Interactions between Residues in Staphylococcal (cid:97) -Hemolysin Revealed by Reversion Mutagenesis*

-Hemolysin ( (cid:97) HL), a pore-forming polypeptide of 293 amino acids, is secreted by Staphylococcus aureus as a water-soluble monomer. Residues that play key roles in the formation of functional heptameric pores on rabbit red blood cells (rRBC) have been identified previously by site-directed mutagenesis. (cid:97) HL-H35N, in which the histidine at position 35 of the wild-type sequence is re- placed with asparagine, is nonlytic and is arrested in assembly as a heptameric prepore. In this study, second- site revertants of H35N that have the ability to lyse rRBC were generated by error-prone PCR under condi- tions designed to produce single base changes. The analysis of 22 revertants revealed new codons clustered pre- dominantly in three distinct regions of the H35N gene. One cluster includes amino acids 107–111 (four rever- tants) and another residues 144–155 (five revertants). These two clusters flank the central glycine-rich loop of (cid:97) HL, which previously has been implicated in formation of the transmembrane channel, and encompass residues Lys-110 and Asp-152 that, like His-35, are crucial for lytic activity. The third cluster lies in the region spanning amino acids 217–228 (eight revertants), a region previously unexplored by mutagenesis. Single revertants were found at amino acid positions 84 and 169. When compared with H35N, the heptameric prepores formed by the revertants underwent more rapid conversion to fully assembled pores, as determined by conformational analysis by limited proteolysis. The rate of conversion to the fully assembled pore was strongly correlated with hemolytic activity. Previous work has suggested that the N terminus of (cid:97) HL and the central loop cooperate in the final step of assembly. The present study suggests that the key N-terminal residue His-35 operates in con-junction with residues flanking the loop and C-terminal residues in the region 217–228. Hence, reversion mutagenesis extends the linear analysis that has been pro- vided by direct point mutagenesis.

␣-Hemolysin (␣HL), a pore-forming polypeptide of 293 amino acids, is secreted by Staphylococcus aureus as a water-soluble monomer. Residues that play key roles in the formation of functional heptameric pores on rabbit red blood cells (rRBC) have been identified previously by site-directed mutagenesis. ␣HL-H35N, in which the histidine at position 35 of the wild-type sequence is replaced with asparagine, is nonlytic and is arrested in assembly as a heptameric prepore. In this study, secondsite revertants of H35N that have the ability to lyse rRBC were generated by error-prone PCR under conditions designed to produce single base changes. The analysis of 22 revertants revealed new codons clustered predominantly in three distinct regions of the H35N gene. One cluster includes amino acids 107-111 (four revertants) and another residues 144 -155 (five revertants). These two clusters flank the central glycine-rich loop of ␣HL, which previously has been implicated in formation of the transmembrane channel, and encompass residues Lys-110 and Asp-152 that, like His-35, are crucial for lytic activity. The third cluster lies in the region spanning amino acids 217-228 (eight revertants), a region previously unexplored by mutagenesis. Single revertants were found at amino acid positions 84 and 169. When compared with H35N, the heptameric prepores formed by the revertants underwent more rapid conversion to fully assembled pores, as determined by conformational analysis by limited proteolysis. The rate of conversion to the fully assembled pore was strongly correlated with hemolytic activity. Previous work has suggested that the N terminus of ␣HL and the central loop cooperate in the final step of assembly. The present study suggests that the key N-terminal residue His-35 operates in conjunction with residues flanking the loop and C-terminal residues in the region 217-228. Hence, reversion mutagenesis extends the linear analysis that has been provided by direct point mutagenesis.
␣-Hemolysin (␣HL), 1 a polypeptide of 293 amino acids, is secreted by Staphylococcus aureus as a water-soluble monomer and assembles to form a heptameric pore on pure lipid bilayers or on biological membranes such as those of red blood cells (1). Based on biochemical, biophysical, and molecular genetic stud-ies, four stages in the assembly of ␣HL have been defined (Refs. 2, 3, and the accompanying paper (4)). Monomeric ␣HL in solution (Structure 1) comprises two domains connected by a central glycine-rich loop (residues 119 -143) (5,6). ␣HL first binds to the membrane surface as a monomer (Structure 2). A nonlytic oligomer consisting of seven subunits (3,7) is then formed through interactions in which the C-terminal domain may play a predominant role (Structure 3) (2, 4). The subunits then further penetrate the membrane to form the heptameric lytic pore (Structure 4). Recent evidence suggests that the central loop lines part of the lumen of the transmembrane channel in the fully assembled structure (8,9).
In vitro mutagenesis has identified residues that are important for the membrane binding, assembly, and pore forming activity of ␣HL (summarized in Ref. 4). For example, replacement of His-35 by Leu, Ile, Ser, Thr, Arg, Pro (10,11), Cys (12), Asn, Trp, or Gln (13) has been shown to eliminate or greatly reduce hemolytic activity. However, when Cys-35 in H35C was alkylated with iodoacetamide to form the modified residue S-carboxamidomethylcysteine, hemolytic activity was restored (13). Therefore, the volume of the residue at position 35, and perhaps other factors such as the polarity and hydrogen bonding potential of the side chain, play a crucial role, while the ability of the residue to ionize is unimportant (13). Recently, four additional amino acids, Asp-24, Glu-70, Lys-110, and Asp-152, have been identified, which, when individually replaced with cysteine, yield hemolysins with greatly reduced hemolytic activity (4). Three of these mutants (D24C, K110C, and D152C) form oligomers but not functional pores, suggesting that like H35N (3) they are arrested at the prepore stage of assembly (Structure 3). By contrast, the loss of activity in the mutant E70C is attributed to its low affinity for the rRBC membrane.
The aim of the present study was to produce second-site mutations that restore the activity of the H35N mutant, and thus to step beyond the linear analysis of point mutagenesis by obtaining information about the interactions between key residues in different domains of the polypeptide chain.

EXPERIMENTAL PROCEDURES
Error-prone PCR-The template for error-prone PCR (14) was the ␣HL-H35N gene in the plasmid pT7NPH-8S (13), which was linearized with EcoRI. The 5Ј primer was 5Ј-CGGGATCCTAATACGACTCAC-TATAGGG-3Ј, the last 20 nucleotides of which are complementary to the T7 promoter region upstream from the NdeI site (CЈ-ATATG) in pT7NPH-8S that contains the initiation codon for ␣HL. The 3Ј primer was 5Ј-AAACATCATTTCTGAAGCTTTCGGCTAAAG-3Ј, the last 17 nucleotides of which are complementary to a sequence in the proximal 3Ј-untranslated region that contains a HindIII site (15). Random mutations were introduced by: 1) using a large number of PCR cycles (fifty); 2) using a high concentration of dNTPs (400 M), which increases the error rate of tag-polymerase (16). Upon screening, a reversion frequency of 1 in 200 was obtained (see below). The PCR product was digested with NdeI and HindIII, gel purified, and ligated to gel-purified pT7NPH-8S vector that had been digested with the same restriction enzymes. The ligation product was precipitated with ethanol and resuspended in 10 mM Tris-HCl, pH 7.5, containing 1 mM EDTA, for electroporation into Escherichia coli JM109(DE3) cells (Promega).
Electroporation-Electroporation-competent E. coli JM109(DE3) were prepared as described and stored in 10% glycerol at Ϫ80°C (16). Electroporation was carried out using a Bio-Rad Gene Pulser (2.5 kV, capacitance 25 microfarads, resistance 200 ohm). The ligated DNA (20 ng in 2 l) was added to thawed cells (200 l), and the mixture was transferred to a prechilled cuvette with a 0.2-cm electrode gap (Bio-Rad, no. 165-2086). The cuvette was placed in the sample chamber and the pulse applied. The electroporated cells were incubated in 1 ml of SOC medium (16) for 30 min at 37°C, before plating onto LB plates containing ampicillin (50 g/ml).
Screening for Revertants-Small colonies that appeared after ϳ8 h growth at 37°C were replicated onto nitrocellulose filters, which were then placed colony side up onto a second LB-ampicillin plate and incubated overnight at 37°C. The replicate colonies were lysed by placing the nitrocellulose filters (colony side up) for 50 min onto Whatman paper (3MM) soaked with 0.1 M NaHCO 3 , 1% Triton X-100 (v/v) containing 2 mg/ml lysozyme in a closed dish saturated with chloroform vapor (17). The filters were then washed three times, 15 min each wash, with K-PBSA (20 mM KH 2 PO 4 , 150 mM NaCl, pH 7.4 containing 1 mg/ml BSA), by gently floating the filters (colony side up) on the surface of the wash buffer. The filters were then inverted onto rabbit blood agar plates prepared as described (18). Colonies showing distinct zones of hemolysis, 1-4 mm diameter, within 24 h were individually replated, and the screening procedure was repeated two to three times until pure populations of colonies exhibiting hemolytic activity were obtained. Plasmid DNA was recovered from each of the revertants. To ensure that mixed plasmids had not been obtained, each plasmid was used to transform E. coli XL-1 from which fresh plasmid was prepared. Hemolytic activity was confirmed after coupled in vitro transcription and translation (IVTT) followed by lysis of rRBC in microtiter wells (see below). The entire ␣HL gene was then sequenced to identify the mutation(s) responsible for restoring the activity of the H35N mutant. For convenience, the revertants are designated with an asterisk, i.e. D108G * is ␣HL with both the H35N and D108G mutations. Full details are in Table I.
In Vitro Transcription and Translation-␣HL-H35N revertant polypeptides were made by coupled IVTT in an E. coli S30 extract (Promega No. L4500) supplemented with rifampicin, [ 35 S]methionine, and T7 RNA polymerase, with supercoiled plasmid DNA as template (15). To obtain high concentrations (Ͼ10 g/ml) of the mutant polypeptides with a specific activity suitable for rapid autoradiography or PhosphorImager analysis, separate reactions containing either a complete amino acid mix or amino acid mix minus methionine were carried out at 37°C for 1 h and then mixed. The IVTT reaction mix was used directly for hemolytic assays, measurements of binding and oligomerization, and conformational analysis by limited proteolysis.
Hemolysis Assay-IVTT mix (20 l) diluted with K-PBSA (80 l) was placed in well 1 of a microtiter plate and subjected to 2-fold serial dilution with K-PBSA (50 l in each well). Washed rRBC (1% in K-PBSA, 50 l) were then added to the wells, and the plate was monitored for 1 h at 20°C in a Bio-Rad microplate reader (model 3550-UV). Lysis was measured by observing the decrease in light scattered at 595 nm.
Binding of H35N Revertant Polypeptides to rRBC and Subsequent Oligomerization-IVTT mix (10 l) was added to 11.1% washed rRBC (90 l), and the mixture was incubated at 20°C for 1 h. The suspension was divided into two tubes before centrifugation for 5 min at 16,000 ϫ g. The pelleted cells, or membranes where lysis had occurred, were resuspended in K-PBSA (500 l) and centrifuged again. The pellet from one tube was solubilized in 1 ϫ loading buffer (30 l,Ref. 19). The sample was warmed to 45°C for 5 min and subjected to overnight electrophoresis at 60 V in a 12% SDS-polyacrylamide gel. The extent of binding of the H35N revertant to the rRBC membranes and the fraction converted to oligomers was determined by autoradiography of the dried gel. The pellet from the second tube was subjected to limited proteolysis (see below).
Limited Proteolysis of the Oligomerized H35N Revertants-Pelleted cells or membranes to which the mutant ␣HL had bound (see above) were resuspended in K-PBSA (36 l). Half of the sample was treated with proteinase K (1 l, 1 mg/ml) for 5 min at room temperature, while the other half was treated with water (1 l). The reaction was stopped by treatment with phenylmethanesulfonyl fluoride (1 mM final) for 5 min at room temperature followed by the addition of 5 ϫ gel loading buffer. The samples were heated to 95°C for 5 min to dissociate oligomers and subjected to electrophoresis in a 12% SDS-polyacrylamide gel, followed by autoradiography of the dried gel.

RESULTS
Identification of Revertants of ␣HL-H35N Produced by Mutagenesis with Error-prone PCR-Mutations were introduced into the ␣HL-H35N gene by error-prone PCR. Revertants of the inactive H35N were identified by screening colonies "fixed" on nitrocellulose for their ability to lyse rRBC in blood agar plates. A fraction of the expressed ␣HL that is immobilized on the nitrocellulose filter must leach out to cause lysis of the cells. Twenty-two revertant genes were sequenced (Table I). Of the 22, four exhibited the same base substitution and hence the same amino acid change (Asp-108 3 Gly). This over-representation suggests that the mutation occurred during an early PCR cycle. Only the 19 apparently independent revertants were further studied. Seventeen of these had sequence changes that produced single amino acid changes. The remaining two (H35N/D227N/R236S and H35N/T11A/D227A) had a second amino acid substitution. Because these two revertants contain mutations at a common site (Asp-227) and because a neighboring mutation, Phe-228 3 Leu (Table I), also rescues H35N, it is assumed that the changes at position 227 are responsible for the activity of these mutants. One revertant, designated ⌬222 * , was missing three bases (TCT or TTC) in the sequence TCT-TCA that encodes Ser-221 and Ser-222, resulting in replacement of the Ser-Ser doublet by a single serine residue.
Hemolytic Activity of H35N Revertants-To determine the efficiency of translation of the revertants, the 35 S-labeled IVTT products were quantitated by electrophoresis and autoradiography. All the revertants were translated as well as WT-␣HL and the H35N mutant (data not shown). Three revertants had two bases (CA) missing from the NdeI (CЈ-ATATG) cloning site (Table I). This change did not affect the efficiency of translation of these mutants, despite its proximity to the initiator methionine codon.
A portion of each IVTT mix was used to compare the hemolytic activity of the revertants. In a quantitative assay, the 19 mutants showed a wide range of activities (Table I). As expected from the screening procedure, all were more active than H35N and a few had activity comparable with that of WT-␣HL.
Binding of Revertants to rRBC and Subsequent Oligomerization-A portion of the same IVTT mix was used to study the binding and subsequent oligomerization of the revertant polypeptides on rRBC. WT-␣HL forms SDS stable oligomers that are conveniently analyzed by SDS-polyacrylamide gel electrophoresis. Binding was carried out for 1 h at 20°C, which is sufficient for maximal binding and oligomerization in the case of WT-␣HL. All the revertants bound to rRBC and formed oligomers that were stable in SDS (Fig. 1).
Limited Proteolysis of the Revertants on rRBC Membranes-Previous studies have revealed that ␣HL mutants that are arrested as membrane-bound monomer or at the heptameric prepore stage (such as H35N) are susceptible to cleavage at the N terminus by proteinase K, while this site is occluded in the fully assembled pore (2, 3). The proteolytic susceptibilities of all 19 independent revertants were determined after binding to rRBC for 1 h at 20°C (Fig. 2). After this period, WT-␣HL is fully assembled and completely resistant to proteolysis (Fig. 2). Lysis by WT-␣HL is complete after 20 min. By contrast, the nonlytic H35N is completely proteolyzed (Fig. 2). There was a strong correlation between the proteinase K resistance of a mutant after 1 h of binding and the rate of hemolysis (Fig. 2). DISCUSSION Point mutagenesis, especially systematic scanning mutagenesis (20), is valuable for obtaining information about the functional roles of individual residues and short sequences of residues in a polypeptide, provided that supporting evidence confirming the structural integrity of the mutant molecules is obtained. A summary of the results obtained by scanning point mutagenesis of ␣HL is given in the accompanying paper (4), in which residues involved in binding to rRBC, oligomerization, and pore formation are identified. Despite its utility, point mutagenesis does not usually provide definitive information about interactions between residues that lie far apart in the linear sequence of the polypeptide chain. One way of obtaining such information is to obtain second-site revertants of an inactive protein by "random" mutagenesis (21)(22)(23)(24). In this study, we have located 16 residues that interact with the key residue  1. Binding of WT-␣HL, ␣HL-H35N and the revertants of H35N to rRBC and subsequent oligomerization. In vitro translated ␣HL polypeptides, radiolabeled with 35 S, were allowed to bind to rRBC for 1 h at 20°C. Bound monomer and SDS stable oligomers were then detected by electrophoresis of the unheated samples in a 12% SDS-polyacrylamide gel followed by autoradiography of the dried gel. The revertants of H35N are marked with an asterisk ( * ), signifying that they still contain Asn-35 as well as the designated mutation. His-35 (3, 10 -13) by seeking revertants of ␣HL-H35N, a mutant with greatly reduced lytic activity. H35N is defective in the last step of hemolysin assembly: conversion of a heptameric prepore to the fully active pore (3,13). As hoped, we have found mutants in which this step is repaired.
Nineteen independent revertants of H35N involving 16 amino acid residues were obtained by error-prone PCR. It must be noted that this method is far from "definably random" (24). First, only one base in a codon is changed and hence only a subset of the 19 possible amino acid substitutions can occur at each position. Second, Taq polymerase, as used here, is biased toward errors at AT base pairs; 17 out of 22 independent base changes occurred at AT pairs (Table I). Further, transitions (15/22) were more frequent than transversions (7/22), which are twice as likely on purely statistical grounds.
In many cases, rather subtle changes, e.g. Ser-217 3 Cys, Asp-227 3 Asn, restored activity to ␣HL-H35N. This may reflect the fact that H35N is poised at the brink of activity. H35Q, which contains an additional methylene group is weakly active, while H35CamC, which contains an additional -SCH 2group, has substantial activity (13). H35N is correctly folded as demonstrated by limited proteolysis in solution, 2 and the defect is in a single late step of assembly (3). It seems likely that reversion mutagenesis would be favored by such a situation of minimal disablement. We were also aided by the development of a powerful screening procedure and a high reversion frequency provided by the large number of acceptable reversion sites.
Fourteen of the 16 amino acids affected in the revertants are clustered in three regions of the polypeptide chain (Fig. 3). Four revertants had mutations between amino acids 107 and 111 inclusive, while five revertants (at four amino acid positions) had mutations between amino acids 144 and 155. These two clusters flank the central loop, which plays an important role in channel formation (3-5, 8, 9, 25, 26). Further, Lys-110 and Asp-152, which are critical for lytic activity (4), are located in the two clusters (Fig. 3). These findings are in keeping with the demonstration that the N terminus of ␣HL and the central loop cooperate in the final step of assembly (3). Of five residues identified as crucial for pore formation by cysteine scanning mutagenesis (4), three are interconnected by this study (His-35, Lys-110, and Asp-152). The integrity of these three residues and Asp-24, which was not identified here, is required for the final step of assembly. The fifth mutant, E70C, is defective in binding and therefore would not be expected to be linked with His-35. A third cluster of 8 revertants (at six amino acid positions) is located between amino acids 217 and 228, a region that was not explored in previous studies.
If the interactions revealed by reversion mutagenesis take place within a single polypeptide chain, the findings imply that the N-and C-terminal thirds of ␣HL cannot be considered as completely independent domains, although they contain distinctive distributions of functional residues (4). Interactions between the N and C termini of monomeric ␣HL have been demonstrated directly in experiments in which they are synthesized separately and recombined to form a functional hemolysin (25,27). Perhaps the regions around His-35 and residues 217-228 form a point of contact between the two halves. Helix contacts in membrane proteins such as the a subunit of E. coli F 1 F 0 -ATPase (28) and the E. coli lactose permease (29,30) have been proposed, based on the existence of second-site revertants. Alternatively, because the fully assembled pore is a heptamer, it is quite possible that intersubunit interactions are corrected in the revertants.  2. Hemolytic activity and limited proteolysis of the oligomerized ␣HL-H35N revertants. For each ␣HL polypeptide, the window shows a hemolysis assay of the intact molecule, as monitored for 1 h at 20°C in an automated microplate reader. The in vitro translated protein was diluted 40-fold in the assay mix. Below each assay trace is an autoradiogram of a 12% SDS-polyacrylamide gel showing the proteolytic pattern of the 35 S-labeled ␣HL polypeptide after assembly on rRBC for 1 h at 20°C at the same dilution used in the hemolysis assay. Treatments were with water (Ϫ) or proteinase K (ϩ) at 50 g/ml for 5 min. Oligomers were dissociated by heating before SDS-polyacrylamide gel electrophoresis. The revertants are designated as described in Fig. 1 (legend). The last two shown are, in order, F228 * L(A) and F228 * L(B). ␣ 1 , undigested ␣HL; p, proteinase K fragments generated by cleavage near the N terminus (2, 3). Therefore, His-35 and all three clusters may be in close proximity in the prepore (Structure 3), either within individual subunits or at intersubunit contact sites, or they may be brought into proximity during formation of the active pore (Structure 4). Accordingly, the cysteine in H35C becomes unreactive toward a water-soluble sulfhydryl reagent during formation of the prepore (12). Proximity of the residues in question would lend a ready explanation for the revertants as beneficiaries of compensating mutations that through direct interaction repair a defect in the final step of assembly. However, it is by no means certain that the restoration of activity is the outcome of such proximity. For example, while second-site revertants of a defective triose phosphate isomerase were clustered near the primary mutation at the active site (24), reversion of other mutant proteins such as staphylococcal nuclease (23) and phage lambda repressor (22) can be brought about by amino acid substitutions distant from the primary site. Therefore, it will be most interesting to examine the placement of His-35, the three clusters and the two lone mutations at amino acid positions 84 and 169 in a three-dimensional structure of ␣HL.