FAQ  •  Register  •  Login

Strange signal shift

Forum rules
Please be polite and civil. We know that troubleshooting is vexing...
<<

Johanna

Participant

Posts: 8

Joined: Thu Jul 28, 2016 5:23 pm

Post Tue Oct 18, 2016 8:11 pm

Strange signal shift

Hi everyone,

I have encountered a strange phenomenon while recording data from a big barcoded experiment last week, and after hours of trying to figure out what is wrong I still did not come up with a likely explanation.

Here is a short description of my protocol:
I thawed cryopreserved PBMCs, stimulated for 5 hours with PMA+Ionomycin plus BrefeldinA/Monensin or BrefeldinA/Monensin only. After stimulation, I performed cisplatin viability staining, fixed the cells with 1.6% PFA, washed with CSM and left at 4C overnight.
The next day, I performed Palladium barcoding (0.02% Saponin, i.e. 1:50 dilution of eBioscience 10x Permeabilization buffer in PBS), pooled the samples, stained with surface markers, stained for Cytokines using 1x eBioscience Permeabilization buffer, and incubated with Ir intercalator in 2% PFA/PBS for 48 hours.

On the day of data collection, I resuspended the cells in 1 ml of normalization beads dilution and measured cell concentration. I collected data on a CyTOF2 using the supersampler in total 9 runs of about 1 hour each. Directly before each run, I diluted my initial cell/beads mixture about 10-fold to a concentration of about 500 000 cells/ml. At the end of the day, I concatenated all the files together.

What I noticed when I looked at the concatenated files was that for 4 of my channels (148Nd, 150Nd, 152Sm, 154Sm), the signal was shifting upwards during each of my 1-hour runs. Each time after I prepared fresh sample dilution, the signals came down for a short period of time before going up again. Other channels seem not to be affected (except maybe very weakly 149Sm, and 146Nd), the bead signal in all the bead channels (including 151 and 153 in the same mass range as the affected chanels) is steady as well. The affected markers are, except for IL-6 154 Sm, surface markers. I see this phenomenon independent of data processing (i.e. on the raw data, data normalized using the Matlab normalizer, on all my debarcoded files).

I was running a panel including the same surface markers, but stained for phosphorylated intracellular signaling molecules following methanol permeabilization, the day before and did not observe this signal shift. Another person was running samples directly after me and also did not observe something similar.
Before starting the experiment, I tested my cytokine panel (but without the barcoding step and using the loop system instead of the supersampler), and everything was fine then. (I only collected for 2x10 minutes then, so a shift over time might not have been that obvious for that short period of time, but still the signal seems to be very steady in this earlier experiment).

I am attaching a pdf to illustrate this problem. I'm showing plots of time vs. channels 147-155, for ungated files before normalization and debarcoding.

I would be happy for any explanation or suggestions of what tests to run in order to figure out what might be the problem here.
Thanks in advance,

Johanna
Attachments
signal_shift.pdf
(218.52 KiB) Downloaded 796 times
<<

AdeebR

Grand master

Posts: 169

Joined: Thu Mar 13, 2014 5:58 pm

Location: NYC

Post Tue Oct 18, 2016 8:46 pm

Re: Strange signal shift

Hi Johanna,

I unfortunately also have no good explanation for what might cause this, but I wanted to add that we've also come across a similar phenomenon on a few rare occasions (increasing background staining for a few specific channels over the course of the run). The samples in question had been shipped from an external collaborator, so I thought it might have something to do with inadequate fixation of the samples and consequent degradation of the samples upon exposure to water, but I'm not sure why this would cause increasing signals. Perhaps antibodies/metal falling off the sample and binding non specifically to other cells? I also wondered it could somehow be related to residual wash solution contaminant left in the SuperSampler syringe which was mixed into the sample upon initial aspiration and mixing and promoted sample degradation.

I'm curious to see if anyone has other ideas.

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

mleipold

Guru

Posts: 5792

Joined: Fri Nov 01, 2013 5:30 pm

Location: Stanford HIMC, CA, USA

Post Tue Oct 18, 2016 9:01 pm

Re: Strange signal shift

Hi Johanna,

A couple questions, for clarification:
1. You said "At the end of the day, I concatenated all the files together." Did you use the Fluidigm concatenator, or some other concatenator (FlowJo, etc)? Have you gone back to look at your original (separate) files, and seen that there's the same upward trend, just to make sure it's not some weird concatenator artifact?

2. I noticed that in your second "injection", not only is the signal intensity (y-axis) going up, but the cell density (heat color) also appears to be increasing over time. You can also see it in a couple of your other "injections", but the second one is the most striking. Adeeb's question about residual Wash solution is a possibility: I know that even using loops or the Helios PSI, if you're not diligent about that, you can get some degradation/streaking at the beginning of your sample. However, that typically only lasts a couple minutes.....it shouldn't be 20-30min of your 60min sample before you have unaffected sample....

3. Out of curiosity: I'm not that familiar with the behavior of IgA, CD43, or CD21: are these normally "clean", non-streaky antibodies like CD27, or are these antibodies like CD45RA that tend to be streaky anyway?


Mike
<<

BjornZ

Contributor

Posts: 43

Joined: Fri Jul 10, 2015 1:04 am

Post Tue Oct 18, 2016 9:09 pm

Re: Strange signal shift

Interesting -- I wonder if it could be from a gradual reduction in measured background/debris that leads to less background subtraction and consequently higher apparent staining. This fits with only seeing this in some channels and under some conditions, but for the relatively large changes in your PDF the background would have to have been pretty high, I think. You might try re-analyzing one or more IMDs with background subtraction disabled. Did the rain look dirty in the affected channels?

Best,
Zach
<<

Johanna

Participant

Posts: 8

Joined: Thu Jul 28, 2016 5:23 pm

Post Tue Oct 18, 2016 9:31 pm

Re: Strange signal shift

Hi Mike, Adeeb, and Zach, thanks for your replies!

1. You said "At the end of the day, I concatenated all the files together." Did you use the Fluidigm concatenator, or some other concatenator (FlowJo, etc)? Have you gone back to look at your original (separate) files, and seen that there's the same upward trend, just to make sure it's not some weird concatenator artifact?

I used the Fluidigm concatenator, and yes, I see the same thing if I look at the single files.

2. I noticed that in your second "injection", not only is the signal intensity (y-axis) going up, but the cell density (heat color) also appears to be increasing over time. You can also see it in a couple of your other "injections", but the second one is the most striking. Adeeb's question about residual Wash solution is a possibility: I know that even using loops or the Helios PSI, if you're not diligent about that, you can get some degradation/streaking at the beginning of your sample. However, that typically only lasts a couple minutes.....it shouldn't be 20-30min of your 60min sample before you have unaffected sample....

To clarify, what I would call the "unaffected" sample is actually what comes at the beginning of the run. At least in case of CD43, i seem to be "losing" my negative population over time, e.g. for CD19+CD20+ B cells, which should be primarily CD43 neg, these tend to become more positive during the run. (Even at the beginning of the run, the signal seems to be higher than I previously experienced, but at least still lower than for CD3+ T cells for example.) For the other channels affected, this is not as obvious, but in these cases the markers are expressed on a smaller number of cells, so the negative population is much bigger.

3. Out of curiosity: I'm not that familiar with the behavior of IgA, CD43, or CD21: are these normally "clean", non-streaky antibodies like CD27, or are these antibodies like CD45RA that tend to be streaky anyway?

I'm not sure what you mean by "streaky" antibodies, but in general I also experienced higher background/unspecific signal for many channels in the 141-154 mass range this time.

Interesting -- I wonder if it could be from a gradual reduction in measured background/debris that leads to less background subtraction and consequently higher apparent staining. This fits with only seeing this in some channels and under some conditions, but for the relatively large changes in your PDF the background would have to have been pretty high, I think. You might try re-analyzing one or more IMDs with background subtraction disabled. Did the rain look dirty in the affected channels?

Best,
Zach

I am not familiar with how the background subtraction works...How is the background to be subtracted measured?
I will try to do what you suggested... If I look at the processed files, I do see more background later in the run, but I guess this might also be the case if the level of subtracted background changes?
<<

GregBehbehani

Master

Posts: 85

Joined: Tue Apr 12, 2016 10:17 pm

Location: The Ohio State University, Columbus, Ohio

Post Tue Oct 18, 2016 9:32 pm

Re: Strange signal shift

Hi Johanna,

I certainly agree this is pretty strange.

I would also agree with Adeeb's suggestion about residual wash solution. It will take quite a while to normalize the low pH of the wash solution with water alone, this could lead to low signals until the pH comes back to normal, though it's hard to imagine why that would specifically effect only certain metal channels. If this were due to cell breakdown, you should see changes in Ir intensity and cell length. Did you see any evidence of this? If you do have changes in cell length, it will change the integration window of the cell events and this will have wide-ranging effects that will be most notable in dimmer channels with higher antibody background.

The only other thing I would suggest is that this could be due to some change in the oxidation of these metals (Sm and Nd are more oxidation sensitive than Eu). Maybe something to do with the amount of oxygen dissolved in the sample or some small variation in the injection rate of the super sampler. Did you notice anything similar with Pr141 (more sensitive to oxidation)? If oxidation was the issue, you should be able to see this in the bead events that you collected by looking at the 156 channel (this should give a reliable signal from Ce140 oxide that would change with any change in global oxidation).

Please let us know if you figure it out.

Good luck,

Greg
<<

GregBehbehani

Master

Posts: 85

Joined: Tue Apr 12, 2016 10:17 pm

Location: The Ohio State University, Columbus, Ohio

Post Tue Oct 18, 2016 9:40 pm

Re: Strange signal shift

Per Zach's suggestion, you can re-analyze the IMD files (to make new FCS files) with the "background subtraction" box unchecked. You would have seen a lot of streaking in the rain for 150 and 154 for background to have created shifts this large.
<<

Johanna

Participant

Posts: 8

Joined: Thu Jul 28, 2016 5:23 pm

Post Tue Oct 18, 2016 9:55 pm

Re: Strange signal shift

If this were due to cell breakdown, you should see changes in Ir intensity and cell length. Did you see any evidence of this? If you do have changes in cell length, it will change the integration window of the cell events and this will have wide-ranging effects that will be most notable in dimmer channels with higher antibody background.

No, I do not see this. Ir intensity and cell lenght stay pretty constant.

The only other thing I would suggest is that this could be due to some change in the oxidation of these metals (Sm and Nd are more oxidation sensitive than Eu). Maybe something to do with the amount of oxygen dissolved in the sample or some small variation in the injection rate of the super sampler. Did you notice anything similar with Pr141 (more sensitive to oxidation)? If oxidation was the issue, you should be able to see this in the bead events that you collected by looking at the 156 channel (this should give a reliable signal from Ce140 oxide that would change with any change in global oxidation).


I see a very low upward trend for the Pr141 marker, but I only noticed it because I was looking for it, as it is a lot less pronounced than for the other channels mentioned. Unfortunately I do not have the 156 channel in this panel, but I will open it for future tests to look at this.

Per Zach's suggestion, you can re-analyze the IMD files (to make new FCS files) with the "background subtraction" box unchecked. You would have seen a lot of streaking in the rain for 150 and 154 for background to have created shifts this large.


I remember seeing higher than usual background in 150, but not very dramatic (I've seen exemples of far worse background posted here).
<<

mleipold

Guru

Posts: 5792

Joined: Fri Nov 01, 2013 5:30 pm

Location: Stanford HIMC, CA, USA

Post Tue Oct 18, 2016 10:24 pm

Re: Strange signal shift

Hi Johanna,

By "streaky" antibody, I mean that there's free signal that's not cell-associated. See attached JPG for an example of a vertical streak for CD45RA:

101216-Liq vs Lyo-Pat03-Liq.jpg


Especially if it's only happening in one channel, this is usually due to the antibody itself continuing to wash off. CD45RA is notorious for this, even in flow. In some cases, there may be a specific type of debris that is taking up a bunch of only one antibody: I had this happen once with a huge CD56 streak in some brain tissue samples (NCAM=neural cell adhesion molecule...), but not in any other channel.

You said "At least in case of CD43, i seem to be "losing" my negative population over time, e.g. for CD19+CD20+ B cells, which should be primarily CD43 neg, these tend to become more positive during the run". If your cell density is changing over time (remember, the second injection graph appeared to get "hotter" with more cells, as well as higher signal intensity), then your debris and other junk concentration is also changing over time. This may lead to a gradual increase in background in all channels as the "streak" gets darker. I'm not that familiar with CD43: you say it's supposed to be on T cells and not on B cells....is it supposed to be on NK cells or Monocytes? If not, are NK cells and Monocytes also increasing in CD43 signal over time?


Zach: what do you mean by "background subtraction"? Do you mean noise reduction, or cell subtraction?


Mike
<<

Johanna

Participant

Posts: 8

Joined: Thu Jul 28, 2016 5:23 pm

Post Wed Oct 19, 2016 12:27 am

Re: Strange signal shift

Hi Mike,
thanks for explaining. No, I usually don't see a lot of streaking for my antibodies.
For this specific run, as mentioned, I observed some streaking, but not as much as in the CD45RA example you posted. I also did not notice it getting more intense towards the end of each run, but I am not sure whether I would have noticed it.

CD43 is also expressed on monocytes and NK cells (and activated B cells, plasma cells and B1 cells). CD43 depletion is frequently used to isolate untouched, resting B cells.

I have looked at my data again very carefully, and I don't think the number of cells increases over time. I've looked at time histograms (both for total events and gated on cells based on Ir/event length) and it seems very stable over the course of a run, with some variations between runs (due to slightly different dilution of samples.) I've also made a gate to catch a fixed time ranged and moved it around to different time periods during my runs, again no differences in event counts within the gate. So I think what you are referring to as the graph getting "hotter" over time is specific to the intensity range with highest cell densities rather than overall increased event/cell numbers.

Johanna
Next

Return to CyTOF troubleshooting

Who is online

Users browsing this forum: No registered users and 8 guests