It is common in phospho-flow and other single cell signaling experiments to compare responses to stimulation for a given sample.  In addition it can be important to know whether signaling started particularly high or low.  This page addresses some issues around calculating fold change and making basal comparisons in signaling experiments like these.

Why measure both basal and induced signaling?

Measuring both fold change and basal level in healthy and diseased cells will allow you to ask questions like:

1. Is there an abnormally high level of signaling response to an environmental cue in the diseased cells (e.g. gain of oncogenic signaling)?
2. Alternatively, are some signaling pathways failing to activate in diseased cells (e.g. loss of tumor suppressor signaling)?
3. Is there a normal level of signaling observed, but in an abnormal context, such as an extremely low dose of input signal (e.g. hypersensitivity, as in Kotecha et al., Cancer Cell 2008)?
4. Is signaling rerouted (shifted from one effector phospho-protein to another) in the diseased cells?
5. Is a sample’s lack of response to stimulation due to higher than normal basal signaling?
6. Is greater variation in basal phosphorylation seen among the disease samples than there is among the healthy samples?
7. Is a pattern of high basal phosphorylation in diseased cells comparable in magnitude with the signaling response seen in a healthy cell?

We recently reviewed this and other questions in the context of cancer biology for a book chapter:

Doxie DB and Irish JM.  High-Dimensional Single-Cell Cancer Biology. Current Topics in Microbiology & Immunology 2014 Mar 27. (PDF – Doxie and Irish – High Dimensional Single Cell Analysis)

In my Ph.D. thesis study of leukemia cell signaling (Irish et al., Cell 2004), if we had only looked at basal signaling, we would have missed the ‘big finding’ that a profile of signaling stratifies initial chemotherapy response.  Both basal signaling and responses to stimulation were valuable in stratifying outcomes.  An analysis of the data from that study from the point of view of basal signaling only and the full profile is included in this slide deck (2012-05 Journal Club – Irish Cell 2004).

Calculating distance (fold change)

We recommend using a transformed ratio to calculate fold change or other distances between fluorescence intensity values.  In a typical example, the fold change for stimulated vs. unstimulated (resting) cells on a FACSCalibur would be calculated as:

log₁₀ fold change in MFI for stimulated cells vs. unstimulated cells =
log₁₀(stim. MFI) – log₁₀(unstim. MFI)

Note that for this case, this is the same as writing:

log₁₀(stim. MFI / unstim. MFI)

Basal comparisons

Sometimes it it important to know whether signaling started particularly high or low.  How do we compare basal levels of signaling (by phospho-flow)?

Generally, we suggest comparing basal values to each other (between samples) using the same comparison technique and scale as for fold change between stimulated and unstimulated samples.  The key benefit of this is that the basal values will be placed on the same scale as the fold change values, making it easy to compare them.

We recommend using median fluorescence intensity (MFI) for comparing basal vs. basal, although 95th percentile might be good in some situations. The key part is to compare basal values using the same statistic and on the same distance scale as the fold change values. In an experimental sense, the fold change values ask, “When I stimulated this, how far did it move from where it was when it was resting?” For the resting values, the question could be, “How far is the baseline for this disease sample from a healthy control?” The key thing is that the comparison type be the same for both questions so that you can compare between them later.  Also, it gets confusing and tends to send a red flag to reviewers if there are lots of different comparisons / transformations / normalizations in the same study.

For example, we recommend picking one healthy sample’s basal measurement to be the control for all the samples in a particular study (other healthy samples, other cancer samples, etc.). Then do a fold difference between any case’s basal vs. this healthy control sample’s basal.