Structure and Conformational Dynamics of a COMPASS Histone H3K4 Methyltransferase Complex.

Contact for Reagent and Resource Sharing

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Georgios Skiniotis ( yiorgo@stanford.edu ) Experimental Model and Subject Details For recombinant protein expression, we used Spodoptera frugiperda (Sf9) cells at 27 °C and E. coli strain Rosetta™ (DE3) pLysS at 18 °C, respectively. For yeast culture, all knock-out strains were haploid cells and purchased from the Yeast Knock-Out (YKO) collection (Dharmacon) except set1Δ and cps50Δ in the charge-swapping experiment, which were generated with a PCR-mediated gene disruption technique with HIS3 and LEU2 as selection marker, respectively. More details are described in the Method Details section.

Experimental Model and Subject Details

For recombinant protein expression, we used Spodoptera frugiperda (Sf9) cells at 27 °C and E. coli strain Rosetta™ (DE3) pLysS at 18 °C, respectively. For yeast culture, all knock-out strains were haploid cells and purchased from the Yeast Knock-Out (YKO) collection (Dharmacon) except set1Δ and cps50Δ in the charge-swapping experiment, which were generated with a PCR-mediated gene disruption technique with HIS3 and LEU2 as selection marker, respectively. More details are described in the Method Details section.

Method Details Expression, and purification of COMPASS Saccharomyces cerevisiae Set1 starting at Met762 to the C-terminal end (SET762), and full-length Cps60, Cps50, Cps40, Cps30, and Cps25, the components of the extended core COMPASS, were cloned into pBacPAK8 vector with FLAG or His6-FLAG tag on the N-terminus. Each baculovirus was individually prepared and amplified by using BacPAK baculovirus expression system (Clontech), and mixed in a predetermined ratio to achieve stoichiometric expression in cells. Sf9 insect cells were co-transfected at 27 °C for 72–96 h before collection, and lysed in 10 mM Tris-HCl pH 8.0, 1.5 mM MgCl2, 10 mM KCl, supplemented with protease inhibitor cocktail (P8340, Sigma-Aldrich). The supernatant was cleared with centrifugation, supplemented with NaCl to be 500 mM, mixed with anti-FLAG M2 affinity agarose gel (Sigma-Aldrich), and gently stirred for 2 h at 4 °C. COMPASS-bound M2 resin was collected with centrifugation, washed with FLAG buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 10 μM Zn(OAc) 2 10% glycerol), and loaded into a column. COMPASS was eluted with FLAG buffer supplemented with 0.4 mg/ml 3× FLAG peptide (ApexBio), concentrated with Amicon Ultra Centrifugal Filters (EMD Millipore), and loaded onto Superose 6 Increase 10/300 GL column pre-equilibrated with 50 mM Tris-HCl pH 8.0, 50 mM NaCl, 0.5 mM TCEP with AKTA pure chromatography system (GE Healthcare). Protein interaction assays Complementary DNA corresponding to COMPASS subunits of thermophilic filamentous fungus Chaetomium thermophilum (Ct) for CtCps60, CtCps40, and CtCps30, and Myceliophthola thermophila (Mt) for MtCps50 were cloned into pET28a or pGEX4T3 vectors. Truncations and mutants of each thermophilic COMPASS subunit were generated using a site-directed mutagenesis kit (stratagene). Wild-type and mutant proteins were overexpressed as TEV cleavable His or GST tagged proteins in the E. coli strain Rosetta™ (DE3) pLysS using 0.2mM IPTG during 17h at 18 °C. Cells were harvested in 50mM sodium phosphate, 500mM NaCl and 5mM β-mercaptoethanol, 10% glycerol, 1% Triton X-100, lysed by sonication and clarified by centrifugation. His-tagged proteins were purified by Talon Co 2+ affinity chromatography and following TEV cleavage, the proteins were further separated by size exclusion chromatography (Superdex 200). For size exclusion chromatography analysis, purified proteins were mixed at a 1:1.5 mole ratio of Cps50:Cps30, Cps40:Cps50 or Cps60:Cps50 the binding buffer (BB) composed of 50mM Tris pH8.0, 200mM NaCl, 5mM BME and 2% Glycerol for 2h at 4 °C. Complexes were separated by size exclusion chromatography using a Superdex 200 pre-equilibrated in the BB. For GST pull-down assays, cell lysates containing 2μg of GST or GST-tagged proteins were applied onto glutathione-sepharose beads during 1 hour and washed extensively using the BB supplemented with 0.05% Triton-X 100. These beads were then further incubated with 10μg of another COMPASS subunit in 1ml of BB-T buffer during 2h at 4 °C. Beads were collected by centrifugation (1000 rpm, 1min), and washed extensively with BB-T buffer, and subjected to elution using BB supplemented with 10mM reduced glutathione. The input and eluted samples were separated using SDS-PAGE and Coomassie staining.

Yeast strains

Yeast culture was performed using standard methods. All yeast stains used in this work were haploid cells. All knock-out strains were from the Yeast Knock-Out (YKO) collection (Dharmacon) except set1Δ and cps50Δ in the charge-swapping experiment, which were generated with a PCR-mediated gene disruption technique with HIS3 and LEU2 as selection marker, respectively. All truncation and point mutations of COMPASS subunits were generated with site-directed mutagenesis after cloning into pFA6a-kanMX6 (Set1, and Cps50 in the charge-swapping experiments) or pRS306 (Cps60, Cps50, and Cps30) vectors. Cps60, Cps50, and Cps30 were C-terminally FLAG-tagged. ADH1 terminator sequence was introduced after the stop codon of each cloned COMPASS subunit gene. Engineered COMPASS subunit genes were introduced into their original loci through homologous recombination with standard yeast transformation technique and G418 or uracil selection. The introduced gene sequences in the genome were verified with Sanger sequencing.

Western blotting analysis

Over-night grown yeast cultures were harvested, washed, suspended in nuclear isolation buffer (250 mM sucrose, 60 mM KCl, 12.5 mM NaCl, 5 mM MgCl2, 1mM CaCl2, 0.8% Triton X-100, supplemented with protease inhibitor cocktail (P8215, Sigma-Aldrich)), and lysed with glass-beads by using Mini-BeadBeater-1 (BioSpec) or multi-tube vortex mixer. Crude nuclei pellets were collected with centrifugation, mixed with SDS-Laemmli sample buffer, boiled, resolved by SDS-PAGE, transferred to nitrocellulose membrane, and probed with home-made anti-H3, anti-H3K4me1, anti-H3K4me2, anti-H3K4me3, and anti-Set1 antibodies. Crystallization and Structure Determination The WD repeat domain of Cps50 (mtCps50 1−347 ) was overexpressed in E.coli Rosetta cells (Novagen) as a TEV cleavable fusion protein with a hexa-histidine tag, induced by 0.2mM IPTG for 17 hours at 18°C. Cells were pooled and resuspended in 50mM sodium phosphate, 500mM NaCl, 5mM β-mercaptoethanol, 10% glycerol, 1% Triton X-100, lysed by sonication and clarified by centrifugation. Proteins were then purified by Talon Co 2+ affinity chromatography and cleaved subsequently by TEV. Following gel filtration on Superdex 200 pre-equilibrated with a buffer composed of 50mM Tris pH8.0, 200mM NaCl, 5mM β-mercaptoethanol and 2% glycerol, the protein was concentrated and mixed with a mother liquor solution composed of 200 mM sodium citrate, 22% (w/v) polyethylene glycol 3350. Cps50 crystals were harvested, transferred into fomblin and flash-frozen in liquid nitrogen. A full dataset was collected at the 17-ID beamline at the Argonne National Laboratory’s Advanced Photon Source and indexed using HKL-2000 (HKL Research) ( Otwinowski and Minor, 1997 ). A search model generated by I-TASSER ( Roy et al., 2010 ) was used to solve the structure of Cps50 by molecular replacement. Using PHASER, four molecules were placed in the asymmetric unit. Following several rounds of model building and refinement using Coot ( Emsley et al., 2010 ) and PHENIX ( Adams et al., 2010 ). The quality of the model was evaluated using Molprobity ( Chen et al., 2010 ). Final refinement statistics are provided in Table S1 . Cryo-EM data collection 3μl protein aliquotes with a concentration of ~1.5mg/ml supplemented with 0.05% Octyl β-D-glucopyranoside were applied onto glow-discharged Lacey carbon grids, blotted with Vitrobot Mark IV (FEI Company) and flash frozen in liquid ethane. The grids were imaged at liquid nitrogen temperature on a FEI Titan Krios electron microscope operating at 300 kV. Cryo-EM images were recorded with a Gatan K2 Summit direct electron detector at a nominal magnification of 29,000X in counted mode, corresponding to a pixel size of 1.0 Å/pixel. A dose rate of ~9.0 electrons/Å 2 /s and defocus values ranging within −1.5~−3.5 μm were used. Total exposure of 10s per image were dose-fractionated into 50 movie frames, resulting in an accumulated dose of ~90 electrons per Å 2 .

Cryo-EM image processing

Micrograph movie stacks were first subjected to MotionCorr2 ( Zheng et al., 2017 ) for whole-frame and local drift correction, with subsequent dose filtering applied to the integrated images. Visual inspection was conducted to discard micrographs with large carbon area, obvious ice contamination and visible astigmatism, resulting in a working data set of 8298 cryo-EM images. CTFFIND4 ( Rohou and Grigorieff, 2015 ) was used for contrast transfer function (CTF) parameter determination on binned 2X2 images. About 10,000 particles interactively picked with EMAN2 ( Tang et al., 2007 ) were subjected to ISAC ( Yang et al., 2012 ) for reference-free 2D classification, and averages were employed to reconstruct an ab initio 3D model by VIPER ( Penczek et al., 1994 ). Automatic particle picking, 2D classification and 3D classification were conducted in RELION1.4 ( Scheres, 2012 ). A total of ~1.5 million particles selected through 2D classification were subjected to a first round of 3D classification into eight classes ( Figure S2A ). Particles from classes with well-defined structural features from the first 3D classification of the two subsets were combined (731,504 particles) and submitted to a second round 3D classification producing eight classes. A partition of 163,539 particles combined from two classes were refined to 4.0-Å using cisTEM ( Grant et al., 2018 ). The angular distribution of this projections contributing to this reconstruction shown in Figure S2B . Another partition with 92,070 particles with a more compact conformation was refined to 4.4-Å resolution. Resolution estimations was obtained in cisTEM with gold standard Fourier shell correlation using the 0.143 cut-off criterion ( Figure S2C ) and local resolution of 4.0-Å map is estimated by Bsoft ( Figure S2D ). The two conformers and models were aligned in Chimera (Figures 1B and S2E ).

Model building and refinement

The two clearly resolved doughnut-like density modules leaning almost perpendicularly against each other represent the classical seven-blade WD40 domains from Cps30 and Cps50, which have ~30% sequence similarity. To correctly assign the WD40 domains, which would serve as a starting point to build the complete model, we considered the following: First, we observed a long winding density element extending from one of the doughnut lobes of the EM map, suggesting that it corresponds to the unstructured ~80 amino acid C-terminus of Cps50, whereas the entire Cps30 sequence folds into one single WD40 propeller domain. Second, this well-defined density could perfectly accommodate the reported structural fragments of RbBP5, the human homologue of yeast Cps50. Homology models generated by MODELLER ( Eswar et al., 2007 ) for Cps30 and Cps50 WD40 domains derived from WDR5 (PDB 2H14) and our mtCps50 WD40 domain crystal structure were fit into the EM density. To complete the Cps50 C-terminal tail modeling, two fragments (352–376 and 386–391aa) sharing high similarity with homologous RbBP5 in structures of MLL3-RbBP5-ASH2L (PDB ID: 5F6K) and WDR5-MLL1-RbBP5 (PDB ID: 3P4F) were fit with very good agreement into the EM density, which anchored the tracing of the remainder main chain. Models for Set1 SET domain, Cps60 SPRY domain and its C-terminal helix in complex with Cps25 C-terminal a-helical domain were also generated by MODELLER, using crystal structures MLL3-RbBP5-ASH2L (PDB ID: 5F6K), ASH2L-RbBP5 (PDB ID: 4X8P) and ASH2L-DPY30 (PDB ID: 4RIQ) as templates, respectively. For the n-SET region modeling, the secondary structure prediction analysis suggested a α-helix which matches the EM density. The subsequent Win-motif was built into well-resolved EM density guided from the WDR5-MLL1-RbBP5 complex structure (PDB ID: 3P4F). A rod-like density on top of the two WD40 domains ridges corresponds to the Cps40 protein ( Thornton et al., 2014 ). We built a model for scCps40 by combining secondary structure prediction in Psipred ( Buchan et al., 2013 ), 3D de novo model building in Robetta ( Kim et al., 2004 ) and I-TASSER ( Roy et al., 2010 ), and biochemical mutagenesis assays ( Figure 5 ). Given high flexibility, or local resolution not sufficient for de novo modeling, some regions of the subunits were not built in the final model: Cps40 residues 1–116 which containing a PHD domain; Cps60 regions 1–86, 115–141, 243–351, 372–432aa; Cps25 1–124aa; and Set1 segments 762–798 and 851–924aa. All models were docked into the cryo-EM density in Chimera ( Pettersen et al., 2004 ), and iteratively adjusted manually using Coot ( Emsley et al., 2010 ). The models were subjected to global refinement and minimization in real space, by using the module phenix.real space refine from PHENIX ( Adams et al., 2010 ) package with secondary restraints. The quality of the model was evaluated using Molprobity ( Chen et al., 2010 ). The final refinement statistics are provided in Table S2 . For model deposition, the side chains were stubbed in areas of low resolution unless supported by crystal structures.

Quantification and Statistical Analysis

No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment.

Data and Software Availability

The cryo-EM density map has been deposited in the Electron Microscopy Data Bank (EMDB; https://www.ebi.ac.uk/pdbe/emdb/ ): EMD-7303 and the coordinate for COMPASS has been deposited in the Protein Data Bank (PDB; https://www.rcsb.org/ ): 6BX3. The crystal structure of Myceliophthola thermophila Cps50 WD40 domain has been deposited in Protein Data Bank: 6E29.

Supplementary Material 1 Figure S1. COMPASS sample purity and cryo-EM. Related to Figure 1 . (A) Purified COMPASS sample evaluated by sliver staining SDS-PAGE after size exclusion chromatography. (B) Representative cryo-EM raw image. (C) Representative 2D class averages. 2 Figure S2. Cryo-EM data processing. Related to Figure 1 . (A) Data processing for the final 3D reconstruction of COMPASS core complex. Particle selection, 2D classification and 3D classification were conducted in RELION. Refinement of the two 3D volumes were performed with cisTEM. (B) Angular distribution for particle projections of the 4.0-Å map. (C) Gold Standard FSC curves for 4.0-Å and 4.4-Å maps, colored green and yellow. (D) Local resolution of 4.0-Å map is estimated by Bsoft. (E) Refined models against 4.0-Å (green) and 4.4-Å (yellow) maps are aligned on SET domain. 3 Figure S3. COMPASS model determination. Related to Figure 1 . (A) Crystal structure of Myceliophthora thermophila Cps50 WD40 domain. (B) Our de novo Cps40 model fits well into the EM density map. (C) Improved local EM densities using focused refinement of the projections used for the 4.0-Å map. (D) FSC curves for the final refined model versus the final cryo-EM map (full dataset, black), the refined model with a half map versus the same map (green), and the refined model with a half map versus the other half map (red). See also Supplemental Figure 2 . 4 Figure S4. Representative regions of cryo-EM density map and corresponding model for COMPASS. Related to Figures 1 and 2 . EM density map and model are shown for Cps40 elongated helices (pink), the Cps60 C-terminal helix (shown in blue) interacting with Cps25 dimeric helical stacks (orange and red), Cps30 β sheet (green), Cps50 β sheet (blue) and Cps50 C-terminal region (blue). Bulky side chains are indicated for each segment. 5 Figure S5. Interaction of Cps50 with Cps30, Cps40 and Cps60. Related to Figure 2 . (A, C and E) Co-elution upon size exclusion chromatography of Cps50 with Cps30 (A), Cps40 (C) and Cps60 (E). (B, D and F) Cps50 regions important for interactions are characterized by purified Cps50 fragments pulled down by GST-Cps30 (B), GST-Cps40 (D) and GST-Cps60 (F). The asterisk marks contaminants. 6 Figure S6. Mutations of Cps50 and Cps60 affecting COMPASS activity in vivo. Related to Figure 2 . (A) Sequence alignment among Cps50/RbBP5 homologues. (B) Charge reversal mutation of residues forming the interface between Cps50 and Cps60 alter COMPASS activity. 7 Figure S7. Structural comparison between COMPASS and fragmented MLLs. Related to Figure 6 . (A) Overlay of the SET-I interacting Motif (SIM) region of yeast Cps30, with its human homologous WDR5 protein (PDB 2H14). The beginning and ending residues of the SIM loops connecting the flanking β17 and β18 strands are highlighted with red dots. (B) Sequence alignment of the SIM region among Cps30 and its homologues. (C) Structural comparison of SET/Cps50/Cps60 in COMPASS (orange), MLL1mut/RbBP5/ASH2L (PDB 5F6L; green) and MLL3/RbBP5/ASH2L (PDB 5F6K; grey), focusing on the catalytic SET/MLL domain. (D) Zoom-in view of SET/MLL domains aligned in (C). (E) A schematic model of COMPASS recognizes a mono-nucleosome. 8 Table S1. Data collection, refinement statistics for mtCps50 WD40 domain. Related to Figure S3A . 9 Table S2. Data collection, refinement statistics for cryo-EM. Related to Figure 1 .