Nontoxic crystal protein from Bacillus thuringiensis demonstrates a remarkable structural similarity to β‐pore‐forming toxins

A ubiquitous Gram-positive bacterium, Bacillus thuringiensis (Bt), produces parasporal crystalline inclusions, also known as crystals, during sporulation. The crystals contain one or more proteins that are often toxic to certain kinds of insects.1, 2 Therefore, this bacterium has attracted industrial interest as a natural biological insecticide. These toxins historically have been divided into two categories: (1) insecticidal Cry proteins including a number of phylogenetically heterogeneous protein groups, and (2) cytolytic Cyt proteins with broad specificity. On the other hand, it has also been noted that noninsecticidal Bt strains are more widely distributed in various environments than insecticidal ones.3, 4 While the noninsecticidal Bt strains have long aroused little attention, recent screenings of such strains for cytocidal activity have identified several strains that produce novel crystal proteins with preferential cytotoxicity against particular human carcinoma cells.5 This group of crystal proteins has been named parasporins.6 Parasporal inclusions obtained from Bt strain A1470 (formerly called 89-T-34-22) have been shown to contain proteins that are neither insecticidal nor hemolytic but highly cytocidal against human leukemic T cells after proteolysis.7, 8 In proteinase-K digests of the inclusions, a 27-kDa protein has been identified as toxic.9 The 27-kDa protein is proteolytically produced from a 34-kDa precursor and is specifically cytotoxic against MOLT-4 cells. This protein is the fourth parasporin (parasporin-4) idendified thus far10 and is registered as Cry45Aa1 in the comprehensive Bt toxin list (http://www.lifesci.sussex.ac.uk/Home/Neil_Crickmore/Bt/).11 The proteinase-K digests also contain a 26-kDa protein for which no toxicity has been detected but which shares a considerable sequence identity with parasporin-4. Like parasporin-4, this nontoxic 26-kDa protein is produced from a 32-kDa precursor by proteolysis. Marked contrasts in observed toxicity, along with probable structural similarities presumed from the sequence homology and the proteinase-processing patterns, suggest that a detailed comparison between the 26-kDa protein and parasporin-4 would be of great value towards the identification of the structural elements responsible for receptor recognition and toxicity. In this article, we present the X-ray crystal structure of the 26-kDa protein determined at 2.07 Å resolution. This structure closely resembles that of ϵ-toxin from Clostridium perfringens, a very potent β-pore-forming toxin (β-PFT). The source of the 26-kDa protein was a soil isolate, B. thuringiensis strain A1470, which belongs to serovar shandongiensis.7 The bacteria were grown at 28°C for 7 days on 2.0% (w/v) nutrient agar (pH 7.6) containing 1% (w/v) meat extract, 1% (w/v) polypeptone, and 0.2% (w/v) NaCl. Sporulated cells were harvested, washed three times in 1M NaCl, and disrupted in distilled water. The parasporal inclusions were isolated by the biphasic separation method12 and solubilized in 50 mM Na2CO3 (pH 10.8) at 37°C for 60 min in the presence of 2 mM dithiothreitol (DTT) and 1 mM ethylenediaminetetraacetic acid (EDTA). After being clarified by centrifugation, the solution was titrated to pH 10 with 1 M NaOH and treated with proteinase K (final concentration 100 μg/mL) at 37°C for 90 min. The proteolysis was stopped by the addition of phenylmethylsulfonyl fluoride (final concentration 1 mM). The proteinase-treated proteins were applied to a DEAE-Sepharose FF column (Amersham Biosciences) equilibrated with solubilization buffer and eluted with an increasing gradient of NaCl concentrations from 0 to 1M in the same buffer. Fractions eluted with approximately 150 mM NaCl were pooled and dialyzed against 20 mM Na2CO3–NaHCO3 buffer (pH 10.5) containing 2M NaCl. The dialyzed sample was loaded onto a Toyopearl Butyl-650 column (Tosoh) equilibrated with the same buffer. The 26-kDa protein was recovered in the flow-through fraction, which was subsequently concentrated by ultrafiltration for crystallization experiments. Prior to the crystallization experiments, the proteinase-treated protein was dialyzed overnight against 10 mM 3-(cyclohexylamino)propanesulfonic acid (CAPS), pH 12.0, 0.5 mM EDTA, 2 mM β-mercaptoethanol, and then was filtered through a 0.22 μm Ultrafree-MC filter (Millipore). Crystals suitable for data collection were obtained by the hanging-drop vapor diffusion method with a reservoir solution of 10% (v/v) 2-methyl-2,4-pentanediol (MPD), 50 to 100 mM ammonium sulfate, 100 mM sodium citrate, pH 5.8, 1 mM EDTA, and 1 mM DTT. Two microliters of the protein solution (15.0 mg/mL) were mixed with an equal volume of the reservoir solution and equilibrated against 100 μL of the reservoir solution. Wedge-shaped crystals or crystals with an orange section appeared within a week and grew slowly, reaching 0.2 to 0.3 mm in the maximum dimension in 2 to 3 months at 298 K. To prepare isomorphous mercury-derivative crystals, native crystals were soaked in the harvest solution [50 mM Li2SO4, 10% (v/v) MPD, 100 mM 2-(N-morpholino)-ethanesulfonic acid (MES), pH 5.8] containing 5 mM p-chloromercuriphenylsulfonate (PCMPS) for 3 h. MALDI-TOF mass spectrometry was used to verify the crystallized protein molecule. The crystals obtained were crashed and dissolved in 50 mM CAPS, pH 12.0. The solution was mixed with a matrix solution of 10 mg/mL sinapinic acid in 50% acetonitrile/50% water with 0.1% trifluoroacetic acid and examined on an AXIMA-CFR plus mass spectrometer (Shimadzu Biotech). Native and derivative crystals were flash-frozen in a nitrogen gas stream at 100 K after being dipped briefly in the harvest solution supplemented with glycerol [15% (w/v)]. X-ray data were collected from the frozen crystal on beamline NW12 at PF-AR (KEK, Tsukuba, Japan) using a Quantum-210 CCD detector (Area Detector Systems Corp.). HKL200013 and TRUNCATE14, 15 were used for X-ray data processing. Data collection statistics calculated by SCALA15, 16 are summarized in Table I. The experimental phases for the electron density map was obtained by the multiple-wavelength anomalous dispersion (MAD) method using four-wavelength data sets from a mercury-derivative crystal (Table I). Three mercury atoms were identified from anomalous difference Patterson maps and their positions were obtained by the direct method using SHELXS.17 Initial phase calculation at 2.3 Å and subsequent density modification by solvent flattening were performed with autoSHARP.18 The initial model was built by automated chain tracing with ARP/wARP19 and manual editing with Turbo-FRODO.20 The model was refined against the native data set by iterative use of the CNS program21 and manual model fitting with Turbo-FRODO. The final model consists of 498 amino acids and 210 water molecules, and has a crystallographic R-factor of 23.0% and an Rfree of 26.6%. The quality of the structure was checked with PROCHECK.22 The refined coordinates and structure factors have been deposited in the Protein Data Bank with accession code 2D42. The 26-kDa protein crystallized in the trigonal space group P3112 with cell dimensions of a = b = 95.38 Å, c = 130.73 Å, and γ = 120°. The mass spectra of the dissolved crystals showed a single major peak at m/z = 27541.02, which agrees with a calculated mass of 27542.7 Da for Ala2–Ala250 of the protein. Consequently, the proteolytic procedure employed here removed C-terminal 28 residues from the intact protein lacking the N-terminal methionine. Using the molecular mass thus obtained, the Matthews coefficient (VM) of the crystal was 3.1 Å3/Da, corresponding to two molecules in the asymmetric unit and a solvent content of 60.3%.23 The protein structure was determined by the MAD method utilizing a mercury atom as an anomalous scatterer and was subsequently refined at 2.07 Å using a native data set. For each of the two molecules in the asymmetric unit, the final model contains all 249 residues expected from the mass spectra. The 26-kDa protein is an unusually elongated molecule with overall dimensions of 118 Å × 32 Å × 30 Å [Fig. 1(a)]. The structure is dominated by β-strands (67%, S1–12), most of which are remarkably extensive, running all or most of the longer axis of the molecule. The longest strand (S9), comprising 36 residues, runs the whole length of the molecule; three others (S5, S8, S12) span two-thirds of the molecule's length. The molecule can be divided into three domains: domain 1 (Ala2–Pro35, Gly176–Pro210), domain 2 (Ile36–Pro53, Ala81–Asp138, Glu160–Ser175, Gly211–Gln236), and domain 3 (Val54–Asn80, Phe139–Leu159, Thr237–Ala250). Domain 1 is composed of a short N-terminal β-strand (S1), an α/β structure (H1 and S2), a β-hairpin (S9 and S10), and an α-helix (H2). α-Helices are found only in this domain. The two α-helices, H1 and H2, are close together and short, occupying only 8.9% of the molecule. Domains 2 and 3 are both β-sandwiches: the former is made of a two-stranded β-hairpin (S6 and S7) and a curled anti-parallel five-stranded β-sheet including S3, S9, S12, S5, and S8; and the latter is of antiparallel three-stranded and two-stranded β-sheets (S4, S9, and S12; S5 and S8). Structure of the 26-kDa protein. (a) Ribbon representation of the 26-kDa protein structure with secondary structures labeled: H1–2 for α-helices, S1–12 for β-strands. Labeled arrows on the left of the model indicate approximate regions of domains 1 to 3. The polypeptide chain is colored in a rainbow ramp from blue at the N-terminus to red at the C-terminus. (b) Ribbon representation of two molecules in the asymmetric unit. They are labeled A and B and colored yellow and green, respectively. (c) Superposition of the Cα traces of the two molecules. The models are colored as in (b). Arrows indicate hinge-bending motions observed. (d) Comparison of B-factor distributions in the two molecules. Ribbon models of the molecules are color-coded according to averaged main-chain B-factors in a gradient from white for 21.1 Å2 to red for 50.6 Å2. Figures were generated using Molscript33 and Raster 3D.34 Two molecules in the asymmetric unit contact each other at the side of domain 3, forming a V shape [Fig. 1(b)]. The shape may imply the presence of local twofold symmetry, but it was not observed in self-rotation functions. This is probably due to the dislocation of the molecules along the quasi-twofold axis. Association of the two molecules is mediated mainly by hydrogen bonds between main-chain carbonyl and amide groups in two S5 β-strands running parallel to each other in the domain 3 region. Thus, a juxtaposed pair of the two-stranded β-sheets (S5, S8) constitutes an intermolecular four-stranded β-sheet in this region, staggering the two molecules about 3 Å apart. Superimposition of the two molecules reveals considerable deviations in domains 1 and 3, while a good fit is obtained for domain 2 [Fig. 1(c)]. The β-hairpin in domain 1 and the whole structure of domain 3 of one molecule (B in the figure) show hinge-like bending motions of about 10° each relative to their counterparts in the other molecule (A). The two deviating parts are bent at the domain boundaries, where the β-strands change their organization and seem to be less stiff. These motions result in an RMSD of 1.47 Å between the two molecules for 80% of the Cα atoms, while an RMSD of 0.67 Å is obtained for all the Cα atoms except for those in the bending parts. The distributions of peptide chain mobility also reveal conspicuous differences between the two molecules [Fig. 1(d)]. In molecule B, the more bent molecule, the averaged B-factors of the β-hairpin (S6, S7) and its flanking loops are considerably high. Parts of β-strands adjacent to the β-hairpin also have high B-factors in molecule B. These local anomalies are not observed in molecule A. This difference is not ascribed directly to differences in intermolecular contact between the two molecules because domain 2 is scarcely involved in such interaction. Instead, increased chain mobility in domain 2 may be a result of mechanical stress conducted through adjacent domains, which probably also causes the hingelike bending of the molecule. The structure of the 26-kDa protein is distinct from the available structures of Bt crystal proteins, including four Cry proteins (Cry1Aa,24 Cry2Aa,25 Cry3Aa,26 Cry3Bb127) and a Cyt protein (CytB).28 Unexpectedly, the structure of the 26-kDa protein is remarkably similar to the recently reported structure of ϵ-toxin from C. perfringens.29 In spite of the rather low sequence identity (22.7%), secondary structures are organized virtually in the same way in the two proteins. Taking molecule B of the 26-kDa protein dimer for comparison, the Cα traces of the two structures can be superimposed very well, with an RMSD of 1.8 Å for 137 Cα atoms [Fig. 2(a)]. ϵ-Toxin is very potent and is responsible for fatal enterotoxemia in economically important livestock. It is considered as a β-PFT owing to its functional and structural similarities to aerolysin, a β-PFT from the Gram-negative bacterium Aeromonas hydorphila.30 Therefore, it seems fair to assume that the 26-kDa protein is an inactive variant of a certain β-PFT. Comparison of the 26-kDa protein and ϵ-toxin. (a) Superposition of the Cα traces of the 26-kDa protein (molecule B in the asymmetric unit) and ϵ-toxin (PDB ID code: 1UYJ). The 26-kDa protein is in green and ϵ-toxin in blue (the inactivation peptide region in light blue). (b, c) The CPK models of the 26-kDa protein (b) and ϵ-toxin (c) are each viewed from two directions separated by 160°. Atoms of surface residues are colored as follows: aspartic acid, glutamic acid: red; histidine, lysine, arginine: blue; serine, threonine: green; alanine, valine, prorine, isoleucine, leucine, cystine, methionine: yellow; phenylalanine, tyrosine, tryptophan: orange; others: gray. Arrows in (c) indicate patches of hydrophobic residues and of hydroxylated residues on the surface of the ϵ-toxin molecule. In contrast to the general structural similarities described above, there are remarkable differences in the surface distributions of amino acids between the 26-kDa protein and ϵ-toxin [Fig. 2(b,c)]. ϵ-Toxin has a large distinct patch of hydroxylated amino acids, serine and threonine, on the opposite side of the β-hairpin in domain 2. In addition, the removal of the C-terminal inactivation peptide from the protoxin model exposes a small hydrophobic patch on the surface of domain 3 [Fig. 2(c)]. Both of these characteristic patches are supposed to be responsible for the oligomerization of ϵ-toxin within the membrane, which is necessary for pore formation.29 The 26-kDa protein lacks both features. This could be a reason for the loss of toxicity in the protein. A substantial sequence identity (38.0%) between the 26-kDa protein and a cancer-cell–targeting toxin, parasporin-4, suggests the structural similarity of the two proteins. Hence, parasporin-4 probably has a structure similar to that of ϵ-toxin. The rapid cytopathic effects of parasporin-4 on MOLT-4 cells including nuclear decondensation and cell-ballooning9 are also similar to those observed for ϵ-toxin against the Madin–Darby canine kidney cells.31 Thus, parasporin-4 and ϵ-toxin would share a structure and a mode of action as active β-PFTs. Appreciable sequence identity to the 26-kDa protein is observed for a few insecticidal Cry proteins, including Cry23 (26.5% identity), Cry15 (23.5%), and Cry33 (22.5%). These proteins are related to mosquitocidal toxins Mtx2 and Mtx3 from B. sphaericus,2 which have homology to several β-PFTs including ϵ-toxin and aerolysin. Consequently, the Mtx2/3-related Cry proteins presumably have an ϵ-toxin-type structure in general. (The crystal structure of a binary toxin complex of Cry23 and Cry37 has been solved32 but, unfortunately, its atomic coordinates were not public during the preparation of this manuscript. The available figure of the Cry23 structure in the complex exhibits a close similarity to the 26-kDa protein.) We thank Drs. Naohiro Matsugaki, Noriyuki Igarashi, Mamoru Suzuki, and Soichi Wakatsuki for their assistance with data collection at PF-AR (KEK, Tsukuba, Japan). This work was supported by a grant to KH from the New Energy and Industrial Technology Development Organization (NEDO) and by a grant to EM from the Special Coordination Funds for the Promotion of Science and Technology (Leading Research Utilizing Potential of Regional Science and Technology) of the Ministry of Education, Culture Sports, Science and Technology of the Japanese Government.