Phosphorylation of Tet3 by cdk5 is critical for robust activation of BRN2 during neuronal differentiation

Abstract

Tet3 regulates the dynamic balance between 5-methylcyotsine (5mC) and 5-hydroxymethylcytosine (5hmC) in DNA during brain development and homeostasis. However, it remains unclear how its functions are modulated in a context-dependent manner during neuronal differentiation. Here, we show that cyclin-dependent kinase 5 (cdk5) phosphorylates Tet3 at the highly conserved serine 1310 and 1379 residues within its catalytic domain, changing its in vitro dioxygenase activity. Interestingly, when stably expressed in Tet1, 2, 3 triple-knockout mouse embryonic stem cells (ESCs), wild-type Tet3 induces higher level of 5hmC and concomitant expression of genes associated with neurogenesis whereas phosphor-mutant (S1310A/S1379A) Tet3 causes elevated 5hmC and expression of genes that are linked to metabolic processes. Consistent with this observation, Tet3-knockout mouse ESCs rescued with wild-type Tet3 have higher level of 5hmC at the promoter of neuron-specific gene BRN2 when compared to cells that expressed phosphor-mutant Tet3. Wild-type and phosphor-mutant Tet3 also exhibit differential binding affinity to histone variant H2A.Z. The differential 5hmC enrichment and H2A.Z occupancy at BRN2 promoter is correlated with higher gene expression and more efficient neuronal differentiation of ESCs that expressed wild-type Tet3. Taken together, our results suggest that cdk5-mediated phosphorylation of Tet3 is required for robust activation of neuronal differentiation program.

Figure 1.

The catalytic domain of Tet3 is phosphorylated by Cdk5. (A) Schematic depiction of 21 phosphorylation sites on human Tet3 that were identified by mass spectrometry analysis. The green circles denote two conserved serine residues that were also phosphorylated in mouse Tet3 protein. (B) Amino acid sequences of the two conserved SPx(K/R) motifs found in the catalytic domain (CD) of mouse Tet3 protein, the single (AS) and double (AA) alanine mutants generated. (C) Phosphorylation of mouse Tet3CD by Cdk5 at S1318 and S1387 residues in vitro. GST-tagged wild-type (WT), single (AS) and double (AA) alanine mutant Tet3CD proteins were subjected to in vitro kinase assay in the presence or absence of Cdk5/p25. The products were probed with antibodies against GST, phosphor-serine with SPx(K/R) motif (34B2) and phosphorylated S1318 (pTet3). (D, E) Mouse Tet3 is phosphorylated at both S1318 and S1387 residues in vivo. Lysates from HEK293 cells transfected with either Flag-tagged WT, SA or AA Tet3CD constructs were immunoprecipitated (IP) and probed with FLAG, phosphor-serine 34B2 and pTet3 antibodies. Untransfected cells (–) were used as negative control.

Figure 2.

Cdk5-mediated phosphorylation of Tet3 increases its in vitro catalytic activity. (A. B) Cdk5 phosphorylates human Tet3 protein at S1310 in vivo. (A) HEK293 cells expressing Flag-tagged human Tet3 was treated with either DMSO or 20 μM of roscovitine (Ros) for 24 h. (B) HEK293 cells were co-transfected with Flag-tagged human Tet3 and either scrambled (scr) or Cdk5 siRNA for 48 hr. Cell lysates were western blotted with FLAG, pTet3, β-actin, GFP and Cdk5 antibodies. (C) Phosphorylated mouse Tet3CD exhibits increased catalytic activity in vitro. Flag-tagged mouse Tet3CD protein was subjected to in vitro kinase assay in the presence or absence of Cdk5/p25. One third of the proteins was western blotted with FLAG and 34B2 antibodies (Top). Another one third of the proteins were used for in vitro dioxygenase assay with equal amount of 5mC DNA. The DNA products were visualized by ethidium bromide (EtBr) staining on 2% agarose gel, Southern blotted and probed with 5hmC antibody (Bottom). (D) Amino acid sequences showing the two conserved SPx(K/R) motifs in the catalytic domain of human Tet3 protein and the double alanine (AA) mutants generated in this study. (E–G) Alanine substitution at S1310 and S1379 of human Tet3 protein leads to lower 5hmC level in HEK293 cells. (E) FACS sorted HEK293 cells that expressed either wild-type (WT) or phosphor-mutant (AA) full-length Flag-tagged human Tet3 were lysed and immunoblotted with pTet3, FLAG and β-actin antibodies. (F) Genomic DNA from untransfected HEK293 (–) and FACS sorted cells that expressed either WT or AA Flag-tagged human Tet3 were dot blotted and probed with 5hmC antibody. (G) Quantification of 5hmC level in transfected HEK293 cells where data are mean ± S.E.M. (n = 3).

Figure 3.

Correlation of gene expression with 5hmC enrichment at gene bodies in ESCs that expressed either wild-type or phosphor-mutant Tet3. (A) Stable Tet1, 2, 3 triple knockout (TKO) mouse ESC lines that expressed comparable level of either Flag-tagged wild-type (Wt25) or double phosphor-mutants (AA10) human Tet3 were generated by lentiviral infection. Lysates from parental (–), Wt25 and AA10 were immunoblotted with FLAG, pTet3, Lamin B1 (LMNB1) and Oct4 antibodies. (B) Dot blot revealed no significant difference in the global 5hmC level between Wt25 and AA10 mouse ESC lines. Parental Tet TKO line (–) was used as negative control (Top). Quantification data are mean ± S.D. (n = 4) (bottom). (C) Density plot showing high reproducibility in 5hmC-DNA immunoprecipitation (DIP) signals across differential 5hmC regions (DhMRs) between two biological replicates from the same genotype, but lower Pearson correlation between Wt25 and AA10 cells. RPKM, reads per kilobase million mapped reads. (D) Genomic distribution of the annotated peaks of differential 5hmC regions between Wt25 and AA10 mouse ESCs. Blue bars indicate peaks where the 5hmC level is higher in Wt25 cells whereas orange bars indicate peaks where the level of 5hmC is higher in AA10 cells. UTR, untranslated region; TSS, transcription start site; TTS, transcription termination site. (E) Boxplots of mRNA expression of all genes (gray); genes with the presence of higher 5hmC differential level in the intergenic region (blue, IG); at the promoter (green, Pr); at the gene bodies (yellow, Ex: exons, In: introns) and at the TTS (white) in Wt25 and AA10 cells. Promoters were defined as ± 1kb from the TSS. Thick lines indicate mean and whiskers extend to ±1.5 of the interquartile range. Significance levels were calculated by Wilcoxon signed-rank test relative to ‘all genes’ category. WT: * P = 0.0003, ** P = 9.8e–5, *** P = 2.8e–8. AA: * P < 0.008, ** P = 7e–5, *** P = 1.9e–8. FPKM, fragments per kilobase million mapped reads.

Figure 4.

Phosphorylation at S1310 and S1379 leads to higher 5hmC and expression of neuronal genes. (A) Heatmap of differentially expressed genes that were determined by DEseq2 (adjusted P-value < 0.05) between two biological replicates (Rep1 and Rep2) of Wt25 and AA10 mouse ESCs. (B) Venn diagrams showing the overlaps in the number of genes that have higher level of 5hmC and upregulation of mRNA expression in Wt25 (top) and AA10 (bottom) mouse ESCs. (C) Enriched GO terms for genes with higher level of 5hmC and mRNA expression in Wt25 compared to AA10 mouse ESCs. (D) Enriched GO terms for genes with higher level of 5hmC and mRNA expression in AA10 compared to Wt25 mouse ESCs. (E) Validation of differentially expressed genes in Wt25 and AA10 mouse ESCs by quantitative PCR (qPCR) using Gapdh expression as reference. Representative data are presented as mean ± S.D. (triplicate qPCR reactions). Similar expression pattern was observed in biological replicate as well as in independent mouse ESCs lines (Supplementary Figure S5).

Figure 5.

Phosphor-mutant Tet3 ESCs show impaired neuronal differentiation. (A) Detection of S1318 phosphorylation of endogenous Tet3 in the prefrontal cortex (PFC) tissues from C57BL/6 mice but not in mouse E14 ESC. (B) Retinoic acid (RA) was used to induce in vitro neuronal differentiation of mouse ESCs. Lif, Leukemia inhibitory factor; EB, embryoid bodies; NPC, neuronal progenitor cells; +N2 and +B27 indicate the addition of the respective culture medium for the indicated duration (48 h). (C) Detection of S1318 phosphorylation of endogenous Tet3 in neurons that were differentiated from E14 ESC. Neurons were cultured in N2 medium for 48 h. (D) Generation of stable Tet3 knockout (KO) mouse ESC lines that expressed either empty lentiviral vector (Em), Flag-tagged wild-type (Wt14) or double phosphor-mutants (AA13) human full-length Tet3 gene. Lysates from were probed with FLAG, pTet3, Tet3, MAP2, β-actin and LMNB1 antibodies. (E) Expression of Pax6 and BRN2 genes is significantly lower in Tet3 KO NPCs that were rescued with phosphor-mutant AA Tet3. Wt14 and AA13 lines were subjected to qPCR using GAPDH as reference. Data are shown as mean ± S.E.M. (Pax6,n = 6, * P = 0.04; BRN2, n = 5, ** P = 0.014). (F) Representative western blot of lysates from Em, Wt14 and AA13 NPCs using BRN2, cleaved caspase-3 and β-actin antibodies.

Figure 6.

Impaired differentiation of phosphor-mutant NPCs is correlated with unique epigenetic signature and lower BRN2 expression. (A) Phosphor-mutant (AA) Tet3 has higher binding affinity to histone H2A.Z in mouse ESCs. Lysates from Tet3 KO mouse ESCs that expressed either wild-type (Wt) or phosphor-mutant (AA) Tet3 were IP with either IgG or FLAG antibodies. The IP factions were then probed for FLAG and H2A.Z. (B) Quantification of IP blots with data presented as mean ± S.E.M. (four replicates of Wt14/AA13; one replicate of Wt1/AA6 lines; * P = 0.003, paired one-tailed t-test). (C) Schematic of BRN2 gene promoter that was reported to contain differentially DNA methylated sites and H2A.Z nucleosomes. (D) H2A.Z occupancy in AA13 is specifically higher at the BRN2 promoter compared to Wt14 ESCs. Representative ChIP data are presented as mean ± S.D. (triplicate qPCR reactions, P < 0.05). Similar pattern was also observed in another biological replicate (Supplementary Figure S6F). (E) Reduced 5hmC level at BRN2 gene promoter in phosphor-mutant (AA13) NPCs. The 5hmC level in Wt14/AA13 NPCs was determined with 5hmC-DIP followed by qPCR and presented as mean ± S.E.M. (n = 4, * P = 0.004, ** P = 0.0007, paired one-tailed t-test). (F, G) Phosphor-mutant Tet3 (AA13) line has lower level of MAP2 protein in terminally differentiated neurons. (F) Representative western blot of neurons that were cultured in B27 medium for 48 h and probed with MAP2 and β-actin antibodies. (G) Quantification of four independent experiments with data represented as mean ± S.E.M. (n = 4, * P < 0.04, paired one-tailed t-test). (H) Quantification of the density of MAP2-positive neurons after long-term culture in B27 medium. Data are presented as mean ± S.E.M. (Wt14/AA13, 12 days of culture, Figure 6I, * P < 0.04; Wt1/AA6, 13 days of culture, Supplementary Figure S8; Wt2/AA3, 8 days of culture, Supplementary Figure S9; ** P < 0.003, unpaired one-tailed t-test. See also Supplementary Table S7B for detailed analysis). (I) Wt14 Tet3 KO mouse ESCs line differentiated efficiently to form higher density of MAP2-positive neurons when compared to AA13 line. Same number of NPCs were seeded and neurons were stained with MAP2 antibody after 12 days of culture in B27 medium.