The Oct-1 POU homeodomain stabilizes the adenovirus preinitiation complex via a direct interaction with the priming protein and is displaced when the replication fork passes.

Initiation of adenovirus DNA replication is strongly enhanced by two cellular transcription factors, NFI and Oct-1, which bind to the auxiliary origin and tether the viral precursor terminal protein-DNA polymerase (pTP·pol) complex to the core origin. NFI acts through a direct contact with the DNA polymerase, but the mode of action of Oct 1 is unknown. Employing glutathione S-transferase-POU pull-down assays and protein affinity chromatography, we have established that the POU domain contacts pTP rather than pol. The POU homeodomain is responsible for this interaction. The protein-protein contacts lead to increased binding of pTP-pol to the core origin, which is caused by a reduced off-rate. The enhanced formation of a pTP·pol·POU complex on the origin correlates with stimulation of replication. Using an immobilized replication system, we have studied the kinetics of dissociation of the Oct-1 POU domain during replication. In contrast to NFI, which dissociates very early in initiation, Oct-1 dissociates only when the binding site is rendered single-stranded upon translocation of the replication fork. Our data indicate that NFI and Oct-1 enhance initiation synergistically by touching different targets in the preinitiation complex and dissociate independently after initiation.

The adenovirus genome is a linear double-stranded DNA molecule of 36 kbp, 1 which replicates very efficiently, producing high amounts of progeny DNA (10 5 -10 6 ) in infected cells. Origins of replication are located at both ends of the genome, within the inverted terminal repeats, and encompass approximately 50 bp. The main proteins required for efficient replication are encoded by the virus itself and are expressed early in infection. These are the DNA polymerase (pol); the precursor terminal protein (pTP), which is required for correct initiation; and the DNA-binding protein (DBP), which is essential for elongation. The first two are present in infected cells as a stable heterodimer (pTP-pol).
Initiation of DNA replication occurs by a protein-priming mechanism in which a serine residue in pTP covalently binds the first nucleotide, a deoxycytidine residue, via a phosphodi-ester bond (reviewed in Ref. 1). A pTP-trinucleotide intermediate is formed, guided by base pairing with an internal triplet and then jumping back to the very end of the DNA (2). Subsequently, this intermediate is used as a primer for further chain elongation via a strand displacement mechanism requiring the viral DNA-binding protein. The pTP remains bound to the 5Ј terminus and probably serves several functions, such as protection from exonucleases, attachment to the nuclear matrix, and stabilizing the incoming pTP⅐pol complex (3,4). pTP⅐pol binding to the origin requires base pairs 8 -18 for correct positioning of the complex (5).
In addition to the three viral proteins, at least two host proteins are recruited by the virus for maximal origin function. Both proteins, nuclear factor I (NFI) and the octamer-binding protein (Oct-1), are cellular transcription factors that bind independently to a region adjacent to the 18-bp core origin. Together, these two factors stimulate replication up to 200-fold. Their mechanism of stimulation has been studied both in vitro and in vivo (6 -9).
NFI binds as a dimer to a partially symmetric sequence, 5Ј-TGGA(N) 5 GCCAA-3Ј, located between residues 25 and 38 of the Ad5 origin. The position of this NFI site relative to the core origin is critical for efficient replication. In vitro, NFI stimulates initiation in a pTP⅐pol concentration-dependent fashion (7), suggesting that the proteins interact. A direct, DNA-independent contact can indeed be detected with the polymerase in the pTP⅐pol complex (7,10,11). This interaction is functional because mutations in NFI that fail to bind pTP⅐pol are also defective in stimulation of replication (12). Template commitment studies showed a 10-fold increased stability of the pTP⅐pol complex on the origin in the presence of NFI, explaining most of the stimulatory action of NFI (13).
Oct-1 binds immediately next to the NFI site to the sequence 5Ј-TATGATAATGA-3Ј, which is located between residues 39 and 49. Deletion of the Oct-1 binding site in the adenovirus type 5 origin results in poorly growing viruses, with a yield 50-fold lower than that of the wild type (9). As for NFI, the stimulatory activity of Oct-1 resides within the DNA binding domain. This DNA binding domain, the POU domain, consist of two combined helix-turn-helix DNA binding elements, the POU-specific domain (POUs), and the POU homeodomain (POUhd). POUs binds the sequence 39 TATGA 43 , and POUhd recognizes 44 TAATGA 49 (14,15). In vitro Oct-1 stimulates replication 3-7-fold, also depending on the pTP⅐pol concentration and a DNA independent interaction between the pTP⅐pol complex and the POU homeodomain was observed (16). The target in the pTP⅐pol complex and the mechanism of stimulation were not established; these discoveries are the purpose of the present study.
Together with the viral proteins, NFI and Oct-1 assemble a preinitiation complex leading to optimal initiation. After initiation, this complex presumably disintegrates, but the details and kinetics of dissociation are still largely unknown. Employing an immobilized replication system, we previously showed that NFI disembarks from the preinitiation complex already at an early stage (17). Using the same technology, we have now obtained evidence that Oct-1 dissociates later, upon passage of the replication fork through the Oct-1 binding site.
Oct-1 POU Domain Dissociation Assay-Recombinant GST-POU and GST-NFI were purified from bacterial lysates by successive applications of DEAE anion exchange, glutathione agarose affinity, and fast flow Q anion exchange chromatography essentially as described previously (16).
100 ng of GST-POU were incubated with 0.05 pmol of a 817-bp end-labeled Ad5 origin fragment (pHRI digested with EcoRI and ApaLI) in 20 l of buffer B (25 mM Hepes-KOH, pH 7.5, 5 mM MgCl 2 , 55 mM NaCl, 0.5 mM dithiothreitol, 5% glycerol, 5% dimethyl sulfoxide). Subsequently, 5 l of glutathione agarose (GA) beads (50% v/v) in buffer B were added and the suspension was placed on a tumbling wheel at 4°C. After 45 min, beads were spun in a Microfuge (15 s) and washed twice in 150 l of buffer B. Under these conditions, more than 80% of the input DNA was bound. The beads were resuspended in 20 l of buffer B containing 0.9 g of DBP and 15 ng of pTP⅐pol. Replication was started by adding nucleotides at 40 M and placing the reaction tube at 37°C on a tumbling wheel. After 30 min, the beads were spun down. To the supernatant, 4 l of stop mix was added, and the released products were analyzed on 7.5% polyacrylamide gel.
DNA Binding Studies-The Ad5 origin probe used for band shift assays consisted of partially duplex oligonucleotides that lacked the first 14 nucleotides from the 5Ј end of the nontemplate (TD15; Ref. 19). The DNA was end-labeled with T4 polynucleotide kinase and purified by preparative polyacrylamide gel electrophoresis. Binding reactions were carried out for 60 min on ice in 20 l of binding buffer (20 mM HEPES-KOH, pH 7.5, 1 mM EDTA, 1 mM dithiothreitol, 0.025% Nonidet P-40, 4% Ficoll). The concentration of input DNA was 1 nM. Free DNA and protein-DNA complexes were separated on a 7% polyacrylamide gel (37.5:1) run in 0.5 ϫ Tris borate-EDTA at 4°C for 15 h at 100 volts. For the dissociation experiments, after 1 h, a 200-fold excess of unlabeled single-stranded DNA (derived from the template strand of the origin) was added, and samples were loaded onto a running polyacrylamide gel at the indicated time points.
GST-Oct-1 POU Domain Coprecipitation-To detect whether the Oct-1 POU domain and NFI associate with the different replication components, 0.25 g of GST, GST-POU, or GST-NFI fusion proteins were coupled to casein-coated GA beads (10% v/v) by incubation for 2 h at 4°C in 60 l of buffer C (20 mM Hepes-KOH, pH 7.5, 10% glycerol, 100 mM NaCl, 0.5 mM ␤-mercaptoethanol, 0.1% Nonidet P-40, 0.5 mM phenylmethylsulfonyl fluoride, 1 mM Na 2 S 2 O 5 ) and 50 g/ml ethidium bromide on a tumbling wheel. The beads were spun down and washed three times with 400 l of buffer C. Resin-bound proteins were separated on SDS-polyacrylamide gels and detected by immunoblotting (16).
GST Chromatography-2 mg of cytoplasmic extracts from Sf9 cells infected with a recombinant baculovirus expressing pTP (11) were applied to a 0.25-ml column of GA beads coupled with 100 g of GST-POU or GST equilibrated with buffer containing 200 g/ml ethidium bromide. After washing with 5 ml of the same buffer, a 10-ml linear gradient from 75 to 1000 mM NaCl in this buffer was applied. 0.5 ml fractions were collected and 5 l of each were run on an SDS-polyacrylamide gel. pTP was detected by immunoblotting with polyclonal anti-pTP⅐pol serum (16).

RESULTS
The Oct-1 POU Domain Interacts with the pTP-Previous experiments have shown an interaction between the Oct-1 POU domain and the pTP⅐pol complex (16). To study which of the two proteins in the complex is involved in this interaction, we separately expressed pTP and pol using recombinant baculoviruses. Protein-protein interactions were assessed by affinity chromatography with the Oct-1 GST-POU domain fusion protein bound to glutathione agarose beads. A cytoplasmatic extract of insect cells expressing pTP was passed over the resin, and bound proteins were eluted by applying a linear gradient from 75 to 1000 mM NaCl. The presence of pTP in eluted fractions was determined by Western blotting using polyclonal anti-pTP⅐pol antiserum (Fig. 1). Although some of the pTP ran through the column, a significant amount was specifically retained on the GST-POU column and eluted at approximately 225 mM NaCl. A control column in which only GST was bound to the matrix did not retain any pTP, thus excluding nonspecific binding (Fig. 1). Moreover, the column was equilibrated in a buffer containing ethidium bromide to prevent a DNA-mediated interaction. In analogous experiments using crude extracts, we could not detect an interaction between the Oct-1 POU domain and the DNA polymerase.
NFI and the POU Domain Bind Different Subunits in the pTP⅐pol Complex-Because we could not exclude the presence of inhibitors in the crude extracts, we purified pTP and pol separately, as well as the pTP⅐pol complex, and studied binding in a pull-down assay employing GST-POU or GST-NFI. Fig. 2A shows that pTP specifically binds to GST-POU beads (lane 3) but not to GST-NFI beads (lane 4), whereas pol is only retained by GST-NFI (lane 8). A slightly smaller product, presumably a degradation product of pTP, is also bound. The pTP⅐pol complex bound both POU and NFI, as expected (lanes 11 and 12). The levels of pTP⅐pol retained by NFI were slightly higher than for the POU domain, possibly indicating stronger binding, in A consequence of POU and NFI contacting different subunits in the pTP⅐pol complex would be that all four proteins together form a single large complex in the absence of DNA. We tested this by assaying whether NFI could be retained by GST-POU in the presence of pTP⅐pol (Fig. 2B). Indeed, more than 50% of the input NFI was bound under these conditions (lane 2), whereas only 10% was bound when pTP⅐pol was left out. Binding of NFI was reduced in the presence of purified His6-POU domain, which competes with GST-POU, and therefore less is precipitated by GA beads. These data show that a multiprotein complex can be formed between the viral and cellular proteins.
The POU Domain Enhances Binding of pTP⅐pol to the Origin-To investigate the functional consequences of the interaction, we studied the effect of the POU domain on binding of the pTP⅐pol complex to the core origin. Because pTP⅐pol binding to a double-stranded origin is weak and difficult to study using band shifts, we employed a partially duplex origin, TD15, which lacks the first 14 nucleotides from the 5Ј end (19). Using increasing amounts of purified pTP⅐pol, we could detect binding to TD15 starting from 7.5 ng (Fig. 3, lane 10). Addition of the POU domain at this level of pTP⅐pol led to the formation of a POU-pTP-pol-DNA complex that was much stronger (lane 11). The heterotrimer-DNA complex can already be detected with 1.9 ng of pTP⅐pol (lane 17). Quantification of the bound complexes showed a 6-fold increased binding of pTP⅐pol in the presence of the POU domain at low pTP⅐pol concentration, whereas almost no enhanced binding was seen at high amounts of pTP⅐pol (lanes 1 and 2). The concentration-dependent com-plex formation correlates well with the pTP⅐pol concentrationdependent stimulation of replication by the POU domain (16). Enhanced pTP⅐pol binding was also observed when only the POUhd was added (Fig. 3). The pTP-pol-POUhd complex (210 kDa) on DNA had almost the same electrophoretic mobility as the pTP⅐pol complex (200 kDa), which makes it difficult to establish the presence of the POUhd in the complex based on its position in the gel. At equimolar amounts, the increase in binding was lower than for the intact POU domain, most likely reflecting a lower stability.
Enhanced binding of pTP⅐pol to the origin could be due to a increased association or a decreased dissociation of the complex or both. We assayed the on-rate of the pTP⅐pol complex to TD15 DNA but could not detect a difference between the presence and the absence of the POU domain (data not shown). A stabilizing effect of the POU domain was observed when dissociation was measured (Fig. 4). A pTP-pol-origin complex was allowed to form with or without the presence of the POU domain. After equilibrium was reached, a 200-fold excess of unlabeled template strand DNA was added. This singlestranded competitor can be bound efficiently by the pTP⅐pol complex, whereas it is not recognized by the POU domain, thus allowing an accurate measurement of the stabilizing effect. Dissociation was measured as a function of time by analyzing samples on a nondenaturing gel (Fig. 4). Without the POU domain, dissociation of the pTP⅐pol complex was observed within 1 min (lane 10), whereas in the presence of the POU domain, a pTP⅐pol complex is still present after 10 min (lane 5).

FIG. 2. Oct-1 and NFI interact with different components of the pTP⅐pol complex.
A, GST, GST-POU, or GST-NFI bound to GA beads was incubated with purified pTP (lanes 2-4), pol (lanes 6 -8), or the pTP⅐pol complex (lanes 10 -12). 50% of the input used in the precipitation reaction was loaded directly on the gel (lanes 1, 5, and 9). Bound pTP and pol were detected by Western blotting using polyclonal rabbit anti-pTP⅐pol antiserum. The positions of pTP and pol are indicated. B, GST-POU bound to GA beads was incubated with pTP⅐pol and NFI. Bound NFI was detected by Western blotting using polyclonal rabbit anti-NFI antiserum. 50% of the input NFI used in the coprecipitation reaction was loaded directly on gel (lane 1). 0.5 g of purified His6-POU domain was added as a competitor in lane 5. We conclude that the enhanced equilibrium binding of the pTP⅐pol complex in the presence of the POU domain is caused by increased stability of the complex rather than by an increased rate of assembly.
The POU Domain Dissociates when the Replication Fork Passes-In order to establish the moment at which Oct-1 dissociates from the template, we used a modified immobilized replication system developed previously in our laboratory (17). In this system, a functional initiation complex was assembled on GA beads employing GST-NFI. As for NFI, the system enables the study of dissociation of the POU domain during the early stages of replication. A labeled Ad5 origin fragment bound by GST-POU was immobilized on GA beads (Fig. 5A). Dissociation of the POU domain from the origin can be monitored by the release of labeled DNA from the beads because the GST-GA interaction remains. When formation of a pTP-dCMP complex is allowed to take place by adding only dCTP, the template is not released (Fig. 5B). This is in contrast to NFI, which already dissociates when the polymerase binds the first nucleotide (17). Addition of dATP and dTTP allows progression of the polymerase up to 26 nucleotides but does not result in release of the origin from the POU domain. Only when all four nucleotides are added, allowing the polymerase to proceed beyond position 26, is the origin released. We conclude that the POU domain dissociates only when the polymerase passes the POU domain recognition sequence, presumably because this site is rendered single-stranded upon translocation of the replication fork. We did not detect a release of all bound POU domains from the origin, which was probably due to the fact that not all templates are efficiently replicated (17). Reassociation of the released POU domains might also lead to lower release levels. In parallel reactions, we tested the replication products formed under the same conditions using ␣ 32 P-dCTP and unlabeled template. As can be seen in Fig. 5C, replication proceeded in a way similar to a soluble system.
POUhd Stimulates Replication Only on a Small Origin-containing Fragment-Because POUhd stabilizes the binding of pTP⅐pol on the origin, we anticipated that, like the POU domain, POUhd would also stimulate initiation. Previous results indicated, however, that POUhd inhibited replication (20). The system used to assay this effect employed Ad5 TP-DNA predigested with XhoI. We repeated the experiment with a more extensive concentration range, with the same result. In contrast to the POU domain, which stimulated 3-4-fold under the conditions chosen (a relatively high pTP⅐pol concentration), addition of POUhd resulted in a 2-fold reduction (Fig. 6, A and B).
The inability of POUhd to stimulate replication could be caused by its higher dissociation rate, resulting in a less stable complex, or by incorrect targeting of the pTP⅐pol complex. Deletion of the POUs subdomain results in loss of specificity (21), and the large number of other POUhd binding sites on the FIG. 4. The POU domain stabilizes pTP⅐pol binding to the origin. 5 ng of POU domain and 30 ng of pTP⅐pol were bound to a labeled TD15 probe. After 1 h at 0°C, a 200-fold excess of unlabeled template strand DNA was added and samples were loaded onto a running polyacrylamide gel at the indicated time points (lanes 1-8). On the right side (lanes 9 -16), the POU protein was left out. 36-kbp viral genome might lead to incorrect targeting of the pTP⅐pol complex. To test this, we used a small (110 bp) origincontaining fragment as template rather than the whole viral genome. Under the same replication conditions, we observed that POUhd is indeed able to stimulate replication of this small fragment (Fig. 6C). The level of stimulation is not as high as that obtained with the intact POU domain (Fig. 6D), possibly due to the higher off-rate of the POUhd (22). This experiment shows that the observed interaction of POUhd with pTP is functional, although a covalent linkage to the POUs domain is required for accurate targeting and stability on the intact viral DNA. A drop in replication is observed with increasing amounts of POU protein on the 110-bp origin fragment (Fig.  6D). We assume that the pTP⅐pol complex is squelched by the POU protein at high concentrations. The terminal protein, which stabilizes the replication complex, is absent on this plasmid-derived fragment, making it more sensitive to squelching by the POU domain. DISCUSSION We show that the Oct-1 POU domain stabilizes the preinitiation complex via a direct interaction between pTP and POUhd. Enhanced pTP⅐pol binding in the presence of Oct-1 correlates well with the levels of stimulation, with the restriction that DNA binding assays and replication assays are performed under different conditions. Although binding of pTP is essential, it is not necessarily the only function of the POU domain. The POU domain is also capable of bending its recognition sequence slightly (23,24), and such a bend in the origin might facilitate the pTP-pol-POU interaction on DNA.
Detailed mapping of the interaction domains on POUhd or pTP has not yet been achieved. POU proteins from all six classes are able to stimulate replication (25), 2 suggesting that the interaction domain is a conserved region. Mutation of sev-2 A. van der Flier, personal communication.

FIG. 5. The Oct-1 POU domain dissociates when the replication fork passes.
A, experimental set-up: GST-POU bound to a radiolabeled origin fragment was immobilized on glutathione agarose beads. Replication was initiated by the addition of pTP⅐pol, DBP, and nucleotides. Arrows indicate the expected product length when a limited number of nucleotides were added. B, dissociation of the POU domain from the origin fragment was monitored by the release of the radiolabeled fragment in the supernatant after elongation with the consecutive nucleotides. After replication of the immobilized complexes, the beads were spun down and the products were analyzed on a polyacrylamide gel. The bound origin fragment is shown in the upper panel, released fragments in the lower panel. C, replication products formed using a unlabeled origin fragment and [ 32 P]dCTP (indicated with C*) were analyzed on a polyacrylamide gel. The products formed were as indicated. eral surface exposed residues, however, did not reduce the stimulatory action of the POU domain (data not shown). Previously (16) we observed that mutation of Oct-1 POUhd residues Q24 and E29 gave rise to enhanced stimulation, but the mechanism of this is not clear, and this effect might also have been due to a slight contamination of the mutant proteins with NFI.
A Model for Assembly of the Adenovirus Initiation Complex-We envisage the following model for initiation and early elongation (Fig. 7). The four proteins (pTP⅐pol, NFI and Oct-1) can form a complex even in the absence of DNA (Fig. 2B). Whether such a DNA free complex is formed in vivo and is sufficiently stable is presently unknown. The multiprotein complex (altogether, approximately 480 kDa) binds the origin, thereby positioning and stabilizing pTP⅐pol correctly in a preinitiation complex. Stabilization both by Oct-1 (Fig. 4) and NFI (13) is achieved by lowering the off-rate of the pTP-pol-DNA interaction. Because Oct-1 and NFI interact with different subunits of the pTP⅐pol complex they supplement each other, thereby explaining the synergism in assembly and acti-vation observed before (7). Although the four proteins form a complex, they do not all interact with each other. No interaction between NFI and pTP or between Oct-1 and the polymerase was detected ( Fig. 2A). We also do not know whether both NFI subunits interact with the polymerase. It is likely that only one subunit of NFI is involved, but this is difficult to establish because monomeric NFI cannot be isolated (11,26). Although a weak interaction between NFI and Oct-1 was observed (Fig. 2B), we do not think that this is functional because maximal stimulation levels of Oct-1 and NFI do not seem to be influenced by each other (7). Only slight cooperative DNA binding was observed between NFI and Oct-1 on the Ad2 origin (7). Such cooperativity was shown to occur in another sequence context on the human papillomavirus enhancer (27).
Not indicated in Fig. 7 is the role of DBP in initiation. We do not have evidence that DBP forms a complex with the other four proteins in the absence of DNA nor can it be found stably in the preinitiation complex. Nevertheless DBP has a pronounced effect on initiation because it decreases the K m of the polymerase for the initiator dCTP (28) and enhances the bind- ing of NFI (29,30). These effects are most likely due to changes in the origin DNA, which are known to occur by DBP although we can not exclude a direct interaction with the polymerase.
The complete preinitiation complex spans five helical turns. To allow all protein-protein contacts, considerable bending of the DNA has to occur within the first 50 bp (Fig. 7). Several lines of evidence indicate that this is indeed the case. The protein-free origin is already slightly bent, as concluded from circular permutation experiments (31). This intrinsic bend is enhanced by binding of NFI (31). Because Oct-1 also bends the DNA, we assume that a considerable distortion of the DNA structure must exist in a preinitiation complex. We are currently investigating whether simultaneous binding of NFI and Oct-1 causes the DNA to bend toward the same side of the helix. Further bending might also be generated by the covalently bound TP, which induces structural changes in the origin as concluded from the appearance of DNase I hypersensitive sites (32). Finally, the enhancing effect of DBP on initiation could well be related to its property of inducing structural changes in DNA (33). Together, these DNA conformational changes might promote protein-protein interactions and destabilize the origin to facilitate initial opening.
Dissociation of the Adenovirus Initiation Complex-After the initiation, the complex starts dissociating. NFI already dissociates as soon as initiation starts (Fig. 7). Because the NFI site is located between bp 25 and 38 of the origin and is likely to remain double-stranded at this stage, drastic changes in the initiation complex must occur that cause NFI to dissociate. The POU domain, in contrast, remains bound to the origin until the passing polymerase unwinds the octamer recognition site (Fig.  5). This suggests that the POU domain may also stabilize the complex after the first dCTP coupling, during formation of early elongation products. The pTP⅐pol heterodimer dissociates early in elongation. 3 During elongation, DBP binds cooperatively to the displaced strand (34), thereby assisting the polymerase in DNA unwinding. 4 POU Domain Proteins as Regulators of Viral Replication-The use of a POU domain protein for efficient multiplication is not restricted to adenoviruses. Direct protein-protein interaction was reported between Oct-6 and the JC papovavirus T antigen (35) and between Oct-1 and the Herpes simplex transactivator protein (36), both resulting in enhanced viral gene activation. Other viruses that use a POU transcription factor include SV40, Epstein-Barr virus, murine mammary tumor virus, human papillomovirus, and hepatitis B virus (37-41). Presumably, their high level of conservation, in particular of the POU domain, and wide expression have made them attractive targets for invading viruses during evolution. Adenovirus, however, is the only virus known to use Oct-1 for efficient DNA replication. There is some indirect evidence that Oct-1 is involved in eukaryotic replication. Octamer sequences have been found in several chromosomal origins of replication (42,43), and enhanced replication of a transfected plasmid depended on an intact octamer site (42). Whether this is a direct effect on replication remains to be elucidated.