Extrusion versus diffusion: mechanisms for recovery from sodium loads in mouse CA1 pyramidal neurons

Abstract

Key points

• Neuronal activity causes local or global sodium signalling in neurons, depending on the pattern of synaptic activity.

• Recovery from global sodium loads critically relies on Na⁺/K⁺‐ATPase and an intact energy metabolism in both somata and dendrites.

• For recovery from local sodium loads in dendrites, Na⁺/K⁺‐ATPase activity is not required per se. Instead, recovery is predominately mediated by lateral diffusion, exhibiting rates that are 10‐fold higher than for global sodium signals.

• Recovery from local dendritic sodium increases is still efficient during short periods of energy deprivation, indicating that fast diffusion of sodium to non‐stimulated regions strongly reduces local energy requirements.

Abstract

Excitatory activity is accompanied by sodium influx into neurones as a result of the opening of voltage‐ and ligand‐activated channels. Recovery from resulting sodium transients has mainly been attributed to Na⁺/K⁺‐ATPase (NKA). Because sodium ions are highly mobile, diffusion could provide an additional pathway. We tested this in hippocampal neurones using whole‐cell patch‐clamp recordings and sodium imaging. Somatic sodium transients induced by local glutamate application recovered at a maximum rate of 8 mm min⁻¹ (∼0.03 mm min⁻¹ μm⁻²). Somatic sodium extrusion was accelerated at higher temperature and blocked by ouabain, emphasizing its dependence on NKA. Moreover, it was slowed down during inhibition of glycolysis by sodium fluoride (NaF). Local glutamate application to dendrites revealed a 10‐fold higher apparent dendritic sodium extrusion rate compared to somata. Recovery was almost unaltered by increased temperature, ouabain or NaF. We found that sodium diffused along primary dendrites with a diffusion coefficient of ∼330 μm²/s. During global glutamate application, impeding substantial net diffusion, apparent dendritic extrusion rates were reduced to somatic rates and also affected by NaF. Numerical simulations confirmed the essential role of NKA for the recovery of somatic, but not dendritic sodium loads. Our data show that sodium export upon global sodium increases is largely mediated by NKA and depends on an intact energy metabolism. For recovery from local dendritic sodium increases, diffusion dominates over extrusion, operating efficiently even during short periods of energy deprivation. Although sodium will eventually be extruded by the NKA, its diffusion‐based fast dissemination to non‐stimulated regions might reduce local energy requirements.

Key points

• Neuronal activity causes local or global sodium signalling in neurons, depending on the pattern of synaptic activity.

• Recovery from global sodium loads critically relies on Na⁺/K⁺‐ATPase and an intact energy metabolism in both somata and dendrites.

• For recovery from local sodium loads in dendrites, Na⁺/K⁺‐ATPase activity is not required per se. Instead, recovery is predominately mediated by lateral diffusion, exhibiting rates that are 10‐fold higher than for global sodium signals.

• Recovery from local dendritic sodium increases is still efficient during short periods of energy deprivation, indicating that fast diffusion of sodium to non‐stimulated regions strongly reduces local energy requirements.

Abbreviations

K_M

Michaelis–Menten constant

LTP

long term potentiation

MgG‐AM

magnesium‐green‐acetoxymethyl ester

mACSF

modified artificial cerebrospinal fluid

NaF

sodium fluoride

NKA

sodium‐potassium ATPase

ROI

region of interest

SBFI

sodium‐binding benzofuran isophthalate

SBFI‐AM

sodium‐binding benzofuran isophthalate‐acetoxymethyl ester

τ

time constant

V_max

maximum rate for recovery from sodium loads

Introduction

The balance of extra‐ and intracellular ion concentrations is of upmost functional importance for the brain and requires constant cellular transport activity and energy supply (Erecinska & Silver, 1994; Somjen, 2002). Most energy is used for the export of sodium, a task which is mediated by the plasma membrane Na⁺/K⁺‐ATPase (NKA) (Sweadner, 1989; Kaplan, 2002). Estimates indicate that this pump alone requires ∼50% of all ATP used by the brain (Whittam, 1962; Astrup et al. 1981; Ames, 2000). Mutations in NKA are known to be the origin of many neurological problems such as cognitive deficits (Moseley et al. 2007), parkinsonism (De Carvalho Aguiar et al. 2004) or hemiplegic migraine (Capendeguy & Horisberger, 2004).

Inhibition of NKA results in an immediate increase in intracellular sodium of central neurons even at rest, emphasizing the need for constant pump activity as a result of constitutive sodium influx (e.g. through secondary active transporters such as sodium‐calcium exchange or sodium‐proton exchange) (Rose & Ransom, 1997; Kelly & Rose, 2010 a; Kelly & Rose, 2010 b). Additional sodium loads are imposed onto active neurons. Here, an influx of sodium through voltage‐ and ligand‐gated channels represents a major metabolic challenge. In fact, sodium influx following action potentials requires 22–39% of cellular ATP generation, whereas 34–52% is allocated for postsynaptic ion fluxes following synaptic transmission (Lennie, 2003; Hallermann et al. 2012; Harris et al. 2012).

Former studies have firmly established that action potentials and the opening of voltage‐gated sodium channels cause an influx of sodium into axons (Kole et al. 2008; Fleidervish et al. 2010) and, with backpropagating action potentials present, also into dendrites and spines of central neurons (Jaffe et al. 1992; Rose et al. 1999), resulting in significant sodium transients in these cellular compartments. Especially prominent activity‐related sodium transients occur in postsynaptic dendrites in response to excitatory synaptic activity and the influx of sodium through glutamate‐gated ionotropic receptors (Lasser‐Ross & Ross, 1992; Rose & Konnerth, 2001; Bennay et al. 2008). In dendrites of hippocampal CA1 pyramidal neurons, short‐burst synaptic stimulation of glutamatergic fibres (Schaffer collaterals) induced local sodium transients that amounted to 10 mm, whereas, with a typical LTP induction protocol, dendritic sodium rose by even 45 mm (Rose & Konnerth, 2001). In addition to these local sodium signals, arising close to activated synapses, highly synchronized global sodium signalling was reported from the somata of the CA1 pyramidal cell population under epileptiform conditions (Karus et al. 2015 a). Because NKA is the only relevant transporter for the export of sodium under physiological conditions, it was hypothesized that recovery from both local and global activity‐related sodium transients was mainly mediated by activity of the pump (Rose & Konnerth, 2001; Karus et al. 2015 a).

Sodium, however, is an essentially non‐buffered, highly mobile ion for which intracellular diffusion coefficients of ∼600 μm²/s have been reported (Kushmerick & Podolsky, 1969), which considerably exceed those of calcium (Allbritton et al. 1992). Thus, another mechanism that could contribute to efficient recovery from local dendritic sodium increases in addition to NKA is the fast diffusion of sodium ions into non‐activated regions. Indeed, recent work on axon initial segments has provided evidence that fast sodium transients induced by the opening of voltage‐gated sodium channels during action potential generation are not altered upon inhibition of the NKA by ouabain, indicating that diffusion was responsible for their fast decay (Fleidervish et al. 2010).

In the present study, we have addressed this question by studying the properties of recovery from glutamate‐induced sodium transients in cell bodies and dendrites of CA1 pyramidal neurons, employing quantitative, ratiometric imaging with the sodium‐sensitive fluorescent dye sodium‐binding benzofuran isophthalate (SBFI). Our results show that NKA governs recovery from global sodium increases in both cell bodies and dendrites. Recovery from local dendritic sodium loads, however, is predominately mediated by diffusion, exhibiting recovery rates that are more than one order of magnitude higher.

Methods

Ethical approval

The present study was carried out in strict accordance with The Journal of Physiology’s policy (Grundy, 2015), as well as with the institutional guidelines of the Heinrich Heine University Düsseldorf and the European Community Council Directive (86/609/EEC). All experiments 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 number: O52/05) in accordance with the recommendations of the European Commission (Close, 1997).

Tissue dissection and saline composition

Experiments were performed on acute tissue slices (250 μm) of mouse hippocampus (Mus musculus, Balb/C; postnatal days 14–20, both sexes). Animals were anaesthetized with CO₂ and immediately decapitated. The brain was then removed and parasagittal hippocampal slices prepared using standard techniques (Meier et al., 2006) in cold (2–4ºC) modified artificial cerebrospinal fluid (mACSF). mACSF contained (in mm): 125 NaCl, 2.5 KCl, 0.5 CaCl₂, 6 MgCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃ and 20 glucose and was bubbled with 95% O₂ and 5% CO₂, pH 7.4; osmolarity was 310 mosmol l^–1. Slices were cut using a vibratome (Microm HM 650 V; Thermo Scientific, Waltham, MA, USA) and subsequently incubated at 34ºC for 30 min.

After the incubation period, slices were kept at room temperature (20–22ºC) in standard artificial cerebrospinal fluid (ACSF) until they were used for experiments, which were also performed at room temperature. The composition of the ACSF was (in mm): 125 NaCl, 2.5 KCl, 2 CaCl₂, 1 MgCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃ and 20 glucose, bubbled with 95% O₂ and 5% CO₂, pH 7.4. During the experiments, slices were constantly perfused with ACSF to which TTX (0.5 μm), a blocker of voltage‐gated ion channels, was routinely added to prevent action potential generation.

All substances, except glutamate, were applied via bath perfusion. Chemicals were purchased from Sigma‐Aldrich (Munich, Germany), except for tetrodotoxin (Abcam, Cambridge, UK) and ouabain (Calbiochem, Darmstadt, Germany). Glutamate (0.5 mm) was dissolved in Hepes‐buffered saline composed of (in mm): 125 NaCl, 25 Hepes, 3 KCl, 2 MgSO₄, 2 CaCl₂ and 1.25 NaH₂PO₄. Glutamate was applied for 100 ms by a pressure application device (PDES‐02D; NPI Electronic GmbH, Tamm, Germany) coupled to borosilicate glass micropipettes (Hilgenberg, Waldkappel, Germany) placed either 5–10 μm from the cell body or the primary apical dendrite of a selected cell (see Fig. 3 A and 3 C).

Figure 3. Recovery from sodium loads in cell bodies and dendrites .

A, left: scheme of the experimental design: a neuron was held in the whole‐cell patch‐clamp mode, glutamate was focally applied to the soma (as indicated by the green area) and sodium signals were measured from the soma (as indicated by the red area). Centre: somatic sodium signals evoked by repetitive glutamate application to the soma as indicated by the arrowheads. Right: relationship between sodium concentration and decay velocity (V) as obtained during the first few seconds after the sodium peak. The data were fitted with a Hill equation, the results for V _max and V _20–25 are indicated. B, as described in (A) but with AM‐ester loaded neurons (no patch‐pipette attached). V _max and V _20–25 were strongly reduced under this condition. C, left: scheme of the experimental design: a neuron was held in the whole‐cell patch‐clamp mode, glutamate was focally applied to the dendrite (as indicated by the green area) and sodium signals were measured from a section of the dendrite (as indicated by the red area). Centre: dendritic sodium signals evoked by repetitive glutamate application as indicated by the arrowheads. Right: relationship between sodium concentration and decay velocity (V) as obtained during the first few seconds after the sodium peak. The data were fitted with a Hill equation, the results for V _max and V _20–25 are indicated. D, as described in (C) but, in this case, the patch pipette was carefully retracted after loading with SBFI and the cell was allowed to reseal. Values obtained for V _max and V _20–25 were similar to those acquired under the same conditions but with the pipette attached. [Colour figure can be viewed at wileyonlinelibrary.com]

Whole‐cell patch‐clamp

Somatic whole‐cell patch‐clamp recordings were performed from CA1 hippocampal neurons using an upright microscope (BX51WI; Olympus Europe, Hamburg, Germany), equipped with an Achroplan 40× objective (water immersion, NA 0.8; Zeiss, Göttingen, Germany), using an EPC10 amplifier and PatchMaster software (HEKA Elektronik, Lambrecht, Germany). Patch pipettes were pulled from borosilicate glass capillaries (Hilgenberg) using a vertical puller (PP‐830; Narishige, Tokyo, Japan). Pipettes were filled with intracellular saline containing (in mm): 120 K‐MeSO₃, 10 Hepes (N‐(2‐hydroxymethyl) piperazine‐N′‐2‐ethanesulphonic acid) 32 KCl, 4 NaCl, 1 EGTA (ethylene glycol tetracetic acid), 4 Mg‐ATP and 0.4 Na₃‐GTP, as well as the K⁺ salt of SBFI (0.1–2 mm; TEFLabs Inc., Austin, TX, USA), titrated with KOH to pH 7.3. Pipette resistance was 2.3–3.0 MΩ when filled with this saline. Liquid junction potential was not corrected. Cells were routinely held in the voltage clamp mode at a holding potential of –60 mV. Data were sampled at 1 kHz and processed and analysed employing IGOR Pro (WaveMetrics, Lake Oswego, OR, USA).

Wide‐field imaging

For sodium imaging, individual neurons were loaded with the sodium‐selective fluorescent dye SBFI (0.1–2 mm) via a patch pipette. Alternatively, cells were loaded by bolus injection of SBFI‐AM (sodium‐binding benzofuran isophthalate‐acetoxymethyl ester; Invitrogen, Karlsruhe, Germany) into the CA1 region as described previously (Meier et al. 2006; Langer & Rose, 2009). De‐esterification was allowed for 1 h before commencing the imaging experiments.

Wide‐field fluorescence imaging was performed using a variable scan digital imaging system (TILL Photonics, Martinsried, Germany) attached to a microscope (see above) and a CCD camera (Imago VGA; TILL Photonics). SBFI was alternately excited at 340 nm (sodium‐insensitive wavelength) and at 380 nm (sodium‐sensitive wavelength) (cf. Fig. 2 A and B). Images were captured at a frequency of 1 or 4 Hz. SBFI emission was collected above 440 nm from defined regions of interest (ROIs), representing either the soma or the primary apical dendrite. After standard dynamic background correction (Langer & Rose, 2009), the fluorescence ratio (R _340/380) was calculated for each ROI and analysed off‐line using OriginPro, version 8.5G (Origin Lab Corporation, Northampton, MA, USA). Fluorescence ratios were converted into absolute changes in sodium concentration based on in situ calibrations as described in the results (Fig. 1).

Figure 2. Influence of dye concentration on SBFI signals .

A, change in the SBFI fluorescence at 340 nm (F ₃₄₀, top trace) and at 380 nm (F ₃₈₀, centre trace) excitation, as well as the resulting ratio, converted into sodium changes (Δ[Na]_i, lower trace) during dye‐loading after obtaining the whole‐cell mode with a pipette containing 0.1 mm SBFI. During this phase, glutamate (0.5 mm, 100 ms) was applied repetitively as indicated by the arrowheads. Note that glutamate application does not evoke changes in F ₃₄₀, whereas signals at F₃₈₀ increase in amplitude and the ratio signals remain almost unaltered with increasing dye concentrations. B, as described in (A) with a pipette containing 1 mm SBFI. Again, glutamate‐induced changes in the SBFI ratio are virtually independent from the SBFI concentration. C, integrals of the changes in sodium induced by glutamate application as determined at different concentrations of SBFI. Data were pooled for concentrations between 1–200, 201–400, 401–600, 601–800 and 801–1000 μm SBFI as indicated. There is no significant difference between means. D, changes in peak sodium concentration (Δ[Na⁺]_i) induced by glutamate application, normalized to the charge of the respective inward currents, at different SBFI concentrations as indicated. There is no significant difference between means. E, decay time constants (τ) of the recovery phase of glutamate‐induced sodium transients, normalized to the charge of the corresponding currents. There is no significant difference between the means. C–E, boxes represent quartiles; whiskers correspond to 95% confidence interval. Middle lines show the median; squares indicate the mean value.

Figure 1. In situ calibration of SBFI fluorescence and glutamate‐induced sodium signals .

A, normalized changes in the fluorescence ratio of SBFI (ΔR/R _0Na) in response to changes in the sodium concentration in SBFI‐AM loaded cell bodies of CA1 pyramidal neurons (ΔR/R _0Na: ratio R at a given [Na⁺] minus ratio R at [Na⁺]_i = 0 mm (R _0Na), divided by R _0Na and presented as the percentage change relative to R _0Na). Cells were perfused with calibration solutions containing different concentrations of sodium as indicated in addition to gramicidin (3 μm), monensin (10 μm) and ouabain (100 μm). B, relationship between normalized changes in SBFI fluorescence ratio (ΔR/R _0Na) and [Na⁺]_i. Data are shown as the mean ± SD. A fit of the data between [Na⁺]_i = 10 mm and [Na⁺]_i = 40 mm (red line) reveals an almost linear relationship within this range, with a 10% change in ratio corresponding to a change in sodium concentration of ∼8.5 mm. C, change in the SBFI ratio of a bolus‐loaded neuron after obtaining the whole‐cell mode with a patch‐pipette containing 4 mm Na⁺ (upper trace) or 10 mm Na⁺ (lower traces). D, somatic inward current and sodium transient in a CA1 pyramidal neuron upon pressure application of glutamate (0.5 mm, 100 ms). E, relationship between glutamate‐induced peak changes in sodium concentration and the charge of the accompanying inward currents as well as a linear fit of the data (red line). [Colour figure can be viewed at wileyonlinelibrary.com]

To study the effects of metabolic inhibition, slices were bolus‐loaded with the magnesium‐sensitive fluorescent dye magnesium‐green‐acetoxymethyl ester (MgG‐AM; Life Technologies, Eugene, OR, USA). Changes in the emission of this dye, indicating changes in the intracellular free Mg²⁺ concentration, were used as indication for changes in the intracellular ATP concentration as reported previously (Magistretti & Chatton, 2005).

Data analysis and statistical analysis

To obtain the rate of recovery from intracellular sodium loads, the rate of decline in [Na⁺]_i (−d[Na⁺]_i/dt) from a glutamate‐induced sodium load was determined from a linear fit of the data obtained during 1–2 s after the peak sodium concentration had been reached (d[Na⁺]_i/dt = 0). This rate of decline (−d[Na⁺]_i/dt), from now on termed recovery rate V, was plotted vs. [Na⁺]_i and fitted with a Hill equation V = V _max [Na⁺]_i ^nHill/(K _M ^nHill + [Na⁺]_i ^nHill) using OriginPro software to derive the maximum sodium recovery rate (V _max).

Data are presented as the mean ± SD and the results for the parameters obtained from fitting functions are shown as mean [l.l., u.l.], where l.l. and u.l. represent the lower limit and the upper limit of the 95% confidence intervals, respectively. All hypotheses were tested statistically with 95% confidence by one‐way ANOVA followed by Tukey's test for multiple comparisons (P represents error probability, n.s., not significant, ^*0.01 ≤ P < 0.05, ^**0.001 ≤ P < 0.01, ^*** P < 0.001). The statistical analysis was performed using Prism, version 6 (GraphPad Software Inc., San Diego, CA, USA).

Experiments were repeated on at least three different slice preparations obtained from at least three different mice; n represents the number of individual sodium signals evaluated; N represents the number of different cells analysed.

Modelling

Simulations were performed with Mathematica, version 10.3 (Wolfram Research; http://www.wolfram.com) by numerically solving a set of coupled ordinary differential reaction diffusion equations (Helmchen & Tank, 2000; Schmidt & Eilers, 2009). The cell soma was represented by a sphere (r _s = 5 μm), as calculated for a somatic surface of 300 μm², taken from http://neuromorpho.org (Ascoli et al. 2007; Beguin et al. 2013), which was coupled via diffusion to a cylindrical dendrite (r _de = 0.4 μm) (Araya et al. 2006), divided into three diffusionally coupled compartments with length (l_de) of 50 (de1), 20 (de2) and 500 (de3) μm (cf. Fig. 7). All compartments were modelled as well mixed. A Gaussian shaped sodium influx (width 0.5 s) was placed either at the soma or at dendritic compartment de2. The resting concentration of sodium ([Na]_rest) was set to 13 mm. The diffusional flux (J) between compartments i and j was simulated as:

$$J{}_{\mathrm{i},\mathrm{j}}=({\mathrm{D}}_{\mathrm{Na}}\pi {r_{\mathrm{de}}}^2/1_{\mathrm{de}})\left({\left[\mathrm{Na}\right]}_i\left(t\right)-{\left[\mathrm{Na}\right]}_{\mathrm{j}}\left(t\right)\right)$$

where D _Na is the diffusion coefficient of Na⁺ (320 μm² s^–1).

Figure 7. Numerical simulation of sodium recovery .

A, left: scheme of the simulated experiment: sodium influx was placed on the soma (indicated by red area) which was coupled to a dendrite that was divided into three compartments of 50, 20 and 500 μm in length as indicated. Right: changes in sodium concentrations in the different compartments with NKA on (left) or turned off (right). B, left: scheme of the simulated experiment: sodium influx was placed on the second dendritic segment (indicated by red area). Right: changes in sodium concentrations in the different compartments with NKA on (left) or turned off (right). [Colour figure can be viewed at wileyonlinelibrary.com]

Sodium dynamics for somatic influx were simulated as:

$$\begin{array}{ccc}\hfill {(\mathrm{d}\left[\mathrm{N}a\right]/\mathrm{d}t)}_s& =& \left(\mathrm{Influx}-\mathrm{Pump}-J{}_{s,\mathrm{de}1}/V_{\mathrm{s}}\right)/(1+{\mathrm{k}}_{\mathrm{B}})\hfill \\ \hfill {(\mathrm{d}\left[\mathrm{N}a\right]/\mathrm{d}t)}_{\mathrm{de}1}& =& \left(-\mathrm{Pump}+J{}_{s,\mathrm{de}1}/V_{\mathrm{de}1}-J{}_{\mathrm{de}1,\mathrm{de}2}/V_{\mathrm{de}1}\right)/\hfill \\ & & (1+{\mathrm{k}}_{\mathrm{B}})\hfill \\ \hfill {(\mathrm{d}\left[\mathrm{Na}\right]/\mathrm{d}t)}_{\mathrm{de}2}& =& (-\mathrm{Pump}+J{}_{\mathrm{de}1,\mathrm{de}2}/V_{\mathrm{de}2}-J{}_{\mathrm{de}2,\mathrm{de}3}/\hfill \\ & & V_{\mathrm{de}2})/(1+{\mathrm{k}}_{\mathrm{B}})\hfill \\ \hfill {(\mathrm{d}\left[\mathrm{Na}\right]/\mathrm{d}t)}_{\mathrm{de}3}& =& \left(-\mathrm{Pump}+J{}_{\mathrm{de}2,\mathrm{de}3}/V_{\mathrm{de}3}/V_{\mathrm{de}3}\right)/(1+{\mathrm{k}}_{\mathrm{B}})\hfill \end{array}$$

where V denotes the volume of the respective compartment and κ_B the buffer capacitance added by the indicator dye (calculated from a K _D of 24 mm and the dye concentration of 1 mm). Sodium dynamics with dendritic influx was simulated correspondingly but with the influx being moved from the soma to de2.

The pump is a sodium clearance mechanism obeying Michaelis–Menten kinetics and is balanced by a ‘leak’ that keeps the sodium concentration at 13 mm under resting conditions:

$$\begin{array}{ccc}\hfill \mathrm{Pump}& =& {V_{\max }}^{*}{A}_{\mathrm{i}}/{V_{\mathrm{i}}}^{*}\left(\right({\left[\mathrm{Na}\right]}_{\mathrm{rest}}/({\left[\mathrm{Na}\right]}_{\mathrm{rest}}+K_{\mathrm{M}})\hfill \\ & & -({\left[\mathrm{Na}\right]}_{\mathrm{i}}\left(\mathrm{t}\right)/{\left[\mathrm{Na}\right]}_{\mathrm{i}}\left(\mathrm{t}\right)+K_{\mathrm{M}})\left)\right)\hfill \end{array}$$

where K _M is the Michaelis–Menten constant of 19 mm, A/V is the surface to volume ratio of the ith compartment and V _max is the maximum pump velocity of 0.03 mm μm⁻² s⁻¹. The parameters of the pump were set according to the experimentally obtained values (see Results and Fig. 3), leaving the amplitude of the sodium influx as the only free parameter of the simulation. It was adjusted to match the experimental increases in free sodium concentration of 18– 20 mm.

Results

Quantitative sodium imaging and glutamate‐induced sodium signals in CA1 neurons

The goal of the present study was to analyse the mechanisms for recovery from intracellular sodium loads in hippocampal principal neurons in situ. To this end, quantitative ratiometric wide‐field imaging with the sodium‐sensitive fluorescent dye SBFI was performed in somata and apical dendrites of CA1 pyramidal neurons of acute mouse hippocampal tissue slices. SBFI changes its fluorescence properties between a cell‐free aqueous solution and intracellular environments (Harootunian et al. 1989; Minta & Tsien, 1989), an effect mainly caused by differences in viscosity of the medium (Harootunian et al. 1989). For reliable determination of intracellular sodium concentrations ([Na⁺]_i), we therefore performed in situ calibrations of SBFI fluorescence. These essentially consist of recording the changes in the SBFI fluorescence ratio in response to known changes in [Na⁺]_i.

For calibration, CA1 pyramidal neurons were bolus loaded with SBFI‐AM and then subjected to calibration solutions containing different concentrations of sodium, as well as gramicidin (3 μm), monensin (10 μm) and ouabain (100 μm), to enable equilibration of external and internal [Na⁺] (Rose & Ransom, 1997; Meier et al. 2008). When perfused with these calibration solutions, SBFI fluorescence emission changed in response to changes in sodium concentration when excited at 380 nm (F ₃₈₀; sodium‐sensitive wavelength) but remained largely unaltered when excited at 340 nm (F ₃₄₀; sodium‐insensitive wavelength; not shown) (see also Fig. 2 A and B) (Langer & Rose, 2009). Changes in the fluorescence ratio of SBFI were normalized to the ratio at [Na⁺]_i = 0 mm and plotted against the respective sodium concentration (ΔR/R _0Na: ratio R at a given [Na⁺] minus ratio R at [Na⁺]_i = 0 mm (R _0Na), divided by R _0Na and presented as the percentage change relative to R _0Na; eight experiments, N = 54) (Fig. 1 A). A linear fit of the data for Na⁺ concentrations between 10 and 40 mm (R ² = 0.99) revealed that a change in the ratio by 10% corresponded to change in [Na⁺]_i by 8.5 mm (Fig. 1 B).

During the first 10–15 min after starting perfusion with calibration saline, the tissue experienced swelling. The resulting changes in focus precluded a direct comparison of fluorescence ratio values between physiological and calibration solutions and therefore a proper determination of physiological baseline [Na⁺]_i of neurons before switching to calibration salines from these experiments. To determine baseline [Na⁺]_i, we thus followed an alternative strategy. This approach based on (1) the known linear relationship between changes in SBFI fluorescence ratio and sodium concentration determined by the full calibration in bolus‐loaded slices (Figs 1 A and B) and (2) on the assumption that whole‐cell patch‐clamping a neuron causes equilibration of somatic sodium with the sodium concentration of the patch‐pipette.

To this end, neurons were first bolus‐loaded with SBFI‐AM and the fluorescence ratio was determined in physiological saline. Subsequently, an individual neuron was subjected to whole‐cell patch‐clamp with an intracellular saline containing 0.1 mm SBFI and either 4 or 10 mm Na⁺. This resulted in a change in the SBFI ratio because of equalization of the somatic sodium with that of the pipette saline. Based on the performed in situ calibrations (see above), the change in fluorescence ratio after obtaining the whole‐cell mode and dilution of the cell with the known pipette [Na⁺] was then used to deduce baseline [Na⁺]_i in this cell before breakthrough (Fig. 1 C). These measurements revealed a mean value of 13.4 ± 4.7 mm [Na⁺]_i in hippocampal CA1 pyramidal neurons (n = 6).

To study intracellular sodium signals and their recovery, CA1 neurons were subjected to whole‐cell patch‐clamp and loaded with SBFI (0.1–1 mm). Focal pressure application of glutamate (0.5 mm, 100 ms) to the soma resulted in an inward current accompanied by a transient increase in the sodium concentration. The amplitude of sodium transients was linearly correlated (R ² = 0.94) with the total charge of the evoked somatic current (n = 136, N = 33) (Fig. 1 D and E). Taken together, these results demonstrate that glutamate application reliably evokes sodium transients in somata of CA1 pyramidal neurons, the amplitude of which depends on the induced inward current.

Influence of dye concentration

Compared to other ion‐selective fluorescent dyes (e.g. Fura‐2), SBFI exhibits a minor quantum efficiency and a low signal‐to‐noise ratio, necessitating the use of rather high dye concentrations (between 0.5 and 1 mm) when loaded through a patch‐pipette (Schreiner & Rose, 2012). To determine whether the addition of such SBFI concentrations induces any detectable buffering and distortion of intracellular sodium signals, we employed the ‘added buffer approach’ (Neher, 1995; Helmchen et al. 1996). To this end, neurons were subjected to whole‐cell patch‐clamp with intracellular salines containing SBFI concentrations ranging from 0.1 (Fig. 2 A) to 1 mm (Fig. 2 B).

After rupturing the patch, SBFI diffused from the pipette into the soma and the resulting increase in fluorescence emission from the soma was monitored at both excitation wavelengths (F ₃₄₀, F ₃₈₀) until a stable maximum was reached, indicating full equilibration with the pipette saline (Fig. 2 A and B). During this loading phase, which took ∼10 min, glutamate (0.5 mm, 100 ms) was repetitively applied to the soma, resulting in transient changes in the F ₃₈₀, the amplitude of which increased with increasing dye concentration (Fig. 2 A and B) (N = 33). By contrast, the signals as depicted in fluorescence ratio R (F _340/380) were unaltered in amplitude and time course during the loading phase (Fig. 2 A and B). The SBFI concentration did not significantly influence the integral of the sodium signals (Fig. 2 C) and, after normalization of the signals to the charge of the accompanying inward currents, no difference was seen in their peak amplitudes (Fig. 2 D), nor decay time constants (Fig. 2 E) with SBFI concentrations of up to 1 mm.

In summary, these results establish that the presence of the sodium indicator SBFI in concentrations of up to 1 mm does not visibly distort sodium signals as reported by its fluorescence ratio. Furthermore, glutamate‐induced sodium signals can be evoked reliably and repetitively in a given cell without significant alteration in their amplitude or time course.

Sodium extrusion from somata

To study the properties of sodium extrusion from cell bodies, CA1 pyramidal neurons were subjected to whole‐cell patch‐clamp and loaded with SBFI though the patch‐pipette. Sodium signals were induced by focal pressure application of glutamate (0.5 mm, 100 ms) close (5–10 μm) to the soma (Fig. 3 A). The rate of recovery from intracellular sodium loads V (mm min⁻¹) was determined by a linear plot of the data points between 1 and 2 s after the peak sodium concentration had been reached, representing the phase of the steepest decline of the recovery from a glutamate‐induced sodium increase (Fig. 3 A).

When glutamate was applied to the soma in the whole‐cell mode, which is with the patch‐pipette still attached (Fig. 3 A), the recovery from somatic sodium loads followed a mono‐exponential decay as described before (Fig. 1 D), exhibiting decay time constants of τ ∼20 s. Because cellular sodium export is expected to be mediated by the NKA, which follows a Michaelis–Menten relationship (Munzer et al. 1994; Zahler et al. 1997), rates of V were plotted against the respective peak sodium concentrations and fitted using a Hill equation to derive V _max, representing the maximum rate for recovery from sodium loads. Under this condition, V _max was 134.8 (93.8, 175.7) mm min⁻¹ (Fig. 3 A).

To evaluate the influence of sodium diffusion into the patch pipette on recovery characteristics, we next loaded cells by bolus injection of SBFI‐AM and again applied glutamate to the soma (Fig. 3 B). Somatic sodium signals now exhibited a strikingly slower decay (τ  = 63.5 ± 48.7 s) and V _max was only 8.3 (7.7, 9.0) mm min⁻¹ (n = 297, N = 61) (Fig. 3 B). Assuming that this recovery is solely mediated by sodium transport across the membrane, this represents a sodium efflux of 0.028 ± 0.001 mm min⁻¹ μm⁻² (average somatic surface area 300 μm²; data taken from http://neuromorpho.org) (Ascoli et al. 2007; Beguin et al. 2013), The apparent K _M as retrieved from the Hill plot was 19.0 (18.4, 19.5) mm. This latter value denotes that the pump is half‐maximally activated at an intracellular sodium concentration of 19 mm, and thus (at an intracellular baseline sodium concentration of ∼13 mm) by sodium increases of ∼6 mm. Saturation was reached at ∼ 30 mm, suggesting that the pump rate is no longer dependent on the amplitude of sodium increases if these exceed ∼17 mm.

The large difference in V _max between cells held in the whole‐cell mode and bolus‐loaded cells strongly indicated that the recovery from sodium loads was dominated by sodium diffusion from the cell body into the pipette in patch‐clamped cells. In the latter, determination of V _max, which should describe enzymatic NKA activity, was therefore meaningless. We thus additionally calculated the average sodium recovery rate for concentrations between 20 and 25 mm, V _20–25, to enable meaningful comparison between the different conditions. In whole‐cell patch‐clamped cells, V _20–25 was 79.9 ± 30.5 mm min⁻¹ (n = 12, N = 5) (Fig. 3 A), whereas it amounted to 6.1 ± 1.7 mm min⁻¹ in bolus‐loaded neurons (n = 108, N = 32) (Fig. 3 B), again emphasizing the strong influence of sodium diffusion into the pipette.

Taken together, these results suggest that the recovery from somatic sodium transients in whole‐cell patch‐clamped cells is largely governed by fast diffusion of sodium into the patch‐pipette that, if it contains a saline with a low sodium concentration, serves as an infinite sink for sodium. In the undisturbed situation, recovery from somatic sodium transients is considerably slower and the maximum sodium recovery rate more than 15‐fold lower, reaching ∼8 mm min⁻¹. Under this assumption that recovery is solely mediated by NKA activity under the latter condition, this corresponds to a maximum sodium extrusion rate of ∼0.03 mm min⁻¹ μm⁻².

Sodium extrusion from dendrites

To determine the sodium recovery rates from dendrites, cells were loaded with SBFI through a patch pipette. For stimulation of dendrites, application pipettes were positioned such that their tip pointed towards a spot on the dendrite which was ∼50 μm away from the soma. With given settings for pressure application, a dendritic section of ∼20 μm in length was directly exposed to the ejection stream and ROIs were accordingly placed (representing a dendritic segment of ∼20 μm in length) (Fig. 3 C). Focal application of glutamate (0.5 mm, 100 ms) close (∼5–10 μm) to the dendrite resulted in sodium transients that decayed with a time constant of τ  = 21.5 ± 14.4 s. V _max was 118.4 (92.9, 143.9) mm min⁻¹ (n = 110, N = 10) and V _20–25 was 81.5 ± 45 mm min⁻¹ (n = 12, N = 5) (Fig. 3 C).

Loading with SBFI‐AM does not result in sufficiently high dye concentrations and tolerable signal‐to‐noise ratio in dendrites, and such measurements were thus not feasible. Instead, we therefore filled a neuron through a patch‐pipette, which was then withdrawn carefully. After resealing of the cell, we applied glutamate to the dendrite as described before. Under this condition, the decay time constant was τ  = 20.3 ± 17.8 s; V _max was 137.1 (115.1, 159.3) mm min⁻¹ (n = 20, N = 4 cells) and V _20–25 amounted to 79.1 ± 19.2 mm min⁻¹ (n = 3 signals from three cells). The apparent K _M under this condition was 18.5 (15.8, 21.2) mm (n = 20, N = 4 cells) (Fig. 3 D), a value close to the one obtained in the soma (see above). Thus, in the dendrite, recovery was not significantly influenced by the presence of the whole‐cell patch‐pipette in the soma.

These data thus reveal that recovery from local sodium transients in dendrites is much faster compared to the soma and V _max is more than 10‐fold higher. The apparent K _M was similar in both compartments, indicating that the sensitivity of the NKA to sodium increases is similar in both compartments.

Influence of sodium extrusion through NKA

Our experiments so far showed that the recovery from local sodium loads is considerably faster in dendrites than in the soma. To analyse the mechanisms of sodium extrusion from these two compartments, we bath‐applied ouabain (100 μm), a specific blocker of the NKA. Somatic sodium signals were studied in cells bolus‐loaded with SBFI‐AM. For measurement of dendritic sodium signals, neurons were subjected to whole‐cell patch‐clamp with a pipette saline containing 1 mm SBFI. After dye loading, cells were perfused with physiological saline and baseline sodium determined before switching to saline containing ouabain (Fig. 4 A and B).

Figure 4. Role of NKA on recovery from sodium loads .

A, left: scheme of the experimental design: glutamate was focally applied to the soma (green area) and sodium signals were measured from the soma as indicated by the red area in a bolus‐loaded neuron. Right: changes in somatic sodium concentration upon perfusion with the NKA inhibitor ouabain (100 μm) as indicated by the bar. During perfusion with ouabain, glutamate was pressure‐applied (0.5 mm, 100 ms) as indicated by the arrowhead. The area indicated by the dotted box is shown enlarged in (C). B, as described in (A) but with focal glutamate application to the dendrite and measurement of dendritic sodium signals. Note that dendritic sodium transients still recover in the presence of ouabain. Right: relationship between sodium concentration and decay velocity (V) as obtained during the first few seconds after the sodium peak. The data were fitted with a Hill equation (red curve). C, glutamate‐induced sodium signals as shown in (A) (dotted area, top trace) and (B) (dotted area, bottom trace) at an enlarged time scale. [Colour figure can be viewed at wileyonlinelibrary.com]

Within 1–2 min after starting the ouabain application, somatic sodium began to rise by 1.8 ± 0.6 mm min⁻¹ (n = 86 cells), a value that is in accordance with earlier work (Rose & Ransom, 1997; Despa et al. 2002). For an average somatic surface area of 300 μm² (Ascoli et al. 2007; Beguin et al. 2013), this value corresponded to a sodium flux of 0.006 ± 0.002 mm min⁻¹ μm⁻². In dendrites, ouabain caused a sodium increase by 0.5 ± 0.1 mm min⁻¹ (n = 8 cells). For an average dendrite diameter of 0.75 μm (Araya et al. 2006) and a surface of 47.1 μm⁻², this corresponds to a sodium flux of 0.01 mm min⁻¹ μm⁻². These values indicate that basal pump rates and constitutive sodium influx are in the same range in both compartments. In soma, as well as in dendrites, sodium kept increasing until a plateau of 40–50 mm was reached (Fig. 4 A and B).

To evoke local sodium signals, glutamate was focally pressure applied either close to the soma (Fig. 4 A) or close to the primary dendrite (Fig. 4 B). In the presence of ouabain, glutamate‐evoked sodium increases in somata added to the strong on‐going increase in ‘baseline’ sodium (n = 22) (Fig. 4 A and C). Recovery from glutamate‐induced sodium loads was not observed and sodium levels just continued to rise until the plateau was reached (Fig. 4 A and C). Thus, in the presence of ouabain, sodium extrusion from the soma was completely blocked.

In dendrites, the situation was quite different. Here, glutamate application, in the presence of ouabain, evoked sodium transients that recovered quickly on the top of the on‐going increase in baseline sodium (n = 22) (Fig. 4 B and C). Measuring the maximal slope, plotting it against the sodium concentration and fitting it with a Hill equation (Fig. 4 B) revealed a V _max of 116.8 (96.6, 136.9) mm min⁻¹ and V _20–25 amounted to 78.7 ± 29.5 mm min⁻¹ (n = 7 signals, N = 3 cells). These values were quite similar to those obtained in the absence of ouabain (Fig. 3 C). Thus, although inhibition of the NKA blocked sodium extrusion from the soma completely, local sodium increases in dendrites recovered with essentially the same speed as under control conditions.

These results strongly indicate that NKA activity is not per se required for the recovery from local sodium loads in dendrites. An additional pathway in this compartment might be diffusion of sodium into neighbouring, non‐stimulated regions.

NKA activity vs. diffusion

To test for the role of diffusion in recovery from glutamate‐induced sodium transients, we performed experiments in which glutamate was pressure‐applied with a wide‐tip pipette, resulting in non‐focal stimulation of essentially the entire cell (Fig. 5 A). Under this condition, the recovery from sodium signals as determined in the dendrite was dramatically slowed. V _max was reduced to 15.5 (12.8, 18.1) mm min⁻¹ (n = 25, N = 7 cells) and V _20–25 to 15.7 ± 1.8 mm min⁻¹ (n = 3, N = 3) (Fig. 5 A). Maximum dendritic sodium efflux thus amounted to 0.33 ± 0.03 mm min⁻¹ μm⁻². The Hill plot revealed an apparent K _M of 15.4 mm and a saturation of pump activity at ∼30 mm. Although flux rates thus differed by a factor of ∼10 compared to the soma, dendritic values for V _max, K _M and saturation were similar to somatic ones (Fig. 3 B), indicating a similar sodium sensitivity of the NKA in both compartments.

Figure 5. Role of diffusion on recovery from sodium loads .

A, left: scheme of the experimental design: glutamate was applied non‐focally with a wide‐tip pipette and the sodium signals were measured in the dendrite (red area). Centre: changes in dendritic sodium concentration upon glutamate application. Right: relationship between sodium concentration and decay velocity (V) as obtained during the first few seconds after the sodium peak. The data were fitted with a Hill equation, the results for V _max and V _20–25 are indicated. B, left: scheme of the experimental design: glutamate was focally applied to the soma (green area) and sodium signals were measured in the soma as indicated by the red area. Top right: changes in somatic sodium concentration at room temperature (21°C) compared to 34°C. Bottom right: overlay of the sodium transients with amplitudes normalized. C, left: scheme of the experimental design: glutamate was focally applied to the dendrite (green area) and sodium signals were measured as indicated by the red area. Top right: changes in dendritic sodium concentration at room temperature (21°C) compared to 34°C. Bottom right: overlay of the sodium transients with amplitudes normalized. D, left: wide‐field‐fluorescence image of a pyramidal neuron, filled with SBFI through a patch‐pipette. Both soma and primary dendrite are clearly visible. Centre: inverted image indicating regions of interest and position of glutamate application pipette for experiment shown on the right. Right: dendritic sodium transients induced by brief glutamate application (arrowhead) close to the stimulation pipette (ROI1) and more proximal to the soma (ROI2 and 3, respectively). Sodium signals in ROI2 and 3 are delayed, reduced in amplitude and exhibit slower kinetics than those in ROI1, indicating sodium diffusion. The red line represents a curve fit of the data to reveal the peak. E, left: scheme of the experimental design: glutamate was focally applied to the soma (green area) and sodium signals were measured in the dendrite as indicated by the red area. Right: changes in dendritic sodium concentration upon application of glutamate to the soma. The red line represents a curve fit of the data to reveal the peak. [Colour figure can be viewed at wileyonlinelibrary.com]

To obtain further evidence for the differential influence of NKA activity on the recovery from local sodium transients in soma vs. dendrites, we performed experiments at near physiological temperature (34°C). Switching to 34°C caused an acceleration of the initial recovery rate from glutamate‐induced sodium transients by a factor of 2.56 in somata, resulting in a Q ₁₀ of 1.97 (n = 36, N = 4) (Fig. 5 B). With local application of glutamate to dendrites, in contrast, recovery was unaltered when switching to high temperature (Q ₁₀ = 0.94; n = 8, N = 8 for both temperatures) (Fig. 5 C).

We next probed for lateral diffusion of sodium along dendrites. To this end, we performed experiments in which glutamate was applied focally to a dendrite and resulting sodium transients were determined in two to three different dendritic ROIs: one ROI located directly within the ejection stream of the pressure application pipette, and further ROIs positioned more proximal to the soma (n = 17, N = 3) (Fig. 5 D). As observed before, a fast rise in sodium followed by a mono‐exponential recovery phase was detected close to the application pipette. In dendritic regions located further away from the stimulation pipette, rise times increased, peak amplitudes decreased and the delay between the stimulation events and peaks increased (Fig. 5 D). Based on the assumption that this phenomenon was largely a result of the intracellular diffusion of sodium, its apparent diffusion coefficient was calculated. To this end, the distance between the two ROIs was determined using the centre of each as anchor point. The delay between peak sodium concentrations was determined and the parameters time and distance were introduced in the one‐dimensional diffusion equation (D∼x ²/2t), resulting in a D _Na+ of 328 ± 20 μm² s^–1 for sodium diffusion along dendrites (n = 17; N = 3).

Finally, we probed for diffusion of sodium from somata into dendrites. To this end, glutamate was applied locally to the cell body of a patch‐clamped cell and a region of interest set on the dendrite. Upon glutamate application, sodium transients were also detected in dendrites (n = 4, N = 4) (Fig. 5 E). These, however, exhibited slow rise times, small amplitudes and blunted peaks, indicating that they were mostly a result of the diffusion of sodium from out of the soma along the dendrites.

Taken together, these results suggest that, although the NKA is the major mechanism responsible for the recovery from sodium loads from cell bodies, lateral diffusion is the predominant pathway for the recovery from local sodium increases in dendrites. Also, lateral diffusion provides for a much faster sodium extrusion from dendrites with a more than 10‐fold higher rate compared to NKA‐mediated extrusion observed in the soma.

Recovery from sodium loads under conditions of energy deprivation

Whereas diffusion is only dependent on the existence of ion gradients and does not primarily break down energy, NKA activity is a major energy consumer in neural cells. Consequently, and based on our results obtained so far, reduction of cellular ATP levels should not have an immediate impact on the recovery of local sodium transients in dendrites, but should strongly affect sodium extrusion from somata.

To test this hypothesis, we perfused slices with saline containing 4 mm sodium fluoride (NaF). NaF decreases cellular pyruvate and ATP levels by inhibition of enolase, which catalyses the second to last step of the glycolysis (Cox & Bachelard, 1982; Lees, 1991; Bizzozero et al. 1999). To gain evidence that NaF resulted in a decrease in neuronal ATP concentration in our preparation, we monitored changes in intracellular free Mg²⁺ concentration using the fluorescent indicator dye MgG, as described previously (Chatton et al. 2003; Magistretti & Chatton, 2005; Fernandez‐Moncada & Barros, 2014). Free magnesium provides an indirect evaluation of intracellular ATP levels because ATP exhibits ∼10‐fold higher affinity for Mg²⁺ than ADP and binds a large proportion of cellular magnesium (Leyssens et al. 1996). To this end, neurons were bolus‐loaded with MgG‐AM and baseline fluorescence was monitored before adding NaF.

Upon perfusion with NaF, MgG fluorescence emission started to increase until it reached a stable level, which was 25 ± 21% higher than the initial baseline fluorescence, indicating an increase in the intracellular free magnesium concentration (n = 11; N = 3) (Fig. 6 A). Removal of extracellular magnesium during this phase caused a drop in MgG fluorescence by 10 ± 7% (n = 12, N = 3) (Fig. 6 A). Although this suggests that influx of magnesium from the extracellular space might have contributed to the NaF‐induced increase, this result also demonstrated that the greater part of the signal was the result of an intracellular magnesium source, most probably its unbinding from ATP in the course of a decrease in intracellular ATP levels.

Figure 6. Sodium extrusion under metabolic stress .

A, left: schematic drawing indicating mechanism of inhibition of glycolysis by NaF. Centre: increase in Mg‐Green fluorescence relative to baseline fluorescence upon bath perfusion with NaF as indicated by the bar. Right: changes in MgG fluorescence induced by bath perfusion with NaF and during additional removal of extracellular magnesium (0 Mg²⁺) as indicated by the bars. B, left: scheme of the experimental design: glutamate was focally applied to the soma and sodium signals were measured from the soma as indicated by the red area in a neuron loaded with SBFI‐AM. Centre: somatic sodium concentration recorded 20 min after starting perfusion with the glycolysis inhibitor NaF. Arrowheads indicate repetitive focal application of glutamate (0.5 mm, 100 ms) to the soma. Right: sodium extrusion rate (V) as determined during the first few seconds after the sodium peak under control conditions and 20 and 25 min after starting perfusion with NaF. C, as in (B) but with focal glutamate applications in the proximity of the dendrite and recorded in the red area; the patch pipette was retracted from the soma. There is no significant difference in extrusion rates between control and after perfusion with NaF. D, as in (C) but glutamate was applied to a large area of a cell. B–D, boxes represent quartiles, whiskers correspond to the 95% confidence interval. Middle lines show the median and squares the mean value. n.s., not significant, ^*0.01 ≤ P < 0.05, ^*** P < 0.001. [Colour figure can be viewed at wileyonlinelibrary.com]

To analyse recovery from sodium loads under this condition, neurons were filled with SBFI via a patch pipette upon which the pipette was carefully retracted. After resealing of the cell, slices were perfused with NaF for 20 min and then subjected, repetitively, to puffs of glutamate to evoke sodium transients. When glutamate was focally pressure‐applied to the soma, the first one to two applications to neuronal somata resulted in well detectable sodium transients with an average amplitude of 4.8 ± 2.5 mm ([Na⁺]_i = 18.3 ± 2.5 mm). Sodium transients recovered only slightly slower than those in the same range of concentrations under control conditions [maximum rate in the absence of NaF: 3.7 ± 1.5 mm min⁻¹ (n = 192, N = 50); at 20 min of NaF perfusion: 3.2 ± 2.0 mm min⁻¹ (n = 72, N = 28)] (Fig. 6 B). However, with repeated glutamate application, peak amplitudes decreased, recovery slowed and sodium started to accumulate (Fig. 6 B). Accordingly, the somatic recovery rate dropped to 1.0 ± 1.0 mm min⁻¹ after 25 min in the presence of NaF (n = 133, N = 36) (Fig. 6 B).

Again, the situation was completely different when glutamate was locally applied to dendrites. Here, and as observed before, recovery from sodium loads was much faster than in the soma (rate at the same range of sodium concentrations (14.2 ± 1.8 mm) was 35.5 ± 25.9 mm min⁻¹ in the control (n = 13, N = 8) (Fig. 6 C). Moreover, transients remained largely unaltered even in continued presence of NaF and no increase in the sodium baseline was seen [extrusion rate: 26.9 ± 19.5 mm min⁻¹ after 20 min (n = 6, N = 3) and 28.8 ± 13.7 mm min⁻¹ after 25 min (n = 12, N = 3)] (Fig. 6 C).

Finally, we repeated these experiments with non‐focal application of glutamate to evoke sodium transients throughout the entire cell, preventing the generation of intracellular sodium concentration gradients and diffusion. As opposed to focal glutamate application in the presence of NaF, dendritic sodium transients now dramatically changed with repeated stimulation and an increase in baseline sodium was observed (n = 9, N = 3) (Fig. 6 D). Although the recovery rate was 26.8 ± 23.8 mm min⁻¹ after 20 min of NaF perfusion (n = 9), it dropped significantly to 4.9 ± 7.7 mm min⁻¹ after 30 min (n = 6) (Fig. 6 D).

Taken together, our data show that impairment of energy metabolism significantly influences the recovery from sodium transients in the soma, in agreement with the demonstrated main role of NKA, and its dependence on cellular ATP, for sodium extrusion. In dendrites, in contrast, recovery from local sodium loads is unaffected by periods of up to 30 min of metabolic impairment, supporting its independence on ATP on this time frame and emphasizing the predominant role of sodium diffusion in this process.

Numerical modelling of somatic and dendritic Na⁺ extrusion

To test the validity of our experimental findings, we applied a numerical simulation (see Methods), employing the parameters for NKA pump rate, K _M, sodium flux and diffusion determined experimentally. Sodium influx was first placed onto the soma to obtain an increase by ∼18–20 mm in this compartment (Fig. 7 A, left). With NKA active, somatic sodium recovered mono‐exponentially from this increase. In addition, diffusion‐based sodium increases developed in the two adjacent dendritic segments, whereas, in the third (quasi infinite) dendritic compartment, almost no sodium increase occurred (Fig. 7 A, left). Recovery from the somatic sodium load was strongly dependent on NKA. With pumps turned off, sodium recovery towards baseline was extremely slowed in the soma and the two adjacent dendritic compartments, accompanied by a slow sodium increase in the third dendritic compartment (Fig. 7 A, right).

Next, sodium influx was placed on the second dendritic compartment to mimic the local application of glutamate onto this compartment. Dendrites recovered quickly from this sodium load, and sodium diffusion resulted in a detectable increase in the neighbouring proximal dendritic compartment (Fig. 7 B, left). By stark contrast to what was seen in the soma, dendritic sodium also recovered efficiently with pumps turned off, albeit recovery was slowed especially in the later phase (Fig. 7 B, right).

In summary, these simulations support our conclusion that NKA‐mediated extrusion is critical for recovery from somatic sodium loads. Although diffusion of sodium from the soma into the proximal apical dendrite is visible, it does not mediate efficient clearance of somatic sodium when NKA is blocked. On the other hand, diffusion is the dominating mechanism for the initial fast recovery from local sodium loads from dendrites, whereas NKA‐mediated extrusion shapes and accelerates recovery at later phases.

Discussion

Using ratiometric imaging with the sodium indicator SBFI, we demonstrate that local puff application of glutamate induces transient sodium increases in mouse CA1 pyramidal neurons. Glutamate‐evoked changes in sodium as reported by SBFI are not dependent on the concentration of the dye, demonstrating that it does not visibly buffer, nor distort sodium transients. Recovery from somatic sodium transients followed a mono‐exponential time course. A Hill plot revealed a maximum sodium recovery rate of 8 mm min⁻¹, corresponding to a flux of 0.03 mm min⁻¹ μm⁻², and an apparent K _M of 19 mm. Somatic sodium extrusion was completely blocked in the presence of the NKA blocker ouabain and significantly hampered upon inhibition of glycolysis by NaF. In addition, somatic recovery rates were doubled by an increase in temperature by 10°C. In dendrites, maximum sodium recovery rates were more than 10‐fold faster than in the soma with local glutamate application. Moreover, dendritic recovery was largely unaltered at elevated temperature and, in the time frame studied, by ouabain or during inhibition of glycolysis. With non‐focal application of glutamate, in contrast, dendritic sodium recovery rates were reduced to those found in the soma and were diminished during application of NaF. Numerical modelling supports the conclusion that NKA activity is required for recovery from somatic sodium loads, whereas diffusion is the dominating mechanism for recovery from local sodium loads in dendrites.

SBFI as a reporter of intracellular sodium concentration

Ratiometric imaging with SBFI is a reliable tool for the quantitative measurement of intracellular sodium concentrations (Minta & Tsien, 1989; Levi et al. 1994; Rose & Ransom, 1997; Chatton et al. 2000; Diarra et al. 2001). Although the relationship between SBFI fluorescence and sodium concentration generally follows Michaelis–Menten kinetics (Donoso et al. 1992; Rose et al. 1999), there is an acceptable linear correlation in the concentration range between ∼5 and 40 mm, as reported previously (Meier et al. 2006) and in the present study. Within this concentration range, such linearization enables the calculation of absolute changes in sodium concentration without the need to know the baseline concentration of each individual cell (and cellular compartment, respectively).

In the present study, we also employed a new approach for the determination of baseline sodium concentration in cell bodies of CA1 neurons. By comparing SBFI ratio levels before and after loading the cell with a known sodium concentration through dialysis with the intracellular saline of a patch‐pipette, we obtained a value of ∼13 mm, which is in good accordance with earlier studies (Donoso et al. 1992; Langer & Rose, 2009; Kelly & Rose, 2010 a; Azarias et al. 2012; Karus et al. 2015 a). Chemical ion indicator dyes bind an ion of choice selectively and therefore may exert a buffering effect which distorts ion signals. This is especially relevant for high‐affinity calcium indicator dyes such as Fura‐2 or Oregon‐Green (Neher & Augustine, 1992; Zhou & Neher, 1993; Maravall et al. 2000). SBFI has a K _D of ∼24 mm (Donoso et al. 1992; Jung et al. 1992; Rose et al. 1999; Sheldon et al. 2004; Meier et al. 2006). Relevant buffering of sodium is therefore not expected because applicable dye concentrations (0.1–1 mm) are far below this value (Sabatini et al. 2001). However, because this has not yet been confirmed experimentally, we tested this theoretical prediction by investigating whether glutamate‐induced sodium signals were altered during loading of the dye to concentrations of up to 1 mm, an approach that has been introduced previously (Neher, 1995; Helmchen et al. 1996). Although the amplitudes of single‐wavelength fluorescence emission signals (F ₃₈₀) increased with increasing dye concentrations, the calculated ratio signals (F _340/380) were not altered. Our data thus clearly demonstrate that the presence of SBFI does not imply a relevant distortion of intracellular sodium transients, a finding that is of critical importance for the correct interpretation of the further results of the present study.

Sodium extrusion from somata

To study extrusion of sodium in neurons, we imposed transient intracellular sodium increases by puff application of glutamate, which results in influx of sodium through AMPA and NMDA receptor channels (Rose & Konnerth, 2001; Lamy & Chatton, 2011). As reported previously (Rose et al. 1999; Bennay et al. 2008), recovery from transient sodium loads followed a mono‐exponential decay and was characterized by large decay time constants. In whole‐cell patch‐clamped cells, the maximum velocity at which sodium recovered was ∼135 mm min⁻¹. This value was more than 15‐fold higher than in bolus‐loaded neurons (see below), indicating a rapid diffusional exchange between the soma and the pipette as described previously (Pusch & Neher, 1988). In AM‐ester loaded cells, recovery from somatic sodium transients was considerably slower, reaching a maximum rate of ∼8 mm min⁻¹ under control conditions. This peak rate is in the same range as those reported from cultured myocytes (Despa & Bers, 2003; Despa et al. 2004).

The recovery rate was dependent on the intracellular sodium concentration and a Hill plot revealed an apparent K _M of 19 mm with saturation above 30 mm. Recovery was completely blocked in the presence of ouabain, indicating that it was mediated by the export of sodium through NKA. Because NKA activity (but not diffusion) is highly temperature‐sensitive, the dominating role of the NKA in sodium extrusion from somata was supported by the observation that an increase in the temperature by 10 °C accelerated the recovery rates by a factor of two (den Hertog & Ritchie, 1969; Skou & Esmann, 1992; Rose et al. 1999). Our simulation is in line with these experimental findings, predicting that NKA activity is required for recovery from somatic sodium loads.

Hippocampal neurons express two α‐isoforms of the NKA, namely α1 and α3 (Sweadner, 1992). When expressed in HeLa cells, α1 has a low K _M for sodium (12 mm), whereas the K _M for sodium is 33 mm for α3 (Zahler et al. 1997). From these data, it was concluded that α1 sets baseline sodium concentrations and α3 specifically handles rapid recovery from large sodium increases in small compartments following neuronal activity (Munzer et al. 1994; Zahler et al. 1997). A recent study has provided evidence that α1 and α3 mediate functional sodium export from cultured rat neurons (Azarias et al. 2012), and the apparent K _M as determined in the present study (19 mm) probably reflects the activity of both isoforms. For dendrites, similar values for V _max, K _M and saturation were found, indicating a similar isoform expression profile as in somata.

NKA activity is strongly dependent on the availability of intracellular ATP and it is estimated that the pump consumes ∼50% of all the energy needed by the CNS (Ames, 2000). To challenge cellular metabolism, we inhibited glycolysis with NaF. This resulted in an increase in cellular MgG fluorescence, indicating an increase in free magnesium and suggesting a decrease in ATP as reported before (Chatton et al. 2003; Magistretti & Chatton, 2005; Fernandez‐Moncada & Barros, 2014). NaF perfusion also resulted in a gradual impairment of recovery from glutamate‐induced transient sodium loads, as well as in a gradual build‐up of intracellular sodium in cell bodies, indicative of a reduction in NKA activity.

At an intact mitochondrial respiration, the prominent influence of inhibition of glycolysis on sodium export is at first surprising. Glycolysis produces only ∼5% of the cellular ATP (Erecinska & Silver, 1989), which will certainly not be sufficient to feed NKA. Inhibition of glycolysis, on the other hand, will ultimately result in a reduction of pyruvate and thereby also decrease the availability of substrates for the tricarboxylic acid cycle and for mitochondrial respiration. Along those lines, a previous study provided evidence that the NKA is mainly fuelled by ATP produced by oxidative phosphorylation (Fernandez‐Moncada & Barros, 2014), which, in the light of the high energy demand of the NKA, indeed appears to be conclusive.

Recovery from sodium loads in dendrites

With focal puff application of glutamate, the maximum rate for sodium recovery in dendrites was in the same range as that of somata of whole‐cell patch‐clamped cells, and 10‐fold higher that of the somata of non‐patched cells. Although perfusion with ouabain resulted in an immediate increase in intracellular sodium in dendrites, indicating inhibition of NKA, recovery from local sodium transients was still largely intact and V _20–25 under this condition was almost unaltered. By contrast to what was seen in the soma, recovery from dendritic sodium transients was also almost unchanged upon an increase in temperature to 34°C. Our numerical simulation demonstrated the feasibility of this experimental finding, predicting a rapid return of locally‐induced dendritic sodium transients to baseline even with pumps turned off.

These results strongly indicate that recovery from local sodium loads in dendrites is largely governed by lateral diffusion along the dendrite and independent from NKA activity. A similar result has been obtained in axon initial segments of layer 5 pyramidal neurons. Here, the recovery from action‐potential induced sodium transients was not significantly altered by ouabain, indicating that sodium clearance was mainly mediated by passive diffusion (Fleidervish et al. 2010). Our experiments also revealed an estimate for sodium diffusion along primary dendrites. The resulting value of ∼330 μm² s^–1 is around half that reported from frog muscle fibres (600 μm² s^–1), and significantly smaller than that of large lizard axons (1300 μm² s^–1) (David & Barrett & Barrett, 1997), indicating that diffusion in rodent dendrites is considerably hindered compared to the cell types and compartments described above.

The notion that fast diffusion dominates over NKA activity for recovery from local sodium transients in dendrites was also supported by experiments in which glutamate was applied non‐focally to eliminate sodium gradients and diffusion. Under this condition, dendritic V _max dropped from 120 to ∼16 mm min⁻¹ and V _20–25 decreased from 80 to 15 mm min⁻¹, probably now primarily reflecting dendritic NKA activity. Calculated peak sodium fluxes in dendrites were ∼10‐fold higher than those obtained with glutamate application to the soma (0.33 vs. 0.03 mm min⁻¹ μm⁻²). Thus, although maximum extrusion rates (mm min^–1) are two‐fold higher in dendrites compared to somata, this difference multiplies for sodium fluxes, indicating that the much larger surface‐to‐volume ratio in dendrites also bring about much higher absolute sodium export rates by the sodium pump.

In cultured rat hippocampal neurons, a recent study reported initial extrusion rates of 34 mm min⁻¹ from dendrites obtained after removal of extracellular potassium to block NKA (Azarias et al. 2012), a value more than twice that obtained in the present study. This difference in the concentration‐based extrusion rates might result from a different surface‐to‐volume ratio of the compartment under study. Thus, a given membrane transport activity will result in a slower recovery rate in a larger compared to a smaller cellular compartment (e.g. a finer dendrite).

A direct consequence of the dominance of diffusion vs. local extrusion in dendrites was the primary independence of local dendritic sodium transients on an intact glycolysis. In dendrites, application of NaF (which resulted in gradual failure of sodium export from cell bodies), neither affected the amplitude, nor the recovery of sodium transients induced by repetitive focal glutamate application induced within the same time frame. Again, when diffusion was hampered by non‐focal application of glutamate, dendrites experienced a similar build‐up of sodium and failure of sodium extrusion than somata. These data indicate that reduced ATP availability does not hamper recovery from local sodium loads in dendrites, at least in the time frame investigated in the present study. As a reverse conclusion, local sodium transients will not per se induce a local increase in ATP consumption and will thus not represent a significant local challenge for metabolism. Of course, and as indicated by our measurements, sodium ions entering the cytosol will ultimately be exported through the NKA if there is no diffusion gradient for sodium and/or if there is an overall increase in cytosolic sodium.

Functional consequences and conclusions

Excitatory neuronal activity causes large sodium transients in central neurons as a result of the opening of sodium‐permeable voltage‐ and ligand‐gated ion channels (Rose, 2002). These sodium transients can either be local, as is the case for local activation of glutamatergic synapses (Rose & Konnerth, 2001), encompass the entire dendritic tree as is the case for back‐propagating action potentials (Jaffe et al. 1992; Rose et al. 1999), or even occur as global network oscillations as described during epileptiform activity (Karus et al. 2015 a). The results reported in the present study show that recovery from these sodium transients is mediated by two central mechanisms: lateral intracellular diffusion to neighbouring, non‐stimulated areas and extrusion through the plasma membrane by the NKA. The relative weight and the contribution of these two mechanisms depend on the form of activity and the cellular compartment.

Our data emphasize that the NKA is the central mechanism for export of sodium from neurons. For recovery from sodium transients, this general principle, is, however, only valid for cell bodies and for global sodium loads imposed on the entire cell. When its function is compromised, either by ouabain or by energy depletion, sodium homeostasis and recovery from additional sodium loads fail. In dendrites, after localized sodium influx, sodium is mainly removed by fast lateral diffusion. Consequently, our data also suggest that sodium increases in cellular microdomains, such as those generated during excitatory synaptic activity (Lasser‐Ross & Ross, 1992; Knopfel et al. 2000; Rose & Konnerth, 2001; Bennay et al. 2008), will probably not require a local increase in ATP consumption, nor energy metabolism.

The concept of ‘diffusion over extrusion’ also implies that sodium recovery in dendrites is one order of magnitude faster than if it was mediated by local NKA activity only. Although sodium ions will be pumped out by the NKA at some point, our study thus suggests that energy requirements will not surge locally after a local sodium increase but rather be disseminated to and shared with the entire cell. However, when subjected to global cellular sodium increases, such as during synchronized epileptiform discharges, sodium recovery will be mediated by NKA only and even small‐amplitude sodium increases will thus be maintained for a relatively long time as shown recently (Karus et al. 2015 a). It can be assumed that, under such conditions, support of neuronal metabolism and sodium homeostasis by astrocytes will be especially critical (Allaman et al. 2011; Barros, 2013; Karus et al. 2015 a). Moreover, an additional weakening or impairment of energy metabolism during global sodium increases will have direct and immediate consequences on the capacity of neurons to maintain low intracellular sodium (Karus et al. 2015 a; Karus et al. 2015 b) and thereby promote additional fatal consequences such as the reversal of sodium‐calcium exchange and calcium load (Hertz, 2008).

Additional information

Competing interests

The authors declare that they have no competing interests.

Author contributions

All imaging experiments were performed at the Institute of Neurobiology, Heinrich Heine University Duesseldorf, Germany. Modelling was performed at the Carl‐Ludwig‐Institute for Physiology, Medical Faculty, University of Leipzig, Liebigstrasse 27, 04103 Leipzig, Germany. MAM, HS and CRR were responsible for the conception and design of the experiments. MAM, HS, JL, CK, KWK and CRR were responsible for the collection, assembly, analysis and interpretation of data. MAM, HS, JL, CK, KWK and CRR were responsible for drafting the article or revising it critically for important intellectual content. All authors contributed to critical revision of the manuscript for intellectual content and final approval of version to be published. All persons designed as authors qualify for authorship and all those who qualify for authorship are listed.

Funding

This work was supported by the Deutsche Forschungsgemeinschaft (German Research Foundation; Ro2327/6‐1).