ABSTRACT
The [sigma] subunit of RNA polymerase determines promoter recognition and catalyzes DNA strand separation. The -35 promoter region is recognized by a helix-turn-helix motif in region 4, while the -10 region is specified, at least in part, by an amphipathic helix in region 2. We have proposed that conserved aromatic residues in [sigma] region 2.3 interact with the non-template strand of the -10 element to drive open complex formation. We now report that Bacillus subtilis[sigma]A holoenzyme, but neither core nor [sigma]A alone, binds with high selectivity to single-stranded (ss) DNA containing the non-template -10 consensus sequence. UV irradiation of holoenzyme-ssDNA complexes efficiently crosslinks [sigma]A to DNA and protease mapping supports a primary contact site in or near region 2. Several mutations in [sigma]A region 2.3, shown previously to impair promoter melting, affect ssDNA binding: Y184A decreases binding selectivity, while Y189A and W193A decrease the efficiency of photocrosslinking. These results support a model in which these aromatic amino acids are juxtaposed to ssDNA, consistent with their demonstrated role in stabilizing the open complex.
The [sigma] subunit of RNA polymerase (RNAP) determines promoter selectivity (1 ,2 ). Genetic and biochemical experiments have established that selective recognition of the -35 and -10 elements is mediated, at least in part, by conserved regions 4 and 2 of [sigma]. The -35 element is recognized by a conserved helix-turn-helix DNA binding motif. Although no structural information for this region of [sigma] is yet available, this interaction is likely to be similar to the sequence-selective binding of other helix-turn-helix proteins to double-stranded (ds) DNA targets. In contrast, recognition of the -10 consensus element is much more complex and is poorly understood (3 ).
The -10 region recognized by primary [sigma] factors, such as Escherichia coli [sigma]70 and Bacillus subtilis [sigma]A, is the sequence 5'-TATAAT-3' or a close variant. Mutations in region 2.4 of both [sigma]70 and [sigma]A can affect -10 recognition, particularly of the start site distal T, as judged by allele-specific suppression of promoter mutations (4 -6 ). Similar genetic experiments support an interaction of the corresponding region 2.4 residues in [sigma]H (7 ,8 ) and [sigma]E (9 -11 ) with their cognate -10 consensus elements. Nearby amino acids in region 2.3 may also participate in promoter recognition and are implicated in stabilization of the open complex (12 -14 ). Much or all of the -10 element is converted from dsDNA to ssDNA during maturation of the open complex and it is not yet clear whether sequence recognition occurs before, during or after this transition.
We have characterized a series of [sigma]A mutants altered in aromatic amino acids in region 2.3. Holoenzymes containing [sigma] factors bearing mutations at Y189, W192 or W193 are impaired in open complex formation and are cold sensitive for transcription on linear but not supercoiled DNA, as expected for a melting defect (12 ,13 ). These mutations are dominant negative, consistent with a defect subsequent to core binding (14 ). A [sigma]A Y184A mutant also leads to cold-sensitive transcription in vitro, but this mutant [sigma] can support cell growth, therefore this residue does not play an essential role in [sigma] function. Together, these residues define a `melting motif' thought to stabilize the open complex by interaction with the non-template strand within or near the -10 element (1 ,3 ).
This model is supported by several other observations. Mutagenesis of B.subtilis [sigma]E reveals that Y113 (equivalent to [sigma]A Y189) is critical for in vivo function and a missense mutation at the [sigma]E residue corresponding to [sigma]A W193 is dominant negative in vivo and blocks promoter melting in vitro (15 ,16 ). In addition, the recent crystallographic determination of the atomic structure of a [sigma]70 fragment containing region 2 indicates that the corresponding aromatic residues (Y420, Y425, W433 and W434) project into solution from an extended [alpha]-helix which also includes region 2.4 residues (Q437 and T440) which interact with the T-A base-pair at -12 (17 ). This extended binding surface, covering several turns of [alpha]-helix, appears to be key to the process of both -10 recognition and DNA melting.
RNAP holoenzyme binds tightly and specifically to ssDNA containing a -10 region consensus sequence (18 ). We report a series of studies aimed at identifying the amino acid determinants of this binding. Our data suggest that the region 2.3 `melting motif' contributes to the selectivity of binding and may also be a site of protein-nucleic acid photocrosslinking. The roles of region 2 in -10 recognition and DNA melting are discussed in the light of these results.
Bacillus subtilis RNAP, core enzyme and [sigma]A mutants were purified as described previously (12 ,19 ). The modification of [sigma]A to incorporate a protein kinase A (PKA) recognition site at the N-terminus will be described in detail elsewhere (F.L.de S. and J.D.H., unpublished results). The modified N-terminus of PKA-[sigma]A contains a substitution of MLRRASLG in place of the initiating methionine and can be labeled to high specific activity with [[gamma]-32P]ATP and the catalytic subunit of cAMP-dependent protein kinase (New England Biolabs; no. 6000).
Six oligonucleotides were used in these studies: C, 5'-ATTGGGTATAATTGACTCA-3'; C', 5'-TGAGTCA
Electrophoretic mobility shift assays (EMSAs) were performed by incubation of end-labeled DNA with core RNAP and [sigma]A at the concentrations indicated in the relevant figure legends. All reactions (20 [mu]l) were incubated at room temperature (22oC) for 30 min in binding buffer (20 mM Tris-HCl, pH 8.0, 10 mM NaCl, 10 mM MgCl2, 0.1 mM EDTA, 1 mM dithiothreitol, 5% v/v glycerol, 100 [mu]g/ml BSA). For equilibrium competitition DNA binding assays, the oliognucleotides were pre-mixed prior to addition of RNAP. After incubation, 2 [mu]l loading buffer (0.1% bromophenol blue, 0.1% xylene cyanol FF and 50% glycerol) were added and the reactions fractionated by 4% PAGE run in 1* TAE buffer at 180 V for 1 h at 4oC. Gels were pre-electrophoresed for 60 min prior to sample addition. After electrophoresis, gels were dried and visualized by autoradiography. To quantify the amount of DNA bound, a phosphorimager (Molecular Dynamics) was used together with ImageQuant data analysis software.
Reactions contained 0.4 pmol end-labeled oligonucleotide C, 2 pmol core RNAP and 10 pmol [sigma]A in 20 [mu]l binding buffer. Samples were incubated at room temperature for 30 min and then illuminated with a FotoDyne Foto/PrepI transilluminator (with the filter removed) for 15 min on ice. An aliquot of 10 [mu]l SDS-PAGE loading buffer was then added, samples were incubated in a boiling water bath for 3 min and separated by 7.5% SDS-PAGE. The crosslinked protein was visualized by autoradiography of the dried gel.
To map the site(s) of crosslinking by a fragment release assay, 100 pmol 32P-labeled PKA-[sigma]A were incubated with 50 pmol core RNAP and 50 pmol oligonucleotide CB in 100 [mu]l binding buffer for 30 min and irradiated with UV light for 15 min as above. Samples of 300 [mu]l 1 mg/ml streptavidin magnesphere paramagnetic beads (Promega Corp.) were prepared by washing three times with SA buffer (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 100 [mu]g/ml BSA) and then added to the photocrosslinking reaction. The resulting suspension was incubated at room temperature for 30 min with occasional mixing and the beads were then washed three times with 500 [mu]l SA buffer and suspended in 100 [mu]l SA buffer. Samples of 20 [mu]l were removed and treated with 10 [mu]l proteases (0.01 [mu]g/ml) at room temperature for 5 min prior to separation into supernatant (15 [mu]l) and bound (15 [mu]l) fractions. Each fraction was treated with an equal volume of SDS-PAGE loading buffer, incubated in a boiling water bath for 5 min and fractionated by 12.5% SDS-PAGE. The released protease fragments were visualized by autoradiography.
We have demonstrated previously that mutations in any of four residues in region 2.3 of [sigma]A (Y184, Y189, W192 and W193) lead to a specific promoter melting defect (12 ,13 ,21 ). These residues are close to a region 2.4 amino acid (Q196) known to recognize T12 (Fig. 1 ) (4 ). In the recently determined atomic structure for this region of the highly similar E.coli [sigma]70 protein (17 ), these residues appear to form an extended binding surface thought to interact with DNA. We postulated that those [sigma]A region 2.3 mutations which impair DNA melting might also affect the affinity or selectivity of binding to ssDNA representing the -10 non-template consensus sequence.
Holoenzyme containing [sigma]A binds to a short oligonucleotide containing the non-template strand of a consensus -10 element (C) with high affinity and selectivity (Fig. 2 A), consistent with the findings of Marr and Roberts (18 ). Oligonucleotide C is quantitatively shifted by 100 nM [sigma]A holoenzyme (lane 6). In contrast, [sigma]A holoenzyme does not detectably bind an oligonucleotide (A) which lacks a -10 consensus element (lane 5). Binding is specific for holoenzyme, as neither core nor [sigma]A alone can bind tightly to oligonucleotide C under these conditions.
We used equilibrium competition binding assays to determine the effect of several region 2.3 [sigma]A mutations on sequence-selective binding of holoenzyme to oligonucleotide C. The binding affinity and selectivity for the seven tested mutants varied greatly (Fig. 3 A). Of the seven mutant holoenzymes tested, five were able to bind oligonucleotide C as well as the wild-type: only the F178A and F186A [sigma]A proteins were grossly defective and this defect could be at least partially overcome by increasing the concentration of mutant [sigma]A in the reaction (see below). When binding to oligonucleotide C was tested in the presence of a 5-, 20- or 100-fold molar excess (over RNAP) of oligonucleotide D, each of the five most active mutant proteins still recognized and bound oligonucleotide C with a preference similar, but not identical, to that observed for wild-type [sigma]A. The ability of holoenzyme to bind to C in the presence of excess D was impaired by the Y184A mutation and, surprisingly, enhanced by the W193A mutation (Fig. 3 A). In contrast, when the competing DNA was double stranded (annealed oligonucleotides C and C'), these same three mutant [sigma]A proteins displayed a selectivity similar to the wild-type (Fig. 3 B). Thus, the primary effect of these mutations is on the ability of the holoenzyme to discriminate between competing oligonucleotides, rather than on the ability to distinguish ssDNA from dsDNA. This is further supported by dsDNA binding studies, which demonstrate that all three mutants are as selective as the wild-type when the double-stranded [sigma]A -10 consensus (C/C') is competed against the [sigma]D consensus -10 region (D/D') (data not shown).
Labeled oligonucleotide C can be crosslinked to [sigma]A by UV irradiation of holoenzyme-oligonucleotide complexes (18 ). Using this assay, we demonstrated that all seven [sigma]A mutants could be UV crosslinked to oligonucleotide C. The efficiency of crosslink formation varies between mutants, however, and there is a remarkable disparity between binding affinity and efficiency of crosslink formation (Fig. 4 ). Specifically, the Y189A and W193A mutants both display a signficant and reproducible decrease in the yield of crosslinked product (Fig. 4 ), despite the fact under these conditions the corresponding holoenzymes bind oligonucleotide C with an affinity and selectivity at least as high as the wild-type (Fig. 3 A and data not shown). This decrease in efficiency of crosslink formation is observed at [sigma]A protein concentrations of both 0.5 and 2 [mu]M. Using this same assay we find that the binding defects of both the F178A and F186A [sigma]A mutants can be suppressed at elevated concentrations of [sigma]A; crosslink formation is restored to near wild-type levels at 2-2.5 [mu]M F186A [sigma]A and at 5 [mu]M F178A [sigma]A.
The observation that two mutations (Y189A and W193A) in region 2.3 affect the efficiency of photocrosslink formation suggests that this region may constitute part of the observed ssDNA binding site (Fig. 1 ). We used protease footprinting to test the hypothesis that ssDNA binding is mediated by residues from region 2. For this analysis we prepared a derivative of [sigma]A containing a PKA recognition site at the N-terminus (PKA-[sigma]A). This protein was end-labeled with the catalytic subunit of cAMP-dependent protein kinase and the kinetics of chymotrypsin cleavage were monitored for free PKA-[sigma]A and the corresponding E[sigma]A- and E[sigma]A-ssDNA complexes. The rate of disappearance of full-length PKA-[sigma]A (band 1) is decreased between 2- and 3-fold in the presence of core enzyme, with a small additional decrease noted in the presence of oligonucleotide C (Fig. 5 ). The patterns of cleavage were very similar in both the presence and absence of core RNAP, suggesting that the major sites of chymotrypsin attack are surface exposed in the holoenzyme complex. However, when oligonucleotide C was present the rate of formation of bands 2 and 3 was decreased several-fold, while the rate of accumulation of band 4 was essentially unchanged, as judged by phosphorimage analysis. This suggests that the bound ssDNA differentially protected these sites against chymotrypsin attack. Since only N-terminal fragments were visualized in this experiment, the corresponding sites of cleavage could be assigned to region 4 (band 2), near the end of region 2 (band 3) and upstream of region 2 (band 4). The apparent molecular masses of these proteins, as judged by SDS-PAGE, were 55 (intact PKA-[sigma]A; band 1), ~47 (band 2), ~22 (band 3) and ~15 kDa (band 4).
To further define the position of the crosslink within the [sigma]A protein we incubated holoenzyme with 5'-biotinylated oligonucleotide C (oligo CB), treated the holoenzyme-ssDNA complexes with UV light and recovered the crosslinked protein using streptavidin paramagnetic beads. In this experiment the DNA was unlabeled and PKA-[sigma]A was labeled with 32P. The resulting protein-ssDNA complexes were treated with various proteases and the released N-terminal protein fragments were separated by SDS-PAGE and visualized by autoradiography (Fig. 6 ). The largest labeled protein fragments efficiently released from these covalent complexes were near 22 kDa, while several larger protease products were preferentially retained with the paramagnetic beads. This suggests that the predominant site of crosslinking is located near the end of region 2. As noted above, the ~22 kDa chymotrypsin cleavage product appeared to result from cleavage at the 2.2/2.3 boundary (the end of helix 13). Since this site of chymotrypsin cleavage is less accessible when ssDNA is bound to holoenzyme (Fig. 5 ) yet the corresponding product is efficiently released from the paramagnetic beads (Fig. 6 ), we suggest that the primary site of crosslinking is located C-terminal of this cleavage site, within or distal to helix 14. This also coincides with the location of the amino acid residues (Y189 and W193) which are implicated in UV photocrosslinking by the mutational studies (Fig. 4 ).
The [sigma] subunit of RNAP is essential for the process of promoter recognition and determines the preferred base sequences for both the -35 and -10 promoter elements (1 ,2 ,25 ). This suggests that [sigma] directly contacts the DNA within these regions. This inference is supported by the observation that specific amino acid changes in [sigma] can partially suppress the deleterious effect of certain promoter mutations (4 -8 ,10 ). This allele-specific suppression is consistent with recognition of the -35 promoter region by a helix-turn-helix unit in [sigma] region 4.2 and suggests that amino acids in region 2.4 contact the start site distal portion of the -10 element (Fig. 1 ). A role for [sigma] in start site melting is inferred from the observations that: (i) [sigma] can be photochemically crosslinked to the non-template strand within the open complex (26 ,27 ); (ii) point mutations within [sigma] impede open complex formation (12 ,13 ).
The interactions between [sigma] and DNA have been difficult to study experimentally because [sigma] has little or no affinity for DNA in the absence of the core subunits. This difficulty can be at least partially circumvented by the use of truncated [sigma]70 polypeptides lacking conserved region 1. In this case, specific complex formation between [sigma]70 fragments and promoter DNA can be detected by an equilibrium competition nitrocellulose filtration assay (28 ). Alternatively, some [sigma]70 family members naturally lack conserved region 1 and bind promoter DNA in vitro, albeit with low affinity (24 ,29 ). In solution the binding interaction between [sigma] and promoter DNA retains some, but not all, of the features characteristic of promoter recognition. First, this interaction has much lower sequence selectivity than observed with RNAP holoenzyme. Second, binding is apparently not affected by -10 region mutations at start site-proximal positions, even though these mutations impair promoter activity (30 ). Third, [sigma]70 fragments partially bind heteroduplex (bubble) templates independent of the sequence within the single-stranded region (30 ) while the holoenzyme retains sequence selectivity even in single-stranded regions (31 ,32 ). We have found that [sigma]A fragments lacking region 1 bind to both dsDNA and ssDNA with low affinity (Kd values between 1 and 10 [mu]M) and little sequence selectivity. Further, binding to ssDNA is not significantly affected by region 2.3 mutations that are known to affect promoter melting in the context of the holoenzyme (unpublished data). Thus, we concluded that [sigma]A fragments lacking region 1 were not a good model system for analysis of the sequence-selective interactions between [sigma] and promoter DNA.
We now report that [sigma]A RNAP binds selectively and with high affinity to short oligonucleotides containing a consensus -10 sequence, as also noted for E.coli [sigma]70 RNAP (18 ). This provides a simple model system for investigation of the role of [sigma] in -10 recognition. High affinity, sequence-selective recognition of the -10 region requires both core and [sigma]A and is affected by region 2.3 mutations known to impair promoter melting. Specifically, a Y184A mutant [sigma]A has a reduced ability to select the -10 consensus oligonucleotide from a mixture of two oligonucleotides and the Y189A and W193A mutant [sigma]A proteins have a reduced efficiency of photocrosslink formation, even though they bind tightly and selectively to the consensus oligonucleotide.
The ssDNA binding and UV crosslinking results reported here can be interpreted in the light of our previous functional assessment for the various aromatic amino acids in region 2.3. The Y180A mutant [sigma]A is most similar to wild-type in vitro and is capable of supporting growth in vivo (12 ,14 ). As expected, the Y180A mutant behaves like the wild-type in the assays reported here. We demonstrated previously that mutations at two conserved Phe residues (F178 and F186) led to [sigma] proteins defective in supporting open complex formation. Since the transcriptional defects imparted by these [sigma] factors could not be suppressed by negative supercoiling and, unlike other defective mutants, they were not dominant negative, we postulated a primary defect in protein folding or core binding. We now report that these mutant [sigma]A proteins are defective in ssDNA binding, although they can be efficiently crosslinked to ssDNA at elevated concentrations of [sigma]A. These two Phe residues form part of the hydrophobic core for the region 2 domain (17 ), consistent with the idea that substitutions at these positions destabilize [sigma] structure.
The Y184A mutant [sigma]A protein imparts a promoter melting defect in vitro and Y184A mutant cells have a reduced growth rate at low temperature (12 ,14 ,21 ). This mutant [sigma]A is also least able to bind the C oligonucleotide selectively from a mixture of C and D oligonucleotides, suggesting a modest defect in sequence recognition. However, since Y184 is not conserved among [sigma]70 family proteins (33 ) and mutants at this position are functional in vivo in both [sigma]A and [sigma]E (14 ,16 ), this residue does not play an essential role in either promoter recognition or melting.
Remarkably, two holoenzymes known be defective in promoter melting (containing [sigma]A Y189A or W193A) are able to bind tightly and selectively to a consensus -10 oligonucleotide (C) but are reduced in their ability to form protein-nucleic acid crosslinks upon irradiation with UV light. The simplest interpretation of these results is that Y189 or W193 participate directly in UV photocrosslinking. We cannot rule out, however, indirect effects of these mutations on the proximity of other amino acids to DNA. The chemistry of photocrosslinking between DNA and proteins is complex and many amino acids can form photoproducts with DNA bases (34 ,35 ). Formation of a protein-nucleic acid crosslink between [sigma] region 2 and ssDNA is also observed in studies with E.coli RNAP; truncated [sigma]70 derivatives containing region 2 still support photocrosslink formation, suggesting a site of crosslinking between [sigma]70 amino acids 360 and 448 and therefore near or within region 2 (18 ). In addition, binding of holoenzyme to a -12T -> C mutant oligonucleotide is enhanced by an allele-specific suppressor mutation in [sigma]70 region 2.4, Q437H. Thus, [sigma]70 Q437 is implicated in the sequence-selective recognition of the -12T position in ssDNA (18 ).
Interpretation of these mutant studies is greatly facilitated by the availability of an atomic structure for a fragment of [sigma]70 containing region 2 (17 ). This structure reveals a tightly packed cluster of three [alpha]-helices, designated 12-14. Conserved region 2.1 forms the end of helix 12a and helix 12b, region 2.2 forms helix 13, while region 2.3 and 2.4 together form helix 14. Amino acids implicated in both -10 recognition and binding (Q437, R441) and in promoter melting (Y425, Y439, W433 and W434) project from five adjacent turns of helix 14 and together are postulated to mediate -10 recognition and DNA strand separation (Fig. 1 ). The precise nature of the protein-nucleic acid interactions involved in these processes have not yet been determined, but our data suggest that this region contributes to the sequence-selective interaction of holoenzyme with ssDNA. This interaction is not easily competed by the corresponding duplex fragment, suggesting that ssDNA is a favored substrate for binding to this site. This supports our previous proposal that recognition of the -10 sequence can occur in a single-stranded site (12 ).
We thank M.Marr and J.Roberts for helpful discussions. This work was supported by NIH grant GM 47446.
REFERENCES