Assembling a COMPASS

Abstract

Post-translational modifications of histone proteins lie at the heart of the epigenetic landscape in the cell’s nucleus and the precise regulation of gene expression. A myriad of studies have showed that several histone-modifying enzymes are controlled by modulatory protein partner subunits and other post-transcriptional modifications deposited in the vicinity of the targeted site. All together, these mechanisms create an intricate network of interactions that regulate enzymatic activities and ultimately control the site-specific deposition of covalent modifications. In this Point-of-View, we discuss our evolving understanding on the assembly and architecture of histone H3 Lys-4 (H3K4) methyltransferase COMPASS complexes and the techniques that progressively allowed us to better define the molecular basis of complex formation and function. We further briefly discuss some of the challenges lying ahead and additional approaches required to understand mechanistic details for the function of such complexes.

Keywords: :

• H3K4 methylation
• chromatin modification
• COMPASS
• SET
• structural biology
• RbBP5
• WDR5
• ASH2

Introduction

In the eukaryotic cell’s nucleus, chromatin is the host of several complex regulatory processes associated with DNA transcription, replication, and repair. Studies in the past 20 years have shown that several evolutionary conserved protein assemblies regulate these events, including the displacement of histone octamers as well as the covalent modifications of histone tails. These same studies highlighted that intricate networks of intermolecular interactions at the core of these protein assemblies often determine their subunits’ activities. However, owing to the exquisite requirement of each individual subunit to stably interact with other components of the complex, our understanding of the biochemical and structural determinants underlying the assembly of these protein complexes has lagged behind. Here we provide a brief historical perspective on identifying the structural elements controlling the formation of the histone H3K4 methylase Set1/COMPASS and its human homolog MLL1/COMPASS-like complexes, and we briefly present the challenges that lie ahead to fully characterize their function.

What is COMPASS

In the early 2000s, five laboratories independently characterized the protein SET1,^1-5 a protein sharing high sequence homology with trithorax, an important regulator of homeotic genes in drosophila.^6,7 Initial purifications by the Shilatifard laboratory revealed that at least seven different polypeptides formed a complex associated with SET1 (COMPASS).⁴ While the substrate of SET1 was still obscure at the time, the presence of a Su(var)3–9, Enhancer of Zeste and Trithorax (SET) domain in its C-terminus suggested that SET1 was able to methylate lysine residues. Shortly after, it was demonstrated that Set1/COMPASS was capable of methylating Lys-4 of histone H3, suggesting that SET1 was a genuine histone H3 Lys-4 (H3K4) methyltransferase.^8,9 Combined with genome-wide studies, these findings attracted a great deal of attention since H3K4 mono- and di-methylation are now commonly known to be enriched in transcriptional enhancer elements at active and potentially active genes, respectively, while trimethylation of the same residue is a mark associated with the transcriptional start sites of actively transcribed genes.^10,11

Additional analysis of the COMPASS complex revealed that several of its protein subunits (Cps) were evolutionary conserved. Cps60 (also referred to as Bre2) shared high sequence homology with drosophila ASH2, a trithorax protein required for the controls of genes important for disc patterning in flies¹² (Table 1). The mammalian homolog of Cps40 (also referred to as Spp1), Cfp1, was found to be homologous to a protein now known to bind unmethylated CpG islands^13,14 and to prevent spurious H3K4 trimethylation in embryonic stem cells.¹⁵ Initially identified as interacting proteins in a yeast two-hybrid system,¹⁶ the WD40-motif containing proteins Cps50 and Cps30 (also referred to as Swd1 and Swd3) were found to share high sequence homology with RBQ-3, a protein identified to bind Retinoblastoma (Rb),¹⁷ and a 7-bladed β-propeller protein named BIG-3. RBQ-3 and BIG-3 are now commonly referred as RbBP5 and WDR5, respectively. Another predicted β-propeller protein, Cps35 (also referred to as Swd2), was found to be an essential gene in budding yeast and link H3K4 methylation to H2B ubiquitylation. It was later established that Cps35 is homologous to WDR82 in mammals.^18,19 One of the smallest subunit of the complex, Cps25 (also referred to as Sdc1), shared high sequence homology with DPY30 which was initially found to play a role in dosage compensation of gene expression on the X-chromosome in C. elegans^20,21 (Table 1). Finally, Roguev et al. identified Shg1 as a subunit of COMPASS, however no mammalian homologs of this protein has been identified thus far⁸ and the loss of Shg1 has no effect on COMPASS stability or the pattern of H3K4 methylation. Along with studies on mammalian COMPASS-like complexes,^22-27 these findings illustrated the complexity of these multi-subunit protein machineries and established the foundation to perform detailed molecular analysis on the role of each subunit in the assembly and function of COMPASS.

Table 1. Budding yeast COMPASS subunits and their corresponding homologs in mammals

Budding yeast	Mammals
Set1	SET1A/B, MLL1–4
Cps60 (Bre2)	ASH2L
Cps50 (Swd1)	RbBP5 (RBQ-3)
Cps40 (Spp1)	CFP1 (CXXC1)
Cps35 (Swd2)	WDR82
Cps30 (Swd3)	WDR5 (BIG-3)
Cps25 (Sdc1)	DPY30
Shg1	-

Subunit Deletion: The First Approach to Dissect Protein Complex Formation

Budding yeast is a powerful model system to study evolutionary conserved protein complexes.²⁸ In fact, non-lethal homozygous deletion strains are commercially available and can be tailored, when combined with tandem affinity purification (TAP) tagging approaches, to verify protein complex formation. Using a TAP-tagging approach in combination with strains harboring genetic deletion of COMPASS components, Roguev et al. determined that Cps30 and Cps50 remained bound upon deletion of S. cerevisiae (Sc) SET1 but lost interaction with the other COMPASS subunits, suggesting that these two β-propeller domain containing proteins formed a heterodimer.⁸ Subsequent studies confirmed that this heterodimer was critical for maintaining global levels of H3K4 mono- di- and tri-methylation in vivo,^24,29-31 demonstrating the importance of the global organization of the ScSET1 complex for its function. Similar to the Cps30-Cps50 heterodimer, deletion of ScSET1 resulted in the dissociation of a Cps60-Cps25 heterodimer,⁸ thus supporting previous genome wide yeast two-hybrid studies showing that Cps60 and Cps25 stably associated in budding yeast.^16,32 These in vivo findings were also recently supported by three independent studies with the human homologs, showing that ASH2L (Cps60) and DPY30 (Cps25) interact independently of MLL1.^33-35 Finally, recent studies in yeast COMPASS demonstrated that histone H2B monoubiquitination regulates the incorporation of Cps35 in COMPASS.³⁶ Overall, these initial deletion studies provided the first insights into the assembly of COMPASS.

Protein Complex Reconstitution: A second Challenging Step

Studying protein complex assembly by homozygous deletion in budding yeast, although informative, can be hampered by compensatory mechanisms or the instability of the dissociated proteins. This is exemplified by the deletion of Cps50 and Cps30 which decreases the overall level of ScSET1.^2,29,31,37 To tackle these problems, in vitro reconstitution of protein complexes with purified components from a homologous or heterologous system, such as baculovirus-infected insect cells, have proven to be well-suited approaches.

In human, the first lines of evidence assigning roles of MLL1 regulatory subunits originated from the Roeder laboratory.²³ Using recombinant baculovirus expressing a FLAG-tagged WDR5, Roeder and coworkers reconstituted and purified the WDR5-ASH2L-RbBP5-MLL1 complex. In addition, using combinations of recombinant baculoviruses, they determined the hierarchical organization of the MLL1 complex, showing that RbBP5 directly interacts with both ASH2L and WDR5. The Cosgrove group further delineated the MLL1 complex assembly using analytical ultracentrifugation analysis. By iterative addition of each subunit, they confirmed the directionality of the interactions as MLL1-WDR5-RbBP5-ASH2L-DPY30 (MLL1-WRAD).³⁵ Utilizing sequence alignment, isothermal titration calorimetry, and GST pull-down experiments with proteins harboring deletions, the Cosgrove and Kingston laboratories showed that WDR5 bound to a region preceding the SET domain of MLL1, which they coined as WDR5 INteracting (WIN) motif.³⁸ Using a similar line of experiments, the Wilson and Couture groups identified a region of RbBP5 bound by WDR5 [referred to as WDR5 Binding Motif (WBM)].^39,40 Mutations of WBM resulted in the loss of stimulation of MLL1 methyltransferase activity, probably through the disruption of the complex. Overall, by combining several biochemical approaches, these studies provided important information of the regions of each member allowing the assembly of COMPASS or COMPASS-like complexes. However the precise molecular determinants defining the interactions between the Cps subunits were still elusive.

X-ray Crystallography Helps Identifying Structural Elements Underpinning COMPASS Assembly

X-ray crystallography has traditionally been an invaluable method for obtaining high-resolution snapshots of macromolecular complexes and for delineating the structural principles of their assembly. Coupled with site-directed mutagenesis and binding assays, X-ray crystallography has also provided crucial mechanistic insights into the molecular basis for the formation of mammalian COMPASS-like complexes. Interestingly, the crystal structure of the MLL1 catalytic SET domain in complex with the product cofactor S-Adenosyl-L-homocysteine (AdoHcy) and a histone H3 peptide revealed that the MLL1 catalytic domain has a more open active site when compared with the structure of other SET domain methyltransferases which do not require the presence of protein cofactors for optimal activity (Fig. 1A).⁴¹ This open conformation, which might possibly impair optimal recognition of the substrate, gave rise to the hypothesis that members of the WRAD complex are required to orient regions involved in the binding of the peptide substrates or to position active site residues in an orientation amenable for catalysis.

Figure 1. COMPASS structure and network of interactions. (A) Crystal structure models of Cps domains. Note that SET938 was co-crystallized with an N-terminal H3 peptide (pdb:2W5Z), while WDR5 was co-crystallized with RbBP5 and WIN (SET1) peptides (pdb:2XL2). The model shown for the WD40 domain of Cps50/RbBP5 is from the homologous domain of WDR5 (pdb:2XL2). The SPRY domain crystal structure shown is the one determined from human ASH2L (pdb:3TOJ). The arrows indicate interactions between Cps subunits as revealed by biochemical and structural studies. (B) Cryo-EM 3D map of yeast Set1-Cps50-Cps30- Cps60-Cps25 complex showing the relative location of Cps subunits as suggested by EM and biochemical studies.

Several co-crystal structures of WDR5 with peptides from its interacting partners, MLL1 and RbBP5, highlighted the importance of the β-propeller protein in the formation of the MLL1 complex. The crystal structure of WDR5 in complex with the WIN motif of MLL1 revealed that a conserved arginine residue (R3765) of MLL1 is positioned between two key phenylalanine residues (F133 and F263) of the WDR5 peptidyl arginine-binding cleft.^42,43 Recently, the crystal structures of WDR5 in complex with different WIN regions showed that the WDR5 peptidyl arginine binding cleft enables the binding of the WIN motifs of other members of the SET1 family of methyltransferases including MLL2–4 and SET1A/B (Fig. 1A).^44,45 Additional structural studies revealed that WDR5 binds RbBP5 via a V-shaped cleft located on the opposite side of the arginine binding cleft.^39,40 The same studies also demonstrated that a valine-aspartate-valine (VDV) motif of RbBP5 is necessary for binding to WDR5. Identification of WDR5 peptide binding pockets confirmed the initial findings showing that WDR5 plays a pivotal role in scaffolding the MLL1 complex by bridging together MLL1 with the RbBP5-ASH2L regulatory subunits. Finally, the crystal structure of the SPRY domain of ASH2L revealed that several arginine and lysine residues form a positively charged surface important for binding a cluster of acidic residues on RbBP5 (Fig. 1A).³⁴ Overall, these structural studies have provided detailed information regarding the specificity of MLL1 complex formation and offered a better understanding of the regulation of MLL1 by the core complex Cps members. However, given that the recent structural studies have focused on protein/peptide complexes, the full extent of the interactions underlying the formation of the WRAD complex and its interaction with SET1 members has remained unclear. Consequently, the crystal structures reported so far may have not shown the entire spectrum of interactions controlling MLL1 complex assembly. This fact is well illustrated by recent observations by the Cosgrove laboratory showing that WDR5 binds a construct encompassing the WIN and SET domains of MLL1 with ~14-fold more affinity compared with a peptide corresponding to the WIN motif.³⁵ This finding stresses the needs to consider full-length subunits of SET1 complexes in structural studies, either using X-ray crystallography or other structure determination techniques.

Electron Microscopy and the 3D Puzzle of COMPASS Complexes

Single particle electron microscopy (EM) analysis, applied on negative stained or vitrified specimens (cryo-EM), is a powerful approach for characterizing the 3D architecture of multi-subunit complexes, albeit at typically lower resolution compared with X-ray crystallography or NMR. The application of this methodology recently started shedding light on the architecture of in vitro reconstituted yeast and human COMPASS complexes that included the methyltransferase C-terminal SET domain of Set1/MLL, Cps60/ASH2L, Cps50/RbBP5, Cps30/WDR5 and Cps25/DPY30. These subunits are all present in Set1 and MLL COMPASS complexes and were been shown to form a catalytically fully active complex.⁴⁶ By combining 3D cryo-EM reconstructions with 2D negative stain analysis of individual sub-complexes, Takahashi et al. showed that yeast core COMPASS adopts a Y-shaped architecture with Cps50 and Cps30 localizing on the top two adjacent lobes and Cps60-Cps25 forming the base at the bottom (Fig. 1B).⁴⁶ Fab labeling experiments localized the SET domain of Set1 in a region adjacent to a central channel at the juncture of Cps50, Cps30 and the Cps60-Cps25 module. Interestingly, the human core COMPASS-like complex revealed high architectural similarity with its yeast counterpart, further strengthening the notion that yeast is a representative model system to study the regulation of the SET1 family of methyltransferases.⁴⁶ Moreover, the observed subunit arrangement is compatible with the available X-ray crystallography data, and elegantly highlights the requirement of all these subunits for the regulation of mono-, di-, and tri-methylation. It is worth noting that the position of the SET domain, immediately adjacent to each component of core COMPASS, suggests a centralized organization that reflects the importance of the SET motif in catalysis, together with complex stabilization by its closely associated subunits.

Where to Go from Here

All of these studies have provided an initial framework for understanding the overall assembly of COMPASS complexes and the molecular basis of methylation activity. Building upon these findings, we need to expand our structural and functional understanding of COMPASS complexes by determining the crystal structures of full-length proteins alone and in complex with other COMPASS subunits. It is clear however that the application of these approaches will not be so forthcoming due to the biochemical challenges in obtaining full-length proteins and also stable complexes, both compositionally and conformationally. To this end it will be important to fully explore the use of heterologous expression systems, such as insect cells and mammalian cells, for the expression and purification of COMPASS proteins. Such protein preparation platforms will also allow the simultaneous co-expression of several Cps subunits, which might be key to obtain stable complexes for structural studies. Along the same lines, it might be necessary to probe the use of antibody fragments (Fabs), nanobodies,⁴⁷ or even chemical cross-linkers like glutaraldehyde in order to limit the conformational variability of the COMPASS complex⁴⁶ and render it more amenable to crystallization.

Obtaining more homogeneous preparations will also be necessary for determining cryo-EM structures at higher resolution, which will facilitate accurate docking of the X-ray structures in the EM envelopes of the full complex. Perhaps cryo-EM will be crucial for determining the structure of COMPASS in complex with the nucleosome, through which we will gain insights into the mechanisms underlying substrate recognition by the SET1 family of methyltransferases. In this direction, we also need to develop COMPASS activity assays that are based on nucleosomal preparations and not just histone peptides which, though informative, do not necessarily reflect a native substrate when isolated from the nucleosome structure.

But crystal structures or even high resolution EM densities may tell only one part of the COMPASS story, as recapitulated in static pictures of a protein entity that is by nature highly conformationally “active.” Another key to obtaining insights into dynamic aspects of COMPASS subunits and their assembly may lie in the application of more recently developed experimental approaches, such as deuterium exchange mass spectroscopy (DXMS),⁴⁸ electron paramagnetic resonance (EPR) spectroscopy,⁴⁹ and FRET measurements.⁵⁰ These methods can accurately monitor the conformational dynamics of protein regions at a range of resolutions and permit the spatiotemporal probing of changes associated with substrate binding and dissociation. Such enabling technologies can offer remarkable information on COMPASS conformational transitions that, combined with crystallographic or EM data, will start to delineate the molecular underpinnings of COMPASS function. Altogether, these studies should guide detailed mutagenesis work to probe the role of specific regions important for both directing the assembly and also for establishing the different levels of methylation capacities built into the protein complex.

Acknowledgments

We thank Ali Shilatifard and Raymond Trievel for valuable comments on this manuscript. J.-F.C. acknowledges an Early Research award (Ontario) and a Canada Research Chair in Structural Biology and Epigenetics. J.-F.C. is also supported by a grant from the Canadian Institutes of Health Research. G.S. is supported as a University of Michigan Biological Sciences Scholar and as a Pew Scholar of Biomedical Sciences.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.