INTRODUCTION

The enteropathogenic Gram-negative bacterium Yersinia enterocolitica has developed strategies to resist the host immune defense. This enables extracellular survival and multiplication of the bacteria in host lymphoid tissue after infection and invasion of the intestinal mucosa. It is becoming increasingly evident that Yersinia sp. evade host defense mechanisms by disrupting key functions of the host cell. This ability is linked to the expression of a set of released plasmid-encoded proteins, termed Yersinia outer proteins (Yops)¹ (1, 2). Export of Yops is triggered by attachment of Yersinia sp. to the host cell (3-5). Eleven Yops have been described so far (2). At least four of them, i.e. YopE, YopH, YopM, and YopO (the homolog of YpkA in Yersinia pseudotuberculosis), are translocated across the host cell membrane to their putative intracellular targets (3, 4, 6-10). YopE disrupts actin filaments (3, 4, 11) and acts synergistically with the protein-tyrosine phosphatase YopH (12) to inhibit phagocytosis and to suppress the oxidative burst of professional phagocytes (11, 13-16). YopH and also YopO, which displays serine/threonine kinase activity (17), share homologies with eukaryotic proteins, and both are supposed to interfere or block host cell signal transduction pathways (12, 17-20).

Y. enterocolitica, like other pathogens (Brucella sp. (21, 22), Bacillus anthracis (23), or Leishmania donovani (24)), also interferes with cytokine production. It suppresses chemokine interleukin-8 secretion of epithelial cells (25) and prevents production of the macrophage proinflammatory cytokine TNF (26-29). Released TNF enhances the activation of cells involved in the immune defense (i.e. macrophages, polymorphonuclear leukocytes, NK cells, and T lymphocytes) and thus contributes in overcoming bacterial infection. Previous studies already demonstrated that TNF also plays an important role in limiting the severity of Y. enterocolitica infection (30). However, the impact of Y. enterocolitica on signaling pathways of mammalian cells, leading to suppression of cytokine release, is still completely unknown. Since LPS itself stimulates macrophage secretion of TNF, it seems reasonable to assume that Y. enterocolitica interferes with LPS-stimulated pathways.

LPS from Gram-negative bacteria was reported to activate the three different MAPK families, i.e. ERK, JNK/SAPK, and p38, in macrophages (31-35). The mechanism of MAPK activation by LPS remains unclear (36). On the one hand, ceramide seems to play an important role (37), since LPS was shown to activate a ceramide-dependent kinase (38) and ceramide itself stimulates the JNK/SAPK pathway (39). Moreover, c-Raf, the upstream kinase activator of MEK1/2 and ERK1/2, was recently shown to be activated by ceramide (40). On the other hand, LPS, through its fatty acid chains, has structural homology with ceramide, and it was suggested that LPS stimulates a ceramide-dependent kinase by mimicking the ceramide molecule (38). However, irrespective of the mechanism of LPS-induced MAPK stimulation, it is well established that among MAPKs, p38 plays an important role in LPS-induced TNF production in macrophages (41-42).

In the present study, we analyzed possible alteration of MAPK activation during infection with Y. enterocolitica. We thus chose the macrophage-like J774A.1 cell line as a well established infection model to study Yersinia sp.-macrophage interactions (13-14, 18-19, 43). Interestingly, there is a nonvirulent Y. enterocolitica strain that is virulence plasmid-cured, thus providing an ideal control for comparison experiments with virulent wild-type or mutated Y. enterocolitica strains (Table I). Here, we report that virulent Y. enterocolitica indeed strongly interferes with macrophage signal transduction, resulting in blockade of ERK, JNK, and p38 MAPK activities. This MAPK inhibition correlates with the suppression of TNF production but is not required for the inhibition of macrophage phagocytosis and oxidative burst.

Table I. Y. enterocolitica strains used in this study Strain Relevant characteristics Former designation Reference Virulent Serogroup O8; clinical isolate harboring virulence plasmid pYVO8 WA-314 44 Nonvirulent Plasmidless derivative of the virulent strain WA-C 44 YopH(1) Mutant strain, deficient in YopH secretion; insertional inactivation of sycH, the gene for the YopH-specific chaperone SycH WA-C(pYV-7146) 16 WA-C(pYV O8::Tn7) 45 YopH(1)/H+ YopH(1) strain, complemented with sycH and yopH; YopH secretion-positive WA-C(pYV-7146, pB8-64) 16 YopH(2) YopH mutant; insertional inactivation of the yopH gene WA-C(pYV 08H) 46 Yop secr. Mutant strain, deficient in secretion of Yops; insertional inactivation of lcrD, the gene encoding LcrD, which is essential for Yop secretion WA-C(pYV-515) 16 YopD,B,N,V+ Strain harboring plasmid pLCR encoding the secretion apparatus of Y. enterocolitica including the genes for YopD, YopB, YopN, and the V antigen WA-C(pLCR) 46 YopD,B,N,V,H,E,YadA+ Strain YopD,B,N,V+ harboring an additional plasmid encoding the genes for YopH, YopE, and YadA WA-C(pLCR, pB8-23) 46

EXPERIMENTAL PROCEDURES

The bacterial strains used in this study are listed in Table I. Overnight cultures grown at 26 °C were diluted 1:20 in fresh Luria-Bertani broth and grown for 2 h at 37 °C as described previously (16). Bacteria were then washed once and resuspended in phosphate-buffered saline. Cells were infected at a ratio of 50 bacteria/cell. The desired bacterial concentration was adjusted by measuring the optical density at 600 nm and checked by plating serial dilutions from every sample on agar and counting colony-forming units after incubation at 26 °C for 20 h.

The murine macrophage-like cell line J774.A1 (ATCC TIB 67) was cultured in RPMI 1640 medium (Life Technologies, Cergy, Pontoise, France) supplemented with 10% heat-inactivated fetal calf serum and 5 mM L-glutamine at 37 °C in a humidified atmosphere (5% CO₂). Before treatment with bacteria or 10 µg/ml LPS from Escherichia coli (Sigma), cells were scraped, washed, and resuspended in complete culture medium. Cell stimulation occurred at 37 °C for different periods of time as indicated. Cell viability was more than 90% after 90 min of bacterial infection, as determined by trypan blue exclusion.

Antibodies directed against the C-terminal end (peptide KPLDQEEMES) of p38 kinase were raised in New Zealand rabbits as described previously (47). Antibodies were purified from immune sera by ammonium sulfate precipitation followed by overnight recycling through Affi-Gel 10 (Bio-Rad) to which the antigen peptide had been linked. After acidic elution, neutralized affinity-purified antibodies were dialyzed against phosphate-buffered saline and thereafter against a glycerol/phosphate-buffered saline solution (1:1) before storage at 20 °C.

5 × 10⁶ cells were treated with bacteria and/or LPS for different periods of time and lysed with radioimmune precipitation buffer (150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 50 mM Tris, pH 8.0, 20 mM -glycerophosphate, 10 mM p-nitrophenyl phosphate, 0.1 mM Na₃VO₄, 0.5 µg/ml leupeptin, 2 µg/ml aprotinin, 1 µg/ml pepstatin, 0.5 mM PMSF). For immunoprecipitation, cell lysates were incubated with polyclonal anti-p38 antibody at 4 °C for 1 h. Immune complexes were then collected with protein A-Sepharose (Pharmacia Biotech, Uppsala, Sweden) and washed three times with radioimmune precipitation buffer. Both whole cell lysates and immunoprecipitates were mixed with 4 × or 2 × Laemmli buffer, respectively. Proteins were separated by 10% SDS-PAGE, electrotransferred to PVDF membrane (Polyscreen, DuPont NEN), blocked with 3% bovine serum albumin, and probed with appropriate antibodies. Phosphotyrosine immunostaining was performed with the monoclonal antibody 4G10 (Upstate Biotechnology, Inc., Lake Placid, NY). Immunoblotting for p38 and ERK1/2 was performed using rabbit polyclonal anti-p38 antibodies (1:3000 dilution) or goat polyclonal anti-ERK1/2 antibodies (1:30,000 dilution; Santa Cruz Biotechnology, Inc., Santa Cruz, CA), respectively. Immunoreactive bands were visualized by incubation (1 h) with rabbit anti-mouse (1:10,000 dilution; Sigma), goat anti-rabbit (1:20,000 dilution; Sigma), or rabbit anti-goat (1:10,000; Santa Cruz Biotechnology) antibodies conjugated to horseradish peroxidase using enhanced chemiluminescence reagents (Renaissance; DuPont NEN). When required, membranes were stripped in 62.5 mM Tris, pH 6.7, 0.1 mM 2-mercaptoethanol, and 2% SDS for 30 min at 50 °C after film exposure. Thereafter, membranes were reprobed with appropriate primary and secondary antibodies and developed by chemiluminescence.

5 × 10⁶ cells treated with bacteria or LPS for different periods of time were lysed in 200 µl of cell extract buffer (25 mM Hepes, pH 7.7, 0.3 M NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA, 0.1% Triton X-100, 20 mM -glycerophosphate, 0.1 mM Na₃VO₄, 2 µg/ml leupeptin, 10 µg/ml benzamidin, 2 µg/ml aprotinin, 1 µg/ml pepstatin, 100 µg/ml PMSF). After centrifugation, the supernatant was diluted with 600 µl of dilution buffer (20 mM Hepes, pH 7.7, 2.5 mM MgCl₂, 0.1 mM EDTA, 0.05% Triton X-100, 20 mM -glycerophosphate, 0.1 mM Na₃VO₄, 2 µg/ml leupeptin, 10 µg/ml benzamidin, 2 µg/ml aprotinin, 1 µg/ml pepstatin, 100 µg/ml PMSF), incubated on ice for 10 min, and centrifuged again. Lysates were then mixed with GST fusion protein kinase substrates (8 µg of each, as indicated) and glutathione-agarose (20 µl, Sigma) and incubated overnight at 4 °C. Experiments were performed as described (48) with four different GST fusion protein kinase substrates obtained from M. Karin (GST-c-Jun-(1-222)), B. Dérijard (GST-c-Jun-(1-79) and GST-ATF2), and A. Nordheim (GST-Elk-1). The substrate-agarose complexes were washed four times with binding buffer (20 mM Hepes, pH 7.7, 50 mM NaCl, 25 mM MgCl₂, 0.1 mM EDTA, 0.05% Triton X-100), and in vitro phosphorylation was carried out for 20 min at 30 °C in the presence of 20 mM Hepes, pH 7.6, 20 mM MgCl₂, 2 mM dithiothreitol, 20 mM -glycerophosphate, 0.1 mM Na₃VO₄, 20 mM p-nitrophenyl phosphate, 20 µM ATP (4 µCi of [-³²P]ATP) (total volume, 30 µl). The reaction was stopped by a single wash with binding buffer and by adding 30 µl of 2 × Laemmli buffer. Proteins were fractionated by 10% SDS-PAGE, electrotransferred to PVDF membrane, and subjected to autoradiography or quantified with a PhosphorImager (Molecular Dynamics, Inc.).

Raf-1 kinase activity was assayed, after immunoprecipitation, by phosphorylation of exogenously added MEK1 and ERK2 (49). 1 × 10⁷ cells, treated with bacteria or LPS for 90 min, were lysed with lysis buffer (20 mM Tris, pH 7.5, 150 mM NaCl, 50 mM -glycerophosphate, 1% Triton X-100, 50 mM NaF, 2 mM dithiothreitol, 100 µM Na₃VO₄, 5 mM benzamidin, 1 mM PMSF, 10 µg/ml leupeptin, 10 µg/ml aprotinin). Cell lysates were precleared with an irrelevant polyclonal rabbit antibody and protein A-Sepharose (Pharmacia) before incubation with a rabbit anti-Raf-1 antibody (Sc133; Santa Cruz Biotechnology). Immune complexes were collected with protein A-Sepharose and washed twice with lysis buffer and twice with washing buffer (20 mM Hepes, pH 7.4, 10 mM MgCl₂). Immunoprecipitated Raf-1 was incubated with 1 µg of purified histidyl-tagged recombinant MEK1 in the presence of 100 µM ATP (5 µCi of [-³²P]ATP), 20 mM Hepes, pH 7.4, 10 mM MgCl₂, 10 mM p-nitrophenyl phosphate, and 1 mM dithiothreitol at 30 °C for 30 min in a total volume of 50 µl. Thereafter, 1 µg of histidyl-tagged recombinant kinase-inactive ERK2 was added to each sample, and incubation was carried out for another 10 min at 30 °C. The reaction was stopped by adding 15 µl of 4 × Laemmli buffer. Proteins were separated by 10% SDS-PAGE, electrotransferred to PVDF, subjected to autoradiography, and analyzed with a PhosphorImager (Molecular Dynamics, Inc.). Purification of recombinant histidyl-tagged wild-type MEK1 and kinase-inactive ERK2 was performed as described (49).

Quantitation of TNF and Analysis of TNF Expression by Reverse Transcriptase-PCR

Cells dispatched in plastic culture plates (1 × 10⁶ cells/sample for TNF quantitation and 1 × 10⁷ cells/sample for analysis of TNF mRNA expression) were treated with bacteria or LPS at 37 °C in a humidified atmosphere (5% CO₂). After 60 min, extracellular bacteria were killed by the addition of 100 µg/ml gentamicin. In some experiments, cells were treated with both LPS and gentamicin after 60 min of bacterial infection. For TNF quantitation, the cell culture supernatants were removed after a final 120-min incubation. The TNF cytokine level in the culture supernatant was evaluated by a cytotoxic assay performed with the TNF-sensitive murine fibroblast cell line L929 as described (21, 22). L929 viability was colorimetrically determined using the CellTiter 96 AQ Assay (Promega, Madison, WI) according to the manufacturer's instructions. For analysis of TNF expression, total RNA was extracted with Trizol (Life Technologies), as described by the manufacturer, after a final 120-min incubation. The reverse transcription reaction was performed at 42 °C for 50 min on 5 µg of total RNA, using the murine Moloney leukemia virus reverse transcriptase (Life Technologies) and oligo(dT)_(12-18) (Life Technologies). 1 µl of each cDNA was amplified using 1 unit of Gold Star polymerase (Eurogentec, Seraing, Belgium) and 0.5 µM specific primers. Primer pairs specific for TNF (sense, 5-TCT CAT CAG TTC TAT GGC CC-3; antisense, 5-GGG AGT AGA CAA GGT ACA AC-3; PCR product, 212 base pairs) and for ₂m (sense, 5-TGA CCG GCT TGT ATG CTA TC-3; antisense, 5-CAG TGT GAC CCA GGA TAT AG-3; PCR product, 222 base pairs) were designed and purchased from Eurobio (Les Ulis, France). PCR was performed with 20 cycles. Amplification of ₂m was used as a control. The PCR products were run on a 1.5% agarose gel supplemented with ethidium bromide.

We analyzed the oxidative burst of J774.A1 cells in response to opsonized zymosan after pretreatment of the cells with different Y. enterocolitica strains. The oxidative burst was measured as luminol-enhanced chemiluminescence in an automatic luminescence analyzer (Lumicon, Hamilton, Bonaduz, Switzerland) as described (50). 5 × 10⁵ cells were infected with bacteria for 1.5 h. Thereafter, cells were resuspended in 0.95 ml of phosphate-buffered saline containing 5 µg/ml luminol (Boehringer, Mannheim, Germany). Stimulation was started by the addition of 50 µl of opsonized zymosan (Sigma), and chemiluminescence was recorded for a total of 30 min. Assays were repeated at least three times.

5 × 10⁵ cells/well were infected with bacteria in 24-well culture plates for 1 h. To discriminate between intra- and extracellularly located bacteria, cells were then stained using a double-immunofluorescence technique as described (16, 51). This technique allows determination of the numbers of both cell-associated (red and green fluorescence) and phagocytosed (exclusively green fluorescence) bacteria. For every strain investigated, three separate experiments were performed, and 100 cells from each experiment were analyzed under a fluorescence microscope. Mean percentages of phagocytosed versus total numbers of bacteria per cell were determined.

RESULTS

To determine possible differences between the virulent and nonvirulent Y. enterocolitica strain during infection, we compared the patterns of tyrosine-phosphorylated proteins in cells stimulated with LPS from E. coli or with the two Y. enterocolitica strains. Cell lysates prepared after different stimulation times were immunostained with the monoclonal anti-phosphotyrosine antibody 4G10. After 15 min of stimulation with LPS, nonvirulent and virulent yersiniae, there was a dramatic increase in tyrosine phosphorylation of two proteins at the 38- and 42-kDa level (Fig. 1A). The phosphorylated 38-kDa protein was shown, by stripping and reprobing the membrane with the anti-p38 peptide antibody, to have the same electrophoretical mobility as p38, also known as RK, reactivating kinase (52, 53), or for cytokine-suppressive anti-inflammatory drug binding proteins (41) (data not shown). Reprobing the membrane in the same manner with a pan-anti-ERK antibody demonstrated that the tyrosine-phosphorylated protein at 42 kDa was ERK2 (Fig. 1B). This immunoblot also allowed us to distinguish between the unphosphorylated and phosphorylated forms of ERK2 and ERK1, since the phosphorylated forms exhibited slower electrophoretical mobilities. In unstimulated cells (Fig. 1B, lane 4), ERK labeling corresponded to unphosphorylated forms of ERK2 (lower band) and ERK1 (upper band). Cell treatment with LPS and Y. enterocolitica strains induced a total upward shift in the ERK proteins, in accordance with the strong tyrosine phosphorylation of ERK2 at about 42 kDa in panel A. It was not clear from panel A whether or not ERK1 (44 kDa) was also phosphorylated, because of intense phosphotyrosine labeling at about 44 kDa that was not regulated by LPS or bacterial stimulation. Nevertheless, Fig. 1B revealed a electrophoretical shift that also occurred at the ERK1 level, demonstrating that ERK1/p44^MAPK was phosphorylated over the same time course as ERK2. We thus referred to these proteins as ERK1/2, since both ERK proteins behaved similarly. Interestingly, in the 46-55-kDa region, where the JNK subtypes migrate, no substantial tyrosine phosphorylation change could be detected by immunoblotting with the anti-phosphotyrosine antibody 4G10 (Fig. 1A).

Phosphorylation of p38 and ERK1/2 tyrosine residues remained unchanged for at least 30 min in each stimulating condition. Thereafter, a slow decrease in phosphorylation was observed under stimulation with LPS and the nonvirulent strain (Fig. 1, lanes 1 and 3), whereas almost complete dephosphorylation of p38 and ERK1/2 occurred after a 90-min infection with the virulent Y. enterocolitica strain (Fig. 1, lane 2). The decrease in tyrosine phosphorylation of p38 might have been faster than that of ERK1/2, since a preferential decrease in p38 tyrosine phosphorylation was already visible after only 60 min of infection. After a 90-min infection, inhibition of tyrosine phosphorylation induced by the virulent Y. enterocolitica strain affected p38 and ERK1/2 and also some other proteins. This phenomenon, i.e. interference of Yersinia sp. with macrophage tyrosine phosphorylation, has already been described (18, 19, 54, 55) and can at least partially be attributed to the tyrosine phosphatase of Yersinia sp., named YopH. In any case, the virulent Y. enterocolitica strain selectively decreased p38 and ERK1/2 tyrosine phosphorylation levels, indicating that their kinase activities, conferred by dual phosphorylation of the Thr-X-Tyr motif, should be concomitantly reduced.

To directly measure the activities of ERK and p38 and to determine whether JNK activity was also affected by the inhibitory effect of Y. enterocolitica, we analyzed the ability of cytosolic extracts to phosphorylate the transcription factors Elk-1, ATF2 and c-Jun (Fig. 2). Although Elk-1 and ATF2 cannot be considered as selective substrates for the kinases ERK and p38, respectively (48), c-Jun appears to be specifically phosphorylated by JNK. A 60-min incubation of J774A.1 cells with LPS (Fig. 2, lanes 2) or the nonvirulent Y. enterocolitica strain (Fig. 2, lanes 3) induced a substantial (3-10-fold) increase in phosphotransferase activities toward the different GST fusion proteins as compared with basal levels. For example, GST-c-Jun-(1-79) phosphorylation was increased 6-fold with LPS and 11-fold with nonvirulent yersiniae. This clearly indicated that Y. enterocolitica also stimulated kinase activity of the JNK protein. However, after infection with the virulent Y. enterocolitica strain (Fig. 2, lanes 4), phosphorylation of all substrates was markedly reduced as compared with the nonvirulent strain, since the substrate phosphorylation was only 1.5-3-fold that of the control level. The parallel alteration of the different GST fusion proteins, including the two GST-c-Jun substrates, indicated that, in addition to the reduction in ERK and p38 kinase activities, the virulent Y. enterocolitica strain also inhibited JNK activity, as revealed by the weaker phosphorylation of both GST-c-Jun-(1-79) and GST-c-Jun-(1-222) substrates.

Inhibition of TNF Production Is Associated with a Reduction in MAPK Activities

Since the virulent Y. enterocolitica strain abolished MAPK activation, we wondered whether MAPK deactivation was related to the inhibition of macrophage TNF secretion. Fig. 3 (lanes 2-4) confirms that LPS and nonvirulent yersiniae induced a strong TNF-response in J774A.1 cells, while the virulent Y. enterocolitica strain, which prevented p38 and ERK1/2 tyrosine phosphorylation, completely blocked TNF secretion (Fig. 3A) as well as TNF mRNA expression (Fig. 3B). Furthermore, when cells were first infected with the virulent Y. enterocolitica strain, further stimulation with LPS from E. coli could trigger neither TNF production (data not shown) nor tyrosine rephosphorylation of p38 and ERK2 (Fig. 3C, bottom panel). To gain further insight into a possible relation between the lack of TNF production and the decreased p38 and ERK1/2 tyrosine phosphorylation, we analyzed defined Y. enterocolitica mutants. The YopH(1) strain, a mutant with selectively impaired secretion of the protein-tyrosine phosphatase YopH, prevented tyrosine phosphorylation and TNF production to a similar extent as the virulent wild-type strain (Fig. 3, lanes 4 and 5). On the contrary, the Yop secretion-negative LcrD mutant with defective secretion of all Yops (Yop secr.; Fig. 3, lane 6), did not decrease p38 and ERK2 tyrosine phosphorylation and induced strong TNF release, similar to the nonvirulent strain. Fig. 3 also shows the results obtained with two other mutants expressing a restricted repertoire of yop genes. The YopD,B,N,V⁺ strain harbors the fragment of the Y. enterocolitica virulence plasmid encoding the Yop secretion machinery, including the genes coding for YopD, YopB, YopN, and the V antigen, which are necessary for Yop expression, secretion, and translocation. The second strain, referred to as YopD,B,N,V,H,E,YadA⁺ expresses, in addition to yopD, yopB, yopN, and lcrV (encoding the V antigen), yopH and yopE, which encode the translocated proteins YopH and YopE, and yadA, encoding the cell adhesin YadA. Analysis of these two mutants indicated that they were able neither to reduce p38 and ERK1/2 phosphorylation nor to block TNF-production of J774A.1 cells (Fig. 3, lanes 7 and 8).

Correlation between inhibition of TNF production and reduction of p38/ERK1/2 tyrosine phosphorylation.

To compare the action of various mutants on MAPK activities, a time course study was performed with GST-Elk-1, GST-ATF2, and GST-c-Jun-(1-79) as in Fig. 2, except that the three substrates were added together in the kinase assay. In agreement with the phosphorylation of p38 and ERK1/2 seen in Fig. 3B, LPS of E. coli and all Y. enterocolitica strains induced strong phosphorylation of the three substrates within 30 min of stimulation (Fig. 4). Thereafter, only cells infected with the virulent Y. enterocolitica strain and with the YopH(1) mutant exhibited almost complete disappearance of kinase activities within 60-90 min. The reduction in phosphorylation occurred over a similar time course for the three substrates, suggesting that the virulent and the YopH(1) strain decreased the activities of MAPK cascades simultaneously. MAPK activities were also inhibited within 90 min, when virulent yersiniae were killed after 30 min of infection by the addition of 100 µg/ml gentamicin (data not shown). Taken together, these results indicate the existence of a relation among blockade of p38/ERK1/2 tyrosine phosphorylation, inhibition of p38/ERK1/2/JNK kinase activities, and suppression of TNF-production.

Inhibition of TNF Production Is Distinct from Other Y. enterocolitica Virulence Properties

Resistance of Yersinia sp. to phagocytosis and inhibition of the oxidative burst of professional phagocytes is known to be conferred by YopH and by YopE (11, 13-16). This was confirmed by strain YopD,B,N,V,H,E,YadA⁺, which effectively resisted phagocytosis and inhibited the J774A.1 cell oxidative burst, in contrast to strain YopD,B,N,V⁺ lacking YopH and YopE (Table II). However, since these two strains induced strong TNF release (Fig. 3 and Table II), the ability of Y. enterocolitica to resist phagocytosis and to suppress the oxidative burst is obviously completely independent of its ability to block p38/ERK1/2 tyrosine phosphorylation and TNF production. Interestingly, the YopH protein, which possesses tyrosine phosphatase activity (12), seems not to be involved in p38 and ERK1/2 dephosphorylation and TNF inhibition (Fig. 3, lane 5). In the YopH(1) strain, the mutation actually involves the YopH-specific chaperone sycH, and thus residual secretion of YopH by leaky cells cannot be excluded. To definitely rule out a possible inhibitory role of YopH, we analyzed another YopH strain affected in YopH expression (YopH(2)) and the complemented YopH(1) strain, secreting YopH (YopH(1)/H⁺). Cells were infected with bacteria for 60 min and thereafter were restimulated by LPS treatment. The p38 protein was immunoprecipitated and immunoblotted with the anti-phosphotyrosine antibody 4G10. As expected, the nonvirulent strain and the LcrD mutant, impaired in Yop secretion (Yop secr.), did not prevent p38 tyrosine phosphorylation (Fig. 5, lanes 3 and 8). In contrast, the virulent wild-type strain, the two YopH mutants, and the complemented mutant (YopH(1)/H⁺), which all inhibited TNF production (data not shown), blocked tyrosine phosphorylation of p38. These results demonstrate that the inhibition of tyrosine phosphorylation of p38 (Fig. 5) ERK1/2 (Fig. 3) and probably that of JNK occur independently of the presence of YopH but depend on a functional Yop secretion/translocation apparatus (e.g. LcrD). This suggests that one or several secreted Y. enterocolitica virulence proteins are involved.

Table II. Comparison of the effects of four different Y. enterocolitica strains on J774A.1 cell TNF production, phagocytosis, and oxidative burst Y. enterocolitica strain TNF releasea Phagocytosisb Oxidative burstc Nonvirulent 87  ± 8 91  ± 1 65  ± 2 Virulent 8  ± 1 5  ± 1 5  ± 4 YopD,B,N,V+ 64  ± 6 87  ± 3 58  ± 16 YopD,B,N,V,H,E,YadA+ 97  ± 10 5  ± 2 5  ± 5 a Results for TNF release are the values from the experiment depicted in Fig. 3, expressed as percentages of released TNF induced by 10 µg/ml LPS (551 ± 20 pg/ml = 100%). b Cells were incubated with bacteria for 1 h and then stained by a double-immunofluorescence technique to discriminate between intra- and extracellularly located bacteria. Mean percentages of ingested bacteria with respect to the total number of bacteria per cell were determined by counting 100 cells from each experiment. c Cells were preexposed to Y. enterocolitica for 90 min and then treated with opsonized zymosan. Chemiluminescence responses were recorded for a total of 30 min. Mean percentages of the zymosan-induced chemiluminescence response of cells not preexposed to bacteria (100%) are shown. Values ± S.E. shown in columns three and four are from three independent experiments.

Since the reduction in MAPK tyrosine phosphorylation and activation cannot be attributed to the phosphatase YopH, we analyzed Raf-1 kinase activities to determine whether the virulent Y. enterocolitica strain modulates signaling pathways upstream of MAPKs. Fig. 6 demonstrates that the ability of Raf-1 to activate MEK1, which then in turn phosphorylated ERK2, was markedly reduced when macrophages were infected with the virulent Y. enterocolitica strain, in contrast to treatment of macrophages with the nonvirulent strain or with LPS. Suppression of the Raf-1 kinase activity by the virulent Y. enterocolitica strain may indicate that inhibition of the MAPK signaling cascades occurs at least partially via reduction of upstream kinase activities already at the level of Raf-1.

DISCUSSION

Yersinia sp., like a number of other microbial pathogens, are supposed to modulate eukaryotic signaling pathways for their own benefit (20). In this study, we analyzed the impact of Y. enterocolitica on macrophage MAPK signaling pathways using J774A.1 cells as an infection model. Infection with Y. enterocolitica was found to stimulate p38 and ERK1/2 MAPK pathways, as detected by tyrosine phosphorylation. By contrast, direct tyrosine phosphorylation of JNK was not obvious, but this has been attributed to the fact that JNK phosphotyrosine cannot be easily detected by immunoblotting using an anti-phosphotyrosine antibody (56, 57). The patterns of tyrosine-phosphorylated proteins (Figs. 1 and 3) and kinase activities (Figs. 2 and 4) obtained with LPS and virulent and nonvirulent Y. enterocolitica strains were very similar over the first 30 min. Thereafter, the virulent strain harboring the Y. enterocolitica virulence plasmid induced a substantial reduction in kinase activities, as revealed by a decrease in the phosphorylation of ERK1/2 and p38 kinases along with their substrates, the transcription factors Elk-1 and ATF2, as well as that of the JNK-specific substrate c-Jun. The fact that the reduction in kinase activities occurred after only 1 h of cell infection can be explained by the delay necessary for the Yops to reach their targets and to exert their effects on the host cell (5, 16, 19). The initial stimulation of the three types of MAPKs followed by selective inhibition with the virulent Y. enterocolitica strain was also observed with macrophages derived from human monocytes (data not shown).

A link between MAPK activation and TNF production induced by LPS has been widely documented (41-42, 58-60). Deactivation of the MAPKs p38, JNK, and ERK1/2 induced by virulent Y. enterocolitica, therefore, might be related to its inhibitory effect on macrophage TNF secretion. Indeed, we found that all investigated Y. enterocolitica strains capable of inhibiting MAPK activities also prevented TNF production, and reciprocally, all strains inhibiting TNF release also deactivated MAPKs. This finding strongly supports the hypothesis that inhibition of TNF release by Y. enterocolitica originates from shortening p38, ERK1/2, and JNK activation by reducing their levels of tyrosine phosphorylation. Furthermore, our evaluation of TNF mRNA levels by reverse transcriptase-PCR indicated that LPS stimulation and infection with the nonvirulent Y. enterocolitica strain dramatically enhanced the amount of TNF mRNA, while no accumulation of this messenger occurred after infection with the virulent strain. This indicates that the inhibitory effect of the virulent Y. enterocolitica strain on TNF release is probably not due to alteration of TNF maturation or secretion but rather to a lack of TNF mRNA accumulation. The absence of mRNA may be due to inhibition of TNF gene transcription or due to mRNA instability. A role of p38 in post-transcriptional control of TNF gene expression has been clearly shown by the group of Lee (41, 42) using the anti-inflammatory drug SB203580. It is thus possible that the accelerated dephosphorylation of p38 is partially responsible for the inhibition of TNF synthesis.

The fact that not only p38, but also ERK1/2 and JNK, are deactivated by the virulent Y. enterocolitica strain raises the question of their potential role in TNF suppression. Indeed, it was recently shown that blockage of the ERK pathway by the MEK inhibitor PD98059 prevents TNF mRNA synthesis induced by FcR stimulation (58). No specific drugs for the JNK pathway are available yet. However, TNF gene expression is stimulated by AP-1 (61, 62), a transcription factor composed of c-Jun and c-Fos, which are activated through phosphorylation by ERK and JNK (63). Deactivation of p38, ERK, and JNK induced by the virulent Y. enterocolitica strain may thus together contribute to inhibition of TNF synthesis. The similar time courses of deactivation of all the MAPK, ERK1/2, JNK, and p38 suggests that yet unidentified bacterial virulence factors might act at a step that is common to the three pathways, i.e. upstream of MAPKs. When investigating this possibility, we found that Raf-1 activity was lowered after infection with the virulent Y. enterocolitica strain compared with the nonvirulent strain. This finding indicates that at least part of the TNF-inhibitory action takes place upstream of the MAPKs. However, it remains to be determined whether Y. enterocolitica inhibits MAPK signaling cascades via reduction of upstream kinase activities only or whether it also causes dephosphorylation of MAP kinases themselves. It cannot be ruled out that bacterial factors trigger or accelerate the expression of an endogenous macrophage phosphatase, such as the specific MAPK phosphatase-1 (64, 65) or HVH1 or HVH2, two human homologs of the vaccinia virus dual specific phosphatase VH1 (66, 67).

In an attempt to identify the potential virulence factors involved in MAPK deactivation and TNF inhibition, we compared the characteristics of several Y. enterocolitica mutants. Analysis of mutants with impaired tyrosine phosphatase YopH expression or secretion clearly excluded participation of YopH in the inhibitory effect on MAPK tyrosine phosphorylation and TNF production. Furthermore, experiments on a mutant with defective secretion of all Yops (LcrD mutant) demonstrated that indeed one or several released Y. enterocolitica proteins other than YopH mediate inhibition of MAPK activities and TNF production. Analysis of Y. enterocolitica strains capable of producing individual virulence factors revealed, in agreement with our previous data on granulocytes (16), that YopH and YopE confer resistance to phagocytosis and suppression of the J774A.1 cell oxidative burst to Y. enterocolitica; a strain, capable of producing YopH, YopE, and the adhesin YadA, as well as YopD, YopB, YopN, and the V antigen, suppressed macrophage phagocytosis and oxidative burst, in contrast to a strain secreting only the latter proteins, which are necessary for expression, secretion, and translocation of active Yops (2, 4, 7-8, 10). The fact that both strains were unable to inhibit MAPK activities and TNF production clearly demonstrates that the anti-TNF effect is not a consequence of the ability of Y. enterocolitica to inhibit phagocytosis and to suppress the oxidative burst. Furthermore, this finding implies that the Yops released by this strain (YopD, YopB, YopN, YopH, YopE, V antigen) are not, or at least not solely, responsible for the inhibition of TNF production, although such an effect was previously attributed to YopB (28) and the V antigen (26, 27).

In summary, we demonstrated for the first time that virulent Y. enterocolitica mediates disruption of eukaryotic signal transduction. Moreover, our study highlights a correlation between the inhibition of macrophage TNF production by Y. enterocolitica and deactivation of MAPK pathways. The virulence factors responsible for these inhibitory effects are released Y. enterocolitica proteins other than YopH or YopE. The cellular target from which the different MAPK pathways are affected seems to be located at the MAPK kinase kinase level, i.e. Raf-1, or upstream. These characteristics point to the Ras superfamily of small G proteins, among which Cdc42, Rac, and Rho appear to be activated in cascade, with subsequent activation of multiple pathways including MAPK modules (68). Studies presently under way address this question and should provide new insight into the pathogenesis of yersiniosis.