DNA Packaging Specificity of Bacteriophage N15 with an Excursion into the Genetics of a Cohesive End Mismatch

During DNA replication by the λ-like bacteriophages, immature concatemeric DNA is produced by rolling circle replication. The concatemers are processed into mature chromosomes with cohesive ends, and packaged into prohead shells, during virion assembly. Cohesive ends are generated by the viral enzyme terminase, which introduces staggered nicks at cos, an approx. 200 bp-long sequence containing subsites cosQ, cosN and cosB. Interactions of cos subsites of immature concatemeric DNA with terminase orchestrate DNA processing and packaging. To initiate DNA packaging, terminase interacts with cosB and nicks cosN. The cohesive ends of N15 DNA differ from those of λ at 2/12 positions. Genetic experiments show that phages with chromosomes containing mismatched cohesive ends are functional. In at least some infections, the cohesive end mismatch persists through cyclization and replication, so that progeny phages of both allelic types are produced in the infected cell. N15 possesses an asymmetric packaging specificity: N15 DNA is not packaged by phages λ or 21, but surprisingly, N15-specific terminase packages λ DNA. Implications for genetic interactions among λ-like bacteriophages are discussed.


Introduction
Large DNA viruses, such as tailed bacteriophages and herpes viruses, use an ATP-powered motor to translocate viral DNA into the preformed empty shell, called the prohead or procapsid [1][2][3]. Recent structural and bioinformatic studies demonstrate that the DNA packaging machinery of these viruses is descended from that of an ancient common ancestor [4]. For example, the prohead shell is an icosahedral lattice principally constructed of many copies of the major capsid protein whose fold is conserved [5]. Similarly, one of the prohead's 5-fold Packaging is initiated when terminase binds and nicks a cos on a concatemer. Following cosN nicking and cohesive end separation, terminase forms a tight complex, Complex I, on the cosB-containing chromosomal end [37,38]. Complex I then docks on the portal protein, gpB, of the prohead [39,40]. Following docking of Complex I on the portal, TerL λ 's ATPase is activated and ATP hydrolysis-powered translocation of the DNA into the shell ensues [41,42]. Terminase remains docked on the portal during translocation. When the next cos along the concatemer is encountered, terminase (1) nicks cosN, (2) dissociates from the portal, and (3) remains bound to the cosB-containing end of the next chromosome along the concatemer [43]. By remaining bound to the next chromosome along the concatemer as a Complex I, terminase processively packages downstream chromosomes. A third sub-site, cosQ, is essential for recognition and nicking of the downstream cos during termination of packaging [44,45]. In sum, cosN and cosB are required to initiate λDNA packaging, cosQ and cosN are required for termination, and all three subsites are required for processivity [46].
In this paper, we refer to the λ-like phages as those with λ-like cohesive ends, i.e., 12 baselong, 5' extensions with significant identity to λ's cohesive ends. This group includes λ, 434, ϕ80, 21, N15, Monarch, and gifsy-1 [54]. In the cosNs of this group, bp differences in the right half of the cohesive end sequences, i.e., bp 7-12, are common. For example N15's cohesive Fig 1. Comparison of the coses and terminase small subunit genes of λ, 21 and N15. A. Complex cos structure of phages λ, 21, Monarch. R3, R2 and R1 sequences are TerS binding sites, and I1 is an IHF binding site. B. Simple cos of N15 and relatives [58]. rR2 is a proposed accessory TerS binding site in the cos of N15 and relatives. ends differ from λ's at bp 9 and 12, and gifsy-1's differ at positions 8, 9, and 11. In contrast, the 7 bp cosQ site is highly conserved. The cosBs of the λ-like phages differ both in structural complexity and in packaging specificity, as follows.
With respect to cosB structure, DNA sequence analysis suggest that many λ-like phages have a complex cosB similar to cosB λ , with three R sequences with R2 oriented opposite to R3 and R1, and with a likely IHF binding site between R3 and R2. Recently, N15 was shown to have a different, simpler cosB consisting of a single critical TerS binding site located approximately at the position of the R3 in complex cosBs. A second, inverted sequence that closely matches R3 N15 , called rR2, is located within the TerS gene, about 200 bp distal to R3 N15 , and though not critical for DNA packaging, is proposed to play a non-essential, accessory role in DNA packaging. The R1 sequence of cosB λ plays such a role [55]. Examination of prophages in Genebank identified an additional five prophages that have the R3 and rR2 sequences found in N15's cosB [56]. In sum, two strikingly different cosB structures, complex and simple, are found in the λ-like phages.
In addition to structural differences, packaging specificity differences are found in the λ-like phages, as exemplified by the well-characterized specificities of λ and 21. λ and 21 have identical cosQ and cosN sequences, yet each package self-DNA >10 3 -fold more efficiently than that of the other phage. Examination of the recognition helix and cosB R site sequences shows significant differences that account for the specificity difference (Fig 2) [20,57]. Genetic analysis identified recognition helix:R site contacts that differ between λ and 21, plus clashes involved in discrimination by TerS λ and TerS 21 against the heterologous DNA [56]. Looking at TerS N15 and R3 N15 suggests that N15, and its relatives with simple cosBs might have a novel packaging specificity distinct from the λ and 21 specificities (Fig 2) [58]. Interestingly, a second group of N15-like prophages, exemplified by Monarch, have complex, tripartite cosBs. TerS Monarch and TerS N15 have identical recognition α-helixes and strong R site identity, suggesting that both subgroups share the same packaging specificity. Here we present a study of N15's packaging specificity. Because of the strategy used, a preliminary analysis of the genetics of phage chromosomes with mismatched cohesive ends, λ versus N15, was carried out.

Results
Genetics of a cohesive end mismatch: Phages with mismatched cohesive ends are viable; the mismatches can persist to post-DNA replication Strategy for studying DNA packaging specificity. To ask about the effects of cosN and cosB differences on DNA packaging, we used a helper phage/passive prophage approach (Fig  3). In these experiments, the DNA packaging substrate, a passive prophage, is provided to the DNA packaging machinery of a lytically growing "helper" phage, as follows. Two prophages integrated in tandem at attB contain between them a chromosome that mimics a chromosome in a concatemer. Terminase from a helper phage can initiate packaging at the cos of the upstream prophage, translocate DNA and terminate packaging at the cos of the downstream prophage. In our experiments, the helper phage was a heat-inducible (cI857) derivative of a λ strain that forms a plasmid prophage. Helpers with λ, 21 or N15 DNA packaging specificity were used. The 21-and N15-specific helpers were viable hybrid derivatives of λ in which a left chromosome end, the DNA segment including cosB λ and much of the TerS λ -encoding Nu1 gene, was replaced with the functionally analogous segment from phages 21 or N15, respectively. The 21-and N15-specific helpers are called λ21 hy51 [59] and λN15 hy4 [58], respectively. These helper phages are viable hybrid phages in which the left λ DNA end is replaced with the corresponding left end of the 21 and N15 chromosome, respectively. The left end segments, each about 600 bp in length, include 21's or N15's cosB and the 5' end of the TerS-encoding gene which contains the helix-turn-helix DNA binding specificity motif. The corresponding chimeric TerS proteins are called TerS hy51 and TerS hy4 , respectively. When a helper prophage is heat-induced to carry out lytic growth, the tandem prophages remain repressed because the prophages have thermo-stable imm 21 and imm 434 immunity repressors that differ in specificity from the imm λ helper. The maximum number of packageable passive prophages equals the number of bacterial chromosomes in the cell, so efficient packaging of a prophage gives a yield of 1-4 virions/infected cell; the helper yield was also low, at~10-20/induced lysogen.
Genetics of mismatched cohesive ends. We did preliminary helper packaging experiments asking if the helper packaging approach was appropriate for N15 specificity studies. First, using in vivo phage crossing techniques, we made λ att + N15 hy4 , phage ϕ1182 (Table 1),  [56]. The CG!TA change at R3 λ bp 58 (highlighted in green) renders λ IHF-dependent for plaque formation [49,70] [62]. B. Alignments of the left ends of TerS subunits. Proposed recognition helixes are highlighted in grey and residues 20 and 24 of TerS 21 and TerS λ , known to be involved in packaging recognition and discrimination, are highlighted in green. for dilysogen construction. ϕ1182 was used to lysogenize MF532, which contains a partial λ imm 434 indprophage which includes the prophage segment from attL through cos λ . MF532 (ϕ1182) contains the passive prophage structure: imm 21 cosQ N15 N N15 B N15 ! imm 434 ind -cosQ λ N λ B λ , where the arrow indicates the direction of DNA packaging. Packaging of the passive prophage by the λ N15 hy4 helper was expected to package a passive chromosome with mismatched cohesive ends. That is, initiation at the cos N15 of the upstream prophage and termination at the cos λ of the downstream prophage generates a helper-packaged chromosome with an N15 left cohesive end (designated as ceL N15 ) and a λ right cohesive end (ceR λ ; Fig 4). It was uncertain whether phages carrying the mismatched passive chromosome would be viable. Would the chromosome with a 2-bp cohesive end mismatch be able to circularize? Would the λ N15 hy4 terminase, after initiating packaging at the cos N15 of the upstream prophage, be able to terminate packaging at the cos λ of the downstream prophage? Although it was unknown whether N15-specific terminase would terminate packaging at cos λ , previous work showed that λ and 21 terminases could terminate packaging at either cos λ or cos 21 . In fact, previous work indicated that termination appears to not depend on the presence of cosB [46]. Another termination issue was whether the helper's N15-specific terminase would show preference for the downstream cosN sequence. Previous work demonstrated that TerL λ efficiently nicks a cosN with bp changes in the right half of the cohesive end sequence [60], indicating cosN specificity was also not a likely issue.  Packaging of MF532 (ϕ1182)'s passive prophage by the λN15 hy4 helper produced a modest yield (0.1 phages/induced cell) bearing passive prophages with mismatched cohesive ends. This yield is about 10% of the yield one might expect. The low yield of this ceL N15 ! ceR λ chromosome might be due to inefficiency at any of the stages of DNA packaging, i.e., initiation, termination, or reduced viability of the virions carrying chromosomes with mismatched cohesive ends, or the infirm recA -uvrBhost cells. Plaque-formation by ceL N15 /ceR λ mismatch phages indicates that the cohesive end mismatch is not lethal, i.e., the mismatch does not block cyclization of the injected DNA. Finding that phages with mismatched cohesive ends are viable indicated that none of the confounding possibilities prevented packaging of mismatched chromosomes into virions that were competent to carry out successful infections. Below we report experiments directly testing both cosN specificity and termination efficiency of λand N15-specific terminases at heterologous coses.
The cyclized DNA of the mismatch phages from MF532 (ϕ1182) contains a cosN with mismatches at positions 9 and 12 ( Fig 4A). The mismatched bp might or might not be subject to repair prior to DNA replication. If the mismatches are repaired, then all the progeny of an infection by a mismatch phage will either be cosN λ or cosN N15 , depending on the direction of the repair process. If the cosN mismatch is not repaired, then a mixed burst of both cosN λ and cosN N15 progeny phages are expected from an infected cell. To ask if the mismatched bp were or were not repaired, four well-isolated plaques produced by mismatch phages were suspended in buffer, replated, and individual plaques were subjected to DNA sequence analysis to Helper phage with λ packaging specificity [55] λ-P1:5R Kn R 21 hy51 cI857 nin5 = ϕ1081 Helper phage with 21 packaging specificity [64] λ-P1:5R Kn R N15 hy4 cI857 nin5 = ϕ1187 Helper phage with N15 packaging specificity [58] λ-P1:5R Kn R cI857 nin5 Nu1ms1 = ϕ1082 Helper phage [55,62] λ-P1:5R Kn R cI857 nin5 cos2 cosN deletion phage [55] λ imm 21 = ϕ72 cos λ / dilysogen constructions Feiss collection Dilysogen constructions This work determine the cohesive end sequence. We found that in the progeny of two of the four infections by mismatch phages, both cosN λ and cosN N15 phages were present ( Table 2). We conclude that in at least some infections, the mismatch is not repaired and DNA replication produces a progeny phage burst in which some are cosN λ and some are cosN N15 . In practical terms, the MF532 (ϕ1182) helper packaging produced phages that were imm 434 ind -cosQ λ cosB N15 and either cosN λ (ϕ1197) or cosN N15 (ϕ1199). A second dilysogen, MF713 (ϕ1199), was constructed. MF711 contains a partial λ b511 prophage containing the prophage segment extending rightward from cos λ . MF713 (ϕ1199) has the prophage structure: cosQ λ N λ B λ !cosQ λ N N15 B N15 . Packaging MF713 (ϕ1199)'s passive prophage by the λ helper is expected to produce phages bearing chromosomes with the reciprocal mismatch, i.e., a ceL λ /ceR N15 mismatch (Fig 4B). Using the λ-specific helper, the yield was 0.67 phages bearing passive prophage chromosomes/induced cell, indicating that the reciprocal cohesive end mismatch is also not lethal. To ask about persistence of the mismatch, the progeny of five of these ceL λ /ceR N15 mismatch phages were examined. Four isolates produced only progeny with cosN λ , and a fifth isolate produced a burst containing both cosN λ and cosN N15 phages ( Table 2). As was the case for ceL N15 /ceR λ mismatch phages, some ceL λ /ceR N15 mismatch phages give rise to both cosN λ phages and cosN N15 phages. In practical terms, the MF713 (ϕ1199) helper packaging experiment produced a λ imm434 ind -b511 phage pair with cosQ λ cosB λ and with either cosN λ (ϕ1203) or cosN N15 (ϕ1201) that were useful for dilysogen constructions for experiments presented below.
Although the data are limited, there is a preponderance of cosN λ progeny; the reason for this asymmetry is unknown. In both orientations, both cohesive end alleles are found in progeny phages, indicating that at least in some infections, mismatches are not resolved when the cohesive ends anneal and are ligated. Rather the mismatch persists until the mismatch is resolved by DNA replication. In sum, the cohesive end mismatch does not block cyclization, and in at least some cases, the mismatch persists until after the mismatch-containing parental chromosome has replicated. The limited data also are consistent with the scenario that in some infections by a mismatch chromosome, repair may resolve the mismatch prior to replication.
Quantitating effects of a cohesive end mismatch on infectious phage production. To ask about the quantitative effects of a cohesive end mismatch on infectious phage production,  MF4942, with the prophage structure imm 21 cosQ λ N λ B λ ! λ imm 434 cosQ λ N N15 B, and MF4943, with the prophage structure λ imm 21 cosQ λ N λ B λ ! λ imm 434 cosQ λ N λ B λ , were constructed and used in helper packaging experiments (see S1 Table for details). With the λ-specific helper, packaging the ceL N15 /ceR λ mismatched chromosome, the yield was about 25% that of a chromosome with matched ceL λ /ceR λ cohesive ends (Table 3, compare lines 1 and 3), indicating that the mismatch reduced the level of viable virion production by four-fold. Remarkably, the λ-N15 hy4 helper phage was able to package λ DNA rather efficiently, at about 50% the efficiency of the λ-specific helper ( Table 3, compare lines 1 and 2 and lines 3 and 4). As with the λ-specific helper, the yield of viable ceL N15 /ceR λ mismatched chromosome-carrying phages was about 25% that of phages with matched cohesive ends (Table 3, compare lines 2 and 4).
To summarize, the λ/N15 cohesive end mismatch reduced the recovery of viable packaged chromosomes roughly fourfold, and the N15-specific helper packaged λ-specific DNA at about 50% efficiency.
N15's Packaging Specificity cosN specificity. One explanation for the reduction in yield for mismatched phages is that packaging is inefficient due to reduced efficiency of cos cleavage. For example, since both the λ and N15 hy4 terminases have TerL λ , both terminases might cleave cosN λ more efficiently than cosN N15 . Alternatively, the TerS N15hy4 :cosB λ interaction might affect TerL λ positioning and the efficiency of cosN cleavage. In either case, λ terminase might act more efficiently in phages with cosB λ than N15 hy4 terminase. To investigate this point, the yields of phage pairs expressing λ or N15 hy4 terminase, with cosB λ and either cosN λ or cosN N15 , were determined. The yields were found not to depend on cosN: for both λ and λ N15 hy4 phages, the yield did not significantly differ between cosN λ and cosN N15 (Table 4). These results do not support the idea that λand N15-specific terminases are cosN-specific. An alternative explanation for the yield reduction caused by a cohesive end mismatch is that DNA packaging is fully efficient, but the infectivity of the virions with mismatched cohesive ends is reduced. The reduction in yield might be due to less efficient cyclization or a lethal event caused by attempted mismatch repair. Finding a lack of cosN specificity is not surprising, as follows. First, the λ and N15 TerLs are quite similar, with 64.9% identity and 78.5% similarity. Second, the base pair differences in the cohesive end sequences are located in the right half of the cohesive end sequence, where sequence changes do not have measurable effects on cos cleavage [60].
Termination specificity. Previous work showed that λand 21-specific terminases efficiently terminate DNA packaging at the downstream cos regardless of cosB specificity [61]. Later work showed that cosQ, cosN and I2, the latter a segment between cosN and cosB, were required for efficient packaging termination [46]. To ask if λ and N15 hy4 terminases were also cosB-non-specific for packaging termination, dilysogens were constructed in which the initial cos was cosB N15 and the downstream cos was cosB N15 (MF4944) or cosB λ (MF4945). In a second pair of dilysogens, the initial cos was cosB λ , and the downstream cos was either cosB λ (MF4946) or cosB N15 (MF4947). These dilysogens were then further lysogenized with the λand N15-specific helpers and used in helper packaging experiments. Neither helper displayed any cosB specificity for termination at the downstream cos (Table 5). These results extend the earlier observations that packaging termination is not dependent on cosB specificity [46]. These results are not unexpected given the strong sequence conservation for cosQ, cosN and I2 in λ, 21 and N15.
Packaging specificities of λ, 21 and N15. The packaging specificities of λ and 21 are due to the TerS-cosB interaction at the start of DNA packaging. To study N15's packaging initiation specificity, a set of three dilysogens was constructed in which the passive prophage's upstream initiation cos was cosQ λ N λ B λ (MF4948), cosQ λ N λ B 21 (MF4949), or cosQ λ N λ B N15 (MF4950). In each strain the downstream termination cos was Q λ N λ B λ . These dilysogens were then lysogenized with the λ-, 21-, or N15-specific helper phages. As in previous helper packaging experiments, the λand 21-specific terminases failed to package each other's chromosomes (Table 6). Further, neither the λ-specific or 21-specific helpers packaged the N15-specific chromosome, suggesting that N15 has a unique packaging specificity. The inability of λ and 21 to package N15-specific chromosomes is not unexpected, given the apparent simplicity of cos N15 compared to the elaborate structures of cos λ and cos 21 . The possibility remained that N15 and λ might have the same packaging specificity, but that λ terminase was unable to utilize cosB N15 because of the lack of I1, R2 and R1. In previous work, it was shown that several Nu1 missense mutations affecting TerS λ (gpNu1) suppressed a variety of cosB λ defects. The Nu1ms1 mutation  [62]. Because Nu1ms1 terminase sponsors packaging of DNA containing only R3 of cosB λ , we wondered if Nu1ms1 terminase would sponsor packaging of N15-specific DNA. We used the N15-specific, λ-specific, and Nu1ms1 helper phages to ask if Nu1ms1 terminase could sponsor packaging of N15-specific chromosomes in MF4950. While the N15-specific helper gave a yield of 3.5 prophages/induced lysogen, the prophage yields were 2.9 x 10 -4 and 1.35 x 10 -4 for the λ wild type and λ Nu1ms1 phages, respectively. As observed above with other dilysogens, the N15-specific helper packaged λ-specific chromosomes with reasonable efficiency, i.e., about 50% the efficiency of packaging self-specific N15 chromosomes. In sum, N15 has an asymmetrically unique packaging specificity, with N15 DNA's specificity being distinct from those of λ and 21, but surprisingly, N15-specific terminase is able to package λ DNA reasonably efficiently.

Discussion
The goal of this work was to use helper packaging experiments to study N15's packaging specificity, using λ N15 hy4 [58]. For helper packaging experiments, the λ-versus-N15 cohesive end difference was a complication. That is, the 12 base-long cohesive ends of λ and N15 differ at bp 9 and 12, and initial helper packaging experiments generated chromosomes with a λ/N15 cohesive end mismatch. We found that chromosomes with mismatched cohesive ends are able to cyclize, and the mismatch persists past replication of the chromosome, in at least some infections. The cohesive end mismatch is correlated with a roughly fourfold reduction in viable phage yield. Burst size studies suggest that TerL λ acts with equal efficiency at cosN λ and cosN N15 , and that cos N15 is efficiently utilized for packaging termination. The preliminary helper packaging experiments produced progeny phages, enabling helper packaging experiments on N15's packaging specificity which did not generate mismatched chromosomes. N15's DNA packaging specificity was found to be distinct from those of λ and 21. That is, neither λ nor 21 packages N15 chromosomes and N15-specific terminase does not sponsor 21 DNA packaging. Surprisingly, N15-specific terminase packages λ DNA at about 50% efficiency.
Functional asymmetry of the cohesive ends of λ-like phages cosN λ , while showing significant two-fold rotational symmetry, is remarkably asymmetric functionally, as follows. An early study of cosN mutations found that many mutations affecting bp 7-12 in the right half of cosN had minimal phenotypic effects, whereas mutations affecting symmetric bp in the left half of cosN have severe phenotypic effects [60]. For example, the G2C mutation reduced virus yield about 10-fold, but the symmetrical C 11 G change had no phenotypic effect. Subsequent biochemical cos cleavage experiments confirmed that cosN is functionally asymmetric, and further showed that the asymmetry reflected the proximity of cosN's right half to cosB [34]. Studies on the importance of the λ cosN-to-cosB spacing also indicate that the TerS-R3 interaction precisely positions TerL with respect to the cosN nicking sites for proper nicking by the TerL endonuclease [33,36]. These results suggest that anchoring of TerL λ to cosB makes bottom strand nicking relatively insensitive to base pair changes. The functional asymmetry of cosN correlates well with natural cosN sequence variation, as follows. The cohesive end sequence of ϕD326, determined long ago [63], differs from cosN λ at positions 9 and 12, as does N15. Similarly, Salmonella phage Gifsy-1 differs from λ at cosN bps 8, 9, and 11. In contrast, the two base pairs flanking the nick positions, 5'-CG-3' and 5'-CG-3', are highly conserved, indicating that there is selection against changes of these bps. Similarly, natural variation in cosQ, the packaging termination signal, is not apparent. In sum, the results indicate that sequence maintenance for cosQ and the left half of the cohesive end sequence is under selection but maintenance of bp positions 7-12 is not.

Genetic consequences of the λ versus N15 cohesive end mismatch
Our experiments show phages carrying λ chromosomes with mismatched cohesive ends, at bp positions 9 and 12, are able to circularize and replicate. For such phages, there is a modest, four-fold decrease in yield. The reduced yield is not due to effects of cosN bp changes on terminase efficiency, as both λ and N15 hy4 terminases sponsor robust virus yields with either cosN λ or cosN N15 . We speculate that the reduced yield of viable phages is due to either inefficient cyclization or lethal attempts at mismatch repair. One candidate for repair of the mismatch is the 5'-to-3' exonuclease of DNA polymerase I, which could carry out nick translation through the mismatched bp. In at least some of the successful infections by phages with mismatched cohesive ends, the mismatch persists through replication of the parental chromosome, since some singly infected cells yield both cosN λ and cosN N15 progeny phages ( Table 2). These results are surprising, since bp position 12 is adjacent to the bottom strand nick which presumably must be ligated prior to replication fork passage. The λ and N15 hy4 terminases, with TerS λ and chimeric TerS N15hy4 subunits, respectively, do not show cosN specificity. For both, the large subunit is TerL λ which must be appropriately positioned to nick cosN. Given the functional asymmetry of cosN λ , efficient cleavage of both cosN λ and cosN N15 is as expected.
Packaging specificity in the N15-like phages N15 is an unusual λ-like phage in having a simple cosB containing a single critical TerS binding site, R3 N15 . R3 N15 extends from N15 bp 48-59, and is positioned similarly to R3 λ . In cosB N15 , there is a proposed additional remote site, "rR2", in opposite orientation to the critical R3 N15 sequence and located in the 1 gene at bp 249-260 [58]. The rR2 sequence strongly matches that of R3 N15 and is found in all N15-like prophages with the simple cosB [58]. The rR2 sequence is not critical for N15 growth. Though not essential for plaque formation, rR2 is speculated to contribute to viral fitness for N15 and its close relatives [58]. Another group of prophages, typified by Monarch, has a strong match to R3 N15 and TerS N15 's recognition helix, suggesting a shared packaging specificity (Fig 2). The Monarch group, however, does not have the simple cosB N15 . Rather cosB Monarch is analogous to cosB λ , with three R sequences and an IHF binding site between R3 and R2, and the rR2 sequence is not present.
Comparing the R3 sequences of 21, N15, Monarch, and λ leads to some conclusions about TerS-R3 interactions (Fig 2). The strong identity of R3 N15 and R3 Monarch is apparent, as is the lack of similarity of R3 N15 and R3 Monarch to R3 λ and R3 21 . In R3 λ and R3 21 , bp 56 and bp 59 are critical for TerS recognition [56]. R3 λ bp 58 is also important for TerS λ recognition [55]. R3 N15 was identified by the packaging effects of two 6 bp-long block mutations, m1 and m2. In m1 and m2, bp 48-53 and 54-59, respectively, were scrambled, as follows. In m1, all six bp were changed-GAGGTT!TGTAGG, and in m2, four were changed-GTTGTT!TGTTGT [58]. The cosB N15 m1 and m2 mutations reduced λ N15 hy4 's yield by about 7-fold and 100-fold, respectively, indicating bp 54-59 are most important for recognition. That bp 54 to 59 include bp important for TerS interactions agrees with the locations of cosB λ and cosB 21 bp 56 and 59, which are important bp for recognition by TerS [56].
Why can N15-specific terminase package λ DNA, but not 21 DNA? Looking at the R3 alignment gives hints, but first it is important to consider the validity of the alignment in Fig 2A. In the alignment, each sequence is equidistant from bp 1, the first base of the left cohesive end. It is possible to shift the N15 and Monarch sequences to increase the sequence identity to R3 λ and R3 21 . However in λ, the cosN-to-R3 distance is critical, because shifting the distance by inserting or deleting more than one or two bp lowers virus yield and results in improper nick introduction in cosN [33,36]. These results indicate that when N15-specific terminase sponsors packaging of λ DNA, TerS N15hy4 anchors TerL λ in a manner that closely matches TerS λ anchoring of TerL λ . In the cosB alignment, R3 N15 and R3 Monarch have a match to the R3 λ TA bp 56, an important TerS λ contact, R3 λ bp 56 also plays a role in λ/21 phage specificity, as follows. Genetic analysis indicates that Glu24 of the TerS λ recognition helix clashes with GC bp 56 of R3 21 , accounting in part for the inability of λ terminase to recognize 21-specific DNA (Fig 2B). The genetics also indicated that Glu24 of TerS λ makes a favorable contact with TA bp 56 of R3 λ . We speculate that, in the case of N15's recognition helix, perhaps the Val22 residue, at the position analogous the TerS λ 's Glu24, makes a favorable contact with the bp 56 TA bp of cosB λ , but is unable to make a favorable contact with the GC bp 56 of R3 21 . It is also not clear whether the inability of λ terminase to package N15 DNA is due to a defect in cosN cutting, or in a postcleavage defect such as formation of complex I. More research is required to understand these relationships.

cosB architecture and function
The likely shared packaging specificity of phages N15 and Monarch, and the finding that N15 packages λ DNA raise further interesting questions about cos structure and terminase recognition. The interaction of TerS λ with cosB λ includes a proposed IHF-assisted DNA hairpin that juxtaposes the major grooves of R3 and R2 for docking by dimeric gpNu1 [20]. The strong similarity of cosB λ and cosB Monarch suggests a hairpin-containing nucleoprotein structure also forms at cosB Monarch . It is interesting that the TerS N15 DNA binding domain forms a tight dimer [58], even though cosB N15 is much simpler than cosB Monarch . Does TerS N15 form a similar hairpin structure at cosB N15 , even though there is neither an IHF site nor an R2-equivalent? Elsewhere we propose that cosB N15 is derived from cosB Monarch by unequal crossing over between R3 and R1. If so, the TerS N15 would be derived from TerS Monarch , and might retain the ability to interact with cosB Monarch in a manner similar to TerS λ docking at cosB λ . A related issue is packaging of λ DNA by N15-specific terminase. Does TerS N15hy4 utilize all three R sequences of cosB λ when packaging λ DNA? These interesting questions await further studies.
What evolutionary steps account for the packaging systems of the λ-like phages? It seems clear that λ and 21 are descended from a common ancestor phage, as the basic cos structure is preserved. We earlier presented one possible path, in an early ancestor virus where a single R sequence and a terminase with low sequence specificity acquired greater packaging specificity by acquisition of three R sequences [64]. A scenario for N15 is that N15 has undergone a simplification process, by which a Monarch-like ancestor phage with three R sequences has lost two, and the IHF site, by unequal crossing over between R3 and R1. One would think that such an event would result in a major fitness reduction. It is possible however that the simplification and subsequent refinement, such as the acquisition of rR2, might take place in the prophage state, where fitness for lytic growth is moot [58,65].
Genetic consequences of the asymmetric, shared packaging specificities of λ and N15 The chromosomes of many bacteriophages, including the λ-like phages, are mosaics of DNA segments, indicating that extensive horizontal exchange is part of a phage's evolutionary history [66]. Enteric bacteria frequently harbor λ-like prophages, so that an infecting λ-like phage has a strong likelihood of replicating in the presence of another λ-like phage chromosome. Of course several significant consequences may result from such an interaction. If a prophage shares the packaging specificity of the infecting phage that enters the lytic cycle, then the packaging machinery of the infecting phage will initiate packaging at the prophage cos. For the λ-like prophages, cos is located roughly in the center of the prophage, so that the packaging initiation encapsidates a DNA segment that will include adjacent bacterial DNA, stalling the motor when the head shell is full, which in turn produces a tailless, non-infectious particle. Production of such inactive phage structures reduces the production efficiency of viable progeny phages. One might think therefore, that it would be advantageous for λ-like phages to have diverse packaging specificities. So far, only modest diversity is observed, though not many examples have been studied. Among characterized λ-like phages with related cos structures, the three packaging specificities of λ, 21 and N15 have been documented. Other natural isolates, such as phages 434 and ϕ80, share λ's packaging specificity. N15 is an interesting case, as N15's packaging machinery retains the ability to package λ DNA, even though N15 has a significantly divergent DNA recognition system.
A second interaction between λ-like phages is recombination. If a packaging event initiated at a prophage cos, is accompanied by a downstream crossover between prophage and lytic phage DNAs, the recombinant is a novel progeny phage that contributes to genetic diversity. Some such genetic exchanges generate recombinants that are more robust in a changing environment. In the case of a cohesive end mismatch, such as in the present λ versus N15 situation, recombination may produce less fit progeny phages. Because the cohesive end mismatch does not persist, the reduced viability of a recombinant phage is transient.

DNA recognition by TerS
TerS molecules interact with their recognition sites through their N-terminal DNA binding domains, and with TerL through their C-terminal domains. These two specificity domains are separated by the long α-helical core domain. In principal, phage diversity might be generated by swapping of the TerS DNA recognition domain for that of any other DNA binding protein with an asymmetric recognition site. In the present case, we note that both the cos and TerS of N15 are recognizably evolutionarily related to those of phage λ [58]. Phage 21's packaging specificity determinants also share a common ancestor with λ [56]. N15 and 21 were chosen for study as interesting variants of the λ paradigm, and as such these phages do not shed light on extreme possibilities for specificity swapping. We note that in a more systematic study in P22-like phages, considerable horizontal transfer of TerS gene segments has occurred, not only between highly diverged members of the P22 family, but also has occurred between P22-like phages and phages outside the P22-like group [16]. Acquisition of a TerS segment from a nonphage source such as a transcriptional regulatory system has not been observed. Surely further sequence acquisition and study will add to what we know about this issue.

Bacteria, phages and plasmids
These are listed in Table 1. A list of dilysogens constructed for this work is in S1 Table.

Microbiological methods
Phage crosses and infections were done using standard protocols [67].

Helper phages
The helper phages used were in the thermo-inducible λ-P1:5R cI857 Kn R nin5 genetic background. In λ-P1:5R cI857 Kn R nin5, the DNA segment containing the λ site-specific recombination system is replaced by the phage P1 plasmid replication and partitioning segment, so the helper phage forms a plasmid prophage [68]. The Kn R marker in λ-P1:5R cI857 Kn R nin5 is a substitution for λ DNA between the two λ SalI sites [69]. The SalI sites are located in (1) the bet gene and (2) distal to gam, so that the helper phages are redgam -. When growing in a recAhost, redgamphages are unable to generate concatemers through rolling circle replication or recombination [29]. Consequently, the helper phage yield is low, about 10-20 phages/cell in the experiments reported here. For simplicity, λ-P1:5R cI857 Kn R nin5 will simply be designated λ; this phage was used as the helper with λ packaging specificity. λ 21 hy51 [59] and λ N15 hy4 [58] were the 21-and N15-specific helpers, respectively.

Dilysogens
Dilysogens were constructed with two recAstrains with deleted prophages, MF713 and MF532 ( Table 2). The λ b511 prophage of MF713 is deleted for a DNA segment extending from the bacterial gal operon thru the prophage early genes, including cI, while the segment containing cos, the head and tail genes, the b511 marker and attΔ•B are present. The genetic structure of MF713 is: D½galÀ À ÀattL À cl À S À cos l À == À b511 attD BÀÀbioÀ where λ genes are underlined, slashes indicate the long DNA stretch of the λ head and tail genes, dashed lines indicate E. coli. Lysogenization of MF713 by λ imm 434 indcos i generates the following genetic structure: D½galÀ À ÀattLÀcl À SÀcos l À==ÀattPÀimm 434 À cos i À == À attRÀÀbioÀ; where cos i indicates that the downstream cos, at which termination occurs, may have λ, 21 or N15 sequences. The passive packageable prophage is double-underlined. Care was taken to exclude isolates with two copies of the added prophage. For MF713 (λ imm 434 cos i ), we confirmed that the chromosomes produced by helper packaging of the passive prophage carried the b511 marker. Generally, 10 isolates were tested by a PCR assay or a test for a b511 phage's inability to lysogenize.