Structural Insights to the Heterotetrameric Interaction between the Vibrio parahaemolyticus PirAvp and PirBvp Toxins and Activation of the Cry-Like Pore-Forming Domain

Acute hepatopancreatic necrosis disease (AHPND) is a newly emergent penaeid shrimp disease which can cause 70–100% mortality in Penaeus vannamei and Penaeus monodon, and has resulted in enormous economic losses since its appearance. AHPND is caused by the specific strains of Vibrio parahaemolyticus that harbor the pVA1 plasmid and express PirAvp and PirBvp toxins. These two toxins have been reported to form a binary complex. When both are present, they lead to the death of shrimp epithelial cells in the hepatopancreas and cause the typical histological symptoms of AHPND. However, the binding mode of PirAvp and PirBvp has not yet been determined. Here, we used isothermal titration calorimetry (ITC) to measure the binding affinity of PirAvp and PirBvp. Since the dissociation constant (Kd = 7.33 ± 1.20 μM) was considered too low to form a sufficiently stable complex for X-ray crystallographic analysis, we used alternative methods to investigate PirAvp-PirBvp interaction, first by using gel filtration to evaluate the molecular weight of the PirAvp/PirBvp complex, and then by using cross-linking and hydrogen-deuterium exchange (HDX) mass spectrometry to further understand the interaction interface between PirAvp and PirBvp. Based on these results, we propose a heterotetrameric interaction model of this binary toxin complex. This model provides insight of how conformational changes might activate the PirBvp N-terminal pore-forming domain and should be helpful for devising effective anti-AHPND strategies in the future.


Introduction
Vibrio parahaemolyticus is a common halophilic Gram-negative bacterium that can be found in estuarine, marine and coastal environments. Recently, however, some new virulent strains of this opportunistic marine pathogen were identified as the causative agent of acute hepatopancreatic necrosis disease (AHPND) in shrimp. AHPND has a high mortality rate in shrimp (70-100% in Penaeus monodon and Penaeus vannamei), leading to catastrophic drops in shrimp production and enormous economic losses that impact the whole industry. Shrimp infected with AHPND show some readily observable symptoms, such as lethargy, an empty stomach and midgut, and a pale to white atrophied hepatopancreas [1], while the characteristic histological symptom of AHPND is the sloughing of HP tubule epithelial cells into the HP tubule lumens [1,2]. However, due to the absence of obvious bacterial colonies in the hepatopancreas tube lumens in the initial stage of acute infection of V. parahaemolyticus [1,3,4], Tran et al. proposed that AHPND symptoms were not induced by the bacteria themselves, but the toxins secreted by the bacteria [1]. Subsequent reverse gavage experiments further confirmed this hypothesis [1,5].
In addition to the specific strains of V. parahaemolyticus, several other Vibrio species such as V. harveyi, V. campbelli, V. owensii, and V. punensis were also found to cause AHPND [6][7][8][9]. These and other reports further showed that all these AHPND-causing pathogens harbor plasmids that contain the pirA and pirB toxin genes which are homologs of the Photorhabdus insect-related (Pir) binary toxins [6][7][8][9][10]. Durán-Avelar et al. even reported that a non-Vibrio bacterium, Microccocus luteus, also harbors the pirA and pirB toxin genes [11]. These two toxin proteins were confirmed to be the key factors that cause AHPND symptoms by Lee et al., who demonstrated that PirA vp /PirB vp are sufficient to induce the typical symptoms of AHPND by feeding shrimp with either the recombinant PirA vp /PirB vp proteins or with E. coli that expressed both PirA vp and PirB vp [10]. In addition to showing that PirA vp and PirB vp form a complex, we also used structural analysis (i.e., X-ray crystallography) to show that the assembled PirA vp and PirB vp structure was similar to that of the Bacillus thuringiensis Cry insecticidal toxins: the N-terminal and C-terminal of PirB vp correspond to the pore-forming domain I and the receptor-binding domain II of Cry protein, respectively, while PirA vp corresponds to Cry toxin domain III, which is the sugar-binding domain [10,12]. However, we failed to obtain the crystal of the PirA vp /PirB vp complex for subsequent structural analysis, and the interaction model between PirA vp and PirB vp therefore remained unclear. Here, we use alternative methods, such as isothermal titration calorimetry (ITC), gel filtration, cross-linking mass spectrometry, and hydrogen-deuterium exchange (HDX), to investigate the interface between PirA vp and PirB vp . We expect that with a better understanding of this interface, it will be possible to develop more effective strategies against AHPND in the future.

PirA vp and PirB vp Have a Low Binding Affinity
Since PirA vp and PirB vp need to bind to each other to achieve their cytotoxic activity [10], an understanding of their binding mode should be useful for designing an anti-AHPND strategy, for example by using a small compound, peptide or antibody to block the interfaces of PirA vp and PirB vp to prevent formation of the toxic complex. In our previous report, we confirmed that PirA vp could interact with PirB vp [10]. Unfortunately, we were unable to obtain any crystal for the PirA vp /PirB vp complex, and in fact we found crystals only for PirB vp ; neither PirA vp nor the complex was detected [10]. This result suggested that the complex is unstable. In the present study, to confirm this result, we used isothermal titration calorimetry (ITC) to determine the binding affinity of PirA vp and PirB vp and obtained a value of 7.33 ± 1.20 µM (Figure 1). Other thermodynamic parameters from the ITC assay are shown in Table 1. In general, this binding affinity is not good enough to maintain a stable complex for a long time, which might explain why we were unable to obtain a useful crystal of this complex for X-ray crystallography. In addition, the calculated binding stoichiometry ratio (N) was 0.74 ± 0.01, suggesting that some protein molecules may be inactive. For these reasons, we decided to use alternative methods to investigate the binding model for PirA vp and PirB vp . and obtained a value of 7.33 ± 1.20 μM (Figure 1). Other thermodynamic parameters from the ITC assay are shown in Table 1. In general, this binding affinity is not good enough to maintain a stable complex for a long time, which might explain why we were unable to obtain a useful crystal of this complex for X-ray crystallography. In addition, the calculated binding stoichiometry ratio (N) was 0.74 ± 0.01, suggesting that some protein molecules may be inactive. For these reasons, we decided to use alternative methods to investigate the binding model for PirA vp and PirB vp .   To determine the PirA vp /PirB vp binding ratio, we used gel filtration chromatography to evaluate the native molecular weights of PirA vp , PirB vp and the PirA vp /PirB vp complex. The observed molecular weights of PirA vp and PirB vp were 15.75 kDa and 56.12 kDa, respectively ( Figure 2 and Table 2). These values are close to their theoretical molecular weights and suggest that PirA vp and PirB vp both appear as monomers in the solution. The matching theoretical and evaluated molecular weights of PirA vp and PirB vp also indicate that the results of this gel filtration analysis are reliable. We also found that the observed molecular weight of the PirA vp /PirB vp complex (136.08 kDa) was similar to the theoretical molecular weight of the heterodimer/heterodimer interaction (132.59 kDa) ( Table 2), suggesting that PirA vp and PirB vp may interact with each other to form a heterotetramer. We note that, while we used a molar ratio of the input PirA vp and PirB vp of more than 1:1 (about 1.7:1), there were still unbound, free-form PirA vp and PirB vp monomers left over, as evidenced by the asymmetry of the PirA vp /PirB vp Figure 1. Determination of the binding affinity between PirA vp and PirB vp by an isothermal titration calorimetry (ITC) assay. The dissociation constant (K d ) between PirA vp and PirB vp was determined as 7.33 ± 1.20 µM. Other thermodynamic parameters for the PirA vp /PirB vp interaction are shown in Table 1. The data were collected from triplicate experiments. All three experiments produced very similar results; only a single experiment is shown in the Figure. To determine the PirA vp /PirB vp binding ratio, we used gel filtration chromatography to evaluate the native molecular weights of PirA vp , PirB vp and the PirA vp /PirB vp complex. The observed molecular weights of PirA vp and PirB vp were 15.75 kDa and 56.12 kDa, respectively ( Figure 2 and Table 2). These values are close to their theoretical molecular weights and suggest that PirA vp and PirB vp both appear as monomers in the solution. The matching theoretical and evaluated molecular weights of PirA vp and PirB vp also indicate that the results of this gel filtration analysis are reliable. We also found that the observed molecular weight of the PirA vp /PirB vp complex (136.08 kDa) was similar to the theoretical molecular weight of the heterodimer/heterodimer interaction (132.59 kDa) ( Table 2), suggesting that PirA vp and PirB vp may interact with each other to form a heterotetramer. We note that, while we used a molar ratio of the input PirA vp and PirB vp of more than 1:1 (about 1.7:1), there were still unbound, free-form PirA vp and PirB vp monomers left over, as evidenced by the asymmetry of the PirA vp /PirB vp complex peak (Figure 2A). This result was consistent with our ITC results, which also implied that some of the purified recombinant PirA vp and/or PirB vp failed to interact to form the PirA vp /PirB vp complex. Both experiments therefore suggest that some of the PirA vp and PirB vp might somehow have been rendered inactive. The presence of free-form PirA vp and PirB vp monomers might also have been due to the dissociation of the complex, because as noted above, the PirA vp /PirB vp complex is unstable and its components have a low binding affinity.    We further investigated the possible stoichiometry between PirA vp and PirB vp by using an alternative method. As shown in Figure S1A, two dilution series of known amounts of PirA vp and PirB vp were separated by SDS-PAGEs, and by using Image J software (https://imagej.nih.gov/ij/download.html), the intensities of the protein bands were quantified and used to produce standard curves. Serial dilutions of the complex that was collected from the gel filtration assay described above were separated by another SDS-PAGE ( Figure S1B), and the intensities of the protein bands were quantified using the same procedures. In Figure S1B, only Lane  We further investigated the possible stoichiometry between PirA vp and PirB vp by using an alternative method. As shown in Figure S1A were separated by SDS-PAGEs, and by using Image J software (https://imagej.nih.gov/ij/download.html), the intensities of the protein bands were quantified and used to produce standard curves. Serial dilutions of the complex that was collected from the gel filtration assay described above were separated by another SDS-PAGE ( Figure S1B), and the intensities of the protein bands were quantified using the same procedures. In Figure S1B, only Lane 5 contained amounts of PirA vp and PirB vp proteins that were located within the confidence intervals given by the standard curves. The protein quantities in Lane 5 were then used with the observed molecular weights of PirA vp and PirB vp ( Table 2) to calculate the number of PirA vp and PirB vp molecules, and the molar ratio of PirA vp and PirB vp was determined as 1.1 ( Figure S1C and Table S1). This binding stoichiometry was close to the 2:2 ratio proposed above.

Determination of the Interface between PirA vp and PirB vp Using Cross-Linking Coupled Mass Spectrometry Analysis
Cross-linking coupled mass spectrometry analysis is widely used as an alternative way to investigate the interaction between macromolecules [13][14][15][16][17][18], and we used this technique to study the interface between PirA vp and PirB vp . The crosslinker used in this assay is bis(sulfosuccinimidyl)suberate BS3, which is an amine-to-amine crosslinker with an arm length of 11.4 Å. BS3 crosslinks the amine groups on two lysines when the distance between them is shorter than BS3 s arm length. Since 11.4 Å is such a short distance, the BS3 crosslinked lysines must be very close to each other, and thus probably part of the interface of the two interacting proteins [16,17].
As shown in Figure 3, shifted protein bands of around 70 kDa, 85 kDa and 140 kDa (asterisks) were found in the lane of the crosslinked PirA vp and PirB vp . These protein bands were close to the theoretical molecular weights of the heterodimer (1 PirA vp /1 PirB vp , 66.31 kDa), heterotrimer (2 PirA vp / 1 PirB vp , 81.48 kDa) and heterotetramer (2 PirA vp /2 PirB vp , 132.59 kDa), respectively. This result also suggests that the PirA vp /PirB vp oligomers might be unstable, perhaps because of the low binding affinity. The major protein band (proposed heterodimer) was excised and subjected to in-gel digestion procedures using trypsin plus chymotrypsin. The digested peptides were further analyzed by a NanoLC-nanoESI-MS/MS (LTQ-Orbitrap Elite) coupled with MassMatrix software analysis. The result revealed that the residues Lys 67 and Lys 70 of PirA vp were crosslinked with Lys 394 of PirB vp ( Table 3) and implies that these lysine residues should be localized in the dimeric interface of PirA vp and PirB vp . molecules, and the molar ratio of PirA vp and PirB vp was determined as 1.1 ( Figure S1C and Table S1). This binding stoichiometry was close to the 2:2 ratio proposed above.

Determination of the Interface between PirA vp and PirB vp Using Cross-Linking Coupled Mass Spectrometry Analysis
Cross-linking coupled mass spectrometry analysis is widely used as an alternative way to investigate the interaction between macromolecules [13][14][15][16][17][18], and we used this technique to study the interface between PirA vp and PirB vp . The crosslinker used in this assay is bis(sulfosuccinimidyl)suberate BS3, which is an amine-to-amine crosslinker with an arm length of 11.4 Å. BS3 crosslinks the amine groups on two lysines when the distance between them is shorter than BS3′s arm length. Since 11.4 Å is such a short distance, the BS3 crosslinked lysines must be very close to each other, and thus probably part of the interface of the two interacting proteins [16,17].
As shown in Figure 3, shifted protein bands of around 70 kDa, 85 kDa and 140 kDa (asterisks) were found in the lane of the crosslinked PirA vp and PirB vp . These protein bands were close to the theoretical molecular weights of the heterodimer (1 PirA vp /1 PirB vp , 66.31 kDa), heterotrimer (2 PirA vp /1 PirB vp , 81.48 kDa) and heterotetramer (2 PirA vp /2 PirB vp , 132.59 kDa), respectively. This result also suggests that the PirA vp /PirB vp oligomers might be unstable, perhaps because of the low binding affinity. The major protein band (proposed heterodimer) was excised and subjected to in-gel digestion procedures using trypsin plus chymotrypsin. The digested peptides were further analyzed by a NanoLC-nanoESI-MS/MS (LTQ-Orbitrap Elite) coupled with MassMatrix software analysis. The result revealed that the residues Lys 67 and Lys 70 of PirA vp were crosslinked with Lys 394 of PirB vp ( Table 3) and implies that these lysine residues should be localized in the dimeric interface of PirA vp and PirB vp .     In addition to the cross-linking mass spectrometry analysis, we also used HDX mass spectrometry to investigate the interface between PirA vp and PirB vp . HDX has been used in structural biology research for nearly 30 years [19], and its principles and other information have been reviewed in detail [20]. The basic principle is that the closer the location of the hydrogen/deuterium to the surface of the molecule, the higher the exchange rate. Conversely, the exchange rate of interacting regions in a protein complex is lower because the participating residues are buried within the complex [20]. Thus, a comparison of the hydrogen-deuterium exchange rates of the PirA vp /PirB vp complex with the exchange rates of PirA vp and PirB vp alone should reveal which areas of PirA vp and PirB vp interact with each other or undergo structural changes during formation of the PirA vp /PirB vp complex. Since the exchange between hydrogen and deuterium was close to equilibrium in the later stages of the reaction ( Figure S2), we focused on analyzing the data at 10, 40, and 80 s after the addition of deuterium. Our criteria were as follows: (1) if two-thirds of these three time points showed more than a 1.4-fold difference in the hydrogen-deuterium exchange rates between the single protein and complex, we considered this region to be heavily involved in the PirA vp /PirB vp complex formation; (2) peptides showing a difference of between 1.2-1.4-fold at two or three time points were considered to be interacting regions closer to the surface; (3) peptides with less than 1.2-fold difference were considered not to be involved in the interaction. Results are shown in Table 4. In Figure 4, the interacting regions are colored blue, and the regions not showing significant differences in hydrogen-deuterium exchange rates are colored red. Figure 4A shows that there are two major regions of PirA vp that interact with PirB vp . This model is based on the fact that the peptide 52-TIQYQWGAPFMAGGWK(67)VAK(70)SHVVQRDET-79 from PirA vp showed decreased deuterium incorporation in PirA vp /PirB vp complex compared to stand alone PirA vp (Table 4A), suggesting that this region may be involved in the PirA vp /PirB vp interaction. As mentioned above, this region also includes the Lys 67 and Lys 70 of PirA vp that can be cross-linked with Lys 394 on PirB vp . In addition, the relatively small difference in exchange rate (~1.2X) implies that this region may be located on the edge of the interface between PirA vp and PirB vp , and may not be embedded deeply in the complex. Another PirA vp peptide, 15-WTVEPNGGVTEVDSKHTPIIPEVGRS-40, showed distinctly lower hydrogen-deuterium exchange rates (differences of more than 1.4X), suggesting that this region may also be involved in the interaction between PirA vp and PirB vp and may be embedded deep in the interior of the complex.  The putatively pore-forming domain in the N-terminal region that is thought to become exposed due to conformational changes after formation of the heterodimer is colored orange.

Proposed PirA vp /PirB vp Binding Model Using Cross-Linking Coupled Mass Spectrometry and HDX Analysis
We next used the information from cross-linking coupled mass spectrometry and HDX analysis to generate a PirA vp /PirB vp binding model. After using the Z-DOCK server (http://zdock.umassmed.edu/), to generate possible models based on the three cross-linked lysines, these candidate models were evaluated in terms of the HDX analysis. Figure 5A shows a proposed PirA vp /PirB vp heterodimer that was the best fit to these requirements. In this heterodimeric model, the β-sheet 2 of PirA vp (20-GGVTEDSKH-30) was predicted to interact to β-sheet 9 region of PirB vp (395-PSVRLQ-400) ( Figure 5B, left). Additionally, the β-sheet 5 of PirA vp region (66-WKVAKSHV-73) is close to the loop region between β-sheet 8 and 9 of PirB vp (390-ENSGKP-395), which suggests an interaction between them ( Figure 5B; right). Notably, in this proposed PirA vp /PirB vp heterodimer, there are still some available binding regions on PirA vp (i.e., the β-sheet 4-α-helix 1-β-sheet 5 regions of PirA vp ; 52-TIQYQWGAPFMAGGWKVAKSHVVQRDET-79, as determined by HDX analysis). These regions may be involved in higher orders of PirA vp /PirB vp oligomeric formation. In PirA vp , orange cylinders and yellow arrows represent α-helices and β-sheets, respectively. In PirB vp , magenta cylinders and cyan arrows represent the α-helices and β-sheets. The BS3-crosslinked lysines are colored green. The regions thought to be involved or not involved in the interaction are colored blue and red, respectively. (Right) The crystal structures of PirA vp and PirB vp . The putatively pore-forming domain in the N-terminal region that is thought to become exposed due to conformational changes after formation of the heterodimer is colored orange.
Meanwhile, the PirB vp -derived peptide 386-FVVGENSGK(394)PSVRLQL-401 also showed a decreased deuterium incorporation (~1.2X) after the complex formation. This region also contains the Lys 394 that can be crosslinked with PirA vp , which further supports its probable involvement in the interface between PirA vp and PirB vp . The significantly decreased hydrogen-deuterium exchange rates of other PirB vp -derived peptides (214-WADNDSYNNANQD-226, 290-DEIPQPLKPNM-300, 322-YNRVGRLKL-330, 409-MLADQEGSDKVAA-421 and 426-YELFHPDEF-434) in the PirA vp /PirB vp complex imply that these regions too may be involved in the complex formation and/or embedded in the depths of the complex. Most of these peptides are localized on the antiparallel β-sheets (β-8, 9, 10 and 11) found in the C-terminal domain of PirB vp , suggesting that this PirB vp domain mediates the major interaction to PirA vp . It is also worth noting that the hydrogen-deuterium exchange rate of peptide 116-TIENFGYAAAKDDYIGL-132 derived from the N-terminus of PirB vp was progressively increased after complex formation, suggesting that this region may be exposed after PirA vp /PirB vp interaction (Figure 4). Since the N-terminus of PirB vp was similar to the pore-forming domain I of Cry toxin [10,12], this exposed region (colored orange in Figure 4B) perhaps acts as a fusion peptide that participates in membrane insertion or pore formation on the host cell membrane.

Proposed PirA vp /PirB vp Binding Model Using Cross-Linking Coupled Mass Spectrometry and HDX Analysis
We next used the information from cross-linking coupled mass spectrometry and HDX analysis to generate a PirA vp /PirB vp binding model. After using the Z-DOCK server (http://zdock.umassmed.edu/), to generate possible models based on the three cross-linked lysines, these candidate models were evaluated in terms of the HDX analysis. Figure 5A shows a proposed PirA vp /PirB vp heterodimer that was the best fit to these requirements. In this heterodimeric model, the β-sheet 2 of PirA vp (20-GGVTEDSKH-30) was predicted to interact to β-sheet 9 region of PirB vp (395-PSVRLQ-400) ( Figure 5B, left). Additionally, the β-sheet 5 of PirA vp region (66-WKVAKSHV-73) is close to the loop region between β-sheet 8 and 9 of PirB vp (390-ENSGKP-395), which suggests an interaction between them ( Figure 5B; right). Notably, in this proposed PirA vp /PirB vp heterodimer, there are still some available binding regions on PirA vp (i.e., the β-sheet 4-α-helix 1-β-sheet 5 regions of PirA vp ; 52-TIQYQWGAPFMAGGWKVAKSHVVQRDET-79, as determined by HDX analysis). These regions may be involved in higher orders of PirA vp /PirB vp oligomeric formation.
Since the structural topology of PirA vp /PirB vp complex is very similar to Cry toxins [10], we speculate that PirA vp /PirB vp toxin might use a similar mechanism to damage host cells. For example, during the pore-formation by Cry 1A toxin, the GalNAc sugar on the aminopeptidaseN (APN) receptor is firstly recognized by Cry domain III, and the receptor is bound by Cry domain II [21][22][23]. Subsequently, this receptor-bound Cry toxin undergoes a proteolytic cleavage which induces the oligomerization of Cry toxin, and finally, the toxin inserts to cell membrane and forms a pore on the membrane [21][22][23]. According to the proposed heterotetramer structure, the PirA vp /PirB vp heterodimer may form a heterotetramer through PirA vp -PirA vp interaction ( Figure 5C). The other side of PirA vp may participate in receptor recognition and binding. After the receptor binding, and possibly after the PirA vp /PirB vp complex forms a higher-order oligomer, PirB vp may be pulled toward the cell membrane, where it inserts into the membrane using its α-helix to efficiently form a transmembrane pore (Figure 6). We noted above that the low binding affinity between PirA vp and PirB vp may directly affect the stability of the heterotetramer. However, the stability may be improved after the PirA vp /PirB vp complex recognizes and interacts with its receptor. In addition, all of our experiments were performed in the absence of any membranes, and it is possible that, as suggested above, after the PirA vp /PirB vp complexes undergo receptor recognition/binding with an actual cell membrane, the heterotetramer might form a larger oligomer and its subunit stoichiometry may even be different from that of the free complex.  In the gel filtration analysis (Figure 2), we found that PirA vp and PirB vp form a tetramer in solution. The proposed PirA vp /PirB vp heterodimer was subsequently used to predict their tetrameric conformation. A predicted PirA vp /PirB vp heterotetramer that uses a similar docking strategy is shown in Figure 5C. In the resulting PirA vp /PirB vp heterotetramer, two PirA vp /PirB vp heterodimers bind to each other by using the β-sheet 4-α-helix 1-β-sheet 5 regions of PirA vp (52-TIQYQWGAPFMAGGWKVAKSHVVQRDET-79) ( Figure 5D).
Since the structural topology of PirA vp /PirB vp complex is very similar to Cry toxins [10], we speculate that PirA vp /PirB vp toxin might use a similar mechanism to damage host cells. For example, during the pore-formation by Cry 1A toxin, the GalNAc sugar on the aminopeptidaseN (APN) receptor is firstly recognized by Cry domain III, and the receptor is bound by Cry domain II [21][22][23]. Subsequently, this receptor-bound Cry toxin undergoes a proteolytic cleavage which induces the oligomerization of Cry toxin, and finally, the toxin inserts to cell membrane and forms a pore on the membrane [21][22][23]. According to the proposed heterotetramer structure, the PirA vp /PirB vp heterodimer may form a heterotetramer through PirA vp -PirA vp interaction ( Figure 5C). The other side of PirA vp may participate in receptor recognition and binding. After the receptor binding, and possibly after the PirA vp /PirB vp complex forms a higher-order oligomer, PirB vp may be pulled toward the cell membrane, where it inserts into the membrane using its α-helix to efficiently form a transmembrane pore ( Figure 6). We noted above that the low binding affinity between PirA vp and PirB vp may directly affect the stability of the heterotetramer. However, the stability may be improved after the PirA vp /PirB vp complex recognizes and interacts with its receptor. In addition, all of our experiments were performed in the absence of any membranes, and it is possible that, as suggested above, after the PirA vp /PirB vp complexes undergo receptor recognition/binding with an actual cell membrane, the heterotetramer might form a larger oligomer and its subunit stoichiometry may even be different from that of the free complex. Step 1), after which the newly-exposed N-terminus region of PirB vp (orange) is pulled toward the cell membrane ( Step 2) where it inserts into the membrane using its α-helix and initiates the process of pore formation.
Since both PirA vp and PirB vp have been reported as secreted proteins [10,24], they are therefore good targets for anti-AHPND drug design. For pore-forming toxins, structural insights into the protein components can provide useful information on conformation rearrangement, the binding interface between receptor and ligand, and the oligomerization of the toxin [25]. Depending on the structural information, different strategies can be used to prevent pore formation. For example, in developing a drug to treat infection with multidrug-resistant strains of Staphylococcus aureus, the interaction of natural compounds (e.g., oroxylin A, oroxin A and oroxin B) with S. aureus Hla may prevent the loop transition during the pore formation, and therefore inhibit the haemolytic activity of Hla [26,27]; an antibody against the receptor binding domains of S. aureus Hla also prevents the toxin from recognizing the cell membrane and blocks its binding with the receptor [28]. Meanwhile, as part of a strategy against necrotic enteritis caused by Clostridium perfringens, recombinant toxoids can be used as vaccines that trigger immune responses, and structural information on the toxin or pore architecture will be helpful in the design of site-directed mutations [29,30]. In the case of PirA vp /PirB vp , based on the interface and heterotetrameric binding model proposed here, another possible option is to use in silico screening to identify small compounds that may be able to block the Figure 6. Schematic representation of the proposed binding mechanism of the heterotetrameric PirA vp /PirB vp toxin with its receptor. The PirA vp /PirB vp heterotetramer first uses PirA vp to recognize and bind with a receptor on the host cell membrane (Step 1), after which the newly-exposed N-terminus region of PirB vp (orange) is pulled toward the cell membrane (Step 2) where it inserts into the membrane using its α-helix and initiates the process of pore formation.
Since both PirA vp and PirB vp have been reported as secreted proteins [10,24], they are therefore good targets for anti-AHPND drug design. For pore-forming toxins, structural insights into the protein components can provide useful information on conformation rearrangement, the binding interface between receptor and ligand, and the oligomerization of the toxin [25]. Depending on the structural information, different strategies can be used to prevent pore formation. For example, in developing a drug to treat infection with multidrug-resistant strains of Staphylococcus aureus, the interaction of natural compounds (e.g., oroxylin A, oroxin A and oroxin B) with S. aureus Hla may prevent the loop transition during the pore formation, and therefore inhibit the haemolytic activity of Hla [26,27]; an antibody against the receptor binding domains of S. aureus Hla also prevents the toxin from recognizing the cell membrane and blocks its binding with the receptor [28]. Meanwhile, as part of a strategy against necrotic enteritis caused by Clostridium perfringens, recombinant toxoids can be used as vaccines that trigger immune responses, and structural information on the toxin or pore architecture will be helpful in the design of site-directed mutations [29,30]. In the case of PirA vp /PirB vp , based on the interface and heterotetrameric binding model proposed here, another possible option is to use in silico screening to identify small compounds that may be able to block the interaction of PirA vp and PirB vp . Further, the structural biology approach used here should be useful not only for designing anti-AHPND strategies in the future, but also as a platform to study other Pir toxins and provide structural insights that can be applied to the control of other pests or vector mosquitoes.

Construction and Recombinant Protein Purification
The coding regions of pirA vp and pirB vp (accession no. KP324996, regions 35028 to 35363 and 33699 to 35015, respectively) were constructed into pET16b and pET21b vectors, and the resulting recombinant plasmids were named pirA vp -pET16b and pirB vp -pET21b, respectively. These plasmids were then transformed into E. coli BL21(DE3) strain to express the N-terminal tagged His 10 -PirA vp and the C-terminal tagged His 6 -PirB vp . The cultures of pirA vp -pET16b or pirB vp -pET21b transformed BL21(DE3) were incubated overnight and then subcultured into fresh LB medium with dilution ratios of 1:250 and 1:50, respectively. The cultures were then grown at 37 • C until the OD 600 reached to 0.4. To induce protein expression, IPTG was added to a final concentration of 1 mM and the cultures were incubated at 16 • C for 20 h. The cells were then collected and subjected to protein purification procedures. For the N-terminal tagged His 10 -PirA vp , cells were resuspended with lysis buffer (20 mM Tris, pH 7.4, 100 mM NaCl, 20 mM imidazole, 100 µg/mL lysozyme, 10 µg/mL DNase I, 1 mM PMSF) and homogenized with sonication. After removing the cell debris, the supernatant was filtered through a 0.45 µm filter and loaded onto a 5 mL HisTrap HP column (GE Healthcare, Chicago, IL, USA). The column was washed with wash buffer (20 mM Tris, pH 7.4, 100 mM NaCl, 20 mM imidazole), and the His-tagged protein was eluted with a imidazole gradient (from 20 mM to 500 mM). The eluted His 10 -PirA vp was further purified with a Superdex 75 column using the gel filtration buffer (30 mM Tris, pH 7.4, 100 mM NaCl, 1 mM DTT, 5% glycerol). The C-terminal tagged His 6 -PirB vp protein was purified using the same procedures, except that the buffers were as follows: lysis buffer (20 mM Tris, pH 8.0, 300 mM NaCl, 20 mM imidazole, 100 µg/mL lysozyme, 10 µg/mL DNase I, 1 mM PMSF), wash buffer (20 mM Tris, pH 8.0, 300 mM NaCl, 20 mM imidazole), elution buffer (20 mM Tris, pH 8.0, 300 mM NaCl, 20-500 mM imidazole), and gel filtration buffer (20 mM Tris, pH 8.0, 300 mM NaCl). Protein concentrations were determined by the Bradford method.

Determination of the Binding Affinity between PirA vp and PirB vp by Isothermal Titration Calorimetry (ITC)
After the N-terminal tagged His 10 -PirA vp and the C-terminal tagged His 6 -PirB vp were both dialyzed against the same reaction buffer (30 mM Tris, pH8.0, 100 mM NaCl, 5% glycerol, 1mM DTT), they were respectively loaded into the sample cell and syringe of an iTC200 (GE Healthcare) instrument. For the single proteins, N-terminal tagged His 10 -PirA vp or C-terminal tagged His 6 -PirB vp were separately diluted with the reaction buffer (30 mM Tris, pH8.0, 100 mM NaCl, 5% glycerol, 1mM DTT) to a final concentrations of 65.83 µM (1 mg/mL) and 39.12 µM (2 mg/mL), respectively. The proteins were then loaded onto an Enrich SEC 650 10 × 300 column (BioRad, Hercules, CA, USA), eluted with reaction buffer, and a continuous series of 0.5 mL eluted samples were collected. For the PirA vp /PirB vp complex, N-terminal tagged His 10 -PirA vp and C-terminal tagged His 6 -PirB vp were mixed in reaction buffer to final concentrations of 65.83 µM (1 mg/mL) and 39.12 µM (2 mg/mL), respectively, incubated at 25 • C for 15 min and analyzed with the same column under the same conditions. The fractions corresponding to the peaks in the diagrams were analyzed with SDS-PAGE. To create a standard curve, native proteins from two gel filtration calibration kits (GE Healthcare), including aldolase (A; 158 kDa), conalbumin (C; 75 kDa), ovalbumin (O; 43 kDa), carbonic anhydrate (CA; 29 kDa), and ribonuclease A (R; 13.7 kDa) were also analyzed with the same column, and the logarithm of molecular weight (log MW) of each protein was plotted against its Kav using the calculation: where Ve is the elution volume, Vo is the column void volume (determined using blue dextran 2000), and Vt is the total column bed volume (24 mL for the Enrich SEC 650 10 × 300 column).

Determination of the Binding Stoichiometry of PirA vp and PirB vp by Densitometric Analysis
To achieve this analysis, we first collected the PirA vp /PirB vp complex from the gel filtration fractions that eluted at 12.5 to 13 mL to avoid any possible contamination of the PirB vp monomer. The complex sample was then serially diluted and separated in SDS-PAGE. Different quantities of PirA vp (1-6 µg) and PirB vp (2-8 µg) were also separated by SDS-PAGE. Computer software Image J (https://imagej.nih.gov/ij/download.html) was used to quantify the signal intensities of all of the protein bands. Standard curves were generated for PirA vp and PirB vp and used to calculate the amounts of PirA vp and PirB vp that dissociated from the PirA vp /PirB vp complex. Only serially diluted complex samples for which the amount of protein of both PirA vp and PirB vp were located within the confidence intervals were used for these calculations. Finally, the amounts of PirA vp and PirB vp were divided by their observed molecular weights (as obtained from the gel filtration analysis) to give the molar ratio of PirA vp and PirB vp in the complex.

Cross-Linking Coupled Mass Spectrometry Analysis of PirA vp and PirB vp
Before being subjected to a cross-linking reaction, the recombinant PirA vp and PirB vp were dialyzed against 1X PBS to remove Tris, the presence of which can inhibit the activity of bissulfosuccinimidyl suberate (BS3). PirA vp was then mixed with PirB vp in PBS at a concentration of 13.2 µM each. To serve as a control, PirA vp and PirB vp were also diluted with PBS separately. After incubating for 15 min at 25 • C, BS3 was added to the mixtures to a final concentration of 1 mM. The mixtures were then incubated at 25 • C for an additional 60 min and separated in SDS-PAGE. As loading controls, the PirA vp , PirB vp and PirA vp + PirB vp were incubated without BS3 and separated with the same SDS-PAGE.
For mass spectrometry analysis, the shifted bands of the crosslinked PirA vp and PirB vp were excised from the SDS-PAGE and digested with trypsin coupled with chymotrypsin in accordance with standard in-gel digestion procedures. The digested peptides were then analyzed with a NanoLC-nanoESI-MS/MS (LTQ-Orbitrap Elite, Thermo Fisher Scientific, Waltham, MA, USA) using the standard protocol of the Common Mass Spectrometry Facilities of the Institute of Biological Chemistry at Academia Sinica [31,32] and subjected to data analysis using the Massmatrix software [33].

Hydrogen-Deuterium Exchange (HDX) Mass Spectrometry Analysis of PirA vp and PirB vp
To initiate the hydrogen-deuterium exchange, the recombinant PirA vp , PirB vp (15 pmol each) and PirA vp /PirB vp protein complex (15 pmol: 15pmol) were separately diluted in the exchange buffer (99.9% D 2 O in PBS, pH 7.4) at a ratio of 1:10 at room temperature. To quench the HD exchange, 3.5 µL sample (1.5 pmol of target protein) was mixed with 6.5 µL pre-chilled quenching buffer [to a final concentration of 1.5 M guanidine hydrochloride, 150 mM tris(2-carboxyethyl)phosphine, and 0.8% formic acid] at 10, 40, 80, 180, 600, 1800, and 3600 s after deuterium was added. The mixture was immediately loaded onto a pepsin column for online digestion, and the digested peptides were then desalted using a reverse-phase column (Zorbax 300SB-C18, 0.3 × 5 mm; Agilent Technologies, Wilmington, DE, USdA). The desalted peptides were further separated on a customized HydroRP column using a linear gradient of 8-95% HPLC buffer (99.9% acetonitrile, 0.1% formic acid, 0.025% trifluoroacetic acid) for 14.5 min with a flow rate of 0.5 µL/min. The LC apparatus was coupled with a 2D linear ion trap mass spectrometer (Orbitrap Classic; Thermo Fisher Scientific, Waltham, MA, USA) and the full-scan MS was performed in the Orbitrap over a m/z range of 350 to 1600 Da and a resolution of 60,000 at m/z 400. The ion signal of [Si(CH3) 2 O]6H + at m/z 536.165365 was served as the lock mass for internal calibration. The peptides were ionized at an electrospray voltage of 1.8 kV, and the temperature of the capillary was set to 200 • C. To control the accumulated time or ions, the automatic gain control of MS and MS/MS were 1,000 ms (full scan) and 150 ms (MS/MS), or 1 × 10 6 ions (full scan) and 2 × 10 3 ions (MS/MS), respectively.

Peptide Identification and HDX Data Analysis
The computer software Proteome Discoverer (Version 1.4, Thermo Fisher Scientific, Waltham, MA, USA, 2012) was used for peptide identification, and the SEQUEST search engine was used for the MS/MS spectra searching against the single protein database (i.e., PirA vp or PirB vp ). For peptide identification, the mass tolerance was 10 ppm for intact peptide masses and 0.5 Da for CID fragment ions. Peptide-spectrum matches (PSM) were then filtered based on high confidence and a peptide identification search engine rank of 1 to ensure an overall false discovery rate below 0.01. For HDX profile analysis, the peptide identification template was first generated based on the LC-MS/MS result of target protein identification. The template was then preloaded in ExMS module installed in the MATLAB environment. To calculate the deuterium atom number in each peptide, the HDX MS spectra were loaded and analyzed. The result was then presented as an average value of deuterium incorporation based on two independent experiments.

Molecular Docking Analysis of PirA vp /PirB vp Complex
To identify the binding mode of PirA vp and PirB vp , we performed a protein-protein docking analysis. First, potential binding poses were generated by uploading the crystal structures of the two proteins to the ZDOCK server (http://zdock.umassmed.edu/) [34]. A PirA vp /PirB vp complex that satisfied the cross-linking distance criteria from the predicted poses was selected as the binding mode. Next, the complex structure was uploaded to the ZDOCK server to generate potential binding poses for the PirA vp /PirB vp heterodimer complexes. Finally, the complex structure was evaluated by the HDX results: the binding regions where the hydroge-deuterium exchange rate became lower after PirA vp /PirB vp complex formation should be found inside the complex, while regions where the exchange rate was unchanged should not be involved in the interaction. The resulting pdb files of PirA vp /PirB vp heterodimer and heterotetramer were included in the supporting files for a reference.