Unique Structural Platforms of Suz12 Dictate Distinct Classes of PRC2 for Chromatin Binding

Summary

Developmentally regulated accessory subunits dictate PRC2 function. Here, we report the crystal structures of a 120KDa heterotetrameric complex consisting of Suz12, Rbbp4, Jarid2 and Aebp2 fragments that is minimally active in nucleosome binding, and of an inactive binary complex of Suz12 and Rbbp4. Suz12 contains two unique structural platforms that define distinct classes of PRC2 holo complexes for chromatin binding. Aebp2 and Phf19 compete for binding of a noncanonical C2 domain of Suz12; Jarid2 and EPOP occupy an overlapped Suz12 surface required for chromatin association of PRC2. Suz12 and Aebp2 progressively block histone H3K4 binding to Rbbp4, suggesting that Rbbp4 may not be directly involved in PRC2 inhibition by the active H3K4me3 histone mark. Nucleosome binding enabled by Jarid2 and Aebp2 is in part accounted for by the structures, which also reveal that disruption of the Jarid2–Suz12 interaction may underlie the disease mechanism of an oncogenic chromosomal translocation of Suz12.

Graphical abstract

Introduction

Polycomb repressive complex 2 (PRC2) mediates trimethylation of histone H3K27 (H3K27me3) to maintain repressive chromatin domains. The core PRC2 complex comprises Ezh1/2, Eed, Suz12 and Rbbp4/7. A plethora of developmentally regulated accessory subunits profoundly impact function and chromatin recruitment of PRC2 and display mutual exclusivity in holo complex formation (Grijzenhout et al., 2016; Simon and Kingston, 2013; Vizan et al., 2015)

Accessory subunits Jarid2 and Aebp2 define a distinct class of PRC2 holo complex (Grijzenhout et al., 2016). Both of them show colocalization with PRC2 on chromatin and play pleiotropic roles in embryonic development (Simon and Kingston, 2013; Vizan et al., 2015). Loss of Jarid2 hampers PRC2 chromatin occupancy and gene repression in embryonic stem cells (Landeira et al., 2010; Li et al., 2010; Pasini et al., 2010; Shen et al., 2009). Mutation of Abep2 or its fly homolog Jing causes developmental defects (Grijzenhout et al., 2016; Kim et al., 2011; Liu and Montell, 2001). Jarid2 and Aebp2 also enhance the enzymatic activity of PRC2 on nucleosomes (Cao and Zhang, 2004; Kalb et al., 2014; Li et al., 2010; Son et al., 2013; Zhang et al., 2011).

Human homologs of Drosophila polycomblike (PCL), including Phf1, Mtf2 and Phf19, form another class of PRC2 holo complexes that are incompatible with Aebp2 (Grijzenhout et al., 2016). The tudor domain of these accessory subunits recognizes the active H3K36me3 histone mark and recruits PRC2 to control a subset of developmental genes (Ballare et al., 2012; Brien et al., 2012; Cai et al., 2013; Musselman et al., 2012). Furthermore, EPOP (a.k.a. C17orf96) coexists with Phf1, Mtf2 or Phf19 in PRC2 holo complexes in the absence of Jarid2, and regulates transcription by also interacting with the elongation factors elongin BC (Alekseyenko et al., 2014; Beringer et al., 2016; Grijzenhout et al., 2016; Liefke et al., 2016).

Local epigenetic context modulates PRC2 function as well. H3K27me3 is bound by Eed and the stimulation-responsive motif (SRM) of Ezh2 to stimulate PRC2 catalysis such that the repressive H3K27me3 histone mark is propagated on chromatin (Hansen et al., 2008; Jiao and Liu, 2015; Margueron et al., 2009). In addition, the ubiquitin-interacting (UI) motif of Jarid2 interacts with the monoubiquitinated histone H2AK119 (H2AK119ub) to promote H3K27 methylation and allow PRC2 targeting by polycomb repressive complex 1 (PRC1) (Cooper et al., 2016; Kalb et al., 2014). Conversely, H3K36me3 that is associated with actively transcribed genes inhibits PRC2 catalysis (Musselman et al., 2012; Yuan et al., 2011). The active H3K4me3 histone mark greatly impairs Rbbp4 binding to the unmethylated histone H3K4 tail and antagonizes H3K27 methylation by PRC2, which conceivably prevents PRC2 invasion into active chromatin domains. However, what remains puzzling is that Rbbp4 appears to be dispensable for the H3K4me3-mediated PRC2 inhibition (Schmitges et al., 2011).

Functional analysis of the increasing collection of the “substoichiometric” accessory subunits of PRC2 has so far been mostly limited to proteomics and genomics studies of individual accessory subunits and calls for mechanistic studies with a defined, fully reconstituted system. Understanding of the regulation of PRC2 recruitment to chromatin is also largely hindered due to the lack of structures of any of these accessory subunits caught in the complex with PRC2 that allow the building of reliable molecular models. We present here the 2.9Å X-ray crystal structure of a reconstituted 120KDa heterotetrameric complex of Suz12, Rbbp4, Jarid2, and Aebp2 that is active in nucleosome binding, and the 3.2Å crystal structure of a corresponding binary complex of Suz12 and Rbbp4 that is inactive. The thorough structural and biochemical analysis reveals that two unique structural platforms of Suz12 not only dictate formation of different classes of PRC2 holo complexes that are indeed stoichiometric in reconstitution, but also define different modes of chromatin binding. Together, this work provides a clear structural framework to begin to clarify the mechanism of functional control of PRC2 by accessory subunits that exhibit a high degree of complexity during development, explaining a number of biochemical, genetic and disease data as well.

Results

Overall structure of a Suz12–Rbbp4–Jarid2–Aebp2 heterotetrameric complex

We and others have previously determined the crystal structures of a minimally active catalytic module of PRC2, containing Ezh2, Eed and Suz12(VEFS) (Fig. 1A) (Jiao and Liu, 2015; Justin et al., 2016). Rbbp4 and the N-terminal portion of Suz12 (residues 76–545) [Suz12(N)] are folded together to form the nucleosome-binding module of PRC2 (Ciferri et al., 2012; Nekrasov et al., 2005) (Fig. 1A). In the current study, we solved the crystal structures of the nucleosome-binding module of PRC2 (i.e. Suz12–Rbbp4; hereafter referred to as “S12R4”) (Fig. 1A and S1A and Table 1), and of S12R4 simultaneously bound to both Jarid2 and Aebp2 fragments minimally required for nucleosome binding (i.e. Suz12–Rbbp4–Jarid2–Aebp2; hereafter referred to as “S12R4J2A2”) (Fig. 1A, 1B and 1C and Table 1). Model building was facilitated by single-wavelength anomalous dispersion (SAD) signals of sulfur atoms in the complex (Fig. S1B and S1C).

Fig. 1. Overall structure of the Suz12–Rbbp4–Jarid2–Aebp2 heterotetrameric complex active in nucleosome binding.

(A) Domain architecture of the Jarid2- and Aebp2-containing holo-PRC2 complex is outlined. Compositions of the catalytic module and nucleosome-binding module of the core PRC2 complex are indicated. Protein domains of Suz12, Rbbp4, Jarid2 and Aebp2 that are included in the crystal structure of the heterotetrameric complex are color coded and labeled. Domains not included in the structure are in gray.

(B) Top View of the cartoon representation of the heterotetrameric complex. Structure of individual domains is colored and labeled according to the schematic in (A). Disordered loops that are missing from the structure are indicated by dotted lines. Zinc atom coordinated by the C2H2-type zinc finger of Suz12 is represented as an orange sphere. A rotation matrix is provided to guide the transition of the views from (B) to (C).

(C) Side view of the complex.

Table 1.

Data collection and refinement statistics.

Crystal	Suzl 2-Rbbp4-Jarid2-Aebp2	Suz12-Rbbp4
Diffraction data
Wavelength (Å)	0.97946
Space group	P212121	P212121
a, b, c (Å)	95.3, 111.4, 253.2	106.0, 107.7, 107.8
α, β, γ (°)	90.0, 90.0, 90.0	90.0, 90.0, 90.0
Resolution range (Å)	50.00–2.90 (2.95–2.90)a	50.00–3.20 (3.26–3.20)
Rsym or Rmerge	0.150 (2.851)	0.067(0.000)
Mean I/σ (I)	22.4 (1.5)	23.6 (1.1)
Rpim (%)	4.2 (82.8)	3.6 (88.0)
CC1/2 (%)	99.9 (47.2)	99.9 (78.8)
Completeness (%)	99.1 (99.8)	100.0 (100.0)
Redundancy	13.1 (12.6)	7.3 (7.5)
Refinement
Bumber of reflections	67,533	15,396
Rwork/Rfree	0.18/0.22	0.18/0.25
Number of non-hydrogen atoms	13,018	4,766
Suz12	4,978	1,637
Rbbp4	6,288	3,128
Aebp2	1,490
Jarid2	260
Ligand (zinc ion)	2	1
Protein residues	1,601	588
Average B-factor (Å2)	62.3	75.1
Suz12	73.5	83.8
Rbbp4	49.2	70.6
Aebp2	83.4
Jarid2	47.9
Ligand (zinc ion)	30.2	52.8
R.m.s deviations
Bond lengths (Å)	0.010	0.010
Bond angles (°)	1.22	1.25
Ramachandran
Favored (%)	95.87	95.79
Allowed (%)	4.00	3.91
Outliers (%)	0.13	0.30

^aStatistics for the highest-resolution shell are shown in parentheses.

Four regions of Suz12, including ZnB, WDB1, Zn, and WDB2, are present in the structures of both S12R4 and S12R4J2A2 (Fig. 1 and S1A). They are scattered on the surface of the WD40 β-propellers of Rbbp4 (Fig. 1 and S1A). The zinc finger-binding (ZnB) helix forms an intramolecular complex with the C2H2-type zinc finger (Zn) (Fig. 1 and S1A). Two separate WD40-binding domains, WDB1 and WDB2, are closely associated with Rbbp4 (Fig. 1 and S1A). The WDB1 includes a minimal Rbbp4-binidng epitope formerly characterized in the fly homolog of Suz12, Su(z)12 (Fig. S1D) (Schmitges et al., 2011).

The Jarid2 and Aebp2 fragments captured in the crystal structure are known to interact with PRC2 in vitro and the Jarid2 fragment is also necessary for transrepression in vivo (also see below) (Cao and Zhang, 2004; Pasini et al., 2010). Comparing the two structures, an additional domain of Suz12 becomes better ordered in S12R4J2A2 due to Aebp2 binding (Fig. 1 and S1A). Remarkably, it adopts an eight-strand β-sandwich structure, similar to the C2 domain typically involved in phospholipid binding (Fig. 1B and 1C and see below). The Jarid2 transrepression (TR) domain (residues 147–165) binds to the ZnB–Zn complex of Suz12 (Fig. 1). The Aebp2 fragment (residues 407–503) is divided into the C2-binding (C2B) domain and the histone H3K4 displacement (H3K4D) domain; it spans from the ZnB to the C2 of Suz12 and ends on the central top surface of Rbbp4 (Fig. 1).

The overall structure of S12R4J2A2 fits well into the negative stain EM map of a holo-PRC2 (Fig. S1E) (Ciferri et al., 2012). The binding of Rbbp4 to many other cellular factors such as histone H4, FOG1, PHF6, and MTA1 is blocked in S12R4J2A2 such that the bound Rbbp4 is dedicated to PRC2 function (Fig. S1F–S1I) (Alqarni et al., 2014; Lejon et al., 2011; Liu et al., 2015; Millard et al., 2016; Murzina et al., 2008). In particular, in the case of Rbbp4 binding to metastasis-associated protein MTA1 of the NuRD histone deacetylase complex, the WDB1, Zn, and WDB2 domains of Suz12 all impose steric clashes with MTA1 on the Rbbp4 surface (Fig. S1I) (Alqarni et al., 2014; Millard et al., 2016).

Suz12 structure orchestrates different classes of PRC2 holo complexes that display distinct modes of chromatin binding

Proteomics and genomics studies have identified a growing number of developmentally regulated accessory subunits that modulate chromatin recruitment of PRC2 in vivo. The current crystal structures implicate that Suz12 functions to structurally organize different PRC2 holo complexes. We first tested the hypothesis that the S12R4 nucleosome-binding module of PRC2 drives stable association of the accessory subunits, such as Jarid2, Aebp2, Phf19 and EPOP (Fig. 1A, S2A–S2D). We further mapped the minimal regions of these accessory subunits sufficient for stable binding to S12R4, which include residues 141–170 of Jarid2 harboring the TR domain (Fig. S2A) (Pasini et al., 2010; Son et al., 2013), residues 407–478 of the C2B domain of Aebp2 (Fig. S2B) (Cao and Zhang, 2004), residues 531–580 of the “reversed chromodomain” (RC) of Phf19 (Fig. S2C) (Ballare et al., 2012), and residues 311–359 of the C-terminal domain of EPOP (Fig. S2D) (Liefke and Shi, 2015).

Notably, the isolated C2 domain of Suz12 was sufficient to form a stoichiometric complex either with an Aebp2 fragment harboring the C2B and H3K4D domains or with a Phf19 fragment harboring the RC domain by size-exclusion chromatography (Fig. 1A, 2A, 2B and 2C). Accordingly, the C2 domain of Suz12 may serve as a determinant for formation of respective holo complexes. In addition, either Aebp2 or Phf19 fragment prebound to the S12R4 nucleosome-binding module eliminated association of the other accessory subunit in a competitive GST pull-down assay (Fig. 2D and 2E). In contrast, Jarid2 binding to S12R4 is independent of either Aebp2 or Phf19 (Fig. 2D and 2E). The crystal structures together with these biochemical data indicate that the C2 domain of Suz12 mediates competitive binding of the C2B domain of Aepb2 and the RC domain of Phf19, likely dictating the mutual exclusivity of Aebp2 and Phf19 in PRC2 holo complex formation (Fig. 2F). Furthermore, this conclusion likely applies to Phf1 and Mtf2 based on the sequence homology between Phf19 and these two proteins (Ballare et al., 2012).

Fig. 2. Suz12 structure organizes distinct classes of PRC2 holo complexes.

(A) A schematic highlighting the domains of Suz12 that define different classes of PRC2 holo complexes. Schematics of Phf19 and EPOP are also shown.

(B) and (C) The C2B-H3K4D domain of Aebp2 (residues 407–503) and the RC domain of Phf19 (residues 500–580) form stoichiometric complexes with the C2 domain of Suz12 (residues 146–363). SDS-PAGE gels of the binary complexes are shown along with the respective elution profiles of size exclusion chromatography.

(D) Aebp2(C2B-H3K4D) but not Jarid2 (residues 119–232) that harbors the TR domain competes with Phf19(RC) for S12R4 binding. S12R4 was preincubated with Aebp2, Jarid2 or BSA prior to pull-down by GST-tagged Phf19(RC).

(E) This is a reciprocal assay to that shown in (D). S12R4 was preincubated with Phf19, Jarid2 or BSA prior to pull-down by GST-tagged Aebp2(C2B-H3K4D).

(F) A schematic illustrating that the C2 domain of Suz12 mediates the mutually exclusive binding of the C2B domain of Aebp2 and the RC domain of Phf19.

(G) A truncation of the Zn domain of Suz12 (residues 76–447) abolished the binding. A FLAG-tagged Jarid2 (residues 119–232) harboring the TR domain was used for the pull-down assay.

(H) An intact Zn domain of Suz12 is also necessary for EPOP binding to S12R4. A FLAG-tagged EPOP (residues 285–379) was used for the pull-down assay.

(I) The binding of Jarid2 and EPOP to S12R4 is mutually exclusive. Jarid2 lost binding to S12R4 prebound to EPOP.

(J) A schematic illustrating that the ZnB-Zn surface of Suz12 mediates the competitive binding of the TR domain of Jarid2 and the C-terminal domain of EPOP.

(K) The S12R4 nucleosome-binding module of human PRC2 does not bind to a 147-bp ‘601’ mononucleosome in an EMSA assay.

(L) S12R4J2A2 is active in nucleosome binding. Both Jarid2(TR) and Aebp2(C2B-H3K4D) fragments were required for nucleosome binding with a low micromolar binding affinity. The S12R4J2A2-bound mononucleosome ran as a well-defined shifted band on the native gel, indicative of formation of a specific supercomplex.

(M) S12R4P19EPOP exhibits poor nucleosome binding.

We next showed that an intact but not truncated C2H2-type Zn domain of Suz12 was necessary to mediate stable binding of the TR domain of Jarid2 (Fig. 1A, 2A and 2G). The same domain of Suz12 was also responsible for stable binding of the C-terminal domain of EPOP (Fig. 2A and 2H). Jarid2 lost binding to the S12R4 nucleosome-binding module in the presence of EPOP (Fig. 2I), indicative of direct competition between these two accessory subunits. In combination with the crystal structures, we concluded that the ZnB–Zn structural platform of Suz12 governs the mutual exclusive binding of Jarid2 and EPOP in PRC2 holo complexes, through the TR domain of Jarid2 and the C-terminal domain of EPOP (Fig. 2J). Collectively, our data indicate a central role of Suz12 in structurally organizing different classes of PRC2 holo complexes.

To confirm that our conclusion based on the purified protein fragments and PRC2 subcomplex is applicable to the full-length proteins and PRC2 holo complexes in vivo in a quantitative manner, we overexpressed the core and accessory subunits of PRC2 in HEK293T cells and carried out coimmunoprecipitation (Co-IP) assays. While the full-length Aebp2 formed a nearly stoichiometric complex with the core PRC2 complex, Aebp2 lacking the C2B-H3K4D region did not associate with it (Fig. S2E). Similarly, the RC domain was required for the binding of Phf19 to PRC2 in vivo (Fig. S2F). When the C-terminal domain was removed, the binding of EPOP to PRC2 was largely eliminated (Fig. S2G). Due to the low expression of Jarid2 in our system, we were not able to assess the quantitative binding of the PRC2 core subunits in a Coomassie blue-stained SDS-PAGE gel; nonetheless the Western blot result clearly indicated an indispensable role of the TR domain of Jarid2 in PRC2 binding (Fig. S2H). These data are in line with the results from the earlier protein binding studies as well (Ballare et al., 2012; Cao and Zhang, 2004; Liefke and Shi, 2015; Pasini et al., 2010; Son et al., 2013).

To investigate the modes of chromatin binding, we reconstituted the S12R4 nucleosome-binding module of human PRC2 in complex with different classes of accessory subunits for a series of electrophoretic mobility shift assays (EMSA) using mononucleosomes assembled with Widom ‘601’ sequences (Lowary and Widom, 1998). We found that S12R4 exhibited poor nucleosome binding by itself (Fig. 2K and S3A). In stark contrast, S12R4J2A2 displayed robust binding towards mononucleosomes, with a low micromolar binding affinity (Fig. 2L and S3B). Both the Jarid2 and Aebp2 fragments captured in the crystal structure of S12R4J2A2 were essential for its nucleosome binding activity. Neither S12R4J2 nor S12R4A2 ternary subcomplex displayed efficient nucleosome binding under the same condition (Fig. S3C–S3F). The catalytic module of PRC2 barely bound mononucleosomes too (Fig. S3G). In addition, the 4-subunit core PRC2 complex also required the minimal Jarid2 and Aebp2 fragments for mononucleosome binding, with a comparable binding affinity (Fig. S3H). These data suggest that chromatin recruitment of the core PRC2 complex and the accessory subunits Jarid2 and Aebp2 may become interdependent at least in a certain scenario, given that the S12R4J2A2 assembly is minimally active in nucleosome binding. Indeed, Jarid2 and the core PRC2 complex were known to be recruited to some PRC2 target genes as one entity in embryonic stem cells (Pasini et al., 2010).

Finally, as an equivalent of S12R4J2A2, a heterotetrameric complex of the S12R4 nucleosome-binding module bound to the RC domain of Phf19 and the C-terminal domain of EPOP that represents another class of PRC2 holo complex (hereafter referred to as “S12R4P19EPOP”) was inactive in nucleosome binding (Fig. 2M), suggesting that a different mode of chromatin binding may be used by this class of holo complex in vivo. Indeed, the region N-terminal of Phf19 not included in the reconstituted S12R4P19EPOP was implicated in PRC2 targeting in vivo, through direct recognition of either the active H3K36me3 histone mark by the tudor domain or CpG island DNAs by the extended homologous (EH) domain (Ballare et al., 2012; Brien et al., 2012; Cai et al., 2013; Choi et al., 2017; Li et al., 2017; Musselman et al., 2012). In comparison, both Aebp2 and Jarid2 have been implicated in H2AK119ub binding (Cooper et al., 2016; Kalb et al., 2014), and the zinc finger-containing N-terminal region of Aebp2 is also important for PRC2 binding to chromatin by mediating direct interaction with linker DNAs (Kim et al., 2009; Wang et al., 2017). In these regards, Suz12 appears to dictate linker DNA binding, histone mark binding as well as linker DNA and histone mark-independent nucleosome core binding, three contributing factors for the chromatin recruitment of PRC2, by selectively associating with different classes of accessory subunits (also see the discussion section below).

We will next focus on the structure of S12R4J2A2 that is minimally active in nucleosome binding to gain additional insights into the function and regulation of both the core PRC2 complex and the class of Jarid2 and Aebp2-containing PRC2 holo complex.

Suz12 contains a noncanonical C2 domain

The discovery of the C2 domain in Suz12 is unexpected (Fig. 3A and 3B). The C2 domain was originally identified in protein kinase C (PKC) and was subsequently found in a large number of proteins to mediate phospholipid binding and protein-protein interactions (Nalefski and Falke, 1996). The C2 domain of Suz12 contains over 200 amino acids, considerably larger in size than other C2 domains. Nonetheless, it shares notable structural homology with the classical C2 domains but differs from the latter by harboring longer interstrand loops (Fig. S4A–S4D and 3C) (Holm and Laakso, 2016). The topology of the C2 domain of Suz12 resembles the type II C2 domain exemplified in PLCδ, PTEN, and PI3Kα (Fig. 3B and S4A–S4C). A similar C2 domain may also be present in Su(z)12, Drosophila homolog of Suz12, based on sequence homology (Fig. 3C).

Fig. 3. The noncanonical C2 domain of Suz12.

(A) A zoom-in view of the C2 domain of Suz12 shown in Fig. 1C. The β-strands and interstrand loops from the β-sandwich structure are labeled.

(B) Schematic depiction of the topology of the C2 domain.

(C) Sequence alignment of the C2 domain of human Suz12 and Drosophila Su(z)12. Both the β strands and interstrand loops are labeled.

(D) The C2 domain of Suz12 in S12R4J2A2 sits on an edge of the top surface of Rbbp4. The view is related to the Side view of the heterotetrameric complex shown in Fig. 1C by a 240° rotation along the y-axis. Only the C2 d omain of Suz12 and Rbbp4 from the complex are shown in cartoon for clarity. Residue R196 of the C2 domain that functions as an arginine anchor on the Rbbp4 surface is highlighted as sticks in a dotted box.

(E) Zoom-in view of an acidic patch on the Rbbp4 surface that captures the R196 arginine anchor of the C2 domain. Extensive interactions on the binding interface are indicated by the black dotted lines. The blue mesh represents the well-ordered 2F_o−F_c electron density map of the binding interface contoured at 1σ.

In S12R4J2A2 the C2 domain of Suz12 maintains the structural integrity of the complex. While the β7 strand of the C2 domain is associated with the C2B of Aebp2, the L2 and L4 loops are seated on an edge of the top surface of Rbbp4 (Fig. 3D). In particular, the C2 domain uses an arginine anchor, residue R196, to interact with one of the acidic patches on Rbbp4 (Fig. 3D, 3E and S4E). The L1, L3, L5, and L7 loops on the distal side away from Rbbp4 are all disordered, implicating that they may be involved in other functions (Fig. 3D). Intriguingly, the interstrand loops at the equivalent positions in some C2 domains mediate phospholipid binding (Nalefski and Falke, 1996) (Fig. S4D).

It is not unprecedented that a chromatin complex contains a phospholipid-binding domain (Hamann and Blind, 2018). Indeed, phospholipids and lipid headgroups are known to modulate the enzymatic activity and chromatin association of certain histone modifying and chromatin remodeling complexes (Hamann and Blind, 2018; Kutateladze, 2012; Watson et al., 2012). However, the potential interplay of phospholipid binding and PRC2 function awaits future investigation.

The C2 domain of Suz12 impedes histone H3K4 binding to Rbbp4 in the core PRC2 complex

Notably, the C2 domain of Suz12 is a movable domain. Different from its location in S12R4J2A2, the C2 domain appears to occupy the central part of the top surface of Rbbp4 in S12R4 in the absence of Jarid2 and Aebp2 (Fig. S5A). The corresponding F_o−F_c difference electron density map indicates the position of the C2 domain in S12R4, and yet it is too flexible to allow model building (Fig. S5A). Nonetheless, the position of the C2 domain in S12R4 appears to overlap with residues R2, T3, and K4 of the histone H3K4 tail critical for the binding of Nurf55, the fly homolog of Rbbp4 (Fig. S5A) (Schmitges et al., 2011). Furthermore, the WDB2 domain of Suz12 presents steric clashes with the histone H3K4 tail too (Fig. S5A).

Congruent with the structural observation, Suz12 competes with H3K4 for Rbbp4 binding in solution. The binding affinity of an H3K4 peptide for Rbbp4 was measured by ITC to be 2.0 μM, comparable to that for Nurf55 reported previously (Fig. 4A) (Schmitges et al., 2011). In contrast, the binding affinity of the same H3K4 peptide for S12R4 was notably over 30-fold weaker at 70.1 μM (Fig. 4B). In vitro pull-down assay confirmed that H3K4 binding was severely impeded in S12R4 and in the 4-subunit core PRC2 complex compared with Rbbp4 alone (Fig. 4C). Both C2 and WDB2 domains of Suz12 contribute to the inhibition of H3K4 binding to Rbbp4 (Fig. 4D and S5B). We further showed that the unmethylated H3K4 was unable to pull down PRC2 from HeLa nuclear extract; as a positive control it associated with components of the NuRD complex in an H3K4 methylation-sensitive manner as shown previously (Fig. S5C) (Nishioka et al., 2002; Zegerman et al., 2002).

Fig. 4. Competitive binding of Suz12 and the histone H3K4 tail to Rbbp4 in the core PRC2 complex.

(A) and (B) ITC measurement of a synthetic histone H3K4 peptide (residues 1–19) titrated into Rbbp4 and the S12R4 binary complex.

(C) Competitive pull-down assay using a biotinylated histone H3K4 peptide harboring residues 1–21 of histone H3.

(D) Domain contribution to the inhibition of H3K4 binding to Rbbp4. As illustrated by the schematic, progressive deletion mutants of Suz12 were used in the context of the S12R4 binary complex. The presence of the C2 domain in Suz12_76–363 was sufficient to cause a notable inhibition compared to Rbbp4 alone. Inclusion of the WDB2 domain in Suz12_76–545 further enhanced the inhibition.

An important implication for our new finding that Suz12 greatly inhibits H3K4 binding to Rbbp4 is that Rbbp4 may not play a direct role in PRC2 inhibition by the active H3K4me3 histone mark. This interpretation is further supported by the fact that Abep2 completely blocks H3K4 binding to Rbbp4 in the context of the S12R4A2 ternary complex, a part of the Aebp2-containing PRC2 holo complex (see the next section below). Given the evidently specific and methylation-sensitive binding of H3K4 to Rbbp4 (Schmitges et al., 2011), it has been tempting to think that the Rbbp4 subunit of PRC2 may play a direct role in preventing invasion of PRC2 into the active H3K4me3 chromatin domain in vivo. However, with the more complete structure of PRC2 in which the H3K4 binding site on Rbbp4 is blocked, our results disfavor this model. Indeed, an alternative model of PRC2 inhibition by this active histone mark was already proposed by the previous study: H3K4me3 can directly antagonizes the enzymatic activity of PRC2 in the absence of Rbbp4 (Schmitges et al., 2011) (also see the discussion section below).

Isoform- and species-specific C-termini of Aebp2 block H3K4 binding to Rbbp4 and regulate nucleosome binding

Aebp2 is required for normal development in mice, and it is exclusively associated with a class of PRC2 holo complex that also contains Jarid2 in mouse embryonic stem cells (mESCs) (Grijzenhout et al., 2016; Kim et al., 2011). Different isoforms of Aebp2 differ in N-terminal domains and contain characteristic C-termini as well (Kim et al., 2009). There are three human Aebp2 isoforms in the UniProt database, two longer forms (isoforms 1 and 2) and one shorter form (isoform 3) corresponding to the somatic and embryonic isoforms of mouse Aebp2, respectively (Fig. S6A) (Kim et al., 2009).

The Aebp2 fragment used in the crystal structure of S12R4J2A2 contains the last 97 residues of human Aebp2 isoform 2, including the C2B and H3K4D domains (Fig. 1A). The C2B of Aebp2 is conserved from Drosophila to humans (Fig. S6A). It binds to the C2 domain of Suz12 and relocates the latter to a unique position in S12R4J2A2 to promote nucleosome binding (Fig. S5A).

The H3K4D domain harboring the last 9 residues of Aebp2 is extended to the acidic central cavity of Rbbp4, through a path orthogonal to that for the histone H3K4 tail (Fig. 5A) (Schmitges et al., 2011). Remarkably, a close examination reveals that residues K502 and R503 of Aebp2 are engaged in the same set of Rbbp4 interactions as residues K4 and R2 of H3K4, respectively (Fig. 5B). The C-terminus of human Aebp2 isoform 2 thus appears to displace H3K4 from Rbbp4 through direct competition. The R196 arginine anchor of the C2 domain of Suz12 partially blocks the binding site of the N-terminal amine of H3K4 as well (Fig. 3E and S5A) (Schmitges et al., 2011). In agreement with the structure, H3K4 binding to the S12R4 nucleosome-binding module in the presence of Aebp2 (i.e. the S12R4A2 ternary complex) was completely blocked according to the ITC result (Fig. S6B). H3K4 binding to Rbbp4 is thus progressively eliminated by Suz12 and Aebp2 (Fig. 4A, 4B and S6B), a mechanism that may prevent PRC2 from competing with the NuRD repressor complex as a direct sensor for H3K4 methylation. Intriguingly, NuRD was previously shown to enable the chromatin recruitment of PRC2 to a subset of genes in embryonic stem cells by mediating histone H3K27 deacetylation (also see the discussion section below) (Reynolds et al., 2012).

Fig. 5. Isoform- and species-specific C-termini of Aebp2 regulate nucleosome binding by a class of PRC2 holo complex.

(A) A zoom-in view of the Top view shown in Fig. 1B. The structure of S12R4J2A2 is compared to the structure of the Nurf55-H3K4 binary complex (PDB: 2YBA). The aligned Rbbp4/Nurf55 proteins from these two complexes are shown in pink and gray, respectively. The WDB2 domain and the L2 loop of the C2 domain of Suz12 are shown in purple and green. The H3K4D domain of Aebp2 is in gold and the histone H3K4 peptide is in magenta. The C-terminus of the H3K4D domain and the N-terminus of the histone H3K4 peptide are both inserted into the axial channel of Rbbp4/Nurf55. The dotted arrows represent the opposite polarity and show the orthogonal spatial relationship of the H3K4D domain of Aebp2 and the histone H3K4 peptide.

(B) Zoom-in view of the direct competition between the H3K4D domain of Aebp2 and the histone H3K4 peptide. Residues K502 and R503 of the H3K4D domain occupy the same binding pockets and interact with the same set of Rbbp4 residues as residues K4 and R2 of the histone H3K4 peptide, respectively.

(C) The reconstituted 147-bp mononucleosomes are shown together with the results of the EMSA assays. The wild-type and tailless H3-containing mononucleosomes display comparable binding affinities to S12R4J2A2.

(D) Quantitative comparison of the mononucleosome binding affinity of S12R4J2A2 containing Aebp2 variants. Aebp2^a (residues 407–517 of isoform 1) contains the C-terminus of human Aebp2 isoforms 1 and 3, and Aebp2^d (residues 407–498 of isoform 2) mimics the Aebp2 homologs from some lower eukaryotes, where the H3K4D domain is missing. Three independent EMSA assays for each of the S12R4J2A2 complexes were performed. A representative EMSA gel is shown in Fig. S26.

It was impossible to assess the contribution of the histone H3K4 tail to nucleosome binding by the S12R4 nucleosome-binding module since the binary complex displays little nucleosome binding activity. However, consistent with the binding data for the histone H3K4 peptide shown above, we were indeed able to show that the entire histone H3 tail (residues 1–27) does not contribute to mononucleosome binding by the minimally active S12R4A2J2 nucleosome-binding module (Fig. 5C and S6C), further supporting that the H3K4 binding site on Rbbp4 is not accessible in PRC2.

The H3K4D domain of Aebp2 is conserved in mammals but not in lower eukaryotes (Fig. S6A). In particular, the Jing protein from Drosophila contains a domain homologous to the C2B of Aebp2 but lacks the H3K4D (Fig. S6A). We disturbed the H3K4D by either deleting the last 5 residues to mimic Aebp2 from some lower eukaryotes (A2^d) or adding the 14 residues found in isoforms 1 and 3 (A2^a) (Fig. S6A). Notably, while S12R4J2A2^d lost binding to mononucleosomes, S12R4J2A2^a exhibited a 2-fold enhancement of nucleosome binding compared to the wild-type counterpart (Fig. 5D and S6D), suggesting that the H3K4D domain of Aebp2 is involved in nucleosome binding and that the distinct C-termini of Aebp2 found in different isoforms and species may tune chromatin association of PRC2 in vivo.

Jarid2 recognizes a composite Suz12 surface that is disrupted in an oncogenic chromosomal translocation

The TR domain of Jarid2 fragment used in S12R4J2A2 is necessary for nucleosome binding in vitro and Jarid2-mediated gene repression in vivo (Fig. 2L, S3B, S3D and S3F) (Pasini et al., 2010). These functional observations can be accounted for by the crystal structure of S12R4J2A2. The ZnB and Zn domains of Suz12 form a unique structural platform of an intramolecular complex that provides a composite concave surface for Jarid2 binding (Fig. 6A). An overall similar Suz12 surface is present in the S12R4 binary complex in the absence of Jarid2 (Fig. S7A) and may serve as the binding site for EPOP as suggested above (Fig. 2H).

Fig. 6. A composite Suz12 surface for Jarid2 binding.

(A) Surface representation of the composite Suz12 surface formed by the Zn (violet) and ZnB (cyan) domains. The TR domain of Jarid2 is shown in cartoon with a limon color. The rotation matrix relative to the Top view of the heterotetrameric complex in Fig. 1B is provided.

(B) The view is the same as in (A). The Zn and ZnB domains of Suz12 and the TR domain of Jarid2 are shown in cartoon representation. The breakpoint on the ZnB domain of Suz12 from the oncogenic chromosomal translocation is indicated by a red arrow.

(C) Jarid2 binding was abolished for the Δ79–106 deletion mutant and the F86A/F90A double mutant of the ZnB domain of Suz12.

(D) Suz12 and the JAZF1-Suz12 oncogenic fusion protein were co-expressed to a comparable level in HEK293T cells that also ectopically expressed the other core subunits of PRC2 and Jarid2. Jarid2 binding was dominated by Suz12 but not JAZF1-Suz12. Suz12 and JAZF1-Suz12 were run separately next to the marker lane as the gel migration control.

The C-terminal amphipathic helix of the Jarid2 fragment is responsible for the stable association with Suz12 (Fig. 6B). A cluster of positively charged residues was identified in the N-terminal loop region (residues 149–153) of the Jarid2 fragment; the charge reversal or alanine mutations of these residues eliminated nucleosome binding by S12R4J2A2 (Fig. 6B and S7B). The N-terminal loop region of the TR domain of Jarid2 thus represents another nucleosome binding surface in S12R4J2A2, in addition to the H3K4D domain of Aebp2 (Fig. 5D and Fig. S6D).

The C2H2-type Zn domain of Suz12 is highly conserved across species ranging from fungi to plants and humans (Rai et al., 2013). Although Drosophila PRC2 harboring a mutated Zn domain of Su(z)12 retains a comparable enzymatic activity as the wild type counterpart in vitro, it is defective in chromatin association in vivo (Rai et al., 2013). This phenomenon can be understood in the following way. Mutations of the Zn domain do not hamper PRC2 catalysis per se since this domain is spatially separated from the catalytic SET domain of Ezh2 (Fig. S1E) (Rai et al., 2013). In contrast, the Zn domain of Suz12 is an integral component of the nucleosome-binding module of PRC2 and its integrity is required for Jarid2 binding (Fig. 2G). Therefore, the lack of Jarid2 binding caused by the Zn domain mutations may explain the observed loss of PRC2 recruitment to chromatin in vivo (Pasini et al., 2010; Rai et al., 2013).

Remarkably, the structure of S12R4A2J2 also indicates that disruption of the interaction between Jarid2 and Suz12 may underlie the disease mechanism of an oncogenic chromosomal translocation of Suz12. A recurrent chromosomal translocation found in endometrial stromal tumors generates the JAZF1-SUZ12 fusion oncogene (Koontz et al., 2001). We found that the breakpoint for translocation is located in the middle of the ZnB helix of Suz12 (Fig. 6B); the first half of the ZnB is replaced by the fusion partner in the chromosomal translocation product. We reasoned that Jarid2 binding to PRC2 might be impaired by the translocation based on the structure (Fig. 6B). In a pull-down assay with a biotinylated TR domain peptide of Jarid2, we showed that deletion of the ZnB helix of Suz12 largely diminished Jarid2 binding to the 4-subunit PRC2 and additionally that alanine mutations of residues F86 and F90 located in the first half of the ZnB were sufficient to abolish Jarid2 binding (Fig. 6C). As a control, these Suz12 mutations did not compromise the integrity of the core PRC2 complex (Fig. S7C).

We made a JAZF1-Suz12 construct mimicking the actual oncogenic fusion protein as previously reported (Ma et al., 2017). We co-expressed an equal amount of wild-type Suz12 and JAZF1-Suz12 in HEK293T cells together with Jarid2 as well as the other core subunits of PRC2. The assembly of the core PRC2 complex was minimally affected by the fusion protein in our system (Fig. S7D). However, Jarid2 binding was largely diminished for JAZF1-Suz12 compared with the wild-type Suz12 (Fig. 6D). Collectively, our results strongly suggest that impaired Jarid2 binding to PRC2 and the resulted loss of chromatin association of PRC2 may underlie the oncogenic phenotype associated with this chromosomal aberration. In a similar assay, EPOP binding to JAZF1-Suz12 was greatly compromised as well (Fig. S7E), raising the possibility that dysregulation of EPOP function may also be related to the disease mechanism.

Discussion

A major highlight of the current study is that it provides a clear structural framework to understanding the formation, organization and function of distinct classes of PRC2 holo complexes. Notably, our results reveal that Suz12 domains play central roles in orchestrating PRC2 holo complex formation. The C2 of Suz12 mediates the mutually exclusive, stoichiometric binding of Aebp2 and Phf19, and the ZnB–Zn of Suz12 interacts with both Jarid2 and EPOP that also display mutual exclusivity in PRC2 binding. Two classes of PRC2 holo complexes, PRC2–Jarid2–Aebp2 and PRC2–Phf19/Mtf2/Phf1–EPOP, are thus structurally defined in this way (Fig. 7). Furthermore, the Suz12-centered analysis of the structural platforms of PRC2 may serve as a paradigm for elucidating the mode of action of new PRC2-interacting proteins.

Fig. 7. A schematic model of chromatin binding by distinct classes of PRC2 holo complexes dictated by Suz12.

(A) The structure of Suz12 is shown based on Fig. S4. The binding partners of the each domain of Suz12 in the Jarid2 and Aebp2-containing 6-subunit PRC2 holo complex are indicated.

(B) Three structural surfaces of Suz12, including the VFES, C2 and ZnB-Zn, are highlighted in colors. Suz12-interacting domains of the accessory subunits are also colored. Accessory subunits play essential roles in linker DNA binding (not shown), nucleosome core binding and histone tail binding. S12R4J2A2 is sufficient to bind a mononucleosome core. The C2 domain of Suz12 and the H3K4D domain of Aebp2 block histone H3K4 binding to Rbbp4. The UI motif of Jarid2 contributes to PRC2 binding to H2AK119ub-containing nucleosomes. The tudor domain of Phf19, Mtf2, or Phf1 binds the H3K36me3 histone tail directly. Other regulatory histone tails that may also facilitate the chromatin recruitment of PRC2 such as H3K27me3 are not shown for clarity.

How PRC2 is recruited to chromatin is a long-standing question. Distinct mechanisms may be used for different gene loci in different cells. At least three modes of direct binding may contribute to the chromatin recruitment of PRC2: linker DNA binding, histone mark binding, and nucleosome core binding independent of both linker DNAs and histone marks. While the mechanism for the former two modes that are largely mediated by functional domains of the accessory subunits started to emerge from the recent studies (Ballare et al., 2012; Brien et al., 2012; Cai et al., 2013; Choi et al., 2017; Cooper et al., 2016; Kalb et al., 2014; Kim et al., 2009; Li et al., 2017; Musselman et al., 2012; Wang et al., 2017), we demonstrated here that the binding between nucleosomes and the S12R4J2A2 nucleosome-binding module of the Jarid2 and Aebp2-containing PRC2 holo complex largely belongs to the third mode. This binding occurs with an affinity in the low micromolar range, which is comparable to, if not higher than, the affinity for the histone mark binding by some core and accessory subunits of PRC2 (Margueron et al., 2009; Musselman et al., 2012). Taking together, our finding that Suz12 structurally orchestrates different PRC2 holo complexes helps generate a more complete picture regarding the different modes of PRC2 targeting in vivo (Fig. 7).

Another notable discovery from this study is that a recurrent oncogenic chromosomal translocation of Suz12 compromises Jarid2 binding to PRC2 (Fig. 6). JAZF1-SUZ12 has been long known as a frequent chromosomal translocation in endometrial stromal tumors (Koontz et al., 2001). The molecular basis for the connection between the disease and the translocation product was unclear. Our structural and biochemical results suggested a possible mechanism: the Jarid2 binding surface on Suz12 is disrupted in the JAZF1-Suz12 fusion protein, which leads to loss of Jarid2 binding to PRC2. Given that Jarid2 plays an essential role in the chromatin recruitment of PRC2 (Landeira et al., 2010; Li et al., 2010; Pasini et al., 2010; Shen et al., 2009), we propose that dysregulation of the Jarid2-mediated PRC2 targeting may underlie the oncogenesis in patients diagnosed to have this chromosomal translocation.

In addition, we observed in the crystal structures that PRC2 and in particular Abep2-containing PRC2 evolve to specifically impede H3K4 binding to Rbbp4 (Fig. 4 and 5). NuRD is known as a major protein complex that associates with the unmethylated H3K4 tail (Nishioka et al., 2002; Zegerman et al., 2002), and for a subset of gene loci NuRD facilitates the chromatin recruitment of PRC2 and H3K27 methylation by deacetylating histone H3K27 (Reynolds et al., 2012). In this scenario, the different intrinsic enzymatic activities of the two repressor complexes naturally define that NuRD may be recruited before PRC2 as a primary sensor of H3K4 methylation during gene repression. Therefore, blocking H3K4 binding to PRC2 may be viewed as a structural mechanism to allow the sequential and cooperative action of NuRD and PRC2 in gene silencing.

Finally, the current work also suggested new directions for future studies. First, although we reconstituted the mononucleosome-binding activity in S12R4J2A2, other PRC2 surfaces may bind to additional nucleosomes on chromatin. Second, while the Jarid2 and Aebp2 fragments displayed strict interdependence in mononuclesome binding in the context of the minimally active S12R4J2A2 nucleosome-binding module, the full length Aebp2 or Jarid2 alone may be sufficient under certain circumstances to confer chromatin binding of PRC2 (Son et al., 2013; Wang et al., 2017). Third, the isoform- and species-specific H3K4D domain of Aebp2 appeared to modulate nucleosome binding in our vitro system (Fig. 5D), however the corresponding biological consequence and phenotype remain to be shown in a model organism. Finally, in addition to its obvious role in PRC2 holo complex formation, whether the C2 domain of Suz12 is a bona fide phospholipid-binding domain awaits further investigation.

STAR METHODS

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-6X His tag® antibody [HIS.H8]	abcam	Cat# ab18184
Monoclonal ANTI-FLAG® M2 antibody	Sigma-Aldrich	Cat# F3165
RbBP4 Antibody	Bethyl Laboratories Inc.	Cat# A301-206A-T
HDAC1 (10E2) Mouse mAb	Cell Signaling	Cat# 5356T
Ezh2 (D2C9) XP® Rabbit mAb	Cell Signaling	Cat# 5246S
JARID2 (D6M9X) Rabbit mAb	Cell Signaling	Cat# 13594T
Monoclonal Anti-HA antibody produced in mouse	Sigma-Aldrich	Cat# H9658
c-Myc Antibody (9E10)	Santa Cruz Biotechnology	Cat# sc-40
Bacterial and Yeast Strains
Rosetta 2(DE3)	Novagen	Cat# 71402
S. cerevisiae CB010 strain (MATa pep4∷HIS3 prb1 ∷LEU2 prc∷HISG can1 ade2 trp 1 ura3 his3 leu2–3,112)	(Jiao and Liu, 2015)	N/A
Chemicals, Peptides, and Recombinant Proteins
Anti-FLAG M2 affinity gel	Sigma-Aldrich	Cat# A2220-5ML
Streptavidin Agarose	ThermoFisher	Cat# 20347
Strep-Tactin® Superflow® high capacity 50% suspension	iba-lifesciences	Cat# 2-1208-010
Pierce™ Avidin Agarose	ThermoFisher	Cat# 20219
Pierce™ Glutathione Agarose	ThermoFisher	Cat# 16101
HIS-Select® Nickel Affinity Gel	Sigma-Aldrich	Cat# P6611-25ML
d-Desthiobiotin	Sigma-Aldrich	Cat# D1411-1G
SuperSignal™ West Pico PLUS Chemiluminescent Substrate	ThermoFisher	Cat# 34580
SYBR Gold	ThermoFisher	Cat# S11494
Mouse IgG-Agarose	Sigma-Aldrich	Cat# A0919-2ML
H3 peptide (residues 1–19) ARTKQTARKSTGGKAPRKQ	Elim Biopharmaceuticals Inc.	N/A
Jarid2(147–165) Ac-LSKRKPKTEDFLTFLCLRG-NH2	Elim Biopharmaceuticals Inc	N/A
Histone H3 (1–21), Biotinylated ARTKQTARKSTGGKAPRKQLA - GGK(BIOTIN) - NH2	AnaSpec	Cat# AS-61702
[Lys(Me3)4] - Histone H3 (1–21) - GGK(Biotin), H3K4(Me3), biotin labeled ART - K(Me3) - QTARKSTGGKAPRKQLA - GGK(Biotin) - NH2	AnaSpec	Cat# AS-64192
Deposited Data
Crystal structure of S12R4	This study	PDB: 5WAK
Crystal structure of S12R4J2A2	This study	PDB: 5WAI
Original gel images	This study and Mendeley Data	http://dx.doi.org/10.17632/59z763frkw.1
Experimental Models: Cell Lines
HEK293T	ATCC	Cat# CRL-3216
Recombinant DNA
pCS2-Ezh2-Myc	This paper	N/A
pCS2-Flag-Eed	This paper	N/A
pCS2-His6-Eed	This paper	N/A
pCS2-HA-Rbbp4	This paper	N/A
pCS2-Suz12-Flag	This paper	N/A
pCS2-HA-Suz12-Flag	This paper	N/A
pCS2-HA-Jazf1 (1-129)-Suz12(93-739)-Flag	This paper	N/A
pCS2-ProteinA-3C-Aebp2	This paper	N/A
pCS2-ProteinA-3C-Aebp2(ΔC2B-H3K4D)	This paper	N/A
pCS2-ProteinA-3C-PHF19	This paper	N/A
pCS2-ProteinA-3C-PHF19(ΔRC)	This paper	N/A
pCS2-ProteinA-TEV-EPOP	This paper	N/A
pCS2-ProteinA-TEV-EPOP(ΔCT)	This paper	N/A
pCS2-ProteinA-3C-Jarid2	This paper	N/A
pCS2-ProteinA-3C-Jarid2(ΔTR-1)	This paper	N/A
pCS2-ProteinA-3C-Jarid2(ΔTR-2)	This paper	N/A
pGEX-4T1-TEV-Aebp2(407–503)	This paper	N/A
pET-28a-Sumo-Aebp2(407–517)	This paper	N/A
pET-28a-Sumo-Aebp2(407–498)	This paper	N/A
Software and Algorithms
Pymol	The PyMOL Molecular Graphics System, Schrödinger, LLC.	N/A
PHENIX	https://www.phenix-online.org	N/A
Coot	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/	N/A

CONTACT FOR REAGENT AND RESOURCE SHARING

Further inquiries and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Xin Liu (xin.liu@utsouthwestern.edu)

METHOD DETAILS

Protein expression and purification

Expression and purification of the Suz12-Rbbp4 (S12R4) binary complex

Human Rbbp4 and different truncated versions of Suz12 proteins were expressed in Sf9 insect cells utilizing the Bac-to-Bac baculovirus expression system (Invitrogen) according to the manufacturer’s instructions. To generate the different S12R4 complexes, Sf9 cells were co-infected with baculoviruses encoding His_6-Thrombin-tagged full-length Rbbp4 and Strep-tag II-tagged Suz12_76–545, Suz12_76–447, or Suz12_76–336, and were harvested 48 hours post infection. For purification, cell pellets were resuspended in the lysis buffer containing 50mM HEPES, pH 7.4, 150mM NaCl, 1mM PMSF, 0.1% NP40, 10% glycerol and 5mM 2-mercaptoethanol and lysed by sonication followed by ultracentrifugation. The S12R4 complexes were purified from the supernatant using Ni-NTA resin followed by Strep-Tactin resin (IBA). The complexes eluted from the Strep-Tactin column were pooled, concentrated and loaded onto a Superdex S200 preparative size exclusion chromatography column (GE Healthcare) in a buffer containing 20mM Tris-HCl, pH 8.0, 100mM NaCl, and 2mM DTT. The peak fractions were pooled, concentrated, flash-frozen in liquid nitrogen and stored at −80°C. Protein purity was assessed by SDS-PAGE.

Expression and purification of Ezh2-EED-Suz12(VEFS) ternary complex

Human Ezh2-Eed-Suz12(VEFS) ternary complex was expressed and purified essentially the same as previously described for the Chaetomium thermophilum PRC2 complex (Jiao and Liu, 2015). Briefly, the fused Ezh2 (residues1 to 746)-LVPRGS-Suz12 (residues 546 to 695) cDNA was tagged at the 5′ end with a TAP tag cassette composed of 2×Protein A and His₆ tags separated by a TEV protease cleavage site and subcloned into p416GAL1 vector. Full-length Eed (residues 1 to 441) was subcloned into a modified p416GAL1 vector, in which the selection marker was replaced by TRP1. To generate the Ezh2-Eed-Suz12(VEFS) ternary complex, S. cerevisiae BY4741 strain was co-transformed with the two plasmids and induced by galactose for 20–24 hours. The complex was purified by IgG column followed by size exclusion chromatography, and the purity was assessed by SDS-PAGE.

Expression and purification of the 4-subunit core PRC2 complex

The core PRC2 complex (Ezh2-Eed-Suz12-Rbbp4) was co-expressed in Sf9 insect cells utilizing the Bac-to-Bac baculovirus expression system (Invitrogen). These constructs include FLAG-Ezh2, His₆-FLAG-Suz12-TEV-Strep-tag II, His₆-FLAG-Eed-3C-Strep-tag II and His₆-Thrombin-Rbbp4. 48 hours after co-infection, Sf9 insect cells were harvested and washed once with PBS prior to freezing at −80°C. For purification, cell pellets were resuspended in lysis buffer containing 50mM HEPES, pH 7.4, 150mM NaCl, 1mM PMSF, 0.1% NP40, 5 mM 2-mercaptoethanol and 10% glycerol and lysed by sonication. The complex was captured by a Ni-NTA column, cleaved by TEV protease at 4°C for 12 hours, and loaded onto a Strep-Tactin column (IBA) for further purification. The protein complex eluted from the Strep-Tactin column was pooled, concentrated and loaded onto a superose 6 10/300 GL column (GE Healthcare) in a buffer containing 20mM Tris-HCl, pH 8.0, 100mM NaCl, and 2mM DTT. The peak fractions were pooled, concentrated and stored at –80°C.

Expression and purification of Aebp2_407–503

This truncated version of human Aebp2 was used for making S12R4J2A2 crystals. The cDNA sequence encoding human Aebp2_407–503 was subcloned into PGEX-4T1 vector and a TEV protease cleavage site was also inserted between GST tag and Aebp2. E.coli Rosetta 2(DE3) cells transformed with the expression plasmid were grown in LB medium at 37°C to an OD 600 of 0.8 and induced with 0.5mM isopropyl-β-d-thiogalactoside (IPTG) at 20°C for 16 hours. After cell harvesting, cell pellets were resuspended in GST binding buffer containing 50mM Tris-HCl, pH 8.0, 150mM NaCl, 10% glycerol, and 5mM DTT supplemented with 1mM PMSF and lysed by sonication. The cell lysate was clarified by ultracentrifugation and the supernatant was incubated with GST resin for 2 hours at 4°C with mi xing. The Aebp2 protein bound to GST resin was incubated with TEV protease overnight at 4°C. The flow-through fraction was collected, concentrated and loaded onto a gel filtration column equilibrated with 50mM Tris-HCl, pH 8.0, 150mM NaCl and 2 mM DTT.

Expression and purification of Aebp2_407–517 and Aebp2_407–498 and reconstitution of the S12R4A2 ternary complexes

Human Aebp2_407–517 and Aebp2_407–498 were both subcloned into a pET-28a-Sumo vector. His₆-SUMO-Aebp2_407–498 was grown in LB medium at 37°C to an OD ₆₀₀ of 0.8 and induced with 0.5mM IPTG at 20°C for 16 hours. The cells were lysed by sonication in lysis buffer containing 50mM Tris-HCl pH 8.0, 150mM NaCl, and 2mM 2-mercaptoethanol supplemented with 1mM PMSF. The protein was purified on a Ni-NTA column and the His₆-SUMO tag was removed by SUMO protease. Aebp2_407–498 was further purified on a size exclusion chromatography column in a buffer containing 20mM Tris-HCl, pH8.0, 100mM NaCl and 2mM DTT. The Suz12_76–545-Rbbp4-Aebp2_407–498 ternary complex was reconstituted by mixing Suz12_76–545-Rbbp4 with Aebp2_407–498 that was in excess and was purified by size exclusion chromatography.

His₆-SUMO-Aebp2_407–517 was grown in LB medium at 37°C to an OD600 of 0.8 and induced with 0.5mM IPTG at 37°C for 5 hours. Cell pellets were resuspended in a denaturing buffer containing 50mM Tris-HCl, pH 8.0, 150mM NaCl, 8M urea and 2mM 2-mercaptoethanol supplemented with 1mM PMSF and lysed by sonication. The clarified cell lysate was incubated with Ni-NTA for 1 hour at 4°C with mixing and the bound protein was eluted . His₆-SUMO-Aebp2_407–517 protein was refolded by dialysis against a refolding buffer containing no urea at 4°C for overnight. The purified Rbbp4-Suz12_76–545 binary complex was incubated with the refolded His₆-Sumo-Aebp2_407–517 in 20mM Tris-HCl, pH8.0, 150mM NaCl, 2mM DTT at 4°C for 1 hour. SUMO protease was then a dded to remove the His_6-SUMO tag from the resulting complexes. The reconstituted Suz12_76–545-Rbbp4-Aebp2_407–517 ternary complex was purified via size exclusion chromatography in a buffer containing 20mM Tris-HCl pH8.0, 100mM NaCl, 2mM DTT.

Expression and purification of PRC2 complexes in HEK293T cells

HEK293T cells (ATCC) were also used for expression and purification of wild-type and mutant core PRC2 complexes. Full-length Ezh2, Eed, Suz12 and Rbbp4 were subcloned in to a pCS2+ vector with the following modification. The Ezh2 cDNA was fused to a N-terminal Myc-tag by PCR and cloned into the pCS2+ vector. Full-length Eed was subcloned into the pCS2+ vector with an N-terminal His₆ tag. Suz12 cDNA was tagged at the 5′ end with a TAP tag cassette composed of a Protein A and a FLAG separated by a TEV protease site. The deletion mutant Suz12(Δ79-107) and the double mutant Suz12(F86A/F90A) were introduced by site-directed mutagenesis. pCS2+-HA-Rbbp4 vector carrying a human Rbbp4 ORF in frame with the HA tag. HEK293T cells were grown in DMEM supplemented with 10% (v/v) fetal bovine serum (FBS) and were transiently co-transfected with the desired combination of plasmids using Lipofectamine 2000. Cells were harvested 48 hours post-transfection, lysed in 50mM Tris-HCl, pH 7.5, 500mM NaCl, 10% Glycerol, 0.5% NP40 and 2mM DTT (supplemented with protease inhibitors) on ice for 30 min, and centrifuged at 17,000 × g for 30min. The core PRC2 complexes were captured on a rabbit IgG-agarose column. The protein-bound resin was washed stepwise by buffer A (50mM Tris-HCl, pH8.0, 500mM NaCl, 2.5mM DTT, 10% glycerol, and 0.1% NP40) and buffer B (50mM Tris-HCl, pH 8.0, 150mM NaCl, 2.5mM DTT, and 10% glycerol). Proteins were released from IgG-agarose column after incubation with TEV protease overnight. Protein purity was assessed by SDS-PAGE.

Crystallization and structure determination

For crystallization of the Suz12_76–545-Rbbp4 binary complex, the purified Suz12_76–545-Rbbp4 binary complex at 15 mg/ml was screened by sitting drop vapor diffusion method at 22°C. Crystals were refined by mixing 1 μl protein solution at 10 to 15 mg/ml with 1 μl crystallization well solution containing 15% polyethylene glycol (PEG) 3350, 100 mM ammonium citrate tribasic and 100 mM Tris-HCl, pH 8.6. Crystals were harvested after 2 or 3 days, cryoprotected by the crystallization buffer supplemented with 25% glycerol, and flash frozen in liquid nitrogen.

For crystallization of the Rbbp4-Suz12-Aebp2-Jarid2 heterotetrameric complex, the Suz1276-545-Rbbp4-Aebp2 ternary complex was reconstituted by incubating Rbbp4-Suz12_76–545 with a ~3 molar excess of Aebp2_407–503 for 1 hour on ice and the resulting ternary complex was loaded onto a Superdex S200 preparative size exclusion chromatography column (GE Healthcare) equilibrated with 20mM Tris-HCl, pH 8.0, 100mM NaCl, and 2mM DTT to remove the excess Aebp2_407–503. Prior to crystallization, Jarid2 peptide (residues 147–165; LSKRKPKTEDFLTFLCLRG) was added to the reconstituted ternary complex in a molar ratio of 1:5. Crystals were grown by mixing 1 μl protein solution at 10 to 15 mg/ml with 1 μl crystallization well solution containing 10% PEG6000, 5%v/v(+/−)-2-Methyl-2,4-pentaneediol (MPD) and 100mM HEPES pH 7.5. Crystals were cryoprotected in a crystallization buffer containing 25% glycerol and subsequently flash frozen in liquid nitrogen.

Native X-ray diffraction data were collected at Advanced Photo Source (APS) beamline 19ID, Stanford Synchrotron Radiation Lightsource (SSRL) beamlines 9-2 and 12-2, and Advanced Light Source (ALS) beamline 5.0.2. Reflection data were indexed, integrated, and scaled with HKL2000 package and further processed with the CCP4 suite of programs (Otwinowski and Minor, 1997; Winn et al., 2011). The Suz12_76–545-Rbbp4 structure was determined in P2₁2₁2₁ space group, with one complex in an asymmetric unit. The Rbbp4-Suz12-Aebp2-Jarid2 heterotetrameric complex was crystallized in space group P2₁2₁2₁, with two complexes in an asymmetric unit. The structures were solved by molecular replacement using Phaser with a previously determined crystal structure of Rbbp4 as a search model (McCoy et al., 2007). Iterative model building and structure refinement were performed using Coot, Phenix, and autoBUSTER (Bricogne G, 2010; Emsley et al., 2010). Statistics for data collection, phase calculation, and refinement are summarized in Table 1. Structure images were rendered using PyMOL software (The PyMOL Molecular Graphics System).

Nucleosome Assembly

Mononucleosomes were reconstituted as previously described (Luger et al., 1999). Briefly, histone octamers and 147-bp ‘601’ DNA or 216-bp ‘601’ DNA were mixed together in a buffer containing 2M NaCl, 10mM Tris-HCl, pH 8.0, 0.1mM EDTA and 1mM 2-mercaptoethanol, and nucleosomes were reconstituted by the salt dialysis method. The resulting 147-bp and 216-bp ‘601’ nucleosomes were further purified by Model 491 Prep cell (Bio-Rad), from which the fractions containing nucleosomes were pooled and concentrated by using an Amicon 10 kDa cut-off concentrator. The assembly of the tailless H3-containing mononucleosomes followed the same procedure.

Eletrophoretic Mobility Shift Assay

For nucleosome binding assay, the wild-type and mutant Jarid2_147–165 peptides were used for the heterotetrameric complex (Rbbp4-Suz12-Aebp2-Jarid2) reconstitution. The mononucleosomes at the concentration of 20nM were incubated with 0–10 μM wild-type or mutant Rbbp4-Suz12-Aebp2-Jarid2 heterotetrameric complex in a 10μl binding reaction in a binding buffer containing 10mM Tris-HCl, pH 8.0, 50mM NaCl, 10% glycerol for 1 hour at 4°C. Binding reactions were analyzed by electrophoresis at 100V for 60 min on a 4% native polyacrylamide gel (Acrylamide/Bis 60:1) in 1× TGE buffer (25mM Tris-HCl, 190mM glycine, 1mM EDTA at pH 8.3). Thee gel was stained by SYBR Gold (Molecular Probes). The EMSA for the core PRC2 complex was performed under the same condition.

Isothermal Titration Calorimetry (ITC)

Isothermal titration calorimetry (ITC) experiments were carried out using a VP-ITC Microcal calorimeter (Microcal, Northhampton, MA, USA) at 25°C. The Rbbp4 protein and the Suz12_76–545-Rbbp4 or Suz12_76–545-Rbbp4-Aebp2_407–503 protein complex were dialyzed against ITC buffer (20mM Tris-HCl, pH 8.0, 150mM NaCl and 2mM DTT) overnight before the titration. The synthetic histone H3 peptide (residues 1–19; ARTKQTARKSTGGKAPRKQ) was resuspended in the same ITC buffer. During titration, 0.5 mM H3 peptide in the calorimeter injection syringe were delivered as a series of 2 μL injections every 2 min to the calorimetric cell containing 50μM Rbbp4, Suz12_76–545-Rbbp4 or Suz12_76–545-Rbbp4-Aebp2_407–503. The thermodynamic binding parameters were estimated by using a one-site binding model with a stoichiometry of 1:1.

Pull-down assays

GST pull-down assay

GST pull-down assays were carried out at 4 °C in the binding buffer containing 50mM Tris-HCl, pH 8.0, 150mM NaCl, 2mM DTT and 0.1% NP40. Briefly, different truncated GST-tagged Aebp2 or Jarid2 or GST alone (control) were pre-incubated with GST beads, and then mixed with the Suz12_76–545-Rbbp4 binary complex in the binding buffer for 1 hour at 4°C with mixing. After extensive wash with the same binding buffer, the supernatants were removed and the beads were boiled with SDS loading dye and subjected to SDS-PAGE analysis.

FLAG pull-down assay

FLAG pull-down assays were performed at 4 °C in the binding buffer containing 50mM Tris-HCl, pH 8.0, 150M NaCl, 2mM DTT and 0.1% NP40. Briefly, Jarid2_119–232-FLAG was expressed and purified from Rosetta 2 (DE3) cells. This protein was then pre-incubated with Anti-FLAG beads (Sigma) and mixed with the Suz12_76–545-Rbbp4 or Suz12_76–447-Rbbp4 binary complex in the binding buffer for 1 hour at 4°C with mixing. The b ound proteins were eluted by FLAG peptide and analyzed by SDS-PAGE.

Histone H3 and Jarid2 peptide pull-down assay

The following protein or protein complexes were used for H3 peptide pull-down assay: Rbbp4, Suz12_76–545-Rbbp4, Suz12_76–545-Rbbp4-Aebp2_407–503, Suz12_76–545-Rbbp4-Aebp2_407–498, Suz12_76–545-Rbbp4-Aebp2_407–517; The following protein complexes were used for Jarid2_147–165 peptide pull-down assay: the 4-subunit core PRC2 complex containing Suz12(WT), Suz12(F86A/F90A), or Suz12(ΔN79–107). Biotinylated histone H31-21 (ARTKQTARKSTGGKAPRKQLA) or Jarid2_147–165 (LSKRKPKTEDFLTFLCLRG) was pre-incubated with avidin resin beads (Thermo Fisher), and then mixed with the different proteins listed above in the binding buffer containing 50mM Tris-HCl, pH 8.0, 150mM NaCl, 2mM DTT and 0.1% NP40 for 1 hour at 4°C with mixing. The supernatant s were removed before the avidin beads were boiled with SDS loading for analysis by SDS-PAGE stained by Coomassie blue or Western blot.

HeLa nuclear extract pull-down assay

HeLa nuclear extract was prepared following a standard protocol. 5μg C-terminal biotinylated histone H3K4 or H3K4me3 peptide (AnaSpec) was pre-bound to streptavidin beads for 2 hours and the excessive unbound peptides were removed. The beads were then incubated with 200μl a total of 1mg HeLa nuclear extract for 1 hour and subjected to wash for 5 times. The bound proteins were released by boiling the beads in 1X SDS loading dye and were detected by Western blot.

Coimmunoprecipitation assay in HEK293T cells

Binding of accessory subunits to the PRC2 core complex

The full-length and mutated accessory subunits of interest were co-expressed with the four core subunits of PRC2 ectopically in HEK293T cells. All genes were cloned into the pCS2+ vector for overexpression under a CMV IE94 promoter. Each of the core subunits contained an epitope tag as indicated. The accessory subunits were expressed as protein A fusion proteins. HEK293T cells were transiently transfected with the corresponding plasmids. The accessory subunits were captured by IgG beads and washed in a buffer containing 50mM Tris-HCl, pH 8.0, 150mM NaCl, 2mM DTT, 10% Glycerol and 0.1% NP40. The bound core subunits of PRC2 were released by TEV or 3C protease and subjected to quantification on a Coomassie blue-stained SDS PAGE gel. In the case for Jarid2, the bound core subunits of PRC2 were examined by Western blot.

Binding of Jarid2 and EPOP to the Suz12 chromosomal translocation product

the binding assay was performed essential the same as above. The integrity of the core PRC2 complex containing Suz12 or JAZF1-Suz12 was first examined by Co-IP using an anti-FLAG antibody with HEK293T cells that co-expressed the corresponding four subunits of PRC2. To assess the binding of Jarid2, wild-type Suz12 and the JAZF1-Suz12 fusion protein were expressed together Jarid2 and the other three core subunits of PRC2 such that Suz12 and JAZF1-Suz12 were expressed in a similar level. Jarid2 was fused to an N-terminal protein A tag followed by a 3C protease cleavage site. The assay was done by capturing Jarid2 with IgG beads and the bound PRC2 subunits were released by 3C protease. The released Suz12 and JAZF1-Suz12 were examined by Western blot using an anti-FLAG antibody. EPOP binding to JAZF1-Suz12 was examined essentially by the same Co-IP experiment, except that TEV was used to release EPOP and the bound PRC2 from IgG beads.

QUANTIFICATION AND STATISTICAL ANALYSIS

Graphs display the mean ± SD from three independently performed binding assays. GraphPad Prism 6.0 was used for statistical analysis.

DATA AND SOFTWARE AVAILABILITY

The accession numbers for the coordinates and structure factors in this study are PDB: 5WAK (S12R4) and 5WAI (S12R4J2A2). Original gel images are uploaded to Mendeley Data: http://dx.doi.org/10.17632/59z763frkw.1