Revisiting the Lysogenization Control of Bacteriophage λ

Upon infection to the Escherichia coli cell, the genome of bacteriophage λ either replicates to form new progenies (lytic growth) or integrates into the host chromosome (lysogenization). The λ CII protein is a key determinant in the lysis-lysogeny decision. It is a short-lived transcription activator for the λ genes essential for lysogeny establishment. In this study, we isolated a new class of hfl (high frequency lysogenization) mutants of E. coli, using a new selection for enhancement of CII-stimulated transcription. The gene affected was termed hflD, which encodes a protein of 213 amino acids. AnhflD-disrupted mutant indeed showed an Hfl phenotype, indicating that HflD acts to down-regulate lysogenization. HflD is associated peripherally with the cytoplasmic membrane. Its interaction with CII was demonstrated in vitro using purified proteins as well as in vivo using the bacterial two-hybrid system. Pulse-chase examinations demonstrated that the HflD function is required for the rapid in vivo degradation of CII, although it interfered with FtsH-mediated CII proteolysis in an in vitro reaction system using detergent-solubilized components. We suggest that HflD is a factor that sequesters CII from the target promoters and recruits it to the membrane where the FtsH protease is localized.

Upon infection to the Escherichia coli cell, the genome of bacteriophage either replicates to form new progenies (lytic growth) or integrates into the host chromosome (lysogenization). The CII protein is a key determinant in the lysis-lysogeny decision. It is a short-lived transcription activator for the genes essential for lysogeny establishment. In this study, we isolated a new class of hfl (high frequency lysogenization) mutants of E. coli, using a new selection for enhancement of CIIstimulated transcription. The gene affected was termed hflD, which encodes a protein of 213 amino acids. An hflD-disrupted mutant indeed showed an Hfl phenotype, indicating that HflD acts to down-regulate lysogenization. HflD is associated peripherally with the cytoplasmic membrane. Its interaction with CII was demonstrated in vitro using purified proteins as well as in vivo using the bacterial two-hybrid system. Pulsechase examinations demonstrated that the HflD function is required for the rapid in vivo degradation of CII, although it interfered with FtsH-mediated CII proteolysis in an in vitro reaction system using detergent-solubilized components. We suggest that HflD is a factor that sequesters CII from the target promoters and recruits it to the membrane where the FtsH protease is localized.
When bacteriophage infects to the Escherichia coli cell, it undergoes either lytic growth or lysogenization (1). The establishment of lysogenization requires sufficient accumulation of the repressor protein, CI, before the lytic cycle is initiated. Transcription of the cI gene from the pRE promoter is positively regulated by the CII protein, which also activates the pI promoter for the int gene expression required for the prophage integration, as well as the pAQ promoter for the antisense RNA that inhibits the synthesis of the Q protein required for late lytic gene expression. Thus, increased activity of CII leads to preferential lysogenization.
It is known that the intracellular concentration of CII is controlled at the levels of not only transcription but also protein stability. CII is short-lived, having a half-life of about 2 min in -infected wild-type cells (2,3). Host hfl (high frequency lysogenization) mutations that stabilize CII have been studied. The mutation first isolated by Belfort and Wulff (4) and an additional five mutations isolated by Gautsch and Wulff (5) were mapped at the hflA locus of the chromosome, which was later shown to comprise three genes, hflX, hflK, and hflC (6). The hflA mutations affect either hflK or hflC (7,8), but the function of hflX is unknown (6). HflK and HflC are membrane proteins, forming a binary complex, HflKC (8 -10). Although HflKC was once believed to be a protease degrading CII (10), we showed that it is not a protease; rather, it associates with the true protease, FtsH (11)(12)(13), and somehow modulates the activity of the latter (11,12). Both HflK and HflC have a signal-anchor sequence at the N terminus, which is followed by the main body of protein exposed to the periplasmic space (8,12).
The remaining mutation (hflB29) of Gautsch and Wulff (5) defined the hflB gene (14), now known to be identical with ftsH (15). Its product acts as the primary proteolytic enzyme against CII (12,13). FtsH is a zinc metalloprotease, which is membrane-bound and ATP-dependent (reviewed in Ref. 16). Its cytoplasmic domain includes an evolutionarily conserved AAA ATPase domain (reviewed in Ref. 17). Proteolytic substrates of this enzyme include both soluble and membrane proteins. The actions of FtsH against soluble and membrane-bound substrates are differentially affected by the HflKC complex (18). Some substrates may require additional factors for optimal proteolysis. For instance, 32 requires the DnaK chaperone for its degradation in vivo (19). Thus, FtsH may be subject to modulation by multiple factors that interact either with the substrates or FtsH itself. In this study, we addressed the question of whether any additional host factors exist that affect the CII protein with respect to its stability or function. A new factor, termed HflD, was thus identified as a negative regulator of lysogenization. HflD is a peripheral membrane protein, which interacts with CII. It is positively required for the degradation of CII in vivo. We propose that HflD holds CII on the membrane surface, away from the target promoters but close to FtsH.
Construction of the hflD::tet Strain-A blunt-ended 1.5-kb XbaI-AvaI tet fragment of pACYC184 was inserted into hflD (BseRI site) within the 7.2-kb SalI fragment, originally from pKH422 but now on pTWV228 (this plasmid was named pKH443). The 7.4-kb PvuII-XhoI fragment of pKH443 was then introduced into FS1576 (recD Ϫ ; Ref. 24), and tetracycline (6.25 g/ml)-resistant recombinants were selected. One of them was confirmed for the disrupted hflD gene by polymerase chain reaction, and its hflD::tet marker was P1-transduced into AD16, yielding AK2149.
Other Methods and Materials-Hfl phenotypes were assessed by measuring lysogenization frequency of ϩ (12) as well as by examining bacterial growth after infection with c17 phage (12). Transduction using P1vir was done by the established procedures (22). Pulse-chase and immunoprecipitation (28), immunoblotting (8), and cross-linking (11) were done essentially as described previously. Rabbit antiserum against HflD was prepared using a synthetic peptide for residues 76 -91 and affinity-purified using the immobilized peptide. Anti-CII serum was prepared using a synthetic peptide for residues 77-93. Antisera against a proton-ATPase subunit, F 0 a (29), and GroEL (30) were described previously.

Isolation of a New Class of Mutations that Elevates CII-dependent
Transcription-In the original mutant isolation, only one hfl allele (hflB29) was isolated at the hflB locus (4, 5), raising a question as to whether genes involved in lysogenization control have been all identified. We revisited this question using a "modernized" approach. A plasmid (pKH394) was constructed in which the tetracycline resistance gene (tet) was placed under the control of a CII-controlled promoter, pI (Fig.  1A). This plasmid also contained the cII gene itself and the lacZ␣ gene, which are expressed by read-through transcription from the chloramphenicol resistance gene (cat). The cII gene contained the silent cy42 base substitution (31), which reduces the CII binding to the reverse-oriented pRE promoter within cII (31), thus minimizing complication due to the presence of two CII-responding promoters of different orientations. Model experiments showed that the zgj-525::IS1A mutation, which reduces the expression level of ftsH (21), indeed conferred tetracycline resistance to the pKH394-bearing cells (Fig. 1B).
To isolate a new class of hfl mutations, pKH394 was introduced into mutagenized cells, and tetracycline-resistant mutants were selected. Among 15 mutants, three had mutations at 69 min, where ftsH is located, and five had mutations at 95 min, where hflK-hflC are located. The remaining seven muta- tions, named hflD33, hflD11, hflD13, hflD24, hflD28, hflD45, and hflD48, were examined further. After mutant cells had been cured of the plasmids, they were examined for an Hfl phenotype. They indeed failed to support growth of c17 mutant phage (14).
Identification of hflD-To map one of the mutations, hflD33, we isolated a Tn5 insertion (termed zcg-2002::Tn5) that was P1 co-transducible with hflD33 at a frequency of about 50% (Fig.  2). We then cloned a chromosomal segment together with the kan region of the Tn5 insertion. Sequence analysis of one such plasmid (pKH402) showed that the Tn5 insertion was located at 26.2 min on the E. coli genome. In P1 transduction, zcg-2002::Tn5 was co-transducible with dsbB::cat (located at 26.6 min; 32) at a frequency of about 40%, but hflD33 was not. Thus, the order of hflD33-zcg-2002::Tn5-dsbB was suggested. Although some E. coli K-12 strains possess a defective prophage e14 (33), in this chromosomal region (Fig. 2), the strains we used in this study did not contain e14 (data not shown).
The remaining six hfl mutations also proved to be P1 cotransducible with zcg-2002::Tn5. We determined the nucleotide sequence for each of the mutant hflD genes ( Table I). Four of them (hflD24, hflD28, hflD45, and hflD48) contained a nonsense mutation, whereas two (hflD33 and hflD11) contained a missense mutation within the hflD open reading frame. The remaining mutant (hflD13) had a base change in the putative Shine-Dalgarno sequence for hflD. We thus identified hflD as a new gene affecting the CII function of bacteriophage .
HflD Down-regulates Lysogenization-The hflD gene is expected to encode a protein of 213 amino acids (35). Its homologs exist in several bacterial species including Hemophilus influenzae, Vivrio cholerae, and Pseudomonas aeruginosa (Gen-Bank TM accession numbers, respectively, I64155, B82237, and D83317). No physiological role has been assigned for this dispensable gene (35). We also constructed a strain in which chromosomal hflD was disrupted by tet. No obvious growth phenotype was observed for any of the hflD mutants without infection. Four non-sense (hflD24, hflD28, hflD45, and hflD48) and one missense (hflD33) mutation as well as the tet disruption of hflD increased the lysogenization frequency of by 25-50-fold (Table I, Experiment I), whereas zgi-525::IS1A did so by about 100-fold (Table I, Experiment II). One missense mutation (hflD11) and the noncoding region mutation (hflD13) gave ϳ5-fold increases in the lysogenization frequency. When hflD was overexpressed from a plasmid in a wild-type strain, the lysogenization frequency was lowered by 10-fold (Table I, Exp. III). These results indicate that the hflD gene product actively participates in keeping the lysogenization frequency low.
We then examined whether overexpression of hflD could suppress the Hfl phenotypes associated with other classes of hfl mutations. When a plasmid overexpressing hflD 15-30-fold was introduced into the hflD::tet, the zgj-525::IS1A and the ⌬hflK-hflC::kan strains, these bacteria were converted to be completely sensitive to c17 (data not shown). Thus, HflD overproduction channels the cell to a more lysis-oriented status.
Identification of the HflD Protein and Its Peripheral Association with the Membrane-The wild-type and the mutant hflD genes were expressed in the hflD::tet strain, and their products were examined by immunoblotting, using antiserum against an HflD synthetic peptide. A protein with an apparent molecular mass of 23 kDa was detected for the wild-type gene (Fig.  3A, lane 1). This product was missing for the non-sense mutants (Fig. 3A, lanes 4, 5, and 8) except for hflD45, which produced a smaller fragment (Fig. 3A, lane 7). The hflD33 missense mutation also gave a band of faster electrophoretic mobility (Fig. 3A, lane 6), suggesting its instability. The hflD13 mutation markedly lowered the expression level of HflD (Fig.  3A, lane 2), consistent with a lowered translation initiation. In wild-type cells without plasmid, HflD was detected as a faint band at the 23-kDa position (data not shown).
Upon cell fractionation, HflD was recovered from the membrane fraction (Fig. 3B, lane 3), which is consistent with a previous report (35). It was extractable with 0.1 M NaOH (Fig.  4B, lane 4), in contrast to the ATP synthase F 0 a subunit used as an integral membrane protein control (Fig. 3B, lane 5). HflD was inaccessible to externally added proteinase K (Fig. 3C, lane  2), unless the spheroplasts were broken by a detergent (Fig. 3C,  lane 3). Thus, HflD is peripherally associated with the cytosolic surface of the cytoplasmic membrane.
HflD Interacts with CII-We purified CII and HflD (Fig. 4A,  lanes 2 and 5). HflD was purified also as an N-terminal glutathione S-transferase (GST) fusion (Fig. 4A, lane 3), which was functional as its expression complemented the hflD::tet strain with respect to the Hfl phenotype (data not shown). No activity to degrade CII was detected for the purified HflD preparations, arguing against its being a CII-degrading protease. We failed to detect any interaction between HflD and FtsH (data not shown).
When GST-HflD and CII were mixed in vitro and subjected to affinity isolation using a glutathione-Sepharose column, they were co-eluted by 20 mM glutathione (Fig. 4B, lane 8). CII alone did not bind to the column (Fig. 4B, lane 2). Thus, HflD has an ability to bind to CII. Independent evidence for the HflD-CII interaction was obtained by cross-linking experiments. HflD was partially converted by treatment with 3, 3Јdithiobis-(succinimidyl) propionate, a primary amine-reactive homobifunctional cross-linker, to a form expected for a dimer (ϳ41 kDa; Fig. 5, lanes 4 and 5). When a mixture of HflD and CII was treated similarly, an additional product was observed at ϳ32 kDa (Fig. 5, lanes 8 -10). This new band most likely represented an HflD-CII cross-linked product, because it possessed both the HflD (Fig. 5) and the CII (data not shown) antigenicity. CII itself, which did not react with the anti-HflD, produced homodimeric, trimeric, and tetrameric products upon cross-linking (data not shown). A tetrameric state of CII was reported previously (37).
To examine whether HflD and CII interact mutually in vivo, the E. coli two-hybrid system (25) was used. CII was fused to the N-terminal domain of adenylate cyclase from Bordetella pertussis, and HflD was fused to the independently cloned C-terminal domain of this enzyme. In the presence of both   4 g (120 pmol), lanes 1-4) and a mixture of CII (1.4 g) and GST-HflD (11.7 g (240 pmol), lanes [5][6][7][8] were incubated in 300 l of 50 mM Tris⅐HCl (pH 8.1) containing 10% glycerol, 50 mM NaCl, 0.1% Nonidet P-40, and 1 mM DTT at 0°C for 30 min and applied to a glutathione-Sepharose column (load, lanes 1 and 5; flow-through, lanes 2 and 6). The column was washed (lanes 3 and 7) and eluted with 20 mM glutathinone in the same buffer (lanes 4 and 8). Proteins were separated by SDS-PAGE and stained with Coomassie Brilliant Blue. plasmids, the expression of lacZ, which is dependent on cyclic AMP, the product of adenylate cyclase, was increased about 3-fold over the control. This result suggests that HflD and CII interact with each other in vivo.
HflD Enhances CII Degradation in Vivo but Interferes with CII Degradation in Vitro-A purified preparation of FtsH-His 6 -Myc, in detergent-solubilized states, can degrade CII in the presence of ATP (Ref. 12; Fig. 6A, open circles). When increasing concentrations of HflD were added to the reaction, degradation of CII was increasingly inhibited (Fig. 6A). Thus, HflD binding to CII results in the protection of the latter from FtsH-mediated proteolysis. This observation, however, was in apparent contradiction with the in vivo ability of HflD to downregulate lysogenization.
We then examined the in vivo effects of the hflD mutations as well as of HflD overproduction on the stability of CII as it was expressed from a plasmid. CII was degraded with an initial half-life of about 3 min in wild-type cells (Fig. 6B, open squares). It was stabilized in the hflD::tet cells although not completely; the initial half-life was prolonged to ϳ10 min (Fig.  6B, closed squares). This residual degradation, observed in the absence of HflD, should have been FtsH-mediated, because loss-of-function mutations of ftsH effectively stabilizes CII (12,13). In contrast, overproduction of HflD destabilized CII; the initial half-life became ϳ1 min (Fig. 6B, open circles). These results indicate that HflD contributes positively to the degradation of CII in vivo.

DISCUSSION
The CII protein of offered a classical example in which stability control of a regulatory protein serves as a developmental switch. Several factors affect the stability of CII. These include FtsH (HflB), the primary enzyme responsible for proteolysis (12,13), and the HflKC complex, which modulates the proteolytic function of FtsH against different classes of substrates (11,12,18). The -encoded CIII protein also affects CII stability, as it interferes competitively with the rapid degradation of CII (3,14,39). In this study, we identified a new host element, HflD, which directly interacts with CII and functions to reduce the lysogenization frequency of .
Although HflD is not absolutely required for FtsH-mediated degradation of CII, it significantly stimulates the degradation in vivo (Fig. 6B). HflD is not itself a protease. It is not simply an activator of FtsH either, because its in vitro effect on the FtsH-dependent CII degradation is negative (Fig. 6A). The HflD effect on protein degradation is specific for CII, as in vivo degradation of 32 or SecY was not affected by the hflD disruption (data not shown). We have shown that HflD directly interacts with CII both in vivo and in vitro.
Under the in vitro reaction conditions, HflD-binding to CII resulted in an inhibition of CII degradation. In this regard HflD does not appear to be a specificity-enhancing factor like SspB (40). How can an in vivo stimulator act as an in vitro inhibitor? Although FtsH is membrane-associated in vivo, it was in the solubilized states in vitro. CII is predominantly DNA-bound in vivo (37), but no DNA was involved in the in vitro reaction. These differences between the in vivo and in vitro reaction conditions may explain the observed discrepancy. Thus, a conclusion that HflD⅐CII complex is a less favored substrate of FtsH than free CII holds only for the in vitro reactions, in which every component has been solubilized. We propose that HflD acts in vivo to change localization of CII from DNA to the membrane. Then, in the in vivo situation, the HflD⅐CII complex  1 and 6) or with 3,3Јdithiobis-(succinimidyl) propionate (DSP) at 3.1 g/ml (lanes 2 and 7), 6.3 g/ml (lanes 3 and 8), 12.5 g/ml (lanes 4 and 9), and 25.0 g/ml (lanes 5 and 10) at 4°C for 1 h. Proteins were separated by SDS-PAGE using the Weber and Osborn (38) system and visualized by anti-HflD immunoblotting.

FIG. 6. Contrasting effects of HflD on in vitro and in vivo degradation of CII.
A, effects of purified HflD on FtsH-mediated in vitro degradation of CII. Proteolysis of CII (1.86 g or 160 pmol) by detergent-solubilized and purified FtsH-His 6 -Myc (1.48 g or 20 pmol) was carried out as described previously (11) in a total volume of 60 l and in the presence of ATP (open circles). Added to the reaction mixture was HflD at 0.9 g (40 pmol; closed circles), 1.8 g (80 pmol; closed squares), 3.7 g (160 pmol; closed triangles), and 7.4 g (320 pmol; closed diamonds). Reaction was at 37°C for 0, 0.5, 1, and 2 h. CII in each sample was visualized by immunoblotting, and the relative intensities were determined by a lumino-image analyzer (LAS1000, Fuji Film). Values relative to the initial amount of CII are shown. B, stability of CII in cells with defective or overproduced HflD. Plasmid pKH479 expressing CII was introduced into strains AD16 (wild-type; open squares), AK2149 (hflD::tet; solid squares), AD16/pTWV228 (vector control; closed circles), and AD16/pKH441 (overexpressing hflD; open circles). Plasmid-encoded genes were induced at 37°C with 1 mM IPTG and 3 mM cyclic AMP for 10 min and pulse-labeled with [ 35 S]methionine for 30 s followed by chase for the indicated time periods. Radioactive CII was immunoprecipitated, separated by SDS-PAGE, and quantitated using a phosphorimager (BAS1800 , Fuji Film). Values relative to those at the 0 min chase point are shown for each sample. might be a better substrate than the DNA-bound form of CII. The reason for this may be 2-fold. First, HflD will bring CII to the membrane surface where the active site domain of the FtsH protease resides (41), increasing the frequency of the CII-FtsH contacts. Second, the DNA-bound state of CII may be less susceptible to proteolysis than the dissociated state even though the latter might always be in the HflD-bound form in vivo.
Although the affinity of CII to HflD seems to be higher than to FtsH under the in vitro conditions used, the relative affinities could change by modification such as phosphorylation of a component (42). It is tempting to speculate that HflD is a part of the host system that controls the activity and the level of CII in response to environmental conditions. It should also be noted that an additional factor could participate in vivo, in conjunction with HflD, to stimulate CII degradation, although we do not understand why such a factor, if it does exist, escaped our mutant selection.
In gel filtration in the presence of a detergent, HflD migrated in two peaks (data not shown). Cross-linking experiments identified a cross-linked dimer of HflD. Thus, HflD may partly be in a dimeric or larger homocomplex. Cross-linking in the presence of CII yielded an HflD⅐CII complex, whereas CII alone produced tetrameric homo-cross-linking (data not shown for the CII cross-linking). These results suggest that HflD is in equilibrium between the monomeric and oligomeric states and that the monomeric HflD may interact with CII. Although HflD is hydrophilic overall, some segments of it could form amphipathic ␣-helices. In particular, 8 leucines and 1 isoleucine can be aligned on one side of an ␣ helix for the Leu 85 -Leu 120 interval, and the amphipathic nature seems to continue up to Tyr 138 . These regions could mediate membrane association, CII association, and/or dimerization of HflD.
Although our results indicate that HflD enhances degradation of CII by bringing it to the vicinity of membrane-bound FtsH, it might participate in lysogeny control also by sequestering CII from the target promoter (43). In this latter respect, it could be regarded as similar to RseA, a membrane-bound anti-sigma E factor (44). The role played by HflD in the uninfected host cell is an interesting subject which will be left for future studies.