I came across a paper recently that might cause us to rethink in vivo GCaMP recording, by validating fibre photometry. Or at least to rethink our interpretation of the data based on assumed fidelity of the GCaMP signal to neuronal spiking activity.
I’m talking about a recent submission to the BioArchive preprint server from the lab of Lex Kravitz1. Its title “Fiber photometry does not reflect spiking activity in the striatum” speaks for itself, and counters what one might suppose is the whole point of doing photometry recordings in the first place.
Before we get on to their paper validating fibre photometry, a quick bit of background about neuronal calcium signalling. Depolarisation of a neurone opens the voltage-dependent calcium channels, causing influx of Ca2+ and an increase in intracellular calcium – this is in fact what triggers vesicle fusion at the nerve terminal for release of neurotransmitters. Therefore, it makes sense to use [Ca2+] as a marker of neuronal activity.
However, it is easy to forget when using a surrogate marker in this way, such as using c-fos as a marker for neuronal activity following a stimulus, that there is no guarantee of a causal relationship. It is important, therefore, that these assumptions are put to the test, as Legaria et al. have done in this work. They only present 2 figures, so I will quickly go through the salient points.
To achieve this, Legaria et al. transfected mice with GCaMP6 and implanted a Mightex OASIS implant for calcium imaging of individual cells (Figure 1; the Mightex system looks interesting in that it can do optogenetics, multiunit photometry and miniscope imaging, I’ll try and do a blog post on it in the future).
Anyway, Legaria et al. analysed three signals from each recording: the GCaMP-transfected cell bodies, the surrounding background from non-transfected tissue (“neuropil”), and the overall signal (which would correlate to the signal seen from photometry which simply takes a global readout from the illuminated area; Figure 1C). Analysing these three signals alongside, they found that the “photometry” signal correlated extremely well with the “background” calcium signal, but not at all with the “cell body” signal (Figure 1D).
This is somewhat alarming for someone who performs photometry recordings, that your photometry signal should be detecting fluctuations in “background” rather than identifiable cells. To check whether this (mis)correlation also extended to the electrical activity of the cell, Legaria et al. next performed photometry combined with multi-electrode array recordings of the GCaMP-transfected cells. They saw essentially zero correlation between the two (Figure 2).
Now, before we go and chuck our £60k-worth of photometry recording equipment in the incinerator, I will say that this is a pre-print, so hasn’t been peer reviewed yet. The major questions I would pose as a referee to this work are:
- The electrophysiology array looks to be very widespread (and lopsided) compared to the photometry fibre (Figure 2A) – are we sure the two methods will be picking up signals from the same neurones?
- The electrophysiology records non-discriminantly from any nearby firing cell, so how well can we expect that to correlate with the GCaMP signal, which always label only a subset of those cells (even with non-cre-dependent expression and a global promoter, you will never transfect 100% of cells)
Having said that, the data looks solid to me, so I think it’s safe to assume the conclusions are sound. My further thoughts I had while looking at the data were also brought up in the text:
“There are also some limitations to our study. First, the results presented here are limited to recordings in the striatum. Striatal neurons have extensive dendritic arbors, which may accentuate the neuropil contribution to the fiber photometry signal. Fiber photometry signals from other brain structures may result in different conclusions concerning the relative contributions of somatic versus neuropil calcium. Second, our recordings were performed with GCaMP6s, a variant of GCaMP that has slow dynamics. Different relationships may be observed with GCaMP variants with faster kinetics, or those that target GCaMP to specific cellular compartments”1
This highlights exactly what also occurred to me, that these results may not extend to other brains areas, although I think it would be unsafe to assume they don’t. Also, if we want to mitigate the issues raised here, we should think about using miniscopes and/or use cell body-targeted GCaMP variants. In the meantime, however, I will continue doing photometry recordings as before, but I will keep this work in mind when analysing photometry data. I’m glad this work was done though, it’s good to see someone validating fibre photometry.
1. Legaria et al., BioArchive (2021) Fiber photometry does not reflect spiking activity in the striatum. doi: https://doi.org/10.1101/2021.01.20.427525