Viral delivery of iGluSnFR in neonates for multi-synapse glutamate imaging in situ
In our previous studies, we introduced optical glutamate sensors in the hippocampal neuropil via stereotaxic viral delivery in young animals [20] or via biolistic transfection in organotypic brain slices [21,23]. However, brain injections in adults face challenges, such as potential interference with the tissue designated for acute slices, whereas the functional morphology of organotypic slices might not fully represent that of intact tissue. We, therefore, sought to explore viral transduction in vivo via neonatal intracerebroventricular (ICV) injections (Fig. 1A), aiming at efficient transgene expression in neurons, for up to six weeks post-infection for subsequent ex vivo imaging.
We employed the new generation of AAV-based sensor variant with a relatively high off-rate, AAV9.hSynap.iGluSnFR.WPRE.SV40, but also used the recently described low off-rate sensor variant, SF-iGluSnFR.A184S [24] for comparison. Although AAV9 appeared to penetrate more readily after ICV administration [25] than did AAV2/1, at three to four weeks post-injection, both methods provided efficient labelling of Schaffer collateral fragments in s. radiatum (Fig. 1C-E). The robust level of expression was maintained for at least six weeks post-infection, which made it suitable for ex vivo experiments in acute slices from young adult animals.
To validate the method, we set out to monitor iGluSnFR fluorescence intensity integrated across the region of interest (ROI, the area incorporating several axonal boutons) during electric stimulation of Schaffer collaterals (five stimuli 50 ms apart; imaging settings as described earlier [20]). For time-lapse imaging, we employed frame-scanning mode providing rapid sampling rate (pixel dwell time 0.5 µs, frame time ~25 ms) across the area of interest (256 x 96 pixels, Fig. 1C). The recorded data sets were arranged as T-stacks, consisting of multiple frame scans (typically 35 to 50, depending on the duration of recording). We thus achieved reliable imaging of the dynamics of glutamate release across the sampled tissue fragment (using a galvo mirror scanhead), with clear separation of five responses to individual electric stimuli (Fig. 1C; fEPSP and ΔF/F0 signal traces; one-trial example). Our attempts to achieve a faster frame rate using a continuous resonant-scanner mode (using a Femtonics Femto-SMART scope) could not obtain a suitable trade-off between laser power and pixel dwell time to generate satisfactory signals without tissue damage, at least under the current protocol. Specific (non-continuous) regimes for resonant-scanner imaging may be required to achieve that.
A similar experiment using the slow-unbinding A184S sensor variant (Fig. 1D) revealed robust stimulus-evoked rises in the iGluSnFR intensity (Fig. 1E). However, this sensor variant did not seem to provide reliable separation between individual responses to five stimuli applied at 20 Hz (Fig. 1E; ΔF/F0 trace, one-trial example), thus pointing to the corresponding limitations in temporal resolution.
Multi-synapse imaging of glutamate release at individual axonal boutons
We next asked if the chosen frame-scanning method is sufficiently sensitive to document glutamate release at individual axonal boutons. We therefore used the recorded image-frame stacks to analyse fluorescence dynamics at small ROIs associated with individual axonal boutons (Fig. 2A). The fluorescence dynamics at individual selected boutons showed that recording sensitivity and signal-noise ratios were sufficient, in principle, to document individual glutamate releases (Fig. 2B; ΔF/F0 traces, four-trial average), at least in baseline conditions. For comparison purposes, we recorded a fragment of the same axon (as Fig. 2A) in linescan mode, which provides high temporal resolution (~1.45 ms). The fluorescence dynamics thus recorded from three boutons of interest (Fig. 3A, bouton numbers as in Fig. 2A; one-trial example) was qualitatively similar to that obtained in the frame-scanning mode (compare boutons 5-7 in Fig. 2B and Fig. 3B).
Imaging glutamate release during LTP induction
One of the main advantages of the frame-scan mode (with galvo mirrors), as opposed to various linescan options, is relatively low overall laser exposure per pixel yet sufficient pixel dwell time to generate enough photons. Firstly, this lowers the propensity for irreversible photo-damage that may occur in cellular structures under intense laser light. Secondly, it reduces photobleaching of the fluorescent indicator, which has been a key prerequisite for stable longer-term imaging. As pointed out above, available parameters of the continuous resonant scanning fell outside the optimal range for the present protocol.
We therefore set out to explore our imaging method in an attempt to document changes, if any, of glutamate release during the high-frequency stimulation (HFS)-induced LTP. The classical LTP induction protocol in iGluSnFR-expressing acute slices produced a reliable increase in the fEPSP slope, lasting for up to 90 min post-induction (example in Fig. 4A). In the selected areas of s. radiatum, we thus identified groups of candidate axonal boutons that responded to afferent stimulation but also remained firmly in focus during the experiment, to reduce any bias associated with focal drift (Fig. 4B). The boutons selected based on this mandatory criterion, were not necessarily the boutons showing the best signal-to-noise ratios of their ΔF/F0 responses (this may also relate to varied iGluSnFR expression). Whilst individual boutons displayed varied effects of LTP induction on the fluorescence dynamics of iGluSnFR, they nonetheless appeared to indicate a clear trend towards an increase in the ΔF/F0 signal amplitude (example in Fig. 4C).
This trend was more prominent when the area-integrated ΔF/F0 signals (as in Fig. 1C, E) were compared (Fig. 5A). To evaluate this quantitatively, we first measured the iGluSnFR signal amplitude {ΔF/F0}, the mean ΔF/F0 value measured over 300 ms after the first stimulus onset (Fig. 5A, traces), 1-5 min prior to LTP induction, and 30 min and 90 min after LTP induction. Comparing these {ΔF/F0} values within individual slices revealed a significant increase after LTP induction (from 3.1 ± 0.9% to 7.2 ± 2.2% at 30 min post-HFS, p < 0.035; to 6.5 ± 1.6% at 90 min post-HFS, p < 0.005; n = 7 slices, paired t-test). Second, we compared full ΔF/F responses at the same time points. To achieve paired comparison, we normalised post-HFS traces by the {ΔF/F0} value of the pre-LTP response, within each individual preparation (slice), and then re-scaled all the traces to match the sample-average {ΔF/F0} value in baseline conditions (Fig. 5C). Again, this paired-comparison design revealed a prominent increase in the ΔF/F0 signal at 30 and 90 min after LTP induction (Fig. 5C). Whether such an increase necessarily indicates a greater amount of evoked glutamate release is discussed below.
Blocking astroglial glutamate transport saturates iGluSnFR signal
Because the ΔF/F0 signal we record reflects glutamate binding to iGluSnFR molecules, it may compete with other (invisible) binding sites for glutamate in the neuropil. It has been well established that, once released from presynaptic boutons in the hippocampus, >90% of glutamate molecules are bound and taken up by high-affinity astroglial glutamate transporters [26]. These transporters will therefore compete with iGluSnFR for glutamate binding and removal from the extracellular space, prompting a hypothesis that their inhibition could boost the iGluSnFR signal. To test this, we added the transporter blocker TFB-TBOA to the bath (50 µM), after recording a reliable ΔF/F0 response 90 min after LTP induction. Within 3 min after TBOA application, afferent stimulation induced a large, virtually irreversible increase in the iGluSnFR ΔF/F0 signal (Fig. 5D). The signal has become undetectable within the next few minutes, most likely due to the progressive saturation of iGluSnFR by the excess of extracellular glutamate in TBOA (Fig. 5D). At the same time, TBOA had little effect on the fast fEPSPs (Fig. 5D, fEPSP traces), reflecting no detectable influence on glutamate release, in line with earlier reports [27,28]. These results indicate that the LTP-associated increase in the iGluSnFR ΔF/F0 signal could potentially be related to the reduced presence of astroglial glutamate transporters in the perisynaptic environment.