All experimental procedures were in accordance with guidelines issued by the Hamamatsu University School of Medicine on the ethical use of animals for experimentation and approved by the Committee for Animal Care and Use (No. 23-049). All efforts were made to minimize the number of animals used and their suffering.

The CRF-VenusΔNeo mouse was produced as described previously (Nagai et al., 2002; Itoi et al., 2014; Kono et al., 2017) and kindly gifted from RIKEN BioResource Research Center (BRC No. RBRC09893). CRF-VenusΔNeo mice express Venus under control of the CRF (CRH) promoter. These mice were bred on a C57BL/6J background. Genotyping of CRF-VenusΔNeo mice was carried out by PCR of mouse tail DNA using the following primers: 5’-AGGACGACGGCAACTACAAG-3’ and 5’-TCTCGTTGGGGTCTTTGCTC-3’. In the present study, male CRF-VenusΔNeo mice were placed with female C57BL/6J mice (CLEA Japan, Tokyo, Japan) overnight in a cage. The day when a vaginal plug was identified was defined as embryonic day (E) 0 and gestational day (G) 0 for the mother. Mice were housed with a 12-h light/dark cycle (lights on at 07:00 am) and access to food and water ad libitum unless otherwise noted.

Coronal brain slices (300 μm in thickness) including the hypothalamus were obtained from male and female mice at E15–E17 and postnatal days (P) 0–2 and 7–9. Under deep anesthesia with ketamine (150 mg/kg body weight)/xylazine (30 mg/kg body weight), mice were decapitated, and their brains were removed. Slices were made using a Vibrating Microtome 7000 (Campden Instruments, Loughborough, Leicestershire, UK) in an ice-cold oxygenated sucrose solution containing (in mM) 220 sucrose, 120 NaCl, 2.5 KCl, 0.5 CaCl₂, 1.25 NaH₂PO₄, 1 MgCl₂, 26 NaHCO₃, 30 glucose, and 10 MgSO₄ (pH 7.4). Slices were maintained in standard artificial cerebrospinal fluid (ACSF) consisting of (in mM) 126 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 2 MgSO₄, 2 CaCl₂, 26 NaHCO₃, and 20 glucose (pH 7.4) equilibrated with 95% O₂ and 5% CO₂ at room temperature before recording. For recording, slices were transferred to a recording chamber that was perfused with oxygenated ACSF.

Electrophysiological recordings were performed using a MultiClamp 700B amplifier (Molecular Devices, San Jose, CA, USA) and pClamp10 software (Molecular Devices, San Jose, CA, USA). Currents were filtered at 2 kHz and digitized at 10 kHz using DigiData1440A. Data were analyzed offline using Clampfit10 (Molecular Devices, San Jose, CA, USA). To estimate the reversal potential of GABA (E_GABA) of CRH neurons in hypothalamic slices, we performed gramicidin-perforated patch-clamp recording to acquire GABA-evoked responses with native intracellular Cl^– concentrations. CRH neurons expressing Venus were selected under epifluorescent illumination. Patch electrode pipettes (4 to 6 megaOhms) were pulled from borosilicate glass capillaries on a P-97 puller (Sutter Instrument, Novato, CA, USA) and filled with pipette solution composed of 150 mM KCl and 10 mM HEPES (pH 7.2) supplemented with gramicidin (Sigma-Aldrich, St. Louis, MO, USA). Gramicidin was dissolved in methanol to prepare a stock solution of 10 mg/mL and then diluted in pipette solution to a final concentration of 30 μg/mL. Recordings were made in ACSF supplemented with 10 μM CNQX (Sigma-Aldrich, St. Louis, MO, USA), 40 μM D-AP5 (Cayman Chemical, Ann Arbor, MI, USA), 3 μM CGP55845 (Sigma-Aldrich, St. Louis, MO, USA), and 500 nM tetrodotoxin (WAKO, Osaka, Japan) to block AMPA- and NMDA-type glutamate receptors, GABA_B receptors, and voltage-gated Na⁺ channels, respectively. The reversal potential of 100 μM GABA-induced current was measured at −70 mV holding potential (V_H), and 1-sec voltage ramps from −90 to −10 mV applied before and during GABA application. E_GABA was estimated by measuring the voltage at which the I-V relationships before and during GABA application intersected (Saitsu et al., 2016). Series resistance (R_s) was compensated by 70%. Membrane potential values were corrected for the voltage drop across the uncompensated R_s using the following equation: V_corr = V_com – I_clamp × R_s, where V_com is the command potential and I_clamp is the clamp current. To determine E_GABA, these values were plotted as a function of the R_s-corrected membrane potential. GABA puffs (5-sec duration) were applied through a patch pipette approximated to the soma using an IM-300 Programmable Microinjector (Narishige, Tokyo, Japan). To study the effect of GABA on membrane potential, current-clamp recording was performed on CRH neurons of P0-P2 mice in gramicidin-perforated patch mode. The bridge balance circuit was applied while recording voltage changes. Recordings were made in ACSF supplemented with 10 μM CNQX, 40 μM D-AP5, and 3 μM CGP55845. GABA puffs (3-sec duration) were applied through a patch pipette approximated to the soma and GABA-induced potential was recorded. Resting membrane potential (RMP) was calculated from traces at zero current injection (I = 0) levels. The calculated liquid junction potential of −3.6 mV was not corrected because the higher K⁺ ion concentration in the electrode solution compared with the cytosol would be negated by the E_K of approximately +4 mV (Kim and Trussell, 2007).

Under deep anesthesia with ketamine/xylazine, E15 C57BL/6J mice were transcardially perfused with 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer. Mouse brains were rapidly removed and post fixed for 2 h in 4% PFA at 4°C, followed by 20 and 30% sucrose in phosphate-buffered saline (PBS) at 4°C overnight. Brains were sectioned coronally (30 μm) with a cryostat (HM560R; Zeiss Microm). Subsequently, sections were incubated for 1 h in blocking solution [10% normal goat serum in 0.1% Tween 20 (Sigma-Aldrich, St. Louis, MO, USA) in PBS (PBS-T)] at room temperature before incubation for 48 h at 4°C with guinea pig anti-CRH (1:800; Peninsula Laboratories, San Carlos, CA, USA) and rabbit anti-KCC2 (1:500; Merck Millipore, Burlington, MA, USA) diluted in PBS-T. Subsequently, sections were washed several times with PBS-T and incubated with the following secondary antibodies: Alexa Fluor 488-conjugated goat anti-guinea pig and Alexa Fluor 594-conjugated goat anti-rabbit (1:1,000; Molecular Probes, Eugene, OR, USA). After several washes with PBS-T, the slides were mounted with PermaFluor aqueous mounting medium (Thermo Fisher Scientific, Wilmington, DE, USA) and coverslipped. Slices were imaged using a confocal laser-scanning microscope (FV1000-D, Olympus, Tokyo, Japan).

CRF-VenusΔNeo pregnant mice were subjected to 70% food restriction stress from G10 to G18. For the food restriction experiment, daily food consumption of pregnant mice was measured from G10 to G18 and average food consumption per gram of body weight was calculated. Pregnant mice were given a food pellet equal to 70% of their average daily food intake from G10 to G18. Body weights were measured during 70% food restriction. For hormone assays, pregnant mice and fetuses at E18 were used. For electrophysiology experiments, pups were raised by naive surrogate mothers with the same delivery date at P0 and recordings were performed using P0–P2 mice.

Pregnant mice at E18 were sacrificed by cervical dislocation. Mice were decapitated and the trunk blood was collected into polyethylene tubes containing EDTA-2K (Becton, Dickinson and Company, Franklin Lakes, NJ, USA). Samples were collected between 09:00 and 11:00 am Samples were centrifuged at 800 g for 20 min at 4°C, and the serum was collected and stored at −80°C until further use. As fetuses at E18 provide an insufficient amount of blood, the fetus body was homogenized with a homogenizer in 500 μL of 1 N HCl, followed by centrifugation at 9,100 g for 15 min at 4°C. The resulting supernatant was collected and stored at −80°C. For analysis, 100 μL of each supernatant was lyophilized and reconstituted in 300 μL of radioimmunoassay (RIA) buffer [0.1 M PB (pH 7.4) containing 0.05% NaN₃ and 0.1% Triton X-100]. Corticosterone levels were determined by RIA, as previously reported (Yesmin et al., 2022). Briefly, a 25-μL sample of serum was boiled at 98°C for 5 min. Ice-cold RIA buffer was used for dilution and ¹²⁵I-corticosterone (Institute of Isotopes) was used as the label. A mixture comprising 100 μL of corticosterone standard or sample, 100 μL of corticosterone antiserum, and 100 μL of ¹²⁵I-labeled corticosterone was incubated for 24 h at 4°C. Antibody-bound and antibody-free corticosterone were separated via incubation with 100 μL of a secondary antibody (bovine γ-globulin) and 400 μL of 25% polyethylene glycol, followed by centrifugation at 800 g for 15 min at 4°C. The radioactivity of bound antibodies was counted using a γ-counter (ARC-7010, Aloka, Tokyo, Japan). The assay did not cross-react with other corticosteroids and its sensitivity was 2 pg/tube.

All statistical analyses were performed using GraphPad Prism (version 8; GraphPad Software). The data was evaluated for normal distribution based on Shapiro–Wilk test and measure of variance was based on F-Test. If the data was found to be normally distributed, we performed unpaired t-test. When the data failed tests for normality, nonparametric tests for comparison between two groups (Mann–Whitney U test) and multiple groups (Kruskal–Wallis one-way ANOVA followed by Dunn’s multiple comparison test) were used, respectively. Data are presented as mean ± standard error of the mean (SEM) as well as median in the text. When the data is statistically compared using nonparametric test, graphical representation of results in figures show box plot that indicates median, 25 and 75th percentile, and whiskers extending from each end of a box depict the minimum and maximum ranges. The individual data points are shown with dots. The confidence interval for statistics was set at 95% for all data comparisons. P-value < 0.05 is indicated as * and < 0.01 as **. No difference between groups was shown as ns.

The activity of CRH neurons is reportedly regulated by GABA in adult mice (Camille Melón and Maguire, 2016). GABAergic inputs exert inhibitory action on the somata of CRH neurons through activation of GABA_A receptors (GABA_ARs) (Kakizawa et al., 2016). To investigate when GABAergic regulation of CRH neurons is established and the development of chloride homeostasis, we measured E_GABA determined primarily by the Cl^– equilibrium potential [reflecting intracellular [Cl^–]_i (Medina et al., 2014)] using gramicidin-perforated patch-clamp technique, which avoids disruption of native neuronal [Cl^–]_i. Using CRF-VenusΔNeo mice, CRH neurons were identified by Venus fluorescence. We recorded E_GABA of CRH neurons in hypothalamic slices from E15–E17, P0–P2, and P7–P9 mice (Figure 1A). The mean E_GABA of CRH neurons was −61.3 ± 1.4 mV at E15–E17 and −60.6 ± 2.4 mV at P0–P2 (Figure 1B). We observed no significant differences in mean E_GABA values between E15–E17 and P0–P2. The E_GABA of P7–P9 CRH neurons was significantly shifted to more hyperpolarized values (P7–P9: −68.8 ± 1.5 mV) compared to E15–E17 and P0–P2 CRH neurons (p < 0.05 by Kruskal–Wallis test). The corresponding median values for E15-E17 and P0-P2 were −60.8 and −60.7 mV, respectively, whereas for P7-P9 was −67.9 mV. These results suggest that the developmental shift of E_GABA occurs between P0 and P7, and E_GABA of perinatal CRH neurons is hyperpolarized compared to neurons in other brain regions, such as the cortex and hippocampus. During gramicidin-perforated patch-clamp recording, the neuronal membrane occasionally ruptured and spontaneous inhibitory postsynaptic currents (sIPSCs) were observed from E15 CRH neurons when voltage-clamped (V_H) at −70 mV (Figure 1C), suggesting that E15 CRH neurons express synaptic GABA_ARs and receive GABAergic inputs at E15.

Whether activation of GABA_A receptors results in depolarization or hyperpolarization of membrane potential depends on the relationship of E_GABA to the RMP. Next, we performed current-clamp recordings from CRH neurons in hypothalamic slices obtained from P0 mice and examined the effect of GABA on membrane potential changes. Most CRH neurons showed hyperpolarizing responses to brief application of GABA (92%; 13 out of 14 neurons; Figures 2A, C). Only one CRH neuron displayed a depolarizing response to GABA (8%: 1 out of 14 neurons). Two CRH neurons showed spontaneous action potentials that were inhibited by GABA application (Figure 2B). The RMP of CRH neurons at P0–P2 was −41.7 ± 1.3 mV (Figure 2D) and positive to E_GABA. These results suggest that GABA action is already inhibitory at P0–P2.

The chloride cotransporter KCC2 is essential for GABAergic inhibition. In the cortex and hippocampus, the developmental shift of GABA action from excitatory to inhibitory, which results from KCC2 dependent Cl^– extrusion, occurs between P7 and P14 (Ben-Ari, 2002; Blaesse et al., 2009). Therefore, to examine if KCC2 is expressed in CRH neurons of E15 wild-type (WT) mice, we performed immunohistochemistry for CRH and KCC2. We observed CRH expression in the cell bodies of CRH neurons in the PVN and CRH terminals in the ME (Figures 3A, B), suggesting that CRH neurons already exist in the PVN and CRH axons project to the ME at E15. KCC2 expression in the somata of CRH neurons in the PVN was detected at E15, albeit weakly (Figure 3A), suggesting that it is expressed and functional at this age. In the ME, KCC2 expression in the CRH neuron terminals was not detected, similar to our report that KCC2 is not expressed in the CRH neuron terminals of adult mice (Figure 3B; Kakizawa et al., 2016). Cl^– homeostasis in early developmental CRH neurons attains mature [Cl^–]_i levels with GABA acting in a predominantly inhibitory manner during their early development.

Stress reportedly inhibits KCC2 function in adult CRH neurons of the PVN (Hewitt et al., 2009; Sarkar et al., 2011). Therefore, we examined whether prenatal stress affects chloride homeostasis in CRH neurons during early development. We used a maternal food restriction stress model that subjected pregnant mice to 70% food restriction (FR) from G10 to G18. Prenatal stress began on G10 because it corresponds with development of the fetal central nervous system (Clancy et al., 2001) and minimizes premature termination of the pregnancy as a result of excessive stress. First, we confirmed that FR affects the body weights and corticosterone levels of both the mother and fetus. Body weights of FR mothers were significantly decreased from G13 to G18 compared with control mothers (Figure 4A; p < 0.05, p < 0.01 by Mann–Whitney U test; median value: Control: 28.9 g and FR: 25.8 g at E13, Control: 35.85 g and FR: 30.6 g at E17). Fetal body weights from FR mothers tended to decrease, although not significantly (Figure 4B; ns by Mann–Whitney U test; Control: 1.261 ± 0.03 g; median value: 1.217 g, FR: 1.196 ± 0.02 g; median value: 1.19 g). Maternal serum corticosterone levels were significantly increased in FR mothers compared with control mothers (Figure 4C; p < 0.01 by unpaired t-test). Pups from FR mothers showed increased corticosterone levels compared with control pups (Figure 4D; p < 0.01 by unpaired t-test). These results confirm that FR activated the HPA axis of mothers and pups. To examine the effects of FR on chloride homeostasis in CRH neurons, we examined the effects of FR on E_GABA of CRH neurons from P0–P2 pups. E_GABA of CRH neurons from FR mothers was shifted to more depolarized values (Figures 4E–G; Control: −60.6 ± 2.4 mV, FR: −51.8 ± 2.0 mV; p < 0.01 by unpaired t-test) compared with previously observed E_GABA values from control pups (shown in Figure 1), suggesting that FR alters the chloride homeostasis of CRH neurons and reduces their inhibitory control by GABA. Such alterations may contribute to increased CRH neuron activity and cause activation of the HPA axis.