2.1. 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 His₆-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 C-terminal 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 C-terminal sequence is cleaved but remains associated with the complex are entirely consistent with YenC3 functioning as a bona fide TcC protein.

2.2. The YenB/YenC3 complex structure

The complex consisting of YenB and both the N- and C-terminal 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, 2016). An RMSD of 0.77 Å 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 C-terminal 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).

Figure 1 (a) Crystal structure of YenB/YenC3NTD, coloured from blue at the N-terminus to red at the C-terminus. YenB is highlighted in green and YenC3NTD in blue. (b) Strands of the β-sheet shell around the central cavity, coloured as in (a). The β-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 YenC3NTD 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 Å above the plane of a phenyl­alanine residue. (b) and (c) Areas of tube-like electron density (Fo − Fc map, σ = 3.0) occupying hydro­phobic clefts on the interior surface.

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 ∼3.9 Å from the plane of a phenyl­alanine residue (Y1220), possibly representing aromatic residues of the cargo protein forming π-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 multi-angle 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.

Table 2 SAXS data collection statistics Data collection parameters Instrument SAXS/WAXS beamline at the Australian Synchrotron Detector Pilatus 1M Beam geometry point Wavelength (Å) 1.0332 Camera length (mm) 3252 q range (Å−1) 0.005-0.292 Exposure time (s) 1 Concentration range (mg ml−1) 0.15–2.4 Temperature (K) 285   Structural parameters† I(0) (cm−1) [from P(r)] 0.48 Rg (Å) [from P(r)] 43.99 I(0) (cm−1) [from Guinier] 0.48 Rg (Å) [from Guinier] 43.85 Dmax (Å) 153.06 Calculated molecular mass‡ [from I(0)] (kDa) 276.9 Calculated molecular mass (from SAXS MoW2) (kDa) 269.0 Molecular mass from sequence (kDa) 273.8   Software employed Primary data reduction scatterBrain Data processing PRIMUS Ab initio analysis DAMMIF/DAMMIN Validation and averaging DAMAVER Computation of model intensities CRYSOL§ Three-dimensional graphics representation PyMOL †Calculated for a 2.4 mg ml−1 concentration. ‡Calculated from data placed on an absolute scale and using a partial specific volume of 0.7425 cm3 g−1. §Svergun et al. (1995).

Figure 3 (a) Small-angle X-ray scattering bead model of YenB/YenC3NTD/YenC3CTD (blue mesh) with the YenB/YenC3NTD crystal structure superimposed. (b) Experimental small-angle X-ray scattering of YenB/YenC3NTD/YenC3CTD (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: χ2 = 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).

2.3. 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 RHS-repeat-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 proline-rich 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).

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 β-sheet is indicated with a blue arrow. All structures are coloured from blue at the N-terminus to red at the C-terminus.

2.4. Analysis of the YenC3^CTD structure

A DALI (Holm, 2019) search of the PDB revealed that the structure of YenC3^CTD is most similar to a structurally conserved family of mono-ADP-ribosyltransferase toxins (CATH Superfamily 3.90.176.10). The closest structural homologues are the catalytic domain of the Salmonella typhimurium virulence protein SpvB (PDB entry 2gwl; Margarit et al., 2006), which ADP-ribosylates actin (Margarit et al., 2006); and the C3 secreted toxin from Clostridium botulinum (PDB entry 1r4b; Margarit et al., 2006), which ADP-ribosylates Rho GTPases (Aktories & Frevert, 1987; Ménétrey et al., 2008). The core ADP-ribosyltransferase folds of YenC3^CTD, SpvB and C3 toxin are all very similar, with an N-terminal helical domain separated from a β-sheet domain 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).

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, solvent-exposed phenyl­alanine residue (F208 in C3 toxin) in a loop of the central β-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 α-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 β-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).

2.5. 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.

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.

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 sub­units (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 label-free 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.

4.1. 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 co-expression. 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).

4.2. 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⁻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 Å 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).

4.3. Crystallization and structure determination of YenC3^CTD

Crystal screens were performed in 96-well sitting drop format with protein at 17.4 mg ml⁻¹. Conditions were fine-screened, 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 K₂PtCl₄ 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 1.07219 Å) 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).

4.4. SEC-MALLS

To assess protein homogeneity and oligomeric state, size-exclusion chromatography with multi-angle laser light scattering (SEC-MALLS) was used. A dilution series consisting of 100 µl protein samples at 3.7, 1.8 and 0.9 mg ml⁻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.

4.5. 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.

4.6. Molecular graphics

Figs. 1–4 were prepared using The PyMOL Molecular Graphics System, version 1.8, Schrödinger, LLC (DeLano, 2002).

4.7. Reverse transcriptase PCR

For preparation of Y. entomophaga RNA, Y. entomophaga was grown in LB media to a cell density of 4.4 × 10⁹ colony forming units ml⁻¹. Two 200 µl aliquots were independently aliquoted into 1.5 ml micro centrifuge tubes containing 400 µl 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.

cDNA synthesis was performed using the SuperScript IV First-Strand Synthesis System (Invitrogen). 4 µl of 2.5 µM dNTP mix, 1 µl of 50 µM Random Hexamers (Promega), 0.5 µl template RNA (1474 ng µl⁻¹) and 7.5 µl DEPC-treated H₂O was added and the sample incubated at 65°C for 5 min, followed by 1 min on ice. A solution containing 4 µl 5× Super Script IV Buffer, 1 µl 100 mM DTT, 1 µl RNaseOUT (Invitrogen) recombinant RNase inhibitor and 1 µl SuperScript IV Reverse Transcriptase were then added and the sample was mixed and incubated at 23°C for 10 min. The reactions were transferred to 55°C for 15 min, followed by a 10 min heat inactivation at 80°C. The RT control comprised the same protocol as above but with 1 µl DEPC-treated H₂O used in place of SuperScript IV Reverse Transcriptase.

50 µl PCR reactions were performed using Platinum Taq PCR Kit (Invitrogen). Combined 5 µl 10× PCR Buffer, 2 µl 50 mM MgSO₄, 1 µl 2.5 µM dNTPs (Invitrogen), 0.2 µl Taq DNA polymerase, 0.5 µl each primer (40 µM), 1 µl Template (either cDNA, -RT control or untreated RNA control) and 39.8 µl H₂O. Primer sequences: YenB-F: 5′-CTC ATC GCG TCC AAT AGC CT-3′, YenB-R: 5′-CGT ACT ACG CTG GCT GAG AG-3′, YenC3-F: 5′-AGC AAC TTG ACC GAA AAC GC-3′, YenC3-R: 5′-GCT TTC CGT TTC CGT ATC GC-3′. PCR cycles as follows: 95°C, followed by 32 cycles (95°C 30 s, 57°C 30 s, 72°C 60 s) with a final 5 min at 72°C.

4.8. 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 (ThermoFisher 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%.

4.9. 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-β-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.