Report

Disulfide Isomerization After Membrane Release of Its SAR Domain Activates P1 Lysozyme

See allHide authors and affiliations

Science  07 Jan 2005:
Vol. 307, Issue 5706, pp. 113-117
DOI: 10.1126/science.1105143

Abstract

The P1 lysozyme Lyz is secreted to the periplasm of Escherichia coli and accumulates in an inactive membrane-tethered form. Genetic and biochemical experiments show that, when released from the bilayer, Lyz is activated by an intramolecular thiol-disulfide isomerization, which requires a cysteine in its N-terminal SAR (signal-arrest-release) domain. Crystal structures confirm the alternative disulfide linkages in the two forms of Lyz and reveal dramatic conformational differences in the catalytic domain. Thus, the exported P1 endolysin is kept inactive by three levels of control—topological, conformational, and covalent—until its release from the membrane is triggered by the P1 holin.

In phage infections, the timing of host lysis is strictly regulated (1, 2). For double-stranded DNA phages, lysis requires degradation of the cell wall by a phage-encoded endolysin, or lysozyme. In the classic systems of phages λ and T4, premature cell wall degradation is prevented by keeping the bacteriolytic enzyme sequestered in the cytoplasm (3). At a genetically programmed time, the holin, a small phage-encoded membrane protein, disrupts the membrane, which allows access of the lysozyme to the cell wall; lysis follows within seconds (1, 4). However, endolysins are not always dependent on holins for export (57). The endolysins from enterobacteriophage P1 (Lyz) and lambdoid coliphage 21 (R21), although orthologs of T4 lysozyme (8, 9), are exported by the host sec system by virtue of an N-terminal transmembrane domain (TMD) (Fig. 1) that functions as a type II signal anchor, or uncleaved signal peptide (7, 10). This leaves the secreted protein in the periplasm but tethered to the membrane in an inactive form. When the membrane is depolarized by the holin, this N-terminal TMD exits the bilayer, resulting in activation of the endolysin and cell lysis. Because of the ability to escape from the membrane, the Lyz and R21 TMDs, as well as similar motifs in many other phage lysozymes, including the well-studied coliphages Mu (7) and T1 (11), have been designated as SAR (“signal arrest and release”) domains. The capacity for membrane release probably derives from a high content of the small hydrophobic and polar residues Gly, Ala, Ser, and Thr, but the mechanism for extraction from the bilayer is unknown.

Fig. 1.

SAR endolysins. P1 Lyz and R21, the endolysin from lambdoid phage 21, are aligned with the homologous soluble endolysins T4 gpe and P22 gp19 (15). Residue numbering (italics) is for P1 Lyz. Coloring: purple, Cys; green, catalytic Glu, Thr; blue, catalytic Asp; yellow: predicted SAR TMDs. The positions of Cys substitutions made in the C13S allele are indicated; purple represents changes that suppressed the lysis defect. Below the Cys substitutions, the R21ΦLyz chimera is shown. Positions where a Cys substitution suppresses (Gly13) or fails to suppress (Ala16) the lysis defect of the chimera are underlined and underlined italic, respectively. The signal sequence of PhoA is shown in blue at the position where it replaces the N terminus of P1 Lyz, in the PhoAΦLyz chimera. The G2H6G2 shaded in gray in LyzHis replaces residues 124 to 134 (underlined) of Lyz.

It was also not understood how the tethered SAR endolysins are kept inactive until holin triggering. To address this issue, we aligned Lyz with T4 lysozyme, gpe (Fig. 1). Members of the T4 lysozyme family have an essential catalytic triad corresponding to Glu11, Asp20, and Thr26 of gpe (12). Although Lyz has two of these conserved catalytic residues, Glu42 and Thr57, the alignment suggests that Cys51 in Lyz fulfills the catalytic role of Asp20 in T4 lysozyme. Consistent with this, the gpeD20C (13) variant is functional (14). The functionality of LyzC51D and the sensitivity of Lyz, but not LyzC51D, to thiol reagents confirmed the catalytic role of Cys51 (Fig. 2A and table S1). In addition to the catalytic Cys51, Lyz has six other cysteine residues, one of which, Cys13, is located in the N-terminal SAR domain (Fig. 1) and is expected to be embedded in the bilayer in its sulfhydryl form. To test this, chemical cleavage was used to determine the positions of reduced cysteines in Lyz present in induced cultures before and after the onset of lysis (15). Reduced cysteines, but not cysteines in disulfide bonds, can be cyanylated, and the cyanylated cysteines can be sites of partial cleavage when treated with ammonium hydroxide (16). In samples taken soon after induction, only a single cleavage product was detected, with an apparent mass expected from cleavage at Cys13 (Fig. 3A). This indicates that, in the membrane-associated, inactive form, Cys13 is present in its sulfhydryl form, but all of the six periplasmic cysteines, including the catalytic Cys51, are in disulfides. In samples taken after the onset of lysis, two cleavage products were detected, one generated by cleavage only at Cys13, the other by cleavage only at Cys51. Peptides corresponding to cleavage at the other five cysteine residues were never detected. Because none of the cysteines in Lyz are involved in intermolecular disulfide bonds (fig. S1), cleavage at Cys13 and Cys51 in the post lysis sample indicates the presence of two isomeric forms of Lyz differing in the arrangement of intramolecular disulfide bonds. By this interpretation, the membrane-associated form of Lyz would have a single free cysteine, Cys13, in the membrane-embedded SAR domain. By contrast, the soluble and active form of Lyz would have a sulfhydryl at Cys51 whereas all other cysteines, including Cys13, would exist in disulfide linkages.

Fig. 2.

The roles of Cys13, Cys44, and Cys51 in Lyz function. XL1-Blue cells carrying plasmids with the indicated lyz allele were induced with IPTG and monitored for turbidity. All plasmids were isogenic to pJFLyz (lyz+). Residue numbering refers to positions in Lyz unless otherwise indicated. Vertical arrows indicate time of addition of 1 mM DTT. (A) lyz+ (◼), lyzC51S (△), and lyzC51D (◯). (B) lyzC13S (◼); lyzC13S with the addition of DTT (▢); lyzC44S (⚫); and lyzC13,44S (△). (C) R21Φlyz (⚫), R21G13CΦlyz (▢), R21A16CΦlyz (▾), and R21ΦlyzC44S (♦). (Positions here refer to the position in the chimera, not R21; see Fig. 1.) (D) lyz+ (◼) lyzC13S,G9C (▢): lyzC13S,G10C (▾); lyzC13S,A11C (⚫); lyzC13S,A14C (△); and lyzC13S,I15C (◯). (E) lyzC13S,I12C (⚫); lyzC13S,A16C (▢); and lyzC13S,V17C (△). (F) phoAΦlyz (▢); phoAΦlyz, with the addition of DTT (⚫); and phoAΦlyzC51D(◼).

Fig. 3.

(A) The free thiols and subcellular localization of Lyz. Lanes 1 to 4, thiol cleavage. Cells harboring pJFLyzHis were induced and samples were taken from cells carrying pJFLyzHis at 30 min (before lysis) and 80 min (after lysis) after induction, precipitated with trichloroacetic acid, subjected to cyanylation and NH4OH cleavage, and analyzed by SDS–polyacrylamide electrophoresis (SDS-PAGE) (15). Lane 1, untreated Lyz; lanes 2 and 3, cleavage products from 30-min and 80-min samples, respectively; lane 4, same as lane 3 except that the cyanylation was reversed by treatment with DTT before exposure to 1 M NH4OH. Asterisks adjacent to lanes 2 and 3 indicate bands corresponding to cleavages at Cys13 (top) and Cys51 (bottom). Lanes 5 to 10, subcellular fractionations. Total protein (T: lanes 5 and 8), soluble protein (S: lanes 6 and 9), and membrane protein fractions (M: lanes 7, 10) were prepared from induced culture of XL1-Blue cells carrying pPhoAϕLyz (lanes 5 to 7), pLyz (lane 8), or pLyzC13S (lanes 9 to 10), and analyzed by SDS-PAGE and immunoblotting (15). In all cases, the Mr of the major band was identical, except for lanes 5 and 6, where the signal sequence is cleaved from the PhoAΦLyz chimera. (B) Model for intramolecular disulfide bond isomerization after the release of the SAR domain from the membrane. Left, Lyz with its SAR TMD embedded in the fully energized membrane. Cys13 is in the sulfhydryl state, whereas Cys44 and Cys51 are in a disulfide linkage that prevents formation of the active site cleft. Right, Lyz after the SAR domain has been released from the depolarized membrane, undergoes refolding and disulfide bond isomerization, forming the active site (star).

These results suggest a model for Lyz activation involving disulfide bond isomerization with no net change in overall oxidation state and in which a critical feature is the liberation of the Cys13 sulfhydryl from the bilayer (Fig. 3B). Accordingly, although mutants of Lyz in which Cys13 is replaced by Ser or Ala (C13S and C13A, respectively) are still released from the cytoplasmic membrane (Fig. 3A), they are lytically inactive. Moreover, the addition of low concentrations of dithiothreitol (DTT) to cultures expressing the C13S mutant causes lysis (Fig. 2B). Here, DTT probably acts by substituting for Cys13 in the disruption of the inactivating disulfide. The lysis defect of C13S could also be suppressed intragenically by a second change at Cys44, C44S; the C13S,44S double mutant of Lyz was not only lytically active but it caused host lysis with kinetics that were slightly faster than seen with wild-type Lyz (Fig. 2B), presumably because one step in the activation, disulfide bond isomerization, is unnecessary. This implicates Cys44 as the cysteine that forms a disulfide with Cys51 in the inactive, nascent form of the endolysin. The Lyz C44S single mutant is inactive (Fig. 2B), presumably because, lacking Cys44, its normal disulfide partner, Cys51 is involved in incorrect disulfide bonds before and after membrane release.

Further support for this model of Lyz activation is derived from the behavior of the chimera R21ΦLyz, in which the SAR domain of Lyz is replaced by that of R21, the SAR-type lysozyme of phage 21 [(Fig. 1); also, see (7)]. The SAR domain of R21 has no significant sequence similarity to the Lyz SAR and does not contain cysteine, presumably because R21 has the canonical Glu-Asp-Thr catalytic triad and thus lacks a catalytic Cys residue (Fig. 1). The R21 SAR domain supports export and release of the R21ΦLyz chimera (7), but the released protein is inactive. The activity could be restored by the introduction of C44S into the chimera (Fig. 2C) or the addition of reducing agent to the induced culture (17), which indicates that the inhibitory Cys44 to Cys51 disulfide bond is retained when the chimera is released from the membrane. Moreover, introduction of a Cys into the R21 SAR domain at position 13 (but not position 16) also restores the lytic activity of the chimera (Fig. 2C). Hence, on release from the bilayer, a heterologous SAR domain can provide the thiol required for reduction of the Cys44 to Cys51 disulfide bond. The lack of a strict sequence context for the activating thiol was also observed in intragenic suppressors of the Lyz C13S mutant, where substitution of a cysteine at positions 9 to 11, 14, or 15 (but not 12, 16, or 17) resulted in enzymatically active double mutants (Fig. 2, D and E). Although there is some steric constraint in the position of the activating sulfhydryl group, these results suggest that there is little requirement for specific contacts between the SAR domain, after its extraction from the bilayer, and the periplasmic domain of Lyz.

These results also imply that, other than providing an export signal and supplying the thiol necessary for reduction of the inhibitory disulfide bond, the Lyz SAR domain is not essential to the structure of the active endolysin. To test this idea, we replaced the entire N-terminal segment of Lyz, including the SAR domain, with the cleavable signal sequence of a well-characterized secreted protein, alkaline phosphatase (PhoA) (18) (Fig. 1). When this construct was expressed in E. coli, the PhoAΦLyz protein was secreted to the periplasm with its signal peptide removed (Fig. 3A), but lysis required the addition of DTT (Fig. 2F). Note that the C51 replaced by Asp (C51D) allele of phoAFlyz caused host lysis even in the absence of a reductant. Hence, the SAR domain of Lyz is not essential for the function of the mature form of Lyz.

To dissect the structural basis of Lyz activation, we determined the crystal structures of the full-length, soluble, active endolysin, Lyza, and also of the inactive, secreted, and processed form of PhoAΦLyz, Lyzi, which consists of residues 29 to 185. Crystals of Lyza and Lyzi were obtained in different space groups (table S2). In the Lyza structure (Fig. 4A), which was solved using multiwavelength anomalous diffraction (MAD) methods (19) to 1.8 Å resolution, the entire protein is visible except residues 1 to 8. With the exception of the N-terminal SAR domain and two β hairpins protruding from the C-terminal domain (Fig. 4A; fig. S2; SOM text), the structure of Lyza is very similar to T4 gpe lysozyme, composed of two domains connected by a long interdomain helix (residues 74 to 96) (supporting online text). Part of the SAR TMD that has been released from the membrane persists in an α helix (residues 15 to 25), whereas its N-proximal residues are in an extended conformation. The crystal structure confirms that Cys51 of the active form of Lyz is in sulfhydryl form and exposed to solvent, occupying the site corresponding to Asp20 in gpe. As predicted, Cys13 of the SAR domain and Cys44 are in a disulfide linkage, which holds the SAR helix close to the body of the endolysin; also as predicted, there are few specific contacts between the SAR domain and the rest of Lyza.

Fig. 4.

Crystal structures of Lyza and Lyzi: domain organization (A and B) and conformational change (C and D). (A) Structure of Lyza with distinct structural domains. Colors: α helices, red; β strands, yellow; loops, green; disulfides, blue. The N- and C-terminal domains are connected by an interdomain connecting helix, α3. (B) Structure of Lyzi. Colors as in (A). The residues 1 to 28 are truncated. Lyzi has three additional helices (α3 to α5), lacks strands β1 to β4 of Lyza, and has a different N-proximal disulfide (44 to 51 instead of 13 to 44). (C and D) Only the N domain and α3 [colored orange; other colors as in (A and B)] are shown. The catalytic residues Glu42 (slate), Cys51 (blue), and Thr57 (pink) are shown with side chains. (C) The ∼60° bend in the middle of α3 is marked with a circle on Lyzi. The catalytic Cys51 is in disulfide linkage with Cys44. Electron density for the residues 62 to 64 is missing. (D) Creation of the active site by N-terminal domain movements and disulfide isomerization. The ∼60° bend relieved in α3 is marked by a dotted box. The new disulfide (Cys13 to Cys44) and the rearrangement of catalytic residues at the active site cleft are visible. A different view of (C and D) shows changes in α2 due to the 112° rotation at the end of that helix (fig. S6).

The Lyzi structure was solved by molecular replacement using the Lyza structure as a search model and refined to 1.1 Å (Fig. 4B). The two structures are very similar for residues 29 to 39 and the entire C-terminal domain (residues 88 to 185), with a root mean square deviation (rmsd) of 0.65 Å. However, the bulk of the N-terminal domain (residues 40 to 73) and part of the interdomain connecting helix (residues 74 to 85) are very different in the two structures (rmsd for residues 40 to 85 = 7.62 Å) (fig. S3). The Lyzi structure confirms that Cys44 and Cys51 are in a disulfide linkage, which shows clear electron density in the refined structure. This disulfide bond holds the N-terminal domain in a compact, closed conformation in which the entire active site cleft is absent, and the distinct separation of the N- and C-terminal domains is missing (Fig. 4, C and D; fig. S4). In Lyzi, two of the catalytic residues, Glu42 and Thr57, are on the surface of the protein with the side chains facing outwards (Fig. 4, B and C).

On the basis of these structures, it is clear that major movements within the region Gly39 to Asn84 are required for the transition from the inactive to active forms (Fig. 4, C and D; figs. S3 and S5). The transition is punctuated by a rotation of 112° between Gly39 and Asn40 that adds another turn at the end of α2; causes α helices 3, 4, and 5 to unwind; and allows three β strands to form a β sheet (Fig. 4, C and D; fig. S5). Also the apparently strained 60° bend at Asn84 in the interdomain connecting helix, α6, of Lyzi is straightened in Lyza. Importantly, these changes do not just unblock the active site by moving or removal of rigid domains, as in zymogen activation (20). Instead, formation of the catalytic cleft is accomplished by remodeling discrete secondary structure elements of the inactive form (Fig. 4, C and D; figs. S5 and S6; movie S1).

The SAR domain of Lyz has two functions. First, it mediates association of Lyz with the cytoplasmic membrane by acting as a secretory signal and a TMD. Second, when released from the membrane, its cysteine residue participates in an isomerization event necessary for Lyz to assume its active conformation. Although changes in the oxidation state of cysteine residues are known to regulate the activity of numerous proteins (21, 22), the distinguishing feature here is that Lyz activation is redox-independent and proceeds without a change in its overall oxidation state.

It remains to be determined whether host proteins are involved in the membrane release, conformational changes, and disulfide bond isomerization required for the maturation of Lyz. In E. coli, periplasmic disulfide bond isomerases (PDI) DsbC and DsbG function by using free sulfhydryls to attack disulfide bonds in nascent or mis-folded substrate proteins (23). Lyz-mediated lysis requires neither DsbC or DsbG (24), which suggests that, once liberated from the membrane, Cys13 may act in cis analogously to the thiol of a PDI. However, in the structure of the periplasmic domain of inactive Lyz, the Cys44 to Cys51 disulfide linkage is so buried that it would seem to be inaccessible to attack by Cys13. The conformational changes required to expose this disulfide to attack and for the subsequent formation of active Lyz may be facilitated by periplasmic foldases or chaperones. A genetic selection for host mutants insensitive to the lethal lytic effect of Lyz could reveal which, if any, of these factors are involved.

Lyz is activated by relief of topological, covalent, and conformational constraints. Although many other SAR lysozymes appear to share this regulatory scheme, with a Cys residue in the predicted SAR domain and a catalytic Cys equivalent to Lyz Cys51, others, like R21, do not (7) (fig. S7). In R21 and other SAR lysozymes lacking the disulfide bond regulation, the Glu residue of the catalytic triad is much closer to the predicted boundary of the SAR TMD, which suggests that in these lysozymes, membrane proximity may sterically block the active site. Moreover, in R21 there are 10 fewer residues between the putative membrane boundary and the active site, which suggests that the SAR domain, once extracted from the membrane, may have to provide much of or the entire α2 element in the active form and thus may interact more intimately with the rest of the catalytic domain. In any case, the redundancy and diversity of the regulatory strategies for the SAR lysozymes reflect how critical lysis timing is in the phage infection cycle.

Supporting Online Material

www.sciencemag.org/cgi/content/full/307/5706/113/DC1

Materials and Methods

SOM Text

Figs. S1 to S7

Tables S1 and S2

References

Movie S1

References and Notes

View Abstract

Navigate This Article