ClpV recycles VipA/VipB tubules and prevents non-productive tubule formation to ensure efficient type VI protein secretion


  • Nicole Kapitein,

    1. Center for Molecular Biology of the University of Heidelberg and German Cancer Research Center, DKFZ-ZMBH Alliance, Universität Heidelberg, Heidelberg, Germany
    Search for more papers by this author
  • Gabriele Bönemann,

    1. Center for Molecular Biology of the University of Heidelberg and German Cancer Research Center, DKFZ-ZMBH Alliance, Universität Heidelberg, Heidelberg, Germany
    Search for more papers by this author
  • Aleksandra Pietrosiuk,

    1. Center for Molecular Biology of the University of Heidelberg and German Cancer Research Center, DKFZ-ZMBH Alliance, Universität Heidelberg, Heidelberg, Germany
    Search for more papers by this author
  • Fabian Seyffer,

    1. Center for Molecular Biology of the University of Heidelberg and German Cancer Research Center, DKFZ-ZMBH Alliance, Universität Heidelberg, Heidelberg, Germany
    Search for more papers by this author
  • Ingrid Hausser,

    1. Department of Infectious diseases and Core Facility Electron Microscopy, Universität Heidelberg, Heidelberg, Germany
    Search for more papers by this author
  • Jacomine Krijnse Locker,

    1. Department of Infectious diseases and Core Facility Electron Microscopy, Universität Heidelberg, Heidelberg, Germany
    Search for more papers by this author
  • Axel Mogk

    Corresponding author
    • Center for Molecular Biology of the University of Heidelberg and German Cancer Research Center, DKFZ-ZMBH Alliance, Universität Heidelberg, Heidelberg, Germany
    Search for more papers by this author

For correspondence. E-mail; Tel. (+49) 6221 546863; Fax (+49) 6221 545894.


The multicomponent type VI secretion system (T6SS) mediates the transport of effector proteins by puncturing target membranes. T6SSs are suggested to form a contractile nanomachine, functioning similar to the cell-puncturing device of tailed bacteriophages. The T6SS members VipA/VipB form tubular complexes and are predicted to function in analogy to viral tail sheath proteins by providing the energy for secretion via contraction. The ATPase ClpV disassembles VipA/VipB tubules in vitro, but the physiological relevance of tubule disintegration remained unclear. Here, we show that VipA/VipB tubules localize near-perpendicular to the inner membrane of Vibrio cholerae cells and exhibit repetitive cycles of elongation, contraction and disassembly. VipA/VipB tubules are decorated by ClpV in vivo and become static in ΔclpV cells, indicating that ClpV is required for tubule removal. VipA/VipB tubules mislocalize in ΔclpV cells and exhibit a reduced frequency of tubule elongation, indicating that ClpV also suppresses the spontaneous formation of contracted, non-productive VipA/VipB tubules. ClpV activity is restricted to the contracted state of VipA/VipB, allowing formation of functional elongated tubules at a T6SS assembly. Targeting of an unrelated ATPase to VipA/VipB is sufficient to replace ClpV function in vivo, suggesting that ClpV activity is autonomously regulated by VipA/VipB conformation.


The secretion of bacterial effector proteins into the extracellular space represents a crucial step in pathogenesis and interbacterial virulence. In Gram-negative bacteria various specialized secretion system mediate the transport of effectors in a single step across inner and outer membranes. Type VI secretion systems (T6SSs) are the most widespread specialized secretion systems and have been identified in 25% of all sequenced Gram-negative bacteria (Bingle et al., 2008). T6SSs are encoded by a gene cluster comprising 15–20 conserved open reading frames. They mediate the translocation of effector proteins to kill target cells in a cell-to-cell contact-dependent process (Pukatzki et al., 2006; Hood et al., 2010; MacIntyre et al., 2010; Russell et al., 2011).

T6SS components have been localized to the cytosol, inner and outer membrane and the periplasm, suggesting the formation of a multicomponent T6SS assembly spanning the entire cell envelope (Aschtgen et al., 2008; 2010; Ma et al., 2009; Felisberto-Rodrigues et al., 2011). Bioinformatic and structural analyses suggest that T6SSs represent contractile injection systems, using a similar infection mechanism as Myoviridae, bacteriophages (e.g. T4 phage) that infect, e.g. Escherichia coli cells by injecting their viral genome (Pukatzki et al., 2007; Leiman et al., 2009; Pell et al., 2009; Veesler and Cambillau, 2011). These bacteriophages employ a syringe-like macromolecular nanomachine to puncture the membrane of bacterial host cells. The infectious device is composed of a baseplate, a contractile sheath that is harbouring an internal non-contractile tube and an associated tail spike complex (Leiman et al., 2010). Upon contact of the phage with the host cell the elongated tail sheath contracts, causing the ejection of the internal tube and the tail spike complex and resulting in penetration of the host envelope (Leiman et al., 2010).

The T6SS exoproteins Hcp and VgrG, whose presence in cell culture supernatants represents a hallmark of T6SS activity, are structurally similar to viral proteins playing key roles in the infection process. The structure of Hcp resembles viral tail tube proteins (Leiman et al., 2009). Hcp forms hexameric rings, which can form tubular stacks in vitro if stabilized by artificial disulfide bonds (Mougous et al., 2006; Ballister et al., 2008). The structure of trimeric VgrG is highly similar to the viral tail spike complex, sharing an elongated β-helix at the tip and functioning as the structural device to puncture target membranes (Leiman et al., 2009). The conserved T6SS components VipA and VipB form a tubular complex that is reminiscent in shape and dimension to viral tail sheath proteins (Bonemann et al., 2009; Leiman et al., 2009). It has been speculated that VipA/VipB tubules might engulf Hcp tubes and, in analogy to the viral infection mechanism, eject Hcp and associated VgrG upon contraction (Leiman et al., 2009; Bonemann et al., 2010). Recently, Mekalanos and colleagues (Basler et al., 2012) identified tubular VipA/VipB structures in intact Vibrio cholerae by cryo-electron tomography. Such structures were only formed in wild-type but not in T6SS mutant cells. The same authors showed that a plasmid-encoded VipA–sfGFP fusion protein formed dynamic elongated structures, providing further evidence that T6SSs resemble contractile nanomachines (Basler et al., 2012).

In addition, there is raising evidence that the ATPase ClpV is crucial for VipA/VipB dynamics in V. cholerae. VipA/VipB tubules are rapidly disassembled by the ATPase ClpV in vitro and its ATPase activity is crucial for type VI protein secretion (T6S) (Bonemann et al., 2009). Since VipA–sfGFP dynamics are lost in V. cholerae ΔclpV mutant cells, it was recently suggested that ClpV is required to recycle VipA/VipB tubules after successful contraction (Basler et al., 2012).

Here, we confirm and further elucidate the model of this dynamic interplay by investigating the localization of VipA/VipB and ClpV in V. cholerae V52 cells by electron and fluorescence microscopy. We show that VipA/VipB tubules form in vivo and match properties of viral tail sheath proteins. VipA/VipB tubules exhibit dynamic cycles of polymerization and disassembly and are positioned near-perpendicular to the membrane. The formation of elongated and dynamic VipA/VipB tubules depends on various T6SS core components. ClpV exerts dual activities by preventing the spontaneous formation of non-productive, contracted VipA/VipB tubules, which have no function in T6S and by recycling tubules after productive contraction. ClpV activity can be taken over by a different AAA+ protein, targeted to the tubules, indicating that ClpV activity is specifically controlled by the conformational state of VipA/VipB tubules.


VipA/VipB form dynamic tubular structures in V. cholerae cells

The recent analysis of a VipA–sfGFP fusion protein in V. cholerae revealed the formation of punctuate and elongated structures, indicating that VipA/VipB tubules form at distinct cellular positions in vivo (Basler et al., 2012). In order to localize the tubules directly, we determined VipA/VipB position in V. cholerae V52 wild-type (WT) cells by immunofluorescence microscopy using affinity-purified VipB-specific antibodies. 40% of V. cholerae wild-type cells showed a fluorescence signal, whereas no signal was detected in V. cholerae ΔvipB mutant cells (Fig. 1, data not shown), indicating that the signal specifically reports on cellular VipB distribution. Among those cells exhibiting fluorescence signals, most V. cholerae wild-type cells harboured single (60%, n = 900) or two (25%, n = 900) fluorescent VipB foci, whereas multiple foci or a diffuse VipB staining was barely observed (Fig. 1A). To analyse whether VipB foci represent VipA/VipB assemblies, we monitored VipB localization in V. cholerae ΔvipA cells. Here, we observed a strong increase in the number of cells exhibiting a diffuse staining (33%, n = 600), a phenotype that could be complemented by expression of plasmid encoded vipA (Fig. 1A). The detection of VipB foci in immunofluorescence microscopy could therefore be largely attributed to association with VipA, suggesting that the observed foci represent tubular complexes.

Figure 1.

VipA/VipB form dynamic tubular assemblies in V. cholerae V52 cells.

A. V. cholerae wild-type (WT) and ΔvipA cells harbouring either the empty vector pMPM-A4 or pMPM-A4–vipA were grown at 37°C to mid-logarithmic growth phase in the presence of arabinose. VipB localizations were determined by immunofluorescence microscopy using specific antibodies. Overlays of fluorescent signals and DIC images are given. Scale bar: 1 μm. The fraction of cells showing either diffuse or punctuate fluorescence was determined (n > 200).

B. Cryosections of V. cholerae WT cells were immunogold-labelled with VipB-specific antibodies. Elongated, tubular-like structures are marked by an arrow. An electron-dense region is encircled by dashed lines. Scale bar: 100 nm.

C. V. cholerae vipAyfp cells were grown at 37°C to mid-logarithmic growth phase and VipAYFP localization was analysed. Borders of V. cholerae cells are indicated by dashed lines. Scale bar: 2 μm.

D. Two independent time-lapse experiments of V. cholerae vipAyfp cells. Punctuate and elongated VipAYFP structures are indicated by arrows. Borders of V. cholerae cells are indicated by dashed lines. Scale bar: 2 μm.

To provide direct evidence for the existence of VipA/VipB tubules in V. cholerae cells we employed electron microscopy of thawed cryosections that were immunogold-labelled using VipB-specific antibodies (Fig. 1B). In 2.9% of all wild-type sections (n = 550) we observed single elongated structures in the cytosol, which were decorated by gold particles. Tubular structures were not observed in ΔvipB cells (Fig. 1, data not shown). The structures varied in lengths from 250 to 500 nm and had a diameter of 30 nm and appeared similar in dimensions to purified VipA/VipB tubules and tubules observed in cryo-electron tomography images (Bonemann et al., 2009; Basler et al., 2012). The structures were always oriented near-perpendicular towards the inner membrane, with a possible direct contact via an electron-dense region (Fig. 1B). Together these findings prove the existence of VipA/VipB tubules in V. cholerae cells and confirm that tubular structures identified previously in intact V. cholerae cells represent VipA/VipB assemblies (Basler et al., 2012).

To analyse tubule formation in living V. cholerae cells, we genomically integrated vipAyfp or vipByfp fusions at the authentic locus, thereby replacing the corresponding wild-type genes and ensuring expression of the fusion protein to wild-type levels. We tested for functionality of both fusions by determining Hcp levels in cell culture supernatants as readout for a functional T6SS (Fig. S1A). VipA–YFP allowed for Hcp secretion in contrast to VipB–YFP. The integration of vipByfp did not cause a polar effect on downstream-located T6SS-encoding genes as confirmed by unchanged ClpV protein levels compared to wild-type cells (Fig. S1A). Purified VipA/VipB–YFP still formed tubular complexes, which were disassembled by ClpV (Fig. S1B and C), excluding major structural defects as basis for its non-functionality in vivo. EM analysis of VipA/VipB–YFP, however, revealed the presence of electron density within the central channel, which was not observed for VipA/VipB tubules (Fig. S1B). The extra density can be attributed to the fused YFP moiety. This finding indicates that blocking the central channel of VipA/VipB tubules abolishes T6S.

Next, we analysed the localization of functional VipA–YFP fusions in V. cholerae vipAyfp cells. We ensured by Western blot analysis using YFP-specific antibodies that only the full-length VipA–YFP fusion is produced, allowing to faithfully track the cellular localization of VipA via fused YFP (Fig. S1A). Most V. cholerae vipAyfp cells (83%, n = 240) exhibited a diffuse cytosolic staining of VipA–YFP, while in other cells one (11%) or two fluorescent foci (3%) were detected in addition to diffuse fluorescence (Fig. 1C). The majority (61%, n = 96 cells) of VipA–YFP foci were located close to the cell poles: 26% of cells contained foci at quarter-cell position and 13% of cells showed foci at mid-cell position. The degree of living V. cholerae cells exhibiting VipA–YFP foci was significantly lower compared to fixed cells showing VipB foci in immunofluorescence microscopy (Fig. 1A). This is unexpected given the documented complex formation between VipA and VipB. The variation in foci numbers might be attributable to differences in staining procedures and is at least partially caused by the fixation protocol, since the fraction of fixed V. cholerae vipAyfp cells showing VipA–YFP foci increased to 32% (n = 160) upon fixation (data not shown).

The pronounced impact of cell fixation on the degree of VipA–YFP foci formation could be explained by the dynamic nature of VipA/VipB, which has been documented for a plasmid-encoded VipA–sfGFP fusion protein (Basler et al., 2012). To reassess if such dynamic behaviour is also observed for our chromosomally integrated VipA–YFP fusion, we monitored VipA–YFP localization by time-lapse microscopy in living V. cholerae vipAyfp cells (Fig. 1D). VipA–YFP also exhibited an enormous dynamic behaviour, including initial foci formation followed by elongation of foci to linear structures spanning almost the entire cell length or width (Fig. 1D). Elongation of foci typically took 60–165 s, depending on the final length. Shortening of elongated structures occurred between two consecutive shots (15 s) and was accompanied by an intensification of VipA–YFP fluorescence. Shortened VipA–YFP structures subsequently vanished within 30–90 s. The nature and degree of VipA–YFP dynamics are similar to the ones reported recently for plasmid-encoded VipA–sfGFP in V. cholerae (Basler et al., 2012).

Upon disintegration new VipA–YFP foci frequently formed in the same cell. We determined the number of foci positions and the frequency of foci elongation at individual positions over a period of 20 min, representing one generation time. Within one generation we observed 2–3 cycles of VipA–YFP foci elongation and disintegration. All cells harboured one elongated VipA–YFP structure at a given time. After disintegration VipA–YFP foci formed either at the original position or at alternative sites (54% formed at the same site, 46% at another position, n = 20). These findings suggest that more than one T6SS assembly exists per cell, of which usually only one is used at a given time.

ClpV decorates VipA/VipB tubules in vivo

The ATPase ClpV decorates and disassembles VipA/VipB tubules in vitro, but direct evidence for such activity was lacking in V. cholerae cells. We first analysed whether ClpV, like VipA and VipB, reveals a specific localization in vivo by immunofluorescence of fixed V. cholerae cells using affinity-purified ClpV-specific antibodies.

The majority of V. cholerae WT cells (60%, n = 1400) exhibited single ClpV foci, whereas a diffuse fluorescence was only observed in a minority (< 10%) of cells (Fig. 2A). The degree of ClpV foci formation was almost identical to the one determined for VipB, suggesting colocalization of ClpV with VipA/VipB tubules, which we confirmed by immunofluorescence of fixed V. cholerae vipAYFP cells using ClpV-specific antibodies (Fig. 2B). Colocalization of ClpV and VipA/VipB can be explained either by direct interaction or by their independent recruitment to the same cellular site, representing a T6SS assembly. To distinguish between these scenarios, we monitored ClpV localization in V. cholerae ΔvipA or ΔvipB mutants (Fig. 2A). In contrast to WT cells a major fraction of both mutant cells (> 40%, n > 900) revealed a diffuse ClpV distribution. A high degree of ClpV foci formation was restored upon expression of plasmid-encoded vipA or vipB (Fig. 2A). These findings demonstrate that cellular ClpV localization depends on VipA/VipB tubules, suggesting that ClpV is recruited by VipA/VipB to a T6SS assembly.

Figure 2.

ClpV decorates VipA/VipB tubules in vivo.

A. V. cholerae V52 wild-type (WT), ΔvipA and ΔvipB cells harbouring either the empty vector pMPM-A4 or pMPM-A4–vipA or pMPMA4–vipB were grown at 37°C to mid-logarithmic growth phase in the presence of arabinose. ClpV localizations were determined by immunofluorescence using specific antibodies. Overlays of fluorescent signals and DIC images are given. Scale bar: 1 μm. The fraction of cells showing either diffuse or punctuate fluorescence was determined (n > 500).

B. V. cholerae vipAyfp cells were grown at 37°C to mid-logarithmic growth phase. Colocalization of ClpV with VipAYFP/VipB tubules was monitored in fixed cells by immunofluorescence microscopy. Scale bar: 2 μm.

C. Cryosections of V. cholerae WT cells were immunogold-labelled with ClpV-specific antibodies. An elongated, tubular-like structure, which is decorated by gold particles, is marked by an arrow (top). In a thin section that does not reveal a tubular-like structure gold particles are randomly distributed (bottom). Scale bar: 100 nm.

D. Cryosections of V. cholerae WT cells were immunogold-labelled with ClpV- (5 nm gold particles, red arrows) and VipB-specific antibodies (10 nm gold particles, blue arrow). An elongated, tubular-like structure is marked (black arrow). Scale bar: 100 nm.

In vitro ClpV entirely decorates VipA/VipB tubules in presence of non-hydrolysable ATPγS (Bonemann et al., 2009). While our findings indicate ClpV binding to VipA/VipB tubules in V. cholerae cells, it was unclear whether ClpV entirely decorates the tubules as determined in vitro or whether the presence of a functional T6SS restricts ClpV association to a specific site. To address this issue, we monitored ClpV localization by immunoelectron microscopy of V. cholerae WT thin sections using ClpV-specific antibodies and gold-labelled secondary antibodies. While gold particles were randomly distributed throughout the cytosol in most sections, they were organized in a linear array surrounding an elongated structure in some sections (Fig. 2C). Such observation was not made in V. cholerae ΔvipB cells (data not shown) suggesting that the staining represents ClpV decoration of VipA/VipB tubules. Next, we performed a double labelling experiment using ClpV- and VipB-specific antibodies and gold particles of different sizes (5 and 10 nm) coupled to the secondary antibody. Whenever elongated, tubular-like structures were identified in V. cholerae sections, they were labelled with gold particles of both sizes, demonstrating that ClpV decorates VipA/VipB tubules in V. cholerae cells (Fig. 2D). These findings are in complete agreement with recent data, showing colocalization of a ClpV–mCherry fusion protein with elongated VipA–sfGFP structures in V. cholerae cells (Basler and Mekalanos, 2012).

VipAYFP localization differently depends on T6SS components

Our observation that ClpV decorates VipA/VipB tubules in vivo suggests that ClpV acts in tubule disassembly during T6S. In agreement with such possibility a VipA–sfGFP fusion protein formed static foci in V. cholerae ΔclpV knockout cells (Basler et al., 2012). Accordingly, we observed that the majority of V. cholerae vipAyfp ΔclpV mutant cells exhibited one or two VipA–YFP foci (65% and 22% respectively; n = 200), which appeared more intense compared to VipA–YFP foci observed in WT cells. Consequently, ΔclpV cells did almost not display diffuse fluorescence, in contrast to WT cells (Fig. 3 and C). Time-lapse experiments in this mutant revealed that VipA–YFP foci do not show cycles of elongation and disintegration, indicating that the foci represent static VipA–YFP/VipB assemblies (Fig. S2A). Next, we analysed whether the changes in VipA–YFP localization and dynamics caused by missing ClpV activity are similar or distinct compared to other T6SS mutants. The mutants, namely Δhcp1hcp2, Δvca0109, Δfha and ΔicmF, were selected based on the following criteria. Hcp shows structural similarity to viral tail tube proteins and is predicted to be engulfed by VipA/VipB tubules (Leiman et al., 2009). VCA0109 exhibits sequence similarity to gp25, a component of the bacterio-phage T4 base plate (Leiman et al., 2009). Fha has been shown to control ClpV localization in Pseudomonas aeruginosa (Mougous et al., 2007). IcmF represents the second essential ATPase component of T6SSs (Ma et al., 2009; 2012). All these mutants abrogate T6S as shown for Hcp export and could be complemented by plasmid-encoded wild-type copies of the respective genes, excluding polar effects of the gene deletions (Fig. S3A). VipA–YFP was exclusively produced as full-length protein to the same level in all mutants, allowing for direct comparison of fluorescent stainings (Fig. S3B). VipA–YFP exclusively exhibited a diffuse localization in all these mutant cells (Fig. 3 and C). Time-lapse experiments did not reveal any changes in VipA–YFP localization over time, indicating that all tested essential T6SS core components are required for VipA–YFP foci formation and elongation (Fig. S2B–E). These findings were additionally confirmed by restoring VipA–YFP dynamics in Δhcp1hcp2 and Δvca0109 mutant cells through arabinose-controlled expression of plasmid-encoded hcp1 and vca0109 (Fig. S4A and B). The phenotype of V. cholerae vipAyfp Δhcp1hcp2, Δvca0109, Δfha and ΔicmF mutants is therefore opposite to ΔclpV, which reveals quantitative and static VipA–YFP foci formation. Based on our analysis we therefore define two distinct classes of T6SS components. At this stage, we hypothesize that the first class of proteins, including Hcp, VCA0109, Fha and IcmF, is required for VipA/VipB tubule assembly, whereas ClpV, representing the second class, is acting at later stages to remove the formed tubules. If this holds true we expect V. cholerae vipAyfp double knockout cells lacking ClpV and a class I member to display the phenotype observed in class I mutant cells: diffuse fluorescence of VipA–YFP. However, all double mutant cells exhibited VipA–YFP foci in all cells and did not display significant diffuse fluorescence, indicating that the functional consequences of ΔclpV cells on VipA–YFP localization are dominant (Fig. 3 and C). Preexisting VipA–YFP foci in V. cholerae ΔclpV Δhcp1hcp2 were cleared upon arabinose-induced expression of plasmid-encoded clpV, suggesting that the foci still represent tubular assemblies (Figs 3C).

Figure 3.

Localization of VipAYFP is differently affected in T6SS mutant cells.

A and B. V. cholerae WT and T6SS mutant cells expressing vipAyfp were grown at 37°C to mid-logarithmic growth phase and VipAYFP localization was analysed by fluorescence microscopy. Borders of V. cholerae vipAyfp cells are indicated by dashed lines. Scale bar: 2 μm.

C. The fraction of cells showing either diffuse or punctuate VipAYFP fluorescence was determined (n > 500).

D. V. cholerae vipAyfp ΔclpV Δhcp1/2 vipAyfp cells containing either pMPMA4–clpV or the empty vector cells were grown in the presence of 10% (w/v) glucose to mid-logarithmic growth phase. Cells were washed and clpV expression was induced by addition of arabinose [0.0004% (w/v)]. VipAYFP localization was analysed prior and post ClpV production (left) and the fraction of cells showing either diffuse or punctuate VipAYFP fluorescence was determined (n = 100).

Together these data suggest that VipA/VipB tubules can in principle form in the absence of a functional T6SS (class I mutants) but are immediately disintegrated in presence of ClpV. This explains why VipA–YFP displays a diffuse staining in class I mutant cells, but quantitatively forms foci upon additional deletion of clpV.

VipA/VipB tubules mislocalize in ΔclpV cells

The lack of ClpV activity has a profound effect on VipA–YFP localization, causing quantitative and static foci formation. To test whether ClpV also influences the position of VipA/VipB tubules within cells, we monitored tubule localization in thin sections of ΔclpV cells by immunoelectron microcopy using VipB-specific antibodies (Fig. 4A). In ΔclpV cells the number of visible tubules was twofold increased in comparison to WT (5.7%, n = 1050). VipA/VipB tubules also showed a different cellular localization, as they were frequently no longer oriented near-perpendicular to the membrane, but were located either randomly in the cytosol or parallel to the membrane (Fig. 4 and D). In addition, 25% of the positive sections showed two or more tubules, contrasting sections of WT cells, which always displayed a single tubule. In some cases, bundles of VipA/VipB assemblies were noticed with individual tubules lying adjacent to one another (Fig. 4A). These findings demonstrate that the lack of ClpV function does not only stabilize VipA/VipB tubules as judged from increased VipA–YFP foci formation (Fig. 3 and C), but also leads to tubule mislocalization (Fig. 4A). We next performed the respective analysis in Δhcp1hcp2 and Δfha mutant cells, which show a diffuse fluorescent staining of VipA–YFP (Fig. 3A). The number of long VipA/VipB tubules located near-perpendicular to the membrane was 12.6-fold reduced in Δhcp1hcp2 cells compared to WT and long tubules were no longer detectable in Δfha cells (Fig. 4B, C and E). In both mutant cells the majority of identified VipA/VipB tubules were significantly shorter (< 100 nm) and were distributed throughout the cytosol without apparent connection to the inner membrane (Fig. 4B, C and E). The existence of a small fraction of long VipA/VipB tubules in Δhcp1hcp2 cells contradicts at first glance the diffuse VipA–YFP fluorescence in intact Δhcp1hcp2 cells (Fig. 3A) but can be explained by the fixation protocol used for immunoelectron microscopy, since VipA–YFP foci were also observed in a subpopulation of fixed Δhcp1hcp2 cells (15%, n = 250) (Fig. S5).

Figure 4.

VipA/VipB tubules mislocalize in V. cholerae ΔclpV, Δhcp1/hcp2, Δfha and ΔclpV Δhcp1/hcp2 mutant cells.

A–D. Cryosections of cells were immunogold-labelled with VipB-specific antibodies and revealed the indicated different cellular locations of VipA/VipB tubules compared to WT cells. (A) V. cholerae ΔclpV cells: bundles of adjacent tubules (left), tubules located parallel to the membrane (middle) and two tubules per cell (right). Scale bar: 100 nm. (B) V. cholerae Δhcp1/hcp2: longer tubules (left), bundles of tubules with more than one tubule per cell (middle) and short tubular fragments (right). (C) V. cholerae Δfha cells only showed the presence of short VipA/VipB fragments. (D) V. cholerae ΔclpV Δhcp1/hcp2 cells: parallel and piled stacks of tubules. Tubular-like structures are marked by arrows. Scale bar: 100 nm.

E. The fractions of thin sections showing the indicated VipB localization were determined.

We also analysed VipA/VipB localization in ΔclpV Δhcp1hcp2 mutant cells, which show quantitative VipA–YFP/VipB foci formation (Fig. 3B). Here, the number of elongated VipA/VipB structures located near-perpendicular to the membrane was 4.8-fold reduced compared to WT cells (Fig. 4 and E). The majority of VipB-stained structures formed parallel or piled assemblies, positioned either randomly in the cytosol or parallel to the inner membrane.

Summarized, the EM analysis of VipB localization in ΔclpV, Δhcp1hcp2, Δfha and ΔclpV Δhcp1hcp2 cells is matching VipA–YFP localization in the respective knockout cells (Fig. 3 and B) and additionally indicates mislocalization of VipA/VipB tubules. Both data sets point to a function of ClpV, Hcp and Fha in proper VipA/VipB localization.

VipAYFP foci elongation is compromised in ΔclpV daughter cells

The results presented so far indicate that a lack of ClpV activity has two functional consequences on T6S. First, VipA/VipB tubules are no longer disassembled after contraction. Second, the tubules are frequently not correctly positioned implying that they are no longer linked to a T6SS assembly. Since 40% of VipA/VipB tubules identified in ΔclpV mutant cells were still oriented near-perpendicular to the membrane (Fig. 4D), such tubules might still be functional in a single round of T6S. To determine whether a single round of T6S is possible in ΔclpV mutant cells, we monitored the dynamics of VipA–YFP in V. cholerae vipAyfp WT and ΔclpV daughter cells immediately after cell division. The observation of elongated VipA–YFP structures in new-born daughters served as readout for a newly initiated T6S cycle. Formation and clearance of VipA–YFP foci and elongated structures were frequently observed in daughters of WT cells (Fig. 5A). In total 131 foci dynamically formed in WT daughters (n = 30) within a time frame of 60 min and 61% of such foci appeared elongated. Thirty-one static VipA–YFP foci formed in ΔclpV daughter cells (n = 30) during the same period of time and only 16% of those foci appeared elongated (Fig. 5B). These findings suggest that single, productive rounds of T6S take place in ΔclpV cells but are infrequent events. This points to an additional function of ClpV beyond VipA/VipB tubule recycling by increasing the efficiency of secretion-competent VipA/VipB tubule formation.

Figure 5.

V. cholerae ΔclpV daughter cells initiate new rounds of T6S at low frequency. Time-lapse experiments of V. cholerae WT (A) and ΔclpV (B) cells expressing vipAyfp. Punctuate and elongated VipAYFP structures forming in mother or daughter cells are indicated by white (mother cell) and yellow (daughter cell) arrows. The upper time lapse of ΔclpV vipAyfp cells shows the formation of an elongated structure in one daughter cell. Borders of V. cholerae vipAyfp mother and daughter cells are indicated by white and yellow dashed lines respectively. Scale bar: 2 μm.

A hybrid AAA+ protein, harbouring the N-terminal specificity domain of ClpV, is functional in T6S

The crucial function of ClpV in T6S of V. cholerae given, an activity control can be assumed that guarantees for efficient and timely VipA/VipB tubule disassembly. As shown, VipA–YFP transiently forms foci and elongated structures in V. cholerae vipAyfp WT cells, which are subsequently removed (Fig. 1D). This implies that VipA/VipB tubules exist in different structural states that are either resistant or vulnerable to disassembly by ClpV. Such states could involve elongated and contracted conformations, in which only the contracted form is susceptible to ClpV-mediated disintegration. The formation of elongated VipA/VipB tubules seems to rely on multiple T6SS components including Hcp, VCA0109, IcmF and Fha, as neither foci nor elongated VipA–YFP structures were detected in respective mutant cells (Fig. 3 and C). Quantitative VipA–YFP foci formation was observed upon additional loss of ClpV activity, indicating that VipA/VipB tubules form in absence of a functional T6SS (Fig. 3 and C). We suggest that these tubules exhibit a contracted conformation, which would be immediately subjected to disintegration provided the presence of ClpV. Evidence for exclusive targeting of ClpV to contracted VipA/VipB tubules was recently provided (Basler and Mekalanos, 2012; #6512). Such mechanism would represent an in-built activity control of the AAA+ protein as it would restrict its disassembly activity to a specific step of the T6S cycle. Such substrate-mediated ClpV activity control could supersede the involvement of additional regulatory players. On the other hand, ClpV harbours next to its AAA domains two extra domains, which often control AAA+ protein activity (Fig. 6A). The N-terminal domain (N-domain) mediates ClpV binding to VipA/VipB tubules (Bonemann et al., 2009; Pietrosiuk et al., 2011), while the function of the middle domain (M-domain) is currently unknown. M-domains of related AAA+ proteins (e.g. ClpB, ClpC) control ATPase and chaperone activity through binding to partner proteins (Kirstein et al., 2006; Wang et al., 2011; Oguchi et al., 2012). To address a potential role of the M-domain in ClpV activity control, we first generated a ClpV-ΔM deletion variant. ClpV-ΔM exhibited folding deficiencies and aggregated in vivo, thereby obstructing the analysis of the ClpV M-domain via this approach (data not shown).

Figure 6.

A hybrid AAA+ protein can replace ClpV function in vivo.

A. Schematic domain organization of V. cholerae ClpV, ClpB and the hybrid ClpVB fusion proteins. The position of the Y520D M-domain mutation is indicated by a star.

B. In vitro disassembly of VipA/VipB complexes by the indicated AAA+ proteins was monitored as a decrease of sample turbidity over time.

C. V. cholerae vipAyfp ΔclpB ΔclpV mutant cells expressing pMPM-A4 encoded clpV, clpB, clpVB or clpVBY520D were grown at 37°C in presence of 0.0004% (w/v) (ClpV WT level, left) or 0.01% (w/v) arabinose (ClpV overproduction, right) leading to low or high AAA+ protein production levels. Total levels of AAA+ proteins, VipB and Hcp and export of Hcp (sup) were determined using specific antibodies. Anti-ClpV-N, polyclonal antibody raised against the ClpV N-domain.

D. Time-lapse experiments of V. cholerae vipAyfp ΔclpV cells expressing plasmid-encoded clpVBY520D or clpVB were grown in the presence of arabinose [0.0004% and 0.01% (w/v) respectively]. Punctuate and elongated VipAYFP structures are indicated by arrows. Borders of V. cholerae vipAyfp cells are indicated by dashed lines. Scale bar: 2 μm.

We therefore pursued an alternative strategy, by determining the ability of hybrid AAA+ proteins to replace ClpV function in V. cholerae cells. Hybrid AAA+ proteins harbouring the ClpV N-domain have been demonstrated to disassemble VipA/VipB tubules in vitro (Pietrosiuk et al., 2011; Seyffer et al., 2012). However, it remained unclear whether such in vitro activity is sufficient to restore T6S in V. cholerae ΔclpV mutant cells. Here, we generated and analysed a fusion protein between V. cholerae ClpV and V. cholerae ClpB, termed ClpVB, harbouring the N-domain of ClpV and the ATPase and M-domains of ClpB (Fig. 6A). ClpB solubilizes aggregated proteins in cooperation with the DnaK chaperone system and is not involved in T6S (Winkler et al., 2012). The ClpVB fusion protein is no longer subjected to a potential activity control through ClpV M-domains but is instead controlled by ClpB M-domains, which downregulate ClpB unfolding activity (Oguchi et al., 2012). Accordingly, the disassembly activity of purified ClpVB was strongly reduced compared to ClpV, in agreement with recent findings (Fig. 6B) (Seyffer et al., 2012). To overcome the constraint of reduced ClpVB unfolding power, we made use of a characterized ClpB M-domain mutation (Y503D), which causes ClpB derepression and leads to strongly increased unfolding activity (Oguchi et al., 2012). The corresponding hybrid ClpVB–Y520D variant exhibited high VipA/VipB disassembly activity in vitro that was comparable to ClpV activity (Fig. 6B). We next tested for the ability of ClpVB and ClpVB–Y520D to restore the appearance of Hcp in cell culture supernatants of V. cholerae ΔclpV cells. To exclude the formation of mixed oligomers between hybrid ClpVB proteins and V. cholerae ClpB, we generated ΔclpB ΔclpV double mutant cells. In addition, we integrated the VipA–YFP fusion at the authentic genomic site, allowing us to use VipA–YFP localization as a second functional readout. Plasmid-encoded clpV, clpB and clpVB fusion constructs were expressed in V. cholerae vipAyfp ΔclpB ΔclpV cells from arabinose-controlled promoters. All AAA+ proteins were first produced at ClpV wild-type levels. Low levels of secreted Hcp were determined in supernatants of V. cholerae ΔclpB ΔclpV cells expressing plasmid-encoded clpVB compared to clpV expressing mutant cells (Fig. 6C). Hcp levels were slightly higher compared to control cells expressing clpB, indicating partial restoration of T6S (Fig. 6C). Expression of clpVBY520D led to increased secreted Hcp levels that were similar to those observed upon clpV expression, indicating that the highly active hybrid protein can complement the T6S defect of V. cholerae ΔclpB ΔclpV cells (Fig. 6C). Next, we tested whether ClpVB overproduction can compensate for its reduced disassembly activity. Overproduction of ClpVB increased the amount of secreted Hcp, confirming the ability of hybrid AAA+ proteins to replace ClpV function in vivo.

We also monitored the localization of VipA–YFP in V. cholerae vipAyfp ΔclpB ΔclpV expressing clpV, clpB, clpVB and clpVBY520D to comparable levels. Expression of either clpVB or clpVBY520D partially restored the occurrence of cytosolic VipA–YFP fluorescence, which was not observed upon clpB expression, indicating partial activity of both hybrid proteins (Fig. S6). The degree of diffuse VipA–YFP fluorescence was higher in ClpVB–Y520D expressing cells and was already observed at lower expression levels. Expression of ClpVB–Y520D at moderate levels and ClpVB at high levels also restored VipA–YFP dynamics in V. cholerae vipAyfp ΔclpB ΔclpV cells, indicating that the hybrid proteins allow for multiple cycles of VipA/VipB tubule formation and disintegration (Fig. 6D). Together these findings demonstrate that the fusion of the ClpV N-terminal domain to an AAA+ protein not engaged in T6S is sufficient to at least partially restore T6S in ΔclpV mutant cells.


We report on the existence and characterization of VipA/VipB tubules in V. cholerae V52 cells and the physiological meaning of tubule disassembly by the ATPase ClpV. VipA/VipB tubules span almost the entire length or width of WT cells and are always oriented near-perpendicular to the membrane. Our findings are consistent with a recent study reporting on the identification of tubular structures in intact V. cholerae WT but not in ΔvipB cells by cryo-electron tomography (Basler et al., 2012). The authors purified the tubular structures and identified VipA and VipB via mass spectrometry as the two major proteins. In line with this finding, we here provide direct evidence that such structures represent VipA/VipB tubules by their specific labelling with VipB antibodies in immunoelectron microscopy (Fig. 1B). Using a genomically integrated functional VipA–YFP fusion we show that VipA–YFP/VipB tubules are highly dynamic and exhibit cycles of foci formation, elongation, shortening and disassembly. These findings are also in agreement with a recent report describing the dynamics of a plasmid-encoded VipA–sfGFP construct in V. cholerae (Basler et al., 2012). Together these results support a model in which VipA/VipB assemblies are functioning as contractile structures in analogy to viral tail sheath proteins to eject engulfed Hcp tubes upon contraction. Extended and contracted tubular structures were observed in cryo-electron tomographic images of V. cholerae cells (Basler et al., 2012). Extended structures exhibited central density whereas contracted ones appeared hollow, providing indirect evidence for Hcp engulfment (Basler et al., 2012). Here, we show that VipA/VipB–YFP tubules are non-functional in vivo, despite forming tubules that can be disassembled by ClpV in vitro. The fused YFP moiety blocks the central channel of VipA/VipB–YFP tubules (Fig. S2B), indicating that channel integrity is required for T6S, presumably for accommodating an Hcp tube.

VipA/VipB tubules were originally identified as substrate of the AAA+ protein ClpV but the physiological relevance of VipA/VipB disassembly in the context of T6S remained unclear. Here, we show that the cellular localization of V. cholerae ClpV depends on VipA/VipB, which attract the ATPase upon tubule formation. In agreement with our findings, it was recently shown that the localization of a ClpV–GFP fusion protein remained diffuse in V. cholerae and P. aeruginosa ΔvipA mutant cells (Basler and Mekalanos, 2012). Other T6SS components, which are required to form a functional T6SS and to allow for elongated VipA/VipB tubule formation, are expected to indirectly affect ClpV localization. Accordingly, ClpV was randomly distributed in the cytosol of V. cholerae Δfha cells, which also did not display long VipA/VipB tubules (Figs 4). In P. aeruginosa Fha has been reported to directly control ClpV localization (Mougous et al., 2007). Further experiments are required to elucidate the potential differences of V. cholerae and P. aeruginosa Fha proteins, concerning their role in controlling VipA/VipB and ClpV localization.

ClpV decorates the entire VipA/VipB tubule in vivo (Fig. 2C), indicating a role in tubule disassembly during T6S. Accordingly, ΔclpV cells exhibited quantitative and static foci formation of VipA–YFP foci. These findings are in line with a recycling function of ClpV, removing contracted VipA/VipB tubules after contraction and allowing for a new round of T6S. Evidence for specific binding of ClpV to contracted but not elongated VipA/VipB tubules was recently provided (Basler and Mekalanos, 2012). The formation of elongated VipA/VipB tubules, which are protected from ClpV activity, seems to be dependent on a functional T6SS assembly, as no punctuate or elongated structures of VipA–YFP were observed in all analysed knockouts of T6SS core components (Fig. 3A). A mere recycling activity of ClpV would allow for single rounds of T6S in the absence of ClpV. Consistent with such possibility, V. cholerae ΔclpV cells exhibit low killing activity towards E. coli cells (Zheng et al., 2011) and form elongated VipA–YFP structures after cell division at low frequency (Fig. 5B).

Next to its role in removing VipA/VipB tubules after ejection of Hcp, we identified a second function of ClpV, which we suggest to be of equal importance. We provide independent evidences that ClpV prevents the formation of non-productive VipA/VipB tubules, which are not linked to a T6SS assembly. First, the absence of ClpV causes mislocalization of VipA/VipB tubules in vivo. The majority of tubules identified by immunoelectron microscopy in ΔclpV cells are no longer oriented near-perpendicular to the membrane (Fig. 4 and D). Second, V. cholerae vipAyfp cells lacking T6SS core components solely exhibit a diffuse VipA–YFP fluorescence in presence of ClpV, but quantitative foci formation upon additional clpV deletion. This finding indicates that contracted VipA/VipB tubules do form in the absence of a functional T6SS. Accordingly, VipA and VipB form tubular complexes if produced in E. coli cells lacking a T6SS (Bonemann et al., 2009).

We suggest that VipA/VipB tubules that form spontaneously without connection to a T6SS adopt a contracted conformation and are immediately removed by ClpV. This ClpV activity indirectly directs VipA/VipB to a T6SS assembly, which triggers the polymerization of VipA/VipB into its extended conformation, which is insensitive to ClpV activity. ClpV activity also ensures the presence of a pool of non-assembled VipA and VipB subunits required for tubule elongation. In line with such activity V. cholerae vipAyfp wild-type cells exhibit a large degree of diffuse VipA–YFP even in the presence of initially formed foci, in contrast to ΔclpV cells.

The importance of the newly identified role of ClpV in preventing non-productive VipA/VipB tubule formation became obvious when we monitored VipA/VipB dynamics in V. cholerae vipAyfp WT and ΔclpV daughter cells after cell division. Whereas new-born wild-type cells predominantly harboured dynamic elongated VipA–YFP structures, the vast majority of ΔclpV daughters contained static VipA–YFP foci and only a minority appeared elongated (Fig. 5). This finding underscores the physiological relevance of ClpV in preventing random, non-productive tubule formation and indicates that the frequency of T6SS-linked elongated VipA/VipB tubule formation in absence of ClpV is low.

The exclusive binding of ClpV to contracted VipA/VipB tubules restricts its activity to a specific step of the T6S cycle and allows for autonomous regulation of ClpV activity (Basler and Mekalanos, 2012). In agreement with such mechanism, we show that targeting a different AAA+ protein to VipA/VipB tubules via fusion of ClpV N-domains allows for T6S. Increased levels of ClpV affected neither Hcp secretion nor VipA–YFP localization (Fig. 6 and D), also arguing against the crucial involvement of a regulatory factor, which would become out titrated under such conditions. This most simple mechanism of ClpV activity control in V. cholerae does not exclude the presence of additional regulatory circuits, which have been proposed in P. aeruginosa (Mougous et al., 2007). Here, opposing activities of a kinase (PpkA) and phosphatase (PppA) pair are controlling the phosphorylation status of Fha, which is suggested to recruit ClpV to the T6SS assembly through direct interaction (Mougous et al., 2007; Hsu et al., 2009). It is currently unknown whether such interaction involves the ClpV M-domain. It is tempting to speculate that the extra domain might allow for factor binding to regulate ClpV activity under specific conditions.

Taken together, we demonstrate that ClpV-mediated disassembly of VipA/VipB is required at two distinct steps of T6S. First, it allows for the formation of elongated VipA/VipB tubules at a T6SS assembly by suppressing the formation of non-productive contracted tubules. Second, ClpV removes correctly positioned tubules after contraction to initiate a new secretion process.

Experimental procedures

Strains and plasmids

Table S1 lists all strains and plasmids used in this study. All bacterial strains were grown in LB media at 37°C if not stated otherwise. Antibiotics were used at the following concentrations: 2 μg ml−1 (V. cholerae) and 20 μg ml−1 (E. coli) chloramphenicol, 100 μg ml−1 (V. cholerae) and 50 μg ml−1 (E. coli) streptomycin, 200 μg ml−1 (V. cholerae) and 100 μg ml−1 (E. coli) ampicillin. PCR reactions and cloning procedures followed standard protocols.

Vibrio cholerae in frame gene deletions and vipAyfp or vipByfp integrations were performed as described before by using the pDS132 suicide plasmid for allelic replacement of the gene of interest (Philippe et al., 2004). Deletions and integrations were identified by colony PCR and confirmed by immunoblot analysis using specific antibodies. V. cholerae deletion mutants were complemented by pMPM-A4 harbouring the gene of interest, allowing for arabinose-controlled gene expression. Total protein levels were adjusted to wild-type levels by titration of arabinose levels and comparing the signal intensities upon immunoblot analysis.

For transformation of plasmids into V. cholerae electrocompetent cells were prepared as described (Hamashima et al., 1995). Plasmids were transformed as described (Bonemann et al., 2009).


C-terminal His6-tagged V. cholerae ClpV, ClpB and hybrid ClpVB constructs were purified after overproduction from E. coli ΔclpB mutant cells following the instructions of the manufacturer (Macherey-Nagel). C-terminal His6-tagged VipB–YFP was purified accordingly after co-overproduction with VipA from E. coli XL1 Blue cells. Protein concentrations were determined with the Bio-Rad Bradford assay using bovine serum albumin as standard.

Biochemical assays

The analysis of the V. cholerae secretome was performed as described (Pukatzki et al., 2006). Disassembly of VipA/VipB tubules or VipA/VipB–YFP tubules (0.5 μM each) by ClpV or hybrid ClpVB constructs (0.125 μM each) was analysed with a Perkin-Elmer LS50B spectrofluorimeter by monitoring the decrease of turbidity (excitation and emission wavelength: 550 nm). The reactions were performed at 25°C in reaction buffer (50 mM Tris pH 7.5, 150 mM KCl, 20 mM MgCl2, 2 mM DTT) containing an ATP Regenerating System (2 mM ATP, 3 mM phosphoenolpyruvate, 20 ng ml−1 pyruvate kinase).

Fluorescence microscopy

For microscopy cell cultures were grown to mid-log phase in LB media at 37°C. When appropriate, gene expression was induced by addition of arabinose at concentrations indicated (Table S2).

For conventional fluorescence microscopy, V. cholerae cells of mid-log phase were diluted in LB medium to OD600 ∼ 0.1 and immobilized on agarose pads containing 0.2% glucose or 10% LB. Agarose pads were sealed with silicone paste and covered with coverslips.

For immunofluorescence, mid-log phase V. cholerae cells were harvested by centrifugation. The pellet was resuspended in PBS and fixed in pre-cooled methanol for 10–20 min at −20°C (Teleman et al., 1998). Fifty to 200 μl of cell suspension were distributed on poly-l-lysine coated coverslips and dried until the methanol had evaporated. Adhered cells were incubated for 30 min with 2 mg ml−1 lysozyme in GTE buffer (50 mM glucose, 25 mM Tris, 1 mM EDTA, pH 7.5). Next, cells were washed with PBS and subsequently blocked with 2% (w/v) BSA in PBS for 12 h. Incubations with primary and secondary antibodies (Table S3) were performed at room temperature for 1 h with extensive intermediate washing steps [PBS + 0.05% (v/v) Tween-20]. Before imaging coverslips where mounted on conventional microscopy slides in the presence of PBS + 1 mg ml−1 p-phenylenediamnie + 88% (v/v) glycerol (Sourjik and Berg, 2000).

Images were taken using an Olympus IX81 inverted system equipped with a PLAPO 100×/1.45 NA oil objective or a UAPO 100× TIRF/1.49 NA oil objective, a Hamamatsu Orca-R2 or EMCCD C9100-02 camera and appropriate filter sets [Dualband CFP/YFP sbx HC filter set, Quadband DAPI/FITC/Cy3(TexasRed)/Cy5 sbx HC filter set].

Images were analysed using the ImageJ software ( Processing includes cropping and scaling of images, adjustment of brightness and contrast and creation of animated sequences. Spectral unmixing of respective fluorophores was performed using the Olympus xcellence software.

Electron microscopy

Vibrio cholerae cells were grown to mid-log phase and fixed with 4% paraformaldehyde, 0.1% glutaraldehyde and 0.1 M PHEM buffer (60 mM Pipes, 25 mM HEPES, 2 mM MgCl2, 10 mM EGTA, pH 6.9) for 60 min at room temperature.

After fixation, cells were washed twice with PBS followed by washing with PBS/50 mM glycine. Cell pellets were resuspended in PBS + 10% (w/v) gelatin in PBS. After solidification the cell pellet was cut into small cubes, infiltrated with 2.3 M sucrose (in PBS) and flash-frozen in liquid nitrogen. Forty to 60 nm cryosections were generated using a Leica UC6 cryo-microtome with Diatome ultra 35° diamond knifes at −110°C.

For immunogold labelling grids with thawed sections facing the surface were incubated on PBS with 2% gelatine followed by PBS only at 37°C to remove remaining gelatine. Grids were washed with PBS/50 mM glycine and blocked for 30 min at room temperature [50 mM glycine, 0.8% (w/v), 0.1% (w/v) fish skin gelatine in PBS]. Incubations with primary and secondary antibodies were performed at room temperature for 30 min using the antibody dilutions indicated in Table S3. Prior to contrasting with uranyl acetate in methylcellulose on ice, grids were washed with double-distilled water.

For double labelling an additional fixation step using 0.1% glutaraldehyde in PBS followed by extensive washing with PBS/50 mM glycine was added after the first incubation with primary and secondary antibody (Raposo et al., 1997; Ghosh et al., 2003). VipA/VipB–YFP complexes were analysed by transmission electron microscopy. Images were recorded at 10 000–50 000 × magnification with a Zeiss EM900 microscope.


We thank Eva Kummer and Juliane Winkler for critical reading of the manuscript. We thank Bernd Bukau for continuous support. This work was supported by a grant from the Deutsche Forschungsgemeinschaft to A. M. (MO970/3). N. K., A. P. and F. S. were supported by the Hartmut Hoffman-Berling International Graduate School of Molecular and Cellular Biology (HBIGS).

The authors have no conflict of interest to declare.