Recognition of modification status on a histone H3 tail by linked histone reader modules of the epigenetic regulator UHRF1

Structure of the Linked TTD and PHD Histone Reader Modules Bound to the Histone H3 Tail.

The linked histone reader modules of UHRF1 (designated “TTD-PHD” hereafter) consist of the N-terminal TTD and the C-terminal PHD finger separated by a 17-aa intermodule linker (Fig. 1A). In agreement with previous reports (19, 20, 23, 25), our binding assay using isothermal titration calorimetry (ITC) showed that isolated TTD and PHD bound to the N-terminal H3 peptide depending on K9me3 and unmodified R2, respectively (Table 1 and SI Appendix, SI Results). We also observed 1:1 stoichiometric binding of TTD-PHD to the H3 tail bearing unmodified R2 and K9me3 with significantly higher affinity (K_d = 0.37 μM) compared with TTD (K_d = 1.75 μM) or PHD finger (K_d = 1.47 μM) alone (Table 1 and SI Appendix, Fig. S3).

To understand the molecular basis for bivalent recognition of the H3 tail by UHRF1, we determined the 2.9-Å crystal structure of TTD-PHD (residues 134–366, deleting residues 167–175) in a complex with a single N-terminal H3-K9me3 tail (residues 1–12) (SI Appendix, Table S1). The crystal structure of the isolated PHD finger bound to the unmodified H3 tail was also solved at 1.4 Å (SI Appendix, Fig. S5, Table S1, and SI Results). The entire structure of TTD in the complex is essentially identical to that of isolated TTD (23) (Fig. 1B and SI Appendix, Fig. S6B). The structure of the PHD finger in TTD-PHD was nearly identical to that in the PHD:H3 complex (SI Appendix, Figs. S5 and S6A and SI Results). In the TTD-PHD:H3-K9me3 complex, TTD and the PHD finger are adjacent to each other, forming a ring-shaped architecture with a central hole between the modules (Fig. 1B). The association of modules is maintained by extensive contacts between TTD and the intermodule linker (Fig. 1B). The extended linker is packed against a cleft between the two Tudor domains in TTD, except for its three C-terminal residues (299–301), which serves as an intermodule junction between the two reader modules (see Fig. 4A). No apparent interface is observed between the TTD and PHD finger, indicating the importance of the conformation of the linker for maintaining the higher-order structure of TTD-PHD (Fig. 1B). The P(r) function obtained from small-angle X-ray scattering (SAXS) measurements of the TTD-PHD:H3-K9me3 complex showed broader distribution to some extent than that calculated from the crystal structure (SI Appendix, Fig. S4C). Therefore, it is not likely that the domain orientation is static in solution, which is consistent with poor electron density of PHD fingers observed in two of four complexes in the crystal asymmetric unit (SI Appendix, Fig. S12 A–C and SI Results). In addition, the free and H3-bound forms showed overall similar P(r) profiles, suggesting that the ring-shaped architecture of TTD-PHD and its intrinsic dynamic motion are mainly defined by the structure around the intermodular junction rather than by the H3 binding (Fig. 1C and SI Appendix, Fig. S4 A and B and SI Results).

Specific Recognition of the Histone H3 Tail by TTD-PHD.

The PHD finger and TTD make dense contacts with two stretches of histone H3 from A1 to K4 in cassette 1 and from K9me3 to S10 in cassette 2, respectively, at the inner walls of the negatively charged central hole (Fig. 2A). Binding of the H3-K9me3 tail to TTD-PHD excludes about 650 Ų of solvent accessible area on the TTD-PHD surface, which is provided equally by TTD and the PHD finger.

The H3 recognition mode of the isolated PHD finger is well-conserved in the TTD-PHD:H3-K9me3 complex (Fig. 2B and SI Appendix, Fig. S5B) (19, 20). Residues 2–4 of H3 cassette 1 make main-chain hydrogen bonds to form an intermolecular β-sheet with the PHD finger (residues 330–333) (Fig. 2B and SI Appendix, Fig. S7). The N-terminal amino group of the H3 tail donates direct hydrogen bonds to the backbone carbonyl oxygens of E355 and D356 of the PHD finger (Fig. 2B and SI Appendix, Fig. S7 and SI Results). The side chain of H3-R2 fits into the acidic cavity in the PHD finger, in which the guanidino nitrogen atoms are strictly recognized by hydrogen bonds with the UHRF1 residues, C333, D334, and D337 (Fig. 2B and SI Appendix, Fig. S7). Indeed, methylation of H3-R2 decreased the binding affinity of the isolated PHD finger for the unmodified H3 peptide (K_d = 12.70 μM) and also interfered with the interaction of H3 with TTD-PHD regardless of H3-K9 methylation (Table 1). In the isolated TTD:H3-K9me3 complex, the cleft between the Tudor domain accommodates the side chain of H3-K4 (23), whereas in the TTD-PHD:H3K9me3 complex, the same region forms an extended interface with the inter modular linker (Fig. 1B). No significant interaction between the H3-K4 side chain and protein residues is observed in the TTD-PHD:H3K9me3 complex (Fig. 2C). Thus, the H3-binding mode of isolated TTD does not represent that in the context of the linked TTD-PHD.

The C-terminal two residues in cassette 2, H3-K9me3 and H3-S10, form contacts only with the first Tudor domain region. The aliphatic side chain of K9me3 fits into a surface groove, and the trimethyl ammonium group is trapped in an “aromatic cage” motif formed by F152, Y188, and Y191 of the first Tudor domain (Fig. 2C and SI Appendix, Fig. S7), as commonly observed for other methylated lysines in reader modules (7, 26). The hydroxyl group of H3-S10 donates a hydrogen atom to the carboxyl group of D230 of UHRF1, which is not observed in the structure of the TTD:H3 complex (Fig. 2C and SI Appendix, Fig. S6B). No binding of isolated TTD to the unmodified H3 tail was observed in the ITC-binding assay. Nonmethylation of H3-K9 decreased the affinity of TTD-PHD for H3 down to the same order as that of the isolated PHD finger (Table 1). TTD-PHD simultaneously recognizes unmodified R2 in cassette 1 and K9me3 in cassette 2 on a single histone H3 tail.

Induced Helix Formation of the Histone H3 Tail.

In contrast to the cassette regions, residues 5–8 of the H3 tail make no amino acid–specific contacts with TTD-PHD and form an α-helix stabilized by an N-capping hydrogen bond between T3 Oγ and T6 NH (Fig. 3A). The same segment in the isolated PHD:H3 complex adopts an extended conformation (SI Appendix, Fig. S5); thus, the helix formation seems to be induced by the bivalent interaction with TTD-PHD. This conformational change upon binding to TTD-PHD was confirmed by comparison of the ¹H-¹⁵N correlation spectra of the ¹³C, ¹⁵N-labeled H3-K9me3 peptide in its free and bound forms (Fig. 3A). We assigned the chemical shifts of the backbone ¹HN, ¹⁵N, ¹³C_α, and ¹³CO nuclei of the H3-K9me3 peptide in complex with TTD-PHD, except for A1, K4, Q5, and K9me3, and its secondary structure was predicted on the basis of its backbone ¹³C_α and ¹³CO chemical shifts. In agreement with the crystal structure, at least residues 6–8 are suggested to form an α-helix, whereas the C-terminal residues 11–20 are structurally random (Fig. 3 A and B). Such a helical conformation in the H3 tail has not been observed in other histone-binding complexes (7, 26), except that a H3_1–16 tail–like peptide inhibitor bound to lysine specific demethylase 1 (LSD1), which bears Met in place of H3-K4, has been reported to form a similar helical structure (27).

On basis of the structural findings, we reasoned that modification of the H3 tail with negative charge, such as phosphorylation, would have a negative effect on its interaction with TTD-PHD. The ITC data showed that phosphorylation of H3-T3 or H3-S10 significantly repressed binding of TTD-PHD to the H3-K9me3 tail (Table 1). Thus, both the shape and charge complementarities are important for TTD-PHD:H3 interactions (Fig. 2A).

Linker:Tudor Contacts Define the Histone H3–Binding Mode of TTD-PHD.

The structural importance of the TTD:linker contacts was further assessed by SAXS analysis using a linker mutant (TTD-PHD_R295A/R296A), in which residues R295 and R296, whose side chains make contacts with TTD (Fig. 4A), were substituted with alanine. TTD-PHD_R295A/R296A exhibited a P(r) distribution with a shoulder, which is a characteristic of noncompact flexible proteins composed of two domains (Fig. 1C). This P(r) distribution, which deviated from that of the wild type, was observed regardless of H3 binding (Fig. 1C). Combined with the results of NMR experiments (SI Appendix, Fig. S8), these data suggest that the R295A/R296A mutations disrupt the compact ring-shape structure.

TTD-PHD_R295A/R296A bound the H3-K9me3 tail with an affinity ∼10-fold lower than that of the wild type and with an N value of 2.06, which suggests loss of the 1:1 stoichiometric interaction (Fig. 4B and Table 1). Furthermore, in the ¹H-¹⁵N correlation spectrum of the H3-K9me3 peptide bound to TTD-PHD_R295A/R296A, the chemical shifts presumably attributed to helix formation were not observed (Figs. 3B and 4D). Notably, the interaction of the linker mutant with the H3-K9me3 tail exhibited thermodynamic parameters similar to those calculated from titration with an equimolar solution of the isolated TTD and PHD finger (Fig. 4B and Table 1). This is consistent with the NMR spectrum of the H3-K9me3 peptide bound to TTD-PHD_R295A/R296A, which was superimposed on those of the peptide titrated with isolated TTD and PHD finger (Fig. 4D), These data suggest that the R295A/R296A mutations abrogate the simultaneous binding of TTD-PHD to H3-R2 and H3-K9me3, resulting in an independent H3-binding mode of each reader module.

Phosphorylation of S298 at the Linker Region in UHRF1.

S298, which is located at the junction and contacts the indole ring of W238 on the surface of TTD (Fig. 4A), is a target of PKA phosphorylation (28). The corresponding phosphorylation site in homologous proteins is highly conserved among mammals, birds, and reptiles, suggesting its functional importance (SI Appendix, Fig. S9). We prepared TTD-PHD phosphorylated at S298 (TTD-PHD_S298ph) using a bacterial coexpression system with PKA (Fig. 4C and SI Appendix, Fig. S10 and SI Results) and analyzed its solution structure by SAXS. The SAXS data indicated that, unlike TTD-PHD_R295A/R296A, TTD-PHD_S298ph exhibited only small differences from wild-type TTD-PHD in the R_g value and the profile of the P(r) function (Fig. 1C and SI Appendix, Fig. S4 A and B). Comparing the ¹H-¹⁵N correlation spectra between wild-type TTD-PHD and TTD-PHD_S298ph, phosphorylation of S298 caused chemical shift differences in 15 signals (SI Appendix, Fig. S11A), implying local conformational changes at or near the junction. Nevertheless, phosphorylation of S298 decreased the apparent binding affinity for the H3-K9me3 tail ∼30-fold and abrogated the binding stoichiometry (n = 1.49) (Fig. 4B and Table 1). Indeed, the NMR spectrum of the H3-K9me3 peptide bound to TTD-PHD_S298ph did not show peaks corresponding to those of the H3 residues that presumably form the helix in the nonphosphorylated complex (Figs. 3B and 4D). Collectively, either local perturbation of the linker:TTD contacts or alteration of the junction conformation by S298ph disrupts the simultaneous recognition of multiple modifications in the histone H3 tail by TTD-PHD.

Protein Preparation.

DNA fragments encoding the PHD finger (299–366), TTD (123–285), TTD-linker (123–300), and TTD-PHD (123–366) of human UHRF1 were each amplified by PCR and cloned into modified a pGEX4T-3 plasmid (GE Healthcare Bio-Sciences) engineered for protein expression with an N-terminal GST and small ubiquitin like modifier (SUMO)-1 fusion tag. For ITC and NMR experiments, a TTD-PHD_R295A/R296A mutant was generated by the QuikChange method (Stratagene). We also prepared the TTD-PHD, residues 134–366, with a deletion of residues 167–175, for crystallography. Each protein was overexpressed in Escherichia coli strain Rosetta (DE3) and purified by affinity, ion-exchange, and size-exclusion chromatography (GE Healthcare Bio-Sciences). More details are provided in SI Appendix, SI Methods.

Preparation of TDD-PHD_S298ph Using an E. coli Phosphorylation System.

Phosphorylated TTD-PHD was prepared by coexpression with rat cAMP-dependent protein kinase (rPKA) (residues 1–350) in E. coli BL21 (DE3). Phosphorylation efficiency and specificity was examined by a gel shift-mobility assay using Phos-tag SDS/PAGE and liquid chromatography-electrospray ionization-tandem mass spectroscopy (LC-ESI-MS/MS) analysis. More details are provided in SI Appendix, SI Methods and SI Results.

Peptide Preparation.

The histone H3 peptides H3_1–10 and H3_1–12-K9me3 for crystallography and a series of H3 peptides for ITC measurements were purchased from Toray Research Center. Histones H3_1–20 and H4 peptides harboring a site- and degree-specific methylated lysine analog were prepared by cysteine alkylation as reported previously (36, 37).

ITC Measurements.

A MicroCal calorimeter, itc200, was used for the ITC measurements. Protein solutions were dialyzed into buffer [10 mM Hepes (pH 7.5), 150 mM NaCl, 0.1 mM tris(2-carboxyethyl)phosphine (TCEP), 30 μM zinc acetate], and lyophilized histone peptides were dissolved in the same buffer. All measurements were carried out at 293 K. The protein solution (30–60 μM) in the calorimetric cell was titrated with the peptide solution (400–800 μM). The data were analyzed with the software ORIGIN (MicroCal) using a one-site model. The first data point was excluded from the analysis. For each interaction, at least three independent titration experiments were performed. A summary of the ITC measurement and results is given in SI Appendix, Table S2 and Figs. S1–S3.

X-Ray Crystallography of TTD-PHD in Complex with the H3-K9me3 Peptide.

TTD-PHD (residues 123–366), with residues 123–132 deleted in the N-terminal region and 167–175 deleted in a loop region, was used for crystallization. The TTD-PHD:H3 complex was prepared by adding a 1.5-molar excess of H3_1–12K9me3 peptide to the protein before concentration. The crystal was obtained by using a 20 mg/mL concentration of the complex at 4 °C and the hanging drop vapor diffusion method with a reservoir solution containing 100 mM Bis-Tris propane (pH 6.5), 200 mM sodium citrate, and 20% PEG3350. The crystal was directly frozen in liquid nitrogen using a cryoprotectant containing 20% (vol/vol) ethylene glycol. The X-ray diffraction data were collected at a wavelength of 1.0000 Å on an ADSC Q315 CCD detector in beam-line BL-5A at Photon Factory (PF) (Tsukuba, Japan) and scaled at 2.9 Å resolution with the program HKL2000 (38). The structure was solved by molecular replacement using the coordinates of TTD (PDB ID code 3db3) and the PHD finger of UHRF1 as search models. After several cycles of refinement by the program PHENIX (39) with noncrystallographic symmetry restraints in the four TTD moieties, the model was converged, resulting in a crystallographic R value of 24.2% and a free R value of 28.6% for all diffraction data up to 2.9-Å resolution. Four TTD-PHD complexes, designated as complexes A, B, C, and D, are included in an asymmetric unit and show different average temperature factors, in particular for the PHD moiety (see SI Appendix, SI Results and SI Methods). The structure of complex A is described in the main text. The peptide bond between Gly-299 and Pro-300 in molecules A and B is in the cis conformation. The crystallographic data and refinement statistics are given in SI Appendix, Table S1. More details are provided in SI Appendix.

SAXS.

SAXS measurements were performed with a Rigaku NANO-VIEWER system mounted on a Rigaku FR-D X-ray generator. The scattering intensities were recorded with a PILATUS 100K detector. SAXS data were analyzed with the software package ATSAS 2.4. The radius of gyration R_g was estimated from the Guinier plot (40) of I(q). The distance distribution function P(r) was calculated in the program GNOM (41). More details are provided in SI Appendix, SI Methods.

NMR Spectroscopy.

The ¹H-¹⁵N band-selective optimized-flip-angle-short-transient (SOFAST)–heteronuclear multiple quantum coherence (HMQC) spectra (42) of wild-type TTD-PHD, TTD-PHD_R295A/R296A, TTD-PHD_S298ph, TTD, TTD linker, and PHD finger were acquired at 30 °C on a Bruker DRX-700 spectrometer equipped with a TCI cryogenic probe. To assign the backbone ¹NH, ¹⁵N, ¹³C_α, and ¹³CO chemical shifts of the H3-K9me3 peptide, ¹H-¹⁵N heteronuclear single quantum coherence (HSQC), HNCO, HNCA, and HN(CO)CA spectra were measured with a Bruker Avance 800-MHz spectrometer equipped with a TCI cryogenic probe at 30 °C. More details are provided in SI Appendix, SI Methods and SI Results.