Expression and activity-dependent function of Px1 channels in hippocampal astrocytes
Consistent with previous reports, we observed Px1 to be strongly expressed in the hippocampus of juvenile mice (P30) (Fig 1A). To investigate expression of Px1 specifically in astrocytes, we first performed PCR on hippocampal astroglial RNA isolated using the TRAP technique on Aldh1l1:L10a-eGFP mice (see Materials and methods; Fig 1B) and found Px1 mRNA expression in astrocytes at various developmental stages, from postnatal days 10 to 50 (P10, P30, P50) (Fig 1A). Next, quantitative PCR (qPCR) revealed that Px1 is developmentally regulated in astrocytes, as its transcription levels were increased at P50 compared to P10 (n = 3, p = 0.027; Fig 1A). Px1 expression in astrocytes was confirmed by Fluorescent In Situ Hybridisation (FISH), which showed the presence of Px1 mRNAs in both astrocytes (stained with S100β) and neurons (stained with NeuN; n = 3 mice; Fig 1C) in wild-type (+/+) mice. We then investigated whether in +/+ mice Px1 channels from hippocampal CA1 astrocytes and pyramidal neurons were functional and their opening activity-dependent. For that purpose, we performed Ethidium Bromide (EtBr) uptake assay in acute hippocampal slices displaying population activity generated spontaneously in a pro-bursting artificial cerebrospinal fluid (ACSF) (Fig 1D and 1E). EtBr being fluorescent and small enough to travel through Px1 channels, the state of opening of these channels can be inferred from fluorescence intensity. In both neurons and astrocytes, population activity significantly increased EtBr uptake compared to basal conditions (neurons, 137 ± 11%, n = 9 mice, p = 0.0423; astrocytes, 155 ± 7%, n = 7 mice, p = 0.007; Fig 1E and 1F). This effect was inhibited by blocking selectively Px1 channels with 10Panx (400 μM, 15 min) (population activity-induced EtBr uptake normalised to basal ACSF, 10Panx: neurons, 95 ± 6%, n = 9, p = 0.0278; astrocytes, 110 ± 9%, n = 7, p = 0.0318; Fig 1E and 1F), which resulted in similar EtBr uptake compared to basal condition (p = 0.9826 and 0.9103 for neurons and astrocytes, respectively), but not by applying the scramble control peptide scPanx (400 μM, 15 min) (population activity-induced EtBr uptake normalised to basal ACSF, scPanx: neurons, 141 ± 19%, n = 6, p = 0.9933; astrocytes, 171 ± 8%, n = 6, p = 0.7095; Fig 1F), thus pointing to an activity-dependent uptake through neuronal and astroglial Px1 channels. Consistent with these data, we found using electrophysiology that astroglial Px1 channels are open in population activity regime, since their inhibition by the 10Panx peptide in +/+ mice reduced the astrocyte whole-cell conductances recorded at positive potentials, with no change in resting membrane potential or membrane resistance measured at negative membrane potential (S1 Fig; n = 4 cells from 3 mice). Altogether, these data indicate an activity-dependent activation of Px1 channels in astrocytes.
Fig 1. Developmental expression and activity-dependent function of Px1 channels in hippocampal astrocytes.
(A) Scheme illustrating the protocol used for hippocampal astroglial RNA extraction from Aldh1l1:L10a-eGFP mice using TRAP technique. (B) Upper panel: Px1 expression obtained by PCR at different developmental stages (postnatal days 10, 30, and 50) in hippocampal astrocytes (first three lanes) from Aldhl1:L10a-eGFP mice and in total hippocampus (postnatal day 30, last lane). Lower panel: Px1 transcription level normalised to RNA18s quantified by qPCR in P10, P30, and P50 mice in hippocampal astrocytes (n = 3 mice; p = 0.027; Friedman test followed by Dunn posttest). (C) Left: Schematics of the hippocampus showing the CA1 regions of interest from which the representative images are taken. s.p., stratum pyramidale; s.r, stratum radiatum. Right: Representative confocal images of Px1 mRNA detected in the CA1 region of the hippocampus by FISH by RNAscope on brain sections from P20-P30 C57BL6 mice. Neuron nuclei are immunolabelled with NeuN (top images) and astrocytes with S100β (bottom images). Scale bar, 10 μm. (D) Representative EtBr uptake (red) in stratum pyramidale (s.p.) neurons immunolabelled with NeuN (blue) and stratum radiatum (s.r.) astrocytes immunolabeled with S100β (grey) in hippocampal slices. Scale bar, 20 μm. (E) EtBr uptake in basal or population activity conditions without or with 10Panx (400 μM) applied 15 min prior and during EtBr uptake assay. Scale bar, 20 μm. (F) Quantification of activity-dependent neuronal and astroglial EtBr uptake normalised to control conditions in slices from +/+ mice treated or not with 10Panx (neurons, n = 9 mice; astrocytes, n = 7 mice) and scPanx (neurons, n = 6 mice; astrocytes, n = 6 mice; Repeated measures one-way ANOVA). Asterisks indicate statistical significance (*p < 0.05, **p < 0.01). The data underlying this figure can be found in the S1 Metadata A tab.
Astroglial Px1 channels limit sustained neuronal population activity
To date, our understanding of Px1 physiological and pathological relevance in the brain mostly relies on genetic and pharmacological disruption of Px1 functions in all brain cells. Here, to investigate the role of astroglial Px1 channels in neuronal network activity, we used molecular approaches, including transgenic mice and viral vectors targeting Px1 in astrocytes. We first engineered a Px1 conditional mutant mouse lacking Px1 expression in astrocytes. To do so, mice with floxed exon 3 (Panx1tm1c(KOMP)Wtsi conditional allele) were crossed with transgenic mice expressing the Cre recombinase under the promoter of the human glial fibrillary acidic protein (hGFAP-cre mice ; S2A Fig). To investigate the selectivity and efficiency of Px1 deletion in astrocytes from hGFAP-Cre-Px1fl/fl mice, we first performed FISH using a probe for Px1 exon 3, which is expected to be deleted solely in astrocytes. In hGFAP-Cre-Px1fl/fl mice, while this Px1 probe was detected in the hippocampal pyramidal layer at similar levels compared to +/+ mice (43,969 ± 7,163 versus 55,752 ± 9,104 dots/mm2, p = 0.3428, n = 3 and 3 mice for hGFAP-Cre-Px1fl/fl and +/+ mice; S2B Fig, upper panel), it was significantly reduced in astrocytes (3,520 ± 1,120 versus 22,475 ± 6,698 dots/mm2, p = 0.0257, n = 3 and 3 mice for hGFAP-Cre-Px1fl/fl and +/+ mice; S2B Fig, lower panel), thereby indicating astroglial-specific deletion of Px1.
Consistent with this finding, we found in astroglial Px1-deficient mice a loss of Px1 function selectively in astrocytes, but not in neurons, as assessed by activity-dependent uptake. Indeed, the activity-dependent EtBr uptake induced by sustained network activity was inhibited in astrocytes (population activity-induced EtBr uptake normalised to basal ACSF, 125 ± 12%, n = 15 for basal and n = 14 for population activity, p = 0.2079; S2C and S2D Fig), while it persisted in neurons (139 ± 8%, n = 15, p = 0.003; S2C and S2D Fig), to a similar level as that found in +/+ mice (p = 0.8524; S2D Fig). In addition, inhibiting Px1 channels with the 10Panx peptide still significantly decreased EtBr uptake in neurons (106 ± 9%, p = 0.0035, n = 15; S2D Fig), but not in astrocytes, from hGFAP-Cre-Px1fl/fl mice (117 ± 21%, p = 0.3671, n = 14; S2D Fig). In contrast, the control scPanx peptide did not affect activity-dependent EtBr uptake in neurons or astrocytes from hGFAP-Cre-Px1fl/fl mice (neurons, 143 ± 23%, p = 0.8450, n = 8; astrocytes, 128 ± 23%, p = 0.1829, n = 8; S2D Fig). Consistent with these data, Px1 disruption in hGFAP-Cre-Px1fl/fl mice decreased astrocyte conductances similarly to Px1 inhibition by 10Panx peptide in +/+ mice (S1 Fig; n = 5 cells from 3 hGFAP-Cre-Px1fl/fl mice).
Importantly, hGFAP-Cre-Px1fl/fl hippocampi did not present developmental defects. They indeed showed no gross anatomical alterations and presented normal architecture and layered structure, with similar density of CA1 pyramidal cells and astrocytes, as assessed by NeuN and GFAP staining, respectively (S3A and S3B Fig), as well as equivalent stratum pyramidale thickness compared to +/+ mice (S3A–S3C Fig; n = 9 slices from 3 mice for both +/+ and hGFAP-Cre-Px1fl/fl animals).
Using this transgenic mouse, we investigated the role of astroglial Px1 channels in neuronal population activity. To do so, we recorded population activity generated spontaneously in a pro-bursting ACSF in hippocampal slices from +/+ and hGFAP-Cre-Px1fl/fl mice using the Multi-Electrode Array (MEA) technique  (Fig 2A). Hippocampal slices from +/+ mice exhibited spontaneous bursts with an incidence of 6.69 ± 1.01 bursts/min and a duration of 1.43 ± 0.07 s (n = 13 slices from 6 mice; Fig 2B–2D). Strikingly, disrupting astroglial Px1 switched the pattern of spontaneous discharges to paroxysmal events in 77.8% of recorded slices (p < 0.001; Fig 2D; paroxysmal events frequency: 0.70 ± 0.22 events/min; duration: 105.22 ± 19.27 s; n = 14 out of 18 slices from 8 mice; Fig 2E) and increased delta (0.5 to 4 Hz) activity (p < 0.001; Fig 2F). In addition, 41.6% of hGFAP-Cre-Px1fl/fl slices with paroxysmal events displayed occasional interparoxysmal event bursts.
Fig 2. Astroglial Px1 deficiency induces paroxysmal activity.
(A) Top panel, hippocampal slice on a MEA chamber (DG: dentate gyrus). Bottom panel, illustration of bursting activity recorded using a MEA device (200 μm interelectrode distance) in hippocampal slices. Scale bar: 500 ms, 0.3 mV. (B) Representative traces of bursting activity in +/+ mice (upper trace) and paroxysmal activity in hGFAP-Cre-Px1fl/fl mice (lower trace). The corresponding time-frequency plots are shown under the traces. Scale bar: 30 sec, 0.2 mV. (C) Magnification of bursting activity in +/+ mice and paroxysmal activity in hGFAP-Cre-Px1fl/fl mice highlighted by the red rectangles in panel B. Scale bar: 10 s, 0.2 mV. (D) Proportion of bursts and paroxysmal events recorded in +/+ and hGFAP-Cre-Px1fl/fl mice (+/+, n = 13 slices from 6 mice; hGFAP-Cre-Px1fl/fl, n = 18 slices from 8 mice; Fisher exact test). (E) Quantification of bursts and paroxysmal events frequency and duration (+/+, n = 13 slices; hGFAP-Cre-Px1fl/fl, n = 14 slices). (F) Left, power spectral density of activity (normalised to the percent of the total PSD) recorded in +/+ (grey) and hGFAP-Cre-Px1fl/fl (black) slices. The area highlighted by the red rectangle is zoomed in the inset. Right, power spectral density of left panel binned according to different brain rhythms: delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–3 Hz), beta (13–30 Hz), gamma (30–80 Hz), fast (80–200 Hz), and >200 Hz oscillations (+/+, n = 13 slices; hGFAP-Cre-Px1fl/fl, n = 12 slices; p = 0.002, repeated measures two-way ANOVA). Asterisks indicate statistical significance (*p < 0.05; ***p < 0.0001). The data underlying this figure can be found in the S1 Metadata B tab.
Recombination driven by the hGFAP promoter is not necessarily restricted to astrocytes in hGFAP-Cre-Px1fl/fl mice, due to transient expression of GFAP in other cell types during development. We thus tested whether the switch from bursting to paroxysmal activity observed in hGFAP-Cre-Px1fl/fl animals was also present in conditional and inducible Px1 knockout mice (hGFAP-CreERT2-Px1fl/fl), where Px1 deletion is induced postnatally specifically in astrocytes. In these mice, in which tamoxifen (TF)-induced CreERT2 expression leads to recombination in 58.6 ± 2% of GFAP-expressing cells (n = 6 slices from 2 mice; S4A Fig), Px1 expression in neurons assessed by FISH is comparable to that observed in +/+ and hGFAP-Cre-Px1fl/fl mice (44,311 ± 9,455 dots/mm2; p = 0.7192 and p > 0.9999 in comparison with +/+ and hGFAP-Cre-Px1fl/fl mice), while it is strongly reduced in astrocytes (6,739 ± 1,308 dots/mm2; p = 0.0353), to similar levels as in hGFAP-Cre-Px1fl/fl mice (p = 0.9372; n = 3, 3, and 3 for +/+, hGFAP-Cre-Px1fl/fl and hGFAP-CreERT2-Px1fl/fl mice, respectively; S4B Fig). MEA recordings of slices from hGFAP-CreERT2-Px1fl/fl revealed the same pattern of activity as the one observed in hGFAP-Cre-Px1fl/fl mice (p = 0.4569; S4 Fig). Indeed, the majority of slices (62.5%; n = 10 out of 16 slices from 4 mice) displayed paroxysmal activity, while bursting activity was observed in only 37.5% of the slices (n = 6 out of 16 from 4 mice). Furthermore, paroxysmal activity recorded in slices from hGFAP-CreERT2-Px1fl/fl mice displayed similar frequency and duration compared to hGFAP-Cre-Px1fl/fl slices (frequency: 0.39 ± 0.12 /min, p = 0.312; duration: 68.09 ± 12 s, p = 0.156; S1 Table). In contrast, control mice (+/+ and hGFAP-CreERT2 treated with TF) mostly displayed bursting activity (86%, n = 12 out of 14 slices from 3 mice for +/+ + TF; 77%, n = 20 out of 26 slices from 4 mice for hGFAP-CreERT2 + TF; p = 0.0106 and 0.027, respectively; S4C–S4G Fig). These results thus indicate that postnatal deletion of Px1 specifically in astrocytes limits neuronal population activity.
Px1 is expressed in both neurons and astrocytes. To evaluate the potential differential role of neuronal versus astroglial Px1, we then compared the effect of global versus astroglial Px1 disruption on neuronal population activity. To assess the effect of global Px1 inhibition, we used either the 10Panx peptide in +/+ mice (S5A and S5B Fig) or a constitutive Px1−/− mouse (S5C–S5E Fig), in which Px1 is deleted both in neurons and in astrocytes, as assessed by FISH (neurons: 13,395 ± 4,238 dots/mm2; astrocytes: 3,211 ± 1,215 dots/mm2; p = 0.0027 and 0.0057 for neurons and astrocytes, respectively; n = 3 Px1−/− and 3 +/+ mice; S5C Fig). 10Panx peptide, as well as Px1 deficiency in constitutive Px1−/− mice, did not induce paroxysmal activity and had no effect on neuronal bursting pattern compared to control condition in +/+ mice (Control (before 10Panx): frequency, 7.72 ± 2.45 bursts/min; duration, 1.94 ± 0.25 s; 10Panx: frequency, 7.91 ± 2.91 bursts/min, p = 0.9193; duration, 2.74 ± 1.01 s, p = 0.4770, n = 5 slices from 3 mice; S5B Fig; +/+: frequency, 6.68 ± 1.01 bursts/min; duration, 1.43 ± 0.07 s; n = 13 slices from 6 mice; Constitutive Px1−/− mice: frequency, 6.90 ± 0.89 bursts/min, p = 0.87; duration, 1.25 ± 0.10 s; n = 16 slices from 6 mice; p = 0.100; S5E Fig). Altogether, these data indicate that ubiquitous Px1 deletion has no effect on activity pattern and suggest that astroglial Px1 differentially regulate network activity compared to neuronal Px1.
Paroxysmal activity recorded in hippocampal slices from hGFAP-Cre-Px1fl/fl mice could translate in vivo into increased susceptibility to seizures. To investigate this, we performed recordings of electroencephalogram (EEG) after intraperitoneal (IP) injection of the proconvulsant agent pilocarpine and found that hGFAP-Cre-Px1fl/fl mice display increased seizure susceptibility in vivo. Indeed, hGFAP-Cre-Px1fl/fl mice had a shorter first seizure onset delay (+/+: 15.73 ± 1.42 min; hGFAP-Cre-Px1fl/fl: 10.94 ± 1.04 min; n = 16 and 13 mice for +/+ and hGFAP-Cre-Px1fl/fl, respectively; p = 0.0136; Fig 3A and 3B) and a lower survival rate (+/+: 68.75%; hGFAP-Cre-Px1fl/fl 30.77%; n = 16 and 13 mice for +/+ and hGFAP-Cre-Px1fl/fl respectively; p = 0.0426; Fig 3B) compared to +/+ mice.
Fig 3. Astroglial Px1-deficient mice are more susceptible to pilocarpine-induced seizures.
(A) Left, schematic representation of a mouse implanted with wireless ETA-F10 transmitters and EEG electrodes for EEG recordings. Right, representative traces of EEG recordings in +/+ (upper trace) and hGFAP-Cre-Px1fl/fl mice (lower trace). Red arrows indicate pilocarpine injection. Scale bar: 2 min, 500 μV. (B) Quantification of first seizure onset delay and percent survival (+/+, n = 16 mice; hGFAP-Cre-Px1fl/fl, n = 13 mice; Student t test and probability of survival analysis with log-rank (Mantel–Cox) test). Asterisks indicate statistical significance (*p < 0.05). The data underlying this figure can be found in the S1 Metadata C tab.
In all, these data show that astroglial Px1 channels inhibit paroxysmal activity.
Astroglial Px1 channels control excitability of pyramidal cells
How do astroglial Px1 modulate neuronal network pattern? To examine the contribution of single neurons to the altered network activity in hGFAP-Cre-Px1fl/fl mice, we characterised the electrophysiological properties of CA1 pyramidal cells during bursts and paroxysmal events in control and astroglial Px1-deficient mice by performing simultaneous field potential and patch clamp recordings (Fig 4A). While in +/+ mice, neurons displayed low frequency bursts of activity, as observed extracellularly, and action potentials (AP) firing in between bursts, neurons from hGFAP-Cre-Px1fl/fl mice displayed paroxysmal activity and AP firing between seizures (Fig 4B). These results suggest enhanced hippocampal pyramidal neuron excitability in hGFAP-Cre-Px1fl/fl mice.
Fig 4. Pyramidal cells from astroglial Px1-deficient mice are more excitable.
(A) Schematic illustration of simultaneous recording of fEPSP (1) and single neuron (2) in the hippocampus. St Pyr., Stratum Pyramidale; St Rad., Stratum Radiatum. (B) Representative traces of fEPSP (1) and single neuron spontaneous activity (2) in +/+ mice (upper traces) or hGFAP-Cre-Px1fl/fl mice (lower traces). The time-frequency plots corresponding to single neuron recordings are shown under the traces. Scale bar: +/+, 2 s; hGFAP-Cre-Px1fl/fl, 5 s; fEPSP, 0.2 mV; single neuron, 20 mV. (C) Schematic illustration of single neuron recording in CA1 hippocampal area. (D) Representative traces of CA1 pyramidal cell firing pattern from +/+ and hGFAP-Cre-Px1fl/fl mice in basal ACSF containing synaptic blockers (picrotoxin 100 μm, NBQX 10 μM, CPP 10 μM) during depolarisation by 10 pA injection for 500 ms. Cells were held at −60 mV. Scale bar: 50 ms, 10 mV. (E) Number of APs evoked by current injection from 10 pA to 100 pA (+/+, n = 10 neurons from 5 mice; hGFAP-Cre-Px1fl/fl, n = 13 neurons from 5 mice; two-way ANOVA and multiple comparisons test). (F) Quantification of rheobase, time to first spike and membrane potential, (+/+, n = 10 neurons from 5 mice; hGFAP-Cre-Px1fl/fl, n = 14 neurons from 5 mice; unpaired Student t test). Asterisks indicate statistical significance (*p < 0.05, **p < 0.01). The data underlying this figure can be found in the S1 Metadata D tab.
We therefore investigated whether astroglial Px1 alters intrinsic membrane properties and excitability of pyramidal cells in basal conditions using whole-cell patch clamp recordings while blocking synaptic activity (Fig 4C). Neuronal excitability was increased, as shown by the enhanced firing rate of pyramidal cells in response to depolarising current pulses (+/+: n = 10 neurons from 5 mice, hGFAP-Cre-Px1fl/fl: n = 13 neurons from 5 mice; Fig 4D and 4E), and by the reduction of both the rheobase (approximately −40%), i.e., the minimal current necessary to evoke an AP, and the delay to first spike (rheobase, +/+: 28.8 ± 2.3 pA, hGFAP-Cre-Px1fl/fl: 16 ± 4 pA; p = 0.009; first spike delay, +/+: 144.9 ± 31.4 ms, hGFAP-Cre-Px1fl/fl: 58.5 ± 14 ms; p = 0.025; +/+: n = 10 neurons from 5 mice, hGFAP-Cre-Px1fl/fl: n = 14 neurons from 5 mice; Fig 4F). These changes were not due to alterations in pyramidal cell resting membrane potential and membrane resistance in hGFAP-Cre-Px1fl/fl mice (Vm, +/+: −62.9 ± 2.4 mV, n = 10; hGFAP-Cre-Px1fl/fl: −59.9 ± 3.8 mV, n = 14; Fig 4F; Rm, +/+: 223.1 ± 15.1 MΩ, n = 10; hGFAP-Cre-Px1fl/fl: 245.6 ± 25.8 MΩ, n = 13). We therefore show that astroglial Px1 tunes excitability but not intrinsic properties of pyramidal neurons.
Astroglial Px1 regulates neuronal network activity and excitability via A1R signalling
We next further addressed the mechanism implicated in astroglial Px1-dependent modulation of neuronal network activity and single neuron excitability. Astrocytes can regulate neuronal activity via release of neuroactive molecules  through various pathways including Px1 channels [9,10]. We therefore hypothesised that Px1 channels, when activated during population activity, release molecules that limit neuronal excitability and prevent paroxysmal events. To test this postulate, we compared extracellular levels of ATP, previously described to be released by Px1 channels , during basal and population activity in +/+ and hGFAP-Cre-Px1fl/fl mice, and found that ATP extracellular levels, measured using a luciferin–luciferase assay (Fig 5A, left panel), significantly increased during bursting activity compared to basal conditions in +/+ mice (Basal: 1.07 ± 0.33 nM, Population activity: 2.55 ± 0.61 nM; n = 7 mice; p = 0.026; Fig 5A, right panel). Notably, slices from hGFAP-Cre-Px1fl/fl mice displayed a marked decrease in ATP extracellular concentration in conditions of sustained activity compared to +/+ mice (0.53 ± 0.17 nM; n = 8 mice; p = 0.001; Fig 5A, right panel), suggesting a role for purinergic release in the astroglial Px1-mediated regulation of neuronal network activity. Through which target does Px1-released ATP control neuronal activity? ATP signalling is multifold, in that it leads to both excitation and inhibition of neuronal activity, depending on its targets. ATP is an agonist of P2X and P2Y receptors and can be cleaved by enzymatic hydrolysis to adenosine, which in turn binds A1 and A2 receptors (A1R and A2R) . Further, ATP also modulates the activity of ATP-sensitive potassium channels (KATP) . To examine whether mimicking the strong depletion of ATP measured in slices from hGFAP-Cre-Px1fl/fl mice can induce a switch from bursting activity to paroxysmal activity, we applied ATP or adenosine receptors antagonists in +/+ slices. Antagonists for P2X and P2Y receptors (PPADS and RB2, respectively), KATP channels (tolbutamide), and A2R (SCH58261) did not induce paroxysmal activity (S6 Fig). However, paroxysmal events were induced by either acutely blocking A1R pharmacologically (8-CPT) in all +/+ slices recorded (n = 9 slices from 3 mice; p < 0.0001, n = 9), or by genetic deletion using A1R −/− mice in 70% of the slices (n = 10 slices from 3 mice; Fig 5B and S1 Table), thus mimicking the network activity pattern recorded in hGFAP-Cre-Px1fl/fl mice. Consistently, antagonising A1R in conditions of sustained activity increased excitability of CA1 pyramidal neurons from +/+, but not from hGFAP-Cre-Px1fl/fl mice, as the number of APs elicited by membrane depolarisation in the presence of synaptic blockers was increased (+/+: n = 7 neurons from 6 mice, hGFAP-Cre-Px1fl/fl: n = 7 neurons from 3 mice; p = 0.039; Fig 5C and 5D). To confirm the specific role of astroglial Px1 channels and A1Rs in the network inhibition process, we restored in vivo postnatally Px1 expression selectively in hippocampal astrocytes of hGFAP-Cre-Px1fl/fl mice using adeno-associated viral vectors (Fig 6A), or applied the A1R agonist CPA (Fig 6E). We found that restoring Px1 expression in astrocytes from hGFAP-Cre-Px1fl/fl mice recovered activity-dependent EtBr uptake induced by sustained network activity (population activity-induced-EtBr uptake normalised to basal ACSF: astrocytes from hGFAP-Cre-Px1fl/fl mice, 117 ± 5%; astrocytes from hGFAP-Cre-Px1fl/fl mice + AAV Px1, 165 ± 16%, p = 0.0243, n = 5 mice; Fig 6B and 6C). Conversely, this effect was not observed using a control AAV driving the expression of GFP selectively in astrocytes (GFP AAV; astrocytes from hGFAP-Cre-Px1fl/fl mice + GFP AAV, 103 ± 8%, p = 0.9763, n = 5 mice; Fig 6B and 6C). In addition, treatment with CPA or postnatal viral expression of Px1 in astrocytes, but not control GFP, rescued a +/+ bursting pattern in hGFAP-Cre-Px1fl/fl mice (Fig 6D–6F). Taken together, our data indicate that astroglial Px1 channels limit neuronal network activity and excitability via A1R signalling.
Fig 5. A1R signalling mediates the astroglial Px1 regulation of network activity and excitability.
(A) Schematic diagram of the experimental design used to measure extracellular ATP concentration (left panel) and quantification of extracellular ATP concentration in 500 μl of ACSF in basal and population activity conditions (+/+: Basal, n = 7 mice, Population activity, n = 6 mice; hGFAP-Cre-Px1fl/fl: Basal, n = 8 mice, Population activity, n = 8 mice; right panel; one-way ANOVA and Bonferroni post hoc test). (B) Representative traces of neuronal network activity in +/+ mice before and during application of the A1R antagonist 8-CPT (1 μM; n = 9 slices from 3 mice; scale bar, 20 s, 50 μV) and in A1R−/− mice (n = 10 slices from 3 mice; scale bar, 10 s, 100 μV). The corresponding time-frequency plots are shown under the traces. (C) Representative traces of CA1 pyramidal cell firing pattern from +/+ and hGFAP-Cre-Px1fl/fl mice in condition of population activity before and after application of 8-CPT during depolarisation by 10 pA injection for 500 ms. Scale bar: 50 ms, 10 mV. (D) Number of APs induced by 10 pA current injection, after application of the A1R antagonist 8-CPT, normalised to control (+/+, n = 7 neurons from 6 mice; hGFAP-Cre-Px1fl/fl, n = 7 neurons from 3 mice; Student t test). Asterisks indicate statistical significance (*p < 0.05). The data underlying this figure can be found in the S1 Metadata E tab.
Fig 6. Restoring Px1 expression in astrocytes or pharmacological activation of A1Rs rescue bursting patterns in hGFAP-Cre-Px1fl/fl mice.
(A) Representative confocal images of GFP (green), S100 (grey), and DAPI (blue) immunolabelling in hippocampal slices from hGFAP-Cre-Px1fl/fl mice infected with AAV-GFAP-GFP-Px1 (Px1 AAV; see Materials and methods). Scale bars, upper panel, 200 μm; lower panel, 50 μm. (B) Astroglial EtBr uptake in basal or population activity conditions in hGFAP-Cre-Px1fl/fl astrocytes without or with GFP AAV and Px1 AAV expression. Scale bar, 10 μm. (C) Quantification of astroglial activity-dependent EtBr uptake normalised to control conditions in slices from +/+ and hGFAP-Cre-Px1fl/fl mice injected or not with GFP AAV and Px1 AAV (n = 5 and 5 mice, respectively; Student t test). (D) Representative trace of neuronal network activity recorded in hippocampal slices from hGFAP-Cre-Px1fl/fl mice infected with GFP AAV (top; n = 19 slices from 4 mice) and Px1 AAV (bottom; n = 21 slices from 4 mice). The corresponding time-frequency plots are shown under the traces. Scale bar, 10 s, 100 μV. (E) Representative traces of neuronal network activity in hGFAP-Cre-Px1fl/fl mice before and during application of the A1R agonist CPA (300 nM; n = 6 slices from 3 mice). The corresponding time-frequency plots are shown under the traces. Scale bar, 25 s, 100 μV. (F) Proportion of bursts and paroxysmal events recorded in hippocampal slices from +/+, hGFAP-Cre-Px1fl/fl, hGFAP-Cre-Px1fl/fl + GFP AAV, hGFAP-Cre-Px1fl/fl + Px1 AAV, and hGFAP-Cre-Px1fl/fl + CPA mice. Asterisks indicate statistical significance (*p < 0.05, **p < 0.01). The data underlying this figure can be found in the S1 Metadata F tab.
Astroglial Px1 restrains neuronal network activity via A1R-mediated regulation of HCN channels
Several A1R-dependent mechanisms have been described to impact neuronal excitability. G protein-coupled inwardly rectifying potassium channels (GIRK) activation by A1R-dependent intracellular processes was reported to decrease neuronal excitability . We thus investigated whether inhibiting GIRK channels in +/+ mice can mimick the paroxysmal activity observed in hGFAP-Cre-Px1fl/fl mice. However, we found that GIRK inhibition by SCH23390 failed to induce paroxystic activity, and only increased the frequency of bursts (n = 5; p = 0.005; S6 Fig). Alternatively, A1R-mediated partial inhibition of HCN-gated channels has been reported to decrease neuronal excitability [29,30]. To determine whether HCN channels are indeed endogenously inhibited by A1R signalling during sustained activity in +/+ mice, we measured the voltage sag ratio, which reflects the Ih current mediated by HCN channels activation, using whole-cell patch clamp recording of pyramidal cells. We found that blockade of A1R with 8-CPT indeed increased the voltage sag ratio in +/+ mice (+/+: n = 5 neurons from 4 mice; p = 0.039; Fig 7A and 7B).
Does this pathway set neuronal network pattern? To evaluate whether blockade of HCN channels can restore a bursting phenotype in hGFAP-Cre-Px1fl/fl mice, we inhibited HCN channels with ZD7288 and indeed found that this switched the activity pattern from paroxysmal to bursting activity (n = 5 slices from 4 mice; Fig 7C), while it had no effect on the bursting pattern in +/+ mice (n = 5 slices from 4 mice; p = 0.513 and p = 0.486 for burst frequency and duration, respectively; Fig 7D and 7E and S1 Table). Interestingly, the bursting pattern induced by ZD7288 in hGFAP-Cre-Px1fl/fl mice (n = 5 slices from 4 mice) did not differ from that observed in +/+ mice (n = 6 slices from 2 mice; p = 0.339 and 0.441 for burst frequency and duration, respectively; S1 Table). Lastly, to ensure that the A1R-mediated negative control of HCN channels can limit population activity, we induced paroxysmal activity in all tested hippocampal slices from +/+ mice by inhibiting A1Rs with 8-CPT (n = 5 slices from 4 mice, p < 0.0001) and subsequently blocked HCN channels with ZD7288. Consistently, HCN channel antagonism inhibited paroxysmal events caused by A1R blockade, systematically reverting the electrophysiological phenotype to a bursting pattern (n = 5 slices from 4 mice, p < 0.0001), which was similar to the one displayed in control condition (Fig 7F; p = 0.3 and p = 0.189 for burst frequency and duration, respectively, n = 5 slices from 4 mice). Altogether, these data suggest that Px1 channels in astrocytes tune down population activity through purinergic signalling-mediated regulation of HCN channels (Fig 8).
Fig 7. Astrocytic Px1 limits network activity through A1R-mediated modulation of HCN channels.
(A) Representative traces of the voltage sag ratio in +/+ mice before and after application of the A1R antagonist 8-CPT. Scale bar: 50 ms, 10 mV. (B) Quantification of voltage sag ratio, defined as ((Vmin − Vend)/Vmin) × 100 (n = 5 neurons from 4 mice, paired Student t test). (C-D) Representative traces of neuronal network activity recorded in hGFAP-Cre-Px1fl/fl mice (C) and +/+ mice (D) before and during application of the HCN channel antagonist ZD7288 (10 μM; hGFAP-Cre-Px1fl/fl, n = 5 slices from 4 mice; +/+, n = 6 slices from 2 mice). The corresponding time-frequency plots are shown under the traces. Scale bars, in c: 10 s, 200 μV; in d: 10 s, 50 μV. (E) Quantification of the change in burst frequency in +/+ mice after application of ZD7288 (paired Student t test). (F) Representative traces of neuronal network activity recorded in the same hippocampal slice before and after subsequent applications of 8-CPT and ZD7288 in +/+ mice (n = 5 slices from 4 mice). The corresponding time-frequency plots are shown under the traces. Scale bars, control and 8-CPT + ZD7288: 10 s, 25 μV; 8-CPT: 25 s, 25 μV. Asterisks indicate statistical significance (*p < 0.05). The data underlying this figure can be found in the S1 Metadata G tab.
Fig 8. Mechanism of Px1-mediated modulation of network activity.
Schematic diagram depicting the proposed mechanism through which astroglial Px1 channels signalling decreases population activity: Px1 channels release ATP, converted to adenosine, which binds neuronal A1 receptors. Subsequent intracellular signalling induces inhibition of HCN channels, leading to decreased excitability and population activity. Ado, Adenosine.