Acetylcholine-Activated Ion Channels In Embryonic Cockroach Neurones Growing In Culture

Neuronal cultures from embryonic cockroaches, Periplaneta americana, have been shown to possess binding sites for the snake toxin, α-bungarotoxin (α-BTX) and to be depolarized by pressure-applied acetylcholine (Lees et al. 1983; Beadle et al. 1984), suggesting that they possess nicotinic acetylcholine receptors and are therefore suitable for a detailed physiological analysis of insect central acetylcholine receptors. Acetylcholine is a major neurotransmitter in the central nervous system of both vertebrates (Krnjevic, 1974) and insects (Callec & Boistel, 1967; Kerkut et al. 1969; Gerschenfeld, 1973; Sattelle, 1980; Sattelle & Breer, 1987) and is the neurotransmitter at the vertebrate muscle endplate (Katz & Miledi, 1972; Anderson & Stevens, 1973). Acetylcholine receptors with a pharmacological profile characteristic of nicotinic receptors that are antagonized by α-BTX have been identified on giant interneurones GI2 and GI3 of the sixth abdominal ganglion of the cockroach (Harrow et al. 1980,1982; Sattelle et al. 1980) and on the slow (D_s) and fast (D_f) coxal depressor motoneurones of P. americana (Carr & Fourtner, 1980; David & Sattelle, 1984), and receptors with a nicotinic profile that are insensitive to α-BTX have been identified on dorsal unpaired midline (DUM) neurones of the grasshopper, Schistocerca nitens (Goodman & Spitzer, 1980). A detailed kinetic analysis of these receptors is made difficult, however, by the location of these neurones within the central nervous system which hinders the use of high-resolution patch-clamp techniques (Hamill et al. 1981).

Preliminary experiments with cultured cockroach neurones have resulted in the successful recording of unitary currents evoked by acetylcholine (Beadle et al. 1985, 1986) and similar recordings have been obtained from cultured Drosophila neurones (Wu et al. 1983) and freshly dissociated adult cockroach neurones (Sattelle et al. 1986). In addition, nicotinic acetylcholine receptors purified from the central nervous system of the locust have been reconstituted in planar lipid bilayer membranes and acetylcholine-evoked unitary currents detected (Hanke & Breer, 1986).

We present here the results of a detailed analysis of the acetylcholine-activated ion channels of cultured cockroach neurones using the patch-clamp technique in both the cell-attached and the whole-cell voltage-clamp mode. The analysis was carried out on neurones that had been growing in vitro for at least 14 days, at which stage the cells have become fully differentiated and express acetylcholine receptors (Beadle et al. 1982; Lees et al. 1983, 1985).

Neuronal cultures were prepared from the brains of 21- to 23-day-old embryos of P. americana, as described elsewhere (Dewhurst & Beadle, 1985). The cultures were initiated in a medium consisting of five parts of Schneider’s revised Drosophila medium and four parts of Eagle’s basal medium containing streptomycin and penicillin. After 7 days growth they were transferred to a medium containing equal parts of Leibovitz’s L-15 and Yunker’s modified Grace’s medium containing streptomycin and penicillin. The cells were grown at 29°C in air in 50 mm Falcon Petri dishes using a modification of the hanging column method. Fig. 1 shows embryonic cockroach neurones after 14 days growth in vitro.

For electrophysiological experiments the growth medium was removed and cells were allowed to equilibrate in a saline solution modified from that of Pitman (1979) to meet the osmotic requirements of the cultured cells. The solution contained 210mmoll⁻¹ NaCl, 10 mmoll⁻¹ CaCl₂, 3·l mmoll⁻¹ KC1 and 10 mmol l⁻¹ Hepes buffer at pH 7·2. The experiments were performed at room temperature (22–24°C). For experiments using the cell-attached configuration the electrodes were filled with this saline containing low concentrations of either acetylcholine (ACh) or carbamylcholine (CCh). For voltage-clamp experiments the whole-cell clamp configuration of the patch-clamp technique was used (Hamill et al. 1981). The electrodes were filled with a solution containing 114 mmol l⁻¹ KC1, 5 mmol l⁻¹ EGTA, l·6mmoll⁻¹MgCl₂, 0·2 mmoll⁻¹ CaCl₂ and buffered at pH 7·2 with 10 mmol l⁻¹ Hepes. Electrodes were pulled in two stages from 1·5 mm haematocrit tubing on a modified Kopf 150C vertical puller and were polished and coated with Sylgard. The resistance ranged from 2 to 7 MΩ. The electrode was advanced towards the cell until gentle contact was made. A small amount of suction applied to the electrode usually resulted in a high-resistance seal of several gigaohms. When needed, rupture of the cell membrane was obtained by further suction and the resting potential of the cell was recorded.

The patch electrodes were connected to either a List (L/M-EPC5) or Dagan (8900) patch-clamp amplifier. The results were displayed on either a Tektronix or a Gould oscilloscope and stored either on analogue tape (Ampex PR 500) or on video cassettes using a modified Betamax video recorder and pulse code modulator at full amplifier bandwidth (Lamb, 1985), and were reproduced photographically from Gould 220 Brush chart records. Current fluctuations were analysed off-line using a spectrum analyser (HP3582A) connected to a desk computer (HP9825T) and a digital plotter (HP9872A). The fluctuations were analysed for bandwidths between 0 and 500 Hz using the method described by Shimahara et al. (1987). For analysis of single-channel activity, long stretches of stable recording are needed and the results from three cells were selected for detailed statistical analysis. The data were digitized on an IBM PC clone using a Data Translation (DT2801A) A/D converter. The original data were filtered at 5 kHz using a Tehebicheff filter and acquired at a 10 kHz sampling frequency. The sample used for analysis consisted of about 100 000 12-bit strings stored on the hard disc. The data were analysed using a computer program called IPROC-2 originally developed by Sachs et al. (1982) and donated by C. J. Lingle. The output of the program consists of a series of files containing information on the amplitude and duration of each valid event and on-time, off-time and duration histograms. A curve-fitting program linked to a statistical library was subsequently used to analyse the histograms except for on-time histograms of very short events (see next paragraph).

In cultured cockroach neurones, as in other preparations such as the frog muscle endplate (Colquhoun & Sakmann, 1981, 1985) and snake muscle fibre (Dionne & Leibovitz, 1982), the apparent single openings of the ion channel associated with the cholinergic receptor are interrupted by brief closed periods (see Fig. 4). These closed periods are usually too short to be resolved under our experimental conditions [by analogy with the results of Colquhoun & Sakmann (1985) using frog muscle endplate, the time constant of the briefest gap component for both ACh and CCh may be of the order of 10–20 μs]. This results in an overestimation of the open time of the channels and an underestimation of the single-channel amplitude (Sachs, 1983). These errors were minimized in the analysis by setting the parameters of the minimum closed time in a burst to its minimum value (100μs). Another limitation was the short duration of many of the openings, especially with ACh, and the existence of ‘triangular’ openings. Here again, the minimum duration for an event to be accepted as an opening was set to the minimum value (100 μs). Under these conditions, the amplitude of the short events was underestimated and the amplitude histograms distorted and shifted towards smaller values (see Figs 5, 6). Furthermore, the first two bins of the on-time histograms (0·1 and 0·2 ms) were underestimated. This resulted in a substantial error in the fit of the on-time histograms. To overcome this difficulty, a different program was used in which the slow exponential components were calculated first from a semilogarithmic plot of the data and then subtracted from the fast component.

When 10–50 μmol l⁻¹ ACh or CCh was applied from a pressure pipette onto the soma of cultured embryonic cockroach neurones held under whole-cell voltageclamp conditions, an inward current was evoked that reached a peak value of 200–250 pA with the higher concentrations. The inward current was typically accompanied by an increase of current noise (Fig. 2). Spectral analysis of this noise revealed the existence of one (CCh) or two (ACh) Lorentzian functions. At rest, the mean corner frequency of the Lorentzian component in CCh was 35 ±13 Hz, corresponding to a mean open time of 4·6 ms; the mean corner frequencies of the two Lorentzian functions in ACh were 10 ±0·6 Hz and 116·8 ± 9 Hz, corresponding to mean open times of 15·9 ms and 1·36 ms, respectively (Fig. 3). Membrane hyperpolarization resulted in a shift of the spectrum towards lower frequencies, suggesting an increase of the mean open time of the single-channel events (the mean corner frequency with CCh decreased from 35 ± 13 Hz at rest to 16·5 ± 2 Hz for an 80 mV hyperpolarization, corresponding to a lengthening of the mean open time from 4-6ms to 9·7ms).

When added to the patch pipette at micromolar concentrations, both ACh and CCh induced small inward unitary currents. At the resting potential (RP), the mean amplitude and mean duration of these channels were, respectively, 1·57 ± 0·5 pA and 0·342 ± 0·28 ms for ACh and 1·53 ± 0·64 pA and 0·415 ± 0·53 ms for CCh. Membrane hyperpolarization revealed the existence of two categories of unitary currents (Fig. 4A,B): large events with a mean current of 4·77pA with ACh and 5·09 pA with CCh and small events with a mean current of 1·92pA with ACh and 1·72pA with CCh (for a 60mV hyperpolarization). These events could be classified into three categories from their time course: long bursts of openings (several tens of milliseconds), short openings (a few milliseconds) with occasional closings and substates and very short triangular openings (less than 0·5ms). The bursting behaviour and triangular events were more frequent with ACh than with CCh. The mean open time of the short channels was longer with CCh than with ACh.

Amplitude histograms were constructed for the two agonists for membrane potentials between +100mV and −40mV relative to the resting potential (i.e. H100 to D40). In all experiments, the histograms could be reasonably well fitted with two Gaussian functions reflecting the existence of the two populations of events. Individual values of the amplitude and standard deviations of these Gaussian functions in two patches are presented in Table 1. It can be seen that the standard deviations are larger for the ACh-induced events than for those induced by CCh, reflecting the existence in ACh of a larger proportion of very short, partly unresolved triangular openings. Examples of such fits are given for the two agonists at four potential levels in Figs 5 and 6. It can be observed that the short duration of the events distorts the distribution and shifts the fitted curve to the left (i.e. towards lower conductance values).

Fig. 7 shows the effects of membrane hyperpolarization on the activity of channels induced by 5 μmol l⁻¹ CCh: both amplitudes and durations are increased but this change is not associated with a significant increase of the noise during the openings (substates or flicker). For membrane potential values between 100mV (H100) and −30 mV (D30) the current-voltage relationships were linear for the two agonists (Figs 8, 9). For ACh, the reversal potentials calculated from the linear portion of the curves were –69·7 ± 2·5 mV (N = 19) for the large channels, –73·4 ± 6·2 mV (N = 9) for the small channels and –67·9 ± 1·84mV (N= 15) for the mean amplitude of all the channels before curve fitting. For CCh, the corresponding values were −68·2 ±4·15 mV (N = 8) for the large channels –93·9 ±10·5 mV (N = 4) for the small ones and -78·6 ±3·7 mV (N=8) for the mean amplitude. Single-channel conductance values obtained from the same curves were 37·4±1·90pS (N =19) for the large channels in ACh, and 19·0 ± 2·0 pS (N = 9) for the small channels in ACh, with a mean of 28·8 ± l·4pS (N = 15). For CCh, the single-channel conductance values were 51·8±4·4pS (N =8) for the large channels, and 14·8 ± 2·8 pS (N =4) for the small channels with a mean of 33·6±4·8pS (N = 8). For a given ACh or CCh concentration, changes in membrane potential were found to alter the duration of the singlechannel events: hyperpolarizations increasing the duration and depolarizations decreasing it. The effect of hyperpolarization is particularly striking in Fig. 7 between resting potential (RP) and 30H (30mV hyperpolarization).

On-time histograms were constructed for the two agonists and tentatively fitted with one, two and three exponential functions. Within the limitations of the analysis, it was found that the best fit was obtained with two exponential functions. In ACh, the duration of the single channels was usually so short that it was not possible to obtain a reasonable estimate of the time constant of the fast exponential component but, as illustrated in Fig. 10, the data could not be fitted with a single exponential. The situation was slightly better with CCh and the histograms illustrated in Fig. 11 were fitted with two exponential components. No systematic study of the off-time histograms was performed. Preliminary analyses with CCh at 60H suggested, however, the existence of three exponential components.

Examination of the cross-correlation histograms in both ACh and CCh revealed that events of both sizes contributed to the two time constants although separate analysis of the channels of the two sizes revealed on one occasion that the large events were somewhat faster than the small ones.

The voltage-dependency of the open time was studied for membrane potential values between H100 and D40. Fig. 12 illustrates the voltage-dependency of the mean open time for two patches in ACh (Fig. 12A) and one patch in CCh (Fig. 12B). In the three cases, the voltage-dependency was found to increase considerably when the membrane was depolarized. The data were thus fitted with two exponentials, a fast one for depolarized potentials and a slow one for potential values more positive than −20mV. For the three illustrated patches, e-fold changes in duration for the fast component were observed for potential changes of 36·2mV (open triangles in Fig. 12A), 24·1 mV (filled triangles in Fig. 12A) and 25·5 mV (open triangles in Fig. 12B). For the same three patches, e-fold changes in duration for the slow component were observed for potential changes of 603 mV (open triangles in Fig. 12A), 258mV (filled triangles in Fig. 12A) and 1420mV (open triangles in Fig. 12B).

When measurable, both fast and slow components of the on-time histograms were voltage-dependent. Thus, for the patch illustrated in Fig. 12B, e-fold changes in the time constant of the fast component of the on-time histogram were observed for potential changes of 7·89 mV (depolarizations) and 355 mV (hyperpolarizations), whereas an e-fold change in the slow component of the on-time histogram was observed for a 103mV change in potential (not illustrated).

The mean frequency of the events induced by micromolar concentrations of the two agonists was low (around 10–15 s⁻¹ for ACh and 30 s⁻¹ for CCh at the resting potential level), corresponding to a mean open-time probability of 0·003 for ACh and 0·01 for CCh. The actual values are probably lower since there is evidence, that will be discussed below, that at least two channels were probably present under the patch pipette. The effects of membrane potential on the relative frequency of opening and the relative open-time probability were determined for membrane potential values between 100mV (H100) and −40mV (D40). In all cases, membrane hyperpolarization was associated with a statistically significant increase in the number of openings per unit time. For the three experiments illustrated in Fig. 13, an e-fold change in the frequency was obtained for 168·7 mV, 50·5 mV (Fig. 13A) and 80·66 mV (Fig. 13B). Since both open times and frequencies of opening increased with membrane hyperpolarization, the (apparent) open-time probability was also strongly voltage-dependent. Thus, in the experiment illustrated in Fig. 14, e-fold changes in the relative open-time probabilities were obtained for 115·1 mV and 46·4mV (Fig. 14A) and 55mV (Fig. 14B).

The results presented here indicate that the somata of neurones from the brains of embryonic cockroaches growing in culture possess ion channels that are activated by acetylcholine and carbamylcholine. This confirms the results of previous microelectrode studies that have demonstrated that the somata of many insect central neurones are cholinoceptive despite their lack of synapses (Kerkut et al. 1969; Pitman & Kerkut, 1970; David & Pitman, 1979; Lees et al. 1983; Suter & Usherwood, 1985). Recent work involving the mapping of acetylcholine receptors on insect neurones using the radiolabelled ligand [¹²⁵I] α-BTX has also demonstrated their presence on neuronal cell bodies (Sattelle, 1980; Lees et al. 1983). The functional significance of these extrajunctional receptors has yet to be determined.

For both agonists, there is an apparent discrepancy between single-channel data and the results of spectrum analysis of the current fluctuations recorded in the whole-cell configuration. Amongst other possibilities, this difference could indicate that the ACh channels are not homogeneously distributed over the cell surface. Because of the specific limitations of these two methods of analysis (limited bandwidth, unequal representation of the different spectral components, importance of the background noise), randomly selected single-channel data were re-examined and it was found that the two categories of data could be reconciled to some extent. For example, in ACh (Fig. 4A) the mean duration of the short channels was 1·17 ± 0·87 ms, corresponding to a corner frequency of 136 Hz, and the duration of the burst was 12·5 ms, corresponding to a corner frequency of 12·7 Hz. Similarly, with carbamylcholine (Fig. 4B), the mean duration of the large events was 2·17 ± 1·8 ms, corresponding to a mean corner frequency of 73·4 Hz, and that of the small ones was 4·07 ±0·9 ms, corresponding to a mean corner frequency of 39·1 Hz.

The conductance values calculated from the I/V curves were 37 pS (large) and 19 pS (small) for ACh and 52 pS (large) and 15 pS (small) for CCh. These are in reasonable agreement with our previously published values for CCh of 48 pS (large) and 18 pS (small) (Beadle et al. 1985). The larger value of conductance in CCh than in ACh is at variance with most reports on other preparations (see, for example, Colquhoun & Sakmann, 1985; Gardener et al. 1984) where the conductance was found to be independent of the nature of the agonist. As mentioned in Materials and methods, such a difference may not be genuine but may result from an underestimation of the current amplitude of the single channel in ACh because of the short open time (triangular events were not eliminated) or bursting behaviour. To test this hypothesis, single channels with durations longer than 1 ms and no visible burst or substate were selected for further analysis. Under these conditions, the mean single-channel conductance for the large events in ACh was found to approximate 50 pS (i.e. not significantly different from that induced by CCh). The larger conductance value is similar to the value of 40pS for ACh channels in freshly dissociated adult cockroach neurones (Sattelle et al. 1986) but is considerably lower than the value of 75 pS reported by Breer (1986) for locust ACh channels reconstituted in planar lipid bilayers. The smaller conductance value is similar to that of 9–25 pS for ACh channels in cultured larval Drosophila neurones (Wu et al. 1983). Acetylcholine-activated ion channels with conductances of 38·42 pS have been reported in chick ciliary ganglion neurones grown in culture (Ogden et al. 1984) and of 34 pS in embryonic rat muscle in culture and 30 pS in adult frog muscle endplate (Gardner et al. 1984).

Do the two conductance states reported here correspond to two distinct populations of channels or two substates of the same channel? Substates have been reported for ACh-activated ion channel at the frog endplate (Colquhoun & Sakmann, 1985). In this case precise kinetic studies associated with modelling of channel activity clearly indicate that they are substates. Substates can also be seen in recordings of unitary ACh currents from dissociated adult cockroach neurones (Sattelle et al. 1986), although only one conductance value is given. The existence of substates is based on the frequent occurrence of conductance changes from one level to the other. Analysis of large numbers of channels in cultured cockroach neurones with both ACh and CCh failed to reveal such frequent conductance changes, with transitions from the small to the large conductance state occurring only very rarely. From this we conclude that embryonic cockroach neurones growing in culture possess two populations of ACh-activated ion channels. For embryonic frog muscle, two conductance states of 25pS and 35pS as well as a substate of 10pS have been reported (Hamill & Sakmann, 1981), and cultured rat myocytes express an embryonic ACh-activated ion channel with a conductance state of 33 pS and an adult one of 58 pS (Jaramillo & Schuetze, 1988). Rat skeletal muscle fibres express at least two different types of ACh receptor differing in their conductance state and their gating properties (Hamill & Sakmann, 1981; Siegelbaum et al. 1984; Jamarillo & Schuetze, 1988). Recent results indicate that these two types differ in their subunit composition (Mishima et al. 1986), the gamma subunit of the low-conductance channel in the embryo (alpha2-beta-gamma-delta) being replaced by an epsilon subunit in the adult. It may be that one of the conductance states reported here represents an embryonic receptor and the other an adult receptor. Developmental studies using this culture system may resolve this question.

Bursting behaviour of the type illustrated in Fig. 4 was frequently observed, being more common with ACh, and this probably accounts for the apparently smaller conductance value for ACh channels than for CCh channels. More precise studies are needed to understand the bursting behaviour of these channels, although it appears to be very similar to that described by Colquhoun & Sakmann (1985) at the frog muscle endplate. Unitary events never appear in clusters with the range of concentrations (1–50μmol l⁻¹) used in our experiments, suggesting that the receptor does not desensitize. The open-time probability was very low with both agonists, less than 1%, in contrast with that of adult dissociated cockroach neurones where the recordings show a high probability of channel openings (Sattelle et al. 1986). This could represent a difference between cultured neurones and those occurring in situ or between embryonic and adult neurones. The voltage-sensitivity of the mean channel open time is a classical feature of the N-ACh receptors in vertebrate preparations (see Jamarillo & Schuetze, 1988, for references). The observed increase in the open-time probability (which results from this effect and from the voltage-dependent increase in the frequency of events per unit time) should be reflected in the I/V relationship for a given agonist concentration, as in the fast coxal depressor motoneurones (D_f) in the metathoracic ganglion of the cockroach (Sattelle et al. 1986).

In conclusion, ACh receptors in cockroach neurones after 2 weeks in culture resemble in some aspects the nicotinic receptors at the vertebrate endplate. The stability of this preparation and its ease of use suggest that it may act as a useful model for the study of the basic properties and development of the neuronal nicotinic receptor. It would be very well suited for a study of the effects of various factors and other cell types such as glia on the development and characteristics of these receptors.

This work was supported by a grant from the Agricultural and Food Research Council.