Characterization of DNA-protein complex formation in nuclear extracts with a sequence from the herpes simplex virus thymidine kinase gene.

The biochemical characteristics of complex formation in nuclear extracts from mock-infected and herpes simplex virus (HSV)-infected Vero and HeLa cells with a sequence downstream of and adjacent to the promoter for the HSV thymidine kinase gene were studied using the mobility shift electrophoresis assay. This region is bound by host cell proteins, as evidenced by the formation of complexes after incubation in extracts from mock-infected cells. Unique virus-specific complexes form in extracts prepared from infected cells, and these complexes contain ICP4, the major regulatory protein of HSV. Examination of the salt requirements for assembly and the stability of preformed DNA-protein complexes to added salt demonstrate the distinct nature of the complexes that form in each extract. This finding is supported by analyses of the relative association and dissociation rates of these complexes which show that complexes formed in extracts prepared from infected cells are kinetically labile. After depletion with chelators, the divalent cation requirements for complex formation were assayed by supplementation with various metal salts. Addition of Mn2+ restored binding activity in extracts from both mock-infected and infected HeLa cells. Finally, footprinting assays revealed that sequences on each strand throughout this region of the thymidine kinase gene were involved in complex formation only in extracts from mock-infected cells. These experiments suggest that one consequence of virus gene expression is to alter the interaction of cell proteins with virus DNA.

The biochemical characteristics of complex formation in nuclear extracts from mock-infected and herpes simplex virus (HSV)-infected Vero and HeLa cells with a sequence downstream of and adjacent to the promoter for the HSV thymidine kinase gene were studied using t,he mobility shift electrophoresis assay. This region is bound by host cell proteins, as evidenced by the formation of complexes after incubation in extracts from mock-infected cells. Unique virus-specific complexes form in extracts prepared from infected cells, and these complexes contain ICP4, the major regulatory protein of HSV. Examination of the salt requirements for assembly and the stability of preformed DNA-protein complexes to added salt demonstrate the distinct nature of the complexes that form in each extract. This finding is supported by analyses of the relative association and dissociation rates of these complexes which show that complexes formed in extracts prepared from infected cells are kinetically labile. After depletion with chelators, the divalent cation requirements for complex formation were assayed by supplementation with various metal salts. Addition of Mn'+ restored binding activity in extracts from both mock-infected and infected HeLa cells. Finally, footprinting assays revealed that sequences on each strand throughout this region of the thymidine kinase gene were involved in complex formation only in extracts from mock-infected cells. These experiments suggest that one consequence of virus gene expression is to alter the interaction of cell proteins with virus DNA.
Regulation of eukaryotic gene expression is a consequence of complex interactions between &acting regulatory elements, sequence-specific DNA-binding proteins, and proteinprotein interactions.
Cells infected with herpes simplex virus (HSV),' a double-stranded DNA-containing virus, provide a model system for examining these interactions.
Gene regulation in cells infected with HSV can be divided into three temporally regulated phases (Honess and Roizman, 1974;Honess and Roizman, 1975), termed a, p, and 7. The promoters for the corresponding genes are recognized by both cell and virus proteins. a genes are activated in response to OTF-1 (O'Hare et al., 198& Gerster and Roeder, 1988), a * These studies were supported by Grant CA17477 from the National Institutes of Health (to S. J. S.) The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "uduertisement" in accordance with 18 U.S.C. Section 1'734 solely to indicate this fact. ' The abbreviations used are: HSV, herpes simplex virus; bp, base pair(s); OP, l,lO-phenanthroline. cellular protein which recognizes the sequence ATGCAAAT, and a-TIF, a virion-associated protein that does not bind directly to DNA but recognizes at least one and probably more host proteins McKnight et ul., 1987;Preston et al., 1988;Triezenberg et ul., 1988). The promoters for the b genes are activated in response to synthesis of an a gene product, ICP4, the major regulatory protein of HSV. Genetic and biochemical experiments have demonstrated that functional ICP4 is required throughout the virus replication cycle to both activate and repress expression of the three temporally regulated gene families (DeLuca et al., 1984;Dixon and Schaffer, 1980;Gelman and Silverstein, 1986;Preston, 1979aPreston, , 1979bWatson and Clements, 1980). True 7 genes are transcribed only after the initiation of DNA synthesis (Clements et ul., 1977;Harris-Hamilton and Bachenheimer, 1985;Honess and Roizman, 1974;Stringer et al., 1977;Swanstrom et al., 1974) and require both ICP4 and another a gene product., ICPZ7, for synthesis of late proteins in stoichiomet.ric amounts (McCarthy et ul., 1989;Rice and Knipe, 198% Sacks et ul., 1985;Sekulovich et ul., 1988).
To study control of these three classes of temporally regulated genes, the response of defined promoter-regulatory elements to known or suspected virus regulatory proteins was examined in both stable transformants and short-term transient expression assays (Cordingly et ul., 1983;DeLuca and Schaffer, 1985;Eisenberg et ul., 1985;ElKareh et ul., 1985;Everett, 1983;Gelman and Silverstein, 1985;Homa et ul., 1986;Kristie and Roizman, 1984;O'Hare and Hayward, 1985;Post et ul., 1981;Sandri-Goldin et ul., 1983). These studies have revealed that when plasmids containing either a or /3 promoters, linked to reporter molecules, are introduced into eukaryotic cells, they respond to virus trans-acting factors in the same manner as their cognates in the HSV genome. Furthermore, it was shown that recognition of some of these promoters exhibited cell type specificity (Everett, 1988;Gelman and Silverstein, 1987). The late, or 7, promoters are not faithfully regulated in stable or transient expression systems because they are activated in the absence of virus DNA replication (Dennis and Smiley, 1984;Mavromara-Nazos et ul., 1986).
Virus and cellular factors that affect HSV gene expression are in the process of being identified.
Their isolation and characterization will make it possible to determine the biochemical basis for temporal regulation.
The promoter region for the HSV-1 thymidine kinase gene, a prototype of the @ kinetic class, is perhaps the best characterized among ICP4inducible genes. The role of specific sequences that compose this promoter and how various virus gene products affect its expression have been studied in great detail (Coen et ul., 1986;Eisenberg et ul., 1985;ElKareh et ul., 1985;Gelman and Silverstein, 1986;McKnight et ul., 1981;McKnight and Kingsbury, 1982;Zipser et al., 1981). While earlier reports suggested that induction-specific sequences exist (ElKareh et ul., 1985;Zipser et ul., 1981), more recent studies (Coen et ul., 1986;Eisenberg et ul., 1985) have demonstrated that the same promoter domains required for basal expression are required during viral infection. As a result of these studies, it has been postulated that thymidine kinase induction is mediated through the action of ICP4 (and perhaps other viral a gene products) on cellular transcription factors, leading to an increased transcription rate of this gene. Identifictition of the target sequences involved in regulation of the thymidine kinase gene has proven elusive. In one instance (Zipser et ul., 1981), sequences within the mRNA leader were shown to be important for induction. Similarly, leader sequences for the HSV a-TIF and VP5 genes are required for regulation and stabilization of the mRNAs encoding these proteins (Blair and Wagner, 1986;Blair et ul., 1987). In the present study, the interaction of crude nuclear extracts from two different cell types (mock-infected and HSVinfected Vero and HeLa cells) with a sequence that overlaps the cap site and leader (-16 to +56, TKB) for the HSV-1 thymidine kinase gene is examined in detail using the mobility shift electrophoresis assay (Fried and Crothers, 1981;Garner and Revzin, 1981). Our results demonstrate that ICP4 participates in complex formation in extracts from infected cells and reveal a relationship between the appearance of virusspecific complexes and the disappearance of cell-specific complexes and an absolute requirement for Mn*+ in complex formation in extracts from both mock-infected and infected cells. They also show that complexes formed in nuclear extracts prepared from infected cells are kinetically labile compared with those formed in extracts from mock-infected cells. Moreover, DNase I footprinting and missing contact analyses supported the conclusion from the kinetic studies that complexes formed in extracts from infected cells rapidly dissociate and slowly reform.

MATERIALS AND METHODS
CelLs, Viruses, and Preparahn oj Extracts-HeLa and Vera (African Green Monkey kidney) (Yasumura and Kawakita, 1963) cells were propagated as previously described (Gelman and Silverstein, 1985). To prepare extracts, 5 x 106 actively growing cells were seeded in 150.mm culture dishes in 20 ml of Dulbecco's modified Eagle's medium containing 8% calf and 2% fetal calf serum (HyClone).
The medium was replaced every 24 h and when the monolayer was confluent, the cells were either mock-infected or infected at a multiplicity of infection of 5 plaque-forming units with HSV-1 strain F (Ejercito et al., 1968). Virus was adsorbed for 30 min at 37 'C, after which fresh medium was added. Infections were allowed to proceed for 4.5 h (or for 2 and 10 h, respectively, in the time course-experiments presented in Fig. 2) at 37 'C before the cells were harvested for extiact preparatioi Nuclear extracts were prepared as described (Dignam et al., 1983) except that 1) all of the buffers employed after the phosphate-buffered saline washes contained 0.2 mM L-l-chloro-3-(4-tosylamido)-4-phenyl-2-butanone and 0.2 mM L-l-chloro-3-(4tosylamido)-7-amino-2-heptanone (Boehringer Mannheim) as additional proteinase inhibitors and 2) buffer D (used in the final dialysis step) contained 0.1 M NaCl rather than 0.1 M KC1 as originally specified.
The nuclear extracts were frozen as aliquots in liquid nitrogen and stored at -70 "C. The protein concentration varied between 6 and 8 mg/ml and was determined using a Bio-Rad protein assay kit.
Cloning and Preparation of Probe and Competitor DNAs-The HSV-1 TKB DNA fragment shown in Fig. 1

TKB-Protein
Complex Formation in Nuclear Extracts from Mock-infected and Infected Cells-The origin and the boundaries of the region downstream from and adjacent to the promoter for the thymidine kinase gene (TKB) used in all experiments are shown in Fig. 1. This fragment was endlabeled, isolated, and incubated with crude nuclear extracts prepared from mock-infected or HSV-l-infected (for different periods of time) HeLa and Vero cells, as described under "Materials and Methods." The reaction mixtures were then analyzed for DNA-protein complex formation using the mobility shift electrophoresis assay (Fried and Crothers, 1981;Garner and Revzin, 1981). In the absence of extract, the probe migrated as a single band (data not shown). When extracts from mock-infected cells were used, two complexes (Vero, Vl and V2; HeLa, Hl and H2, numbered in order of decreasing electrophoretic mobility) were detected in each extract (Fig.  2, A and B, lane a). These may have resulted from binding of a cellular factor to several sites within the probe, oligomerization of a single factor on the same template, or interactions between DNA-binding and non-DNA-binding proteins. Although the number of the retarded species appeared to be the same for both extracts, a shorter exposure of the gel revealed that the faster migrating complex formed in extracts from Vero cells (Vl) was actually a doublet ( Fig. 2A, lane a). Moreover, the electrophoretic mobility and abundance of the complexes formed in the extracts from the two cell types differed, demonstrating the unique nature of the participants The effect of virus gene expression on TKB complex formation was examined by preparing extracts from infected cells and harvesting them after 2.5,5, and 10.5 h. The amount of cell-specific complexes decreased as a function of the time postinfection at which the extracts were prepared (Fig. 2

Herpesvirus Infection
Alters DNA-Protein Interactions and B, lanes d, g, and;). In extracts from Vera cells, formation of the slower migrating cell-specific species (V2) and the lower band of the cell-specific doublet (Vl) was abolished at 5 h postinfection. The remaining complex (V3) persisted but was reduced in amount at 5 h postinfection and increased its abundance at 10.5 h postinfection. The Hl and H2 complexes could not be detected in extracts prepared from HeLa cells infected for 10.5 h (Fig. 2B, lune j). Novel complexes were detected after incubation in all of the extracts prepared from infected cells (Vera, V4; HeLa, H4, H5, H6, numbered in order of decreasing electrophoretic mobility). The 5-h extracts were the most active (Fig. 2, A and B, lczne g). To determine if ICP4 was involved in the formation of these novel complexes, reaction mixtures were prepared with each of the infected cell extracts and incubated with monoclonal antibody specific for ICP4. If the antibody recognizes the protein in a complex, it will (depending on the location of the specific epitope) either block complex formation or be tethered to it and further reduce the mobility of the complex in the gel Roizman, 1986a, 1986b;Muller, 1987). The de nouo appearance of a slowly migrating species (urrowheuds) and the corresponding loss of a band (&z& circles) in the 5-h samples demonstrated that ICP4 was present in at least one of the complexes (Fig. 2, A and B, lune h). The low abundance of virus-specific complexes in the 2.5and 10.5-h HeLa cell extracts made detection of the tethered species difficult. However, the virus-specific bands disappeared (black circles) after addition of antibody, suggesting that ICP4 was a component of these complexes as well (Fig.  2B, lunes e and k). ICP4 was also demonstrated in the complexes formed in the 2.5-and 10.5-h Vero cell extracts, although the amount of virus-specific complex V4 in these extracts was very low ( Fig. 2A, arrowheads in lunes e and k). A control antibody specific for glycoprotein C (a 7 gene product) had no effect on the mobility of the complexes formed after incubation in extracts from infected HeLa or Vero cells (Fig. 2, A and B, lunes f, i, and 1). Incubation of these antibodies with complexes formed in extracts from mock-infected cells demonstrated the specificity of the supershift produced when the anti-ICP4 antibody was included in the binding reactions with extracts from infected cells (Fig. 2,A and B,lunes b and c). The very fast migrating species present in all lanes were deemed to be nonspecific because their appearance varied with each extract preparation and they were readily competed with heterologous DNAs.
These data demonstrate a relationship between the abundance of the complexes which form and the time postinfection at which the extract was prepared. Furthermore, recognition of this sequence by extracts from infected cells appears to first involve binding of cell proteins, with the subsequent participation of at least ICP4 affecting the formation of stable DNA-protein complexes.

Effect of Ionic Strength on TKB-Protein
Complex Formution-The polyelectrolyte nature of nucleic acids makes the interactions of proteins with DNA quite sensitive to salt concentration (Lohman, 1986). Several groups have reported that the ionic strength of the binding buffer can alter the yield and pattern of ICP4-containing DNA-protein complexes when crude extracts from HSV-infected cells are used as a source of ICP4 (DeLuca and Schaffer, 1985;Kristie and Roizman, 1986;Muller, 1987). In these experiments, optimum complex formation occurred in the presence of 15-25 mM NaCl. These reports and the apparent differences in the patterns of the TKB-protein complexes observed between extracts derived from HeLa and Vero cells prompted us to examine the effect of NaCl concentration on formation of complexes in extracts from either mock-infected or infected HeLa and Vero cells. When binding reactions were conducted with extracts from Vero cells, the yields of the slower migrating species (V2 and its mobility homologue V5) and the virus-specific complex V4 increased until 50 mM NaCl and then decreased until they were eliminated at 200 mM NaCl (Fig. 3A, lunes u-d and g-j). The faster migrating species (Vl and its mobility homologue V3) were more resistant to increasing salt concentrations. Complex formation was maximal at 5 mM NaCl (contributed by the extract) and then gradually decreased at elevated salt concentrations (Fig. 3A, lunes u-f and g-l). However, because Vl is a doublet, it may be that the observed differences in salt effects on the amount of Vl result from loss of only one of the components in this doublet. In any case, it was the only complex detected in both extracts when NaCl concentrations exceeded 100 mM. Moreover, although its abundance differed in the two extracts, its sensitivity was very similar throughout the entire salt range tested.
Optimal yield of TKB-protein complexes in extracts from HeLa cells occurred at 50 mM NaCl in those from mockinfected cells and at 5 mM NaCl (contributed by the extract) in extracts from infected cells (Fig. 3B, lunes b and g). In contrast to the "bell-shaped" salt dependence behavior seen in extracts from Vero cells (at least for V2 and V4), there was a sharp inhibition of complex formation at NaCl concentrations other than the optima (Fig. 3B). The labeled material appearing at the top of the gel (Fig. 3B, lunes d-f) probably results from high molecular weight aggregates formed at elevated ionic strength between the DNA probe and specific as well as nonspecific binding proteins which, under these conditions, adopt an overall positive surface potential. This precipitation and trapping near the origin was not detected when the probe was exposed at high salt in the absence of extract (data not shown).
These experiments demonstrate that increasing the Na+ concentration above 50 mM results in decreased abundance of the cell-specific complexes present in extracts from mock- A. abcdefg hi i kl ,cz.-+ -, ." Labeled TKB DNA was allowed to complex with proteins in extracts prepared from either mock-infected (&e~ a-/) or 5-h-infected (lanes a-0 Vero CA) and HeLa fI3) cells in the nresence of differing NaCl infected Vera cells and their mobility homologues in extracts from infected cells. The ICP4-containing complex (V4) showed a very similar salt sensitivity profile to that of the slower migrating species (V2) formed in extracts from mockinfected cells, suggesting that ICP4 interacts with complexes containing proteins with similar ionic requirements for binding in both extracts. In contrast, the complexes that form in extracts from mock-infected and infected HeLa cells exhibit distinct Na+ requirements. However, the lower abundance of these complexes precluded a direct comparison with those formed in extracts from Vero cells.

Effect of Ionic Strength on Stability of Preformed TKB
Complexes-The stability of the complexes to increasing ionic strength was measured by incubating the probe with each extract for 30 min in binding buffers containing the optimal NaCl concentration appropriate for complex formation. Then, a constant volume of varying concentrations of NaCl was added. After an additional 5 min of incubation, the reaction mixtures were analyzed for stable complexes by the mobility shift electrophoresis assay. Although an excess of unlabeled TKB fragment should be added to each reaction before the salt chase to bind any proteins that might dissociate from the template during exposure to increased NaCl concentrations, we have demonstrated that the salt stability of cell-and virusspecific complexes at equilibrium would be misinterpreted because of marked differences in the relative dissociation rates of the complexes formed in extracts from mock-infected and infected cells (see below). The rapidly migrating species formed in extracts from both mock-infected and infected Vero cells (Vl and V3) were sensitive to the salt chase ( Fig. 4.4, lanes f-a and g-l). However, comparison of their stability profile in these extracts with their salt sensitivity (Fig. 3A, lanes a-f and g-l) indicated that increased ionic strength had a less severe effect on the stability of the complex than on its formation. The slower migrating cell-specific complex (V2) and the ICP4-containing virus-specific species (V4) were par- tially resistant to the effect of increasing NaCl concentration. Although these complexes failed to form when the NaCl concentration exceeded 100 mM, they were stable to elevated levels of salt (Fig. 4A, lanes f-a and g-l). The refractile nature of these complexes to increased salt concentration might reflect a multi-step binding or assembly mechanism (Lehman, 1986) because each complex is thought to arise through multiple protein-protein interactions involving either host proteins (V2) or host proteins and ICP4 (V4).
When binding reactions with extracts from HeLa cells were challenged with increasing NaCl concentrations, the results were very similar to those seen in the salt sensitivity experiments (Fig. 4B, lanes a-f and g-l). However, the ICP4-containing complex (H4) which forms only at 5 mM NaCl was stable after addition of NaCl to 100 mM. The more rapidly migrating cell-specific complex (Hl) that continued to form (although the yield decreased) in the presence of up to 200 mM NaCl (Fig. 3B, lanes a-d) was unstable to the NaCl chase above 50 mM. These data suggested that the presence of ICP4 in complexes formed in extracts from infected cells alters the nature of the ionic interactions that stabilize these complexes and established the unique cell-specific features of complex formation and stability in extracts from mock-infected and infected HeLa cells.

Relative Association Kinetics of TKB-Protein
Complex Formation-To analyze the association of both cell and virus proteins with the target DNA, we first determined the time required to reach equilibrium. In these experiments, control binding reactions were allowed to equilibrate for 30 min. Preincubation of the extracts alone or incubation in the presence of the TKB probe at room temperature for more than 30 min led to decreased complex formation. However, because complexes were "stable" for at least 30 min, it was possible to compare relative association rates for all complexes. Accordingly, nuclear extracts were incubated with probe, and at various times reactions were quenched by addition of a 250-fold molar excess of cold TKB competitor fragment to limit further association of factor(s) with the probe. Quenched mixtures were then applied directly to a running polyacrylamide gel to separate the specific complexes. Although the quenching procedure prevents formation of additional complexes within a few seconds, it does not cause dissociation of the preexisting complexes during the time required for the complex to enter and be stabilized by the gel matrix (Fried and Crothers, 1984).
In extracts prepared from mock-infected Vero cells, formation of the slower migrating complex (V2) was apparently complete by 30 s, whereas formation of the faster migrating complex (Vl) was maximal only after 5 min (Fig. 5A, lanes af). In extracts prepared from infected cells, formation of the V3 complex continued for 30 min. In contrast, formation of the ICP4-containing V4 complex was maximal by 1 min and thereafter it dissociated slowly (Fig. 5A, lanes l-g). The two cell-specific complexes (Vl and V2) are clearly distinguishable by their different observed rates of formation. The more rapidly migrating complex (Vl) forms slower than the other species (Fig. 5A, lanes b and d). The V3 complex appeared to form slower than its Vl homologue did in extracts from mockinfected cells (clearly seen on a shorter exposure), suggesting that virus gene expression affected its kinetic behavior. In contrast, V4 formed rapidly but had a short half-life (Fig. 5A).
These experiments were repeated with extracts derived from mock-infected HeLa cells (Fig. 5B, lanes a-f). They revealed that the relative rates of formation of these complexes were slower than for the analogous complexes formed in extracts from Vero cells (completion required 1 min for the A. abcdefa The TKB probe was incubated in standard binding reactions containing extracts prepared from either mock-infected (&es a-f) or 5 h-infected (lanes g-l) Vero (A) and HeLa (B) cells for the times indicated. Lanes a and l, 15 s; lanes b and /z, 30 s; lanes c and j, 1 min; lanes d and i, 5 min; lanes e and h, 10 min; lanes f and g, 30 min. Unlabeled TKB fragment (250-fold molar eq) was then added to each reaction, and the samples were immediately loaded onto running low ionic strength 4% polyacrylamide gels. Only the regions of the gels containing complexes are shown.
slower migrating species (H2) and 10 min for the rapidly migrating complex (Hl)). In contrast to the differences in apparent association rates for the two complexes formed in extracts from mock-infected cells, all of the complexes formed in extracts from infected cells accumulated at the same rate ( Fig. 5& lcznes l-g). Complex formation was linear with time, and the yield for all complexes continued to increase during the 30-min incubation period. These data suggest that initiation of complex formation in extracts from infected HeLa cells is rate-limiting.
The marked differences in the observed association rates of probe with proteins in the extracts from the mock-infected and infected cells indicate that the presence of virus proteins alters protein-protein interactions that catalyze complex formation. Furthermore, the linear increase in complex formation over time suggests a stoichiometric relationship between infected cell proteins and their target.
Relative Dissociation Kinetics of TKB Complexes-The rate of dissociation of preformed TKB-protein complexes was studied after allowing reactions to come to equilibrium. Dissociation was initiated by addition of a 250-fold molar excess of unlabeled TKB competitor DNA. The vast majority of any released protein(s) will bind this DNA rather than rebind the probe. Samples taken at intervals after addition of the competitor DNA were analyzed by the mobility shift electrophoresis assay in a running polyacrylamide gel.
In extracts from mock-infected Vero cells, both complexes started dissociating by 10 min; however, even after 20 min the majority of each complex remained (Fig. 6A, lanes u-e).
In contrast, the complexes formed in extracts from infected cells began to dissociate in less than 1 min and were completely dissociated by 2 min (Fig. 6A, lunes 1 and !z). Similar results were obtained with HeLa cell extracts. However, dissociation of complexes formed in extracts from mock-infected HeLa cells was more rapid than for the comparable complexes in Vero cell extracts (Fig. 6, A and B, compare lune a to lune f). In these extracts, complexes began to dissociate at 2 min and were completely dissociated by 20 min (Fig. 6B, lunes cze). Again, complexes formed in extracts from infected cells dissociated much more rapidly (Fig. 6B, lunes 1 and /z).
From these results we infer that the TKB-protein interactions are modified as a result of the presence of virus proteins (at least ICP4) so that the observed dissociation rates of the TKB-protein complexes were allowed to form under standard reaction conditions (30 min) in extracts prepared from either mock-infected (lanes a-f) or 5-h-infected (lanes g-1) Vero (A) and HeLa (B) cells. Reaction mixtures were then chased by the addition of 250-fold molar eq of unlabeled competitor TKB DNA. At the times indicated after the chase, samples were applied directly to running low ionic strength 4% polyacrylamide gels. Only the regions of the gels containing complexes are shown. Lanes a and l, 1 min; lanes b and k, 2 min; lanes c and j, 5 min; lanes d and i, 10 min; lanes e and h, 20 min; lanes f and g, control reactions, not chased.
complexes formed in extracts from infected cells are much more rapid than for the corresponding complexes in extracts from mock-infected HeLa or Vero cells. Requirement of Divalent Cations for TKB Complex Formation-Recent reports that certain DNA-binding proteins require divalent cations, especially Zn2+, for activity (Berg, 1986;Blumberg et ul., 1987;Hanas et ul., 1983;Johnston, 1987;Kadonaga et al., 1987) directed us to investigate the dependence of TKB-protein complex formation on divalent ion(s). If these metal ion(s) are required, then the binding activity of the putative protein(s) might be inhibited by addition of l,lOphenanthroline (OP), a potent metal chelator. To test this, nuclear extracts were incubated for 20 min in binding buffer with different concentrations of l,lO-phenanthroline, and aliquots of each of the treated samples were assayed in binding reactions. Pretreatment of extracts from mock-infected and infected Vero cells with increasing concentrations of l,lOphenanthroline (0.5-2 mM) slightly decreased complex formation. This effect was most pronounced for the slower migrating species (V2) formed in extracts from mock-infected cells (Fig. 7A, lanes e-c and h-j). Identical experiments using extracts from HeLa cells revealed that l,lO-phenanthroline abolished formation of all complexes (Fig. 7B, lunes e-c and /r-j). Formation of cell-specific complexes was always more sensitive to inhibition by the chelator (compare lunes e and !z in Fig. 7B). Preincubation of the extracts with a nonchelating structural isomer, 1,7-phenanthroline, did not affect complex formation under conditions where the 1,lO derivative was strongly inhibitory (Fig. 7B, lunes b and h). Two other metal chelators tested, 2,2'-dipyridyl and EDTA, also prevented complex formation although they were required at higher concentrations (Fig. 8, A and B, lanes a and b). The inhibition profile of 2,2'-dipyridyl was identical to l,lO-phenanthroline in extracts from both mock-infected and infected cells (compare lanes a in Fig. 8, A and B, to lanes e and h in Fig. 7B). However, neither compound had any effect on complex formation when tested in extracts from Vero cells (data not shown). These experiments demonstrated that metal ion(s) are stringently required for complex formation in extracts from HeLa cells but not in extracts from Vero cells.
In analogous studies on the RNA polymerase III transcrip- OP was added to nuclear extracts prepared from either mock-infected (&es a-f) or 5-h-infected (lanes g-l) Vero (A) and tion factor TFIIIA, Zn*+ could be chelated from the free factor but not when TFIIIA was complexed to RNA within a 7 S ribonucleoprotein particle (Hanas et ul., 1983). This observation led us to test the stability of preformed TKB-protein complexes after addition of l,lO-phenanthroline. Complexes were allowed to form and after 30 min, l,lO-phenanthroline was added to 1 mM, and incubation was continued for 20 min. When TKB-protein complexes formed in extracts from either mock-infected or infected Vero cells were challenged with l,lO-phenanthroline, they remained resistant to the chelator (Fig. 7A, lunes a and l). Thus, if metal ion(s) are involved in binding, they are sequestered by proteins that are complexed to DNA. This presumably occurs only if the metal ion(s) are actually a component of the DNA-binding domain(s). The same experiment performed with binding reactions containing extracts from HeLa cells revealed that the DNA-protein complexes were unstable after addition of the chelator (Fig. 7&  lanes a and 1).
To identify ions that restore the ability to form TKBprotein complexes in metal-depleted HeLa cell extracts, a series of reconstitution experiments was performed. After preincubation of the extracts with 1 mM l,lO-phenanthroline in binding buffer for 20 min, various divalent ions were added to a final concentration of 1.25 mM, and incubation was continued for 20 min. Treated samples were then assayed for binding activity. Apparently, both cell-and virus-specific complexes were formed by the subsequent addition of MnClz (Fig. 8, A and & lune i). No other cations tested restored complex formation after l,lO-phenanthroline treatment, apart from a weak activity seen when CdC12 was added to the treated extracts from infected cells (Fig. 8B, lane k). Interestingly, of allethe divalent metal ions te$ted, the ionic radius of Cd*+ (0.97 A) is closest to Mn'+ (0.91 A) (Pauling, 1960). Ionic size undoubtedly plays a role in the high selectivity of some proteins for their target species of metal ion. We cannot conclude from this result which ion is associated with the cell factor(s) in uiuo or if one ion is able to functionally substitute for the other when provided after chelation. However, the amount of cadmium contaminant present in the MnC& preparation (~5 nM) is far below that needed to weakly restore some binding activity in our assay. Incubation of the metaldepleted extracts with Mg+ (an attractive candidate for substituting Mn2+ because of their similar coordination properties) led to degradation of the DNA probe. This reconstitution experiment demonstrates that Mn2+ is required for complex formation in extracts from both mock-infected and infected HeLa cells and suggests that common factors are used to generate these complexes.

TKB-Protein
Znteractiom-To determine the effect of the observed differences in kinetic stability (between complexes formed in extracts from mock-infected and infected cells) on DNA-protein interactions over the TKB region, we attempted to identify the site(s) of protein binding on this sequence by performing DNase I footprinting (Galas and Schmitz, 1978) of both free and complexed DNAs (Singh et ul., 1986). The TKB fragment was first labeled in the noncoding strand and incubated with extracts from either mock-infected or infected Vero cells, as described under "Materials and Methods." The reaction mixtures were then briefly treated with DNase I and resolved by electrophoresis in a nondenaturing polyacrylamide gel. Free DNA (from each reaction) and DNA derived from cell-(Vl) and virus-specific (V4) complexes (see Fig. 2) were eluted and analyzed by electrophoresis through a sequencing gel. The gel patterns demonstrated that single bases as well as short nucleotide stretches on both DNA strands throughout the length of the probe were partially protected from nuclease activity (Fig. 9A). An A at +3, three Cs at +9, +24, and +25, and a T at +28 (all indicated by arrowheads) were protected in both extracts when the noncoding strand was examined. The pattern obtained after incubation with the extract from mock-infected cells was more complex and revealed additional protection or enhancement of nucleotides from cleavage, not detected after incubation with the extract from infected cells. across the sequence and included the region from +28 to +45 (Fig. 9A). When labeled in the coding strand, the DNase I cleavage pattern of the probe incubated with extracts from mock-infected cells was again one of scattered protected nucleotides without enhancements (Fig. 9A). An A at +29, a T at +31, and a G at +51 (indicated by arrowheads) were the only nucleotides protected in both extracts. The pattern after incubation with the extract from mock-infected cells revealed that the region from +1 to +42 was protected (Fig. 9A).
There are few differences between the footprinting patterns of free and bound DNAs isolated from reactions containing infected cell extracts (complex V4). However, both of these patterns appear to be very similar to the one seen with bound DNA isolated from reactions containing extracts from mockinfected cells (complex Vl). Identical results were obtained when the V2 and V3 complexes were subjected to this analysis (data not shown). This observation suggests that the so-called "free" DNA from reactions containing extracts from infected cells reflects rapid dissociation of proteins from the DNA target. This could occur if kd>>k= and is predicted by the kinetic experiments, which demonstrated that complexes formed in extracts from infected cells are kinetically much more labile compared with those formed in extracts from mock-infected cells. Because it was not possible to generate similar footprinting patterns when the TKB fragment was incubated with HeLa cell extracts, we turned to the missing contact method (Bru-nelle and Schleif, 1987). In this technique, the DNA was sparingly methylated and subsequently depurinated, then bound, separated, and analyzed after chemical cleavage to identify guanines involved in complex formation. Gel retardation analysis of the binding reactions after treatment demonstrated no significant effect on the intensity or mobility of the complexes after methylation and depurination. When the TKB fragment was subjected to depurination and then incubated with extracts from HeLa cells, we observed that G residues present between +34 and +53 were underrepresented only when the extracts were prepared from mock-infected cells (complex Hl; Fig. 9B). It was again unexpected that no differences in the pattern between free and bound DNAs were detected in infected cell extracts by this analysis (complex H4; Fig. 9B), indicating that guanines located in this region participate in specific noncovalent interactions with cell factors which are altered by the presence of virus proteins (at least ICP4).

DISCUSSION
Mobility shift analysis was used to demonstrate that DNAprotein complexes can form between a sequence downstream from and adjacent to the promoter for the HSV-1 thymidine kinase gene (composed mostly of sequences encoding the leader RNA) when it is incubated in nuclear extracts prepared from either mock-infected or infected Vero and HeLa cells. Our studies included extracts from each of two cell types because of the different response patterns of virus a promoters when introduced into these cells (Everett, 1988, Gelman andSilverstein, 1987). The complexes that form are characteristic of the cell type from which the extract was prepared and differ both qualitatively and quantitatively depending on the extract used to generate them. Infected cell extracts showed altered patterns of complex formation; novel virus-specific complexes are present. The ability to form complexes with the same mobility as those formed in extracts from mockinfected cells diminished as a function of the time postinfection. Thymidine kinase transcripts can be detected as early as 2 h postinfection.
They accumulate to maximum levels at 5 h of infection and, in contrast to the a mRNAs, decline steadily to nearly undetectable levels by 10 h late in infection (Harris-Hamilton and Bachenheimer, 1985;Weinheimer and McKnight, 1987). Although the time course of synthesis of thymidine kinase RNA was not verified in this study, temporal analysis of extracts prepared at different times postinfection showed that those prepared from the 5-h infected cell samples were the most active in binding the @ gene sequence. The 2.5-h extracts formed predominantly complexes with the same mobility as those formed in extracts from mock-infected cells. V4 and H4 were barely detectable in the 10.5-h extracts. In each instance ICP4, the major virus regulatory protein, was demonstrated to be a component of these complexes. These late extracts preferentially bind 7 gene promoters and form specific complexes.3 The nature of the complexes formed in extracts from mockinfected and infected cells was examined in greater detail. The electrostatic interactions required to form and maintain these complexes were examined in a qualitative way by studying the effects of increasing ionic strength on complex assembly and stability. These analyses demonstrated that complex formation in extracts from mock-infected and infected cells was similar between cell types. In each instance formation of one of the two complexes was sensitive to 200 mM NaCl, while the other complex could still form at this elevated salt con-centration.
We note that the complexes formed in extracts from Vera cells (Fig. 3A) are considerably more abundant than those formed in extracts from HeLa cells (Fig. 3B). Therefore, the presumed differences in salt effects on complex formation might only be qualitative. Complexes formed in extracts from mock-infected HeLa cells were unstable when exposed to NaCl concentrations other than their formation optima (50 mM). The ICP4-containing complex formed in extracts from infected HeLa cells (H4) was stable when exposed to moderately higher ionic strength. However, this complex did not form at these elevated salt concentrations. In contrast, the complexes formed in extracts from infected Vero cells were stable to increasing NaCl concentrations over the entire range examined. The apparent differences in salt optima for complex formation and the altered stability profiles of complexes between the two extracts from infected cells demonstrate that there are distinct requirements for recognition and interaction of sequences within the probe by proteins in these extracts. Once formed, the infected cell complexes are resistant to increases in ionic strength. This suggests that infection leads to changes in the electrostatic interactions between host proteins and the target DNA in the formed complexes, rendering them resistant to the effects of high salt.
TKB complex formation was also examined by probing the relative association and dissociation kinetics of the complexes formed in the various extracts. The rates of formation of Vl and VZ were similar. In contrast, V3 formed more slowly in extracts from infected cells, requiring 30 min to complete the reaction. Formation of the ICP4-containing species (V4) occurs more rapidly than V3. However, the V4 complex begins to decay between 1 and 5 min of incubation, suggesting that the rate of its formation declines with time. This might occur if it was converted to another labile complex. The dissociation studies revealed that the complexes formed in extracts from mock-infected cells are relatively stable. However, the complexes formed in extracts from infected Vero cells rapidly dissociate. In extracts from mock-infected HeLa cells the rate of complex formation is slower than in extracts from Vero cells. In contrast to the results obtained with infected Vero cell extracts, incubation in extracts from infected HeLa cells reveals that all of the complexes continue to accumulate throughout the time period examined. This observation might reflect cooperative assembly of proteins on the DNA template. The results of the dissociation studies were identical to those observed with extracts from Vero cells. Because these assays were carried out with crude nuclear extracts, we have not attempted to interpret the results quantitatively in terms of actual binding affinities.
Although DNase I protection and missing contact analysis address different aspects of DNA-protein interactions, the two techniques gave similar results when used to examine the complexes formed between the TKB sequence and extracts from mock-infected Vero and HeLa cells. A region between +28 and +53 on the noncoding strand was preferentially protected by host proteins (Fig. 9, A and J3). In contrast, the coding strand was completely protected (Fig. 9A). These results show that binding is occurring across the entire TKB sequence (at least in extracts from Vero cells) and suggest that the 3' region is involved in more intimate contacts (in both extracts). There were very few (DNase I protection) or no (missing contact analysis) differences in the patterns obtained with free and bound DNAs after incubation in extracts from infected Vero or HeLa cells, respectively. However, both of these patterns were most like those seen with the complexed fraction after incubation in extracts from mock-infected cells and not with the pattern of the corresponding free DNA. To explain this observation we suggest that if DNA-protein complexes are slowly formed, rapidly dissociate, are hit by DNase I, and then reassemble, a mixed population of free and bound molecules with similar digestion patterns would result.
The kinetics of dissociation of the complexes formed in infected cell extracts suggests a biochemicai basis for the failure to generate DNase I digestion patterns that differ between the free and bound probe. We believe that this results from alteration of or modification to host proteins, leading to weaker interactions between them and the DNA sequences they recognize. Therefore, because of the very short lifetimes of the complexes in infected cell extracts (less than 1 min), bound proteins can rapidly detach from the protected DNA fragment and slowly reassociate to other fragments that have already been nicked by DNase I or depurinated, rendering it difficult to observe a clear footprint.
Nevertheless, "stable" complexes are still detected by the mobility shift electrophoresis assay; this may result from an inherent stabilizing effect of the gel. Previous studies suggested that the exclusion volume of the gel increases the local concentration of the reactants, shifting the equilibrium toward reassociation. Therefore, even kinetically labile complexes can be detected by this technique (Fried and Crothers, 1981;Garner and Revzin, 1981).
In support of our hypothesis that common cell-specific factors recognize the DNA probe in extracts from mockinfected and infected cells, we note that the metal requirements for complex formation in both extracts from HeLa cells are very similar. In each instance they are sensitive to three different chelators, and their binding activities are restored by supplementation with Mn'+. Moreover, both cell-and virus-specific complexes formed in extracts from HeLa cells are susceptible to the addition of l,lO-phenanthroline.
These observations indicate that a common Mn*+-requiring protein(s) is involved in forming the DNA-protein complexes in both extracts and that the metal is important, for complex stability, perhaps by catalyzing a protein multimerization process occurring after binding to the DNA template. A role for Zn*+ and Cd*+ in protein dimerization was recently described for the human immunodeficiency virus Tat protein (Frankel et ul., 1988). It is also possible that the metal is important for maintaining the overall protein(s) structure and perhaps nonspecific interactions with DNA, whereas distinct domains are involved in direct contact with DNA. Furthermore, the resistance to inhibition of complex formation by l,lO-phenanthroline and 2,2'-dipyridyl in extracts from infected cells suggests that conformational changes in the DNA-bound metalloprotein(s), induced by proteins found in these extracts (i.e. ICP4), interfere with the accessibility of the metal-binding site(s). The refractile nature of the complexes formed in extracts from Vero cells emphasizes the differences between the two cell types. It is conceivable that if analogous metalloproteins are involved in complex formation in these extracts, the metal(s) are more tightly bound by the protein(s) and therefore inaccessible to the chelating action of l,lO-phenanthroline.
Alternatively, steric hindrance might interfere with the penetration of the chelators to the metal-binding site(s). From these results we infer that newly synthesized virus proteins modify and/or interact with preexisting host proteins to alter the binding properties of complexes that form between these cell proteins and the TKB region. We propose that this interaction reflects changes that occur during the course of the HSV infection cycle and can be explained hy a model where cellular proteins recognize and bind to a site(s) in @ promoters