The ABC toxin complex from Yersinia entomophaga can package three different cytotoxic components expressed from distinct genetic loci in an unfolded state: the structures of both shell and cargo

A cytotoxin encoded by an ‘orphan’ genetic locus in the insect pathogen Yersinia entomophaga is shown to be a functional part of an ABC toxin complex, and the structures of both the RHS-repeat-containing ‘shell’ and the free toxin alone are determined by X-ray crystallography. The structure of the toxin indicates that it most likely functions by directly modifying actin in the target cell.


Introduction
The ABC toxins are a class of multi-subunit toxin complexes present in several entomopathogenic bacteria that have three main protein components: TcA, TcB and TcC (ffrench-Constant & Waterfield, 2006).The structures of several ABC toxin complexes have been determined, and these structures have enabled the mechanism of action of this class of toxin to be elucidated (Landsberg et al., 2011;Busby et al., 2013;Gatsogiannis et al., 2013Gatsogiannis et al., , 2016Gatsogiannis et al., , 2018;;Meusch et al., 2014;Piper et al., 2019;Leidreiter et al., 2019;Roderer et al., 2019a).ABC toxins exist in two sub-types: type I, where five copies of a single TcA protein form a pentameric assembly that acts as a nano-injection device to penetrate the cell membrane of target cells and deliver the toxic payload (Gatsogiannis et al., 2013); and type II, where the TcA protein is split into two halves, transcribed from separate genes, and in some cases also incorporate other accessory proteins (Landsberg et al., 2011;Piper et al., 2019).In both cases, the pentamer is topped with a single copy of each of the TcB and TcC proteins that come together to form a hollow shell, encapsulating the cytotoxic C-terminal domain of TcC (TcC CTD ) as 'cargo' inside the shell (Busby et al., 2013;Meusch et al., 2014).
The hollow shell formed by TcB and TcC consists primarily of repeated copies of the rearrangement hot-spot (RHS) repeat sequence, also described as 'YD' repeats because they contain a highly conserved tyrosine-aspartate dipeptide motif.RHS repeats are found in many different bacterial species, both in ABC toxin complexes and in other genes involved in pathogenesis or inter-strain competition (Poole et al., 2011;Koskiniemi et al., 2013;Jure ˙nas et al., 2021;Gu ¨nther et al., 2022), and are also found in the teneurins, which play a key role in inter-cellular communication in eukaryotes (Ferralli et al., 2018;Jackson et al., 2018Jackson et al., , 2019;;Li et al., 2018;Arac ¸& Li, 2019;del Toro et al., 2020;Meijer et al., 2022).
Bacterial RHS-repeat-containing proteins belong to a broad family of proteins known as polymorphic toxin systems (Zhang et al., 2012;Jamet & Nassif, 2015).Polymorphic toxins are a class of bacterial toxin delivery system with a characteristic genomic layout, usually consisting of a signal peptide or secretory domain, multiple copies of a repeating unit, a releasing peptidase, and a highly variable C-terminal toxin domain.RHS cassettes can be swapped with other orphaned cassettes in the genome by recombination, allowing for a variety of different toxin cargoes to share a common delivery vehicle.Due to the high degree of variation in their sequences, the toxin cargoes have also been described as 'hypervariable regions', or HVRs (Roderer & Raunser, 2019).
The bacterium Yersinia entomophaga is pathogenic to a variety of insect species (Hurst et al., 2011a).One of the primary determinants of its pathogenicity is a type II ABC toxin complex, Yen-Tc, encoded by two tcA genes (each encoding half of the TcA component); a single tcB gene; two distinct tcC genes, each encoding different C-terminal toxin domains; and two accessory chitinases (Hurst et al., 2011b).On the basis of their amino acid sequences, the two TcC toxins are predicted to be homologous to cytotoxic necrotizing factors (that constitutively activate Rho GTPase) and nucleotide deaminases, respectively (Busby et al., 2013).
The genome sequence of Y. entomophaga (Hurst et al., 2016) contains several genes encoding RHS-repeat-containing proteins at different genomic locations.The rhs2 region, located �900 000 base-pairs from the Yen-Tc-encoding gene cluster in the genome, contains a gene (PL78_18780) that shows remarkable sequence similarity to the other tcC-type genes.An amino acid sequence alignment (Sievers et al., 2011) of YenC1, YenC2 and PL78_18780 shows a high level of similarity within the N-terminal domain (�60% pairwise sequence identity over �680 residues), but the C-terminal domains show no detectable sequence similarity (see the supporting information), consistent with the presence of a hypervariable region.'Orphaned toxin cassettes' have been described for other polymorphic toxin systems (Koskiniemi et al., 2014) and so we speculated that PL78_18780, which has been re-annotated as yenC3 (Paulson et al., 2021), may indeed encode a third TcC protein, carrying a different C-terminal cargo, that may in turn form a complex with YenB in a similar manner to YenC1 and YenC2 (Fig. S1 of the supporting information).The rhs2 locus also encodes other potential virulence effector proteins, including a putative YopT-type cysteine protease and a putative vertebrate C-lysozyme inhibitor (Hurst et al., 2016), suggesting that this region of the genome is a pathogenicity island required in the infection process.Congruent with this idea, the YenC3 transcript is upregulated during the early infection stage of infection of the wax moth Galleria mellonella by Y. entomophaga (Paulson et al., 2021).
We therefore set out to characterize the predicted RHSrepeat-containing protein YenC3 to determine if it is an additional TcC protein associated with Yen-Tc and whether it functions as a toxin.

Co-expression and purification of YenB and YenC3
To test whether YenC3 could form a complex with YenB, the yenB and yenC3 gene products were co-expressed in E. coli from a single expression plasmid.YenB was expressed with an N-terminal His 6 -tag, whereas YenC3 was left untagged.The two proteins formed a complex that bound to immobilized metal affinity chromatography (IMAC) resin, and they remained associated with each other throughout further purification by size-exclusion chromatography (SEC).As with other TcC proteins (Busby et al., 2013), YenC3 underwent self-cleavage at the interface between its N-and Cterminal domains, with both domains remaining associated with the complex (Fig. S2).The observations that YenC3 is able to form a stable complex with YenB, and that its Cterminal sequence is cleaved but remains associated with the complex are entirely consistent with YenC3 functioning as a bona fide TcC protein.

The YenB/YenC3 complex structure
The complex consisting of YenB and both the N-and Cterminal domains of YenC3 (YenC3 NTD and YenC3 CTD ) was crystallized, X-ray diffraction data were collected, and the structure was determined by molecular replacement using the structure of YenB/YenC2 NTD (PDB entry 4igl; Busby et al., 2013) as the search model (Table 1, Fig. 1).Overall, the structure is extremely similar to the structure of YenB/ YenC2 NTD which we had determined previously (Busby et al., 2013(Busby et al., , 2016).An RMSD of 0.77 A ˚over 2092 superimposed residues was calculated (Krissinel, 2012), with the major point of difference being that the YenC2 CTD had to be removed from the YenB/YenC2 complex before it could be crystallized (Busby et al., 2013).In contrast, the complete YenB/YenC3 sequence was crystallized, with YenC3 CTD still present in the complex.Despite this, no continuous electron density could be observed for YenC3 CTD .We therefore concluded that the Cterminal domain is most likely encapsulated in an unfolded or disordered state, explaining why it is not visualized in the spatially averaged structure produced by X-ray crystallography (Busby et al., 2013;Meusch et al., 2014;Roderer et al., 2019b).Close inspection of the electron density delivered further support for this conclusion, with tubes and patches of electron density packing against several regions of the interior surface of the shell observed, which could not be accounted for by the protein residues encoding for the shell itself (Fig. 2).
In our previous structure of the YenB/YenC2 CTD shell, we observed multiple hydrophobic patches, derived from copies of the RHS repeat, distributed throughout the mostly positively charged interior shell surface.We proposed that this interior surface could stabilize an unfolded, encapsulated cargo protein in a chaperone-like fashion (Busby et al., 2013;Meusch et al., 2014).Equivalent hydrophobic patches are also a feature of the interior YenB/YenC3 shell surface, and are usually found in close proximity to the putative cargo protein density described above, supporting this hypothesis that they stabilize the unfolded cargo.For example, an area of flat electron density was seen �3.9A ˚from the plane of a phenylalanine residue (Y1220), possibly representing aromatic residues of the cargo protein forming �-stacking interactions [Fig.2(a)].Multiple examples of tubes of electron density packing into hydrophobic clefts on the inner surface of the shell were also visible in the structure reported here [Figs. 2(b) and 2(c)].These observations are consistent with previous reports that the HVR of the TccC3 toxin from Photorhabdus luminescens is unfolded within its cognate shell (Gatsogiannis et al., 2018;Belyy et al., 2022).
To independently confirm that YenC3 CTD was encapsulated within the YenB/YenC3 NTD complex, we subjected the YenB/ YenC3 complex to small-angle X-ray scattering analysis.The protein sample was purified to homogeneity by SEC (Fig. S2).A concentration series was analysed using SEC with multiangle light scattering (SEC-MALS) to confirm that the complex formed a 1:1:1 ratio of the expected size, with no aggregation or concentration-dependent oligomerization (Fig. S3).The sample showed no change in elution volume or molecular mass across the concentration range, and the observed molecular mass calculated at the leading edge of the peak was consistent with a 1:1:1 ratio of YenB:YenC3 NTD : YenC3 CTD in the complex.
Small-angle X-ray scattering data were subsequently collected from this sample (Fig. S4, Table 2) and an ab initio bead model built from the solution scattering data showed similar dimensions to the crystal structure (Fig. 3), the main difference being the presence of density inside the centre of the hollow 'shell'.In contrast, bead models built from the YenB/YenC2 NTD complex lacking the C2 CTD domain show a similar overall shape with a large internal cavity (Busby et al., 2013).We therefore conclude from the X-ray crystal structure and SAXS analyses that YenC3 CTD is encapsulated within the  cage in an unfolded state, and further analysis of the SAXS data (see the supporting information) supported this conclusion.This situation is likely to be the case for other TcB-TcC complexes as well.

Determination of the YenC3 CTD HVR crystal structure
We hypothesized that the C-terminal HVR domain of YenC3 is most likely a toxin active against insects, as is the case with other characterized TcC proteins, which cause apoptotic cell death in the epithelial cells of the insect midgut by disrupting the actin cytoskeleton (Hurst et al., 2011b;Aktories et al., 2015).The amino acid sequence of YenC3 CTD shows two distinct regions.Immediately downstream of the self-cleavage site are four copies of a proline-rich repeat with the consensus sequence PPPPPPMMGGN, which are likely to form extended poly proline helical segments (Adzhubei et al., 2013).These repeats are followed by a predicted globular domain of �230 amino acids.A BLAST search of the Genbank non-redundant protein database with the full YenC3 CTD HVR sequence identified a small family of RHSrepeat-containing proteins, with orthologues in a range of Gram-negative bacterial species, including Morganella morganii, Vibrio parahaemolyticus and several Pseudomonas species.None of the identified orthologues are functionally or structurally characterized.
In order to better understand the function of the HVR of YenC3, we expressed, purified and crystallized YenC3 CTD , excluding the poly proline repeat region, as extended prolinerich sequences tend to crystallize poorly (Williamson, 1994).We determined its crystal structure (PDB entry 6aqk) using anomalous diffraction from a platinum derivative (Fig. 4, Table 1).by the NAD-binding active site.Consistent with this structural similarity, the inferred active site cleft of YenC3 CTD contains the key catalytic arginine, serine and glutamate residues (R839, S881 and E919) that are conserved in the cholera toxin (CT) group of ADP-ribosyltransferases (Simon et al., 2014).

Analysis of the YenC3
The toxic HVR regions of the well characterized ABC toxins from P. luminescens are also ADP-ribosyltransferases. Like the C. botulinum C3 toxin, the HVR of TccC5 from P. luminescens ADP-ribosylates Rho proteins, causing constitutive activation and dysregulation of the actin cytoskeleton (Lang et al., 2010).In contrast, the HVR of TccC3 is more similar to S. typhimurium SpvB, as it ADP-ribosylates actin directly (Aktories et al., 2012), although TccC3 functions by stabilizing filamentous actin, whereas SpvB modifies monomeric actin.
The C. botulinum C3-like ADP-ribosyltransferase toxins recognize Rho as their substrate via a conserved, solventexposed phenylalanine residue (F208 in C3 toxin) in a loop of the central �-sheet of the enzyme.The loop (indicated by a red arrow in Fig. 4) reaches across the active site to interact with the helical domain, supported by the N-terminal �-helix.During substrate recognition, the exposed phenylalanine recognizes a hydrophobic pocket on the surface of Rho, and positions a conserved glutamine residue (Q211 in C3 toxin) to interact with the residue in Rho (N41) that is the target for  ADP-ribosylation (Han et al., 2001).Neither F208 nor Q211 is conserved in the YenC3 CTD sequence, and neither the arrangement of the loop nor the N-terminal helix are conserved in the YenC3 CTD structure.This makes it unlikely that Rho is the cellular target for YenC3 CTD .
In contrast, there are several features of the YenC3 CTD structure that are similar to S. typhimurium SpvB, including the lack of an N-terminal helix, the presence of an extra helix in the helical domain and a characteristic helical insertion in a loop of the central �-sheet (indicated by a blue arrow in Fig. 4).We therefore predict from structural similarity that YenC3 CTD is most likely an ADP-ribosyltransferase toxin, with actin as its likely cellular target.SpvB is a member of a large family of ADP-ribosyltransferases that modify R177 of actin, consequently inhibiting its polymerization by introducing a steric clash that prevents actin monomers interacting to form a normal filament (Margarit et al., 2006).This disruption of actin filament formation then leads to an apoptotic mechanism of cell death (Browne et al., 2002).

Toxin expression profiles in Y. entomophaga
Having established that YenC3 is able to form a complex with YenB in vitro, we investigated whether YenB and YenC3 were co-expressed and/or formed functional complexes that could be detected in vivo.We first assessed whether mRNA transcripts from both the yenC3 and the yenB genes were present under the same growth conditions.RT-PCR was carried out using gene-specific primers and mRNA prepared from a Y. entomophaga culture at the late log phase of growth, a growth phase at which the expression of Yen-Tc had previously been observed (Hurst et al., 2011b).We observed the presence of both yenB and yenC3 transcripts [Fig.5(a)], demonstrating that the genes are contemporaneously expressed, and that YenC3 and YenB are potentially available to combine to form a mature Yen-Tc complex within the cell.
It is known that genetic loci that encode multiple TcC-like proteins give rise to mixed toxin populations containing the same A and B subunits, but combined with different C sub- units (Hurst et al., 2011b).To confirm that proteins derived from the detected transcripts were indeed able to form detectable complexes in vivo, we carried out mass spectrometry analysis of Yen-Tc purified from Y. entomophaga culture supernatant [Fig.5(c)].Validating the presence of a subpopulation of Yen-Tc complexes that incorporated YenC3 was challenging due to the high sequence identity between YenC3 and the two known TcC proteins derived from the Yen-Tc pathogenicity island, YenC1 and YenC2 [Fig.5(b)].Regardless, we found clear and unambiguous evidence for the presence of four peptide sequences unique to YenC3 in our mass spectrometry data, confirming that YenC3 is indeed secreted as part of a complete Yen-Tc holotoxin assembly, albeit at a much lower level than YenC1 and YenC2.A labelfree analysis using the summed intensities of all the unique peptides identified by mass spectrometry across all the gel bands analysed suggested that YenC3 is the least abundant of the three TcC proteins present in Yen-Tc, with a relative abundance of approximately less than 1% [Fig.5(d)].We therefore conclude that YenC3 is a bona fide Yen-Tc subunit, despite the fact that the gene coding for it is located a considerable distance from the Y. entomophaga pathogenicity island that encodes the cognate components required for it to form a functional, secreted toxin complex, and is a minor, but nonetheless potent cytotoxic component of the mature Yen-Tc holotoxin population.

Discussion
Y. entomophaga has a single characterized toxin complex locus containing two open reading frames encoding the TcA protein, a single tcB gene, two tcC genes and two chitinaseencoding genes (Hurst et al., 2011b).This work describes the identification of an orphan tcC gene at a different locus in the Y. entomophaga genome.The YenC3 protein produced from this orphan gene is able to form a complex with YenB, encapsulate its C-terminal cargo and perform the self-cleavage necessary to enable subsequent delivery when incorporated into a functional holotoxin assembly.
Organisms containing polymorphic toxin systems (Zhang et al., 2012;Jamet & Nassif, 2015) often have orphan modules (incomplete gene fragments consisting of the Cterminal toxin-containing portion) at other locations in their genome, allowing them to rapidly generate different toxin complexes by genetic rearrangement.In this case, however, the orphan tcC gene (yenC3) is expressed at the same time as yenB, suggesting that it forms a complex with the TcA and TcB components during assembly of the complete ABC toxin complex.This would lead to a mixed population of toxin complexes comprising at least three different toxic payloads, and presumably increasing its overall observed potency.This is in agreement with our previous work, showing that individual TcC proteins are unable to cause the wide variety of cellular effects seen with the natively produced toxin complex (Marshall et al., 2012).Bioinformatic analysis of YenC1 CTD suggests that it contains a cytotoxic necrotizing factor domain (Hurst et al., 2011b), which would modify Rho.YenC2 CTD , on the other hand, contains a predicted nucleic acid deaminase (Busby et al., 2013) and presumably causes cell death by DNA and/or RNA damage rather than disruption of the cytoskeleton.We speculate that the presence of a third tcC gene encoding a third, different Cterminal effector cargo targeting actin may in part explain the high potency of the complete Yen-Tc -it shows an LD 50 of 1.59 fmol per diamond-back moth larva (Landsberg et al., 2011).
The crystal structure of YenB/YenC3 shows clear electron density for the majority of YenB and YenC3 NTD .Electron density for the remaining, otherwise unaccounted for protein was patchy, and despite the fact that a model of the YenC3 CTD HVR could not be built into this density, SAXS analysis confirmed that the YenC3 CTD is contained within the hollow shell formed by YenB and YenC3 NTD .The interior surface of this shell is mainly positively charged with hydrophobic patches, features that likely maintain the cargo in an unfolded state (Busby et al., 2013;Meusch et al., 2014).Direct evidence for this was found on closer inspection of the crystal structure, with areas of electron density attributable to YenC3 CTD packing against hydrophobic grooves or stacking on aromatic residues.
The structure of the YenC3 CTD HVR shows it adopts a mono-ADP-ribosyltransferase fold, with structural similarities to the S. typhimurium virulence protein SpvB, suggesting that the YenC3 CTD HVR modifies actin directly.Targeting actin would mean that YenC3 CTD is an approximate functional analogue of the HVR of TccC3 from P. luminescens, despite the proteins being undetectably similar in amino acid sequence.The structure of TccC3 has recently been determined using solution NMR spectroscopy and its mechanism of action established (Belyy et al., 2022).Like YenC3 CTD , TccC3 is a two-domain protein, with an N-terminal helical domain and a C-terminal ADP-ribosyltransferase domain.However, TccC3 has a non-canonical structure, with both domains showing considerable topological deviation from other structurally characterized ADP-ribosyltransferases, and a different substrate-binding mechanism.However, it does retain a functional core NAD + -binding region, and exerts cellular toxicity by ADP-ribosylating residue T148 of F-actin, preventing the severing of actin filaments by the actin depolymerizing factor cofilin.
The N-terminal poly proline repeat region of YenC3 CTD is unlikely to play a direct catalytic role in the toxin mechanism.However, it does bear superficial similarity to the formin homology region 1 (FH1) domain (PFAM PF06346).The FH1 domain binds to profilin which in turn localizes it to the fastgrowing barbed end of actin filaments (Evangelista et al., 2003).We speculate that the poly proline region of YenC3 CTD may function in a similar manner, helping to localize the toxin to its actin target.
In summary, this study describes the structure and proposed function of a previously uncharacterized toxic payload of the ABC toxin complex from the entomopathogenic bacterium Y. entomophaga.This toxin is expressed at the same stage as other components of the ABC complex and can bind to the rest of the toxin complex machinery to reconstitute a full ABC complex.Hence, apparently orphaned toxin genes present throughout bacterial genomes can be incorporated into complete ABC complexes, further expanding the repertoire of toxins that can be delivered by a shared mechanism.This situation may also be the case for other RHS-repeat containing proteins that are found in many bacterial genomes.

Cloning, expression and purification
The genes yenB and yenC3 from Y. entomophaga (GenBank accession Nos.PL78_03760 and PL78_18780, respectively) were cloned into the pET-Duet1 plasmid for coexpression.yenC3 CTD (consisting of residues 740-965) was cloned into the pDEST17 expression plasmid.Expression was performed in ZYM-5052 auto-induction medium (Studier, 2005), and cell lysis and protein purification were carried out as previously described (Busby et al., 2012).

Crystallization and structure determination of YenB/ YenC3
Initial crystallization screens were performed in 96-well sitting-drop format with protein at 10.6 mg ml À 1 .Fine screening was performed, with the best crystals growing in 22%(w/v) PEG 6000, 0.2 M TAPS pH 8.5.Crystals were cryoprotected by briefly soaking in mother liquor containing 20%(v/v) glycerol and snap-cooled in liquid nitrogen.Diffraction images were collected over 472 � of rotation to a resolution of 2.4 A ˚using a Rigaku MicroMax 007HF X-ray generator and a mar345dtb image plate detector.The crystal was maintained at 110 K by a cryostream of cold nitrogen gas.Data were integrated using XDS (Kabsch, 2010) and scaled and merged using AIMLESS (Evans, 2011).Molecular replacement was performed with Phaser (McCoy et al., 2007), using the structure of YenB/YenC2 NTD (PDB entry 4igl) as a model.The structure was refined by rounds of manual model building with Coot (Emsley et al., 2010) and refinement with REFMAC5 (Murshudov et al., 2011).Data processing and model statistics are presented in Table 1, and the structure has been deposited in the Protein Data Bank (PDB entry 5kis).

Crystallization and structure determination of YenC3 CTD
Crystal screens were performed in 96-well sitting drop format with protein at 17.4 mg ml À 1 .Conditions were finescreened, with the best being 29%(w/v) PEG 2000 MME, 0.15 M KBr.Crystals were derivatized by transferring into a drop containing 30%(w/v) PEG 2000 MME, 0.15 M KBr, 10 mM K 2 PtCl 4 and incubating for 10 min.Crystals were then briefly back-soaked in 30%(w/v) PEG 2000 MME, 0.15 M KBr, 25%(v/v) glycerol and snap-cooled in liquid nitrogen.Diffraction data were collected on the MX2 beamline of the Australian Synchrotron in two positions along the crystal, with 360 � collected at each position.Crystals were maintained at 100 K by a stream of cold nitrogen gas.Diffraction data were collected at the platinum L III -edge (wavelength 1.07219A ˚) with an EIGER X 16M photon counting detector in 0.1 � wedges.Data were integrated and scaled using XDS (Kabsch, 2010) and merged using AIMLESS (Evans, 2011).Heavy atom sites were found and SAD phasing was performed using SHELX (Sheldrick, 2010).Density modification was performed with DM (Cowtan & Main, 1998) and this produced electron density sufficient to allow automatic model building with ARP/wARP (Langer et al., 2008).This was followed with rounds of manual model building in Coot (Emsley et al., 2010) and refinement with REFMAC5 (Murshudov et al., 2011).Data processing and model statistics are presented in Table 1 and the structure has been deposited in the PDB (PDB entry 6aqk).

SEC-MALLS
To assess protein homogeneity and oligomeric state, sizeexclusion chromatography with multi-angle laser light scattering (SEC-MALLS) was used.A dilution series consisting of 100 ml protein samples at 3.7, 1.8 and 0.9 mg ml À 1 was used.Samples were run through a Superdex S200 Increase 10/ 300GL column, and a Dionex HPLC, PSS SLD7000 7-angle MALLS detector and Shodex RI-101 differential refractive index detector were used.Molecular weight calculations were performed using PSS winGPC Unichrom software.

SAXS
Protein was purified to homogeneity as determined by SEC-MALLS and SDS-PAGE (Fig. S3).Protein was exhaustively dialysed against 20 mM HEPES pH 7.5, 150 mM NaCl, 0.5 mM TCEP, with a sample of the dialysate used as the solvent blank.Small-angle scattering data were collected for a concentration series at the SAXS/WAXS beamline of the Australian Synchrotron.Data collection and processing statistics are presented in Table 2 and Fig. S4.Data were integrated using scatterBrain (Australian Synchrotron, 2019) and subsequent analysis performed using the ATSAS suite (Petoukhov et al., 2012).Scattering data were placed on the absolute scale by normalization against water (Orthaber et al., 2000).Data were extrapolated to zero concentration using PRIMUS (Konarev et al., 2003) prior to ab initio modelling.
Model building was performed using 20 runs of DAMMIF, followed by model superposition and selection using DAMSUP, DAMSEL, DAMAVER, DAMSTART and a final model refinement run with DAMMIN.The 20 initial models had a mean NSD of 0.546 and a standard deviation of 0.041; 19 of the models were selected to be used in the averaging process.

Reverse transcriptase PCR
For preparation of Y. entomophaga RNA, Y. entomophaga was grown in LB media to a cell density of 4.4 � 10 9 colony forming units ml À 1 .Two 200 ml aliquots were independently aliquoted into 1.5 ml micro centrifuge tubes containing 400 ml of RNAprotect Bacteria Reagent and RNA was then prepared using the RNeasy Mini Kit (Qiagen), following the manufacturer's instructions.The concentration of the resultant RNA was measured using a Nanodrop ND-1000 spectrophotometer and the integrity was confirmed by 1% agarose gel electrophoresis.

Mass spectrometry
To assess the presence and abundance of YenC3 in native toxin complexes, native Yen-Tc was purified from the supernatant of Y. entomophaga cultures grown in LB broth, essentially as described previously (Jones and Hurst, 2016).Following Superose 6 SEC, proteins were in-gel digested with trypsin according to Shevchenko et al. (2006) with minor alterations.Liquid chromatography conditions were as described previously (O'Brien et al., 2020) except that peptides were separated with a gradient of 5 to 50% solvent B over 34 min, then 50-98% B over 6 min.The samples were analysed on an Orbitrap Elite mass spectrometer (Thermo-Fisher Scientific).Survey scans were acquired in the Orbitrap at an m/z of 350 to 1800 with a resolution of 60 K using a maximum injection time of 200 ms.The ten most intense precursors with charge states above two were selected for MS2 in the ion trap using an isolation window of 2 Da, a maximum injection time of 150 ms and normalized collision energy of 35%.

Data analysis
The Sequest HT node in Proteome Discoverer (version 2.5.0.400;ThermoFisher Scientific) was used to search the combined spectra from all excized gel slices.The protein database contained eight Y. entomophaga protein sequences with a custom contaminants database.Cleavage specificity was set as trypsin, enzyme specificity was set as specific and a maximum of two missed cleavages were allowed.Mass tolerances of 15 p.p.m. and 0.6 Da were applied to precursor and fragment ions, respectively.Cys-S-�-propionamide was set as a static modification, and dynamic modifications were set to deamidation of Asn and mono-oxidized Met.Confident peptide-to-spectrum matches (PSMs) were assigned using the Percolator node with default settings.For label-free quantification, precursor peak intensities were obtained using the 'Minora Feature Detector' node in the processing workflow and the 'Feature Mapper' and 'Precursor Ions Quantification' nodes in the consensus workflows.Protein abundance was calculated by dividing the summed intensities of unique PSMs for a single protein by the summed intensities of unique PSMs from all Y. entomophaga proteins.

Figure 1 (
Figure 1 (a) Crystal structure of YenB/YenC3 NTD , coloured from blue at the N-terminus to red at the C-terminus.YenB is highlighted in green and YenC3 NTD in blue.(b) Strands of the �-sheet shell around the central cavity, coloured as in (a).The �-propeller domain is highlighted in orange.Surface representations showing (c) the exterior and a vertical slice through showing (d) the internal cavity.YenB is shown in green and YenC3 NTD is shown in blue.
Figure 2 Electron density features on the interior face of the hollow shell.(a) Flattened volume of electron density stacking 3.9 A ˚above the plane of a phenylalanine residue.(b) and (c) Areas of tube-like electron density (F o À F c map, � = 3.0) occupying hydrophobic clefts on the interior surface.

Figure 4 (
Figure 4 (a) The C3 exoenzyme from Clostridium botulinum (PDB entry 1r4b).The substrate recognition loop is indicated with a red arrow.(b) The crystal structure of YenC3CTD (PDB entry 6aqk).(c) The structure of SpvB from S. typhimurium in complex with NAD (shown as sticks) (PDB entry 2gwl).The helical insertion in the central �-sheet is indicated with a blue arrow.All structures are coloured from blue at the N-terminus to red at the C-terminus.

Figure 3 (
Figure 3 (a) Small-angle X-ray scattering bead model of YenB/YenC3 NTD / YenC3 CTD (blue mesh) with the YenB/YenC3 NTD crystal structure superimposed.(b) Experimental small-angle X-ray scattering of YenB/ YenC3 NTD /YenC3 CTD (circles) and the theoretical scattering of the bead model (blue line).The residual plot demonstrates a very good fit of the theoretical scattering to the experimental data: � 2 = 0.7548.(c) SAXS bead model (blue mesh) with the crystal structure (green mesh) superimposed.(d) Vertical slice-through of (c) showing the large internal cavity present in the crystal structure (green) but absent in the SAXS model (blue).

Figure 5 (
Figure 5 (a) RT-PCR of Y. entomophaga culture at the late log phase, showing contemporaneous amplification of both yenB and yenC3 genes.Amplicons were visualized on a 1% agarose gel alongside the GeneRuler 1 kb DNA ladder (Thermo Scientific).(b) Multiple sequence alignment of YenC1, YenC2 and YenC3, highlighting the diagnostic sequences used to establish the presence of YenC3.(c) SDS-PAGE of purified Yen-Tc used to identify the constituent protein subunits.(d) Relative abundance of Yen-Tc component proteins, calculated using the summed intensities of unique peptides per protein.

Table 1
Crystallographic data collection and refinement statistics.
† Friedel pairs treated as different reflections.

Table 2
SAXS data collection statistics.