Membrane protein TM domain has contained by micelle

Hi, Everyone!

I am struggling with a membrane protein of Mw 100 kDa in detergent. I could see good 2D classes in the top and side view. Moreover, after reconstruction (ab-intio, hetero-refine and Nu-refine), we are never been able to get high-resolution maps of the transmembrane domain.

I can see the top view, but can’t see an obvious side view. Which one do you think represents the side view?


Dear @olibclarke
Thank you very much for your reply.

I have labeled 8 class which I guess it is a side view. In fact, this protein has a big intracellular domain and a transmembrane domain. In the begaining, after 2D classification, this protein was found to have an orientation advantage (almost top view). So, I added detergent to eliminate orientation.
When I select all top view to further refine (ab-initio and NU-refine), I can not find transmembrane domain in the map. So, I chioce side view particles stacks in the 2D class (e.g., figure below) to run further refine.

I can see a extra density (NU-refine).

But the map is still bad.
Is it possible that the detergent is wrapping the transmembrane domain of my protein so I can not get a high resolution map of that region? and how can I fix this proble?

1 Like

By the way, I am not a native English speaker, so there may be some grammatical errors in my replies, so please forgive me!

Hi Wu,

These do not look like side views to me I’m afraid - they look like mis-centered picks where you have two particles close together. I would consider the possibility that the detergent has not worked yet to eliminate the orientation bias - often we need to screen multiple detergents, or use a substrate like graphene, in order to find something that works for a problem case. You could also try extensively sub-classifying the non-top views, to see if something more obvious pops up?

I would also try running ab initio just with the high res top views - the results will be very stretched, but perhaps will give you some idea what to look for if there are a few side view mixed in there - worth a try!


1 Like

Hi oli,

Thank you so much for your reply!

In fact, I used a self-made graphene oxide layer to keep the protein stay in the hole.

(We can clearly find that there is no sample without GO layer in this figure)

I also perform many batch of 2D classiciftion to sub-classifying non-top views particles, but thry still seem to be the same (figure below).

Nearly 20,000 particles (non-top view parcticles represent a small percentage in total)

As you advice, I run the ab-initio for the high res of top views and I will post it here as soon as the results are come in.

By the way, I have a reference map from other’s published article with a large micelle belt between intracellular and transmembrane domain, I load this map in cryoSPARC and create template, but it has a large difference from my non-top view (below fig). Can this be a evidence that this non-top view is not a atual side view?

create template form import volume.

It can. Your sample is far too concentrated - it is clearly overlapping where protein is present on your micrograph and the 2D confirms this with how the classes look.

In your original post you’ve got some nice looking top views…

If Topaz doesn’t work to find any, you could try picking with only a side view (from the templates you created from that EMDB map) and try cleaning it up that way; that sometimes works (leave the low pass filter at default, no matter how tempting it is to try picking higher frequencies!)

But to be honest, when I see micrographs that crowded, I usually go back to playing with grid conditions as it’s often more successful than trying to fight with the data.

1 Like

Thank you for you reply!

I put the non-top view particles to extra 2D classification, I got the result like this (below).

(almost 10,000 particles)

I found it hard to believe that this is not what I wanted, because it really does look like a side view (through it only few thousand)

As you advice, how can I use just side view as template (create template will create 50 equally-spaced templates), how can I use part of them to do template picking.

You can run the Select 2D job on the templates and just select one or two classes. :slight_smile:

1 Like

Thank you very much, I have never executed select 2D job like this. :grinning:

1 Like

Can be pretty useful; especially since template picking speed scales on number of input templates.