Structure-function relationships of a membrane pore forming toxin revealed by reversion mutagenesis

Abstract

Cyt2Aa1 is a haemolytic membrane pore forming toxin produced by Bacillus thuringiensis subsp. kyushuensis. To investigate membrane pore formation by this toxin, second-site revertants of an inactive mutant toxin Cyt2Aa1-I150A were generated by random mutagenesis using error-prone PCR. The decrease in side chain length caused by the replacement of isoleucine by alanine at position 150 in the αD-β4 loop results in the loss of important van der Waals contacts that exist in the native protein between I150 and K199 and L203 on αE. 28 independent revertants of I150A were obtained and their relative toxicity can be explained by the position of the residue in the structure and the effect of the mutation on side-chain interactions. Analysis of these revertants revealed that residues on αA, αB, αC, αD and the loops between αA and αB, αD and β5, β6 and β7 are important in pore formation. These residues are on the surface of the molecule suggesting that they may participate in membrane binding and toxin oligomerization. Changing the properties of the amino acid side-chains of these residues could affect the conformational changes required to transform the water-soluble toxin into the membrane insertion competent state.

• Bacillus thuringiensis
• cytolytic δ-endotoxin
• Cyt2Aa1 toxin
• reversion mutagenesis
• transmembrane pore

Introduction

Bacillus thuringiensis (Bt) insecticidal protein toxins are divided into two families (Cry and Cyt) with unrelated amino acid sequences [1], [2]. Although the X-ray crystallographic structures of Cry and Cyt toxins are entirely different [3], [4], the mechanism of action of these two families is thought to be the same. Both Cry and Cyt toxins form small pores approximately 1–2 nm in diameter in the larval midgut epithelial cell membranes leading to colloid-osmotic lysis [5]. Cyt2Aa1 which is the subject of this study is produced as a 29 kDa protoxin which is processed at both ends by gut proteases to generate the active 21–23 kDa toxin [6].

Cyt2Aa1 exists as an inactive protoxin dimer linked by intertwined N-terminal strands in a continuous 12-stranded β-sheet [4]. The high sequence similarity between members of the Cyt toxin family (see Figure 1) indicates that they have the same structural fold as Cyt2Aa1 [4].

Figure 1. Multiple sequence alignment of amino acids from Cyt δ-endotoxins. The alignment was performed using Clustal X. The number on the right of each sequence indicates the residue number at the end of each line. The symbols (*), (:) and (.) represent identical, highly conserved and less conserved amino acids, respectively. The graph under the sequences indicates the degree of amino acid identity for all toxins (from 0% at the bottom to 100% at the top). The secondary structures of Cyt2Aa1 are shown on top of the sequence. The ruler number under the sequences is shown for indication of the alignment position only. The residues in Cyt2Aa1 which were selected for mutagenesis are marked by |. This Figure is reproduced in colour in Molecular Membrane Biology online.

Protease processing by insect gut enzymes cleaves the intertwined N-terminal β1 strands responsible for dimerization to release the active toxin as a monomer and simultaneously eliminate several structural barriers to pore formation [4], [6]. Activated toxin molecules can form pores in cell membranes [7], phospholipid vesicles and bilayers [8–10]. Lytic pore formation is believed to require toxin monomers to assemble into an oligomeric complex [4], [11], [12].

The Cyt toxin pore is not likely to be formed by the six α-helices in the structure, as they are too short to span the 30 Å width of the lipid bilayer [4]. In contrast, strands β5, β6 and β7 are sufficiently long to span the bilayer. Over these three strands the sheet shows an amphiphilic or hydrophobic character and therefore they could oligomerise to form a β-barrel pore [4].

The aim of this study was to investigate structure-function relationships associated with membrane pore formation by Cyt2Aa1. A non-haemolytic mutant of Cyt2Aa1 was generated and used as a template for reversion mutagenesis. Analysis of the revertant proteins should provide information about residues that are important in pore formation. This method proved valuable in the study of Staphylococcal α-hemolysin [13]. Cyt2Aa1 was chosen in preference to other Cyt δ-endotoxins because the availability of its crystal structure [4] enables the position of mutated residues to be mapped on the structure of the molecule.

Materials and methods

Plasmids and oligonucleotides

Plasmid pMSV1Cyt2Aa1 (wild type) is based on pALTER (Promega) and carries the full-length cyt2Aa1 gene from Bt subsp. kyushuensis[14]. pUC18 was described by Yanisch-Perron et al. [15]. Two oligonucleotides were designed to change the sequence of Cyt2Aa1 from isoleucine at position 150 and from threonine at position 221 to alanine (I150A and T221A). To facilitate identification of the mutants, the oligonucleotide for I150A was designed so as to introduce a new Hae III restriction site and the T221A oligonucleotide was designed to abolish a Hind III site. DNA sequences of oligonucleotides for I150A and T221A were 5′-TTTAGACACTCAGGCCGATGAAGCATG-3′ and 5′-TAAAATGAAAGCTCTTGCTTTAGTTCAAG-3′, respectively. Oligonucleotides for error-prone PCR were universal M13 forward and reverse sequencing primers. All oligonucleotides were produced by the Oligonucleotide Synthesis Facility, Department of Biochemistry, University of Cambridge, UK.

Site-directed mutagenesis and cloning

Site-directed mutagenesis was performed using the Altered Sites II In Vitro Mutagenesis System (Promega) as described by the manufacturer. Single-stranded template was prepared from pMSV1Cyt2Aa1 using R408 helper phage. Mutagenesis reactions were used to transform the mismatch repair deficient E. coli strain ES1301 mutS by electroporation [16]. Transformed cells were grown overnight in 5 ml LB (Luria-Bertani) medium containing 100 µg/ml ampicillin. Plasmid DNA was then isolated and re-transformed into E. coli JM109 or TG1 cells. Transformants were analysed for introduced mutations by restriction endonuclease digestion and DNA sequencing. Wild type and mutant genes were also sub-cloned into plasmid pUC18 for high efficient expression in E. coli.

Error-prone PCR

Error-prone PCR was performed as described by Panchal and Bayley [13]. The template for PCR was the cyt2Aa1-I150A gene in plasmid pUC18. The 3′- and 5′- universal primers were as described above. Random mutations were introduced using a large number of PCR cycles (50) and a high concentration of dNTPs (0.4 mM each). PCR reactions were carried out in standard PCR buffer (Pharmacia) plus 1.0 ng of plasmid template, 0.8 mM of each primer, 3.0 mM MgCl₂ and 2.5 units of Taq DNA polymerase (Pharmacia) in 50 µl reaction volume for 50 cycles at 94°C, 30 sec; 50°C, 30 sec; 72°C, 2 min (10 min for the last cycle). PCR products were then cloned into plasmid pUC18 and transformed into E. coli JM109 or TG1.

Screening for revertants

The screening method was modified from the method described by Panchal and Bayley [13]. Small colonies that appeared after 8 h growth at 37°C were replicated onto nitrocellulose membranes (Gelman Sciences), which were then placed colony side up onto a second LB-ampicillin plate previously spread with 50 µl of 0.1 M IPTG and 50 µl of 2% X-gal, and incubated overnight at 37^oC. The replicate colonies were lysed by removing the nitrocellulose membranes and placing them (colony side up) for 1 h on Whatman paper (no. 3) soaked with 0.1 M NaHCO₃, 1% Triton X-100 (v/v) containing 2 mg/ml lysozyme. The toxin inclusions were solubilized and processed by removing the membranes and placing them on Whatman paper (no. 3) soaked with 50 mM Na₂CO₃ pH 10.5 containing 0.1 mg/ml proteinase K for 1 h at 37°C. The membranes were then washed with PBS buffer (8 mM Na₂HPO₄, 1.5 mM KH₂PO₄, 140 mM NaCl, 2.7 mM KCl pH 7.4) plus 1 mg/ml BSA by gently floating the membranes (colony side up) on the surface of the wash buffer for 15 min. The membranes were then inverted onto blood agar plates prepared by adding 2.5 ml of washed human blood to 100 ml of tryptone soya agar (Oxoid) held at 50°C. Colonies showing distinct zones of haemolysis within 24 h at room temperature were individually picked from the master plates for further analysis.

Protein preparation and processing

Cyt2Aa1 inclusions were released from E. coli cells by sonication as described previously [17]. Prior to use, the inclusions were solubilized at 1–2 mg/ml at 37°C for 1 h in 50 mM Na₂CO₃ pH 10.5 with occasional shaking. Solubilized toxin was separated from insoluble materials by centrifugation at 12,000 g for 5 min and stored at −20°C in 1 ml aliquots. Concentrations of solubilized toxin were determined by the method of Bradford [18] using the BioRad protein assay kit and BSA as standard. For proteolytic processing, the solubilized material was mixed with 0.1–10% (w/w) proteinase K and incubated at 37°C for 10 min to 3 h.

Circular dichroism analysis

Protein samples in 50 mM Na₂CO₃ pH 10.5 in a 0.02 cm cylindrical quartz cuvette were analysed using a Jasco J-715 spectropolarimeter at 25^oC. The instrument was calibrated with camphorsulfonic acid (CSA) and continuously purged with nitrogen gas. Scanning was set to accumulate for 4 averaged spectra in the 180–320 nm region and subtracted with baseline spectra from buffer. All spectra were analysed and compared at the same protein concentration.

Intrinsic fluorescence spectroscopy

Cyt toxin fluorescence spectra were obtained by emission scanning of protein solution in a 10 mm path length square quartz cell. The protein samples were excited at 280 nm using a Jasco FP-6500 spectrofluorometer. The emission spectra were recorded from 300–500 nm with medium scanning speed using 3 nm of excitation and emission slit width. All spectra were subtracted with baseline spectra from buffer.

In vitro haemolysis assay

Haemolysis assays in microtitre plates were as described by Thomas and Ellar [19] except that a 4% human blood (National Blood Transfusion Service) suspension in haemolysis buffer (50% PBS, 2.25% glucose, 0.05% gelatine) was used. Processed toxin samples were diluted in two-fold serial dilutions with haemolysis buffer. 0.1 ml aliquots of the diluted toxin samples were mixed with 0.1 ml of diluted blood cells and left at room temperature. The haemolytic end-point was recorded after 24 h of incubation.

Mosquito larvicidal assay

Mosquito larvicidal assays were performed as described by Koni and Ellar [6] with modification. Solubilized toxin samples in 50 mM Na₂CO₃ pH 10.5 were precipitated by adding 0.1 volume of 11% (w/v) citric acid to lower the pH to approximately 4.5 and the samples stored at −20^oC for at least 2 h until required. The material was then used directly without further manipulation. The assays were carried out 3 times in duplicate using 10 larvae of 3-day-old Aedes aegypti (reared from eggs supplied by Zeneca Agrochemical, UK and the Institute of Molecular Biology and Genetics, Mahidol University, Thailand) in 1.5 ml microtubes containing 0.9 ml of water in which the larvae were reared and 0.1 ml of precipitated toxin samples. The final concentration of toxin in each tube was a 2-fold serial dilution from 250 µg/ml to 0.25 µg/ml. The negative control contained 90 µl of 50 mM Na₂CO₃ pH 10.5 and 10 µl of 11% citric acid. Mortality was recorded after 24 h at room temperature.

Results

Expression and activity of primary mutants

Ward et al. [20] demonstrated that replacing K154 or K225 with alanine in the closely related toxin Cyt1Aa (Figure 1) caused a loss of activity without affecting expression. Therefore, it was decided to substitute alanine at the equivalent residues I150 and T221 in Cyt2Aa1. Both mutant proteins expressed at levels similar to that of the wild type and were soluble in 50 mM Na₂CO₃ pH 10.5 without the addition of a reducing agent suggesting that the mutations did not affect dimerization and crystal packing. No differences between solubilized mutant proteins and wild type protein were detectable on SDS-PAGE (data not shown). Proteinase K treatment of the mutant proteins gave identical products to the wild type (data not shown) indicating that the processing sites in the mutants had not changed. The haemolytic activity of Cyt2Aa1-T221A was similar to the wild type, but Cyt2Aa1-I150A was not haemolytic even at very high concentrations (250 µg/ml; Table I). Finally analysis of both proteins using Circular Dichroism (CD) and intrinsic fluorescence revealed no significant differences in structure between the inactive mutant and the wild type (data not shown).

Table I.  Activity of Cyt2Aa1-wt, Cyt2Aa1-T221A, Cyt2Aa1-I150A and revertants of Cyt2Aa1-I150A. The revertants are indicated in square bracket [].

Cyt2Aa1 toxin	Haemolysis end-point (µg/ml)^a	Mosquito larvicidal activity LC50 (µg/ml)^b
wt	0.25	1–2
T221A	0.50	2–4
I150A	>250	>250
[Q50R]	250	>250
[N74D]	4	4–8
[F77S]	250*	128–256
[F77L]	2	4–8
[I83T]	250*	>250
[V105A]	2	8–16
[T129A]	32	32–64
[T148I]	1	1–2
[A150V]	0.25	1–2
[S166G]	2	2–4
[N170S]	250*	128–256
[E178G]	250	128–256
[V195A]	2	8–16
[N8S,F47S]	250*	>250
[F47L,N259H]	16	16–32
[S9F,N74D]	4	4–8
[F77L,S210N]	16	64–128
[I68M,S166G]	16*	64–128
[D69A,I171T]	250*	>250
[K79R,I83T]	250*	>250
[V109A,A185V]	16	4–8
[V111I,K218R]	250	>250
[V125A,A185T]	1	4–8
[S166G,I171V]	2	2–4
[A185V,F205L]	4	2–4
[K14R,V49A,S130N]	250	128–256
[Q119R,T180A,I207T]	250*	128–256
[D39G,S108P,I171T,N259D]	16*	64–128

^aHaemolysis end-point of toxin activated with 1% (w/w) proteinase K at 37^oC for 1 h (*activated with 0.5% (w/w) proteinase K at 37^oC for 30 min). ^bConcentration of precipitated toxin that cause 50% mortality (LC₅₀) of 3-day-old Aedes aegypti laevae. Probit analysis was not applied because the change in toxicity from one dose to the next was dramatic.

The Cyt2Aa1-I150A mutant was therefore selected as the template for reversion mutagenesis. The equivalent mutation in Cyt1Aa (K154A) was one of 6 mutations showing low toxicity and reduced phospholipid binding, but no change in expression levels [20]. All these mutations mapped to loops lying along the edge of the β-sheet adjacent to the helix pair αC-αD [4]. Because alanine replaced a charged residue in 4 of these mutants and because these charged residues are not conserved between Cyt1Aa1 and Cyt2Aa1, Li et al. [4] proposed that this region is required to remain exposed to the aqueous phase when β5, β6 and β7 insert into the membrane. In the remaining two low toxicity mutants, alanine was replaced with proline suggesting that main chain flexibility as well as the appropriate hydrophobicity is needed in this region. In the model for toxin insertion proposed by Li et al. [4] these properties were important in allowing helix pair αC-αD to peel away from the β-sheet to lie on the membrane surface while the sheet itself rearranges to form the oligomeric trans-membrane pore. On the basis of these suggestions the loss of toxicity when the highly hydrophobic isoleucine is replaced by alanine in Cyt2Aa1-I150A could result from a disruption of the hydrophobicity which positions the surface-located αC-αD helix pair adjacent to the edge of the inserted β-sheet. Because monomer interaction following membrane insertion is crucial to the assembly of the oligomeric pore it is important to keep in mind the possibility that interference with monomer recognition could also result in loss of toxicity. If the activity of the Cyt2Aa1-I150A mutant could be recovered by substitution at other residue positions in the molecule, the location of these substitutions could therefore reveal regions important in the steps of membrane recognition, toxin insertion and pore formation.

Reversion mutagenesis

Reversion mutagenesis of the inactive Cyt2Aa1-I150A was performed using error-prone PCR. Random mutations were introduced into PCR products by using Taq polymerase which lacks 3′-5′ exonuclease activities [21] and has an error rate of 2×10⁻⁴ errors/base [22]. This error rate was further increased by using a large number of PCR cycles and a high concentration of dNTPs. PCR products were then cloned into pUC18, expressed in E. coli and colonies screened on blood agar plates. This screen is a powerful method to identify a few haemolytic revertants among thousands of non-haemolytic colonies [13]. In the work reported here, 40 positive colonies were obtained after screening approximately 10,000 colonies. This reversion frequency of 0.4% is very similar to that reported for α-haemolysin (0.5%) [13]. DNA sequencing of the positive clones revealed 28 different revertants. Of these, 13 revertants were single mutations, 12 were double mutations, 2 were triple mutations and 1 was a quadruple mutation (see Table I). Four single revertants were found repeatedly: [V105A] (7 colonies), [N170S] (4 colonies), [A150V] (3 colonies) and [T148I] (2 colonies). The positions of revertant residues in the primary sequence and in the toxin structure are shown in Figures 2A and 2B, respectively.

Figure 2. Amino acid sequence and structure of wild type Cyt2Aa1. (A) The position of the original inactive mutant (I150A) is marked by * and the positions of revertant residues are shown in bold. Proteinase K processing sites are indicated by arrows. (B) 3-D structure of Cyt2Aa1 from different views showing the positions of wild type residues that were replaced in the revertants. The images were produced by MolScript v2.0.2 [33]. This Figure is reproduced in colour in Molecular Membrane Biology online.

Solubilization and proteolytic processing of revertant toxins

All revertant proteins expressed in E. coli at the same level as the wild type and formed typical inclusion bodies that could be solubilized in 50 mM Na₂CO₃ pH 10.5 without the need for a reducing agent (data not shown). This result was not unexpected since none of the revertants has a residue changed to cysteine. Cyt2Aa1 contains a single cysteine residue located near the N-terminus which does not appear to play an important role in crystal packing [4]. After solubilization, all revertant proteins showed SDS-PAGE profiles similar to the wild type (data not shown).

Most revertants gave similar products to the wild type when treated with proteinase K to simulate protoxin activation (data not shown). However, 9 revertants: [F77S], [I83T], [N170S], [N8S,F47S], [I68M,S166G], [D69A,I171T], [K79R,I83T], [Q119R,T180A,I207T] and [D39G,S108P,I171T, N259D] were unusually sensitive to proteinase K digestion and were almost completely digested by 0.5% (w/w) proteinase K at 37^oC for 1 h. These proteins were only partially digested by 0.1% (w/w) proteinase K under the same conditions (data not shown). Most of the mutated residues in the proteinase K-sensitive revertants were located on the surface of the molecule (see Figure 2B). The remaining two mutated residues in this group (N170S and I171T) are located in the middle of β5. Interestingly 8 of these 9 protease sensitive revertants which all showed much reduced mosquitocidal and haemolytic activities compared to the wild type toxin, have a residue mutated to serine or threonine. The protease sensitivity of the relatively toxic revertant [I68M,S166G] is likely to be the result of I68M because the active form of revertant [S166G] is not sensitive to proteinase K digestion.

Activity of the revertants

The results of haemolysis assays (see Table I) show that none of the revertant toxins was as active as the wild type, except [A150V]. Larvicidal assays using Aedes aegypti generally showed very similar results to the haemolytic assays, but there were some revertants (e.g., [T148I] and [A185V,F205L]) with significantly less haemolytic activity but comparable larvicidal activity to the wild type (Table I).

The revertants could be divided into three groups according to their activity. The first group, [N74D], [F77L], [V105A], [T148I], [S166G], [V195A], [S9F,N74D], [V109A,A185V], [V125A,A185T], [S166G,I171V], and [A185V,F205L], had relatively high activity similar to wild type. A second group, [T129A], [F47L,N259H], [I68M,S166G], and [F77L,S210N], had moderate activity, while a third group, containing the remaining revertants all had low activity. The revertant [D39G,S108P,I171T,N259D] may be classified into the moderate or high activity group, although its activity is very difficult to quantitate because of its protease sensitivity.

Positions of revertant residues in the structure of molecule

Two revertants ([A150V] and [T148I]) contain mutations close to or at the position occupied by A150 in the inactive primary mutant. Both of these return a hydrophobic residue to the loop at the edge of the β-sheet adjacent to the helix pair αC-αD from which isoleucine was removed in the primary inactive mutant. The fact that both revertants showed very high activity (Table I) indicated that hydrophobicity is required in this segment.

Seven of the revertants ([N74D], [F77S], [F77L], [S9F,N74D], [F77L,S210N], [I68M,S166G], [D69A,I171T]) contained mutations in the loop separating helices αA-αB. (Revertant [F77L] appears to be more active than revertants [F77S] and the double revertant [F77L,S210N], but the intrinsic activities of these three mutants are difficult to measure because their low activities are undoubtedly partly due to their increased protease sensitivity.) In the protoxin, residues in this loop play a role in stabilizing the packing of helix pair αA-αB against the β-sheet [4]. It was suggested above that an impairment of the hinge function of the loop linking helix αD to the β-sheet causes the loss of activity in the Cyt2Aa1-I150A mutant. The location of these seven revertants suggests that mutations in the helix αA-αB loop can compensate for this. Since the report by Du et al. [23] indicates that helices αA and αB are exposed on the membrane surface, it is possible that disruption to the connectivity between helix pair αA-αB and the β-sheet may compensate for the change in the connectivity between helix pair αC-αD and the sheet that has occurred in the Cyt2Aa1-I150A mutant. The two revertants located in β2 ([N8S,F47S], [F47L,N259H], or the adjacent β2-αA loop ([K14R,V49A,S130N], [Q50R]) may compensate by a similar mechanism.

Three revertants ([T129A], [V125A,A185T] and [K14R,V49A,S130N]) have replacements in the loop joining helices C and D. This may indicate that disruption to the connectivity between this region and the sheet can compensate for the inactivating I150A change in the segment between helix αD and the sheet. It is noteworthy that the addition of the A185T replacement in the double revertant [V125A,A185T] dramatically increases the activity. If as suggested by Li et al. [4] and Du et al. [23] β5, β6 and β7 from several monomers span the membrane in the oligomeric pore, these strands would have alternate residues facing the membrane lipid or the channel lumen. A185 in β6 faces the membrane lipid [4] and a change to a more hydrophobic residue may favour sheet insertion. This may be a partial explanation for the activity of revertant [A185V,F205L]. However A185 is hydrogen bonded to I175 in β5 and a strengthening of this interaction in A185T may also promote β5, β6 and β7 insertion. The hydrophobicity of the residue at position 205 in the loop between helix αE and β7 is highly conserved (see Figure 1) and any restoration of activity caused by the F205L change should be due to other reasons.

Four revertants ([V109A,A185V], [V111I,K218R], [Q119R,T180A,I207T], [D39G,S108P,I171T,N259D]) are found with replacements in helix αC and one in helix αB ([I83T]). The important intra-molecular salt bridge between E192 and K218 [4] is preserved in [V111I,K218R] and therefore restoration of activity is attributable to the V111I change. The finding that these helices are not membrane-protected [23] suggests that in addition to their importance in connectivity to the sheet, their surface location may also be important in the monomer recognition that precedes oligomerization. This is supported by studies using synthetic peptides [24], [25], reporting that αA and αC had the ability to bind the membrane and self-associate. Thus changes to both the amphipathic character and sequence in the surface exposed helices may also be compensatory.

The revertants [S166G] and [S166G,I171V] show similar levels of activity, suggesting that the conservative change from isoleucine to valine in the double revertant may have no effect on the toxin activity. [S166G] is located at the N-terminal end of β5 where it forms a hydrogen bond with H158 in β4. Du et al. [23] have shown that β2 remains on the membrane surface after insertion of β5. Therefore one suggestion to explain restoration of activity in this revertant would be a decrease in β2–β5 connectivity caused by substitution of glycine for serine. A similar explanation could be proposed for restoration of activity in the mutant [N170S] since N170 forms a hydrogen bond with N44 in β2. (The increased protease sensitivity of N170S makes it difficult to compare its intrinsic activity with that of S166G.).

The revertant [E178G] is located in the β5–β6 loop which is proposed to be oriented towards the cell interior during sheet insertion [4]. Although 5 of the loop residues are highly conserved (Figure 1) the lack of conservation at this position suggests that the negative charge on E178 is not important during insertion. The ability of a glycine substitution at this position to compensate for the distal I150A substitution may suggest that loop flexibility favours membrane insertion.

Discussion

The α/β architecture of the Cyt2Aa1 toxin monomer is unusual in that the two outer layers sandwiching the β-sheet consist of helical hairpins that can lift upwards and away from the sheet without unraveling it [4]. Helices αA, αB, αC and αD are amphipathic (see Figure 3) with hydrophobic residues packed against the β-sheet. The model for membrane insertion and pore formation proposes that upon membrane contact, helix pair αC-αD swings away from the sheet to lie on the membrane surface allowing β5, β6 and β7 to move into the lipid bilayer with the loop between β5 and β6 oriented towards the cell interior. Subsequent association of these inserted strands from several monomers creates an oligomeric pore [4]. The appearance of new proteolytic sites upon toxin binding to the membrane and the location of these sites [23] are consistent with the model. In a variation on their proposal Li et al. [4] suggested that helix hairpin αA-αB might remain packed against the sheet during insertion rather than remaining on the surface. This appears to be ruled out by the recent finding that only the segment containing β4, β5, β6, αE and β7 remains tightly associated with the membrane after proteolysis of membrane bound Cyt2Aa toxin [23].

Figure 3. Helical wheels of helices αA, αB, αC and αD. All helices show amphipathic character. Residues on the hydrophobic face could interact with the lipid bilayer of the membrane while residues on the hydrophylic side may interact with residues from other monomers to form the oligomer. This Figure is reproduced in colour in Molecular Membrane Biology online.

As measured by their solubility, protease sensitivity, expression levels and reaction with wild-type antibody (data not shown), the non-toxic Cyt2Aa1-I150A and the wild type toxin were identical. In addition no significant structural differences between the proteins were detectable using Circular Dichroism (CD) and intrinsic fluorescence. The dramatic difference in toxicity between the two proteins despite this similarity in properties and structure, suggests that replacement of isoleucine by alanine may affect the conformational changes required for transforming the water-soluble toxin into the membrane-bound pore form. Li et al. [4] proposed a key role in membrane insertion for the hinge region between helix pair αC-αD and the β-sheet both in terms of surface-seeking ability and flexibility. Because residue 150 is located in this region, the loss of activity in the Cyt2Aa1-I150A mutant may originate in changes in one or both of these properties. By reversion mutagenesis we have located residues proximal and distal from this mutation that can restore toxin activity in this mutant. Comparative analysis of these revertants may therefore help us to understand the insertion mechanism.

Proteolytic activation of the Cyt2Aa1 protoxin removes 33–37 amino acids at the N-terminus and 22-31 amino acids at the C-terminus [6] to generate either the fully-activated (S38 to S228) or the partially activated toxin (T34 to F237 or S38 to F237 or T34 to S228). Mutation of residues upstream or downstream of these regions should have no effect on toxin activity. Therefore, residues N8S, S9F, K14R, N259H and N259D contained in revertants [N8S,F47S], [S9F,N74D], [K14R,V49A,S130N], [F47L,N259H] and [D39G,S108P,I171T,N259D] are not considered to be responsible for restoring the activity of these revertants.

Combining our results with the findings of Du et al. [23] lead us to refine the model proposed by Li et al. [4] to describe the events taking place during pore formation (see Figure 4). The refined model proposes that upon membrane contact both helix hairpins αA-αB?, αC-αD swing away from the sheet. These helices together with αF, β2, β3, β4 and the loops between them locate on the membrane surface with β5, β6 and β7 inserting into the membrane. After insertion, this places alanine in the primary inactive Cyt2Aa1-I150A mutant on the membrane surface.

Figure 4. Schematic diagram of a possible model for insertion of Cyt2Aa1 into the cell membrane. The helices αA, αB, αC and αD bi nd the membrane and form an oligomer with other protomers. The regions between αD and β5 and between β6 and β7 act as hinges during the conformational changes following membrane binding. Sheets β5, β6 and β7 insert into the membrane and form an oligomer with similar sheets from other protomers to form a β-barrel pore. Residues in helices and sheets are coloured according to their hydrophobicity (red for hydrophobic, yellow for polar or charged and green for tyrosine). Sheet β3 which links helices αB and αC is not shown. The positions of the primary inactive mutant (I150A) and the high activity revertants are shown in ball and stick as the original residues in the wild type. I175, A182 and A225 are the residues at the C or N termini of β5, β6 and β7, respectively.

Experiments to investigate the effect of temperature on pore formation of Cyt2Aa1 and studies on membrane insertion and pore formation of acrylodan-labelled mutants of Cyt2Aa1 [17], [26] show that the toxin binds and inserts into the membrane as a monomer. Oligomerization occurs when toxin molecules bind in close proximity to each other and the pores are formed when oligomers reach a critical number. Cyt2Aa thus exists in three different configurations (water-soluble monomer, membrane inserted monomer and membrane inserted oligomer) in any or all of which a residue may have an important role. Toxicity may therefore be restored in a revertant because of compensatory changes in the steps of membrane insertion or oligomerization or both. Using the refined model, it could be suggested that in addition to a possible impairment of the hinge region between helix pair αC-αD and the sheet, the loss of toxicity in Cyt2Aa1-I150A could be due to the reduction of binding between the toxin molecule and the membrane surface.

If the monomer-monomer recognition that precedes pore formation involves recognition/association between the surface exposed elements, then clearly toxicity may be affected by a broad range of residue changes in the elements. For example αA, αB, and αC are amphipathic (see Figure 3) and the model proposes that residues on the hydrophobic face interact with the lipid bilayer while residues on the other side of the helix are available to recognize residues from other monomers to form the oligomer. V109 is the only hydrophobic residue located on the hydrophilic face of helix C (Figure 3) which is known to be capable of self-association in a membrane [24], [25]. Substitution of V109 by alanine at this position in revertant [V109A,A185V] may therefore further stabilize the surface location of this element and/or increase its ability to oligomerize with other monomers. Many of the other high activity revertants that map to the loops or helices exposed on the surface in the model e.g., N74D, F77L, V105A, V109A, V125A, and T148I, may compensate by similar mechanisms.

The model shows β4 to be on the membrane surface rather than inserted in the bilayer. Proteolysis of membrane-bound Cyt2Aa1 occurred at a number of cleavage sites in the αD-β4 loop to leave β4 associated with β5, β6 and β7 in the membrane protected fragment [23]. However proteolysis of membrane bound Cyt1Aa revealed cleavage sites in β4 including one at the C-terminal end. This indicates that in Cyt1Aa, β4 is membrane bound but not inserted and we suggest this is also likely to be true for Cyt2Aa1.

Li et al. [4] proposed that β5, β6 and β7 from several toxins form a pore based on a β-barrel structure. Experiments using polarity sensitive acrylodan-labeled toxin have confirmed that the pore is oligomeric with each monomer contributing three strands (β5, β6, β7) [17]. Strands β5 and β7 show a pattern of alternating hydrophobic and hydrophilic residues similar to that of the known membrane spanning regions of α-hemolysin [27] and anthrax toxin protective antigen [28]. Strand β6 is highly hydrophobic and the pattern is less obvious than in β5 and β7. The amino acid sequences of β5, β6 and β7 are not conserved but the alternating hydrophobic and hydrophilic pattern is highly conserved in all Cyt δ-endotoxins (Figure 1).

The crystal structures of the transmembrane domains of several proteins have been determined. In many of these structures, bands of aromatic residues (Tyr, Phe and Trp) lie at the hydrophobic/hydrophilic interface of the lipid bilayer and are considered to be a general characteristic of transmembrane proteins such as α-hemolysin [27], OmpA [29] and porin from Rhodobacter capsulatus[30]. Although the three β-barrel structures are diverse in terms of subunit stoichiometry and architecture, each shows the band of aromatic residues (Tyr, Phe and Trp) at the hydrophobic/hydrophilic interface of the lipid bilayer. These residues have intermediate polarity and are considered to interact favorably with the membrane at the hydrophobic/hydrophilic interface. In Cyt2Aa1, there are three tyrosine residues near the N-terminus of β5 and β7 (Y167, Y169 in β5 and Y213 in β7) which are conserved in all Cyt δ-endotoxins (Figure 1). After the Cyt toxin monomers have assembled and formed a pore in the membrane, these residues could form a band around the β-barrel at a point that the model proposes to be the point of entry of β5 and β7 (Figure 4). This band would be very similar in position to the band in α-hemolysin.

The recent solution of the crystal structure of a fungal pore forming toxin, volvatoxin A2 (VVA2), has shown it to be almost identical to Cyt2Aa1 [31]. Electron microscopy of this toxin suggested that it could form an oligomer with 18 subunits [31]. Investigation of truncated fragments of VVA2 suggested that the N-terminal fragment domain which consists mostly of α-helices is responsible for oligomerization and the C-terminal fragment domain consists of three β-sheets inserted into the membrane [32]. In Cyt2Aa1, sheets β5, β6 and β7 were reported to insert into the membrane and are protected from protease digestion [23]. Experiments using synthetic peptides corresponding to helices αA and αC [25] demonstrated that these helices could form oligomers. We therefore propose that Cyt2Aa1 forms a β-barrel transmembrane pore by inserting β5, β6 and β7 from 18 toxin molecules into the membrane with helices αA and αC playing a role in the oligomerization step.

We thank Dr Jade Li for valuable suggestions and discussion and Dr Paul Davis for assistance with graphics and multiple sequence alignment and Zeneca Agrochemicals, UK and the Institute of Molecular Biology and Genetics, Mahidol University, Thailand, for supplying mosquito eggs. Support to B.P. from the Thailand National Center for Genetic Engineering and Biotechnology is gratefully acknowledged.