Specialized Nucleoprotein Structures at the Origin of Replication of Bacteriophage X PROTEIN ASSOCIATION AND DISASSOCIATION REACTIONS RESPONSIBLE FOR LOCALIZED INITIATION OF REPLICATION*

Binding of the 0 protein of phage X to the replication origin (oriX) results in the formation of an organized nucleoprotein structure termed the 0-some. The 0- some serves to localize and initiate a six-protein sequential reaction that provides for localized unwinding of the origin region, the critical prepriming step for precise initiation of DNA replication. By the use of electron microscopy of gold-tagged antibody complexes, we have defined four stages of protein associ- ation and dissociation reactions that are involved in the prepriming pathway. First, as defined previously, 0 protein binds to multiple DNA sites and self- associates to form the 0-some. Second, AP and host DnaB proteins add to generate an O*P*DnaB.oriX and containing and acts a gle replication of disassembly activity of Bacteriophage

functional analogs of Escherichia coli DnaA and DnaC proteins, serving to divert and direct the host replicative machinery to orih via a series of protein-protein interactions (3). The 0 protein associates with the repeated regions of oriX ( 5 , 16), most likely binding as a dimer at its amino terminus to each inverted repeat (6, 17, 18). The interaction of 0 with orih results in more extensive bending of the repeated region and induces helix destabilization of the adjacent A/T-rich region (15). The P protein interacts with the carboxyl terminus of the 0 protein, and P also associates with the host DnaB protein to form a P .DnaB complex (17, 19-25). The interaction of the P . DnaB complex with 0 directs other host proteins, including DnaG, DnaJ, and DnaK, to the initiation of DNA replication at oriX (24, 26-31). Hence, the association of 0 with oriX initiates a complex series of protein-protein interactions that culminate in origin specific initiation of DNA replication.
In vitro DNA replication systems that depend upon these proteins to replicate plasmids containing the X origin have been developed (27, 28). More recently, a "prepriming" reaction has been defined with purified 0, P, DnaB, DnaJ, DnaK, and single strand DNA-binding protein (SSB)' that generates a DNA-protein complex capable of initiating DNA replication upon the addition of DnaG primase and DNA polymerase 111 holoenzyme (31, 32). We have recently used electron microscopy as a means of elucidating this prepriming pathway (33, 34). Based upon the morphological appearances of nucleoprotein structures formed at oriX, we have defined three stages for the prepriming reaction. First, dimeric 0 binds multiply to the directly repeated sequences of oriX and self-associates to form a nucleoprotein structure termed the 0-some. Such DNA-wound multiprotein complexes have been termed specialized nucleoprotein structures, or "snups" (35). Second, P and host DnaB proteins interact with the 0-some to generate a larger nucleoprotein entity that includes additional DNA from the A/T-rich domain of oriX. Third, the addition of DnaJ, DnaK, and SSB proteins and ATP results in an originspecific unwinding reaction, presumably catalyzed by the helicase activity of DnaB (30,34). The unwinding is unidirectional, proceeding "rightward" from the origin. The minimal DNA sequence competent for unwinding consists of two 0 binding sites and the adjacent A/T-rich region to the right of the binding sites. The three-stage pathway to unwinding of oriX appears to resemble that of the E. coli host, in which DnaA, DnaC, and DnaB proteins interact at the E. coli origin to generate an unwound prepriming structure (36,37). However, the two systems differ in that unwinding from the host 'The abbreviations used are: SSB, single strand DNA-binding protein; Hepes, 4-(2-hydroxyethyl)-l-piperazineethanesulfonic acid.
origin, oriC, is bidirectional and requires neither DnaJ or DnaK (36,37). Specialized nucleoprotein structures (snups) appear to function in many instances to localize and control replication origins and site-specific recombination (35).
The simple rotary shadowing technique that successfully permitted visualization of the localized three-stage prepriming reaction does not reveal directly the actual protein composition of the latter two stages. Nor does it provide any direct clue as to the function of DnaJ and DnaK in generating the locally unwound prepriming structure, other than to show that these proteins are required for unwinding to occur. These uncertainties must be addressed in order to formulate a comprehensive model describing the pathway to the locally unwound prepriming complex.
In this study we combine electron microscopy with antibody labeling to analyze the protein composition of the nucleoprotein structures at each stage of the prepriming pathway. We show that the second stage in the pathway leading to localized unwinding is generated by the interaction of a P . DnaB protein complex with the 0-some. We also show that at least one more additional stage, prior to unwinding of the DNA, occurs by the addition of DnaJ and DnaK proteins to the second stage 0. P . B . oriX structure and that DnaJ and DnaK act to remove P from the nucleoprotein complex, presumably freeing DnaB to catalyze origin-specific unwinding. A possible general function for DnaJ and DnaK in E. coli is also considered. The accompanying papers by Alfano and McMacken (66,67) describe a complementary analysis of the prepriming pathway that leads to essentially the same conclusions.
DNA-The plasmid pRLM4, which contains the replication origin of phage A, has been described (38). Ml3oriAright and its deletion derivatives have also been described (10); the deletions used in this work are at positions 39,160 and 39,131 on the A map (44). Plasmid attP DNA, which contains the binding sites for the A Int protein, and pSVO1, which contains the origin and binding sites for SV40-T antigen, have been described (45,46).
Binding of Proteins to DNA and Electron Microscopy-The reaction mixtures (20 pl) contained the following: Hepes/KOH at pH 7.6, 25 mM; magnesium acetate, 11 mM; KCl, 50 mM; NaCI, 30 mM; ATP, 2 mM; supercoiled pRLM4 or Ml3oriA DNA, 215 ng; A 0 protein, 51 ng; AP protein, 108 ng; DnaB protein, 103 ng; SSB protein, 617 ng; DnaJ protein, 60 ng; DnaK protein, 5 pg. The protein concentrations and DNA-protein stoichiometry are sufficient for the initiation of in vitro replication (32). The X Int or T antigen binding reactions were carried out in the same buffer used for the A replication proteins; reaction mixtures contained either 240 ng of pSVOl DNA, 1.6 pg of T antigen, and 4 mM ATP, or 180 ng of attP plasmid and 120 ng of Int protein. All reaction mixtures were assembled on ice. Mixtures were incubated for 10 min at 30 "C, followed by the addition of glutaraldehyde to 0.1% and incubation for an additional 15 min at 30 'C. After glutaraldehyde fixation, the reaction mixtures were passed through a Sepharose 4B column (3 X 0.5-cm) that had been previously equilibrated with NsiI restriction enzyme buffer (Tris-HCl, pH 8.4, 10 mM; NaCl, 150 mM; magnesium chloride, 10 mM). Fractions (7 pl) were collected, and peak fractions were pooled (40 pl) followed by the addition of 5 units of NsiI restriction enzyme and between 0.5-2.6 pg of IgG. (The quantity of IgG added to the mixtures was such that it did not cause aggregation of the nucleoprotein complexes; the correct quantity was determined empirically for each antibody). The mixtures were then incubated for 30 min at 37 'C and passed again through a Sepharose 4B column (3 X 0.5-cm) that had been equilibrated with protein A-binding buffer (Hepes/KOH, pH 7.6, 40 mM; KCl, 100 mM; magnesium acetate, 11 mM), and peak fractions were pooled (40 pl). One p1 of protein A-colloidal gold particles (7 pg/ml, 10-nm diameter gold) was added, and the mixtures were incubated for 1 h at room temperature. The mixtures were passed again through a Sepharose 4B column (4 X 0.5-cm) (previously equilibrated with Hepes/KOH at pH 7.6, 40 mM; magnesium acetate, 11 mM) and examined by electron microscopy using the polylysine technique of Williams (47).

RESULTS
An 0 . P . D m B . oriX Complex Results from the Interaction of a P -D m B Complex with the 0-some-0 protein specifically interacts with orih to generate a specialized nucleoprotein structure (snup) termed the 0-some (33). P and DnaB proteins interact with the 0-some to generate a different snup that is larger and more asymmetric than the 0-some. Genetic and biochemical evidence has indicated that P protein interacts with both 0 and DnaB (3). Therefore, the nucleoprotein structure formed by the interaction of 0, P, and DnaB at oriX was presumed to contain all three of these proteins (33). However, this presumption cannot be demonstrated directly by electron microscopic examination of the rotary shadowed nucleoproteins. Since P is a small protein (Mr = 26,500), an 0. P. oriX complex, visibly indistinguishable from the 0-some (Mr = 400,000), might be formed prior to the addition of DnaB to the structure (Fig. 1). Alternatively, P and DnaB might displace 0 from oriX and generate a P .DnaB. oriX complex or even a DnaB.oriX complex. Finally, P might deliver DnaB to the 0-some and generate an 0. DnaB . oriX structure ( Fig. 1).
In order to determine which of the above possibilities is correct, an immunoelectron microscopic technique was used to analyze qualitatively the protein composition of the specialized nucleoprotein structures formed by the interactions of various combinations of 0, P, and DnaB proteins with oriX DNA. This analysis of protein composition was accomplished by first generating the nucleoprotein structures and then probing with antibody specific for the protein in question, followed by tagging the antibody with protein A-colloidal gold conjugate (e.g. Refs. 36,48). Fig. 2 illustrates the technique. The 0 protein was reacted with oriX DNA on the plasmid pRLM4 to form 0-somes, which were then probed with anti-0 IgG and protein A-gold conjugate. On each plasmid molecule a cluster of gold was bound at a single site only (Fig. 2a). Cleavage of the plasmid with NsiI restriction enzyme, which cuts at a site approximately 700 base pairs from orih, demonstrated that the gold was bound at the origin and nowhere else on the DNA (Fig. 2b). The gold labeling was specific and efficient. Under the conditions used, greater than 95% of the 0-somes were bound by gold when probed with IgG specific for 0 protein (Table I, line 1). Probing 0-somes with IgG specific for either P or DnaB proteins yielded a signal close to the background level observed for mixtures in which antibody was omitted (Table I, line 1). In the absence of DnaB, there was no clear interaction of P with the 0-some (Table I, line 2), thus ruling out the formation of a stable 0 .P .or& intermediate. In the absence of P, DnaB protein also did not associate stably with the 0-some (Table I,  Possible nucleoprotein structures generated from the interaction of 0 , P, and DnaB proteins at oriX. The interaction of these three proteins at the origin of replication could generate one of several possible nucleoprotein entities whose protein composition would be indiscernible by electron microscopic examination of the rotary shadowed complexes. P and DnaB were both added to reaction mixtures containing 0 and oriX was a clear response seen for each protein when probed with specific antibody (Table I, line 4). (The lower response for probes specific for P and DnaB as compared with 0 is consistent with the observation that formation of 0.P. DnaB.oriX structures is only 60% efficient (33).) 0-somes formed with 0 protein truncated for the carboxyl domain failed to bind DnaB ( Table  I, Table I, we conclude that the second stage of the pathway to a prepriming structure is a snup of 0, P, and DnaB at oriX (Fig. 3).
Formation of 0. P . DnaB.oriX Complexes on DNA Substrates Deleted for the A/T-rich Region-The minimal region of oriX competent for generating a locally unwound prepriming structure consists of a t least two of the direct repeats plus the adjacent A/T-rich region (34). As judged by electron microscopy, deletion of the adjacent A/T-rich region from oriX abolishes the unwinding reaction (33) but does not adversely affect the first stage in the unwinding pathway, the genesis of an intact 0-some (data not shown). The second stage 0 . P. DnaB . oriX structure, which arises from the association of one or two P.DnaB complexes with the 0-some, includes much of the A/T-rich region (33). T o study whether the A/T-rich region contributes to the genesis of the second stage 0 . P . DnaB . oriX structure, we used immunoelectron microscopy to quantitate the extent of 0 . P . DnaB . oriX snup formation on two different oriX substrates deleted for part or nearly all of the A/T-rich region. The smaller of the two deletions was still completely competent for the assembly of FIG. 2. Tagging of the 0-some with anti-0 IgC and Staphylococcus protein A-gold conjugate. a, gold-tagged 0-somes on supercoiled pRLM4 plasmid DNA. b, gold-tagged 0-somes on pRLM4 DNA cut with NsiI restriction enzyme.

TABLE I
Characterization of protein composition of specialized nucleoprotein structures by immunoelectron microscopy 0, P, and DnaB proteins were mixed with plasmid DNA containing the X origin of replication. After glutaraldehyde fixation, the resultant nucleoprotein structures were probed with either anti-0, anti-P, or anti-DnaB IgG and then secondarily labeled with protein A-gold conjugate. The percentage of nucleoprotein molecules bound by gold was determined by electron microscopy. Between 100 and 300 molecules were scored for each category. Nonspecific background binding of the protein A-gold conjugate to nucleoproteins was 7% or less in control experiments in which the antibody was omitted. ATP included in reaction mixture. 0 / 2 is 0 protein lacking its carboxyl-terminal domain.
an 0 . P.DnaB .oriX snup; this nucleoprotein complex supported formation of a locally unwound prepriming structure and in uitro DNA synthesis (10, 34) (   Effect of deletions in the A/T-rich region of oriX on 0. P . DnaB . oriX complex formation 0, P, and DnaB proteins were mixed with DNA containing either a wild type X origin (Ml3orih right, uppermost bar) or with X origins deleted for part of the A/T-rich region (M13oriXA39160, M13oriXA39131, lower two bars). The extent of 0.P.DnaB.oriX complex formation was assessed by probing the glutaraldehyde-fixed nucleoprotein structures with anti-DnaB IgG and protein A-gold conjugate. The data are compared with in vitro replication and unwinding for the same substrates (10,34). The number of complexes scored was 400 for each DNA construct. Separate counts of 100 were done and averaged. Data are presented as averages of those separate counts k standard deviation. Nonspecific background binding of the protein A-gold conjugate to nucleoprotein structures was 2 +-1% when antibody was omitted from reaction mixtures.
Origin of replication  ' ATP included in reaction mixture.
' ATPyS included in reaction mixture.

TABLE IV
Association of DnaK with 0-, P-, and DnuB-mediated nucleoprotein structures Plasmid DNA containing oriX DNA was mixed with the various combinations of 0, P, DnaB, DnaJ, and DnaK proteins and allowed to form nucleoprotein structures. The glutaraldehyde-fixedcomplexes were then probed with anti-DnaK IgG and tagged with protein Agold conjugate. Separate counts of 100 were done and averaged.   and DnaK proteins and the mixtures were supplemented with either SSB protein or ATP or both. The glutaraldehyde-fixed complexes were probed with either anti-0, anti-P, or anti-DnaB IgG, tagged with protein A-gold conjugate, and then examined by electron microscopy. Separate counts of 100 were done and averaged.

Additions % snups
No. Two classes of nucleoproteins are observed when ATP and SSB are included in the reaction mixture. Data for the spherically shaped complexes.
Lower row is the data for the unwound prepriming complexes.

TABLE VI Association of DnaK with various nucleoprotein structures
Nucleoprotein structures were formed and fixed with glutaraldehyde. The structures were then treated with anti-DnaK IgG and tagged with protein A-gold conjugate. 012 is the 0 protein truncated for its carboxyl domain. T antigen is the SV40 virus-encoded large T antigen. oriSV40 is the origin of DNA replication for SV40. Int is the integrase protein of phage X. att P is the recombination and binding site for Int protein. "ATP was included in the reaction mixture. This was necessary for the formation of a large nucleoprotein structure at the SV40 origin.
for either 0, P, or DnaB (Table V, lines 3 and 4). Among the unwound structures (line 4), the number that scored positive for P protein was similar to that for nucleoprotein complexes formed in the absence of SSB. However, the number that scored positive for DnaB was approximately 90%, suggesting that P is probably ejected from the nucleoprotein complex by DnaJ and DnaK, allowing DnaB to carry out its helicase activity. For the non-unwound structures (line 3), the DnaB appears to have been largely lost, indicating that unwinding fails to occur because of an abortive release of the helicase.
Interestingly, there is a substantial drop in the number of molecules scoring positive for 0 protein in the unwound prepriming structure. We conclude that DnaJ and DnaK act to partially disassemble the O.P.DnaB snup, freeing DnaB to mediate localized unwinding.
DnaK Binds to Other Nucleoprotein Structures-A possible general function for heat shock proteins is the elimination of protein aggregates that accumulate in the cell either normally or during periods of stress (53). The 0-some and 0 . P. DnaB .
oriX complex can be thought of as such an aggregate, requiring disassembly for the initiation of DNA replication (and possibly for normal transcription and packaging). The data of Tables IV and V indicate that DnaK can associate with the 0-some before the addition of the P and DnaB proteins that are the eventual targets of the disassembly reaction. Thus, nonspecific billding of DnaK to a multiprotein aggregate might be the first stage in the disassembly pathway. To test this hypothesis, the binding of DnaK to different nucleoproteins was tested (Table VI). DnaK associates with O-somes formed by the truncated 0/2 protein, though not as well as with 0-somes generated from full-size 0. Although there was a high background of nonspecific binding of the protein Agold to the Intasomes, there was a 2-fold increase in the binding of gold when DnaK was included in the reaction mixture. Thus, DnaK appears to bind snups generated by the interaction of X Int protein with attP. Although the extent of interaction is slightly less than for the other nucleoprotein complexes tested, DnaK also associates with the snup generated by the interaction of SV40-T antigen with the SV40 origin of replication. Thus, DnaK may interact generally with large protein aggregates. However, our data do not exclude the possibility that DnaK simply interacts with many proteins.

Stages Leading to the Formation of a Locally Unwound
Prepriming Structure-Based on electron microscopic evidence, we have recently proposed a sequential three-stage pathway leading to the formation of a locally unwound prepriming structure competent for the initiation of DNA replication (34,35,54). However, due to the limitations imposed by rotary shadowing of the nucleoprotein complexes, we were unable to discern the actual protein composition of the complexes by ordinary electron microscopy. Knowledge of the protein composition of these structures is crucial for assignment of function to each of the proteins participating in the proposed pathway leading to the prepriming structure. Immunoelectron microscopy has proven useful in revealing the sequence of protein association and dissociation reactions and has allowed us to define a fourth stage in the pathway. The combination of gel electrophoresis and antibody labeling described in the accompanying article has yielded similar information and conclusions. The 0-some is believed to localize the initiation of DNA replication by serving as a recognition site and foundation for the assembly of a multiprotein initiation complex. Direct evidence in support of this idea is the observation that the addition of P and DnaB proteins along with 0 protein results in the formation of a nucleoprotein complex localized at oriX containing all three proteins. The genesis of this second-stage snup presumably occurs via interaction of the exposed carboxyl domains of the oriX-bound 0 proteins with one or more P .DnaB complexes. At least one DnaB is then positioned over the A/T-rich region, where the DNA is presumably partially unwound to give rise to an "open complex" that is gripped by the single-stranded binding domain of DnaB protein (30, 55, 56) (Fig. 3). Although the A/T-rich region is required for unwinding and DNA synthesis to occur, it is not required for assembly of the 0 . P . DnaB . oriX complex. This observation further suggests that the 0-some itself is the primary recognition site for the assembly of a multiprotein initiation complex. However, the observation that either a superhelical substrate or an intact A/T-rich region enhances formation of the 0 . P. DnaB .oriX complex suggests that DnaB is associating with single-stranded DNA in the A/Trich domain of oriX. Zahn and Blattner (4) have proposed that the free energy of bending of oriX DNA is trapped in the 0-some and is used to locally unwind the adjacent A/T-rich domain. Schnos et al. (15) have recently adduced evidence that supports this idea. They observed that after formation of the 0-some, the adjacent A/T-rich domain of oriX became sensitive to cleavage by single strand specific nucleases. Furthermore, this nuclease sensitivity required DNA supercoiling and was reduced in mutants deleted for portions of the A/Trich region. Thus, the assembly of an 0-some apparently serves to partially open the A/T-rich domain of oriX prior to addition of the P. DnaB complex.
After formation of the 0 . P . DnaB . oriX snup, the DnaB is presumably trapped in an inactive state because of its tight association with P (22)(23)(24)(25). If the locally unwound prepriming structure is to be generated, the linkage between P and DnaB must be broken so that helicase activity is restored to DnaB, which then initiates origin-specific unwinding. This task is performed by the host DnaJ and DnaK proteins. In the third stage of the pathway to the prepriming structure, DnaK associated with 0 protein joins DnaJ, which associates with either P or P and DnaB (Fig. 3). These events are then followed by an ATP-dependent disassembly of the 0 . P. DnaB.oriX complex in which P, and possibly some 0, are ejected. Recently Liberek et al. (65) have used a complementary approach with radioactive P protein to demonstrate the release of P from the 0 . P. DnaB . oriX complex through the action of DnaJ and DnaK. The liberated DnaB, which has associated with single-stranded DNA from within the A/Trich region, begins hydrolyzing ATP and migrates away, unwinding duplex DNA as it moves. SSB protein maintains the single-stranded state so that primase and polymerase I11 holoenzyme can add to the prepriming structure and initiate replication. Gel electrophoresis experiments described in the accompanying article and our preliminary electron microscopic experiments indicate that DnaJ is also ejected from the locally unwound complex.
In E. coli the prepriming pathway is highly similar to that of X. The binding of DnaA protein to multiple sites in oriC generates a snup at which there is also an adjacent A/T-rich region (57-59). The DnaA-mediated nucleoprotein complex serves as an assembly target for formation of a DnaA/DnaC/ DnaBloriC structure which is the precursor for the prepriming unwinding reaction (57-59). After formation of the DnaA snup, the A/T-rich region is sensitive to cleavage by a single strand DNA nuclease, which suggests that the DnaA-mediated nucleoprotein assembled at oriC somehow alters the conformation of the adjacent A/T-rich DNA. However, the DnaAloriC complex differs from that of oriX in that ATP is required for helix destabilization of the A/T-rich region (58). Function of DnaK in E. coli-The role that DnaK plays in replication of host DNA, if any, is not clear. DNA replication from oriC in vitro does not require DnaK (60). However, studies of DnaK mutants suggest an involvement of DnaK in initiation at oriC (61). Our data, garnered from the study of in vitro replication of X, suggest that a possible general function for DnaK in the host may be to dissociate protein aggregates that may arise either normally or during periods of stress (53). There is an extraordinary degree of homology between DnaK and the hsp70 family of heat shock proteins in eukaryotic cells (53). One of the eukaryotic heat shock proteins, hsc70, binds to clathrin cages and disassembles them into clathrin trimers in an ATP-dependent reaction (62). This process may be similar to the reaction observed above for DnaK and DnaJ in disassembling the 0 . P. DnaB ' oriX complex. Pelham (53) has proposed that the hsp7O heat shock proteins associate with the hydrophobic portions of a protein aggregate and, after hydrolyzing ATP, undergo a conformational change which may in turn alter the conformation of the protein substrate. The resulting conformational change in the substrate then promotes disassembly of aggregates and/ or renaturation of denatured protein. A possible unfolding function has been supported by the observation that hsp7O heat shock proteins facilitate protein translocation across 15.