using CRISPR design tool and cloned into PX458 plasmid (Ran et al, 2013) upon testing with surveyor mutation detection kit (IDT). JM8 mESCs were transfected using lipofectamine 2000 (Thermo Fisher Scientific). Cells positive for GFP expression were sorted using flow cytometry and plated at serial dilutions to obtain isolated clonal colonies. DNA of 12 genetic clones was PCR-amplified (5′-TCTGCCTGGCTTATTCTTCC-3′ and 5′-AGTGCTGCTGGCCATTTAAG-3′) and sequenced upon pGEM-T (Promega) cloning. Four genetic clones showed mutations in both alleles and in the one containing mutation that produced a shorter version of Bmal1 protein was further characterized by RT-qPCR and Western blot.

RNA was isolated using Trizol reagent (Thermo Fisher Scientific), reverse transcribed using RevertAid Frist Strand cDNA Synthesis kit (Thermo Fisher Scientific), and analysed by SYBRG real-time PCR using GoTaq qPCR Master Mix (Promega). Primers used are provided in Table S1.

Table S1 Sequences of primers used in RT-qPCR and CRISPR-Cas9 experiments.

Immunofluorescence analysis of mESCs colonies was carried out as described previously (Landeira et al, 2015). Briefly, the cells were fixed for 20 min in 2% paraformaldehyde, permeabilized 5 min in 0.4% Triton-X100, and blocked for 30 min in phosphate buffer saline supplemented with 0.05% Tween 20, 2.5% bovine serum albumin, and 10% goat serum. Primary antibodies used were Ms anti-Oct4 (611202; BD Biosciences), Rb anti-Nanog (RCAB002P-F; Cosmo Bio), and Ms anti-Nestin (MAB353; Merck). Secondary antibodies used were goat antimouse Alexa Fluor 488 (A-11001; Thermo Fisher Scientific) and goat antirabbit Alexa Fluor 555 (A-21429; Thermo Fisher Scientific). Slides were imaged by widefield fluorescence microscope Zeiss Axio Imager and images were analysed using Image J.

Separation of cytosol and nucleus fractions was carried out by resuspending 4 × 10⁶ cells in 80 μl of cell lysis buffer, pH 7.9 (10 mM Hepes, 10 mM KCl, 100 μM EGTA, 100 μM EDTA, and freshly added 1 mM DTT, 1 mM PMSF, and protease inhibitors). After 15-min incubation on ice, 10 μl of cell lysis buffer with NP40 10% was added. Cytoplasmic supernatant was recovered by spinning at 16,200g, 4°C for 5 min. Pelleted nuclei were resuspended in 80 μl of nuclear lysis buffer (20 mM Hepes, 400 mM NaCl, 1 mM EGTA, 1 mM EDTA, and freshly added 1 mM DTT, 0.5 mM PMSF, and protease inhibitors). Soluble nuclear supernatant was recovered by spinning at 13,000 rpm, 4°C for 5 min. Cytoplasmic and nuclear fractions were mixed with equal volumes of 2× Laemmli loading buffer. In experiments where cell fractionation was not carried out, Western blots were carried out in whole-cell extracts lysed with 2× Laemmli loading buffer and using standard procedures. The following primary antibodies were used: Ms anti-Oct4 (611202; BD Biosciences), Rb anti-Nanog (RCAB002P-F; Cosmo Bio), Gt anti–Lamin B (SC-6216; Santa Cruz), Ms anti-ActB (A4700; Sigma-Aldrich), Rb anti-Bmal1 (ab93806; Abcam), Ms anti-MAP Kinase 2 (05-157; Merck), and Rb anti-Ezh2 (C15410039; Diagenode). Secondary species-specific antibodies conjugated to horseradish peroxidase were used: anti-rabbit-HRP (NA931; GE Healthcare), anti-mouse-HRP (NA931; GE Healthcare), and anti-goat-HRP (ab97110; Abcam). Clarity Western ECL reagents (Bio-Rad) was used for detection.

Total RNA of JM8 wild-type and Bmal1−/− mESCs growing in the 2i + LIF medium was isolated using Trizol (Thermo Fisher Scientific), and mRNA library preparation was carried out using TruSeq Stranded mRNA kit (Illumina). 30 million 75-bp paired end reads were analysed per sample. Two biological replicates were carried out for each cell line. Library preparation and Illumina sequencing was carried out at National Center for Genomic Analysis-Center for Genomic Regulation (CNAG-CRG). Raw FASTQ files were quality filtered and aligned to the mouse genome (version mm9) using STAR 2.5.2 (Dobin & Gingeras, 2015). Gene expression was normalized with trimmed mean of M values (Robinson & Oshlack, 2010) implemented in NOISeq R package (Tarazona et al, 2015). Differential expression analysis was performed with DESeq2 package (Love et al, 2014). MA plot was generated with edgeR (McCarthy et al, 2012) and heat maps were generated with pheatmap package. mRNA of deregulated genes (log2 fold change greater than 1 or smaller than −1 and adjusted P-value < 0.05) were used as input gene lists in DAVID bioinformatic gene enrichment analysis. List of differentially expressed genes is provided in Table S1.

Single-cell RNA expression of Bmal1 and Oct4 in developing blastocysts was extracted from published data (Deng et al, 2014). The cells were classified by the level of expression of Oct4 in four quartile groups (Oct4 low correspond to bottom quartile and Oct4 high corresponds to top quartile), and average expression of Bmal1 was plotted.

Datasets are available at GEO-NCBI with accession number (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE134884↗).