FtsK translocase multimers are active in vivo
Most previous research on FtsK translocation in vitro
has used a protein, FtsK50C
, in which a 50 aa segment derived from the FtsK N-terminus has been added to the αβγ translocase domain (Aussel et al, 2002
; Ip et al, 2003
; Saleh et al, 2004
; Pease et al, 2005
; Bigot et al, 2006
). These 50 aa facilitate hexamerization and are required for significant in vitro
catalytic activity (Aussel et al, 2002
). Nevertheless, FtsK50C
has a high propensity to aggregate, making quantitative biochemistry and mechanistic interpretation difficult. We reasoned that a minimal FtsK translocase, lacking the N-terminal 50 aa of FtsK50C
and derived by using the available structure (Massey et al, 2006
), might demonstrate in vitro
activity if the individual monomers were covalently linked. Therefore, FtsK aa 840–1329 were linked together with a 14 aa linker (GGGSEGGGSEGGSG), thus forming covalent multimers of the translocase, the first subunit being tagged with a 6-His tag and/or biotin-tagged peptide ().
Figure 1 (A) Schematic of the FtsK proteins used. FtsK depicts the wild-type protein, with four transmembrane helices in the N-terminus region (dark green), a 639 amino-acid linker (blue line) and the C-terminus motor domain (light blue boxes) is drawn, containing (more ...)
In initial experiments, we showed that the gene, encoding a 320 kDa FtsK covalent hexamer, expressed well enough to give high levels of in vivo translocation-dependent XerCD-dif recombination (data not shown). Nevertheless, the level of expression was not sufficient to enable ready purification of the protein. Therefore, we chose to work with monomers, covalent dimers and covalent trimers (hereafter described as ‘trimers'), which were expressed well. This provided the opportunity to introduce specific mutations into defined individual subunits of a covalent multimer. We then tested the in vivo activity of these translocases.
First, we exploited the observation that the overexpression of active FtsK translocase (e.g. FtsK50C
when expressed from a para
promoter with 0.2% arabinose) is toxic and cells stop growing and die within ~15 min of induced expression. Toxicity is correlated with the ability to hydrolyse ATP (Massey et al, 2006
). Expression of monomer, dimer and trimer of the new FtsK translocase derivatives was toxic, consistent with them having in vivo
activity (). Cells stopped growing ~45 min after induced expression of the monomer, ~30 min after dimer expression and ~15 min after trimer expression, consistent with multimerization by covalent linking enhancing specific activity, because hexamers are the active species. Second, we tested whether the proteins could support in vivo
recombination on a reporter plasmid containing two dif
sites (Recchia et al, 1999
). The results of this assay mirrored the toxicity results; trimers were more active than covalent dimers, which were more active than monomers (). These semi-quantitative assays do not take into account any differential levels of these proteins in cells. Nevertheless, our extensive experience with these assays using FtsK50C
derivatives (Sivanathan et al, 2006
and data not shown) has shown them to give a good indication of in vitro
specific activity. We conclude that covalent trimers and dimers are highly active in vivo
In vivo activities of FtsK multimers
In vitro activities of FtsK translocase multimers
The recombinant FtsK monomer, covalent dimer and trimer were then purified and assayed in vitro
for ATPase activity and for two independent assays of DNA translocation: in vitro
recombination and triplex displacement assays (; Aussel et al, 2002
; Sivanathan et al, 2006
). ATPase activity was DNA dependent (not shown), and at a concentration of 50 nM hexamer, the trimer hydrolysed >1.7 × 103
ATP/min/hexamer. At this protein concentration, the covalent dimers displayed ~80% of trimer specific activity, and the monomers ~1%, because of less efficient formation of active hexamers. The relatively low steady state level of ATP hydrolysis by the trimers indicates that most molecules are not translocating at any given time, presumably because they are not loaded onto DNA, have collided with other translocases, or are simply not active.
Figure 2 In vitro activity of FtsK multimers. (A) ATP hydrolysis over time is shown for monomer, covalent dimer and trimer, at a concentration of 50 nM hexamer equivalent. The black arrow indicates the 1 min time point, which was used in later experiments ( (more ...)
In XerCD-dif recombination and triplex displacement assays, the trimers were again the most active species, with covalent dimers showing ~40% of the activity of trimers. In the triplex displacement assays, optimal activity was obtained with trimers at 50 nM (hexamer), but required 80 and 150 nM for covalent dimers and monomers, respectively. These results confirm that covalent dimers and trimers are more active in vitro than the constituent monomers, with dimers and monomers forming hexamers at high and very high concentrations, respectively. In conclusion, trimers dimerize to form active hexamers efficiently at 50 nM, and their activity is higher than that observed for FtsK50C at the same concentration ( and data not shown). Furthermore, the trimers were still responsive to the presence of KOPS in the triplex displacement and recombination assays (data not shown).
FtsK translocase trimers form hexamers on DNA
Despite the trimers being highly active in translocation, we needed to be confident that the activity was resulting from dimerization of two trimers into a hexamer, rather than the activity residing in higher-order structures (, right panel), which could compromise interpretation of data once defined mutations had been introduced into the linked multimers. FtsK subunits were rendered non-functional by mutation in the highly conserved ATP-binding pocket. We used two mutations that have been shown to be catalytically defective in an extensive range of studies of RecA-fold proteins (ClpB: Watanabe et al, 2002
; Dynein: Reck-Peterson and Vale, 2004
; Rad51: Wiese et al, 2006
and archeal MCM: Moreau et al, 2007
). A K997A substitution in the Walker A box (WA) should prevent ATP binding, whereas a D1121A substitution in the Walker B box (WB) may allow ATP binding, but should be defective in ATP hydrolysis. Binding of trimers, containing 0 to 3 WA mutated subunits, to short KOPS-containing DNA was analysed by gel electrophoresis. Two shifted bands were observed in the presence of 200 nM protein, consistent with binding of one and two trimers on DNA (data not shown). Protein–DNA complexes containing wild-type, or three mutated WA subunits, were also observed by electron microscopy using negative staining. On a 2.7 kb DNA, particles whose shape and size were consistent with ‘side-view' FtsK hexamers were the majority species (; Massey et al, 2006
). The trapezium shape particles had diameters of 120–130 Å (large side) and 60–70 Å (small side). Aggregates or higher-order structures comprised <2% of particles. On a short DNA (44 bp), when ‘end-on' view (top view) particles are expected, rounded particles of diameter ~130 Å were observed (), the size expected for a hexamer obtained by dimerization of trimers. However, we did not observe any lack of density corresponding to the central hole, or any six-fold symmetry (Massey et al, 2006
). This may be because the γ-subdomains are attached to the αβ motor by a flexible linker, and therefore may occlude the central hole, as well as mask the symmetry of the protein complexes.
We then analysed whether the trimers could adopt alternative configurations when they dimerize into hexamers. The 14 aa linker may be long enough that it does not restrict the assembly into only a single configuration, although we expect the linker length to restrict the arrangement of subunits such that subunit 1 is never adjacent to subunit 3 (). The existence of alternative subunit configurations potentially complicates the interpretation of experiments using mutant subunits. To address this issue, we labelled His-tagged covalent trimers, pre-loaded onto short DNA to give preferential end-views, with 2 nm Ni-NTA-bound gold beads and assessed the relative locations of the beads within the hexamers by electron microscopy, followed by image classification (). The 1702 particles analysed were classified into 10, 50 or 100 groups on the basis of protein shape and gold position. Classes with two adjacent golds and with the two golds in trans were observed for each group, indicating that trimers can dimerize in both head–tail and head–head configurations ( and ). However, some classes were ambiguous, with a very low signal–noise ratio, thereby precluding quantitative estimates of the two configurations. Given the likelihood that both configurations can form, we chose symmetrical configurations of mutations to avoid any ambiguity in the configurations of the subunits in the assembled hexamers. We focused most of our experiments on trimers containing a mutated subunit in the central position, so that hexamers should carry two mutated subunits each separated by two wild-type subunits. Taken together, these results give us confidence that most biological activities assayed were the result of dimerization of trimers on DNA.
Activities of translocases with mutated subunits
To study coordination between FtsK subunits during DNA translocation, we compared the activities of presumptive hexamers derived from trimers with unmutated and/or mutated subunits. The mutations were introduced separately into a given subunit of the trimers; in the central position to get a single mutant; in the first and the last subunits to get a double mutant and in all three subunits to get a triple mutant.
The mutated trimers were first tested in vivo using the toxicity and XerCD-dif recombination assays (). In the toxicity assay, cells expressing the wild-type trimer showed a loss of exponential growth 15 min after induction of expression. Trimers containing three WA or three WB mutated subunits showed no toxicity and continued to grow. Trimers containing single WA or WB mutations were more toxic than trimers carrying two WA or WB mutations, with the trimer with two mutated WA subunits appearing slightly less toxic than its WB counterpart. The in vivo FtsK XerCD-dif recombination assays broadly mirrored the toxicity results, although in these assays, trimers with a single WA or WB mutant showed similar activity to wild-type trimers, whereas trimers with two WA or WB mutations appeared as deficient as trimers with three mutations. The residual level of recombination from the trimers with three mutated subunits may result from translocation-independent stimulation of recombination by the γ-subunits present on the trimers (Grainge I and Sherratt DJ, unpublished data).
Wild-type trimers were more active than wild-type covalent dimers in both of these assays, whereas wild-type monomers showed little if any activity. This result reassured us that any monomers or covalent dimers with one wild-type subunit and one mutated subunit, derived from proteolysis of trimers in vitro or in vivo, are unlikely to have significant activity that could confound the interpretation of the above experiments. These results gave us the confidence to purify the mutated trimers, giving the single, double and triple mutants in the WA and WB motifs, respectively ().
The FtsK translocation mechanism is neither stochastic nor strictly concerted
FtsK-dependent XerCD-dif site-specific recombination and displacement of a triplex-forming oligonucleotide were used to assay DNA translocation by the trimer variants and their combinations (). For trimer mutants with 0, 1, 2 or 3 mutated subunits, we followed the decrease of recombination as a function of the number of mutant subunits. Assuming that the only potentially active species are hexamers formed by trimer dimerization, this gives data for hexamers with 0, 2, 4, or 6 mutated subunits. To obtain results for hexamers containing 1, 3 or 5 mutant subunits, different trimer mutants were mixed in equimolar amounts (). For example, to obtain a hexamer population with only one mutated subunit, equal amounts of wild-type trimer and single-mutant trimer were mixed. In this mixture, it is predicted that 25% of hexamers are wild type; 50% carry a single mutated subunit; and 25% carry two mutated subunits. As the activity of each homohexamer is known, we were able to deduce the activity of the heterohexamer.
Figure 3 In vitro activity of covalently linked mutated hexamers. (A) XerCD-dif recombination as a function of the number of mutated subunits. Recombination reactions were performed for 1 min on a dimeric plasmid containing two dif sites. In all, 50 nM hexamers (more ...)
For the in vitro recombination assays that are dependent on DNA translocation by FtsK for activation of recombination, hexamers in which all of the subunits carry WA or WB mutations showed no activity, as expected, and gave us confidence that the mutations abrogate DNA translocation. By comparison, the wild-type subunits showed ~30% recombinant product in a 1 min reaction (; normalized to 100% in ).
Hexamers with two WA or WB mutations showed substantial recombination activity (~65% of wild type in 1 min reactions), whereas the mixed subunit population that gives hexamers with a single mutation (50%) showed at least wild-type activities after correction for the wild-type (25%) and double-mutated (25%) population. Some residual recombination activity (<20%) was observed in the assays in which 50% of the population should contain hexamers with three mutated subunits. No activity was observed with four mutated subunits. These observations are inconsistent with a stochastic firing model, in which a linear decrease in activity as a function of increasing mutated subunit number would be expected (Moreau et al, 2007
). Similarly, obligate sequential rotary models, in which each subunit must be catalytically active in turn, can be ruled out, because a single-mutated subunit would block translocation. In parallel, ATPase activity was assayed in each of the recombination reactions, giving an estimate of the number of ATP consumed as a function of the level of recombination for each mutant hexamer (). Note that the absolute level of ATPase activity in the recombination assays cannot be compared directly with ATPase activity measured in the absence of recombinases, because the ATPase activity is downregulated when FtsK encounters XerCD bound to dif
(Graham et al, 2010
). ATPase activity decreased less sharply, with the four mutated subunit population retaining >40% of wild-type activity. This observation is consistent with an uncoupling of ATPase and translocation. Indeed by independent assays of ATPase activity under optimal conditions, with DNA in excess, there was an approximately linear decrease in ATPase activity as a function of the number of mutated subunits (data not shown). These results contrast to those from experiments in which wild-type P. aeruginosa
FtsK (PAK4) hexamers were doped with increasing amounts of a WA mutant subunit and the ATPase activity measured (Massey et al, 2006
). In that paper, there was a rapid, greater than linear reduction in ATPase activity as a function of mutated subunit concentration, leading to the conclusion of a rigorous sequential (or concerted) mechanism. However, we note that the absolute levels of ATPase activity were low, perhaps because a 16 bp dsDNA was used, which would not have supported significant DNA translocation. Furthermore, it is possible that the mutated subunits formed homohexamers and mixed hexamers more avidly than the wild-type protein on the 16 bp substrate, leading to an over-representation of hexamers containing
3 mutant subunits for a given input ratio.
In triplex-displacement assays, which provide an independent measure of translocation, hexamers with three or more mutant subunits all failed to show significant displacement activity in assays that required 60 bp of translocation after loading at KOPS, before encountering the triplex (). Hexamers with two WA or two WB mutated subunits showed
20% of wild-type activity, a marked contrast to their higher activity when activating Xer recombination. ATPase activities measured in the same reactions gave results similar to those in (data not shown). These data suggest a quantitative difference in ‘readout' as compared with the recombination assays, perhaps because hexamers with mutated subunits are less able to displace a triplex when they have translocated up to it, as compared with being able to activate recombination after translocation up to the XerCD-dif
complex. It is unlikely that the difference between assays reflects reduced processivity of translocation with mutated subunits, given that FtsK has to translocate only 60 bp in the triplex assay, whereas most random loadings of FtsK onto DNA in the recombination assay will leave >60 bp of translocation needed to encounter XerCD-dif
. To test this directly, we assayed translocation by wild-type and mutated trimers in a single-molecule assay.
Translocases with two catalytically inactive subunits show wild-type translocation velocity
We used magnetic tweezers with a DNA molecule (17 kb, 5.7 μm) tethered between the surface and a magnetic bead. The DNA was maintained in an extended conformation by magnets applying a constant force (Strick et al, 1996
), to study translocation by hexamers derived from wild-type and mutated trimers. The FtsK variant and ATP were then added and the position of the bead was observed. Looping during translocation, probably arising when the translocase attaches to either the bead or the surface, causes the bead to move towards the surface during translocation. Either release of the loop or a reverse in translocation direction causes the bead to move away from the surface (). These events can be distinguished by the rate of reversal of the bead. Between 360 and 930 events per protein were recorded, and the distributions of burst speeds were determined. Translocases derived from the wild-type trimer were first compared with FtsK50C
. Both proteins showed the same translocation velocity (data not shown), with translocation events of several microns being common.
Figure 4 Single-molecule analysis of covalent multimer translocation. Magnetic tweezers were used to study translocation of wild-type and double mutant hexamers, containing WA or WB in the central subunit of the trimers. Upper panels show examples of events obtained (more ...)
Translocases derived from trimers containing a mutated (WA or WB) central subunit were analysed in the same way and the average burst speed was deduced. Surprisingly, they translocated along DNA at rates similar to the wild-type covalent trimer against forces of 1–35 pN. The data shown were obtained at 5 pN and reveal a translocation velocity of ~3.3 kb/s. We are confident that these events arise from the translocation of hexamers formed by dimerization of the mutated trimers, because our EM and gel-shift analysis provided no evidence for significant numbers of higher form multimers that could assemble six wild-type subunits into a hexamer ( and ). Also note that monomers and covalent dimers were not active at this concentration. Furthermore, the observation of substantial translocation activity by the covalent trimers with a single-mutated subunit in the XerCD-dif recombination reaction, where specific activity can be estimated, reinforces our conclusion that hexamers with mutated subunits in the 2 and 5 positions exhibit wild-type translocation rates. This led us to wonder whether the low activity in the triplex-displacement assays results from a failure to displace the triplex efficiently rather than an impaired ability to translocate up to the triplex. This could mean that the covalent trimer with a central mutant subunit would have an impaired ability to generate the mechanical work required to displace the triplex, despite a normal translocation velocity.
Magnetic tweezers are not a good tool to measure and compare FtsK stall forces as both processivity and events rate dramatically decrease at high force (Saleh et al, 2004
; Pease et al, 2005
), while modification of DNA structure at high forces may affect the results. This led us to use another strategy.
Trimers with a single catalytic mutation give translocases impaired in displacing a DNA roadblock
To calibrate the relative ability of molecular motors to strip proteins from DNA, we have generated a panel of mutated streptavidins with a wide range of off-rates from biotin (Chivers et al, 2010
). The streptavidin–biotin interaction is one of the strongest non-covalent interactions characterized, with a KD
of ~4 × 10−14
M (Green, 1990
). Here, we focused on one stronger mutant and one weaker mutant of streptavidin. The stronger streptavidin mutant had a 10-fold lower off-rate from biotin (<2% spontaneous dissociation from biotin-4-fluorescein in 12 h) than wild-type streptavidin (12% spontaneous dissociation in 12 h). The weaker mutant had accelerated dissociation from biotin conjugates (Supplementary Figure S1
), but binding was still strong enough to survive gel electrophoresis.
Therefore, to examine the ability of FtsK to do mechanical work to displace roadblocks on DNA during translocation, we compared the ability of wild-type and mutated trimers to displace wild-type and mutated streptavidins from biotinylated DNA. A 597 bp DNA fragment biotinylated at one end and with two KOPS sites positioned 230 bp upstream was bound by streptavidin and used in FtsK translocation assays. After 2 min reactions with translocase and ATP, agarose gel electrophoresis was used to analyse the fraction of DNA molecules where streptavidin had been displaced. Excess free biotin was used as a ‘sink' to prevent rebinding of displaced streptavidin and 0.1% SDS was used to stop translocation, conditions that do not denature streptavidin (). Wild-type hexamers displaced >70% of streptavidin in 2 min, whereas translocases with two WA or WB mutations displaced <45% of streptavidin. In all, 30% of the tight-binding streptavidin was displaced in 2 min by the wild-type trimer, whereas
10% was displaced by the singly mutated trimers. In contrast, the weaker streptavidin was displaced to comparable levels (
82%) by all three proteins, showing that they have comparable abilities to translocate up to, and displace, a weak roadblock. Taken together, these data show that trimers with one catalytically inactive subunit form translocases as rapid as those formed by wild-type trimers, but which are impaired in displacing strong roadblocks that include triplex DNA and streptavidin bound to biotin.
Figure 5 Displacement of the streptavidin ‘roadblock' during translocation. Streptavidin displacement activity was followed on a 597 bp biotinylated DNA. Excess streptavidin (wt, strong, or weak binding derivatives) was bound for 30 min and then excess (more ...)