Neuronal splicing of the unmethylated histone H3K4 reader, PHF21A, prevents excessive synaptogenesis

Abstract

PHF21A is a histone-binding protein that recognizes unmethylated histone H3K4, the reaction product of LSD1 histone demethylase. PHF21A and LSD1 form a complex, and both undergo neuron-specific microexon splicing. The PHF21A neuronal microexon interferes with nucleosome binding, whereas the LSD1 neuronal microexon weakens H3K4 demethylation activity and can alter the substrate specificity to H3K9 or H4K20. However, the temporal expression patterns of PHF21A and LSD1 splicing isoforms during brain development and their biological roles remain unknown. In this work, we report that neuronal PHF21A isoform expression precedes neuronal LSD1 expression during human neuron differentiation and mouse brain development. The asynchronous splicing events resulted in stepwise deactivation of the LSD1-PHF21A complex in reversing H3K4 methylation. An unbiased proteomics survey revealed that the enzymatically inactive LSD1-PHF21A complex interacts with neuron-specific binding partners, including MYT1-family transcription factors and post-transcriptional mRNA processing proteins such as VIRMA. The interaction with the neuron-specific components, however, did not require the PHF21A microexon, indicating that the neuronal proteomic milieu, rather than the microexon-encoded PHF21A segment, is responsible for neuron-specific complex formation. Finally, by using two Phf21a mutant mouse models, we found that Phf21a neuronal splicing prevents excess synapse formation that otherwise would occur when canonical PHF21A is expressed in neurons. These results suggest that the role of the PHF21A microexon is to dampen LSD1-mediated H3K4 demethylation, thereby containing aberrant synaptogenesis.

Keywords: LSD1; PHF21A; histone demethylase; histone methylation; microexon splicing.

Figure 1

The expression of PHF21A-n increases faster than that of LSD1-n.A, schematic representation of LSD1 and PHF21A neuronal splicing events. B, mRNA levels of LSD1 and PHF21A isoforms in LUHMES cells. Cells were differentiated into neurons as indicated and harvested from day 3 to day 12, analyzed by RT-PCR (n = 2). C, expression of PHF21A and associated proteins in LUHMES cells and 293T cells examined by Western blot analysis using antibodies as indicated. D, mRNA levels of LSD1 and PHF21A isoforms in the developing mouse brain. Whole-cell lysates prepared from mouse whole brains (E12.5 and E13.5) and cortices (from E14.5 to P2) at indicated periods were analyzed by RT-PCR. E, expression of PHF21A and LSD1 proteins in the developing mouse brain. Whole-cell lysates were subjected to Western blot analysis. F, quantification of Western signals for PHF21A-c, PHF21A-n, and LSD1 normalized by histone H3. PHF21A-c, and LSD1 levels were further normalized to E12.5. G, the ratio of PHF21A protein isoforms in the developing mouse brain based on the Western signals. F & G, day 12.5 to 13.5: n = 1, pooled embryonic brain is used. Day 14.5-P2: Mean ± S.E.M., n = 3, technical replicates.

Figure 2

A weaker H3K4 demethylation activity of the PHF21A-n:LSD1-c complex.A, schematic of the demethylation assay. The PHF21A-containing complex was immunoprecipitated from LHUMES cell nuclear extracts and incubated with designer nucleosomes with specific lysine di-methylations. B-D, Western blot using the antibodies for indicated histone methylations to detect demethylase activity. The reactions were carried out using H3K4me2- (B), H3K9me2- (C), and H4K20me2- (D) nucleosomes. The appearance of mono-methylated lysine indicates the demethylation activity. The abundance of LSD1, PHF21A, and total H3 (c-term) were examined with specific antibodies. H3K4me1/me2 designer nucleosomes and 293T nuclear extract serve as specificity and positive control for the histone antibodies, respectively.

Figure 3

No detectable demethylation activity of the mature neuronal complex with PHF21A-n and LSD1-n. The demethylation assay of immunoprecipitated PHF21A complexes from the P0 mouse cortices using the designer nucleosomes carrying H3K4me2 (A), H3K9me2 (B), and H4K20me2 (C). Neither a reduction of di-methylation nor the appearance of mono-methylation was observed with any demethylation reactions.

Figure 4

Proteomics analysis of PHF21A-containing complex in MEFs and neurons.A, silver staining of proteins co-precipitated by an anti-PHF21A antibody. Nuclear extracts from MEFs and cortical neurons (DIV7) were used. Asterisks indicate proteins specifically found in PHF21A-immunoprecipitates. B-C, Co-IP-MS analysis of PHF21A-associated proteins in MEF (B) and cortical neurons (C). Volcano plots of identified proteins (n = 3). Green dots: the bait protein, PHF21A. Black and red dots: the interactors (Cutoff: log2FC > 1.5, p-value < 0.1, and peptide ≥ 6). Red dots: neuron-specific interactors with the same cut-off. Gray dots: proteins that did not pass the cutoff. The x-axis shows the log₂ FC (a-PHF21A antibody/control IgG), and the y-axis denotes -log₁₀p-values. D, Venn diagram of PHF21A interactor in MEFs (orange) versus neurons (blue). E, domain organization of neuron-specific PHF21A interactors. F, Co-IP-Western assays to test the interaction between PHF21A, BRAF35, or iBRAF. Asterisks indicate nonspecific IgG bands. G, expression of neuron-specific PHF21A interactors in the developing mouse brains. Whole-cell lysates prepared from mouse whole brains (E12.5 and E13.5) and cortices (from E14.5 to after birth 2) at indicated periods were subjected to Western blot analysis using antibodies as indicated. The H3 blot was re-used from Fig. 1E.

Figure 5

Binding partners of PHF21-n and ectopic PHF21A-c in neurons are highly similar.A, generation of Phf21a-neuronal exon KO allele (Δn). Black arrows: PCR primers. Red triangles: CRISPR cut sites. B, genotyping PCR analysis of Phf21a^+/+, Phf21a^+/Δn, and Phf21a^Δn/Δn mice using the primer sets shown in (B). C, representative picture of P15 Phf21a^+/+ (left) and Phf21a^Δn/Δn (right) mice. Scale bar represents 1 cm. D, Western blot analysis of lysates prepared from cortical neurons (DIV7) isolated from Phf21a^+/+, Phf21a^+/Δn, and Phf21a^Δn/Δn mice using antibodies as indicated. PHF21A-c instead of PHF21A-n is expressed in Phf21a-n^Δn/Δn neuron. E, silver staining of proteins co-precipitated by anti-PHF21A antibody using the nuclear extracts from P0 cortices of the indicated genotypes. Asterisks indicate proteins specifically found in PHF21A-immunoprecipitates. F, functional enrichment analysis of PHF21A-interacting proteins with Metascape (41). The statistically significant PHF21A interactors in Phf21a^+/+ cortices were used. Log₁₀p-value < −10. The molecular networks that contain proteins validated by reciprocal IP experiments are presented. G, reciprocal Co-IP-Western assays to validate the interaction between PHF21A and newly identified interaction partners. H and I, volcano plots of Co-IP-MS analysis of PHF21A-interacting proteins (n = 3–4) using Phf21a^+/+ (H) and Phf21a^Δn/Δn (I) cortices. Green dots: the bait protein, PHF21A. Black and red dots: statistically significant interactors (Cutoff: FC > 2, Padj-value < 0.01, and peptide ≥ 6). Red dots: neuron-specific proteins identified in Fig. 4C that pass the cutoff. Gray dots: proteins that do not pass the threshold. The x-axis shows the log₂ FC (a-PHF21A antibody/control IgG), and the y-axis denotes -log₁₀p-values. J, Venn diagram of PHF21A interactor in Phf21a^+/+ cortex (blue) versus Phf21a^Δn/Δn cortex (orange). K, scatter plot comparing the log2FC (a-PHF21A antibody/control IgG) of all PHF21A-associated proteins between WT and Phf21a^Δn/Δn cortices. The R² value and the fit of the linear regression are indicated. Green dots: the bait protein, PHF21A. Black and red dots: statistically significant interactors in Phf21a^+/+ a-PHF21A and Phf21a^Δn/Δn (Cutoff: Padj-value < 0.01, and peptide ≥ 6). Red dots: neuron-specific proteins identified in Fig. 4C that pass the cutoff. Blue dots: statistically significant interactors in Phf21a^+/+ a-PHF21A. Orange dots: statistically significant interactors in Phf21a^Δn/Δn. The x-axis shows the log₂ FC of Phf21a^+/+ (a-PHF21A antibody/control IgG). The y-axis shows the log₂ FC of Phf21a^Δn/Δn (a-PHF21A antibody/control IgG).

Figure 6

Characterization of synapse and dendrite growth in Phf21a-mutant mice.A-F, synapse density in cultured cortical neurons. Cortical neurons from E16.5 embryos (2–6 per genotype) were cultured for 14 days in vitro (DIV14). A and D, the representative images of excitatory synapses (A), visualized by the immunofluorescent colocalization of postsynaptic marker PSD95 (green) and presynaptic marker vGlut (red), formed on cortical pyramidal neurons. Inhibitory synapses (D) were visualized by the colocalization of postsynaptic Gephyrin (green) and presynaptic vGat (red). Nuclei were stained with DAPI (blue). Scale bars represent 10 μm. B-F, quantification of excitatory (B and C) and inhibitory (E and F) synapse density in Phf21a-null (B and E) and Phf21a-Δn (C and F) mutants (mean ± S.E.M., n = 12, ∗p < 0.05, ∗∗∗∗p < 0.001, one-way ANOVA with Tukey post hoc analysis). G-L, dendritic morphology analysis of CA1 and PFC pyramidal cells of Phf21a mutants. G and J, the representative Golgi staining images of CA1 and PFC pyramidal cells of Phf21a^+/+, Phf21a^+/−, and Phf21a^Δn/Δn mice. The upper panel shows the dendritic morphology of pyramidal cells, and the lower panel shows the dendritic spines of a dendritic segment. Scale bars represent upper: 100 μm, lower: 5 μm. H and I, quantification of dendritic length in Phf21a-null (H) and Phf21a-Δn (I) mutants (n = 72). K and L, quantification of dendritic spine density in Phf21a-null (K) and Phf21a-Δn (L) mutants (n = 72). The boxes show the lower and upper quartile values, respectively. Both apical and basal dendrites from 12 neurons per animal were quantified: three of 2-month-old males per each genotype were used. The median is indicated with a bar within the box, and the whiskers denote the 1.5 × data range of the box. (∗p < 0.05, ∗∗∗∗p < 0.001, Unpaired Student’s t test).