A minimal set of RNA polymerase II transcription protein interactions.

All pairwise interactions of RNA polymerase II and general transcription factors (TF) IIB, E, F, and H have been quantitated by surface plasmon resonance with the use of a Ni2+ chelate on the sensor surface where necessary to attain higher sensitivity. Only 4 of 10 possible interactions were found above the detection limit: TFIIB, -E, and -F binding to RNA polymerase II and TFIIE binding to TFIIH. These four interactions constitute a minimal set for the formation of a transcription initiation complex and may represent the primary interactions involved in assembly of the complex. Point mutations in TFIIB that alter the location of transcription start sites in vivo markedly diminished the affinity of TFIIB binding to RNA polymerase II. Protein blotting revealed an interaction between the largest subunit of TFIIE and third largest subunit of TFIIH, which may be responsible for TFIIE binding to TFIIH.

All pairwise interactions of RNA polymerase II and general transcription factors (TF) IIB, E, F, and H have been quantitated by surface plasmon resonance with the use of a Ni 2؉ chelate on the sensor surface where necessary to attain higher sensitivity. Only 4 of 10 possible interactions were found above the detection limit: TFIIB, -E, and -F binding to RNA polymerase II and TFIIE binding to TFIIH. These four interactions constitute a minimal set for the formation of a transcription initiation complex and may represent the primary interactions involved in assembly of the complex. Point mutations in TFIIB that alter the location of transcription start sites in vivo markedly diminished the affinity of TFIIB binding to RNA polymerase II. Protein blotting revealed an interaction between the largest subunit of TFIIE and third largest subunit of TFIIH, which may be responsible for TFIIE binding to TFIIH.
An RNA polymerase II transcription system can be reconstituted from the polymerase and five general initiation transcription factors (TF) 1 IIB, D, E, F, and H (1). These proteins have been resolved to homogeneity from yeast and mammalian cells, and genes for all 28 polypeptides of the minimal system have been cloned from yeast (2)(3)(4)(5)(6)(7)(8). As the proteins assemble in a large complex at a promoter prior to the initiation of transcription, the analysis of interactions among them should be informative about the initiation mechanism. Interactions of purified transcription proteins have been investigated in the past by sedimentation, gel electrophoretic, and affinity chromatographic approaches (3, 6, 9 -11). Many pairwise interactions among the polymerase and general transcription factors have been observed, but the results have remained largely qualitative and incomplete.
We have extended previous analyses to the quantitative determination of pairwise interactions among all transcription proteins except TFIID. Our findings distinguish between strong and comparatively weak interactions and reveal a simple pattern of protein-protein contacts in the transcription initiation complex. This pattern is consistent with previous func-tional evidence (12) and with an ordered pathway of initiation complex assembly derived by others (1,(13)(14)(15).
RNA polymerase II was purified as described previously with the addition of 10 M ZnCl 2 to all buffers (18). C-terminal domain (CTD)less RNA polymerase II was generated as described (16).
For analysis by gel filtration, 10 l of 0.1 mg/ml TFIIE was injected onto a Bio-Sil Sec 400 300 ϫ 7.5 mm column (Bio-Rad) at 0.2 ml/min in running buffer (40 mM Hepes, pH 7.6, 7.5 mM MgCl 2 , 120 mM potassium acetate, and 0.005% Surfactant P-20). The apparent molecular mass was 105 kDa on the basis of comparison with gel filtration standards (Bio-Rad).
TFIIB, TFIIF, and TFIIH-Recombinant TFIIB was prepared as * This work was supported in part by National Institutes of Health Grants GM36659 and AI21144 (to R. D. K.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ** To whom correspondence should be addressed. Tel.: 415-723-6988; Fax: 415-723-8464. 1 The abbreviations used are: TF, transcription factor(s); CTD, Cterminal domain; TBP, TATA-binding protein.
Surface Plasmon Resonance-Surface plasmon resonance measurements of protein-protein interactions were performed with the BIAcore Biosensor (Pharmacia Biotech Inc.). All measurements were at room temperature in running buffer (40 mM Hepes, pH 7.6, 7.5 mM MgCl 2 , 120 mM potassium acetate, 0.005% Surfactant P-20) with a flow rate of 15 l/min. Proteins were immobilized on CM5 research grade sensor chips with the use of an amine coupling kit (Pharmacia) (23,24). Polymerase was immobilized in 100 mM sodium acetate, pH 5.0; TFIIE was immobilized in 100 mM sodium acetate, pH 5.0; TFIIB was immobilized in 100 mM sodium acetate, pH 6.0; and TATA-binding protein (TBP) was immobilized in running buffer. Blank surfaces were prepared by the introduction of buffer without protein. Noncovalently bound proteins were removed by washing with running buffer containing 1 M potassium acetate. All proteins were diluted in running buffer from concentrated stock solutions prior to use.
A Ni 2ϩ chelate sensor chip was prepared as described (25). For immobilization of RNA polymerase II and TFIIE, the hexahistidinetagged proteins (50 g/ml) were passed over the surface at 1 l/min in running buffer containing 2 mM imidazole. All binding measurements were performed at 15 l/min in the same buffer. Data were analyzed with BIAevaluation, Version 2.1, using the AB ϭ A ϩ B and A ϩ B ϭ AB models for dissociation and association (Pharmacia).

RESULTS
Protein-protein interactions were quantitated by surface plasmon resonance, which detects the change in refractive index near a surface bearing one adsorbed protein due to interaction with another protein. The surface is typically coated with a dextran layer, adsorbing the first protein nonspecifically and therefore in a variety of orientations. The refractive index change on binding of the second protein depends on its size and proximity to the surface; and for a single site, as in the cases studied here, the change is proportional to the amount of bound protein. The refractive index change, expressed in resonance units, may be recorded as a function of time after the addition of a solution of the second protein or following the removal of this solution for determination of rates of association and dissociation, respectively. These rates are related through simple exponentials to the fundamental rate constants k a and k d for the binding reaction, with k a /k d ϭ K a , the equilibrium association constant. A test of validity of the analysis is the linearity of ln(R 0 /R) versus time for the dissociation reaction, where R is the detector response and R 0 is its value at the start of dissociation. In most cases studied here, linearity was observed, except for deviations due possibly to bulk refractive index changes during the first 30 s of association reactions and more persistent deviation for RNA polymerase II-TFIIE interaction discussed below (Fig. 1).

RNA Polymerase II and General Transcription Factor
Interactions-Yeast RNA polymerase II was coupled to the dextran surface of a BIAcore sensor chip by cross-linking between protein amino groups and carboxymethylated dextran with the use of N-hydroxysuccinimide and N-ethyl-NЈ-(dimethylaminopropyl)carbodiimide as described. Yeast general transcription fac- tors tested for interaction with RNA polymerase II exhibited little, if any, nonspecific binding to the dextran surface in the absence of polymerase with the exception of TBP, the source of transcription factor IID activity in a minimal transcription system. TBP bound so strongly to a blank dextran surface that no measurements of its interaction with other bound proteins could be taken.
RNA polymerase II-TFIIB interaction, previously demonstrated by affinity chromatography (3), was readily measurable, although the standard deviation of multiple determinations of k a was comparable to k a (Fig. 2). This high standard deviation could be due to batch to batch variation in TFIIB preparations or possible aggregation of TFIIB. The specificity of the interaction was shown by the effects of point mutations in TFIIB. Two point mutants, sua7-1 (E62K mutation) and sua7-3 (R78C mutation), isolated from a screen for altered locations of transcription start sites in vivo (21) were expressed in recombinant form and found to be inactive in transcription reconstituted with other pure transcription proteins in vitro. With the possibility in mind that the inactivity of the mutants might be due to diminished affinity for RNA polymerase II, the recombinant proteins were tested by surface plasmon resonance. Neither mutant protein proved to interact with the polymerase to a significant extent ( Fig. 3; data not shown).
RNA polymerase II-TFIIE interaction was of particular interest since previous investigation of the yeast proteins by glycerol gradient sedimentation failed to detect a polymerase-TFIIE complex, whereas human TFIIE was reported to cosediment with polymerase (26,27) and to bind preferentially to polymerase with a nonphosphorylated CTD in co-immunoprecipitation experiments (9). Yeast RNA polymerase II-TFIIE interaction was readily detectable by surface plasmon resonance. A ln(R 0 /R) plot for the dissociation reaction was nonlinear ( Fig. 1), with a variation in slope corresponding to a 3-fold range of k d values. This variation could not be accounted for by the presence of multiple TFIIE-binding sites but might reflect conformational mobility, heterogeneity, or some other aspect of the proteins under study (see "Discussion"). In view of the dependence of human TFIIE binding on the phosphorylation state of the polymerase CTD, we tested yeast TFIIE binding to polymerase lacking a CTD, as well as to polymerase lacking subunits 4 and 7, and obtained apparent K a values of 1.4 ϫ 10 7 and 7.4 ϫ 10 6 M Ϫ1 , respectively, which differs little from K a for the native polymerase (Fig. 4). TFIIE binding to CTD-less polymerase and ⌬ 4/7 polymerase showed the same variance in dissociation as TFIIE binding to normal RNA polymerase II (data not shown). We did note an influence of the nature of the salt upon RNA polymerase II-TFIIE interaction, with potassium acetate most favorable, followed by ammonium sulfate and sodium chloride (Fig. 5). It may be significant that the same order of preference for these salts has been found in transcription.
For the studies of TFIIE interaction reported here, recombinant protein was isolated from a bacterial strain co-expressing the large (Tfa1) and small (Tfa2) subunits. A mass of 105 kDa was estimated by gel filtration. In order to determine the affinities of the individual subunits for RNA polymerase II, they were separately expressed and purified. Tfa1 bound native, CTD-less, and subunits 4 and 7-deficient polymerases with K a values of 7.8 ϫ 10 6 , 1.1 ϫ 10 7 , and 1.1 ϫ 10 7 M Ϫ1 , respectively, which are comparable to the affinities measured for intact TFIIE. Tfa2 exhibited no appreciable binding at all.
RNA polymerase II-TFIIF interaction, well documented in many previous studies (6, 10, 11), could not be assessed with polymerase bound to a dextran surface because the concentrations of pure yeast TFIIF available and the k d were low, and the surface plasmon resonance effect was consequently near the lower limit of detection. We turned to the use of a sensor chip derivatized with a Ni 2ϩ chelate for binding hexahistidinetagged proteins (23). Omission of the dextran layer brings bound proteins closer to the chip surface, enhancing the reso- RNA polymerase II-TFIIH interaction was investigated with preparations of core TFIIH, a 5-subunit protein that included a trace of a sixth subunit, Ssl2 (22). No evidence of interaction was detected. Taking into account the concentration of TFIIH used, k a for RNA polymerase II-TFIIH must be less than about 10 4 s Ϫ1 M Ϫ1 , corresponding to K a less than or on the order of 10 6 M
General Transcription Factor-Factor Interactions-The only interaction among the general transcription factors detectable above the approximate limit was between TFIIE and TFIIH ( Figs. 1 and 4). The binding of TFIIH (concentration range, 5-100 nM) to immobilized TFIIE occurred with a k a value of 3.3 ϫ 10 5 s Ϫ1 M Ϫ1 (standard deviation was 1.2 ϫ 10 5 ) and a k d value of 1.1 ϫ 10 Ϫ3 s Ϫ1 (standard deviation was 1.9 ϫ 10 Ϫ4 ) (results of 15 determinations). To identify the subunits of these proteins involved in the interaction, Tfa1 and Tfa2 were individually immobilized on dextran chips. Tfa1 bound TFIIH with nearly the same affinity as intact TFIIE, whereas Tfa2 failed to bind TFIIH at all. Tfa1 interacted specifically with the Tfb1 subunit of TFIIH in a protein (Western) blot (Fig. 6).

DISCUSSION
The analysis by surface plasmon resonance reported here has revealed 4 of 10 possible pairwise interactions among RNA polymerase II transcription proteins (Fig. 4): TFIIB, -E, and -F bind RNA polymerase II, and TFIIE binds TFIIH. No other interaction was detectable above an approximate limit for an equilibrium association constant of 10 6 M Ϫ1 . The four interactions observed constitute a minimal set for formation of a complete transcription initiation complex since these interactions connect all five proteins studied here, and the remaining component of the complex, TFIID, is incorporated through contacts with both TFIIB and promoter DNA. The initiation complex formation guided by four primary interactions is consistent with the assembly pathway defined by others (1,(13)(14)(15). TFIID, and in particular its TBP component, binds to the TATA box in promoter DNA that creates a context for TFIIB binding; TFIIB then binds RNA polymerase, as do TFIIE and TFIIF, followed by TFIIH binding to TFIIE. Additional interactions, while dispensable for complex formation, are not ruled out. They may occur in the context of a ternary or higher complex either as a consequence of conformation changes or because they are of much lower affinity, below the limit of detection in this study, but persist when they are bridged by the primary interactions reported here.
The four primary interactions are all likely to be specific in view of the effects of mutations in TFIIB reported here and the results of functional and structural studies presented elsewhere. TFIIB-RNA polymerase II and TFIIE-TFIIH interactions in yeast are required for the initiation of transcription in vitro (12). TFIIE-TFIIH interaction has also been reported to stimulate TFIIH kinase activity, and TFIIF-RNA polymerase II interaction influences template binding (12, 20, 28 -30). Electron crystal structures of TFIIB-and TFIIE-RNA polymerase II complexes show single sites of transcription factor binding to the polymerase. 2 FIG. 5. Effect of various salts upon RNA polymerase II-TFIIE interaction. The binding of TFIIE (1 M) to immobilized RNA polymerase II in the presence of 120 mM potassium acetate (top curve), 40 mM ammonium sulfate (middle curve), and 120 mM sodium chloride (bottom curve) is shown.
FIG. 6. Subunit-subunit interactions of TFIIE and TFIIH. Following electrophoresis of TFIIH (1 g) in an SDS-10% polyacrylamide gel, proteins were transferred to nitrocellulose, denatured, renatured, and then allowed to bind the 35 S-labeled Tfa1 subunit of TFIIE as described (7). An autoradiograph (lane A) obtained with the use of a PhosphorImager plate (Molecular Dynamics, Inc.) and a silver-stained gel (lane B) are shown.