Acyltransferase-mediated selection of the length of the fatty acyl chain and of the acylation site governs activation of bacterial RTX toxins

In a wide range of organisms, from bacteria to humans, numerous proteins have to be posttranslationally acylated to become biologically active. Bacterial repeats in toxin (RTX) cytolysins form a prominent group of proteins that are synthesized as inactive protoxins and undergo posttranslational acylation on e -amino groups of two internal conserved lysine residues by co-expressed toxin-activating acyltransferases. Here, we investigated how the chemical nature, position, and number of bound acyl chains govern the activities of Bordetella pertussis adenylate cyclase toxin (CyaA), Escherichia coli a -hemolysin (HlyA), and Kingella kingae cytotoxin (RtxA). We found that the three protoxins are acylated in the same E. coli cell background by each of the CyaC, HlyC, and RtxC acyltransferases. We also noted that the acyltransferase selects from the bacterial pool of acyl – acyl carrier proteins (ACPs) an acyl chain of a specific length for covalent linkage to the protoxin. The acyltransferase also selects whether both or only one of two conserved lysine residues posttranslationally Functional assays revealed that RtxA has to be modified by 14-carbon fatty acyl chains to be biologically active, that HlyA remains active also when modified by 16-carbon acyl chains, and that CyaA is activated exclusively by 16-carbon acyl chains. These results suggest that the RTX toxin molecules are structurally adapted to the length of the acyl chains used for modification of their acylated lysine residue in the second, more conserved acylation site. by 16-carbon acyl chains. These results reveal the selection of acyl chains of appropriate lengths by the respective cognate RTXC acyltransferase enzymes and suggest a structural adaptation of the RTXA toxin molecules to the length of the acyl chains used for modification of their crucial acylated residue in the second, more conserved acylation site.

In a wide range of organisms, from bacteria to humans, numerous proteins have to be posttranslationally acylated to become biologically active. Bacterial repeats in toxin (RTX) cytolysins form a prominent group of proteins that are synthesized as inactive protoxins and undergo posttranslational acylation on e-amino groups of two internal conserved lysine residues by co-expressed toxin-activating acyltransferases. Here, we investigated how the chemical nature, position, and number of bound acyl chains govern the activities of Bordetella pertussis adenylate cyclase toxin (CyaA), Escherichia coli a-hemolysin (HlyA), and Kingella kingae cytotoxin (RtxA). We found that the three protoxins are acylated in the same E. coli cell background by each of the CyaC, HlyC, and RtxC acyltransferases. We also noted that the acyltransferase selects from the bacterial pool of acyl-acyl carrier proteins (ACPs) an acyl chain of a specific length for covalent linkage to the protoxin. The acyltransferase also selects whether both or only one of two conserved lysine residues of the protoxin will be posttranslationally acylated. Functional assays revealed that RtxA has to be modified by 14-carbon fatty acyl chains to be biologically active, that HlyA remains active also when modified by 16-carbon acyl chains, and that CyaA is activated exclusively by 16-carbon acyl chains. These results suggest that the RTX toxin molecules are structurally adapted to the length of the acyl chains used for modification of their acylated lysine residue in the second, more conserved acylation site.
Acylation of the RTX toxins appears to be crucial for all of their known cytotoxic activities (2,5,6,8,17,18). However, the precise molecular mechanism by which the acyl chains contribute to membrane insertion and formation of pores by the toxins remains poorly understood. The presence of the acyl chains was shown to play a structural role in the folding of CyaA into a biologically active conformation (19,20) and in a productive and irreversible interaction of CyaA with cells expressing the complement receptor 3 (CR3; also known as the integrin a M b 2 , CD11b/CD18, or Mac-1) (17,21). Acylation was also shown to be required for the irreversible insertion of HlyA to target membrane (22) and for protein-protein interaction in HlyA oligomerization within the membrane microdomains (23).
This article contains supporting information. ‡ These authors contributed equally to this work. * For correspondence: Radim Osicka, osicka@biomed.cas.cz.
The toxin-activating acyltransferase genes are highly conserved between the rtx loci of various bacterial genera, and some of these acyltransferases were reported to activate also heterologous protoxins. For example, the HlyC-modified Actinobacillus pleuropneumoniae hemolysin ApxIA, as well as the ApxC-modified HlyA expressed in E. coli, exhibited a hemolytic activity on erythrocytes (24,25). Similarly, the heterologously HlyC-or CyaC-activated Pasteurella hemolytica leukotoxin LktA exhibited the same activity and target cell specificity as the LktA activated by its cognate LktC acyltransferase. Nevertheless, the activation was not reciprocal, as the LktC-activated HlyA and CyaA produced in E. coli were neither hemolytic nor cytotoxic (26,27). However, it has not been determined why some RTX protoxins are efficiently cross-activated by heterologous acyltransferases and some are not.
Here, we analyzed the activation of the CyaA, HlyA, and RtxA toxins, each acylated by one of the three CyaC, HlyC, or RtxC acyltransferases and produced in the same E. coli cell background, so as to eliminate the potential impact of differences in acyl-ACP pool composition of the original producer bacteria. The results reveal that it is the RTXC acyltransferase enzyme that selects the type of the acyl chain of adapted length that is covalently linked to the proRTXA protein, and the acyltransferase also selects whether a single lysine residue or both modification sites of proRTXA will be posttranslationally acylated, thereby conferring the biological activity on the RTX toxin.

Results
CyaC selects C16 fatty acyl chains, whereas HlyC and RtxC select C14 acyl chains and differ in recognition of acylation sites To test the acyl residue and acylation site selectivity of the three homologous RTX toxin-activating acyltransferases, we produced the nine pairwise combinations of the three proto-xins proCyaA, proHlyA, and proRtxA with the three acyltransferase enzymes CyaC, HlyC, and RtxC, respectively, in the same E. coli cell background. For this purpose, the expression signals of the pT7CACT1 plasmid for production of CyaC-activated CyaA toxin (28) were employed. The cyaC ORF in pT7CACT1 was replaced from start to stop codon by the coding sequences of the hlyC or rtxC genes, and the cyaA ORF was replaced by hlyAor rtxA-coding sequences, respectively (Fig. 1A). In addition, the hlyA and rtxA genes were fused in frame at the 39 terminus to a sequence encoding a double-hexahistidine purification tag. The obtained set of nine constructs was used to produce the nine resulting RTXC-acylated RTXA toxins in E. coli BL21 cells. These were purified close to homogeneity from urea-solubilized inclusion bodies by affinity chromatography on calmodulin-Sepharose (CyaA proteins) or Ni-NTA agarose (HlyA and RtxA proteins) (Fig. 1B).
To analyze the acylation pattern of the nine RTXA variants, digests of the purified proteins were analyzed by liquid chromatography coupled to ultrahigh-resolution Fourier transform ion cyclotron resonance MS (LC FT-ICR MS). As summarized in Table 1, the CyaA CyaC toxin activated by its cognate acyltransferase CyaC exhibited a predominant acylation with palmitoyl (C16:0) and palmitoleyl (C16:1) chains at the Lys-860 (;66%) and Lys-983 (;88%) residues. A small proportion of the CyaA CyaC molecules was also modified by myristoyl (C14:0) and octadecenoyl (C18:1) chains linked to the Lys-860 (;3%) and Lys-983 (;11%) residues (Table 1). Only ;31% of the Lys-860 residues remained unacylated by CyaC, whereas the Lys-983 residue of CyaA CyaC was acylated nearly completely (;99%). In contrast, when proCyaA was produced in the presence of HlyC or RtxC, only minimal (1% in CyaA HlyC ) or no (0% in CyaA RtxC ) acylation of the Lys-860 residue in the first acylation site of CyaA was detected (Table 1). Hence, the heterologous acyltransferases recognized the first acylation site of CyaA ( Fig. 1C) with negligible efficacy. In contrast, the second To generate the constructs, the pT7CACT1 plasmid harboring the cyaC and cyaA genes was used. The cyaC ORF in pT7CACT1 was replaced from its start to stop codon by hlyC or rtxC, and similarly, the cyaA ORF was replaced by hlyA or rtxA, respectively. B, a set of the nine pT7CACT1-derived constructs was used for the production of the RTXC-activated RTXA toxins in E. coli BL21/pMM100 cells. The proteins were purified close to homogeneity from urea-solubilized inclusion bodies by affinity chromatography on calmodulin-Sepharose (CyaA variants) or Ni-NTA agarose (HlyA and RtxA variants). The samples were analyzed on 7.5% polyacrylamide gels and stained with Coomassie Blue. St, molecular mass standards. C, ClustalW sequence alignment of acylated sites of CyaA (UniProt code: P0DKX7), HlyA (UniProt code: P08715), and RtxA (K. kingae isolate PYKK081) with two conserved internal lysine residues (in boldface type and underlined) whose e-amino groups are posttranslationally acylated. *, identity; :, strongly similar; ., weakly similar.
All of these results demonstrated that the different RTXC acyltransferases exhibit a varying selectivity for acyls of various lengths. CyaC selected from the E. coli acyl-ACP pool almost exclusively the C16:0 and C16:1 acyl chains for acylation of the cognate proCyaA or of the heterologous proHlyA and proRtxA substrates. In contrast, the HlyC and RtxC acyltransferases selected nearly exclusively the shorter C14:0 and C14:0-OH chains for acylation of all three proRTXA substrates ( Table 1). This reveals that it is the RTXC acyltransferase enzyme that selects the length of the acyl chain that it transfers to the RTXA substrate protein. Moreover, the data showed that the RTXC enzymes promiscuously and quite efficiently recognize and modify the second, more conserved acylation sites (Fig. 1C) of heterologous proRTXA substrates, whereas the recognition and acylation of the first acylation sites in heterologous pro-RTXA substrates was nil or inefficient.

Acylation patterns determine the levels of biological activity of the RTX toxins
The set of nine toxins with defined numbers and lengths of attached acyl chains enabled us to assess how the single or double acylation and the length of the attached acyl chains affect the biological activities of these proteins. For the three differently acylated CyaA variants, we first determined their capacity to bind, penetrate, and lyse sheep erythrocytes, respectively, using red blood cells as surrogate target cells lacking the CyaA receptor CR3. As documented in Fig. 2A, compared with double acylation of the CyaA CyaC protein by predominantly the C16:0 and/or C16:1 chains, monoacylation of the Lys-983 residue by predominantly the C14:0 or C14:0-OH chains reduced the relative capacity of CyaA HlyC (by ;62%) and CyaA RtxC (by ;70%) to bind erythrocytes. The modification by the C14 acyls decreased even more the relative cell-invasive capacity of both toxin variants (by ;94%) ( Fig. 2A). Moreover, at the rather high concentration of 10 mg/ml that was used, the CyaA HlyC and CyaA RtxC proteins were unable to provoke any lysis of erythrocytes over time, whereas complete erythrocyte lysis was provoked by the same amount of CyaA CyaC within 5 h of incubation (Fig. 2B). The relative AC-translocating and lytic capacities of the CyaA HlyC and CyaA RtxC proteins remained low even when their input concentration was increased 2.5-fold over that of CyaA CyaC to achieve binding of equal amounts of the three CyaA variants to erythrocyte cells (Fig. 2, C and D). The C14:0 or C14:0-OH monoacylated CyaA HlyC and CyaA RtxC toxins bound with a reduced capacity (;53 and ;60% of CyaA CyaC binding) also to mouse J774A.1 macrophages that express the CyaA receptor CR3 (Fig. 2E). However, the binding of CyaA HlyC  None  31  1  99  20  100  7  93  0  10  0  73  0  100  1  97  0  98  0  C12:0  2  2  2  2  2  2  2  2  3  2  1  2  2  2 The RTXA variants were produced in the presence of the RTXC acyltransferases in E. coli BL21/pMM100 cells, purified close to homogeneity and analyzed by MS. Percentage distributions of fatty acyl chains linked to the e-amino groups of the lysine residues were estimated semiquantitatively, from the relative intensities of selected ions in reconstructed ion current chromatograms. Average values are calculated from determinations performed with two different toxin preparations. 2, the acyl chain was not detected.
and CyaA RtxC to J774A.1 cells was CR3 receptor-specific, as it could be blocked by the competing antibody M1/70 that binds the CD11b subunit of CR3 (29, 30) (Fig. 2E). Nevertheless, compared with the C16 biacylated CyaA CyaC , the C14 monoacylated CyaA HlyC and CyaA RtxC proteins were strongly impaired in their capacity to translocate the AC domain into the cytosol of J774A.1 cells to elevate the intracellular cAMP levels (Fig. 2F). This functional defect on J774A.1 cells was most likely not due to the lack of acylation of the Lys-860 residue of the CyaA HlyC and CyaA RtxC toxins, as a CyaA-K860R mutant acylated by CyaC only on the Lys-983 residue (17) exhibited similar capacity to bind (Fig. 2E) and translocate its AC domain (Fig. 2F) into J774A.1 cells as the intact doubly acylated CyaA CyaC toxin. Hence, the low cell-invasive activity of the monoacylated CyaA HlyC and CyaA RtxC proteins was rather due to the modification of the Lys-983 residue by the shorter C14 acyl chains, which conferred a much lower specific membrane penetration capacity on CyaA than the modification of the Lys-983 residue by the C16 acyl chains.

EDITORS' PICK: Specificity of toxin-activating acyltransferases
In agreement with the residual cytolytic capacity on erythrocytes, the CyaA HlyC and CyaA RtxC variants exhibited also a very low overall membrane activity on artificial lipid bilayers made of 3% asolectin (Fig. 3A). However, as calculated from singlepore recordings (Fig. 3B), the most frequent conductances of pores formed by CyaA HlyC (12 pS) and CyaA RtxC (12 pS) were quite comparable with that of CyaA CyaC (11 pS) (Fig. 3C). Similarly, the most frequent lifetimes of pores formed by CyaA HlyC (1140 ms) and CyaA RtxC (1129 ms) were comparable with that of CyaA CyaC (1083 ms) (Fig. 3D). It can thus be concluded that the difference in the number (one versus two acylated lysine residues) and the length and chemical nature (C16:0/C16:1 versus C14:0/C14:0-OH) of attached acyl chains affected strongly the propensity of formation but not the overall characteristics of the individual pores generated by the differently acylated CyaA toxin variants.
In line with the comparable cytolytic capacities on erythrocytes, all three HlyA variants displayed comparable overall membrane activities on artificial lipid bilayers (Fig. 5A). As calculated from single-pore recordings (Fig. 5B), the HlyA HlyC variant formed pores with the most frequent conductance of 405 pS that was similar to that of pores formed by the HlyA CyaC (322 pS) and HlyA RtxC (384 pS) proteins (Fig. 5C). The HlyA CyaC and HlyA RtxC toxin variants also formed pores with most-frequent pore lifetimes similar to those of the pores formed by the HlyA HlyC (1788, 1654, and 1599 ms, respectively) (Fig. 5D).
All of these results demonstrate that the C16 acyl chains attached to the Lys-690 residue conferred on the HlyA CyaC toxin a similar pore-forming and cytotoxic activity as did the naturally HlyC-mediated acylation by C14 acyl chains attached to both Lys-564 and Lys-690 residues in HlyA HlyC or the partial acylation of Lys-564 and full acylation of Lys-690 by C14 acyl chains in the HlyA RtxC toxin. This indicates that in contrast to CyaA, monoacylation of the single Lys-690 residue by either C14 or C16 acyl chains was sufficient for biological activity of HlyA.
Similarly, single acylation of the Lys-689 residue of RtxA by C14:0 or C14:0-OH acyl chains (Table 1) was sufficient for full cytolytic activity of the RtxA RtxC and RtxA HlyC toxins, acylated by RtxC or HlyC, respectively (Fig. 6). However, in contrast to HlyA, the RtxA toxin was not activated by CyaC despite almost complete modification of the Lys-689 residue by the C16:0 or C16:1 acyl chains. Hence, the biological activity of RtxA was supported only upon modification by C14 and not C16 acyl chains. In line with that, the RtxA RtxC and RtxA HlyC toxins displayed comparable overall membrane activities on planar lipid bilayers, whereas the RtxA CyaC protein exhibited a low membrane activity (Fig. 7A). Nevertheless, the formed RtxA CyaC pores (Fig. 7B) exhibited a comparable most-frequent conduct-ance of 487 pS, like RtxA HlyC (454 pS) and RtxA RtxC (479 pS) pores, respectively (Fig. 7C). Similarly, the most frequent values of single-pore lifetimes of RtxA CyaC (968 ms) were similar to those of RtxA HlyC (1002 ms) and of RtxA RtxC (1288 ms) (Fig.  7D). These results indicate that the RtxA CyaC variant was impaired in its capacity to insert into the lipid bilayer and/or in its propensity to form oligomeric pores due to the modification by the longer C16 acyl chains.

Discussion
We report that of the three examined RTXA toxins, the CyaA toxin is only activated by predominant modification of its Lys-983 residue by C16 fatty acyl chains and the RtxA protein only by the modification of its Lys-689 residue by C14 acyl chains. Intriguingly, the HlyA toxin can be fully activated by modification with either C14 or C16 acyl chains linked to its Lys-690 residue. These results further show that it is the acyltransferase activating the RTX protoxin that selects the acyl chain of the functionally adapted length from the acyl-ACP pool of the producing bacterium. Furthermore, the acyltransferase also determines whether both or only one of the two conserved acylation sites in the respective RTX protoxin will be recognized and covalently modified by the linked acyl chains.
In line with these findings, we found here that palmitoylation (Lys-860, ;32%; Lys-983, ;45%) and palmitoleylation (Lys-860, ;34%; Lys-983, ;43%) are the two major posttranslational modifications of the Lys-860 and Lys-983 residues also when the CyaC-acylated recombinant CyaA is produced in the E. coli B strain BL21. In contrast, the HlyC-or RtxC-modified CyaA variants produced on the same genetic background were predominantly myristoylated and hydroxymyristoylated almost exclusively on the Lys-983 residue (CyaA HlyC , ;77%; CyaA RtxC , ;85%). Modification by C16:0 and C16:1 acyl chains was negligible in CyaA HlyC (;3%) and CyaA RtxC (;8%). The CyaA HlyC and CyaA RtxC variants then exhibited substantially lower biological activities than CyaA CyaC , and this was most likely not due to the missing acylation at the Lys-860 residue of CyaA HlyC and CyaA RtxC . Indeed, we have demonstrated that acylation of the Lys-983 residue of CyaA CyaC was necessary and sufficient for biological activities of the toxin on both CR3-negative erythrocytes (13) as well as on CR3-positive J774A.1 cells (Fig. 2  (E and F)) (17). It is therefore plausible to conclude that the 14- carbon myristoyl and hydroxymyristoyl chains are unable to functionally replace the 16-carbon palmitoyl and palmitoleyl chains at the Lys-983 residue of CyaA. Thus, not only the absence of acylation itself, but also the length of the covalently linked acyl chains appears to play a crucial role in activation of CyaA and in conferring of biological activities on this toxin.
The specific binding of the myristoylated and hydroxymyristoylated CyaA HlyC and CyaA RtxC variants to CR3-positive J774A.1 cells was affected only mildly. However, compared with C16-acylated CyaA CyaC , the C14 acyl-modified CyaA HlyC and CyaA RtxC variants were largely impaired in the capacity to deliver the AC enzyme across target cell membrane into the cytosol and to intoxicate cells by cAMP production. Hence, the two-carbon unit shorter C14 acyl chains were unable to adequately support the membrane insertion and translocation of the CyaA polypeptide. Indeed, the CyaA HlyC and CyaA RtxC variants exhibited a negligible overall membrane activity also on artificial lipid bilayers. This suggests that a 16-carbon-long acyl chain has to be attached to the e-amino group of the side chain of the Lys-983 residue of CyaA to impose the necessary structure on the toxin molecule and enable it to effectively interact with the lipid bilayer of the target cell or of the artificial membrane. However, once the toxin has inserted into the lipid bilayer at a low frequency, the modification of CyaA with the C14 acyl chains does not affect anymore the conductance and lifetime of single pores formed by the CyaA HlyC and CyaA RtxC variants, because these exhibited quite the same properties as those formed by the C16-acylated CyaA CyaC . This shows that differences in the number and length/chemical nature of the acyl chains linked to the CyaA molecule impact only the propensity of the toxin to insert into the membrane and form oligomeric pores and not on the structure of the pores itself.
We further demonstrated that CyaC-modified CyaA was completely acylated on the Lys-983 residue, whereas ;31% of toxin molecules remained unacylated at the Lys-860 residue (Table 1). In contrast, only residual (;1%) or no acylation was observed on Lys-860, when CyaA was modified with HlyC or RtxC, respectively. This suggests that CyaC may have a higher affinity for the Lys-860 acylation site of CyaA than the HlyC and RtxC acyltransferases. All of these results indicate that the CyaC acyltransferase co-evolved in B. pertussis with CyaA to modify it by C16 acyl chains and cannot be simply replaced by toxin-activating acyltransferases of other Gram-negative bacteria that preferentially select C14-bearing ACP for modification of proRTXA substrates.
Previously, fatty acyl modification of HlyA activated in vivo in two different uropathogenic isolates of E. coli (with a chromosomal (J96) and an extrachromosomal (pHly152) hly locus) was analyzed, and the Lys-564 and Lys-690 residues of HlyA were found to be mostly acylated by a myristoyl chain (;68%). The remaining linked acyls were then identified as the very rare C15:0 (;26%) and C17:0 (;6%) odd-carbon fatty acyl chains (9). Here, we demonstrated that the recombinant HlyC-modified HlyA toxin produced in the E. coli strain BL21 was acylated mostly by the C14:0 and C14:0-OH chains both at the Lys-564 (;84%) and Lys-690 (;93%) residue and partially by the C12:0, C12:0-OH, C16:0, and C16:1 chains. In contrast to the results of Lim et al. (9), we were unable to identify any C15:0 and C17:0 acyl groups attached to the HlyC-modified HlyA molecule despite the extreme sensitivity and accuracy of the current state-of-the-art analytical LC FT-ICR MS technology used. This indicates that uropathogenic E. coli isolates may have an acyl-ACP pool composition different from that of the E. coli B strain used here.
Intriguingly, the activation of HlyA exhibits a flexibility as to the length and nature of the attached acyl chains. When the Lys-690 residue of HlyA was predominantly acylated with the C16:0 and C16:1 (;90%) acyl chains by CyaC, the HlyA CyaC protein exhibited a similar capacity to reduce viability of THP-1 cells or to lyse erythrocytes, and it exhibited similar membrane properties on planar lipid bilayers as upon C14 modification on both Lys-564 and Lys-690 residues. Thus, in contrast to CyaA, the HlyA toxin can be modified by acyl chains of varying length (C14 versus C16) and still gains biological activity.
Previously, using deleted HlyA protoxin variants and peptides as substrates in an in vitro acylation assay, Stanley et al. (31) demonstrated that HlyC possessed an about 4 times higher affinity for the segment encompassing the Lys-564 residue of HlyA than for that harboring the Lys-690 residue, resulting in acylation of 80% Lys-564 residues and only 20% of Lys-690 residues. Moreover, substitutions of the Lys-564 and Lys-690 residues revealed that both sites were required for hemolytic activity of HlyA (8). Interestingly, our results indicate that the affinity of the CyaC and RtxC acyltransferases in vivo was much higher to the segment harboring the Lys-690 residue, as this was quantitatively acylated by both enzymes in vivo, whereas only ;7 or ;27% of HlyA molecules were modified by CyaC or RtxC on the Lys-564 residue. Despite a substantially lower extent of acylation of the Lys-564 residue, the complete acylation of the Lys-690 residue conferred on the HlyA CyaC and HlyA RtxC proteins a full capacity to lyse erythrocytes. Moreover, the HlyA CyaC and HlyA RtxC proteins also exhibited an equally high membrane activity on planar lipid bilayers as the EDITORS' PICK: Specificity of toxin-activating acyltransferases native doubly acylated HlyA HlyC protein. Hence, rather than acylation of Lys-564 being essential, the activity of HlyA was likely lost upon substitution of Lys-654 because of a structural role of the Lys-564 residue in toxin activity. Indeed, a similar conclusion was reached upon substitution of the homologous Lys-860 residue of CyaA, the acylation of which is by itself dispensable for CyaA activity on both erythrocytes, as well as on CR3-expressing cells (13,17). However, the CyaA-K860R mutant is importantly affected in its specific membrane penetration activity on cells lacking the CR3 receptor (13,17).
Recently, we reported that the recombinant RtxA produced together with RtxC in the E. coli B strain BL21 was primarily modified by C14:0 and C14:0-OH acyl chains (;89%) on the Lys-689 residue (5). Only a minor proportion of the RtxA mole-cules (;8%) was found to be acylated on Lys-689 by C16:1 palmitoleyl chains (5). A small proportion of the RtxA molecules (;23%) was also found to be modified by C14:0 and C14:0-OH acyl chains on the Lys-558 residues (5). Here, we confirmed the type and extent of acylation of recombinant RtxA toxin on the Lys-689 residue, whereas a lower level (;2%) of modification of the Lys-558 residue was detected. This indicates that the extent of Lys-558 acylation may vary as a function of the physiological state of the producing bacteria, as discussed previously for the acylation of the Lys-860 residue of CyaA (13). Alternatively, differences in LC-MS configuration used for quantification of protein peptides between the reports may have accounted for the variation in the detected quantities of the acylated peptide (5). Indeed, reproducibility of MS-based peptide quantitation Figure 7. The CyaC-acylated RtxA variant has substantially reduced overall membrane activity but forms pores with similar single-pore conductance and pore lifetime as RtxA HlyC and RtxA RtxC . A, overall membrane activities of the RtxA variants on asolectin/decane:butanol (9:1) membranes in the presence of 250 pM purified proteins. The aqueous phase contained 150 mM KCl, 10 mM Tris-HCl (pH 7.4), 2 mM CaCl 2 ; the applied voltage was 50 mV; the temperature was 25°C; and the recording was filtered at 10 Hz. B, single-pore recordings of asolectin membranes in the presence of 10 pM purified RtxA variants under conditions otherwise identical to those in A. C, KDE of single-pore conductances calculated from single-pore recordings (.500 events) acquired on several different asolectin membranes with 10 pM RtxA RtxC or its variants at the same conditions as in A. The numbers represent the most frequent conductances 6 S.D. of pores formed by the RtxA variants. D, for lifetime determination, ;400 individual pore openings were recorded on several different asolectin membranes with 10 pM RtxA RtxC or its variants under the same conditions as in A, and the logarithmic histogram of dwell times was fitted with a double-exponential function. The error estimates of lifetimes were obtained by bootstrap analysis. The numbers in each panel represent the most frequent values 6 S.D. was shown to vary by up to 20%, depending on the sample preparation, unique characteristics of reversed-phase columns used for separation of peptides, and LC-MS instrument configuration, respectively (32).
Similarly as RtxA RtxC , the HlyC-modified RtxA was predominantly acylated by C14:0 and C14:0-OH acyl chains (;96%) on the Lys-689 residue, and only residual modification by C14:0 and C14:0-OH chains (;3%) was observed on the Lys-558 residue. The RtxA RtxC and RtxA HlyC variants with a similar acylation pattern then exhibited comparable capacities to lyse erythrocytes and similar membrane properties on planar lipid bilayers. However, unlike the fully biologically active C16-acylated HlyA CyaC toxin, the RtxA CyaC protein modified by CyaC at the Lys-689 residue by the C16:0 and C16:1 chains (;89%) was unable to lyse erythrocytes and exhibited only a residual overall membrane activity on planar lipid membranes. C16acylated RtxA CyaC was most likely impaired in binding/insertion to the lipid bilayer and/or in the propensity to form oligomeric pores, as once inserted into the membrane, it formed pores exhibiting single-pore conductances and lifetimes similar to those of RtxA RtxC . In this respect, the RtxA CyaC protein was similar to the unacylated proCyaA, proHlyA, and proRtxA protoxins, which despite highly decreased overall membrane activity, once inserted into the lipid bilayer, formed pores with properties similar to those of the acylated toxins (5,17,33,34).
In conclusion, we report here that RtxA has to be modified by 14-carbon fatty acyl chains to be biologically active, whereas HlyA remains active also when modified by 16-carbon acyl chains and CyaA is only activated by 16-carbon acyl chains. These results reveal the selection of acyl chains of appropriate lengths by the respective cognate RTXC acyltransferase enzymes and suggest a structural adaptation of the RTXA toxin molecules to the length of the acyl chains used for modification of their crucial acylated residue in the second, more conserved acylation site.

Bacterial strains
The E. coli strain XL1-Blue (Stratagene, La Jolla, CA) was used throughout this work for DNA manipulations and was grown in Luria-Bertani medium at 37°C. The E. coli strain BL21 (Novagen, Madison, WI) carrying the plasmid pMM100 (encoding LacI and tetracycline resistance) (35) was used for expression of the RTX proteins.

Standard techniques
Determination of protein concentration and SDS-PAGE were performed according to standard protocols (36). PageRu-ler Unstained Protein Ladder (catalog no. 26614, Thermo Fisher Scientific) was used as a size standard in SDS-PAGE.

Plasmid construction
The pT7CACT1 plasmid (28), harboring the cyaC and cyaA genes under control of the isopropyl-b-D-thiogalactopyranoside-inducible lacZp promoter, was used to generate constructs for the expression of the CyaA, HlyA, and RtxA toxins activated by the acyltransferase CyaC, HlyC, or RtxC, respectively. For this purpose, the cyaC ORF in pT7CACT1 was replaced from its start to stop codon by coding sequences of hlyC or rtxC, and similarly, the cyaA ORF was replaced by hlyA or rtxA ORFs, respectively (Fig. 1A). The hlyC and hlyA genes were PCR-amplified from the plasmid pTZHly11 (37), and the rtxC and rtxA genes were amplified from the plasmid pT7rtxC-rtxA (5). In addition, the hlyA and rtxA genes were fused in frame at the 39 terminus to a sequence encoding a double-hexahistidine purification tag (38).

Protein production and purification
The RTXC-activated RTXA toxins were produced in E. coli BL21/pMM100 cells transformed with the appropriate plasmids. 500-ml cultures were grown with shaking at 37°C in MDO medium (20 g/liter yeast extract, 20 g/liter glycerol; 1 g/ liter KH 2 PO 4 , 3 g/liter K 2 HPO 4 , 2 g/liter NH 4 Cl, 0.5 g/liter Na 2 SO 4 , 0.01 g/liter thiamine hydrochloride) containing 150 mg/ml ampicillin and 12.5 mg/ml tetracycline. When cultures reached OD 600 = 0.8, protein production was induced with 1 mM isopropyl-b-D-thiogalactopyranoside for an additional 4 h. For protein purification, the cells were harvested by centrifugation, washed twice with 50 mM Tris-HCl (pH 8.0), and disrupted by sonication at 4°C, and the homogenate was centrifuged at 20,000 3 g for 30 min at 4°C. The inclusion bodies collected in the pellet were washed with 50 mM Tris-HCl (pH 8.0) containing 4 M urea and then solubilized with 50 mM Tris-HCl (pH 8.0) containing 8 M urea, and the urea extract was cleared at 20,000 3 g for 30 min at 4°C.
The urea extracts containing the HlyA and RtxA variants were loaded on an Ni-NTA agarose column (Qiagen, Germantown, MD) equilibrated with TNU buffer (50 mM Tris-HCl (pH 8.0), 200 mM NaCl, and 8 M urea). The column was washed with TNU buffer containing 20 mM imidazole, and the HlyA and RtxA variants were eluted with TNU buffer containing 600 mM imidazole. The eluted fractions of HlyA and RtxA were diluted 4 times in ice-cold 50 mM Tris-HCl (pH 8.0) containing 1 M NaCl and loaded on a phenyl-Sepharose CL-4B column (Sigma-Aldrich) equilibrated with the same buffer. The column was then washed with 50 mM Tris-HCl (pH 8.0), and the HlyA and RtxA variants were eluted with TUE buffer (50 mM Tris-HCl (pH 8.0), 8 M urea, and 2 mM EDTA).
The urea extracts containing the CyaA variants were diluted 4 times in ice-cold washing buffer (50 mM Tris-HCl (pH 8.0), 500 mM NaCl, and 2 mM CaCl 2 ) and loaded at 4°C on a calmodulin-Sepharose 4B column (GE Healthcare) equilibrated with the same buffer. The column was washed with washing buffer, and the CyaA variants were eluted at room temperature with TUE buffer.

LC-MS analysis
The proteins were dissolved in 50 mM ammonium bicarbonate buffer (pH 8.2) to reach 4 M concentration of urea and digested with trypsin (Promega (Madison, WI), modified sequencing grade) at a trypsin/protein ratio of 1:50 for 6 h at 30°C. The second portion of trypsin was added to a final ratio of trypsin/protein of 1:25, and the reaction was carried out for another 6 h at 30°C. When the reaction was complete, the concentration of the resulting peptides was adjusted by 0.1% TFA to 0.1 mg/ml, and 5 ml of the sample were injected into the LC-MS system. The LC separation was performed using a desalting column (ZORBAX C18 SB-300, 0.1 3 2 mm) at a flow rate of 40 ml/min (Shimadzu, Kyoto, Japan) of 0.1% formic acid (FA) and a separation column (ZORBAX C18 SB300, 0. RtxA: UniprotKB code A0A1X7QNC7) using the home-built Linx software (RRID:SCR_018657). The Linx algorithm was set for fully tryptic restriction with a maximum of three missed cleavages and variable modification for methionine oxidation along with lysine acylation ranging from C12 to C18, including monosaturated and hydroxylated variants. The mass error threshold was set to 62 ppm, and all assigned peptides used for quantification were verified manually. The acylation status of lysine residues was determined by comparison of the relative intensity ratio between acylated peptide ions and their unmodified counterparts. Only lysine residues modified at specific positions (CyaA: 860,983; HlyA: 564,690; RtxA: 558,689) according to the sequence of the full-length proteins were investigated. All assigned peptide sequences, including posttranslational modifications, along with corresponding FASTA formats used within the search algorithm are listed in the supporting MS Data.

Planar lipid bilayers
Measurements on planar lipid bilayers (black lipid membranes) (39) were performed in Teflon cells separated by a diaphragm with a circular hole (diameter 0.5 mm) bearing the membrane. The RTX proteins were prediluted in TUC buffer (50 mM Tris-HCl (pH 8.0), 8 M urea, and 2 mM CaCl 2 ) and added into the grounded cis compartment with a positive potential. The membrane was formed by the painting method using soybean lecithin in n-decane-butanol (9:1, v/v). Both compartments contained 150 mM KCl, 10 mM Tris-HCl (pH 7.4), and 2 mM CaCl 2 , and the temperature was 25°C. The membrane current was registered by Ag/AgCl electrodes (Theta) 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, ;400 individual pore openings were recorded, and the dwell times were determined using QuB software (40) with a 100-Hz low-pass filter. The kernel density estimation was fitted with a double-exponential function using Gnuplot software. The relevant model was selected by the x 2 value.

Cell-binding and cell-invasive activities on sheep erythrocytes
AC enzyme activities of the CyaA variants were measured in the presence of 1 mM calmodulin as described previously (41). One unit of AC activity corresponds to 1 mmol of cAMP formed/min at 30°C, pH 8.0. Cell-invasive AC activity was determined in TNC buffer (50 mM Tris-HCl (pH 7.4), 150 mM NaCl, and 2 mM CaCl 2 ) as the amounts of the AC enzyme protected against inactivation by externally added trypsin upon internalization into sheep erythrocytes, as described previously (42). Erythrocyte binding of the CyaA variants was determined in TNC buffer as described previously (42). Activity of CyaCactivated CyaA was taken as 100%.

Hemoglobin release assay
Sheep erythrocytes stored in Alsever's solution (Sigma-Aldrich) were repeatedly washed with TNC buffer. Washed erythrocytes (5 3 10 8 /ml) were then incubated with various acylated CyaA, HlyA, and RtxA variants in 1 ml of TNC buffer, and hemolytic activity was measured in time by photometric determination (A 541 ) of the hemoglobin release.
Binding of CyaA to J774A.1 cells and determination of cAMP levels Prior to assays, RPMI was replaced with DMEM (which contains 1.9 mM Ca 21 ) without fetal calf serum, and the cells were allowed to rest in DMEM for 1 h at 37°C in a humidified 5% CO 2 atmosphere (43). J774A.1 cells (1 3 10 6 ) were incubated in DMEM with 1 mg/ml of the CyaA variants for 30 min at 4°C, prior to removal of unbound toxin by three washes in DMEM. After the transfer to the fresh tube, the cells were lysed with 0.1% Triton X-100 for determination of cell-bound AC enzyme activity. For intracellular cAMP assays, 1.5 3 10 5 cells were incubated at 37°C with the CyaA variants for 30 min in DMEM, the reaction was stopped by the addition of 0.2% Tween 20 in 100 mM HCl, samples were boiled for 15 min at 100°C and neutralized by the addition of 150 mM unbuffered imidazole, and cAMP was measured by a competitive immunoassay (42). Activity of CyaC-activated CyaA was taken as 100%.

Cell viability
Cell viability following exposure to the toxin was determined as the capacity of mitochondrial reductases to convert the tetrazolium salt WST-1 to formazan, using the WST-1 assay kit (Roche Applied Science) according to the protocol of the manufacturer.

Data availability
The MS data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD018859 (44).