Characterization of C3larvinA, a novel RhoA-targeting ADP-ribosyltransferase toxin produced by the honey bee pathogen, Paenibacillus larvae

Abstract C3larvinA is a putative virulence factor produced by Paenibacillus larvae enterobacterial-repetitive-intergenic-consensus (ERIC) III/IV (strain 11-8051). Biochemical, functional and structural analyses of C3larvinA revealed that it belongs to the C3-like mono-ADP-ribosylating toxin subgroup. Mammalian RhoA was the target substrate for its transferase activity suggesting that it may be the biological target of C3larvinA. The kinetic parameters of the NAD+ substrate for the transferase (KM = 75 ± 10 µM) and glycohydrolase (GH) (KM = 107 ± 20 µM) reactions were typical for a C3-like bacterial toxin, including the Plx2A virulence factor from Paenibacillus larvae ERIC I. Upon cytoplasmic expression in yeast, C3larvinA caused a growth-defective phenotype indicating that it is an active C3-like toxin and is cytotoxic to eukaryotic cells. The catalytic variant of the Q187-X-E189 motif in C3larvinA showed no cytotoxicity toward yeast confirming that the cytotoxicity of this factor depends on its enzymatic activity. A homology consensus model of C3larvinA with NAD+ substrate was built on the structure of Plx2A, provided additional confirmation that C3larvinA is a member of the C3-like mono-ADP-ribosylating toxin subgroup. A homology model of C3larvinA with NADH and RhoA was built on the structure of the C3cer-NADH-RhoA complex which provided further evidence that C3larvinA is a C3-like toxin that shares an identical catalytic mechanism with C3cer from Bacillus cereus. C3larvinA induced actin cytoskeleton reorganization in murine macrophages, whereas in insect cells, vacuolization and bi-nucleated cells were observed. These cellular effects are consistent with C3larvinA disrupting RhoA function by covalent modification that is shared among C3-like bacterial toxins.


Introduction
Many insect species are indispensable for the pollination of wild and cultivated plants and therefore essential for both the survival of natural ecosystems and for a sufficiently diverse human diet [1,2]. Among these insect species, honey bees play a prominent role because managed colonies are invaluable agents for targeted pollination in agriculture [3]. In fact, commercial pollination of fruit and crops is a pillar to agricultural prosperity extolling the economic value of honey bee colonies in many regions of the world [4,5]. Given the importance of honey bees in human life and prosperity, it is not surprising that they have received much scientific attention. Honey bee pathogens, however, are still a niche existence in science. Exceptions to this rule are those pathogens that cause considerable colony losses worldwide [6][7][8], like entry and toxin activity remains elusive so far, particularly because it is not essential for Plx2A cytotoxicity in the insect cell culture assays.
Recently, yet another C3-like exoenzyme of P. larvae, (C3larvin) herein called C3larvin trunc , was discovered via genome mining, identified as a mART toxin that targets RhoA, and suggested to be a virulence factor for P. larvae ERIC I and ERIC II [38]. However, C3larvin trunc was shown to lack N-terminal sequences responsible for cell-entry activity, and indeed, the toxin was unable to invade mouse macrophages [38]. Consistent with this observation, P. larvae ERIC I and ERIC II gene inactivation mutants lacking C3larvin trunc expression did not cause larval mortality compared with WT strains when used for experimental infection [39]. These data suggested that despite its enzymatic activity in biochemical assays [38], C3larvin trunc does not influence the virulence of P. larvae [39]. Further in silico analyses then revealed that in P. larvae ERIC I and ERIC II, the C3larvin trunc gene is part of a binary AB toxin locus which had been annotated as non-functional due to several disruptions of the open-reading frames coding for the A-and B-subunits [34]. In this context, a full-length C3larvinAB locus comprising non-interrupted genes for the A-(C3larvinA) and the B-subunits (C3larvinB) was found in a singular P. larvae ERIC III/IV strain. Remarkably, C3larvinA contains the N-terminal sequences [39] not present in the originally described, inactive C3larvin trunc [38].
Herein, we now report the biochemical and functional characterization of C3larvinA as a full-length, C3-like toxin virulence factor (a full-length version of C3larvin trunc ) [38] produced by P. larvae ERIC III/IV strains. We show that C3larvinA has 55% sequence identity with Plx2A, the other functional C3-like toxin of P. larvae produced by ERIC I strains. C3larvinA binds and hydrolyzes NAD + as substrate, has glycohydrolase (GH) activity, and enzymatically modifies RhoA with ADP-ribose. Hence, it is a classical mART enzyme and contains the hallmark catalytic motifs and residues conserved among C3-like toxins. Recombinantly expressed C3larvinA was used to characterize its functional and biochemical properties. Purified C3larvinA enters host cells and interferes with actin remodeling in mouse macrophages and cytokinesis in insect cells. Therefore, C3larvinA not only resembles Plx2A in that it has an associated B-subunit, but also shows the same phenotype in mammalian and insect cells as previously described for Plx2A [37]. Additionally, C3larvinA represents a full-length version of the previously characterized C3larvin trunc and can enter host cells. C3larvin trunc was unable to enter macrophages because it is an inactive toxin at the cell entry level due to an N-terminal truncation (missing most of the α-1 helix). Furthermore, C3larvinA and Plx2A are functionally and structurally similar virulence factors that originate from different P. larvae strains; C3larvinA is produced by P. larvae ERIC III/IV strains, while Plx2A is produced by ERIC I strains.

Experimental procedures Transformation, expression, and purification
The C3larvinA gene was cloned into a pET-28a + vector with an N-terminal hexa-histidine tag. This plasmid was used to transform chemically competent Escherichia coli BL21 λDE3 cells using the heat-shock method. The cells were then grown in 4 l of 2× YT media containing 30 μg/ml kanamycin at 37 • C. When the culture reached an OD 600 of 0.6, expression of C3larvinA was induced with 1 mM IPTG. The culture was grown for an additional 3 h at 37 • C before being harvested by centrifugation at 3000×g for 15 min at 4 • C. The cell pellet was resuspended in buffer containing 500 mM NaCl and 50 mM Tris/HCl at pH 7.5. The cells were then lysed using an Emulsiflex-C3 high pressure homogenizer (Avestin Inc., Ottawa, Canada) in the presence of 120 μM PMSF, 50 μg/ml CHAPS, 1 mM EDTA, and 100 μg/ml DNase. Next, the homogenate was centrifuged at 23700×g for 55 min at 4 • C to remove insoluble cell debris. The supernatant was then incubated with 10 mM MgCl 2 with mixing for 30 min at 4 • C. This solution was passed over a Ni 2+ -charged chelating FastFlow™ Sepharose column, which was then washed twice, first with lysis buffer containing 25 mM imidazole, and second, with lysis buffer containing 40 mM imidazole. A final wash with lysis buffer containing 250 mM imidazole was used to elute C3larvinA. Fractions were analyzed with SDS/PAGE, and fractions showing bands at the correct molecular weight for C3larvinA were dialyzed overnight in lysis buffer and further purified using a HiLoad 16/60 Superdex-200 column (GE Healthcare) in size-exclusion chromatography (SEC). Fractions showing pure protein after being analyzed via SDS/PAGE were pooled and concentrated.

Site-directed mutants of C3larvinA
Point mutations were introduced into the C3larvinA gene using the QuikChange ® Mutagenesis method (Stratagene, La Jolla, CA, U.S.A.) according to the manufacturer's instructions. The Gln and the Glu of the Q 187 -X-E 189 motif (catalytic center and major site) were exchanged individually in single variants (Q187A and E189A), or together in a double variant at residues Gln and Glu (A-X-A) by substitution with Ala. The STS motif (residues Ser 149 -Thr 150 -Ser 151 ) (NAD + substrate binding site) was substituted with three Ala residues (A-A-A). The catalytic Arg residue was also substituted with Ala (R104A). The C3larvinA variants (Table 1) were expressed and purified as described for the WT C3larvinA.

Circular dichroism spectroscopy
C3larvinA and catalytic variants were dialyzed into buffer containing 250 mM NaF and 10 mM Tris/HCl, pH 7.5. A JASCO J-815 circular dichroism (CD) spectropolarimeter was used to acquire the CD spectra of C3larvinA WT and variant proteins (0.16 mg/ml) at 25 • C in a 1-mm pathlength cuvette by scanning from 250 to 190 nm for a total of nine scans from which an average spectrum was calculated.

Expression and purification of RhoA CAAX-GST
RhoA does not express well in soluble form unless as a GST-fusion protein. Therefore, RhoA-GST fusion proteins are typically used for toxin kinetic analyses when RhoA is the putative cellular target [36,40]. Constitutively active human GST-RhoA ( CAAX) was recombinantly expressed in E. coli TG1 from a plasmid obtained as a gift from Dr. Joseph Barbieri (Medical College of Wisconsin) and purified essentially as described previously [38]. Briefly, RhoA CAAX (C = cysteine, A = aliphatic amino acid, X = any amino acid) -GST expression was started by inoculating a 2× YT culture including ampicillin (100 μg/ml) with transformed E. coli TG1. Bacteria were grown until an OD 600 of 1 was reached at 37 • C. Expression was induced with 1 mM IPTG at 27 • C overnight. The cultures were centrifuged at 3000×g at 4 • C for 12 min. Afterward, the pellets were dissolved in lysis buffer (10 mM HEPES, pH 7.5, 150 mM NaCl, 2.5 mM MgCl 2 , 1 mM DTT) including freshly added DNase. The pellet was lysed by sonication or high-pressure homogenization, and cell debris was removed by centrifugation. The supernatant was passed three times over a column containing glutathione resin (GenScript, Piscataway, NJ, U.S.A.) calibrated with lysis buffer. The column was washed five times with lysis buffer and the recombinant protein eluted with lysis buffer containing 20 mM reduced glutathione. To remove the glutathione, the purified protein was dialyzed into lysis buffer at 4 • C overnight.

Homology model of C3larvinA
A homology model of C3larvinA was built based on the 1.65Å Plx2A crystal structure (PDB: 5URP, apo/substrate-free; 55% sequence identity, Figures 1B and 2A) using Phyre2 (Protein Homology/Analog/Y Recognition Engine) [41] and was reported with 100% confidence. The Phyre2 method uses template detection by HHpred 1.51 [42], secondary structure prediction with Psi-pred 2.5 [43], disorder prediction using Disopred 2.4, and multi-template modeling and ab initio with Poing 1.0 [44]. The Plx2A structure was chosen as the template rather than C3larvin trunc because the latter lacks a full helix 1 sequence whereas the former has a functional helix 1 [37]. The homology model includes the entire polypeptide chain of C3larvinA, but not the recombinant His-tagged, N-terminal region) and shares similar topology with the other C3-like toxins ( Figure 2B). It is identical (99% sequence identity) with C3larvin trunc from ERIC I and II P. larvae strains except for an extended N-terminal region which includes a full-length helix 1 ( Figures 1A and 2B). In order to correctly model the helix 1 region within C3larvinA, Plx2A was chosen to provide a suitable template to model this helix (Arg 27 -Trp 41 in C3larvinA) and to position the nearly conserved Phe 23 and semi-conserved Glu 25 residues located further upstream of helix 1 and which are found in the C3-subgroup ( Figure 1A) except the C3larvin trunc sequence. The C3bot1-NAD + complex (PDB: 1GFZ.A) was used as the basis to model in the NAD + substrate into the active site of C3larvinA [40].

Force-field settings and structure preparation
Protein preparation and molecular mechanics (MM) calculations were performed using the computational suite Molecular Operative Environment (MOE) release 2018.10 (Chemical Computing Group Inc, Montreal, CA). The force field employed was the MOE Amber12:EHT, with AMBER12 parameters set (ff12) for protein, and parameters Multiple sequence alignment of selected C2 and C3 toxins, and C3larvinA using the T-Coffee web server to align the sequences and ESPript to generate the figure [70]. Identical (or nearly so) residues between both C2 and C3 toxins are printed in red text and highlighted in yellow; identical (or nearly so) residues shared among the C2 toxins with P. larvae toxins, C3larvin trunc , Plx2A and C3larvinA, are bound by red rectangles. Abbreviations: IMAC, immobilized-metal-affinity chromatography. C3larvinA is shown in light green and RhoA in salmon color. The NADH inhibitor is bound in the active site and is shown in stick format colored with standard element colors, Gln41 in RhoA is circled and is shown in stick format colored magenta; Mg 2+ is shown as a cyan sphere and GDP-γS is shown in stick format with standard element colors. (D) Sequence alignment of C3cer and C3larvinA C3-like toxins using the T-Coffee web server to align the sequences and ESPript to generate the figure [70]. Identical residues between both C3-like toxins are printed in red text and highlighted in yellow; the conserved loop regions that form the critical interactions with the RhoA substrate based on the C3cer-NADH-RhoA crystal complex (PDB:4XGS) are bounded by red rectangles [40].
calculated from the Extended Hückey Theory for the NAD + molecule. For the implicit solvent model, the Generalized Born-Volume Integral (GB/VI) formalism was employed, with dielectrics ε pro = 1 for the interior of the protein.
The MOE Protonate3D module was used to assign the ionization states and tautomers of side-chains at T = 300 K, pH = 7.4 and 0.1 M of ionic strength, along with the GB/VI solvation model and MMFF94 partial charges. The molecular surfaces are solvent-excluded surfaces obtained by rolling a probe sphere of 1.4Å diameter (water radius) and colored by several schemes. The van der Waals-interaction surfaces correspond to zero-potential contours of the van der Waals potential, E vdw = 1, between the specific set of atoms and a water O-atom as mobile probe, using a standard 12-6 Lennard-Jones definition.

Modeling the C3larvinA-NAD + complex
The NAD + molecule was taken from the C3bot1-NAD + complex (PDB:2C8C) and docked (translated) into the protonated apo C3larvin homology model built on the X-ray structure of Plx2A (PDB: 5URP) after optimal superposition of both proteins by their C α -atoms of pocket residues. Backbone atoms of this initial C3larvinA-NAD + complex were fixed, and the system was energy minimized (root-mean-square (RMS) gradient ≤ 0.001 kcal/mol/Å 2 ) in an implicit solvent model (ε sol = 80). Then, the energy function was updated to vacuum (ε sol = 80), and the MOE Solvate module was used to (i) locate the center of mass of the toxin at the center of a periodic box of 69.17 × 60.16 × 47.75Å 3 (edge lengths), (ii) solvate the complex with 6485 TIP3P water molecules at a density of 1.023 g/cm 3 , and (iii) neutralize the system by incorporating nine Cl − ions at optimal locations. The system was energy minimized in a stepwise fashion (each to a RMS gradient ≤ 0.01 kcal/mol/Å 2 ) as follows: first, the complex was fixed and the solvent (water and ions) was relaxed, then backbone atoms were fixed and side-chains, NAD + , and solvent were energy minimized, and finally the full system was minimized. With this molecular system, a molecular dynamics (MD) simulation was performed by the Scalable MD (NAMD) simulator release 2.9 [45], under periodic boundary conditions by wrapping protein and solvent, with an integration time of 1 fs and recording each 5 ps under the following sequential steps: (i) 1000 ps heating from 0 to 295 K; (ii) 4000 ps equilibration at 295 K; and finally (iii) an NPT ensemble at 295 K and 1 atm of production phase for 100 ns. Then, 20000 frames of the MD trajectory were stripped off solvent molecules and calculated the potential energy of the C3larvinA-NAD + decoys under an implicit solvent with ε sol = 80. All decoys with potential energy lower than the average value were selected to an energy minimization (RMS gradient ≤ 0.01 kcal/mol/Å 2 ). The decoy with the lowest potential energy after this geometry optimization step was saved and reported as the 'C3larvinA-NAD + complex' model.

Differential-scanning fluorimetry
The thermal stability of C3larvinA WT and variants was measured in triplicate measurements by melting-curve analysis in a StepOnePlus Real-time PCR system (Applied Biosystems, Foster City, CA, U.S.A.) using Protein Thermal Shift dye, Sypro Orange ® according to the manufacturer's instructions (Applied Biosystems) adopted from a previous method [46,47]. Melting curve analysis of the purified proteins established that C3larvinA and all active-site variants had a single, distinct melting point and similar melting temperature values (T M ) ( Table 1) indicating that all purified proteins were stable and properly folded.

GH activity
The GH activity of C3larvinA against etheno-NAD + (ε-NAD + ) was measured on a Cary Eclipse fluorescence spectrometer (Agilent Technologies, Mississauga, Canada) with 305 nm excitation and 405 nm emission wavelengths, and bandpasses of 5 nm. C3larvinA at 20 μM and ε-NAD + concentrations ranging from 0 to 500 μM in reaction buffer (50 mM NaCl and 20 mM Tris, pH 7.9) were mixed in a total volume of 75 μl. The reaction was monitored for 5 min, and the resulting slope was converted from relative fluorescence units into μM concentrations using a standard etheno-AMP (ε-AMP) curve. All Michaelis-Menten kinetics values were calculated from initial rate data using GraphPad ver 8 Software (La Jolla, CA, U.S.A.).

NAD + substrate binding
The affinity between C3larvinA and β-NAD + was assessed in a tryptophan fluorescence-quenching assay using a Cary Eclipse fluorescence spectrometer with 295 nm excitation and 340 nm emission, and 5 nm bandpasses. A solution of 1.25 μM C3larvinA protein in 600 μl buffer (20 mM Tris, pH 7.9, 50 mM NaCl) was titrated with β-NAD + concentrations between 1 and 1000 μM. The average of all the standard curve slopes was considered for the conversion of fluorescence units/min of the initial sample slope into [ε-ADP-ribose] formed/min. The converted slopes were plotted against the ε-NAD + concentration and fitted to the Michaelis-Menten model using GraphPad Prism Ver 8 software to calculate the kinetic parameters. The assay was repeated in triplicate with three technical replicates for each sample.

Transferase activity
The ADP-ribosylation activity of C3larvinA against RhoA-GST was measured using an end-point assay. It was not possible to monitor the modification of RhoA-GST with ADP-ribose since this reaction product is unstable. The second product of the C3larvinA-catalyzed transferase reaction to RhoA-GST, nicotinamide, was measured using an Agilent high-performance liquid chromatography (HPLC) system. To measure the kinetic parameters in relation to RhoA-GST, β-NAD + was held at 300 μM to ensure saturation while varying the concentration of RhoA-GST from 0 to 80 μM. Conversely, to measure parameters in relation to β-NAD + , RhoA-GST was held at 20 μM to ensure saturation while varying the concentration of NAD + from 0 to 500 μM. The transferase reaction was conducted in buffer containing 5 mM MgCl 2 , 150 mM NaCl and 20 mM Tris/HCl, pH 7.4. The reaction was initiated with the addition of 1 μM C3larvinA and was stopped after 5 min with the addition of mobile-phase solution, including an internal standard (5% acetonitrile and 95% of 20 mM monobasic phosphate buffer pH 5.5, and 2.5 μg/ml para-4-nitrobenzoic acid; PABA). This produced a final ratio of 25% sample to 75% mobile phase (v/v). The solutions were passed through a Captiva filtration 96-well plate (Agilent Technologies) to remove the enzyme protein and then the sample was injected on to a Zorbax RX-C18, 5 μm, 4.6 mm × 12.5 mm reversed-phase column operating at 0.8 ml/min at 85 bar with 259 nm detection (Agilent Technologies, Mississauga, ON, Canada). An isocratic run of 10 min proved successful at separating the reaction components. A product standard curve was generated using various concentrations of nicotinamide (0-750 μM in mobile phase buffer) that were injected into the HPLC system. The area under the nicotinamide peak was determined using the peak analysis function in Origin 8.0 (Northampton, MA) and was standardized with the PABA internal standard. Background GH activity was corrected for each sample, and the calibrated area was converted into pmole of nicotinamide using a standard nicotinamide curve. Kinetic values were calculated using GraphPad version 8 software.

Yeast cytotoxicity assay
Toxicity of C3larvinA WT and catalytic variants was tested on Saccharomyces cerevisiae BY4741 (MATa, his3Δ1 leu2Δ0 met15Δ0 ura3Δ0) using a yeast growth-deficiency assay as previously described [37,48]. Briefly, electrocompetent S. cerevisiae cells were prepared according to a modified protocol [49]. Macrophage cytotoxicity assay J774A.1 murine macrophage cells were grown in Dulbecco's Modified Eagle's Medium with 10% fetal bovine serum in the presence of penicillin-streptomycin antibiotic. Cells were grown in 25-cm 2 breathable flasks with 5% CO 2 and passaged at 80-90% confluence. Cells were lifted through scraping and diluted ten-fold into the above medium. Confluent J774A.1 cells were used to assess the effect of C3larvinA cellular processes. Cells were diluted to 2.5 × 10 5 cells/ml in the presence of either 30 or 300 nM toxin concentrations. One hundred and fifty microliters of cell suspension was added in triplicate to a 96-well plate and allowed to incubate for 20 h. After this time period, cells were viewed under a microscope and photographed.

Insect cell culture and actin staining
In the present study, the cell line Sf9 derived from Spodoptera frugiperda (Lepidoptera) was used to analyze the effect of C3larvinA on insect cells. Sf9 cells were cultured at 27 • C in Insect-XPRESS w/L-Gln medium (Lonza, Verviers, Belgium) supplemented with 5% heat inactivated fetal bovine serum (Gibco, Thermo Fisher Scientific). Before the start of the assay, cells were grown to confluence. The cell number was determined in a Neubauer improved counting chamber. Cells were diluted to 0.5 × 10 5 cells/ml in medium supplemented with penicillin (250 IU/ml)-streptomycin (250 μg/ml) (Carl Roth GmbH and Co. KG, Karlsruhe, Germany). The cells were premixed with sterile filtered, purified C3larvinA with a final concentration near 1.0 μM in reference to the C3-like toxin Plx2A as previously described [37]. The negative control contained sterile filtered NaCl-Tris-buffer with an equal volume to the purified C3larvinA. A volume of 2.0 ml of each mixture was added to wells of a sterile Cellstar six-well tissue culture plate (Greiner Bio-One GmbH, Kremsmünster, Austria) which were previously equipped with a sterile coverslip.

P. larvae ERIC III/IV strain, 11-8051, encodes a C3-like mART toxin
The 248-residue (28.6 kDa) C3larvinA protein produced by P. larvae strain ERIC III/IV strain, 11-8051 [39] is a single-domain mART toxin that possesses a 26-residue leader sequence. The mature protein is 222 residues (25.7 kDa). It has a B-subunit partner that encodes a 562-residue, 62.2 kDa protein. C3larvinA has the hallmark catalytic Q-X-E signature of C3-like mART toxins ( Figure 1A) [39]. The ten C3-like exotoxins known so far are shown in the sequence identity matrix ( Figure 1B) and share a minimum of 27% sequence identity, and the most similar pair includes C3larvin trunc and C3larvinA at 99% identity ( Figure 1A,B). Unlike C3larvin trunc from ERIC I and II P. larvae strains, C3larvinA possesses an intact Helix 1 and is the full-length and biologically active C3-like toxin [39] ( Figure  1A,B). All C3-like enzymes are A-domain only toxins, except for Plx2 toxin, a major virulence factor exclusively expressed by P. larvae ERIC I strains [50]. Plx2A (A-domain only) shares 55% sequence identity with C3larvinA ( Figure 1B) and was recently characterized for enzymatic activity, RhoA target recognition, cell invasion, and its crystal structure was also determined [37]. The C3-like proteins all share a conserved α-3 motif, catalytic Arg, S-T-S motif, PN-loop, ARTT loop and catalytic Q-X-E motif ( Figure 1A). Additionally, the C3-like enzymes consist of a small mART domain (<25 kDa) and have both transferase and GH activity, although GH activity is considerably weaker than transferase activity [36,51]. Most C3-like toxins modify the small G-proteins, RhoA, RhoB, and RhoC, at the Asn 41 residue [36,51]. RhoA ADP-ribosylation occurs at the highest rate, followed by that of RhoB and then RhoC [52]. Rac and Cdc42 are targets of several C3-like toxins, but show weaker ADP-ribose substrate activity than RhoA [36,53]. C3stau1, 2, and 3 are C3-like toxins which modify RhoE and Rnd3 as well as RhoA, B, and C; however, C3stau toxins show weaker transferase activity against all other protein substrates compared with RhoA [36,53].

Expression of C3larvinA in E. coli
The C3larvinA (LarA) gene product was overexpressed in Rosetta E. coli BL21(λDE3) cells and the WT protein was purified at a yield of 1 mg/l culture using immobilized-metal-affinity chromatography (IMAC) followed by SEC. The purity level was assessed by SDS/PAGE ( Figure 1C, lanes 6 and 8 for IMAC and SEC, respectively), and the recombinant 27.8 kDa protein (N-terminal His 6 tag with TEV8 cleavage site) was positively identified by MALDI-TOF mass spectrometry ( Figure 1C spectrum).

C3larvinA structure
A homology model of C3larvinA with NAD + substrate bound in the active site was built based on the 1.65Å Plx2A crystal structure (PDB: 5URP, apo/substrate-free; 55% sequence identity, Figure 1B) (Figure 2A) using Phyre2 [41] The homology model includes the entire polypeptide chain of C3larvinA (but not the recombinant His-tagged, N-terminal region) and shares similar topology with the other C3-like toxins (Figure 2A,B). It is identical (99% sequence identity) with C3larvin trunc from ERIC I and II P. larvae strains except for an extended N-terminal region which includes a full-length helix 1 (Figures 1A and 2A,B). It folds into a mixed α/β structure with a β-sandwich core and displays a characteristic mART fold, containing two perpendicular β-sheets next to the P-N loop, responsible for binding the NAD + substrate (Figure 2A,B). Superposition of C3larvinA with the apo conformations of C3bot1 (PDB: 1G24), C3stau2 (PDB: 1OJQ), C3lim (PDB: 3BW8), Plx2A (PDB: 5URP) and C3larvin trunc (PDB: 4TR5) revealed low root mean square deviations (r.m.s.d) of 1.44, 1.39, 1.48, 1.82, and 0.63Å, respectively. However, C3larvinA has an extended N-terminus that protrudes beyond helix 1 and is longer than the N-terminus of both C3bot1 and C3bot2. A conserved residue, Phe 9 (C3bot1 numbering; Phe 23 in C3larvinA) in this extended N-terminus is found in nearly all C3-like toxins except for C3larvin trunc and C3cer. A second residue, Thr 10 (C3bot1 numbering; Glu 25 in C3larvinA) is conserved among all C3-like toxins except those from P. larvae ( Figure 1A). Thr 10 is substituted with Ile in Plx2A and a Glu residue in C3larvinA. Multiple sequence alignment between the N-termini of both C2-like and C3-like toxins show distinct clusters of conserved residues ( Figure 1D). These C3/C2-like toxins consist of enzymes that have C3-like activity, but for which a B-domain has been found. Despite their differences, there are highly conserved residues among all groups. This is interesting, since the N-terminal adaptor domain of the iota I a -subunit (PDB: 4GY2) is required to bind to the translocating I b -domain (B-domain) via its N-terminal Ca 2+ -binding motif [54,55]. This may point to a divergent evolution of C3-like toxins from an ancestral C2-like toxin where the binding machinery evolved to accommodate a single-domain enzyme [56].
Overall, the active site of C3larvinA is remarkably like C3bot1, with the location of the β-strands appearing nearly identical between the two structures. An interesting feature in C3larvinA is the presence of extended β-strands (β-5, β-6, β-7), the most significant being the β-sheet (β-5) on the N-terminal region of the ARTT-loop and the β-sheet (β-6) containing the Q-X-E motif (data not shown). As a result, the ARTT-loop has comparatively fewer residues than most C3-like toxins and may provide more stability within the active site. Also, the solvent-exposed loops, i.e., the ARTT-and PN-loops in C3larvinA, adopt a more open conformation compared with other C3-like structures (Figure 2A). The catalytic Glu 189 is in a similar position to that of other C3-like toxins. However, the catalytic Gln 187 of C3larvinA like C3larvin trunc (PDB: 4TR5) has a significantly different orientation than the conserved Gln residues found in other NAD + -bound conformations of C3-like toxins (Figure 2 B). The Gln 187 in its current position would clash with the residues of the PN loop of other C3-like toxins. It is rotated along the axis of the main chain, moving ∼4 A farther than the similar residue (Gln 172 ) in C3bot1. The source of this change in orientation may lie in the residue separating the Gln and Glu (Q-X-E motif). In other solved structures, the residue which separates these catalytic residues is a Leu (or another Gln, in the case of C3staus 1, 2, and 3). C3larvinA is like C3larvin trunc in which the Q-X-E intervening residue has been substituted with a Tyr (Tyr is also present in this position in C3cer); the intervening residue in Plx2A is a Leu. This Tyr residue found in C3larvinA may exert some steric hindrance with residues in nearby areas of the protein and cause a displacement of the backbone structure in this area. It is also possible that there is an induced-fit mechanism in this region of the structure, in which binding of a substrate would cause the Glu 189 to shift to a more catalytically relevant position. Upon binding an NAD + molecule, the ARTT-loop changes its conformation from a solvent-exposed environment to a more buried conformation in C3bot1. In particular, the Gln 172 residue in C3bot1 makes a large shift (∼8Å) toward the interior of the NAD + -binding cleft. Interestingly, the position of Gln 189 residue in C3larvinA is also in a similar location and orientation ( Figure 2B).

C3larvinA-NADH-RhoA model
A homology model of C3larvinA-NADH-RhoA was built based on the template of the C3cer-NADH-RhoA crystal structure (PDB: 4XSG) using Phyre2 based on hidden Markov models and detection by HHpred 1.51 [41] ( Figure  2C). The surface contact area between C3larvinA and RhoA is large (∼1200Å 2 ) and the toxin recognizes RhoA via the switch I, switch II, and interswitch regions as seen in the C3cer-RhoA crystal structure. C3larvinA is highly conserved (nearly identical) with C3cer in the loop regions that make contact with RhoA ( Figure 2D); these loop regions include L2 (residues 45-52, active-site loop), L3 (residues 100-110, adenine loop); L4 (residues 148-156, PN-loop), L5 (residues 175-183, ARTT-loop), L6 (residues 204-209). As observed for the C3cer-RhoA complex structure, there is no L1 (Tyr 60 -Tyr 62 ) in C3larvinA as seen in the iota I a complex with actin (PDB: 3BUZ) [57]. In the C3cer-RhoA structure, Tyr 180 in turn 1 of the ARTT-loop recognizes RhoA via a hydrophobic patch around the ADP-ribose acceptor residue, Asn 41 in RhoA. This residue is conserved in C3larvinA (Tyr 184 ) indicating that it interacts and catalyzes the modification of Asn 41 in RhoA in an identical manner to C3cer ( Figure 2D). Gln 183 in turn 2 (Gln-X-Glu motif) also interacts with Asn 41 in RhoA in the C3cer complex and the same residue and motif is found in C3larvinA (Gln 187 -X-Glu 189 ). Interestingly, the intervening 'X' residue in the Gln-X-Glu motif in both C3cer and C3larvin is a Tyr residue further cementing their identical catalytic signatures and detailed mechanism for RhoA modification. As observed for C3cer, the Tyr 184 in C3larvinA interacts with a patch on RhoA composed of Val 43 , Ala 56 , and Trp 58 . The hydroxyl group of Tyr 184 forms an H-bond with the main-chain carbonyl group of Leu 57 in RhoA. RhoA Asn 41 forms an H-bond with Gln 187 in the Q-X-E motif with the ARTT-loop of C3larvinA. Clearly, Gln 187 and Glu 189 participate in the ADP-ribose transferase reaction to RhoA Asn 41 based on the mutagenesis of this motif in C3larvinA, which resulted in a loss of enzyme function that restored the growth-deficient phenotype in yeast (variant C3larvinA AXA in Figure 3A). Additionally, Asp 175 in C3cer interacts with the critical Arg 5 in RhoA at 2.7 A and this residue is also conserved in C3larvinA (Asp 179 ).

C3larvinA shows strong yeast cytotoxicity
To test the toxicity of C3larvinA to eukaryotic cells, a yeast-based growth-inhibition assay was employed [48]. In this method, C3larvinA gene expression is under control of the CUP1 (copper-inducible) promoter, and an active toxin will cause a growth-defect phenotype in yeast as previously shown for two other C3-like toxins from P. larvae [37,38]. The effect of C3larvinA WT and active-site variants on yeast growth is shown in Figure 3A. This test The C3larvinA variants are shown as C3larvinA AXA (Q187A-X-E189A), C3larvinA AAA (S149A-T150A-S151A) and C3larvinA R104A. (B) CD spectra of C3larvinA WT (red), Q187A-X-E189A (green), S149A-T150A-S151A (black) and R105A (blue) were collected in aqueous solution (25 • C) containing 250 mM NaF and 10 mM Tris, pH 7.5 buffer. The concentration of the proteins was at 0.16 mg/ml and each protein CD consensus spectrum was obtained by scanning from 250 to 190 nm and is the average of nine independent spectra. The measurements of the kinetic and substrate-binding affinity parameters for C3larvinA are described in the 'Experimental procedures' section. The parameters for NAD + substrate binding and GH enzyme activity represent the mean + − SD of at least three different measurements.
showed that a weak C3larvinA WT gene-induction by copper in the yeast culture (0.5 mM) was highly toxic to yeast cells, even more so than ExoA (catalytic domain) from P. aeruginosa (positive control toxin) ( Figure 3A). The R104A variant did not recover the growth-defect phenotype when expressed in yeast, whereas the STS motif variant (S149A/T150A/S151A) partially restored the growth defect ( Figure 3A). Notably, the C3larvinA Q-X-E catalytic motif variant (Q187A-X-E189A) fully restored the C3larvinA growth defect in yeast. This pattern was also observed for the Q-X-E variant for C3larvin trunc [38]; however, the STS and NAD + -binding Arg motifs were not tested in this assay for C3larvin trunc . In the case of Plx2A, a major virulence factor of P. larvae ERIC I genotype that causes contemporary AFB outbreaks worldwide, the pattern of growth restoration in the catalytic signature is different from C3larvinA [37]. First, the Q-X-E catalytic motif variant was similar for both the Plx2A and C3larvinA toxins. However, the STS motif and NAD + -binding Arg motif Plx2A variants both completely recovered the growth defect in yeast [37]. The origin of these differences in the catalytic signature variants is not known, but may be related to the extended N-terminus in C3larvinA compared with Plx2A. C3cer toxin was previously shown to lose all transferase activity when either the catalytic Gln or Glu were substituted with an Ala residue [53]. This suggests that the three P. larvae C3-like toxins possess a highly specialized catalytic site, where both the Gln and Glu residues in the Q-X-E catalytic motif are essential for proper function. Thus, these results indicate that C3larvinA is a bona fide mART toxin and confirmed its cytotoxicity (caused by its mART activity) in a model (yeast) eukaryotic system. The C3larvinA enzyme is a more stable protein in aqueous solution with a T m value of 63 • C (Table 1) compared with 55 • C for Plx2A [37] and 51 • C for C3larvin trunc (unpublished data). These thermal stability data indicate that a full-length helix 1 (Plx2A; Figure 1A) adds folded stability to the enzyme compared with the truncated helix 1 in C3larvin trunc and the ADPRT fold is further stabilized in C3larvinA which harbors an additional 19 residues extending past helix 1 compared with Plx2A ( Figure 1A).

C3larvinA binds and hydrolyzes NAD + as substrate
C3larvinA has only two Trp residues (Trps 33 and 41; Figure 1A) and both are located within Helix 1 with Trp 33 being a conserved Trp found in both C2-and C3-like toxins ( Figure 1D). Trp 41 in C3larvinA is conserved in the P. larvae C3-like toxins ( Figure 1D). Both Trp residues face inwardly in the protein folded structure and likely provide stability to the ADPRT fold through numerous molecular packing interactions (data not shown). The Trp fluorescence was exploited to characterize the NAD + substrate binding to the active site for C3larvinA WT and variants ( Table 2). In these experiments, C3larvinA proteins were titrated with NAD + substrate which caused quenching of the intrinsic Trp fluorescence (data not shown) and binding isotherms (single-site model) were used to calculate the K D values for NAD + of approximately 56 μM for the WT protein (Table 2). This affinity for the NAD + substrate compares well with the affinity shown by other C3-like toxins such as C3bot1, Plx2A and C3larvin trunc of 60, 33, and 21 μM, respectively [37,38,58].

C3larvinA GH activity
GH activity is present as a secondary enzymatic activity in most mART enzymes, is likely not biologically relevant and represents an alternative reaction where OH − serves as the nucleophile in the absence of a target protein [59]. C3larvinA GH activity was characterized with a fluorescence-based assay developed previously (Table 3) [60]. C3larvinA GH activity showed Michaelis-Menten behavior and gave a K M value of 107 + − 20 μM and a k cat of 261 + − 20 × 10 −3 min −1 (Table 3). C3 enzymes show a range of kinetic parameters for GH activity; for example, C3lim had a K M (NAD + ) = 160 μM and a k cat = 2 × 10 −3 min −1 [61]. C3-like toxins from P. larvae, C3larvin trunc and Plx2A, had K M values of 120 and 176 μM, respectively and k cat values of 1.3 × 10 −3 and 58 × 10 −3 min −1 , respectively [37,38]. Kinetic parameters for transferase activity are shown for the RhoA target substrate. 1 The measurements of the kinetic parameters for C3larvinA GH and transferase activity are described in the 'Experimental procedures' section. The parameters represent the mean + − SD of at least three different measurements.

C3larvinA interferes with actin remodeling in macrophages
C3larvinA cell entry was tested against the J774A.1 macrophage cell line derived from mice. C3bot1 and C3lim can enter these cells at nanomolar concentrations [64]. Previous work suggested that the N-terminal helices of C3bot1 and C3lim may be important for their cell entry [64]. Both toxins were previously shown to cause distinct morphology changes in macrophages as seen by enlarged cells with filopodia-like protrusions, with the most obvious changes occurring in J774A.1 cells [64]. We previously showed that Plx2A caused cell morphological changes in murine macrophages indicative of interference with actin cytoskeletal processes [37]. Under identical conditions, WT C3larvinA caused macrophages to elongate in a toxin-dependent manner, indicative of toxin commandeering of RhoA-dependent cytoskeletal functions, such as actin remodeling (Figure 4). The three catalytic variants of C3larvinA, S149A/T150A/S151A, Q187A-X-E189A, and R105A, showed marginal effects on macrophage morphology, as expected. This demonstrates that C3larvinA causes actin remodeling in the target macrophage cells via its ADP-ribosyltransferase activity against RhoA.

C3larvinA interferes with cytokinesis in insect cells
We also tested the effect of purified C3larvinA on the insect cell line Sf9 from the fall armyworm, Spodoptera frugiperda (Lepidoptera). We chose this model system because to the best of our knowledge, there is no viable honey bee cell line currently available. In this assay, an effect of C3larvinA (final concentration approximately 1.0 μM) on Sf9 cells was observed similar to the effect of Plx2A on the same cell line [37]. Sf9 cells treated with purified C3larvinA exhibited a phenotype with slightly enlarged, binucleated cells indicating an interference in cytokinesis ( Figure 5, white arrows). Like Plx2A, C3larvinA did not cause an effect on the actin cytoskeleton of insect cells as visualized with phalloidin staining ( Figure 5). These results point to a remarkable difference in RhoA signaling in mammalian versus insect cells.

Discussion
C3larvinA and B were recently identified as a binary toxin pair that function as a virulence factor in P. larvae ERIC III/IV strain 11-8051 [39], which was originally isolated from Chilean honey [65]. In the present study, C3larvinA (29 kDa) was shown to be a typical C3-like mART toxin with the classical ADPRT-fold and catalytic signatures. Recombinant C3larvinA was expressed and purified from E. coli and was shown to possess both GH (OH − from water as the nucleophile) and transferase activities (Asn 41 from RhoA as the nucleophile) ( Figure 1C, Tables 2 and  3). C3larvinA is similar to previously characterized C3-like mART toxins and is most similar to the P. larvae C3-like toxins, all of which modify RhoA at Asn 41 [37,38,52,53]. The characterization of the GH activity using a fluorescent NAD + analog revealed that C3larvinA follows Michaelis-Menten kinetics with respect to the NAD + substrate. Kinetic parameters were determined for the GH activity, and were similar to those previously determined for C3-like toxins [37,38,52]. Transferase kinetic parameters to the RhoA substrate were also determined quantitatively using a novel HPLC-based method. Site-directed mutants of the active-site architecture of C3larvinA confirmed its structure and function as a C3-like mART enzyme, including a Q-X-E motif involved in the transferase reaction of ADP-ribose to Asn 41 in RhoA, a catalytic Arg residue involved in docking/orientation of the NAD + substrate and an STS motif required for stabilization of the NAD + -binding pocket [66]. Amino-acid residue substitution of each of these catalytic signature motifs caused a near total loss of C3larvinA GH activity ( Table 2). The expression of WT C3larvinA in the cytoplasm of yeast demonstrated that it is cytotoxic to a eukaryotic host as previously shown for both C3larvin trunc and Plx2A toxins [37,38]. Notably, the catalytic variant (Q187A-X-E189A) was not cytotoxic as expected; furthermore, the NAD + -binding variants of C3larvinA showed a surprising level of cytotoxicity toward yeast, with the S149A/T150A/S151A variant only showing a 30% reduction in cytotoxicity compared with the WT. Remarkably, the R105A variant was as cytotoxic as the WT toxin in yeast ( Figure 3A) which does not correlate with the GH activity data ( Table 2). The basis for this unexpected finding is currently not known and will require further investigation.
Previously, we determined the crystal structures of two P. larvae C3-like toxins, an N-terminal truncated C3larvin trunc (PDB:4TR5) [38] and Plx2A, an important virulence factor in P. larvae ERIC I strains (PDB: 5URP) [37]. C3larvinA from P. larvae ERIC III/IV strain 11-8051 shares 55% sequence identity with Plx2A ( Figure 2B). Consequently, we built a homology model of C3larvinA with NAD + substrate based on the Plx2A structure (PDB: 5URP), which clearly showed that C3larvinA has an extended N-terminus compared with Plx2A ( Figure 2A). The role of the N-terminal helix 1 in these P. larvae toxins was further supported by macrophage cell entry experiments. Previously, the truncated C3larvin trunc toxin was unable to enter macrophages because it lacks part of helix 1 ( Figure  1A,D), whereas a C3larvin-C3bot1 chimera possessing the C3bot1 N-terminal sequence (Tyr 2 -Trp 18 ) was capable  of entering macrophages [38]. In contrast, Plx2A from P. larvae was fully functional and entered macrophages causing the expected cell enlargement and elongation with filipodia-like extensions [37]. C3larvinA with its elongated N-terminus entered macrophages and gave the expected C3-like phenotype ( Figure 4) at a similar dose as seen for Plx2A. This suggests that the extended N-terminus in C3larvinA may play an additional role in its cell function since it does not modulate cell entry compared with the shorter Plx2A toxin. In this context, C3larvinA with its extended N-terminus is considerably more stable than either C3larvin trunc or Plx2A, but the role of the N-terminal extension is currently not known.
A homology model of C3larvinA in complex with RhoA ( Figure 2C) built on the C3cer-RhoA crystal structure [40] suggested that C3larvinA shares identical catalytic features for ADP-ribose transfer to RhoA and a highly similar catalytic mechanism. Furthermore, sequence alignment of C3larvinA with C3cer ( Figure 2D) showed that their catalytic signatures, including the contact loop regions (L2-L6) are highly conserved (∼50% identity; 98% similarity). The key catalytic residues involved with ADP-ribose transfer to Asn 41 in RhoA are identical between C3cer and C3larvinA indicating that these two C3-like toxins share an identical catalytic mechanism in the C3-like toxin subgroup.
Furthermore, C3larvinA was able to enter lepidopteran Sf9 cells as was previously shown for Plx2A [37]. The addition of purified C3larvinA to the Sf9 cells caused an enlargement of the cells and the presence of two nuclei indicative of perturbation of cytokinesis as previously observed for Plx2A [37]. It is known that members of the Rho subfamily of low molecular weight GTP-binding proteins participate in cell cycle progression. In particular, switching of RhoA between the activated and inactivated state is required for completion of cytokinesis [67][68][69]. In contrast with murine macrophage cells, C3larvinA did not influence the actin cytoskeleton as was also shown for Plx2A [37]. The ADP-ribosyltransferase toxins from P. larvae, C3larvinA and Plx2A, both were able to enter non-phagocytic insect cells pointing to a difference in cell entry to the other C3-like toxins, which have mammalian cells as their natural host. Interestingly, C3larvinA and Plx2A are the only functional C3-like toxins with a B-subunit [39,50]. This deserves further investigation. The kinetic data and cell culture assays in this study show that C3larvinA is more active than C3larvin trunc and show that C3larvinA can enter cells on its own in contrast with C3larvin trunc . This substantiates the hypothesis that only C3larvinA from P. larvae ERIC III/IV strain, 11-8051, is the active form of this toxin. Furthermore, C3larvinA has an intact B-subunit partner which also has been shown to play a role in virulence in P. larvae 11-8051 [39]. Notably, the interaction between A-and B-subunits is an interesting subject for future studies.