Alix is required for activity-dependent bulk endocytosis at brain synapses

Abstract

In chemical synapses undergoing high frequency stimulation, vesicle components can be retrieved from the plasma membrane via a clathrin-independent process called activity-dependent bulk endocytosis (ADBE). Alix (ALG-2-interacting protein X/PDCD6IP) is an adaptor protein binding to ESCRT and endophilin-A proteins which is required for clathrin-independent endocytosis in fibroblasts. Alix is expressed in neurons and concentrates at synapses during epileptic seizures. Here, we used cultured neurons to show that Alix is recruited to presynapses where it interacts with and concentrates endophilin-A during conditions triggering ADBE. Using Alix knockout (ko) neurons, we showed that this recruitment, which requires interaction with the calcium-binding protein ALG-2, is necessary for ADBE. We also found that presynaptic compartments of Alix ko hippocampi display subtle morphological defects compatible with flawed synaptic activity and plasticity detected electrophysiologically. Furthermore, mice lacking Alix in the forebrain undergo less seizures during kainate-induced status epilepticus and reduced propagation of the epileptiform activity. These results thus show that impairment of ADBE due to the lack of neuronal Alix leads to abnormal synaptic recovery during physiological or pathological repeated stimulations.

Fig 1. Alix is recruited presynaptically during synaptic activation.

(A) Western blot analysis of cortical neurons cultured for 3 to 15 DIV demonstrates the increase in Alix expression correlating with synaptogenesis as illustrated by the increase in PSD95 expression. (B) Western blot analysis of the increase of Alix in synaptosome-enriched neuronal membranes upon neuronal stimulation by Bic/4AP. Synaptophysin and PSD95 were used as pre- and postsynaptic markers, respectively. The phosphorylated form of ERK (p-ERK) assessed the efficiency of the stimulation. (C) Images from time-lapse video microscopy of 15 DIV hippocampal neurons expressing both mche-Alix and Syp-pH stimulated between 2 and 7 min with Bic/4AP. White arrowheads indicate presynaptic boutons where Alix is recruited during stimulation. Scale bar: 10 μm. (D) mche-Alix and Syp-pH fluorescence variation at presynaptic boutons during Bic/4AP incubation (blue line). (E) Number of synapses responding to Bic/AP stimulation (Syp-pH increase) and recruiting Alix (mche-Alix increase). (F) 15 DIV hippocampal neurons expressing Alix-YFP (green) were stimulated for 5 min with Bic/4AP before fixation and stained with anti-synapsin-1 antibody (Syn, magenta). Dashed white squares indicate insets 1 and 2. Scale bars: 5 and 1 μm (inset). (G) Graph shows the presynaptic increase in Alix-YFP upon stimulation. Presynaptic Alix-YFP corresponds to the ratio of YFP fluorescence between synapsin-positive and synapsin-negative axonal regions. (H) Selective recruitment of Alix to the presynaptic part on synapses: 15 DIV hippocampal neurons expressing Alix-YFP (yellow) were stimulated for 5 min before fixation and stained with anti-synapsin-1 antibody (Syn, magenta) and anti-PSD95 (postsynaptic, cyan). Dashed white square indicates the inset on a single synaptic bouton. Scale bars: 5 and 1 μm (inset). (I) Quantification of the distance between the peak of fluorescence between Alix and synapsin (magenta) or Alix and PSD95 (cyan) shows that Alix is significantly closer to synapsin than to PSD95. (J) Colocalization coefficient (Pearson’s correlation coefficient) showing a higher level of colocalization between Alix and synapsin (magenta) than that between Alix and PSD95 (cyan). Average +/‒ SEM, N, statistical analysis: (B) 2.07 +/‒ 0.34; 1.38 +/‒ 0.21; 1.00 +/‒ 0.01 for Alix, synaptophysin, and PSD95, respectively. N = 4 independent experiments, Alix versus PSD95, p = 0.0187, 1-way ANOVA. (D) N = 25 and 16 synapses for Syp-pH and mChe-Alix, respectively, from 2 independent experiments. (E) Active synapses: 12.5 +/‒ 0.7; recruiting Alix: 10 +/‒ 1.4. N = 25 synapses from 2 independent experiments. (G) 1.63 +/‒ 0.12; 3.27 +/‒ 0.44 for no stim and stim, respectively. N = 12 neurons per condition from 4 independent experiments, p = 0.0017, unpaired t test. (I) Syn: 143.6 +/‒ 46 nm, PSD95: 430.9 +/‒ 133.7 nm. N = 5 independent experiments (40 synapses). Syn vs. PSD95, p = 0.0019, unpaired t test. (J) Syn: 0.49 +/‒ 0.1, PSD95: 0.21 +/‒ 0.06. N = 14 neurons from 5 independent experiments. The data underlying all the graphs shown in the figure can be found in S1 Data. Alix, ALG-2-interacting protein X; DIV, day in vitro; mche-Alix, mCherry-Alix; PSD95, postsynaptic density protein 95; Syp-pH, synaptophysin-pHluorin.

Fig 2. Interplay between Alix, ALG-2, and endophilin recruitments at activated synapses.

All experiments made use of 15 DIV hippocampal neurons expressing the indicated constructs. Neurons were stimulated with Bic/4AP for 5 min before fixation and stained with anti-synapsin-1 (Syn, magenta). (A) GFP-ALG-2 (green) is recruited presynaptically upon stimulation. The yellow dashed line indicates the position of the axon in the synapsin-I channel (upper panel). Scale bar: 5 μm. (B) Quantification of synaptic recruitments of GFP-ALG-2 in Alix wt (black) and Alix ko (gray) neurons shows that ALG-2 presynaptic recruitment does not depend on Alix. (C) Quantification of synaptic recruitments of GFP-ALG-2ΔCa in Alix wt neurons (black) shows that ALG-2 presynaptic increase depends on its capacity to bind calcium. (D) Quantification of synaptic recruitments of GFP-AlixΔALG-2 in Alix ko neurons shows that Alix recruitment depends on its capacity to bind ALG-2 as GFP-AlixΔALG-2 is not recruited upon stimulation of Alix ko neurons. (E) Endophilin-mCherry (endo, green) concentrates at presynaptic parts of stimulated neurons. The yellow dashed line indicates the position of the axon in the synapsin-I channel (upper panel). Scale bar: 5 μm. (F) Synaptic recruitment of endophilin-mCherry following synaptic activation requires Alix as it is impaired in Alix ko neurons. (G) Quantification of synaptic recruitments of GFP-AlixΔendo in Alix ko neurons shows that Alix recruitment does not depend on its binding to endophilins. Average +/‒ SEM, N, statistical analysis: (B) 1.41 +/‒ 0.13; 2.24 +/‒ 0.11; 1.36 +/‒ 0.06; 2.06 +/‒ 0.12 for Alix wt no stim, Alix wt stim, Alix ko no stim, Alix ko stim, respectively. N = 7, 10, 17, 17 neurons for GFP-ALG-2 in Alix wt no stim, Alix wt stim, Alix ko no stim, Alix ko stim, respectively, from 3 experiments. p = 0.0001 for all the conditions tested, 1-way ANOVA. (C) 1.8 +/‒ 0.05; 1.18 +/‒ 0.16 for GFP-ALG2ΔCa in no stim and stim conditions, respectively. N = 12 and 6 neurons for no stim and stim, respectively. (D) 1.54 +/‒ 0.05; 1.48 +/‒ 0.08 for GFP-AlixΔALG-2 in no stim and stim conditions, respectively. N = 18 and 15 neurons for no stim and stim, respectively, from 3 independent experiments. (F) 1.92 +/‒ 0.11; 3.48 +/‒ 0.33; 2.34 +/‒ 0.12; 2.61 +/‒ 0.08 for endophilin-mche in Alix wt no stim, Alix wt stim, Alix ko no stim, Alix ko stim, respectively. N = 10, 12, 13, 19 neurons for Alix wt no stim, Alix wt stim, Alix ko no stim, Alix ko stim, respectively, from 3 independent experiments. Alix wt no stim vs. Alix wt stim, p < 0.0001, 1-way ANOVA. (G) 1.56 +/‒ 0.06; 2.43 +/‒ 0.32 for GFP-AlixΔendo in no stim and stim conditions, respectively. N = 10 and 9 neurons for no stim and stim, respectively, from 3 independent experiments. No stim vs. stim, p = 0.0004, Mann–Whitney test. The data underlying all the graphs shown in the figure can be found in S1 Data. Alix, ALG-2-interacting protein X; DIV, day in vitro; ko, knockout; wt, wild type.

Fig 3. Alix is necessary for activity-dependent bulk endocytosis.

(A, C) Normalized traces of Syp-pH fluorescence in synaptic boutons of Alix wt and ko hippocampal neurons stimulated with 200 APs applied at 5 Hz (A) or 40 Hz (C). (B, D) The exponential fit of fluorescence decay in the imaged fields after stimulations show an increased decay in Alix ko neurons at 40 Hz (D) but not at 5 Hz (B). (E, H) Electron micrographs of cerebellar granule neurons stimulated in presence of free HRP to label newly formed SV (blue arrowheads) and bulk endosomes (red arrowheads). (E) FIB-SEM orthogonal views from different planes (xy, xz, yz) extracted from a stack used for the 3D reconstruction of a wt presynaptic terminal shown in (F). Scale bar: 500 nm. (F) Two different views of the reconstructed synapse are shown where the membrane is represented in transparent gray and HRP-positive structures in red. Scale bar: 500 nm. (G, H) Transmission electron microscopy images of Alix wt and Alix ko cerebellar granule neurons incubated with HRP. Scale bars: 500 nm. (I, J) Quantification of the number of bulk endosomes (I) and SVs (J) in Alix wt and ko neurons from images as shown in (G). (K) Quantification of the diameter of SV (blue dots) and bulk endosome (red dots) in Alix wt and Alix ko neurons in basal or stimulated condition showing that bulk endosomes of Alix ko synapses are smaller than in Alix wt neurons. Average +/‒ SEM, N, statistical analysis: (B, D) 6.47 +/‒ 0.78; 6.72 +/‒ 0.53; 9.71 +/‒ 0.88; 6.45 +/‒ 0.44 for Alix wt 5 Hz, Alix ko 5 Hz, Alix wt 40 Hz, Alix ko 40 Hz, respectively. N = 16, 35, 29, 46 fields of view per condition from 4 experiments for wt and 5 for ko mice. Alix wt vs. Alix ko, p = 0.99 (B) and p = 0.002 (D), Mann–Whitney test. (I) 0.52 +/‒ 0.17; 0.64 +/‒ 0.16; 3.28 +/‒ 0.19; 1.82 +/‒ 0.07 for Alix wt no stim, Alix ko no stim, Alix wt stim, Alix ko stim, respectively. N = 4, 3, 4, 4 independent experiments Alix wt no stim, Alix ko no stim, Alix wt stim, Alix ko stim, respectively. Alix wt no stim vs. Alix wt stim, p < 0.0001; Alix ko no stim vs. Alix ko stim, p = 0.0015; Alix wt stim vs. Alix ko stim, p = 0.0001, 1-way ANOVA. (J) 4.43 +/‒ 1.02; 4.26 +/‒ 0.59; 7.46 +/‒ 0.49; 9.64 +/‒ 0.67 for Alix wt no stim, Alix ko no stim, Alix wt stim, Alix ko stim, respectively. N = 4, 3, 4, 4 independent experiments Alix wt no stim, Alix ko no stim, Alix wt stim, Alix ko stim, respectively. Alix wt stim vs. Alix ko stim, p = 0.039, unpaired t test. (K) 112.2 +/‒ 2.37 nm; 191.7 +/‒ 3.39 nm; 152.4 +/‒ 4.1 nm; 135.2 +/‒ 1.79 nm for Alix wt no stim, Alix wt stim, Alix ko no stim, Alix ko stim, respectively. N = 304, 1,326, 285, 2,415 vesicles for Alix wt no stim, Alix wt stim, Alix ko no stim, Alix ko stim, respectively, from 3 independent experiments. p < 0.0001 in all conditions tested, Kruskal–Wallis test. The data underlying all the graphs shown in the figure can be found in S1 Data. Alix, ALG-2-interacting protein X; AP, action potential; FIB-SEM, focused ion beam-scanning electron microscopy; HRP, horseradish peroxidase; ko, knockout; SV, synaptic vesicle; Syp-pH, synaptophysin-pHluorin; wt, wild type.

Fig 4. Alix-driven bulk endocytosis requires its binding to ALG-2 and endophilin but not to CHMP4.

(A) Confocal images of Alix wt and Alix ko hippocampal neurons stimulated with Bic/4AP in the presence of 10 kDa dextran. Scale bar: 50 μm. (B) Dextran uptake triggered by stimulation is strongly reduced in Alix ko neurons. (C) Representative images of dextran uptake by GFP-expressing Alix wt and Alix ko neurons (Alix wt, top and Alix ko, middle) or Alix ko neurons expressing both GFP and Alix (Alix ko + Alix, bottom). Scale bars: 10 μm. (D) Dextran uptake is rescued in ko neurons expressing Alix wt and AlixΔChmp4B (AlixΔ4B), but not AlixΔALG-2 or AlixΔendo. The % of dextran uptake corresponds to the number of dextran spots per μm expressed as percentages of the control (Alix wt) for each experiment. Average +/‒ SEM, N, statistical analysis: (B) 16.26 +/‒ 1.08; 11.86 +/‒ 2.05; 100 +/‒ 1.76; 15.10 +/‒ 1.76 for Alix wt no stim, Alix ko no stim, Alix wt stim, Alix ko stim, respectively. N = 4 experiments, p = 0.0001, 1-way ANOVA. (D) 100 +/‒ 6.31; 33.27 +/‒ 3.32; 94.78 +/‒ 5.81; 45.89 +/‒ 9.36; 80.05 +/‒ 7.32; 44.59 +/‒ 1.85 for Alix wt, Alix ko, Alix ko + Alix, Alix ko + AlixΔChmp4B, Alix ko + AlixΔendo, Alix ko + AlixΔALG-2, respectively. N = 17 neurons for Alix wt, N = 13 neurons for Alix ko, N = 11 for Alix ko + Alix, N = 12 for Alix ko + AlixΔChmp4B, N = 12 neurons for Alix ko + AlixΔendo, and N = 11 for Alix ko + AlixΔALG-2, all from 3–4 independent experiments, p = 0.62 (Alix wt vs. Alix AlixΔ4B), p = 0.0001 for the other conditions, 1-way ANOVA. The data underlying all the graphs shown in the figure can be found in S1 Data. Alix, ALG-2-interacting protein X; ko, knockout; wt, wild type.

Fig 5. Alterations of Alix ko synapses revealed in hippocampal slices and in an in vivo model of epilepsy.

(A) Emx1 ^IREScre (Emx-Cre) and Alix ^fl/fl mouse lines were crossbred to delete Alix in neocortical and hippocampal excitatory neurons (Alix cko). (B) No difference in sEPSC amplitude was detected between control and cko mice. (C) The frequency of sEPSC in cko neurons is lower compared to control. (D) Representative traces showing short-term depression in response to 10 Hz stimulation trains in control (black) and cko (gray) mice. (E) Cumulative EPSC amplitudes in response to 10 Hz stimulation. Train-extrapolation is illustrated by the dashed line. (F) The size of the RRP, estimated by the train-extrapolation method, was significantly decreased in Alix cko mice. (G) Representative traces showing recovery from depression in response to 10 Hz stimulation trains in control (black) and Alix cko (gray) mice. (H) To evaluate the speed of recovery, all inward signals were normalized to the first inward current. Recovering signals with a single-exponential function revealed the plateaus at which the capacities of recovery were saturated. (I) Alix cko neurons recover slightly faster than controls. (J) Representative EEG traces from control and Alix cko mice. (K) Total duration of SE did not differ between Alix cko and control mice. (L) Alix cko mice experience about 66% fewer seizures during SE than controls. (M) Mean seizure duration during SE was not affected in Alix cko mice compared to controls. Average +/‒ SEM, N, statistical analysis: (B) 17.93 +/‒ 1.81; 16.49 +/‒ 1.35 for controls and Alix cko, respectively. N = 13 neurons from 5 control mice and N = 15 neurons from 6 Alix cko mice. (C) 0.31 +/‒ 0.09; 0.12 +/‒ 0.03 for controls and Alix cko, respectively. N = 10 neurons from 5 control mice and N = 15 neurons from 6 Alix cko mice. Control vs. cko, p = 0.015, Mann–Whitney test. (F) 1.23 +/‒ 0.15; 0.67 +/‒ 0:08 for controls and Alix cko, respectively. N = 12 neurons from 5 control mice and N = 10 neurons from 4 Alix cko mice. Control vs. cko, p = 0.0052, unpaired t test. (I) 8.34 +/‒ 0.90; 5.56 +/‒ 0.69 for controls and Alix cko, respectively. N = 5 and 6 neurons from 3 mice from control and Alix cko, respectively. Control vs. cko, p = 0.0336, unpaired t test. Median (IQR), N, statistical analysis: (K) Control: 1.6 h (IQR = 2.5); cko: 1.7 h (IQR = 0.9). N = 11 mice for both genotypes. (L) Control: 18 seizures (IQR = 9); cko: 8 seizures (IQR = 9). N = 11 mice for both genotypes. Control vs. cko, p = 0.023, Mann–Whitney test. (M) Control: 32.9 s (IQR = 18.6); cko: 25.7 h (IQR = 12.7). N = 11 mice for both genotypes. The data underlying all the graphs shown in the figure can be found in S1 Data. Alix, ALG-2-interacting protein X; cko, conditional ko; IQR, interquartile range; RRP, readily releasable pool; SE, status epilepticus; sEPSC, spontaneous excitatory postsynaptic current.