LARGE protein drives activity-induced homeostatic resetting

Abstract

In the brain, memory can be coded as relative differences in synaptic strength produced by Hebbian plasticity [e.g., long-term potentiation (LTP)]. However, changes in neuronal activity, including the saturation of synaptic strength by the positive-feedback nature of Hebbian plasticity, could deteriorate the encoded memory. Homeostatic plasticity is thought to contribute to the stability of the encoded memory by maintaining the relative differences in synaptic strength against persistent destabilizing changes in neuronal activity. However, it remains unclear how and when these two types of plasticity work together in the context of memory. Here, we have demonstrated that LARGE, a protein associated with intellectual disability, drives homeostatic resetting several hours after LTP by down-regulating AMPA-receptor trafficking via the Golgi apparatus. LARGE deficiency impairs long-term memory formation in mice. Our study reveals a potential molecular mechanism underlying the stability of memory mediated by cross-talk between Hebbian and homeostatic plasticity.

Fig. 1.. cLTP induced the decrease in surface GluA1 accompanied by the increase in LARGE expression.

(A) Biochemical and confocal analyses in the indicated time point after glycine-induced cLTP in hippocampal neurons (C57BL/6 mice) at DIV 14. qPCR, quantitative polymerase chain reaction. (B) An increase in surface GluA1 was observed 20 min after glycine treatment (Gly, 200 μΜ) and stayed until 3 hours (h) (n = 3; one-way ANOVA; *P < 0.05, **P < 0.01, and ***P < 0.001), which was reduced at 6 and 9 hours (n = 3; one-way ANOVA; #P < 0.05 and ##P < 0.01). LARGE expression increased at 6 and 9 hours (n = 3; one-way ANOVA; §§§P < 0.001). No significant difference in total GluA1 (n = 3, one-way ANOVA; F_5,12 = 0.1134, P = 0.8179). (C) Colocalization of LARGE with GluA1 increased 6 hours after cLTP (n = 19, 19, and 19; one-way ANOVA; F_2,54 = 34.303, ***P < 0.001). (D) Hippocampal neurons were transfected with plasmids expressing Scramble shRNA, LARGE shRNA, or LARGE rescue (LARGE shRNA/LARGE cDNA) with EGFP for 1 day at DIV 13. LARGE knockdown (KD) decreased the colocalization of GluA1 with LARGE, which was reversed by LARGE rescue (upper images; n = 10, 10, and 10; **P < 0.01). An elevated colocalization of GluA1 with LARGE 6 hours after cLTP was reduced in LARGE KD neuron but was restored by LARGE rescue (bottom images; n = 10, 10, and 10; ###P < 0.001). (E) After transfection, the surface GluA1 intensity on the EGFP-positive dendrite was analyzed separately in distal and proximal dendrites. LARGE KD increased the surface GluA1 level, whereas this phenomenon was reversed by LARGE rescue (upper images; n = 10, 10, and 10; distal dendrite, **P < 0.01; proximal dendrite, §§§P < 0.001). Overexpression of LARGE by rescue attenuated the increase in GluA1 1 hour after cLTP (bottom images; n = 10, 10, and 10; distal dendrite, ###P < 0.001; proximal dendrite, $$$P < 0.001). Scale bars, 30 μm (whole image) and 10 μm (dendrite).

Fig. 2.. LARGE returned the cLTP-induced increase in surface GluA1 to basal levels.

(A) Schema illustrating the lentiviral vector expressing IPTG-inducible Scramble shRNA or LARGE shRNA with LacI that was packaged in HEK293FT cells. (B) Schema showing the experimental timeline. The cultured hippocampal neurons (C57BL/6 mice) were transduced with a lentivirus for 5 days at DIV 9. At DIV 14, IPTG (500 μM) was applied to express Scramble shRNA or LARGE shRNA, and glycine (200 μM) was used to induce cLTP. The biochemical analysis was performed in the indicated time point. WB, Western blot; m, minutes. (C) In neurons expressing IPTG-inducible Scramble shRNA, glycine and IPTG treatment induced the increase in surface GluA1, which was reduced at 6 and 9 hours (n = 3; one-way ANOVA; **P < 0.01 and ***P < 0.001, compared to that before treatment). LARGE expression increased at 6 and 9 hours (n = 3; one-way ANOVA; §§§P < 0.001, compared to that before treatment). There was no significant difference in total GluA1 expression (n = 3; one-way ANOVA; F_5,12 = 0.1682, P = 0.7597; n.s., no significance). (D) In neurons expressing IPTG-inducible LARGE shRNA, the glycine-induced increase in LARGE expression was temporarily suppressed by IPTG treatment at 6 and 9 hours (n = 3; one-way ANOVA; F_5,12 = 9.763, §§§P < 0.001, compared to that before treatment). The increase in surface GluA1 was maintained until 9 hours (n = 3; one-way ANOVA F_5,12 = 6.307, **P < 0.01 and ***P < 0.001, compared to that before treatment). There was no significant difference in total GluA1 expression (n = 3; one-way ANOVA; F_5,12 = 0.1271, P = 0.9567). All experiments were performed in triplicate and repeated three times with similar results. h, hours.

Fig. 3.. cLTP induced the homeostatic resetting of synaptic weights and return of increased neuronal firing rate.

(A) Schema showing the electrophysiological analyses such as whole-cell patch clamp and MEA after glycine-induced cLTP in cultured hippocampal neurons (C57BL/6 mice) at DIV 14. (B) After cLTP (Gly, 200 μΜ), the mEPSC amplitude increased around 20 min and 3 hours. The increase was reduced around 6 and 9 hours [n = 18, 19, 17, 17, and 17; two-tailed t test; **P < 0.01, compared to that before glycine (Cont); #P < 0.05 and ##P < 0.01, compared to that around 20 min and 3 hours]. (C) MEA analysis demonstrated the increase in neuronal firing by cLTP induction (Gly, 200 μΜ), which was maintained until 3 hours (n = 13; two-tailed t test; **P < 0.01) but decreased 6 and 9 hours after cLTP compared to that at 20 min and 3 hours (n = 13; two-tailed t test; #P < 0.05). Neuronal firing from all 64 channels of a MEA single-well plate was presented and quantified. All experiments were performed in triplicate and repeated three times with similar results. h, hours.

Fig. 4.. The pool of AMPA-R associated with LARGE at the Golgi increases during homeostatic scaling-down.

(A) Schema showing biochemical and confocal analyses after Bicu (20 μΜ, 48 hours) or TTX (1 μM, 48 hours) treatment in cultured hippocampal neurons (C57BL/6 mice). (B) Series of confocal microscopy images of cultured hippocampal neurons double-stained with LARGE and GluA1 48 hours after Bicu treatment at DIV 12. The colocalization of GluA1 and LARGE in the perinuclear region increased in response to Bicu (n = 9 and 10; two-tailed t test; t₁₇ = −4.728; ***P < 0.001, compared to that before treatment). (C) The cultured hippocampal neurons were transfected with plasmids expressing Scramble shRNA, LARGE shRNA, or LARGE rescue (LARGE shRNA/LARGE cDNA) with EGFP for 1 day at DIV 11. LARGE knockdown (KD) decreased the colocalization of GluA1 with LARGE, which was reversed by LARGE rescue (upper images; n = 10, 10, and 10; *P < 0.05 and **P < 0.01, compared to Scramble shRNA/before Bicu). An elevated colocalization of GluA1 with LARGE 48 hours after Bicu was reduced in the LARGE knockdown (KD) neuron but was restored by LARGE rescue (bottom images; n = 10, 10, and 10; ##P < 0.01, compared to Scramble shRNA/Bicu 48 hours). (D) The binding of LARGE to GluA1 increased 48 hours after Bicu treatment in cultured hippocampal neurons at DIV 12, as demonstrated by the increased coimmunoprecipitation of LARGE and GluA1 (n = 3; two-tailed t test; **P < 0.01, compared to that before treatment). (E) Density gradient fractionation of subcellular organelles revealed increases in the relative amounts of GluA1 and LARGE in the Golgi fraction following treatment with Bicu for 48 hours in cultured hippocampal neurons at DIV 12 (n = 4; two-tailed t test; *P < 0.05 and **P < 0.01). All experiments were performed in triplicate and repeated three times with similar results. h, hours.

Fig. 5.. LARGE is necessary for homeostatic scaling-down.

(A) Biochemical and electrophysiological analyses in hippocampal neurons (C57BL/6 mice) treated with Bicu or TTX. DIC, differential interference contrast. (B) Neurons were transduced with an AAV expressing Scramble shRNA, LARGE shRNA, or LARGE shRNA with wt LARGE cDNA (LARGE shRNA + wt rescue) at DIV 7, followed with or without Bicu or TTX treatment at DIV 12 for 48 hours. Pictures show neurons 5 days after AAV transduction at DIV 7. Scale bar, 60 μm. LARGE KD increased surface GluA1, which was reversed by rescue (n = 3; one-way ANOVA; *P < 0.05, **P < 0.01, and ***P < 0.001). A Bicu-induced decrease in surface GluA1 was blocked by LARGE KD and returned by rescue (n = 3; one-way ANOVA; §P < 0.05 and §§P < 0.01). The TTX-induced increase was induced in all groups (n = 3; one-way ANOVA; #P < 0.05, ##P < 0.01, and ###P < 0.001). (C) Neurons were transfected with plasmids carrying Scramble shRNA, LARGE shRNA, or LARGE shRNA with wt LARGE cDNA (LARGE shRNA + wt rescue) for 1 day at DIV 11, followed by Bicu or TTX for 48 hours, and recorded for mEPSC. (D) EGFP-positive and EGFP-negative neurons at DIV 14 in a whole-cell patch clamp. Scale bar, 10 μm. (E) Current traces, cumulative plots, and scattered plots from mEPSC in neurons. Bicu or TTX induced scaling-down or scaling-up in the control group (1343, 1340, and 1324 events from n = 17, 16, and 13 neurons, respectively; one-way ANOVA; *P < 0.05 and **P < 0.01; Cum. Pro., cumulative probability; upper panel). A LARGE KD–mediated increase in mEPSC amplitude was affected by TTX but not by Bicu (1330, 1321, and 1311 events from n = 18, 15, and 12 neurons; one-way ANOVA; *P < 0.05; middle panel). The mEPSC amplitude in the rescue group remained at the decreased level with or without Bicu but was increased by TTX (1297, 1300, and 1305 events from n = 15, 13, and 13 neurons, respectively; *P < 0.05; bottom panel). No significant changes in the frequency of mEPSC among all groups.

Fig. 6.. TBS-LTP in the hippocampal slice induced homeostatic resetting through LARGE accumulation.

(A) After LTP induction with TBS, hippocampal slices were incubated for 1 hour, 3 hours, and 6 hours, followed by PSD fractionation and immunoblotting. (B) The brain was microdissected (400 μm) 2 weeks after injecting the hippocampal CA1 region of 3-week-old male mice with a lentivirus expressing IPTG-inducible Scramble shRNA or LARGE shRNA with LacI. Hippocampal slices were transferred to a recovery chamber containing ACSF with IPTG (500 μM). After recovery for 1 hour, the slices were transferred to a recording chamber containing IPTG. LTP was induced by TBS 20 min after extracellular recordings, which was monitored for 1 hour. Stimulated slices were incubated at the indicated time points and used for mRNA extraction and PSD fractionation. (C) fEPSP showed potentiation of synapse and was maintained for 60 min, confirming the induction and maintenance of LTP. (D) In hippocampal slices expressing IPTG-inducible Scramble shRNA, TBS increased phosphorylation on serine-831 (pS831) of GluA1 at 1 hour and 3 hours in PSD fraction (n = 3; one-way ANOVA; *P < 0.05 and **P < 0.01) but decreased it at 6 hours (n = 3; one-way ANOVA; #P < 0.05). Total LARGE expression increased at 3 hours after TBS (n = 3; one-way ANOVA; §P < 0.05 and §§§P < 0.001). There was no significant difference in total GluA1 expression (n = 3; one-way ANOVA; F_3,8 = 0.1412, P = 0.7365). (E) In hippocampal slices expressing IPTG-inducible LARGE shRNA, accumulation of LARGE following TBS-LTP was blocked by the IPTG incubation (n = 3; one-way ANOVA; F_3,8 = 0.7164, P = 0.5694). The increase in GluA1 pS831 in PSD fraction was maintained until 6 hours (n = 3; one-way ANOVA; **P < 0.01 and ***P < 0.001). There was no significant change in total GluA1 expression (n = 3; one-way ANOVA; F_3,8 = 0.2551, P = 0.6291). All experiments were performed in triplicate. h, hours.

Fig. 7.. LARGE deficiency impairs hippocampus-dependent long-term memory but not short-term memory.

(A) Schema showing the timeline of a series of behavioral tasks 4 weeks after AAV injection. (B) During the “exposure training,” mice were given the opportunity to explore the start arm and the other arm over five trials, each lasting 2 min, while the novel arm was blocked. In the novelty preference test, mice were allowed to explore all three arms—start arm, other arm, and the previously restricted novel arm—for 2 min. The interval between exposure trials, as well as the delay before the novelty preference test, was set at either 1 min (1-min ITI) or 24 hours (24-hour ITI). LARGE knockdown (KD) mice exhibited a preference in the short term (1 min; n = 12,12, Mann-Whitney rank-sum test; entries, T = 168, U = 54, P = 0.308; time, T = 141, U = 57, P = 0.601) but not in the long term (24 h; n = 12 and 12, two-tailed t test; entries, t₂₂ = 2.727; **P < 0.01; time, t₂₂ = 2.483, *P < 0.05). The novel arm preference was calculated as [Novel/(Novel + Other)] × 100%. (C) In the novel object recognition test, control and LARGE KD mice similarly explored two identical objects during training (n = 19 and 20; two-tailed t test; t₃₇ = 0.556, P = 0.582). Both groups exhibited similar preferences during a short-term memory test (5 min; n = 19 and 20; Mann-Whitney rank-sum test, T = 393, U = 177, P = 0.725). In a long-term memory test, LARGE KD mice exhibited no preference for the novel object (24 hours; n = 19 and 20; two-tailed t test; t₃₇ = 3.773, ***P < 0.001). The novel object preference was calculated as [Novel/(Novel + Familiar)] × 100%.

Fig. 8.. Schema of our working model for the regulation of AMPA-R trafficking by LARGE during homeostatic resetting in response to the increase in neuronal activity by LTP.