Disturbed development of the parvalbumin-positive fast-spiking (FS) interneurons in the prefrontal cortex (PFC) is closely associated with many neuropsychiatric disorders such as schizophrenia and autism. FS interneurons form at least 2 microcircuits in the PFC: one with pyramidal neurons (FS–PN) through chemical synapses; the other with other FS interneurons (FS–FS) via chemical and electrical synapses. It is currently unknown when and how these circuits are established in the PFC during early development. Here, we used G42 mice, in which FS interneurons are specifically labeled with enhanced green fluorescent protein, to make dual whole-cell recordings from postnatal day 3 (P3) to P30 to study the development of FS interneuronal networks in the PFC. We found that FS interneurons were poorly developed in terms of the membrane and network properties during the first postnatal week, both of which exhibited an abrupt maturation during the second postnatal week. The development of FS interneuronal microcircuits lasted throughout early adulthood. Thus, our data suggest that FS interneurons might not be involved in generating cortical oscillatory activity and γ oscillations during the first postnatal week. Our data also indicate an independent development of electrical and chemical synapses among FS interneuronal networks during the early period.
The prefrontal cortex (PFC), which is responsible for higher cognitive functions such as working memory, is impaired in patients with neuropsychiatric disorders such as schizophrenia and autism (Zikopoulos and Barbas 2010; Deserno et al. 2012). The disturbed maturation of parvalbumin (PV)-positive fast-spiking (FS) interneurons is closely associated with these diseases (Lewis et al. 2012). FS interneurons form at least 2 microcircuits in adult PFC. First, they are extensively connected with each other via chemical and electrical synapses (Fukuda and Kosaka 2000; Galarreta and Hestrin 2002). Second, while innervating the soma and proximal dendrites of pyramidal neurons (PNs) (Kawaguchi and Kubota 1997), FS interneurons also receive extensive glutamatergic synapses from PN cells (Angulo et al. 1999). In addition, FS interneurons also receive abundant synaptic inputs from thalamocortical relay neurons (Daw et al. 2007), and project to and receive inputs from somatostatin-expressing interneurons (Beierlein et al. 2000). Despite the putative relevance of FS interneurons for neurodevelopmental disease processes in humans, to the best of our knowledge, little effort has been exerted to study when and how FS interneurons develop their networks in the PFC.
In rodents, FS interneurons mainly arise from the medial ganglionic eminence as early as embryonic day 13.5 (Butt et al. 2005). After they are born, FS interneurons undergo sequential tangential and radial migration to populate specific cortical layers in appropriate brain areas (Anderson et al. 2001; Wonders and Anderson 2006). After a protracted period of postnatal development lasting well into early adulthood, FS interneurons acquire mature characteristics, both morphophysiologically and neurochemically, and become functionally incorporated into the cortical network (Du et al. 1996; Chattopadhyaya et al. 2004; Daw et al. 2007; Huang et al. 2007; Doischer et al. 2008; Okaty et al. 2009; Wang and Gao 2010; Goldberg et al. 2011; Pangratz-Fuehrer and Hestrin 2011).
In this study, by taking advantage of G42 transgenic mice, in which the most enhanced green fluorescent protein (EGFP)-expressing neurons are FS interneurons (Chattopadhyaya et al. 2004), we made dual whole-cell recordings from postnatal day 3 (P3) to P30 to study the development of 2 FS interneuron microcircuits in the PFC (FS–FS and FS–PN). Our data showed that most features of the FS interneuronal phenotype were poorly developed during the first postnatal week. An abrupt maturation of the membrane properties emerged during the second postnatal week. Neither FS–FS nor FS–PN microcircuit was established until the second postnatal week. The development of these networks continued throughout adolescence. Our data indicated an independent development of gap junctions and GABAergic synapses during the early period. We also found that interneurons tend to innervate PN cells preferentially before targeting other interneurons.
Materials and Methods
Preparation of In Vitro Brain Slices
All animal use procedures were reviewed and approved by the Animal Advisory Committee at Zhejiang University following the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals. G42 mice of both sexes ranging from P3 to P30 were deeply anesthetized and decapitated (P0 = day of birth). No sexually dimorphic observations were made. The brain was quickly removed and placed in ice-cold artificial cerebrospinal fluid (ACSF) consisting of (in mM): 125 NaCl, 2.5 KCl, 11 d-glucose, 26 NaHCO3, 1.25 NaH2PO4, 2 CaCl2, and 2 MgCl2. The solution was bubbled with 95% O2/5% CO2 to maintain a pH around 7.4. Coronal slices (300 μm) containing PFC (1–2.5 mm anterior to bregma) were cut using a microtome (Leica VT1000S, Germany). The slices were stored for 30–45 min at 36°C in oxygenated ACSF and then kept at room temperature.
Whole-Cell Paired Recordings and Data Analysis
Fluorescent cells were visually identified under an upright microscope (Nikon, ECLIPSE FN1) equipped with a 40× water-immersion lens and infrared differential interference contrast optics and illuminated with a mercury lamp. Individual slices were transferred to a recording chamber and fully submerged in continuously perfused (2–3 mL/min) oxygenated ACSF and maintained at 32 ± 2°C. We used intracellular solution containing (in mM): 110 K-gluconate, 40 KCl, 10 HEPES, 3 Mg-ATP, 0.5 Na2-GTP, and 0.2 EGTA. pH was adjusted to 7.25 with 10 M KOH. We simultaneously recorded from 2 EGFP-positive cells (<100 µm apart) or 1 EGFP-expressing cell and 1 PN neuron (<20 µm apart) from layers III to VI (P3–8) or V/VI (P9–30) in the PFC (Table 1). dl-2-Amino-5-phosphonopentanoic acid (AP-V) (50 µM, Tocris Bioscience) and 6,7-dinitroquinoxaline-2,3(1H, 4H)-dione (DNQX) (20 µM, Tocris Bioscience) were present in the bath solution when studying synaptic connections between FS interneurons. In a subset of experiments, we applied the gap junction blocker carbenoxolone (CBX, 100–200 µM, Tocris Bioscience) in the bath solution for at least 20 min before assessing the properties (i.e., amplitude, failure rate, and paired-pulse ratio [PPR]) of chemical synapses among developing FS interneurons (Fig. 5C–E). The coupling probability of chemical connections between FS interneurons with CBX was not significantly different from that without CBX (Supplementary Fig. S1), thus these data were pooled (Figs 5B and 7 R-IPSCs). Tight seals (>2 GΩ before breaking into whole-cell mode) were obtained using patch electrodes that had an open-tip resistance of 3–8 MΩ. Series resistance (<20 MΩ) was not compensated and continually monitored throughout each experiment. Recordings were terminated whenever series resistance increased >30%. To detect chemical synaptic transmission, the presynaptic neuron was held in the current-clamp mode near resting membrane potential (RMP) and stimulated at a frequency of 0.5 Hz. Action potentials (AP) were generated by brief current pulses (duration 2 ms, amplitude 20–2000 pA). The postsynaptic cell was held in the voltage-clamp mode (holding potential, −70 mV). GABAergic/glutamatergic postsynaptic currents (I/EPSCs) under these conditions are inward. For detecting electrical connections, we injected prolonged pulses (500–600 ms) of hyperpolarizing current (–10 to –500 pA) into each cell in an alternating manner and measured the steady-state voltage deflections of both the injected cell (ΔV1, 20–40 mV) and its coupled neighbor (ΔV2). A coupling coefficient (CC) was then calculated as ΔV2/ΔV1. Cell pairs were considered to be electrically coupled if CC ≥0.01. To measure electrical coupling conductance (GJ), both cells were initially held at −70 mV under voltage-clamp conditions, each cell was hyperpolarized to −120 mV for 300–400 ms in turn, and GJ was estimated by dividing the gap junction-mediated currents generated in each cell by 50 mV. To isolate miniature EPSCs (mEPSCs), tetrodotoxin (1 µM) and picrotoxin (100 µM) were present in the bath solution. To isolate miniature IPSCs (mIPSCs), tetrodotoxin (1 µM), AP-V (50 µM, Tocris Bioscience), and DNQX (20 µM, Tocris Bioscience) were present in the bath solution. Recording from putative FS interneurons was verified after each experiment by the backfill of GFP in the tip of the recording pipette (P3–12) or their firing properties (P13–30), that is, FS firing in a nonadaptive manner. Signals were acquired using an Axon Instruments MultiClamp 700B amplifier (Molecular Devices, Sunnyvale, CA) controlled by Clampex 10.2 software via a Digidata 1440A interface (Molecular Devices). Responses were filtered at 2 kHz, digitized at 10 kHz and analyzed using Clampfit 10.2 (Molecular Devices) and Mini Analysis 6.0 software (Synaptosoft, Decatur, GA). The root-mean-square noise level was 2–3 pA and a threshold of 10–12 pA was used to detect and measure mI/EPSCs.
|FS–FS||50 ± 4||51 ± 3||55 ± 7||54 ± 3||51 ± 4||56 ± 5||54 ± 5||58 ± 8|
|FS–PN||11 ± 0.8||12 ± 0.9||11 ± 0.8||10 ± 1.2||10 ± 1.1||11 ± 0.8||13 ± 1.0||11 ± 1.2|
|FS–FS||50 ± 4||51 ± 3||55 ± 7||54 ± 3||51 ± 4||56 ± 5||54 ± 5||58 ± 8|
|FS–PN||11 ± 0.8||12 ± 0.9||11 ± 0.8||10 ± 1.2||10 ± 1.1||11 ± 0.8||13 ± 1.0||11 ± 1.2|
For analysis of the passive and active membrane properties of FS interneurons, data gathered from recordings with DNQX and AP-V were used. The RMP was the average membrane potential within 2 min after obtaining the whole-cell configuration. Input resistance (Rin) was calculated from small (2–6 mV) voltage deflections induced by rectangular hyperpolarizing current injections (1–50 pA). Membrane time constant (τm) was obtained by fitting a single exponential function to these same hyperpolarizing voltage deflections. Membrane capacitance (Cm) was calculated by dividing τm by Rin. We used the first spike evoked by a small suprathreshold current step applied from RMP to quantify spike properties. AP amplitude was calculated as the voltage difference between AP threshold and AP peak. AP duration was measured as the duration at half-maximal amplitude. Afterhyperpolarization (AHP) amplitude was defined as the voltage difference between AP threshold and the depth of the AHP. Spike adaptation was referred to as the ratio of the first to the 10th interspike interval (ISI) (ISI1/ISI10) during a sustained train of APs at 30–50 Hz. The maximal firing rate was defined as the inverse of ISI1 during 600-ms depolarizing current injections prior to AP failure (i.e., AP peak amplitude < 0 mV).
Layers III/VI EGFP-expressing cells lacking prominent apical dendrites were selected for dye injection. Dye coupling was done at 32 ± 2°C. The pipette solution contained 0.6% neurobiotin (Vector Laboratories, Burlingame, CA) (pH 7.25, 3–8 MΩ electrode resistance). After we obtained whole-cell access, the neuron (one per slice) was held at −60 to −70 mV under current clamp and was injected with the tracer by passing subthreshold depolarizing rectangular pulses of 100 ms duration at 4 Hz for 10 min. Then the neuron was held at −70 mV under voltage clamp for 20 min. After the removal of the recording pipette, the slice was incubated in the recording chamber for an additional 30–45 min to allow dye diffusion. Incubation for longer periods did not seem to improve dye coupling. The slice was then fixed in 4% paraformaldehyde in phosphate buffer (pH = 7.4) at 4°C for 4–7 days. After washing, slices were incubated in avidin–biotin horseradish peroxidase complex containing 0.5% Triton X-100 (ABC kit; Vector Laboratories) for 2 h at 37°C. Visualization was completed by using a DAB method.
Mice were sacrificed for immunoblot analysis on P16 and P30. Protein samples were run on 10% SDS-polyacrylamide gels, and then transferred to polyvinylidene difluoride membranes. After blocking in 3% nonfat milk for 1 h at room temperature, membranes were incubated with the respective primary antibodies at 4°C overnight (rabbit anti-Cx36, 1:200, Invitrogen; rabbit anti-β-actin, 1:1000, Cell Signaling). After washing 3 times in 0.1% Tween20 in TBS, the membranes were incubated with horseradish peroxidase-conjugated secondary antibodies in 3% nonfat milk for 1 h at room temperature and visualized with an ECL kit (Thermo).
Unless otherwise specified, data are expressed as mean ± SEM and statistical analyses were made with one-way ANOVA and independent Student's t-test. Differences were considered significant if P < 0.05.
Intrinsic Properties of FS Interneurons
In GAD67-GFP (G42) transgenic mice, most EGFP-expressing neurons become PV-positive FS interneurons in adults and these neurons were observed as early as P0 in the PFC (Chattopadhyaya et al. 2004). A total of 302 FS–FS interneuron pairs and 198 FS–PN cell pairs were recorded from P3 to P30 in G42 mice (Table 2). Most of the passive and active membrane characteristics of FS interneurons changed dramatically during the first 2 postnatal weeks (Figs 1 and 2). We found that the mean RMP was more depolarized at P3–4 than in older mice (Fig. 1B). The most significant intrinsic change was more than the 12-fold decrease in Rin as animals matured (Fig. 1A,C). The change in Rin was paralleled by a nearly 4-fold decrease in τm (Fig. 1A,D) and a 4-fold increase in Cm (Fig. 1E). Interestingly, we found that Cm decreased sharply from 65.21 ± 2.67 pF at P15–16 to 46.33 ± 2.69 pF at P21–30 (P < 0.0001 vs. P15–16, t-test) (Fig. 1E). Several firing properties of FS interneurons also became more adult-like during postnatal development (Fig. 2). The AP amplitude increased significantly between P3–4 and P7–8 (Fig. 2A–C). AP duration was more than 4-fold longer at P3–4 than at P21–30 (Fig. 2B,D). We observed only a modest change in AHP amplitude with age (Fig. 2B,E). In addition, we found that most FS interneurons were able to fire APs at 30–50 Hz induced by rectangular supra-threshold current injections as early as P3–4, the earliest age studied (Fig. 2A,F); however, spike adaptation of FS interneurons became more adult-like by P15–16 (Fig. 2A,F). The maximal firing rate developed throughout adolescence (Fig. 2G,H, t-test).
|FS–FS # pairs||22||30||36||58 (24)a||53 (11)||39 (18)||32 (10)||32 (10)||302|
|FS–PN # pairs||25||24||22||22||23||22||21||39||198|
|FS–FS # pairs||22||30||36||58 (24)a||53 (11)||39 (18)||32 (10)||32 (10)||302|
|FS–PN # pairs||25||24||22||22||23||22||21||39||198|
aValues in parentheses are the numbers of cell pairs recorded with CBX for each age group.
Development of Electrical and Chemical Synapses Between FS Interneurons
FS interneurons are extensively interconnected by both electrical and chemical synapses (Galarreta and Hestrin 1999; Fukuda and Kosaka 2000). When 2 cells formed an electrical connection (Fig. 3A), injection of hyperpolarizing current into either cell induced downward voltage responses in both cells, a hallmark of electrical coupling. At P3–4, we did not detect electrical synapses among FS interneurons (0 of 22 pairs) (Fig. 3C). At P5–6, we found only 1 of 30 FS pairs to be electrically coupled (Fig. 3C). From that time, an age-dependent increase in coupling progressed until P13–14, after which the incidence of electrical coupling remained at a relatively high level (Fig. 3C). We found electrical coupling in 4 of 36 pairs at P7–8 (Fig. 3B,C), which more than doubled at P9–10 (9/34 pairs) (Fig. 3C). From P11 to P30, electrical coupling was expressed in more than half of the FS–FS pairs (Fig. 3B,C).
The amplitude of junctional conductance (GJ), which is frequently used to estimate the strength of gap junction coupling, also showed an age-dependent increase from P5–8 to P13–14 (P = 0.018 vs. P5–8, t-test) (Fig. 3D). Surprisingly, the amplitude of GJ decreased steeply from 468.28 ± 104.72 pS at P15–16 to 206.07 ± 32.37 pS at P21–30 (P = 0.032 vs. P15–16, t-test) (Fig. 3D). We observed a similar developmental change in CC (Fig. 3E, t-test), another means of measuring the strength of electrical coupling. The protein expression level of connexin36, which is the most prominent connexin underlying electrical synapses in adult neocortex (Venance et al. 2000), showed a significant decrease at P30 (Fig. 3F,G, t-test).
Dye coupling is widely used to demonstrate gap junction coupling (Connors et al. 1983; Peinado et al. 1993). We then injected FS interneurons with neurobiotin and counted the number of cells to which a single FS neuron can be coupled based on 2 criteria: with a round/oval shape and located up to 100 µm from the darkly stained injected cell (Amitai et al. 2002). In accordance with our results from paired recordings (Fig. 3B), the incidence of dye coupling increased during the first 2 postnatal weeks (Fig. 4B, t-test), suggesting a developmental increase in neuronal gap junction coupling. At P3–4, we found only 1 of 8 successful injections showed dye-coupled neurons (Fig. 4). At P7–8, each injected neuron dye-coupled to 1–13 cells (Fig. 4). The numbers of neurons dye-coupled to the injected cell ranged up to 23 for P15–16 (Fig. 4).
When 2 cells were chemically connected, suprathreshold stimuli applied to the presynaptic cell induced in the postsynaptic cell hyperpolarizing currents characteristic of unitary IPSCs (uIPSCs) (Fig. 5A), which were blocked by the GABAA receptor antagonist bicuculline (data not shown). At the earliest ages studied (P3–6), none of 52 cell pairs possessed chemical synapses, which became detectable only at P7–8 (2/36 pairs) (Fig. 5B). We found a nearly 3-fold increase in the incidence of chemical synapses from P9–10 (12/58 pairs) to P11–12 (32/53 pairs; P = 0.004 vs. P9–10, Fisher's Exact test) (Fig. 5B). From P13 to P30, GABAergic connections were expressed in more than half of the FS–FS pairs (Fig. 5B).
Most FS interneurons are interconnected by both electrical and GABAergic synapses (Galarreta and Hestrin 1999), and the low-pass filtering properties of gap junctions enable them to transmit signals at low frequency such as AHP more efficiently (Bennett and Zukin 2004); thus, the postsynaptic responses mediated by both synaptic types tend to overlap, leading to the distortion of pure GABAergic responses (Supplementary Fig. S2). Thus, to analyze the amplitude, failure rate, and PPR of uIPSCs, we used data gathered from recordings with CBX (100–200 µM) in the bath solution throughout experiments or from those expressing only chemical synapses. Data with or without CBX were not different, thus were pooled. We found an age-dependent increase in the amplitude of uIPSCs between P7–10 and P15–16 (P = 0.001, t-test) (Fig. 5C). This change was accompanied by a decrease in the failure rate (P7–10 vs. P15–16; P < 0.001, t-test) (Fig. 5D) and an increase in PPR (P7–10 vs. P15–16, P = 0.041, t-test) (Fig. 5E), suggesting a functional maturation of GABAergic synapses among developing FS interneurons.
FS interneurons preferentially synapse on the somata and proximal dendrites of their target neurons (Chattopadhyaya et al. 2004) and most mIPSCs originate from proximally located synapses (Soltesz et al. 1995), which thus may reflect their FS interneuronal origin. We then analyzed the developmental profile of mIPSCs on FS interneurons (Fig. 5F–H). At P3–4, 8 of 10 putative FS interneurons expressed mIPSCs. By P6, mIPSCs could be detected in all FS interneurons. Both the frequency and amplitude of mIPSCs significantly increased between P5–6 and P11–12 (Fig. 5G,H, t-test). A further increase in mIPSC frequency was observed at P21–22 (Fig. 5F,G, t-test).
In juvenile and adult rodents, FS interneurons are extensively coupled by electrical and chemical synapses; however, this does not seem to be the case before P10 (Fig. 6). We found that at P5–10, only a minority of electrically coupled FS–FS pairs coexpressed chemical synapses; by contrast, most FS–FS pairs were electrically and chemically connected at P11–12 (2/14 vs. 18/23 pairs; P < 0.0001; χ2 test) (Fig. 6A). The probability of encountering gap junctions among chemically connected pairs was lower at P5–10 than at P11–12 (2/7 vs. 18/25 pairs; P = 0.018; χ2 test) (Fig. 6B). At P5–8, the earliest age at which both synapses were detectable, of the 66 FS–FS pairs we found 5 to be electrically coupled (Fig. 3B) and 2 chemically connected (Fig. 5B). Thus, our data indicated an independent development of gap junctions and GABAergic synapses during the early period.
All together, we found 46 of 75 electrically coupled FS–FS pairs to be also chemically connected (Fig. 6A), and 46 of 70 chemically coupled pairs to be also electrically connected (Fig. 6B). Following the onset of electrical connections among FS interneurons at P5–6, the probability of identifying gap junctions (75/229 pairs at P5–30) (Fig. 3B) was not statistically different from that of finding GABAergic synapses (107/280 pairs at P5–30) (Fig. 5B) (P = 0.386, Fisher's Exact test).
The probability of encountering reciprocal chemical communications between FS interneurons is reported to be high (Gibson et al. 1999). All together, we encountered 41 cell pairs that formed reciprocal GABA-mediated connections (R-IPSCs) (Fig. 7). The development of R-IPSCs was significantly correlated with postnatal age (R = 0.990; P < 0.0001, n = 302 FS–FS pairs, linear regression) (Fig. 7). At P21–30, 11/32 FS–FS pairs formed R-IPSCs (Fig. 7). The probability of finding both gap junctions and R-IPSCs also increased with postnatal age (R = 0.959; P = 0.010; n = 229 FS–FS pairs, linear regression) (Fig. 7).
Synaptogenesis Between FS Interneurons and Pyramidal Neurons
In addition to forming a microcircuit via GABAergic and electrical synapses with neurons of the same type, FS interneurons also project to and receive inputs from PN neurons (Kawaguchi and Kubota 1997; Angulo et al. 1999). To characterize the development of the microcircuit between FS interneurons and PN cells, we first performed dual whole-cell recording from EGFP-expressing FS interneurons and neighboring GFP-negative PN cells in G42 mice. We found that in the PFC, FS interneurons formed functional GABAergic synapses on PN cells (FS → PN) as early as P3–4 (Fig. 8A,B), with a low probability (1/25 pairs) (Fig. 8B). We found a steep increase in coupling probability from P3–4 to P9–10 (Fig. 8B). This developmental change was accompanied by increases in uIPSC amplitude and PPR and a decrease in failure rate (Fig. 8C–E, t-test), suggesting a developmental maturation of the GABAergic synapses in FS → PN connections. At P11–16, GABAergic connections were detected in ∼25% of FS → PN pairs (Fig. 8B). By P21–30, we identified 16 of 39 pairs that established GABAergic synapses (Fig. 8B). Consistent with the notion that interneurons innervate PN cells before targeting other interneurons (Ben-Ari et al. 2007), our data showed that FS → PN connections appeared at least 2 days earlier (P3–4; Fig. 8B) than GABAergic FS → FS connections (P7–8; Fig. 5B), and reached a peak at P9–10 (Fig. 8B), a time when the FS → FS counterparts were poorly developed (Fig. 5B). At P3–4, 6 of 7 PN neurons expressed mIPSCs. By P6, we detected mIPSCs in all PN neurons. While the frequency of mIPSCs significantly increased between P3–4 and P21–22 (Supplementary Fig. S3A,B, t-test), a decrease in mIPSC amplitude was observed at P21–22 (Supplementary Fig. S3C, t-test).
As was found in FS → PN pairs, we detected glutamatergic synapses in PN → FS pairs at P3–4 (Fig. 9A,B). However, in contrast to FS → PN connections (Fig. 8B), the coupling probability of PN → FS pairs was relatively low between P3–4 and P9–10 (Fig. 9B). An abrupt increase occurred between P11–12 and P15–16 (Fig. 9B). The amplitude of unitary excitatory postsynaptic currents showed a nearly 4-fold increase from P3–10 to P15–16 (P = 0.033 vs. P3–10, t-test) (Fig. 9C), paralleled by a decrease in failure rate (P3–10 vs. P15–16, P < 0.0001, t-test) (Fig. 9D), again revealing the postnatal maturation process of excitatory PN → FS connections. By P21–30, 11 of 39 PN → FS pairs expressed excitatory synapses (Fig. 9B); however, the failure rate markedly increased (P = 0.027 vs. P15–16, t-test) (Fig. 9D), and PPR almost doubled between P3–10 and P21–30 (P < 0.0001, t-test) (Fig. 9E). We detected mEPSCs in 9/10 FS interneurons at P3–4. Afterward, all FS interneurons expressed mEPSCs. Analysis of the mEPSCs revealed a sharp increase in the frequency of mEPSCs between P5–6 and P11–12 (Fig. 9F,G, t-test), whereas the mEPSC amplitude decreased significantly from P15–16 to P21–22 (Fig. 9H, t-test).
The probability of encountering both glutamatergic and GABAergic synapses between FS interneurons and PN cells was significantly correlated with age (R = 0.917; P = 0.028, n = 198 FS–FS pairs, linear regression) (Fig. 10). All together, we found 13 mutually connected FS–PN pairs from P7 to P30 (Fig. 10), which was markedly less than the probability of reciprocal chemical FS–FS connections during the same developmental period (41/250 FS–FS pairs vs. 13/149 FS–PN pairs, P = 0.037, Fisher's Exact test) (Fig. 7A). Of the 198 PN-FS pairs, we found none to be electrically coupled.
In a subset of experiments, we used CBX to block gap junctions (Fig. 5C–E). However, cautions should be made that CBX may have multiple effects in addition to blocking gap junctions, and previous studies have reported an inhibition of GABAergic currents by CBX (Tovar et al. 2009; Beaumont and Maccaferri 2011). It could be inferred from these observations that the prevalence and developmental time-course of inhibition among FS interneurons might be underestimated with CBX. However, this does not seem to be the case under our experimental conditions; the coupling probability of chemical connections between FS interneurons with CBX was not significantly different from that without CBX (Supplementary Fig. S1). As the block of gap junctions also increases GABAergic currents via an increase in input resistance (Deans et al. 2001), we speculated that both the increase and decrease of IPSCs by CBX may partially counteract each other, leading to a nonapparent change in the properties of inhibition among FS interneurons.
GABAergic actions have long been hypothesized to be excitatory during early development and to play an important role in generating oscillatory activity, thus contributing to the proper construction of cortical networks (Ben-Ari 2002; Moody and Bosma 2005; Ben-Ari et al. 2007). We found both GABAergic and glutamatergic synapses between FS interneurons and PN cells as early as P3–4 (Figs 8B and 9B), and chemical synapses among FS interneurons became detectable at P7–8 in mouse PFC (Fig. 5B). Because of the low probability of connection and the small amplitude and high failure rate of synaptic responses between FS interneurons and PN cells and among FS interneurons (Figs 5B–E, 8B–E, and 9B–E), it is less likely that these newly formed synapses contribute to synchronizing network activity during the first postnatal week. Instead, it has been suggested that subplate-driven (Dupont et al. 2006; Yang et al. 2009) or septum-dependent (Conhaim et al. 2011) mechanisms are involved. In addition, a few “hub” neurons such as low-threshold spiking (LTS) interneurons may act as a “pacemaker” during early life as they pioneer cortical development (Long et al. 2005; Bonifazi et al. 2009; Cossart 2011). It would be interesting to know whether the relatively more mature FS interneurons are involved in generating oscillatory activity during the second postnatal week.
Because of their specific membrane and synaptic properties, PV-expressing FS interneurons play an essential role in the generation and maintenance of γ oscillations (for review, see Bartos et al. 2007). We found that although most FS interneurons were able to fire APs at low-frequency γ oscillation rates (30–50 Hz) at P3–4 (Fig. 2), and the presence of γ oscillations in this low-frequency range in newborn cortex (P0–6) has also been reported in vivo (Yang et al. 2009; Minlebaev et al. 2011), the poor development of membrane and synaptic properties of FS interneurons may prevent them from participating in the genesis of the early form of γ oscillations in the first postnatal week. It has been shown that the morphological, electrophysiological, and synaptic properties of basket cells, the putative cellular substrate of γ oscillations in the hippocampus, also undergo dramatic changes from P6 to P25 (Doischer et al. 2008). Computational simulations revealed that young basket cells contribute more to the generation of the lower frequency γ oscillations than their mature counterparts (Doischer et al. 2008). Future studies are required to determine when FS interneuron networks in the PFC are mature enough to take over the role in generating “mature” γ oscillations (Tamas et al. 2000; Bartos et al. 2007; Doischer et al. 2008; Cardin et al. 2009).
Unlike LTS interneurons, which reach a plateau at P8 in the expression of electrical synapses (Long et al. 2005), we found that electrical synapses among FS interneurons were poorly developed during the first postnatal week (Fig. 3). The lack of electrical connections at young ages could be caused by systematic differences in the intercellular distance. Although the intersomatic distances of the recorded pairs are quite similar at all ages (Table 1), this does not represent the same proportion of the network at P3 and at P30 given the considerable increase in the size of the brain and of the PFC in particular between these 2 time-points. Another likely possibility is that the dendritic spread of the immature cells is much less than mature cells (Fig. 4A), and the apparent lack of coupling in young neurons was caused by sampling the cells at distances too far apart to allow overlap, and gap junction contact, between immature dendrites. Although the dye method has been hypothesized to be a more sensitive way than electrophysiology for detecting very weak gap junctions, our data showed that putative FS interneurons were poorly dye-coupled during early development (Fig. 4), thus a more likely possibility of the lack of electrical connections at young ages could be the sparse expression of gap junctions among FS interneurons.
Several factors could lead to the change of the amplitudes of junctional conductance (GJ) and CC, such as the total number of gap junction channels, the single gap junction channel conductance, the location of gap junctions, and the membrane properties of the coupled cells (Galarreta and Hestrin 2001). The decrease in both GJ and CC at P21–30 (Fig. 3D,E) might be due to the difference in intersomatic distances (Table 1); another possibility is the decreased protein expression of connexin36 at P30 (Fig. 3F,G). A recent study from Connors' lab demonstrates a positive correlation between Cm and GJ from neurons in the thalamic reticular nucleus (Parker et al. 2009) (Fig. 5B, their article). In accord with these observations, we found that Cm of FS interneurons exhibited a significant decrease between P15–16 and P21–30 (Fig. 1E). The mechanism underlying the decrease in Cm requires future investigation.
Paired recording is effective for identifying pairs of neurons that are electrically coupled since electrical synapses require close proximity between 2 cells (Amitai et al. 2002). However, chemical coupled neurons can connect across greater distances, and it is possible that the detection of chemically coupled pairs is underestimated by a bias of selecting cells in close proximity. This seems to hold true in the case of PN → FS connections. Analysis of the development of mEPSCs in FS neurons showed that excitatory connections on FS interneurons were abundant at P11–12 (Fig. 9F,G), a time when paired recording revealed still very few glutamergic connections between PN cells and FS interneurons (Fig. 9B). However, paired recording may be more effective for identifying chemical synapses among FS interneurons, as analysis of mIPSCs showed that inhibitory synapses in FS interneurons were poorly developed at P11–12 (Fig. 5F,G), while paired-recording revealed an abrupt increase in the coupling probability among FS interneurons between P9–10 and P11–12 (Fig. 5B). Thus, it is very likely that paired-recording is more suitable for detecting connections mediated by locally projecting interneurons, whereas miniature analysis is more suitable for identifying transmissions mediated by long projection PN neurons.
Several previous studies show that during early postnatal development (<P14), FS interneurons are not only electrically coupled among themselves but also with PN cells (Venance et al. 2000; Bittman et al. 2002; Meyer et al. 2002). We found that none of the 138 FS–PN pairs at P3–14 had detectable electrical synapses. This could be due to potential bias of the G42 mouse line, as G42 mice express GFP only in a subset of FS cells (∼50% of mature PV neurons express GFP in G42 cortex) (Chattopadhyaya et al. 2004). Thus, it is possible that a subpopulation of FS interneurons has been unintentionally selected for study in this work, and that the development of synapses in other subtypes of FS cells could be different.
Taken together, we systematically studied the developing features of the PV-expressing FS interneuronal phenotype itself and 2 FS interneuron-mediated microcircuits in the PFC. Our data showed that both the membrane and network properties of FS interneurons were poorly developed during the first postnatal week, and an abrupt maturation was observed during the second postnatal week, with circuit development lasting throughout early adulthood. Our data suggest that the poor development of FS interneurons might not allow them to participate in the genesis of cortical oscillatory activity and γ oscillations during the first postnatal week. A better understanding of when and how the FS interneuron-driven γ oscillations are set up will shed light on both the cellular mechanisms resulting in the generation and maintenance of γ oscillations and the pathological mechanisms leading to such developmental neuropsychiatric disorders as schizophrenia and autism.
All authors contributed to the design of the experiments. J.M.Y. and J.Z. performed experiments. All authors contributed to the analysis and interpretation of data. J.M.Y. and J.Z. performed the statistical analyses. J.M.Y. wrote the manuscript. X.M.L. supervised the work.
This work was supported by grants from the National Natural Science Foundation of China (91132714, 30970916, 31070926 and 81221003), the Major Research Program from the State Ministry of Science and Technology of China (2010CB912004), the Zhejiang Provincial Natural Science Foundation of China (Z2090127), the Zhejiang Provincial Qianjiang Talent Plan (2010R10057), PCSIRT, the Fundamental Research Funds for the Central Universities (2011XZZX002) and the Zhejiang Province Key Technology Innovation Team (2010R50049).
We thank Dr Josh Huang (CSH, United States of America) and Dr X.H. Zhang (ION, China) for kindly providing the G42 mice; Dr T.M. Gao and Dr L. Mei for insightful suggestions; Dr I.C. Bruce for critical reading of this manuscript; and the anonymous reviewers for critical and insightful suggestions. Conflict of Interest: None declared.