PD-L1 expression on tolerogenic APCs is controlled by STAT-3



During infection, TLR agonists are released and trigger mature as well as differentiating innate immune cells. Early encounter with TLR agonists (R848; LPS) blocks conventional differentiation of CD14+ monocytes into immature dendritic cells (iDCs) resulting in a deviated phenotype. We and others characterized these APCs (TLR-APC) by a retained expression of CD14 and a lack of CD1a. Here, we show in addition, expression of programmed death ligand-1 (PD-L1). TLR-APCs failed to induce T-cell proliferation and furthermore were able to induce CD25+Foxp3+ T regulatory cells (Tregs). Since PD-L1 is described as a key negative regulator and inducer of tolerance, we further analyzed its regulation. PD-L1 expression was regulated in a MAPK/cytokine/STAT-3-dependent manner: high levels of IL-6 and IL-10 that signal via STAT-3 were produced by TLR-APCs. Blocking of STAT-3 activation prevented PD-L1 expression. Moreover, chromatin immunoprecipitation revealed direct binding of STAT-3 to the PD-L1 promoter. Those findings indicate a pivotal role of STAT-3 in regulating PD-L1 expression. MAPKs were indirectly engaged, as blocking of p38 and p44/42 MAPKs decreased IL-6 and IL-10 thus reducing STAT-3 activation and subsequent PD-L1 expression. Hence, during DC differentiation TLR agonists induce a STAT-3-mediated expression of PD-L1 and favor the development of tolerogenic APCs.


DC are initiators and modulators of the adaptive immune response 1. They are able to induce T-cell activation as well as T-cell tolerance. During infection, DCs are confronted with pathogen-associated molecular patterns (PAMP), which in turn trigger effector functions in innate immune cells. For example, immature DCs (iDCs) generated from monocytes by in vitro culture with GM-CSF and IL-4 (G4) mature and become fully activated upon stimulation with TLR agonists. Mature DCs (mDCs) in turn activate most efficiently naïve T cells 2. However, during infection induction of inhibitory immune pathways can also be observed 3, 4. Here, we investigate an alternative TLR-induced APC phenotype, which inhibits immune reactivity. It has been shown that encounter of monocytes with LPS during the very beginning of the differentiation process blocks conventional differentiation to iDCs. A phenotypically distinct APC type (TLR-APC) is generated, characterized by a CD1aCD14+ phenotype 5–7. Activation of p38 MAPK, the secretion of IL-10 and the inactivation of ERK and NF-kB 7 have been correlated with the generation of TLR-APCs. LPS-treated cells showed in addition an intense STAT-3 phosphorylation.

Differentiation processes of DCs are plastic and can be influenced by various factors, e.g. cytokines. Many cytokines mediate their cellular response via the JAK/STAT signaling pathway thereby controlling the status of transcription and cellular differentiation. For instance, during the maturation of DCs, a switch occurs from constitutive activated STAT-6 in iDCs to a pre-dominant activation of STAT-1 in mDCs 8. This indicates that the activation pattern of STATs critically determines the phenotype and function of DCs. It has been shown that STAT-3 activation is often associated with tolerogenic functions 9–11. Indeed IL-6 and IL-10 are able to direct DC differentiation toward a tolerogenic phenotype 12–14. Several other means that induce tolerogenic DCs have been described: e.g. vitamin D3-derived DCs 15, TGF-β-induced DCs 16, TNF-α-induced semi-mature DCs 17 or iDCs 18. They all share the ability to negatively regulate T-cell responses, yet their phenotypes, cytokine profiles and thus their mode of action are divergent. IL-6- or IL-10-derived DCs for example have a similar phenotype as TLR-APCs 19–21. But differences in respect of CD86 13, 20 and IL-12 have been identified 14, 22.

Programmed death ligand-1 (PD-L1) is mainly described as a negative regulatory molecule and it has been shown frequently that the expression of PD-L1 is linked with the ability of DCs to induce tolerance 23–25. PD-L1 belongs to the co-stimulatory/co-inhibitory B7 family and is expressed on a variety of tissues and cells. So far, no general pathway is known which controls PD-L1 expression. Depending on stimulus and cell type, the expression of PD-L1 was found to correlate with various signaling molecules: p44/42 and/or p38 MAPKs 26, 27 or STAT-1, STAT-3 and IRF-1 28–30.

Here, we characterize the phenotype and function of APCs induced by an early TLR-mediated block of conventional differentiation of iDC. These TLR-APCs had a tolerogenic phenotype and could be induced by different classes of TLR-agonists (TLR7/8 R848 and TLR4 LPS). PD-L1 expression correlated with the functional properties of these APCs. Furthermore, we show that TLR-induced expression of PD-L1 is regulated in an IL-6-, IL-10- and STAT-3-dependent manner.


Phenotype of TLR-APCs

In a preceding publication, we have shown that cytokine-driven differentiation of DCs from monocytes can be deviated by simultaneous stimulation with TLR agonists. When isolated CD14+ monocytes were stimulated with GM-CSF and IL-4 (G4) in the presence of LPS, cells failed to upregulate the DC marker CD1a and retained CD14 expression 5, which contrasts the phenotype obtained with G4 stimulation alone. When we tested other TLR agonists, we found that the TLR7/8 small molecular weight agonist R848 influences the differentiation of DCs in a comparable manner (Fig. 1B and C). R848 inhibitory effects on CD1a expression were dose dependent with an optimum of 1 μg/mL (Supporting Information Fig. 1A). The time frame of inhibitory effects was limited until three days after addition of GM-CSF and IL-4 (Supporting Information Fig. 1B).

Figure 1.

Comparison of CD1a, CD14 expression. (A) Gating strategy for all flow cytometric analyses. Cells were gated on R1 population using FSC and SSC and then gated on CD1a, CD14 or other parameters. Measurement of CD1a (B) and CD14 (C) expression by flow cytometry in iDCs that were classically differentiated by addition of GM-CSF and IL-4 (G4) or R848-APCs that had been stimulated at day 0 with R848 prior to differentiation with G4. Analysis was done at day 6. The data are representative for at least three independent experiments.

Functional analysis of TLR-APCs

To test the functional properties of R848-generated TLR-APCs, we first analyzed their ability to induce proliferation in a mixed leukocyte reaction with allogeneic responder cells. TLR-APCs proved to be only weak stimulators of PBMCs in comparison to iDCs (Fig. 2A). To examine how TLR-APCs affect T-cell subset responses, we performed mixed leukocyte reactions with allogeneic CD4+ or CD8+ responder T cells. TLR-APCs induced only weak proliferative responses in CD4+ T cells (Fig. 2B). However, CD8+ T-cell proliferation, as compared to the proliferation induced by iDCs, was not significantly changed (Fig. 2C). Thus, the deficit of TLR-APCs to induce proliferative responses seems to be linked to CD4+ T cells. Since CD8+ T cells failed to respond in cultures with CD4+ cells, it was suggestive that TLR-APCs might induce CD4+ T cells with suppressive properties like CD4+CD25+Foxp3+ Tregs. To check this hypothesis we analyzed whether T cells cultured with TLR-APCs express CD25 and Foxp3 after allogeneic stimulation. Indeed, we could detect a CD4+CD25+T-cell population that expressed FoxP3 (Fig. 2D). CD4+CD25 T cells in contrast failed to express significant amounts of FoxP3 (Fig. 2E).

Figure 2.

R848-APCs fail to induce T-cell proliferation and additionaly induce Tregs. (A–C) Proliferation of responder cells (A, PBMCs; B, CD4+ T cells; C, CD8+ T cells) was measured by [3H]-thymidine-uptake after 96 h of stimulation with allogeneic iDCs (G4), R848-APCs or responder cells alone (no). Shown are the respective mean values of three individual experiments. *p<0.05 by Student's t-test. (D and E) Isolated CD4+ T cells were co-cultured with R848-APCs. After 5 days, FoxP3 expression of CD4+CD25+ (D) or CD4+CD25- (E) cells was measured by flow cytometry (gray, isotype; black line, FoxP3 expression). (F and G) Transfer experiments of Tregs generated in the presence of iDCs (G4) or R848-APCs. Freshly isolated CD4+ T-cells were co-cultured with iDCs (G4) or R848-APCs for 5 days. At day 5, CD25+ T cells were isolated (indicated as Tr) and cultured with CFSE-labeled CD4+ T cells (indicated as Te) from the same donor in the presence of activation beads. Loss of CFSE was measured after 5 days by flow cytometry. Presented data are representative for two independent analyses. (F) The overlay depicts an example of an analysis demonstrating the suppression of CD4 T-cell proliferation after addition of CD25+ T cells from the co-culture with TLR-APCs (Te+Tr, solid line) compared to the proliferation of bead-activated T cells alone (Te, in gray). (G) Titration of CD25+ and CD25- T-cells added to CFSE-labeled CD4+ T cells. Cells were activated with beads. Displayed is the percentage of cells proliferating compared to bead activated cells alone (reflecting 100%). The Control (noTr) are bead-activated, CFSE-labeled CD4+ T cells alone.

To confirm the functionality of Tregs induced by TLR-APCs, we performed transfer experiments: allogeneic CD4+ T cells were co-cultured for 7 days with TLR-APCs. Thereafter, CD25+ and CD25- cells from each culture were isolated and added at graded amounts to indicator cultures. These consisted of responder CD4+ T cells from the same donor (thawed), which were labeled with carboxyfluoroscein succinimidyl ester (CFSE) and stimulated with a mixture of antibodies (CD3/CD28/CD2). After 5 days, CFSE staining was measured. The overlay in Fig. 2F depicts an example of an analysis demonstrating the suppression of T-cell proliferation after addition of CD25+ T cells from the co-culture with TLR-APCs. The complete titration is given in Fig. 2G revealing a clear dose-dependent inhibition of proliferation. Thus, the data demonstrated clearly that the CD25+ cells isolated from the co-culture with TLR-APCs inhibited effectively primary T-cell responses. CD25+ T cells isolated from cultures with iDCs showed less regulatory properties (Fig. 2G). CD25 T cells were not able to block T-cell proliferation independent from which co-culture they were isolated from. Thus, TLR-APCs are not only weak stimulators of MLC but are further capable to induce CD4+CD25+ Tregs.

In addition to the functional assays, we analyzed IL-2 production, since IL-2 is required for expansion of Tregs and their suppressive function 31. The co-cultures of T cells and R848-APCs showed higher amounts of IL-2 compared to the co-cultures of T cells and iDCs (Supporting Information Fig. 2).

Surface expression of co-stimulatory and co-inhibitory molecules on TLR-APCs

Next, we analyzed the co-stimulatory and co-inhibitory properties of TLR-APCs. We compared the expression of the co-stimulatory and co-inhibitory B7 family members (PD-L1, PD-L2, B7-H3, B7-H4, CD80, CD86 and ICOS-L; Fig. 3A) of iDCs and TLR-APCs. The differences of PD-L1 expression were remarkable. R848 generated cells showed very high expression levels of PD-L1 (Supporting Information Fig. 3). To exclude that PD-L1 expression is exclusively linked to the TLR7/8 agonist R848 we additionally measured PD-L1 expression in LPS generated TLR-APCs (Supporting Information Fig. 3). In general, LPS-generated TLR-APCs showed a similar but less pronounced phenotype.

Figure 3.

Expression of co-stimulatory and co-inhibitory molecules. (A and B) The expression of different surface molecules on R848-APCs (R848) compared to iDCs (G4) was measured by flow cytometry at day 6, except for MHC II expression (measured at day 3). Displayed are at least three independent experiments. *p<0.05 by Student's t-test. (A) Analysis of the B7 family. (B) Expression of CD40, CD252 (OX40L) and MHCII. (C) Allogeneic co-culture experiments of CD3+ T cells with R848-APCs in the presence of 10 μg/mL anti-PD-L1. T-cell proliferation was determined by [3H]-thymidine-uptake after 96 h. Shown are the respective mean values of three individual experiments *p<0.05 by Student's t-test.

Additionally, we analyzed the expression of CD40, CD252 and MHCII, which are important for the activation of T cells (Fig. 3B). MHC class II molecules are required for effective antigen presentation. While G4-stimulated cells showed high expression, R848-APC had a reduced number of MHC class II molecules, which could explain their low stimulatory potency. However, since PD-L1 is correlated with tolerance induction 32, we also tested, whether PD-L1-dependent signaling contributes to the weak T-cell proliferation observed. Blockade of PD-L1 was effective to enhance T-cell proliferation in the presence of R848-APCs (Fig. 3C). Thus, reduced MHC class II expression and upregulation of PD-L1 are characteristics for TLR-APCs and their changed functional capacities.

Cytokine release and mimicking of TLR-APCs phenotype with cytokines

To further analyze the mechanisms of induction of the tolerogenic APC phenotype, we next analyzed release of cytokines upon initial TLR trigger. APCs generated in the presence of R848 secreted high amounts of pro-inflammatory cytokines (IL-6, TNF and IL-12p40) as well as immunosuppressive cytokines (IL-10) (Fig. 4A–D). Secretion of IL-6 was remarkably high (Fig. 4A). In order to determine whether auto- or paracrine active cytokines directly mimic the effect of R848 we added cytokines alone or cytokine mixtures to G4-stimulated cell cultures. While single addition of cytokines (IL-6 or IL-10) only partially induced the TLR-APC phenotype, a combination of both was almost similar effective to stimulation with R848 (Supporting Information Fig. 4).

Figure 4.

Cytokine release. (A–D) Supernatants of iDCs (G4) and R848-APCS were harvested after 2 h, 1, 3, and 6 days and were analyzed for IL-6 (A), IL-10 (B), IL-12p40 (C) and TNF (D) production by ELISA. Displayed is one of three independent experiments (mean±SD).

Pattern of MAPK activation and their role in cytokine production

In order to further define the signal requirement for induction of TLR-APCs, we analyzed the pattern of MAPKs, known to be involved in TLR-mediated cytokine release 33. MAPKs are in addition important for differentiation processes. It was striking that the pattern of MAPK activation was clearly different between R848-APCs and conventional iDCs. Each MAPK exhibited a special pattern of activation (Fig. 5A): differentiation of monocytes in the presence of G4 and R848 showed an early and prolonged phosphorylation of p38, whereas in G4-generated cells p38 phosphorylation was only detectable within the first 30 min. The activation pattern of p44/42 differed completely from p38 phosphorylation. p44/42 phosphorylation was only visible during the initial 15 min in R848-APCs and in contrast for 24 h in iDCs. Phosphorylation of SAPK/JNK was only detectable in R848-APCs and only for a short period.

Figure 5.

Analysis of MAPKs activation and cytokine detection after p38 and p44/42 MAPK inhibition. (A) For western blot analysis equal amounts of protein lysates were blotted and probed with un- and phosphorylated antibodies for p38, p44/42 and SAPK/JNK. Cells were lysed in RIPA buffer 5, 15, 30 min and 1 day after stimulation with GM-CSF and IL-4 (G4) or with GM-CSF and IL-4 plus R848 (R848); n.s., non-stimulated. (B and D) Purified monocytes were pre-treated for 1 h with 10 μM of p44/42 inhibitor (UO126, UO), p38 inhibitor (SB203580, SB) alone or in combination. Afterwards, they were stimulated with GM-CSF and IL-4 plus R848. Supernatants were taken at day 1 and analyzed for IL-6 (B), IL-10 (C) and IL-12p40 (D) release (mean±SD). Displayed is one of three independent experiments.

Inhibition of the two MAPK pathways (p38, p44/42) with pharmacological p38 (SB203580, SB) and p44/42 inhibitors (UO126, UO) resulted in markedly reduced secretion of IL-6 (Fig. 5B) and IL-10 (Fig. 5C), at least when both MAPKs p38 and p44/p42 were blocked. Similar results were obtained when the cells were stimulated with LPS plus G4 (data not shown). IL-12p40 release in contrast was not diminished (Fig. 5D) but even slightly increased.

Role of MAPKs for the surface expression of CD1a, CD14 and PD-L1

The reduced cytokine release after MAPK inhibition correlated with reduced surface expression of CD14 and PD-L1. FACS analyses revealed that preservation of CD14 expression was blocked almost completely by the addition of SB and UO (Fig. 6A). In addition, upregulation of PD-L1 expression was also blocked by MAPK inhibition (Fig. 6B). This was not due to the toxicity of the inhibitors, since cellular viability as measured with the dye MTT was not affected (Supporting Information Fig. 5A). CD1a expression was not altered (data not shown).

Figure 6.

CD14 and PD-L1 expression in TLR-APCs are reduced after inhibition of MAPK p38 and p44/42. Purified monocytes were pre-treated for 1 h with 10 μM of p44/42 inhibitor (UO126, UO), p38 inhibitor (SB203580, SB) alone or in combination. Afterwards, they were stimulated for 1 day with GM-CSF and IL-4 plus R848. 0.2% DMSO was used as solvent control. Modifications of CD14 (A) and PD-L1 (B) expression were detected by flow cytometry. Mean values are shown as box blot (n=5). *p<0.05 by Student's t-test.

Activation pattern of STATs

The results so far indicated that IL-6 and IL-10 are important for the induction of the TLR-APC phenotype. Both cytokines are known to signal via STAT-3. We therefore analyzed expression and phosphorylation of STAT molecules (STAT-1, -3, -5 and -6). The STAT activation pattern of iDCs and TLR-APCs differed significantly (Fig. 7): differentiation of DCs in the presence of R848 resulted in an almost constitutive activation of STAT-3. In contrast, STAT-1 tyrosine phosphorylation was much shorter compared to STAT-3 phosphorylation (1 h–day 1). Regarding STAT-6 activation no significant differences between TLR-APCs and iDCs were detected (data not shown). In contrast, during the whole differentiation process, STAT-5-activation dominated in iDCs and was much lower in TLR-APC. Hence, the comparison of the STAT activation pattern in iDCs and TLR-APCs revealed a prevailing STAT-5 activation in iDCs and a dominant STAT-3 activation in TLR-APCs.

Figure 7.

STAT activation pattern. Analysis of un- and phosphorylated STAT-1, 3 and 5 from iDCs (G4) and TLR-APCs (R848). Cells were lysed in RIPA buffer after 1 h, 1 day, 4 days and 6 days for western blot. n.s., non-stimulated. Equal amounts of protein lysates were blotted and probed with antibodies for STAT-1, 3, and 5. For un- and phosphorylated STAT molecules the same lysates were used but different membranes. The data are representative for three independent experiments.

Role of STAT-3 for the surface expression of CD1a, CD14 and PD-L1

To further corroborate the link between STAT-3 activation and expression of CD14 and PD-L1, we performed blocking experiments of STAT-3 with the chemical inhibitor JSI-124. After addition of JSI-124 expression of CD14 was not sustained (Fig. 8A) and upregulation of PD-L1 expression was prevented (Fig. 8B). CD1a expression was unaffected (data not shown). Treatment with the inhibitor JSI-124 did not compromise cell viability (Supporting Information Fig. 5B).

Figure 8.

Inhibition of PD-L1 expression after STAT-3 blockade and direct binding of STAT-3 at the PD-L1 promoter. (A and B) Purified monocytes were pre-treated for 2 h with 200 nM of STAT-3 inhibitor (JSI-124). Afterwards, they were stimulated for 1 day with GM-CSF and IL-4 plus R848. 0.02% EtOH was used as solvent control. Modifications of CD14 (A) and PD-L1 (B) expression were detected by flow cytometry. Mean values are shown as box blot (n=3). *p<0.05 by Student's t-test. (C and D) Binding quality of STAT-3 and STAT-1 at the PD-L1 promoter. CD14+ monocytes were stimulated with GM-CSF and IL-4 (G4) or additionally with R848 (R848) for 4 h, thereafter cells were fixed with formaldehyde. For ChIP antibodies against STAT-1 and STAT-3 were used. Binding of STAT-3 and STAT-1 was analyzed by quantitative PCR with promoter-specific primers for PD-L1 (percentage of input; mean±SD). PD-L1 promoter results are representative for three independent analyses. NC, negative control without antibodies; unst, non-stimulated.

To close the link between STAT-3 activation and induction of PD-L1 expression we used chromatin immunoprecipitation (ChIP) assay to determine the ability of STAT-3 to bind to the PD-L1 promoter. We found that STAT-3 was rapidly recruited to the PD-L1 promoter (Fig. 8C). Since STAT-1 is known to be involved in PD-L1 expression too and since STAT-1 was also activated we checked the binding activity of STAT-1 to the PD-L1 promoter (Fig. 8D). However, we found that STAT-1 binding was minor compared to STAT-3 and nearly no differences in STAT-1 binding between iDCs and TLR-APCs were detectable.

STAT-3 phosphorylation after blocking of the two MAPK p38 and p44/42

From the results so far, we concluded that STAT-3 has a central role for the formation of the TLR-APC phenotype. On the other hand, inhibition of MAPKs with the pharmacological inhibitor SB203580 (MAPK p38) and UO126 (MAPK p44/42) had the same effect as STAT-3 inhibition: the failure to sustain expression of CD14 and the prevention of PD-L1 expression. To link both effects with each other, we tested whether suppression of cytokine production (especially of IL-6 and IL-10) after MAPK inhibition influenced the status of STAT-3 activation.

After combined blockade of p38 and p44/42 tyrosine phosphorylation of STAT-3 was reduced markedly. The same pattern was found when LPS instead of R848 was used to induce TLR-APC (Fig. 9A). In contrast, STAT-5 phosphorylation was not altered (Fig. 9B). Consequently, the reduction of STAT-3 tyrosine phosphorylation after inhibition of p38 and p44/42 MAPKs could be prevented by the addition of exogenous IL-6 and IL-10 (Fig. 9C).

Figure 9.

Inhibition of MAPK p38 and p44/42 prevents STAT-3 phosphorylation. (A–C) Purified monocytes were pre-treated for 1 h with 10 μM of p44/42 inhibitor (UO126, UO), p38 inhibitor (SB203580, SB) alone or in combination. Afterwards, they were stimulated with GM-CSF and IL-4 plus R848/LPS or additionally with IL-6 (100 ng/mL) plus IL-10 (50 ng/mL). 0.2% DMSO was used as solvent control. Samples for western blot were prepared after 12 h. n.s., non-stimulated. Displayed is one representative experiment out of three independent measurements. Shown are the corresponding loading controls of STAT-3 tyrosine phosphorylated blots. (A) TLR-APC lysates were analyzed for STAT-3 activation after p38 and p44/42 MAPKs inhibition. (B) STAT-5 activation was measured after p44/42 and p38 inhibition. (C) SB and UO pre-treated cells were stimulated with GM-CSF, IL-4, R848±IL-6 and IL-10. Lysates were analyzed for STAT-3 activation.


It has been shown previously that the TLR4 ligand LPS added at early time points during the GM-CSF and IL-4-driven differentiation of monocytes into iDCs alter the differentiation process 5–7. APCs (TLR-APC) are generated that express no CD1a, but remain CD14 positive. We found that other TLR ligands especially the TLR7/8 small molecular weight agonist R848 influences the differentiation of DCs in a comparable manner (Fig. 1). By using allogeneic MLRs we show that R848-APCs were weak stimulators for CD4+T cells (Fig. 2B). However, CD8+ T cells were activated almost equally by iDCs and TLR-APCs (Fig. 2C). This suggested that TLR-APCs might induce inhibitory T cells in the CD4+ T-cell population. Indeed, the experiments revealed that TLR-APCs generated Tregs (Fig. 2D–G). Thus, TLR-APCs display a tolerogenic APC phenotype. During induction of TLR-APCs, we found a strong IL-6 production, which is at first glance conflicting to our finding that TLR-APCs induce Tregs. It is known that both Tregs and Th17 cells are induced by TGF-β, yet in the presence of IL-6 the balance between Th17 cells and Tregs is shifted toward Th17 cells 34, 35. However, other cytokines counteract the IL-6-driven induction of Th17 cells. IL-2 for example has been shown to block Th17 differentiation in the presence of TGF-β and IL-6 36. In that context, it is interesting, that cultures of T cells with TLR-APCs contained high amounts of IL-2 (Supporting Information Fig. 2), suggesting that this mixture of cytokines indeed promotes induction of Tregs.

Several studies link PD-L1 expression directly to the development and function of Tregs 37, 38. As TLR-APCs express high levels of PD-L1 (Fig. 3A), this could explain in turn their ability to induce Tregs. While PD-L1 expression might favor Treg generation, the reduced MHC II expression on TLR-APCs (Fig. 3B) could account for their inability to induce effectively primary T-cell responses. Interestingly, it has been shown in DCs that the expression of MHC II can be negatively influenced by the IL-6/STAT-3 pathway 39, which seems to be also important in R848-APCs. Other members of the B7 family in addition to PD-L1 are described as co-inhibitory and are also increased in R848-APCs: PD-L2 (B7-DC) 25, B7-H3 40 and B7-H4 41 (Fig. 3A). The role of PD-L2 seems to be of particular interest, since the genes for PD-L2 and PD-L1 are closely linked 42 and both molecules bind the same receptor (PD-1). Besides co-inhibitory also co-stimulatory molecules like CD80 (Fig. 3A) and CD40 (Fig. 3B) are upregulated. However, co-inhibitory molecules seem to be expressed preferentially in R848-APCs. This is in accordance with recent evidences that the ratio between co-inhibitory and co-stimulatory molecules critically determines the functionality of APCs 32, 43.

Even though the inhibitory abilities of R848-APCs might not be exclusively linked to PD-L1 (Fig. 3C), PD-L1 is an interesting tool to manipulate immune responses. It has been shown that the PD-1/PD-L1 pathway controls graft versus host reactive T cells 44 and that PD-L1 knockout mice have a stronger allostimulatory reactivity compared to WT mice 45. Hence, we were especially interested in the regulation of PD-L1 expression. We identified a MAPK-dependent production of IL-6 and IL-10 that cause a long-lasting STAT-3 activation as a central hallmark of TLR-APCs and accordingly to PD-L1 expression.

The TLR-stimulus led to the production of two cytokines that mainly signal via STAT-3: IL-6 and IL-10 (Fig. 4A and B). Both cytokines are able to alter the phenotype of iDCs toward the TLR phenotype: no CD1a expression, retained CD14 expression and high expression levels of PD-L1 (Supporting Information Fig. 4). To verify the importance of IL-6 and IL-10 we compared the activation of different STAT molecules (Fig. 7). As expected, TLR-APCs show an almost constitutive STAT-3 activation. In contrast, STAT-5 was activated in iDCs and diminished in TLR-APCs. Therefore, TLR-APCs and iDCs show clear differences in STAT-3 and STAT-5 activation pattern. Our results indicate that TLR agonists added at an early time point of iDC differentiation block STAT-5 activation and shift the STAT activation pattern toward STAT-3. Indeed, blocking of STAT-3 signal transduction had an eminent effect on the TLR-APC phenotype. STAT-3 inhibition repressed CD14 and PD-L1 (Fig. 8A and B). In accordance with our data, Barton et al. 11 suggested that stimulatory or tolerogenic function of APCs depends on their STAT-3 activation level. To further support the role of STAT-3, we performed ChIP assays and detected that STAT-3 binds to the PD-L1 promoter (Fig. 8C). STAT-1 was also able to bind PD-L1, but less effectively (Fig. 8D). There were only few quantitative differences in the magnitude of STAT-1-binding between iDCs and TLR-APCs, indicating a minor role for STAT-1 in the initial differentiation process of TLR-APCs.

Induction of cytokine expression can be regulated by different mechanisms controlled by the stimulus. For TLR signaling, NF-κB and MAPKs have been described as major signaling pathways. We revealed that IL-6 and IL-10 were not released after blocking p38 (SB) and p44/42 (UO) MAPKs (Fig. 5B and C) and that CD14 and PD-L1 expression was reduced (Fig. 6A and B). Blocking p38 (SB) alone influenced the production of IL-10 but had no effect on IL-6 production. In contrast, the inhibition of p44/42 (UO) affected IL-6 expression. Similar preferences were also discernible in regulation of CD14 and PD-L1 surface expression: inhibition of p44/42 affects to a greater extent expression of CD14, while the inhibition of p38 is related more to the expression of PD-L1. In spite of the short duration of p44/42 activation and the dominant activation of p38 it is hard to decide which MAPK is preferentially involved in the generation of the TLR-APC phenotype. Anyway the combined inhibition of p38 and p44/42 had the greatest impact on the cytokine secretion and the TLR-APC phenotype.

Blocking experiments show that STAT-3 and MAPKs are essential for the TLR-APC phenotype. To connect the MAPK and STAT-3 findings, we checked STAT-3 activation after MAPK inhibition to find that after blocking p38/p44/42 almost no tyrosine phosphorylation of STAT-3 was detectable (Fig. 9A). This effect could be overcome by the addition of exogenous IL-6 and IL-10 (Fig. 9C). Thus, the TLR-APC phenotype is dependent on the p38 and p44/42 MAPK-induced cytokine production and the resulting STAT-3 activation. An involvement of p38 and p44/42 in the activation of STAT-3 after TLR stimulation has been observed also from others 46. Xie et al. 7 suggest that MAPK p38 activity might be responsible for the impaired differentiation of monocytes into iDCs after LPS stimulation. One day after LPS stimulation, p38 is activated and p44/42 not. Due to the late time point (d1), the initial and short activation of p44/42 was not seen, thus the link between p44/42 MAPK, IL-6 production and STAT-3 activation was missed.

Our results indicate that TLR agonists added at an early time point of iDC differentiation induce a shift from STAT-5 toward STAT-3 activation and thus critical determine the functional phenotype of the APCs. We have shown before, that the addition of LPS during the differentiation of murine bone marrow cells into myeloid DCs led to a reduced CD11c expression 5. The effect on CD11c could be traced back to a SOCS-1 dependent blockade of STAT-5 phosphorylation. Additionally, we could show that SOCS-3 is also able to reduce STAT-5 phosphorylation 5. Since TLR-APC upregulate preferentially SOCS3 (data not shown) we suppose that in the human system the block of STAT-5 might be SOCS-3-dependent. Hence, two different mechanisms seem to balance STAT-5/STAT-3 and thus regulate the expression of CD14, PD-L1 and CD1a.

During infection, pathogen-derived TLR-agonists might bypass conventional iDCs differentiation and induce PD-L1-expressing tolerogenic APCs in a STAT-3-dependent manner. Studies investigating organs and tissues with close contact to microbial TLR agonists provide indications of the in vivo relevance of TLR-APC. For example, the liver has to deal with gut-derived portal blood that contains high concentrations of bacterial products. It has been demonstrated that liver DCs have reduced T-cell stimulatory capacities 47, 48. The data of Lunz et al. 49 support these findings. They could show that gut-derived bacterial products induce IL-6/STAT-3 signaling and thereby inhibit the hepatic DC activation/maturation.

In summary, we show here that STAT-3 is responsible for the regulation of PD-L1 expression, triggered via IL-6 and IL-10. TLR agonists potently induce STAT-3 activation and thus direct DC differentiation to tolerogenic APCs.

Materials and methods

Recombinant cytokines were purchased from R&D (Wiesbaden, Germany). The TLR agonist LPS from Salmonella Minnesota was provided by U. Seydel (Borstel, Germany) and the TLR agonist R848 was purchased from ALEXIS (Lausen, Switzerland). MAPK inhibitor SB203580 and STAT-3 inhibitor JSI-124 were bought from Calbiochem (Schwalbach, Germany), p44/42 inhibitor UO126 from Cell Signaling Technology (Danvers, MA, USA). FACS antibodies were acquired from BD (Heidelberg, Germany) except PD-L1, PD-L2, B7-H3, B7-H4 and ICOS-L antibodies (Natutec, Frankfurt/Main, Germany). Western blot antibodies were purchased from Cell Signaling Technology except for unphosphorylated STAT-5 and STAT-1 (Santa Cruz Biotechnology, Heidelberg, Germany).

Generation of human APCs

PBMCs were isolated from fresh blood or buffy coat by density gradient centrifugation (Biocoll seperating solution 1.077 g/mL; Biochrom AG, Berlin, Germany) and washed three times in PBS. CD14+ cells were positively selected by magnetic-associated cell sorting (AutoMACS: program possel; Miltenyi Biotec, Bergisch-Gladbach, Germany). Sorted cells were seeded in 24-well plates (Greiner bio-one, Frickenhausen, Germany) at a density of 2×106 cells/mL in RPMI 1640 medium (Biochrom AG) supplemented with 10% FBS (BioWest, East Sussex, UK) and 1% penicillin and streptomycin (PAA, Pasching, Austria). Cultures were supplemented with 1000 IU/mL rhGM-CSF and IL-4 to generate iDCs. For generation of TLR-APCs 1 μg/mL R848 or 30 ng/mL LPS were added. Cells were cultured at 37°C in a humidified atmosphere in the presence of 5% CO2.

Isolation of T cells

PBMCs were isolated from fresh blood or buffy coat by density gradient centrifugation and washed three times. Desired T-cell population (CD3+, CD4+ and CD8+) were obtained by positive selection (AutoMACS: program possel; Miltenyi Biotec). T cells were seeded for the respective co-culture experiments. T cells isolated from co-culture experiment were also positive selected by AutoMACS.

Mixed lymphocyte reaction

MLRs were performed in allogeneic settings: purified 2×105 T cells or 5×105 PBMCs (CD4+ or CD8+) were co-cultured with 1×104 of Mitomycin C-pre-treated APCs. Cells were cultured for 4 days and exposed to [3H]-thymidine (Amersham Pharmacia Biotech GmbH, Freiburg) during the last 18 h of culture. Thymidine uptake was measured by using a liquid scintillation counter.


After differentiation 1×104 cells/200 μL of R848-APCs were seeded in 96-well plates (Greiner bio-one) and 1×105 fresh isolated, allogeneic CD3+ T cells were added. Afterwards, the cells were treated with 10 μg/mL anti-PD-L1 antibody (eBioscience, Vienna, Austria). Cells were cultured for 4 days and exposed to [3H]-thymidine during the last 6 h of culture.

Tregs and their functionality

For the determination of CD25 and FoxP3 1×106 CD4+ T cells were incubated with 5×104 APCs for 5 days. Activation beads (Anti-BiotinMACSiBead Particles plus biotinylated antibodies against CD2, CD3 and CD28; Miltenyi Biotec) were used to mimic APC stimulation and to activate resting T cells. Beads were loaded following the manufacturer's protocol. Concentrations used were 100 μL beads diluted 1:100 in RPMI.

To confirm the generation of Tregs, we performed transfer experiments: CD4+ cells were isolated from PBMCs. One half of the cells were differentiated into Tregs by co-stimulation with different APC types for 6 days. The other half was frozen at −80°C. On day 6, T cells from cultures were separated in CD25+ and CD25- cells. They were added at a ratio of 1:10 or 1:30 in 96-well flat-bottom plates to thawed CD4+ T cells, which were labeled with CFSE. Afterwards, the cell mixture was stimulated with activation beads. Cell proliferation was measured after 5 days by flow cytometry.

For CFSE-labeling cells were incubated 10 min at room temperature in 0.3 μM CFSE/PBS (MolecularProbes, San Diego, CA, USA) and thereafter intensively washed.

Flow cytometry

Cells were analyzed on a FACS Canto (BD). CD1a, PD-L1, CD14, ICOS-L1, PD-L2, B7-H3, B7-H4, CD80, CD86, MHCII CD40 and CD252 were stained at the cell surface. Therefore, cells were washed in PBS and stained directly with FITC, PE or APC-labeled antibodies. Overlays were done with the Weasel v2.5 software (WEHI, Melbourne, Australia). FoxP3 expression in T cells was assessed using an anti-human FoxP3 Staining Kit (e-Biosciences, San Diego, CA, USA), including corresponding isotype controls.


Cell-free supernatants were harvested and analyzed for IL-6, IL-12p40, IL-10 and TNF by commercial available ELISA kits (OptEIA; BD).

Western blotting

About 8×106 cells were stimulated and subsequently lysed in RIPA buffer (50 mM Tris-HCL, pH7.4; 1% Igepal; 0.25% sodium deoxycholate; 150 mM NaCl; 1 mM EDTA; 1 mM PMSF; 1 μg/mL each aprotinin, leupeptin and pepstatin; 1 mM Na3VO4; and 1 mM NaF). Lysates were cleared by centrifugation at 4° for 20 min at 14 000×g. Equal amounts of the lysates were fractionated by 12% SDS-PAGE and electrotransferred to nitrocellulose membranes (Whatman Protran nitrocellulose membrane; neoLab, Heidelberg, Germany). The membranes were blocked with TBS/0.05% Tween-20/3% BSA and were blotted with the indicated antibodies. Detection was by enhanced chemiluminescence (ECL; Perkin Elmer, Groningen, Netherlands). For the analyses of the un- and phosphorylated proteins the same lysates but different membranes were used.


The ChIP assay was carried out as described by Natoli and co-workers 50 modified by Bode et al. 51. One-twentieth of the immunoprecipitated DNA was used in quantitative PCR. Results were shown as percentage of input. STAT-3, STAT-1 and STAT-5 antibodies used for ChIP were acquired from Santa Cruz Biotechnology. The following primers were used for DNA quantification: PD-L1 promoter fw TGGACTGACATGTTTCACTTTCT and rev CAAGGCAGCAAATCCAGTTT.

Statisitical analysis

The comparison of two data groups were analyzed by Student's t-test.


We appreciate the discussions and help of Dr. K. Kubatzky and Dr. K. A. Bode and the help of Judith Bauer. This work was supported by the Collaborative Research Center (SFB) 405 (Bartz/Heeg).

Conflict of interest: The authors declare no financial or commercial conflict of interest.