Infection of Myeloid Dendritic Cells with Listeria monocytogenes Leads to the Suppression of T Cell Function by Multiple Inhibitory Mechanisms¹ Free

Phagocytic cells play an important role in the defense against infectious pathogens including intracellular bacteria, for example, Listeria monocytogenes (L.m.)⁴ (1) and Mycobacterium tuberculosis (2). Macrophages and neutrophils play a major role during early immune responses against these intracellular pathogens (3, 4), and yet, for a complete clearance of infection, an efficient adaptive immune response involving dendritic cells (DC) is required (1). As shown in listeriosis, DC are involved in pathogen elimination (5) and induction of Ag-specific CD8⁺ T cell responses (6). In vitro studies demonstrated an induction of costimulatory molecules on human DC after Listeria infection (7, 8). However, there is also evidence that macrophages and DC infected by intracellular bacteria, viruses, and parasites can acquire regulatory phenotype and exert inhibitory function (9, 10, 11, 12). Infection of human DC with L.m. leads to up-regulation of the immune inhibitory enzyme IDO, a key enzyme of tryptophan metabolism (10). Moreover, in vivo, an increase in DC numbers resulted in impaired protective immunity to subsequent infection of mice with L.m. (6), suggesting that DC-mediated inhibitory mechanisms might play a role in host-pathogen interaction.

During recent years, it has been suggested that DC are not only the most stimulatory cells for the induction of Ag-specific T cell responses (13), but that they are also capable of inhibiting T cell activation or even inducing T cell tolerance (14, 15). Numerous mechanisms have been linked to the phenotype of such cells, which are termed “tolerogenic” or “regulatory” DC (15, 16, 17). Down-regulation of costimulatory molecules (18) and stimulatory cytokines such as IL-12 (19), induction of coinhibitory molecules (e.g., PDL1 or TGF-β) (20, 21), as well as enzymatic mechanisms such as tryptophan catabolism (22, 23) have been identified as major inhibitory effectors of regulatory DC. The detrimental effect of such DC on Ag-specific immune responses is best illustrated in malignant disease (24), while their utility is explored in induction of tolerance in organ transplantation (15).

In infectious diseases, however, the existence of regulatory DC does not seem to be intuitively productive. We therefore examined regulation, expression, and function of immunoregulatory pathways in relation to granulomatous infection in vivo and in vitro primarily using L.m. and human DC as a model system. Under these conditions regulatory myeloid DC and macrophages seem to play a dual role within granuloma by impairing T cell-mediated immune responses against the infected cells in the granuloma while at the same time inhibiting pathogen growth. Altogether, the induction of regulatory DC and macrophages especially in granulomatous infections might be favorable to the host.

Blood samples were collected from healthy blood donors at the Center for Transfusion Medicine after informed written consent was obtained. All experiments were approved by the University of Cologne Institutional Review Board.

Lymph node specimens from patients with clinically and serologically confirmed cervicoglandular listeriosis and tuberculosis were obtained from the Institute of Pathology (University of Cologne). As a control, tonsils and lymph nodes of uninfected patients were used. Immunohistochemical and immunofluorescence analyses of paraffin-embedded tissue samples were performed with Abs for S100, CD4, CD8, and CD83 (DakoCytomation), CD11c and CD25 (NovoCastra), CD56 (Zytomed), IDO (Serotec), cyclooxygenase 2 (COX-2) (IBL International), and FoxP3 (eBioscience) as described before (10).

Flow cytometry (FACSCanto, BD Biosciences) was performed using the following mAbs: CD3, CD11c, CD14, CD16, CD19, CD25, CD54, CD56, and anti-HLA-DR (BD Biosciences), CD11b, CD40, CD80, and CD86 (BD Pharmingen), anti-TNF-RI, anti-TNF-RII, and CCR7 (R&D Systems), CD58 and CD83 (Immunotech), and CD68 (Serotec) with appropriate isotype controls. Intracellular staining for IDO (monoclonal mouse anti-human IDO Ab) (25) and IFN-γ (IFN-γ-PE, Immunotech) was performed using Cytofix/Cytoperm Plus kit with GolgiStop (BD Biosciences) according to the manufacturer’s instructions. All flow cytometric data were assessed with the BD CellQuest 3.3 software (BD Biosciences).

DC were generated according to standard protocols, as previously described (10, 26). For maturation, DC were incubated with TNF (Sigma-Aldrich), anti-CD40 mAb (BD Pharmingen) with or without prostaglandin E₂ (PGE₂, Sigma-Aldrich), as previously described (26). For some experiments, DC were treated with TNF in combination with PGE₂ and Pam₃Cys-Ser-(Lys)₄.trihydrochloride (Pam₃, Axxora, 1 μg/ml), TNF and PGE₂, TNF and Pam₃, or incubated with LPS (Sigma-Aldrich, 1 μg/ml). Alternatively, 50% of supernatants derived from immature DC (immDC), TNF-matured DC (matDC), or infected DC (infDC) were added to allogeneic immDC or matDC, respectively. Additionally, immDC were incubated with matDC or infDC in a 0.4-μm polycarbonate transwell system (Nunc). Macrophages were generated from monocytes, as previously described (10).

Wild-type strain (EGD) of L.m. was processed as previously described (10). Heat-killed Listeria were obtained by incubating wild-type bacteria at 65°C for 1 h. Monocytes, DC, and macrophages were infected with FITC-labeled Listeria or incubated with heat-killed Listeria at multiplicity of infection of 10 and infection efficiency was controlled by flow cytometry as previously described (10). Following the infection phase, cells were recultivated in fresh medium; for prolonged cultures (>12 h), gentamicin (50 μg/ml, Sigma-Aldrich) was added. For some experiments, DC treated with heat-killed Listeria were pulsed with recombinant human IL-2 (rhIL-2, Chiron) 24 h after the treatment onset and supernatants were collected after additional 48 h for functional assays.

Listeria-infected DC were incubated with appropriate neutralizing mAbs or inhibitors, as previously described (10). The following Abs were used: anti-TNF mAb (20 μg/ml, BD Pharmingen), clinically applied TNF-neutralizing Ab infliximab (0.001–10 μg/ml, Centocor), anti-IFN-γ mAb (0.1–1 μg/ml, BD Pharmingen), anti-IL-10 mAb (5 μg/ml, R&D Systems), and anti-TNF-RI and anti-TNF-RII mAbs (10–100 μg/ml). The COX-2 inhibitor rofecoxib (a gift of Drs. K. Schrör and J. Meyer-Kirchrath) was used at 1–10 μM.

Immature DC were harvested on day 7 and mature DC were harvested 72 h upon start of maturation. Infected and corresponding control monocyte-derived DC (mo-DC) were harvested 2, 4, 6, and 24 h after infection. RNA and cRNA preparation, microarray hybridization (HG-U133A, Affymetrix), data analysis, and visualization were performed as previously described (10, 26). Microarray data are accessible at the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database (accession no. GSE9946 at http://www.ncbi.nlm.nih.gov/ geo/query/acc.cgi?acc).

Quantitative analysis of real-time PCR was performed using LightCycler3 and RelQuant software, version 1.0 (Roche Diagnostics), as previously described (10). Primers used (Roche Diagnostics) included: IL2RA forward, ACTGCTCACGTTCATCATGG, reverse, GATCTCTGGCGGGTCATC, Universal ProbeLibrary probe no. 13; B2M forward, TTCTGGCCTGGAGGCTAT, reverse, TCAGGAAATTTGACTTTCCATTC, Universal ProbeLibrary probe no. 42.

COX-2 expression was assessed by anti-COX-2 polyclonal Ab (IBL International) as previously described (10). Alternatively, Western blots for COX-2 or IDO (25) were performed on the LI-COR Odyssey System (LI-COR Biosciences) according to the manufacturer’s instructions.

Soluble CD25, IL-2, TNF, IL-10, and IFN-γ in cell supernatants were measured by sIL-2R, IL-2, IFN-γ, IL-10, and TNF-α Eli-Pair kits (Diaclone Research) according to the manufacturer’s instructions. All samples were analyzed at least in duplicates.

Assessment of kynurenine levels in the supernatants was performed as previously described (10, 25). Samples were run in triplicates against a standard curve of l-kynurenine concentrations (Sigma-Aldrich).

MLR was performed as previously described (26). Briefly, freshly isolated allogeneic CD4⁺ T cells were labeled with CFSE and incubated with DC at different ratios. DC were either incubated for 24 h in 96-well plates before T cells were added or they were additionally washed before MLR for the assay to be performed in fresh medium. In some experiments, magnetic beads, coated with anti-CD3 mAb or with anti-CD3 and anti-CD28 mAbs (27) at ratios of 1:1 or 1:10 (beads/T cells) were added. Alternatively, CD3/CD28-activated T cells were incubated with matDC or infDC in a transwell system. To neutralize the suppressive effect of infected DC, 1-methyl-tryptophan (10 μM, Sigma Aldrich), anti-IL-10 mAb (10 μg/ml), rofecoxib (10 μM), and rhIL-2 (20 U/ml) were given separately or in combination. After 3 days of culture, T cell proliferation was assessed by flow cytometry.

To study the effect of DC-derived soluble factors on CD3/CD28-induced T cell proliferation, CFSE-labeled CD4⁺ T cells were incubated with l-kynurenine (Sigma-Aldrich), IL-10 (R&D Systems), or PGE₂ in a range of concentrations in either RPMI 1640 (Invitrogen) or in tryptophan-free RPMI 1640 medium (BioWhittaker) for 3 days until T cell proliferation was assessed by flow cytometry. Alternatively, CD3/CD28-activated T cells were incubated with supernatants derived from matDC or infDC (in a dilution range from 1 to 100%). In some experiments, T cell proliferation was assessed by incorporation of BrdU (Roche Diagnostics): cells were pulsed with BrdU on day 2 of culture and harvested 24 h thereafter; subsequently, BrdU incorporation was assessed according to the manufacturer’s instructions.

The murine IL-2-dependent T cell line CTLL-2 was obtained from the American Type Culture Collection (TIB-214) and cultured in supernatants derived from DC supplemented with rhIL-2 as previously described (26). Cell proliferation was determined by cell counting and viability assessed by flow cytometry with propidium iodide (Sigma-Aldrich).

Immediately after infection, DC were lysed and the number of viable Listeria (initial bacterial burden) was determined in a CFU assay on blood agar. Consequently, bactericidal activity of DC was analyzed after various periods of time. Additionally, L.m. was incubated for 6 h in a tryptophan-free RPMI 1640 medium supplemented with different concentrations of l-tryptophan and l-kynurenine (Sigma Aldrich) and Listeria viability was controlled by CFU assay.

Statistical analysis was performed using R software package (version 2.3.0). To calculate statistical significance, a two-sample two-tailed t test was used. Comparisons with a p-value <0.05 were called statistically significant.

We have recently demonstrated that infection of human DC with L.m. leads to the induction of the immunoinhibitory enzyme IDO (10), while in other studies L.m. infection was associated with a stimulatory DC phenotype (7, 8). To clarify this apparent discrepancy we performed detailed comparative microarray analysis on myeloid DC infected with L.m. (infDC) (10), mature inhibitory DC stimulated with prostaglandin E₂ (PGE₂-DC) and expressing IDO (26), mature stimulatory DC lacking the expression of IDO (matDC), and immDC. For presentation of differentially expressed transcripts associated with inhibitory or stimulatory DC function (reviewed in detail in Refs. 16 ,17) expression values were standardized using Z score transformation at the probeset level and visualized as a heat map.

As shown in Fig. 1,A, infDC are enriched for numerous transcripts associated with stimulatory as well as inhibitory DC function. Among inhibitory genes PTGS2 (COX-2), IL10 (IL-10), and IL2RA (CD25), enzymes of tryptophan catabolism and inhibitory Ig-like transcripts were most strongly induced in infDC (Fig. 1,A). Furthermore, transcripts for TNF, IFN-β, and IL-23, which can exert antiinflammatory action (28), were up-regulated only in infDC. Stimulatory genes induced by infection included various chemokines, cell adhesion (ICAM1 and CD58) and costimulatory molecules (CD40, CD80, CD86, and TNFSF9). However, other important stimulatory molecules, such as CD83, CCR7, CCL19, TNF superfamily members, and CD1 family molecules were only up-regulated in noninfected stimulatory matDC (Fig. 1 A). Altogether, infDC present a rather specific transcriptional “signature” that distinguishes them from immDC, stimulatory matDC, and inhibitory PGE₂-DC: they simultaneously express numerous inhibitory as well as stimulatory molecules.

Assessment of surface receptor expression by flow cytometry demonstrated that infDC and stimulatory matDC expressed comparable amounts of costimulatory and adhesion molecules but differed in expression of CD83 (Fig. 1,B). To corroborate transcriptional regulation of inhibitory molecules, expression of CD25, IL-10, and COX-2 was assessed using IDO as a control marker for regulatory DC (17). Cell-surface CD25 is up-regulated in virtually all infDC and is coexpressed with intracellular IDO, while only 70% of inhibitory PGE₂-DC expressed CD25 (Fig. 1,C). ImmDC and matDC did not express CD25 and IDO. Supernatants from infDC contained significant amounts of kynurenine as a consequence of functional IDO expression (Fig. 1,D). Moreover, infDC secreted large amounts of soluble CD25 (sCD25) (Fig. 1,E). The onset of CD25 mRNA expression, as assessed by quantitative real-time PCR, was faster and the amount of secreted sCD25 after L.m. infection was higher (data not shown) as compared with PGE₂-DC (30). In contrast to CD25 and IDO, IL-10 and COX-2 were only induced by infDC but not by other DC subsets (Fig. 1, F and G). The kinetics of IL-10 secretion resembled that of sCD25 (data not shown). Altogether, infection of myeloid DC by L.m. induces stimulatory molecules together with numerous inhibitory pathways previously associated with regulatory DC function (16, 17).

In light of expression of inhibitory and stimulatory molecules by infDC, we next studied the impact of infDC on T cell activation. First, the stimulatory capacity of infDC in comparison to other DC subsets was assessed in an allogeneic MLR using purified CD4⁺ T cells and beads coated with low-dose anti-CD3 mAb, thus providing equal signal 1 for all conditions (27). Treatment of T cells with anti-CD3 beads alone did not induce T cell proliferation or T cell cytokine production (Fig. 2,A and data not shown). Co-cultures of T cells and matDC in the presence of anti-CD3 beads induced significant T cell proliferation (Fig. 2,A) comparable to beads coated with anti-CD3 and anti-CD28 Abs (Fig. 2,A). However, when T cells were cocultured with DC without anti-CD3 beads, proliferation was very low (Fig. 2 A), and therefore the following MLR experiments were performed with anti-CD3 coincubation.

To study the effect of infection on DC stimulatory capacity, matDC and infDC were simultaneously coincubated with allogeneic CFSE-labeled CD4⁺ T cells in the presence of anti-CD3 beads, and after 3 days T cell proliferation was assessed by flow cytometry. Significant reduction in T cell proliferation (up to 60%) was observed when T cells were coincubated with infDC, pointing out that infDC display an impaired ability to stimulate CD4⁺ T cells compared with matDC (Fig. 2,B). Reduced T cell proliferation was accompanied by reduced IFN-γ production (data not shown). In the second set of experiments, DC were preincubated for 24 h (to allow induction of inhibitory mechanisms after infection) and prior T cells were added for a coculture (Fig. 2,C). Under these conditions T cell proliferation was also significantly reduced (up to 75%) in the presence of infDC (Fig. 2,C), and this reduction was even more profound as in a direct coculture (Fig. 2 B).

Subsequently, we determined whether infected DC can suppress the proliferation of activated T cells when provided with a strong stimulatory signal (anti-CD3 plus anti-CD28 beads). For this purpose, DC were added to T cell/bead cocultures at the onset of stimulation (Fig. 2,D). T cells strongly proliferated when stimulated by anti-CD3/anti-CD28 beads, and addition of matDC did not significantly increase T cell proliferation (Fig. 2,D). In contrast, proliferation of CD3/CD28-activated T cells was significantly lower when cocultured with infected DC (Fig. 2,D). The suppression of CD3/CD28-induced T cell proliferation was even more evident when the stimulus was suboptimal (1:10 ratio of beads/ T cells, Fig. 2 D). Interestingly, when this assay was performed in a transwell system, where DC were separated from T cell/bead cocultures by a 0.4-μm polycarbonate membrane, the difference between matDC and infDC was substantially less prominent (∼10%, data not shown).

Timing of interaction between DC and T cells has an impact on T cell activation (29). Therefore, in the next set of experiments we studied whether pretreatment of T cells with a stimulatory signal (anti-CD3/anti-CD28 beads) for 24 h prior to MLR would have an impact on the suppressive effect of infDC (Fig. 2,E). In contrast to the experiments presented in Fig. 2,D, addition of infDC to prestimulated T cells did not alter the proliferation rate of the latter, even when the CD3/CD28 stimulus was suboptimal (Fig. 2 E). These data indicate that infDC cannot exert inhibitory effects upon previously activated T cells. Altogether, the outcome of encounter between T cells and infDC depends on the timing and T cell preactivation status.

Next we assessed whether infDC-derived soluble factors are sufficient to suppress the CD3/CD28-activated T cell proliferation (Fig. 3). For this purpose, CD3/CD28-activated T cells were incubated with a concentration range of DC-derived supernatants (0–100%) or with kynurenine, IL-10, and PGE₂; the latter assay was also performed in a tryptophan-free medium. Addition of infDC-derived supernatants to activated T cells resulted in a significant suppression of T cell proliferation, starting from 20% dilution, when compared with the same amount of matDC-derived supernatants (Fig. 3 A).

Furthermore, addition of kynurenine to CD3/CD28-activated T cells was sufficient to suppress the T cell proliferation, and this effect was even more profound when the assay was performed in a tryptophan-free medium (Fig. 3,B). Combination of low tryptophan and kynurenine (as a model of IDO enzymatic activity), as well as addition of IL-10 or PGE₂, led to a comparable and significant reduction of CD3/CD28-induced T cell proliferation (Fig. 3 B).

While inhibitory functions of IL-10 (30), IDO (14), and COX-2-mediated PGE₂ (27) have been well established (Fig. 3,B), the role of cell-surface CD25 and soluble CD25 is less clear (26, 31). IL-2-dependent CTLL-2 cells and heat-killed L.m. were used to address this question. Heat-killed L.m. induce similar amounts of CD25 in infDC (data not shown) but no IDO and COX-2 (10), so it was chosen to study the function of CD25 expression on infected DC. As shown in Fig. 3,C, the number of viable CTLL-2 cells was significantly reduced in cultures incubated with supernatants from infDC in comparison to matDC. As a consequence, an increase of dead CTLL-2 cells in supernatants from infDC was observed (Fig. 3 D). This was associated with a 30% reduction of IL-2 in culture supernatants as measured by ELISA (data not shown).

To determine the impact of individual inhibitory mechanisms, CD4⁺ T cells were cocultured with infDC together with IL-2 (as a CD25 antagonist) and inhibitors against COX-2, IL-10, or IDO. Blockade of IDO, COX-2, and IL-10 individually did not significantly change T cell proliferation in presence of infDC, and IL-2 was also without effect (Fig. 3,E). Only with inhibiting COX-2, IL-10, and IDO in the presence of IL-2 was an increase in T cell proliferation detected (Fig. 3 E). Overall, these data indicate that multiple inhibitory mechanisms that are capable to turn infDC into regulatory DC are induced during infection.

In addition to IFNs (14), we have recently revealed TNF as a potent inducer of the inhibitory molecule IDO (10). Here, we assessed the role of IFN-γ, TNF, and TNF-RI and TNF-RII on the induction of the IDO-associated inhibitory molecules CD25, IL-10, and COX-2 during DC infection. TNF-RI and TNF-RII are expressed in infDC (data not shown), while TNF and IFN-γ are secreted by infected DC with TNF production preceding IFN-γ production (10). Secreted TNF was completely sequestered by anti-TNF mAb (infliximab) and was not blocked by Abs against TNF receptors (anti-TNF-RI, anti-TNF-RII) or IFN-γ (Fig. 4,A). In contrast, IFN-γ was significantly reduced by neutralization of TNF or blocking of TNF-RI/RII, indicating that IFN-γ acts downstream of TNF receptor signaling in infDC (Fig. 4,B). To control the experiments addressing regulation of CD25, IL-10, and COX-2, we assessed IDO and kynurenine induction. Blockade of TNF or IFN-γ significantly reduced IDO expression and abrogated kynurenine production and this was mediated by both TNF-RI and TNF-RII signaling (Fig. 4, C and D). Expression of cell-surface and particularly of soluble CD25 was similarly reduced by the blockade of TNF and TNF-RI/II signaling (Fig. 4, E and F). However, in contrast to IDO, blockade of IFN-γ did not alter expression of CD25 and secretion of sCD25. This was similarly true for IL-10 secretion (Fig. 4,G) and COX-2 protein induction (Fig. 4 H), which were inhibited by blockade of TNF but not by IFN-γ. Dose dependency of TNF-mediated induction of CD25, IL-10, COX-2, and also IDO was demonstrated by increasing doses of infliximab or anti-TNF-R Abs (data not shown). Inhibition of IL-10 or COX-2 did not alter expression of CD25 or IDO (data not shown). Altogether, this places TNF as a central mediator of inhibitory molecules in DC during L.m. infection. TNF receptor-mediated signaling is required for induction of IFN-γ, COX-2, IL-10, CD25, and IDO, whereas IFN-γ is only necessary for IDO induction.

Based on up-regulation of molecules involved in tryptophan metabolism (IDO, KMO, KYNU; see Fig. 1,A), we postulated that tryptophan reduction might play a role in defense against L.m. (32). To assess Listeria uptake and bactericidal capacity, we first searched for an appropriate model system. The models were defined by a comparison of comprehensive transcriptional signatures of infected DC, described herein, and of differentially treated DC performed by ourselves or derived from publicly accessible databases, and by a regulatory DC (DCreg) phenotype, which we defined to be CD25/IDO/COX-2 and IL-10. Treatment of immDC with a combination of TNF and Pam₃, as well as with TNF, PGE₂, and Pam₃ (A. Popov, unpublished observations) or LPS (33) (data accessible at NCBI GEO database, accession no. GSE2706, at link provided above) resulted in up-regulation of IL2RA, INDO, IL10, and PTGS2 as well as genes coding for bactericidal peptides, toxic oxygen, and nitrogen radicals’ donors and acidic proteases (data not shown). Moreover, TNF/PGE₂/Pam₃ and LPS treatment resulted in the induction of a DCreg phenotype (Fig. 5,A). To address if a comparable phenotype can be induced in human DC, which are associated with infection foci but are not infected themselves, we incubated immDC or matDC with the supernatants derived from the allogeneic immDC, matDC, or infDC, respectively. All supernatants were double sterile filtered (0.2 μm) and seeded on agar plates to assure the absence of bacteria before the experiment (data not shown). As shown in Fig. 5 B, incubation of immDC with infDC-derived supernatants, but not with the supernatants derived from immDC or matDC, resulted in the induction of the DCreg phenotype. Comparable results were obtained in a transwell system when immDC (lower chamber) were cocultured with infDC (upper chamber) and when mature DC were treated with the supernatants derived from infected DC (data not shown).

For additional experiments we mostly used TNF/PGE₂/Pam3-induced DC as a model for IDO⁺ DCreg. Different DC subsets (immDC, matDC, DCreg) were infected with FITC-labeled L.m. at a multiplicity of infection of 10 and bacterial burden was analyzed over time after infection. As determined by uptake of bacteria, susceptibility to infection with L.m. differed between various DC populations (Fig. 5,C). As expected, immDC were more easily infected than IDO⁻ matDC and IDO⁺ DCreg. Among the latter two, IDO⁺ DCreg were most resistant against infection and phagocytosed less Listeria than did matDC. Within the first 30 min after infection, all DC subsets substantially reduced the initial bacterial load; however, on a cell-to-cell basis, IDO⁺ DCreg were significantly more efficient in controlling the phagocytosed Listeria than IDO⁻ immDC or IDO⁻ matDC (Fig. 5,D, relative number of intracellular viable bacteria derived 30 min after infection and normalized to the initial time point (t = 0) set to 100% is shown). Moreover, only IDO⁺ DCreg were able to significantly restrict the number of viable intracellular bacteria over time, while the number of viable L.m in matDC and immDC significantly increased during the course of infection (Fig. 5,E). Importantly, DC expressing a regulatory phenotype were equally efficient in “managing” intracellular Listeria independently of the factors used for inducing the DCreg phenotype, such as TNF/PGE₂/Pam₃, TNF/Pam₃, LPS, or supernatant-treated DCreg (Fig. 5,E and data not shown). In contrast, immature DC and DC matured with TNF or anti-CD40 appeared to lose the control over the intracellular bacteria (Fig. 5 E and data not shown). Overall, fewer IDO⁺ DCreg are infected by L.m. and those infected are more sufficient in reducing and controlling bacterial load, independently of the agents that induce the regulatory phenotype.

To assess the effect of IDO-mediated tryptophan catabolism on L.m., the survival of L.m. was determined in vitro under different concentrations of tryptophan and one of its key downstream metabolites, kynurenine. Growth of L.m. was influenced both by tryptophan starvation and toxic metabolites (as exemplified for kynurenine), with kynurenine being more potent than tryptophan reduction in suppressing bacterial growth (Fig. 5 F).

To address the in vivo relevance of our findings, we examined CD25 and COX-2 expression in lymph node specimens of patients with serologically confirmed cervicoglandular-type listeriosis with suppurative granuloma. These granulomas consist mostly of S100⁺CD11c⁺ DC and, to a lesser extent, of CD68⁺CD11c⁺ macrophages (10). In fact, most of the cells forming the outer border of the granuloma expressed the DC marker S100 and substantial amounts of IDO, CD25, and also COX-2 (Fig. 6,A). Of note, granuloma-forming DC did not express the DC activation marker CD83 (Fig. 6,A), supporting the transcriptional data and the results of cell-surface staining (see Fig. 1,B). Morphological assessment, as well as FoxP3 staining, excluded a massive infiltration of CD25⁺ regulatory T cells; however, some FoxP3⁺ cells were present within the granuloma wall and more FoxP3⁺ cells were detected between the granuloma (Fig. 6,A). Staining with mAbs specific for CD4, CD8, and CD56 demonstrated that T cells were displaced in the area between the granuloma, whereas NK cells were almost absent. Although we did not perform the costaining of FoxP3 and CD4, based on the spatial localization of both markers, FoxP3⁺ cells might be regulatory T cells. Using two-color immunofluorescence analysis, we demonstrated that CD25 and IDO are coexpressed by S100⁺ DC (Fig. 6 B). Double staining for CD11c and CD25 confirmed CD25⁺ DC to be of myeloid origin (data not shown). CD68⁺ macrophages within the granuloma ringwall also expressed CD25, but to a lesser extent as DC (data not shown). Macrophages generated from monocytes and infected with L.m. expressed surface CD25, secreted sCD25, and expressed COX-2, albeit to a lesser extent than infDC (data not shown). Altogether, myeloid DC and macrophages within the granuloma express multiple inhibitory molecules associated with a regulatory phenotype of DC (15).

If formation and maintenance of granuloma rely on inhibitory pathways induced in myeloid cells, we hypothesized that other pathogens associated with granuloma formation would induce comparable molecular programs. To address this question, we applied the transcriptional signature of stimulatory and inhibitory genes established for L.m. infection (see Fig. 1,A) to a publicly available dataset describing the genome-wide response of human DC and macrophages to phylogenetically distinct pathogens (data accessible at NCBI GEO database, accession no. GSE360, at link provided above) (34). Infection of DC with M. tuberculosis and macrophages with Leishmania major most closely resembled the expression pattern observed for L.m. infection. Induction of transcripts for TNF, IFN-γ, CD25, COX-2, IL-10, and enzymes of the tryptophan catabolism (IDO, KMO, KYNU) was apparent in DC infected with either of these three pathogens (Fig. 7,A). Similarly, genes encoding stimulatory molecules (e.g., CD40, CD80, CD86, ICAM1, and CD58) induced in human DC by L.m. infection were also induced by M. tuberculosis and L. major. Other infections assessed in this experiment (Fig. 7,A) clearly differed in the regulation of stimulatory and inhibitory pathways. To assess the in vivo relevance of these findings, we assessed expression of IDO, CD25, and COX-2, as well as DC marker S100 and macrophage marker CD68, in granuloma in tuberculosis. The structure of tubercular granuloma is regulated within time: as the granuloma matures, collagen-rich fibers substitute the cells at the periphery, which leads to granuloma organization (35). To determine whether expression of regulatory proteins by granuloma-forming cells is also regulated over time, we examined their expression in the early as well as late-stage granuloma. In contrast to listeriosis, CD68⁺ macrophages were the major cellular component of early granuloma in tuberculosis, while only few S100⁺ DC were present; expression of IDO was very prominent and CD25 was moderately expressed (Fig. 7,B), whereas COX-2 staining was rather faint (data not shown). Costaining experiments revealed that CD25⁺ cells do express IDO as well as CD11c, confirming their myeloid origin (data not shown). However, regulatory cells seem to be an early event during granuloma formation, since CD25⁺IDO⁺ cells were almost absent in late-stage granuloma, although macrophages and DC were still present (Fig. 7 C).

Tryptophan-catabolizing DC are a major component of granuloma in human listeriosis (10). Herein we demonstrate that granuloma-forming myeloid cells, either DC or macrophages, in human listeriosis and tuberculosis are characterized by coinduction of multiple inhibitory pathways, including CD25 secretion, IL-10 expression, COX-2-dependent mechanisms, as well as tryptophan catabolism. Using listeriosis as a model we linked the induction of this regulatory phenotype of myeloid DC to infection and established TNF as the major mediator of inhibitory proteins, while IFN-γ, which is downstream of TNF, is only required for induction of the tryptophan-catabolizing enzyme IDO. Regulatory DC induced during infection with L.m. are strong inhibitors of T cell activation, and their regulatory function can be reversed only by simultaneous blockade of several inhibitory proteins. The modest recovery of T cell proliferation when blocking four regulatory proteins at once (Fig. 3,E) underscores the predominance of the regulatory DC phenotype over the stimulatory one and points out that even more regulatory mechanisms might be involved (Fig. 1 A). Regulatory DC are not only endowed with the ability to suppress T cell function, but also to suppress bacterial infection. While tryptophan metabolism in regulatory DC seems to play an important role in reducing bacterial burden during infection, other unknown factors must account for the increased resistance of regulatory DC to infection. Altogether, these findings advocate that regulatory myeloid cells involved in granuloma formation during listeriosis and tuberculosis are provided with multiple inhibitory mechanisms evolved to protect the host from disseminating infection while at the same time inhibiting granuloma destruction by T cells.

Accumulating data suggest that DC, in addition to macrophages, play an important role in the pathogenesis of granulomatous diseases such as listeriosis, tuberculosis, or cat-scratch disease (4, 10, 36, 37). Infection of DC with various pathogens was associated with induction of stimulatory effects on DC function by these pathogens or their components (7, 8, 38). We corroborate these data, demonstrating that important stimulatory molecules (e.g., CD40, CD80, and CD86) are induced in DC and macrophages during infection with L.m. and M. tuberculosis (Figs. 1 and 7). However, simultaneous induction of several inhibitory pathways (IL-10, COX-2, CD25, and IDO) does counterbalance these stimulatory pathways toward inhibition of T cell function. Furthermore, IL-10 was reported to inhibit pathogen-specific T cell responses (39), and prostaglandins, products of COX-2-mediated arachidonic acid metabolism, were shown to suppress cellular immunity to Listeria (40). At the same time both factors directly suppress T cell proliferation (27, 30) and are also involved in the induction of regulatory T cells (41, 42). For CD25, we demonstrated that cell-surface and soluble CD25 induced by infection can function as an IL-2 scavenger receptor. This is in line with clinical evidence linking increased levels of sCD25 to immunosuppression in infectious (43, 44, 45) and malignant diseases (46, 47). Of note, induction of CD25 and sCD25, as well as IDO, is not restricted to mo-DC, but has been observed in BDCA-1⁺ primary myeloid DC stimulated ex vivo with lipoteichoic acid derived from L.m. (A. Popov, unpublished observation) or with PGE₂ (26).

The balance between the stimulatory and inhibitory phenotype of regulatory DC does not result only from the expression of stimulatory or inhibitory molecules, but also from the kinetics of interaction with CD4⁺ T cells as well as the activation status of the T cells. If the inhibitory phenotype of regulatory DC is elicited before encounter of T cells, their suppressive effect will be clearly more pronounced. Conversely, preactivation of T cells via TCR and costimulatory signals can neutralize the inhibitory effect of regulatory DC.

The bactericidal effect of tryptophan depletion in different cells expressing the key enzymes of the tryptophan pathway has been recognized for several pathogens (32). For L.m., conflicting data concerning the role of tryptophan metabolism have been reported. Although L.m. was shown to be auxotrophic for tryptophan (48), growth of the virulent strains of Listeria is regarded to be tryptophan independent (49) due to expression of tryptophan synthase genes (trp) enabling autonomous tryptophan production (50). Herein we demonstrate that IDO⁺ DC are most effective in reducing bacterial burden and infection. Moreover, bacterial growth appears to be predominantly regulated by the accumulation of its toxic metabolite kynurenine. A lower sensitivity to tryptophan starvation compared with the effect of toxic metabolites is in line with previous findings (50). Gram-positive bacteria can produce tryptophan autonomously, making them insensitive to fluctuations in tryptophan concentration in the environment but still vulnerable to the accumulation of toxic metabolites (50). However, IDO⁻ DC were capable of initially reducing bacterial burden postinfection, suggesting that other mechanisms are utilized by regulatory DC during pathogen containment. Altogether, these data emphasize the importance of tryptophan metabolism for regulatory DC function; however, both T cell suppression as well as pathogen inhibition rely on a multitude of inhibitory mechanisms acting in concert.

An intriguing finding was the hierarchy of signals necessary to induce inhibitory molecules after Listeria infection. Clearly, TNF is a main mediator inducing CD25, COX-2, IL-10, and IDO in DC. Furthermore, both TNF receptors are essential for TNF signaling since both had to be blocked to achieve maximum reduction of inhibitory proteins. Most striking, however, was the finding that IFN-γ is clearly downstream of TNF signaling and only governs induction of IDO but not the other inhibitory pathways. The major function of granulomatous structures is the containment of pathogens that otherwise cannot be eradicated by the immune system, thereby preventing an uncontrolled systemic spreading of the pathogen (51, 52). TNF has been recognized as an important factor governing formation and maintenance of granuloma containing intracellular pathogens such as M. tuberculosis or L.m. (53, 54). However, strict TNF dependency of multiple important inhibitory mechanisms in regulatory DC came somewhat as a surprise since this might be the granuloma’s Achilles’ heel. Clinical evidence is in line with our experimental data. In patients with rheumatoid arthritis elevated sCD25 serum levels are significantly reduced after treatment with the TNF-neutralizing drug infliximab (55). Suppression of IDO and IL-10 during infliximab therapy was recently reported for patients with Crohn’s disease (56, 57). More dramatic, a severe side effect of anti-TNF therapy is in fact exacerbation of granulomatous listeriosis or tuberculosis mediated by T cells destroying the granulomas (58, 59).

Collectively, this study provides strong evidence that intracellular pathogens such as M. tuberculosis and L.m. induce a specific transcriptional program in myeloid DC and macrophages characterized by a functional preponderance of multiple inhibitory mechanisms. On the one hand, these myeloid regulatory cells are equipped to suppress unwanted T cell attacks against granulomatous structures; on the other hand, they prohibit pathogens from spreading throughout the host. The exact mechanisms, however, should be studied in the proper animal models. Of particular interest for further research will be the exploitation of the yet unknown pathways of myeloid regulatory cells conferring resistance to infection. This might lead to the discovery of novel strategies protecting other cells from overwhelming infection with these devastating intracellular pathogens.

We thank our colleagues from the Center for Transfusion Medicine for providing us with peripheral blood products, and we are grateful to all blood donors. We thank Mirela Stecki and Julia Claasen for technical assistance. We are in debt to Dr. Alexander Poyarkov for his invaluable help during the study design and manuscript preparation. We also thank Drs. K. Schrör and J. Meyer-Kirchrath from the Institute for Pharmacology and Clinical Pharmacology, University of Düsseldorf, Germany, for providing rofecoxib, and Daniela Eggle for critical reading of the manuscript.

The authors have no financial conflicts of interest.