For more information on our CAN-ASC Webinar Series, visit our Webinars page.
Q&A Summary
“A step-by-step guide to generating high parameter flow cytometry data” by Florian Mair, PhD
Questions compiled and answered by Sabine Ivison (from CAN-ASC) and Florian Mair. Note: these answers are based on the verbal Q&A of the webinar and are not meant to be fully exhaustive references.
Standardization
When you said that you adjust your voltage to obtain the same MFIs for your beads, what is your “acceptable” range? Is +\- 10% is ok? This is a difficult question to answer as the acceptable range of variation depends on the particular instrument. I would refer to Perfetto et al, 2012 for more details – I try to go as precise as possible and would rather say +/- 5% or less.
Setting gates
How useful are FMOs for setting gates? Many times the FMO negative population boundary is much lower than in the fully stained sample. Isotype controls look like they work well (no jumps in negative popn boundary), but it's not recommended? FMOs are very useful to determine if a specific signal is caused by spreading error in an underlying population, but FMOs cannot account for non-specific signal increase which accompanies every addition of a fluorophore-linked antibody. As for isotype controls: the level of background will depend on the specific antibody used. As an alternative, one option is to find reference populations with known expression patterns within your cellular mixture to set gates, and ensure these are of the same cell types (eg. don’t compare a negative monocyte with a positive lymphocyte).
Single cell controls
For your single cell gating to determine the compensation matrix, do you use dot plots (FSCA or SSC vs. fluorochrome) or do you use a histogram. How tight do you make the gates? Typically, after tightly gating on the bead population in a FSC vs SSC plot it is sufficient to use histograms for positive and negative gates. It is important to ensure that there are enough events recorded in each gate (preferably over 5000).
Is it possible to use bead controls for all the fluorochromes and cells for Fixable viability dye? This is fine, as long as the negative and positive controls within one parameter are using the same material (eg. the same cell type or beads).
Can you calculate SSM with single stained beads or does one need to use the same cell type of the experiment? In my experience, for the reference SSM single-stained beads are sufficient.
Do you compensate for your live/dead stain? We routinely never do so, so I'm curious how important you think this is. Per se this can be done, but I would advise against it. While for example FACSDiva allows to include a parameter without having a single stained control for it, and would still correctly calculate the spill-over of all other channels into the L/D channel, some other commonly used analysis programs would not allow you to do so. The assumption is correct that the spillover of the Live/Dead stain does not matter, because you gate out the dead cells – however, I would still include it in my comp matrix.
Does the median FI of the beads positive need to be as bright or more than sample or the whole range of positive beads? The goal is that the median fluorescence intensity of the positive bead population is as bright or brighter as the MFI of the brightest population in the actual sample.
If you want to be sure that the comp control positive is BRIGHTER than the one in your panel then I guess you would need to use beads- this would not be possible with cells? unless you use much more Ab? It is only important that the positive compensation control is AS bright, not brighter. So cells are fine.
Re: fixing controls, do you also mean that this is needed for bead controls? Like PFA fixing the beads? Yes, exactly that. As exposure to the fixative can impact the flourochromes, single-cell controls (be i beads or cells) should be treated the same way as the cells in the samples you are analysing.
If I determine the SE and brightness for various fluorophores on my instrument with beads, how much do the specific antibodies I use to stain those beads affect my results? (Different antibodies can have varying levels of conjugation per ab). That is very true and that is why the official recommendation is to use beads and the same Abs used in your panel. The answer will always depend on the specific antibodies used. Sometimes there may be little difference, but in other cases there may be significant differences. Practically, however, we find that comp matrices usually require some tweaking no matter which method is used. Although this is always very dangerous and should be done with caution and the final results for key markers checked with an FMO.
Antibody titrating helps with spread problems. However, if you compensate your panel with comp beads and excess of antibodies are the SSM data reliable? Titrating antibodies can help dealing with spreading error because you can specifically lower a signal while still reliably separating positive and negative populations. The lower the signal, the lower the spreading error – a caveat is that working with subsaturating concentrations of antibodies requires similar cell numbers and staining conditions to avoid variability in the staining. The SSM data are reliable as long as the signal in the positive control is as bright or brighter than the signal in your sample – having brighter controls is never a problem, as long as they remain in the linear range of the detector.
What causes over- and or/ under-compensation? The most common causes are likely: using a single cell control that is either too dim or conjugated to a different fluorochrome (eg. FITC vs. AF488). Or using beads for the positive population and cells for negative peaks, or using positive and negative controls that have different levels of autofluorescence. Inaccurate compensation can also arise when you have treated the single stained-controls differently than your sample (eg. different exposure to fix/perm buffers, for instance).
Spillover spreading matrix and panel design
Is there an online resource that will tell you the spreading error for any given combination of fluorophores for panel design? Spreading error is highly dependent on the specific cytometer configuration and the performance of the optics/detectors so there would need to be the ability to choose all of these variables to determine the SSM. To our knowledge there is not a website out there which offers this.
So a good flow core needs to post their SSMs. And how often should it be updated? Rather than posting their SSM, the flow core should post the raw data from single-colour controls (in every possible fluorochrome) acquired on each cytometer using its optimal configuration as determined by voltration. That way the users can pull the fcs files from the parameters/fluorochromes they are considering into FlowJo and determine the SSMs themselves.
As to how often it should be updated, that depends on the level of Swiss-ness of the SRL manager :) . Assuming that everyone is using the same cytometer settings (eg. the optimal voltages) and these voltages are standardized (updated using standardized beads), they should not require updating very often. If the voltages are not being standardized, or the users are adjusting them, they will no longer be absolutely reliable- but will still provide a good idea where major spillover problems are. Perhaps once a year/after major maintenance. The SSM is currently the best thing out there to help you design your panel, but much will depend on the individual markers you are using.
Is this correct: if a detector tends to ACCEPT a lot of spillover then it should be on a very strong marker (to rise above the spread)? Yes.
Does voltration help minimize spreading error at all? Voltration is done to ensure you can achieve the optimal separation of signal in a given detector (usually SRLs will pick one ‘representative’ colour in each channel to do this). The historic notion that “voltage balancing” helps with spreading error is incorrect, as demonstrated here: Cytometry tutorial: The impact of adjusting PMT voltages on spillover and compensation
Are there instruments known to have lower spreading error that you’d recommend? Not that we know of.
If the spreading error is instrument-specific, would repairs to the cytometer or re-calibration of the machine change the spreading error in a way that would change the practical outcomes (i.e. panel design)? Recalibration of the machine could have some impact on the SSM but is unlikely to results in changes so serious that it would impact your choice of fluorochromes, unless the samples acquired to make the SSM were acquired on a really sub-optimally calibrated cytometer. Configuration changes (e.g. change of optics, lasers, detectors), would require a new SSM.
When we do not have an idea about a marker on a specific population or we do not know the expression after treatment, then the gating tree would not be very useful. What would be your strategy in these conditions? In this case I would treat these markers as “unknown/dim” and assign fluorophores accordingly.
So you always go for the lowest spreading error? That depends on how bright the markers are that you need to resolve on a specific cell. If you have a very strong marker that can rise above the spread, then a larger spreading error can be tolerated.
Do we need to consider only the spread donated and received for the markers in our panel, or the total sum like you are doing in this example? It is only important to consider the fluorophoes that are actually in your panel. The purpose is to ascertain the contribution of individual fluorochromes if these are co-expressed on the celltype you want to analyse. So the sum of all markers that are co-expressed will create the overall spread on any given detector. In the specific example shown during the presentation, the goal was always to design a 28c-panel, so all fluorochromes were considered. If on that particular instrument you e.g. want to do a 10-color panel, you could actually pick 10 fluorochrome-detector pairs that will have almost no SE.
Are the relative SSM values comparable across different SSM matrixes? Yes they to some extent they are. It is an arbitrary, unitless number, but larger SSM values will always equal more spread.
Would you recommend picking a fluorochrome for key lineage markers (eg. CD45 and CD4) just sticking with them? Yes that is a decent strategy if you have the flexibility in the other Abs to accommodate.
Other questions
When you are looking at published data is there any quick way to know if people had a spillover issue and incorrectly interpreted data? This is an excellent question and there is no easy answer. It is suspicious if at least some samples of key data is not shown clearly in the paper or supplemental figures. Good papers will include two-dimensional plots clearly showing the gating strategy. It is very difficult to know and there is still a lot of trust involved in accepting people’s data, although journals are increasingly asking for raw data and even analyses (eg. wsp files) to be uploaded to publicly-accessible servers. Be on the alert for the ‘super-negative’ populations which often (but not always!) indicate overcompensation.
For microparticles it is recommended to use H instead of A, but for compensation controls the system accepts A... can I compensate in A, but analyze in H? This is related to the fact that for particles that are smaller than the laser beam diameter the Signal-height is more representative for the total signal, while for particles that are larger than the laser beam diameter (e.g. leukocytes) the Area is considered more representative of the total signal. For the compensation, this does not make a difference (to best of my knowledge).