Technical Comments

Comment on “Structural basis of histone H3K27 trimethylation by an active polycomb repressive complex 2”

See allHide authors and affiliations

Science  23 Dec 2016:
Vol. 354, Issue 6319, pp. 1543
DOI: 10.1126/science.aaf6236

Abstract

Jiao and Liu (Research Articles, 16 October 2015, aac4383) reported the crystal structure of the protein complex polycomb repressive complex 2 from Chaetomium thermophilum. This landmark structure has brought invaluable insights into the activation mechanism of this essential methyltransferase. However, the analysis of the x-ray data discussed below suggests that the description of oncogenic H3K27M peptide binding to the active site is incorrect.

Polycomb repressive complex 2 (PRC2) represses the expression of target genes through the methylation of associated nucleosomes on histone H3K27 (1). The catalytic subunit, EZH2, is a SET [su(var)3-9, enhancer-of-zeste and trithorax] domain lysine methyltransferase, which has a characteristic hydrophobic channel at its active site that positions the target substrate lysine side chain for methylation (2, 3). Mutations that disrupt the normal function of EZH2, both activating and deactivating, are associated with several cancer pathways (4, 5). In addition, the histone variant H3K27M is a dominant-negative somatic mutation that is implicated in high-grade pediatric gliomas (6, 7). Even low-level expression of this variant was found to effectively inhibit methylation of the wild-type histone by PRC2 (8). The recent structure of the catalytically active subcomplex of Chaetomium thermophilum PRC2, in complex with a H3K27M-derived peptide, provides a major breakthrough in understanding the basis of the potency of H3K27M and the disruption of normal EZH2 function (9). Unexpectedly, in the proposed model, the side chain of arginine-26, rather than the methionine at position 27, occupies the substrate “lysine” channel in the active site of the EZH2. We were surprised that the mutant side chain had no role in recognition and therefore reexamined the C. thermophilum x-ray data, Protein Data Bank (PDB) entry 5CH1.

The C. thermophilum EZH2 SET domain map features electron density that is characteristic of a protein chain at the active site in the groove where substrate peptide is expected to bind (2). In the Jiao and Liu model, the histone H3K27M peptide (corresponding to residues 21 to 29) is built into this electron density (9). Surprisingly, there is no feature corresponding to the variant methionine-27 side chain in the electron density map. Previous biochemical data indicated that properties of the side chain at position 27 dictate the potency of peptide inhibitors (8, 10). These analyses indicated that an unbranched hydrophobic side chain is favored at the 27 position, and it is difficult to reconcile this with a model in which the methionine-27 side chain is not ordered. Further inspection of the C. thermophilum data highlights additional issues (Fig. 1A). Flanking side chains both show a poor fit to the electron density and have poor complementarity to the surface of the enzyme. For example, the lysine side chain at the –3 position has been truncated by two atoms and is inserted into a hydrophobic pocket, and the serine at the +2 position has been built in two conformations to account for the shape of the observed density, one of which forces the hydroxyl into a second hydrophobic pocket.

Fig. 1 Fitting of the H3 and VEFS sequence into the C. thermophilum x-ray data.

(A) The side chains of the H3 sequence (KAARMS) do not fit well into the electron density map (refined 2Fo – Fc density). Residues K(–3) and M(+1) required truncation, and A(–1) is too large. To fill the observed density at the +2 position, a serine has been placed in two conformations. In addition, the electrostatic surface of the C. thermophilum EZH2 substrate binding cleft is not complementary to the H3 sequence—for example, requiring that K(–3) amine and the S(+2) hydroxyl would sit in hydrophobic pockets. The missing atoms for K(–3) are indicated by a dotted line. (B) The SUZ12 sequence (LPGRGV) is a better fit for the electron density map. The L(–3) and V(+2) side chains, in addition to fitting the density, are complementary to the two hydrophobic pockets. We have deposited coordinates for this alternative assignment of the substrate peptide with PDB code 5M5G.

We propose an alternative interpretation of the C. thermophilum data. Inspection of the 2Fo – Fc and omit maps at each peptide side chain position suggested that that the sequence LPGRGV would better fit the electron density and would be complementary to the binding surface (Fig. 1B). A search of the database entries with this sequence identified a perfect match in the C. thermophilum SUZ12 component of the PRC2 complex (residues 533 to 538) amino terminal to the VEFS [Vrn2-Emf2-Fis2-Su(z)12] domain. These residues are present in the crystallization fragment but were not accounted for in the atomic model. Furthermore, the molecular packing of the yeast PRC2 crystal suggests that this motif could be contributed from the SUZ12 subunit of a neighboring lattice molecule (Fig. 2). The LPGRGV sequence of SUZ12 is positioned between the VEFS domain and EZH2 on an artificial loop that arises from the engineered linker between EZH2 C-terminus and the SUZ12 VEFS domain in the crystallographic construct. There are sufficient unbuilt residues to account for the distance from the N- and C-terminal ends of the LPGRGV motif to the neighboring molecule. Although it is likely that this sequence has lower affinity for EZH2 than the H3K27M peptide that was included for co-crystallization, the higher “effective” local concentration in the crystalline environment would allow it to outcompete the free peptide.

Fig. 2 The peptide chain occupying the SET domain active site most likely originates from a neighboring crystal lattice copy.

(A) Schematic diagram of the EZH2-SUZ12 fusion construct in the asymmetric unit (blue) and symmetry related molecule (gray). The LPGRGV sequence is located N-terminal to the VEFS domain but immediately after an engineered loop on the crystallographic construct. The 28 residues preceding and 10 following this sequence are not built in the structure. (B) Cartoon representation of the relevant region of PRC2 (light blue) and a neighboring crystal lattice copy (gray). The SUZ12 protein is colored in green and the C terminus of the neighboring copy EZH2 in orange. The LPGRGV region is shown as a green stick representation in the active site of EZH2. The distance to the C terminus of the SET domain and the N terminus of the SUZ12 region is compatible with the number of unbuilt amino acids.

The binding mode reported for other SET domain structures with repressive peptides has the target lysine side chain occupying the substrate channel, but, notably, there is a conserved arginine in the –1 position, which has an important role in recognition (1113). This arginine side chain sits in a complementary pocket on the surface of the SET domain and makes a series of salt bridges. This well-conserved arrangement was also observed in our recent structure of the human PRC2 complex with oncogenic H3K27M peptide (14). To support their assignment of the H3K27M peptide with noncanonical binding of the arginine, Jiao and Liu have presented activity data that shows that the R26A/K27M mutant peptide no longer inhibits methylation (9). Although consistent with their model, it is important to note that these data are also consistent with the binding mode found in other repressive peptide structures, where R26 makes essential contacts for peptide recognition. It would therefore be expected for the R26A/K27M to have reduced potency.

Our analysis therefore suggests that the proposed molecular model of the oncogenic histone peptide in the C. thermophilum PRC2 structure is incorrect, arising from the misidentification of the peptide bound to the SET domain. Thus, although we agree that there is an arginine side chain in the crystal structure of yeast PRC2, it is an artifact of the crystallization environment, and in physiological conditions the substrate channel would be occupied by the mutant methionine at position 27. This is important because it suggests that C. thermophilum PRC2 likely binds its histone substrate in a canonical manner. Perhaps more important, this reevaluation leads to a molecular basis for the potency of the H3K27M mutation that is consistent with earlier studies (6, 8, 10).

References and Notes

Acknowledgments: This work was supported by the Francis Crick Institute, which receives its core funding from Cancer Research UK (FC001078), the UK Medical Research Council (FC001078), and the Wellcome Trust (FC001078).
View Abstract

Navigate This Article