Concentration-dependent formation of H3K4me3-associated condensates by NUP98-KDM5A.

A. Schematic of NUP98-KDM5A construct, with the N-terminal region of NUP98 (amino acids 1-513) fused to the C-terminal region of KDM5A (1486-1690), which includes the H3K4me3-binding PHD3 domain, and NSPARC images of NUP98-KDM5A-mEGFP in U2OS nuclei at different concentrations showing sub-resolution-limit puncta. B. Schematic of the NUP98-KDM5A-mCherry construct and NSPACR images in U2OS nuclei at different concentrations. C. Fraction of signal in condensates (fluorescent protein signal within condensates relative to total nuclear signal) and condensate density (number of condensates per µm2) as a function of nuclear protein concentration in individual cells (n = 282 cells for mEGFP tagging and n = 208 for mCherry tagging). D. Fixed U2OS nuclei expressing NUP98-KDM5A-mEGFP and immunostained for H3K4me3 with Alexa Fluor 546 (AF546), showing images with different mean nuclear mEGFP intensities (INUP98-KDM5A), 2D histograms of pixel intensities, and the Pearson correlation, R, between the images of the two channels. E. Schematic of the chromatin-binding-deficient NUP98-KDM5A (W1625A, W1635A) construct, and the same set of analyses as in D. F. Image Pearson correlation, R, as a function of expression level, INUP98-KDM5A, for U2OS cells expressing NUP98-KDM5A-mEGFP (n = 134 cells) or NUP98-KDM5Amut-mEGFP (n = 132).

Binding of the PHD3 domain of NUP98–KDM5A to H3K4me3 promotes the formation of NUP98–KDM5A condensates.

A. Fluorescence polarization (FP) assays to determine the KD of NUP98-KDM5A WT or W1625A mutant for a 10-mer FAM-labelled H3K4me3 peptide (10 nM). Error bars represent standard deviations from technical triplicates. KD values show means and 95% C.I.. N.D.: Not determined because no binding was observed. B. Competitive displacement of 2 µM NUP98-KDM5A (WT) or MBP-PHD3 (KDM5A residues 1601-1662) bound to 10-mer FAM-H3K4me3 (10 nM) by increasing concentrations of unlabeled 10-mer H3K4me3 peptide. Error bars represent standard deviations from technical triplicates. IC50 values show means and 95% C.I.. C-F. Representative confocal images of C 100 nM AF647 labelled H3K4me3 or H3K4me0 nucleosomal arrays, D AF488 labelled NUP98-KDM5A (800 nM) with AF647 H3K4me3 nucleosomal arrays (0 or 100 nM), E AF488 NUP98-KDM5A (800 nM) with AF647 H3K4me0 nucleosomal arrays (0 or 100 nM), or F AF488 NUP98-KDM5A W1625A (800 nM) with AF647 H3K4me3 nucleosomal arrays (0 or 100 nM). All images are maximum intensity projections (scale bar = 10 µm). G: The fraction of protein fluorescence signal in condensates relative to total signal in the image for the conditions in D-F. n = 3 technical replicates; ns: P > 0.05; *: P ≤ 0.05; **: P ≤ 0.01, parametric unpaired t-test.

NUP98-KDM5A condensate formation at loci with varying H3K4me3 density.

A. Schematic of the 100 kb genomic region surrounding each FISH-labeled locus. Bars indicate genes. Black peaks at the bottom show H3K4me3 ChIP-seq signal in HEK293 cells. FISH probe targeted regions are marked in magenta (Cy5-labeled probes) and yellow (Cy3-labeled probes). B. Images of combined FISH (magenta and yellow) and NUP98-KDM5A immunofluorescence (green) in HEK293 cells for the HOXB and MYC loci, showing a single z-slice with white lines marking the nucleus based on DAPI signal segmentation. Circles in the zoomed-in views of the boxed regions mark the radii for maximum and background signal calculation. C. Local NUP98-KDM5A signal (intensity around the locus) across expression levels as measured by mean NUP98-KDM5A nuclear intensity, INUP98-KDM5A, (HOXA, n = 135 loci; HOXB, n = 319; HOXD, n = 105; TNF, n = 340; MUC4, n = 111; MYC, n = 42). D. Boxplots of local NUP98-KDM5A signal for cells with low (INUP98-KDM5A = 0 ∼ 20 au) and high (INUP98-KDM5A = 20 ∼ 200 au) expression levels. The median is shown in red. The box extends from the first to the third quartile, and whiskers extend to the most extreme data points located within 1.5x the interquartile range.

Degree of expression correlates with H3K4me3 intensity for the dysregulated genes in NUP98:KDM5A leukemia patients.

A. Volcano plots indicate a change in gene expression in patients compared to healthy donors, with color encoding the degree of H3K4me3 intensity. Rows correspond to patients. B. Correlation plot of change in gene-expression with H3K4me3 intensity in patients with genes binned by methylation signal and expression fold change. Blue: intensities computed across TSS-centered windows, and Gray: across random sites for comparable window size. Error bars: SEM. C. Rank-ordered H3K4me3 intensity plot with color representing the fold-change in the gene-expression. Top ranked HOXA and HOXB protein coding genes with log2 fold change ≥ 0.5 are indicated. Inset: boxplot represents the fold-change in gene-expression for HOXA/B coding genes compared to all the other genes in the ranked set (Patient 1: n =16 for HOXA/B and 9205 for the rest; Patient 2: n =12 for HOXA/B and 4125 for the rest).

Live imaging of U2OS cell nuclei expressing NUP98-KDM5A-related constructs.

1. Nucleus expressing full-length KDM5A construct fluorescently tagged on the N-terminus. 2. Chromatin-binding-deficient version of the construct in (1). 3. Condensates formed by NUP98-KDM5A tagged with mEGFP on the N-terminus. 4. Nuclei expressing only the chromatin-binding C-terminal fragment of the NUP98–KDM5A fusion, derived from KDM5A (1486–1690). 1-3 show cells with comparable expression levels, whereas 4 shows two cells with high expression but no chromatin-associated puncta formation. All images are a single z-slice acquired on a standard confocal microscope.

NSPARC enhanced-resolution live imaging of NUP98-KDM5A condensates in U2OS cells.

A. Concentration-dependent formation of (Left) NUP98-KDM5A-mEGFP and (Right) NUP98-KDM5Amut-mEGFP condensates, showing single z-slices from 3D stacks. B. Quantification of NUP98-KDM5A-mEGFP and NUP98-KDM5Amut-mEGFP condensate formation across expression levels for the fraction of signal in condensates and the number density of condensates. C. Calibration between mEGFP and mCherry fluorescence intensities under the NSPARC microscopy settings used for live-cell imaging (see Materials and Methods). D. Calibration between EGFP fluorescence intensity and EGFP molar concentration in solution under the NSPARC microscopy settings used for live-cell imaging (see Materials and Methods).

Estimation of the relative abundance of KDM5A to NUP98 in patient and healthy donors.

A. Group level relative abundance of KDM5A: NUP98 transcripts quantified as the ratio of the mean expression across cells in each group. B. Distribution of per-cell KDM5A: NUP98 ratios for cells with detectable NUP98 expression (NUP98 > 0). Squares represent ‘mean’. Wilcoxon rank sum test, ns: not significant, *p<0.05, ***p< 0.001.

In vitro reconstitution of NUP98-KDM5A condensates.

A. Spinning disk confocal images of condensates formed in vitro with increasing concentration of the following AF488 labelled constructs: NUP98 (1-513)-KDM5A (1486-1690), the full length oncofusion; NUP98 (1-513)-KDM5A (1601-1662), which contains only the PHD3 domain of the KDM5A fusion partner; NUP98 (1-513)-KDM5A (1486-1690), where the C-terminal truncation of the KDM5A fusion partner removes the PHD3 domain; NUP98 (1-513), the NUP98 segment alone. B. In vitro phase separation assay of AF488 labelled NUP98-KDM5A (750 nM) with chicken erythrocyte polynucleosomes (0 or 2 µg). C. In vitro phase separation assay of AF488 labelled NUP98-KDM5A W1625A (750 nM) with chicken erythrocyte polynucleosomes (0 or 2 µg). All images in A-C show maximum intensity projections of Z-stacks (scale bar = 10 µm; n = 3). D. Fraction of signal in condensates for reactions with 750 nM AF488 labelled NUP98-KDM5A with 0 or 2 µg chicken erythrocyte polynucleosomes (left panel) or 750 nM AF488 labelled NUP98-KDM5A W1625A with 0 or 2 µg chicken erythrocyte polynucleosomes (right panel). The fraction of signal in condensates corresponds to the fluorescence intensity measured inside condensates relative to the total image intensity, reflecting the proportion of molecules partitioned into condensates. A parametric unpaired t-test was carried out to compare the average for the fraction of signal in condensates (n=3; ns, P>0.05; *, P≤0.05; ** P≤0.01).

Dynamics of NUP98-KDM5A condensates.

A. Representative images of the fluorescence recovery (inside yellow circle) of AF488 labelled NUP98-KDM5A (1 µM) mixed with 0 or 2 µg chicken erythrocyte polynucleosomes following photobleaching. Fluorescence recovery (inside yellow box) of U2OS cells expressing NUP98–KDM5A constructs tagged with mEGFP on either the C- or N-terminus after photobleaching. Normalized fluorescence recovery curves of photobleached regions (inside yellow marks) are plotted. Error bars are standard deviations of the means. B. SoRa-enhanced resolution live imaging of U2OS cells expressing NUP98-KDM5A constructs with an N- or C-terminal mEGFP tag. C. Quantification of mEGFP-NUP98-KDM5A and NUP98-KDM5A-mEGFP condensate formation across expression levels (C NUP98-KDM5A - Log 10) for the fraction of signal in condensates and condensate density (µm-2). D. Calibration between EGFP fluorescence intensity and EGFP molar concentration in solution under the SoRa microscopy settings used for live-cell imaging (see Materials and Methods).

Representative images of in vitro NUP98-KDM5A and chromatin co-phase separation assays.

A. AF488 labelled NUP98-KDM5A (0, 50 or 200 nM) with AF647 H3K4me3 nucleosomal arrays (0, 12.5 or 100 nM nucleosomes). B. AF488 labelled NUP98-KDM5A (0, 50 or 200 nM) with AF647 H3K4me0 nucleosomal arrays (0, 12.5 or 100 nM nucleosomes). C. AF488 labelled NUP98-KDM5A W1625A (0, 50 or 200 nM) with AF647 H3K4me3 nucleosomal arrays (0, 12.5 or 100 nM nucleosomes). Micrographs are presented as maximum intensity projections (scale bar = 10 µm; n = 3).

Negative control for the local signal analysis.

A. Intensity around the FISH-labeled loci (HOXA, HOXB, HOXD, TNF, MUC4, and MYC) on the DAPI channel across a range of NUP98-KDM5A-FLAG expression levels in HEK293FT cells. B. Boxplots showing the DAPI channel intensity surrounding FISH-labeled loci (HOXA, HOXB, HOXD, TNF, MUC4, and MYC) across binned NUP98-KDM5A-FLAG expression levels in HEK293FT cells.

Selection of window size for H3K4me3 quantification.

Window size selection for the quantification of gene-specific H3K4me3 intensity.