When the global pandemic hit and Chicago’s shelter in place orders prevented me from going into the lab, I started doing some experiments in my kitchen instead. I decided to take on baking sourdough bread (like many other quarantined individuals). And while I learned to perfect my bread in my free time, I spent a lot of my working hours on training materials for new flow cytometrists – that’s when I started to realize some similarities between these two tasks. It’s been a while since I learned flow cytometry, but baking bread reminded me of the process of learning a new skill: follow a protocol, assess the results and identify problems, determine how to fix the problems, and repeat the protocol with adjustments to address the problems. As I created my flow cytometry training materials I realized that I didn’t cover all of these steps. Flow cytometry training materials usually focus on the protocol of how to perform a flow cytometry experiment, but they don’t always cover how to troubleshoot. The step that is often skipped is how to assess your results to identify problems. With baking this is easy for everyone, even a novice – you taste your food and immediately know it doesn’t taste good. But for novice flow cytometrists this step of identifying problems can be very difficult. Looking back, I realized that no one really ever told me what bad data looks like. It was just a skill that I eventually picked up. So in this post I’d like to go over some examples of bad data to assist novice flow cytometrists.
There are a lot of ways to get bad flow cytometry data – in this post I’ll focus on the reasons that I feel are the easiest to identify and fix. From this post, you should be able to identify if there are major issues with the instrument settings (which can only be fixed by re-recording the samples) or if the samples need to be cleaned up in analysis software with gates or a new compensation matrix.
FSC and SSC Settings
FSC and SSC should be set so that all of your cells of interest are within the plot. If FSC is set too low, it will be difficult to use the FSC to remove debris or noise. If FSC is too high, it can also be difficult to cleanly gate on single cells. The same rules apply for SSC. The example above shows a cell line with a distinct cell population. In this case, it should be relatively easy to set the FSC and SSC. However, tissue digests containing multiple cell types can be a bit more difficult if you are unfamiliar with the sample.
The above images show data from a tissue digest. The first plot shows the FSC and SSC for ungated cells and at first glance, these FSC and SSC voltages look great. However, I’ve mentioned that it can be more challenging to set FSC and SSC for digested tissue. To determine if the settings are correct, you can use a technique called backgating. In this case, I gated on CD4+ T cells or F4/80+CD11b+ granulocytes and then examined the FSC and SSC of the gated cells. The second plots shows that my T cell population is in the middle of the FSC SSC plot, but the third plot shows that the granulocyte population is on the top axis. In this scenario it is best to consider the goal of the experiment – if T cells or lymphocytes are the only cells of interest, these voltages work well. However, if the goal of the experiment is to look at granulocytes, the SSC should be decreased. The images below show a different experiment with different voltages using the same tissue. Using the backgating technique to first gate on SiglecF+ granulocytes and then look at the FSC and SSC, you can determine that these voltages for FSC and SSC are appropriate for this cell population.
Fluorophore Voltage Settings
For best results, avoid saturating the fluorophore detectors. Similar to the FSC and SSC, all fluorophore signals should be within the plot. The two examples below show data with incorrect voltage settings. Note that in the right plot, 71% of the events are saturating the detector.
Before proceeding with data analysis you should always make sure compensation is correct. My last blog post went into detail about compensation – read it here if you missed it! However a good flow cytometrist will always be on the lookout for compensation errors. You can easily identify compensation errors by looking at the negative portion of the axis – populations that are not symmetrical and below zero are concerning and can indicate compensation errors. The teardrop shape can sometimes be ok or it may be compensation or autofluorescence (third panel). However symmetrical spreading error is perfectly acceptable (read more on the trumpet effect here).
Another unusual pattern you may find in your data is caused by antibody aggregates. This pattern is a bit harder to identify because the flurorophores used to find the pattern are unique to the panel. However if you are finding a pattern of super bright events similar to the one below you may have antibody aggregates. For a completed data set you can remove them with a gate. To prevent them from happening you can centrifuge your antibodies for 10,000 RPM for 3 minutes prior to using.
Clogs and Other Issues with Cytometer Fluidics
Before analyzing your data, it’s a good idea to look at the time parameter to examine the cytometer fluidics. If there were no issues with the fluidics, you should see an even signal for the duration of your sample. The images below show problematic data – the gaps suggest there may have been clogs. It may be possible to salvage the data by gating on the portion of data where the signal was steady. In some cases you can draw a gate manually – for example all of the data after 10 in the last panel below. Alternatively algorithms like FlowClean and FlowAI could also be used to clean the data. Please note: algorithms do not like slashes, so if any of the parameter names contain slashes, extra steps should be taken to remove them. If you are using FlowJo, you can use the export function to create new fcs files where all slashes are replaced with underscores. However if clogging is a recurring problem, you may consider filtering your samples prior to running them.
There is one situation where you will expect to find interruptions in the time parameter: data acquired by syringe injection. In this case, you will find the data is interrupted at regular intervals as the syringe empties and refills. Examples of cytometers with syringes include the Attune, BD fortessa HTS, and Helios. Any cytometers that require a specific tube so that it can be sealed and pressurized does not utilize a syringe. It is up to you if you want to remove the interruption of refilling the syringe, but it is recommended.
To set up the cytometer voltages/gains properly, all data (or as much data as possible) should be inside of the plot. Once data has been recorded, there is no way to alter voltages/gains. If it is determined that voltages/gains are incorrect, samples must be re-recorded on new settings.
Before analyzing data, you should check for correct compensation and inconsistencies in the time parameter. If compensation is incorrect, it is simple to generate a new compensation matrix to apply to the samples. If there are problems with the time parameter or if there are antibody aggregates, these can usually be solved during analysis with additional gating or a cleaning algorithm.
If you have questions about your data quality, I offer consultation services to help identify and troubleshoot problems. Also, this “bad data” post is the first in a series, stay tuned for the next one!
Update: Part 2 can be found here.