A mouse model of autism implicates endosome pH in the regulation of presynaptic calcium entry

Psychoactive compounds such as chloroquine and amphetamine act by dissipating the pH gradient across intracellular membranes, but the physiological mechanisms that normally regulate organelle pH remain poorly understood. Interestingly, recent human genetic studies have implicated the endosomal Na+/H+ exchanger NHE9 in both autism spectrum disorders (ASD) and attention deficit hyperactivity disorder (ADHD). Plasma membrane NHEs regulate cytosolic pH, but the role of intracellular isoforms has remained unclear. We now find that inactivation of NHE9 in mice reproduces behavioral features of ASD including impaired social interaction, repetitive behaviors, and altered sensory processing. Physiological characterization reveals hyperacidic endosomes, a cell-autonomous defect in glutamate receptor expression and impaired neurotransmitter release due to a defect in presynaptic Ca2+ entry. Acute inhibition of synaptic vesicle acidification rescues release but without affecting the primary defect due to loss of NHE9.

Middle panel shows an immunoblot of the same samples probed with HA antibody, confirming the identity of the protein as NHE9. Since western blotting revealed no specific bands in WT relative to KO brains (data not shown), we adsorbed the NHE9 antibody with an acetone extract made from the brains of NHE9 KO mice. Right panel, Membrane-enriched extracts (25 or 100 µg protein) from the brains of the same WT and KO mice used for qPCR transcript analysis in Figure 1 were immunoblotted with affinity-purified NHE9 antibody and detected with femto ECL substrate. Adsorbed antibody detects NHE9-HA from HEK cells but shows no specific signal in the brain lysates. We similarly adsorbed several commercial NHE9 antibodies as well as those generated in other labs, all of which could detect NHE9-HA over-expressed in HEK cells, but did not find any capable of detecting a protein in the WT that was not also present in the KO (data not shown). Immunofluorescence similarly failed to identify a signal in the WT that was not also present in the KO (data not shown).
(d) qPCR amplification of exon 2-5 from brain cDNA shows a single product for WT mice, two for HET and one for cKO (left panel). Deletion of exon 5 results in bands 115 bp smaller. PCR from exon2-5 illustrates the absence of a detectable transcript present in cKO brains (right panel).   (a) cKO brain weight (n=16) is the same as WT (n=22  Interhemispheric EEG recordings were made over the prefrontal cortex of two freely moving NHE9 knockout (962 left and 3 middle) and one wild type C57Bl/6Tn mouse (right). (a) Sample EEG records at low temporal resolution show intermittent high voltage activity that lasts up to 10 minutes in the knockouts but not wild type.

Arrows and numbers indicate segments of the trace shown at expanded temporal resolution in (c) and (d). (b)
In knockout mice, the corresponding spectrograms show that the bursts of high amplitude activity coincide with periods of increased low frequency (<20 Hz) and decreased high frequency (>30 Hz) power. These episodes also correlate with immobility in the knockouts. Baseline (c) and high voltage (d) EEG activity are shown at high temporal resolution for the knockout mice (left and middle). Two segments of the trace from wild type are shown for comparison to the right. (e) Spectrogram of the high voltage activity in (d) shows a peak frequency at 3-4 Hz for both knockout mice, but the wild type shows a peak at 0.  (c) BCECF was calibrated using live neurons loaded with BCECF-AM. External solutions at the pH indicated was used to equilibrate vesicle with buffer pH in the presence of ionophores nigericin and valinomycin to equilibrate vesicular with buffer pH. The lines were fit by linear regression, and r2=0.995 for WT, 0.992 for KO. WT, n=9 neurons/3cultures; KO, n=9 neurons/3 cultures (d) A cDNA encoding the FLAG-tagged, G protein-coupled delta-opioid receptor (DOR) was transfected into hippocampal neurons and the cells at DIV14 stimulated with the agonist DADLE (10 µM) to monitor degradation in the endolysosomal pathway. Western blotting for the FLAG epitope shows that DOR rapidly internalizes in response to DADLE and then undergoes degradation. WT and KO, n=6 wells/3 cultures. Bars indicate mean ± s.e.m.

Supplementary Figure 5. Loss of NHE9 Does Not Affect Spine Number, Morphology or Vesicular Glutamate Transport Activity.
(a) Nissl stain of WT and NHE9 KO brain slices shows grossly normal brain architecture. Scale bar 250 µm. (b) Representative Golgi stain of CA1 dendritic spines and branching shows no difference between NHE9 KO and WT. Scale bars indicate 5 µm for dendritic spines (left) and 25 µm for dendritic branching (right). (c) The analysis of spine density and shape by electron microscopy shows no difference between WT and KO hippocampal CA1 pyramidal neurons. WT, n=54 neurons/4 brains; KO, n=54 neurons/4 brains (d) Synaptic vesicles purified from the whole brain of WT and NHE9 KO mice were incubated for 10 minutes with radiolabeled glutamate or GABA in the presence or absence of K + . The background in Evans Blue was subtracted for glutamate uptake, and in the inophores valinomycin and nigericin for GABA. K + stimulation for each LP2 was normalized to the uptake in choline. n=3 independent LP2 preparations.   (a) Representative trace from control experiment to verify that bafilomycin loaded into vesicles expressing VGLUT1-pHluorin increases the pH of these vesicles and the more neutral pH is stable for 10 minutes.
(b) Analysis of the single exponent time constant for Ca ++ decay after 10 Hz stimulation shows that bafilomycin accelerates decay in the KO neurons. ***, p<0.001 by one-way ANOVA with Bonferroni. WT, n=22/21 coverslips/3 cultures; KO, n=21/21 coverslips/3 cultures genotype per condition (c) Analysis of the single exponent time constant for Ca ++ decay in response to a single action potential before and after synaptic vesicle loading with bafilomycin shows that bafilomycin accelerates the decay of Ca ++ fluorescence for both WT and KO neurons (right panel). **, p<0.01; ****, p<0.0001 by one-way ANOVA with Bonferroni. WT, n=15/15 coverslips/3 cultures; KO, n=14/14 coverslips/3 cultures genotype per condition.