CKO of Nlg3 in the MS impairs social memory and reduces sleep. Sleep disturbances and social deficits are commonly observed in patients with ASD (1). The MS has been implicated in regulating sleep-wake and social behaviors (10, 11). To investigate the potential role(s) of autism-associated NLG3 in the MS related to sleep-wake and social behaviors, we generated CKO mice by injecting adeno-associated virus (pAAV-hSyn-Cre) into the MS of Nlg3fl/fl mice (Figure 1, A and B), resulting in the specific deletion of Nlg3 in the MS (Nlg3-CKO mice). Six weeks after virus injection, we assessed the mRNA and protein levels of NLG3 using FISH, real-time reverse transcription polymerase chain reaction (RT-PCR), and Western blotting, respectively. Both the mRNA and protein levels of NLG3 were significantly reduced in the MS of Nlg3-CKO mice (Figure 1, C–E, and Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/JCI176770DS1).

Conditional knockout of Nlg3 in the MS impairs social memory and reduces sleep. (A) Schematic of the experimental procedure. (B) pAAV-hSyn-Cre-WPRE was injected into the MS of Nlg3fl/fl mice to conditionally knock out Nlg3 (Nlg3-CKO mice). (C–E) Representative FISH images (C) and quantification of NLG3 mRNA and protein in MS between control and Nlg3-CKO mice 6 weeks after virus injection (D and E, mean ± SEM, n = 5 mice). Scale bar: 200 μm. (F–H) Representative heatmaps (F) and quantification of sniffing time and sociability index between control (n = 10 mice) and Nlg3-CKO mice (n = 13 mice) during the sociability test (G, F[1,21] = 5.71; and H, P = 0.483). In this and the following violin plots, data are presented as median (red line) with 25th and 75th percentiles (dashed line). (I–K) Representative heatmaps (I) and quantification of sniffing time and social novelty index between control and Nlg3-CKO mice during the social novelty test (J, F[1,21] = 76.6; and K). (L) Twenty-four hours of continuous EEG spectrogram, EMG trace, and brain states (color-coded). ZT, zeitgeber time. (M–O) The graphs illustrate the average (mean ± SEM) percentages of wake, NREM, or REM sleep during the day and night for control (n = 10 mice) and Nlg3-CKO mice (n = 10 mice). (P–R) Nlg3-CKO mice exhibited significantly more wakefulness (F[1,18] = 10.63), significantly less NREM (F[1,18] = 10.88), and a nonsignificant difference in REM sleep (F[1,18] = 0.403, P = 0.534) during the day and night. *P < 0.05; **P < 0.01; ***P < 0.001; ###P < 0.001 by 2-tailed Mann-Whitney test (D, E, H, and K) or 2-way ANOVA with Bonferroni’s post hoc test (G, J, and P–R). NS, not significant. See the Supporting Data Values file for statistical details.

Social behavior and social memory were assessed in modified 3-chamber tests of sociability and social novelty, respectively. Sociability was initially tested for the subject mouse in a partner-object trial, where it was paired with a partner (stranger 1, S1) in one cage and an object (O) in the other cage (Figure 1F). Both control and Nlg3-CKO mice spent more time sniffing S1 compared with the object (Figure 1, F and G). Nlg3-CKO mice spent significantly less time sniffing S1 in comparison with the control mice (Figure 1G). However, both groups of mice exhibited a similar sociability index (Figure 1H), suggesting normal sociability in Nlg3-CKO mice. In the social novelty test, another unfamiliar mouse was introduced as a novel social partner (stranger 2, S2) to replace the object (Figure 1I). The control mice showed a preference for S2, with a normal social novelty index, whereas Nlg3-CKO mice displayed no clear preference for S2 and had a significantly lower social novelty index (Figure 1, J and K). Nlg3-CKO and control mice travelled similar distances in 3-chamber tests (Supplemental Figure 2A). These findings suggest that NLG3 deficiency specifically in the MS results in an impairment of social memory.

To assess sleep-wake amounts in Nlg3-CKO mice, we implanted cortical electroencephalographic (EEG) and nuchal electromyographic (EMG) electrodes, followed by conducting 24-hour baseline home-cage sleep recordings (Figure 1L). Nlg3-CKO mice exhibited a notable reduction in NREM and a partial decrease in rapid eye movement (REM) sleep, along with a complementary increase in wakefulness as compared with the control mice (Figure 1, M–R). Thus, social memory deficits and sleep disturbance coexisted in Nlg3-CKO mice.

To investigate whether NLG3 deletion in the MS influences social interaction, sucrose preference, basal locomotion, anxiety-like behaviors, and cognitive ability for novel objects, we performed a battery of behavioral tests. In the social interaction test, Nlg3-CKO mice spent significantly less time in anogenital sniffing and pursuit than the control mice (Supplemental Figure 2, B and C). In the sucrose preference test, Nlg3-CKO and control mice exhibited similar sucrose preferences (Supplemental Figure 2D). Additionally, Nlg3-CKO mice showed normal locomotor activity and anxiety-like behaviors in the open field and elevated plus maze tests (Supplemental Figure 2, E–M), respectively. In the novel object recognition test, which assesses novel object recognition memory, both Nlg3-CKO and control mice showed comparable preference for the novel object (Supplemental Figure 2, N–S). These findings suggest that Nlg3-CKO mice have normal levels of locomotion, anxiety-like behaviors, sucrose preference, and object recognition memory, but deficits in social interaction.

Together, these findings reveal that CKO of Nlg3 in the MS impairs social interaction behaviors and reduces sleep.

Nlg3 knockout in the MS causes hyperactivity of MSGABA neurons during social avoidance and wakefulness. c-Fos has been widely recognized as a reliable indicator of neuronal activity (19). In order to explore the potential activation of neurons in the MS resulting from the CKO of Nlg3, we conducted a FISH experiment for c-fos mRNA. The levels of c-fos mRNA in the MS of Nlg3-CKO mice were found to be significantly higher than those in the control mice (Figure 2, A–C). Additionally, we performed double FISH of c-Fos with Vgat. The majority of c-fos–expressing neurons in the MS of Nlg3-CKO mice coexpressed Vgat, whereas a considerably smaller proportion of neurons in the MS of control mice expressed Vgat (Figure 2D). These results show that the majority of active neurons are GABAergic neurons following Nlg3 CKO in the MS.

Nlg3 knockout in the MS causes hyperactivity of MSGABA neurons during social avoidance and wakefulness. (A–C) Representative FISH images (A and B) and quantification of c-Fos expression in MS between control and Nlg3-CKO mice (C, n = 5 mice). Scale bars: 200 μm (left) and 20 μm (right). (D) Quantification of the overlap between c-Fos+ cells and Vgat (n = 5 mice). (E and F) Schematic diagram for experimental protocols and optrode recording. (G) Representative images of the MS from an Nlg3-CKO mouse. Scale bars: 200 μm (left) and 20 μm (right). (H) Optogenetic tagging and identification of MSGABA neuron. Waveforms (top), a representative raw trace (middle), and raster plots (down) from an identified MSGABA neuron. (I–K) Example recording of spontaneous spikes (I) and mean firing rates (J, shading represents ±SEM) and quantification of the identified MSGABA neurons from control and Nlg3-CKO mice (K, n = 38 units from 7 mice, F[1,74] = 157.9) during the approach and avoidance of a novel mouse. (L and M) Representative trace of firing rate from a MSGABA neuron (L) and quantification of the identified MSGABA neurons in control mice and Nlg3-CKO mice (M, n = 48 units from 9 mice, F[2,188] = 59.83) during wakefulness, NREM, and REM sleep. (N) Firing rate modulation of identified MSGABA neuron from both control and Nlg3-CKO mice. W, wake; R, REM; NR, NREM. (O) Mean firing rates of identified MSGABA neurons from both mice during different brain state transitions. Shading represents ±SEM. **P < 0.01; ***P < 0.001; ###P < 0.001 by 2-tailed Mann-Whitney test (C and N) or 2-way ANOVA with Bonferroni’s post hoc test (K and M). NS, not significant.

We next wondered how these neurons behaved during social and sleep-wake behaviors. To address this question, we first labeled these neurons with channelrhodopsin-2 (ChR2) by injecting an AAV vector (AAV2-GAD67-hChR2-eGFP) into the MS of both control and Nlg3-CKO mice (Figure 2, E–G, and Supplemental Figure 3A). As expected, AAV expressing ChR2 fused with the enhanced green fluorescent protein (eGFP) selectively labeled GABAergic neurons in the MS (Figure 2G). Subsequently, we performed optogenetic tagging and optrode recordings (11, 20) to monitor ChR2-expressing GABAergic neurons in the MS of control and Nlg3-CKO mice during social behaviors and the sleep-wakefulness cycles (Figure 2, E–G). Optogenetic stimulations (473 nm, 1-ms pulse duration, 10 Hz × 1 second) were applied to screen and define the ChR2-expressing GABAergic neurons according to the following 3 criteria: (a) laser stimulations should evoke spikes reliably (>70% of firing rate), (b) laser-evoked spikes should have short latencies (1–3 ms) and low jitters (<3 ms) to rule out network implication, and (c) laser-evoked and spontaneous spikes should display similar waveforms (Figure 2H and Supplemental Figure 3, B–D).

To monitor the activity of MSGABA neurons in both Nlg3-CKO and control mice during object and social exploration, we recorded MSGABA neuron activity while the subject mouse was approaching and avoiding a novel inanimate object, a novel mouse, as well as the same, now familiar mouse. Notably, the activity in MSGABA neurons of Nlg3-CKO mice decreased significantly when approaching a novel mouse, but increased remarkably when avoiding the same mouse (Figure 2, I–K). No significant changes were detected in MSGABA neuron activity of control mice during the approach and avoidance of a novel mouse. Additionally, the firing rate of MSGABA neurons in Nlg3-CKO mice was significantly higher when the mice were in the “avoidance zone” than in the “approach zone,” and also higher than that of control mice (Figure 2K). However, the activity of MSGABA neurons did not change appreciably in both Nlg3-CKO and control mice when they approached and avoided a novel object or the familiar mouse (Supplemental Figure 4, A–F). Notably, the firing rate of MSGABA neurons in Nlg3-CKO mice was found to be higher than that of control mice (Supplemental Figure 4, B and E). Moreover, no significant changes were observed in the firing rate of unidentified MS neurons from Nlg3-CKO mice during social approach-avoidance of a novel mouse (Supplemental Figure 4, G–I). These findings suggest that MSGABA neurons of Nlg3-CKO mice exhibited increased activities in response to the avoidance of a novel mouse.

Subsequently, we recorded the activity of these MSGABA neurons in both mice across sleep and wakefulness (Figure 2L). In Nlg3-CKO mice, the average firing rate of MSGABA neurons was considerably higher during wakefulness in comparison with NREM and REM sleep (Figure 2, L and M). Furthermore, the firing rate of these neurons in Nlg3-CKO mice was significantly higher than that in control mice during wakefulness, NREM, and REM sleep (Figure 2M). In contrast with the MSGABA neurons in Nlg3-CKO mice, the MSGABA neurons from control mice exhibited similar mean firing rates across awake, NREM, and REM states (Figure 2, L and M).

We further quantified the brain-state preference of recorded neurons by calculating their NREM-wake modulation [(RNREM – Rwake)/(RNREM + Rwake)], where R represents the average firing rate within each brain state and REM-wake modulation [(RREM – Rwake)/(RREM + Rwake)] (Figure 2N). In comparison with the MSGABA neurons from control mice, the MSGABA neurons in Nlg3-CKO mice exhibited reduced NREM-wake and REM-wake modulations. Additional analysis was conducted on MSGABA neurons in both mice during the transitions between 2 distinct brain states. Specifically, during transitions from NREM or REM sleep to wakefulness in Nlg3-CKO mice, the averaged firing rate of these neurons increased sharply, whereas the high firing rate of these neurons gradually decreased during transitions from wakefulness to NREM sleep (Figure 2O). In contrast, no such activity change was observed in MSGABA neurons from control mice. Therefore, the increased activity of MSGABA neurons from Nlg3-CKO mice during wakefulness may contribute to the reduced NREM sleep and increased wakefulness.

Taken together, these results indicate that Nlg3 CKO in the MS leads to hyperactivity of MSGABA neurons during social avoidance and wakefulness, suggesting a potential relationship between the increased MSGABA neuron activity and social memory impairments as well as sleep loss.

Activation of MSGABA neurons induces social memory deficits and sleep loss in C57BL/6J mice. The above correlative results prompted us to explore the potential causal roles of hyperactivity in MSGABA neurons in regulating social behavior and sleep/wakefulness. Subsequently, we investigated whether the hyperactivity of MSGABA neurons could be responsible for the observed social memory impairment and sleep loss (Figure 3A). In C57BL/6J mice, we selectively expressed ChR2 in MSGABA neurons by injecting AAV2-GAD67-hChR2-eGFP (hereafter referred to as MSGAD67-ChR2 mice, Supplemental Figure 5A). The MSGABA neurons were optogenetically activated using a blue laser (473 nm) at 10 Hz (Figure 3, B–D). During the sociability test, the consistent activation of MSGABA neurons significantly reduced the sniffing time on S1, but did not affect the sociability index in MSGAD67-ChR2 mice (Figure 3, E–G). During the social novelty test, the persistent activation of MSGABA neurons led to a significant decrease in sniffing time on S2 and social novelty index in MSGAD67-ChR2 mice (Figure 3, H–J). In the MSGAD67-eGFP control mice, in which AAV2-GAD67-eGFP was injected into the MS of C57BL/6J mice, the same blue laser stimulation had no effect on sociability and social memory (Supplemental Figure 5, B–G). These results indicate that optogenetic activation of MSGABA neurons causes impairments in social memory.

Activating MSGABA neurons induces social memory deficits and sleep loss in C57BL/6J mice. (A) Schematic diagram of the experimental procedure. (B) AAV2-GAD67-hChR2-eGFP was injected into the MS and blue laser light (473 nm, 10 Hz) was applied to the MS. (C) Representative image showing selective transduction of hChR2-eGFP in the MS where an optical fiber is located. Scale bar: 200 μm. (D) Representative images showing hChR2-eGFP–expressing neurons colocalized with Vgat. Scale bar: 20 μm. (E–G) Representative heatmaps (E) and quantification of sniffing time (F; n = 12 mice, F[2,33] = 5.811) and sociability index (G, F[1.26,13.9] = 2.692) in MSGAD67-ChR2 mice during the sociability test. (H–J) Representative heatmaps (H) and quantification of sniffing time (I; F[2,33] = 116.8) and social novelty index (J, F[1.74,19.13] = 132.3) in MSGAD67-ChR2 mice during the social novelty test. (K) Representative EEG spectrogram (top), EMG trace (middle), and brain states (bottom) from a MSGAD67-ChR2 mouse. Blue stripe indicates laser stimulation (473 nm, 10 Hz, 120 seconds). (L) Percentage of time in different brain states before, during, and after blue laser (473 nm, 10 Hz, 120 seconds) activation of MSGABA neurons (n = 12 mice, NREM, REM, and Wake). Shading represents ±SEM. (M) The changes in transition probability between each pair of brain states in MSGAD67-ChR2 mice during blue laser stimulation. (N and O) No significant change in percentage of time in different brain states (N, n = 10 mice, NREM, P = 0.189; wake, P = 0.196; REM, P = 0.13) and transition probability (O, P > 0.05) in MSGAD67-eGFP control mice. *P < 0.05; ***P < 0.001; #P < 0.05; ##P < 0.01; ###P < 0.001 by 2-way (F and I) or 1-way (G and J) repeated-measures ANOVA with Bonferroni’s post hoc test, or bootstrap test (L–O). NS, not significant.

On the other hand, we applied the same 10 Hz blue laser stimulations to MSGABA neurons in the regulation of sleep-wake. Optical activation of MSGABA neurons induced immediate transitions from NREM sleep to wakefulness (Figure 3K), a notable reduction in NREM and REM sleep, and a complementary increase in wakefulness (Figure 3L). We further quantified the changes in transition probability between each pair of brain states during the activation of MSGABA neurons. A significant increase was observed in the NREM→wake and wake→wake transitions during the activation (Figure 3M and Supplemental Figure 5H). Conversely, a complementary decrease was detected in the wake→NREM and NREM→NREM transitions in MSGAD67-ChR2 mice, indicating an increase in both initiation and maintenance of wakefulness, as well as a reduction in NREM sleep. However, no significant changes in brain state percentage and transition probability were found in MSGAD67-eGFP control mice during laser stimulation (Figure 3, N and O, and Supplemental Figure 5I). Collectively, these findings demonstrate that the activation of MSGABA neurons induces social memory deficits and sleep loss in C57BL/6J mice.

Both sleep deprivation and social isolation reduce NLG3 expression and increase MSGABA neuron activity. We next asked whether direct sleep deprivation and social isolation could mimic the effects of optogenetic activation of MSGABA neurons to regulate sleep and social behaviors. To address this question, we performed a 6-hour sleep deprivation in C57BL/6J mice (Figure 4A). This sleep deprivation induced deficits in social memory (Figure 4, B–G), with effects similar to that resulting from optogenetic activation of MSGABA neurons. Furthermore, we checked the mRNA level of NLG3 and c-Fos. There was a significantly lower level of Nlg3 mRNA in MS of mice subjected to sleep deprivation than in control mice (Figure 4, H and I). Moreover, the majority of c-fos–expressing neurons in the MS of sleep-deprived mice coexpressed Vgat mRNA (Figure 4, J–L), which is consistent with high colocalization of c-fos and Vgat mRNA in the MS of Nlg3-CKO mice (Figure 2, A–D).

Both sleep deprivation and social isolation reduce NLG3 expression and increase MSGABA neuron activity. (A) Experimental scheme. (B) Representative heatmaps showing occupancy time in a control and sleep-deprived (SD) mouse during sociability test. (C and D) Quantification of sniffing time (C, control, n = 11 mice; SD, n = 12 mice; F[1,21] = 14.52) and sociability index (D, P = 0.833) in the sociability test. (E) Representative heatmaps of occupancy time during social novelty test from a control and SD mouse. (F and G) Quantification of sniffing time (F, F[1,21] = 128.8) and social novelty index (G) in the social novelty test. (H and I) Representative FISH images (H) and quantification of Nlg3 mRNA in the MS between control and SD mice (I, n = 5 mice). (J–L) Representative FISH images (J and K) showing coexpression of c-fos and Vgat mRNA in the MS, and quantification between control and SD mice (L, n = 5 mice). Scale bars: 200 μm (left) and 20 μm (right). (M) Schematic showing experiment protocol. (N–P) The average (mean ± SEM) percentages of wake, NREM, or REM sleep during the day and night for control (n = 12 mice) and socially isolated (SI) mice (n = 13 mice). ZT, zeitgeber time. (Q–S) SI mice exhibited significantly more wakefulness (F[1,23] = 21.98), less NREM (F[1,23] = 9.567), and REM sleep (F[1,23] = 20.37) during the day and night. (T and U) Representative FISH images (T) and quantification of Nlg3 mRNA in the MS between control and SI mice (U, n = 5 mice). (V–X) Representative FISH images (V and W) showing coexpression of c-fos and Vgat mRNA in the MS and quantification between control and SI mice (X, n = 5 mice). *P < 0.05; **P < 0.01; ***P < 0.001 by 2-way ANOVA with Bonferroni’s post hoc test (C, F, and Q–S) or 2-tailed Mann-Whitney test (D, G, I, L, U, and X).

On the other hand, we performed 4 weeks of social isolation in C57BL/6J mice (Figure 4M). Socially isolated mice exhibited considerably less NREM sleep and more wakefulness than control (group housed) mice (Figure 4, N–S). Similarly, the mRNA level of Nlg3 in the MS from socially isolated mice was consistently lower than that from control mice (Figure 4, T and U). Additionally, most of the MS neurons expressing c-fos in socially isolated mice also coexpressed Vgat (Figure 4, V–X). Collectively, these findings demonstrate that sleep deprivation impairs social memory and chronic social isolation reduces sleep, implicating potential roles of NLG3 protein and MSGABA neurons in social and sleep behaviors.

Inactivation of MSGABA neurons increases NREM sleep and ameliorates social memory deficits in Nlg3-CKO mice. In contrast with activation, we next sought to determine whether the inactivation of MSGABA neurons could ameliorate sleep loss and social memory deficits in Nlg3-CKO mice. Specifically, we injected mixed viruses (pAAV-hSyn-Cre and AAV2-GAD67-eNpHR-eGFP) into the MS of Nlg3fl/fl mice (MSGAD67-eNpHR-CKO mice; Figure 5, A and B, and Supplemental Figure 6I). Six weeks after virus injection, optogenetic inhibition (589 nm, 8 seconds on/2 seconds off, 2 minutes) of MSGABA neurons induced a significant increase in NREM sleep, with a consequent reduction in wakefulness in Nlg3-CKO mice (Figure 5, C–E). Further analyses reveal that these changes were attributable to the increased NREM→NREM and wake→NREM transitions, as well as the decreased wake→wake and NREM→wake transitions in MSGAD67-eNpHR-CKO mice (Figure 5E and Supplemental Figure 6A). In the MSGAD67-eGFP-CKO control mice, in which pAAV-hSyn-Cre and AAV2-GAD67-eGFP were injected into the MS of Nlg3fl/fl mice, the optogenetic inhibition had no discernible effect on either the brain state percentages or the transition probability (Figure 5, F and G, and Supplemental Figure 6B).

Inactivating MSGABA neurons ameliorates sleep loss and social memory deficits in Nlg3-CKO mice. (A) Schematic diagram of the experimental procedure. (B) Expression of eNpHR-eGFP in MSGABA neurons of Nlg3-CKO mice (MSGAD67-eNpHR-CKO mice). Scale bar: 200 μm. (C) Baseline conditions and yellow laser inactivation of MSGABA neurons from an MSGAD67-eNpHR-CKO mouse with representative EEG spectrogram (top), EMG trace (middle), and brain states (bottom). Yellow stripe indicates laser stimulation (589 nm, yellow laser, 8 seconds on/2 seconds off, 120 seconds). (D) Percentage of time in different brain states before, during, and after yellow laser inactivation of MSGABA neurons from MSGAD67-eNpHR-CKO mice (n = 9 mice, NREM and Wake; REM, P = 0.252). Shading represents ±SEM. (E) The changes in transition probability between each pair of brain states in MSGAD67-eNpHR-CKO mice during yellow laser stimulation. (F and G) No significant change in percentage of time in different brain states (F, n = 8 mice, NREM, P = 0.25; wake, P = 0.154; REM, P = 0.174) and transition probability (G, P > 0.05) in MSGAD67-eGFP-CKO control mice. (H–J) Representative heatmaps (H) and quantification of sniffing time (I; n = 9 mice, F[2,24] = 0.051, P = 0.951) and sociability index (J, F[1.27,10.17] = 1.719, P = 0.224) in MSGAD67-eNpHR-CKO mice during the sociability test. (K–M) Representative heatmaps (K) and quantification of sniffing time (L; F[2,24] = 38.29) and social novelty index (M, F[1.896,15.17] = 39.65) in MSGAD67-eNpHR-CKO mice during the social novelty test. *P < 0.05; ***P < 0.001; #P < 0.05 by bootstrap test (D–G), or 2-way (I and L) or 1-way (J and M) repeated-measures ANOVA with Bonferroni’s post hoc test. NS, not significant.

Moreover, silencing MSGABA neurons in MSGAD67-eNpHR-CKO mice significantly decreased the sniffing time and total distance travelled in sociability and social novelty tests (Supplemental Figure 6, C–H). These results may be due to a significant increase in NREM sleep induced by inhibition of MSGABA neurons. Therefore, to avoid dozing off or falling asleep in MSGAD67-eNpHR-CKO mice during social tests, we applied a repetitive yellow laser stimulation paradigm (589 nm, 8 seconds on/2 seconds off, 120 seconds) that inhibited MSGABA neurons for 1 hour prior to the social tests (Figure 5A). After application of yellow laser stimulation, MSGAD67-eNpHR-CKO mice exhibited no significant difference in social preference for S1 and sociability index during the sociability test (Figure 5, H–J). However, there were significant increases in the time spent sniffing S2 and the social novelty index in MSGAD67-eNpHR-CKO mice during the social novelty test (Figure 5, K–M). Meanwhile, the firing rate of MSGABA neurons in MSGAD67-eNpHR-CKO mice decreased significantly after 1 hour of yellow laser stimulation (Supplemental Figure 7, A–C). Following the repetitive yellow laser stimulation, MSGAD67-eGFP-CKO control mice showed no noticeable changes in sociability and social memory (Supplemental Figure 7, D–G). These data suggest that the inactivation of MSGABA neurons increases NREM sleep and ameliorates social memory deficits in Nlg3-CKO mice, indicating that dual-functioning MSGABA neurons regulate both social memory and sleep.

MSGABA neurons project to both POA and CA2. We next investigated the downstream projections that regulate 2 distinctly different physiological modalities. We conducted anterograde tracing of MSGABA neurons in Nlg3-CKO mice and identified a series of downstream brain regions that are known to regulate either sleep/wakefulness or social memory. These regions include the preoptic area (POA) (21–23), lateral hypothalamus area (LHA) (24–26), medial habenula (MHb) (27), hippocampal CA2 (28–33), ventral tegmental area (VTA) (6, 34, 35), ventrolateral periaqueductal gray (vlPAG) (36), supramammillary region (SuM) (37–39), and dorsal raphe (DR) (40) (Figure 6, A and B, and Supplemental Figure 8, A and B). It has been postulated that the activation of GABAergic neurons could potentially reduce sleep by inhibiting sleep-promoting neurons (41–43), and impair social memory through inactivation of social memory–encoding neurons (44, 45).

MSGABA neurons project to both the POA and CA2. (A) Schematic drawing showing the axonal distributions of MSGABA neurons in Vgat-Cre mice. MHb, medial habenula; HP, hippocampus; vlPAG, ventrolateral periaqueductal gray; DR, dorsal raphe; VTA, ventral tegmental area; SuM, supramammillary region; LHA, lateral hypothalamus area; POA, preoptic area. (B) Viral expression of mGFP and mRuby in MSGABA neurons of Vgat-Cre mouse. MSGABA neurons send projections to a variety of brain regions. Scale bar: 200 μm. ox, optic chiasm; 3V, 3rd ventricle; Aq, cerebral aqueduct; fr, fasciculus retroflexus. (C) Schematic of viral injection for simultaneous retrograde tracing from the POA and CA2 in Nlg3-CKO mice. (D) Representative images displaying retro-AAV injection sites in the POA (red) and CA2 (green). Scale bar: 200 μm. (E) Quantification of the percentage of labeled cells in the MS. Note that POA-projecting and CA2-projecting MSGABA neurons are partially overlapping. n = 5 mice. POA+, POA-projecting cells; CA2+, CA2-projecting cells; Vgat+, MSGABA neurons expressing Vgat. (F) Fluorescence images of MS showing the retrograde-labeled POA-projecting cells (in red), CA2-projecting cells (in green), and Vgat (in cyan) neurons. Scale bars: 300 μm (top) and 20 μm (bottom).

It is well known that the POA is enriched in sleep-promoting neurons (21–23), while the CA2 contains the neurons that encode social memory (28–33), among the downstream projection regions of MSGABA neurons. We injected rAAV2 retroviruses into the POA and CA2 to express mCherry and eGFP, respectively, in projection fibers. This procedure allowed for the retrograde labeling of projection neurons in the MS of Nlg3-CKO mice (46) (Figure 6, C–F). Notably, the overlap between mCherry-MSGABA neurons and eGFP-MSGABA neurons was approximately 30% (31.53% ± 2.35%), suggesting a portion of MSGABA neurons divergently projected to both the POA and CA2. What are the roles of downstream targets of the POA and CA2 in regulating sleep and social memory?

Inhibition of CA2 or POA neurons in C57BL/6J mice innervated by MSGABA neurons selectively impairs social memory or reduces sleep. To explore the functional roles of relevant downstream circuits, the conventional approach involves the activation of ChR2-expressing axon terminals in a specific target (47). However, this terminal activation has been shown to induce “antidromic spikes” and unwanted activation of collateral targets through antidromic stimulation (47). Therefore, to avoid “antidromic stimulation” and selectively inhibit CA2 or POA neurons innervated by MSGABA neurons, we employed a dual virus approach and injected 2 separate vectors (Figure 7, A, B, and H). Specifically, we first injected anterograde trans-synaptic AAV (pAAV2/1-GAD67-EGFP-P2A-Cre-WPRE) into the MS of C57BL/6J mice, and subsequently injected another AAV (AAV8-Ef1a-DIO-eNpHR-mCherry) into either the CA2 or POA 1 week later (Figure 7, B and H). As a result, eNpHR-mCherry was selectively expressed in the CA2 or POA neurons that were innervated by MSGABA neurons in C57BL/6J mice. Moreover, after injection of only AAV8-Ef1a-DIO-eNpHR-mCherry into the CA2 and POA we did not detect any eNpHR-mCherry in the CA2 and POA (Supplemental Figure 8, C and D).

Inhibiting MSGABA-innervated CA2 or POA neurons selectively impairs social memory or reduces sleep, respectively, in C57BL/6J mice. (A) Schematic of the experimental procedure. (B) Schematic for labeling MSGABA→CA2 neurons in C57BL/6J mice. (C) Raster plot (top) and peristimulus time histogram (bottom) of a CA2 neuron innervated by MSGABA neurons. (D and E) Quantification of sniffing time (D, n = 12 mice, F[2,33] = 128.8) and social novelty index (E, F[1.84,20.23] = 175.1) by inactivation of MSGABA-innervated CA2 neurons in C57BL/6J mice during the social novelty test. (F) Representative EEG spectrogram (top), relative EMG trace (middle), and brain states (bottom) from a C57BL/6J mouse. Yellow stripe indicates laser stimulation (589 nm, 8 seconds on/2 seconds off, 120 seconds). (G) No significant change in NREM sleep (P = 0.192), wakefulness (P = 0.349), and REM sleep (P = 0.376) during laser stimulation (n = 12 mice). Shading represents ±SEM. (H) Similar to B, but for labeling MSGABA→POA neurons. (I) Similar to C, but for an MSGABA-innervated POA neuron. (J and K) Quantification of sniffing time (J, n = 13 mice, F[2,36] = 3.865, P = 0.03) and social novelty index (K, F[1.52,18.25] = 0.643, P = 0.496) by inactivation of MSGABA-innervated POA neurons in C57BL/6J mice during the social novelty test. (L) Similar to F, but for inactivation of MSGABA-innervated POA neurons. (M) Optogenetic inactivation of MSGABA-innervated POA neurons (589 nm, yellow laser, 8 seconds on/2 seconds off, 120 seconds) significantly decreased in NREM sleep (n = 13 mice), increased in wakefulness, and did not affect REM sleep (P = 0.336). Shading represents ±SEM. ***P < 0.001; ###P < 0.001 by 2-way (D and J) or 1-way (E and K) repeated-measures ANOVA with Bonferroni’s post hoc test, or bootstrap test (G and M). NS, not significant.

One month after virus injection in C57BL/6J mice, the optogenetic inhibition (589 nm, 8 seconds on/2 seconds off) of CA2 neurons innervated by MSGABA neurons did not have an impact on sociability (Figure 7C and Supplemental Figure 8, E and F), but significantly impaired social memory (Figure 7, D and E). However, this inhibition of CA2 neurons did not affect the sleep/wakefulness rhythm (Figure 7, F and G). In contrast, when the same optogenetic inhibition was applied to POA neurons innervated by MSGABA neurons (Figure 7, H and I), we did not find a significant change in either sociability or social memory (Figure 7, J and K, and Supplemental Figure 8, G and H), but there was a notable decrease in NREM sleep and an increase in wakefulness (Figure 7, L and M). These collective findings demonstrate that the inhibition of the CA2 or POA neurons innervated by MSGABA selectively impairs social memory or reduces sleep, respectively, in C57BL/6J mice.

Activation of MSGABA-innervated CA2 or POA neurons selectively ameliorates social memory deficits or recovers lost sleep in Nlg3-CKO mice. Finally, we investigate whether social memory deficits and sleep loss in Nlg3-CKO mice could be rescued by activation of CA2 or POA neurons innervated by MSGABA neurons. To this end, we created Nlg3-CKO mice by injecting pAAV-hSyn-Cre into the MS of Nlg3fl/fl mice (Figure 8, A, B, H, and I). Simultaneously, we labeled CA2 or POA neurons innervated by MSGABA neurons with hChR2-EGFP by an injection of anterograde trans-synaptic AAV (pAAV2/1-GAD67-hChR2-EGFP-3FLAG-WPRE).

Activating MSGABA-innervated CA2 or POA neurons selectively ameliorates social memory deficits or recovers lost sleep, respectively, in Nlg3-CKO mice. (A) Schematic of the experimental procedure. (B) Schematic for labeling MSGABA→CA2 neurons in Nlg3-CKO mice. (C) Raster plot and peristimulus time histogram of a CA2 neuron innervated by MSGABA neurons. (D and E) Quantification of sniffing time (D, n = 8 mice, F[2,21] = 141.3) and social novelty index (E, F[1.45,10.14] = 99.7) by activation of MSGABA-innervated CA2 neurons in Nlg3-CKO mice during the social novelty test. (F) Representative EEG spectrogram, relative EMG trace, and brain states from an Nlg3-CKO mouse. Blue stripe indicates laser stimulation (473 nm, 10 Hz, 120 seconds). (G) No significant change in NREM sleep (n = 8 mice, P = 0.446), wakefulness (P = 0.125), and REM sleep (P = 0.465) during laser stimulation. Shading represents ±SEM. (H) Similar to A, but with a repetitive photostimulation paradigm to activate POA neurons for 1 hour before social tests. (I) Similar to B, but with fiber implantation into the POA. (J) Similar to C, but for an MSGABA-innervated POA neuron. (K and L) Activation of MSGABA-innervated POA neurons (473 nm, 10 Hz, 120 seconds) significantly increased in NREM sleep (n = 9 mice), decreased in both wakefulness and REM sleep. Shading represents ±SEM. (M and N) Quantification of sniffing time (M, n = 9 mice, F[2,24] = 0.127, P = 0.881) and social novelty index (N, F[1.95,15.57] = 0.81, P = 0.46) by activation of MSGABA-innervated POA neurons in Nlg3-CKO mice during the social novelty test. ***P < 0.001; ###P < 0.001 by 2-way (D and M) or 1-way (E and N) repeated-measures ANOVA with Bonferroni’s post hoc test, or bootstrap test (G and L). NS, not significant.

Six weeks after virus injections, we performed optical activation (473 nm, 10 Hz) of the CA2 neurons innervated by MSGABA neurons. Optical activation of the CA2 neurons did not change sociability of Nlg3-CKO mice (Figure 8C and Supplemental Figure 8, I and J), but ameliorated social memory impairment (Figure 8, D and E). However, this activation of CA2 neurons did not affect the sleep/wakefulness rhythm (Figure 8, F and G). In contrast, when the same optogenetic stimulation was applied to POA neurons innervated by MSGABA neurons (Figure 8, H–J), we detected a notable increase in NREM sleep and a decrease in wakefulness and REM sleep (Figure 8, K and L). To prevent dozing off or falling asleep in Nlg3-CKO mice during social tests, we employed a repetitive photostimulation paradigm to activate the POA neurons for 1 hour before social tests (Figure 8H). After application of blue laser stimulation to the POA neurons, Nlg3-CKO mice did not show any significant change in sociability (Supplemental Figure 8, K and M), and their social memory remained impaired (Figure 8, M and N). These findings demonstrate that activating MSGABA-innervated CA2 or POA neurons selectively ameliorates social memory deficits or recovers lost sleep in Nlg3-CKO mice, respectively.

Taken together, all these results indicate that MSGABA-innervated CA2 and POA neurons are 2 downstream targets that diverge the regulatory function of MSGABA neurons toward social memory and sleep/wakefulness, respectively.