We observed that SipA remains associated with internalized bacteria (Figure S1A in the Supplemental Data available with this article online), while other SPI1 effectors are degraded (data not shown; Kubori and Galan, 2003). However, as we initially employed a wild-type S. typhimurium strain expressing enhanced levels of SipA from a plasmid (sipA⁺⁺; Cain et al., 2004), we next examined the localization of SipA^FLAG expressed from the endogenous chromosomal context. SipA^FLAG similarly surrounded ∼60% internalized bacteria 2 hr postinfection and after 8 hr was associated with a similar proportion of microcolonies (Figure 1A). To establish which T3SS delivers this formerly unrecognized SipA^FLAG pool, we equivalently engineered S. typhimurium invG⁻ and ssaV⁻ mutants, which respectively lack essential components of the SPI1 and SPI2 T3SSs, rendering them inactive (Crago and Koronakis, 1998; Hensel et al., 1997). While SipA^FLAG was never detected after infection with S. typhimurium invG⁻, localization using the ssaV⁻ mutant mirrored the wild-type (Figure 1A). SipA^FLAG staining was unchanged following bafilomycin treatment that disrupts the SPI2 T3SS (Steele-Mortimer et al., 2000), and in SPI2 effector mutants (Figure S1B). These findings demonstrate that SipA persistence requires the SPI1 T3SS post-Salmonella entry.

To ascertain whether SCV-associated SipA is exposed in the host cytosol, we exploited the finding that delivered SipA can be visualized during entry using its export chaperone (InvB) fused to GFP as a cytosolic reporter (Schlumberger et al., 2005). GFP-InvB transfectants were infected with wild-type S. typhimurium or an isogenic sipA⁻ mutant. GFP-InvB distributed throughout the cytosol of control cells or those infected with the sipA⁻ mutant (Figure 1B). By contrast, GFP-InvB was recruited to SCVs post-wild-type infection (Figure 1B and Figure S1C), demonstrating that SipA is exposed on the SCV and potentially poised to engage host or bacterial targets.

We next investigated whether SipA influences intracellular multiplication. Replication of the sipA⁻ mutant was significantly attenuated, whereas that of the sipA⁺⁺ strain was reciprocally enhanced in epithelial-like cells and fibroblasts (Figure 2A). This dose-dependent response shows that SipA is central to Salmonella multiplication in nonphagocytic cells. As SPI2 effectors also govern replication in macrophages, we analyzed any comparable role for SipA using cultured and primary macrophages. Unexpectedly, the sipA⁻ mutant was again significantly attenuated, while increasing SipA levels bolstered bacterial replication (Figure 2A). These surprising findings illustrate that SipA also influences replication in macrophages.

Attenuated sipA⁻ mutants adopt a dispersed intracellular distribution biased toward the cell periphery (Figure 2B), reminiscent of bacteria lacking SPI2 effectors that direct SCV-organelle tethering or impede host motor protein activity (Boucrot et al., 2005; Henry et al., 2006; Salcedo and Holden, 2003). To evaluate “scattering,” every bacterium was categorized as nuclear proximal or distal. The proximal zone typically encompasses the GA (Figure S2A). Concomitant with the onset of the replicative defect, wild-type and sipA⁺⁺ strains appeared predominantly perinuclear, whereas ssaV⁻ and sipA⁻ mutants were dispersed (Figure 2B). To assess whether scattering reflected increased SCV-GA uncoupling (Abrahams et al., 2006; Salcedo and Holden, 2003), infected cells were treated with brefeldin A (BFA) that induces cis-Golgi redistribution into the endoplasmic reticulum (Chardin and McCormick, 1999; Lippincott-Schwartz et al., 1990). While BFA reduced wild-type and sipA⁺⁺ replication, the sipA⁻ mutant was not additionally attenuated, and BFA induced scattering of both wild-type and sipA⁺⁺ strains, whereas positioning of the sipA⁻ mutant was unchanged (Figure S2B). Indeed, the sipA⁻ mutant is seldom coincident with the cis-Golgi (Figure S2A). Taken together, these findings implicate SipA as a positioning determinant upstream of GA association.

Scattering and replicative attenuation of the sipA⁻ mutant might reflect bacterial release into the host cytosol. To assess SCV integrity, LAMP1 distribution was examined in epithelial cells and fibroblasts infected with wild-type S. typhimurium, sipA⁻, and sipA⁺⁺ strains. Bacteria lacking the SPI2 effector SifA that are released into the host cytosol at late time points were also examined (Beuzon et al., 2000). In each infected cell, multiple wild-type bacteria were observed within continuous perinuclear SCVs, from which Sifs extended (Figure 2C and Figure S3; Movie S1). SCVs formed by the sipA⁻ mutant remained intact but always unusually encapsulated only individual or sometimes pairs of bacteria (Figure 2C and Figure S3; Movie S2). Furthermore, Sifs rarely extended from these SCVs, although occasional “stunted” protrusions were evident (<5% SCVs). Even more unexpectedly, the sipA⁺⁺ strain formed continuous perinuclear SCVs that lacked Sifs, in which the membrane tightly apposed encapsulated bacteria. These tight SCVs seemed inherently unstable, as bacteria frequently became cytosolic (in ∼50% infected cells) (Figure 2C). By comparison, 30%–40% sifA⁻ mutants were cytoplasmic at this time point (Figure 2C; (Boucrot et al., 2005). These findings demonstrate that SipA is not essential for SCV integrity, but that relative SipA concentration influences SCV morphology and positioning.

Salmonella replication and SCV integrity require balanced activity of the microtubule motors dynein and kinesin, which respectively transport cargo toward the nucleus and cell periphery (Boucrot et al., 2005; Guignot et al., 2004; Harrison et al., 2003; Marsman et al., 2004). As SipA manipulates SCV positioning, we examined dynein and kinesin distribution in infected epithelial cells and fibroblasts. After infection with wild-type S. typhimurium, both dynein and kinesin accumulated diffusely around the SCV periphery but were only occasionally coincident with bacteria (Figure 3A; Boucrot et al., 2005). However, kinesin and tubulin frequently colocalized with peripheral sipA⁻ SCVs, whereas dynein remained infrequently associated (Figure 3B).

When dynein function was disrupted with p50/dynamitin, which induces endosome redistribution to the periphery by uncoupling dynactin (data not shown; Burkhardt et al., 1997), an equivalent subtle increase (∼10%) in wild-type and sipA⁻ at the cell periphery was observed, together with mild replicative attenuation (Figure 3C and Figure S4). These data verify that dynein contributes to perinuclear SCV positioning (Guignot et al., 2004) and demonstrate that this is SipA independent. Intriguingly, both positioning and replication of the sipA⁺⁺ strain are resistant to p50/dynamitin expression (Figure 3C and Figure S4). Kinesin activity was inhibited with aurintricarboxylic acid (ATA), which impeded replication of wild-type and the sipA⁺⁺ strain (Figure S4; Guignot et al., 2004), and although both remained perinuclear, peripheral migration of the sipA⁻ mutant was blocked (Figure 3C). However, despite positional rescue, sipA⁻ replication was not restored (Figure S4). Taken together, these data demonstrate that scattering of sipA⁻ SCVs is associated with aberrant kinesin recruitment.

Cytoskeletal dynamics underpin endogenous vacuole biogenesis and trafficking, and both F-actin and microtubules accumulate around the SCV (Guignot et al., 2004; Meresse et al., 2001). Two SPI2 effectors influence SCV-kinesin interaction, possibly antagonistically; SifA negatively regulates kinesin activity, whereas PipB2 triggers kinesin recruitment (Boucrot et al., 2005; Henry et al., 2006). To gain further insight into the role of actin-binding SipA in SCV positioning, we examined the localization of SifA^HA, PipB2^HA, and F-actin after infection with wild-type, sipA⁻, and sipA⁺⁺ strains. In fibroblasts and epithelial cells, nuclear-proximal SCVs containing wild-type bacteria were enriched with SifA^HA, which additionally decorated Sifs as expected (Figure 4A), whereas SifA^HA present on SCVs containing sipA⁻ mutants was markedly reduced, and consequently Sifs seldom formed. SifA^HA nevertheless localized to compartments distinct from the SCV (Figure 4A). By contrast, SifA^HA was present but apparently dormant on “tight” Sif-devoid SCVs formed by the sipA⁺⁺ strain (Figure 4A). PipB2^HA localized to SCVs, peripheral vesicles, and extended tubular structures after wild-type infection and remained localized with SCVs containing the sipA⁻ mutant, consistent with its role as a kinesin linker (Figure 4B). Strikingly, as with SifA^HA, PipB2^HA localized only to SCVs formed by the sipA⁺⁺ strain and was unable to disseminate within the infected cell (Figure 4B). These findings demonstrate that SipA imbalance induces mislocalization of SPI2 effectors SifA and PipB2, leading indirectly to aberrant positioning and morphological defects.

Given that SipA binds F-actin (Zhou et al., 1999), we next investigated any link with F-actin-SCV association, and how this might potentially impact on SifA and PipB2 localization. By 6 hr postinfection, F-actin “nests” surrounded the SCVs of wild-type bacteria, which were more evident in fibroblasts than epithelial-like cells (Figure 4C; Meresse et al., 2001). F-actin staining appeared more indistinct following equivalent infection of both cell lines with the sipA⁻ mutant, suggesting that SipA stabilizes phagosomal F-actin (Figure 4C), akin to its role during cell entry (McGhie et al., 2004). However, F-actin was robustly enriched on the “tight” SCVs formed following infection of epithelial cells with the sipA⁺⁺ strain (Figure 4C). Furthermore, F-actin colocalized with both SifA^HA and PipB2^HA (Figure 4D). Excessive SCV-F-actin accumulation therefore impedes SPI2 effector activity. Nevertheless, as with SipA staining during entry (Schlumberger et al., 2005), SCV-SipA^FLAG association occurred when infected cells were treated with cytochalasin D (CD) or latrunculin B (LB), which prevent actin assembly and inhibit bacterial replication (Figure 4D; Meresse et al., 2001), and was unaltered even after actin assembly was reinitiated by LB washout (Figure 4D). Thus, SipA-induced F-actin stability modulates localization of key SPI2 effectors, but SipA targeting to the SCV is actin independent.

SipA lacks lipid affinity in vitro (Hayward and Koronakis, 1999). To delineate the region(s) of SipA involved in vacuole targeting, we transfected cells with SipA, the C-terminal actin-binding fragment (SipA-C) and the remainder that encodes no recognized activity (SipA-N) as C- or N-terminal fusions to YFP or CFP, respectively. SipA and SipA-C colocalized with F-actin, although SipA-C was more peripheral and SipA distributed throughout the cell body (Figure 5A). SipA-N fusions were never coincident with F-actin but instead exhibited punctate perinuclear distribution (Figure 5A). Fractionation revealed that, while SipA partitioned in the internal membrane/cytoskeleton fraction (Cain et al., 2004), actin-binding SipA-C was located exclusively in the plasma membrane fraction, which additionally contains ∼10% cellular actin (Cain et al., 2004), and SipA-N was predominantly in the internal membrane fraction, with ∼15% detected in the cytosolic fraction (Figure 5A).

As SipA localizes to LAMP1-rich SCVs independently of F-actin during infection, we visualized LE/Lys distribution in cells expressing SipA, SipA-N, and SipA-C. Remarkably, although neither SipA nor SipA-N trigger obvious rearrangement of the actin or microtubule networks, both induced dramatic centripetal aggregation of LAMP1-positive LE/Lys compartments toward the microtubule-organizing center (MTOC; Figure 5B). Additionally, these compartments colocalized with SipA-N (Figure 5B), demonstrating that SipA-N autonomously targets LE/Lys and induces their relocalization. Significantly, despite SipA localizing to the actin cytoskeleton, it also retained the capacity to induce similar LE/Lys redistribution, whereas actin-binding SipA-C exhibited no comparable activity.

Given this previously uncharacterized ability of SipA to redistribute LE/Lys toward the nucleus, we investigated whether SipA, SipA-N, or SipA-C could complement the sipA⁻ mutant in infected cells. Expression of SipA or either derivative had no significant effect on intracellular positioning of wild-type S. typhimurium (Figure 6), although SipA expression attenuated intracellular replication (Figure 6B). However, expressed SipA and SipA-N both restored perinuclear positioning to the sipA⁻ mutant (Figure 6), with ∼75% of bacteria lacking SipA shifting into the perinuclear zone in SipA-N transfectants (Figure 6B). Yet despite these effects on positioning, expression in trans failed to restore replicative proficiency or Sif formation (Figure 6B; data not shown).

As SCV-localized SipA stabilizes SifA that in turn inhibits kinesin activity (Figure 4), we examined the effect of in trans SipA expression on intracellular replication and positioning of sifA⁻, sifA⁻sipA⁻, and control sseI⁻ mutants. Unlike with the sipA⁻ mutant that delivers SifA, expression of SipA, SipA-N, or SipA-C did not restore positioning or replication of the sifA⁻ mutant (Figure 6; Figure S5). On the contrary, SipA-C expression increased peripheral positioning of the sifA⁻ mutant. Identical results were obtained postinfection of transfectants with the double mutant lacking SipA and SifA (data not shown). These data illustrate that SipA and SifA must cooperate to ensure perinuclear SCV positioning during infection.

By genetically manipulating Salmonella, we have demonstrated that SipA concentration profoundly influences intracellular replicative proficiency, SCV positioning, and SPI2 effector localization. We therefore predicted that SipA concentration must be tightly controlled during Salmonella infection. To assess this, we initially investigated the effect of inhibiting bacterial protein synthesis after entry using chloramphenicol. Only ∼15% internalized bacteria associated with SipA^FLAG 1 hr after antibiotic treatment, with residual immunostaining appearing fragmented and distal from the bacteria, which as expected also failed to replicate. No signal was detectable 6 hr postinfection (Figure 7A). These data show that bacterial protein synthesis is required for SipA persistence and illustrate that SipA is apparently actively degraded. Some delivered SPI1 effectors are differentially targeted by the cellular proteasome (Kubori and Galan, 2003). To confirm whether intracellular SipA is similarly susceptible to host-mediated degradation, infected cells were treated with a proteasome inhibitor (MG132). Although MG132 promoted cytosolic Salmonella replication (data not shown; Perrin et al., 2004), SipA^FLAG accumulation increased significantly (Figure 7A). This was not a function of increased bacterial load, as a pool of delivered SipA^FLAG could be captured following treatment with chloramphenicol and MG132 (Figure 7A). Intriguingly, inhibiting proteasome activity induced increased F-actin accumulation around the SCV and vacuole instability (Figure 7A). This phenotype mirrors that following infection with the sipA⁺⁺ strain, which suggests that aberrant F-actin accumulation triggers SCV instability.

Bacterial strains and plasmid construction are described fully in the Supplemental Data. Bacteria were maintained on Luria-Bertani agar or cultured in tryptone-yeast (TY) medium supplemented with 10 μgml⁻¹ tetracycline, 50 μgml⁻¹ kanamycin, 8 μgml⁻¹ chloramphenicol, or 50 μgml⁻¹ ampicillin.

Mammalian cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (v/v) fetal calf serum (FCS), L-glutamine, and antibiotics (Sigma). Bone marrow-derived macrophages were cultured from mouse bone marrow in medium containing 20 ngml⁻¹ recombinant macrophage colony-stimulating factor (see the Supplemental Data). Cells were incubated at 37°C, 5% CO₂.

Drug stock solutions in dimethyl sulphoxide or ethanol (Cm) were diluted at least 1:1000 in DMEM to working concentrations: 100 μgml⁻¹ Cm, 5 μgml⁻¹ BFA, 1 μM bafilomycin A1, 10 μM ATA and MG132, 1 μgml⁻¹ CD and LB. Cells were pretreated with ATA for 3 hr, MG132 and BafA were added 30 min prior to infection, Cm and BFA were added 1 hr postinfection, and CD and LB were added 3 hr postinvasion.

Infection of NIH3T3 and HeLa cells with S. typhimurium was performed as described (Garner et al., 2002). Macrophages were infected with opsonized bacteria as described (Beuzon et al., 2000).

NIH3T3 cells were transfected using Lipofectamine, according to the manufacturer's instructions (Invitrogen).

Samples were paraformaldehyde fixed, permeabilized in 0.2% Triton X-100, incubated with appropriate primary and secondary antibodies, and analyzed using a fluorescence microscope (Leica DM IRBE), as fully detailed in the Supplemental Data.

Fractionation was performed as described (Cain et al., 2004).