Cryo-EM structure of human SAGA transcriptional coactivator complex

Dear Transcription initiation is an essential regulatory step of gene expression that requires coordinated functions of transcription factors and chromatin regulators to provide accessible chromatin environment for the assembly of transcription pre-initiation complex. In this process, the transcription regulatory proteins are often organized into a number of modular and multi-functional coactivator complexes to orchestrate their different activities. Evolutionarily conserved with yeast SAGA (Spt – Ada – Gcn5 acetyltransferase) complex, human SAGA (abbreviated as hSAGA) complex is a well-established transcription coactivator that regulates the transcription of numerous genes 1 . Several distinct functional modules have been characterized for the 1.4-MDa SAGA complex, including a histone acetyltransferase (HAT) module and a histone deubiquitinase (DUB) module that establish histone modi ﬁ cation signatures for transcription activation, the largest subunit TRRAP that directly interacts with transcription activators, the core module that serves as a structural scaffold, and the spliceosome U2 snRNP factors SF3B3 and SF3B5 1,2 (Supplementary Fig. S1). Recent cryo-electron microscopy (cryo-EM) structures of yeast SAGA complex elucidated the structural organizations of the SAGA modules and the mechanism of TATA box-binding protein (TBP) recruitment and delivery for the nucleation of pre-initiation complex 3,4 . However, TBP was barely identi ﬁ ed in the interactome of hSAGA complex 5,6 ,

150 mM NaCl, 12 mM MgCl2, 10% glycerol, 5 mM 2-mercaptoethanol, 1 mM PMSF) and the protein complex was eluted in wash buffer supplemented with 500 ng/µl Flag peptides (WHSTbio). The eluent was then incubated with Strep-Tactin XT Superflow high-capacity resin (IBA Lifesciences) overnight. The resin was washed extensively and the complex was eluted in wash buffer containing 50 mM D-biotin (YEASEN). The hSAGA complex was concentrated to about 0.6 mg/ml and was used immediately.

Sample preparation and cryo-EM data collection
Purified hSAGA complex was incubated with 0.1% glutaraldehyde for 20 minutes on ice and was then purified using the method of GraFix 1 . In brief, a continuous 10-50% gradient of glycerol with a 0-0.1% gradient of glutaraldehyde was generated on a BioComp gradient master.
Then the hSAGA complex was loaded on top of the glycerol gradient in the buffer of 20 mM HEPES pH 7.5, 150 mM NaCl, 5 mM MgCl2 and 1 mM tris (2-carboxyethyl) phosphine (TCEP), and the ultracentrifugation was performed at the speed of 37,000 rpm at 4 ºC for 15 hours using a SW 60 Ti rotor (Beckman). The Grafix fractions containing the hSAGA complex was collected and the cross-linking was quenched by 50 mM Tris-HCl pH 7.5. The samples were then dialyzed against 25 mM HEPES pH 7.5, 150 mM NaCl and 1 mM DTT, and were concentrated to about 0.4 mg/ml. 2 μl aliquots of the samples were applied to lacey-carbon grids with an ultra-thin carbon film (Ted Pella) preprocessed with glow-discharging at 0.42 mbar and 25 mA for 25 s, and the grids were blotted for 3 s and were then plunged into liquid ethane cooled by liquid nitrogen using a Vitrobot Mark IV (FEI).
Cryo-EM images were acquired on a Titan Krios transmission electron microscope (FEI) operated at 300 kV. The images were collected on a K3 direct electron detector (Gatan) with a pixel size of 1.10 Å. Automated image acquisition was performed with EPU software (FEI). The dataset was collected at a defocus varying between -0.8 to -3.0 µm, and each micrograph was dose-fractioned to 32 frames with 0.1-s exposure time for each frame. The total accumulated dose of each micrograph is 50.0 e -/Å 2 . The imaging conditions were also listed in Table S1.

Image processing and model building
A total of 5,036 cryo-EM images of hSAGA complex were collected on a K3 detector, and motion correction was performed on the dose-fractioned image stacks using MotionCor2 with dose weighting 2,3 . The contrast-transfer function (CTF) parameters of each image were determined with Gctf 4 . Particle picking, 2D classification, 3D initial model, 3D classification, 3D auto-refine, CTF refinement, and Bayesian polishing were performed with RELION-3.1 5 . We manually picked about 1000 particles to generate the templates for particle auto-picking using Gautomatch (http://www.mrc-lmb.cam.ac.uk/kzhang/Gautomatch/) for further processing. An overview of the data processing procedure was shown in Supplementary Fig. S2a. After two rounds of 2D classification and two rounds of 3D classification, a total of 378,168 particles that belong to the hSAGA complex were processed with 3D auto-refinement and solvent-masked post-processing, which generated a cryo-EM density map with an overall resolution of 3.7 Å. To improve the map densities of TRRAP, the core module, and the SF3B3/5 module, the particles were further processed through focused 3D refinement and masked 3D classifications with partial signal subtraction, respectively 6 . The resolution estimation was based on the goldstandard Fourier shell correlation (FSC) 0.143 criterion and the local resolution was estimated with ResMap 7,8 .
Model building was carried out by fitting the available structures of yeast SAGA (PDB codes: 6T9I, 6TBM and 6MZD) in the electron microscopy density map of the hSAGA complex using UCSF Chimera 9 . The model was then manually built in Coot and real-space-refined with secondary structure restraints in Phenix 10,11 .

Cross-linking mass spectrometry
Samples for cross-linking mass spectrometry were prepared in the same way as those for cryo-EM. The purified hSAGA complex was cross-linked by bis (sulfosuccinimidyl) suberate (BS 3 ) (Thermo Fisher) with 1:1 mass ratio at room temperature for 1 hour. 50 mM Tris-HCl pH 7.5 was used to terminate the reaction after incubation. Cross-linked complexes were precipitated with cooled acetone and were dried by using a Speedvac. The pellet was dissolved in 8 M Urea, 100 mM Tris-HCl pH 8.5, followed by TCEP reduction, iodoacetamide alkylation, and overnight trypsin digestion. Tryptic peptides were desalted with MonoSpin TM C18 column (GL Science, Tokyo, Japan) and analyzed by a home-made 30 cm-long pulled-tip analytical column (75 µm ID packed with ReproSil-Pur C18-AQ 1.9 µm resin, Dr. Maisch GmbH, Germany) coupled to an EASY-nLC 1200 nano HPLC (Thermo Scientific, San Jose, CA). The peptide mixture was separated by applying a 120-minutes step-wise gradient of 5-100% buffer B (80% acetonitrile (ACN) in 0.1% formic acid) at a flow rate of 300 nl/min. Peptides eluted from the LC column were directly electrosprayed into the mass spectrometer and analyzed in positive mode using a data-dependent top-20 acquisition methods. Peptides containing the isopeptide bonds were identified and evaluated using pLink2 software 12 . Protein-protein cross-links were filtered with 5%FDR and plotted using xVis 13 .