Two Frameshift Products Involved in the Transposition of Bacterial Insertion Sequence IS629*

IS629 is 1,310 bp in length with a pair of 25-bp imperfect inverted repeats at its termini. Two partially overlapping open reading frames, orfA and orfB, are present in IS629, and two putative translational frameshift signals, TTTTG (T4G) and AAAAT (A4T), are located near the 3′-end of orfA. With the lacZ gene as the reporter, both T4G and A4T motifs are determined to be a –1 frameshift signal. Two peptides representing the two transframe products designated OrfAB′ and OrfAB, are identified by a liquid chromatography-tandem mass spectrometric approach. Results of transposition assays show that OrfAB′ is the transposase and that OrfAB aids in the transposition of IS629. Pulse-chase experiments and Escherichia coli two-hybrid assays demonstrate that OrfAB binds to and stabilizes OrfAB′, thus increasing the transposition activity of IS629. This is the first transposable element in the IS3 family shown to have two functional frameshifted products involved in transposition and to use a transframe product to regulate transposition.

The insertion sequence IS629 is a member of the IS3 family of transposable elements. It was initially isolated from the chromosome of Shigella sonnei (1) and has been detected in many other enteric bacteria, including S. dysenteriae, S. flexneri, S. boydii, Escherichia coli C, E. coli O157:H7, Enterobacter cloacae MD36, and Serratia marcescens (2). IS629 is 1,310 bp in length and has a pair of 25-bp imperfect inverted repeats at its termini (3). Similar to the genetic organization of other members of the IS3 family, two consecutive and partially overlapping open reading frames, designated orfA and orf B, are present in IS629 (see Fig. 1). The coding potential of orfA (nucleotides 55-381) is 108 amino acids, and that of orf B (nucleotides 378 -1,268) is 296 amino acids (3). The stop codon ( 379 TGA) of orfA overlaps the initiation codon ( 378 ATG) of orf B (see Fig. 1). A putative promoter and the Shine-Dalgarno sequence are found upstream from the initiation codon of orfA, but no such sequences are present in the upstream region of orf B (3).
Two putative Ϫ1 translational frameshift signals, TTTTG (T 4 G) and AAAAT (A 4 T), are located near the 3Ј-end of the orfA at nucleotide positions 342-346 and 375-379, respectively ( Fig. 1) (4), suggesting the existence of two frameshifted products. In this study, we demonstrated that both of these two putative frameshift signals are functional, causing a Ϫ1 translational frameshift and resulting in the production of two transframe products designated OrfABЈ and OrfAB. OrfABЈ was shown to be the transposase of IS629, and OrfAB was demonstrated to bind and stabilize OrfABЈ.

EXPERIMENTAL PROCEDURES
Cloning of IS629-A fragment containing both orfA and orf B sequences of IS629 was amplified by PCR from the chromosome of S. sonnei (ATCC 9290) or E. coli O157 with primers F NdeI-55 and R Ecl136II-Term (Table 1). The PCR product thus generated was cloned into pGEMT-Easy (Promega) to produce pGEMT629. Subsequently, the 1.2-kb NdeI-Ecl136II fragment containing orfA and orf B without terminal repeats was isolated from pGEMT629 and then inserted into the corresponding sites of pET-29a(ϩ) (Novagen), generating pET629. DNA fragments containing different portions of IS629 for various experiments were generated from pGEMT629 or pET629 by PCR using oligonucleotide primers listed in Table 1. Recombinant plasmids used in this study are described below in Table 2.
Detection of Translational Frameshifting-To detect Ϫ1 translational frameshifting in IS629, the lacZ gene was fused to a DNA fragment containing the two putative frameshift motifs, referred to as the frameshift window (fsw), 2 so that the lacZ gene is expressed only when a Ϫ1 frameshift occurs and that the frameshifting can be detected by measuring ␤-galactosidase activity. The 3.2-kb SmaI-PstI fragment containing the lacZ gene from pMC1871 (5) was cloned into the corresponding sites of pUCD1752X (6) to generate pUCDlacZ. To investigate the function of the two putative frameshift motifs, a DNA fragment (IS629 nucleotides 55-425) containing the two motifs was amplified from pGEMT629 using primers F XbaI-55 and R RsrII-425 and cloned between XbaI and SmaI sites of pUC18, generating pUC629-21. The 450-bp XbaI-Acc65I fragment from pUC629-21 was then inserted into the corresponding sites of pUCDlacZ to generate pF1wF2wIw, thus making the expression of the lacZ gene dependent on Ϫ1 frameshifting. For this and subsequent plasmid designations, F1, F2, and I represent frameshift signal 1 (T 4 G), frameshift signal 2 (A 4 T), and initiation codon for orf B, respectively, whereas * This work was supported by research grants (NSC90-2320-B-010-058 and NSC92-2320-B-010-063) from the R.O.C. National Science Council (to S.-T. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 To whom correspondence should be addressed. "w" denotes wild-type sequence, and "m" means mutated sequence. Because this 450-bp XbaI-Acc65I fragment contained the initiation codon of orf B, which may render the lacZ gene constitutively expressed, this initiation codon was mutated, generating pF1wF2wIm. To investigate the function of the first frameshift motif T 4 G (F1), the sequence AAAAT (F2) was mutated to TCGAT to create pF1wF2mIm. Similarly, the sequence TTTTG (F1) was changed to TACTG to investigate the function of the second frameshift motif, generating pF1mF2wIm. The plasmid that contained mutations of both motifs and the orfB initiation codon was called pF1mF2mIm, whereas the one containing the two mutated motifs with the wildtype orfB initiation codon was referred to as pF1mF2mIw. Mutations were created by PCR on the 450-bp XbaI-Acc65I fragment using primers listed in Table 1, and the new 450-bp XbaI-Acc65I fragment with a certain mutation was used to replace  the XbaI-Acc65I fragment on pF1wF2wIw to generate various  plasmids (Table 2 and Fig. 2

TABLE 1 Oligonucleotide primers used in this study
The underlined letters denote restriction sites, and lowercase letters indicate built-in mutations.

Primer Sequence (5 3 3)
ACGTTTCCCGTTGAATATGGCTCAT F BamHI-367 and R AscI-458 . The resulting PCR product was digested with BamHI and AscI, and plasmid pET629A 4 TMBP was generated by substituting the BamHI-AscI fragment on pET629T 4 GMBP with the digested PCR product. The plasmid pET629T 4 GMBP or pET629A 4 TMBP was then introduced into E. coli BL21(DE3). Overnight cultures of cells containing the plasmid were diluted 1:100 in Luria-Bertani (LB) broth containing 50 g/ml kanamycin and grown to an A 600 of 0.8. Isopropyl 1-thio-␤-D-galactopyranoside (IPTG) was added to the culture to a final concentration of 1 mM to induce protein expression. After 2.5 h of induction, the cells were harvested and lysed with the B-PER bacterial protein extraction reagent (Pierce) and lysis buffer (50 mM NaH 2 PO 4 , 10 mM Tris-HCl, 8 M urea, and 500 mM NaCl, pH 8.0). The whole cell lysate was clarified by centrifugation, and the His-tagged recombinant protein was purified by affinity column chromatography with the ProBond resin (Invitrogen). The proteins bound to the resin were washed with 50 ml of lysis buffer and then with 50 ml of wash buffer (500 mM NaCl and 20 mM NaPO 4 buffer, pH 6.0). Finally, the bound proteins were eluted with elution buffer (150 mM imidazole, 500 mM NaCl, and 20 mM NaPO 4 buffer, pH 6.0). The eluted proteins were concentrated and washed extensively with wash buffer using an Amicon Ultra PL-30 filter (Millipore). The purified proteins were resolved on 10% SDS-polyacrylamide gels for visualization by Coomassie Blue staining or Western blotting. For Western blotting, the proteins on the gel were electrotransferred onto Immobilon-P transfer membranes (Millipore). The membranes were then probed with a 1:1,000 dilution of the anti-MBP antibody (7) and a 1:1,000 dilution of horseradish peroxidase-conjugated anti-mouse immunoglobulin G (IgG) antibodies (Sigma). The signals on the blots were visualized using the enhanced chemiluminescence system (PerkinElmer Life Sciences).
Mass Spectrometry-The gel piece containing the transframe protein was subjected to reduction, pyridylethylation, and ingel digestion with trypsin or Asp-N as described by Tsay et al. (7). The digested products were separated by an Agilent 1100D high-performance liquid chromatography system, which was interfaced with the ThermoFinnigan LCQ Deca XP ion trap tandem mass spectrometer. A 150-ϫ 0.3-mm Agilent 300SB C18 column (3-m particle diameter, 300-Å pore size) with mobile phases of 0.1% formic acid in water and 0.1% formic acid in acetonitrile was used to separate peptides.
The occurrence of frameshifting was confirmed by detecting peptides that were produced by a Ϫ1 frameshift by matching the mass spectra of the peptides against two databases created based on the frameshift model of Jacks et al. (8). These two databases contained computer-generated nucleotide sequences of different peptides that may be produced by trypsin or Asp-N digestion of fsw(T 4 G)-MBP-His 6 and fsw(A 4 T)-MBP-His 6 , respectively, translated in both the 0 and Ϫ1 frames. Because the region located between the most downstream peptide sequence derived from the 0 translation frame and the most upstream peptide sequence derived from the Ϫ1 translation frame is where frameshifting may occur, this region is referred to as the "frameshift region." Two additional new databases were then created. The first one contained the nucleotide sequences of different peptides derived from a Ϫ1 frameshift that occurs at each codon within the first frameshift region (fsw(T 4 G)-MBP-His 6 ), and the second one contained those derived from the second frameshift region (fsw(A 4 T)-MBP-His 6 ). Each nucleotide sequence corresponded to a resultant peptide from one Ϫ1 frameshifting event within a frameshift region.
The collision-induced dissociation spectra of a peptide were acquired as three successive scans as described by Tsay et al. (9). The acquired collision-induced dissociation spectra were interpreted using a ThermoFinnigan software package, the Turbo-SEQUEST browser, which matches the tandem mass spectrum with those in the databases described above. The MS/MS data that matched the peptide sequences with appropriate cleavage sites at the right positions were subjected to manual analysis using another computer program (EverNew Biotech) to confirm the results.
Transposition Assays-To investigate effects of IS629-encoded proteins on IS629 transposition in vivo, a mini-IS629 with the kanamycin resistance gene was first constructed, and proteins that may affect IS629 transposition were supplied in trans. The left terminal repeat (IRL) sequence of IS629 was amplified from the chromosome of S. sonnei ATCC 9290 with primers F IRL-BamHI and R AscI-375 , and the PCR product was ligated into pGEMT-Easy (Promega) to generate pGEMT-IRL. The right terminal repeat (IRR) was amplified with primers F 316-Bst1107I and R IRR-BamHI and similarly cloned into pGEMT-Easy to generate pGEMT-IRR. The 1.9-kb ScaI-BamHI fragment containing the IRL from pGEMT-IRL and the 1.2-kb ScaI-BamHI fragment containing the IRR from pGEMT-IRR were joined together to obtain pGEMT-mini629. The 1.3-kb BamHI fragment containing the kanamycin resistance gene from pUC4K (Amersham Biosciences) was then inserted into the BamHI site of pGEMT-mini629 to generate pGEMT-mini629Km. Finally, pMini629 was constructed by inserting the 1.3-kb NotI fragment containing the mini-IS629 with the kanamycin resistance gene from pGEMT-mini629Km into pET-22b(ϩ) (Novagen). The 1.2-kb NdeI-Ecl136II fragment containing the orfA and orfB sequences of IS629 from pGEMT629 was then inserted into the corresponding sites of pMini629 to generate pMini629ABЈ-AB-A-B, which would express OrfABЈ, OrfAB, OrfA, and OrfB. The 370-bp NdeI-RsrII DNA on pMini629ABЈ-AB-A-B was then replaced with the 370-bp PCR-generated NdeI-RsrII DNA fragment encoding OrfABЈ, OrfAB, and OrfA to produce pMini629ABЈ-AB-A. Similarly, pMini629ABЈ-A that would express OrfABЈ and OrfA was constructed by replacing the same NdeI-RsrII DNA on pMini629ABЈ-AB-A-B with the 370-bp PCR-generated NdeI-RsrII DNA fragments encoding OrfABЈ and OrfA ( Table 2).
The transposition activity of IS629 was determined by the standard mating-out assay as described previously (10). Derivatives of pMini629 (Km r ) carrying various IS629 genes were transformed into E. coli DH1(DE3) cells (Str s ) harboring an F-derived conjugative plasmid pCJ105 (Cm r ), which served as the target for IS629 transposition. Because pCJ105 carries a chloramphenicol resistance gene, transposition of mini629Km onto pCJ105 will render the host resistant to both kanamycin and chloramphenicol. To determine the transposition frequency of IS629, pCJ105::mini629Km was mated out from E. coli DH1(DE3) to E. coli HB101(Str r ) at 37°C for 90 min. Appropriate dilutions of the conjugation mix were plated on LB agar plates containing both chloramphenicol (50 g/ml) and streptomycin (150 g/ml) as well as on plates containing kanamycin (50 g/ml), chloramphenicol (50 g/ml), and streptomycin (150 g/ml). Colonies that appeared on these plates were counted, and the transposition frequency was determined as the ratio of the number of Cm r Km r Str r colonies to that of the Cm r Str r colonies. To confirm transposition, some of the transposition products (pCJ105:: mini629Km) were isolated and examined for direct repeat sequences flanking the mini-IS629. The direct repeat sequence adjacent to IRR was identified by nucleotide sequencing using primer FP-1 (Table 1), which anneals to the 3Ј-end of the kanamycin resistance gene 140 bp upstream from IRR. To detect the direct repeat sequence adjacent to IRL, primer PRP-1 (Table 1), which anneals to the 5Ј-end of the kanamycin resistance gene 164 bp downstream from IRL, was used for sequencing.
Pulse-Chase Experiments-Pulse-chase experiments were performed to investigate the half-life of OrfABЈ and OrfAB. To express OrfABЈ, the 1.2-kb NdeI-Ecl136II fragment from pMini629ABЈ-A was cloned into the corresponding sites on pET-29a (ϩ) (Novagen) to generate pET629A-ABЈ. Similarly, the NdeI-Ecl136II fragments from pMini629AB-A and pMini629ABЈ-AB-A were inserted between NdeI and Ecl136II sites on pET-29a (ϩ) to generate pET629A-AB and pET629A-ABЈ-AB that express OrfAB and OrfABЈϩOrfAB, respectively.
Overnight cultures of E. coli DH1(DE3) cells containing pET629A-ABЈ, pET629A-AB, or pET629A-ABЈ-AB were diluted 1:50 with fresh M9 minimal medium containing kanamycin (50 g/ml) and grown to an A 600 of 0.3. The cells in the culture were pelleted, washed with M9 buffer (11), and suspended in M9 minimal medium containing 2% methionine assay medium (Difco Laboratories). After 100-min incubation at 37°C, IPTG was added to the culture to a final concentration of 1 mM to induce the synthesis of the T7 RNA polymerase. Forty minutes later, rifampin (200 g/ml) was added, and the culture was incubated for another 40 min. The cells were then labeled with [ 35 S]methionine (20 Ci/ ml, Amersham Biosciences) for 10 min and subsequently chased with an excess amount of non-radioactive methionine (final concentration, 2.5 mg/ml). Samples were taken at different time points, pelleted, and suspended in electrophoresis sample buffer (50 mM Tris-HCl, pH 6.8, 2% SDS, 7 mM ␤-mercaptoethanol, 10% glycerol, 0.1% bromphenol blue). The samples were boiled for 5 min and electrophoresed on a 12% SDS-polyacrylamide gel. The gels were scanned with a PhosphorImager (Amersham Biosciences), and quantification of the protein was performed with the program ImageQuant (Amersham Biosciences).
Bacterial Two-hybrid Assay-To assess the interaction between OrfABЈ and OrfAB, an E. coli two-hybrid experiment was performed. In this system (BacterioMatch Two-Hybrid System, Stratagene), the bait plasmid pBT (Cm r ) encodes the bacterial phage cI protein under the control of the lac-UV5 promoter. This cI protein contains the N-terminal DNAbinding domain and the C-terminal dimerization domain. The protein of interest, the bait, is fused to the cI protein. The target plasmid pTRG (Tc r ) contains an RNAP␣ gene, which is driven by the E. coli lipoprotein promoter (lpp) and regulated by the lac operator. The target protein is fused to the N-terminal domain of the RNA polymerase ␣ subunit. DNA fragments encoding OrfABЈ or OrfAB were cloned into these plasmids to fuse OrfABЈ or OrfAB to cI or RNAP␣ to generate cI-OrfABЈ, cI-OrfAB, RNAP␣-Orf-ABЈ, and RNAP␣-OrfAB. When interaction between cI-Orf-ABЈ and RNAP␣-OrfABЈ, cI-OrfAB and RNAP␣-OrfAB, cI-OrfABЈ and RNAP␣-OrfAB, or cI-OrfAB and RNAP␣-OrfABЈ had occurred, these complexes would bind to the operator (O R 2) located upstream from the reporter cassette containing the lacZ genes in E. coli XL-1 Blue MRFЈ (Km r ). This binding would recruit and stabilize the binding of RNA polymerase at the promoter and activate the transcription of the reporter gene. Thus, the protein-protein interaction between OrfAB and OrfABЈ or between themselves could be determined by the levels of ␤-galactosidase activity.
The bait and target plasmids were constructed as follows. The IS629 fragment containing an insertion of a thymine residue within the T 4 G motif was amplified from pET629A-ABЈ with primers F T5 and R Ecl136II-Term . The PCR product was digested with EcoRV and HindIII and then cloned into the corresponding sites of pGEMT629, resulting in pGEMT629T5. A 1.2-kb NotI-EcoRI fragment containing IS629(T5) without the terminal repeats was then amplified from pGEMT629T5 with primers F NotI-55 and R T7 and then cloned into the corresponding sites of pBT and pTRG to yield pBT-ABЈ and pTRG-ABЈ. For construction of pBT-AB and pTRG-AB, the IS629 fragment (nucleotides 55-425) with an adenine insertion within the A 4 T motif was amplified from pET629A-AB with primers F NotI-55 and R A5 . This PCR product was ligated into pGEMT-Easy vector (Promega) to generate pGEMT629A5NR. The RsrII-SphI fragment containing the IS629 nucleotides 420 -1,269 from pGEMT629 was then inserted into the corresponding sites of pGEMT629A5NR to obtain pGEMT629A5. The NotI-EcoRI fragment of pGEMT629A5 was inserted between NotI and EcoRI sites on pBT and pTRG to generate pBT-AB and pTRG-AB, respectively.
E. coli XL-1 Blue MRFЈ was co-transformed with various combinations of recombinant bait and target plasmids to examine interaction between OrfABЈ and OrfAB or with non-recombinant pBT and pTRG vectors to serve as negative controls for the interaction analysis. Transformants were selected on LB agar plates containing 12.5 g/ml tetracycline, 34 g/ml chloramphenicol, and 50 g/ml kanamycin. In the presence of 20 M IPTG, cells were cultured at 30°C to mid-log phase, and then assayed for ␤-galactosidase activity by the method of Miller (12) using o-nitrophenyl-␤-D-galactopyranoside as the substrate.

RESULTS
Two Functional Ϫ1 Frameshift Signals-To determine whether the T 4 G or the A 4 T motif was able to provoke a Ϫ1 frameshift, a lacZ reporter gene was fused to the 3Ј-end of a DNA fragment (nucleotides 55-425) containing the entire orfA, the first 48 bp of orf B, and the wild-type or mutated putative frameshift motifs. The lac promoter and E. coli RNA polymerase were used to express the hybrid gene on these constructs ( Table 2). Translation of mRNA derived from each of these plasmids would start at frame 0 of orfA, and the ␤-galactosidase would be expressed only when a Ϫ1 frameshift had occurred, because lacZ was fused to the Ϫ1 reading frame of orfA. The ␤-galactosidase activity in cells harboring a certain construct after IPTG induction was measured (Fig. 2). As the positive control for ␤-galactosidase production, the lacZ gene on pF1mF2mLacZ was fused in-frame to orfA. The ␤-galactosidase activity conferred by this plasmid was determined to be 3,896 Ϯ 243 units (Fig. 2) and designated as 100%. pF1wF2wIw, which contained the wild type of both T 4 G and A 4 T sequences fused to the lacZ gene, conferred 665 Ϯ 50 units (17.1% of control) of ␤-galactosidase activity, suggesting that a Ϫ1 frameshift had occurred. To determine whether both T 4 G and A 4 T motifs were essential for frameshifting, they were mutated to TACTG and TCGAT, respectively. Surprisingly, pF1mF2mIw, which harbors these mutations, still conferred 504 units (12.9% of control) of ␤-galactosidase activity (Fig.  2). A careful analysis of the nucleotide sequence revealed the presence of two tandem translation initiation codons ( 78 ATGATG 83 ) located at the beginning of orf B. To determine whether the lacZ activity conferred by pF1mF2mIw was due to translation initiated from one of these two codons, the sequence ATGATG on pF1mF2mIw was changed to ATAATC to generate pF1mF2mIm. As expected, pF1mF2mIm with both the two putative frameshift signals and both the two ATG codons mutated conferred very little ␤-galactosidase activity (1.69 Ϯ 0.15 units, 0.04% of control). This result indicates that the majority of ␤-galactosidase activity conferred by pF1wF2wIw was derived from translation initiated from the initiation codons of orf B.
To assess the frameshifting function of T 4 G, the TACTG sequence on pF1mF2mIm was changed back to TTTTG, generating pF1wF2mIm. pF1wF2mIm was found to confer 148 Ϯ 14 units of ␤-galactosidase (3.8% of control), indicating that the T 4 G sequence is a functional Ϫ1 frameshift signal. Similarly, the TCGAT sequence on pF1mF2mIm was changed back to AAAAT, generating pF1mF2wIm to examine the frameshifting ability of A 4 T, and pF1mF2wIm was found to confer 173 Ϯ 11 units of ␤-galactosidase (4.4% of control). This result indicates that A 4 T is also a functional frameshift signal. When both of the mutated frameshift motifs were changed back to wild type, the plasmid pF1wF2wIm, which carries these changes, conferred 174 Ϯ 14 units of ␤-galactosidase (4.5% of control). Because pF1wF2wIm had a mutated initiation codon for orf B, this result indicates that the ␤-galactosidase activity derived from frameshifting was ϳ4.5%.
Identification of Frameshift Sites-To confirm that frameshifting indeed occurred, the transframe products were identified. A DNA fragment containing nucleotides 334 -375 of IS629 with a portion of orfA and the T 4 G motif was fused outof-frame to the sequence encoding His 6 -tagged maltose binding protein (MBP-His 6 ), generating pET629T 4 GMBP (Fig. 3A). On this plasmid, the MBP-His 6 would be translated only when a Ϫ1 frameshift had occurred. The fused gene was driven by the T7 promoter under the control of the lac operator. After IPTG induction, proteins were purified with nickel affinity column chromatography, electrophoresed on an SDS-polyacrylamide gel (Fig. 3B, left panel), and immunoblotted with anti-MBP antibodies (Fig. 3B, right panel). The 32-kDa band of the expected frameshifted product and two additional bands were seen. These two additional bands could be degradation products of fusion protein. The 32-kDa band was isolated from the gel and then subjected to LC-MS/MS analysis. A peptide with the sequence GSoMADIGSAYFCEGGVRPPLEIHR was identified (Fig. 3C). This sequence was the resultant product from a Ϫ1 frameshift that took place at the T 4 G motif.
A similar experiment was performed to examine whether a Ϫ1 frameshift indeed occurred at the A 4 T motif. Nucleotides 367-458 containing the A 4 T motif of IS629 was fused out-of-frame to the sequence encoding MBP-His 6 to generate pET629A 4 TMBP (Fig. 4A). pET629A 4 TMBP was then introduced into E. coli, and the expressed protein was analyzed by gel electrophoresis (Fig. 4B,  left panel), immunoblotting (Fig. 4B, right panel), and LC-MS/MS analysis. When the expected 34-kDa band was analyzed, a peptide with the sequence DIGSLWKK-MMPLL was found (Fig. 4C), indicating that a Ϫ1 frameshift had occurred at the A 4 T motif.
Effects of IS629-encoded Proteins on IS629 Transposition-The results described above indicate that both the T 4 G and A 4 T motifs can mediate a Ϫ1 frameshift, suggesting that in addition to OrfA and OrfB, two transframe proteins, OrfABЈ and OrfAB, are also produced. Experiments were then performed to investigate effects of these proteins on IS629 transposition. A plasmid (pMini629) carrying a mini-IS629 composed of the kanamycin resistance gene flanked by terminal repeats of IS629, IRL (nucleotides 1-29) and IRR (nucleotides 1,280 -1,310), was first constructed. DNA fragments containing various IS629 genes without terminal repeats, including OrfA, OrfB, OrfAB, and OrfABЈ, were then inserted immediately upstream from the mini-IS629 (Fig. 5).
To determine the function of OrfABЈ, the T 4 G sequence was changed to T 5 G by inserting an extra thymine residue to generate pMini629ABЈ(T5A4) so that OrfABЈ would be produced without frameshifting. Similarly, an adenine residue was inserted into the A 4 T motif, changing it to A 5 T to generate pMini629AB(T4A5) so that OrfAB would be produced without frameshifting. No other changes in IS629 sequence on these two plasmids were made; therefore, the OrfABЈ and OrfAB proteins produced by pMini629ABЈ(T5A4) and pMini629-AB(T4A5), respectively, were of native form.
In both pMini629AB(T4A5) and pMini629ABЈ(T5A4), the translation initiation codon for orf B is intact and may produce the OrfB protein. To avoid effect of OrfB on IS629 transposi-tion, the two ATG codons located at the beginning of orf B in these two plasmids was changed to ATAATC. In addition, the A 4 T motif was mutated to TCGAT, resulting in pMini629ABЈ(T5) so that OrfABЈ would be produced without frameshifting. Similarly, the T 4 G motif was changed to TACTG, generating pMini629AB(A5), which would produce OrfAB without frameshifting. The transposition frequency of the mini-IS629 on pMini629ABЈ-(T5) was determined to be (42.2 Ϯ 5.9) ϫ 10 Ϫ7 very similar to that of pMini629ABЈ(T5A4), which had a transposition frequency of (43.5 Ϯ 3.8) ϫ 10 Ϫ7 (Fig. 5). The mini-IS629 on pMini629AB(A5) transposed at a frequency of (0.8 Ϯ 0.1) ϫ 10 Ϫ7 very similar to that on pMini629AB-(T4A5), which had a transposition frequency of (0.6 Ϯ 0.1) ϫ 10 Ϫ7 (Fig.  5). These results indicated that these mutations did not affect the functions of OrfAB and OrfABЈ and confirmed that OrfABЈ plays a major role in IS629 transposition and that OrfAB alone does not mediate IS629 transposition.
In the experiments described above, OrfAB and OrfABЈ were artificially produced without frameshifting. To investigate effects of these two proteins that were produced by frameshifting on IS629 transposition, the transposition frequency of the mini-IS629 on pMini629ABЈ-AB-A-B, which expresses all four IS629 proteins, was examined and determined to be (70.3 Ϯ 12.2) ϫ 10 Ϫ7 (Fig. 5). To abolish the synthesis of OrfB, the two ATG codons located at the beginning of orfB was changed to ATAATC, generating pMini629ABЈ-AB-A. The transposition frequency of the mini-IS629 on pMini629ABЈ-AB-A was increased by 1.8-fold (from (70.3 Ϯ 12.2) ϫ 10 Ϫ7 to (128.9 Ϯ 20.9) ϫ 10 Ϫ7 ) (Fig. 5) when the OrfB protein was not expressed, suggesting that OrfB negatively regulates IS629 transposition. To abolish the synthesis of OrfAB, the A 4 T motif was mutated to TCGAT, resulting in pMini629ABЈ-A. Similarly, the T 4 G motif was changed to TACTG to prevent the synthesis of OrfABЈ, generating pMini629AB-A. The transposition activity of the mini-IS629 on pMini629ABЈ-A was profoundly diminished (from (128.9 Ϯ 20.9) ϫ 10 Ϫ7 to (2.6 Ϯ 0.9) ϫ 10 Ϫ7 ) (Fig. 5) when the OrfAB protein was not expressed. This result suggests that OrfAB enhances IS629 transposition, although OrfAB itself did not mediate transposition as evidenced by a background transposition frequency ((0.6 Ϯ 0.1) ϫ 10 Ϫ7 ) (Fig. 5) when the transposition activity of the mini-IS629 on pMini629AB-A was assayed.
Stabilization of OrfABЈ by OrfAB-The experiments described above demonstrated that the OrfAB protein is not a transposase but has the ability to enhance transposition of IS629. Because transposase stability may affect the transposition activity of a certain transposable element (13,14), the effect of OrfAB on the stability of OrfABЈ was examined by pulse-chase experiments. After IPTG induction, IS629 proteins encoded by pET629A-ABЈ, pET629A-AB, and pET629A-ABЈ-AB were labeled with [ 35 S]methionine for 10 min and then chased with excess amounts of non-radioactive methionine. The half-life of OrfABЈ in the presence and absence OrfAB was then measured by determining the ratio of radioactive OrfABЈ to that of OrfA every 30 min up to 120 min (Fig. 6). OrfA was used as the reference, because it is constitutively expressed from all plasmids used in this experiment.
In the absence of OrfAB, the ratio of OrfABЈ to OrfA was 0.53% at the zero time point, and 0.2, 0.17, 0.15, and 0.14% at the 30-, 60-, 90-, and 120-min time points, respectively (Fig. 6A). From these data, the half-life of OrfABЈ was determined to be ϳ30 min. The ratios of OrfAB to Orf A were 2.0, 1.99, 1.38, 0.9, and 0.69% at 0-, 30-, 60-, 90-, and 120-min time points (Fig. 6B), indicating that the half-life of OrfAB was ϳ90 min. Because OrfABЈ and OrfAB are identical in size, they cannot be separated by the gel electrophoresis used in this experiments. Therefore, the ratio of OrfABЈϩOrfAB to OrfA was measured and was determined to be 4.53, 4.38, 4.3, 3.5, and 2.7% at 0-, 30-, 60-, 90-, and 120-min time points, respectively (Fig. 6C). From this result, the half-life of OrfABЈϩOrfAB was determined to be 120 min. Therefore, the half-life of OrfABЈ was much longer in the presence of OrfAB, indicating that OrfAB has the ability to stabilize OrfABЈ. The results suggested that OrfAB may bind to OrfABЈ to form hetero-multimers that are more stable than homo-multimers of OrfABЈ or OrfAB.
Interaction of OrfAB with OrfABЈ-The possibility that OrfAB binds and stabilizes OrfABЈ was then examined by an E. coli two-hybrid assay in which interaction between bait and target proteins activates the lacZ reporter gene enabling the cells to produce ␤-galactosidase. In this experiment, OrfABЈ or OrfAB was fused to the bait protein on pBT or to the target protein on pTRG. To allow expression of OrfABЈ, the T 4 G sequence was changed to T 5 G, and the 1.2-kb DNA fragment containing this mutation without the terminal repeats of IS629 was cloned into the NotI and EcoRI sites on pBT or pTRG fusing OrfABЈ to the cI protein or to RNAP␣, generating pBT-ABЈ or pTRG-ABЈ. Similarly, the A 4 T sequence was changed to A 5 T to allow expression of OrfAB without frameshifting, and the 1.2-kb DNA fragment containing this mutation was cloned into the NotI and EcoRI sites on pBT or pTRG to generate pBT-AB or pTRG-AB. Different pairs of pBT-and pTRG-derived plasmids were then introduced into E. coli XL-1 Blue MRFЈ, and the co-transformants were assayed for ␤-galactosidase activity. As shown in Fig. 7, negative control cells containing no plasmids had a basal level of ␤-galactosidase activity of 17.5 Ϯ 0.4 units. Cells containing plasmid pairs (pBT/pTRG, pBT/pTRG-ABЈ, pBT/pTRG-AB, pBT-ABЈ/pTRG, and pBT-AB/pTRG) that did not express both OrfABЈ and OrfAB or expressed only one of the two did not have significantly elevated levels of ␤-galactosidase activity. In contrast, cells containing plasmid pairs (pBT-ABЈ/pTRG-ABЈ, pBT-ABЈ/pTRG-AB, pBT-AB/pTRG-ABЈ, and pBT-AB/pTRG-AB) that expressed OrfABЈ, OrfAB, or both OrfABЈ and OrfAB had ϳ3 times as much ␤-galactosidase activity as the negative controls. These results indicate that OrfABЈ and OrfAB can bind to each other or to themselves.

DISCUSSION
In this study, we showed that the two putative translational frameshift signals, T 4 G and A 4 T, located near the 3Ј-end of orfA of IS629 are functional. Using the lacZ gene as a reporter, we demonstrated that each of these two motifs can mediate a Ϫ1 translational frameshift (Fig. 2), resulting in the production of two transframe proteins OrfAB and OrfABЈ. This Ϫ1 translational frameshift mediated by the T 4 G or the A 4 T motif was confirmed by the existence of the transframe products (Figs. 3 and 4). Therefore, IS629 has the potential to encode four different proteins, including OrfAB, OrfABЈ, OrfA, and OrfB. These proteins were expressed either alone or in combinations in E. coli to examine their ability to mediate IS629 transposition. The transframe protein OrfABЈ alone was sufficient for IS629 transposition (Fig. 5), indicating that OrfABЈ is the transposase of IS629. Simultaneous production of both OrfABЈ and OrfAB increased the transposition activity of IS629, whereas production of OrfAB alone did not mediate IS629 transposition. These results suggest that the OrfAB protein is not a transposase but can enhance the transposition of IS629. OrfAB was shown by the bacterial two-hybrid assay to have the ability to bind OrfABЈ (Fig. 7), and binding of OrfAB to OrfABЈ was shown to increase the half-life of OrfABЈ (Fig. 6). Increase in the half-life of the transposase has been shown to enhance the transposition of IS903 (15). Therefore, the stabilization of OrfABЈ  by OrfAB would be a mechanism by which IS629 positively regulates its transposition. This type of regulation has not been found in other members of the IS3 family.
The ability of OrfAB and OrfABЈ to bind to each other or to themselves would allow them to form multimers. In many transposable elements, multimerization of transposase forms a stable transpososome, which is essential for transposition (16 -20). For phage Mu, the transposase MuA is a monomer in solution, but its active form in the transpososome is a tetramer (21,22). In Tn7, the transposase is a hetero-multimer of TnsA and TnsB (23,24). The transposase (OrfAB) of IS911 has a leucine zipper motif that is involved in the multimerization of the transposase and in the binding of the transposase to terminal repeats (20). Many IS3 family members have a leucine zipper motif located near the C-terminal end of OrfA (13,20). The leucine zipper motif was predicted to be present with a very high probability (0.99) in both OrfABЈ and OrfAB of IS629 in the same region (residues 65-95) by the programs COILS and PEPCOIL (25,26). It is possible that OrfABЈ and OrfAB of IS629 interact with each other through the leucine zipper motifs. The predicted region (nucleotides 247-337) that can form a coiled-coil leucine zipper is located 5 bp upstream from the T 4 G frameshift motif; therefore, the protein-protein interaction ability of OrfABЈ and OrfAB mediated by this leucine zipper motif would not be changed by frameshifting.
A putative promoter and the Shine-Dalgarno sequence have been located upstream of orfA at nucleotides 1-32 and 40 -47 of IS629, respectively (3). Although no such sequences are present in the upstream region of orf B (3), results of this study suggest that OrfB is produced (Fig. 2). Several members of the IS3 family. including IS150, IS911, and IS3, have been shown to produce the OrfB protein by various mechanisms. In IS150, the OrfB protein is produced by a Ϫ1 frameshifting event (27), whereas the OrfB protein of IS911 is translated using an unusual initiation codon AUU (28). In IS3, the OrfB protein is produced by translational coupling, which is triggered by a pseudoknot structure located in the overlapping region between orfA and orf B (29). Because a secondary structure similar to this pseudoknot is also present in IS629 at the corresponding position, it is possible that the OrfB protein of IS629 is translated by the same mechanism.
The initiation codon for orf B is essential for IS3 transposition (30). When this initiation codon is changed to ATA, the transposition activity of IS3 is abolished due to decreased stability of the pseudoknot, leading to the loss of the transposase OrfAB (29). In contrast, the transposition ability of IS629 was not affected when the initiation codon of orf B was mutated (Fig. 5). The OrfB protein of IS3 is known to enhance the inhibitory activity of OrfA in transposition, although IS3 OrfB itself has no inhibitory activity (31). In this study, IS629 OrfB was found to have a negative effect on transposition (Fig. 5). Sequence analysis reveals that the OrfB of IS3 family members carry a DD(35)E motif, which is also present in many retroviral integrases and various other transposases (32). This DD(35)E motif is the catalytic domain of most transposases and integrases. It is required for the strand transfer reaction during transposition or inte-gration. The region containing the DD(35)E motif has been shown to mediate multimerization of a retroviral integrase (33), the OrfB protein of IS911 (20), and the transposase of IS50 (19). Because OrfB, OrfABЈ, and OrfAB of IS629 share the same DD(35)E motif, it is possible that OrfB exerts its negative regulatory effect on IS629 transposition by binding to OrfABЈ or OrfAB. Binding of OrfB to OrfABЈ would interfere with the formation of transpososome or the catalytic reaction in strand transfer during transposition. On the other hand, OrfAB will not bind and stabilize OrfABЈ if OrfB is bound to OrfAB. Because the initiation codon of orf B overlaps the second frameshift signal, which is responsible for production of OrfAB, it is also possible that translation initiation of orf B adversely affects the frameshifting, and thus less OrfAB is available to stabilize the transposase of IS629. The function of IS629 OrfA was not investigated in this study. In IS911, OrfA forms hetero-multimers with the transposase OrfAB via a leucine zipper to enhance the inter-molecular transposition (34). Because IS629 OrfAB contains the same leucine zipper as OrfA, it is possible that OrfAB acts similarly to that of IS911 OrfA to modulate IS629 transposition.
Analyses of nucleotide sequences of IS629 from various organisms revealed two different types of IS629 sequences. These two sequences differ by one base located at nucleotide position 360 between the T 4 G and A 4 T motifs. The presence of a T residue at this position creates a TGA stop codon in the Ϫ1 reading phase, which will render IS629 unable to produce the transposase OrfABЈ; such IS629 would not be transposable. Among the 60 IS629 sequences we have analyzed, 55 sequences have a C and 5 sequences have a T residue at this position (35)(36)(37)(38)(39)(40)(41)(42). Therefore, the majority of IS629 elements in nature are transposable. The significance for the existence of nonfunctional IS629 elements remains to be investigated.