Research ArticleNEUROSCIENCE

Depolarizing GABAA current in the prefrontal cortex is linked with cognitive impairment in a mouse model relevant for schizophrenia

See allHide authors and affiliations

Science Advances  31 Mar 2021:
Vol. 7, no. 14, eaba5032
DOI: 10.1126/sciadv.aba5032


Cognitive impairment in schizophrenia (CIAS) is the most critical predictor of functional outcome. Limited understanding of the cellular mechanisms of CIAS hampers development of more effective treatments. We found that in subchronic phencyclidine (scPCP)–treated mice, an animal model that mimics CIAS, the reversal potential of GABAA currents in pyramidal neurons of the infralimbic prefrontal cortex (ILC) shifts from hyperpolarizing to depolarizing, the result of increased expression of the chloride transporter NKCC1. Further, we found that in scPCP mice, the NKCC1 antagonist bumetanide normalizes GABAA current polarity ex vivo and improves performance in multiple cognitive tasks in vivo. This behavioral effect was mimicked by selective, bilateral, NKCC1 knockdown in the ILC. Thus, we show that depolarizing GABAA currents in the ILC contributes to cognitive impairments in scPCP mice and suggest that bumetanide, an FDA-approved drug, has potential to treat or prevent CIAS and other components of the schizophrenia syndrome.


Cognitive impairment associated with schizophrenia (CIAS) is a core feature of schizophrenia (SCZ) (1, 2) and the strongest predictor of poor functional outcome (3) in this heterogeneous syndrome that is the result of multiple genetic and environmental influences (4). Atypical antipsychotic drugs (AAPDs) are the main treatment for SCZ (5), but they only produce a partial remediation of CIAS and only in some patients (6, 7). Moreover, the cellular mechanisms underlying CIAS and the site of action of AAPDs to ameliorate CIAS are poorly understood. Acute or repeated administration of the N-methyl-d-aspartate receptor (NMDAR) noncompetitive antagonist phencyclidine (PCP) to rodents and nonhuman primates impairs numerous cognitive tasks, which are also affected in CIAS, as well as produces deficits in social interaction and lowered threshold for psychotic-like behaviors (5, 8, 9). Many types of evidence suggest that impaired inhibitory transmission and an excitation/inhibition (E/I) imbalance play critical roles in CIAS (1012) and that the beneficial effects of AAPDs in CIAS may result from normalization of such imbalance (5). The mechanisms of γ-aminobutyric acid–releasing (GABAergic) system involvement in CIAS, however, remain controversial, including the direction of the E/I imbalance and whether it is selective for specific brain areas (1315).


To fill the knowledge gap concerning the cellular mechanisms of CIAS, we combined ex vivo cellular analysis with in vivo behavioral studies to clarify the role of GABAA signaling in the medial prefrontal cortex (mPFC) of subchronic PCP (scPCP)–treated mice, an intensively studied animal model that mimics CIAS symptoms (16, 17). The functional effects of the GABAA current critically depend on its reversal potential (18), and even small alterations of this parameter would heavily shift the E/I balance (19). The polarity of GABAA currents in the CNS is dynamically modulated, as it shifts from depolarizing to hyperpolarizing during early postnatal development, critically shaping brain circuitry (20, 21). Because chloride is the main permeant ion for GABAA channels under physiological conditions (22), the current reversal potential is largely determined by the activity of the potassium chloride cotransporters NKCC1 (which imports chloride) and KCC2 (which exports chloride) in postsynaptic neurons (23). The developmental shift in GABAA current polarity is driven by the increased expression of KCC2 (24, 25). Intriguingly, it was recently found that depolarizing GABAA currents are also a feature of some pathological states of the adult brain (2629), and decreased KCC2 expression was reported in postmortem samples from the hippocampus of some patients with SCZ (30). On the basis of these findings and prior studies of the scPCP model of CIAS (12,13), we decided to test whether depolarizing GABAA current might also contribute to the pathological mechanism in scPCP mice. To this end, we performed perforated patch clamp recordings (31) in acute mPFC slices. These measurements showed that in the infralimbic cortex (ILC; the more ventral division of the mPFC; fig. S1) pyramidal cells from scPCP mice, the reversal potential of the GABAA current is depolarizing (it was ~9 mV more positive compared with cells in slices from vehicle-treated controls; Fig. 1, A to C). The shift in the current reversal potential was present in pyramidal cells of both layer 5 (Fig. 1) and layer 2/3 (fig. S2, A to C) of the ILC, while no change was detectable in the prelimbic (PLC) division of the mPFC (Fig. 1, D to F, and fig. S2C). Recordings of spontaneous GABAA currents (fig. S3) performed using the conventional whole-cell patch-clamp configuration did not reveal any change in current frequency or amplitude in pyramidal neurons of scPCP mice (fig. S4, A to C). This suggests that the changes are largely limited to the current reversal potential without detectable modifications in presynaptic GABA release or in the density of postsynaptic GABAA receptors in mPFC pyramidal neurons of scPCP mice. Next, we used cell-attached recordings (which allow recordings without affecting intracellular chloride level and resting membrane potential) to assess the net effect of depolarizing GABA on ILC pyramidal neurons excitability. We reasoned that depolarizing GABA should facilitate the generation of epileptic-like discharges following sustained synaptic stimulation due to GABA-mediated depolarization and decreased feedback inhibition. In keeping with this prediction, we found that following a train of synaptic stimulations (20 Hz, 1 s), ILC pyramidal neurons in slices from scPCP mice generated twice as many action potentials as cells from vehicle-treated mice (Fig. 2, A and C). When the recordings were performed while blocking ionotropic glutamatergic transmission, sustained firing was virtually abolished in slices from control (vehicle-treated) mice, but cells from scPCP mice still generated numerous action potentials (Fig. 2, B, D, and E) that were blocked by picrotoxin (fig. S5), supporting the idea that firing was facilitated by depolarizing GABAA currents. Parallel sets of recordings revealed no significant differences in the properties of excitatory synaptic currents in either the ILC or the PLC of scPCP mice compared with vehicle-treated controls (fig. S4, D to F), in agreement with the idea that increased cortical excitability in scPCP mice is caused by depolarizing GABA and may be the basis for cortical hyperactivity in SCZ.

Fig. 1 GABAA current is depolarizing in ILC pyramidal neurons of scPCP-treated mice.

(A and B) Current/voltage (I/V) curves of the peak GABAA currents recorded in L5 pyramidal neurons in slices from vehicle (veh)– and scPCP-treated mice. The dotted lines represent polynomial fits to the data. Currents (insets) were recorded in perforated patch clamp in the presence of 3 mM kynurenic acid. (C) Average GABAA reversal potentials for vehicle- and scPCP-treated mice in ILC L5 cells. Two-tailed Mann-Whitney U test, P = 0.04 and u = 29. The box plots dots represent outliers and + represent means. Orange lines represent the average resting potentials (measured in whole-cell configuration in separate experiments). Numbers above the boxes show the sample size in each group. (D to F) Same as (A) to (C) for prelimbic (PLC) cells; P = 0.86 and u = 39.5.

Fig. 2 ILC pyramidal neurons of scPCP-treated mice display increased excitability.

Cell-attached recordings of layer 2/3 ILC pyramidal cells from a vehicle-treated (left trace) and an scPCP-treated mouse (right trace) without synaptic blockers (A) and in the presence of kynurenic acid (3 mM; B). Slices were electrically stimulated (0.15 to 0.35 mA, 20 Hz, 1 s; red lines) with an electrode in layer 1. Stimulation artifacts are truncated in the figure. (C and D) Scatter plots representing the number of spikes recorded in the 10 seconds following the end of the stimulus without synaptic blockers (C; n = 12 for vehicle-treated mice, and n = 10 for scPCP-treated mice; P = 0.18 and u = 39 by Mann-Whitney U test), and in the presence of kynurenic acid (D; P = 0.07 and u = 58 by Mann-Whitney U test; n = 15 in vehicle versus n = 13 in scPCP). The lines in the plots represent means ± SEM. (E) In the presence of kynurenic acid, the fraction of neurons that kept firing after the end of the stimulus was significantly larger in the scPCP group compared with control (P = 0.029; Fisher’s exact test).

Because, as previously discussed, GABAA current polarity is largely determined by the activity of NKCC1 and KCC2 in postsynaptic cells, we quantified the amount of NKCC1 and KCC2 transcripts in mPFC slices using the “RNAscope” technique, a quantitative in situ hybridization protocol that allows single-cell resolution (32). In keeping with the electrophysiological findings, NKCC1 expression was increased in both layer 5 (P = 0.0087) and layer 2/3 (P = 0.026) of the ILC, but not the PLC (P = 0.31 and 0.48, for layers 5 and 2/3, respectively), of scPCP mice (Fig. 3 and figs. S2D and S6), while no significant change was detected in KCC2 expression. Thus, GABAA function in ILC pyramidal cells of adult scPCP mice mimics that in early brain development. However, in contrast with the physiological developmental regulation that depends on KCC2 expression, in scPCP mice, only NKCC1 expression is altered, which is reminiscent of other pathological states of the adult brain that are similarly characterized by the presence of depolarizing GABAA current (2629). This is important because of the availability of drugs that can inhibit NKCC1 activity. Thus, we tested the effect of the NKCC1 inhibitor bumetanide in slices from scPCP mice and found that it completely normalized the GABAA current reversal potential (Fig. 4, A to C, and fig. S2, E and F). In addition, we found that an intraperitoneal injection of bumetanide 30 min before initiating behavioral testing markedly ameliorated scPCP mice performance on multiple mPFC-dependent cognitive tasks assessing declarative memory, working memory, and executive function (Fig. 4, D to F), while no behavioral effects were noted when bumetanide was given to control mice (fig. S7). A similar cognitive rescue in scPCP mice was observed with focal bumetanide injection into the ILC (Fig. 5A). In agreement with the electrophysiological and in situ hybridization findings that showed no evidence for depolarizing GABAA in the PLC, focal bumetanide injection into the PLC did not ameliorate cognitive performance (Fig. 5B). While bumetanide is a well-known NKCC1 antagonist, it is impossible to rule out that the behavioral effects are due to some unknown pharmacological effect independent of the NKCC1 antagonism. To address this possibility, we investigated whether selective knockdown of NKCC1 expression in the ILC of scPCP mice mimics the behavioral effects of bumetanide. To this end, we used an adeno-associated viral construct containing mouse NKCC1–shRNA (short hairpin RNA), which reduces NKCC1-expression by ~70% 3 weeks after cortical injection (fig. S8). Figure 6 shows that bilateral injection of this construct into the ILC (Fig. 6, A and B) replicated the bumetanide effects on both the novel object recognition (NOR) and the Y-maze task, while injection of a scrambled construct had no effect (Fig. 6, C and D). Thus, all these different approaches concur with the idea that depolarizing GABAA current in the ILC has a direct causal role in the cognitive impairment of scPCP mice. One last question concerned the potential sexual dimorphism of this mechanism, as all our data were obtained in male mice. To address this possibility, we repeated the most critical experiments in female mice. Figure S9 shows that the same pathogenic mechanism is present in female scPCP mice, as NKCC1 expression is selectively increased in the ILC and bumetanide has the same marked effect in rescuing cognitive performance.

Fig. 3 NKCC1 expression is selectively increased in the ILC of scPCP mice.

(A) In situ hybridization images for NKCC1 (Slc12a2) and KCC2 (Slc12a5) in sections of L5 ILC from vehicle- and scPCP-treated mice. Brown signal [3,3′-diaminobenzidine (DAB)] represents NKCC1 and KCC2 mRNAs; methyl green was used to stain cell bodies. (B) Data quantification. Top: NKCC1; P = 0.0087 and u = 2 for ILC and P = 0.31 and u = 11 for PLC (Mann-Whitney U test). Bottom: KCC2; P = 0.81 and u = 16 for ILC and P = 0.82 and u = 16 for PLC (Mann-Whitney U test). Data from six control and six scPCP mice.

Fig. 4 Bumetanide restores GABAA reversal potential and rescues cognitive performance of scPCP mice in multiple cognitive tasks.

(A) GABAA reversal potential measured in an ILC L5 pyramidal cell from a vehicle and (B) scPCP mouse in the presence of bath-applied bumetanide (BMN). (C) GABAA reversal potentials recorded in ILC slices from vehicle-treated (black box) mice and in slices from scPCP-treated mice recorded either in the absence (blue box) or in the presence (purple box) of bumetanide; P = 0.0057 for vehicle versus scPCP (+saline), P = 0.0058 for vehicle + bumetanide versus PCP (+saline), and P = 0.0032 for scPCP (+saline) versus bumetanide-treated scPCP; F = 6.54, one-way analysis of variance (ANOVA) followed by Bonferroni test. The number of neurons in each group is indicated above the boxes. Orange lines represent the average resting potentials. (D) Discrimination index calculated from NOR tests. Bumetanide (0.1 mg/kg, in saline) was administered by intraperitoneal injection 30 min before testing. P < 0.0001 for scPCP versus either vehicle- or bumetanide-treated scPCP group. Each group, n = 10 (F = 71.49). (E) Y-maze spontaneous alteration behavior test. P < 0.0001 for scPCP versus either vehicle- or bumetanide-treated scPCP group and n = 10 for all three groups (F = 35.78). (F) Operant reversal learning test. The plot shows correct responses in reversal test phase; P < 0.0001 for scPCP versus either vehicle- or bumetanide-treated scPCP group. n = 10 for all groups (F = 95.08). We observed no statistical difference between groups in the initial phase test (not shown).

Fig. 5 Focal administration of bumetanide to the ILC, but not the PLC, improves cognitive performance in scPCP mice.

Bumetanide (0.3 mg/ml; 1 μl) was acutely injected into either the ILC or the PLC of vehicle- and scPCP-treated mice. (A) Effects of bumetanide on cognitive performance (NOR and Y-maze alternation tests) when injected into the ILC; six mice in each group. For NOR test, P < 0.0001 for scPCP versus either vehicle- or bumetanide-treated scPCP group (one-way ANOVA followed by Bonferroni test; F = 73.07). For Y-maze test: P < 0.0001 for scPCP versus either vehicle- or bumetanide-treated scPCP group (F = 39.44). (B) Bumetanide injections into the PLC had no detectable effects on cognitive performance; six mice in each group. NOR: P < 0.0001 for vehicle versus either scPCP- or bumetanide-treated scPCP group (F = 31.66). Y-maze test: P < 0.0001 for vehicle versus either scPCP- or bumetanide-treated scPCP group (F = 28.72).

Fig. 6 NKCC1 knockdown in the ILC improves cognitive performance in scPCP mice.

Adeno-associated viral constructs containing either mouse Slc12a2-shRNA [short hairpin RNA; knockdown (KD)] or a scrambled RNA sequence (SCR) were injected into the ILC. (A) Microphotograph (2×) of the ILC of an injected mouse; green fluorescent protein (GFP) expression identifies the injection sites. (B) Map indicating the injection sites for every mouse treated. Black circles identify vehicle/KD mice, blue squares identify scPCP/SCR, and green circles indicate scPCP + KD. (C) Results of the NOR test. The sample size was five, six, and five mice, respectively. P = 0.0293 for vehicle/KD versus scPCP/SCR and P = 0.0039 for scPCP/KD versus scPCP/SCR, one-way ANOVA followed by Bonferroni test (F = 8.16). (D) Y-maze spontaneous alternation test. Same mice as in (C). P = 0.0001 for vehicle/KD versus scPCP/SCR and P < 0.0001 for scPCP/KD versus scPCP/SCR (F = 28.14).


The results of the present study identify depolarizing GABAA current in the ILC as a novel mechanism mediating cognitive deficits in scPCP mice of both sexes. The finding that GABAergic dysfunction plays a central role in mediating cognitive impairment in scPCP mice is consistent with previous findings showing that scPCP disrupts NOR, working memory, and reverse learning in rodents and that these deficits are rescued by acute treatment with other drugs affecting GABAergic transmission (33). The most likely explanation for the marked cognitive effect of cortical depolarizing GABA is provided by the central role that GABAergic inhibition has in brain oscillations, which require coordinated activity across different brain areas (34). GABAergic inhibition is essential to the generation of oscillatory networks (35), and oscillations may even emerge naturally from reciprocal interactions between populations of excitatory and inhibitory neurons (36). Brain oscillations, in turn, organize neuronal activity in space and time (37), leading to the dynamic assembly of functional networks that is critical for cognition. Accordingly, aberrant oscillations (in particular in the gamma frequency) have been observed in SCZ (38) and related disorders such as Fragile X syndrome, a common form of autism spectrum (39). Impaired oscillations had been suggested to depend on a loss of parvalbumin-positive interneurons (10), which has often, but not always (40), been reported in cortex (41) or hippocampus (42) following subchronic treatment with PCP or other NMDAR noncompetitive antagonists.

If similar differences in the role of GABAergic signaling also exist between the different domains of pathology that along with CIAS comprise the SCZ phenotype, then bumetanide may be effective for only some types of behavioral abnormalities, in a subset of patients for whom depolarizing GABA represents the most important pathogenic mechanism. Because SCZ is a heterogeneous disorder that affects ~1% of humanity, it is possible that in other cases, pathogenic mechanisms unrelated to the dysregulation of the chloride transporters will prevent the correction of the functional deficits. Further, the finding that NKCC1 expression was increased in the ILC, but not in the PLC (nor the hippocampus; see fig. S10), of scPCP mice highlights the importance of the ILC for CIAS. However, functional impairments in other brain regions involved in various cognitive functions, such as the perirhinal, retrosplenial, and anterior cingulate cortices as well as the dorsal striatum, may also be involved and will require future studies. Different functional changes are likely present in other cortical areas in the scPCP model, as activation of the PLC and anterior cingulate cortex is altered in scPCP mice (43). The ILC specificity of the observed changes in scPCP mice has also practical consequences when trying to reconcile mouse models and human data as it shows that biochemical deficits in CIAS or other cognitive impairments can be highly localized. This observation suggests that large-scale brain examinations may fail to identify possible abnormalities, stressing the need for single-cell resolution approaches to study the pathogenesis of cognitive impairment.

In conclusion, this study provides the first evidence that the cognitive impairments in scPCP model similar to those in CIAS are caused by depolarizing GABA in the ILC and can be rescued by inhibiting NKCC1 activity. Identical or similar mechanisms may be present in patients with SCZ, opening up many avenues of therapeutic intervention, including the use of bumetanide. Intriguingly, bumetanide also ameliorates cognitive performance in a rodent model of Down syndrome (26), and it may have beneficial effects for psychosis and hallucinations in SCZ (4446), supporting the idea that depolarizing GABAA current in the adult brain may represent a convergent mechanism and a shared pharmacological target across several pathological conditions.


Animal model

C57BL/6J mice (3 weeks old) were intraperitoneally injected with either saline (vehicle, n = 154; 142 male and 12 female mice) or PCP (10 mg/kg in saline; n = 181, 12 of which were female) two times a day for seven consecutive days and were then rested for another 7 days without injections. PCP was supplied by the National Institute of Drug Abuse (Bethesda, MD). Animals were group housed (five per cage) in a controlled environment held at 21° ± 2°C and 50 ± 15% relative humidity, with a 14:10-h light-dark cycle and food and water available ad libitum. All experiments were conducted during the light phase and in accordance with guidelines of the institutional animal care and use committee of the Northwestern University.

Electrophysiological recordings

Slice preparation. Mice (vehicle, n = 86; scPCP, n = 97) were deeply anesthetized with isoflurane and euthanized by decapitation 7 days after the last PCP injection (5 weeks old at the time of euthanasia). The brains were removed from the skull, and 300-μm-thick coronal brain slices of the mPFC (coordinates from bregma, +1.845 to +1.145 mm) were cut using a vibro-slicer (Leica VT-1200) in ice-cold artificial cerebrospinal fluid (ACSF) containing 125 mM NaCl, 25 mM NaHCO3, 2.5 mM KCl, 1.25 mM NaH2PO4, 25 mM glucose, 2 mM CaCl2, 1 mM MgCl2, and 3 mM kynurenic acid, saturated with 95% O2 and 5% CO2 to pH 7.4. Slices were then stored for ~20 min at 35°C and allowed to recover at room temperature (22° to 24°C) for at least 30 min in the same solution.

Whole-cell patch-clamp recording. All electrophysiological measurements were performed using an Axopatch 200B amplifier. Signals were filtered at 2 kHz and sampled at 5 kHz for voltage-clamp recordings and filtered at 10 kHz and sampled at 20 kHz for current-clamp recordings. Data were acquired using pClamp9 software (Axon Instruments) running on a personal computer. Data and statistics were analyzed using software written in Matlab (MathWorks). Recordings were performed at 31° to 32°C in the presence of 3 mM kynurenic acid to block fast glutamatergic synaptic currents. Neurons were visualized using an upright microscope (Scientifica) with oblique illumination, a 60× water-immersion objective (Olympus) and a digital camera (DVC). Pyramidal cells were visually identified according to their location, size, and shape. Pipettes were pulled from thick-walled borosilicate glass (1.5-mm outer diameter; Sutter Instruments) using a horizontal puller (Sutter Instruments, P-97) and were filled with a KCl-based internal solution consisting of 148 mM KCl, 6 mM NaCl, 2 mM MgATP, 0.2 mM Na3GTP, 0.1 mM EGTA, and 10 mM Hepes (pH 7.3 with KOH). Pipette resistances in the working solutions ranged from 4 to 6 megohms yielding series resistances of 20 to 30 megohms for whole-cell recordings. Resting membrane potential was measured in whole-cell configuration immediately after breaking into the membrane. Input resistance was calculated from the peak voltage responses to hyperpolarizing current injections (−200 to −40 pA, 40-pA steps). Spontaneous inhibitory postsynaptic currents (IPSCs) were recorded at −85 mV over a 3-min period. The total number of IPSCs in each recording was counted using the Mini Analysis Program 6 (Synaptosoft). For evaluation of IPSC amplitudes, the events identified using the Mini Analysis software were analyzed using custom-made code written in Matlab. Spontaneous excitatory postsynaptic current (EPSC) recordings were also 3 min long and were performed using a potassium-methylsulfate internal solution (146 mM K-methylsulfate, 8 mM NaCl, 2 mM MgATP, 0.2 mM Na3GTP, 0.1 mM EGTA, and 10 mM Hepes, pH 7.3 with KOH) at −65 mV and analyzed as reported for IPSCs. The external solution for EPSCs recordings contained 50 μM picrotoxin and no kynurenic acid.

Cell-attached recordings. Giga-seals (1.5 to 8.5 gigohms) were obtained using 6- to 8-megohm pipettes filled with modified ACSF (NaHCO3 was substituted with 10 mM Hepes and titrated to pH 7.4 with NaOH). We counted the number of spikes elicited by electrical stimulation (in layer 1; 0.15 to 0.35 mA, 20 Hz for 1 s) within a 10-s segment, in the absence and in the presence of kynurenic acid in the bath. Each neuron was only stimulated ≤2 times, and these recordings were limited to ≤5 min to prevent spontaneous break-in. If we observed any change in the leak current or if the cell became intrinsically active, then the neuron was discarded as these signs suggest partial breaking of the cellular membrane.

Perforated patch recording. Our protocol was adapted from Heigele et al. (47). Fresh gramicidin (Sigma-Aldrich) stock solution [20 mg/ml in dimethyl sulfoxide (DMSO)] was prepared daily. Stock solution (2.5 μl) was added to 1 ml of warm (35°C) KCl internal (final concentration, 50 μg/ml). The solution was then vortexed for 1 min, sonicated for 15 min, and, lastly, filtered using a 0.2-μm filter. The solution was used within 2 hours. Recording pipettes (4 to 6 megohms) were tip filled with gramicidin-free KCl internal solution and then backfilled with the gramicidin solution. When approaching cell, small positive pressure was applied until the cell was patched. Seal properties were monitored measuring the current responses to square voltage steps (−5 mV, 150 ms; 0.1 Hz). Eighty- to 150-megohm access resistances could be attained within 10 to 15 min. Lucifer yellow (0.1%) was included in the internal solution to visually confirm in real time the integrity of the membrane. If any lucifer yellow staining of the cell body was observed during or after completion of the recordings, then data were discarded. All perforated patch recordings were performed at 27° to 30°C in the presence of 3 mM kynurenic acid. GABAergic synaptic currents were elicited by extracellular electrical stimulation (0.2 to 1.0 mA, 0.2 ms) using a bipolar electrode positioned in layer 1 of the mPFC, 100 to 300 μm apart from the recorded cell. Current/voltage (I/V) curves were obtained measuring the current at different holding voltages (−100 to −20 mV, 20-mV steps). For each voltage, 5 to 10 traces were averaged. If unclamped action potentials were recorded during any sweep, then that sweep was excluded from the analysis. The current peaks were plotted versus the holding potentials, and the I/V curves were obtained by fitting the data points using second-degree polynomial functions (using Clampfit 10, Axon Instruments).

Drug applications in slice recordings

Working drug solutions were freshly prepared from stock solutions on the day of the experiments. Kynurenic acid stock solution (500 mM; Sigma-Aldrich) was prepared in 1 N NaOH and stored at 4°C; the drug was used at a final concentration of 3 mM. Picrotoxin stock solution (100 mM; Tocris) was prepared in DMSO and kept at −20°C. The final working solution was 50 μM. Bumetanide (Tocris) stock solution (10 mM in DMSO) was prepared fresh every day. For slice recording of bumetanide-treated mice, the mice received intraperitoneal injections (0.1 mg/kg in saline) 30 to 40 min before being euthanized, and bumetanide (10 μM) was also added to the ACSF in the slice incubation and recording chambers. Simply adding bumetanide to the recording solution did not produce notable effects on the GABAA reversal potential when recording slices from scPCP mice, most likely because bumetanide requires ≥20 min to significantly affect eGABA (48), and in our experience, this time span is close to the maximum safe duration of perforated patch recordings.

Data quantification and statistical analyses

Whisker plots show median (lines inside boxes), 25th and 75th percentile (box margins), 10th to 90th percentile (whiskers), and outliers (points); the crosses inside the boxes represent the means. Both P and U values calculated using two-tailed Mann-Whitney U test are reported when comparing two independent groups. For cognitive tests to assess the effects of bumetanide or NKCC1 knockdown, one-way analysis of variance (ANOVA) was followed by a post hoc Bonferroni test. The sample size of every experimental group is provided in the figures or the figure legends. Statistical calculations were performed using Prism 8 software (GraphPad Software).

In situ hybridization assay

Section preparation. Mice (n = 27, 14 vehicle and 13 scPCP) were deeply anesthetized with Euthasol and intracardially perfused with 4% freshly depolymerized paraformaldehyde in 0.12 M phosphate buffer. Brains were postfixed in the same solution overnight and cryoprotected in 30% sucrose in phosphate-buffered saline with 0.1% diethyl pyrocarbonate. Coronal sections (14 μm) were cut using a freezing stage microtome and mounted on positively charged glass slides. We cut a total of 48 coronal sections from each brain along the anterior-posterior axis as, according to the Allen Brain Atlas stereotaxic coordinates, the ILC and PLC regions of the mPFC lay between +1.845 and +1.145 mm from bregma. Three sections spaced 336 μm apart from each other and corresponding to the anterior, middle, and posterior parts of the mPFC were sampled from each animal. Sections of both the left and right hemispheres were used for hybridization and for counting as no difference between the hemispheres was detected.

Probe hybridization. Hybridization was performed using a single probe assay kit (RNAscope 2.5 High-Definition BROWN Assay kit) and probes from Advanced Cell Diagnostics (Newark, CA) following the manufacturer’s user manual with minor modifications. Briefly, sections were treated with 0.9% H2O2 and 10% methanol in phosphate-buffered saline for 10 min in 10 mM sodium citrate buffer solution (pH 6.0) for 2 min at 91°C. After incubation in proteinase K (1 μg/ml; Sigma-Aldrich, REF03115887001) for 15 min at 40°C, probes were hybridized to sections for 2 hours at 40°C. Signals were amplified using the single probe assay kit reagents. Before the first amplification step, sections were washed in 0.1× saline-sodium citrate (SSC) buffer at 33°C. Sections were also washed in the same solution after each of the following six amplification steps: the first three times at 33°C, and then at room temperature. Probes for NKCC1 (Mm-SLC12A2) and KCC2 (Mm-SLC12A5) as well as positive (POLR2A) and negative (dapB) probes were tested on different (consecutive) sections. Signals were detected using DAB (3,3′-diaminobenzidine; developed for ~6 min), and cell bodies were counterstained using 25% methyl green (developed for 10 to 30 s).

Data quantification. NKCC1 and KCC2 expression was quantified in four regions of interest (layer 2/3 and layer 5 in ILC and PLC) by counting the number of DAB-stained mRNA puncta in each. The signal density was estimated using Stereo Investigator (MicroBrightField Bioscience) in three coronal mPFC sections from each animal sampled at each of the three levels (anterior, middle, and posterior) described above. Counting was performed on monochrome images acquired using an Optronic camera mounted on a Zeiss upright microscope and using a 63× oil immersion objective lens.

For counting, a sampling grid with 100-μm (for NKCC1) or 80-μm (for KCC2) interline distances was randomly superimposed over the mPFC section image, and we counted the number of puncta within the counting frames (40 × 40 μm for NKCC1 and 20 × 20 for KCC2). The frame size was optimized following preliminary analysis to fulfill the following criteria: (i) approximately less than 60 puncta per division and (ii) more than eight divisions per region. Last, we calculated the signal density in each region by dividing the total number of puncta in all counting frames by the total area within the counting frames. The results obtained in the three sections (at each level) counted from each animal were pooled and regarded as an individual sample point.

Behavioral assessments

A total of 125 (54 vehicle and 71 scPCP-treated) mice were assessed using cognitive tests.

NOR test. This protocol was adapted from Hashimoto et al. (49) and slightly modified as described previously (12). In particular, we used white background to the walls of the box instead of black reflective surfaces, as previous studies established that when black reflective surfaces were used for the inner surfaces of the NOR box, the animals failed to explore the objects (12). The same was true when large objects were used. Hence, white walls and small objects for exploration were used throughout. The NOR apparatus consisted of an open box made of Plexiglas (52 cm by 52 cm by 31 cm) with white walls and a solid floor. The box was positioned approximately 30 cm above the floor centered on a table such that the overhead lights could not provide a spatial cue.

Following the 7-day washout from subchronic drug treatment, mice were habituated to the empty NOR arena for 1 hour/day for each of the three days before the acquisition trial. During the acquisition trial, the animals were allowed to explore two identical objects for 10 min. This was followed by a 24-hour intertrial interval when the animals were returned to the home cage. During the retention trial, the animals were allowed to explore a familiar object from the acquisition trial and a novel object. The location of the novel object in the retention trial was randomly assigned for each test mouse using a pseudorandom schedule to reduce the effects of object and place preference. In addition, we made sure to avoid any form of bias or olfactory trails. Behavior was recorded on video for blind scoring of object exploration. Object exploration was defined as an animal licking, sniffing, or touching the object with the forepaws while sniffing. If the animal failed to explore one or both objects for more than 5 s in the acquisition or retention trial, then it was excluded from the final data analysis. We did not exclude any data in the studies presented in this article, because all the animals explored >5 s. The exploration time (in seconds) of each object in each trial was recorded manually. The discrimination index [(time spent exploring the novel object − time spent exploring the familiar object)/total exploration time] was then calculated for retention trials. When the animals were retested, new objects were used. Furthermore, mice showed stable exploration throughout the study period.

Y-maze spontaneous alteration behavior test. Testing was done in a Y-shaped maze with three black, plastic arms at a 120° angle from each other. After introduction to the center of the maze, the animal was allowed to freely explore the three arms. Over the course of multiple arm entries, mice show a tendency to enter a less recently visited arm. The number of arm entries and the number of triads (any sequence consisting of entries into three different arms) were recorded to calculate the percentage of alternation. An entry is defined when all four limbs are within the arm (50). Between sessions, the maze was cleaned with a 70% ethanol solution. As a measure for spatial egocentric memory, the percentage of alternations that the mouse made was calculated, being the number of triads divided by the maximum possible alternations (i.e., the total number of entries minus 2 × 100). If a mouse scored significantly above 50% alternations (the chance level for choosing the unfamiliar arm), then this was indicative of a functional spatial egocentric memory.

Operant reversal learning task. The experiments were conducted in eight operant chambers (exterior, 17.8 cm by 15.2 cm by 18.4 cm; interior, 15.9 cm by 14.0 cm by 12.7 cm; base, 40.6 cm by 29.2 cm by 1.3 cm; Med Associates, St. Albans, VT, USA) placed in sound- and light-attenuating enclosures. Each animal was trained and tested in the same operant chamber throughout the study. The stimulus lights, house lights, and reinforcements were controlled, and responses were recorded by Med-PC software (Version 2.0 for Windows, Med Associates, VT, USA). Programs controlling the schedule of reinforcement were written using Med-State notation. In training phase, mice were first trained to press both left and right levers to obtain sugar pellets as food reward. When they reached 90% baseline learning, they were trained to respond for food on a fixed ratio 10 (FR10) schedule of reinforcement, i.e., there was a reward only if they pressed the levers 10 times in a row. After stabilization of responding to a FR10, an animal would receive reward only if they press the lever 10 times in a row and only if they press the lever that has cue light on it. Then, the reverse training was started, i.e., they would get their reward only if they pressed the lever that did not contain cue light, 10 times. The testing phase consisted of the same protocol described in the training phase. The house and stimulus lights were illuminated signaling the beginning of a 10-min period (the initial phase), during which the reinforcement contingency matched that of the previous day’s training session. After the initial phase, the house and stimulus lights were extinguished, and a 10-min time-out began. After the 10-min time-out elapsed, the house light was illuminated, and the next phase of the test session began (the reversal phase), during which the reinforcement contingency was reversed. The reversal phase consisted of trials across a 10-min period. The test session was terminated after the end of the 10-min reversal. The primary dependent measure of reversal learning was the percent correct responses. Data for percent correct responses were calculated using the number of lever presses on the correct lever divided by the total number of presses multiplied by 100.

Bumetanide administrations for behavioral tests

Intraperitoneal injection. Acute intraperitoneal injections of bumetanide (0.1 mg/kg in saline) were performed 30 min before the behavioral tests [the half-life of bumetanide in mice is ~30 min (51)]. The pharmacokinetic parameters of bumetanide in rodents have been extensively studied. In mice, they were plasma binding of 98.9%, brain:plasma distribution of 0.015, and volume distribution of 0.2 liter/kg (51). On the basis of these parameters, the dose we used (0.1 mg/kg, intraperitoneally) yields brain concentrations between 5 and 7 μM. Linear extrapolation from data obtained in rat brain 30 min after intraperitoneal injections of 0.5 mg/kg (52) yields concentrations about one order of magnitude lower (~250 μM). Both these concentrations are within the range in which bumetanide is an effective and relatively specific blocker of NKCC1, as the IC50 of bumetanide for NKCC1 is in the range between 200 and 330 nM (53, 54). The concentration we used is also the same as that previously used in mice to evaluate cognitive effects attributable to NKCC1 block (26).

Intracortical injections. Before surgery, all instruments were heat sterilized, and the area around the surgical site was disinfected with betadine and 70% alcohol. Mice were given buprenorphine (0.05 mg/kg) and then anesthetized with 1 to 3% isoflurane (0.8 to 1.5 liters/min of flow rate using a vaporizer) and placed in a stereotaxic apparatus (Kopf Instruments, Tujunga, CA, USA). A guide cannula (PlasticsOne, Roanoke, VA, USA), with the dummy as an obturator was surgically implanted either in the ILC [anterioposterior (AP), +1.7 mm; mediolateral (ML), ±0.3 mm; dorsoventral (DV), −2.8 mm; relative to bregma] or PLC (AP, +1.7 mm; ML, ±0.3 mm; DV, −2.2 mm). Two or 3 days after cannulation, mice were again anesthetized with 1 to 3% isoflurane (0.8 to 1.5 liters/min of flow rate), and then an injector was inserted through the guide cannula. The injector was removed after completing the bumetanide injection (0.3 mg/ml; total volume, 1 μl) and then the dummy was inserted again into the cannula. Normally, the mice were fully awake in 3 to 5 min after the injections. Injected mice were used for behavioral studies starting 30 min following infusion procedure.

Intracortical viral infusion for NKCC1 knockdown

Viral delivery. Mice were anesthetized with isoflurane inhalation and secured in a stereotaxic frame as described above (intracortical bumetanide injection). Injections were performed using a 35-gauge blunt-end Nanofil syringe (1 μl). Adeno-associated virus constructs containing either mouse Slc12a2-shRNA [AAV9-GFP-U6-mSlc12a2-shRNA; titer of stock, 1.6 × 1013 genome copies (GC)/ml] or scrambled-shRNA (AAV9-GFP-U6-scrmb-shRNA; titer of stock, 3.1 × 1013 GC/ml) were produced by Vector Biolab (Malvern, PA, USA) and kept at −80°C until used. Control experiments to assess the efficacy of the viral knockdown were performed in three naïve mice by selectively injecting the construct into the mPFC of one hemisphere and using the contralateral cortex as control to be analyzed within the same slices. Expression of NKCC1 transcript in ILC (layers 2 and 5 combined) was then quantified 3 weeks after stereotaxic injection. This analysis showed that at this time point, NKCC1 mRNA in the injected site was decreased by 71.1 ± 3.3% compared with the contralateral side (fig. S8).

We infused the virus (250 nl in each side) bilaterally into the ILC (AP, +1.7 mm; ML, ±0.3 mm; DV, −2.8 mm; relative to bregma) at a rate of 0.15 μl/min; the final amount of virus delivered to each hemisphere was 4.0 × 109 GC of Slc12a2-shRNA and 7.8 × 109 GC of the scrambled construct. At the end of the infusions, we rested the needle in place for 5 min before retracting it. The skin incision was closed with veterinary tissue adhesive, and analgesics were administered after operation: single subcutaneous injection of buprenorphine sustained release (SR) (1 mg/kg) and daily subcutaneous injection of meloxicam (1 mg/kg) for up to 3 days. Behavioral tests were performed starting on day 21 after the virus infusion.

Anatomical verification of injection sites. Mice were euthanized once behavior tests were completed, and the injection sites were identified by assessing virus-driven green fluorescent protein (GFP) expression. Brains were dissected and immersed in 4% paraformaldehyde (at 4°C, overnight) and then cryoprotected in 30% sucrose in phosphate-buffered saline for 2 days. Coronal sections (50 μm) were prepared on a freezing stage microtome. Floating sections were incubated in blocking solution (3% normal goat serum, 1% bovine serum albumin, 0.2% Triton X-100 in tris-buffered saline) for 1 hour and then moved to 1% NGS solution (1% normal goat serum, 1% bovine serum albumin, and 0.2% Triton X-100 in tris-buffered saline) containing chicken anti-GFP primary antibody (1:1000, Abcam) at 4°C, overnight. After washing four times with tris-buffered saline, the sections were incubated with secondary antibody (Alexa Fluor 488, 1:600, Invitrogen) for 1 hour at room temperature. The center of GFP expression in each brain was marked on a brain atlas as the injection site. One mouse (of 17) in which both injection sites were off target was excluded from the analysis of behavioral data.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We are grateful to G. Sekerková for optimization of the in situ hybridization protocols, to M. Huang for implanting cannulas for intracortical drug infusions, to J. Do for virus injections, and to P. Penzes and G. Maccaferri for critical reading of the manuscript. Funding: This work was supported by NIH grants DA044121 and NS112292d (to M.M.), MH109466 (to H.Y.M.), and by a donation from the Weisman family (to H.Y.M.). H.Y.M. also discloses grant support from Janssen Pharmaceuticals, Allergan, Eli Lilly, Lundbeck, Neurocrine, Takeda, Sumitomo Dainippon, and Sunovion. Author contributions: H.R.K. performed the electrophysiological recordings and analysis, the in situ hybridization, designed the knockdown experiments, and provided critical readings of the manuscript. L.R. performed the PCP/vehicle injections and the behavioral tests and data analysis. H.Y.M. developed the project, designed the behavioral experiments, and cowrote the manusrcipt. M.M. developed the project, designed the patch clamp, in situ hybridization and the knockdown experiments, and wrote the manuscript. Competing interests: M.M. and H.Y.M. are inventors on a patent application related to this work filed by INVO, METHODS FOR TREATING PSYCHIATRIC DISEASES AND DISORDERS AND THE SYMPTOMS THEREOF IN A SUBJECT BY ADMINISTERING AN ANTAGONIST OF THE NA+-K+-2CL-CATION-CHLORIDE COTRANSPORTER ISOFORM 1 (NKCC1), with filing number 16/915,746. The other authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article