Bt Toxin Resistance from Loss of a Putative Carbohydrate-Modifying Enzyme

See allHide authors and affiliations

Science  03 Aug 2001:
Vol. 293, Issue 5531, pp. 860-864
DOI: 10.1126/science.1062441


The development of resistance is the main threat to the long-term use of toxins from Bacillus thuringiensis (Bt) in transgenic plants. Here we report the cloning of a Bt toxin resistance gene,Caenorhabditis elegans bre-5, which encodes a putative β-1,3-galactosyltransferase. Lack of bre-5 in the intestine led to resistance to the Bt toxin Cry5B. Wild-type but notbre-5 mutant animals were found to uptake toxin into their gut cells, consistent with bre-5 mutants lacking toxin-binding sites on their apical gut. bre-5 mutants displayed resistance to Cry14A, a Bt toxin lethal to both nematodes and insects; this indicates that resistance by loss of carbohydrate modification is relevant to multiple Bt toxins.

Crystal toxins produced by B. thuringiensis are used worldwide in transgenic crops to control caterpillars and beetles, are an important tool of organic farming, and have made important contributions to the control of insect-borne diseases such as African river blindness. Once ingested by an insect, Bt toxins are proteolytically activated in the midgut and bind to membrane gut receptors, leading to pore formation and death (1,2). Although Bt toxins are safe to vertebrates and are considered beneficial to the environment relative to chemical pesticides, Bt toxin effectiveness is threatened in the long term by the development of insect resistance (3). Bt-resistant variants of the diamondback moth have been identified in the field, and resistant strains of at least 11 insect species have been documented in the laboratory (4, 5). Understanding the molecular mechanism of toxin action and identifying the genes that can mutate to yield resistance are important steps in developing strategies to help delay or circumvent this problem.

Some Bt toxins are toxic to the nematode C. elegans(6). The best characterized of these nematicidal toxins, Cry5B, falls into a phylogenetic group of eight Bt toxins, at least two of which, Cry5A and Cry14A, are toxic to insects (2). Cry5B has ∼24% sequence identity to commercially important insecticidal toxins, such as Cry1Ac, in the toxin domain. Cry5B contains four of the five protein sequence blocks conserved among most Bt toxins and may fold into a three-domain structure related to those of insecticidal toxins (1, 7). Like insects, nematodes fed Bt toxin rapidly cease feeding and incur intestinal damage (6, 8). These similarities in toxin sequences and response suggest that the general mechanism of Bt toxicity in nematodes and insects is conserved. By studying the process in C. elegans, we can take advantage of the molecular, genetic, and cell biological tools available in this model organism. In addition, nematicidal Bt toxins are important to study because of their potential to control plant-parasitic nematodes. These widespread agricultural pests cause ∼$80 billion per year in crop damage (9) that is likely to be exacerbated by an upcoming worldwide ban of methyl bromide, the main chemical currently used to control them.

We previously reported the identification of five C. elegans genes, called bre (for Bt resistance), that mutate to Cry5B resistance (6). One of these genes, bre-5, was mapped to the right end of the chromosome IV cluster. We transformedbre-5 mutant animals with cosmids in this area and rescued toxin resistance to toxin susceptibility in these animals with the cosmid T12G3 (10). We then narrowed down bre-5rescue to a 4.3-kb fragment within T12G3 (Fig. 1, A and B). This fragment did not contain any genes predicted in the genome database, but it did contain a single potential gene with extensive sequence similarity to mammalian glycosyltransferases. The cDNAs corresponding to this gene were isolated and a complete sequence assembled (11); it encodes a 322–amino acid protein (Fig. 1C). To confirm that this gene is bre-5, we sequenced the complete bre-5coding region from each of our two bre-5 mutant alleles. Both alleles show alterations in this gene consistent with a loss or reduction of function (Fig. 1C).

Figure 1

BRE-5 encodes a putative galactosyltransferase that is required in the C. elegans gut for Bt toxin action. (A and B) Rescue experiments. In (A), a bre-5(ye17) animal fed Bt toxin for 24 hours shows a healthy, resistant gut. In (B), abre-5(ye17) animal transformed with the 4.3-kb rescuing fragment fed Bt toxin for 24 hours shows a damaged, susceptible gut. Anterior is to the left. The posterior pharynx and anterior intestine are shown for each animal. Scale bar, 50 μm. (C) CLUSTALW (version 1.81) alignment of BRE-5 protein with human β-1,3-galactosyltransferase polypeptide 5 (hB3T5), mouse β-1,3-galactosyltransferase polypeptide 3 (mB3T3), andDrosophila BRAINIAC (Brn). Blue, absolutely conserved residues; green, conserved amino acid groups. The putative transmembrane domain is underlined; the DXD and DDVFTG motifs are double-underlined (single-letter abbreviations for amino acid residues: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr). The locations of the two arginines mutated in thebre-5 alleles are indicated: ye107 alters an arginine conserved in all β-1,3-galatcosyltransferases; ye17 introduces a premature stop codon. (D and E) Mosaic analysis showing that lack of BRE-5 in the gut leads to resistance to Bt toxin. Orientations are as in (A) and (B); scale bar, 50 μm. Arrows point to the anterior end of the gut. Exposures were identical for each pair of wild-type andbre-5(ye17) images. Shown in (D) are differential interference contrast (DIC) and deconvolved fluorescein isothiocyanate (FITC)–channel images of a bre-5(ye17) animal transformed with the extrachromosomal array that contains the 4.3-kb rescuing fragment (the animal is therefore sensitive to toxin) and SUR-5GFP(NLS), which results in GFP expression in the nuclei. Five large gut nuclei are in focus and brightly express GFP (arrowheads). In (E), DIC and deconvolved FITC-channel images of a mosaic animal that expressed GFP in some cells (and therefore contained the array) but was resistant to toxin (and therefore had lost the array in cells required for toxicity) are shown. In all 67 resistant animals, the gut nuclei never express GFP, indicating that the array was lost in the gut lineage. In the animal shown, the array is present in pharyngeal terminal bulb cells, descendants of the MS lineage.

BLAST and protein domain searches indicated that BRE-5 is a member of the β-1,3-galactosyltransferase family that transfers galactose onto proteins and lipids (12). BRE-5 is most similar in sequence to the Drosophila protein BRAINIAC. Over a 200–amino acid stretch that includes the catalytic domain, BRE-5 shows 37%, 27%, and 25% sequence identity to BRAINIAC, mouse β-1,3-galactosyltransferase 3, and human β-1,3-galactosyltransferase 5, respectively (Fig. 1C). BRAINIAC has been implicated in Notch signaling, perhaps by influencing ligand-receptor interactions, as has been speculated (13,14). BRE-5 contains all the hallmarks of β-1,3-galactosyltransferases, including a putative NH2-terminal transmembrane domain, an Asp-X-Asp motif, and a conserved variant of the Glu-Asp-Val-Tyr-Val-Gly motif. To confirm that loss of this galactosyltransferase gene leads to Bt toxin resistance, we injected hermaphrodite gonads with double-stranded (ds) RNA derived from the bre-5 cDNA (15). Injection of dsRNA is known to deplete gene function in the progeny of injected hermaphrodites via RNA interference (RNAi) (16). After injections of dsRNA at 1.5 and 3.0 mg/ml, we found that, respectively, 45% (n = 60) and 73% (n= 60) of the progeny were resistant to Cry5B. These results confirm that Bt toxin resistance is the bre-5 loss-of-function phenotype. As previously reported for bre-5(ye17)(6), we did not detect lethality or other obvious phenotypes in bre-5 RNAi animals.

On the basis of the identification of BRE-5 as a putative galactosyltransferase and numerous in vitro studies that pointed to the importance of carbohydrates in the binding of insecticidal Cry1Ac to receptor and membrane (17, 18), we hypothesized that BRE-5 functions in forming a carbohydrate structure, present on proteins or lipids exposed at the gut surface, that is necessary for toxin binding. In the absence of bre-5–dependent carbohydrates, Bt toxin cannot bind, resulting in resistance. Such a requirement for carbohydrates in microbial toxin recognition would not be without precedent. For example, cholera toxin binds to host cells via carbohydrates (19).

To understand BRE-5 function better, we performed experiments involving mosaic animals to determine the anatomical focus of the gene's function with respect to Bt toxin susceptibility. Homozygousbre-5(ye17) hermaphrodites were injected with a cocktail of plasmids that included the dominant rol-6 marker (which causes animals to roll), SUR-5GFP(NLS) [which is expressed in the nuclei of many somatic cells, including the intestine (20)], and the 4.3-kb rescuing fragment ofbre-5. A stable line was established that transmitted all three transgenes as an extrachromosomal array in 50% of the progeny. As a result of bre-5 rescue and expression of SUR-5GFP, rolling worms were sensitive to Cry5B, as expected, and displayed nuclear green fluorescent protein (GFP) (Fig. 1D). Of 2060 worms that were rolling (and therefore carrying the array) and were transferred to toxin plates, 67 rare, toxin-resistant animals were identified. These animals were resistant presumably because the extrachromosomal array had been lost during somatic divisions (20) in the tissue(s) where BRE-5 expression is needed for toxin to be effective. When these resistant, mosaic animals were examined for fluorescence, all 67 had lost GFP signal in the gut (Fig. 1E), indicating that the array and BRE-5 function were missing in the gut lineage, derived exclusively from the E blastomere. Furthermore, in 19 of these 67 animals, GFP staining still was present in posterior pharyngeal cells derived from MS, the sister of E (Fig. 1E), excluding the possibility that the array also had to be lost in other cells leading up to the birth of the E cell. We have verified that nonrolling mosaic animals also lack GFP fluorescence in the gut (20 of 20 animals). Thus, loss of the wild-typebre-5 gene in the gut causes animals to be resistant to toxin. Consistent with this finding, we have performed immunofluorescence with a recently purified BRE-5 antibody that indicates expression in the gut (21).

To test directly whether BRE-5 is required for toxin to interact with the nematode gut in vivo, we fed fluorescently labeled Bt toxin to L4-staged hermaphrodites and followed its fate in wild-type andbre-5 mutant animals (22). In wild-type animals, labeled toxin was internalized by gut cells, where it colocalized with autofluorescent gut granules, probably the site of the intestinal lysosome (23) (Fig. 2, upper panels). This uptake into wild-type gut cells was seen as rapidly as 20 min after feeding was initiated. These data suggest that toxin binds to the nematode gut via receptors and is then endocytosed. In contrast, when bre-5(ye17) animals were fed labeled toxin, toxin remained in the intestinal lumen and was not internalized by gut cells (Fig. 2, lower panels).

Figure 2

bre-5 mutant animals are defective in internalizing Bt toxin into gut cells. Wild-type andbre-5(ye17) animals were fed rhodamine-labeled Cry5B toxin for 1.5 hours and then imaged with DIC, in the rhodamine channel to visualize toxin, and in the FITC channel to visualize autofluorescent gut granules. Toxin was detected inside the wild-type gut cells and often colocalized with lysosomal gut granules. Toxin was not detected inside the bre-5 mutant gut cells but was confined to the lumen. Anterior is to the left in each panel; arrows point to the anterior end of the gut. Exposures are identical for each pair of wild-type and bre-5(ye17)images.

To rule out the possibility that the bre-5mutant gut was generally defective in endocytosis, we fed wild-type andbre-5(ye17) animals the lipophilic dye FM4-64, a marker for membrane-mediated endocytosis, and rhodamine-labeled bovine serum albumin (BSA), a marker for fluid-phase endocytosis. Within 20 min, FM4-64 dye entered gut cells and colocalized with lysosomal gut granules in both wild-type and mutant animals, with indistinguishable kinetics. Rhodamine-BSA also entered wild-type and mutant gut cells with similar kinetics, but took much longer to detect than did FM4-64 or toxin. These data indicate that membrane-mediated and fluid-phase endocytosis occur relatively normally in thebre-5 mutant. Moreover, the finding that the rapid uptake of toxin into gut cells more closely resembles uptake of FM4-64 than that of rhodamine-BSA is consistent with toxin entering gut cells by membrane association rather than by fluid-phase endocytosis.

We have also ruled out the idea that bre-5 mutant animals have altered feeding behaviors that might affect the ability of toxin to interact with the gut membrane. We found that pharyngeal pumping rate (243 ± 9 pumps per min in wild type, 242 ± 11 inbre-5(ye17); n = 10 for both) and defecation rate (48 ± 5 s per cycle in wild type, 49 ± 7 s per cycle in bre-5(ye17); n = 10 for both) are not affected in the bre-5 mutant.

To address how widespread the bre-5resistance mechanism might be, we tested whether bre-5mutants were resistant to other Bt toxins. We took advantage of the fact that there is a known Bt toxin, Cry14A, that is part of the same phylogenetic subgroup of Bt toxins as Cry5B and is toxic to both nematodes and insects (2, 24, 25). Cry14A is 23% identical to Cry1Ac and 34% identical to Cry5B in the toxin domain. As with Cry5B, wild-type C. elegans fed Cry14A rapidly show gut damage. We found that bre-5(ye17) animals were sick on plates expressing high levels of Cry14A but were healthy on plates expressing lower levels of Cry14A that were still toxic to wild-type animals. To quantitate this dose-dependent resistance, we performed brood size assays for wild-type and bre-5(ye17) animals in the presence of variable amounts of Cry14A toxin (Fig. 3) (26). These data indicate that, relative to the wild type, bre-5(ye17) is 19 times as resistant to Cry14A.

Figure 3

Relative to the wild type,bre-5(ye17) is 19 times as resistant to Cry14A. The 3-day brood sizes of wild-type and bre-5 animals were counted in varying doses of Cry14A toxin. Percent of control indicates the percent progeny relative to the 3-day brood sizes of wild-type andbre-5(ye17) in the absence of toxin, which are 132 ± 38 and 99.2 ± 28, respectively. The dose at which their brood sizes are reduced to 50% is 11.2 ng/ml in wild-type and 210 ng/ml inbre-5(ye17).

Because bre-5 mutants show resistance to two divergent Bt toxins that share only 34% identity, this mechanism of resistance is likely to be applicable to other Bt toxins as well. Our Cry14A data also suggest that this mechanism is relevant for insects, because Cry14A is toxic to the beetle Diabrotica spp. It is likely that Cry14A recognizes the same carbohydrate structure in the beetle as in the nematode, and that bre-5–mediated resistance could develop in this insect with this toxin.

Our identification and characterization ofbre-5 as a Bt resistance gene provides evidence in vivo for the importance of carbohydrates in Bt toxicity and the development of resistance. It is noteworthy that in the commercially important insecticidal toxins Cry1Aa and Cry3A, subdomain II folds into a β prism structure similar to that found in the plant lectins jacalin andMaclura pomifera agglutinin (27, 28). Both of these lectins are highly specific for binding the carbohydrate galactose β-1,3-N-acetylgalactosamine, precisely the type of structure made by the BRE-5 family of enzymes. Thus, it is conceivable that the binding of these insecticidal toxins requires a carbohydrate structure similar to that putatively formed by BRE-5.

Our results potentially explain a dilemma in the Bt field, namely that a single toxin can bind to at least two receptors that are completely unrelated in sequence (2). We hypothesize that these disparate receptors are able to bind the same toxin through a common carbohydrate motif. Resistance by loss of a carbohydrate-modifying enzyme is thus particularly dangerous and more threatening than mutation of a single receptor. Loss of a single general modifier such as BRE-5 could affect the binding of multiple Bt toxins to multiple receptors, leading to a high level of resistance to a single toxin and cross-resistance to other toxins as well. Determining how widespread this type of resistance is among different invertebrates—and how to deal with it—is vital for the long-term effectiveness of this important technology.

  • * To whom correspondence should be addressed. E-mail: raroian{at}


View Abstract

Stay Connected to Science

Navigate This Article