Double-stranded DNA translocase activity of transcription factor TFIIH and the mechanism of RNA polymerase II open complex formation

TFIIH Has Ssl2-Dependent dsDNA Translocase Activity.

We first measured DNA helicase activity of purified yeast TFIIH using a substrate consisting of a 100-base oligonucleotide with a labeled 25-base oligonucleotide annealed at either the 3′ or 5′ end (Fig. 1A). Consistent with the function of human TFIIH, we found that yeast TFIIH contains 5′ → 3′ helicase but not 3′ → 5′ helicase function (Fig. 1B). The 5′ → 3′ helicase activity requires both hydrolysable ATP and the Rad3 ATPase, because a TFIIH derivative containing a mutation in the Rad3 Walker B motif (Rad3 E236Q) that is predicted to be defective in ATP hydrolysis (27) is defective in helicase function. In contrast, both WT and Rad3 mutant TFIIH preparations are equally active in supporting transcription with purified factors and Pol II from the yeast HIS4 promoter (Fig. S1).

As an initial test for dsDNA translocase function, we assayed whether TFIIH can displace a 22 nucleotide DNA triplex, a well-established assay for dsDNA translocase function (28–30). The DNA triplex was formed by annealing a polypyrimidine triplex-forming oligonucleotide (TFO) to a 142-bp dsDNA containing a complementary polypurine sequence at one end (Fig. 1A). The triplex is disrupted by heating, but is stable at 26 °C with or without the addition of ATP (Fig. 1C, lanes 1–3). Incubation of WT TFIIH, ATP, and the triplex substrate results in near complete dissociation of the TFO (Fig. 1C, lanes 4–6). The triplex displacement activity of TFIIH is unaffected by the Rad3 E236Q ATPase mutation but is abolished by the equivalent Walker B ATPase mutation in Ssl2 (E489Q; Fig. 1C, lanes 7–11). To assay whether TFIIH can simply bind and displace the triplex without translocation, we used a 22 nucleotide triplex substrate that did not contain additional dsDNA (Fig. 2A). The short triplex substrate is less stable than the longer triplex, likely due to the lack of stabilizing dsDNA. Nevertheless, as predicted for a dsDNA translocase, TFIIH cannot displace the TFO from this shorter substrate lacking dsDNA (Fig. 2B). The TFIIH triplex displacement activity was not dependent on a free DNA end, because TFIIH can also displace the TFO annealed to a 3.2-kb double-stranded plasmid, although with much slower kinetics (Fig. 2C). Both ATP and dATP function to promote OC formation (7) and, as expected, both nucleotides can promote triplex displacement (compare Fig. 1C and Fig. 2C, Right). Our combined results are consistent with the prediction that TFIIH contains an Ssl2-dependent dsDNA translocase activity.

Ssl2 Tracks in the 5′ → 3′ Direction Along the DNA Duplex.

dsDNA translocases are thought to track along one strand of the DNA backbone (31). If true for Ssl2, discerning the polarity of translocation will reveal which promoter strand is used by Ssl2 during DNA unwinding. As an initial test of polarity, we assayed triplex displacement using substrates where the TFO was annealed to either the top or bottom strand of the 142-bp duplex DNA (Fig. 3A). As shown above, TFIIH can readily displace the TFO from the bottom strand triplex substrate (Fig. 3B, lanes 4–6). In contrast, the top strand triplex substrate is almost completely resistant to TFO displacement (Fig. 3B, lanes 1–3). One model consistent with this result is that Ssl2 tracks along one strand of duplex DNA in the 5′ → 3′ direction and that this tracking is blocked by the annealed TFO. As a further test of this model, we introduced biotin as a tracking barrier on the triplex substrates (Fig. 3A). These substrates contain a single-strand DNA nick with biotin attached to the 5′ end of one strand of duplex DNA. We found that the top-strand biotin was a much stronger block to TFO displacement, with only one third of the TFO displaced after 6 h (Fig. 3 C and D). Although the bottom-strand biotin modestly inhibited TFO displacement, the TFO was >80% displaced after 6 h, similar to the nonbiotinylated template. As a further test of whether the integrity of the top strand is the most critical for TFO displacement, we generated triplex substrates with 5-bp single-stranded gaps on either the top or bottom duplex strand, directly adjacent to the triplex (Fig. S2). Consistent with the above results, we found that TFO displacement was most inhibited by the gap on the top strand. Under conditions of the assay, 80–90% of the TFO was displaced from the nongapped template, 45% displaced from the template with the gap on the bottom strand, and only 20–25% displaced from the template with the gap on the top strand. Our combined results strongly suggest that Ssl2 primarily tracks along one strand of duplex DNA in the 5′ → 3′ direction.

Blocking Ssl2 Translocation Inhibits Transcription Initiation.

The above results predict that blocking Ssl2 translocation should inhibit transcription initiation. Mapping the location of Ssl2/XPB-DNA binding in PICs has suggested Ssl2 binds 30–36 (yeast) and 40–50 bp (human) downstream from TATA (20–23). To test the role of Ssl2 translocase function in transcription initiation, we created three promoter derivatives with Cy3 dye inserted in the phosphodiester backbone of the nontemplate DNA strand 37, 41, or 46 bp downstream from the HIS4 TATA (Fig. 4A). Cy3 positioned in the DNA backbone is a strong inhibitor of Ssl2 translocation in the TFO displacement assay (Fig. S3).

The modified and unmodified DNA templates were tested for in vitro transcription activity using the reconstituted system (32), and transcription from these templates was completely TFIIH dependent (Fig. 4B, compare lanes 4–7 and 8–11). We found that Cy3 positioned on the nontemplate strand 41 and 46 bp from TATA inhibited transcription approximately fivefold (Fig. 4B, lanes 4, 6, and 7), consistent with the dsDNA translocase model. Cy3 positioned 37 bp downstream from TATA inhibited transcription approximately twofold (Fig. 4B, lanes 4–5). We speculate that transcription escaping this more upstream Cy3 insertion may be due to inherent flexibility in the Ssl2–DNA interaction, allowing Ssl2 to escape the Cy3 block by binding DNA just downstream of Cy3 37 in a position nearly equivalent to the observed XPB-human promoter interaction (23).

To confirm that the observed transcription inhibition by Cy3 could not be explained by an inhibition of transcription elongation, we created HIS4 promoter derivatives containing a 12 nucleotide single-strand bubble beginning 21 bp downstream from TATA (Fig. 4C). Pol II initiating from this bubble in the absence of any general factors transcribed past the Cy3 block with 47–77% efficiency compared with a bubble template lacking Cy3 (Fig. 4D, lanes 4–7). Taken together, our results show that the integrity of the nontemplate strand is important for transcription initiation and is consistent with the dsDNA translocase model for open complex formation.

Ssl2 Translocates with Low Processivity.

To further characterize the Ssl2 motor within the TFIIH complex, we next investigated stimulation of the ATPase by nucleic acid. Measurements of how the steady-state rate of ATP hydrolysis depends on DNA template length and concentration can be used to assay for translocation and to evaluate different models of translocation (33–35). For these measurements, we used the TFIIH preparation containing Rad3 E236Q so that Ssl2 was the only functional ATPase. TFIIH and DNA were preincubated for 40 min, and then ATP was added, and phosphate release was quantitated at different times (1–20 min). The steady-state rate of ATP hydrolysis was determined by a linear fit of these data (Fig. 5A). ATP hydrolysis in the absence of nucleic acid was undetectable, and dsDNA consistently stimulated ATPase activity four- to fivefold higher than that of single-stranded DNA (Table 1), consistent with the enzyme being a dsDNA translocase.

The steady-state rate of ATP hydrolysis was measured as a function of DNA concentration and DNA template length over a range of DNA lengths from 30 to 80 bp and on a circular plasmid representing an infinitely long template (Fig. 5 A and B). As the dependence of rate on DNA concentration follows a Michaelis–Menten curve, we were able to determine both a Michaelis constant (K_M) and a maximal velocity (V_max) for each length of DNA (Fig. 5B). Although K_M does not show a length dependence (Fig. 5C), V_max is clearly sensitive to DNA length (Fig. 5D). The dependence of V_max on template length is a clear indication that Ssl2 is a dsDNA translocase as nontranslocating enzymes do not show this behavior (33–35) (Fig. S4A).

Models of translocation can be used to fit the dependence of V_max on DNA length to extract quantitative values for parameters such as occluded site size, kinetic step size, and motor processivity, i.e., the probability that the motor takes a step forward instead of dissociating (Materials and Methods) (33–35). In the simplest model for translocation, the motor binds DNA at any site and steps forward coupled to ATP hydrolysis. At each site, the motor may either hydrolyze ATP and take a step forward or dissociate (Fig. S4B). At the end of the template, the motor either dissociates or translocates off the end of the DNA. This model predicts that K_M will vary with the length of the DNA template, whereas V_max will not (Fig. S4B). This behavior is inconsistent with our observed results (Fig. 5 C and D).

However, in contrast to the simple translocation model above, models that incorporate a slow kinetic step between unbound and the active DNA-bound motor states result in a length-dependent V_max and a length-independent K_M (33–35). Two limiting cases of this behavior are (i) a slow step after binding and before translocation or (ii) a slow step before dissociation from the end of the DNA (Fig. S4C). These two cases predict different pre–steady-state behavior in the ATPase rate (34). A slow step at the end of the DNA predicts a burst of rapid ATP hydrolysis on ATP addition compared with the steady-state rate. This burst phase arises because initial rounds of hydrolysis do not encounter the rate-limiting step that takes place at the DNA end. In contrast, a slow step only between DNA binding and translocation predicts a lag in ATP hydrolysis rate. This lag phase arises because the rate-limiting step occurs before translocation and before ATP hydrolysis occurs. Our data are consistent with the latter case, as there is a lag in ATP hydrolysis after adding ATP to TFIIH prebound to DNA (Fig. 5A). The observed lag suggests that Ssl2, in complex with TFIIH, first binds DNA in an inactive conformation and then, after ATP addition, isomerizes to the active state before translocation.

To extract quantitative measures of translocase parameters, we analyzed the DNA length dependence of the Ssl2 ATPase V_max with an expression derived from the isomerization model (SI Materials and Methods) (34). Assuming a translocation step size of 1 bp and a contact size on the DNA of 16 bp (Table S1 and Materials and Methods), we determined a processivity of Ssl2 in the context of TFIIH of 0.90 with a fit error of 0.01 corresponding to an average of 10.0 ± 0.9 bp translocated before dissociation from the DNA. Because we have no independent measure of step size, we note that a larger step size would lead to a lower processivity. For example, analyzing the same data with a step size of 2 bp would correspond to 5.3 ± 0.3 bp translocated before dissociation from the DNA.

DNA Helicase Assay.

TFIIH helicase activity was monitored using a fluorescent dye-labeled oligonucleotide (IR700-TTCACCAGTGAGACGGGCAACAGCC) annealed to PAGE-purified 100-base oligonucleotides, with the resulting templates having 5′ or 3′ overhangs of 75 bases (Fig. 1). The assay was performed as previously described (36), with the following modifications: 10-μL reactions contained 10 mM Hepes (pH 7.6), 100 mM potassium glutamate, 10 mM magnesium acetate, 3.5% (vol/vol) glycerol, 1 mM DTT, 1 μg BSA, 60 fmol holo-TFIIH, and 30 fmol template DNA. ATP or ATPγS was added to 1 mM, and reactions were incubated 1–2 h at 26 °C. Controls received TE buffer (10 mM Tris, pH 7.5, 1 mM EDTA) in lieu of ATP or were heated to 95 °C for 30 s. Reactions were quenched by the addition of 10 μL 33% glycerol, 40 mM EDTA, and 0.5% SDS and analyzed by PAGE using 10% acrylamide gels in 0.5× TBE buffer (134 mM Tris, 44 mM boric acid, 2.5 mM EDTA, pH 8.8) plus 0.1% SDS. Following electrophoresis, gels were visualized using an Odyssey IR scanner (LI-COR).

In Vitro Transcription Assay.

In vitro transcription using recombinant and purified factors was performed similarly to previously described assays (32). See SI Materials and Methods for additional information.

TFIIH Purification.

WT TFIIH was purified from strain SHY869 (RAD3-(HA)1-TAP tag, tfb6Δ) as previously described (32) except that, following the ultracentrifugation step, potassium acetate was added to the extract to a final concentration of 0.6 M before binding to IgG-Sepharose (GE Healthcare). TFIIH with ATPase-defective Rad3 was purified by the same method from strain SHY887 (leu2Δ rad3Δ::KanMX, tfb6Δ::HPH) carrying plasmid pJF82 [ars cen LEU2 RAD3 (E236Q)-(HA)1-TAP tag]. Because the Ssl2 E489Q mutation is lethal, this TFIIH derivative was purified from a WT strain containing the Tap-tagged Ssl2 mutation on a plasmid. Strain SHY861 (leu2Δ SSL2) carrying plasmid pJF62 [ars cen LEU2 ssl2 (E489Q)-(FLAG)1-TAP tag] was grown in glucose complete media lacking leucine, and TFIIH was purified by the method described above.

Triplex Disruption Assay.

Triplex DNA template formation and disruption reactions were performed as previously described (28), with the following modifications: templates were assembled from duplex DNA containing a triplex target sequence (AAGAAAAGAAAGAAGAAAGAAA) and a fluorescent or ³²P-labeled TFO (TTCTTTTCTTTCTTCTTTCTTT). The DNA and TFO were combined at 1 μM final concentration in 25 mM Mes (pH 5.5) and 10 mM MgCl₂ and heated to 57 °C for 15 min and then cooled at 1 °C/min over 35 min to allow annealing. Triplex DNA was stored at -20 °C and diluted in 50 mM Tris⋅HCl, pH 8.0, 10 mM MgCl₂, and 1 mM DTT before the assay. Ten-microliter reactions contained 10 mM Hepes (pH 7.6), 100 mM potassium glutamate, 10 mM magnesium acetate, 3.5% glycerol, 1 mM DTT, 1 μg BSA, 0.01% Nonidet P-40, 15 fmol holo-TFIIH, and 30 fmol triplex DNA. ATP or dATP was added to 1 mM, and reactions were incubated at 26 °C for the indicated times before stopping with 2.5 μL 5× GSMB (15% glucose, 3% SDS, 250 mM Mops, pH 5.5, and 0.04% bromophenol blue). The reactions were analyzed by PAGE using 6% acrylamide gels with buffer: 40 mM Tris-acetate (pH 5.5), 5 mM sodium acetate, and 1 mM magnesium chloride. Gels were visualized using either an Odyssey IR scanner (LI-COR) or dried and visualized by PhosphorImager (Molecular Dynamics).

ATPase Assay.

DNA-dependent ATPase activity of TFIIH (Rad3-E236Q) was measured using a colorimetric assay kit (Innova; 601-0120). Forty-microliter reactions contained 10 mM Hepes (pH 7.6), 100 mM potassium glutamate, 10 mM magnesium acetate, 3.5% glycerol, 1 mM DTT, 4 μg BSA, 40 fmol holo-TFIIH, and 1.25 nM to 1.5 μM template DNA. After 40 min at room temperature, purified ATP was added to 0.5 mM, and reactions were incubated 1–20 min at 26 °C. Reactions were stopped by the addition of 10 μL gold mix and, after 4 min, 4 μL stabilizer 2. After 30 min at room temperature, absorbance was measured at 635 nm and plotted against DNA concentration. A standard curve was established for every experiment using the kit-included phosphate standard and used to determine TFIIH catalyzed ATP hydrolysis. Templates from 30 to 80 bp were tested at nine concentrations in triplicate, spanning a 1,200-fold range of DNA concentration. See SI Materials and Methods for information on extracting translocation kinetic parameters from the ATPase data.