Biochemical and Genetic Evidence for Three Transmembrane Domains in the Class I Holin, λ S*

λ S, the prototype class I holin gene, encodes three potential transmembrane domains in its 107 codons, whereas 21 S, the class II prototype spans only 71 codons and encodes two transmembrane domains. Many holin genes, including λS and 21 S, have the “dual-start” regulatory motif at the N terminus, suggesting that class I and II holins have the same topology. The primary structure of 21 S strongly suggests a bitopic “helical-hairpin” topology, with N and C termini on the cytoplasmic side of the membrane. However, λ S chimeras with an N-terminal signal sequence show Lep-dependent function, indicating that the N-terminal domain of S requires export. Here the signal sequence chimera is shown to be sensitive to the missense change A52V, which blocks normal S function. Moreover, cysteine-modification studies in isolated membranes using a collection of S variants with single-cysteine substitutions show that the positions in the core of the 3 putative transmembrane domains of λ S are protected. Also, S proteins with single-cysteine substitutions in the predicted cytoplasmic and periplasmic loops are more efficiently labeled in inverted membrane vesicles and whole cells, respectively. These data constitute direct evidence that the holin Sλ has three transmembrane domains and indicate that class I and class II holins have different topologies, despite regulatory and functional homology.

Host lysis for double-stranded DNA bacteriophages involves active degradation of the host peptidoglycan by enzymes designated as endolysins, or phage-encoded muralytic activities (1). Diverse enzymatic activities fulfill this role in different phages but a common feature is that, despite the fact that their substrate is outside the cytoplasmic membrane, endolysins have neither a secretory signal sequence nor a specialized secretory system. Instead, endolysins require the function of a membrane protein called a holin to exit the cytosol and gain access to the cell wall. Active endolysin accumulates in the cytosol during the vegetative cycle until, at a genetically programmed time, the holin proteins somehow permeabilize the membrane and allow escape of the endolysin. Holins thus control the timing of host lysis and the progeny yield of the infective cycle. Nothing is known about the nature of the "hole," the membrane lesion produced by the holin, except that it is apparently nonspecific in that different holins can function with different endolysins.
Many holin genes have been identified and most, by inspection of their predicted primary structure, can be grouped into two classes. Class I holins, the prototype for which is S, are typically 90 to 125 amino acid residues in length and have three hydrophobic domains which could serve as integral transmembrane helices (TM) 1 (Fig. 1A) (2,3). In most cases of class I holins, all three potential TMs have residues with charged or hydrophilic side chains. Class II holins, the founding member of which is the phage 21 S 21 protein (4), are typically 65 to 85 residues in length and have two hydrophobic TMs, generally lacking in charged or hydrophilic residues (Fig. 1A). A feature common to both classes of holins is the highly charged Cterminal domain, generally rich in basic residues, which has been shown to be non-essential in S (5). Class I and II holin genes complement and appear to be wholly interchangeable, irrespective of the nature of the endolysin (2).
Holin function terminates the infective cycle and thus it is not surprising that it is subject to sophisticated regulation, which has been extensively studied for the holin. Lysis normally occurs 50 min after induction of a lysogen, but missense mutants of S are known which exhibit lysis times from 19 min to greater than 2 h (6, 7). Moreover, lysis can be triggered prematurely by addition of an energy poison, like cyanide, to the medium (8). Deletion of the C-terminal charged domain of S causes more rapid onset of lysis, whereas replacement of the normal sequence with a heterologous sequence with more positive charge retards lysis (5). A remarkable feature of the S gene, and other holin genes, is the "dual start motif"; i.e. two translational start codons, at positions 1 and 3 (Fig. 1A), which give rise to two products, called S107 and S105 (9 -11), with opposing functions. That is, S105 acts as the lethal holin, while S107 acts as an inhibitor of S105 (1,9,11,12). Within certain ranges, the proportion of S107 and S105, which is determined by an RNA stem-loop structure overlapping the translational initiation region (9 -11) and which is normally 1:2, is an important determinant of the lysis "clock" (13) (Fig. 1B). The operative difference between S107 and S105 is the positively charged residue at position 2; replacement of this residue with a neutral or acidic residue ablates inhibitor function and converts S107 into a lethal holin (11). Structural information is limited and restricted to S, which in the detergent octyl glucoside has a CD spectrum consistent with approximately 40% ␣-helical structure (14). The topology of S is as yet undefined. The C terminus has been shown to be cytoplasmic, by the criterion of protease sensitivity and resistance in inverted membrane vesicles (IMV) and spheroplasts, respectively (5). CD analysis of detergent-solubilized S indicates that about 40 -45 residues are in helical conformation, suggesting two TM helices, assuming the detergent environment preserves the secondary structure (14). The distal hydrophobic domain (residues 65-83) has a proportion of hydroxylated residues which would be unusual for a TM domain. These data suggested a model with hydrophobic domains 1 and 2 acting as TM domains (Fig. 1C, model  3). However, gene fusion studies in which a secretory signal sequence was fused to S indicated that the N terminus must be externalized for lysis to occur, suggesting a topology with 3 TM domains (Fig. 1C, model 5) (15). In contrast, the topology of S 21 is to be evident from inspection of the primary structure, with the two uniformly hydrophobic domains of the right length for TM helices and the C terminus rich in basic residues, strongly suggesting a "C-inside, N-inside" orientation ( Fig. 1C, model 4).
Recently, it has been demonstrated that S 21 has a dual start motif functionally homologous to S, giving rise to two S 21 protein products of 71 (S 21 71 ) and 68 (S 21 68 ) amino acids that function as inhibitor and effector, respectively (16) (Fig. 1, A  and B). Phylogenetic analysis shows that many holin genes, of both classes, appear to have a dual start motif. Thus the simplest expectation is that the dual start motif in class I and II holins should function from the same side of the membrane, a notion which would support models with two TM domains for S ( Fig. 1, model 3).
Here we report a cysteine-scanning analysis of the S gene, which has only a single-Cys codon in the parental allele (Fig.  1A). The properties of the mutant alleles are examined and the accessibility of the Cys residues of the membrane-embedded protein products is determined by chemical modification. The results allow an unambiguous determination of the membrane topology of S, and by extension, all class I holins.
Standard DNA Manipulation, Polymerase Chain Reaction, Site-directed Mutagenesis, and DNA Sequencing-These methods have been described previously (14,18). Site-directed mutagenesis was performed using pairs of synthetic oligonucleotides purchased from the Gene Technology Laboratory of the Department of Biology, Texas A&M University. In each case, 17-30 nucleotides of the parental sequence flanked the 1-3 nucleotides mutated sequence. All constructs were checked by automated fluorescent sequencing (14). For the construction of the A52V alleles of the VIIIS fusions, the pL vector plasmids pVIII-S105 or pVIII-S107 (15) were used as templates, and the products were designated pVIII-S105 A52V and pVIII-S107 A52V , respectively. For all other mutageneses, the parental plasmid was pS105. This plasmid carries the S M1L allele, which produces only the S105 translational product (14), as part of the SRRzRz1 lysis cassette, under the control of the pЈR promoter. This plasmid and its derivatives are carried in a lysogenic host, MC4100[⌬(SR)], and expression is obtained by thermal induction of the prophage, which supplies the late gene activator Q to transactivate the pЈR promoter on the plasmid (14). Most of the plasmids used carry a cysteine-less S allele (S C51S ) and ⌬R. The latter deletion spans the RRzRz1 gene cluster and its construction is described elsewhere. 2 Twenty-two single-cysteine alleles were created by altering individual codons of the cysteine-less S gene in pS105 C51S ⌬R, another was created by appending a Cys residue to the C terminus of the S105 reading frame in the same plasmid, and a further 18 singlecysteine alleles were created in the S C51S,A52V context.
Induction of VIIIS Fusion Genes-The induction has been described (15). Briefly, the plasmids carrying the VIIIS chimeras were transformed into pop2135, the transformants grown in LB-ampicillin at 30°C to maintain repression of the pL promoter of the plasmids, and induced by shifting to 42°C and continuing aeration. Culture mass was monitored as A 550 in a Gilford Stasar II sipping spectrophotometer. To check for endolysin expression, 1% chloroform was added to a portion of the induced culture and the optical density monitored for lysis.
Membrane Sample Preparation and Analysis of VIIIS Fusion Protein-Detergent-solubilized preparations of inner membrane proteins were obtained as described previously (13,18). Proteins were separated on a 10 -20% Tris glycine precast gradient gel (Xcell II Minicell; Novex, San Diego, CA) following the manufacturer's instructions.
Preparation of IMV and Whole Cells for IASD Labeling-For the IASD labeling reaction, the lysogen MC4100[⌬(SR)] carrying a pS105⌬R plasmid with a single-cysteine allele of S was aerated at 30°C until A 550 ϭ 0.4. After aeration for 15 min at 42°C, the cultures were aerated another 35 to 45 min at 37°C and then immediately placed on ice. 20 ml of induced culture were collected by centrifugation at 1,100 ϫ g for 10 min at 4°C. The cell pellet was resuspended in 1 ml of TBS buffer (25 mM Tris, 150 mM NaCl, pH 7.2). Cells were disrupted by a single passage through an SLM-Aminco French pressure cell (Spectronic Instruments; Rochester, NY) at 16,000 p.s.i. to produce IMV. Unbroken cells and large cell debris were removed by centrifugation at 5,000 ϫ g for 5 min at room temperature. This procedure results in at least 95% inverted membranes, as judged by the protease accessibility of the cytosolic C terminus of S in vesicles prepared in the same way (5). For the preparation of whole cells, 20 ml of induced culture were collected as described above. The cell pellet was resuspended in TBS at room temperature and treated with 10 mM EDTA for 15 min to permeabilize the outer membrane prior to the labeling reaction.
Labeling of IMV or Whole Cells with 5 mM IASD-IASD, purchased from Molecular Probes (Eugene, OR), was freshly dissolved in TBS buffer, pH 7.2, to a final concentration of 100 mM. 50 to 200 l of IMV or whole cell samples were labeled with 5 mM IASD for the indicated time in the dark. The reactions were stopped with 0.15 M cysteine and incubation for 15 min at room temperature. A stock solution of 1 M cysteine was prepared just prior to use in TBS. As a negative control for labeling, the cysteine was added to one sample before addition of IASD. IMV were collected by ultracentrifugation at 100,000 ϫ g for 1 h at 18°C. Whole cells were disrupted by sonication in a cup horn attachment of a sonicator model W-375 (Heat Systems, Ultrasonics, Plainview, NY). Cells were sonicated 6 ϫ 30 s in the continuous mode and kept on ice for at least 1 min between the cycles. Following sonication, the cell debris were collected by ultracentrifugation. Membrane pellets were solubilized in 50 l of membrane extraction buffer (1% Triton X-100, 10% glycerol, 0.5 M NaCl, 35 mM MgCl 2 , 20 mM Tris-HCl, pH 8.0) supplemented with 0.2 M cysteine for 6 to 8 h at 37°C with shaking (19,20). Afterward detergent-insoluble material was removed by centrifugation at 100,000 ϫ g for 45 min at 18°C to produce a detergentsolubilized preparation of inner membrane proteins. For SDS-PAGE electrophoresis the detergent soluble fraction were diluted 1:1 with 2ϫ protein sample buffer containing 2.8 M ␤-mercaptoethanol (13). Protein samples were boiled for 5 min and centrifuged at 14,000 ϫ g for 5 min at room temperature just before loading on an SDS-PAGE gel.
Labeling under Denaturing Conditions in 2% SDS with 15 mM IASD at 100°C-IMV were prepared as described above, with the exception that the cell pellet of the induced culture was resuspended in 1 ml of TBS supplemented with 0.5 mM DTT. A 100 mM stock solution of DTT was prepared in TBS just prior to use. 50 to 200 l of IMV were collected by ultracentrifugation at 100,000 ϫ g for 1 h at 18°C. The membrane pellet was solubilized in 50 l of 2% SDS, 0.5 mM DTT, TBS buffer. Membrane proteins were extracted for 6 to 8 h at 37°C with shaking. Insoluble material was removed by ultracentrifugation. Shortly before the labeling reaction, 0.5 mM DTT was added and the samples were kept for 10 min at room temperature. For the labeling reaction under denaturing conditions, 15 mM IASD was added from a 100 mM stock solution dissolved in TBS buffer and samples were incubated for 15 min at 100°C. The reactions were stopped with 0.4 M cysteine from a freshly prepared 1 M stock solution in TBS buffer and incubated for 15 min at room temperature. Protein samples were mixed with the same volume of 2ϫ protein sample buffer containing 2.8 M ␤-mercaptoethanol. The samples were boiled for 5 min at 100°C, centrifuged for 5 min at 14,000 ϫ g at room temperature, and loaded on the SDS gel.
Labeling of IMV with 15 mM IASD in High Salt Conditions-20 ml of induced culture were resuspended in 1 ml of TBSHS (0.75 mM NaCl, 25 mM Tris, pH 8.0), supplemented with 1 mM DTT. The labeling reaction was performed for 60 min at room temperature with 15 mM IASD, dissolved as a 100 mM stock solution in TBS8 (TBS buffer, pH 8.0). Reactions were stopped with 0.4 M cysteine and inner membrane protein extracts were prepared for SDS-PAGE and Western blot analysis as described above.

SDS-PAGE, Western
Blotting, and Immunodetection-For immunodetection of S protein, protein samples and prestained molecular mass standards (Life Technologies, Inc., Gaithersberg, MD) were resolved on 16% Tris/Tricine gels. Tris/Tricine 16% SDS-PAGE was performed according to the method of Schä gger and Jagow (21). Proteins were then transferred to a 0.1-m nitrocellulose membrane (Schleicher & Schuell) with a semi-dry blotting apparatus from Enprotech (Natick, MA) for 2 to 3 h at 100 mA. Western blots were developed as described by Towbin et al. (22). An antibody raised in rabbits against a synthetic C-terminal peptide from the S sequence (13) was used as primary antibody at a dilution of 1:1,000. The secondary antibody (goat anti-rabbit immunoglobulin G conjugated to horseradish peroxidase) was purchased from Pierce and used at a dilution of 1:1,000. Blots were developed according to the manufacturer's instructions. Images of blots were digitized with a Hewlett-Packard ScanJet IIcx scanner (Palo Alto CA). Quantitative analysis of immunoblots was performed using the image analysis program NIH image (version 1.54) public domain software suit (W. Rasband, National Institutes of Health). The labeling efficiency of the single cysteines was calculated with the formula, Intensity of modified band (Intensity of modified band ϩ intensity of unmodified band) (Eq. 1)

RESULTS
Signal Sequence-dependent Lysis Abolished by Holin Defect-A common feature of the inhibitor and putative inhibitor products of holin genes with dual start motifs is that there is at least one positively charged residue in the N-terminal extension (Fig. 1B). The ability of the S107 product of S to exert its inhibitory function is abolished by energy poisons (11). Moreover, the inhibitor form alone can be triggered to cause lysis by cyanide, unless two or more positively charged residues are present at the N terminus (11,23). These observations suggested that the N terminus of both S products might require transit of the membrane, after more distal segments of the protein have integrated into the bilayer (Fig. 1C, model 1), and that this transit is sensitive to the charge at the amino terminus, as demonstrated for the N-terminal domain of Lep (24 -26). According to this view, the final topology of both forms of S would be N-outside, C-inside, with three TM domains (Fig. 1C,  model 2). Strong support for this notion was supplied by the results of Graschopf and Blä si (15), who fused the M13 gene VIII signal sequence to both inhibitor and effector forms of S and demonstrated that lysis was obtained concomitant with and dependent on signal-sequence cleavage. However, another interpretation of this result is that the high level of VIII-S fusion production, resulting from the combination of the pL promoter and the very efficient gene VIII ribosome-binding site, creates a nonspecific insult to the membrane, unrelated to the normal S permeabilization function, which has been shown to occur when there are only about 10 3 S proteins per cell (18). This is an important distinction, because, if the signal-sequence S chimera actually functions along the same pathway as the normal S holin, it would be very strong evidence that the N terminus of S, and thus other class I holins, must be exported to the periplasm. To distinguish between these interpretations, we altered the S sequence in the VIIIS fusion gene to contain the A52V missense change. This change confers an absolute defect in lysis to S without altering the ability of the S proteins to insert in the membrane (3,6), and thus, if the same change conferred a lysis defect on the VIII-S chimera, it would be a strong indication that the chimera depends on an S-specific lysis pathway instead of some nonspecific membrane toxicity. Fig. 2A shows that this alteration abolishes the lytic function of the fusion gene. Western blots of membrane samples taken after induction show clearly that the mutant chimera accumulates in the membrane (Fig. 2B), demonstrating that the lysis defect is a functional defect rather than a failure of localization or loss of stability. Moreover, the lysis supported by parental fusion is extremely saltatory (Fig. 2A), which is characteristic of holin function (15). Together, these observations indicate that the lysis supported by the VIIIS fusion gene is fundamentally the same as that supported by S, and thus strongly support the idea that S imbeds in the membrane with three TM domains in an N-out, C-in topology (Fig. 1C, model 2).
Chemical Probing of S Topology-The parental S sequence has a single-Cys codon at position 51 (Fig. 1A). The C51S allele is fully lytic, and a large collection of alleles with single-Cys substitutions have been created, all of which generate stable, membrane-inserted proteins and most of which retain lytic capacity. 2 To probe the topology of S in the membrane, each of these alleles was expressed in bacterial cells and IMV were prepared from each induced culture. These samples were treated with IASD, a cysteine-specific modification reagent, and the extent of modification of the different single-cysteine S proteins assessed by monitoring SDS-PAGE mobility (Fig. 3). Three hydrophobic regions of the S sequence were resistant to modification, including residues 23-26, 49 -54, and 74 -81. These clusters of resistant regions map to the core of predicted TM domains in the 3 TM model and would define 1, 1.5, or 2 turns of ␣-helix, respectively (Fig. 1A). The chemical reactivity of these sites was confirmed by treating the samples with boiling SDS to solubilize the S protein (Fig. 4). These results strongly suggest that S has three membrane-imbedded domains centered on these protected regions. The 3 TM model would predict that there would be 4 aqueous domains (N terminus and loop 2 in the periplasm, loop 1 and C terminus in the cytoplasm). Given the membrane-impermeant nature of IASD, the simplest expectation would be that the N terminus and loop 2 would be sensitive in spheroplasts and resistant in IMV, whereas the C terminus and loop 1 would show a complementary pattern. However, multiple sites in the FIG. 2. The A52V substitution blocks lysis caused by the VIII-S chimera but not its accumulation in the membranes. A, pop2135 cells harboring plasmids with VIIIS fusion genes were thermally induced at time 0 and monitored for culture mass as A 550 . Plasmids: pVIII-S105, induced (q) or uninduced (OE); pVIII-S105 A52V , induced (E), pVIII-S107, induced (f), pVIII-S107 A52V , induced (Ⅺ). To check for endolysin production, CHCl 3 was added (vertical arrow) to a portion of the induced pVIII-S105 A52V and pVIII-S107 A52V cultures (dashed lines). B, Western blot analysis of the accumulation of the hybrid protein in the membrane. Inner membrane protein samples were prepared from induced and uninduced cultures. Samples of induced cultures were taken after cell lysis was completed, or, for cultures which did not undergo lysis, at 20 min after induction. Samples from the uninduced cultures were taken at 23 min after induction. Lane 1, prestained molecular weight marker; lanes 2-5, uninduced cultures; lanes 6 -9, induced cultures; cells carrying pVIII-S105 (lanes 2 and 6), pVIII-S107 (lanes 3 and 7), pVIII-S105 A52V (lanes 4 and 8), and pVIII-S107 A52V (lanes 5 and 9). Arrows indicate positions of S monomers and dimers. Numbers to the left of the panel are the sizes of molecular mass standards in kDa. N terminus and both putative inter-TM loops were sensitive in IMV (Fig. 3). Previous studies with these IMV preparations showed that at least 95% of the vesicles are inverted, as judged by the accessibility of the cytosolic C terminus of S to proteases (5). Thus, the simplest explanation for the labeling on both sides of the membrane is that the "hole-forming" ability of S makes the membrane permeable to IASD. In this case, the requirement for the reagent to pass through the S-mediated lesions might impose a kinetic difference in the modification reaction, as observed for cis and trans sites within the poreforming toxin of Saccharomyces aureus (27). Fig. 5 shows that, as predicted, sites in the N-terminal domain and in loop 2 are more efficiently modified in spheroplasts, whereas a site in loop 1 is better labeled in IMV, again supporting a 3 TM model with N-out and C-in (Fig. 1C). Interestingly, the ability to label on either side of the membrane, albeit with different efficiencies, persists even with the inclusion of the A52V mutation. This indicates, that even a non-lytic S protein confers permeability defects, at least for the IASD reagent.
The most unexpected result was that several sites in the C-terminal domain, which has been demonstrated to reside in the cytoplasmic compartment by protease-accessibility studies (5), were resistant to modification in vesicles (Fig. 3). The C-terminal region is rich in charged residues, especially basic residues, and is non-essential for S function (5,28). Moreover, replacing the C-terminal domain with unrelated sequences of comparable size but increased positive charge results in a dominant-negative phenotype (5,28). We reasoned that these properties and the IASD resistance of the C-terminal sites in vesicles might reflect an intimate association of the C-terminal domain with the negatively charged inner surface of the cytoplasmic membrane. To test this idea, IASD modification was repeated on membrane vesicles in the presence of high salt to disrupt the putative electrostatic interactions. Under the high salt conditions, the C-terminal sites (positions 94 and 95), but not a site in the core of the putative TM3 (position 78), are quantitatively modified in IMV (Fig. 6), strongly supporting the idea that the C terminus is located on the cytoplasmic side of the bilayer and probably intimately associated with the membrane surface by virtue of ionic interactions.
An Oligomerization Defect Has No Effect on the Pattern of IASD Modification-S has been shown to oligomerize in the membrane (3,29,30). A possible alternative interpretation of the resistance of the three hydrophobic domains to IASD modification is that intimate protein-protein contacts, rather than the impermeant core of the bilayer, block access of the reagent to some of the sites. Considering that the resistance spans one or more entire helical turns of the putative TM domains and the fact that S is such a small protein, such extensive proteinprotein contacts would require an oligomeric structure. As noted above, the A52V substitution abolishes S lytic function and cross-linking studies have demonstrated that this defect is associated with a failure to oligomerize. 3 Fig. 7 shows that the pattern of IASD modification is unaffected by the A52V substitution. This indicates that the patterns of resistance within the hydrophobic domains reflect sequestration in the core of the bilayer rather than within an oligomeric assemblage, and that there is no gross alteration of topology concomitant with the lytic function of S.

DISCUSSION
The Topology of Class I Holins-The topology of S, the prototype class I holin, is ambiguous from examination of the primary structure. Although there are three domains with net neutral charge, only the putative TM1 and TM2 domains have a preponderance of hydrophobic residues, and TM3 is unusually rich in hydroxylated amino acids for a TM domain. Clusters of mutations with lysis-defective phenotypes involving changes in the size of side chains have been found in both TM1 and TM2, but not TM3. CD spectroscopy of purified, detergentsolubilized S supports a 2 TM model (14). Moreover, elementary primary structure constraints (i.e. approximately 20 amino acid residues per TM helix) make it highly likely that S 21 , the prototype class II holin, has only two TM domains (Fig.  1, A and C). If for no other reason than intellectual economy, the presence of functionally homologous dual start motifs in both class I and class II holins (2,12,16) would suggest that the N-terminal domain of both classes of holins are on the same side of the membrane; i.e. the cytoplasmic side. However, the results presented here constitute convincing evidence for a model in which the S holin of phage is integrated in the membrane with 3 TM domains, with N-out and C-in (Fig. 1C,  model 2).
First, we showed that the lytic capability of a gene fusion between the signal sequence domain of M13 VIII and S is sensitive to the same missense change, A52V, which blocks lysis in the native holin context. Given the conservative nature of this substitution and the specific molecular defect associated with it in the S context (i.e. failure to oligomerize 3 ) this strongly suggests that the presence of a cleavable signal sequence does not direct S along a fundamentally different pathway to lysis. Taken with the finding of Graschopf and Blä si (15) that the presence of the signal sequence bypasses the block associated with the presence of positive charge at the N terminus and with the requirement for concomitant signal sequence cleavage, these results strongly indicate that the N-out, 3 TM topology for S must be considered to be at least an active form of the holin at the end of the pathway which leads to "hole-formation." These findings, while persuasive, do not rule out the possibility that a complete re-orientation of the S protein domain occurs after cleavage of the signal sequence in the chimeric protein, allowing S to assume a different topology. For example, after cleavage, the N terminus could in principle slip back 3 A. Grü ndling and R. Young, manuscript in preparation.
FIG. 4. All single-Cys S proteins can be quantitatively labeled in denaturing conditions. Extracts of inner membrane samples containing S proteins were prepared in 2% SDS and treated with 15 mM IASD for 15 min at 100°C. After stopping the reactions with 0.4 M cysteine, the samples were analyzed by Western blot. Lanes are labeled as described in the legend to Fig. 3. In each case, the ϩ sample is flanked by Ϫ samples to ensure that mobility differences can be detected. out of the bilayer and thus allow S to re-insert with a 2-TM, N-in, C-in topology (Fig. 1C, models 1 and 3). The cysteine modification studies allow us to rule out this possibility and provide convincing evidence that the central region of three potential TM domains are imbedded in the bilayer. TM domains in the cytoplasmic membrane are always found to be ␣-helical (31). In view of the presence of a strong helical signal, and the absence of ␤-sheet signal, in the CD spectrum of purified S (14), one would expect each of the three regions of IASD insensitivity to extend approximately 20 residues. Using similar hydrophilic cysteine modification reagents to probe other oligotopic membrane proteins, other workers have found that only a central core of about 10 residues is protected in TM helices (32,33). Given the asymmetry of these reagents, which, After preparation of inner membrane protein extracts, cysteine modification was analyzed as described in the legend to Fig. 3. Lanes are labeled as in Fig. 3. Note that the mobility produced by modification of positions 94 and 95 is significantly smaller than with other positions, as shown in Fig. 4. As in Fig. 4, each ϩ lane is flanked by Ϫ lanes for ease in detecting mobility shifts.
FIG. 7. The A52V substitution has no effect on IASD labeling of S proteins in IMV. A, IMV containing S A52V proteins with single-Cys substitutions were treated with 5 mM IASD for 60 min and analyzed by SDS-PAGE and Western blot. Modification of each single-Cys S protein was seen as reduced SDS-PAGE mobility. Lanes are labeled as described in the legend to Fig. 3. B, graphical representation of labeling efficiency in the S A52V context. Position of each of the 18 single-Cys substitutions within the S sequence is indicated on the x axis, as is the position of the A52V substitution.
in the case of IASD, results in a 7.5-Å separation between the charged groups conferring the hydrophilicity and the reactive group, it is likely that the reactive moieties are able to enter the bilayer to a considerable extent. Alternatively, membrane proteins, and especially S, may destabilize the bilayer locally and allow partial penetration of the reagents. In any case, it is clear that this type of methodology is capable of establishing the existence of a TM domain by mapping out the protected central core, but it is not capable of determining the boundaries of the TM and aqueous domains. For the purposes of this study, however, the existence of three protected cores clearly indicates the existence of three TM domains for this model class I holin.
It is not clear what the nature of the S hole is, but whatever it is, it was not unexpected that domains predicted to be in either aqueous compartment would be labeled by IASD treatment of membrane vesicles, because most of the single-cysteine mutants are functional as lytic proteins and thus would be expected to allow the reagent through the membrane. The difference in rate and efficiency of labeling of putative N-terminal, loop 1, and loop 2 sites in spheroplasts and inverted vesicles is consistent with this interpretation (Fig. 5). Nevertheless, in one case where the cysteine substitution abolishes S lytic function, the mutant with the S68C and A52V substitutions, the single cysteine is labeled in both spheroplasts and inverted vesicles (Fig. 5), albeit more slowly in the latter. This is difficult to rationalize, especially considering the fact that expression of these single-Cys S alleles is at the normal in vivo levels. However, it should be noted that preparation of vesicle and spheroplast samples requires time intervals which are relatively long compared with the normal kinetics of lysis after induction of a thermo-sensitive lysogen. Alleles which are defective in supporting the saltatory lysis characteristic of S might still yield S products which would damage the integrity of the membrane on these long time scales. It is also possible that S protein, even if blocked in the steps leading to release of the endolysin to the periplasm, may still have a significant destabilizing effect on the membrane, especially after the manipulations involving severe physical stress which are required to form IMV.
Implications for the Pathway of Holin-mediated Lysis and Its Regulation-A fundamental characteristic of holin function is that it is distinctly saltatory; that is, holins accumulate in the membrane throughout the late protein synthesis phase of the vegetative cycle, without apparent affect on host physiology, until suddenly, at a programmed time, holin function triggers, the membrane is permeabilized, host respiration stops, and the cytosolic endolysin gains access to the peptidoglycan (1). The fact that holin function can be triggered prematurely by the addition of energy poisons suggests that the energized membrane somehow opposes spontaneous triggering. A key issue in understanding how holin action is regulated is the ability of S107 to act as an inhibitor of S105 (12). The fact that S105 has a 3-TM, N-out, C-in membrane topology, and the success of Graschopf and Blä si (15) in demonstrating that the presence of a signal sequence largely dissipates the difference between S105 and S107, clearly evokes the perspective that triggering may reflect the penetration of the N terminus of S across the membrane. This process may be blocked or severely inhibited for S107, with its basic N terminus, as observed for other membrane proteins which require externalization of an Nterminal domain (24,26,34). It should be noted that the membrane topology experiments reported here all deal with single-Cys derivatives of S105. Moreover, it is not likely that this type of approach would be useful for determining whether S107 has 2 TM domains until triggering, because, for example, the manipulations of the membranes required to prepare inner mem-brane vesicles are incompatible with maintaining a fully energized membrane. This caveat also applies to interpreting our finding that the A52V change, which blocks the lethal action of S, does not alter the 3 TM topology of S. The S A52V proteins could be fully triggered during the preparation of the membrane samples, so there are no data about the topology of the mutant S proteins before triggering.
The data are not relevant to issues of whether tertiary or quaternary changes occur at the instant of triggering, however. Moreover, the salt-dependent sensitivity of the C-terminal domain to the cysteine modification reagent in inverted vesicles is consistent with the idea that the last 16 residues of S, while non-essential for lytic function, serve as a regulatory domain. According to this hypothesis, the positively charged residues of the C terminus interact with the anionic surface of the inner membrane and this ionic interaction somehow inhibits the conformational change required for hole-formation. This notion is buttressed by the fact that scrambled C-terminal sequences of comparable length but increased positive charge can cause severe retardation of S function, implying that the interaction does not rely on a particular protein structure. However, a previous suggestion, that positive charge at both N and C termini basically function in the same way (by interacting with the inner surface of the cytoplasmic membrane (12)) may not be correct, given the evidence that N terminus resides in the periplasm for S. In any case, the regulation of 21 S, the prototype class II holin, which almost certainly has N-in, C-in topology, may be only superficially similar to that of S. Although S 21 also has an N-terminal dual start motif with analogous function (16), the predicted topology of its protein products would require that the positively charged residue on the inhibitor form, S 21 71 , function on the cytoplasmic side of the membrane (Fig. 1C). Recently, in preliminary experiments, it has also been shown, that C-terminal truncations of S 21 lacking the positively charged hydrophilic domain are lysis-defective, 4 in contrast to the situation with S, where such truncations cause an acceleration of lysis triggering (5). Thus, there appear to be unexpectedly fundamental differences between the class I and class II holins despite their genetic interchangeability.