Different roles of conserved tyrosine residues of the acylated domains in folding and activity of RTX toxins

Pore-forming repeats in toxins (RTX) are key virulence factors of many Gram-negative pathogens. We have recently shown that the aromatic side chain of the conserved tyrosine residue 940 within the acylated segment of the RTX adenylate cyclase toxin-hemolysin (CyaA, ACT or AC-Hly) plays a key role in target cell membrane interaction of the toxin. Therefore, we used a truncated CyaA-derived RTX719 construct to analyze the impact of Y940 substitutions on functional folding of the acylated segment of CyaA. Size exclusion chromatography combined with CD spectroscopy revealed that replacement of the aromatic side chain of Y940 by the side chains of alanine or proline residues disrupted the calcium-dependent folding of RTX719 and led to self-aggregation of the otherwise soluble and monomeric protein. Intriguingly, corresponding alanine substitutions of the conserved Y642, Y643 and Y639 residues in the homologous RtxA, HlyA and ApxIA hemolysins from Kingella kingae, Escherichia coli and Actinobacillus pleuropneumoniae, affected the membrane insertion, pore-forming (hemolytic) and cytotoxic capacities of these toxins only marginally. Activities of these toxins were impaired only upon replacement of the conserved tyrosines by proline residues. It appears, hence, that the critical role of the aromatic side chain of the Y940 residue is highly specific for the functional folding of the acylated domain of CyaA and determines its capacity to penetrate target cell membrane.

Except for the presence of the unique ' AC-to-Hly-linking segment' , the domain structure of the Hly moiety of CyaA exemplifies the general domain organization of other known RTX hemolysins/cytolysins that share a fair extent of sequence homology and differ primarily in the number of repeat sequences forming their RTX domains 1 .

Results
The aromatic ring of the tyrosine residue 940 plays a crucial role in the biological activity of Bordetella CyaA toxin. We have previously observed that alanine or proline substitutions of the tyrosine residue 940 (Y940) in the acylated domain of CyaA did not interfere with the CyaC-mediated posttranslational acylation of the toxin, but still ablated the capacity of recombinant CyaA toxin to penetrate target cell membrane and deliver its AC domain into cells 12 . It was thus important to exclude that this loss of toxin activity might be an artifact caused by extraction of the recombinant CyaA proteins from E. coli-produced inclusion bodies by denaturating solutions of 8 M urea and subsequent calcium-driven refolding of the recombinant toxin variants in target cells suspensions. Therefore, we examined whether the Y940 substitutions will affect also the activities of CyaA toxin variants secreted by Bordetella pertussis and B. bronchiseptica. Towards this aim, mutations encoding the various Y940 substitutions were introduced by allelic exchange into the cyaA genes on B. pertussis and B. bronchiseptica chromosomes and the corresponding variants of secreted CyaA toxins were extracted from bacterial cell surface and used in toxin activity assays. Both B. pertussis and B. bronchiseptica intact CyaAs showed a similar capacity to deliver the AC domain across the plasma membrane of sheep erythrocytes and human THP-1 monocytes (Supplementary Figure S1). As further shown in Fig. 1a,b, the wild-type bacteria producing intact CyaA, or the mutant secreting the CyaA-Y940F toxin variant, exhibited a hemolytic phenotype on Bordet-Gengou (BG) agar plates. In contrast, the B. pertussis and B. bronchiseptica mutants secreting the CyaA-Y940A or CyaA-Y940P toxins formed non-hemolytic colonies indicative of loss of hemolytic activity the Y940A and Y940P CyaAs. Indeed, when the respective secreted toxins were extracted from the surface of producing bacteria, only the CyaA and the CyaA-Y940F toxin variant were able to bind and penetrate erythrocytes or CR3expressing THP-1 macrophages in the in vitro activity assay and the CyaA-Y940A and CyaA-Y940P proteins were inactive (Fig. 1c-f). Finally, the loss of biological activity of CyaA upon Y940A and Y940P substitutions was examined by in vivo intranasal infection experiments in Balb/cByJ mice. As shown in Table 1, the mutant B. pertussis strains secreting the CyaA Y940A and CyaA Y940P toxin variants exhibited a decreased virulence with about an order of magnitude higher LD 50 value (lethal dose for 50% of infected animals) than CyaA or CyaA-Y940F -secreting B. pertussis strains. All these results thus demonstrate that the presence of the aromatic ring as the side chain of residue 940 of the CyaA protein is critical for the cytotoxic activity of CyaA on target cells in vitro and for its biological function as a key virulence factor of B. pertussis in vivo in the mouse model of lung infection.   www.nature.com/scientificreports/ The aromatic side chain of the tyrosine residue 940 is essential for folding of the acylated domain of CyaA. We hypothesized that substitutions of Y940 by alanine or proline residues affects the folding and formation of a functionally important structure in the CyaA molecule. To test this hypothesis and analyze the structural consequences of Y940 substitution, we constructed a truncated CyaA-derived polypeptide RTX719 (Fig. 2a), in which the fatty acyl-modified segment of CyaA comprising the Y940 residue was fused to a truncated RTX domain capable of vectorial calcium-driven folding 22 . Subsequently, the Y940A and Y940P substitutions were introduced into the RTX719 construct, the proteins were produced in E. coli cells in the presence of the CyaC acyltransferase and purified close to homogeneity from urea extracts by chromatography on DEAE-Sepharose (Supplementary Figure S2). The calcium-dependent folding of the RTX719 variants was then analyzed by circular dichroism (CD) spectroscopy. The urea-denatured proteins were refolded in the Ca 2+ -free buffer and titrated by stepwise addition of Ca 2+ ions. The far-UV CD spectra revealed that the RTX719 construct as well as its Y940A and Y940P variants undergo a Ca 2+ -dependent structural transition from an unfolded to a folded conformation that is characterized by a prominent negative peak in the spectra at 218 nm, which corresponds to a parallel β-roll structure (Fig. 2b). Intriguingly, the Ca 2+ -induced assembly of the Y940A and Y940P mutants was initiated and reached completion at lower Ca 2+ ion concentrations, as that needed for completion of folding of the intact RTX719 construct (Fig. 2b, Supplementary Figure S3). Moreover, the on-column refolding characteristics of the RTX719 variants during Superdex HR200 gel permeation chromatography differed substantially. As documented in Fig. 2c, a major fraction of the on-column refolded RTX719 eluted as monomeric protein with a retention time of 32 min, while the Y940A and Y940P proteins formed predominantly oligomers eluting with retention time of 25 min. Moreover, the far-UV CD spectrum of the eluted RTX719 monomers exhibited a single negative peak at 218 nm typical of the β-stranded protein skeletons (Fig. 2d). In contrast, the spectra of the monomeric forms of the Y940A and Y940P protein variants eluted in the minor fraction at 32 min (black arrows over chromatograms in Fig. 2c and black curves in Fig. 2d) exhibited two negative peaks in the spectra at 206 and 218 nm, revealing the presence of a mixture of α-helices and β-sheets in their structures. Taken together, these data clearly indicated that the conserved tyrosine 940 residue plays a key role in Ca 2+ -induced folding of the acylated domain of CyaA and its replacement by a non-aromatic residue results in misfolding and aggregation of the RTX719 construct.
Substitutions of the conserved tyrosine residues in the acylated segments of other RTX hemolysins do not affect their posttranslational acylation but proline substitutions impair their toxin activites. The Y940 residue is located in a predicted β-strand structure within the acylated segment of CyaA and appears to be highly conserved within the homologous segments of other pore-forming RTX hemolysins (Fig. 3). To analyze whether the conserved tyrosine residues might play a role also in the functional folding and cytotoxic activities of other hemolysins, we replaced the Y642 residue of RtxA from Kingella kingae, the Y643 residue of HlyA from Escherichia coli, and the Y639 residue of ApxIA from Actinobacillus pleuropneumoniae with phenylalanine, alanine, and proline residues, respectively. The toxin variants were produced in E. coli BL-21 cells and purified close to homogeneity from urea extracts by Ni-NTA chromatography using the purpose-introduced 6xHis affinity purification tags. As verified by tandem mass spectrometry (Supplementary  Table S1), the substitutions of the conserved tyrosine residues did not affect the ability of the co-expressed cognate acyltransferase RtxC, HlyC, and ApxIC to recognize and posttranslationally acylate the respective segments of proRtxA, proHlyA, and proApxIA, which were quantitatively modified by myristoylation and hydroxymyristoylation on the corresponding lysine residues. As shown in Fig. 4a, the HlyA Y643A and HlyA Y643F variants exhibited a near intact hemolytic activity on sheep erythrocytes, whereas the activity of the HlyA Y643P construct was strongly reduced over a wide range of toxin concentrations. Similarly, the cytotoxicity of the HlyA Y643P variant towards LFA-1-positive THP-1 cells was markedly reduced and its cytotoxic effect was observed only at the highest concentration used, whereas the cytotoxic capacity of the HlyA Y643A and HlyA Y643F variants was similar to that of intact HlyA toxin (Fig. 4b). Consistent with this, the HlyA, HlyA Y643A, and HlyA Y643F proteins exhibited similar overall membrane activity on artificial lipid bilayers prepared from crude asolectin, whereas the HlyA Y643P variant formed pores with a significantly lower overall membrane activity than intact HlyA (Fig. 4c). As shown in Fig. 4d,e, the porecharacteristics, such as mean single pore conductance and lifetime of pores formed by all three HlyA mutants were comparable to those of intact HlyA. These data indicated that the presence of the aromatic ring in the side chain of the conserved Y643 residue of HlyA was as such not essential for the formation of the α-hemolysin pores, since the HlyA Y643A variant and the intact HlyA toxin exhibited comparable pore-forming membrane activities. www.nature.com/scientificreports/ We further analyzed whether the low pore-forming (hemolytic) and cytotoxic potency of the HlyA Y643P variant was due to a reduced membrane-binding capacity. Towards this aim, we prepared a series of HlyC-activated chimeric molecules in which the full-length HlyA (residues 1-1024) carrying the substitutions Y643A, Y643F, or Y643P was fused to the AC domain and the AC-to-Hly linker segment of CyaA (N-terminal residues 1-501), (Fig. 4f). The highly active N-terminal AC enzyme domain in the hybrid proteins then allowed the quantification of the specific capacity of the molecules to associate tightly with the cell membrane. As shown in Fig. 4g, the www.nature.com/scientificreports/ cell binding capacity of CyaA 1-501 /HlyA 1-1024 with the Y643F and Y643A substitutions was almost indistinguishable from the cell binding capacity of the intact chimera. In contrast, the cell binding capacity of the CyaA 1-501 / HlyA 1-1024 Y643P construct was reduced by ~ 70% compared to the intact chimera. These data strongly suggest that the low pore-forming and cytolytic activity of the HlyA Y643P mutant was due to the reduced membrane association capacity of the construct.
A tyrosine residue pair plays a role in membrane penetration of HlyA. In contrast to CyaA, where the conserved Y940 is adjacent to a hydrophilic S939 residue, a pair of hydrophobic aromatic residues is present in the acylated segment of HlyA (Fig. 3b). To analyze whether the adjacent Y642 residue synergizes with the conserved Y643 residue in maximizing the lytic activity of HlyA on target cells, we produced HlyA-derived constructs carrying both tyrosine residues substituted by a pair of alanine (HlyA Y642A + Y643A) or proline (HlyA Y642P + Y643P) residues, and a further construct had the Y642 replaced by a proline residue (HlyA Y642P). As shown in Fig. 5, the hemolytic and cytotoxic activities of intact HlyA and of the doubly substituted HlyA Y642A + Y643A variant were similar, suggesting that there was no functional impact of a combined tyrosine-toalanine substitution at the positions 642 and 643 of HlyA. However, the hemolytic and cytotoxic activities of the HlyA Y642P protein were reduced over a range of toxin concentrations (Fig. 5). Moreover, the Y642P + Y643P double substitution completely abolished the cytolytic activity of the fully acylated HlyA Y642P + Y643P protein (Supplementary Table S1) on both erythrocytes and THP-1 macrophages (Fig. 5). In conclusion, the nil effect of the double Y642A + Y643A substitutions on HlyA cytotoxic activities strongly suggests that the aromatic side chains of Y642 and Y643 are not critical for membrane insertion and subsequent lytic activity of HlyA. Intriguingly, the complete loss of toxin activity occurred only upon a double Y642P + Y643P substitution, presumably destroying the local structure of the HlyA segment.
A single substitution of the tyrosines does not affect the formation of secondary structures of the acylated segment of HlyA. To assess the structural impact of tyrosine residue substitutions in the acylated domain of HlyA, we constructed a truncated HlyA-derived polypeptide carrying the C-terminal residues 419-1024 (HlyA419, Fig. 6a). Subsequently, the Y642P, Y643P and Y643A substitutions were introduced into the HlyA419 construct and the corresponding proteins were produced in E. coli in the presence of HlyC acyltransferase and purified from the urea extracts on Ni-NTA Sepharose (Supplementary Figure S4). Far-UV CD spectra of the Hly-derived constructs revealed that all four proteins undergo a similar Ca 2+ -dependent structural change from disordered conformation to β-roll secondary structures (Fig. 6b). These data show that in contrast to a similar CyaA719 construct, the substitutions of the tyrosine residue of the acylated segment of www.nature.com/scientificreports/ HlyA with alanine or proline residue do not affect any importantly the Ca 2+ -dependent formation of secondary structures in the truncated HlyA419 fragment.
Proline, but not alanine substitutions of the conserved tyrosine residues strongly impair the hemolytic activity of the RtxA and ApxIA hemolysins. Finally, the hemolytic activity of the RtxA and ApxIA hemolysins bearing susbtitutions of the conserved tyrosine residues of the acylated domains were analyzed. As shown in Fig. 7a, while the Y642F variant of RtxA was fully active and the Y642A substitution affected the hemolytic activity of RtxA only modestly, the capacity of the RtxA Y642P construct to lyse erythrocytes was strongly impaired. Similarly, the ApxIA Y639P construct was unable to lyse erythrocytes even at high toxin concentrations, while the hemolytic activity of the ApxIA Y639A variant was reduced by a factor of ~ 2 compared to the intact toxin or its ApxIA Y639F variant (Fig. 7b).
In summary, these data indicate that the aromatic side chain of the conserved Y642 and Y639 residue is important for the full hemolytic activity of RtxA and of ApxIA hemolysins on sheep erythrocytes.

Discussion
We show here that the aromatic ring of the side chains of the conserved tyrosine residues located in the acylated segments of four RTX toxins play a different roles in their biological activities. While replacement of the aromatic ring of Y940 residue of CyaA by the alanine or proline residues led to self-aggregation of the truncated RTX719 protein and loss of toxin activities of the full-length CyaA, the elimination of the tyrosine residue at position 643 of HlyA (Y643A) did not significantly affect the folding of the acylated segment and subsequent cytotoxic activity of the α-hemolysin. Interestingly, for the quite homologous RtxA and ApxIA hemolysins, the presence of the aromatic ring of the conserved tyrosine residues were still required for a full hemolytic activity of the RtxA (Y642A) and of ApxIA (Y639A) toxins.
First, we re-examined the toxin activities of the CyaA variants secreted by corresponding mutant strains of the closely related B. pertussis and B. bronchiseptica This was important to do, as CyaA is translocated from the cytosol of Bordetella directly into the external medium through a T1SS channel-tunnel assembly spanning the bacterial envelope and undergoes a vectorial co-secretional folding initiated by the first-emerging C-terminal folding scaffold upon exit from the bacterial cell due to binding of extracellular calcium ions 19 . It was thus crucial to verify that also such excreted and vectorially folded molecules of CyaA bearing the Y940A and Y940P substitutions will be devoid of membrane penetrating activity, like the recombinant forms of the toxin produced and acylated in E. coli cells. Indeed, as shown in Fig. 1, this was the case for the corresponding mutant forms of CyaA secreted by both Bordetella species.
Recently, Fukui-Miyazaki and coworkers presented that B. bronchiseptica CyaA, in contrast to B. pertussis CyaA, exhibits weak cAMP intoxication of nucleated cells due to phosphorylation of the S375 residue that is bound by the host factor 14-3-3 and results in abrogation of the AC penetration activity 53 . Working with the same human THP-1 cell line and parental B. bronchiseptica RB50 strain as Fukui-Miyazaki et al., we were unable to reproduce the reported striking difference in the specific capacity of B. pertussis and B. bronchiseptica CyaA proteins to increase cAMP in target cell cytosol and we found that the two proteins exhibited comparable cAMPelevating activity in macrophage cells, as previously also reported by Henderson and colleagues 54 . Moreover, using LC FT-ICR-MS analysis, we checked the amino acid residues at position 375 of purified CyaAs by peptide mass mapping (Supplementary Figure S5) and found that B. pertussis CyaA carried F375, whereas B. bronchiseptica CyaA carried S375, as previously described 53 . Our mass spectrometry analysis also identified other substitutions in the CyaA sequences, namely at positions 370, 800, 808, 910, and 978, located in the N-terminal part of B. pertussis and B. bronchiseptica CyaAs (Supplementary Figure S5), confirming that we are working with the same CyaA proteins as Fukui-Miyazaki et al. We have further shown that the binding of B. bronchiseptica CyaA on myeloid cells is inhibited by the monoclonal antibody M1/70 against the CD11b subunit of CD11b/ CD18, as has been repeatedly demonstrated for recombinant CyaA 23,24 . All these data indicate that B. pertussis and B. bronchiseptica CyaAs use the same mechanisms required for receptor binding, membrane insertion, and subsequent penetration into CD11b/CD18-positive cells.
We have previously shown that the recombinant CyaA Y940A and CyaA Y940P variants overproduced in E. coli exhibit very low membrane binding and membrane penetration capacity similarly as non-acylated proCyaA 49,55 . However, analysis of the acylation status of the CyaA Y940A and CyaA Y940P variants showed that the residues K860 and K983 were correctly modified by the CyaC acyltransferase as intact CyaA 12 . This excluded the possibility that the absence of fatty-acyl chains induces destabilization of the apolar segments of the hydrophobic domain and of most of the acylation region 56 . Similarly, all three other RTX hemolysins characterized here and their corresponding mutant variants were properly acylated by a combination of myristoyl and hydroxymyristoyl chains, ruling out the possibility that the reduced cytotoxic capacity of the hemolysin mutants was due to their aberrant acylation. This is also the first report showing that A. pleuropneumoniae ApxIA hemolysin, like other bona fide hemolysins 2 , is acylated by a mixture of myristoyl and hydroxymyristoyl fatty acyl chains.
The role of aromatic residues having a specific affinity for a membrane region near the lipid carbonyl has been shown in the membrane insertion of numerous bacterial toxins [57][58][59] . In case of CyaA, however, the presence of the aromatic side chain at position 940 plays rather a role in the proper folding of secondary structures of the acylated segment, preventing the formation of inactive aggregates 60 , than in the direct anchoring of the toxin into the lipid bilayer. While the Y940 residue could be functionally replaced by an aromatic phenylalanine residue, lacking hydroxylation in the para position of the aromatic ring, its replacement by an aromatic tryptophan residue resulted in reduction of the cell-binding and membrane penetration capacity of the CyaA Y940W variant (Supplementary Figure S6). This could be due to the presence larger indole ring of the tryptophan residue that might interfere with functional folding of the acylated domain of the toxin. The membrane penetration process of CyaA  www.nature.com/scientificreports/ may further proceed by the insertion of several putative transmembrane α-helices located in the N-terminal part of the hemolysin moiety of the toxin, comprising the hydrophobic pore-forming domain and the ' AC-to-Hly' linker segment 6,7,10,12-15,61 . Consistent with our previously published results with recombinant CyaA 12 , our recent results confirm that the tyrosine-to-phenylalanine substitution of the conserved tyrosine of the acylated domain plays no role in membrane insertion and penetration of different RTX hemolysins. However, unlike for CyaA, the tyrosine-to-alanine substitution does not reduce the cytolytic capacity of RtxA and ApxIA hemolysins no more by a factor of ~ 2. A negligible, if any, reduction of the pore-forming and cytotoxic capacity of HlyA Y643A also fits well with the results showing that membrane binding of HlyC-activated CyaA 1-501 /HlyA 1-1024 Y643A resembles the membrane binding capacity of the intact chimeric CyaA 1-501 /HlyA 1-1024 construct. These data together with data from CD spectra show that the presence of the aromatic ring in position 643 is not essential for proper folding of the acylated segment of HlyA and subsequent insertion of the HlyA into the lipid bilayer. However, the tyrosine-to-proline substitution most likely disrupts the secondary structure(s) of the acylated segment involved in membrane insertion, as the specific hemolytic and/or pore-forming activity of HlyA Y643P, ApxIA Y639P, or RtxA Y642P was drastically reduced compared to the intact HlyA, ApxIA or RtxA. This is most likely due, at least in the case of HlyA, to the reduced membrane binding capacity of the HlyA Y643P construct, since the pore properties, namely pore conductance and pore lifetime, of the HlyA mutant variants were similar to those of intact HlyA. Comparable pore characteristics may also indicate that the membrane-interacting pore-forming domain of HlyA Y643P is properly folded. Surprisingly, the large reduction in the ability of HlyA Y643P to insert into the plasma membrane was not reflected in the calcium-dependent folding of the truncated variant of HlyA analyzed by CD spectroscopy. This may indicate that the changes in the secondary structure of the HlyA419 Y643P construct are not as dramatic as in the similar CyaA719 Y940P construct. The vicinity of the conserved Y940 residue of the acylated segment of CyaA is different compared to other RTX hemolysins. In contrast to CyaA, where the conserved Y940 is adjacent to a hydrophilic S939, the Y638-Y639, Y642-Y643, and F641-Y642 pairs are present in the acylated segment of ApxIA, HlyA, and RtxA, respectively. The complete loss of hemolytic and cytotoxic activity of the HlyA Y642P + Y643P variant compared with the HlyA Y642P and HlyA Y643P constructs may indicate that the non-conserved Y642, together with the conserved Y643, may maximize the cytolytic and cytotoxic effects of HlyA. Our data also suggest that bona fide hemolysins, such as HlyA, RtxA, and ApxIA, may interact with the plasma membrane of target cells in a different manner than CyaA and that, unlike Y940 in CyaA, the aromatic rings of Y639 in ApxIA, Y643 in HlyA or Y642 in RtxA are not involved in the folding of the acylated segment preceding the membrane insertion of the hemolysin molecule. Sequence alignments revealed a number of other conserved or non-conserved aromatic residues within the acylated segment of RTX hemolysins 62 , which may be involved in the efficient folding of the acylated segment and the correct positioning of hemolysin molecule towards the lipid bilayer, followed by the insertion of the acylated segment or of its part into the target cell membrane.   Construction of RtxA, HlyA, ApxIA, RTX719, HlyA419 and CyaA 1-501 /HlyA 1-1024 variants. Plasmid pT7rtxC-rtxA was used for co-expression of the rtxC and rtxA genes for production of the recombinant RtxC-activated RtxA toxin equipped with a C-terminal double 6xHis purification tag 35 . To produce HlyCactivated HlyA with a C-terminal double 6xHis tag, the plasmid pT7hlyC-hlyA was used 65 . For production of the ApxIC-acylated ApxIA toxin with 6xHis tags on both the N-terminal and C-terminal ends, the pET28bap-xIC-apxIA construct was used 38 . Plasmid pT7hlyC-cyaA 1-501 -hlyA 1-1024 was used to produce HlyC-activated CyaA 1-501 /HlyA 1-1024 hybrid molecule 66 . The expression vectors encoding the RTX719 protein was derived from pT7CT7ACT1-ΔNdeI, a bicistronic vector encoding the structural cyaA gene and the cyaC gene for the dedicated acyltransferase 67 . For construction of the pT7CT7-RTX719, the PCR fragments amplified from pT7CT7ACT1-ΔNdeI using the forward 5'-ATA CAT ATG CAT CAT CAT CAT CAT CAT GAA AAG CTG GCC AAC GAT TAC -3' and the reverse 5'-CCA GAG CTC GTT GTC CTG G-3' primers. Oligonucleotide-directed PCR mutagenesis was performed to construct: (i) pT7rtxCrtxA-derived plasmids for the expression of RtxC-acylated RtxA mutant variants with single substitutions Y642A, Y642F, or Y642P; (ii) pT7hlyC-hlyA-derived plasmids for the expression of HlyC-activated HlyA mutants with single/double substitutions Y642P, Y643A, Y643F Y643P, or Y642A + Y643A and Y642P + Y643P; (iii) pET28bapxIC-apxIA-derived constructs for the production of ApxIC-acylated ApxIA variants harboring single substitutions Y639A, Y639F, or Y639P; and (iv) pT7hlyC-cyaA 1-501 -hlyA 1-1024 -derived plasmids for the expression of HlyC-activated CyaA 1-501 /HlyA 1-1024 hybrid molecules carrying single substitutions Y643A, Y643F, or Y643P. The pT7CT7ACT1 plasmid 67 , harboring the cyaC and cyaA genes, was used to generate a construct for the expression of the HlyC-activated HlyA fragment harboring residues 419-1024 (HlyA419) and equipped with an N-terminal hexahistidine (6xHis) purification tag. For this purpose, the cyaC ORF in pT7CT7ACT1 was replaced from its start to stop codon by a coding sequence of hlyC and similarly, the cyaA ORF was replaced by a hlyA sequence encoding the residues 419-1024 of HlyA. The hlyC and hlyA sequences were PCR-amplified from the plasmid pT7hlyC-hlyA 65 and in addition, the hlyA sequence was fused in frame at the N-terminus to a sequence encoding a 6xHis purification tag, to yield the pT7CT7hlyC-N-6xHis-hlyA419-1024 plasmid. Site . The aromatic ring of the conserved tyrosine is essential for the full hemolytic activity of RtxA and ApxIA. The hemolytic activity of RtxA (a) and ApxIA (b) was analyzed as described in Fig. 4. Each point represents the mean ± SD of three independent determinations performed in duplicate with two independent toxin preparations. Significant differences are indicated by asterisks (**p < 0.01; ***p < 0.001; ****p < 0.0001). www.nature.com/scientificreports/ directed mutations (Y642P, Y643P and Y643A) were introduced into the hlyA fragment of pT7CT7hlyC-N-6xHis-hlyA419-1024 by site-directed PCR mutagenesis as previously described 10 .
cAMP determination. THP-1 cells were incubated at 37 °C with CyaA for 30 min in D-MEM, the reaction was stopped by addition of 0.2% Tween-20 in 50 mM HCl, samples were boiled for 15 min at 100 °C, neutralized by addition of 150 mM unbuffered imidazole and cAMP was measured by a competitive immunoassay as previously described 70 . Activity of CyaA was taken as 100%.
Hemolytic activity on sheep erythrocytes. Hemolytic   The instrument was operating in survey LC-MS mode and calibrated online using Agilent tuning mix, which results in mass accuracy below 2 ppm. After analysis, spectra were processed using the Data Analysis 4.4 software package (Bruker Daltonics) and extracted data were searched against the FASTA of single corresponding toxin molecule (ApxIA, UniProtKB: P55128; HlyA, UniProtKB: P08715; RtxA, UniProtKB: A0A1X7QMH9) using Linx software (RRID:SCR_018657). The Linx algorithm was set for fully tryptic restriction with a maximum of 3 missed cleavages and variable modification for methionine oxidation along with lysine acylation ranging from C12 to C18, including monosaturated and hydroxylated variants. The acylation status of lysine residues was determined by comparison of relative intensity ratios between acylated peptide ions and their unmodified counterparts 2 .
CD spectroscopy. The far-UV CD spectra were recorded at 25 °C on a Chirascan-plus spectrometer (Applied Photophysics, USA) in rectangular quartz Suprasil cells of 1-mm path length (110-QS, Hellma, Germany). Protein samples were diluted in 5 mM Tris-HCl (pH 8.0) and 20 mM NaCl in the absence or presence of CaCl 2 and measured at wavelengths from 203 to 260 nm at a scan speed of 1 nm/s. Ca 2+ -induced structural changes were monitored by stepwise titration of protein samples with increasing concentrations of CaCl 2 . The spectra of the buffers were subtracted from the protein spectra and the molar residue ellipticity (Θ) was expressed in milidegrees square centimeter per decimole [mdeg cm 2 dmol −1 ].
Gel filtration. Gel filtration chromatography was performed on a Superdex 200HR gel filtration column (GE Healthcare, UK) connected to AKTAprime Plus liquid chromatography system (GE Healthcare, UK). The column was equilibrated with a buffer containing 20 mM Tris-HCl (pH 8.0), 150 mM NaCl and 2 mM CaCl 2 before the denatured proteins (400 µg) in 50 mM Tris-HCl (pH 8.0) and 8 M urea were loaded onto the column and refolded at a flow rate of 0.5 ml/min.

Planar lipid bilayers.
Measurements on planar lipid bilayers were performed in Teflon cells separated by a diaphragm with a circular hole (diameter 0.5 mm) bearing the membrane 15 . RTX proteins were prediluted in TUC buffer (50 mM Tris-HCl (pH 8.0), 8 M urea, and 2 mM CaCl 2 ) and added to the grounded cis compartment with positive potential. The membrane was formed by the painting method using soybean lecithin in n-decane-butanol www.nature.com/scientificreports/ with salt bridges (applied voltage, 50 mV), amplified by LCA-200-100G and LCA-200-10G amplifiers (Femto, Berlin, Germany), and digitized by use of a LabQuest Mini A/D convertor (Vernier, Beaverton, OR). For lifetime determination, approximately 400 of individual pore openings were recorded and the dwell times were determined using QuB software with 100 Hz low-pass filter. The kernel density estimation was fitted with exponential function using Gnuplot software. The relevant model was selected by the χ2 value.
Statistical analysis. Results were expressed as the arithmetic mean ± standard deviation (SD) of the mean.

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary Information file).