All procedures were approved by Chungnam National University Hospital, Daejeon, South Korea (CNUH-017-P0016). BTBR mice (Jackson laboratory, Bar Harbor, ME, United States) and C57BL/6J mice (RRID: MGI:5659186, Damul Science, Daejeon, South Korea) were housed in a controlled environment (temperature, 22°C; humidity, 40%) on a 12-h reverse light/dark cycle and allowed free access to food and water. Mice were weaned at PND21 and housed in groups of 3–5 per cage. All studies were performed in male mice, as the prevalence of autism is much higher in males.

BTBR mice were exposed to sevoflurane at PND16/17 as previously described (Ju et al., 2020a). Mice were randomly divided into four groups: a naïve C57BL/6J (B6-control) group, a B6-sevoflurane group, a BTBR-control group, and a BTBR-sevoflurane group. Mice in the B6/BTBR-sevoflurane group exposed to three cycles of 2.5% sevoflurane (Ilsung, Seoul, South Korea), delivered at a rate of 5 L/min at 36°C [inspired oxygen fraction (FiO₂), 0.4]. Sevoflurane was administered for 1 h at each cycle, and there was a 2-h interval between cycles. BTBR-control and B6-control mice were exposed to the same constant fresh of gas but without sevoflurane. Sevoflurane and carbon dioxide (CO₂) levels were monitored continuously (Dräger Vamos 2; Draegerwerk AG & Co., Germany).

Whole-cell patch-clamp recordings of layer II and III pyramidal neurons in the medial prefrontal cortex (mPFC) were performed 5 days after sevoflurane exposures, as previous described (Han et al., 2014; Chung et al., 2017a). Coronal sections of the mPFC (300 μm) were cut in ice-cold, continuously aerated (95% O₂/5% CO₂) dissection buffer (sucrose-based) using a VT1200S vibratome (Leica, Switzerland). For recovery, the slices were transferred to chamber filled with aCSF warmed to 32°C for 30 min. Glass capillaries were filled with an internal solution whose composition depended on the experiment. To avoid obtaining too much data from a single slice or animal, the number of whole-cell recordings was limited to ≦2 cells per slice (containing one side of mPFC) and ≦5 cells per animal. For spontaneous inhibitory postsynaptic currents (sIPSCs) recordings, the solution consisted of 115 mM CsCl, 10 mM tetraethylammonium chloride, 8 mM NaCl, 10 mM HEPES, 5 mM QX-314-Cl, 4 mM Mg-ATP, 0.3 mM Na-guanosine triphosphate, and 10 mM EGTA. For spontaneous excitatory postsynaptic current (sEPSC) recordings, the solution consisted of 117 mM CsMeSO4, 10 mM tetraethylammonium chloride, 8 mM NaCl, 10 mM HEPES, 5 mM QX-314-Cl, 4 mM Mg-ATP, 0.3 mM Na-guanosine triphosphate, and 10 mM EGTA. Whole-cell voltage recordings were acquired at a holding potential of −70 mV in the presence of 50 μM picrotoxin (for sEPSCs) or 10 μM NBQX and 50 μM D-AP5 (for sIPSCs). Neurons were visually identified (BX50WI; Olympus, Japan), and recordings were performed using a MultiClamp 700B amplifier (Molecular Devices, United States). Data were acquired with Clampex 11.0.3 (Molecular Devices) and analyzed using Clampfit 11.0.3 (Molecular Devices).

Mitochondria were isolated from the cerebral cortex 5 days after sevoflurane exposures, as previously described (Lee et al., 2021). Mitochondrial fraction was acquired from the cerebral cortex, and seeded in an XF-24 plate (Seahorse Bioscience, United States). Plates were centrifuged (2,000 × g, 20 min, 4°C) in a swinging bucket microplate adaptor (Eppendorf, Germany), after which 450 μl of mitochondrial assay buffer containing substrate (10 mM succinate plus 2 μM rotenone) was added to each well. The XF-24 plate was then placed in a CO₂-free incubator and maintained at 37°C for 8–10 min, then loaded into the Seahorse XF-24 extracellular flux analyzer for OCR measurements.

Western blot was performed as previously described (Ju et al., 2020a). Protein samples of the cerebral cortex were obtained 5 days after sevoflurane exposures. The indicated primary antibodies against the following proteins were used: BDNF (Abcam, United Kingdom, CAT#ab226843), tropomyosin receptor kinase B (TrkB; Abcam, CAT#ab18987), phosphorylated TrkB (Sigma-Aldrich, Germany, CAT# ABN1381), OXPHOS (Abcam, CAT#ab110413), and β-actin (Santa Cruz Biotechnology, United States, CAT# sc-8432). Specific antibody-labeled proteins were detected using an enhanced chemiluminescence system (WEST-ZOL plus; intROn BioTechnology, South Korea).

All behavioral tests were performed blindly in adult mice (8–10 weeks) and video recorded. Time spent in each chamber in the three-chamber test was automatically measured using a tracking software (Ethovision XT; Noldus Information Technology, Netherlands). Grooming behavior was manually analyzed by a blinded observer.

A three-chamber apparatus (60 cm width × 40 cm height × 22 cm depth) containing a cage in each side chamber was used to measure sociability. The experiment was composed of three 10-min sessions. In the first session, mice were restricted to the center chamber; in the second session, mice were allowed free access to explore all three chambers. After these two habituation sections, a novel object, consisting of a blue cube measuring 3.0 × 3.0 × 4.0 cm or a novel stranger mouse (age-matched male mouse of the same strain) was placed inside each side chamber. The subject mouse was allowed to freely explore the apparatus for 10 min, and the time spent sniffing the cages in each side chamber (containing a stranger mouse or novel object) was measured. The positions of the object and stranger were alternated each time to prevent side preference. Preference index (PI) was calculated using the time spent in each chamber (Mo, time spent in the mouse chamber; Ob, time spent in the object chamber): PI (%) = (Mo – Ob)/(Mo + Ob) × 100.

A home cage without any bedding was used for measuring grooming behavior (Chung et al., 2015). Mice were placed in the center of the cage for 20 min. Grooming was measured manually for 10 min after a 10-min adaptive phase. Cages were changed for each experiment.

Sample size was based on previous studies (Chung et al., 2017a; Ju et al., 2020a). Data were analyzed using the R statistical software package (4.0.3; R Core Team, Austria). All continuous variables were tested for conditions of normality and homogeneity of variance. When both conditions were met, a one-way analysis of variance (ANOVA) with post hoc Tukey’s HSD test was performed; when homogeneity of variance was unmet, Welch’s ANOVA with post hoc Tukey’s HSD test was performed; and when normality was unmet, the Kruskal–Wallis test with post hoc Dunn’s test was performed. Statistics are presented as Supplementary Data.

E/I imbalance during neurodevelopment is a major mechanism underlying autism and a reported feature in BTBR mice (Han et al., 2014). As we previously reported that sevoflurane exposures induce long-lasting E/I imbalance (Ju et al., 2020a), changes in E/I balance were evaluated by measuring spontaneous synaptic transmission in pyramidal neurons of the mPFC (layer II, III), a brain region involved in autism behavior (Chung et al., 2015; Kim et al., 2016). Excitatory synaptic transmission was increased in BTBR mice compared with B6 mice, as evidenced by the significantly increased frequency (but not amplitude) of sEPSCs (Figures 1A,B; Welch’s ANOVA with post hoc Tukey’s HSD test, p < 0.001). Such differences in excitatory synaptic transmission were not affected by early sevoflurane exposures (Figures 1A,B). We also discovered that inhibitory synaptic transmission was decreased in BTBR mice, as shown by a significant decrease in sIPSC amplitude and frequency compared with B6 mice (Figures 1C,D) (Amplitude, Kruskal–Wallis test with post hoc Dunn’s test, p = 0.004; Frequency, ANOVA with post hoc Tukey’s HSD test, p < 0.001). Interestingly, early sevoflurane exposures significantly increased both amplitude and frequency of sIPSCs in BTBR mice (Figures 1C,D; Amplitude, Kruskal–Wallis test with post hoc Dunn’s test, p = 0.005; Frequency, ANOVA with post hoc Tukey’s HSD test, p = 0.043). Collectively, these results suggest that early sevoflurane exposures improve pre-existing E/I imbalances in BTBR mice.

Previous studies suggest multiple mechanisms for the autistic behavior in BTBR mice, including suboptimal mitochondrial function (Ahn et al., 2020). Since sevoflurane can also affect mitochondrial function (Chung et al., 2017a; Ju et al., 2020b), we next evaluated the effects of early sevoflurane exposures on mitochondrial respiration (OCR). Consistent with previous studies, mitochondrial OCR was reduced in BTBR-control mice compared with that in B6-control mice (Figure 2A; one-way ANOVA with post hoc Tukey’s HSD test, p < 0.001). Interestingly, early sevoflurane exposures significantly increased mitochondrial respiration in BTBR mice (BTBR-control vs BTBR-Sevoflurane) (Figure 2A; one-way ANOVA with post hoc Tukey’s HSD test; Stage I, p = 0.005; II, p < 0.001; III, p < 0.001; IV, p < 0.001). We also measured expression levels of mitochondrial oxidative phosphorylation (OXPHOS) complex subunits. The expression of NDUFB8 [NADH dehydrogenase (ubiquinone) 1 beta subcomplex subunit 8, complex I] and ATP5A (mitochondrial α-F₁-ATP synthase subunit, complex V) were significantly decreased in BTBR mice compared with B6 mice (Figure 2B; NDUFB8, Kruskal–Wallis test, p = 0.005; ATP5A, Kruskal–Wallis test, p = 0.03). However, only the expression of SDHB (succinate dehydrogenase (ubiquinone) iron-sulfur subunit, complex II) was increased in BTBR mice after early multiple exposures to sevoflurane (Figure 2B; One-way ANOVA with post hoc Tukey’s HSD test, p = 0.036). Unlike BTBR mice, sevoflurane exposures did not affect expression of mitochondrial oxidative phosphorylation (OXPHOS) complex subunits in B6 mice (Figure 2C).

While our results suggest that early sevoflurane exposures correct both E/I imbalance and mitochondrial dysfunction, it raises the question of a common upstream signaling pathway. In this context, the BDNF/TrkB signaling pathway may be involved, since it not only regulates both mitochondrial function and synaptic transmission (Markham et al., 2004, 2012; Gottmann et al., 2009), but is also reduced in BTBR mice (Scattoni et al., 2013; Cristiano et al., 2018; Rhine et al., 2019). Consistent with previous studies, BDNF expression was significantly decreased in the cerebral cortex of BTBR-control mice compared with B6-control mice (Figures 3A,B; One-way ANOVA with post hoc Tukey’s HSD test, p = 0.01). Notably, early sevoflurane exposures in BTBR mice significantly increased this low basal BDNF expression, restoring BDNF to levels similar to those in B6-control mice (Figures 3A,B). We next measured TrkB phosphorylation, the main receptor of BDNF. Whereas phospho-TrkB/total-TrkB ratios were not significantly different between B6-control and BTBR-control mice, early sevoflurane exposures also increased the phospho-TrkB/total-TrkB ratio in BTBR mice (Figures 3A,B; Welch ANOVA with post hoc Tukey’s HSD test, p = 0.02). Similar to the mitochondrial results, sevoflurane exposures did not affect BDNF signaling in B6 mice (Figures 3C,D).

Since our results suggest that early sevoflurane exposures correct pathophysiological mechanisms involved with autistic behavior in BTBR mice, we next investigated for possible changes in autistic behavior. We first tested sociability using the three-chamber test. We initially discovered that both B6 and BTBR mice spent a significant amount of time sniffing the cage containing the stranger mouse (Supplementary Figure 1), an unexpected result given that impaired sociability is a well-established feature of BTBR mice. Previous studies have also shown that BTBR mice may express significant sociability under certain experimental conditions (Bales et al., 2014; Segal-Gavish et al., 2016). A possible explanation for our initial, apparently anomalous, results may be differences in the three-chamber test protocol. We initially used a relatively small three-chamber apparatus (Chung et al., 2015), and used aged-matched B6 mice as a stranger mouse regardless of the strain of the subject mouse. However, most previous studies performed the three-chamber test using a much larger apparatus (Moy et al., 2007), and several studies suggest that the strain of the stranger mouse may have significant effects (Ryan et al., 2019). Thus, we repeated the experiments using a larger three-chamber apparatus and changed the stranger mice to the same strain as the subject mouse. Under these conditions, B6-control mice continued to prefer sniffing the cage containing the stranger mouse (Figure 4A; One-way ANOVA, p < 0.001), whereas BTBR-control mice showed no significant difference in the time spent sniffing the cage containing the stranger mouse and that containing an inanimate object (Figure 4A). However, BTBR-sevoflurane mice displayed normal sociability by spending significantly more time in the chamber containing the stranger mouse (Figure 4A; One-way ANOVA, p = 0.002). The preference index in the three-chamber test was also increased after early sevoflurane exposures in BTBR mice (Figure 4B; One-way ANOVA with post hoc Tukey’s HSD test, p = 0.009), further suggesting that sevoflurane exposures during neurodevelopment may improve sociability in BTBR mice. Self-grooming is often increased in animal models of ASD and has been suggested to represent restricted, repetitive behavior. Similar to previous results, grooming behavior was significantly increased in BTBR-control mice compared with B6-control mice (Figure 4C; One-way ANOVA with post hoc Tukey’s HSD test, p < 0.001). Interestingly, this increased grooming behavior in BTBR mice was also reduced by early sevoflurane exposures (Figure 4C; One-way ANOVA with post hoc Tukey’s HSD test, p = 0.018). Our behavioral results thus suggest that early sevoflurane exposures may reduce autistic behavior by correcting the pathophysiological mechanisms involved in BTBR mice.