Inhibition of Connexin 36 attenuates HMGB1‐mediated depressive‐like behaviors induced by chronic unpredictable mild stress

Abstract Background High mobility group box 1 (HMGB1) released by neurons and microglia was demonstrated to be an important mediator in depressive‐like behaviors induced by chronic unpredictable mild stress (CUMS), which could lead to the imbalance of two different metabolic approaches in kynurenine pathway (KP), thus enhancing glutamate transmission and exacerbating depressive‐like behaviors. Evidence showed that HMGB1 signaling might be regulated by Connexin (Cx) 36 in inflammatory diseases of central nervous system (CNS). Our study aimed to further explore the role of Cx36 in depressive‐like behaviors and its relationship with HMGB1. Methods After 4‐week chronic stress, behavioral tests were conducted to evaluate depressive‐like behaviors, including sucrose preference test (SPT), tail suspension test (TST), forced swimming test (FST), and open field test (OFT). Western blot analysis and immunofluorescence staining were used to observe the expression and location of Cx36. Enzyme‐linked immunosorbent assay (ELISA) was adopted to detect the concentrations of inflammatory cytokines. And the excitability and inward currents of hippocampal neurons were recorded by whole‐cell patch clamping. Results The expression of Cx36 was significantly increased in hippocampal neurons of mice exposed to CUMS, while treatment with glycyrrhizinic acid (GZA) or quinine could both down‐regulate Cx36 and alleviate depressive‐like behaviors. The proinflammatory cytokines like HMGB1, tumor necrosis factor alpha (TNF‐α), and interleukin‐1β (IL‐1β) were all elevated by CUMS, and application of GZA and quinine could decrease them. In addition, the enhanced excitability and inward currents of hippocampal neurons induced by lipopolysaccharide (LPS) could be reduced by either GZA or quinine. Conclusions Inhibition of Cx36 in hippocampal neurons might attenuates HMGB1‐mediated depressive‐like behaviors induced by CUMS through down‐regulation of the proinflammatory cytokines and reduction of the excitability and intracellular ion overload.

increased levels of inflammatory mediators like interleukin (IL)−1β, IL-6, tumor necrosis factor alpha (TNF-α), the acute phase reactant (CRP), and high mobility group box 1 (HMGB1) were seen in both patients and rodents with depressive disorder (Cernackova et al., 2020;Liu et al., 2019;Xie et al., 2021;Zolfaghari et al., 2021). Our previous work focused mainly on the role of HMGB1, a late mediator of inflammation in depressive-like behaviors (Wang et al., 1999). We first discovered that HMGB1 was significantly increased in mice under acute or chronic stress, while application of glycyrrhizinic acid (GZA), an inhibitor of HMGB1, could markedly improve depressive symptoms (Lian et al., 2017;Wu et al., 2015). These findings were later demonstrated by other teams (Fu et al., 2019;Yang et al., 2020). The increased HMGB1 might subsequently induce the imbalance of two different metabolic approaches in kynurenine pathway (KP) in microglia and astrocytes, thus resulting in the abnormal glutamatergic neurotransmission and contributing to the development of depressive-like behaviors Wang, Lian et al., 2019). It was interesting that our results also showed the increased HMGB1 mainly originated from neurons and microglia in the central nervous system (CNS) (Wang, Huang et al., 2020), which then brought out important questions about how HMGB1 communicated among cells and the possible mechanisms.
There was evidence that the release of HMGB1 could be regulated by connexins (Cxs). Cxs are fundamental structures of intracellular communications, which can form two functional forms, namely gap junctions (GJ) and hemichannels (HC) (Takeuchi & Suzumura, 2014). In a mouse model of acute lung injury (ALI), application of a broad-spectrum Cxs blocker named P5 significantly reduced the endotoxin-induced release and accumulation of HMGB1 . Recent evidence also strongly indicated that Cxs dysfunction might be involved in the pathogenesis of depressive disorder (Ren et al., 2018). In a study of depressive-like behaviors in mice induced by inescapable stress, knock-down of Cxs could result in significant anti-depressant effects as well as the improvement of cognitive functions (Quesseveur et al., 2015). Similarly, injection of a nonselective Cxs inhibitor named carbenoxolone (CBX) into the hippocampus of mice exhibited the same therapeutic effects in a depression model of CMS (Chen, Wang, Zuo et al., 2016). Therefore, it is probable that abnormal expressions or functions of Cxs are involved in the development of depressive disorder. On these foundations, we conjectured that Cxs might be involved in HMGB1 signaling during the process of depressive-like behaviors. Since there is coexpression of Cx36 in both neurons and microglia, (Medina-Ceja et al., 2019;Orellana et al., 2009) and based on our previous work, we mainly focus on the role of Cx36 in HMGB1 mediated depressive-like behaviors.
Actually, Cx36 is the main connexin between neurons, which not only plays an important role in the development of neurons, but also participates in glutamate-induced neuronal excitotoxicity and injury (Arundine & Tymianski, 2004;Belousov et al., 2017). Since neuronal injury, glutamate excitotoxicity, and enhanced HCs activity were closely related to the development of depressive disorder, we further speculated that Cx36 might be the mediator in the process of HMGB1 mediated depressive-like behaviors based on our previous work. However, the underlying mechanisms were not fully elucidated. In this study, we sought to explore the change of Cx36 under chronic unpredictable mild stress (CUMS) and its role in HMGB1 signaling.

Animals and treatments
All the animals used in this study were 8-week-old male BALB/c mice purchased from the experimental animals center (Second Military Medical University, Shanghai, China). The mice were raised in standard conditions of constant temperature (21-23 • C) and humidity (50%−52%) on a light/dark cycle of 12 h with food and water available ad libitum. All experiments were conducted in accordance with the standards established by the experimental animal ethics committee of Second military Medical University. Before the experiments, mice could adapt to new feeding conditions for 2 weeks.
In the experiment of GZA (G2137, Sigma-Aldrich, USA) treatment, mice were randomly assigned to four groups, namely control group, CUMS group, CUMS+GZA group, and GZA group. In the experiment of quinine (V900715, Sigma-Aldrich, USA) treatment, mice were randomly assigned to four groups, namely control group, CUMS group, CUMS+quinine group, and quinine group. GZA and quinine were dissolved in 0.9% sterile saline and injected intraperitoneally every day from the day before CUMS. According to previous studies (Nassiri-Asl et al., 2009;, GZA was applied at a dose of 20 mg/kg and quinine at 50 mg/kg. The final volume of all drugs injected into each mouse was 200 μl. Mice of control and CUMS group were intraperitoneally administered of equal volume of saline every day.

Chronic unpredictable mild stress
According to our previous studies (Leng et al., 2018;Lian et al., 2017), the following mild stimuli were applied randomly to mice: (i) 5 min of exposure to high temperature (45 • C); (ii) 10 min of cage shaking; (iii)

Behavioral tests
After four-week stimuli, behavioral tests were performed in mice to evaluate depressive-like behaviors. All the tests were conducted between 18:00 and 21:00.

Sucrose preference test
By reference to previous studies (Lian et al., 2017;, sucrose preference test (SPT) was conducted to assess the affection of mice for sweets, which reflected anhedonia, one of the key symptoms in depressed patients. Mice first learned to choose 1% sucrose solution or water from two bottles for 2 weeks. During this period, bottles were switched every 3 days to avoid preference for either side. Sucrose preference was measured before and after 4 weeks of CUMS. Mice were deprived of food and water for 12 h, and then allowed free access to 1% sucrose and water in two separate bottles. An hour later, the bottles were weighed and sucrose preference was calculated. Sucrose preferences (%) = sucrose consumption/(sucrose+water consumption) × 100%.

Open field test
Open field test (OFT) was conducted to assess the spontaneous activities of mice , which often reflected the anxious performance of mice (Chen, Wang, Liang et al., 2016

Tissue preparation
After behavioral tests, mice were all sacrificed and tissues were quickly obtained.

For western blot analysis
The hippocampus and prefrontal cotex were weighed and homoge-

Immunofluorescence (IF)
The samples of hippocampus were dried and embedded with optimum cutting temperature (OCT) compounds and frozen at −20 • C. The samples were cut into slices of 10 μm using a Leica CM-1900 slicer.
Then sodium citrate solution (PH 6.0) was added for antigen repairing and slices were boiled in the water bath at 95 • C for 10 min. After

Enzyme-linked immunosorbent assay (ELISA)
The levels of HMGB1, TNF-α, and IL-1β in the serum and hippocampus were measured by ELISA kits bought from Shanghai Westang Biotech and IL-1β were measured using the double antibody sandwich method.
After the immunoreactions, the absorbance value at 450 nm was measured within 30 min and the concentration of cytokines was calculated according to the optical density (OD) values. The sensitivity of these kits was 2 ng/ml for HMGB1, 4 pg/ml for TNF-α, and 8 pg/ml for IL-1β.

Primary neuronal cultures
The isolation and culture of hippocampal neurons from postnatal (P0-P1) mice were referred to previous studies (Beaudoin et al., 2012;Gardner et al., 2012;Kaech & Banker, 2006). Briefly, mice were first disinfected with 75% alcohol thoroughly and euthanized by decapitation.
The brains were quickly isolated and placed in a 60 mm dish containing dissection medium. The hippocampus was gently separated under a dissecting microscope, and the meninges were removed as completely as possible. The hippocampus was transferred into a new dish with 1 ml of fresh dissection medium and cut into pieces with fine scissors.
Then two drops of DNase solution (DN25, Sigma-Aldrich, USA) and 1 ml of 0.25% trypsin solution (#25300-054, Gibco, USA) were added into the dish. The tissues were incubated at 37 • C in a cell culture incubator for 10 min and gently shaked every 5 min. After that, 2 ml of plating medium (Neurobasal-A containing 2% B27; #21103-049 and #12587-010, Gibco, USA) was immediately added to inactivate DNase and trypsin. After centrifuged at 800 rpm for 5 min, the sediment was resuspended in 3 ml of plating medium and gently dissociated to obtain a homogenous cell suspension. The viable cell density was estimated on a hemocytometer and the suspension was plated on 60-mm diameter dishes coated with poly-L-lysine (P9155, Sigma-Aldrich, USA). Then the neurons were incubated in the cell culture incubator at 37 • C for 4-6 h.

Whole-cell patch-clamp recording
The electrophysiological signalings were recorded by Clamp 700B amplifier (Axon, USA) and transformed by Digidatal 1440 (Axon, USA) at room temperature. The procedures were referred to previous reports (Due et al., 2012;Gomez et al., 2019

Statistical analysis
All data were expressed as mean ± standard error of the mean (SEM).
One-way or two-way analysis of variance was performed for comparison among multiple groups followed by a least significant difference (LSD-t) post hoc test. Nonparametric test (Kruskal-Wallis test, K-W test) was adopted under nonnormal distribution or heterogeneity of variance among groups. The difference was considered statistically significant when p < .05. All statistical analyses were performed with SPSS 22 (SPSS, USA).

3.1.1
Cx36 was increased by CUMS, while decreased by GZA Evidence indicated that Cx36 took part in several pathophysiological activities, like injury and inflammation , which was closely related to depressive disorder. Therefore, we first sought to see how Cx36 changed in depressive model induced by CUMS. After 4-week chronic stress, behavioral tests including SPT, TST, FST were performed to confirm whether depressive model was successfully constructed. As shown in Figure S1a-c, mice in CUMS group displayed significantly decreased sucrose preference and prolonged immobility time, which were important features of depressive disorder. After treatment with GZA, the above depressive symptoms were obviously relieved (Figure S1a, F = 5.606, p = .003, analysis of variance; Figure   S1b, F = 2.740, p = .059, analysis of variance; Figure S1c, H = 17.574, p = .001, K-W test). GZA was a Chinese herbal extract known as the direct inhibitor of HMGB1. Our results showed that GZA could not only inhibit HMGB1 in both serum and hippocampus, but also reduce downstream cytokines like TNF-α and IL-1β ( Figure S2a, F = 5.182, p = .006, analysis of variance; Figure S2b, F = 6.568, p = .002, analysis of variance; Figure S2c, H = 18.218, p < .001, K-W test; Figure S2d

3.1.2
Cx36 in hippocampal neurons was increased by CUMS, while decreased by GZA Since the expression of Cx36 in hippocampus was greatly changed, we sought to further nflamm its cellular localization using triplelabeling immunofluorescence. Neurons, astrocytes and microglias were labelled in green by NeuN, GFAP, and Iba-1, respectively. Nuclei were labelled in blue by dapi, and Cx36 in red by its specific antibody.
As shown in Figure 2a, Cx36 was mainly expressed in neurons and microglia, while hardly expressed in astrocytes. This was consistent with other studies (Dobrenis et al., 2005;Medina-Ceja et al., 2019). In order to see how Cx36 changed in hippocampal neurons after different treatments, immunofluorescence continued to be adopted as men-

3.2.1
Qunine alleviated depressive-like behaviors induced by CUMS On the basis of the above findings, the depressive-like behaviors of mice induced by chronic stress could be obviously alleviated by F I G U R E 1 GZA decreased the expression of Cx36 in hippocampus induced by CUMS. (a) Expression of Cx36 in hippocampus was significantly increased by CUMS, while decreased after treatment with GZA (*p < .05, **p < .01, n = 6-8 for each group). (b) There was no significant difference of Cx36 expression among groups (p > .05, n = 6-8 for each group). GZA, glycyrrhizinic acid; Cx36: Connexin36; CUMS: chronic unpredictable mild stress inhibition of HMGB1 by GZA, which might be through down-regulation of Cx36. Therefore, we went on to see whether direct inhibition of

3.2.2
Qunine decreased pro-inflammatory cytokines in the serum and hippocampus induced by CUMS According to our findings, inhibition of Cx36 by quinine could play a therapeutic role in depressive-like behaviors induced by CUMS. In addition, studies have indicated that cytokines could be regulated by Cx36 . Therefore, to further investigate the possible underlying mechanisms, ELISA was adopted to detect three classical proinflammatory cytokines in the serum and hippocampus of mice, which might be associated with our study, namely HMGB1, TNF-α, and IL-1β. As shown in Figure 4, either in the serum or hippocampus, the levels of HMGB1, TNF-α and IL-1β were all significantly increased in CUMS group, while treatment of quinine markedly decreased them.

Inhibition of HMGB1 and Cx36 altered the electrical activities of hippocampal neurons induced by LPS
Since depressive-like behaviors were ultimately related to the elec- Then, the firing frequency of action potentials was recorded (Figure 6a). When the frequency was plotted against injected current steps, there were significant differences among groups. LPS could markedly increase the firing frequency within the current range of 150-300 pA, while treatment with GZA or qunine could significantly F I G U R E 2 Changes of Cx36 in hippocampal neurons of mice exposed to CUMS and GZA. (a) Cx36 (red) was colabeled with neurons (NeuN, green), astrocytes (GFAP, green), microglia (Iba-1, green), and nuclei (DAPI, blue) in hippocampal slices of control mice. Cx36 was expressed in hippocampal neurons and microglia, but not in astrocytes (Scale bar: 50 μm). (b) Cx36 (red) was colabeled with neurons (NeuN, green) and nuclei (DAPI, blue) in the hippocampus of different groups (Scale bar: 50 μm). The quantity (c) and proportion (d) of Cx36 positive neurons in the hippocampus were both significantly increased by CUMS while decreased after treatment with GZA (*p < .05, ***p < .001, n = 6 for each group). GZA Glycyrrhizinic acid; Cx36, Connexin36; CUMS, chronic unpredictable mild stress; NeuN, neuronal nuclei antigen; GFAP, Glial Fibrillary Acidic Protein; Iba-1, ionized calcium binding adapter molecule 1; DAPI, 4,6-diamidino-2-phenylindole decrease it. Application of GZA or quinine alone made no difference ( Figure 6b, F = 103.861, p < .001, analysis of variance). Similarly, the threshold and rheonobase current of the action potentials were all decreased when mice were exposed to LPS treatment, while application of GZA or quinine could significantly increase them. There was no significant change when GZA or quinine was applied alone (Figure 6c,

DISCUSSION
Our findings indicated that the increase of Cx36 might be involved in HMGB1-mediated depressive-like behavious induced by CUMS.
Inhibition of HMGB1 might alleviate depressive symptoms through F I G U R E 3 Quinine alleviated depressive-like behaviors induced by CUMS. CUMS significantly decreased the sucrose preference (a) and extended the immobility duration of mice in TST (b) and FST (c), while treatment with quinine could markedly alleviate these symptoms (a-c) (*p < .05, **p < .01, ***p < .001, n = 8-10 for each group). However, there was no significant difference in the total distance (d) and central distance (e) among the treated groups in OFT (p > .05, n = 8-10 for each group

F I G U R E 4
Quinine decreased the release of pro-inflammatory cytokines in serum and hippocampus induced by CUMS. The hippocampal levels of HMGB1 (a), TNF-α (b), and IL-1β (c) were significantly increased by CUMS, but decreased significantly after treatment with quinine (a-c) (*p < .05, n = 8-10 for each group). Similarly, the serum levels of HMGB1 (d), TNF-α (e), and IL-1β (f) were also increased significantly by CUMS and decreased after quinine administration (*p < .05, **p < .0l, ***p < .001, n = 8-10 for each group). CUMS, chronic unpredictable mild stress; HMGB1, high mobility group box 1; TNF-α, tumor necrosis factor alpha; IL-1β, interleukin-1β down-regulation of Cx36 and subsequent inflammatory cytokines The drug itself could not bind to Cxs to exert biological effects (Mitrou et al., 2016;Ozog et al., 2002a;Patrone et al., 2014). However, according to our previous work, GZA could alleviate depressive-like behaviors of mice through inhibition of HMGB1 . Evidence also indicated that GZA might directly bind to HMGB1 or form DNA complex so that the nuclear translocation or release of HMGB1 was obstructed (Bailly & Vergoten, 2020). In this way, HMGB1 could not interact with its receptors like advanced glycation end products receptors (RAGE), toll-like receptors (TLR) 2 and TLR4, which also prevented the follow-up increase of other inflammatory cytokines like TNF-α and IL-1β (Lian et al., 2017). Hence, the physiological activity of HMGB1 was decreased and GZA exerted anti-inflammatory effects (Bailly & Vergoten, 2020). Therefore, we conjectured that increase of Cx36 in CUMS might be mediated by HMGB1. This was also seen in other studies that expression of Cx36 was significantly elevated by injection of capsaicin, CFA or IL-6 in different models of injury, while inhibition of Cx36 could alleviate the inflammatory symptoms and motor deficits (Garrett & Durham, 2008;. However, in a study of cytokine-induced oxidative stress, pro-inflammatory cytokines like IL-1β, TNF-α, and IFN-γ could markedly down-regulate the expression of Cx36 through activation of AMPK (Allagnat et al., 2013). These obviously contradictory results might be probably due to the different functional constructions of Cx36, namely GJs and HCs (Quesseveur et al., 2015). Studies indicated that the roles of HCs and GJs in depressive disorder might be totally opposite. Promotion of GJs communication between neuroglial cells might increase neurotrophic supplies and enhance brain plasticity, thus playing an antidepressant role (Jeanson et al., 2016). However, enhanced activities of HCs might result in the increase of glutamate-induced excitotoxicity and neural injuries (David et al., 2009;Hanstein et al., 2009;Orellana et al., 2015;Rainer et al., 2012). Although Cx36 inhibitors or geneediting of Cx36 were found to alleviate or exacerbate sick behaviors in different animal models, their effects on Cx36 protein itself or the functions of HCs and GJs were not yet clear. In our study, the functions of Cx36 HCs and GJs were not covered as well, so we were going to arrange more researches to further explore the underlying mechanisms.
Both clinical and experimental evidence showed that the emotional circuits associated with depressive disorder consisted of different parts of the brian, such as PFC, ventral hippocampus, amygdala and nucleus accumbens (Nac) . In our study, we found the increase of Cx36 induced by chronic stress mainly appeared in the hippocampus, but not PFC. One possibility was that the expression of Cx36 might be time-and region-dependent. In a study of an inflammatory CNS disease, the expression of Cx36 began to change in the hippocampus four weeks after inflammation was induced, while it did not change in the cortex until 8 weeks later (McCracken et al., 2005). Therefore, the findings that Cx36 in PFC did not change obvi- Since the increase of Cx36 might aggravate depressive-like behaviors and GZA could not block Cx36 directly, we then sought to choose a relatively specific inhibitor, namely quinine. Quinine was reported to be the inhibitor of both Cx36 and Cx50. However, Cx50 was hardly expressed in mammals and only seen in lens (Srinivas et al., 2001).
There might be two possible explanations for these paradoxical observations. On one hand, the role of quinine might depend on the mutual interaction between chemical and electric synapses and to what extent they were involved in different models of CNS diseases (Ghanbarabadi & Sayyah, 2013). On the other hand, it might be due to the off-target effects of quinine at a higher dose (e.g., 400 μM), which was considered as a less likely explanation but could not be completely ruled out (Voss et al., 2009). Additionally, neither chronic stress nor quinine treatment produced an impact on the locomotor activity and anxious status of mice in OFT (Figure 3), indicating that chronic stress or quinine treatment made no difference to the excitability of central nervous system, which was consistent with other studies (Frisch et al., 2005;Su et al., 2018). In a recent study, neither 5 weeks nor 9 weeks of CUMS changed the locomotor activity of mice in OFT . As to quinine, similar results showed that knockout of Cx36 did not affect the locomotor activity of mice in OFT (Frisch et al., 2005;Steffensen et al., 2011). However, quinine might be able to improve the impaired locomotor activity in neuropsychiatric diseases (Nassiri-Asl et al., 2009).
Inhibition of Cx36 by quinine could subsequently reduce the release of pro-inflammatory cytokines like HMGB1, IL-1β, and TNF-α in the serum and hippocampus of mice exposed to CUMS, which was consistent with other studies. On the circumstances of injury, infection or stress, the innate immune system could be activated and induce inflammation afterwards to exert protective physiological effects (Vezzani et al., 2016). In most CNS diseases, a common underlying mechanism might be the inflammatory cascade reaction triggered by the release of inflammatory cytokines (Vezzani et al., 2016). Clinical evidence showed that inflammatory mediators were elevated in patients with depressive disorder, such as HMGB1, IL-1β, TNF-α, CRP, IL-6, and so on (Leighton et al., 2018). In addition, animal studies also showed that in different models of depressive disorder, the inflammatory cytokines in the serum and CNS were increased significantly (Leighton et al., 2018;Miller et al., 2009). Furthermore, both clinical and experimental evidence indicated that administration of certain proinflammatory cytokines like IL-1β, TNF-α, or IFN-α could induce depressive symptoms or depressive-like behaviors, such as cognitive impairment, psychomotor retardation, reduced social, or exploratory behaviors (Dantzer et al., 2008;Eggermont et al., 2008;Friebe et al., 2010;Kim et al., 2016;Reichenberg et al., 2005). On the contrary, treated with anti-inflammatory drugs or conditional knockout of inflammatory factors like HMGB1 could improve these symptoms (Coleman et al., 2018;Kappelmann et al., 2018;Musumeci et al., 2014;Tyring et al., 2006;Ye et al., 2019). There was also evidence that stress could enhance the activity of Cx36 HCs in neurons by direct or indirect overactivation of microglias, thus promoting the release of cytokines like TNF-α and IL-1β (Orellana et al., 2009;Sánchez et al., 2020). Treatment with quinine or GZA could in turn attenuate inflammatory symptoms through inhibition of certain proinflammatory cytokines, such as TNF-α, IL-1β, GZA might inhibit HMGB1 at first and subsequently reduce the levels of other inflammatory cytokines like TNF-α and IL-1β (Lian et al., 2017).
However, the mechanisms underlying down-regulation of inflammatory cytokines by quinine were complicated. In spite that quinine was a relatively specific inhibitor of Cx36, it could also bind to other proteins or ion channels to exert off-target effects (Grassin-Delyle et al., 2019; Maruyama et al., 1994). Therefore, more studies were needed to further invesitigate the mechanism.
In our study, an interesting finding drew our attention that HMGB1  . The common mechanism might be the altered permission of Cx HCs in these processes mediated by the direct interaction with HMGB1 or its vectors. On the other hand, HMGB1 could also promote the release of other cytokines like TNF-α and IL-1β through regulation of Cx HCs in a study of osteocytes and osteoclasts (Davis et al., 2019). These results indicated that the possible mutual promotion of HMGB1 and Cx36 with each other in depressive disorder might promote the process, which were in need of further experiments.
Our results also showed that exposure to LPS could increase the excitability and influx of ions in hippocampal neurons, which was consistent with many studies. LPS was the important constituent in the outer membrane of Gram-negative bacteria. It has been widely used in animal studies in vivo and vitro to explore the mechanisms underlying inflammatory CNS diseases (Laflamme et al., 2003;Rivest, 2003). Studies in vivo showed that intraperitoneal or intracerebroventricular injection of LPS could induce acute depressive-like behaviors in rodents, which might be through the increase of proinflammatory mediators like HMGB1, IL-1β, cyclooxygenase 2 (COX-2), nitric oxide (NO), and prostaglandin (PG) (Leighton et al., 2018  In our study, we also observed that treatment with GZA or quinine could decrease the amplitude and frequency of APs, increase the threshold and rheonobase currents, and decrease inward currents, while application of GZA or quinine alone did not made any differences to the electrical characteristics of hippocample neurons. As mentioned above, GZA was an inactive analog of CBX that could not directly inhibit Cxs. Evidence showed that GZA itself did not affect the action potential, voltage-gated Na + current (I Na ), or voltage-gated K + current (I K ) evoked by depolarizing voltage steps (Elsen et al., 2008;Kimura et al., 1985). However, in the presence of other stimulators like glutamate or sustained depolarization (SD), GZA could diminish the inward Ca 2+ current (Cherng et al., 2006;Kimura et al., 1985). As to quinine, evidence also showed that in normal artificial cerebrospinal fluid (ACSF), it did not produce influences on the resting membrane potential, input resistance (R in ), AP threshold or amplitude of evoked population spikes (PS) in hippocampal CA1 pyramidal neurons (Bikson et al., 2002). The effects of quinine on the elecrical characteristics of neurons lied in the increase of the duration and decrease of the firing frequency of APs in the process of SD, but it produced no effects on normal resting membrane potentials (Bikson et al., 2002).
Furthermore, quinine could inhibit extracellular transient potassium current and reduce the firing frequency in a voltage-dependent manner (Cummings et al., 2008). There was also evidence that under normal membrane potential, quinine had no effect on the evoked current, but it could significantly inhibit the calcium influx during depolarization (Seemann et al., 2018). In spite of these observations, the in depth researches were still in need to further expound the underlying mechanisms.

CONCLUSIONS
In summary, inhibition of Cx36 in hippocampal neurons might attenuate HMGB1 mediated depressive-like behaviors induced by CUMS through down-regulation of proinflammatory cytokines and reduction of the excitability and intracellular ion overload, thus playing an antidepressant role (Figure 8). Cx36 and HMGB1 might mutually promote each other in the development of depressive-like behaviors so that inhibition of them might be of benefit to depressive disorder.

ACKNOWLEDGMENTS
This study was supported by Natural Science Foundation of China (NSFC, Nos. 81171124 and 81771301).

CONFLICT OF INTEREST
The authors declare no conflict of interest.

ETHICAL APPROVAL
This study was approved by the experimental animal ethics committee of Second military Medical University.

DATA AVAILABILITY STATEMENT
All data included in this study are available upon request by contact with the corresponding author.