Unusual Amino Acid Determinants of Host Range in the Mtx2 Family of Mosquitocidal Toxins*

Five different mosquitocidal toxin ( mtx2 ) gene ho- mologs have been cloned from eight Bacillus sphaericus strains. Pairwise comparisons of the predicted amino acid sequences show between four and eight substitutions compared with the prototype Mtx2 from B. spha- ericus strain SSII-1. Mtx2 from strain SSII-1 was (cid:59) 7-fold more toxic to Culex mosquito larvae than the Mtx2 hom- olog from B. sphaericus strain 31-2. Conversely, Mtx2 from strain 31-2 was (cid:59) 100-fold more toxic to Aedes mosquito larvae than Mtx2 from strain SSII-1. Lys 224 in Mtx2 was found to be the most important amino acid for tox- icity to Culex larvae, and substitution of Lys 224 with threonine abolished the toxicity of Mtx2 from strain SSII-1 to these larvae. In complete contrast, Thr 224 was found to be crucial for the toxicity of Mtx2 from strain 31-2 to Aedes larvae, and substitution of Thr 224 with lysine caused a (cid:59) 100-fold drop in toxicity to these larvae. Thus, amino acid 224 in the Mtx2 family of mosqui- tocidal toxins is an unusual and important determinant of mosquito larvicidal activity and host range. Gif sur Yvette, and all other B. sphaericus strains were obtained from H. de Barjac, Pasteur Institute, Paris. All B. sphaericus strains were grown either in L-broth or NYSM medium (5). Cloning, Mutagenesis, and Sequence Analysis— The protein coding regions of mtx2 from eight different strains of B. sphaericus (31-2, IAB59, Kellen Q, 2297, 2362, 1691, 1593M, and 2317.3) was amplified by PCR from genomic DNA (9) using synthetic oligonucleotides based on the sequence of B. sphaericus SSII-1 mtx2 (4). The sequences of these oligonucleotides were: 5 (cid:57) -CCCCCCATGGATCCAATGAAAAGGAC- CAAATTACTTTTTTATATT-3 (cid:57) (TT119) for the 5 (cid:57) (upstream) primer, and 5 (cid:57) -CCCGACGTCATCGATGCATGCCTTAAGTTATTTAAAA- GAAATTTCTTTAACATCTATTA-3 (cid:57) (TT120) for the 3 (cid:57) (downstream) primer. Artificial Bam HI, Nco I, Cla I, Sph I, and Afl II restriction en- zyme sites were incorporated into these sequences for cloning purposes. 2297, 31-2, 2362,

Five different mosquitocidal toxin (mtx2) gene homologs have been cloned from eight Bacillus sphaericus strains. Pairwise comparisons of the predicted amino acid sequences show between four and eight substitutions compared with the prototype Mtx2 from B. sphaericus strain SSII-1. Mtx2 from strain SSII-1 was ϳ7-fold more toxic to Culex mosquito larvae than the Mtx2 homolog from B. sphaericus strain 31-2. Conversely, Mtx2 from strain 31-2 was ϳ100-fold more toxic to Aedes mosquito larvae than Mtx2 from strain SSII-1. Lys 224 in Mtx2 was found to be the most important amino acid for toxicity to Culex larvae, and substitution of Lys 224 with threonine abolished the toxicity of Mtx2 from strain SSII-1 to these larvae. In complete contrast, Thr 224 was found to be crucial for the toxicity of Mtx2 from strain 31-2 to Aedes larvae, and substitution of Thr 224 with lysine caused a ϳ100-fold drop in toxicity to these larvae. Thus, amino acid 224 in the Mtx2 family of mosquitocidal toxins is an unusual and important determinant of mosquito larvicidal activity and host range.
Bacillus sphaericus is an aerobic, Gram-positive, spore-forming bacterium which is widespread in soil and aquatic environments. Some strains of B. sphaericus produce protein toxins which are lethal to mosquito larvae (1,2). The best studied mosquitocidal strains of B. sphaericus are divided into a high toxicity group (e.g.. 2362, 2297, and IAB59) and a low toxicity group (e.g.. SSII-1, 31-2 and Kellen Q) (1). The high toxicity but not the low toxicity strains encode 51.4-and 41.9-kDa proteins, which together form a binary toxin expressed at high levels during sporulation. Most mosquito pathogenic B. sphaericus tested also harbor a 100-kDa toxin gene (mtx) and a 31.8-kDa toxin gene (mtx2) 1 (3,4).
Toxin production in the low toxicity strain B. sphaericus SSII-1 begins in the vegetative phase of growth before the onset of sporulation (5,6). The mtx2 gene of this strain encodes a polypeptide of 292 aa (Mtx2) with a molecular mass of 31.8 kDa, which is detected in the vegetative phase of growth (4). Mtx2 is unrelated to the binary and 100-kDa toxins but has regions of significant homology with the 33-kDa ⑀ toxin of Clostridium perfringens and the 31.68-kDa cytotoxin of Pseudomonas aeruginosa (4). In this study, we demonstrate that the Mtx2 toxins from six strains of B. sphaericus have few sequence differences, but some of these toxins exhibit major differences in larvicidal activities against two species of mosquitoes. The results permitted the design and assay of hybrid toxins and the identification of aa residues in Mtx2 that are unusual determinants of larvicidal activity and mosquito host range.
Bacterial Strains and Media-B. sphaericus SSII-1 was a gift from E. W. Davidson, Arizona State University, Tempe. Strains Kellen Q and 31 were obtained from A. A. Yousten, Virginia Polytechnic Institute and State University, Blacksburg. Strain 1593M was a gift from J. Szulmajster, C.N.R.S., Gif sur Yvette, and all other B. sphaericus strains were obtained from H. de Barjac, Pasteur Institute, Paris. All B. sphaericus strains were grown either in L-broth or NYSM medium (5).
Cloning, Mutagenesis, and Sequence Analysis-The protein coding regions of mtx2 from eight different strains of B. sphaericus (31-2, IAB59, Kellen Q, 2297, 2362, 1691, 1593M, and 2317.3) was amplified by PCR from genomic DNA (9) using synthetic oligonucleotides based on the sequence of B. sphaericus SSII-1 mtx2 (4). The sequences of these oligonucleotides were: 5Ј-CCCCCCATGGATCCAATGAAAAGGAC-CAAATTACTTTTTTATATT-3Ј (TT119) for the 5Ј (upstream) primer, and 5Ј-CCCGACGTCATCGATGCATGCCTTAAGTTATTTAAAA-GAAATTTCTTTAACATCTATTA-3Ј (TT120) for the 3Ј (downstream) primer. Artificial BamHI, NcoI, ClaI, SphI, and AflII restriction enzyme sites were incorporated into these sequences for cloning purposes. * This work was funded by the Institute of Molecular and Cell Biology, Singapore. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM 1 The abbreviations used are: mtx2, mosquitocidal toxin type 2 gene; Mtx2, mosquitocidal toxin type 2 protein; aa, amino acid(s); GST, glutathione S-transferase; LC 50 , lethal concentration of protein or cells theoretically required to kill 50% of mosquito larvae; PCR, polymerase chain reaction; bp, base pair(s); kb, kilobase pair(s). 2 Protein mutations are abbreviated using the following convention; the residue number is preceded by the symbol (in the one-letter code) for the wild-type amino acid and followed by the symbol for the mutant amino acid. Thus D31H denotes a mutation from Asp to His at position 31.
The 0.8-kb PCR products were digested with NcoI and ClaI, and cloned into the NcoI and ClaI sites of the large (vector) fragment of plasmid pTH26 (7), which was propagated in Escherichia coli DH5␣. Another set of recombinants were generated in the same way, and both sets were completely sequenced to ensure mutations were not artefactually introduced into the mtx2 genes during the PCR. Cloning and sequencing were carried out in the standard manner (10). Protein sequence analysis and alignments were performed using DNASTAR software.
Expression and Purification of Mtx2 Proteins-Synthesis and purification of Mtx2␦ fusion proteins was carried out essentially by a published procedure (8), except that E. coli DH5␣ were grown at 30°C for 2 h after induction of Mtx2␦ synthesis.
Antibodies and Western Blotting-Polyclonal antibodies against recombinant Mtx2 were raised in a previous study (4) and used to perform Western blotting as described (4) to detect recombinant Mtx2 homologs expressed in recombinant E. coli.
Mosquito Larvicidal (Toxicity) Assays-All plasmids expressing Mtx2 fusion proteins were transformed into E. coli DH5␣ for larvicidal assay. The toxic activities of different E. coli clones synthesizing the Mtx2 fusion proteins (or purified Mtx2␦ fusion proteins) (4) were measured on laboratory-reared 1st instar larvae of Culex quinquefasciatus and Aedes aegypti mosquitoes. Recombinant E. coli cultures were harvested, washed once with water, resuspended in 0.1 the original volume with water, and the A 600 was measured. The cultures were adjusted to the desired range of concentrations with water and added to 10 1st instar larvae in 1 ml of water with 50 l of a 100 mg/ml suspension of yeast. Assays using the desired range of concentrations of Mtx2␦ fusion proteins were performed in a similar manner, except that 50% PBS was used throughout. All assays were performed at least twice, and each concentration of bacteria (or purified Mtx2␦ fusion protein) was assayed in triplicate. Surviving larvae were counted at 24 h. The LC 50 values were calculated from the average mortality observed at 24 h by using the CA Cricket Graph III program. Mortality of control larvae fed with yeast suspension alone, with 50% PBS, or with E. coli transformed with plasmid vector alone was less than 10% in all assays.

RESULTS
Cloning and Comparison of mtx2 Genes-Using PCR primers based on the sequences at the extremities of the region coding for the mtx2 gene from B. sphaericus SSII-1 (4), 0.8-kb PCR products were generated from the genomic DNA of eight strains of B. sphaericus, namely 31-2, IAB59, Kellen Q, 2297, 2362, 1691, 1593M, and 2317.3. The amplified DNAs were cloned in a plasmid vector, pTH26 (7), and two independent clones of each mtx2 gene were completely sequenced to exclude the possibility of PCR-derived mutations.
The predicted aa sequences of the eight Mtx2 homologs were compared with each other and with the known sequence of B. sphaericus SSII-1 Mtx2, and aa variations were found at 10 positions between aa 31 and aa 279 (Fig. 1). The Mtx2 homologs from B. sphaericus 2362, 1691, 1593M, and 2317.3 were identical to each other (Table I). Overall, there were between four and eight aa substitutions when the Mtx2 sequences from strains 31-2, IAB59, Kellen Q, 2297, and 2362 were compared in a pairwise fashion with the prototype Mtx2 sequence from SSII-1 (Table I). Mtx2 from SSII-1 was unique in having Ser 37 , Ser 67 , Lys 224 , and Thr 279 , while Phe 239 was only found in Mtx2 from Kellen Q, and Thr 171 was unique to Mtx2 from strain 2297 (Table I).
Synthesis and Mosquito Larvicidal Activity of Mtx2 Homologs in E. coli-In order to compare the mosquitocidal activities of the six Mtx2 homologs, the mtx2 coding regions were fused in frame to GST (see "Experimental Procedures"). Recombinant E. coli cells were incubated with isopropyl-␤-D-thiogalactopyranoside to induce the synthesis of GST-Mtx2, and cell extracts were examined by Western blotting using a polyclonal antiserum to Mtx2 from B. sphaericus SSII-1. All recombinant cells produced a protein of ϳ58 kDa, the expected size of GST-Mtx2 (Fig. 2). The synthesis in E. coli of Mtx2 from B. sphaericus strains 2362 and Kellen Q was lower than that of Mtx2 from SSII-1, 31-2, IAB59, and 2297, but expression of all fusion proteins did not vary by more than 2.7-fold (Fig. 2).
Intact E. coli cells were fed to larvae of the mosquito C. quinquefasciatus, and larvicidal activities of the various recombinant E. coli clones harboring Mtx2 fusion proteins were quantitated (Table II), taking into account the different levels of Mtx2 fusion proteins (Fig. 2 legend). All clones exhibited significant toxicity to C. quinquefasciatus, and E. coli synthesizing Mtx2 from strains SSII-1 and 2362 were the most toxic. E. coli synthesizing Mtx2 from SSII-1 were ϳ7 to ϳ14 times more toxic than E. coli synthesizing Mtx2 from strains 31-2, IAB59, 2297, and Kellen Q (Table II). E. coli synthesizing Mtx2 from SSII-1 were ϳ7 times more toxic to C. quinquefasciatus larvae than clones synthesizing Mtx2 from strain 31-2 (Table  II). Surprisingly, the opposite result was obtained when E. coli synthesizing Mtx2 from SSII-1 and 31-2 were assayed against larvae of the mosquito, A. aegypti. Mtx2 from strain 31-2 was about 100-fold more toxic to A. aegypti than Mtx2 from SSII-1 (Table III). SSII-1 and 31-2 Mtx2 differ in four aa positions, namely 37, 67, 224, and 279 (Table I), and any or all of these aa could contribute to the significant differences in the toxicity and mosquito host range of these toxins.
Single or double amino acid substitutions were introduced by  site-directed mutagenesis into selected Mtx2 homologs to test the contribution of individual aa to toxicity and host range. Changes were made to render the Mtx2 proteins more or less like Mtx2 from strain SSII-1 ( Table I). None of the single substitutions in Mtx2, including IAB59 (D31H or R50K), 31-2 (P67S), 2297 (T171K), 2362 (K40Q), SSII-1 (S37I), and SSII-1 (S67P) ( Table I) had any detectable effect on toxicity toward C. quinquefasciatus larvae (data not shown).
Completely opposite results were obtained when the aa 224 and 279 mutants were assayed against A. aegypti larvae. SSII-1 (T279A) was about as toxic to these larvae as the weakly toxic SSII-1 Mtx2 protein, but SSII-1 (K224T) and the double mutant SSII-1 (K224T,T279A) were ϳ50 -150-fold more toxic than the parental SSII-1 Mtx2. The excellent A. aegypti toxicity of 31-2 Mtx2 was not significantly affected by the A279T mutation, but the double mutant 31-2 (T224K,A279T) was only weakly toxic to A. aegypti due to a drop in larvicidal activity of ϳ100-fold compared with 31-2 (A279T) ( Table III).
Assays against C. quinquefasciatus and A. aegypti larvae were performed with purified Mtx2␦ fusion proteins in which the N-terminal 15 amino acids containing the putative signal sequence of Mtx2 were deleted from GST-Mtx2 (Fig. 3) to exclude the possibility that the different larvicidal activities of E. coli expressing SSII-1 and 31-2 Mtx2 were due to differences in toxin solubility or stability in the bacteria. Table IV shows that SSII-1 Mtx2␦ was over 4-fold more toxic than 31-2 Mtx2␦, and that the 31-2 (T224K,A279T) mutant was about 17-fold more toxic than parental 31-2 Mtx2␦ to C. quinquefasciatus larvae. Both SSII-1 mutants with the K224T substitution were nontoxic to C. quinquefasciatus (Table IV). Conversely, purified toxins with T224 were the most toxic to A. aegypti, and toxins with K224 (SSII-1 and 31-2 (T224K,A279T)) were much less toxic to these larvae (Table IV). These results mirror the larvicidal activities of these toxins in live E. coli (Table III), suggesting that the different toxicities of the recombinant bacteria are due to variations in the aa sequences of the Mtx2 homologs, particularly in residue 224.

DISCUSSION
The Mtx2 homologs from six strains of B. sphaericus differ in only a few aa positions, and some of the homologs were found to be more toxic to mosquito larvae than others. These observations allowed us to design mutagenesis experiments which pinpointed Lys 224 as an important aa in the toxicity of Mtx2 to C. quinquefasciatus larvae. The SSII-1 (T279A) mutant was 5-6-fold more toxic to C. quinquefasciatus larvae than SSII-1 Mtx2, but the double mutant SSII-1 (K224T,T279A) was inactive, showing that, although Ala 279 is more favorable than Thr 279 in the context of SSII-1 Mtx2, Lys 224 is an overriding determinant of larvicidal activity. Why did 31-2 Mtx2 exhibit significant toxicity to C. quinquefasciatus despite the absence of Lys 224 ? The highly active 31-2 Mtx2 toxin differs from the inactive SSII-1 (K224T,T279A) mutant in having Ile 37 and Pro 67 , suggesting that one or both of these aa contributes to the toxicity of 31-2 Mtx2 and compensates for the absence of Lys 224 . This assumption is strengthened by the observation that the 31-2 (T224K,A279T) mutant, which is identical to SSII-1 Mtx2 except for the presence of Ile 37 and Pro 67 , was over 7-fold more toxic to C. quinquefasciatus larvae than SSII-1 Mtx2. Therefore, it was surprising that the single substitutions 31-2 (P67S), SSII-1 (S37I), and SSII-1 (S67P) had no effect on toxicity. Further reciprocal constructs (e.g. 31-2 (I37S), SSII-1 (S37I,K224T), and SSII-1 (S67P,K224T)) are needed to resolve the issue of the relative importance of Ile 37 and Pro 67 in the context of 31-2 Mtx2.
When E. coli synthesizing Mtx2 from SSII-1, 31-2, and their respective mutants were assayed against A. aegypti larvae, even greater differences in toxicity were observed. However, in complete contrast to the results of larvicidal assays against C. quinquefasciatus, SSII-1 Mtx2 was found to be ϳ100-fold less toxic to A. aegypti than was 31-2 Mtx2; and surprisingly, the difference was also due to aa position 224. Parental or mutant Mtx2 proteins with Thr 224 were always ϳ50 -150-fold more toxic to A. aegypti than their counterparts with Lys 224 (Table III).
How can a single aa substitution in a toxin substantially increase toxicity to one species of mosquito and virtually abolish toxicity to another species of mosquito? Although this is a new phenomenon, there are earlier studies on lepidopteran and dipteran toxins which are instructive and allow us to speculate on the possible mechanism of action of aa 224 (11)(12)(13)(14)(15)(16). Important genetic determinants of mosquito host range in the binary toxin from B. sphaericus have been localized to aa positions 99 and 104 in the 41.9-kDa subunit, but other aa in the 51.4-kDa subunit also contribute to toxicity (16). This is analogous to aa 224 in Mtx2 playing a major role in mosquito host range and aa 37 and/or 67 playing an accessory role. Several other studies have also concluded that one or a very few aa are major determinants of larvicidal activity and host range in a variety of toxins active against Lepidoptera and/or Diptera (11)(12)(13)(14)(15). However, in many cases insect specificity could not easily be attributed to particular aa, as different toxin segments were found to determine specificity by interacting in an undefined manner (11)(12)(13). Nevertheless, it is evident that different and sometimes overlapping segments of many toxins carry determinants of specificity for different insects (11)(12)(13).
In one interesting study, Ile 545 of a dual specificity larvicidal protein from Bacillus thuringiensis aizawai ICI was found to be essential for toxicity to A. aegypti larvae, but not to larvae of the caterpillar Pieris brassicae; conversely, the single substitution I568T abolished toxicity to P. brassicae larvae but not to A. aegypti larvae (12). Together, the results suggested that the I568T mutation destroyed proteolytic cleavage activation of the 130 ϫ 10 3 M r protoxin to a known ϳ55 ϫ 10 3 M r Lepidopteraactive toxin, while the I545P mutation inhibited proteolytic activation at a different site of the protoxin and prevented the formation of a known ϳ53 ϫ 10 3 M r Diptera-active toxin (12).
Although proteolytic cleavage by larval gut proteases activates many insecticidal toxins (1,2,12), it is unclear whether this also occurs in Mtx2 as the mechanism of action of the recently discovered Mtx2 family of mosquitocidal toxins is unknown (4). However, it is worth speculating that K224 in SSII-1 Mtx2 may be a major site of proteolytic cleavage activation by C. quinquefasciatus gut trypsin-like proteases (1,2), and that other as yet unknown neighboring site(s) in 31-2 Mtx2 (which has Thr 224 ) are exposed for cleavage activation by the compensating aa, Ile 37 and Pro 67 . By analogy with the dual specificity larvicidal toxin from B. thuringiensis aizawai ICI, we would have to postulate that in the A. aegypti larval gut a different protease is responsible for cleavage activation of Mtx2 at a distinct site from the one cleaved in the C. quinquefasciatus gut, and that the T224K mutation in 31-2 Mtx2 adversely affects substrate specificity or denies access to the putative A. aegypti protease.
An alternative and perhaps more plausible hypothesis is simply that there are subtle (but vital) aa differences in the C. quinquefasciatus and A. aegypti gut receptors in the domains which interact with position 224 on the surface of Mtx2. For example, a crucial electrostatic interaction might form only with Thr 224 in A. aegypti, while in C. quinquefasciatus a similar crucial interaction might form only with Lys 224 . Position 224 is predicted to lie at the peak of 1 of 11 predicted hydrophilic regions in Mtx2, which is consistent with a surface location. Regardless of the mechanism, it is clear that aa position 224 in the Mtx2 family of mosquitocidal toxins is an important and unusual determinant of toxicity and mosquito host range.
Finally, it is worth noting that some of the mutant Mtx2 toxins were significantly more toxic to C. quinquefasciatus  a Variation between replicates within one assay Ϯ 10%. Values are means of two or more assays using equivalent amounts of GST-Mtx2␦ fusion proteins based on SDS-polyacrylamide gel electrophoresis analysis (Fig. 3). b Ͼ20 g/ml is nontoxic.