The Last r Locus Unveiled: T4 RIII Is a Cytoplasmic Antiholin

ABSTRACT

The latent period of phage T4, normally ∼25 min, can be extended indefinitely if the infected cell is superinfected after 5 min. This phenomenon, designated lysis inhibition (LIN), was first described in the 1940s and is genetically defined by mutations in diverse T4 r genes. RI, the main effector of LIN, has been shown to be secreted to the periplasm, where, upon activation by superinfection with a T-even virion, it binds to the C-terminal periplasmic domain of the T4 holin T and blocks its lethal permeabilization of the cytoplasmic membrane. Another r locus, rIII, has been the subject of conflicting reports. In this study, we show that RIII, an 82-amino-acid protein, is also required for LIN in both Escherichia coli B strains and E. coli K-12 strains. In T4ΔrIII infections, LIN was briefly established but was unstable. The overexpression of a cloned rIII gene alone impeded T-mediated lysis temporarily. However, coexpression of rIII and rI resulted in a stable LIN state. Bacterial two-hybrid assays and pulldown assays showed that RIII interacts with the cytoplasmic N terminus of T, which is a critical domain for holin function. We conclude that RIII is a T4 antiholin that blocks membrane hole formation by interacting directly with the holin. Accordingly, we propose an augmented model for T4 LIN that involves the stabilization of a complex of three proteins in two compartments of the cell: RI interacting with the C terminus of T in the periplasm and RIII interacting with the N terminus of T in the cytoplasm.

IMPORTANCE Lysis inhibition is a unique feature of phage T4 in response to environmental conditions, effected by the antiholin RI, which binds to the periplasmic domain of the T holin and blocks its hole-forming function. Here we report that the T4 gene rIII encodes a cytoplasmic antiholin that, together with the main antiholin, RI, inhibits holin T by forming a complex of three proteins spanning two cell compartments.

INTRODUCTION

The r genes of the T-even phages, first identified by laboratories of the Phage Group in the 1940s (1, 2), have a special place in the history of molecular biology. Detailed studies of the first three loci discovered—rI, rIIAB, and rIII—were foundational in working out the fundamentals of inheritance, genetic code, mutation, recombination, DNA repair, and gene structure (3–7). These mutable loci were originally discovered by their distinctive plaque morphology: large, clear, sharply defined plaques, easily distinguished from the small, fuzzy-edged, turbid plaques of the parental phages (1). The “r” designation meant “rapid lysis,” which refers to the observation that the mutant phages isolated from the r-type plaques caused rapid, culture-wide lysis at ∼25 min after infection, whereas cultures infected with the parental phages continued to increase in mass and accumulate progeny virions intracellularly for hours, in a state called “lysis inhibition”(LIN) (8). In the ensuing decades, more loci were classified as r genes based on mutant plaque phenotypes; at one point, r genes numbering up to rVI were assigned map positions (9, 10). In 1998, Paddison et al. (11) reviewed this field and concluded that only rI, rIII, and rV were directly involved in LIN, while the other genes caused lysis phenotypes through indirect physiological pathways. The rV mutants were shown to be missense alleles of gene t, which encodes T, the holin of phage T4 (12). Holins are the master lysis control proteins of the Caudovirales (13), acting to terminate the infection cycle by permeabilizing the cytoplasmic or inner membrane (IM) at a programmed time. It followed that the simplest operational model to explain the involvement of the remaining loci associated with direct LIN defects, rI and rIII, would be that the RI and RIII proteins were required to inhibit the lethal function of T and thus establish the LIN state (9, 11).

More-recent studies on T and RI have confirmed aspects of this operational model for LIN and provided molecular details for the lysis pathway of T4 (14–16). Like other holins, including the well-studied S105 holin of phage lambda, the T holin accumulates harmlessly in the host IM until it suddenly forms lethal, micrometer-scale membrane lesions at an allele-specific time. This event, which is defined as holin triggering, results in the escape of the cytoplasmic endolysin E (the product of gene e) (13) into the periplasm, where it rapidly degrades the cell wall. In turn, the loss of cell wall activates the spanin complex (the product of pseT.2 and pseT.3) (17), which then disrupts the outer membrane (OM) and completes the release of the progeny. In single infections, T4 completes this three-step pathway in ∼25 min (1). However, if the T4-infected cells are superinfected by other T4 (or T-even phages) after the first 5 min of the infection cycle, LIN is imposed (18). There has been progress on the molecular basis of LIN (15, 19–21). While most holins have two or more transmembrane domains (TMDs) and only short soluble loops connecting them (13), the T holin has a unique structure, with only a single TMD and significant N-terminal (34-amino-acid [34-aa]) and C-terminal (163-aa) cytoplasmic and periplasmic domains, respectively (15, 22) (Fig. 1). Moreover, the RI protein has been shown to have a SAR (signal anchor release) domain, which is a TMD that can escape from the membrane (19). By virtue of this domain, RI is secreted initially as a membrane-tethered periplasmic protein and is then released into the periplasm, where, in single infections, it is degraded rapidly (19, 20). However, under LIN conditions (i.e., when there is superinfection with a second T4 phage particle), RI is stabilized and accumulates in the periplasm, where it forms an equimolar complex with the cytoplasmic domain of T and inhibits triggering, thus imposing the LIN state. Additionally, if the SAR domain of RI is replaced by cleavable signal peptidase I, the processed RI protein overaccumulates in the periplasm in a stable, mature form, forms the complex with T, and imposes LIN without requiring activation by superinfection (19, 20).

FIG 1.

Topology of T4 holin T-antiholin RI interaction. T is an inner membrane protein with a single TMD (shown as a filled cylinder) and an amphipathic helix (shown as an open cylinder). RI has a SAR (signal anchor release) domain (shown as a rectangle with a dashed outline) that allows RI to be released spontaneously into the periplasm (19). If stabilized by the LIN signal, periplasmic RI binds to the C-terminal globular periplasmic domain of T. IM, inner membrane.

Because RI is a specific inhibitor of T, it is formally a member of a diverse class of proteins designated antiholins (23–26). Moreover, since RI inhibits T only under certain physiological conditions, it is the only antiholin known that transduces environmental information to effect real-time control of holin function and thus of the length and fecundity of the phage infection cycle (21). However, despite these conceptual and mechanistic advances with T and RI, the genetic basis of the LIN phenomenon remains incomplete, some decades after the genetics of the r genes were first published, because no role has been found for rIII (1, 3). Although it was reported that RIII was not required for LIN on some K-12 strains (6), rIII shares with rI the features that neither locus can suppress t lysis-null mutations and both loci are transcribed from both early and late promoters (11, 27). Recently, rIII was suggested to play a role in the propagation of T4 in slow-growing host cells (28). Here we present the preliminary results of in vivo and in vitro characterization of rIII. The results are analyzed in terms of a model that suggests the direct molecular involvement of RIII in LIN as a new class of antiholin.

MATERIALS AND METHODS

Bacterial growth and induction.

Table 1 provides a full list of phages and bacterial strains used in this study. Bacterial strains were plated on a standard LB agar plate supplemented with the appropriate antibiotics (ampicillin, 100 μg ml⁻¹; chloramphenicol, 10 μg ml⁻¹; kanamycin, 40 μg ml⁻¹). A single colony from an LB plate was used to inoculate a 3-ml overnight culture at 30°C for λ lysogens and at 37°C for nonlysogenic Escherichia coli strains, as described previously (21). Overnight cultures were diluted to an A₅₅₀ of ∼0.03 and were grown at 30°C or 37°C with aeration. Bacterial growth and lysis were monitored as described previously (21) using a Gilford Stasar III sipping spectrophotometer (Gilford Instrument Inc., Oberlin, OH). The λ lysogens were induced as described previously (14, 21). All plasmid-cloned genes were induced with 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG).

TABLE 1.

Phages, strains, and plasmids used in this study

Phage, bacterial strain, or plasmid	Descriptiona	Source or reference(s)
Phages
T4wt	Bacteriophage T4D	Laboratory stock
T4rIII	T4r67; H42-to-R (CAU-to-CGU) mutation in rIII locus	Laboratory stock
T4ΔrI	Complete deletion of rI from nt 59204 to nt 59496 in T4D genome	19
T4ΔrIII	Complete deletion of rIII from nt 130779 to nt 131042 in T4D genome	This study
λ-t	λ with holin gene S replaced with T4 holin gene t	14
λSA52G	λcI857 carrying Ala52Gly early-lysis allele of S holin gene	31
T4rBB9	W16-to-stop (UGG-to-UGA) mutation in rIII locus	Laboratory stock
T4rES35	H42-to-Q (CAU-to-CAA) mutation in rIII locus	Laboratory stock
T4rES40	K82-to-E (AAG-to-GAG) mutation in rIII locus	Laboratory stock
Bacterial strains
CQ21	E. coli K-12 ara leu lacIq purE gal his argG rpsL xul mtl ilv	Laboratory stock
CQ21 λ-t	CQ21 lysogen carrying λ-t prophage	14
CQ21 λSA52G	CQ21 lysogen carrying λSA52G prophage	This study
BL21(DE3) fhuA::Tn10	E. coli B ompT rB− mB− (PlacUV5::T7 gene 1) slyD::kan fhuA::Tn10	Laboratory stock
B834	E. coli B ompT rB− mB− met	Laboratory stock
MG1655	E. coli F− λ− ilvG rfb-50 rph-1	Laboratory stock
MDS12 tonA::Tn10 lacIq1	MG1655 with 12 deletions, totaling 376,180 nt including cryptic prophages	19
DHP1	E. coli F− cya-99 araD139 galE15 galK16 rpsL1 (Strr) hsdR2 mcrA1 mcrB1	36
Plasmids
pZE12	ColE1 origin; PLlacO-1 (PL promoter with three lacO operators); Ampr	30
pZE12-luc	Luciferase gene luc cloned under the control of PLlacO-1	30
pZE12RI	T4 rI cloned under the control of PLlacO-1 with native SD	21
pZE12RIIIo	T4 rIII cloned under the control of PLlacO-1 with native SD	21
pZE12RIIIs	T4 rIII cloned under the control of PLlacO-1 with plasmid SD	This study
pZE12RI-RIII	Tandem clone of rI–rIII inserted between KpnI and XbaI sites	This study
pET11a-RIII	pBR322 origin, T7 promoter, carrying codons 1–82 of rIII	This study
pET11a-RIIIH42R	H42-to-R (CAU-to-CGU) mutation in rIII	This study
pTB146	bla lacIq PT7::h-sumo	32
pTB146-RIII	Codons 2–82 of rIII gene inserted between SapI and XhoI sites	This study
pTB146-nT	Codons 2–34 of t gene inserted between SapI and XhoI sites	This study
pCH364	T18-empty (Ampr); N-terminal tag	36
pKNT25	Empty-T25 (Kanr); C-terminal tag	35, 36
pKT25	T25-empty (Kanr); N-terminal tag	35, 36
pCH364RIII	Codons 2–82 of rIII gene inserted between BamHI and EcoRI sites	This study
pKNT25RIII	Codons 2–82 of rIII gene inserted between XbaI and EcoRI sites	This study
pKT25nT	Codons 2–34 of t gene inserted between BamHI and EcoRI sites	This study

^aSD, Shine-Dalgarno sequence.

Phage infection and preparation of phage lysates.

Phage lysates were prepared by adding 10% CHCl₃ (vol/vol) to the E. coli cell culture after lysis, either in induced lysogens or by liquid culture infections, as described previously (19). The lysate was cleared by centrifugation at 5,000 × g, and the supernatant was filtered through a 0.22-μm syringe filter. Phage infection experiments were carried out as described previously (19, 21). For liquid culture infections, host E. coli cells were grown to an A₅₅₀ of ∼0.3 and were infected at a multiplicity of infection (MOI) of ∼5. For the observation of plaque morphology, 100-μl overnight cultures of host cells were added to 3 ml of LB top agar, and the mixture was immediately poured onto standard LB agar plates. Five microliters of phage lysates with proper dilutions was spotted onto the top agar. For the complementation experiment, BL21(DE3) fhuA::Tn10 cells carrying pET11a vectors were grown to an A₅₅₀ of ∼1 at 37°C and were induced with 1 mM IPTG for 2 h before being mixed with LB top agar and poured onto LB plates containing proper antibiotics and 1 mM IPTG. All plates were incubated for ∼16 h at 37°C. The plaque sizes were analyzed using ImageJ software (NIH, Bethesda, MD).

Standard DNA manipulations and sequencing.

All plasmids used in this study are listed in Table 1. Isolation of plasmid DNA, DNA amplification by PCR, DNA transformation, and DNA sequencing were performed as described previously (15, 22, 29). The DNA sequences of oligonucleotides (primers) are listed in Table 2. All purified oligonucleotides (primers) were purchased from Integrated DNA Technologies (Coralville, IA). Restriction and DNA-modifying enzymes were purchased from New England BioLabs (Ipswich, MA). The manufacturer's instructions for performing reactions were followed. The DNA sequences of all constructs were verified by a sequencing service provided by Eton Bioscience (San Diego, CA).

TABLE 2.

List of oligonucleotides (primers) used in this study^a

Primer name	Sequence
RIIIS CLONING F	CGGTACATTAAACAATTACAACACGCTC
RIIIS CLONING R	GGCTCTAGATTACTTCAGTGTTACCACAAAGTG
RIIIS PET F	GGAATTCCATATGATTAAACAATTACAACACGCTC
RIIIS PET R	GCGGGATCCTTACTTCAGTGTTACCACAAAGTG
RIII DEL +500 F	GGGGTACCCATCTGTTAACAAAAAGGAAAAACG
RIII DEL −500 R	GCTCTAGAGCGTTCAGATTAATCGTTTTCA
RIII DEL MIX F	TTTTAATCTCTAACGAGGGAGATTCACTGCCTTAGTGTGAGC
RIII DEL MIX R	CCGAGTTTTAATCTCTAACGAGGGAGATTCACTGCCTTAGT

^aAll oligonucleotides used were designed for this study.

PCR and plasmid construction.

T4D phage lysate was used directly as the PCR template for cloning out T4 genes. Pfu DNA polymerase was used for all PCRs by following standard protocols provided by Promega (Madison, WI). Site-directed mutagenesis was performed as described previously (22). The rIII gene, either with its native ribosome binding site (GAG) or with a stronger ribosome binding site (AGGAG), was cloned into the medium-copy-number, IPTG-inducible vector pZE12 (30). Plasmid pZE12-RIII_o and pZE12-RIII_s were constructed by inserting T4 DNA from nucleotide (nt) 130738 to nt 131080 (RIII_o), or from nt 130785 to nt 131033 (RIII_s), into pZE12 between the KpnI and Xbal sites. Plasmid pET11aRIII has the same insertion as pZE12RIII_s between its Ndel and BamHI sites. Plasmid pZE12RI-RIII was made by inserting a tandem clone of rI and rIII genes with their original ribosome binding site into plasmid pZE12. These plasmids were transformed into a CQ21λ-t lysogen, in which the λ holin gene S has been replaced by T4 gene t (14). In this system, T can be expressed from the λ-t prophage and RI and/or RIII can be expressed in trans from pZE12 plasmids by adding 1 mM IPTG after lysogenic induction. A λS_A52G lysogen was used as a control, since S_A52G confers a ∼20-min lysis time, similar to that with the t gene (31). Plasmid pTB146 is a derivative of plasmid pET11a encoding an N-terminal His₆-Sumo tag (29, 32). Plasmids encoding His-Sumo-tagged versions of RIII and nT (the N-terminal domain of T)—pTB146-RIII and pTB146-nT—were constructed by inserting codons 2 to 81 of the rIII gene, or codons 2 to 34 of the t gene (nt 160218 to nt 160322 of the T4 genome), respectively, into the pTB146 plasmid between its SapI and XhoI sites.

Constructing a T4 rIII deletion mutant.

T4ΔrIII was constructed by homologous recombination between pZE12-ΔrIII and T4D, as described previously for T4ΔrI (19). pZE12-ΔrIII was made by deleting the rIII gene from plasmid pZE12-rIII-flank, which contains T4 DNA from nt 130231 to nt 131541 between its KpnI and Xbal sites, using our previously described method (33). Plasmid pZE12-ΔrIII was transformed into E. coli strain MDS12 tonA::Tn10 lacI^q1, and the transformants were grown to an A₅₅₀ of ∼0.4. The culture was infected with T4D phage at an MOI of 10 for 3 h at 37°C with aeration and was then lysed by adding 10% (vol/vol) CHCl₃. T4 rIII recombinants in this lysate were enriched three times for early lysis as described previously (19). The enriched lysate was plated onto E. coli B834 and was screened for r plaque morphology. The ΔrIII deletion was confirmed by PCR and sequencing.

SDS-PAGE and Western blotting.

SDS-PAGE and Western blotting were conducted as described previously (22) Ten percent trichloroacetic acid (TCA) was used to precipitate proteins from the whole-cell samples. Reducing sample loading buffer (SLB) supplemented with β-mercaptoethanol was used for resuspending protein samples unless otherwise indicated. RIII proteins transferred to polyvinylidene difluoride (PVDF) membranes were detected using a rabbit polyclonal anti-RIII antibody purchased from GenScript (Piscataway, NJ). The monoclonal anti-His tag antibody was purchased from Sigma-Aldrich (Carlsbad, CA). To detect proteins, blots were incubated overnight at 4°C with anti-RIII or anti-His at a dilution of 1:4,000 in 3% milk–Tris-buffered saline (TBS) buffer. Blots were developed with the SuperSignal West Femto chemiluminescence kit, purchased from Thermo Fisher Scientific (Rockford, IL). The chemiluminescence signal was detected using a Bio-Rad ChemiDoc XRS system (Bio-Rad Laboratories, Hercules, CA). Images were obtained and analyzed by Quantity One 1-dimensional (1-D) analysis software (Bio-Rad Laboratories, Philadelphia, PA).

B2H assay.

Bacterial two-hybrid (B2H) assays were conducted as described previously (34–36). Plasmids were constructed by inserting codons 2 to 81 of the rIII gene, or codons 2 to 34 of the t gene, into plasmids pCH363, pKNT25, pCH364, and pKT25, as described previously (36). Different pairs of plasmids were transformed into strain DHP1 and were grown to an A₅₅₀ of ∼0.3 in LB with 0.2% glucose and appropriate antibiotics (ampicillin, 50 μg ml⁻¹; kanamycin, 25 μg ml⁻¹). For the plate assay, 5 μl of cell cultures was spotted onto M9 minimal medium plates supplemented with 0.2% glucose, 40 μg ml⁻¹ 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal), 150 μM IPTG, and proper antibiotics and was incubated for 48 h at 25°C.

Pulldown assays.

Plasmid pET11a and pTB146 derivatives, described above, were transformed into BL21(DE3) fhuA::Tn10 strains. Pulldown assays were conducted according to the instructions in the manufacturer's protocol from the Dynabeads His-Tag Isolation and Pulldown kit (Thermo Fisher Scientific, Rockford, IL). All incubations and reactions were carried out at 4°C, and beads were collected using a DynaMag-2 magnet (Thermo Fisher Scientific). All samples were resuspended in SLB and were boiled for 5 min to elute proteins, which were then analyzed by SDS-PAGE and Western blotting, performed as described above.

RESULTS

rIII is required for LIN in both E. coli B and E. coli K-12 backgrounds.

The role of rIII in LIN has been ambiguous, with reports differing as to whether rIII is required on E. coli B but not K-12 strains (6, 9, 37). In our hands, the classic rIII allele T4r67, which was used as the standard allele in the early T4 genetic map studies, formed r-type plaques on the lawns of both E. coli B834 and MG1655 (Fig. 2A); the plaques were somewhat smaller than those formed by T4rI but significantly larger than wild-type (wt) T4 plaques (Table 3). It was also reported that different T4 rIII mutants differed in plaque size, suggesting a possible correlation between the location of mutations on the rIII locus and plaque morphology (38). However, when we compared the plaque morphologies of four different T4 rIII-defective mutants (T4rIII, T4rBB9, T4rES35, and T4rES40) on B834, we did not observe significant differences (Table 3). Nevertheless, as the first step for interrogating the role of rIII in LIN, we constructed an rIII deletion allele, T4ΔrIII, to eliminate the potential for partial reversion. As shown in Table 3 and Fig. 2A, T4ΔrIII formed r-type plaques that were larger than wt plaques but smaller than those of T4ΔrI. Moreover, the wt plaque morphology could be complemented by a plasmid-borne rIII gene (Fig. 2B). In infections of both E. coli B and E. coli K-12 cultures under conditions where the wt T4 exhibited LIN, T4ΔrIII infections showed lysis at ∼25 min, reproducibly later than the lysis time of T4ΔrI (∼18 min), (Fig. 3A). The simplest notion, based on the established role of RI in LIN, is that RI expressed in the T4ΔrIII infection causes transient LIN and that, by extension, RIII is required for stable LIN in both E. coli B and K-12 strains.

FIG 2.

Plaque morphologies of T4 and its r mutants. (A) Plaque morphology of wt T4 (T4D) and T4 mutants on either an E. coli B strain (B834) or an E. coli K-12 strain (MG1655). Bars, 2.5 mm. The average plaque sizes of T4D, T4ΔrIII, T4rIII, and T4ΔrI on E. coli B834 or E. coli MG1655 are given in Table 3. (B and C) Complementation of r plaque morphology. (B) Differences in plaque sizes are shown as the ratio of the average phage plaque radius (r) to the average plaque radius of T4D plated on B834 (r₀). Black bars, T4D; hatched bars, T4ΔrIII; open bars, T4rIII; gray bars, T4ΔrI. Strains are indicated by numbers as follows: 1, MG1655; 2, BL21(DE3) fhuA::Tn10 with no plasmid; 3, BL21(DE3) fhuA::Tn10/pET11a-RIII; 4, BL21(DE3) fhuA::Tn10/pET11a-RIII_H42R. (C) T4 rIII mutants plated onto E. coli strains expressing wt RIII restored wt T4 plaque morphology, whereas the expression of RIII_H42R did not.

TABLE 3.

Mean diameters of phage plaques

Phage	Mean (±SD) diam (mm) of phage plaques in the following host:
	B834	MG1655	BL21(DE3)	BL21(DE3)/pET11aRIII	BL21(DE3)/pET11aRIII-H42R	BL21(DE3)/pTB146	BL21(DE3)/pTB146-nT
T4D	0.57 (±0.04)	0.67 (±0.04)	0.63 (±0.06)	0.53 (±0.05)	0.57 (±0.05)	0.68 (±0.04)	1.07 (±0.05)
T4rIII	1.09 (±0.09)	1.43 (±0.07)	1.07 (±0.09)	0.62 (±0.03)	0.92 (±0.08)	1.32 (±0.08)	1.38 (±0.06)
T4ΔrIII	0.92 (±0.08)	1.12 (±0.06)	0.95 (±0.07)	0.52 (±0.06)	0.86 (±0.08)	1.13 (±0.08)	1.17 (±0.07)
T4ΔrI	1.29 (±0.09)	1.88 (±0.11)	1.24 (±0.07)	1.21 (±0.03)	1.34 (±0.06)
T4rBB9	0.89 (±0.15)
T4rES35	1.02 (±0.05)
T4rES40	1.05 (±0.05)

FIG 3.

(A) Lysis in infections of E. coli B strain B834 (left) or E. coli K-12 strain MG1655 (right) with T4 and derivatives. ×, no phage; ●, T4D (wt); ▲, T4ΔrI; △, T4ΔrIII. Cultures were grown to an A₅₅₀ of ∼0.25 at 37°C and were then infected with T4 at an MOI of ∼5. (B) Inductions (at time zero) of CQ21λ-t (left) or CQ21λS_A52G (right) lysogens carrying the indicated genes cloned under IPTG control in the context of the pZE12 plasmid. Plasmids were also induced by the addition of 1 mM IPTG at time zero. Multiplication signs, luc (negative control); triangles, RI; circles, RIII_s; diamonds, RI-RIII. (C) RIII missense mutants exhibit an intermediate LIN phenotype. CQ21λ-t lysogens carrying pZE12 plasmids with the indicated genes were induced at time zero. ×, luc; ▲, RI; ●, RIII_s; ◆, RI-RIII. (Left) □, RIII_G24D; ♢, RIII_H42R; ▽, RIII_A70V. (Right) □, RI-RIII_G24D; ♢, RI-RIII_H42R; ▽, RI-RIII_A70V.

Identification of the RIII protein.

rIII encodes an 82-aa polypeptide without any secretion signals (Fig. 4A) and had not been identified as a protein species in T4-infected cultures. We raised a polyclonal antibody against an RIII oligopeptide sequence predicted to be highly immunogenic (Fig. 4A). The RIII protein could be visualized by immunoblotting in samples taken from cells infected with wt T4 but not with T4ΔrIII (Fig. 4B); the mobility of this RIII species corresponded to a molecular mass slightly lower than that predicted (8.9 kDa versus 9.3 kDa predicted), presumably due to the high content of charged residues (24 out of 82) (Fig. 4A).

FIG 4.

(A) Primary structure of RIII and the N terminus of T4 holin T (nT), with LIN-defective and lysis-defective alleles indicated by downward-pointing arrows. Conserved residues are underlined. The shaded area represents the oligopeptide used to raise the anti-RIII antibody. The predicted secondary structure is indicated as follows: open rectangles, helices; lines, turns; arrows, β-sheets; shaded rectangle, amphipathic helix. (B) RIII protein accumulates during infection. For each sample, 1 A₆₀₀ equivalent of cells was loaded. The anti-RIII antibody was used in Western blotting. The arrow indicates the predicted molecular mass (9.3 kDa) for the RIII monomer.

Recapitulating the role of RIII in LIN in the λ context.

To address the role of RIII in LIN, we used a convenient system based on the inducible lambda prophage λ-t, in which the λ holin gene S is replaced by T4 t (14). Not only was this hybrid phage previously shown to recapitulate T4 lysis timing and LIN at physiological levels of expression; it allows the coexpression of selected T4 genes cloned in inducible plasmid vectors without the confounding effects of T4-mediated host DNA degradation and translational repression (14, 15, 21, 39). This system mimics the T-dependent lysis in the λ context, where effects of T4 genes other than t are excluded. To provide RIII in trans, the rIII gene was cloned into the medium-copy-number plasmid vector pZE12 (30). Two isogenic clones were constructed with different Shine-Dalgarno (SD) sequences serving the rIII cistron: the relatively weak native sequence (GAG) and a stronger near-consensus sequence (AGGAG). The resulting plasmids, pZE12RIII_o (where the subscript letter “o” indicates the original SD sequence) and pZE12RIII_s (where the subscript letter “s” indicates the strong SD sequence), were transformed into the λ-t lysogen. Induction of the λ-t lysogen resulted in reproducible and sharply defined lysis at ∼ 20 min; induction of pZE12RIII_s with IPTG conferred a mild but reproducible lysis delay of ∼5 min (Fig. 3B). In contrast, lysis was much more severely affected by the induction of an isogenic clone of the T4 antiholin gene rI, as shown previously (21). Induction of pZE12RIII_o did not affect the timing of lysis, probably due to the lower protein expression level (data not shown).

We next asked if RIII can extend RI-mediated LIN. An isogenic plasmid with rI and rIII cloned in tandem was constructed and introduced into the λ-t lysogen. Induction of this plasmid, pZE12RI-RIII, led to drastically delayed LIN relative to that with induction of either pZE12RIII_s or pZE12RI (Fig. 3B); under these conditions, the LIN state lasted as long as 80 min and then gradually deteriorated. Using this system, we tested three rIII missense alleles isolated by UV mutagenesis: the G24D, H42R, and A70V alleles (9) (Fig. 4A). In the absence of RI, two of the alleles, the G24D and H42R alleles, exhibited a slight but reproducible LIN defect, although the phenotype was subtle due to the relatively small effect of the parental rIII allele under these conditions (Fig. 3C, left). Coinduction of these rIII alleles with rI, however, resulted in distinct intermediate LIN defects, with lysis times ranging from 40 min to ∼60 min, indicating that these are partially defective alleles, at least in the lambda context (Fig. 3C, right). The lysis blockage was t specific, as evidenced by the fact that isogenic experiments with a lambda holin allele, S_A52G, which has an early lysis phenotype that matches the normal t lysis time (31), did not show lysis delays in induction of pZE12RIII_s, pZE12RI, or pZE12RI-RIII (Fig. 3B). Taken together, these data indicated that RIII has T-specific antiholin activity.

RIII binds to the cytoplasmic N terminus of T.

Next, we addressed the molecular basis of RIII participation in LIN. The simplest hypothesis is that, like other antiholins, including RI, RIII affects holin triggering by binding directly to holin T and blocking hole formation in the IM. Since RIII has no membrane or export signals (Fig. 4A), the only possible target for RIII is the N-terminal cytoplasmic domain of T (nT), which has 34 aa and is required for holin function (15). To test this idea, we used the bacterial two-hybrid system, based on intragenic complementation of CyaA function (34). We fused the nT, wt RIII, and four RIII mutant allele sequences to various combinations of T25 and T18 fragments of CyaA. As shown in Fig. 5A, this system revealed a strong self-interaction of RIII in vivo, which was abolished by the G24D allele, partially affected by H42R, and unaffected by L43Q or R75C. Transformants carrying plasmids expressing T18-RIII and T25-nT resulted in light but reproducible signals (Fig. 5B), suggesting a relatively weak interaction between RIII and nT. Significantly, none of the four RIII mutant fusion constructs retained nT-binding ability (Fig. 5B). These results correlate with the liquid culture lysis results (Fig. 3C) and indicate that the nT-RIII interaction is affected by the changes in the lysis-defective RIII alleles.

FIG 5.

Bacterial two-hybrid results showing self-interaction of RIII (A) and interaction of RIII and the N terminus of T (nT) (B). T18, protein fused to the T18 fragment of CyaA protein; T25, protein fused to the T25 fragment of CyaA. T18 or T25 fragments without RIII or nT fusion serve as negative controls (−).

To address the nT-RIII interaction in vitro, we constructed versions of nT and RIII tagged at the N terminus with the His₆-Sumo moiety (see Materials and Methods). After induction in a T7-based overexpression system, both His₆-Sumo-nT and His₆-Sumo-RIII accumulated as soluble forms (Fig. 6, top, lanes 2 to 5). To detect complexes formed in vitro, the Sumo-tagged nT and RIII proteins were bound to magnetic beads, mixed with cell lysates containing wt RIII or mutant RIII_H42R protein, fractionated (bound and unbound fractions), and then analyzed by immunoblotting. The results showed that both the wt RIII and mutant RIII_H42R proteins form complexes with His₆-Sumo-RIII but that only wt RIII forms complexes with His₆-Sumo-nT (Fig. 6, bottom). The simplest interpretation is that the H42R mutation completely abrogates the RIII-nT interface but not the RIII homo-oligomerization interface, a pattern consistent with the results of the bacterial two-hybrid experiments.

FIG 6.

In vitro interaction between nT and RIII. His-Sumo-tagged nT or RIII was bound to anti-His Dynabeads, and RIII protein pulled down by Dynabeads was analyzed by Western blotting. (Top) Western blotting using an anti-His antibody. Lane 1, His-Sumo tag only (dashed arrow); lanes 2 and 4, His-Sumo-tagged nT (filled arrowhead), lanes 3 and 5, His-Sumo-tagged RIII (open arrow). (Bottom) RIII protein (arrow) is visualized by use of an anti-RIII antibody.

The cytoplasmic N-terminal domain of T can block lysis inhibition in an RIII-specific manner.

The finding that RIII binds nT in vitro and in E. coli in the context of the two hybrids suggested that the r phenotype could be imposed in vivo by titrating the RIII produced in a T4 infection with the Sumo-tagged nT derivative. To test this idea, we plated T4 onto lawns of cells in which His₆-Sumo-nT overexpression had been induced; under these conditions, wt T4 generated plaques distinctly larger and clearer than those generated on the isogenic control strain expressing the His₆-Sumo tag (Fig. 7). Neither T4rIII nor T4ΔrIII plaque morphology was affected by the overexpression of nT (Fig. 7).

FIG 7.

Rescue of r plaque morphology by overexpression of the N terminus of T (nT). (A) T4 phages were plated onto lawns of E. coli BL21(DE3) fhuA::Tn10 carrying a control plasmid expressing His-Sumo-nT (pTB146-nT) or His-Sumo (pTB146; negative control). Bar, 2.5 mm. (B) The quantification of plaque sizes is shown as the ratio of the average plaque radius (r) to the average radius of T4D plaques plated on E. coli BL21(DE3) fhuA::Tn10 (r₀). Filled bars, T4D; hatched bars, T4ΔrIII; open bars, T4rIII.

DISCUSSION

Among the Caudovirales, the lysis timing effected by the holin defines the length and fecundity of the phage infection cycle. Mutational analysis has shown that holin-mediated lysis timing can be drastically altered by single missense changes (15, 39–41), leading to the suggestion that this extreme mutational sensitivity is an evolutionary fitness factor, allowing phages to mutate rapidly to a radically different life cycle length in response to altered environmental conditions (42). Despite the implied correlation between lysis timing and the environment, the T4 LIN phenomenon remains the only documented example of real-time regulation of lysis timing. Genetic analysis has shown that mutations in two of the classic T4 plaque morphology loci, rI and rV, the latter allelic to the t holin gene, confer an absolute defect in LIN. Our work had shown that RI is a secreted protein that is initially synthesized as a periplasmic protein tethered to the membrane with an N-terminal signal anchor release (SAR) domain (19). The presence of the SAR domain allows RI to be released into the periplasm and also confers extreme proteolytic instability on RI. Overexpression of the wt rI gene was shown to impose a delay on T-holin triggering in the lambda context (21). A chimeric rI gene in which the SAR domain was replaced by a cleavable signal sequence generated a proteolytically stable periplasmic RI and, expressed in trans with t, imposed a stable LIN state. A model has been proposed in which an unknown LIN signal is generated by a superinfecting T4 virion. Under these conditions, it is suggested that the T4 Spackle and Imm proteins force the superinfecting virion to eject its capsid contents ectopically into the host periplasm (43). Some component of these virion contents, which include both the T4 genomic DNA and ∼1,000 protein molecules (11, 37), acts as a signal to stabilize the periplasmic RI protein. In this model, RI accumulates and binds to the periplasmic domain of the T4 holin T in such a manner that T triggering is inhibited (19). Significant progress has been made on the RI-T interaction. The soluble domain of RI, sRI, has been purified and shown to be largely α-helical in structure (22). In addition, the sRI molecule was able to bind the soluble domain of T, sT (22, 44) and prevent it from aggregating. Crystal structures of sRI and the sRI-sT complex have been determined (44). However, major gaps remained in our understanding of the LIN phenomenon. First, the signal provided by the superinfecting phage is completely unknown. In addition, the possible role of other r loci, most notably rIII, was not reflected in the model.

In this study, we have shown that rIII, as well as rI, is unambiguously required for LIN on both E. coli K-12 and B hosts, resolving a long-standing controversy (9, 11). Moreover, we have shown that the rIII gene, expressed in trans with the T4 holin gene t, can effect a small but reproducible lysis delay in a T-specific manner. In addition, rIII expression significantly stabilized the LIN state imposed by overexpression of wt rI, which otherwise imposes a lysis delay that collapses after ∼45 min. Since RIII is a cytoplasmic protein, the simplest notion is that RIII acts by binding to the short cytoplasmic domain of T (nT). Evidence supporting this notion was obtained from bacterial two-hybrid analysis and pulldown assays, which revealed a specific interaction between nT and the full-length RIII polypeptide. Importantly, known dysfunctional rIII missense mutations caused a defect in the RIII-mediated stabilization of RI-LIN. Finally, bacterial two-hybrid evidence was provided showing that RIII has a dimerizing or oligomerizing propensity, which may be functionally important in view of the fact that one of the known rIII-defective missense mutations abrogates the response.

Taken together, these results indicate that both RI and RIII are, strictly defined, specific antiholins of the T4 holin T, and they suggest an expansion of our previous model to include inhibitory interactions on both sides of the cytoplasmic membrane (Fig. 8). In this scenario, RI acts as the LIN master regulator by receiving the signal generated by the superinfecting virion. Stabilization of RI leads to the formation of RI-T complexes that prevent the T protein from participating in the holin-triggering pathway. The available evidence indicates that holin triggering occurs when the holin reaches a critical 2-dimensional concentration and forms large oligomers, or rafts, within which the lethal holes are formed (42, 45). The simplest notion is that RI may simply block the homo-oligomerization, and thus the triggering, of T, a notion consistent with the ability of sRI to prevent the aggregation of sT (20). In our new model, we suggest that RIII participates in LIN by stabilizing the RI-T complexes. Indeed, the sRI-sT crystal structures were in the form of sT-sRI-sRI-sT heterotetramers (44). Thus, an attractive notion is that upon the onset of LIN, T-RI-RI-T heterotetramers are formed, providing a symmetric binding site at which RIII dimers can bind to the cytoplasmic nT domains (Fig. 8). It should be noted that, from this perspective, RIII is the first example of an antiholin with no secretory or membrane signal and also that the RI-RIII combination is the first example of a multiple-antiholin system.

FIG 8.

Current model of LIN involving two antiholins. Both RI and RIII are required for stable LIN. (A) In a single-phage infection, the antiholin RI will be degraded by the periplasmic protease DegP after spontaneous release into the periplasm. Cell lysis occurs at ∼25 min. (B) In a superinfection, the DNA of a superinfecting T4 phage will be ectopically ejected into the periplasm, generating the “signal” to stabilize the periplasmic antiholin RI. This leads to the accumulation of RI, which then binds the periplasmic domain of T in a T-RI-RI-T heterotetramer. This facilitates the binding of the cytoplasmic antiholin RIII to the N terminus of T. This unique sandwich-like structure spanning two cell compartments robustly blocks the participation of T in hole formation.

As noted above, major gaps remain in our understanding of the T4 LIN phenomenon, which deserves attention not only because of its historical status but also as a richly documented phenomenon that may be important for our understanding of phage propagation in liquid culture and of environmental scenarios that may be relevant to phage-based therapeutics. Immediate future efforts will be directed at determining the nature of the RIII-nT interaction at the structural level.