“Analysis Consistency in Flow Cytometry” — How to use Cytobank functionalities to achieve consistency in gating, display, analysis iterations, and data communication
“Making Beautiful Plots: Data Display Basics” — Choosing appropriate plot types, labeling, compensation, and how to properly set scale settings in flow cytometry experiments
“Future Proofing Your Experiments and Files: The Importance of Annotation” — An article detailing the importance and power of annotating your datasets and ensuring annotations remained linked to the raw data
This time around, we’ll delve into another round of issues to consider when designing and running a flow cytometry experiment. These themes have emerged out of our personal benchwork experience, our experiences assisting Cytobank users with their analyses, and insights we gained from analyzing large clinical datasets. In this post, we’ll do a brief overview, so stay tuned for future posts that expand on each of these issues.
1. Titrate your antibodies
Before undertaking an experiment, always get to know your reagents. Titrate your antibodies to maximize signal and minimize background staining. Adding too little antibody may result in antibody not staining staining your target, or you may have poor separation from a negative population. Adding too much antibody will raise the background staining level of cells that are actually negative for your target. You’ll want to stain in step-wise increasing concentrations of antibody to determine how to optimally detect a positive signal.
2. Make an antibody mastermix
When staining multiple tubes that share at least portions of a staining panel, always make an antibody mastermix to distribute among the samples. This will minimize pipetting variance and ensure consistency across samples, allowing you to make sample-to-sample comparisons. Avoid adding individual antibodies to individual tubes when possible.
3. Prepare appropriate controls
Think about the questions you’re asking and the controls you’ll need to answer these questions. Consider what you’ll be able to conclude with the controls you choose. For example, comparing unstained/unstimulated cells to stained/unstimulated cells (or using fluorescence-minus-one (FMO) controls) won’t allow you to draw conclusions about basal signaling, as signal in this case is influenced heavily by background staining (and is greatly affected by the results of your antibody titration). To address biological questions like basal signaling, you’ll need biological controls. We suggest combinations of the following:
- Add inhibitors and lower the signal (caveat: the inhibitors may be acting on something other than what you think).
- Add phosphatases and lower the signal (caveat: if you remove all the phosphos in your cells, off target binding, non-specific binding, and background will go down too).
- Add siRNA and remove the actual target of the antibody (great control, but knock down is never 100%, they are difficult to get into cells, and can influence cell biology in unintended ways).
- Stimulate long enough that endogenous phosphatases come on and your signal goes down to biological zero (caveat: there’s no way to know that you have hit the very bottom).
- Compare to genetically matched cells that are missing only the one gene of interest (transgenic knock out + add back of the target). This is the only unassailable control. The main caveat is that most people don’t actually make them genetically matched (which is critical) — instead they might find two cell lines with differential expression of the gene and assume that nothing else about the cells is different.
- Look at many samples and show that some samples have very low basal and others are very high (use natural variation to argue for basal differences; caveat: people will suspect it’s because of staining differences from sample to sample, so it’s important to show that some things don’t change from sample to sample and others do).
Want to quantify signaling induced by a stimulation? Include unstimulated/stained to compare to stimulated/stained.
Need to normalize statistical values to make comparisons among samples within a study? Choose one sample to be the control for fold change comparisons (e.g., choose one healthy sample’s basal measurement to be the reference point for other healthy samples and other cancer samples).
Choosing appropriate controls will save you time, money, and effort down the road. Give this step a lot of thought! (Stay tuned for a more in-depth post on this topic in the future.)
4. Including a predictable sample for reference
When running a new experiment, it’s a good idea to include a sample with a known outcome so that you can use it as a reference point to assess the efficacy of your stimulation and detection. This will help in understanding surprising results, and troubleshooting if something goes wrong.
5. Not switching reagents
If you intend to compare files within an experiment, or experiments within a group using a normalization reference, make sure not to switch reagents. While you may be inclined to substitute fluorophores with similar emissions spectra (or use entirely different ones) if you run out of a reagent, your files will no longer be directly comparable (especially true when switching to a tandem dye or a completely different fluorophore/channel). Different fluorophores achieve different intensities, have different stability, and bleed differently into other channels.
6. Using beads for compensation controls
Often researchers like to use the same antibodies used in their experiment for single stained compensation controls. Sometimes the cells used in the study express the target bound by a given antibody in abundance, and other times binding is weak or nonexistent. You can take the guesswork out of generating single stain compensation controls by using beads that bind antibodies (e.g., IgG-binding beads will bind all IgG antibodies, guaranteeing you signal). Antibody-binding beads ensure consistency and uniformity among compensation controls.
Some extra tips, geared at multi-experiment datasets (e.g., large patient cohorts)…
7. Name experiment files consistently
Naming files and experiment with consistent nomenclature helps keep your data organized, prevents analysis errors, and facilitates downstream bioinformatics approaches that can manipulate data in bulk using consistent keywords found in experiment file names.
8. Cytometer calibration
Calibrating the flow cytometer lasers prior to each run using a template and beads, with the goal being to increase the laser power to the point where the beads land in the same plot template box before each run. Taking 5 minutes to do this before each run will help you generate more comparable figures down the line. You’re more likely to be able to use the same scale settings across all experiments, and may even be able to use the same compensation matrix, although this is not guaranteed. Read the “Scale settings” section of our Making Beautiful Plots: Data Display Basics post to see how to use single stain compensation controls to set scale settings; you can combine compensation files from multiple experiments, apply a test compensation matrix, and adjust scale settings to see if they can be applied to multiple experiments if cytometer calibration settings were similar enough from day to day.