Cross-regulation of cytokine signalling: pro-inflammatory cytokines restrict IL-6 signalling through receptor internalisation and degradation

The inflammatory response is a highly coordinated answer of the body to trauma, tissue injury or infection. It involves a complex interplay of different cell types releasing cytokines which act in an auto- or paracrine manner to induce the so-called acute phase response. This reaction starts with the release of pro-inflammatory cytokines such as interleukin (IL)-1β and tumour necrosis factor (TNF) α. Later, the major mediator of hepatic acute phase protein expression IL-6 is expressed and exerts pro- as well as anti-inflammatory activities (Bode and Heinrich, 2001).

It is mandatory that the function of these cytokines is stringently regulated since dysregulated IL-6 signalling, involving the Janus kinase/signal transducer and activator of transcription (JAK/STAT), MAPK (mitogen activated protein kinase) and the PI3K/Akt pathways, leads to severe inflammatory diseases such as rheumatoid arthritis, inflammatory bowel disease, multiple sclerosis, and also cancer (e.g. plasmacytoma). IL-6 signal transduction is controlled by the IL-6-induced feedback inhibitor SOCS3 (suppressor of cytokine signalling) (Endo et al., 1997; Naka et al., 1997; Starr et al., 1997; Croker et al., 2003) and the protein tyrosine phosphatase SHP2 (SH2-domain containing protein tyrosine phosphatase 2) (Lehmann et al., 2003). Both proteins are recruited to the activated receptor subunit gp130 to control the initial, membrane proximal steps in IL-6 signal transduction (Schmitz et al., 2000). In addition, the nuclear protein phosphatase TC-PTP inactivates the major IL-6-induced transcription factor (signal transducer and activator of transcription 3) STAT3 (Yamamoto et al., 2002).

Apart from this, several lines of evidence indicate a negative influence of pro-inflammatory cytokines. TNFα induces SOCS3 in macrophages (Bode et al., 1999) and additionally leads to binding of the protein tyrosine phosphatase SHP2 to gp130 (Bode et al., 2003). IL-1β counteracts signal transduction of IL-6-type cytokines at different levels: it affects IL-6-induced gene expression through acting on target gene promoters (Zhang and Fuller, 1997; Zhang and Fuller, 2000; Bode et al., 2001a) or by enhancing the IL-6-induced expression of the SOCS3 feedback inhibitor (Bode et al., 1999; Yang et al., 2004). IL-1β also rapidly counteracts IL-6-mediated STAT3 activation (Ahmed and Ivashkiv, 2000; Ahmed et al., 2002). This rapid inhibition of IL-6-initiated STAT3 activation has been suggested to be independent of de novo protein synthesis and SOCS3 induction. Even activation of Janus family tyrosine kinases, essential for IL-6-mediated signalling, appears to be dispensable. Instead, a contribution of the p38 stress kinase-dependent pathway was proposed (Ahmed et al., 2002). However, the underlying molecular mechanisms involved in this rapid stress-induced effect on IL-6-type cytokine signalling have not been identified so far.

An indispensable prerequisite for any cytokine stimulation of a given cell type is the availability of the receptor complex on the cell surface. During differentiation of cells, as well as during cellular responses to exogenous stimuli, the expression of cytokine receptors on the cell surface is tightly regulated. Unlike a number of growth factor receptors, which are internalised only in response to ligand binding (Hari and Roth, 1987; McClain et al., 1987; Glenney et al., 1988; Sorkin et al., 1991), the common receptor subunit of all IL-6-type cytokines, gp130, appears to internalise constitutively (Thiel et al., 1998) through a di-leucine motif-dependent mechanism (Dittrich et al., 1994) and is subsequently degraded (Wang and Fuller, 1994; Blanchard et al., 2001). Gp130 is expressed ubiquitously. Therefore it has so far been a generally accepted idea that the responsiveness of individual cell types to IL-6-type cytokines is mainly determined by the expression of a specific alpha-receptor (i.e. IL-6Rα, IL-11Rα, CNTFRα) or by the expression of a second specific signal-transducing subunit (i.e. LIF-R, OSM-R, WSX1).

Here we describe a novel molecular mechanism for the restriction of IL-6-type cytokine signal transduction by pro-inflammatory cytokines and cellular stress: we show that activation of the stress-activated protein kinase p38 by IL-1β, TNF α or environmental stress leads to the MAPK-activated protein kinase 2 (MK2)-dependent serine phosphorylation of gp130 on serine 782 and increased receptor internalisation and degradation resulting in the net loss of surface-bound gp130 molecules and consequently in reduced IL-6-dependent STAT3 activation and gene induction. The identification of gp130 as a new target of the p38-MK2 axis unravels a novel mechanism of cytokine crosstalk.

To study the mechanisms involved in the rapid inhibition of IL-6 signalling by IL-1β we used HepG2 hepatoma cells, which are very responsive to IL-6 since they express the membrane-bound IL-6 receptor as well as the signal transducer gp130. Preincubation of HepG2 cells with IL-1β results at any time point and at any dose tested in a reduction of IL-6-induced STAT3 tyrosine phosphorylation (Fig. 1A). Binding of activated STAT3 to an optimised STAT-recruiting DNA probe is severely compromised in IL-1β-pretreated cells (Fig. 1B, lanes 2 and 4). To determine whether the IL-1β-activated transcription factor NF-κB is involved in this fast inhibitory effect on STAT3 activation, we made use of HepG2 cells stably expressing a non-degradable IκBα mutant. These cells have been extensively characterised and lack NF-κB activation in response to IL-1β [Fig. 1C, compare lanes 1-5 with lanes 6-10 and Yang et al. (Yang et al., 2004)]. Fig. 1D shows that expression of the non-degradable IκBα in HepG2-IκBα cells did not affect the early inhibition of STAT3 DNA-binding after IL-1β treatment (Fig. 1D, compare lanes 3 with 4 and 7 with 8, and supplementary material Fig. S1 for quantification). We recently reported that IL-1β enhances IL-6-induced SOCS3 expression by stabilising SOCS3 mRNA (Yang et al., 2004). However, we could show that IL-1β was able to inhibit IL-6-mediated STAT3 activation independent of de novo protein synthesis (Fig. 1E).

In addition to activating the NF-κB pathway, IL-1β is a known activator of the stress-activated MAP kinase p38 (Dunne and O'Neill, 2003). Therefore, we analysed whether p38 MAPK activation plays a role in the negative regulation of IL-6-induced STAT3 activation. In these experiments, STAT3 Y705 phosphorylation was determined by intracellular FACS analyses (Fig. 2A) and by western blotting of whole cell lysates (Fig. 2B). Whereas IL-1β efficiently inhibited the IL-6-induced STAT3 phosphorylation by approx. 50% in DMSO-treated cells, the inhibitory effect of IL-1β was absent in cells treated with the p38-specific inhibitor SB202190 (Fig. 2A,B). Efficient inhibition of p38 kinase by SB202190 was demonstrated by lack of phosphorylation of the downstream target protein MK2 in the presence of the inhibitor (Fig. 2B, lower panels, lanes 5-8).

In an alternative approach we determined IL-6-induced translocation of STAT3 from the cytoplasm to the nucleus by confocal microscopy. As expected, treatment of HepG2 cells with IL-6 for 20 minutes led to an obvious nuclear accumulation of STAT3 (Fig. 2C, first panel). Pretreatment with IL-1β for 15 minutes blocked IL-6-induced STAT3 translocation almost completely (third panel) and again, in the presence of SB202190 the inhibitory action of IL-1β was efficiently prevented and thereby accumulation of STAT3 in the nucleus restored (Fig. 2C, fourth panel).

Initiation of any cytokine signalling cascade requires appropriate availability of cytokine receptors at the cell surface. Since the negative regulatory effect of IL-1β described above occurred rapidly we focussed on the influence of IL-1β on the expression of IL-6-type cytokine receptors on the plasma membrane. Indeed, within the first 15 minutes IL-1β induced a fast and persistent reduction of gp130 cell surface localisation to approximately 50% of the initial amounts present (Fig. 3A). Simultaneous presence of IL-6 did not further influence gp130 plasma membrane localisation (not shown). Of note, IL-1β specifically affected gp130 and not the IL-6 α-receptor, the oncostatin M (OSM) receptor (Fig. 3B) or the LIF receptor (not shown). In accordance with data presented above, this negative effect of IL-1β was not dependent on NF-κB activation, since HepG2 cells stably expressing the non-degradable IκBα mutant or lacking the IκB kinase subunit IKKγ, due to siRNA knock down, displayed a similar loss in gp130 expression (supplementary material Fig. S2A,B). However, pharmacological inhibition of p38 activity significantly reduced the downregulation of surface bound gp130 by IL-1β in HepG2 (Fig. 3C). Pre-treatment with a specific JNK inhibitor had no effect (supplementary material Fig. S2C).

The suppressive effect of IL-1β on gp130 cell surface expression was not restricted to HepG2 cells, but was also detectable in HeLa cells as well as primary human dermal fibroblasts (Fig. 3D). In addition to IL-1β, other pro-inflammatory cytokines such as TNFα, as well as cellular stress, e.g. exposure to UV light or arsenic-containing chemicals (phenyl arsine oxide, PAO) activate p38 (supplementary material Fig. S2D). Therefore, we investigated whether these stimuli also induce gp130 downregulation. Indeed, TNFα (Fig. 3D) as well as PAO treatment and UV irradiation (Fig. 3E) potently downregulated gp130 cell surface expression. As observed previously for IL-1β, these properties were significantly reduced in the presence of the p38 inhibitor SB202190 (Fig. 3D,E). Again, OSMR expression was not significantly affected (supplementary material Fig. S2E).

The loss of gp130 surface expression as a result of stress could be explained by an increased gp130 internalisation or by altered post-endocytic processes, i.e. degradation rather than recycling. To monitor receptor internalisation we carried out classical antibody uptake experiments. Receptor uptake was quantified in a FACS-based assay; either indirectly by measuring loss of surface receptor bound antibodies (Fig. 4A) or alternatively by measuring intracellular accumulation of prebound FITC-labelled gp130 antibodies (Fig. 4B). Both approaches revealed an accelerated internalisation of gp130 in response to IL-1β. Again, IL-1β did not significantly increase uptake of OSMR antibodies (supplementary material Fig. S3A).

The increased uptake of FITC-labelled gp130 antibody in response to IL-1β was also confirmed by live-cell imaging: cell-surface-expressed receptors were stained by FITC-labelled antibodies added to the culture media of living cells. This resulted in a clearly visible signal which colocalised with Trypan Blue, which is a membrane stain. Only a minor proportion of the antibodies was constitutively taken up into vesicles over time (Fig. 4C, upper panel). Upon IL-1β treatment, however, we observed a striking loss of gp130 membrane staining, documented in steady-state pictures (Fig. 4C, middle panel) and multiple time series (supplementary material Movie 1). Consequently, hardly any colocalisation with Trypan Blue staining was detectable.

To examine the role of p38 MAPK in this process, we conducted live-cell imaging in the presence of SB202190. The p38 inhibitor efficiently prevented IL-1β-induced gp130 uptake and thus plasma membrane staining was retained (Fig. 4C, lower panel). This result was confirmed by the FACS-based internalisation assay (supplementary material Fig. S3B). A general effect of IL-1β on the internalisation machinery could be excluded by determination of the ratio of FITC-gp130-positive vesicles over constitutively internalised Cy3-transferrin-positive vesicles (supplementary material Fig. S3C).

To elucidate the fate of gp130 we analysed colocalisation of internalised gp130 (FITC-labelled antibodies) with endocytosed transferrin (Alexa-Fluor-555-labelled antibodies; Fig. 4D), the early endosomal marker EEA1 (Fig. 4E) as well as the lysosomal marker LysoTracker Red DND-99 (Fig. 4F). After IL-1β stimulation we observed some colocalisation of gp130-positive vesicles and transferrin-positive vesicles (Fig. 4D), however, a much larger number of vesicles co-localised with the early endosomal marker EEA1, a process that was IL-1β dependent (Fig. 4E). Additionally, we observed a broad co-localisation with the lysosomal marker (Fig. 4F) indicating that IL-1β induced both internalisation and lysosomal degradation of gp130.

To corroborate these findings biochemically, we examined whether IL-1β affects the decay of gp130 overall. Therefore, HepG2 cells were incubated with cycloheximide to block de novo protein biosynthesis. The amount of gp130 remaining after 1-6 hours in the absence or presence of IL-1β was examined by western blotting (Fig. 4G). Indeed, we observed accelerated degradation of gp130 in the presence of IL-1β, further supporting a role for IL-1β, not only in internalisation but also in subsequent degradation of gp130.

Next, we determined the receptor requirements contributing to p38-dependent internalisation. Overexpression of a dominant-negative mutant of dynamin 2 (Dyn2-K44A) — which inhibits clathrin-dependent internalisation (Damke et al., 1994) — prevented stress-induced loss of coexpressed gp130 in COS-7 cells (Fig. 5A). It is well established that the constitutive internalisation of gp130 depends on a di-leucine motif (L786 L787) in its cytoplasmic tail (Dittrich et al., 1994; Thiel et al., 1998). Indeed, mutation of this amino acid sequence leads to an internalisation-deficient receptor which is no longer downregulated upon treatment with PAO (Fig. 5B). Previous studies on LIF signalling suggested that serine residues, most prominently S782, located upstream of the di-leucine motif in gp130 (Fig. 5B, upper panel) might be involved in the regulation of the internalisation process (Gibson et al., 2000). As shown in Fig. 5C, phosphorylation of S782 was indeed induced by the pro-inflammatory cytokine IL-1β: treatment of HepG2 cells with IL-1β for 5 minutes (Fig. 5C lane 3) or 10 minutes (lane 7) led to a strong phosphorylation of endogenous gp130 at S782. Importantly, this IL-1β-induced phosphorylation could be blocked by the p38 inhibitor SB202190 (lanes 5 and 9). By contrast, IL-6 itself did not lead to S782 phosphorylation of gp130 (lanes 2 and 6). Also, in HepG2 cells phosphorylation of gp130 in response to LIF could not be observed (supplementary material Fig. S4A).

Based on these observations we speculated that S782 in gp130 serves as a stress sensor, limiting gp130 cell surface expression. To prove this hypothesis, we constructed new receptor mutants in which S782 was replaced by aspartate or alanine to mimic constitutive phosphorylation of S782 or to prevent phosphorylation of S782, respectively. In order to allow assessment of these mutants in cells endogenously expressing gp130, we generated chimeric receptors containing the extracellular region of the erythropoietin receptor fused to the gp130 transmembrane and intracellular part containing the above mentioned mutations of S782 or of the di-leucine internalisation motif. Additionally, these chimeric receptors contained an N-terminal HA-tag to allow detection. Cell surface expression of each gp130 construct was monitored in stably transfected HepG2 cells. In response to IL-1β the surface expression of the wild-type construct was markedly decreased, whereas the internalisation-deficient gp130 (LL/AA) construct as well as the non phosphorylatable S782A mutant were resistant to IL-1β treatment (Fig. 5D). By contrast, the expression of the phosphorylation mimic S782D was reduced already in the absence of IL-1β and did not decrease further upon IL-1β treatment (Fig. 5D). As expected, only the wild-type receptor and the LL/AA mutant were phosphorylated in response to IL-1β (Fig. 5D, lower panel). Furthermore, similar observations were made for PAO- or TNFα-treated HEK293 cells stably expressing IL-5Rα-gp130 chimeric receptors (supplementary material Fig. S4B,C).

Since lower steady state expression levels of the S782D mutant could indicate a faster internalisation or an overall lower expression of this mutant, we carried out internalisation assays using live-cell imaging and flow cytometry. In live-cell imaging experiments uptake of Alexa-Fluor-555-labelled anti HA-antibody was assessed 15 minutes after transfer of labelled cells from 4°C to 37°C. Single cell analysis at the 0- and 15-minute time points indicated a constitutive internalisation of wild-type gp130 (Fig. 5E, first panel), the S782D mutant (third panel) and to a minor extent also of the S782A mutant (fourth panel). The internalisation-deficient mutant gp130 LL/AA remained at the cell surface (second panel). Quantitative fluorescence analysis at the cell surface and the overall fluorescence analysis was performed for the same cell by single cell analyses at 0 and 15 minutes. Cumulative results for 20-25 cells per cell line are presented in supplementary material Fig. S4D. Flow cytometry allowed quantification of a larger number of cells and confirmed a faster constitutive internalisation of the S782D mutant compared with wild-type gp130 or the gp130 S782A mutant (Fig. 5F).

Taken together, these data clearly show that inflammatory cytokines and stress stimuli lead to a p38-mediated phosphorylation of S782 within the cytoplasmic part of gp130 and that phosphorylation of this serine is crucial for reducing the surface expression of gp130.

Since S782 in gp130 is not part of a p38 consensus site, it is very unlikely that p38 phosphorylates this residue directly. However, p38 itself is an activator of kinases belonging to the family of MAPK-activated protein kinases (MK), e.g. MK2, MK3 and MK5 (for a review, see Gaestel, 2006). These enzymes typically phosphorylate substrates containing a ϕxRxxS/T consensus sequence, where ϕ represents a bulky, hydrophobic amino acid. As demonstrated in Fig. 6A, S782 is part of such a consensus sequence. Furthermore, we could demonstrate that IL-1β or PAO treatment results in MK2 activation in HepG2 and HeLa cells (supplementary material Fig. S5A,B). We generated two new IL-5Rα-gp130 chimeric proteins containing mutations within the MK2 consensus sequence. In these constructs alanine replaced either phenylalanine 777 (F777A) or arginine 779 (R779A; Fig. 6A). HEK293 clones stably expressing these mutants were treated with PAO and the surface expression of the chimeras was monitored by FACS analyses. In two clonal cell lines expressing the wild-type construct, PAO treatment resulted in a pronounced receptor downregulation, whereas the receptor surface expression in gp130-F777A- or -R779A-expressing cell lines remained unchanged (Fig. 6A). Additionally, no serine phosphorylation of gp130 was observed when these mutants were stimulated with PAO (not shown).

These results strongly suggest that an MK family kinase phosphorylates gp130 in response to cellular stress. To further narrow down the potential kinase involved, we examined murine embryonic fibroblasts (MEFs) deficient for MK2 or MK5 (Shi et al., 2003). MEFs from control mice responded as expected with a strong serine phosphorylation of gp130 after IL-1β and PAO treatment (Fig. 6B). Preincubation of the cells with SB203580 strongly reduced the serine phosphorylation initiated by IL-1β and PAO (lanes 3 and 5) as clearly demonstrated by quantitative analysis of the western blots (supplementary material Fig. S5C) indicating that also in murine cells the effect is mediated by p38 MAPK. The inhibitor was slightly less efficient at dampening PAO-induced serine phosphorylation of gp130, indicating that additional, yet-to-be-identified serine kinases might be activated in response to PAO. In MK2-deficient MEFs gp130 serine phosphorylation was strongly impaired (Fig. 6B, lanes 7 and 9; supplementary material Fig. S5C) whereas it remained clearly detectable in MK5-deficient MEFs (Fig. 6B, lanes 12 and 14; supplementary material Fig. S5C) after both stimuli. This correlated well with the intact MK2 phosphorylation in MK5-deficient MEFs compared to wild-type MEFs (Fig. 6B, lower panels).

Consistent with the strongly impaired gp130 serine phosphorylation in MK2-deficient MEFs we also observed a reduced IL-1β- and PAO-mediated loss of gp130 cell surface expression in these cells whereas no difference could be detected between wild-type and MK5^−/− MEFs (Fig. 6C,D). Internalisation assays in wild-type MEFs, MK2^−/− cells and MK2^−/− cells reconstituted with MK2 confirmed these results (supplementary material Fig. S5D). Additionally, a low molecular mass inhibitor of MK2 strongly impaired IL-1β-dependent internalisation of gp130 and retained plasma membrane localisation (Fig. 6E, lower panel). Taken together, these experiments provide strong evidence that MK2 is the kinase responsible for the downregulation of gp130 upon stress and inflammatory cytokine stimulation.

After having shown that pro-inflammatory cytokines and cellular stresses result in a MK2-dependent phosphorylation of gp130 at S782, which initiates rapid loss of surface expressed gp130, we assumed that this molecular mechanism contributes to the rapid IL-1β-mediated inhibition of IL-6-type cytokine-induced STAT3 tyrosine phosphorylation. We therefore tested whether IL-1β-mediated inhibition of gp130-dependent STAT3 activation is abrogated in MK2-deficient fibroblasts. Since fibroblasts do not express membrane-bound IL-6R the cells had to be stimulated with the agonistically acting soluble IL-6R. Indeed, IL-1β reduced IL-6- and sIL-6R-induced STAT3 activation in wild-type MEFs (Fig. 7A, lanes 1-4), but not in MK2-deficient MEFs (lanes 5-8), whereas in MK5-deficient cells the IL-1β effect was unaffected (lanes 9-12).

The observation that the S782D mutant of gp130 displays a much lower cell surface expression than the S782A mutant (Fig. 5D) led us to test whether decreased cell surface expression is reflected in reduced gene induction through this receptor mutant. We utilized EpoR-gp130 receptor chimeras which upon Epo stimulation induce signalling similar to the wild-type IL-6 receptor complex independent of the presence of endogenous gp130. As shown in Fig. 7B an α1-antichymotrypsin promoter-driven as well as an α2-macroglobulin promoter-driven luciferase gene is much better transcribed in response to activation of the gp130-S782A construct than the gp130-S782D mutant, indicating that decreased cell surface expression indeed reduces the responsiveness of cells to the respective ligand.

To check whether our observation also holds true for the regulation of IL-6-induced endogenous gene expression by IL-1β we analysed induction of endogenous γ-fibrinogen mRNA by real-time PCR in HepG2 cells. As shown in Fig. 7C (columns 1-4) IL-1β significantly reduces γ-fibrinogen expression which was initiated through the EpoR-gp130 chimeric receptor. To test whether this inhibitory activity of IL-1β is mediated through the newly identified MK2-dependent pathway we knocked down MK2 expression by siRNA (Fig. 7C, left panel). In HepG2 cells lacking MK2 expression through siRNA knock-down, no inhibitory effect of IL-1β on IL-6-dependent γ-fibrinogen expression could be observed (columns 5-8).

These data strongly suggest that phosphorylation of S782 within gp130 by MK2 contributes to IL-1β-dependent inhibition of gp130-mediated STAT3 activation and gene induction.

The present study identifies a previously unobserved phosphorylation of gp130, the common signal transducing receptor subunit of IL-6-type cytokines, by pro-inflammatory cytokines (IL-1β, TNFα) and cellular stress (UV, PAO), which results in accelerated internalisation and degradation of gp130. Therefore, our finding significantly contributes to the understanding of how pro-inflammatory stimuli and cellular stress rapidly downregulate IL-6-type cytokine-mediated signal transduction and gene expression (Andus et al., 1988; Ahmed and Ivashkiv, 2000; Bode et al., 2001b; Ahmed et al., 2002).

The regulatory phosphorylation occurs on S782 in the intracellular region of gp130, close to the di-leucine motif crucial for internalisation (Dittrich et al., 1994). We demonstrate that phosphorylation of gp130 at S782 induces loss of up to 50% of cell surface expressed gp130 within 30 minutes (Fig. 3A). Both phosphorylation of S782 and the presence of the di-leucine motif are essential for the downregulation. Our studies clearly show that IL-1β initiates the following processes: gp130 internalisation is significantly accelerated (50% internalisation in response to IL-1β after 30 minutes versus 25% constitutive internalisation; Fig. 4A) and gp130 translocates from the plasma membrane to early endosomes and finally to the lysosomal compartment (Fig. 4E,F) where it is degraded (Fig. 4G).

Interestingly, one important prerequisite for di-leucine-mediated internalisation through the membrane adaptor AP-2 is an acidic amino acid in position −4 to the di-leucine motif (Pieters et al., 1993; Pond et al., 1995; Dietrich et al., 1997). Gp130 does not contain such an acidic residue at position −4, but instead the serine residue 782. Thus, phosphorylation of this serine might generate the negative charge required for increasing the affinity for the AP-2 adaptor complex. Indeed, replacement of S782 by aspartate to mimic the phosphorylation decreased the steady state cell surface expression level of gp130 by 50% compared with the S782A mutant, which can no longer be phosphorylated (Fig. 5D).

Serine 782 in gp130 has previously been shown to be phosphorylated by stimulation of 3T3-L1 pre-adipocytes with leukaemia inhibitory factor (LIF) (Gibson et al., 2000). The authors suggested that this phosphorylation contributes to LIF-mediated receptor internalisation. Mutation enhanced expression of target genes. No similar observation has been reported so far for any other IL-6-type cytokine. Of note, we show that in HepG2 hepatoma cells neither IL-6 itself (Fig. 5C) nor LIF (supplementary material Fig. S4A) induces the serine phosphorylation of gp130. These data are in line with the finding that mutation of S782 in gp130 does not influence the gp130-mediated endocytosis of IL-6 (Dittrich et al., 1994). Whether this indicates that ligand-mediated serine phosphorylation of gp130 is restricted to LIF and cell-type specific needs to be determined. Nathanson and colleagues postulated in a recent study that a calmodulin-dependent protein kinase (CaMK), possibly type II (CaMKII) was responsible for the LIF-mediated phosphorylation step (Gibson et al., 2005).

Here we identified MK2, a downstream target of p38 and a CaMK superfamily member, as an executing kinase mediating the serine phosphorylation of gp130. To our knowledge gp130 is the first transmembrane substrate of this kinase to been shown to be crucial for stabilisation of a number of short-lived mRNAs (Gaestel, 2006). Thus, cells isolated from MK2-knockout mice are deficient in the LPS-induced biosynthesis of TNFα and IL-1β (Kotlyarov et al., 1999; Neininger et al., 2002; Wang et al., 2002). Additionally, MK2^−/− mice show increased susceptibility to infection by Listeria monocytogenes (Lehner et al., 2002) and show exacerbated TNFα-mediated inflammatory bowel disease (Kontoyiannis et al., 2002). Therefore, this enzyme is appreciated as an important mediator in inflammatory responses.

Two recent publications suggested the involvement of the p38 pathway in the internalisation of the epidermal growth factor receptor (Vergarajauregui et al., 2006; Zwang and Yarden, 2006). Of note, for the receptor tyrosine kinase family this mechanism seems to be specific for the EGFR since neither the insulin receptor nor the HGF receptor c-Met were internalised in response to TNFα, IL-1β or UV (Zwang and Yarden, 2006). It will be interesting to investigate whether MK2, as a downstream substrate of p38, plays a role in growth factor internalisation as well. Here we demonstrate that within the IL-6-type cytokine receptor family only gp130 is susceptible to p38 and/or MK2-mediated internalisation whereas the IL-6R, the OSMR (Fig. 3B) and LIFR (not shown) were not affected. This is in line with the fact that the serine 782 motif within an MK2 target sequence is not conserved in these receptors although the internalisation motif in the LIFR is also neighboured by a serine at position −4. So far, no involvement in receptor internalisation has been attributed to this serine residue. It is conceivable, however, that similar mechanisms might apply for other cytokine receptor families.

Fig. 7 demonstrates a significant contribution of gp130 internalisation to the restriction of IL-6-mediated signal transduction: MK2-deficient cells are resistant to IL-1β-mediated inhibition of STAT3 tyrosine phosphorylation (Fig. 7A) and receptors harbouring a serine 782 to aspartate mutation show a severely impaired reporter gene induction upon stimulation (Fig. 7B). Most importantly, knockdown of MK2 renders cells resistant to IL-1β-mediated inhibition of endogenous IL-6-induced gene expression (Fig. 7C).

Recently, in vivo studies identified the threshold of STAT3 signalling initiated by IL-6-type cytokines as a critical determinant for pathological consequences. Gp130-dependent STAT3 hyperactivation correlates with splenomegaly, exaggerated hepatic acute-phase response and dysregulated haematopoietic homeostasis. These effects could be completely rescued by reducing STAT3 expression levels by knockdown of one STAT3 allele (Jenkins et al., 2005b; Jenkins et al., 2005a).

Thus, on the one hand the inhibitory activity of IL-1β on IL-6 signalling dampens potential additive pro-inflammatory effects of both cytokines, resulting in prevention of overshooting immunological reactivity such as autoimmune arthritis or inflammatory bowel disease (for a review, see Kubo et al., 2003). On the other hand, IL-1β delays the anti-inflammatory effects of IL-6 to strengthen the pro-inflammatory processes in the initial phase of an inflammation.

Since the negative effect of IL-1β and TNFα on IL-6-induced acute phase protein gene expression was first described in 1988 by Andus et al. (Andus et al., 1988), many studies on the underlying molecular mechanisms of IL-1β-dependent repression of IL-6 signalling have been performed. Our novel findings can explain various recent observations for the inhibition of gp130-mediated signalling: Ahmed et al. reported that IL-1β inhibits IL-6-mediated STAT3 activation in a very rapid manner, independent of de novo protein synthesis (SOCS), independent of the Y759 residue in gp130 and independent of modifications of the STAT3 serine phosphorylation (Ahmed and Ivashkiv, 2000; Ahmed et al., 2002). Interestingly, IL-1β-mediated inhibition of IL-6 signalling could be strongly impaired by inhibiting p38 activity. Additionally, overexpression of a constitutive active MKK3 or MKK6 almost completely abolished IL-6-mediated gene induction (Bode et al., 2001b; Ahmed et al., 2002). These observations suggested that IL-1β acts very early in the signalling pathway, however, the exact target could not be identified. Here, we finally identify gp130 itself as the target of pro-inflammatory cytokine intervention in IL-6 signal transduction.

Recombinant OSM and TNFα were purchased from Peprotech (London, UK). Recombinant IL-6 and sIL-6R were prepared as described by Arcone et al. and Weiergräber et al. (Arcone et al., 1991; Weiergräber et al., 1995). The specific activity of IL-6 was 2×10⁶ B cell-stimulatory factor-2 units/mg protein. Recombinant IL-1β was purchased from Roche. Recombinant erythropoietin (Epo) was a generous gift from B. Hilger and K. H. Sellinger (Roche, Mannheim, Germany).

Antibodies against Y705-phosphorylated STAT3 (pY-STAT3; 9131, 9138), MK2 (3042), T222-phosphorylated MK2 (pT-MK2; 3044) and tubulin (2148) were obtained from Cell Signaling Technology (Frankfurt/Main, Germany). Antibodies against gp130 (M20, sc-656), the OSMR (AN-A2, sc-9992), p38 MAPK (C-20, sc-535) and EEA1 (C-15, sc-6414) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Antibodies against S782-phosphorylated gp130 (pS-gp130) were from Santa Cruz Biotechnology or BD Biosciences (558096). The mouse monoclonal antibody against STAT3 (610189) was purchased from BD Biosciences. The antibody used for FACS analysis to monitor human gp130 (B-R3) and IL-6Rα (B-R6) cell surface expression was obtained from Diaclone (Hölzel Diagnostika GmbH, Cologne, Germany) and for murine gp130 (D023-3) from MBL (Japan). The antibody against the extracellular region of the IL-5Rα was a kind gift from Jan Tavernier (Ghent, Belgium). Antibodies against active double-phosphorylated p38 (T180/Y182; pTpY-p38) were from Promega (Madison, WI, USA). HRP-coupled secondary antibodies as well as IgG1 antibodies for control were from DAKO (Hamburg, Germany). Monoclonal HA antibodies to the HA-tagged extracellular domain of the EpoR and gp130 receptors were from Covance (HA.11, MM-101R).

Aprotinin, pepstatin, and gentamycin were from Sigma-Aldrich (Taufkirchen, Germany), leupeptin was from MP Biomedical Inc. (Illkirch, France), Pefabloc was purchased from Roth (Karlsruhe, Germany). Cycloheximide was from Sigma-Aldrich (Taufkirchen, Germany). DMSO, SB202190, SB203580 and Jak-Inhibitor I was from Calbiochem (Merck, Darmstadt, Germany). Phenylarsine oxide (PAO) was from Sigma-Aldrich. The MK2-inhibitor (2,4-diamino-5H-chromeno-[2,3-β]pyridine-3-carbonitrile; U.S. patent 2004/0127511) was used as described by Vertii et al. (Vertii et al., 2006). Poly(vinylidene difluoride) (PVDF) membrane was from PALL (Dreieich, Germany). The enhanced chemiluminescence kit was from GE Healthcare (Munich, Germany). Protein concentrations were determined with the Bio-Rad reagent (Bio-Rad, München, Germany). The oligonucleotides used were synthesised by MWG-Biotech (Martinsried, Germany). The FACS used was a FACS Calibur and FACS Canto (Becton Dickinson, Heidelberg, Germany). Live cell imaging was performed with a Zeiss LSM510 Meta confocal laser scanning microscope. Transferrin Alexa Fluor 555 and LysoTracker Red DND 99 were purchased from Molecular Probes (Carlsbad, CA, USA). Real-time PCR was performed with the Rotor-Gene 6000 Rotary Analyzer (Corbett Research Pty Ltd, Mortake, NSW, Australia).

RIPA buffer: 50 mM Tris-HCl (pH 7.5) 150 mM NaCl, 0.5% NP-40, 15% glycerol, 1 mM EDTA, 1 mM NaF, 1 mM Na₃VO₄, protease inhibitor cocktail (10 μg/ml of each aprotinin, pepstatin (Sigma-Aldrich), and leupeptin (MP Biomedicals, Illkirch, France). Triton cell lysis buffer: 20 mM Tris-HCl (pH 7.5) 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate, 1 mM Na₃VO₄,1 μg/ml leupeptin, 1 mM PMSF (phenylmethyl sulphonyl fluoride). Washing buffer: 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 10 mM NaF, 0.1% Triton X-100. TBE: 20 mM Tris base, 20 mM boric acid, 0.5 mM EDTA (pH 8.0). Binding medium: Phenol Red-free DMEM-F12 medium, supplemented with 0.2% BSA and 10 mM HEPES buffer.

Dulbecco's modified Eagle's medium (DMEM) was from Invitrogen-Gibco (Paisley, UK) and foetal calf serum (FCS) from PAA (Pasching, Austria).

HepG2 and HepG2 derivatives were grown in DMEM-nut mix F12; normal human dermal fibroblast (HDF), HeLa, HEK293 and COS-7 cells were grown in DMEM. Medium was supplemented with 10% FCS, streptomycin (100 mg/l) and penicillin (60 mg/l). MEF cells were cultured in DMEM (High Glucose, GlutaMAX) supplemented with 10% heat inactivated FCS.

HepG2, COS7 and HeLa cells were purchased from DSMZ (Braunschweig, Germany). IκBα-expressing HepG2 cells were supplied by Stephan Ludwig, University of Münster, Germany. HEK293 cells were purchased from the ATCC (Wesel, Germany). Primary human dermal fibroblasts were kindly supplied by Hans F. Merk and Yvonne Marquardt from the Department of Dermatology (Medical School RWTH Aachen).

HepG2-mock and HepG2-IκBα were prepared as described previously and generously supplied by S. Ludwig (Yang et al., 2004). Murine embryonal fibroblasts (MEF) deficient for MK2 or MK5, as well as the corresponding wild-type cells were described previously (Kotlyarov et al., 1999). Primary human dermal fibroblast (HDF) cells were isolated from foreskin as described previously (Dreuw et al., 2005). Stably transfected HepG2 cells were generated by retroviral infections.

For immunoprecipitation, HepG2 cells from a 60% confluent culture on a10 cm plate were lysed in 500 μl RIPA buffer. Before adding antibodies, protein amounts were adjusted to be equal (1 mg protein). Buffers were supplemented with a protease inhibitor cocktail. For MEF cells a modified Triton-X-100-containing lysis buffer was used. Immune complexes were washed in washing buffer and subsequently separated by SDS-PAGE and transferred to a poly(vinylidene difluoride; PVDF) membrane. Antigens were detected by incubation with specific primary antibodies and horseradish-peroxidase (HRP)-coupled secondary antibodies and developed with an enhanced chemiluminescence kit. For the analysis of whole cell extracts, cells were lysed as described above and 20 μg protein separated by SDS-PAGE. For detection of activated STAT3, the protein amount was reduced to 10 μg, whereas for detection of S782-phosphorylated gp130 the amount was increased to 30 μg. Blotting and staining were carried out as for immunoprecipitated proteins.

Preparation of nuclear extracts and EMSAs were performed as described previously (Wegenka et al., 1993). The protein concentration of the nuclear extracts was determined and 6 μg protein was incubated with a surplus of double-stranded ³²P-labelled κB site (5′-GATCCATGGGGAATTCCCCATG-3′) for binding NF-κB or a labelled optimized STAT3-binding site from the cfos-promoter m67SIE (5′-GATCCGGGAGGGATTTACGGGAAATGCTG-3′). The protein-DNA complexes were separated on a 4.5% polyacrylamide gel containing 7.5% glycerol in 0.25-fold TBE at 20 V/cm for 4 hours. Gels were fixed in 10% methanol, 10% acetic acid and 80% water for 1 hour, dried, and autoradiographed.

Vectors for transfection of COS-7 cells (Fig. 5B): pSVL-gp130-eGFP encodes wild-type human gp130 fused at the C-terminus to the enhanced GFP protein, whereas pSVL-gp130-LL/AA-eGFP results from an exchange of the two leucine residues, 786 and 787, for alanine within the di-leucine internalisation motif of gp130.

Vectors for transfection of HEK293 cells (Fig. 6A, supplementary material S4B,C): pCAG-GS-IL5Rα-gp130 encodes a chimeric receptor composed of the extracellular part of the IL-5Rα and the transmembrane and cytoplasmic part of gp130. Variants of this vector, pCAG-GS-IL5Rα-gp130-LL/AA and pCAG-GS-IL5Rα-gp130-S782A, harbour a mutated di-leucine motif (L786 L787 to AA) or a mutation of the newly identified MK2-phosphorylation site (S782 to A), respectively. pCAG-GS-IL5Rα-gp130-F777A and pCAG-GS-IL5Rα-gp130-R779A harbour single amino acid substitutions of phenylalanine 777 or arginine 779 to alanine in the MK2 consensus sequence of gp130.

Expression vectors for transient transfection of HepG2 cells (Fig. 7B): pCDNA3.1-EpoR/gp130-S782A and pCDNA3.1-EpoR/gp130-S782D encode chimeric receptors composed of the extracellular part of the EpoR and the transmembrane and cytoplasmic part of gp130. The vectors encode receptors which carry a mutation of the MK2 phosphorylation target serine 782 to alanine or aspartate, respectively. The chimeric receptor systems have been carefully characterised in previous studies in our laboratory.

Vectors to express dominant negative dynamin (Fig. 5A) were kindly supplied by S. L. Schmid (Scripps Research Institute, La Jolla, CA).

COS-7 and HepG2 cells were transiently transfected with 1-2 μg receptor expression plasmid using either the DEAE/dextran method as described previously (Hermanns et al., 1999) or FuGene6 according to the manufacturers' instructions. HEK293 Flp-In™ T-Rex™ (Invitrogen, Karlsruhe, Germany) were transfected using FuGene6 (Roche Applied Science, Mannheim, Germany) according to the manufacturer's instructions.

Stably transfected HepG2 cells expressing the EpoR-gp130 chimeric receptors (Fig. 5D,E and Fig. 7C) were generated by retroviral gene transfer. Retrovirus was prepared in HEK293T cells. The day before transfection, cells were seeded on collagen-coated plates. Immediately before transfection, cell medium was changed to OptiMEM plus 5% FCS. For transfection, pVPack-Ampho (Stratagene; 2 μg), pVPack-GP (Stratagene; 2 μg) and 5 μg of pQCXIP (BD Biosciences) encoding the expression cassette for HA-tagged EpoR-gp130 chimeric receptors (wt or S782A, S782D, LL/AA) were mixed with 250 μl OptiMEM. In parallel, 250 μl OptiMEM (Gibco, Karlsruhe, Germany) was mixed with 15 μl Lipofectamine (Invitrogen, Karlsruhe, Germany) and equilibrated at room temperature (RT) for 5 minutes. After mixing both vials and additional incubation at RT for 20 minutes, the solution was added to the HEK293T cells and left overnight in the incubator at 37°C. The next day medium was changed to DMEM-5% FCS and the temperature in the incubator reduced to 32°C. The retrovirus-containing supernatant was harvested 48 hours post-transfection and filtered (45 μm, low protein binding filter). Afterwards the filtrated supernatant was added together with 8 mg/ml polybrene (hexadimethrine bromide) to HepG2 cells (30-40% confluent). The following day medium was changed to DMEM-F12 plus 10% FCS. Next day the selection for transduced cells was started in the presence of puromycine.

siRNA knock-down was performed with ON-TARGETplus SMART pools siRNA from Dharmacon (Schwerte, Germany). The pool L-003516-00-0020 was used to knock down human MK2, whereas the pool L-003767-00-0010 was used to knock down IKKγ. For a control the negative siRNA control duplex OR-0030-neg05 from Eurogentec (Seraing, Belgium) was used. siRNA was transfected using the DharmaFECT 1 transfection reagent T-2001-03 as suggested by the supplier. In detail, 3×10⁵ cells were seeded per well. The following day, 75 nM siRNA was dissolved in 100 μl OptiMEM (serum free). DharmaFECT1 was diluted 1:50 in OptiMEM at room temperature and equilibrated for 5 minutes before mixing the 100 μl diluted DharmaFECT1 and the 100 μl siRNA solution. While incubating the siRNA for 20 minutes, cells were washed with serum-free OptiMEM and dispensed with 800 μl of medium per well. siRNA solution was added and incubated for 5 hours at 37°C in the CO₂ incubator before adding 500 μl OptiMEM and 30% FCS. Before beginning the experiment, the cells were cultivated for additional 72 hours.

RNA was isolated by using the RNeasy kit from Qiagen (Hilden, Germany). cDNA was produced with help of the Omniscript RT Kit from Qiagen and mRNA content determined by real-time PCR. The probes Hs00241037 FGG and Hs99999909-m1 HPRT from Applied Biosystems (Carlsbad, CA) were selected to detect γ-fibrinogen and HPRT mRNA.

5×10⁵ cells were harvested in ice-cold PBS-EDTA and resuspended in 100 μl ice-cold FACS buffer (PBS containing 5% FCS, 0.1% NaN₃) containing 0.2-1.0 μg monoclonal antibody raised against the extracellular part of the respective receptors or the HA-tag (anti-human gp130, anti-mouse gp130, anti-OSMR, anti-IL6Rα, anti-IL-5Rα, anti-HA). After 30 minutes cells were washed in 3 ml FACS buffer and subsequently stained with mouse IgG-specific R-phycoerythrin-conjugated goat antibodies for an additional 30 minutes. Cells were analysed using the FACSCalibur or FACSCanto (Becton Dickinson).

For intracellular FACS analysis cells were stimulated in 6 cm plates and afterwards detached in 1 ml PBS with 10 mM EDTA. The cells were fixed by incubating them with 1 ml 4% PFA-PBS for 10 minutes. After centrifugation the cell pellet was resuspended in 0.5 ml 2% PFA-PBS and rested for an additional 10 minutes. Cells were permeabilised by adding 4.5 ml ice-cold methanol for 30 minutes. Afterwards cells were blocked in 300 μl of 0.5% BSA-PBS and aliquoted for primary antibody incubation. Primary antibody [phospho-STAT3 (1:100, Cell Signaling #9138) or isotypic control mouse IgG1 (1:100, Dako X0931)] was applied for 1 hour at RT. Cell were washed in 3 ml 0.5% BSA-PBS, centrifuged and resuspended in 100 μl PE-conjugated goat anti-mouse IgG (1:100) in 0.5% BSA-PBS. Incubation was continued for 1 hour at RT in the dark. After final washing steps cell were analysed using FACSCalibur.

Cells were preincubated in ice-cold binding medium with or without 1 μg/ml B-R3 (anti-gp130) or HA.11 (anti HA-tag) antibody for 45 minutes. Then, cells were washed three times with binding medium to remove unbound antibodies and subsequently left untreated or they were stimulated with IL-1β (100 IU/ml) for 15 or 30 minutes at 37°C. Control cells were left on ice during this incubation period to monitor initial gp130 cell surface localisation. After stimulation, the cells were harvested in ice-cold PBS-EDTA and incubated in FACS buffer with an R-phycoerythrin-conjugated goat anti-mouse antibody for 30 minutes on ice. Non-internalised gp130 was analysed by FACS. To monitor the appearance of intracellular, endocytosed gp130, cells were preincubated in ice-cold binding medium containing 1 μg/ml FITC-labelled control antibody or B-R3 (anti-gp130) antibody for 45 minutes. The cells were incubated with or without IL-1β (100 IU/ml) for 15 or 30 minutes at 37°C. After stimulation, the cells were harvested in ice-cold PBS-EDTA and washed for 5 minutes using 2 M glycine buffer (pH 2.5) to remove all cell-surface-bound antibodies.

HepG2 cells grown on coverslips were processed as described for the internalisation assays, however, binding medium also contained fluorescently labelled transferrin. After incubation at 37°C for 20 minutes, cells were fixed in 3.7% formaldehyde and mounted with Mowiol. For detection of EEA1-positive vesicles, prior to mounting, cells were permeabilised in 0.1% Triton X-100 and stained using goat anti-EEA1, followed by Cy3-labelled anti-goat secondary antibody.

HepG2 cells were seeded on coverslips 1 day before investigation. During investigation cells were cultured in binding medium. Incubation was performed in the presence of FITC-labelled anti-gp130 antibody at 10% CO₂ and 37°C for 40 minutes with or without IL—1β (100 IU/ml). Recycling endosomes and lysosomes were stained with transferrin—Alexa-Fluor-555 and LysoTracker, respectively, according to the manufacturers' instructions. To visualise the plasma membrane the incubation medium was changed, and cells were stained with Trypan Blue (red fluorescence) and confocal laser scanning analysis was carried out. When multiple time series were examined all pictures were taken using the autofocus.

Internalisation assays were performed as described above. Briefly, cells were incubated for 1 hour at 4°C with Alexa-Fluor-555-labelled anti-HA antibody in binding medium. Subsequently, cells were transferred into the live-cell chamber of the LSM and analysed for 15 minutes under cell culture conditions.

HepG2 cells were transiently transfected with 0.75 μg receptor expression plasmid, 0.8 μg pGL3α2M-Luc or pGL3-α1ACT-Luc reporter and 0.4 μg β-galactosidase plasmid (pCH110, GE Healthcare) using FuGene6 transfection reagent (Roche) according to the manufacturer's instructions. pGL3-α2M-Luc contains the promoter region −215 to +8 of the rat α2-macroglobulin promoter and pGL3-α1ACT-Luc the promoter region −379 to +25 of the human α1-antichymotrypsin promoter upstream of the luciferase-encoding sequence of plasmid pGL3 (Promega). One day after transfection, cells were split onto 12-well plates and 1 day later stimulated with Epo (7 IU/ml) for 16 hours. Luciferase activity was determined with the Luciferase Assay System (Promega) and normalised to the activity of coexpressed β-galactosidase.

Unless stated differently, all data are expressed as medians + s.e.m. for at least three separate experiments (one-tailed, paired Student's t-test). Probabilities of P<0.05, P<0.01 and P<0.001 was considered significantly (*), very significantly (**) and extremely significantly (***) different, respectively. All analyses were performed using the statistical software GraphPad Prism (GraphPad Prism for Windows, Version 3.00, GraphPad Software, Inc., CA).

This work was supported by the Deutsche Forschungsgemeinschaft (DFG, SFB542, FZ82) to F.S. and H.M.H. We thank Stefan Ludwig for generously supplying IκB-expressing HepG2 cells, Hans F. Merk and Yvonne Marquardt for primary dermal fibroblasts, Jan Tavernier for the antibodies against IL-5Rα and S. L. Schmid (Scripps Research Institute, La Jolla, CA) for the expression plasmid encoding dominant-negative dynamin. We thank Uli Heuter and Daniela Kraemer for technical assistance.