Assembly of the Bi-component Leukocidin Pore Examined by Truncation Mutagenesis*

Staphylococcal leukocidin (Luk) and α-hemolysin (αHL) are members of the same family of β barrel pore-forming toxins (βPFTs). Although the αHL pore is a homoheptamer, the Luk pore is formed by the co-assembly of four copies each of the two distantly related polypeptides, LukF and LukS, to form an octamer. Here, we examine N- and C-terminal truncation mutants of LukF and LukS. LukF subunits missing up to nineteen N-terminal amino acids are capable of producing stable, functional hetero-oligomers with WT LukS. LukS subunits missing up to fourteen N-terminal amino acids perform similarly in combination with WT LukF. Further, the simultaneous truncation of both LukF and LukS is tolerated. Both Luk subunits are vulnerable to short deletions at the C terminus. Interestingly, the N terminus of the LukS polypeptide becomes resistant to proteolytic digestion in the fully assembled Luk pore while the N terminus of LukF remains in an exposed conformation. The results from this work and related experiments on αHL suggest that, although the N termini of βPFTs may undergo reorganization during assembly, they are dispensable for the formation of functional pores.

␤ barrel pore-forming toxins (␤PFTs) 3 are secreted by bacteria as water-soluble polypeptides that bind to the surfaces of susceptible cells and assemble into oligomeric transmembrane pores. One family of ␤PFTs includes staphylococcal ␣-hemolysin (␣HL) and the leukocidins (Luk) (1)(2)(3)(4). The pathophysiological effects of these proteins have been attributed to pore formation on target cells, leading to cell permeation or lysis. Leukocidin, in particular, has received widespread attention as an important virulence factor in wound and soft tissue infections (5,6).
␣HL, a ␤PFT comprising a single polypeptide of 293 residues, forms a homo-heptameric pore (7). In contrast, leukocidin is a bi-component toxin (3,4); two distinct proteins are required to form a functional pore, one component from class F (LukF) and the other from class S (LukS). The leukocidins primarily attack polymorphonuclear cells, monocytes, and macrophages but also assemble on erythrocytes (8,9), liposomes (10), and planar lipid bilayers (9). There are at least six class-F proteins (LukF-PV, LukF-R, LukD, LukFЈ-PV, LukF, and LukF-I) and seven class-S proteins (LukS-PV, LukS-R, LukE, LukM, ␥HLII, LukS, and LukS-I) associated with various strains of Staphylococcus aureus (2,11,12). The F and S proteins share a common ancestor (11). Proteins within each class (F or S) share ϳ70% identity at amino acid level, whereas the identity drops to Ͻ27% between members of the two different classes (1,2). No member of either class is Ͼ30% identical to ␣HL (2,13). It is believed that the various F and S components are capable of mixing and matching, thereby creating a diverse repertoire of pores (2,14). Although hexameric and heptameric structures have been proposed for the Luk oligomer (10,(15)(16)(17)(18), recent work suggests that Luk pores are octamers with four LukF and four LukS subunits arranged in alternating fashion around the central axis (19,20).
Although the structure of a Luk oligomer is yet to be solved, the three-dimensional structures of the water-soluble monomeric forms of two class F components, LukF (hlgB gene product) and LukF-PV, and one class S component, LukS-PV, have been determined by x-ray crystallography. The LukF structure has been solved at 1.8-Å and 2.3-Å resolution ( Fig. 1) (16). Excluding the amino latch and putative pre-stem regions, the protein displays a fold that is closely similar to the fold of an individual ␣HL protomer within the structure of the heptameric ␣HL pore ( Fig. 1) (7). The LukF-PV structure has been solved at 2.0-Å resolution and is almost identical to LukF (17). The LukS-PV structure has been determined recently at 2.0-Å resolution and has a similar fold to that of LukF, although the rim domain is in a significantly different conformation ( Fig. 1) (12). Together, the structures of the leukocidin monomers and the ␣HL heptamer represent beginning and end points in ␤PFT assembly. Additional evidence suggests that these proteins bind to membranes as monomers, associate on the membrane surface to form oligomeric prepores, and finally insert into the lipid bilayer (7,16,(21)(22)(23)) (see Fig. 1B in the preceding paper (36)).
It has been proposed that the N termini of these ␤PFTs are crucial for the formation of functional pores (21). The N termini are believed to reorganize during the insertion step to form a "latch" that reinforces the interactions between neighboring subunits (7,16). However, recent work from our laboratory suggests that, whereas this can occur, it is not required for pore formation in the case of ␣HL (Jayasinghe et al., preceding report (36)). Here, we use truncation mutagenesis to examine the contributions of the N and C termini of both LukF and LukS to the assembly of the leukocidin pore. We find the following: 1) Neither N terminus is required for pore formation; therefore, latch formation is not necessary for assembly of the pore. 2) The N termini of LukF and LukS reside in different conformation in the WT Luk pore: one is in the latch conformation and the other is not.
3) The C termini of LukF and LukS must be intact for efficient oligomerization.

* This work was supported by the Medical Research Council/Engineering and Physical
Sciences Research Council and the Office of Naval Research. Work at Texas A&M was supported by Defense Advanced Research Project Agency, the DoD Tri-Service Technology Program, the U.S. Department of Energy, NASA, the National Institutes of Health, and the ONR. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1

EXPERIMENTAL PROCEDURES
Deletion Mutagenesis-All constructs were made in the pT7-SC1 expression vector (9,24) and verified by DNA sequencing of the entire genes. Genes encoding a series of N-and C-terminal truncation mutants of LukF (hlgB gene product) ( Table 1) and LukS (hlgC gene product) ( Table 2) (9) were made by PCR mutagenesis and ligation-free in vivo recombination as described elsewhere (25,26). Each forward mutagenic primer for N-terminal truncation contained an NdeI site at the initiation codon followed by an alanine codon, except for LukF-N⌬1 (see Table 1) and LukS-N⌬1 (see Table 2).
In Vitro Transcription and Translation-All proteins were generated by coupled in vitro transcription and translation (IVTT) by using an Escherichia coli T7-S30 extract optimized for circular DNA (Promega, no. L1130). The complete 1 mM amino acid mixture minus methionine (2.5 l) or the complete 1 mM amino acid mixture minus cysteine (2.5 l) that was supplied in the kit was mixed with premix solution (10 l), [L-35 S]methionine (1 l, ICN Biomedicals, Inc., Irvine, CA, 1175 Ci/mmol) or [ 35 S]cysteine (1 l, Amersham Biosciences, 1200 Ci/mmol), plasmid DNA (4 l, 400 ng/l), and T7-S30 extract (7.5 l) supplemented with rifampicin (20 g/ml final) to generate "hot" radiolabeled polypeptides (25 l) (24). Synthesis was carried out for 60 min at 30°C, and reactions were terminated with chloramphenicol (100 M final). The complete amino acid mixture minus cysteine and the complete amino acid mixture minus methionine were mixed in equal volumes and replaced the individual amino acid mixture minus methionine (or cysteine) (2.5 l) to yield "warm" radiolabeled proteins. Unlabeled (cold) proteins were generated by replacing the [L- 35

Nomenclature and sequences of LukF N-and C-terminal truncation mutants
If the polypeptides are deformylated, the initial Ala (in all the N-terminal truncation mutants except LukF-N⌬1) would ensure the efficient co-translational removal of the N-terminal methionine (21). However, we do not know whether our IVTT proteins contain fMet or Met at the N terminus.  (12). Two striking differences between the Luk and ␣HL structures are in the N-terminal and stem domains. The N termini of the soluble forms of LukF and LukS-PV are ␤ strands, which are hydrogen-bonded to the cap domain. In the ␣HL pore, the N terminus extends away from the cap and forms extensive contacts with the inner ␤ sheet of an adjacent (n ϩ 1) and subsequent (n ϩ 2) subunit. The pre-stems of LukF (residues 110 -144) and LukS-PV (residues 104 -139) form antiparallel ␤ sheets, which are folded against the ␤ sandwich (cap) domain. In contrast, the ␣HL stem is composed of two extended ␤ strands, which are part of a 14-stranded ␤ barrel.
Quantitative Hemolysis Assay-LukF, LukS, and the truncated polypeptides (IVTT, 5 l of each LukF and LukS component) were diluted into MBSA (90 l) in the first well of each column or row of a 96-well microtiter plate. The proteins were then subjected to eight or twelve 2-fold serial dilutions in the same buffer down the columns or across the rows (final volume 50 l per well). An equal volume of 1% washed rabbit erythrocytes (rRBC) in MBSA was added to each well. Hemolysis was recorded for 2 h at 25°C by monitoring the decrease in light scattering at 595 nm with a Bio-Rad microplate reader (Model 3550-UV) and the Microplate Manager 4.0 software.
Trypsin Treatment of Polypeptides on Membranes-Trypsin solutions (5.0, 0.5, and 0.05 mg/ml in water, Sigma, T-7309) were prepared by dilution of an enzyme stock (10 mg/ml in water) and used immediately. Wild-type LukF, wild-type LukS, truncated variants, and ␣HL were allowed bind to rRBC membranes before limited proteolysis was performed. The membranes were resuspended in MBSA (0.19 mg of membrane protein per ml) and divided into four tubes (18 l in each). Trypsin or water (2 l) was added to each tube. After 5 min at room temperature, the reactions were stopped by treatment with phenylmethylsulfonyl fluoride (9 mM final, added in 2 l of isopropanol) for 5 min at room temperature, followed by the addition of 2ϫ Laemmli loading buffer. The samples were subjected to electrophoresis in 10% SDS-polyacrylamide gels.

Hemolytic Activity and Oligomerization Properties of LukF N-terminal Truncation
Mutants-Genes encoding truncation mutants of LukF in which the first 22 residues were removed one-by-one (Table 1) were generated by ligation-free in vivo recombination. The corresponding [ 35 S]Met-labeled "warm" proteins were prepared by coupled IVTT by using an S30 extract from E. coli (29). The truncated LukF polypeptides were tested for the ability to form SDS-stable oligomers on rRBC membranes with unlabeled (cold) wild-type LukS (9). The lytic activity of leukocidin toward rRBC is normally at least 500-fold lower than that of ␣HL (2). Accordingly, the extent of oligomerization of leukocidin is also significantly lower. Nevertheless, the N-terminal truncation mutants of LukF lacking up to 19 amino acids were capable of producing SDSstable hetero-oligomers ( Fig. 2A). Although, the extent of oligomerization of LukF-N⌬2 and LukF-N⌬3 is lower than that of the WT protein, mutants with more extensive deletions (LukF-N⌬5 to LukF-N⌬19) produced enhanced bands of leukocidin oligomer ( Fig. 2A). Only a very weak oligomer band was detected (after prolonged exposure of film to the gels) after deletion of 20 or 21 residues (data not shown). We were unable to detect any SDS-stable oligomers on rRBC membranes after the removal of 22 amino acids or more (N⌬27, N⌬30, N⌬32, and N⌬43), even after prolonged exposure. All constructs (up to N⌬43), however, retained the ability to bind to the same extent to rRBC membranes as monomers (data not shown).
The activities of wild-type LukF and the LukF deletion mutants were also compared in a quantitative hemolytic assay. Mutant LukF polypeptides with N-terminal truncations of up to 19 residues exhibited hemolytic activity comparable to WT LukF when mixed with WT LukS (Fig. 2B). Mutants with an additional one or two deleted residues, N⌬20 and N⌬21, showed a marked reduction in hemolytic activity (at least 30-fold less than the WT protein). LukF truncation mutants missing 22 amino acids or more (N⌬27, N⌬30, N⌬32, and N⌬43) did not lyse rRBC with WT LukS, even after 24-h incubation (data not shown).
Hemolytic Activity and Oligomerization of LukS N-terminal Truncation Mutants-In a similar fashion, a series of N-terminal truncation mutants of LukS labeled with [ 35 S]Met (warm, see "Experimental Procedures") were produced with deletions ranging from 1 to 36 amino acid residues ( Table 2). The truncated LukS polypeptides were tested for their ability to form SDS-stable oligomers on rabbit erythrocyte membranes with unlabeled WT LukF (9). As in the case of the LukF truncations, all of the LukS mutants were capable of binding to rRBC membranes. SDS-stable oligomers were observed with N-terminal deletions of up to 14 residues, with the exception of LukS-N⌬8 (Fig. 3A). However, no oligomers could be seen with deletions of 15 or more amino acids (LukS-N⌬15 to N⌬18).
Hemolytic activity similar to WT LukS was seen with N-terminal truncations of up to 14 residues, with the exception of LukS-N⌬8, in which case the activity was decreased at least by 15-fold (Fig. 3B). LukS-N⌬15 displayed a more pronounced lag in lysis (ϳ3-fold increase in lag time, defined as the time to lysis of 5% of a red blood cell suspension). No hemolysis was recorded with truncations of 16 or more residues, even when the microtiter plates were examined 24 h later (data not shown).
Hemolytic Activity and Oligomerization of Mixed LukS and LukF N-terminal Truncation Mutants-To determine whether leukocidin could tolerate truncations of both components simultaneously, we studied the pore formed by LukF-N⌬16 and LukS-N⌬14. LukF-N⌬16 assembled with LukS-N⌬14 to form stable oligomers, albeit less efficiently than the WT subunits or one truncated subunit with its WT partner (Fig. 4A, lane 2). The hemolytic activity of the pair was at least 8-fold less than that of the WT components and the initial lag time was increased significantly (Fig. 4B).
Conformations of Leukocidin Polypeptides Probed by Limited Proteolysis-Trypsin has previously been shown to cleave ␣HL in solution at Lys-8 near the N terminus and at Lys-131 in the pre-stem domain, and has been used to probe the conformational changes at various stages of pore assembly (30,31). The N terminus of ␣HL remains accessible in the membrane-bound monomer and in the nonlytic prepore state (21,22). Upon insertion of the stem domain to form the transmembrane ␤ barrel, the N terminus becomes occluded within the heptameric pore and is protease resistant (22).
However, the conformations of the N termini of Luk proteins in the assembled pore have not been studied thoroughly. We therefore probed the LukF and LukS polypeptides by proteolysis with trypsin and compared the results to those obtained with ␣HL. The N terminus of LukS was labeled by generating the LukS-D3C mutant by IVTT in the presence of [ 35 S]Cys. The first trypsin site in this mutant is positioned after the cysteine, at position 9 (Fig. 5, A and B). Therefore, the radiolabeled [ 35 S]Cys at position three acts as a marker for an intact N terminus in LukS (24). Sequence analysis and comparison of the crystal structures of the monomers revealed that the position equivalent to LukS-D3 is occupied by Ser-9 in LukF (Fig. 5A). Therefore, we also labeled the N terminus of LukF by generating the mutant LukF-S9C by IVTT, in the presence of [ 35 S]Cys. Although the first trypsin cleavage site of LukF lies before the marker (Lys-4), cleavage sites at positions 11, 12, and 16 are positioned after it (Fig. 5, A and B). Therefore, the radiolabeled [ 35 S]Cys at position 9 acts as a marker for an intact N terminus in LukF. As a control, we also generated [ 35 S]Cys␣HL-S3C and observed the cleavage at Lys-8 (Fig. 5, A and B). WT ␣HL, WT LukF, and WT LukS contain no cysteine residues, and cysteines placed at the designated sites have no adverse effects on activity or oligomerization.

TABLE 2 Nomenclature and sequences of LukS N-and C-terminal truncation mutants
If the polypeptides are deformylated, the initial Ala (in all the N-terminal truncation mutants except LukF-N⌬1) would ensure the efficient co-translational removal of the N-terminal methionine (21). However, we do not know whether our IVTT proteins contain fMet or Met at the N terminus.
With increasing concentrations of trypsin, a gradual loss of radioactivity was observed in the monomeric form of ␣HL bound to red cell membranes, which is consistent with cleavage near the N terminus (lanes 9 -12, Fig. 6). Similarly, both the LukF and LukS membranebound monomers were digested with the increasing concentrations of trypsin (lanes 1-8, Fig. 6). In agreement with previous studies, the radiolabel at position 3 was retained in the membrane-bound ␣HL oligomer, even at a trypsin concentration of 500 g/ml. This result confirms that occlusion of the N terminus inside the pore (i.e. formation of the amino latch) makes the N terminus resistant to proteolysis (7). However, when LukF-S9C* was assembled with WT LukS on rRBC membranes, the radiolabel at position 9 of LukF was completely lost from the oligomer even at the 5 g/ml trypsin concentration (lanes 1-4,  Fig. 6). By contrast, when LukS-S3C* was assembled with WT LukF, the radiolabel at position 3 of LukS was retained in the oligomeric state even at the 500 g/ml trypsin concentration ( lanes 5-8, Fig. 6). In this case, a shift in the band of oligomers to higher mobility does indicate partial digestion of the assembled pore, which might be attributed to cleavage of the N terminus of LukF. We cannot exclude the possibility that the C terminus of one or both subunits is also digested.

Properties of LukF and LukS C-terminal Truncation
Mutants-The C terminus of ␣HL is vital for the stability of the protein assembled on membranes (21). Mutants missing three or five amino acids at the C terminus are inefficient at oligomerization, but do lyse rRBCs, albeit extremely slowly. However, the role of the C terminus of both LukF and LukS has not been investigated in detail. When the sequence and monomeric structures of ␣HL and LukF are compared, LukF is found to have ten additional residues at the C terminus compared with ␣HL (Fig. 5, A  and B). Therefore, five C-terminal deletion mutants were constructed for LukF involving the removal of 1, 4, 6, 7, and 10 residues (Table 1), and their abilities to form SDS-stable oligomers and to lyse rRBCs were evaluated. A loss of up to 4 residues from the C terminus yielded similar amounts of SDS-stable oligomer as WT LukF, when the truncated LukF subunits were assembled with WT LukS (Fig. 7A, lanes 1-3). Although the mutants had close to WT hemolytic activity, only a trace of the Luk oligomer was observed after the removal of 6 or 7 amino acids (Fig. 7A,   5, Fig. 7B). Neither oligomerization nor hemolytic activity was detected after the deletion of 10 amino acids from the C terminus of LukF (LukF-C⌬291) (data not shown). Although the C terminus of LukF-C⌬291 is the same length as that of ␣HL, the deletion most likely resulted in an incorrectly folded polypeptide as the mutant was markedly susceptible to proteolysis (data not shown).
The C terminus of LukS is six residues shorter than that of LukF and four residues longer than that of ␣HL (Fig. 5, A and B). Therefore, three C-terminal deletion mutants were constructed for LukS involving the removal of 4, 6, and 8 residues ( Table 2). The removal of 4 amino acids (LukS-C⌬283) produced an SDS-stable oligomer with WT LukF and the hemolytic activity of the pair resembled that of WT leukocidin (Fig.  7A, lane 7 and Fig. 7B). However, after deletions of 6 or more C-terminal residues, SDS-stable oligomers were not detected on rRBC membranes when the LukS mutants were assembled with WT LukF (Fig. 7A, lanes 8  and 9). Despite its inability to form a stable oligomer, LukS-C⌬281 (a deletion of 6 amino acids) did exhibit hemolytic activity of about 15-fold less than WT leukocidin (Fig. 7B). A pronounced lag in lysis, comparable to LukS-N⌬15, was also observed with this mutant. After the removal of eight residues (LukS-C⌬279), only very weak hemolysis was observed. We cannot exclude the possibility that the observed hemolytic activity of the latter two mutants is the result of undetected amounts of SDS-stable oligomers.
Premature Oligomerization and Co-oligomerization of N-and C-terminal Truncation Mutants-In the previous manuscript, we showed that mutants of ␣HL in part oligomerize in solution in the absence of rRBC membranes when more than three residues are removed from the N terminus. Therefore, we explored the possibility of premature oligomerization with the N-and C-terminal truncation mutants of LukS and LukF and found no detectable spontaneous oligomerization with their respective WT counterparts in the absence of membranes. Like WT LukF and WT LukS, none of the mutants were able, by themselves, to make homo-oligomers on rRBC membranes. In addition, neither the WT Luk polypeptides nor the N-terminal truncation mutants of LukF and LukS were able to co-oligomerize with ␣HL monomers (32).

DISCUSSION
Staphylococcal ␣HL and leukocidin are homologous ␤PFTs with similar structural features. An early analysis of truncation mutants of ␣HL suggested that the N terminus plays a crucial role in the transformation of the prepore to an active pore (21). However, as documented in the preceding report, by a thorough examination of a new series of truncation mutants with up to 22 residues deleted, we demonstrated that the N terminus of ␣HL is in fact not necessary for pore formation (Jayasinghe et al.,preceding paper (36)). During these studies, we found that the N terminus of ␣HL interacts indirectly with position 217 during the transformation of the prepore to a functional pore. Although a Ser 3 Asn mutation at this position does not affect pore formation by the full-length protein, the mutation arrested the assembly of the N-terminal truncation mutants at the prepore stage. This mutation was present in the original series of truncations (21).
The role of the N termini of LukF and LukS during pore formation has not been studied fully. The combination of LukF and LukS, as used in the present report, forms the Luk pore, which has leukocytolytic activity toward human and rabbit polymorphonuclear leukocytes and hemolytic activity toward rabbit erythrocytes (2). The combination of LukF and ␥HLII, a homologous class S component, forms the ␥-hemolysin pore, which has hemolytic activity toward human and rabbit erythrocytes (33,34). In an earlier study, Kaneko and coworkers (34) generated six LukF truncation mutants in which up to 22 residues were deleted from the N terminus. They reported a reduction to Ͻ10% leukocytolytic activity toward human leukocytes when a LukF mutant lacking the first 20 residues was assembled with WT LukS. The same LukF mutant also failed to show any detectable hemolytic activity on human erythrocytes with WT ␥HLII. It was further reported that a mutant lacking the first 22 residues of LukF was unable to bind to human erythrocytes as monomers. However, the study did not include any LukS truncation mutants or examine the hemolytic activity of the combined LukF and LukS subunits on rRBCs.
In the present study, we describe the effects of truncation mutagenesis on the oligomerization and pore-forming activity of the bi-component leukocidin toxin. We demonstrate that large deletions at the N terminus can be made on either component, LukF or LukS, without significantly altering the extent of formation or activity of the oligomer, which in the case of leukocidin is an octamer with alternating F and S subunits (19,20). The results are in general agreement with the properties of ␣HL truncation mutants, in which up to 17 N-terminal residues could be removed with retention of hemolytic activity (Jayasinghe et al., prededing paper (36)), although in the case of leukocidin the activity is weaker to begin with and does not drop off so rapidly with truncation. A deletion of 17 residues removes the residues that constitute the N-terminal latch of ␣HL (Fig. 5, A and B). When mapped onto the structures of monomeric LukF and LukS-PV (LukS-PV has an N terminus of the same length as LukS), deletions of 16 and 10 amino acid residues, respectively, constitute the removal of the N-terminal domains. When assembled with their wild-type counterparts, truncation mutants of up to 19 amino acid residues of LukF and 14 of LukS retained near wildtype lytic activity. Importantly, the leukocidin also tolerated the simultaneous removal of the entire N-terminal domains from both LukF and LukS; although the activity was lowered by at least 8-fold, a truncation mutant missing 16 residues from the N terminus of LukF was able to form a functional pore with a LukS mutant missing 14 residues from its N terminus. Therefore, our results indicate that, as in the case of ␣HL, the N termini of the Luk proteins are not required for pore formation. Functional pores could be formed after partial or complete loss of the N terminus in one or both of the components. Additionally, the finding that these ␤PFTs can tolerate large extensions at their N termini (20), without disruption of oligomerization or lytic activity, further strengthens the argument that the N terminus is not required to form an amino latch with the neighboring subunit in the oligomer.
Removal of residues beyond the first 19 of LukF or the first 14 of LukS resulted in significantly diminished extent of oligomer formation and activity. In both proteins, these deletions extend into the ␤1 strand comprising the inner face of the ␤-sandwich domain (Fig. 1) and are likely to cause misfolding. We examined this possibility by limited proteolysis and found that LukF mutants with more than 22 N-terminal residues removed and LukS mutants with more than 18 N-terminal residues removed were more susceptible to protease digestion (data not shown). Based on subunit-subunit interactions apparent in the structure of the ␣HL heptamer, the ␤1-␤2-␤3 strands of both LukF and LukS  (12). The N terminus of each polypeptide is colored red, the C terminus is blue. Radiolabeled sites used in the limited proteolysis experiments are colored green. Underlined residues indicate potential trypsin cleavage sites near the N termini. B, structures of ␤PFTs with corresponding regions colored as in A. The LukS cleavage sites were mapped onto the LukS-PV structure (12). Arrows indicate potential trypsin cleavage sites.
form critical interfaces in the assembled pore, which when disrupted would prevent toxin assembly.
Limited proteolysis has been used to detect changes in the conformation of ␣HL. Trypsin, which digests polypeptides at the C-terminal side of Lys and Arg, cleaves the N terminus of ␣HL at Lys-8 in the water-soluble monomer. The cleavage site remains solvent-exposed and susceptible to trypsin until latch formation, which in the WT protein is coincident with membrane insertion of the stem domain (Fig. 8A, stage  4). However, in the preceding report (36), we demonstrated that in various mutants of ␣HL the formation of a protease-resistant latch from Freshly translated LukF-S9C, LukS-D3C, and ␣HL-S3C labeled with 35 S[cysteine] were allowed to bind to rRBC membranes. In the cases of the leukocidins, the unlabeled WT counterpart was included to ensure pore formation. The washed membranes were then treated with trypsin as follows (final concentrations): lanes 1, 5, and 9: 0 g/ml; lanes 2, 6, and 10: 5 g/ml; lanes 3, 7, and 11: 50 g/ml; lanes 4, 8, and 12: 500 g/ml. An autoradiogram of a 10% SDS-polyacrylamide gel is shown. the N terminus is not required to make lytic pores. In the present study, by site-specific radiolabeling and limited proteolysis, we have shown that the N termini of LukF and LukS reside in different conformations within the assembled Luk oligomer: the N terminus of LukS is proteaseresistant (Figs. 6 and 8B), whereas the N terminus of LukF is proteasesensitive (Fig. 6). One possibility is that the N terminus of LukS forms a latch while the N terminus of LukF remains exposed to solvent (Fig. 8B). Alternatively, the N terminus of LukF could remain folded onto the ␤-sandwich domain in a protease-sensitive conformation.
In the previous report (36), we showed that the N terminus of ␣HL helps to prevent premature oligomerization of the monomer in solution. A gradual increase in the extent of premature oligomerization was observed when residues were sequentially removed from the N terminus. We therefore explored the possibility of premature oligomerization of leukocidin. We generated truncated LukF and LukS mutants by IVTT and analyzed their ability to oligomerize in the absence of membranes with WT LukS and WT LukF, respectively, or by themselves. Unlike ␣HL, we did not observe premature oligomerization with either LukF or LukS (data not shown). However, even with the WT proteins, the efficiency of oligomerization of leukocidin on membranes is much lower than that of ␣HL. Therefore, a low relative extent of premature oligomerization would be difficult to detect in the Luk proteins.
Compared with ␣HL, both Luk F and Luk S have extended C termini. The 10 additional residues of LukF and the 4 of LukS form ␤ strands that are hydrogen-bonded to the backbones of the ␤-sandwich domains. The functional significance of these extensions remains unclear, but they are likely to contribute to the stability of the ␤-sandwich domains. LukS-C⌬283, with 4 residues removed, a structural equivalent of wildtype ␣HL, produced SDS-stable lytic pores, but the removal of an additional 2 residues abolished the SDS-stability of the pore while leaving appreciable hemolytic activity with an increased lag phase. In the case of LukF, up to 7 residues of the extended C terminus could be removed (LukF-C⌬294) without affecting activity. Traces of oligomer were observed, which were no longer detectable after the deletion of 10 C-terminal amino acids (LukF-C⌬290), when activity was completely lost. These results are in keeping with previous studies on ␣HL, which reported that the removal of 3 or 5 amino acids at the C terminus greatly inhibited both oligomer formation and rRBC lysis (21). A 4-residue C-terminal deletion mutant of ␣HL (␣HL:1-289) was extremely susceptible to inclusion body formation and had compromised tertiary structure as deduced from the fluorescence emission of bound 1-anilino-8-naphthalene sulfonate (35). On the other hand, constructs of ␣HL, LukF, and LukS with very large C-terminal extensions (Ͼ94 amino acids) form SDS-stable oligomers and have unaltered hemolytic activity (19).
In summary, we have examined the assembly of leukocidin by truncation mutagenesis and limited proteolysis. Our data show: 1) In the WT proteins, the conformations of the LukF and LukS N termini differ FIGURE 8. Comparison of the assembly models for the staphylococcal ␤PFTs. A, mechanism of assembly for ␣HL. The N terminus is shown as a dashed red line and the (pre)stem as a solid green line. Four principle stages have been defined (16,(21)(22)(23) as follows: 1) ␣HL exists as a monomer in solution. Both the N terminus and the pre-stem are susceptible to proteolysis. 2) Upon membrane binding, the pre-stem becomes resistant to proteolysis (perhaps because the enzyme cannot approach the membrane surface), but the N terminus remains susceptible to cleavage. 3) In the prepore intermediate, the pre-stem domains fill the cavity within the oligomer, where they remain resistant to proteolysis. The N terminus remains accessible, and either it is released as shown or residues such as Lys-8 are still accessible while the strand remains hydrogen-bonded to the ␤-sandwich domain. 4) In the fully assembled pore, the N terminus is buried and forms a latch within the cap domain (7), where it is protease-resistant. B, a model for the assembly of leukocidin. In LukS, the N terminus is represented by a dashed red line, and the stem is in solid black. In LukF, the N terminus is a solid red line and the stem is solid green. 1) Both LukF and LukS subunits exist as monomers in solution. 2) Little is known about the membrane-bound monomers in the case of leukocidin, although they are sensitive to proteolysis (Fig. 6). 3) In the pre-pore intermediate, the pre-stem domains are presumed to fill the cavity within the oligomer, as in the case of ␣HL, although this has not been directly demonstrated for leukocidin. 4) In the functional pore, the N terminus of LukS becomes protease-resistant and most likely forms a latch within the cap domain. By contrast, the LukF N terminus is protease-sensitive, and either it remains free as shown or cleavage occurs while the strand is hydrogen-bonded to the ␤-sandwich domain. Although the N termini of LukF and LukS reside in different conformations in the assembled leukocidin oligomer, neither is required for pore formation, as shown in this report.
in the assembled state (trypsin proteolysis experiment). The N terminus of LukF is in a protease sensitive conformation, whereas the N terminus of LukS is protease-resistant. 2) In agreement with work on the ␣HL pore (see previous report (36)), the N termini of the LukF and LukS subunits are not required for pore formation. This was demonstrated in experiments where both LukF and LukS were truncated, as well as in experiments where a truncated subunit was assembled with its fulllength counterpart.
3) The N termini of LukF and LukS are not required to prevent the premature assembly of the Luk pore.
The structures of the leukocidin monomers and the ␣HL heptamer represent the starting and end points for the assembly of ␤PFTs. Our new results emphasize the similarities and demonstrate subtle differences between the assembly pathways for the two pores. In the future, several key issues concerning the assembly mechanism remain to be addressed (23). The pathway to the prepore is one unresolved issue. The sequential addition of monomers to the growing pore, random encounters between small oligomers, or a mixed pathway could produce the prepore. Membrane insertion to form the ␤ barrel is another issue that could be a stepwise or concerted process with respect to the individual subunits. A crystal structure of a prepore could provide valuable information about the conformation of the pre-stem domain inside the central cavity, and the observation of pore formation by real-time single molecule fluorescence with labeled proteins could reveal information about transient intermediates (23).