Biochemical Journal

Research article

The secreted Salmonella dublin phosphoinositide phosphatase, SopB, localizes to PtdIns(3)P-containing endosomes and perturbs normal endosome to lysosome trafficking

Joseph D. Dukes, Huailo Lee, Rachel Hagen, Barbara J. Reaves, Abigail N. Layton, Edouard E. Galyov, Paul Whitley


Invasion and survival in mammalian cells by Salmonella enterica is mediated by bacterial proteins that are delivered to the host cell cytoplasm by type III secretion systems. One of these proteins, SopB/SigD, is a phosphoinositide phosphatase that can hydrolyse a number of substrates in vitro including PtdIns(3,5)P2. These substrates are, however, likely to be restricted in vivo by the localization of SopB, as different phosphoinositides have distinct spatial distributions in mammalian cells. In the present study, we show that heterologously expressed SopB localizes almost exclusively to endosomes containing the lipid PtdIns(3)P, and on which ESCRT (endosomal sorting complexes required for transport) proteins assemble. Furthermore, we present evidence that SopB can inhibit trafficking of activated epidermal growth factor receptor to the lysosome. These results provide further evidence that PtdIns(3,5)P2, a lipid involved in endosomal maturation, may be a relevant in vivo substrate of SopB. We hypothesize that reduction of PtdIns(3,5)P2 levels in cells by the action of SopB may perturb the function of a subset of ESCRT proteins that have previously been shown to bind to this lipid.

  • endosome
  • endosomal sorting complexes required for transport (ESCRT)
  • lysosome trafficking
  • PtdIns(3,5)P2
  • Salmonella dublin
  • SopB


Salmonella enterica serovars are facultative intracellular bacteria that can target a variety of eukaryotic hosts and are the causative agents of diseases such as food-borne gastroenteritis in humans [1]. These bacteria have developed sophisticated strategies to regulate host cell responses, which allow them to be internalized and survive/replicate in membrane-bound compartments, known as SCVs (Salmonella-containing vacuoles) [2]. Upon bacterial entry into host cells, SCVs undergo a remodelling that resembles that of the host cell's endocytic pathway, but progression down the endocytic pathway is arrested prior to fusion with degradative lysosomes [3].

The formation and remodelling of SCVs is dependent on the secretion, during and immediately after internalization, of bacterial effector proteins into the host cell cytoplasm via TTSSs (type III secretion systems) encoded within pathogenicity islands 1 and 2 [SPI-1 (Salmonella pathogenicity island-1) and SPI-2] [2,4]. Among the SPI-1 effectors is SopE, which is a RhoGEF (where GEF is guanine nucleotide-exchange factor) that initiates actin rearrangement in the host cell, leading to plasma membrane ruffling and nascent SCV formation [5,6]. Another protein secreted by the SPI-1 TTSS is SopB, which is a phosphoinositide phosphatase [710]. Recent evidence suggests that SopB may contribute to the arrest of progression of the SCV down the endosomal trafficking pathway leading to fusion with lysosomes through its PtdIns(3,5)P2 5-phosphatase activity [11]. This activity is responsible for the conversion of PtdIns(3,5)P2 to PtdIns(3)P by removing a phosphate from the 5-position of the inositol lipid headgroup. It should be noted that SopB has other phosphoinositide and inositol phosphate substrates in addition to PtdIns(3,5)P2 in vitro, and this protein has been implicated in a variety of cellular responses such as Akt activation and modulation of chloride secretory responses [9,12,13]. Furthermore, deletion of SopB abrogates diarrhoea and inflammatory events associated with Salmonella infection [7]. Thus the role of SopB in salmonellosis is complex and unlikely to be due to a single activity.

The phosphoinositide PtdIns(3,5)P2 is produced from PtdIns(3)P by homologous PtdIns(3)P 5-kinases called Fab1p and PIKfyve (PhosphoInositide Kinase for five position containing a Fyve finger domain) in Saccharomyces cerevisiae and mammals respectively. PtdIns(3,5)P2 has been implicated in multiple membrane-trafficking events. In Sacch. cerevisiae, PtdIns(3,5)P2 is involved in retrograde trafficking from vacuoles to late endosomes and ubiquitin-dependent sorting of endocytosed receptors into MVBs (multivesicular bodies) prior to their degradation [1417]. In mammalian cells in culture, the role of PtdIns(3,5)P2 in specific membrane-trafficking events has not been as extensively characterized, although it seems to be required for both endomembrane homoeostasis [18] and insulin-stimulated glucose transporter 4 (GLUT4) isoform trafficking [19]. There is conflicting evidence as to whether PtdIns(3,5)P2 is required for efficient lysosomal sorting and degradation of ubiquitylated membrane proteins in mammalian cells [34,40].

Delivery of endosomal cargoes to lysosomes in mammals or vacuoles in yeast is dependent on the sequential assembly of at least three evolutionarily conserved heteromeric protein complexes known as ESCRT-I (endosomal sorting complexes required for transport-I), ESCRT-II and ESCRT-III in addition to phosphoinositides and other lipids [20]. Previous work in our laboratory has identified a protein component of the mammalian ESCRT machinery as being a PtdIns(3,5)P2-binding protein and thus a strong candidate to be an effector of this lipid [21]. The protein, mVps24p (also known as CHMP3p; where Vps is vacuolar protein sorting and CHMP is charged MVB protein), is homologous with Sacch. cerevisiae Vps24p, a component of the ESCRT-III complex. This complex is made up of multiple proteins belonging to the CHMP family, including others that may also bind phosphoinositides [22]. The exact role of each of the ESCRT complexes, especially ESCRT-III in the endosomal maturation process, remains to be established. Multiple rounds of membrane association/dissociation and assembly/disassembly of ESCRT complexes are however important for efficient endosome to lysosome trafficking. The membrane dissociation and disassembly of ESCRT complexes are catalysed in an ATP-dependent manner by an AAA-type ATPase called Vps4 [23,24]. Expression of a dominant-negative, ATPase-defective Vps4 in mammalian cells results in a swollen endosomal phenotype and defective trafficking to lysosomes.

In the present study, we show that SopB localizes almost exclusively to endosomes containing PtdIns(3)P on which the ESCRT machinery can assemble and that SopB function perturbs normal endosome to lysosome trafficking. The results are consistent with a requirement for PtdIns(3,5)P2 in efficient endosome to lysosome trafficking and support the hypothesis that SopB uses PtdIns(3,5)P2 as a substrate in vivo to divert SCVs away from lysosomal fusion.


Bacterial strains

wt (wild-type) and ΔSopB Sal. dublin strains have been described previously [7].

DNA manipulations and constructs

Standard methods for recombinant DNA manipulations were used. Plasmid constructs for the expression of GFP (green fluorescent protein)–Vps4p E235Q and NTmVps24p–GFP have been described previously [21]. Plasmids for the expression of Myc-tagged SopB wt and SopB C460S in mammalian cells under the control of the CMV (cytomegalovirus) promoter were pRK5myc SopB and pRK5myc SopB C460S respectively. To construct these pRK5myc SopB, a DNA fragment encoding SopB was amplified by PCR using SopBBam: 5′-ctcggatccatgcaaatacagagcttctatcact-3′ and SopBXba: 5′-ctctctagatcaagatgtgattaatgaagaaat-3′ primers with Sal. dublin chromosomal DNA as a template. The DNA fragment was digested with BamHI and XbaI and ligated into the pRK5myc plasmid [25]. For the construction of pRK5myc SopB C460S, the only difference was that the template for PCR amplification was chromosomal DNA from Sal. dublin SBc/s mutant [8]. The GFP–iFYVE FENS-1 (FYVE domain containing protein localized to endosomes-1) construct [26] was a gift from Dr Michael Clague (University of Liverpool, Liverpool, U.K.). The GFP–PIKfyve construct was a gift from Dr Pete Cullen (University of Bristol). The GFP expression plasmid used to transform Sal. dublin strains for use in immunofluorescence microscopy was pSU2007 [27].


Mouse monoclonal anti-Myc antibodies (4A6) were purchased from Upstate and anti-FLAG tag antibodies were purchased from Sigma. Mouse monoclonal anti-EGFR [EGF (epidermal growth factor) receptor] antibodies were purchased from BD Transduction Laboratories. Sheep anti-TGN46 (where TGN is trans-Golgi network) antibodies were purchased from Serotec. The rabbit polyclonal anti-EEA1 (early endosome antigen 1) antibody was a gift from Dr Michael Clague. The rabbit anti-CI-M6PR antibody was a gift from Dr Paul Luzio (University of Cambridge, Cambridge, U.K.). Species-specific fluorophore (Alexa Fluor™ 488 and Alexa Fluor™ 568)-conjugated anti-IgG secondary antibodies were all purchased from Molecular Probes. Horseradish peroxidase-conjugated goat anti-mouse IgG and goat anti-rabbit IgG were purchased from Sigma.

Cell culture, transfection and immunofluorescence microscopy

COS-7 (African green monkey kidney), A431 (a human epidermoid carcinoma cell) and HeLa (human cervix carcinoma) cells were maintained at 37 °C, 5% CO2 (except A431 cells, which were maintained at 10% CO2) in DMEM (Dulbecco's modified minimal essential medium; Cambrex) containing 10% (v/v) fetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin and 100 μg/ml streptomycin. CHO (Chinese-hamster ovary)-K1 cells were maintained at 37 °C, 5% CO2 in Ham's F12 medium (Cambrex) containing 10% fetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin, and 100 μg/ml streptomycin. Cells were plated on to glass coverslips in 35 mm wells and grown until approx. 60% confluent prior to transfection using FuGENE 6 transfection reagent (Roche Diagnostics) according to the manufacturer's instructions. Cells grown on glass coverslips were transfected 16 h prior to fixation (unless stated otherwise) and processed for immunofluorescence microscopy as follows. Cells were rinsed with PBS and fixed with 4% (w/v) paraformaldehyde for 20 min and were permeabilized using methanol at −20 °C for 5 min. Primary antibodies were diluted in PBS 2% NCS [PBS containing 2% NCS (newborn calf serum)] and incubated with the cells for 2 h at 18 °C. Appropriate fluorophore-conjugated secondary antibodies (2 μg/ml) diluted in PBS 2% NCS were incubated with the cells for 1 h at 18 °C. Cells were washed five times for 5 min in PBS 2% NCS following all antibody incubations. Coverslips were mounted in Mowiol (Calbiochem, San Diego, CA, U.S.A.). Cells were examined and images were obtained on a Zeiss LSM510 laser scanning confocal microscope. For co-transfection of CHO-K1 cells with GFP–PIKfyve and SopB C460S, cells were grown on glass coverslips and transfected with GFP–PIKfyve 24 h prior to a second transfection of SopB C460S. Cells were fixed and processed for immunofluorescence microscopy 16 h following the second transfection.

Bacterial infection of cell cultures

A431 cells were seeded at approx. 1×105 cells/well into 35 mm tissue culture wells containing a glass coverslip and were maintained for 3 days at 37 °C in a 10% CO2 incubator. Salmonella strains were grown overnight in LB (Luria–Bertani) medium (supplemented with 30 μg/ml kanamycin when required) and subcultured (1:25 dilution) for approx. 3 h in a fresh medium. Bacteria were pelleted (10000 g, 2 min) and resuspended in PBS. A431 cells were infected with approx. 5×107 bacterial cells (multiplicity of infection of ∼100:1) for 1 h. Cells were washed with PBS prior to fixation and processed for immunofluorescence (see above).

EGFR degradation assay

A431 cells were infected with Salmonella as described above, with the following modifications. Cells were grown in 35 mm tissue culture wells without a coverslip. The A431 cells were washed twice with PBS 16 h prior to infection, and the medium was changed to a reduced serum medium lacking antibiotics (DMEM containing 1% BSA and 2 mM L-glutamine). Following infection for 1 h, the media were removed, cells were washed twice in PBS, and fresh DMEM, 1% BSA and 2 mM glutamine were added. Cells were immediately stimulated with 500 ng/ml EGF (Calbiochem) for different times (0, 8 and 16 min). Following EGF stimulation, cells were washed in ice-cold PBS and then harvested in 100 μl of RIPA buffer (50 mM Tris/HCl, pH 7.5, 150 mM NaCl, 0.5% sodium deoxycholate, 0.1% SDS and 0.1% Nonidet P40) containing protease inhibitor cocktail for mammalian cell extracts (Sigma). The protein concentrations of the cell lysates were determined using bicinchoninic acid reagent assay (Pierce). A 40 μg aliquot of each sample was separated by SDS/PAGE, transferred to nitrocellulose and immunoblotted for the presence of EGFR and β-tubulin. Signals were detected by ECL® (enhanced chemiluminescence) and quantified using an Optichem detector with associated software (Ultra Violet Products). For quantification, the amount of EGFR was normalized to the amount of β-tubulin (loading control) in the same sample.


SopB localizes to PtdIns(3)P containing membrane compartments

To investigate the subcellular location of SopB, COS-7 cells were transiently transfected with pRK5myc SopB or pRK5myc SopB C460S. These plasmids encode a Myc-tagged SopB wt protein (SopB-wt) and a catalytically inactive ‘phosphatase-dead’ Myc-tagged SopB protein (SopB C460S) respectively. Expression of SopB-wt was very toxic to cells, causing dramatic membrane vacuolation, altered cell morphology (Figures 1a and 1b) and cell detachment from coverslips. SopB-wt associated with the membranes of the swollen organelles in some cells (Figure 1a), presumably at a stage prior to the cells rounding up and detaching. In contrast, the catalytically inactive, SopB C460S showed a punctate localization throughout the cytoplasm in most of the transfected COS-7 cells (Figure 1c). The punctate distribution resembled that of early endosomes (Figure 1d). Due to the toxicity of SopB-wt, the remainder of the localization studies were performed with the SopB C460S protein, which did not result in rapid detachment of cells from coverslips. It should be noted that although the active SopB-wt protein, once membrane-associated, had a dramatic effect on the morphology of the membranes it associated with, SopB C460S did not.

Figure 1 Transient expression of SopB-wt and SopB C460S in COS-7 cells

COS-7 cells were transfected with (a, b) pRK5myc SopB or (c) pRK5myc SopB C460S. Cells were fixed after 16 h and immunostained with anti-Myc antibodies, followed by Alexa Fluor 568-conjugated anti-mouse IgG secondary antibodies. Untransfected cells (d) were immunostained with anti-EEA1 antibodies followed by Alexa Fluor 568-conjugated anti-rabbit IgG secondary antibodies.

To determine whether the observed SopB C460S-containing punctae were associated with specific cellular organelles, double immunostaining with markers for different organelles was performed (Figure 2). These studies revealed that SopB C460S localized primarily to early endosomes, as there was a high degree of co-localization with EEA1 (Figures 2a–2c). There was no evidence that SopB associated with other internal membranes such as late endosomes or the TGN as there was only very limited colocalization of SopB C460S with CI-M6PR or TGN46 respectively (Figures 2d–2i). The co-localization with EEA1, which is directed to early endosomes via an interaction with Rab5 and PtdIns(3)P [28], gives an indication that SopB is targeted primarily to membranes containing PtdIns(3)P. To address more directly whether SopB localizes primarily to PtdIns(3)P-containing cellular compartments, we co-expressed SopB C460S together with GFP–iFYVE FENS-1, which is a probe for PtdIns(3)P (Figure 3) [26]. Exogenous expression of GFP–iFYVE FENS-1 caused swelling and vacuolation of early endosomal compartments, as described previously [26]. SopB C460S showed substantial co-localization with GFP–iFYVE FENS-1 on these swollen early endosomes, clearly demonstrating that SopB localizes to PtdIns-(3)P-containing membranes (Figures 3a–3c). To determine whether SopB localizes to PtdIns(3)P-containing endosomes via a direct interaction with PtdIns(3)P, SopB C460S-transfected COS-7 cells were treated with the PI3K (phosphoinositide 3-kinase) inhibitor wortmannin. This treatment did not result in dissociation of SopB C460S from membranes, whereas GFP–iFYVE FENS-1 expressed in the same cells did not remain membrane-associated (Figures 3d–3f).

Figure 2 SopB C460S localizes to early endosomes

COS-7 cells were transfected (ai) with pRK5mycSopB C460S. Cells were fixed after 16 h and immunostained with anti-Myc antibodies, followed by (af) Alexa Fluor 568-conjugated anti-mouse IgG secondary antibodies or (gi) Alexa Fluor 488-conjugated anti-mouse IgG secondary antibodies. Cells were co-immunostained with (ac) anti-EEA1, (df) anti-CI-M6PR or (gi) anti-TGN46 antibodies followed by (af) Alexa Fluor 488-conjugated anti-rabbit IgG or (gi) Alexa Fluor 568-conjugated anti-sheep IgG antibodies. Fluorescence corresponding to (Myc) SopB C460S is shown in (a, d) (red) and (g) (green). Fluorescence corresponding to EEA1, CI-M6PR and TGN46 is shown in (b, e) (green) and (h) (red) respectively. Images of merged fluorescence are shown in (c, f, i) (yellow fluorescence indicates co-localization). Insets in (ac) are magnifications of boxed areas.

Figure 3 SopB C460S is present on PtdIns(3)P-containing membranes

COS-7 cells were co-transfected (af) with pRK5mycSopB C460S and GFP–iFYVE FENS-1. Cells in upper panels (ac) were mock treated and cells in lower panels (df) were treated for 3 h with 100 nM wortmannin (WM) in a serum-free medium. The media containing the wortmannin was changed every hour during treatment: cells were fixed and immunostained with (af) anti-Myc antibodies followed by Alexa Fluor 568-conjugated anti-mouse IgG secondary antibodies. Fluorescence corresponding to (Myc) SopB C460S is shown in (a, d) (red). Fluorescence corresponding to GFP–iFYVE FENS-1 is shown in (b, e) (green). Images of merged fluorescence are shown in (c, f) (yellow fluorescence indicates co-localization).

SopB co-localizes to an ESCRT-containing compartment

We have previously shown that an ESCRT-III component, mVps24p, binds to PtdIns(3,5)P2 and acts as an effector of this phosphoinositide [21]. Thus the action of SopB [by preventing accumulation of PtdIns(3,5)P2 on ESCRT-containing membranes] may manifest itself by compromising ESCRT function. To examine this possibility, SopB C460S was co-expressed together with the dominant-negative ESCRT components NT-Vps24p–GFP or GFP–Vps4p E235Q (Figure 4). NT-Vps24p–GFP associates with intracellular membranes of endocytic origin and induces their vacuolation. Vps4p is not strictly a component of ESCRT-I, -II or -III but catalyses the disassembly and release of ESCRT components from membranes [21,23,24,29]. The dominant negative GFP–Vps4p E235Q is ATPase-defective, associates with intracellular membranes and is not able to catalyse ESCRT disassembly, also resulting in vacuolation. Co-localization of SopB C460S with NT-Vps24p–GFP (Figures 4a–4c) and GFP–Vps4p E235Q (Figures 4d–4f) was observed. This co-localization is entirely consistent with the action of SopB having an effect on ESCRT function.

Figure 4 Dominant-negative ESCRT components assemble on SopB-containing membranes

COS-7 cells were co-transfected with (af) pRK5mycSopB C460S and either a construct for the expression of (ac) NT-Vps24p–GFP or (df) GFP–Vps4p E235Q. Cells were fixed and immunostained with anti-Myc antibodies followed by Alexa Fluor 568-conjugated anti-mouse IgG antibodies. Fluorescence corresponding to (Myc) SopB C460S is shown in (a, d) (red). Fluorescence corresponding to NT-Vps24p–GFP and GFP–Vps4p E235Q is shown in (b) and (e) respectively (green). Images of merged fluorescence are shown in (c, f) (yellow fluorescence indicates co-localization). Insets (ac) are magnifications of boxed areas. CHO-K1 cells were co-transfected with (gi) pRK5mycSopB C460S and GFP–PIKfyve. Cells were fixed and immunostained with anti-Myc antibodies followed by Alexa Fluor 568-conjugated anti-mouse IgG antibodies. Fluorescence corresponding to (Myc) SopB C460S is shown in (h) (red). Fluorescence corresponding to GFP– PIKfyve is shown in (g) (green). A merged image is shown in (i).

Experiments to co-transfect SopB C460S and GFP–PIKfyve into COS-7 cells were attempted but were not successful as GFP–PIKfyve expression could not be detected in these cells. GFP–PIKfyve could, however, be expressed in CHO-K1 cells [30], in which co-expressed SopB C460S showed significant punctate co-localization with GFP–PIKfyve (Figures 4g–4i).

Cells infected with Salmonella expressing SopB have defective receptor degradation kinetics

It has been observed that PtdIns(3)P accumulates on SCVs of Henle-407 cells infected with wt Salmonella but not a Salmonella deletion mutant lacking SopB [11]. To see if this is a general phenomenon for other mammalian cell types, A431, COS-7 and HeLa cells were infected with wt Salmonella (wt Sal. dublin) or an isogenic deletion mutant lacking SopB (ΔSopB Sal. dublin). In order to visualize the intracellular bacteria, the wt Sal. dublin and ΔSopB Sal. dublin were transformed with a GFP-expressing plasmid pSU2007 (see the Experimental section). The presence or absence of PtdIns(3)P on SCVs was monitored by immunostaining fixed cells for endogenous EEA1. Observation of a large number of cells clearly showed that SCVs from wt Sal. dublin-infected cells had large amounts of EEA1 associated with them (Figures 5a–5c), whereas those from ΔSopB Sal. dublin-infected cells had comparatively low amounts of EEA1 associated with them (Figures 5d–5f). Similar results were obtained upon bacterial infection of HeLa and COS-7 cells (results not shown).

Figure 5 A431 cells infected with wt Sal. dublin have EEA1-enriched SCVs

A431 cells were infected with (ac) wt Sal. dublin (‘WT S. dublin’) or (df) ΔSopB Sal. dublin (‘ΔSopB S. dublin’), both transformed with the GFP-expression plasmid, pSU2007, for visualization purposes. Following infection, cells were washed, fixed and immunostained with anti-EEA1 antibodies followed by Alexa Fluor 568-conjugated anti-rabbit IgG secondary antibody. GFP-expressing (b) wt Sal. dublin and (e) ΔSopB Sal. dublin and (c, f) merged images are shown. Insets in (ac) are magnifications of boxed areas. Fluorescence corresponding to EEA1 is shown in (a, d; red). Fluorescence corresponding to wt Sal. dublin and ΔSopB Sal. dublin is shown in (b) and (e) respectively (green). Images of merged fluorescence are shown in (c, f).

The accumulation of PtdIns(3)P, due to SopB antagonizing its transition to PtdIns(3,5)P2, on SCVs has been proposed to divert SCVs away from the endocytic pathway that leads to fusion with destructive lysosomes. In order to test whether SopB could divert endocytic trafficking away from lysosomes, an EGFR degradation assay was used. Following EGF stimulation of uninfected A431 cells, EGFR was rapidly degraded with a half-life of approx. 8 min (Figure 6). If, however, the A431 cells were infected with wt Sal. dublin 60 min prior to EGF stimulation, the kinetics of EGFR degradation was much reduced. Even 16 min after EGF stimulation, 87% of the EGFR remained intact. The kinetics of degradation of EGFR in cells pre-infected with ΔSopB Sal. dublin 60 min prior to EGF stimulation was virtually identical with that in uninfected cells. The localization of EGFR following EGF stimulation of Sal. dublin-infected A431 cells was examined using immunofluorescence microscopy. Even 60 min post-stimulation, EGFR was clearly visible on the limiting membranes of large vesicles within wt Sal. dublin-infected A431 cells (Figure 6c), whereas in ΔSopB Sal. dublin-infected cells, it was barely detectable by immunofluorescence (results not shown). Interestingly, in most wt Sal. dublin-infected cells, EGFR did not localize to SCVs, but to vesicles devoid of any bacterium.

Figure 6 Cells infected with wt Sal. dublin have impaired EGFR trafficking to the lysosome

A431 cells were infected with wt Sal. dublin (‘WT S. dublin’), ΔSopB Sal. dublin (‘ΔSopB S. dublin’) or mock treated. Following infection, cells were washed and incubated in fresh media containing 500 ng/ml EGF. Cells were harvested at the times indicated, and protein extracts were made and analysed by Western blotting for EGFR and β-tubulin. (a) Immunoblots from a representative experiment are shown. (b) A graphical representation of quantifications of immunoblots from three separate experiments. Significant differences (t test; *P<0.05) are indicated. (c) A431 cells infected with wt Sal. dublin were incubated with EGF as described above. After 1 h the cells were washed, fixed and immunostained with anti-EGFR antibodies followed by Alexa Fluor 568-conjugated anti-mouse IgG secondary antibody. Panels (i) (ii) and (iii) show EGFR staining GFP-labelled wt Sal. dublin, and merged images respectively. Note: a neighbouring cell not infected with bacteria does not accumulate an intracellular pool of EGFR.


SopB is known to have a phosphoinositide phosphatase activity that can utilize a variety of phosphoinositides and inositol phosphates as substrates in vitro [8,9]. Furthermore, SopB contributes to the establishment of an intracellular protective niche in which Salmonella can survive and replicate in its host. How it does this remains uncertain, although SopB has been implicated in modulating a wide range of host cell processes such as actin rearrangement, Akt activation and chloride secretory responses. In Salmonella-infected cells, the biologically relevant substrates of SopB are likely to be restricted by its subcellular localization, together with the availability of particular phosphoinositides at that location. Recently, it has been shown that an in vivo substrate of SopB is PtdIns(3,5)P2, a phosphoinositide required for multiple membrane-trafficking pathways [1416,19,31].

It has been proposed that the removal of PtdIns(3,5)P2 from, and maintenance of, high levels of PtdIns(3)P on SCVs, may prevent endosomal maturation required for fusion with lysosomes [11]. In the present study, we have examined this possibility further by initially investigating the subcellular localization of SopB. Localization is relevant to the function of SopB as different phosphoinositides are thought to be spatially restricted on different compartments within eukaryotic cells [32]. We also investigated the effect that SopB has on endocytic trafficking to the lysosome.

In order to investigate the subcellular localization of SopB, it was expressed exogenously in mammalian (COS-7) cells in the absence of bacterial invasion. Most of the COS-7 cells expressing the ‘phosphatase-active’ SopB wt became detached from the coverslips or had rounded morphology, while the remainder were highly vacuolated. SopB wt was present on the limiting membranes of the ‘vacuolated’ structures within the cells. We suspect that vacuolation occurs prior to cell death although, as our observations were made on fixed cells, this could not be confirmed. Vacuolation of cells expressing SopB has been observed previously, although whether or not SopB associated with the vacuolated membranes was not investigated [33]. A number of studies in which accumulation of PtdIns(3,5)P2 has been prevented, either by the expression of a dominant-negative PIKfyve [18] or the lipid phosphatase myotubularin [34], report vacuolation of endosomal compartments. The reason for the vacuolation is not clear but a reduction in PtdIns(3,5)P2 levels might inhibit membrane recycling or cause a defect in the invagination of the limiting membranes of MVBs. In either case, if the rate of membrane addition to the organelle exceeds the rate of removal, the result would be an increase in surface area of the organelle.

Due to the cellular toxicity of SopB wt and its dramatic effects on membrane morphology, most of the SopB localization experiments were performed with a phosphatase-dead SopB C460S mutant. Previous studies with SopB have revealed that its membrane association is independent of an active phosphatase domain [35]. Thus any interactions that determine the initial membrane targeting of SopB-wt and SopB C460S are likely to be identical. Exogenously expressed SopB C460S localized almost exclusively to intracellular membranes containing the EEA1 protein but not markers of late endosomes or the TGN. Interestingly, we did not observe SopB C460S at the plasma membrane of transfected cells, a site proposed to be enriched for certain phosphoinositides such as PtdIns(4,5)P2, an in vitro substrate of SopB. Our findings are in contrast with a recent study, in which SopB was shown to be present on the plasma membrane of fibroblasts infected with Salmonella [36], but in agreement with other observations that place SopB at intracellular punctae and on SCVs [7,35]. These discrepancies may be due to the cell types used in the different studies. The dramatic vacuolation seen upon expression of SopB-wt was not observed with SopB C460S. A likely explanation for this is that the inactive SopB C460S can localize to the endosomal membrane but not hydrolyse its substrate to cause the vacuolation phenotype.

When co-expressed with GFP–iFYVE FENS-1, a PtdIns(3)P reporter, SopB C460S co-localized on the GFP–iFYVE FENS-1-containing membranes. The EEA1 and GFP–iFYVE FENS-1 co-localization data indicate that SopB associates with PtdIns(3)P-containing membranes. In order to establish whether SopB associates with membranes via an interaction with PtdIns(3)P, the production of this lipid was inhibited with wortmannin, a PI3K inhibitor. SopB does not appear to associate with PtdIns(3)P directly, as upon wortmannin treatment of COS-7 cells co-expressing SopB C460S and GFP–iFYVE FENS-1, the GFP–iFYVE FENS-1 dissociated from membranes, whereas SopB C460S remained membrane-associated. The basis for the association of SopB with endosomal membranes remains to be established, although a predicted coiled-coil domain in SopB appears necessary for membrane targeting [35].

We believe that the localization of SopB on PtdIns(3)P-containing membranes is significant to PtdIns(3,5)P2 being a relevant in vivo substrate of SopB, as PtdIns(3,5)P2 can only be produced on PtdIns(3)P-containing membranes. The only known route of PtdIns(3,5)P2 synthesis is from PtdIns(3)P. PIKfyve, the enzyme responsible for converting PtdIns(3)P into PtdIns(3,5)P2, localizes to PtdIns(3)P-containing membranes via an FYVE domain [37]. Furthermore, we have shown that, in CHO-K1 cells, SopB C460S and GFP–PIKfyve co-localize on intracellular punctae when co-expressed. Co-localization of an active SopB to the same membranes as PtdIns(3)P places SopB in a location where it can be antagonistic to PIKfyve function.

In addition to the requirement of phosphoinositides for endocytic trafficking via MVBs to lysosomes, a proteinaceous machinery is needed. This protein machinery, known as the ESCRT machinery, is made up of at least three protein complexes, ESCRT-I, -II and -III [38]. Previous studies from our laboratory have revealed that mVps24p (CHMP3), a component of ESCRT-III, is a possible PtdIns(3,5)P2 effector [21]. Thus we speculated that, if SopB is a PtdIns(3,5)P2 phosphatase, it should have effects on the downstream targets of this lipid, namely mVps24p, and may be present on membranes where the ESCRT machinery assembles. To investigate this possibility, we co-expressed SopB C460S together with dominant-negative components of the ESCRT machinery, NT-Vps24p–GFP and GFP–Vps4p E235Q, that are constitutively membrane-associated [21]. There was significant co-localization of SopB C460S with both NT-Vps24p–GFP and GFP–Vps4p E235Q, a finding that may be relevant to SopB having an effect on ESCRT function. Interestingly, SopB is ubiquitylated but not rapidly degraded by the proteasome of mammalian cells [35]. As ESCRT components bind to ubiquitylated proteins, it is possible that a direct interaction between the ESCRT machinery and ubiquitin–SopB is responsible for membrane association of SopB.

It is well established that the ESCRT machinery is involved in sorting membrane receptors, such as activated EGFR, for degradation in lysosomes [29,39] although there is conflicting evidence for a role of PtdIns(3,5)P2 in receptor degradation in mammals. In mammalian cells, expression of a dominant-negative PIKfyve, that blocks PtdIns(3,5)P2 production, does not seem to have a dramatic effect on EGFR degradation [40], whereas overexpression of myotubularin, a phosphatase that hydrolyses PtdIns(3,5)P2, does have an effect [34]. In A431 cells infected with wt Sal. dublin-expressing SopB, we found that degradation of activated EGFR was dramatically inhibited compared with cells infected with the isogenic ΔSopB Sal. dublin strain. This, together with EGFR accumulation on the limiting membranes of large intracellular vesicles of wt Sal. dublin-infected cells, resembles the mammalian ‘class E’ phenotype typified by ESCRT dysfunction. If, as has been proposed, PtdIns(3,5)P2 is a substrate for SopB, our results are consistent with a role for this lipid in receptor degradation. The results indicate that SopB, in addition to diverting SCVs away from lysosomal targeting, perturbs normal endosome to lysosome trafficking by promoting PtdIns(3)P and preventing PtdIns(3,5)P2 accumulation. This hypothesis is strengthened by our observation that EEA1, an endogenous marker for PtdIns(3)P-containing membranes, accumulates on SCVs of A431 cells infected with wt but not ΔSopB Sal. dublin.

The present study does not rule out other inositol phosphates and phosphoinositides as in vivo substrates for SopB. However, we believe that a number of the cellular effects attributed to SopB may be explained by it causing a defect in endosome to lysosome trafficking due to its PtdIns(3,5)P2 phosphatase activity. For example, the phenomenon of prolonged Akt activation by SopB may be explained as follows. Akt can be activated/phosphorylated as a consequence of receptor activation at the plasma membrane. Normally, activated receptors would be rapidly down-regulated by targeting them to lysosomes for degradation, thus attenuating Akt activation. However, if activated receptors are not targeted for degradation, but continue to signal from endosomes even after internalization [41] as a consequence of SopB action, Akt activation would be prolonged.

As mentioned previously, PtdIns(3,5)P2 has been implicated in multiple membrane trafficking events and in regulating vacuole morphology [17]. Its depletion in yeast contributes to a complex phenotype involving multiple effector proteins [16,42]. Thus, although our results implicate SopB/PtdIns(3,5)P2 in endosome–lysosome trafficking in mammalian cells, it is likely that PtdIns(3,5)P2 depletion will effect other cellular processes not investigated in the present study. For example, it has recently been shown that membrane virus budding, a process also requiring the host cells ESCRT machinery, is inhibited upon small interfering RNA-mediated depletion of PIKfyve from cells [43].

In summary, our results provide further evidence that SopB may effect endosome to lysosome trafficking by preventing PtdIns(3,5)P2 accumulation on endosomal/SCV membranes. Our working hypothesis is that PIKfyve functions downstream of ESCRT assembly. We speculate that, upon generation of PtdIns(3,5)P2, a conformational change occurs in ESCRT. This conformational change, driven by an interaction of the negatively charged PtdIns(3,5)P2 with the positively charged N-termini of CHMPs, may provide the force necessary to cause membrane invagination/scission of MVBs. Further investigation into the role of PtdIns(3,5)P2 and ESCRT-III in endosome to lysosome trafficking will be required to gain a better understanding of the function of SopB.


This work was supported by the Wellcome Trust (project grant 070085 to P.W. and B.J.R.).

Abbreviations: MVB, multivesicular body; CHMP, charged MVB protein; CHO, Chinese-hamster ovary; CI-M6PR, cation independent mannose 6-phosphate receptor; DMEM, Dulbecco's modified minimal essential medium; EEA1, early endosome antigen 1; EGF, epidermal growth factor; EGFR, EGF receptor; ESCRT, endosomal sorting complexes required for transport; FENS-1, FYVE domain containing protein localized to endosomes-1; GFP, green fluorescent protein; NCS, newborn calf serum; PI3K, phosphoinositide 3-kinase; PIKfyve, PhosphoInositide Kinase for five position containing a Fyve finger domain; SCV, Salmonella-containing vacuole; SPI, Salmonella pathogenicity island; TGN, trans-Golgi network; TTSS, type III secretion system; Vps, vacuolar protein sorting; wt, wild-type


View Abstract