FAQ  •  Register  •  Login

Transient signal drop

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

AMitch

Contributor

Posts: 21

Joined: Mon Nov 24, 2014 12:40 am

Post Thu Jan 19, 2017 5:21 am

Transient signal drop

Hi all,

I recently ran some samples for a user and in a number of them saw an artifact that I haven't noticed previously. Basically, there were transient sharp drops in signal and then then a rapid ramping up to the "normal" signal. This was best seen using a bead channel and I've attached an example image.

My first thought was, obviously, a transient clog and there were indeed a number of clogs on the day, and the event rate did vary at times. However this doesn't match well with previous experience where - as far as I have observed - transient clogs come out looking like a brief break in the signal rather than a drop in intensity. I went back and looked at a number of recent experiments and noted the "break" phenomenon but not the "drop" one I'm describing here.

Another theoretical argument I had against this being a clog was that if a particle can make it through the nebulizer and spray chamber into the plasma then I'd assume that it should form a coherent cloud of ions that would be analyzed efficiently (and therefore not show decreased signal). So - my best guess so far is that the effect originates somewhere downstream of the plasma. Any comments on this appreciated...

Image
<<

mleipold

Guru

Posts: 5796

Joined: Fri Nov 01, 2013 5:30 pm

Location: Stanford HIMC, CA, USA

Post Thu Jan 19, 2017 4:12 pm

Re: Transient signal drop

Hi Andrew,

Did this happen in *all* your channels in that time interval, or just a few?

If it happened in *all* your channels, then it's possible that you had a momentary "flicker" in how your sample was being delivered into the machine.

For example, we've had 1-2 cases where someone *really* screwed up their processing and wound up introducing a *LOT* of salt into the instrument. Over the course of a couple hours, they completely salted up the injector tip, partially blocking it. This caused a deflection in the sample delivery out of the tip of the injector, effectively causing an XY misalignment. So, the cells were getting thrown into the plasma, being burned, but a substantial chunk of the ion cloud was "missing" the Sampler orifice (when we checked, we could see a salt spot at about 10 o'clock relative to the orifice).

So, in theory, anything that affects sample introduction through and including the skimmer-reducer could manifest in a similar fashion: I could even see if there was a momentary fluctation on the applied Current for the skimmer-reducer could affect how efficiently ions get into the machine.


Note: I have a lot more experience with a transient nebulizer clog causing a deflection within the spray chamber. In *that* case, the cell transmission efficiency drops for a second, but those cells/beads that *do* make it into the injector/plasma/cones are counted correctly.


In short: I would keep an eye on it, and see if it happens again (especially if it starts happening regularly). If it does, you might need a service call (though diagnosing an intermittent problem is really really difficult).


At least in your case, if *all* channels were affected, then maintaining Ir-bright and Bead-bright gating should effectively "excise" the mis-measured Events from your downstream analysis.

Though this is one case where I think the Fluidigm normalization would give you a problem: since it "forces" all the beads (and therefore cells) back to an externally-determined value, it would force that transient dip back as well, unless there's a subtlety to the algorithm that I'm not aware of. The MATLAB normalizer should be more immune to it.


Mike
<<

GregBehbehani

Master

Posts: 85

Joined: Tue Apr 12, 2016 10:17 pm

Location: The Ohio State University, Columbus, Ohio

Post Thu Jan 19, 2017 5:23 pm

Re: Transient signal drop

Hi Andrew,

I agree with everything Mike posted.
I would add two other possibilities:
1. You could have had a transient fluctuation in the airflow through your plasma exhaust; this can happen on a windy day or if there is a fire alarm (which typically triggers exhaust fans that try to vent the smoke from the building). You probably have a flow regulator valve on your plasma exhaust, but it may not normalize the gas flow instantly. This essentially also creates a transient XY misaligment.
2. If you're using some type of syringe pump (CyTOF1&2 or supersampler) a transient clog might cause some pressure build up that might increase the sample flow rate for a few seconds after the clog passes. If this were the case, the plasma would transiently cool down when the flow rate went up. You should be able to rule this in or out by comparing the ratio of 156 to 140 signal from your beads. If the plasma cooled down, the 156 signal will increase relative to the 140 signal (though the absolute 156 signal may still be lower).

Please let us know if you sort it out.

best of luck,

Greg
<<

AMitch

Contributor

Posts: 21

Joined: Mon Nov 24, 2014 12:40 am

Post Fri Jan 20, 2017 5:42 am

Re: Transient signal drop

Hi Mike and Greg, thanks for the great suggestions/comments.

Just following up with some more relevant information/data/interpretation in case anyone else sees this phenomenon:

-I went through data from several experiments prior to the affected one as well as the two full days running after the affected experiment and the signal drops weren't apparent..

-We have a Helios with a pressurized sample loader. I keep an incident log and when I re-checked this I noted a couple of apparent incidents of increased sample flow through the fluidics. I don't know if these corresponded with the signal drops and I might have missed seeing others.

-The samples from the affected run were stained, fixed and stored in methanol as per Sumatoh (2016) and were shipped to the facility on dry ice. This is the first time I've run samples processed like this. I did the final washes so know there should be little/no salt in the samples.

-Signal for most channels dropped simultaneously with the 140 bead signal *except* the 156 signal from gated beads, this increased slightly and it was even more apparent when the 156:140 ratio was plotted (see link to image below).

-The event length seemed to also drop transiently during the signal drops (also in linked image).

So, FWIW my interpretation...

The increase in the 156:140 ratio suggests that the plasma cooled, probably as a result of transient clogs. The Helios sample introduction system monitors sample flow rate and has a feedback system that adjusts the pressurization of the sample loader to keep sample flow rate constant. A partial/transient clog could have caused the sample flow rate to drop which could then in turn have led to increased pressurization of the sample loader. If the clog cleared by itself then the higher pressure would have released a burst of sample/water into the spray chamber until the flow corrected. This burst would have cooled the plasma leading to the increased 156/140 signal ratio. The only thing that doesn't quite make sense to me is the shortened event length. Only thought is that the decreased signal causes it to be above background for a shorter period(?)

As for the data, we can filter out the "bad" bits using Boolean gating in Flowjo so can salvage the experiment.

https://drive.google.com/file/d/0BzJCE_ ... sp=sharing

cheers,

Andrew
<<

mleipold

Guru

Posts: 5796

Joined: Fri Nov 01, 2013 5:30 pm

Location: Stanford HIMC, CA, USA

Post Fri Jan 20, 2017 4:24 pm

Re: Transient signal drop

Hi Andrew,

Very interesting....thanks for catching that the EL value changed too. I wouldn't have expected that either: Looking at some recent cloggy samples, it appears we see both a drop in EL and and increase in EL that corresponds with bursts-through of a clog (see attached). We're using a Helios with the PSI unit as well.

The only thing I can think of to explain the lower EL values: EL is a function of how much the ion cloud expands during its transit through the plasma and into the machine. The strongest driver of this is total metal content and the associated charge-charge repulsion once the positively charged metal ions can "see" each other when chelators and such are burned away. But it's also a function of plasma transit *time*: from what I understand, that's why the Helios injectors yield cell events with lower EL (~17 EL compared to ~25 EL for v1 and v2 injectors, if I recall), since they do a better job of focusing the event into the heart of the plasma and effectively the shortest route through the plasma.

So: if a clog burst through, the pressure causes a momentary faster flow rate (faster rate of sample introduction). Therefore, cells are running at a higher velocity and might shorten their transit time, thereby shortening the EL.


I think my clogs might still be a bit different than yours: while I do see a spike in M+O16 signal that corresponds fairly well with the spike in events per second (Time histogram), I don't see much evidence for a *decrease* in M signal at the same moment. Neither in Ce140 (bright bead channel) nor Ce142 (dimmer bead channel). So, momentary deflection might still be an issue for you?


Mike
Attachments
011717 samples.pdf
(1.87 MiB) Downloaded 334 times
<<

GregBehbehani

Master

Posts: 85

Joined: Tue Apr 12, 2016 10:17 pm

Location: The Ohio State University, Columbus, Ohio

Post Fri Jan 20, 2017 8:22 pm

Re: Transient signal drop

Hi Andrew,

Thanks for the follow-up. Very interesting. I thought that might be what was happening, but I'm a bit surprised that it happened with the PSI; that's good to know.

I also agree with your interpretation about the Event Length. I'm not at all surprised that your EL decreased when the signal went down. The total signal intensity (across all channels) is used to determine the boundaries of the when a cell event starts and stops, so any cell that was small or not positive for many markers is likely to see a decrease in its event length when there is a global decrease in signal intensity (as the potion of the event cloud containing enough ions to trigger the event start and event stop will be smaller). I bet this change in event length was fairly minimal for the beads (which are very bright and small) and much more pronounced for cell events that were dim or positive for only a few of the markers in the panel. If this is a problem in your data, you should be able to re-extract the IMD with a decreased lower convolution threshold and the effect should be diminished or eliminated.

best,

Greg

Return to CyTOF troubleshooting

Who is online

Users browsing this forum: No registered users and 10 guests