3D structure of the entomophaga toxin complex and implications for insecticidal activity

Michael J. Landsberga,1, Sandra A. Jonesb, Rosalba Rothnagela, Jason N. Busbyc, Sean D. G. Marshallb, Robert M. Simpsond, J. Shaun Lottc, Ben Hankamera, and Mark R. H. Hurstb,1

aInstitute for Molecular Bioscience, University of Queensland, St. Lucia, Queensland 4072, Australia; bInnovative Farming Systems, AgResearch, Lincoln Research Centre, Christchurch 8140, New Zealand; cAgResearch Structural Biology Laboratory, School of Biological Sciences, University of Auckland, Auckland 1142, New Zealand; and dNew Zealand Institute for Plant and Food Research, Palmerston North 4474, New Zealand

Edited by Pamela J. Bjorkman, California Institute of Technology, Pasadena, CA, and approved October 19, 2011 (received for review July 20, 2011)

Toxin complex (Tc) proteins are a class of bacterial protein toxins other species (or from other Tc clusters) can combine to cause an that form large, multisubunit complexes. Comprising TcA, B, and effect with altered host specificity (12, 13), suggesting the TcA C components, they are of great interest because many exhibit component is a determinant of target host range. Further evi- potent insecticidal activity. Here we report the structure of a novel dence for this is provided by the observation that the TcA com- Tc, Yen-Tc, isolated from the bacterium Yersinia entomophaga ponents of an insecticidal X. nematophila Tc bind to solubilized MH96, which differs from the majority of bacterially derived Tcs insect midgut brush border membranes (14, 15). By comparison, in that it exhibits oral activity toward a broad range of insect pests, TcC-like proteins are thought to represent the main toxin com- including the diamondback moth (Plutella xylostella). We have de- ponents of the Tc family (7, 16). Two P. luminescens TcC-like termined the structure of the Yen-Tc using single particle electron toxins, TccC3 and TccC5, trigger ADP ribosylation and RhoA microscopy and studied its mechanism of toxicity by comparative GTPase activation in Galleria melonella haemocytes and cultured analyses of two variants of the complex exhibiting different toxi- HeLa cells, inducing redistribution and clustering of actin, inhi- city profiles. We show that the A subunits form the basis of a five- biting phagocytosis and leading to cell death (16). A model for fold symmetric assembly that differs substantially in structure and Tc-induced toxicity has thus been proposed in which the TcAcom- subunit arrangement from its most well characterized homologue, ponent selectively makes first contact with the target cell wall, the Xenorhabdus nematophila toxin XptA1. Histopathological and following which the TcC components are internalised to the quantitative dose response analyses identify the B and C subunits, cytosol (16). Further mechanistic insights have been hampered by which map to a single, surface-accessible region of the structure, the absence of structural data describing a fully assembled Tc. In- as the sole determinants of toxicity. Finally, we show that the deed, the only structure reported to date is that of the oligomeric assembled Yen-Tc has endochitinase activity and attribute this to TcA-like X. nematophila protein XptA1. Solved in the absence of putative chitinase subunits that decorate the surface of the TcA its TcB and TcC partners to moderate resolution using single par- scaffold, an observation that may explain the oral toxicity asso- ticle EM (14), the structure reveals a tetrameric cage-like assem- ciated with the complex. bly, proposed to encapsulate the TcB and TcC components. The genome of the recently identified bacterium Y. entomo- bacterial toxins ∣ insecticides ∣ single particle analysis ∣ protein assembly ∣ phaga MH96 contains an approximately 32-kb pathogenicity protein–protein interactions — island (designated PAIYe96) harboring five tc-like genes yenA1 (tcA-like), yenA2 (tcA-like), yenB (tcB-like), yenC1 (tcC-like), and he use of such as Bacillius thuringiensis (Bt) and their yenC2 (tcC-like)—as well as two putative chitinase-encoding Ttoxins in crop protection and pest eradication has proven genes (chi1, chi2) (17). Following the nomenclature of ffrench- widely successful, but the emergence of Bt-resistant insects (1, 2) Constant and Waterfield (6) the A component is composed of has motivated an urgent need to identify novel biopesticides two smaller protein subunits (A1 and A2) that combine with the (3, 4). Tc proteins comprise a class candidate class of molecules. B and two C subunits (C1 and C2) to form a single large complex First identified in the nematode-associated bacterium Photorhab- that, together with the coassociated chitinases (Chi1 and Chi2), is dus luminescens (5), Tc proteins characteristically belong to one referred to hereafter as Yen-Tc. The complex is secreted at tem- of three distinct classes with at least one subunit from each of the peratures equal to or below 25 °C and injection of semipurified three classes combining to form a complex with potent insectici- complex directly into the larval haemocoel of G. mellonella is as- dal activity. Subsequent to their initial discovery, the Tc classes sociated with potent toxicity (five day LD50 ¼ 35 ng per larva) have been redesignated A, B, and C, and the P. luminescens Tc (18). More importantly, the semipurified toxin exhibits broad proteins can thus be classified as either TcA-like (TcaA1, TcaB1, host range oral toxicity, causing rapid mortality in many insect TcbA, TccA and TcdA1), TcB-like (TcaC1 and TcdB1), or TcC-like pests belonging to the orders Coleoptera (e.g., beetles) and (TccC1 and TccC2) (6). Genes encoding Tc-like proteins have sub- Lepidoptera (e.g., moths) as well as the orthopteran Locusta sequently been identified in many insect-active bacteria including migratoria (18). Histological analyses show a general dissolution Serratia entomophila (tcA, sepA; tcB sepB; tcC sepC) (7) and the of the larval midgut within 48 h postingestion (18), consistent nematode-associated bacterium Xenorhabdus nematophila (tcA, xptA1, xptA2; tcB, xptC1; tcC, xptB1) (8), while the completion of bacterial genome sequencing projects such as Author contributions: M.J.L., J.S.L., B.H., and M.R.H. designed research; M.J.L., S.A.J., R.R., J.N.B., S.D.G.M., and R.M.S. performed research; M.J.L., S.A.J., J.N.B., S.D.G.M., J.S.L., B.H., C092 (9) and Pseudomonas syringae pv. tomato DC3000 (10) and M.R.H. analyzed data; and M.J.L. and M.R.H. wrote the paper. has revealed the presence of additional putative insecticidal tc The authors declare no conflict of interest. genes. This article is a PNAS Direct Submission. The general mode of action of Tc proteins has yet to be fully Data deposition: The Yen-Tc K∷9 density map has been deposited at the Electron elucidated. In some cases, expression of individual tcA genes is Microscopy Data Bank, http://emdatabank.org (accession no. EMD-1978). sufficient to cause toxicity (8, 11), but full toxicity generally re- 1To whom correspondence may be addressed. E-mail: [email protected] or quires all three Tc components, with the B and C components [email protected]. providing a potentiation of toxicity (12, 13). Interestingly, TcB This article contains supporting information online at www.pnas.org/lookup/suppl/ and TcC components coexpressed with TcA components from doi:10.1073/pnas.1111155108/-/DCSupplemental.

20544–20549 ∣ PNAS ∣ December 20, 2011 ∣ vol. 108 ∣ no. 51 www.pnas.org/cgi/doi/10.1073/pnas.1111155108 Downloaded by guest on October 1, 2021 with reported effects of other Tc toxins including Tca from P. luminescens (5). We set out to purify and structurally characterize Tc complexes isolated from Y. entomophaga MH96 as well as the derivative strain Y. entomophaga MH96∷9, which carries a mutation near the 5′ end of the yenB gene and secretes a complex lacking the B and C subunits (Yen-Tc K∷9). We observed substantial struc- tural differences between Yen-Tc K∷9 and its best characterized homologue, XptA1. We found that Yen-Tc has endochitinase activity, representing to our knowledge the earliest reported exam- ple of a Tc with associated chitinase activity. We also identified the likely location of the subunits responsible for this activity. Histopathological and comparative toxin dose response analyses directed against P. x y l o s t e l l a suggest that the Yen-Tc may be more potent than the most widely used Bt toxins and together with com- parative single particle EM analyses have additionally allowed us to structurally and functionally map the determinants of toxicity and address possible mechanisms of insecticidal activity. Results Isolation and Electron Microscopy of a Y. entomophaga Tc. Prelimin- ary transmission EM (TEM) revealed a strong preferential orien- tation of the mature Yen-Tc applied to glow-discharged carbon coated TEM grids. The related Yen-Tc K∷9 complex, composed of the A1 and A2 subunits as well as the much smaller Chi1 and ∷ Chi2 proteins, showed no preferential orientation and was there- Fig. 1. Single particle TEM analysis of Yen-Tc K 9. (A) Side views (outlined with trapezoids) dominate the field of view of a representative micrograph fore chosen for initial biophysical and structural analyses. Yen-Tc ∷ with limited end views also observed (circled). (Scale bar, 50 nm.) (B) Boxed K 9 offered the additional advantage that it was expected to be out particle projection images were subjected to maximum likelihood ana- structurally analogous to the previously characterized XptA1 lysis to obtain nonbiased averaged projection images. (Box size, 40 nm.) complex, providing a good basis for structural comparison. Approximately 0.5 mg of Yen-Tc K∷9 was routinely obtained The molecular weight estimated from the dynamic light scattering from a 100-mL culture of Y. entomophaga MH96∷9 grown in LB data was consistent with a pentameric assembly comprising all broth at 25 °C. Partially purified Yen-Tc K∷9 was then subjected four subunits in stoichiometric equivalents (A1 ¼ 130 kDa, A2 ¼ to Sephacryl S-400 gel filtration chromatography (Fig. S1A). Tox- 156 kDa, Chi1 ¼ 60 kDa, Chi2 ¼ 70 kDa; ½130 þ 156 þ 60 þ 70× in-containing fractions comprised four major protein components 5 ¼ 2080 kDa) and a k-means clustering analysis of rotationally (Fig. S1B), confirmed using a combination of mass spectrometry averaged spectra calculated from the raw particle images inde- and/or Edman sequencing as the Yen-Tc A1, A2, Chi1, and Chi2 pendently confirmed the inherent fivefold symmetry of Yen-Tc subunits. Minor bands that migrated between the A1 subunit and K∷9(Fig. S2). Chi2 had previously been identified as proteolytic fragments of the A2 subunit (18). The monodispersity and hydrodynamic ra- Three-Dimensional Structure of the Pentameric Yen-Tc K∷9 Heteroli- dius of purified Yen-Tc K∷9 was then measured using dynamic gomer. The reference-free, averaged 2D projection images of light scattering (Fig. S1C). Fitting the raw scattering data to a Yen-Tc K∷9 were used to generate a preliminary 3D reconstruc- one-component model yielded a hydrodynamic radius measure- tion of the complex. This initial model was then iteratively refined ment for Yen-Tc K∷9of15.4ð0.1Þ nm, which in turn yielded an until the reconstruction converged to a stable, final structure BIOCHEMISTRY estimate for its molecular mass of 2;000ð18Þ kDa. Residuals (Fig. S3). Based on the identified symmetry, C5 point symmetry were for the most part randomly distributed, and the calculated was imposed throughout structure refinement to enhance the sig- polydispersity of less than 1% confirmed that the complex was nal-to-noise ratio. A high degree of consistency between indivi- predominately monodisperse and that a single scattering compo- dual particle projection images (Fig. S3A) and the corresponding nent model adequately described the data. class averages (Fig. S3B) supported the self-consistency of the Purified Yen-Tc K∷9 was negatively stained with uranyl for- structure. Further structure validation was provided by the high mate and imaged by TEM (Fig. 1A). Typical micrographs were degree of consistency between reference-free (Fig. 1B) and dominated by lightbulb-shaped particles (outlined with trape- model-biased class averages (Fig. S3B). The final structure had zoids in Fig. 1A) with a height estimated from the raw micro- a resolution of 17 Å according to the 0.5 Fourier shell correlation graphs of approximately 31 nm and a width of approximately criterion (Fig. S3E). 22 nm. A lesser number of particle projection images having a The 3D structure of Yen-Tc K∷9 is shown in detail in Fig. 2. shape and dimensions consistent with an orthogonal view of this Based on its subunit composition, estimated molecular weight, same complex were observed (diameter approximately 22 nm, and symmetry, the complex is a 20-subunit heteroligomer that outlined with circles in Fig. 1A). From approximately 300 digital incorporates five copies each of A1, A2, Chi1, and Chi2. The micrographs, an initial dataset of 10,604 particle projection complex is 26 nm tall when viewed orthogonal to the axis of five- images was obtained and subjected to maximum likelihood ana- fold symmetry and is 22 nm in diameter at its widest point toward lysis (19). The resulting reference-free, averaged 2D projection the base of the complex (Fig. 2). The discrepancy in height images confirmed the overall shape of the nominal side view between the reference-free 2D averages and the 3D structure and, furthermore, were suggestive of a structure featuring an is explained by the absence of variable density at the base of the outer shell of electron density and an inner pore that traverses complex, which is not sufficiently reinforced by averaging to be the length of the complex (Fig. 1B). Unexpectedly, given the retained at the chosen visualization threshold. The structure nar- tetrameric structure previously reported for the homologous rows toward the base and to a greater extent toward the top of the XptA1 complex (14), nominal top views of Yen-Tc K∷9 exhibited complex. The tip of the complex has an external diameter of 8 nm a clear fivefold rotational symmetry (Fig. 1B, row 1, column 6). and a pore with an internal diameter of 2.5 nm.

Landsberg et al. PNAS ∣ December 20, 2011 ∣ vol. 108 ∣ no. 51 ∣ 20545 Downloaded by guest on October 1, 2021 Fig. 3. Mapping the chitinases within Yen-Tc K∷9. The X-ray crystallographic structure of Chi1 resampled at a resolution of 20 Å is shown fitted into a re- presentative lobe from each of the upper and lower rings of density (Fig. 2 “*”). Ninety percent of the Chi1 amino acid sequence, including the core chit- inase domain, is conserved with Chi2 (see Fig. S5). Based on the better fit of Chi1 within the upper lobe density it is thus predicted that five copies of Chi1 constitute the upper ring and that the lower ring comprises five copies of Chi2. The fitted models are colored to reflect the scheme in Fig. 5F. Fig. 2. Three-dimensional structure of Yen-Tc K∷9. Dimensions of the com- plex are annotated on side cross-sectional (upper left) and surface (upper was indicative of a good fit, whereas a visually more moderate right) views of the complex. Marked lobes (*) indicate putative locations fit was observed in the larger, lower ring density, although the of the Chi1 and Chi2 proteins. Nominal top (center left) and bottom views correlation score of 0.9765 was still indicative of a good fit. (lower left) are shown as viewed along the fivefold symmetry axis. Fenestra- For both lobes, the orientation with the highest correlation re- tions in the outer shell density of the cage (arrows) are visible in tilted views sulted in the active site being solvent exposed. As a result, and of the complex (center right, lower right). Lateral cross-sections viewed along having established that the Yen-Tc is an active chitinase, we there- the fivefold axis (far right) are displayed at four different depths (indicated fore concluded that the lobe-shaped densities decorating the ex- by solid black lines). Protein density is colored from dark blue (closest) through to yellow according to proximity to the central cavity. ternal surface of the Yen-Tc structure most likely correspond to extrinsically bound and active chitinase subunits. The amino acid A striking feature of the structure is a large internal cavity sequences of Chi1 and Chi2 predict a high degree of structure separated into two apparently distinct compartments (Fig. 2). conservation (Fig. S5), and so it seems reasonable to suggest A channel-like cavity occupies the center of the complex, having further that the lower lobe-shaped densities are contributed by a diameter approximately equal to the pore at the tip of the com- the slightly larger Chi2 protein (which contains an additional N-terminal domain of 79 amino acids, contributing an extra plex. A second compartment is formed between the density lining 8.7 kDa to the protein molecular weight) while the upper lobes the central channel and the outer shell of the complex. A number correspond to the Chi1 subunits. of fenestrations are observed in the outer shell density, suggesting parts of the cavity may be solvent accessible. The central channel, The B and/or C Components Are the Main Determinants of Toxicity. We on the other hand, is completely inaccessible from the base and is investigated the insecticidal activities of purified Yen-Tc K∷9 and separated from the convex pore within the tip of the complex by a – the native Yen-Tc. Bioassays directed against the diamond back constriction located 5 6 nm below the pore opening. moth P. xylostella revealed a potent toxicity associated with the wild-type complex only (Fig. 4A). The LD50 for Yen-Tc was Yen-Tc Is an Endochitinase. We investigated the chitinase activity determined to be 3.91ð0.03Þ ng per larva (equivalent to 1.59 of the purified Yen-Tc using fluorophore-labeled poly-N-acetyl- fmol per larva, assuming a molecular weight of 2,463 kDa for the glucosamine substrates. An endochitinase activity of 2;515 ∷ 222 : −1 : −1 native toxin). Conversely, Yen-Tc K 9 displayed no detectable mol fluorophore released s mol Yen-Tc was observed toxicity, even at levels as high as 225 ng (108.2 fmol) per larva. at pH 8, indicating a relatively high level of chitinolytic activity This strongly suggested that one or more of the Yen-Tc B and C (Fig. S4). As an exochitinase, Yen-Tc is approximately 10-fold less subunits are responsible for the toxicity of Yen-Tc. 370 49 : −1: −1 active ( mol s mol ). At lower pH a slightly enhanced Yen-Tc-treated P. xylostella were characterized histopathologi- activity was observed, but the pattern remained relatively con- cally by a disappearance of the peritrophic membrane and break- ¼ 3698 173 : −1: −1 stant (pH 6: endochitinase activity mol s mol ; down of the basement membrane within 24 h (Fig. 4B). Microvilli ¼ 320 25 : −1: −1 exochitinase activity mol s mol ). Considered to- typically found on the luminal face of the gut epithelial cells were gether with preliminary reports of chitinolytic activity associated also absent. The goblet cells, still lined with microvilli, appeared with partially purified Yen-Tc (18) this finding clearly establishes enlarged, whereas cell-like bodies (nuclei-containing vesicles) that Yen-Tc is an endochitinase and to our knowledge, represents sloughed off into the gut lumen. By 32 h posttreatment (Fig. 4C) the earliest reported example of a Tc with associated chitinase the basement membrane had largely disappeared with large num- activity. bers of cell-like bodies present in the gut lumen. A more profound We identified two symmetrically arranged rings of five globu- breakup of the columnar and goblet cells was observed along with lar, surface-accessible densities as likely locations of the Chi1 and an eventual, complete disintegration of the larval midgut. Chi2 subunits (Fig. 2, *). This hypothesis was evaluated by down- Having identified the B and/or C subunits as the main deter- sampling the recently solved structure of the Chi1 core chitinase minants of toxicity, we next conducted a comparative structural domain (PDB ID code 3OA5) (20) to a resolution of 20 Å and analysis of Yen-Tc and the nontoxic derivative, Yen-Tc K∷9. The docking the molecular envelope into one density from each of the presence of the additional B and C subunits in the native Tc iso- upper and lower rings that decorate the Yen-Tc K∷9 surface lated from wild-type cultures was confirmed by SDS/PAGE (Fig. 3). The correlation score of 0.9795 with the upper ring (Fig. 5A). TEM analysis revealed a complex estimated to be

20546 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1111155108 Landsberg et al. Downloaded by guest on October 1, 2021 Fig. 4. Toxicity of Yen-Tc and Yen-Tc K∷9. (A) Mortality curves for P. xylos- tella larvae dosed with either Yen-Tc or Yen-Tc K∷9 and observed for up to six days (see Table S1 for raw bioassay data). (B–D) Longitudinal midgut sections taken from third instar P. xylostella larvae after Yen-Tc treatment for 24 (B)or 32 h (C) show a progressive disintegration of the midgut, characterized by a rapid breakdown of the peritrophic membrane (PM) shortly after treatment, as well as breakdown of the basement membrane (BM), the disappearance of the columnar cells (CC), enlarged goblet cells (G), an absence of microvilli Fig. 5. Mapping the determinants of Yen-Tc toxicity. (A) Protein composition (MV) from the lumenal surface of the gut epithelium, and the appearance of the mature Yen-Tc (left lane) compared to nontoxic Yen-Tc K∷9 (right lane) of cell-like bodies (nuclei-containing vesicles; NV) in the gut lumen (L). (D) visualized by SDS/PAGE. The band designated C1/C2 (*) contains both C Representative mock-treated cell (V, vesicle body). (Scale bar, 50 μm.) (Full subunits as well as a comigrating proteolytic fragment of the A2 subunit that methods described in SI Experimental Procedures). persists in the K∷9 lane. (B) Yen-Tc applied to TEM grids exhibits a strongly preferred side view orientation (trapezoid outlined). (Scale bar, 50 nm.) Comparative structural analysis of averaged 2D projections of Yen-Tc (C) approximately 47 nm in length and similar in diameter to Yen-Tc ∷ ∷ and Yen-Tc K 9(D) reveals differences that are highlighted in the merged K 9 (approximately 22 nm; Fig. 5B). A flexible density located image (E). Densities exclusive to the toxic Yen-Tc complex (magenta) indicate at the narrow end of the complex immediately distinguished the the probable location of the B and C subunits. (Scale bar, 20 nm.) (F) Struc- native toxin from the K∷9 mutant, adding an extra 16 nm of tural model of the native Yen-Tc with the B/C subcomplex (red) rendered length to particles representing the native complex (compare based on the projection data shown in E. Orange (Chi1, 60 kDa) and yellow Fig. 1A with Fig. 5B). (Chi2, 70 kDa) densities indicate locations of the chitinase-like proteins. The strong preferential orientation of Yen-Tc and apparent conformational flexibility of the tip density complicated the to almost double the molecular weight of the complex (383 × 5 ¼ 1915 ∷ 9 ¼ 2080 determination of a full 3D reconstruction. Regardless, it was pos- kDa; molecular weight of Yen-Tc K kDa), a sible to assess in detail structural differences between Yen-Tc and difference that would almost certainly be accompanied by a de- Yen-Tc K∷9. A total of 10,224 particle projection images of tectable increase in the hydrodynamic radius of the complex. Con- Yen-Tc were boxed out from approximately 500 digital micro- sequently, it was concluded that the B, C1, and C2 subunits were

graphs and subjected to the same maximum likelihood analysis present in substoichiometric quantities. Semiquantitative densito- BIOCHEMISTRY as Yen-Tc K∷9 (Fig. 1B) generating a set of high signal-to-noise, metric analyses of Coomassie-stained protein bands separated by SDS/PAGE supported this hypothesis (Fig. S6 and Ta b l e S 3 )—the reference-free class averages for comparative analysis (Fig. 5C). ∷ It was immediately confirmed that the most significant difference intensity of the four protein subunits comprising Yen-Tc K 9was ∷ roughly equal, while the ratio of the B subunit to each of these was between Yen-Tc and Yen-Tc K 9 was the flexible density located at 0 9∶5 the end of Yen-Tc. This difference was emphasized visually when estimated at . and the combined intensity of both C subunits represented an approximate ratio of 2∶5 with each of the Yen-Tc Yen-Tc class averages were merged with correlated Yen-Tc K∷9 K∷9 subunits. class averages in the cyan-magenta color spectrum (Fig. 5 C–E). We modeled the density corresponding to a putative cocom- Little or no noticeable density exclusive to Yen-Tc K∷9was plex of B, C1, and C2 as an ellipsoid anchored within the pore observed (cyan), while the most significant density exclusive to located at the tip of Yen-Tc K∷9 (Fig. 5F), with dimensions based Yen-Tc (magenta) was localized to the tip of the complex. Notably, upon those derived from unbiased class averages (Fig. 5C). When no significant differences in structure were observed within the thresholded to the same extent as Yen-Tc K∷9, the molecular central cavity. This was unexpected given prior suggestions that weight associated with this density was estimated to be 369 kDa. bacterial Tc complexes utilize a cavity (such as that identified in ∷ This estimate compared well with the theoretical mass corre- Yen-Tc K 9; see Fig. 2) as a store for the toxin payload prior sponding to one copy of each of the B and C subunits (383 kDa) to reaching its target, in effect shielding the payload and/or main- and further supported the conclusion that these are present in taining its solubility (14). Our analysis clearly indicated that, for substoichiometric equivalents. It was therefore concluded that Yen-Tc at least, this is not the case and that the B and C subunits the overall subunit composition of the native toxin is a 5∶5∶1∶ are localized at the tip of the complex. 1∶1∶5∶5 heteroligomer of A1, A2, B, C1, C2, Chi1 and Chi2, The hydrodynamic radius of Yen-Tc could not be distinguished respectively. from that of Yen-Tc K∷9 by dynamic light scattering within the reliable limits of the technique. Given that the B and C subunits Discussion contribute a combined molecular weight of 383 kDa (theoretical Yen-Tc is the earliest reported example of an orally active Tc from molecular weight of B ¼ 167 kDa, C1 ¼ 107 kDa, C2 ¼ 109 kDa), the Yersinia genus. Here we have reported the 3D structure of stoichiometric equivalents of these proteins would be expected Yen-Tc K∷9, a mutant lacking the B and C subunits, as well as

Landsberg et al. PNAS ∣ December 20, 2011 ∣ vol. 108 ∣ no. 51 ∣ 20547 Downloaded by guest on October 1, 2021 a structural model of the native Yen-Tc toxin. The B and C sub- tiple sites. Studies on the effects of the B. thuringiensis Cry1A9(c) units added to the pentameric Yen-Tc K∷9 complex results in a protein have shown that approximately 2.42 × 1010 and 7.30 × 109 structure with a roughly elliptical density attached at the narrow molecules of toxin are needed to achieve lethality (LD50) toward end of the complex. Our findings counter the previous suggestion M. sexta and Heliothis virescens, respectively (26). Yen-Tc at the that the determinants of Tc toxicity are accommodated within a cellular level therefore appears to be more potent, an observation cavity created by assembly of a cage-like TcA oligomer (14). supported by the relative steepness of the LD50 curve (Fig. 4A). While both Yen-Tc and Yen-Tc K∷9 have internal cavities (Figs. 2 To our knowledge, the endochitinase activity (this study and and 5), toxic (Yen-Tc) and nontoxic (Yen-Tc K∷9) variants of the ref. 18) and incorporation of endochitinases seen in the structure complex studied here exhibit no discernible differences in density described here for Yen-Tc has not previously been reported in a within the cavity, and the main determinants of toxicity—in this Tc. The oral activity of Yen-Tc (this study and ref. 17) also distin- case, single copies of the B, C1, and C2 subunits—are localized guishes it from the majority of other bacterially derived Tcs,which within a density attached to the narrow end of Yen-Tc K∷9 cage. are for the most part released directly into the haemocoel of the Whether all, or a subset, of these proteins are required for full host by nematode-associated bacterial symbionts. It is plausible toxin activity remains to be determined. that the Chi1 and Chi2 proteins may play an essential role in Yen-Tc contains two TcA-like subunits with a combined mole- either degrading or binding to chitin within the peritrophic mem- cular mass approximately equivalent to that of the larger, single brane of the insect midgut, exposing the cell wall to the toxin and TcA components found in a subset of bacterial Tcs that includes enhancing the relative efficacy of the Yen-Tc through greater X. nematophila XptA1 (287 kDa), XptA2 (280 kDa) and accessibility to its target site (Fig. 6). Chitinases may also have S. entomophila SepA (262 kDa). The A1 and A2 subunits of Yen- a role in post cell death liquefaction of the insect cadaver similar Tcare homologous to the N- and C-terminal halves of these larger to chitinase-encoding, insect-active nucleopolyhedro viruses (27). TcAcomponents (Fig. S5). It was therefore surprising to find sub- Many aspects of the mechanism underlying Tc activity remain stantial structural differences between the fivefold symmetric to be defined, although studies of the P. luminescens toxins Yen-Tc K∷9 assembly and its best characterized structural homo- suggest a model for pathogenesis (16) that largely mirrors that logue, the tetrameric XptA1 complex (14). As well as differences seen in well characterized binary bacterial toxins (e.g., diphtheria, in symmetry and subunit number, Yen-Tc K∷9 is a more enclosed anthrax, and toxins) (28, 29). These binary systems are complex with only small fenestrations observed in the outer den- characteristically composed of an active toxin component (desig- sity (Fig. 2). The narrow end of the Yen-Tc K∷9 complex (i.e., the nated A) and a binding moiety (designated B). The B component, binding interface for B/C1/C2) is approximately the same size as composed of several homomerically arranged subunits, mediates that observed in the XptA1 structure (8 nm vs. 9 nm), but the specific, receptor-mediated binding to the cell surface. Following pore is much smaller (2.5 nm vs. 6.5 nm). endocytosis and internalization of the binary toxin, the A compo- We hypothesize that poorly resolved densities at the base of nent traverses a B-mediated pore from the endosome to the the Yen-Tc structure (Figs. 1B and 2 and Fig. S3) may indicate sites of surface-bound membrane-derived material, consistent with previous studies linking TcA subunits to host selectivity (12, 13, 15) and a role for TcA “scaffolds” in cell surface recogni- tion. Density resulting from heterogeneously associated material would be expected to reinforce poorly during averaging, and so the observation that these densities drop below the display threshold of the final structure is consistent with our hypothesis. The toxicity of Yen-Tc toward coleopterans (C. zealandica) (17) and lepidopterans (P. xylostella, G. mellonella) (this study; (18)) suggests the cellular target is conserved across species. The C-terminal domain of the C1 subunit shares high sequence similarity with the Rho-activating domain of cytotoxic necrotizing factor 1 (CNF1) (Fig. S5). CNF1 is impli- cated in deamination of Rho GTP-binding proteins, leading to rearrangement of the actin cytoskeleton (21–23). Similar func- tional characteristics have been previously attributed to the CNF1-like P. luminescens TccC5 protein (16). This observation is therefore consistent with the C1 subunit of Yen-Tc being re- sponsible (solely or in part) for Yen-Tc toxicity. Importantly C1 differs from CNF1 in that it contains no putative cell-targeting or translocation domain, consistent with other components of the Yen-Tc (e.g., the A subunits) fulfilling this role. The function of the B subunit is less obvious, but its sequence does contain a putative RCC1-like motif (Fig. S5), a domain associated with nuclear-localized chromatin binding proteins (24, 25), and studies utilizing a myc-tagged P. luminescens TcB construct have demon- strated that while the N-terminal domain of TcdB drives nuclear translocation, the C-terminal domain causes contraction of actin (13). It therefore appears that both the TcB-like and TcC-likesub- Fig. 6. Functional model for Yen-Tc toxicity. Following ingestion of the Tc, units harbor potential determinants of toxicity and internalization it is likely that the surface-bound chitinases bind to and/or degrade the of both may be required in order to induce a physiological effect. chitin-rich peritrophic membrane of the insect midgut. Cell surface recogni- Effects of Yen-Tc on the P. xylostella larval mid gut (Fig. 4 B–D) tion is likely facilitated by motifs within the A subunits prior to internaliza- tion. Similarities to the well characterized bacterial binary toxin systems mirror effects of the P. luminescens Tca toxin on M. sexta (5). The (e.g., anthrax, cholera, diphtheria) suggest a mechanism involving recep- LD50 of purified Yen-Tc toward P. xylostella, found to be 1.59 tor-mediated endocytosis followed by pore formation and translocation of 8 (0.01) fmol per larva, equates to approximately 9.56 × 10 mo- the B and/or C components into the cytosol (I), although alternative mechan- lecules per insect, suggesting that the toxin acts by targeting mul- isms (e.g., II, III) cannot be ruled out.

20548 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1111155108 Landsberg et al. Downloaded by guest on October 1, 2021 cytosol, inducing toxicity toward the target cell (28, 29). In ana- subunits while a more general mechanism of endosome dissolution logy to this model, TcA-like proteins are suggested to mimic the also cannot be ruled out (Fig. 6). Ultimately, further experiments function of the binary toxin B component while TcB-like and TcC- are required to establish the validity of these and/or other models. like subunits may together functionally substitute for the binary toxin A component. This model (Fig. 6) seems plausible given Experimental Procedures the implication of TcA-like proteins in cell surface recognition Purification of Y. entomophaga Toxin Complexes. Yen-Tc and Yen-Tc (14, 15) and that the pore-forming activity of the P. luminescens K∷9 were isolated from bacterial cultures of Y. entomophaga TcdA1 protein has been demonstrated (16). Many binary toxins MH96 and Y. entomophaga MH96∷9(Table S2) essentially target the actin cytoskeleton, a mode of action for Yen-Tc sug- according to established procedures (18). gested by the CNF1-like domain of the C1 subunit. Assuming Yen-Tc employs a receptor-mediated endocytic me- Chitinase Activity Assays. Chitinolytic activity was measured using chanism, what remains unclear is the mechanism by which the a fluorescent end point assay. Exochitinase and endochitinase toxin is released. In the case of the anthrax binary toxin, a pH- activities were assayed using 4-methylumbelliferyl β-D N,N′- dependent conformational change induces formation of the diacetylchitotrioside and 4-methylumbelliferyl β-D N,N′,N′′ tria- B-mediated pore through which the A toxins (approximately cetylchitotrioside (Sigma) as substrates, respectively. 90 kDa) are translocated into the cytosol. The Yen-Tc K∷9struc- ture bears some similarity to the pore-forming anthrax protective Electron Microscopy and Image Processing. Negative stain TEM was antigen (30) and as for the protective antigen, bioinformatic ana- β performed on a Tecnai F30 FEGTEM operating at a high tension lyses suggest a putative transmembrane -barrel structure in the voltage of 300 kV. A total of 10,224 and 10,604 particle projection N-terminal half of the Yen-Tc A2 subunit (Fig. S5). The disparity images of Yen-Tc and Yen-Tc K∷9, respectively, were extracted in calculated pI between the C1 (6.32), C2 (6.19), and other sub- from digital micrographs using the SWARM algorithm (31) units of Yen-Tc (A1 ¼ 4.72,A2 ¼ 4.78,B¼ 4.80,Chi1 ¼ 4.65, PS Chi2 ¼ 4.93) also appears consistent with a pH-driven mechanism as implemented in EMAN2 and subjected to 2D image processing of toxin release and/or conformational change. It is worth noting, using XMIPP (19) and EMAN (32). Three-dimensional structure however, that our preliminary studies have indicated that Yen-Tc is refinement was subsequently carried out in EMAN. structurally stable over a wide pH range (1–12) in vitro. The solid Full experimental procedures and associated references are base of the Yen-Tc structure also represents a substantial point of available as SI Experimental Procedures. difference when comparing Yen-Tc to both the anthrax protective antigen and the X. nematophila XptA1 toxin. XptA1 in particular ACKNOWLEDGMENTS. We thank Richard Townsend (AgResearch) for rearing has a much more obvious channel, suggesting an alternative and provision of the diamondback moth (P. xylostella) and Chikako van Koten (AgResearch) for assistance with statistical analysis. Histological staining mechanism may be employed by Yen-Tc. A mechanism involving was carried out by Gribbles Veterinary Services, Christchurch, New Zealand. direct penetration of the Yen-Tc tip across the endocytic or cell This study was supported by Contract C10X0804 from the New Zealand membrane might explain the proximal location of the B and C Foundation for Research, Science and Technology.

1. Tabashnik BE (1990) Implications of gene amplification for evolution and manage- 17. Hurst MR, Becher SA, Young SD, Nelson TL, Glare TR (2011) Yersinia entomophaga sp ment of insecticide resistance. J Econ Entomol 83:1170–1176. nov., isolated from the New Zealand grass grub . Int J Syst Evol 2. Tabashnik BE, Finson N, Johnson MW, Moar WJ (1993) Resistance to toxins from Microbiol 61:844–849. Bacillus thuringiensis subsp. kurstaki causes minimal cross-resistance to B. thuringiensis 18. Hurst MR, Jones SA, Binglin T, Harper LA, Glare TR (2011) The main virulence subsp. aizawai in the Diamondback Moth (Lepidoptera: Plutellidae). Appl Environ determinant of Yersinia entomophaga MH96 is a broad host range insect active, Toxin Microbiol 59:1332–1335. Complex. J Bacteriol 193:1966–1980. 3. Bravo A, Soberon M (2008) How to cope with insect resistance to Bt toxins? Trends 19. Scheres SH, et al. (2005) Maximum-likelihood multi-reference refinement for electron Biotechnol 26:573–579. microscopy images. J Mol Biol 348:139–149. 4. Moar WJ (2003) Breathing new life into insect-resistant plants. Nat Biotechnol 20. Busby JN, et al. (2011) Structural analysis of Chi1 Chitinase from Yen-Tc: The multisu- – 21:1152 1154. bunit insecticidal ABC toxin complex of Yersinia entomophaga. J Mol Biol, 10.1016/j 5. Bowen D, et al. (1998) Insecticidal toxins from the bacterium Photorhabdus lumines- .jmb.2011.11.018. – cens. Science 280:2129 2132. 21. Buetow L, Flatau G, Chiu K, Boquet P, Ghosh P (2001) Structure of the Rho-activating BIOCHEMISTRY 6. ffrench-Constant R, Waterfield N (2005) An ABC guide to the bacterial toxin domain of Escherichia coli cytotoxic necrotizing factor 1. Nat Struct Biol 8:584–588. – complexes. Adv Appl Microbiol 58:169 183. 22. Caprioli A, Falbo V, Roda LG, Ruggeri FM, Zona C (1983) Partial purification and 7. Hurst MR, Glare TR, Jackson TA, Ronson CW (2000) Plasmid-located pathogenicity characterization of an Escherichia coli toxic factor that induces morphological cell determinants of Serratia entomophila, the causal agent of amber disease of grass alterations. Infect Immun 39:1300–1306. grub, show similarity to the insecticidal toxins of Photorhabdus luminescens. J Bacter- 23. Schmidt G, Selzer J, Lerm M, Aktories K (1998) The Rho-deamidating cytotoxic necro- iol 182:5127–5138. tizing factor 1 from Escherichia coli possesses transglutaminase activity. Cysteine 866 8. Morgan JA, Sergeant M, Ellis D, Ousley M, Jarrett P (2001) Sequence analysis of insec- and histidine 881 are essential for enzyme activity. J Biol Chem 273:13669–13674. ticidal genes from Xenorhabdus nematophilus PMFI296. Appl Environ Microbiol 24. Dasso M (1993) RCC1 in the cell cycle: The regulator of chromosome condensation 67:2062–2069. takes on new roles. Trends Biochem Sci 18:96–101. 9. Parkhill J, et al. (2001) Genome sequence of Yersinia pestis, the causative agent of 25. Moore JD (2001) The Ran-GTPase and cell-cycle control. Bioessays 23:77–85. . Nature 413:523–527. 26. Höfte H, Whiteley HR (1989) Insecticidal crystal proteins of Bacillus thuringiensis. 10. Buell CR, et al. (2003) The complete genome sequence of the Arabidopsis and Microbiol Rev 53:242–255. tomato pathogen Pseudomonas syringae pv. tomato DC3000. Proc Natl Acad Sci USA 100(18):10181–10186. 27. Hawtin RE, et al. (1997) Liquefaction ofAutographa californica nucleopolyhedrovirus- 11. Liu D, et al. (2003) Insect resistance conferred by 283-kDa Photorhabdus luminescens infected insects is dependent on the integrity of virus-encoded chitinase and cathepsin – protein TcdA in Arabidopsis thaliana. Nat Biotechnol 21:1222–1228. genes. Virology 238:243 253. 12. Sergeant M, Jarrett P, Ousley M, Morgan JA (2003) Interactions of insecticidal toxin 28. Barth H, Aktories K, Popoff MR, Stiles BG (2004) Binary bacterial toxins: biochemistry, gene products from Xenorhabdus nematophilus PMFI296. Appl Environ Microbiol biology, and applications of common Clostridium and Bacillus proteins. Microbiol Mol – 69:3344–3349. Biol Rev 68:373 402. 13. Waterfield N, Hares M, Yang G, Dowling A, ffrench-Constant R (2005) Potentiation 29. Zaretzky FR, Gray MC, Hewlett EL (2003) Direct penetration of bacterial toxins across and cellular phenotypes of the insecticidal Toxin complexes of Photorhabdus bacteria. the plasma membrane. Bacterial Protein Toxins, eds DL Burns, JT Barbieri, BH Iglewski, Cell Microbiol 7:373–382. and R Rappuoli (ASM Press, Washington), pp 149–172. 14. Lee SC, et al. (2007) Structural characterisation of the insecticidal toxin XptA1, reveals 30. Katayama H, et al. (2010) Three-dimensional structure of the anthrax toxin pore in- a 1.15 MDa tetramer with a cage-like structure. J Mol Biol 366:1558–1568. serted into lipid nanodiscs and lipid vesicles. Proc Natl Acad Sci USA 107:3453–3457. 15. Sheets JJ, et al. (2011) Insecticidal toxin complex proteins from Xenorhabdus nema- 31. Woolford D, et al. (2007) SwarmPS: Rapid, semi-automated single particle selection tophilus: Structure and pore formation. J Biol Chem 26:22742–22749. software. J Struct Biol 157:174–188. 16. Lang AE, et al. (2010) Photorhabdus luminescens toxins ADP-ribosylate actin and RhoA 32. Ludtke SJ, Baldwin PR, Chiu W (1999) EMAN: Semiautomated software for high- to force actin clustering. Science 327:1139–1142. resolution single-particle reconstructions. J Struct Biol 128:82–97.

Landsberg et al. PNAS ∣ December 20, 2011 ∣ vol. 108 ∣ no. 51 ∣ 20549 Downloaded by guest on October 1, 2021