One Gene and Two Proteins: a Leaderless mRNA Supports the Translation of a Shorter Form of the Shigella VirF Regulator

ABSTRACT VirF, an AraC-like activator, is required to trigger a regulatory cascade that initiates the invasive program of Shigella spp., the etiological agents of bacillary dysentery in humans. VirF expression is activated upon entry into the host and depends on many environmental signals. Here, we show that the virF mRNA is translated into two proteins, the major form, VirF30 (30 kDa), and the shorter VirF21 (21 kDa), lacking the N-terminal segment. By site-specific mutagenesis and toeprint analysis, we identified the translation start sites of VirF30 and VirF21 and showed that the two different forms of VirF arise from differential translation. Interestingly, in vitro and in vivo translation experiments showed that VirF21 is also translated from a leaderless mRNA (llmRNA) whose 5′ end is at position +309/+310, only 1 or 2 nucleotides upstream of the ATG84 start codon of VirF21. The llmRNA is transcribed from a gene-internal promoter, which we identified here. Functional analysis revealed that while VirF30 is responsible for activation of the virulence system, VirF21 negatively autoregulates virF expression itself. Since VirF21 modulates the intracellular VirF levels, this suggests that transcription of the llmRNA might occur when the onset of the virulence program is not required. We speculate that environmental cues, like stress conditions, may promote changes in virF mRNA transcription and preferential translation of llmRNA.

The relevance of virF activation for the invasive program is further supported by posttranscriptional regulation of icsA. RnaG is a cis-encoded antisense RNA that promotes premature termination of the icsA mRNA (19). VirF binds the RnaG promoter and decreases rnaG expression (14). Thus, VirF plays a dual role: (i) it relieves H-NS-mediated repression to activate icsA transcription, and (ii) it represses RnaG transcription, thus increasing the level of icsA mRNA (14). VirF also globally activates the expression of chromosomal genes in both Shigella and Escherichia coli. In particular, VirF appears to play a role in shaping the Shigella transcriptional program to better match the requirements of an effective intracellular life (20)(21)(22).
Like other members of the AraC family of transcriptional regulators, VirF has a conserved, carboxy-terminal DNA-binding domain with two helix-turn-helix (HTH) motifs. AraC-like proteins are typically insoluble and, accordingly, problems with VirF purification have hampered biochemical studies (23). Mutagenesis experiments indicated that the N-terminal domain of VirF promotes dimerization while C-terminal HTH2 motif mutants are nonfunctional (24).
While attempting a thorough characterization of VirF, we found that the virF mRNA (R1) is subject to differential translation, giving rise to two forms of VirF. VirF 30 activates the virulence system and some chromosomal genes, whereas VirF 21 exerts negative feedback control on virF expression itself.
Moreover, we identified a second virF mRNA species (R2) with a 5= end at position nucleotide (nt) ϩ309/ϩ310. This leaderless yet translation-competent mRNA is transcribed from a geneinternal promoter. Possible implications in an interplay between environmental sensing and virulence gene expression are discussed.

RESULTS
The virF gene encodes two independently translated proteins, VirF 30 and VirF 21 . Earlier experiments on E. coli minicells carrying the virF gene on recombinant plasmids from Shigella flexneri and Shigella sonnei indicated two main VirF protein forms of about 30 and 21 kDa and a minor form of 27 kDa (25,26). The significance of the 27-and 21-kDa forms remained unclear, and it seemed possible that they were degradation products of fulllength VirF (27). To analyze which VirF forms are present in Shigella, a 3ϫFLAG tag sequence was inserted at the 3= end of the S. flexneri M90T virF ORF. Western blot analysis (Fig. 1A) confirmed that two VirF proteins, VirF 30 (30 kDa) and VirF 21 (21 kDa), are expressed by S. flexneri. The 27-kDa form was not detected.
The sequence of the virF gene contains three putative start codons, all in the same frame, for VirF 30 and an internal ATG codon, consistent with independent translation of VirF 21 (25). Thus, we determined at which ATG codons VirF 30 and VirF 21 translation initiates. In the absence of a recognizable Shine-Dalgarno (SD) sequence upstream, prediction of the ATG encoding the N-terminal methionine of VirF 30 was difficult. Thus, each of the ATG codons (codons 1, 2, and 4; codon 3 encodes Asp) (Fig. 1B) was tested for translation initiation activity by using plasmids carrying the virF promoter followed by a virF-lacZ translational fusion. Plasmid pFL-4A is fused in frame after the fourth virF codon (third Met codon), and pFL-1A is fused after the first ATG ( Fig. 2A). ␤-Galactosidase activities indicated that the construct with all three ATGs has Ϸ5-fold-higher activity than the one fused after ATG 1 . Thus, ATG 2 and/or ATG 4 appear to be required for high translation of VirF 30 , and ATG 1 gives a minor contribution. ATG 4 , which has a short upstream SD-like (GAA) sequence, was tested by introducing an ATG 4 ¡ GGG (Gly) mutation into pFL-4A. This plasmid, pFL-M4G, in which ATG 1 and ATG 2 are still present, gave very low reporter gene activity ( Fig. 2A), suggesting ATG 4 as the main VirF 30 start codon. Western blot analysis supported this. VirF 30 was produced only from the wild-type (wt) virF gene, but not when ATG 4 had mutated (Fig. 2B, cf. pMYSH6504 and pF-M4G). To corroborate this finding in vitro, we used a toeprinting assay to analyze the formation of ribosomal initiation complexes on virF mRNA (28). A predominant toeprint was seen 17 nt downstream of AUG 4 and a minor one 16 nt downstream of AUG 1 (Fig. 2C), in line with our in vivo results ( Fig. 2A and B). Additional bands downstream of position ϩ17 of AUG 4 implicated possible 30S binding-driven structure changes resulting in reverse transcription pauses. In conclusion, translation of VirF 30 initiates predominantly at ATG 4 . Throughout the remainder of this paper, codon positions are accordingly renumbered, with ATG 4 as codon 1.
While searching for a VirF 21 translation start site, we noticed an in-frame ATG codon within virF at position 311 to 313 (relative to ϩ1 of virF) (Fig. 1B), consistent with translation of the minor form of VirF. To validate ATG 81 (formerly ATG 84 ) as the start codon for VirF 21 , two mutations were introduced into virF, generating a codon change and a frameshift, respectively. To mutate ATG 81 to a different codon that would retain VirF 30 function, we changed the ATG (mRNA position 311 to 313) to CTG (Met to Leu; pF-M81L) (Fig. 3A) or to ATC (Met to Ile; pF-M81I). Neither mutation should affect VirF 30 translation but should abolish independent translation of VirF 21 . Both mutant VirF 30 proteins were tested for activated expression of virB in a virF-defective S. flexneri strain (M90TFd) (see Table S1 in the supplemental material) carrying plasmids expressing wt VirF, VirF M81L , or VirF M81I . VirF M81L but not VirF M81I activated virB to a level comparable to wt (see Fig. S1 in the supplemental material). Thus, the substitution in VirF M81I impairs VirF 30 functionality, and therefore only pF-M81L was used in subsequent experiments. Moreover, the exclusive expression of VirF 30 upon Met ¡ Leu substitution (Fig. 3B) identified ATG 81 as the start codon for VirF 21 .
To uncouple the translation of VirF 30 and VirF 21 , we inserted a  Figure 4A shows that VirF 30 alone (pF-M81L) induced the expression of both lacZ fusions to a level similar to that in the presence of both VirF 30 and VirF 21 (pMYSH6504). VirF 21 alone (pF-FS) failed to activate (Fig. 4A). Quantitative reverse transcription-PCR (qRT-PCR) results with the S. flexneri strain M90T Fd (virF defective) carrying the same three plasmids supported this conclusion (Fig. 4B). Thus, a role for VirF 21 in the activation of the virulence cascade of Shigella is ruled out. A qRT-PCR experiment also confirmed that the previously shown VirF-dependent activation of some chromosomal heat shock genes (20) cannot be carried out by VirF 21 (see Fig. S3 in the supplemental material).
To address putative functions of VirF 21 , we investigated its role in positive or negative autoregulation of the virF gene. An E. coli K-12 strain harboring a P virF -lacZ fusion (DH10b pvirF-lacZ) was transformed with plasmids that expressed either Ptac promoterdriven VirF 30 (pAC-30) or VirF 21 (pAC-21). Figure 5A clearly shows that VirF 21 , but not VirF 30 , strongly repressed virF expression, and qRT-PCR on the same samples showed corresponding decreases in lacZ mRNA levels in the presence of VirF 21 (Fig. 5B). To validate VirF 21 -mediated repression of virF transcription in Shigella, we asked whether increased VirF 21 levels would reduce the expression of the VirF-activated virB gene. qRT-PCR experiments in the virF-defective strain M90T Fd expressing VirF 30 from pF-M81L confirmed severely reduced virB transcription upon induction (isopropyl-␤-D-thiogalactopyranoside [IPTG]) of VirF 21 (pAC-21) (Fig. 5C). To monitor the VirF 21 induction-dependent effect on the VirF protein level, we introduced pAC-21 in the S. flexneri strain that contained the 3ϫFLAG virF gene (M90T F3xFT; see above). This setup permitted us to assess the levels of VirF 21 and VirF 30 encoded by pINV by FLAG-tagged antibodies as a function of increasing levels of untagged VirF 21 expressed from pAC-21 (monitored via a halon anti-VirF antibody). Figure 5D shows that increasing the VirF 21 concentration resulted in a decrease in VirF 30 , confirming that VirF 21 negatively autoregulates virF expression.
In addition, we performed DNase I footprinting by in vitrotranslated VirF 21 on both strands of the virF promoter region. VirF 21 was translated in an in vitro system (PureSystem) (see Materials and Methods), using a PCR-generated DNA template for virF 21 -only transcription and translation. VirF 21 translation was verified by Western blotting (see Fig. S4 in the supplemental material). Figure 5E indicates that VirF 21 binding conferred protection of the virF promoter region between positions Ϫ90 and Ϫ20 on the plus strand and approximate positions Ϫ60 to Ϫ10 on the minus strand and enhanced minus-strand cleavage from about positions Ϫ70 to Ϫ90. This result, together with data from in vivo experiments ( Fig. 5A and B), strongly suggests that the transcriptional repression of virF by VirF 21 depends on its direct binding to the consensus virF promoter elements.
Identification of a VirF 21 -encoding leaderless mRNA. The above results showed that two VirF proteins are independently translated. Whether both are translated from the same mRNA, or different versions of virF transcripts, was unknown. The possibility of different mRNAs was suggested by two virF mRNA variants detected in a Northern blot assay performed with total RNA from strain M90T Fd complemented with the virFencoding pMYSH6504 and with plasmid-free M90T Fd (Fig. 6A). An Ϸ960-nt band (full-length virF mRNA; R1) and an Ϸ680-nt mRNA that might support translation of VirF 21 (R2) were visible. To test whether R2 virF mRNA is transcribed from a virF internal promoter or generated by processing, virF-lacZ transcriptional fusions and primer extension (PE) analyses were used. We constructed four virF-lacZ fusions starting at positions ϩ70, ϩ145, ϩ205, and ϩ305; all were fused at ϩ405. The ␤-galactosidase activities clearly indicated the presence of a promoter between ϩ205 and ϩ305; truncation up to position ϩ305 produced back- ground values (Fig. 6B). A promoter was indeed predicted by Pro-moterHunt (29), with consensus Ϫ10 (CATTAT; ϩ298 to ϩ303) and Ϫ35 elements (TTGACA; ϩ276 to ϩ289) (Fig. 6C). After mutagenesis of the Ϫ10 box [CATTAT to CGTTAT; pRS-F(ϩ205 Ϫ10mut)], we observed a severe reduction (Ϸ7-fold) in the ␤-galactosidase level. This new promoter was further delineated by PE analysis on RNA extracted from E. coli cells harboring the different virF-lacZ plasmids. This showed 5= ends at positions ϩ309, ϩ310 (major band), and ϩ311. All three bands were absent in the PE on pRS-F(ϩ305), while with pRS-F(ϩ205 Ϫ10mut) the ϩ309/310 bands were not detected. The weaker band at ϩ311 is consistent with a shifted Ϫ10 box (data not shown). Thus, the R2 virF mRNA is transcribed from a second virF promoter, with a transcription start site at position ϩ309/ϩ310.
The 5= ends at ϩ309 to 311 and the start codon at ϩ311 to 313 imply that the R2 mRNA is leaderless (Fig. 6C). To test whether the llmRNA is VirF 21 translation competent, we cloned the sequence corresponding to R2 mRNA, and also the entire R1 mRNA as a control, downstream of a T7 promoter. To ensure a correct 5= end of the R2 mRNA in vivo (5= U ϩ309 as ϩ1) (Fig. 6C), a hammerhead ribozyme sequence downstream of the T7 promoter (see Fig. S5 in the supplemental material) was introduced to generate an R2 mRNA starting at position ϩ309. The plasmids carrying the R1 or R2 transcripts, pAC-T730-FT (R1; virF ϩ1 to ϩ888) and pAC-T7-HH-21-FT (R2; virF ϩ309 to ϩ888) also harbored 3= FLAG tags in virF. Upon IPTG induction, virF mRNA transcription from the T7 promoter was induced in E. coli BL21(D3) harboring either plasmid. PE analysis verified the expected 5= ends of both transcripts (Fig. 7A).
VirF 21 translation from the leaderless R2 mRNA was tested by immunoblot analysis on protein extracts after induction. VirF 21 was detected in cells carrying pACT7-HH-21-FT, confirming that R2 is a leaderless translation-proficient mRNA (Fig. 7B, right panel). Translation of both VirF forms was observed in cells harboring pAC-T730-FT (Fig. 7B, left panel). In vitro translation in the PureSystem (30) was tested on R1 and R2 mRNAs carrying FLAG tag sequences. Translation products were analyzed with anti-FLAG antibodies. In agreement with the in vivo results, R1 mRNA supported translation of both VirF forms, whereas the leaderless R2 transcript only produced VirF 21 (Fig. 7C). Furthermore, toeprint experiments on in vitro-transcribed virF R1 mRNA (start, ϩ1) showed a strong RT stop near the 5= end, indicating initiation complex formation at AUG 4 (compare with Fig. 2C). In contrast, a specific toeprint was observed at position ϩ326 for the llmRNA R2 (start, ϩ309) (Fig. 7D). This toeprint was absent on R1 mRNA, indicating a strong preference for VirF 30 translation from the full-length mRNA. Together, these results suggest that a new virF promoter generates a llmRNA variant (R2 mRNA) dedicated to the exclusive translation of VirF 21 .

DISCUSSION
The complex regulatory cascade for activation of the Shigella virulence genes depends on the VirF protein (7). VirF is at the heart of the switch from the noninvasive to the invasive phenotype. Thus, it is not surprising that its expression is triggered by many environmental signals and that it is controlled at several levels (2,4,10,17). Since its discovery, VirF was known to be present in three forms that differ in size: 30, 27, and 21 kDa (25). The smaller forms were ignored as likely degradation products. Here, we report that the VirF 21-kDa form is translated as an independent polypeptide. Our results address how the VirF 21 variant is produced and suggest an autoregulatory role in virF expression. As a first step, we identified the translation start sites of VirF 30 and VirF 21 . Of the three Met codons among the first four codons of the predicted virF ORF, only ATG4 was essential for VirF 30 translation ( Fig. 1 and 2). Replacement with GGG drastically reduced VirF, as monitored by Western blotting or ␤-galactosidase activity of virF-lacZ translational fusions ( Fig. 2A and B). The identification of ATG4 as a start codon was further supported by toeprint analysis (Fig. 2C). The start codon consistent with the size of VirF 21 is ATG81; accordingly, replacement with CTG blocks VirF 21 production (Fig. 3B).
Interestingly, while the wt virF mRNA is translated into both VirF 30 and VirF 21 in vivo, a frameshift mutation upstream of ATG81 affects only the production of VirF 30 , and not that of VirF 21 . Thus, the two forms are independently translated (Fig. 3B); consequently, a derivative with either the FS mutation or the M81L substitution gives only VirF 21 or only VirF 30 , respectively. ␤-Galactosidase fusion and immunoblot analyses (Fig. 3C) showed that the expression level of VirF 21 under our experimental conditions is generally lower than that of VirF 30 . VirF 21 is clearly not functionally redundant with VirF 30 . Unlike VirF 30 , it does not restore the expression of VirF-regulated genes in a virF-defective S. flexneri mutant (Fig. 4). Instead, overexpres-   30 and VirF 21 were translated from virF R1-3XFT, but only VirF 21 was translated from virF R2-3XFT mRNA. Asterisk, unspecific cross-hybridization with a protein in the extract. In the blot on the right, we included ompF mRNA as an internal canonical, SD-dependent translation control. (D) Toeprint assay results with ϩ1 (full-length) and ϩ309 (leaderless) virF mRNAs. The mRNAs were incubated alone (control; lanes 1 and 4), with 30S (lanes 2 and 5), or with 30S and tRNA ϪfMet (lanes 3 and 6). A specific toeprint was observed on full-length mRNA (ϩ1) near the 5= end (black circle) (compare with Fig. 2C). A second toeprint, specific to the llmRNA, is at position ϩ326 (black asterisk). Sequencing ladders were generated with the same 5=-end-labeled primer. sion of VirF 21 negatively autoregulates virF expression, reducing intracellular levels of VirF 30 and causing reduced virB expression (Fig. 5B). This negative autoregulation is likely due to VirF 21 binding to the virF promoter, as indicated by the position of a DNase I footprint (Fig. 5E) which is predicted to interfere with RNA polymerase access.
An arrangement based on a smaller protein controlling a larger one, with both of them encoded by the same gene, applies to Tn5 transposase (31). The form of Tn5 transposase lacking the first 55 amino acids posttranslationally forms nonproductive complexes with transposase, thus blocking its activity at IS50 inverted repeats (31). Superficially similar in setup, the shorter VirF 21 also lacks a large N-terminal portion of the longer VirF 30 protein, but here, the shorter form alone is sufficient to exert control at the level of virF transcription (Fig. 5A). Though the known C-terminal DNA-binding domain is present in both VirF variants, our data suggest different DNA recognition preferences. Further work will test whether N-terminal sequences affect binding properties of VirF 30 and whether protein folding differences in the shared domain can account for the observed specificity differences.
Since VirF 30 and VirF 21 originate from differential translation, we investigated the virF transcripts in more detail. Long (R1) and shorter (R2) virF mRNAs of lengths compatible with VirF 30 and VirF 21 were detected (Fig. 6A). Evaluation of deletions in the region upstream of the VirF 21 ORF, along with PE analyses, identified a new gene-internal virF promoter that drives the transcription of the virF R2 mRNA. In vivo and in vitro data support that the leaderless R2 is translated into VirF 21 ; plasmid vectors encoding R2 (start site, ϩ309) support in vivo translation of VirF 21 (Fig. 7B). Moreover, leaderless translation of VirF 21 by R2 also occurs in vitro (Fig. 7C), and initiation complex formation occurs at the appropriate position (Fig. 7D).
In recent years, noncanonical translation initiation mechanisms have been reported, including so-called leaderless transcripts, i.e., those lacking a 5=-untranslated region (UTR) and an SD sequence (32)(33)(34). Most leaderless genes identified so far in E. coli reside in mobile DNA, including , P2, and Tn1721. The virF gene is also located within an IS-rich region on an extrachromosomal element, the large Shigella/EIEC invasive plasmid (9). Sequencing data for bacteria and archaea suggest that a leaderless model may not be uncommon (35,36).
The mechanisms underlying synthesis and translation of llm-RNAs are not yet fully understood. Vesper et al. (37) showed that induction of the MazEF toxin-antitoxin (TA) system in E. coli produces a leaderless mRNA population and, simultaneously, specialized "stress" ribosomes with a preference to translate proteins from llmRNAs. The endoribonuclease MazF cleaves singlestranded mRNAs, sometimes at ACA sequences upstream of AUG start codons, generating llmRNA. MazF also cleaves 16S rRNA, removing the anti-SD sequence required for translation on canonical mRNAs. Thereby, a subpopulation of ribosomes is generated for selective translation on llmRNA (37). It is well established that Shigella bacteria sense and respond to environmental conditions within and outside the host, with corresponding reprogramming of transcription. Since VirF 21 modulates the intracellular level of VirF, this suggests that the transcription of the leaderless R2 mRNA could occur under conditions where the activation of the virulence program is undesirable. A possible coupling between stress conditions that might promote changes in R2 virF mRNA transcription and/or preferential translation of leaderless R2 mRNA and effects on virulence gene regulation is an exciting possibility that we intend to pursue. In particular, the environmental cues that may regulate transcription of the shorter virF mRNA, and the translation of VirF 21 from the llmRNA under stress and infection-relevant conditions, will be addressed. In summary, this study has added new, entirely unexpected elements to the complex regulation of the Shigella virulence system and of its major regulator, the VirF protein.

MATERIALS AND METHODS
Oligodeoxyribonucleotides. Oligodeoxynucleotides used in this study (see Table S1 in the supplemental material) were purchased from Metabion.
Bacterial strains and general methods. Strains used in this study are listed in Table S2 in the supplemental material. Cloning was performed wtih strain DH10b. E. coli strain P90CB was obtained by transferring a P virB -lacZ fusion from plasmid pRS415 via homologous recombination to the lac transducing phage RS45 and then integration (38) into the the att site of E. coli P90C. P90CA was previously described (see Table S2). Strains M90T-F3xFT and M90T Fd(⌬virF) were previously constructed (21).
Bacteria were grown aerobically in LB medium at 37°C. Antibiotics and chemicals were used at the following concentrations: ampicillin, 50 g/ml; chloramphenicol, 25 g/ml; kanamycin, 30 g/ml; streptomycin, 10 g/ml; tetracycline, 5 g/ml; 5-bromo-4-chloro-3-indolyl-␤-Dgalactopyranoside, 20 mg/ml. ␤-Galactosidase assays were performed as described elsewhere (39). Reported values represent the means of at least three separate measurements. DNA isolation, PCR, restriction digests, cloning, and other DNA manipulation methods were performed as described previously (39). Plasmids are listed in Table S3 in the supplemental material. In addition, plasmid constructions are detailed in Text S1 in the supplemental material.
Analysis of virF mRNA. S. flexneri M90T Fd (⌬virF) (Table S2) cells with or without pMYSH6504 were grown in LB broth at 37°C to an optical density at 600 nm of 0.4 to 0.5. Total RNA extraction and Northern blot assays with an ␣-32 P-labeled virF-specific probe were performed as described elsewhere (21). Loading controls entailed rRNA staining. Radioactivity was quantified using a Typhoon 9200 instrument (GE Healthcare).
qRT-PCR was performed using Power SYBR green PCR master mix on a 7300 real-time PCR system (Applied Biosystems) as described previously (19). The levels of virB, icsA, and lacZ transcripts were analyzed using the 2Ϫ ⌬⌬CT (cycle threshhold [C T ]) method (40), and results are reported that the fold increase relative to the reference. Primers for mdh (endogenous control) and for virB, icsA, and lacZ transcripts were designed by using Primer Express software v2.0 and validated. The following oligonucleotides were used (see Table S1 in the supplemental material): mdhQF/mdhQR, virBQF/virBQR, icsAQL/icsAQR, and lacZQF/lacZQR. Primer extensions. Total RNA from exponentially growing plasmidcarrying E. coli strains was extracted (41). Total RNA (10 to 20 g) was hybridized with 5=-32 P-labeled ML-512 and ML-1314 primers. Reverse transcription experiments were done at 42°C using the reverse transcriptase ImProm-II (Promega). Reaction products were analyzed on an 8% polyacrylamide gel in parallel with sequencing reaction products obtained using the same primers.
DNase I footprinting. Supercoiled plasmid pMYSH6504 (42) (200 ng/sample) was preincubated for 20 min at room temperature with the indicated volumes of the translation mixture, which contained VirF 21 or control (no-template) PureSystem reagent in 30 l of binding buffer (40 mM Tris-HCl [pH 8.0], 50 mM KCl, 10 mM Mg-acetate, and 0.5 mM dithiothreitol). The DNA-protein complex was incubated with 1 U of DNase I for 40 s. After stopping the reaction, the DNA was precipitated and separately analyzed by primer extension on either DNA strand with 3 pmol of 5=-end-labeled primers ML-U30 or ML-U29 as described pre-viously (14). The extension products and corresponding sequencing reactions were run on 7% sequencing gels and then fixed for 5 min (10% ethanol-6% acetic acid) and dried. Signals were detected using a phosphorimager screen.
Immunodetection of VirF protein. Western blot assays were carried out as described in reference 21. Incubation with primary antibodies (polyclonal halon anti-VirF, anti-FLAG [Sigma F1804]) was at 4°C in phosphate-buffered saline-Tween (PBS-T) containing 2% dried skim milk. Membranes were washed and incubated at room temperature for 1 h with a secondary anti-rabbit (1:10,000) or anti-mouse (1:5,000) horseradish peroxidase-conjugated antibody in PBS-T. After washing with PBS-T, membranes were developed for 5 min for enhanced chemiluminescence and visualized on a ChemiDoc XRSϩ system.
RNA in vitro transcription. The virF-3XFT mRNAs R1 and R2 were transcribed for in vitro translation and toeprint assays. For virF mRNA R1-3XFT (start, ϩ1), DNA templates contained a T7 promoter (PCR with primers ML-U1/ML-982). For virF mRNA R2-3XFT (start, ϩ309), a fragment with a T7 promoter and a hammerhead ribozyme sequence in front of the virF sequence was produced by PCR (primers ML-U20/ML-982) on pAC-T7-HH-21-FT as the template (for the hammerhead sequence, Fig. S5 in the supplemental material). DNA templates were in vitro transcribed as described in reference 43. To obtain virF R2-3XFT, an additional ribozyme self-cleavage step was performed after in vitro transcription according methods described previously (44).
Toeprint assay. Toeprint assays were performed as in reference 45. Aliquots of 0.2 pmol of unlabeled virF-3xft mRNAs R1 and R2 were annealed with 0.5 pmol 5=-end-labeled ML-U25 or ML-U26 primer in water at 95°C for 1 min and chilled on ice for 2 min. After addition of renaturing buffer (20 mM Tris-HCl [pH 7.5], 20 mM MgCl 2 , 100 mM NH 4 Cl) and incubation for 10 min at 37°C, 2 pmol of 30S ribosomal subunits was added. After 15 min, 4 pmol of tRNA-fMet was added, and incubation continued for 20 min before cDNA synthesis with avian myeloblastosis virus reverse transcriptase (7.5 U; Invitrogen) and deoxynucleoside triphosphates (100 nM). Reactions were stopped by phenol-chloroformisoamyl alcohol extraction followed by ethanol precipitation. The cDNAs and sequencing reactions were run on 8% denaturing polyacrylamide gels that were fixed for 5 min (10% ethanol-6% acetic acid) and dried for 1 h at 80°C. Signals were detected using a phoshorimager screen.
In vitro translation. To generate VirF 21 for DNase I footprinting, 500 ng of a PCR product containing a T7 promoter and the virF 21 coding sequence was used as the template in the PureSystem Express (New England BioLabs [NEB]) transcription-translation system at 37°C for 4 h. VirF 21 translation was analyzed by immunoblotting using anti-VirF antibodies (see Fig. S4 in the supplemental material). For the in vitro translation of different virF mRNAs (Fig. 7C), each purified transcript was denatured for 2 min at 95°C, chilled for 1 min on ice, diluted in TMN (20 mM Tris-HCl [pH 7.5], 10 mM MgCl 2 , 150 mM NaCl), and incubated for 15 min at room temperature. In vitro translation (mRNA at 50 nM) was performed with the PureSystem Express (NEB) translation system at 37°C. Translation products were analyzed by immunoblotting with anti-FLAG antibodies.