FAQ  •  Register  •  Login

Doublet discrimination based on Ir191/193

Forum rules
Please be as geeky as possible. Reference, reference, reference.
Also, please note that this is a mixed bag of math-gurus and mathematically challenged, so choose your words wisely :-)
<<

AdeebR

Grand master

Posts: 169

Joined: Thu Mar 13, 2014 5:58 pm

Location: NYC

Post Mon Aug 17, 2015 5:55 pm

Doublet discrimination based on Ir191/193

Hello CyTOFers,

Hopefully most of you are already aware of this, but I wanted to share a few gating examples illustrating some points of concern with traditional doublet discrimination based on Ir191/193+ DNA content and event length; namely that S-phase cells fall in the same region as true doublets, and also that different cell types have different DNA content and that certain cell types (e.g., granulocytes, macrophages and some stromal cells) may be accidentally excluded in a standard doublet gate. Just some points to keep in mind.

Doublet gating concerns.pdf
(143.9 KiB) Downloaded 1142 times


When dealing with tricky samples of this sort, I think that barcoding samples and using mathematical doublet filtering to remove invalid barcode combinations is the most reliable way to accurately remove doublets.

Adeeb
Adeeb Rahman
Icahn School of Medicine at Mount Sinai, NYC
<<

mleipold

Guru

Posts: 5796

Joined: Fri Nov 01, 2013 5:30 pm

Location: Stanford HIMC, CA, USA

Post Mon Aug 17, 2015 6:10 pm

Re: Doublet discrimination based on Ir191/193

Hi Adeeb,

Thanks for the post!

Clarification question: On page 1 (S-phase), what is the parent of the CD4 vs IdU (right) plot? The 165lo-169hi box from the left plot?

Do you see more variation in Ir193 signal in tissue samples (like your EpCAM example on page 2) compared to something like PBMCs, or do you think it's present to the same extent, regardless of sample type?


Mike
<<

billog

Participant

Posts: 15

Joined: Wed Sep 17, 2014 10:53 pm

Post Mon Aug 17, 2015 6:50 pm

Re: Doublet discrimination based on Ir191/193

I guess the question is whether the Helios upgrade might solve this problem... assuming this is CyTOF II data...

Random question Adeeb, what clones of EpCAM and CD68 are you using?

Thanks, - Bill
<<

mleipold

Guru

Posts: 5796

Joined: Fri Nov 01, 2013 5:30 pm

Location: Stanford HIMC, CA, USA

Post Mon Aug 17, 2015 7:37 pm

Re: Doublet discrimination based on Ir191/193

Hi Bill,

If it's an inherent signal intensity variation due to cell cycle phase or other physical properties, then I don't think the platform would matter much.

A platform might make a difference in the *overlap* between discrete populations of true singlets, but the center of each distribution would remain in the same place. In other words, take the final graph of Adeeb's example: the *separation* between the clouds of the macrophages and T cells could improve, but the almost log difference between the center of the two populations should remain.


I know that you have done a lot of whole blood work: have you seen similar Ir signal differences between Macrophages and T cells in PBMCs?


Mike
<<

billog

Participant

Posts: 15

Joined: Wed Sep 17, 2014 10:53 pm

Post Mon Aug 17, 2015 8:01 pm

Re: Doublet discrimination based on Ir191/193

I did a double check for whole blood looking at CD66+ granulocytes vs. every other leukocyte bulked together and at least based on the one sample I looked at it didn't seem like I was disproportionately losing CD66+ singlet events because they might stain more brightly with IR although I suppose one can assume CD66+ cells that are not in cycle might have similar DNA content as CD66- cell. Again, assuming IR is only staining DNA or nucleic acid at low concentrations which might be a risky assumption since at high concentrations it can stain cells w/o permeabilization...

Those Macrophages Adeeb is looking at though would only be present in a single-cell suspension derived from a tissue homogenate (looks like lung based on his legend) one would NOT see those in the peripheral blood. So I suppose the inference here is that macrophages and epithelial cells bind more IR; however, I'm not sure its safe to assume this is definitely because of more DNA content... it doesn't matter what the reason is a solution like bar-coding is more important.

Generally this is like gating general cell populations exclusively based on scatter characteristics in flow cytometry: its risky because T cell blasts for example are more granular than naive T cells. Its risky but people still do it all the time...
<<

AdeebR

Grand master

Posts: 169

Joined: Thu Mar 13, 2014 5:58 pm

Location: NYC

Post Mon Aug 17, 2015 8:05 pm

Re: Doublet discrimination based on Ir191/193

Mike, yes the parent gate for the CD4 vs IdU plot is the 165lo 169hi gate from the left plot. The parent gate of that plot is just total "cells - beads". I should note that the two CD45 population (169 and 165) were barcoded, which is why the double positive events can be unambiguously assigned as true doublets.

I think some of the cell based heterogeneity is more apparent in tissues than in blood/PBMCs partly because of the cell types involved. I find that PBMCs are usually pretty uniform in DNA content, however with whole blood, granulocytes (neutrophils and eosinophils) certainly have higher Ir193 signal than lymphocytes. However this difference is not as dramatic as the example of tissue macrophages and epithelial cells that I presented in those example plots, and is usually distinguishable from true doublets.

Also, as you noted, these are inherent differences in signal intensity with, I believe, valid biological basis behind them so the front end acquisition system shouldn't make a difference in their incidence. However, the Helios software is also going to be collecting additional features of the ion cloud distribution besides "event length", and in theory additional information about this distribution should potentially allow a better separation of ion cloud fusion events (probably wouldn't make a difference for true cell aggregate doublets). As far as I know, this is still theoretical though.

Bill, the Epcam staining in this example was with mouse, and I was using clone G8.8. The CD68 data was with human tissue and in that case I was using clone Y1/82A.

Adeeb
Adeeb Rahman
Icahn School of Medicine at Mount Sinai, NYC
<<

AdeebR

Grand master

Posts: 169

Joined: Thu Mar 13, 2014 5:58 pm

Location: NYC

Post Mon Aug 17, 2015 8:38 pm

Re: Doublet discrimination based on Ir191/193

Here's and example of the difference in Ir193 signal between whole blood granulocytes and lymphocyte (T cells). It is a more subtle difference, but if you're worried about it an easy way to address it is to use a tailored singlet gate for granulocyte and non-granulocyte populations as I have done in this example.

Granulocyte_Ir193.jpg
Adeeb Rahman
Icahn School of Medicine at Mount Sinai, NYC
<<

billog

Participant

Posts: 15

Joined: Wed Sep 17, 2014 10:53 pm

Post Mon Aug 17, 2015 8:55 pm

Re: Doublet discrimination based on Ir191/193

I should qualify and say IR will stain fixed cells at high concentrations and fixed cells are compromised enough in cell membrane integrity that IR could be staining nucleic acid specifically.

I guess the key control data I have never seen is IR staining of live cells at different concentratios, washing, and then fixation & acquisition. This may be in some of the early CyTOF papers perhaps...

In some places Fluidigm's literature says Cell-ID stains nucleic acid in some places DNA, I'm not sure if anyone knows how much Cell-ID staining of RNA is impt... if it's not and Cell ID is purely reading out DNA then its a little bit of a head scratcher why epithelial cells and macrophages have more DNA than leukocytes unless they are all in cycle somehow..
<<

AdeebR

Grand master

Posts: 169

Joined: Thu Mar 13, 2014 5:58 pm

Location: NYC

Post Mon Aug 17, 2015 10:09 pm

Re: Doublet discrimination based on Ir191/193

Hi Bill,

Very true, and the live cell Ir stain would be an interesting control to try. I can say that when using cisplatin as viability dye with single cell tissue suspensions I've seen a lot of heterogenity within the "live" (cisplatin-low) population, and the cells that stain more intensely for cisplatin usually also stain more intensity for Ir, so perhaps this does reflect differing levels of RNA content?

In the case of epithelial cells (and other highly-adherent stromal cells), I suspect that a lot of single cell dissociation protocols are not 100% effective in completely dissociating the tissue to a single cell suspension, so the high-Ir193 events may actually reflect mini-aggregates of cells rather than true single cell events. This might also help to explain why these kinds of samples are more prone to clogging the CyTOF.
Adeeb Rahman
Icahn School of Medicine at Mount Sinai, NYC

Return to CyTOF data analysis

Who is online

Users browsing this forum: No registered users and 8 guests