cos

The cos site is a ≈200 bp long segment that is required to both initiate and terminate the packaging of a monomeric genome from concatemeric DNA. The site where terminase introduces staggered nicks to generate the cohesive ends is called cosN (fig. 2). Early during the study of λ, it was thought that cosN was both necessary and sufficient for DNA packaging. Later studies showed, however, that cos is complex and consists of three and perhaps four distinct subsites.^11-15 Both initiation and termination of packaging require duplex nicking at the cosN site; additionally, efficient initiation requires the presence of the cosB subsite, which is directly downstream from cosN. Conversely, efficient termination requires the presence of cosQ, a subsite that is located upstream of cosN (fig. 2).¹⁶ The I2 sequence is located between cosN and cosB, and also plays a distinct role in efficient DNA packaging.¹¹ Thus, the complete cos sequence consists of several subsites, each of which plays a specific role in the recognition, processing, and packaging of viral DNA. Each of these subsites is discussed in detail below.

Figure 2

The cos region of a lambda concatemer. Upper panel: The cosQ, cosN and cosB subsites within a cos site in concatemeric DNA. The cosB subsite is composed of the I1 and R-elements, as indicated. The I2 region lies between cosN and the R3 element. Middle (more...)

cosN

The terminase enzyme introduces nicks into the duplex at the cosN site to generate the cohesive ends of mature virion DNA. Many of the base pairs (bp) within cosN show two-fold rotational symmetry, which extends over 22 bp if one includes purine-purine and pyrimidine-pyrimidine symmetry (fig. 2); this has been used as evidence that a symmetrically disposed enzyme complex (i.e., a gpA dimer) is responsible for duplex nicking. This argument is further supported by (i) analogies to the interactions of type II restriction endonucleases with their palindromic recognition sequences, and (ii) the presence of a leucine-zipper motif in the primary sequence of gpA.¹⁷ We presume here that a gpA dimer is responsible for symmetric duplex nicking at cosN.

cosB

Terminase nicking of an isolated cosN site is error-prone, and the overall rate of nicking is slow.^18,19 The presence of cosB and the I2 element (discussed below) is required for efficient and accurate duplex nicking. The cosB subsite contains three 16 bp R elements that are specifically recognized by gpNu1 (fig.2);^12,33,43,55 specific cosB-gpNu1 interactions are likely required to properly position a gpA dimer at cosN for the initial cos cleavage event.^18,19 The cosB subsite also contains a consensus sequence for Escherichia coli integration host factor (IHF). This I1 site introduces an intrinsic bend into the duplex, and is also specifically recognized by IHF.^20,21 The role of IHF in DNA packaging and virus assembly is discussed more fully below.

cosQ

This seven base pair subsite is located 17 bp upstream of cosN (fig. 2), and is essential for proper termination of the packaging reaction.^16,22-25 Severe cosQ mutations do not significantly affect the initial cos cleavage reaction, but abolish nicking of the bottom strand at the terminal cos site. Moreover, packaging is not arrested in the absence of cosQ, and additional DNA, including the downstream cos, is packaged until the capsid shell is filled to capacity.²⁶ This suggests that cosQ is required to stop the packaging machinery for appropriate cleavage at the terminal cosN site; however, cosQ alone is incapable of arresting DNA packaging. Rather, cosQ acts in concert with cosN and I2 to promote efficient termination.²⁶

The I2 subsite, originally identified by sequence homology to the IHF binding site consensus sequence, is located between cosN and cosB (fig. 2). I2 is not a functional IHF binding site, however, as evidenced from mutagenesis studies,27 sequence information content analysis²⁸ and direct IHF-DNA binding studies.²⁹ Nevertheless, the I2 region appears to play a role in both initial and terminal cos-cleavage events as follows.^b (i) Mutation of individual bases within the I2 site does not affect the initial cos-cleavage reaction, while small deletions reduce nicking accuracy in vitro and packaging initiation in vivo.^27,30 Based on these data, it was proposed that the role of I2 in the initial cos cleavage reaction may be to simply provide the proper spacing between the cosN and cosB subsites.^27,30 (ii) The role of I2 in the late steps of DNA packaging is more interesting. While the cosB sequence is not required for the terminal cos-cleavage reaction, efficient nicking of a downstream cosN site requires the I2 region.¹¹ Moreover, it has recently been found that replacing the I2 sequence with a randomized sequence that preserves the cosN-cosB spacing is profoundly lethal (B. Charbonneau and M. Feiss, unpublished). Thus, the I2 sequence serves a specific role in terminating the packaging reaction. In other words, cosQ and cosN alone are insufficient for terminal cos cleavage, and efficient termination requires that the entire cosQ-cosN-I2 segment be present.

We note that terminase remains bound to the concatemer after the DNA-filled capsid has been released (see fig. 1B). This binary enzyme-DNA complex binds an empty procapsid to initiate a second round of packaging, and terminase thus packages successive genomes in the concatemer in a processive manner. While the terminal cos-cleavage reaction requires only the I2 element, processive packaging requires that the cosB subsite also be present.¹¹

gpNu1

The phage λ small terminase subunit (181 amino acids) is responsible for the assembly of the packaging machinery, and for the stability of the packaging initiation complexes.³² gpNu1 assembles at cosB, and direct interactions with the three R-elements have been demonstrated by DNase protection assays.³³ ATP increases gpNu1 binding to cos containing DNA substrates without affecting nonspecific DNA binding (M. Ortega and C.E. Catalano, unpublished). Nucleotides do not affect the DNase protection pattern, however.³³ While the isolated subunit can bind ATP, significant hydrolysis by gpNu1 is observed only in the context of the holoenzyme (H. Gaussier and C.E. Catalano, unpublished).^9,34

The Three Dimensional Structure of gpNu1

A high-resolution NMR solution structure for the gpNu1 DNA binding domain (gpNu1-DBD) has recently been solved (see fig. 4A).⁴³ The structure, combined with NMR-monitored titration studies and genetic experiments, confirm sequence analysis predictions that DNA binding is mediated by a helix-turn-helix (HTH) DNA binding motif in this region of the protein (residues Lys5 — Glu24).^45-47 Consistently, substituting the recognition helix from the small subunit of phage 21 terminase into gpNu1 created a chimeric terminase specific for packaging of 21 DNA. Structural analysis of gpNu1-DBD further classified the motif as a winged helix-turn-helix (wHTH) DNA binding motif. The wHTH is present in a diverse family of DNA binding proteins, including rat hepatocyte nuclear factor-3,⁴⁸ Drosophila homeotic fork-head protein,⁴⁹ the bacterial OmpR response regulator protein,⁵⁰ and the phage Mu C repressor and transposase proteins.^51,52 These proteins bind DNA utilizing classical HTH-major groove binding interactions, with additional affinity and/or specificity modulated by interactions between wing residues and the minor groove of DNA.⁵³

Figure 4

Structure of gpNu1-DBD and model for cooperative gpNu1-IHF DNA binding. Panel A: The High Resolution Structure of the DNA Binding Domain of gpNu1 (gpNu1-DBD). The average of the 20 lowest energy structures obtained from NMR is shown in ribbon representation. (more...)

Escherichia coli Integration Host Factor (IHF)

IHF is a site-specific DNA binding protein that introduces a ≈180° bend into duplex DNA.^56,57 IHF often exerts its biological effect by forming nucleoprotein complexes that are conducive to the assembly of other DNA binding proteins at that site;⁵⁸ direct IHF-protein interactions in these higher-order complexes have not been demonstrated. IHF modulates λ development in vivo, with virus yields lowered 3 to 4-fold in its absence.^c56,59 Importantly, in vitro DNA packaging is also lowered ≈4-fold in the absence of IHF,^60,61 consistent with a direct role for the host factor in virus assembly. An IHF consensus sequence, I1, has been identified between R3 and R2 in cos and direct IHF-I1 binding interactions have been demonstrated (fig. 4B).^20,56,59

Genetic and mutagenesis studies show that any of a number of mutations affecting the R-elements of cosB renders the virus IHF-dependent for growth,⁶² and a direct role for IHF in gpNu1 assembly at cosB has been implicated.^63,64 It is presumed that cooperative assembly of gpNu1 at cosB is efficient under normal conditions, and IHF plays a supportive role only; however, when gpNu1-DNA interactions are attenuated (i.e., by mutation), gpNu1 assembly requires additional interactions, presumably provided by structural alterations in the DNA induced by IHF.

Model for an IHF-gpNu1 Nucleoprotein Complex

The high-resolution structure of gpNu1-DBD, while initially perplexing, led to an attractive model that nicely accommodates both gpNu1 assembly at cosB, and the supportive role of IHF in virus assembly (fig. 4B,C). In this model, IHF binding to I1 introduces a strong bend in the duplex, which juxtaposes the R3 and R2 gpNu1 binding sites in the appropriate orientation. A single gpNu1 dimer spans these two “half-sites” that are separated by 44 bp in the duplex. This model predicts that both the DNA sequence (R-elements) and duplex structure (bend) are critical determinants to gpNu1 assembly. Importantly, an intrinsic bend is observed at the I1 site,^20,65 which would promote gpNu1 binding in the absence of IHF. Mutation of an R-element decreases intrinsic gpNu1 binding affinity, which places increased emphasis on the bend; IHF binding provides the prerequisite DNA structure. While this model is fully consistent with all of the genetic, biochemical, and structural data available, it does not directly address the role of the R1-element in gpNu1 assembly. Of note, however, is that this element is dispensable in the presence of IHF in culture^d.^55,62,64

gpA

The large terminase subunit is composed of 641 amino acids. Genetic and biochemical studies have defined several functional domains of the protein, including an N-terminal domain responsible for gpNu1 interactions in the holoenzyme complex,^39,40 and a C-terminal domain responsible for procapsid binding during initiation of DNA packaging^21,38,66 (fig. 3). Other domains are described in detail below.

Terminase Holoenzyme

Lambda terminase is isolated as a heteromultimer with a subunit ratio of gpA₁•gpNu1₂.³¹ As outlined above, the holoenzyme possesses a site-specific endonuclease activity, a so-called helicase activity, and a DNA packaging activity that act in concert to package viral DNA. The holoenzyme also possesses multiple ATPase catalytic sites that play unique roles along the packaging pathway. Each of these catalytic activities is discussed in turn below.

Cooperative Assembly of Terminase at cos

A subtle interaction between gpNu1 assembly at cosB and gpA assembly at cosN has been exposed through mutagenesis studies, as follows. Introduction of mutations into the right cosN half site (cosNR, fig. 2) has little effect on the cos cleavage reaction in an otherwise wild-type background.⁸³ Conversely, introduction of a symmetric mutation into the left cosN half site (cosNL) has dramatic effects, significantly decreasing the rate of the cos cleavage reaction in vitro⁸³ and virus yield in vivo.⁸⁴ Introduction of secondary mutations in cosB, or complete deletion of cosB removes this asymmetry, and cosNR and cosNL mutations have identical effects. These data suggest that intrinsic DNA binding interactions between gpA and cosNL are critical to efficient duplex nicking, and that introduction of mutations into this DNA site strongly affect the reaction. Conversely, gpA binding to the cosNR half site is supported by cooperative interactions with gpNu1 assembled at cosB (fig. 5). Thus, introduction of mutations into the cosNR half-site are “masked” by supportive interactions in the holoenzyme-DNA complex. Deletion (or mutation) of cosB DNA abrogates gpNu1 assembly, which in turn abolishes cooperative gpNu1-gpA binding interactions. The outcome is that gpA binding at both half-sites relies exclusively on intrinsic binding interactions, and symmetric cosN mutations have identical results.

Figure 5

Model for cooperative assembly of gpA (red rectangle) and gpNu1 (blue rectangle) at cos. A symmetric gpA dimer binds to cosN. Binding of gpA to the cosNL half-site relies strictly on intrinsic binding interactions. Binding of gpA to the cosNR half-site (more...)

Summary of the Allosteric Interactions between Terminase Catalytic Sites

Allosteric interactions between the multiple catalytic sites of terminase holoenzyme have been observed, and are summarized in Figure 6.^g (i) ATP binding (but not hydrolysis) increases the affinity of gpNu1 for cos containing DNA.³² This interaction is likely responsible for ATP-stimulated DNA binding by the holoenzyme.⁸² (ii) The ATPase activity of the individual subunits is mutually stimulated in the holoenzyme complex; ATP hydrolysis by the gpNu1 subunit is dependent upon gpA, and the ATPase activity of gpA is stimulated by gpNu1.¹⁰ Furthermore, the endonuclease activity of gpA is stimulated by interactions with the gpNu1 subunit.¹⁰ (iii) ATP binding to the gpA subunit stimulates the rate of the nuclease reaction in the holoenzyme, but not by the isolated gpA subunit.82 Similarly, (iv) DNA stimulates the ATPase activity of both terminase subunits in the holoenzyme complex, but not by the isolated subunits.^10,67,70 The above data suggest that quaternary interactions in the holoenzyme complex are required for full expression of catalytic activity, and provide a critical link for communication between the catalytic sites. (v) GTP binding to gpNu1 stimulates ATP hydrolysis at the gpA subunit of the holoenzyme.⁹² Conversely, (vi) ATP binding to the gpA subunit of the holoenzyme inhibits ATP hydrolysis at gpNu1.⁹²

Figure 6

Summary of allosteric interaction between the catalytic sites of terminase holoenzyme. The catalytic activity of the isolated gpA (red rectangle) and gpNu1 (blue sphere) subunits is shown at top. Assembly into the holoenzyme activates catalytic activity (more...)

It is thus clear that there are significant interactions between the multiple catalytic sites of terminase holoenzyme. It is likely that these interactions are central to the assembly of the packaging machinery onto viral DNA, to promoting stability of intermediate packaging complexes, and to modulating the procapsid-dependent transition to a mobile packaging machine.^93,94 It is further likely that ATP binding and hydrolysis play central roles in these processes. These concepts are incorporated into a model for terminase assembly into a DNA packaging machine, described below.

Processing of the λ Portal

The λ portal, unlike the simple portals found in other viruses, is quite complex both in terms of protein composition and protein modification. First, “processing” of the initiator involves proteolysis of gpB to yield gpB^*, a protein in which the N-terminal 20 residues of gpB have been deleted.⁹⁷ The molecular weight of gpB is 59.4 kDa (based on sequence) and proteolytic digestion of the N-terminal 20 residues would yield a protein with mass 57.2 kDa. The mass of gpB^* is 53-56 kDa (based on SDS-PAGE),^97,98,99 suggesting that additional residues are removed from the C-terminus of gpB and/or that gpB^* migrates anomalously on SDS-PAGE.

A second step in processing of the λ portal is formation of the pX1 and pX2 proteins (referred to collectively as pX proteins). Remarkably, these proteins appear to derive from an uncharacterized covalent fusion product of gpC and gpE proteins.^100,101 A transient intermediate in portal processing (the pY protein, ≈87 kDa) has also been described. This intermediate has a molecular weight consistent with a direct fusion of gpC and gpE;¹⁰² presumably, proteolysis of pY yields both pX1 (31 kDa) and pX2 (29 kDa). The chemical nature of these proteins and the mechanism of their formation remain completely unknown. It is noteworthy that the gpC protein shares sequence identity with the S49 family of serine proteases,¹⁰³ and we postulate that gpC is the viral protease responsible for the formation of both gpB^* and the pX proteins. The temporal relationship between gpB proteolysis, cross-linking of gpC and gpE, proteolysis of pY to yield pX1 and pX2, and gpE assembly into a procapsid is unclear. Moreover, the roles of gpB^*, pX1, and pX2 in the structure and function of the portal complex remains unknown.

The final step in procapsid maturation is proteolysis of the gpNu3 scaffolding protein and exit of the products from the procapsid (fig. 7). The mature procapsid contains 12 copies of gpB^*, ≈6 copies each of pX1 and pX2, and ≈420 copies of gpE.¹⁰⁴

Assembly of the Packaging Machinery

gpA and gpNu1 cooperatively assemble at cosN and cosB, respectively. This involves binding of a gpNu1 dimer to the R3 and R2 elements of cosB, supported by bending of the duplex via IHF interactions with I1 (shown in fig. 4). As discussed above, the role of R1 in this initial assembly is unclear. Interestingly, however, model-building studies show that the R1 element of cosB is juxtaposed to the cosNR half site in the bent duplex (C.E. Catalano, unpublished). Whatever the case, gpNu1 bound at cosB serves to anchor a symmetric gpA assembly at the cosN half-sites yielding a prenicking complex (fig.8). We presume that a gpA dimer assembles at cosN, but the structural details and stoichiometry of the subunits assembled at cos remain speculative. The presence of a bZIP protein dimerization motif in gpA suggests that an even number of subunits is involved, however.

ATP modulates the assembly and stability of the terminase subunits bound at cos as follows. First, ATP increases the affinity of gpNu1 and the holoenzyme for cos DNA (but does not affect gpA binding).^32,82 Second, ATP affects the DNase protection pattern of terminase assembled at cos, yielding an “extensive footprint” that extends from cosQ through the R-elements of cosB.¹⁰⁵ Once assembled, the complex is extraordinarily stable, with a half-life of greater than 12 hours.³² Finally, ATP stimulates the rate and fidelity of the nuclease complex.^6,82 In aggregate, the data are consistent with an ATP-induced reorganization of the proteins assembled in the prenicking complex to yield an activated nicking complex, perhaps inducing wrapping of the DNA by terminase (fig. 8).

Duplex Nicking and Strand Separation

In the presence of Mg²⁺, the gpA subunit(s) introduce symmetric nicks into the duplex at cosN. Separation of strands by the so-called “helicase” activity of gpA leads to release of the right cohesive end (D_R) from the complex (fig.8).³² It is clear that this reaction requires ATP hydrolysis, but an increase in the rate of ATP hydrolysis during strand separation has not been detected. The products of the helicase reaction have not been characterized in detail and it is interesting to consider the fate of the terminase proteins bound at cos upon strand separation. A priori, there is no reason to expect that the structure or stoichiometry of IHF and gpNu1 bound to cosB would change upon strand separation. Conversely, it is feasible that separation of the cohesive ends requires remodeling of gpA assembled at the two cosN half-sites. The two obvious outcomes are that (i) gpA bound to the cosNL half-site separates from the complex along with the D_R fragment (as shown in fig.8), or that (ii) gpA initially bound to the cosNL half-site is retained in the complex after strand separation. While the structural details of this transition must await further biochemical analysis, we propose that strand separation ultimately yields two products in which terminase proteins remain bound at each chromosomal end. This is discussed further below.

What Is Complex Iλ

This stable intermediate was originally described in vivo, and was defined as terminase bound to concatemeric DNA at uncut cos sites.^89,106 This definition was based on the properties of the complex that was formed by incubation of terminase with concatemeric DNA and analyzed using sucrose gradients. That is, the terminase•DNA complex migrated at a rate similar to that of the input concatemeric DNA, and at a position distinct from slower sedimenting monomeric virion DNA. Based on these data, it was concluded that the so-called “complex I” was composed of terminase bound to concatemeric DNA, and it was presumed that the cos-sites were not yet cut; these authors proposed that cos cleavage occurred subsequent to procapsid binding.^89,106 We suggest that these early sedimentation experiments could not discriminate between cut and uncut DNA. We further propose that “complex I” formed in vivo is in fact identical to the stable complex that we have characterized in vitro; specifically, terminase tightly bound to the DL fragment formed by duplex nicking and strand separation, as shown in Figure 8.³² This posit is supported by the following lines of evidence. (i) The cos-cleavage reaction is stoichiometric with respect to terminase, and the reaction stalls after a single round of duplex nicking and strand separation; catalytic turnover by the enzyme is not observed.³² In the absence of packaging, the cos-cleavage reaction is thus limited by the concentration of terminase present, both in vitro and in vivo. (ii) Terminase is poorly expressed during lytic infection,¹⁰⁷ and it is likely that the in vivo concentration of terminase is low relative to the number of cos sites in concatemeric DNA. Stoichiometric cos-cleavage would thus result in only a limited number of cos-sites cleaved by the enzyme, assuming that packaging were interrupted in vivo. (iii) Terminase enzymes with mutations in the procapsid-binding domain of gpA have been constructed. The mutant enzymes fail to package DNA, presumably a result of attenuated procapsid binding interactions.^21,65,66,108 The mutations are specific for packaging and do not affect the cos-cleavage activity of the enzyme in vitro. Nevertheless, the extent of cos-cleavage by these packaging-defective terminases in vivo is reduced to roughly one third the wild type level. This discrepancy may be explained as follows. Processive packaging by terminase results in the sequential packaging of ≈3 genomes per DNA binding event.¹⁰⁹ Thus, three cos sites, on average, are cut in a concatemer once the enzyme has initiated packaging in a wild-type infection. The 30% decrease in cos-cleavage activity observed with the packaging-defective terminases in vivo closely matches the value expected from an enzyme that can efficiently cut the duplex, but that cannot not processively package concatemeric DNA.

The ensemble of data fully support a model where stoichiometric duplex nicking and strand separation by terminase in vivo yields a stable complex in a manner identical to what is observed in vitro; we refer to this intermediate as complex I. It is likely that the stability of this complex is required to protect the single-stranded left cohesive end (D_L) from nuclease damage in vivo, though this has not been directly demonstrated. Conversely, the D_R complex is relatively unstable, and dissociation of the terminase protein(s) leads to degradation of the right DNA end by host nucleases (fig. 8), including the RecBCD nuclease.¹¹⁰

cos Clearance

One of the most fascinating and yet ill-characterized aspects of the λ packaging pathway is the transition from the stable complex I to a highly mobile packaging machine that translocates DNA into the capsid. This transition is analogous to promoter clearance by RNA polymerase enzymes, a prelude to active RNA synthesis.¹¹¹ In the λ packaging pathway, “cos clearance” requires activation of the packaging ATPase in gpA, and a switch from a terminase complex specifically-bound to cos to a motor complex that binds tightly, but nonspecifically to DNA. This series of events is likely initiated by interactions of complex I with the portal proteins in the procapsid, and is also modulated by the phage gpFI protein, as discussed in detail below.

Interaction between the C-terminal 32 residues of gpA and the portal gpB^* protein has been demonstrated, and it is likely that these interactions play a direct role in “docking” terminase to the procapsid.^21,38,112 Additional interactions between gpA and gpE on the procapsid surface may also be involved prior to the docking of complex I at the portal vertex.¹¹³ Whatever the case, these initial interactions lead to a pretranslocation complex, which we propose is related to complex II previously described in vivo (fig. 9).⁸⁹ The structural details of the complex remain unknown; however, it is likely that this interaction is mediated by gpA in complex I and gpB^* in the portal complex (vide infra). The next step in the packaging pathway is the transition to a translocation complex, a reaction that is also mediated in an ill-characterized manner by the phage gpFI protein.

The Roles of gpFI and Procapsids in cos Clearance in Vivo

The gpFI protein is abundantly produced during the λ lytic cycle, but it is not a component of the virion. Mutations in the FI gene are lethal to the virus. FI mutant phage produce normal amounts of viral DNA, but the concatemers are not processed to mature length (i.e., cos cleavage is not observed in vivo).¹¹² Interestingly, this phenotype is also observed in phages carrying mutations in any of the procapsid assembly genes, including the host genes for the GroESL chaperonin (see fig. 7).^112,114,115 These data have been interpreted to indicate that gpFI and mature procapsids are required for efficient cos cleavage in vivo, and several models have been proposed in which gpFI and/or procapsid proteins directly control the cos cleavage reaction. One model invokes a negative regulation mechanism where unassembled capsid proteins directly inhibit the endonuclease activity of terminase; assembly of these proteins into a procapsid effectively removes the inhibitor(s) from solution, relieving the repression of cos cleavage.¹¹⁶ The negative regulation model has been eliminated by the following experiments. Phage with amber mutations in FI and/or any of the procapsid genes (i.e., mutations that result in elimination of the putative regulatory proteins) are still defective in nuclease activity.^112,114,115 Indeed, deletion of the entire block of genes including FI and all of the procapsid genes yields phage that similarly possess an in vivo cos cleavage defect.¹⁰⁸ Thus, negative regulation of terminase activity by gpFI and/or procapsid precursor proteins in vivo is unlikely. A second model invokes positive regulation and proposes that gpFI and/or mature procapsids directly stimulate the endonuclease activity of terminase. This model suggests that terminase binds cos and is poised to nick the duplex, but requires gpFI and procapsids to stimulate the reaction. Central to this model is the presumption that complex I formed in vivo is composed of terminase bound to an uncut cos site in the concatemer,^89,106,117 a model which we disfavor (vide supra).

The in vivo cos cleavage results described above are in direct contrast to in vitro studies showing that terminase can efficiently cut cos-containing DNA in the absence of gpFI and procapsids.^{6,18,31,118,119} In fact, gpFI only modestly stimulates the cos cleavage reaction in vitro, presumably by promoting enzymatic turnover rather than by direct stimulation of the nicking reaction; procapsids alone or in combination with gpFI have little, if any effect.¹²⁰ Thus, neither procapsids nor gpFI affect duplex nicking by terminase in vitro, and we propose that terminase may similarly cut cos in the absence of these factors in vivo (see the discussion of complex I above).

A Reversibility Model for the Procapsid Requirement for cos-Cleavage in Vivo

If terminase does not need to be activated to cut cos, how might the in vivo requirement for procapsids be explainedλ We suggest that the in vitro results accurately reflect the activity of terminase in vivo; that is, the enzyme can efficiently assemble and nick the duplex at cos in the absence of procapsids. This predicts that during an infection by a mutant phage unable to produce procapsids, duplex nicking and strand separation should proceed normally. A recently proposed reversibility model accounts for the apparent procapsid dependence of terminase in vivo.¹⁰⁸ Here we present a variation of that model in which the cos-cleavage reaction yields an intermediate where the nicked-annealed duplex is bound by terminase in a strand separation complex (T•D_R•D_L, fig. 10). Physical separation of the strands requires binding of the 12 base single-stranded D_L and D_R ends by individual gpA subunits in the complex, a process that is driven by an ATP-dependent conformational change. We propose that the strand separation step is reversible within this nucleoprotein complex, and that the nicked, annealed duplex may dissociate from the enzyme if packaging does not proceed in timely fashion. In the absence of procapsids in vivo, the packaging pathway stalls at the strand-separation complex and the nicked duplex is slowly released from the enzyme (fig. 10). Repair of the nicks by host ligase results in an apparent lack of cos-cleavage in vivo. We suggest that procapsid binding to T•D_L in the strand separation complex draws the intermediates towards packaging by mass action, thus avoiding significant dissociation of nicked DNA from the complex. This model may also explain why complex I (T•D_L) is efficiently formed in vitro in the absence of procapsids. A key point is that elevated concentrations of terminase are utilized in the in vitro assays. This would have the effect of driving the reaction forward by mass action, yielding complex I.

Figure 10

Model for gpFI and Procapsids in the cos-Cleavage Reaction. Details are described in the text.

Reversibility of the cos cleavage reaction has been previously invoked to explain related phenomena. For instance, certain lethal mutations in the R sequences of cosB were found to completely block cos cleavage in vivo, and yet only mildly affect cos cleavage in vitro.⁶² To resolve the discrepancy, it was proposed that the cosB mutations primarily affect the formation and/or stability of a post-cleavage intermediate; failure to form the intermediate leads to DNA dissociation from terminase, reannealing of the cohesive ends, and religation of the nicked duplex in vivo.⁶² The post-cleavage intermediate is presumably the strand-separation complex shown in Figure 10. To further emphasize this point, a series of suppressor mutations to the cosB mutant phage were isolated and characterized; all of the pseudorevertants contained mutations in gpNu1.⁶³ Though it was anticipated that the mutant gpNu1 terminase enzymes would be more efficient in utilizing the mutant cosB DNA as a substrate in the cos-cleavage reaction, this was not the case. The mutant enzymes were no more efficient than wild type terminase at cutting cosB mutant DNA; rather, the mutant enzymes were more efficient than the wild type enzyme in packaging of the cosB mutant DNA. Here again it was argued that the Nu1 mutations increased the efficiency of a post-cleavage transition that was made inefficient by the cosB mutations.¹²¹

The proposed reversibility of cos cleavage in vivo depends on ligation to reseal the nicks of annealed cohesive ends. It is interesting that in phage P2, which also has cohesive ends, efficient production of cohesive ends is similarly procapsid-dependent.¹²² In contrast are phages P22 and SPP1 which use a head-full packaging mechanism and do not produce cohesive ends. For concatemer processing by these phages, an initial cut is made at a defined site called pac. Following pac cleavage to initiate packaging, downstream cutting occurs nonspecifically upon filling the procapsid with DNA; the resulting virion chromosomes lack cohesive ends. For both P22 and SPP1, pac site cutting is procapsid-independent.^123,124 Thus, based on this limited sample, there is a correlation that only in phages with cohesive ends is concatemer cutting procapsid dependent. DNA cutting may be irreversible when the concatemer cleavage products lack cohesive ends.

Model for gpFI Requirement in Vivo

As described above, phage with mutations in the FI gene also exhibit an apparent lack of cos-cleavage activity in vivo. The gpFI protein stimulates virion assembly in vitro, specifically when terminase and procapsids are at dilute concentrations.^117,125,126 The authors suggested that gpFI increases the efficiency of procapsid binding to the terminase proteins assembled at cos.^90,117 We have come to a related conclusion by examining the effect of gpFI on the cos-cleavage reaction in vitro, and have suggested that gpFI promotes cos-clearance.¹²⁰ In this model, procapsids are necessary, but not sufficient to promote cos-clearance and gpFI stimulates the process, as shown in Figure 10.

While the mechanism for gpFI action remains obscure, genetic studies provide insight into the nature of the protein-protein interactions between terminase, gpFI and procapsids. λFI_amber mutants are “leaky”, which means that the virus yield is 10^-3 compared to wild type virus. This is in contrast to the ≈10^-5 reduction caused by nonsense mutations that inactivate a gene encoding a virion structural component. Bypass mutations, called fin mutations, allow λFI to form plaques in the absence of gpFI. Two types of fin mutations have been described. The first are finA mutants that have changes affecting terminase, and generally cause the production of increased levels of the gpA protein.^112,113 An increase in the concentration of gpA is expected to increase the in vivo concentration of terminase, which may increase the flux towards packaging, independent of gpFI (see fig.10). This is similar to the mechanism proposed above for the formation of complex I in vitro. The second are finB mutations that alter a 26-residue-long segment of gpE, which has been termed the Efi domain.¹¹² It has been proposed that the Efi domain is located on the surface of the procapsid and that complex I directly interacts with the Efi domain en route to docking at the portal. In this model, gpFI directly modulates this initial interaction and guides the procapsid-terminase docking step.¹¹³ An alternative model suggests that the Efi domain is, in fact, part of the gpE protein retained in the portal pX proteins (see fig. 7). In this model, the Efi domain represents a portal docking site that makes direct contacts with terminase in complex I. The finB mutations presumably allow this interaction in the absence of gpFI.

The Fate of IHF and gpFI during cos-Clearance

One final point to consider is the fate of IHF and gpNu1 assembled at cos during cos-clearance. Both of these proteins bind to specific sites in cosB, and are likely central to the stability of complex I. It thus clear that a major alteration in protein-DNA binding interactions must take place prior to translocation. IHF most likely dissociates from complex I to allow passage of gpA and the procapsid, as shown in Figure 9. Indeed, based on the effect of gpFI on the cos-cleavage reaction in vitro, we have suggested that gpFI may act antagonistically to IHF in modulating the stability of complex I.¹²⁰ The fate of gpNu1 is less clear. One possibility is that gpNu1 switches from a site-specific DNA binding protein to a nonspecific DNA binding protein that is an active part of the packaging motor (shown in fig. 9). This model is consistent with the observation that small terminase subunits from bacteriophage T4 and SPP1 form oligomeric rings in solution, which may indicate their role as a “sliding clamp” during translocation.^127,128 It is also feasible that gpNu1 releases DNA, but remains bound to gpA and simply “goes along for the ride”. A final possibility is that gpNu1 is ejected from the translocating complex. This model suggests that the role of gpNu1 is to site-specifically assemble gpA at cosN, and maintain the integrity of complex I until the procapsid arrives. Thus, gpNu1 plays a role analogous to that of sigma factors and transcription factors in the assembly of the transcription complex. Once the appropriate nucleoprotein complex has been assembled, movement of the protein machinery results in ejection of the assembly protein(s) from the DNA. Again, the answers to these questions must await further experimentation.

The Translocation Complex

Subsequent to cos-clearance, the packaging motor translocates DNA into the confines of the procapsid interior, a reaction fueled by the hydrolysis of ATP. The components and stoichiometry of the λ translocation machine remain unknown, but we speculate that a hexameric gpA ring is arrayed in a polar fashion in direct contact with the gpB^* dodecamer of the portal complex. A hexameric gpA structure is based on a variety of data; (i) studies in phages T3 and Φ29 have demonstrated the presence of 6 copies of the respective large terminase subunit in the packaging complex,^71,129 (ii) electron microscopy has shown that the small terminase subunits of T4 and SPP1 terminase enzymes form oligomeric ring structures,^127,128 and (iii) recent electron micrographs of λ holoenzyme clearly demonstrate that the purified enzyme forms rings in solution (E. Bogner, unpublished). A hexameric gpA ring is pleasing in that it accommodates symmetric interactions with a dodecameric portal complex, and shows mechanistic similarity to the translocating hexameric helicase enzymes.^86,88 It is noteworthy that a gpA hexamer in the packaging complex has important implications for the nature of the activated nuclease complex described above. It has been presumed that a symmetrically disposed gpA dimer is responsible for duplex nicking. This would require recruitment of additional gpA subunits to complex I prior to translocation in order to complete the gpA hexamer. Alternatively, a symmetrically disposed gpA hexamer may be the relevant nuclease complex. The answers to this dilemma must await further experimentation.

The Packaging ATPase

It is presumed that ATP hydrolysis provides the energy required for inserting DNA into the confines of the capsid, and purified terminases from a number of viruses possess ATPase activity.¹³⁰ There is general agreement that, for the tailed dsDNA phages, the ATPase center resides in their respective large terminase subunits. Consistently, sequence analysis of terminase enzymes from phage to the herpesviruses indicates the presence of conserved ATPase motifs in these proteins.^69,78 In λ, a number of studies link the ATPase site located in the N-terminus of gpA to DNA translocation (summarized above).^74,75,77

Active DNA packaging in the phage T3 and Φ29 systems consumes about one ATP per two base pairs packaged. The observed stoichiometry in the λ system is two ATP's consumed per base pair packaged, though this may represent a significant overestimation.⁶¹ If we take the conservative estimate, packaging of a λ genome (48,502 bp) would require the hydrolysis of ≈24,250 moles of ATP. Packaging requires 2-3 minutes in vivo,^14,60 so the estimated ATPase rate would be ≈10,000 min^-1 per packaging motor. This is significantly greater than the basal rate of 50 min^-1 for gpA in the holoenzyme.^60,70 Recent studies have demonstrated a packaging-specific ATPase activity that hydrolyzes ATP at a rate of 600 min^-1.⁶¹ While this value is still lower than expected, it is clear that assembly of the translocation complex must activate a translocation ATPase catalytic site, presumably the N-terminal ATPase site in gpA; however, present data do not rigorously exclude the presence of a cryptic ATPase site that is located elsewhere in terminase holoenzyme, in the portal complex, or that is formed by interaction of the two structures.

Models for DNA Translocation

Once released from the cos site, the packaging machinery translocates along the duplex, actively inserting DNA into the capsid. How might such a translocation machine workλ The goal of understanding a biological motor at the molecular level challenges us to mechanistically link the energy of ATP hydrolysis to physical changes in protein structure that lead to translocation. Significant progress has been made towards our understanding of a number of motors, including myosins, kinesins, and the rotary F₁F₀ and flagellar motors.¹³¹ Viral DNA packaging motors have been particularly difficult to dissect at the molecular level, however, for the following reasons. (i) With very few exceptions, a complete list of the components of the motors remains speculative. It is likely that both terminase proteins and portal proteins assemble to complete a translocating motor, but the exact nature of the complex is unknown in all cases. (ii) Once assembled, the motors act quickly and transiently. Dissociation of the components occurs upon completion of the packaging process. (iii) There is little structural information available for any of the components of these machines, and none on the actively packaging motors.

Nevertheless, it is clear that there are three parts to the motor: terminase, the portal vertex, and the procapsid shell. There are a number of creative models describing how the components of the packaging complex sponsor DNA movement. The first proposes that the terminase enzyme is directly responsible for translocation of DNA into the capsid. These models propose that the terminase subunits physically translocate via flexible, DNA-contacting domains that cyclically contract, and then undock from the DNA.^94,130,132 The conformational changes required in this model are presumably driven by the hydrolysis of ATP. These models find mechanistic similarity to the “inch worm” mechanism proposed for translocation of hexameric helicases.^86,87 A second class of models is portal-centric, which rely on the concept of a symmetry mismatch between the twelve-fold symmetry of the portal and the five-fold symmetry of the capsid shell vertex to which it is attached. In one version, which is actually the earliest model, ATP hydrolysis drives rotation of the portal protein with respect to the capsid shell, which “screws” the DNA into the capsid.¹³³ A more recent portal model is inspired by the structure of the Φ29 portal (Anderson and Grimes, Chapter 7 of this work). This model proposes that the procapsid-terminase complex (which possesses ATPase activity) acts as a stator, the DNA as a spindle, and the portal complex as a ball-race. The DNA helix is proposed to convert the rotary action of the portal into translocation of the DNA.¹³⁴ In the third class of models, the capsid shell acts as a reciprocating gated pump that pulls the DNA into the procapsid interior¹³⁵ (Serwer, Chapter 4 of this work).

For the λ system, available evidence is most consistent with the first class of models in which terminase subunits constitute the DNA translocating motor, though other models are not rigorously excluded. The strongest evidence comes from genetic studies in which lethal mutations in the virus were screened to identify those that affected DNA packaging.⁷⁴ Ten of these mutants were characterized, and all had mutations that mapped to a gene A segment extending from codons 18 to 349 (summarized above). These mutants provide evidence for the involvement of the terminase gpA subunit in translocation and active DNA packaging. A detailed biochemical analysis of these mutants in vitro will undoubtedly provide further insight into the role of gpA in the packaging motor.

Capsid Expansion and the Phage gpD Protein

Translocation of the packaging machinery along the duplex results in packaging DNA into the procapsid. Upon packaging of 10-50% of the viral genome, the procapsid undergoes an expansion process whereby the 25 nm spherical capsid expands to a radius of 32 nm, roughly doubling its volume and acquiring its mature icosahedral shape.^136,137 The molecular basis for expansion is not known. The viral gpD protein, which is a monomer in solution, adds to the surface of the expanded icosahedral capsid as a trimer, and in numbers similar to the major capsid protein (fig. 9).^138,139 The presumed role of gpD is to stabilize the partially filled capsid, and prevent DNA release.^136,138-140 Indeed, D phages are viable if the viral chromosome is less than 80% of wild type chromosome length, and if the virions are kept in medium supplemented with 10^-2 M Mg⁺⁺.¹⁴⁰ Within this context, an interesting chimeric phage is λ shp, a phage in which the D gene has been replaced by the analogous shp gene of phage 21. The Shp protein has a lower affinity than gpD for the λ gpE shell, so that 15-20% of the binding sites on λ shp capsid shells are unfilled. λ shp has a stability intermediate between λ wild type and λ D, such that λ shp virions with full-length chromosomes require supplemental Mg⁺⁺ ions (10^-2 M) for viability. The intermediate stability of λ shp virions indicates that the shp protein imparts considerable stability to the capsid shell, even though the shell is not completely populated by Shp.¹⁴¹

cosQ and Recognition of the Terminal cos Sequence

Termination is a complex process that likely involves significant modification of the translocation complex. Proper termination requires that the downstream cos contain cosQ, cosN and the I2 segment.¹¹ When cosQ is mutant, there is a failure to nick the bottom strand of cosN, and translocation does not arrest. Rather, packaging continues so that the singly-nicked duplex is further inserted until the capsid is filled to capacity; this is a lethal event because the protruding, uncut DNA prevents virus maturation (vide infra).²⁶ Our interpretation of the bottom strand nicking failure of λ cosQ^- is based on symmetry issues, as follows. We presume that the gpA oligomer (hexamerλ) responsible for translocation in the packaging motor is arranged in a unidirectional manner through interactions between the C-termini of gpA subunits and the gpB^* portal protein.^24,26 Conversely, the symmetry of cosN suggests that symmetrically arranged (head-to-head) gpA subunits are required to nick each cosN half-site.^24,26 Thus, reorganization of the unidirectional gpA complex in the packaging motor is required to assemble the symmetric nicking complex required to cut the terminal cos site. We propose that cosQ plays a direct role in this reorganization, and is required to sponsor the presentation of a properly oriented terminase complex to the cosN half-sites. The mechanism for this is obscure, but could involve a bone fide reorganization of the proteins in the translocation complex, and/or recruitment of additional proteins from solution.^24,26

The packaging protein that binds cosQ has not been identified, despite extensive studies of pseudorevertants of cosQ mutants. While not revealing the cosQ binding protein, studies of nonallele specific cosQ suppressors indicate a relationship between the extent of DNA packaged and the efficiency of recognition of mutant cosQ sites, as follows. Mild cosQ mutations can be suppressed by (i) lengthening the chromosome, and/or (ii) missense mutations altering the portal protein, gpB.^22,23 It is argued that chromosome lengthening or altering gpB slows the translocation rate at the time the downstream cos is encountered by the packaging motor; this may increase the efficiency of recognition and cleavage of a mutant downstream cos site. While there is no direct information about the relationship between chromosome length and the rate of λ DNA translocation, data obtained in the phage Φ29 system indicates that translocation rate declines dramatically as the end of the chromosome is approached.¹⁴²

The Extent-of-Packaging Sensor

The efficiency with which a wild type downstream cos is cut to terminate packaging also depends on the length of DNA to be packaged. When the genome is shortened to lengths <80% of wild type, cos cleavage efficiency in vivo declines sharply.²² The cos-cleavage reaction is insensitive to duplex length in vitro,¹⁰ suggesting that the terminal cos cleavage reaction in λ is affected by the amount of DNA packaged into the capsid. This is similar to phages such as T4, P22 and SPP1 that use a head-full activated DNA cleavage mechanism for chromosome processing (Chapters 3, 5 and 6, respectively, of this work). Genetic observations in the phage P22 and SPP1 systems have lead these researchers to propose that the packaging motor has a “sensor” that detects when the capsid is full of DNA, which then activates the endonuclease catalytic site.^143,144 This putative sensor might detect (i) the extent of DNA packaged, (ii) the rate of DNA translocation, and/or (iii) the energy required for translocation. For λ terminal cos cleavage, we propose that the efficiency of cosQ recognition, and hence cleavage, is inversely dependent on translocation rate. Thus, cosQ recognition is proposed to represent the sensor.

Processive Genome Packaging

Following the downstream cos cleavage event, the terminase proteins dissociate from the portal complex of the DNA-filled capsid, but remain bound to the matured (D_L) end of the next genome in the concatemer (fig. 11). This terminase-DNA complex, which is functionally related to complex I, captures a procapsid to sponsor the packaging of the next chromosome in the concatemer. While the cosB subsite is not essential for the terminal cos cleavage reaction, it is required for processive genome packaging. Surprisingly, however, some mutations of cosB that are defective in the initial cos cleavage reaction retain a processive packaging phenotype; these include a triple point mutant cosB R3 R2 R1 allele, and cosBΦ21, a λ genome with the cosB of phage 21.¹⁶ These data indicate that the terminase-DNA interactions required for processive packaging are “relaxed” in comparison with the requirements for the initial assembly of terminase at cos (see packaging initiation, above). Structural studies on these complexes are needed to understand the DNA site requirements for each of the packaging complexes.