IHF-binding sites inhibit DNA loop formation and transcription initiation.

Transcriptional activation of enhancer and sigma(54)-dependent promoters requires efficient interactions between enhancer-binding proteins (EBP) and promoter bound sigma(54)-RNA polymerase (E sigma(54)) achieved by DNA looping, which is usually facilitated by the integration host factor (IHF). Since the lengths of the intervening region supporting DNA-loop formation are similar among IHF-dependent and IHF-independent promoters, the precise reason(s) why IHF is selectively important for the frequency of transcription initiation remain unclear. Here, using kinetic cyclization and in vitro transcription assays we show that, in the absence of IHF protein, the DNA fragments containing an IHF-binding site have much less looping-formation ability than those that lack an IHF-binding site. Furthermore, when an IHF consensus-binding site was introduced into the intervening region between promoter and enhancer of the target DNA fragments, loop formation and DNA-loop-dependent transcriptional activation are significantly reduced in a position-independent manner. DNA-looping-independent transcriptional activation was unaffected. The binding of IHF to its consensus site in the target promoters clearly restored efficient DNA looping formation and looping-dependent transcriptional activation. Our data provide evidence that one function for the IHF protein is to release a communication block set by intrinsic properties of the IHF DNA-binding site.

INTRODUCTION

The expression of σ54-dependent promoters requires efficient communication between enhancer and promoter via DNA loop formation, and in most cases the distance between enhancer and –12 promoter region is around 100–140 bp (1–3). It is generally believed that protein induced DNA bending should be used to lower the energetic cost of DNA looping and to facilitate the enhancer–promoter communication (4–6). For example, the regulatory protein NifA from Klebsiella pneumoniae binds to enhancer and activates σ54-dependent transcription from the nifH promoter; in addition, the integration host factor, IHF, binds between the nifH promoter and enhancer (7). IHF bends the DNA in the nifH promoter regulatory region and greatly stimulates NifA-mediated activation of nifH transcription in vitro and in vivo. NifA and IHF are functionally synergistic. IHF induces DNA bending up to 160° (6) and facilitates DNA looping formation that leads to productive interactions between NifA and RNAP-σ54 holoenzyme which results in transcription initiation (7). An IHF-induced bend is also needed for the DNA looping at glnHp2, Pu and pspA (8–11). However some σ54 promoters such as glnAp2 and nifLA have no obvious IHF-binding site and can transcribe well in the absence of IHF in vitro (9,12), indicating that the DNA looping can occur without any DNA bender in vitro. The distance between enhancer and promoter is similar for IHF-dependent and IHF-independent promoters, and why promoters lacking an IHF-binding site can achieve DNA looping in the absence of IHF in vitro is not fully understood. It was proposed that IHF-independent promoters such as glnAp2 and nifLA contain an intrinsic DNA bend (9). However, most IHF-dependent promoters also contain several sequences AAAA, AATT and TTTT (Table 1), each of which can contribute to an 18–19° bend (13). Therefore, the following two issues remain: (i) whether an intrinsic bend between enhancer and promoter is sufficient for DNA loop formation; (ii) what is the exact role of IHF in DNA looping.

Table 1.

Sequences of the native σ54-dependent promoters and their derivatives

The native (as previously reported) or introduced IHF-binding sites (if present) were boxed and the consensus sequences were shadowed. The centre of the IHF binding site is the first nucleotide of the consensus sequence. The IHF site of the nifJ promoter is located in the opposite orientation. The –24 regions of σ54-dependent promoters were double underlined. The IC-constructs in this table were constructed by inserting IHF-binding site at precise positions of E. coli glnAp2 promoter. The whole sequences of these fragments and other sequences used in this study were shown in Supplementary Table 1. If not specifically indicated, each fragment used in this study is 216 bp in length and contains the entire enhancer and core promoter. After digestion with EagI and ligation, these fragments can form 210-bp circular monomers in the ligation reactions.

IHF can recognize its binding sites by both direct and indirect readout. It has been proposed that IHF-binding sites (although some of them are not consensus) have common DNA structural properties, allowing them to be recognized by IHF through indirect readout (5). Therefore, it is reasonable to explore whether the DNA structural properties of IHF-binding sites can themselves affect looping. In this study, we measured the looping-formation ability of the DNA fragments from eight native promoters (six of them contain IHF-binding sites, and two of them not) and one generic control DNA, using the ligase-mediated cyclization method (14–17). Results showed that the DNA fragments containing an IHF binding site have much less looping-formation ability than those having no IHF-binding site. When an IHF consensus-binding site (IC site, the H′ site of phage λ) was introduced into glnAp2 and generic DNA, it strongly inhibited loop formation. Furthermore, in vitro transcription assays were carried out using wild-type glnAp2 and its IC insertion derivatives. Results show that IHF-binding site can generally inhibit the enhancer-dependent expression of glnAp2 in a position-independent manner in the absence of IHF protein, and the inhibitory role is manifest through reducing DNA looping between enhancer and core promoter. In agreement with previous studies, IHF-binding to the IC sites leads to a general restoration of the transcription initiation of all glnAp2 derivatives. Clearly, the IHF target site and by inference its flanking sequences are not passive but have a negative regulatory role in DNA loop formation, and one role of IHF is to restore DNA loop formation.

MATERIALS AND METHODS

DNA

The 216-bp DNA fragments containing entirely the enhancer and promoter regions were amplified by PCR reactions from genomic DNA. PnifB, PnifE, PnifH, PnifJ, PnifLA and Pu were from Klebsiella pneumoniae; PnifH was from Rhizobium meliloti, glnAp2 was from Escherichia coli while the tandem glnAp1 promoter was silenced as in ref. 18 by removing the cAMP receptor protein (CRP)-binding site. All fragments contain EagI at both ends (introduced by PCR primers). The fragments were then cloned into pBR322 individually. The IC-constructs were constructed by inserting IHF-binding site at precise positions of target DNA by site-directed mutagenesis PCR reactions. The sequences are shown in Table 1 and Supplementary Table 1. All plasmids were sequenced and maintained in E. coli DH-5α cells. When these pBR322-derived plasmids were digested by EagI, the small fragments could form 210-bp circular monomers if they were further treated with T4 DNA ligase. The glnAp2 derived base pair insertion or deletion mutants were also constructed by site-mutagenesis PCR reactions. They can form circular monomers from 207 to 213 bp in ligation assays.

The templates for in vitro transcription assays were constructed by inserting the promoters in front of the two strong transcriptional terminators T1 and T2 of the transcriptional vector pJCD01 (19). All promoters were PCR amplified from the corresponding PBR322 derived plasmids. The upstream primer contains a BamHI site and the downstream primer contains a SalI site. Each fragment was cut by BamHI and SalI and the BamHI-SalI fragment was ligated to BamHI-SalI digested pJCD01.

Radioactive labeling

The mixture of two EagI-digested fragments, about 210 bp and the rest of the plasmids, was labeled by incubation with [γ−32P]-ATP and T4 DNA polynucleotide kinase (New England Biolabs). Unincorporated ATP was removed by using Sephadex G-50 spin column and the samples were then used for the ligation experiments.

Ligation-mediated cyclization assays

Ligation experiments were carried out with 0.2 pmol 210-bp DNA fragment at 22°C in ligation buffer [50 mM Tris–HCl (pH 7.6), 10 mM MgCl2, 10 mM DTT, 1 mM ATP and 100 µg/ml BSA]. The concentration of T4 DNA ligase (New England Biolabs) was 0.25 U/μl. At specific time intervals, the same amount of reaction mixture was withdrawn from the reaction solution and quenched with 1 μl 0.5 M EDTA and heated (65°C, 10 min). The ligation products were separated and analyzed as described previously (13). When E. coli DNA ligase (New England Biolabs) was used as in Supplementary Figure 4B, the ligation experiments were carried out at 22°C in its suggested ligation buffer.

Protein purification

These were conducted as we have described previously (18).

In vitro single-round transcription assays

Assays were done in buffer A (18) at 30°C using 40 nM Eσ54, 50 nM IHF, 50 nM NtrC-P, 2 nM DNA and 2 mM ATP. Elongation was started by adding a nucleotides-heparin mixture (final concentrations 200 µM CTP and GTP, 20 µM UTP and 2.5 µCi [α-32P]UTP and 120 µg/ml heparin). Relaxed templates were obtained by digesting supercoiled plasmids with AseI for 2 h. When PspFΔHTH was used as the activator, STA buffer with PEG (25 mM Tris-acetate pH 8.0, 8 mM Mg-acetate, 100 mM KCl and 3.5% W/V PEG-6000) was used as the reaction buffer. The transcript of glnAp2 has a length of 121 nt which contain 20 UTPs. The absolute intensities for each of the transcripts from the above assays were quantified on a Molecular Dynamics Storage Phosphorimager using ImageQuant software.

RESULTS

Ligation-mediated cyclization assay for DNA Cyclizability

DNA elasticity as well as the efficiency of DNA loop formation of a particular fragment can be determined by measuring its j-factor, a measure of the efficiency of the DNA cyclization. DNA conformational changes that bring the two self-complementary ends into transient base-paired contact can occur spontaneously (17,20). The j-factor equals the molar concentration of one end of a linear DNA molecule in the vicinity of its own other end in the appropriate angular and torsional orientation. Higher j-factors therefore indicate a higher probability of loop formation, or increased flexibility in DNA conformation while lower j-factors indicate a reduced ability in loop formation, or increased rigidity in DNA conformation.

To obtain j-factors experimentally, we measured the ratio of the covalently formed closed monomer circles, C(t), to the covalently joined dimeric species, D(t), during the early stage of fragment ligation (17). The j-factor value is obtained by extrapolating 2M0C(t)/D(t) to zero ligation time, where M0 is the initial concentration of the fragments.

Two factors need to be considered when choosing the length of DNA fragments for the experiments (13). On one hand, shortening the DNA fragment length increases the sensitivity of the j-factor to the DNA flexibility. On the other hand, the initial DNA concentration should be about the value of j-factor to obtain comparable amounts of C(t) and D(t). One has to deal with a very low amount of DNA if the fragments are too short because the j-factor drops very fast with shortening DNA length. In many cases, the length of DNA fragments was chosen to be around 200 bp (13). To satisfy all the parameters mentioned above, we chose 210 bp as the DNA length, which is sufficient to cover all the enhancer and core promoter regions of the native promoters and contains around 20 DNA helical repeats assuming the helical pitch of DNA is about 10.5 bp.

In this study, eight σ54 promoters and one generic DNA were studied first. The DNA fragments were prepared as described in ‘Materials and Methods’ section and the sequences are given in Table 1. The generic DNA is from λ phage and does not contain any known curved DNA sequences such as AAAA, AATT and TTTT as investigated previously (13). Two of the promoters (glnAp2 from E. coli., nifLAp from K. pneumoniae) as well as the generic DNA do not contain IHF-binding sites while the others contain their native IHF-binding sites.

Typical distributions of ligation products for different ligation times are shown in Figure 1A. The j-factor is determined by quantitative analysis of the gel as shown in Figure 1B. As explained previously (13), the rate of the cyclization is much higher for the dimers than for the original fragments because the j-factor for 420-bp fragments is much higher than that for 210-bp fragments. As a result, the majority of the dimers are present in circular form except at the very beginning of the ligation reaction. When D(t) is determined, it included both linear and circular dimers to make more reliable the extrapolation to zero reaction time. As shown in Supplementary Figure 4A and in agreement with previous observations (21), the j-factor value does not depend on the concentrations of DNA ligase within the appropriate range. Moreover it does not vary using either T4 or E. coli DNA ligases, suggesting that it is independent of the type of DNA ligase.

Figure 1.

Determination of j-factor from ligation time course. (A) A typical gel shows the LM (linear monomers), LD (linear dimers), CM (circular monomers) and CD (circular dimers) obtained by ligation of the 210-bp DNA fragments. PhosphorImaging was used to quantify the intensities of the bands. The bands at the top of the gel correspond to the ligation products of the large fragment of the plasmid DNA. (B) The j-factor is obtained by extrapolating the ratio 2M0C(t)/D(t) to zero reaction time. Both LD and CD were included in D(t).

As shown in Table 2, the j-factor of the generic DNA is about 7 nM, in agreement with the previous values (13). The j-factor of nifLAp is similar to the generic DNA, while that of glnAp2 is higher. Significantly, the j-factors of all promoters containing IHF-binding site are much lower than the generic DNA (3–9-fold lower). These results suggest that IHF-binding sites might play a role in decreasing the flexibility of the DNA fragments containing them.

Table 2.

Measurements of j-factor (nM) of native promoters and generic DNA

Generic DNA	6.9 (0.3)
Generic-IC	1.9 (0.4)
K.p nifLA	7 (0.5)
K.p nifB	0.8 (0.2)
K.p nifE	1.1 (0.1)
K.p nifH	1.15 (0.2)
K.p nifJ	1.2 (0.1)
K.p nifU	1.5 (0.5)
R.m nifH	3.2 (0.2)
E.c glnAp2	11.5 (0.5)
IC-44	3.8 (0.3)
IC-50	2.1 (0.2)
IC-53	2.2 (0.2)
IC-54	3.5 (0.3)
IC-55	2.9 (0.2)
IC-57	1.8 (0.1)
IC-59	1.6 (0.1)
IC-61	4.2 (0.3)
IC-64	3.6 (0.2)
IC-69	2.7 (0.2)
IC-55&87	1.1 (0.1)
nifU Mutant-1	3.0 (0.3)
nifU Mutant-2	2.4 (0.2)

All promoters contain an IHF-binding site except glnAp2, nifLAp, generic DNA and two nifU mutants. The 210-bp generic DNA was chosen from λ DNA from nucleotide 29 853–30 057. The location of the center of the IHF-binding site has been moved from –50 to –69, in the series of IC-glnAp2 derivatives. IC-55&87 contains two IHF-binding sites centered respectively at –55 and –87. In NifU mutant-1 and 2, the IHF-binding site of nifU promoter was replaced by DNA from generic DNA or glnAp2. The j-factor value represents the average of three independent assays, and the standard deviations are shown in the parenthesis. E.c, Escherichia coli; K.p, Klebsiella pneumoniae; R.m, Rhizobium meliloti.

To test whether IHF-binding site can affect the cyclizability of a given DNA fragment, we replaced the middle part of the generic DNA by a 22-bp IC site. The construct was named generic-IC (sequence as in Supplementary Table 1). Here, this 22-bp IC sequence was chosen from the H′ site of phage λ (5′-ttgctTATCAAtttgTTGcacc), one of the best-characterized IHF-binding sites (22, 23) which is fully consistent with the IHF consensus binding sequence WATCARXXXXTTR (W is A or T; X is A, T, C or G; R is A or G) (6). As shown in Table 2, the IC insertion decreases the j-factor of generic DNA by 3.6-fold, from 6.9 to 1.9 nM. As described previously (13), this generic DNA contains no known curved short DNA elements. Its j-factor is similar to the values from other reported ‘typical’ DNA fragments with similar length and can be viewed as a standard j-factor value for a B-type DNA. Insertion of an IC sequence into this generic DNA significantly decreases its j-factor.

To further test whether an IHF binding site could affect loop formation between enhancer and core promoter in a native context, we replaced part of the glnAp2 intervening DNA region between enhancer and core promoter by the same 22-bp IC sequence. The IC is centered at –55 and the construct was named IC-55 (sequence as in Table 1). As shown in Table 2, the j-factor of glnAp2 dropped from 11.5 to 2.9 nM. To further verify the effects of IHF binding site on the DNA looping formation, we inserted a second IC sequence into the constructed IC-55. The second IC sequence is centered at –87 and the new construct was named as IC-55&87 (sequence as in Supplementary Table 1). The j-factor then decreased from 2.9 (IC-55) to 1.1 nM (IC-55&87) (Table 2).

Although the proposed intrinsic DNA bending sequence of glnAp2 (CCCTTTT) is only 7-bp long and centred at –73 (9), it is possible that other intrinsic DNA-bending sequences are located around –55, and the decreased j-factors of IC-55 and IC-55&87 may be due to the loss of an intrinsic DNA-bending sequence. In order to eliminate this possibility, we constructed nine glnAp2 derived promoters containing the 22-bp IC at different locations between the enhancer and the core promoter, without changing the linear distance between these two elements (for sequences, see Table 1). In total, these 11 IC-constructs have respective mutations between the enhancer and core promoter of glnAp2, occurring every precise location. Since wild-type glnAp2 does not contain phased A tracts which are intrinsically bent, it is very unlikely that the entire DNA sequence between enhancer and core promoter is required to direct intrinsic DNA bending. Therefore, these IC-constructs can be used to identify whether the inhibition of the DNA loop formation is a direct effect of the IHF-binding site. If IC-insertions decrease the j-factor by replacing intrinsic DNA bending element, one would expect the inhibitory effect only happens at specific position(s). On the other hand, if the inhibition is a direct effect due to the physical properties of IHF-binding site itself, it should be position independent.

As shown in Table 2, every IC-construct has a much-reduced j-factor compared with glnAp2 fragment. For the constructs containing one IC sequence, the measured j-factor varies from 1.6 (IC-59) to 3.8 nM (IC-44) while the wild-type glnAp2 has a j-factor 11.5 nM. Since all the j-factor values of IC-constructs are lower than the generic DNA, results are unlikely to be due to a replacement of an intrinsic bending sequence. Rather, results indicate that the inhibitory effect of IC on the DNA loop formation could dominate over the stimulatory effect of intrinsic bend on the DNA looping, likely a direct effect due to its special physical properties.

If the IHF-binding site inhibits the DNA loop formation in native promoters, the substitution of the IHF binding site with other DNA sequences is expected to increase the j-factors of the target DNA fragments. In order to test this hypothesis, the IHF-binding sites of K.pneumoniae nifU promoter and E.coli glnHp2 were mutated (sequences as in Supplementary Table 1). Table 2 shows that these mutations increased the j-factor of the DNA containing nifU promoter (from 1.6 to 2.4 or 3.0 nM), indicating that the native IHF-binding site sequence reduced the ability of DNA loop formation. However, all of the four kinds of mutations of glnHp2 did not increase its j-factor as much (from 1.1 to up to 1.8 nM, sequence as in Supplementary Table 1). The restoration of j-factor is clearer on nifU derivatives compared with glnHp2 derivatives. In both cases, even the j-factors of the mutated sequences are lower than the generic DNA, indicating that the DNA segments flanking IHF-binding site may also contribute to DNA loop formation.

Since IHF-binding sites might have special secondary structural architecture used to be readout indirectly by IHF protein, it is possible that IHF-binding sites have specific helical pitch other than typical B-type DNA. If the helical pitch of the 22-bp IHF-binding site could be as high as 12.0, the DNA length needed to form 20 helical repeats will increase to 213 bp; if the helical pitch of the 22-bp IHF-binding site could be as low as 9.0, the DNA length needed to form 20 helical repeats will decrease to 207 bp. Since DNAs longer or shorter than an optimal length need to under- or overtwist to allow end-base pairing, they will cyclize with increased energetic cost and decreased equilibrium probability (17,23,25), leading to lower j-factor compared to the DNA with optimal length. To exclude the possibility that the IHF-binding site changed the optimal length for monomer circles formation, we measured j-factor for two sets of five fragments from glnAp2 and IC-55. The longest fragments consisted of 213 bp and the shortest fragments consisted of 207 bp. They were constructed by adding or removing one, two or three base-pairs from the end of the glnAp2 or IC-55 (sequences as in Supplementary Table 1). As shown in Figure 2, the peak of j-factor from the glnAp2-derived set of DNA locates at 210 bp, in agreement with the proposal that the helical pitch is 10.5. The peak of j-factor from the IC-55-derived set of DNA is slightly different from the glnAp2 set and locates around 209 bp. However, all j-factors from the IC-55-derived set of DNA are still much lower than the one of original glnAp2, indicating that the decreased cyclization of IC-55 compared to native glnAp2 is not due to the change of the DNA helical pitch. It is interesting to notice that the j-factor of any IC-55 derivative is always 2–4-fold lower than the one from a same size glnAp2 derivative, the reason is not clear but is very likely due to the specific physical properties of the IHF-binding site itself.

Figure 2.

Dependence of j-factors on the length of DNA fragments derived from wild-type glnAp2 and IC-55. glnAp2-derived fragments were marked by open square and IC-55-derived fragments were marked by filled square. Value represents the average of three independent ligation assays, and error bars are shown.

In vitro transcription assays

Ligation-mediated cyclization assays showed that IHF-binding site could inhibit the DNA loop formation when located between enhancer and core promoter of the glnAp2. We then checked whether the reduction in DNA looping formation could decrease transcription initiation at the glnAp2 promoter. We first examined whether the binding of RNAP to core promoter and the binding of EBP to enhancer are affected by the introduction of IHF-binding sites on glnAp2 derivatives. Results obtained from DNase I footprints and gel shift assays showed that IHF protein binds to its binding site and the closed complex formation as well as activator binding were unaffected on IC-constructs (see Supplementary Figure 1 for sample gels, and data not shown), providing us the opportunity to investigate the effect of an IHF-binding site on open complex formation at glnAp2 IC constructs.

In vitro single round transcriptional assays with glnAp2 derivatives were carried out on supercoiled and relaxed templates (26). When Nitrogen Regulatory protein C-phosphate (NtrC-P) was used as the activator at low concentrations, it needs to bind enhancer DNA and contact the promoter bound RNAP-σ54 via DNA looping. The transcription results showed that NtrC-P dependent activities of all supercoiled IC-constructs in the absence of IHF protein were reduced to 15% to 21% compared to the wild-type glnAp2, regardless the location of the IHF-binding site (Table 3, and Supplementary Figure 2A). On relaxed templates, under the same conditions, transcripts could hardly be detected from the IC-constructs (0–4% compared to the expression from supercoiled wild-type glnAp2 template, see Table 3, and Supplementary Figure 2B). As a control, transcription from relaxed wild-type glnAp2 template is about 10% compared to its supercoiled counterpart (Table 3, and Supplementary Figure 2B). Taken together, the presence of IHF-binding site reduces NtrC-P mediated transcriptional activation of glnAp2 derivatives on both supercoiled and relaxed templates. The effect observed is independent of the location of IHF-binding site.

Table 3.

Quantification of enhancer-dependent, and enhancer-independent, single round transcription on supercoiled or relaxed DNA templates

	Supercoiled template				Relaxed template
	NtrC-P		PspFΔHTH		NtrC-P
	−IHF	+IHF	−IHF	+IHF	−IHF	+IHF
glnAp2	100	112	100	93	10	27
IC-44	15	36	83	35	2	130
IC-50	20	36	75	73	2	40
IC-53	15	34	60	65	1	24
IC-54	15	60	95	75	0	127
IC-55	21	84	130	73	0	140
IC-57	19	66	80	78	2	25
IC-59	22	33	75	75	2	33
IC-61	20	55	76	75	4	120
IC-64	17	82	105	45	3	140
IC-69	20	94	90	92	2	115

Transcription initiation was activated by 50-nM enhancer-dependent activator NtrC-P or 160-nM enhancer-independent activator PspFΔHTH. When present, the concentration of IHF was 150 nM. The transcription activities of supercoiled glnAp2 template are the same using 50 nM NtrC-P or 160 nM PspFΔHTH and the absolute value under these two conditions was taken as standard (100%). The relative percentages of transcription activities under other conditions were shown. All experiments were at least triplicated and the standard deviation of each set of experiments is <15% of the mean. The sample transcription gels were shown in Supplementary Figure 2.

To further verify that IHF-binding site inhibits the transcription initiation via inhibiting DNA loop formation, the DNA-binding defective form of PspF (PspFΔHTH), a homologue of NtrC (18,27), was used in the transcription assays. Importantly, PspFΔHTH cannot bind to the enhancer and it can activate transcription initiation via direct contact with the closed complex from solution in a DNA-looping-independent manner. Results showed that, in the absence of IHF, PspFΔHTH-mediated transcriptional activities of IC-constructs were at the same level as that from the wild-type glnAp2 on supercoiled and relaxed templates respectively (Table 3). The NtrC-P and PspFΔHTH results indicate that the presence of consensus IHF binding site, regardless of its location, inhibits DNA loop formation between enhancer and promoter and prevents enhancer–promoter communication at glnAp2 derivatives. As the inhibitory effect is independent of the IHF-binding site's location, the flanking DNA sequence does not play the major role in preventing DNA loop formation, strongly indicating that the IHF-binding sequence itself is the negative regulatory element in this case. Taken together, our data using purified components demonstrate that the IHF-binding site can itself inhibit enhancer-dependent transcription initiation.

The role of IHF-mediated DNA bending

The j-factors of the 210-bp DNA fragments containing the enhancer and promoter of IC-55 and native glnAp2 were measured in the presence of different concentrations of IHF. Results showed that the cyclization of IC-55 is increased up to 3-fold by IHF while the wild-type glnAp2 is not affected by IHF (data not shown). This indicates that IHF restores the cyclization of DNA fragment containing its target site and helps the DNA loop formation, in agreement with the previous observation that short fragments bent by IHF readily form a DNA loop or circular monomer (28,29). One simple prediction is that IHF should stimulate transcription activation of the IC-constructs.

To test the prediction, single-round transcription assays of IC-constructs together with their native glnAp2 counterpart were carried out on both supercoiled and relaxed templates respectively. The results showed that, regardless the location of the IHF-binding site, IHF provided a general restoration of enhancer-binding (NtrC-P)-dependent transcriptional activation of all the IC-constructs, when compared with their native glnAp2 counterpart (up to 5-fold activation for supercoiled templates and up to 100-fold for relaxed templates, see Table 3 and Supplementary Figure 2).

DISCUSSION

Transcriptional activation at σ54-dependent promoters requires looping of the upstream DNA between enhancer and core promoter to allow productive contacts between enhancer bound activator and promoter bound RNAP-σ54. Therefore, the loop formation ability of the DNA upstream of the RNAP binding site at the promoter determines the efficiency of DNA loop formation in the absence of a DNA bender such as IHF.

It has been proposed that IHF-independent promoter glnAp2 contains an intrinsic DNA-bending sequence CCCTTTT centered at –73 (9). Here, the j-factor of glnAp2 is higher than the generic DNA, supporting this proposal. This DNA fragment is essential for in vitro transcription with relaxed glnAp2 template, but plays no role when transcription is initiated from supercoiled glnAp2 template (9). In agreement, when Huo et al. (18) replaced the native glnAp2 fragment at a series of precise positions between enhancer and promoter by a fixed size (22 bp) CRP consensus binding site, the transcriptional activity was not affected by the DNA replacement in the absence of CRP with all supercoiled glnAp2 derivatives both in vitro and in vivo. Since the distance between enhancer and promoter is similar for IHF-dependent and IHF-independent promoters, some unknown elements among those IHF-dependent promoters must inhibit the DNA loop formation in the absence of IHF.

In this study, the cyclizability of the 210-bp DNA fragments from eight native promoters, 10 glnAp2-derived promoters containing an IHF-binding site, one generic DNA and several IHF binding site variants were measured by the ligase-mediated cyclization method. Results showed that the upstream DNA from IHF-dependent promoters has much less cyclizability than those from IHF-independent promoters, suggesting that the overall cyclizability of the intervening region between enhancer and promoter is a key factor in determining whether the promoter is IHF dependent or not. Furthermore, for the first time our results showed that the IHF-binding site itself plays a negative regulatory role in DNA loop formation. Therefore, an IHF-binding site could block the enhancer-promoter communication in the absence of IHF whereas the presence of IHF provided a general restoration of the DNA loop formation and so increased enhancer–promoter communication. The reason why IHF-binding site itself inhibits the DNA loop formation is not clear. One possibility is that the IHF-binding site has a strong distortion or writhe [based on the reports that DNA deformation energy is low for IHF-binding sites (30,31)], which guides its flanking DNA fragments into different planes, potentially reducing enhancer–promoter communication.

To establish the relevance of our in vitro results to in vivo promoter activities we also investigated the effect of IHF-binding site on the activities of glnAp2 IC derivatives in vivo. In both himA (IHF subunit alpha) and himD (IHF subunit beta) minus strains, the activities of the glnAp2 derivatives are about 60–80% of the native glnAp2 (see Supplementary Figure 3), not as greatly reduced as in vitro. On the other hand, the in vivo binding of IHF protein to the IHF-binding sites generally increases the promoter activities of all IC-constructs as observed in vitro, although an up- and down regulatory pattern was observed (see Supplementary Figure 3). The precise reasons for the difference between in vivo and in vitro results are not clear. One possibility could be that the DNA segments flanking consensus IHF-binding site play an important, but undefined role in vivo. Another possibility could be that other histone-like proteins, such as HU (a protein in the same protein family of IHF) may display a partial complementation of the DNA-bending functionality of IHF. The partial overlapping functionality of chromatin-related proteins has been observed previously (9), making their functional roles difficult to demonstrate genetically. Although HU is not reported to be able to replace IHF for activating the expressions of σ54-dependent promoters in vivo, HU protein might partially replace IHF for the general restoration effect of the glnAp2 derivatives under in vivo conditions (see Supplementary Figure 3). Biochemistry and in vitro assays are likely more powerful tools to gain insight into the precise mechanisms associated with these chromosome organising proteins.

At some σ54-dependent promoters, IHF can also work as an active suppressor (restrictor) of the promiscuous activation (cross-activation) by heterologous EBPs, and thus increase the specificity and fidelity of these promoters (32,33). In this study, we noticed that when IHF binding was centered at –44 and –64, it repressed the PspFΔHTH-dependent expression of glnAp2 IC-constructs by a factor of 2 (Table 3). Since the insertion of IHF-binding site at these two positions does not affect the closed complex formation (data not shown), our results might be explained by the proposal that the sharp IHF-induced DNA bending could direct a segment of upstream DNA to occlude the access of promiscuous EBP to the promoter bound RNA polymerase (32,34).

Taken together, we propose that one role of IHF at σ54-dependent promoters is to antagonize the negative regulatory effects of IHF target sites and their flanking DNA sequence in DNA loop formation, thus providing a general restoration of the productive enhancer–promoter communication. One extension of our results is that the antagonism between IHF and the DNA segments it binds puts the expression of most σ54-dependent promoters under the control of the intracellular concentration of IHF protein, which is only accumulated maximally in stationary phase.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.