Protective Role of Raf-1 in Salmonella-Induced Macrophage Apoptosis

Salmonellae are facultative intracellular pathogens that cause a variety of enteric diseases, ranging from self-limiting gastroenteritis (mainly due to Salmonella typhimurium) to the more severe systemic typhoid fever (caused by Salmonella typhi). After consumption of contaminated food or water, Salmonellae reach the intestine and adhere to specialized epithelial cells (M cells). Salmonella exploits the host signal transduction cascades to induce the formation of membrane ruffles at the contact point between the bacterium and host cell, and is ultimately taken up in large vacuoles 1. Once the infected M cells are destroyed, the bacteria reach the mesenteric lymph follicles and are confronted by the host's macrophages.

For Salmonella, and for many other facultative intracellular pathogens, surviving this encounter is the key to a successful infection. Invasive Salmonella can persist within the macrophages in spacious vacuoles, which do not acquire lysosomal markers and may represent a relatively safe intracellular site where the bacteria can survive and multiply 2,3,4. Furthermore, invasive Salmonella induces phagocyte apoptosis in vitro 5,6,7,8 and in vivo 9. Apoptosis is mediated by a cell-intrinsic suicide program, the activation of which is regulated by different signals originating from both the intracellular and extracellular milieu. Salmonella shares the ability of inducing macrophage apoptosis with Yersinia 10,11 and Shigella 12, suggesting that this may represent a hallmark of, and perhaps a selective advantage for, the establishment of enterobacterial infections.

Both epithelial cell invasion and the induction of macrophage apoptosis depend on a functional type III secretion system. Type III secretion systems translocate bacterial proteins directly into host cells and play a pivotal role in the interaction between pathogenic bacteria and their hosts 13. The type III secretion genes essential for epithelial cell invasion and apoptosis induction are clustered at centisome 63 of the Salmonella chromosome, in a region denominated Salmonella pathogenicity island (SPI)-1. The SPI-1–encoded bacterial proteins interact directly with eukaryotic signal transducers to activate signaling pathways of host epithelial cells 14. A functional type III secretion system is a prerequisite for the activation of the mitogen-activated protein kinase (MAPK) subgroups extracellular signal–regulated kinase (ERK), c-Jun NH₂-terminal kinase (JNK), and p38, and for the production of proinflammatory cytokines by epithelial cells infected with Salmonella 15.

The molecular mechanisms operating during the interaction of S. typhimurium with macrophages are less well characterized than those accompanying epithelial cell invasion. The SPI-1–encoded protein SipB has been shown to bind to and activates caspase-1, thereby causing apoptosis 16. Furthermore, we have previously addressed the question of Salmonella-mediated ERK 17 and JNK stimulation 18. In both cases, the mechanisms of activation used by the pathogen differed radically from those operating in epithelial cells and were not dependent on the function of the SPI-1–encoded type III secretion system 17,18.

The Raf-1 kinase plays a key role in relaying proliferation signals but has been also implicated in inflammatory signaling induced by LPS 19,20. Furthermore, depending on the cell type and the stimulus used, Raf-1 can exert a proapoptotic 21,22,23 or an antiapoptotic function 24,25,26,27,28,29. In this study, we show that Raf-1 is activated and degraded during infection with apoptosis-inducing Salmonella. Raf-1 degradation is a consequence of apoptosis induction, but kinase activation is not. Remarkably, Raf-1–deficient macrophages are hypersensitive towards pathogen-induced apoptosis, and this hypersensitivity correlates with enhanced caspase-1 activation. These data imply that Raf-1 activation upon infection with invasive Salmonella is part of the defensive response of the cells to the pathogen, and demonstrate for the first time that the Raf-1 kinase plays a role in antagonizing caspase-1 activation during pathogen-induced apoptosis.

S. typhimurium strains SR11, SL1344, LT2 (wild-type [wt]), SB111 (invA⁻), and SB169 (sipB⁻) were grown in 5 ml Luria-Bertani broth (1% Bacto Tryptone, 0.5% yeast extract, and 1% sodium chloride) at 37°C overnight for 16–20 h under agitation. To obtain highly invasive bacteria, overnight cultures were diluted to an OD₆₀₀ of 0.02 in 50 ml TYP broth (1.6% Bacto Tryptone, 1.6% yeast extract, 0.5% sodium chloride, 0.25% dipotassium phosphate) and incubated for 5 h under agitation 8.

A genomic DNA clone corresponding to c-raf-1 was isolated from a 129/Sv mouse genomic λ fix library. A 8.5-kb 5′-XbaI/BglII-3′ fragment containing exon 3 and surrounding sequences was used to assemble the targeting construct in pBSIISK⁻. loxP sites were inserted as double-stranded oligonucleotides in the HindIII site 3′ of exon 3 and in the BamHI site 5′ of exon 3. The loxP oligonucleotide 3′ of exon 3 contains a PstI site that serves as a marker for the floxed (flanked by loxP site) allele. A Neo/TK cassette containing an upstream loxP site was excised from plasmid pGH1 and cloned into an XbaI-HindIII site contained in the loxP site 5′ of exon 3. E14.1 embryonic stem (ES) cells grown on γ-irradiated embryonic fibroblasts were transfected with AscI-linearized targeting vector and selected with 0.2 mg/ml G418. Homologous recombinants (c-raf-1^+/floxNeo/TK) were obtained with a frequency of 1 in 35, as detected by nested PCR and Southern blot analysis. Positive clones were transfected with a plasmid expressing the Cre recombinase 30. Cre expression led to the deletion of either the floxed exon 3 or the floxed Neo/TK cassette, or both. The latter two were enriched by negative selection with gancyclovir. Two clones harboring the floxed exon 3 (c-raf-1^flox/+) were injected into C57BL/6 blastocyst-stage embryos and transferred to pseudopregnant B/6CBAF1 mice for further development. Chimeric mice were mated to C57BL/6 animals and agouti offspring were genotyped. Germline transmission of the floxed allele was detected either by Southern blot or PCR analysis of tail DNA. c-raf-1^flox/flox mice were mated to mice expressing Cre under the control of the inducible Mx1 promoter 31.

Bone marrow–derived macrophages were isolated from Mx-Cre; Raf^flox/flox mice and Raf^flox/flox mice after induction of Cre expression by poly inosinic:cytidyllic acid (poly I:C) treatment in vivo (400 μg intraperitoneally, every other day, three injections total) or from caspase-1–deficient 32 and wt C57BL/6 mice. 2 d after the last injection, the bone marrow cells were collected and cultured in Dulbecco's modified Eagle's medium supplemented with 10% FCS and 20% L-conditioned medium as a source of CSF-1 for 1 wk. Confluent cells (∼5 × 10⁶ cells per 100-mm-diameter tissue culture dish) were cultured for 16–20 h in medium without CSF-1 and then infected with bacterial cultures. A multiplicity of infection (moi; bacteria per macrophage) of 25 was used unless indicated otherwise. MAPK or ERK kinase (MEK) activation was inhibited by overnight pretreatment with 50 μM PD098059 (Calbiochem). Proteasome function was inhibited by pretreatment with 10 μM MG-115 and MG-132 for 90 min (Calbiochem).

Cells from one 100-mm-diameter cell culture dish were washed twice with PBS and lysed in 300 μl solubilization buffer (10 mM Tris-HCl, pH 7.0, 50 mM sodium chloride, 30 mM sodium pyrophosphate, 1% Triton X-100). Insoluble material was removed by centrifugation at 20,000 g for 30 min before immunoprecipitation. A rabbit polyclonal antiserum raised against a COOH-terminal peptide of v-raf (SP63, CTLTTSPRLPVF) was used to immunoprecipitate Raf-1. Immunocomplexes were collected after incubation with Protein A–Sepharose beads (Sigma-Aldrich). For Western blot analysis, cell lysates (30 μg/lane) were separated by SDS-PAGE and transferred onto nitrocellulose membranes. After 1 h blocking in TTBS (10 mM Tris-HCl, pH 8.0, 150 mM sodium chloride, 0.1% Tween 20) supplemented with 5% milk powder, the membranes were probed with the appropriate primary antibodies (CSF-1R, rabbit serum generated against GST-CSF-1R [1-313] fusion protein; actin, caspase-1, -2, -3, and IκB, all from Santa Cruz Biotechnology, Inc.; caspase-8, Chemicon; cytochrome-c, BD PharMingen; cytochrome-c-oxidase-subunit IV, Molecular Probes; MEK-1 and panErk, Transduction Laboratories; pMAPK, New England Biolabs, Inc.; Raf-1 [NH₂-terminal], Transduction Laboratories; and Raf-1 kinase domain 33 diluted in 1% BSA [fraction V, Sigma-Aldrich]) in TTBS before incubation with peroxidase-conjugated secondary antibodies and detection by an enhanced chemiluminescence system (Pierce Chemical Co.).

Nuclear extracts were prepared as described previously 34. In brief, 2 × 10⁶ cells, either untreated or infected with S. typhimurium, were washed twice with PBS and resuspended in 400 μl of buffer A (10 mM Hepes, pH 7.8, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM dithiothreitol, and 1 mM PMSF). After incubation on ice for 5 min, NP-40 was added to a final concentration of 0.6%. Nuclei were pelleted and the cytoplasmic proteins were carefully removed. The nuclei were then resuspended in buffer C (20 mM Hepes, pH 7.9, 0.4 M NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol, and 1 mM PMSF). After 30 min at 4°C, the samples were centrifuged and the nuclear proteins in the supernatant were transferred to a fresh vial. 10 μg of nuclear extract was incubated with an end-labeled, double-stranded NF-κB–specific oligonucleotide probe (5-AATTCGGCTTGGAAATTCCCCGAGCG-3). As specificity controls, extracts were incubated with unlabeled wt or mutated (5-AGCTTAGATTTTACTTTCCGAGAGGA-3) probe before the addition of labeled oligo. The binding reaction was performed in a total of 20 μl of binding buffer (5 mM Hepes, pH 7.9, 50 mM KCl, 0.5 mM dithiothreitol, 1 μg poly [dI:dC], and 10% glycerol) for 20 min at room temperature (RT). After incubation, samples were fractionated on a 5% polyacrylamide gel and visualized by autoradiography.

Cells were scraped in Mito buffer (250 mM sucrose, 20 mM Hepes, 10 mM KCl, 1.5 mM MgCl₂, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol, 10 μM PMSF, 1× protease inhibitor cocktail [Boehringer]). After incubation on ice for 30 min, cells were disrupted at 4°C in a 1-ml syringe fitted with a 25G hypodermic needle (15 strokes). Nuclei and unbroken cells were removed by centrifugation at 700 g for 5 min at 4°C. Supernatants were then further centrifuged at 13,000 g for 20 min at 4°C. The resulting pellet was defined as the heavy membrane fraction (HMF).

3 × 10⁵ macrophages were seeded on a coverslip in a well of a 6-well cell culture dish. Chromatin condensation in infected macrophages was determined by staining with 0.5 μg/ml 4′,6-diamidino-2-phenylindole (DAPI; Sigma-Aldrich) for 1 min. Cells were washed twice with PBS and fixed with 3% formaldehyde in PBS (10 min at RT). Coverslips were mounted in 20% Mowiol (Sigma-Aldrich) in PBS. Chromatin condensation was assessed in randomly chosen areas of the sample by independent experimenters (300–500 cells/sample).

3 × 10⁵ macrophages were seeded on a coverslip in a well of a 6-well cell culture dish. Mitochondria were stained with 50 nM Chloromethyl-X-Rosamine (CM-X-ROS, Mitotracker red; Molecular Probes) for 10 min, a potential-sensitive fluorochrome that withstands fixation and permeabilization of cells 35. After infection, cells were permeabilized with 0.01% Triton X-100 (Pierce Chemical Co.) for 1 min and fixed in 4% paraformaldehyde. Fixed cells were washed with quenching/washing solution (Q/W: 50 mM NH₄Cl, 10 mM Pipes, pH 6.8, 150 mM NaCl, 5 mM EGTA, 5 mM glucose, 5 mM MgCl₂) and incubated with primary Raf-1 antibodies (anti-SP63, anti–NH₂-terminal, and anti-kinase domain, respectively) in blocking solution (2% gelatin, 0.3% Triton X-100 in Q/W) for 60 min at RT. Cells were then washed in Q/W solution and incubated with fluorescein-conjugated secondary antibody (Alexa 488) in blocking solution for 60 min at RT. Cells were washed with Q/W solution, mounted (ProLong Antifade kit; Molecular Probes), and examined by confocal microscopy.

Raf-1 kinase activity was measured as the ability of immunoisolated Raf-1 to activate recombinant MEK-1 in coupled assay using myelin basic protein as the endpoint of the assay 36.

A functional SPI-1–encoded type III secretion system is essential for the ability of Salmonella to induce apoptosis. Salmonella grown to stationary phase 8, invA⁻ (lacking an essential membrane component of the type III secretion apparatus) and sipB⁻ mutants are incapable of doing so (Fig. 1 A). SipB causes apoptosis by binding to caspase-1 16. Consistently, primary bone marrow–derived macrophages from caspase-1–deficient mice fail to undergo apoptosis within the 30 min of infection with wt Salmonella (Fig. 1 A).

To explore the connection between Raf-1 activation and apoptosis induction, we infected primary macrophages with invasive and noninvasive Salmonellae and compared their ability to activate Raf-1. Infection of quiescent primary bone marrow–derived macrophages with wt Salmonella in the late logarithmic phase of growth (invasive) caused moderate Raf-1 activation. However, wt bacteria in the stationary phase (noninvasive) were incapable to do so. Several different wt strains (SR11, SL1344, and LT2) behaved in identical manner (data not shown). Consistent with an involvement of the SPI-1–encoded type III secretion system in Salmonella-induced Raf-1 activation, the invA⁻ bacteria in the late logarithmic phase did not stimulate the Raf-1 kinase, and caused a slight, but reproducible decrease in its basal activity (Fig. 1 B). In contrast, the sipB⁻ mutant activated Raf-1 with identical kinetics and to identical extents as wt Salmonella (Fig. 1 C). The amount of immunoprecipitated Raf-1 decreased during the late phase of infection with invasive wt, but not with noninvasive wt or with the invA⁻ or sipB⁻ strain. From these results, we can conclude that SipB is not required for early Raf-1 activation, and that the decline in Raf-1 activation 20 min after infection is not attributable to the degradation of the protein. Therefore, Raf-1 activation is not a consequence of apoptosis.

Next we investigated if the decrease in the amount of Raf-1 observed during the late phases of infection was dependent on caspase-1, and whether it was a reflection of the general demise of the cell undergoing apoptosis or rather a specific phenomenon involving this kinase selectively. Triton X-100 extracts from control (+/+) or caspase-1–deficient macrophages infected with Salmonella were immunoblotted with a Raf-1 antiserum. The amount of kinase was progressively reduced in +/+ macrophages starting from 5 min after infection with invasive Salmonella, but remained constant throughout infection with a sipB⁻ mutant (Fig. 2 A) or of caspase-1–deficient macrophages with wt bacteria (Fig. 2 B). In contrast to Raf-1, the amount of MEK (data not shown) or ERK remained constant throughout infection. Thus, Salmonella induced caspase-1–dependent degradation of Raf-1, but not of MEK or ERK (Fig. 2A and Fig. B).

Loss of Raf-1 from the Triton X-100 soluble fraction was paralleled initially by an increase in Triton X-100 insoluble Raf-1 (Fig. 2 C). Accumulation in the Triton X-100 insoluble fraction is a general feature associated with protein ubiquitinylation, which normally precedes proteasome-mediated degradation. To gain some insight into the mechanism of Raf-1 degradation, we treated macrophages with the proteasome inhibitors MG-115 and MG-132 before infection with invasive Salmonella. Both inhibitors efficiently stabilized Raf-1, as did the caspase inhibitor Z-VAD-fmk. The proteasome inhibitors, but not the caspase inhibitor, caused the appearance of higher molecular forms of Raf-1, particularly in the Triton X-100 insoluble fraction of lysates, probably because of Raf-1 ubiquitinylation (Fig. 2 D). None of the inhibitors had any effect on the amount or solubility of Raf-1 in uninfected cells.

Disruption of the c-raf-1 gene is embryonic lethal at midgestation and is accompanied by fetal liver apoptosis (unpublished observations). To obtain Raf-1–deficient macrophages, conditional inactivation of the c-raf-1 gene was achieved by the insertion of loxP sites cloned 5′ and 3′ of exon 3. A selection cassette (a neomycin resistance gene for positive selection and the thymidine kinase gene of herpes virus for negative selection) was positioned between two loxP sites upstream of the floxed exon 3 (Fig. 3 A). The mutation was introduced into ES cells by homologous recombination. After transient Cre expression, ES cell clones in which the Neo/TK gene cassette, but not exon 3, had been excised were identified by Southern blot analysis (Fig. 3 B). Germline-transmitting chimeras were obtained and bred to C57Bl/6 mice. Mice homozygous for the c-raf-1^flox allele were phenotypically indistinguishable from wt animals. To allow inducible inactivation of Raf-1, c-raf-1^flox/flox animals were crossed to mice expressing the Cre recombinase under the control of the Mx1 promoter 31. Injection of mice with poly I:C resulted in the efficient deletion of the floxed exon 3 in liver (data not shown) and bone marrow (Fig. 3 C, c-raf-1^Δ/Δ). Macrophages derived from these bone marrow cells were devoid of Raf-1 protein, as shown by Western blot analysis, while poly I:C treatment of c-raf-1^flox/flox animals that did not carry the Mx-Cre transgene did not have any effect on Raf-1 expression (Fig. 3 D).

To determine whether activated Raf-1 functions as a pro- or antiapoptotic molecule in the context of Salmonella-induced cell death, we infected c-raf-1^Δ/Δ and c-raf-1^flox/floxmacrophages with different amounts of bacteria and compared the number of cells undergoing apoptosis. The c-raf-1^Δ/Δ macrophages proved more sensitive to pathogen-mediated apoptosis than c-raf-1^flox/flox cells at all moi investigated (Fig. 4A and Fig. B). In addition, apoptosis proceeded with faster kinetics in c-raf-1^Δ/Δ macrophages than in c-raf-1^flox/flox cells (Fig. 4 C). However, infection with wt Salmonella in stationary phase, with invA⁻ or with sipB⁻ bacteria, failed to induce apoptosis in c-raf-1^Δ/Δ macrophages, and these cells were not more susceptible than c-raf-1^flox/flox cells to cell death induced by Listeria monocytogenes, a process resembling delayed necrosis (reference 37, data not shown). Thus, Raf-1 plays a specific protective role in Salmonella-induced macrophage apoptosis.

Unlike Raf-1, ERK was activated equally well and with identical kinetics upon infection with invasive Salmonella and with an invA⁻ mutant (Fig. 5 A). This suggests that Raf-1 is not involved in the activation of the ERK cascade by Salmonella. However, it is still formally possible that wt and invasion-deficient bacteria use distinct signal transduction pathways to stimulate ERKs, and that Raf-1 is involved in ERK activation by invasive bacteria selectively. To investigate this, we monitored ERK activation in c-raf-1^flox/flox and c-raf-1^Δ/Δ bone marrow–derived macrophages infected with wt or invA⁻ Salmonella. ERK activation occurred normally in the Raf-1–deficient cells (Fig. 5 B). These results show that Raf-1 is not essential for ERK activation in Salmonella-infected macrophages. Consequently, the protective effect of Raf-1 against pathogen-mediated apoptosis is not mediated via this pathway.

Antiapoptotic effects ascribed directly or indirectly to ERK activation have been described 38,39. To ascertain whether or not the ERK cascade had any protective effect on pathogen-mediated apoptosis independently of Raf-1, we treated c-raf-1^flox/flox macrophages with the MEK-1 inhibitor PD98059 before infection with invasive bacteria. Although it successfully abrogated ERK activation by Salmonella (Fig. 5 C), the MEK inhibitor had no effect on pathogen-mediated macrophage apoptosis (Fig. 5 D).

A further downstream target of Raf-1 implicated in protection from apoptosis is the transcription factor NF-κB 40. Raf-1 activates NF-κB by inducing IκB phosphorylation and degradation. This pathway is distinct from MEK and ERK activation but involves MEKK-1 upstream of the IκB kinase complex 41. Infection of macrophages with invasive Salmonella caused rapid IκB degradation, whose extent and kinetics were identical in c-raf-1^flox/flox and c-raf-1^Δ/Δ macrophages (Fig. 6 A). In addition, infection with Salmonella resulted in the rapid stimulation of NF-κB binding activity in nuclear extracts from cells of both genotypes (Fig. 6 B).

Thus, Raf-1 is not essential for phosphorylation-induced IκB degradation or NF-κB binding in Salmonella-infected macrophages.

The Raf kinases have been proposed to modulate mitochondrial integrity by regulating the activity of Bcl-2 family members 26,27,29. Stimulus-induced translocation of the Raf-1 kinase to the mitochondrial compartment, which would bring it in the proximity of the putative substrates, has been postulated based on biochemical fractionation experiments 27,42. We addressed the question of mitochondrial translocation of Raf-1 in Salmonella-infected macrophages by biochemical fractionation and confocal microscopy. Biochemical fractionation showed that a significant portion (5–10% in different experiments) of Raf-1 could be recovered from the HMF of macrophages, commonly referred to as “mitochondria-enriched.” In addition, the amount of Raf-1 in this fraction increased after infection with invasive Salmonella (Fig. 7 A). The presence of Raf-1 in the HMF has been previously taken as an indication of the mitochondrial localization of this protein. This fraction contained mitochondrial proteins such as cox-IV and cytochrome c, and was completely devoid of cytosolic proteins such as caspase-3. However, it also contained the glycosylated form of the CSF-1 receptor, and was therefore contaminated with plasma membrane proteins (possibly from portions of the membrane associated with the cytoskeleton). Thus, in spite of the enrichment for mitochondria, the HMF obtained by this method is not pure (Fig. 7 B). We used confocal microscopy to confirm or dispute Raf-1 mitochondrial localization by an independent method. Confocal analysis of macrophages stained with a Raf-1 antiserum showed a punctuate Raf-1 pattern, but did not reveal any colocalization with living mitochondria (visualized with Mitotracker red). Treatment with invasive Salmonella for 10 (data not shown) and 15 min did not alter Raf-1 localization (Fig. 7 C). The same results were obtained with antibodies directed against three different domains of the molecule. Raf-1 knockout macrophages were used to control for specificity.

Failure to detect mitochondrial localization of Raf-1 might be because of technical reasons; alternatively, localization might be extremely transient and therefore elusive. In an attempt to assess whether mitochondrial function was altered in knockout cells, we assayed cytochrome c release from wt and Raf-1–deficient macrophages. Rapid Salmonella-mediated apoptosis did not lead to any appreciable cytochrome c release in c-raf-1^flox/flox or in c-raf-1^Δ/Δ macrophages (data not shown). Thus, the apoptotic phenotype of Raf-1–deficient macrophages cannot be ascribed to mitochondrial fragility.

To investigate whether the ablation of Raf-1 had any effect on caspase-1, c-raf-1^flox/flox and c-raf-1^Δ/Δ macrophages were infected with invasive Salmonella. Caspase-1 activation was monitored by immunoblotting with an antiserum that recognizes the long subunit of the active enzyme (p20). p20 appeared with faster kinetics and in larger amounts in Raf-1–deficient macrophages than in c-raf-1^flox/flox controls. In contrast, the kinetics and strength of caspase-2 activation were indistinguishable. Cleavage (and therefore activation) of the zymogens procaspase-3 and -8 did not occur during rapid Salmonella-induced apoptosis in c-raf-1^flox/flox or c-raf-1^Δ/Δ macrophages (Fig. 8).

The interaction of Salmonella with the host's macrophages is an important event in the early phases of infection. Still, the signaling steps taking place during this interaction are largely unknown. In this paper, we show that Raf-1 is the only kinase of the MAPK pathway to be selectively activated and degraded by apoptosis-inducing Salmonella. Furthermore, conditional inactivation of Raf-1 in macrophages showed that these cells are more sensitive towards Salmonella-mediated apoptosis, implying that Raf-1 has a protective function in this process. This hypersensitivity correlates with, and is probably attributable to, an increase in caspase-1 activation.

Macrophages activate Raf-1 in response to several different stimuli, notably LPS 19,20,43. It was therefore somewhat surprising that infection with invA⁻ Salmonella actually decreased Raf-1 activity. Distinct reactions to soluble and particulate LPS have been observed before; notably, soluble LPS is more efficient in eliciting cellular responses 44. In the specific context of MAPK cascades, our own work has shown that LPS is likely to be the major determinant of ERK activation 17. In contrast, Salmonella stimulates JNKs by a mechanism distinct from LPS, but in this case the type III secretion apparatus is not involved 18,45. Thus, Raf-1 is the first example of a signal transducer whose activation by Salmonella in macrophages requires the type III system. Interestingly, a sipB mutant was still able to cause Raf-1 activation. Consistent with this, activation of Raf-1 is normal in caspase-1–deficient macrophages that lack the only target of SipB identified so far in macrophages (data not shown). sipB mutants are noninvasive 46. They are able to secrete proteins in culture media 46, but cannot transfer them into cultured epithelial cells 47. Therefore, SipB may function as a translocase. If SipB does perform this function in macrophages, it is possible that Raf-1 activation does not need translocation of a Salmonella protein into the cell; activation may be engendered either by the recognition of a structural component of the secretion system (needle complex) or by a protein secreted into the medium. Alternatively, a yet unidentified protein whose translocation does not require SipB might be responsible for Raf-1 activation.

The sipB⁻ Salmonella mutants and the caspase-1–deficient macrophages allowed us to discriminate among three events that follow macrophage infection by Salmonella: Raf-1 activation, which is not affected by these mutations; and Raf-1 degradation and apoptosis, which require both SipB and caspase-1. Thus, it seems plausible that Raf-1 stimulation is a reaction of the cell to the encounter with potentially pathogenic Salmonella. Kinase activation is not an emergency response to apoptotic stress, since it occurs normally in sipB⁻ infected macrophages, which do not undergo apoptosis. In contrast, Raf-1 degradation is a feature of apoptosis. Although Raf-1 is not cleaved in the absence of caspase activation, we failed to observe degradation products that would be diagnostic of a caspase-mediated cleavage. Besides the caspases, several other serine proteases have been implicated in the apoptotic process 48. In particular, the proteasome reportedly acts downstream of caspases 49,50,51. Among the targets of the proteasome are the transcription factor c-Fos 52, the antiapoptotic protein Bcl-2 53, the proapoptotic protein Bid 54, and the natural caspase inhibitors of apoptosis (IAPs) 51. In Salmonella-infected cells, pretreatment with proteasome inhibitors protected Raf-1 from degradation. This was paralleled by the appearance of higher molecular forms of Raf-1, indicative of ubiquitinylation. Thus, Raf-1 degradation in the course of Salmonella-mediated apoptosis appears to be mediated by the proteasome.

Several signal transducers are degraded in apoptotic cells. In some cases, the proteins are activated by caspase cleavage and contribute to the apoptotic response 55,56,57,58; in others, it has been postulated that cleavage disables proteins that would counteract apoptosis 59. To determine the role of Raf-1 in pathogen-mediated apoptosis, we performed the conditional inactivation of the c-raf-1 gene in bone marrow cells. The c-raf-1^Δ/Δ macrophages from these bone marrow cells are significantly more prone to Salmonella-mediated apoptosis than c-raf-1^flox/flox cells. Therefore, Raf-1 appears to play an antiapoptotic role in this process, and its degradation might be part of a bacterial strategy to weaken the eukaryotic cell.

The protective function of Raf-1 is not mediated via the NF-κB or the MEK/ERK pathway. In fact, Raf-1 is not essential for the activation of these pathways in this context. In the case of MEK/ERK, this finding was not totally unexpected, since we have shown before that the activation of these enzymes can be uncoupled from Raf-1 in macrophages by several criteria 43,60. Furthermore, chemical inhibition of ERK stimulation does not have any impact on Salmonella-mediated apoptosis.

The mitochondria could have represented a likely target site of Raf-1 action during Salmonella-mediated apoptosis. We initially considered the possibility that in the absence of Raf-1 an increased fragility of these organelles might accelerate apoptosis. Biochemical fractionation experiments in other systems have indicated that Raf-1 is associated with mitochondria and that further translocation to this compartment accompanies stimulation with a growth factor 27 as well as with UV 42. The cell fractionation experiments performed here suggested a similar picture in Salmonella-infected macrophages. However, in confocal microscopy we did not observe any significant colocalization between Mitotracker red (used to stain living mitochondria) and the Raf-1 signal. Consistent with this result, A-Raf, but not Raf-1, can be visualized associated with mitochondria by immunogold staining followed by transmission electron microscopy 61. In addition, c-raf-1^Δ/Δ cells did not show increased mitochondrial damage (as measured by cytochrome c release) after exposure to invasive Salmonella.

The experiments discussed above rule out MEK/ERK and NF-κB as downstream effectors, and mitochondrial integrity as the target of the antiapoptotic function of Raf-1 in Salmonella-infected macrophages. The only parameters altered in the Raf-1–deficient macrophages were the kinetics and strength of caspase-1 activation. This suggests that the kinase is somehow able to restrain the activation of this protease. Interestingly, lack of Raf-1 did not have any effect on the kinetics or extent of caspase-2 cleavage, which becomes activated slightly earlier than caspase-1 and participates in its activation 62. Cleavage of the zymogens procaspase-3 and -8, which are not activated in the course of rapid Salmonella-mediated apoptosis, did not occur in Raf-1–deficient cells. These data indicate that Raf-1 acts via a specific mechanism targeting caspase-1 selectively, and not by simply restraining caspase activation in general. This is remarkable in light of the fact that caspase-1 is needed to initiate proteasome-dependent Raf-1 degradation during apoptosis. Thus, the relationship between Raf-1 and caspase-1 is reciprocal, with the protease directing the degradation of the kinase that restrains its activation. At present, we do not know how Raf-1 controls caspase-1. It is possible that Raf-1, or one of its unknown downstream effectors, directly modifies this protease. Phosphorylation of human caspase-9 by the antiapoptotic kinase protein kinase B has been shown to inhibit protease activity 63. Alternatively, Raf-1 might directly or indirectly modulate the expression or activity of natural caspase inhibitors such as the IAPs 64. A specific caspase-1 inhibitor, ICEBERG, has been recently described 65. Also in this context, genetic evidence in Drosophila shows that activated D-Ras and D-Raf can inhibit apoptosis by antagonizing Hid 66,67, whose recently cloned human orthologues have been shown to antagonize IAPs 68,69. Raf-1–deficient cells generated by conditional gene inactivation will be an invaluable tool in future studies aimed at identifying the mechanism underlying the antiapoptotic function of Raf-1.

We thank Dr. Jorge E. Galan (Yale University, New Haven, CT) for the gift of the sipB⁻Salmonella strain. We are indebted to Lucia Kucerova and Zvenislava Husak (Baccarini laboratory) for maintaining the Mx-Cre/Raf^flox/flox colony; Karin Paiha and Lukas Huber (Research Institute of Molecular Pathology) for help with confocal microscopy; Rainer de Martin for providing the NF-κB oligos; Thomas Decker and Pavel Kovarik (Vienna Biocenter); Michael Naumann (Max-Planck-Institute for Infection Biology, Berlin, Germany); and Philippe Sansonetti (Institute Pasteur, Paris, France) for critically reading this manuscript.

This work was supported by Austrian Research Fund grants P13252-MOB and P-12279-MOB (to M. Baccarini). V. Jesenberger was the recipient of a fellowship from the Austrian Academy of Sciences.