Molecular Coupling of Histone Crotonylation and Active Transcription by AF9 YEATS Domain

Abstract

Recognition of histone covalent modifications by chromatin-binding protein modules ("readers") constitutes a major mechanism for epigenetic regulation, typified by bromodomains that bind acetyllysine. Non-acetyl histone lysine acylations (e.g., crotonylation, butyrylation, propionylation) have been recently identified, but readers that prefer these acylations have not been characterized. Here we report that the AF9 YEATS domain displays selectively higher binding affinity for crotonyllysine over acetyllysine. Structural studies revealed an extended aromatic sandwiching cage with crotonyl specificity arising from π-aromatic and hydrophobic interactions between crotonyl and aromatic rings. These features are conserved among the YEATS, but not the bromodomains. Using a cell-based model, we showed that AF9 co-localizes with crotonylated histone H3 and positively regulates gene expression in a YEATS domain-dependent manner. Our studies define the evolutionarily conserved YEATS domain as a family of crotonyllysine readers and specifically demonstrate that the YEATS domain of AF9 directly links histone crotonylation to active transcription.

Figure 1. Identification of AF9 YEATS Domain as a Histone Crotonyllysine Reader

(A) Cut-away view of AF9 YEATS in complex with H3K9ac peptide (Li et al., 2014). AF9 YEATS is represented as green surface and histone peptide is shown in yellow. Pink arrow denotes the wide opening of Kac-reader pocket. (B) Chemical structures of known histone lysine acylations (abbreviations and numbers correspond to data shown in (C) and are used throughout text with emphasis on acetyl- versus crotonyl-lysine). (C) Isothermal titration calorimetry (ITC) fitting curves of wild type AF9 YEATS titrated by a series of H3_1–15 peptides containing K9 acylations shown in (B). (D) Representative peptide array screening assay is shown in triplicate using wild type AF9 YEATS against a subset of acetylated versus crotonylated H3 peptides (for full peptide array see Figure S1). (E) ITC fitting curves comparing Kcr and Kac binding preference at sites H3K18 (left) and H3K27 (right). (F) ITC titration and fitting curves of H3K9ac or H3K9cr bound by YEATS domains from human ENL and yeast Yaf9, Taf14. See also Figure S1.

Figure 2. Molecular Basis for H3K9cr and H3K18cr Readout by AF9 YEATS domain

(A) Crystal structure of AF9 YEATS domain shows the insertion of the H3K9 containing Kcr (left) and Kac (right) into its aromatic sandwich cage. K9ac and K9cr are shown as space-filling spheres with the two additional hydrocarbon atoms of the crotonyl group shown in green. (B) Left, detailed interaction map of the K9cr readout by AF9 YEATS shows a rotation of Y78 and F28 when compared to K9ac (pink arrows). The K9cr-bound structure (green, with key residues highlighted in light pink) is superimposed with K9ac-bound structure (white). Hydrogen bonds are shown as dashes. K9cr peptide is covered by 2Fo-Fc omit map (blue meshes) contoured at 2.5 σ. A sharpening B-factor of 75 was applied to obtain the most informative maps. Right, π-aromatic stacking between planar H3K9cr and F59 phenyl ring. Distances are denoted by magenta dashes. (C) Overall structure of H3K9cr bound to AF9 YEATS. AF9 YEATS is shown as green ribbons with key residues highlighted in pink. Histone H3 peptide is shown in yellow. (D) Superimposition of H3K9cr- and H3K18cr-bound complexes. H3K9cr complex is colored white as a reference and H3K18cr peptide is shown in yellow. Structure-based sequence alignment between H3K9cr and H3K18cr is shown below. (E) Mutagenesis and ITC using mutant and wild type AF9 YEATS with H3K9cr (left) or H3K18cr (right) peptides. (F) Sequence conservation analysis of YEATS reader pocket residues from yeast to human. See also Figure S2.

Figure 3. YEATS Domain Prefers Kcr to Kac Marked Nucleosomes

(A) Graphical illustration of AF9-FLAG MNase immuno-precipitation experiment done under native conditions. Subunits of the Super Elongation Complex (SEC) are shown with a FLAG-tagged wild type (WT) or mutant (F59A) AF9 subunits. (B) AF9 co-immunoprecipitates nucleosomes marked by H3K9cr and H3K18cr in a YEATS-dependent manner while the interaction with the SEC components is YEATS-independent. (C) Graphical illustration of nucleosome pulldown assay using synthetic Kcr versus Kac nucleosomes. (D) YEATS domain proteins have a greater affinity for nucleosomes marked with Kcr than Kac. The indicated pre-modified nucleosomes generated by amber suppression were used in pull down assays. Immunoblot analysis with anti-AF9 and anti-ENL antibodies is shown. Direct Blue staining of the membrane demonstrates comparable input material. The streptavidin monomer co-migrates with histone H4 at this resolution. See also Figure S3.

Figure 4. Bromodomains Lack Crotonyllysine Preference

(A) Summary of Kcr- and Kac-binding affinities by selected bromodomains from seven phylogenetic families (Filippakopoulos et al., 2012). Cognate modification sites are listed in the last column of the table. (B) Comparison of chemical structures of lysine acetylation (Kac), propionylation (Kpr), butyrylation (Kbu) and crotonylation (Kcr). The planar part due to π-conjugation is shown in purple. (C) ITC fitting curves comparing Kac, Kpr, Kbu and Kcr binding affinities of H4K8 readout by the first bromodomain of BRD4 (BRD4_BrD1), H3K14 readout by BAZ2A bromodomain, H3K18 readout by the second bromodomain of BRD3 (BRD3_BrD2) and AF9 YEATS. (D) Histone H3K18ac binding by BRD3_BrD2. BRD3_BrD2 is shown as ribbon. H3K18ac peptide is shown in yellow. Water molecules are shown in cyan. (E) Recognition of H3 “R17-K18ac” by BRD3_BrD2. BRD3_BrD2 is shown as surface view. Note the spatial restraints around the K18 acetylamide group caused by the side-open pocket. (F) Steric clash between F334 of BRD3_BrD2 and a modelled K18cr. The experimental K18ac group is shown in yellow and overlaid for reference. The extended hydrocarbon group of crotonylation is colored green. Red disk indicates steric clash. (G) Recognition of H3 “R17-K18cr” by AF9 YEATS. Note the position of crotonylamide group in the extended and end-open pocket. See also Figure S4.

Figure 5. AF9 is Recruited to LPS Stimulated Genes and Correlates with H3K18cr

(A) AF9 is recruited to LPS stimulated genes. Average profile of FLAG-AF9 ChIP-seq data from unstimulated or LPS-stimulated RAW264.7 cells plotted ± 5kb TSS of LPS-stimulated genes (log₂(fold)>2; 890 genes). (B) H3K18cr and AF9 co-localize at genes marked by AF9. Average profile of FLAG-AF9 and H3K18cr ChIP-seq data from LPS-stimulated cells plotted ± 5kb TSS of genes occupied by AF9 (4735 genes). (C) Venn diagram showing the overlap of AF9 occupied (purple) and H3K18cr marked (blue) genes. (D) Genome-browser view of FLAG-AF9 (purple) and H3K18cr (blue) at three representative AF9-occupied genes. (E-F) FLAG-AF9 levels correlate with levels of H3K18cr. (E) Average profile of FLAG-AF9 ChIP-seq data plotted over quintiles ranked by H3K18cr RPKM ± 1kb TSS. Q1 represents the group with highest H3K18cr RPKM and Q5 the lowest (1696 genes per quintile). (F) Using the same groups, FLAG-AF9 RPKM for each gene ± 1kb within a group are plotted as a box and whisker plot (10–90^th percentile) for each group. ****: t-test derived p-value < 0.0001. (G) H3K18cr levels correlate with levels of FLAG-AF9. Similar analysis as in (F), except H3K18cr RPKM for each gene ± 1kb within quintile groups ranked by FLAG-AF9 RPKM ± 1kb TSS are plotted as a box and whisker plot (10–90^th percentile) for each quintile (947 genes per quintile). ****: t-test derived p-value < 0.0001. See also Figure S5.

Figure 6. AF9 Co-Localizes with H3K18cr to Positively Regulate Gene Expression in a YEATS-Dependent Manner

(A) AF9 is further recruited to genes transcriptionally responsive to crotonate pre-treatment. Average profile of FLAG-AF9 ChIP-seq data from RAW264.7 cells unstimulated (green), LPS-stimulated (orange), or pre-treated with 10mM crotonate and then LPS-stimulated (purple) ± 5kb TSS of genes either responsive to crotonate (log₂(fold-FPKM)>1.0; 52 genes) (left), or genes unresponsive to crotonate (log₂(fold-FPKM)<0.01; 233 genes) (right). (B) Box and whisker plot (10–90^th percentile) of fold change due to crotonate pre-treatment in FLAG-AF9 RPKM ± 1kb TSS of crotonate responsive and crotonate unresponsive genes. ****: t-test derived p-value < 0.0001. (C) Genome-browser view of FLAG-AF9 ChIP-seq (purple) and RNA-seq (black) data from cells unstimulated, LPS-stimulated, or crotonate pre-treated and LPS-stimulated for a representative responsive gene (Rsad2) and representative unresponsive gene (Ccl3). (D) The increased recruitment of AF9 due to crotonate pre-treatment is YEATS-Kcr-dependent. FLAG-ChIP from RAW264.7 cells expressing a wild type AF9 (FLAG-AF9(WT)) construct or an AF9 construct with the YEATS-Kcr abrogating F59A mutation (FLAG-AF9(F59A)) under the conditions indicated followed by qPCR analysis of ChIP product and appropriate inputs. Data are plotted as mean percent input + standard deviation. ***: t-test derived p-value = 0.0001; ****: <0.0001; ns >0.05. (E) The transcriptional response to increased crotonyl-CoA is dependent on AF9 and the YEATS-Kcr interaction. The fold changes in mRNA abundance for Rsad2, Il6, Ifit1, and Cmpk2 due to crotonate pre-treatment prior to LPS stimulation are compared across four RAW264.7 cell lines: 1) wild type control (no guide), 2) CRISPR-Cas9 mediated AF9 genetic knockout (AF9 KO), 3) AF9 KO expressing a wild type construct of FLAG-AF9 (AF9KO + WT), and 4) AF9 KO expressing a F59A mutant construct of FLAG-AF9 (AF9KO + F59A). Data are plotted as mean fold change, as measured by qRT-PCR, of two biological replicates + standard deviation. ****: t-test derived p-value < 0.0001; ns: >0.05. See also Figure S6.