Human RNA Polymerase II Elongation in Slow Motion: Role of the TFIIF RAP74 α1 Helix in Nucleoside Triphosphate-Driven Translocation

Abstract

The role of the RAP74 α1 helix of transcription factor IIF (TFIIF) in stimulating elongation by human RNA polymerase II (RNAP II) was examined using millisecond-phase transient-state kinetics. RAP74 deletion mutants RAP74(1-227), which includes an intact α1 helix, and RAP74(1-158), in which the α1 helix is deleted, were compared. Analysis of TFIIF RAP74-RAP30 complexes carrying the RAP74(1-158) deletion reveals the role of the α1 helix because this mutant has indistinguishable activity compared to TFIIF 74(W164A), which carries a critical point mutation in α1. We report adequate two-bond kinetic simulations for the reaction in the presence of TFIIF 74(1-227) + TFIIS and TFIIF 74(1-158) + TFIIS. TFIIF 74(1-158) is defective because it fails to promote forward translocation. Deletion of the RAP74 α1 helix results in increased occupancy of the backtracking, cleavage, and restart pathways at a stall position, indicating reverse translocation of the elongation complex. During elongation, TFIIF 74(1-158) fails to support detectable nucleoside triphosphate (NTP)-driven translocation from a stall position and is notably defective in supporting bond completion (NTP-driven translocation coupled to pyrophosphate release) during the processive transition between bonds.

Transient-state (or pre-steady-state) kinetic analysis allows an enzyme reaction to be tracked through individual catalytic events (21, 22). Using rapid chemical quench-flow technology, our laboratory analyzed the formation of multiple specific phosphodiester bonds during elongation by human RNA polymerase II (RNAP II) (27, 28, 43, 44). The RNAP II reaction is tracked with millisecond precision, allowing characterization of all but the very highest rates of bond formation. These studies provide significant insight into active site isomerization, phosphodiester bond formation, translocation, processive elongation, transcriptional stalling, pausing, RNA cleavage, and restart from a shortened RNA 3′ end.

In recent studies of poliovirus RNA-dependent RNA polymerase, Cameron and colleagues (1, 2) demonstrated that EDTA and HCl quenched elongation at different reaction stages, when the reaction was run in the presence of Mn²⁺ as the divalent cation. This result indicated that the elongation complex (EC) isomerized to sequester two active-site Mn²⁺ atoms from EDTA chelation and that, after EDTA addition, the isomerized EC continued on the forward pathway to complete bond formation. HCl, by contrast, was expected to quench the reaction instantly, demonstrating the time of chemistry. The human RNAP II elongation mechanism can similarly be divided into stages, according to the time of Mg²⁺ sequestration (isomerization; EDTA quench) and the time of phosphodiester bond synthesis (chemistry; HCl quench) (43). Applying this double-quench approach, we obtain a detailed view of the human RNAP II mechanism.

Transcription factor IIF (TFIIF) is a heterodimer (or heterotetramer) of RAP74 and RAP30 subunits (4, 7, 8, 37, 38). In this work, we compared RNAP II elongation in the presence of TFIIF RAP74-RAP30 complexes in which the RAP74 subunit is truncated from the C terminus (Fig. 1). RAP74 is a 517-amino-acid protein. TFIIF 74 (1-227) has the C-terminal, DNA binding, winged-helix-turn-helix domain of RAP74 removed but includes an intact α1 helix (amino acids 157 to 183) (11, 26, 31). TFIIF 74(1-158) has the α1 helix removed. The α1 helix is very important but not absolutely essential for RNAP II initiation and elongation stimulation. TFIIF 74(1-158) supports the formation of preinitiation complexes containing TATA-binding protein (TBP), RNAP II, TFIIB, and promoter DNA under conditions that require TFIIF for assembly (26, 31). Critical point mutations in and around the α1-helix are found mostly in hydrophobic residues. Substitutions in the α1 region with the greatest effects on transcription include I155A, W164A, N172A, I176A, and M177A (10, 26, 31). The activities of these mutant proteins in initiation and elongation stimulation are very similar to those of α1 deletion mutants, including TFIIF 74(1-158). In this paper we compare the activities of TFIIF 74(W164A) and TFIIF 74(1-158) in transient-state elongation assays to confirm that TFIIF 74(1-158) behaves like a critical α1 point mutation in elongation. TFIIF stimulates elongation by RNAP II 5- to 10-fold in vitro (3, 10, 18, 26, 32, 37), and TFIIF is normally required during the initiation phase of the transcription cycle to bring RNAP II into a stable association with the promoter DNA (7, 8, 37).

FIG. 1.

TFIIF mutant proteins. Wild-type TFIIF, TFIIF 74(1-227), TFIIF 74(1-158), and TFIIF 74(W164A) are shown. The RAP30 interaction region, the α1 helix, and the winged-helix-turn-helix (W-HTH) regions of RAP74 are indicated.

TFIIS rescues stalled and arrested RNAP II ECs. In the presence of substrate nucleoside triphosphates (NTPs), TFIIS helps restart stalled elongation by inducing cleavage of the nascent RNA chain, primarily in dinucleotide increments. TFIIS can also stimulate the cleavage of larger RNA segments from previously backtracked and arrested ECs (15, 16, 19, 20). Our work, however, indicates that TFIIS has an activity that does not depend on RNA cleavage, because TFIIS also can suppress transient transcriptional pausing (43, 44). Functions of TFIIS were recently illuminated by an x-ray crystal structure of yeast RNAP II bound to TFIIS (23). In the present study, TFIIS was included in transcription elongation reactions because it improves the efficiency of RNAP II elongation from a stall position. TFIIS may stimulate RNAP II to backtrack, although this issue is not yet fully resolved. TFIIS supports backtracking, RNA cleavage, and restart pathways for RNAP II.

The mushroom toxin α-amanitin blocks translocation by RNAP II (14) and, as such, is a powerful probe for the human RNAP II translocation mechanism. In this work, α-amanitin was used to discriminate the pre-translocated EC from the post-translocated EC. In work to be published elsewhere, we have found that a combination of α-amanitin and a high concentration of the next templated NTP (i.e., 2.5 mM) can result in isomerization reversal, that is, release of a substrate NTP that was tightly held within the RNAP II active site (X. Q. Gong, C. Zhang, M. Feig, and Z. F. Burton, unpublished data). Observation of isomerization reversal demonstrates simultaneous binding of two templated NTP substrates: one in the RNAP II active site and one in the RNAP II main channel. Demonstration of isomerization reversal therefore proves the major tenet of the NTP-driven translocation mechanism: NTP substrates load to DNA template sites prior to translocation into the active site for chemistry. We use the isomerization reversal assay to compare RAP74 wild type, RAP74(1-227) and RAP74(1-158).

These studies show the importance of the RAP74 α1 helix in supporting NTP-driven translocation by RNAP II. When the next templated NTP substrate is withheld, RNAP II is forced to stall. In the presence of TFIIF 74(1-158), RNAP II fails to efficiently maintain the post-translocated EC at a stall position. Instead, RNAP II preferentially populates the pre-translocated EC and tends to fall into the backtracking, RNA cleavage, and restart pathways, supported by TFIIS.

MATERIALS AND METHODS

Cell culture, extracts, and proteins.

HeLa cells were purchased from the National Cell Culture Center (Minneapolis, Minn.). Extracts of HeLa cell nuclei were prepared as described previously (34). Recombinant human TFIIF and TFIIF mutant proteins were prepared by reconstitution of the RAP74-RAP30 complex, as described previously (40, 41). Recombinant human TFIIS was purified by phosphocellulose chromatography followed by MonoS chromatography (our unpublished procedure).

NTP stocks.

Ultrapure NTP sets were purchased from Amersham Pharmacia Inc. Based on our experience using these reagents, CTP and UTP stocks appear to be free of detectable ATP and GTP contamination. The ATP stock is substantially free of GTP contamination. The GTP preparation appears to be lightly contaminated with ATP, which does not complicate our experiment because GTP is the final addition to the reaction mixture, after the addition of ATP.

Preparation of RNAP II ECs.

Protocols for rapid quench-flow experiments have been published previously (10, 27, 28, 44). Briefly, ³²P-labeled C40 (40-nucleotide RNA ending in a 3′-CMP) RNAP II ECs were formed on metal bead-immobilized templates, by incubation in a HeLa transcription extract. Initiation was from the adenovirus major late promoter using a modified downstream sequence (+1-ACTCTCTTCCCCTTCTCTTTCCTTCTCTTCCCTCTCCTCC-+40). The purpose of the 39-nucleotide CT cassette is to synthesize C40 with ApC dinucleotide, dATP, [α-³²P]CTP, and UTP, bypassing the requirement for addition of ATP and GTP. C40 ECs were washed with 1% Sarkosyl-0.5 M KCl buffer to dissociate initiation, elongation, pausing, and termination factors, contributed by the HeLa extract. Moderate salt has been shown to be sufficient to dissociate TFIIF and TFIIS from RNAP II (35). ECs prepared in this manner are highly responsive to the addition of TFIIF and TFIIS, indicating that Sarkosyl and salt-washed ECs are substantially free of contaminating TFIIF and TFIIS. ECs prepared in this fashion are responsive to the addition of TFIIF 74(1-158) (10, 26), which has a very low but detectable stimulatory activity. After being washed, ECs are reequilibrated with transcription buffer containing 8 mM MgCl₂ and 20 μM each CTP and UTP. TFIIF (10 pmol per 15-μl reaction mixture) and TFIIS (3 pmol per reaction mixture) were added as indicated in particular protocols. In work to be published elsewhere (C. Zhang and Z. F. Burton, unpublished), we have determined that Sarkosyl-washed RNAP II ECs behave very similarly to ECs prepared by washing in 60 mM or 1 M KCl transcription buffer, supporting the use of Sarkosyl-washed ECs to analyze human RNAP II elongation.

Running-start, two-bond, double-quench protocol.

The running-start, two-bond protocol has been described previously (10, 27, 28, 44). We improve the procedure by reporting data from experiments in which the reactions were quenched with EDTA or with HCl (1, 2). Rapid-quench experiments were done using the Kintek Rapid Chemical Quench-Flow (RQF-3) instrument. All steps were done at 25°C. Elongation was through the sequence 40-CAAAGGCC-47 (G44 and G45 are underlined because we measure incorporation at these positions). Other DNA templates we have tested produce very similar elongation kinetics (data not shown). C40 ECs were incubated with 20 μM CTP and UTP. Then 100 μM ATP, 500 μM CTP, and 500 μM UTP were added to bead-immobilized C40 ECs on the bench top, to extend the EC to the A43 position. CTP and UTP were added to prevent backtracking and RNA cleavage for reaction mixtures containing TFIIS. During the next 30 to 120 s (depending on the protocol), ECs were loaded into the left sample port of the RQF-3 instrument. GTP (in transcription buffer) at twice its working concentration was loaded into the RQF-3 right sample port. Programmed, equal-volume mixing in the RQF-3 first combined ECs with GTP substrate with or without α-amanitin (0.5 mM), as indicated. The reactions were then quenched with 0.5 M EDTA or 1 M HCl after a precisely timed delay (≥0.002 s). For EDTA-quenched reactions, ECs on beads were collected with a magnetic particle separator and processed by electrophoresis, as described previously (10, 27, 28, 44). HCl quenching dissociates the transcript from the RNAP II EC. The products of HCl-quenched reactions were delivered into collection tubes containing a sufficient volume of 1 M KOH and 300 mM Tris base (∼70 μl) to neutralize the pH of the solution. Beads were removed using a magnetic particle separator. The solution was extracted with an equal volume of phenol-chloroform. The aqueous phase was adjusted to 0.3 M sodium acetate containing 20 μg of glycogen carrier and ethanol precipitated. After being vacuum dried, RNA samples were analyzed by electrophoresis. Gel bands were quantified using a Molecular Dynamics PhosphorImager. At the A43 stall point, the EC fractionates into multiple conformational states, which are revealed by their distinct elongation kinetics to G44. Formation of the G44 bond, therefore, provides detailed insight into mechanism and informs about recovery from the transcriptional stall at A43. The transition from G44 to G45 provides information about processive elongation.

Model-independent analysis.

Model-independent analysis (6, 9) was done with the program Microcal Origin version 7.0, fitting rate data to single, double, or triple exponential rate curves, as appropriate. A significant advantage of the human RNAP II elongation system is that these kinetic phases are so easily distinguished, giving a clear demonstration of distinct rates, which relate to distinct conformations of the RNAP II EC at the time of substrate GTP addition. Because transcripts are labeled at multiple CMP positions near the 5′ end and because the signal is determined as a ratio of bands within a single gel lane, there is little contribution of error due to gel sample loading, and rate determinations are highly consistent in independent experiments.

Global kinetic modeling.

Global kinetic modeling was done using the program DYNAFIT (25). Model-independent analysis was used to make the best initial estimates of individual rate constants in the mechanism. Models were constrained to be initially in thermodynamic balance, and thermodynamic balance was required around cycles. For reactions in the presence of TFIIF plus TFIIS, the RNA cleavage and restart pathway cannot be brought into thermodynamic balance with the pre-translocated EC, maintaining an adequate curve fit (A43a⇋A43/A41c is not in thermodynamic balance). A more detailed analysis of the backtracking, cleavage, and restart pathway has been published elsewhere (44). GTP dissociation rates were constrained to be no greater than 10,000 s⁻¹, although generally DYNAFIT would select higher rates. Estimated rate constants were selected so that the rate constants for G44 synthesis are as similar as possible to the rate constants for G45 synthesis. As far as we can determine, simpler models cannot describe the human RNAP II elongation mechanism (27).

RESULTS

One of the motivations for the present study was to observe human RNAP II elongation in slow motion to help elucidate the RNAP II mechanism for transcription. We knew that human RNAP II elongates slowly in the absence of stimulatory elongation factors or in the presence of defective factors, but we did not understand the mechanistic basis for slow RNA chain extension. We considered that a detailed kinetic comparison of RNAP II running through two bonds at higher and lower rates should reveal details of the elongation mechanism. In this paper, we show that this comparison gives insight into translocation, escape from a transcriptional stall, and coupling of NTP-driven translocation to pyrophosphate release.

Running-start, two-bond, double-quench protocol.

Our laboratory developed the running-start, two-bond, double-quench protocol for RNAP II kinetic analysis (43). The purpose of the running start is to obtain a consistent fraction of RNAP II ECs in an active state for forward elongation (27, 28, 43, 44). We found that short-duration (30- to 120-s) stalls at a defined template position yielded consistent preparations of RNAP II ECs for kinetic analyses. We made the judgment that at least two phosphodiester bonds must be tracked to obtain the most relevant elongation rate data. Initially, we did not know the status of RNAP II ECs at the stall position. We presumed that pyrophosphate would be released with a 30-s stall, but we could not initially guess what the translocation state of the EC might be or how the stalled EC might relate to ECs that arise during ongoing processive elongation. Formation of a second bond, however, must reflect processive elongation, including the translocation and pyrophosphate release steps. The double-quench method, that is, quenching reactions either with EDTA or with HCl, was adopted because of the demonstration by Cameron and colleagues that the RNA-dependent RNAP from poliovirus can be quenched at different stages of formation of each phosphodiester bond (1, 2). EDTA quenches after NTP loading into the active site followed by isomerization, and HCl quenches instantly, defining the time of phosphodiester bond synthesis. We were able to verify the utility of the double-quench method for human RNAP II (43).

The running start commences from the C40 position (40-nucleotide RNA ending in 3′-CMP), using bead-immobilized DNA templates. Elongation is through the sequence 40-CAAAGGCCUUU-50, concentrating on synthesis of G44 and G45 (underlined). CTP and UTP (500 μM each) are added to ECs to support extension to C40, in case of RNA cleavage from C40 to shorter positions. 100 μM ATP is added to extend the EC to A43, at which position the EC is stalled for 30 to 120 s, depending on the protocol. For reactions with no stimulatory factors added or in the presence of TFIIS as the sole stimulatory factor, the stall duration was 120 s. In all other cases, the stall duration was 30 s. During the stall, ECs were loaded into the left sample port of the KinTek RQF-3 Rapid Chemical Quench-Flow instrument. GTP was previously loaded into the right sample port. Rapid mixing combines the EC with GTP to begin G44 and G45 synthesis. Mixing with quench solution (EDTA or HCl) stops the reaction. The shortest start-stop time using the RQF-3 is 0.002 s.

TFIIF mutants.

Figure 2 illustrates the defect of TFIIF RAP74-RAP30 complexes containing RAP74 in which the α1 helix is deleted. TFIIF 74(1-158) supports an elongation activity that is very similar to the reaction with no stimulatory factor. In the presence of TFIIS, TFIIF 74(1-158) supports an elongation activity that is very similar to the reaction stimulated by TFIIS alone. TFIIF 74(1-158) stimulates much slower elongation than does TFIIF 74(1-227), which has an intact α1 helix and which is very comparable in activity to wild-type TFIIF, despite a large truncation of the RAP74 C terminus (deletion of amino acids 228 to 517). The defect of the α1 deletion mutant TFIIF 74(1-158) is obvious in formation of the G44 bond but appears to be even more severe in formation of the G45 bond (compare Fig. 2A and B), indicating a greater defect in processive synthesis than in escape from a long-term stall. To reenforce this initial impression that TFIIF 74(1-158) may have a defect in supporting processive synthesis, we plot the efficiency of the processive transition from G44 to G45 as a function of time (Fig. 2C and D). To do this, we calculate (G45+ all longer transcripts)/(G44+ all longer transcripts). Before G44 is synthesized, this quantity is not defined, and so for these times we assign G45+/G44+ the value of zero. Because G45 cannot be synthesized before G44, the G45+/G44+ ratio represents the efficiency of the G44-to-G45 processive transition and provides a more fair comparison between mutant and wild type than does the comparison shown in Fig. 2B, which is not normalized for available G44 substrate for G45 synthesis. TFIIF 74(1-158) shows a severe defect in the processive transition, which is comparable to elongation with no elongation factor or with TFIIS as the sole elongation factor (Fig. 2C). The defect of TFIIF 74(1-158) is even more evident at 10 μM GTP (Fig. 2D). Interestingly, at 10 μM GTP, TFIIF 74(1-227) supports RNAP II elongation with a slight delay in G45 synthesis relative to elongation supported by wild-type TFIIF (Fig. 2D).

FIG. 2.

Role for TFIIF in processive RNA synthesis. The reaction protocol is shown at the top of the figure. Reactions were quenched with EDTA. (A) G44 synthesis rates (G44+%: G44 plus all longer transcripts expressed as percentage of total ECs) for TFIIF, TFIIF 74(1-227), and TFIIF 74(1-158), with appropriate controls. Times are in seconds. (B) G45 synthesis rates. (C) Efficiency of the processive transition from G44 to G45 (G45+/G44+ %: before G44+ is detected, the function is assigned a value of zero) at 2.5 mM GTP. (D) Efficiency of the processive transition from G44 to > G45 at 10 μM GTP.

To emphasize that the elongation defect of TFIIF 74(1-158) is due primarily to the deletion of the α1 helix of RAP74, we compare elongation rates in the presence of TFIIF 74(1-158) and TFIIF 74(W164A), a critical α1 point mutant (26, 31). The comparison is shown for G44 synthesis at 2.5 mM and 10 μM GTP (Fig. 3), in the presence of TFIIS. Elongation rates are indistinguishable for the deletion and point mutant. We conclude that the defect of TFIIF 74(1-158) can be attributed to deletion of the critical α1 helix. This conclusion is supported by earlier work using steady-state RNAP II elongation and also initiation assays (10, 26, 31).

FIG. 3.

TFIIF 74(1-158) is indistinguishable from TFIIF 74(W164A) in elongation. (A) G44 synthesis rate at 2.5 mM GTP. (B) G44 synthesis rate at 10 μM GTP. The reaction protocol is as shown in Fig. 2. Reactions were quenched with EDTA.

Defects caused by α1 helix deletion.

A gallery of gel electrophoresis data demonstrating RNAP II elongation rates is shown in Fig. 4. Many interesting insights come from this qualitative comparison, some of which are discussed here and some of which are discussed below, to complement the quantitative analyses. RNAP II elongation in the presence of TFIIF 74(1-227) plus TFIIS or TFIIF 74(1-158) plus TFIIS is compared at 10, 100, and 2,500 μM GTP, using EDTA quenching (top three panels). HCl quench data are shown for elongation at 2,500 μM GTP (fourth panels from the top). Inhibition in the presence of α-amanitin is shown (2,500 μM GTP, EDTA quench) (fifth panels from the top). Sufficient α-amanitin (0.5 mM) is added to maximize RNAP II inhibition in the rapid-quench experiment (14). A high concentration of α-amanitin is necessary to obtain sufficiently rapid binding of the toxin to RNAP II. As we have previously shown, the post-translocated RNAP II EC is resistant to α-amanitin inhibition but the pre-translocated EC is sensitive, indicating that α-amanitin blocks translocation. Gel data are also shown for TFIIS alone and TFIIF 74(1-158) alone (2,500 μM GTP, EDTA quench) at the bottom of the figure, for comparison.

FIG. 4.

Gallery of gel electrophoresis elongation data with relevant comparisons. The reaction protocol is shown at the top of the figure. GTP concentrations are indicated (μM). Reactions were quenched with EDTA except where noted (HCl). α-Amanitin was added where noted (α-A). Times are in seconds. 0* indicates no ATP pulse, no GTP chase. 0 indicates no GTP chase.

TFIIS was included in these studies because it improves the efficiency of elongation from the initial C40 stall position (43, 44). Without the addition of TFIIS, C40 ECs are not efficiently extended in the presence of TFIIF 74(1-158) (lower right panel). TFIIS does, however, support the backtracking, RNA cleavage, and restart pathways of RNAP II. RNA cleavage occurs primarily in dinucleotide increments, but it appears that RNAP II and TFIIS may cooperate to cleave mononucleotides also, albeit at a lower frequency. At the C40 stall position, with addition of TFIIS, U38 is detected as a prominent cleavage product, indicating dinucleotide cleavage from C40. At the A43 stall position, dinucleotide cleavage is indicated by detection of A41. Particularly in the presence of TFIIF 74(1-158) (top five right panels), C39, C37, A42, and C40 RNA cleavage products are detected, indicating mononucleotide cleavage of the nascent RNA. Detection of A42, A41, C39, U38, and C37 is dependent on the presence of TFIIS in the reaction mixture. Because these positions are not stall positions, these bands most probably represent TFIIS-dependent RNA cleavage products. A42, A41, C40, U38, and C37 detection is much more evident in the presence of TFIIF 74(1-158) than in the presence of TFIIF 74(1-227), indicating cooperation between a highly active TFIIF and TFIIS to suppress the backtracking, RNA cleavage, and restart pathway. We are not certain about the extent to which higher detection of mononucleotide products in the presence of TFIIF 74(1-158) reflects loss of suppression of the backtracking, RNA cleavage, and restart pathways or selective blocking of formation of mononucleotide products by TFIIF 74(1-227), i.e., through the α1 helix supporting a conformation of the RNAP II EC that resists mononucleotide cleavage. Doing the kinetic analyses in the presence of TFIIS allows us to obtain consistent and efficient elongation in the presence of the defective TFIIF 74(1-158) mutant.

Model-independent analyses.

The core of the evaluation of complex kinetic reactions is the model-independent analysis (Fig. 5; summarized in Table 1). Model-independent analysis indicates the minimal complexity of the kinetic mechanism and the dominant rate constants of the mechanism. When the reaction was run at high substrate concentrations, observed reaction rates approached rate constants for key steps in the reaction mechanism. Three different assays were used, with side-by-side comparisons of the activities of TFIIF 74(1-227) and TFIIF 74(1-158). In the experiments shown, the GTP concentration was 2.5 mM to support G44 and G45 synthesis (40-CAAAGGCC-47 template). Reactions were quenched with EDTA to observe GTP loading and active-site isomerization (Fig. 5A to F). HCl quenching was done to demonstrate the time of phosphodiester bond synthesis (chemistry) (Fig. 5A and B and Fig. 5G and H). Note that longer timescales are shown for the slower reactions supported by TFIIF 74(1-158). In the experiment in Fig. 5I and J, α-amanitin was added to reaction mixtures along with GTP. The model-independent analyses are used to obtain the best initial estimates for rate constants to be applied in a global kinetic simulation (see Fig. 6 and 7) and to gain insight into the probable elongation mechanism.

FIG. 5.

The α1 helix of RAP74 is important for phosphodiester bond completion and NTP-driven translocation. Model independent analyses comparing TFIIF 74(1-227) (left panels A, C, E, G, and I) and TFIIF 74(1-158) (right panels B, D, F, H, and J). Note the differences in timescales (x-axes; times in seconds). Reactions containing α-amanitin (α-A) in panels I and J are indicated (open triangles).

TABLE 1.

Values for apparent reaction rates k_a, k_b, k_c, and k_d, and *A43 · GTP → A43 isomerization reversal (IR) are indicated for TFIIF, TFIIF 74(1-227), and TFIIF 74(1-158)

Factor	ka (s−1)	kb (s−1)a	kc (s−1)	kd (s−1)	IR (%)c
TFIIF + TFIISb	33 ± 4	1,450 ± 240	1.0 ± 0.2	27 ± 4	41
TFIIF74(1-227) + TFIIS	35 ± 6	1,300 ± 180	1.0 ± 0.1	31 ± 4	21
TFIIF74(1-158) + TFIIS	7.6 ± 0.9	760 ± 250	0.3 ± 0.04	7.6 ± 0.9	21

^ak_b rates are too high to be measured reliably with the RQF-3 at 25°C, and so apparent differences in k_b may not be meaningful (43).

^bFrom reference 43.

^cX. Gong, C. Zhang, M. Feig, and Z. Burton, unpublished data.

FIG. 6.

Global kinetic analyses of RNAP II elongation stimulated by TFIIF 74(1-227) and TFIIF 74(1-158). Kinetic data are shown for a range of GTP concentrations (μM). Solid red symbols indicate HCl quench data; open symbols indicate EDTA quench data. Residuals (differences between the kinetic simulation and the data points) are indicated beneath each graph. Data were fit to the kinetic models shown in Fig. 7B and C by using the program DYNAFIT (25).

FIG. 7.

Adequate kinetic models of RNAP II elongation. (A) Elongation in the presence of TFIIF and TFIIS. (B) Elongation in the presence of TFIIF 74(1-227) and TFIIS. (C) Elongation in the presence of TFIIF 74(1-158) and TFIIS. Black triangles indicate GTP-dependent steps, for which rate constants have units of micromolar⁻¹ seconds⁻¹. Other rate constants have units of seconds⁻¹. The percentage of total A43 (pretranslocated, posttranslocated, and paused), A42, A41, C40, and C39 RNAP II ECs, at the time of GTP addition are indicated within rectangles. Rate constants that are strongly indicated from model-independent analyses (Fig. 5; Table 1) are shown within ovals. Estimated K_ds for GTP binding are indicated within rectangles. Gray shading indicates the primary elongation pathways.

Qualitative comparison of the EDTA quench and HCl quench data reveals two distinct events in each bond addition cycle: (i) GTP loading into the active site followed by isomerization (EDTA quench) and (ii) phosphodiester bond synthesis (HCl quench). Isomerization and chemistry serve as signposts to track synthesis of the G44 and G45 bonds (43). Note that the EDTA and HCl quench curves have nearly converged within 0.2 s for the reaction stimulated by the TFIIF 74(1-227) deletion mutant. For the TFIIF 74(1-158) mutant, however, these curves have not converged within 0.5 s. This comparison shows that TFIIF 74(1-158) has a defect in bond completion, which we argue represents pyrophosphate release coupled to NTP-driven translocation, and indicates a defect of TFIIF 74(1-158) in translocation. Failure of the HCl quench curve to merge with the EDTA quench curve therefore indicates a defect in NTP-driven translocation slowing pyrophosphate release.

Fitting the EDTA quench data to exponential curves indicates the complexity of the RNAP II elongation mechanism. Fitting these data up to 0.1 or 0.5 s indicates two high rates of elongation (k_b and k_a) (Fig. 5C and D; Table 1). Rate data obtained in this time interval cannot adequately be fit to single exponential curves. This result indicates the presence of two distinct routes to NTP-Mg²⁺ sequestration through isomerization, which we have interpreted as elongation from two distinct translocation states of RNAP II, the post-translocated and pre-translocated ECs (14, 27, 28, 43, 44). k_b indicates the very high rate of isomerization from the initially post-translocated A43 EC. k_b is too fast to measure accurately using the RQF-3 instrument at 25°C, because the fast reaction is mostly complete in 0.002 s, the shortest reaction time of the instrument (43). k_a indicates the much lower rate of isomerization from the initially pre-translocated A43 EC. Fitting data to longer times (5 or 10 s) indicates an additional lower rate of elongation (k_c). This is the low rate of escape from the backtracking, RNA cleavage, and restart pathway. The distributions of A43 ECs in fractions a, b, and c are indicated in Fig. 5C to F. These distributions are very different for the reactions supported by TFIIF 74(1-227) and TFIIF 74(1-158). TFIIF 74(1-158) appears to have a defect in maintaining the post-translocated state (fraction b) of the A43 EC.

HCl quench data, on the other hand, can be fit to single-phase exponential curves for G44 synthesis in the 0.2- or 0.5-s time intervals (Fig. 5G and H). The rate of G44 chemistry is not distinguishable from k_a, which is the rate for isomerization of the initially pre-translocated A43 EC. For TFIIF 74(1-227), k_d probably provides a reasonable estimate of the rate of chemistry because the post-translocated fraction (fraction b) is dominant and the rate of isomerization (k_b) is high, making k_d rate limiting for G44 synthesis. For TFIIF 74(1-158), however, the apparent rate of chemistry (k_d) is probably limited by the prior low rate of translocation (k_a). Little A43 EC is found in the post-translocated state (fraction b), and pre-translocated ECs (fraction a) must translocate before chemistry, making the rate of translocation (k_a) rate limiting for the rate of chemistry (k_d). Once again, the primary defect of the TFIIF 74(1-158) mutant protein appears to be in supporting translocation.

Reactions in the presence of α-amanitin provide two types of information (Fig. 5I and J). First, α-amanitin discriminates the pre-translocated EC (fraction a) from the post-translocated EC (fraction b), because the pre-translocated EC is sensitive to α-amanitin inhibition but the post-translocated EC is resistant (14). Therefore, α-amanitin confirms the identification of the pre- and post-translocated A43 ECs (fractions a and b). Second, in work to be published elsewhere (Gong et al., unpublished), we have determined that the combination of high incoming NTP substrate (GTP to support G45 synthesis) and α-amanitin can induce isomerization reversal (IR) for a significant fraction of A43 ECs. IR is estimated as [b − (G44+ at 0.2 s)]/b or the fraction of *A43 · GTP ECs formed that are reversed/*A43 · GTP ECs formed (the asterisk indicates an isomerized A43 EC). Isomerization reversal converts *A43 · GTP → A43, an EDTA-sensitive form of the A43 EC. *A43 · GTP → A43 IR was determined as 41% for wild-type TFIIF (Gong et al., unpublished), 21% for TFIIF 74(1-227), and 21% for TFIIF 74(1-158) (Fig. 5I and J; Table 1). Apparently, both mutants have a defect in the isomerization reversal reaction. For the reaction supported by TFIIF 74(1-227) some transplantation occurred in the presence of α-amanitin because some G45 synthesis is detected (Fig. 4, left panel, fifth from top; Fig. 5I). For the reaction supported by TFIIF 74(1-158), no translocation was detected in the presence of α-amanitin (Fig. 4, right panel fifth from top: Fig. 5J), showing that TFIIF 74(1-158) has a defect in translocation that is rendered more severe by the strong α-amanitin block to translocation.

Global kinetic analyses.

The global kinetic analyses of RNAP II elongation in the presence of TFIIF 74(1-227) plus TFIIS and TFIIF 74(1-158) plus TFIIS are shown in Fig. 6 and 7. Rate curves at many GTP concentrations are shown in Fig. 6 for G44 and G45 synthesis (40-CAAAGGCC-47 template). The reaction mechanisms used for curve fitting are shown in Fig. 7B and C. Figure 7A is the previously published reaction mechanism for wild-type TFIIF + TFIIS, for comparison (43). In Fig. 6, open symbols indicate EDTA quench data and solid red symbols indicate HCI quench data. Residuals, shown beneath each set of rate curves, indicate the difference between experimental points and the kinetic simulation. From the quality of the data fit, we conclude that the simulations shown in Fig. 7 are adequate to describe the RNAP II elongation reaction. For the reaction stimulated by TFIIF 74(1-227), the reaction is highly GTP dependent for both G44 and G45 synthesis (Fig. 6A and B). For the reaction stimulated by TFIIF 74(1-158), the reaction is more highly GTP dependent for G45 synthesis than for G44 synthesis (Fig. 6C and D). Low rates of elongation (k_a) appear very similar at both high and low GTP concentrations for TFIIF 74(1-158), showing the low GTP dependence of k_a. For the most part, it is the amplitude of the fastest reaction phase (fraction b) that appears to decrease with decreasing GTP substrate, not the slower forward rate (k_a). The transition to processive RNA synthesis (G44 → G45), however, shows notable GTP dependence (Fig. 6D), indicating that bond completion (pyrophosphate release) is GTP dependent.

A defect in NTP-driven translocation.

With very minor adjustments, the simulation for RNAP II elongation in the presence of wild-type TFIIF plus TFIIS is the same as that in the presence of TFIIF 74(1-227) plus TFIIS (Fig. 7A and B). We conclude that TFIIF 74(1-227) has almost identical activity to wild-type TFIIF. Rate constants within ovals are those that were predicted from model-independent analyses (Fig. 5; Table 1). Only the rate constant for the chemical step appears slightly faster than the value predicted from model-independent analysis (43 versus 31 s⁻¹ [Fig. 5G; Table 1]). Other apparent rates measured at 2.5 mM GTP can be accommodated without modification as rate constants in the model. There are two slight differences between the models for TFIIF 74(1-227) and wild-type TFIIF. First, there is an indication of slight reversibility in the chemical step (*A43dGTP ⇋ *G44e · PPi). Second, the rate of NTP-driven translocation coupled to pyrophosphate release (*G44e · PPi · GTP → G44b · GTP) is slightly lower for the TFIIF 74(1-227) mutant compared with wild type (31 and 45 s⁻¹, respectively [Fig. 7A and B]). Reversibility of the chemical step and slower bond completion with TFIIF 74(1-227) appear to explain why the processive transition between bonds (G44 → G45) is slightly slower for TFIIF 74(1-227) than for wild-type TFIIF at low GTP concentration (Fig. 2D).

Unlike TFIIF 74(1-227), the mechanism for the RNAP II elongation reaction in the presence of TFIIF 74(1-158) shows dramatic defects. Most notably, evidence for NTP-driven translocation is lost from the A43 stall position, so that the slow NTP-independent pathway (A43a ⇋ A43b) is favored over the NTP-driven pathway (A43a ⇋ A43aGTP ⇋ A43bGTP), which predominates for wild-type-TFIIF- and TFIIF 74(1-227)-stimulated reactions. Qualitatively, the loss of NTP-driven translocation is indicated by the low GTP dependence of low elongation rates (k_a) for G44 synthesis with the TFIIF 74(1-158) mutant protein (Fig. 6C). At the processive transition from G44 → G45, however, the transition must be GTP dependent (Fig. 6D). We think that processive elongation (G44 → G45) differs from escape from a stall (A43 → G44) because processive synthesis requires displacement of pyrophosphate from a tightened RNAP II active site (43), which is likely to depend on a conformational change of the RNAP II EC. At the A43 stall position, pyrophosphate release is likely to be complete after a 30-s stall. During processive synthesis, pyrophosphate release is coupled to NTP-driven translocation, resulting in the slow GTP-dependent transition from G44 to G45 with the TFIIF 74(1-158) mutant protein (Fig. 6D). The 4.6 s⁻¹ rate constant for the NTP-driven translocation step (*G44e · PP_i · GTP → G44b · GTP) is slower than the 7.5 s⁻¹ rate constant for the NTP-independent translocation step (A43a → A43b), indicating why NTP-driven translocation is not detected from the A43 stall position. If NTP-dependent translocation occurs from the A43 stall position with the TFIIF 74(1-158) mutant protein, it is too slow to be distinguished from NTP-independent translocation. We suggest that this difference between elongation from the A43 stall and processive synthesis from G44 to G45 results from the prior pyrophosphate release at A43, due to the 30-s stall at this position. As indicated from model-independent analyses, TFIIF 74(1-158) has a defect in supporting translocation, particularly NTP-driven translocation. G44 → G45 processive RNA synthesis differs from escape from a transcriptional stall at A43. Processive elongation (G44 → G45) is more highly GTP dependent and appears to be GTP driven. In Fig. 7C, this feature of the kinetics is reflected in the 4.6 s⁻¹ rate constant for the *G44e · PP_i · GTP → G44b · GTP step, which couples pyrophosphate release to GTP-driven translocation. Every indication shows that TFIIF 74(1-158) has a defect in translocation, NTP-driven translocation, and bond completion, which is pyrophosphate release coupled to NTP-driven translocation.

DISCUSSION

NTP-driven translocation coupled to pyrophosphate release.

Studies with the TFIIF 74(1-158) mutant protein support the NTP-driven translocation mechanism, previously proposed by our laboratory. In particular, the NTP-driven translocation model proves robust to describe the severe defects of TFIIF 74(1-158) in supporting translocation, NTP-driven translocation, and pyrophosphate release coupled to NTP-driven translocation. TFIIF 74(1-158) shows a defect in translocation at the A43 stall position. In the presence of the mutant factor, the pre-translocated RNAP II EC is more fully occupied than the post-translocated EC. The opposite relation holds for elongation stimulated by wild-type TFIIF and TFIIF 74(1-227), both of which strongly promote forward translocation. TFIIF 74(1-158) shows multiple defects in supporting NTP-driven translocation. From the A43 stall position, TFIIF 74(1-158) supports an EC that appears incapable of GTP-driven translocation, and the GTP-driven pathway disappears from the mechanism, presumably because GTP-driven translocation is too slow to be detected in competition with the GTP-independent translocation rate. During the processive transition between bonds (G44 → G45), however, TFIIF 74(1-158) supports an EC that is very slow and inefficient at bond completion. This transition is severely restricted at low substrate GTP concentration. We conclude that, at the A43 stall position, the GTP-independent pathway is favored because pyrophosphate was previously released during the 30-s transcriptional stall. During the processive transition from G44 to G45, however, pyrophosphate release is obligatorily coupled to GTP-driven translocation, forcing the EC through a highly GTP-dependent bond completion step, which can be traversed only with difficulty by the impaired RNAP II EC. So, for elongation in the presence of the TFIIF 74(1-158) factor, escape from a stall shows less GTP dependence than the processive transition between bonds. We posit that processive RNA synthesis is more highly dependent on NTP-driven translocation than is escape from a stall, because processive bond completion couples a conformational change of the RNAP II EC, required for pyrophosphate release, to NTP-driven translocation. TFIIF 74(1-158) fails to stimulate NTP-driven translocation, indicating that the α1 helix of human RAP74, which is deleted in this mutant protein, plays a key role in supporting the NTP-driven translocation mechanism for human RNAP II.

NTP loading by RNAP II.

It has been suggested that the secondary pore, an extensive, narrow channel, is the sole route for NTP loading into the RNAP II active site (3, 5, 12, 38, 41, 44). Recently, Kornberg and colleagues presented a detailed model for secondary pore substrate loading, with a final fidelity check regulated by a novel base-flipping mechanism (42). Kinetic analyses of RNAP II elongation, however, including this paper, are inconsistent with the secondary pore being the only, or even the major, route for NTP loading into the RNAP II active site. We posit that the main RNAP II channel is the primary route for NTP loading, not the secondary pore (27, 43). Recent work from our laboratory clearly demonstrates that at least two template-specified NTP substrates, one in the active site and one in the main channel “allosteric” site, pair to DNA simultaneously (Gong et al., unpublished). Isomerization reversal (Fig. 5I and J) indicates simultaneous templated occupation by GTP of the RNAP II active site and the main-channel “allosteric” site. Work from the Erie laboratory also identifies the main enzyme channel of E. coli RNAP as an “allosteric” NTP-binding site (17).

X-ray crystal structures of the yeast RNAP II EC present contrary views of the extent of template opening at the downstream edge of the transcription bubble (13, 24, 42). Opening of the downstream bubble is a critical issue for the NTP-driven translocation model, because NTP-driven translocation requires a minimum of one unpaired DNA template base in the main channel. Two EC structures from the Kornberg laboratory indicate that at least three single-stranded DNA template sites (n + 2, n + 3, n + 4, where n is the RNA length, for the post-translocated EC) are available for NTP base pairing at the downstream edge of the transcription bubble (13, 42). A fourth DNA template base (n + 5) is partially unpaired in these structures. By contrast, an EC structure from the Cramer laboratory (24) shows that the downstream transcription bubble can collapse to pair at the n + 3 and n + 4 template bases, although the DNA helix remains strained at these positions, indicating that a change in the EC conformation could separate these bases, as shown in the Kornberg structures (13, 42). In the present work, we show that TFIIF can shift the human RNAP II EC from a conformation that supports NTP-driven translocation [TFIIF wt or TFIIF 74(1-227)] to a conformation that is defective for NTP driven translocation [TFIIF 74(1-158)]. The Kornberg and Cramer structures provide a potential structural explanation for our observations. Perhaps the Kornberg structures, with a more open downstream transcription bubble, represent the EC conformation supported by TFIIF (capable of NTP-driven translocation) and the Cramer structure represents the EC conformation in the presence of the defective TFIIF 74(1-158) mutant (defective for NTP-driven translocation). The yeast RNAP II EC structures are not crystallized in the presence of stimulatory elongation factors and therefore might in some cases revert to lower activity conformations of the EC.

NTP-driven translocation is consistent with single-stranded DNA template bases in the main enzyme channel being involved in NTP loading into the RNAP II active site. These single-stranded DNA template bases cannot be accounted for by models that exclude NTP loading through the main RNAP II channel. We conclude that, during processive RNA synthesis, NTPs load through the main enzyme channel, not the secondary pore. NTPs load through the secondary pore only to a stalled and post-translocated EC. We would argue, therefore, that the NTP loading structures (“A” site, “E” site, and “pre-insertion” site) described recently by Kornberg, Sosunov (36), and Cramer represent less favored NTP-loading routes, not processive NTP loading. Some of these EC NTP structures (i.e., “E” site and “pre-insertion” site) may represent mechanisms for excluding misloaded NTPs from the active site. According to our view, the main functions of the secondary pore are in excretion of pyrophosphate after chemistry and expulsion of inappropriate NTP substrates from the RNAP II active site.

Backtracking and pausing.

In the presence of TFIIF 74(1-158), RNAP II overpopulates the backtracking, cleavage, and restart pathways and shows an increased tendency to support TFIIS-mediated mononucleotide RNA cleavage. Normally, TFIIS supports RNA cleavage in dinucleotide increments. As we have previously argued (44), RNA cleavage in dinucleotide increments may be favored over mononucleotide cleavage, because, if cleavage occurs in mononucleotide increments, the pre-translocated RNAP II EC becomes sensitive to cleavage and the normal elongation pathway merges with the RNA cleavage pathway. Cleavage of RNA in dinucleotide increments separates normal elongation from the backtracking and RNA cleavage pathways. Here we show, however, that in the presence of TFIIF 74(1-158) or in the absence of factors other than TFIIS, mononucleotide products can be discerned. Dinucleotide cleavage is favored over mononucleotide cleavage, but transcript patterns provide evidence for both in the presence of TFIIF 74(1-158). It appears that wild-type TFIIF helps to specify dinucleotide cleavage by TFIIS, thus supporting separate elongation and RNA cleavage pathways.

Others have suggested that a paused RNAP II EC might be backtracked by 1 nucleotide relative to the pre-translocated EC (30). Our laboratory had previously argued against backtracking by as much as a full nucleotide during transient pausing because backtracking by 1 nucleotide from the pre-translocated state should poise the EC for TFIIS-mediated dinucleotide cleavage of the nascent transcript. Paused ECs, however, are not necessarily sensitive to TFIIS-mediated dinucleotide cleavage, indicating that pausing may not require backtracking by a full nucleotide (44). In light of our present results, however, we concede that this issue must receive more attention. TFIIF 74(1-158) has a defect in translocation, and in the presence of TFIIF 74(1-158), more backtracking and RNA cleavage is observed than in the presence of TFIIF 74(1-227) or wild-type TFIIF. This result is consistent with involvement of limited backtracking, which is reverse translocation, in transcriptional pausing. We do not have a current resolution to the question of how much reverse translocation is associated with transient pausing. Single-molecule studies of E. coli RNAP elongation from the Block laboratory indicate that transient pausing can occur without substantial backtracking (29, 33).