I’m processing a relatively small GPCR in LMNG micells. The ICL3 region has been replaced with BRIL, so there is some additional protein density protruding from the micelle.
The 2D classification results look reasonably good, and I can see particle features that appear to correspond to the protein rather than just the detergent micelle. However, once I proceed to Ab-initio, the secondary-structure feature of the TM domain are no longer apparent.
In the Ab-initio and subsequent Heterogeneous Refinement maps, the micelle density seems to become increasingly dominant. When I increase the contour threshold, the TM density disappears first, while the micelle density remains relatively strong.
This makes me wonder whether the particle orientations are being driven mainly by the low-resolution detergent micelle signal rather than by the GPCR/TM domain.
I would really appreciate any advice on the following:
Would using higher-resoultion information during Ab-initio be appropriate? For example, I have seen suggestions to try initial/final resolutions such as 9/7 Å or even 7/5 Å for small membrane proteins.
Once a rough protein-containing volume is obtained, would Non-Uniform Refinement or TM-focused Local Refinement be a reasonable way to improve alignment of the TM?
Are there any other processing strategies you would recommend for preventing micelle-dominated alignment?
I can provide screenshots of the 2D classes, Ab-initio volumes, as well as the processing parameters if helpful.
Nice looking 2Ds there, specifically for the “top” views. If you can post what your current data processing workflow/scheme is it would be helpful in offering advice. Did you attempt to apply the workflow described here?
The main issue is that the 2D classes look quite good, but once I move to Ab-initio Reconstruction, the TM secondary structure features are no longer apparent and the detergent micelle becomes much more dominant.
I will post a screenshot of my current processing scheme as well. I have not yet tried the automated GPCR workflow you linked, but I will take a look at it.
Hi. The top-view 2D classes look good. However, there are not enough high-quality side-view classes, suggesting a potential preferred orientation issue. You may want to try the Rebalance Orientations → Ab-initio Reconstruction → Homogeneous Refinement workflow, or perform 3D classification to obtain a more complete reconstruction first. Based on the 2D classification results you shared, I’m not sure whether it will be possible to obtain an intact map.
I will first try Rebalance Orientations and see whether it improves the reconstruction. If there is still no significant improvement, I may need to consider preparing the sample again.
In a previous attempt with this sample, I encountered a severe top-view-biased preferred orientation problem. To reduce this issue, for the current sample I added 0.01% digitonin immediately before vitrification and used a 2 nm thin carbon film.
If you have any suggestions or experience with other approaches that are effective for improving preferred orientation, I would be very grateful to hear them. Any advice would be helpful for my future work on this project.
--You might want to try screening several different types of detergents and different types of grids during cryo-EM grid preparation. -Considering the relatively low molecular weight of the complex, you could also try strategies to increase its effective molecular weight, such as introducing a universal scaffold like an anti-BRIL Fab. Fingers crossed!
I think you will be able to get this to resolve to a reasonable resolution. It does look challenging but here is what I would do. You have plenty of particles so I would try this: https://www.biorxiv.org/content/10.1101/2025.09.08.674935v1 using 3-4 classes and resolution range between 5-3 Ans. Run this multiple times, picking out the best particles each time. It will take some time.
Next take the ab initio maps that you have got and go back to the beginning, denoise the micrographs, and repick with the template picker projected from the best map. Then run the junk detector. Run inspect picks with aggressive thresholds so you only get good picks. This will get you nearly all of the particles from the dataset.
Next use your best ab initio map and several decoy maps to run het refine, set the initial resolution to around 8 Ans. Run several sequential het refines until most of the particles go into a single class, you can also increase the box size from 128 to 256 as needed in later rounds, particularly if the resolution increases as you sort.
Then run another HAIR Ab initio on the best particles. The best set from this should have much better TMD features and refine well with NU refine. You can then do more classifications at that point as needed.
The quality of the micelle-like density is reminiscent of bad classes you can get in DDM containing samples. GDN at concentrations between 0.01 to 0.06 %, and Au300 holey mesh grids are worth trying out.
Thank you very much for your kind and detailed advice.
Interestingly, I attended a seminar last week where HAIR Ab-Initio was also discussed, so it is quite a coincidence that you suggested the same approach here.
I will definitely try following the workflow you described and see whether I can obtain a better particle set and improve the TMD feature.
Thank you again for taking such an interest in my dataset and for sharing these suggestions. I plan to post updates as the processing progresses, so if you happen to have some time later, I would really appreciate it if you could stop by and take a look at the results.
Just to clarify, would you recommend adding GDN to the sample immediately before vitrification at a final concentration in the rang of 0.01-0.06%?
Also, is the main purpose of adding GDN to improve preferred orientation, or is it primarily intended to improve the quality and homogeneity of the micelle-like density? I apologize if this is a basic quation - I still have a lot to learn about optimizing membrane-protein cryo-EM samples.
My SEC buffer already contains 0.00025% GDN. although this is obviously much lower than the concentration range you suggested.
As additives immediately before vitrification, what do you think about detergents other than GDN. such as CHAPSO or fluorinated FOS-Choline-8? Would these be worth screening as well for this type of GPCR sample?
I will definitely keep your suggestions in mind when preparing the next sample. Thank you again for your advice!
GDN is most useful as a replacement to fix the presence of empty micelles and improving particle qulaity, as I cannot say for certain it assists in fixing preferred orientation. The CMC of GDN is 0.0021%, and depending on salt conditions it may be worth using the 0.01 to 0.06% range. If you require stabilizers such as CHS in buffer, it may be troublesome to solubilize in GDN, so you can opt to only use the GDN buffer in the final SEC before vitrification and continue using existing buffers in earlier stages.
