Role of Interdomain Salt Bridges in the Pore-forming Ability of the Bacillus thuringiensis Toxins Cry1Aa and Cry1Ac*

The four salt bridges (Asp222–Arg281, Arg233–Glu288, Arg234–Glu274, and Asp242–Arg265) linking domains I and II in Cry1Aa were abolished individually in α-helix 7 mutants D222A, R233A, R234A, and D242A. Two additional mutants targeting the fourth salt bridge (R265A) and the double mutant (D242A/R265A) were rapidly degraded during trypsin activation. Mutations were also introduced in the corresponding Cry1Ac salt bridge (D242E, D242K, D242N, and D242P), but only D242N and D242P could be produced. All toxins tested, except D242A, were shown by light-scattering experiments to permeabilize Manduca sexta larval midgut brush border membrane vesicles. The three active Cry1Aa mutants at pH 10.5, as well as D222A at pH 7.5, demonstrated a faster rate of pore formation than Cry1Aa, suggesting that increases in molecular flexibility due to the removal of a salt bridge facilitated toxin insertion into the membrane. However, all mutants were considerably less toxic to M. sexta larvae than to the respective parental toxins, suggesting that increased flexibility made the toxins more susceptible to proteolysis in the insect midgut. Interdomain salt bridges, especially the Asp242–Arg265 bridge, therefore contribute greatly to the stability of the protein in the larval midgut, whereas their role in intrinsic pore-forming ability is relatively less important.

During sporulation, Bacillus thuringiensis produces a parasporal crystal body composed of one or more proteins that are toxic to a number of insect larvae (1) or to other invertebrates (2). After solubilization in the insect midgut and activation by intestinal proteases, these proteins bind to specific receptors at the surface of the apical brush border membrane of epithelial columnar cells, insert into the membrane, and form pores that disrupt midgut cellular functions (3)(4)(5).
Elucidation of the crystal structure of the coleopteranspecific Cry3A toxin (6) and the lepidopteran-specific Cry1Aa toxin (7) revealed a similar three-domain structure for both proteins. Domain I, composed of eight amphipathic ␣-helices, is thought to be involved in membrane insertion and pore formation (8 -15). Domain II, composed of three ␤-sheets and two short ␣-helices, is involved in the binding of the toxin to its receptor on the epithelial cell surface (16 -23). Domain III, composed of two ␤-sheets forming a face-to-face ␤-sandwich, appears to be involved in the stability (6), specificity (24 -26), and binding (27)(28)(29)(30)(31)(32)(33)(34) of the toxin.
These domains are closely packed together with the largest number of interdomain contacts found between domains I and II (6,7). In Cry1Aa, domains I and II are linked by four salt bridges: Asp 222 -Arg 281 , Arg 233 -Glu 288 , Arg 234 -Glu 274 , and Asp 242 -Arg 265 (7) (Fig. 1). Three salt bridges, structurally equivalent to Arg 233 -Glu 288 , Arg 234 -Glu 274 , and Asp 242 -Arg 265 , are also found in Cry3A (6,7). In addition, for all four salt bridges at least one of the amino acid residues forming the bridge is located within block 2, a sequence that is highly conserved among Cry toxins (1,5,36). However, only in the case of the Asp 242 -Arg 265 salt bridge are both amino acids located within block 2 ( Fig. 1). Salt bridges appear to play an important role in toxin stability and function. Mutations preventing the formation of the Asp 242 -Arg 265 (37,38), Arg 233 -Glu 288 (38), or Arg 234 -Glu 274 (38) salt bridges in Cry1Ab resulted in substantial losses of stability or activity. The importance of these salt bridges is further supported by the observation that domain I exchanges between different Cry1 proteins can lead to inactive recombinant proteins (39), possibly caused by the absence of one or more essential salt bridges. In contrast, disulfide bond engineering experiments with Cry1Aa have demonstrated that pore formation requires domain I to swing away from the rest of the molecule (14), implying that the interdomain salt bridges must be broken during toxin insertion into the membrane.
In the present study, each of the four salt bridges in Cry1Aa was eliminated by site-directed mutagenesis. Because others have shown that Cry1A mutations at Asp 242 or Arg 265 may result in unstable proteins (37,38), various amino acid alterations were created at Asp 242 including a double mutant D242A/R265A. Because the highly alkaline pH of the lepidopteran midgut, which ranges between 8 and 12 (40,41), is thought to play an important role in toxin function (5), all stable mutants were tested at pH values of 7.5 and 10.5. Although all mutations caused a substantial loss in toxicity, several mutants retained a capacity to form pores in midgut brush border membrane vesicles that was comparable with that of the respective parental toxins.

EXPERIMENTAL PROCEDURES
Mutagenesis-Cry1Aa mutants were created by oligonucleotide-directed in vitro mutagenesis using the Altered Sites II kit (Promega, Madison, WI) as recommended by the manufacturer. Mutated genes were subcloned in the pBU4 Escherichia coli-B. thuringiensis shuttle vector (42) and electroporated in the B. thuringiensis HD-1 acrystalloferous strain Cry Ϫ B as described (43). Cry1Ac mutants were created in the expression plasmid pMP39 (44) using the CLONTECH Transformer kit (CLONTECH, Palo Alto, CA). All mutants were sequenced using an Applied Biosystems (Foster City, CA) model 370A automated fluorescent sequencer.
Toxin Activation and Purification-Wild-type Cry1Aa and Cry1Aa mutant protoxins were produced as parasporal crystals in B. thuringiensis grown at 27°C in BP medium (45) containing 10 mg/ml glucose. Crystals were purified with Renografin gradients, solubilized, and trypsin-activated as described previously (46). Wild-type Cry1Ac and Cry1Ac mutant protoxins were produced as insoluble inclusions in E. coli and solubilized and activated as described (44). Activated toxins were purified by fast protein liquid chromatography using a Mono Q ion exchange column (Amersham Pharmacia Biotech), and bound toxin was eluted with a 50 -500 mM NaCl gradient in 40 mM carbonate buffer, pH 10.5 (44). The purity and integrity of all proteins were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (47), and protein concentrations were determined by the Bradford method (48) using bovine serum albumin as a standard.
Bioassays-Manduca sexta fertilized eggs were obtained from the Carolina Biological Supply Company (Burlington, NC). Toxicity assays were performed on neonate larvae with trypsin-activated toxins. Toxins were diluted in phosphate-buffered saline (8 mM Na 2 HPO 4 , 2 mM KH 2 PO 4 , and 150 mM NaCl, pH 7.4). Toxin samples (100 l) were layered onto artificial diet in 2-cm 2 wells and allowed to dry. A single larva was placed in each well and was reared at 27°C and 70% relative humidity with a photoperiod of 12 h of light and 12 h of darkness. Five groups of at least 30 larvae were tested at each toxin concentration. Mortality was recorded after 7 days and adjusted for mortality of control larvae that were reared in the absence of toxin. Cry1Aa and Cry1Ac were tested at five different toxin concentrations ranging from 25 to 200 ng/ml for Cry1Aa and from 12.5 to 75 ng/ml for Cry1Ac, and the data were analyzed by probit analysis (49). Mutant proteins were tested at 2 g/ml and depending on the mortality rate observed at this concentration, the tests were repeated at either 50 ng/ml or 50 g/ml.
Preparation of Brush Border Membrane Vesicles-Whole midguts were isolated from fifth-instar M. sexta larvae and freed of the attached Malpighian tubules. The midguts were then transected longitudinally to remove the peritrophic membranes and gut contents, rinsed thoroughly in ice-cold 300 mM sucrose, 5 mM EGTA, and 17 mM Tris/HCl, pH 7.5 and stored at Ϫ80°C until use. Brush border membrane vesicles were prepared by Mg 2ϩ precipitation and differential centrifugation (50). Vesicles were resuspended at 0.44 mg of protein/ml in 10 mM Hepes/KOH, pH 7.5 or CAPS/KOH, 1 pH 10.5 and allowed to equilibrate overnight at 4°C.
Light-scattering Assay-Brush border membrane permeability was analyzed with an osmotic swelling assay (51). At least 1 h prior to the experiments, the vesicles were further diluted to a final concentration of 0.40 mg of membrane protein/ml by the addition of 10 mg/ml bovine serum albumin to the buffer in which they were equilibrated. Toxins were assayed after a 60-min preincubation period with the vesicles to analyze the properties of the pores after having ample time to form or without preincubation to monitor the rate at which the toxins increased membrane permeability. The assays were initiated by rapidly mixing the vesicles with an equal volume of 10 mM Hepes/KOH, pH 7.5 or CAPS/ KOH, pH 10.5, 1 mg/ml bovine serum albumin, and 150 mM KCl or 300 mM sucrose or raffinose using a Hi-Tech Scientific (Salisbury, U K) stopped-flow rapid kinetics apparatus. For rate assays, 150 pmol of toxin/mg of membrane protein were added to the KCl substrate solution before mixing with the vesicles. Scattered light intensity was monitored at an angle of 90°at 23°C in a Spex Fluorolog CM-3 (Jobin Yvon Horiba, Edison, NJ) spectrofluorometer with monochromators set at 450 nm.
Data Analysis-Percent volume recovery was defined as 100 (1 Ϫ I t ), where I t is the relative scattered light intensity at time t. For rate assays, percent volume recovery was calculated for every data point, and values obtained for control vesicles assayed without toxin were subtracted from the experimental values measured in the presence of toxin. Data are reported as means Ϯ S.E. of at least three experiments, each performed in quintuplicate with different vesicle preparations. Statistical significance (p Ͻ 0.05) was determined with the two-tailed unpaired student's t test using the Instat version 1.13 program (Graph-PAD Software for Science, San Diego, CA).

RESULTS
Protoxin Expression and Activation-Whereas the production and trypsin-activation of most mutants were essentially equivalent to those of the respective parental toxins, the Cry1Ac mutants D242K and D242E could not be expressed in sufficient quantities for bioassays or light-scattering experiments. It was also found that the Cry1Aa single mutant R265A and the double mutant D242A/R265A were almost immediately degraded upon exposure to trypsin. Therefore, these mutants could not be further analyzed.
Toxicity-In agreement with previous studies (52,53), both Cry1Aa and Cry1Ac were strongly toxic to M. sexta larvae with Cry1Ac being twice as toxic as Cry1Aa (Table I). In comparison, all mutants tested were substantially less toxic (Table II). The loss of toxicity was most evident for the Cry1Aa mutant D242A and the Cry1Ac mutant D242P and least pronounced for the Cry1Aa mutant D222A and the Cry1Ac mutant D242N.
Pore-forming Ability-The ability of each mutant to form channels in the apical membrane of the insect midgut was analyzed with a light-scattering assay (51). When brush border  membrane vesicles, previously incubated for 60 min with either of the mutant toxins, were rapidly mixed with a hypertonic KCl solution, two distinct responses were observed (Fig. 2). Following a rapid shrinking of the vesicles (due to the hypertonic shock as evidenced by a rapid increase in scattered light intensity) the vesicles either swelled rapidly, at a rate dependent on toxin concentration as was the case for D222A ( Fig. 2A), or at a rate comparable with that observed for control vesicles, independent of toxin concentration as observed for D242A (Fig. 2B).
Although a contribution of buffer efflux during vesicle shrinking cannot be excluded, rapid swelling demonstrates a substantial increase in the KCl permeability of the vesicles when exposed to an active toxin.
To summarize the results, percent volume recovery after 3 s (derived from traces like those shown in Fig. 2) was plotted as a function of toxin concentration (Fig. 3). In contrast with D242A, the other Cry1Aa mutants, D222A, R233A, and R234A, as well as the Cry1Ac mutants, D242N and D242P, caused a dose-dependent increase in the permeability of the vesicles comparable with that observed with the respective parental toxins (Fig. 3). Similar results were obtained at pH values of 7.5 and 10.5, although at the latter a slightly but significantly higher permeability than with Cry1Aa was observed with the three Cry1Aa active mutants at the lower toxin concentrations.
All active toxins tested at 150 pmol of toxin/mg of membrane protein also permeabilized the membrane for sucrose and raffinose (Table III). Cry1Aa and its three active mutants caused a greater increase in the permeability of the vesicles to sucrose at pH 10.5 than at pH 7.5 (Fig. 4, A and C). This was also true for Cry1Ac but not for its two mutants (Fig. 4, B and D). A similar pattern was observed in the presence of raffinose although percent volume recovery at both pH values only became significantly different for Cry1Aa, Cry1Ac, and the Cry1Aa mutants when measured after more than ϳ5 s (Fig. 4, E-H).
Rate of Pore Formation-To investigate the kinetics of pore formation following toxin exposure, vesicles were mixed simultaneously with toxin and KCl. Under these conditions the vesicles began to swell after a short delay (Fig. 5). Cry1Ac caused the vesicles to swell more rapidly and after a shorter delay than Cry1Aa. For both toxins, however, increasing the pH from 7.5 to 10.5 resulted in a slightly longer delay and reduced swelling rate. The Cry1Aa mutants caused a pH-dependent increase in the rate of vesicle swelling relative to that observed in the presence of Cry1Aa. Although at pH 7.5 a significantly higher swelling rate was observed only for D222A (Fig. 5A), at pH 10.5 all three mutants caused a significantly more rapid swelling of the vesicles than Cry1Aa (Fig. 5C). On the other hand, at both pH values the rate of vesicle swelling was similar in the presence of Cry1Ac or either one of its mutants, D242N and D242P (Fig. 5, B and D). DISCUSSION The objective of the present study was to investigate the role of the salt bridges linking domains I and II in Cry1Aa and Cry1Ac. Although these two toxins, along with Cry1Ab, share a high level of homology and are therefore expected to have similar pore-forming properties, mutations preventing the formation of the Asp 242 -Arg 265 salt bridge had diverse effects on expression, stability, and function. Whereas the Cry1Aa mutant D242A could be produced normally but had very poor toxicity and pore-forming ability, mutants in which the arginine at position 265 was replaced by an alanine residue in the same protein were extremely sensitive to degradation by trypsin. These results correlate well with those from previous studies showing that mutations at either Asp 242 , Arg 265 , or both positions in Cry1Ab resulted in unstable and inactive mutant proteins (37,38).
Cry1Ac was modeled using the published Cry1Aa (7) and Cry3A (6) three-dimensional structures as templates (54). Interestingly, although Cry1Aa possesses four salt bridges in the interdomain region, there were only two salt bridges in Cry1Ac located in the same positions as those in Cry1Aa; no salt bridge was apparent between residues Asp 222 and Arg 281 or between residues Arg 234 and Glu 274 . The predicted atomic distances of the Cry1Ac salt bridges are as follows: Arg 233 -Glu 288 , 2.85 Å, and Asp 242 -Arg 265 , 2.27 Å and 2.96 Å. The former salt bridge is identical to that found in Cry1Aa, whereas the latter appears to be stronger.
The properties of the Cry1Ac Asp 242 mutants depended greatly on the nature of the mutation. Whereas D242E and  D242K could not be produced, D242N and D242P retained a wild-type pore-forming ability in brush border membrane vesicles. Although an effect on gene expression cannot be excluded, the extremely poor accumulation of D242E or D242K in the bacterial cell was likely due to a structural defect causing these mutant proteins to be rapidly degraded by endogenous proteases. In the case of D242K, normal polypeptide chain folding would result in the juxtaposition of two positive charges at a position normally occupied by a salt bridge. The replacement of the aspartic acid residue by a glutamic acid residue in D242E, however, does not introduce a modification in the charge of this residue but simply increases the length of its side chain by a single methylene group. This modification appears to be sufficient either to prevent the formation of the salt bridge by improper folding of the protein or, if the salt bridge is indeed formed, to introduce an angle or a gap between ␣-helix 7 and the ␤-sheet of domain II. This in turn may prevent the formation of the other salt bridges because the Asp 242 -Arg 265 salt bridge is located nearest to the hinge region linking domains I and II (Fig. 1).
The fact that D242N and D242P retained good pore-forming activity was surprising because a D242N mutant in Cry1Ab was among those previously reported to be unstable (37). Moreover, the introduction of a helix-interrupting proline residue at this position was expected to have profound effects on the structure of the protein by destabilizing the structure of ␣-helix 7. However, when examined by mutational analysis conducted with the Swiss PDB viewer software on the published structure of Cry1Aa and on that of Cry1Ac obtained by Swiss Model modeling, a D242P mutation in ␣-helix 7 was predicted to have very little effect on the structure of the helix and no overall effect on the Cry1Aa or Cry1Ac toxin structures. Following energy minimization, the single-site mutation resulted in a maximum displacement of 0.11 Å of the C ␣ atom at amino acid position 242. It should be noted, however, that of all the mutants tested in the present study D242N retained the strongest toxicity, and D242P was, along with D242A from Cry1Aa, the least toxic to M. sexta larvae.
Mutations that prevented the formation of any single salt bridge linking domains I and II in Cry1Aa or the Asp 242 -Arg 265 salt bridge in Cry1Ac clearly resulted in a substantial decrease in toxicity of the protein. At 2 g of toxin/ml, a concentration 32-fold higher than the LC 50 measured for Cry1Aa and 61-fold higher than that of Cry1Ac, only D222A and D242N were able to kill more than 50% of the larvae. The toxicity level retained by the Cry1Aa mutants was directly related to the position of the abolished salt bridge relative to the hinge region linking domains I and II (7). Removal of the salt bridge located farthest from the hinge region (D222A, Fig. 1) resulted in a toxin with the highest toxicity, whereas elimination of the salt bridge located closest to the hinge region (D242A) rendered the mutant protein essentially non-toxic. R233A and R234A, in which the salt bridges located near the middle of ␣-helix 7 were abolished, retained an intermediate level of toxicity.
Except for D242A, a reduction in toxicity could not be attributed to a loss in the pore-forming ability of the mutants. The increase in vesicle permeability to KCl, sucrose, and raffinose observed with each of the active mutants was comparable with that observed with the respective parental toxins following a 60-min preincubation of the vesicles with the toxin. Moreover, the rate at which membrane permeability increased following exposure to the toxin was at least as high for the mutants as for the wild-type toxins. These results contrast with those of another study in which the R233A mutant of Cry1Ab was found to have a significantly reduced ability to inhibit the short circuit current through the epithelium of midguts isolated from Lymantria dispar (38).
With the Cry1Aa mutant D222A at pH 7.5 and all three Cry1Aa active mutants at pH 10.5, membrane permeability  increased faster upon exposure to the mutants than in the presence of the parental toxin. This result suggests that the removal of one of the salt bridges facilitates the insertion of the toxin into the membrane, at least under in vitro conditions. Interestingly, at pH 10.5 the rate at which vesicle permeability increased in the presence of these mutant toxins correlated with toxicity and, as mentioned above, with the position of the missing salt bridge relative to the hinge region. A faster rate of membrane insertion may increase the toxicity of the mutants by reducing the time they are exposed to proteases in the midgut. Cry1Aa was nevertheless considerably more toxic than any of its mutants, and removal of a salt bridge is very likely to have increased the susceptibility of the mutants to midgut proteases. Cases of toxins with a strong capacity to form pores in vitro despite a relatively poor toxicity have been reported previously (25,55,56). As was discussed extensively (55), this may result from a poorer stability or a stronger sensitivity to proteolysis under the harsh conditions encountered in the insect midgut than in the somewhat idealized in vitro conditions. Among the intestinal factors that could affect toxin activity, the strongly alkaline pH characteristic of the larval midgut of lepidopteran insects (40,41) had only minor effects on the in vitro activity of Cry1Aa and Cry1Ac. A slightly higher permeability to sucrose and raffinose was nevertheless observed at pH 10.5 than at pH 7.5, consistent with a slight increase in pore size with increasing pH as was previously reported for Cry1Ac (57). A similar effect of pH was observed for all three Cry1Aa mutants, but it was not detected with either of the Cry1Ac mutants.
In conclusion, our results along with those of other laboratories (37,38) clearly show that the interdomain salt bridges linking domains I and II play an important role in toxin stability and toxicity. Presumably, these salt bridges are necessary to maintain the toxin in a tightly packed conformation in the midgut until, after binding to its membrane receptor, the toxin undergoes a conformational change leading to insertion into the membrane and pore formation. In the absence of pro- teases or other intestinal factors yet to be identified, removal of either one of these salt bridges does not necessarily alter the toxin's capacity to form a pore in the midgut membrane. On the other hand, under in vitro conditions removal of a salt bridge can even increase the rate of pore formation by Cry1Aa.  A and B) or CAPS/ KOH, pH 10.5 (C and D) and 150 pmol/mg membrane protein of Cry1Aa or either one of the Cry1Aa mutants (A and C), or with Cry1Ac or one of its mutants (B and D), without preincubation with the toxin. Percent volume recovery was calculated for each experimental point, and control values were subtracted from those obtained with the toxins. For clarity, error bars are only shown every 51st data point.