Nereistoxin: Actions on a Cns Acetylcholine Receptor/ion Channel in the Cockroach Periplaneta Americana

Nereistoxin (NTX), a naturally occurring substance which is active at cholinergic synapses (cf. Narahashi, 1973), was first isolated from the marine annelid Lumbriconereis heteropoda by Nitta (1934). Okaichi & Hashimoto (1962a,b) synthesized the toxin and determined its chemical structure as 4-N,N-dimethylamino-l,2-dithiolane (Fig. 1). Much of the interest in the mechanism of action of NTX stems from the high toxicity to insects of the natural toxin (Sakai, 1964, 1966a) and synthesized analogues such as the commercial insecticide ‘Cartap’ (the active form of which is NTX, Sakai & Sato, 1971). Nevertheless, little is known of the mechanism of action of NTX on insect excitable tissues. The present study, therefore, examines the effects of NTX on identified neurones in the central nervous system of the cockroach Periplaneta americana.

Three distinct synaptic actions of NTX were reported in the earlier studies on a variety of vertebrate preparations. First, a d-tubocurarine-like antagonism of nicotinic cholinergic receptors was demonstrated (Nitta, 1941; Sakai, 1966b; Deguchi, Narahashi & Haas, 1971). Secondly, the frequency of miniature end-plate potentials (MEPPs) recorded from frog muscle decreased in the presence of NTX (Deguchi et al. 1971), indicating an additional inhibitory effect of NTX on the mechanism of acetylcholine release. Finally, a number of stimulatory effects of NTX were reported in several mammalian tissues containing muscarinic cholinergic receptors (Nitta, 1941 ). Recently, Eldefrawi et al. (1980) have applied a multidisciplinary approach to the study of synaptic actions of NTX on vertebrate muscle and Torpedo electroplax tissues. They could find no evidence for a presynaptic effect of NTX on acetylcholine release. In the same investigation, electrophysiological experiments on the neuromuscular junctions of frog (sartorius) and rat (diaphragm) muscle were performed in parallel with receptor binding studies using ¹²⁵I-αbungarotoxin, and binding to the ion channel using [³H]perhydrohistrionicotoxin.

NTX failed to inhibit binding of [³H]perhydrohistrionicotoxin and did not modify either the linearity of the current voltage relationships, or the time-course of the end-plate current. It did, however, inhibit the binding of ¹²⁵I-α-bungarotoxin to Torpedo membranes and inhibited both postsynaptic and extrajunctional acetylcholine sensitivity of muscle. It was concluded that NTX inhibition of vertebrate neuromuscular transmission was due to an action at the receptor rather than the ion channel of the acetylcholine receptor/ion channel complex.

The site of action of NTX in insects is not clear. They are rapidly paralysed by NTX but subsequently revive from the poisoning. Using extracellular, hook-electrode recording techniques, Sakai (1967, 1970) showed that concentrations above 2·0 × 10⁻⁶ mol 1⁻¹ NTX suppressed postsynaptic discharges at cockroach cereal-afferent, giant-interneurone synapses. Microelectrode recordings from giant axons in the same preparation revealed that 1 ·0 × 10⁻⁵ mol 1 ⁻¹ Cartap blocked synaptic transmission and depolarized the postsynaptic membrane (Bettini, D’Ajello & Maroli, 1973). By contrast, neuromuscular transmission in the femur of the hindleg of Periplaneta was not affected by 2·0 × 10⁻⁴moll⁻¹ NTX(Sakai, 1970).

Although the excitatory neurotransmitter at the cereal afferent, giant interneurone synapses has not been established, acetylcholine is the strongest candidate (Pitman, 1971; Callec, 1974; Sattelle, 1980). In support of this suggestion are experiments which show that of all molecules tested, a-bungarotoxin and other nicotinic cholinergic ligands are the most effective on excitatory postsynaptic potentials and acetylcholine-induced depolarizations of giant interneurones (Shankland, Rose, & Donniger, 1971; Callec, 1974; Sattelle, 1978; Sattelle, David, Harrow & Hue, 1980; Sattelle et al. 1983; Harrow, Hue, Pelhate & Sattelle, 1979; Harrow, David & Sattelle, 1982). An a-bungarotoxin-binding component exhibiting a similar pharma-cological profile to that of postsynaptic receptors at cereal afferent, giant interneurone synapses has been characterized in extracts of abdominal ganglia of the same insect (Gepner, Hall & Sattelle, 1978). Interpretation of the action of NTX on insects would be facilitated by the use of experimental material such as the CNS of the cockroach Periplaneta americana that is well suited to the application of both electrophysiological and radiolabelled ligand-binding techniques.

In the present investigation electrophysiological methods are used to quantify the actions of NTX on synaptic, cell body and axonal membranes of GI 2 in the sixth abdominal ganglion of Periplaneta americana. In addition, the capacity of NTX to inhibit ¹²⁵I-α-bungarotoxin binding to abdominal nerve cord extracts of. the same species is tested. Two separate ways in which NTX interacts with an insect acetylcholine receptor/ion channel complex are detected. A brief account of some of these findings has appeared in Conference proceedings (Sattelle & Callee, 1977; Sattelle, 1981).

Adult male cockroaches (Periplaneta americana) were used in all physiological experiments and for binding experiments. Drosophila melanogaster extracts were prepared from heads of Canton-S wild-type flies. Nereistoxin hydrogen oxalate (MW 239), from Takeda Chemical Industries Ltd, was used in all the present investigations and was bath-applied in saline in all physiological experiments.

The actions of NTX on the isolated axonal membrane of giant interneurone 2 (GI 2) were studied using the oil-gap, single-fibre method. Details of the axon dissection, experimental chamber and the electrical circuitry for current-clamp and voltage-clamp have been reported elsewhere (Pelhate & Sattelle, 1982). Electrical responses of the axon were monitored on a storage oscilloscope and simultaneously stored using a Histomat S multichannel analyser which enabled display of the ionic currents after correction for leakage and capacity currents by the method of Hille & Campbell (1976). The physiological saline used for the axon was as follows (in mmol 1⁻¹): NaCl, 200·0; KC1, 3·1; CaCl₂, 5·4; MgCl₂, 5·0; NaHCO₃, 2·0; Na₂HPO₄, 0·1 (pH 7·2). Most experiments were performed at 12 °C to prolong the viability of the isolated axon, although a number of experiments were also performed at 22—24 °C.

The actions of NTX on transmission at the cereal afferent, giant interneurone synapses were monitored using an oil-gap, single-fibre synaptic recording technique (Pichón & Callec, 1970) and a sucrose-gap method (Callec & Sattelle, 1973). A preparation comprising the cerci, the cereal nerves, the terminal (sixth) abdominal ganglion and the remainder of the abdominal nerve cord was first excised from the animal, and the sixth ganglion was desheathed to facilitate the access of perfusing fluids to the synapses. Preparations of this type were used for sucrose-gap experiments. For single-fibre experiments, one giant axon (GI 2) was dissected from one of the connectives linking the fifth and sixth abdominal ganglia and the other connective was cut. Electrical and mechanical stimulation of several sensory neurones and the identification of GI 2 were performed as described previously (Sattelle et al. 1983). The experimental chamber for recording changes in membrane potential, input resistance and excitatory postsynaptic potentials, together with the electrical circuitry have been described elsewhere (Pichón & Callec, 1970; Callec, 1972). The ionic composition of the physiological saline used for experiments on synaptic transmission was as follows (in mmol 1⁻¹): NaCl, 208·6; KC1, 3·1; CaCl₂, 5·4; NaHCO₃, 2·0; Na₂HPO₄, 0·1 (pH 7·2). All experiments were performed at 21 °C.

The response to ionophoretically-applied acetylcholine of the cell body membrane of GI 2 in the sixth abdominal ganglion and D_f (the fast coxal depressor motoneurone) in the metathoracic ganglion were recorded as described earlier (Sattelle et al. 1980; David & Pitman, 1982; Harrow & Sattelle, 1983). The actions of NTX on acetylcholine-induced current were monitored using voltage-clamp techniques described elsewhere (David & Sattelle, 1984).

Membrane extracts, containing the particulate form of a putative nicotinic cholinergic receptor, were prepared by differential centrifugation as described elsewhere for Drosophila melanogaster (heads) (Schmidt-Nielsen, Gepner, Teng & Hall, 1977) and Periplaneta americana (abdominal nerve cords) (Gepner et al. 1978). Aliquots of the extract [equivalent to 5 mg heads (Drosophila) starting material, or 20 mg wet weight Periplaneta abdominal nerve cords] were preincubated for 30 min at 21 °C in Krebs original Ringer phosphate buffer pH 7·4 (Dawson, Elliott, Elliott & Jones, 1969) containing 1·7 times the final concentration of NTX in a total volume of 70󐎼l. Then 50/21 of 4 nmol 1^{−1 125}I-α-bungarotoxin (specific activity 27 Ci mmol⁻¹) in Krebs-Ringer buffer containing 0·1 % bovine serum albumin was added and the mixtures were incubated for an additional 30 min. The mixtures were then centrifuged for 30 min at 30 000 × g and the amount of ¹²⁵I-α-bungarotoxin bound was determined by a centrifugal assay. The centrifugal assay and preparation of ¹²⁵I-α -bungarotoxin have been de-scribed by Schmidt-Nielsen et al. (1977).

At concentrations of 1·0 × 10⁻⁴mol 1⁻¹ and below, no effects of bath-applied NTX were detected on the resting potential of the isolated axon. At higher concentrations, a dose-dependent depolarization of the axonal membrane was observed (Fig. 2A). The depolarization induced by 2·0 × 10⁻⁴moll⁻¹ NTX was accompanied by a drop in input resistance as shown by the application of hyperpolarizing current pulses (Fig. 3). The membrane depolarizations induced by the application of depolarizing current pulses were increased in the presence of NTX (2·0 × 10⁻⁴ mol 1⁻¹), a possible explanation for which is a change in the delayed rectification of the axon. NTX at 5·0 × 10⁻³moll⁻¹ depolarized the axon by 15—25 mV. Similar NTX-induced depolarizations were observed in the presence of tetrodotoxin (TTX) at 5·0 × 10⁻⁷ mol 1⁻¹, a concentration which completely blocks axonal sodium channels of cockroach giant axons (Sattelle, Pelhate & Hue, 1979).

Action potentials of normal amplitude (90-100 mV) were observed under current-clamp conditions even after a 30-min exposure to 1·0 × 10⁻⁴mol 1⁻¹ NTX. In the concentration range 2·0 × 10⁻⁴-5·0 × 10⁻⁴moll⁻¹ NTX (six experiments), a slowing down of the repolarization phase was noted (Fig. 4). At these concentrations NTX did not suppress action potentials even after a 30-min application, In fact after a 30-min exposure at 22-24°C to 2·0 × 10⁻⁴moll⁻¹ NTX, a repetitive firing of action potentials was detected in response to long-lasting depolarizing current pulses. In contrast, at higher NTX concentrations of 1·0 × 10⁻³ mol 1⁻¹, it was usually impossible to elicit action potentials from fibres after 15 min of perfusion.

Using voltage-clamped axons, NTX in the concentration range 1·0 × 10⁻⁶-l·0 × 10⁻⁴ mol 1⁻¹ had no effect on membrane currents after 30 min of bath application (five experiments). However, at 2·0 × 10⁻⁴mol 1⁻¹ and 5·0 × 10⁻⁴molU¹ NTX, all the currents were slowly and irreversibly modified. Both the steady-state outward (potassium) current (I_K) and the peak inward (sodium) current (I_Na) were reduced, whereas the leak current (I_L) was increased. The actions of NTX (1·0 × 10⁻³moll⁻¹) on axonal currents are shown in Fig. 5.

Compound EPSPs were recorded from cereal afferent, giant interneurone synapses by the sucrose-gap method, following electrical stimulation of cereal nerve XI. Following a 15-to 20-min exposure to NTX at 3·0 × 10⁻⁸moll⁻¹ and higher concentrations, compound EPSPs were reduced in amplitude (Fig. 2B). A reduction of the EPSP amplitude to approximately 40 % of its normal value in physiological saline (3—8mV) was achieved at l·0 × 10⁻⁶moll⁻¹ NTX. Increasing the toxin concentration to 1·0 × 10 ⁴moll⁻¹ produced no further reduction of EPSP amplitude (Fig. 2B). The EPSP attenuation was observed after 15 min exposure to 10x 10”⁶ mol U¹ NTX and was not further modified by continued perfusion of NTX for another 15 min. This rapid, partial suppression of EPSP amplitude was not accompanied by changes in either input resistance or membrane potential. At concentrations of NTX above 2·0 × 10⁻⁵moll⁻¹, depolarization of the postsynaptic membrane was observed (Fig. 2C). Using 15 min exposure to saline containing NTX, maximum depolarization was achieved at concentrations in the range 1·0 × 10⁻³ to 1·0 × 10⁻² mol 1⁻¹ NTX. Depolarization of the postsynaptic membrane was accompanied by a block of the EPSP and a reduction in input resistance.

The oil-gap, single-fibre technique was used to test the effects of NTX on EPSPs recorded from GI 2 following mechanical stimulation of a single cereal mechanoreceptor, and corresponding in amplitude to the unitary EPSPs previously recorded from this pathway (Callec, 1974). The amplitudes of such EPSPs were in all cases (N = 18) progressively reduced to about 30—40% of the value in normal saline by NTX at concentrations in the range 1·0 xl0⁻⁶ to 1·0 × 10 ⁻⁴ mol 1⁻¹. As shown in Fig. 6, the suppression by 2·0 × 10”⁵ mol 1⁻¹ NTX of the EPSP resulting from the stimulation of a single cereal mechanoreceptor was only partially reversed by bathing in normal saline. Compound EPSPs, recorded by the oil-gap, single-fibre technique from GI 2 in response to electrical stimulation of many cereal afferents, were partially blocked by NTX at concentrations similar to those reported in sucrose-gap experiments where the response is derived from many postsynaptic fibres (see Fig. 2B). The residual component of the EPSP that was not blocked by NTX was rapidly blocked by d-tubocurarine or atropine (1·0 × 10⁻⁴moll⁻¹).

The actions of NTX on the cell body membrane of GI 2 and the fast coxal depressor motoneurone (D_f) were examined. In voltage-clamp experiments on the cell body of D_f, currents resulting from ionophoretic application of acetylcholine were recorded at membrane potentials (E_m) in the range — 120 mV to — 60 mV, in normal saline, and following a 25-min exposure to l⁻0x 10⁻⁷moll⁻¹ NTX (Fig. 7A). The degree of inhibition of the acetylcholine-induced inward current increased with membrane hyperpolarization up to —100 mV and thereafter ( — 110 mV and —120 mV) declined somewhat, indicating a voltage-dependence of the action of NTX (Fig. 7B) with a maximum block at — l00mV. The threshold concentration for an action of the toxin on the cell body membrane of Df was in the range 1·0 × 10⁻⁷ to 1·0 × 10⁻⁶moll⁻¹.

The cell body membrane of GI 2 exhibited an increased percentage block of acetylcholine-induced current with membrane hyperpolarization over the potential range ( — 120 mV to —60 mV) when exposed to 1·0 × 10⁻⁶ mol 1⁻¹ NTX. (Fig. 7C). Both GI 2 (Harrow & Sattelle, 1983) and D_f (Sattelle et al. 1980; David & Sattelle, 1984) possess a-bungarotoxin-sensitive nicotinic acetylcholine receptors on their cell body membranes. For both cells, inhibition of the acetylcholineinduced current by NTX is voltage-dependent in the range —120 mV to — 60mV. At concentrations of 1·0 × 10⁻⁴moll⁻¹ NTX and higher, the cell body membranes of D_f and GI 2 were depolarized.

The inhibition by NTX of ¹²⁵I-α-bungarotoxin binding to extracts from Drosophila (heads) and Periplaneta (abdominal nerve cords) was investigated (Fig. 8). Each point on the binding curves represents the average of three incubation mixtures at that particular toxin concentration. From the Drosophila data, the concentration of NTX required to inhibit ¹²⁵I-α-bungarotoxin-binding by 50 % (I50) was found to be 6·6 × 10⁻⁵ mol 1⁻¹. In an independent repetition of this experiment on Drosophila extract the 150 for NTX was determined to be 5·6 × 10⁻⁵moll⁻¹. From the Periplaneta data, an I50 of 1·7 × 10⁻⁴moll⁻¹ was estimated. Thus the ability of NTX to inhibit ¹²⁵I-α-bungarotoxin binding is similar for Drosophila and Periplaneta extracts. This inhibition is only seen at the higher concentrations of NTX where electrophysiological studies on Periplaneta neurones show depolarization.

The data presented here provide a quantitative description of the actions of NTX on axons, synapses and cell bodies of identified neurones together with its effects on an ¹²⁵I-μ-bungarotoxin binding site in the central nervous system of the insect Periplaneta americana. It is possible to consider potential sites for the central actions of this toxin based on these results. These include: voltagesensitive ion channels (e.g. sodium and potassium); non-specific membrane actions; acetylcholine release; acetylcholine receptor/ion channel.

The axonal membrane is relatively insensitive to NTX at concentrations below 1 ·0 × 10 ^{− 4} mol 1⁻¹. At concentrations higher than 2·0 × 10 ⁻⁴ mol 1⁻¹ several modifications of axonal membrane properties are detected including suppression of the sodium current, decrease in potassium current and increase in leak current. The decrease in potassium current is consistent with the repetitive activity sometimes induced during prolonged exposure to low concentrations of NTX. The increase in the leak current may explain the depolarizing axonal effects of NTX seen under current-clamp conditions. Depolarization in this case is not due to a specific increase in resting sodium conductance since it is noted in the presence of tetrodotoxin (TTX), a specific sodium channel blocker of cockroach axons (Pichón, 1974; Sattelle et al. 1979). Thus NTX at a concentration 1·0x10⁻³ moll⁻¹ blocks voltage-dependent sodium and potassium channels and induces a non-specific, irreversible increase in ion permeability of the axonal membrane.

NTX, a weak inhibitor of insect acetylcholinesterase (Sakai, 1966c) and choline acetyltransferase (A. C. Baillie, K. Wright & D. B. Sattelle, unpublished observations), nevertheless exhibits several potent actions at cereal afferent, giant interneurone synapses. An action of NTX as an antagonist at the postsynaptic acetylcholine receptor recognition site could account for changes in EPSP amplitude. However, this site is probed by a-bungarotoxin and the concentration of NTX producing half-maximal, non-depolarizing block of transmission (maximum suppression of EPSP amplitude = 60 %) is 8·0 × 10⁻⁷moll⁻¹, whereas 50 % inhibition of ¹²⁵I-a-bungarotoxin binding is produced by NTX concentrations of 1·7 × 10⁻⁴moll⁻¹(Periplaneta americana extract) and 6·6 × 10⁻⁵moll⁻¹(Drosophila melanogaster extract). In studies on fly head (Musca domestica) extracts, Eldefrawi & Eldefrawi (1980) showed that l·0x 10⁻⁵moll⁻¹ NTX inhibited by 50% the specific binding of ¹²⁵I-α-bungarotoxin. The discrepancy between the concentrations of NTX effective as a non-depolarizing blocking agent and as an inhibitor of ¹²⁵I-α-bungarotoxin binding suggests that if NTX is interacting with the receptor recognition site of the acetylcholine receptor/ion channel, the site of its blocking action is distinct from the a-bungarotoxin binding site.

Only partial blockade of both compound EPSPs and EPSPs evoked by stimulation of a single cereal mechanoreceptor, was achieved by micromolar NTX. In the case of compound EPSPs, increasing the toxin concentration by about two orders of magnitude failed to cause further block of synaptic potentials. Since even single cereal mechanoreceptor afferents terminating on GI 2 make many synaptic contacts (J. M. Blagburn, D. J. Beadle & D. B. Sattelle, unpublished observations), it is possible that some of the synapses between cereal afferents and GI 2 are insensitive to NTX. Alternatively, all synapses may be partially blocked. The underlying mechanism remains to be determined.

A possible explanation for the synaptic blocking effect is an inhibition of the mechanism of transmitter release from presynaptic terminals. Although Deguchi et al. (1971) noted an inhibitory effect of NTX on acetylcholine release at the frog neuromuscular junction, Eldefrawi et al. (1980), following an analysis of quantal content and frequency of miniature endplate potentials, failed to confirm this observation. No comparable studies of possible presynaptic actions on transmitter release mechanisms are available for the cockroach cereal afferent, giant interneurone synapses. This possibility of a presynaptic action in insects cannot be discounted at the present time, though it should be recalled that acetylcholine responses of cell body membranes are highly sensitive to NTX and synaptic inputs to neuronal cell bodies of adult cockroaches have not been detected (cf. Smith & Treherne, 1965).

The voltage-dependence of the suppression by NTX of acetylcholine-induced currents in the cell body membranes of GI 2 and D_f, at a concentration of 5·0 × 10 ^{− 7} mol 1⁻¹, indicates that the site of this action of NTX is either the channel site of the acetylcholine receptor/ion channel or another component of this complex that is membrane-potential sensitive. The less pronounced voltagedependence of the block of acetylcholine-induced current at more negative (hyperpolarizing) potentials may result from the following possible actions of NTX: (a) an increase in lifetime of the acetylcholine receptor/ion channel; (b) a reduction in voltage-dependence of acetylcholine desensitization; (c) movement of positively charged NTX through the acetylcholine receptor/ion channel at large negative potentials. In preliminary experiments (cf. Sattelle & David, 1983) we have been unable to find evidence for voltage-dependent desensitization of acetylcholine responses in D_f. Further experiments are needed to distinguish between these possible explanations for the voltage-dependent behaviour of acetylcholine currents in the presence of NTX. In a recent study on D_f it was also shown that d-tubocurarine, an ion channel blocker at other neuronal nicotinic cholinergic receptors (see for example Ascher, Large & Rang, 1979), was voltagedependent in its blocking of acetylcholine-induced currents, whereas a-bungarotoxin, a probe of the acetylcholine receptor recognition site (Heidmann & Changeux, 1978), was voltage-independent (David & Sattelle, 1984).

The finding that NTX exhibits a voltage-dependent block of the acetylcholine response of insect neurones at submicromolar concentrations contrasts with the observations on vertebrate peripheral cholinergic receptors. Eldefrawi & Eldefrawi (1980) provide clear evidence that NTX does not alter the linearity of the current-voltage relationship, nor does it affect the time-course of the end-plate current in frog sartorius muscle. In the same study, the binding of [³H]perhydrohistrionicotoxin, a selective probe of the channel of the vertebrate nicotinic acetylcholine receptor/ion channel complex (Albuquerque et al. 1973; Masukawa & Albuquerque, 1978), was found to be unaffected by NTX at concentrations up to 1·0 × 10⁻³moll⁻¹.

The present study on Periplaneta reveals a postsynaptic depolarization by NTX accompanied by EPSP block, at concentrations of 2·0 × 10⁻⁵moll⁻¹ and above, and axonal depolarization at concentrations of 2·0 × 10⁻⁴moll⁻¹ and above. The concentrations effective in depolarizing the synaptic membrane are also effective in blocking the binding of ¹²⁵I-α-bungarotoxin to Periplaneta CNS extracts. One possible explanation is an agonist action on the recognition site of the insect nicotinic acetylcholine receptor/ion channel complex; however, actions at other voltage-sensitive channels and non-specific effects may also account for the postsynaptic depolarizing actions of NTX. NTX is a poor inhibitor (150: 1·0 × 10⁻³moll⁻¹) of flyhead acetylcholinesterase (Sakai, 1966c), so an action on this enzyme resulting in elevation of extracellular synaptic acetylcholine levels is unlikely to account for the postsynaptic depolarization. A depolarizing action of NTX has been observed at vertebrate neuromuscular junctions (Eldefrawi et al. 1980). The insect studies reveal a voltage-dependent blocking action at concentrations well below those required for depolarization.

The differences in the degree of voltage-dependence of the blocking actions of NTX at vertebrate peripheral nicotinic acetylcholine receptors and insect central nicotinic acetylcholine receptors provide further evidence for differences in the properties of these two receptors. Striking differences in the actions of décaméthonium at these receptors have already been noted (cf. Neher & Sakmann, 1975; Schmidt-Nielsen et al. 1977; David & Sattelle, 1984). The disulphide bond of NTX may be reduced to a dithiol which could covalently react with the acetylcholine receptor / ion channel complex in a manner similar to that shown for dithiothreitol, which is known to modify cholinergic receptor properties (cf. Karlin, 1969). The structure of this invertebrate neurotoxin therefore lends itself to chemical manipulation in pursuing further the molecular basis of its action. From the data presented here, an acetylcholine receptor/ion channel complex of the insect CNS appears to be one site of action of nereistoxin.

The authors thank Dr Michitiko Sakai of Takeda Chemical Industries Ltd, Kyoto, Japan for the supply of nereistoxin hydrogen oxalate. DBS acknowledges the support of the Agricultural Research Council, the Royal Society European Exchange Programme and the British Council. This research was supported in part by an award from the Council for Tobacco Research, U.S.A., Inc. to LMH, and a Wellcome Travel Fellowship to JAD.