Transforming Growth Factor (TGF)-β-activated Kinase 1 Mimics and Mediates TGF-β-induced Stimulation of Type II Collagen Synthesis in Chondrocytes Independent of Col2a1 Transcription and Smad3 Signaling*

Transforming growth factor (TGF)-β, bone morphogenetic protein (BMP), and interleukin-1β activate TGF-β-activated kinase 1 (TAK1), which lies upstream of the p38 MAPK, JNK, and NF-κB pathways. Our knowledge remains incomplete of TAK1 target genes, requirement for cooperative signaling, and capacity for shared or segregated ligand-dependent responses. We show that adenoviral overexpression of TAK1a in articular chondrocytes stimulated type II collagen protein synthesis 3–6-fold and mimicked the response to TGF-β1 and BMP2. Both factors activated endogenous TAK1 and its activating protein, TAB1, and the collagen response was inhibited by dominant-negative TAK1a. Isoform-specific antibodies to TGF-β blocked the response to endogenous and exogenous TGF-β but not the response to TAK1a. Expression of Smad3 did not stimulate type II collagen synthesis or enhance that caused by TGF-β1 or TAK1a, in contrast to its effects on its endogenous targets, CTGF and plasminogen-activated inhibitor-1. TAK1a, overexpressed alone and immunoprecipitated, phosphorylated MKK6 and stimulated the plasminogen-activated inhibitor-1 promoter following transient transfection; both effects were enhanced by TAB1 coexpression, but type II collagen synthesis was not. Stimulation by TAK1a or TGF-β did not require increased Col2a1 mRNA, and TAK1 actually reduced Col2a1 mRNA in parallel with the cartilage markers, SRY-type HMG box 9 (Sox9) and aggrecan. Thus, TAK1 increased target gene expression (Col2a1) by translational or posttranslational mechanisms as a Smad3-independent response shared by TGF-β1 and BMP2.

TGF-␤ and BMP exhibit both common and specific effects. Both induce chondrogenesis and associated Col2a1 expression in limb bud mesenchyme and mesenchymal stem cells (9,11,12,21,22). Conversely, TGF-␤ inhibits and BMP enhances chondrocyte maturation and type X collagen expression in the growth plate (14,16,19). In general, such specificity is mediated by different ligands, similar but distinct receptors with integral serine/threonine kinase activity, and recruitment and phosphorylation of distinct subsets of receptor Smads (23)(24)(25). TGF-␤ binding to the type II TGF-␤ receptor (T␤RII) leads to recruitment, phosphorylation, and activation of the type I receptor (T␤RI). The activated T␤RI kinase phosphorylates Smad2 or -3, enabling association with the universal common Smad, Smad4, prior to nuclear translocation. Heteromeric Smads then activate or suppress transcription in cooperation with other transcription factors, coactivators, or repressors (1,26,27). Similarly, BMP signals through the kinase activity of its high affinity type I receptors, BMPR1a and BMPR1b, in combination with BMPRII. Receptor activation in this case leads to phosphorylation of a different subset of receptor Smads, Smad1, -5, and -8, before interaction with Smad4 (1,26).
High level expression of type II collagen is a phenotypic marker for chondrogenesis and a hallmark of the load-bearing structure of adult articular cartilage. Col2a1 gene expression is developmentally regulated by the HMG domain transcription factor Sox9 (43) in cooperation with L-Sox5 and Sox6 (44) through binding to sequences in the first intron enhancer (45). Other functional regulatory sites are present in the Col2a1 promoter/enhancer (46); however, the Sox9 binding domain is sufficient for tissue-specific expression (45). Although TGF-␤ has been shown to have varied effects on collagen synthesis in chondrocyte culture (47,48), Chadjichristos et al. (49) have recently shown that TGF-␤ suppresses expression driven by the Col2a1 proximal promoter, extending similar earlier conclusions using the first intron enhancer (50,51).
Here we demonstrate, in rabbit articular chondrocytes, enhanced synthesis of type II collagen protein by TGF-␤ and BMP2. This stimulation activated endogenous TAK1, was blocked by dominant-negative TAK1a, and was mimicked by adenoviral overexpression of TAK1a or TAK1a and TAB1 together. The response to TGF-␤ or TAK1a did not increase Col2a1 message and was not replicated by adenoviral expression of Smad3. Thus, stimulation of type II collagen synthesis represents a response shared by both TGF-␤ and BMP2 that does not require increased transcription or stability of Col2a1 message or cooperation with Smad3 signaling.
Plasmids-Plasmids were obtained from the following sources: pGEX-MKK6 (R. Davis) (52), p800LUC (D. Loskutoff) (53), pUC18 (ATCC), rabbit Col2a1 3Ј-untranslated region (E. Vuorio) (54), and pNF-B-Luc (Clontech). To create vectors for TAK1 and TAB1 expression, the adenoviral cosmid AxCAwt (55) (TakaRa; containing the chicken ␤-actin promoter, first intron enhancer, and the CMV enhancer (56)) was cut with SalI and NcoI to remove adenoviral sequences and the CMV enhancer prior to ligation with a SalI-NcoI adapter. The first intron enhancer was then removed with HinfI and ClaI and resealed with a HinfI-NotI-ClaI adapter to produce pA3. A SalI site at the 3Ј-end of the CMV enhancer of pCMV6-X4 (Origene) was generated from a cryptic site by QuikChange site-directed mutagenesis (Stratagene). The SalI-NotI fragment was inserted in the same site in pA3 to produce pC containing only the CMV promoter. The 376-bp SalI-NcoI fragment from AxCAwt containing the CMV enhancer was inserted in the same site in pA3 to generate pCEA3 containing the minimal ␤-actin promoter preceded by the CMV enhancer. An arrayed human heart cDNA library (Origene) was screened for full-length cDNA for hTAK1a and hTAB1 by PCR using primers designed from published sequences (29,57). The 3Ј-untranslated region was truncated, and KpnI, XbaI, and NotI sites were inserted 3Ј to the wild type stop codon in the library clone pCMV6-X4-hTAK1a by Excite deletional mutagenesis (Stratagene). The resultant BspHI-Xba fragment was inserted into the unamplified library clone. Similarly, pCMV6-X4-hTAB1 was shortened, and a FLAG epitope was inserted using the following primers: forward, 5Ј-GACGG-TACCAACTGATCTAGATTGCGGCCGCGGTCATAGC-3Ј; reverse, 5Ј-GCTGGTACCGATCTTATCGTCGTCATCCTTGTAATCTTGTGCCGG-TGCTGTCACCACGCTCTGC-3Ј. The PCR product was cut with KpnI and ligated prior to transfer of the BsrBI-XbaI fragment to the same site in the unamplified clone. The shortened TAK1a and TAB1 constructs where subcloned into NotI sites to yield expression vectors pC-hTAK1a and pCEA3-hTAB1. Cloning was monitored and verified by sequencing. A tandem vector was made by inserting the blunted SalI-PmeI pC-hTAK1a expression cassette into PmeI-cut pCEA3-hTAB1. Dominant negative mutants of TAK1 were generated by QuikChange (Stratagene) mutagenesis using 40-bp complementary primers centered at the mutation site. Sequenced mutant fragments for K63A and K63W replaced the KasI-MluI fragment to produce pC-hTAK1a-KA and pC-hTAK1a-KW. Similarly, the S192A fragment replaced the RsrII-MluI fragment to create pC-hTAK1a-SA. The double mutant pC-hTAK1a-KWSA was generated sequentially.
Adenoviral Vector Construction-The adenoviral vectors Ad-pC, Ad-pCEA3, Ad-pCEA3(pC), Ad-pC-hTAK1a, Ad-pCEA3-TAB-FL, and pCEA3-TAB-FL(pC-hTAK1a) (Tandem) were constructed using components obtained from TakaRa. Expression cassettes were excised with SalI and PmeI, blunted, and ligated to SwaI-cut promoterless transfer cosmid pAxcw (55). Following packaging (Promega), recombinant cosmids were used to generate recombinant ⌬E1 and ⌬E3 adenoviruses by homologous recombination in 293 cells transfected with cosmid and restricted viral DNA-terminal protein complexes (58). Adenoviruses were isolated by plaque purification and two rounds of dilution to 0.5 plaque-forming units/well and propagated in 293 cells. Viral stocks were stored as lysates or CsCl-purified particles in aliquots at Ϫ80°C in 10 mM Tris, pH 8.0, 2 mM MgCl 2 , 5% sucrose (59). Titers were determined in 293 cells by TCID 50 assays and corrected to plaque-forming units/ml. The Adeno-X Rapid Titer protocol (BD Clontech) was recently used and produced titers 10-fold higher than paired analysis by TCID 50 ; these were normalized to TCID 50 values for data consistency. Ad-CMV-LacZ was obtained from Quantum. Adenovirus expressing wild type Smad3 was generously provided by Wei Shi (Children's Hospital Los Angeles) and repurified before use.
Cell Culture and Collagen Synthesis-Chondrocytes were released from the articular cartilage of 7-week-old New Zealand White rabbits as previously described (60), except that overnight enzymatic digestion was performed in a spinner flask. Cells were plated at 1.5-2 ϫ 10 6 cells/T75 flask and cultured in 10% heat-inactivated (HIA; 55°C, 30 min) fetal calf serum or HIA-dialyzed fetal calf serum in DMEM (high glucose) for 10 days. Chondrocytes were washed in PBS and released by brief incubation with a film of 0.05% trypsin, 0.53 mM EDTA. Cells were plated for analysis of collagen synthesis in 10% HIA fetal calf serum/ DMEM in 48-well plates at confluent density (5.5 ϫ 10 4 cells/well) by layering beneath 5% HIA fetal calf serum/DMEM. Cells were conventionally plated at proportional densities in 60-and 100-mm dishes for RNA isolation. Typically, chondrocytes were allowed to attach for 24 h, washed with DMEM, and cultured or infected with viral vectors for 24 h in 0.125 ml of 0.3% ITSϩ/DMEM (BD Biosciences). The duration of viral exposure, referred to in the text and figure legends, includes this period of infection. Subsequently, cells were fed with the same medium, treated with recombinant human TGF-␤1 or BMP2 (R&D Systems) for 24 h and then labeled in fresh medium with tritiated proline for 24 h. Labeling medium was 0.3% ITSϩ/DMEM, 62.5 g/ml ␤-aminoproprionitrile, 25 g/ml ascorbate, 40 Ci/ml [5-3 H]proline (Amersham Biosciences), and growth factors. Cultures (medium and cell layer, together) were terminated by freezing, thawed, adjusted to 0.5 N acetic acid and 0.5 mg/ml pepsin, and rocked overnight at 4°C. Following two cycles of lyophilization in the culture plate, samples were dissolved in 0.250 ml of SDS-PAGE sample buffer without reducing agent at 55°C for 30 min before SDS-PAGE on gradient gels and fluorography (60). Triplicate samples were analyzed separately or as a physical pool by digital image analysis of ␣1(II) collagen bands. For analysis of the collagen phenotype, labeled collagen was precipitated in 80% acetone at 4°C for 45 min, washed twice with 40% ethanol, digested with CNBr, and mapped in two dimensions as previously described (61,62).
Transfection and Luciferase Assays-Primary chondrocytes were released and plated beneath 5% HIA fetal calf serum/DMEM at 1.7 ϫ 10 4 cells/well in 48-well plates. After 18 -24 h, cells were washed with serum-free medium, fed with ITSϩ/DMEM, and transfected for 3 h with Lipofectamine Plus (Invitrogen). Each well contained 1 l of Plus reagent, 0.5 l of Lipofectamine, 25 ng of p800Luc or 12.5 ng of pNF-B, and test plasmids, and the DNA content was adjusted to 200 ng/well with pUC18. Cells were washed with ITSϩ/DMEM and fed this medium until termination 48 h later. Substrate buffer (50 l) was used for lysis, and 10 l were transferred to 20 l of 4 mM CaCl 2 , 4 mM MgCl 2 , 50 mM Hepes, pH 7.5, followed by the addition of 10 l of 2ϫ Luclite substrate (Packard). Replicate wells (n ϭ 4) were analyzed for luciferase activity using a scintillation counter and single photon counting (Beckman).
Western Blotting-Cellular proteins were solubilized in SDS-PAGE sample buffer containing 5% ␤-mercaptoethanol, separated by SDS-PAGE, and transferred to nitrocellulose in 25 mM Tris base, 0.2 M glycine, 20% methanol (pH 8.5), and 0.02% SDS. Blots were blocked for 2 h at 23°C with 5% nonfat dry milk in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.1% Tween 20 (TBS-Tw), and incubated with primary antibodies overnight at 4°C in 5% bovine serum albumin in TBS-Tw. After brief washing in TBS-Tw, blots were incubated at 23°C for 40 min with horseradish peroxidase-coupled second antibodies in 5% nonfat dry milk in TBS-Tw. Washed blots were visualized with chemiluminescence using SuperSignal West Dura substrate (Pierce). Molecular weight markers for chemiluminescent detection (PA-Markers) were obtained from Chemicon.

TAK1a
Stimulates PAI-1 Promoter Activity in Chondrocytes-Transient transfection of secondary cultures of differentiated chondrocytes with hTAK1a in combination with the TGF-␤-responsive p800-PAI-1-Luc reporter stimulated this promoter as well as 5 ng/ml TGF-␤ ( Fig. 1). This 10-fold stimulation was only reached at the highest concentrations of TAK1a, whereas a similar maximum stimulation was reached by constitutively active TAK1a-⌬N at much lower concentrations. In both cases, the response was saturated, and only cotransfection with TAB1 exceeded this plateau. At TAK1a concentrations below 2 ng/well, PAI-1-Luc activity was similar for TAK1a-⌬N, TAK1a ϩ TAB1, and tandem expression of both TAK1a and TAB1. Consequently, TAK1a activated by deletion (⌬N) or TAB1 overexpression appear to exhibit similar specific activities. In contrast, the slower increase in PAI-1-Luc activity produced by TAK1a alone suggests that higher concentrations were required for TAK1a to activate itself or be activated by another kinase or molecular interaction. Fig. 1 defines an optimal ratio of expression for exogenous TAK1a and TAB1 using these constructs when TAK1a was increased against a constant TAB1 concentration. However, when an optimal concentration for TAK1a was evaluated with increasing TAB1, luciferase activity declined dramatically (data not shown). Both results demonstrate the importance of the TAK1a/TAB1 expression ratio for biological activity, which was not optimal in the tandem vector governed by the relative strength of the small dual promoters. A further 1.5-2-fold increase in each of the optimal activities for plasmids containing TAK1a or FIG. 1. TAK1a alone or in combination with TAB1 stimulates PAI-1 promoter activity in chondrocytes. Secondary chondrocytes were transfected and relative PAI-luciferase activity was determined 48 h after transfection. Promoter-only vectors were pC (for TAK1a) and pCEA3(pC) (for tandem). Expression vectors were pCEA3-TAB1, pC-hTAK1a, pC-hTAK1a-⌬N (constitutively active), and pCEA3-hTAB1(pC-hTAK1a) (tandem, Tdm). The plasmid dose represents pC-hTAK1a content; all other plasmids amounts were adjusted so that the molar equivalent of the pC-hTAK1a dose was transfected. The concentration of pCEA3-TAB1 was always 0.5 ng/well. *, level of stimulation induced by 5 ng/ml TGF-␤1. Error bars, S.E. (n ϭ 4).
TAK1a and TAB1 was observed if the transfected cultures were simultaneously treated with 5 ng/ml TGF-␤1 (data not shown) and may reflect parallel activation of Smad2/3 or other postreceptor signals.
Adenoviral Overexpression of TAK Stimulates Type II Collagen Synthesis-We have constructed adenoviral expression vectors for TAK1a and TAB1, because the transfection efficiency of chondrocytes is low. Infection with adenoviral vectors produced transduction efficiencies of Ͼ90%, based on ␤-galactosidase expression, and permitted the use of conventional [ 3 H]proline labeling and SDS-PAGE to monitor the effects of TAK1a on type II collagen synthesis at the final protein level rather than relying on mRNA levels or reporter constructs.
In secondary chondrocytes plated at confluent density, TGF-␤1 stimulated type II collagen synthesis 6-fold, whereas infection with TAK1a and TAK1a-⌬N adenoviral vectors pro-duced 4 -5-fold stimulation ( Fig. 2A). Thus, TAK1a mimicked TGF-␤1 in this response. Infection with the tandem vector, expressing both TAK1a and TAB1, produced the same response as TAK1a alone. This suggests that the interaction with TAB1 is not necessary under conditions of overexpression but may still be required for endogenous TAK1 activation. Simultaneous treatment with TAK1a and TGF-␤1 produced little change relative to TAK1a alone, suggesting minimal involvement by parallel signaling pathways downstream of T␤RI. Collagen produced by TAK1a-⌬N treatment was predominantly ␣1(II), as judged by two-dimensional CNBr peptide mapping (Fig. 2B). This indicates that the response to TAK1a overexpression was not due to a change to type I collagen (␣1(I)-chain) synthesis but rather reflects enhanced production of collagens of the differentiated phenotype, predominantly type II collagen.
Autocrine TGF-␤ and Smad3 Are Not Required for TAK1adependent Stimulation-TGF-␤1 has been shown to activate its own promoter and establish a positive autocrine feedback loop (63). To test whether TAK1a mimicked this role of TGF-␤1 and caused stimulation of type II collagen synthesis by inducing TGF-␤ production, chondrocytes were cultured in the presence and absence of a mixture of blocking antibodies directed against all three isoforms of TGF-␤. Pretreatment of TGF-␤1containing medium with antibodies and their continued presence during the following 48-h culture period with TGF-␤ not only blocked the 4-fold stimulation of ␣1(II) synthesis caused by TGF-␤1 but reduced synthesis to below basal levels (Fig.  3A). Thus, basal synthesis was partially due to an autocrine response to endogenous TGF-␤ production, and the antibody mix effectively blocked signaling by both endogenous and exogenous TGF-␤1. In contrast, the 3-fold stimulation produced by Ad-pC-hTAK1a or ⌬N was not blocked by the antibody mix but rather stimulated by it. TAK1a-dependent stimulation, therefore, did not require endogenous TGF-␤ or its postreceptor signals, including Smad activation. The cause of the antibodydependent enhancement is not known, but it may be due to release from effects of inhibitory Smads.
The requirement for Smad signaling in TAK1a or TGF-␤1 stimulation of type II collagen synthesis was addressed directly using an adenoviral vector expressing Smad3 (Fig. 3B). TGF-␤1 or TAK1a alone (5-20 MOI) produced a 3-fold stimulation of collagen synthesis, whereas Smad3 had no effect at its maximally effective dose of 20 MOI (Fig. 3B, lane 7). Signaling by Smad3 did not cooperate with TAK1a, since coinfection did not elevate the response above that of TAK1a alone but actually completely blocked the effects of TAK1a (lanes 8 and 9) when TAK1a expression was still within the effective range (compare TAK1 expression in lanes 4 and 5 with lanes 8 and 9). Similarly, overexpressed Smad3 also blocked stimulation by TGF-␤1, whereas TAK1a did not (lanes 10 and 11). This inhibitory effect of Smad3 was still present when both TAK1a and TGF-␤1 were part of the treatment (lanes 12 and 13).
The expected functionality of Ad-Smad3 in this experiment was verified by monitoring expression of Smad3-responsive endogenous targets, PAI-1 and CTGF, by Western blotting (Fig. 3B). TGF-␤1 enhanced PAI-1 expression 9-fold, whereas TAK1a alone and Smad3 alone exceeded this level, producing 110-and 83-fold stimulation, respectively. In addition, Smad3 cooperated with TAK1a, generating a 350-fold increase, with TGF-␤1 (340-fold increase), and with both to generate a 2600fold stimulation. CTGF expression was increased 25-fold by TGF-␤1 and 13-and 17-fold by TAK1a and Smad3, respectively. The combined response was additive and was only slightly increased by costimulation with TGF-␤1. Thus, substantial responses of these marker genes to Smad3 alone or in combination with TAK1a or TGF-␤1 distinguish PAI-1 and CTGF signaling mechanisms from those involved with stimulating type II collagen synthesis.
TAK1a Mediates the Stimulation of Type II Collagen Synthesis by TGF-␤ and BMP2-The involvement of TAK1 in TGF-␤dependent stimulation of ␣1(II) synthesis was addressed directly by treating chondrocytes with adenoviral vectors expressing dominant negative TAK1a-K63W or K63A before exposure to TGF-␤1 (Fig. 4A). Both constructs provided dosedependent inhibition of ␣1(II) synthesis induced by TGF-␤1; TAK1a-K63W provided maximal inhibition, reducing synthesis to basal levels. In the absence of added TGF-␤, both TAK1a-K63W and K63A also suppressed basal synthesis, again supporting participation of endogenous TGF-␤ in basal synthesis. Because BMP also has been shown to activate TAK1 (28), we evaluated its capacity to stimulate type II collagen synthesis by a TAK1-dependent mechanism. BMP2 produced a 5-fold stimulation of synthesis that was reduced by 60% in the presence of TAK1a-K63A. Thus, two members of the TGF-␤ superfamily, TGF-␤1 and BMP2, share the capacity for TAK1-mediated stimulation of type II collagen synthesis.
Kinase Activity of Overexpressed TAK1-The kinase activity of endogenous and recombinant TAK1 was determined by immune complex kinase assays using a TAK1a C-terminal antipeptide antibody and MKK6 as substrate. Both adenovirally expressed TAK1a and TAK1a⅐TAB1, expressed from the tandem vector, produced TAK1a autophosphorylation and MKK6 phosphorylation, whereas the double mutant, TAK1a-KWSA, produced neither (Fig. 4B, upper panels). When immunoprecipitates were exposed to kinase conditions in the presence of only unlabeled ATP, washed, and then submitted to the radiolabeled kinase assay (Fig. 4B, lower panels), TAK1a expressed alone lost the capacity to phosphorylate MKK6 as well as itself. In contrast, coexpression of TAK1a and TAB1 maintained a strong capacity to phosphorylate MKK6, whereas autophosphorylation was lost due to prior incorporation of unlabeled ATP. These results and those from Fig. 4C demonstrate that (a) only 10% of TAK1a overexpressed alone is activated to participate in downstream signaling in the absence of exogenous TAB1 and (b) such activation apparently does not utilize TAB1 or does not utilize TAB1 through the usual mechanism that provides stabilization of kinase activity in vitro.
Activation of Endogenous TAK1 and TAB1-The capacity of dominant-negative TAK1a to block collagen synthesis by TGF-␤ and BMP2 strongly supports mediation by TAK1. This role for TAK1 also requires activation of endogenous TAK1 by these ligands. Immune complex kinase assays based on the TAK-ct antibody did not detect activation of TAK1 kinase activity or autophosphorylation (Fig. 4B) of endogenous TAK1 from control cells and those treated for 10 min with TGF-␤1 and IL-1␤. In subsequent studies, immunoprecipitation with TAK-m or antibodies raised to the whole TAK1 molecule also failed to detect TAK1 kinase activity. However, only small amounts of endogenous TAK1 were present in immunoprecipitates when compared by Western blotting to direct cell lysates (data not shown). Thus, the absence of TAK1 kinase activity may have been due to the lack of immunoprecipitation of activated TAK caused by interaction with scaffold proteins and/or signaling partners. Activation was therefore addressed directly by Western blotting of whole cell lysates (Fig. 4C). Treatment with 100 ng/ml BMP2 for 10 min increased the intensity of endogenous TAK1 and generated two new forms of TAK1 with retarded mobility characteristic of phosphorylation-dependent activation (30). TAB1 also exhibited a BMP-dependent phosphorylation pattern. TGF-␤ treatment (20 ng/ml) did not induce phosphorylation after 10 min, but after 30 min, the slowest migrating phosphorylated form of TAK1 was readily detected. The same amount of this phosphorylated form of TAK1 was also present when TAK1a or TAK1a and TAB1 were overexpressed (compare Fig. 4C, lanes 4 -6), indicating that generation of this form is controlled by a limiting endogenous cofactor(s). Following 30 min of TGF-␤1 stimulation, the endogenous unstimulated form of TAB1 was completely replaced by two slower migrating phosphorylated forms. Thus, a TGF-␤-depend- Adenoviral transduction with TAK1a and TAB1 together for 48 h completely modified all overexpressed TAK1a but only generated phosphorylated TAK1a with intermediate mobility (with the exception above). Thus, TAB1 binding to TAK1a was not sufficient to generate the most highly phosphorylated form of TAK1a in this cellular system. Similarly, only the interme-diate phosphorylated form of TAB1 was produced by tandem expression. Endogenous TAB1 was completely resistant to activation by TAK1a overexpressed alone (Fig. 4C, lane 5), despite the substantial kinase activity of TAK1a (Fig. 4B). Such resistance is consistent with the activity of a proposed inhibitor of activation of endogenous TAK1⅐TAB1 complexes (30).
RNA Levels Do Not Reflect the Stimulation of Type II Collagen Synthesis by TGF-␤ or TAK1a-The mechanism driving the stimulation of type II collagen synthesis was investigated by comparing mRNA levels and collagen synthesis in the same experiment. Treatment with TGF-␤1 for 48 h induced a 8.7-fold increase in ␣1(II) chain synthesis, whereas its mRNA, by ribo- FIG. 5. RNA levels do not reflect the stimulation of type II collagen synthesis by TGF-␤ and decline in response to adenoviral expression of TAK1a. A, chondrocytes were treated with TGF-␤1 (5 ng/ml) or infected at different MOI with adenoviral vectors expressing TAK1a constructs using the protocol in Fig. 2A. At termination of labeling for collagen synthesis, RNA was isolated from parallel cultures and submitted to RPA analysis with correction for L32 expression as a loading control. Triplicate cultures were separately analyzed for collagen synthesis to provide the quantitative data, and a physical pool was used for the fluorograph presented. The arrows indicate a change in lane labels for RPA analysis. Kinase-negative TAK1 (KN-TK), Ad-pC-hTAK1a-KWSA. B, the change in efficiency of type II collagen generation relative to the amount of specific mRNA was calculated from the data in A by dividing the -fold stimulation for type II collagen by the relative mRNA content. The black bars indicate treatment for 48 h with 5 ng/ml TGF-␤1.

FIG. 4. TAK1a mediates the stimulation of type II collagen synthesis by TGF-␤ and BMP2 and exhibits TAB1-independent kinase activity.
A, chondrocytes were infected 24 h after plating with dominant-negative TAK1 adenoviral vectors, Ad-pC-hTAK1a-K63A, and Ad-pC-hTAK1a-K63W. Seventy-two hours after initiating infection, cells were treated for 24 h with 5 ng/ml TGF-␤1 (T) or 100 ng/ml BMP2 (B) in the presence of [ 3 H]proline. Samples from triplicate wells were pooled before SDS-PAGE. B, for immune complex kinase assays, chondrocytes were infected with Ad-pC-hTAK1a (TK), kinase-negative (KN) Ad-pC-hTAK1a-KWSA, and the tandem vector (TK,TB) pCEA3-hTAB1(pC-hTAK1a). Forty-eight hours after beginning infection, cells were treated with TGF-␤1 (40 ng/ml; T) or IL-1␤ (40 ng/ml; IL) for 10 min prior to lysis, immunoprecipitation with TAK-ct antibody, and TAK1a immune complex kinase assay using bacterially expressed GST-MKK6 as substrate. Lysates were equally divided prior to immunoprecipitation to permit direct kinase assays in the presence of [␥-32 P]ATP (upper panels) or kinase assay after pretreatment with cold ATP (lower panels). The right panels are 7-fold shorter exposures of the last two lanes. C, Western blots (IB) demonstrating activation of endogenous TAK1 and TAB1 and their overexpressed counterparts. Chondrocytes were treated for 10 or 30 min with TGF-␤1 (5 ng/ml) or BMP2 (100 ng/ml) and lysed with SDS sample buffer at the same time as cells exposed to 48 h of adenoviral expression of TAK1a or coexpressed TAK1a and TAB1 (tandem construct). TAK-ct and TAB-m primary antibodies were used for detection. Long dashes, unmodified TAK1 or TAB1 bands; short dashes, activated/phosphorylated bands. Right panel, shorter exposures of lanes 5 and 6. nuclease protection analysis, remained unchanged compared with controls (Fig. 5). Adenoviral expression of TAK1a and TAK1a-⌬N for 72 h after the initiation of infection produced 8and 6-fold increases in type II collagen synthesis while causing a 5-fold reduction in Col2a1 mRNA. This response was doseresponsive with MOI, independent of simple viral infection or the presence of the pC promoter, and only slightly influenced by cotreatment with TGF-␤1. Kinase-negative TAK1a did not alter the basal levels of either type II collagen synthesis or its mRNA. When collagen synthesis stimulated by TAK1a was determined in the 24-h period following the above RNA isolations, type II collagen synthesis remained high (8-fold), although its mRNA at termination was further decreased (data not shown). Thus, increased mRNA levels were not responsible for increased synthesis of type II collagen by TGF-␤1 or TAK1a. In fact, in response to TAK1a, type II collagen synthesis increased, whereas its mRNA levels substantially declined, creating a 30 -80-fold greater ratio of expressed protein/mRNA (Fig. 5B).
RPA analysis of other important cartilage molecules, aggrecan and Sox9, demonstrated no increase in mRNA following treatment with TGF-␤ and a decline in mRNA under the influence of TAK1a that was similar to that in Col2a1 mRNA (Fig.  4A). These data support the coordinated regulation of all three molecules suggested by Murakami et al. in studies with IL-1␤ (64). In contrast, the mRNA level of CTGF, a TGF-␤-responsive growth factor, was increased by TGF-␤1 and TAK1a and further increased by the combination. This cooperative positive response was concurrent with the effects on collagen, aggrecan, and Sox9 and demonstrated that a general inhibition of transcription did not occur under the influence of TAK1a overexpression. The greater stimulation of CTGF in Fig. 4B by Western analysis is consistent with accumulation during transient CTGF transcription prior to isolation of RNA for Fig. 5.
TAK1a Overexpression Activates NF-B but TGF-␤ Does Not-Murakami et al. (64) have shown that decreased expression of ␣1(II), aggrecan, and Sox9 in response to IL-1␤ is due to activation of the transcription factor NF-B. Since TAK1 plays an important role in IL-1␤ signaling (41,42), we evaluated NF-B activation to determine whether it was responsible for the decline in cartilage-specific mRNA levels caused by TAK1a. TAK1a and TAK1a-⌬N increased activation of a cotransfected pNF-B-Luc reporter 5-fold, whereas TAK1a and TAB1 ex-pressed together from separate plasmids or from the tandem plasmid increased activation to 20-fold (Fig. 6). In contrast, TGF-␤1 partially suppressed basal and all TAK1a-dependent increases in activation of NF-B. Thus, exogenous TAK1a, but not TAK1 activated by TGF-␤1, activated NF-B and mimicked the suppressive effects of IL-␤ on cartilage-specific mRNAs. Importantly, both sources of TAK1 stimulate type II collagen synthesis at the protein level. DISCUSSION Regulation of type II collagen synthesis is important for normal skeletogenesis as well as maintenance of the structural and functional properties of growth plate and articular cartilage. Here we describe a TAK1-dependent mechanism that is responsible for enhanced production of type II collagen in adult, differentiated chondrocytes and does not require increased transcription of the Col2a1 gene. TAK1 mimics both TGF-␤ and BMP2 in this regard, and its effects are not dependent on Smad3 signaling. Enhanced production thus occurs in an environment of established type II collagen synthesis and should be distinguished from Col2a1 expression induced during development or activation of chondrogenic precursor cells. The latter are transcriptional processes mediated by Sox9 and relevant Smads. They are also likely to involve TAK1 signaling, because Smad transactivation utilizes DNA-dependent and proteinprotein interactions with other transcription factors (1,26,27) that can be activated by kinases downstream of TAK1.
We have shown that TAK1 is predominately responsible for stimulation of type II collagen synthesis in differentiated chondrocytes in response to TGF-␤ and BMP2 by (a) blocking their effects with dominant negative TAK1-K63W and TAK1-K63A, (b) demonstrating their activation of endogenous TAK1 and TAB1, and (c) mimicking their effects using adenoviral overexpression of wild type TAK1a. In the latter case, we have shown that overexpressed TAK1a can signal and yield a response on its own without simultaneous overexpression of its activating protein TAB1. Similarly, TAK1 can act alone at high concentrations to stimulate the PAI-1 promoter. This agrees with the report of Li et al. (65) that overexpressed TAK1 alone can activate MKK4 and JNK in HEK293 cells but contrasts with the original and common observation that overexpressed TAB1 is required for TAK1 activation and signaling (29,30,36). Such apparent TAB1 independence may result from TAK1 saturation of endogenous regulatory proteins and its subsequent selfassociation and autoactivation by intermolecular phosphorylation (33). Similarly, overexpression of TAK1 may lead to its exposure to kinases that usually have no access to it. Alternatively, ceramide (36) or Sef (similar expression to fibroblast growth factor genes) (66) can activate TAK1 and induce JNK activation; however, the involvement of phosphorylation or association with TAB1 as activating mechanisms in these situations remains unclear.
The nature of the TAK1 activated state appears different among overexpressed TAK1a alone, TAK1a overexpressed with TAB1, and TGF-␤-activated endogenous TAK1. This was most clearly demonstrated by the complete loss of kinase activity when TAK1a, overexpressed alone, was preincubated under kinase conditions with unlabeled ATP before kinase assay. Under the same conditions, TAK1a overexpressed with TAB1 retained its kinase activity (Fig. 4B). Thus, even TAK1a that initially exhibits autophosphorylation and transphosphorylation activity in vitro and biological activity in vivo (TAK1a alone) is not stable in vitro without interaction with TAB1. The lack of stability also suggests that in vivo activation of exogenous TAK1a alone is not mediated by interaction with endogenous TAB1, a conclusion consistent with the absence of endogenous TAB1 activation detected by Western blotting (Fig.   FIG. 6. TAK1a overexpression activates NF-B, and TGF-␤ suppresses this activation. The NF-B-responsive reporter, pNF-B-Luc, was cotransfected with promoterless pUC18 (NP), promoter-only plasmids (pC and pCEA3(pC)), or TAK1a plasmids, alone or in combination with TAB1. Both separate and tandem expression vectors were used. Luciferase assays were performed 48 h after transfection and TGF-␤1 addition (black bars). Error bars, S.E.; n ϭ 4. 4C, lane 5). TAK1a overexpressed alone also must be activated in vivo in such a way that autophosphorylation sites remain open for subsequent incorporation of label during the initial kinase assay. The presence of two functional sites (30,33) meets this requirement. In addition, an activated conformation or associated factor must be lost during immunoprecipitation or assay so that TAK1 is incapable of reactivating itself by autophosphorylation. Denaturation in the absence of TAB1 during the kinase assay or inactivation by an associated phosphatase could yield this result.
Two phosphatases, PP2C␤-1 (67) and PP2A⑀ (65), have been shown to dephosphorylate and inactivate active TAK1 or TAK1⅐TAB1 complexes both in vitro and in vivo. PP2A⑀ binds to both inactive and active TAK1 (65), and thus a ternary complex is expected both before and after ligand-dependent TAK1 activation. The results from the immune complex kinase assays (Fig. 4B) are most consistent with the absence of phosphatase or its inhibition in the immunoprecipitates derived from TAK1⅐TAB1-infected cells. Continuous phosphatase activity during pretreatment with unlabeled ATP should have reduced substrate phosphorylation in the subsequent kinase assay. In addition, phosphatase activity balanced by TAB1dependent reactivation of TAK1a should have allowed TAK1a autophosphorylation in the final kinase assay. Neither of these results was observed. In contrast, phosphatase activity could explain the loss of both auto-and transphosphorylation by TAK1a alone following incubation with cold ATP, because TAK1a reactivation in this TAB1-deficient environment is unlikely.
Activation of TAK1 overexpressed in the presence of TAB1 is mediated by intramolecular autophosphorylation of Ser 192 (30) and/or intermolecular autophosphorylation of Thr 187 (33) in the activation loop. Both sites participate in the capacity of TAK1 for transphosphorylation and activation of the NF-B pathway, and both lead to phosphorylation of associated TAB1 and the appearance of multiple phosphorylated forms of both TAK1 and TAB1 that exhibit retarded mobility on Western blots (30,33). In agreement with these data, we showed that stimulation of chondrocytes by TGF-␤ or BMP generates two retarded, presumably phosphorylated, forms of both endogenous TAK1 and TAB1 (Fig. 4C). However, when overexpressed with TAB1, TAK1a migrates almost entirely as the intermediate form, which is greatly reduced or absent when TAK1a is overexpressed alone. This pattern of abundance of the intermediate form matches that of TAK1a kinase activity (Fig. 4B) and suggests that this is a catalytically active form of TAK1a. However, the biological activity of TAK1a in stimulating type II collagen synthesis does not require the high level of kinase activity produced by tandem expression. TAK1a alone and TGF-␤1 produce the same level of stimulation with minimal presence of the intermediate form. Alternatively, the slowest migrating, most phosphorylated form of TAK1 may be responsible for its effects on collagen synthesis. TGF-␤1, TAK1a overexpressed alone, and tandem expression of TAK1a and TAB1 (Fig. 4C, lanes 4 -6) generate very similar amounts of this form, suggesting that its formation is controlled by limited quantities of other endogenous interaction partners. Importantly, it can be induced by overexpressed TAK1a without TGF-␤1 treatment and without the apparent involvement of TAB1, since no slowly migrating, activated forms of endogenous TAB1 were detected.
Conversely, TGF-␤ treatment phosphorylated all cellular TAB1 to its intermediate or slowest form at the same time that only a fraction of endogenous TAK1 was activated. It is uncertain whether this result is due to endogenous TAB1 content that is insufficient to bind all TAK1 or to the exist-ence of distinct TAK1⅐TAB1 complexes that permit TAB1 phosphorylation without prior or subsequent phosphorylation of TAK1. The latter possibility gains support from the detection of a kinase in TAK1 immune complexes that phosphorylates TAB1 without IL-␤ stimulation (Fig. 7 in Ref. 32). Regardless of the mechanism, complete phosphorylation of TAB1 by a single cytokine (i.e. TGF-␤1) suggests that selective signaling by TAK1 in response to this and other cytokines requires the function of additional components of TAK1 signaling complexes. Models of endogenous TAK1 activation/ signaling based on the IL-1␤ pathway (including tumor necrosis factor receptor-associated factor 6, TAB1, -2, and -3, and polyubiquitination) (31,32,68,69), the Wnt pathway (70,71), and the Sef pathway (66) are complex but likely to be conceptually similar to TAK1 activation in response to TGF-␤ and BMP2.
In addition to the role of TAB1 in activating and stabilizing TAK1 kinase activity, TAB1 appears to exhibit a targeting function. This was most clearly demonstrated by the fact that after stimulation at a low dose, increased levels of constitutively active TAK1a-⌬N did not produce increased expression of the PAI-1 luciferase reporter, whereas coexpression with TAB1 established a new higher level of expression shared by coexpression of wild type TAK1a and TAB1 (Fig. 1). Thus, more abundant active TAK1 could not substitute for the association of TAK1a with TAB1. This may be due to the potential of TAB1 to increase substrate binding and turnover or to a capacity to recruit and activate binding partners that subsequently facilitate downstream signaling, as seen with TAB2 and TAB3 in IL-1␤ signaling (68). Similarly, TAB1 enhanced TAK1a activation of NF-B signaling but failed to increase TAK1 stimulation of type II collagen synthesis, perhaps characterizing two different types of TAK1 signaling. Smad2, -3, and -4 are the principal transcriptional mediators of TGF-␤ signaling (25), whereas Smad1, -4, -5, and -8 mediate the effects of BMP (1). We have used overexpressed Smad3 to demonstrate the absence of Smad3 involvement in TGF-␤1and TAK1a-mediated stimulation of type II collagen synthesis (Fig. 3B). Overexpressed Smad3 was unable to stimulate synthesis on its own or enhance stimulation by TAK1a or TGF-␤1. In addition, no Smad2 or Smad3 signals are expected in the presence of effective TGF-␤1-neutralizing antibodies (Fig. 3A), a situation where stimulation by TAK1a remained. This is consistent with the lack of Smad3 binding to the proximal promoter of Col2a1 (49). Independence from receptor Smads is also supported by (a) the similar stimulation of type II collagen synthesis caused by both TGF-␤1 and BMP2, which signal through completely different sets of receptor Smads, (b) activation of TAK1a by both TGF-␤1 and BMP2 (28) (Fig. 4C), and (c) the parallel, nearly complete inhibition of collagen synthesis by dominant negative TAK1a (Fig. 4A). Despite this evidence, the involvement of receptor Smads other than Smad3 cannot be ruled out, since overexpressed Smad3 blocked collagen synthesis induced by both TAK1a and TGF-␤1 (Fig. 4B). The simplest interpretation of this result is that Smad3 competitively displaced another necessary Smad from its binding site in the L45 loop of T␤RI or altered the function of the Smad adaptor, SARA (72,73). Such displacement by Smad3 has been used with dominant negative Smad2/3 to effectively distinguish Smad3 and Smad2 effects (74).
Horton et al. (51) and Chadjichristos et al. (49) have demonstrated TGF-␤-dependent reduction of Col2a1 transcription mediated by the enhancer (51) or Sp1/Sp3 binding elements in the 63-bp proximal promoter (49). In contrast, the results presented here from a variety of protocols demonstrate substantial TGF-␤-induced increases in type II collagen synthesis without alteration of Col2a1 mRNA and are in agreement with the stimulation of type II collagen synthesis reported in rabbit chondrocytes following adenoviral expression of TGF-␤ or treatment with recombinant protein (47). These differences may result from changes in chondrocyte phenotype during culture (75). During this process, cell shape-and density-dependent signaling and responsiveness may change (49,60,62,76) with or without overt modulation of the collagen phenotype. The present studies also utilized only the endogenous promoter in its normal genomic and chromatin context.
TAK1a stimulation of type II collagen synthesis and its concurrent sharp reduction in Col2a1 mRNA are entirely consistent with the capacity of TAK1 to activate NF-B following TAK1 overexpression or IL-1␤/tumor necrosis factor-␣ treatment (41,42) and the capacity of NF-B to inhibit Col2a1 transcription by reducing Sox9 expression and its binding to the Col2a1 first intron enhancer (64). We have verified TAK1dependent activation of NF-B in this chondrocyte system and demonstrated that, although TGF-␤1 activates TAK1, TGF-␤1 inhibits NF-B activation due to overexpressed TAK1a. This may represent a mechanism for the antagonism of IL-1␤ by TGF-␤. The down-regulation of Col2a1 mRNA by TAK1a also demonstrates that TAK1 signaling is sufficient for this response and does not require support from upstream participants in the IL-1␤ pathway. Tan et al. (77) have reported a rapid, IL-1␤-dependent, EGR-1-mediated down-regulation of constitutive transcription from the Col2a1 proximal promoter. This response utilized both EGR-1 activation and transcription, did not alter Sox9 expression, and did not require the intron enhancer. Since Murakami et al. (64) reported no IL-1␤ effect using a similar 89-bp proximal promoter, these reports are difficult to reconcile. However, both the NF-B and EGR-1 effects could be mediated by TAK1, since (a) TAK1 is essential for IL-1␤-dependent NF-B signaling (42), (b) p38 MAPK has been implicated in IL-␤-dependent suppression of Col2a1 expression (78), (c) EGR-1 is activated by p38 (79), and (d) p38 is activated indirectly by TAK1. Despite this potential for shared mediation by TAK1, the present data most closely match the NF-B/Sox9 pathway, because TAK1 decreased both Sox9 and aggrecan mRNA in parallel with Col2a1 mRNA, and aggrecan and Col2a1 expression share Sox9 dependence during development (43).
Enhanced efficiency of the translational apparatus through increased initiation, elongation, or message recycling provides a rapid mechanism for regulating protein expression without changes in transcription (80). This is a possible explanation for the TGF-␤1-dependent increase in type II collagen synthesis that occurs without increased Col2a1 mRNA. That such increased synthesis occurs following TAK1 overexpression, in the face of decreased message levels, further reinforces the absence of a Col2a1 transcriptional mechanism. The increased efficiency of message utilization depicted in Fig. 6 also could be due to changes in the complex posttranslational processing and export of procollagen. The TAK1 effects reported here may result directly from phosphorylation of its substrates or those of its effector kinases or indirectly through transcriptional regulation of targets other then Col2a1 that influence translation or downstream processes. In either case, TAK1-dependent mechanisms may significantly contribute to the functions of TGF-␤ and BMP in chondrogenesis and maintenance of adult articular cartilage attributed to these factors by genetic studies using dominant negative T␤RII (19) and knock-out or conditional knock-out of essential expression of Bmpr1b (4) and Bmpr1a (17).