Bacillus Thuringiensis δ-Endotoxin And Larval Manduca Sexta Midgut Brush-Border Membrane Vesicles Act Synergistically To Cause Very Large Increases In The Conductance Of Planar Lipid Bilayers

During sporulation, Bacillus thuringiensis produces parasporal inclusions with insecticidal activity. These parasporal inclusions consist of one or more δ-endotoxins, classified as Cry proteins (Hoefte and Whiteley, 1989). Type-I Cry proteins are active only against the larvae of certain lepidopteran insects. CryI inclusions dissolve in the alkaline midguts of lepidopteran larvae and the 130 000–140 000 M_r CryI proteins are digested to serine-protease-resistant toxins of 55 000–70 000 M_r. These toxins diffuse through the peritrophic membrane and bind to specific receptors on the brush-border membrane of midgut columnar cells. Within a few minutes, this binding becomes irreversible and soon thereafter the permeability of the cell membrane increases. This increase in cell membrane permeability leads eventually to cell lysis, disruption of gut integrity and finally to the death of the insect larva from starvation or septicemia (Knowles and Dow, 1993).

Although the overall course of a successful intoxication of a lepidopteran larva by a CryI B. thuringiensis δ-endotoxin is well established at the level described above, many questions remain unanswered at the molecular level. The concentration and affinity of toxin binding sites on its midgut brush-border membranes often, but not always, correlates positively with the susceptibility of a larva to a toxin (Wolfersberger, 1990). Very high concentrations of CryI toxins are able to form pores in phospholipid bilayers (Slatin et al. 1990). The irreversible step that follows binding of toxin to the brush-border membrane is thought to be associated with insertion of all or part of the toxin into the membrane. This insertion of toxin molecules into the membrane is believed to be necessary for the formation of a pore that mediates the potentially lethal increase in membrane permeability (Knowles and Dow, 1993). However, the size, net charge and composition of this membrane pore are unknown.

Membrane pores formed by CryI toxins in the presence of insect receptor proteins could consist of insect membrane molecules alone, toxin molecules alone or some combination of toxin and insect membrane molecules (Knowles and Dow, 1993). In the first case, the membrane would contain a pre-formed but closed pore that would open only upon binding by toxin or some natural insect ligand. The fact that there is no strict positive correlation between binding and toxicity, coupled with evidence that toxin not only binds to a receptor on the membrane surface but also enters the membrane bilayer during pore formation, makes the ligand-gated pore mechanism seem highly unlikely.

A comparison of the permeability properties of membrane pores formed by toxins in the presence and absence of receptors might be useful in differentiating between the two remaining possibilities. Pores consisting of toxin molecules alone might be expected to exhibit the same permeability properties whether or not receptor molecules are also present. However, pores composed of both receptor molecules and toxin molecules might differ significantly in their permeability properties from pores formed by toxin alone. The permeability properties of pores formed by CryIA(c) toxin in planar phospholipid bilayers at pH 9.6 have been determined quantitatively (Slatin et al. 1990). Several studies of the effects of CryI toxins on the permeability of lepidopteran insect tissue culture cells (Knowles and Ellar, 1987; Schwartz et al. 1991) and brush-border membrane vesicles (BBMVs) prepared from midguts of lepidopteran larvae (Sacchi et al. 1986; Hendrickx et al. 1989; Wolfersberger, 1989; Uemura et al. 1992; Carroll and Ellar, 1993) have been published. The results of studies with tissue culture cells must be interpreted with caution since the cell lines are not derived from midgut tissue and, although they show differential sensitivity to B. thuringiensis toxins, effects are seen only when toxin concentrations are much greater than those effective against susceptible insect larvae and do not necessarily parallel in vivo results (Witt et al. 1986). Midgut BBMVs have the advantage of being the true target of the toxins and contain authentic toxin-binding proteins. Although Hendrickx et al. (1989) attempted to measure the effect of toxin on alanine permeability, all but the most recent of the studies with BBMVs cited above used indirect methods that limited them to detecting changes in membrane permeability only for certain inorganic ions. With their light-scattering assay, Carroll and Ellar (1993) were able to study the effects of CryIA(c) toxin on the permeability of larval Manduca sexta midgut BBMVs to a variety of solutes. However, they were able to draw only qualitative conclusions about relative changes in permeability for the different solutes. By incorporating larval M. sexta midgut BBMVs into planar phospholipid bilayers, we have been able to measure directly and quantitatively the conductance of pores formed by CryIA(c) toxin in the presence of insect midgut proteins that interact specifically with this bacterial protein.

Bacillus thuringiensis subspecies kurstaki strain HD-73 was obtained from the Bacillus Genetic Stock Center (Columbus, OH). This strain contains genetic information for only one δ-endotoxin protein designated as CryIA(c) (Hoefte and Whiteley, 1989). The primary structure of this protein is known and its activity as a Manduca sexta larvicide is well established (Adang et al. 1985). Growth of bacteria and isolation of parasporal crystals as well as generation and purification of toxin were as described previously (Wolfersberger, 1989) except that, instead of using size exclusion chromatography, the toxin was separated from trypsin and other polypeptides by ion-exchange chromatography using a Mono-Q column eluted with a NaCl gradient as described by Hofmann et al. (1988). The purified toxin ran as a single band of M_r 67 000 on SDS–PAGE (Laemmli, 1970).

Manduca sexta were reared from eggs at 27 °C with constant light on an artificial diet purchased from Carolina Biological Supply (Burlington, NC). Midguts were isolated and brush-border membrane vesicles (BBMVs) were prepared from second- and third-day fifth-instar larvae, as described previously (Wolfersberger, 1989).

Crude granular soybean lecithin (formula no. 215, Vitamin Specialties Co., Wyncote, PA) was used in the lipid preparation. 15 ml of acetone (Fisher, Pittsburgh, PA) was added to 200 mg of the crude lecithin dissolved in 2 ml of n-decane (Sigma, St Louis, MO). The mixture was allowed to stand at 25 °C overnight. After the acetone/decane solution had been decanted or removed by aspiration, the phospholipid precipitate was resuspended in 15–20 ml of fresh acetone and allowed to reprecipitate overnight. The acetone was again removed by decantation or aspiration and the phospholipid residue was allowed to dry in air. When dry, the washed phospholipids were dissolved in 5 ml of n-decane and stored at 4 °C. This stock lipid preparation had a nominal lipid concentration of 3–4 mg per 100 ml and remained usable for at least 1 month.

The buffer solution was the same as that used by Slatin et al. (1990): 300 mmol l⁻¹ KCl, 5 mmol l⁻¹ CaCl₂, 0.5 mmol l⁻¹ EDTA and 5 mmol l⁻¹ Caps adjusted to pH 9.6 with KOH, unless stated otherwise in a figure legend. Buffer pH was checked daily.

5 ml Teflon beakers (cups) were used to support the lipid bilayers. These beakers were machined according to the pattern used by Mueller et al. (1962). The wall thickness on one side of the beaker was reduced and a 1 mm diameter hole was bored through the remainder of the wall in the center of the thinned portion. The beaker was positioned in the center of a 4 cm×4 cm×1 cm Lucite chamber. A depression was milled into the center of the bottom of the chamber to accept the Teflon beaker. Not more than 4 days before each use, the beakers and chambers were cleaned thoroughly following the procedures described by Alverez (1986).

All experiments were performed in an aluminum Faraday cage on an antivibration table (Technical Manufacturing Co., Peabody, MA). The chamber–cup assembly was positioned on a Lucite stage containing a small magnetic stirring motor (Instech Laboratories, Plymouth Meeting, PA). The stage itself was placed on a plastic laboratory jack (Bel-Art Products, Pequannock, NJ) to allow for vertical positioning of the chamber assembly. This arrangement permitted the hole in the Teflon cup to be viewed through a horizontally mounted stereo dissecting microscope. The hole in the Teflon cup was illuminated by a small, battery-powered light. This illumination allowed the thinning of the membranes to be followed visually. A controller for the stirring motor and a foot switch to operate the illuminator were both positioned outside the Faraday cage, the leads from these devices being shielded to prevent them from carrying electrical noise into the cage.

Voltage-clamping of the membrane and all electrical measurements were effected through the use of a patch-clamp amplifier (Dagan 3900A with a 3910 expansion unit, Dagan Corp., Minneapolis, MN). The control units were mounted outside the Faraday cage. A head stage (Dagan 3903) for the amplifier was mounted on a manipulator unit and connected to the amplifier via its shielded lead. The mount for the head stage also held two electrodes constructed from 1 ml disposable plastic pipette tips containing 1 mol l⁻¹ KCl in 3 % agar. The agar electrodes were back-filled with 1 mol l⁻¹ KCl and connected to the head stage via Ag/AgCl electrodes. In order to suppress oscillations in the head stage, a low-pass L-filter was inserted into the reference electrode lead. The whole assembly could easily be positioned as a unit (head stage, agar electrodes and their interconnections) so that the tips of the reference and test electrodes were in the chamber and cup compartments, respectively.

Output signals from the amplifier corresponding to membrane voltage and membrane current were connected to a strip chart recorder (model 1200, Linear Instruments, Reno, NV), a dual-trace oscilloscope (model D15 with 5A18N dual-trace amplifier and 5B10N time base/amplifier, Tektronics, Beaverton, OR) and, via a data aquisition interface (Flash-12, Strawberry Tree, Sunnyvale, CA), to a personal computer. These provisions allowed real-time monitoring of the experiments at different time resolutions. The output of the amplifier was also connected to a modified (Bezanilla, 1985) digital audio processor (Sony PCM-501ES, Unitrade, Philadelphia, PA) that allowed the digitized data to be stored on video tape cassettes for playback and analysis.

The Teflon beaker was pretreated by applying a small quantity of lipid stock solution directly to the hole in its side. The lipid was then allowed to dry either in air or under a stream of argon. The cup was placed into a chamber and both were filled with buffer solution. A magnetic stirring bar was place in the cup and the entire assembly was positioned on the stage over the stirring motor. A very small quantity of lipid stock solution adhering to the narrow end of a 10 μl pipette tip was placed near the hole in the Teflon cup and the small rubber bulb fitted to the large end of the pipette tip was squeezed gently. The air bubbles that formed carried the lipid to the Teflon cup and, after several bubbles had passed over the hole, a lipid seal was usually formed.

The thinning process leading to bilayer formation was followed visually through the microscope and electrically by monitoring the capacitance of the membrane. The capacitance was monitored by applying a 100 Hz, 5 mV peak-to-peak triangular wave to the membrane. The resulting current waveform was passed through a low-pass filter with a cut-off frequency at 100 Hz and processed by appropriate computer software (Workbench PC, Strawberry Tree, Sunnyvale, CA) to produce a sinusoidal waveform proportional to the membrane capacitance.

The seal of the bilayer lipid membrane (BLM) to the Teflon cup and the conductance of the membrane were determined by applying a four-step change in membrane potential, from –10 to +10 mV, and monitoring the changes in the membrane current. The steps lasted 5 s, which allowed sufficient time for membrane charge/discharge transients to dissipate completely. Experiments were conducted only on membranes with a seal resistance greater than 9 GΩ.

Fusion of BBMVs with a BLM was usually attempted only after a bilayer had formed and remained stable for at least 15 min. A 10 mV holding potential (inside positive with respect to the outside of the cup) was applied, the stirrer was turned on, and 5 μl of an 8–10 μg μl⁻¹ BBMV suspension were added to the buffer solution inside the cup. Addition of the vesicles was followed immediately by the addition of about 100 mg of KCl to the cup solution (final [KCl] in cup approximately 0.6 mol l⁻¹). After 30–60 s, the stirrer was turned off. Evidence for a successful fusion of a vesicle to the planar bilayer was taken to be a step-like change in membrane current. Following observation of one or more of these step-like increases in membrane current, the amplifier gain was reduced, the stirrer was restarted and 5 μl of 0.1 μg μl⁻¹ toxin solution was added to the cup. The stirrer was turned off 30–60 s after the toxin had been added.

A typical thinning process followed a time course of 10–30 min. It started with a gradual increase in membrane capacitance, followed by a sharp increase as black patches appeared, grew and merged, until the entire central portion of the membrane was black and the capacitance stabilized. Membranes with capacitances greater than 2.5 nF had very small annuli and were generally unstable. Membranes with capacitances less than 1.5 nF had large annuli and were not used.

Conductance measurements made on many thinned membranes revealed a relationship between seal conductance and the stability of the membrane. If the seal conductances were less than 50 pS, the membranes generated a relatively high-frequency noise indicative of membrane vibration. If the seal conductances were greater than 200 pS, the membranes gave indications of spontaneous bouts of temporary seal breakage. Membranes with seal conductances between 80 and 110 pS (12.5–9.1 GD seal resistances) were almost completely free of vibrational noise or spontaneous increases in membrane conductance for periods of at least 2 h. Only bilayer membranes with initial conductances between 60 and 110 pS were used in the studies reported in this paper.

The results of four experiments are combined in Fig. 1. Soon after addition of BBMVs and KCl to the stirred cup, step increases in membrane current were observed. Each step shown in Fig. 1A corresponds to a conductance increase of approximately 200 pS. Within 15 min after the addition of toxin to a BBMV-containing BLM, very large step increases in membrane current were observed. Each step shown in Fig. 1B–D corresponds to a membrane conductance increase of 13–56 nS. Additional step increases in membrane current often occurred with time; each adding to the total membrane current. In one experiment, membrane conductance eventually increased to a level exceeding that measurable with the recording equipment (>260 nS), while the continued presence of an intact BLM was confirmed by visual observation. Decreases in the current across toxin-treated BBMV-containing BLMs were never observed. At the concentrations used in these experiments, CryIA(c) toxin had no measurable effect on BLMs that had not been exposed to BBMVs.

Fig. 2 depicts the results of seven independent experiments in which BLMs were sequentially exposed to BBMVs and toxin. Data from portions of experiments A–D are also displayed in Fig. 1. In each experiment, the membrane conductance was less than 1.3 nS before addition of toxin. Experiments A–E were conducted at pH 9.6, whereas experiments F and G were conducted at pH 8.8. Experiments A–C were terminated after the first membrane current increase following toxin addition. These current increases corresponded to changes in membrane conductance of 13, 26 and 27 nS, respectively. Experiments D–G were allowed to run long enough to permit recording of more than one current increase. The smallest step increase in membrane current recorded in experiment D corresponded to a 56 nS increase in membrane conductance, whereas the largest corresponded to a 164 nS increase in membrane conductance. This very large increase in membrane conductance occurred in several steps. However, because of adjustments in amplifier gain required to accommodate large increases in membrane current, we were unable to resolve the individual steps. The smallest and largest increases in membrane conductance in experiment E were both 38 nS, whereas the smallest and largest increases in membrane conductance in experiments F and G, both conducted at pH 8.8, were 2 and 3 nS, respectively.

Careful examination, with high amplification and greatly expanded time scales, failed to reveal any changes other than random fluctuations in the magnitude of current passing through BBMV-containing BLMs bathed by pH 9.6 salt solution. However, addition of CryIA(c) toxin to the salt solution bathing such membranes gave rise to step increases in membrane conductance ranging from 13 nS to greater than 50 nS, with changes of approximately 26 nS and 39 nS occurring most frequently. All of the various step increases in membrane conductance were, within experimental error, some multiple of the smallest increase measured. The quantal nature of membrane conductance increases suggests that the conductance of a single membrane pore could be 13 nS.

13 nS is approximately one order of magnitude greater than the conductance, under similar conditions, of the largest pores formed in planar phospholipid bilayers by CryIA(c) toxin alone (Slatin et al. 1990; F. G. Martin and M. G. Wolfersberger, unpublished observations). Furthermore, the channels formed in BLMs composed only of phospholipids alternate between open and closed states on a time-scale of seconds, whereas those formed in BLMs that contain insect midgut BBMVs have never been observed to close. Such major differences in the properties of pores formed in the presence and absence of insect toxin-binding proteins favor active involvement of insect proteins in pore formation.

Assuming that the insulating hydrocarbon portion of the bilayer is 2.3 nm thick (Hille, 1992), a cylindrical pore with a conductance of 13 nS when filled with the solution used in the experiments reported here would have a diameter of 2.2 nm. This is approximately twice the diameter of the pore required to allow passage of the largest solutes known to enter toxin treated cells or BBMVs (Knowles and Ellar, 1987; Carroll and Ellar, 1993).

M. sexta larvae maintain a standing pH gradient in the lumen of their midgut. The pH in the most anterior portion of the midgut lumen is slightly greater than 10. The pH gradually increases down the lumen to a maximum of 11.3 in the central portion of the midgut and then decreases to near 8 in the most posterior region of the midgut. However, the pH is in excess of 9.5 in all but the most distal 15–20 % of the midgut lumen (Dow, 1984). Knowles and Dow (1993) cite unpublished work suggesting that both the conductance and the ion selectivity of the pores formed in BLMs by CryI toxins can change dramatically with pH. The magnitude of conductance increases following exposure of larval M. sexta midgut BBMVs containing BLMs to CryIA(c) toxin also changes dramatically with pH (Fig. 2). Carroll and Ellar (1993) conducted all of their experiments at pH 7.5. The pores that we observed at pH 8.8 may be more similar in size and permeability to those found by Carroll and Ellar (1993) in toxin-treated BBMVs at pH 7.5 than are the pores we found in toxin-treated BBMV-containing BLMs at pH 9.6. However, considering the pH in the larval midgut, the larger pores seen at higher pH may resemble more closely those formed in the apical membranes of midgut columnar cells in vivo.

We thank Drs P. Mueller and J. C. Tanaka for their help and encouragement in setting up a functional planar bilayer system. We thank D. A. Kozikowsky, D. D. Spaeth and J.-Y. Wang for their assistance in rearing of larvae, culture of bacteria, isolation of parasporal crystals, preparation of BBMVs and purification of toxin. We thank Dr W. R. Harvey for critical comments and useful suggestions on our manuscript. This research was supported by grants from the US National Institutes of Health (GM-41766) and Entotech, Inc.