Role of the Amino Latch of Staphylococcal α-Hemolysin in Pore Formation

Staphylococcal α-hemolysin (αHL) is a β barrel pore-forming toxin that is secreted by the bacterium as a water-soluble monomeric protein. Upon binding to susceptible cells, αHL assembles via an inactive prepore to form a water-filled homoheptameric transmembrane pore. The N terminus of αHL, which in the crystal structure of the fully assembled pore forms a latch between adjacent subunits, has been thought to play a vital role in the prepore to pore conversion. For example, the deletion of two N-terminal residues produced a completely inactive protein that was arrested in assembly at the prepore stage. In the present study, we have re-examined assembly with a comprehensive set of truncation mutants. Surprisingly, we found that after truncation of up to 17 amino acids, the ability of αHL to form functional pores was diminished, but still substantial. We then discovered that the mutation Ser217 → Asn, which was present in our original set of truncations but not in the new ones, promotes complete inactivation upon truncation of the N terminus. Therefore, the N terminus of αHL cannot be critical for the prepore to pore transformation as previously thought. Residue 217 is involved in the assembly process and must interact indirectly with the distant N terminus during the last step in pore formation. In addition, we provide evidence that an intact N terminus prevents the premature oligomerization of αHL monomers in solution.

). The N terminus of the prepore remains proteasesensitive (22,25). In our hands, no intermediates have been observed in the prepore to pore conversion, which may therefore be a cooperative process (stage 4, Fig. 1B) (25). In the case of wild-type ␣HL, formation of the transmembrane ␤ barrel is accompanied by re-organization of the N terminus, which latches onto the neighboring subunit (15,27). In the fully formed pore, the amino latch is hidden within the cavity of the cap domain and resists proteolysis ( Fig. 2A) (22,25).
The ␣HL pore has been extensively engineered for applications in biotechnology (28). For example, in stochastic sensing a wide variety of analytes can be detected at the single molecule level through the modulation of the current flowing through a single pore (29 -31). Further, the ␣HL pore has recently been used as a nanoreactor to examine chemical reactions at the single molecular level (32)(33)(34)(35). Accordingly, the ␣HL pore is emerging as a useful tool in both basic science and biotechnology (28,31). Therefore, it is especially important to understand the assembly and structure of ␣HL and related ␤PFTs in detail.
In the present report, we shed additional light on the role of the N terminus in the assembly of ␣HL. In an early study, the analysis of truncation mutants suggested that the N terminus plays a crucial role in the transformation of the prepore to an active pore (22). Truncating the polypeptide by just two residues at the N terminus rendered ␣HL completely inactive and arrested assembly at the prepore stage (stage 3, Fig.  1B). However, a later study suggested that truncated polypeptides have limited pore-forming activity (36). In the present work, we resolve this issue through an examination of the original truncation mutants and a new series in which up to 22 residues are deleted. We find that: 1) up to 17 residues can be deleted from the N terminus of ␣HL without complete loss of activity, 2) latch formation is not necessary to form functional pores, 3) residue 217 cooperates with the N terminus during pore formation, and 4) the N terminus prevents the premature oligomerization of monomers in solution.
Synthetic LukS (hlgC) and LukF (hlgB) genes were used to produce point mutants of the leukocidin proteins: LukF-P215N, LukS-P201N, LukS-P202N, and LukS-P201N/P202N. The codon usage in these constructs differs from the natural genes, but the encoded amino acid sequences are unchanged. 4 The PCR mutagenesis/in vivo recombina-FIGURE 1. Proposed assembly pathway for ␣-hemolysin. A, structure of the staphylococcal ␣-hemolysin (␣HL) heptamer (15). Each subunit is shown in a different color. The position of the membrane is shown in tan; B, proposed model for the assembly of ␣HL (15,20,22,25). The N terminus is shown in magenta, and the pre-stem domain is shown in green (stage 1). Water-soluble monomer: both the N terminus and the pre-stem region are protease-sensitive (stage 2). Membrane-bound monomer: the N terminus remains protease-sensitive at this stage, but the pre-stem region becomes protease-resistant (indicated by the dashed line) (stage 3). Prepore: the N terminus remains protease-sensitive. One possibility is that it is exposed to the aqueous phase, as depicted (stage 4). Functional pore: latch formation makes the entire pore protease-resistant. Assembly of the heptameric ␣HL pore is shown, but a similar pathway is proposed for leukocidin, which forms an octameric pore containing four copies each of two subunits LukF and LukS. tion technique was again used to generate these mutants (38,39 Coupled in Vitro Transcription and Translation-Proteins were generated by coupled in vitro transcription and translation (IVTT) by using an Escherichia coli T7-S30 extract system for circular DNA (Promega, no. L1130). The complete amino acid mixture (1 mM) minus cysteine and the complete amino acid mixture (1 mM) minus methionine, supplied in the kit, were mixed in equal volumes to obtain the working amino acid solution required to generate high concentrations of the proteins. The amino acids (2.5 l) were mixed with premix solution (10 l), L-[ 35 S]methionine (1 l, MP Biomedicals, no. 51001H, 1175 Ci/mmol, 10 mCi/ml), plasmid DNA (4 l, 400 ng/l), and T7-S30 extract (7.5 l) supplemented with rifampicin (20 g/ml final) (37). Synthesis was carried out for 1 h at 37°C (for ␣HL mutants) or 30°C (for leukocidin mutants) to produce 25 l of radiolabeled IVTT protein.
Hemolytic Assay-All ␣HL and Luk proteins (WT or mutants) were obtained by IVTT. For the hemolytic assay, in the case of ␣HL, IVTT protein (5 l) was diluted with MBSA buffer (95 l) in the first well of each row (Figs. 3B and 5C) or column (Fig. 4B) of a microtiter plate. In the case of the leukocidins, LukS (5 l) and LukF (5 l) proteins were diluted in MBSA (90 l) in the first well of each row (Fig. 3D). The proteins were then subjected to 2-fold serial dilution with the same buffer across each row (Figs. 3B, 3D, and 5C) or down each column (Fig.  4B), leaving 50 l in each well. Hemolysis was initiated by the addition of an equal volume of washed rRBC (1% in MBSA) to each well and monitored by observing the decrease in light scattering at 595 nm with a Bio-Rad microplate reader (Model 3550-UV) (40).
Limited Proteolysis Assay-WT ␣HL and mutant proteins (IVTT, 25 l) were incubated with rRBC membranes (5 l, 4.2 mg of protein/ml) and MBSA (30 l) for 1 h at 37°C. The washed membrane pellets were subjected to SDS-polyacrylamide gel electrophoresis in 5% gels. Oligomers were obtained from the gels and concentrated as described elsewhere (30 l) (18,41). They were then divided into two portions of 15 l each. One portion was treated with water (1 l), and the other portion was treated with proteinase K (1 l, 0.1 mg/ml final, New England Biolabs, no. P8102S). After incubation for 5 min at room temperature, both samples were treated with phenylmethanesulfonyl fluoride (in 1 l of isopropanol, 10 mM final, Sigma, no. P7626) for 5 min at room temperature. After the addition of 2ϫ sample buffer (17 l) (42), the samples were heated at 95°C for 5 min and subjected to electrophoresis in a 12% SDS-polyacrylamide gel (Fig. 3C). The gel was fixed for 30 min prior to drying and autoradiography.
Preoligomerization of Truncated Mutants-WT ␣HL and mutant proteins were expressed by IVTT (12.5 l). Without the addition of rRBC membranes, the samples were centrifuged (25,000 ϫ g, 10 min), and the supernatants were separated. Supernatant (3 l) was mixed with MBSA (22 l) and incubated at room temperature for 1 h. Sample buffer (25 l, 2ϫ) (42) was then added to each tube. The unheated samples were subjected to electrophoresis in a 10% SDS-polyacrylamide gel (Fig.  6A). The gel was fixed for 30 min prior to drying and autoradiography.
For limited proteolysis, the IVTT supernatants from WT ␣HL and the mutants (4 l) were mixed with MBSA (36 l) and incubated at room temperature for 1 h. The samples were then divided into four tubes (9 l each), treated with proteinase K as described earlier, and subjected to electrophoresis in a 12% SDS-polyacrylamide gel (Fig. 6B).

Construction of ␣HL Mutants to Analyze the Effect of the Residue at
Position 217 on the Conversion of the Prepore to the Active Pore-The role of the N terminus of ␣HL in pore formation was originally investigated before the availability of structural data. In the first study, Walker and coworkers in our laboratory generated four truncation mutants of ␣HL by deleting 2, 11, 22, and 38 residues from the N terminus (22). All formed SDS-stable oligomeric complexes on rRBC but were incapable of lysis. Therefore, these mutants were proposed to be arrested at a "prepore" stage, and it was inferred that the N terminus of ␣HL plays a crucial role in the conversion of the prepore to the active pore. However, as reported (22), all the mutants used in that study carried a point mutation at codon 217 (AGT 3 AAT) that had occurred during the construction of the expression vector. The altered codon encodes an Asn residue instead of a Ser residue. The mutation has no effect on the activity of the full-length protein as discussed in the report (22). Nevertheless, a later study from another laboratory suggested that truncated polypeptides might have limited activity (36). Further, Panchal and coworkers (43) from our laboratory had discovered that a mutation at position 217 could partially reactivate the non-lytic mutant ␣HL-H35N. Therefore, we decided to investigate whether the S217N mutation affects the activity of truncated proteins. In addition, in the Luk polypeptides, the corresponding position is occupied by a Pro residue, which is well conserved in all Luk variants: Pro 215 in LukF and Pro 201 in LukS (Fig. 2, B-D) (19). In consideration of these issues, we made seven mutants of ␣HL by changing Ser 217 in the wild-type protein and selected truncation mutants (␣HL-N⌬3 and ␣HL-N⌬12; Table 1) (22). Similarly, (␣HL-N⌬12)-S217N is the same as ␣HL(A12-293) (22).

Membrane Binding, Oligomerization, and Activity of Ser 217
Mutants of ␣HL-The mutant and WT polypeptides were tested for the ability to bind to rRBC membranes, to make SDS-stable oligomeric complexes, and to lyse rRBC. All seven mutants were able to bind to rRBC membranes as monomers (Fig. 3A). However, the S217P mutation affected the electrophoretic mobility of the monomers, both in the full-length ␣HL-S217P and truncated (␣HL-N⌬3)-S217P proteins. Of the three full-length proteins, ␣HL-S217P migrated more quickly in an 8% SDSpolyacrylamide gel compared with WT ␣HL and ␣HL-S217N. In contrast, (␣HL-N⌬3)-S217P showed a decreased mobility relative to the other two proteins with the same deletion. Therefore, the introduction of Pro at position 217 may have affected the folding of the monomers, which may not be fully denatured in SDS-polyacrylamide gels. All the mutants, except ␣HL-S217P, assembled on rRBC membranes as SDSstable oligomeric complexes.
The ability of ␣HL polypeptides to lyse rRBC is characteristic of whether they are able to convert from non-lytic prepores to functional pores. The hemolytic activity of ␣HL-S217N was only ϳ2-fold less than the WT protein (Fig. 3B). Therefore, as stated by Walker and coworkers (22), the S217N mutation does not have a significant effect on the fulllength protein. However, in contrast to the previous studies, the deletion of two residues (␣HL-N⌬3) and eleven residues (␣HL-N⌬12) from the true WT ␣HL did not inactivate the protein. The ␣HL-N⌬3 mutant is only 2-4-fold less active than WT, whereas the activity of the ␣HL-N⌬12 mutant is ϳ30-fold lower (Fig. 3B). Nonetheless, in complete agreement with the previous study, deleting two or eleven residues from the N terminus of the ␣HL-S217N mutant ((␣HL-N⌬3)-S217N and (␣HL-N⌬12)-S217N, respectively) did abolish activity completely (  22)). Therefore, the S217N mutation affects the activity of truncated mutants but not the full-length protein. Remarkably, the  (20). Each region is colored as in A. Pro 215 is in stick form in cyan. C, LukS-PV watersoluble monomer (19). Each region is colored as in A. Pro 199 (Pro 201 in the corresponding position in LukS) is shown in stick form in cyan. All images were generated using PyMOL version 0.97. D, sequence comparison between ␣HL, LukF, and LukS in the region around Ser 217 (␣HL).

TABLE 1 Nomenclature of N terminal truncation mutants of ␣HL
An Ala codon was placed at the first position of each mutant gene in all the mutants except ␣HL-N⌬1. If the polypeptides were deformylated, the initial Ala would ensure the efficient co-translational removal of the N-terminal methionine (22). However, we do not know whether our IVTT proteins contain fMet or Met at the N terminus. ␣HL-N⌬3 is the same deletion as ␣HL(A3-293) (22), and ␣HL-N⌬12 is the same as ␣HL(A12-293), except that the earlier truncation mutants carried the additional mutation S217N.
S217P mutation produced a different effect. Although the full-length ␣HL-S217P did not make SDS-stable oligomers, the mutant was active on rRBC (Fig. 3, A and B). The activity was at least 100-fold lower when compared with WT protein and, therefore, we cannot exclude the possibility that the observed activity is the result of an undetectable amount of SDS-stable oligomer. By contrast, deleting two residues from the N terminus of ␣HL-S217P ((␣HL-N⌬3)-S217P) increased its activity by ϳ16-fold (Fig. 3B). This observation is paralleled by an increase in the amount of SDS-stable (␣HL-N⌬3)-S217P oligomer.
Limited Proteolysis of ␣HL S217 Mutants-According to the present model for ␣HL assembly, the N terminus takes up different conformations in the prepore and the fully assembled pore (15,22). As a result, the prepore and the pore show different sensitivities toward attack by proteinase K (22). Therefore, a limited proteolysis assay was carried out to evaluate the conformations of the N termini of the various truncation mutants. Gel-purified oligomers of ␣HL, ␣HL-S217N, ␣HL-N⌬3, (␣HL-N⌬3)-S217N, and (␣HL-N⌬3)-S217P were digested with proteinase K and heated to disrupt subunit-subunit interactions before analysis by electrophoresis (Fig. 3C). ␣HL-S217P was not tested, because it did not make SDS-stable oligomers. Similarly, the elevenresidue truncation mutants were omitted, because the protease digestion site is removed with the deletion of eleven residues (22). WT ␣HL heptamer was completely resistant to proteinase K digestion (Fig. 3C), confirming previous studies (22). Similar resistance was observed with ␣HL-S217N, suggesting that latch formation has been completed in this mutant and the N terminus is buried inside the cavity. However, all three truncation mutants, ␣HL-N⌬3, (␣HL-N⌬3)-S217N, and (␣HL-N⌬3)-S217P, were digested with proteinase K (Fig. 3C).
Effect of Amino Acid Substitution at the Position Corresponding to 217 in ␣HL on the Assembly of the Leukocidin Pore-According to sequence alignment and a structural comparison between ␣HL, LukS, and LukF, the position equivalent to 217 in ␣HL is occupied by a Pro residue in the Luk proteins (215 in LukF and 201 in LukS; Fig. 2D). Position 202 in LukS is also a Pro residue. To determine whether these positions in the Luk proteins affect the assembly of the Luk pore, four Luk mutants were generated: LukF-P215N, LukS-P201N, LukS-P202N, and LukS-P201N/ P202N. The N termini of LukF and LukS, respectively, are naturally one and seven residues shorter than that of ␣HL (19). The mutant proteins were tested for their ability to lyse rRBC (Fig. 3D). All the combinations of Luk mutants tested were hemolytic. No significant difference was observed in any of the mutant combinations with respect to the WT proteins (Fig. 3D). It should be noted that the hemolytic activity of the WT Luk subunits on rRBC is normally at least 500-fold lower than WT ␣HL (1).
Analysis of the N-terminal Truncation Mutants of Wild-type ␣HL-From the experiments described above, it was clear that the N terminus of the WT ␣HL is in fact not critically important for the formation of the active pore. Partial activity was preserved after two-and eleven-residue deletions. To further analyze the extent of deletion tolerated by WT ␣HL, a complete series of truncated mutants missing up to 22 residues was generated ( Table 1). All 22 mutants were able to bind to rRBC membranes as monomers (Fig. 4A). All 22 mutants were also able to make SDS-stable oligomers on rRBC membranes (Fig. 4A). A hemolytic assay with rRBC was carried out to test for function (Fig. 4B). Within the 2-h assay period, mutants with up to 17 missing residues (␣HL-N⌬1 to ␣HL-N⌬17) showed detectable activity. Across the series from WT ␣HL to ␣HL-N⌬17, the activity decreased by at least 1000-fold and was accompanied by an increase in the initial lag time. Within the assay period, we did not observe any lysis for the mutants ␣HL-N⌬18 to ␣HL-N⌬20. However, ␣HL-N⌬21 had very weak activity and ␣HL- B, hemolytic assay of ␣HL mutants. IVTT proteins were added into the first well of each row of a microtiter plate and subjected to 2-fold serial dilution across the row with MBSA buffer. The proteins were expressed at closely similar levels as determined by SDS-PAGE of the monomers after translation. Hemolysis was initiated by the addition of washed rRBC, and the decrease in light scattering was monitored at 595 nm for 2 h. C, limited proteolysis of ␣HL mutants. Gel-purified oligomers of IVTT proteins of each mutant were digested with proteinase K at 0.1 mg/ml concentration (see "Experimental Procedures"). Samples were heated to release monomers before electrophoresis in a 12% SDS-polyacrylamide gel: (Ϫ), before proteinase K treatment; (ϩ), after proteinase K treatment. D, hemolytic assay of leukocidin mutants. The effect of changes at the position corresponding to 217 in ␣HL is examined. Both components (LukF and LukS) were synthesized by IVTT, mixed in the first well of the row of a microtiter plate, and subjected to 2-fold serial dilution across the row with MBSA buffer. Hemolysis was initiated by the addition of rRBC, and the decrease in light scattering was monitored at 595 nm for 2 h.
N⌬22 was as active as ␣HL-N⌬13 (Fig. 4B). To investigate these observations further, we made the ␣HL-N⌬21 and ␣HL-N⌬22 mutants again, together with seven additional mutants with deletions up to residue 29 (␣HL-N⌬23 to ␣HL-N⌬29). The activities of ␣HL-N⌬21 and ␣HL-N⌬22 were confirmed (data not shown). The activity of ␣HL-N⌬23 was ϳ8-fold lower than ␣HL-N⌬22, and ␣HL-N⌬24 to ␣HL-N⌬29 did not show any activity during the 2-h assay period (data not shown). In accord with the observed activity, ␣HL-N⌬23 oligomerized on rRBC membranes. Although ␣HL-N⌬24 also showed very weak oligomerization, ␣HL-N⌬25 to ␣HL-N⌬29 did not (data not shown).

Sequence Specificity of the N Terminus of ␣HL in Its Role in Pore Formation-
The previous experiments showed that, although the N terminus of ␣HL is not required for pore formation, truncations reduced activity and increased the initial lag time (Fig. 4B). To determine whether these findings are sequence specific, we generated two additional mutants of ␣HL by exchanging its N terminus with that of LukS or LukF. ␣HL showed a significant level of activity when 22 residues were removed from the N terminus (␣HL-N⌬22, Fig. 4B). However, the activity of ␣HL-N⌬21 was very low. Therefore, we chose to replace missing residues in this mutant with the corresponding residues of LukF and LukS to check whether the activity could be increased. In the first mutant, the first 21 residues of ␣HL were replaced by the first 21 residues of LukF to generate LukF(1-21)HL. Therefore, this mutant has an identical length to WT ␣HL, but it has a different sequence at the N terminus. In the second mutant, the first 21 residues of ␣HL were replaced with the first 18 residues of LukS to generate LukS(1-18)HL. This mutant has both a different length and a different sequence at the N terminus compared with WT ␣HL. The ability to oligomerize and the lytic activity of the mutants were compared with WT ␣HL. Both produced SDS-stable oligomers on rRBC membranes. However, the extent of oligomerization of LukS(1-18)HL is lower than that of WT ␣HL and that of LukF(1-21)HL lower still (Fig. 5A). Both oligomers also migrated faster on SDS-PAGE compared with the WT oligomer (Fig. 5A). A limited proteolytic assay was carried out to explore the conformations of the N termini in the LukS(1-18)HL and LukF(1-21)HL oligomers. Proteins were oligomerized on rRBC membranes, digested with proteinase K and heated to disrupt subunit-subunit interactions (Fig. 5B). Under these conditions, monomers that do not oligomerize are digested into small fragments. WT ␣HL was not digested, even with 0.5 mg/ml proteinase K (Fig. 5B). However, despite its similar length to WT ␣HL, LukF(1-21)HL was completely digested at the N terminus at 0.05 mg/ml proteinase K. At this concentration, LukS(1-18)HL was partly resistant to proteolysis. When the lytic activities of these proteins were measured, the full-length chimeric protein, LukF(1-21)HL, showed ϳ60-fold lower activity compared with WT ␣HL (Fig. 5C). Interestingly, the shorter chimeric protein, LukS(1-18)HL, was only 4-fold less active (Fig. 5C). For both mutants, the initial lag time was decreased, and the lytic activity increased compared with the truncated protein, ␣HL-N⌬21 (Figs. 4B and 5C).
Preoligomerization of Truncated Mutants-In the crystal structure of the ␣HL heptamer, the N terminus of each protomer extends away from the ␤-sandwich domain to form a protease-resistant latch with a neighboring subunit inside the cap domain. However, the protease sensitivity of the water-soluble monomer suggests that the N terminus is in an exposed conformation (22). Therefore, it is likely that the N terminus relocates during pore formation. Consistent with these structural considerations, fluorescent probes attached to the N terminus of ␣HL undergo a change in environment during pore formation (27). Further, in the structures of the LukF and LukS-PV monomers, the N termini adopt a ␤-strand conformation and thereby extend the inner ␤ sheet of the ␤-sandwich domains by one strand (Fig. 2B) (19,20). Because of the overall similarities between the folds in the ␣HL, LukF, and LukS-PV structures, it seems likely that the N terminus of the ␣HL monomer also folds against the ␤-sandwich domain. It is possible that this conformation of the N terminus prevents the oligomerization of ␣HL in solution by masking the surfaces of individual subunits (20). Therefore, we analyzed the ability of ␣HL to oligomerize in solution when the N terminus is truncated (Fig. 6A). The truncation mutants (Table 1) were expressed by IVTT and incubated at room temperature without the addition of red cell membranes (see "Experimental Procedures"). Although the IVTT reaction mixture itself contains ϳ1-2 g/l of E. coli lipids (including phosphatidylethanolamine and cardiolipin), 5 the extent of premature oligomerization of WT ␣HL was found to be very low. However, as residues were removed from the N terminus, a gradual increase in the extent of oligomerization was observed up to ␣HL-N⌬18 as analyzed by SDS-PAGE. From ␣HL-N⌬19 to ␣HL-N⌬21, premature oligomerization decreased and only a very weak band of oligomers was seen for ␣HL-N⌬22. No oligomers were observed for truncation mutants between ␣HL-N⌬23 and ␣HL-N⌬29 (Fig. 6A).
Monomers of ␣HL in solution are susceptible to proteolysis at two sites: at the N terminus and in the pre-stem domain. Oligomerization occludes the pre-stem inside the prepore, protecting it from proteolysis. Because the N terminus is already removed in the truncation mutants, an increased resistance to proteolysis in the stem domain would be  (Table 1). IVTT proteins were oligomerized on rRBC membranes, and the proteins on the washed membranes were subjected to electrophoresis in an 8% SDS-polyacrylamide gel. M, protein molecular weight markers. B, hemolytic assay of ␣HL truncation mutants. IVTT proteins were added into the first well of each column of a microtiter plate and subjected to 2-fold serial dilution down the column with MBSA buffer. Hemolysis was initiated by the addition of rRBC, and the decrease in light scattering was monitored at 595 nm for 2 h. expected if these mutants prematurely oligomerize in solution. To analyze this possibility, a limited proteolysis assay was carried out on ␣HL-N⌬17 to ␣HL-N⌬21 after the described treatment (Fig. 6B). The WT protein was completely digested at 50 g/ml of proteinase K showing the distinctive pattern of digestion at both the N terminus and the stem domain (22). By contrast, the truncated mutants were less susceptible to proteolysis and undigested polypeptides were apparent even at 500 g/ml of protease in accord with the notion of premature oligomerization (Fig. 6B).

DISCUSSION
Knowledge about how ␣HL converts from its water-soluble monomeric form into the functional heptameric pore has contributed greatly to an understanding of the assembly and function of both small and large proteins in the ␤PFT family. For example, truncation mutagenesis of ␣HL produced the first evidence for a prepore intermediate in the assembly of the ␤PFT (22). The existence of prepores in small ␤PFT, including ␣HL, was later verified in other studies (25, 26, 44 -46). Similar prepore intermediates were more recently reported for members of the family of cholesterol-dependent cytolysins, such as perfringolysin (47)(48)(49)(50) and pneumolysin (51). This suggests that all ␤PFT share a common mechanism of pore formation (24).
Even before the crystal structure of the ␣HL pore was solved, the N terminus of ␣HL was thought to be vital for pore formation. Walker and co-workers (22) reported that ␣HL is unable to form lytic pores and arrests at the prepore stage when a deletion of two residues is made at the N terminus. Although Vandana and co-workers (36) later reported that ␣HL is active after the removal of four residues, the hemolytic activity of their truncation mutant was low. Nevertheless, we have reexamined this question, paying particular attention to the fact that our original mutants (22) contained the mutation S217N, which was introduced during the construction of the expression plasmid. Remarkably, as shown in the present work, this mutation, which is distant from the N terminus of ␣HL in both sequence and space, plays an important role in pore formation only when the N terminus of polypeptide is truncated.
In the present study, we have demonstrated that the true WT ␣HL is able to make SDS-stable oligomers even after the removal of 22 residues from the N terminus (Fig. 4A). This indicates that these mutants reach at least the prepore stage in assembly. To our surprise, mutants with up to 17 residues deleted were also able to lyse rRBC by forming functional pores, albeit with diminished activity compared with WT ␣HL (Fig. 4B). Therefore, we have demonstrated that the N terminus of ␣HL is not in fact necessary for pore formation. Our results also show that the N terminus of ␣HL can exist in different conformations in the fully functional pore ( Table 2): 1) If the full-length N terminus is intact, it forms an amino latch, as in the cases of the WT polypeptide, corroborated by the crystal structure ( Fig. 2A), and ␣HL-S217N. In this conformation, the N terminus is inaccessible to proteases (Fig. 3C); 2) If the N terminus is partly deleted, it does not form the amino latch. What is left of the N terminus, or more strictly the site of proteolytic cleavage, remains accessible, as in the cases of ␣HL-N⌬3 and (␣HL-N⌬3)-S217P (Fig. 3C); 3) If the N terminus is completely removed, certain mutants still form active pores, as in the cases of ␣HL-N⌬21 and ␣HL-N⌬22 (Fig. 4B).
The prepores of ␤PFT are defined by two properties: 1) They are fully formed oligomers with subunits arranged around a central axis; 2) The lipid bilayer is not breached by them (22,24). The central axis of the prepore lies perpendicular to the membrane surface (9,50,51), and the residues that will become the transmembrane ␤ barrel have not yet undergone the required conformational change (25,26,46,48). In the case of ␣HL, the prepore was believed to feature a protease-sensitive N terminus ( Table 2) that was needed to form a protease-resistant latch upon transformation to the fully assembled pore (22). Therefore, protease sensitivity in the oligomeric state was regarded by us as a diagnostic test for prepore formation and to be absent in the fully assembled functional pore. Although this remains the case for WT ␣HL, the findings reported here show that the N termini of certain active ␣HL mutants remain accessible to proteases in the fully assembled pore (Fig.  3, B and C). Therefore, a protease-sensitive N terminus is no longer solely diagnostic of the prepore structure. For example, ␣HL-N⌬3 forms SDS-resistant oligomers with about the same efficiency as WT-␣HL (Fig. 3A) and has ϳ25-50% of the activity (Fig. 3B). However, the oligomerized protein is completely (at least 95%) digested near the N terminus with proteinase K (Fig. 3C).
As documented above, when the N terminus of ␣HL is completely removed, the polypeptide can still form active pores (Fig. 4B). Mutants in which the N terminus was replaced also showed pore-forming activity (Fig. 5C). Latch formation was inhibited in a full-length protein made by the substitution of the first 21 residues of ␣HL with the corresponding residues of LukF, LukF(1-21)HL (Fig. 5B). The ability of this mutant to oligomerize was poor (Fig. 5A) and, accordingly, the activity was low (Fig. 5C), and we cannot exclude the possibility that it derives from an undetected fraction of protease-resistant oligomer. In LukS(1-18)HL, the first 21 residues of ␣HL are replaced by the first 18 residues of LukS. Latch formation was also absent (Fig. 5B), and this mutant showed increased oligomerization (Fig. 5A) and higher lytic activity compared with LukF(1-21)HL (Fig. 5C, ϳ25% of WT ␣HL). Therefore, LukS(1-18)HL is another example of an ␣HL mutant that exhibits lytic activity without latch formation.
Although the N terminus of ␣HL is not necessary for pore formation, deletions at the N terminus of the WT protein both increased the initial lag time and decreased the overall extent of hemolysis (Fig. 4B). The reduced activity could arise for one or more reasons: 1) Binding of the truncated mutants to red cell membranes is affected; 2) Assembly of monomers to form the prepore is inhibited; 3) The prepore to pore transformation is retarded; 4) The protein oligomerizes prematurely after synthesis, so that less is available for pore formation. Possibilities 1 and 2 cannot have a significant effect as truncation mutants up to ␣HL-N⌬23 bind to red cells and oligomerize efficiently (Fig. 4A). Possibility 4 may contribute to the reduced activity of some of the mutants. ␣HL-N⌬10 to ␣HL-N⌬21 do show significant premature oligomerization (Fig. 6A), but the extent of premature oligomerization is not sufficient to account for the observed reduction in activity. Therefore, it is reasonable to infer that extended truncations at the N terminus of ␣HL retard the unfolding of the pre-stem domain during the prepore to pore transformation, possibility 3. This argument is further supported by experiments with LukS(1-18)HL and LukF(1-21)HL, in which 21 missing residues at the N terminus of ␣HL are replaced with residues from the N termini of LukS and LukF (Fig. 5C). In both cases, pore-forming activity is regained, and in the case of LukS(1-18)HL it is ϳ25% of the value for

TABLE 2
Properties of the prepore and the pore of ␣HL WT ␣HL. In the latter case, by contrast with WT ␣HL, the N-terminal latch is not formed. Therefore, the stimulation of the prepore to pore transition by the N terminus of ␣HL must be independent of the final conformation of the N terminus and its sequence. In an earlier study, Valeva and coworkers (52) provided evidence that, during heteroheptamer formation by certain pairs of mutant ␣HL subunits, the N terminus in one class of subunits can activate its own pre-stem intramolecularly and that this activation is transmitted intermolecularly within the prepore to the pre-stem of a normally defective subunit. Further, when the part of the pre-stem that becomes the lower half of the ␤ barrel is deleted, the resulting mutant assembles spontaneously in the absence of membranes to form heptameric structures in which the N termini have formed latches (37), again indicating a cooperative interaction between the N terminus and the pre-stem.
We have shown here that activity is completely lost in N-terminal truncation mutants of ␣HL only when the truncations are coupled with the S217N mutation (for example, see Fig. 3B). This observation suggests that the nature of the side chain at position 217 controls the prepore to pore transition together with the N terminus. Ser 217 lies in a short loop in the rim domain that connects short stretches of 3 10 -and ␣-helix (stretches C and D, Fig. 2A) (15). Both in the monomer and in the oligomer, the loop lies distant from the N terminus ( Fig. 2A, Ala 1 C␣ to Ser 217 C␣: monomer 30 Å, heptamer 36 Å (same subunit), 26 Å and 45 Å (adjacent subunits)). Therefore, it is likely that indirect, long distance interactions operate between the N terminus and position 217 during assembly. This effect is further illustrated by the properties of ␣HL-S217P, which has weak hemolytic activity that is strongly enhanced after the removal of two residues at the N terminus in (␣HL-N⌬3)-S217P. Interestingly, the WT Luk proteins, which already have Pro at the positions corresponding to Ser 217 in ␣HL, are already "shortened" compared with ␣HL: LukF by one residue, and LukS by seven. However, unlike truncated ␣HL mutants, the Luk proteins remain active when the Pro residues are replaced by Asn (Fig. 3D).
Indirect interactions between residues, over both short and long distances, have been observed in many proteins. For example, in the metalloenzyme carbonic anhydrase II, His 94 , His 96 , His 119 , and a water molecule act as primary ligands for the active site zinc ion. However, the affinity of zinc(II) binding can be fine-tuned by manipulating a set of second shell residues, Gln 92 , Asn 244 , Glu 117 , and Thr 199 (53). These residues do not interact directly with the zinc ion but form a hydrogenbond network with the primary shell residues. Although, short distance indirect interactions are usually readily defined, long distance interactions are often difficult to trace through a protein structure (54). For example, the I60V mutation stabilizes the loop 64 -74 in p-nitrobenzyl esterase when present in the mutant background of H322R, M358V, and Y370F (55)(56)(57). Similarly, in the same background, L144M stabilizes the loop 265-275. Neither mutation is effective in the context of WT p-nitrobenzyl esterase. Both Ile 60 and Leu 144 are located at least 20 Å away from residues 322, 358, and 370, and the interactions causing loop stabilization are not readily traced within the protein structure.
Long distance interactions have been observed previously in ␣HL (43). His 35 is a crucial residue, which is directly involved in subunitsubunit interactions involving Ile 98 and Tyr 101 in the fully assembled ␣HL pore (15). Mutations at this position eliminate or greatly reduce the hemolytic activity of the protein (58 -61). One such mutant, ␣HL-H35N, is arrested in assembly at the prepore stage (61). However, second-site mutations are able to re-activate H35N (43). These mutations are clustered predominantly in three distinct regions: within the sequences 107-111 and 144 -155, which constitute part of the triangle region that flanks the stem domain ( Fig. 2A), and within the sequence 217-228, which is in the rim domain. These sequences are far from His 35 in the structure of the pore, and therefore, they "communicate" with position 35 through long distance interactions (Fig. 2A).
These observations provide insight into the present study, as S217C is one of the second-site mutations that re-activates ␣HL-H35N (43). To create a functional pore, the pre-stem region in the prepore must unfold to form the transmembrane ␤ barrel. This transformation is accompanied by conformational changes within the stretches of amino acids that flank the pre-stem domain in the monomer and become the triangle region, which connects the barrel to the cap domain in the fully assembled pore (20) (Fig. 2A). Therefore, the N terminus, position 35, the triangle region, and position 217, as well as the pre-stem residues, participate together in the prepore to pore transition ( Fig. 2A).
In the LukF monomer, the N terminus adopts a ␤-strand conformation and extends the inner ␤ sheet of the ␤-sandwich domain by one strand (20). Because of the structural similarities between the two proteins, it is reasonable to assume that the N terminus of ␣HL adopts a similar conformation in the water-soluble monomer. This configuration of the N terminus has been proposed to mask the protein surface required for oligomerization to the prepore (20,62). This idea is supported by the present study, in which we found that ␣HL oligomerizes in solution when the N terminus is truncated (Fig. 6A). Prematurely assembled oligomers cannot form functional pores on red cell membranes (62). Therefore, premature oligomerization reduces the amount of active monomers available for pore formation, which is in part (but not fully) responsible for the reduction of the activity of the truncated ␣HL polypeptides (Figs. 4C and 6A).
In summary, we have shown: 1) The N terminus of ␣HL, and therefore latch formation in the fully assembled pore, is not essential for pore formation, as previously believed; 2) Full-length ␣HL polypeptides with N termini with altered sequences are active without latch formation; 3) Residue 217 is critical in the assembly process and interacts indirectly with the N terminus during the last step of pore formation; and 4) The N terminus prevents the premature oligomerization of ␣HL monomers in solution. These findings fill gaps in our knowledge of the assembly pathway of ␤PFTs and open up new avenues for the protein engineering of ␣HL. For example, we have recently showed that the LukF and LukS proteins can be chained together by connecting the C terminus of LukS to the N terminus of LukF via a 15-residue Ser/Gly linker (18). The dimer showed increased activity compared with WT LukF and LukS. The new knowledge about the N terminus of ␣HL suggests that ␣HL subunits might also be linked together in a head-to-tail fashion, which will permit us to carry out more sophisticated engineering of the pore than has been possible previously (28). Concatenated subunits should also be useful in further studies of the assembly of ␤PFTs, because they might resemble assembly intermediates (24). In the accompanying paper (63), we examine mutants of the LukF and LukS subunits of the leukocidin pore and, in accord with the present study, show that, when the N termini of both subunits are deleted, functional pores are still formed.