Fiber photometry is an increasingly popular technique to measure bulk fluorescent signals from a range of biosensors (e.g. GCaMP, dLight, iGluSnFR, etc). Whereas in the early days, photometry rigs were hacked together with parts from various vendors, this method’s popularity has incentivized several commercial entities to offer turnkey solutions for photometry recording. As straightforward as the technical implementation is, it has several limitations (discussed elsewhere or will be addressed here in the future). One challenge I will discuss here is the issue of tuning LED powers for fiber photometry (aka one-pixel imaging). How would you pick that?
Let me just start by saying that you absolutely need to measure the power at the fiber tip. Do not rely on the voltage setting of your LED drivers. Get a decent power meter (I am using this one from Thorlabs), dial the knob on your LED driver, set it on the software, and regularly measure the light power at the end of your patch cable. LED drivers can go bad, so that 1.2V that used to pump 50uW of 470nm light out of your line may now only send 20uW. You would think that your signal is fading, whereas it’s just the LED drivers that are crapping out. Patch cables can also go bad and their transmission coefficients may change in time as you (or your favourite rodents) use them. LEDs themselves are not always stable and need to be checked out once in a while. Long story short: measure your LED powers at least once a week, please!
Back to our conversation. It is tempting to jack up the power until you see a signal. So long as you are not saturating the amplifier on your photodetector, that seems like an ok solution. But not really. High powers can cause tissue heating, will lead to bleaching the fluorophore in your sensor, and may interfere with Optogenetics constructs if you are doing a combinatorial experiment where you are stimulating (a subset of) neurons and recording the neural activity in another subset. But if your excitation power is too low, you may not excite your sensor enough to be able to detect changes. There must be a sweet spot where your power is not causing significant bleaching yet eliciting large enough signals. This depends on several factors such as the type of sensor, region/cell type, duration of your experiment, etc. Here I demonstrate one way I did this for one of my experiments and suggest you do it for each. Find that sweet spot and stick to it for your experiment. Keep it constant across all your subjects to make it possible to do between-subject comparisons and report that number in your publications!
I like Dopamine. So I am using DA measurement using dLight1.3b in the ventral striatum as a poster child here. Unpredicted reward deliveries elicit significant and rapid changes in DA concentration in the ventral striatum. I use this feature to compare this response under different excitation powers. The task details are discussed in a published paper (Mohebi et al. 2019). Briefly, the rat is sitting in a box and at random intervals (15-30s), receives a 45mg sucrose pellet, a food cup that he immediately collects from. A hopper click cues the reward delivery and elicits DA transients. A typical session may look like the one below, where I am recording for ~2 hours.
Aligned to hopper click (reward delivery cue) dopamine transients look like this:
So I repeated this experiment under different LED powers (15, 20, 30, and 40 uW measured at the tip of the patch cable) during the same session and measure DA response to 20 hopper clicks in each condition. The raw output of my photodetector looks something like this:
You might have noticed that I am wasting a large chunk of dynamic range in my photodetector amplifier and that’s fine. My amplifier range goes from 0-5V and I am not even using 10% of it at most, which is completely fine. The signal quality is already quite good. No need to worry here. Folks typically like to quantify photometry response in terms of fractional fluorescence. Let’s use the 405 isosbestic (purple line in Fig. 3). It will look like the below:
Did you also notice the profound decay in the signal over ~10 minutes under the 40uW excitation? Imagine imaging for two hours in that range :). Also, did you pick the initial decay in the first five minutes? I suspect that we are mostly bleaching the tissue immediately underneath the fiber, not the biosensor… At 30uW, I still see some bleaching that is not present under 15 or 20uW excitation. Transients in all four blocks seem comparable. So it turns out we are not losing much when we dial down the power. Let’s take a closer look.
Response magnitudes are comparable and seem not to depend much on the excitation power. Any variability in the response magnitudes seems to be related to behavioral differences between trials (what are these behavioral differences? stay tuned, I am working on the manuscript). I should add that I am not doing proper statistical analysis here, which requires way more data and sound setup, this is just an example. I suggest you do something like this for your own experiments.
So do we have some recommendations? Maybe.
Let me know if I am missing something. I will come back and edit.
Cheers – A