Multivalent Histone and DNA Engagement by a PHD/BRD/PWWP Triple Reader Cassette Recruits ZMYND8 to K14ac-Rich Chromatin

Abstract

Elucidation of interactions involving DNA and histone post-translational-modifications (PTMs) is essential for providing insights into complex biological functions. Reader assemblies connected by flexible linkages facilitate avidity and increase affinity; however, little is known about the contribution to the recognition process of multiple PTMs because of rigidity in the absence of conformational flexibility. Here, we resolve the crystal structure of the triple reader module (PHD-BRD-PWWP) of ZMYND8, which forms a stable unit capable of simultaneously recognizing multiple histone PTMs while presenting a charged platform for association with DNA. Single domain disruptions destroy the functional network of interactions initiated by ZMYND8, impairing recruitment to sites of DNA damage. Our data establish a proof of principle that rigidity can be compensated by concomitant DNA and histone PTM interactions, maintaining multivalent engagement of transient chromatin states. Thus, our findings demonstrate an important role for rigid multivalent reader modules in nucleosome binding and chromatin function.

Keywords: DNA damage; chromatin binding; histone marks; masking of chromatin binding; multivalency; plasticity; protein network assembly; structural rigidity.

Graphical abstract

Figure 1

Structure of the ZMYND8 Triple Reader Module (A) ZMYND8 has a modular architecture, including an N-terminal PHD/BRD/PWWP reader cassette and a C-terminal MYND interaction domain. A recombinant construct expressing residues Q83–S406 was structurally characterized. (B) The crystal structure of the ZMYND8 reader modules is shown, with major secondary structural elements highlighted on the ribbon (left) and surface (right) models. The conserved Kac peptide docking residue (N228) of the BRD and the aromatic cage of the PWWP domain (F288, W291, and F307) are highlighted. The domain topology is highlighted on the right, with domains individually colored. (C) Electrostatic distribution of charge displaying the highly charged surface of the protein. Rotation reveals the insertion of key aromatic residues (F360, Y362, F365, and Y369), found on the C-terminal portion of the construct, into the back of the PWWP domain, stabilizing this module. (D) Detail of the C-terminal tail packing onto the surface of the PWWP domain, revealing three large hydrophobic pockets that accommodate the C-terminal peptide. The first pocket (P279, A285, V303, I318, C321, and L323) accommodates Y369 (left). The second pocket (I353, L280, V350, M324, and M346) is capped by R354 and K326 and accommodates F360 and Y362 (right). The third pocket (W282, Y322, and P329) is capped by E327 and accommodates F365 (right).

Figure 2

ZMYND8 Binds to Histone H3 In Vitro (A) Focused peptide SPOT array spanning lysine acetylation/methylation combinations within the first 21 residues of N-terminal histone H3 peptides. The presence of K14ac increases binding intensity (thick white circles). (B) Focused peptide SPOT array spanning residues 22–42 of all histone H3.x variants, carrying different methyl states of K36, in the absence or presence of pT32. (C) BLI of the recombinant triple reader ensemble profiled against 20-aa-long N-terminal histone H3 peptides carrying single PTMs as indicated in the inset (left). Unmodified H3 peptides bind to the triple reader modules (shown in gray). K9 modifications are tolerated, whereas K4 methylations break the interaction, and K14 acetylation greatly enhances it. Removal of the N-terminal sequence from peptides results in loss of binding irrespective of additional modifications (K9ac, K18ac, K23ac, and K27ac) present together with the central K14ac mark (right), suggesting that the N-terminal portion of H3, which engages the PHD domain, is essential for binding to ZMYND8. Experiments were carried out twice (n = 2) for each condition/peptide. (D) In-solution evaluation of histone H3.1 binding by ITC. Raw injection heats for titrations of modified peptides (carrying specific modifications as indicated in the inset) into a solution of ZMYND8 are shown. The inset shows the normalized binding enthalpies corrected for the heat of peptide dilution as a function of binding site saturation (symbols as indicated in the figure). Solid lines represent a nonlinear least-squares fit using a single-site binding model. Histone H3.1 lacking any modifications binds weakly (red triangle, K_D = 19.0 μM), whereas the addition of the phosphomimetic T32E has little effect (blue pentagon, K_D = 20.7 μM). Binding to K14ac is also weak (dark green circles, K_D = 15.1 μM), and the co-existence of the phosphomimetic T32E shows little effect (orange square, K_D = 17.5 μM). ITC titrations were carried out in triplicate (n = 3), and representative curves are shown. (E) In solution evaluation of histone H3.3. Data are presented as in (D). Histone H3.3 peptides lacking any modifications bind weakly (red triangle, K_D = 19.6 μM), whereas the addition of the phosphomimetic T32E has little effect on binding (blue pentagon, K_D = 17.4 μM). Binding is mainly driven by K14ac (dark and light green circles, K_D = 6.8 μM or 6.6 μM 40-mer and 20-mer, respectively), and the co-existence of the phosphomimetic T32E shows little effect (orange square, K_D = 6.2 μM). (F) In-solution evaluation of ZMYND8 carrying specific domain mutations to histone H3.3 binding by ITC. Data are presented as in (D). Mutation of the conserved asparagine (N228) on the bromodomain abolishes the interaction, as do mutations of conserved PHD residues (N87A, E104A, and D124A). Combination of PHD/BRD mutations has the same effect. Mutation of the conserved residues forming the PWWP cage (F288A and W291A) result in loss of affinity (WT ZMYND8, 6.8 μM; PWWP mutant, 23 μM).

Figure 3

ZMYND8 Binds to Histone H3 in Cells through Its Reader Domains (A) Surface representation of the electrostatic potential distribution onto the triple reader ZMYND8 ensemble (scale as indicated in the inset), highlighting potential docking sites of histone peptides spanning the PHD, BRD, and PWWP domains (see also Figures S7A and S7B). Dotted circles highlight the cavities where histone residues would insert. Details shown on the right highlight the residues implicated in peptide binding as well as critical residues (highlighted in blue) that were further mutated to probe ZMYND8-histone interactions. (B) HEK293 cells transiently transfected with 3×FLAG-ZMYND8 WT or mutants (BRD mutant, N228F; PWWP mutant, F288A/W291A; PHD+BRD+PWWP mutant, N87A/E104A/D124A/N228F and F288A/W291A) were used to interrogate binding to histone H3. The WT protein recognizes all marks tested, whereas mutations on the interacting interface of the three modules result in gradual loss of binding. Mutated residues are annotated in (A). (C) Western blot quantification demonstrating the effect of ZMYND8 mutations on recognition of histone H3 marks from (B). Relative intensity is as annotated in the inset. (D) Quantification of the effect of BRD mutation (left), PWWP mutations (center), or PHD/BRD/PWWP mutations (right) on histone H3 marks. Although BRD mutations primarily affect recognition of acetylation at H3K9/K14ac and PWWP mutations affect H3K36me_x states, simultaneous mutation of all three reader modules results in more than 85% loss of histone H3 binding for all marks tested. Data represent mean ± SEM from three biological replicates of the WBs shown in (B). (E) HEK293 cells transiently transfected with 3×FLAG-ZMYND8 WT or BRD mutant (N228F) were used to interrogate binding to histone H4 marks. Acetylation of H4 can be recognized by the protein, but mutation of the BRD module results in total loss of the interaction.

Figure 4

ZMYND8 Binds to DNA and Nucleosomes (A) Detail of the PWWP domains of human ZMYND11 (PDB: 4NS5) and ZMYND8, showing positively charged areas (highlighted with orange dotted circles). Structures are oriented so that the bottom of the PWWP domain is visible, with residue electrostatic potential plotted on the surface of each protein, as indicated in the inset. (B) Detail of the boxed regions shown in (A) in ribbon representation, as indicated in the inset, highlighting the structural implications of the C-terminal portion of the hZMYND8 structure. The loop region connecting β8 and α4 in the PWWP domain (hZMYND11, orange; hZMYND8, pale yellow) adopts a helical turn in hZMYND8 as a result of the C-terminus packing under the PWWP domain via aromatic residues (Y362, F365, and Y369 shown in blue), resulting in a small extension of helix α4. Repositioning of the loop between β8 and α4 results in K338 (in hZMYND8) overlaying with K334 (in hZMYND11), and the loop between β3 and β4 (K287/K289 in hZMYND8 and K284/K286 in hZMYND11) suggests a role in maintaining contacts with DNA. Rotation of the superimposed structures by 90° (right) reveals a different orientation of the positive residues in each structure, suggesting different topologies when interacting with nucleosomes. (C) DNA binding interfaces identified on the PHD/BRD modules (R96, K233, K239, and K243) and PWWP module (K284, K286, and K334) mapped onto the charged surface of the reader ensemble (scale as indicated in the inset). (D) The recombinant WT reader ensemble can shift a radioactive AT-rich (top) or GC-rich (bottom) DNA probe. Alanine mutations (guided by the model shown in A) introduced on the PHD/BRD (RKKK/AAAA) or the PWWP (KKK/AAA) interface abolish DNA interactions in EMSAs. (E) Full-length ZMYND8 associates with chromatin, whereas mutations of its readers or its DNA-binding face result in loss of binding to chromatin. Full-length 3×FLAG ZMYND8 WT or mutants (as indicated in the inset) were transiently transfected into HEK293 cells, and the whole-cell extract and chromatin-associated fraction were analyzed for 3×FLAG-ZMYND8 by western blot.

Figure 5

ZMYND8 Co-localizes with H3K14ac at Transcriptional Start Sites and Enhancers (A) Average profile of ZMYND8, FLAG-ZMYND8, and H3K14ac ChIP-seq signals on ± 3 kb around transcriptional start sites in HEK293 cells. Representative distributions from biological replicates (n = 2) are shown. (B) Heatmaps of ChIP-seq signals on ± 3 kb surrounding the TSS of genes bound to ZMYND8 or FLAG-ZMYND8 and H3K14ac in HEK293 cells. (C) Distribution of tag densities for ZMYND8 and K14ac around enhancer (top) and promoter (bottom) regions showing significant overlap according to p values calculated relative to local background. (D) ChIP-seq analysis of ZMYND8 and H3K14ac in HEK293 cells for selected genes. Signal enrichment is compared with the published enrichment signals for H3K14ac (GSM521883) and H3K27ac (GSM521887) in IMR90 cells as well as ZMYND8 in ZR-75-30 cells (GSE71323). (E) FRAP experiments following masking of H3K14ac in U2OS cells by transfection with mCherry-3xBAZ2B (WT or N/F mutant) constructs. Competition for H3K14ac by BAZ2B^WT results in faster recovery of ZMYND8, whereas the inactive BAZ2B^N/F mutant has no effect on ZMYND8 recovery, demonstrating direct competition for K14ac in cells. Data represent the mean ± SEM (n = 15). Right: quantitative comparison of time with half-maximal fluorescence recovery for GFP-ZMYND8. ^∗∗∗p < 0.005. (F) ChIP-qPCR validation of FLAG-ZMYND8 over K14ac-bound loci in HEK293 cells stably expressing 3×FLAG-ZMYND8^WT following competition with HA-3xBAZ2B^WT or HA-3xBAZ2B^NF. Masking of H3K14ac by HA-3xBAZ2B^WT resulted in significant reduction of ZMYND8 occupancy at these loci. Data represent mean ± SD from biological triplicates (n = 3). ^∗∗p < 0.01, ^∗p < 0.05. ns, not significant.

Figure 6

ZMYND8 Reader Interactions Control Recruitment to DNA Damage Sites (A and B) U2OS cells transfected with GFP-tagged ZMYND8 WT or mutants were photobleached (small area annotated with a white circle) in FRAP experiments (A). The WT protein is rapidly recruited to the bleached site, whereas mutations of the reader modules result in gradual loss of the protein’s ability to recruit onto the damaged site (B). (C) Detail of the PHD/BRD (left) and PWWP (right) interfaces highlighting residues that can potentially bind to DNA. (D and E) U2OS cells transfected with GFP-tagged ZMYND8 WT or DNA-face mutant (involving simultaneous mutation of the seven residues highlighted in C) were photobleached in FRAP experiments (D). The WT protein is rapidly recruited to the bleached site, whereas mutation of DNA-interacting residues results in significant loss of the proteins ability to recruit onto the damaged site (E). The data in (B) and (E) represent the mean ± SEM from multiple experiments in different cells (n = 15).

Figure 7

ZMYND8 Reader Interactions Control the Assembly of Transcriptional Complexes (A) AP-MS of full-length 3×FLAG WT or BRD^N/F ZMYND8 from HEK293 cells highlighting the changes in the ZMYND8 interactome following disruption of binding to acetylated histones by mutating the conserved BRD asparagine (N228) to a phenylalanine. The data represent the mean of biological replicate experiments (n = 2). Spectral counts of each prey proteins are displayed as a gray scale in the legend. (B) Proximity biotinylation coupled to mass spectrometry of full-length BirA^∗-FLAG WT (top) or BRD mutant (bottom) ZMYND8 in HEK293 cells. Mutation of the BRD module results in partial loss of enrichment of many components found to interact by AP/MS as well as gain of interaction with several new proteins. The data represent the mean of biological replicate experiments (n = 2). (C) Human ZMYND8 mutations and expression in cancer. Mutations annotated by the TCGA consortium found on the triple reader modules of ZMYND8 are annotated on the structure, colored according to the domain to which they belong, and classified by cancer type. (D) Domain organization of proteins that contain multiple reader domains in serial arrangements, potentially adopting multivalent assemblies that can engage distinct histone or nucleosome states.