neuroscience copyright © 2020 cell type–differential ......liu et al., sci. adv. 2020 6 :...

Liu et al., Sci. Adv. 2020; 6 : eaay4073 22 July 2020

S C I E N C E A D V A N C E S | R E S E A R C H A R T I C L E

1 of 13

N E U R O S C I E N C E

Cell type–differential modulation of prefrontal cortical GABAergic interneurons on low gamma rhythm and social interactionLing Liu1*, Haifeng Xu1*, Jun Wang1*, Jie Li1*, Yuanyuan Tian1*, Junqiang Zheng1, Miao He2, Tian-Le Xu3, Zhi-Ying Wu1, Xiao-Ming Li1, Shu-Min Duan1, Han Xu1†

Prefrontal GABAergic interneurons (INs) are crucial for social behavior by maintaining excitation/inhibition balance. However, the underlying neuronal correlates and network computations are poorly understood. We identified distinct firing patterns of prefrontal parvalbumin (PV) INs and somatostatin (SST) INs upon social interaction. Moreover, social interaction closely correlated with elevated gamma rhythms particularly at low gamma band (20 to 50 Hz). Pharmacogenetic inhibition of PV INs, instead of SST INs, reduced low gamma power and impaired sociability. Optogenetic synchronization of either PV INs or SST INs at low gamma frequency improved sociability, whereas high gamma frequency or random frequency stimulation had no effect. These results reveal a functional differentiation among IN subtypes and suggest the importance of low gamma rhythms in social interaction be-havior. Furthermore, our findings underscore previously unrecognized potential of SST INs as therapeutic targets for social impairments commonly observed in major neuropsychiatric disorders.

INTRODUCTIONSocial behavior is the cornerstone of our daily life and the society. Unfortunately, social dysfunctions are commonly observed in many neuropsychiatric disorders, including autism, schizophrenia, social anxiety disorder, and depression (1–3). Both human and animal studies have implicated the medial prefrontal cortex (mPFC) as a key brain region in the control of social interaction behavior (4–6). To date, however, the prefrontal neuronal and network mechanisms responsible for social interaction remain elusive.

Cortical network activities underlying normal animal behaviors crucially depend on an exquisite interplay between excitation and inhibition. Consistently, disruption in excitation/inhibition balance within mouse mPFC has recently been shown to cause profound social deficits (6). In line with this finding, decreases in cortical inhibition are routinely found in animal models of schizophrenia, autism, or depression (7–10). Abnormalities in cortical -aminobutyric acid–releasing (GABAergic) interneurons (INs) have been commonly observed in neuropsychiatric patients with shared social impairments (11, 12). Together, these studies strongly suggest that maintaining proper inhibition within mPFC neuronal network is essential to ensure normal social interaction.

Synaptic inhibition within cortical microcircuitry is mediated by a large diversity of GABAergic INs (13, 14). This diversity essentially contributes to the ability of the cerebral cortex to perform complex operations through multiple circuit wiring strategies such as feed-forward inhibition, feedback inhibition, and disinhibition (15–17). However, we know little about the functional differentiation among

cortical IN subtypes and their resultant network computations in the control of social behaviors. Particularly, it is not known whether different IN subtypes contribute differentially to social interaction, and what kind of prefrontal network activities that they generated is important for appropriate social interaction. Answers to these questions are crucial to understand how mPFC regulates social interaction, behavior. Moreover, they are helpful to know precisely what impair-ment in circuit elements could cause social deficits associated with neuropsychiatric disorders.

In the present study, we found distinct firing patterns and manipulation effects of two major populations of prefrontal GABAergic INs, parvalbumin (PV) INs and somatostatin (SST) INs, during social interaction in mice. In addition, social interaction closely correlated with elevated gamma rhythms of the prefrontal local field potentials (LFPs), particularly at low gamma band (20 to 50 Hz). These results reveal distinct roles of PV INs and SST INs in prefrontal network computations for top-down control of social behavior. In addition, our data suggest a critical dependence of social interaction behavior on mPFC low gamma rather than high gamma rhythms.

RESULTSIdentification and recording of mPFC neurons in freely moving miceTo explore the mPFC neuronal correlates of real-time social inter-action, we used chronic electrophysiological recordings in adult male mice. Specifically, mice were implanted with microdrives contain-ing four to eight adjustable tetrodes aimed at the mPFC, and spiking activities were recorded in freely moving animals (fig. S1A). On the basis of spike features, we first classified the well-isolated neurons (218 from 13 mice) into narrow-spiking (NS; n = 44; trough-to-peak duration, 280 ± 11 s) putative inhibitory INs and wide-spiking (WS; n = 174; trough-to-peak duration, 454 ± 3 s) putative pyra-midal neurons (fig. S1, B to D). Cortical inhibitory INs are highly heterogeneous, of which fast-spiking PV (FS-PV) INs represent the largest population exhibiting landmark features of narrow-spike

1Center for Neuroscience and Department of Neurology of the Second Affiliated Hospital, NHC and CAMS Key Laboratory of Medical Neurobiology, Zhejiang University School of Medicine, Hangzhou 310058, China. 2Institute of Brain Science, State Key Laboratory of Medical Neurobiology, and Collaborative Innovation Center for Brain Science, Fudan University, Shanghai 200032, China. 3Center for Brain Sci-ence and Department of Anatomy and Physiology, Shanghai Jiao Tong University School of Medicine, Shanghai 200025, China.*These authors contributed equally to this work.†Corresponding author. Email: [email protected]

Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. Distributed under a Creative Commons Attribution NonCommercial License 4.0 (CC BY-NC).

on February 13, 2021

http://advances.sciencemag.org/

Dow

nloaded from




2 of 13

waveform and high discharge rate. On the basis of these features, we further classified putative inhibitory INs into FS-PV INs and non-FS NS neurons (18).

Optogenetics allows precise in vivo identification of a genetically defined population of neurons (18, 19); therefore, we also measured the spiking activity of FS-PV and SST INs with optrodes consisting of one optic fiber surrounded by multiple tetrodes (Fig. 1A). For optical tagging of PV- or SST-expressing INs, we delivered Cre- inducible adeno-associated virus (AAV) carrying the gene for channelrhodopsin-2 (ChR2-mCherry) into the mPFC of PV-Cre (20) or SST-Cre (21) knock-in mice, respectively. Histological examina-tion revealed an efficient and selective expression of ChR2 in PV INs (83% of PV-immunopositive neurons expressed ChR2-mCherry, 110 of 132 neurons; 97% of the ChR2-mCherry–expressing neurons were also PV immunopositive, 103 of 106 neurons; n = 3 mice; Fig. 1B) and SST INs (89% of SST-immunopositive neurons expressed ChR2-mCherry, 93 of 104 neurons; 90% of the ChR2-mCherry–expressing neurons were also SST immunopositive, 99 of 110 neurons; n = 3 mice; Fig. 1C). Blue light stimuli (470 nm, 0.08 to 1.35 mW, 1 to 2 ms) at high frequencies (20 and 40 Hz) were applied at the end of each recording session. Single units exhibiting reliable light-evoked spikes (spike probability, >90%) with short latencies (PV, 1.96 ± 0.03 ms; SST, 4.55 ± 0.60 ms) and low spike jitters (the SD of spike latency; PV, 0.62 ± 0.12 ms; SST, 0.85 ± 0.18 ms) were identified as PV or SST INs (PV, n = 14; SST, n = 15; Fig. 1, D to I, and fig. S2). Moreover, light-evoked PV or SST spikes were consistently followed by action potential suppression of WS neurons recorded on the same tetrode (Fig. 1, J and K), demonstrating robust temporal control of inhibitory INs upon surrounding principal cells. To compare inhibitory potency and temporal feature caused by PV IN and SST IN activation, we quantified WS neuronal firing suppression with multiple measurements. First, to quantify the magnitude of the overall WS firing suppression, we measured the firing frequency of WS neurons before (Pre rate) and after (Post rate) light stimulation and calculated a suppression index as (Pre rate − Post rate)/Pre rate. We found that WS firing suppression index is significantly larger upon PV than SST IN activation [PV IN activation: 0.50 ± 0.01 (n = 22); SST IN activation: 0.39 ± 0.02 (n = 25); P < 0.0001, unpaired t test], suggesting that PV IN activation suppressed WS neurons to a larger degree than SST neurons did. Second, to know how quick the WS firing suppression took place in response to PV and SST IN activa-tion, we measured the time point when maximal suppression (i.e., WS neuron’s firing was completely suppressed) starts. We found that the maximal WS suppression happened significantly earlier upon PV IN activation than SST IN activation [PV IN activation: 6.73 ± 0.28 ms (n = 22); SST IN activation: 9.92 ± 0.29 ms (n = 25); P < 0.0001, unpaired t test]. Third, to determine how long the WS suppression lasted upon PV and SST IN stimulation, we measured the duration during which WS neurons were completely suppressed. This analysis did not reveal a significant difference between PV INs and SST INs [PV IN activation: 11.55 ± 0.32 ms (n = 22); SST IN activation: 11.16 ± 0.24 ms (n = 25); P = 0.32, unpaired t test]. Together, PV INs had a stronger and faster inhibitory impact upon WS neurons firing than SST INs did. This observation is consistent with the well-described differential connectivity in subcellular domains where WS neurons receive PV inputs (perisomatic regions) and SST inputs (dendritic regions).

PV INs identified by either tetrodes or optrodes were indistin-guishable in terms of spike waveforms and spike rates; they were

therefore pooled for later analysis (Fig. 1L). Consistent with previous studies (18, 19), we observed that the baseline firing frequency of both PV INs and SST INs varied in a wide range (Fig. 1L). To know whether this heterogeneity corresponded to their physical locations, we sorted the firing rates of PV and SST INs to their respective re-cording depths. This analysis did not reveal a correlation between firing frequency and recording depth for either PV INs (r = −0.23, P = 0.25) or SST INs (r = 0.07, P = 0.82), suggesting that recording depth was not a major contributing factor (fig. S1, E and F). However, note that we could not exclude the possibility that the firing fre-quency diversity could be partially contributed by laminar effects.

PV INs and SST INs respond differentially to social interactionWe next measured the spiking activity of prefrontal neurons in mice subjected to a three-chambered social approach task, which is widely used for testing social behavior in rodents (22). Specifically, electrode- implanted mice were allowed to explore two opposing chambers with an empty cage (neutral stimulus) or a cage containing a conspecific mouse (social stimulus), respectively (Fig. 2A). Animal movement along the three chambers was simultaneously registered with a video tracking system to allow temporally precise correlation of neuronal activities with periods of social interaction (Fig. 2, C, F, and I).

We first examined the responses of WS pyramidal population between social versus neutral chambers. The mean firing rate of a particular neuron was first plotted as a function of the location of the subject mouse in the three chambers (Fig. 2C). To precisely determine neuronal activities when animals engaged in social inter-action (23), we next quantified their discharge rates during social interaction within social and neutral zones (Fig. 2D). A scatterplot summarizing the mean spike rates of individual WS neurons (Fig. 2E) clearly revealed a mixed modulation of the pyramidal population. A majority of WS neurons maintained their activity levels upon social interaction (60%, gray dots), while a small population showed increased (26%, orange dots) or decreased (14%, green dots) firing rates (Fig. 2E).

Next, we examined the responses of PV population between so-cial versus neutral chambers. Heat maps of the mean firing rate for individual PV IN were plotted as a function of the location of the subject mouse in the three chambers as described for WS neurons. In notable contrast to a mixed modulation property observed in WS neurons, the social zone was warmer than the neutral zone for almost all PV INs recorded (Fig. 2F), indicating the tendency for the social stimulus to increase firing rates of PV neurons. Quantification of the mean discharge rates within social and neutral zones (Fig. 2G) confirmed the above observation as the majority of PV INs (74%, 20 of 27) displayed a significant increase in firing rate when test animals were actively interacting with stimulus mouse (Fig. 2H). We found that a subpopulation of PV INs (30%, 8 of 27) also increased their firing rates as the test animals were approaching the stimulus animal. However, the proportion of PV INs exhibited increased firing rates during approach is significantly less than that during social inter-action (approach: 30%, 8 of 27; interaction: 74%, 20 of 27; P < 0.01, Fisher’s exact test). As expected, when the test animals were leaving the social chamber, the firing rates of all recorded PV INs returned to their baseline firing levels before approaching. With fiber photo-metry, a recent study demonstrated an increase in PV IN activity upon social interaction in mice (24). Therefore, our results corroborate the previous study by confirming that PV INs predominantly increased their discharge rates upon social interaction at the level of single unit.



Dow

nloaded from




3 of 13

A

OpticalRecordingelectrode

BPV ChR2-mCherry

PrL

IL

PV-Cre

DAPI AAV–ChR2-mCherry SST ChR2-mCherryDAPI AAV–ChR2-mCherry

PrL

IL

SST-Cre

C

D E

SpontaneousLight

Opto-tagged PV IN

200 ms

F

100 ms0.4 ms

J L

G

200 ms

H

0.4 ms

SpontaneousLight

I

100 ms

0 20 4030100

100

Tria

ls

Time (ms)0 20 403010

Tria

ls

0

100

0 20 4030100

100

Tria

ls

0 20 403010

Tria

ls

0

100

Frequency (Hz)

K

Opto-tagged SST IN

0 10 20 30 40 50 600

100

200

300

400

500

600

700

WS SSTPV

Trou

gh-t

o-pe

ak d

urat

ion

(µs)

Fig. 1. Optogenetic tagging and unit classification of mPFC neurons. (A) Diagram showing optogenetic tagging and electrophysiological recording using an optrode. (B) Location of unilateral optic fiber and ChR2 expression in PV INs. PrL, prelimbic; IL, infralimbic; DAPI, 4′,6-diamidino-2-phenylindole. Scale bars, 200 m (left) and 30 m (right). (C) The same as (B) but for SST INs. (D) Example recording of spontaneous and light-evoked spikes from an opto-tagged PV IN. Blue ticks, light pulses (20 Hz). (E) Over-lay of light-evoked (blue) and averaged spontaneous (red) spike waveforms from the example unit. (F) Raster plot of multiple trials showing spike responses to light stimuli at 20 and 40 Hz of the example unit. (G to I) The same as (D to F) but for an opto-tagged SST IN. (J) Raster plot of a light-activated PV IN (top) and an inhibited WS neuron (bottom) recorded from the same tetrode, both aligned to light onset. (K) The same as (J) but for an SST IN. (L) Summary of three types of mPFC neurons studied in the present study. Dark triangles, WS putative pyramidal cells (n = 174); blue circles, putative PV INs and opto-tagged PV INs (n = 27); green diamonds: opto-tagged SST INs (n = 15).



Dow

nloaded from




4 of 13

When heat maps of the mean firing rate for individual SST INs were plotted, their discharge rates seemed to be evenly distributed when the subject mouse moved around the three-chamber apparatus (Fig. 2I). Quantification of the mean firing rates within social and

neutral zones revealed that a large majority of SST INs (93%, 14 of 15) maintained their activity levels upon social interaction (Fig. 2, J and K). Further analysis revealed that the spiking activity of SST INs did not change either when the test animals were approaching or leaving

Soci

al12345

Neu

tral

5

1234

PV

F H

C EDWS

G

A B

SST

I KJ

12345

Neu

tral

12345

Soci

al

0.4 ms

200 μV

WS

PV

SST

Increase

No changeDecrease

0

30

60

90

0 30 60 90

74%

26%

Soci

al (�

ring

rate

, Hz)

250 ms

Neu

tral

5

1234

Soci

al

5

1234

Neutral (f iring rate, Hz)

0

10

20

30

40

0 10 20 30 40

Soci

al (�

ring

rate

, Hz)

93%

7%

0 5 10 15 20 25

25

0

5

10

15

20

Soci

al (�

ring

rate

, Hz)

60% 26%

14%

Firing rate

Min

MaxNeutral Social

Firing rate

Min

MaxNeutral Social

Firing rate

Min

MaxNeutral Social

Fig. 2. Firing responses of WS, PV, and SST neurons during social interaction. (A) Schematic illustration of electrophysiological recording paradigm. (B) Average spike waveforms of a WS (top), PV (middle), and SST neuron (bottom) recorded through four tetrode channels. Shaded areas indicate SEMs. (C) Heat map showing the firing rate of an example WS neuron. Warmer colors indicate higher firing rates. (D) Raster plot of spikes of the example WS neuron shown in (C) during five individual interactions in neutral zone (top) or social zone (bottom). (E) Correlation of firing rate in social zone versus neutral zone for individual WS neurons (n = 174). Orange and green circles indicate individual units with significantly higher or lower firing rates in social zone, respectively. Gray circles indicate neurons with no significant difference in firing rates. Inset: Proportions of WS neurons with significantly increased rates, decreased rates, or no change in rates upon social interaction (Student’s t test). (F to H) The same as (C to E) but for PV INs. Open circles (n = 13) and filled circles (n = 14) indicate putative and opto-tagged PV INs. (I to K) The same as (C to E) but for SST INs. All circles indicate opto- tagged SST INs (n = 15).



Dow

nloaded from




5 of 13

the stimulus mouse. To know whether SST firing in the center chamber might correspond to a preparatory phase before approaching the neutral zone or to a residual response after visiting the social zone, we also analyzed the center chamber firing rate of SST INs. First, we measured the center chamber firing rate before the experimental animal approached neutral zone and compared this center chamber firing rate to that at the neutral zone. The population results indicated that the center chamber firing rate did not significantly differ from that at the neutral zone [center chamber: 9.96 ± 2.53 Hz; neutral zone: 10.07 ± 2.63 Hz; n = 15; P = 0.976, Student’s t test; fig. S1G]. Second, we measured the center chamber firing rate after the experi-mental animal visited the social zone and compared this center chamber firing rate to that at the social zone. Again, we did not observe a significant difference between center chamber firing rate and that at the social zone [center chamber: 9.53 ± 2.36 Hz; social zone: 9.61 ± 2.27 Hz; n = 15; P = 0.983, Student’s t test; fig. S1H]. Therefore, as distinct from PV INs, the activities of SST INs did not change during social interaction.

Together, these results demonstrated that prefrontal PV INs and SST INs exhibited distinct firing patterns upon social interaction. Specifically, PV INs mostly increased their discharge rates, while SST INs predominantly maintained their activities upon social interaction.

Social interaction enhances mPFC low gamma oscillationsIt is well documented that PV INs play a crucial role in the generation of cortical network oscillations in the gamma band (20 to 80 Hz) (25, 26), which is essential for a variety of cognitive functions (27). Given that PV INs mostly increased their spiking activity upon social interaction, we wondered whether specific gamma power changes in mPFC were associated with mouse behavior of social interaction.

Therefore, we analyzed mPFC LFPs when mice performed a three-chambered social approach task. Frequency spectrogram demonstrated an apparent elevation in LFP gamma power during periods when the experimental mouse actively interacted with the stimulus mouse (Fig. 3, A to C). Statistical analysis further revealed that a significant change in gamma power occurred specifically at low gamma frequencies (20 to 50 Hz; neutral: 0.072 ± 0.008; social: 0.106 ± 0.011; n = 8, P < 0.001; Fig. 3D). In contrast, we did not see a difference in power at high gamma frequencies (50 to 80 Hz; neutral: 0.043 ± 0.006; social: 0.047 ± 0.007; n = 8, P = 0.21; Fig. 3D). Thus, social interaction induced an increase in mPFC gamma oscillations, particularly at low gamma frequencies. To disentangle potential nonspecific activation by nonsocial stimulus, we also examined the prefrontal cortical gamma rhythms when the test animal explored a nonsocial stimulus, that is, a novel object. Specifically, the stimulus mouse was replaced with a Lego in a three-chambered novel object task, and the prefrontal LFPs were recorded when the test mouse explored an object cage or a neutral cage. In contrast to social exploration, object exploration slightly, but not significantly, increased low gamma power under our experimental conditions (20 to 50 Hz; neutral: 0.079 ± 0.006; object: 0.086 ± 0.009; n = 6 sites from three mice, P = 0.09; Fig. 3, E to H). Together, these results suggest that low rather than high gamma rhythms could orchestrate network computations essential for cortical control of social interaction behavior.

Inhibition of PV INs impairs social interactionThe strong network and behavioral correlates of mPFC PV IN activities suggest a functional role of this population in top-down control of social interaction behavior. To directly test this hypothesis,

we used a pharmacogenetic inhibition method to selectively suppress the PV INs during social interaction test. Specifically, PV-Cre mice were injected bilaterally in mPFC with a Cre-dependent AAV ex-pressing mCherry-tagged hM4D, a designer receptor activated exclusively by the otherwise inert agonist clozapine N-oxide (CNO). Immunohistochemical analysis showed that hM4D expression extensively overlapped with PV INs (86% of PV-immunopositive neurons expressed hM4D-mCherry, 125 of 145 neurons; 95% of hM4D-mCherry–expressing neurons were PV immunopositive, 111 of 117 neurons; n = 3 mice), indicating efficient and selective ex-pression of hM4D in PV INs (Fig. 4A). To verify the effectiveness of pharmacogenetic inhibition, we directly measured the effect of CNO on PV IN firing in vivo in mice implanted with tetrodes. As predicted, we observed that the spike rates of putative PV INs decreased significantly following CNO administration (baseline: 27.90 ± 2.27 Hz; CNO: 14.66 ± 3.13 Hz; n = 6 neurons from three mice; P < 0.001; Fig. 4B). Meanwhile, analysis of LFPs revealed a significant reduction in low gamma (20 to 50 Hz) power as well (pre-CNO: 0.109 ± 0.024; post-CNO: 0.055 ± 0.013; n = 6 sites from three mice; P < 0.01; Fig. 4C).

Next, we assessed mice sociability following CNO administration with a three-chambered social approach task (Fig. 4D). As com-pared to the control PV-Cre mice expressing enhanced yellow fluo-rescent protein (EYFP), those expressing hM4D exhibited a significant decrease in both the time spent in the social chamber [hM4D: 303.40 ± 14.26 s (n = 9); EYFP: 381.33 ± 10.72 s (n = 9); P < 0.0001] and the social index [hM4D: 0.24 ± 0.05 (n = 9); EYFP: 0.46 ± 0.04 (n = 9); P < 0.01; Fig. 4E)]. A closer examination of social and neutral zones confirmed the reduced sociability as reflected as a reduction in both interaction time [hM4D: 211.95 ± 15.62 s (n = 9); EYFP: 281.02 ± 14.02 s (n = 9); P < 0.01] and social index [hM4D: 0.28 ± 0.06 (n = 9); EYFP: 0.48 ± 0.05 (n = 9); P < 0.05; Fig. 4F]. The impair-ment in social behavior was not due to an alteration in locomotion or general anxiety because CNO treatment did not affect travel distance, time in center, or center entry times in an open-field test (fig. S3, A to D). To assess the specificity of the pharmacogenetic manipulation on social behavior, we also conducted an object interaction test. Follow-ing CNO administration, no difference was seen in interaction time with an inanimate object (plastic Lego) between hM4D-expressing mice and EYFP-expressing control mice (fig. S3, E to G). These results therefore revealed that inhibition of mPFC PV INs impairs normal social behavior and indicated that PV IN activity is necessary for the top-down control of naturally occurring social interaction. Note that a previous work using optogenetic technique did not observe an effect of mPFC PV IN inhibition on social approach in mice (6). This discrepancy could be due to a low transfection effi-ciency of halorhodopsin in PV neurons, or alternatively, it could be that only a small number of NpHR3.0-expressing PV neurons were inhibited as a result of limited dispersion of light in brain tissue.

To ascertain whether the observed social impairment was specific to PV IN inhibition or whether it was a nonselective effect due to a general reduction in cortical inhibition, we also examined the con-sequence of inhibition of SST INs, the second largest population of inhibitory INs in the neocortex (14). To this end, a Cre-dependent hM4D virus was bilaterally targeted to mPFC of SST-Cre mice. Immunohistochemical examination verified the efficiency and spec-ificity of hM4D expression on SST INs (94% of SST-immunopositive neurons expressed hM4D-mCherry, 97 of 103 neurons; 92% of hM4D-mCherry expressing neurons were SST immunopositive,



Dow

nloaded from




6 of 13

159 of 172 neurons; n = 3 mice; fig. S4A). To verify the effectiveness of pharmacogenetic inhibition, we transduced mPFC SST neurons with both ChR2 and hM4D in a separate group of animals implanted with optrodes. As expected, we observed significantly decreased spike rates of opto-tagged SST INs following CNO administration (baseline: 5.92 ± 1.80 Hz; CNO: 3.30 ± 1.45 Hz; n = 5 neurons from three mice; P < 0.05; fig. S4B). These data indicated successful manipulation of SST neurons by pharmacogenetic inhibition.

In contrast to the detrimental effect of PV IN inhibition, however, we did not observe a difference between hM4D group and EYFP control group in terms of either interaction time or social index irrespective of whether the chambers [time: 357.00 ± 14.52 s for hM4D (n = 11) and 359.30 ± 11.50 s for EYFP (n = 11), P > 0.99; social index: 0.40 ± 0.05 for hM4D (n = 11) and 0.39 ± 0.04 for EYFP (n = 11), P = 0.86] or zones [time: 277.90 ± 13.71 s for hM4D (n = 11) and 290.50 ± 12.57 s for EYFP (n = 11), P = 0.91; social index: 0.45 ± 0.04 for hM4D (n = 11) and 0.45 ± 0.05 for EYFP (n = 11), P = 0.99] were analyzed (fig. S4, D to F). Consistent with no effect of SST IN inhibition on social behavior, analysis of LFPs revealed no change in gamma (20 to 50 Hz) power by CNO administra-tion (pre-CNO: 0.091 ± 0.009; post-CNO: 0.083 ± 0.012; n = 6 sites

from three mice; P = 0.31; fig. S4C). Together, these results revealed that inhibition of mPFC PV INs, but not SST INs, impairs normal social behavior and suggested that PV IN activities are specifically necessary for the top-down control of naturally occurring social interaction.

Activation of either PV INs or SST INs at low gamma frequency produces prosocial effectSince social interaction was accompanied by an increase in PV IN firing (Fig. 2) and gamma power (Fig. 3), we wondered whether an elevation in gamma power by synchronized activation of PV INs would have a prosocial effect. To address this issue, we injected PV-Cre mice with either ChR2-mCherry or a control EYFP virus and implanted optical fibers in the mPFC bilaterally for light delivery (Fig. 5, A and B). We chose 40-Hz light stimulation for our experi-ments because social interaction enhanced spectral power specifi-cally at low gamma frequencies (20 to 50 Hz; Fig. 3). As revealed in recordings with optrode described earlier, light stimuli effectively recruited PV INs at 40 Hz, as indicated by the reliable increase in PV IN firing following each stimulus (97 ± 1% of light pulses evoked an action potential; Fig. 1F). Moreover, PV IN activation generated synchronized gamma frequency inhibition, as indicated by periodic

C

A B

0

Freq

uenc

y (H

z)

Neutral

0 0.5 1 1.5 220

40

60

80

Time (s)

Raw LFP

30- to 40-Hzband-pass

Time (s)0.5 1 1.5 2

Min

Max

400 ms

50 μV1 m

V

Social

D0.15

0.10

0.05

0

***

20–50 50–80Frequency (Hz)

Rela

tive

pow

er

Rela

tive

pow

er

0.000

0.005

0.010

0.015

0.020

Frequency (Hz)0 20 40 60 80

NeutralSocial

Min

Max

50 μV1 m

V

G

E F

0

Freq

uenc

y (H

z)

Neutral

0 0.5 1 1.5 220

40

60

80

Time (s)

Raw LFP

30- to 40-Hzband-pass

Time (s)0.5 1 1.5 2

400 ms

Object

Rela

tive

pow

er

0.000

0.005

0.010

0.015

0.020

Frequency (Hz)0 20 40 60 80

NeutralObject

H0.15

0.10

0.05

020–50 50–80

Frequency (Hz)

Rela

tive

pow

er

Social interaction Object interaction

Fig. 3. Social interaction is characterized by an elevated low gamma power of mPFC LFPs. (A and B) Time course of mPFC raw LFP (top), band-filtered LFP (30 to 40 Hz; middle), and spectrogram (20 to 80 Hz; bottom) from a mouse exploring the neutral zone (A) or engaging in social interaction (B) during three-chambered social approach task. The time “0” refers to the onset of an interaction event. Warmer colors indicate enhanced power. (C) Averaged relative mPFC LFP power when test mice explored the neutral zone (black) or interacted with a stimulus mouse (red). Shaded areas indicate SEM. (D) Statistics of LFP power at different gamma frequency bands. Note the power at the low gamma band (20 to 50 Hz) is selectively enhanced when mice engaged in social interaction. Error bars indicate means ± SEM (n = 8 mice). ***P < 0.001, paired t test. (E to H) The same as (A to D) but for object interaction test. Note that there is no significant difference in LFP power between neutral cage explorations and object cage explorations.



Dow

nloaded from




7 of 13

suppression of firing in pyramidal cells (Fig. 1J). Besides, consistent with previous report (26), 40-Hz light stimuli induced a significant increase in the gamma power of mPFC network activities (baseline: 0.057 ± 0.013; light: 0.099 ± 0.021; n = 6 sites from three mice; P < 0.01; Fig. 5C).

We evaluated the effect of 40-Hz blue light (470 nm) stimuli on mice sociability with a three-chambered social approach task. The test consisted of two consecutive 5-min sessions, with the first session under no light stimuli and the second session under light illumina-tion at 40 Hz, respectively (Fig. 5D). As expected, in the first session without light stimuli, no differences were found in social interaction time or social index between the ChR2 group and the EYFP control group when either chambers or zones were considered (Fig. 5E). In

the second session, upon light stimuli, mice in the ChR2 group dis-played a significant increase in the time spent in the social chamber [ChR2: 174.73 ± 5.84 s (n = 7); EYFP: 137.87 ± 8.12 s (n = 9); P < 0.001] and in social index [ChR2: 0.41 ± 0.04 (n = 7); EYFP: 0.18 ± 0.07 (n = 9); P < 0.05] relative to EYFP controls (Fig. 5F, top). This prosocial effect was confirmed in terms of both interaction time [ChR2: 132.37 ± 5.92 s (n = 7); EYFP: 98.37 ± 6.84 s (n = 9); P < 0.001] and social index [ChR2: 0.45 ± 0.04 (n = 7); EYFP: 0.23 ± 0.07 (n = 9); P < 0.05] when zones were analyzed (Fig. 5F, bottom). The prosocial effect in PV-Cre::ChR2 mice was not due to an alteration in locomotor activity or general anxiety, as revealed with an open-field test (fig. S5, A to D). In addition, the effect was specific to social behavior because the same stimuli did not change

A B

PrL

IL

PV-Cre

DAPI AAV–hM4D-mCherry PV hM4D-mCherry

CNOBase

line

0

10

30

Firin

g ra

te (H

z)

20

40***

CNO

Baseline

200 μV

200 ms

DC

EYFP + CNO hM4D + CNO

FE

0.0

0.2

0.4

0.6

*

EYFPhM4D

Soci

al in

tera

ctio

n in

dex

0

100

200

300

400

Social

Neutral

Social

Neutral

EYFP hM4D

Tim

e in

zon

e (s

) **

Soci

al in

tera

ctio

n in

dex

EYFP0.0

0.2

0.4

0.6

hM4D

**

Tim

e in

cha

mbe

r (s)

EYFP hM4D

0

100

200

300

400****

*

Social

Middle

Neutral

Social

Middle

Neutral

Rela

tive

pow

er0.15

0.10

0.05

0.00

**

Post-CNO

Pre-CNO

Frequency (Hz)

Rela

tive

pow

er

0 20 40 60 800.000

0.005

0.010

0.015

0.020 Pre-CNOPost-CNO

Fig. 4. Pharmacogenetic inhibition of mPFC PV INs impairs social interaction. (A) Location of viral infection (left) and hM4D expression in PV INs (right). Scale bars, 200 (left) and 30 m (right). (B) Comparison of spontaneous spikes of PV INs before and after CNO administration. Error bars indicate means ± SEM (n = 6). ***P < 0.001, paired t test. (C) Relative mPFC LFP power upon social interaction before and after CNO administration. Shaded areas indicate SEM. Error bars indicate means ± SEM (n = 6 sites from three mice). **P < 0.01, paired t test. (D) Heat maps showing the locations of an EYFP-expressing control mouse (left) and an hM4D-expressing mouse (right) following CNO administration. (E and F) Quantification of time spent by hM4D and EYFP mice in each chamber (E) or each zone (F). Note that social interaction index was significantly smaller in hM4D mice. Error bars indicate means ± SEM (EYFP, n = 9; hM4D, n = 9). *P < 0.05, **P < 0.01, and ****P < 0.0001. Time in chamber or zone: two-way analysis of variance (ANOVA) and Bonferroni multiple comparison post hoc tests; interaction index: unpaired t test.



Dow

nloaded from




8 of 13

F

Social

Social

Neutral

Neutral

EYFPChR2

EYFP ChR2

0

50

100

150

200

Tim

e in

zon

e (s

) *** *

Soci

al in

tera

ctio

n in

dex

0.7

0.1

0.3

0.5

EYFPChR2

0.1

0.3

0.5

0.7

*

Soci

al in

tera

ctio

n in

dex

EYFP ChR2

Social

Middle

Neutral

Social

Middle

Neutral

*** *

Tim

e in

cha

mbe

r (s)

0

50

100

150

200

250

2nd 5 min with 40-Hz light stimuli

A B C

Opticalf ibers

DAPI AAV–ChR2-mCherry

PrL

IL

PV-Cre

Frequency (Hz)

Rela

tive

pow

er

0 10 20 30 40 50 60 700.000

0.005

0.010

0.015

0.020 Baseline40-Hz light

D

ChR2

OFF (1st 5 min) ON (2nd 5 min) OFF (1st 5 min) ON (2nd 5 min)

EYFP

E

EYFPChR2

Tim

e in

zon

e (s

)

0

50

100

150

200

Social

Social

EYFP ChR2

Neutral

Neutral

0.7

Soci

al in

tera

ctio

n in

dex

0.1

0.3

0.5

EYFPChR2

0.1

0.3

0.5

0.7

Soci

al in

tera

ctio

n in

dex

Tim

e in

cha

mbe

r (s)

0

50

100

150

200

250

Social

Middle

Neutral

Social

Middle

Neutral

EYFP ChR2

1st 5 min without light stimuli

0.15

Rela

tive

pow

er

Light

Baselin

e

0.10

0.05

0.00

**

Fig. 5. Activation of PV INs at low gamma frequency produces prosocial effect. (A) Diagram showing bilateral optogenetic manipulation in freely moving mice. (B) Placement of optic fibers for photostimulation of mPFC FS-PV INs in a PV-Cre mouse bilaterally injected with ChR2-mCherry. Scale bar, 200 m. (C) Left: Example spec-tral power of LFP with (blue) and without (black) light stimuli of PV INs at 40 Hz. Shaded areas indicate SEMs. Right: Comparison of relative LFP gamma power between baseline (black) and light stimuli (blue). Error bars indicate means ± SEM (n = 6 sites from three mice). **P < 0.01, paired t test. (D) Heat maps showing the locations of an EYFP-expressing control mouse (left) and a ChR2-expressing mouse (right) in a three-chambered social approach task. (E and F) Quantification of time spent by ChR2 mice and EYFP mice in each chamber (top) or zone (bottom) in the first 5-min test (E) and the second 5-min test (F). Error bars indicate means ± SEM (EYFP, n = 9; ChR2, n = 7). *P < 0.05 and ***P < 0.001. Time in chamber or zone: two-way ANOVA and Bonferroni multiple comparison post hoc tests; interaction index: unpaired t test.



Dow

nloaded from




9 of 13

the interaction time that PV-Cre::ChR2 mice spent with an inanimate object (fig. S5, E to G).

In the above social approach behavioral test, the EYFP control mice spent less time in the social chamber and social zone during the second 5 min of the test compared to the first (Fig. 5), likely because of social habituation (28). Because of this time effect, there exists a possibility that 40-Hz stimulation of PV INs altered cortical function to reduce habituation to social stimuli. To clarify this possibility, we also delivered 40-Hz stimulation to PV INs from the very beginning to the very end in a 5-min three-chambered social approach task. We found that the stimulated ChR2-expressing animals spent significantly more time with stimulus mouse as compared to EYFP-expressing control animals (fig. S6, A to C). Therefore, low gamma frequency stimulation of mPFC PV INs indeed increases sociability in mice.

To determine whether the prosocial effect was associated with an increase in inhibition or an elevation in gamma rhythms, we next stimulated PV INs with a nonfixed frequency but maintained the total number of light pulses (40 pulses per second). Although the nonrhythmic stimuli reliably recruited PV INs (baseline: 21.08 ± 4.47 Hz; light: 41.98 ± 1.11 Hz; n = 5; P < 0.05; 98 ± 1% of light pulses evoked an action potential; fig. S7A), it only caused a negligible in-crease in gamma (20 to 50 Hz) power (baseline: 0.056 ± 0.012; light: 0.068 ± 0.013; n = 6 sites from three mice; P = 0.10; fig. S7B). At behavioral level, the nonrhythmic stimuli produced only a slight, but not significant, increase in mice sociability (fig. S7, C to E). This result is indeed consistent with previous studies that demonstrate pharma-cogenetic activation of mPFC PV INs compromised social interaction rather than produced prosocial effect (17, 29). Therefore, social interaction is not simply determined by the gross level of PV IN firing activities. Instead, the resultant enhancement in cortical gamma oscillations by rhythmic PV IN activation seems to play a key role.

Given that 40-Hz light stimuli of SST INs induced rhythmic sup-pression of WS neurons as PV IN activation did (Fig. 1, J and K), we wondered whether rhythmic stimuli of SST INs could also produce a prosocial effect. To test this, we injected SST-Cre mice with either ChR2-mCherry or a control EYFP virus and implanted optical fibers in the mPFC bilaterally for light delivery (Fig. 6, A and B). Light stimuli at 40 Hz reliably activated SST INs (94 ± 1% of light pulses evoked an action potential) and induced a significant increase in the gamma (20 to 50 Hz) power of mPFC LFPs (baseline: 0.057 ± 0.008; light: 0.077 ± 0.009; n = 5 sites from three mice; P < 0.05; Fig. 6C). Similar to PV IN activation, optogenetic synchronization of SST INs at 40 Hz significantly increased the time of social interaction in SST-Cre::ChR2 mice (Fig. 6, D to F). The effect was not associated with a change in mice locomotion or anxiety and was specific to social interaction (fig. S5, H to N). Again, when 40-Hz light stimulation was continuously delivered to mPFC SST INs during a 5-min three-chambered social approach task, the stimulated ChR2-expressing animals spent significantly more time with stimulus mouse as compared to EYFP-expressing control animals (fig. S6, D to F). In contrast, nonrhythmic light stimulation (40 pulses per second) activated SST INs (74 ± 9% of light pulses evoked an action potential) but did not increase gamma (20 to 50 Hz) power (baseline: 0.049 ± 0.007; light: 0.049 ± 0.006; n = 5 sites from three mice; P = 0.94) or sociability (fig. S8). It even reduced the social interaction index. One possibility is that nonrhythmic stimulation of SST INs disrupted the temporal fidelity of PV IN spiking and, thereby, the synaptic inhibition they provided onto pyramidal cells. The decreased temporal precision

ultimately compromised the prefrontal inhibition required for normal social information processing (6). Consistently, pharma-cogenetic activation of mPFC SST INs did not produce a prosocial effect in a three-chambered social approach task either (fig. S9, A to D). These results further supported the notion that the rhythmic network activities generated by synchronized activation of inhibitory INs, but not the gross inhibition per se, is important to modulate social behavior.

Next, we further determined whether the prosocial effect of the rhythmic network activities is frequency dependent. To this end, we stimulated PV INs at a high gamma frequency (80 Hz) since light stimuli of PV INs at 80 Hz successfully recruited PV INs (baseline: 23.11 ± 5.04 Hz; light: 86.60 ± 11.62 Hz; n = 5; P < 0.01; 95 ± 2% of light pulses evoked an action potential; fig. S10A) and significantly enhanced spectral power at around 80 Hz (baseline: 0.020 ± 0.006; light: 0.042 ± 0.010; n = 7 sites from three mice; P < 0.01; fig. S10B). However, in contrast to 40-Hz stimulation, light stimulation at 80 Hz did not increase the time of social interaction (fig. S10, C to E).

Together, these results demonstrated that synchronized activation of either PV INs or SST INs and the resultant enhancement in pre-frontal gamma rhythms produce prosocial effect. Moreover, the prosocial effect of the prefrontal rhythmic network activities is frequency dependent and relies on low rather than high gamma rhythms.

DISCUSSIONIn the present study, using chronic single-unit recording with optogenetic-tagging technique, we directly measured prefrontal PV and SST neuronal activities during real-time social interaction in mice. We found that PV INs displayed elevated discharge rates, whereas SST INs maintained their activities in response to social stimuli. Moreover, the spectral power of LFP oscillations, specifically at low but not high gamma frequencies, was enhanced during bouts of social interaction. In addition, specific inhibition of mPFC PV INs, but not SST INs, with pharmacogenetics reduced low gamma power and impaired sociability. Furthermore, optogenetic activation of either PV INs or SST INs at low gamma frequency enhanced low gamma power and produced a prosocial effect. However, the prosocial effect was absent when PV INs were synchronized at high gamma frequency. Together, these results demonstrated that PV INs and SST INs function differentially in prefrontal network computations for top-down control of social behavior. Moreover, our data suggest a critical dependence of social interaction on prefrontal low rather than high gamma rhythms.

The role of mPFC IN subtypes in social interaction behaviorThe complex functions of the cerebral cortex depend on neuronal circuits of highly interconnected glutamatergic pyramidal neurons and GABAergic INs. Previous studies aiming to dissect prefrontal neuronal substrate of social interaction have mainly focused on pyramidal neurons (30, 31). Aside from excitatory cells, there is a large diversity of cortical GABAergic INs based on differences in their anatomical features, physiological properties, and the expres-sion of specific neurochemical markers (13, 14). Understanding IN diversity and its functional consequences is essential to understand cortical functions.

In rodent neocortex, PV INs and SST INs account for around 40 and 30% of total GABAergic INs, respectively, and represent the two largest populations of GABAergic INs (14). PV INs preferentially target perisomatic regions of pyramidal cells and frequently form



Dow

nloaded from




10 of 13

PrL

IL

DAPI AAV–ChR2-mCherrySST-Cre

A B C

Opticalf ibers

D

OFF (1st 5 min) ON (2nd 5 min)

EYFP

OFF (1st 5 min) ON (2nd 5 min)

ChR2

Frequency (Hz)

Rela

tive

pow

er

0 10 20 30 40 50 60 700.000

0.005

0.010

0.015

0.020 Baseline40-Hz light

Rela

tive

pow

er

Light

Baselin

e

0.15

0.10

0.05

0.00

*

E F

EYFPChR2

0.1

0.3

0.5

0.7

Soci

al in

tera

ctio

n in

dex

Tim

e in

cha

mbe

r (s)

0

50

100

150

200

250

Social

Middle

Neutral

Social

Middle

Neutral

EYFP ChR2 EYFP ChR2

Social

Middle

Neutral

Social

Middle

Neutral

Tim

e in

cha

mbe

r (s)

0

50

100

150

200

250

**

EYFPChR2

0.1

0.3

0.5

0.7

Soci

al in

tera

ctio

n in

dex

*

EYFP ChR2

Tim

e in

zon

e (s

)

0

50

100

150

200

Social

Social

Neutral

Neutral

EYFP ChR2

Social

Social

Neutral

Neutral

0

50

100

150

200

Tim

e in

zon

e (s

) **

EYFPChR2

0.7

Soci

al in

tera

ctio

n in

dex

0.1

0.3

0.5

EYFPChR2

*

Soci

al in

tera

ctio

n in

dex

0.7

0.1

0.3

0.5

1st 5 min without light stimuli 2nd 5 min with 40-Hz light stimuli

Fig. 6. Activation of SST INs at low gamma frequency also produces prosocial effect. (A) Diagram showing bilateral optogenetic manipulation in freely moving mice. (B) Placement of optic fibers for photostimulation of mPFC SST INs in an SST-Cre mouse bilaterally injected with ChR2-mCherry. Scale bar, 200 m. (C) Left: Example spec-tral power of LFP with (blue) and without (black) light stimuli of SST INs at 40 Hz. Shaded areas indicate SEMs. Right: Comparison of relative LFP gamma power between baseline (black) and light stimuli (blue). Error bars indicate means ± SEM (n = 5 sites from three mice). *P < 0.05, paired t test. (D) Heat maps showing the locations of an EYFP-expressing control mouse (left) and a ChR2-expressing mouse (right) in a three-chambered social approach task. (E to F) Quantification of time spent by ChR2 mice and EYFP mice in each chamber (top) or zone (bottom) in the first 5-min test (E) and the second 5-min test (F). Error bars indicate means ± SEM (EYFP, n = 8; ChR2, n = 11), *P < 0.05 and **P < 0.01. Time in chamber or zone: two-way ANOVA and Bonferroni multiple comparison post hoc tests; interaction index: unpaired t test.



Dow

nloaded from




11 of 13

contacts with each other both chemically and electrically (32, 33). In comparison, SST INs preferentially target distal dendrites of pyra-midal cells and typically do not form chemical contacts with each other (32, 33). In addition, these two IN populations have distinct membrane properties (32, 33). These notable differences in neuronal structures, synaptic targets as well as membrane properties strongly suggest their functional differentiation in cortical information pro-cessing and, hence, behavioral control. Recent studies gained evidence for functional correlates between specific IN subtypes and distinct behaviors ranging from attention and working memory to fear ex-pression (17–19). Our findings demonstrate that mPFC PV INs and SST INs play distinct roles in support of social interaction and therefore provide evidence that there is a functional differentiation among IN subtypes in social behavior.

A recent optogenetical study demonstrates elegantly that small changes in key parameters of the neuronal networks, such as spon-taneous neuronal activity, could produce a paradoxical conclusion as to the computational functions of INs in auditory cortex (34). This finding raises an important concern in studies that aim to dissect causal roles of IN subtypes in cortical computations by bidirectional manipulations of neuronal activities. For the present study, for ex-ample, a lack of behavioral effect on social interaction following SST IN inhibition could be a result of low baseline activity in mPFC SST population. However, our electrophysiological recordings offered a precise measurement of neuronal spiking activities under physio-logical conditions and revealed no modulation effect of this IN sub-type by social interaction. This observation alleviates the concern regarding possible limitations inherited from pharmacogenetic or optogenetic manipulations. Note that, in addition to PV and SST INs, there is another population of GABAergic INs expressing the ionotropic serotonin receptor 5HT3a (5HT3aR) (14). 5HT3aR INs are heterogeneous and include several subtypes, and their possible roles in social behavior are to be tested in future studies.

The role of gamma rhythms in social interaction behaviorGain-of-function and loss-of-function studies using optogenetics demonstrated that PV INs are causally involved in the generation of gamma rhythms in the neocortex (25, 26). In agreement with these findings, our data show that inhibition and activation of mPFC PV INs decreased and increased gamma power, respectively (Figs. 4 and 5). Although PV INs have been well accepted to be essential for gamma rhythms, most recent evidence indicates that SST INs are also inti-mately involved in gamma rhythms, particularly at low gamma range, in the visual cortices of awake mice (35). Consistently, we found that photostimulation of mPFC SST INs at 40 Hz was sufficient to enhance low gamma power (Fig. 6). Together, these findings sug-gest that there exist multiple inhibitory neuronal mechanisms for the generation of cortical gamma rhythms.

Rhythmic neuronal activity, particularly in the gamma frequency band, is believed to play a key role in various cognitive functions including attention, sensory perception, and working memory (27). As for social behavior, human studies with magnetoencephalography recordings also found increases in prefrontal gamma power when a subject is performing social interaction task (36). Consistently, we found increased gamma power when mice explored a social target as compared to a neutral object (Fig. 3). These data suggest a sound correlation between social interaction and prefrontal gamma rhythms. Our study further demonstrated a bidirectional modulation effect of gamma power on social interaction behavior. Specifically, when

gamma powers were reduced by PV IN inhibition, social interaction was compromised (Fig. 4). Conversely, when gamma power was augmented by synchronized activation of either PV INs or SST INs at low gamma frequency, social interaction was enhanced (Figs. 5 and 6). These data indicate a causal link between prefrontal gamma oscillations and social interaction behavior.

An excellent study by Yizhar et al. (6) demonstrates that elevation in mPFC excitation and inhibition balance by raising excitatory neurons’ activity increases gamma power and impairs social behavior. This result seems controversial to ours but could be explained by the difference in the frequency range where gamma power enhancement was induced. Constant activation of excitatory cells results in a marked power increase at high gamma frequencies peaked at 80 Hz (6). As comparison, social improvement observed in the present study was associated with a low gamma power enhancement (Figs. 5 and 6). In contrast, when PV INs were activated at high gamma frequency (80 Hz), no prosocial effect was observed (fig. S10). Given our finding that social interaction increased gamma power specifically at low (20 to 50 Hz), but not high (50 to 80 Hz), gamma range (Fig. 3), the observed dependence of social interaction on low gamma rhythms is reasonable.

How could prefrontal gamma rhythms contribute to the behavioral benefits of sociability? It is hypothesized that gamma oscillations pro-vide a means for information coding via the formation of coactive cell assemblies (37). Thus, the increased temporal precision of py-ramidal firing could help relay mPFC outputs to downstream brain regions important for social interaction, such as ventral tegmental area and nucleus accumbens (31, 38–40).

Implications for neuropsychiatric disorders with shared social deficitsImpairments in prefrontal gamma oscillations have been reported in a number of neuropsychiatric disorders with shared social deficits (41–43). Consistently, postmortem examinations find abnormalities in cortical GABAergic INs, particularly PV INs, in brain tissues from subjects with schizophrenia or autism (11, 12). In addition, functional alterations of the mPFC PV INs and aberrant gamma activities have been observed in mouse models of autism and schizophrenia (7, 8, 10). Our results demonstrate that PV activities and resultant gamma oscillations play causal roles in orchestrating the operations of local mPFC network to support social interaction behavior. On the basis of these findings, it is conceivable that deficits in mPFC PV INs interfere with the generation of proper gamma rhythms and thus prevents temporal precision of local pyramidal firing. As a con-sequence, pyramidal cells required for social control are not well assembled to drive downstream targets, and therefore, social dys-functions emerge.

On the other hand, the prosocial effect of synchronized activa-tion of PV INs at gamma frequency is encouraging (Fig. 5). It points to the potential of improving sociability or curing social impair-ments in major neuropsychiatric disorders including schizophrenia and autism by targeting PV INs in the mPFC. A recent study demon-strates that increasing prefrontal PV IN activity rescues social defi-cits in a mouse model of autism (24). We found that activation of SST INs at low gamma frequency could also boost low gamma pow-er and produced a prosocial effect (Fig. 6). This finding has an im-portant clinical implication that SST INs could serve as potential targets for manipulation to improve social performance. Besides, our data indicated a critical dependence of social interaction on low



Dow

nloaded from




12 of 13

but not high, gamma rhythms. This observation is also of important clinical relevance. When transcranial electrical/magnetic stimula-tion are used as treatment options to improve sociability, choosing low gamma stimulation frequency could be a key factor to achieve a beneficial effect.

MATERIALS AND METHODSA brief summary is provided here. For more detailed information, please refer to Supplementary Methods. Adult male mice (2 to 4 months old) were used in the present study. All experimental pro-cedures were conducted in accordance with the Guidelines for the Care and Use of Laboratory Animals of Zhejiang University. Animal surgery, virus injection, immunostaining for PV/SST INs, in vivo single-unit recordings, and optogenetic stimulation were carried out as described in our previous studies (17, 44) and in the Supplementary Materials. Electrophysiological recordings were performed with a Plexon amplifier and analyzed offline with NeuroExplorer (Plexon, USA). For optogenetic stimulations, PV-Cre mice received stereotaxic injections of AAV viruses carrying ChR2 in the mPFC. Bilateral activation of ChR2 was performed with implanted optic fibers con-nected to a laser beam. For pharmacogenetic manipulation experi-ments, to avoid potential confounding effect of CNO metabolite (45), EYFP-infected animals received the same amount of CNO administration as hM4D-infected animals did and served as controls. Animal behaviors were video recorded and analyzed with EthoVision XT (Noldus). All data are shown as means ± SEM unless otherwise specified. Statistical analysis was done with Prism 7 (GraphPad) or MATLAB and described in detail in Supplementary Methods.

SUPPLEMENTARY MATERIALSSupplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/6/30/eaay4073/DC1

View/request a protocol for this paper from Bio-protocol.

REFERENCES AND NOTES 1. L. K. Bicks, H. Koike, S. Akbarian, H. Morishita, Prefrontal cortex and social cognition

in mouse and man. Front. Psychol. 6, 1805 (2015). 2. R. Hari, M. V. Kujala, Brain basis of human social interaction: From concepts to brain

imaging. Physiol. Rev. 89, 453–479 (2009). 3. R. Hari, L. Henriksson, S. Malinen, L. Parkkonen, Centrality of social interaction in human

brain function. Neuron 88, 181–193 (2015). 4. L. Schilbach, A. M. Wohlschlaeger, N. C. Kraemer, A. Newen, N. J. Shah, G. R. Fink,

K. Vogeley, Being with virtual others: Neural correlates of social interaction. Neuropsychologia 44, 718–730 (2006).

5. J. Sliwa, W. A. Freiwald, A dedicated network for social interaction processing in the primate brain. Science 356, 745–749 (2017).

6. O. Yizhar, L. E. Fenno, M. Prigge, F. Schneider, T. J. Davidson, D. J. O'Shea, V. S. Sohal, I. Goshen, J. Finkelstein, J. T. Paz, K. Stehfest, R. Fudim, C. Ramakrishnan, J. R. Huguenard, P. Hegemann, K. Deisseroth, Neocortical excitation/inhibition balance in information processing and social dysfunction. Nature 477, 171–178 (2011).

7. K. K. A. Cho, R. Hoch, A. T. Lee, T. Patel, J. L. R. Rubenstein, V. S. Sohal, Gamma rhythms link prefrontal interneuron dysfunction with cognitive inflexibility in Dlx5/6+/− mice. Neuron 85, 1332–1343 (2015).

8. S. Han, C. Tai, R. E. Westenbroek, F. H. Yu, C. S. Cheah, G. B. Potter, J. L. Rubenstein, T. Scheuer, H. O. de la Iglesia, W. A. Catterall, Autistic-like behaviour in Scn1a+/− mice and rescue by enhanced GABA-mediated neurotransmission. Nature 489, 385–390 (2012).

9. J.-F. Sauer, M. Strüber, M. Bartos, Impaired fast-spiking interneuron function in a genetic mouse model of depression. eLife 4, e4979 (2015).

10. W. Cao, S. Lin, Q.-Q. Xia, Y.-L. Du, Q. Yang, M.-Y. Zhang, Y.-Q. Lu, J. Xu, S.-M. Duan, J. Xia, G. Feng, J. Xu, J.-H. Luo, Gamma oscillation dysfunction in mPFC leads to social deficits in neuroligin 3 R451C knockin mice. Neuron 97, 1253–1260.e7 (2018).

11. D. A. Lewis, T. Hashimoto, D. W. Volk, Cortical inhibitory neurons and schizophrenia. Nat. Rev. Neurosci. 6, 312–324 (2005).

12. E. Hashemi, J. Ariza, H. Rogers, S. C. Noctor, V. Martínez-Cerdeño, The number of parvalbumin-expressing interneurons is decreased in the medial prefrontal cortex in autism. Cereb. Cortex 27, 1931–1943 (2017).

13. G. Fishell, B. Rudy, Mechanisms of inhibition within the telencephalon: "Where the wild things are". Annu. Rev. Neurosci. 34, 535–567 (2011).

14. B. Rudy, G. Fishell, S. Lee, J. Hjerling-Leffler, Three groups of interneurons account for nearly 100% of neocortical GABAergic neurons. Dev. Neurobiol. 71, 45–61 (2011).

15. J. S. Isaacson, M. Scanziani, How inhibition shapes cortical activity. Neuron 72, 231–243 (2011).

16. R. Tremblay, S. Lee, B. Rudy, GABAergic interneurons in the neocortex: From cellular properties to circuits. Neuron 91, 260–292 (2016).

17. H. Xu, L. Liu, Y. Tian, J. Wang, J. Li, J. Zheng, H. Zhao, M. He, T.-L. Xu, S. Duan, H. Xu, A disinhibitory microcircuit mediates conditioned social fear in the prefrontal cortex. Neuron 102, 668–682.e5 (2019).

18. H. Kim, S. Ährlund-Richter, X. M. Wang, K. Deisseroth, M. Carlén, Prefrontal parvalbumin neurons in control of attention. Cell 164, 208–218 (2016).

19. D. Kim, H. Jeong, J. Lee, J.-W. Ghim, E. S. Her, S.-H. Lee, M. W. Jung, Distinct roles of parvalbumin- and somatostatin-expressing interneurons in working memory. Neuron 92, 902–915 (2016).

20. S. Hippenmeyer, E. Vrieseling, M. Sigrist, T. Portmann, C. Laengle, D. R. Ladle, S. Arber, A developmental switch in the response of DRG neurons to ETS transcription factor signaling. PLOS Biol. 3, e159 (2005).

21. H. Taniguchi, M. He, P. Wu, S. Kim, R. Paik, K. Sugino, D. Kvitsiani, Y. Fu, J. Lu, Y. Lin, G. Miyoshi, Y. Shima, G. Fishell, S. B. Nelson, Z. J. Huang, A resource of Cre driver lines for genetic targeting of GABAergic neurons in cerebral cortex. Neuron 71, 995–1013 (2011).

22. S. S. Moy, J. J. Nadler, A. Perez, R. P. Barbaro, J. M. Johns, T. R. Magnuson, J. Piven, J. N. Crawley, Sociability and preference for social novelty in five inbred strains: An approach to assess autistic-like behavior in mice. Genes Brain Behav. 3, 287–302 (2004).

23. J. L. Silverman, M. Yang, C. Lord, J. N. Crawley, Behavioural phenotyping assays for mouse models of autism. Nat. Rev. Neurosci. 11, 490–502 (2010).

24. A. Selimbeyoglu, C. K. Kim, M. Inoue, S. Y. Lee, A. S. O. Hong, I. Kauvar, C. Ramakrishnan, L. E. Fenno, T. J. Davidson, M. Wright, K. Deisseroth, Modulation of prefrontal cortex excitation/inhibition balance rescues social behavior in CNTNAP2-deficient mice. Sci. Transl. Med. 9, eaah6733 (2017).

25. V. S. Sohal, F. Zhang, O. Yizhar, K. Deisseroth, Parvalbumin neurons and gamma rhythms enhance cortical circuit performance. Nature 459, 698–702 (2009).

26. J. A. Cardin, M. Carlén, K. Meletis, U. Knoblich, F. Zhang, K. Deisseroth, L.-H. Tsai, C. I. Moore, Driving fast-spiking cells induces gamma rhythm and controls sensory responses. Nature 459, 663–667 (2009).

27. P. J. Uhlhaas, W. Singer, Abnormal neural oscillations and synchrony in schizophrenia. Nat. Rev. Neurosci. 11, 100–113 (2010).

28. D. E. Dluzen, J. D. Kreutzberg, 1-Methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) disrupts social memory/recognition processes in the male mouse. Brain Res. 609, 98–102 (1993).

29. B. R. Ferguson, W.-J. Gao, Thalamic control of cognition and social behavior via regulation of gamma-aminobutyric acidergic signaling and excitation/inhibition balance in the medial prefrontal cortex. Biol. Psychiatry 83, 657–669 (2018).

30. B. Liang, L. Zhang, G. Barbera, W. Fang, J. Zhang, X. Chen, R. Chen, Y. Li, D.-T. Lin, Distinct and dynamic ON and OFF neural ensembles in the prefrontal cortex code social exploration. Neuron 100, 700–714.e9 (2018).

31. M. Murugan, H. J. Jang, M. Park, E. M. Miller, J. Cox, J. P. Taliaferro, N. F. Parker, V. Bhave, H. Hur, Y. Liang, A. R. Nectow, J. W. Pillow, I. B. Witten, Combined social and spatial coding in a descending projection from the prefrontal cortex. Cell 171, 1663–1677 (2017).

32. J. R. Gibson, M. Beierlein, B. W. Connors, Two networks of electrically coupled inhibitory neurons in neocortex. Nature 402, 75–79 (1999).

33. G. Di Cristo, C. Wu, B. Chattopadhyaya, F. Ango, G. Knott, E. Welker, K. Svoboda, Z. J. Huang, Subcellular domain-restricted GABAergic innervation in primary visual cortex in the absence of sensory and thalamic inputs. Nat. Neurosci. 7, 1184–1186 (2004).

34. E. A. Phillips, A. R. Hasenstaub, Asymmetric effects of activating and inactivating cortical interneurons. eLife 5, e18383 (2016).

35. J. Veit, R. Hakim, M. P. Jadi, T. J. Sejnowski, H. Adesnik, Cortical gamma band synchronization through somatostatin interneurons. Nat. Neurosci. 20, 951–959 (2017).

36. M. Pavlova, M. Guerreschi, W. Lutzenberger, I. Krägeloh-Mann, Social interaction revealed by motion: Dynamics of neuromagnetic gamma activity. Cereb. Cortex 20, 2361–2367 (2010).

37. G. Buzsáki, B. O. Watson, Brain rhythms and neural syntax: Implications for efficient coding of cognitive content and neuropsychiatric disease. Dialogues Clin. Neurosci. 14, 345–367 (2012).



Dow

nloaded from

http://advances.sciencemag.org/cgi/content/full/6/30/eaay4073/DC1

http://advances.sciencemag.org/cgi/content/full/6/30/eaay4073/DC1

https://en.bio-protocol.org/cjrap.aspx?eid=10.1126/sciadv.aay4073




13 of 13

38. L. A. Gunaydin, L. Grosenick, J. C. Finkelstein, I. V. Kauvar, L. E. Fenno, A. Adhikari, S. Lammel, J. J. Mirzabekov, R. D. Airan, K. A. Zalocusky, K. M. Tye, P. Anikeeva, R. C. Malenka, K. Deisseroth, Natural neural projection dynamics underlying social behavior. Cell 157, 1535–1551 (2014).

39. G. Dölen, A. Darvishzadeh, K. W. Huang, R. C. Malenka, Social reward requires coordinated activity of nucleus accumbens oxytocin and serotonin. Nature 501, 179–184 (2013).

40. L. W. Hung, S. Neuner, J. S. Polepalli, K. T. Beier, M. Wright, J. J. Walsh, E. M. Lewis, L. Luo, K. Deisseroth, G. Dölen, R. C. Malenka, Gating of social reward by oxytocin in the ventral tegmental area. Science 357, 1406–1411 (2017).

41. F. Farzan, M. S. Barr, A. J. Levinson, R. Chen, W. Wong, P. B. Fitzgerald, Z. J. Daskalakis, Evidence for gamma inhibition deficits in the dorsolateral prefrontal cortex of patients with schizophrenia. Brain 133, 1505–1514 (2010).

42. R. Y. Cho, R. O. Konecky, C. S. Carter, Impairments in frontal cortical gamma synchrony and cognitive control in schizophrenia. Proc. Natl. Acad. Sci. U.S.A. 103, 19878–19883 (2006).

43. A. E. Richard, R. R. Lajiness-O'Neill, S. M. Bowyer, Impaired prefrontal gamma band synchrony in autism spectrum disorders during gaze cueing. Neuroreport 24, 894–897 (2013).

44. H. Xu, H.-Y. Jeong, R. Tremblay, B. Rudy, Neocortical somatostatin-expressing GABAergic interneurons disinhibit the thalamorecipient layer 4. Neuron 77, 155–167 (2013).

45. J. L. Gomez, J. Bonaventura, W. Lesniak, W. B. Mathews, P. Sysa-Shah, L. A. Rodriguez, R. J. Ellis, C. T. Richie, B. K. Harvey, R. F. Dannals, M. G. Pomper, A. Bonci, M. Michaelides, Chemogenetics revealed: DREADD occupancy and activation via converted clozapine. Science 357, 503–507 (2017).

46. J. Courtin, F. Chaudun, R. R. Rozeske, N. Karalis, C. Gonzalez-Campo, H. Wurtz, A. Abdi, J. Baufreton, T. C. M. Bienvenu, C. Herry, Prefrontal parvalbumin interneurons shape neuronal activity to drive fear expression. Nature 505, 92–96 (2014).

47. B. N. Armbruster, X. Li, M. H. Pausch, S. Herlitze, B. L. Roth, Evolving the lock to fit the key to create a family of G protein-coupled receptors potently activated by an inert ligand. Proc. Natl. Acad. Sci. U.S.A. 104, 5163–5168 (2007).

Acknowledgments: We thank D. Anderson, G. Feng, M. Luo, and L.-j. Wu for scientific discussions and the anonymous reviewers for valuable comments and suggestions. We thank Z. J. Huang for SST-Cre mice. We thank W. Xi, Z. Shi, and M. Song for help with electrophysiological recording and data analysis and the technical support by the Core Facilities of Zhejiang University Institute of Neuroscience and the Core Facilities of Zhejiang University School of Medicine. Funding: This research was supported by grants from the National Key R&D Program of China (2016YFA0501000), the National Natural Science Foundation of China (31471025, 91432110), the Zhejiang Provincial Natural Science Foundation of China (LR17H090002), the Fundamental Research Funds for the Central Universities (2019QNA5001), the Chinese Ministry of Education Project 111 Program (B13026), and the Non-profit Central Research Institute Fund of Chinese Academy of Medical Sciences (2017PT31038 and 2018PT31041) to Han Xu. Author contributions: L.L. and J.W. conducted all tetrode and optrode recordings. Haifeng Xu and Y.T. performed optogenetics and pharmacogenetics experiments. J.L. performed immunohistochemistry experiments. J.Z. performed part of LFP recording experiments. M.H., T.-L.X., Z.-Y.W., X.-M.L., and S.-M.D. provided resources. Han Xu supervised the project and wrote the manuscript with help from all contributing authors. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Submitted 21 July 2019Accepted 5 June 2020Published 22 July 202010.1126/sciadv.aay4073

Citation: L. Liu, H. Xu, J. Wang, J. Li, Y. Tian, J. Zheng, M. He, T.-L. Xu, Z.-Y. Wu, X.-M. Li, S.-M. Duan, H. Xu, Cell type–differential modulation of prefrontal cortical GABAergic interneurons on low gamma rhythm and social interaction. Sci. Adv. 6, eaay4073 (2020).



Dow

nloaded from


rhythm and social interactiondifferential modulation of prefrontal cortical GABAergic interneurons on low gamma−Cell type

Shu-Min Duan and Han XuLing Liu, Haifeng Xu, Jun Wang, Jie Li, Yuanyuan Tian, Junqiang Zheng, Miao He, Tian-Le Xu, Zhi-Ying Wu, Xiao-Ming Li,

DOI: 10.1126/sciadv.aay4073 (30), eaay4073.6Sci Adv

ARTICLE TOOLS http://advances.sciencemag.org/content/6/30/eaay4073

MATERIALSSUPPLEMENTARY http://advances.sciencemag.org/content/suppl/2020/07/20/6.30.eaay4073.DC1

REFERENCES

http://advances.sciencemag.org/content/6/30/eaay4073#BIBLThis article cites 47 articles, 6 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.Science AdvancesYork Avenue NW, Washington, DC 20005. The title (ISSN 2375-2548) is published by the American Association for the Advancement of Science, 1200 NewScience Advances

License 4.0 (CC BY-NC).Science. No claim to original U.S. Government Works. Distributed under a Creative Commons Attribution NonCommercial Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of



Dow

nloaded from

http://advances.sciencemag.org/content/6/30/eaay4073

http://advances.sciencemag.org/content/suppl/2020/07/20/6.30.eaay4073.DC1

http://advances.sciencemag.org/content/6/30/eaay4073#BIBL

http://www.sciencemag.org/help/reprints-and-permissions

http://www.sciencemag.org/about/terms-service


neuroscience copyright © 2020 cell type–differential ......liu et al., sci. adv. 2020 6 :...

Documents