New Insights into the Mechanism of Initial Transcription

Background: RNA polymerases must couple the energetics of nucleotide addition to the timed release of promoter contacts. Results: A mutant that aborts less during initial transcription releases the promoter at longer RNA lengths. Conclusion: Hybrid growth remodels the protein to disrupt promoter contacts; the protein, in turn, pushes back on the hybrid, leading to abortive instability. Significance: The model extends to multisubunit RNA polymerases. RNA polymerases undergo substantial structural and functional changes in transitioning from sequence-specific initial transcription to stable and relatively sequence-independent elongation. Initially, transcribing complexes are characteristically unstable, yielding short abortive products on the path to elongation. However, protein mutations have been isolated in RNA polymerases that dramatically reduce abortive instability. Understanding these mutations is essential to understanding the energetics of initial transcription and promoter clearance. We demonstrate here that the P266L point mutation in T7 RNA polymerase, which shows dramatically reduced abortive cycling, also transitions to elongation later, i.e. at longer lengths of RNA. These two properties of the mutant are not necessarily coupled, but rather we propose that they both derive from a weakening of the barrier to RNA-DNA hybrid-driven rotation of the promoter binding N-terminal platform, a motion necessary to achieve programmatically timed release of promoter contacts in the transition to elongation. Parallels in the multisubunit RNA polymerases are discussed.

RNA polymerases carry out a complex process that involves much more than the simple phosphoryl transfer of nucleotide addition. At the onset of initiation, an RNA polymerase must be highly sequence specific in its binding to promoter DNA. During elongation, however, it should show no sequence specificity. During initiation, an RNA polymerase must melt the DNA strands to expose the templating bases, and during subsequent initial transcription, the enzyme must maintain the resulting bubble at least until the RNA-DNA hybrid is of sufficient length to resist competition from DNA collapse (1). In de novo initiation, the active site must add a nucleoside monophosphate onto a single priming mononucleoside triphosphate and then grow that to at least an 8-bp hybrid during initial transcription. Critically, a mechanism and a driving force must exist to drive the positionally timed release of initially strong promoter contacts to allow the transition to a stable elongation complex. Finally, during elongation, the active site must maintain a constant sized (at least eight base) RNA-DNA hybrid as the accepting species, dissociating RNA from one end of the hybrid, whereas adding nucleotides to the other.
Given these different requirements, it is perhaps not surprising that RNA polymerases undergo significant structural rearrangements in transitioning from an initially transcribing complex to a stably elongating complex (2,3) and that initially transcribing complexes are relatively unstable, producing abortive RNA transcripts, whereas elongation complexes are (and must be) extremely stable during processive elongation (4,5). Indeed, short abortive products have been observed both in vitro and in vivo (6), and the transition from initially transcribing to stably elongating complexes is a target of gene regulation in many systems, most notably in promoter proximal pausing (7).
It is tempting to imagine that RNA polymerases release short abortive products simply because short RNA-DNA hybrids are expected (in the absence of protein, at least) to be unstable. In the model RNA polymerase from bacteriophage T7, however, a single point mutation (P266L), distant from the hybrid, dramatically reduces the relative production of released abortive transcripts (8), pointing to a more complex process involving the protein. Another popular model for abortive complex instability is the energetic stress of DNA bubble expansion and/or the accommodation of increasing DNA bulk within the protein: scrunching (9 -11). The P266L mutation similarly lies distant from nucleic acid, and we have recently shown that such stress contributes at most a small amount to abortive instability in T7 RNA polymerase (12).
Understanding the mutation P266L is key to understanding initial transcription. The initial report of this mutation indicated that the mutant bound the promoter more weakly than wild type enzyme, suggesting that promoter release was a key element of abortive instability. It was proposed that an earlier release of promoter contacts allowed an earlier transition to a stable elongation complex (8).
We now revisit this mutant, showing that the mutant does not bind promoter more weakly than wild type (at least in the binary enzyme-DNA complex). Using a combination of qualitative proteolytic assays and quantitative, globally fit fluorescence and transcription kinetic data, we present evidence that the mutant P266L instead transitions to the elongation configuration later than wild type RNA polymerase. Finally, we propose a model for abortive cycling that accounts for this new observation, coupling it to a structural model for the transition to elongation.

EXPERIMENTAL PROCEDURES
T7 RNA Polymerase-T7 RNA polymerase with a histidine tag was prepared from Escherichia coli strain BL21 carrying the plasmid HB161 (kindly supplied by W. T. McAllister), in which RNA polymerase is expressed under inducible control of the lac UV5 promoter (13). The T7 RNA polymerase P266L mutant was produced by QuikChange mutagenesis (Agilent Technologies, Santa Clara, CA). To confirm that no other mutations were inserted, the entire gene was sequenced using custom sequencing primers. Both native and mutant enzymes were purified in two steps, first by nickel affinity chromatography and subsequently by ionic exchange chromatography. Protein concentration was determined (⑀ 280 ϭ 1.4 ϫ 10 5 M Ϫ1 cm Ϫ1 ) as described previously (14). The purity of the enzyme was characterized by SDS-PAGE electrophoresis.
Oligonucleotides-Oligonucleotides containing 2-aminopurine were synthesized using phosphoramidite chemistry on a PerSeptive Biosystems Expedite 8909 nucleic acid synthesis system. Standard and 2-aminopurine phosphoramidites were purchased from Glen Research (Sterling, VA). Non-modified oligonucleotides were synthesized by Eurofins MWG Operon (Huntsville, AL), and TAMRA 3 label DNA was synthesized by Integrated DNA Technologies (Coralville, IA). A complete listing of oligonucleotides used in this study is presented in supplemental Fig. S1.
Single-stranded oligonucleotides were purified by electrophoretic separation in gels containing 15% polyacrylamide/6 M urea. Oligonucleotides of the desired length were extracted from gel slices using an electroelution separation chamber (Elutrap, Schleicher & Schuell, Keene, NH).
Fluorescence Anisotropy Measurements-Fluorescence anisotropy measurements were carried out using a Photon Technologies International T-format fluorometer with a 75-watt arc lamp with automated large aperture Glan-Thompson polarizers on the excitation and both emission channels, using a 500-l micro cell from Starna Cells Inc. (Atascadero, CA), as described previously (15). Fluorescence anisotropy G-factors were measured directly for each experiment, according to standard procedures. In binding assays, T7 RNA polymerase was titrated into 2.5 nM TAMRA-labeled DNA (see supplemental Fig. S1) in fluorescence buffer: 30 mM Tris acetate, pH 7.8, 15 mM magnesium acetate, 25 mM potassium glutamate, 0.25 mM EDTA, and 0.05% Tween 20 (Calbiochem 10% protein grade; EMD Chemicals, Billerica, MA).The compartment was thermostatted at 25°C.
For each of the mutant and wild type enzymes, anisotropy data (A) from three independent titrations of enzyme (E t ) into 2.5 nM DNA (D t ) were combined in a fit to Equation 2, using the exact (quadratic) form of the binding function represented in Equation 1.
Each data point in Fig. 1A corresponds to the average of the three measurements at that enzyme concentration, with standard deviations as shown.
Steady-state Fluorescence Measurements-Steady-state fluorescence measurements were carried out in the above fluorometer in L-format mode and both emission and excitation monochromators, using a 50-l (3-mm light path) ultra-micro cell from Starna Cells. Excitation and emission slits were both set to 2 mm; fluorescence was excited at 320 nm, and emission was monitored at 370 nm. Samples in fluorescence buffer (above), thermostatted at 25°C, were stirred with a magnetic flea, and additions were achieved via direct injection with a Hamilton syringe, secured for reproducibility and lack of optical interference. Mixing times were estimated to be less than 5 s.
For the measurements in Limited Proteolysis Assays-Limited proteolysis assays were carried out as described previously, with some modifications (13). Briefly, 10 l of 10 M enzyme in fluorescence buffer was mixed with 10 l of 13 M DNA followed by the addition of 2 l of a mixture of NTPs (stock concentrations 16 mM each of ATP, GTP, and UTP) or of a similar stock of ATP, GTP, and UTP plus 4 mM 3Ј-dCTP (TriLink BioTechnologies, Inc.). After a 30-s reaction at 25°C, 2 l (0.1 g) of sequence grade trypsin (Promega) was added. The cleavage reaction was then stopped after 30 s by the addition of 6ϫ SDS-loading buffer with 2-mercaptoethanol. The samples were resolved on 9% SDS-polyacrylamide gels and stained with Coomassie Brilliant Blue.
Global Fitting of Kinetic Data-Data from Fig. 7, C-E, were globally fit to a single set of best-fit kinetic and fluorescence parameters for the wild type protein and then separately for the mutant protein. Differential equations describing the reduced kinetic model in Fig. 7A were numerically integrated (using an integration time of 0.02 s Ϫ1 ) and fit with the data analysis software pro Fit (QuantumSoft, Uetikon am See, Switzerland). Doubling the integration time produced no significant change in the best fit parameters. To allow computationally accessible fitting, fluorescence data from Fig. 7, C and E, were collected at higher frequency than shown, and data were grouped (locally averaged) to yield the smaller set of data shown. An example of this data reduction can be seen comparing the collected data in Fig. 1B with the reduction of the same data shown in Fig. 7E.

RESULTS
The current study aims to understand the energetic forces that underlie initial transcription and abortive cycling by RNA polymerases. Although simple notions of energetic stresses focus on the nucleic acid, for example, DNA scrunching or the inherent instability of short hybrids, mutation in the protein of proline to a leucine at position 266 (P266L), distant from both the hybrid and the promoter binding interface, has been discovered to decrease dramatically the abortive profile of T7 RNA polymerase on otherwise highly abortive sequences (8). How does the single point mutant P266L alter the energetics of initial transcription?
P266L and Wild Type T7 RNA Polymerase Have Similar Promoter Binding Affinities-The initial report of the P266L mutant employed a nuclease protection assay to assess binding and suggested that the mutant enzyme binds more weakly to the native promoter (8). Weakening of promoter binding would predict an earlier transition to elongation, and by some models, to stability.
We have now used a more sensitive fluorescence anisotropy method to measure the binding affinity of the mutant P266L to its promoter (15). The titrations presented in Fig. 1A use promoter DNA (see supplemental Fig. S1) that has a TAMRA fluorescent label at the 3Ј end of the template strand. The data reveal similar values for the dissociation constants (K d ) of the mutant (62 nM) versus wild type (56 nM) enzyme on DNA that is double-stranded in the promoter region (dsPromoter in supplemental Fig. S1). Given the scatter in the data, we do not consider these values to be significantly different, but in any case, the results clearly do not show the mutant binding more weakly to the promoter.
Previous characterization of the T7 consensus promoter using this approach has shown that a partially single-stranded (pss) DNA promoter has much stronger binding affinity to the wild type protein (15). The results in Fig. 1 demonstrate that both mutant and wild type enzyme bind (at least 10-fold) more tightly to the pss promoter, as expected, but not significantly different from each other, with K d values of 4.1 and 4.5 M, respectively.
Abortive release is an irreversible (beyond RNA lengths of 2) kinetic process, and so it is promoter release kinetics that are FIGURE 1. Wild type T7 RNA polymerase and mutant P266L have similar binding affinities and dissociation rates. A, fluorescence anisotropy assays measure binding to consensus promoter DNA that is double-stranded (dsPromoter) and partially single-stranded (pssPromoter) in the binding and initially melted regions (pssPromoter is essentially "premelted" and therefore is bound more tightly). Enzyme was titrated onto 2.5 nM DNA end-labeled with TAMRA at 25°C as described under "Experimental Procedures". In each case, data from three independent titrations of 2.5 nM DNA were combined in a fit to Equation 2, as described under "Experimental Procedures." Dashed lines indicate the 68% joint confidence intervals of the fitted parameters (shown as a range in parentheses for each value of K d ). B, promoter release monitored by fluorescence from 2-aminopurine at position Ϫ4 of the template strand. The addition of enzyme to promoter DNA leads to an equilibrium (steady-state) population of open complexes; the addition of an excess of a more tightly binding promoter sink traps enzyme, returning DNA to the lower double-stranded fluorescence. Data were fit to the integrated rate equations of the indicated kinetic model, holding k bind to previously determined levels and fitting k off promo , yielding very similar values for the dissociation rates of the wild type and mutant enzymes. most relevant. The results shown in Fig. 1B directly measure the rate of (double-stranded) promoter release for the mutant and wild type enzymes and again reveal no significant difference between the two.
If abortive dissociation necessarily involved promoter release, one would expect a sharp decrease in abortive fall off on the pssPromoter construct, which binds much more tightly and dissociates more slowly (16,17). As shown in Fig. 2, transcription from this construct shows behavior very similar to that observed with the double-stranded construct; wild type enzyme yields substantial abortive fall off, whereas the mutant produces much less. Together, these results argue against a model for reduced abortive cycling in which a proposed weak promoter binding affinity allows the mutant P266L to transition to elongation at shorter RNA (hybrid) lengths.
Kinetics of Initial Transcription-The complete kinetic scheme presented in Fig. 3 describes the possible outcomes available to an initially transcribing RNA polymerase. Recent studies have shown that the transition to elongation does not occur at a discrete RNA length, but rather begins with very low probability at position ϩ8, increasing in probability as the complex translocates to longer RNA lengths (18,19). To characterize kinetics at a specific position, we can halt transcription by leaving out a specific NTP from the mix. Thus, on a sequence encoding the transcript GGGAUAUACC⅐⅐⅐, we can halt transcription at position ϩ8 by supplying only GTP, ATP, and UTP. Intermediate incorporation events (k i , for i Ͻ 8) are much faster than events at the halt site, and in transcription through sequences that are known not to produce significant abortive products, we can neglect the intermediate k i,off init steps (verified by gel analysis in Fig. 5). Thus we can dramatically simplify the equation in Fig. 3, as shown in subsequent figures.
Finally, we can probe the state of the halted complexes by incorporating fluorescent base analogs in the nontemplate strand near the halt site (within the expected melted bubble). The DNA in that region will remain double-stranded, and hence the fluorescence will be low, until the complex approaches the halt site. Occupancy of the halt position will yield melted DNA, with increased fluorescence. The overall level of fluorescence at steady state will then reflect the relative population of the free DNA, the binary ED complex, both "dark," and the halted complex, which is "bright." Once steady state is achieved, we can then add an excess of a promoter sink (partially single-stranded promoter DNA encoding no RNA).  In this reasonably complete description of initial transcription, the lower path represents promoter-bound species, whereas the upper path represents promoter released species (vertical paths represent transitions from the former to the latter, promoter release). In the schematics, the pink element represents the rotating N-terminal platform, which undergoes a dramatic structural rearrangement on promoter release. Based on current understandings, less likely kinetic paths and species are illustrated in gray and by shorter arrows. An initially transcribing complex (ED⅐R i , lower manifold) has, in principle, three kinetic options: productive elongation (k i ), abortive release of RNA (k i,off init ), or release of promoter contacts and transition to the elongation configuration (k i trans ).
As the dynamics proceed, any enzymes that were originally promoter-free or that transiently revert back to a promoter-free state will be competitively (and efficiently) bound by the nonfluorescent promoter sink, added in 10-fold excess. The freed labeled DNA then reverts to the low fluorescent value characteristic of the double-stranded form. This expected behavior can be seen in the first panel of Fig.  4D (Halt at ϩ8). An expected distribution of initial free DNA and enzyme-DNA complexes (0 -60 s) shows low fluorescence as the probe at ϩ8 is expected to be in a fully duplex environment. The addition of GTP, ATP, and UTP at 60 s leads to a sharp increase in fluorescence as the complexes rapidly walk to the halt at position ϩ8 and establish a steady state. Subsequent addition (at 120 s) of an excess of competing promoter sink now sequesters free enzyme, leaving free (and therefore low fluorescence) DNA. As 8-mer RNA is released, the complex reverts to the binary promoter-bound (ED) state, allowing the downstream DNA to spontaneously collapse, with an associated decrease in fluorescence. At this point, the enzyme-DNA binary complex has two kinetic options; it can initiate transcription and rapidly reestablish the halted 8-mer complex, or it can dissociate. In the absence of competing promoter sink, the freed enzyme simply rebinds promoter and initiates another round of transcription, but in the presence of an excess of competing promoter sink, this is now disfavored. Over time, all of the enzyme is inhibited, and all of the DNA reverts to the low fluorescence state. Note that the native and P226L mutant protein behave the same in a halt at position ϩ8.
Halting transcription at position ϩ9 yields qualitatively different behavior for the native enzyme. The rapid rise in fluorescence following NTP addition is followed not by a stable steady state, but rather by a slower phase increase in fluorescence. This is explained by the kinetic model shown in Fig. 4B. We propose that as for the halt at position ϩ8, the system rapidly establishes a steady state, but then with time, halted complexes slowly transition to the long-lived elongation state, also characterized by high fluorescence. Thus, mass action eventually draws all of the enzyme and DNA into the elongation state. For the native enzyme halted at position ϩ9 for 60 s after the addition of NTPs, about half of the complexes have converted to the elongation state, and thus the remaining half are sensitive to the addition of the sink. Fluorescence does not return to the fully free DNA base line. Halting at positions ϩ10 and beyond yields a more complete transition, indicating faster rates of transition to elongation, as seen previously (20).
At position ϩ9, the P266L mutant, by contrast, shows little evidence of a transition to elongation. As a function of halt position (RNA length), the profile for the P266L mutant is very similar to that of wild type, but is shifted. Halting at positions ϩ8, ϩ9, and ϩ10 shows no evidence of a transition to elongation, but the transition is observed in complexes halted at position ϩ11 and beyond. In contrast to earlier proposals (19), the P266L mutant is transitioning to elongation later (at longer RNA lengths) than wild type.
The sequences used in this study (see supplemental Fig.  S1) were designed to yield relatively few abortives in the run up to the halt site, thus enabling the simple kinetic models presented in Fig. 4. To illustrate that complexes are halting as predicted and to confirm that these sequences yield relatively low amounts of abortive products for both wild type and mutant enzymes, we carried out parallel transcription reactions, shown in Fig. 5. With the exception that the halted P266L complexes turn over more slowly (are more stable), the profiles are the same for both mutant and wild type enzymes.
Limited Proteolysis Confirms a Delayed Transition to Elongation for the P266L Mutant-The slow phase increase in 2aminopurine fluorescence above has been modeled as a transi- tion from the initiation conformation to the more stable elongation conformation. Previous studies have exploited a proteolysis assay to follow the protein structural change associated with this transition (18,21,22). The proteolysis data presented in Fig. 6 complement the above fluorescence measurements and test the proposed assignment.
The protease-based assay is founded on the differential susceptibility of the initiation and elongation configurations of T7 RNA polymerase to trypsin. More specifically, residues 172-180 (and 95-99) are exposed to solvent and susceptible to trypsin in the initiation structure. Cleavage in the 172-180 region (with or without cleavage in the 95-99 segment) yields a readily detectable 80-kDa tryptic fragment (23). In the elongation configuration, the 172-180 region moves over 70 Å, refolds, and becomes a part of the RNA exit channel, protected from trypsin (2,3). Although residues 95-99 interact directly with promoter DNA in the promoter-bound initially transcribing complexes, they are exposed to trypsin in the promoter-free and elongation states (10). Thus trypsinolysis in the elongation phase yields a fragment of about 88 kDa.
In the limited proteolysis results shown in Fig. 6, for complexes halted at initially transcribing positions ϩ6, ϩ7 and ϩ8, both wild type and mutant enzymes show a 80-kDa band indicative of the initially transcribing phase. In complexes halted at position ϩ9, wild type enzyme reveals the appearance of a small amount of an 88-kDa band, indicating the onset of the transition to elongation, whereas the mutant shows no such evidence. At position ϩ10, wild type enzyme shows a more significant amount of the 88-kDa band, whereas the mutant reveals only a trace, indicating that wild type enzyme is transitioning significantly to elongation, whereas the mutant is only beginning the transition. Finally, at position ϩ11, both enzymes show similar amounts of the 88-kDa band.
The qualitative measurements in Fig. 6 provide independent evidence to support our interpretation of the data in Fig. 4 that the mutant P266L is delayed with respect to RNA length in its transition to elongation (it transitions to elongation at longer RNA lengths). In contrast to predictions of previous models, the mutant P266L does not release its promoter contacts more readily than wild type, but rather retains those contacts through longer RNA lengths than wild type.
More Complete Tests of the Kinetic Model-The fluorescence data of Fig. 4 are interpreted as providing kinetic signatures for halted complexes that are either stable or labile with respect to the probability of transitioning to elongation. In Fig. 7, we more fully explore the behavior of complexes halted at position ϩ9. The experiments presented in Fig. 7C mirror those of the ϩ9 halt data in Fig. 4, but in different experiments, inhibitory promoter sink is added at different times following the addition of NTPs. Thus, the transition to elongation is allowed to proceed to different extents prior to challenge with the inhibitor. As expected, longer incubations with NTPs show a leveling off of the slow phase fluorescence increase. Also, as predicted by the model, the inhibitor sink challenge leads to a decrease in fluorescence that decreases in overall magnitude with later additions of sink as more of the complexes transition to the stable elongation state.
In contrast, for complexes of the P266L mutant enzyme halted at position ϩ9, there is no slow phase increase in fluorescence following the fast rise to an initial steady state, and challenges with promoter sink lead to a complete return to the low fluorescence state. These results indicate clearly that at this position, the mutant does not transition to a conventional elongation state. Note that there is a slow decrease in fluorescence that can be fit with a kinetic step that leads to an unknown dark state. That this state may not convert readily to an active   Fig. 4) to allow halting of complexes at the indicated positions. After 1 min, trypsin was added and allowed to react for 30 s prior to quenching by the addition of an equal volume of SDS gel loading buffer. C, as evidenced by the appearance of the 88-kDa band, wild type protein (W) begins the transition to elongation at position ϩ9, whereas the P266L mutant (P) begins the transition at position ϩ10 to ϩ11. Controls show uncleaved protein and trypsin cleavage of the initially bound enzyme-DNA complex without added NTPs. OCTOBER 26, 2012 • VOLUME 287 • NUMBER 44 enzyme-DNA complex is suggested by the fact that transcription in the mutant does slow with time (panel D). Note that we observe this decay even for wild type protein halted at position ϩ8, suggesting that this minor behavior is not a unique property of the mutant.

Energetics of Initial Transcription
Global Kinetic Analysis of Transcription and Fluorescence Data-The fluorescence data of Fig. 7C report on the relative population of the halted complex, which defines the ratios of the underlying kinetic rate constants. To restrict the magnitudes of those rate constants, we have followed the time course of RNA production, adding these data to the global fit. To provide independent assays to support our model assignment and to help constrain global fitting, the kinetics of RNA (R 9 ) production was followed quantitatively as a function of time and is presented in Fig. 7D. Although RNA production levels off with time for the wild type enzyme, production levels off much less for the mutant. Leveling off of synthesis is consistent with a slow transition to a relatively stable elongation state that turns over very slowly, if at all. These data constrain the global fit below and are an essential complement to the fluorescence data.
The binary complex binding data in Fig. 1 clearly show no difference in the thermodynamics of promoter binding, but it is actually the kinetic dissociation constant k off promo that would be the key determinant in the mutant phenotype were promoter release altered in the mutant. Simple measurements of binary complex dissociation in the absence of NTPs (using the same sink challenge approach), shown in Fig. 7E and fit independently in Fig. 1B, confirm that the dissociation kinetics of the mutant are unchanged and at the same time place independent constraints on the rate of dissociation of the binary enzyme-DNA complex. For the native enzyme, the globally fit promoter dissociation rate of 0.04 s Ϫ1 and the independently fit value of 0.12 s Ϫ1 are similar to previous studies (15). Note that the former is impacted by other data in the global fit, whereas the latter reflects the data more directly. The data for the P266L mutant are indistinguishable from wild type, and the fit parameters fall also into this general range.
Combining the data from independent kinetic measurements in Fig. 7, C-E, into a single global fit provides a very powerful test of the kinetic model and assessment of the individual rates reported in Fig. 7A. The only parameter that was fixed in each fit is the promoter binding (association) rate (k bind ), assessed iteratively by knowledge of K d and k off promo . Phenomenologically, this parameter should affect the initial rise to steady state following the addition of NTPs and should be involved in the tradeoffs that establish the level of the initial steady state. The initial rise is not well resolved in these assays, and thus the data impose only a lower limit on this parameter.
In mutant and wild type fits, most of the resulting parameters are similar for wild type and mutant, indicating that the P266L mutation does not significantly impact the initial kinetic events (at least on this sequence). The parameter k cat , which represents the rate-limiting step(s) in the series of events starting at initial dinucleotide formation and ending at establishment of the halted complex, appears to be 2-fold reduced in the mutant relative to the wild type enzyme. It is unclear whether this difference is statistically significant, but it is not a large difference.

DISCUSSION
The genetically isolated mutation P266L in T7 RNA polymerase presents a valuable tool in understanding the energetics of initial transcription in this model system (8). In particular, the fact that this mutant shows dramatically reduced levels of abortive cycling provides an opportunity to understand this fundamental property of RNA polymerases specifically, and the mechanisms of initial transcription more generally.
Abortive Cycling Does Not Arise from Failure to Release the Promoter-Previous mechanistic proposals for abortive cycling have proposed that initially transcribing complexes release abortive transcripts because they retain promoter contacts at the expense of forward progression (24). The initial report that the P266L mutant might have weakened promoter binding was consistent with predictions of that model (8). Thus, the P266L mutant was proposed to release promoter contacts and transition to elongation earlier, escaping the promoter-retained abortive pathway shown in the lower half of Fig. 3.
The binding data presented in Fig. 1 clearly show that the mutant P266L has a promoter binding affinity that is similar to that of wild type T7 RNA polymerase. This is in contrast to earlier studies that used a less quantitative restriction enzyme protection assay (8). The latter may be complicated by the kinetics of nuclease digestion relative to the rates of polymerase binding and release. The titration of enzyme onto fluorescently labeled DNA provides a more direct thermodynamic measure of binding (15,25).
That binary complex stability does not directly contribute to abortive release during initial transcription is also evidenced by three other findings. First, weakening of promoter contacts has indeed been shown to lead to an earlier promoter release, yet does not generally yield reduced abortive cycling (18). Second, transcription from partially single-stranded promoter constructs, which bind polymerase with at least 10-fold greater affinity, yields abortive fall off percentages similar to transcription from fully double-stranded promoters (15,16,26). Similarly, the covalent cross-linking of promoter DNA to the protein would predict a decrease in abortive cycling, yet normal abortive profiles are observed (27). Third, because initial transcription is inherently controlled by kinetics, the dissociation rate is more relevant to the kinetic model. However, the data presented in Fig. 7D (and elaborated in Fig. 1B) show very similar rates of binary complex dissociation for the P266L mutant and wild type proteins. The above considerations rule out a model for abortive cycling in which the mutant P266L escapes abortive cycling by releasing the promoter and transitioning to elongation earlier than wild type, at shorter RNA lengths.
Qualitative and Quantitative Evidence of a Delayed Promoter Release in the Mutant-The placement of 2-aminopurine in the nontemplate strand at position ϩ8 provides a measure of the total population of halted complexes at steady state. The reduced kinetic model presented in Fig. 4A effectively describes the observed fluorescence changes of 2-aminopurine in complexes that are not transitioning to elongation. The population of complexes poised at position ϩ8 rapidly reaches a constant, steady-state level. The addition of promoter sink captures free enzyme in the cycle and rapidly draws all DNA, via mass action, to the low fluorescence, unbound state.
At position ϩ9, the wild type enzyme shows two different behaviors that are both consistent with a slow, but significant, transitioning to a stable elongation complex. First, after the rapid rise to an initial steady state, fluorescence continues to increase, consistent with the kinetic model in Fig. 4B. As halted complexes in the initially transcribing configuration convert to the stable elongation configuration, mass action draws free DNA into the steady-state cycle, increasing (more slowly) the ensemble fluorescence. Second, challenging that population with the promoter sink no longer returns fluorescence to the double-stranded base line, consistent with the proposal that a population has converted to a long-lived and stable elongation complex.
Global analysis of multiple data sets supports the kinetic assignments of the observed fluorescence changes. Including real time transcription data (accumulation of halted product) and separate promoter dissociation measurements with the above fluorescence changes allows the global fitting of all three data sets to a single set of kinetic parameters. Independent fits of wild type and mutant proteins yield similar values for promoter dissociation rates, for the rates of approach to the halt position, and for the rates of RNA release from the initially transcribing configuration. Only the nature of the transition to elongation at the halt position is different. These data provide strong support for the kinetic model presented in Figs. 4 and 7.
Most importantly, this approach defines separate kinetic signatures for complexes halted at positions at which the enzyme is or is not transitioning to elongation at significant rates. The halt position-dependent profiles reveal that the wild type enzyme begins transitioning to elongation at about position ϩ8. This behavior is consistent with recent single molecule measurements that show a very slow transition to elongation commencing at position ϩ8, and increasing for complexes halted at positions ϩ9 and ϩ10 (20). Importantly, these data show that the P266L mutant is not transitioning to elongation significantly at position ϩ9, but instead the onset of its transition is delayed to about position ϩ11.
Finally, an independent, biochemical assay provides still further support for the conclusions reached from the kinetic data. A wealth of structural data in this system predicts a conformational change in the protein that accompanies the transition to elongation. Previous studies provide a convenient biochemical assay for the known structural change in the protein that accompanies the transition to elongation (18, 21, 22). Fully consistent with the interpretation above, the proteolysis assays pre-sented in Fig. 6 show exactly the change predicted. In complexes halted at positions up to ϩ8, the data indicate a population primarily in the initially transcribing configuration (results from complexes halted at positions ϩ6 and ϩ7, not shown, are very similar). Limited proteolysis of wild type complexes halted at position ϩ9 (and beyond) show evidence for a population of enzyme that has converted to the elongation configuration, with the rate of accumulation of that population increasing with halt position.
Limited proteolysis data from the P266L mutant protein show a clear delay in the appearance of the proteolytic cleavage pattern characteristic of the elongation complex. The onset of the transition has shifted to longer RNA lengths, fully supporting conclusions from the fluorescence data.
Abortive Cycling Does Not Arise from Premature Release of the Promoter-The fact that the mutant that shows reduced abortive cycling is also delayed in promoter clearance might suggest an alternate model in which abortive RNA synthesis occurs when the RNA polymerase releases the promoter too soon. In this model, which would include elements of the scrunching model, premature collapse of the bubble would competitively displace an RNA-DNA hybrid that is too short to compete, releasing RNA via the upper, promoter-released pathway in Fig. 3.
However, the arguments against a role for delayed promoter release also rule out this model. Transcription from much stronger binding partially single-stranded promoters does not lead to the predicted decrease in abortive cycling (nor does covalent cross-linking of the promoter to the protein (27)). Moreover, transcription from constructs that are partially single-stranded in the promoter region, and so can provide no destabilizing bubble collapse, shows no significant reduction in abortive synthesis. We propose that the delayed promoter release correlates with, rather than directly drives, the reduction in abortive RNA release characteristic of the P266L mutant.
A Protein "Push-back" Model for Abortive Complex Instability-Crystal structures of complexes with ϩ3, ϩ7, and ϩ8 base RNA-DNA hybrids provide a clue to a new model for abortive instability (10, 28). It seems likely that growth of the RNA-DNA hybrid during initial transcription pushes against the N-terminal promoter binding platform (residues 80 -149 and 202-242), inducing a rotation that serves ultimately to weaken promoter interactions. This provides both a timing of, and the energetic driving force for, promoter release and the transition to elongation.
It is a necessary consequence that the pushing of the hybrid against the N-terminal platform is accompanied by an equal and opposite pushing of the platform against the hybrid. This provides an alternate proposal for abortive instability. As illustrated in Fig. 8, pushing of the N-terminal platform against the hybrid could induce backtracking (shortening the hybrid) or otherwise destabilize its structure, leading to instability of the bound RNA and abortive dissociation (but with retention of promoter contacts, as in our kinetic modeling).
The results presented here suggest that the P266L mutant is characterized by a positionally delayed barrier to rotation, leading to reduced stress in initial transcription and the observed delay in the transition to elongation. Reciprocally, the hybrid would also experience less destabilizing stress and allow passage through otherwise abortive transcribed sequences.
A positionally delayed or shifted barrier to rotation of the N-terminal platform is consistent with the observation that crystallization of initially transcribing halted intermediates at positions ϩ7 and ϩ8, which are significantly rotated, has been successful with the P266L mutant, but has not been reported with the wild type protein (28). A positionally delayed barrier to rotation is also consistent with the observation that the P266L mutant binds preassembled RNA-DNA elongation complex scaffolds faster than does wild type (29).
Why might the P266L mutation shift the barrier to rotation of the N-terminal promoter binding platform? This residue lies in a segment of the protein (residues 242-272) that connects the rotating N-terminal platform to the nonmobile C-terminal catalytic domain at the base of the O helix. This linkage extends away from the O helix as the RNA polymerase translocates to positions ϩ7 and ϩ8 (28). Mutation of Pro-266 may weaken interactions with the O helix that restrict this extension, allowing for an initially lower barrier to rotation of the N-terminal platform. Tests of this specific model are in progress.
Parallels with the Multisubunit RNA Polymerases-The coupling of hybrid growth to structural rearrangements leading to FIGURE 8. Hybrid push-back model for abortive cycling. Growth of the RNA-DNA hybrid (red-blue) against the N-terminal platform (pink) drives a rotation of the platform that ultimately weakens promoter binding, allowing promoter release and the transition to elongation. However, the platform necessarily "pushes back" on the hybrid, leading to (i) distortion of the hybrid structure and/or (ii) backtracking that could lead to abortive loss of the RNA transcript from the promoter-bound complex.
weakened promoter binding appears to be a common mechanism in RNA polymerases, providing both the necessary energetics and a timing (with respect to translocation/RNA length) mechanism for this key step in transcription. In the structurally unrelated bacterial RNA polymerase, the 3.2 linker region lies in the path of the nascent RNA-DNA hybrid, suggesting that growth of the hybrid serves to dislodge this specificity factor and trigger promoter release (30 -32). Intriguingly, single point mutants S506F and P504L in this linker lead to a dramatic reduction in abortive cycling (33,34), and it has been proposed that these mutations allow more facile (lower energy) displacement of the linker during initial transcription (30,31). It appears that the analog of this linker in eukaryotic RNA polymerase II is the B-finger, which also lies in the path of the hybrid and has similarly been associated with the transition to elongation (35,36). Consistent with our model, the reduction in steric clash between these elements and the hybrid should lead not only to a (positional) delay in promoter release (not yet tested), but reciprocally, to the observed decrease in abortive cycling.