Layer-Specific Properties of the Transient K Current (I_A) in Piriform Cortex

Abstract

Piriform cortex in the rat is highly susceptible to induction of epileptiform activity. Experiments in vivo and in vitro indicate that this activity originates in endopiriform nucleus (EN). In slices, EN neurons are more excitable than layer II (LII) pyramidal cells, with more positive resting potentials and lower spike thresholds. We investigated potassium currents in EN and LII to evaluate their contribution to these differences in excitability. Whole-cell currents were recorded from identified cells in brain slices. A rapidly inactivating outward current (I_A) had distinct properties in LII (I_A,LII) versus EN (I_A,EN). The peak amplitude ofI_A,EN was 45% smaller thanI_A,LII, and the kinetics of activation and inactivation was significantly slower forI_A,EN. The midpoint of steady-state inactivation was hyperpolarized by 10 mV forI_A,EN versusI_A,LII, whereas activation was similar in the two cell groups. Other voltage-dependent potassium currents were indistinguishable between EN and LII. Simulations using a compartmental model of LII cells argue that different cellular distributions ofI_A channels in EN versus LII cells cannot account for these differences. Thus, at least some of the differences are intrinsic to the channels themselves. Current-clamp simulations suggest that the differences between I_A,LIIand I_A,EN can account for the observed difference in resting potentials between the two cell groups. Simulations show that this difference in resting potential leads to longer first spike latencies in response to depolarizing stimuli. Thus, these differences in the properties of I_Acould make EN more susceptible to induction and expression of epileptiform activity.

Because of its high sensitivity to convulsant drugs and to electrical kindling in vivo (Piredda and Gale, 1985; McIntyre and Racine, 1986; Racine et al., 1988), piriform cortex has received much attention as a model system for studying temporal lobe epilepsy. In slices of piriform cortex, sustained bursting during treatments with Mg²⁺-free or low-Cl⁻ saline induces long-lasting alterations in synaptic responses. After induction, weak synaptic stimulation evokes epileptiform excitatory postsynaptic potentials (eEPSPs), which are all-or-none events corresponding to simultaneous firing of large numbers of neurons (Hoffman and Haberly, 1989). Induction of eEPSPs occurs in endopiriform nucleus (EN) and possibly in adjacent “deep” structures. When evoked, eEPSPs originate in EN and spread outward to involve the pyramidal cells in more superficial layers (Hoffman and Haberly, 1991a). These eEPSPs strongly resemble epileptiform activity induced by kindling in vivo (McIntyre and Wong, 1986;Hoffman and Haberly, 1991b; Haberly and Sutula, 1992), suggesting a common mechanism for the two phenomena. Although the reasons are not clear as to why eEPSPs arise in EN rather than the superficial layers, the recurrent excitatory connections that are prominent in EN are likely to play a role (Hoffman and Haberly, 1993).

Differences in intrinsic membrane properties also could contribute to the high seizure susceptibility of EN. Cells in EN have more depolarized resting potentials, lower spike thresholds, and higher input resistances than layer II (LII) pyramidal cells (Tseng and Haberly, 1989a,b). K currents are known to be important regulators of cell excitability (Hille, 1992) and, thus, differences in the properties of K currents could contribute to these differences in excitability. Studies using in situ hybridization have demonstrated laminar patterns in K channel expression in cortex (Drewe et al., 1992; Tsaur et al., 1992; Weiser et al., 1994), but the properties of K currents have not been compared in identified cortical cells from different laminae. We have investigated K currents in cells in EN and LII and have found differences in the properties of a rapidly inactivating K current (A current, or I_A) that could contribute to the higher susceptibility of EN to induction and expression of epileptiform activity. A preliminary account of this work has appeared in abstract form (Banks and Jackson, 1995).

MATERIALS AND METHODS

Preparation of slices. Young rats (9–18 d old) were decapitated under CO₂ narcosis, and the head immersed immediately in a slicing solution composed of 90% sucrose ACSF (artificial CSF) [composition (in mm): sucrose 250, KCl 1.8, KH₂PO₄ 1.2, NaHCO₃ 26, Mg₂SO₄ 10, glucose 10] and 10% normal ACSF [composition (in mm): NaCl 125, KCl 1.8, KH₂PO₄ 1.2, NaHCO₃ 26, Mg₂SO₄ 1, CaCl₂ 2, glucose 10] saturated with 95% O₂/5% CO₂ at 4°C. A block of tissue containing the piriform cortex was dissected out with the brain immersed in slicing solution, and the tissue glued to a vibratome tray with cyanoacrylate glue. Slices (400 μm) were cut in a plane ∼15° off the frontal plane to allow for optimal visualization of the cortical laminae and stored at room temperature in normal ACSF saturated with 95% O₂/5% CO₂. After a 30 min recovery period, a slice was transferred to the recording chamber where it was perfused at 4 ml/min with normal ACSF at 30–32°C.

Patch-clamp electrophysiology. Cells in LII and EN were visualized using an upright microscope (Reichert Jung Diastar) with Nomarski optics and a long working-distance water-immersion objective (Zeiss Achroplan 40×, 0.75 numerical aperture). Patch pipettes were fabricated from thin-walled aluminosilicate glass, fire polished, and coated with Sylgard to reduce electrode capacitance. Tight-seal whole-cell recordings were obtained using standard techniques (Hamill et al., 1981; Edwards et al., 1989). An Axopatch 200 patch-clamp amplifier was used for all recordings. Data were sampled at 5 kHz using pClamp (Axon Instruments, Foster City, CA) interfaced to an IBM-compatible PC.

Recording pipettes were filled with solution containing (in mm): KCl 130, MgATP 4, Na₂GTP 0.3, EGTA 0.4, NaCl 10, HEPES 10, pH 7.2. Series resistance and capacitance were determined by optimal cancellation of the capacitative transient, which for a “Rall motoneuron” (Rall, 1977) gives the capacitance of the cell body (Jackson, 1992). After compensation of 60–90%, series resistances were 1.45 ± 0.27 MΩ (mean ± SEM; n = 18) for recordings from LII, and 1.38 ± 0.39 MΩ (n = 22) for recordings from EN. At 0 mV, peak current often exceeded 1 nA/pF (e.g., Fig.4B). In the majority of cells, the series resistance error exceeded 10 mV for voltage commands beyond +15 mV, and data for these voltages were discarded. Average series resistance errors for pulses to −15 mV were 3.66 ± 0.76 mV for LII and 2.91 ± 0.56 mV for EN. The liquid junction potential between the pipette and bath solutions was less than 3 mV, and the data were not corrected for this voltage. Average age of the animals for the two cell groups was 11.8 ± 0.7 d for LII and 11.4 ± 0.6 d for EN.

Fig. 4.

A, B, Outward currents recorded from a cell in endopiriform nucleus. Depolarizing voltage steps to −65, −55,  … , +5 mV were applied for 200 msec (A) or 2 sec (B) from a holding potential of −90 mV. Biexponential fit to the 2 sec response to +5 mV (B,top trace) gave time constants of 376 msec (16.4% of total current) and 1530 msec (54.0% of total current). C, Activation functions estimated from the data in A andB. Early and late peak values were divided by the driving force (assuming E_K = −98 mV) and normalized using the technique described in Materials and Methods. The early peak was distinguishable only for the responses to voltage steps up to −15 mV. Voltages were corrected for series resistance errors. Boltzmann fit parameters: V_0.5 = −29.2 mV,k = 9.35 mV (early peak); V_0.5 = −16.0 mV, k = 8.08 mV (late peak).

Isolation of K currents. Potassium currents were isolated using 1 μm tetrodotoxin (TTX) to block Na currents and low-Ca²⁺ ACSF [composition (in mm): NaCl 125, KCl 1.8, KH₂PO₄ 1.2, NaHCO₃ 26, Mg₂SO₄ 2.9, CaCl₂ 0.1, glucose 10] to reduce contamination by Ca currents. In some experiments, a small residual Na current remained unblocked (e.g., Fig. 5), but was fast enough so that it did not affect the analysis of K currents. In initial experiments, Ca currents were blocked using 200 μmCd²⁺, but it was found thatI_A also was partially blocked. Blocking Ca currents by replacing all extracellular Ca²⁺ with Mg²⁺ resulted in large, rapid, and irreversible decreases in the somatic capacitance. Changes in capacitance and contamination by Ca currents were not observed with 0.1 mm extracellular Ca²⁺, and this concentration was used in the standard recording medium.

Fig. 5.

Activation and kinetics ofI_A. A,I_A currents recorded in an LII pyramidal cell in response to depolarizing voltage steps from −90 mV to −55, … , −15 mV. Currents were isolated using the protocol illustrated in Figure 1. Monoexponential fits are superimposed on the data. Time constants: −45 mV, 9.79 msec; −35 mV, 9.03 msec; −25 mV, 8.71 msec; −15 mV, 8.07 msec. B, IsolatedI_A obtained by the same protocol from a cell in EN. Exponential fits are superimposed on the data (dotted lines). Time constants: −45 mV, 11.4 msec; −35 mV, 11.5 msec; −25 mV, 12.2 msec; −15 mV, 10.5 msec. C, Pooled activation data for 11 cells in LII (triangles) and 14 cells in EN (squares). Data were obtained by normalizing the peak currents in response to the activation protocol used in Aand B, and dividing by the driving force (assumingE_K = −98 mV). Boltzmann functions were fitted to the data for each cell, then averaged to give the plotted functions (solid lines). Boltzmann fit parameters are given in Table 2. D, E, Voltage dependence ofI_A activation (D) and inactivation (E) kinetics for LII (striped bars) and EN (gray bars). Activation kinetics was measured by computing the time to reach 90% of peak (t_0.9pk). Inactivation kinetics was measured by fitting single exponentials to decaying currents, as inA and B. For D, n = 18 for EN and 13 for LII. For E, n = 17 for EN and 11 for LII. Statistical comparisons were made between LII and EN at each voltage point using the Student’s t test. Significance level was p ≤ 0.01 for all voltages.

The most reliable method for isolating I_Awas the subtraction protocol shown in Figure 1, which exploited differences in the kinetics of inactivation between the high- and low-threshold currents. The protocol used for studying activation of I_A (see Fig. 5) is illustrated, but a similar procedure also was used for studying inactivation (see Fig. 6). Currents were recorded in response to depolarizing voltage steps from a holding potential of −90 mV (Fig. 1, trace a). The depolarizing voltage steps were repeated, but preceded by 50 msec prepulses to −30 mV to inactivate I_A (Fig.1, trace b). The responses to the steps with prepulses were subtracted from the responses to the steps alone to yield the difference current (Fig. 1, trace a, b), which had an appearance typical of I_A in other preparations. In all cells studied in detail, this procedure was reliable for voltages at least up to −15 mV, and in the majority of cells to −5 mV. Beyond these voltages, noninactivating or slowly inactivating components often were apparent in the records possibly because of contamination by other outward currents.

Fig. 1.

Subtraction protocol used to isolateI_A. Shown is the procedure used to study activation of I_A. The cell was held at −90 mV and stepped to depolarized voltages (trace a; response at −15 mV). The protocol then was repeated, but with a 50 msec prepulse to −30 mV preceding the depolarizing voltage steps (trace b). The difference current yielded I_A(traces a, b). Data from this cell are shown in Figure 5A.

Fig. 6.

Steady-state inactivation ofI_A. A,I_A in a LII cell in response to the inactivation protocol. Conditioning voltage steps varying from −100 mV to −40 mV were applied for 500 msec, followed by test steps to 0 mV.I_A was isolated using a procedure similar to that illustrated in Figure 1, with the prepulse interposed between the conditioning and test pulses. B,I_A recorded from a cell in EN using the same protocol as in A. C, Steady-state inactivation functions for 10 cells in LII (triangles) and 10 cells in EN (squares). Data were obtained by normalizing the peak currents in response to the inactivation protocol used in Aand B. Boltzmann functions were fitted to the data for each cell, and then averaged to give the plotted functions (solid lines). Boltzmann fit parameters are given in Table 2.

Extracellular recordings. Bipolar tungsten electrodes (resistance 5 MΩ) were used to stimulate afferent and association fibers at the border of LIa and Ib or association fibers in LIII. Field potentials were recorded in LII or EN using patch pipettes with tip diameters of ∼10 μm and filled with 1 m NaCl (pipette resistance, 0.2–0.8 MΩ).

Biocytin labeling. Biocytin (0.3%) (Sigma, St. Louis, MO) was included in the patch pipettes to determine the morphologies of cells studied in LII and EN. The procedure for tissue processing was similar to that of Horikawa and Armstrong (1988). After terminating the recording, slices were fixed overnight (4% paraformaldehyde in 0.1 m PBS), then rinsed in PBS and cryoprotected by passage through a series of glycerol–sucrose solutions of increasing concentration. Frozen sections (60 μm) were cut using a microtome, the cryoprotection procedure reversed, and the slices rinsed in PBS. Sections were incubated for 20 min in 0.5% H₂O₂ in PBS, rinsed in PBS, and incubated overnight in the avidin–biotin–horseradish peroxidase (HRP) reagent (ABC Kit, Vector Laboratories, Burlingame, CA). The sections then were rinsed in PBS and the HRP reacted using the standard DAB-nickel/cobalt intensification method (Adams, 1981). Finally, the sections were mounted, counterstained with cresyl violet, and coverslipped. The recovery rate of stained cells was low (<50%), most likely because the stained cells frequently were near the surface of the slice and were lost during resectioning.

Data analysis. Data were leak-subtracted off-line using pClamp v6.02. Data were fitted to exponentials and Boltzmann functions ({1 + exp[(V − V_0.5)/k]}⁻¹) using pClamp or Origin v3.5 (MicroCal). Because it was not possible to depolarize the cells to maximal activation ofI_A, g_max was allowed to vary in the Boltzmann fits for these data (in addition toV_0.5 and k). Attempts were made to fit isolated A currents to functions of the form A · [1 − exp(−t/τ_activ)]ⁿ· [exp(−t/τ_inact)], but the fits to the activation phase often were frustrated by residual fast Na currents remaining in 1 μm TTX. As an alternative, activation kinetics was estimated from the time to 90% of the peak current (t_0.9pk). Inactivation was analyzed by fitting the decaying phase to a single exponential. All data are presented as mean ± SE.

Modeling. To test whether the differences betweenI_A,EN andI_A,LII could be accounted for solely by different channel distributions on the dendrites and somata of the two cell groups, a compartmental model of an LII pyramidal cell was implemented using the program Nodus v3.1 (De Schutter, 1989) on a Macintosh computer (IIsi). The model consisted of a spherical soma connected to a single, variable diameter dendrite with uniform membrane resistivity, divided into 20 compartments of equal electrotonic length (ΔL = 0.05). This scheme has been used successfully in a previous modeling study of LII pyramidal cells (Kapur, 1995). Because the latter model was developed for cells from older animals, the metric dimensions of the tapered dendrite were scaled by a factor of 0.75 for the simulations presented here, which yielded dendritic charging transients similar in size to those observed in the LII cells in our study. The soma diameter (22 μm) was adjusted to give a capacitance of 15 pF, close to the average of LII cells studied here. The passive electrical parameters used were membrane resistivity, 14.5 kΩ/cm²; axial resistivity, 100 Ω-cm; membrane capacitance, 1 μF/cm².

I_A was modeled using a Hodgkin–Huxley scheme, with the conductance g_A computed as:

$$g_A=g_{max}·m^n(V,t)·h(V,t).$$

g_max was adjusted to give currents recorded at the soma an amplitude of 265 pA/pF, for a pulse from −100 mV to 0 mV, similar to those observed in whole-cell recordings from LII cells. n was chosen to be 3, the value that gave best fits in the majority of the data with little residual Na current.m and h correspond to the activation and inactivation functions, respectively, and were computed as:

$$\frac{\text{dm}}{\text{dt}}={\alpha }_m(V)-[{\alpha }_m(V)+{\beta }_m(V)]·m(V,t),$$

$$\frac{\text{dh}}{\text{dt}}={\alpha }_h(V)-[{\alpha }_h(V)+{\beta }_h(V)]·h(V,t),$$

$${\alpha }_m={\alpha }_{m,0}·exp[\frac{{\gamma }_m}{k_m} (V-V_{0.5,m})],$$

$${\beta }_m={\beta }_{m,0}·exp[\frac{{\gamma }_m-1}{k_m} (V-V_{0.5,m})],$$

$${\alpha }_h={\alpha }_{h,0}·exp[\frac{-{\gamma }_h}{k_h} (V-V_{0.5,h})],$$

$${\beta }_h={\beta }_{h,0}·exp[\frac{1-{\gamma }_h}{k_h} (V-V_{0.5,h})].$$

The values for these parameters are given in Table1. For this form of the Hodgkin–Huxley model, the “base rates” α_m,0 = β_m,0 = [2τ_activ(V_0.5,m)]⁻¹and α_h,0 = β_h,0 = [2τ_inact(V_0.5,h)]⁻¹. Measurements for τ_activ and τ_inact were not available at these voltages, so γ_m, γ_h, and the base rates were adjusted to give reasonable values fort_pk and τ_inact in the voltage range over which these parameters were measured. The values for k_m andV_0.5,m were determined from the cube root of the measured steady-state activation curve (m_∞(V)), and the values ofk_h and V_0.5,hwere determined from the measured steady-state inactivation curve (h_∞(V)).

Table 1.

Model I_Aparameters

αm,0, βm,0	0.5 msec−1
γm	0.5
km	15.4 mV
V0.5,m	−39.3 mV
αh,0, βh,0	0.04 msec−1
γh	0.9
kh	6.86 mV
V0.5,h	−65.7 mV

Note that V_0.5,m andk_m are different from the experimental values (Table 2), because they correspond to m(V) instead of mⁿ(V).

Simulations also were run in current-clamp mode to assess the effects of the differences between I_A,LII andI_A,EN on cell excitability (see Fig. 11). For these simulations, when I_A was located in the dendrites as well as the soma, the current measured at the soma under voltage clamp had slower kinetics than in recordings from LII and EN cells and also exhibited small shifts in the activation and inactivation functions. Because we wished to assess the effects on cell excitability for different distributions ofI_A, assuming that these distributions gave rise to the currents recorded in our experiments, we changed the parameters of the underlying model in these simulations to approximate the experimental data. For example, forI_A,LII uniformly distributed in the model, it was necessary to increase α_m,0 from 0.6 to 1.28 msec⁻¹ and α_h,0 from 0.1 to 0.21 msec⁻¹ to match the kinetic parameters given in Table 2. In addition, V_0.5,mand V_0.5,h both were shifted by −3 mV to match the corresponding parameters in Table 2.

Fig. 11.

Effect of I_A on cell excitability. I_A first was placed in the soma, then was added to the first two dendritic compartments, then the first five dendritic compartments, and so on, until it was distributed uniformly throughout the cell. In each case,g_max, V_0.5,m,V_0.5,h, α_0,m, and α_0,h were adjusted to give currents measured at the soma under voltage clamp comparable to the experimental data (Table 2). For all distributions, the current decays were reasonably well fit by single exponentials. A, Effect onE_rest. The resting potential of the model was measured after setting the leak withI_A,EN to giveE_rest = −68 mV, then switching toI_A,LII with the same leak.E_rest became increasingly hyperpolarized asI_A was added to more and more dendritic compartments. The horizontal dotted lines at −68 and −74 mV mark E_rest as measured in EN and LII, respectively. B. Effect on first spike latency and interspike interval of switching from I_A,EN (solid line) to I_A,LII (dashed line) in a model with currents distributed uniformly along the dendritic tree. The model was stimulated with a 25 msec current pulse (0.3 nA) to elicit action potentials. First spike latency increased from 3.83 (EN) to 7.48 (LII) msec, whereas interspike interval increased from 9.1 (EN) to 9.35 (LII) msec. C, Effect on spike width. The traces from B were shifted by ∼3.6 msec to align the first action potentials. Spike width at −10 mV decreased from 0.66 (EN) to 0.57 (LII) msec. Spike amplitude was slightly smaller (4 mV) with I_A,LII.

Table 2.

Activation and inactivation parameters forI_A

	Endopiriform	Layer II
Activation
V0.5,m(mV)	−22.3  ± 1.5 (14)	−24.4  ± 1.8 (11)
km (mV)	11.9  ± 1.2 (14)	11.0  ± 0.8 (11)
t0.9pk(ms)	1.89  ± 0.08 (18)	1.44  ± 0.10* (13)
Ipk (pA)	130  ± 10 (10)	191  ± 25** (11)
Inactivation
V0.5,h(mV)	−65.4  ± 0.9 (10)	−55.5  ± 1.0*** (10)
kh (mV)	6.15  ± 0.47 (10)	6.40  ± 0.36 (10)
τinact(ms)	10.7  ± 1.1 (18)	6.52  ± 0.35*** (13)

Parameters are presented as mean ± SE, with the number of cells in parentheses. t_0.9,pk,I_pk, and τ_inact were measured at −15 mV. Statistical comparisons were between LII and EN using Student’s t test. *p < 0.0025; **p < 0.025; ***p < 0.0005.

The parameters for the fast, inactivating Na current (I_Na) and the delayed rectifier K current (I_KDR) were taken from the model of Traub et al. (1991) of CA3 hippocampal pyramidal neurons. The conductance density was set at 20 mS/cm² forI_Na and 10 mS/cm² forI_KDR, which were the smallest values that gave >30 mV action potentials and repetitive spikes in response to depolarizing current pulses when the currents were distributed uniformly in the model cell.

RESULTS

Cell identification

The photomicrograph of Figure 2Aillustrates that LII and EN are readily located in these slices. Healthy cells had a characteristic appearance (Fig.2B,C, arrowheads) and were typically visualized 1–2 cell diameters below the surface. Although somatic and (usually) proximal dendritic morphologies were clearly visible in the slice, identification of cell type is unreliable using these criteria. To address this issue, a portion of the cells studied were filled with biocytin to determine their morphologies. Of those stained and recovered, all of the cells in LII (8/8) were pyramidal cells (Fig.3A), whereas most of the cells in EN (4/5) had spiny dendrites (Fig. 3B,C). The fifth cell in EN was too lightly stained to determine whether it, too, had spiny dendrites. In LI and LII, as in hippocampus and neocortex, the absence of dendritic spines putatively identifies a cell as GABAergic (Ribak, 1978; Haberly, 1983; Houser et al., 1983; McCormick et al., 1985;Somogyi et al., 1985; Haberly et al., 1987), but this has yet to be shown for EN. It is likely that spiny multipolar cells in EN are glutamatergic and responsible for the recurrent positive feedback that precedes eEPSPs in EN (Hoffman and Haberly, 1991a).

Fig. 2.

A, Low-power view of a typical piriform cortex slice used in these experiments. LI, LII,LIII, and EN are indicated. LOT, Lateral olfactory tract. B, Pyramidal cells in LII in a live piriform cortex slice. Healthy cells (arrowheads) had a three-dimensional appearance, with smooth, flat surfaces and no noticeable vacuoles. C, Multipolar cells in EN in a live piriform cortex slice. Arrowheads indicate healthy cells. Scale bars: A, 500 μm; B, C, 20 μm.

Fig. 3.

A, Biocytin-filled pyramidal cell in LII of piriform cortex. Distal basal and apical dendrites were out of the plane of focus. The border between LI and LII is visible at thetop right of the picture. B, Biocytin-filled multipolar cell in endopiriform nucleus. Dendrites can be seen extending both within and orthogonal to the plane of section.C, Higher-power view of a distal dendrite from the cell inB, showing dendritic spines (arrowheads). Scale bars: A, B, 50 μm; C, 20 μm.

Properties of outward currents

Recordings were obtained from 30 cells in LII and 49 cells in EN. Average input resistance, somatic capacitance, and specific input resistance for the two cell groups were 178 ± 32 MΩ, 14.3 ± 1.2 pF, and 15.7 ± 3.8 MΩ/pF (LII); 212 ± 35 MΩ, 18.8 ± 1.2 pF, and 13.3 ± 3.1 MΩ/pF (EN). Somatic capacitance was significantly different between LII and EN (p < 0.01), but the differences in input resistance and specific input resistance were not significant (p > 0.1). The latter observation differs from that reported by Tseng and Haberly (1989b) and may be related to differences between the patch and sharp electrode techniques or the age of the animals used.

Outward currents elicited in LII and EN neurons were composed of multiple components distinguished by their threshold and rate of inactivation (Fig. 4). Depolarizing voltage steps from −90 mV elicited a low threshold, rapidly inactivating outward current (I_A), as well as higher threshold currents (Fig. 4A). In both EN and LII, the higher threshold current was composed of noninactivating and slowly inactivating components (Fig. 4B), the latter typically with biexponential decay. For the data in Figure 4A and B, approximate activation curves were derived from the early and late peak current amplitudes (Fig. 4C). For the cell illustrated, the two currents differed in their midpoints of activation by ∼13 mV.

Activation of I_A

I_A was investigated in detail in 18 cells in LII and 22 cells in EN. Depolarizing voltage steps from a holding potential of −90 mV were used to study the voltage dependence of activation and the kinetics of activation and inactivation (Fig.5). Currents from a cell in LII are shown in Figure5A, whereas Figure 5B shows data from a cell in EN. I_A was substantially larger in LII than in EN, the difference exceeding 45% at −15 mV (Fig.5A,B; Table 2). Activation functions (mⁿ_∞(V)) were estimated using peak currents divided by the driving force (E_K = −98 mV) for voltages in the range of −65 to −5 mV (Fig. 5C). No differences were observed between LII and EN for the normalized activation curve (Table 2).

Kinetics of activation and inactivation

In response to depolarizing voltage steps, currents peaked rapidly (<5 msec) and decayed back to zero within ∼50 msec (Fig.5A,B). For voltage steps in the range −55 mV to −15 mV, currents inactivated monoexponentially by >80% (dashed lines in Fig. 5A,B). Beyond this voltage, a slower kinetic component sometimes was observed, possibly reflecting contamination by other outward currents. Because this phenomenon was similar for LII and EN cells, it was not studied further.

I_A activated and inactivated more rapidly in LII cells than in EN cells. t_0.9pk was used as a measure of the activation kinetics, and the decaying currents were fitted to single exponentials to estimate τ_inact (Fig. 5A,B). At every voltage compared, both t_0.9pk and τ_inact were significantly faster in LII cells than in EN cells (Fig. 5D,E; Table 2) (significance level p < 0.01 or smaller for each voltage, using Student’s t test). Relative to cells in EN, faster activation kinetics in LII cells would result in a largerI_A in response to transient stimuli, consistent with the differences in steady-state properties discussed above, and this effect would be magnified by the larger peak current density observed in LII cells. The faster inactivation kinetics observed for LII cells would tend to counteract this effect.

Steady-state inactivation of I_A

Steady-state inactivation of I_A was studied using a standard inactivation protocol. Conditioning pulses of 500 msec to voltages ranging from −100 to −40 mV were followed by 50 msec test pulses to 0 mV. The resulting currents for a cell in LII are shown in Figure 6A, and those for a cell in EN are shown in Figure 6B. Conditioning pulses to −40 and −45 mV resulted in a small but easily discernibleI_A in LII (Fig. 6A, bottom two traces), but in EN the current was inactivated completely (Fig. 6B, bottom two traces). This is better illustrated in Figure 6C, in which the normalized peak currents are plotted as a function of voltage to give an estimate ofh_∞(V), the steady-state inactivation function for I_A. The data for LII (triangles) and EN (squares) were well fitted by single Boltzmann functions, and the midpoints of inactivation differed for the two cell groups by ∼10 mV (Fig. 6C; Table2). The slope of h_∞(V) did not differ for the two groups (p > 0.25).

To compare the steady-state current between the two cell groups, the product mⁿ_∞(V) · h_∞(V) was computed for LII and EN cells using the Boltzmann fits shown in Figures 5C and6C. This function is the steady-state open probability at a given voltage. Between −70 and 0 mV,mⁿ_∞(V) · h_∞(V) is larger for LII cells than for EN cells, the ratio exceeding 720% at −40 mV (Fig.7). However, because the function peaks closer to the resting potential (E_rest) for EN than for LII (EN: peak = −65 mV, E_rest = −68 mV; LII: peak = −53.5 mV, E_rest = −74 mV) (Tseng and Haberly, 1989b), it is slightly larger atE_rest for EN than for LII (EN, 0.010; LII, 0.0072). These results imply that for small stimuli near the resting potential, I_A will hyperpolarize the cell to a similar extent in LII and EN. However, for ongoing depolarizations (e.g., during bursting), I_A in LII cells will be larger and will tend to oppose and shorten the depolarization to a greater extent than in EN. These differences in steady-state behavior will be magnified by the larger peak current density observed in LII (Fig. 5; Table 2).

Fig. 7.

mⁿ_∞(V) · h_∞(V) in LII and EN. The product of the best-fitting activation and inactivation functions from Figures 5C and 6C are shown for LII (solid line) and EN (dashed line) to give the steady-state fraction of active I_A channels. Resting potentials are indicated (arrows) on the respective curves for LII (−74 mV) and EN (−68 mV). Resting potential values are fromTseng and Haberly (1989b).

Development of I_A and induction of eEPSPs in piriform cortex

Piriform cortex is not fully mature physiologically and anatomically until the end of the third postnatal week (Schwob and Price, 1984; Schwob et al., 1984; Moriizumi et al., 1995). In addition, neurogenesis differs between LII and EN by ∼1.5 d, with EN cells originating first (Bayer, 1986; Bayer and Altman, 1991), implying that at any given age during the first 3 postnatal weeks, cells in LII and EN could be at different developmental stages. Several studies have shown that the properties of voltage-gated channels change during development (e.g., Bader et al., 1985; Aguayo, 1989; Hamill et al., 1991; Beck et al., 1992; Spigelman et al., 1992; Costa et al., 1994). However, technical limitations necessitated the use of younger animals for these experiments. Thus, it is possible that the differences observed between LII and EN in I_A could be attributable to temporal expression patterns in cells at two different developmental ages. To address this issue, theI_A data were divided into two age groups, 10–11 d and 14–16 d, and activation and inactivation parameters computed within each group. None of the parameters tested (i.e., those appearing in Table 2) varied with age. Data for τ_inact, t_0.9pk, andV_0.5,h are shown in Figure 8. Although τ_inact decreased slightly with age in both LII and EN, the changes were not statistically significant (p > 0.1). These results suggest that the differences between the two cell types cannot be attributed to differences in developmental state.

Fig. 8.

Development of I_A in LII and EN. Shown are τ_inact (left four bars), t_0.9pk (middle four bars), and V_0.5,h (right four bars) for recordings from animals 10–11 d old (gray bars) and 14–16 d old (striped bars). Numbers of cells are given in parentheses. None of the parameters varied significantly with age (p ≥ 0.1). τ_inact and t_0.9pkwere measured at −15 mV.

Another potential concern with using young animals is the relevance to induction and expression of epileptiform activity, which has been studied previously in adult animals. Because the exact roleI_A plays in the induction process is unclear, it is difficult to address this issue directly. We found, however, that induction and expression of eEPSPs in young animals shared many similarities with the process in adult animals. Slices from animals ranging from 7 d to 20 d were bathed for 20–30 min in either 0 Mg²⁺ (n = 14) or low-Cl⁻ ACSF (NaCl replaced by Na-isethionate;n = 17). eEPSPs always were present on returning to normal ACSF (data not shown). Before and after induction, synaptic responses in LII and EN were quite similar to those recorded in adult animals (Hoffman and Haberly, 1989; 1991a), although an important difference was observed for treatments with 0 Mg²⁺, where induction was always transient, lasting on average <1 hr. In slices treated with low Cl⁻, induction was maintained for the duration of the experiment (typically >4 hr), similar to adult animals after either treatment (Hoffman and Haberly, 1989).

High-threshold K currents

Although other K currents were not studied in detail, comparisons were made between LII and EN cells with regard to activation and inactivation of the total high-threshold current, i.e., the current remaining after the inactivation of I_A was complete (Fig. 4). No effort was made to separate the different components of the high-threshold current. Figure9A shows the average current–voltage relationships for 31 cells in EN (squares) and 16 cells in LII (triangles), using the amplitude of the total outward current at the end of a 200 msec pulse. No significant differences were observed between the two cell groups. Differences also were not observed in steady-state inactivation of the slowly inactivating component of the high-threshold current, as illustrated in Figure9B (LII: n = 5; EN: n = 9). Although these two measures do not exclude the possibility of more subtle differences between LII and EN in the high-threshold components, they argue against any disparities that could substantially influence cell excitability in the two cell groups.

Fig. 9.

Properties of high-threshold K currents in LII and EN. A, Current–voltage relationships derived from responses at the end of 200 msec depolarizing voltage steps from −90 mV for cells in EN (squares; n = 31) and LII (triangles; n = 16). Current was averaged over the final 5 msec of the response. Voltages were corrected for series resistance errors and the data binned at approximately equal voltage increments. Voltage within each bin was averaged. In all cases, the horizontal SE bars were smaller than the symbol size. B, Steady-state inactivation data for the slowly inactivating K current recorded from cells in EN (n = 9) and LII (n = 5). The inactivation protocol used was similar to that illustrated in Figure 1, trace b, but the conditioning pulse duration was 10 sec, and the test pulse duration was 200 msec. Boltzmann fit parameters are LII: V_0.5 = −57.0 ± 6.8 mV, k = 11.0 ± 2.8 mV; EN: V_0.5 = −55.2 ± 3.9 mV, k = 11.2 ± 2.1 mV.

Simulations: inactivation of I_A

The voltage-clamp experiments presented here were done on cells with intact dendrites and thus under limited voltage control. Modeling current flow has been used previously to aid in the interpretation of voltage-clamp data obtained under these conditions (Muller and Lux, 1993; White et al., 1995). It is important to investigate whether the differences in I_A observed between LII and EN cells could be explained simply by different distributions ofI_A channels along the somatic and dendritic membrane. In a passive cell, voltage applied at the soma will be attenuated in the dendrites, and the degree of attenuation will increase with electrotonic distance from the soma. To illustrate the problem with an extreme case, assume that the channels underlyingI_A in EN and LII have identical properties, with the I_A channels in EN located only at the soma, and the I_A channels in LII located in a distal dendritic segment. If the voltage attenuation in this segment is 10 mV for a conditioning pulse to −40 mV, the voltage seen by the channels would be only −50 mV. Such a voltage error could potentially explain the difference observed between the two cell groups in V_0.5,h ofh_∞(V) (Fig.6C).

This issue was investigated using a simplified compartmental model of an LII cell. The goal was to determine whether placing “EN-like”I_A channels (see Materials and Methods, Table 1) on dendritic compartments would produce anI_A at the soma similar to that measured in LII cells. The implicit assumption is that the channels in EN all are located at the soma, but the results hold independent of this assumption. The simulations used an inactivation protocol similar to that used in our experiments. Voltage was applied to the soma, with 500 msec conditioning pulses to voltages ranging from −100 mV to −40 mV followed by test pulses to 0 mV. I_Achannels were placed in a single compartment, the position for which varied from the soma to the most distal dendritic compartment. A uniform distribution also was tested, consisting ofI_A channels located in every compartment. Channel density was adjusted for each configuration to give current amplitudes recorded at the soma for a conditioning pulse of −100 mV comparable to those seen in recordings from LII cells. The results are illustrated in Figure 10.

Fig. 10.

Simulated I_Afrom a compartmental model of an LII cell (see Materials and Methods, Table 1). Simulations mimicked the inactivation protocol used in Figure6. A, Response of the model withI_A confined to the somatic compartment (L = 0). B, Steady-state inactivation functions derived from the data in A (circles) and C(diamonds), as well as four otherI_A distributions. Boltzmann fit parameters (in mV; V_0.5,h, k_h) are as follows: Uniform: −63.3, 7.15; Soma(L = 0): −65.8, 6.88; L = 0.05: −64.6, 6.93; L = 0.1: −61.5, 7.14; L = 0.15: −57.3, 7.87; L = 0.2: −47.0, 12.9. C, Same asA, but with I_A confined to the third dendritic compartment (L = 0.15). Note that decays no longer appear exponential.

As expected, with I_A channels located only at L = 0 (i.e., at the soma), the model produced currents in response to the inactivation protocol similar to those recorded in EN cells (Fig. 10A; compare with Fig. 6B), witht_0.9pk = 1.60 msec and τ_inact = 10.0 msec for a conditioning pulse of −100 mV. The estimated h_∞(V) function had V_0.5,h = −65.8 mV andk_h = 6.86 mV (Fig. 10B,circles; compare with Fig. 6C,squares). Placing the I_Achannels in progressively more distal dendritic compartments yieldedhr_∞(V) functions with more depolarized values for V_0.5,h. The uniform case produced an h_∞(V) function with an intermediate value of V_0.5,h (−63.3 mV) (Fig. 10B, squares). A shift of 8.5 mV, slightly smaller than the difference between EN and LII, was produced by placingI_A at L = 0.15 (i.e., dendritic compartment 3) [Fig. 10B (diamonds), C]. However, the slope of h_∞(V) also was altered, with k_h = 7.87 mV, 14% larger than for the L = 0 case. The kinetics ofI_A recorded at the soma was distorted substantially by placing the I_A channels in dendritic compartments. For the currents shown in Figure10C, t_0.9pk = 2.5 msec, 56% longer than in the L = 0 case. The decay ofI_A was no longer exponential for the conditioning pulse of −100 mV, but the time to decay 63% from peak was 31.8 msec, 318% slower than for L = 0. The slower kinetics associated with a depolarizing shift inh_∞(V) is in contrast with the data recorded in LII and EN; a depolarizing shift inh_∞(V) in LII relative to EN was associated with faster activation and inactivation kinetics (Figs.5D,E, 6C). Furthermore, our data always displayed exponential decays, in contrast to the simulation in Figure 10C.

Simulations: activation of I_A

We also performed simulations using an activation protocol similar to that used for Figure 5, withmⁿ_∞(V) estimated from the normalized peak conductance. For I_A located at the soma (L = 0), the currents produced were similar to those recorded in EN (data not shown) (V_0.5,m = −20.4 mV; k_m = 10.7 mV). The Boltzmann parameters differed slightly from the experimental data (compare with Table 2) because of the limitation of using the peak current as an estimate of maximal activation. Although in our recordingsmⁿ_∞(V) did not differ between EN and LII (Fig. 5C; Table 2), moving theI_A channels to more distal locations in the model altered both V_0.5,m andk_m. Interestingly, the shift inmⁿ_∞(V) was in the hyperpolarizing direction, contrary to the case forh_∞(V). For example, withI_A located at L = 0.15, V_0.5,m = −26.8 mV, andk_m = 13.5 mV. This hyperpolarizing shift occurred because in a partially clamped dendritic segment, activation of I_A drove the voltage towardE_K (−98 mV) and away from the command voltage at the soma.

Thus, both inactivation and activation parameters were altered by placing I_A in dendritic compartments, but the changes were inconsistent with our recordings from LII and EN. For this model, it was not possible to produce currents similar toI_A,LII using a model based onI_A,EN, arguing strongly that the differences observed between EN and LII were not solely attributable to different spatial distributions of channels in the two cell groups.

Simulations: role of the high-threshold K current

Another issue investigated with the model was whether an unequal distribution of other voltage-activated K channels, in addition toI_A, could contribute to the differences between I_A,EN andI_A,LII. We performed simulations with bothI_A and the high-threshold K current shown in Figure 9 located in the same compartments. The subtraction protocol illustrated in Figure 1 was used to isolateI_A. With both currents located at L = 0.15, I_A never had exponential decay kinetics, contrary to what we observed in LII, and the resulting currents often were inward even for small depolarizing steps (data not shown). The reason for the latter observation is similar to that for the hyperpolarizing shift inmⁿ_∞(V). Activation of bothI_A and the high-threshold K current drives the membrane potential further from the command potential than activation of the high-threshold current alone (because the total current is larger) and, thus, results in a smaller high-threshold current. When the subtraction protocol is used to isolateI_A, the resulting current is inward. These results indicate that other K currents located in the dendrites cannot account for the differences we found betweenI_A,EN andI_A,LII.

Simulations: I_A and cell excitability

To assess the contribution of I_A to cell excitability in LII and EN, we also ran simulations with the model in current-clamp mode. Cells in EN and LII are known to have to have different resting potentials (EN: E_rest = −68 mV; LII: E_rest = −74 mV) (Tseng and Haberly, 1989b), and we first determined whether the differences between I_A,EN andI_A,LII could account for this observation. For each channel distribution, simulations initially were run usingI_A,EN, and the leak adjusted to setE_rest = −68 mV.I_A,LII then was substituted forI_A,EN and the model allowed to reach a steady resting potential. Switching toI_A,LII always resulted in a hyperpolarizing shift in the resting potential, but the magnitude of the shift was highly dependent on the channel distribution (Fig.11A). With channels located only at the soma (i.e., no dendritic compartments with I_A), the difference in resting potential solely attributable to switching from I_A,EN toI_A,LII was ∼0.5 mV, <10% of the difference in E_rest between EN and LII. However, with channels distributed uniformly throughout the model cell (i.e., I_A in the soma and in all 20 dendritic compartments), the difference in resting potential was >6 mV, thus accounting for the difference observed inE_rest between LII and EN. We note that a uniform distribution of I_A altered only slightly the simulated h_∞(V) function (Fig. 10B) and gating kinetics (data not shown). Thus, the uniform distribution is consistent with our voltage-clamp data and will be used further in evaluating the effect of channel differences on excitability.

The previous simulations show that the differences betweenI_A,EN andI_A,LII could have substantial effects on the membrane potential of these cells in the steady state. BecauseI_A is a rapid, inactivating current, its role usually is associated with more dynamic situations such as action potential repolarization, and we assessed howI_A,EN andI_A,LII might differentially affect spike width, latency, and interval during excitatory stimuli. There is little information available on the properties of other voltage- and Ca²⁺-dependent ion channels present in these two cell types. Therefore, we constructed the simplest model that was able to generate trains of action potentials in response to current pulses, using a fast, inactivating Na current (I_Na) and a delayed rectifier K current (I_KDR) from a model of CA3 hippocampal pyramidal neurons (Traub, 1991) (see Materials and Methods). With the I_A,EN,I_Na, and I_KDRdistributed uniformly throughout the model cell, the resting potential was set to −68 mV by adjusting the leak conductance, and then the model stimulated with a 25 msec current pulse (0.3 nA) (Fig.11B, solid line).I_A,LII then was substituted forI_A,EN, the model was allowed to reach its new resting potential, and the current stimulation was repeated (Fig.11B, dashed line). The largest effect was on the first spike latency, which increased from 3.8 msec (EN) to 7.5 msec (LII) because of the difference in E_rest. Interestingly, the interspike interval was increased by <3% when switching from I_A,EN toI_A,LII, and the effect on action potential width also was modest, with I_A,EN resulting in a 17% increase in duration (width at half amplitude) relative toI_A,LII (Fig. 11C).

These current-clamp simulations suggest that the presence ofI_A,LII versusI_A,EN will have the strongest effect when the channels are distributed widely in the cell. The primary influence on cell excitability is via steady-state behaviors such as the resting potential. We wish to emphasize, however, that no attempt was made to investigate the model dependence of the simulations onI_Na and I_KDR, and the effects on action potential parameters should be interpreted only qualitatively until more complete biophysical descriptions can be made of these two cell types.

DISCUSSION

We have observed differences in I_Abetween cells in LII and EN in slices of piriform cortex. Biocytin labeling in a subset of these experiments confirms that most, if not all, of the recordings came from superficial pyramidal cells in LII and deep, spiny multipolar cells in EN. Because EN is the site of origin for eEPSPs, the observed layer-specific properties ofI_A may be functionally relevant to epileptogenesis in piriform cortex. I_A,ENhas a smaller peak amplitude than I_A,LII, and its steady-state inactivation function is shifted to hyperpolarized potentials. These differences combine to make the steady-stateI_A,EN smaller in amplitude at any given voltage, and also make it less available to counteract transient depolarizing stimuli. Both of these effects would contribute to the enhanced excitability in deep cells described in previous studies on piriform cortex (Tseng and Haberly, 1989b). Simulations with a simplified action potential generator show that the differences in kinetics may be less important, because the effects of slower activation tend to counteract the effects of slower inactivation. The result with τ_inact (Fig. 5E) was unexpected within the framework of the Hodgkin–Huxley model, in which a hyperpolarizing shift in steady-state inactivation for EN (Fig. 6) would be accompanied by faster inactivation kinetics, not slower as observed here. This suggests either that recovery from inactivation is much slower in EN than in LII or that the process of inactivation is not a two-state process, as in the Hodgkin–Huxley model.

Comparisons to other systems

I_A has been found in numerous cell types in nearly all invertebrates and vertebrates studied (Rogawski, 1985), including most cells in hippocampus and neocortex. Steady-state inactivation of I_A,LII was similar to that observed in neocortical pyramidal cells, in whichV_0.5,h ranged from approximately −50 to −60 mV (Spain et al., 1991; Foehring and Surmeier, 1993), but was different from that observed in CA1 pyramidal neurons, in whichV_0.5,h was more negative than −80 mV (Numann et al., 1987; Ficker and Heinemann, 1992; Wu and Barish, 1992). The value for V_0.5,h for I_A,EN(−65.2 mV) was similar to that reported for dentate gyrus neurons (−65 mV) (Beck et al., 1992). Conversely, the kinetics of inactivation was slower in both LII and EN neurons than in neocortical cells (Spain et al., 1991; Foehring and Surmeier, 1993), but similar to hippocampal (Numann et al., 1987; Wu and Barish, 1992), thalamic (Huguenard et al., 1991), and sensory neurons (McFarlane and Cooper, 1991), when differences in the temperature at which the experiments were performed are taken into account.

Previous studies have provided evidence for layer-specific differences in the membrane properties of cortical neurons under current clamp (McCormick et al., 1985; Kawaguchi, 1993, 1995; Connors, 1994), but there have been very few studies of voltage-gated channels in identified cortical cells. In isolated neocortical cells, Hamill et al. (1991) found differences between pyramidal and nonpyramidal neurons in sodium, potassium, and calcium currents. Interestingly, the most striking difference observed was that nonpyramidal cells did not express I_A. Thus, variations inI_A may constitute an important mechanism for differential regulation of membrane excitability in cortical neurons.

Space-clamp issues

Immunohistochemical and in situ hybridization studies of the intracellular distribution of I_Achannels cloned from mammalian brain suggest that the channels can be localized to axonal and dendritic processes (Sheng et al., 1992;Maletic-Savatic et al., 1995), regions that are under variable voltage-clamp control in our recordings. Compounding this problem,I_A has relatively fast kinetics, and would be subject to substantial distortion attributable to dendritic filtering in regions electrically remote from the soma (Rall, 1977). For these reasons, the quantitative analysis presented here may not reflect accurately the properties of I_Achannels in the two cell types. However, the differences observed strongly imply that there are differences in the channels underlyingI_A,LII andI_A,EN. Simulations showed that the observed properties were inconsistent with these differences being solely attributable to differences in space-clamp errors in the two cell types (Fig. 10). Although moving the I_A channels to more distal locations in the model resulted in a depolarizing shift in the steady-state inactivation curve, as observed forI_A,LII relative toI_A,EN, the shift was associated with slower kinetics and nonexponential current decays, as well changes in the activation function mⁿ_∞(V). This stands in contrast to what we observed in LII neurons. We cannot exclude the possibility that differences in channel distribution also contributed to the observed differences inI_A. For example, ifI_A,LII had even faster activation and inactivation kinetics than we observed, but was located distally, a 10 mV difference in steady-state inactivation might be possible. Under these conditions, however, we would not expect the recorded currents to have the observed exponential inactivation kinetics (Fig.10C)

Molecular basis of channel differences

The differences between I_A,LII andI_A,EN could arise from many possible sources, including subunit composition and post-translational modification. At least three α subunit mRNAs have been cloned from mammalian brain that encode fast-inactivating K channels when expressed as homomultimers (Stuhmer et al., 1989; Tseng-Crank et al., 1990;Baldwin et al., 1991; Pak et al., 1991; Schroter et al., 1991). Heteromultimeric combinations of α subunits within K-channel families can form fast-inactivating channels with properties distinct from their corresponding homomultimeric components (Isacoff et al., 1990;Ruppersberg et al., 1990). Coexpression of α and β subunits also can dramatically alter the inactivation properties of K channels (Rettig et al., 1994; McCormack et al., 1995; Morales et al., 1995). Potential mechanisms of post-translational modification include phosphorylation (Walsh and Kaas, 1988; Augustine and Bezanilla, 1990;Moran et al., 1991; Covarrubias et al., 1994), cysteine oxidation (Ruppersberg et al., 1991), and direct modulation by lipid metabolites (Ordway et al., 1989; Bowlby and Levitan, 1995), all of which have been shown to modulate K channels in other systems. Additional experiments on the laminar distribution of K channels in piriform cortex and electrophysiological studies using patches pulled from cells in EN and LII would help to distinguish among these different mechanisms.

Implications for circuit behavior

The properties of I_A,LII andI_A,EN could have relevance to epileptogenesis in piriform cortex, because intrinsic membrane properties are believed to play a role in determining epileptic foci elsewhere in the brain (Connors, 1994). For example, burst firing is more pronounced in hippocampal cells in regions prone to generating epileptiform behavior (Jefferys, 1990), and alterations in intrinsic excitability may contribute to the development of epileptiform behavior in neocortex and thalamus in vivo and in vitro(Mody et al., 1990; Vreugdenhil and Wadman, 1992; Prince and Tseng, 1993; Barkai et al., 1994; Tsakiridou et al., 1995). Although no substantial differences were observed in I_Abefore and after induction of eEPSPs in piriform cortex (M. Banks, L. Haberly, and M. Jackson, unpublished observations), differences in the properties of I_A,LII andI_A,EN could contribute to the susceptibility of EN to induction. I_A has been shown to play prominent roles in regulating excitability in other systems, from invertebrates to mammalian central neurons (Connors and Stevens, 1971a,b; Neher, 1971; Daut, 1973; Segal et al., 1984; Kaang et al., 1992). Particularly relevant are studies in sympathetic ganglia of guinea pigs, in which differences in steady-state inactivation ofI_A were shown to contribute to differences in the firing properties and synaptic responses of phasic and tonic cells (Cassell and McLachlan, 1986; Cassell et al., 1986). In tonic cells, I_A is more prominent at rest, allowing for regular, repetitive firing in response to sustained stimuli. I_A also shortens excitatory synaptic potentials significantly, reducing the excitability of tonic neurons in response to synaptic stimuli.

Our simulation results indicate that the primary effects of the differences in the steady-state properties ofI_A (Fig. 6C) and the smaller current density observed in EN (Table 2) are via alterations in steady-state behavior (e.g., resting potential). Although not tested with the model, we also would expect larger and more sustained steady-state depolarizations in EN relative to LII, for example, during the bursting behavior that induces eEPSPs. Induction of eEPSPs is dependent on Ca²⁺ influx through NMDA receptors (Hoffman and Haberly, 1989), and depolarization enhances this influx because of the intrinsic voltage dependence of NMDA receptors. Thus, the differences between LII and EN in the density and steady-state inactivation of I_A would result in greater influxes of Ca²⁺ in EN cells than in LII cells during bursting. If intracellular [Ca²⁺] must exceed some threshold level for induction to occur, then EN could be more susceptible to induction than LII because of a less prominentI_A at depolarized voltages. Expression of eEPSPs also would be enhanced in EN, in whichI_A,EN would be less able to counter the reverberative excitation that immediately precedes eEPSPs at their site of origin. The more depolarized resting potentials in EN cells, in part attributable to I_A,EN, would allow for more rapid responses to excitatory input, as reflected in the shorter latencies observed in our model (Fig. 11B). These results suggest that biophysical differences between cell types may contribute to layer-specific plasticity in cortex.