Bumetanide, an Inhibitor of Cation-chloride Cotransporter Isoform 1, Inhibits γ-Aminobutyric Acidergic Excitatory Actions and Enhances Sedative Actions of Midazolam in Neonatal Rats

IN the immature hippocampus and neocortex, γ-aminobutyric acid (GABA) causes neuronal excitation via activation of GABA_A receptors,¹  resulting in increased intracellular concentration of Ca²⁺ ([Ca]²⁺i) through opening of voltage-dependent Ca²⁺ channels and N-methyl-d-aspartate receptor channels.^2–5  GABA-induced [Ca]²⁺i increase is considered to play important roles in the development of neuronal circuits.^6,7  The underlying mechanisms for this neuronal excitation by GABA include the balance of Na⁺–K⁺–2Cl⁻ cotransporter isoform 1 (NKCC1) and K⁺–Cl⁻ cotransporter isoform 2 (KCC2) levels.^8,9  In the immature brain, NKCC1 is increased while KCC2 is low, and this balance is reversed during early postnatal development. Because NKCC1 acts to uptake Cl⁻ into the cytoplasm and KCC2 primarily works as an extruder of Cl⁻, immature neurons have significantly increased intracellular concentration of Cl⁻ ([Cl⁻]i) compared with mature neurons. Therefore, the equilibrium potential for Cl⁻ (E_Cl) becomes more positive than the membrane resting potential.^1,8  Thus, the activation of GABA_A receptors during early brain development induces Cl⁻ efflux and neuronal depolarization. In addition, E_Cl shift close to the resting membrane potential may reduce hyperpolarization, resulting in attenuation of inhibitory effects of GABA.

Peak expression of NKCC1 in the rodent neocortex is considered to occur around postnatal days 5–7 (P5–P7), and KCC2 expression levels gradually increase during the second postnatal week.¹⁰  Developmental changes of the expression patterns of NKCC1 and KCC2 occur early in the caudal parts of the central nervous system and extend to the rostral parts.¹¹  These changes underlie the developmental switch of GABA_A signaling from excitation to inhibition,^8,9  which is considered to occur during the second postnatal week in the rodent hippocampus and neocortex.^12–14 

Previous studies using cultured neonatal neurons showed that propofol induces [Ca]²⁺i increase¹⁵  and isoflurane enhances spontaneous Ca²⁺ oscillations¹⁶  and that these effects can be blocked by bicuculline, a GABA_A receptor antagonist. However, it has not been well defined whether anesthetic agents with GABA_A receptor stimulating actions induce excitation or potentiate excitatory effects of GABA in immature brain slices or in vivo.

GABA-induced depolarization in immature neurons is blocked by bumetanide, a specific inhibitor of NKCC1.^8,10  Bumetanide was shown to attenuate epileptogenic and neurotoxic effects of sevoflurane in neonatal rat brain¹⁷  and to enhance the anticonvulsant efficacy of phenobarbital in a neonatal seizure model.¹⁸ 

Consequently, we hypothesized that midazolam, a GABA-mimetic hypnotic agent, causes neuronal excitation in brain slices and in vivo, and that this is dependent on NKCC1. To test these hypotheses, we studied the effects of midazolam and bumetanide on [Ca]²⁺i and the expression level of a neuronal activity-dependent marker in brain slices.

Furthermore, it has been shown that midazolam lacks sedative effects in P3 rats, in contrast to its overt sedative actions in P21 rats.¹⁹  If this loss of sedation in P3 rats is due to GABAergic excitation mediated by NKCC1, it is predicted that bumetanide would interfere with loss of sedation and may reveal sedative effects of midazolam. In this study, we also tested this hypothesis using righting reflex in rats.

All animal care procedures in this study were in accordance with the standards approved by Yokohama City University Institutional Animal Care and Use Committee (Yokohama, Japan). The experimental protocol number issued by our institution was F–A–12–032. Sprague–Dawley rats at the ages of 4, 7, and 28 days were used in this study. All animals were obtained from Japan SLC Corporation (Shizuoka, Japan) and were kept in cages with their littermates and mothers in a temperature-controlled animal care facility room with a 12-h light/dark cycle. We made all efforts to minimize animal suffering and the number of animals used.

We performed a Ca²⁺ imaging study in hippocampal slices prepared from Sprague–Dawley rats at P4, P7, and P28. Rats were anesthetized deeply by inhalation of isoflurane (Escain; Mylan Inc., Canonsburg, PA) before decapitation. After decapitation, the brain was removed quickly from the skull and transferred into an ice-cold (0–4°C) solution containing 220 mm sucrose, 26 mm NaHCO₃, 10 mm d-glucose, 2 mm sodium pyruvate, 2 mm KCl, 10 mm MgCl₂, 0.5 mm CaCl₂, 1.25 mm NaHPO₄, 4 mm dl-lactate, and 2 mm kynurenic acid, bubbled with 5% CO₂–95% O₂, and then recovered for approximately 5 min in the buffer. After trimming, sagittal brain slices of 150-µm thickness were cut on a Vibratome (LinearSlicer PRO7; Dosaka EM, Kyoto, Japan) while submerged in the same ice-cold buffer. Then, the slices were kept in artificial cerebrospinal fluid (ACSF) bubbled with 5% CO₂–95% O₂ for at least 1 h at room temperature before undergoing fura-2 fluorescent staining. For the experiments using bumetanide, the slices were kept in ACSF containing 20 μm of bumetanide. ACSF contains the following: 125 mm NaCl, 3.5 mm KCl, 1.25 mm NaHPO₄, 26 mm NaHCO₃, 10 mm d-glucose, 2 mm CaCl₂, and 1 mM MgCl₂.

We measured [Ca]²⁺i in the CA3 hippocampal area using fura-2 microfluorometry at room temperature. After 1-h recovery, brain slices were incubated for 45 min at room temperature with 15 µm fura-2-acetoxymethyl ester and 0.01% Pluronic F127 in ACSF bubbled with 5% CO₂–95% O₂. After the slices were rinsed again for 10 min in ACSF, they were transferred to a bath on the stage of an IX70 microscope (Olympus, Tokyo, Japan). The slices were perfused with ACSF bubbled with 5% CO₂–95% O₂ at an approximate rate of 6 ml/min in a bath volume of approximately 2 ml. The somata of individual pyramidal cells of the hippocampal CA3 area in the brain slices were illuminated with a xenon lamp at 10-s intervals, and images were obtained with a charge-coupled device camera (ORCA-ER^®; Hamamatsu Photonics, Hamamatsu, Japan). We recorded the fluorescence intensity of the somata at 510 nm elicited by 340 and 380 nm and determined the fluorescence ratio (i.e., 340/380 nm excitation) using C-imaging software (Hamamatsu Photonics).

We measured changes in fluorescence ratios in response to the bath application of isoguvacine (Sigma, St. Louis, MO) at 10 μm or midazolam (Sandoz, Holzkirchen, Germany) at 0.01, 0.1, and 1 μm, with or without pretreatment with 20 μm of bumetanide (Sigma), 50 μm of bicuculline methobromide (Sigma), or 10 μm of nicardipine (Wako, Tokyo, Japan). After determining regions of interest, fluorescence ratios were recorded before application of isoguvacine or midazolam for 60 s as the baseline. Then, the recording was continued during isoguvacine or midazolam application for 420 s. For the experiments without pretreatment, the perfusate was plain ACSF (ACSF group). For those involving pretreatment with bumetanide, bicuculline, or nicardipine, the perfusate contained one of these agents (bumetanide, bicuculline, or nicardipine groups), and isoguvacine or midazolam was applied in the continuous presence of these agents. We also recorded fluorescence ratios without isoguvacine or midazolam application for 480 s to obtain time control data instead of recording the washout process (control group) because washout was extremely slow.

To standardize baseline ratios of individual pyramidal cells from different samples, the fluorescence ratio (340/380 nm) was normalized to the mean ratio calculated by averaging the values during baseline. We calculated both the relative ratios at each time point at 10-s intervals and the mean relative ratios during baseline (0–60 s), the early phase (120–180 s), and the late phase (420–480 s) of drug application in all groups. We tested three slices from three animals (one by one) in each condition. Approximately 10 regions of interest were chosen in each slice, resulting in approximately 30 regions of interest in each condition. For the experiments using nicardipine, approximately seven regions of interest were chosen in each slice.

Twenty-four rats were randomly divided into equal-sized groups according to the four combinations of the drugs to be administered, resulting in four groups of six rats each for P4, P7, and P28 rats, respectively. All drugs were administered by intraperitoneal injection. The four drug combinations were (1) midazolam (20 mg/kg) 1 h after bumetanide (10 µmol/kg; bumetanide + midazolam group), (2) midazolam 1 h after saline (same volume as the vehicle for bumetanide; saline + midazolam group), (3) saline (same volume as the vehicle for midazolam) 1 h after bumetanide (bumetanide group), and (4) saline (same volume as the vehicle for midazolam) 1 h after saline (same volume as the vehicle for bumetanide; control group). We compared righting reflex latencies between the four groups. The righting reflex latency was measured by recording with a stopwatch the time required for the rat to turn over and stand upright after placing it in the supine position. The latencies were also measured at the time points of 15 and 30 min after administration of the second drugs. Three measurements were made, and the mean latency was calculated at each time point. If the rats failed to right themselves within 60 s, we recorded the righting latency to be 60 s and turned them to the upright position and continued the measurements. During the intervals between the measurements, each animal was placed in the cage holding its mother and littermates.

Eighteen rats were randomly divided into equal-sized groups according to the three combinations of the drugs to be administered, resulting in three groups of six rats each for P7 and P28, respectively. All drugs were administered by intraperitoneal injection. For the behavior study, the three combinations of the drugs were the same as those for the bumetanide + midazolam, saline + midazolam, and control groups. At 45 min after the second drugs were administered, rats were anesthetized deeply by inhalation of isoflurane for approximately 10 s and perfused transcardially with 4% paraformaldehyde in 0.05 m phosphate-buffered saline before decapitation. After decapitation, the brain was removed quickly from the skull and placed into 4% paraformaldehyde in 0.05 m phosphate-buffered saline at 4°C for 72 h. The brains were then transferred into a solution of 20% sucrose in distilled water and incubated at 4°C for 72 h. After incubation, the brains were transferred into optimal cutting temperature compound (Tissue-Tek^®; SAKURA, Tokyo, Japan) and frozen at −80°C. Then, 20-µm thick frozen coronal sections of the brains were cut at −20°C with a LEICA 1800 Cryostat Microtome (Leica Microsystems, Wetzlar, Germany) and placed on the slide glass. We chose two slices from each brain that were comparable with the sections of figures 35 and 36 in the atlas of Paxinos and Watson.²⁰  Immunostaining of phosphorylated cyclic adenosine monophosphate–response element–binding protein (pCREB) was performed as described previously.^21,22  Briefly, after permeabilization, suppression of endogenic peroxidase, and blocking of nonspecific binding of proteins, the slices were incubated overnight at 4°C with a primary rabbit antibody against pCREB. The next day, the slices were incubated with a goat anti-rabbit biotinylated secondary antibody. Then, the secondary antibody was detected by avidin/biotinylated enzyme complex technique, and peroxidase was developed by 3,3’-diaminobenzidine.

We selected the hippocampal CA3 area and thalamus in the right hemisphere of all samples to count the number of pCREB-positive cells. The images were photographed and the pCREB-positive cells were counted with a Biorevo BZ 9000^® microscope (Keyence Corporation, Osaka, Japan) with ×4 and ×10 magnification lenses by an operator blinded to the treatments.

Data are expressed as mean with SEM. In the Ca²⁺ imaging study, we first examined the full interactive term of the pretreatment, the drugs, and the time points in P4 and P7 rats by the three-way repeated measure ANOVA. Then, we compared the mean relative ratios at three time periods (baseline, early phase, and late phase) within the same pretreatment group (ACSF, bumetanide, or bicuculline group) of each age group by two-way repeated ANOVA. When the interactive term of the drugs and the time points was significant, the values at each period were compared with the corresponding values of the control group by Dunnett post hoc test. In assessment of sedation levels, we compared righting reflex latencies between the four groups within each age group with nonparametric statistical methods, the Kruskal–Wallis test followed by the Mann–Whitney U test with Bonferroni correction. Therefore, values of P less than 0.0083 (0.05/6) were considered statistically significant. We compared righting reflex latencies of the saline + midazolam group between three age groups with the same methods used for comparison within each age group. We also compared the latencies of the bumetanide + midazolam group in P4 and P7 rats with those of the saline + midazolam group in P28 rats in the same way. In these comparisons, values of P less than 0.0167 (0.05/3) were considered statistically significant. Comparisons of the number of pCREB-positive cells among the three groups within each age group were analyzed by one-way ANOVA followed by Tukey post hoc tests. We employed two-tailed tests in all comparisons. Unless stated otherwise, values of P less than 0.05 were considered statistically significant. Statistical analysis was performed with SPSS 11.0J for Windows (SPSS Inc., Chicago, IL).

First of all, we performed the three-way repeated ANOVA (pretreatment × drugs × time points) in P4 and P7 rats. In the ANOVA, the interactive term of the pretreatment and the time points was significant for the relative ratios in P4 rats (P = 0.018; F = 5.44) and P7 rats (P = 0.003; F = 10.18). The results of the ANOVA suggest that the differences in the pretreatment affect the results of the ratios. Then, we performed the two-way repeated ANOVA (drugs × time points) to detect which drugs significantly changed the ratios compared with the control while fixing the condition of pretreatment. In the two-way repeated ANOVA of the relative ratios, the interactive term of the drugs and the time points was significant in the P4 (P < 0.001; F = 14.5) and P7 age groups (P = 0.002; F = 5.6) of the ACSF group. There was no significant interaction in other pretreatment groups (bumetanide, bicuculline, or nicardipine groups). In the control of the ACSF group, the relative ratios gradually increased in the P4 and P7 age groups; however, the changes were not statistically significant. The application of 0.1 µm midazolam or 10 µm isoguvacine induced a significant increase in the relative ratio compared with the control in the ACSF group at P4. Significant increases were observed at the early phase for isoguvacine and at the late phase for midazolam. However, 0.01 or 1.0 µm midazolam induced no significant increases in P4 rats. The increases in the relative ratio were not observed when slices were pretreated with 20 µm bumetanide, 50 µm bicuculline, or 10 µm nicardipine in P4 rats (fig. 1).

In P7 rats, 0.1 µm midazolam induced a significant increase in the relative ratio at the late phase. Both 20 µm bumetanide and 50 µm bicuculline inhibited the increases during application of 0.1 µm midazolam in P7 rats (fig. 2). In contrast to the results of P4 and P7 rats, neither midazolam nor isoguvacine significantly changed the relative ratio in P28 rats (fig. 3).

As shown in figures 1C and 2C, the changes in relative fluorescence ratios in response to midazolam or isoguvacine application did not seem to reach steady-state levels in the presence of bumetanide pretreatment. These findings raised the possibility that bumetanide slowed the changes in relative ratios but did not prevent them. We additionally measured the relative ratios in response to a longer period of application of midazolam from 60 to 660 s in P7 slices with or without bumetanide pretreatment. The results of the additional study indicated that the changes in relative ratios reached a stable level during the measurement in bumetanide-pretreated slices and that bumetanide prevented the midazolam-induced increase in relative ratios. We also found that bumetanide pretreatment prevented isoguvacine-induced increase in relative ratios during the extended period of isoguvacine application (data not shown).

Midazolam exerted sedative effect in P4 and P7 rats compared with the control groups. Latency to righting was significantly longer in the bumetanide + midazolam group versus the saline + midazolam group at both P4 and P7 (fig. 4). The latencies were 15.7 ± 4.3 and 59.3 ± 0.5 s for the saline + midazolam and bumetanide + midazolam groups of P4 rats and were 7.3 ± 1.4 and 54.7 ± 3.7 s for the saline + midazolam and bumetanide + midazolam groups of P7 rats, respectively. In P28 rats, although midazolam showed a sedative effect, there were no differences in latencies to righting between the saline + midazolam group (47.8 ± 8.9 s) and the bumetanide + midazolam group (51.2 ± 5.5 s; fig. 4). There were no differences between the control and bumetanide groups in the latencies at each age. In addition, we compared the latencies of the saline + midazolam group between P4, P7, and P28 rats. Latency to righting in the P28 rats was significantly longer than that in the P4 and P7 rats 15 min after the administration of midazolam (fig. 5). There was no significant difference in the latencies at P4 and P7 in the bumetanide + midazolam group and at P28 in the saline + midazolam group. Taken together, we found that midazolam exerted less sedative effects in neonatal rats compared with P28 rats, and pretreatment with bumetanide enhanced sedative effects of midazolam in neonatal rats but not in P28 rats, resulting in reversal of the reduction of sedation.

The number of pCREB-positive cells in the hippocampal CA3 area was significantly higher in the saline + midazolam group than in the bumetanide + midazolam and control groups of P7 rats (771.9 ± 86.4, 269.5 ± 44.4, and 148.9 ± 27.4 for the saline + midazolam, bumetanide + midazolam, and control groups, respectively). There were no significant differences in the number of pCREB-positive cells in the thalamus between the three groups at P7 (fig. 6). In contrast to the results of neonates, there were no significant differences in pCREB expression in the hippocampal CA3 area between the three groups in P28 rats. The numbers of cells were 120.2 ± 19.5, 133.4 ± 22.9, and 119.3 ± 15.7 for the saline + midazolam, bumetanide + midazolam, and control groups, respectively. Interestingly, the saline + midazolam and bumetanide + midazolam groups showed lower expression of pCREB in the thalamus than did the control group in P28 rats (fig. 7).

In this study, we showed that midazolam induced increases in [Ca]²⁺i and pCREB expression in the hippocampal CA3 area of the neonatal rat brain, and these actions were inhibited by bumetanide, an NKCC1 inhibitor. Moreover, we found that bumetanide enhanced the sedative effects of midazolam in neonatal rats, resulting in reversal of the reduction of sedative effects. Bumetanide was found to be without effect on midazolam-induced sedation in young rats. Ca²⁺ imaging study showed that isoguvacine- and midazolam-induced increases in [Ca]²⁺i were abolished by nicardipine, a blocker for L-type voltage-dependent Ca²⁺ channels, suggesting that isoguvacine and midazolam may induce the membrane depolarization leading to Ca²⁺ influx through voltage-dependent Ca²⁺ channels. It is likely that depolarization by midazolam induces excitation in this region because depolarization induced by GABA agonists is shown to cause excitation in immature hippocampal and cortex neurons.^1,3,7,13  Taken together, these results suggest that sedative actions of midazolam may be hampered by excitatory effects of GABA_A receptor stimulation in immature brain, in which NKCC1 is dominant relative to KCC2.

We selected the hippocampal CA3 area for Ca²⁺ imaging study because this area under development has been frequently used to investigate excitatory effects of GABA.^3,5,12  We found that 0.1 µm midazolam significantly increased [Ca]²⁺i of this area in P4 slices. Typical plasma concentrations of midazolam are reported to be 1.66–2.55 µm during sedation for mechanical ventilation in neonatal patients.²³  The concentration in the cerebrospinal fluid is considered as a surrogate measure to predict brain interstitial fluid concentration, which corresponds to the concentration in the perfusate.²⁴  The concentration ratio of cerebrospinal fluid/plasma of midazolam was reported to be 0.056 in mice.²⁵  If we assume that the cerebrospinal fluid/plasma ratio in humans is comparable with that in mice, the range of the bath concentrations of clinical relevance would be 0.09–0.14 µm. Therefore, we considered 0.1 µm as the clinically relevant concentration, and we chose concentrations 10 times lower and higher to study dose-dependent changes.

The time course of changes in relative ratios was much slower during midazolam application than that during isoguvacine application. Whereas isoguvacine directly activates GABA_A receptors,^26,27  midazolam is considered to enhance the affinity of endogenous GABA to the receptors. The difference in the mode of the action, together with the slow rate of wash in, may account for the slow time course of the changes, at least in part. The high concentration of midazolam failed to induce an increase in [Ca]²⁺i in P4 slices. One possible reason may be the inhibitory effect of high concentrations of benzodiazepines on voltage-dependent Ca channels.²⁸  In P7 slices, 0.1 µm midazolam induced a significant increase in [Ca]²⁺i; however, isoguvacine-induced increase did not reach statistical significance. The reason for this is not clear; however, the limited time resolution of our recording system might obscure a transient raise in [Ca]²⁺i induced by isoguvacine. Midazolam-induced increase in [Ca]²⁺i was blocked by bicuculline and bumetanide in slices from neonatal rats. Moreover, isoguvacine- and midazolam-induced increases in [Ca]²⁺i were completely blocked by nicardipine in slices from P4 rats. This result suggests that isoguvacine- and midazolam-induced [Ca]²⁺i increases are mediated by the opening of voltage-dependent Ca²⁺ channels associated with the membrane depolarization. Both isoguvacine and midazolam failed to increase [Ca]²⁺i in slices from P28 rats. These findings suggest that midazolam-induced increases in [Ca]²⁺i may be mediated by GABA_A receptor activation followed by depolarization due to NKCC1-dependent changes in E_Cl.

Next, we tested the hypothesis that excitatory actions of midazolam interfere with sedative effects and result in attenuation of sedative effects of midazolam in neonatal rats. To our knowledge, it has not been clarified how immature GABA_A receptor signaling influences hypnotic actions of anesthetics that modulate GABA_A receptors, or how bumetanide alters hypnotic actions of these anesthetics in neonatal rodents. We found that 20 mg/kg of intraperitoneal midazolam induced less sedation in P4 and P7 rats than in P28 rats. We also found that pretreatment with bumetanide reversed the reduction in sedative effects in P4 and P7 rats and was without effect in P28 rats. These results suggest that sedative effects of midazolam may be hampered by GABA_A receptor–mediated excitation sensitive to bumetanide in neonatal rats.

A previous report showed that 10 mg/kg subcutaneous administration of midazolam had sedative effects in P21 rats but not in P3 rats.¹⁹  Preliminarily, we examined the righting reflex latency in P28 rats after intraperitoneal administration of 10 mg/kg of midazolam, but found no prolongation of the latency. We tested several different doses of intraperitoneal midazolam and used a 20 mg/kg dose because this dose induced consistent sedation in P28 rats. The larger dose used in our study may explain the difference in the results of these two studies. When bumetanide was administered without midazolam in our study, it did not induce significant changes in latency. Therefore, it is unlikely that diuresis caused by bumetanide would affect the normal righting reflex of the rats.

Cyclic adenosine monophosphate–response element–binding protein is a representative activity-dependent transcription regulator.²⁹  CREB is activated by phosphorylation at the serine-133 site, resulting in up-regulation of pCREB through signal transduction secondary to [Ca]²⁺i increase induced by various stimuli such as excitatory neurotransmission or growth factor treatment.^29,30  Therefore, pCREB is one of the major markers of Ca²⁺ influx and neuronal activity. GABA induces CREB phosphorylation in developing neurons.³¹  Mantelas et al.³²  showed that diazepam caused significant up-regulation of pCREB in the cortex of P5 rats, which is dependent on activation of GABA_A receptors and L-type Ca channels. We found that midazolam induced significant up-regulation of pCREB at the hippocampal CA3 in P7 but not P28 rats and that bumetanide inhibited these changes. These findings are consistent with the results of the Ca²⁺ imaging study and provide further support to our hypothesis.

In contrast to the results of the hippocampus, no excitatory effects were induced by midazolam in the thalamus from P7 rats. Glykys et al.³³  found that the effects of GABA_A receptor activation are inhibitory in the ventroposterior thalamus but excitatory in the neocortex of neonatal rats because the ventroposterior thalamus has lower [Cl⁻]i than does the neocortex during early postnatal development due to the earlier maturation of the expression pattern of cation-chloride cotransporters in this region. Therefore, the differential maturation of GABA_A receptor signaling seems to underlie the differential effects of midazolam on pCREB expression in the hippocampus and thalamus, although other factors such as differences in sensitivities of GABA_A receptor subtypes to midazolam between two regions might be also involved. It has been shown that the hippocampus participates in sedative or hypnotic actions of general anesthetics.³⁴  Our results suggest the possibility that reduced sedative actions of midazolam might be attributed to excitatory actions on certain regions of immature brain including the hippocampus, where neuronal [Cl⁻]i is increased by NKCC1, at least in part.

Our study has several limitations. First, the Ca²⁺ imaging study lacks data regarding reversibility of the effects of the drugs and has limited time resolution. These factors potentially affect the results. Second, interpretation of the results is largely dependent on the assumption that effects of bumetanide are specific to inhibition of NKCC1 in the brain. However, both in vitro and in vivo experiments showed consistent effects of bumetanide, and there has been no evidence that bumetanide acts on other targets at the dose used in this study. Third, we did not directly study effects of midazolam on the membrane potential or excitability of neurons. Although the results of the Ca²⁺ imaging study with nicardipine suggest that midazolam may cause the membrane depolarization in the immature hippocampus, further studies are needed to confirm and characterize the excitation induced by midazolam.

In the human cortex, the developmental change in the dominant type of cation-chloride cotransporters is considered to occur later in the 41st postconceptional week.¹⁰  Therefore, midazolam may show reduced sedative effects in preterm and early neonatal patients via mechanisms suggested in this study. Sensitivity to midazolam may also be modulated by other mechanisms, such as differences in pharmacokinetic properties and different subunit compositions of GABA_A receptors in this population.^35,36  It is possible that benzodiazepines preferentially inhibit the caudal central nervous system rather than the rostral brain because of different timings of the developmental switch of GABA_A receptor signaling in these regions. This phenomenon of discrepancy in inhibitory effects on the caudal and rostral central nervous system is analogous to the clinical observation that barbiturates and benzodiazepines are less effective in suppressing electroencephalogram activity rather than the motor component of clinical neonatal seizures.^37,38  Bumetanide is known to enhance antiepileptic effects of these drugs as judged by electroencephalogram activity in neonatal seizure models of rodents.¹⁸  It is plausible that bumetanide may potentiate sedative effects of these hypnotic drugs in neonatal patients; however, future study is needed to confirm this speculation.

In summary, our results suggest that midazolam may exert excitatory actions in the developing hippocampal CA3 area and reduced sedative effects in neonatal rats, both in a NKCC1-dependent manner. These findings suggest that GABA_A receptor–mediated excitation may underlie the attenuated sedative effects of midazolam in neonatal rats.