Escherichia coli K-12 carries a gene for a protein denoted ClyA or SheA that can mediate a cytolytic phenotype. The ClyA protein is not expressed at detectable levels in most strains of E. coli, but overproduction suitable for purification was accomplished by cloning the structural gene in an hns mutant strain. Highly purified ClyA protein was cytotoxic to macrophage cells in culture and caused detachment and lysis of the mammalian cells. Results from osmotic protection assays were consistent with the suggestion that the protein formed pores with a diameter of up to 3 nm. Using Acholeplasma laidlawii cells and multilamellar liposomes, we studied the effect of ClyA on membranes with varying compositions and in the presence of different ions. ClyA induced cytolytic release of the fluorescent tracer from carboxyfluorescein-loaded liposomes, and the release was stimulated if cholesterol was present in the membranes whereas addition of calcium had no effect. Pretreatment of the ClyA protein with cholesterol inhibited the pore formation, suggesting that ClyA could bind to cholesterol. Efficient coprecipitation of ClyA with either cholesterol or 1,2,3-trioctadecanoylglycerol in aqueous solutions showed that ClyA directly interacted with the hydrophobic molecular aggregates. We tested the possible functional importance of selected ClyA protein regions by site-directed mutagenesis. Defined mutants of ClyA were obtained with alterations in postulated transmembrane structures in the central part and in a postulated membrane-targeting domain in the C-terminal part. Our results were consistent with the suggestion that particular amphiphilic segments are required for ClyA activity. We propose that these domains are necessary for ClyA to form pores.
Haemolysins are cytolytic toxins that can be found in a wide range of microorganisms. Although these molecules are named for their ability to lyse erythrocytes, many are also toxic to other cell types. Based on the mechanism of action against target cell membranes, cytolytic toxins can be divided into three subtypes (Rowe and Welch, 1994): (i) hose showing enzymatic mechanisms, in which cytolysis is caused by an enzymatic disruption of the target cell membrane; (ii) the detergent-like action of surfactant haemolysins; and (iii) the pore-forming cytolysins, which insert into the target cell membrane and form transmembrane pores. Such cytolytic toxins presumably contribute to the virulence of many pathogenic bacteria. However, recently it became evident that even the common laboratory strain of Escherichia coli K-12 can express a haemolytic/cytolytic phenotype. Bacteria expressing a 34 kDa protein, which we denoted ClyA (and which is also known as SheA), show haemolytic/cytolytic toxicity towards erythrocytes and towards different cultured mammalian cells, e.g. HeLa cells and murine-derived macrophages (Oscarsson et al., 1996; del Castillo et al., 1997; Y. Mizunoe, J. Oscarsson and B. E. Uhlin, unpublished). Apparently, the latent/silenced structural gene in E. coli K-12 can be activated by several means, e.g. by mutation of the hns locus, by overproduction of the slyA gene product, by overproduction of the mprA gene product and by overproduction of the HlyX protein from Actinobacillus pleuropneumoniae (Libby et al., 1994; Uhlin and Mizunoe, 1994; Ludwig et al., 1995; Gómez-Gómez et al., 1996; Oscarsson et al., 1996; del Castillo et al., 1997; Green and Baldwin, 1997). The actual function of the 34 kDa protein in E. coli remains to be elucidated. Because it confers a cytolytic activity towards different cells and membranes, we will here refer to it as a cytolysin using the abbreviation ClyA as before (Oscarsson et al., 1996). The ClyA polypeptide appeared as a novel cytolysin with little apparent resemblance to previously characterized cytolysins, and it prompted studies to elucidate its mode of action and the structural/functional features of the protein.
Isolation of native ClyA cytolysin
As reported in brief previously, the chromosomal DNA region containing the clyA structural gene from E. coli K-12 was cloned and expressed using the pUC18 plasmid vector (Uhlin and Mizunoe, 1994; Oscarsson et al., 1996). Expression of the cloned clyA gene was tested in different genetic backgrounds and was found to be at the highest levels in the hns mutant strain YMZ4/pYMZ80 (Fig. 1A). In SDS–PAGE, the ClyA protein migrates with an apparent molecular weight of ≈34 kDa (Oscarsson et al., 1996; del Castillo et al., 1997). Using the overproducing strain, we developed a rather quick and efficient procedure for purification of the ClyA protein. The crude preparation of native cytolysin A protein was exposed to electrophoresis on a 6% cylindrical PAGE Bio-Rad Prep Cell (see Experimental procedures ), and several of the resulting fractions showed haemolytic activity when spotted on blood agar. Selected fractions containing the major activity were pooled together, and this preparation exhibited one major protein band of approximately Mr 34 000 according to Coomassie blue staining of SDS polyacrylamide gels (data not shown). A few additional very faint bands could be detected by silver staining of the gels (Fig. 1B), but such proteins were also present in fractions that showed no haemolytic activity and we concluded that they did not contribute to the cytolysin activity. Furthermore, we analysed the N-terminal amino acid sequence of the ClyA protein preparations obtained from polyacrylamide gels and from the purification. The sequencing revealed the following N-terminal amino acid residues: Thr–Glu–Ile–Val–Ala–Asp–Lys–Thr–Val–Glu–Val–Val–Lys–Asn–Ala–Ile–Glu. The sequence was in agreement with the prediction made on the basis of DNA sequencing. It also showed that the structural gene sequence starts at the second of two possible ATG codons (Blattner et al., 1997; del Castillo et al., 1997; this study). The initial Met residue was presumably removed soon after synthesis because no such residue was detected in any of the protein preparations analysed. We concluded that the translated ClyA polypeptide contains 303 amino acid residues rather than the 305 predicted on the basis of genome sequencing (Blattner et al., 1997). The final protein product thereby contained 302 amino acid residues. Studies in vitro with the purified ClyA protein supported the previous findings of del Castillo et al. (1997) with S30 extracts separated on non-dissociating discontinuous polyacryalmide gels, and the studies provided direct evidence that this polypeptide by itself had lytic activity. The indication of a pore-forming activity was previously reported from studies with extracts of E. coli carrying the cloned slyA locus (Ludwig et al., 1995). However, the actual lytic component in such extracts was not directly identified at that time. Furthermore, when the present manuscript was under revision, Ludwig et al. (1999) reported that a ClyA preparation obtained by excising the protein from a SDS–PAGE gel generated pores in planar lipid bilayer membranes. As shown below, our results with native purified ClyA are in accordance with the suggestion about pore-forming activity made by Ludwig et al. (1995; 1999).
Osmotic protection assays
Using the procedure described by van Leengoed and Dickerson (1992), we measured the specific activity of ClyA on horse erythrocytes. Expressed as arbitrary haemolytic units (HU), the ClyA protein showed an activity of ~ 200 HU mg−1 according to our measurements with two independent cytolysin preparations. We used an osmotic protection assay to assess the putative pore-forming activity of purified ClyA protein. For this assay, it has been reported that 30 mM raffinose affords no protection against HlyA haemolysin-induced lysis in a 45 min haemolysis assay, whereas dextran 4, a larger molecule, prevents lysis (Bhakdi et al., 1986). We performed haemolysis assays using the native preparation of ClyA, and as osmotic protectants we tested glucose, maltose, raffinose, dextrin 20, dextrin 15 and dextran 4. As shown in Fig. 2, our results fully supported the suggestion concerning pore-forming activity that was made by Ludwig et al. (1995; 1999), i.e. the cytolytic action of the purified ClyA was significantly reduced by dextran 4 but not by the other sugars. Lobo and Welch (1994) suggested that the molecular size of the protecting agent has to be larger than the size of the pore formed by the toxin for protection to occur. The present findings with ClyA indicate that the pores formed are smaller than 3.0–3.5 nm, the molecular diameter of dextran 4, but bigger than 2.2 nm, the molecular diameter of dextrin 15.
Macrophage cytotoxicity and detachment
We studied the effect of purified ClyA protein on the murine macrophage-like cell line J774, as explained in Experimental procedures, and results from cytotoxicity and cell detachment experiments correlated well. As monitored with the LDH assay, ClyA protein at a concentration of 0.9 μg ml−1 showed a clear cytotoxic activity; it induced an approximately 70% release of the LDH. The cytotoxic effects were also monitored by microscopy, and at high concentration we observed substantial detachment of the target cells (Fig. 3). Swanson (1989) showed that phorbol esters stimulate macropinocytosis and solute flow through macrophages. We found a considerable increase of detachment of the J774 target cells after the addition of 20 μg ml−1 of ClyA protein to cells treated with phorbol 12-myristate 13-acetate (PMA) at a final concentration of 150 ng ml−1 (Fig. 3). Taken together, we found that purified ClyA showed both cytotoxic and cell-detaching activity on J774 murine macrophage-like cells.
ClyA-mediated permeabilization of Acholeplasma laidlawii cells
Acholeplasma laidlawii represents a usable cell model in which the lipid bilayer composition can be altered by varying the growth conditions of the organism; the membranes are also enriched in glucolipids. Because of the lack of a protective cell wall mesh, osmotic swelling of the cells can be easily monitored (McElhaney, 1993). To test whether the presence of cholesterol in the target cell membranes affects the cytolysin activity, a preparation of ClyA was added to growing cultures of cells with 18:1c acyl chains and 0, 20 or 50 μM medium cholesterol. Aliquots of growing A. laidlawii culture cells, containing different membrane acyl chains and cholesterol amounts, were mixed with ClyA protein and incubated at room temperature. At intervals, the morphology and the size of the A. laidlawii cells were analysed in a phase-contrast microscope. The effect of the ClyA cytolysin on the three cultures was different, as monitored by cell swelling under the microscope. The effect was increasingly severe on cells with increasing cholesterol content. Such cultures showed a rather high frequency of enlarged cells (approximately half of the cells had about twice the normal cell diameter) when incubated with the purified cytolysin. The observed swelling is consistent with increased osmotic permeability caused by the action of the ClyA protein. When purified SlyA protein (the regulatory protein here used as a control) from E. coli K-12 was added in the same buffer to the cultures instead of ClyA, no effect on any of the cultures was detected (data not shown). The effect of ClyA on substantially thinner A. laidlawii lipid bilayers, i.e. with 14:0 plus 14:1c acyl chains (Wieslander et al., 1995), was essentially similar to the thicker 18:1c ones. Taken together, these results suggest that ClyA is active on glucolipid-enriched membranes, and may be stimulated by the presence of cholesterol in the target cell membrane.
Cholesterol-stimulated action of ClyA in liposomes
To investigate whether native cytolysin A could act on artificial membranes of various composition, we utilized multilamellar liposomes loaded with the fluorescent chromophore carboxyfluorescein (CF) in an assay to monitor release after treatment with the preparation of ClyA protein at room temperature. As shown in 4Fig. 4A, ClyA protein induced CF release from liposomes made from total lipid extracts from human erythrocytes, and with a composition of phospholipids/cholesterol at a molar ratio of 0.7/0.3. The assay was performed with ClyA concentrations of 1.2 μg ml−1 and 2.4 μg ml−1, and the activity was clearly dose dependent. As shown in 4Fig. 4B, liposomes composed of synthetic di18:1c-phosphatidylcholine (PC)/di18:1c-phosphatidylglycerol (PG) in a molar ratio of 0.9:0.1 were almost unaffected compared with liposomes composed of di18:1c-PC/di18:1c-PG/cholesterol in a molar ratio of 0.675:0.075:0.25. Similar results with respect to the stimulatory effect of cholesterol, however, with lower levels of CF release were achieved using liposomes composed of total lipid extracts from A. laidlawii grown in a broth with and without 25 μM of cholesterol (Fig. 4C). A. laidlawii membranes with 18:1c acyl chains contain ≈35 mol per cent anionic lipids (including PG), the balance being glucolipids. This anionic fraction is increased somewhat when cholesterol is present (see Wieslander et al., 1995); here, the cholesterol concentration is ≈25 mol per cent. We attribute the generally lower efficiency of ClyA on the A. laidlawii-derived liposomes to their increased content of anionic lipids or to the presence of the glucolipids. No effect on the ClyA action was noted in the presence of 10 mM CaCl2 in the working buffer (data not shown). Taken together, the above results suggest that the ClyA activity was stimulated by the presence of cholesterol in the target cell membranes, but the purified toxin did not seem to require Ca2+ for activity.
Cholesterol interaction by ClyA
Some cytolytic bacterial toxins are known as cholesterol-binding toxins and the biological properties of such toxins may be irreversibly lost in the presence of very low concentrations of cholesterol and other related 3β-hydroxysterols, which interfere with toxin binding to target cells (Alouf and Geoffroy, 1991). To test whether the presence of cholesterol might affect the ClyA-mediated cytolytic activity, a haemolytic blocking assay as described by Jarvill-Taylor and Minion (1995) was used, and we also studied 1,2-dioleoyl-phosphatidylcholine (DOPC) as an alternative to cholesterol. For this purpose, we monitored the cytolytic activity of ClyA-expressing bacterial cells on erythrocytes. As shown in 5Fig. 5A, a cholesterol dose-dependent decrease of haemolytic activity was observed when ClyA extract and cholesterol were mixed and incubated for 30 min at room temperature before the addition of erythrocytes. Such a decrease was not seen when erythrocytes and cholesterol were first mixed and incubated for 30 min at RT before the addition of the ClyA preparation, or when DOPC was used instead of cholesterol. These findings indicate that the cytolytic action of ClyA can to some extent be blocked by the addition of free cholesterol.
To test whether cholesterol could directly bind the free ClyA in sonic extracts, we performed the assay so that we could monitor whether coprecipitation with cholesterol occurred in the aqueous solution. After allowing the cholesterol to precipitate, we measured the remaining ClyA activity. As shown in 5Fig. 5B, the cytolytic activity of the supernatants of the strain MC1061/pYMZ80 (obtained as described in the Experimental procedures section) could be more or less completely removed by treatment with increasing concentrations of cholesterol. Furthermore, the ClyA activity could also be removed to a large extent by treatment with 1,2,3-trioctadecanoylglycerol (tristearate). Analysis by Western immunoblotting with ClyA-specific antisera confirmed that there was coprecipitation; both the ClyA–cholesterol and ClyA–tristearate pellets showed a reactive band of 34 kDa corresponding to the ClyA protein (data not shown). The results clearly indicated that the ClyA protein by itself has the ability to bind to hydrophobic aggregates of cholesterol or tristearate. This binding, therefore, did not appear to be very specific, but the results were consistent with the observation that cholesterol in membranes may stimulate ClyA lytic action.
Potential functional domains of ClyA
The finding that the ClyA protein, by itself, interacted with lipid bilayer/membranes prompted us to consider whether there are sequence features in the protein that may explain its activity. Homology searches using updated blast and FASTA algorithms (Pearson and Lipman, 1988; Altschul et al., 1990) revealed that the 303-amino-acid ClyA protein has no large similarity to existing bacterial toxins (data not shown). However, the score lists were headed by proteins strongly enriched in α-helices and coiled structures, such as tropomyosin and apolipophorin. Fold prediction analyses also suggested similarities to α-helix-enriched proteins (e.g. to cytochrome c, bacteriorhodopsin, GroEl, and annexin XII), but slightly below the significance thresholds. Analysis of the hydrophobicity and the amphilicity (hydrophobic moment) (Eisenberg et al., 1984) along the sequence proposed two, hydrophobic, potential transmembrane (TM) segments, also shown by del Castillo et al. (1997): TM-1 at positions 83–103 (‘putative’) and TM-2 at positions 179–205 (‘certain’), and seven amphiphilic segments of at least 10 amino acids in length. The last had average hydrophobic moments from 0.50 to 0.71 according to the Eisenberg scale (Fig. 6A). This is typical for amphiphilic peptides or segments with one hydrophobic and one polar (or charged) longitudinal face, and the range of hydrophobic moments is similar to that for several lytic amphiphilic antibacterial peptides and three helices in the membrane-active Colicin A (Eisenberg and Wesson, 1990).
The mechanism by which ClyA may be released from the bacterial cell is at present unknown. The ClyA amino acid sequence contains no typical N-terminal signal peptide (also noted by del Castillo et al., 1997) or cleavage site, no N-terminal lipid modification site (according to prosite), no conserved Gly–Lys sequences for fatty acid modification typical for bacterial RTX toxins such as the E. coli HlyA haemolysin (Stanley et al., 1996), nor the Gly-rich Ca2+-binding repeat domain of HlyA. Furthermore, the C-terminal secretion signal in HlyA (Stanley et al., 1996) is not similar to the sequence at the ClyA C-terminus.
It has been proposed from analysis of several recently sequenced genomes that cellular proteins are based upon a limited set of domains or modules (Riley and Labedan, 1997; Teichmann et al., 1998). Division of the ClyA amino acid sequence into seven partly overlapping peptide fragments of similar size yielded slightly better statistics for the sequence similarity searches using the FASTA procedure, finding several well-known helix/coil proteins as described above. However, included among these were also several helical and amphiphilic peptide segments with established structures and membrane-binding properties. Most striking was a good correspondence (58% similarity) between positions 266 and 291 in the C-terminal of ClyA to the lytic 26-amino-acid peptide delta-toxin from Staphylococcus aureus (Fig. 6B). This is a secreted peptide, forming an amphiphilic α-helix on a target membrane. The distribution of the negative and positive amino acid charges in the N-terminus and the C-terminus, respectively, is evidently essential for the lytic, pore-forming properties of the delta-toxin (Dhople and Nagaraj, 1995). In addition to the striking similarity, we noted that there was a more pronounced bias of negatively and positively charged amino acids along this domain of the ClyA protein than in the delta-toxin (Fig. 6C). Although the sequence similarity to other toxins was rather low when analysed over their full lengths, a one-by-one alignment of a sublibrary with 30 bacterial toxin sequences of ClyA and its seven overlapping fragments (using the pileup program) grouped all of the last together with the S. aureus peptide delta-toxin, the pore-forming RTX haemolysins (cf. HlyA above) and the lytic peptide melittin (data not shown).
To test the potential functional significance of the predicted transmembrane helices and the amphiphilic C-terminus, the structural gene was subjected to site-directed mutagenesis and the cytolytic activities of the resulting variants were measured with the erythrocyte assay. We used a PCR-based strategy (see Experimental procedures ), and all constructs were confirmed by DNA sequence analysis. All mutant cytolysins were found to be expressed at detectable levels, and the cellular amounts were in most cases similar to that of wild-type ClyA as shown by Western immunoblotting (Figs 7 and 8). First, we investigated the potential transmembrane helices at positions 83–103 (TM-1) and 179–205 (TM-2) by introducing the mutations G88 → D and V89 → D (plasmid pJON75), A185 → D and G186 → D (pJON70), and A189 → D and G190 → D (pJON71). As shown in Table 1, the amino acid changes in the TM-1 region led to a reduced (only 25%) activity in the sonicated, whole cell extracts, and virtually no activity was exhibited by intact bacteria. Similar results were obtained after alteration of the amino acids N143 → D and A144 → D (pJON76) in the loop between the two predicted transmembrane helices. The amino acid changes within the TM-2 region resulted in the most drastic effect. Substitution to aspartic acid residues at positions A185 → D and G186 → D (pJON70) or at positions A189 → D and G190 → D (pJON71) led to total loss of cytolytic activity. These findings were consistent with the suggestion that the predicted transmembrane helices may be essential for the pore-forming activity of ClyA.
Table 1. . Effect of mutations in a predicted transmembrane domain. a. Relative lytic activity was determined as described in Experimental procedures.
We subsequently investigated the ClyA C-terminus and obtained a number of different mutant variants (Table 2). We made substitutions at several positions in the region of residues 268–294 by introducing alanine residues, and, fortuitously, we also obtained double or triple substitutions and variants that produced truncated ClyA proteins. The most pronounced effect was seen in cases in which part of the C-terminal region was deleted. By removing the last 23 amino acids (pJON66), the cytolytic activity was totally lost, in comparison with assays with whole cell extracts. A mutant lacking the last 12 amino acids (pJON63) retained only about 15% of the activity. Of the different alanine substitutions created in the region 268–294, most mutants retained cytolytic activity at the level of the wild type (Table 2). However, in some cases, the cytolytic activity expressed by intact bacteria appeared reduced whereas the total activity detected in sonic extracts was virtually unaltered (e.g. D268 → A, K275 → A and H292 → A), indicating that the mutant peptides were not effectively translocated in the cells. In the case of H292 → A (pJON69), there was slightly less protein in the sonicated preparation, which could explain the reduced lytic activity (86%). However, we conclude that the reduced relative activity seen with intact bacteria must mainly be a direct effect of the mutation. Reduced activity in both sonic extracts and intact bacteria was found in the case of a mutant with two substitutions near the immediate C-terminus (pJON74; G293 → A and K294 → A), and there was only about 40% of the total activity remaining. Similarly, a mutant with both substitutions and a deletion in the 288–291 region (pJON50) was defective in total activity (about 50% remaining), and there was no cytolytic activity seen with intact bacterial cells. We also assessed the activity of mutant cytolysins as a function of time. In addition to the reduced total activity observed after 120 min, there seemed to be a distinct delay in the onset of cytolysis by several of the tested mutants (Fig. 9). Taken together, our results provide evidence that the predicted transmembrane helix at position 179–205 (TM-2) is essential for the cytolytic activity, and that the amphiphilic C-terminus may have an important role in pore formation by ClyA.
Table 2. . Effect of mutations in the C-terminal region of ClyA. a. Relative lytic activity was determined as described in Experimental procedures.
The discovery that even E. coli K-12 strains, normally viewed as non-pathogenic, carry the potential to be cytolytic towards mammalian cells has raised many questions regarding the function and mode of action of the cytolytic protein. That the cytolysin could be readily overproduced for purification and for in vitro studies enabled us to assess several aspects of importance for our understanding of ClyA function. It was evident that purified ClyA protein has cytolytic activity by itself. Native ClyA protein showed both a cytotoxic and a cell-detaching activity towards murine macrophage-like cells, and, as judged by osmotic protection assays, it formed pores of a size between 2.2 and 3.0–3.5 nm in erythrocytes. The purified ClyA showed a specific activity of ~ 200 HU mg−1 on horse erythrocytes, which interestingly is similar to the specific activity of delta-toxin from S. aureus (~ 200 HU mg−1 on human and horse erythrocytes) reported previously by Alouf et al. (1989). However, the activities of these cytolysins are ~ 50 times lower than the haemolytic activity of HlyA, as reported previously by Eberspächer et al. (1989).
Studies with A. laidlawii cells and with liposomes suggest that the presence of cholesterol in membranes facilitates the cytolytic activity of ClyA, and this pore-forming cytolysin does not require Ca2+ ions for its activity against target cell membranes. This is in contrast to the HlyA protein that, when overexpressed from the plasmid pANN202–203, showed a strong CF release when incubated with liposomes, which was dependent upon the presence of Ca2+ (J. Oscarsson, unpublished). Results from structure and function analyses by directed mutagenesis of the putative membrane-associated domains in ClyA supported the hypothesis that certain domains play a role in pore formation.
Although the ClyA protein caused lysis of erythrocytes and other mammalian cells, it caused little or no lysis of the bacterial cells from which it is produced. The finding that cholesterol stimulated the lytic activity may explain this apparent tropism for mammalian cells. The molecular mechanism by which the presence of cholesterol in membranes may increase the cytolytic action of toxins is not entirely clear, but in the present case there may be an increased binding to the membrane-targeting portion of ClyA when cholesterol is present. Indications that there could be such an interaction came from the observation that pretreatment of the ClyA preparation with cholesterol blocked a large part of the cytolytic activity (Fig. 5A). In a similar experiment using the HlyA-producing strain MC4100/pANN202–312, a similar effect, i.e. reduction of haemolytic activity, was noted (J. Oscarsson, unpublished), which is in accordance with previous findings (Bakas et al., 1996).
That there may be direct binding of ClyA to cholesterol was evident from the coprecipitation occurring in aqueous solutions (Fig. 5B). Cholesterol interaction and inhibition of lytic activity is a well-established feature common among the thiol-activated toxins from several Gram-positive bacteria (Alouf and Geoffroy, 1991; Boulnois et al., 1991). Although the interaction may contribute to membrane binding per se, the cholesterol may also affect subsequent steps in the formation of pores by such toxins. For example, in the case of listeriolysin O, recent evidence suggests that inhibition of lysis by addition of cholesterol is not due to decreased binding but rather to inhibition of some step important for polymerization of the toxin during pore formation (Jacobs et al., 1998). Preincubation of listeriolysin O with cholesterol does not influence binding of the toxin complex to red blood cells or to artificial membranes (Jacobs et al., 1998). It is known that the binding of several different surface-seeking proteins is enhanced by an increase in the surface hydrophobicity of the membrane, given by, for example, cholesterol and similar substances. Hence, the effect of cholesterol and tristearate on ClyA shown here may be a general, i.e. not specific, binding to a hydrophobic surface. Although our present findings with ClyA were consistent with the suggestion that cholesterol binding might prevent subsequent binding to red blood cells, it remains a possibility that some additional later step in pore formation could also be affected.
Sequence analyses and fold predictions for ClyA suggest a protein rich in α-helices with several/many amphiphilic peptide segments, some of which are similar to established membrane-binding and pore-forming parts in other proteins, including some lytic bacterial toxins. The segment with the strongest average hydrophobic moment near the ClyA C-terminus (Fig. 6; from residue 266 to residue 291) showed the greatest resemblance to the delta-toxin of S. aureus. However, the sequence is also similar to an amphiphilic region of the HlyA haemolysin at the edge of the proposed HlyA pore structure (Ludwig et al., 1991). These similarities are consistent with the present experimental findings of the membrane-active and pore-forming abilities of the purified ClyA toxin protein. Both the C-terminal domain and the postulated transmembrane helices in the middle of the protein appeared to be essential for activity. Several substitution mutations were tolerated in the C-terminus, but in some cases the mutations affected the surface exposure of ClyA by intact bacteria. The protein does not have a regular N-terminal signal sequence, and how the ClyA protein is secreted/surface located in the E. coli cells is not clear at this time. It seems feasible that the C-terminal region may be directly involved in such translocation, in addition to participating in membrane targeting and pore formation. We therefore suggest that this domain may serve dual functions.
Bacterial strains and growth conditions
The relevant genotypes of the bacterial strains used in this study are listed in Table 3. For overexpression of the ClyA protein, E. coli strains were transformed with plasmids encoding the clyA structural gene (e.g. pYMZ62 and pYMZ80). E. coli strains were grown aerobically, with vigorous shaking, at 37°C in Luria–Bertani (LB) broth or on LB broth solidified with 1.5% (w/v) agar. Blood agar plates consisted of 5% (v/v) horse erythrocytes solidified with 1% (w/v) Columbia agar base (Oxoid). Antibiotic selection was made using 50 μg ml−1 carbenicillin. The mollicutes (mycoplasma) Acholeplasma laidlawii, strain A-EF22, were grown in a lipid-depleted BSA/tryptose medium supplemented with (i) 75 μM myristic acid (14:0) plus 75 μM myristoleic acid (14:1c), (ii) 150 μM oleic acid (18:1c), or (iii) 150 μM 18:1c plus 20 or 50 μM cholesterol (all added from sterile ethanol stocks). The supplied fatty acids are covalently incorporated as acyl chains into the membrane lipids during growth, whereas cholesterol is passively taken up by the lipid bilayer of the cell depending upon growth medium concentrations. The size and morphology of A. laidlawii were observed using a Zeiss Standard WL phase-contrast microscope (magnification 1000×, numerical aperture 1.30), equipped with a movable eyepiece micrometer ruler (calibrated; Zeiss). Measurements of A. laidlawii cell diameters using this technique have been described previously (Wieslander et al., 1995).
ClyA was purified from YMZ4/pYMZ80 at 4°C as follows: a 250 ml overnight culture was suspended in 10 ml of A50 buffer [50 mM tris-HCl, 1 mM EDTA, 50 mM (NH4)2SO4, 10% glycerol, 1 mM dithiothreitol (DTT), 0.1 mM phenylmethylsulphonyl fluoride, pH 8.0]. Three hundred microlitres of sonic extract was applied on a cylindrical 6% PAGE (60 mm length and 28 mm diameter) cast in a Model 491 Prep Cell (Bio-Rad), using the continuous buffer system of McLellan (1982) (50 mM Tris, 25 mM boric acid) at 300 V constant voltage. A flow of 80 μl min−1 was provided by the elution pump and, after 90 min, 125 fractions, each of 160 μl, were collected and further analysed on SDS–PAGE as described previously (Laemmli, 1970), and these fractions were then assayed for cytolytic activity. Silver staining and protein concentration determination were carried out using the Silver Stain Plus Kit and BCA microprotein assay (Bio-Rad) following the protocols of the manufacturer. Determination of the N-terminal sequence of ClyA protein preparations was achieved by automated Edman degradations. Samples were blotted onto filters essentially as described previously by Matsudaira (1987).
Western blotting analysis
Rabbit anti-ClyA antibodies were raised against the ClyA protein overexpressed from the plasmid pYMZ80. The 34 kDa protein band was excised after gel electrophoresis and was then used for rabbit immunization. An antiserum taken 6 weeks after the fourth injection was affinity purified according to the method of Taraseviciene et al. (1994) before being used in Western blotting analysis at a final dilution of 1:1000. Western blotting was performed as described earlier by Johnston and Zabriskie (1986), using alkaline phosphatase-conjugated secondary antibody at a final dilution of 1:1000.
For determination of the amount of ClyA protein in cellular extracts, aliquots were run on SDS–PAGE gels as previously described (Laemmli, 1970), together with known amounts of standard markers, and were analysed after staining with Coomassie blue. Approximately 1.5 μg of each ClyA protein was loaded onto SDS–PAGE gels for subsequent Western blotting analysis (see Figs 7 and 8).
Assays for cytolytic activity
Phenotypically, the bacteria expressing haemolytic activity showed a clearance zone on blood agar plates. For quantitative assessments, contact haemolytic assays were performed, essentially as described previously (Sansonetti et al., 1986) with the following modifications: when used, bacteria were grown in LB rather than tryptic-soy broth, and horse erythrocytes instead of sheep erythrocytes were used. Values represent the means of at least three separate measurements. One haemolytic unit of ClyA protein was defined as the concentration evoking 50% haemolysis of an erythrocyte suspension (final density 5 × 108 cells ml−1) within 120 min at 37°C, essentially as described previously by van Leengoed and Dickerson (1992). For our measurements, we used horse erythrocytes. For comparing the lytic activity of different ClyA mutant peptides, a relative lytic activity of 1.0 corresponded to assays using sonic extracts of the strain MC1061/pYMZ80 in PBS buffer at a final concentration of ≈10 μg ml−1 of wild-type ClyA protein in the reaction wells for 120 min at 37°C (see Tables 1 and 2). For osmotic protection, we used 30 mM sugar solutions in PBS. The molecular diameters of saccharides used were taken from published data (Scherrer and Gerhardt, 1971; Schönherr et al., 1994): glucose, 0.7; maltose, 1.0; dextrin 20, 1.6; and dextrin 15, 2.2. In accordance with Bhakdi et al. (1986), the molecular diameters of raffinose and dextran 4 were assumed to have the mean molecular diameters of 1.2–1.4 nm and 3.0–3.5 nm, respectively, on the basis of published data (Scherrer and Gerhardt, 1971).
The determination of the effects of cholesterol or 1,2-dioleyl-phosphatidylcholine (DOPC) on haemolytic activity was essentially as described earlier (Jarvill-Taylor and Minion, 1995): a 1.2% (w/w) cholesterol solution in 99.5% ethanol or a 2.4% (w/w) DOPC solution in 99.5% ethanol was added directly to the assay wells (0.2, 0.5, 1.0, 2.0, 5.0 and 10 μl). Controls consisted of erythrocytes incubated with PBS, ethanol, cholesterol or DOPC respectively.
Binding of ClyA from culture supernatants was carried out essentially as described earlier (Vazquez-Boland et al., 1989): 1 ml aliquots of sonic extracts of MC1061/pYMZ80 in PBS buffer, containing ≈5 μg ml−1 ClyA protein, were filtered under sterile conditions and were incubated with 1.0, 5.0 and 20 μl of a 10 mg ml−1 solution of cholesterol, or a 23 mg ml−1 solution of tristearate, for 45 min at 37°C with shaking. After centrifugation, the haemolytic activity of the supernatants was analysed and the cholesterol or tristearate pellets were washed in PBS and subjected to SDS–PAGE and Western immunoblotting using an antiserum raised against ClyA.
Cytotoxicity assays with macrophage cells
The murine macrophage cell line J774 was routinely grown in Ham's F-10 medium supplemented with 10% fetal calf serum, 1 mM glutamine and 100 U ml−1 of penicillin; culturing was performed at 37°C in a humidified atmosphere with 5% CO2. One day before infection with bacteria, about 2 × 104 J774 cells were seeded onto flat-bottomed 96-well plates (Falcon, Becton-Dickinson Labware). Cells were washed twice with PBS the next day and were covered with cell culture medium lacking penicillin. In PMA-related experiments, cells were pretreated with PMA at a final concentration of 150 ng ml−1 for 15 min at 37°C and remained so until the end of the experiment. J774 cells were infected with bacteria at a multiplicity of infection (MOI) of 100, or were treated with purified and filtered ClyA in 200 μl total volume of cell medium. At 2 h post infection (PI), cells were washed and incubated with 100 μg ml−1 gentamicin. For PMA-treated cells, there were no further washes and no addition of antibiotics. Supernatants of the infected or ClyA-treated macrophages from designated time points were sampled and assayed for the activity of released intracellular enzyme lactate dehydrogenase (LDH) (Korzeniewski and Callewaert, 1983), using the cytotox 96 kit (Promega) according to the instructions from the manufacturer. The percentage of cytotoxicity was calculated as 100 × (experimental release — spontaneous release)/(total release — spontaneous release), in which spontaneous release was the amount of LDH activity in supernatants of cells incubated in medium alone and total release was the activity in macrophage lysates after treating with Triton X-100 provided with the kit. The strains that we used do not have endogenous LDH activity when grown aerobically.
The detachment assay was basically as described previously (Gunzburg et al., 1993; Vanmaele et al., 1995; Elliott et al., 1998). Briefly, about 1.0 × 105 cells were added to each well of the 24-well tissue culture plates, and ClyA was added in 300 μl total volume of cell medium. Twenty-four hours after treatment, the monolayers were washed three times with PBS to remove non-adherent cells. J774 cells remaining in each well were fixed for 10 min with 70% methanol and were then stained with Giemsa stain for 30 min. The monolayers were then washed three times with water to remove excess Giemsa stain, and the stained cells were lysed with 2% sodium dodecyl sulphate (SDS). A portion of the lysates was transferred to 96-well microtitre plates, and the absorbance of the contents from each well was recorded using a multiscan plate reader at 620 nm. The percentage of monolayer detaching was calculated as follows: 100 − [(A620 of inoculated well/A620 of uninoculated well) × 100].
Lipid extraction and liposome preparation
Membrane lipids were extracted from A. laidlawii and human erythrocytes using a modified Bligh and Dyer method (Kates, 1972). Synthetic lipids were purchased from Avanti. Large, hand-shaken multilamellar liposomes were prepared basically as described by New (1990). Two micromoles of lipids in chloroform/methanol (2:1 v/v) were evaporated under reduced pressure to form a lipid film on the round-bottomed flask wall using a rotavapor and then under vacuum overnight. The lipids were hydrated at room temperature in 0.2 ml buffer containing 50 mM tris-HCl, 20 mM NaCl, 0.1 M carboxyfluorescein (CF), pH 8.7, for 2 h. At this CF concentration, there is self-quenching of the fluorescence emission. To remove the non-encapsulated marker, the liposome suspensions were applied on small Sephadex G-25 columns (Pharmacia) and eluted with buffer containing 50 mM tris-HCl, 140 mM NaCl, pH 8.7. Phospholipid concentrations of the liposomes were determined according to a modified Bartlett method (Kates, 1972).
Assay of fluorescent marker release from liposomes
Release of the marker was monitored by measuring the increase in fluorescence intensity upon release, and hence dilution of CF. An amount of liposomes corresponding to ≈0.2 nmol ml−1 phospholipids in Tris-buffered saline in a cuvette was incubated at RT for 5 min, then native ClyA in working buffer (50 mM Tris, 25 mM boric acid) was added to a concentration of 1.2 or 2.4 μg ml−1. The intensity of fluorescence at 515 nm was measured for 10 min at RT in samples excited at 490 nm. Excitation and emission beam slits were 5 nm. The complete (100%) release of CF was defined as the intensity of fluorescence that was gained upon complete solubilization of the liposomes using 1% (w/v) Triton X-100.
Computer-aided sequence analyses
Homologues to the full-length ClyA amino acid sequence were searched using the blast and FASTA algorithms (Pearson and Lipman, 1988; Altschul et al., 1990) at the (www) molecular biology servers at ExPASy, University of Geneva, and at EMBL, Heidelberg respectively. Fold assignment analyses were carried out with the methods PhD and PSCAN and the method described by Fischer and Eisenberg (1996) through the servers at EMBL, Heidelberg, the Department of Biochemistry, Stockholm University and UCLA-DOE respectively. At the second site, potential hydrophobic, transmembrane segments were also searched using TopPred 2 (von Heijne, 1992). The hydrophobic moment and potential transmembrane segments (Eisenberg et al., 1984) were analysed using the GCG program moment. A possible domain organization of ClyA was investigated by dividing the amino acid sequence into seven overlapping segments and searching the swissprot database (release 34) for each of these using the FASTA algorithm within the GCG computer analysis package. Likewise, a comparison of these seven with a sublibrary of toxin sequences was carried out using the GCG program pileup.
Mutants with alterations in the ClyA amino acid sequence were obtained by using the QuikChange Site-Directed Mutagenesis Kit of Stratagene, following the instructions of the manufacturer, and plasmid pYMZ80 was used as the template. All polynucleotides used were synthesized by DNA Technology, and the desired mutations were always placed in the middle of the primer with 15 bases of correct sequence on each side. All recombinant plasmids were introduced by transformation into strain MC1061, and mutations were verified by DNA sequencing using the ABI PRISM Dye Terminator Cycle Sequencing Ready Reaction Kit with AmpliTaq DNA Polymerase, and an ABI PRISM 377 DNA Sequencer.
†Present address: Department of Bacteriology, Faculty of Medicine, Kyushu University, Fukuoka 812–82, Japan
We thank Elisabet Pålsson for skilful technical assistance. This work was supported by grants from the Swedish Natural Science Research Council, the Swedish Medical Research Council, and the Göran Gustafsson Foundation for Research in Natural Sciences and Medicine.