The Use of ATP and Initiating Nucleotides during Postrecruitment Steps at the Activated Adenovirus E4 Promoter*

Permanganate probing has been used to follow the progress and ATP dependence of promoter opening during activated adenovirus E4 initiation and clearance. Using templates designed to restrict synthesis to de-fined positions, formation of a 3-nucleotide-long RNA was found to be sufficient to trigger expansion of the initial transcription bubble. This occurred by a discrete transition that expanded the downstream limit of melting from position 1 to 15. Subsequent clearance of the bubble from the promoter region also occurred without detectable intermediates. Thus, initial opening, extension, and the clearance of the promoter bubble appear to occur as discrete, unique transitions. The apparent K m values for these three steps were determined to be near 5, 9, and 50 m M , respectively. Comparison of these values with ATPase activities within known transcription factors raises the possibility that different activities could be responsible for each step.

Transcription initiation by RNA polymerase II can be divided into a number of steps (reviewed in Ref. 1). Initially, the polymerase is recruited to the promoter in the closed complex state. In this state, the DNA remains fully double-stranded. Transcription cannot initiate because the template strand cannot be read in double stranded form. In a series of post-recruitment steps, the DNA is opened over the transcription start site, allowing initiation to begin. During the subsequent promoter clearance steps, the polymerase elongates away from the promoter along with the transcription bubble of melted DNA (2). Each round of initiation requires that a new polymerase be recruited into closed complexes and that the DNA be reopened.
Transcription initiation has long been known to require hydrolysis of the ␤-␥ bond of ATP (3,4). The hydrolysis was shown to be in a postrecruitment step that preceded elongation (5). Subsequently, permanganate probing was used to identify open complexes, and ATP was shown to be required to open the DNA within closed complexes at a variety of promoters (6,7). Promoter clearance was also shown to depend on ATP (2,8,9).
More recently the postrecruitment pathway has been subdivided into several steps. Parallel studies in two different systems identified three consecutive postrecruitment steps (10,11). In the first step, the DNA upstream from the start site is opened (to form the preinitiation open complex (POC)). 1 In the second step, the melted bubble is enlarged as DNA downstream from the start site is opened (to form the initiation open complex (IOC)). Formation of the IOC is triggered after initiation of RNA synthesis. When ATP is depleted, the POC decays rapidly, but the IOC does not, implying that the IOC has much higher stability. Finally, the polymerase and the bubble exit the promoter region as the complex enters elongation mode. In at least one of these systems, ATP hydrolysis is required for all of these postrecruitment steps and also for the formation of the first bond of the RNA (11).
Because each of these steps has the potential to be regulated, some effort has been applied to identify the activities that catalyze them. General transcription factors TFIIE and TFIIH have been implicated in these postrecruitment steps (8,12). Mutations that inhibit a helicase activity of TFIIH impair the ability to reach the IOC stage (13). Several lines of evidence implicate the C-terminal domain kinase activity of TFIIH in the final conversion to the elongation stage (2, 14 -16). Both helicase and kinase activities use ATP, which supports their involvement in transcription initiation.
Several important issues surrounding these postrecruitment steps remain to be resolved. These issues center on how the promoter is opened initially to form the POC and how the POC is converted to the IOC during initial RNA synthesis. We showed previously that all postrecruitment steps require ATP hydrolysis (11). This includes the initial opening of the DNA and the formation of the first RNA bond. There have not been studies to identify which activities use ATP in these steps. Although the TFIIH helicase is an obvious candidate activity for steps prior to the IOC stage, it has not been implicated specifically in POC formation. Moreover, helicases rarely if ever initiate opening from unmodified fully double-stranded DNA without assistance from other activities (17). Thus, this activity is likely required to extend the initial bubble, but what causes the initial bubble to form is not known.
In addition, the POC to IOC conversion appeared to occur differently in the two systems studied. In one case, the conversion occurred in a discrete step triggered by the initiation of RNA synthesis (11). In the other case the conversion was gradual and was not complete until a substantial length of RNA had been produced (10). The more rapid and discrete stabilization process was observed in an activated unfractionated system, allowing for the possible involvement of factors absent from highly purified basal system. Thus, it is not yet clear how many activities are involved in forming and clearing the open DNA during initiation. Significant uncertainty surrounds how DNA opening is initiated and how RNA synthesis is involved in extending and stabilizing the open complex.
This study focuses on how ATP is used in order to define how many activities might be involved in these processes. In addition, it explores the applicability of the discrete model for conversion to the IOC using an experimental system with tighter regulation of RNA synthesis than used previously (11).
The results lend further support for the postulated pathway for formation of the IOC and also provide constraints on the types of activities that may be involved in the use of ATP during initiation.

EXPERIMENTAL PROCEDURES
General Materials-Ultrapure NTPs, dNTPs, and 3-O-methyl-GTP were from Amersham Pharmacia Biotech. Ribodinucleotide UpA and ATP␥S were from Sigma. The DNA template G9E4T contains nine Gal4 binding sites upstream of a truncated adenovirus E4 promoter (18).
Potassium Permanganate Footprinting-Permanganate probing in a HeLa extract was done as described (6,11). To measure the K m for the POC formation, various concentrations of dATP were added 2 min before permanganate treatment. To measure the K m for the IOC formation, 2 mM UpA and 25 M CTP were added before the dATP and permanganate treatment. To measure the K m for the promoter clearance, various concentration of dATP were added for 2 min, followed by the addition of the elongation mixture (50 M GTP, CTP, UTP, and 2.5 M ATP) for 2 min and permanganate treatment.
Quantitation of results used ImageQuant software for PhosphorImager results (Molecular Dynamics, Inc., Sunnyvale, CA). Background subtraction used a common region of each lane, not associated with melting. Signals were normalized to the strongest POC or IOC signal or the maximum amount of clearance in appropriate experiments. Data from several experiments were averaged and fit into the classic Michaelis-Menten equation using SigmaPlot (version 4.0, SPSS Inc.). The K m values and S.D. values were calculated using the same software.

Full Bubble Extension Is Complete when a 3-Nucleotide-long
RNA Is Made-The first experiments address the role of initial RNA synthesis in extending the POC bubble to form the stable IOC. In our prior study of the activated adenovirus E4 promoter, this transition was sudden and discrete (11); i.e. when formation of a 3-mer RNA was triggered by the addition of dinucleotide UpA and CTP, the downstream edge of the transcription bubble moved from near position 1 to near position 15. This result contrasted with results from a basal adenovirus major late system, where the downstream edge extended gradually as the initial bonds of the RNA were made (10). Another difference was that only the latter study used a series of constructs that stop synthesis definitively at set points by incorporating the chain terminator 3Ј-O-methyl-GTP. By including G at different template locations, one can use such templates to eliminate potential nucleotide misincorporations (10). We created a series of modified templates to test the discrete model in the situation where chains can be terminated definitively with 3Ј-O-methyl-GTP.
The parent G9E4T template contains nine Gal4 binding sites upstream of an adenovirus E4 promoter. Transcription starts within the T 6 A region approximately 25 base pairs downstream from the TATA box. For convenience, we designate the last thymidine in the T6 sequence as ϩ1. Fig. 1A shows the series of templates constructed and used in these studies. Each variant has a G substituted in a different position, including positions 2, 3, 4, and 5. As discussed elsewhere, the use of 3Ј-Omethyl-GTP allows synthesis to be restricted to RNAs terminated at these positions (10). We confirmed that each of these templates can be transcribed (data not shown).
The transition through the different stages of open complex formation can be visualized using permanganate footprinting (6). Permanganate can selectively modify the thymidine residues in the single strand region of the DNA template, thus causing a stop when copied with Taq polymerase in a subsequent primer extension protocol. The region that has been melted is detected as a series of hypersensitive bands upon autoradiography of the primer extension products separated on gels. In this set of experiments, permanganate is used to detect the downstream limits of bubble formation on templates where G nucleotides are present to ensure termination at different points.
The experiment of In the next experiment, the DNA with the initial sequence TAGAC was used to restrict synthesis in conjunction with the chain terminator 3-O-methyl-GTP. Lane 5 shows the permanganate signal obtained when this template is used with the dinucleotide UpA and 3Ј-O-methyl-GTP (expected product of UpApG). The addition of these nucleotides has triggered the IOC pattern (compare with lane 3) rather than the POC pattern (compare with lane 2). The overall signal is somewhat weaker, perhaps indicating that the conversion is not quite complete. The downstream edge of the bubble appears to have reached its full extension, previously shown to approach position 15.
Although the presumptive UpApG product in this experiment contains two RNA bonds, only a single one has formed due to the use of a dinucleotide primer. Another experiment indicates that the existence of an RNA product of 3 nucleotides in length may be more important than the synthesis of a single bond. When ATP and 3Ј-O-methyl-GTP are used with this same template (lane 4) there is, at best, only weak extension of the bubble to form the IOC. Under these circumstances the expected product is only a 2-mer, pppApG. The comparison indicates that formation of a 3-nucleotide-long RNA is sufficient for formation of an IOC with the bubble extended fully to near position 15.
This discrete transition is also observed in protocols allowing extension to positions 4 and 5 without dinucleotide primers (lanes 6 and 7). These used ATP, CTP, and 3Ј-O-methyl-GTP on templates TACAG (potential product pppApCpApG) and TA-CACG (potential product pppApCpApCpG). The patterns in lanes 6 and 7 are slightly more hypersensitive in the downstream region (compare with lane 5), indicating a slightly higher efficiency of IOC formation. The downstream limit of the extended bubble appears to be the same in all three circumstances.
It should be emphasized that these experiments use a high concentration of 3Ј-O-methyl-GTP, so the possibility of producing longer transcripts due to misincorporation is very low. Direct observation of RNA did not show such products (Ref. 19 and data not shown), but they would not necessarily be detectable in the crude extract system. In the system with purified factors where longer products could be seen, 3Ј-O-methyl-GTP was able to arrest chain elongation at the first G position (10). We further confirm the termination efficiency of 3Ј-O-methyl-GTP by using a TAGAG template. The combination of ATP and GTP (with the expected product pppApGpApG) is sufficient to cause IOC formation (lane 8). However, the substitution of 3Ј-O-methyl-GTP for GTP blocks IOC formation (lane 9), demonstrating that it has prevented any read-through due to misincorporation.
The data indicate that when the IOC forms the opening always extends as far as position 15 (lanes 3 and 5-8), whereas the POC opening never extends beyond the start site (lanes 2, 4, and 9). The results indicate that the transition from POC to IOC is not gradual and is accompanied by a sudden extension of the transcription bubble. This extension happens when an RNA as short as a 3-mer is formed and appears to be somewhat more efficient when longer RNA is made. The results contrast with a prior study in a different experimental system in which the leading edge of the bubble extended gradually as the RNA became longer (Ref. 10; discussed below).
The Apparent K m Values for POC and IOC Are in the Range of 4 -10 M-It is apparent from these and prior data that the formation of the POC and IOC are different processes. Formation of both complexes requires ATP. It is not yet known if the same activities are involved in the creation of these two complexes. Recent results using mutant forms of TFIIH measured the formation of the IOC only; the effect of mutation on the POC was not assessed (see Introduction). In order to place constraints on which activities might be involved, we measured the K m for these steps. We use dATP rather than ATP to avoid potential complications from the use of ATP as an elongation substrate. The result will allow these K m values to be compared with each other and with values for various activities that are candidates for triggering formation of the POC and the IOC.
To measure the apparent K m for dATP, its concentration was varied, and the extent of POC formation was assayed. dATP concentrations up to 125 M were added to reactions. Two minutes later, permanganate was added, and opening was assayed in the standard manner. The result ( Fig. 2A) shows the increasing POC signal that is observed with increasing dATP concentration.
The extent of POC formation was quantified by PhosphorImager analysis. The average of four trials was used in the data display of Fig. 2B. The results indicate an apparent K m for dATP of about 4.8 M with an S.D. value of 0.9 M. This value is consistent with previous indications that the K m of POC formation should be under 10 M for dATP (20).
The apparent K m for IOC formation was determined using a protocol modified to trigger bubble extension. In this protocol, UpA and CTP are included as substrates for UpApC formation. Varying concentrations of dATP are added to trigger IOC formation. Fig. 3A shows the resulting permanganate patterns. Part of the IOC-specific signal (bar at right) was quantified and analyzed as for the IOC data. The results of five trials were averaged and are shown plotted in Fig. 3B. The apparent K m for dATP for IOC formation is about 8.5 M, with a S.D. of 1.6 M.
This IOC K m is approximately 2-fold higher than that obtained for POC formation. Qualitatively, the raw data in the autoradiograph also confirm that it takes less dATP to trigger the upstream opening typical of POC formation than to trigger bubble extension to form the IOC. For example, the upstream melted DNA signal in the fifth lane from the left is nearly saturated as judged by comparison with the seven lanes to its right. However, in this same lane, the IOC-specific part of the signal is significantly below saturation (see bracket at right). This comparison shows directly that bubble extension requires a higher dATP concentration than upstream opening, although the difference may be modest.
IOC and POC Formation Are Inhibited Similarly by ATP␥S-It was shown previously that a large excess of the nonhydrolyzable ATP␥S inhibited formation of both the POC and the IOC (11,20), consistent with prior reported effects during the early stage of transcription initiation (5, 21). If the activities catalyzing both steps are similar, they would be expected to be affected similarly by this inhibitor. In this exper- iment, we used a low concentration of inhibitor to compare the responses in formation of the POC and the IOC.
The assays used were like those just described but performed in the presence of 25 M ATP␥S. The amount of dATP was varied from 0 to 125 M. Separate experiments determined the amount of POC formed (Fig. 4A) and IOC formed (Fig. 4B). The inhibition is apparent in both cases; i.e. at 12.5 M dATP there is little formation of POC (Fig. 4A, lane 3) as well as little formation of IOC (Fig. 4B, lane 3). This is in contrast to the extensive formation of both complexes at 12.5 M dATP seen when no inhibitor was present (see Fig. 3). Qualitatively, both complexes are inhibited similarly. Quantitative analysis (not shown) indicates that the apparent K m for each process is raised by roughly 1 order of magnitude in the presence of 25 M ATP␥S. We infer that the activities that catalyze POC and IOC respond similarly to this inhibitor.
The Promoter Clearance Step Has an Apparent K m Near 50 M-The step following IOC formation is the escape of RNA polymerase from the promoter into elongation phase. The escaped polymerase is accompanied by the transcription bubble, and both clear the promoter region together. We devised an assay to follow the disappearance of the bubble from the promoter region. The disappearance is monitored at different dATP concentrations to evaluate the K m for the clearance step.
Such an assay is complicated by the need for ATP as an elongation substrate for transcription. Thus, ATP must be present to allow the polymerase to move away from the pro-moter. But because the steps prior to clearance also require ATP, the aggregate effects of low ATP could be confusing. This complexity can be bypassed if the K m for clearance is substantially higher than that for the prior steps. This turned out to be the case, and the following protocol proved to be successful.
In this protocol a mixture of 50 M each of CTP, GTP, and UTP and 2.5 M ATP is added to reaction mixtures. This amount of ATP is sufficient to allow POC formation (Fig. 5A, leftmost lane). However, it is too low to allow efficient clearance, as indicated by the retention of the permanganate signal over the promoter region. If 1 mM dATP is added (rightmost lane), the permanganate signal is substantially depleted. This demonstrates that a high concentration of dATP can trigger clearance of the bubble.
The experiment is then to simply vary the concentration of dATP and follow the clearance of the bubble from the promoter region. Fig. 5A shows the resulting autoradiographs, which demonstrate progressive degrees of clearance triggered by increasing dATP concentrations. The transition to clearance is a discrete rather than a gradual process; i.e. there is no evidence in the patterns that inefficient clearance at low dATP concentrations gives complexes where the bubble has changed its location. Instead, the patterns at low and high dATP concentrations are essentially similar; there is just a lower degree of melting upstream from the start site at the high concentrations at which clearance has occurred efficiently.
The data were analyzed quantitatively by PhosphorImager analysis, and four trials were averaged (Fig. 5B). The apparent K m for promoter clearance is 53 M, with a S.D. of 12 M. This is obviously greater than the K m values for the prior steps, as measured above.
Overall, the data show that the three consecutive postre- cruitment steps occur via activities that are associated with progressively higher K m for dATP. Below we discuss the relevance of this to the postrecruitment pathway.

DISCUSSION
Prior studies of postrecruitment processes at the promoter have shown that there are three critical consecutive steps (2,10,11). First, the DNA is opened, then the bubble is extended and stabilized, and finally the polymerase and the bubble clear the promoter as the transcript is elongated. The current data on an activated adenovirus E4 promoter indicate that the transitions between the steps occur discretely rather than gradually. The bubble extension to form a stable complex occurs reasonably efficiently when a 3-nucleotide-long RNA is made. The full number of base pairs is melted at this stage. The transition to clearance is also discrete. The three consecutive steps use activities associated with progressively higher K m values for dATP. The first two steps, initial opening and bubble extension, respond similarly to an inhibitor. These data have both similarities and differences to data from the literature. Below we attempt to place the data in context.
The Activated Transcription System Involves Early and Discrete Bubble Extension during Initiation-A primary area of prior uncertainty involves the relationship between initial unstable promoter opening and the extension of the bubble to form a fully stable open complex. Two unresolved types of questions have arisen. 1) How is the bubble extension triggered and maintained? 2) What protein factors are involved, and does the same activity catalyze the initial opening and the extension? We consider the transition first.
The current data indicate that the transition is virtually complete when a 3-nucleotide-long RNA is made. The opening extends a full helix turn downstream from the edge of the RNA, and the entire downstream region appears to open at once. Thus, the transition appears to be between one discrete complex and another.
Aspects of these data differ from a prior study at the adenovirus major late promoter (10). Some of these differences were seen previously (11). In the major late case, the transition occurred gradually and apparently was not between two fully discrete complexes. Instead, as the RNA grew progressively longer the bubble also grew progressively longer. There are several differences between the two experimental systems, and the new data reported here help to decide which differences are the most relevant. One difference in the prior studies was that only the major late templates were engineered to prevent unintended read-through via placement of sites for the chain terminator 3Ј-O-methyl-GTP. In the current study, we engineered an analogous series of templates at the E4 promoter. The use of these showed that this was not the source of the different results in the two systems; the more detailed studies presented here still demonstrated a discrete transition early in the initiation process. Other data indicate that the different promoter used was also unlikely to be the source of the difference; i.e. prior less detailed study of the major late promoter showed that formation of a 3-nucleotide-long RNA was sufficient to catalyze opening downstream the start site (12,22).
The remaining system difference is that the current E4 system relies on activated transcription from unfractionated extracts, whereas the comparison major late promoter study used a highly purified basal system. We suggest that factors present in the unfractionated activated system allow the transition to occur earlier and more efficiently. Thus, the activated system can trigger the discrete transition as soon as a very short RNA is made. We suggest that the normal pathway is likely to proceed in this manner.
Potential Activities Used in the Three-step Postrecruitment Pathway-The current data show that the three postrecruitment steps have a progressively higher K m for dATP. The transcription machinery has multiple ATPase activities within factor TFIIH, two helicases and a protein kinase (23,24). TFIIH is indeed required in the postrecruitment steps along with factor TFIIE (8,13,25,26). The properties of these TFIIH activities may be discussed in terms of the use of ATP in the postrecruitment pathway.
The available data give strongest support to the involvement of the TFIIH kinase in clearance and weakest support for the involvement of TFIIH helicases in the initial opening. The involvement of the kinase in clearance is supported by diverse experiments (Refs. 2, 14, and 15; but see Refs. 27 and 28). The K m value for the use of ATP in a protein kinase assay was found to be 40 M for the kinase component, CAK, of TFIIH (29). This is essentially identical to the K m measured here for the promoter clearance reaction, lending additional support to the involvement of the TFIIH kinase in promoter clearance.
By contrast, there is very poor agreement between the K m values measured here for POC and IOC formation and that for the TFIIH helicases. The K m value measured for the TFIIH "transcription" helicase is approximately 200 M (30 -32), which is in the typical range for helicases (17). This is 20 -50 times higher than the values found here for initial opening and subsequent extension and stabilization. The contrast is supported by prior experiments reporting very low K m values for formation of sarkosyl-resistant initiated complexes (5). The contradiction is not definitive, because the K m value for a helicase in a replication system can be dramatically reduced in the context of a larger macromolecular complex (33). An ATPase activity within TFIIH with a low K m value has been reported, and its coupling to helicase activity was implied (34). Thus, these considerations suggest that TFIIH could be involved, but they do not identify the appropriate activity within it.
Other data support some involvement of the TFIIH transcription helicase. The helicase subunit has been mutated, and the re-assembled TFIIH was shown to be deficient in reaching the stage of a fully extended stable open complex (IOC formation) (13). Initial opening (POC formation) was not assayed. Thus, TFIIH is required to reach the IOC stage in this system. The mutations destroyed the presumptive ATP binding sites of the polypeptide and thus inactivated the associated ATP-dependent helicase activity. This helicase is known to be involved in DNA repair and is a typical repair helicase in that it uses templates at which melting has been initiated by mispairing and then extends the length of the melted region. This activity is analogous to the transition from POC to IOC, where a melted template has its open region extended. Thus, despite the lack of agreement between the measured K m values, the involvement of the TFIIH transcription helicase in the transition to IOC is plausible.
By contrast, there are no data available concerning which activity opens the DNA initially to form the POC. The involvement of a helicase activity would be very unusual in this step. Typically, helicases require a single strand DNA region or an end of double strand DNA to function (17). They rarely, if ever, initiate an unassisted melting reaction within intact duplex DNA. This view is supported by recent structural studies, which show that helicases typically require an initial single strand region to catalyze their melting reactions (35). Moreover, the K m for initial opening measured here is far below that measured for the TFIIH helicase (30 -32) but does correspond to an ATPase activity within TFIIH (34). Therefore, if the TFIIH helicase participates in this step, it must be using an unusual and modified helicase activity. These considerations suggest that other activities should be considered along with the TFIIH helicase for triggering the initial opening of the DNA.
Our K m data show that the initial opening and the extension steps have similar but distinguishable K m values and that the two steps respond similarly to the inhibitor ATP␥S. This is the type of behavior that one might expect from a single ATP binding site used in two different reactions. Thus, it is possible that TFIIH binds ATP and uses nonhelicase determinants to trigger initial opening. This could be coupled to subsequent triggering of the helicase activity, which could use the same ATP binding site to extend the opening in a conventional helicase reaction.
Thus, a key unresolved issue is how opening is initiated. In prokaryotes, this requires a fork junction binding activity that binds double strand-single strand junctions and nucleates melting (36). Such activities have not been assayed in the mammalian machinery and could function in conjunction with TFIIH. In this speculative model, an ATPase activity of TFIIH would work in conjunction with other polypeptides to initiate DNA opening. This initial opening would trigger the helicase activity of the same TFIIH polypeptide, which then extends the bubble as in a typical helicase reaction. Presumably, single strand DNA binding activities within the transcription machinery would be used to stabilize the open complexes. To be fully harmonious with the data of this report, the helicase activity would need to be assisted by the formation of a DNA-RNA hybrid with a length of at least 3 base pairs. Whether or not this speculation proves to be correct, the data presented here should provide significant constraints on any models proposed.