Transcription Initiation at the Flagellin Promoter by RNA Polymerase Carrying ς28 from Salmonella typhimurium *

The ς subunit of RNA polymerase is a critical factor in positive control of transcription initiation. Primary ς factors are essential proteins required for vegetative growth, whereas alternative ς factors mediate transcription in response to various stimuli. Late gene expression during flagellum biosynthesis inSalmonella typhimurium is dependent upon an alternative ς factor, ς28, the product of the fliA gene. We have characterized the intermediate complexes formed by ς28 holoenzyme on the pathway to open complex formation. Interactions with the promoter for the flagellin gene fliCwere analyzed using DNase I and KMnO4 footprinting over a range of temperatures. We propose a model in which closed complexes are established in the upstream region of the promoter, including the −35 element, but with little significant contact in the −10 element or downstream regions of the promoter. An isomerization event extends the DNA contacts into the −10 element and the start site, with loss of the most distal upstream contacts accompanied by DNA melting to form open complexes. Melting occurs efficiently even at 16 °C. Once open complexes have formed, they are unstable to heparin challenge even in the presence of nucleoside triphosphates, which have been observed to stabilize open complexes at rRNA promoters.

The subunit of RNA polymerase is a critical factor in positive control of transcription initiation. Primary factors are essential proteins required for vegetative growth, whereas alternative factors mediate transcription in response to various stimuli. Late gene expression during flagellum biosynthesis in Salmonella typhimurium is dependent upon an alternative factor, 28 , the product of the fliA gene. We have characterized the intermediate complexes formed by 28  Prokaryotic RNA polymerase (RNAP) 1 is a multimeric enzyme consisting of the core subunits, ␣ 2 ␤␤Ј (E), and a factor (1). The factor is responsible for directing the recognition and binding of specific promoter DNA consensus sequences by holoenzyme (␣ 2 ␤␤Ј) and for facilitating transcription initiation (2,3). Many different factors have been identified, and they fall into two major categories, those similar to Escherichia coli 70 and those similar to 54 (4). The 70 family can be further subdivided into primary and alternative factors (5,6). Primary factors are essential and required for the expression of housekeeping genes in vegetative cells, whereas alternative factors are activated in response to environmental changes and in programming developmental pathways (4,6,7). Promoters recognized by the 70 family consist of conserved hexamers at Ϫ35 and Ϫ10 with respect to the start site of transcription (8).
Additional DNA contacts are made at some promoters by the C-terminal domain of the ␣ subunit of RNAP at an AT-rich region between Ϫ38 and Ϫ59 called the UP element (9).
Comparison of the amino acid sequences of members of the 70 family reveals four highly conserved regions with subdomains that have been implicated in specific functions ( Fig. 1) (4). Region 1.1 inhibits 70 from binding to the DNA in the absence of the core subunits (10) and is also required for efficient progression from the earliest RNAP⅐DNA complex to a transcriptionally active complex during initiation (11). Deleting regions 1.1 and 1.2 of 70 results in transcriptional arrest after initial binding of RNAP to the promoter (11). Regions 2.1, 2.2, and 3.2 are important for core binding (12)(13)(14), and region 2.3 has been implicated in promoter melting (15)(16)(17)(18). Both holoenzyme and the factor interact with non-template bases in the Ϫ10 element to stabilize the open complex (19 -21). Regions 2.4 and 4.2 are responsible for contacting the -10 and -35 promoter recognition elements, respectively, and for positioning holoenzyme for initiation (10,(22)(23)(24)(25)(26). Regions 1.2, 2, and 4 are found in almost all factors, but region 1.1 is found only in the primary factors (4). Interestingly, both regions 1.1 and 1.2 are absent in the Salmonella typhimurium alternative factor required for flagellum biosynthesis, 28 , hinting at potential variation in the structure of E 28 ⅐DNA complexes or in the mechanism of transcription initiation.
Initiation can be described as a series of sequential steps, which have been well characterized for both E 70 and E 32 from E. coli (27). By varying the conditions, several intermediates can be visualized using footprinting methods. RNAP (R) binds to the promoter (P) to form an initial closed complex, RP C1 , which protects the DNA from approximately Ϫ60 to Ϫ5 relative to the start point of transcription. RP C1 isomerizes to a second closed complex (RP C2 ) that maintains the upstream contacts and extends further downstream of the transcription start site to ϩ20. RP C2 then undergoes strand opening to form an open complex, RP O , whereas the length of the footprint is unchanged. There is evidence for more than one open complex, which is dependent on the presence or absence of Mg 2ϩ (28,29). RP O enters the initiation stage, or RP init , in the presence of initiating nucleotides. Short transcripts of 2-12 nucleotides in length (abortive products) are synthesized while RNAP remains at the promoter. RNAP then enters into the elongation phase of transcription upon promoter clearance and factor release. Several of these intermediate complexes have been visualized by performing DNase I and KMnO 4 footprinting experiments over a range of temperatures, with the rationale that temperature-dependent complexes represent timedependent events (30 -32).
The S. typhimurium flagellar operon can be divided into three classes of genes (I-III) based on their transcriptional hierarchy in flagellar assembly (33). 28 , encoded by the fliA gene, is expressed late in flagellum biosynthesis and is re-quired for transcription of all class III genes including fliC, encoding flagellin, the primary component of the flagellar filament (34), and flgM, encoding the 28 anti-factor (35,36). In this report, we have characterized transcription initiation by RNA polymerase carrying the S. typhimurium factor, 28 , on variants of the flagellin (fliC) promoter. We detected intermediate complexes formed by E 28 during initiation that are distinct from those characterized for other holoenzymes. Initial binding to the promoter does not require the Ϫ10 element, but further extension of the RNA polymerase-DNA contacts and isomerization to the open complex demand the presence of the Ϫ10 sequence. We propose an alternative mode of promoter recognition and binding as compared with other well characterized holoenzymes.

EXPERIMENTAL PROCEDURES
Overproduction and Purification of 28 and Reconstitution of Holoenzyme-The fliA gene, encoding 28 , was inserted into plasmid pET15b (Novagen, Inc.) to generate pKH439 (a gift from K. Hughes), which resulted in the addition of six histidines at the amino terminus. Hexahistidine-tagged 28 was overproduced and purified using the method described by Wilson and Dombroski (11). Holoenzyme was reconstituted by adding 1.0 pmol of E. coli core RNA polymerase (E) (Epicentre Technologies Corp.) to 5.0 pmol of 28 in protein dilution buffer (10 mM Tris-HCl (pH 8.0), 10 mM KCl, 10 mM ␤-mercaptoethanol, 1 mM EDTA, 0.4 mg/ml bovine serum albumin, and 0.1% Triton X-100) and incubating on ice for 15 min. Because of the amino acid sequence conservation among the core subunits, heterologous holoenzymes have been used to assess the behavior of alternative factors (37)(38)(39)(40). Additionally, we compared the amino acid sequences of the ␣ and ␤ subunits of E. coli and S. typhimurium RNA polymerases (11) (data not shown). There is 100% identity between the ␣ subunits and 98% identity between the ␤ subunits.
Constructions and Generation of Promoter Fragments-pfliC⌬35 was constructed by annealing two oligonucleotides (Integrated DNA Technologies) of 122 bases in length (41). The DNA was designed to incorporate HindIII and BamHI restriction sites near the 5Ј-and 3Ј-ends, respectively. The double-stranded DNA was digested with HindIII and BamHI, ligated into the same sites in pBluescriptII KS ϩ (Stratagene), and transformed into E. coli strain DH5␣ (Life Technologies, Inc.). The clones were sequenced to confirm the deletion using the fmol TM DNA sequencing system (Promega).
pfliC⌬10 was constructed using an oligonucleotide with a deletion of the Ϫ10 element as a polymerase chain reaction primer with plasmid pMC72, containing the wild-type fliC promoter, as the template. The resulting DNA fragment was ligated into pCR2.1 and transformed into Inv␣FЈ One Shot competent cells from the original TA cloning kit (Invitrogen). The clones were sequenced to confirm the deletion using the fmol TM DNA sequencing system. 32 P-5Ј-End-labeled primers were generated for use in synthesizing labeled fliC promoter DNA (10). Oligonucleotide primers were from BioServe Biotechnologies, Genosys Biotechnologies, Inc., or Integrated DNA Technologies. 2 Plasmid pMC72 (a gift from K. Hughes) and the two plasmids containing pfliC⌬10 and pfliC⌬35, as described above, were used as the template DNA to generate the fliC promoter DNA and derivatives. Both radiolabeled and unlabeled fliC promoters were synthesized using the polymerase chain reaction to generate a 230-base pair fragment. Each polymerase chain reaction contained 10ϫ Taq Buffer A (Fisher), 50 pmol of each primer, 40 ng of template DNA, and 2.5 units of Taq polymerase (Fisher) in a final volume of 100 l. A Perkin-Elmer Thermocycler was set for 35 cycles with 95°C for denaturation, 50°C for annealing, and 72°C for extension. The products were purified using the Qiaquick polymerase chain reaction DNA purification kit (QIAGEN Inc.).
DNase I Footprinting-32 P-End-labeled DNA promoter fragment and DNase I buffer (20 mM sodium Hepes (pH 7.5), 10 mM MgCl 2 , 100 mM NaCl, 0.1 mM EDTA, 1 mM dithiothreitol, and 200 g/ml bovine serum albumin) were combined in a volume of 50 l. Holoenzyme was added in 10-fold excess over the DNA unless noted otherwise. Additional processing of these samples has previously been described (40). Several variations of the DNase I footprinting are described below.
Heparin Competition-The E 28 -promoter complexes were allowed to form at 37°C for 15 min. In some cases, nucleoside triphosphates (NTPs) were added to a final concentration of 0.2 mM for 1 min, and then heparin was added (25 g/ml final concentration), followed by incubation for another minute. The complexes were treated with 0.5 units of DNase I (Promega) for 30 s and processed as already described.
Temperature Variation-The E 28 -promoter complexes were allowed to form at 0, 4, 16, 25, and 37°C for 15 min. The complexes formed at 0 and 4°C were subjected to DNase I digestion (2 units of DNase I) for 35 and 30 min, respectively. Complexes formed at 16°C were digested for 4 min (1.5 units of DNase I), and the 25 and 37°C samples were digested for 1 min (1.0 unit of DNase I).
KMnO 4 Footprinting-E 28 and 0.1 pmol of end-labeled DNA promoter fragment were combined in KMnO 4 buffer (20 mM sodium Hepes (pH 7.5), 10 mM MgCl 2 , 100 mM NaCl, 0.1 mM EDTA, 0.2 mM dithiothreitol, and 200 g/ml bovine serum albumin) to a final volume of 50 l and then treated with 2.5 l of 50 mM KMnO 4 (Sigma) for 2 min at the temperatures indicated. Additional processing has been described (42).
Nucleotide Stabilization Assay in Vitro-The nucleotide stabilization assay was performed as outlined by Wilson and Dombroski (11) with the following modifications. The E 28 -promoter complexes were allowed to form for 15 min, and then 0.1 mM ATP, CTP, and GTP were added to the preformed complexes for 30 s, prior to filtration and washing with 0.8 M NaCl.

RESULTS
Promoter Binding and Transcription in Vitro-RNAP was formed by mixing hexahistidine-tagged 28 with purified E. coli core RNAP. Because of the high degree of similarity between the core subunits of E. coli and S. typhimurium RNAPs, a heterologous system was used as reported by others to assess the behavior of alternative factors (37)(38)(39)(40). Runoff transcription assays were performed to analyze the overall efficacy of transcription initiation by E 28 . The promoter chosen for this analysis, pfliC, drives the expression of flagellin, one of the late flagellar gene products. A transcript of the expected size was observed. No difference in behavior was noted in the presence or absence of the hexahistidine tag (data not shown). Deletion of the AT-rich sequence just upstream of the Ϫ35 element had a slight effect on transcription (2-fold reduction), but was not attributable to the presence of an UP element since holoenzyme containing a truncation of the carboxyl-terminal domain of the ␣ subunit resulted in the same transcriptional behavior as the wild-type enzyme, independent of the upstream DNA sequence (data not shown).
DNase I footprinting was used to characterize the interaction of E 28 with pfliC (Fig. 2). A DNA fragment of 230 base pairs in 2 The sequences are available upon request. length containing the fliC promoter was 5Ј-end-labeled on the template strand with 32 P. Increasing amounts of reconstituted E 28 were added, followed by DNase I digestion (Fig. 3). Continuous protection was observed from Ϫ24 to ϩ17, with partial protection from Ϫ46 to Ϫ24 relative to the transcription start site. Additional weak interactions between ϩ17 and ϩ20 could be discerned. Overall, this footprint is shorter in the upstream region than those typical for E 70 or E 32 under similar conditions.
Intermediate Complexes Formed during Transcription Initiation-We characterized the intermediate E 28 ⅐DNA complexes during the process of transcription initiation that can be visualized by manipulating the temperature of incubation. E 28 was incubated with 5Ј-end-labeled pfliC (template strand) for 15 min, at 0, 4, 16, 25, and 37°C. Protection was observed from Ϫ65 to Ϫ54 and from Ϫ48 to Ϫ19 at 0 and 4°C (Fig. 4A). At 16°C, a hypersensitive band appeared at Ϫ46, and partial protection began to extend downstream toward the Ϫ10 region and the start site (Ϫ46 to ϩ17), with the reappearance of bands at Ϫ33 and -24. By 25°C, E 28 fully occupied the region from Ϫ46 to ϩ17. At 37°C, the upstream contacts from Ϫ65 to Ϫ54 disappeared, but strong protection from Ϫ46 to ϩ17 remained. Thus, E 28 appeared to initially bind primarily to the upstream region of the promoter since Ϫ10 protection was absent at lower temperatures and then shifted downstream to contact the Ϫ10 element up to the ϩ17 region as the temperature was increased. We cannot rule out the possibility that some minor contacts in the upstream region are maintained. However, it is clear that a major rearrangement takes place as a function of temperature. The same pattern of contacts was observed with footprinting performed using DNA labeled on the non-template strand (data not shown).
Open Complex Formation-Extended DNase I footprints at 37°C are usually indicative of open complexes for E 70 (27,(43)(44)(45)(46). We used KMnO 4 sensitivity to determine which E 28promoter complexes in the DNase I temperature series were open complexes. KMnO 4 chemically modifies thymine residues in single-stranded DNA, which renders modified positions sensitive to cleavage upon piperidine treatment (47). E 28 -promoter complexes were formed for 15 min at 0, 4, 16, 25, or 37°C and subjected to KMnO 4 treatment and piperidine cleavage (Fig. 4B).
From 0 to 4°C, the DNA remained base-paired as demonstrated by lack of reactivity to KMnO 4  Promoter Variants Lacking the Ϫ10 or Ϫ35 Element-If E 28 initially binds to the Ϫ35 region, releases upstream interac-tions, and shifts its contacts to include the Ϫ10 and downstream regions, then removal of the Ϫ35 element should preclude any binding by E 28 . Likewise, removal of the Ϫ10 consensus sequence should permit binding of only E 28 in the Ϫ35 region. We constructed deletions of the TAAA sequence at Ϫ35 (pfliC⌬35) and the GCCGATA sequence at Ϫ10 (pfliC⌬10) (Fig. 2) and analyzed the DNase I footprints over a range of temperatures. Deletion of the Ϫ10 element still allowed normal binding of E 28 at 0 and 4°C, as expected based on the presence of the predicted initial binding site at Ϫ35 (Fig. 5A). However, upon raising the temperature, binding to pfliC⌬10 was gradually abolished, consistent with the idea that polymerase shifts its contacts downstream during open complex formation to interact with the Ϫ10 element and downstream sequences. In the case of pfliC⌬10, these sequences have been removed, and E 28 is unable to remain stably bound to the upstream sequences and results in dissociation.
Also as expected, E 28 was unable to bind to pfliC⌬35 even at 0°C, presumably due to lack of the initial binding site in the Ϫ35 region and upstream portions of the promoter (Fig. 5B). KMnO 4 footprinting showed no open complexes at either pfliC⌬10 or pfliC⌬35 at any temperature (data not shown). Taken together, these results support the idea that initial binding occurs from Ϫ65 to Ϫ19, followed by isomerization to relinquish major contacts between Ϫ65 and Ϫ54 and establishment of new contacts extending to ϩ20 with concomitant DNA melting.
Stability of E 28 -Promoter Complexes-Previous studies have shown that E 70 typically forms a heparin-stable open complex (48 -51). One exception is the ribosomal operon promoter, rrnBP1, which requires the addition of NTPs to confer heparin stability (52). We used heparin challenge as a tool to examine the stability of E 28 ⅐DNA complexes. The binding of E 28 to pfliC was sensitive to 25 g/ml heparin, even in the presence of initiating NTPs (Fig. 6) (11,50,54). We used an NTP stabilization assay to determine whether RP init complexes formed by E 28 on pfliC were stable to salt challenge. E 28 ⅐fliC complexes were formed in the absence and presence of NTPs and were then filtered through nitrocellulose and subjected to a 0.8 M NaCl wash. In the presence of NTPs, 52% of the complexes were retained on the filter, as expected since RP init is in equilibrium with RP O , whereas in the absence of NTPs, only 5% were retained. Therefore, E 28 RP init complexes on pfliC, like E 70 ⅐DNA complexes, are stable to 0.8 M NaCl, indicating a equivalent conformational change and stabilization upon NTP binding.

FIG. 2. Variants of the fliC promoter used in this study.
The promoters used in this study, wild-type pfliC, pfliC⌬35, and pfliC⌬10, are shown. The Ϫ10 and Ϫ35 regions for 28 are indicated in boldface type with underlining. The positions where the Ϫ10 and Ϫ35 elements were deleted are indicated by asterisks.

DISCUSSION
Transcription initiation has been characterized as a multistep process in studies in which intermediates on the pathway to formation of an initiated complex have been detected by manipulating the temperature of incubation and varying the level of Mg 2ϩ for the RNAP-promoter interactions (30,31,43,46,50,51). There is evidence for two different closed complexes (RP C1 and RP C2 ), which differ in the extent of their contacts with the promoter DNA. RP C1 , which extends from approximately Ϫ60 to Ϫ5, is only observed at low temperatures (0°C) for E 70 (30) and E 32 (31,32). However, we have recently shown that derivatives of 70 with region 1.1 or both regions 1. 1   FIG. 3. Binding of E 28 to pfliC using DNase I footprinting. 5Ј-Radiolabeled pfliC (template strand) was incubated with increasing amounts of E 28 for 15 min at 37°C and subsequently treated with DNase I. All samples were analyzed by electrophoresis on a denaturing 8% polyacrylamide gel. The extent of each footprint is indicated on the side of each footprinting ladder. and 1.2 removed slow the initiation process such that RP C1 can be detected at 37°C (11).
We have investigated the mechanism of transcription initiation by holoenzyme carrying 28 , the alternative factor required for flagellum biosynthesis in S. typhimurium. Our analysis was conducted using the promoter for fliC, encoding flagellin, a late gene in this pathway. The identity and stability of transcription complex intermediates that we identified during transcription initiation by E 28 are distinctive from those previously identified for E 70 and E 32 .
Based on our results, we suggest the following pathway for the mechanism of E 28 transcription initiation (Fig. 7). E 28 forms a short complex with pfliC at 0°C that protects primarily the upstream region of the promoter (Ϫ65 to Ϫ19) from DNase I digestion. This suggests that the initial binding of E 28 is mediated mainly through interactions in the vicinity of the Ϫ35 consensus sequence. E 28 then isomerizes to make additional contacts that extend to ϩ20. During this progression, some upstream contacts are lost as evidenced by the reappearance of bands from Ϫ65 to Ϫ54 on the DNase I footprints at 25 and 37°C. Thus, E 28 appears to initially bind in the Ϫ35 region and then shifts downstream to contact the Ϫ10 region, start site, and downstream sequences to ϩ20 while relinquishing some of the original upstream contacts. This represents a notable difference in binding intermediates from those observed for E 70 or E 32 .
The identity of RP C versus RP O complexes was assessed by KMnO 4 sensitivity, which probes for DNA strand separation.
Typically, significant open complex formation for E 70 or E 32 requires temperatures above 16°C (27,30,31). One distinctive characteristic of E 28 ⅐fliC complexes is that they are optimally strand-separated between -9 and ϩ1 even at 16°C. Thus, the extended DNase I protection observed at 16°C is representative of RP O , rather than an extended closed complex like RP C2 . Thus, one major difference between E 28 intermediates and those found for E 70 or E 32 from E. coli is the absence of a detectable RP C2 -like complex under the conditions that we used. E 28 appears to form a single closed complex characterized by a short DNase I footprint at 0 -4°C and then progresses to RP O . If an RP C2 complex forms, it may be too rapid or unstable to be detected with the methods employed in this study.
Because E 28 is functionally similar to the B. subtilis holoenzyme that is required for flagellin biosynthesis, E D , we compared the transcription initiation complexes formed by these two enzymes at their respective flagellin promoters. The flagellin gene promoter (phag) from B. subtilis is very similar to pfliC, except it contains an UP element between Ϫ60 and Ϫ40 that provides additional contact sites for RNAP through the ␣ subunit (57). E D forms complexes that are more similar to E 70 and E 32 because, at 4°C, the DNase I footprint occupies from Ϫ73 to ϩ1 (55). This protection then extends to ϩ9 and ϩ21 at 23 and 37°C, respectively. Disappearance of the upstream contacts, which is observed for E 28 , does not occur. Thus, DNA binding primarily in the Ϫ35 and upstream regions in the earliest detectable complex, with lack of contact in the Ϫ10 region, appears to be novel for E 28 .
With respect to open complex formation, however, E 28 (11, 48 -51). E 28 appears to form transcriptionally competent complexes that remain dissociable throughout the initiation process. We cannot rule out that E 28 may form heparin-resistant complexes on other 28 -dependent promoters or that heparin is actively destabilizing these complexes. However, the information from these studies is useful in comparing the relative stability of holoenzymes carrying different factors.
In the case of 70 , it has been shown that amino-terminal region 1.1 is involved in inhibition of DNA binding by (10,56) and is required for efficient open complex formation by holoenzyme. Region 1.2, which is typically present in all primary and alternative factors, may also be involved in open complex formation for E 70 (11). 28 is a very unusual member of the 70 family of proteins since it lacks any homology to region 1.2, leading to some speculation regarding how this difference in structure may translate into a difference in function. The flagellar biosynthesis factor from B. subtilis, D , retains homology to region 1.2. E D forms two closed complexes and isomerizes through several open complex intermediates (55). 28 , on the other hand, lacks both regions 1.1 and 1.2. E 28 forms an unusually short RP C1 , and we did not observe an RP C2 -like intermediate. Additionally, many E 70 open complexes are stable to heparin either in the absence or presence of NTPs, and open complexes do not form efficiently at low temperatures, whereas the opposite is true in both cases for E 28 . The composition of the amino terminus of the factor may affect the nature of the RNAP⅐DNA complexes that can be discerned during initiation. 28 appears to facilitate transcription initiation at pfliC very efficiently by utilizing a mechanism with few intermediate complexes, but with reduced stability of the open complex. Whether this mechanism is a general phenomenon for all 28 -dependent promoters or whether it is characteristic for the flagellin gene promoter remains to be determined. FIG. 7. Schematic of the mechanism for transcription initiation by E 28 . RNA polymerase (R) binds to the promoter DNA (P) to form a closed complex (RP C ), which extends from Ϫ65 to Ϫ19. Extension of the downstream contacts to ϩ17 and release of upstream contacts occur concomitantly with open complex formation (RP O ). Regions of full protection are indicated by black bars. Partial protection is indicated by dashed lines. The thymidine residues sensitive to KMnO 4 modification in the open complex are indicated (‚).