A cis-spreading nucleoprotein filament is responsible for the gene silencing activity found in the promoter relay mechanism.

Transcription-generated DNA supercoiling plays a decisive role in a promoter relay mechanism for the coordinated expression of genes in the Salmonella typhimurium ilvIH-leuO-leuABCD gene cluster. A similar mechanism also operates to control expression of the genes in the Escherichia coli ilvIH-leuO-leuABCD gene cluster. However, the mechanism underlying the DNA supercoiling effect remained elusive. A bacterial gene silencer AT8 was found to be important for the repression state of the leuO gene as part of the promoter relay mechanism. In this communication, we demonstrated that the gene silencer AT8 is a nucleation site for recruiting histone-like nucleoid structuring protein to form a cis-spreading nucleoprotein filament that is responsible for silencing of the leuO gene. With a DNA geometric similarity rather than a DNA sequence specificity, the E. coli gene silencer EAT6 was capable of replacing the histone-like nucleoid structuring protein nucleation function of the S. typhimurium gene silencer AT8 for the leuO gene silencing. The interchangeability between DNA geometrical elements for supporting the silencing activity in the region is consistent with a previous finding that a neighboring transcription activity determines the outcome of the gene silencing activity. The geometric requirement, which was revealed for this silencing activity, explains the decisive role of transcription-generated DNA supercoiling found in the promoter relay mechanism.

DNA supercoiling has been known to play important roles in transcriptional regulation (1)(2)(3)(4)(5)(6)(7). By using a bacterial transcription regulation model system, we have demonstrated that transcription-generated DNA supercoiling is a crucial driven force that triggers the sequential activation of genes in the Salmonella typhimurium ilvIH-leuO-leuABCD gene cluster (8 -12). This rather complex sequential gene activation process was named the promoter relay mechanism (11,12). The exact molecular detail that underlies the effect of transcription-generated DNA supercoiling on the sequential activation of genes at this locus remains unclear. The direct DNA supercoiling effect on activating promoters of genes in this region has been ruled out. Instead, the effect appears to mediate through cis-ele-ments within the locus control regions (LCRs 1 illustrated in Fig. 1) located between genes in the ilvIH-leuO-leuABCD gene cluster (8).
Although not ruling out the possible involvement of other cis-acting elements in the transcription regulation, we have identified two cis-elements in the LCR-I that are important for the promoter relay mechanism as follows: a bacterial gene silencer, termed AT8; and a LeuO protein-binding site, termed AT7 (13,14). The bacterial gene silencer AT8-mediated transcriptional silencing is integral to the gene expression regulation and is responsible for the repressed state of the leuO gene. LeuO protein-mediated derepression, which relieves the repression of leuO gene, is also a crucial part of the promoter relay mechanism. Transcription-generated DNA supercoiling is likely to play its roles in the promoter relay mechanism via modulating the processes of the repression-derepression control (10). To better understand the DNA supercoiling effect, we investigated the basic molecular criteria of the repression element (gene silencer) and derepression element (LeuO-binding site). We revealed that it is the geometric features of the DNA rather than the specific sequences of the transcription elements that are important for their transcription regulatory functions. The striking DNA geometric requirement is consistent with the involvement of transcription-generated DNA supercoiling in the transcription regulatory process.
The revealed DNA geometric features of the gene silencer also provided clues for the possible involvement in the transcription regulation of chromosome architectural proteins (e.g. HU, H-NS, Lap, and IFS, etc.) that usually recognize DNA structure rather than specific DNA sequence for their bindings (15)(16)(17). Indeed, a genetic screening has led to the identification of a histone-like nucleoid structuring protein (H-NS) for its role in the gene silencing mechanism. Functionally, we demonstrated that the gene silencer AT8 is indeed an H-NS nucleation site that triggers the formation of a nucleoprotein filament structure in the region. With the assistance of the neighboring AT-rich DNA in LCR-I, the transcriptional repressive nucleoprotein structure reaches (cis-spreads) to the promoter region of the leuO gene and results in the repression of the gene. Despite the low DNA sequence homology, the Escherichia coli gene silencer EAT6, but not a same size neutral DNA sequence, was capable of replacing the S. typhimurium gene silencer AT8 for its H-NS nucleation function in S. typhimurium LCR-I. Therefore, it is clear that regardless of the low DNA sequence specificity, either the S. typhimurium gene silencer AT8 or the E. coli gene silencer must provide the crucial DNA geometry for triggering the H-NS nucleation. The recruited H-NS, along with other nucleoproteins, appear to form a cis-spreading nucleoprotein filament that represses the activities of promoters located within the proximity.

EXPERIMENTAL PROCEDURES
Plasmids and Bacterial Strains-Plasmid constructs: pAO, pEV101, and pWU204 have been described previously (12,18). A 75-bp DNA (positions ϩ48 to Ϫ27 of ilvIH), including the Ϫ10 sequence of the promoter of ilvIH, was deleted from pWU204 and resulted in pWU205; otherwise pWU205 is identical to pWU204.
The 393-bp E. coli LCR-I was generated using PCR. Primers 5Ј-GTCAACCCTGACGTCATAAAAACGTCC-3Ј and 5Ј-GAATGAGT-CATTTACGACGTCATAATAATCCATAATG-3Ј were used in the PCR. The primers contain mismatches (the underlined sequences) for producing AatII restriction sites on both ends of the PCR product. The AatII-digested E. coli LCR-I was inserted into the unique AatII site on pAO. Other pAO-based testing plasmids were also derived by using similar strategies, and those DNA inserts involved are individually described in the experiments.
The plasmid, pWU802, was derived from pWU804 (8) by deleting a 1.4-kb BamHI-NsiI fragment that includes the coding region of leuO gene and the downstream pilvIH. To replace AT8 DNA in S. typhimurium LCR-I on pWU802 with EAT6 DNA, the following two DNA oligomers were chemically synthesized: 5Ј-taaatatataaattaattattaaata-agcacatttaatcATCATTCACTTG-3Ј and 5Ј-GTGAATGATgattaaatgtgcttatttaataattaatttatatattta-3Ј; the lowercase DNA sequence is EAT6 DNA. The annealed DNA was used to replace the 58-bp HpaI-DraIII fragment containing AT8 in S. typhimurium LCR-I on pWU802. With a similar approach, the following two synthetic DNA oligomers, consisting of a neutral DNA sequence (part of the nucleotide sequences of the coding region of lacZ), were used to replace AT8 on pWU802: 5Ј-aacc-atcgaagtgaccagcgaatacctgttccgtcatagcgataacgATCATTCACTTG-3Ј and 5Ј-GTGAATGATcgttatcgctatgacggaacaggtattcgctggtcacttcgatggtt-3Ј; the lowercase is the neutral DNA sequence used. The procedures resulted in the precise replacement of the AT8 DNA sequence with either the E. coli repression element EAT6 or the neutral DNA sequence to maintain the flanking DNA sequences in LCR-I as intact. For plasmids, pWU802-LA and pWU802-LH were derived from pWU802. DNA oligomers containing the 27-bp lac repressor binding sequence (lac operator) 5Ј-CGGAATTGTGAGCGGATAACAATTTCG-3Јand the flanking restriction sites were synthesized, annealed, and ligated at the AatII site or the HindIII site on pWU802 to generate pWU802-LA or pWU-802-LH, respectively.
Plasmid pWU902OZ was derived from pWU802. The leu-500 mutation (A to G transition) at the Ϫ10 region of the promoter of the leu-ABCD operon was converted back to the wild type Ϫ10 DNA sequence, otherwise the LCR-I DNA, which controls the expression of the downstream leuO promoter, stayed the same as that in pWU802. The entire coding sequence of lacZ gene, along with its upstream ShineDelgarno sequence, was positioned downstream of the transcription-initiation site of the leuO gene so that the leuO promoter can control the expression of lacZ gene product as a reporter for leuO promoter activity.
Primer Extension-Primer extension was carried out as described previously (13) with the following modifications. The following primers were used: (i) 5Ј-TCTGGGTGAGACAAAACAGGAAGGC-3Ј for detecting pbla-mediated transcripts; (ii) 5Ј-GAAACCATTATTATCATGACA-TTAACC-3Ј for detecting E. coli pilvIH-mediated transcripts; (iii) 5Ј-G-CATATAAAATAAGAAAAAGCAAAATGAGTAAAATTCG-3Ј for detecting E. coli pleuO-mediated transcript; and (iv) 5Ј-CGGAAAACATAAA-GACGCTGACAGAGAC-3Ј for detecting S. typhimurium pleuO-mediated transcripts. Each primer extension reaction consisted of 100 g of total RNA and DNA primer(s). Two primers were mixed in the primer extension reactions for simultaneous detection of the bla and ilvIH transcripts in the results as shown in Fig. 1D and the ilvIH and leuO transcripts in the results as shown in Fig. 1E. In other primer extension reactions, only one primer was used for detecting an individual RNA transcript. The primer extension results were visualized and quantified by using a STORM imaging system 840 (Amersham Biosciences). The quantification was normalized against the content of plasmid DNA in each experiment.
Northern Hybridization-The total RNA preparations and the procedures for Northern hybridization were conducted as described previously (11).
Electrophoresis Mobility Shift Assay (EMSA)-As described previously (14), all testing DNA segments were annealed pairs of synthetic DNA oligomers consisting of DNA sequences complementary to each other as illustrated in the figures. All synthetic DNA oligomers were 5Ј-end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase prior to the annealing. The radioactively labeled duplex DNAs (15 pg per reaction; ϳ10,000 cpm/reaction) were then mixed with the purified histidine-tagged S. typhimurium LeuO protein in a binding buffer (40 mM Tris-Cl, pH 8.0; 4 mM MgCl 2 ; 70 mM KCl; 0.1 mM EDTA; and 0.1 mM dithiothreitol). The binding reaction mixtures were incubated at 37°C for 30 min. The mixtures were then analyzed on 5% native polyacrylamide gels in TBE buffer (89 mM Tris borate, pH 8.3; 1 mM EDTA). Results were visualized using a STORM imaging system 840 (Amersham Biosciences).
DNase I Footprinting-PCR was used to generate DNA templates employed for DNase I footprinting studies. Each involved DNA template is described in the legend of each figure. Reaction mixtures (50 l/reaction) containing radioactively labeled DNA (1 ng/reaction equivalent to ϳ20,000 cpm/reaction) plus increasing amounts of either purified (His) 10 -SLeuO or purified H-NS (kindly provided by Dr. Sylvie Rimsky, IGR, Villejuif Cedex, France) were incubated at room temperature for 30 min to a protein-DNA binding equilibrium. After the incubation, 5 l of 1 mM MgCl 2 and 0.5 mM CaCl 2 was added into each reaction for continued incubation at room temperature for 1 min. DNase I (units used per reaction is indicated in each figure legend) was then added to the reactions for a 90-s incubation at room temperature. The DNase I footprinting reactions were stopped by adding 140 l of stop solution consisting of 192 mM sodium acetate, 32 mM EDTA, 0.14% SDS, and 64 g/ml yeast RNA. The samples were then phenol-extracted, ethanol-precipitated, and resuspended in gel loading buffer. To mark precisely the positions of the protein-mediated DNase I protection sites, primers used for PCR were individually used in the DNA sequencing reactions (Sequenase sequencing kit, U. S. Biochemical Corp.) for preparing DNA sequence ladders. By using a chemical cleavage reaction (24), a G ϩ A marker was also prepared from the radioactively endlabeled DNA. Along with one of the position markers (the DNA sequence ladders or the G ϩ A marker), the DNA products prepared from the footprinting reactions were analyzed on 7% acrylamide, 7 M urea denaturing PAGE.
Tn5 Transposon Insertion-mediated Random Mutagenesis and the Reverse Screening Procedure-1 l of EZ::TN TM ͗KAN-2͘] Tnp Transposome TM (Epicenter), which contains the Tn5 transposon and the kanamycin-resistant gene (kan r ), was mixed in a 0.2-cm electroporation cuvette with 40 l of electrocompetent E. coli DH5␣ cells harboring pCH501S-(AT8) 2 that were prepared by serial washes with cold 10% sterile glycerol during the mid-exponential growth phase (A 600 ϭ 0.6). Electroporation was performed in a Bio-Rad Gene Pulser apparatus (set up at 25 microfarads, 200 ohms, and 2.50 kV). After adding 1 ml of SOC medium, the components are 20 g of bactotryptone, 5 g of yeast extract, and 0.5 g of NaCl, pH 7.5, in 1 liter of H 2 O (final concentrations of 20 mM MgSO 4 and 20 mM glucose were added into the medium after autoclaving), the electroporated cells were transferred to a culture tube and incubated at 37°C for 1 h. After incubation, cells were plated on LB agar plates containing 50 g/ml ampicillin and 25 g/ml kanamycin. Once colonies formed, they were replicated onto plates with or without 5% sucrose supplement.
␤-Galactosidase Assay-The ␤-galactosidase assay was measured by hydrolysis of o-nitrophenyl ␤D-galactosidase to produce o-nitrophenol in permeabilized bacterial cells as described previously (25).
Direct Genomic DNA Sequencing-Chromosome DNAs were isolated from the selected Tn5 knock-out strain and used for direct genomic DNA sequencing. Primers that hybridize to the ends of the inserted transposon, Tn5, were used for a bi-directional outward sequencing that reads into the genomic DNA, flanking the Tn5 insert. The obtained DNA sequences were used to pin-point the insertion site on the bacterial genome using the Blast search (NCBI Data Bank). The direct genomic DNA sequencing was performed at the facility of Fidelity Systems Inc. (Gaithersburg, MD).

Genes in the E. coli ilvIH-leuO-leuABCD Gene Cluster Are
Regulated via the Promoter Relay Mechanism-The promoter relay mechanism was found based on the coordinated expression of genes in the S. typhimurium ilvIH-leuO-leuABCD gene cluster (12). By using a plasmid-borne S. typhimurium DNA context in E. coli hosts, we initially demonstrated that transacting protein factors important for the promoter relay mechanism are functionally available in E. coli strains. This is based on the observed activation of a plasmid-borne leu-500 promoter, the hallmark of the promoter relay mechanism, in the heterogeneous (E. coli protein factors acting on S. typhimurium DNA context) assay system (data not shown). The result of this initial test prompted us to directly monitor the mRNAs of ilvIH and leuO in MC4100, an E. coli relA1 strain under an experimental condition that triggers a severe starvation for branched-chain amino acids (b-caa), isoleucine, leucine, and valine, during the log phase of bacterial growth. The leuO gene in E. coli relA1 strain is normally silent and activated in response to the starvation for b-caa during exponential growth in 17-amino acid SSA, a synthetic medium supplemented with all amino acids except the three b-caa (18). The b-caa starvation causes a 2-h growth arrest (the slow-down period between points 2 and 3 shown in the growth curve in Fig. 1) prior to the growth resumption (the growth rate increase after point 3 in the growth curve shown in Fig. 1). We have demonstrated that the LeuO protein is required during the growth stress for cells to resume their growth after the 2-h growth arrest (growth stress) because the leuO knock-out strain, MF1, failed to resume its growth after the arrest (18).
According to the promoter relay mechanism, the production of LeuO in MC4100 cells during growth stress is presumably because of the activation of the leuO gene triggered by the transcription activity of ilvIH promoter. The ilvIH transcription activity-dependent leuO gene activation has been demonstrated previously in S. typhimurium cells that are entering the stationary phase (25); however, this has not been directly demonstrated in E. coli strains. Hence, we monitored the mRNAs of ilvIH and leuO during the growth of MC4100 in 17-amino acid SSA medium. Indeed, both ilvIH and leuO mRNAs were only detectable at the end of the 2-h growth arrest prior to the cell growth resumption (Fig. 1, B and C, lanes 3). Both mRNAs were not detected at the time points during the exponential growth ( Fig. 1, B and C, lanes 1), and at the beginning of the 2-h growth arrest (Fig. 1, B and C, lanes 2). This Northern result indicated that the ilvIH operon is indeed activated during growth stress (2-h growth arrest). The transcriptional activity of the ilvIH operon is expected to subsequently activate the leuO gene via the promoter relay mechanism (11,12).
The co-detection of ilvIH and leuO mRNAs during the 2-h growth arrest can best be explained by the promoter relay mechanism. Therefore, we expected that the deletion of ilvIH promoter activity would also abolish the activity of the leuO gene under the same experimental condition (MC4100 cells under the starvation for b-caa). Hence, the ilvIH transcription activity-dependent activation of leuO was tested using a pair of multicopy plasmids, pWU204 and pWU205, that both carry the E. coli promoter relay DNA sequence (NCBI GenBank/EBI data bank accession number AF106955). These two plasmids are identical except that the ilvIH promoter (-10 sequences of ilvIH operon) has been deleted from the E. coli promoter relay DNA sequence on pWU205, whereas the promoter of the ilvIH operon on pWU204 remains intact (Fig. 1). The plasmids were tested in MC4100 cells grown in 17-amino acid SSA for the activation of the ilvIH promoter in response to the starvation for b-caa ( Fig. 1, D and E). By using a primer extension study, ilvIH promoter activity was detectable on pWU204 (Fig. 1D, lane 1). Under the same testing condition, ilvIH promoter activity was not detectable on pWU205, the ilvIH promoter-less plasmid (Fig. 1D, lane 2). The activity of the promoter of the ␤-lactamase gene (bla) was simultaneously monitored as an internal control in the primer extension studies (Fig. 1D) and found to be identical. This internal control clearly shows that the different ilvIH promoter activities on pWU204 and pWU205 are not due to copy number differences between the two testing plasmids. Further primer extension studies (Fig.  1E) demonstrated that the ilvIH promoter activity is required for activation of the leuO gene. In the presence of the intact ilvIH promoter, the leuO promoter activity was detectable on pWU204 (Fig. 1E, lane 3). In contrast, because of the lack of ilvIH promoter activity on pWU205, the identical leuO gene on that plasmid failed to be activated (Fig. 1E, lane 4). These data clearly support a promoter relay mechanism for the activation of the leuO gene at the ilvIH-leuO-leuABCD gene locus in E. coli.
The Promoter Relay Mechanism Is Regulated by Cis-elements with Little DNA Sequence Specificity-Two cis-elements that are important for the promoter relay mechanism, the 47-bp repression element AT8 and the 25-bp derepression element AT7, were identified in the S. typhimurium ilvIH-leuO-leuABCD gene cluster (14). The two elements are located in the locus control region I (LCR-I) upstream of the divergently arrayed leuO and leuABCD (illustrated in Fig. 1). These two elements are responsible for the repression and the derepression states of the leuO gene as an integral part of the promoter relay mechanism. Because the promoter relay mechanism is also responsible for the transcriptional regulation of the genes in the E. coli ilvIH-leuO-leuABCD gene cluster, we expected to identify similar elements in the LCR-I of the E. coli ilvIH-leuO-leuABCD gene cluster. Therefore, it was puzzling that homologous DNA sequences to the identified S. typhimurium elements AT7 and AT8 could not be found in the E. coli LCR-I. This was, however, consistent with a previous observation that the DNA sequence in the region (LCR-I) is AT-rich but otherwise shares little DNA sequence homology between the two closely related enteric bacteria (26). Because it was impossible to identify the cognate transcription elements in E. coli LCR-I, based on the known DNA sequences of S. typhimurium transcription elements, we searched for transcription elements in E. coli LCR-I based on the known properties of the transcription elements.
E. coli Repression Element-Based on previously established criteria (13,14), we assayed for gene silencing activity (repression element) in the 393-bp E. coli LCR-I (the AT-rich DNA located between the Ϫ20 position and the Ϫ412 position of the E. coli leuO gene) flanked by leuO and leuABCD (illustrated in Fig. 2). The results showed that gene silencing activity is located near the leuO promoter end of the regulatory region, because both EAT1 and EAT3 did not show any significant gene silencing activity in the assay (Fig. 2, lanes 3 and 5). The gene silencing activity was finally narrowed down to EAT6, a 39-bp AT-rich DNA sequence located between Ϫ84 and Ϫ122 positions of the leuO gene. This location is consistent with the possibility that the E. coli repression element EAT6 is also directly responsible for the repression state of the E. coli leuO gene in a manner analogous to that of the S. typhimurium repression element AT8, in control of the S. typhimurium il-vIH-leuO-leuABCD gene cluster (14).
E. coli Derepression Element-Again, by using our previously established criteria (14), we searched for the derepression element (LeuO-binding site) in the region between the transcription start site (ϩ1 position) of the E. coli leuO gene and the repression element EAT6 (illustrated in Fig. 3). Based on the results of EMSA, by tracing down the binding region, a LeuO-binding site, 29-bp EAT16, was identified at a position downstream of the E. coli repression element EAT6 and upstream of the transcription initiation site of the E. coli leuO gene (Fig. 3). A DNase I footprinting experiment (Fig. 4) revealed that the LeuO protein-dependent DNase I protection was indeed initiated at the LeuO-binding site EAT16 (Ϫ38 to Ϫ66 position of E. coli leuO gene) in LCR-I (Fig. 4A,  lane 2). Upon the increase of LeuO protein concentration, the LeuO-mediated DNase I protection was extended toward the regions (zones 1 and 2) flanking EAT16 (Fig. 4A, lanes 3 and  4). Although the LeuO-mediated DNase I protection in the center region of EAT16 is less clear on the complementary strand of the LCR-I DNA (Fig. 4B), a similar DNase I protection in zones 1 and 2 was observed on the complementary strand in the presence of a high concentration of LeuO protein (Fig. 4A, lane 1). Whether the extension of LeuO proteinmediated DNase I protection beyond the LeuO-binding site EAT16 at a very high protein/DNA ratio has any biological significance in the transcription regulation, remains to be tested in the future. Nonetheless, the DNase I footprinting result (Fig. 4) and the EMSA data (Fig. 3) together supported that LeuO binding in E. coli LCR-I was initiated at the LeuO-binding site EAT16. This is consistent with the relative positions of the repression element, derepression element, and the targeting leuO promoter found in the S. typhimurium ilvIH-leuO-leuABCD gene cluster (14). Hence, a similar repression-derepression process may indeed be responsible for the repression and the transient activation (derepression) of the E. coli leuO gene as an indispensable part of the promoter relay mechanism in the E. coli ilvIH-leuO-leuABCD gene cluster (10).
Functional Replacement of the DNA Sequence Heterogeneous E. coli and S. typhimurium Elements in the Repression-Dere- pression Process-The identified E. coli repression-derepression elements shared no obvious DNA sequence homology with the cognate S. typhimurium elements. The DNA sequence homology between the derepression elements is 28%, and the homology between the two repression elements is 31% (Fig.  5A). The similarity observed was that both E. coli and S. typhimurium repression elements are extremely AT-rich (85% A ϩ T for AT8 and 90% A ϩ T for EAT6). The derepression elements, S. typhimurium AT7 and E. coli EAT16, are relatively less AT-rich (about 60% AT). The best possible DNA sequence alignment, allowing gaps for the alignment between sequences (Fig. 5A), revealed some homologous motifs between the cognate S. typhimurium and E. coli elements. However, the limited homologous regions do not explain the transcription activities exerted by the elements because all the identified elements are the minimal DNA sequences that are absolutely required for retaining the transcriptional regulatory functions.
Although the DNA sequence homology is limited, we found DNA geometrical similarity between the elements. In both gene silencers, intrinsic DNA curvatures were predicted using a SDAB computer program that predicts local bending by elastic models that incorporate sequence-dependent anisotropic bendability (27,28). The bend centers of the predicted DNA curvatures of S. typhimurium AT8 and E. coli EAT6 are indicated in Fig. 5A. We also found a similar size DNA palindrome in both gene silencers (Fig. 5A, underlined sequences). We tested whether these transcription elements, with the described DNA geometrical similarity, could functionally replace each other's activity in the repression-derepression process. The repression-derepression regulatory activity was assayed by its effects on pbla, the promoter of the ␤-lactamase gene (bla) on pAO when the LeuO protein was provided in trans (Fig.  5B). Consistent with the previous findings (14), the S. typhimurium repression element AT8 represses the activity of pbla as evidenced in a primer extension assay (Fig. 5B, compare  lanes 1 and 2). The placement of the S. typhimurium derepression element AT7 within the proximity of the S. typhimurium repression element AT8 resulted in the expected transcriptional derepression, thus restoring the activity of pbla (Fig. 5B,  lane 3). Most strikingly, the E. coli derepression element EAT16 was able to replace the function of the S. typhimurium derepression element AT7 to provide almost the same degree of transcriptional derepression (Fig. 5B, lane 4).
The compatibility of the heterogeneous elements for repression was then tested in the following experiments using the E. coli repression element as the primary element (Fig. 5B, lanes 5-7). The E. coli repression element EAT6 represses the activity of pbla (Fig. 5B, compare lanes 1 and 5). The placement of the E. coli derepression element EAT16 within the proximity of the E. coli repression element EAT6 resulted in the expected transcriptional derepression, which restores the activity of pbla (Fig.  5B, lane 6). The S. typhimurium derepression element AT7 was capable of replacing the function of the E. coli derepression element EAT16 to support LeuO protein-mediated transcription derepression (Fig. 5B, lane 7). The nearly perfect repressionderepression functions of the reconstituted heterogeneous groups demonstrated that the heterogeneous transcription elements performed their cognate functions in the repression-derepression process despite the DNA sequence heterogeneity.

The E. coli Repression Element Is Functional for Regulation of leuO Expression in the Context of S. typhimurium LCR-I-In
prior experiments, the activities of the repression-derepression elements were assayed with a reconstituted plasmid DNA context. To confirm the functional compatibility of the transcription elements in a native environment (the DNA context of S. typhimurium LCR-I with the flanking leuO and leuABCD promoters), we replaced the S. typhimurium repression element AT8 on pWU802, either with a same-size neutral DNA sequence (in the case of pWU802NS) or with the E. coli repression element EAT6 (in the case of pWU802ES), and we assayed for the effect of these replacements on the activity of the promoter of the leuO gene (pleuO) by using primer extension (Fig.  6A). As expected, the S. typhimurium repression element AT8 is indeed responsible for the repression of the leuO gene in its native context (S. typhimurium LCR-I), because the replacement of the S. typhimurium repression element AT8 with a same-size neutral DNA abolished the repression of pleuO activity (Fig. 6A, lane 3). Most strikingly, the transcriptional repression was restored when the E. coli repression element EAT6 was used to replace the S. typhimurium repression element AT8 (Fig. 6A, lane 2). This is a strong piece of evidence that the E. coli repression element exerts a reasonable transcriptional repression activity in context of the S. typhimurium ilvIH-leuO-leuABCD gene cluster, despite the lack of apparent DNA sequence homology between the E. coli and S. typhimurium LCR-I. The gene silencing effect appears to extend some distance because the promoter of the ␤-lactamase gene (pbla), which is located 392 bp downstream, was also affected when either the S. typhimurium repression element AT8 or the E. coli repression element EAT6 was present (Fig. 6B).
The results therefore showed that both repression and derepression elements were functionally interchangeable for the control of the expression of genes in the ilvIH-leuO-leuABCD gene cluster. Instead of DNA sequence homology, DNA geometrical similarities may be important for the transcription regulatory functions of the repression and derepression elements. This complex mechanism involving DNA geometrical changes stabilized by transcription-generated DNA supercoiling is likely to be responsible for the repression-derepression process. If so, this is a novel transcription regulatory mechanism that deserves further investigation to define the underlying molecular details. As the first step, we focused on the repression mechanism. Although the cis-acting repression element (gene silencer) has been well characterized thus far, the trans-acting protein factors responsible for gene silencing remain unknown.
A Genetic Approach for the Identification of Genes Required for Bacterial Gene Silencing-A two-step screening procedure involving an initial reverse selection (as described under "Experimental Procedures") followed by a positive screen was used to identify gene(s) important for the bacterial gene silencing mediated by the gene silencer AT8. Bacillus subtilis sacB gene encodes levansucrase, which catalyzes the hydrolysis of sucrose resulting in the synthesis of levans (29). In the presence of 5% sucrose, expression of sacB in Gram-negative bacteria such as E. coli is lethal (30). The E. coli harboring "suicide" plasmid, pCH501S-(AT8) 2 , carrying the coding region of the sacB gene under the control of the leuO promoter, which is repressed by the direct repeat of the bacterial gene silencer AT8, was used in the first step screening for genes that are important for the gene silencing. The AT8 dimer (AT8) 2 provided a very tight gene silencing effect as demonstrated in our previous study (14). Because the promoter of the leuO gene is one of the natural promoters regulated by the gene silencer, expression of the lethal gene sacB on the suicide plasmid is strongly repressed in the presence of (AT8) 2 . To screen for genes required for silencing, Tn5 transposon insertion mutagenesis (as described under "Experimental Procedures") was used to randomly knock-out genes throughout the bacterial genome. If a Tn5 insertion knocked out a gene important for the gene silencing mechanism, E. coli harboring the suicide plasmid would no longer survive on LB plates containing 5% sucrose because of the relief of gene silencing. However, that mutant can still form a colony on LB without sucrose. By using this reverse selection procedure, we obtained 398 clones out of a total 2,296 colonies from this negative screening. Because Tn5 transposon may also target plasmid DNAs, we excluded the clones containing plasmids that carry the Tn5 inserts from the 398 clones. This exclusion resulted in 300 potential positive clones selected from the first-step reverse selection procedure.
Because the reporter gene sacB is very toxic for Gram-negative bacteria used in the reverse selection, we expected many of these 300 selected clones might contain knock-out genes important for sucrose metabolism or sugar transport, rather than genes directly relevant to AT8-mediated gene silencing. A preliminary titration experiment to determine the appropriate concentration of sucrose (titration range, 10 to 0.5% sucrose in LB) to be used for the initial reverse selection procedure had indicated that the sucrose toxicity is very stringent. Whereas 5% sucrose was chosen as a toxicity threshold in the initial reverse selection, we noticed that even a small leakage of the activity of the promoter that controls the expression of sacB gene could result in toxicity in the presence of as little as 0.5% sucrose. To be sure that we did not exclude possible candidate genes, which may have minor effects on the bacterial gene silencing in the first step selection, a relatively high toxicity (5% sucrose) was used for the initial screening that resulted in the 300 potential clones. We then lowered the toxicity to 0.5% sucrose for the second-step screening. According to preliminary testing, we expected that bacterial strains harboring pCH501S-(AT8) 2 should not experience any toxicity at all in LB containing 0.5% sucrose if the Tn5 insertion had affected genes that are irrelevant to the gene silencing mechanism. This is based on the fact that the silencer repeat, (AT8) 2 , is capable of providing sufficient repression (a minimum leakage) on the sacB gene of pCH501S-(AT8) 2 . However, any major relief of the bacterial gene silencing was expected to result in severe toxicity in LB even supplemented with only 0.5% sucrose. Hence, the 300 potential positive clones were grown in LB supplemented with 0.5% sucrose as a second-step screening procedure. Most strikingly, we found that with the exception of one clone, the rest of the 299 clones survived in the culture of LB supplemented with 0.5% sucrose. Although it is possible that some of the 299 excluded clones may be genes that play minor roles in the silencing mechanism, the single positive clone isolated in the second-step screening procedure must contain a knock-out gene crucial for the bacterial gene silencing. This was confirmed using pWU902OZ to report the gene silencing activity in the isolated Tn5 knock-out clone. Plasmid pWU902OZ carries the entire regulatory region (LCR-I) of the S. typhimurium leuO gene and the leuO promoter that controls the expression of the downstream coding sequence of lacZ reporter gene. Because of the effect of the gene silencer located in the LCR-I region, the expression of the reporter lacZ gene is repressed. If the Tn5 insertion had knocked out a gene that was truly important for the silencing activity in the LCR-I, then we expected an increase of the expression of the reporter lacZ gene. Indeed, compared with pWU902OZ harboring DH5␣ (the parental strain of the Tn5 knock-out strain), a 16-fold increase of ␤-galactosidase activity was found in the isolated Tn5 knockout strain harboring pWU902OZ (data not shown).
H-NS Is the Trans-acting Factor Responsible for the Transcriptional Repression Mediated by the Gene Silencers-Direct genomic DNA sequencing was used to identify the site of Tn5 insertion on the chromosome of the positive clone isolated. The DNA sequencing result indicated that the Tn5 insertion is located at the ϩ59 position of the coding region of the hns gene. The insertion is likely to cause either early termination or truncation of the translation product of the gene because it is in FIG. 7. Gene silencer-mediated transcriptional repression is H-NSdependent. The pAO testing plasmid series with or without the insertion of AT4 DNA at the AatII site was used for testing AT4-mediated transcriptional repression in MF1 and SC1, an isogenic hns ϩ /hns Ϫ pair (A). The pWU802 plasmid series used in Fig. 6 was used for testing the S. typhimurium AT8 silencer-mediated or E. coli silencer EAT6-mediated transcriptional repression in the same isogenic hns ϩ / hns Ϫ pair (B). Primer extension was used for detecting the activity of pbla or pleuO in the studies. The quantified data shown at the bottom of each lane for three repeated experiments is expressed as the mean within the range of Ϯ0.03 (S.D.). the N terminus of the coding region. Hence, H-NS, the gene product of hns, is most likely to be the protein factor important for the gene silencing mechanism.
The possibility was further confirmed in experiments using a pair of isogenic hns ϩ /hns Ϫ strains (Fig. 7). Two testing plasmids were used in the experiments. First, two testing plasmids, pAO with or without the gene silencer sequence (AT4), which had been used to characterize the bacterial gene silencer in our previous studies (13,14), were used to confirm the involvement of H-NS in the bacterial gene silencing (Fig. 7A). The results showed that a reduced gene silencing activity was observed when pAO-AT4 was tested in the hns Ϫ strain (Fig. 7A, lane 4) compared with the silencing activity detected in the hns ϩ strain (Fig. 7A, lane 2). The 4-fold reduction on the gene silencer AT4-mediated gene silencing must be due to the absence of H-NS in the hns Ϫ strain because the genetic background of the two bacterial strains is identical except for the hns gene. In a second experimental set, we directly monitored the effects of the gene silencer on the promoter of the leuO gene on the testing plasmid pWU802 series (Fig. 6). The presence of either the S. typhimurium silencer AT8 or the E. coli silencer EAT6 in the S. typhimurium LCR-I was able to repress the promoter activity of the leuO gene on the pWU802 plasmid series (Fig.  7B, lanes 1 and 2). The gene silencing effect was clearly H-NSdependent because the repression of leuO promoter activity was significantly reduced in the hns Ϫ strain (Fig. 7B, lanes 4  and 5). Moreover, the repressive effect of H-NS on the promoter activity of leuO gene must be mediating through either the S. typhimurium silencer AT8 or the E. coli silencer EAT6, because the replacement of the gene silencer with a neutral sequence failed to affect the promoter activity of leuO gene in either testing condition (Fig. 7B, lanes 3 and 6). The gene silencing effect on the relatively distal bla promoter was also H-NS-dependent because a 2-fold reduction on the gene silencing effect was found in the hns Ϫ strain (Fig. 7B, compare lanes  7 and 8 with lanes 10 and 11). The repression activity assay was carried out in the absence of the interference of LeuO protein because the leuO gene was knocked out in the isogenic pair of hns ϩ /hns Ϫ strains used in the experiment.
The Gene Silencer Important for the Promoter Relay Mechanism Is an H-NS Nucleation Site-Because the H-NS-dependent gene silencing activity is absolutely dependent on the presence of either the S. typhimurium gene silencer AT8 or the E. coli gene silencer EAT6 in LCR-I, H-NS must affect the promoter activity with a biological event specifically initiated at the gene silencer. The DNA sequence required for triggering the initiation process of H-NS-mediated gene silencing has been termed the H-NS nucleation site (31,32). Based on this definition, an H-NS nucleation site is distinct from other ordinary H-NS-preferred binding DNA sequences. H-NS nucleation is absolutely required for the formation of a well organized nucleoprotein structure, which is shown to be transcriptionrepressive. Other ordinary H-NS-binding DNA sequences may lead to the binding of H-NS, but in the absence of an H-NS nucleation site, the bound H-NS is not necessarily well organized and hence is not transcriptionally repressive (31).
DNase I footprinting assays were used to investigate whether the identified gene silencer (AT8 or EAT6) is in fact an H-NS nucleation site. The results showed that the S. typhimurium gene silencer AT8 was the first region to be protected by H-NS in the DNase I footprinting experiment (Fig. 8, left  panel, lane 3). With increasing H-NS concentration in the DNase I footprinting experiment (Fig. 8, left panel, lanes 2 and  1), the protected region gradually extended toward the promoter of the leuO gene. A detailed titration of H-NS concentration was carried out to better demonstrate the gradual exten-sion of the H-NS-dependent protection in the DNase I footprinting experiment (Fig. 8, right panel). Clearly, the S. typhimurium gene silencer AT8 is the site occupied by H-NS as the first step. Although the DNase I protection was also slightly extended toward the 5Ј direction (Fig. 8, right panel,  lanes 1-4), the 3Ј end extension was much more significant (Fig. 8, right panel, lanes 1-4), and the DNase I protection eventually reached the promoter of the leuO gene (Fig. 8, left  panel, lanes 1-3). Hence, this DNase I footprinting result demonstrated that the S. typhimurium gene silencer AT8 is indeed an H-NS nucleation site, which initiates the formation of a transcriptionally repressive nucleoprotein structure. This structure has a potential to extend (cis-spread) toward either direction; however, the remaining AT-rich DNA sequence in LCR-I may confer directionality to the cis-spreading nucleoprotein structure. In this case, the transcriptionally repressive nucleoprotein structure preferentially extended toward the promoter of the leuO gene.
Similar H-NS nucleation and unidirectional cis-spreading were also observed when E. coli LCR-I was tested in a similar DNase I footprinting experiment (Fig. 9). H-NS-dependent DNase I protection was also initiated near the E. coli gene silencer EAT6 (Fig. 9, lane 3), was spread in both directions, and eventually extended preferentially toward the promoter of the E. coli leuO gene (Fig. 9, lanes 1-3). Hence, both the S. typhimurium gene silencer AT8 and the E. coli gene silencer EAT6 are important for the promoter relay mechanism and are most likely the H-NS nucleation sites. This possibility was further tested using the replacement plasmid constructs where the S. typhimurium gene silencer AT8 was replaced with either the E. coli gene silencer EAT6 (the case in pWU802ES) or a same size neutral DNA sequence (the case in pWU802NS). The previous functional studies (Figs. 6 and 7B) clearly indicated that EAT6 was able to replace the gene silencing activity of AT8 in the S. typhimurium LCR-I for the repression of the promoter of the downstream leuO gene and even the more distal bla gene. Based on this observed functional replacement, we also anticipated observing similar H-NS nucleation on the replacement plasmid construct, pWU802ES. Indeed, the 39-bp EAT6 DNA sequence was the first region occupied by H-NS as evidenced in the DNase I footprinting experiment (Fig. 10, left  panel). Although the DNase I protection was not as efficient as had been observed in the experiment using the original S. typhimurium LCR-I DNA sequence (Fig. 8), the H-NS-dependent DNase I protection also eventually extended toward the promoter of the leuO gene (Fig. 10, left panel, lane 1). The lower efficiency is most likely due to imperfect compatibility of the E. coli gene silencer EAT6 in the context of S. typhimurium LCR-I DNA. The less-than-perfect compatibility is evidenced on the DNase I-mediated cleavage pattern of the DNA of the replacement plasmid construct pWU802ES. There is a stretch of intensive DNase I-mediated cleavages located near the junction of the 5Ј end EAT6 DNA sequence and the context of the S. typhimurium LCR-I (the dense cleavage pattern shown in Fig. 10, left panel). Nevertheless, the DNase I footprinting result demonstrated that the E. coli gene silencer EAT6 is able to replace the H-NS nucleation function of the S. typhimurium gene silencer AT8 in S. typhimurium LCR-I. The function that is replaced must be H-NS nucleation because the H-NS nucleation was not observed upon the replacement with a same size neutral DNA sequence in the context of S. typhimurium LCR-I in plasmid pWU802NS. Because the gene silencing activity observed in the functional assay (Fig. 6) correlates highly with the H-NS nucleation observed in the DNase I footprinting experiments (Figs. 8 -10), we concluded that the gene silencer AT8 or EAT6, important for the promoter relay mechanism, must be an H-NS nucleation site rather than an ordinary H-NS-preferred binding DNA sequence.
A Cis-spreading Nucleoprotein Filament Is Responsible for the Bacterial Gene Silencing-In addition to the nucleation site, the rest of the AT-rich DNA sequence in S. typhimurium or E. coli LCR-I appears to also possess a certain binding preference for H-NS because, with increasing H-NS concentrations, some of this AT-rich DNA was occupied by H-NS as well (Figs. 8 -10). This H-NS binding preference may be required for the propagation of H-NS polymerization (as the step II suggested in Ref. 33). The AT-rich DNA may also determine the directionality of the transcriptionally repressive nucleoprotein structure. In both cases, the H-NS binding preference led to a cis-spreading preferentially toward the target promoter, the A unique end radioactively labeled DNA segment consisting of the promoter and the upstream region (ϩ40 position to Ϫ269 position) of the S. typhimurium leuO gene was generated by PCR using either pWU802ES or pWU802NS. The silencer AT8 in S. typhimurium LCR-I was replaced with the E. coli silencer in the PCR product when pWU802ES was used as the DNA template (left panel). The silencer AT8 in S. typhimurium LCR-I was replaced with a same size (47 bp) neutral DNA sequence in the PCR product when pWU802NS was used as the DNA template (right panel). The endlabeled DNA was incubated with H-NS at a concentration indicated above each lane and exposed to DNase I (0.5 unit per reaction). The positions of relevant elements and restriction sites in the region are marked.
promoter of leuO gene. Together, these data supported a nucleoprotein filament model for the H-NS effect.
The binding of a foreign protein, such as the lac repressor or repressor within the cis-spreading pathway of the transcriptionally repressive nucleoprotein filament, was capable of blocking the gene silencing effect found in the AT-rich DNA sequences flanking the promoter of the bgl operon (34). To confirm that a continuous cis-spreading nucleoprotein filament is responsible for the silencing activity found in our model system, a lac operator was positioned at the AatII site of the testing plasmid pWU802. Under this testing condition, we expected that the lac repressor should block the gene silencing activity from reaching the downstream bla promoter, whereas the leuO promoter remained as repressed (the model illustrated in Fig. 11) if a continuous cis-spreading nucleoprotein filament indeed simultaneously repressed the activities of both the leuO promoter and the bla promoter on the testing plasmid (as observed in Figs. 6 and 7). Indeed, we found that the presence of a lac operator at the AatII site derepresses the activity of the bla promoter but not the leuO promoter upon providing the trans-acting lac repressor (Fig. 11, lanes 2 and 8). If the inserted lac operator was not positioned between the gene silencer and any one of the two target promoters (lac operator was positioned at the HindIII site of the testing plasmid as shown in Fig. 11, lanes 3 and 9), the lac repressor did not affect the gene silencing effect on either promoter. The binding of the lac repressor at the operator site rather than the inserted operator DNA must be responsible for the derepression of the bla promoter activity. This is concluded because no effect was observed if the lac repressor is not provided (Fig. 11,  lanes 4 -6). Therefore, it is clear that a cis-spreading nucleoprotein filament initiated from the gene silencer AT8 is responsible for the repression of leuO promoter activity. The transcriptionally repressive nucleoprotein filament further extended to the downstream region as a continuous nucleoprotein filament structure (model illustrated in Fig. 11) and resulted in the repression of the downstream ␤-lactamase gene on the testing plasmid. DISCUSSION Clearly, the AT-rich LCRs in the ilvIH-leuO-leuABCD gene cluster are prone to form DNA secondary structures that are functionally important for transcription regulation (the promoter relay mechanism) in the region. Consequently, the DNA geometry rather than the specific DNA sequence of the LCRs was conserved between the two closely related enteric bacteria, E. coli and S. typhimurium. In the present study, the function of one of these conserved DNA geometrical elements was characterized. The characterization led to the finding that an H-NSmediated cis-spreading nucleoprotein filament is responsible for the leuO gene silencing. H-NS is an abundant nucleoid protein important for the architecture of bacterial chromosomes (32,35,36). The protein is known to affect the expression of many genes including the hns gene itself (33,(37)(38)(39). Among them, the silencing mechanisms of proU and bgl genes have been relatively well characterized (34, 40 -44).
It has been known for some time that H-NS binds to DNA in a sequence-nonspecific manner, but it preferentially binds with curved AT-rich DNA (45)(46)(47). Our finding is similar to the gene silencing activity mediated by the AT-rich DNAs found within the proximity of the proU and bgl promoters (34, 48). In these AT-rich DNAs, H-NS nucleation sites and flanking AT-rich DNAs are also important for their gene silencing activities (31,40). In our model system, the flanking AT-rich DNA determines the directionality of the cis-spreading transcriptionally repressive H-NS filament structure toward the promoter of the target gene leuO. This is a perfect example that the exact function of a DNA element (the gene silencer in this case) shall be evaluated in its natural DNA environment (the neighboring AT-rich DNA in LCR-I in this case), as the gene silencer exerts a clear bi-directional transcriptional repressive activity in the testing plasmid pAO series where the gene silencer is surrounded by plasmid DNA context (13,14). The previously demonstrated distance limit (300 bp) for the AT8-mediated gene silencing effect that was determined using pAO plasmids (13) may also be due to the flanking foreign DNA context. In the FIG. 11. A cis-spreading nucleoprotein filament is responsible for gene silencer AT8-mediated transcriptional repression. The plasmid, pWU802, was tested in MC1060, a lac Ϫ strain. The lac repressor was expressed from the pACYC-based pSO1000 in MC1060 when necessary. The activities of the bla and leuO promoters on pWU802 were simultaneously monitored. The quantified promoter activity shown at the bottom of each lane for three repeated experiments is expressed as the mean within the range of Ϯ0.02 (S.D.). Illustrated in the model is the effect of a tetrameric lac repressor located at either the AatII site or the HindIII site for blocking the cis-spreading nucleoprotein filament initiated from gene silencer AT8. present study with the assays performed in the presence of the surrounding LCR-I DNA sequence, AT8-mediated gene silencing was found to be able to extend to such a distance as to affect not only the target leuO promoter but also the promoter of the ␤-lactamase gene, bla, which is located downstream on the testing plasmid of the pWU802 series (Figs. 6, 7, and 11). This could be related to a previous finding that the DNA sequence upstream of the ␤-lactamase gene also shows H-NS binding preference (48). This DNA sequence helps to extend the gene silencing effect to a further distance. Hence, the gene silencer AT8 (an H-NS nucleation site) serves as a crucial driving force for triggering the transcription regulatory effect, but the rest of the DNA sequence in the region determines the distance and the directionality of the effect. This conclusion is basically similar with the finding in a previous study by using the repression of proU or bgl as a model system (31). It seems that nature has evolved the optimal transcription controls in the ilvIH-leuO-leuABCD gene clusters. The optimal transcription controls do not require DNA sequence specificity in the control region for either the E. coli LCR-I or the S. typhimurium LCR-I. The DNA sequence-nonspecific, but functionally interchangeable, DNA elements may similarly be responsible for the transcriptional regulation exerted by the long stretch of ATrich DNA within the proximity of the promoter of proU or bgl.
Studies thus far have clearly demonstrated that DNA geometric similarity rather than the sequence specificity of the gene silencer is responsible for its H-NS nucleation function in this model system. Certain DNA geometric requirements are probably also true for the flanking AT-rich DNA, which determines the cis-spreading (propagation) of the transcriptionally repressive nucleoprotein structure. Transcription-generated DNA supercoiling is very likely to play crucial roles in the sequential activation of genes in the ilvIH-leuO-leuABCD gene cluster via modulating the DNA geometrical changes of ciselements, including the gene silencer in the region. Although the molecular details underlying the DNA supercoiling effect remain to be further explored, the apparent DNA geometric requirement for the H-NS nucleation has provided an explanation for the decisive role of transcription-generated DNA supercoiling found in the promoter relay mechanism (9, 10).