Structurally related Bacillus thuringiensis δ-endotoxins display major differences in insecticidal activity in vivo and in vitro

The intracytoplasmic protein crystal synthesized during sporulation of the grampositive bacterium Bacillus thuringiensis is the major cause of toxicity of strains of this organism for insects. On the basis of their flagellar H-antigens, the 33 varieties of B. thuringiensis have been divided into 22 serotypes (de Barjac et al. 1985). For many of these varieties there are differences in the type of insect that is killed by the (5-endotoxin crystals. Many of the serotypes produce d-endotoxins pathogenic to caterpillars (Lepidoptera); others kill mosquito and blackfly larvae (Diptera). One recent isolate, B. thuringiensis var. tenebrionis kills beetles (Coleóptera; Krieg et al. 1983), while others have not yet been shown to be toxic to any insect tested.

The crystal δ-endotoxin is a protoxin, activated after larval ingestion by the high pH and proteases of the larval gut (Lecadet & Martouret, 1967). Once activated, the toxin attacks the cells of the midgut epithelium, which rapidly swell and lyse (Ebersold et al. 1978; de Barjac, 1978; Endo & Nishiitsusuji-Uwo, 1980; Percy & Fast, 1983). Similar morphological changes are observed in larval cell lines in vitro (Murphy et al. 1976; Nishiitsusuji-Uwo et al. 1979; Thomas & Ellar, 1983a). Susceptible cells become rapidly leaky to small ions, dyes and internal markers (Fast & Donaghue, 1971; Ebersold et al. 1978, 1980; Gupta et al. 1985).

The biochemical basis of the specificity of these toxins is one of the many unanswered questions in the study of these commercially important organisms. Among the possible explanations for this specificity are: the ability of larval gut proteases to solubilize and activate the protoxin, the presence or accessibility of specific toxin receptors in the insect gut, the composition of the insect diet (Luthy et al. 1985) or the structure of the δ-endotoxin itself. It has been shown that in the case of B. thuringiensis var. colmeri, the source of gut proteases plays a critical role in determining the pattern of proteolytic processing of the protoxin and hence the insect specificity of the δ-endotoxin (Haider et al. 1986). The δ-endotoxin of B. thuringiensis var. israelensis has certain ubiquitous phospholipids as its receptors (Thomas & Ellar, 1983d) but even when activated it is not toxic to lepidopteran larvae, whose gut cells apparently contain the receptors in a form or location inaccessible to the toxin. Crystal δ-endotoxins from different serotypes may vary in their polypeptide composition (Tyrell et al. 1981; Ellar et al. 1985). At least three antigenically distinct toxic proteins have been described: B. thuringiensis var. kurstaki HD-1 contains a 130×10³M_r lepidopteran toxin (Pl) and a 63 ×10³M_r protein (P2) toxic to both Lepidoptera and mosquitoes (Yamamoto & McLaughlin, 1981) and B. thuringiensis var. israelensis contains a 27 ×10³M_r mosquito toxin (Ward et al. 1984).

Analysis of the genes of a number of 130 × 10³M_r lepidopteran toxins has revealed extensive protein sequence homology with the Pl toxin of HD-1 (Schnepf et al. 1985; Shibano et al. 1985; Adang et al. 1985). Despite this homology, bioassays show that these proteins differ substantially in specificity and potency. These data suggest that minor modifications in the sequence of the toxins may yield major differences in their ability to recognize and bind to insect-specific receptors. To investigate this possibility we have compared the specificity and potency of three antigenically cross-reacting δ-endotoxins. The results demonstrate that when these similar toxins are activated under identical conditions they do indeed show very different toxicity spectra.

Bacillus thuringiensis var. kurstaki HD-1 was obtained from Dr H. D. Burges (Glasshouse Crops Research Institute, England); B. thuringiensis var. aizawai HD-249 and B. thuringiensis var. thuringiensis HD-350 were from Dr H. T. Dulmage (USDA, Brownsville, Texas). Conditions for growth and sporulation were as described by Stewart et al. (1981) for B. megaterium KM. The crystals were separated from spores and vegetative cell debris by ultracentrifugation on discontinuous sucrose gradients as described by Thomas & Ellar (1983a) for B. thuringiensis var. kurstaki.

Protein concentration was measured by the method of Lowry et al. (1951) using bovine serum albumin (Sigma) as a standard.

SDS-PAGE was carried out by the method of Laemmli & Favre (1973) as modified by Thomas & Ellar (1983a).

Solubilized Pl δ-endotoxin from B. thuringiensis var. kurstaki, processed by endogenous proteases to 54 ×10³M_r (Knowles et al. 1984) was mixed with Freund’s complete adjuvant and administered by subcutaneous injection into New Zealand White rabbits. In subsequent injections at 10-day intervals Freund’s incomplete adjuvant was used. Blood was obtained from the ear vein and serum collected after clotting. Antiserum towards var. kurstaki P2 protein, purified by preparative PAGE, was kindly donated by Mr F. Drobniewski (University of Cambridge).

Crystal proteins separated by SDS-PAGE were transferred electrophoretically to nitrocellulose filters (Schleicher & Schiill) by the method of Towbin et al. (1979) using a BioRad ‘Trans-Blot’ apparatus. Immunoblotting was carried out as described by Hawkes et al. (1982) and detection of bound peroxidase conjugated goat anti-rabbit immunoglobulins (Sigma) was with 4-chloro-l-naphthol and H₂O₂.

Crystals were dissolved by incubation in freshly prepared 50mM-Na₂CO₃/HCl, pH 9 ·5, and 10mM-dithiothreitol (DTT) for 60min at 37°C (Huber et al. 1981). Insoluble material was removed by centrifugation at 10000gfor 5 min. The fractions obtained by this method are referred to as soluble and insoluble crystal proteins, respectively.

The soluble fraction was incubated 10:1 (v/v) with gut extract from Pieris brassicae or Aedes aegypti larvae at 37°C for 30 min. Preparation of P. brassicae gut extract was described by Knowles et al. (1984). A. aegypti gut extract was made from 7-day-old larvae: 50 larval guts were excised and disrupted by sonication in 1ml Na₂CO₃/HCl, pH9 ·5. Particulate material was removed by centrifugation at 10 000g for 5 min and the supernatant filtered through a 0 ·45 μm Milipore filter.

Mosquito assays were conducted by the method of Tyrell et al. (1979) using 7-day-old A. aegypti larvae hatched from eggs kindly provided by Mr D. Funnell (Shell Research Ltd, Kent).

Bioassay of 3rd instar P. brassicae larvae (obtained from Mr B. Gardiner, Cambridge Biotech Services) was carried out as described by Burges et al. (1975).

Choristoneura fumiferana CF1 cells (Lepidoptera, spruce budworm, trypsinized larval tissue) obtained from Dr S. S. Sohi (Canadian Forest Pest Management Institute, Ontario) were grown in Grace’s Medium.

Spodoptera frugiperda (Lepidoptera, fall army worm, pupal ovary), Heliothis zea (Lepidoptera, cotton bollworm, adult ovary) and Mamestra brassicae cells (Lepidoptera, cabbage moth), ob tained from Mrs T. Lescott (NERC Institute of Virology, Oxford), were grown in TC100 medium. Aedes albopictus (Diptera, trypsinized larval tissue) obtained from Mrs T. Lescott, A. aegypti Aa(s) (Diptera, trypsinized larval tissue) and Culex quinquefasciatus C2 cells (Diptera, larval tissue), obtained from Dr D. W. Roberts (Boyce Thompson Institute, Ithaca, U.S.A.), were grown in Mitsuhashi and Maramorosch medium. Drosophila melanogaster cells (Diptera, ovarian tissue), obtained from Mrs T. Lescott, were grown in Schneider’s Drosophila medium.

All media contained 10% foetal calf serum (Gibco), 50 μgml⁻¹ gentamicin (Nicholas Laboratories) and 50 μgml⁻¹ fungizone (Flow Laboratories). Cells were grown at 28°C.

Activated soluble crystal protein was made 10% in foetal calf serum (FCS, Gibco) prior to use in order to neutralize the effect of insect gut enzymes on cell lines. Cells, at 1—2 (×10⁶) ml⁻¹ in tissue culture medium were incubated with activated toxin at 50 μgml⁻¹. Viability was assessed by ability to exclude Trypan Blue (Thomas & Ellar, 1983a). Controls contained the appropriate volumes of buffer, gut extract and foetal calf serum (FCS).

The polypeptide composition of crystal δ-endotoxins was assessed by SDS-PAGE (Fig. 1). Crystals of var. kurstaki and var. aizawai showed almost identical gel profiles, with a major polypeptide of 130×10³M_r and a minor component of 63 ×10³M_r only faintly visible by Coomassie Blue staining, but clearly visualized by immunoblotting, var. kurstaki Pl antiserum cross-reacted with var. aizawai 130 ×10³M_r protein, and antiserum to var. kurstaki 63 × 10³M_r (P2) protein reacted with var. aizawai 63 ×10³M_r protein, visualized by immunoblotting (Fig. 2). var. thuringiensis had a major component of 138 ×10³M_r, cross-reacting weakly to Pl antiserum, and a minor component of 130 ×10³M_r, cross-reacting strongly; neither component bound P2 antiserum.

The polypeptides of 100— 60 (×10³) M_r seen in the SDS-PAGE profiles are proteolytic products of the 130 × 10³M_r component, all of which bind Pl antiserum. These are generated by the endogenous proteases associated with the crystals (Bulla et al. 1977; Chestukhina et al. 1978).

In order to assay toxicity in vitro, the crystal protein protoxins must be solubilized and activated. This is done by simulating the conditions in the insect gut, the site of crystal activation in vivo. Incubation of crystals in SOmM-Na₂CO₃/HCl, pH9 ·5, and 10 mM-dithiothreitol for 60 min at 37 °C resulted in the complete solubilization of the 130 ×10³M_r proteins of var. kurstaki and var. aizawai, and the 130 and 138 (×10³) components (i.e. the entire crystal) of var. thuringiensis (Fig. 3A-C, track 3), leaving the 63 ×10³M_r protein of var. kurstaki and var. aizawai as an insoluble pellet (Fig. 3A,B, track 1). All subsequent in vitro assays were carried out using the soluble proteins, since the insoluble fractions cannot be assayed in vitro.

Activation of the soluble protoxins with P. brassicae or A. aegypti gut extracts resulted in proteolysis of the high M_r proteins, yielding products of 50—60 (× 10³) M_r (Fig. 3A-C, tracks 4 and 5).

Purification of the crystal (δ-endotoxins by discontinuous sucrose density gradient ultracentrifugation yielded preparations with less than 0 ·2% spore contamination for var. kurstaki and var. thuringiensis, assessed by phase-contrast microscopy, but 10% contamination for var. aizawai. In assays in vitro the spores were removed from the solubilized toxin preparation by centrifugation, but for in vivo assays of native crystal and insoluble crystal fraction spores were present. For this reason bioassays were conducted for only 24 or 48 h, since the reduction of feeding of P. brassicae larvae over short exposure periods is caused only by crystals, and is not influenced by the spores (Burges et al. 1975).

Results of bioassays are shown in Table 1: var. thuringiensis proteins of 130 plus 138 (×10³)M_r, and var. kurstaki 130 ×10³M_r protein killed only P. brassicae larvae, whereas var. aizawai 130×10³M_r protein, and var. aizawai and var. kurstaki 63 × 10³M_r proteins, were toxic to both P. brassicae and A. aegypti larvae.

It should be noted that since A. aegypti larvae are filter feeders, and ingest particulate material more efficiently than soluble material, LC₅Q values for soluble toxin will be artificially high, and cannot be compared directly with values for insoluble material.

The results of in vitro assay of activated soluble toxins on a range of lepidopteran and dipteran cell lines are shown in Table 2. The time taken for 50% cell death (assessed by Trypan Blue staining) on addition of 50 μgml⁻¹ toxin was used for a direct comparison of toxicity towards the different cell lines. The control cells, incubated in an equal volume of buffer, gut enzymes and FCS without toxin, showed less than 5% mortality during the course of the assays. This is important in assays in vitro, since cultured cell lines are sensitive to changes in pH and osmotic concentration, and to the presence of digestive enzymes, and are likely to be more susceptible to toxins if stressed in this way.

var. kurstaki activated Pl protein killed only C. fumiferana cells, var. thuring-iensis activated soluble proteins lysed C. fumiferana and H. zea cells, while var. aizawai activated soluble protein killed all of the lepidopteran and most of the dipteran cell lines tested. Toxins activated with P. brassicae gut extract showed qualitatively the same toxic effects as those activated by A. aegypti gut extract (although A. aegypti gut-extract-activated toxins showed quantitatively lower toxicity), indicating that the source of gut extract is not important in determining the insect specificity of these toxins.

The effect of activated toxins on a range of cell lines is shown in Figs 4—5. The toxins cause, sequentially, blebbing of membrane vesicles from cell processes, rounding up, granulation, swelling and lysis. These effects are very similar to those observed in midgut cells in vivo (Endo & Nishiitsusuji-Uwo, 1980; Percy & Fast, 1983).

Equal volumes of var. kurstaki Pl antiserum and 50 μgml⁻¹ solubilized, activated δ-endotoxins from all three crystal varieties were preincubated together for 30 min at 22°C before addition to insect cells in vitro. In all cases the antiserum completely neutralized toxicity: this can be seen in the case of var. aizawai toxin assayed against H. zea cells in Fig. 6.

When Pl antiserum was added to C.fumiferana cells at varying time intervals after addition of activated var. kurstaki toxin, the antiserum was unable to neutralize toxicity 2 min after toxin application, and 1 min of exposure to toxin was sufficient to cause 50% maximal cell damage. This is in agreement with the findings of Murphy et al. (1976) using a similar system.

The results presented here illustrate the complexity of B. thuringiensis δ-endotoxin classification. Thus the similar polypeptide and antigenicity profiles of var. kurstaki HD-1 and var. aizawai HD-249 crystals are not paralleled by a similar toxicity spectrum. Previous workers have reported that different varieties of B. thuringiensis and different isolates of the same variety can produce δ-endotoxins differing in their host spectrum in vivo (Heimpel & Angus, 1960; Dulmage, 1975). However, these in vivo studies are open to criticism on various counts: for instance, it is now known that many crystals comprise at least two distinct toxic moieties (Yamamoto & lizuka, 1983) with very different toxicity spectra, that may act synergistically in some hosts, thus making it difficult to interpret toxicity data from assays involving whole crystals. Also, the diet of the insect can alter its response to a particular toxin, as seen in the case of Heliothis virescens, which is susceptible to δ-endotoxins when fed soya, tobacco or artificial diets, but is not adequately controlled when feeding on cotton (Luthy et al. 1985). This effect of the diet may be due to tannins that inactivate var. kurstaki HD-1 δ-endotoxin (Luthy et al. 1985), or to the possibility that certain plant lectins may inhibit this toxin in vitro (Knowles et al. 1984). Moreover, if the diet of larvae affects their gut pH and, or, protease production this might alter their ability to activate the toxins.

The work described here is the first to compare the toxicity of different δ-endotoxins to a wide range of insect cell lines in vitro. This approach is of value in studying specificity determinants since it enables the assay of different toxins to be carried out in a controlled manner, unaffected by the variable conditions encountered in assays in vivo. This assay system is highly reproducible, and has already provided information on the mechanism of action of B. thuringiensis var. israelensis δ-endotoxin (Thomas & Ellar, 1983b) and the basis of specificity of B. thuringiensis var. colmeri δ-endotoxin (Haider et al. 1986).

However, in vitro assays do have limitations, principally stemming from the unavailability of midgut cell lines. It has not so far proved possible to grow midgut cells in continuous culture, perhaps because of the difficulties involved in providing a culture system that simulates the asymmetry of growth conditions in vivo. For example, steep K⁺ and H⁺ gradients are maintained across the midgut and there are many other differences in the composition of gut contents and haemolymph (Wood, 1972). Thus in vitro assays are restricted to cell lines derived from tissue of various sources such as those produced from ovarian tissue (H. zea, S. frugiperda, D. melanogaster) or whole trypsinized larvae (C. fumiferana, A. albopictus, An. stephensi, An. gambiae). In the latter case mixed cultures of unknown origin are produced. However, it seems that for at least one of the toxins described here, the soluble protein of var. aizawai, the source of insect tissue does not determine its spectrum of activity in vitro, implying that the var. aizawai receptor is not confined to midgut cells. However, unlike the δ-endotoxin of var. israelensis, which binds phospholipids and thus kills all eukaryotic cells once activated (Thomas & Ellar, 1983b), the activated soluble protein of var. aizawai does not lyse a range of erythrocytes tested (data not shown), suggesting that in this case the range of activity might be confined to insects.

B. thuringiensis var. thuringiensis HD-350 was isolated from an unusual source, a dead grasshopper, and although we have not shown it to kill insects of this order (unpublished observation), the possibility remains that the host range of this organism is unique. The 138 ×10³M_r protein present in the crystal does not have a counterpart in other strains studied in this laboratory, but a recent report by Jarrett (1985) indicates that crystals from var. aizawai isolates HD-112, HD-135, HD-137 and HD-282 contain 138 and 130(×10³)M_r components with distinct toxicity spectra encoded in different plasmids. Since many of the δ-endotoxin proteins are encoded on plasmid-bome genes, crystal components may be conveniently studied separately by generating mutants lacking one of the toxin-encoding plasmids as described by Jarrett (1985). The assays described here do not distinguish between the 130 and 138 (× 10³)M_r proteins since they are equally soluble in the buffers used, thus it remains possible that the toxicity observed may be due to only one of the two proteins, or to synergistic effects.

The fact that many B. thuringiensis δ-endotoxin genes are borne on plasmids, which might be transferred from one variety to another (Gonzalez et al. 1982), implies that there is no a priori reason to suppose that toxins produced by different isolates of the same serotype bear any structural or functional relationship to each other. Thus, for example, a recently isolated strain of var. morrisoni, designated PG-14, has a toxicity spectrum totally unrelated to that of the reference strain of var. morrisoni (Padua et al. 1984) and var. darmstadtiensis isolate 73-E-10·2 has a different host range to the reference strain (Padua et al. 1980). It has also become clear from recent toxin sequence data that small changes in amino acid sequence can lead to significant differences in toxicity spectra (Schnepf et al. 1985; Shibano et al. 1985; Adang et al. 1985). This underlines the importance of specifying the isolate number in studies of different δ-endotoxins since the variety name is not sufficient to define the toxin type.

We thank Mrs Jillian Clements and Mrs Margaret Bradley for secretarial assistance and Mrs Audrey Symonds for her meticulous washing up. This work was supported by grants from SERC and AFRC.