SEARCH

SEARCH BY CITATION

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References

DNA processing reactions often involve multiple components acting in concert to achieve the desired outcome. However, it is usually difficult to know how the components communicate and cooperate to orchestrate an ordered series of events. We address this question in the context of the Tn10 transposition reaction, in which the DNA cleavage and joining events occur within a higher-order complex containing a transposase dimer, two transposon ends and the DNA-bending host-factor IHF (Integration Host Factor). Previously it was shown that the complex is asymmetric. The α side consists of an IHF protomer initially immobilized by a DNA-loop, but subsequently used to promote conformational changes required for the cleavage steps. The β side of the complex was considered to fulfil a more passive role. Here we show that the α side of the complex promotes coupled conformational changes at both transposon ends, while the α and β sides communicate and cooperate to dominate different phases of the transposition reaction. Together, these effects provide for a robust response to critical changes in the transposon end. These findings also explain the intriguing genetic phenotypes of a series of previously reported Tn10 mutants and have consequences for the evolution of new elements.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References

Tn10 is a non-replicative composite transposon. It consists of two inverted copies of the insertion sequence IS10, which cooperate to mobilize the intervening tetracycline-resistance genes (Kleckner, 1989; Chalmers et al., 2000; Haniford, 2002). Transposition starts with the assembly of a synaptic complex, or transpososome, containing the IS10-Right-encoded transposase, the host-encoded DNA bending protein Integration Host Factor (IHF) and two transposon ends. Within this complex, both transposon ends are liberated from the flanking DNA by three sequential one-step transesterification reactions. Initially, water acts as the nucleophile to expose a free –OH group at the end of the transposon. This group then attacks the opposite DNA strand forming a hairpin intermediate, which is resolved by a second nucleophilic attack using water. The reaction is completed when the complex captures a new target DNA, followed by integration of the element involving a fourth direct transesterification reaction (Haniford, 2002).

The ends of IS10 are referred to as inside and outside, as defined by their positions in Tn10 (Kleckner, 1989). In both cases, the inverted repeats at each end (bp +1 to +22) encode a binding site for transposase. However, the key difference between the outside-end and the inside-end is the presence of a consensus binding site for IHF at bp +30 to +42 of the outside-end. In the absence of negative supercoiling, assembly of the synaptic complex requires specific IHF binding to at least one of the two partner ends. This is because the DNA bend provided by IHF performs a structural role, facilitating the DNA loop formed by the so-called ‘subterminal’ contacts between transposase and sequences distal to the IHF binding site (Fig. 1A; Crellin and Chalmers, 2001). This DNA-loop is necessary both for efficient assembly of the complex and also to promote some of the later intermediate steps of the transposition reaction (Sewitz et al., 2003; Crellin et al., 2004; Liu et al., 2005).

image

Figure 1. Models for the structural and functional asymmetry of the Tn10 transpososome. A. A schematic representation of the molecular model of the PEC assembled with different transposon ends (Crellin et al., 2004). In this representation, the target binding groove is at the top. The large arrowhead is at the junction between the transposon and the flanking DNA. The vertical ticks are the subterminal transposase contacts determined by hydroxyl radical footprinting, which extended out to bp +80, beyond the protein–DNA contacts depicted in the model. The 73 bp transposon end lacks the most distal set of contacts. The 52 and 56 bp transposon ends lack all of the contacts distal to the target binding groove. B. The assembly and unfolding of the Tn10 transpososome (Crellin et al., 2004). IHF binds specifically to the outside-end of Tn10 and activates assembly of the PEC. In the absence of divalent metal ion, competitor DNA or heparin treatment strips the IHF from the β side of the complex. On the α side of the complex, IHF remains locked in position until released by the addition of divalent metal ion. Further details are given in the text. Arrowhead, transposon end; hatched green oval, IHF; magenta and turquoise ovals, transposase (T′ase); ffbPEC, fully folded bottom-PEC; sfbPEC, semi-folded bottom-PEC; tPEC, top-PEC; Me++, divalent metal ion. C. Multiple pathways for the cleavage of Tn10 transpososome. The cleavage step is biased towards the αSEB isomer (Liu et al., 2005). The extent of the bias measured in different experiments is somewhat variable and ranges from 15% to 30% in favour of the αSEB intermediate. Assuming a 50:50 distribution in the case of random cleavage, a bias of 15–30% will yield 65–80% of the αSEB intermediate.

Download figure to PowerPoint

The outside-end and inside-end sequences have been subjected to extensive mutagenesis to investigate the phenotype of mutations at intermediate stages during progression of the transposition reaction in vivo (Huisman et al., 1989). These genetic studies revealed two classes of mutants corresponding to two physical domains in the transposon end. Mutations at bp +1 to +3 resulted in defects in the reaction chemistry, while mutations at bp +6 to +13 affected an earlier step by reducing binding site recognition by transposase (Haniford and Kleckner, 1994; Sakai et al., 1995; Kleckner et al., 1996). Mutations in both regions dramatically reduced the frequency of transposition when present on both transposon ends. However, many of the mutations could be rescued by a wild-type partner, and many varied in severity depending on whether they were present on an inside or an outside-end (Huisman et al., 1989). The ability of one transposon end to rescue the other suggested that the reactions at the two ends do not proceed independently, and hinted that each side of the complex may be functionally coupled. This may be a general feature of transposition reactions. Earlier genetic evidence, for the distantly related IS903 element, likewise showed that the transposon end was organized in two domains and that a mutation could be rescued by a wild-type partner (Derbyshire et al., 1987; Derbyshire and Grindley, 1992).

If the rescue of mutations by wild-type partner is a general feature of transposons, what function, if any, does this mechanism fulfil? One possibility is that it accelerates the evolution of transposable elements. For example, the diversification of the flanking inverted repeat sequences might otherwise be expected to be a very slow process because it requires the simultaneous acquisition of mutations at both transposon ends and a concomitant change in the specificity of the transposase. The rescue of deleterious mutations by a wild-type partner immediately suggests a mechanism to accelerate this process. Following the acquisition of a mutation in one transposon end, the rescue effect will allow the element to remain active and provide a window of opportunity for the acquisition of suppressor mutations in the partner end and/or the transposase gene. This mechanism would naturally accelerate the otherwise constrained evolution within families of transposons.

Here we provide direct evidence of structural and functional coupling between the two sides of the Tn10 transpososome. We have determined the mechanistic nature of the defects caused by specific mutations on one or other side of the complex, and show how these are rescued by a wild-type partner. We also present evidence that both sides of the complex become dominant in rescue at different stages of the reaction. These findings offer an explanation for the unusual observations previously reported by the Kleckner laboratory (Huisman et al., 1989). Furthermore, these findings provide insight into the mechanism which underlie the robust mechanistic response to ends-mutations that have long been postulated to aid in the evolution of new transposable elements (Sasakawa et al., 1983; Derbyshire et al., 1987).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References

Experimental background

During Tn10 transposition, both transposon ends have identical IHF binding sites adjacent to the inverted repeats. However, we have previously shown that the transpososome is asymmetric and that the IHF sites do not have identical structural or functional roles within the complex (Fig. 1B; Sewitz et al., 2003; Crellin et al., 2004). We therefore refer to these as the α and β IHF binding sites. IHF binding is required only at the α site to stimulate assembly of the paired ends complex (PEC). After assembly of the complex in vitro, IHF is immobilized and locked into the α side of the complex until it is released by the addition of the catalytic metal ion (Sewitz et al., 2003). In contrast, an IHF binding site is not required on the β side of the complex during assembly. However, if one is present, it remains unlocked and IHF is free to associate or dissociate, independently of the presence or absence of divalent metal ion. As the IS10 elements within Tn10 have IHF binding sites at one end only, the α and β sides of the complex therefore appear to represent the outside and inside-ends of IS10 respectively. This structural asymmetry has been shown to affect the cleavage step of the reaction, which is slightly biased towards initiating on the α side of the complex (Fig. 1C; Liu et al., 2005).

We have also described a substantial conformational change in the transpososome, which renders the nucleotides at bp +1 and −1 on the transferred strand of the transposon end hypersensitive to hydroxyl radical footprinting (Crellin and Chalmers, 2001). Although the precise significance of the hypersensitivity is unknown, it almost certainly reflects a conformational change in preparation for the excision step of the reaction. Excision starts with a nick located between the two hypersensitive nucleotides, and continues with formation and resolution of the hairpin intermediate. This conformational change was originally observed in response to the addition of Ca++, an analogue of the catalytic metal ion (Fig. 2, panels 1 and 2). However, further experiments showed that hypersensitivity could not be attributed directly to the effects of Ca++ (Crellin et al., 2004). Instead, Ca++ was required to license the release of IHF from the α side of the complex, converting the bottom-PEC (bPEC) to the top-PEC (tPEC) as the transposon arm(s) unfolded. It was therefore the unfolding of the folded α transposon arm that produced the conformational change, rather than the divalent metal ion per se. This mechanical view of the reaction is supported by the phenotype of point mutations in transposase that alter transpososome unfolding and inhibit the hairpin formation step (Humayun et al., 2005). The amino acid residues in question are distant from the active site and, based on the Tn5 cocrystal structure, are predicted to interact directly with the folded transposon arm distal to the IHF binding site.

image

Figure 2. Coupled conformational changes on either side of the PEC. The indicated complexes were assembled and footprinted with hydroxyl radicals as described in Experimental procedures. The experiment reveals that the conformational change responsible for the hyperreactivity is present on both sides of the complex and is produced by unfolding of the α transposon arm. The radioactive label was either present on both sides of the complex or was placed specifically on the α or the β side of the complex, as indicated by the asterisk. The even-end cannot form a PEC on its own because it cannot interact with IHF. In reactions containing a labelled even-end and a cold outside-end, only the mixed complex is therefore detected. In reactions containing a labelled outside-end and a cold even-end, mixed complexes are formed along with complexes containing two outside-ends. In this case, the different complexes are resolved at the gel shift stage of the experiment because the DNA fragments are of different sizes. The asterisk indicates the location of radioactive label.

Download figure to PowerPoint

Structural coupling between opposite sides of the transpososome

As described above, the α and β transposon arms have different structural and functional roles during assembly and subsequent unfolding of the Tn10 PEC. These observations raise further interesting questions regarding whether the asymmetry of the transposon arms is reflected in the structure of the active site. For example, whether or not the structural change responsible for the appearance of hydroxyl radical hypersensitivity is symmetric on each side of the complex. We are able to address this question using a mixed complex in which only one of the two transposon arms contains an IHF binding site (Fig. 2).

We have previously developed the technique for assembly of mixed complexes using an artificial transposon end in which the DNA between bp +19 and +47 was replaced by tandem repeats of the bases 5′-CTGA (Crellin and Chalmers, 2001). We refer to this as an ‘even-end’ because the hydroxyl radical cleavage profile in the presence of a saturating concentration of IHF is completely even, indicating that IHF binding is negligible. The even-end fails to form PEC on its own because it lacks an IHF binding site. However, it is efficiently recruited into a mixed PEC when the assembly reaction is supplemented with an outside-end fragment. In these experiments the even-end is the equivalent of the inside-end of IS10 which also lacks an IHF binding site. In such mixed complexes, the outside-end DNA fragment, with its IHF binding site, is always located on the α side of the PEC.

As alluded to above, the conformational change responsible for the hydroxyl radical hypersensitivity takes place after unfolding of the α transposon arm, as the bPEC is converted to the tPEC. When the α transposon arm is ≤ 73 bp, this can be achieved in the absence of divalent metal ion by using heparin instead of competitor DNA to remove the IHF (Fig. 1A). As the aim of the experiment was to determine whether the hypersensitivity is symmetric or asymmetric, aliquots of the mixed complexes were treated with hydroxyl radicals either before or after treatment with heparin to promote IHF dissociation and unfolding of the α transposon arm. The complexes were then purified using the standard electrophoretic mobility shift assay (EMSA) and the hypersensitivity of the transposon ends was determined by analysis on a DNA sequencing gel (Fig. 2). The experiment was performed twice with the radioactive label present either on the outside-end or on the even-end (equivalent to the α and β sides of the complex respectively). When the radioactive label was on the α side of the complex, the hypersensitivity was identical to the standard bPEC with two outside-end fragments (compare panels 2 and 4). Likewise, identical hypersensitivity was observed when the radioactive label was present on the β side of the complex (compare panels 2 and 6). Taken together, these results demonstrate that unfolding of the α transposon arm promotes a conformational change on both sides of the complex. This suggests that there is communication or structural coupling between the active sites at each transposon end. This idea is supported by further evidence presented below.

Functional coupling between opposite sides of the transpososome

Hydroxyl radical footprints have shown that transposase contacts the DNA up to 80 bp inside the transposon arm (Fig. 1A; Crellin and Chalmers, 2001). These extensive contacts are facilitated by the 180° IHF-induced bend that folds the transposon arm during PEC assembly. Previous structural modelling suggested that the subterminal contacts between bp +48 and bp +80 were located on either side of the transposase dimer interface, providing a possible mechanism to couple opposite sides of the complex (Fig. 1A; Crellin and Chalmers, 2001; Sewitz et al., 2003).

When a highly truncated 52 bp outside-end is used in the transposition reaction, PEC assembly is very inefficient (Crellin and Chalmers, 2001). Once formed, the PEC is also functionally defective and the reaction can progress only as far as the single end break (SEB) stage. This is a particularly interesting phenotype as the hairpin intermediate on the cleaved transposon end remains unresolved (Crellin et al., 2004). These defects appear to be the result of abolishing the contacts between the transposon arm and the transposase monomer most distal to the dimer interface (illustrated in Fig. 1A). The location of the missing contacts on one side of the dimer interface, taken together with the generation of a SEB product, suggests that this phenotype may result from a failure of communication or coupling between the active sites within the complex. We were again able to address this problem using the mixed complex in which only one transposon end has an IHF binding site. In this way we are able to ask whether a short transposon end imparts the SEB phenotype only when present on one specific side of the complex, or whether it imparts an intrinsic defect irrespective of its location.

When the PEC is assembled using two standard outside-ends, addition of Mg++ induces cleavage of the flanking DNA and the PEC is converted to a double end break (DEB) (Fig. 3, panel 1) (Sakai et al., 1995). If a mixed complex is assembled by replacing one of the outside-end partners with an even-end that is unable to interact with IHF, the PEC behaves in the same way and addition of Mg++ produces DEB (Fig. 3 panel 4). In the method used to assemble these mixed complexes, a radioactively labelled even-end is supplemented with a cold outside-end. Note that complexes containing two outside-ends are also formed in such a mixture, but that these are invisible because they contain no radioactive label.

image

Figure 3. A hairpin resolution defect reveals functional coupling on either side of the PEC. Transposition reactions were analysed using the standard EMSA and visualized by autoradiography. To form the mixed complexes, the unlabelled outside-end DNA fragment was added in excess, which accounts for the apparently low level of PEC in some lanes. Transposition reactions were incubated for 3 h which is sufficient time for the standard reaction, with full-length transposon arms (panel 1), to reach completion. Reactions with truncated outside-end fragments on the α side of the complex fail to reach completion (panels 2 and 3). Long OE, outside-end > 85 bp; long EE, even-end > 85 bp; short OE, short outside-end < 56 bp; short EE, short even-end < 56 bp. The asterisk indicates the location of the radioactive label.

Download figure to PowerPoint

As explained above, a PEC with two short (52 bp) outside-ends fails to complete the cleavage step of the reaction when Mg++ is added (Fig. 3, panel 2). In this case no DEB complex is produced. However, some flanking DNA is released owing to the cleavage of one transposon end to produce the SEB complex (Crellin et al., 2004).

If the two active sites are in communication, or are mechanistically coupled, then the cleavage defect of the short outside-end PEC in panel 2 may be rescued by a full-length partner. Furthermore, the rescue may also depend on whether the full-length partner is on the α or β side of the complex. Of course, the structural considerations outlined above would suggest that rescue is more likely for a full-length partner that can fold and unfold normally on the α side of the complex.

To address this issue, a mixed PEC was assembled in which a short outside-end is paired with a long even-end partner (Fig. 3, panel 3). As the even-end is unable to interact with IHF, it will always be present on the β side of the complex. When Mg++ is added, no DEB is produced and the reaction is almost identical to that in panel 2 where the complex contains two short (52 bp) outside-ends. This result demonstrates that a full-length transposon arm on the β side of the complex is unable to rescue the cleavage defect produced by the truncated transposon arm on the α side. Flanking DNA is again detected owing to the cleavage at one transposon end that produces the SEB.

The symmetry of the SEB complex was examined by reversing the labelling strategy (not shown). This experiment revealed that the product of this reaction is a mixture of α and βSEBs. Thus, even though the defect is produced by truncation of the α transposon arm, the SEB phenotype can be manifest on either side of the complex.

Finally, a mixed PEC was assembled in which a short (52 bp) even-end on the β side of the complex was paired with a full-length α outside-end (Fig. 3, panel 5). When Mg++ is added, this complex behaves normally and completes the cleavage steps of the reaction, releasing the flanking DNA to produce the DEB complex. This is an important result because the presence of the 52 bp even-end in the complex shows that a truncated transposon arm is not sufficient in itself to confer a cleavage defect. This experiment therefore demonstrates functional coupling between the two active sites in the PEC that depends on a full-length IHF-folded transposon arm on the α side of the complex.

The base pair 1G mutation

Based on the genetic phenotypes of mutations, the inverted repeat at the end of Tn10 can be divided into two domains (Huisman et al., 1989; Haniford and Kleckner, 1994; Sakai et al., 1995; Kleckner et al., 1996). Transposase binding is mediated by sequence-specific interactions with bp +6–13. In contrast, mutations at bp +1–3 cause defects in the chemical steps of the reaction, subsequent to synapsis of the transposon ends. Earlier genetic evidence had demonstrated that some of these mutations could be partially rescued by a wild-type partner (Huisman et al., 1989). We were therefore interested to discover whether the physical basis for these intriguing genetic observations is provided by the asymmetry and communication between the α and β sides of the complex. As a starting point we chose the 1C to G mutation (subsequently referred to as 1G). We chose this mutation because it has a strong phenotype in vivo when present at both ends, reducing transposition to 0.07 of the wild-type level. However, when present at only one transposon end, the wild-type partner rescues the activity back to 0.7 of the wild-type level.

As a reference point, a standard bPEC was assembled with two wild-type outside-end fragments (Fig. 4A, left panel). This is prefixed by the term fully folded (ff) because both transposon arms remain bound by IHF throughout the reaction (see illustration in Fig. 1B). When the reaction is initiated by addition of Mg++, the SEB intermediate appears early and is gradually converted into DEB product during the time-course. The DEB complex is the end-point of the reaction as no suitable DNA was included in these reactions to act as a target or IHF competitor. When both of the transposon ends contain the 1G mutation (Fig. 4A, right panel), the kinetics of SEB formation are also slow. After 15 min, SEB is clearly visible in the wild-type reaction, but very little has appeared in the 1G reaction. At later time points no DEB is produced, and the reaction appears to stall at the SEB intermediate stage (Fig. 4A, indicated by the letter X). This is consistent with previous work with this mutation showing that it is proficient for the first nick in the reaction, but partially defective for hairpin formation (Allingham and Haniford, 2002).

image

Figure 4. Cleavage and hairpin resolution defect of the 1G transposon end mutation. A and B. The indicated complexes were assembled using different combinations of wild-type and mutant transposon ends as indicated above and below the panels. The reactions were initiated by the addition of Mg++, stopped at the indicated time point by addition of EDTA, and visualized using the standard EMSA. All of the DNA substrates have full-length transposon arms (> 85 bp), which either have the wild-type sequence or have the 1G mutation at bp +1 of the transposon end, as indicated. In (A), the complexes contain a pair of identical outside-ends. In (B), mixed complexes were assembled with outside-end and even-end fragments on the α and β sides of the complex respectively. The asterisk indicates the location of the radioactive label. C. The SEB complexes labelled as X and Y in (A) and (B), respectively, were recovered from the gel by the crush and soak method and deproteinated by SDS treatment. The DNA component of the complexes was analysed on a denaturing DNA sequencing gel. Complex X contains the unreacted OE.1G substrate plus the cleavage product and the hairpin intermediate. Complex Y contains the unreacted EE.1G substrate. The cleaved transposon end on the opposite side of complex Y is not detected because it is unlabelled. D. A diagram representing the structures of the SEB complexes X and Y deduced from the DNA components visualized in (C).

Download figure to PowerPoint

Although the SEB product in this experiment indicates that the 1G defect may involve communication between the two active sites, it provides no information regarding the symmetry because it is not possible to distinguish the α and β sides when the complex contains two identical outside-ends. The experiment was therefore repeated using a mixed complex in which an outside-end is paired with an even-end partner (Fig. 4B). As the radioactive label is present only on the even-end, only mixed complexes are detected in this experiment. In this case, the mixed bPEC is prefixed by the term ‘semi-folded’ (sf) because only the α transposon arm is bound by IHF.

When the sfbPEC contains two wild-type transposon ends, the SEB intermediates appear early and are converted to the DEB product at later time points (Fig. 4B, left panel). In this experiment, with the semi-folded complex, the EMSA is able to resolve the α and β isomers of the SEB intermediate which co-migrate in the gel when using the fully folded form of the complex (the αSEB and βSEB are illustrated in Fig. 1C; Sewitz et al., 2003). The slight bias in the reaction towards the αSEB intermediate (Fig. 1C) can be seen quite clearly at the 15 min time point. When the outside-end and even-end partners both contain the 1G mutation, no DEB is produced and the reaction again stalls at the SEB intermediate stage (indicated by the letter Y in Fig. 4B, right panel). The location in the gel suggests that this is the αSEB intermediate, as will be shown in more detail below.

The SEB complexes X and Y from Fig. 4A and B were recovered from the gel, deproteinated by SDS treatment and analysed on a denaturing DNA sequencing gel (Fig. 4C). As expected, the SEB represented by band X contains both the unreacted and the cleaved outside-end of the transposon. In addition, it contains a substantial amount (83%) of the hairpin intermediate, which will accumulate if the hairpin resolution step is blocked following the double strand cleavage to release the flanking DNA. This indicates a catalytic defect because the hairpin intermediate has a short half-life in wild-type reactions, and very little would normally persist after 120 min (Kennedy et al., 1998).

The analysis of band X provides no information about the symmetry of the complex because the transposon ends on both sides are identical and cannot be distinguished. However, band Y is derived from a mixed PEC in which the radioactively labelled even-end is by definition always present on the β side of the complex. Figure 4C shows that this SEB complex contains only the unreacted even-end, with no hairpin intermediate or cleavage product detected. The cleaved transposon end in this SEB complex must therefore be on the α side of the complex. Taken together, these results show that bands X and Y represent the αSEB complex (summarized in Fig. 4D). Furthermore, the cleaved transposon end on the α side of the complex is 83% hairpin. As the hairpin intermediate is barely detectable in equivalent wild-type reactions, this indicates that the hairpin resolution step of the reaction has been severely compromised.

It is noteworthy that the SEB-hairpin product obtained with the 1G mutation is very similar to that obtained using the truncated 52 bp outside-end substrate (Fig. 3; Crellin et al., 2004). The key difference is that the 1G mutation produces only αSEB, whereas the short α transposon arm produces a mixture of α and βSEBs. Thus, although the SEB-hairpin represents a mechanistically significant intermediate, the reaction is sufficiently flexible that this point can be reached by more than one pathway. (For another example of the flexibility of the reaction pathway see Fig. 7 of Liu et al., 2005.)

Rescue of the 1G mutation

Genetic experiments had previously demonstrated that a wild-type partner could rescue the phenotype of the 1G mutation (above). We were therefore interested to determine whether rescue was dependent on the location of the wild-type partner, on one side of the complex or the other. To address this point we assembled a mixed PEC in which a wild-type transposon end was paired with the 1G mutation either on the α or β side of the complex (Fig. 5A). These mixed complexes are referred to as α1G and β1G respectively.

image

Figure 5. Rescue of the 1G mutation cleavage defect. A. Mixed PECs were assembled in which wild-type and 1G mutant transposon ends were placed on either the α or the β side of the complex. Reactions were initiated by the addition of Mg++, stopped at the indicated time point with EDTA and visualized in the standard EMSA. In the left panel, the reaction fails to reach completion and produces an αSEB. In the right panel, the reaction reached completion via the βSEB intermediate. In these gels the αSEB migrates only slightly faster than the βSEB. However, the identities of these complexes are confirmed in (B). The asterisk indicates the location of the radioactive label. B. The αSEB (A, left panel) and the βSEB (A, right panel) were recovered from the gel by the crush and soak method and deproteinated by SDS treatment. The DNA component of the complexes was analysed on a denaturing DNA sequencing gel. The identities of the complexes assigned in (A) were confirmed because an unreacted even-end was recovered from the αSEB, and a cleaved even-end fragment was recovered from the βSEB complex. No hairpin intermediate is detected in either complex. C. Identical complexes to those in (A) were assembled, except that the labelling scheme was reversed, placing the radioactive label on the outside-end fragment. The complexes were purified using the standard EMSA, located by autoradiography and excised from the gel. The gel slices containing the respective complexes were soaked in reaction buffer. After incubation under reaction conditions, complexes were recovered from the gel by the crush and soak method and deproteinated by SDS treatment. The DNA component of the complexes was analysed on a denaturing DNA sequencing gel. The extent of the reaction is low because of the inherent inefficiencies of the in-gel cleavage strategy. However, both samples contained cleaved transposon ends but no signal from the hairpin intermediate, indicating that hairpin resolution had been achieved. The question mark indicates an unidentified cleavage product. D. A diagram summarizing the structures of the αSEB and the DEB complexes in (A), as deduced from the DNA components identified in (B) and (C). No hairpin intermediates are present.

Download figure to PowerPoint

In the β1G situation, the reaction stalls at the SEB stage (Fig. 5A, left panel). A wild-type partner on the α side of the complex is therefore unable to rescue the reaction. The location in the gel suggests that the product is an αSEB and this will be demonstrated in more detail below.

In the α1G situation, when the wild-type partner is placed on the β side of the complex, the mutant phenotype is rescued and the complexes complete the cleavage steps of the reaction, producing DEB (Fig. 5A, right panel). A wild-type partner on the β side of the complex is therefore able to rescue the reaction. This was surprising as it is the opposite placement of the wild-type partner needed to rescue reactions with the truncated 52 bp transposon end (Fig. 3). There are also two notable differences compared with the wild-type reactions presented in Fig. 4B. First, the kinetics of the reaction are much slower, which is the reason the time-course of the reaction has been extended in this experiment. Second, the vast majority of the SEB intermediate appears to be the β isomer. It therefore appears that the normal slight bias in the reaction towards the αSEB intermediate has been completely reversed in favour of the βSEB. Presumably, both of these effects arise from an inhibition of cleavage on the α side by the 1G mutation, and the promotion of cleavage on the β side of the complex by the wild-type sequences.

The slightly unusual nature of these results prompted us to confirm the identities of the SEB isomers assigned in Fig. 5A. The respective bands were therefore recovered from the gel, deproteinated by SDS treatment and analysed on a DNA sequencing gel (Fig. 5B). Note that in both cases the radioactive label was present only on the even-end which is always on the β side of the complex. As expected, only unreacted even-end is present in the αSEB lane. Also, > 95% of the signal in the βSEB lane corresponds in size to the cleaved even-end. These results confirm the identities of the respective complexes. Furthermore, as no hairpin intermediate is detected in the βSEB lane, the resolution step has been achieved.

In these experiments with mixed complexes, the radioactive label is usually present only on the β side of the complex. It is therefore unknown whether the αSEB and the DEB products, shown in Fig. 5A, are trapped at the hairpin intermediate stage. This question can be addressed by reversing the labelling strategy for the mixed complexes, placing the radioactivity on the outside-end present on the α side of the complex. However, this approach is problematic because, in addition to the desired mixed complexes, homogeneous complexes containing two labelled outside-ends will also be detected. Although the two types of complexes can be resolved in the standard EMSA, owing to the different length of the DNA fragments employed, the products of the reaction after addition of Mg++ cannot be resolved completely or identified unambiguously. To circumvent this problem, the unreacted mixed complexes were resolved in the standard EMSA and excised in a gel slice. The gel slice was then soaked in reaction buffer to initiate catalysis. After extended incubation, the reactions were stopped by addition of EDTA, deproteinated by SDS treatment, and the DNA was recovered and analysed on a DNA sequencing gel (Fig. 5C). The in-gel reactions produced the expected αSEB and the DEB products, as judged by the presence of the unreacted and cleaved outside-end fragments. Furthermore, there was no signal from the hairpin intermediate, indicating that resolution had been achieved. Taken together these results show that the β1G and α1G complexes yield fully resolved DEB and αSEB products respectively (summarized in Fig. 5D).

The defect and rescue of the 8C mutation

Transposase binding to the transposon end is mediated primarily via sequence-specific interactions with bp +6–13 (Huisman et al., 1989; Haniford and Kleckner, 1994; Sakai et al., 1995; Kleckner et al., 1996). Unfavourable mutations at these positions are thought to affect assembly of the transpososome, prior to the first catalytic step. We have examined the behaviour of the bp +8A to C mutation (subsequently referred to as 8C). When present on both transposon ends in vivo, this mutation reduces transposition to 0.003 of the wild-type frequency. However, a wild-type partner end rescues the activity back to 0.1 of the wild-type level (Huisman et al., 1989).

A mixed bPEC was assembled with the 8C mutation present on both transposon ends (Fig. 6, panel 1). When the reaction is initiated by Mg++, most of the complexes dissociate, although a low level of SEB and DEB products are detected (compare with the wild-type control in Fig. 4B, left panel). Dissociation probably results from a combination of the weakened end binding caused by the 8C mutation and the mechanical forces that appear to operate during the cleavage step of the reaction (Crellin et al., 2004).

image

Figure 6. The cleavage defect and rescue of the 8C mutation. The indicated complexes were assembled using different combinations of wild-type and mutant transposon ends, as indicated above and below the panels. The reactions were initiated by the addition of Mg++, stopped at the indicated time point by addition of EDTA, and visualized using the standard EMSA. All of the DNA substrates have full-length transposon arms (> 85 bp), which either have the wild-type sequence or have the 8C mutation at bp +8 of the transposon end as indicated. The asterisk indicates the location of the radioactive label.

Download figure to PowerPoint

We next assembled a mixed PEC in which a wild-type transposon end was paired with the 8C mutation either on the α or β side of the complex. These are referred to as α8C and β8C when the mutant end is present on the α and β side of the complex respectively (Fig. 6, panels 2 and 3). DEB production in both of these mixed complexes is rescued compared with the fully mutant situation (compare the relative amounts of PEC and DEB at the last time point in each of the three sets of reactions). This is immediately different from the 1G mutation in which rescue is specific to a wild-type partner on the β side of the complex (Fig. 5A).

Visual assessment of the time-courses suggests that α8C is more rescued than β8C (Fig. 6, panels 2 and 3). Quantification of the gel further reveals that this is due to the fact that a larger proportion of the β8C complex present at the zero time point dissociates at the start of the reaction. This effect is particularly noticeable in the double mutant situation (panel 1). Indeed, even in the wild-type situation, a significant fraction of the PEC dissociates before cleavage in response to the addition of Mg++. Although never specifically addressed, this is a well-documented effect, observed since the earliest work in this experimental system (Sakai et al., 1995). However, four different classes of metal ion-sensitive PECs, including the 8C mutation used here, have been addressed specifically, and can all be construed as a result of weakened end-binding (Sakai et al., 1995; Kennedy and Haniford, 1996; Crellin et al., 2004). The present results suggest that although this effect operates on both sides of the complex, the β side is more sensitive.

A further unusual aspect of this experiment is that the normal bias of the reaction towards the αSEB is exaggerated in β8C (Fig. 6, panel 3), and almost completely reversed α8C (panel 2). The simplest explanation is that the SEB complex is even more sensitive than the PEC to metal ion-dependent dissociation when the 8C mutation is present on the side of the complex that cleaves first. Thus, α and βSEBs are under represented in α8C and β8C respectively. Finally, the products of the reactions with the 8C-mutant complexes were examined on a DNA sequencing gel (not shown). No signal corresponding to the hairpin intermediate was detected, showing that hairpin resolution had been achieved. This demonstrates that the defect caused by the bp +8 mutation does not affect the catalytic steps and is likely caused by a simple reduction in transposase affinity for the transposon end.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References

Rescue of ends mutations; the physical basis for some unusual genetic observations

Kleckner and colleagues previously used genetic techniques to investigate the phenotypes of transposon ends mutations in Tn10 (Huisman et al., 1989). They were particularly interested in whether the ends of the transposon functioned independently. As a measure of the independence they defined the term B2/A, where A is the rate of transposition when the mutation is present at both ends of the transposon, and B is the rate when it is present at only one transposon end. For mutations at almost all positions, the value of B2/A was close to 1, indicating that the ends functioned independently. The greatest exceptions were the mutations at bp +1–3, which all had values ranging between 5 and 10. The defect in these mutants was interpreted as stemming from a structural problem in the active site, and the high B2/A value as an indication of rescue by a wild-type partner. Although the underlying mechanism could not be addressed using the methods available at that time, there was clearly some appreciation of the structural and functional coupling between opposite sides of the Tn10 transpososome.

Communication between opposite sides of the transpososome during the cleavage step

Using a mixed PEC in which a 73 bp outside-end was paired with an even-end partner, the conformational change responsible for hydroxyl radical hypersensitivity at bp +1 and −1 of the transposon end was shown to be coupled between each side of the transpososome (Fig. 2). Furthermore, this effect was shown to be independent of the divalent metal ion cofactor necessary for catalysis. These results show that a cycle of folding and unfolding of the α transposon arm only is sufficient to produce a conformational change on both sides of the complex.

Structural coupling was further reflected by functional coupling between the active sites on either side of the transpososome (Fig. 3). When the PEC was assembled using truncated (52 bp) transposon arms, the complex failed to complete cleavage even after extensive incubation with Mg++ (Fig. 3, panels 2 and 3). The long even-end, which cannot interact with IHF, does not rescue this phenotype (panel 4). However, the 52 bp even-end substrate was rescued by a long outside-end partner (Fig. 3, panel 5). These results show that activation of the complex for cleavage requires one cycle of IHF-driven folding and unfolding of a single full-length outside-end located on the α side of the complex.

Processing of the DNA hairpin intermediate also depends on the symmetry of the complex, and communication between opposite sides. When present on the α side of the complex, the truncated 52 bp transposon arm slows down hairpin formation and resolution, yielding a SEB product in which most of the hairpin intermediate persists (Fig. 3 and Crellin et al., 2004). The physical connection between arm unfolding and hairpin processing is further supported by the point mutations RA182 and RA184, both of which have altered unfolding properties and block formation of the hairpin intermediate (Humayun et al., 2005).

The cocrystal structure of the related Tn5 transpososome (Davies et al., 2000) provides structural insight into the basis of the communication between each side of the Tn10 complex. There is an extensive dimer interface, and both of the transposase monomers make extensive contacts with both of the transposon ends. Indeed, two-thirds of the dimer interface is contributed by the protein–DNA interactions. Another key aspect of the structure is that each transposase subunit performs catalysis on the transposon end opposite the one with which it has the greatest amount of sequence-specific interactions. This trans-catalysis arrangement provides for an intimate structural and functional relationship between each side of the complex.

In Tn10, sequence-specific recognition by transposase is mediated primarily by bp 6–13 of the transposon ends (Haniford and Kleckner, 1994; Sakai et al., 1995; Kleckner et al., 1996). The 8C mutation was selected for the present study because it has the highest level of rescue by a wild-type partner (Huisman et al., 1989). Tn5 also provides insight into the structural basis of this effect. Base pair 8 is at the boundary between the cis and the trans interactions with the transposon end, and contacts both transposase subunits (Davies et al., 2000). The dissociation of the 8C mutant complexes, and the partial rescue by a wild-type partner, can therefore be understood in terms of its location at the heart of the dimer interface.

The 1G mutation reveals two conformational intermediates involved in hairpin resolution

When the 1G mutation is present at both ends of the transposon, the product of the reaction is an αSEB in which the hairpin intermediate persists on the broken end (Fig. 4). To understand the significance of this it is necessary to recall the four chemical steps of the reaction at each transposon end. In the first step water acts as the nucleophile to cleave the phosphodiester bond to generate the 3′-OH group at the end of the transposon. In the second step the 3′-OH group attacks the opposite strand cleaving the transposon from the flanking DNA and generating the hairpin intermediate (Kennedy et al., 1998). Hairpin resolution and integration of the transposon at the target site are essentially a reiteration of the first two steps when water and the 3′-OH again act sequentially as the nucleophiles. Although, it should be noted that the stereo-selectivity of the transesterification reaction changes before the final step (Kennedy et al., 2000). The 1G mutation therefore appears to partially block a conformational change necessary to reset the active site for the reuse of water as the nucleophile. This structural transition is presumably identical or related to that disrupted when using truncated 52 bp transposon arms which produces a very similar phenotype (Fig. 3 and Crellin et al., 2004).

In the β1G complex, when the 1G mutation is paired with a wild-type α partner, the reaction again yields an αSEB product (Fig. 5). However, analysis on a DNA sequencing gel revealed that the reaction had been partially rescued and that hairpin resolution had been achieved. This suggests that there is an additional conformational stage in the reaction that takes place between hairpin resolution on one side of the complex and the initiation of catalysis on the other.

In the α1G complex, when the 1G mutation is paired with a wild-type β partner, the reaction is fully rescued, producing DEB (Fig. 5). It is unusual that this reaction proceeds via the β isomer of the SEB intermediate. This was unexpected because in the wild-type situation cleavage is slightly biased towards the α SEB intermediate (illustrated in Fig. 1C; Liu et al., 2005). The α1G complex must produce the βSEB because the favourable wild-type sequences on the β side of the complex promote cleavage. Successful cleavage on the β side of the complex must then be communicated to the α side, promoting cleavage and hairpin resolution at the 1G mutant end.

Together with the data discussed above, these results show that the α and β sides of the complex can both dominate different aspects of the transposition reaction. Folding and unfolding of the α transposon arm is important for assembly of the complex and for promoting the coupled conformational changes required for efficient hairpin resolution. The β side of the complex, in contrast, plays a more passive role in these aspects of the reaction. However, when cleavage is blocked by an unfavourable mutation at the transposon end, the reaction can be rescued by a wild-type partner on the β side of the complex. Indeed, in this circumstance the reaction is sufficiently versatile that it can proceed via the βSEB intermediate, reversing the usual αSEB bias.

Experimental procedures

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References

Chemicals were from Sigma. Restriction enzymes and the exo Klenow fragment of DNA polymerase used for 32P-labelling were from New England Biolabs. The 32P-radiolabelled nucleotides used for end filling of restriction fragments were from Amersham Biosciences.

DNA, proteins and assembly of complexes

Transposase and IHF were expressed and purified as described previously (Chalmers and Kleckner, 1994; Lynch et al., 2003). The expression plasmids were as follows: wild-type transposase was from pRC60, which contains the transposase gene on an NdeI-BamHI fragment cloned into pET11a (Novagen); IHF was expressed from pRC188, which is identical to pPR204 obtained from Phoebe Rice. This plasmid contains the IHF operon cloned downstream of the bacteriophage T7 promoter in pET27b (Novagen).

The transposon end fragments were prepared from the following plasmids (complete sequences available on request). pRC98 encodes 87 bp of the standard IS10 outside-end sequence. It was created by cloning the BglII-SalI fragment from pNK1935 into BamHI-SalI digested pBluescript SK+ (Stratagene). pRC173 also encodes a standard outside-end of IS10. The HindIII-NdeI fragment of pLO129, encoding 64 bp of outside-end sequences, was blunt ended and cloned into EcoRV cleaved pBluescript SK+. pRC100 contains the even-end, an artificial transposon end in which the DNA between bp +19 and +47 of the IS10 outside-end was replaced by tandem repeats of the bases 5′-CTGA (Crellin and Chalmers, 2001). The following plasmids were derived from pRC98: pRC351 has a XhoI site introduced by PCR at bp +51 of the transposon end; pRC843 has the 1G mutation introduced by PCR at bp +1 of the transposon end, and the flanking DNA between the transposon end and bp −35 was deleted; pRC842 has the 8C mutation introduced by PCR at bp +8 of the transposon end, and the flanking DNA between bp −13 and BamHI site was deleted. The following plasmids were derived from pRC100: pRC845 has the 1G mutation introduced by PCR at bp +1 of the transposon end, and the flanking DNA between the transposon end and bp −16 was deleted; pRC844 has the 8C mutation introduced by PCR at bp +8 of the transposon end, and the flanking DNA between bp −13 and −16 was deleted.

The DNA fragments used in each of the Figures were prepared from the plasmids by restriction endonuclease digestion as follows: Fig. 2: in panels 1 and 2 the outside-end had a transposon arm of 88 bp, prepared by digesting pRC98 with HincII and XbaI; in panels 3–6 the outside-end had a transposon arm of 73 bp, prepared by digesting pRC173 with EcoRI and ClaI. Figure 3: the long OE was prepared by digesting pRC98 with XbaI and XhoI; the long EE was prepared by digesting pRC100 with XbaI and XhoI; the short outside-end was prepared by digesting pRC351 with XbaI and XhoI; the short even-end was prepared by digesting pRC100 with SpeI and HindIII. Figures 4 and 5: OE.wt is the wild-type outside-end, prepared by digesting pRC98 with BamHI and AccI; OE.1G is the outside-end with the 1G mutation, prepared by digesting pRC843 with XbaI and AccI; EE.wt is the wild-type even-end, prepared by digesting pRC100 with BamHI and AccI; EE.1G is the even-end with the 1G mutation, prepared by digesting pRC845 with SacII and AccI. Figure 6: OE.wt is the wild-type outside-end, prepared by digesting pRC98 with BamHI and AccI; EE.wt is the wild-type even-end, prepared by digesting pRC100 with BamHI and AccI; OE.8C is the outside-end with the 8C mutation, prepared by digesting pRC842 with XbaI and AccI; EE.8C is the even-end with the 8C mutation, prepared by digesting pRC844 with XbaI and AccI.

Following restriction enzyme digestion, the DNA was labelled by end filling using a single 32P-labelled deoxynucloetide and the Exo Klenow fragment of DNA polymerase. The transposon end DNA fragments were then purified by electrophoresis in TAE-buffered 5% polyacrylamide gels and recovered by the crush and soak method as described previously (Sakai et al., 1995; Chalmers and Kleckner, 1996).

Transposase-DNA complexes were assembled and visualized using the EMSA as described (Sakai et al., 1995; Crellin and Chalmers, 2001). The standard reaction was 20 μl and contained 50 fmol of radioactively labelled transposon end, 20 fmol transposase and 300 fmol IHF. When mixed complexes containing an outside-end paired with an even-end were assembled, 50 fmol of the labelled transposon end was mixed with 200 fmol of the unlabelled end. IHF was the standard 300 fmol, but transposase was increased to 100 fmol. The reactions were assembled at room temperature and the components were mixed before addition of transposase. Assembly of the complexes is very rapid and they can be visualized in the EMSA after 1 min or after overnight incubation. Where indicated the following additions were made to reactions: MgCl2, 4 mM; CaCl2, 4 mM; EDTA, 25 mM and heparin, 500 ng ml−1. Quantification was performed using a Fuji phosphorimager.

Hydroxyl radical footprinting

DNA footprints were generated by hydroxyl radical treatment of protein–DNA complexes as previously described (Crellin and Chalmers, 2001). The resultant ladders were compared with Maxam-Gilbert G+A sequence ladders and plotted using NIH image software available from the US National Institutes of Health web server (http://rsb.info.nih.gov/nih-image/).

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References

This work was funded by grants from The Wellcome Trust to RC.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  • Allingham, J.S., and Haniford, D.B. (2002) Mechanisms of metal ion action in Tn10 transposition. J Mol Biol 319: 5365.
  • Chalmers, R.M., and Kleckner, N. (1994) Tn10/IS10 transposase purification, activation, and in vitro reaction. J Biol Chem 269: 80298035.
  • Chalmers, R.M., and Kleckner, N. (1996) IS10/Tn10 transposition efficiently accommodates diverse transposon end configurations. EMBO J 15: 51125122.
  • Chalmers, R., Sewitz, S., Lipkow, K., and Crellin, P. (2000) Complete nucleotide sequence of Tn10. J Bacteriol 182: 29702972.
  • Crellin, P., and Chalmers, R. (2001) Protein-DNA contacts and conformational changes in the Tn10 transpososome during assembly and activation for cleavage. EMBO J 20: 38823891.
  • Crellin, P., Sewitz, S., and Chalmers, R. (2004) DNA looping and catalysis; the IHF-folded arm of Tn10 promotes conformational changes and hairpin resolution. Mol Cell 13: 537547.
  • Davies, D.R., Goryshin, I.Y., Reznikoff, W.S., and Rayment, I. (2000) Three-dimensional structure of the Tn5 synaptic complex transposition intermediate. Science 289: 7785.
  • Derbyshire, K.M., and Grindley, N.D. (1992) Binding of the IS903 transposase to its inverted repeat in vitro. EMBO J 11: 34493455.
  • Derbyshire, K.M., Hwang, L., and Grindley, N.D. (1987) Genetic analysis of the interaction of the insertion sequence IS903 transposase with its terminal inverted repeats. Proc Natl Acad Sci USA 84: 80498053.
  • Haniford, D.B. (2002) Transposon Tn10. In Mobile DNA II. Craig, N.L., Craigie, R., Gellert, M., and Lambowitz, A.M. (eds). Washington, DC: American Society for Microbiology, pp. 457483.
  • Haniford, D., and Kleckner, N. (1994) Tn10 transposition in vivo: temporal separation of cleavages at the two transposon ends and roles of terminal basepairs subsequent to interaction of ends. EMBO J 13: 34013411.
  • Huisman, O., Errada, P.R., Signon, L., and Kleckner, N. (1989) Mutational analysis of IS10s outside end. EMBO J 8: 21012109.
  • Humayun, S., Wardle, S.J., Shilton, B.H., Pribil, P.A., Liburd, J., and Haniford, D.B. (2005) Tn10 transposase mutants with altered transpososome unfolding properties are defective in hairpin formation. J Mol Biol 346: 703716.
  • Kennedy, A.K., and Haniford, D.B. (1996) Isolation and characterization of IS10 transposase separation of function mutants: identification of amino acid residues in transposase that are important for active site function and the stability of transposition intermediates. J Mol Biol 256: 533547.
  • Kennedy, A.K., Guhathakurta, A., Kleckner, N., and Haniford, D.B. (1998) Tn10 transposition via a DNA hairpin intermediate. Cell 95: 125134.
  • Kennedy, A.K., Haniford, D.B., and Mizuuchi, K. (2000) Single active site catalysis of the successive phosphoryl transfer steps by DNA transposases: insights from phosphorothioate stereoselectivity. Cell 101: 295305.
  • Kleckner, N. (1989) Transposon Tn10. In Mobile DNA. Berg, D.E., and Howe, M.M. (eds). Washington, DC: American Society for Microbiology, pp. 227268.
  • Kleckner, N., Chalmers, R.M., Kwon, D.K., Sakai, J., and Bolland, S. (1996) Tn10 and IS10 transposition and chromosome rearrangements: mechanism and regulation in vivo and in vitro. In Transposable Elements, Vol. 204. Saedler, H., and Gierl, A. (eds). Berlin: Springer, pp. 4982.
  • Liu, D., Crellin, P., and Chalmers, R. (2005) Cyclic changes in the affinity of protein–DNA interactions drive the progression and regulate the outcome of the Tn10 transposition reaction. Nucleic Acids Res 33: 19821992.
  • Lynch, T.W., Mattis, A.N., Gardner, J.F., and Rice, P.A. (2003) Integration host factor: putting a twist on protein-DNA recognition. J Mol Biol 330: 493502.
  • Sakai, J., Chalmers, R.M., and Kleckner, N. (1995) Identification and characterization of a pre-cleavage synaptic complex that is an early intermediate in Tn10 transposition. EMBO J 14: 43744383.
  • Sasakawa, C., Carle, G.F., and Berg, D.E. (1983) Sequences essential for transposition at the termini of IS50. Proc Natl Acad Sci USA 80: 72937297.
  • Sewitz, S., Crellin, P., and Chalmers, R. (2003) The positive and negative regulation of Tn10 transposition by IHF is mediated by structurally asymmetric transposon arms. Nucleic Acids Res 31: 58685876.