Research ResourceHost-Pathogen Interactions

Phosphoproteomic Analysis of Salmonella-Infected Cells Identifies Key Kinase Regulators and SopB-Dependent Host Phosphorylation Events

See allHide authors and affiliations

Science Signaling  20 Sep 2011:
Vol. 4, Issue 191, pp. rs9
DOI: 10.1126/scisignal.2001668


Salmonella enterica is a bacterial pathogen that causes gastroenteritis and typhoid fever. Virulence is achieved by two type III secretion systems that translocate effector proteins into host cells, where they mimic or block host protein function. Effectors translocated into host cells by the first type III secretion system facilitate invasion and stimulate intracellular signaling cascades leading to inflammation. Here, we performed global temporal analysis of host signaling events induced during the initial stages of Salmonella infection and identified the dynamics of host protein phosphorylation as well as differences between growth factor–stimulated and bacteria-induced signaling. Informatics analysis predicted that sites with altered phosphorylation in infected cells were targeted by the serine-threonine kinases Akt, protein kinase C, and Pim and that up to half of the host phosphorylation events detected after Salmonella infection required the effector protein SopB. Our data reveal extensive manipulation of host phosphorylation cascades by this Salmonella effector and provide a detailed map of the events leading to intestinal inflammation, which is the crucial host response that enables Salmonella to proliferate in the intestine.


Salmonella enterica is a foodborne, facultative intracellular bacterium that causes gastroenteritis and typhoid fever leading to considerable human morbidity and mortality (1). The various serovars (antigenically distinct members of the same species) of S. enterica have evolved intricate mechanisms for evading host immunity, whereby the pathogen induces its own internalization into a membrane-bound vacuole termed the Salmonella-containing vacuole, which evades degradation within host cells (2, 3). Effector proteins, translocated into host cells by two type III secretion systems (T3SSs) encoded on Salmonella pathogenicity islands 1 and 2 (SPI-1 and SPI-2), control the formation and maturation of the Salmonella-containing vacuole (4).

Whereas T3SS-2 acts at later stages of infection to direct trafficking of and bacterial replication within the Salmonella-containing vacuole, T3SS-1 effectors are translocated into the host cytosol immediately upon contact with the host cell. T3SS-1 effectors target Rho family guanosine triphosphatases (GTPases) and actin-binding proteins to induce membrane ruffling, macropinocytosis, and nuclear responses leading to inflammation (57). They also regulate bacterial replication and trafficking of the Salmonella-containing vacuole and host functions such as cell death, ion secretion, and cell signaling through phospholipids, as well as protein phosphorylation, acetylation, and ubiquitylation (811).

The phosphoinositide phosphatase SopB and the guanine nucleotide exchange factors SopE and SopE2 (SopE/E2) activate cell division control protein 42 (Cdc42) to induce downstream signaling through c-Jun N-terminal kinase (JNK) and mitogen-activated protein kinase (MAPK) p38 isoforms, leading to proinflammatory cytokine production (12). SopB also activates protein kinase B (also known as Akt), which induces antiapoptotic signaling, while also stimulating p21-activated kinase 4 (PAK4) and inactivating a GTPase-activating protein (GAP) for Ras-related protein Rab-14 (Rab14) (13, 14). Conversely, the effector SptP (which is a GAP and protein-tyrosine phosphatase) inactivates the kinase RAF and dephosphorylates transitional endoplasmic reticulum adenosine triphosphatase (ATPase) (VCP), and SspH1 binds protein serine-threonine kinase N1 (PKN1) to inhibit nuclear factor κB (NFκB)–dependent gene expression (1517). Finally, AvrA also inhibits MAPKs through acetylation and prevents NFκB activation by deubiquitylating NFκB inhibitors (IκBs) (18, 19). SipA also induces protein kinase C (PKC) activation, leading to transepithelial migration of polymorphonuclear lymphocytes (2022). However, activation of PKC by SipA is indirect, and the signaling components of this pathway remain largely uncharacterized.

Although reductionist approaches have been used successfully to identify these signaling events, advances in global proteomics techniques, such as metal oxide chromatography (MOC), now make systems-level tracking of protein phosphorylation possible (23, 24). These methods, in combination with stable isotope labeling of amino acids in cell culture (SILAC), can resolve protein phosphorylation events in space and time (25). Here, we apply quantitative phosphoproteomics to measure more than 9500 phosphorylation events in an epithelial cell line during the initial 20 min after Salmonella infection. We have identified host pathways that are regulated during infection, as well as important distinctions between ligand receptor–induced and bacteria-induced signaling dynamics. Many of the changes induced by wild-type S. enterica serovar Typhimurium (S. Typhimurium) were functionally validated with a strain lacking the T3SS-1 effector SopB. Hundreds of dynamic host phosphorylation events were dependent on SopB, reflecting its wide-reaching impact on bacteria-induced signaling.


Quantitative phosphoproteomics analysis of host cell signaling during Salmonella invasion

For phosphoproteomics analysis, three HeLa cell populations were labeled with light, medium, and heavy isotopologs of lysine and arginine for SILAC quantitation (fig. S1). Light-labeled populations were mock-infected, whereas medium- and heavy-labeled populations were infected with wild-type or ΔsopB strains for the various times indicated. Cells were lysed; separated into nuclear, membrane, and cytosolic fractions; and solubilized in detergent. Equal masses of protein from each SILAC label and fraction were combined, digested with trypsin, and fractionated by in-solution isoelectric focusing (IEF), and phosphopeptides were enriched by MOC (24). Biological replicates of each experiment each yielded 30 samples for liquid chromatography–tandem mass spectrometry (LC-MS/MS). Three independent biological replicates were analyzed after infection with wild-type S. Typhimurium, and two independent biological replicates were analyzed after infection with the ΔsopB strain.

Together, 1973 phosphorylated proteins were identified with a false discovery rate of 1%, corresponding to 6893 nonredundant phosphopeptides and 9508 nonredundant phosphorylation sites (table S1; The dynamic profiles measured appeared to be accurate because the average coefficient of variation (XCV) across biological replicates was <15% (table S1). More than 85% of phosphopeptides were quantified in at least one experiment, and 61, 43, and 49% of these were identified in the nuclear, membrane, and cytosolic fractions, respectively. The distribution of phosphorylation events between Ser, Thr, and Tyr was 82:15:3, similar to a previous report (26), and highly correlated with the overall 80:16:4 ratio for all phosphorylation sites observed in PhosphoNET (

At least 24% of detected phosphopeptides showed altered phosphorylation during infection. Fuzzy c-means (FCM) clustering of the regulated profiles generated four clusters containing peptides with increased phosphorylation and two containing peptides with decreased phosphorylation (Fig. 1A). The clusters of peptides with increased phosphorylation were different, peaking or beginning to increase in abundance at early or late time points. Similarly, the clusters containing peptides with decreased phosphorylation began to decrease either early or quite late, indicating differences in phosphopeptide profiles across the data set. Differences among the Gene Ontology (GO) terms enriched in the clustered proteins with increased or decreased phosphorylation were detected. Terms for regulation of apoptosis, transmembrane transport, nuclear organization, and cell proliferation were enriched among proteins with increased phosphorylation, whereas terms for cytoskeleton organization, protein complex assembly, and cell polarity were enriched among proteins with decreased phosphorylation (Fig. 1B).

Fig. 1

Global analysis of phosphoproteomics data. (A) FCM clustering of profiles of phosphopeptides with altered phosphorylation during infection (P < 0.01, Benjamini-Hochberg false discovery rate test). Each profile is color-coded according to membership values shown on the color bar. The number of SDs from the mean (log10-transformed) is shown on the y axis and time after infection is shown on the x axis. (B) GO analysis of proteins with altered phosphorylation during infection. All proteins with altered phosphorylation (containing at least one phosphopeptide with P < 0.01), proteins with increased phosphorylation [containing at least one phosphopeptide with membership value >0.6 from clusters of increased phosphorylation in (A)], and proteins with decreased phosphorylation [containing at least one phosphopeptide with membership value >0.6 from clusters of decreased phosphorylation in (A)] were analyzed for enrichment in GO terms. GO terms for biological process are shown on the x axis, and P = 0.01 is indicated by a dashed line. (C) Distribution of changes in phosphorylation within each subcellular fraction. Sites with decreased phosphorylation are shown in magenta, and those with increased phosphorylation are shown in cyan. Time after infection is shown in the z axis.

Overall trends within the nuclear, membrane, and cytosolic fractions were also considered. Although the percentages of phosphopeptides with altered phosphorylation were similar (22, 18, and 20%) across each fraction, peptides with decreased phosphorylation were enriched in the membrane and nuclear fractions (Fig. 1C). Although quicker dynamics might be expected in the membrane and cytosolic fractions (due to cascades initiating at the cell periphery at sites of invasion), overall rates were similar across each fraction. This may be due to the fact that the T3SS-1 pool of effectors is quickly injected into host cells, and several directly target host proteins in the cytosol and nucleus (17, 19, 27).

Validation and confirmation of phosphoproteomics data

To validate the phosphoproteomics data, we confirmed profiles for several phosphorylation sites by immunoblotting. Phosphosite-specific antibodies were used to probe phosphorylation dynamics within nuclear, membrane, and cytosolic fractions prepared from cells that were either mock-infected or infected for 2, 5, 10, and 20 min with wild-type S. Typhimurium. These profiles compared favorably to those generated by quantitative LC-MS/MS (Fig. 2A).

Fig. 2

Validation of phosphoproteomics data. (A) Validation of SILAC profiles by immunoblotting with phosphospecific antibodies. (B) Phosphoproteomics data for host proteins that are targeted by Salmonella. Phosphosites are indicated in magenta in each peptide sequence. In each case, equal protein amounts were analyzed from each time point. MARCKS, myristoylated alanine-rich C-kinase substrate; CFL1, cofilin-1; PKR2, cAMP-dependent protein kinase type II-α regulatory subunit; CDK2, cyclin-dependent kinase 2; ARAF, serine/threonine-protein kinase A-Raf.

The SILAC data were also compared to protein phosphorylation signaling events previously reported in host cells during Salmonella invasion. In each case in which a phosphosite within a host target of Salmonella was identified in the LC-MS/MS data, previously reported alterations in phosphorylation were confirmed by the SILAC profiles. Phosphorylation of the TxY motif in extracellular signal–regulated kinase 1 and 2 (ERK1/2) increased 10-fold during Salmonella infection (Fig. 2, A and B), which is consistent with previous work showing activation of these kinases downstream of SopB and SopE/E2 (12, 28). Similarly, increased phosphorylation of threonine pairs 69/71 and 51/53 was observed in the cAMP (adenosine 3′,5′-monophosphate)–dependent activating transcription factor 2 (ATF2) and ATF7, which are components of the activator protein 1 (AP-1) complex (29, 30). Downstream of SopB, we observed increased phosphorylation of Ser474 from Akt2, as well as Ser588 of an Akt effector, AS160 (Fig. 2B) (14, 31). Phosphorylation of Ser181 in PAK4 was also increased, which is downstream of Akt and follows activation of Cdc42 by SopB and SopE/E2 (14, 32). Changes in the phosphorylation of Ser260 in RafA were also expected because SptP inhibits signaling through the ERK1/2 pathway (15).

Phosphorylation dynamics after epidermal growth factor treatment compared to Salmonella infection

Signaling through the epidermal growth factor receptor (EGFR) represents a relatively well-characterized pathway that is initiated by extracellular ligand-receptor binding and that proceeds through an intracellular MAPK cascade to induce transcription of genes encoding factors that promote cell growth and proliferation. Conversely, a host cell receptor for Salmonella has not been identified, and induction of signaling cascades has not been detected in epithelial cells as a result of bacteria binding to the cell surface (7). Instead, the array of effector proteins injected into host cells localizes to various compartments within the host cell, in some cases inducing MAPK signaling for cytokine production or stimulating an innate immune response (or both). Given the different spatial modes of action, one might expect that signaling downstream of Salmonella would not necessarily conform to that of canonical cellular signaling pathways.

Olsen et al. have previously reported global temporal profiles for changes in the phosphorylation of Tyr, Ser, and Thr sites during EGF stimulation (26). Although a slightly different time course was used and lysates were separated into only nuclear and cytosolic fractions, their data were also collected from HeLa cells, and the experimental design and data set size are similar to those of our study. Olsen et al. observed that changes in the phosphorylation of Tyr sites occur earlier than that of Ser or Thr sites during stimulation with EGF (Fig. 3A). Treating our data similarly, we observed similar rates for all three sites, indicating that, unlike EGF, Salmonella does not trigger rapid changes in the phosphorylation of Tyr sites in proteins at the cell membrane (Fig. 3A). However, the magnitude of changes in the phosphorylation of Tyr sites induced by Salmonella is higher than that for Ser or Thr sites. Although we would not necessarily expect a great degree of overlap between phosphorylation sites regulated by Salmonella and EGF, a more detailed comparison of individual sites is illuminating. Seventy-two regulated phosphopeptides were common between the two data sets, with only half of them showing vaguely similar patterns of regulation (Fig. 3B). Thus, even individual proteins can play different roles in signaling pathways. In addition, among peptides with similar profiles (such as ERK2), changes in phosphorylation were almost always delayed during Salmonella infection compared to EGF stimulation. This indicates that stimulation of the various host targets by SPI-1 delivered effectors occurs more slowly than signaling events initiated by receptor activation at the cell membrane. We asked whether specific functional classes of proteins are used in a similar way. Enriched GO terms for cellular compartment were similar after EGF treatment and Salmonella infection, but terms for biological processes relating to transcriptional regulation were enriched by EGF stimulation, whereas those relating to small GTPase-mediated signaling transduction and apoptosis were enriched during Salmonella infection (Fig. 3C).

Fig. 3

Comparison of EGF- and Salmonella-induced signaling. (A) Average SILAC profiles for serine, threonine, and tyrosine residues showing increased phosphorylation. Dashed lines represent data from EGF stimulation and solid lines represent data from Salmonella infection. (B) Hierarchical clustering of all peptides with altered phosphorylation that are similar between the EGF and the Salmonella data sets. Time after EGF treatment or Salmonella infection is indicated, as well as a color scale showing log fold change within each profile. Profiles corresponding to the peptide VADPDHDHTGFLpTEpYATR from ERK2 are indicated by an arrow. Peptides with identical sequence and the same number of phosphosites were considered common between the two data sets. (C) GO analysis of proteins showing increased phosphorylation after EGF stimulation or Salmonella infection (Fig. 1B). GO terms for biological processes are shown on the x axis, and P = 0.01 is indicated by a dashed line.

Host phosphorylation sites regulated during Salmonella infection

A main trigger of innate immune responses is the recognition of pathogen-associated molecular patterns (PAMPs) by Toll-like receptors (TLRs) and nucleotide oligomerization domain–like receptors (NLRs), leading to MAPK signaling and the production of proinflammatory cytokines. These proteins are abundant in professional phagocytes, but epithelial cells, which are continually exposed to large numbers of bacteria, down-regulate signaling through these receptors to prevent uncontrolled inflammation (33). Salmonella actively stimulates innate immune responses in epithelial cells through mechanisms dependent on effector proteins SopE/E2 and SopB, but which do not involve TLRs or NLRs (7). Salmonella use tetrathionate derived from reactive oxygen species generated during inflammation as an electron acceptor for respiration (34). This gives them a growth advantage over commensal microbes, and inflammation enhances transmission and growth of Salmonella in the intestinal lumen (35). Thus, signaling events observed during Salmonella invasion of these cells likely represent active manipulation of host processes by the bacteria, such as cytoskeletal dynamics and immune signaling, both of which are central to pathogenesis and disease.

Within the wild-type data set, several sites showing changes in phosphorylation likely represent previously unknown host targets of T3SS-1 effectors (Table 1). For example, increased phosphorylation of activating sites in Ras GTPase-activating–like protein IQGAP1 and PKD3 (protein serine-threonine kinase D3) was observed. Because PKC phosphorylates both of these sites, these represent potential signaling targets downstream of the T3SS-1 effector SipA, which promotes PKC activation (9, 20, 36, 37). IQGAP1 promotes bacterial invasion and activation of Rac1 and Cdc42 during Salmonella infection (38). Similarly, dynamically phosphorylated sites in the transcriptional regulators histone deacetylase 1 and non–histone chromosomal protein HMG-14 imply a role for casein kinases 1 and 2 (39, 40).

Table 1

Selected host phosphosites regulated during Salmonella infection.

View this table:

Although the upstream kinases and functions of the sites in Table 1 have been previously described, most phosphosites regulated during infection have not been characterized. Using a kinase prediction algorithm, we scored all phosphosites whose phosphorylation increased more than twofold during infection with wild-type S. Typhimurium against the substrate specificity determinants for 493 human kinase domains. Akt and PKC (which are targets of SopB and SipA) were well represented among the top 20 scoring kinases, as were the Pim family kinases p70S6K (70-kD ribosomal protein S6 kinase) and mTOR (mammalian target of rapamycin) (Table 2). In addition, among the total 241 phosphosites identified from 66 different kinases (table S1), we observed increased phosphorylation at known or predicted activating sites in 10 of them, including PKCα, Src, Akt2, ERK1, ERK2, MSK2 (ribosomal protein S6 kinase α-4), RSK1 (ribosomal protein S6 kinase α-1), CLK1, PKD3, and CRK7. Some of these, such as the ERKs, likely have somewhat restricted substrate repertoires and do not rank at the top of Table 2, but others, such as PKC and Akt, may be used by Salmonella to globally regulate host cell function. Furthermore, the emergence of Pim consensus motifs among Salmonella-regulated sites suggests an immediately testable hypothesis, that inhibition of Pim kinases should affect one or more endpoints of Salmonella infection.

Table 2

Kinases most frequently predicted to be upstream of sites showing increased phosphorylation in infected cells.

View this table:

To identify roles during Salmonella infection for the kinases predicted from our phosphoproteomics data, we assayed interleukin-8 (IL-8) concentrations in supernatants from wild-type S. Typhimurium–infected cells treated with different kinase inhibitors. Treatment of infected cells with kinase inhibitors directed at Akt, MEK (mitogen-activated or extracellular signal–regulated protein kinase kinase), PKC, or Pim family kinases identified a dominant role for Pim family kinases in the release of IL-8 from Salmonella-infected epithelial cells (Fig. 4A) and confirms the involvement of PKC in this process (22). An Akt inhibitor did not affect IL-8 secretion, consistent with a transcriptional study of epithelial cell responses to Salmonella (7). Small interfering RNA (siRNA) targeting of Pim-2 or Pim-3 not only decreased the abundance of the intended target but also appeared to affect the abundance of the other isoform. Whether this is cross-reactivity of the siRNA, an indirect compensatory effect, or an artifact of partial cross-reactivity of antibodies is unclear, but it makes the impact of these knockdowns on IL-8 secretion (Fig. 4B) difficult to interpret. Regardless, the inhibitor and proteomic data suggest that Pims are an important mediator of Salmonella-induced signaling.

Fig. 4

Effects of kinase inhibitors and siRNA treatment on IL-8 secretion by infected cells. (A) Mean IL-8 concentrations determined by ELISA of the supernatants of HeLa cells infected with wild-type Salmonella. Kinase inhibitors are indicated on the x axis, and the target families of the inhibitors are indicated above each column. CEC, chelerythrine; Pim Inh, Pim-1 inhibitor 2. Inhibitors that significantly affected IL-8 production relative to the control (P < 0.05, Bonferroni post hoc test from one-way ANOVA) are indicated with an asterisk. Relative cell viability is also shown, as determined by reduction of XTT by viable cells. A450, absorbance at 450 nm. (B) siRNA targeting of each Pim isoform is indicated on the x axis. Immunoblots confirming knockdown of Pim-2 or Pim-3 are shown. Pim-1 was not detectable in HeLa cells. Error bars represent SEM; all experimental conditions were repeated at least three times.

Phosphorylation dynamics after infection with wild-type compared to ΔsopB Salmonella

SopB regulates various processes within host cells. These include formation and trafficking of the Salmonella-containing vacuole and activation of Rho GTPases (Cdc42 and RhoG) and Akt, largely through the action of two phosphoinositide phosphatase domains located near its C terminus (6, 13, 4144). SopB also acts in a redundant manner with SopE/E2 to induce innate immune signaling detected within host epithelial cells (7). Thus, we hypothesized that SopB is responsible, through its actions on Cdc42 and Akt, for a large fraction of host signaling that is induced by Salmonella infection.

Globally, infection with the ΔsopB strain resulted in a 35% reduction in the number of proteins with altered phosphorylation relative to wild type (Fig. 5A), a result that was consistent for phosphorylated Ser and Thr sites in all three subcellular fractions tested. Infection with ΔsopB resulted in a reduced number of membrane fraction proteins with altered Tyr phosphorylation and an increased number of cytosolic fraction proteins with altered Tyr phosphorylation. SopB localizes to the host cell membrane and to the Salmonella-containing vacuole and likely locally alters signaling at these locations (42, 45). SopB can also recruit proteins to these compartments, a possible explanation for the reduced regulation observed in the cytosol (10).

Fig. 5

Host phosphorylation events dependent on the T3SS-1 effector SopB. (A) Global impact of SopB on host phosphorylation. The percentage of phosphoserine (pSer), phosphothreonine (pThr), and phosphotyrosine (pTyr) sites that show changes (P < 0.01, Benjamini-Hochberg false discovery rate test) in phosphorylation during wild-type (WT) and ΔsopB infection is shown for each subcellular fraction. (B) MAPK regulation dependent on SopB. SILAC profiles are shown for the indicated phosphopeptides from RSK1, MSK2, and ERK5 during infection with WT and ΔsopB S. Typhimurium. (C) Amino acid sequences surrounding phosphosites with SopB-dependent changes in phosphorylation that feature an RxRxxS motif. Gene names, UniProt identification numbers, and location of the phosphosite are shown, as well as kinases that target each site. (D) Host sites whose change in phosphorylation after infection depends on SopB. SILAC profiles are shown for the indicated phosphopeptides containing the RxRxxS motif. In (A) and (D), phosphorylated residues are shown in magenta.

Initially, SILAC profiles for known host phosphorylation targets were compared during infection with the wild-type and ΔsopB strains. As expected, due to redundancy in SopB and SopE/E2 function, changes in the phosphorylation of ERK2, ATF2, ATF7, and PAK4 were reduced, but not eliminated, after infection with the ΔsopB strain. Phosphorylation of activating sites in both RSK1 and MSK2 was also increased 4- and 10-fold during infection with wild-type but not the ΔsopB strain (Fig. 5B). Ser368 in RSK1 is an activating site phosphorylated by ERK1/2, and Ser343 and Ser347 in MSK2 are phosphorylated by p38 (46, 47). In contrast to the proteins mentioned above, our data suggest that activation of RSK1 and MSK2 during Salmonella infection may depend on SopB. This suggests that SopB and SopE/E2 may regulate the activity of MAPKs differently, perhaps through signaling downstream of Akt.

The four MAPK cascades are the ERK1/2 pathway, the JNK [same as stress-activated protein kinase (SAPK)] pathway, the p38 pathway, and the ERK5 pathway (48), all of which regulate cellular growth and proliferation. However, compared to the other three pathways, less is known about the signaling pathways leading to ERK5 activation, which has not been implicated in Salmonella infection (49). We found that the phosphorylation of two activating autophosphorylation sites (Ser731 and Thr733) in ERK5 was increased more than fourfold during infection with the ΔsopB strain (Fig. 5B) (50). Although these data show that Salmonella infection can alter the activation status of ERK5, this kinase was not identified in the wild-type data set, so it is currently unclear whether ERK5 activity is increased by wild-type Salmonella or decreased by SopB.

Presumably not all the proteins that show decreased phosphorylation with ΔsopB infection are actually direct targets of SopB, either through its phosphatase activity or by other means. Rather, it is more likely that SopB acts through a small number of kinases or phosphatases to alter the phosphorylation of a wider range of host proteins indirectly. To identify kinases that might be regulated downstream of SopB, we first filtered out the phosphopeptides that showed at least a twofold difference between wild-type and ΔsopB infections and that were significantly different from mock infections. Among the 348 phosphopeptides meeting these criteria, one sequence motif, RxRxxS, was significantly enriched, occurring in 15 peptides (Fig. 5C). This motif is a consensus recognition phosphorylation site for Akt, RSK, and p70S6K isoforms (51), and at least 7 of the 15 peptides are substrates of one or more of these kinases (Fig. 5C) (51). These data suggest that SopB targets Akt and RSK1 (Fig. 5B), which in turn targets the sodium-hydrogen exchanger 1 (Fig. 5C) (14, 52).

SopB can decrease apoptosis through an Akt-dependent mechanism that has not been fully characterized. Accordingly, we detected a 2.5-fold increase in phosphorylation of Ser99 in Bcl2 antagonist of cell death (BAD) with infection by wild-type bacteria, an increase that is abrogated with ΔsopB (Fig. 5D). Ser99 in BAD is phosphorylated by Akt, resulting in the sequestering of BAD in the cytosol, thereby inactivating its proapoptotic activity by preventing its association with prosurvival proteins at mitochondria (55). Thus, SopB- and Akt-dependent antiapoptotic signaling may be facilitated through BAD inactivation.

Salmonella effector proteins induce actin rearrangement during the invasion process, which requires the activity of SopB and SopE/E2 on small G proteins [heterotrimeric guanosine triphosphate (GTP)–binding proteins]. Akt phosphorylation of Rac1 at Ser71 inhibits GTP binding and, thus, Rac1 activity (53). When active, however, Rac1 can signal through PAK and LIM domain kinase 1 (LIMK) to phosphorylate Ser3 of cofilin and thereby inhibit its actin depolymerizing activity (58, 59). Infection with wild-type Salmonella induced a 3.5-fold increase in Ser71 phosphorylation of Rac1 and a 2-fold decrease in Ser3 phosphorylation of cofilin (Fig. 5D) (see also the immunoblot in Fig. 2A). Both of these effects depended on SopB, indicating that SopB can inactivate Rac1 through Akt to promote cofilin activity.

Recruitment of vesicle-associated membrane protein 8 (VAMP8)–positive endocytic vesicles to Salmonella-induced membrane ruffles is required for efficient invasion and dependent on the phosphatase activity of SopB (60). However, although these vesicles likely provide membranes to the forming macropinosome, targets of VAMP8 during infection are currently unknown. Syntaxin 7 forms a SNARE (soluble N-ethylmaleimide–sensitive factor attachment protein receptor) complex with VTI1b, syntaxin 8, and VAMP8 to fuse membranes within the endosomal pathway (61), and formation of this complex can be increased by Akt-mediated phosphorylation of syntaxin 7 at Ser126 and Ser129 (54). Salmonella induced a SopB-dependent, threefold increase in phosphorylation on both Ser126 and Ser129 (Fig. 5D). Thus, SopB may promote SNARE complex formation through VAMP8 recruitment, as well as Akt-dependent phosphorylation of syntaxin 7 to regulate vesicle fusion during formation of the Salmonella-containing vacuole.


The mechanism or mechanisms through which Salmonella effector proteins bypass innate immune receptors to exploit host signaling pathways represent an exceptional degree of adaptation by Salmonella to its host. Here, we present a global analysis of host phosphorylation events induced during the initial phase of S. Typhimurium challenge of a well-characterized human epithelial cell line, enabling temporal modeling of the infection-induced changes in the phosphorylation of more than 5800 phosphopeptides. Given that several Salmonella effectors activate signaling proteins directly and at various stages downstream of host cell receptors, it is not surprising that signaling induced by Salmonella infection differs from classical receptor-induced signaling.

Several comparisons have been made between signaling induced in host cells by Salmonella infection and by EGF treatment. Initially, it was proposed that Salmonella use the EGFR during invasion, but neither the data presented here nor that in a previous study shows that tyrosine phosphorylation of EGFR changes after Salmonella invasion (62, 63). At a gross level, EGF and Salmonella induce some similar effects and use similar modes of action: Both activate canonical MAPK signaling pathways in cells and induce internalization into an early endosome or similar compartment. However, our data demonstrate that, when compared to EGF-stimulated signaling, Salmonella-induced increases in phosphorylation of Tyr sites are slower, increasing within the same time frame as that of Ser or Thr sites (26). We expect that this is a result of the initiation of Salmonella-induced signaling at intracellular sites where effectors act, rather than at the membrane, as is the case for receptor tyrosine kinase pathways. Indeed, comparison of individual peptides reveals that there is little similarity between the two signaling systems, and, where similar trends are seen, they are almost always delayed in Salmonella-infected cells compared to EGF-treated cells.

Our data provide detailed mechanistic insight into several phenomena associated with Salmonella invasion (Fig. 6). SopB increases Akt-mediated phosphorylation of an inhibitory site (Ser99) in the proapoptotic protein BAD, which supports the notion that SopB decreases apoptosis through sequestration of BAD (13, 55). Likewise, during infection, Rac1 activation and deactivation are controlled by SopE/E2 and SptP to regulate membrane ruffling for internalization of bacteria (12). We observe that SopB promotes Akt-mediated phosphorylation of an inhibitory site in Rac1. Because Rac1 can inactivate the actin-depolymerizing activity of cofilin by inducing LIMK-mediated phosphorylation of cofilin, SopB may preserve cofilin activity through Rac1 inactivation to exert an effect on actin similar to that of SptP (58, 59). This is inconsistent with a report from Patel and Galán (12), in which similar Rac1 activity was detected 20 min after infection with wild-type or ΔsopB S. Typhimurium. However, similar to SptP, SopB has a longer half-life in host cells compared to SopE, suggesting that this effect may be most pronounced at later times after infection (42, 64). Also, given the potential redundancy in SopB and SptP function, this discrepancy needs to be addressed using a ΔsptP background. Lastly, VAMP8 is recruited to Salmonella-induced membrane ruffles where it promotes invasion, likely by providing membranes from endocytic compartments (60). However, the t-SNARE (target membrane SNARE) target of VAMP8 during Salmonella invasion was unknown. Here, we report SopB-dependent phosphorylation of syntaxin 7 at two sites, and phosphorylation of these sites promotes complex formation with VAMP8 (54). Thus, syntaxin 7 may be the t-SNARE target of VAMP8 during formation of the Salmonella-containing vacuole.

Fig. 6

Activity flow diagram of phosphorylation-based signaling cascades initiated by T3SS-1 effectors. Salmonella effectors are shown in blue, and proteins phosphorylated during infection are shown in magenta. Biological activities, typically labeled with the gene name, are shown in boxes, and the influences of each biological activity are indicated by arcs (lines) directed at subsequent biological activities or at phenotypes (hexagonal boxes). Positive influences are indicated by arrowheads on an arc, negative influences by a short perpendicular line at the end of an arc, and unknown influence by a diamond shape at the end of an arc (SspH1, for example). A necessary influence (one required for the subsequent biological activity) is indicated by a perpendicular line followed by an arrowhead (IL-8, for example). Modifiers added to arcs include Boolean terms and the Greek letter tau, which indicates that the biological activity is delayed. This diagram is designed to conform to the systems biology graphic notation (80).

Epithelial cells typically decrease innate immune signaling to avoid uncontrolled inflammation, but Salmonella effectors SopB and SopE/E2 initiate innate signaling pathways themselves, independent of TLRs and NLRs (7). Thus, Salmonella-induced signaling events in HeLa cells are more likely to be the result of the bacteria’s actions rather than undesirable side effects of its invasion, and the proteins with altered phosphorylation detected here may be potential host targets of Salmonella (Table 2). In addition to our initial large-scale phosphoproteomics studies with wild-type bacteria, we validated more than a third of these with a ΔsopB strain. In addition to demonstrating the validity of the initial data set, the results with the mutant strain suggest that SopB targets the kinases Akt and RSK1, thus extending the functions of this effector protein beyond preventing apoptosis.

The phosphoproteomics analysis presented here shows how signaling in host cells responds to a bacterial pathogen. Low biological variability (XCV <15%) and rigorous statistical limits mean that the data presented here are a reliable resource for generating new testable hypotheses about the actions of Salmonella. Indeed, the combination of these data with refined bioinformatics predictions of kinase substrates led us to identify an unanticipated role for the kinase Pim as a major player in Salmonella-induced host cell signaling. Furthermore, by using a ΔsopB strain to attribute specific events to a particular effector, we have established connections between previously orphaned observations of the impact of Salmonella on host cells. These data provide a systems-level view of a signaling system other than canonical receptor-mediated cascades, and, given the importance of bacterial-induced inflammation, they are a valuable resource for understanding host-pathogen interactions.

Materials and Methods

Cell culture and Salmonella infection

HeLa cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) containing glucose (4500 mg/liter) and 4 mM l-glutamine (Thermo Fisher Scientific) supplemented with 10% (v/v) qualified fetal bovine serum (FBS) (Invitrogen), an additional 2 mM l-glutamine (Thermo Fisher Scientific), and penicillin and streptomycin (0.1 U/liter; Thermo Fisher Scientific). For SILAC labeling, cells were split from normal growth media into arginine- and lysine-free DMEM (Caisson Laboratories Inc.) supplemented with 10% (v/v) dialyzed FBS (Invitrogen), penicillin and streptomycin (0.1 U/liter), and l-lysine (36.5 mg/liter) and l-arginine (21 mg/liter) for light-labeled cells (Sigma-Aldrich), D4-lysine (37.5 mg/liter) and 13C6-arginine (21.75 mg/liter) for medium-labeled cells, and 13C615N2-lysine (38.5 mg/liter) and 13C615N4-arginine (22.25 mg/liter) for heavy-labeled cells (Cambridge Isotope Laboratories). Cells were then maintained in labeling medium for at least six cell divisions, as previously described (65). Eight 15-cm culture dishes of each SILAC label were seeded, each with 5 × 106 cells, about 40 hours before infection.

Twenty hours before infection, SILAC DMEM was removed, cells were washed twice with 10 ml of phosphate-buffered saline (PBS), and then 16 ml of serum-free and antibiotic-free SILAC DMEM was added to each plate. An overnight Salmonella culture [wild-type S. Typhimurium SL1344 or S. Typhimurium SL1344 ΔsopB (31)] was subcultured (1:33) in LB broth for 3 hours. The bacterial inoculum was prepared by pelleting bacteria at 10,000 relative centrifugal force (RCF) for 2 min at room temperature. Bacterial pellets were resuspended in serum-free, antibiotic-free DMEM, and HeLa cells were infected at a multiplicity of infection (MOI) of 200 and incubated at 37°C and 5% CO2 (light-labeled cells were mock-infected in medium only).

At the desired infection times, cells were removed from the incubator and immediately placed on ice. The medium was removed and plates were washed three times with 10 ml of cold PBS and harvested with a scraper. Cells carrying the same SILAC label were pooled, pelleted for 4 min at 600 RCF at 4°C, and resuspended in 3.2 ml of cold lysis buffer [1 mM Na3VO4, 10 mM NaF, 5 mM Na4P2O7, 20 mM β-glycerophosphate, 0.5 mM pervanadate, 100 μM deltamethrin, 100 nM calyculin A, and Roche complete protease inhibitor cocktail (Roche Diagnostics) in PBS]. Cells were lysed on ice by five or six passages through a 22-gauge needle, and nuclei were pelleted by centrifugation for 4 min at 600 RCF at 4°C. The supernatant was retained, and the pellet (nuclear fraction) was resuspended in 1.6 ml of digestion buffer [1% (w/v) sodium deoxycholate/50 mM NH4HCO3] and immediately heated at 99°C for 5 min. Membranes were pelleted from the supernatant for 30 min at 16,000 RCF at 4°C and the supernatant was retained. The pellet and supernatant from this step were mixed with 800 μl and 1.6 ml of digestion buffer, respectively, and both were heated at 99°C for 5 min. Samples were removed and cooled to room temperature, MgCl2 was added to a final concentration of 1.5 mM, and benzonase (2.5 × 10−3 U/μl; Novagen) was added to cleave DNA and decrease viscosity. Protein concentrations were determined for each sample with a standard Bradford assay (Pierce). Eight milligrams of protein from each nuclear fraction was combined, 8 mg from each cytosolic fraction was combined, and 2 mg from each membrane fractions was combined.

Sample generation for phosphoproteomics

Samples were subjected to tryptic digest, desalting, in-solution IEF, and phosphopeptide enrichment conditions as previously described (24).

Mass spectrometry

For LC–ESI (electrospray ionization) MS, samples were analyzed on an LTQ-OrbitrapXL (Thermo Fisher Scientific) coupled online to an Agilent 1100 Series nanoflow high-performance liquid chromatography (HPLC) instrument with a nanospray ionization source (Proxeon Biosystems) as previously described (66).

LC-MS/MS data analysis and clustering

LC-MS/MS data were analyzed with MaxQuant software (v1.0.1.13) as previously described, except that only N-terminal acetylation and phosphorylation of serine, threonine, and tyrosine residues were selected as variable modifications (67). Data were searched against a combined database containing both forward and reversed protein sequence to determine false discovery rate, and P values were generated by MaxQuant on the basis of the biological variation between all samples analyzed. At this stage, all peptides that showed a statistically significant (P < 0.01) change in at least one time point compared to the uninfected control (at zero time) were considered “regulated” by Salmonella. For global analyses, all identified phosphopeptides with distinct phosphosites or phosphosite probabilities were considered separate peptides. In cases where specific peptides are reported, assignments were made to the site(s) with the highest probability reported by MaxQuant. All identified phosphosites from this study are publicly available at, along with analyses of their evolutionary conservation in 20 other species and predictions of the kinases that most likely target these phosphosites. FCM clustering was done with the MFuzz toolbox with c (number of clusters) and m (fuzzification parameter) values of 6 and 2, respectively, applied after the software normalized all data within each time point to have a mean of 0 and SD of 1 (68). Clustering for GO analysis was done with the DAVID Functional Annotation Tool (69, 70). Common motifs were searched with Motif-X, and default settings were used except that MS/MS was selected as a foreground format, phosphorylated serine was selected as the central character, and International Protein Index Human Proteome was selected as the background data set (71). Predictions of upstream kinases were performed with the Kinase Predictor algorithm in PhosphoNET (72).

High-throughput Western blotting

Custom Kinetworks KCPS multi-immunoblotting analyses were performed as described previously ( (73) with 300 μg of detergent-solubilized HeLa lysates prepared as above. The Kinetworks analysis involves resolution of proteins in a single lysate sample by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and subsequent immunoblotting overnight at 4°C with one to three primary phosphosite-specific antibodies per channel in a 20-lane Immunetics multiblotter. Phosphosite antibodies were typically sourced from Invitrogen, Cell Signaling Technologies, and Millipore. The antibody mixtures were carefully selected to avoid overlapping cross-reactivity with target proteins. The membranes were later rinsed with TBST buffer [50 mM tris base, 150 mM NaCl, and 0.5% (v/v) Triton X-100 (pH 7.4)] and then incubated with the relevant horseradish peroxidase–conjugated secondary antibodies for 45 min at room temperature. The immunoblots were developed with ECL Plus reagent (Amersham), and signals were captured by a Fluor-S MultiImager and quantified with Quantity One software (Bio-Rad). Background was fewer than 100 counts per minute for these analyses.

IL-8 enzyme-linked immunosorbent assay

Infections for IL-8 enzyme-linked immunosorbent assay (ELISA) were carried out in 24-well plates, with HeLa cells seeded at 1 × 105 cells per well the afternoon of the day before infection. Three hours before infection, the cells were serum-starved, and serum-free DMEM was used at all subsequent steps of the experiment. Inhibitors were added to the cells 30 min before the cells were inoculated with bacteria and prepared as described above, at an MOI of 100, and the inhibitor treatment was continued for the duration of the infection. At 30 min after inoculation, the cells were washed four times with PBS followed by the addition of serum-free DMEM containing gentamicin (100 μg/ml) to kill extracellular bacteria. Culture supernatants were collected at 4 hours after inoculation and centrifuged for 2 min at 10,000 RCF to pellet any cellular material, and the supernatant was aliquoted and stored at −80°C until IL-8 concentrations were assayed.

Inhibitors were dissolved in dimethyl sulfoxide (DMSO), and cells treated with DMSO served as the control. MK 2206 and SGI 1776 (Selleck) were used at 1 and 3 μM, respectively; PD 98059 and chelerythrine (EMD Chemicals) were used at 50 and 2 μM, respectively; and Pim-1 inhibitor 2 (Tocris Bioscience) was used at 100 μM. The IL-8 ELISA used to determine the concentration of IL-8 in culture supernatants was supplied in kit form (BD) and was performed according to the provided instructions. All infections were performed at least three times and assayed independently, and statistical significance was assessed with a one-way analysis of variance (ANOVA) and a Bonferroni post hoc test.

Cell viability was assayed according to the manufacturer’s instructions with an In Vitro Toxicology Assay Kit, XTT based (Sigma). Cells were treated as described above except that DMEM without phenol red was used and the XTT stock solution was added to cells after the initial 30-min inoculation. Culture supernatants were then assayed for cell viability according to the manufacturer’s instructions.

For siRNA treatment, HeLa cells were seeded at 2.5 × 104 cells per well in 24-well plates the afternoon before transfection of the siRNA pools. Cells were transfected with 300 ng per well of control nontargeting siRNA pool (D-001810-10-05) or pools targeting Pim-1 (L-003923-00-0005), Pim-2 (L-005359-00-0005), or Pim-3 (L-032287-00-0005) with Dharmafect1 according to the manufacturer’s instructions (Dharmacon). The medium was changed 24 hours after transfection. At 48 hours after transfection, cells were infected with S. Typhimurium for 4 hours, and IL-8 concentration in cell culture supernatants was assayed by ELISA as described above. Remaining cells were washed three times with PBS and lysed in 1× SDS sample buffer for subsequent immunoblotting with antibodies against Pim-2 (HPA000285, Sigma) and Pim-3 (AP7171a, Abjent).

Supplementary Materials

Fig. S1. Flow diagram outlining phosphoproteomics method for Salmonella infection.

Fig. S2. Sample loading controls for phosphosite-specific immunoblotting.

Table S1. Phosphopeptides detected in infected HeLa cells (Excel file).

Table S2. Matching of kinases to sites showing increased phosphorylation in Salmonella-infected HeLa cells (Excel file).

Table S3. Frequency of each kinase listed in table S2 (Excel file).

References and Notes

  1. Acknowledgments: We thank B. Finlay for the ΔsopB strain and for reading the manuscript, as well as other members of our group for discussions, advice, and technical support, particularly H. Zhang and J. Safaei for their contributions. Funding: A Canadian Institutes of Health Research (CIHR) operating grant (MOP-77688) to L.J.F. supported this work. Infrastructure used in this work was supported by the Canada Foundation for Innovation, the British Columbia (BC) Knowledge Development Fund, and the BC Proteomics Network. L.J.F. is the Canada Research Chair in Quantitative Proteomics. L.D.R. is supported by a CIHR CGS-D award and a Michael Smith Foundation for Health Research Trainee Award. Author contributions: L.D.R. and L.J.F. designed the phosphoproteomics experiments. L.D.R. conducted the phosphoproteomics experiments and analyzed the data, with assistance from Y.F. N.F.B. and L.D.R. conducted the IL-8 ELISA experiments, and N.F.B. constructed Fig. 6. Kinase predictions were done by S.P. L.D.R. wrote the initial draft of the manuscript. Competing interests: S.P. owns Kinexus Bioinformatics, a company involved in analyzing phosphorylation. Data availability: Raw MS/MS spectra acquired in this study are available at the PRoteomics IDEntification (PRIDE) database ( under experiment accession numbers 18477 and 18485.
View Abstract

Navigate This Article