DNA binding is not sufficient for H-NS-mediated repression of proU expression.

H-NS is a major component of bacterial chromatin and influences the expression of many genes. H-NS has been shown to exhibit a binding preference for certain AT-rich curved DNA elements in vitro. In this study we have addressed the factors that determine the specificity of H-NS action in vitro and in vivo. In bandshift studies, H-NS showed a slight binding preference for all curved sequences tested whether GC-based or AT-based; the specific architecture of the curve also influenced H-NS binding. In filter retention assays little difference in affinity could be detected for any sequence tested, including the downstream regulatory element (DRE) a downstream curved DNA element required for H-NS to repress transcription of the Salmonella typhimurium proU operon in vivo. A Kd of 1-2 microM was estimated for binding of H-NS to each of these sequences. In vivo, the distance between the proU promoter and the DRE, their relative orientations on the face of the DNA helix, and translation of the DRE had no major effect on proU regulation. None of the synthetic curved sequences tested could functionally replace the DRE in vivo. These data show that differential binding to curved DNA cannot account for the specificity of H-NS action in vivo. Furthermore, binding of H-NS to DNA per se is insufficient to repress the proU promoter. Thus, the DRE does not simply act as an H-NS binding site but must have a more specific role in mediating H-NS regulation of proU transcription.

The binding of H-NS to DNA is relatively nonspecific (3,5,10). This raises a paradox. How does H-NS influence transcription at only a subset of promoters? Part of the explanation may come from the observation that H-NS exhibits a binding preference for certain intrinsically curved DNA sequences in vitro (11)(12)(13)(14) and that, at least at some promoters, curved DNA elements are required for H-NS action in vivo (13). Indeed, it may be that curved DNA sequence elements are required for the action of H-NS at all promoters (15)(16)(17). However, for the limited number of DNA curves studied, the increased preference for curved DNA, compared with nonspecific binding to generic DNA, is only a few-fold (5,12,13,18). Furthermore, as curved DNA sequences are found near most promoters, including many promoters that are not responsive to H-NS (for review see Ref. 19), it is not clear how specificity is achieved. Finally, the distance between the curve and the RNA polymerase binding site can be varied up to 200 bp. 1 Thus, it is unclear whether DNA curvature per se is the principal factor that determines H-NS action or how the interaction of H-NS and curved sequences might influence promoter function.
To address the mechanisms by which H-NS affects transcription we have used the proU promoter of Salmonella typhimurium as a model. proU encodes a transport system for the osmoprotectant glycine betaine, and its expression is increased when cells are exposed to high extracellular osmolarity (20 -23). proU appears to be one of the simplest H-NS-responsive promoters as no trans-acting factor, other than H-NS itself, appears to be required for its regulation. proU expression is derepressed in hns mutants (6,7,24), although residual osmoregulation is retained (13,25,26). Repression of proU by H-NS requires an intrinsically curved region of DNA located around 200 bp downstream of the transcription start site, within the first structural gene of the proU operon (the downstream regulatory element, DRE) (12, 26 -28). However, the binding preference of H-NS for the DRE in vitro is, at most, only a few-fold greater than for generic DNA, and the precise distance between the DRE and the proU promoter does not appear to be critical (13,25,26). It is not known how interaction of H-NS with the DRE might influence transcription.
In this paper, we have addressed the factors that determine the specificity of H-NS action both in vivo and in vitro. Consistent with previous reports, a binding preference for curved DNA was detected under certain conditions, although this preference was found to be limited and cannot account for the specificity of action of H-NS in vivo. Furthermore, the binding of H-NS to DNA per se is insufficient to account for the effects of H-NS on proU expression. Thus, the DRE cannot simply provide a site for H-NS binding but must serve a more specific role in H-NS-mediated gene regulation.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-Wild-type S. typhimurium LT2 and its congenic derivatives CH1838 (hns-1::Kan r ) and CH1833 (zde-1710::mTn10) were used for expression studies (8). Note that zde-1710::mTn10 has no phenotype but was used solely for strain construction (8). All plasmids were constructed and maintained in Escherichia coli recA strains prior to electroporation into the appropriate S. typhimurium strain. Cells were grown with aeration at 37°C in LB (29).
Plasmids Containing Curved and Noncurved DNA Inserts-To generate defined curved and noncurved DNA fragments, complementary oligonucleotides (Table I) were annealed, ligated together, and separated by electrophoresis in a 3% NuSieve agarose gel. Ligated fragments of an appropriate size (100 -200 bp) were isolated from the gel using a QIAEX gel extraction kit (Qiagen). The overhanging ends of the fragments were filled in with the Klenow fragment of DNA polymerase and ligated into the EcoRV site of pBluescript I SK(Ϫ) (Stratagene). Three plasmids were generated in this manner, containing the GC curve (pAF1), GC noncurve (pAF2), and AT planar curve (pAF3). Two closely related plasmids containing the AT curve (pTP4) and AT noncurve (pTP5) have been described previously (13). A sixth plasmid (pAF6) was constructed by ligating a 200-bp fragment encompassing the S. typhimurium proU DRE (base pair ϩ99 through ϩ299; Ref. 30) between the NotI and SpeI sites of pBluescript II SK(Ϫ). The 200-bp fragment was generated by PCR using primers ϩ99 (5ЈGGGCGGCCG-CATATTTGGAGAGCATCCG-3Ј) and ϩ299 (5ЈGGACTAGTTTCAAT-CAGGCGATTGAGAAGG-3Ј) which incorporate sites for NotI and SpeI to facilitate cloning. The nature and size of the DNA inserts in these plasmids are given in Table II. To clone these curved/noncurved sequences between the proU promoter and the lacZ reporter gene, a multiple cloning site was created in plasmid pAV399 (13). Plasmid pAV399 contains the proU promoter region (Ϫ217 through ϩ100, where ϩ1 is the start of the transcription) (30), directing transcription of the lacZ reporter gene. A 51-bp polylinker was inserted into the EcoRI site between the proU promoter and lacZ gene. The polylinker was constructed by annealing two complementary oligonucleotides, 5ЈAATTCATACGCGGCCGCAAGTCTA-GACTTGACTAGTACTGGTACCTATGCG-3Ј and 5ЈAATTCGCATAGG-TACCAGTACTAGTCAAGTCTAGACTTGCGGCCGCGTAT-3Ј, containing NotI, SpeI, XbaI, and KpnI restriction sites and an EcoRI site overhang at each end. The resulting plasmid, pAF450, was used as a vector for the construction of a further six plasmids in which curved or noncurved fragments were inserted into the polylinker. Five of the inserts were the curved and noncurved fragments from pAF1, pAF2, pAF3 (excised with EcoRI and HindIII), pTP4 and pTP5 (excised with BamHI and EcoRI). The sixth was a 200-bp fragment, generated by the PCR, encompassing the S. typhimurium proU DRE (bp ϩ99 through ϩ299) and incorporating terminal NotI and SpeI sites.
Plasmids with Altered Distances between the proU Promoter and the Downstream Curve-pSJ4 contains the same 317-bp proU promoter fragment (bp Ϫ217 to ϩ100) as pAV399, cloned into the EcoRI site upstream of the luciferase reporter genes in pSB71 (Table II). A SalI-SacI fragment from pSJ4, containing the proU promoter coupled to the luxAB reporter genes, was cloned into the pALTER vector (Promega). Using the Altered Sites in vitro mutagenesis system (Promega), unique sites for the restriction endonucleases NcoI and BglII were introduced at bp ϩ17 and ϩ31, respectively (Fig. 1). The mutated SalI-SacI fragment was recloned into pSJ4 generating plasmid pBJ1. Plasmids pBJ2, pBJ4, pBJ5, and pBJ6 were generated from pBJ1 by inserting appropriate DNA fragments between the NcoI andBglII sites (Fig. 1). Plasmid pBJ2 contains a 210-bp fragment NcoI-BglII fragment obtained by PCR amplification of the trp-lacZ fusion from plasmid pCH133 (13,31). Plasmids pBJ4, pBJ5, and pBJ6 were generated by ligating the following oligonucleotides (and their complementary strands) between the NcoI and BglII sites, increasing the distance between the proU promoter and DRE by 5, 10, and 15 bp, respectively: pBJ4, 5ЈGAATTCCATGGCTATATCGCCGATAGATCTGAATTC 3Ј; for pBJ5, 5ЈGAATTCCATGGCTATATCGCCGATGAAAGAGATCTGAAT-TC 3Ј; for pBJ6, 5ЈGAATTCCATGGCTATATCGCCGATGAAAGCG-GCGAGATCTGAATTC3Ј.
Anomalous Migration of DNA Fragments-Curved and noncurved DNA fragments were isolated by restriction enzyme digestion and gel purification. Fragments were electrophoresed on a 12% polyacrylamide gel at 5°C in TBE buffer (32), visualized by ethidium bromide staining, and relative mobility measured compared with standard markers.
DNA Bandshift Studies-H-NS protein from S. typhimurium was purified as described (13) and stored at 1-2 mg/ml at Ϫ20°C in 50% glycerol. 1 g of plasmid DNA digested with HaeIII was incubated in 9 l of binding buffer (10 mM Tris-Cl pH 7.5, 15 mM KCl, 2 mM spermidine, 15% glycerol, 0.1 mM EDTA), together with 1 l of H-NS diluted (in binding buffer) to the appropriate concentration, for 2 min at 20°C. The samples were electrophoresed on a 3% Metaphor agarose gel at room temperature at 4 V/cm. The DNA was visualized by staining with ethidium bromide.
For competition bandshift studies, appropriate curved or noncurved DNA fragments were isolated from plasmids by restriction enzyme digestion and gel purification. A typical bandshift reaction was carried out in a volume of 15 l and contained 1 g of H-NS protein, 10 ng of radiolabeled DNA, the indicated amount of competitor DNA, in buffer (10 mM KH 2 PO 4 /K 2 HPO 4 , pH 7.6, 15 mM KCl, 0.1 mM EDTA). The reaction was incubated at room temperature for 30 min, and 5 l of loading buffer (50% glycerol, 2.5% bromophenol blue) was added to the reaction. Samples were loaded onto a 5% polyacrylamide gel in TBE and electrophoresed at 8 V/cm at 5°C for 2 h. After electrophoresis the gel was dried and radiolabeled DNA detected by autoradiography.
Filter Retention Assays-Radiolabeled DNA fragments, extending from 77 bp upstream to 51 bp downstream of the polylinker, were amplified from plasmids pAF451-456 using Vent DNA polymerase (New England Biolabs) as directed by the manufacturer, but using only 100 M each dNTP and with [␣-32 P]dCTP included at 2 Ci/nmol. The amplified fragments were purified from 5% polyacrylamide gels and quantitated by liquid scintillation spectroscopy. DNA binding reactions (20 l) were done in FR buffer (10 mM Tris-Cl, pH 7.5, 15 mM KCl, 10 mM dithiothreitol, 5 mM MgCl 2 , 1 mM spermidine, 0.1 mM EDTA) using 5-10 ng of radiolabeled DNA and amounts of H-NS ranging from 0.17 to 1.67 g. Reactions were incubated at room temperature for 10 min and then filtered under vacuum on Biotrace NT filters (Gelman). The filters were washed with 5 volumes of FR buffer minus spermidine, dried, and the radioactivity retained on the filters determined using a Molecular Dynamics PhosphorImager.
DNA Manipulations-Standard methods were used for gel electrophoresis and the construction of recombinant plasmids (32). Plasmid DNA was electroporated into S. typhimurium using a Gene Pulser® (Bio-Rad) according to the manufacturer's instructions. One pulse was applied, at capacitance 25 microfarads, resistance 200 ⍀, and voltage 2.45 kV. Electrocompetent S. typhimurium were prepared as described (13).
Reporter Gene Assays-Cells were grown in low osmolarity medium (see below) to an A 600 between 0.3 and 0.5. The culture was then split and NaCl added to one-half of the culture to a final concentration of 0.3 M to give an osmotic upshock. The other half of the culture provided a low osmolarity control. The cells were grown for a further 40 min and ␤-galactosidase or luciferase activities assayed as described previously Details of the fragments introduced between these two sites to create pBJ2 and pBJ4-pBJ6 are given under "Experimental Procedures." (13). Note that using transient osmotic shock, rather than growing cells to steady state at different osmolarities, circumvents possible effects of osmolarity on plasmid copy number and on mRNA stability. For luciferase assays, NB (Difco) was used as the low osmolarity medium; for ␤-galactosidase assays LB with NaCl omitted was used as the low osmolarity medium. All data are the result of multiple independent determinations. For ␤-galactosidase assays the data are presented after correction for background expression using data obtained from an identical plasmid lacking a promoter upstream of the reporter gene (pAV250).

Construction of Plasmids Containing Synthetic Curved and Noncurved DNA Fragments
Previous studies have shown that H-NS exhibits a binding preference for curved DNA resulting from phased A tracts. To ascertain whether H-NS recognizes DNA curvature per se or the A tracts responsible for curvature, a series of plasmids was constructed containing curved and noncurved DNA fragments ( Table I). The base composition of each curved and noncurved pair was similar, but the order of the bases resulted in distinct architectures. One pair of fragments was the AT curve and AT noncurve described previously; H-NS binds to the AT curve preferentially (12,13). A second pair of curved and noncurved fragments (the GC curve and GC noncurve) were constructed which did not contain A tracts. The GC curve contains the motif "CCGG" in phase with the pitch of the DNA helix (an average of once every 10.5 bp) and has been shown previously to be curved by anomalous electrophoretic mobility and ligation assays (33)(34)(35). The GC noncurve has a similar base composition to the GC curve but is designed such that the DNA is much less curved. A fifth DNA fragment (AT planar curve) contained an AT-rich curve designed to be planar (36), in contrast to the nonplanar AT curve described above. The sixth fragment was the DRE from downstream of the S. typhimurium proU promoter (bp ϩ99 to ϩ299). Five of the synthetic DNA fragments were generated from the appropriate oligonucleotides (Table I), whereas the sixth fragment, the DRE, was amplified by the PCR. All fragments were cloned into pBluescript. These plasmids, and the sizes of the inserted curved and noncurved DNA fragments, are described in Table II.

Characterization of Curved and Noncurved DNA Fragments
Computer Predictions- Fig. 2 shows the predicted structures of 120-bp fragments formed by ligation of the five different oligonucleotides determined by the CURVATURE program (37). Panel A shows the predicted helical path of the AT curve. Note that this curve is "out of plane" with the axis of the DNA helix with the end of the fragment pointing into the plane of the page. Panel B shows the AT noncurve in which the A tracts are out of phase. This sequence has several small "snake-like" curves although, on average, the helical path is straight. Panel C shows the predicted path of the GC curve. The curvature is not predicted to be as great as for the AT curve, but it is "in plane" (i.e. the path of a longer piece of DNA would form a flat mini-circle as opposed to the superhelical path of the curve in panel A). Panel D depicts the GC noncurve, of similar base composition to the GC curve ( Table I). The fragment is slightly curved, although less so than the GC curve, and is probably closest to "generic" DNA of all the fragments depicted. The AT planar curve depicted in panel E is in plane, such that a 126-bp fragment will form a flat mini-circle (36). Panel F shows the predicted curvature of the DRE. Note that the curvature of the DRE is less than that of the synthetic AT curves; the DNA is curved by roughly 60°over about 100 -120 bp.
Gel Mobility-Although the DNA fragments were predicted to be either curved or noncurved, it was necessary to demonstrate this experimentally. One commonly used method for assessing DNA curvature is anomalous mobility on polyacrylamide gels (38). Fig. 3 shows the relative mobility of the five synthetic DNA fragments and the DRE. The apparent sizes of the fragments under these conditions were estimated from their mobilities with respect to molecular weight markers. The "R L values," the ratios of the apparent sizes to the actual sizes of the fragments, were as follows: GC noncurve, 1.0; GC curve, 1.3; AT curve, 2.8; AT noncurve, 1.0; AT planar curve, 2.1, DRE, 1.2. The curvatures measured experimentally under these conditions were consistent with the computer predictions.

Binding of H-NS to Curved and Noncurved DNA Fragments by Bandshift
The relative affinity of H-NS for curved and noncurved DNA fragments was initially assessed by bandshift assays. Plasmids containing the curved and noncurved DNA fragments (pAF1-3; pAF6; pTP4 -5) were digested with HaeIII, incubated with different amounts of H-NS, and electrophoresed in agarose gels. Fig. 4 shows that, in general, the larger the DNA fragment, the lower the H-NS concentration required to initiate a bandshift. This is consistent with H-NS binding nonspecifically to generic DNA. However, the fragment containing the AT planar curve was "shifted" at an approximately 5-fold lower H-NS concentration than the other fragments of similar size, demonstrating a small binding preference for this curve. In contrast, no significant preference could be detected for the nonplanar AT curve (Fig. 4). As the AT curve and AT planar curve differ in the planarity of the curve, this observation demonstrates a role for the specific architecture of the curve in determining binding preference. Importantly, in the same assay, H-NS did not show a detectable binding preference for any of the other curved or noncurved fragments, including the DRE (data not shown).
An alternative assay that has been used previously to show that H-NS has a binding preference for curved compared with noncurved DNA fragments is a competition bandshift. DNA fragments were isolated and end-labeled with [ 32 P]dATP. In each assay, one DNA fragment was radiolabeled and incubated with sufficient H-NS to form a discrete bandshifted protein-DNA complex. Excess unlabeled competitor DNA fragment was then added to 1, 10, or 100 ϫ the mass of labeled fragment. Addition of competitor DNA reduced the amount of "bandshifted" complex that could be quantitated by densitometry. An example of such an experiment is shown in Fig. 5. When the AT curve fragment was labeled (lanes a-h), addition of 10-fold excess of the AT curve competitor fragment abolished the bandshifted complex (lane d). In contrast, 10-fold excess of the AT noncurve competitor fragment (lane g) had little effect, and 100-fold excess was required to prevent complex formation (lane h). Thus, consistent with previous data (12,13,18) H-NS has a greater affinity for the AT curve fragment than for the AT noncurve fragment, although the difference in affinity is less than an order of magnitude.
A similar experiment (Fig. 5) shows that H-NS has a higher  affinity for the GC curve than the GC noncurve; this difference in affinity is similar to the difference between the AT curve and AT noncurve. Thus, the preferential affinity for curved DNA is not restricted to A tract curves; and curvature, rather than base composition, determines H-NS binding preference. As H-NS binds the AT planar curve better than the nonplanar AT curve, the specific architecture of the curve is also important. However, in no case was the affinity for any sequence, includ-ing the DRE, greater than a few-fold higher than for noncurved or generic DNA.

H-NS-DNA Interactions by Filter Retention
To obtain a more quantitative comparison of the apparent affinity of H-NS for the different curved and noncurved DNA fragments, we used a nitrocellulose filter retention assay. In this assay no significant binding preference for any of the DNA ligands tested was detected; all the DNA fragments were retained by H-NS to a similar extent (nearly 100%) and with a similar concentration dependence (Fig. 6). To confirm that the assay was measuring concentration dependence and not the stoichiometry of the H-NS-DNA interaction, we repeated the binding studies with different amounts of DNA ligand (10 and 50 ng) and obtained indistinguishable binding curves (data not shown). The apparent K d values, assuming that the H-NS preparation was 100% active and that H-NS interacts with DNA as a monomer, were calculated by curve-fitting analysis. The apparent K d values for the interaction of H-NS with all of the sequences were similar, although the AT planar curve bound slightly better (K d ϭ 1.0 M) than the other fragments (K d ϭ 1.6 -2.3 M).
Thus, no significant difference in affinity of H-NS for any sequence, including the DRE, could be detected by filter retention. This is in apparent contrast to bandshift studies that showed slightly enhanced affinity for curved sequences. This difference is likely to be due to differences in the rates of association and dissociation of the protein-DNA complexes during electrophoresis which may contribute to the apparent specificity of the interaction under bandshift conditions. However, measurement of differences in the on and off rates of the H-NS⅐DNA complexes with curved and noncurved DNA ligands using the filter retention assay was precluded by the rapid interaction kinetics of H-NS with DNA: the complex forms and dissociates within 10 s (not shown).

The Distance between the proU Promoter and the Downstream Curve Is Unimportant in Vivo
To ascertain whether the relative efficiency with which the curved and noncurved sequences bind H-NS in vitro reflects their activity in vivo, the ability of these sequences to replace the function of the DRE in vivo was assessed. The DRE is located approximately 200 bp downstream of the proU pro- moter and is required for H-NS-mediated repression of transcription. Before replacing the DRE with heterologous sequences, it was necessary to ensure that the precise distance between the promoter and downstream curve is not important. Thus, defined oligonucleotides were introduced between the proU promoter and the lux reporter gene in plasmid pBJ1, increasing the distance between the proU promoter and the downstream curve by 5, 10, 15, or 210 bp (Fig. 1). proU promoter activity of each of these plasmids was monitored by measuring lux reporter gene activity in S. typhimurium hns ϩ strain CH1833 (Table III). Increasing the distance between the proU promoter and the downstream curve had no major effect on osmoregulation of transcription, although the 210-bp fragment did reduce the absolute magnitude of expression somewhat. As the insertion of a nonintegral number of turns of the DNA helix (5 or 15 bp) will change the "face of the helix" of the curve relative to the proU promoter, it appears that the orientation of the curve with respect to the promoter is relatively unimportant. Thus, the distance between the promoter and the DRE, and their relative orientations with respect to the face of the helix, can be varied with little effect on expression or regulation. This is consistent with other reports (26). Importantly, this result ensures that interpretation of the following in vivo experiments is not influenced by the precise position at which the curves were inserted.

Ability of Curved and Noncurved DNA Fragments to Replace DRE Function in Vivo
Plasmid pAF450 contains a segment of the proU promoter extending from Ϫ219 to ϩ100 fused to the lacZ reporter gene. This plasmid lacks the DRE, and hence, the proU promoter is derepressed and exhibits minimal osmoregulation. Curved and noncurved DNA fragments were inserted into the polylinker between the proU promoter and the lacZ reporter gene in this plasmid (see "Experimental Procedures" and Table II). The resultant plasmids were electroporated into wild-type S. typhimurium (LT2) and ␤-galactosidase expression measured at low osmolarity and after osmotic upshock (0.3 M NaCl). The ratio of expression at high to low osmolarity, the induction ratio, indicates the degree of osmoregulation.
Plasmid pAV260 which contains the proU promoter region from Ϫ207 to ϩ736, and therefore includes the DRE, served as a positive control. Promoter activity was tightly osmoregulated (induction ratio ϭ 63.0) ( Table IV). The proU promoter in plasmid pAF450, from which the DRE was deleted, was derepressed at low osmolarity such that the osmotic induction ratio was severely reduced (induction ratio ϭ 7.3). These results with the control plasmid were as expected (13). It should be noted that, as reported previously, residual osmotic regulation was observed when the DRE was deleted showing that sequences other than the DRE play some role in regulation (13,26,28). Re-insertion of the DRE between the promoter and lacZ restored normal osmoregulation (pAF456; induction ratio of 71.0). Although the DRE is translated in its natural location within the first structural gene of the proU operon, in this plasmid the DRE is not translated. As normal osmoregulation was observed, this demonstrates that the influence of the DRE on proU transcription is independent of translation. In contrast to the DRE, none of the other curved or noncurved fragments, including the AT planar curve, restored osmoregulation when inserted at the same site in pAF450.
The plasmids were also introduced into the congenic hns strain CH1838. As expected for the positive control plasmid pAV260, expression at low osmolarity was severely derepressed such that the induction ratio was reduced from 63.0 to 2.0 (Table IV). A similar derepression at low osmolarity, reducing the induction ratio (71.0 to 1.3), was seen for plasmid pAF456 into which the DRE had been reinserted. Thus, as reported previously, repression of the proU promoter at low osmolarity requires both H-NS and the DRE.
Importantly, for the plasmid with the AT planar curve inserted (pAF453), proU expression was derepressed in the hns strain compared with the hns ϩ strain. Thus, it appears that H-NS interacts with this curve in vivo. However, in the hns ϩ strain the AT planar curve did not lead to full repression, in contrast to the DRE. Thus, the binding of H-NS to the AT planar curve only partially represses transcription, whereas H-NS binding to the DRE results in full repression.
It is important to note that, in the absence of H-NS, the nature of the sequence inserted downstream of the proU promoter influenced the basal level of expression. Most notably, the DRE and AT planar curve led to maximum derepression at both low and high osmolarity (19,280 and 18,974 units, respectively, at low osmolarity). In contrast, the other sequences achieve a degree of repression even in the absence of H-NS. This is most obvious for the GC curve and AT curve (pAF452 and pAF454, respectively) where expression at low osmolarity in the presence of H-NS was 1226 and 2466 units, respectively, which only increased to 2114 and 2121 units in the absence of H-NS. This implies that the overall configuration of the downstream region is important for optimal transcription, irrespective of the presence or absence of H-NS. DISCUSSION To understand the role of H-NS in the bacterial cell, and specifically the mechanism by which it influences transcription, it is necessary to understand how this protein interacts with DNA. It has previously been reported that, although H-NS binds nonspecifically to any DNA molecule, it has enhanced binding to intrinsically curved DNA. Furthermore, at least some of the in vivo sites of action of H-NS are curved. This led to the hypothesis that the interaction of H-NS with curved DNA is important for its mode of action. In this study we have tested this hypothesis, asking two questions: whether DNA curvature per se determines the specificity of action of H-NS, and whether the binding of H-NS to curved DNA is sufficient to account for the effects of H-NS on gene expression.
H-NS has previously been shown, by bandshift studies, to exhibit a binding preference for intrinsically curved DNA elements generated by phased A tracts (11)(12)(13). We have shown here that H-NS has a similar binding preference for non-A tract curves (GC curves). Thus DNA curvature, rather than the specific presence of A tracts, appears to be responsible for this H-NS binding preference. In addition, H-NS bound more strongly to a planar curve than to a nonplanar curve of similar base composition, showing that the specific architecture of the curve also influences H-NS binding. However, in these bandshift studies the maximum preference for any curved sequence tested, including the DRE, was only a few-fold greater than for generic DNA. This is consistent with previous reports that H-NS shows a binding preference for curved DNA; these other studies also used similar bandshift assays, and all demonstrated less than a 10-fold preference for any curved DNA fragment compared with generic DNA (11)(12)(13)(14)18).
In apparent contrast to the bandshift experiments, filter retention experiments failed to reveal any major binding preference of H-NS for curved DNA or for the DRE. Although the AT planar curve did bind with slightly higher K d than the other fragments tested, this was only by about 2-fold. The difference between bandshift and filter-retention studies probably reflects differences in on/off rates for the different sequences which during electrophoresis could influence apparent binding affinities in bandshift experiments. However, the on/off rates were too fast to permit measurement of any differences in these parameters. Despite the apparent difference in conclusions reached using the two methodologies it is more important to emphasize the similarities: in neither type of assay was more than about a 5-fold binding preference for any sequence observed. Furthermore, the binding preference for the DRE was slight and less than for the AT planar curve. Thus, it is clear that the binding preference for curved DNA, or the native DRE, cannot account for the specificity of H-NS action.
The filter retention experiments allowed the K d of H-NS binding to DNA ligands to be estimated as 1-2 M, assuming that the H-NS preparation was 100% active and that H-NS interacts with DNA as a monomer. As H-NS is able to form oligomers (15), 2 the actual K d value may be somewhat lower. Thus, the affinity of H-NS for DNA is relatively poor in comparison with classical sequence-specific regulatory proteins, although better than nonspecific binding of proteins to DNA. For example, the integration host factor interacts with its specific sites with high affinity (K d Ϸ 50 nM) and with generic DNA with a K d of about 20 M (39). Instead, H-NS interacts with DNA with similar affinity to HU, the other major chromatin-associated protein in E. coli, which has a K d for generic DNA molecules of about 1 M (40). The low affinity of the H-NS-DNA interaction cannot be reconciled with a classical repressor-type mechanism for modulating gene expression.
At the proU promoters of E. coli and S. typhimurium, H-NS represses transcription by binding to the DRE, located some 100 -200 bp downstream of the transcription start site (13,25,26). To determine whether H-NS binding per se is sufficient to maintain repression of the proU promoter, defined sequences were cloned between the proU promoter and a downstream reporter gene, replacing the DRE. As a control we showed that the precise distance between the proU promoter and the DRE, and their relative orientations on the face of the helix, had little effect on H-NS-mediated repression of transcription. This is consistent with previously published data (26,28). Insertion of the DRE between the proU promoter and lacZ reporter gene restored H-NS-mediated repression at low osmolarity, as expected. However, none of the other curved or noncurved sequences tested could replace the DRE (although it should be noted that we have previously shown that a heterologous intrinsic curve from the luxAB genes can replace the DRE; 13). Even the AT planar curve, to which H-NS binds with higher affinity than it does to the DRE, could not functionally replace the DRE. Previous studies have indicated that H-NS binds to the AT planar curve in vivo (17), and in this study the effect of the AT planar curve is altered in an hns mutant implying that H-NS binds to this curve in vivo. Thus, the binding of H-NS to a downstream curved sequence per se is insufficient to repress proU transcription. This argues strongly against classical repressor-type models in which H-NS directly occludes RNA polymerase binding at the proU promoter. These findings suggest that the DRE is not simply a specific binding site for H-NS but, instead, acts as an effector site for H-NS action. One possible mechanism by which the DRE might act as an effector is by undergoing specific topological changes in response to the interaction of H-NS. This is consistent with the findings reported here that, even in the absence of H-NS, the nature of the downstream sequence influences promoter activity, with published data showing that the proU promoter is sensitive to topological changes (6,13,41) and with observations that H-NS influences DNA topology (5-9). The simplest model consistent with available data is that, in the absence of H-NS, the DRE allows the promoter to adopt an architecture leading to maximum promoter activity. When H-NS binds to the DRE it alters the architecture of the promoter region and, hence, represses transcription. These architectural changes could involve DNA flexibility (the proU promoter is activated by increased DNA flexibility; 41) and/or DNA looping. There is accumulating evidence that H-NS distorts DNA, although the nature of this distortion is unclear: H-NS can constrain DNA supercoils in vitro (5), it can act at a distance in vivo (13,17,28), and it generates DNase I-hypersensitive sites on binding (5,13,16). Although proU is simpler than other H-NS-dependent promoters, in that it does not appear to require additional trans-acting factors, there is no a priori reason why similar mechanisms should not operate at many other H-NS-dependent promoters; positive regulators may alter local promoter architecture to overcome H-NS repression (42).
Most H-NS-dependent genes are regulated in response to environmental signals. DNA curvature changes significantly in response to altered environmental conditions (19), 2 and it is tempting to postulate that, at simple promoters such as proU, the structure of the DRE itself may act as an "environmental sensor"; environmentally induced changes in DNA configura- tion influence the action of H-NS, leading to an alteration in promoter configuration and, hence, transcription. Furthermore, as different DNA sequences respond to different environment cues (19), 3 it is possible to envisage a mechanism by which H-NS can regulate specific genes in response to different environmental signals. The possibility that environmental signals may directly affect promoter architecture and hence transcription, through influencing the interaction of architectural proteins with DNA, may be an important concept in understanding the environmental regulation of gene expression in bacteria.