Early growth response 2 in the mPFC regulates mouse social and cooperative behaviors

Adolescent social neglect impairs social performance, but the underlying molecular mechanisms remain unclear. Here we report that isolation rearing of juvenile mice caused cooperation defects that were rescued by immediate social reintroduction. We also identified the transcription factor early growth response 2 (Egr2) in the medial prefrontal cortex (mPFC) as a major target of social isolation and resocialization. Isolation rearing increased corticosteroid production, which reduced the expression of Egr2 in the mPFC, including in oligodendrocytes. Overexpressing Egr2 ubiquitously in the mPFC, but not specifically in neurons nor in oligodendroglia, protected mice from the isolation rearing-induced cooperation defect. In addition to synapse integrity, Egr2 also regulated the development of oligodendroglia, specifically the transition from undifferentiated oligodendrocyte precursor cells to premyelinating oligodendrocytes. In conclusion, this study reveals the importance of mPFC Egr2 in the cooperative behavior that is modulated by social experience, and its unexpected role in oligodendrocyte development. Using behavioral assays, transcriptome analysis and viral approaches to manipulate gene expression in the mPFC, Zhang et al. identified an important role for early growth response 2 (Egr2/Krox-20) in the development of social and cooperative behaviors in mice.

S ocial behaviors promote information exchange and action coordination among a group of individuals, and eventually improve the survival fitness of the species. As a kind of high-level social behavior, social cooperation enables the participants to accomplish tasks that are difficult or impossible for an individual 1 . Indeed, the social and cooperative behaviors of species ranging from ants and honeybees to rodents and primates have contributed to their ecological success. Cooperative behaviors also lay the basis for modern human society. Impaired social cooperation is frequently observed in patients with psychiatric or neuronal disorders such as depression, autism, schizophrenia, bipolar disorder, obsessive-compulsive disorder and Alzheimer's disease [2][3][4][5][6][7] . However, the biological and molecular bases underlying social cooperation remain largely undetermined. Addressing this issue will help to explore effective therapeutic strategies for these neuropsychiatric disorders.
Recent studies have identified the medial prefrontal cortex (mPFC) as a critical brain region for social and cooperative behaviors 8 . In humans, childhood neglects such as social isolation (SI) are associated with altered white matter tracts that are connected to the mPFC 9,10 . Oligodendrocytes are responsible for the generation and maintenance of the central nervous system (CNS) myelin sheath. The CNS myelin sheath displays dramatic plasticity during development and myelination continues into the third decade of life in humans 11 . The myelin thickness, the length of the axon initial segment, and the internodal length along the axon are all substrates of experience 12,13 . These parameters combined together affect nerve conduction velocity, and therefore influence the orchestration and temporal integration of neural activities 13 . In agreement with human studies, isolation rearing of weanling mice delays oligodendrocyte maturation and myelination in the mPFC, which is rescuable by rehousing the mice with group-housed (GH) mice but not with SI mice 14,15 . Treating SI mice with clemastine, an antimuscarinic compound that promotes oligodendrocyte differentiation and myelination, successfully reverses the isolation-rearing-induced social avoidance behavior in mice 16 . However, it is unclear whether cooperative behaviors are also affected by social neglect and resocialization. Moreover, the molecular machinery regulating myelin plasticity in the mPFC following SI and/or reintroduction remains largely unknown 17,18 .
In this Article, by the combined use of mouse behavioral assays, transcriptome sequencing (RNA-seq) and adeno-associated virus (AAV)-mediated gene expression manipulations, we identified an important role for early growth response 2 (Egr2/Krox-20)─the established master transcriptional regulator in Schwann cell myelination 19 ─in the development of social and cooperative behaviors. Our results indicate that Egr2 is required for normal oligodendrocyte differentiation and myelination in the mPFC, which is critical for myelin plasticity and the development of social behaviors. This work identified an unanticipated role for Egr2 in the CNS, which could be explored as a potential target for future therapeutic development treating social behavioral defects.

Impaired social interaction and cooperative behaviors in isolation-reared mice.
To evaluate the consequence of isolation rearing on juvenile mice, we randomly divided 3-week-old weanling mice into two groups: SI (one mouse per cage) and GH (five littermates per cage) groups. The social interaction, cooperative behavior and other behaviors were examined 3 weeks later (Fig. 1a). The three-chamber test revealed significantly reduced sociability and social memory in SI mice (Fig. 1b,c) compared with GH mice, which is consistent with previous reports 14 . The mutualistic cooperation was evaluated by a water-reward cooperative task assay that was recently established by our group 20 . The water-reward cooperative assay contains a training phase and a testing phase. The upfront training period is an instrumental learning assay, in which a mouse is expected to complete a position task to turn on a switch and get drinking water as a reward ( Fig. 1d and Supplementary Video 1). In the testing period, two trained mice are put into one cage, where they can get water only if they accomplish the two position tasks simultaneously ( Fig. 1e and Supplementary Video 2). In the training phase, SI mice improved their performance more slowly compared with GH ones in the first 4 days, but they gradually caught up and displayed no significant difference eventually, which means SI mice were equally capable as GH ones to perform the task at the end of the training (day 7, Fig. 1d). In the subsequent testing phase, however, SI mice performed dramatically worse than their GH counterparts. Their co-drinking latency was about three times that of GH mice (F (4, 68) = 70.621, P = 0.0001); their co-drinking number was also drastically reduced (F (4, 68) = 104.823, P = 0.0001), as well as their cumulated co-drinking time (F (4, 68) = 130.235, P = 0.0001) ( Fig. 1e, Supplementary Videos 2 and 3). In the 5 day testing period, with regard to latency, times or total time standing on the drinking platform for each individual mouse, we did not observe any significant difference between the two groups; therefore, the poor cooperative performance of SI mice is a consequence of their defective synchronization, and not of a reduced willingness to drink ( Supplementary Fig. 1a-c). Compared with GH mice, SI mice also displayed impaired spatial working memory, impaired fear learning and memory, reduced locomotion and exploration, and more anxiety-like behaviors (Extended Data Fig. 1a-d), which are all consistent with previous studies [21][22][23] .
Immediate social reintroduction rescues the sociability and cooperation defects of isolation-reared mice. We next explored whether resocialization could rectify isolation-rearing-induced cooperation defects and other behavioral abnormalities. After    3 weeks of SI, we transferred the 6-week-old SI mice into cages containing age-matched GH mice. After 4 additional weeks, these re-socialized SI mice (GH-SI mice) were tested with the same set of behavioral assays (Extended Data Fig. 2a). We found that both social interaction and cooperative behaviors (elevated co-drinking latency (F (1, 80) = 12.933, P = 0.0001), reduced co-drinking number (F (1, 80) = 17.267, P = 0.0001) and reduced co-drinking time (F (1, 80) = 11.907, P = 0.001)) were almost completely rescued by the resocialization treatment (Extended Data Fig. 2b-e).
Resocialization also counteracted the effects of SI on locomotion, exploration, anxiety-like behaviors, spatial working memory, and fear learning and memory (Extended Data Fig. 2f-i).
The mPFC is known to be an important target of isolation rearing [21][22][23] . Therefore, we examined whether myelin and synaptic development was affected in the mPFC of SI mice. Electron microscopy (EM) analysis on ~12-week-old SI mice revealed reduced sheath thickness compared with GH mice, as reflected by an elevated g ratio (0.797 ± 0.009 versus 0.699 ± 0.01). The heterochromatin proportion of chromatin was also reduced in mPFC oligodendrocytes of SI mice, indicating the myelin development might be impaired, given that oligodendrocytes are rich in heterochromatin 24 . Both myelin sheath thinning (g ratio 0.703 ± 0.009) and heterochromatin reduction were rescued by social reintroduction (Extended Data Fig.  3a-c). Similarly, the maturation state of synapses was also affected by isolation rearing and normalized by resocialization, as revealed by EM analysis (Extended Data Fig. 3a,d). The expression levels of the mature myelin marker protein myelin basic protein (MBP) and the postsynaptic marker postsynaptic density protein-95 (PSD-95) were reduced in the mPFC of SI mice, both of which were normalized by the resocialization treatment as revealed by immunostaining and western blot analyses (Extended Data Fig. 3e-h).
Egr2 is a major target of SI and resocialization. To explore the molecular mechanisms behind SI-induced behavioral alterations, we performed RNA-seq on mPFC samples collected from ~8-week-old GH or SI mice. We identified 1,265 differentially expressed genes (DEGs) that were enriched in four pathways: cognitive behavior, myelin development, synaptic development and ion channels ( Fig. 2a-c). We validated the most affected DEGs with quantitative real-time PCR (qPCR) and western blot, and identified four genes consistently showing reductions in expression in SI mice versus GH mice: Egr2, Arc, Egr1 and neurotensin (Nts) (Fig. 2d-i). We further found that the expression levels of Egr1, Egr2 and Arc were restored in GH-SI mice ( Fig. 2j-l). Intriguingly, the promoter regions of both Arc and Egr1 contain potential binding sites for Egr2, according to the transcription factor binding site prediction tool JASPER (http:// jaspar.genereg.net). We also performed dual-luciferase reporter experiments in 293 cells, which further validated that Egr2 could transcriptionally upregulate the expression of Arc and Egr1 (Fig. 2m,n). Therefore, we decided to focus on Egr2 in the subsequent experiments, to explore whether it mediated the effects of SI and/or resocialization on mouse sociability and cooperation.
Egr2 encodes a zinc finger transcription factor that is important for Schwann cell myelination and hindbrain development 25,26 . Co-labeling of Egr2 with neuronal marker class III β-tubulin (Tuj1) on brain sections of GH mice showed that Egr2 was expressed in some but not all mPFC neurons ( Supplementary Fig. 2a). To date, Egr2 has been regarded as not involved in the CNS myelination process. However, recent transcriptome sequencing studies have shown that Egr2 is also expressed in certain oligodendroglia-lineage cells 27-29 . In agreement with these studies, our immunofluorescence analysis in the mPFC of GH mice revealed that Egr2 was also expressed in cells positive for many oligodendrocytes marker proteins, including the oligodendrocyte glial lineage marker SOX10, the early and middle oligodendrocyte marker O4, and mature oligodendrocyte markers CC1 and CNPase ( Supplementary Fig. 2b). By contrast, Egr2 was not or rarely expressed in astrocytes or microglia ( Supplementary  Fig. 2c). The NG2 proteoglycan is routinely employed as a marker for oligodendrocyte precursor cells (OPCs). In the mPFC, we found reduced numbers of Egr2 + cells and Egr2 + NG2 + cells in SI mice compared with GH or SI-GH mice, suggesting Egr2 in OPCs is possibly targeted by the isolation rearing ( Supplementary Fig. 3a,b).

Reducing Egr2 expression specifically in the mPFC impairs mice's social and cooperative behaviors.
To examine whether Egr2 is really required for mice's social and cooperative behaviors, we generated an AAV9 virus expressing a shRNA targeting all Egr2 isoforms under the U6 promoter, together with a separate GFP expression cassette. The 3-week-old weanling mice received virus injection into the bilateral mPFC and were then reared in isolation (SI-Egr2-RNAi) or group-housed (GH-Egr2-RNAi) for 3 weeks, followed by a series of behavioral tests and pathological analyses (Fig. 3a). We observed clear GFP signals in both neurons and oligodendrocytes, indicating successful virus transduction (Fig. 3b). Immunohistology and western blot analyses revealed significantly reduced Egr2 expression in GH-Egr2-RNAi mice, but not in SI-Egr2-RNAi mice compared with their respective controls ( Supplementary Fig. 4a-c). This could be explained by the fact that isolation rearing had already reduced Egr2 expression to a very low level in SI mice.
GH-Egr2-RNAi mice displayed reduced performance in sociability and social memory as revealed by the three-chamber assay (Fig. 3c,d). In the cooperation test, they did not show any abnormality in the training period (Extended Data Fig. 4a); however, in the subsequent testing period, their performance was obviously worse than that of control GH-GFP mice, performing similarly to the SI-GFP group. Specifically, the co-drinking latency of GH-Egr2-RNAi mice was almost doubled (F (4, 156) = 11.907, P = 0.001), and their co-drinking number (F (4, 156) = 16.948, P = 0.0001) and co-drinking time (F (4, 156) = 9.786, P = 0.003) were reduced by more than 50% at the end of the test compared with those of GH-GFP mice (Fig. 3e). By contrast, knocking down Egr2 did not affect the locomotion or exploration behaviors of the mice, nor did it change anxiety-like behaviors, spatial working memory and fear memory (Extended Data Fig. 4b-e). , which identified 1,265 differential genes with q < 0.001. The red points represent q < 0.001, log 2 fold change >0.7 or log 2 fold change <−0.7, count read >100 genes. b, Clustering of the top 40 groups of DeGs, which are enriched in four pathways: myelin development, synaptic development, learning and memory, and ion transport. c, For each pathway, the ten most altered genes are shown. d,e, qPCR confirmation of gene expression differences in mPFC between GH and SI mice (n = 6 for each group). f-i, Western blot showing protein expression for candidate genes showing differential expression by qPCR (f and h; shown in d and e) and the corresponding quantification (n = 4 for each group) (g and i). j, qPCR analysis on mPFC samples from GH, SI and GH-SI (resocialization) mice at P87 (n = 4 for each group) and evaluating expression of candidate genes. k,l, Western blot evaluating protein expression in GH, SI and GH-SI mice (n = 6-8 for each group) of the four genes that showed differential expression by qPCR (shown in j). m,n, Arc promoter-driven (m) and Egr1 promoter-driven (n) luciferase activities in HeK 293 cells with the indicated transfections (n = 5 per group). Data are presented as median and IQR. Data in a-i were analyzed by two-tailed Student's t-tests; data in j and l were analyzed by one-way aNOVa followed by Tukey's post-hoc test, and data in m and n were analyzed by two-way aNOVa followed by Tukey's post-hoc test. GH: group housing (P59 mice in a-i and P87 in j-l); SI: social isolation (P59 mice in a-i and P87 in j-l); GH-SI: resocialization of SI mice with GH mice (P87 mice in j-l).       Overexpressing Egr2 in the mPFC protects mice from isolation rearing-induced sociability and cooperation defects. Next, we examined whether overexpressing Egr2 could protect mice from the detrimental effect of isolation rearing, by using an AAV9 virus with a CMV promoter driving the expression of a full-length Egr2 fused with EGFP (Fig. 4a). Immunohistology and western blot analyses revealed a significant increase in Egr2 expression following virus transduction in the mice reared in isolation (SI-Egr2), but not in GH mice (GH-Egr2) (Supplementary Fig. 4d-f). GFP expression was spotted in cells expressing the oligodendroglia markers NG2, PDGFRα or CC1, as well as in cells positive for Tuj1, suggesting successful Egr2 overexpression in both oligodendroglia and neurons (Fig. 4b). The three-chamber assay revealed that Egr2 overexpression improved sociability performance of SI-Egr2 mice, compared with SI-GFP controls (Fig. 4c,d). In the cooperation test, SI-Egr2 mice did not show any obvious alterations in performance during the training period (Extended Data Fig. 5a), but they displayed a robust improvement in performance during the testing period compared with SI-GFP mice (co-drinking latency (F (4, 192) = 5.506, P = 0.023); co-drinking number (F (4, 192) = 22.172, P = 0.0001); co-drinking time (F (4, 192) = 16.916, P = 0.0001)) ( Fig. 4e). By contrast, SI-Egr2 mice showed no improvement in the exploring behavior, spatial working memory or fear memory test (Extended Data Fig. 5b,d,e), but they displayed mildly less anxiety-like behavior as revealed in the elevated plus-maze (EPM) test (Extended Data Fig. 5c).
In contrast to the RNAi results where Egr2 knockdown affected only myelin in GH mice, overexpressing Egr2 mitigated both myelin and synaptic impairments in the mPFC of SI mice, as supported by the EM results as well as the quantitative analyses of MBP and PSD-95 expression levels (  Fig. 4d-f). Therefore, our results suggest that Egr2 is not essential for normal synaptic activity in juvenile mice, but when overexpressed it might be able to attenuate the synaptic impairment caused by isolation rearing.
Neither neuron-nor oligodendroglia-specific overexpression of Egr2 is sufficient to protect mice from isolation-rearing-induced cooperation defects. The results reported above showed that ubiquitous Egr2 overexpression in the mPFC protected the mice Weaning GFP/DAPI  from isolation rearing-induced social and cooperation abnormalities, but the exact cellular basis for these effects remained unclear. Given that oligodendrocytes are known to be important player in social behaviors 14 , we examined whether oligodendroglia-specific Egr2 overexpression would be sufficient to protect the mice from the effects of SI rearing (Fig. 5a). We overexpressed Egr2 under the CNP promoter via AAV9 virus injection in mice reared in isolation (SI-CNP-Egr2) (Extended Data Fig. 6a-d). Our immunofluorescence results confirmed that GFP was expressed in NG2/ PDGFRα/CC1-positive oligodendrocytes, but not in Tuj1-positive neurons (Fig. 5b).
Isolation rearing increases corticosteroid production, which might consequently reduce Egr2 expression. We next investigated how social treatments might regulate Egr2 expression. In human Jurkat T cells, the glucocorticoid-induced leucine zipper (GILZ) protein inhibits Egr2 transcription 30 . As corticosterone is a rodent glucocorticoid, we investigated whether corticosterone regulates mouse Egr2 expression and examined serum corticosterone levels in GH, SI and GH-SI mice 31 . Primary mouse OPCs-derived oligodendrocytes were exposed to 1 μM corticosterone for 48 h as described previously 32 , which resulted in reduced Egr2 mRNA and protein levels (Fig. 6a-c). When measuring serum corticosterone levels in mice, we found that isolation rearing of juvenile mice stimulated corticosterone production (Fig. 6d). Serum corticosterone level was significantly increased in SI mice compared with GH mice from the second week and was restored to a normal level after ~6 weeks of resocialization treatment (Fig. 6d,e). In addition, in ~12-week-old SI mice, serum corticosterone level was inversely correlated with Egr2 expression in mPFC at both mRNA and protein levels (Fig. 6f,g). To further demonstrate that the effects of corticosterone on Egr2 expression were mediated by oligodendrocytes, we showed that glucocorticoid receptor (GR) was expressed in cells that were also expressing various oligodendroglia markers, including the OPC marker NG2, the glial lineage marker Olig2 and the early-stage marker PDGFRα in the mPFC (Fig. 6h). Ser211 phosphorylation is a marker for activated GR in vivo 33 . In the mPFC of ~12-week-old SI mice, we also found a significant increase in the number of NG2 + phosphor GR (P-GR + ) cells compared with GH mice, indicating increased GR activation in oligodendrocytes; the increase in NG2 + P-GR + cells was diminished in the GH-SI group (Fig. 6i,j). Together, these results strongly suggest that increased corticosterone signaling in oligodendrocytes after isolation rearing leads to Egr2 reduction in the mPFC.

Egr2 regulates oligodendrocyte differentiation and maturation in culture.
To examine the contribution of Egr2 to oligodendrocyte differentiation and maturation, we analyzed Egr2 transcript expression levels in oligodendrocytes at different stages. We cultured primary OPCs and then differentiated them into oligodendrocytes. In the differentiation process, we evidenced a one-pulse elevation of Egr2 on DM2 (2 days after the initiation of differentiation), as seen by immunofluorescence and qPCR analysis (Fig. 7a,b). By preventing this pulse of Egr2 expression in primary oligodendrocytes via shRNA targeting Egr2 ( Supplementary Fig. 4g), we observed a significant reduction in Nkx2.2 and Myrf expression, while both Id2 and Id4 expression were increased (Fig. 7c). Id2 and Id4 are known to maintain the progenitor state and inhibit premature differentiation 34 , while Nkx2.2 and Myrf are required for the transition to and survival of premyelinating oligodendrocytes 35,36 . Therefore, our results suggest that Egr2 knockdown dampens the transition from OPCs into premyelinating oligodendrocytes. Consistent with the qPCR result, small interfering RNA (siRNA)-mediated knockdown of Egr2 dramatically reduced process complexity in cultured mature oligodendrocytes (Fig. 7d-f). In addition, we observed a clear positive association between Myrf levels and Egr2 levels in the mPFC of our mouse models (Fig. 7g,h). The above data revealed a necessary role of Egr2 in the differentiation and maturation of oligodendrocytes.
Taken together, as illustrated in Fig. 7i, our findings showed that isolation rearing of juvenile mice led to social and cooperation defects that could be rescued by immediate social reintroduction. Egr2 in the mPFC seems to be a major target of these social treatments, as reducing or increasing Egr2 expression impaired or improved social interaction and cooperation performance in juvenile mice respectively. In addition, we found that, at least in some mPFC OPCs, Egr2 expression is required for the differentiation into premyelinating oligodendrocytes. The altered glucocorticoid-Egr2 signaling probably mediates some of the adverse effects of long-term SI on mPFC myelination.

Discussion
Rodents are prosocial animals that cooperate with one another. Our three-chamber test indicates that isolation rearing of juvenile mice impairs sociability, which can be rescued by immediate resocialization; these results are consistent with previous reports 14, 15 . Here we demonstrate that cooperative behavior is also a target of SI and resocialization. Traditionally, cooperation behavior is studied in rat models, but the mouse offers the advantage that more disease-related models are available. Recently, Shin et al. evaluated mice's cooperation performance with a Shank2/Shank3 knockout mouse model by using the degree of synchronization as a proxy and a 20% sucrose solution as a reward 37 . Another recent study used a coordinated platform-locating task that is conceptually similar to our assay for assessing the cooperation ability of Lister Hooded rats 38 , with food pellets served as the reward. We found that the addition of sugar in drinking water does not affect mice's cooperative behavior; however, blocking the physical contact between mice with a grille partition dampens cooperation performance 20 . Altogether, recent work from our group and others indicate that it is possible to develop mouse behavioral assays for evaluating mutualistic cooperation, which might be particularly useful for the study of neurological disorders with deficits in social cooperation.
Previous studies have established a strong causal chain that links SI to mPFC myelination defect and further to social behavioral deficits. Recent work from our lab showed that microRNA 124 (miR-124) is involved in this process and has a suppressing role in oligodendroglia development 39 . Another study revealed that an antifungal agent miconazole could improve APP/PS1 mice's cooperative behavioral performance and promote mPFC myelination 40 , which also supports a close relationship between mPFC myelination and mouse cooperation.
Egr2 has an established role in the myelination process in peripheral Schwann cells. In these cells, Egr2 expression is promoted by Oct6 and SOX10 binding to its promoter and enhancer regions, respectively 26 , and Egr2 promotes the transition of Schwann cells from the premyelinating stage to a terminally differentiated myelination stage 41 . The myelination in CNS has been regarded as using a different and Egr2-unrelated molecular machinery. However, recent RNA-seq studies have revealed that Egr2 is expressed in some oligodendroglia-lineage cells [27][28][29] . In particular, using single-cell sequencing, Marques et al. identified a specific Egr2-expressing population of oligodendrocytes in the CA1 hippocampus 29 .  a and b) and qPCR analysis (n = 4) (c) showing reduced egr2 expression after corticosterone treatment of primary mouse oligodendrocytes. egr2 expression is normalized to GaPDH. d, eLISa measurement of corticosterone levels in the serum of mice at weeks 1, 2 and 3 in GH or SI treatment (aged 4, 5 and 6 weeks, respectively, at the time of measurement; GH: n = 9, 9 and 11 mice, at weeks 1, 2 and 3 respectively; SI: n = 9, 9 and 21 mice, at weeks 1, 2 and 3 respectively). e, eLISa measurement of corticosterone levels in the serum in mice at week 4 of GH, GH-SI or SI treatment (aged 7 weeks at the time of measurement; GH: n = 7 mice; GH-SI: n = 6 mice; SI: n = 7 mice). Samples in each group were collected at postnatal day 87 (P87) for eLISa analysis. f,g, Linear fitting showing negative correlation of serum corticosterone levels with egr2 protein levels (f, n = 8) or mRNa levels (g, n = 12) in the mPFC of SI mice at P59. h, Representative immunofluorescence images showing GR expression in NG2-labeled, Olig2-labeled and PDGFRα-labeled oligodendrocytes in the mPFC of P59 SI mice. The fluorescence intensity diagrams on the right show the co-localization of GR and NG2, GR and Olig2 or GR and PDGFRα. Scale bars, 50 µm. i,j, Representative immunofluorescence images and quantified analysis of NG2 and P-GR expression in the mPFC of GH, SI and GH-SI (resocialization) mice at P87 (n = 6). Scale bars, 50 µm. Data are represented as mean ± s.e.m. in b and d; and as medians and IQR in c, e and j. Data in b-d were analyzed by two-tailed Student's t-tests; data in e and j by one-way aNOVa followed by Tukey's post-hoc test; and data in f and g by Pearson correlation coefficient test. GH, group housing; SI, social isolation; GH-SI, resocialization of SI mice with GH mice. O4 and MBP serve as markers for early-and late-differentiated oligodendrocytes, respectively (n = 28 cells for siNC; n = 35 cells for siegr2). g,h, Representative western blot bands and the corresponding quantification graphs showing Myrf protein expression in 2-month-old mPFC of GH and SI mice with egr2 overexpression or knockdown (n = 5-6 for each group). i, Illustration showing that SI on weanling mouse leads to increased production of glucocorticoids, which acts on differentiating OPCs in mPFC and reduces egr2 expression in these cells; egr2 regulates the development of oligodendrocytes and therefore mPFC myelination, which explains why egr2 depletion dampens the development of cooperative behaviors in mice. Data are presented as median and IQR. Data in b were analyzed by one-way aNOVa followed by Tukey's post-hoc test; data in c and f were analyzed by two-tailed Student's t-tests; and data in g and h were analyzed by two-way aNOVa followed by Tukey's post-hoc test. DM1, DM2 and DM3: the first, second and third day of OPC culture in differentiation medium. GH-GFP: GH mice with aaV-GFP virus injection (P61); GH-egr2: GH mice with aaV-egr2 virus injection (P61); SI-GFP: SI mice with aaV-GFP virus injection (P61); SI-egr2: SI mice with aaV-egr2 virus injection (P61); GH-egr2-RNai: GH mice with aaV-egr2-RNai virus injection (P61); SI-egr2-RNai: SI mice with aaV-egr2-RNai virus injection (P61); OPCs: oligodendrocyte progenitor cells; PM9: the ninth day of OPCs in proliferation medium. These studies agree with our finding that Egr2 is expressed at least in some oligodendroglia-lineage cells in the mPFC (Fig. 3b and Supplementary Figs. 2b and 3a). Egr2 transcriptionally upregulates myelin proteins and enzymes to produce myelin sheath in Schwann cells 26,42 . Overexpressing Egr2 in 3T3 fibroblast activates the expression of myelin genes such as periaxin and myelin protein zero 43 . As revealed by the present work, Egr2 expression in oligodendroglia-lineage cells can produce a similar effect. We hypothesize that oligodendroglial Egr2 might also produce some 'peripheral nervous system-specific' myelin proteins in the CNS, resulting in a mixed myelin that can still function efficiently. In adult CNS demyelination models, OPCs not only produce oligodendrocytes, but also Schwann cells, which are functional and are actively investigated as cell sources for transplant therapy 44,45 .
In 3-week-old mice, the synapse development is completed while CNS myelination occurs later and lasts until the end of adolescence 46 . This temporal difference could partly explain why reducing Egr2 expression in the mPFC of juvenile mice did not affect synaptic development but inhibited myelination. On the other hand, the Egr family contains five members with overlapping functions; the effect of Egr2 reduction might have been possibly compensated by other family members such as Egr1 (ref. 47). In the present study, overexpressing Egr2 in the mPFC, either ubiquitously or specifically in neurons, improved PSD-95 level and other synaptic parameters (Fig. 4f,i, Supplementary Fig. 6 and Extended Data Fig. 10). This enhancement of synaptic strength agrees with a previous study, in which researchers suggested that Egr2 is required for long-term potentiation stabilization 48 .
In the EPM test, both CNP-Egr2 and hSyn-Egr2 overexpression significantly increased SI mice's time spent in the open arm, suggesting reduced anxiety level. Anxiety level is usually associated with one's performance on social behaviors 49 ; indeed, patients with generalized social anxiety disorder display fear or avoidance of social places, which seriously affects their social life 50,51 . However, in our work, hSyn-Egr2 overexpression did not improve SI mice's performance on sociability. Two factors might contribute to this seeming discrepancy. (1) The reduction in anxiety reduction is not sufficient; in the open field test, AAV-hSyn-Egr2 mice did not spend more time in the center zone, suggesting the amount of anxiety reduction revealed by the EPM test is probably small. (2) There are two forms of anxiety: social anxiety and non-social anxiety, and the EPM test measures only the non-social form 52 ; social anxiety, which should be the one closely associated with social behavioral performance, might not change at all.
Contrary to hSyn-Egr2 overexpression, CNP-Egr2 overexpression rescued the sociability defect in SI mice, confirming that oligodendrocytes have a key role in social behavior. This is consistent with previous reports that mPFC myelin formation is necessary for the establishment of social interaction and social memory [14][15][16] .
However, social cooperation behavior could only be improved by the ubiquitous CMV-Egr2 overexpression, while neither neuronal nor oligodendroglial Egr2 overexpression was sufficient to improve cooperation in SI mice. These results suggest that synapse and myelin integrity are both required for the establishment of social cooperative behavior. The cooperative drinking water assay used in the present study is an instrumental learning task, for which coordinated neuronal activity is required 38 . Indeed, using a similar assay in a rat model, researchers identified highly associated local field potentials during cooperative behaviors 38 . Here the cooperative drinking water reward task lasts 12 days, with a relatively high difficulty coefficient. By contrast, the three-chamber sociability test is obviously simpler, which partly explains why correcting mPFC hypomyelination successfully rescues sociability impairments but not the cooperation defects.
In the present study, we evidenced a gradual increase in serum glucocorticoid levels in the first 3 weeks of isolation rearing, which was counteracted by resocialization. Egr2 transcript and protein levels in the mPFC were inversely correlated with SI mouse's serum glucocorticoid concentration; the maturation of cultured mouse oligodendrocytes was impaired when exposed to glucocorticoid, along with a reduction in Egr2 expression. These results suggest that SI-induced glucocorticoids overproduction could be a key factor leading to Egr2 reduction, which would further impair mPFC myelination. A regulation of Egr2 by glucocorticoid has been shown before. In Jurkat T cells, glucocorticoid-induced expression of GILZ reduces Egr2 expression by inhibiting the NFAT/AP-1 signaling 30 ; in osteoblast cultures, glucocorticoids decrease Egr2 expression by inhibiting the Wnt pathway 53 . It has been reported that corticosterone exacerbates the loss of white matter tract in a stroke mouse model 54 , and that prolonged corticosterone treatment inhibits the proliferation of OPCs in adult rats 55 . Our results suggest that corticosteroid might also target OPC differentiation by reducing Egr2 expression, which leads to a myelination defect as observed in SI mice.
Egr2 is expressed in various brain regions, including the neocortex, hippocampus, amygdala, olfactory bulb, striatum, cerebellum, diencephalic and brainstem structures 56,57 . Whether there is a brain region specificity in the role of Egr2 needs to be clarified. A recent study has reported that the neuropeptide Tac2 coordinates complex behavioral effects after chronic SI stress in a distributed manner 58 . Whether Egr2 similarly regulates the different behavioral effects of SI in different brain regions is worth exploring.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41684-022-01090-0.

Animals and treatment.
Male C57BL/6J mice were employed in the present study. Weanling mice (3 weeks old) were randomly divided into GH and SI groups. Five mice (for GH group) or a single mouse (for SI group) were housed in a plexiglass cage (31 cm × 22 cm ×15 cm) for 3 weeks as the social treatments (Fig. 1a). For resocialization, two SI mice were then put in a cage containing three GH mice (GH-SI) for 4 weeks; similar to the GH-SI group procedure, two GH mice from one cage were housed together with three GH mice from another cage in the GH group. For the SI group, however, the SI mouse was always reared in a separate cage (Extended Data Fig. 2a). Thus, when performing resocialization experiments, for a GH mouse, the cumulated group-housing time is 66 days (21 days of initial GH + 45 days of co-housing with three other GH mice including 17 days of a series of behavioral tests described later); for an SI mouse, the cumulated isolation-rearing time is also 66 days; for a GH-SI mouse, the isolation-rearing time is 21 days, and thereafter there are 45 days co-housing with three other GH mice cumulatively.
To increase Egr2 expression, the mPFC region of 3-week-old weanling mice was injected with AAV viruses overexpressing Egr2 under the ubiquitous CMV promoter (AAV-Egr2-EGFP), the oligodendrocyte-specific CNP promoter (AAV-CNP-Egr2-EGFP-3Flag) or the neuron-specific hSyn promoter (AAV-hSyn-Egr2-EGFP-3Flag), respectively; to knock down mPFC Egr2 expression, an AAV9 virus expressing an shRNA targeting Egr2 under the U6 promoter (AAV-Egr2-RNAi) was employed (for timeline of experiments, see Figs. 3a, 4a and 5a and Extended Data Fig. 8a). Three weeks later, behavioral tests, which lasted for 17 days, were performed on these mice. During this period, the mice in each group were maintained in their original housing conditions. Mice were maintained on a 12 h light/dark cycle with ambient temperature (18-22 °C) and humidity (30-50%), with free access to food and water. All animal experiments were conducted following the international standards on animal welfare and the guidelines of the Institute for Laboratory Animal Research of Nanjing Medical University (approval no. IACUC: 2004013).
Behavioral tests. Data on mouse activity in the following behavioral tests were collected using a computer-connected digital video camera (TopScan, CleverSys). Room lights, except a dim light, were switched off during each experimental session for mouse comfort. Before each test, the apparatuses were cleaned with 70% alcohol to eliminate possible remaining olfactory cues. All tests were performed by two independent experimenters, who were each blinded to the treatment schedule.
Three-chamber test. The sociability and social novelty behaviors were assessed using an apparatus with three compartments (40 cm × 40 cm × 30 cm), with two open squares (8 cm × 8 cm) allowing access into each chamber 59,60 . In the sociability test period, a mouse (mouse A) was first placed in the middle compartment, spent 5 min to explore the apparatus before being taken out; then, another 3-week-old male mouse (stranger 1) in a circular wire mesh cage (diameter 9 cm, height 15 cm) was placed into one of the side compartments, and mouse A was put back into the middle compartment. Mouse A's exploring behavior in the following 5 min was recorded and analyzed. The social novelty test begins 10 min after finishing the sociability test. In this period, an additional 3-week-old male mouse (stranger 2) was placed into the other side compartment. Then mouse A was put back into the middle compartment, and its respective time interacting with the two mice was recorded and analyzed.
Cooperative test. The cooperative ability of mice was evaluated by a cooperative drinking device developed by our laboratory (Fig. 1d,e and Supplementary Videos 1 and 2) 20,40 . The device is a plexiglass cage (60 cm × 40 cm × 35 cm) equipped with two water valves from which drop shape water could be delivered by turning on two serial photoelectric switches. During the first phase of training (the training period), one photoelectric switch was defaulted to ON; a mouse was placed into the chamber, and, during a 5 min period, it was able to explore and turn on the other switch to get water rewards. The training was conducted over 7 consecutive days, with two trials per day. Data on the latency for the first successful drinking (drinking latency), as well as the number of times mice were drinking (drinking number) and total time spent drinking (drinking time) in each trial, were collected. During the second phase (the testing period), both switches were defaulted to OFF. Two trained mice were placed in the chamber and allowed to move freely for 10 min. They could get the water reward together only if they turned on both switches at the same time. The co-drinking latency, times and time of co-drinking were recorded. The test was carried out once a day for 5 days. The mice were returned to their original cages after each trial. To increase their incentive for getting drinking water, we made water unavailable from 6 h before each trial.
Open-field test. The open-field test was used to evaluate the locomotion and anxiety-like behaviors 61 . The open field consisted of a square blue box (60 cm × 60 cm × 25 cm), with an outlined center area (30 cm × 30 cm). Each mouse was placed in the center of the box and allowed to move freely for 5 min within the box. The time spent in the center area, times of entering the center area and total distance traveled were calculated during the test.
EPM test. The anxiety-like behavior was also evaluated by the EPM, which was composed of four arms (50 cm × 10 cm) connected by a central square (10 cm × 10 cm) and elevated 100 cm above floor 23 . Two opposite arms were open, while the two remaining were enclosed with 40-cm-high walls. The mouse was placed into the enclosed arm and allowed to freely explore the maze for 5 min. The percentage of time spent and the frequency of entries into the open arms were calculated.
Y-maze test. The spatial working memory was examined using a Y-shaped maze 23 . The maze consisted of three arms (8 cm × 30 cm × 15 cm), with an angle of 120° in-between. The procedure included a 5 min training stage and a 5 min testing stage, with an interval of 2 h. During the first stage, one arm (the new arm, NA) was blocked by a baffle, allowing the mouse to move freely in two other arms for 5 min. During the second stage, the NA was opened and the mouse could freely explore all three arms for 5 min. The percentage of time traveled in the NA and the number of entries into the NA were analyzed.
Fear conditioning test. The mouse was placed in the conditioning chamber for 3 min as a habituation period followed by one tone-foot-shock (tone, 30 s, 70 dB, 1 kHz; foot-shock, 2 s, 0.8 mA) (ref. 62). The stimulation was given every 2 min within 6 min. The mouse was placed in the chamber again with tone-shock pairing (tone, 30 s, 70 dB, 1 kHz) for 3 min after 24 h. Freezing behavior, defined as the absence of all visible movement of the body except the movement necessitated by respiration, was scored.
RNA sequencing. After the behavioral tests, the mice were killed and total RNA from the mPFC tissues of P59 GH and SI mice was extracted using RNAiso Plus (Takara, #9109) according to the manufacturer's instructions. The RNA quality and quantity were measured using NanoDrop 2000 (Thermo Scientific). RNA library was constructed using the BGISEQ-500 platform according to the manufacturer's instruction; then, adapters were ligated to each end of the RNAs and subsequently reverse transcribed to create single-stranded cDNA, which were sequenced on the BGISEQ-500 platform with a read length of 50 bps. Stringent criteria were set to determine significantly dysregulated genes: adjusted q value < 0.01 and log 2 FC >0.7. mPFC samples from SI and GH mice were analyzed as two independent datasets. Heatmaps representing z-scores were generated using the seaborn package in Python. All RNA-seq datasets are available at GEO (GEO number GSE162343).

Stereotaxic injection.
Deeply anesthetized mice were fixed in a stereotaxic instrument with the skull surface exposed. A total of 1.5 µl of AAV-Egr2, AAV-CNP-Egr2, AAV-hSyn-Egr2 and AAV-Egr2-RNAi (1 × 10 12 v.g./ml) or control virus (AAV-GFP, AAV-CNP-GFP and AAV-hSyn-GFP) were infused bilaterally into the mPFC at the following coordinates: +1.8 mm anterior/posterior, −1.5 mm dorsal/ventral and ± 0.4 mm medial/lateral relative to Bregma use 33-gauge syringe needles (Hamilton) 63 . The infusion rate was 0.2 µl/min, and the cannula was left in place for 5 min following completion of the infusion. Two days after the injection, the mice were group-housed or subjected to isolated housing. All behavioral experiments were performed 23 days after the injection.
Corticosterone treatment. OPCs were differentiated for 24 h as described above, and then treated with 1 μM corticosterone (Sigma-Aldrich, #C2505) for 48 h before harvesting for qPCR or western blotting.
Luciferase reporter assay. The relevant plasmids were transfected into HEK 293 cells with Lipofectamine 2000 (Invitrogen, #52887). The luciferase reporter assay for Arc/ Egr1 promoter activity was performed according to the manufacturer's instructions (Promega, E1910). Briefly, 0.05 μg luciferase reporter plasmid, 0.2 μg Egr2 expression plasmid and 1.25 ng Renilla were co-transfected into the 80-90% confluence 293 cells; 48 h after the transfection, cells were harvested and resuspended in 50 μl passive lysis buffer, and then placed on a micro-oscillator and split for 15 min. The supernatant was used to measure luciferase activity; the normalized values (Renilla/ firefly activity) were used for analysis. Experiments were performed in triplicate.
Corticosterone assay. After the behavioral tests, orbital blood was collected from mice, placed into ice-cooled centrifugal tubes and centrifuged (1,700g, 10 min, 4 °C) to separate serum, which was stored at −80 °C until the assay. Blood sampling took place between 09:00 and 12:00 (ref. 65). Corticosterone was quantified by enzyme-linked immunosorbent assay (ELISA; Parameter Corticosterone Assay R&D Systems, #KGE009) according to assay instructions. All samples were run in duplicate, counterbalanced across plates, and were within the standard curve. Final values were determined by averaging the results of duplicated samples.

Brain sample preparation.
After finishing the behavioral tests, mice were subjected to anesthetization and cardiac perfusion of paraformaldehyde (PFA; 0.9% saline for 2 min, then 4% PFA for 7 min). Brain tissues were then fixed overnight in 4% PFA. For immunohistochemical staining, brain tissues were dehydrated in a series of graded ethanol solutions, embedded in paraffin and continuously sectioned to 5 μm thickness using a paraffin slicing machine (Leica RM2245). For immunofluorescence, brain tissues were dehydrated overnight in PBS containing 30% sucrose, and then embedded in optimal cutting temperature compound (Sukura, #4583). Tissues were cut at a thickness of 15 μm on a Leica CM1950 cryostat with super frost plus slides installed and then stored at −20 °C. For EM, mice were perfused with 0.01 M PBS for 2 min, followed by 2.5% glutaraldehyde plus 2% PFA in 0.1 M PB for 5 min after anesthesia. The mPFC was rapidly dissected from brain and post-fixed (4% PFA and 2.5% glutaraldehyde in 0.1 M phosphate buffer containing 0.5% NaCl) for at least 1 week at 4 °C. For qPCR and western blot, the anesthetized mice were perfused with 0.9% saline, and then the mPFC was rapidly dissected out and stored in −80 °C until use.
For immunofluorescence staining, frozen sections were air dried for 1 h at room temperature and rinsed with 1× PBS, blocked and permeabilized in blocking solution (5% bovine serum albumin in PBS) containing 0.1% Triton X for 1 h at room temperature, incubated overnight at 4 °C with primary antibodies anti-Tuj1 EM. The mPFC samples were stained with 70% ethanol containing 0.5% uranyl acetate for 1 h, and dehydrated in a serial dilution of ethanol, then cleared in propylene oxide, embedded in Epon and incubated at 60 °C for 24 h. The tissues in Epon blocks were then trimmed and reoriented so that ultrathin (70 nm) cross-sections were cut using an ultramicrotome (Leica EM UC7).
Image capture and analysis. EM was performed on a FEI Tecnai G2 electron microscope (FEI Company). The layer II/III pyramidal neurons in the anterior cingulate of mPFC were imaged at 10,000×. For each mouse, 25-40 myelinated axons were analyzed. In EM images, fusiform cells depicting a thin rim of perinuclear cytoplasm and containing microtubules but no intermediate filaments nor glycogen granules, were regarded as oligodendrocytes 15,67,68 . For myelin analysis, the heterochromatin was selected using the threshold tool and reported as the percentage of total nuclear area 67,69 ; the g ratio was obtained via the diameter of the axon divided by the diameter of the entire myelinated fiber as previously described 15 . For synapse analysis, an average of ten measurements of synaptic cleft width evenly spaced across the synapse was taken as a reading for each synapse, where the ends of the synapse were defined as the ends of the electron-dense postsynaptic density (PSD); the synaptic curvature was then calculated as the ratio of the arc length of the active zone to its corresponding chord length 70 .
Immunohistochemistry images were captured on a fluorescence microscope Leica DM4000 B (Leica Microsystems) with a 5×/0.12 objective or a 40×/0.75 objective. Immunofluorescence images were from a LSM710 confocal microscope (Zeiss) with a 10×/0.45 objective or a 40×/0.95 objective. Pixel size was 110 nm on xy plane; pinhole diameter was 90 μm; four (for cultured oligodendrocytes) or six (for brain sections) z-stacks were acquired at a 0.97 μm interval. All light and electron micrographs were analyzed with ImageJ (NIH). The mean integrated optical density was measured to assess the immunostaining intensity of MBP, PSD-95 and Egr2 on immunohistochemical staining sections. The amounts of P-GR + , P-GR + NG2 + , Egr2 + and Egr2 + NG2 + cells in the mPFC were manually counted and normalized to the area. The center of a cell was defined manually, and Sholl analysis was performed by starting with a circle with a diameter of 0.05 μm and expanding the circles by 0.05 μm each time. Intersections per circle were counted using ImageJ Sholl analysis 71 . Data from three sections per mouse, and four to six mice per group, were averaged to provide a mean value for each group in immunohistochemical and immunofluorescence analyses. All images were analyzed by a person blind to the experimental condition.
qPCR. Total RNAs from the mPFC were extracted with TRIzol according to the manufacturer's instructions. cDNAs were synthesized from 1 µg total RNA using the Maxima First Strand Synthesis Kit for RT-qPCR (Takara, #RR047B). qPCR was performed by amplifying cDNA for 40 cycles using the SYBR Green PCR master mix. Relative expression of mRNA for the target genes was calculated by the comparative ∆∆Ct method using GAPDH as control reference genes. All primers were synthesized by TsingKe, and the related information is listed in Supplementary  Table 1. Data from six (for RNA extraction) or four (for qPCR) mice per group in duplicate experiments were averaged to provide a mean value for each group.
Western blot. The brain samples were lysed in the RIPA buffer (Beyotime, #P0013B) containing protease inhibitors (Beyotime, #st506) and phosphatase inhibitors (Roche, #04906837001). Cell lysates were cleared of cellular debris by centrifugation (12,000 rpm for 15 min), and equal amounts of protein were loaded in the SDS-PAGE on 8-12% gels and then transferred to PVDF membranes. After blocking for 1 h in TBST containing 5% nonfat milk, the membranes were incubated at 4 °C overnight with one of the following primary antibodies: anti-Egr1 (1:1,000, Santa Cruz, #sc-189), anti-Egr2 (1:1,000, Santa Cruz, #sc-293195), anti-Arc (1:1,000, Santa Cruz, #sc-17839), anti-Lef1 (1:500, Proteintech, #14972-1-AP), anti-Nab2 (1:1,000, Proteintech, #19601-1-AP), anti-NTs (1:500, Immunoway, #YT5535), anti-Pcsk9 (1:500, Proteintech, #55206-1-AP), anti-Htr1d (1:250, Novus, #NB100-56349SS), anti-PSD-95 (1:1,000, Abcam, #ab18258), anti-MBP (1:1,000, Abcam, #ab7349) and anti-GAPDH (1:3,000, Proteintech, #60004-1-Ig). HRP-conjugated secondary antibodies are from Vector Laboratories; bands were visualized using ECL plus detection system (Imagequant LAS4000 mini). Quantification was performed using Gelpro32. Western blot uncropped data are presented in Supplementary Information. Experiments were performed in triplicate. Fig. 2 | The sociability, cooperation, locomotion, exploration, spatial working memory, anxiety-like behaviors, and fear memory defects of Si mice are rescued by resocialization. a, Diagram showing the timeline of the experiment, including weanling, social isolation, resocialization, behavioral tests (shown in b-i), and then qPCR (shown in Fig. 2j), Western blot (shown in Fig. 2k, l and extended Data Fig. 3g,h), eM (shown in extended Data Fig. 3a-d), immunohistochemistry (shown in extended Data Fig. 3e, f), immunofluorescence (shown in Fig. 6i, j and Supplementary Fig. 3a, b), or eLISa analysis (shown in Fig. 6e). b, c, The social behavior performance and social memory of mice in social test. d, e, Graphs showing the time before the first drinking (drinking latency), the times of drinking (drinking number), and the total time spent on drinking (drinking time) each day in the training (d) and testing (e) periods. f, Locomotion, and exploration behavior in the open field. g, The anxiety-like behavior examined by the elevated plus-maze test. h, The spatial working memory performance in the Y-maze testing. i, Fear memory performance in the fear conditioning test. (GH: n = 15 mice; GH-SI: n = 15 mice; SI: n = 14 mice); Data are represented as medians and IQR in c, f-i; and mean + SeM in d and e (n = 7 pairs of mice per group). Data in c and f-i were analyzed by One-way aNOVa followed by Tukey's post hoc test; and data in d and e were analyzed by Repeated Measures aNOVa with post hoc Student-Newman-Keuls test. GH: group housing; SI: social isolation; GH-SI: resocialization of SI mice with GH mice. Fig. 3 | The mPFC impairment of Si mice is rescued by resocialization. a, Representative eM images showing myelin (the top panel), oligodendrocyte nuclear heterochromatin (the middle panel), and synaptic morphology (the bottom panel) in the mPFC, respectively. Scale bar, 500 nm. b, The scatter plots of g ratios with linear least squares fitting (g ratio vs. axon caliber) in GH (106 axons), GH-SI (107axons), and SI (110 axons) mice (n = 4 per group). c, Graph showing the area percentage of nuclear heterochromatin in GH (17 nuclei), GH-SI (19 nuclei), and SI (15) mice (n = 4 per group). d, Postsynaptic density thickness, synaptic cleft width, presynaptic active zones length, and synaptic curvature in GH (70 synapses), GH-SI (68 synapses) and SI (63 synapses) mice (n = 4 per group). e-h Representative immunostaining images (e) and Western blot bands (g) showing MBP and PSD-95 expression in the mPFC of GH, GH-SI, and SI mice; corresponding quantification analyses are shown in f for e (n = 4 per group) and h for g (n = 8); scale bar, 50 µm. Data are represented as mean ± SeM in c and d; and as medians and IQR in f and h. Data in b-h were analyzed by One-way aNOVa followed by Tukey's post hoc test. GH: group housing (P87); SI: social isolation (P87); GH-SI: resocialization of SI mice with GH mice (P87). Fig. 4 | Knocking down Egr2 in the mPFC of GH mice does not affect exploring activities, spatial working memory, anxiety-like behavior, and fear memory. a, Graphs showing the time before the first drinking (drinking latency), the times of drinking (drinking number), and the total time spent on drinking (drinking time) each day in different groups at P46-57 in the 7 consecutive training days. b, Locomotion, and exploration behavior in different groups at P58 in the open field. c, The anxiety-like extent examined in different groups at P60 by the elevated plus-maze test. d, The spatial working memory performance of mice at P59 in the Y-maze testing. e, Fear memory performance of mice at P60-61 in fear conditioning test. Data are represented as mean + SeM in a; and as medians and IQR in b-e (n = 20 per group). Data in a were analyzed by repeated-measures aNOVa with post hoc Student-Newman-Keuls test; and in b-e by Two-way aNOVa followed by Tukey's post hoc test. GH-GFP: GH mice with aaV-GFP virus injection; GH-egr2-RNai: GH mice with aaV-egr2-RNai virus injection; SI-GFP: SI mice with aaV-GFP virus injection; SI-egr2-RNai: SI mice with aaV-egr2-RNai virus injection. Fig. 5 | Overexpression of Egr2 in the mPFC of Si mice mitigates abnormal anxiety-like behavior, but not exploring activities, spatial working memory and fear memory. a, Graphs showing the time before the first drinking (drinking latency), the times of drinking (drinking number), and the total time spent on drinking (drinking time) each day in different groups at P46-57 in the 7 consecutive training days. b, Locomotion and exploration behavior of P58 mice were tested in the open field. c, The anxiety-like behavior of P60 mice was examined in the elevated plus-maze test. d, The spatial working memory performance of P59 mice was tested in the Y-maze test. e, Fear memory performance in different groups at P60-61 in fear conditioning test. Data are represented as mean + SeM in a; and as medians and IQR in b-e (data from 18 mice in GH-GFP group, 22 mice in GH-egr2 group, 30 mice in SI-GFP group and 28 mice in SI-egr2 group, respectively). Data in a were analyzed by repeated-measures aNOVa with post hoc Student-Newman-Keuls test; and data in b-e were analyzed by Two-way aNOVa followed by Tukey's post hoc test. GH-GFP: GH mice with aaV-GFP virus injection; GH-egr2: GH mice with aaV-egr2 virus injection; SI-GFP: SI mice with aaV-GFP virus injection; SI-egr2: SI mice with aaV-egr2 virus injection.