We used a gene knockout approach to elucidate the specific roles played by the Jun‐N‐terminal kinase (JNK) and NF‐κB pathways downstream of TNF‐α in the context of α(2) type I collagen gene (COL1A2) expression. In JNK1−/−‐JNK2−/− (JNK−/−) fibroblasts, TNF‐α inhibited basal COL1A2 expression but had no effect on TGF‐β‐driven gene transactivation unless jnk1 was introduced ectopically. Conversely, in NF‐κB essential modulator−/− (NEMO−/−) fibroblasts, lack of NF‐κB activation did not influence the antagonism exerted by TNF‐α against TGF‐β but prevented repression of basal COL1A2 gene expression. Similar regulatory mechanisms take place in dermal fibroblasts, as evidenced using transfected dominant‐negative forms of MKK4 and IKK‐α, critical kinases upstream of the JNK and NF‐κB pathways, respectively. These results represent the first demonstration of an alternate usage of distinct signaling pathways by TNF‐α to inhibit the expression of a given gene, COL1A2, depending on its activation state.
Through their ability to modulate the expression of extracellular matrix (ECM) components and ECM‐degrading enzymes, cytokines and growth factors orchestrate the balance between ECM destruction and neosynthesis and therefore play an important role in the control of tissue homeostasis and repair (Mutsaers et al., 1997; Uitto and Kouba, 2000). Disruption of the fragile equilibrium between anabolic and catabolic cytokines in favor of TGF‐β may lead to excessive collagen deposition, the hallmark of fibrotic conditions.
TGF‐β signals via serine/threonine kinase transmembrane receptors that phosphorylate cytoplasmic mediators of the Smad family. Phospho‐Smads are translocated into the nucleus, where they function as transcription factors, binding DNA either directly or in association with other proteins (Attisano and Wrana, 2000). Smad signaling may be blocked by TNF‐α, via mechanisms that implicate either c‐Jun (Verrecchia et al., 2000) or NF‐κB (Bitzer et al., 2000).
Jun‐N‐terminal kinases (JNKs), members of the family of mitogen‐activated protein kinases (MAPKs), are activated upon exposure of cells to cytokines, growth factors and environmental stresses (Davis, 2000). Three distinct genes encode JNKs: jnk1, jnk2 and jnk3, the former two being ubiquitously expressed. Dual threonine/tyrosine phosphorylation of JNK by the MAPK kinases MKK4 and MKK7 results in its activation and nuclear translocation and subsequent phosphorylation of transcription factors, such as c‐Jun or ATF2 (Davis, 2000).
In the human COL1A2 promoter, a Smad3/4 gene target downstream of TGF‐β (Chen et al., 1999; Verrecchia et al., 2001a), a region between nucleotides –313 and –235 and containing AP‐1, NF‐κB, Sp1, Smad3/4 and C/EBP binding sites, has been shown to participate in the regulatory programs activated by TGF‐β and TNF‐α to control type I collagen expression (Ghosh, 2002).
In this study, we demonstrate the critical role of JNK activation for TNF‐α in antagonizing TGF‐β‐induced COL1A2 transactivation, whereas NF‐κB activation is essential for inhibition of basal COL1A2 expression. These results provide the first demonstration of an alternate usage of distinct signaling mechanisms by TNF‐α to repress the COL1A2 gene depending on its activation state.
Results and discussion
Modulation of JNK and NF‐κB activities by TNF‐α in JNK−/− and NEMO−/− fibroblasts
We examined the endogenous modulation of the COL1A2 gene in immortalized fibroblast lines derived from JNK−/− and NF‐κB essential modulator−/− (NEMO−/−) mouse embryos (Sabapathy et al., 1999; Schmidt‐Supprian et al., 2000). First, to validate our experimental system, we examined the modulation of the JNK and NF‐κB pathways by TNF‐α in these cell lines. Using NF‐κB‐ and AP‐1‐ pathway‐specific reporter constructs in transient cell transfection experiments, together with western analysis of endogenous phospho‐c‐Jun content, we determined that JNK−/− fibroblasts are devoid of JNK activity but exhibit ‘normal’ NF‐κB response, whereas NEMO−/− fibroblasts lack NF‐κB activity and exhibit a functional JNK pathway, allowing efficient c‐Jun phosphorylation and AP‐1‐dependent transactivation in response to TNF‐α (data not shown). Notably, TGF‐β activated neither the JNK nor the NF‐κB pathways in any fibroblast line tested (data not shown).
Respective roles of JNK and NF‐κB in the regulation of COL1A2 gene expression by TNF‐α in mouse embryo fibroblasts
The effect of TNF‐α on TGF‐β‐induced COL1A2 expression was examined by measuring the modulation of COL1A2 steady‐state mRNA levels by TGF‐β and TNF‐α. Experiments were carried out in medium supplemented with 1% fetal calf serum, a condition that allows potent COL1A2 upregulation by TGF‐β (Chung et al., 1996), whereas higher serum concentrations may interfere with TGF‐β, due to the ability of α2‐macroglobulin to complex and inhibit growth factors (O'Connor‐McCourt and Wakefield, 1987). Strong enhancement of COL1A2 mRNA steady‐state levels (3.5‐fold) was observed in response to TGF‐β in the three cell types (Figure 1A, lanes 2, 6 and 10). TNF‐α antagonized TGF‐β in both wild‐type (wt) and NEMO−/− fibroblasts (Figure 1A, lanes 4 versus 2 and 12 versus 10, respectively), but not in JNK−/− fibroblasts (Figure 1A, lane 8 versus 6). Type I collagen production paralleled the modulation of COL1A2 mRNA steady‐state levels (Figure 1B, upper panel). Specifically, TGF‐β treatment resulted in increased type I collagen production by wt, JNK−/− and NEMO−/− fibroblasts (Figure 1B, lanes 2, 6 and 10). TNF‐α efficiently blocked TGF‐β‐induced accumulation of immunoreactive type I collagen in wt and NEMO−/− fibroblasts (Figure 1B, lanes 4 and 12, respectively) but not in JNK−/− fibroblasts (Figure 1B, lane 8). Actin levels showed no modulation by cytokines in either cell type (Figure 1B, lower panel). These data indicate a critical role of JNK, but not NF‐κB, in mediating the antagonistic effect of TNF‐α against TGF‐β‐induced type I collagen gene expression. Notably, despite the lack of NF‐κB activation, no apoptosis was detected in TNF‐α‐treated (0, 1, 10 and 100 ng/ml up to 24 h) immortalized NEMO−/− fibroblasts (data not shown).
Under experimental conditions most favorable for high basal COL1A2 expression, e.g. medium containing 10% serum (Chung et al., 1996) adequate for studying inhibitory mechanisms by cytokines (Higashi et al., 1998; Kouba et al., 1999; Czuwara‐Ladykowska et al., 2001), TNF‐α inhibited COL1A2 mRNA levels (Figure 2A) and type I collagen production (Figure 2B) in both wt and JNK−/− fibroblasts but had no effect in NEMO−/− fibroblasts.
Next, we performed transient transfection experiments with pMS3.5CAT, a plasmid that contains 3.5 kb of 5′ regulatory sequences of the COL1A2 gene, driving the expression of the CAT gene. Under low serum conditions, TNF‐α induced (i) a 30% reduction in COL1A2 promoter activity in wt and JNK−/− fibroblasts exclusively and (ii) a marked inhibitory activity against TGF‐β in wt and NEMO−/− fibroblasts but not in the JNK−/− fibroblasts (Figure 3A). In contrast, in medium containing 10% serum, a 70% repressive effect of TNF‐α on basal COL1A2 promoter activity was observed in wt and JNK−/− fibroblasts but not in NEMO−/− fibroblasts (Figure 3B). These results, consistent with the data presented in Figure 2 for endogenous collagen type I expression, indicate that JNK and NF‐κB exert distinct effects on COL1A2 expression at the transcriptional level.
c‐Jun requires functional JNK to antagonize TGF‐β‐induced COL1A2 promoter transactivation
c‐Jun, the prototypic JNK substrate, is a key effector of TNF‐α inhibitory activity on Smad signaling (Verrecchia et al., 2000) and blocks TGF‐β‐induced COL1A2 promoter transactivation (Chung et al., 1996). In addition, an antisense c‐jun expression vector completely prevents TNF‐α antagonism against TGF‐β‐induced COL1A2 promoter transactivation (data not shown). To determine whether c‐Jun inhibitory potential against TGF‐β required functional JNK, c‐Jun and pMS3.5CAT were co‐transfected into wt, JNK−/− and NEMO−/− fibroblasts subsequently treated with TGF‐β. c‐Jun expression abolished TGF‐β effects in wt and NEMO−/− fibroblasts but had no effect in JNK−/− fibroblasts (Figure 4A). The lack of c‐Jun effects in JNK−/− fibroblasts was actually due to the jnk gene knockout and not the non‐specific altered cytokine responsiveness in this cell type, as ectopic expression of jnk1, which allowed for c‐jun phosphorylation in response to TNF‐α (see Supplementary figure 1A and B available at EMBO reports Online), entirely rescued the antagonistic effect of c‐Jun against TGF‐β (Figure 4B). Similarly, jnk1 expression restored the ability of TNF‐α to interfere with TGF‐β‐induced collagen gene expression in JNK−/− fibroblasts (see Supplementary figure 1C).
Respective roles of JNK and NF‐κB in the regulation of COL1A2 gene expression by TNF‐α in human dermal fibroblasts
The JNK and NF‐κB cascades were targeted in human dermal fibroblasts by means of transfected dominant‐negative (D/N) mutant forms of the upstream kinases MKK4 and IKK‐α, respectively (see Supplementary figure 2). As shown in Figure 5A, and as expected from our previous work using this cell type (Chung et al., 1996), TNF‐α prevented 90% of the TGF‐β response. D/N MKK4 blocked the inhibitory effect of TNF‐α against TGF‐β, whereas D/N IKK‐α had no effect, indicating that the MKK4/JNK cascade is critical for TNF‐α to interfere with TGF‐β‐induced COL1A2 promoter activation, whereas the NF‐κB cascade is not. Indeed, in NEMO−/− fibroblasts, transfection of D/N MKK4, and that of an antisense c‐jun vector, prevented TNF‐α antagonism against TGF‐β‐induced COL1A2 promoter transactivation (data not shown).
Under high serum conditions, and consistent with the effects observed with the knockout fibroblasts, inhibition of COL1A2 promoter activity by TNF‐α was abolished by D/N IKK‐α, not by D/N MKK4 (Figure 5B). Notably, the slight inhibitory effect exerted by TNF‐α on basal promoter activity under low serum conditions (Figure 5A) was abolished by D/N IKK‐α but not by D/N MKK4.
Using independent approaches to interfere with the JNK and NF‐κB pathways, we have provided compelling evidence for a central role of JNK in allowing the antagonistic activity of TNF‐α and c‐Jun against TGF‐β‐induced type I collagen gene expression, whereas NF‐κB activity, although critical for the inhibitory effect of TNF‐α on basal COL1A2 expression, plays no role in mediating the former phenomenon. A schematic representation summarizing our findings is provided in Figure 6: TNF‐α activates distinct signaling pathways that regulate a given target gene independently, according to its activation state. Direct NF‐κB‐driven repression occurs via a proximal NF‐κB binding site within the COL1A2 promoter (Kouba et al., 1999). COL1A2 is a direct Smad3/4 gene target (Chen et al., 1999; Verrecchia et al., 2001a). Our data indicate that JNK blocks TGF‐β‐induced type I collagen gene expression via c‐Jun phosphorylation, certainly by preventing Smad signaling. Mechanisms underlying c‐Jun inhibitory activity include (i) physical association of Jun and Smad3, not compatible with Smad–DNA complex formation (Shi et al., 1998; Verrecchia et al., 2001b), and (ii) sequestration of the shared transcriptional co‐activator p300 (Verrecchia et al., 2000). In accordance with our findings, overexpression of constitutively active forms of either MEKK1 or MKK4, kinases involved in JNK activation, enhances Smad–Jun associations and represses Smad‐dependent transcription (Dennler et al., 2000).
With regard to JNK function in the context of ECM turnover, our findings are complementary to a recent study indicating that a synthetic inhibitor of JNK, SP600125, suppresses interleukin‐1‐induced phospho‐Jun accumulation, Jun–DNA interactions and interstitial collagenase (MMP‐1) gene expression in synovial fibroblasts (Han et al., 2001). It appears therefore that the benefit of JNK targeting in degenerative inflammatory diseases such as rheumatoid arthritis results not only from blocking degradative events induced by interleukin‐1 or TNF‐α but also by maintaining the anabolic functions of TGF‐β on ECM deposition, otherwise inhibited by these cytokines. Conversely, means to activate the JNK pathway may be of interest in pathological situations where interfering with TGF‐β signaling is critical, such as in fibrosis.
Immortalized fibroblast cell lines were derived from wt, JNK1−/−‐JNK2−/− (Sabapathy et al., 1999) and NEMO−/− (Schmidt‐Supprian et al., 2000) mouse embryos in which targeted disruption of the jnk1 and jnk2 or the nemo genes, respectively, had been performed. Human dermal fibroblasts were established by explanting neonatal foreskins. Cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% heat‐inactivated fetal calf serum, 2 mM glutamine and antibiotics (100 U/ml penicillin, 50 μg/ml streptomycin‐G and 0.25 μg/ml Fungizone). Human recombinant TGF‐β1 (R&D Systems, Minneapolis, MN), referred to as TGF‐β throughout the text, and human recombinant TNF‐α (Roche Diagnostic, Indianapolis, IN) were both used at a concentration of 10 ng/ml.
Total RNA was obtained using an RNeasy kit (Qiagen, Hilden, Germany) and analyzed by northern hybridization (20 μg/lane) with 32P‐labeled cDNA probes for COL1A2 and GAPDH, as described previously (Mauviel et al., 1991). The hybridization signal was quantified with a phosphoimager (Storm 840, Amersham‐Pharmacia Biotech, Uppsala, Sweden).
Whole cell lysates from fibroblasts were prepared by washing cells twice in 1× PBS, followed by scraping into Laemmli buffer (62.5 mM Tris–HCl pH 6.8, 2% SDS, 10% glycerol, 0.5 mM PMSF). Protein (25–100 μg) was denatured by heating at 95°C for 3 min prior to resolution by SDS–PAGE. After electrophoresis, proteins were transferred to Hybond ECL nitrocellulose filters (Amersham‐Pharmacia), immunoblotted with either anti‐type I collagen (Southern Biotech, Birmingham, AL) or anti‐actin (Sigma, St Louis, MO) antibodies, incubated with a horseradish peroxidase‐conjugated goat‐anti‐rabbit secondary antibody (Santa Cruz Biotechnology, Santa Cruz, CA), developed according to ECL protocols (Amersham‐Pharmacia) and revealed with a phosphoimager (Storm 840).
pMS3.5/CAT, a plasmid containing ∼3.5 kb of human COL1A2 promoter linked to the CAT reporter gene, has been described previously (Boast et al., 1990). pRSV‐β‐galactosidase (Promega, Madison, WI) was used to control transfection efficiency. pNF‐κB‐lux and pAP1‐TA‐lux (Clontech, Palo Alto, CA) were used to determine NF‐κB‐ and AP‐1‐driven transcription, respectively. For c‐Jun expression, we used a full‐length human cDNA cloned into pRSV (Chiu et al., 1989). D/N MKK4, D/N IKK‐α and jnk1 expression vectors have been described previously (Atfi et al., 1997; Sabapathy et al., 2001).
Transient cell transfections and reporter assays.
Transient cell transfections were performed with the calcium phosphate/DNA co‐precipitation procedure using a commercial assay kit (Promega). Reporter activities were determined as described previously (Verrecchia et al., 2000).
Supplementary data are available at EMBO reports Online.
Drs A. Atfi (INSERM U482, Paris, France) and G. Courtois (Pasteur Institute, Paris, France), S. Dennler and J.‐M. Gauthier (Glaxo‐Wellcome, Les Ulis, France) and M. Pasparakis (EMBL, Rome, Italy) provided cellular and molecular reagents essential for these studies. Dr L. Michel provided invaluable help in the apoptosis experiments. This work was supported by INSERM, the Association pour la Recherche contre le Cancer (ARC), Ligue Nationale contre le Cancer, Comité de Paris and Electricité de France (Service de Radioprotection).
- Copyright © 2002 European Molecular Biology Organization