Toll-like receptor 2 (TLR2) and TLR4 differentially activate human dendritic cells.

Toll-like receptors (TLRs) mediate cell activation by various microbial products. Here, we demonstrate that activation of dendritic cells by TLR2 or TLR4 agonists, although it led to comparable activation of NF-kappa B and mitogen-activated protein kinase (MAPK) family members, resulted in striking differences in cytokine and chemokine gene transcription, suggesting that TLR2 and TLR4 signaling is not equivalent. A TLR4 agonist specifically promoted the production of the Th1-inducing cytokine interleukin (IL) 12 p70 and the chemokine interferon-gamma inducible protein (IP)-10, which is also associated to Th1 responses. In contrast, TLR2 stimulation failed to induce IL-12 p70 and interferon-gamma inducible protein (IP)-10 but resulted in the release of the IL-12 inhibitory p40 homodimer, producing conditions that are predicted to favor Th2 development. TLR2 stimulation also resulted in preferential induction of IL-8 and p19/IL-23. Involvement of phosphatidylinositol 3-kinase and p38 MAPK in the TLR-mediated induction of several cytokine and chemokine messages was demonstrated using specific inhibitors. Thus, TLRs can translate the information regarding the nature of pathogens into differences in the cytokines and chemokines produced by dendritic cells and therefore may contribute to the polarization of the acquired immune response.

Toll-like receptors (TLRs) mediate cell activation by various microbial products. Here, we demonstrate that activation of dendritic cells by TLR2 or TLR4 agonists, although it led to comparable activation of NF-B and mitogen-activated protein kinase (MAPK) family members, resulted in striking differences in cytokine and chemokine gene transcription, suggesting that TLR2 and TLR4 signaling is not equivalent. A TLR4 agonist specifically promoted the production of the Th1-inducing cytokine interleukin (IL) 12 p70 and the chemokine interferon-␥ inducible protein (IP)-10, which is also associated to Th1 responses. In contrast, TLR2 stimulation failed to induce IL-12 p70 and interferon-␥ inducible protein (IP)-10 but resulted in the release of the IL-12 inhibitory p40 homodimer, producing conditions that are predicted to favor Th2 development. TLR2 stimulation also resulted in preferential induction of IL-8 and p19/IL-23. Involvement of phosphatidylinositol 3-kinase and p38 MAPK in the TLR-mediated induction of several cytokine and chemokine messages was demonstrated using specific inhibitors. Thus, TLRs can translate the information regarding the nature of pathogens into differences in the cytokines and chemokines produced by dendritic cells and therefore may contribute to the polarization of the acquired immune response.
Cells of the innate immune system possess germ line-encoded receptor molecules that enable them to recognize structural components conserved among classes of microorganisms (1,2). This recognition event serves two purposes. First, it alerts the immune system to the presence of pathogens so that an immediate response can be mounted to contain the infection. Later, during the establishment of acquired immunity, this recognition event is thought to provide information regarding the nature of the invading microorganism that will contribute to determine the differentiation of T helper (Th) 1 cells into either Th1 cells, which promote cell-mediated immunity, or Th2 cells, which encourage humoral responses. Cytokines play a pivotal role in this process, with IL-12 committing cells to Th1 lineage differentiation and IL-4, in the absence of IL-12 and IFN-␥, promoting Th2 development (3). However, the mechanism by which the nature of the pathogens is translated into differences in the cytokines produced remains poorly understood.
Recognition of several microbial products is mediated by members of the Toll-like receptor (TLR) family (4,5). These receptors are characterized by an extracellular leucine-rich domain that shows considerable divergence and is probably involved in recognition of different agonists, and a highly conserved cytoplasmic Toll-IL-1 receptor domain that connects the receptor to the signaling machinery shared by IL-1 and IL-18 (6,7). Although all TLRs for which an agonist has been identified appear to activate similar signaling pathways, including NF-B, p38 MAPK, stress-activated protein/JNK, and MAPK (Erk1/2) kinases (8 -12), it is not clear whether different TLRs differ in their ultimate function, the activation of innate immune responses. Evidence that TLRs are not functionally equivalent is accumulating. Homodimerization of the cytoplasmic domain of TLR4 is sufficient to initiate signaling that leads to TNF-␣ production by a macrophage cell line. In contrast, heterodimerization of TLR2 cytoplasmic domain with either TLR1 or TLR6 cytoplasmic tails is required for induction of this cytokine (13). These findings indicate that despite their high sequence homology, the cytoplasmic tails of different TLRs are not functionally equivalent and suggest that different signals may emanate from distinct receptors. The concept that distinct TLR may differ in their signaling ability is also supported by studies of Drosophila TLRs. Despite sharing homologous cytoplasmic domains, toll and 18 wheeler, which mediate the immunity of the fly to fungi and bacteria, respectively, are known to activate very different and nonoverlapping gene expression (14,15). In this study, we explored the possibility that TLR2 and TLR4 may differentially activate human DCs.

EXPERIMENTAL PROCEDURES
Reagents-LPS (Escherichia coli K12 LCD25) was from List Biological Laboratories (Campbell, CA). It was purified from contaminant lipoproteins normally found in commercially available LPS preparations by double phenol extraction, exactly as described in (16). Staphylococcus aureus peptidoglycan (PGN) was from Fluka (Milwaukee, WI). Zymosan was from Sigma. Synthetic lipopeptide Pam 3 -Cys-Ala-Gly is derived from an E. coli lipoprotein and was purchased from Bachem (Torrance, CA). OspB-L is a synthetic lipopeptide (Cys-Ala-Gln-Lys-Gly-Ala) derived from Borrelia burgdorferi, and OspB is the non-acylated hexapeptide that served as a control as it possesses no biological activity. They were both kindly provided by Dr. Juan C. Salazar, University of Connecticut Medical School. LY-294,002 and SB203580 were purchased from Alexis Biochemicals (San Diego, CA). U0126 was purchased from Cell Signaling Technology (Beverly, MA).
Plasmids and Cell Lines-TLR2 and TLR4 sequences were gener-* This work was supported by National Institutes of Health Grant 5R35-CA47554 (to J. L. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ated by reverse transcription-polymerase chain reaction from human monocyte total RNA using Superscript II reverse transcriptase (Life Technologies, Inc.) and Pfu Turbo DNA polymerase (Stratagene, La Jolla, CA) using the following primers: TLR4 -5Ј, GCGCGCGGCCGCG-GAAAGCTGGGAGCCCTGCGTGGAGG; TLR4 -3Ј, GCGCGGATCCT-CAGATAGATGTTGCTTCCTGCC; TLR2-5Ј, GCGCGCGGCCGCGAA-GGAAGAATCCTCCAATCAGGC; and TLR2-3Ј, GCGCGGATCCCTA-GGACTTTATCGCAGCTCTCAG. The TLR2 sequence was cloned in the NotI-BamHI sites of pFLAG-CMV-1 (Sigma). TLR4 sequence was cloned as a NotI-BamHI fragment into a modified pFLAG-CMV-1 that contained the c-Myc epitope tag in place of the FLAG tag. C-terminal FLAG-tagged MD-2-containing vector was kindly provided by Dr. Miyake (Saga Medical School). The MD-2 sequence was excised as a XhoI-NotI fragment and inserted in the SalI-NotI sites of vector pBOS-H2Bgfp, which also contains the blastidicin resistance gene. A subclone of the HeLa cell line was transfected using Superfect reagent (Life Technologies, Inc.) with pFLAG-CMV-TLR2 and pPuro or with pMyc-CMV-TLR4 and pBOS-MD-2, and stable transformants were selected in puromycin or blastidicin containing medium, respectively. Several colonies were isolated, and TLR2 or TLR4 expression was tested by Western blot and FACScan analysis using M2-FLAG or Myc monoclonal antibodies. Two clones with comparable TLR2 and TLR4 expression were expanded and further characterized.

FIG. 1. TLR2 and TLR4 agonists activate NF-B and members of the MAPK family in HeLa and DCs to the same extent. A,
HeLa-TLR2 and HeLa-TLR4/MD-2 stable clones were transiently transfected with the endothelial leucocyte adhesion molecule-luciferase reporter construct CMV-CD14 (kindly provided by Dr. Golenbock, Boston University), and CMV-␤-galactosidase and stimulated for 6 h with zymosan, PGN, Pam-Cys, OspB-L, LPS, or TNF-␣ at the indicated concentrations (expressed as g/ml). LPS was used at 10 g/ml on HeLa-TLR2 and at 0.1, 1, and 10 ng/ml on HeLa-TLR4/MD-2. NF-B activation was measured by luciferase assay. B, immature human DCs were treated with LPS or PGN at the indicated concentrations and incubation times, and NF-B activation was measured by EMSA. C-E, immature human DCs were treated with LPS or PGN at the indicated concentrations and incubation times, and JNK activity (C) was measured in cell lysates using glutathione S-transferase (GST)-c-Jun as substrate. MAPK (Erk1/2) (D) and p38 MAPK (E) activation was detected using phosphospecific antibodies. The bottom panel in E shows detection of p38 MAPK. OspB is the non-acylated hexapeptide that serves as a control as it possesses no biological activity. The smeared bands in this sample lane are due to RNA degradation. B, DCs were stimulated with LPS (10 ng/ ml), PGN (10 g/ml), or Pam-Cys (10 g/ ml) for 36 h. IP-10 was measured in supernatant by ELISA. Measurements were done in duplicate. Data (mean Ϯ S.D.) are from three donors. The detection limit of the assay was 31 pg/ml. C, RNase protection assay analysis of transcripts for chemokines in HeLa. HeLa-TLR2 and HeLa-TLR4/MD-2 were stimulated with LPS (10 ng/ml), PGN (10 g/ml), or zymosan (50 mg/ml) for 4 h.
TLR2 and TLR4 Differentially Activate Human Dendritic Cells luciferase and 1 g of pcDNA-CD14 (kindly provided by Dr. Golenbock, Boston University) and 0.1 g of CMV-␤-galactosidase. Luciferase assay was performed using Promega (Madison, WI) reagents according to the manufacturer's recommendations. Efficiency of transfection was normalized by measuring ␤-galactosidase in cell lysates.
Jun Kinase Assay-Cells were lysed in 25 mM Hepes, pH 7.5, 300 mM NaCl, 0.1% Triton, 0.2 mM EDTA, 1.5 mM MgCl 2 , 20 mM ␤-glycerophosphate, 1 mM Na3 VO4, 0.5 mM DTT, and 1 M phenylmethylsulfonyl fluoride. Cell lysates containing 100 g of proteins were incubated for 3 h at 4°C with glutathione-agarose beads coated with glutathione S-transferase-c-Jun recombinant protein. Beads were then washed three times with lysis buffer and twice with kinase buffer (20 mM Hepes, pH 7.5, 20 mM ␤-glycerophosphate, 10 mM PNPP, 10 mM MgCl 2 , and 10 mM DTT). The kinase reaction was performed at 37°C for 30 min in kinase buffer supplemented with 10 M ATP and 5 Ci of [␥-32 P]ATP per sample. Proteins were resolved by PAGE, and the dried gel was exposed for autoradiography. Phospho-MAPK antibodies were from Cell Signaling (Beverly, MA).
RNase Protection Assay-Total RNA was isolated using Trizol rea-gent (Life Technologies, Inc.). RNase protection assay was performed using 4 -6 g of total RNA using the BD Biosciences-PharMingen Riboquant kit according to the manufacturer's recommendations. The hCK-2b, hCK-5, and hCK3 multiprobe template sets were used. The templates for IL-23 p19 and for GAPDH were generated by reverse transcription-polymerase chain reaction using the following primers: IL-23a, GCAGAGCTGTAATGCTGCTG; IL-23b, CCGATCCTAGCAGC-TTCTC. The polymerase chain reaction product was cloned into pCR2.1-TOPO, and the p19 sequence was excised by EcoRI-BamHI restriction and cloned into the EcoRI-BamHI sites of pcDNA3 (Invitrogen, Carlsbad, CA). For GAPDH, the following primers were used: GAPDHa, CGCGCTCGAGCCCAGAAGACTGTGGATGG; GAPDHb, CGCGGAATTCGGCAGGTTTTTCTAGACGG. The polymerase chain reaction product was digested with XhoI-EcoRI and cloned into the EcoRI-XhoI sites of pcDNA3. The templates were linearized by EcoRI or XhoI restriction, phenol-extracted, and precipitated, and 100 ng were used in a standard RNase protection assay. ELISA-DCs were plated in RPMI 1640 medium/10% fetal calf serum at a concentration of 2 ϫ 10 6 cells/ml in 48-well plates and stimulated as described in the figure legends. IP-10, IL-8, IL-12 p70, and IL-12 p40 were measured by ELISA using matched pairs of antibodies from R&D Systems according to the manufacturer's recommendations. The IL-12 p40 ELISA is specific for the p40 subunit with less than 5% cross-reactivity with the IL-12 p70.

RESULTS
To determine whether TLR2 and TLR4 differ in their ability to activate DCs, several microbial products that have been previously demonstrated to activate different TLRs were tested. Highly purified E. coli LPS was used as TLR4 agonist FIG. 4. TLR2 agonist is a more powerful IL-8 inducer than TLR4 agonist. A, RNase protection assay analysis of transcripts for chemokines. DCs were stimulated with LPS (10 ng/ml) or PGN (10 g/ml) for 14 h. B, DCs were stimulated with LPS (1 or 10 ng/ml) or PGN (1 or 10 g/ml) for 24 h. IL-8 was measured in supernatants by ELISA after dilution to 1:40 for LPS stimulation and to 1:100 for PGN stimulation. Measurements were done in triplicate. Data are mean Ϯ S.D. One representative experiment of three is shown. The detection limit of the assay was 31 pg/ml. (17,18). S. aureus PGN (19,20) and yeast zymosan (21) were used as agonists of the TLR2/TLR6 heterodimer (13). In addition, two synthetic lipopeptides derived from bacterial lipoproteins, OspB-L and Pam-Cys, were used. Similar peptides have been shown to activate cells through TLR2 paired to a yet unidentified TLR (22,23). Preliminary experiments were conducted in order to confirm specificities and characterize dose responses using HeLa clones stably transfected with TLR2 or TLR4/MD-2. Increasing concentrations of zymosan, PGN, and OspB-L and Pam-Cys lipopeptides were able to activate NF-B-dependent luciferase production in HeLa-TLR2 but not in HeLa-TLR4/MD-2 (Fig. 1A). As expected, the activation of NF-B by LPS was detected only in HeLa-TLR4/MD-2. This response was MD-2-dependent (data not shown) and was not detected in HeLa-TLR2 even at concentrations several logs higher than those that elicited a maximal response in HeLa-TLR4/MD-2. Similar doses of LPS or PGN were able to activate NF-B to the same degree in human immature DCs, as assessed by EMSA (Fig. 1B). Supershift experiments demon-strated that for both agonists, the main NF-B species induced was a p65/p50 dimer (data not shown). In addition to NF-B activation, LPS and PGN are known to activate p38 MAPK, stress-activated protein/JNK, and MAPK (Erk1/2) kinases in a TLR-dependent fashion. The activation state of these kinases was tested in DCs treated with LPS and PGN. In these cells, both LPS and PGN activated stress-activated protein/JNK (Fig. 1C), MAPK (Erk1/2) (Fig. 1D), and p38 MAPK (Fig. 1E) to the same extent. MAPK (Erk1/2) activation was also similarly affected by both agonists in HeLa-TLR2 and HeLa-TLR4/MD-2 (data not shown). Finally, we compared the effects of these microbial products on the induction of maturation of DCs (Fig.  2). Treatment with either LPS or PGN markedly induced equivalent expression of mature DC surface markers CD40, CD83, CD86, and human leucocyte antigen-DR. Thus, optimal and comparable cell activation in HeLa and DCs occurred at doses of 1-10 ng/ml LPS, 10 g/ml PGN, 50 g/ml zymosan, 2 g/ml lipopeptide Pam-Cys, and 10 g/ml lipopeptide OspB-L.
Activation of cells by microbial products through TLR2 and FIG. 5. TLR4 but not TLR2 agonists induce IL-12 p70 production in human immature DCs. A, RNase protection assay analysis of transcripts for proinflammatory cytokines. DCs were stimulated with TLR2 or TLR4 agonists for 4 h. Two representative experiments are shown. Generally, visualization of IL-12 p35 and p40 transcript required a longer time of exposure than the rest of the cytokine transcripts. For the gel on the right, 4-and 24-h-exposure autoradiograms are shown. The asterisks indicate bands that are derived from spillover of the unprotected riboprobe. Lines connect the unprotected probe fragments to the corresponding protected mRNA species. OspB is the non-acylated hexapeptide that serves as a control as it possesses no biological activity. B, DCs were stimulated with LPS (10 ng/ml), PGN (10 g/ml), or Pam-Cys (10 g/ml) in the presence or absence of IFN-␥ (50 ng/ml) for 36 h. Cytokines were measured in supernatants by ELISA. One representative experiment of three is shown. Values are mean Ϯ S.D. The detection limit of the assay was 15 pg/ml for IL-12 p70 and 31 pg/ml for IL-12 p40.
TLR4 is known to induce the secretion of various cytokines and chemokines. Therefore, we investigated whether TLR2 or TLR4 agonists differentially regulate the expression of these factors in human immature DCs. The pattern of expression of several chemokines, including Ltn, RANTES, IP-10, MIP1␣, MIP1␤, MCP-1, IL-8, and I-309, was analyzed using RNase protection assays (Fig. 3A). Housekeeping genes L32 and GAPDH provided internal controls. Whereas the majority of the chemokines were induced to the same extent by both TLR2 and TLR4 agonists after 4 h of stimulation, one transcript, the message for IP-10, was specifically induced by LPS but not by the TLR2 agonists PGN, zymosan, or lipopeptides. This effect was consistently observed using DCs derived from several donors and was seen at different time points and for all the effective doses of agonists tested. The lowest effective dose of LPS (0.1 ng/ml) was still able to induce IP-10 transcript, whereas maximally potent doses of PGN and zymosan never induced IP-10. The IP-10 message was not observed even after 18 h of stimulation with PGN (see Fig. 4) or zymosan and with prolonged overexposure of the autoradiograph. Lipopolysaccharide-stimulated DCs secreted large quantities of IP-10 protein in the media (Fig. 3B). This chemokine was completely absent in PGN-stimulated, DC-conditioned media. IP-10 message induction by TLR4 agonist was confirmed in HeLa TLR4/MD-2 ( Fig. 3C), thus ruling out the possibility that the induction of this transcript, rather than being TLR4-mediated, was indirectly stimulated by IFN-␥ released by few contaminating cells.
In some experiments, TLR2 stimulation appeared to preferentially induce the transcripts of specific chemokines, such as IL-8 and MIP-1␣. Initially, we interpreted this effect as a result of the supramaximal concentration of TLR2 agonists used to rule out the possibility that insufficient cell stimulation was the cause of the lack of IP-10 transcript. Surprisingly, when RNAs of cells stimulated for 14 h (instead of 4 h, as in previous experiments) with optimal agonist concentrations were analyzed, the IL-8 transcript was detected exclusively in cells stimulated with TLR2 agonist (Fig. 4A). Thus, 4-h stimulation resulted in comparable IL-8 message induction, whereas during prolonged stimulation, IL-8 message was detected exclusively in cells stimulated with TLR2 agonists. Note that in the HeLa cell lines, 4-h stimulation also resulted in comparable level of IL-8 transcript (Fig. 3C). Measurement of IL-8 in DC culture media revealed that cells stimulated for 24 h with TLR2 agonist released 100 times more IL-8 than cells stimulated with TLR4 agonist (400 Ϯ 32 ng/ml versus 4.4 Ϯ 0.26 ng/ml) regardless of the concentration of agonist used (Fig. 4B).
To test whether stimulation through TLR2 or TLR4 also resulted in differences in the expression pattern of proinflammatory cytokines, transcript levels for IL-12 p35 and p40 subunits, IL-10, IL-1␣ and IL-1␤, IL-1ra, IL-6, IL-18, and IFN-␥ were analyzed (Fig. 5A). As was observed for chemokine expression, most cytokine transcripts were induced to the same extent by both TLR2 and TLR4 agonists, but one transcript, IL-12 p35, was specifically induced only by the TLR4 agonist LPS. Similarly to IP-10, IL-12 p35 message was not observed even at supramaximal doses of PGN or zymosan or for prolonged incubation times (18 h). Interestingly, the transcript of IL-12 p40 subunit was equally induced by both TLR agonists. This protein can form homodimers that have been shown to act as an IL-12 receptor antagonist (24). Measurements of IL-12 p70 and IL-12 p40 protein in the cell culture supernatants of stimulated DCs confirmed the RNase protection data (Fig. 5B). Interleukin-12 p70 was produced at low but detectable levels only in the media from cells stimulated with LPS. In contrast, IL-12 p40 was produced at comparable levels in the media of cells stimulated with either LPS or PGN. Note that in these experiments, IL-12 p40 was not produced in such excess of IL-12 p70 to be able to block its activities. In fact, for this to happen, a 500 -1000-fold excess of p40 should be produced (24). Optimal IL-12 p70 production is known to require IFN-␥ priming. Cotreatment of DCs with IFN-␥ restored the ability of PGN and lipopeptides to induce IL-12 p70, as previously reported by others (23). However, even under these extreme conditions, LPS appeared to be a more potent stimulus for IL-12 p70 secretion. It should be noted that the primary encounter between a DC and a microbe is likely to occur in an IFN-␥-poor environment. Under these conditions, the low amounts of IL-12 p70 induced by TLR4 agonist might be enough to initiate the positive feedback loop between IL-12 and IFN-␥, thus producing conditions that would favor Th1 lineage commitment. It has recently been shown that IL-12 p40 can pair to a protein called p19 to form IL-23, a novel cytokine with biolog- ical activities similar but not identical to those of IL-12 (25). We therefore analyzed the expression of p19 using RNase protection assay (Fig. 6). Surprisingly, we found that the message for p19 was preferentially induced by TLR2 agonist. The reagents to test whether the p19 message is translated into protein and whether bioactive IL-23 is formed are at present not available.
Using the same experimental approach, we next analyzed the pattern of expression of more proinflammatory cytokines, including TNF-␤, Lt␤, TNF-␣, IFN-␥, IFN-␤, transforming growth factor ␤3, transforming growth factor ␤2, and transforming growth factor ␤1. Again, we found that the message of one of these cytokines, IFN-␤, was preferentially induced by TLR4 agonist (Fig. 7).
In order to investigate the molecular mechanism responsible for the differences in cytokines and chemokines gene induction observed we used pharmacological inhibitors of MAPK/Erk1/2, of p38MAPK, and of PI3K. As shown in Fig. 8 induction of chemokine and cytokine transcripts were most potently blocked by LY-294002, an inhibitor of PI3K, and by SB203580, an inhibitor of p38MAPK, whereas MAPK(Erk1/2) pathway inhibitor U0126 (or PD98059, data not shown) was less effective. DISCUSSION Once activated by microbial products, DCs start producing a variety of soluble factors that direct T helper cell differentiation toward the type of immune response, humoral versus cellmediated, that is better suited to fight the invading pathogen. The mechanism by which the nature of the pathogen determines the panoply of factors produced by DCs, and therefore the type of adaptive immune response, is poorly understood.
Here, we report that activation of DCs by microbial agonists of TLR2 or TLR4, although leading to comparable activation of NF-B and MAPK family members, resulted in striking differences in cytokine and chemokine gene transcription, suggesting that signals emanating from different TLRs are not equivalent and that these pattern recognition receptors may differentially contribute to the polarization of adaptive immu-nity. A TLR4 agonist specifically promoted the production of the cytokine IL-12 p70, the chemokine IP-10, and the transcript for IFN-␤. Conversely, TLR2 stimulation resulted in preferential expression of IL-8 and of IL-23 p19.
Among these factors, IL-12 and IP-10 appear to be particularly interesting for their ability to contribute to T helper cell polarization. Due to its ability to stimulate IFN-␥ production in T cells, IL-12 is the key cytokine in directing the development of Th1 cells (26). It is believed that in its absence, T cells development proceeds by default toward the Th2 lineage. Bioactive IL-12 is produced by activated APC as a heterodimer composed of the p40 and p35 subunits. The p40 subunit is expressed in large excess of the p35 subunit, which is the limiting factor in controlling IL-12 p70 production (27,28). IL-12 p40 can also form homodimers that have been shown to act as IL-12 receptor antagonist (24). It is therefore particularly interesting that TLR2 stimulation, although it fails to induce IL-12 p70 release due to defective p35 gene induction, is nevertheless still able to stimulate release of the inhibitory p40 homodimer, thus producing conditions that are predicted to favor Th2 lineage commitment. On the other hand, TLR2 stimulation preferentially induces p19 gene transcription. p19 is a recently identified protein that can form heterodimers with p40, creating a novel cytokine, IL-23, with activities similar to as well as different from IL-12. For example, compared with IL-12, IL-23 has a diminished ability to induce IFN-␥ secretion from activated T cells (25). It is also unclear which role IL-23 plays in T cells polarization. Due to the uncertainties regarding IL-23 biological activities, it is impossible at present to argue whether the lack of IL-12 production by TLR2-stimulated DCs can be compensated for by the higher p19 gene transcription in these cells.
Differential induction of IL-12 by TLR2 and TLR4 agonists has recently been reported in mouse macrophages (29). It is interesting to note that in that study, the lack of IL-12 production by TLR2 stimulation was due to absence of p40 transcripts rather than to a failure to transcribe p35, as in our study. Whether this reflects differences between human and mouse in the regulation of IL-12 biosynthesis is not clear.
Our finding that TLR4, but not TLR2, agonists stimulate production of IP-10 is also remarkable. IP-10 is a CXC chemokine produced by different cell types in response to IFN-␥ and microbial products. IP-10 is a chemoattractant for monocytes, NK cells (30,31), and, importantly, Th1 cells, which have been shown to preferentially express the IP-10 receptor CXCR3 (32,33). The inability of TLR2-stimulated DCs to produce detectable amounts of IP-10, in contrast to the high level obtained by TLR4 stimulation, may result in a defective recruitment of Th1 cells and is a further indication that signaling emanating from these receptors can differentially contribute to the polarization of the adaptive immune response.
Other factors were found to be differentially induced in our study. For example, the chemokine IL-8, a chemoattractant for neutrophils, was produced in much greater amounts by cells stimulated with TLR2 agonists than by cells stimulated with TLR4 agonists; conversely, the IFN-␤ gene was preferentially induced by TLR4 stimulation.
The use of pharmacological inhibitors of p38 MAPK and PI3K allowed us to establish the involvement of these kinases in the induction of several cytokine and chemokine genes in response to both TLR2 and TLR4 agonists. Interestingly, the transcripts of certain cytokines and chemokines, such as IL-12 p40, IL-10, and IP-10, appeared to be more affected by these inhibitors than others. Although our data cannot indicate whether either of these kinases is the TLR4-specific signaling molecule responsible for the differences between TLR2 and TLR4 stimulation, the involvement of PI3K in the induction of IP-10 appears particularly interesting because recruitment of PI3K and activation of AKT, a kinase that depends on PI3K activity, have been demonstrated for TLR2 but not for TLR4 (34). Here we show that this kinase plays an important role in the induction of chemokine and cytokine genes by both TLR2 and TLR4. IP-10 is an IFN-␥ inducible gene. At present, it is not known whether members of TLR family are able to activate IFN-␥-specific signaling pathways, such as the interferon-responsive factors. This area is being investigated.
In conclusion, we have provided evidence that TLRs can translate the information regarding the nature of pathogens into differences in the cytokines and chemokines produced by DCs. The observed patterns of gene induction would be predicted to differentially polarize adaptive immune responses.