Dealing with symmetry of nucleosome while processing

[Anjitha]

Hi,
I am working with a dataset of my protein of interest bound to nucleosomes. While processing, I observed that my protein was binding to both faces of the nucleosome. I was using the default C1 symmetry in all jobs, and was wondering if this is an artifact caused by the symmetry of nucleosomes, or a real observation. I tried subtracting the bound density on one side, and did a non-uniform refinement with the subtracted particles, and I see the density on the other side more clearly.
Does anyone with experience working with nucleosomes, or symmetrical particles have a way to tell whether it is real or not.

Many thanks,
Anjitha

[hickschadw]

Hi Anjitha,

I have some experience with processing cryoEM datasets of nucleosomes and protein+nucleosome complexes, so I may be able to help.

Nucleosomes with equal length linker DNA and without bound protein generally have C2 symmetry. You can apply C2 symmetry during your refinement jobs to “duplicate” the information into a single face of the nucleosome by combining the signal from both faces of the nucleosome. This effectively doubles your particle stack. If you do this, you can no longer interpret any potential asymmetry between the two faces because they will look the same. It is usually better to start with C1 symmetry, as you described, to determine if the nucleosome is truly C2-symmetric at the resolution you care about. If it is, then you can begin applying C2 symmetry to boost your signal to noise and increase resolution.

For protein+nucleosome complexes, you can have different stoichiometries of protein bound to the nucleosome depending on how many binding sites there are for the protein. If your protein binds to the nucleosome face (2 binding sites) then your particle stack may contain three different complexes (compositional heterogeneity) – free nucleosome, 1:1 protein:nucleosome, or 2:1 protein:nucleosome (protein bound to both faces of the nucleosome). However, if your protein binds to the nucleosome face but, in doing so, occludes binding to the opposite face of the nucleosome, then you will only have free nucleosome and 1:1 protein:nucleosome species in your particles stack. CryoEM density of bound protein at both faces of the nucleosome may indicate that you have nucleosomes with two proteins bound to a single nucleosome or it may indicate that the refinement job mask only included nucleosome density and did not include the bound protein(s), so it did not consider aligning the bound protein so that the bound protein was located at the same face of the nucleosome.

To determine if you have nucleosomes with protein bound to both faces, run a 3D classification job on the aligned particle stack. Use low resolution (10-15 Å), and ask for 4 output volumes, which would correspond to free nucleosome, 1 protein bound to the upper face of the nuc, 1 protein bound to the lower face of the nuc, and 2 proteins bound to each face of the nuc. You should use a focus mask that includes regions from both faces of the nuc where the protein could bind. Make sure the solvent mask overlaps with the focus mask in these regions. If you get a volume showing protein density bound to both faces of the nucleosome, then this is evidence showing that your particle stack contains nucleosomes with two bound proteins. In any case, you will want to exclude free nucleosome particles from your particle stack because it will reduce the visibility of the bound protein in your final map.

If you have 2:1 protein:nucleosome complexes, then I would recommend performing symmetry expansion on the particle stack with C2 symmetry to combine the information from both faces of the nucleosome into a single face and duplicate the particle stack. If you do this make sure to use local refinement jobs for your refinements, not homogenous refinement, heterogenous refinement, or non-uniform refinement. Make sure your mask includes only one of the bound proteins, not both. If you C2 symmetry expand you should only consider and interpret information from one face of the nucleosome. Also, make sure to leave the symmetry as C1 as a parameter in the local refinement job of C2 symmetry-expanded particles.

If you do not find evidence of 2:1 protein:nucleosome complexes, then I would recommend removing any free nucleosome particles using 3D classification and running regular C1 refinement jobs – homogenous refinement then local refinement. Make sure your masks include the region of density with your bound protein.

I don’t think it is necessary to perform signal subtraction of one of the bound proteins to boost resolution for the other bound protein if the bound proteins are small relative to the nucleosome, but it may help if the bound proteins are large.

Hope this helps!
-Chad Hicks

[hbridges1]

Hi @Anjitha!

That’s a great question! Seeing a binding partner appearing on both faces during C1 refinement might mean that there are two copies bound, or it may mean that there had been misalignment of some particles during refinement (or a combination of the two).

As nucleosomes can have C2 pseudo-symmetry, one test you can do is to run refinements with C2 symmetry with symmetry relaxation enabled (either maximisation or marginalisation). Using these settings, both of the plausible symmetry-related orientations for particles will be explicitly tested, and you can then compare the resulting map to your C1 result to see if the density location(s) for your binding partner appear the same. If with C2 and symmetry relaxation the density for your binding partner is in only one location and has better quality density than in the C1 refinement, then there may only be one copy there.

[Guillaume]

Hi Anjitha,

If the DNA sequence is symmetric (which is the case if you assembled nucleosomes with the alpha-satellite repeat sequence, for example), and both copies of each histone are the same and unmodified, then the nucleosome is truly C2 symmetric. The histone tails can adopt different conformations on both sides and likely always break the symmetry, but they are too dynamic to contribute to the reconstruction in most cases. Exceptions to this trend are observed with nucleosome-binding factors interacting with histone tails, but in such cases the whole complex is likely asymmetric (depending on binding stoichiometry and whether two copies of the binding factor can sterically accomodate each other).

If the DNA is asymmetric, like the Widom 601 sequence, then there is truly no C2 symmetry, but you need base-pair resolution to be able to align these images correctly (i.e. correctly assign “heads” versus “tails” projections, so to speak; I like this terminology because the nucleosome kinda has the shape of a thick coin). So in practice, you have C2 pseudo-symmetry and get blurry nucleobases, in the absence of larger asymmetric features.

One trick that greatly helps aligning images of nucleosomes is to artificially make them asymmetric in a way that is distinguishable at lower resolution. And one easy way to achieve this, other than with a binding factor stably bound only on one side (which is not always easy to ascertain, as you pointed out) is to design the DNA such that the nucleosome-positioning sequence has an additional 10 bp only on one side.
For an example of this, see EMPIAR-11618. I collected this dataset when our Glacios in Uppsala had just been installed and needed test samples, and it is suboptimal in many ways (1-year old nucleosomes kept in the fridge, so more prone to unwrapping; grid prepared quickly without much optimization; microscope setup done by a less experienced version of me from 2021), but if you process it you should see the 10-bp linker DNA pop up easily in your reconstructions (a reconstruction I did quickly is at EMD-17944; it was necessary to be allowed to deposit to EMPIAR, but I didn’t have time to do it very carefully, so you can probably get a better result).

Of course, it is not always possible to use this trick, depending on where your binding factor binds. Or it may not even help much, for example with proteins that can bind 2:1 on both sides of the nucleosome (like many chromatin remodelers do at SHL2), because the larger binding factor drives alignment much more than the extra 10 bp of DNA (which end up less visible because averaged out).