Astrocytic neuroligin 3 regulates social memory and synaptic plasticity through adenosine signaling in male mice

Abstract

Social memory impairment is a key symptom of many brain disorders, but its underlying mechanisms remain unclear. Neuroligins (NLGs) are a family of cell adhesion molecules essential for synapse development and function and their dysfunctions are linked to neurodevelopmental and neuropsychiatric disorders, including autism and schizophrenia. Although NLGs are extensively studied in neurons, their role in glial cells is poorly understood. Here we show that astrocytic deletion of NLG3 in the ventral hippocampus of adult male mice impairs social memory, attenuates astrocytic Ca²⁺ signals, enhances the expression of EAAT2 and prevents long-term potentiation, and these impairments are rescued by increasing astrocyte activity, reducing EAAT2 function or enhancing adenosine/A2a receptor signaling. This study has revealed an important role of NLG3 in astrocyte function, glutamate homeostasis and social memory and identified the glutamate transporter and adenosine signaling pathway as potential therapeutic strategies to treat brain disorders.

Fig. 1. Astrocytic deletion of NLG3 impairs social memory.

a Strategy for generation of astrocytic NLG3 KO mice. b Significantly reduced NLG3 protein in NLG3 ^f/Y; hGFAP-CreERT2 mice treated with tamoxifen compared to NLG3 ^f/Y; hGFAP-CreERT2 mice without tamoxifen or NLG3 ^f/Y mice with tamoxifen (n = 4 mice for each group, p = 0.000412). c Three-chamber social test consisting of Stage 1, Stage 2 and Stage 3. d Both GFAP-NLG3 WT (n = 10 mice) and GFAP-NLG3 KO (n = 13 mice) mice displayed no preference for either chamber 1 (C1) or chamber 2 (C2) during Stage 1. e Both GFAP-NLG3 WT (n = 10 mice, p < 0.001) and GFAP-NLG3 KO (n = 13 mice, p < 0.001) mice preferred stranger 1 (S1) over empty cage (E) during Stage 2. f GFAP-NLG3 WT (n = 10 mice, p < 0.001), but not GFAP-NLG3 KO mice (n = 13 mice), preferred stranger 2 (S2) over stranger 1 (S1) during Stage 3. g AAV virus constructs and timeline of virus injections and tests. h Significantly reduced NLG3 protein in NLG3 ^f/Y mice injected with GFAP-EGFP-Cre virus compared to those injected with GFAP-EGFP virus (n = 3 mice, p = 0.00923). i AAV virus injections to both ventral and dorsal hippocampus used for (j, k). j Both GFAP-EGFP (n = 8 mice, p < 0.001) and GFAP-EGFP-Cre (n = 9 mice, p < 0.001) injected NLG3 ^f/Y mice preferred S1 over E during Stage 2. k GFAP-EGFP (n = 8 mice, p < 0.001), but not GFAP-EGFP-Cre (n = 9 mice) injected NLG3 ^f/Y mice preferred S2 over S1 (p = 0.00113 for memory index) during Stage 3. Data represent mean ± SEM; two-tailed t test for (b, h), right panel of (f) and (k); two-way ANOVA with Tukey post hoc multiple comparisons for (d, e, j), and left panel of (f, k); **p < 0.01, ***p < 0.001. Source data are provided as a Source Data file.

Fig. 2. Ventral hippocampal astrocytic NLG3 is necessary for social recognition memory.

a AAV virus injections to ventral hippocampus. b Both WT (n = 11 mice, p < 0.001) and NLG3 ^f/Y (n = 8 mice, p < 0.001) mice injected with GFAP-EGFP-Cre virus in ventral hippocampus preferred S1 over E during Stage 2 of three-chamber test. c WT (n = 11 mice, p = 0.004), but not NLG3 ^f/Y (n = 8 mice) mice injected with GFAP-EGFP-Cre virus in ventral hippocampus preferred S2 over S1 during Stage 3 (memory index p = 0.0151). d AAV virus injections to dorsal hippocampus. e Both WT (n = 10 mice, p < 0.001) and NLG3 ^f/Y (n = 15 mice, p < 0.001) mice injected with GFAP-EGFP-Cre virus in dorsal hippocampus preferred S1 over E during Stage 2. f WT (n = 10 mice, p = 0.002) and NLG3 ^f/Y (n = 15 mice, p < 0.001) mice injected with GFAP-EGFP-Cre virus in dorsal hippocampus preferred S2 over S1 during Stage 3. g AAV virus constructs and injection strategy for re-introducing WT NLG3-HA in ventral hippocampal astrocytes of GFAP-NLG3 KO mice. h Increased NLG3 protein in GFAP-NLG3 KO mice expressing NLG3-HA compared to those expressing EGFP (n = 3 mice, p = 0.0382). i AAV virus (same as in g) injections to ventral hippocampus used for (j, k). j Both NLG3-HA (p = 0.001) and EGFP (p < 0.001) expressed GFAP-NLG3 KO mice preferred S1 over E (n = 8 mice for each group) during Stage 2. k NLG3-HA, but not EGFP expressing GFAP-NLG3 KO mice preferred S2 over S1 during Stage 3 (p = 0.002 for sniffing time, p = 0.0106 for memory index, n = 8 mice for each group). Data represent mean ± SEM; two-tailed t test for right panels of (c, f, h, and k); two-way ANOVA with Tukey post hoc multiple comparisons for (b, e, j), and left panels of (c, f, and k); *p < 0.05, **p < 0.01, ***p < 0.001. Source data are provided as a Source Data file.

Fig. 3. Astrocytic deletion of NLG3 impairs hippocampal astrocyte activity.

a, b AAV virus constructs, injection and astrocyte Ca²⁺ imaging in hippocampal slices. c, d Sample traces and summary data of increased GCaMP6f signals in hGFAP-CreERT2 (top; n = 20 cells, 4 mice; p = 0.00203), but not in GFAP-NLG3 KO (bottom; n = 20 cells, 4 mice) astrocytes after HFS. e Reduced activated astrocytes in GFAP-NLG3 KO (n = 10 slices, 4 mice) compared to hGFAP-CreERT2 (n = 11 slices, 4 mice) in response to HFS (p = 0.00000208). f, g Representative traces and summary data of increased GCaMP6f signals in hGFAP-CreERT2 hippocampal astrocytes treated with ACSF (n = 15 cells, 4 mice; p = 0.00732) or AP5 (n = 8 cells, 3 mice; p = 0.0240), but not with PPADS + suramin (n = 15 cells, 4 mice) or MCPG (n = 15 cells, 3 mice) after HFS. h Reduced activated astrocytes in hGFAP-CreERT2 slices treated with PPADS + suramin (n = 6 slices, 3 mice; p < 0.001) and MCPG (n = 5 slices, 3 mice; p < 0.001) compared to those treated with ACSF (n = 6 slices, 3 mice) or AP5 (n = 6 slices, 3 mice) after HFS. i, j AAV virus construct and injections for in vivo fiber photometry Ca²⁺ imaging. k, l Representative traces and summary data of increased astrocytic GCaMP6f signals during social sniffing episodes (pink lines) in GFAP-NLG3 WT (n = 8 recordings, 4 mice; p < 0.001), but not in GFAP-NLG3 KO mice (n = 8 recordings, 4 mice). Data represent mean ± SEM; two-tailed t test for (e); two-tailed paired t test for (d, g); one-way ANOVA with Holm-Sidak test for multiple comparisons for (h), and repeated two-way ANOVA with Holm-Sidak test for multiple comparisons for (l); *p < 0.05, **p < 0.01, ***p < 0.001. Scale bars: 10 μm in left panels of (c); 0.5 dF/F₀ /5 s in right panels of (c, f); 0.2 dF/F₀/60 s in (k). Source data are provided as a Source Data file.

Fig. 4. Enhancing astrocyte activity in ventral hippocampus improves social recognition memory.

a, b AAV hM3D-mCherry virus constructs and injection to ventral hippocampus. c Sample image of viral expression of hM3D-mCherry in ventral hippocampus (repeated 5 times with similar results). Magnified images (d) showing greater >90% hM3D-mCherry expression colocalized with GFAP (e, n = 8 sections, 5 mice) and approximately 50% GFAP+ cells expressing hM3D-mCherry (f, n = 7 sections, 5 mice). g Systemic CNO injection and behavior tests for (h and i). h GFAP-NLG3 KO mice expressing hM3D-mCherry (n = 9 mice, p = 0.004) or mCherry (n = 8 mice, p = 0.001) preferred S1 over E during Stage 2 of three-chamber test. i GFAP-NLG3 KO mice expressing hM3D-mCherry (n = 9 mice, p = 0.010), but not mCherry (n = 8 mice) preferred S2 over S1 during Stage 3 of three-chamber test (p = 0.010 for sniffing time, p = 0.0199 for memory index). j–l Sample images, representative traces and summary data of Ca²⁺ imaging showing that CNO increased astrocytic GCaMP6f signals in hippocampal slices of hM3D-mCherry virus injected GFAP-NLG3 KO mice (DMSO n = 13 cells, 4 mice; CNO n = 13 cells, 4 mice; p = 0.0113). m Increased activated astrocytes in CNO treated hM3D-mCherry virus injected GFAP-NLG3 KO hippocampal slices compared to DMSO treatment (DMSO n = 10 slices, 4 mice; CNO n = 10 slices, 4 mice; p = 0.0000000384). Data represent means ± SEM; two-tailed t test for (l, m), and right panel of (i); two-way ANOVA with Tukey post hoc multiple comparisons for (h), and left panel of (i); Scale bars: 200 μm in panel (c), 50 μm in panel (d), 10 μm in panel (j). Scale bar: 1 dF/F₀/30 s in (k). Source data are provided as a Source Data file.

Fig. 5. Enhanced EAAT2 expression and impaired extracellular glutamate in astrocytic NLG3 KO mice.

a Increased EAAT2 protein level in GFAP-NLG3 KO (n = 6 mice) compared to GFAP-NLG3 WT mice (n = 6 mice, p = 0.0494). b, c Increased EAAT2 fluorescence intensity in total hippocampus (top, n = 10 sections, 3 mice, p = 0.00754) and GFAP+ cells (bottom, p = 0.0000453) in GFAP-NLG3 KO (n = 36 cells, 3 mice) compared to GFAP-NLG3 WT (n = 35 cells, 3 mice) mice. d Reduced EAAT2 protein level in GFAP-NLG3 KO mice virally expressing WT NLG3-HA compared to control EGFP (n = 4 mice, p = 0.0265). e, f Increased surface EAAT2 fluorescence intensity in total hippocampus (n = 10 sections, 3 mice, p = 0.0260) and in GFAP+ cells (n = 40 cells from 3 mice, p = 0.00000871) in GFAP-NLG3 KO compared to GFAP-NLG3 WT mice. g, h AAV virus construct, iGluSnFR expression and normalized data (h) showing increased iGluSnFR signals in response to social sniffing episodes in GFAP-NLG3 WT (n = 5 recordings, 5 mice), but not in GFAP-NLG3 KO mice (n = 4 recordings, 4 mice). i Reduced iGluSnFR signals after initial social sniffing in GFAP-NLG3 KO mice compared to GFAP-NLG3 WT (n = 5 mice for each genotype, p = 0.0437). Data represent mean ± SEM; two-tailed t test for (c, d, f and i); *p < 0.05, **p < 0.01, ***p < 0.001. Scale bar: 100 μm in left (b, e), 10 μm in right (b and e); Source data are provided as a Source Data file.

Fig. 6. Enhanced EAAT2 expression underlies impaired astrocyte activity and memory in astrocytic NLG3 KO mice.

a, b Increased GCaMP6f fluorescence in both GFAP-NLG3 WT (n = 26 cells, 3 mice, p = 0.00000169) and KO (n = 26 cells, 3 mice, p = 0.00198) astrocytes after (post) glutamate treatment compared to before (pre). c Activated astrocytes in GFAP-NLG3 WT (n = 6 slices from 3 mice) and KO (n = 6 slices from 3 mice) slices treated with glutamate. d, e Increased GCaMP6f signals in response to HFS in GFAP-NLG3 WT (n = 24 cells, 3 mice, p = 0.00540) and KO (n = 24 cells, 3 mice, p = 0.00000000212) astrocytes treated with TBOA, but not in GFAP-NLG3 KO astrocytes treated with ACSF (n = 28 cells, 4 mice). f Activated astrocytes in TFB-TBOA treated GFAP-NLG3 KO (n = 6 slices, 3 mice, p < 0.001) and WT slices (n = 5 slices, 3 mice), compared to ACSF treated GFAP-NLG3 KO (n = 6 slices, 3 mice), g Local TFB-TBOA injections to ventral hippocampus before behavior tests. h GFAP-NLG3 KO mice treated with saline (n = 8 mice, p < 0.001) or TFB-TBOA (n = 8 mice, p < 0.001) preferred S1 over E during Stage 2  of three-chamber test. i GFAP-NLG3 KO mice treated with TFB-TBOA (n = 8 mice), but not with saline (n = 8 mice), preferred S2 over S1 during Stage 3 of three-chamber test (sniffing time p < 0.001, memory index p < 0.001). Data represent mean ± SEM; two-tailed paired t test for (b and e); two-tailed t test for (c) and right panel of (i); one-way ANOVA test with Holm-Sidak test for (f); two-way ANOVA with Tukey post hoc multiple comparisons for (h) and left panel of (i); **p < 0.01, ***p < 0.001. Scale bar: 0.5 dF/F₀ /5 s in (a and d). Source data are provided as a Source Data file.

Fig. 7. Normal basal synaptic transmission but impaired LTP in astrocytic NLG3 KO mice.

a Input/output curve of fEPSPs in GFAP-NLG3 KO (n = 18 slices, 8 mice) and GFAP-NLG3 WT mice (n = 13 slices, 6 mice). b Input/output curve of fEPSPs in NLG3 ^f/Y mice injected with GFAP-EGFP-Cre (n = 7 slices, 5 mice) or GFAP-EGFP virus (n = 8 slices, 5 mice). c Paired pulse ratio (PPF) in GFAP-NLG3 KO (n = 10 slices, 6 mice) and GFAP-NLG3 WT mice (n = 11 slices, 6 mice). d PPF in NLG3 ^f/Y mice injected with GFAP-EGFP-Cre (n = 9 slices, 5 mice) or GFAP-EGFP virus (n = 9 slices, 5 mice). e mEPSCs in GFAP-NLG3 KO (n = 9 cells, 3 mice) and GFAP-NLG3 WT mice (n = 10 cells, 3 mice). f AMPAR/NMDAR EPSC ratio in GFAP-NLG3 KO (n = 8 cells, 3 mice) and GFAP-NLG3 WT mice (n = 8 cells, 3 mice). g Impaired LTP in GFAP-NLG3 KO (n = 5 slices, 3 mice) compared to GFAP-NLG3 WT mice (n = 5 slices, 3 mice, p = 0.00000249). h Impaired LTP in NLG3 ^f/Y mice injected with GFAP-EGFP-Cre (n = 6 slices, 4 mice; p = 0.0175) compared to those injected with GFAP-EGFP virus. i Impaired dorsal hippocampal LTP in GFAP-NLG3 KO (n = 5 slices, 3 mice) compared to GFAP-NLG3 WT mice (n = 5 slices, 3 mice, p = 0.00505). j 50 nM TFB-TBOA (p = 0.009) restored LTP in GFAP-NLG3 KO mice whereas 150 nM TFB-TBOA did not (n = 5 slices, 3 mice for each group). Data represent mean ± SEM; two-tailed t test for (e, f, g, h, and i); repeated two-way ANOVA test with Holm-Sidak test for multiple comparisons for (c and d); One-way ANOVA for (j); *p < 0.05, *p < 0.01, ***p < 0.001. Scale bars: 0.5 mV/25 ms in (a and b), 30 pA/10 s in (e), 50 pA/50 ms in (f), 0.25 mV/50 ms in (g–j). Source data are provided as a Source Data file.

Fig. 8. Activation of adenosine signaling restores LTP in astrocytic NLG3 KO mice.

a Adenosine restored LTP (ACSF n = 5 slices, 3 mice; adenosine n = 7 slices, 4 mice; p = 0.00394). b D-serine had no effect on LTP (ACSF n = 5 slices, 3 mice; D-serine n = 6 slices, 3 mice). c A2a agonist CGS 21680, not A1 agonist CPA, restored LTP (ACSF n = 5 slices, 3 mice; CPA n = 5 slices, 3 mice; CGS 21680 n = 5 slices, 3 mice; ACSF vs CGS 21680, p = 0.018; CPA vs CGS 21680, p = 0.014). d LTP rescue by adenosine blocked by A2a antagonist SCH 58261 (ACSF n = 5 slices, 3 mice; adenosine + SCH 58261 n = 5 slices, 3 mice). e ATP restored LTP (ACSF n = 5 slices, 3 mice; ATP n = 5 slices, 3 mice; p = 0.0139). f LTP rescue by ATP blocked by A2a antagonist SCH 58261 (ATP n = 5 slices, 3 mice; ATP + SCH 58261 n = 5 slices, 3 mice; p = 0.0311). g ATPγS or bzATP had no effects on LTP (ACSF n = 7 slices, 3 mice; bzATP n = 6 slice, 3 mice; ATPγS n = 5 slices, 3 mice). h Adenosine had no effect on LTP in WT mice (ACSF n = 5 slices, 3 mice, adenosine n = 5 slices, 3 mice). i A2a antagonist SCH 58261 blocked LTP in WT mice (ACSF n = 5 slices, 3 mice; SCH 58261 n = 5 slices, 3 mice; p = 0.000465). j A2a agonist CGS 21680 had no effect on LTP in WT mice (ACSF n = 5 slices, 3 mice; CGS 21680 n = 5 slices, 3 mice). Data represent mean ± SEM; two-tailed t test for (a, b, d, e, f, h, i, and j); one-way ANOVA with Holm-Sidak test for multiple comparisons for right panels of (c and g); *p < 0.05, **p < 0.01, ***p < 0.001. Scale bars: 0.25 mV/50 ms. Source data are provided as a Source Data file.

Fig. 9. Adenosine/A2a activation improves social memory.

a, b Microdialysis measurement of ventral hippocampal adenosine. c Increased extracellular adenosine in GFAP-NLG3 WT (n = 5 mice, p = 0.0144), not in GFAP-NLG3 KO (n = 5 mice), after social stimulation. RFU: Relative Fluorescence Units. ΔRFU: Post-social RFU minus Pre-social RFU. d AAV virus construct and expression of GRAB_ado. e Increased adenosine signals after sniffings in WT (n = 8 recordings, 4 mice), not in GFAP-NLG3 KO mice (n = 8 recordings, 4 mice). f Reduced GRAB_ado signals after sniffings in GFAP-NLG3 KO mice (n = 8 recordings, 4 mice, p = 0.0325). g Systemic CGS 21680 injections. h GFAP-NLG3 KO mice treated with saline (n = 6 mice, p < 0.001) or CGS 21680 (n = 10 mice, p < 0.001) preferred S1 over E. i GFAP-NLG3 KO mice treated with CGS 21680 (n = 10 mice, p < 0.001), not with saline (n = 6 mice), preferred S2 over S1 (memory index p = 0.0111). j, k AAV virus construct and injection for ventral hippocampal neuronal A2a deletion. l CaMKIIα-EGFP-2A-Cre expression in ventral hippocampus and CA1 neurons (repeated 5 times). m WT and A2a ^f/f mice injected with CaMKIIα-EGFP-Cre virus preferred S1 over E (n = 7 mice, p < 0.001). n CaMKIIα-EGFP-Cre virus injected WT (p < 0.001), not A2a ^f/f mice, preferred S2 over S1 (n = 7 mice, memory index p = 0.000161). o Reduced A2a receptor protein in A2a ^f/f mice injected with CaMKIIα-EGFP-Cre virus (n = 5 mice, p = 0.00174). p Impaired LTP in A2a ^f/f mice injected with CaMKIIα-EGFP-Cre virus (n = 5 slices, 3 mice, p = 0.00117). Data represent mean ± SEM; two-tailed t test for (c, f, o), and right panels of (i, n, and p); two-way ANOVA with Tukey post hoc multiple comparisons for (h, m), and left panels of (i and n); *p < 0.05, **p < 0.01, ***p < 0.001. Scale bars: 200 μm in left panel of (l), 50 μm in right panel of (l), 0.25 mV/50 ms in (p). Source data are provided as a Source Data file.