10x Genomics claims that the Xenium provides sub-30nm resolution for single molecule RNA detection. I can’t find any data to back up their claim, or even any technical specs on their optics. Has anyone here had any experience with this machine, or similar, and tested the spatial resolution?
My instincts are screaming at me that this can’t be true because true SMLM systems need oil objectives, long acquisitions, and lots of post-processing, but maybe something about the spatial transcriptomics process lets them avoid that?
I believe that they are claiming that they can localize flourescently labeled RNA spots with 30nm precision, not that they can resolve spots that are present in the same imaging round and separated by as little as 30nm. This isn’t necessarily that difficult depending on the value of “high numerical aperture”. My understanding is that they use an amplification approach so each RNA will have many dyes associated with it, which helps to increase the localization precision.
1 Like
This is claim that probably would need more context.
Xenium uses Rolling Circle Amplification (RCA) to achieve 1000 of repetitions of the same sequence that will roll-up to a ssDNA nanoball.
Diffraction-limited signals are then achieved by hybridizing fluorescent oligos onto the repeated sequence.
RCA amplicons can differ slightly in size but somewhere around 100 nm to 1 um is something to expect depending on conditions. Shaping Rolling Circle Amplification Products into DNA Nanoparticles by Incorporation of Modified Nucleotides and Their Application to In Vitro and In Vivo Delivery of a Photosensitizer
Here is an example from my lab where we directly compare the diffraction-limited smFISH signal of RNA with the RCA product signal from the same RNA targets in budding yeast, focusing on two cyclins.
This would give you quite a good feeling of what kind of resolution and sensitivity you could expect relative to smFISH (Gold standard).
3 Likes
Thank you both for your replies!
What I’m getting from these two posts is that the Xenium can plausibly localize the center of an isolated mRNA strand to <30 nm, due to the large number of photons generated by an RCA amplicon. Even if the diffraction spots are very wide due to the RNA being outside the optimal focal plane, the large photon count should rescue it?
However, if two strands from the same gene are near each other, the results might be inaccurate? Since they are only taking a single image of each cycle (I assume?), the Xenium likely cannot differentiate two overlapping PSFs from two strands? Would the amplicons interfere with each other if they were too close together? In a situation like this from @Daniel_Furth 's fantastic example image, would it be able to get a trustworthy count and localization?
Thank you again!
1 Like
@smcardle this is a very good point that requires some further explanation.
In the case of smFISH overcrowding is a real phenomena where a molecule is so abundant och localized to a single space that it becomes impossible to distinguish them. Prime example of this would be if you make smFISH probe against 28S or 18S rRNA, you wont get diffraction limited spots but just a signal that looks like acridine orange staining of RNA.
For RCA amplicons this is actually less of a concern.
First, there are multiple steps to amplify up the signal so sensitivity is much lower than smFISH (you will not get all molecules but a random sample of them).
Secondly, the amplicons from RCA are single-stranded and very negatively charged. RCA from two molecules close to each other tend to not work, only one of the molecules will successfully amplify. The way we check these things is that we examine RCA amplicons for clonality – that there should be no mixed signal within a spot. This is very true for RCA amplicons and a good thing with them.
To get back to your case with the image you show of CLB2 in the bud. You can see that there are 5 amplicons in total inside the bud. 4 of them are in a cluster and the fifth is on the right side of the bud. When imaging with z-resolution this would be quite clear since what you see there is a maxintensity projection.
Again back to the original issue: can Xenium localize RNA down to 30 nm resolution. Probably not, or at least I would never be comfortable talking in these terms. This measurement would be based on merely segmenting out the amplicon and taking the centroid of the amplicon. Even in the best of worlds there are still other factors involved such as diffusion of RCA products etc. assuming the physics of the optics in terms of resolution are ideal, which is also a big assumption.
But these kind of claims are common within the spatial transcriptomic literature. For example, the other product 10x have - Visium - is usually described as having spots with a certain resolution. In practise you could get these spots down to 2 um or smaller. Problem is that vendors market this as “resolution” ignoring both the Nyqvist theorem as well as that these capture based technologies would have quite high diffusion rate of transcripts from when the sample is lysed to how RNA falls down on the array. This is very clear when you plot read data from spots outside of the tissue - there are a lot of molecules out there that could only possibly come there through diffusion.
2 Likes
Ooh, this is very helpful. Thank you so much for helping me understand this!
