Relation between activity-induced intracellular sodium transients and ATP dynamics in mouse hippocampal neurons

Excitatory neuronal activity results in the inﬂux of Na + through voltage- and ligand-gatedchannels.RecoveryfromaccompanyingincreasesinintracellularNa + concentrations ([Na + ] i ) is mainly mediated by the Na + /K + -ATPase (NKA) and is one of the major energy-consuming processes in the brain. Here, we analysed the relation between different patterns of activity-induced [Na + ] i signalling and ATP in mouse hippocampal CA1 pyramidal neurons by Na + imaging with sodium-binding benzofurane isophthalate (SBFI) and employing the genetically encoded nanosensor ATeam1.03 YEMK (ATeam). In situ calibrations demonstrated a sigmoidal dependence of the ATeam F¨orster resonance energy transfer ratio on the intracellular ATP concentration ([ATP] i ) with an apparent K D of 2.6 m M , indicating its suitability for [ATP] i measurement. Induction of recurrent network activity resulted in global [Na + ] i oscillations with amplitudes of (cid:2) 10 m M , encompassing somata and dendrites. These were accompanied by a steady decline in [ATP] i by 0.3–0.4 m M in both compartments. Global [Na + ] i transients, induced by afferent ﬁbre stimulation or bath application of glutamate, caused delayed, transient decreases in [ATP] i as well. Brief focal glutamate application that evoked transient local Na + inﬂux into a dendrite, however, did not result in a measurable reduction in [ATP] i . Our results suggest that ATP consumption by the NKA following global [Na + ] i transients temporarily overrides its availability, causing a decrease in [ATP] i . Locally restricted Na + transients, however, do not result in detectable changes in local [ATP] i , suggesting that ATP production, together with rapid intracellular diffusion of both ATP and Na + from and to unstimulated neighbouring regions, counteracts a local energy shortage under these conditions.

r Calibration of ATeam in situ enabled a quantitative estimate of changes in intracellular ATP concentrations.
r Different paradigms of stimulation that induced global Na + influx into the entire neuron resulted in decreases in [ATP] in the range of 0.1-0.6 mM in somata and dendrites, while Na + influx that was locally restricted to parts of dendrites did not evoke a detectable change in dendritic [ATP].
r Our data suggest that global Na + transients require global cellular activation of the Na + /K + -ATPase resulting in a consumption of ATP that transiently overrides its production.
r For recovery from locally restricted Na + influx, ATP production as well as fast intracellular diffusion of ATP and Na + might prevent a local drop in [ATP].

Introduction
Excitatory electrical signalling of neurons and the generation of action potentials and excitatory postsynaptic potentials mainly relies on the influx of Na + , which is driven by a large inwardly directed Na + gradient. In addition, many other cell functions (e.g. regulation of intracellular Ca 2+ or pH) are directly or indirectly related to the Na + gradient. The Na + gradient across neuronal plasma membranes, however, is not stable, but changes under physiological and pathophysiological conditions (Rose, 2002;Somjen, 2004). Indeed, as opposed to the long-held view that electrical signalling does not cause relevant fluctuations in intracellular Na + concentrations ([Na + ] i ), research over the past years has established that active central neurons experience transient [Na + ] i elevations. These are predominately caused by Na + influx through voltage-gated channels and ionotropic glutamate receptors and are especially prominent in small cellular compartments such as axons and spiny dendrites (Rose & Konnerth, 2001;Miyazaki & Ross, 2017;Ona-Jodar et al. 2017).
Recovery from global [Na + ] i increases is mainly accomplished by the plasma membrane Na + /K + -ATPase (NKA), which is the major cellular consumer of ATP (Erecinska & Silver, 1994). NKA exchanges three intracellular Na + for two extracellular K + (Sweadner, 1995;Kaplan, 2002). Another primary transporter at the plasma membrane is the Ca 2+ -ATPase, which mediates recovery from activity-induced influx of Ca 2+ from pre-or postsynaptic compartments (Clapham, 2007). These and other transport processes that maintain ion gradients across the plasma membrane are responsible for most of the brain's energy consumption. Altogether, estimations indicate that ion movements directly related to action potentials require 20-30% of cellular ATP generation, while about 50% is allocated for postsynaptic ion fluxes following synaptic transmission (Lennie, 2003;Howarth et al. 2012;Engl & Attwell, 2015).
Despite their considerable need for ATP, neurons lack significant energy stores and are therefore dependent on a constant supply with blood-derived glucose and oxygen to feed their ATP production by mitochondrial respiration and glycolysis. Unsurprisingly, former work established a direct link between cellular energy metabolism, NKA activity and Na + regulation, demonstrating that metabolic failure results both in large shifts in neuronal [Na + ] i and a substantial decrease in intracellular ATP concentrations ([ATP] i ) with secondary breakdown of other ion gradients and detrimental consequences (Hansen, 1985;Erecinska & Silver, 1994;Leng et al. 2014;Gerkau et al. 2017). In addition to ATP-dependent ion transporters, many neurons express K ATP channels that couple their metabolic state to their excitability and which are activated by a more moderate decline in cellular [ATP] (Proks & Ashcroft, 2009;Tanner et al. 2011;Mollajew et al. 2013;Martinez-Francois et al. 2018).
The conditions under which excitatory activity and accompanying Na + influx may actually affect levels of ATP or other metabolites in neurons in the brain are, at present, only partly understood. To answer these questions and gain more insight into brain metabolism, protein-based, genetically encoded fluorescent indicators are increasingly employed (Tantama et al. 2012). Here, we have used the Förster resonance energy transfer (FRET)-based nanosensor ATeam1.03 YEMK (ATeam) (Imamura et al. 2009), which has already been successfully employed to analyse ATP dynamics in neurons and astrocytes in cell culture and in brain tissue (Trevisiol et al. 2017;Winkler et al. 2017;Baeza-Lehnert et al. 2019;Lerchundi et al. 2019). Using this tool, we studied the relationship between different patterns of activity, Na + transients and [ATP] in somata and dendrites of CA1 pyramidal neurons in mouse hippocampal tissue slices. In our study, we also provide the first calibration of ATeam inside neurons, suggesting that this sensor is suitable to detect changes in neuronal [ATP].

Ethical approval
The present study was carried out in strict accordance with the institutional guidelines of the Heinrich Heine University Düsseldorf as well as the European Community Council Directive (2010/63/EU). All experiments using brain slices were communicated to and approved by the Animal Welfare Office at the Animal Care and Use Facility of the Heinrich Heine University Düsseldorf (institutional act no. O50/05). In accordance with the recommendations of the European Commission (Close et al. 1997), animals up to 10 days old were killed by decapitation, while older mice were anaesthetized with CO 2 and then quickly decapitated.
To induce disinhibition and recurrent network activity, slices were perfused with a saline nominally devoid of Mg 2+ , to which the GABA A receptor blocker bicuculline was added at a concentration of 10 µM. Electrical stimulation of Schaffer collaterals was performed using a glass micropipette filled with HEPES-buffered saline. The pipette was positioned into the stratum radiatum at a distance of 100-150 µm from the cells of interest and a train of stimuli (10 pulses at 50 Hz) was delivered using a stimulus generator (Isolated Stimulus Generator Model 2100, A-M Systems, Carlsborg, WA, USA). Bath application of glutamate was performed by perfusing the slices with ACSF to which 1 mM glutamate was added. For focal pressure application onto dendrites, glutamate was dissolved at 1 mM in Hepes-buffered saline composed of (in mM): 125 NaCl, 2.5 KCl, 2 CaCl 2 , 2 MgCl 2 , 1.25 NaH 2 PO 4 and 25 HEPES, pH 7.4 (adjusted with NaOH). This solution was filled into a fine-tipped glass pipette and delivered using a pressure application system (PDES nxh, npi electronic, Tamm, Germany).
All chemicals were purchased from Sigma-Aldrich (Munich, Germany), unless otherwise noted. β-Escin for permeabilization of plasma membranes was obtained from mpbio (Illkirch, France). J Physiol 597.23 Field potential recordings and determination of extracellular sodium Field potential recordings were performed at 1 kHz in the stratum radiatum of the CA1 region using sharp glass microelectrodes filled with ACSF coupled to an Axopatch 200B amplifier and a Digidata 1322A (Molecular Devices, Sunnyvale, CA, USA). Data traces were filtered using a Savitzky-Golay filter and analysed with OriginPro 8.5G software (OriginLab Corp., Northhampton, MA, USA).
Extracellular Na + concentrations ([Na + ] o ) were determined using concentric, Na + -sensitive microelectrodes as described earlier (Haack et al. 2015). Briefly, Na + -sensitive microelectrodes were prepared from thin-walled borosilicate glass capillaries with filament. The Na + -sensitive capillary was filled with liquid neutral ion carrier based on ETH 157 (Ionophore II, Cocktail A, Fluka, Buchs, Switzerland). The reference electrode was filled with HEPES-buffered saline. Electrodes were positioned ß50 µm below the slice surface into the CA1 stratum radiatum. They were calibrated in the experimental bath directly before and after each individual experiment using HEPES-buffered saline containing defined sodium concentrations ([Na + ]) ranging from 70-160 mM (with Na + replaced by N-methyl-D-glucamine chloride (NMDG-Cl) to maintain iso-osmolarity).

Sodium imaging
For imaging of [Na + ] i in CA1 pyramidal neurons, the membrane-permeant form of SBFI (SBFI-AM; 200 µM, Teflabs, Austin, TX, USA) was pressure-injected into the stratum radiatum and stratum oriens as reported before (Meier et al. 2006;Langer et al. 2017). Alternatively, individual CA1 neurons were subjected to whole-cell patch-clamp and loaded with the membrane-impermeant salt of SBFI (Teflabs) through the patch pipette for at least 15 min before starting imaging experiments. The pipette solution contained (in mM): 145 KMeSO 3 , 40 KCl, 12.5 Hepes, 10 NaCl, 5 Mg-ATP, 0.5 Na 3 -GTP, and 1 SBFI; pH was adjusted to 7.3. Neurons were held in the voltage-clamp mode at −70 mV (liquid junction potential not corrected) using an EPC10 amplifier (HEKA Elektronik, Lambrecht, Germany) and PatchMaster software (HEKA Elektronik). In some experiments, the patch pipette was carefully retracted after sufficient dye-loading of fine dendrites and Na + imaging was commenced after ß15 min to allow re-sealing of the cell membrane.
Multiphoton imaging of SBFI was performed employing a custom-build laser-scanning system based on a FluoView300 (Olympus Europe, Hamburg, Germany) coupled to a femtosecond-pulsed infrared laser (MaiTai; Spectra Physics, Darmstadt, Germany), equipped with a water immersion objective (NIR Apo ×60/NA 1.0, Nikon Instruments Europe, Düsseldorf, Germany). Excitation wavelength was 800 nm, fluorescence emission was collected <700 nm. Images were acquired at 1-5 Hz and fluorescence emission from selected regions of interest (ROIs) was analysed offline using FluoView 5.0 (Olympus Europe) and OriginPro software. After each experiment, Z-stacks were generated at a step size of 1 µm for reconstruction of entire cells and at 0.2 µm for reconstruction of dendrites. Images were post-processed employing deconvolution software to delineate cellular morphology (Huygens Professional, SVI imaging, Hilversum, Netherlands).

Determination of intracellular ATP
For expression of the genetically encoded sensor ATeam (Imamura et al. 2009) in cultured organotypic slices, an adeno-associated viral vector (AAV2/5) containing the sequence of the sensor under the human-synapsin1 promoter (packaged by the Viral Vector Facility of the University/ETH Zürich, Switzerland) was employed as described previously (Lerchundi et al. 2019b). Briefly, 0.5 µl of the vector (virus titre: 1.375 × 10 12 viral genome ml −1 ) was directly applied on top of slices at 1-3 days in vitro, and then kept in culture for at least 7 more days before use. Alternatively, for imaging of ATP levels in acute tissue slices, transgenic mice expressing ATeam under the murine Thy1.2 promoter were used (Trevisiol et al. 2017).
Fluorescence emission was detected from cellular ROIs expressing the sensor using NIS-Elements software (Nikon GmbH Europe). For background correction, a region of interest (ROI) in the field of view apparently devoid of cellular structures expressing the sensor was chosen and its signals subtracted from those derived from defined cellular ROIs. The ATeam ratio was calculated by dividing the fluorescence emission of Venus (FRET acceptor) by the eCFP fluorescence (FRET donor) and normalizing it to the baseline under control conditions (Fig. 1D).
Changes in the ATeam ratio were converted into changes in intracellular [ATP] based on a calibration procedure performed in organotypic slice preparations as described in the Results section. To mimic intracellular ion composition, a saline composed of 25 mM NaCl, 129 mM potassium gluconate, 10 mM HEPES and 10 mM MgCl 2 , titrated to pH 7.3, was used during these calibrations.

Data analysis and statistics
Unless otherwise specified, data are presented in Tukey box-and-whisker plots indicating median (line), mean (square), interquartile range (IQR; box) and standard deviation (whiskers). In addition, all individual data points are shown in grey underneath the Tukey plots. Data were statistically analysed by either Student's t-test or one-way ANOVA followed by post hoc Bonferoni test.
The following symbols are used to illustrate the results of statistical tests in the figures: * : 0.01 ࣘ P < 0.05; * * : 0.001 ࣘ P < 0.01; * * * : P < 0.001. n represents the number of cells or individual dendrites analysed; N represents the number of individual experiments/slice preparations. Each series of experiments was performed on at least three different animals.

Quantitative determination of [Na + ] and [ATP] in neurons
The goal of the present study was to analyse the relation between changes in [Na + ] i and [ATP] i evoked by different patterns of activity in principal neurons. To this end, we employed quantitative multi-photon Na + imaging with the fluorescent dye SBFI in CA1 pyramidal neurons of mouse hippocampus. Calibration of SBFI fluorescence was performed in CA1 neurons of organotypic slice culture (n = 20-46 individual cells and N = 3 tissue slices for each [Na + ] tested; see Fig. 1A) and followed established procedures as described in detail previously (Rose et al. 1999;Meier et al. 2006;Ona-Jodar et al. 2017). The relation between normalized changes in SBFI fluorescence and [Na + ] followed Michaelis-Menten kinetics, revealing an apparent K D of 42.1 mM (Fig. 1B). A baseline [Na + ] i of 10 mM, corresponding to the [Na + ] of the intracellular saline, was assumed for cells held in whole-cell patch-clamp. For neurons in bolus-loaded slices, we assumed a baseline [Na + ] i of 13.4 mM according to our earlier study (Mondragao et al. 2016). Based on these parameters and the in situ calibration, changes in SBFI fluorescence observed upon different pattern of stimulations were normalized to baseline levels and converted into changes in [Na + ] i in the following. For measurement of [ATP] i in CA1 pyramidal neurons, we employed the genetically encoded sensor ATeam (Imamura et al. 2009). While in former work (e.g. Trevisiol et al. 2017;Winkler et al. 2017;Lerchundi et al. 2019b), ATeam was employed to monitor relative changes in cellular ATP levels, we aimed to establish a more quantitative approach by performing an intracellular calibration. To this end, ATeam was expressed under the human-synapsin1 promoter in organotypic tissue slices. As reported before (Lerchundi et al. 2019a,b), this resulted in a specific transduction and expression of ATeam in neurons. Cell bodies of individual CA1 pyramidal neurons and adjacent primary dendrites could be distinguished based on the fluorescence emission of the donor (eCFP) and the acceptor (Venus) (Fig. 1C). Their normalized ratio ( ATeam ratio) was then calculated as described below to reveal changes in cellular ATP following a procedure introduced previously ( Fig. 1D and E) (Lerchundi et al. 2019b).
For calibration of ATeam, cultured slices expressing the sensor were first superfused with a saline containing 50 µM β-escin, a saponin which induces pore formation and permeabilization of plasma membranes in a dose-dependent manner (Konishi & Watanabe, 1995;Akagi et al. 1999). Subsequently, slices were exposed to ATP-free saline containing low [Na + ] and high [K + ] to mimic intracellular concentrations of these major cations. In addition, this saline contained NaN 3 (5 mM; an inhibitor of the mitochondrial complex IV), 2-deoxy-D-glucose (2-DG, 2 mM; a non-metabolizable glucose analogue) and Na 3 VO 4 (0.1 mM; a non-selective ATPase inhibitor) to minimize the cellular production and consumption of ATP. Afterwards, defined concentrations of ATP ranging from 1 to 10 mM were added to the nominally ATP-free saline. Changes in Venus and eCFP fluorescence were recorded from cell bodies of CA1 pyramidal neurons and converted into changes in the ATeam ratio (n = 14-20 individual cells and N = 3-5 tissue slices for each ATP concentration tested; see Fig. 1D and E). These were normalized to the ATeam ratio at 0 mM ATP, and the resulting ATeam ratio values plotted against the [ATP], after which data were fit with a Michaelis-Menten function, revealing an apparent K D of 2.63 mM (Fig. 1F).
Addition of β-escin and exposure to calibration salines caused a swelling of neurons and resulted in tissue movement and changes in focus. The required re-focusing prevented a reliable deduction of cellular baseline [ATP] in individual cells from these calibrations. Based on earlier studies, which determined baseline [ATP] of ß2 mM in different neurons (Fukuda et al. 1983;Ainscow et al. 2002;Mollajew et al. 2013;Rangaraju et al. 2014;Toloe et al. 2014;Pathak et al. 2015), data derived from calibrations were fit linearly for [ATP] between 1 and 3 mM (R 2 = 0.76) (Fig. 1F). Of note, for decreases in [ATP] by a maximum of 0.5-0.6 mM as detected in the present study (see below), this linearization is applicable for baseline ATP levels between ß1.5 and 3 mM. This enabled a conversion of the ATeam signals into changes in [ATP] in the following, with a 10% change in the ATeam ratio corresponding to a change in [ATP] by 0.57 mM.
Taken together, these results confirm that SBFI is a suitable dye for the imaging of activity-related changes in [Na + ] i as established before. In addition, our data demonstrate that ATeam can be calibrated in situ and, when expressed in pyramidal neurons, exhibits an apparent K D of 2.6 mM for ATP. This implies that ATeam can be employed for quantitative imaging of changes in neuronal [ATP].

Network [Na + ] oscillations induce a decrease in neuronal [ATP]
To study if activity-related changes in neuronal [Na + ] i are accompanied by changes in intracellular [ATP] in our preparation, we first employed an established paradigm for induction of recurrent network activity. To this end, acute hippocampal slices were perfused with nominally Mg 2+ -free saline containing 10 µM bicuculline methiodide (0Mg 2+ /BIC) to unblock NMDA receptors and reduce inhibition by GABA A receptors. As reported before (Karus et al. 2015), this resulted in regular changes in the field potential recorded from the stratum radiatum, indicating burst activity of the neuronal network within a few minutes upon perfusion with 0Mg 2+ /BIC saline (N = 5) ( Fig. 2A). Large epileptiform bursts lasted ß5 s, hosted several population spikes and occurred at a frequency of 1-2 per minute. In between, smaller and shorter biphasic events occurred every ß20 s ( Fig. 2A).
Regular electrical bursting activity was accompanied by rapid transient dips in the extracellular [Na + ] ([Na + ] e ) in the stratum radiatum as determined using concentric Na + -sensitive microelectrodes. These averaged 2.2 ± 0.7 mM in the time window between 10 and 15 min after starting the perfusion with 0Mg 2+ /BIC (range: 1.0-4.5 mM; median: 2.1 mM; n = 46 signals, N = 5) (Fig. 2B). Moreover, apparent baseline [Na + ] e steadily decreased by about 3 mM during a 30 min exposure to 0Mg 2+ /BIC saline (Fig. 2B). The slow decrease in [Na + ] e essentially mirrored the slow and steady increase in neuronal [Na + ] i by ß2 mM reported to occur under these conditions (Karus et al. 2015).
In acute tissue slices derived from transgenic mice, expressing ATeam in neurons (N = 7) (Trevisiol et al. 2017), perfusion with 0Mg 2+ /BIC resulted in a slow decline in the ATeam ratio in both somata and primary apical dendrites of CA1 pyramidal neurons, which partly recovered upon reperfusion with standard ACSF (Fig. 2F). After 30 min of 0Mg 2+ /BIC, [ATP] in cell bodies had decreased by 0.30 ± 0.16 mM (range: 0.04-0.58 mM, median: 0.29 mM; n = 103) (Fig. 2G). In dendrites, [ATP] decreased by 0.39 ± 0.16 mM (range: 0.13-0.64 mM, median: 0.42 mM; n = 36 dendrites), which is a significantly larger decline as compared to somata (P = 0.007) (Fig. 2G). Notably the ATeam ratio does not change under control conditions within this time frame (Lerchundi et al. 2019a), indicating that the slow decline observed in the present study was indeed related to the ongoing activity induced by the dis-inhibition.
In summary, these experiments demonstrate that recurrent network activity and global network Na + oscillations in somata and dendrites of CA1 pyramidal neurons are accompanied by a slow and steady decrease in [ATP] in these compartments that amounts to 0.3-0.4 mM after 30 min.

Changes in [Na + ] i and [ATP] i evoked by afferent fibre stimulation
With recurrent network activity and ongoing global Na + oscillations, a slow decrease in neuronal [ATP] over time was observed. Next, we analysed if distinct transient increases in [Na + ] i are accompanied by detectable changes in [ATP] i . Earlier work has shown stimulation of glutamatergic afferents results in transient increases in [Na + ] i in cell bodies and dendrites of CA1 pyramidal neurons mainly due to Na + influx through ionotropic glutamate receptor channels (Rose & Konnerth, 2001;Langer & Rose, 2009;Miyazaki & Ross, 2017). To induce transient Na + influx into neurons, we thus performed a burst stimulation of afferent Schaffer collaterals (10 stimuli at 50 Hz) at a distance of approximately 100-150 µm to the analysed cells or ROIs in acute hippocampal slices.
An overlay and peak normalization of evoked [Na + ] i transients and changes in [ATP] i visualizes the time courses of both signals (Fig. 3F). It illustrates that changes in [ATP] basically start in parallel to those in [Na + ] in a given compartment. Their peaks, however, are considerably delayed and reached only after [Na + ] i has already partially recovered to baseline. Recovery of [ATP] i is essentially not seen before [Na + ] i has fully recovered, and thus lags many minutes behind.
Overall, these results show that a single burst stimulation of afferent fibres results in a transient increase in [Na + ] in CA1 pyramidal neurons, which exhibits a significantly larger amplitude and faster kinetics in dendrites as compared to cell bodies. Moreover, the stimulation evokes a transient decrease in [ATP] in dendrites and somata. Amplitude and time course of the drop in [ATP] i essentially mirror those in [Na + ] in the respective compartment, albeit with much slower kinetics.

Changes in [Na + ] i and [ATP] i induced by application of glutamate
The results presented so far demonstrated that somatic and dendritic [Na + ] i transients resulting from synchronized excitatory input into CA1 pyramidal neurons are accompanied by delayed decreases in [ATP] i in both compartments. We next tested the effect of a direct application of glutamate on intracellular [ATP] i . These experiments were performed in the presence of TTX (0.5 µM) to prevent the generation of action potentials.
Again, an overlay of peak-normalized averaged traces revealed that changes in [ATP] i induced with bath application of glutamate began with those in [Na + ] i in each compartment, but exhibited a significantly delayed peak and displayed much slower kinetics. To better visualize the apparent coincidence of the first phase of both signals, we calculated the first derivative of the normalized changes in [ATP] i . The resulting trace (green trace in Fig. 4E) clearly delineates that the decline in [ATP] i was indeed coincident with the rise in [Na + ] i and slowed down with the recovery of [Na + ] i to baseline, indicative of a declining consumption of ATP in both compartments.
In a second set of experiments, glutamate (1 mM) was applied focally through a fine glass pipette positioned   close (ß10 µm) to apical dendrites of CA1 neurons in organotypic slices to evoke local Na + influx (Fig. 5A). ROIs, encompassing dendritic sections of ß10 µm in length were set to study resulting [Na + ] i transients along dendrites (Fig. 5A). Pressure application for 50 ms (n = 14, N = 6) resulted in a [Na + ] i increase of 12.0 ± 4.6 mM (range: 5.0-18.16 mM, median: 12.6 mM) in ROIs in the immediate vicinity (ß10 µm) of the application pipette, while in sections further away, changes in [Na + ] i could hardly be resolved (Fig. 5A). Increasing the duration of the pressure application (n = 9-15, N = 5) resulted in increased peak amplitudes of [  (Fig. 5B). At the same time, their spatial restriction was lost and at 500 ms application, all dendritic sections in the field of view showed [Na + ] i transients (Fig. 5A). The same stimulation paradigm and protocol were then employed while imaging [ATP] with ATeam in dendrites (n = 31 cells, N = 3). To determine the detection threshold, all data points obtained from a given ROI during a 1-2 min period of baseline were averaged to obtain a mean value and their standard deviation was calculated. The latter was multiplied by two and subtracted from the mean, defining the minimum threshold for signal detection as shown in Fig. 5C ('2SD'). With a typical detection threshold of 0.02-0.03 mM in this set of experiments, pressure application of glutamate to dendrites for 50 ms did not result in a detectable change in [ATP] in the vast majority (29/31) of dendritic segments recorded from ( Fig. 5C and D). Upon application for 100 ms, a decrease in [ATP] by 0.03-0.06 mM could be resolved in 7 out of 31 dendritic sections (overall average peak decrease: 0.01 ± 0.019 mM; range: 0-0.057 mM; median: 0 mM); 19 out of 31 dendrites responded to application for 200 ms (overall mean: 0.030 ± 0.026 mM, range: 0-0.074 mM, median: 0.034 mM). All dendritic sections (31/31) showed a decrease in [ATP] to pressure application of glutamate for 500 ms, averaging 0.071 ± 0.022 mM (range: 0.040-0.120 mM, median: 0.068 mM).
In summary, these results show that large, global [Na + ] i transients, evoked by bath application of glutamate in both dendrites and somata, result in a delayed decline in [ATP] in both compartments. In contrast to this, glutamate-induced local [Na + ] i transients that are restricted to small sections of dendrites are not accompanied by detectable changes in dendritic [ATP].

Discussion
In the present study, we analysed the relation between activity-induced Na + transients and intracellular ATP levels in CA1 pyramidal neurons of the mouse hippocampus using the Na + -sensitive indicator dye SBFI and the nanosensor ATeam. Calibration of ATeam in neurons revealed an apparent K D of the sensor of 2.6 mM. Disinhibition of slices, burst stimulation of glutamatergic afferent fibres, or bath application of glutamate induced global [Na + ] i signals that comprised dendrites as well as somata of CA1 neurons. Global [Na + ] i transients were accompanied by transient decreases in [ATP] in both compartments. The relative peak amplitude and kinetics of these [ATP] i changes largely correlated with the properties of the changes in [Na + ] i in the given compartment. In contrast to global [Na + ] i transients, induction of a locally restricted Na + influx into dendrites by focal application of glutamate did not result in a detectable change in dendritic [ATP].

Properties of activity-related neuronal Na + elevations
Influx of Na + is a major mechanism for the generation of plasma membrane depolarization. Depending on the pattern of activity, this can be accompanied by transient changes in the extracellular [Na + ]. Earlier work performed in the cat cortex in vivo as well as in rat hippocampal tissue slices has shown that electrical stimulation at 15-20 Hz for 10 s results in a drop in the extracellular [Na + ] by 5-10 mM (Dietzel et al. 1982;Zanotto & Heinemann, 1983). In the present study, burst-like neuronal activity induced by 0Mg 2+ /BIC saline was accompanied by transient decreases in extracellular [Na + ] by 1-4 mM. These bursts only lasted ß5 s, and the magnitude of [Na + ] decreases in relation to activity pattern is thus rather comparable between the different studies.
[Na + ] i transients are, for example, generated upon opening of voltage-gated Na + channels in axons (Fleidervish et al. 2010;Baranauskas et al. 2013;Ona-Jodar et al. 2017). Similarly, bursts of back-propagating action potentials cause transient [Na + ] elevations in dendrites and spines, albeit of much lower peak amplitude (Jaffe et al. 1992;Rose et al. 1999;Ona-Jodar et al. 2017). A main route for Na + influx is postsynaptic ionotropic glutamate receptors activated by synaptic release or exogenous application of glutamate Knopfel et al. 2000;Rose & Konnerth, 2001;Meier et al. 2006;Bennay et al. 2008;Miyazaki & Ross, 2017).   Ionotropic glutamate receptors also represent the most important influx pathways underlying synchronous oscillations of somatic [Na + ] i during recurrent neuronal network activity in hippocampal tissue slices as shown here and described in detail previously (Karus et al. 2015). In the present study, we moreover demonstrate that [Na + ] i oscillations also comprise the apical dendritic tree of CA1 neurons, indicating that they occur in the entire cell. In addition, global cellular [Na + ] i transients, ranging from 20-40 mM, were induced by burst stimulation of the Schaffer collateral pathway and by a 10 s bath application of glutamate.
As opposed to such global events, [Na + ] i transients can also be local and occur in parts of a dendrite and adjacent spines only, e.g. upon more locally restricted stimulation of excitatory fibres and/or postsynaptic sites Knopfel et al. 2000;Rose & Konnerth, 2001;Meier et al. 2006;Bennay et al. 2008;Mondragao et al. 2016;Miyazaki & Ross, 2017). In the present study, brief (50-100 ms) pressure application of glutamate close to a dendrite induced [Na + ] i increases that were largely confined to dendritic sections within a distance of 10-15 µm from the tip of the application pipette.
The spatial extent of activity-related [Na + ] i elevations as reported by chemical Na + indicators can thus vary strongly, depending on their underlying mechanism and on the pattern of activity. High-affinity Ca 2+ -sensitive dyes like Oregon Green or Fura-2 exert a considerable buffering of Ca 2+ , thereby causing a distortion of Ca 2+ signals influencing both their amplitude and their spatial occurrence (Maravall et al. 2000). This, however, is not the case for Na + transients determined using SBFI. The K D of SBFI is well above the baseline [Na + ] i (present study: 42 mM and Rose et al. 1999;Sheldon et al. 2004;Meier et al. 2006). Even for whole-cell patch-clamp measurements, where SBFI is routinely used at concentrations of around 1 mM, it does not visibly buffer Na + inside cells (Mondragao et al. 2016).
Upon its influx, Na + has to be extruded from neurons to maintain low baseline [Na + ] i and thereby maintain driving forces for Na + -based electrical signalling. Na + export is achieved by the NKA, which is activated by rises in [Na + ] i (Thompson & Prince, 1986;Sweadner, 1995;Therien & Blostein, 2000;Larsen et al. 2016). Recovery from global [Na + ] i signals is, therefore, strictly dependent on NKA and on an intact energy metabolism (Lees, 1991;Erecinska & Silver, 1994;Somjen, 2002;Mondragao et al. 2016). For recovery from local [Na + ] i transients, an additional mechanism was demonstrated to come into play, namely the intracellular diffusion of Na + to non-stimulated areas, a phenomenon described from axon initial segments (Fleidervish et al. 2010) as well as dendrites and dendritic spines (Rose & Konnerth, 2001;Mondragao et al. 2016;Miyazaki & Ross, 2017). In primary apical dendrites largely devoid of spines, it was estimated that Na + diffused with a diffusion coefficient of approximately 330 µm 2 s −1 (Mondragao et al. 2016). This is much faster as compared to (buffered) diffusion of calcium (ß13 µm² s −1 ) (Yuste et al. 2000), and enables a rapid lateral redistribution Na + from the site of influx solely driven by its concentration gradient (Mondragao et al. 2016).

Different activity patterns neuronal ATP levels
In the present study, we employed the FRET-sensor ATeam for detection of intracellular ATP. Since the description of ATeam and its variants in 2009 (Imamura et al. 2009), ATP sensors have been constantly adapted and/or newly designed (Nakano et al. 2011;Yaginuma et al. 2014;Arai et al. 2018;Mendelsohn et al. 2018;Lobas et al. 2019), also for detection of ATP in organelles (Imamura et al. 2009;Rueda et al. 2015;Suzuki et al. 2018). Moreover, a ratiometric sensor for imaging of the ATP-to-ADP ratio (PercevalHR) is available (Tantama et al. 2013).
ATeam exhibits a relatively low K D in in vitro calibrations (around 0.25 mM at room temperature and 1.2 mM at 37°C), suggesting that it might be saturated inside cells (Imamura et al. 2009). Our recent work performed in hippocampal tissue slices provided evidence that this is not the case by demonstrating that ATeam reports an increase in astrocytic ATP levels in response to an increase in the extracellular potassium concentration (Lerchundi et al. 2019b). The in situ calibration of ATeam presented here supports this notion, revealing an apparent K D for ATP of 2.6 mM under our experimental conditions. Such a change in properties does not seem to be exclusive to ATeam, since an apparent decrease in binding affinity inside cells was, for example, also reported from the ratiometric fluorescent ATP indicator Queen (Yaginuma et al. 2014). Neither the relative increase in ATP levels induced by high K + nor their decrease upon blocking cellular metabolism reported by ATeam was altered when elevating the temperature by 10°C (Lerchundi et al. 2019b). The latter observation indicates that its K D is not strongly temperature-dependent inside cells, which is also different from the results obtained in vitro (Imamura et al. 2009).
Together with the reported [ATP] of ß2 mM in neurons (Fukuda et al. 1983;Chinopoulos et al. 2000;Ainscow et al. 2002;Mollajew et al. 2013;Rangaraju et al. 2014;Pathak et al. 2015), these results strongly suggest that ATeam is well suited to monitor changes in [ATP] i . As a rule of thumb, the apparent K D of 2.6 mM indicates that ATeam will be able to detect changes in the ATP concentration at baseline ATP levels between 0.26 and a theoretical maximum of 26 mM (but see Fig. 1 it is practically close to saturation at ß10 mM). At the presumed baseline ATP concentration of 2 mM, the sensor works close to its K D and is thus in its optimal 'working range' , responding to changes in ATP with relatively large changes in fluorescence (see Fig. 1). There are also other ATeam variants available that enable detection of very small changes in ATP (e.g. ATeam3.10; Imamura et al. 2009). However, under physiological conditions, lower K D sensors are already completely saturated inside cells and thereby not applicable in this kind of experiment and preparation.
All stimulation paradigms employed in the present study, except brief local pressure application of glutamate onto dendrites, resulted in well-detectable decreases in somatic and dendritic [ATP]. With a detection threshold of 0.02-0.03 mM, changes in [ATP] i were roughly in the range of 0.1-0.6 mM. For a given stimulation there was a large variance in the peak amplitude of [ATP] i decreases between individual cells and also between dendritic regions of a given cell. This remarkable variability may reflect a different density and intracellular distribution of mitochondria (and capacity for rapid ATP production): cells or dendrites that experienced a relatively large decrease in [ATP] i in response to a given stimulation might have hosted fewer mitochondria than those showing a small decrease. Mitochondria in neurons and astrocytes generally exist in two pools, one of them stationary and the other one in motion (Misgeld & Schwarz, 2017). The latter pool apparently serves to enable a directed movement and arrest of mitochondria to sites of increased energy demand (Jackson & Robinson, 2018). Based on these findings, one could speculate that a repeated stimulation with glutamate will increase the number of mitochondria close to Na + (and Ca 2+ ) influx sites, thereby increasing the local capacity for rapid ATP production. Subsequent challenges with glutamate might then result in smaller decreases in [ATP] i .
In dendrites, ATP levels generally dropped to lower levels as compared to cell bodies. Moreover, dendritic [ATP] decreased significantly faster than somatic [ATP] in response to afferent stimulation and bath application of glutamate. This suggests that the ATP availability in dendrites is more restricted and/or that dendritic ATP consumption is more pronounced than that of somata. In both compartments, [ATP] i decreases were fully reversible upon fibre stimulation or local pressure application of glutamate for 200-500 ms. This was in contrast to recurrent network activity or bath application of glutamate after which ATP levels did not always fully return to initial baseline, but remained at a slightly decreased level (around −0.1 mM) for the duration of the recording. A similar phenomenon was reported from astrocytes in culture upon bath application of glutamate , suggesting that with strong stimulation (and Na + influx, respectively), cells establish a new steady-state.
Upon disinhibition of slices and neuronal network activity accompanied by [Na + ] i oscillations, we observed a slowly developing decline in neuronal [ATP] by 0.3-0.4 mM. This is somewhat lower as reported before based on estimates of [ATP] using the sensor ATeam1.03 (Toloe et al. 2014). A reduction in this range might alter the conductance of K ATP channels, resulting in a dampening of neuronal excitability, thereby executing an anti-convulsant effect (Proks & Ashcroft, 2009;Mollajew et al. 2013;Toloe et al. 2014;Martinez-Francois et al. 2018).
Bath application of glutamate for 10 s had the strongest impact on neuronal [ATP], resulting in a drop by 0.5-0.6 mM. At an assumed neuronal [ATP] of 2 mM, even such a relatively large decrease in ATP levels would probably not directly restrict the activity of the NKA. The NKA has both a high-affinity binding site (K D 0.1-0.2 µM) and a regulatory binding site for ATP (K D 0.2-0.4 mM) (Lees, 1991;Blanco & Mercer, 1998;Kaplan, 2002) and earlier studies have estimated that its activity declines only when ATP levels fall below 0.4 mM (Dagani & Erecinska, 1987). This is also evident from the efficient recovery from the large [Na + ] i increases that were seen under these conditions. Notably, and as mentioned above, the changes in [ATP] i as detected during the different stimulation paradigms might, however, exert an effect on neuronal membrane potentials by increasing the opening of K ATP channels.
Local pressure application of glutamate for 50 ms close to a dendrite, as opposed to all other stimulation paradigms, generally did not result in a detectable change in [ATP] i , while at the same time inducing a transient [Na + ] i increase by about 12 mM. Increasing the duration of the pressure application increased the ratio of responding dendrites, until at 500 ms, all dendritic sections tested showed a decrease in [ATP]. This suggests that localized activation of glutamate receptors changes dendritic [ATP] i by less than 0.02-0.03 mM, which is not expected to exert a significant effect on the activity of the NKA or of K ATP channels.

Relation between sodium and ATP
Previous work has demonstrated a link between Na + influx and ATP levels in different cell types including neurons (e.g. Chinopoulos et al. 2000;Mollajew et al. 2013). Here, such a correlation was found for patterns of activity which induced global [Na + ] i transients that comprised both neuronal somata and dendrites. For a given stimulation paradigm, the amplitude of [Na + ] i transients and the change in [ATP] i were positively correlated. Moreover, we found that the kinetics of changes in [ATP] i generally followed those in [Na + ] i . However, while [ATP] i essentially began to drop as soon as changes [Na + ] i began to rise (see Fig. 4E), the peak of [ATP] i changes was delayed as compared to the peak of [Na + ] i transients and recovery to baseline [ATP] took significantly longer. These results indicate that [Na + ] i loads and activation of the NKA are indeed major J Physiol 597.23 determinants of neuronal ATP consumption. Activation of glutamatergic signalling as performed in the present study, however, not only causes a rise in [Na + ] i , but will, for example, also evoke an elevation of intracellular calcium. This will most likely result in the activation of other ATP-consuming processes, including the plasma membrane Ca 2+ -ATPase and/or the SERCA (Clapham, 2007), contributing to the observed drop in [ATP] i . At presynaptic terminals, ATP will additionally be required by the vacuolar-type H + -ATPase to regenerate the H + gradient required for neurotransmitter accumulation in presynaptic vesicles (Cotter et al. 2015;Sobieski et al. 2017).
The time course and temporal relation of changes in [Na + ] i and [ATP] i are likely to be influenced by the temperature. While experiments presented here were performed at room temperature (20-22°C), our earlier work has shown that recovery of CA1 pyramidal neurons from global [Na + ] i transients is accelerated by a factor of ß2 when increasing the temperature by 10°C (Rose et al. 1999;Mondragao et al. 2016), indicating that the NKA indeed plays a dominating role in this process. At the same time, mitochondrial generation of ATP will be accelerated at higher temperature, suggesting that both, global [Na + ] i transients and the accompanying changes in [ATP] i will be speeded up at physiological temperature (37°C). Recovery from local [Na + ] i signals, in contrast, is largely driven by diffusion, and consequently, not altered by changing the temperature (Mondragao et al. 2016).
Our results strongly suggest that activation of NKA to recover from global [Na + ] i increases requires, at least temporarily, more ATP than is available and/or produced. As mentioned above, [ATP] will most likely not be rate-limiting for the NKA under the conditions tested here. However, for periods of energy restriction, such as during ischaemia or hypoxia, insufficient ATP availability has been shown to slow down NKA, resulting in a failure of Na + extrusion and a rise in [Na + ] i that aggravates damage by, for example, secondary loading of neurons with protons or Ca 2+ (Dagani & Erecinska, 1987;Somjen, 2002;Gerkau et al. 2017).
In contrast to activity and stimulation patterns that involved global [Na + ] i elevations, spatially restricted local [Na + ] i influx as observed upon brief pressure application of glutamate onto dendrites did not evoke a detectable decrease in [ATP] i . A recent study that failed to observe changes in somatic ATP levels in response to electrical activity and Na + influx concluded that ATP consumption by the NKA is matched by a direct increase in mitochondrial ATP production (Baeza-Lehnert et al. 2019). Earlier work by us and other laboratories, together with the results presented here, provides an alternative and/or additional explanation for the lack of changes in neuronal [ATP] levels despite relevant increases in [Na + ] i and the efficient recovery thereof.
Peak amplitudes of local dendritic [Na + ] i increases determined in the present study were around 12 mM, which represents a transient doubling of [Na + ] i (Karus et al. 2015;Mondragao et al. 2016). In CA1 neurons in acute tissue slices, recovery from a global transient increase in [Na + ] i in the same range (10-15 mM) was dependent on Na + extrusion by the NKA (Mondragao et al. 2016). Notably, the former study also demonstrated that recovery from locally restricted dendritic [Na + ] i transients of similar amplitude was not primarily dependent on NKA activity, as it was essentially unaltered during application of ouabain or during energy restriction. Based on these results, it was concluded that the NKA is not involved in the recovery from locally restricted Na + influx. Instead the study demonstrated that rapid lateral diffusion of Na + from regions with elevated [Na + ] i to unstimulated parts of the cell is the main mechanism for the recovery of [Na + ] to baseline under these conditions (Mondragao et al. 2016). This has also been described for axon initial segments and dendritic spines, where diffusion-driven, intracellular redistribution of Na + is the main mechanism for recovery from local Na + influx (Fleidervish et al. 2010;Miyazaki & Ross, 2017).
The results presented here are in agreement with this latter conclusion. Lateral diffusion of Na + does not per se involve a consumption of ATP, and we did not detect a change in local [ATP] with local Na + transients. In addition to diffusion of Na + , intracellular diffusion of ATP and/or phosphocreatine, estimated to occur at 500 and 640 µm 2 s −1 , respectively (de Graaf et al. 2000), might counteract a decline in ATP resulting from local activation of the NKA. Combined with a direct stimulation of ATP production (Baeza-Lehnert et al. 2019) and with diffusion of ATP itself, diffusion-driven recovery from activity-related [Na + ] i transients, which does not require direct local ATP consumption, will spread the metabolic burden from activated neuronal microdomains to neighbouring, unstimulated regions.