The Conserved Lysine 860 in the Additional Fatty-acylation Site of Bordetella pertussis Adenylate Cyclase Is Crucial for Toxin Function Independently of Its Acylation Status*

The Bordetella pertussis RTX (repeat in toxin family protein) adenylate cyclase toxin-hemolysin (ACT) acquires biological activity upon a single amide-linked palmitoylation of the ε-amino group of lysine 983 (Lys983) by the accessory fatty-acyltransferase CyaC. However, an additional conserved RTX acylation site can be identified in ACT at lysine 860 (Lys860), and this residue becomes palmitoylated when recombinant ACT (r-Ec-ACT) is produced together with CyaC in Escherichia coli K12. We have eliminated this additional acylation site by replacing Lys860 of ACT with arginine, leucine, and cysteine residues. Two-dimensional gel electrophoresis and microcapillary high performance liquid chromatography/tandem mass spectrometric analyses of mutant proteins confirmed that the two sites are acylated independently in vivo and that mutations of Lys860 did not affect the quantitative acylation of Lys983 by palmitoyl (C16:0) and palmitoleil (cis Δ9 C16:1) fatty-acyl groups. Nevertheless, even the most conservative substitution of lysine 860 by an arginine residue caused a 10-fold decrease of toxin activity. This resulted from a 5-fold reduction of cell association capacity and a further 2-fold reduction in cell penetration efficiency of the membrane-bound K860R toxin. These results suggest that lysine 860 plays by itself a crucial structural role in membrane insertion and translocation of the toxin, independently of its acylation status.

RTX (repeats in toxin) protein family (6) and has the capacity to form small cation-selective membrane channels, which account for its weak hemolytic activity (8 -11). The major cytotoxic activity of the 1706-residue-long protein, however, consists in its capacity to invade a variety of eukaryotic cells directly across their cytoplasmic membrane (12)(13)(14) and to deliver into cells a catalytic adenylate cyclase (AC) domain. This intoxicates cells by unregulated conversion of ATP to cAMP (15-18) and causes impairment of microbicidal functions of immune effector cells and apoptosis of lung macrophages (19).
The capacity of ACT to penetrate into target cell membranes and to intoxicate cells depends on a posttranslational activation by the accessory protein, CyaC (20,21). It was first established for the Escherichia coli ␣-hemolysin (HlyA) that the activation of RTX toxins consists in amide linked fatty-acylation (22), and Hackett et al. (23) have demonstrated by mass spectrometric analysis that native ACT produced by Bordetella (Bp-ACT) is mono-acylated by a palmitoyl residue at the ⑀-amino group of lysine 983. In contrast, the HlyA from E. coli was found to be acylated at two lysines, both in vitro and in vivo (24,25). Moreover, two highly conserved RTX acylation sites (25), corresponding to lysine 983 (Lys 983 ) and lysine 860 (Lys 860 ) are also found in ACT, and for an unknown reason, the CyaCactivated recombinant ACT produced in E. coli (r-Ec-ACT) is palmitoylated also at Lys 860 , in addition to acylation of Lys 983 (26). When compared with the native mono-acylated Bp-ACT, the doubly acylated r-Ec-ACT exhibits about four times lower specific hemolytic activity on sheep erythrocytes (20) and about 10 times lower specific channel-forming activity in artificial planar lipid bilayers (10). At the same time, however, the characteristics of the channels formed by both proteins are identical, and both proteins have identical capacity to insert into the target membranes and to deliver the invasive AC domain into cells (5,20,26). In contrast to membrane insertion and AC translocation into cells, the channel-forming activity of ACT exhibits a cooperative concentration dependence curve, indicative of an oligomerization step involved in ACT channel formation (5,8,26). On this basis, a hypothesis was formulated that the acylation of Lys 860 might interfere selectively with the oligomerization of ACT within target membranes (26).
Interestingly, the capacity of ACT to induce protective immune response against Bordetella infection in mice also depends on the CyaC-mediated acylation, and the doubly acylated r-Ec-ACT exhibits significantly lower protective antigenicity than the mono-acylated Bp-ACT (5). These differences in immunological properties of native and recombinant ACT also indicate that acylation at Lys 860 may alter the conformation of the toxin.
We hypothesized that preventing acylation at Lys 860 by sub-stitution of the lysine 860 residue might result in production of native-like and fully active recombinant ACT in E. coli. In this report we demonstrate, however, that lysine 860 is by itself important for toxin action.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Growth Conditions, and Plasmids-The E. coli K12 strain XL1-Blue (Stratagene) was used throughout this work for DNA manipulation and for expression of ACT. Bacteria were grown at 37°C in LB medium supplemented with 150 g/ml of ampicillin. pT7CACT1 is a construct derived from pCACT3 (5), allowing enhanced production of recombinant CyaC-activated ACT in E. coli (r-Ec-ACT) under control of the IPTG-inducible lacZ p promoter. To construct pT7CACT1, first the cyaC open reading frame was amplified by PCR, using purpose-designed primers (5Ј-GCGCGCGCCATATGCTTCCGTC-CGCCCAAG and 5Ј-CCCGGGGGATCCTTAGGCGGTGCCCCGCGGT-CG, respectively). The PCR product was fused to the translational enhancement and initiation signals of gene 10 from phage T7 by cloning into the NdeI and BamHI sites of pT7-7 (27). Absence of undesired mutations was verified by sequencing, and the XbaI fragment containing the gene together with the expression signals was inserted into a blunted HindIII site of pDLACT1 (5), to substitute the cyaC allele in pCACT3, thereby placing it under lacZ p control and yielding pT7CACT1.
Site-directed Mutagenesis of cyaA-The substitutions of the codon for Lys 860 were introduced into the cyaA gene by PCR mutagenesis using three pairs of PCR primers, consisting of one common forward primer 5Ј-GTGGTCCTCGACGTCGCCGCC and one mutation-specific backward primer. These were 5Ј-TGTGGTGAATTCAGATCTGCCCGTTTT-CGTGCG for the K860R substitution, 5Ј-TGTGGTGAATTCGCTCAG-GCCTGTTTTCGTGCG for the K860L substitution, and 5Ј-TGTGGTG-AATTCGCTGCAGCCCGTTTTCGTGCG, for the K860C substitution, respectively. Using the proofreading Vent DNA polymerase (New England Biolabs) and appropriate primer pairs, 144-base pair fragments of the cyaA gene with the incorporated mutations were PCR-amplified, purified on agarose gels, and cloned as ClaI-EcoRI fragments into appropriately digested pT7CACT1. The presence of the desired substitutions and the absence of any unrelated mutations were verified by DNA sequencing of the cloned PCR product. In a second step, the complete mutant cyaA alleles were assembled by addition of a 2617base pair EcoRI fragment encoding the COOH-terminal part of ACT, downstream to the mutagenized residue 860.
Plasmids for expression of the truncated mutant ACT were obtained by replacement of the XhoI-ScaI fragments of the respective mutant pT7CACT1 derivatives by the KpnI-ScaI fragment of pACT⌬C1322 (36). This resulted in insertion of a TAA stop codon at the XhoI site, leading to production of truncated ACT proteins (ACT⌬) with a COOHterminal Asp 1008 residue.
The nonacylated variants of the intact and mutant proteins were expressed from plasmids from which the cyaC gene was deleted as an NdeI-BamHI fragment.
Production and Purification of the CyaA-derived Proteins-The wildtype ACT and the different mutant derivatives were produced in the presence, or in the absence of the activating protein CyaC, using the E. coli strain XL1-Blue (Stratagene) bearing the appropriate plasmid derived from pT7CACT1. Exponential 500-ml cultures were induced with IPTG (1 mM), and the extracts of insoluble cell debris after sonication were prepared in 8 M urea, 50 mM Tris-HCl, pH 8.0, 0.2 mM CaCl 2 , as described previously (20). The proteins were further purified by ionexchange chromatography on DEAE-Sepharose and phenyl-Sepharose (Amersham Pharmacia Biotech) as described previously (28). In the final step, the proteins were eluted with 8 M urea, 50 mM Tris-HCl, pH 8.0, and stored frozen. Alternatively, the truncated proteins were purified directly from urea extracts by affinity chromatography on calmodulin-agarose, as described previously (20). For the K860C mutant 10 mM ␤-mercaptoethanol was added into all extraction and purification solutions to avoid cysteine oxidation. SDS-PAGE analysis and determination of protein concentration were performed according to standard protocols (29).
High Resolution Two-dimensional Gel Electrophoresis-The whole cell extracts of the respective clones expressing the truncated proteins were prepared from exponentially growing cultures induced by IPTG (1 mM) for 3 h. Total protein samples (20 g) were analyzed by large format two-dimensional (IEF/SDS-PAGE) gel electrophoresis (30) using the Investigator system (Oxford Glycosystems). In the first dimension, proteins were separated according to their isoelectric point under denaturing conditions, for 16 h at 18,000 V-h, on a high-resolution 18-cmlong capillary tube gels, containing 2.3% ampholites (Sigma) admixed in pH ranges ratio (3-10):(4 -6):(5-7) ϭ 1:2:2. In the second dimension, the proteins were separated by SDS-PAGE on 7.5% acrylamide slab gels (22 ϫ 22 cm). The gels were stained with Coomassie Blue.
Upon initial experiments with purified proteins, the use of whole cell extracts for two-dimensional electrophoresis was preferred, in order to avoid prolonged incubation of ACT in 8 M urea solutions during purification. This was because products of urea decomposition may cause secondary covalent modifications, such as carbamylation, leading to uncontrolled changes of the proteins isoelectric points. Moreover, the whole cell extract samples gave better reproducibility of resolution and protein positioning in two-dimensional electrophoresis of the differently acylated ACT⌬ protein forms. ACT⌬ constituted generally about 5-10% total cell proteins in the sample and could be unambiguously identified on the gels by their characteristic size and by comparison with samples devoid of the ACT⌬-derived material. Moreover, the spots corresponding to the invariant cell proteins could be conveniently used as internal positioning standards for definition of charge/migration differences and relative positioning of the acylated ACT⌬ proteins.
Mass Spectrometric Analysis of Protein Acylation-The location and identity of the acyl modifications were analyzed by liquid chromatography/MS/MS and MALDI-TOF MS techniques essentially as described previously (26). Tryptic and Asp-N fragments of the analyzed proteins were generated according to standard protocols and separated on 50-m inner diameter ϫ 12-cm capillary column packed with Magic 5 m 200 Å C18 material (Michrom BioResources, Auburn, CA), at a flow rate of 150 nl⅐min Ϫ1 , with a 0 -100% acetonitrile gradient (1% acetic acid) over 40 min. Peptide analysis by microcapillary HPLC coupled to a Finnigan TSQ 7000 electrospray tandem quadrupole mass spectrometer was performed as described in detail elsewhere (31,32).
Positive ion MALDI mass spectra were measured on a PerSeptive Voyager DE linear time-of-flight mass spectrometer equipped with a gridless delayed extraction ion source. Ion acceleration voltage was 20 kV, and guide wire voltage was set to 0.02%. For delayed extraction, an 8% kV potential difference between the probe and the extraction lens was applied with a time delay of 20 ns after each laser pulse. Samples in ␣-cyano-4-hydroxycinnamic acid (premix sample preparation technique) (33) were irradiated by 337-nm photons. Typically 20 -50 shots were summed into a single mass spectrum. Spectra were calibrated externally using the average [M ϩ H] ϩ ion of a peptide standard (insulin, Aldrich).
Identities of the C16:0 and C16:1 cis ⌬D9 fatty-acyl groups were confirmed by capillary gas chromatography/MS with electron impact ionization after a one-step extraction and derivatization procedure (34).
Assay of AC, Cell Binding, and Cytotoxic and Hemolytic Activities-Adenylate cyclase activities were measured as described previously in the presence of 1 M calmodulin (35). One unit of AC activity corresponds to 1 mol of cAMP formed per min at 30°C, pH 8.0. Cellinvasive AC was determined as the amount of AC that becomes protected against externally added trypsin upon internalization into erythrocytes in 30 min of incubation (11), and the hemolytic activity was measured as the hemoglobin released upon incubation of washed sheep erythrocytes (5 ϫ 10 8 /ml) with the toxins for 270 min, respectively (11). Erythrocyte binding of the toxins was determined as described in detail previously (36).

RESULTS AND DISCUSSION
Independent Acylation of the Lys 860 and Lys 983 Residues of ACT in Vivo-In contrast to the essential acylation of lysine 983, the additional CyaC-mediated acylation of lysine 860, occurring in E. coli K12 host background, is not required for ACT activity and interferes with the hemolytic (channel-forming) activity of r-Ec-ACT (20,26). It was important to determine whether this aberrant acylation can be prevented by substitutions of lysine 860 and whether the activities of the mutant mono-acylated toxins produced by E. coli would resemble those of the mono-acylated wild-type ACT from Bordetella (Bp-ACT). Therefore, the lysine 860 was replaced with arginine, leucine, and/or cysteine residues by PCR mutagenesis, and the mutant proteins were produced in an E. coli K12 strain (XL1) in the presence, or in the absence, of CyaC.
The acylation status of the obtained mutants was analyzed by two-dimensional electrophoresis (two-dimensional IEF/ SDS-PAGE) by the method of O'Farrell (30). This approach was already successfully used for the analysis of HlyC-dependent acylation of HlyA (24). It allows to detect fatty-acylation of the respective lysines as loss of one or two positive charges, leading to shifts in the isoelectric point of the toxin. Indeed, nonacylated ACT, bi-acylated ACT, and mono-acylated K860R mutant ACT could be resolved by this technique, although less well than reported previously for HlyA (data not shown). Despite use of 18-cm-long high-resolution IEF tube gels, with a pH gradient set from pH 4 to 6, quantitative and reproducible resolution of proACT from the mono-and bi-acylated ACT forms was difficult to achieve. This could be expected, because the size of the ACT protein is almost twice the size of HlyaA (177 versus 110 kDa), and the theoretical isoelectric points of the three ACT forms under denaturing conditions differ only by 0.01 pH unit (4.27, 4.26, and 4.25 for the non-, mono-, and bi-acylated ACT, respectively). To circumvent this problem, the 70-kDa-long, acidic repeats were deleted from the r-Ec-ACT and its mutant variants. The resulting ACT⌬ proteins, with a carboxyl-terminal aspartate residue at position 1008, had theoretical isoelectric points of 5.41, 5.35, and 5.29 in the non-, mono-, and bi-acylated forms, respectively. These proteins could be well resolved by the two-dimensional electrophoresis, as demonstrated in Fig. 1.
The proACT⌬ protein, produced in the absence of CyaC, migrated as a single spot at a position of the nonacylated form, while the ACT⌬, produced in the presence of CyaC, resolved in two spots corresponding to the major, bi-acylated form (about 80% total) and a mono-acylated form (about 20% total) of ACT⌬, respectively (Fig. 1). When the proACT⌬ and ACT⌬ proteins were mixed in a roughly equimolar ratio, two major spots of the bi-acylated and nonacylated forms were resolved with the minor spot of the mono-acylated ACT⌬ protein positioned between them. This observation was in good agreement with the semiquantitative estimations from mass spectrometric analysis of CyaC-activated intact r-Ec-ACT (26), where about 60% r-Ec-ACT was estimated to be bi-acylated and the rest was mono-acylated. These results demonstrate that removal of the repeat portion downstream to residue 1008 did not have any effect on the efficiency of the CyaC-mediated acylation of the Lys 983 and Lys 860 residues of ACT. Therefore, the two-dimensional electrophoresis of the COOH-terminally truncated proteins could be used as a suitable assay for acylation of the Lys 860 mutants of ACT. It is noteworthy, however, that in some gels trailing of a small portion (Ͻ10%) of the ACT⌬derived material was observed and yielded two very minor satellite spots, corresponding to one more and/or one less positive charge than the major protein species in the sample, respectively. It was not determined whether occurrence of these minor spots was due to handling of samples or whether it was due to intrinsic microheterogeneity of the ACT⌬ proteins in the preparations and could be observed due to high resolving power of the applied separation technique.
When the K860R mutant forms of ACT⌬ were produced in the absence (proACT⌬-K860R) and in the presence (ACT⌬-K860R) of CyaC, each migrated as a single major spot in two-dimensional electrophoresis (Fig. 1). When the two proteins were mixed, they resolved quantitatively in two distinct spots separated by a distance corresponding to a single charge difference. Because the K860R substitution does not alter the net charge of the nonacylated protein, this result is consistent FIG. 1.

Analysis of ACT acylation by two-dimensional electrophoresis.
Fifty microgram of total protein extracts of IPTG-induced cells producing the ACT⌬ proteins were prepared and analyzed by two-dimensional (IEF/SDS-PAGE) electrophoresis, as described under "Experimental Procedures." In general, binary or tertiary mixtures of extracts were prepared by directly mixing equal volumes of extracts containing similar amounts of total cell proteins. For extracts with important differences in the relative content of the ACT⌬-derived proteins (i.e. proACT⌬-K860R versus ACT⌬-K860R), the volumes of extracts in the mixture were adjusted to yield comparable levels of the ACT⌬-derived proteins in the mixture. The gels were stained with Coomassie Blue and gel sections comprising only the spots of the various forms of ACT⌬-derived proteins are shown. The sections were aligned in columns so that the position of the spots relative to the invariant cell proteins (positioning standards) in the original individual gels was respected. Therefore, the relative position of the spots in the columns reflects their relative position in respect to the anode (ϩ) and cathode (Ϫ) in the original pH gradients after isoelectric focusing.
with the conclusion that the substitution of lysine 860 by an arginine prevented the acylation at residue 860 and that quantitative acylation of Lys 983 of the mutant protein still occurred in presence of CyaC. Indeed, as expected due to equal charges, the proACT⌬-K860R and proACT⌬ proteins produced in the absence of CyaC co-migrated as a single spot, while ACT⌬-K860R produced in the presence of CyaC resolved quantitatively also from the nonacylated proACT⌬ (Fig. 1). Furthermore, ACT⌬-K860R, if acylated on Lys 983 , is expected to bear one positive charge more than bi-acylated ACT⌬. Indeed, when ACT⌬ and ACT⌬-K860R were mixed, they resolved in two spots, with the ACT⌬-K860R mutant migrating at the position of the mono-acylated form of ACT⌬. Finally, when ACT⌬-K860R protein was mixed with proACT⌬ and ACT⌬, it migrated at the position of mono-acylated form of ACT⌬, in between the nonacylated proACT⌬ and the bi-acylated form of ACT⌬ (Fig. 1). This demonstrated that the K860R mutant was still quantitatively mono-acylated by CyaC, most likely on Lys 983 .
The same acylation pattern was found also with the K860L mutant (Fig. 1). The ACT⌬-K860L protein expressed in the presence of CyaC migrated at the same position as the biacylated ACT⌬. This is consistent with its mono-acylation on lysine 983, because the replacement of lysine 860 by leucine removes one positive charge, as does the acylation of lysine 860. Indeed, the nonacylated proACT⌬-K860L and proACT⌬ resolved as two spots differing by one charge, as did proACT⌬-K860L and ACT⌬-K860L. In contrast, the ACT⌬-K860L produced in the presence of CyaC and the nonacylated proACT⌬ resolved as two spots at a distance corresponding to a difference of two charges. Consistently, the ACT⌬-K860R resolved as a third spot between ACT⌬-K860L and proACT⌬. These results clearly indicate that also the ACT⌬-K860L mutant was quantitatively acylated on lysine 983.
Both Palmitic and Palmitoleic Acids Are Used for Acylation of Lys 983 of ACT-It was important to prove that the acylation of the ACT-K860R mutant was indeed located on lysine 983 and that identical fatty-acyl residues were used for modification of the wild-type and mutant proteins. For this, the same approach was used as that originally applied for determination of the chemical nature and location of the fatty-acyl modification of ACT from Bordetella (23). Tryptic and Asp-N proteolytic fragments of the full-length r-Ec-ACT and of the truncated ACT⌬ forms of both the wild-type and the K860R proteins, respectively, were analyzed for acylation by MALDI-TOF MS and/or microcapillary HPLC coupled to electrospray ionization tandem mass spectrometry, as described previously (26). Representative mass spectra are shown in Fig. 2, and the results of the MS analysis of the various proteins are summarized in Table I. These results fully confirm the acylation patterns observed by two-dimensional gel electrophoresis. As expected, no acylation was detected on the proACT and proACT⌬ proteins produced in the absence of CyaC, while acylation was found on both the Lys 983 and Lys 860 residues of the r-Ec-ACT and ACT⌬ proteins produced in presence of CyaC. More importantly, no acylation of the arginine residue Arg 860 of the ACT-K860R and ACT⌬-K860R mutants could be detected, and the acyl residues in this mutant proteins were entirely located on the Lys 983 residue. Finally, both wild-type and K860R mutant proteins were acylated by palmitoyl (C16:0) and palmitoleil (cis ⌬9 C16:1) fatty-acyl groups. No evidence of myristoylation was observed at either lysine in any of the proteins, quite unlike our previously reported results (26).
The sum of this data demonstrates that the CyaC-mediated acylation of truncated ACT⌬ protein in E. coli is identical to that of the full-length ACT protein. The results show further that two-dimensional electrophoresis of whole cell samples containing the ACT⌬ protein can be used as a straightforward and reliable assay for the acylating activity of CyaC in vivo. Collectively, these results demonstrate that the substitution of the  lysine 860 by arginine did not have any effect on CyaC-mediated acylation of Lys 983 and that acylation of Lys 860 and Lys 983 residues of ACT occurs independently in vivo. This is similar to the previously observed independent acylation at the two sites of HlyA, where only about 20 -30 flanking residues around each acylated lysine were required for productive modification to take place both in vitro and in vivo (24,25). Functional Characterization of the Mutants-For comparison of the biological activities, the mutant proteins and intact r-Ec-ACT were purified close to homogeneity, as shown in Fig.  3. Interestingly, all substitutions, even the most conservative K860R replacement, caused a strong decrease in toxin activities, reducing to about 10% the cell-invasive AC activity of the K860R mutant and to about 5% those of the K860L and K860C mutants, as shown in Table II. The difference between the activities of the K860R mutant and those of K860L and K860C mutants was small, but reproducible. This suggests that presence of a positively charged amino acid at position 860 is important for toxin function, albeit even an arginine residue could not functionally replace the lysine 860.
The mutant proteins were defective in both the target cell association and in membrane translocation of the bound toxin, as shown further in Table II. Tight cell association was reduced to 23% for the K860R mutant and to 15% for the other two mutants, as compared with intact r-Ec-ACT. Moreover, while under given conditions about 60% cell-bound r-Ec-ACT had the AC domain translocated into cells and had the AC domain protected against externally added trypsin, only 30% bound K860R and 17% K860L and K860C ACT, respectively, was translocated under identical conditions. Hence, the translocation efficiency of K860R ACT was half, and that of both K860L and K860C was about 28%, of the translocation efficiency of the intact r-Ec-ACT, respectively. These data suggested that lysine 860 plays an important role in both the membrane insertion of ACT and translocation of its AC domain into cells.
The acylation of Lys 860 affects selectively the hemolytic (channel-forming) activity of the toxin, which exhibits different calcium requirements as the invasive AC activity (37)(38)(39)(40). It was, therefore, important to compare the calcium and concentration dependencies of the K860R mutant with those of intact r-Ec-ACT. No substantial difference of the two proteins in function of free calcium concentration could, however, be observed when the amount of the K860R protein in the comparative assays was raised seven times in respect to intact ACT, in order to measure both toxin activities in a similar readout range (Fig. 4).
It appears, therefore, that the K860R mutant toxin still exhibits similar calcium requirements for action as the intact ACT. This indicates that the K860R substitution did not affect the toxin structure in a gross way, causing generalized misfolding of the protein. Such effect of a conservative substitution appears unlikely. Moreover, in contrast to what would be expected for a misfolded protein, no difficulties with mutant toxin production and/or reduced stability during purification were observed (Fig. 3), indicating that the sensitivity of the mutant ACT to the endogenous proteases of E. coli was not increased and the yields of intact and mutant toxins were also very similar (data not shown). The residual activity of the K860R mutant reaching up to 10% that of intact ACT also suggests that the mutation caused a specific defect of membrane interaction efficiency of ACT.
Conclusions-The results reported here suggest that the lysine 860 (Lys 860 ) residue is by itself an important functional residue involved in membrane insertion and translocation of ACT, regardless of its acylation status. It is intriguing that even a conservative substitution of Lys 860 by an arginine resi-due caused a strong decrease in all toxin activities, while modification of Lys 860 in E. coli by a bulky and highly hydrophobic palmitoyl residue causes only a selective reduction of the channel-forming activity of r-Ec-ACT and does not affect the membrane insertion and AC delivery capacities of the toxin (5, 10, 20, 26).
Acylation of Lys 860 is not required for ACT function (23,26). FIG. 4. Effect of calcium on biological activity of the ACT-K860R mutant toxins. Sheep erythrocytes (5 ϫ 10 8 /ml) were incubated at 37°C with purified ACT (ࡗ, 0.5 unit/ml) or ACT-K860R (Ⅺ, 3.3 units/ml) in 10 mM Tris-HCl, pH 8.0, 150 mM NaCl, and various free calcium concentrations. After 30 min aliquots were taken for determination of tightly cell-associated AC activity (A) and of the AC activity internalized into erythrocytes and protected against externally added trypsin (B). The hemolysis (C) was measured photometrically at 541 nm, as hemoglobin released after 270 min of incubation. Each point is the mean of duplicate determinations, and this graph is representative of three separate experiments. In order to compensate for the lower biological activity of the mutant toxin and to obtain for both toxins a readout in the same range, the input concentration of the ACT-K860R mutant was 7-fold higher as that of intact ACT (3.3 units/ml versus 0.5 unit/ml, respectively).
In the E. coli ␣-hemolysin (HlyA), however, the corresponding Lys 563 residue (Lys 564 in some alleles) at the first acylation site of HlyA was shown to be preferentially acylated by HlyC in vitro (25), and it is quantitatively acylated also in vivo (24). Interestingly, like with the ACT-K860R mutant described here, some of the Lys 563 substitutions in HlyA, such as Lys 563 3 Cys, still allowed rather high hemolytic (20%) and especially leukotoxic activities of HlyA (60%) (41). Moreover, some reported substitutions at the second acylation site of HlyA on Lys 689 (Lys 690 in some HlyA variants), such as Lys 689 3 Arg, also allowed substantial hemolytic (11%) and leukotoxic (41%) activities of HlyA (41). The requirement for quantitative acylation of HlyA on both the Lys 563 and Lys 690 was also recently questioned by the analysis of HlyA acylated by mutant HlyC (42). Guzmá n-Verri et al. (42) demonstrated that HlyA preparations, which contain only trace amounts of doubly acylated HlyA and consist predominantly of a mixture of mono-acylated HlyA with the nonacylated proHlyA, still exhibited significant hemolytic activities, ranging from 27 to 60% of the activity of preparations containing exclusively the doubly acylated HlyA. It was, however, not investigated whether only one of the two potential acylation sites was preferentially modified in the mono-acylated HlyA or whether the produced HlyA was a mixture of toxins acylated at either the Lys 564 or the Lys 690 residues. These observations indicate, nevertheless, that as in the case of Lys 860 substitution in ACT, the loss of activity upon substitution of Lys 563 (Lys 564 ) in HlyA might also be due to structural effect of replacement of an essential residue, rather than a mere consequence of the loss of the modification at the first acylation site of HlyA (Lys 563 ).
It remains unclear why double acylation of HlyA at Lys 563 and Lys 689 is optimal for hemolytic and cytotoxic activity of HlyA, while single acylation at the corresponding Lys 983 is optimal for full hemolytic activity of B. pertussis ACT. The different acylation patterns of the recombinant r-Ec-ACT and the native Bp-ACT show that maximal caution and detailed characterization of fatty-acylated recombinant RTX toxins are required prior to their use in functional studies. Nevertheless, understanding of reasons why the second CyaC-mediated acylation of ACT on Lys 860 occurs only in E. coli K12 would undoubtedly contribute important new insights into the process of fatty-acyl activation of RTX toxins. It will be important do decipher the mechanism by which the acylation of Lys 860 interferes selectively with only the hemolytic (channel-forming) activity of the toxin and not with its membrane insertion and translocation capacities. Detailed analysis of this intriguing artifact is likely to advance the understanding of mechanism of action of ACT within target membranes. It is intriguing that despite identical culture protocols, expression plasmids, and E. coli strain used for production of ACT, myristoylation of Lys 983 was not present in the new batches of toxins analyzed here, while myristoyl was found to constitute about 13% fatty acid residues bound to Lys 983 in previous batches of recombinant ACT (26). Similarly, presence of palmitoleil (cis ⌬9 C16:1) fatty-acyl groups was not found in previously analyzed batches of ACT (26). It can only be speculated that subtle differences in the physiological state of the producing culture, due to limited reproducibility of different batches of complex culture media and/or culture conditions, may affect the exact chemical nature of acyl groups used for modification of ACT by CyaC. This hypothesis would deserve a systematic testing. Our results suggest a remarkable flexibility of the modifying enzyme (CyaC) in using fatty-acyl substrates and indicate that various acyl residues may also be attached to naturally produced RTX toxins. It will be important to evaluate the significance of such mixed acylation for biological activity and function of these toxins in bacterial pathogenesis.