Insect blood-brain barrier: an electrophysiological investigation of its permeability to the aliphatic alcohols

Pharmacological and toxicological studies on the insect central nervous system are complicated by the presence of an ionic diffusion barrier between the neuronal surfaces and the bathing medium, the existence of which was first demonstrated by Hoyle (1953) in the locust, and subsequently in the American cockroach by Twarog & Roeder (1956). In both cases the method was electrophysiological, it being shown that high-potassium saline did not immediately abolish axonal conduction unless the nerve sheath was removed. Ultrastructural studies using extracellular tracers (Lane & Treherne, 1970) have suggested that the barrier is as a consequence of lateral occlusions between the perineurial cells, which form an epithelial layer below the fibrous zone of the sheath. Toxicological data (O’Brien & Fisher, 1958) have also been interpreted in terms of such a barrier, and a fuller understanding of the permeability of this barrier to organic molecules is of considerable relevance to insecticidal research, as many of these agents act within the central nervous system (Narahashi, 1971).

The most common method of studying the movements of substances is to use radioisotopes, but although such experiments are simple to perform, they are less easy to interpret, as they generally provide no indication of the distribution of the activity within the tissue. In contrast to the electrophysiological findings, Treherne (1961 a-e, 1962 a, b), using radioisotopes, demonstrated a relatively rapid exchange of both monovalent and divalent cations between the bathing medium and the central nervous tissues of Periplaneta americana, and similar results were later obtained by Tucker & Pichon (1972), who demonstrated that 45% of the exchangeable sodium could be washed out of the central nervous connectives of this insect with a half-time of 30 s. The electrophysiological method measures the accessibility of substances only to the neuronal surfaces, whereas the radioisotope method measures the accessibility to all the compartments of the central nervous system. Thus the discrepancy between the results obtained from the two methods suggests that the accessibility of ions to the neuronal surfaces is considerably lower than to other compartments of the central nervous system. Presumably, the difference is due (at least in part) to the existence of active transport processes, and although it is unreasonable to postulate such mechanisms for the majority of other substances, there could well be other effects (such as binding) which may cause the accessibility of organic molecules to the neuronal surfaces to differ markedly from that to other compartments of the tissue. It was this possibility which prompted the present studies.

The most comprehensive study so far made of the permeability of the insect central nervous system to organic molecules is that of Eldefrawi, O’Brien and co-workers. Using the tracer technique, they studied the movements of fatty acids (Eldefrawi & O’Brien, 1966), quaternary ammonium salts (Eldefrawi & O’Brien, 1967 a), alcohols (Eldefrawi & O’Brien, 1967b) and also the effect of polarity (Eldefrawi et al. 1968). Both influx and efflux of all compounds studied were slow. The efflux was always biphasic, with respective half-times of a few minutes and a few hours, and the effect of de-sheathing the ganglia was comparatively small, although it was greater for positively charged compounds. On the basis of the data obtained from quaternary ammonium salts they concluded that increasing liposolubility ‘tended to increase influx’, and increasing size decreased it. However, influx of the fatty acids was faster than that of the analogous alcohols, even though they are less liposoluble.

These findings suggest a very complex and somewhat paradoxical situation, and it was hoped that an electrophysiological study of accessibility would go at least some way towards its clarification, as well as being of greater pharmacological and toxicological relevance. The technique involves the use of the anaesthetic effect of a substance as a measure of its concentration at the neuronal surfaces, so it is limited to those substances for which the effects are simple and replicable. The properties of the aliphatic alcohols are, however, ideal for the technique.

These molecules have long been known to have anaesthetic action, and belong to the class of biological depressants, i.e. those compounds which have a biological activity unrelated to their structure (cf. Albert, 1973). The action of such compounds was independently suggested by H. Meyer (1899) and Overton (1901) to be in some way related to their solubility in biological lipids, but their precise mode of action is still unclear. K. Meyer & Hemmi (1935) suggested that isonarcotic effects were produced by such substances when their concentrations in the cell lipids were identical, but Ferguson (1939) pointed out that thermodynamic activity was a more valid parameter, on the grounds that narcosis was an equilibrium condition and hence the activity of the anaesthetic in all the phases of the tissue must be identical. This provides an explanation for highly liposoluble substances not having anaesthetic properties. As a homologous series is ascended, the thermodynamic activity for an isonarcotic effect increases, eventually reaching that of a saturated solution, so higher homologues have little or no effect. The member of the series at which this ‘cutoff’ occurs depends on the sensitivity of the tissue, but for nerve conduction blocking by the alcohols, this point appears to be reached at about C10 (see, for example, the data in Seeman, 1972), so only the lower homologues were used in these experiments.

The experiments were performed on isolated abdominal nerve cords of the cockroach, Periplaneta americana. These individuals were reared on a standard diet in this Laboratory, and only adult male specimens were used. The design of the experimental chamber is described elsewhere (Thomas & Treherne, 1975). It allowed simultaneous microelectrode and ‘sucrose-gap ‘(Stämpfli, 1954) recordings to be made from the penultimate connectives. The chamber consisted of five adjacent compartments which were isolated from each other by seals of ‘electrical’ silicone grease. The two left-hand compartments contained normal saline solution and were connected via platinum wires and an RF-coupled isolating unit to a Farnell pulse-generating system. The central compartment contained the experimental solution flowing at a regulated rate and was connected to the reference electrode (which was common to both recording systems) via a KCl-Agar bridge. The fourth compartment contained a continuously-flowing isotonic mannitol solution. The fifth compartment was filled with normal saline and was connected via a KCl-Agar bridge to a high-impedance amplifier with input capacity compensation (Unwin & Moreton, 1974). Microelectrodes were pulled from Corning capillary glass tubing, and were filled with 3 m KC1; tip resistances were approximately 15 MΩ. They were connected to a second high-impedance amplifier, and both amplifiers were connected to a Tektronix 502 A oscilloscope. The output of the microelectrode amplifier was connected to a Servoscribe RE520.20 pen recorder and to a peak level measuring device (Thomas, 1974, 1976) the output of which was connected to the other channel of the recorder. The microelectrode signal was also sent to a differentiator circuit, the output of which could be displayed on the oscilloscope in place of the ‘sucrose-gap’ signal.

The preparation was laid across the five compartments so that the connectives between the fourth and fifth ganglia were continuously irrigated by the test solution. The nerve cord was supported in this region by a small block of rubber, to facilitate impalement by a microelectrode. A small vinyl ring fitted into the top of this compartment, its purpose being to ensure that the connectives were at all times fully immersed and to hold them against the rubber block.

The choice of experimental solution and its flow rate were accomplished by control valves of the design shown in Fig. 1. The two opposing 6BA screws crimp the rubber tubing, the screw on the left being set so that the tube is totally occluded when the control lever is in a fully clockwise position. Moving back this lever unscrews the thread to which it is attached, allowing solution to pass down the tube, the flow rate being continuously variable. This design enabled the solutions to be changed very rapidly, whilst avoiding any violent changes in the rate of flow, which could dislodge a microelectrode. The possibility of any mixing of the experimental solutions during the changeover periods was avoided by use of a multiway non-retum valve, similar to the design of Holder & Sattelle (1972), which was situated immediately prior to the test compartment. Test solutions were perfused at approximately 3 cm³ min⁻¹, and experiments with dye solutions showed that exchange of solution in the test compartment was effectively complete in just under 1 s. Although short, this delay was nevertheless taken into account during analysis of the results, and it is estimated that the residual uncertainty in the timing of the solution changes was about 0-2 s.

The normal saline (Yamasaki & Narahashi, 1959) contained 214 mM Na⁺, 3·1 mM K⁺, 1·8 mM Ca⁺⁺, 216·9 mM C⁻ 0·2 mM Hg₂PO₄^{− −} and 1·8 mM HPo₄^{− −}. This saline has higher sodium and lower potassium ion concentrations than those in cockroach haemolymph (Pichon, 1970), and it has been shown to cause weight loss in intact nerve cords (Evans, 1975). Alternative salines for use with intact nerve cords have, therefore, been suggested (Evans, 1975; Treherne, Buchan & Bennett, 1975) but it appears that the Yamasaki & Narahashi saline is a close approximation to the composition of the extraneuronal fluid (Thomas & Treherne, 1975). The electrical responses of intact nerve cords are apparently normal in this saline, and its use in all experiments simplified comparison between the results from intact and desheathed preparations. The high-potassium saline, which was used to test the effectiveness of the blood-brain barrier, contained 214 mM K⁺ and 3·8 him Na⁺, but its composition was otherwise similar. The alcohols were dissolved directly into the saline, and these solutions were frequently made up.

Desheathing of the penultimate connectives was performed with electrolytically sharpened tungsten needles, immediately before transfer of the nerve cords to the experimental chamber. All experiments were performed at room temperature, which was 24 ± 1 °C.

Preliminary results were obtained with ethanol and n-butanol, using only the ‘sucrose-gap’ technique. This has the advantages of simplicity and avoidance of possible ‘pinholing’ of the perineurium by a microelectrode. In many of these experiments, sub-blocking concentrations of the alcohols caused some instability of the action potential amplitude, which made interpretation difficult. The instability was apparently related to the occurrence of spontaneous and repetitive activity in the nerve cords, and similar behaviour under these conditions has been reported for axons of the squid (Moore, Ulbricht & Takata, 1964) and the lobster (Houck, 1969). It was nevertheless possible to make approximate estimates of the apparent influx and efflux half-times, and Fig. 2 shows the results of such an experiment with n-butanol on an intact preparation. The influx half-time is given by the time required for the action potential amplitude in 100 mM butanol to decline to the final value observed in a 50 mM solution, and the efflux half-time is determined in a complimentary manner. In spite of the inherent inaccuracies of the technique, it is clear that n-butanol moves extremely rapidly to and from its site of action in the nerve cord, the half-times being of the order of 10 s.

Since the movements are so rapid, they cannot be significantly affected by ‘pinholing’ by a microelectrode. Intracellular recording was used for all subsequent experiments, as there were no instability effects with this technique (although repetitive firing still sometimes occurred). It is greatly preferable to measure the steady-state axonal responses to a series of concentrations of each alcohol, so that a dose-response relation can be obtained. This can then be used in conjunction with the time-course of the effect of a high concentration of the alcohol, to obtain the complete time-course of the concentration changes.

Although the effects on any one preparation were fully replicable, there was a considerable variation in sensitivity between preparations. It was therefore impossible to establish a reliable dose-response relation by statistical means, so one had to be obtained separately for each preparation. Although it added considerably to the technical difficulties, the procedure had the advantage that each experiment gave a totally independent result. As the procedure was essentially identical for all experiments, it is illustrated by showing in detail the results of an experiment with n-butanol on an intact preparation.

The effect of a sub-blocking concentration (75 mM) is shown in Fig. 3. This result clearly shows that, at least on the time-scale relevant to these experiments, butanol reduces the action potential amplitude of a giant axon to a stable value, and that the effect is fully reversible. The same applies to the maximum rates of rise and fall of the action potential (although the former is rather more affected), which are expected to be closely related to the peak sodium and potassium conductances. In this respect the finding is in agreement with that of Armstrong & Binstock (1964), who reported that the effect of several alcohols on the squid giant axon was to reduce the peak sodium conductance by rather more than the peak potassium conductance.

The complete dose-response relation obtained from this experiment is shown in Fig. 4. The steady-state action potential amplitude has been plotted, as a fraction of its value in normal saline, for a series of butanol concentrations. The experiment followed the normal procedure, which was to apply the concentrations in ascending order, allowing the preparations to recover in normal saline between each. The first concentration was then repeated as a check of reproducibility. Any shift was normally small, and none was detectable in this experiment. In some experiments, the action potential did not always recover exactly to its original value between exposure to the alcohol solutions, instead being up to 2–3 mV higher or lower, and in these cases the relative action potential amplitude was expressed as the fraction of the mean of the initial and final values.

The points in Fig. 4 have been joined by a smooth curve, which has no theoretical significance, but the relation is very nearly linear for butanol concentrations up to 75 mM. At concentrations sufficient to reduce the action potential amplitude to below about 75 % of its normal value, its amplitude was found to be rather less stable, and to decline far more sharply with increasing alcohol concentration. The instability was associated with conduction failure, and estimates from this region of the dose-response relation were less reliable. The resting potential was scarcely affected. There was a small depolarization which was approximately proportional to the butanol concentration, it being only 4 mV at 125 him. The effects of the other alcohols on the resting potential were in general even smaller, except for ethanol, where they were approximately twice as great, but were nevertheless fully reversible.

The effect of 125 mM on the action potential amplitude is shown in Fig. 5. Each point of the graph was expressed as a fraction of the normal amplitude, and converted into an equivalent butanol concentration by reference to the dose-response relation. The influx kinetic for this experiment is shown in Fig. 6a. A plot of log {1 – (Ct/C test)} (where Ct is the calculated concentration at the giant axon surfaces at time t) was used, as it yields a linear relation if the influx follows first-order kinetics. The data from the experiment fitted the relation remarkably well, apart from the last two points on the graph (open circles). They correspond, however, to butanol concentrations in excess of 75 mM, which is on the non-reliable region of the dose-response relation. For this reason they were ignored in calculation (by least squares) of the best fitting straight line, which corresponds to a half-time of arrival at the giant axon surface of 10·4 s.

It is possible to perform a similar analysis for the efflux of the alcohol during recovery, by plotting log (Ct/C test) against time, and this is shown in Fig. 6b. Although the fit is less good, it is clear that the rates of influx and efflux are not significantly different. Apart from this observation, it would be unwise to place as great an importance on the efflux estimates, because exposure of the connectives in the test compartment for a period sufficiently long to allow complete equilibration of the alcohol is expected to result in a significant diffusion of the alcohol into the adjacent regions of the cord, which would reduce the rate of recovery on return to normal saline. The effect is expected to be greater for less permeant molecules, and for that reason, efflux estimates were made only for n-butanol, as it was the most permeant alcohol. As expected, the mean efflux half-time was slightly longer than that for influx (Table 1).

The same procedure was followed for all the other experiments, except those for octanol. The rate of action of this alcohol was too slow for a microelectrode impalement to be maintained for sufficiently long to determine a complete dose-response relation, so the simpler technique shown in Fig. 2 had to be used.

The rates of arrival of the other alcohols at the axonal surfaces did not fit a first-order relation quite so well, and the results with ethanol showed the most extreme departure, a typical one being shown in Fig. 7a. There is an initial lag of several seconds before there is any significant change, but after that a first-order kinetic is followed quite closely. To obtain the best straight-line fit in the later section, the first few points (open circles) were ignored. Although a more complex analysis could, perhaps, have been performed, the present one always gave a good fit around the half-time (corresponding to log {1 – (Ct/C test) =–0·3}, which is the most useful point of comparison. It had the further advantage that the departure from a first-order relation was clearly indicated by the duration of the ‘lag’ period. For the other alcohols the lag period was far shorter, and for comparison, the result of an experiment with hexanol is shown in Fig. 7b.

The results of all the experiments are summarized in Table 1. Although the lag periods are shown separately, they are also included in the half-time estimates, so they should be subtracted to obtain the first-order rate constants. This presentation has been used to allow direct comparison with the octanol results, as the simpler analysis technique (cf. Fig. 2) which had to be employed for this alcohol gives no indication of the length of any lag period. The most striking aspect of the results is that, particularly for the lower alcohols, the half-times are extremely short and are little affected by removal of the nerve sheath. Two main lines of evidence support the validity of the results.

The first concerns the alcohol concentrations used in these experiments. The concentrations necessary to cause conduction block are expected to be similar to those observed for other systems, and to be closely related to their lipid/water partition coefficients. In view of the shape of the dose-response relation in these experiments (cf. Fig. 4), a rigorous definition of the blocking concentration is impossible, but because of the sharp decline in the relation for all the alcohols below a relative amplitude of about 0·75, this level was chosen as a basis for comparison. The results are shown in Table 2. There was no significant difference in sensitivity between intact and desheathed preparations, so all the results have been pooled together. The values are very similar to those observed for the frog sciatic nerve at a similar temperature (22 °C), as quoted in the review by Seeman (1972). The octanol/water partition coefficients shown in this table are taken from Leo, Hansch & Elkins (1971), and the coefficients for membrane lipid are expected to be about five times lower (Seeman, 1972). The relation between blocking concentration and partition coefficient for the present results is shown in Fig. 8. The slope of the regression line is –0·92, and the departure from a 1:1 relationship may not be significant, since there are only four data points and graphs of this type usually show considerable scatter.

As no dose-response relation was obtained for octanol, it was not possible to include this alcohol in the analysis. Preliminary experiments with this substance suggested, however, that the blocking concentration was approximately 1 mM, which is in reasonably close agreement with the value predicted by extrapolation of the graph.

The second line of evidence concerns the effects of high-potassium saline on anaesthetized cords. Treherne, Schofield & Lane (1973) have shown that a short| exposure to 3·0 M urea solution considerably reduces the ionic diffusion barrier in this preparation, so it could be argued that the alcohols may have a similar effect, which would explain why the nerve sheath apparently offered such a small barrier to the diffusion of these molecules. It was shown, however, that intact connectives were still unaffected by high-potassium saline, and developed normal extraneuronal potentials (cf. Treherne et al. 1970) after exposure to the alcohols.

This does not, however, rule out the possibility of reversible damage of the barrier, and the possible occurrence of such an effect was investigated by experiments of the type shown in Fig. 9. The experimental procedure was to expose an intact preparation to a high-potassium saline, and then to a high concentration of the alcohol (200 mM n-butanol in this case) in the continued presence of high potassium, so if the barrier were damaged by the alcohol, potassium ions would be able to enter. The preparation was then again exposed to high-potassium saline in the absence of the alcohol. Thus if potassium ions had entered during exposure to the alcohol, they would have remained there, preventing the neurones from recovering. Fig. 9 clearly show this not to be the case, and for comparison, the rapid effect of this saline on a desheathed preparation is also shown. This result shows that n-butanol does not disrupt the barrier, even at nearly three times the mean blocking concentration. Similar results were obtained for the other alcohols, although concentrations nearer to those required to cause conduction block were used.

The most surprising aspect of these results is that the insect nerve cord is apparently far more permeable to the alcohols than was suggested by the results of Eldefrawi & O’Brien (1967b), who used radioisotopes. According to these authors, efflux of the alcohols from intact cords was biphasic, the half-times of the two phases for n-butanol being 2·5 and 220 min, corresponding to 22 % and 78 % of the total uptake, and very similar results were reported for the other alcohols. The reasons for the discrepancy are more fully discussed elsewhere (Thomas, 1974, 1975), where evidence is presented suggesting that these slow components are artifactual, resulting from metabolism and possibly binding. The experimental technique employed by these authors was unsuitable for the detection of the more rapid movements observed in the present experiments, but by use of a more appropriate technique, radioisotope efflux components with half-times of a few seconds have been demonstrated (Thomas, 1974, 1976b).

The alcohol concentrations used in the present experiments were somewhat higher than those used by Eldefrawi & O’Brien, but the experiment shown in Fig. 9 demonstrates that they do not damage the nerve cord. There have been reports that alcohols markedly alter membrane permeability, but they have been based on the use of still higher concentrations. For example, Burt & Green (1971) reported that 400 mM n-butanol increased sodium ion influx in human erythrocytes 20–40 fold (cf. 200 mM in Fig. 9), but this concentration irreversibly blocks the frog sciatic nerve (Skou, 1954), suggesting severe damage of the axonal membranes. In contrast, Roth & Seeman (1972) reported that potassium ion leakage from erythrocyte ghosts was not increased by the alcohols at concentrations sufficient to cause conduction block in nerve, and that for n-butanol the concentration had to be at least 350 mM for there to be any significant increase.

Although the precise mode of action of the alcohols and other general anaesthetics is still unclear, it presumably results directly from their dissolution in the membrane, and the evidence is strongly against the involvement of any specific receptor. It appears that the functioning of the sodium and potassium channels is adversely affected as a result of an increase in fluidity and disordering of the axonal membranes by the anaesthetics ; such increases have been demonstrated by nuclear magnetic and electron spin resonance techniques (Metcalfe & Burgen, 1968; Paterson et al. 1972). The alcohols are expected to dissolve in the axonal membrane very rapidly on arrival at the axonal surfaces, as the process is exergonic by about 800 cal mole⁻¹ per methylene group (Schneider, 1968) and there is no reason to suggest a significant activation energy. It is therefore unlikely for there to be any significant delay between arrival of the alcohols at the axonal surfaces and the onset of conduction block. Even if the delay were significant, its effect would be to increase the calculated half-times beyond the actual ones, so the values in Table 1 are maximum estimates.

Although the experimental half-times are very short, they are, nevertheless, significantly longer than those predicted on a basis of unimpeded aqueous diffusion, suggesting that the movements of the alcohols are significantly retarded by the tissue. Hill (1928) has calculated that the average degree of saturation of a cylindrical tissue is 50% when (Dt/r²) = 0·06 (Fig. 5 of his paper), where D is the diffusion coefficient, which for n-butanol is 7·7 × 10⁻⁶ cm² s⁻¹ (Davson & Danielli, 1952), and r is the radius of the tissue, which is about 70 µm. The calculated time t for 50 % saturation of 0·4 s is more than an order of magnitude faster than that observed, and the difference for the other alcohols is even greater.

A comparison of the present results with those obtained from other systems is complicated by the fact that the movements of very liposoluble substances have been little investigated, the partition coefficients of the majority of molecules tested being below unity. The results obtained suggested that permeability was closely related to the partition coefficient, notably those with giant algal cells (Collander, 1949) and rabbit gallbladder (Diamond & Wright, 1969). The main departures from the relation were for substances with molecular weights below about 60, which penetrated more rapidly than expected (even after allowance for their higher diffusion coefficients), and for ones which were highly branched, which penetrated more slowly.

The present results for ethanol and the two butanol isomers can be explained satisfactorily in the light of the above findings. The difference in influx half-times in intact connectives for the two butanol isomers is approximately proportional to the difference in their liposolubility, although the branched chain of the tertiary isomer may also contribute to the difference. Ethanol has a molecular weight of only 46 (compared with 74 for butanol), which could explain why its penetration is more rapid than expected from its liposolubility. The effects of desheathing for these three alcohols is small, being significant only for tertiary butanol. The finding is not entirely unexpected, as there remains a significant barrier to cationic diffusion after removal of the nerve sheath, the mean half-time for the arrival of potassium ions at the axonal surfaces being 24 s (Treherne et al. 1970). The small effect of desheathing on the access half-times is thus not surprising, as they are of the same order even in intact connectives. These findings are, however, in very strong contrast to those obtained for more polar molecules such as tetrodotoxin, to which the nerve sheath is an extremely effective barrier (Thomas, 1974).

Wartiovaara (1949) reported that the permeability of giant algal cells increased through the alcohol series from methanol to butanol, with which the present findings are broadly in agreement, but the higher alcohols were not investigated. In fact, the higher primary alcohols have not been investigated in any of the classic permeability studies, so the results from previous investigations are of little help in explaining the observation that beyond butanol, increasing liposolubility increases the access half-time. The finding that the nerve sheath significantly retards the rate of access of hexanol and octanol, although the barrier is lipophilic and these alcohols are more liposoluble than n-butanol (for which the sheath apparently offers no retardation), would also appear to be paradoxical. The access of hexanol and octanol is expected to be slowed by their higher moleculer weights, but since diffusion coefficients are proportional to the square root of the molecular weight, this factor could slow the influx rate of octanol relative to butanol by only about one third, which is far smaller than the observed differences. The extent of the differences after correction for this factor is shown in Fig. 10, and the dotted line represents the relationship predicted for intact connectives assuming a lipid barrier, based on the best fit for the butanols (cf. Fig. i of Collander, 1949). The results can be explained, however, by the following hypothesis.

The membrane concentrations of the alcohols (and other general anaesthetics) necessary to cause conduction block are approximately equal, as judged by the relation between aqueous concentration and partition coefficient (Fig. 8). The membrane/saline partition coefficient of n-butanol is expected to be about five times lower than the octanol/water coefficient (Seeman, 1972) (i.e. about 1·5). Similarly, the membrane coefficients for hexanol and octanol are predicted to be 22 and 280.

The lipids in the central nervous system form a reservoir which the alcohols must fill to cause anaesthesia. Although the size of the reservoir in terms of the number of molecules is about the same for all the alcohols, its size relative to the number of molecules in the aqueous phase necessary to cause anaesthesia is proportional to the partition coefficient of the alcohol. The aqueous concentrations of the higher alcohols, particularly octanol, are thus expected to be buffered very strongly by the uptake into the lipid phase, which would significantly slow the rate of equilibration. The extent of such buffering would depend on the total lipid content of the nerve, and on the basis of estimates for other species (Table 13 of Treherne, 1966) it could be as high as 10% of the fresh weight. In this case, the total lipid uptake of n-butanol is expected to be about 15% of that in the aqueous phase of the nerve cord, whereas the value predicted for octanol is nearly 30 times greater than the aqueous uptake, and the effect could easily outweigh the expected increase in the rate of access predicted from the higher partition coefficient. The significant decrease in access half-times for hexanol and octanol as a result of removal of the nerve sheath can be explained in terms of a reduction in size of the lipid reservoir, rather than of the removal of a discrete diffusion barrier.

The results suggest that substances with membrane/saline partition coefficients near unity have the shortest access half-times. Those with higher coefficients are slowed by the increased reservoir effect of the lipid phase, and those with lower coefficients are slowed because the membranes present a greater barrier to their diffusion.

I thank my supervisor, Dr J. E. Treherne, for helpful discussion during the course of this work. Financial support was from a CAPS studentship, awarded jointly by ICI Plant Protection Ltd and the Science Research Council.