ATRX histone binding and helicase activities have distinct roles in neuronal differentiation

Abstract

ATRX is a chromatin remodeler, which is mutated in ATRX syndrome, a neurodevelopmental disorder. ATRX mutations that alter histone binding or chromatin remodeling activities cluster in the PHD finger or the helicase domain respectively. Using engineered mouse embryonic stem cells that exclusively express ATRX protein with mutations in the PHD finger (PHDmut) or helicase domains (K1584R), we examine how specific ATRX mutations affect neurodifferentiation. ATRX PHDmut and K1584R proteins interact with the DAXX histone chaperone but show reduced localization to pericentromeres. Neurodifferentiation is both delayed and compromised in PHDmut and K1584R, and manifest differently from complete ATRX loss. We observe reduced enrichment of PHDmut protein to ATRX targets, while K1584R accumulates at these sites. Interestingly, ATRX mutations have distinct effects on the genome-wide localization of the polycomb repressive complex 2 (PRC2), with PHDmut and ATRX knockout showing reduced PRC2 binding at polycomb targets and K1584R showing loss at some sites and gains at others. Notably, each mutation associated with unique gene signatures, suggesting distinct pathways leading to impaired neurodifferentiation. Our results indicate that the histone binding and chromatin remodeling functions of ATRX play non-redundant roles in neurodevelopment, and when mutated lead to ATRX syndrome through separate regulatory pathways.

Figure 1.

Characterization of ATRX PHD finger and helicase domain mutant mESCs. (A) Schematic of Flag and V5 tagged ATRX-KI, PHDmut, and K1584R proteins. PHD fingers are shown in grey and the helicase domain in blue. Conserved amino acid residues within each region are shown with the specific mutations in red. (B) Western blot for ATRX, DAXX and EZH2 in nuclear extract from WT, ATRX-KI, PHDmut and K1584R mESCs. V5 antibody was used to detect tagged proteins in ATRX-KI, PHDmut and K1584R mESCs. Actin serves as a loading control. (C) Quantification of ATRXKI, PHDmut, and K1584R distribution between cytosolic, nuclear soluble, and chromatin-bound fractions in mESC. Band intensity was measured with ImageJ for three independent Western blots. Vinculin was used as cytosolic marker. EZH2 and H3K4me3 served as nuclear soluble and chromatin bound fraction markers. Bar charts present mean values ±SEM. (D) Left: immunostaining of WT, PHDmut, and K1584R mESCs for ATRX (red) and DAPI (blue). Scale bar = 10 μm. Right: quantification of nuclei (n = 200–300) that show strong (black), weak (blue) or absent (white) ATRX signal at pericentromeres. (E) Flag immunoprecipitation of ATRX in WT, ATRX-KI, PHDmut and K1584R mESCs. Inputs and immunoprecipitates were analyzed for ATRX, DAXX and EZH2 with antibodies as labeled on the right. (F) Left: representative image of proximity-ligation assay (PLA), showing in situ co-localization between ATRX and DAXX (red) in WT, PHDmut and K1584R mESC. DNA was stained with DAPI (blue). Scale bar = 10 μm. Right: quantification of PLA foci number per nucleus (n = 100). Lines on scatter plots represent 25th-, median and 75th-percentile. Statistical significance was determined by Welsh's test.

Figure 2.

ATRX PHD finger and helicase domain mutants display neurodifferentiation defects. (A) Immunostaining of WT, PHDmut, K1584R and ATRXKO for OCT4 and NANOG (red) and DAPI (blue) in mESC. Scale bar = 10 μm. (B) Bar chart showing expression of Klf4, Pou5f1 (Oct4), and Nanog as transcripts per million (TPM) in the mESC state. Circles indicate individual biological replicates. Data are presented as mean values ± SEM. (C) Schematic of neural differentiation. Withdrawal of LIF/2i and addition of basic Fibroblast Growth Factor (bFGF) results in the generation of EpiLCs. Addition of Smoothened Agonist (SAG) to EpiLCs promotes NPC formation. (D) Representative phase contrast images of WT, PHDmut, K1584R and ATRX KO at the mESC, EpiLC and NPC stages. Scale bar = 100 μm. (E) Representative images of crystal violet staining assay for cell populations in WT, PHDmut, K1584R and ATRX KO at day 6 of NPC differentiation.

Figure 3.

ATRX PHDmut and K1584R mutations deregulate unique gene groups in mESCs and NPCs. (A) MA plot of RNA-seq expression of 13,958 genes between WT and PHDmut mESCs. For all MA plots in this figure, red dots and numbers indicate differentially expressed genes that are upregulated in mutant cells (adjusted P-value ≤0.05, log₂ fold change > 1); blue dots and numbers indicate differentially expressed genes that are downregulated in mutant cells (adjusted P-value ≤ 0.05, log₂ fold change < –1). (B) MA plot of RNA-seq expression of 13,347 genes between WT and K1584R mESCs. (C) Venn diagram showing overlaps between upregulated and downregulated differentially expressed genes in PHDmut mESCs and K1584R mESCs (red and blue dots in Figure 3A and B). Total numbers of differentially expressed genes and overlaps are shown. (D) MA plot of RNA-seq expression of 14,879 genes between WT and PHDmut NPCs. (E) MA plot of RNA-seq expression of 15,079 genes between WT and K1584R NPCs. (F) Venn diagram showing overlaps between upregulated and downregulated differentially expressed genes in PHDmut NPCs and K1584R NPCs (red and blue dots in D and E). Total numbers of differentially expressed genes and overlaps are shown. (G) Immunostaining of WT, PHDmut, K1584R and ATRXKO for β3-Tubulin (red) and DAPI (blue) at day6 of NPC differentiation. Scale bar = 10 μm. (H) Heat map of expression select neuroectoderm related genes in ESC and NPC. Each column represents the mean Z-score for three independent biological replicates from RNA-seq.

Figure 4.

PHDmut and K1584R mutations have distinct effects on genome wide ATRX binding. (A) Genomic distribution of 1,958 ATRX ChIP-Seq peaks called across two replicates in WT mESCs. (B) Heatmap of ATRX ChIP-Seq signal across 1,958 ATRX peaks in WT, PHDmut and K1584R mESCs. Reads per million, RPM. (C) Genome browser view of the Trak2 gene showing ATRX ChIP signal (reads per million, RPM) in WT, PHDmut and K1584R mESCs. (D) Boxplots showing log₂ edgeR-normalized expression fold change of WT ATRX target genes in PHDmut compared to WT (purple); WT ATRX target genes in K1584R compared to WT (gray); and K1584R-only ATRX target genes in K1584R compared to WT (white). Box, 25th percentile—median—75th percentile. Whiskers extend to 1.5× interquartile range; outliers not displayed. (E) MA plot of RNA-Seq expression for 13,347 genes between WT and K1584R mESCs. Purple dots indicate K1584R-only ATRX target genes that overlap an ATRX peak identified only in K1584R mESCs. (F) Top 5 most significantly enriched processes in K1584R-only ATRX targets.

Figure 5.

ATRX PHDmut and K1584R mutations differentially alter PRC2 occupancy. (A) Violin plot of spike-normalized EZH2 CUT&RUN signal in WT, ATRX KD and ATRX KO mESCs. Box, 25th percentile—median—75th percentile. Whiskers extend to 1.5× interquartile range; outliers not displayed. (B) Heatmap of spike-in normalized EZH2 CUT&RUN signal across 19 796 WT EZH2 peaks in WT and PHDmut mESCs. (C) Genome browser view of the Zic5 gene showing spike-in normalized EZH2 CUT&RUN signal in WT and PHDmut mESCs. (D) Heatmap of spike-in normalized EZH2 CUT&RUN signal in WT and K1584R mESCs across EZH2 peaks that are gained (6,029) or lost (4,429) in K1584R compared to WT. (E) Genome browser view of the Fa2h and Zfp960 genes showing spike-in normalized EZH2 CUT&RUN signal in WT and K1584R mESCs. (F) MA plot of RNA-seq expression of 13,347 genes between WT and K1584R mESCs. Red dots indicate differentially expressed genes (adjusted P-value ≤ 0.05, absolute log₂ fold change > 1); blue dots and numbers indicate differentially expressed genes with gained promoter EZH2 in K1584R. (G) Boxplots showing expression of poised enhancer related genes that are activated in NPCs. Box represents 25th, median and 75th percentile. The whiskers correspond to minimum and maximum values. The Wilcoxon signed-rank test was used for calculation significance calculation, ****P < 0.001. (H) Heatmap of expression of 298 poised enhancer related genes that are induced upon differentiation and grouped by k-means clustering of Z-scores at the NPC stage in WT, PHDmut, K1584R and ATRX KO. Each column represents the mean Z-score for three independent biological replicates from RNA-seq. Z-scores were computed across WT, PHDmut, K1584R and ATRX KO mESCs and NPCs. (I) Left: bar chart showing expression of Sox17 and Col4a1 as transcripts per million (TPM) in the NPC state. Circles indicate individual biological replicates. Data are presented as mean values ± SEM. Right: genome browser view of the Sox17 and Col4a1 genes showing spike-in normalized EZH2 CUT&RUN signal in WT, PHDmut and K1584R mESCs.