ZMYND8 mediated liquid condensates spatiotemporally decommission the latent super-enhancers during macrophage polarization

Abstract

Super-enhancers (SEs) govern macrophage polarization and function. However, the mechanism underlying the signal-dependent latent SEs remodeling in macrophages remains largely undefined. Here we show that the epigenetic reader ZMYND8 forms liquid compartments with NF-κB/p65 to silence latent SEs and restrict macrophage-mediated inflammation. Mechanistically, the fusion of ZMYND8 and p65 liquid condensates is reinforced by signal-induced acetylation of p65. Then acetylated p65 guides the ZMYND8 redistribution onto latent SEs de novo generated in polarized macrophages, and consequently, recruit LSD1 to decommission latent SEs. The liquidity characteristic of ZMYND8 is critical for its regulatory effect since mutations coagulating ZMYND8 into solid compartments disable the translocation of ZMYND8 and its suppressive function. Thereby, ZMYND8 serves as a molecular rheostat to switch off latent SEs and control the magnitude of the immune response. Meanwhile, we propose a phase separation model by which the latent SEs are fine-tuned in a spatiotemporal manner.

Fig. 1. ZMYND8 suppresses pro-inflammatory gene expression and alleviates the macrophage-mediated inflammatory response in vivo.

a, b RNA-Seq analysis of WT and Zmynd8 cKO BMDMs stimulated with LPS for 6 h. The heatmap analysis showing the differentially expressed genes with adjusted p value <0.05, false discovery rate <0.05, and log₂ fold-change >1.2 (a). The indicated pro-inflammatory cytokine and chemokine genes are upregulated in Zmynd8 cKO BMDMs (b). Two-tailed Student’s t-test determined p values. c GSEA analysis of RNA-Seq data obtained from WT and Zmynd8 cKO BMDMs treated with LPS. NF-κB pathway gene cluster was enriched in Zmynd8 cKO BMDMs. d Five 6-week-old female WT and Zmynd8 cKO mice were challenged with DSS in drinking water for 7 days, and the body weight loss was measured from day 0. DSS-treated mice were euthanized on day 12. Two-tailed Student’s t-test determined p values. e, f The macroscopic images of colons were shown (e), and colon lengths were measured (n = 5 per group). P values were determined by a two-tailed Student’s t-test (f). g Representative H&E-stained images of proximal colon cross-section of naïve and DSS-challenged WT and Zmynd8 cKO mice at day 12. Bars, 50 µm. h The percentage of the colon-infiltrated immune cells of DSS-challenged WT and Zmynd8 cKO mice (n = 4 per group) was evaluated by flow cytometry on day 6. i ELISA measurement of pro-inflammatory cytokines production in the blood of WT and Zmynd8 cKO mice (n = 5 per group) challenged with DSS for 7 days. j Body weight changes of WT and Zmynd8 cKO mice (n = 10 per group) during HFD feeding. k Depot mass of gonWAT, subWAT, and BAT in WT and Zmynd8 cKO mice (n = 9 per group). l, m GTT (n = 8 per group) (l) and ITT (n = 9 per group) (m) assays were performed in WT and Zmynd8 cKO mice. n, o The absolute numbers (n) and frequencies (o) of gonWAT macrophage in HFD fed WT and Zmynd8 cKO mice (n = 7 per group) based on flow cytometry analysis. All data are means ± SD. Two-tailed Student’s t-test determined p values. All the animal experiments were repeated twice. Source data are provided as a Source Data file.

Fig. 2. LPS-induced ZMYND8 translocation onto the latent SEs regulating pro-inflammatory genes.

a Venn diagram analysis of ChIP-seq peaks of ZMYND8 in non-treated (NT) and LPS-treated BMDMs. LPS-gained, LPS-lost, and constitutively bound ZMYND8 peaks were labeled. b Heatmap analysis of ChIP-seq data obtained from NT and LPS-treated BMDMs. Peak density heatmap showing LPS-gained ZMYND8-binding signals, where the H3K4me1, H3K27ac, and chromatin accessibility (ATAC-seq) signals before and after LPS treatment were also shown. Each row shows ±2 kb centered regions of LPS-gained ZMYND8 peaks. c Based on the RNA-Seq results from WT and Zmynd8 cKO BMDMs, the violin plots summarizing the Zmynd8 deficiency-induced expression change of genes associated with LPS-gained, LPS-lost, and constitutively bound ZMYND8 peaks. P value by one-way ANOVA test. d Ranked plots of enhancers and super-enhancers defined in NT (left) and LPS-treated (right) BMDMs associated with increasing ZMYND8 signals (units: r.p.m.). Enhancers and SEs are defined as regions of ZMYND8 ChIP-seq binding not contained in promoters. The cutoff discriminating TEs from SEs is shown as a dashed line. Genes associated with enhancers that are considered typical or super are colored gray and red, respectively. e Genome Browser tracks showing ZMYND8, H3K4me1, H3K27ac ChIP-seq signals, and chromatin accessibility (ATAC-Seq) in the NT and LPS-treated macrophages at selected SE regions. f Genes associated with LPS-gained ZMYND8 peaks are overlapped with those LPS-induced latent SEs. g Heatmap analysis showing Zmynd8 deficiency-induced expression changes of genes regulated by ZMYND8-enriched latent SEs. h KEGG pathway analysis of genes associated with both LPS-gained ZMYND8 peaks and LPS-gained SEs. Source data are provided as a Source Data file.

Fig. 3. ZMYND8 forms liquid-liquid phase separation in the nucleus.

a Immunofluorescence imaging of endogenous ZMYND8 in BMDMs. ZMYND8 signal (green) is shown merged with DAPI stain. Experiments were repeated three times. b FRAP experiment of GFP-ZMYND8 overexpressed in Raw264.7 cells (left), quantification of FRAP data for GFP-ZMYND8 puncta (right). Bleaching event occurs at t = 20 s. Data are plotted as mean values ± SD (n = 10). c Graphs plotting intrinsic disorder regions (PONDR VSL2B and VL3H) of ZMYND8. PONDR VSL2B and VL3H score (y-axis) and amino acid position (x-axis) are shown. The red bar designates the IDR under investigation. d Schematic of recombinant GFP fusion ZMYND8 fragments tested for LLPS. e ZMYND8-IDR fragment was transfected into HeLa cells and analyzed by FRAP (left), quantification of FRAP data for ZMYND8-IDR fragment puncta (right). Data are plotted as mean values ± SD (n = 10, examined over three independent experiments). f Droplet fusion is highlighted in a higher resolution and extended times frame as indicated. g Representative images of GFP-ZMYND8-IDR droplet formation at different protein concentrations. Data are presented as mean values ± SD (10 µM: n = 107; 20 µM: n = 116; 40 µM: n = 132; examined over three independent experiments). P value by one-way ANOVA test. h 1,6-Hexanediol (HD) was added at the indicated concentrations into solutions containing 40 μM GFP-ZMYND8-IDR. Data are presented as mean values ± SD (0%: n = 169; 5%: n = 160; 10%: n = 166; examined over three independent experiments). P value by one-way ANOVA test. i Representative images of droplet formation at different salt concentrations. Forty micromolar GFP-ZMYND8-IDR was added to the droplet formation buffer. Data are presented as mean values ± SD (62.5 mM: n = 169; 125 mM: n = 160; 250 mM: n = 166; examined over three independent experiments). P value by one-way ANOVA test. j All aspartic acid (D) and glutamic acid (E) residues in IDR were mutated to alanine (A) to generate mutant D/E-to-A (D/E-A). Then full-length (FL) WT (n = 7 samples), ZMYND8 (red line), and mutant (n = 10 samples) (blue line) were transfected into HeLa cells (upper), followed by FRAP quantification assay (lower). Data are presented as mean values ± SD. Cells were examined over three independent experiments. k Representative images of droplet formation of GFP-ZMYND8-IDR (D/E-A mutant) in vitro (upper) and analyzed by FRAP (lower). Red line: WT IDR (n = 10 samples); blue line: D/E-A mutant (n = 9 samples). Data are presented as mean. values ± SD. Source data are provided as a Source Data file.

Fig. 4. The translocation of ZMYND8 compartments onto SEs of pro-inflammatory genes is dependent on its liquidity feature.

a Depiction of Ccl2 SE locus with ZMYND8 ChIP-seq signals (blue histograms) and the location of FISH probes. Colocalization between ZMYND8 protein and the Ccl2 probe in fixed WT BMDMs by IF followed by DNA FISH. Images of the indicated IF and FISH are shown, along with the merged channels (overlapping signal in white). b Left, heatmap analysis showing the total and LPS-gained ZMYND8 ChIP-seq signals from the untreated (−) or 1,6-HD treated (+) BMDMs. Right, averaged profiles of the ZMYND8 ChIP-seq signals. All ChIP-seq signals are displayed from −2 to +2 kb surrounding the center of each annotated ZMYND8 peak. c Genome browser snapshot of ChIP-Seq peaks for ZMYND8 at indicated SE regions without (−) or with (+) 1,6-HD treatment in BMDMs. d Heatmap analysis showing ZMYND8 ChIP signals at the LPS-gained ZMYND8 peaks region in different types of Raw264.7 cells, including Zmynd8 KO rescued with Vector (Zmynd8 KO + Vector), Zmynd8 KO with WT Zmynd8 (Zmynd8 KO + WT Zmynd8), and Zmynd8 KO with D/E-A mutant (Zmynd8 KO + D/E-A mutant), before (NT) and after LPS treatment (LPS). Averaged profiles of the ZMYND8 ChIP-seq signals are shown in the right panel. All ChIP-seq signals are displayed from −2 to +2 kb surrounding the center of each annotated ZMYND8 peak. e Genome browser view of ChIP-seq peaks for ZMYND8 at indicated SE regions in different types of Raw264.7 cells, including Zmynd8 KO + Vector, Zmynd8 KO + WT Zmynd8, and Zmynd8 KO + D/E-A mutant. f Heatmap analysis of RNA-Seq results represent differentially expressed genes regulated by ZMYND8 enriched, latent SEs (from Fig. 2d) in Zmynd8 KO + WT Zmynd8 and Zmynd8 KO + D/E-A mutant Raw264.7 cells. g The latent SEs’ genes upregulated in D/E-to-A cells are overlapped with latent SE genes upregulated in Zmynd8 cKO cells. Source data are provided as a Source Data file.

Fig. 5. p65 guides the redistribution of ZMYND8 liquid condensates onto NF-κB associated SEs.

a 293 T cells were transfected with Myc-p65 and Flag-ZMYND8 for 24 h, followed by Co-IP and western blotting with FLAG antibody. Experiments were repeated three times. b Endogenous Co-IP and western blotting with anti-ZMYND8 antibody in NT and LPS-treated BMDMs. Experiments were repeated three times. c Immunofluorescence staining of endogenous ZMYND8 (green) in primary mouse BMDMs shows colocalization with endogenous p65 (red) after treatment with LPS for 1 h, DAPI nuclear staining in blue. d Quantization of colocalization: the graph represents Pearson’s coefficient of ZMYND8 and p65 from three independent treatments of primary mouse BMDMs. Colocalization of ZMYND8 to p65 shows a significant difference between non-treated (NT) and LPS-treated (LPS) cells. Data are presented as mean values ± SD (n = 17 cells). Two-tailed Student’s t-test determined p values. e Droplet formation of various TFs with GFP-ZMYND8-IDR in droplet formation buffers containing 125 mM NaCl and 10% PEG8000 in vitro. Ten micromolar of GFP-ZMYND8-IDR was added into 10 µM of mCherry, mCherry-p65, mCherry-MYC, and mCherry-p53, respectively. Representative images were shown. f Statistical analysis of the coalescent droplet size when GFP-ZMYND8-IDR mixed with mCherry, mCherry-p65, mCherry-p53, and mCherry-Myc, respectively (n = 60, examined over three independent experiments). P value by one-way ANOVA test. g 293 T cells were transfected with Flag-p65 and EV, GFP-ZMYND8, GFP-ZMYND8 mutant (D/E-A) for 24 h followed by Co-IP and western blot with anti-FLAG antibody. Experiments were repeated three times. h Representative images of separated droplet formation when mCherry-p65 was incubated with GFP-ZMYND8-IDR mutant (D/E-A). i Statistical analysis of the mCherry-p65 droplet size, when WT-ZMYND8-IDR and GFP-ZMYND8-IDR D/E-A mutant was added in droplet formation buffer respectively (n = 60, examined over three independent experiments). Two-tailed Student’s t-test determined p values. j The colocalization of p65 with WT ZMYND8 or D/E-A mutant in Raw264.7 cells was evaluated by IF. Data are presented as mean values ± SD (n = 20 cells, examined over three independent experiments). Two-tailed Student’s t-test determined p values. k Heatmap analysis of the colocalization of p65 ChIP-Seq signals with LPS-gained and LPS-lost ZMYND8 peaks. l Genome browser view of ChIP-seq peaks of ZMYND8 and p65 at indicated SE regions in NT and LPS treated BMDMs. Source data are provided as a Source Data file.

Fig. 6. LPS-induced p65 acetylation at the K122 residue directs the docking of ZMYND8 liquid condensates onto NF-κB associated SEs.

a Co-IP assay between Myc-tagged ZMYND8 and various p65 acetylation site mutants in 293 T cells followed by western blot. Experiments were repeated three times. b The endogenous K122 acetylation level in the cytosol and nucleus of BMDMs was evaluated by western blot at the indicated time after LPS treatment. Experiments were repeated three times. c In vitro droplet formation by incubating GFP-ZMYND8-IDR with mCherry-p65, mCherry-p65-acetylated, mCherry-p65 (K122R), and mCherry-p65 (K122R)-acetylated respectively. Representative images were shown. d Statistical analysis of droplet size formed by GFP-ZMYND8-IDR with mCherry-p65, mCherry-p65-acetylated, mCherry-p65 (K122R), and mCherry-p65 (K122R)-acetylated respectively (n = 50, examined over three independent experiments). Two-tailed Student’s t-test determined p values. e Genome-wide ChIP-Seq analysis of ZMYND8-binding patterns after LPS treatment in different Raw264.7 cells, including p65 KO cell transduced with Vector (p65 KO), p65 KO transduced with WT p65 (p65 KO + WT p65), and p65 KO transduced with K122R mutant (p65 KO + K122R). The average signal intensity of total ZMYND8 peaks (left) and LPS-gained ZMYND8 peaks (right) was shown separately. f Genome browser tracks show that LPS-gained ZMYND8 peaks were recruited to the indicated SE regions in the p65 KO, p65 KO + p65 WT, and p65 KO + K122R Raw264.7 cells after LPS treatment. g Heatmap analysis of RNA-Seq results represent differentially expressed genes regulated by ZMYND8 enriched, latent SEs (from Fig. 2d) in p65 KO + WT p65, p65 KO + K122R Raw264.7 cells after LPS treatment. h The latent SEs regulated genes enhanced in K122R cells are overlapped with latent SE genes upregulated in D/E-to-A cells. Source data are provided as a Source Data file.

Fig. 7. ZMYND8 liquid condensates activate LSD1 to decommission SEs.

a, b Gel filtration chromatography and western blots analyze the indicated protein level in the nuclear fractions purified from control (EV) and Zmynd8 KO Raw264.7 cells. c Endogenous co-IP by using the anti-LSD1 antibody in EV and p65 KO Raw264.7 cells. d Co-IP with anti-Myc tag antibody evaluated the interaction between endogenous LSD1 and WT or mutated ZMYND8 (Myc tagged) in different Raw264.7 cells, including Zmynd8 KO rescued with Vector (Zmynd8 KO + Vector), WT Zmynd8 (Zmynd8 KO + WT Zmynd8), and D/E-A mutant (Zmynd8 KO + D/E-A mutant) Raw264.7 cells. e–g Representative images of droplet formation containing the indicated proteins, such as BFP-LSD1 only (e), GFP-ZMYND8-IDR with BFP-LSD1 (f), GFP-ZMYND8-IDR, mCherry-p65, and BFP-LSD1 (g). All the experiments above were repeated three times. h Genome-wide LSD1-binding patterns in the WT and Zmynd8 cKO BMDMs after LPS treatment were evaluated by ChIP-Seq. Histogram plot showing the average distribution of total (left) and LSD1 signals at the LPS-gained ZMYND8 peaks region (right). All ChIP-seq signals are displayed from −2 to +2 kb surrounding the center of each annotated LSD1 peak. i Genome Browser tracks show the LSD1-binding signals at the selected SE regions in LPS polarized WT and Zmynd8 cKO BMDMs. j Genome-wide H3K4me1 levels in the WT and Zmynd8 cKO BMDMs after LPS treatment were evaluated by ChIP-Seq. Histogram plot showing the average distribution of total H3K4me1 (left) and H3K4me1 signals at the LPS-gained ZMYND8 peaks region (right). All ChIP-seq signals are displayed from −2 to +2 kb surrounding the center of each annotated H3K4me1 peak. k Genome Browser tracks showing the H3K4me1 signals at the selected SE regions in WT and Zmynd8 cKO BMDMs. l Nascent RNA in LPS polarized WT and Zmynd8 cKO BMDMs were collected, and then the enhancer RNA transcription level near the indicated pro-inflammatory genes was measured by qRT-PCR. Source data are provided as a Source Data file.