Dear community,
I’m working on a small membrane protein complex
(~100 kDa ordered mass) with almost no extramembrane
features. I solved its structure bound to a Fab fragment to
3.4
Å
resolution using a 200 kV Arctica microscope. We recently
also collected a 300 kV Krios dataset (~250k particles, no
phase plate) in an attempt to solve the structure of the
membrane protein alone, without the Fab fragment. The 2D
classes obtained from this dataset look fine and display
membrane helical features within the micelle. Unfortunately,
standard 3D classification and refinement approaches in both
Relion3 and Cryosparc2 only yielded low resolution maps
(worse than 9 Å), in which
membrane helices cannot be distinguished.
At first, this led me to belief that the
reconstruction of the membrane protein alone suffers from
low SNR-induced misalignments due to the large, disordered
micelle surrounding the small protein. However, when I
perform an ab-initio reconstruction in Cryosparc2 with a max
resolution of 6
Å (instead of the 12 Å default), I
get a map in which all helices are clearly visible as
sausages, consistent with a ~6
Å map. The positions of the helices
are identical to the
3.4 Å
Fab-bound map, even though the reconstruction was ab-initio.
When I subject the corresponding map and particle set to a
regular gold-standard homogeneous refinement in Cryosparc or
Relion, I again end up with a map at a worse resolution than
9 Å
without visible helices. Altogether, this tell me that there
is high-resolution information in the data but that the
standard image-processing strategies are not suitable.
My very general question is: Could anyone here
suggest any image-processing strategies that worked in
similar cases that I could still try to obtain a good
reconstruction with this dataset?
Strategies within Relion3 and/or Cryosparc2 I
tried so far, but which didn’t give satisfactory results:
- Increasing the amplitude
contrast of the particles from 0.1 to 0.15/0.2.
- Highpass filtering the
particles to 120
Å.
- Increasing the tau_fudge
parameter in Relion to higher values during refinement.
- Lowpass filtering the
initial model to only 8 – 15
Å during
classifications and refinements.
- Numerous masked
classifications with different mask sizes and softness’s.
- Non-uniform refinements in Cryosparc.
- Finer angular sampling during classifications
and refinements.
Thanks in advance,
Wout
Wout
Oosterheert; PhD Candidate; Crystal and Structural
Chemistry; Bijvoet Center for Biomolecular Research;
Utrecht University; The Netherlands