Pathogenic Mechanism of an Autism-Associated Neuroligin Mutation Involves Altered AMPA-Receptor Trafficking

Mouse husbandry

NL3 R704C KI mice (Figs. 1 and 2) were maintained and genotyped as described. ¹⁷ All mouse work was approved by Stanford University Administrative Panel on Laboratory Animal Care (APLAC).

Primary neuronal cultures

Newborn mice (postnatal day P0-P2) were used to derive primary mixed cultures of hippocampal and olfactory bulb neurons as described (for detailed procedures, see supplementary information).^18,19

Plasmids constructs, virus productions, and infection and transfection of cultured neurons

Neuroligin constructs (empty for control vectors) were cloned after the human synapsin promoter and followed by an IRES-EGFP expression cassette for fluorescence visualization in expressing cells (Fig. S3). Lentiviruses were produced using transfected HEK293 cells, ²⁰ and used as described. ²¹ Neuronal transfections were performed using the Ca²⁺/phosphate method (for detailed protocols, see supplementary information). ²²

Electrophysiology

Whole-cell voltage-clamp recordings were performed essentially as described, using pharmocological blockers to isolate AMPAR-, NMDAR-, and GABA_AR-mediated synaptic currents and extracellular stimulation to elicit evoked responses. ^19,23

Immunocytochemistry

Immunostaining of AMPAR surface levels and AMPAR endocytosis (Fig. 2) was performed by sequential applications of rat, mouse or/and rabbit antibodies (for details, see Fig. S5 and ref. 24). Cell-surface binding assays were performed in HEK293 cells expressing HA-tagged NL3-WT or NL3-R704C (Fig. S5). ^25,26

Immunoprecipitation and immunoblotting

Co-immunoprecipitations of AMPARs with NL3 were performed using HA-antibodies and cultured cortical neurons infected with lentiviruses encoding HA-tagged NL3-WT or NL3-R704C. Samples were analyzed by immunoblotting using fluorescently labeled anti-GLUA2 antibody with synaptophysin as an internal control (Fig. 4).

The R704C substitution in NL3 decreases synaptic AMPAR levels and increases AMPAR endocytosis

We had shown earlier that the NL3 R704C-knockin mutation caused a selective decrease in AMPAR- but not NMDAR-mediated synaptic responses.¹⁷ To explore the functional importance and mechanism underlying this observation, we analyzed hippocampal neurons cultured from littermate WT and NL3-R704C knockin mice. We confirmed a specific, ~60% decrease in AMPAR-mediated evoked EPSCs without a change in NMDAR-mediated EPSCs or in GABA_AR-mediated evoked IPSCs (Figs. 1A–1C; Table S1). To explore whether this phenotype reflected a loss in postsynaptic AMPARs, we measured spontaneous ‘mini’ mEPSCs (which are mediated by AMPARs) and mIPSCs (GABA_AR-mediated). Consistent with the decrease in AMPAR-mediated evoked EPSCs, we observed a significant decrease in mEPSC amplitude (~25% decrease) and frequency (~60% decrease), but no change in mIPSC amplitude or frequency (Figs. 1D and 1E).

The decreased mEPSC amplitude in NL3 R704C-mutant synapses indicates that the postsynaptic concentration of AMPARs per synapse may be decreased, whereas the decreased mEPSC frequency could be due to either a silencing of a subset of synapses, an undersampling of small mEPSCs, or a reduction in synapse numbers. We measured paired-pulse ratios (PPRs), which are indirect measurements of the presynaptic release probability of excitatory and inhibitory synapses, but failed to uncover a change, arguing against a presynaptic effect (Fig. S2). To directly monitor postsynaptic AMPARs, we quantified the surface levels of AMPARs by immunocytochemistry of non-permeabilized neurons using GluA1 antibodies (Fig. 2A). ²⁴ In addition, we electrophysiologically recorded AMPAR EPSCs following an acute application of AMPA, an AMPAR agonist (Fig. 2B). Both methods revealed a large decrease (~40–60%) in surface AMPAR levels in NL3-R704C knockin neurons, suggesting that the R704C substitution decreased postsynaptic AMPAR levels.

To test how the NL3-R704C decreases surface AMPAR levels, we measured the internalization of the AMPAR subunit GluA1. To this end, we labeled live neurons with an antibody to the extracellular GluA1 sequences for 5 min, washed the cells, and incubated them for another 15 min before fixing the neurons and separately staining them for surface AMPARs and internalized AMPARs. ²⁴ We detected a ~50% increase in GluA1 internalization in NL3-R704C knockin neurons, suggesting that the R704C substitution in NL3 accelerates the endocytosis of surface AMPARs (Fig. 2C).

R704C-mutant NL3 acts in a dominant and cell-autonomous fashion

The NL3-R704C knockin produces a different, more severe phenotype (Figs. 1 and 2) than the NL3 knockout, ^27–29 suggesting a possible gain-of-function effect. To explore this hypothesis, we tested whether expression of R704C-mutant NL3 in wild-type neurons produces a phenotype that mimics the R704C knockin mutation.

We first examined the synaptic effects of infecting cultured olfactory bulb neurons with a control lentivirus or with lentiviruses expressing wild-type or R704C-mutant NL3 (Fig. S3). Olfactory bulb neurons were chosen because of their well-defined neuronal composition, ¹⁸ and lentiviral infection because it accomplishes homogeneous gene delivery to all neurons in culture. ²⁴ We observed the same selective impairment in AMPAR-mediated synaptic responses in olfactory bulb neurons expressing R704C-mutant NL3 (but not in neurons expressing NL3-WT) as in hippocampal neurons containing the NL3 R704C-knockin mutation (Fig. 3A). Since wild-type neurons express endogenous NL3-WT, R704C-mutant NL3 must act as a dominant negative in this experiment.

We next inquired if the R704C phenotype is due to a cell-autonomous mechanism. We sparsely transfected olfactory bulb neurons with the plasmids used for lentiviral infection experiments (control, NL3-WT, and NL3-R704C), and measured synaptic responses in the transfected neurons (Figs. 3B and S3). Again, we observed in NL3-R704C expressing neurons a selective AMPAR-impairment similar to that of NL3-R704 knockin neurons (Figs. 3B, 3C, and S4). Thus, R704C-mutant NL3 alters AMPAR trafficking by a postsynaptic, cell-autonomous, and dominant-negative mechanism that is identical between NL3 R704C-knockin neurons and wild-type neurons with exogenously expressed NL3-R704C. Unexpectedly, neurons expressing exogenous NL3-WT did not have an increase in AMPAR- or NMDAR-mediated EPSCs compared to control neurons. Since expression of WT- NL3 and NL3-R704C causes a similarly large increase in synapse numbers, ¹⁷ the fact that NL3-WT does not increase EPSCs suggests that the morphologically observed newly induced synapses are non-functional.

The R704C-mutation enhances the interaction of NL3 with AMPARs

How does the NL3-R704C mutation decrease surface AMPAR levels and increase AMPAR endocytosis? We hypothesized that NL3 may be associated with AMPARs in neurons, and that the R704C mutation changes this association. To test this hypothesis, we used NL3-WT and R704C-NL3 containing an extracellular HA epitope, allowing us to monitor and immunoprecipitate NL3 (Fig. 4A and S5A). We first confirmed in hippocampal neurons that lentiviral expression of HA-tagged NL3-R704C produces the same phenotypes as the R704C knockin in hippocampal neurons and as non-tagged NL3-R704C in olfactory bulb neurons (Fig. 4B; Table S1). We then examined whether AMPARs are co-immunoprecipitated with HA-tagged NL3, using synaptophysin as a negative control (Fig. 4C). In these experiments, we monitored AMPARs with antibodies to the GluA2 AMPAR subunit, and also asked whether NL3-R704C changes the co-immunoprecipitation of AMPARs. We detected significant co-immunoprecipitation of GluA2 with NL3-WT that was dramatically enhanced by the R704C mutation (Fig. 4C). These experiments suggest that NL3 is directly or indirectly associated with AMPARs, possibly via common binding to a scaffolding protein such as PSD-95, ³⁰ and that the R704C-mutation enhances this association.

The most plausible explanation for the observed effects of the R704C mutation in NL3 – the decrease in AMPAR surface levels and the increase in AMPAR endocytosis (Fig. 2) – is that the R704C-mediated increase in the interaction of NL3 with AMPARs is accompanied by accelerated NL3 endocytosis, thereby enhancing internalization of bound AMPARs. To test this hypothesis, we measured the density and size of NL3-containing surface puncta in hippocampal neurons expressing HA-tagged NL3-WT or NL3-R704C (Figs. S5B and S5C). We found that the density of postsynaptic NL3-containing puncta (like that of AMPAR puncta; Fig. 2A) was unchanged by the R704C mutation, but that the size of the puncta, a reflection of the amount of NL3 present in a synapse under our assay conditions, was dramatically decreased (Figs. 4D). We then measured the relative distribution of NL3-WT and NL3-R704C between the cell surface and cell interior by immunocytochemistry, and observed a relative loss of NL3-R704C from the cell surface (Figs. 4E and S5B). The R704C mutation in NL3 did not change the density of vGLUT1-positive puncta, measured to monitor excitatory synapse numbers (Figs. 4F). NL3-vGLUT1 co-localization experiments, however, confirmed that the R704C mutation caused significant loss of NL3 from excitatory synapses (Figs. 4F). Viewed together, these experiments suggest that the R704C-mutation causes NL3 both to bind more strongly to AMPARs (directly or indirectly) and to undergo more endocytosis, thereby depleting AMPARs from postsynaptic specializations.

R704C mutation in NL4 enhances AMPAR-mediated synaptic responses

We initially analyzed the R704C mutation in NL3 despite the fact that this autism-relevant mutation was observed in NL4 because we wanted to examine this mutation using a knockin approach without exogenous expression of mutant proteins. ^17,23 We could not do such a knockin for mouse NL4 because mouse NL4 is expressed at only low levels, and most importantly, its sequence is not yet present in the current mouse genome sequence. ⁴ However, the current finding that the NL3-R704C knockin phenotype is fully reproduced by expression of exogenous NL3-R704C in wild-type neurons – as long as it is compared to the NL3-WT (Fig. 3) – indicates that the effects of the R704C mutation on NL4 can be examined by expression of R704C-mutant NL4 in wild-type neurons.

We infected cultured hippocampal neurons with lentiviruses expressing either EGFP only (control), or both EGFP and NL4-WT or NL4-R704C. We then measured evoked AMPAR- and NMDAR-mediated EPSCs and GABA_AR-mediated IPSCs in these neurons (Fig. 5). Strikingly, we observed that overexpression of NL4-WT – consistent with earlier results ¹⁰ – caused a decrease in both AMPAR- (~38%) and NMDAR-mediated EPSCs (~45%) without affecting IPSCs or the PPR (Figs. 5A, and S6). In contrast, overexpression of NL4-R704C produced an increase in AMPAR- and NMDAR-mediated responses (~66% and ~100%, respectively), compared to NL4-WT (Fig. 5A). Since NL4 expression also increases synapse numbers, ¹⁰ it seems to induce formation of non-functional synapses and additionally to inactivate existing synapses, while the R704C mutation reverses this inactivating effect of NL4. The selective effects of exogenously expressed NL4 on excitatory synapses in all parameters analyzed and the distinct changes induced in these effects by the R704C mutation strongly suggests that NL4 normally also acts in excitatory synapses, and that the R704C mutation impairs NL4 function.

To confirm that the NL4-R704C-induced changes reflect changes in surface AMPARs, we measured the effect of NL4-WT and NL4-R704C on the response to AMPA puffed onto the neurons (Fig. 5B). We found that expression of WT-NL4 had no significant effect on the total AMPAR-induced current, which is due to both synaptic and extrasynaptic AMPARs. Since in the same NL4-WT-expressing neurons the synaptic AMPAR-mediated responses are decreased, this result suggests that WT-NL4 may not change the total surface AMPAR levels, but cause their redistribution out of synapses. Strikingly but consistent with the synaptic physiology, we found that expression of R704C-NL4 increased the AMPA-induced current in neurons expressing NL4 (Fig. 5B). Thus, in NL4 the R704C mutation appears to truly increase surface AMPAR levels.

Because the R704C-mutants of NL3 and NL4 produced opposing effects on AMPARs in neurons, we hypothesized that the R704C mutation may act via different cellular mechanisms in an isoform–specific manner. To test this hypothesis, we extracellularly tagged NL4-WT and NL4-R704C with HA-epitopes (similar to NL3 in Fig. 4A), and probed if the R704C mutation affected the synaptic distribution of NL4. Interestingly, we did not detect a significant difference in the relative surface expression of NL4-WT and NL4-R704C (Fig. 5C) or in their synaptic localization (assessed via their co-localization with vGluT1; Fig. 5D). These data suggest that unlike in NL3, the R704C mutation in NL4 does not cause increased internalization of NL4 and may thereby stabilize surface AMPARs.

Implications

Our observations have potentially important implications. First, they strengthen the notion that different neuroligins, despite extensive homology spanning their entire sequences, have distinct functions. At the same time, the fact that the R704C substitution in both NL3 and NL4produces a functional change in AMPARs, although in opposite directions, indicates that neuroligins are governed by similar mechanistic principles consistent with their homology.

Second, our results suggest that NL4 like NL3acts in excitatory synapses, but differs functionally from NL3. NL4 is likely also present at glycinergic synapses, ⁷ and may like NL3 function in both excitatory and inhibitory synapses.

Third, our observations demonstrate that by altering NL4 function, the R704C mutation is likely pathogenic, and may act by enhancing synaptic transmission at least in a subset of excitatory synapses.

Fourth, our data suggest that neuroligins – at least NL3 and NL4 – physiologically interact with AMPARs as part of their function in conferring particular properties onto synapses.

How is it possible that the R704C mutation has distinct effects in NL3 vs. NL4, especially if it acts via enhancing an interaction of neuroligins with AMPARs? Our data show that the R704C mutation increases the internalization of NL3 but not of NL4, which may explain its distinct effects on NL3 and NL4 (summarized in Fig. S7), but does not account for the different effects of these neuroligins on AMPARs in general. Although our data thus do not provide a definitive final analysis of all neuroligin function, they do support the overall notion that neuroligins are complex transducers of a presynaptic signal into a postsynaptic regulatory effect that differentially operates in various synapse types.