Fluorescence lifetime imaging reveals regulation of presynaptic Ca2+ by glutamate uptake and mGluRs, but not somatic voltage in cortical neurons

Abstract Brain function relies on vesicular release of neurotransmitters at chemical synapses. The release probability depends on action potential‐evoked presynaptic Ca2+ entry, but also on the resting Ca2+ level. Whether these basic aspects of presynaptic calcium homeostasis show any consistent trend along the axonal path, and how they are controlled by local network activity, remains poorly understood. Here, we take advantage of the recently advanced FLIM‐based method to monitor presynaptic Ca2+ with nanomolar sensitivity. We find that, in cortical pyramidal neurons, action potential‐evoked calcium entry (range 10–300 nM), but not the resting Ca2+ level (range 10–100 nM), tends to increase with higher order of axonal branches. Blocking astroglial glutamate uptake reduces evoked Ca2+ entry but has little effect on resting Ca2+ whereas both appear boosted by the constitutive activation of group 1/2 metabotropic glutamate receptors. We find no consistent effect of transient somatic depolarization or hyperpolarization on presynaptic Ca2+ entry or its basal level. The results unveil some key aspects of presynaptic machinery in cortical circuits, shedding light on basic principles of synaptic connectivity in the brain.

. In contrast, Ca 2+ signalling at small central synapses, which are difficult to access in situ, has hitherto been explored mainly by monitoring the fluorescence intensity of Ca 2+ -sensitive indicators. The intensity-based approach has been instrumental in relating dynamic changes in presynaptic Ca 2+ to use-dependent plasticity of neurotransmitter release (reviewed in Regehr, 2012;Zucker & Regehr, 2002). However, intensity measures are prone to uncontrolled concomitants, such as changes in local dye concentration, photobleaching, tissue light scattering, or laser power fluctuations. These limitations could be critical for Ca 2+ concentration ([Ca 2+ ]) measurements whereas the accuracy of ratiometric Ca 2+ indicators in optically turbid media, such as brain tissue, is compromised by the strong dependence between the wavelength and scattering/absorption of light. Thus, monitoring [Ca 2+ ] inside individual axons, in particular the nanomolar range basal Ca 2+ levels, has been a challenge.
A breakthrough came with exploring fluorescence lifetime sensitivity of some Ca 2+ indicators to free Ca 2+ (Wilms & Eilers, 2007;Wilms, Schmidt, & Eilers, 2006). As a time-domain measure, fluorescence lifetime imaging (FLIM) is not influenced by light scattering, dye concentration, focus drift or photobleaching. We have recently advanced and validated an approach that optimizes FLIM-based readout of such indicators in experimental settings in situ (Jennings et al., 2017;Zheng, Jensen, & Rusakov, 2018). This method has enabled dynamic monitoring of presynaptic [Ca 2+ ] in individual axons in situ, with nanomolar sensitivity (Jensen et al., 2019;Jensen, Zheng, Tyurikova, Reynolds, & Rusakov, 2017). Here, equipped with this approach, we asked, first, whether the excitatory synapses supplied by individual axons of cortical neurons show evenly distributed functional features of presynaptic Ca 2+ signalling, or whether these features change along the axon. This quest has been an important line of enquiry into fundamental traits of circuit formation and function (Bakkum et al., 2013;Debanne, Guerineau, Gahwiler, & Thompson, 1997;Guerrero et al., 2005;Kukley, Capetillo-Zarate, & Dietrich, 2007).
Finally, our aim was to establish whether somatic depolarization (or hyperpolarization) of the host neuron affects its axonal Ca 2+ signalling. This issue has long been a subject of debate. It has been shown that depolarizing central neurons can boost glutamate release from distant axonal boutons (Alle & Geiger, 2006;Christie, Chiu, & Jahr, 2011;Scott, Ruiz, Henneberger, Kullmann, & Rusakov, 2008;Shu, Hasenstaub, Duque, Yu, & McCormick, 2006). However, axonal Ca 2+ imaging (using fluorescence-intensity measures) has suggested that, in hippocampal granule cells, somatic depolarization reduces spikeevoked presynaptic Ca 2+ entry in proximal axonal segments (Ruiz et al., 2003;Scott et al., 2014) while having no detectable effect distally (Scott et al., 2008). In contrast, in cortical pyramidal cells, somatic depolarization was proposed to boost spike-evoked presynaptic Ca 2+ entry (Christie et al., 2011;Shu et al., 2006) whereas it was presynaptic hyperpolarization that enhanced transmission between cortical or hippocampal pyramidal cells (Rama et al., 2015). The role of the underlying Ca 2+ mechanisms has therefore remained debateable, mainly because of the limitations imposed by the traditional fluorescence intensity-based Ca 2+ measures. We therefore thought that it was important to explore the FLIM-based approach, in the context. IMSR_JAX: 000664) of both sexes (60% male and 40% female) were group housed in a controlled environment as mandated by the locally approved guidelines, on a 12 hr light cycle and with food and water provided ab libitum. This study was not pre-registered.

| Electrophysiology, axon tracing and Tornado scanning in pre-synaptic boutons
We used a Femto2D-FLIM two-photon excitation (2PE) imag- Cat# O6807) for FLIM recordings. Following whole-cell break-in, 40-60 min were allowed for the dyes to equilibrate across the cell, and then the axonal arbour was traced in frame-scan mode, also using z-axis browsing, until the first axonal bouton had been identified as described previously (Jensen et al., 2017). Pre-synaptic imaging was carried out in current clamp mode (V m ≈ −70 mV) using an adaptation of pre-synaptic Ca 2+ imaging methods previously described (Jensen et al., 2017(Jensen et al., , 2019. Cortical neurons requiring compensation current of > 70 pA were discarded before imaging. In the imaging channels, cells demonstrating trial-to-trial fluctuations in the baseline [Ca 2+ ] or evoked [Ca 2+ ] over ~20% were discarded. Once the bouton was identified, position and size of spiral shaped (Tornado) line scans were adjusted to cover the visible bouton profile, and recorded as described below. Depending on the bouton size, one spiral scan typically takes 1-1.5 ms, thus providing readout of axonal fluorescence with high temporal and spatial resolution. Individual action potentials were evoked by a 2 ms pulse of depolarizing current (0.9-1.5 nA), in current clamp mode, as detailed previously (Scott et al., 2014).

| 2PE Tornado-FLIM readout of Ca 2+ concentration in small axonal boutons
In slice preparations, we thus identified and patched pyramidal neurons located in layer 2/3 of the visual cortex. Cell axons were followed, as described above, to focus on individual boutons; during individual trials (typically lasting 2 s), continuous tornado line scans were collected. The scan data were recorded by the standard analogue integration in Femtonics MES (RRID: SCR_018309), and by TCSPC in Becker and Hickl SPCM (RRID: SCR_018310) using dual HPM-100 hybrid detectors. Next, we used the fast-FLIM analysis procedure described previously (Zheng et al., , 2018 to handle individual Tornado scans. We routinely collected and stored FLIM line scan data in a t × x × y × T data cube representing an x-y image with the distribution of nanosecond decay timestamps (t) of individual photons, pixel-by-pixel over the frame duration (T). However, for the purposes of this study, we collapsed all spatial information thus boosting photon counts per scan cycle.
The FLIM data represented therefore the average signal over the bouton area (approximately the entire profile) covered by the scan.
Post-hoc FLIM analyses were performed in a custom-made data analysis toolbox, which is available online (https://github.com/ zheng kaiyu /FIMAS; RRID: SCR_018311). The fluorescence decay curve (lifetime photon counts) was integrated over the 9 ns period post-pulse, and normalized to the maximum value, as detailed earlier (Zheng et al., 2018). Data from up to 5-10 neighbouring pixels were averaged to ensure that the FLIM decay traces had sufficient counts towards the tail of the decay (8-12 ns post-pulse). Data from a single trial were normally sufficient for boutons located closer to the surface of the tissue; for deeper-located boutons, several trials were required to estimate accurately the Ca 2+ dynamics evoked by an AP.

| Estimating action potential evoked presynaptic Ca 2+ entry
The (steady-state) basal presynaptic [Ca 2+ ] 0 was directly estimated from FLIM readout over the averaging interval of ~500 ms before an action potential. However, the rapid rise of presynaptic [Ca 2+ ] (1-2 ms) was faster than the averaging time of FLIM recording (5-10 ms).
Therefore, to improve the signal-to-noise ratio in measuring presynaptic Ca 2+ entry Δ[Ca 2+ ], the spike-evoked peak presynaptic [Ca 2+ ] peak was estimated using both FLIM and intensity recordings as follows.

| Statistical analysis
During axonal tracing with 2PE imaging, axonal boutons were sampled in an arbitrary manner, as they appeared in the focal plane showing distinct varicose morphology and clear action potential induced Ca 2+ responses. No exclusion criteria were applied to animals or slices; unhealthy patched cells were excluded according to the criteria described above. Blinding was not applicable to experimental manipulations during live recording. Thus no strict randomization procedures were applicable during 3D axonal tracing. In experiments comparing independent samples in control condition (branch order comparisons), both two-way ANOVA and conservative non-parametric Kruskal-Wallis ANOVA tests were applied as The sample size was not predetermined because the variability of measured parameters was not known a priori. Shapiro-Wilks tests for normality produced varied results across raw data samples.
Accordingly, we used either the paired-sample t-test, or the pairedsample non-parametric Wilcoxon Signed Ranks test, as indicated.
The statistical software in use was Origin 2019 (Origin Lab; RRID: SCR_014212).

| Monitoring presynaptic [Ca 2+ ] using FLIMbased readout
To calibrate FLIM readout for absolute [Ca 2+ ] measurements on a designated two-photon excitation (2PE) microscopy imaging system, we employed the protocol established for OGB-1 previously (Zheng et al., , 2018. The procedure uses the ratiometric Normalized Total Count (NTC) method in which photon counts are integrated under the lifetime decay curve (over its Ca 2+ -sensitive span), and the result is related to the peak value (Materials and Methods; Figure 1a). The outcome confirmed high sensitivity of the readout in the 0-300 nM [Ca 3+ ] range, providing a quantitative reference to the microscopy measurements (Figure 1b). This calibration outcome was similar to the data set obtained previously for a different 2PE system (Zheng et al., , 2018, arguing for the robustness of the present approach. We next held individual layer 2/3 pyramidal cells in whole-cell mode dialysing them with 300 µM OGB-1, and traced their axons up to a distance of 250-300 µm from the soma, in two-photon excitation (2PE) mode (Figure 1c). Once focussed on individual axonal boutons, we used spiral (tornado) line scan (at 500-1000 Hz) covering the bouton profile (Figure 1d), to record Ca 2+ -sensitive photon count data, before and after triggering a somatic spike (Jensen et al., 2017(Jensen et al., , 2019. With the averaging of the spatial scan data (Methods), this type of recording provides stable photon count acquisition from a small region of interest during repeated trials over ~20 min (Figure 1e). This was consistent with the previously documented FLIM recording stability, in similar settings, for up to 60 min (Jensen et al., 2019;Zheng et al., 2018). Thus, decoding the recorded FLIM data provided robust traces of resting basal [Ca 2+ ] 0 and spikeevoked presynaptic [Ca 2+ ] dynamics, in the 10-300 nM range, for boutons located at axonal branch orders 1-3, at different distances from the soma (Figure 1f).

| Subthreshold somatic depolarization (or hyperpolarization) has no consistent effect on [Ca 2+ ] 0 or Δ[Ca 2+ ]
To understand the effect of somatic depolarization on presynaptic

| D ISCUSS I ON
In the present study, we employed an imaging method that could detect changes in presynaptic [Ca 2+ ] with virtually nanomolar sensitivity in the concentration range between 10-300 nM (Zheng et al., , 2018. We have documented average [Ca 2+ ] 0 values in baseline conditions between 30 and 60 nM, which is consistent with earlier highsensitivity Ca 2+ measurements in neuronal processes (Canepari, Vogt, & Zecevic, 2008;Helmchen, Imoto, & Sakmann, 1996), including axons (Ermolyuk et al., 2013), that employed alternative Ca 2+ imaging methods. Similarly, the range of Δ[Ca 2+ ] between 50 and 300 nM reported here corresponds to the equilibrated presynaptic [Ca 2+ ] after a very brief (~1 ms) local 'hotspot' entry, and is fully in line with previous estimates based on fluorescence-intensity measures (Ermolyuk et al., 2013;Helmchen et al., 1996;Rusakov, Saitow, Lehre, & Konishi, 2005;Scott & Rusakov, 2006). However, the FLIMbased method has several advantages over previous approaches, which enables us to explore presynaptic [Ca 2+ ] dynamics in greater detail, as discussed earlier (Wilms et al., 2006;Zheng et al., 2018;. The quest to identify a systematic pattern of functional synaptic features along the axon has been an important line of enquiry into fundamental traits of circuit formation and function (Bakkum et al., 2013;Debanne et al., 1997;Guerrero et al., 2005;Kukley et al., 2007). One of the most common questions asked in this context has been whether the increasing sparsity of longer cell-cell connections in the cortex is compensated by their increased synaptic efficacy. We have recently employed multiplexed imaging of glutamate release and presynaptic Ca 2+ in organotypic brain slices to find that [Ca 2+ ] 0 and Δ[Ca 2+ ] are positively correlated with release probability (Jensen et al., 2019). Thus, the present data appear to argue against increased release efficacy with greater distances from the soma, but they do support the idea that in cortical pyramidal cells, axonal branches of higher orders host more efficient release sites (Figure 2). Clearly, imaging glutamate release at individual axonal boutons should provide further clarity on the subject.
However, no known time-resolved (FLIM-based) optical sensors of glutamate are available at present. Therefore, to gauge accurately glutamate release efficacy in the turbid medium of acute cortical slices or in vivo, a special effort would be required to avoid multiple concomitants of the fluorescence intensity signal, for its unbiased interpretation.
We have found that the blockade of the group 1 mGluRs, which occur in cortical axons (Cartmell & Schoepp, 2000;Gereau & Conn, 1995), reduces presynaptic basal [Ca 2+ ], suggesting that these receptors are constitutively active, in a glutamate-independent manner. These receptors are known to trigger a powerful molecular cascade initiating local IP 3 -receptor dependent release from Ca 2+ stores, both in neurons (Pinheiro & Mulle, 2008;Reiner & Levitz, 2018) and in astroglia (Bazargani & Attwell, 2016;Verkhratsky & Kettenmann, 1996), and their ligand-independent persistent activity has long been known documented (Ango et al., 2001 in their axons (Figure 4). Previous studies in cortical pyramidal cells and hippocampal granule cells have shown that somatic depolarization enhances release probability in their axons (Alle & Geiger, 2006;Christie et al., 2011;Scott et al., 2008;Shu et al., 2006). However, in the hippocampus, subthreshold somatic excitation had no effect on Δ[Ca 2+ ] in remote (giant) boutons (Scott et al., 2008) although it did inhibit Δ[Ca 2+ ] in proximal axonal segments (Ruiz et al., 2003;Scott et al., 2014). In contrast, in cortical pyramidal cells, the fluorescent intensity readout of intracellular Fluo-5F (K d ~ 2.3 µM) (Christie et al., 2011), or indirect tests with Ca 2+ chelators in the presynaptic cells (Shu et al., 2006), led to a conclusion that somatic depolarization should boost Δ[Ca 2+ ]. It might be important to establish reasons for the disparity between the present data and the previous observations. All experiments were conducted in compliance with the ARRIVE guidelines.

CO N FLI C T O F I NTE R E S T
The authors declare no known conflict of interest.