Spreading depolarizations exhaust neuronal ATP in a model of cerebral ischemia

Significance

Spreading depolarizations (SD) occur in many brain pathologies, and are suggested to contribute to brain damage through a mismatch in energy supply and demand. However, measurements directly demonstrating an imbalance between adenosine triphosphate (ATP) production and consumption, particularly in individual neurons, have not been reported. Here, we show that SDs lead to a transient decrease in intracellular neuronal ATP even in presence of glucose and oxygen. When the supply of oxygen and glucose was interrupted, ATP declined gradually until SD led to the exhaustion of neuronal ATP. This process would be terminal without the renewed supply of oxygen and glucose. Therefore, therapies targeting SDs could preserve neuronal ATP or delay its loss in brain pathologies.

Abstract

Spreading depolarizations (SDs) have been identified in various brain pathologies. SDs increase the cerebral energy demand and, concomitantly, oxygen consumption, which indicates enhanced synthesis of adenosine triphosphate (ATP) by oxidative phosphorylation. Therefore, SDs are considered particularly detrimental during reduced supply of oxygen and glucose. However, measurements of intracellular neuronal ATP ([ATP]_i), ultimately reporting the balance of ATP synthesis and consumption during SDs, have not yet been conducted. Here, we investigated neuronal ATP homeostasis during SDs using two-photon imaging in acute brain slices from adult mice expressing the ATP sensor ATeam1.03^YEMK in neurons. SDs were induced by application of potassium chloride or by oxygen and glucose deprivation (OGD) and detected by recording the local field potential, extracellular potassium, as well as the intrinsic optical signal. We found that, in the presence of oxygen and glucose, SDs were accompanied by a substantial but transient drop in neuronal ATP sensor signals, corresponding to a drop in ATP. OGD, which prior to SDs was accompanied by only a slight reduction in ATP signals, led to a large, terminal drop in ATP signals during SDs. Subsequently, we investigated whether neurons could still regenerate ATP if oxygen and glucose were promptly resupplied following SD detection, and show that ATP depletion was essentially reversible in most cells. Our findings indicate that SDs are accompanied by a substantial increase in ATP consumption beyond production. This, under conditions that mimic reduced blood supply, leads to a breakdown of [ATP]_i. Therefore, our findings support therapeutic strategies targeting SDs after cerebral ischemia.

Spreading depolarizations (SD) have been electrocorticographically recorded with subdural electrodes during neurocritical care in various situations: during the dying process after cardiac arrest, during the development of brain death with continued systemic circulation, during the evolution of delayed ischemic infarcts after subarachnoid hemorrhage (SAH), following malignant hemispheric stroke (MHS) due to middle cerebral artery occlusion, after traumatic brain injury (TBI), in Moyamoya vasculopathy, and e.g., subdural hematoma and status epilepticus in the context of acute brain injury (1–10). Two clinical trials of larger scale in SAH and TBI have found that SDs are associated with worse patient outcome (5, 11). A recent study in MHS patients has shown that SD-induced depression of neuronal activity is an electrocorticographic indicator of infarct growth (12), consistent with findings from animal studies indicating that interventions antagonizing SDs reduce cerebral damage (13–15). The SD continuum describes the spectrum from transient waves with negative direct current (DC) shifts of short to intermediate duration in adequately supplied or less ischemic tissue, to terminal waves in severely ischemic tissue characterized by long-lasting DC shifts and transition of the neurons from the state of injury to cell death (16, 17). Thus, SD recordings are indicative of the neuronal metabolic state.

SDs are characterized by a near-complete depolarization of neurons and glial cells, and a near complete breakdown of transmembrane ion gradients accompanied by a net influx of cations and water into the neurons (18). From an electrochemical and thermodynamic point of view, SD is the abrupt, passive transition from a state of low entropy to a state of high entropy, which involves the loss of energy stored in electrochemical gradients (19, 20). Subsequently, the recovery of cells from SD, i.e., the restoration of ion gradients, requires chemical energy, e.g., adenosine triphosphate (ATP). There is compelling evidence that activation of Na⁺/K⁺-ATPases is paramount to terminate SD and prevent neuronal death due to prolonged intracellular overload, e.g., with Na⁺ (21, 22) and Ca²⁺ (23). Increases in the intracellular concentration of sodium ions ([Na⁺]_i) from ~10 to ~35 mM and in the extracellular concentration of potassium ions (“[K+]o”) from ~3 to ~10 mM, which is even exceeded during SDs (16, 24), were shown to strongly stimulate Na⁺/K⁺-ATPases to near maximal velocity, while further increases did not result in significant further activation (25–28). Increased Na⁺/K⁺-ATPase activity causes a rise in ATP consumption and demand. In response, ATP production by oxidative phosphorylation is thought to increase, which has been experimentally demonstrated by measurements of higher cerebral metabolic rate of oxygen (CMRO₂) (29–31). At the same time, overall tissue ATP concentrations were shown to decrease by around 50% even in otherwise healthy parenchyma despite increased ATP synthesis in some studies (32, 33). These data support the hypothesis of reduced neuronal intracellular ATP concentrations ([ATP]_i) during SDs, however, corresponding measurements, particularly at single-cell resolution, have not yet been conducted.

Here, we monitored neuronal [ATP]_i semiquantitatively during SDs in acute brain slices from mice expressing the Förster resonance energy transfer (FRET) based ATP sensor ATeam1.03^YEMK under the neuronal promotor Thy1.2 (34, 35): first in the presence of oxygen and glucose, then during oxygen and glucose deprivation (OGD), and finally upon the renewed supply of oxygen and glucose after real-time detection of SDs.

Results

Imaging of ATP Dynamics during SDs.

To monitor neuronal [ATP]_i, we acquired time series of the neuronally expressed ATP sensor (Fig. 1A) (34, 35), either in multiple brain regions (referred to as “large scale” imaging) or in up to a dozen individual neurons at single-cell resolution. The ratio of yellow fluorescent protein (YFP) over cyan fluorescent protein (CFP) of the FRET-based sensor provided a semiquantitative readout of [ATP]_i, hereafter referred to as “ATP signals.” In combination with established methods to induce SDs, i.e., by application of KCl (3M) (Fig. 1A “KCl puff”) or OGD, and to detect SDs, i.e., by electrophysiological recordings of the DC local field potential (LFP), of the extracellular potassium ion concentration ([K⁺]_o), as well as the intrinsic optical signal (IOS, Fig. 1A) (14, 36, 37), we were able to investigate neuronal ATP signals in the course of SDs.

Fig. 1.

SDs induce a transient drop in neuronal ATP. (A) Experimental setup. SDs were evoked by puff application of KCl (3M, Right), while the LFP and the extracellular K⁺ concentration ([K⁺]_o) were recorded by double-barreled electrodes (Left) in acute mouse brain slices. Laser-scanning microscopy with two-photon excitation (“exc”) was used to record the fluorescence of the ATP-sensor ATeam 1.03^YEMK (expressed under the neuronal promoter Thy1.2) as well as the transmission and the IOS (i.e., the %-change in transmission). The FRET signal, i.e., the ratio of the acceptor and donor fluorescence (from yellow and cyan fluorescence proteins, “YFP” and “CFP,” respectively), correlates with the intracellular ATP concentration ([ATP]_i). (B, Left) Transmission image of an acute mouse brain slice (300 µm thickness) prior to induction of an SD (“baseline”) by K⁺ puff application (1s). Black and blue lines outline the combined LFP/extracellular potassium electrode (LFP/[K⁺]_o) and the KCl puff electrode, respectively. Dotted circles denote regions of interest (ROIs), three in the neocortex, one subcortical, and one in the white matter (CTX1-3, SUBC, and WM, respectively). (Middle and Right) IOS images 15, 75, and 195 s after SD induction. The dashed line in the “15s” IOS image denotes the wavefront of the SD. Note how the SD propagates from the induction site (“KCl puff electrode” near CTX1) to the other neocortical and subcortical ROIs, yet leave out the WM. (C) Corresponding color-coded images of the ATeam signal (YFP/CFP) based on the maximum intensity z-projection (7 z-planes at 5 µm intervals) of background-corrected ratio images of the donor (CFP) and the acceptor-domain (YFP). From left to right: ATeam-signal before (baseline) and after SD induction. Note how the transient reduction of the ATeam signal follows the SD (cf. B) first reaching the cortical (75s) ROIs, then the subcortical ROI (195s). Cooler colors indicate lower ATP signals. (Da) Parallel LFP (Top) and [K⁺]_o signals (Bottom), corresponding to the experiment shown in B and C. Note the transient negative shift of the DC potential in the LFP signal as well as the transient increase in [K⁺]_o (Bottom), characteristic for SDs. Also note that the LFP trace is oriented according to the convention of electroencephalography (EEG) with negativity up and positivity down. (Inset) Temporally enlarged, M-shaped LFP signal typical for SDs. The dotted vertical line continues to Db illustrating the relative timing of the shown signals. (Db) ATP signals from the ROIs indicated in B and C during the SD. Note the propagation of the SD from “CTX1” to “SUBC” and that the WM was not affected by the SD. The vertical bars indicate the onset of SD as detected by the IOS per ROI. (Ea) Single-cell resolution, color-coded background-, and gain-corrected ratio images (20 z-planes, 1.5 µm interval) capturing four neurons (labeled 1 to 4) before (Left), during (Middles), and 10 min after an SD (Right) induced by K-puff application. Letters in the Top Right corner (i, ii, iii) mark the timepoint of imaging corresponding to the curves shown in Eb. (Eb) Self-normalized temporal ATeam profiles of the four neurons labeled in Da. Note the synchronized reduction followed by recovery of ATP signals in all four neighboring neurons. (F) Average of ATeam-profiles synchronized to the SD onset from 99 self-normalized cells captured during 21 SDs (median ± IQR/2, 14 slices, 8 mice). (G) Summary boxplot of ATeam-signals shown in E and F at the end of the baseline normalized to the baseline mean (typically 3 to 5 min, but up to ~1 h in a subset of slices (SI Appendix, Fig. S2A), at the SD peak, and 10 min after the SD. Statistics are based on mean values of 2 to 12 cells captured per ROI during individual SDs (n = 21 SDs, 14 slices, 8 mice, ***P < 0.001, Wilcoxon-test & Bonferroni correction). Box plots indicate the median, IQ1, and IQ3, whiskers the extrema; individual data points are indicated by dots.

SDs Induce a Transient Drop in Neuronal ATP.

We first addressed the question of whether ATP consumption exceeds production during SDs in neurons, here in presence of oxygen and glucose. Focal induction of SDs by KCl (Fig. 1 A and B) led to SDs propagating across the neocortex to subcortical regions, as indicated by changes in the IOS, reflecting changes in water distribution between the intra- and extracellular space, which affects refractive tissue properties (Fig. 1B) (36, 37). The DC potential showed a negative shift and [K⁺]_o increased sharply upon SD onset at the site of the recording electrode (Fig. 1Da). Parallel large-scale imaging revealed a substantial but transient reduction of the ATP sensor signal by 8 to 16% in cortical as well as subcortical regions of interest (ROIs, Fig. 1 C and Db). Exemplary CFP and YFP raw transients corresponding to ROI “CTX1” are shown in SI Appendix, Fig. S1A. Notably, the changes in ATP signals follow the propagation pattern of the SD (Fig. 1 C and Db and cf. Fig. 1B). Neither IOS nor ATP signals changed in the white matter ROI, consistent with the finding that SDs are a gray matter phenomenon (38).

Imaging at single-cell resolution confirmed the transient reduction in ATP signals during SDs in individual neurons (Fig. 1 Ea and Eb; CFP and YFP signals of cell 1 are shown in SI Appendix, Fig. S1B). On average, the neuronal ATP signal decreased by 13% [10%, 18%] [median, interquartile 1 (IQ1), and IQ3], but recovered in all experiments (Fig. 1 F and G, n = 21 SDs, 14 slices, 8 mice, 99 individual neurons; P < 0.001, Wilcoxon signed rank & Bonferroni correction). To assure long-term stability of the ATP signal, we performed recordings for ~55 min without experimental intervention (n = 17 cells, 3 slices, 3 mice, SI Appendix, Fig. S2), revealing stable neuronal ATP signals. Notably, inhibition of the Na⁺/K⁺-ATPase by ouabain did not reduce the transient drop in the ATP signal, while increasing baseline [K⁺]_o confirmed drug activity (SI Appendix, Fig. S3). To test the contribution of Ca²⁺-ATPases, we induced SDs in Ca²⁺-free artificial cerebrospinal fluid (aCSF), which significantly reduced the transient drop in the ATP signal due to SDs from −7% [−1%, −15%] to 1% [−4%, 3%] (ctrl and Ca²⁺-free aCSF, respectively, n = 8 slices from 4 and 3 mice; P = 0.004, Mann–Whitney U test & Bonferroni correction; SI Appendix, Fig. S3 C and D). In summary, neuronal ATP signals can drop transiently during SDs, even when oxygen and glucose are available.

ATeam1.03^YEMK has been shown to be partially pH sensitive, particularly for ATP levels above 1 mM when investigated as a purified protein (39). To test whether changes in pH underlie the differences in the ATP signals in Ca²⁺-free solution and in standard aCSF, we imaged intracellular pH in pyramidal neurons electroporated with 8-hydroxypyrene-1,3,6-trisulfonic acid (HPTS) and ATTO-594 for semiquantitative pH detection and volume correction, respectively (SI Appendix, Fig. S4A and Materials and Methods). During SDs, we saw reductions in pH in some but not all neurons, while acidified aCSF (pH 6.8; see Materials and Methods) consistently reduced the pH-signal by 14% [6%, 18%] (n = 10 slices, 11 cells, 5 mice; SI Appendix, Fig. S4). Importantly, while in Ca²⁺-free aCSF the ATP signal drop was largely reduced (SI Appendix, Fig. S3 C and D), there were no differences in the pH-transients between Ca²⁺-free and standard aCSF during SDs (SI Appendix, Fig. S4D), suggesting that pH changes are not the cause of the observed ATP sensor signals.

SDs Deplete Neuronal ATP during OGD.

SDs were shown to occur in conditions of primarily reduced energy supply but could as well reduce [ATP]_i by increasing energy consumption (20, 40, 41). This led us to investigate the effect of SDs on ATP signals in the context of cerebral ischemia, here OGD.

OGD was induced by wash-in of aCSF, wherein glucose was replaced by sucrose and oxygen displaced by nitrogen. Recordings of partial tissue oxygen pressure (pO₂) during wash-in of the OGD-modified aCSF confirmed a critical reduction of the pO₂ below the hypoxia threshold (~8 to 10 mmHg, Fig. 2A) within approximately 1 min (n = 5 experiments), a pO₂ level, at which oxidative phosphorylation was shown to break down (42).

Fig. 2.

SDs deplete neuronal ATP during oxygen-glucose deprivation. (A) Clark-electrode recording of pO₂ 50 µm below the slice surface during wash-in of glucose-free and oxygen-depleted aCSF (OGD). Note, that the hypoxia threshold for the breakdown of oxidative metabolism (42) was reached in just over a minute. (Ba) Recording of [K⁺]_o during wash-in of OGD solution, which shows a typical small early increase in [K⁺]_o (Δ[K⁺]_o pre SD) followed by a plateau. The occurrence of an SD was accompanied by a sharp rise in [K⁺]_o (Δ[K⁺]_o peak). Note the delayed recovery of [K⁺]_o following the SD (cf. Fig. 1Da). (Bb) Summary boxplot of the corresponding [K⁺]_o parameters (n = 18 SDs in 18 slices, 11 mice; ***P < 0.001, Wilcoxon-test & Bonferroni correction). Box plots indicate the median, IQ1, and IQ3, whiskers the extrema; individual data points are indicated by dots. (C) Transmission image (Left) and IOS maps (from Left to Right) captured after the onset of a spontaneously occurring SD in ODG. The black dotted line in the IOS map “8 min of OGD” indicates the wavefront of SD. (D) Color-coded images of background- and gain-corrected ATP signals synchronous to the IOS images (C) of the SD-associated reduction of the ATP signal in OGD. (Ea) LFP- and [K⁺]_o-recording corresponding to the experiment in C and D, showing the prolonged SD (Top) occurring in OGD (cf. Fig. 1Da) as well as a subtle increase in [K⁺]_o early during OGD and the prolonged increase in [K⁺]_o during the SD. (Eb) ATP signals from the ROIs indicated in C and D during the SD in OGD. The vertical bars indicate the onset of SD as detected by the IOS per ROI. Note that the sharp decline in [ATP]_i coincides with the onset of the SD. (Fa) Average of ATeam signal in gray and white matter ROIs (“GM” and “WM,” respectively); synchronized to the SD onset in each ROI (median ± IQR/2; 28 GM ROIs, 7 WM ROIs, 7 slices, 5 mice). Note that SDs started at variable times in OGD (indicated by a dotted line on Top), in this subset of slices after 9.8 min [7.2 min, 12.6 min] (median, IQ1, and IQ3). (Fb) Boxplots summarizing the self-normalized ATeam signals (as shown in Fa) in OGD prior to SDs and 2 and 10 min after SDs. Boxplot shows the median, IQ1, and IQ3, whiskers the extrema; individual data points are indicated by dots (n = 7 slices, 5 mice, n.s.-not significant, *P < 0.05, **P < 0.01, Mann–Whitney U test & Bonferroni correction). (Ga) Single-cell resolution background- and gain-corrected ATeam images showing individual neurons before OGD (Top Left, baseline), following 11 min in OGD (Top Right, “pre SD”), and after SD (Bottom Left and Right). Arrowheads point to neurons that moved out of the focal planes, which were not analyzed. Letters in the Top Right corner (i, ii, iii, iv) mark the timepoint of imaging corresponding to the curves shown in Gb. (Gb) Self-normalized ATP signals of the four neurons labeled in Ga. Note that the ATP signal did not recover from the SD-induced drop. (H) Boxplot summarizing the change in ATP signal in OGD during the 2 min before SD onset (Left), and during the first 2 min after SD onset (Right; n = 9 slices, 6 mice, 45 individual neurons, Wilcoxon test, *P < 0.01). SDs started after 12.8 min [12.3 min, 15.4 min] of OGD.

As an early effect of OGD, [K⁺]_o increased on average by 1.2 mM [0.8 mM, 1.6 mM], (median, IQ1, and IQ3; n = 18 SDs in 18 slices, 11 mice; P < 0.001, Wilcoxon-test & Bonferroni correction; Fig. 2B), typically reaching a plateau (Fig. 2Ba). SDs occurred spontaneously but at varying timepoints in OGD, with a median of 12.9 min [12.0, 16.0] (n = 36 slices, 19 mice, 34 SD-positive slices) and were accompanied by sharp increments in [K⁺]_o by 22.9 mM [13.0 mM, 28.0 mM] (n = 18 SDs in 18 slices, 11 mice; P < 0.001, Wilcoxon-test & Bonferroni correction; Fig. 2 B and Ea) and the typical negative DC-shift (Fig. 2Ea). SDs propagated across the cortex to subcortical gray matter as shown by the IOS, similar to KCl induced SDs (Fig. 2C).

ATP signals decreased gradually in OGD, followed by a substantial, sharp drop upon spontaneous SD onset, which in contrast to well-nourished slices, did not recover (Fig. 2 D and Eb, cf. Fig. 1; see SI Appendix, Fig. S1 C and D for CFP and YFP signals). Before SD onset the ATP sensor signal declined to −5% [−1%, −10%] compared to baseline values before OGD over a time course of 9.8 min [7.2 min, 12.6 min] in regions that eventually underwent SD, i.e., the gray matter (n = 7 slices, 5 mice, Fig. 2 Fa and Fb). Notably, the ATP signal decreased further to −29% [−19%, −32%] and −36% [−33%, −39%] at two and 10 min after SD onset. The time points were chosen to capture the prompt SD-induced decline in neuronal ATP signals as well as the absence of acute recovery. After SD, ATP signals were significantly lower in ROIs that underwent SDs compared with those that did not, i.e., in the white matter (n = 7 slices, 5 mice, P > 0.05, Mann–Whitney U test, Fig. 2 Fa and Fb). Single-cell imaging confirmed a sharp drop in neuronal ATP signals without recovery following SD (Fig. 2 Ga–H). Increasing CFP fluorescence coupled with decreasing YFP fluorescence underlies reduced YFP/CFP ratios (SI Appendix, Fig. S1D), signal changes that are absent in cells that do not express the ATP sensor (SI Appendix, Fig. S1E) or in slices from a wild type mouse (SI Appendix, Fig. S1F). The ATP signal dropped by 3% [2%, 6%] in the 2 min prior to SD compared to a drop of 14% [10%, 38%] in the first 2 min after SD (Fig. 2H, n = 9 slices, 6 mice, 45 cells in total, P < 0.01, Wilcoxon-test). The single-cell data confirm that ATP signals did not recover in OGD.

SD-Induced ATP Depletion under OGD Is Reversible.

We next investigated whether neurons are able to restore [ATP]_i following the significant drop after SDs in OGD, or whether the ATP signal reduction following a period of severe glucose deprivation and hypoxia (<10 mmHg), which led to SDs, indicates the terminal breakdown of neuronal energy metabolism. For this purpose, we once more acquired ATP signals during a period of OGD (Fig. 3A), but this time started to reperfuse oxygen- and glucose-containing aCSF upon real-time detection of SDs in the electrophysiological recordings (Fig. 3Ba).

Fig. 3.

Reversibility of SD-induced ATP depletion during oxygen-glucose deprivation. (A) From left to right: Background- and gain-corrected ATeam maps before OGD (baseline), 2 min post SD in OGD, 15 min following resupply of oxygen and glucose (“+OG”) and 15 min after wash-in of OGD together with sodium azide (inhibitor of complex IV). White dotted circles mark three neocortical ROIs (CTX1-3), one subcortical (SUBC), and one white matter (WM) ROI, analyzed in Bb. (Ba) LFP- and [K⁺]_o-recording corresponding to A, confirming a spontaneously occurring SD in OGD. Note the sharp decline in [K⁺]_o upon resupply of OG, which slightly undershoots baseline levels. Vertical line and “#” denote timepoint 10 min after the [K⁺]_o peak during SD, which was used for quantitative analyses shown in Inset. Brief, spike-like transients (for example those next to #) are typical recording artifacts. (Inset) Boxplot of [K⁺]_o (in mM) 10 min after its peak during SD in persistent OGD (left bar; data derived from experiments for Fig. 2 Ba, Bb, and Ea) and for resupply of OG (right bar, n = 9 slices each, 5 & 6 mice, **P < 0.01, Mann–Whitney U test). (Bb) ATP signals from the ROIs indicated in A. The vertical bars indicate the onset of SD as detected by the IOS per ROI. Note that ATP signals recovered upon resupply of OG in all ROIs. Also note, that the addition of sodium azide during OGD lowered ATP signals in the WM similar to the ROIs from the GM. (Ca) Average of transients in the ATeam signal separated into ROIs from GM and WM synchronized to the SD onset (median ± IQR/2; 28 GM ROIs, 7 WM, 7 slices, 5 mice). SD started 14 min [12.8 min, 15.8 min] after OGD in this subset of slices. (Cb) Boxplots of the self-normalized ATeam signal (as shown in Ca) quantifying the minimum after SD in OGD (“SD_min”), 15 min after resupply of oxygen and glucose (+OG), and in OGD together with sodium azide (SD_min and +OG; n = 7 slices, 5 mice, *P < 0.05, Wilcoxon test & Bonferroni correction; OGD + azide, n = 5 slices, 4 mice, Mann–Whitney U test, P = 0.63, n.s.- not significant). (Da) From top to bottom: Background- and gain-corrected ATeam maps showing individual neurons before OGD, 12 min in OGD (immediately before SD onset), during the SD (SD_min), and 15 min after resupply of OG (+OG). Letters in the Top Right corner (i, ii, iii, iv) mark the timepoint of imaging corresponding to the curves shown in Db. (Db) Self-normalized transients of the ATeam signal of the three neurons labeled in Da. (E) Average of ATeam transients of neurons that showed partial ATP signal recovery upon resupply of OG (median ± IQR/2, n = 11 slices, 8 mice; out of 60 neurons, 39 showed a recovery). Data are synchronized to the start of +OG. In this subset of experiments, SDs started 15.3 min [12.2 min, 18.4 min] after OGD.

These experiments resembled those shown in Fig. 2 up to the point at which oxygen and glucose were resupplied (“+OG”), both at large-scale and single-cell imaging (Fig. 3 A–C and D–E, respectively). Resupply of oxygen and glucose led to a return of the negative shift in the DC potential (Fig. 3Ba) and a full restoration of [K⁺]_o to baseline levels with a transient undershoot (Fig. 3Ba). Ten minutes after the SD-associated peak of [K⁺]_o, [K⁺]_o had returned to baseline levels or even below (4.5 mM [3.5 mM, 5.9 mM]) in slices with renewed supply of oxygen and glucose, whereas [K⁺]_o remained elevated during persistent OGD (7.2 mM [6.9 mM, 7.8 mM]; Fig. 3 Ba, Inset; n = 9 slices from 5 & 6 mice, P < 0.01, Mann–Whitney U test).

Notably, resupply of oxygen and glucose induced a prompt recovery of ATP signals (Fig. 3 A and Bb). Median ATP signals increased significantly toward baseline levels (baseline corresponds to 0% change), namely from −13% [−11%, −15%] to −4% [−2%, −5%] and from −32% [−26%, −38%] to −10% [−6%, −16%] in white matter ROIs and gray matter ROIs, respectively, following the resupply of oxygen and glucose for 15 min (Fig. 3C, n = 7 slices, 5 mice, P < 0.05; Wilcoxon test & Bonferroni correction). To test whether the differences in ATP signals in SD-affected areas compared to unaffected areas were due to different basal ATP levels (43), we applied sodium azide [inhibiting complex IV and increasing permeability of TRPV4 channels (44)] in addition to OGD. Sodium azide has been shown to promptly deplete neuronal ATP in gray matter as well as white matter tracts (35). Consistently, we saw a sharp reduction in ATP signals to similar levels independent of brain regions upon wash-in of sodium azide (Fig. 3 Bb and Cb; Mann–Whitney U test, P = 0.63, n = 5 slices, 4 mice), indicating a similar range of the ATP sensor signal and basal [ATP]_i in all ROIs. Last, we investigated recovery of ATP signals following SDs at the cellular level and found essentially restored ATP signals in about two-thirds of the neurons (Fig. 3 D and E). Taken together, brain tissue affected by SDs undergoes a critical, yet in principle reversible, reduction in neuronal [ATP]_i during OGD.

Discussion

SDs have been suggested to contribute to developing brain injury in various brain pathologies, particularly when involving a mismatch in the supply and demand of energy (20). Therefore, we investigated the effect of SDs on neuronal ATP signals. We provide evidence that SDs increase neuronal ATP consumption beyond production, exhausting neuronal [ATP]_i during OGD, which could be restored by renewed supply of oxygen and glucose upon detection of SDs in most cells.

Increased ATP Consumption and Production during SDs.

There is evidence that both ATP consumption and ATP production increase following SDs, the former via stimulation of Na⁺/K⁺-ATPase activity due to rising [K⁺]_o and [Na⁺]_i (25–27) and the latter via stimulation of glycolysis and oxidative phosphorylation, as concluded from measurements of reduced brain glucose concentrations as well as increased CMRO₂ (29–31, 45, 46). Here, ATP signals were transiently reduced following SDs in presence of oxygen and glucose (Fig. 1). Proper calibration of the fluorescent ATP nanosensor is challenging due to technical issues, including membrane permeabilization as well as the requirement to block ATP-consuming as well as ATP-producing processes (for a detailed discussion see ref. 43). Nevertheless, to estimate the drop of [ATP]_i, we assumed a basal [ATP]_i level of 2 mM (47) and calculated [ATP]_i based on published K_D values and Hill coefficients of ATeam1.03^YEMK, first as a purified protein at 37 °C [K_D: 1.2, Hill coefficient: 2.1 (34)], and second, in hippocampal neurons at room temperature in situ [K_D: 2.6, Hill coefficient: 1.0 (47)] (SI Appendix, Fig. S5). Differences in apparent K_D values and Hill coefficients (in situ vs. purified protein) were suggested to result from a contribution of the cytosolic environment to the sensor properties rather than changes in temperature as ATeam1.03^YEMK was shown to be less temperature sensitive in situ (48). Interestingly, using either calibration, [ATP]_i dropped to a similar minimum of ~1.5 mM. To cover a wider range of reported basal [ATP]_i levels (47, 49, 50), we repeated the calculations using the in situ calibration curve (SI Appendix, Fig. S5A) for basal [ATP]_i of 1 and 3 mM, which yielded reductions of [ATP]_i to 0.8 and 2.0 mM, respectively (SI Appendix, Fig. S5D). These data suggest that Na⁺/K⁺-ATPases and Ca²⁺-ATPases would still obtain sufficient ATP to enable ion transport even at minimal [ATP]_i values, as these ATPases bind ATP and the respective ions with K_D values in the low micromolar range (51–53). It is therefore rather unlikely that the SD-induced transient drop in [ATP]_i slowed down ion transport in the presence of oxygenated aCSF. Nevertheless, we can conclude a transient excess of ATP consumption over production in the course of SDs.

Removal of [Ca²⁺]_o but not inhibition of the Na⁺/K⁺-ATPase reduced the drop in ATP signal due to SDs. This is in line with a shift in ATP consumption to Ca²⁺-ATPases as rising [Na⁺]_i during SDs would no longer be pumped out by the Na⁺/K⁺-ATPase but by the Na⁺/Ca⁺-exchanger operating in reverse mode, i.e., exchanging [Ca²⁺]_o for [Na⁺]_i (24). This in turn, has been shown to cause secondary elevation of neuronal [Ca²⁺]_i in acute brain slices during chemical ischemia, for middle cerebral artery occlusion in mice in vivo (24) and in simulations of SDs subsequently requiring increased activity of Ca²⁺-ATPases (23). ATP signals did not drop during SDs in Ca²⁺-free aCSF despite Na⁺/K⁺-ATPase activity, as indicated e.g., by the “undershoot in [K⁺]_o” (54, 55) (SI Appendix, Fig. S3A and see below). This raises the hypothesis that it is not the workload for the Na⁺/K⁺-ATPase per se that leads to the transient ATP signal reduction. Somatic [Ca²⁺]_i has been reported to rise to ~25 µM during SDs (56), which was shown to depolarize mitochondria during SDs (57), a phenomenon also described for epileptic seizures (58). This could in turn reverse the ATP-synthase (59), thus transiently block mitochondrial ATP-synthesis. Overall, our data suggest an underlying role of Ca²⁺ rather than increased Na⁺/K⁺-ATPase activity per se leading to a mismatch of ATP consumption and production during SDs.

Transient reductions in [ATP]_i upon putatively increased consumption during SDs could also be explained by either delayed or exhausted adaptation of ATP production. It remains unclear whether ATP production is fully exploited in the course of SDs, however, there is evidence that adaptation of ATP production is indeed delayed. Previous experiments have shown that peak CMRO₂, indicating stimulation of oxidative phosphorylation, follows SD onset by ~30 s (31). In addition, in primary neuronal cultures, ATP production by glycolysis and oxidative phosphorylation was coupled to [Na⁺]_i-dependent stimulation of Na⁺/K⁺- ATPase activity and similarly delayed by several seconds (60). In hippocampal slices, neuronal activation, accompanied by a Na⁺ influx (47), has been shown to induce a prolonged yet delayed increase in neuronal glycolysis (61). In addition, reversible fragmentation of mitochondria, as observed during SDs in otherwise healthy animals in vivo, could delay ATP production and thus contribute to a transient neuronal [ATP]_i reduction (62).

Previous studies demonstrated high sensitivity of ATeam1.03^YEMK to [ATP]_i over ATP-independent effects in various conditions. In particular, following the application of N-methyl-D-aspartate, glutamate, as well as increased [K⁺]_o, a mutated version of the ATP sensor with a dysfunctional ATP-binding site showed little or no signal changes compared to the functional sensor (43, 63), indicating that ATP sensor signals here and in previous studies reflect an actual decrease in [ATP]_i rather than an unspecific effect on the fluorophores of the ATP sensor.

[K⁺]_o recordings have provided evidence for increased ATP consumption due to increased Na⁺/K⁺-ATPase activity during SDs (31, 64) and were confirmed here: first, by the rapid return of [K⁺]_o from its peak to baseline levels, and second, by the reduction in [K⁺]_o below baseline levels for several minutes after SD, also called “undershoot” in [K⁺]_o (54, 55). This undershoot can be recorded in acute brain slices when electrodes are inserted at a sufficient depth to not strictly record concentrations of K⁺ in the perfused aCSF. Notably, ATP signals began to recover despite ongoing, increased consumption by the Na⁺/K⁺-ATPase, indicating that ATP production rapidly adapts to and even increases beyond consumption. To allow gradual recovery of [ATP]_i, Na⁺/K⁺-ATPase activity, while shown to activate glycolysis and oxidative phosphorylation (60, 65), could in addition be regulated to operate, i.e., consume ATP, at rates slightly below ATP production. Consistently, experiments on inside-out vesicles indicated an inhibitory feedback-loop on Na⁺/K⁺-ATPase activity, showing reduced Na⁺ transport by the Na⁺/K⁺-ATPase when adenosine diphosphate (ADP) levels were increased (66). ADP levels are expected to increase due to ATP-hydrolysis resulting from aforementioned stimulation of Na⁺/K⁺-ATPase activity in the course of SDs.

At the end of SDs, marked by normalization of the DC potential, neuronal ATP signals were still reduced compared to baseline, but gradually increased further thereafter. A similar time course has been observed when overall brain ATP content was measured spectrophotometrically in lysates of brain samples collected at different timepoints after SDs from rats (32), suggesting that these measurements were in principle representative of neuronal [ATP]_i dynamics in presence of glucose and oxygen. However, not all studies measuring ATP in tissue samples found reduced ATP after SDs (reviewed in ref. 40). We thus provide further and first cell-specific evidence for reductions in neuronal [ATP]_i in the course of SDs.

SDs Exhaust Neuronal ATP in Affected Gray Matter.

Given that SDs were observed in conditions of metabolic compromise while simultaneously increasing ATP consumption (41), we monitored ATP signals during OGD, focusing particularly on the effect of SDs.

Remarkably, SDs occurred during OGD when neurons still contained considerable amounts of [ATP]_i, which, in some cases hardly differed from pre-OGD levels (Fig. 3Db). However, a distinct drop in ATP signals followed the SD onset. Thus, SDs were probably not the consequence of widespread depletion of neuronal [ATP]_i, although we cannot exclude prior depletion of [ATP]_i at the very focal SD initiation site. [ATP]_i depletion during continuous OGD was terminal. Consistently, in rats with middle cerebral artery occlusion, ATP remained lower in tissue samples of the penumbra after spontaneous SDs than in tissue samples from nonischemic animals after induced SDs (33). Prevention of SDs could protect a majority of neurons from both the loss of residual [ATP]_i as well as from the ATP demand necessary to fully restore transmembrane ion gradients. Based on our data, brain regions, which are more resistant to SDs or inherently do not undergo SD, such as white matter, are expected to retain ATP longer during OGD.

During OGD or ischemia in vivo, intracellular neuronal pH has been shown to decrease to 6.7 to 7.0 (67–70). In this pH range, the signal of the ATP sensor ATeam1.03^YEMK, i.e., the YFP/CFP ratio, is partially pH-sensitive, however, predominantly for ATP concentrations above 1 mM. Below 1 mM, ATP signals were mostly pH-insensitive or would even increase slightly with acidification (39), i.e., opposite to what we observed. Neuronal [ATP]_i was estimated to be 1.4 to 2.8 mM (49, 50, 71). For [ATP]_i in this range, intracellular acidification to pH 6.7 could explain decreased ATP signals by 10 to 15%, but not by 70%, as seen in individual neurons (Fig. 2Gb). Importantly, actual drops in [ATP]_i would shift the ATP sensor into the pH-insensitive range (39). Furthermore, intracellular acidification during ischemia or OGD was shown to start immediately (67–70, 72), whereas we detected the major drop in ATP in association with SDs, on average ~13 min in OGD. Importantly, recovery of [ATP]_i due to the resupply of oxygen and glucose (Fig. 3), i.e., increasing ATP signals from low [ATP]_i cannot be explained by pH-normalization as the sensor would not respond to changes in pH but to changes in [ATP]_i for [ATP]_i of 0.5 mM or lower. SDs, in presence of oxygen and glucose, have also been reported to cause a transient intracellular neuronal acidification (57). However, these measurements were not quantitative, which prevents an estimation of the contribution of changes in pH to the observed ATP signal based on pH-dependent calibration curves of purified ATeam1.03^YEMK (39). Notably, the changes in pH were limited to the period of depolarization (57), whereas we detected a reduction in ATP signals beyond the SD duration. Additionally, intracellular pH-imaging showed no differences between SDs induced in Ca²⁺-free vs. standard aCSF (SI Appendix, Fig. S4), indicating that observed differences in ATP signals between these two conditions cannot be attributed to changes in pH. This applies both to the drop in ATP signal in standard aCSF and its absence in Ca²⁺-free aCSF. Consequently, we conclude that SDs induce a transient reduction in [ATP]_i. Furthermore, based on quantitative [ATP]_i calculations we concluded that the [ATP]_i sink during SDs in presence of oxygen and glucose would not reach levels low enough to arrest Na⁺/K⁺- and Ca²⁺-ATPases [(51–53) and see above]. Since the [ATP]_i drop could here be overestimated by a possible acidification, this conclusion is robust despite possible pH influences. In summary, although we cannot exclude that some changes in ATP-sensor signal resulted from changes in pH, the ATP signal kinetics, the differences in standard and Ca²⁺-free aCSF, the magnitude (changes of up to 70%), and the recovery of ATP signals could not be dominated by changes in pH, suggesting that observed ATP sensor signals indeed report changes in [ATP]_i.

In contrast to OGD, neuronal ATP signals decreased directly and sharply upon pharmacological inhibition of oxidative phosphorylation by sodium azide (Fig. 3), which when applied in glucose-free aCSF further containing 2-deoxyglucose to arrest glycolysis is referred to as “chemical ischemia.” Induction of chemical ischemia causes an inward Na⁺-current via TRPV4 channels within seconds (44, 71). In contrast, such an inward current was not recorded during the first 4 min of hypoxia (95%N₂/5%CO₂) in brain slices (73). Furthermore, inhibition of TRPV4 channels reduced the drop in [ATP]_i upon chemical ischemia, confirming the link of TRPV4 activation and ATP consumption, which turns out to be a major difference between chemical ischemia and OGD. Notably, a mutated version of the ATeam sensor (AT1.03^R122K/R126K) previously remained insensitive to the application of sodium azide (43, 63).

Reversibility of Neuronal ATP-Depletion Induced by SDs during OGD.

To mimic reperfusion therapy in acute stroke patients, we resupplied oxygen and glucose upon detection of SD. Remarkably, ATP signals essentially recovered in all of the investigated large-scale ROIs and in about two-thirds of neurons imaged at single-cell resolution (Fig. 3). In vivo, changes in [ATP]_i may follow different kinetics, e.g., due to blood flow responses, which can be impaired after stroke and thus limit the supply of oxygen and glucose during reperfusion (18). Conversely, the inevitable trauma of slicing may have contributed to the absence of recovery of ATP signals in about one third of the neurons in addition to the transient OGD. Nevertheless, our data reveal that even after an anoxic SD, which would have been terminal without resupply of oxygen and glucose, mitochondria are functional or rapidly regain function. This opens up the question, which substrates fuel the renewed ATP synthesis, specifically whether extracellular lactate becomes a substrate of oxidative phosphorylation by neuronal uptake and conversion to pyruvate or whether accumulated intracellular lactate and renewed glycolysis play the predominant role. Lactate was shown to accumulate in the brain parenchyma after various conditions: following SDs and cerebral ischemia models in vivo and in vitro (45, 46, 74), in patients following ischemic stroke (75), and, specifically in response to SDs, after TBI, SAH, and cerebral hematoma (76). Thus, lactate could in principle be used for oxidative phosphorylation if transported into neurons. Overall, the role of extracellular lactate as a substrate of neuronal oxidative phosphorylation remains debated (77–80). Direct measurements of [ATP]_i, as shown semiquantitatively in this study, could be used to identify cell types and substrates that allow neuronal recovery of [ATP]_i after SDs in transient OGD, with potential implications for stroke therapy.

In summary, we show that SDs are accompanied by a substantial increase in neuronal ATP consumption beyond production, which may not be primarily driven by increased Na⁺/K⁺-ATPase activity, and suggest a yet incompletely understood key role of Ca²⁺. Under conditions that mimic stalled blood supply, SDs exhaust neuronal [ATP]_i whereas before SD onset, ATP signals decreased markedly slower. Thus, our findings indicate that the sequence of events leading to neuronal [ATP]_i depletion in severely ischemic tissue, corresponding to the ischemic core, is essentially the same as in the penumbra with partially impaired blood flow. Specifically, [ATP]_i depletion follows SDs. Still, most cells were capable of restoring [ATP]_i if oxygen and glucose were rapidly resupplied following SD. Our findings support therapeutic strategies to prevent SDs after cerebral ischemia as this could prevent not only the acute loss of [ATP]_i but also the remaining need of energy to fully restore ion gradients. Resolving the kinetics of [ATP]_i in different neuronal compartments, cell types and brain regions during SDs, ultimately in vivo, are pending topics for future studies, likewise, addressing the potency of therapeutics to preserve neuronal [ATP]_i by preventing SDs and how metabolic status affects [ATP]_i depletion as well as recovery upon the renewed supply of oxygen and glucose.

Materials and Methods

Animals.

Animals were housed and bred in accordance with the German Animal Welfare Act, the European Communities Council Directive (2010/63/EU), as well as Leipzig University guidelines and with approval of the local authorities (T09/20, T05/21-MEZ). Animals had free access to food and water and were kept in a 12 h/12 h light–dark cycle. The study includes six adult wildtype C57Bl6N mice (4 males, 2 females, median age: 10 wk) and 36 adult transgenic mice (14 females and 22 males, median age 22 wk) expressing the ATP sensor ATeam 1.03^YEMK (34) under the neuronal promotor Thy1.2 on the background of the C57BL/6J mouse strain (B6-Tg(Thy1.2-ATeam1.03^YEMK)AJhi; MGI:5882597; referred to as ThyAT) (35), which were originally generated by J. Hirrlinger’s group.

Preparation.

Animals were anesthetized by isoflurane and decapitated. Brains were removed and cut into coronal slices (thickness 300 µm, −1.5 mm to 2 mm relative to bregma) on a vibratome (Leica VT1200S, Leica Biosystems, Nussloch, Germany) in ice-cooled aCSF, then transferred to warmed aCSF (35 °C) for 30 min and kept at room temperature (20 to 22 °C) at which experiments were performed.

Solutions and Drugs.

Drugs and chemicals were purchased from Sigma Aldrich. ACSF contained in mM: 125 NaCl, 2.5 or 5 KCl, 1.25 NaH₂PO₄, 26 NaHCO₃, 1.8 MgCl₂, 2 CaCl₂, and 20 glucose and was continuously carbogenated (95% O₂, 5% CO₂, pH 7.4, osmolarity ~305 mOsm). For OGD, glucose was replaced by equimolar saccharose and gassed with a mixture of 95% nitrogen (N₂) and 5% CO₂ (pH 7.4, osmolarity ~305 mOsm). Complex IV of the respiratory chain was inhibited by sodium azide (5 mM) in a subset of experiments. Ca²⁺-free contained no CaCl₂ but was otherwise unaltered.

Electrophysiological Recordings.

For electrophysiological recordings and parallel imaging (see below), slices were transferred to an open bath perfusion chamber (Warner Instruments, Holliston, MA, perfusion rate 6 to 8 mL/min). Slices were held in place with a U-shaped platinum wire covered with a nylon grid. DC LFP were recorded ~50 µm below the surface of the slice in layer 2 to 3 of somatosensory and -motor areas using microelectrodes prepared from borosilicate glass (Science Products, Hofheim, Germany; pipette puller PC-10, Narishige, Tokyo, Japan) or from double-barrel theta glass (Fa. H. Kugelstätter, Garching, Germany) to measure the LFP in parallel with K⁺-sensitive Nernst potentials. The potassium-selective barrel was filled with a solution containing 150 mM KCl and Potassium-Ionophore 1 (Valinomycin) at the tip. The LFP and the Nernst potential were amplified (EXT-10 and ION-01 for LFP or LFP and [K⁺]_o, respectively), low-pass filtered (3 kHz, LPBF-01GX, all npi electronic, Tamm, Germany), and digitized (CED-1401 Micro3 with Spike2 9.01 software, Cambridge Electronic Design, Cambridge). To confirm oxygen deprivation, pO₂ was recorded using Clark-style glass oxygen microelectrodes (10 µm tip; Unisense, Aarhus, Denmark) (81) during wash-in of OGD solution in a subset of experiments.

Induction of SD.

SDs were either triggered focally by release of 3 M KCl from glass microelectrodes (2 to 3 MOhm when filled with ACSF) by pressurized air (0.5 bar, 1 s, PDES-2L, npi electronic, at >500 µm from the recording site) or occurred spontaneously during OGD. When perfusing OGD-conditioned aCSF containing 2.5 mM K⁺, SDs occurred in 33% of the brain slices after 12.8 to 18.0 min (n = 15, 5 mice), a low incidence compared to in vivo models of cerebral ischemia, in which most animals developed SDs (see e.g., refs. 6 and 14). To mimic extracellular accumulation of K⁺ as shown in the course of cerebral ischemia in the severely ischemic core but also in the penumbra (see e.g., refs. 82 and 83), we raised K⁺ to 5 mM in the aCSF, an approach which has been implemented by others (84) and counterbalances constant dilution of cellular K⁺ efflux by the perfused aCSF. Importantly, concentration of 5 mM [K⁺]_o is well below the threshold that would induce SDs on its own, i.e., ~15 mM [K⁺]_o (36, 64, 85). Raising K⁺ to 5 mM increased the incidence of SDs to 94% (n = 31 slices, 16 mice).

IOS and FRET.

Images were acquired by two-photon laser scanning microscopy using a Ti:Sapphire laser (810 nm, Mai Tai DeepSee, Spectra-Physics, Milpitas, CA), as previously described for imaging ATeam1.03^YEMK (48) resulting in CFP excitation >60% and YFP excitation <5% of their respective maximal values (https://public.brain.mpg.de/shiny/apps/SpectraViewer/). The laser scanning microscope (Olympus Fluoview 10M, Olympus, Tokyo, Japan) was equipped with a 3.32× (NA:0.14) or 25× objective (XLPlan N 25× W NA:1.05, both Olympus). We recorded time series of vertical stacks (image dimension 320 × 320 to 640 × 640 pixels, xy pixel dimension: 0.2 to 12 µm), up to 20 z-steps (1.5 to 2 µm at cellular resolution and 5 µm/image plane for large scale imaging at lower magnification) and time intervals of 15 to 40 s during SDs and up to 2 min for extended baseline measurements. Slices were first inspected by bright field microscopy to avoid imaging of ROIs containing swollen neurons, an early sign of cell damage (86). Imaging was performed 33 µm to 56 µm (median of upper and lower boundary) below the slice surface and neuronal somata were close to the lower boundary at baseline to anticipate some vertical movement in case of overall slice swelling due to SD or OGD.

To detect onset and propagation of SDs, we measured light transmission, which is affected by refractive tissue properties and shows characteristic changes during SDs and is also known as the IOS (37). The onset of SDs was shown to coincide with a rapid change in IOS (57), which was used to detect SD onset.

To measure neuronal ATP, we used the FRET-based sensor ATeam 1.03^YEMK consisting of a donor domain (CFP) and an acceptor domain (YFP) as well as a bacterial ε-subunit of the F₀F₁-ATP synthase, which contains the ATP-binding site (34). The distance and orientation of the donor and the acceptor fluorophores and, therefore, the subsequent FRET depend on the conformation of the ε-subunit of the F₀F₁-ATP synthase, which is dependent on the ATP concentration (K_D = 1.2 mM) (34). The ratio of the background-corrected signal from CFP and YFP was the semiquantitative readout for the intracellular ATP ([ATP]_i). Thus, percentual signal changes are not strictly equivalent to changes in [ATP]_i.

Fluorescence was low pass filtered (690 nm) and split by a dichroic mirror (D510 nm) to capture CFP and YFP fluorescence separately. CFP emission was further bandpass filtered (median wavelength 480 nm/bandwidth 40 nm, F47-480SG, AHF Analysentechnik AG, Tübingen, Germany). The linear electronic gain was set to 1.0 in the CFP channel and varied between 1.0 and 2.0 in the YFP channel depending on imaging modalities.

Experimental Protocols.

All experiments began with baseline imaging (5 to 10 stacks, duration of at least 2.5 min) before KCl-induction of SDs or OGD. We investigated SDs in three conditions: 1) KCl-induced SDs in presence of oxygen and glucose, 2) SDs occurring during continuous OGD, and 3) SDs occurring during OGD with resupply of oxygen and glucose following visual SD detection on LFP and [K⁺]_o-recordings. Autofluorescence was measured in four slices from one wildtype mouse during KCl-induced SDs.

Data Analysis.

All data were analyzed in Igor Pro 8 (WaveMetrics, Inc., Lake Oswego, OR) and Matlab (The MathWorks, Natick, MA, version R2022a). SDs were visually identified by their characteristic signatures in IOS (post hoc analysis) as well as LFP and [K⁺]_o recordings (post hoc and live analysis), respectively. The IOS during SDs was compared to the baseline IOS (prior to KCl-induced SDs or OGD) by calculating the percentage change.

Nernst potentials recorded with K⁺-sensitive electrodes were converted to [K⁺]_o in mM using the Nernst equation and assuming baseline [K⁺]_o to be equivalent to the K⁺ concentration in the aCSF (81).

The ratio of YFP and CFP fluorescence (YFP/CFP) emitted by the ATP-sensor was used as a semiquantitative parameter for dynamic ATP analysis. As a preprocessing step background signals were subtracted, e.g., derived from sensor-free extracellular compartments in high magnification images or from lateral ventricles in low magnification images. Further, z-planes were compressed into maximum intensity z-projections. Movement in x and y direction was corrected using a registration algorithm based on Fast Fourier Transformation (87). To extract temporal ATP profiles, ROIs of individual cells, cortical and subcortical gray matter and white matter were manually selected. ATP profiles were self-normalized to their baseline average prior to OGD or SD induction, resulting in relative changes in the YFP/CFP ratio (∆%). To ensure comparability of the absolute ratios shown, the electronic gain was corrected to 1.0 when >1.0.

Intracellular pH Imaging.

Acute coronal brain slices were prepared from wildtype C57Bl6N mice as described above for ThyAT mice. Neocortical pyramidal neurons in layer 2 to 5 were electroporated with the pH-indicator HPTS (200 µM, “green”) (88) and ATTO-594 (200 µM, “red”) for volume correction using a glass micropipette and a patch-amplifier [EPC10/2, HEKA, Pfalz, Germany; −4 V, 1 ms, 200 Hz for 2 s, −0.7 V offset current (89)]. Dyes were dissolved in an intracellular solution containing in mM: 150 K-gluconate, 10 NaCl, 3 Mg-ATP, 0.3 GTP, 10 4-(2-Hydroxyethyl)piperazine-1-ethanesulfonic acid. Neurons were imaged during SDs induced by KCl in either standard aCSF or Ca²⁺-free, otherwise equivalent, aCSF (incubated for ~10 min) and during wash-in of aCSF acidified by propionic acid (90). Image analysis was performed on z-projections and involved a linear bleach correction, subtraction of extracellular background, and calculation of green over red ratios to compensate for changes in cell volume during SDs as previously published (57).

Calculation of ATP Concentrations Based on ATeam1.03^YEMK Signal.

Calibration curves to calculate actual concentrations of [ATP]_i were derived from published K_D values and Hill coefficients (H) of ATeam1.03^YEMK, first as a purified protein (34) and second as expressed in neurons of organotypic hippocampal slice cultures in situ (47). The ATP-bound sensor fraction (B) at baseline was estimated by the following equation (eq1): B = [ATP]^H/([ATP]^H + K_D^H) (34, 43) and assuming basal neuronal [ATP]_i of 1 to 3 mM (47, 49, 50). As a second assumption, we set B = 0 for recorded ATeam1.03^YEMK signals following SDs in OGD, i.e., assumed [ATP]_i to be 0 mM at this point. Based on these assumptions, the median of the actual ATeam signal recording of KCl-induced SDs (Fig. 1F) was transformed to B-values (SI Appendix, Fig. S5B). Subsequently, eq1 was solved for “[ATP]” to obtain dynamic [ATP]_i in millimolar units.

Data Reporting and Statistical Analysis.

This is an exploratory study. Due to the lack of prior evidence sample sizes could not be accurately estimated and were selected to comply with likewise research. Slices were excluded from analysis if movement artifacts prevented reliable analysis. Neurons that moved out of the vertical stacks during the acquired time series were excluded from analysis. Neurons with baseline ATP sensor signals similar to those obtained from cells depleted of [ATP]_i following SD under OGD, indicating [ATP]_i depletion at baseline, were also excluded.

Data are reported as median and IQ1 (25th percentile) and IQ3 (75th percentile). Boxes in boxplots display median, and IQ1 and IQ3 mark lower and upper box limits. Whiskers extend to minimal and maximal values unless considered as outliers (“+,” values >1.5*IQR). Data analysis was not blinded. Statistical inference was performed by the Wilcoxon test for paired comparisons and the Mann–Whitney U test for independent sample. Multiple comparisons were corrected by the Bonferroni post hoc test. Differences were considered significant at P < 0.05. Data will be made available upon reasonable request.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Data used for statistical analysis are shown in full. Imaging source data will be made available upon request.