Analysis of Global and Site-Specific Radiation Damage in Cryo-EM.

KEY RESOURCES TABLE

REAGENT or RESOURCE SOURCE IDENTIFIER Chemicals, Peptides, and Recombinant Proteins Proteinase K Sigma-Aldrich Cat#P2308; CAS: 39450-01-6 Ammonium sulfate Sigma-Aldrich Cat#A4418; CAS: 7783-20-2 BIS-TRIS Sigma-Aldrich Cat#B9754; CAS: 6976-37-0 GSNQNNF GenScript N/A Polyethylene glycol 8,000 Hampton Research Cat#HR2-535; CAS: 25322-68-3 MES sodium salt Sigma-Aldrich Cat#M3885; CAS: 71119-23-8 Zinc acetate dihydrate Sigma-Aldrich Cat#96459; CAS: 5970-45-6 Deposited Data Atomic coordinates, proteinase K crystal structure ( Hattne et al., 2016 ) PDB: 5i9s Atomic coordinates and density map of proteinase K at 0.86 e − Å −2 this paper PDB: 6cl7; EMDB: EMD-7490 Atomic coordinates and density map of proteinase K at 2.6 e − Å −2 this paper PDB: 6cl8; EMDB: EMD-7491 Atomic coordinates and density map of proteinase K at 4.3 e − Å −2 this paper PDB: 6cl9; EMDB: EMD-7492 Atomic coordinates and density map of proteinase K at 6.0 e − Å −2 this paper PDB: 6cla; EMDB: EMD-7493 Atomic coordinates and density map of proteinase K at 7.8 e − Å −2 this paper PDB: 6clb; EMDB: EMD-7494 Atomic coordinates and density map of GSQNNF at 0.27 e − Å −2 this paper PDB: 6clc; EMDB: EMD-7495 Atomic coordinates and density map of GSQNNF at 0.81 e − Å −2 this paper PDB: 6cld; EMDB: EMD-7496 Atomic coordinates and density map of GSQNNF at 1.3 e − Å −2 this paper PDB: 6cle; EMDB: EMD-7497 Atomic coordinates and density map of GSQNNF at 1.9 e − Å −2 this paper PDB: 6clf; EMDB: EMD-7498 Atomic coordinates and density map of GSQNNF at 2.4 e − Å −2 this paper PDB: 6clg; EMDB: EMD-7499 Atomic coordinates and density map of GSQNNF at 2.9 e − Å −2 this paper PDB: 6clh; EMDB: EMD-7500 Atomic coordinates and density map of GSQNNF at 0.17 e − Å −2 this paper PDB: 6cli; EMDB: EMD-7501 Atomic coordinates and density map of GSQNNF at 0.50 e − Å −2 this paper PDB: 6clj; EMDB: EMD-7502 Atomic coordinates and density map of GSQNNF at 0.82 e − Å −2 this paper PDB: 6clk; EMDB: EMD-7503 Atomic coordinates and density map of GSQNNF at 1.2 e − Å −2 this paper PDB: 6cll; EMDB: EMD-7504 Atomic coordinates and density map of GSQNNF at 1.5 e − Å −2 this paper PDB: 6clm; EMDB: EMD-7505 Atomic coordinates and density map of GSQNNF at 1.8 e − Å −2 this paper PDB: 6cln; EMDB: EMD-7506 Atomic coordinates and density map of GSQNNF at 2.1 e − Å −2 this paper PDB: 6clo; EMDB: EMD-7507 Atomic coordinates and density map of GSQNNF at 2.5 e − Å −2 this paper PDB: 6clp; EMDB: EMD-7508 Atomic coordinates and density map of GSQNNF at 2.8 e − Å −2 this paper PDB: 6clq; EMDB: EMD-7509 Atomic coordinates and density map of GSQNNF at 3.1 e − Å −2 this paper PDB: 6clr; EMDB: EMD-7510 Atomic coordinates and density map of GSQNNF at 3.4 e − Å −2 this paper PDB: 6cls; EMDB: EMD-7511 Atomic coordinates and density map of GSQNNF at 3.8 e − Å −2 this paper PDB: 6clt; EMDB: EMD-7512 Software and Algorithms TVIPS tools ( Hattne et al., 2015 ) N/A iMosflm ( Leslie and Powell, 2007 ; Battye et al., 2011 ) RRID:SCR_014217 AIMLESS ( Evans and Murshudov, 2013 ) RRID:SCR_015747 MOLREP ( Vagin and Teplyakov, 1997 ) RRID:SCR_007255 XDS ( Kabsch, 2010b ) RRID:SCR_015652 XSCALE ( Kabsch, 2010a ) RRID:SCR_015652 XDSCONV ( Kabsch, 2010b ) RRID:SCR_015652 SHELXD ( Sheldrick, 2008 ) RRID:SCR_014220 SCALEIT ( Howell and Smith, 1992 ) RRID:SCR_007255 EFRESOL ( Urzhumtseva et al., 2013 ) N/A REFMAC ( Murshudov et al., 2011 ) RRID:SCR_014225 AREAIMOL ( Winn et al., 2011 ) RRID:SCR_007255 RIDL ( Bury et al., 2015 ) N/A PyMol ( Schrödinger, 2014 ) RRID:SCR_000305 Other TEM grids Quantifoil N/A easiGlow glow discharge cleaning system PELCO N/A Vitrobot Mark IV plunge-freezer Thermo Fisher N/A Gatan 626 cryo-transfer holder Gatan N/A FEI Tecnai F20 Thermo Fischer N/A TemCam-F416 TVIPS N/A 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, Tamir Gonen ( tgonen@ucla.edu ). METHOD DETAILS Proteinase K Crystal Growth Proteinase K from Engyodontium album (Sigma-Aldrich, St Louis, MO, USA) was prepared by combining 2 ml of protein solution (50 mg ml −1 ) with 2 ml of precipitant solution (1.0–1.3 M ammonium sulfate, 0.1 M Tris pH 8.0) ( Hattne et al., 2016 ).

Sample Preparation

The protein solution was dispensed on a glow-discharged grid (easiGlow; Pelco) and vitrified with force position 24 in an FEI (now Thermo Fisher) Vitrobot Mark IV after blotting for 12 s at an environment humidity of 30% ( Shi et al., 2016 ). Frozen-hydrated grids were loaded onto a Gatan 626 cryo-holder and transferred to the microscope, where the specimen temperature was maintained at ~100 K.

Data Collection

Electron diffraction datasets from separate crystals were collected using an FEI Tecnai F20 transmission electron microscope operated at 200 kV, with the objective aperture fully open to evenly illuminate an area extending beyond the sample and setting the selected area aperture to closely match the size of the crystal. For each crystal of thickness 200–400 nm, the same 23° wedge was repeatedly collected up to five times by continuously rotating the stage from −12° to +11° (−38° to −15° for crystal 3) off its untilted orientation at a constant rate of 0.089° s −1 ( Nannenga et al., 2014b ). The rate of electron exposure was adjusted to 0.007 e − Å −2 s −1 , calibrated using a Faraday cage. The individual datasets, each consisting of 49–50 frames with exposure time 5.1s were recorded at a camera length setting of 1.2 m, corresponding to an effective detector distance of 2.2 m. All diffraction images were acquired using a TVIPS TemCam-F416 CMOS camera and corrected to account for negative pixel values ( Hattne et al., 2015 , 2016 ) prior to further processing.

Data Reduction Proteinase

K data were indexed and integrated in P 4 3 2 1 2 using MOSFLM ( Leslie and Powell, 2007 ) through its graphical interface iMosflm ( Battye et al., 2011 ). Wedges from six different crystals were merged by the order in which they were collected using AIMLESS ( Evans and Murshudov, 2013 ), and the set of free reflections was copied from the molecular replacement search model, PDB entry 5i9s. Neither of these crystals yield a complete dataset on their own, but since proteinase K does not exhibit a pronounced preferred orientation, this produced five reasonably complete datasets, each comprised of frames with a similar degree of exposure. The choice between intensities derived from summation integration and profile fitting was left up to the optimization algorithm implemented in AIMLESS; in all cases this resulted in profile-fitted intensities being used. Relative B-factors were calculated between merged single-crystal datasets with SCALEIT from the CCP4 suite ( Howell and Smith, 1992 ; Winn et al., 2011 ). To quantify the global effects of radiation damage, integrated intensities were averaged for each diffraction image. Averages from different crystals were scaled by a single factor in the range [0.10, 1.0] and simultaneously fit to a common function on the form A × exp(− B × x ) using the BFGS minimizer implemented in scipy ( Oliphant, 2007 ). The mean effective resolution was calculated by EFRESOL ( Urzhumtseva et al., 2013 ) and used as an objective high-resolution cutoff for all datasets.

Phasing and Model Refinement

The first dataset was phased by molecular replacement using MOLREP ( Vagin and Teplyakov, 1997 ) from PDB entry 5i9s, and the solution was reused for all subsequent datasets. Water molecules and ions were excluded from the refined structure: while these improve the quality of the model at high resolution, they are difficult to reliably model once damage degrades the quality of the data. This model was also used to calculate solvent-accessible areas with AREAIMOL ( Winn et al., 2011 ). All models were refined with REFMAC ( Murshudov et al., 2011 ), with electron scattering factors calculated using the Mott-Bethe formula. The occupancies were set to unity for all atoms and no alternate confirmations were used to model partial damage to specific sites of the molecule. Further processing and refinement statistics are given in Table 1 . GSNQNNF Crystal Growth The 7-residue peptide GSNQNNF (>98% purity) was purchased from GenScript, dissolved in water at 10 mg ml −1 , and crystallized by the hanging-drop method in a high-throughput screen. Crystals grew as needle clusters at a 1:1 ratio of peptide solution to mother liquor in a condition containing 10% (w/v) PEG-8000, 0.1 M MES pH 6.0, and Zn(OAc) 2 ( Martynowycz et al., 2017 ).

Sample Preparation

Clusters were broken by pipetting and dispensed onto glow-discharged grids, which were then blotted for 20 s and vitrified with force position 24. Otherwise, GSNQNNF samples were prepared identically to those of proteinase K.

Data Collection

Crystals of the hepta-peptide that were 100–500 nm thick, were tilted over ~60° at a three-fold higher rotation rate (0.3°s −1 ) than was used for proteinase K and up to 12 sweeps were collected from each crystal. To probe the effect of dose rate on radiation damage, peptide data were collected at both 0.0028 e − Å −2 s −1 and 0.0017 e − Å −2 s −1 . These rates were tuned to maximize the number of sweeps collected from an individual crystal. Single crystal datasets comprised of approximately 100 images were collected with an exposure time of 2.1 s and camera length 0.73 m which corresponds to an effective sample to detector distance of 1.2 m. Because two orders of magnitude fewer reflections are typically observed on a diffraction pattern from short segments like GSNQNNF than from proteinase K, intensities were integrated with a higher gain value and averaged for each dataset instead of for each frame when estimating the effects of global damage on the hepta-peptide. Otherwise data collection was performed as detailed for proteinase K. Data Reduction, Phasing, and Model Refinement The datasets were indexed and integrated in P 1 with XDS ( Kabsch, 2010b ) and an isomorphous subset was scaled and merged with XSCALE ( Kabsch, 2010a ). Phases for the GSNQNNF data were determined ab initio by direct methods from the first collected data set using SHELXD ( Sheldrick, 2008 ). XDSCONV ( Kabsch, 2010b ) was used on this dataset to assign a free set of reflections, which was subsequently reused for all later peptide datasets. A ligated acetate, three water molecules, and a single zinc atom were included with the GSNQNNF model, because they constitute a significant fraction of the unit cell contents, and all atoms were fixed at full occupancy. Otherwise processing was performed as detailed for proteinase K; statistics for the datasets at the high and low dose rates are given in Tables 2 and 3 , respectively.

QUANTIFICATION AND STATISTICAL ANALYSIS

Quantification and statistical analyses are given in Tables 1 , 2 , and 3 . These values were extracted from the programs used to merge and refine the respective multi-crystal datasets.

DATA AND SOFTWARE AVAILABILITY

The atomic coordinates have been deposited in the Protein Data Bank under ID codes 6cl7 to 6clt. Density maps have been deposited in the Electron Microscopy Data Bank under ID codes EMD-7490 to EMD-7512.

METHOD DETAILS Proteinase K Crystal Growth Proteinase K from Engyodontium album (Sigma-Aldrich, St Louis, MO, USA) was prepared by combining 2 ml of protein solution (50 mg ml −1 ) with 2 ml of precipitant solution (1.0–1.3 M ammonium sulfate, 0.1 M Tris pH 8.0) ( Hattne et al., 2016 ).

Sample Preparation

The protein solution was dispensed on a glow-discharged grid (easiGlow; Pelco) and vitrified with force position 24 in an FEI (now Thermo Fisher) Vitrobot Mark IV after blotting for 12 s at an environment humidity of 30% ( Shi et al., 2016 ). Frozen-hydrated grids were loaded onto a Gatan 626 cryo-holder and transferred to the microscope, where the specimen temperature was maintained at ~100 K.

Data Collection

Electron diffraction datasets from separate crystals were collected using an FEI Tecnai F20 transmission electron microscope operated at 200 kV, with the objective aperture fully open to evenly illuminate an area extending beyond the sample and setting the selected area aperture to closely match the size of the crystal. For each crystal of thickness 200–400 nm, the same 23° wedge was repeatedly collected up to five times by continuously rotating the stage from −12° to +11° (−38° to −15° for crystal 3) off its untilted orientation at a constant rate of 0.089° s −1 ( Nannenga et al., 2014b ). The rate of electron exposure was adjusted to 0.007 e − Å −2 s −1 , calibrated using a Faraday cage. The individual datasets, each consisting of 49–50 frames with exposure time 5.1s were recorded at a camera length setting of 1.2 m, corresponding to an effective detector distance of 2.2 m. All diffraction images were acquired using a TVIPS TemCam-F416 CMOS camera and corrected to account for negative pixel values ( Hattne et al., 2015 , 2016 ) prior to further processing.

Data Reduction Proteinase

K data were indexed and integrated in P 4 3 2 1 2 using MOSFLM ( Leslie and Powell, 2007 ) through its graphical interface iMosflm ( Battye et al., 2011 ). Wedges from six different crystals were merged by the order in which they were collected using AIMLESS ( Evans and Murshudov, 2013 ), and the set of free reflections was copied from the molecular replacement search model, PDB entry 5i9s. Neither of these crystals yield a complete dataset on their own, but since proteinase K does not exhibit a pronounced preferred orientation, this produced five reasonably complete datasets, each comprised of frames with a similar degree of exposure. The choice between intensities derived from summation integration and profile fitting was left up to the optimization algorithm implemented in AIMLESS; in all cases this resulted in profile-fitted intensities being used. Relative B-factors were calculated between merged single-crystal datasets with SCALEIT from the CCP4 suite ( Howell and Smith, 1992 ; Winn et al., 2011 ). To quantify the global effects of radiation damage, integrated intensities were averaged for each diffraction image. Averages from different crystals were scaled by a single factor in the range [0.10, 1.0] and simultaneously fit to a common function on the form A × exp(− B × x ) using the BFGS minimizer implemented in scipy ( Oliphant, 2007 ). The mean effective resolution was calculated by EFRESOL ( Urzhumtseva et al., 2013 ) and used as an objective high-resolution cutoff for all datasets.

Phasing and Model Refinement

The first dataset was phased by molecular replacement using MOLREP ( Vagin and Teplyakov, 1997 ) from PDB entry 5i9s, and the solution was reused for all subsequent datasets. Water molecules and ions were excluded from the refined structure: while these improve the quality of the model at high resolution, they are difficult to reliably model once damage degrades the quality of the data. This model was also used to calculate solvent-accessible areas with AREAIMOL ( Winn et al., 2011 ). All models were refined with REFMAC ( Murshudov et al., 2011 ), with electron scattering factors calculated using the Mott-Bethe formula. The occupancies were set to unity for all atoms and no alternate confirmations were used to model partial damage to specific sites of the molecule. Further processing and refinement statistics are given in Table 1 . GSNQNNF Crystal Growth The 7-residue peptide GSNQNNF (>98% purity) was purchased from GenScript, dissolved in water at 10 mg ml −1 , and crystallized by the hanging-drop method in a high-throughput screen. Crystals grew as needle clusters at a 1:1 ratio of peptide solution to mother liquor in a condition containing 10% (w/v) PEG-8000, 0.1 M MES pH 6.0, and Zn(OAc) 2 ( Martynowycz et al., 2017 ).

Sample Preparation

Clusters were broken by pipetting and dispensed onto glow-discharged grids, which were then blotted for 20 s and vitrified with force position 24. Otherwise, GSNQNNF samples were prepared identically to those of proteinase K.

Data Collection

Crystals of the hepta-peptide that were 100–500 nm thick, were tilted over ~60° at a three-fold higher rotation rate (0.3°s −1 ) than was used for proteinase K and up to 12 sweeps were collected from each crystal. To probe the effect of dose rate on radiation damage, peptide data were collected at both 0.0028 e − Å −2 s −1 and 0.0017 e − Å −2 s −1 . These rates were tuned to maximize the number of sweeps collected from an individual crystal. Single crystal datasets comprised of approximately 100 images were collected with an exposure time of 2.1 s and camera length 0.73 m which corresponds to an effective sample to detector distance of 1.2 m. Because two orders of magnitude fewer reflections are typically observed on a diffraction pattern from short segments like GSNQNNF than from proteinase K, intensities were integrated with a higher gain value and averaged for each dataset instead of for each frame when estimating the effects of global damage on the hepta-peptide. Otherwise data collection was performed as detailed for proteinase K. Data Reduction, Phasing, and Model Refinement The datasets were indexed and integrated in P 1 with XDS ( Kabsch, 2010b ) and an isomorphous subset was scaled and merged with XSCALE ( Kabsch, 2010a ). Phases for the GSNQNNF data were determined ab initio by direct methods from the first collected data set using SHELXD ( Sheldrick, 2008 ). XDSCONV ( Kabsch, 2010b ) was used on this dataset to assign a free set of reflections, which was subsequently reused for all later peptide datasets. A ligated acetate, three water molecules, and a single zinc atom were included with the GSNQNNF model, because they constitute a significant fraction of the unit cell contents, and all atoms were fixed at full occupancy. Otherwise processing was performed as detailed for proteinase K; statistics for the datasets at the high and low dose rates are given in Tables 2 and 3 , respectively.