Regulation of kinesin-1 activity by the Salmonella enterica effectors PipB2 and SifA

The species Salmonella enterica is composed of a large number of serotypes, including typhoid strains, which cause severe systemic disease (typhoid fever) in humans, and non-typhoid strains responsible for acute gastroenteritis. The latter may also cause systemic infection in hosts already weakened by another viral or parasitic infection (Gordon et al., 2008). The non-typhoid strain S. Typhimurium (Salmonella enterica subsp. enterica serotype Typhimurium) is commonly used as a model for studying host-pathogen interactions (LaRock et al., 2015).

Salmonella is an enteropathogenic bacterium. During the infectious process, bacteria that have reached the gut cross the intestinal barrier, preferably through the M cells of Peyer's patches (Lelouard et al., 2012) and to a lesser extent by infecting enterocytes. They then migrate and are found in a multitude of cell types, especially in cells of the immune system (Aussel et al., 2011), in Peyer's patches, in liver, in spleen and also in epithelial cells, e.g. the gallbladder (Knodler et al., 2010).

The pathogenicity of Salmonella depends on its ability to replicate intracellularly in a membrane-bound compartment called the Salmonella-containing vacuole (SCV). The membrane of the SCV is enriched with lysosomal proteins. Infected cells show a rarefaction of lysosomal vesicles that seem to have merged with the SCVs. This recruitment of lysosomal membranes takes place thanks to the formation of membrane tubules that are formed from the SCVs and extend on the network of microtubules in the cells. These tubular membrane structures, called Salmonella-induced tubules, promote exchange with lysosomal vesicles, membrane recruitment and transport of material to the SCVs (Liss et al., 2017), and ultimately provide the nutrients necessary for intracellular bacterial replication (Noster et al., 2019).

This unique process is orchestrated by a series of bacterial proteins that are injected into the host cell cytosol by a type III secretion system (T3SS), which is encoded by the second pathogenicity island (T3SS-2). The T3SS-2 effector proteins influence a large number of cellular processes, such as gene expression (Mazurkiewicz et al., 2008), localization of membrane proteins (Bayer-Santos et al., 2016) or dynamics of the cytoskeleton (Abrahams et al., 2006; Méresse et al., 2001). By interacting with the microtubule motor protein kinesin-1, the two effector proteins PipB2 (Henry et al., 2006) and SifA (Boucrot et al., 2005) play a particularly important role in the formation and function of Salmonella-induced tubules. Kinesin-1 is a hetero-tetrameric ATPase composed of two heavy chains (KHCs; KIF5A–KIF5C) and two light chains (KLCs; KLC1–KLC4) that promotes the transport of cargoes toward the growing end of microtubules. PipB2 interacts directly with KLCs (Henry et al., 2006), while SifA binds to SKIP (also known as PLEKHM2), an adaptor host protein that interacts with KLCs (Boucrot et al., 2005; Dumont et al., 2010).

In the absence of cargo binding, kinesin-1 adopts a folded, compact conformation that is auto-inhibited and cytosolic (Cai et al., 2007). Cargo binding relieves autoinhibition and results in an extended structure, releasing KHC motor domains that engage with the microtubules (Hirokawa and Noda, 2008) and convert the chemical energy of ATP hydrolysis into motion along microtubules. The mechanism of activation of kinesin-1 is not yet fully understood. It involves interactions between the cargo and both KHCs and KLCs, the latter being involved in the inhibition of the enzymatic activity of the KHCs. In the cellular context, binding of the cargo protein JIP1 (also known as MAPK8IP1) to KLCs is not sufficient to activate kinesin-1 and requires additional interactions of KHCs with FEZ1 (Blasius et al., 2007). The TPR domain of KLC recognizes a peptide sequence composed of a tryptophan (Dodding et al., 2011) or a tyrosine (Pernigo et al., 2018) residue flanked by acid residues, sequences that have recently been named W-acidic and Y-acidic. The W-acidic motif originally characterized in vaccinia virus protein A36 is also found in many adaptor proteins such as SKIP, which contains two of them (Pernigo et al., 2013). A recent study showed that SKIP can also interact with KHCs when already interacting with KLCs, and that this double interaction relieves the auto-inhibition of kinesin-1 (Sanger et al., 2017).

In Salmonella-infected cells, SifA and PipB2 are important for the formation and/or elongation of Salmonella-induced tubules. Cells infected with a mutant ΔpipB2 are characterized by the formation of short tubules (Knodler and Steele-Mortimer, 2005), whereas these are lacking or very rare in the absence of SifA (Stein et al., 1996) or SKIP (Boucrot et al., 2005). The specific impact of PipB2 and SifA on kinesin-1 activation is still not understood. However, based on previous observations and, in particular, because PipB2-bound kinesin-1 accumulates on SCVs in the absence of SifA, we proposed that the SifA/SKIP complex is necessary for the activation of the kinesin-1 recruited by PipB2 (Dumont et al., 2010; Leone and Méresse, 2011). In this study, we have used in vitro and in cellulo approaches to investigate the activation status of kinesin-1 when recruited by PipB2 and SifA effectors.

To better understand the role played by kinesin-1 in the formation of vesicles and tubules emerging from SCVs, we analysed the consequences of the absence of the molecular motor in Salmonella-infected cells. For this purpose, we differentiated macrophages from Hoxb8-immortalized macrophage-committed progenitor cells (Redecke et al., 2013; Wang et al., 2006). These cells were derived from C57BL/6 mice expressing KIF5B (kinesin-1 heavy chain) in hematopoietic cells or not (Munoz et al., 2016). By western blotting, anti-KHC antibody recognized a band between 100 and 150 kDa in KIF5B^+/+ macrophages and mouse brain cytosol that was absent in KIF5B^−/− macrophages (Fig. 1A), thus confirming the absence of functional kinesin-1.

The macrophages were infected with a S. Typhimurium strain (wild-type) expressing GFP and a HA-tagged version of the T3SS-2 effector protein SseJ and observed by confocal microscopy 14 h after infection. GFP was used to detect the presence of bacteria in cells. LAMP1 and SseJ were used as membrane markers of bacterial compartments. While Salmonella-induced tubules form in most infected epithelial cells, these membrane structures are only rarely seen in non-activated macrophages (Knodler et al., 2003; Krieger et al., 2014). In macrophages, effectors tend to be located on discrete vesicles throughout the cell (Freeman et al., 2003; Ohlson et al., 2005). Unsurprisingly, in KIF5B^+/+ macrophages infected with wild-type Salmonella, SseJ was present on SCVs and also at distance from SCVs in association with vesicular structures (Fig. 1B). In contrast, SseJ was essentially localized on SCVs in the absence of kinesin-1 and also in KIF5B^+/+ cells infected with the ΔsifA mutant (Fig. 1B,C) as expected (Zhao et al., 2015). These results indicate that kinesin-1 is required for the mechanisms by which SseJ traffics from SCVs to discrete vesicular structures. They also support the previously proposed hypothesis (Dumont et al., 2010) that associates the absence of vesicles/tubules in cells infected by a ΔsifA mutant with a low level of activation of kinesin-1 that, when recruited by PipB2, accumulates on SCVs (Henry et al., 2006).

We also observed an influence of kinesin-1 on the position of SCVs. Whereas in the KIF5B^+/+ and KIF5B^−/− macrophages, the wild-type SCVs were located in the juxtanuclear region (Fig. 1C), they were grouped very compactly and close to the nucleus in the absence of kinesin-1 (Fig. 1B). This indicates that kinesin-1 has an anterograde transport activity that maintains the SCVs in the juxtanuclear region but at some distance from the nucleus. In KIF5B^+/+ macrophages, two populations of ΔsifA SCVs were distinguished according to whether or not they were positive for SseJ. The positive ones, which are also those that accumulate kinesin-1 (Zhao et al., 2015), showed a localization much further from the nucleus than other vacuoles. This difference was not found in the KIF5B^−/− macrophages where vacuoles indistinctly adopt a juxtanuclear position (Fig. 1B,C), and shows that the kinesin-1 present on ΔsifA SCVs, although not very active, is nevertheless responsible for the anterograde movement of bacterial vacuoles. Overall, these results underline the importance of kinesin-1 in the formation of effector-positive membrane vesicles derived from SCVs, indicating that this molecular motor is also necessary for the anterograde movement of SCVs and confirming the low state of activation of PipB2-bound kinesin-1 present on ΔsifA SCVs.

PipB2 plays a key role in the recruitment of kinesin-1 on SCVs (Boucrot et al., 2005; Henry et al., 2006). However, it is still unknown whether kinesin-1 is the only molecular motor recruited by PipB2 and whether this effector can distinguish between the activation states of kinesin-1 or play a role in the activation of this molecular motor. To gain insights on the motor activity of PipB2-bound proteins, we reconstituted an in vitro microtubule gliding assay. It consists of a streptavidin-coated glass chamber successively filled with biotinylated Δ17PipB2 (Bio-PipB2), a fraction of molecular motors, fluorescent microtubules and a motility buffer containing 1 mM ATP (Fig. S1A). As a positive control, the glass chamber was coated with the biotinylated motor domain of Drosophila kinesin heavy chain (K401-Bio). This constitutively active recombinant protein (Roux et al., 2002; Subramanian and Gelles, 2007) induced a microtubule gliding with a velocity of 0.05±0.01 µm/s and a narrow distribution of gliding velocities (Fig. 2; Fig. S2A and Movie 1). In the presence of Bio-PipB2 and MAP-depleted mouse brain cytosol as a source of molecular motors, we observed persistent attachment and sustained movement of the microtubules after injection of motility buffer (Fig. 2; Fig. S2B,C). The microtubules displayed a gliding motility throughout the recorded period of 15 min (Movie 2) with a velocity of 0.22±0.01 µm/s (Fig. S2A), which is higher than that observed with the motor domain of Drosophila KHC but inferior to the gliding speed of 0.5-0.7 µm/s reported in the literature in a similar assay using bovine kinesin (Grummt et al., 1998). When biotinylated BSA (Bio-BSA), our negative control, was injected instead of PipB2, we noticed the gliding of a small fraction of microtubules (Fig. 2; Fig. S2C) and the rapid decrease in the number of attached microtubules (Fig. S2B, Movie 3). We concluded that the attachment and the movement of the microtubules depend on PipB2.

To determine whether kinesin-1 is the cytosolic factor mediating microtubule gliding, we fractionated a cytosol preparation enriched in microtubule-based motors by centrifugation on a sucrose gradient. Fraction 7, which presented the highest concentration of kinesin-1 (Fig. S3A) was tested in the gliding assay. We observed a velocity of 0.15±0.01 µm/s, of the same order of magnitude as that obtained with MAP-depleted mouse brain cytosol (Fig. 2; Fig. S2A and Movie 4). To confirm the involvement of kinesin-1 in the PipB2-mediated motility, we prepared cytosol fractions from Hoxb8 macrophages. KIF5B^+/+ cytosol promoted the gliding of microtubules with a speed of 0.23±0.02 µm/s (Fig. 2; Fig. S2A and Movie 5). In the presence of KIF5B^−/− cytosol, a small fraction of microtubules that partially attached sometimes showed random mobility over a short period before detaching (Fig. 2; Fig. S2C, S2D and Movie 6). Overall, these results indicate that kinesin-1 is responsible for the gliding of microtubules observed in this assay and that PipB2-bound kinesin-1 is capable of engaging microtubules and inducing their movement in the presence of ATP.

The formation of Salmonella-induced tubules from SCVs requires an intact microtubule network where tubules exhibit a bidirectional movement (Drecktrah et al., 2008; Krieger et al., 2014; Rajashekar et al., 2008). Using a minimal biomimetic membrane system in vitro, we tested whether the PipB2/kinesin-1 complex could provide enough force to generate membrane tubule networks from giant unilamellar vesicles (GUVs). GUVs were made fluorescent by addition of 1% β-BODIPY-DOPC and contained 0.1% biotinylated-DOPE, making them capable of binding biotinylated proteins in the presence of streptavidin (Fig. 3A). We modified the tube pulling assay previously described (Leduc et al., 2004). Briefly, in a flow chamber covered by fluorescent microtubules, we sequentially injected the biotinylated protein in the presence of streptavidin, a motility buffer containing 1 mM ATP and GUVs (Fig. S1B). We used K401-Bio and Bio-BSA as positive and negative controls, respectively. In the presence of K401-Bio, we observed the formation of an extended network of tubules, pulled from the GUVs, which extended along the microtubules network (Fig. S4), whereas no tubule was observed with Bio-BSA (Fig. 3B). These results validated the experimental system. When using Bio-PipB2 pre-incubated with a kinesin-1-containing fraction, we obtained networks of tubules similar to those observed with K401-Bio (Fig. 3B). Taken together, these results confirm that PipB2-bound kinesin-1 is capable of engaging microtubules and demonstrate that the force produced by the complex is sufficient to induce membrane tubules.

The vast majority of cellular kinesin-1 is cytosolic and inactive in a folded state that blocks its interaction with microtubules (Cai et al., 2007). Yet our results indicate that at least a part of PipB2-bound kinesin-1 is active. This can be due to the presence of a significant fraction of active kinesin-1 in the mouse cytosol and the capacity of PipB2 to indifferently bind active or auto-inhibited kinesin-1. Another possibility is the capacity of PipB2 to relieve the autoinhibition of the microtubule motor. To discriminate between these hypotheses, we developed a microtubule-binding protein spin-down assay to evaluate the fraction of cytosolic kinesin-1 that is capable of engaging microtubules (activated kinesin-1) in the presence or absence of PipB2. Briefly, in vitro polymerized microtubules were incubated with the proteins of interest and sedimented through a glycerol cushion by ultracentrifugation. As a control, incubations were performed in the absence of polymerized microtubules.

We first analysed by SDS-PAGE and western blotting the distribution of K401-Bio in the supernatant and pellet fractions. This recombinant protein is constitutively active (Roux et al., 2002; Subramanian and Gelles, 2007). The presence of microtubules significantly increased the proportion of the motor domain in the pellet (∼40% compared to ∼10% in the absence of microtubules) (Fig. 4A). This result indicates that K401-Bio is capable of engaging microtubules but remains in the supernatant in the absence of microtubules. Next, we analysed the consequences of introducing Bio-PipB2 on the activation of kinesin-1 in a mouse brain cytosolic fraction. In the absence of microtubules, kinesin-1 was not detectable in the pellet, whether or not PipB2 was present (Fig. 4B). In the presence of microtubules, ∼40% of total kinesin-1 was found in the pellet. This fraction was increased to 90% in the presence of PipB2 (Fig. 4B), which was also present in greater quantity in the pellet in the presence of microtubules (Fig. 4C). We performed a dose-response experiment and observed a correlation between the amount of PipB2 added and kinesin-1 present in the pellet (Fig. 4D). These results indicate that PipB2 activates kinesin-1 and increases the fraction of molecular motor that engages microtubules.

As both PipB2 and SifA interact with kinesin-1 and influence, as the above results show for PipB2, or may influence (SifA; Schroeder et al., 2011) the activity of kinesin-1, we considered the possibility of an interaction between SifA and PipB2. HeLa cells were transfected with plasmids for the expression of GFP or GFP-tagged PipB2 and Myc-tagged variants of SifA. We used cell lysates to immunoprecipitate GFP proteins and analysed the associated proteins by SDS-PAGE and anti-Myc western blotting. SifA co-immunoprecipitated with GFP-PipB2 but not with GFP (Fig. 5A), indicating a specific interaction between the two effectors. We tested the two SifA domains (Ohlson et al., 2008) and found an interaction of PipB2 with the N-terminal part of SifA (Fig. 5A). Finally, we expressed and purified GST-PipB2 and [His]₆-SifA from E. coli and performed a GST pull-down assay to assess whether or not the interaction is direct. We used raw or GST-coated beads as controls for the specificity and the TPR domain of mouse KLC2 (His₆-KLC) as a positive control for the interaction with GST-PipB2. Results indicate a specific direct interaction between SifA and PipB2 (Fig. 5B).

We checked whether SifA alone or in combination with PipB2 and in the presence of cytosol (a source of kinesin-1 and SKIP) could bind and influence the activity state of the molecular motor. As the gliding and tubulation assays used a form of PipB2 with its first 17 amino acid residues deleted (Bio-PipB2), we verified that this domain is not necessary for interaction with SifA (Fig. 5A). We then tested the ability of SifA to interact with kinesin-1 or to influence the interaction between the molecular motor and PipB2. For this, streptavidin beads were coupled with Bio-PipB2, SifA-Bio or an equimolar mixture of both effectors and incubated in the presence of cytosol. Bead-bound kinesin-1 was analysed by western blotting using an antibody directed against the KHCs (Fig. S3B). Kinesin-1 was present on beads coated with PipB2 but not on those coated with SifA. Furthermore, the presence of these two effectors did not influence the recruitment of kinesin-1 in any way. In the microtubule gliding assay, SifA-Bio induced the binding of a low number of microtubules that quickly detached (Fig. S3C and Movie 7). When mixed in an equimolar amount to PipB2, SifA did not modify the PipB2-induced microtubule binding and gliding (Fig. S3C). In the microtubule-binding protein spin-down assay, SifA-Bio did not increase the pelleted fraction of kinesin-1 (Fig. 4B). The conclusion of these experiments is that, under the experimental conditions used, SifA is not capable of changing the activation state of kinesin-1 or of modulating the ability of PipB2 to activate kinesin-1.

So far, our results indicate that PipB2 is able to increase the activity of kinesin-1 in vitro. We therefore checked whether the same was true in the cellular context. For this, we transfected COS-7 cells with plasmids for the expression of GFP-tagged Salmonella effectors and HA-tagged light (mouse KLC2) and heavy (rat KIF5C) chains of kinesin-1. After 24 h we analysed by fluorescence microscopy the impact of Salmonella effectors on the activation of the molecular motor, i.e. its presence in the cytosol versus on microtubules. Overexpressed KHC, KLC and kinesin-1 (KHC/KLC) were essentially cytosolic. KHC was also occasionally detected on microtubules (Fig. S5). In agreement with previous studies (Henry et al., 2006; Knodler and Steele-Mortimer, 2005), PipB2 was found in the cytosol and on vesicles on which KLCs, which interact with PipB2, were detected, unlike KHCs (Fig. 6A, upper row). Compared with PipB (a homologue of PipB2; Knodler et al., 2003), we observed a dramatic change in the localization of kinesin-1 in cells expressing PipB2, which then adopted a fibrous appearance and colocalized with microtubules (Fig. 6A, middle row). The fraction of cells in which we detected the presence of kinesin-1 on microtubules increases from 2% to about 65% in the presence of PipB2, whereas no impact could be observed in cells expressing PipB. In cells expressing KLC we also noted a significant increase of the association with microtubules in the presence of PipB2 (0 versus 20%). This likely reflects the ability of exogenous KLC2 to form functional hybrid molecular motor with endogenous KHC. Finally, we also analysed the impact of SifA in combination with or without PipB2. As seen in vitro, the expression of SifA had no impact on the localization of kinesin-1 or on the ability of PipB2 to promote the engagement of kinesin-1 on microtubules (Fig. 6A,B; Fig. S5).

In order to validate these microscopic observations, we biochemically analysed the activation of kinesin-1 in the cellular context as recently described (Sanger et al., 2017). To do this, we lysed the transfected cells under microtubule-stabilizing conditions, centrifuged the lysates and studied the distribution of ectopically expressed proteins in the supernatant and microtubule pellet. To validate the experimental system, we expressed an ATPase ‘rigor’ mutant of KHC (KHC-R) that associates very stably with microtubules (Guardia et al., 2016) and that was found almost exclusively associated with microtubules in the pellet (Fig. 7). Compared with GFP or PipB, our negative controls, we found that the expression of PipB2 significantly increased the fraction of kinesin-1 associated with microtubules, as was also the case with SKIP, our positive control (Sanger et al., 2017) (Fig. 7). These results confirm the microscopic observations and indicate that, as seen in vitro, PipB2 promotes the activation of kinesin-1 in this cellular context.

Here, we show that the interaction between the T3SS-2 effector PipB2 and the KLCs leads to the activation of kinesin-1. This is demonstrated in vitro by the activity of PipB2-recruited kinesin-1 in microtubule gliding and GUV tubulation assays, by the ability of PipB2 to increase the fraction of kinesin-1 bound to microtubules, and by the re-localization of kinesin-1 from the cytosol to the microtubule cytoskeleton in cells expressing PipB2.

Kinesin-1 mainly exists in a folded conformation that is auto-inhibited, i.e. inactive for microtubule-based motility (Cai et al., 2007). Several examples show that the interaction of KLCs with cargo participates in the activation of the molecular motor (Blasius et al., 2007; Cai et al., 2007; Sanger et al., 2017). From this point of view, the results obtained with PipB2 are not surprising. However, they are not consistent with observations made in the infectious context. Previous studies have shown that by recruiting kinesin-1 to the SCV, PipB2 induces elongation of tubules emerging from bacterial vacuoles. However, in the absence of SifA, these tubules/vesicles do not form and a PipB2-dependent accumulation of kinesin-1 is observed on SCVs. Based on these observations, we proposed (Boucrot et al., 2005) that: (1) the kinesin-1 present on ΔsifA SCVs is weakly or not active, which prevents the formation of tubules/vesicles from SCVs; (2) the SifA/SKIP complex is necessary for the activation of the kinesin-1 recruited by PipB2. This model is supported by a recent study (Sanger et al., 2017) demonstrating that multiple interactions of the N-terminal part of SKIP with KLC and KHC promote activation of kinesin-1. The data presented here also support this model as they show that the absence of functional kinesin-1 in macrophages is associated with the presence of SseJ on SCVs, a phenotype seen in the absence of SifA. This strongly suggests that the kinesin-1 recruited by PipB2 on ΔsifA SCVs is essentially inactive.

Our results thus show that the control of the activity of the kinesin-1 present on SCV is more complex than initially thought and suggest the involvement of other Salmonella effectors and/or host proteins. Among the effectors (other than SifA and PipB2) known to influence vesicle/tubule formation, two have activities compatible with possible involvement in the regulation of kinesin-1. Although not essential for the formation of Salmonella-induced tubules, SseJ promotes the tubulation of the lysosomal compartment when expressed ectopically with SifA (Ohlson et al., 2008). This event depends on SKIP and microtubules, which suggests that SseJ exacerbates the ability of SifA to induce tubulation (Boucrot et al., 2003) and reveals a functional interaction that is important for the formation of tubules. SopD2, when inactivated together with SifA, results in a strain (ΔsifAΔsopD2) that produces tubules. This phenotype is accompanied by a decrease in kinesin-1 associated with SCVs and the appearance at the cellular periphery of vesicles that are positive for PipB2 and kinesin-1 (Schroeder et al., 2010). This observation can be interpreted as a sign that SopD2 exerts an inhibitory activity on kinesin-1 linked to PipB2. It will be necessary to analyse the impact of these effectors and possibly others also present on SCVs to understand the regulation of kinesin-1 activity in the context of infection.

Kinesin-1 is an abundant molecular motor that is involved in a large number of transport processes in different cell types and cell sites (for reviews, see Dinu et al., 2007; Morfini et al., 2016; Verhey et al., 2001). Regulatory mechanisms for binding/detaching various cargoes are still poorly understood and the protein patterns recognized by KLCs for binding cargoes have only recently begun to be identified. A motif was originally found in the neural protein calsyntenin-1 (CSTN1, also known as CLSTN1) (Konecna et al., 2006). It consists of a peptide sequence containing a tryptophan residue (W) flanked by negatively charged residues (D or E). This W-acidic motif has been identified in pathogen or host proteins. For example, it is duplicated in vaccinia virus protein A36 (Dodding et al., 2011), which is necessary for recruitment of kinesin-1 and transport of viral particles to the periphery of infected cells (Rietdorf et al., 2001). The same pattern is found in two copies in SKIP and is located in the N-terminal part of the protein while the interaction domain with SifA is in the C-terminal half (Boucrot et al., 2005; Pernigo et al., 2013). As no tryptophan residues are used in the composition of PipB2, the motif that allows this effector to interact with KLC must be different.

Several functional domains of PipB2 have been described. The N-terminal part (residues 1 to 225, out of a total of 350) is sufficient for translocation and localization on SCVs and tubules (Knodler et al., 2003), but not enough to cause the accumulation of LAMP1-positive vesicles at the cellular periphery, a phenotype observed upon overexpression of PipB2. This phenotype requires a pentapeptide, located in the last ten residues (L₃₄₁FNEF₃₄₅) (Knodler and Steele-Mortimer, 2005). It should be noted that this pentapeptide partially overlaps a sequence (E₃₄₄FYSE₃₄₈) that has the characteristics of a Y-acidic motif. The latter has recently been identified in the JIP1 protein and is, like the W-acidic motif, recognized by the KLC (Pernigo et al., 2018). However, the involvement of this sequence in the interaction between kinesin-1 and PipB2 is uncertain as the deletion of the pentapeptide, which partially covers this hypothesized Y-acidic motif, has only a partial effect on the interaction between PipB2 and KLC (Henry et al., 2006). All this information leads us to believe that, as already proposed (Knodler and Steele-Mortimer, 2005), other determinants of PipB2 that have not yet been identified are involved in the interaction with kinesin-1 and the regulation of its activity.

The antibodies and reagents used in this study were: mouse anti-Myc (clone 9E10, homemade ascite, 1:1000), mouse anti-HA (clone 16B12, Covance, 1:1000), mouse anti-KLC (clone L1, Chemicon International, 1:50), mouse anti-β-tubulin (clone TUB 2.1, Sigma, 1:1000), mouse anti-histidine (clone HIS.H8; Thermo Fisher Scientific, 1:1000), rat anti-HA (clone 3F10; Roche Molecular Biochemicals, 1:500), goat anti-mouse or anti-rabbit IgG coupled to peroxidase (Sigma, 1:10,000), donkey anti-rat or anti-mouse IgG conjugated to Alexa Fluor 546 or Alexa Fluor 647 (Jackson ImmunoResearch, 1:500), HRP-conjugated streptavidin (Thermo Fisher Scientific, 1:10,000) and GFP-Trap (Chomotek). The rabbit anti-KHC (KIF5B, PCP42) was generously provided by R. Vale (University of California, San Francisco, CA, USA). Unless otherwise stated, all reagents were purchased from Sigma-Aldrich.

HeLa (ATCC CCL-2) and COS-7 cells (ECACC General Cell Collection, 87021302) were routinely grown in Dulbecco's modified Eagle's medium (Gibco-BRL) supplemented with fetal calf serum 10% (Gibco-BRL) and glutamine 2 mM.

Hoxb8-immortalized macrophage-committed progenitor cells were derived from either wild-type C57BL/6 mouse or C57BL/6 not expressing KIF5B in hematopoietic cells following the procedure described by Redecke et al. (Redecke et al., 2013). In brief, bone marrow cells from the femurs were pre-stimulated for 48 h in RPMI 1640-based medium with 250 ng/ml SCF, 10 ng/ml IL3 and 20 ng/ml IL6. Ficoll-purified mononuclear cells were subjected to spinoculation with 1 ml of ER-Hoxb8 retrovirus. Infected progenitors were cultured in Progenitor Outgrowth Medium (RPMI 1640 with 10% FCS, 50 µM 2-mercaptoethanol, 2 mM glutamine, 1 µM β-oestradiol and 2% GM-CSF-conditioned medium from B16 melanoma expressing the Csf2 cDNA). Immortalized myeloid progenitors were selected by transferring nonadherent progenitor cells every 3 days to a new well in a six-well culture plate over 3 weeks to produce immortalized macrophage progenitor lines. Differentiation to macrophages was performed for 7 days in the same base medium free of β-oestradiol and GM-CSF but containing 10% M-CSF-conditioned medium from L929 cells.

A wild-type S. enterica serovar Typhimurium 12023 strain expressing SseJ-HA from the chromosome was transformed with pFPV25.1 (Valdivia and Falkow, 1996) for GFP expression. The ΔsifA mutant (AAG003G) was obtained by P22 transduction of ΔsifA::kan. The oligonucleotides and plasmids used in this study are described in Tables S1 and S2, respectively

Hoxb8-immortalized cells were seeded in six-well plates with 12 mm diameter glass coverslips and differentiated for 7 days. Cells were Salmonella infected and processed for immunolabelling and fluorescence confocal microscopy as described previously (Zhao et al., 2016).

The pDEST17-Bio (V276) vector was obtained by modifying the pDEST17 Gateway destination vector by PCR to insert a biotin acceptor peptide (15-mer GLNDIFEAQKIEWHE) between the [His]6 tag and the attR1 site. Δ17PipB2 and SifAΔ6Bio were amplified by PCR from Salmonella 12023 wild-type genomic DNA using the primer pairs O786/O312 and O299/O806, respectively. The primer SifAΔ6BioRev was designed to add the biotin acceptor peptide sequence at the C-terminal. These PCR products were then inserted into pDONRZeo, resulting in the pDONR-Δ-17PipB2 (C1141) and pDONR-SifAΔ6-Bio (C1176). By recombination of pDONR-Δ17PipB2 with pDEST17-Bio and pDON-SifAΔ6-Bio with pDEST17, the final plasmids p[His]6-Bio-Δ17PipB2 (C1143) and p[His]6-SifAΔ-6Bio (C1177) were obtained. pWC2 (K401-Bio-H6) was a gift from Jeff Gelles (Addgene plasmid 15960) (Subramanian and Gelles, 2007).

Rosetta(DE3)pLysS competent cells (Novagen) containing pDEST17-SifAΔ6Bio or pWC2 (for the production of SifA-Bio or K401-Bio, respectively) were grown overnight at 25°C in MagicMedia E. coli expression medium (Thermo Fisher) supplemented with ampicillin 100 µg/ml and D-biotin 2.5 µg/ml. Rosetta containing pDEST17Bio-Δ17PipB2 (for the production of Bio-PipB2) was grown in 2× YT medium (Fisher Scientific) also supplemented with ampicillin and D-biotin. At OD600nm=0.8, overnight protein expression was induced by adding 0.5 mM IPTG at 25°C. Proteins were purified with Hi-Trap Chelating-Ni (GE Healthcare Life Sciences) on FPLC (ÄKTA Start, GE Healthcare Life Sciences) using a standard procedure. Finally, Bio-PipB2 was dialyzed against 50 mM imidazole (pH 6.7), 50 mM KCl, 4 mM MgCl₂, 2 mM EGTA, 10 mM β-mercaptoethanol and 20% glycerol. K401-Bio was dialyzed against the same buffer with 50 nM ATP. SifA-Bio was dialyzed as previously described (Diacovich et al., 2009). Proteins were snap-frozen in liquid N₂ and were stored at −80°C.

Mouse brains were washed several times with PBS, homogenized in 1 ml/mg of lysis buffer [50 mM imidazole (pH 6.7), 1 mM MgCl₂, 2 mM EGTA, 1 µM DTT and 250 mM sucrose] and centrifuged at 10,000 g, 4°C, for 20 min. Proteinase inhibitors (Thermo Scientific) and 10 µM paclitaxel were added to the supernatant that was incubated for 30 min at 37°C. Finally, the sample was centrifuged at 107,000 g for 30 min at 25°C (Optima MAX-XP; Beckman Coulter) and the supernatant was kept at −80°C.

Mouse brains were homogenized in 1 ml/g PME [100 mM PIPES (pH 6.9), 2 mM EGTA, 1 mM MgSO₄, 1 mM DTT and 1 mM GTP]. The homogenate was centrifuged at 100,000 g for 30 min at 4°C and the supernatant incubated for 30 min at 37°C with 2 units/ml apyrase and 20 µM paclitaxel, and for 30 extra minutes with 0.5 mM adenylyl-imidodiphosphate (AMP-PNP). After centrifugation at 39,000 g for 30 min at 30°C, the pellet was resuspended in PME containing 0.1 mM AMP-PNP and 20 µM paclitaxel and centrifuged again. The pellet was resuspended in a small volume of PME with 20 µM paclitaxel and 10 mM Mg-ATP, incubated for 30 min at 37°C and centrifuged as in the previous steps. The supernatant was loaded on a linear sucrose gradient (5-20% in PME with 1 mM Mg-ATP), centrifuged at 31,500 rpm in a Beckman SW41 rotor for 16 h at 4°C and finally fractionated in aliquots of 0.5 ml that were stored at −80°C.

Polymerized microtubules were prepared by mixing Rhodamine-tubulin (Cytoskeleton) with purified porcine tubulin (Cytoskeleton) at a ratio of 1:100 in BRB80 buffer [80 mM PIPES (pH 6.7), 1 mM MgCl₂ and 1 mM EGTA] with 10% glycerol and 1 mM GTP and incubated at 37°C for 20 min, followed by addition of 10 µM paclitaxel and incubation for a further 20 minutes. Microtubules were kept room temperature and protected from light until use. The gliding assay was performed in a 10 µl flow chamber built with a glass slide, coverslip and parafilm spacers in between, joined by heating at 150°C for a few seconds. It was pre-coated with streptavidin (1 mg/ml) for 5 min and washed three times with wash buffer [BRB80 buffer (pH 6.7), 0.15 mg/ml casein, 5 mM DTT and 10 µM paclitaxel]. Protein buffer [10 µl of 20 mM Tris-HCl (pH 7.6), 300 mM NaCl, 10 mM β-mercaptoethanol, 2 mM EGTA and 20% glycerol] containing 20 µM K401-Bio, 5 µM Bio-PipB2 or 5 µM SifA-Bio were injected, incubated for 5 min and washed three times with the wash buffer. 10 µl of a kinesin-containing fraction was injected into the chambers containing Bio-PipB2 or SifA-Bio, incubated for 5 min and washed three times with the wash buffer. 10 µl rhodamine-microtubules previously diluted 1:100 (0.01 mg/ml) in wash buffer were injected, incubated for 5 min and washed three times. Immediately before recording the microtubule movement, motility buffer [BRB80 buffer (pH 6.7), 0.15 mg/ml casein, 5 mM DTT, 10 µM paclitaxel, 0.18 mg/ml catalase, 0.37 mg/ml glucose oxidase, 10 mM glucose and 1 mM ATP] was introduced into the flow chamber and the chamber was sealed. Microtubules were observed using an epifluorescence microscope (Axio-observer; Zeiss) and time-lapse images (1 images/30 s) were captured for 2 to 30 min. Microtubule tracking was performed using FIJI (MTrackJ plugging, 100 microtubules per condition) and velocity of movement or quantification of bound microtubules was determined.

Giant unilamellar vesicles (GUVs) were prepared by electroformation (Angelova et al., 1992) with a lipid composition of 1,2-Dioleoyl-sn-Glycero-3-Phosphocholine 1 mM (DOPC; Avanti Polar Lipids) supplemented with 1% of BodipyFL-C5-HPC (Fisher Scientific) and 0.1% of the biotinylated lipid 1,2-Dioleoyl-sn-Glycero-3-Phosphoethanolamine-N-(Cap Biotinyl) (Biot-Cap-DOPE; Avanti Polar Lipids). 10 µl microtubules (1 mg/ml) in IMI buffer [50 mM imidazole (pH 6.7), 50 mM NaCl, 2 mM EGTA, 1 mM MgCl₂ and 10 µM paclitaxel] were injected into the chamber and incubated for 10 min before injecting 10 µl casein (7 mg/ml in IMI buffer). 10 µl of protein buffer with K401-Bio (20 µM) and streptavidin (7 µM) or 10 µl of kinesin-containing fraction with Bio-PipB2 (17 µM) or Bio-BSA (20 µM) and streptavidin (7 µM) were injected and incubated for 15 min. Finally, 10 µl of motility buffer [IMI buffer (pH 6.7), 5 mM DTT, 10 µM paclitaxel, 1 mM ATP, 0.18 mg/ml catalase, 0.37 mg/ml glucose oxidase and 10 mM glucose] and GUVs (1:10 v/v) were sequentially injected. The chamber was sealed and kept at 45° from the horizontal for few minutes to enable GUVs to settle before microscopic observation.

10 µl MAP-depleted mouse brain cytosol, 1 µM Bio-PipB2 or 1 µM SifA-Bio, 10 µM paclitaxel and 0.5 mM AMP-PNP in BRB80 buffer (pH 7.2) (final volume of 50 µl) were incubated for 30 min at room temperature in the presence or absence of 5 µl microtubules (1 mg/ml). K401-Bio was used as positive control. Reaction mixes were then ultra-centrifuged through a glycerol cushion (BRB80 buffer, 60% glycerol and 10 µM paclitaxel) for 40 min at 100,000 g at 25°C. Supernatant and pellet were kept and mixed/resuspended with Laemmli sample buffer, and 20 µl of each sample were analysed by SDS-PAGE and western blotting.

HeLa and COS-7 cells were transfected using FuGENE 6 following the manufacturer's protocol.

Transfected HeLa cells were washed with PBS, scraped with a rubber policeman, centrifuged for 5 min at 400 g and resuspended in 400 μl lysis buffer [10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.5 mM EDTA and 0.5% NP-40] supplemented with a protease inhibitor cocktail. After 30 min at 4°C, cell lysates were centrifuged at 20.000 g for 10 min at 4°C. 10% of the lysates were boiled in Laemmli sample buffer. Immunoprecipitations were performed with GFP-Trap beads that contain a recombinant alpaca anti-GFP antibody covalently coupled to agarose beads following manufacturer's protocol. Immunoprecipitated proteins were analysed by SDS-PAGE and western blotting using an appropriate antibody.

Bio-PipB2 or SifA-Bio, or a combination of both proteins in a ratio 1:1, were incubated with streptavidin-agarose beads for 30 min at room temperature. As negative controls, the same protocol without any protein was performed. The beads were pelleted at 2000 g for 2 min, incubated with 10 µM D-biotin and washed three times with BRB80 buffer. Then they were incubated with MAP-depleted mouse brain cytosol for 30 min at room temperature and washed three times as previously described. Finally, beads were resuspended in Laemmli sample buffer and western blotting was performed to analyse the bound proteins.

Statistical analyses were performed using Prism 6 software (GraphPad).

Authors acknowledge Mark P Dodding (University of Bristol, UK) for providing plasmids for expression of HA-tagged light (mouse KLC2) and heavy (rat KIF5C) chains of kinesin-1, Hans Häcker (St Jude Children's Research Hospital, Memphis, TN, USA) for the plasmid MSCV-ERHBD-Hoxb8, Gaël Ménasché (Imagine Institute, Paris, France) for providing bone marrow extracted from C57BL/6 mice not expressing KIF5B in hematopoietic cells, Pauline Brige (CERIMED, Marseille, France) for her help in collecting organs for the extraction of molecular motors, and J. Ruiz-Albert and D. W. Holden (Imperial College of Science, London, UK) for the 12023 strain expressing SseJ-HA from the chromosome. We thank the imaging core facility (ImagImm) of the Centre d'Immunologie de Marseille-Luminy (CIML).