Structural insights into the recognition of histone H3Q5 serotonylation by WDR5.

Serotonylation of histone H3Q5 (H3Q5ser) is a recently identified posttranslational modification of histones that acts as a permissive marker for gene activation in synergy with H3K4me3 during neuronal cell differentiation. However, any proteins that specifically recognize H3Q5ser remain unknown. Here, we found that WDR5 interacts with the N-terminal tail of histone H3 and functions as a "reader" for H3Q5ser. Crystal structures of WDR5 in complex with H3Q5ser and H3K4me3Q5ser peptides revealed that the serotonyl group is accommodated in a shallow surface pocket of WDR5. Experiments in neuroblastoma cells demonstrate that H3K4me3 modification is hampered upon disruption of WDR5-H3Q5ser interaction. WDR5 colocalizes with H3Q5ser in the promoter regions of cancer-promoting genes in neuroblastoma cells, where it promotes gene transcription to induce cell proliferation. Thus, beyond revealing a previously unknown mechanism through which WDR5 reads H3Q5ser to activate transcription, our study suggests that this WDR5-H3Q5ser-mediated epigenetic regulation apparently promotes tumorigenesis.

INTRODUCTION

Nucleosomes comprise 147 base pairs of DNA, wrapped around core histone octamers; these are the fundamental subunits of the chromatin that carries both genetic and epigenetic information to control developmentally and environmentally appropriate gene transcription and genomic regulation (). Posttranslational modifications on histones (hPTMs) are now understood to exert major regulatory impacts on diverse biological processes. A variety of hPTMs such as methylation, acetylation, phosphorylation, and ubiquitination have been identified to be epigenetic marks, driving gene regulatory networks in controlling cell fate decisions. hPTMs are dynamically regulated by “writer” and “eraser” enzymes and are recognized by dozens of “reader” modules, and these proteins variously coordinate to specify many distinct biological outcomes (–). Owing to developments in mass spectrometry technologies, previously unidentified histone modifications such as crotonylation, β-hydroxybutyrylation, lactylation, and serotonylation have been reported to activate gene transcription, and these modifications have been functionally implicated in physiological responses to metabolic stress and neuronal cell differentiation, among many others (–).

Serotonin is a neurotransmitter that activates intracellular signaling pathways through cell surface receptors (, ). In addition, serotonin can covalently modify and thereby regulate the functions of target proteins, including fibronectin, small guanosine triphosphatase, and Rac1, upon transamidation catalyzed by tissue transglutaminases (TGMs) (–). Although histone proteins are known to be suitable substrates for TGMs in vitro, it has been unclear for more than a decade whether histones are serotonylated by TGMs in vivo (). A recent study showed that TGM2 can catalyze serotonylation of the Q5 position of histone H3; the resultant serotonylation of histone H3Q5 (H3Q5ser) is frequently present alongside trimethylation of histone H3K4 (H3K4me3). Of particular note, H3 histones doubly modified with trimethylated K4 (K4me3)/serotonylated Q5 (Q5ser) marks (H3K4me3Q5ser) were implication in the transcription activation of genes that function in neuronal cell differentiation and in the development of the central nervous system ().

Although H3Q5ser has been proposed as permissive modifications thought to generally promote gene transcription, no reader proteins that specifically recognize H3Q5ser to mediate gene activation have been confirmed experimentally. Nevertheless, 77 proteins were shown to associate with histone H3 in the presence of Q5ser and K4me3 dual modifications, including WD repeat-containing protein 5 (WDR5) (). WDR5 functions as the core subunit of the mixed lineage leukemia (MLL) complex, which catalyzes the methylation of H3K4 upon interaction with the N-terminal tail of histone H3 (). Previous studies have implicated WDR5-mediated transcriptional activity in tumor growth, proliferation, differentiation (), and metastasis (). In addition, there are reports that WDR5 binds the promoters of neuronal genes to regulate the transcription of genes affecting cell differentiation (). Given its previously shown interaction with H3Q5ser, and considering the reported neuronal-development impacts of both the WDR5 protein and H3Q5ser, it is possible that WDR5 may be a reader protein for H3Q5ser that deciphers epigenetic signals to activate neurodevelopment-related genetic programs.

In this study, we identified WDR5 as a reader of histone H3Q5ser and demonstrated that serotonylation of Q5 significantly increases the binding affinity of WDR5 for histone H3. We elucidated the atomic mechanism through which WDR5 specifically recognizes H3Q5ser on the basis of two crystal structures of WDR5 in complex with H3Q5ser or H3K4me3Q5ser peptide. Extending these structural insights, we showed that disrupting WDR5’s capacity to read H3Q5ser decreases H3K4me3 level in neuroblastoma cells and restricts cell proliferation. Thus, our study demonstrates how WDR5 reads H3Q5ser and shows how this recognition and subsequent occupancy at selected loci activates target gene transcription. Last, the fact that multiple genes targeted for transcriptional activation based on this WDR5/H3Q5ser mechanism are known cancer-promoting genes may help explain the progression of neuroblastoma and could support the development of innovative antitumor therapeutics.

RESULTS

Identification of WDR5 as a reader of H3Q5ser

H3Q5ser (Fig. 1A) is catalyzed by TGM2 and is now known to mediate permissive gene expression during neuronal development. WDR5, a well-known H3 binding protein, was previously suggested as a possible reader protein of H3Q5ser because it was pulled-down in an assay using a peptide bearing an H3K4me3Q5ser modification as the bait (). Therefore, we conducted streptavidin pull-down assays with either purified recombinant full-length human WDR5 (named WDR5FL) or truncated version (containing residues 22 to 334; named WDR522–334) using synthetic biotinylated histone H3 peptides as interaction ligands; these peptides comprised the N-terminal 14 amino acids of H3 and carried either no modification or modifications including K4me3 and/or Q5ser (table S1). We found that both WDR522–334 and WDR5FL interact with four types of H3 peptides (H3Q5ser, H3K4me3Q5ser, H3K4me3, and unmodified H3 peptides) (Fig. 1B, left). Notably, serotonylation on Q5 enhances the binding ability of WDR5 to the H3 peptide by ~2-fold, regardless of the presence or absence of K4me3 (Fig. 1B, right).

Fig. 1

Serotonylation of H3Q5 facilitates WDR5 binding to H3.

(A) Chemical structure of serotonyl group covalently attached to the Q5 residue of histone H3. The histone octamer core and the wrapped DNA of the nucleosome are respectively shown as a solid yellow cycle and a blue line. The N-terminal 1 to 14 residues of H3 are indicated by their single-letter codes, with Q5 being colored red and the others colored black. Trimethylation on the K4 residue is indicated by “me3.” The chemical structure of the serotonyl group on Q5 is shown in red. (B) Streptavidin pull-down assay of recombinant WDR522–334 (upper left) or full length WDR5FL (middle left), with various biotinylated H31–14 variants, with or without K4 trimethylation or Q5 serotonylation as indicated. The relative intensities of WDR522–334 and WDR5FL bands pulled down by various peptides are quantitated by densitometry and expressed as the fold of the protein level pulled down by H3 peptide. The numbers below the bands are the average of three independent experiments, and the means ± SD are also presented as the bar plots (the right). *P < 0.05; **P < 0.01. FL, full length. (C) ITC analysis to measure the binding affinity of WDR522–334 with an unmodified H3 (gray) ligand peptide or H3 variant peptides bearing modifications including H3K4me3 (blue), H3Q5ser (red), or dual H3K4me3Q5ser (green).

We further conducted isothermal titration calorimetry (ITC) assays to quantify the binding affinity of WDR522–334 to the H3 peptides. WDR522–334 binds to H3Q5ser peptide with a dissociation constant (KD) of 4.02 ± 0.39 μM, which is approximately two times lower than that of the unmodified H3 (KD = 8.71 ± 0.94 μM, Fig. 1C, Table 1, and table S2). WDR522–334 shows similar affinities with both H3Q5ser and H3K4me3Q5ser peptides, showing that the trimethylation of K4 has no obvious noticeable impacts on the interaction of Q5ser with WDR5. Together, the data from ITC and pull-down assays show that WDR5-H3 binding is enhanced by the presence of a serotonylation modification to the H3 peptide’s Q5 residue.

Table 1

Binding affinity of wild-type or mutant WDR522–334 with various H3 peptides determined by ITC.

H31–14 peptide	KD values (μM)
	WDR522–334
	WT	N130A	Y131A
H3	8.71 ± 0.94	11.80 ± 1.40	22.80 ± 2.22
H3Q5ser	4.02 ± 0.39	8.77 ± 0.51	14.60 ± 1.08
H3K4me3	8.22 ± 0.54	11.20 ± 1.66	19.40 ± 2.06
H3K4me3Q5ser	4.33 ± 0.30	9.48 ± 1.59	14.80 ± 1.41

Overall structure of WDR522–334 in complex with serotonylated H3 peptides

To further investigate the underlying molecular mechanism through which H3Q5ser interacts with WDR5, the crystal structures of WDR522–334 in complex with H3Q5ser or H3K4me3Q5ser peptide were solved, both at 1.6 Å resolution (Table 2). In these two structures, the overall structures of WDR522–334 are almost identical, with a root mean square deviation value of 0.1 Å and both exhibiting the typical seven-bladed β-propeller structure of WD40 repeat proteins (Fig. 2A). On the basis of the electron density maps, the N-terminal residues 1 to 7 of these two peptides can be clearly traced, whereas residues 8 to 14 are disordered and not visible. Interpretable electron density was observed for the serotonyl group attached to Q5 in both structures, which was modeled unambiguously (Fig. 2, B and C). No electron density was observed for the trimethyl group of K4 in the structure of the WDR5/H3K4me3Q5ser complex, likely because the side chain of K4me3 is flexible and protrudes toward the solvent.

Table 2

Data collection and refinement statistics.

RMS, root mean square.

	WDR5/H3Q5ser	WDR5/H3K4me3Q5ser
PDB code	7CFP	7CFQ
Data collection
Space group	C2221	C2221
Cell dimensions
a, b, c (Å)	78.75, 99.41, 80.62	78.36, 98.38, 80.34
α, β, γ (°)	90, 90, 90	90, 90, 90
Resolution (Å)	50.00–1.60 (1.63–1.60)*	50.00–1.60 (1.63–1.60)
Rsym or Rmerge (%)†	7.90 (61.20)	7.90 (27.50)
I/σI	26.81 (3.33)	26.64 (8.44)
Completeness (%)	100.00 (100.00)	100.00 (100.00)
Redundancy	12.50 (12.10)	13.10 (12.80)
Refinement
No. reflections	41999	41159
Rwork‡ / Rfree§ (%)	17.43/19.94	16.36/18.98
RMS deviations
Bond lengths (Å)	0.01	0.01
Bond angles (°)	1.51	1.42
No. of protein/ligandmolecules
Protein	1	1
Peptide	1	1
Glycerol		1
Ethylene glycol		1
No. of water	179	179
B-factors (Å2)
Protein	22.25	13.71
Peptide	33.58	24.75
Glycerol		20.03
Ethylene glycol		23.79
Water	27.12	21.12
Ramachandran plotǁ
Most favored regions (%)	94.08	95.05
Allowed regions (%)	5.92	4.95
Generously allowedregions (%)	0	0

*The values in parentheses refer to statistics in the highest shell.

†Rsym = |Ii − < I > |/|Ii| where Ii is the intensity of the ith measurement, and < I > is the mean intensity for that reflection.

‡Rwork = |F|/F

§Rfree was calculated with 5.1% of the reflections in the test set.

ǁStatistics for the Ramachandran plot from an analysis using MolProbity.

Fig. 2

Crystal structures of WDR522–334 in complex with either H3Q5ser or H3K4me3Q5ser peptide.

(A) Overall structures of WDR522–334 in complex with either H3Q5ser or H3K4me3Q5ser peptide. The structures of WDR522–334 are shown in a cartoon diagram; WDR522–334/H3Q5ser is colored in wheat and WDR522–334/H3K4me3Q5ser is colored in gray. The H3Q5ser (green) and H3K4me3Q5ser (yellow) peptides are shown in a cartoon diagram, with the serotonyl group on Q5 shown as sticks. (B and C) Omit Fo – Fc electron density map of the H3Q5ser (B) and H3K4me3Q5ser (C) peptides, contoured at the 2.5σ level. The peptides are shown as sticks and colored green and yellow [following color scheme from (A)]. Trimethyl and serotonyl groups are indicated with red labels. (D) Structural comparison of complexes including WDR522–334/H3 (cyan; PDB code: 2H9M), WDR523–334/H3K4me3 (orange; PDB code: 2H6Q), WDR522–334/H3Q5ser (green), and WDR522–334/H3K4me3Q5ser (yellow). Enlarged view in the right panel shows binding-induced conformational changes in the peptides. (E) Electrostatic potential surface view of WDR5 in complex with the H3Q5ser peptide. The peptide is shown as sticks. (F) The interaction of H3Q5ser (green) and H3K4me3Q5ser (yellow) peptides with WDR522–334. Amino acid residues of WDR5 involved in the peptide interaction are shown as sticks; these are colored wheat in WDR522–334/H3Q5ser and colored gray in WDR522–334/H3K4me3Q5ser. The enlarged views in the left and right panels show the detailed interactions between the H3 peptides and WDR5. Hydrogen bonds are indicated as dashed lines.

Alignment of the WDR522–334/H3Q5ser and WDR522–334/H3K4me3Q5ser complex structures with those of the WDR5/H3 [Protein Data Bank (PDB) code: 2H9M] complex and the WDR5/H3K4me3 complex (PDB code: 2H6Q) revealed that the overall structure of WDR5 in these structures is almost identical (Fig. 2D). The orientations of the first four residues 1 to 4 of histone H3 peptide in these structures are also similar, whereas residues 5 to 7 undergo large conformational changes. Specifically, the side chain of Q5ser is flipped to the opposite side, thus positioning the serotonyl group in a shallow hydrophobic surface pocket (Fig. 2, D and E).

In both the WDR522–334/H3Q5ser and WDR522–334/H3K4me3Q5ser complex structures, the histone H3 N termini are accommodated in the binding cleft of WDR5 via a series of hydrogen bonds and van der Waals contacts, findings similar to previously reported structures of WDR5 in complex with methylated histone H3K4 peptides (–). More specifically, R2 is anchored in a narrow central channel through cation-π interactions with F133 and F263. The side chains of both K4 and K4me3 are solvent exposed and do not engage in any direct interactions with WDR5, helping to explain our findings from the ITC and pull-down assays, which indicated that the enhanced interaction of WDR5 with serotonylated histone H3 is independent of the methylation status of K4 (Figs. 1C and 2, E and F).

Specific recognition between WDR5 and serotonylated histone H3Q5

According to the complex structures, WDR5 recognizes the serotonyl group through a network of hydrogen bonds and van der Waals contacts (Fig. 3, A and B). In details, the amide group of the WDR5 N130 side chain forms a hydrogen bond with the hydroxyl group of serotonyl group attached to Q5. In addition, the WDR5 D172 side chain engages in water-mediated hydrogen bonding with the amide group of serotonin. The aromatic side chains of WDR5’s Y131, F149, and Y191 residues make van der Waals contacts with the hydrophobic moiety of the serotonyl group that help stabilize the WDR5-Q5ser interaction.

Fig. 3

Recognition of serotonyl group by WDR5.

(A and B) Serotonyl group of H3Q5ser (A) and H3K4me3Q5ser (B) peptides bound to WDR522–334. The hydrogen bonds are shown as dashed lines. Amino acid residues involved in serotonyl group binding are shown as sticks and are labeled in black. (C) Streptavidin pull-down assays of WT and five putative binding-deficient mutant variants of WDR522–334 with the aforementioned biotinylated H3 peptides as ligands, as indicated. The relative intensities of WT or mutant variants of WDR5 pulled down by each peptide are quantitated by densitometry and expressed as the fold of the WT WDR5 level pulled down by the same type of peptide. The numbers below the bands are the average of four independent experiments. (D and E) The binding of WDR522–334 N130A (D) and WDR522–334 Y131A (E) to various H3 peptides was evaluated via ITC. Peptides were titrated into sample cell containing WDR522–334 N130A (D) or WDR522–334 Y131A (E) proteins.

To examine impacts of WDR5 residues involved in specific recognition of the serotonyl group, we generated a total of five WDR5 mutants to examine their interaction with histone H3 peptides. Circular dichroism (CD) spectroscopy and size-exclusion chromatography supported that all of the putative binding-deficient WDR5 mutant variants are soluble and properly folded (fig. S1, A to C). Furthermore, pull-down assays showed that the F149A, D172A, and Y191A WDR5 variants had obviously reduced interactions with all four types of H3 peptides compared to wild-type (WT) WDR5 (i.e., regardless of the presence or absence of Q5ser or K4me3). In contrast, the N130A and Y131A WDR5 variants had significantly reduced binding ability for the Q5ser-containing ligands but much less obvious difference for the unmodified H3 or H3K4me3 ligands. Collectively, these findings suggest that the functions of the N130 and Y131 residues are involved in the recognition of serotonylation modifications of histone H3Q5, whereas the F149, D172, and Y191 residues apparently function in interactions with other histone H3 regions (Fig. 3C and fig. S1D).

We next quantified the binding affinity of WT WDR5 and WDR5 variants carrying the N130A or Y131A mutations for H3 peptide ligands by ITC experiments (Fig. 3, D and E, Table 1, and table S2). The dissociation constants of mutant N130A (KD = 11.80 ± 1.40 μM) is ~1.4-fold higher than that of WT WDR5 (KD = 8.71 ± 0.94 μM) with unmodified H3, whereas the KD value of N130A (8.77 ± 0.51 μM) with H3Q5ser peptide was 2.2-fold higher than that of WT WDR5 (4.02 ± 0.39 μM). Compared to WT WDR5, the Y131A variant had a 2.6-fold increase in KD value (22.80 ± 2.22 μM) for unmodified H3 but had a 3.6-fold higher KD value (14.60 ± 1.08 μM) for the H3Q5ser peptide ligand. These results confirm that both N130 and Y131 are involved in the recognition of serotonylation of H3Q5.

Given reports that H3K4me3 frequently occurs alongside H3Q5ser modification, we next attempted to identify residues that may contribute to specific recognition of H3Q5ser modification rather than recognition of H3K4me3Q5ser dual modifications. The ITC data indicated that the binding affinity of the N130A variant (KD = 9.48 ± 1.59 μM) for the H3K4me3Q5ser peptide is ~2-fold lower than that of WT WDR5 (KD = 4.33 ± 0.30 μM), whereas there was only slight difference in their respective binding with the H3K4me3 ligand (KD = 11.20 ± 1.66 μM), confirming the involvement of WDR5’s N130 residue in specific recognition of serotonylation at Q5 site. Similarly, the KD value of the Y131A variant (14.80 ± 1.41 μM) for H3K4me3Q5ser was increased by 3.4-fold compared to WT WDR5 (KD = 4.33 ± 0.30 μM). In addition, the binding affinity of Y131A variant (KD = 19.40 ± 2.06 μM) for H3K4me3 peptide was decreased by about 2.4-fold when compared to WT WDR5 (KD = 8.22 ± 0.54 μM) (Fig. 3, D and E, Table 1, and table S2). These observations suggest that WDR5’s Y131 residue is also required for the recognition of Q5ser modification via hydrophobic interaction. Together, these results identified N130 and Y131 as particularly impactful residue for the specific recognition of serotonyl group at H3Q5 site by WDR5, regardless of the presence or absence of K4me3.

Impacts of WDR5-H3Q5ser recognition on H3K4me3 in neuroblastoma cells

To confirm the existence of H3Q5ser in neuroblastoma cells, lysates of the WT neuroblastoma (SK-N-SH) cells were either directly loaded onto SDS–polyacrylamide gel electrophoresis (PAGE) gel or first immunoprecipitated with an anti-histone H3 antibody and were then immunoblotted with an anti-H3K4me3Q5ser antibody. Serotonylation modification at the Q5 residue was detected in cell lysates; moreover, it was detected in the anti-histone H3 antibody precipitates but not in control immunoglobulin G (IgG) precipitates (Fig. 4A), establishing that the Q5 site of H3 histones carry serotonylation modifications in neuroblastoma cells.

Fig. 4

Interaction of WDR5 with H3Q5ser affects H3K4me3.

(A) Immunoprecipitation-western blot (IP-WB) analysis to detect the existence of H3Q5ser (top) in human neuroblastoma SK-N-SH cells, both in directly loaded cell lysates and in anti-histone H3 antibody enriched precipitates. Histone H3 (bottom) was visualized using an anti-histone H3 antibody (as a control). (B) 3× FLAG-tagged WDR5 (WT or with N130A mutation) were complemented in WDR5 knockout (WDR5−/−) SK-N-SH cells. Colocalization of WT WDR5 or N130A mutant (green, anti-FLAG tag antibody) with H3Q5ser (red, anti-H3K4me3Q5ser antibody) in the SK-N-SH cell nuclei was detected by immunofluorescence microscopy. (C) Disruption of WDR5-H3Q5ser interaction significantly decreased the level of H3K4me3 modification. Cellular levels of WDR5, H3K4me3, and H3Q5ser in various cell lines were visualized by Western blotting after staining with anti-WDR5, anti-H3K4me3, and anti-H3Q5ser antibodies, respectively. The relative intensity of each band is quantitated by densitometry after normalization to β-actin and then expressed as the fold of that in the control cells. *P < 0.05; ***P < 0.001; ****P < 0.0001; n.s., not significant.

To further explore the association of WDR5 with H3Q5ser in cells, a WDR5 knockout (WDR5−/−) SK-N-SH cell line was generated using CRISPR-Cas9 (fig. S2, A and B), and 3× FLAG-tagged WDR5, either WT WDR5 or N130A WDR5 variant, was complemented to WDR5−/− cells (named WDR5−/− + WDR5WT and WDR5−/− + WDR5N130A, respectively). WDR5 and H3Q5ser were monitored in both cell lines by confocal microscopy after staining with anti-FLAG tag and anti-H3K4me3Q5ser antibodies. Both WDR5 (green) and H3Q5ser (red) were exclusively localized to neuroblastoma cell nuclei, with the signals of WT WDR5 and H3Q5ser being extensively colocalized (Fig. 4B, top). In contrast, much reduced colocalization was evident in WDR5−/− + WDR5N130A cells expressing the N130A variant of WDR5, which tended to localize at the peripheral nuclei (Fig. 4B, lower section). Consistently, much less WDR5 N130A protein was coprecipitated with anti-histone H3 antibody despite comparable cellular expression levels (fig. S2, B and C), suggesting a significantly decreased capacity of WDR5 N130A to interact with H3Q5ser.

Given WDR5’s role in mediating MLL1 complex recruitment to promote H3K4 methylation (, ), we sought to explore the impact of WDR5-H3Q5ser interaction on H3K4me3. As expected, WDR5 knockout in neuroblastoma cells significantly decreased the level of H3K4me3 modification (Fig. 4C, middle). When WT WDR5 was complemented to WDR5−/− cells (WDR5−/− + WDR5WT), H3K4me3 returned to normal level. However, when WDR5 N130A variant was complemented to WDR5−/− cells (WDR5−/− + WDR5N130A), H3K4me3 level was still significantly lower than control. Such decrease of H3K4me3 in WDR5−/− + WDR5N130A cells may be caused by the disruption of WDR5-H3Q5ser interaction, which led to less WDR5 and subsequently less MLL1 complex recruitment to gene promoters. Therefore, our data suggest that H3K4me3 can be affected by WDR5-H3Q5ser interaction. Similar results have been observed when TGM2, the known catalytic enzyme for H3Q5ser, was knocked down (fig. S2, F to I). In contrast, although H3K4me3 in WDR5−/− + WDR5N130A cells was significantly decreased, the H3Q5ser level is comparable to that in the control cells or WDR5−/− + WDR5WT cells, suggesting that serotonylation of H3Q5 is likely independent of H3K4 methylation (Fig. 4C, right side).

WDR5 regulates proliferation of neuroblastoma cells by recognition of H3Q5ser

To explore potential pathological impacts of the WDR5-H3Q5ser interaction in neuroblastoma, proliferation rates of WDR5−/−, WDR5−/− + WDR5WT, and WDR5−/− + WDR5N130A cells were detected by bromodeoxyuridine (BrdU) assay. As expected, WDR5 knockout (WDR5−/−) caused significantly decreased neuroblastoma cell proliferation compared to unedited cells (Fig. 5A), consistent with the previously reported clinical findings that high levels of WDR5 are correlated with poor prognosis of neuroblastoma patients (). The decreased cell proliferation phenotype was completely rescued upon complementation of the WDR5−/− cells with WT WDR5. Moreover, we found that complementation with the N130A variant of WDR5 could only partially rescue the neuroblastoma cell proliferation. Thus, our results demonstrate that WDR5 functions to promote the proliferation of neuroblastoma cells and indicate that this proliferation-promoting impact of WDR5 is exacerbated by the presence of serotonylation modifications at the Q5 site of H3 histones.

Fig. 5

WDR5 regulates neuroblastoma cell proliferation by recognizing H3Q5ser.

(A) WDR5−/− significantly decreased the proliferation of SK-N-SH cells assessed by BrdU assays. The decreased cell proliferation phenotype was completely rescued upon complementation of the WDR5−/− cells with WT but not N130A mutant WDR5. (B) mRNA expression levels of three cell proliferation-related genes in WDR5−/− cells or WDR5−/−cells complemented with WT or N130A mutant WDR5, as assessed by qPCR. (C and D) ChIP was performed with an anti-histone H3K4me3 or anti-histone H3K4me3Q5ser antibody in SK-N-SH cells (C) or with an anti-FLAG antibody [(D) for WDR5] in WDR5−/− cells complemented with 3× FLAG-tagged WT WDR5 or N130A variant. qPCR analysis of the ChIP precipitates was performed to assess the co-occurrence of H3K4me3, H3K4me3Q5ser, and WDR5 at the promoters of PDCD6, GPX1, and C-MYC genes shown to regulate tumor cell proliferation. The data are presented as the means ± SD calculated from three independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

WDR5 has been reported to bind to various promoter regions known to function in tumor maintenance (), including genes that regulate tumor cell proliferation such as PDCD6, GPX1, and C-MYC (–). In neuroblastoma cells, we also observed notable decrease of these three genes in WDR5−/− cells, which can be rescued by WDR5WT complementation, but not WDR5N130A (Fig. 5B), indicating the importance of WDR5-H3Q5ser interaction in promoting transcription of these genes. We therefore determined whether there is enrichment for WDR5 occupancy and/or H3Q5ser modifications in the regulatory regions of these loci in neuroblastoma cells. Chromatin immunoprecipitation sequencing (ChIP-seq) was performed with an anti-H3K4me3 antibody, an anti-H3K4me3Q5ser antibody, and an anti-WDR5 antibody, respectively, in SK-N-SH cells (fig. S3A). Co-occupancy of H3K4me3, H3K4me3Q5ser, and WDR5 has been observed in the promoter regions of the three genes, with similar peak pattern but higher enrichment of H3K4me3Q5ser than H3K4me3 (fig. S3B).

To confirm this finding, ChIP–quantitative polymerase chain reaction (qPCR) for the three genes were performed first using anti-H3K4me3 or H3K4me3Q5ser antibody in SK-N-SH cells. Significantly higher amount of DNA was precipitated with anti-H3K4me3Q5ser antibody than that of H3K4me3 (Fig. 5C), suggesting more K4me3Q5ser dual modified histone H3 enriched at the promoter regions of these genes. ChIP-qPCRs of the three genes were then performed in WDR5−/− + WDR5WT and WDR5−/− + WDR5N130A cells with anti-FLAG antibody (for WDR5), and significant less amount of DNA was detected in WDR5−/− + WDR5N130A cells (Fig. 5D). Together, our data revealed the co-occurrence of WDR5 occupancy and H3Q5ser modifications at the promoters of PDCD6, GPX1, and C-MYC. Notably, we observed significantly reduced occupancy of the N130A variant of WDR5 at the promoters of these three loci compared to WT WDR5, with results establishing that H3Q5ser recognition promotes the recruitment of WDR5 to promoter of cancer-promoting genes and thus activates gene transcription to facilitate cell proliferation.

DISCUSSION

Serotonylation of H3Q5 is the first endogenous monoaminyl modification of histones that has been identified (). However, the reader proteins for H3Q5ser and its biological functions remain largely unknown. In the present study, we found that WDR5 functions as an H3Q5ser reader. We show that WDR5 preferentially binds serotonylated H3 over unmodified or H3K4me3 peptides, and both biochemical and structural studies, including solving two crystal structures for WDR522–334/H3Q5ser and WDR522–334/H3K4me3Q5ser complexes, revealed the residues and atomic interactions through which WDR5 recognizes H3Q5ser. Specifically, WDR5’s N130 residue mediates binding with the serotonyl group by contributing a hydrogen bond interaction.

In both structures of WDR522–334/H3Q5ser and WDR522–334/H3K4me3Q5ser, WDR5 binds to the peptides through a series of hydrogen bonds and van der Waals contacts with R2 engaged into a negatively charged central channel. In addition, the serotonyl group attached to the side chain of Q5 is accommodated in a hydrophobic pocket on the surface of WDR5 (Figs. 2, E and F, and 3, A and B). Both the central channel and hydrophobic pocket are the key structural determinants for WDR5 to specifically recognize H3Q5ser. Several other WD40 repeat proteins, such as EED, RBBP4, RBBP5, WDR77, and RACK1, were reported to associate with chromatin (–). Comparison of the structures of EED, RBBP4, RBBP5, WDR77, and RACK1 with WDR5 indicates a lack of either negatively charged central channel or hydrophobic surface pocket in these protein structures (fig. S4), suggesting that H3Q5ser is unlikely to interact with these proteins. Although WDR5 shares similar overall structural feature with other WD40 repeat proteins, the structural determinants of the binding site ascertain the specific recognition between WDR5 and H3Q5ser.

WDR5 is considered to be an oncogene because its overexpression is related to poor prognosis in various human cancers such as neuroblastoma, colon cancer, prostate cancer, gastric cancer, breast cancer, lung cancer, and bladder cancer (, –). Among these cancers, neuroblastoma is of particular interest to pediatric clinicians because it is the most common pediatric extracranial solid tumors, and neuroblastoma is the primary cause of death from pediatric cancers in children between 1 and 5 years old and leads to approximately 13% of pediatric cancer mortality (). The underlying oncogenic mechanisms of WDR5 have to date been largely attributed to its interaction with MLL1 and its subsequent catalysis of trimethylation at H3K4 sites (). In addition to MLL1, WDR5 can be recruited to N-MYC target gene promoters through its physical interaction with N-MYC, doing so in a manner that can promote H3K4 trimethylation in neuroblastoma; this WDR5 recruitment results in elevated transcription of N-MYC target genes to promote tumor formation ().

A clinically relevant aspect of our study is our demonstration from experiments in neuroblastoma cells that WDR5 undergoes a specific interaction with histone H3Q5ser to promote its localization at the promoter regions of PDCD6, GPX1, and C-MYC, genes known to promote the proliferation of tumor cells including neuroblastoma (–), and high level of which is correlated with poor prognosis of neuroblastoma patients (unpublished data). We show that disrupting the interaction between WDR5 and H3Q5ser by N130A mutation trapped WDR5 protein at the peripheral nuclei (Fig. 4B) but not the nuclear membrane. Such prevention of WDR5’s nucleus accumulation may lead to the decrease of the occupancy of WDR5 at target gene promoters, thus reduced the expression levels of target genes and decreased the extent of neuroblastoma cell proliferation. Together, these findings suggest that specific recognition of H3Q5ser by WDR5 should be understood as an activating mechanism for the proliferation of neuroblastoma cells.

Histone H3Q5ser frequent coexists alongside H3K4me3 on euchromatin, and these dual modifications are known to support efficient recruitment of the TFIID complex and to be associated with “permissive gene expression” (). Our work here with neuroblastoma cells also shows that recognition of H3Q5ser facilitates the recruitment of WDR5 to the chromatin (Figs. 4B and 5D). These findings support that H3Q5ser should be understood as an epigenetic mark that is correlated with transcriptional activation. Although serotonylation on histone H3Q5 has no effect on the activity of MLL1 complex to catalyze H3K4me3 in vitro (, ), the molecular mechanism underlying the cross-talk between these two types of modification inside the cells remains elusive. In the present study, take the advantage of N130A mutation that preferentially affects WDR5-H3Q5ser interaction, we have shown that disruption of WDR5-H3Q5ser interaction significantly decreased the H3K4me3 level, probably due to less MLL1 complex recruitment to gene promoters. In contrast, H3Q5ser modification is not significantly affected when H3K4me3 modification is decreased, as evident by the comparable H3Q5ser level in WDR5−/− + WDR5N130A cells and control cells (Fig. 4C). Decrease of H3Q5ser in WDR5−/− cells is largely due to the significantly lowered TGM2 level (fig. S2, D and E).

Besides WDR5, there are several reader proteins for histone H3K4me3 (e.g., TAF3, BPTF, and Spindlin-1) that have been found to interact with histone H3K4me3Q5ser peptides (, ). Previous studies have shown that WDR5, TAF3, BPTF, and Spindlin-1 are involved in activating gene transcription through their interactions with histone H3K4me3 (, –). It should be highly informative to determine whether these proteins can recognize H3Q5ser, whether they impact gene transcription, and to dissect whether their regulatory functions are specifically affected by the simultaneous presence of H3Q5ser and H3K4me3 marks at target loci. Moreover, our study reveals insights about the progression of neuroblastoma and suggests the opportunity to target WDR5 and/or H3Q5ser-related regulatory networks for developing innovative antitumor therapies.

MATERIALS AND METHODS

Reagents and cell cultures

The human neuroblastoma cell line SK-N-SH was obtained from the Cell Bank of Chinese Academy of Sciences. The SK-N-SH cells were cultured in 1:1 Eagle’s minimum essential medium (American Type Culture Collection) and Ham’s F12 medium (Thermo Fisher Scientific) supplemented with 15% fetal bovine serum and incubated at 37°C in a humidified incubator with 5% CO2.

The antibodies used in current study included anti-WDR5 (Abcam, ab178410), anti-Histone H3K4me3 (Abcam, ab213224), anti-Histone H3Q5ser (Millipore, ABE1791), anti-Histone H3K4me3Q5ser (Millipore, ABE2580), anti-FLAG tag (Affinity, T0003), anti-Histone H3 (ImmunoWay, YM3038), anti-Lamin A+C (Abcam, ab108595), anti-TGM2 (Santa Cruz Biotechnology, sc-48387), anti–β-actin (Sigma-Aldrich, A3854), horseradish peroxidase (HRP)–conjugated anti-rabbit IgG (CST, 7074S), and HRP-conjugated anti-mouse IgG (CST, 7076S).

Protein expression and purification

The full-length human WDR5 (WDR5FL) and WDR5 fragments 22 to 334 (WDR522–334) were cloned into the pET-28a vector (that contains a TEV (Tobacco etch virus protease) cleavable N-terminal 8× His tag) and was overexpressed in Escherichia coli Rosetta2 strain in LB medium. Cells were induced with 0.4 mM isopropyl-β-d-thiogalactopyranoside and grown at 16°C for an additional 20 hours for protein expression. Cells were harvested and resuspended in lysis buffer containing 50 mM tris-HCl (pH 7.5), 500 mM NaCl, 5% (v/v) glycerol, 5 mM imidazole, 1 mM phenylmethylsulfonyl fluoride, 2 mM β-mercaptoethanol, and deoxyribonuclease I (20 μg/ml) (shyuanye, 9003-98-9) and lysed with a high-pressure homogenizer. The cell lysate was centrifuged and the supernatant was incubated with Ni-NTA (nitrilotriacetic acid) affinity resin (GE Healthcare). The target protein was eluted with lysis buffer supplemented with 300 mM imidazole. The 8× His tag was removed by TEV protease and tag-free WDR522–334 was further purified by gel filtration using a Superdex 75 (10/300) increase column (GE Healthcare) in 10 mM Hepes (pH 7.4), 150 mM NaCl, 1 mM EDTA, and 0.5 mM TCEP [tris(2-carboxyethyl)phosphine]. The eluted WDR522–334 protein was collected and concentrated for further use. All WDR522–334 mutants were generated by PCR-based site-directed mutagenesis and were purified using the same procedure as the WDR522–334 protein.

Streptavidin pull-down assay

Streptavidin pull-down assays were performed to verify the binding ability of recombinant WDR522–334 or WDR5FL with serotonylated H3 peptides. Protein and peptide concentration were determined based on the absorbance at 280 and 205 nm, respectively.

Synthetic biotinylated H3 peptides (6.6 μM) were incubated with 15 μl of streptavidin-agarose resin in 0.6-ml binding buffer (gel filtration buffer containing 0.2% NP-40) at 4°C for 40 min. The beads were washed three times with binding buffer, and 0.86 μM purified tag-free WDR5FL or WDR522–334 proteins, WT or mutant, were added into the peptide-bound resin and incubated at 4°C for another 50 min. The beads were washed four times with binding buffer and boiled at 100°C. The bound proteins were analyzed by SDS-PAGE followed by Coomassie brilliant blue staining, and the band intensity was quantified by ImageJ.

ITC measurements

ITC experiments were performed using MicroCal PEAQ-ITC (Malvern) at 20°C. All peptides and proteins were solved in gel filtration buffer. For each ITC titration, peptides at a concentration of 0.9 to 1.0 mM were titrated into sample cell containing 80 to 85 μM WT or mutant WDR522–334 proteins by 19 individual injections with 1 μl for the first and 2 μl for the rest. The data were analyzed by PEAQ-ITC analysis software using one set of sites fitting model.

Circular dichroism

Far-ultraviolet CD spectrum signals were detected with an Applied Photophysics Chirascan spectrometer. All measurements were carried out at 298 K in buffer containing 10 mM tris-HCl (pH 7.4) and 50 mM NaCl at wavelengths ranging from 190 to 260 nm. All samples were recorded for three times, and each final spectra curve was the average of three scans.

Crystallization, x-ray data collection, and structure determination

Purified WDR522–334 was mixed with synthetic H3Q5ser or H3K4me3Q5ser peptide at a molar ratio of 1:5 for 2 hours on ice before crystallization. The crystals of the WDR522–334/H3Q5ser complex were grown in 0.1 M sodium citrate tribasic dihydrate (pH 5.5), 22% polyethylene glycol (PEG) 3350, and 0.1% n-Octyl-β-d-glucoside at 22 °C. Crystals of the WDR522–334/ H3K4me3Q5ser complex were obtained in 0.15 M ammonium sulfate, 0.1 M MES (pH 5.5), 25% PEG 4000, and 2% PEG 3350 at 22 °C.

The x-ray diffraction datasets were collected at the National Facility for Protein Science in Shanghai at beamline BL18U1. All data were processed, integrated, and scaled using HKL2000 ().

The crystal structure of WDR5 (PDB code: 2GNQ) was used as the searching model for molecular replacement by Phaser () in Phenix (). Structure refinement, manual model building, and structural analysis were carried out using REFMAC5 () and WinCoot (). Data collection and structure refinement statistics are summarized in Table 2. Figures were generated using PyMOL (www.pymol.org).

Lentivirus-mediated WDR5 knockout and complementation

To generate WDR5 knockout SK-N-SH cells by CRISPR-Cas9–mediated gene editing, two independent guide sequences targeting WDR5 gene were designed (E-CRISP, www.e-crisp.org/E-CRISP/) and cloned into the LentiCRISPR-v2-BSD plasmid, which was derived from the lentiCRISPR v2 plasmid (Addgene plasmid no. 52961) by replacing the puromycin gene with the blasticidin gene. The single guide RNA (sgRNA) sequences targeting WDR5 gene are as follows: sgRNA1: CTGGGACTACAGCAAGGGGA and sgRNA2: CAGAAACTACAAGGCCACAC. The guide RNA–containing lentiviral plasmid was then cotransfected with two packing plasmids psPAX2 and pMD2.G into human embryonic kidney 293 T cells to generate lentivirus. SK-N-SH cells were transduced with the lentivirus for 48 hours and were subjected to blasticidin selection at 3 μg/ml for 14 days to obtain stable WDR5 knockout (WDR5−/−) cell lines.

To complement WDR5 in WDR5−/− cells, full-length complementary DNA (cDNA) encoding WDR5 was amplified from the human brain cDNA library and cloned into a pLL3.7-puro–tagged cloning vector encoding a C-terminal 3× FLAG tag. Mutagenesis of WDR5 N130A was carried out using a Q5 Site-Directed Mutagenesis Kit (New England Biolabs, E0552S). Lentivirus packaging and infection were carried out as described above.

Lentivirus-mediated TGM2 knockdown

The short hairpin RNAs (shRNAs) targeting TGM2 and negative control were synthesized by Cyagen (China) and the sequences are as follows: shRNA1: TATCACCCACACCTACAAATA and shRNA2: GGGCTGAAGATCAGCACTAAG. Lentivirus packaging and infection were carried out as described above, and the cells were subjected to puromycin (Invitrogen) selection at 3 μg/ml for 7 days to obtain stable TGM2 knockdown (TGM2 KD) cell line.

Cell proliferation assay

Cell proliferation was assayed using a BrdU Cell Proliferation Assay Kit (CST, no. 6813) following the manufacturer’s instructions. Briefly, cells were seeded at a density of 4 × 103 cells per well in a 96-well plate. BrdU labeling solution (10 μM) was substituted and incubated for 4 hours; cells were then fixed for 1 hour and incubated with anti-BrdU for 1 hour. After removal of the unbound BrdU, the cells were incubated with an HRP-labeled secondary antibody for 30 min. TMB substrate solution was then added, and the absorbance at 450 nm was measured using a BioTek Synergy™2 microplate reader (BioTek).

IP and Western blotting

SK-N-SH cells (6 × 106), TGM2 knockdown cells, WDR5−/− + WDR5WT, or WDR5−/− + WDR5N130A cells were lysed with RIPA lysis buffer [50 mM tris-HCl (pH 7.4) 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 1% sodium deoxycholate, and 0.1% SDS] supplemented with 1% protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific) at 4°C for 30 min. The cell lysates were immunoprecipitated with anti-histone H3 antibody or control IgG at 2 μg/ml and protein G agarose beads at 4°C for 1 hour. Proteins bound to the agarose beads were separated on a 15% SDS-PAGE gel; transferred to a polyvinylidene difluoride membrane; immunoblotted with an anti-H3K4me3Q5ser, anti-WDR5, or anti-histone H3 antibody; and lastly visualized using an enhanced chemiluminescence substrate kit (Millipore) with an ImageQuant LAS 4000 mini densitometer (GE Healthcare Life Science).

Immunofluorescence staining and confocal microscopy imaging

WDR5−/− + WDR5WT and WDR5−/− + WDR5N130A cells were seeded onto 12-mm slides (Thermo Fisher Scientific) in a 24-well plate 1 day before immunofluorescence staining. The cells were fixed with 4% paraformaldehyde Fix Solution (PFA) in phosphate-buffered saline (PBS) at room temperature for 10 min and then permeabilized with 0.5% Triton X-100 in PBS for 10 min. After blockage with 5% bovine serum albumin (BSA) and 5% goat serum in PBS at room temperature for 30 min, the cells were incubated with anti-histone H3K4me3Q5ser (1:200 diluted in PBS supplemented with 1% BSA) or anti-FLAG tag (1:1000 diluted in PBS/BSA) antibody for 1 hour, followed by incubation with a fluorescently labeled secondary antibody (1:500 diluted in PBS/BSA) for 30 min. After 4′,6-diamidino-2-phenylindole (DAPI) staining, cells were dehydrated with ethanol and mounted with ProLong Diamond Antifade Mountant (Invitrogen). Images were acquired using a Leica SP8 confocal microscope.

Quantitative PCR

Total RNA of various cell samples was extracted using TRIzol (Thermo Fisher Scientific). Hieff qPCR SYBR Green Master Mix (Yeasen) was used for messenger RNA (mRNA) quantification following the manufacturer’s instructions. Briefly, RNA reverse transcription was performed with 5× PrimeScript RT Master Mix (Takara). qPCR was performed using a CFX Connect Real-Time System (Bio-Rad) under the following cycling condition: 95°C for 2 min, 40 amplification cycles of 95°C for 5 s and 60°C for 30 s, followed by a final cycle of 95°C for 5 s and 65°C for 5 s. Relative expression levels of target genes were calculated using the 2−ΔΔCT method, and the levels of these genes in control SK-N-SH cells were set at 1 for normalization. qPCR for each gene was performed in technical triplicates in three independent experiments. Sequences of the primer used are listed in table S3.

ChIP-seq and ChIP-qPCR

ChIP assays were performed following the protocol of the Myers Lab at the Stanford University (www.hudsonalpha.org/myers-lab/protocols/). Briefly, 2 × 107 cells in culture dishes were fixed with 1% formaldehyde at room temperature for 10 min, and the fixation was stopped via addition of 0.125 M glycine at room temperature for 5 min. The cells were washed with PBS and scraped into cold lysis buffer [10 mM tris-HCl (pH 7.4), 10 mM NaCl, 3 mM MgCl2, and 0.5% NP-40]. Chromatin pellets obtained by centrifugation were resuspended with 0.8 ml of radioimmunoprecipitation assay (RIPA) buffer [50 mM tris-HCl (pH 7.5), 300 mM NaCl, 1.0% Triton X-100, 0.5% sodium deoxycholate, and 0.1% SDS) and sonicated at 25% amplitude for a total of 8 min (with intervals). Sonicated chromatin was then incubated with 2 μg of anti-histone H3K4me3Q5ser, anti-histone H3K4me3, anti-FLAG tag antibody, or control IgG precoated on Dynabeads protein G (Life Technologies) at 4°C overnight. The spike-in antibody and spike-in chromatin (active motif) were added as internal control. Immunoprecipitants were washed twice with RIPA buffer and additional three times with LiCl washing buffer [100 mM tris-HCl (pH 7.5), 500 mM LiCl, 1% NP-40, and 1% sodium deoxycholate]. Chromatin precipitated was then eluted by protease K digestion, cross-link reversed at 65°C overnight, and purified using a QIAquick PCR purification kit (Qiagen). The ChIPed DNA fragments were either sent for high-throughput sequencing (Novogene) or assessed qPCR as described above; primer information is listed in table S4.

For ChIP-seq data (table S5) analysis, reads from ChIP-seq were aligned to the hg19 reference genome using Burrows-Wheeler Alignment tool v0.7.17-r1188 (). After filtering ambiguously-mapped and duplicates reads, MACS2 v2.2.6 () was used to call the peaks with the following parameters, -f BAMPE -g hs -B. Results from the ChIP-seq data were visualized in Integrative Genomics Viewer v2.8.13 ().

Statistical analysis

The data from at least three independent experiments were expressed as the means ± SD. All data were analyzed with the Shapiro-Wilk test for normal distribution before testing for differences between groups. For within group comparison, paired t tests were used for normally distributed data, whereas Wilcoxon rank sum tests were used for non-normally distributed data. All graphs were generated using GraphPad Prism 6.0 (GraphPad Software Inc., USA). P < 0.05 was considered statistically significant.