Study of Ren, RexA, and RexB Functions Provides Insight Into the Complex Interaction Between Bacteriophage λ and Its Host, Escherichia coli

Abstract

The phage λ rexA and rexB genes are expressed from the P_RM promoter in λ lysogens along with the cI repressor gene. RexB is also expressed from a second promoter, P_LIT, embedded in rexA. The combined expression of rexA and rexB causes Escherichia coli to be more ultraviolet (UV) sensitive. Sensitivity is further increased when RexB levels are reduced by a defect in the P_LIT promoter, thus the degree of sensitivity can be modulated by the ratio of RexA/RexB. Expression of the phage λ ren gene rescues this host UV sensitive phenotype; Ren also rescues an aberrant lysis phenotype caused by RexA and RexB. We screened an E. coli two-hybrid library to identify bacterial proteins with which each of these phage proteins physically interact. The results extend previous observations concerning λ Rex exclusion and show the importance of E. coli electron transport and sulfur assimilation pathways for the phage.

Introduction

The λ rexA and rexB genes and the λ cI repressor gene are all expressed from the P_RM promoter for repressor maintenance in λ lysogens (Fig. 1). In lysogens, rexB is also transcribed from the constitutive promoter, P_LIT^1–4 located near the 3′ end of the rexA gene (Fig. 1). We have previously identified a role for the RexA and RexB proteins in prophage induction.⁵ In addition, expression of the λ rexA and rexB genes perturbs Escherichia coli cellular energetics^6,7 and increases ultraviolet (UV) sensitivity.^8,9 Both RexA and RexB contribute to these cellular metabolic effects. The severity of the energetic dysfunction is sensitive to the ratio of RexA to RexB, with excess RexA increasing toxicity^6,7; however, RexA without RexB is not toxic. Parma et al⁶ suggested that expression of RexB from P_LIT buffers the cell against RexA effects. Consistent with this idea, a P_LIT promoter mutation, P_LIT-10, causes a defect in UV induction of cI⁺ lysogens, and premature asynchronous lysis of heat-induced cI857 lysogens.¹⁰

FIG. 1.

Genetic map of bacteriophage λ immunity region and P_R operon. (A) The CI and Cro repressors are expressed in the P_RM and P_R operons, respectively, and determine whether the phage is in the lysogenic or lytic state, respectively. The rexA and rexB genes are downstream of cI and are expressed with cI from the P_RM promoter in established lysogens. The P_LIT promoter is embedded near the end of the rexA coding sequence. The rexB gene is transcribed from two promoters, P_RM and P_LIT, whereas cI and rexA are transcribed only from P_RM.¹⁰ Promoter transcripts are indicated by black arrows. Transcription from P_RM and P_LIT stops at the intrinsic transcriptional terminator, T_IMM, immediately downstream of rexB. The phage DNA replication genes O and P are transcribed from P_R, as is ren. (B) In strains LT732 and LT1964, the phage λ P_RM operon with the temperature-sensitive cI857 repressor gene, as shown here, is inserted at the Escherichia coli lac operon with lacZ expression driven from the P_R lytic promoter⁶⁸ and with the lacI gene and the lac promoter having been replaced with the P_R promoter. This leaves the lacZ ribosome-binding site and the remainder of the lacZYA operon intact, but not shown here. LT1964 differs from LT732 by having the promoter mutation P_LIT-10. These strains were used to determine the effects that RexA and RexB functions have on E. coli UV sensitivity and metabolism. Genetic maps are similar to those previously shown.⁵

RexAB expression from the λ prophage inhibits replication of certain phage mutants; a summary of a phage sensitive to RexAB is provided in Table 1. The classic example of a phage subject to Rex exclusion is a T4 rII mutant,¹¹ and T4 rII-infected λ lysogens display defects in energy metabolism.^11–13 The nonessential proteins encoded by the T4 rIIA and rIIB genes associate with the E. coli inner membrane and with the T4 DNA replication complex.^14–16 Colowick and Colowick¹⁷ postulated that during T4 infection, the T4 RIIA and RIIB proteins provide increased energy during glycolysis.

Table 1.

Summary of Phage Mutants Sensitive to Physiological Levels of RexAB

Phage mutant inhibited	Function targeted	Dysfunctiona
T4 rII	Recombination-dependent DNA replication	Loss of PMF ATP hydrolyzed Macromolecular synthesis NAD leakage
Heteroimmune λ red	Recombination	Macromolecular synthesis (probably same as for T4 rII)
Heteroimmune λ ren	Unknown

^a T4 rII^6,7,11,13; λ red and ren.^18–20

ATP, adenosine triphosphate; PMF, proton motive force; NAD, nicotinamide adenine dinucleotide.

Under some conditions, RexAB can also inhibit phage λ development.^18–20 Heteroimmune λ such as imm434 and λ imm21 phage are not blocked by λ CI repressor and are able to replicate well in λ lysogens. However, red or ren mutants of these same phage are inhibited by the combined expression of RexA and RexB.^18–20 The red genes encode the λ generalized recombination system, Red.²¹ Red recombination acts at the replication fork²² and stimulates phage DNA replication.^23,24

The ren gene is adjacent to the phage DNA replication genes O and P (Fig. 1) and encodes a basic 97 amino acid protein (pI 9.9), Ren. Since the Red system is involved in DNA replication and repair, Ren may be as well. We previously reported that λ gam mutants are also excluded by RexAB²⁵; we also find that defects in RecB and RecC functions are extragenic suppressors of this exclusion, as are sites in the DNA called Chi^18–20,26 that activate RecBCD-mediated recombination. Collectively, these results suggest that RexAB-imposed defects in DNA recombination may prevent phage λ red, ren, and gam mutants from growing.

Using bacterial strains expressing the phage immunity region (Fig. 1B), we confirmed the previous observation^8,9 that combined expression of λ RexA and RexB confers increased UV sensitivity to E. coli (Fig. 2A). We find that the degree of UV sensitivity varies with the ratio of RexA to RexB. The expression of Ren can reduce this UV sensitivity phenotype (Fig. 2C, D); Ren can also rescue RexAB-dependent abnormal lysis of heat-induced λcI857 lysogens¹⁰ (Fig. 3). Bacterial adenylate cyclase two-hybrid system (BACTH) analyses demonstrate in vivo physical interactions between proteins.^27,28 Using the BACTH system, we have found evidence⁵ for physical interaction between all combinations of the λ proteins RexA, RexB, Ren, and the phage replication initiator protein O.

FIG. 2.

λ RexA and RexB function sensitizes Escherichia coli to UV irradiation, and λ Ren rescues UV sensitivity. UV survival curves were generated as described in the Materials and Methods section. The phage λ genes present in each strain are indicated in the figure legend. For all strains except those lacking any phage DNA, the phage genes are inserted in single copy at the E. coli lac operon⁵ and the phage cI857 repressor gene is present. The number of independent experiments (n) is 3 unless otherwise indicated, and the SD is indicated by vertical bars. Strain LT351 is MG1655 (●), which provides a wildtype control of survival after UV exposure, and is shown in all graphs (n = 4). (A) Expression of phage λ rexA and rexB genes increases UV sensitivity of E. coli. Strain LT732 (○) bears the λ immunity region with cI857, rexA, rexB, and T_IMM (n = 4) as shown in Figure 1B. Strain LT1964 (▵) has the same DNA as LT732 but with a mutation in the −10 region of the P_LIT promoter, P_LIT-10. Survival of a recA mutant, strain LT1015, is shown for comparison (▼). (B) Strains lacking rexA and/or rexB have UV resistance like that of MG1655. The individual rex mutants LT2154 rexA<>cat P_LIT-10 (▼) and LT2155 rexB<>cat P_LIT-10 (▵) contain the P_LIT-10 mutation. The doubly rex mutant strain, LT1651 cI857 (rexA rexB)<>cat (○), expresses only the phage cI857 gene. (C) The UV sensitivity of strain LT732 (○) caused by concurrent expression of rexA and rexB is attenuated in strain LT1761 (▲) by providing ren gene expression from the P_BAD promoter with the addition of 0.2% arabinose to the overnight culture and the plating agar. (D) Expression of λ ren largely rescues the additional degree of UV sensitivity caused by the P_LIT-10 promoter mutation. LT1767 (○) is LT1651 with P_BAD-ren, grown and plated in the presence of 0.2% arabinose. LT1964 (▵) has the same DNA as LT732 but has the P_LIT promoter mutation, P_LIT-10 (n = 6). Strain LT2439 (▲) is LT1964, with P_BAD-ren induced by the addition of 0.2% arabinose. Note that in the presence of cI857, Ren does not affect UV sensitivity of MG1655 (compare LT1651 in B with LT1767 in D). SD, standard deviation; UV, ultraviolet.

FIG. 3.

λ Ren rescues aberrant early lysis of heat-induced cI857 P_LIT-10 lysogens. High temperature (42°C) inductions of λ cI857 lysogens were performed as described in Thomason et al.¹⁰ Briefly, cultures were grown to an OD₆₀₀ of 0.4 in a 32°C water bath with shaking and at time zero were shifted to 42°C for 15 min, and then they were placed at 37°C until lysis occurred. OD₆₀₀ was measured at 5 min increments. Strain LT447 (●), a MG1655 cI857 lysogen with intact rexA and rexB functions illustrates normal heat induction (n = 5). LT2228 (○) contains the P_LIT-10 mutation but is otherwise identical to LT447 (n = 5). Phage induction is impaired by the P_LIT-10 mutation, causing asynchronous lysis.¹⁰ LT2336 (⋄) is identical to LT2228 except for a single chromosomally located copy of P_BAD-ren, which was activated by adding 0.2 arabinose to the growth medium (n = 4). As seen, ren expression largely counteracts the detrimental effects caused by the P_LIT-10 promoter mutation. No rescue occurs in the absence of arabinose addition (not shown). Average phage yield for the wildtype LT447 culture was 1.4 × 10¹⁰/mL ±3.7 × 10⁹/mL (n = 5), whereas that of the LT2228 culture with P_LIT-10 was 2.5 × 10⁹/mL ±1.0 × 10⁹/mL (n = 5). Average phage yield for strain LT2336 in the absence of arabinose was 3.7 × 10⁹/mL ±1.2 × 10⁹/mL (n = 4); in the presence of arabinose, the same strain gave a phage yield of 1.5 × 10¹⁰/mL ±6.6 × 10⁹/mL (n = 4). SD is shown by the error bars.

Here, we report the screening of an E. coli two-hybrid library²⁹ with RexA, RexB, and Ren and have identified bacterial proteins with which each physically interacts (Fig. 4; Supplementary Table S2A, B).

FIG. 4.

β-galactosidase levels indicate interaction between λ RexA or RexB or Ren and Escherichia coli proteins found in a BACTH two-hybrid library screen. The BACTH system^27,28 was used to look for evidence of in vivo intracellular interaction between the phage λ RexA, RexB, and Ren proteins and E. coli host proteins using plasmids expressing E. coli proteins from a genomic library, as described in the Materials and Methods section. The protein-protein interactions observed for RexA, RexB, and Ren are plotted separately. In each case, the y-axis shows units of β-galactosidase and the x-axis shows the gene identified from the library that allowed growth on minimal maltose when co-transformed with the phage plasmid into a cya mutant strain, BTH101. E. coli library DNA is inserted into a low-copy (Kan^R) vector; the phage genes are in a high-copy (Amp^R) vector. Each experiment was done at least three times, individual measurements are shown, and SD is indicated. The numerical data used to make the bar graphs are presented in Supplementary Table S2A and B. All E. coli genes shown conferred growth on minimal maltose, indicating that their protein products interact with the phage proteins to produce cAMP; however, only a few of the plasmid pairs gave substantial β-galactosidase activity. Previously,⁵ we determined negative control values for the two-hybrid measurements; these negative controls averaged 53 ± 13 (SD, n = 114) β-galactosidase units. The stronger interactions we emphasize here have average values significantly higher than the negative controls. (A) Of the E. coli proteins interacting with RexB, of interest here is the NuoM subunit of the NADH dehydrogenase. (B) RexA associates with the sulfur metabolism enzyme CysN. Another NADH dehydrogenase subunit, NuoI, was identified as a weak interactor in the RexA screen. (C) Several E. coli proteins gave high β-galactosidase activity with Ren protein, and here we focus on the lysine tRNA-ligase LysS. A second tRNA-ligase, HisS, was also identified in the Ren screen. Ren associates with itself, as shown in the last bar. BACTH, bacterial adenylate cyclase two-hybrid system; tRNA, transfer RNA.

Based on our protein-interaction results, we suggest a model to explain RexAB-mediated cellular defects. We note that other models have been proposed.³⁰ We propose that concurrent expression of RexA and RexB biases the oxidation-reduction (redox) state of the cell toward reduction, which results in NAD⁺ limitation. NAD⁺ limitation, in turn, limits functional DNA ligase activity since DNA ligase requires NAD⁺ as a cofactor. Less available active DNA ligase results in replication and recombination defects. We also suggest that Ren protein may counter the effects of RexA and RexB by promoting thiolation of a lysine transfer RNA (tRNA), which is able to counteract a highly reduced redox state.³¹ Our model proposes that these λ functions affect cellular redox in opposing ways and that together they modulate the redox state of the bacterial cell to benefit the phage.

Materials and Methods

Media and bacterial strains

Bacterial strains used in the main paper are listed in Table 2, those used in Supplementary Data in Supplementary Table S1. All bacterial strains are derivatives of MG1655. General bacterial methods are as in Thomason et al.^5,10 Lysis curves were performed as in Thomason et al,¹⁰ and previous two-hybrid analyses are described in Thomason et al.^5,10 Graphpad Prism was used to graph data. Note that this program does not plot error bars that are smaller than the symbols used to indicate individual points. IRB approval was not needed for this work.

Table 2.

Escherichia coli Bacterial Strains

Bacteria	Relevant genotype	Source
BTH101	F-, cya-99 araD139 galE15 galK16 rpsL1 (StrR) hsdR2 mcrA1 mcrB1	Euromedex https://web.euromedex.com
LT351	MG1655 wild type E. coli K-12	B. Bochner
LT447	MG1655 λ(cI857)	Thomason et al10
LT732	MG1655 ΔlacI-kan luc-N pLoL rexB+ rexA+ cI857ind1 pRoR cro-lacZYA	Thomason et al10
LT1015	MG1655 recA<>cat	S. Austin
LT1651	LT732 (rexA rexB)<>cat	This work
LT1761	LT732 PBAD-ren	This work
LT1964	LT732 PLIT-10	Thomason et al10
LT2154	LT732 PLIT-10 rexA<>cat	This work
LT2155	LT732 rexB<>cat PLIT-10	This work
LT2228	MG1655 λ(cI857 PLIT-10)	Thomason et al10
LT2336	MG1655 λ(cI857 PLIT-10) PBAD-ren	This work
LT2439	LT732 PLIT-10 PBAD-ren	This work

UV survival on square plates

Overnight cultures were grown at 32°C from single colonies in L broth. Ten-fold serial dilutions of the cultures were made in M9 salts out to 10⁻⁵, and 10 μL of the undiluted cultures and the dilution series were serially spotted on each of 13 square L agar square petri dishes marked with a 6 × 6 grid (Carolina Biologicals https://www.carolina.com # 741470). The spots of bacteria were allowed to dry, and then each petri plate was exposed over time to UV irradiation in 5 s increments out to 60 s.

This method allowed a complete dilution series for each strain, with one petri plate per UV time point. An unirradiated control was included in the series. After UV exposure, the plates were wrapped in aluminum foil and incubated overnight at 32°C, until small colonies formed. Colonies in each square of the grid were counted, and titers were determined for each culture at each time point.

Expression of Ren from P_BAD

The λ ren gene was inserted under control of the arabinose operon P_BAD promoter as described in Li et al.³² To examine the effects of ren expression on bacterial sensitivity to UV irradiation (Fig. 2B, C), overnight cultures were grown in 0.2% arabinose. Arabinose was also added to the square petri plates used to determine UV survival for ren-expressing strains. For lysis curves (Fig. 3), on the day of the experiment fresh overnight cultures grown in L broth were diluted into L broth containing 0.2% arabinose to express ren.

Two-hybrid library screen

High copy Amp^R plasmids pUT18 and pUT18C containing the rexA, rexB,¹⁰ and ren⁵ genes expressed from P_LAC were transformed into BTH101 using Amp (100 μg/mL) for selection. Next, 12.5 ng of a BACTH E. coli library cloned into the Kan^R pKT25 vector^27–29 was introduced into each of the three resulting Amp^R strains by electroporation. After transformation, cultures were outgrown for 2–3 h in 1 mL L broth containing 1 mM cAMP to stimulate plasmid P_LAC promoter expression. Cells were washed three times in M9 salts to remove cAMP before suspending in 1 mL M9 salts.

One hundred microliters of the washed culture was spread on M63 minimal agar containing 0.2% maltose, kanamycin (Kan) at 30 μg/mL, ampicillin (Amp) at 50 μg/mL, isopropyl β-D-1-thiogalactopyranoside at 0.5 mM, and 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal) at 40 μg/mL. Plates were incubated at 32°C degrees until blue colonies appeared after 3–5 days, indicating the production of cAMP. Blue colonies were purified on the same solid medium and then patched in a grid on the same medium for maintenance.

The clones present in the cAMP positive E. coli library plasmids were sequenced to determine the E. coli genes expressed from P_LAC. Only plasmids with in-frame insertions were considered as possible candidates. Cultures of the candidates were grown in L broth containing Kan (30 μg/mL) and Amp (50 μg/mL) and assayed for β-galactosidase activity as previously described.^5,10 Since the analysis has no obvious internal control, we used negative control data from Thomason et al⁵ for comparison.

Results

λ RexAB expression confers UV sensitivity on E. coli

As first reported by Li⁸ and Fix et al,⁹ we found that concurrent expression of RexA and RexB in E. coli sensitizes the cells to UV irradiation (Fig. 2A). In strain LT732, the RexA, RexB, and the CI857 repressor proteins are expressed from the bacterial chromosome in single copy from the phage P_RM and P_LIT promoters (Fig. 1B), in the absence of any other phage functions. Strain LT1964 is identical to strain LT732 except for a mutation in the −10 region of the P_LIT promoter, thereby reducing the level of RexB relative to RexA in LT1964. This change in the ratio of the RexB and RexA proteins increases LT1964 sensitivity to UV-mediated DNA damage (Fig. 2A). Note that strains with null mutations in either or both rex genes have UV survival curves such as those of wild type E. coli (Fig. 2B). Thus, it is the relative change in RexA and RexB expression levels that causes increased UV sensitivity.

Phage λ Ren expression rescues Rex-mediated defects

We asked whether the ren gene function affects E. coli UV survival when expressed in single copy from the E. coli P_BAD promoter by the addition of arabinose (Fig. 2C, D). This level of Ren expression rescues UV sensitivity of strain LT732, which has normal regulation of rexA and rexB. Ren also largely restores normal UV resistance to strain LT1964 (Fig. 2D), which has the promoter mutation P_LIT-10, and as discussed, an abnormally low level of RexB relative to RexA.

We previously found that heat-induced cI857 P_LIT-10 lysogens undergo RexAB-dependent aberrant lysis and impaired adenosine triphosphate (ATP) generation¹⁰; Ren protein reverses the lysis phenotype (Fig. 3) and restores normal phage yield. Our observations are consistent with those of Gauss et al,³³ who showed that strong Ren expression allows T4 rII deletion phages to grow and form plaques on Rex⁺ lysogens. Thus, Ren expression can generally counteract phenotypes caused by RexAB.

BACTH two-hybrid E. coli library screen with phage λ proteins RexA, RexB, and Ren

We previously showed that λ RexA and RexB interact with Ren, as does the phage replication protein O, by using the BACTH two-hybrid protein-protein interaction system.⁵ Our observations and those of Toothman and Herskowitz^18–20 suggested an involvement of the RexA, RexB, and Ren proteins in phage DNA replication. To elucidate the cellular targets of these phage proteins, we used an existing E. coli two-hybrid library^29,34 and screened for host proteins that are able to interact with RexA, RexB, or Ren.

In the BACTH system, candidate cells containing interacting clones express cAMP, allowing their growth and identification on minimal maltose selective agar, as described in Materials and Methods. Once interacting plasmid pairs were genetically identified, β-galactosidase activity was measured for each pair^5,35 (Fig. 4; Supplementary Table S2A, B). Since cAMP production is required for growth on minimal maltose, and cAMP is only made if the proteins expressed from any two compatible plasmids interact, all colonies identified in the selective screen should contain interacting clonal functions.

However, only a fraction of the host plasmids we isolated by this sensitive method gave activity higher than background when paired with plasmids expressing the phage functions (Fig. 4; Supplementary Table S2A, B). Thus, we have focused on the stronger interactors. Other host proteins, which gave background levels of β-galactosidase activity, are shown but are of unknown significance. Since we did not construct all possible BACTH plasmids for each E. coli protein,^27,28 we could have missed some interactions.^5,27 The host proteins identified in the BACTH screen that display the tightest interactions with RexB, RexA, and Ren are discussed below.

• 1.RexB (Fig. 4A): The inner membrane protein RexB interacts with NuoM, a subunit of the NADH dehydrogenase I (NDH-1). NDH-1 is the first complex in the electron transport chain and composed of 13 different proteins. The complex converts nicotinamide adenine dinucleotide from its reduced form (NADH) to its oxidized form (NAD⁺) while shuttling protons from the cytoplasm to the periplasmic space. NuoM is part of the inner membrane component of NDH-1 and may function in proton translocation.³⁶ Another subunit of NDH-1, NuoI, was a weak interactor in the RexA screen (Fig. 4B). The cytoplasmic subunit NuoI contains iron-sulfur (Fe-S) clusters involved in electron conduction.³⁷ RexB also interacts with two transporters, an acetate permease, ActP, and a ribonucleoside efflux pump, NepI, and possibly with the inner membrane protein HflC. It should be noted that although these are the strongest interactions that we identified between RexB and the E. coli proteins expressed in the genomic library, they are substantially weaker than the previously observed interactions between RexB and the λ CI repressor protein⁵ and the replication protein O.⁵ They are also weaker than the RexA and Ren interactions with host proteins that are shown in Figure 4. Taken at face value, these results suggest that RexB interacts most strongly with other phage functions and associates more loosely with the E. coli proteins it interacts with.
• 2.RexA (Fig. 4B): The cytoplasmic protein RexA interacts with the sulfur metabolism protein, CysN, sulfate adenylyltransferase, which is involved in sulfate activation and when complexed with CysD, converts intracellular sulfate to adenosine 5′ phosphosulfate (APS).³⁸ As noted earlier, the NuoI subunit of NDH-1 was also identified in this screen, but the β-galactosidase activity for the RexA-NuoI pair was low.
• 3.Ren (Fig. 4C): λ Ren protein interacts with the constitutive lysine-tRNA ligase LysS,³⁹ and weakly with a second tRNA ligase, HisS. It also interacts with an uncharacterized conserved protein, YcbX, which contains a sulfur carrier MOCO sulfurase C-terminal domain and Fe-S domains. YcbX functions as a partner with CysJ,⁴⁰ a flavin reductase, and interacts with CysI, and NuoC (also a subunit of NDH-1). Ren also interacts with OPP, the product of the ispB gene, which is involved in ubiquinone synthesis.⁴¹ Ubiquinones are essential components of NDH-1 and the electron transport chain. Ren also associates with MtlA, the mannitol-specific phosphotransferase system (PTS) enzyme II. MtlA transports and phosphorylates mannitol into the cell in preparation for utilization in glycolysis. Mannitol-1-phosphate is oxidized to fructose-6-phosphate by a NAD⁺-dependent enzyme, MtlD. Like RexA or RexB¹⁰ Ren also interacts with itself (Fig. 4C).

Supplemental Results

Other results contained in the Supplemental Data are mentioned here. Supplementary Table S2A and B contain numerical data for the β-galactosidase assays presented in Figure 4. Supplementary Figure S1A and B show that expression of RexAB does not affect the basal level of an SOS reporter, sfi-lacZ, nor does it affect the behavior of the cellular DNA repair response (SOS response) to the DNA damaging agent Mitomycin C. Supplementary Figure S2 shows growth curves for strain LT357, a strain nearly isogenic to LT732, which expresses the phage immunity region (Fig. 2). LT357 differs from LT732 in that it contains several point mutations in the immunity region, including one resulting in a mutant RexA protein (T154A). LT357 displays growth defects on certain carbon sources similar to those of nuo mutants, which are defective in NDH-1 function.⁴² Supplementary Table S3 contains Biolog Phenotypic Microarray data for LT357, showing that it produces more NADH than an isogenic strain not expressing RexA and RexB. Supplementary Table S4 shows that LT357 produces a high level of extracellular acetate when grown in tryptone broth.

Discussion

The decreased survival after exposure to UV light caused by λ RexAB expression (Fig. 2) could be interpreted as concurrent RexA and RexB expression either causing DNA damage or inhibiting some aspect of DNA repair. We favor the latter interpretation, since we found that RexAB function does not activate expression of an SOS reporter (Supplementary Fig. S1) either in the absence of the DNA damaging agent Mitomycin C or when cells were exposed to the drug. Cells that fail to repair DNA damage also experience RexAB-dependent energetic defects, as illustrated by the defective heat induction of a cI857 P_LIT-10 lysogen,¹⁰ also see Figure 3.

When the cI857 P_LIT-10 lysogen is induced, cellular ATP levels fall,¹⁰ as if ATP generation by the proton motive force is compromised. Our results are consistent with published data^6,7 showing that the RexAB system impacts cellular energetics. NDH-1 oxidizes NADH to NAD⁺, coupling this reaction to an electrochemical proton gradient that generates ATP.⁴³ The λ Rex proteins may at times interfere with normal NDH-1 function and thus ATP production, perhaps while acting as an energy sensor for the phage. Under some conditions, RexAB expression causes phenotypes like those of nuo mutants, which have defects in NDH-1⁴² (Supplementary Fig. S2; Supplementary Tables S3 and S4).

Based on our observations and those of others, we suggest a model for the failure of phage T4rII mutants to grow in RexAB expressing strains. Colowick and Colowick¹⁷ proposed that T4 RIIA and RIIB functions provide energy during glycolysis, and that T4 rII mutants would lower energy and ATP levels. Since T4 DNA ligase, gp30, uses ATP as a cofactor,⁴⁴ lower ATP levels would limit ligase activity, and T4 rII mutant phage will rely more on host E. coli DNA ligase. This would not be a problem if there is adequate E. coli DNA ligase. Whereas T4 lig mutants are defective in phage growth,^45,46 T4 rII lig double mutants can grow but rely on E. coli DNA ligase.^46,47 These observations suggest that when the RII proteins are present, they constrain the T4 phage replication machinery to use the T4 phage DNA ligase, gp30. E. coli DNA ligase requires NAD⁺ as a cofactor⁴⁴; the E. coli ligase cannot use NADH.⁴⁸ RexAB expression likely increases NADH levels at the expense of NAD⁺, and NAD⁺ becomes limiting. When NAD⁺ is limiting, E. coli DNA ligase activity may then become limiting, since it requires NAD⁺. If T4 rII mutant phage cannot generate adequate ATP, the activity of the T4 ATP-dependent DNA ligase may be compromised. Similarly, when RexAB is present, and when NAD⁺ is limiting, functional E. coli DNA ligase activity will be in short supply. Under these limiting ligase conditions, T4rII mutants growing on RexA⁺B⁺ λ lysogens would become defective for DNA replication and recombination. Reduced levels of NAD⁺-dependent E. coli DNA ligase activity would also account for the Toothman and Herskowitz^18–20 observation that RexAB excludes red and ren mutants of heteroimmune λ. Reduced DNA ligase activity would also explain the defect in survival of UV irradiated cells expressing RexA and RexB that we (Fig. 2) and others^8,9 have observed.

The λ Ren protein rescues both the DNA damage and lysis problems caused by RexA and RexB (Figs. 2 and 3). One explanation for this result is that rescue occurs through protein-protein interactions, that is, Ren binds to either RexA, RexB, or both proteins to sequester them and reduce their activity. Alternatively, Ren may counteract the effects of Rex activity through some activity of its own. If so, Ren must affect E. coli physiology rather than acting specifically on some aspect of phage λ development such as DNA replication, since Ren can rescue the Rex-dependent defect (Fig. 2) in bacterial strains expressing only RexA, RexB, and CI repressor (the construct shown in Fig. 1B).

We were not surprised to see the NDH-1 proteins turn up in our two-hybrid screen for E. coli proteins that interact with the RexAB system, since we know that Rex inhibits cellular energetics.^6,12,13 We were somewhat surprised to find a strong interaction between RexA and the sulfur assimilation enzyme CysN. This CysN-RexA interaction prompted us to review what is known about phage λ and sulfur metabolism (Fig. 5). E. coli uses sulfur for the synthesis of the sulfur-containing amino acids, cysteine and methionine, as well as for sulfur-containing coenzymes and prosthetic groups, such as the Fe-S clusters involved in redox reactions.⁴⁹

FIG. 5.

Simplified diagram of the Escherichia coli sulfur assimilation pathway. Extracellular sulfate comes in via a permease. The next step is sulfate activation by CysN and CysD, to make APS. Our two-hybrid analysis show that RexA interacts with CysN. The next step in the pathway is phosphorylation of APS by a kinase, CysC, to form PAPS, which is then reduced to sulfite by a PAPS reductase, CysH. Phage λ encodes a PAPS reductase, NinC, which is nearly identical to the E. coli enzyme. In the next two steps, sulfite is reduced to sulfide, which is then converted to cysteine. Cysteine is used by IscS, cysteine desulfurase, as a sulfur donor for either tRNA modification or iron-sulfur [Fe-S] cluster biosynthesis,^55,56 as discussed in the text. Diagram modified from.⁴⁹ ADP, adenosine diphosphate; ATP, adenosine triphosphate; PAPS, 3’-phospho-adenylylsulfate reductase.

Interestingly, the phage λ ninC gene encodes a sulfur metabolism enzyme, a 3′-phosphoadenosine-5′-phosphosulfate (PAPS) reductase.⁵⁰ The λ phage PAPS reductase aligns directly with 80% identity to the MG1655 PAPS reductase, according to National Center for Biotechnology Information basic local alignment search tool.⁵¹ The reason phage λ encodes a PAPS reductase is not understood, but its presence suggests that sulfur is important for λ growth and development. Perhaps the phage has evolved a pathway to scavenge and sequester sulfur for virus production, and if so, RexA may affect that pathway.

The importance of sulfur metabolism for λ was previously noted by Maynard et al.^52,53 They performed a large-scale screen of the Keio collection of E. coli mutants⁵⁴ and found that E. coli sulfur metabolism and tRNA thiolation pathways had competing effects on λ development. For both pathways, sulfur is sequestered by the E. coli cysteine desulfurase, IscS. It is then used for either tRNA modifications or for iron-sulfur [Fe-S] cluster biosynthesis.^55,56 They found that the [Fe-S] dependent pathway inhibits phage growth, whereas the [Fe-S] independent pathway enhances viral production. They demonstrated that the [Fe-S] independent pathway is involved in programmed ribosomal frameshifting, which is important in generating the λ tail proteins gpG and gpGT. Ribosomal frameshifting at a slippery sequence to form the frameshifted gpGT is essential for phage λ development.⁵⁷ However, although both gpG and gpGT are needed for proper tail assembly, the gpG/gpGT ratio must remain high, and too much frameshifting to generate gpGT is detrimental for phage production. Maynard et al.^52,53 found that specific thiolation of a lysine tRNA at position 34 inhibits this programmed ribosomal frameshifting. Thus, the level of lysine tRNA thiolation modulates the ratio of the two proteins, gpG to gpGT.

On λ induction, host-encoded lysine tRNA is increased three-fold relative to uninduced E. coli levels.⁵⁸ In addition, phage λ induction results in the production of a novel thiolated tRNA.⁵⁹ Since the λ genome does not encode any tRNA,^58,60 we believe this is a host-encoded tRNA modified by the phage. Considering these results, we note that the constitutive lysine-tRNA ligase, LysS, was identified in our two-hybrid screen as interacting with Ren protein. Nakayashiki et al³¹ found that the E. coli intracellular redox state is biased toward the reduced state in E. coli mutants defective for tRNA thiolation.

If Ren promotes tRNA thiolation, it could help reverse the highly reduced cellular redox state that we propose RexAB causes. This reversal could increase NAD⁺ levels and thereby DNA ligase activity. As noted earlier, Gauss et al³³ found that Ren expression from the λ P_L promoter rescues Rex exclusion of T4 rII deletion phages. The P_BAD-ren expression system described here does not allow T4 rII deletion phage to grow and form plaques on RexA⁺B⁺ hosts, probably because the E. coli P_BAD promoter is much weaker than the strong λ P_L promoter.

NAD⁺ limitation could have other consequences for Rex⁺ lysogens. Another NAD⁺-dependent enzyme in E. coli is the sirtuin CobB,^61,62 a lysine deacetylase. Known CobB substrates include CheY and the acetyl-coA-synthase Acs. RexAB-expressing cells may be CobB mutant phenocopies, since under some conditions they display perturbations in acetate metabolism (Supplemental Results section and Supplementary Table S3 in Supplementary Data). Perhaps related to this, CobB and other sirtuins are antiviral restriction factors,⁶³ and a null mutation of cobB significantly improves phage λ growth in those cells.

We previously reported⁵ that RexA and RexB expression affects induction of the λ prophage, with RexA promoting lytic growth, and RexB modulating RexA activity. How do these previous observations mesh with the results presented here? The obvious difference is that effects of RexA and RexB on prophage induction do not require expression of both proteins, in contrast to Rex-mediated inhibition of DNA repair, which does. RexA alone destabilizes repression of the lytic promoters, either by binding DNA, by direct contact with CI repressor,⁵ or by some combination of both. Interaction between RexA and RexB¹⁰ and their cellular targets may cause some new activity unrelated to prophage induction.

Our model, that RexB-mediated inhibition of NDH-1 limits NAD⁺ levels and thus DNA ligase activity, does not identify a role for RexA, yet RexA is essential for the DNA repair problem to occur. As noted, screening the BACTH E. coli library with RexA protein identified a second NDH-1 subunit, NuoI, which may interact with RexA, although the average β-galactosidase measurement for interaction between RexA and NuoI proteins was not significantly higher than background (Fig. 4B). However, we were unable to test all possible plasmid combinations^27,28 in our two-hybrid survey of E. coli proteins. Perhaps changing the location of the cyclase tag on the two-hybrid plasmids or putting the E. coli nuoI gene on the high copy vectors would allow better interaction between RexA and NuoI and yield significant β-galactosidase data. This is plausible, since we have previously observed that the cyclase tag orientations and choice of plasmid vector can affect the level of β-galactosidase signal for the same protein pair.^5,10 Thus, it remains a possibility that both Rex proteins interact with NDH-1, which would better account for the essential role of RexA in the DNA repair phenotype.

Another possibility to explain the essential role of RexA in limiting DNA repair is based on the notion that RexA may scavenge sulfur from bacterial components and recycle it for phage use. Sulfur scavenging might reduce bacterial DNA repair through NAD⁺ limitation by damaging NDH-1 function, since this respiratory complex contains as many as 10 Fe-S complexes.⁶⁴ F-S clusters are also necessary for the recently characterized DNA charge transport-mediated DNA repair,⁶⁵ and their removal could directly inhibit this type of DNA repair.

Our BACTH analysis also shows an interaction between λ Ren protein and MtlA, the mannitol PTS permease. MtlA takes up extracellular mannitol and phosphorylates it in preparation for utilization in glycolysis. Mannitol-1-phosphate is oxidized to fructose-6-phosphate by an NAD⁺-dependent enzyme, MtlD. If NAD⁺ levels are low, carbon flow through glycolysis could be compromised. If Ren provides more NAD⁺ to this sugar uptake system, it might stimulate growth of induced lysogens on mannitol. Like Ren, MtlA is reported to interact with LysS.⁶⁶

Based on our experiments and those of others, we hypothesize that the Rex system inhibits the activity of the first complex in the electron transport chain, NDH-1, biasing cellular redox potential toward the reduced state and lowering the available level of NAD⁺, the essential cofactor for DNA ligase. NAD⁺ limitation would result in the observed DNA repair defects and exclusion of recombination-dependent phages, that is, T4 rII and λ red, ren, or gam mutants. NAD⁺ limitation would also allow λ to inhibit the NAD⁺-dependent anti-phage sirtuin, CobB.⁶³ In the context of our model, which is diagrammed in Figure 6, the ability of λ Ren protein to rescue RexAB-mediated inhibition of DNA repair and abnormal cell lysis suggests that Ren expression counteracts this highly reduced cellular state, perhaps by stimulating lysine tRNA thiolation as suggested by our two-hybrid results. When the RexA, RexB, and Ren functions are present at optimal levels, their combined activities may allow λ to co-opt bacterial energy circuitry for phage use and lytic growth; a similar idea was proposed by Campbell and Rolfe.⁶⁷ Healthy phage growth apparently requires a delicate balance of these proteins. Our two-hybrid observation that RexA interacts with CysN strengthens known connections between phage λ and sulfur metabolism.^52,53 Perhaps λ Rex and Ren proteins cooperate in sulfur scavenging, redirecting it for phage use. These proposed phage-directed activities may occur at the expense of long-term bacterial well-being, but this is of little consequence for λ, since the bacterial cell has become a receptacle for phage reproduction.

FIG. 6.

Schematic of model explaining the effects of RexA, RexB, and Ren on the redox state of the host cell. The reaction performed by the NDH-1 is shown vertically at the center of the diagram. NDH-1 is the first enzyme complex in the electron transport chain; it produces NAD⁺, protons (H⁺), and free electrons (e⁻) from NADH formed during glycolysis. This complicated 13-subunit enzyme performs a dehydrogenation reaction while transferring electrons to cytoplasmic quinones and generating a proton electrochemical gradient (PMF) across the cytoplasmic membrane. The NAD⁺ produced during the reaction serves as an essential cofactor for several enzymes, including DNA ligase and the deacetylase CobB. Our BACTH analysis detected interactions between the phage λ RexA and RexB proteins and two different NDH-1 subunits (Fig. 4; Supplementary Table S2A, B). In our model, RexA and RexB (shown on the left side of the diagram) interfere with NDH-1 function, resulting in lower levels of NAD⁺ and thus reducing the activity of NAD⁺-dependent enzymes. According to the model, a reduction in functional DNA ligase causes an increased RexAB-mediated sensitivity to DNA damage (Fig. 2). Similarly, the RexAB-dependent defects in acetate metabolism described in the Supplemental Data (Supplementary Fig. S2) are due to CobB inhibition. The phage λ Ren protein can reverse the RexAB-dependent DNA damage repair phenotype (Fig. 2). Our BATCH analysis shows that Ren interacts with LysS, the constitutive lysine-tRNA. Thiolation of lysine tRNA is important for phage λ,⁵³ and defects in tRNA thiolation bias the redox state of the cell toward reduction.³¹ As illustrated at the right side of the diagram, the model proposes that Ren acts by promoting tRNA thiolation, which counteracts the detrimental effects of RexAB by promoting a more oxidizing cellular redox state. Thus, this interplay of the three λ proteins, RexA, RexB, and Ren, can modulate intracellular redox conditions. NDH-1, NADH dehydrogenase I; PMF, proton motive force.

Our study has not provided us with definitive answers to unanswered questions, and since we are now retired, we have no possibility of obtaining them. We hope that our results will spur further exploration of the tantalizing physiological interactions between phage λ and its host E. coli as well as the phage-specific interference imposed by the λ RexA and RexB functions.

Authors' Contributions

L.C.T.: Conceptualization (lead), investigation, writing - original draft (lead), review and editing (equal). D.L.C.: Conceptualization, funding acquisition, review and editing (equal).

Author Disclosure Statement

The authors confess to no conflict of interest.

Funding Information

This project has been funded in whole or in part with Federal funds from the National Cancer Institute, National Institutes of Health, Department of Health and Human Services, under Contract No. 75N91019D00024.

Supplementary Material

Supplementary Data

Supplementary Figure S1

Supplementary Figure S2

Supplementary Table S1

Supplementary Table S2

Supplementary Table S3

Supplementary Table S4