Treatment with the anti-mitotic agent methylazoxymethanol (MAM) on embryonic day 33 (E33) in ferrets changes features of somatosensory cortex. These include dramatic reduction of cells in layer 4, and altered distributions of thalamocortical afferent terminations and GABAA receptors. To determine the effect of the relative absence of layer 4 on processing of sensory stimuli we used current source-density profiles to assess laminar activity patterns. Nearly synchronous activation occurs across all layers in treated animals, which contrasts with the normal cortical activation pattern of initial sinks in layer 4. This change after MAM treatment is consistent with the absence of layer 4 cells and widespread termination of thalamocortical afferents. Using periodic stimulation at ‘flutter’ frequency, layer 4 neurons in normal somatosensory cortex fire reproducibly to the stimulus rate; the capacity for entrainment is best for layer 4 and weaker in the extragranular layers. The capacity to encode periodic sensory stimuli is disrupted in MAM-treated somatosensory cortex; after an initial response to the onset of periodic stimuli, neurons in all cortical layers show weak entrainment. Neural responses to sensory drive in E33 MAM-treated cortex are also embedded in levels of neural activity substantially above those in normal somatosensory cortex. Sustained stimulation additionally reveals different capacities in each layer for improved signal-to-noise ratios, with layer 4 neurons in normal animals exhibiting the most improved signaling over time. We conclude that normal thalamic terminations, an intact layer 4 and subsequent intracortical processing are integral to proper encoding of stimulus features.
Layer 4 of the cerebral cortex is a key site of thalamic termination, especially in sensory areas. After exciting neurons in layer 4, information transfers to the supragranular and infragranular layers; this has been demonstrated anatomically and functionally by single unit recording coupled with latency measures (Lund et al., 1973; Hubel and Wiesel, 1977; Johnson and Alloway, 1996; Zhu and Connors, 1999; Dantzker and Callaway, 2000; Feldmeyer et al., 2002; Thomson et al., 2002; Hirsch, 2003; Pinto et al., 2003; Thomson and Bannister, 2003). Current source-density profiles (CSD) recorded in sensory regions of cortex elaborate this pattern, showing that initial activity occurs in layer 4, with subsequent activity in the upper and lower layers (Mitzdorf, 1985; Di et al., 1990; Kenan-Vaknin and Teyler, 1994; Schroeder et al., 1995; Aizenman et al., 1996). Despite our knowledge of the general plan of information flow through the cortical layers, little is known about the distinct response of each cortical layer to single or repetitive stimulation.
The idea that inhibition sharpens the ultimate processing of information in the cerebral cortex is supported by observations that responses in layer 4 cells receiving the initial thalamocortical synapses are focused by influence from GABAergic cells, which also receive direct thalamic input (Lubke et al., 2000; Sachdev et al., 2000; Miller et al., 2001; Porter et al., 2001). Direct feed-forward inhibition may dominate incoming excitation by tuning the excitatory neurons that receive principal thalamic input (Miller et al., 2001). Additional evidence supports the idea that GABA influence and the proper balance of excitation and inhibition sculpts responses necessary for distinct perception of sensory information (Galarreta and Hestrin, 2001; McBain and Fisahn, 2001; Hirsch, 2003).
We developed a model in ferrets that interrupts the birth of layer 4 neurons. This leads to adult animals possessing a thin layer 4 in the primary somatosensory cortex, with the other layers remaining relatively normal (Noctor et al., 2001). The model was produced by injecting pregnant ferrets with an antimitotic (methylazoxymethanol, MAM) on a day when many layer 4 cells of area 3b are being produced (embryonic day 33, E33) (Noctor et al., 1997, 2001). This configuration allows us to clarify the importance of layer 4 in orchestrating responses by studying the flow of information through somatosensory cortex with little contribution from layer 4.
Concurrent with the diminution of layer 4 after MAM treatment, a population of GABAA receptors alter their distribution by expanding in density outside of normal layer 4 (Jablonska et al., 2004). In addition, in contrast to their normal heavy termination in layer 4, thalamic projections distribute throughout cortical layers. Our original observations in this model of diminished layer 4 also demonstrated that the overall topographic map in somatosensory cortex is normal when multiple unit activity (MUA) is recorded without regard for laminar distinctions (Noctor et al., 2001). This suggests that the topographic arrangement of the thalamocortical projections is relatively normal in the E33 MAM-treated animals.
We conducted a series of experiments in normal and E33 MAM-treated animals by recording CSD profiles, MUAs and evoked potentials (EPs) in response to single taps and repetitive stimuli to the ferret forepaw. We hypothesized that the normal sequence of cortical activation would be disrupted after MAM treatment at E33 in ferrets and that the cortex of treated animals would exhibit a degraded capacity for sensory encoding. Our results led us to conclude that the flow of information is substantially altered in the somatosensory cortex with diminished layer 4. These findings demonstrate that layer 4 is essential for normal processing of information through cortical layers.
Materials and Methods
MAM Treatment and Disruption of Layer Formation
Timed pregnant ferrets were obtained from Marshall Farms (New Rose, NY). On the appropriate day, a pregnant ferret was anesthetized with 5% halothane and 0.05% N2O. An injection of MAM (12 mg/kg, dissolved in saline; Sigma, St Louis, MO) was administered i.P. on E33. MAM injections on this gestational day disrupted cells undergoing final mitosis that were intended for layer 4 (E33) (Noctor et al., 1997, 2001). Ferrets were closely monitored after injections to ensure proper health.
Experiments were conducted on 16 ferrets of either sex. They were anesthetized with halothane (2–3%); expired CO2 and body temperature (35–37°C) were monitored and maintained. Ophthalmic ointment was placed onto corneal lens surfaces and heart rate monitored periodically during the experiment. An i.v. catheter was inserted in the right external jugular vein to allow continuous infusion of 5% dextrose in lactated Ringer's solution. The scalp and left forelimb were shaved and the animal's head secured in a stereotaxic device. A craniotomy was performed over the right somatosensory cortex and the dura mater removed to expose the pia. The brain was covered with warm mineral oil and photographed (including a scale). An enlarged photograph was used to map recording sites.
Intracortical recordings were made via a platinum-iridium microelectrode that was lowered to a depth of 2000 μm, which was usually below layer 6. The electrode was allowed to settle for 15 min before we began systematically retracting in 100 μm steps until returning to the cortical surface. The first sample was collected at 2000 μm and thereafter at 100 μm intervals. A return to the cortical surface (end of retraction) was established by monitoring the physiological signal for onset of noise artifact, noting the waveform features for peak reversals and considering the distance traveled according to the microdrive. We twice positioned the electrode at the cortical surface: when the microelectrode contacted the cortical surface initially and again at the end of the retraction. At the end of the retraction, we noted the distance measurement on the electrode carrier, positioned the electrode tip at the cortical surface under magnification and noted the scale reading on the electrode carrier again. Any discrepancies between the initial and final measurements were considered a measure of cortical displacement as a consequence of the penetration. This discrepancy was documented and regarded as one of the contributing factors to our determinations of laminar position of each electrode recording level. Averages of 30 trials were made for each run, and two runs were made at a subset of the penetration sites. A reference electrode was attached to scalp muscle lateral to the opening in the cranium. Data were amplified, filtered (DC, 500 Hz for field potentials and DC, 2 kHz for multiple unit activity), fed to an audio monitor and oscilloscope, and converted for computer storage at a rate of 5 kHz. Data collection began 20 ms before the onset of stimulation and terminated 180 ms later or 500 ms later depending on the stimulus condition.
Stimuli consisted of light mechanical stimulation of digit 4 and were presented at a rate of either 1 or 20 per second. Sinusoidal output of 20 cycles per second was delivered to Ling mechanical transducers attached to 1 cm diameter Perspex spheres. Only the most azimuthal surface of the spheres contacted the skin during stimulation. Probe displacement during stimulation was 1.0–1.5 mm, measured using scales and video analysis. Spheres were closely applied to the thick glabrous skin of the forepaw digit, such that dimpling occurred; this was to ensure contact for the duration of the sampling period. When stimulated, the individual digit was isolated from neighbors and elevated to prevent stimulation of the opposite skin compartment. Paws and digits were shaved prior to mechanical stimulation.
Injections of a fluorescent tracer (Fluororuby, at a concentration of 50 mg/ml, conjugated to dextran from Molecular Probes) were made into selected recording sites, into the same penetration made by the recording electrode. These tracers were used to locate the relative position of the penetration with respect to cortical curvature and to the underlying cytoarchitectonic fields. Micropipettes with tip diameters of ∼30 μm were lowered to selected depths and allowed to settle for 2 min before the dye was injected iontophoretically at 3 mA for 4 min with positive alternating current.
After each recording session, animals received an overdose of Na pentobarbital (65 mg/kg i.v.) and were perfused transcardially with 0.9% saline followed by 4% sucrose in 0.1 M phosphate-buffered 4% paraformaldehyde. The brain was removed and placed in the same fixative with 10% sucrose. The next day, the concentration of sucrose was increased to 20% for cryoprotection. After sinking, the brain was frozen and cut on a cryostat at 40 μm thickness in the coronal plane. Alternate sections were saved for Nissl staining or fluorescence microscopy. Contour drawings of Nissl-stained sections were made and morphological landmarks (e.g. postcruciate dimple and coronal sulcus), the location of dye injections and cytoarchitectonic boundaries for areas 4, 3a, 3b, 1 and 2 were included (McLaughlin et al., 1998; McLaughlin and Juliano, 2003).
Digitized data were submitted to Mathematica (Wolfram Research) for digital filtering and analysis. To obtain field potentials, inverse Fourier transforms were used to remove frequencies higher than 200 Hz from the raw data. Records containing artifacts were omitted. As an index of animal and recording stability, at least two datasets were obtained non-consecutively at a subset of penetrations. Discrepancies were noted in ∼1 in 15 penetrations (∼7%). When datasets obtained from the same site did not yield comparable results, we omitted the data for that animal.
CSD analysis was performed following off-line digital filtering. The analysis is based on the second spatial derivative of field potentials across three cortical layers, 200 μm apart as given by the following computation:
Features of the Responses
Innocuous taps to the forepaw digits were used to evoke responses in the somatosensory cortex of normal (n = 9) and MAM-treated (n = 7) ferrets. These responses consist of an initial negativity (N1) followed by a positivity (P2). A representative example can be seen in Figure 1. Temporal features of the depth-recorded responses were similar to those recorded at the pial surface (McLaughlin and Juliano, 2003). Tactile stimulation of digits 3, 4 and 5 results in mean (± SE) peak latencies across layers of 14.7 ± 2.7 ms (n = 17 datasets) for the initial negativity and 24.2 ± 4.3 ms (n = 17 datasets) for the subsequent large positive P2 in normal animals. For animals treated with MAM on E33, mean peak latencies across layers were 15.1 ± 3.1 ms (n = 13 datasets) for the initial negativity and 22.7 ± 5.1 ms (n = 13 datasets) for the subsequent large positive P2. Mean peak latencies are not statistically different for normal and MAM-treated animals (n.s., Student's t-test).
Cortical Activation Sequence
A number of existing studies consistently show a reproducible activation pattern across layers in primary sensory cortex (Mitzdorf, 1985; Di et al., 1990; Kenan-Vaknin and Teyler, 1994; Schroeder et al., 1995; Aizenman et al., 1996). To obtain a view of local synaptic activity across the layers in normal and treated animals, we applied a current source density (CSD) computation to the recorded field potentials. To confirm the presence of synaptic activity, we also recorded extracellular multiple unit activity at the same recording levels as the field potentials. Using this approach, we can localize the origin of neural activity to within about 200 μm. Our CSD and MUA profiles demonstrate that the activation sequence in normal ferret somatosensory cortex (specifically, area 3b) is consistent with previously described patterns in other animals. In the findings presented here, the middle layer responses in normal brains were usually recorded at depths of 800–900 μm, defined by high spontaneous activity, ease of activation and lastly—and less influential—the size of the receptive field. The location of other layers was judged on the relative distance from the determined middle layers and on physiological activity. Both the upper and lower layers have lower spontaneous activity than the middle layers, particularly in area 3b. Intermittent vigorous discharges in the upper and lower layers were indicative of large neurons, presumably pyramidal cells (Dykes and Lamour, 1988). In MAM-treated animals, which have a very thin layer 4, laminar positions in the figures are meant to indicate relative position within the cortex. Figure 2 presents examples of penetrations in the somatosensory cortex of normal and MAM-treated brains.
A single tap to the skin of digit 4 elicits activity in normal area 3b of ferret somatosensory cortex with a cortical activation sequence similar to those reported in studies in other species (Fig. 3). Current sinks in the middle layers of normal animals in response to a single light tap to digit 4 occur at 10.2 ± 0.5 (n = 6 datasets). Next, sinks are observed in the upper layers with a mean latency of 12.8 ± 0.7 ms and then in layer 5 with a mean latency of 19.4 ± 1.5 ms. The latencies in layer 6 are short, similar to those in layer 4, with a latency of 10.4 ± 0.9 ms. The earliest current sinks computed by CSD analyses were localized to the middle input layers in normal animals (Fig. 3). Coincident passive return currents are evident in the nongranular layers as current sources (upward deflections; Fig. 3, upper CSD). Activity subsequent to onset in the middle layers is present at depths appropriate for upper layer 3 and layers 2 and 5. Current sinks in upper layer 3 and layer 2 are typically larger in amplitude and duration compared with those computed for the middle layers. The pattern observed in layer 5 is usually somewhat delayed relative to the upper layers, of low amplitude, and protracted (see the responses for the lower layers in Fig. 3, upper CSD, MUA). In layer 6, on the other hand, latencies are short, similar to those in layer 4. This is not surprising, given the known direct thalamic input to layer 6 (e.g. Jones, 1986). Pairwise-layer comparisons are statistically significantly different, except for comparison of latencies for layers 4 and 6 (analysis of variance [ANOVA], p < 0.01; pairwise comparisons using a post hoc Dunn t test; Fig. 4).
Single taps to the digit skin of normal animals clearly generate spike discharges in the middle layers of area 3b coincident with the earliest sinks, indicating that the current density measures are associated with excitatory events (Fig. 3; see asterisks in the CSD and MUA profiles at the top). Increased discharges are also evident in nongranular layers concomitant with current sinks (Fig. 3, upper CSD, MUA).
In E33 MAM-treated animals, single light taps to digit 4 yield an activation sequence distinct from that described above for normal animals (Fig. 3; profiles from two animals are shown). Current sinks do not appear organized in a systematic spatiotemporal sequence. Response latencies across layers after taps to digit 4 are not as distinct from each other as they are in normal cortex (13.2 ± 1.2 ms for the middle layers, 13.8 ± 0.5 ms for the upper layers and 13.0 ± 2.0 ms for the layer 5; the latency for layer 6, however, was 11.2 ± 0.3 ms; n = 6 datasets; ANOVA, not significant, Fig. 4). Latencies for layer 6 are comparable to those for normal ferrets, whereas latencies for layer 5 are shorter than in normal animals (Fig. 4). Layers 5 and 6 have largely formed before the MAM injection, and layer 6 is likely to receive direct thalamic input.
The most striking feature of the CSD patterns is the relative absence of a clearly recognizable spatiotemporal series of sinks (Fig. 3, bottom two CSDs). There is no initial current sink associated with the middle layers, nor is there a successive series of activity sinks in the upper and lower layers. Since the presence of a sink or source indicates a gradient in local current density, the relative absence of noteworthy shifts from baseline indicates that current (membrane potential) is more uniformly distributed across the layers. A broad initial activation pattern is also evident in the pattern of the spike discharges, which occur nearly simultaneously across all cortical layers (Fig. 3, bottom two MUA profiles).
Encoding of Periodic Stimuli
We hypothesized that the disruption in the laminar activation sequence of animals treated on E33 is likely to interfere with transfer of information about stimulus features in somatosensory cortex. To address this possibility, we presented periodic stimuli as a way to provide feature-rich information. We used tactile stimulation of the fourth digit of the forepaw, which has a representation on the crown of the posterior sigmoid gyrus. Evoked responses in normal and MAM-treated animals contain periodicities related to the stimulation rate (Fig. 5). In normal animals, evoked responses were more pronounced and nearly twice as large as those from E33 MAM animals. Responses from treated animals were embedded in background spontaneous activity and difficult to distinguish (Fig. 5). Activity from normal cortex exhibits good entrainment (representative upper and middle layer responses are shown in Figure 5). Slight differences in the features of the responses for each layer occur, but overall periodicity is evident. In the E33 MAM-treated animal, periodic activity at the stimulation rate is visible in the upper layer trace, but the middle layer response exhibits lapses in ability to entrain at the stimulation rate (Fig. 5).
Figures 5 and 6 demonstrate that in normal animals, the field potentials, CSD profiles and MUA reflect the rate of stimulation. Spectral analysis of the CSD demonstrates accurate representation of the stimulus features across cortical levels, particularly in the middle layers (Fig. 6, left). In the CSD of normal animals, the number of current sinks evident in the CSD profile corresponds well with the number of cycles in the stimulus train (Fig. 6, left). Spectral analysis confirms that the periodicity in the CSD corresponds to the stimulation rate (Fig. 6; see arrowheads and asterisks). This correspondence is strongest for the middle layers, indicating that middle layer neurons are differentially entrained with the incoming activity drive compared with cell populations in other layers. The transfer of information out of layer 4 results in a delay of current sinks in the upper and lower layers (Fig. 6, left; see solid arrow heads). Spectral analyses also reveal that the current sinks give rise to spectral energy occurring at the rate of the stimuli as well as at several subharmonics of the stimulation rate (Fig. 6, left, FFT of CSD and Fig. 7). Spectral analysis demonstrates that activity in frequency bands unrelated to the stimulation rate is relatively minor compared with that linked to stimulation (compare the bar associated with 30Hz, i.e. not related to stimulation rate, to the fundamental and subharmonic components in Fig. 7).
In animals treated with MAM on E33, a clear activation sequence is not evident in response to intermittent stimuli (Fig. 6, right). Current sinks are evident throughout the laminar profile but do not give rise to an identifiable spatiotemporal pattern of activation. Using spectral analysis, we observe that the current sinks in the CSD occur mostly at subharmonic frequencies of the stimulation rate (Fig. 6, right; see gray asterisks). Graphs of the percentage of spectral power contained within each frequency band related to the stimulation rate reveals that MAM-treated animals exhibit overall less stimulus-linked spectral energy compared with normal animals (Fig. 7; compare 5, 10, 20 and 40 Hz bars of E33 MAM with those in normal animals). Although the values of the unrelated frequency content (30 Hz spectral energy) are not substantially greater in the MAM-treated animal than in the normal, they represent proportionally more power in comparison with the other stimulus-related frequency bands (Fig. 7). The distribution of frequencies for each of the four representative levels is statistically significantly different between normal and MAM animals (P < 0.01; χ2-test for distributions).
Multiple Unit Activity is Reflected in the Current Sinks
Further characterization was obtained by multiple unit activity recorded across the cortical layers in response to periodic tactile stimulation. Normal animals displayed clear spike activity to each cycle of stimulation (Fig. 6, left, MUA). The time course of responses can also be seen when the MUAs are viewed on a slightly expanded time scale (Fig. 8). When an intermittent stimulus is delivered, the MUA activity is initially strong in layer 4 and transfers its strength to the upper layers and lower layers, consistent with the single tap observations, as can be seen in the latency information shown in Figure 4. By the third or fourth stimulus cycles, clear entrainment occurs in all layers (Fig. 8). To further quantify the strength of responses to the intermittent stimuli in normal cortex we measured the ratio of the amplitude of response in representative upper, middle and lower layers in relation to the background amplitude at a point midway between each stimulus. Figure 9 demonstrates that although there is variability, the strength of response increases gradually with the increasing numbers of stimuli and levels off around the seventh or eighth stimulus. We also observe that the strength of the response is greatest in the middle layers compared with the lower and middle layers.
Overall, responses to early cycles of the stimulus train were less pronounced than those formed during prolonged stimulation. In the middle layers specifically, cells clearly responded to each cycle of stimulation but early in the train the responses are less defined, whereas with continuous exposure to the stimulus train, the spikes are distinct and highly tuned to the presentation of the stimulus. This can be seen in Figure 10, which demonstrates the cortical response in the initial period after delivery of a repetitive stimulus and in the final period of data collection. Comparison of the MUA background levels between successive stimuli indicates that they are relatively lower for the middle layers than for upper and lower layers. To assess changes in signal-to-noise features of different layers, a ratio of mean spontaneous activity after each stimulus to mean pre-stimulus baseline activity was taken for representative cortical layers. In normal animals, this ratio is initially high, and then decreases over the first half of the stimulus period for all layers. At this point, the background activity remains low for the middle layers of normal animals, but increases in the upper and lower layers (Fig. 11, normal). In MAM-treated animals, while an initial cortical response is evident, the capacity for entrainment is weak and discharges appear buried in background levels of activity (Figs 5 and 10). In MAM-treated animals, a differential laminar response pattern is not evident. Rather, activity of the upper, middle and lower layers is similar and reveals a lack of differential processing across the layers (Figs 10 and 11, E33 MAM). Distributions of background activity levels over time are statistically significantly different for comparisons of the middle layers and non-granular layers in normal animals (P < 0.005; χ2-test for distributions; n = 6 datasets), but not for differences between background activity levels in treated animals (n = 6 datasets). Statistical models indicate that background suppression during the first time segment is readily explained by a second-degree regression model incorporating layer and time as the dependent variables. Parameters of the model are defined in terms of the major identified sources of systematic variance, which gives a model of the form
Information Transfer through Cortical Layers
Disruption of neurogenesis on embryonic day E33 or E34 by injection of MAM results in reduction of cells that normally populate layer 4, causing a conspicuous paucity of spiny stellate cells (Noctor et al., 2001). Lower layer 3 neurons, the stellate neurons of layer 4 and interneurons of layers 3 and 4 normally receive the vast majority of thalamocortical input (White, 1989; Senft and Woolsey, 1991; Johnson and Alloway, 1996). Neurons of lower layer 3 and layer 4 convey their output vertically to upper and lower layers for subsequent processing within the column and in neighboring columns (Armstrong-James et al., 1992; Staiger et al., 2000; Petersen and Sakmann, 2001; Feldmeyer et al., 2002; Laaris and Keller, 2002; Schubert et al., 2003; Thomson and Bannister, 2003). Deep pyramidal neurons also receive contacts from thalamocortical axons but their projections are not regarded as principal cortical input (Katz and Callaway, 1992; Mountcastle, 1997; Buonomano and Merzenich, 1998; Thomson and Bannister, 2003). Thalamic input onto lower layer pyramidal cells generates strong local synaptic activity but weaker synaptic transfer to the upper layers (Benardo, 1997; Rockland, 1998; Thomson and Bannister, 2003). Neurons in layers 2 and 3 respond 2–3 ms after middle layer neurons of the same vertical column (Lubke et al., 2000). This is not to suggest, however, that upper and lower layer cells completely derive their response properties from neurons of the middle layers (Armstrong-James et al., 1992). They are influenced by neighboring cortical input from the upper and lower layers as early as 3 ms after the initial cortical response in the middle layers and exhibit complex properties influenced by surrounding cortical cells and not simply by integrated input from neurons in the granular layers (Armstrong-James et al., 1992). Nonetheless, a normal laminar activation sequence requires initial activation of the middle layers with subsequent transfer to the other layers (Mitzdorf, 1985; Di et al., 1990; Armstrong-James et al., 1992; Schroeder et al., 1995; Staiger et al., 2000).
In the absence of middle layer stellate cells in the E33 MAM-treated animals, thalamocortical afferents do not have access to their normal middle layer targets. While lower layer 3 pyramidal cells are available for contact, these cells do not possess neuronal connectivity comparable to that of spiny stellate cells and, by inference, cannot transfer the features of a sensory signal in a normal manner to extragranular cortical layers (Miller et al., 2001). Thalamic input onto granule cells of the middle layers contributes to a complex circuitry that governs the output characteristics of middle layer cells (Brumberg et al., 1999; Miller et al., 2001; Linden and Schreiner, 2003). The neural signal is highly processed by excitatory and inhibitory circuits converging onto excitatory neurons of the middle layers (Kim et al., 2003). Consequently, output from these cells represents a refined input onto cells of the extragranular layers (Sachdev et al., 2000; Miller et al., 2001). Pyramidal neurons of layers 2/3 and 5 extend axons tangentially and are the main output from a functional column (Burkhalter, 1989; Fitzpatrick, 1996). These projections to neighboring columns sharpen the influence of the neuronal column receiving most of the input by reducing the influence of neighbors. In the absence of normal layer 4 contributions in the MAM-treated animals, neurons of layers 2, 3 and 5 will have unrefined functional output that contributes to poor inter-columnar communication.
The Effect of Redistributed Thalamocortical Afferents after MAM Treatment
In our model, the remaining pyramidal cells are in a position to receive thalamic input but are apparently not capable of directing the thalamic afferents to terminate locally. Rather, after E33 MAM treatment, thalamic afferents terminate widely through the cortical layers (Palmer et al., 2001). Both pyramidal and stellate cell types receive strong excitatory and inhibitory connections from within their corresponding vertical column. Spiny stellate neurons receive local input primarily from layer 4 cells (excitatory and inhibitory), whereas pyramidal neurons receive input from all layers (except layer 1) of a column (Schubert et al., 2003). Since the predominant locus of thalamic drive is to layer 4, the widely distributed input to pyramidal cells in our model suggests that sensory cortical responses in treated animals will be less well tuned to the features of a periodic stimulus. In contrast, comparable intrinsic and passive membrane properties indicate that these two cell types can respond equally well to incoming excitatory drive. In MAM-treated animals, therefore, the initial response to cortical stimulation should appear relatively normal, whereas altered responses are expected to occur to periodic drive that engages intracortical networks of excitatory and inhibitory feed-forward and feedback mechanisms.
In MAM-treated cortex, the sequence of activation consists of nearly synchronous activation across all layers, which contrasts with the normal activity pattern conspicuous by distinct initial sinks in layer 4. This change in activity pattern after MAM treatment is consistent with the diminished presence of layer 4 cells and the widespread termination of thalamocortical afferents previously reported.
It is not known if termination in the lower layers, particularly layer 6, is normal in our model. Agmon et al. (1993) suggest that during development, the termination of thalamic afferents onto layer 6 aids in directing the organization of thalamic afferents onto layer 4, proposing that cues in layer 6 orchestrate topographic order. In addition, Ghosh and Shatz (1992) found that subplate neurons interact with thalamocortical afferents early in corticogenesis and may be involved in development of the sensory map. The somatosensory cortex in animals treated with MAM on E33 exhibits normal topographic organization (Noctor et al., 2001), which confirms the integrity of structures and circuits in the lower cortical plate and subplate in our model.
The Influence of Inhibition in Shaping Cortical Responses
The capacity to encode the rate of periodic sensory stimuli is also disrupted in MAM-treated somatosensory cortex. Layer 4 neurons in ferret area 3b normally fire reproducibly to periodic drive presented at rates within the ‘flutter’ range, i.e. 5–40 Hz, as they did here. In normal animals the capacity for entrainment is best for layer 4 and weaker in the supragranular and infragranular layers. In MAM-treated animals with a diminished layer 4, there was no entrainment to the intermittent stimuli and, after an initial response, all cortical layers failed to follow the stimulus delivery. Finally, neural responses to sensory drive in E33 MAM-treated cortex are embedded in levels of ‘neural noise’ substantially above those in normal somatosensory cortex.
The failure of the remaining layers to entrain to intermittent stimuli may be explained by mechanisms involving inhibition. MAM treatment on E33 leads to changes in ferret somatosensory cortex that include alterations in the distributions of GABAA receptors (Noctor et al., 2001; Jablonska et al., 2004). In normal ferret somatosensory cortex GABAAα receptors are highly concentrated in layer 4. After MAM-treatment, GABAAα receptors are widely distributed throughout the remaining layers (Jablonska et al., 2004). The normal GABA receptor distribution suggests that excitatory input from sensory thalamus makes direct contact with spiny stellate cells and inhibitory interneurons of the middle layers (White, 1989; Miller et al., 2001). Inhibition by interneurons is sufficiently delayed so that its influence on the initial cortical response is limited (Miller et al., 2001). With repeated stimulation, however, inhibition exerts a powerful influence on subsequent responses (Gardner et al., 1984; McLaughlin and Kelly, 1993; Miller et al., 2001; Shapley et al., 2003). One important reason for the powerful influence of inhibitory mechanisms on subsequent responses is the time course of inhibition. Gardner et al. (1984) demonstrated a 40 ms window in which test responses were less than maximal following the onset of the initial response to the conditioning stimulus. McLaughlin and Kelly (1993) established that cortical responses are less able than subcortical responses to recover amplitude with repeated stimulation. This diminished capacity for recovery was attributed to the complex balance of excitation and inhibition that characterizes feature selectivity of cortical neurons in somatosensory cortex (Hicks et al., 1985; Alloway et al., 1989; Juliano et al., 1989; Whitsel et al., 1989, 1991; Lee et al., 1992; McLaughlin and Kelly, 1993; Tommerdahl et al. 1999).
Inhibitory interneurons normally make substantial connections on excitatory neurons, and form recurrent connections and feed-forward contacts onto other subpopulations of inhibitory neurons. In our model of cortical dysplasia, the pattern of GABAA receptors is redistributed so that the normal high density of receptors in layer 4 is extended to layers 2, 3 and 5 (Jablonska et al., 2004). A redistribution of GABAergic receptors suggests that the subpopulations of GABAergic cells activated by thalamic input and engaging in local connections after MAM treatment are different from those in normal animals. A shift in the GABAergic cells activated by incoming thalamic drive may be pivotal in the reduced capacity of neurons in MAM-treated cortex to entrain to periodic stimulation.
The failure of information transfer, however, is equally consistent with a disturbance of temporal integration in the available pools of excitatory and/or, inhibitory neurons. Several studies support alterations in intrinsic neuronal properties in models of cortical dysplasia. Benardete and Kriegstein (2002) developed a model of cortical dysplasia in which pyramidal neurons have a decreased sensitivity to GABA; the authors suggest this may be interpreted as a decrease in the postsynaptic efficacy of GABA (Benardete and Kriegstein, 2002).
A Model of Information Processing in Dysplastic Cortex
In normal animals, thalamocortical input predominantly activates cells and processes in the middle layers (Fig. 12). Spiny stellate neurons and interneurons are rapidly activated by thalamic afferents (Fig. 12, top T1). After a short delay, the interneurons act rapidly on local neurons, including spiny stellate neurons and other interneurons (Fig. 12, top T2) and the information is transmitted to the upper and lower layers (Fig. 12, top T3). Activity in middle layer neurons is modified by local interneurons, so that the response to periodic stimulation undergoes further temporal sculpting, as does the upper and lower layer responses (Fig. 12, top T3–T4). In the upper layers and lower layers, neurons receive input from diverse sources, including pyramidal cells of neighboring cell columns and concurrent input from the spiny stellate neurons and interneurons of the middle layers (Fig. 12, top T3–T4). This diversity of input assures that the neurons of the non-granular layers will exhibit activity that reflects input from sources other than the stellate neurons of the middle layers. While the upper and lower layer neurons exhibit periodic responses, these responses are not as narrowly tuned as those recorded for neurons in the middle layers. In MAM-treated animals, the neural elements receiving the initial thalamic drive populate all layers of the cortex (Fig. 12, bottom, T1). A spatiotemporal activation pattern is no longer evident and neuronal populations respond similarly across all layers. The initial activation is protracted compared with the normal response because the mechanisms underlying normal transfer of information are disrupted. After the initial stimulation, the neurons exhibit weak, or intermittent, entrainment to periodic stimulation (Fig. 12, T2–T4).
The authors express appreciation to Donna Tatham and Laiman Tavedi for their outstanding technical assistance. This work was supported by PHS RO1 NS24014.