May 23, 2022
Processing and symmetry expansion of LRRK2RCKW on microtubules
- Mariusz Matyszewski1
- 1Department of Cellular and Molecular Medicine, University of California, San Diego, La Jolla, CA 92093
Protocol Citation: Mariusz Matyszewski 2022. Processing and symmetry expansion of LRRK2RCKW on microtubules. protocols.io https://dx.doi.org/10.17504/protocols.io.bwwnpfde
License: This is an open access protocol distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Protocol status: Working
We use this protocol and it’s working
Created: July 26, 2021
Last Modified: May 31, 2024
Protocol Integer ID: 51886
Keywords: cryo-em, helical reconstruction, symmetry expansion, microtubule, LRRK2, ASAPCRN, microtubules protocol for symmetry expansion, lrrk2rckw filament, symmetry expansion of lrrk2rckw, microtubules protocol, microtubule, expanded particle, symmetry expansion, popular cryo, cryo, lrrk2rckw
Funders Acknowledgements:
ASAP
Grant ID: ASAP-000519
MJFF
Grant ID: 18321
Abstract
Protocol for symmetry expansion of LRRK2RCKW filaments bound to microtubules. This protocol covers everything from preprocessing to creating a new set of symmetry expanded particles that can be used by any of the popular cryo-EM refinement programs.
Guidelines
Keep in mind that filament information contained within the Relion star files is important in preventing duplicate particles. CryoSPARC recently included helical processing, but I did not have time to verify if helices extracted using it will contain the proper helical information for Relion to work with.
Relion does good job of preventing duplication as long as Helical Reconstruction is on (even if not using helical symmetry). If FSC do not reach 0, there is a good chance that something went wrong and duplicate particles were created along the way. Once symmetry expanded, there is no easy way to check for particle duplication.
Materials
Requires the use of Relion 3.1 for the new particle subtraction job.
Software for mask manipulation. We used Python with mrcfile and numpy packages installed, but similar results can be achieved in Chimera or ChimeraX if creating the masks by hand.
Troubleshooting
Safety warnings
None.
Preprocessing data
Align frames and do CTF fitting
Use your preferred software. The original publication used MotionCor 2 and CTFFIND4, but other programs should work just as long as CTF information can be imported into Relion.
As a note, all of the data processed was collected on lacey carbon grids rather than C-Flats or UltrAuFoils.
Filament Picking
Manual picking to create references
We used Relion 3.0 originally (manual picking in version 3.1 should be identical) to manually pick filaments for about 20 micrographs. Using a helical rise of 30 Å, this resulted in about 1600 particles. This extraction used slightly modified extraction parameters from the rest of the data: 600 px box size, re-scaled to 250 px, resulting in 2.784 Å/px size. For helix settings, we used a tube size of 600 Å.
2D classification to create references
Using the extracted particles, we classified them into 10 classes, using 20 iterations, fast subsets off, and a 600 Å mask. For helix settings, I used a 600 Å tube diameter. Restrict helical offset to rise was not used. We selected 3 different looking classes to act as references for AutoPick.
AutoPick
References were lowpassed to 15 Å. In-plane sampling was set to 3 degrees.
Autopicking threshold was set to 0.05. Minimum average noise was disabled and so was maximum standard deviation of noise. We used GPU acceleration with shrink factor of 1.
Helix settings were: 500 Å Tube size and helical rise was set to 30 Å with 1 asymmetric unit. Maximum curvature was set to the default value of 0.1 and minimum length requirement was not imposed.
These settings results in about 250 particles per micrograph, but were prone to picking carbon edges as well.
Extraction and particle clean-up
Initial extraction was done with a box size of 656 px, rescaled to 164 px (4x binning). This gave us a pixel size of 4.64 Å/px.
To clean up, we classified in Relion 3.1 with 100 classes, using 20 iterations (Fast subsets on), 450 Å mask, and initial angular sampling of 3 degrees.
Since this step is to mainly determine which particles have microtubules, the tube diameter was set to 300 Å and restrict helical offsets to rise was disabled.
Second round of the classification (50 classes, everything else the same) was performed on the good classes (anything that looked like it had a microtubule), removing further contaminates.
Selecting proper LRRK2RCKW filaments
2D classification to select ordered LRRK2RCKW filaments
50 classes, 25 iterations, fast subsets on, 650 Å mask, initial angular sampling of 3 degrees, 500 Å Tube diameter, restrict helical offsets to rise set to off.
Since this dataset contained MLi-2, only filaments showing extremely ordered LRRK2RCKW were selected.
3D classification to determine protofilament size
This step is inspired by MiRP and used their provided references for the different protofilament sizes.
1 round of classification with alignment was done on the 6 references of protofilaments ranging from 11 to 16 PF size. The references were lowpassed to 15 Å, and a mask diameter of 450 Å was used. Sampling interval was set to 1.8 degrees. Helix settings were enabled without applying any helical symmetry. 325 Å tube diameter was used. Tilt and psi had a search range of 3 degrees.
The results split up the particles into different protofilament sizes. Only protofilament size of 11 was picked for this dataset.
Note
Reference:
Alexander D. Cook, Szymon W. Manka, Su Wang, Carolyn A. Moores, Joseph Atherton,
A microtubule RELION-based pipeline for cryo-EM image processing,
Journal of Structural Biology,
Volume 209, Issue 1,
2020,
107402,
ISSN 1047-8477,
Code and references available at: https://github.com/moores-lab/MiRP
Re-extract particles at a lower binning
Initial re-extraction was done at 2x binning, but if computation power and storage is not an issue, you can re-extract at the original pixel size now, as the final symmetry expanded particles work best unbinned. Additionally, it is possible to continue using 4x binned images and unbin them at Step 15 as that is the last step before symmetry expansion.
Signal subtraction of the microtubule
Refine the microtubule section of the filament
Use the reference structure of a 11-PF microtubule lowpassed to 15 Å as a starting structure. We used a mask of 450 Å, and an initial sampling of 1.8 degrees.
Helix settings used a tube diameter of 300 Å, rot and psi search of 3 degrees, keep tilt-prior fixed set to true.
Helical symmetry applied was based on the usual 11-PF symmetry, in this case the starting point was 11 asymmetric units, -32.5 deg rotation and 11.3 Å rise. Twist search was confined to -33 and -32 degrees, but the rise was allowed to float between 10.5 and 12 Å. Both used 0.1 unit search.
If the final microtubule structure does not reach mid-resolution (about 6-8 Å or better), please do a 3D classification search with 3 classes, then use refine for each good class separately. However, the subtraction does not need to be perfect, as we will be reverting back to the original particles before the symmetry expansion.
Create a microtubule mask
Using the output of the refine, create a mask by choosing the lowest threshold that only shows the microtubule. We extended the map using 6 pixels, and added a 6 pixel soft-edge.
Subtraction
Due to changes to the subtraction module in Relion 3.1 (using half maps from the iteration, rather than the final map), we need to mimic Relion 3.0 behavior. This can be achieved in 2 ways:
1. Convert the star files to Relion 3.0 format and use the subtraction job there. Use the newly created mask and the final refine map.
2. Use the new External job type in Relion 3.1 to run subtract in legacy mode:
External executable: relion_project
Params:
i: final map from the 3D Refine
mask: newly created mask
ang: particle data from the 3D Refine job
o: name of folder as would be created by this job (ex. External/job024/subtracted)
angpix: -1 (or set to current pixel size, -1 will attempt to read the header)
subtract_exp
ctf
Old subtract only uses 1 thread, so leave mpi at 1.
Refining LRRK2RCKW filaments
2D classification to choose the best LRRK2RCKW filaments
40 classes (might need to adjust depending on the number of particles at this point, we had 47K for this particular dataset), 650 Å mask, 20 iterations, fast subsets set to off. 3 degree in-plane sampling. Tube diameter of 500 Å and no restriction on the helical offsets.
Only classes with an intact LRRK2RCKW filament were picked.
3D classification of LRRK2RCKW filaments
If no reference is available, run a Class3D job with a featureless cylinder with the expected diameter. Feel free to use a previously solved LRRK2RCKW filament as a starting model if this is a related structure, as long as the protofilament size is the same. If running on a cylinder, only create 1 class. Otherwise 3 classes are better to use.
Settings to use:
Lowpass reference to 15 Å, 20 iterations, 650 Å mask, 1.8 degree angular sampling.
Helix settings:
Tube diameter 250-520 Å (to exclude potential left over microtubule signal)
3 degree tilt and psi search range, keep tilt-prior fixed true.
Helical symmetry settings:
33.3 degree rotation, 31 Å rise (or a previously refined value; starting values based on Villa's tomography paper)
Allow for symmetry search of +/- 2 degrees and 2 A.
This might need to iterated if some classes look less sharp than others. If starting with a cylinder, re-run this with the output but this time request 3 classes.
Note
Reference:
Reika Watanabe, Robert Buschauer, Jan Böhning, Martina Audagnotto, Keren Lasker, Tsan-Wen Lu, Daniela Boassa, Susan Taylor, Elizabeth Villa,
The In Situ Structure of Parkinson’s Disease-Linked LRRK2,
Cell,
Volume 182, Issue 6,
2020,
Pages 1508-1518.e16,
ISSN 0092-8674,
3D Refinement of the LRRK2RCKW filament
Take the best LRRK2RCKW classes and let them refine using Refine3D.
Use the same settings as the Class3D job above, updating the helical symmetry as needed.
Unsubtraction of the microtubule and full refinement
Use the subtract job to revert to the original particles. This should work even if the particles were created with the legacy Subtract job.
If planning on using unbinned data for the final symmetry expansion, make sure to re-extract the reverted particles with no binning at this stage, if you haven't done so before.
Use the reverted (and unbinned, if desired) particles for a final refinement of the filament:
Use the output of Step 14 as the reference, lowpassed to 15 Å and a 650 Å mask.
Make sure to set initial sampling to 1.8 degrees and that local searches is on at 1.8 degrees as well. This should cause Relion to start off with a local search rather than randomizing the poses, as we need keep the particles in roughly the same position as in the last step.
Use the full tube diameter of 520 Å (we are refining the microtubule as well in this step).
Angular search should still be 3 degrees for both tilt and psi.
Since the microtubule does not have the exact symmetry as the LRRK2RCKW filament, we are not applying helical symmetry for this step.
If everything goes well, you will have a map with both LRRK2RCKW and a low resolution microtubule visible.
Symmetry expansion
Create masks for LRRK2RCKW dimers within the filament.
Since in our case there were 3 strands of LRRK2RCKW present that did not follow C3 symmetry, we had to create masks for each LRRK2RCKW dimer we wanted to extract. We chose to focus on the CorB-CorB dimer as the center of the mask and kept a significant portion of the microtubule within the mask as well. Parts of the neighboring LRRK2RCKW were kept as well.
Since each "particle" mrc is just a moving frame of 30 Å along the filament, make sure that the three masks you are creating are relatively planar to each other.
To facilitate creations of the masks, we created spherical "ones" masks for each dimer we wanted to extract and then multiplied it by the refined mask. Note: this behavior is new to Relion 3.1, as previously masks were used to determine what to subtract, but now they are used to determine what to keep, hence the "ones" map.
Each resulting mask was the processed by Relion by lowpassing to 15 Å, extending the map by 12 pixels, and adding a 12 pixel soft-edge. Helix option is not used for mask creation.
Symmetry expansion by subtraction
Using Relion 3.1's new Subtract job, we used the last optimizer from the previous refine job as input and created a Subtract job for each mask created in the last step. We also used centered the images on the mask and changed the output size to be smaller, 300 by 300 pixel particles if using unbinned particles, as this saves on computational processing power later on.
Realign the new particles
Since each mask corresponds to a different orientation of LRRK2RCKW within the filament, they need to be realigned. For best results, use one of maps that was used to create the subtraction masks. It will need to be same size as the new particles. We used python scripts to extract a subsection of the map centered on the mask. Alignment can be done in any processing software. We transitioned almost exclusively to CryoSPARC at this stage, and from this step processing will vary depending on what one wants to accomplish.