Dose-dependent cross-talk between the transforming growth factor-β and interleukin-1 signaling pathways

IRAK, MyD88, TRAF6, and TAK1 Are Required for TGFβ-Dependent Activation of NFκB.

Using a κB reporter-dependent luciferase assay, we find that the activation of NFκB by TGFβ2 is deficient in IRAK-null or MyD88-null cells, compared with parental 293C6 cells (Fig. 1A), indicating that these two major components of IL-1β-dependent signaling are also needed for TGFβ2 to activate NFκB. Furthermore, restoring IRAK or MyD88 to the deficient cells restored both IL-1β- and TGFβ2-dependent NFκB activation (Fig. 1B). The inhibitory effects of dominant-negative constructs for TRAF6 or TAK1 showed that these two proteins are also likely to be required for TGFβ2 to activate NFκB (Fig. 1C). Because four important IL-1 signaling components seem to be required for TGFβ2 to activate NFκB, it is likely that the entire IL-1 pathway is involved.

IL-1β Activates Smads.

To test for reciprocal cross-activation, 293C6 and derived cell lines lacking IRAK or MyD88 were treated with 5 nM IL-1β or 0.2 nM TGFβ2 (positive control). The results show that Smad2 was activated within a very short time in all three cell lines (Fig. 2 A and B). The cells were pretreated with cycloheximide (CHX) for 1 h to show that the phosphorylation of Smad2 in response to IL-1β does not require new protein synthesis (Fig. 2 A and B). Furthermore, as expected, IκBα was not degraded in IL-1β-treated IRAK-null or MyD88-null cells, compared with 293C6 parental cells, but Smad2 was still phosphorylated (Fig. 2 A and B), showing that neither IRAK nor MyD88 is required for IL-1β to activate Smads.

Activation of Smads by IL-1β Is Dose Dependent and General.

Previously, we determined the dose dependence for the activation of NFκB in response to TGFβ2 (3). To obtain comparable data for cross-activation, titrations with IL-1β were done in stable 293C6 cell lines expressing luciferase reporter constructs for NFκB and Smads. Treatment for 4 h activated NFκB half-maximally at ≈0.005 nM IL-1β, and the activation reached a plateau at a concentration of ≈0.5 nM (Fig. 3A). In contrast, the titration curve for Smad activation by IL-1β reached a plateau at ≈5 nM (Fig. 3B). The concentration of IL-1β for half-maximum Smad activation was ≈1.25 nM, 250-fold higher than for NFκB. However, activation of Smads was detected at concentrations of IL-1β as low as 0.2 nM (Fig. 3B and Western blot analysis data, data not shown). Because 293C6 cells overexpress the IL-1R and the IL-1R accessory protein (13, 14), it is necessary to show that the activation of Smads by IL-1β is general. Therefore, we analyzed four additional cell lines. Smads can be activated by IL-1β in human fibroblast BJ and WI38 cells, human glioblastoma T98G cells, and human melanoma Mel29 cells (Fig. 3C). Titrations revealed that 0.2 nM IL-1β is able to activate Smads substantially in BJ and Mel29 cells and detectably in WI38 and T98G cells (Fig. 3C).

Cross-Signaling with TLR2.

Because of the similarity of the IL-1R and TLR signaling pathways, it is logical to ask whether TGFβ can use other members of the receptor superfamily to activate NFκB. To address this question, we used 293 cells that, in contrast to 293C6 cells, have not been transfected with constructs encoding the IL-1R or the IL-1R accessory protein. Also, 293 cells do not express appreciable amounts of TLRs (15). Constructs encoding flag-tagged TLRs 1, 2, 4, 5, or 6 were transfected into 293 cells, and stable pools were checked for expression from the constructs (data not shown). We were not able to get enough expression of TLR4 in this experiment. TLRs 3, 7, 8, and 9 were not investigated because they are not expressed on the cell surface (6, 15). The cells were transfected with a κB reporter and treated with 2 nM TGFβ2 or 0.2 nM IL-1β for 4 h. As shown in Fig. 4, TLR2 and possibly TLR5, but not TLR1 or TLR6, enhance the activation of NFκB in response to TGFβ2. The results indicate that other members of the IL-1R superfamily can interact with TGFβ receptors to catalyze receptor-mediated cross-talk.

Receptor Requirements for Cross-Signaling.

Stable pools of 293C6 cells were generated that express siRNAs corresponding to TβRI, TβRII, Smad2, IL-1R, and Trp (tropomyosin, a control). Also included in the experiment were 293C6 cells expressing the IκB super repressor of NFκB (SR-IκB), which carries serine–alanine mutations at positions 32 and 36 and therefore is not phosphorylated and degraded in response to activation signals. The cells were pretreated with CHX for 1 h to inhibit protein synthesis, so that IκB would not be resynthesized, and then treated with 2 nM TGFβ2 or 5 nM IL-1β for 1 h. Smad2 phosphorylation and IκBα degradation were detected by Western blot analyses (Fig. 5A). The same cells were transfected transiently with κB- or Smad-dependent luciferase reporter constructs and then treated with 2 nM TGFβ2 or 5 nM IL-1β overnight (Fig. 5 B–E). The data show that TβRI and TβRII, but not IL-1R, are required for TGFβ2 to activate Smads (Fig. 5 A, D, and F), and IL-1R, TβRI, and TβRII are all required for TGFβ to activate NFκB (Fig. 5 A, E, and F) and for IL-1β to activate Smads (Fig. 5 A, C, and F). Furthermore, IL-1R, but not TβRI or TβRII, is required for IL-1β to activate NFκB (Fig. 5 A, B, and F). Additionally, Smad2 is not required for IL-1β or TGFβ2 to activate NFκB (Fig. 5 A, B, and E). The SR-κB affected the IL-1β/NFκB and TGFβ2/NFκB pathways (Fig. 5 A, B, and E), but not the IL-1β/Smad or TGFβ2/Smad pathways (Fig. 5 A, C, and D). As summarized in Fig. 5F, cross-signaling requires not only the receptors that bind to the ligand that is applied to the cells, but also the receptors for the cross-signaling ligand.

Cross-Recruitment of Receptors After High-Dose Treatment with TGFβ2 or IL-1β.

Cells were treated with 0.2 or 4 nM TGFβ2 or with 0.25 or 5 nM IL-1β for 15 min, and the receptors were immunoprecipitated with anti-TβRI, anti-TβRII, or anti-IL-1R. IL-1R was strongly recruited to both TβRs at high concentrations of TGFβ2 or IL-1β, and detectable cross-recruitment was also seen at lower concentrations (Fig. 6). Similarly, both TβRI and TβRII were recruited strongly to IL-1R at high concentrations and detectably even at lower concentrations of TGFβ2 or IL-1β. These data reveal that cross-talk between the TGFβ2 and IL-1β signaling pathways is likely to be mediated by ligand-dependent physical association of the receptors.

Dose-Dependent Gene Activation by TGFβ2 and IL-1β.

293C6 cells were treated with TGFβ2 (0.2 or 4 nM) or IL-1β (0.25 or 5 nM) for 4 h, and mRNA expression was analyzed by using microarrays. Low or high concentrations of TGFβ2 or IL-1β induced gene expression differently (Fig. 7A) [a short gene list is found in supporting information (SI) Table 2]. Confirmation by Northern blot analysis of several examples from the microarray data showed that P8 and ELMO3 were induced well by the low concentration of IL-1β and the high concentration of TGFβ2, but were not induced well by the low concentration of TGFβ2 (Fig. 7 B and C). Therefore, these two are likely to be NFκB-responsive genes. Conversely, NUMBL and FAIM2 were induced by the low concentration of TGFβ2 and the high concentration of IL-1β, but were not induced well by the low concentration of IL-1β (Fig. 7 B and C), showing that these genes, which normally respond to TGFβ, are likely to be induced by IL-1β through the activation of Smads. Note that there is significant cross-activation of gene expression (Fig. 7) at cytokine concentrations that are consistent with those that stimulate cross-binding of the receptors (Fig. 6). ECM1 and GAGEB1 were not induced by low concentrations of IL-1β or TGFβ2, but were induced by high concentrations of either cytokine (Fig. 7), suggesting that their expression might depend on the simultaneous activation of both the NFκB and Smad pathways.

IL-1 Is Highly Expressed by Some Glioblastoma Cells.

In previous work, we showed that prostate cancer PC3 cells secrete high concentrations of TGFβ2 and that this secreted cytokine is responsible for the constitutive activation of NFκB seen in these cells (3). To demonstrate that large amounts of IL-1 are also secreted by tumor cells, we examined cell lines and primary isolates derived from glioblastomas. Table 1 shows that IL-1α, IL-1β, or both are secreted at high levels by some glioblastoma cells. In conditioned media from primary CCF3 cells, the concentration of IL-1α was 3.0 nM and that of IL-1β was 1.5 nM. In conditioned media from the U87 cell line, the concentration of IL-1β reached 1.2 nM, but that of IL-1α was lower. Therefore, some tumors secrete levels of TGFβ or IL-1 that are well above the concentrations required to initiate cross-signaling.

Mechanism of Cross-Talk.

TGFβ2 and IL-1β activate their primary pathways at low concentrations and cross-activate the other pathways at higher concentrations. The mechanism involves ligand-dependent receptor–receptor interactions. Association of the two types of receptors might trigger cross-signaling in any of several ways. A major possibility is that aggregation of the receptors brings receptor-bound kinases into close proximity, allowing cross-phosphorylation. In fact, a somewhat similar phenomenon has been reported in the IFN system, where the endogenous mouse IFN-α receptor 1 coimmunoprecipitates with IFN-γ receptor 2 (16). These cytokine receptors might exist in close proximity in a multimeric complex, for which the Taniguchi group proposed the term “receptosome” (17). The components are probably brought together within lipid raft domains in the membrane that are rich in cholesterol and sphingolipids, allowing them to function efficiently in IFN-dependent signaling. It is possible that IL-1R and TβRI/TβRII also associate within lipid rafts. Di Guglielmo et al. (18) suggest that the TβRs are internalized into both caveolin- and early endosome antigen 1-positive vesicles and that they reside in both lipid raft and nonraft membrane domains.

Signaling Interactions Between TGFβ and IL-1.

To date, very few studies have documented connections between TGFβ- and IL-1-dependent signaling. Most of the published work shows that each cytokine antagonizes the other's effects. For example, Park et al. (19) showed that TGFβ reduces the level of IL-1-induced cyclooxygenase-2 mRNA in mouse calvarial bone cells. Other studies report IL-1-mediated inhibition of TGFβ-dependent signaling in cocultured fibroblasts (20) and down-regulation of the expression of IL-1-induced TLR2 by TGFβ in murine hepatocytes (21). Choi et al. (22) reported that Smad6 negatively regulates IL-1R/TLR signaling through a direct interaction with the adaptor Pellino-1. Simultaneous positive and negative regulation of biological responses has been seen before and may provide an opportunity for finely tuned regulation. A well known example is the stimulation by TNF-α of both pro- and antiapoptotic responses (23).

Although TAK1 was first identified in the context of TGFβ-dependent signaling, it functions primarily as an essential component of the pathways activated by IL-1 (24–26). The role of TAK1 in TGFβ-dependent signaling is controversial. For example, TAK1 has been reported to activate p38, allowing it to mediate a Smad-independent response to TGFβ or even to interfere with the Smad function (27, 28). A physical interaction between the inhibitory Smad6 and TAK1 has also been observed (29). Furthermore, Benus et al. (30) suggested that IL-1 can inhibit TGFβ-dependent signaling directly through the phosphorylation of Smad3 by TAK1.

Physiological Relevance of Cross-Talk.

Secretion of high concentrations of IL-1 has been documented in several instances. Elaraj et al. (31) showed that mRNAs encoding both IL-1α and IL-1β are highly expressed in metastases from patients with several different cancers and also in several tumor cell lines. IL-1β is highly expressed in non-small-cell lung carcinoma, colorectal adenocarcinoma, melanoma, and pancreatic carcinoma (31, 32). The supernatant media from these cell lines stimulated a significant increase in the permeability of endothelial cell monolayers, a hallmark of early angiogenesis. Human pituitary adenoma HP75 cells generate ≈1 nM IL-1α when kept in culture for 72 h (33). The same group tested 25 other primary cultures from human pituitary adenomas, removed during routine transphenoidal surgery, and reported that the secretion of IL-1α by these cells ranged from a concentration of 0.14 to 9.6 nM (33). Moreover, Fries et al. (34) reported that cultured monocytes from glioblastoma multiforme patients generated 2.2–2.8 nM IL-1β after 21 days in culture and that these cells survived for >250 days, whereas monocytes derived from controls generated 0.03–0.07 nM IL-1β after 21 days and that these cells survived for only 114 days. The authors suggested that enhanced IL-1β release increases the longevity of glioma-associated peripheral blood monocytes in vitro. Our current data contribute further to this story, showing that in human glioblastoma cells IL-1α and IL-1β are overexpressed. For example, CCF3 cells secreted 3.0 nM IL-1α and 1.5 nM IL-1β (Table 1).

High concentrations of IL-1 may also drive the activation of Smads in inflammation. Persistent IL-1β signaling in glial cells is likely to make a key contribution to chronic inflammation of the brain (35). Also, in a mouse model study, Joshi et al. (36) showed that, after treatment with lipopolysaccharide, IL-1β was induced and released, reaching a concentration of ≈0.14 nM in serum. The local concentration of IL-1β is likely to be much higher at a site of inflammation. TGFβ2 is secreted at a high concentration by a variety of cancer cell lines, including lines derived from prostate cancer, melanoma, and glioblastoma (3, 11, 12). The high concentration of TGFβ2 (1.2 nM in medium from prostate cancer PC3 cells conditioned for 24 h) is responsible for the activation of NFκB in these cells (3). The above examples strongly suggest that the phenomena we report here for both IL-1β and TGFβ have important biological and pathological significance.

Cross-Talk at the Level of Gene Expression.

Our analysis of gene expression suggests a concentration-dependent control by both TGFβ2 and IL-1β. High concentrations of each of these cytokines activated genes that are normally activated by low concentrations of the reciprocal cytokine (Fig. 7A). For example, 93 genes were induced 5-fold or more by a low concentration of TGFβ2. Of those, none was induced by a low concentration of IL-1β, whereas 32 genes were induced by a high concentration of IL-1β, indicating that high IL-1β can induce a subset of, but not all, TGFβ-specific genes. Furthermore, besides those 93 genes, an additional 69 genes were induced by a high concentration of TGFβ2. Of these, 49 genes were also induced by a low concentration of IL-1β. On the other hand, 143 genes were induced 5-fold or more by a low concentration of IL-1β. Of those, none was induced by a low concentration of TGFβ2, whereas 49 genes were induced by a high concentration of TGFβ2, indicating that high TGFβ2 can induce a subset of, but not all, IL-1β-specific genes. Moreover, besides those 143 genes, an additional 62 genes were induced by a high concentration of IL-1β. Of these, 32 genes were also induced by a low concentration of TGFβ2. Our data strongly suggest that cross-talk between the TGFβ and IL-1β signaling pathways leads to dose-dependent cross-control of gene expression, which is likely to have an important effect on biological outcomes. An appreciation of these interactions provides new insight into responses to IL-1 and TGFβ in the proximity of tumors that secrete high concentrations of these factors and probably also at sites of inflammation, where the local concentrations of these cytokines are also likely to be high.

Cell Culture.

All cells were cultured in Dulbecco's modified Eagle's medium (Cleveland Clinic Foundation Medium Laboratory, Cleveland, OH) with 10% FBS. The cell lines used were: human 293 cells, human 293C6 cells [293 cells previously transfected with constructs encoding IL-1R1 and IL-1R accessory protein (13), and newly transfected with E-selectin-driven zeocin-resistance and thymidine kinase genes] and the derived MyD88-null and IRAK-null mutants (7, 11), human BJ and WI38 fibroblasts, human Mel29 melanoma cells, and human glioma cells T98G, U87, U138, U251, and D54 (a gift of Darrell Bigner, Duke University Medical Center, Durham, NC). WM3 cells were obtained from a temporal lobectomy for epilepsy. CCF3, 4, and 52 cells are short-term primary glioblastoma multiforme cell cultures made at the Cleveland Clinic. Experiments were carried out when the cells reached 90% confluence. To collect conditioned media, cells were cultivated as described above, the media were replaced, and the cells were grown for another 48 h. The conditioned media were collected, floating cells were pelleted at 3,000 × g at 4°C for 10 min, and the supernatant media were pipetted into sterile tubes and stored at −80°C.

Plasmids.

To construct siRNAs, sequences corresponding to IL-1R, TβRI, TβRII, and Smad2 mRNAs were cloned in the retroviral vector H1RNA (3). The sequences are available upon request. Approximately 10 days after puromycin selection, cells were pooled and tested for the best knockdown by the Western blot method. The κB-luciferase construct p5XIP10 κB (with five tandem copies of the κB site from the IP10 gene), or the Smad binding element-luciferase construct, was transfected transiently into the above cells. Transfections and luciferase assays were carried out essentially as described in ref. 11.

Flag-tagged TLR1, 2, 4, 5, and 6 were kind gifts from Ruslan Medzhitov (Yale University School of Medicine, New Haven, CT). These constructs were cotransfected with a puromycin-resistance plasmid following the procedure described above, and stable pools were set up after puromycin selection.

EMSA.

The oligomer used for NFκB binding (5′-AGTTGAGGGGACTTTCCCAGGC-3′; Santa Cruz Biotechnology, Santa Cruz, CA) was labeled with [γ-³²P]ATP by the polynucleotide kinase method following the protocol provided by Promega (Madison, WI). The experiments were carried out essentially as described in ref. 11.

Western and Northern Blot Analyses.

Cells treated with TGFβ2 or IL-1β for different times were washed with 1× PBS and pelleted at 3,000 × g at 4°C for 4 min before the cell pellets were lysed with RIPA buffer (11). Samples were assayed by the Western blot method as described in ref. 11. Northern blot analyses were also carried out as described in ref. 11 by using 20 μg of total RNA. Human cDNA probes for P8, FAIM2, ELMO3, NUMBL, ECM1, and GAGEB1 were derived from sequence-verified Human I.M.A.G.E. Consortium cDNA clones (American Type Culture Collection, Manassas, VA).

Coimmunoprecipitations.

Cells treated with TGFβ2 or IL-1β for 15 min were lysed in Triton buffer (0.5% Triton X-100, 20 mM Hepes, pH 7.4, 150 mM NaCl, 12.5 mM β-glycerophosphate, 1.5 mM MgCl₂, 10 mM NaF, 2 mM DTT, 1 mM sodium orthovanadate, 2 mM EGTA, 20 μM aprotinin, and 1 mM phenylmethanesulfonyl fluoride). These cell extracts were incubated with 1 μg of anti-IL-1R, anti-TβRI, or anti-TβRII for 2 h, followed by incubation with protein A-Sepharose beads at 4°C overnight. The beads were resuspended in 1× SDS sample loading buffer (20% glycerol, 10% 2-mercaptoethanol, 5% SDS, 0.2 M Tris·HCl, pH 6.7, 1 mg of bromophenol blue) at a 1:1 ratio. The released proteins were separated by SDS/PAGE for Western blot analysis. The membranes were probed with anti-IL-1R (sc-688; Santa Cruz Biotechnology), anti-TβRI (sc-399; Santa Cruz Biotechnology), or anti-TβRII (sc-1700; Santa Cruz Biotechnology).

Gene Expression Analysis.

293C6 cells were treated with IL-1β or TGFβ2 for 4 h. Total RNAs were purified and analyzed by using CodeLink arrays (GenUs Biosystems, Northbrook, IL). Data were analyzed by using GenUs software. Expression was normalized to the levels of GAPDH mRNAs in all the samples. The levels of mRNAs induced by IL-1β or TGFβ2 were compared with the levels in untreated 293C6 cells.

ELISA for IL-1α and IL-1β Quantification.

Conditioned media were collected as described above. Human IL-1α and IL-1β Quantikine colormetric sandwich ELISA (R&D Systems, Minneapolis, MN) was carried out following the manufacturer's protocol.