Abrupt Maturation of a Spike-Synchronizing Mechanism in Neocortex

Introduction

Synchronous neuronal activity in the cerebral cortex has been implicated in sensory perception, motor control, attention, and plasticity (Ritz and Sejnowski, 1997; Fetz et al., 2000; Steinmetz et al., 2000; Engel et al., 2001; Levy et al., 2001; Sanes, 2003). In addition, correlated activity has been commonly observed in developing systems and is thought to play an important role in the formation of neural circuits (Katz and Shatz, 1996). Despite the relevance of cortical synchrony, its mechanisms and maturation are poorly understood.

Several findings suggest that inhibitory interneurons play a critical role in the generation of rhythmic, synchronized activity (Cobb et al., 1995; Fuchs et al., 2001; Whittington and Traub, 2003). For example, activation of muscarinic or kainite receptors induces synchronous gamma frequency rhythms that involve fast-spiking (FS) inhibitory interneurons (Traub et al., 2005). The low-threshold spiking (LTS) inhibitory interneurons of the neocortex have several characteristics that might allow them to serve as a general spike-synchronizing mechanism. First, LTS cells form frequent inhibitory connections onto local excitatory cells as well as onto the FS type of inhibitory interneurons (Gibson et al., 1999; Beierlein et al., 2003). Second, LTS neurons are interconnected by electrical synapses that can strongly synchronize their spiking activity (Gibson et al., 1999, 2005) and thereby allow them to generate synchronous inhibition on a sizeable domain of surrounding cortex (Beierlein et al., 2000; Deans et al., 2001). Third, several neuromodulators induce a robust depolarization of LTS neurons with minimal excitatory effects on other layer 4 neurons in barrel cortex. These include agonists of muscarinic acetylcholine (Xiang et al., 1998; Beierlein et al., 2000), metabotropic glutamate (mGluR) (Beierlein et al., 2000), noradrenergic (Kawaguchi and Shindou, 1998), and serotonergic (S. J. Cruikshank and B. W. Connors, unpublished observations) receptors.

Although several aspects of LTS function have been considered in relatively mature neocortical circuits, the role of these neurons during development is unknown. Here we describe studies of the developing rat somatosensory (barrel) cortex and show that the synchronizing ability of the LTS cell network matures abruptly at the end of the second postnatal week. Thus, concurrent with the onset of active whisking behavior (Welker, 1964), the neocortex acquires a mechanism for rapidly regulating the degree of synchronous spiking among local pyramidal cells.

Materials and Methods

Thalamocortical slices (350 μm thick) were obtained from Sprague Dawley rats aged postnatal day 8 (P8) to P15 (Agmon and Connors, 1991). Briefly, rats were deeply anesthetized with Nembutal (Abbott Labs, Irving, TX) and intracardially perfused with 10-20 ml of an ice-cold solution containing the following (in mm): 126 NaCl, 3 KCl, 1.25 NaH₂PO₄, 2 MgSO₄, 26 NaHCO₃, 10 dextrose, and 2 CaCl₂ (saturated with 95% O₂/5% CO₂). After perfusion, the brain was rapidly removed, and slices were made with a vibratome while slowly advancing the blade manually (Campden Instruments, Lafayette, IN). The slices were incubated for 45 min at 32°C and then kept at room temperature until they were transferred to a 32°C submersion-style recording chamber.

Paired intracellular recordings of neurons within rat primary somatosensory (barrel) neocortex were the main method used in this investigation. Whole-cell micropipettes were filled with the following (in mm): 130 potassium gluconate, 4 KCl, 2 NaCl, 0-10 sucrose, 10 HEPES, 0.2 EGTA, 4 ATP-Mg. 0.3 GTP-Tris, and 7 phosphocreatine-Tris pH 7.2-7.3 (280 mOsm). All recordings were made in current-clamp mode (Axon Instruments, Union City, CA). Infrared differential interference contrast visualization was performed using a Zeiss (Oberkochen, Germany) Axioskop and a CCD camera (Hamamatsu, Bridgewater, NJ). Before targeting an area for recording, the slice was viewed with a low-magnification objective. This allowed for us to clearly visualize characteristics of layer 4 (i.e., the presence of layer 4 barrels, relatively high contrast with respect to surrounding tissue, and characteristic distance from layer 1). With the high-power immersion lens, the border between layers 5 and 4 was easily visualized. The border between 3 and 4 was less clear, so we restricted our recordings to the middle and deep portions of layer 4. In addition, all of the biocytin-filled LTS cells were confirmed to be within layer 4.

The mGluR (groups I and II) agonist (1S,3R)-1-aminocyclopentane-1,3-dicarboxylic acid (ACPD) (100 μm) was obtained from Tocris Cook-son (Ballwin, MO). Some experiments were performed with 20 μm DNQX (Sigma, St. Louis, MO) and 50 μm APV (Sigma) present to block AMPA- and NMDA-type glutamate receptors, respectively. Data were analyzed with auto- and cross-correlograms and power spectra using Labview software (National Instruments, Austin, TX). Subthreshold membrane potentials were bandpass filtered (1.5-70 Hz) for analysis. The correlograms were first averaged and then normalized by the product of the SD of the amplitude of the signal in each recording, as described previously (Beierlein et al., 2000). The correlation coefficient for each pair was the average cross-correlation of five epochs of 2 s each, for each pair, after administration of ACPD. Action potential cross-correlations were calculated in 10 s epochs (300 ms analysis window; bin size, 20 ms). To compare across cell pairs, the correlograms were normalized. First, spike numbers in each bin were divided by the total number of spikes in the reference cell, and then baseline correlations were zeroed out by subtracting the across-bin average from each individual bin.

To test for the presence of inhibitory connections between LTS and regular-spiking (RS) cells, short (5 ms) current pulses (200-400 pA) were applied to evoke an action potential in LTS cells (0.1 Hz). The amplitude of the IPSP was measured from the averaged responses to 10-50 unitary LTS action potentials at a holding potential of -50 to -53 mV. To estimate the IPSP reversal potential in LTS-to-RS connected pairs, at least 10 RS responses were averaged at -50, -60, -70, -80, and -90 mV. The amplitude of the IPSP was plotted against the holding potential to estimate the reversal potential for each neuron.

For the anatomical reconstructions, cells were filled through the whole-cell pipettes for at least 20 min with biocytin (4 mg/ml). Slices were then incubated in 4% paraformaldehyde solution for at least 24 h, before being transferred to a cryoprotectant solution (30% sucrose/4% paraformaldehyde). Tissue was resectioned into 80-μm-thick slices using a microtome, and biocytin labeling was revealed using 3,3′-diaminobenzidine (Sigma). Neuronal morphology was reconstructed using a camera lucida attached to a light microscope with a 40× objective. We applied three different analyses to these reconstructions to quantify and compare the structural organization of dendritic and axonal arborizations between age groups. Initially, we counted the number of intersections that occurred across a series of concentric circles (10 μm spacing) radiating from the somata (Sholl, 1953). To examine the vertical extent of axonal arborization, we used a grid of parallel lines spaced 20 μm apart and aligned parallel to the orientation of the pial surface (adapted from Xiang et al., 1998). Intersections between these lines and either axonal or dendritic branches were counted. Measurements were gathered into groups by averaging adjacent measurements into bins of 100 μm.

All data are presented as mean ± SD unless otherwise noted. Paired comparisons between groups were tested with a two-tailed Student's t test unless otherwise noted.

Results

Whole-cell recordings were made in layer 4 of primary somatosensory cortex from juvenile rats. All neurons were classified according to their intrinsic membrane properties and synaptic connections (Gibson et al., 1999; Beierlein et al., 2003). Neurons clearly identified as excitatory RS cells and inhibitory LTS cells were included in these analyses.

Once stable paired-cell recordings were obtained, the mGluR agonist ACPD (100 μm) was applied through the bathing solution. At P12, ACPD induced IPSPs that were small, irregular, and poorly synchronized across RS cell pairs (Fig. 1A). In P15 rats, however, ACPD elicited a 3-5 min episode of vigorous, rhythmic IPSPs in RS cells. IPSPs were tightly synchronized across pairs of neighboring neurons (Fig. 1B), as reported previously (Beierlein et al., 2000; Deans et al., 2001). Within a sample of 55 RS-RS cell pairs from P12 to P15, the synchrony of mGluR-induced IPSPs was strongly correlated with age (Fig. 1C)(r = 0.74; p < 0.0001). The maturation of this synchronous inhibitory mechanism was abrupt; it was entirely absent at P12 and fully expressed by P14. At P13, inhibitory synchrony was bimodally distributed, and cells showed either uncorrelated or strongly correlated inhibition (Fig. 1C). The modal frequency of synchronous inhibitory events in ACPD ranged from ∼4 to 12 Hz.

We tested the ability of mGluR-activated IPSPs to synchronize the spiking patterns of excitatory RS cells (which are the targets of inhibitory synapses from LTS cells). RS cell pairs were stimulated to fire with long steps of depolarizing current, and their synchrony was compared in the presence versus absence of ACPD. In P15 cortices, ACPD produced highly correlated action potentials in neighboring RS pairs, presumably mediated by the synchronous IPSPs that were observed between spikes (Fig. 2B,D,F). Note that ionotropic glutamate receptors were blocked (50 μm APV and 20 μm DNQX), preventing any effects of fast excitatory transmission between the RS cells. The central bin of the spike cross-correlograms, which reflects the degree of spike synchrony, increased significantly during ACPD for the P15 pairs (Fig. 2D,F) (p < 0.005, paired t test, ACPD vs predrug). In contrast, ACPD had no effect on spike correlations in P12 cell pairs (Fig. 2A,C,E)(p = 0.94, paired t test).

What accounts for the abrupt maturation of this spike-synchronizing mechanism? Previous studies demonstrated that a network of electrically coupled LTS interneurons can drive synchronous mGluR-induced inhibition (Beierlein et al., 2000; Deans et al., 2001). When LTS cells from P15 were excited by ACPD, their spikes were tightly correlated with rhythmic IPSPs in neighboring RS cells (Fig. 3A). Hyperpolarizing LTS cells with injected current revealed that rhythmic fluctuations of membrane potential underlie mGluR-induced spiking (Fig. 3B). At P15, subthreshold voltage fluctuations of LTS cells were strongly anticorrelated with rhythmic IPSPs recorded from neighboring RS cells (mean peak correlation, -0.68 ± 0.14). However, at P12, the activity of LTS and RS cell pairs was entirely uncorrelated in the presence of ACPD (mean peak correlation, 0.00 ± 0.15), and there was a general paucity of subthreshold activity in both cell populations. The relationship between induced LTS cell activity and RS cell inhibition developed between P12 and P15 (Fig. 3C) (r = -0.84; p < 0.0001), with the same developmental schedule seen for the synchrony of mGluR-induced inhibition (Fig. 1C). This similarity reinforces the notion that the maturation of the spike-synchronizing mechanism is attributable to developmental changes in the LTS network or its inhibitory synapses.

As one measure of the maturation rate of LTS interneurons, we examined their dendrites (n = 12) and axons (n = 6) between P12 and P15. Physiologically identified LTS cells from layer 4 were filled with biocytin and reconstructed using camera lucida (Fig. 4). In general, the dendrites of LTS cells had a vertical bias to their orientation (Amitai et al., 2002), and their axons extending upward from the somata were the most dense. Both dendrites (Fig. 5A) and axons (Fig. 5B) of LTS cells also showed an age-dependent increase in the total number of crossings of concentric circles originating at the soma (Sholl, 1953). Several bins achieved differences that were statistically significant (p < 0.05) in both the axonal (225 and 275 μm) and dendritic (175-275 μm) analyses. The extent of arborization was estimated by the number of times that axons or dendrites crossed evenly spaced lines that were parallel to the pia (Xiang et al., 1998). The total number of crossings of both dendrites (Fig. 5C) and axons (Fig. 5D) increased between P12 and P15 (dendritic crossings, 97 ± 43 to 122 ± 21; axonal crossings, 330 ± 152 to 449 ± 59), implying an increase in the total length of neuronal processes. Most of this growth occurred from 300 and 600 μm below the pia, corresponding to lower layer 2/3 and the border region with layer 4, including a doubling of the number of axonal crossings from 154 ± 140 to 301 ± 79.6.

The synchronization of mGluR-induced IPSPs depends strongly on the presence of electrical synapses between LTS cells (Beierlein et al., 2000; Deans et al., 2001). Because there is evidence that electrical coupling between neurons is strongly regulated during cortical development (Connors et al., 1983; Lo Turco and Kriegstein, 1991; Peinado et al., 1993; Galarreta and Hestrin, 2002; Meyer et al., 2002), it is possible that maturation of electrical coupling underlies the development of LTS-mediated synchrony. To test for changes in electrical synapses, we recorded from LTS cell pairs from ages P8-P15. Figure 6A shows examples of electrically coupled LTS pairs from P12 and P15. Across a sample of 21 pairs of closely spaced (intersomatic distance <50 μm) LTS cells, neither the strength nor the incidence of electrical coupling varied with age (Fig. 6B).

A sudden increase in the strength or number of inhibitory synapses originating from LTS cells could account for the abrupt onset of synchronous inhibition. We tested inhibition directly by recording from 57 pairs of LTS-RS cells from P12 to P15. There was a weak relationship between age and the probability of LTS-to-RS inhibitory connections (0.43 at P12/P13; 0.52 at P14/P15). The strength of these inhibitory connections, defined as the average amplitude of unitary LTS-to-RS IPSPs (Fig. 7A), increased from P12 (-0.52 ± 0.26 mV) to P15 (-0.79 ± 0.37 mV), as reflected in the cumulative probability histogram (Fig. 7B). There was a significant increase in inhibitory synaptic strength across all sampled pairs from P12 to P15 (r = -0.40; p < 0.05).

Nonsynaptic factors did not have a significant effect on the change in unitary IPSP size observed between P12 and P15. The input resistance of RS cells decreased from P12 (280 ± 68 MΩ; n = 9) to P15 (235 ± 79 MΩ; n = 9), which by itself would have slightly reduced the amplitude of IPSPs as cells matured. The IPSP reversal potential was indistinguishable between P12 (-81.3 ± 1.0 mV; n = 5) and P15 (-82.1 ± 2.0 mV; n = 5). In addition, RS neurons were given depolarizing current to bring their membrane potentials to between -50 and -53 mV while testing for inhibitory connectivity. As a result, changes in unitary IPSP size in LTS-RS pairs probably reflected an increase in the mean inhibitory synaptic conductance.

Maturation of the intrinsic properties of LTS cells could contribute to the onset of the spike-synchronizing mechanism. Input resistance (r = -0.59; p < 0.0001) and membrane time constant (r = -0.65; p < 0.0001) dropped, and spike-frequency adaptation rates increased (r = -0.60; p < 0.0001) between P12 and P15 (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). Developing LTS cells also became more sensitive to mGluR activation. In response to ACPD, every LTS cell recorded in the absence of injected current was strongly depolarized and generated spikes (n = 47; ages P12-15), including all 14 cells from age P12. However, the frequency of ACPD-induced spiking was positively correlated with age [r = 0.62; p < 0.001; mean spike rate was 10.4 ± 6.8 Hz at P12 (n = 8) and 22.3 ± 6.8 Hz at P15 (n = 10)]. Furthermore, the coefficient of variation of spike intervals in ACPD-treated LTS cells increased from P12 (0.29 ± 0.13; n = 8) to P15 (0.52 ± 0.23; n = 10). These results could be attributable to age-dependent changes in the expression of mGluRs or their downstream signaling pathways and membrane channel effectors. The subthreshold characteristics of mGluR-induced activity also changed (Fig. 8A). Both the magnitude (Fig. 8B)(r = 0.62; p < 0.0001) and the dominant frequency (Fig. 8C) (r = 0.49; p < 0.0002) of ACPD-induced rhythms increased dramatically from ages P12 to P15. This implies that the maturation of LTS cell physiology is a major contributor to the abrupt appearance of the spike-synchronizing mechanism.

One of the most remarkable features of the LTS cell population is its ability to evoke synchronized inhibition over a large volume of cortical circuitry (Beierlein et al., 2000). This presumably arises from the dense and spatially extended nature of electrical coupling within the LTS cell network (Deans et al., 2001; Amitai et al., 2002). Here we found that pairs of older LTS cells (P14-P15) could generate strongly synchronized subthreshold fluctuations even when there was no measurable electrical coupling between them, as in the example shown in Figure 9A. This widely spaced (250 μm) pair, which had a negligible coupling coefficient, was strongly synchronized when activated by ACPD. Fast chemical synaptic transmission was blocked in these experiments, implying that synchrony likely resulted from interactions via electrical synapses in intervening LTS cells. Similar recordings made from variably spaced LTS-LTS cell pairs showed that electrical coupling was common among the closest cell pairs at all ages tested (Fig. 9B, filled symbols). The mGluR-induced activity of the electrically coupled pairs was highly correlated at P14-P15 (correlation coefficient of 0.82 ± 0.13; triangles) and less strongly correlated at P12 (0.51 ± 0.13). The more distantly spaced cells were not electrically coupled (Fig. 9B, open symbols), and mGluR-induced activity was uncorrelated in the uncoupled LTS cell pairs at P12 (0.03 ± 0.03). At P14-P15, however, a substantial degree of synchrony remained, even between widely spaced, electrically uncoupled pairs (0.53 ± 0.23; p < 0.005).

Discussion

Our results show that a spike-synchronizing mechanism develops rapidly in the adolescent rat neocortex between P12 and P15. At P12, mGluR activation resulted in weak and uncorrelated inhibition within RS neurons. At P15, the same treatment invariably led to episodes of rhythmic, synchronized IPSPs that were capable of synchronizing spikes among RS cells. These mGluR-activated rhythms are mediated by an electrically coupled system of LTS inhibitory interneurons (Gibson et al., 1999; Beierlein et al., 2000). The precise mechanisms underlying the rapid maturation of the spike-synchronizing system remain unclear, but several developmental modifications of the LTS network and its synaptic connections probably contribute.