Interaction of Selective Amino Acid Residues of K Ca Channels with Carbon Monoxide

Abstract

The activation of big-conductance K_Ca channels in vascular smooth muscle cells by carbon monoxide (CO) has been demonstrated previously. One specific target of CO on K_Ca channel proteins is the histidine residue. The roles of other amino acid residues on the functionality of K_Ca channels, as well as their reactions to CO, have been unclear. In the present study, the cell-free single channel recording technique was used to investigate the chemical modification of K_Ca channels by CO and other chemical agents. The modification of negatively charged carboxyl groups and the ε-amino group of lysine did not affect the open probability, but decreased single-channel conductance of K_Ca channels. When sulfhydryl groups of cysteine were modified with N-ethylmaleimide, the open probability of K_Ca channels was decreased, but single-channel conductance was not affected. None of the above chemical modifications affected the CO-induced increase in the open probability of K_Ca channels. However, N-ethylmaleimide treatment reduced the stimulatory effect of nitric oxide (NO) on K_Ca channels. Finally, pretreatment of smooth muscle cells with NO abolished the effects of subsequently applied CO on K_Ca channel proteins. Our study demonstrates that CO and NO acted on different amino acid residues of K_Ca channel proteins. The interaction of CO and NO determines the functional status of K_Ca channels in vascular smooth muscle cells

The physiological importance of carbon monoxide (CO) on the regulation of vascular tone has been generally acknowledged (1, 2). The vasorelaxation induced by CO could result from the activation of the sGC/cGMP pathway and/or the opening of the big-conductance calcium-activated K⁺ (K_Ca) channels in vascular smooth muscle cells (SMCs)(3–5). CO hyperpolarized vascular smooth muscle cell (SMC) membranes and increased the open probability of K_Ca channels (4, 5). This stimulatory effect of CO on K_Ca channels in SMCs was not mediated by intracellular second messenger systems because the absence of the activation of the cGMP pathway (6) and the presence of GTP-γ-S (4) do not interfere with the effect of CO on single K_Ca channel currents recorded from cell-free membrane patches.

As a gasotransmitter, the direct interaction of CO and K_Ca channels has been a mystery. Big-conductance K_Ca channels are stimulated by increased intracellular calcium concentration and by membrane depolarization. Although the primary structure of K_Ca channels in vascular SMCs is known, the contributions of various amino acids to the gating and conducting of K_Ca channels are not yet clear. The modification of one or more amino acid residues may suffice to change the conductance and/or the open probability of K_Ca channels (3, 7–9). Our previous study demonstrated that CO induced a chemical modification of histidine residue located on the extracellular surface of the transmembrane K_Ca channel proteins in vascular SMCs (3). The interaction of CO with other amino acid residues of K_Ca channel proteins has not been delineated.

Vascular tone is under the influence of the family of gasotransmitters, including CO and nitric oxide (NO)(10). Interestingly, NO also stimulated big-conductance K_Ca channels in vascular SMCs (7). The interaction of NO and CO on K_Ca channels in vascular SMCs likely affects the integrated control of vascular tone. Although this interaction may have important physiological implications, the modulation of K_Ca channels by simultaneous treatment with CO and NO has not been studied.

Three objectives were pursued in the present study. First of all, the roles of different amino acid residues in the gating of big-conductance K_Ca channels in vascular SMCs were examined. Second, the interactions of CO with different amino acid residues of K_Ca channel proteins were determined. Third, the interaction of CO and NO on K_Ca channel functions was investigated. Our results showed that cysteine residues participated in the gating while carboxyl groups and lysine residues were involved in the permeation of K_Ca channels. The interaction of CO and NO determines the functional status of K_Ca channels in vascular SMCs.

Materials and Methods

Preparation of SMCs and Recording of Single K_Ca Channel Currents.

Single SMCs from rat tail artery were isolated and identified as described previously (11). Dispersed cells were plated in 35-mm Petri dishes and cultured in Dulbecco’s modified Eagle’s medium containing 10% fetal calf serum in a CO₂ incubator at 37°C. The cells were used 8–36 hr after isolation (11). The inside-out and outside-out configurations of the patch-clamp technique were used to record single K⁺ channel currents. Pipettes with a resistance of 6-8 MΩ were used and the seal resistance was usually greater than 10 GΩ. Membrane patches with no more than three channels were used for these experiments. Single-channel currents were filtered at 2 KHz (8-pole Bessel, -3 dB) and recorded with a 5-μs sampling interval in a gap-free mode. For each concentration of tested agents, at least 60 s of channel activity was directly recorded on the hard disk of a computer. The open probability (NPo), i.e., the fraction of time when the channels stay open within the total observation period with N representing the number of single channels in one patch (3, 12) and single channel conductance were determined from an all-point amplitude histogram using Fetchan and Pstat programs (Axon Instruments, Inc., Union City, CA). The outside surface of membrane patches was bathed in a solution containing KCl 145 mM, HEPES 10 mM, and glucose 10 mM. The inside surface of membrane patches was exposed to a solution containing KCl 145 mM, HEPES 10 mM, MgCl₂ 1.2 mM, glucose 10 mM, EGTA 1 mM, and different amounts of CaCl₂ to reach the desired final free Ca²⁺ concentrations. Unless otherwise specified, [Ca²⁺]_i was 0.5 μM throughout all the experiments.

Chemicals and Data Process.

To prepare the CO solution, 20 ml of stock solution in a sealed glass tube was bubbled with a stream of CO (Canadian Liquid Air Ltd., Oshawa, Ontario, Canada) for 20 min under the pressure of 100 kPa at 37°C (12, 13). One microliter of this CO-saturated solution contained 30 ng of the gas. The stock solution of CO was freshly prepared before each experiment, and then immediately diluted to the desired concentration with the bath solution.

Immediately before the experiment, solid trimethyloxonium (TMO, final concentration of 50 mM) was added to the bath solution (pH 7.0 with 1 M NaOH). The external or internal sides of the membrane patches were superfused with this reaction solution for 30 sec to 1 min. The membrane patches were then washed extensively with a regular bath solution for 5 min before electrical recording began. For the experiment with trinitrobenzenesulfonic acid (TNBS), the membrane patches were superfused with a bath solution containing TNBS (3 mM) for 6 min at pH 9. After that, the patches were re-exposed to a regular bath solution with normal pH 7.4. N-ethylmaleimide (NEM) was dissolved in 1 M KOH as a 500 mM stock solution and dissolved in the bath solution immediately prior to the experiment and the pH was readjusted to 7.4 (14).

In all chemical modification experiments, the pH of reaction solutions was adjusted with the hydroxide of the major cation. Each experiment was bracketed by a control conducted at the same pH. A short period of exposure of the membrane patches to various pH values, such as those used in chemical modification experiments, did not yield significant changes in the behavior of K_Ca channels. Unless specified, chemicals were obtained from Sigma Chemical Company (St. Louis, MO). Osmolalities of all recording solutions were adjusted to 290 mOsm and 7.4, respectively. All electrophysiological experiments were conducted at room temperature.

The data were expressed as mean ± SEM and analyzed using Student’s t test or analysis of variance in conjunction with the Newman–Keuls test where applicable. Group differences were considered statistically significant at the level of P < 0.05.

Results

In the present study, we examined the chemical modification of K_Ca channels in vascular SMCs. Three general classes of reagents were used: (i) those that act on negatively charged carboxyl groups, such as TMO; (ii) those that act on positively charged lysine, such as TNBS; and (iii) those that act on sulfhydryl groups of cysteinyl residues, such as NEM.

Effect of TMO.

Extracellular application of TMO (50 mM) reduced the amplitude of single K_Ca channel currents (outside-out patch) so that all the channel events following TMO treatment were of a smaller amplitude (Fig. 1A). This decrease in unit current amplitude was reflected from all-point amplitude histograms derived from 5 min continuous records before and after TMO treatment, respectively. The single channel I–V relationship for this experiment revealed that the slope conductance of K_Ca channels was reduced from 254 ± 6 pS to 171 ± 7 pS after TMO treatment (n = 3). In contrast, intracellular application of TMO (inside-out patch) did not cause significant changes in the conducting (Fig. 1B ) or gating (open probability, data not shown) behaviors of K_Ca channels. ChTX inhibited K_Ca channel activity by decreasing the open probability of single channels (Fig. 2A ). Extracellularly applied TMO had no effect on the open probability of single K_Ca channels. However, the presence of TMO significantly decreased the blocking effect of ChTX on K_Ca channels (Fig. 2B ), which is consistent with the previous report by MacKinnon and Miller (9). Moreover, the extracellular application of TMO in the same membrane patches (n = 3) did not alter CO-induced increases in open probability of K_Ca channels (Fig. 2C).

Effect of NEM.

Sulfhydryl groups of K_Ca channel proteins were modified with NEM. To better reveal the effect of NEM (5 mM) on single K_Ca channels, the calcium concentration in the solution bathing the inner side of the membrane patch was increased from 0.5 to 5 μM in the experiment shown in Figure 3 . Treatment of inside-out patches with NEM (5 mM) led to a significant decrease in the open frequency of single K_Ca channels by 72 ± 3% (n = 5, P < 0.01). Consequently, NPo of single K_Ca channels was significantly reduced and multiple openings of K_Ca channels were closed (Fig. 3B and C ). In agreement with previous studies on vascular SMCs (8), NEM did not affect the conductance of K_Ca channels in rat tail artery SMCs (not shown). Lastly, the presence of NEM on the extracellular (not shown) or intracellular side (Fig. 4A) did not alter the effect of CO on the open probability of single K_Ca channels.

Effect of TNBS.

Pretreatment with TNBS (3 mM) for 5 min did not alter the characteristics of K_Ca channels recorded from four inside-out membrane patches isolated from rat tail artery SMCs. It has been reported that the modification of lysine by TNBS was not reversible at least one hour after washout of TNBS (15). Therefore, we subsequently applied CO to the same patch. After treatment of K_Ca channels with TNBS, CO (30 μM) still effectively increased the K_Ca channel activity (Fig. 4B). This result was repeated in two additional inside-out membrane patches.

Interaction of CO and NO on K_Ca Channels.

Albeit acting on individual subunits of K_Ca channels, NO and CO together presented an integrated effect on the activity of K_Ca channels in rat tail artery SMCs. After the cell membranes were pretreated with SNP for 10 min, the subsequent application of CO was not able to further increase the NPo of K_Ca channels. The NPo were 0.32 ± 0.05 and 0.30 ± 0.04 before and after the application of CO in the presence of SNP (n = 5, P > 0.05). However, pretreatment of cell membranes with CO for 10 min did not prevent SNP from further increasing the NPo. The mean NPo of K_Ca channels were 0.25 ± 0.04 and 0.53 ± 0.05, before and after the application of SNP in the presence of CO (n = 5, P < 0.05).

Discussion

In the present study, we selectively modified certain amino acid residues of K_Ca channel proteins to probe the relationship between structure and function of these channels in rat tail artery SMCs. Our results provide evidence for the presence of essential sulfhydryl (possibly cysteine), basic amino acids (possibly lysine), and carboxyl groups associated with the normal function of K_Ca channels in rat tail artery SMCs. The gating mechanism of K_Ca channels in this cell preparation is under the influence of multiple amino acid residues, including histidine and cysteine, although CO may only act on histidine residues (3). The carboxylic acid moieties may actively participate in the control of ion permeation. The topography of these amino acid residues is also important in terms of the functioning behavior of K_Ca channels.

Modification of ion channels with TMO has been used to study the functional role of carboxylic acid groups in both voltage-dependent and ligand-gated channels. TMO is a highly reactive agent that specifically esterifies carboxyl groups, such as those contained in aspartic or glutamic acid residues (16). This reaction converts a normally negatively charged hydrophilic group to neutral methyl esters, a more hydrophobic residue. It has been reported that TMO decreased the whole-cell acetylcholine-activated channel current by 60% and the cell-attached single-channel conductance by 10% in a muscle cell line (16). TMO modification also reduced the single-channel conductance of voltage-gated Na channels (17, 18) and K_Ca channels (9). It is worth noting that the hydrolization of TMO is very rapid and the accurate concentration of TMO reaching the cell membrane in our experiments is not known, although TMO powder was added to the reaction solution immediately before superfusing the membrane patches. However, there is little doubt that TMO modified K_Ca channels in rat tail artery SMCs. TMO decreased single-channel conductance, but had no effect on the open probability of K_Ca channels in our study. In contrast, CO was shown to increase the open probability, but had no effect on the conductance of K_Ca channels. Moreover, our preliminary experiments showed that CO increased the calcium sensitivity of K_Ca channels (4), but TMO did not (9). Our present results show that only externally applied TMO (outside-out patch) affected K_Ca channels, suggesting that the K_Ca channel has carboxyl groups on its external surface and that those carboxyl groups may have a profound influence on channel conductance.

We also found that the effect of CO on K_Ca channels was not influenced by TMO pretreatment, but that of ChTX was inhibited. Our control experiments showed that both ChTX and CO decreased the open probability of K_Ca channels. Why does TMO have different effects on the changes of the gating mechanisms of K_Ca channels induced by ChTX and CO? The answer may reside in the different actions of these two molecules. ChTX is a highly basic peptide with eight positively charged residues (four lysines, three arginines, and one histidine), but only two negatively charged groups. ChTX blocks K_Ca channels by physically occluding the pore and preventing ion conduction (19). However, CO may act on K_Ca channels by chemically modifying the channel protein. Carboxyl groups themselves may not be involved in the gating mechanisms of K_Ca channels, as indicated from the direct effect of TMO on K_Ca channels. The modification of carboxyl groups by TMO, however, may change the electrostatic forces, and thus vacillating the affinity of ChTX to K_Ca channels.

TNBS neutralizes the peptide terminal amino groups and the ε-amino group of lysine (20). In giant axons of squid, externally applied TNBS modified delayed rectifier K channels by slowing down the kinetics of macroscopic ionic currents (voltage-clamp, microelectrode); increasing the size of currents at large positive voltages; shifting the voltage-dependent probability of channel openings at more positive potentials; and altering several properties of K channel gating currents (15). TNBS (1 mM) also produced a Mg²⁺-dependent irreversible inhibition of K_ATP channel activity in inside-out patches from an insulin-secreting cell line. Radioligand binding studies showed that TNBS completely inhibited the binding of ³H-glibenclamide (14). In contrast, internal application of TNBS (0.1 to 1 mM) irreversibly increased K_Ca channel open probability in inside-out patches from GH₃ cells without affecting single-channel conductance and voltage sensitivity. The increase in open probability is predominantly due to the loss of long-duration closures of the channel; however, the lengths of long-duration openings are increased (8). In our study, internally applied TNBS did not affect the permeation or open probability of K_Ca channels in rat tail artery SMCs. The presence of TNBS also did not interfere with the effect of CO on K_Ca channels. The discrepancy between our results and the aforementioned could be partially due to the differences in the species and cell types, since even the similar big-conductance K_Ca channels from different species may possess different molecular structures. That only one concentration of TNBS (3 mM) was tested in our study could be another explanation for the ineffectiveness of the reagent to modify K_Ca channels. We also acknowledge that the effect of extracellularly applied TNBS on single K_Ca channels in rat tail artery SMCs was not further investigated. Nevertheless, our results at least suggest that, under the present experimental conditions, the modification of K_Ca channels by CO is unlikely due to the modification of internally located lysine residues of the channel protein.

Sulfhydryl groups of cysteinyl residues of peptides and proteins are generally the most reactive of all amino acid side chain functionalities under normal physiological conditions. Among the irreversible modifying agents for the sulfydryl group, NEM has been used mostly (21). We found that NEM (5 mM) applied to excised membrane patches for 2 min produced a 3-fold decrease in channel activity, suggesting that covalent modification can alter normal K_Ca channel activity. Our study, thus, provides evidence for the structural or functional involvement of cysteines (sulfhydryl group) in the K_Ca channels of rat tail artery SMCs, which is consistent with many reports in literature (7, 14, 22). For example, NEM included in the pipette solution was found to decrease mean open time by 15-fold and slightly reduced single-channel conductance of single acetylcholine receptor channels of BC3H-11 cells in the ‘cell-attached’ configuration (22). Lee K et al. (14) showed that NEM produced an irreversible inhibition of K_ATP channel activity when applied to the intracellular surface of excised inside-out patches. Bolotina et al. (7) reported that the pretreatment of the membrane patches with NEM prevented NO-induced activation of K_Ca channels. Our results showed that the presence of NEM did not modify the effect of CO, but blocked the effect of SNP, on K_Ca channels. Therefore, the modification of K_Ca channels by NO or CO may have quite different mechanisms. One gas may act on cysteine residues whereas the other does not.

It is also well known that ionic strength and pH values of reaction solutions are important for the effect of chemical reagents on ion channels (9, 16, 23). However, in the present study, all reaction solutions have normal ionic strength and no comparison of the effects of different pH values was made for individual chemical reagents. This is because the ultimate goal to use chemical reagents in our study was to modify certain amino acid residues, thus facilitating the identification of reactive sites on K_Ca channels for CO. Although the identification of various factors that influence the chemical modification of K_Ca channels is important, the most we need to do for the present study is to make sure that the desired amino acid residues have been modified. We believe that this goal has been achieved. With the current experimental conditions (ionic strength, pH, temperature and reagent concentrations), the modification of K_Ca channels by different reagents are obvious and relatively specific. In future studies, the ionic strength and pH of reaction solutions will be taken into account to better characterize the effect of CO and other chemical reagents.

Figure 1.

Effect of TMO on the conductance of single K_Ca channels in rat tail artery SMCs. Solid lines on the right of the current traces denote the close state of the channels. For the clarity of presentation, the baseline counts (close state of channels) were omitted from all-point amplitude histograms. (A) TMO reduced the amplitude of single-channel currents recorded in an outside-out patch held at a membrane potential of −20 mV. Before and after the application of TMO, single-channel current amplitude was decreased from 7.2 ± 0.003 pA to 4.8 ± 0.006 pA, calculated as the mean± SEM of the Marquardt least square fits (Gaussian fit) to the histograms. The single-channel conductance was decreased from 260 pS to 170 pS in the presence of TMO (right panel). (B) TMO had no effect on the amplitude of single-channel currents recorded in an inside-out patch held at a membrane potential of −20 mV. The single-channel conductance was also not affected by TMO treatment (right panel). All-point amplitude histograms resulting from 5 min of digitized records were shown just beneath single-channel current records in (A) and (B).

Figure 2.

Effect of TMO on the open probability of K_Ca channels in outside-out membrane patches held at a membrane potential of −30 mV. (A) ChTX (100 nM) reduced the open probability of K_Ca channels. Similar results were obtained in the other three patches. (B) TMO alone had no effect on the open probability of K_Ca channel. However, the ChTX (100 nM)-induced reduction of the open probability of K_Ca channels was decreased after TMO modification. Similar results were obtained in the other three patches. (C) The CO-induced increase in the open probability of K_Ca channels was not affected by TMO modification. Solid lines on the right of the current traces denote the close state of the channels.

Figure 3.

Effect of NEM (5 mM) on single K_Ca channels recorded from inside-out membrane patches held at a membrane potential of −30 mV. The calcium concentration in the bath solution in this set of experiments was 5 μM. (A) Original K_Ca current traces recorded in the absence and then the presence of NEM. Solid lines on the right of the current traces denote the close state of the channels. (B) The effect of NEM on the dwell-time of K_Ca channels. (C) The effect of NEM on the current amplitude of single K_Ca channels. For clarity of presentation, the baseline counts (close state of channels) were omitted from all-point amplitude histograms.

Figure 4.

The interaction of NEM and TNBS with CO on single K_Ca channels. Each group of traces represents typical recordings from two to five membrane patches with respective treatments and 30 μM of CO. (A) NEM (5 mM, inside-out patch, +30 mV) did not modify the effect of CO on single K_Ca channel currents. (B) The effect of TNBS (3 mM, inside-out patch, +20 mV) on single K_Ca channel currents recorded from the inside-out membrane patch held at a membrane potential of +20 mV.

Footnotes

This work was supported by research grants from Canadian Institutes of Health Research to R.W. and L.W.

1

To whom all requests for reprints should be addressed at Department of Physiology, University of Saskatchewan Saskatoon, SK, Canada S7N 5E5. E-mail:

Acknowledgements

R. Wang is an Investigator of Canadian Institutes of Health Research (CIHR), and L. Wu is a New Investigator of CIHR.

References

Wang R. Resurgence of carbon monoxide: An endogenous gaseous vasorelaxing factor. Can J Physiol Pharmacol 76 : 1–15, 1998.

Zhao W, Wang R. Carbon monoxide-induced vasorelaxation and the underlying mechanisms. In: Wang R, Ed. CO and Cardiovascular Functions. Baca Raton: CRC Press, pp23–44, 2001.

Wang R, Wu L. The chemical modification of K_Ca channels by carbon monoxide in vascular smooth muscle cells. J Biol Chem 272 : 8222–8226, 1997.

Wang R, Wu L, Wang ZZ. The direct effect of carbon monoxide on K_Ca channels in vascular smooth muscle cells. Pflügers Arch 434 : 285–291, 1997.

Cao K, Wang R. Carbon monoxide, Vascular contractility, and K⁺ channels. In: Wang R, Ed. CO and Cardiovascular Functions. Baca Raton: CRC Press, pp83–109, 2001.

Kaide JI, Zhang F, Wei Y, Jiang H, Yu C, Wang WH, Balazy M, Abraham NG, Nasjletti A. Carbon monoxide of vascular origin attenuates the sensitivity of renal arterial vessels to vasoconstrictors. J Clin Invest 107 : 1163–1171, 2001.

Bolotina VM, Najibi S, Palacino JJ, Pagano PJ, Cohen RA. Nitric oxide directly activates calcium-dependent potassium channels in vascular smooth muscle. Nature 368 : 850–853, 1994.

Frace AM, Eaton DC. Chemical modification of Ca²⁺-activated potassium channels of GH3 anterior pituitary cells. Am J Physiol 263 : C1081–C1087, 1992.

MacKinnon R, Miller C. Functional modification of a Ca²⁺-activated K⁺ channel by trimethyloxonium. Biochemistry 28 : 8087–8092, 1989.

10.

Wang R. Two’s company, three’s a crowd – Can H₂S be the third endogenous gaseous transmitter? FASEB J 16 : 1792–1798, 2002.

11.

Wang R, Karpinski E, Pang PKT. Two types of calcium channels in isolated smooth muscle cells from rat tail artery. Am J Physiol 256 : H1361–H1368, 1989.

12.

Wang R, Wang ZZ, Wu L, Hanna ST, Peterson-Wakeman R. Reduced vasorelaxant effect of carbon monoxide in diabetes and the underlying mechanisms. Diabetes 50 : 166–174, 2001.

13.

Wang R, Wang ZZ, Wu L. Carbon monoxide-induced vasorelaxation and the underlying mechanisms. Br J Pharmacol 121 : 927–934, 1997.

14.

Lee K, Ozanne SE, Hales CN, Ashford MLJ. Effects of chemical modification of amino and sulfhydryl groups on KATP channel function and sulfonylurea binding in CRI-G1 insulin-secreting cells. J Membr Biol 139 : 167–181, 1994.

15.

Spires S, Begenisich T. Modification of potassium channel kinetics by amino group reagents. J Gen Physiol 99 : 109–129, 1992.

16.

Pappone PA, Barchfeld GL. Modifications of single acetylcholine-activated channels in BC3H-1 cells. Effects of trimethyloxonium and pH. J Gen Physiol 96 : 1–22, 1990.

17.

Worley JF, French RJ, Krueger BK. Trimethyloxonium modification of single batrachotoxin-activated sodium channels in planar bilayers. Changes in unit conductance and in block by saxitoxin and calcium. J Gen Physiol 87 : 327–349, 1986.

18.

Sigworth FJ, Spalding BC. Chemical modification reduces the conductance of sodium channels in nerve. Nature 283 : 293–295, 1980.

19.

Munujos P, Knaus H, Kaczorowski GJ, Garcia ML. Cross-linking of charybdotoxin to high-conductance calcium-activated potassium channels: Identification of the covalently modified toxin residue. Biochemistry 34 : 10771–10776, 1995.

20.

Cescatti L, Pederzolli C, Menestrina G. Modification of lysine residues of Staphylococcus aureus alpha-toxin: Effects on its channel-forming properties. J Membr Biol 119 : 53–64, 1991.

21.

Kenyon GL, Bruice TW. Novel sulfhydryl reagents. In: Colowick SP, Kaplan NO. Ed. Methods in Enzymology, New York: Academic Press, Vol 47:pp407–430, 1997.

22.

Bouzat CB, Barrantes FJ, Sigworth FJ. Changes in channel properties of acetylcholine receptors during the time course of thiol chemical modifications. Pflugers Arch 418 : 51–61, 1991.

23.

Christensen BN, Hida E. Protonation of histidine groups inhibits gating of the quisqualate/kainate channel protein in isolated catfish cone horizontal cells. Neuron 5 : 471–478, 1990.