Transforming Growth Factor- (cid:1) (TGF- (cid:1) ) Type I Receptor/ ALK5-dependent Activation of the GADD45 (cid:1) Gene Mediates the Induction of Biglycan Expression by TGF- (cid:1) *

We have recently shown that induction of biglycan (BGN) expression by transforming growth factor- (cid:1) 1 (TGF- (cid:1) 1) required sequential activation of both Smad and p38 mitogen-activated protein kinase signaling (Un-gefroren, H., Lenschow, W., Chen, W.-B., and Kalthoff, H. (2003) J. Biol. Chem. 278, 11041–11049). Here, we have analyzed the receptors through which TGF- (cid:1) 1 controls expression of BGN and GADD45 (cid:1) , the latter of which is postulated to link early Smad signaling to delayed activation of p38. Ectopic expression of a dominant-negative mutant of the TGF- (cid:1) type II receptor in PANC-1 cells abrogated TGF- (cid:1) -induced BGN up-regulation. Similarly, inhibition of the TGF- (cid:1) type I receptor/ALK5 with either SB431542 or by enforced stable expression of a kinase-dead mutant greatly attenuated the TGF- (cid:1) effect on both BGN and GADD45 (cid:1) expression in PANC-1 and MG-63 cells. The enhancing effect of ALK5 on TGF- (cid:1) mediated GADD45 (cid:1) and BGN BGN in (pancreatic) epithelial versus mesenchymal cells in qualitative and/or quantitative differences in receptor-proximal signaling, or at the transcriptional/post-transcriptional level. TGF- (cid:1) signaling to BGN has revealed an unexpected degree of complexity. The TGF- (cid:1) effect on BGN is strictly dependent on activation of Smads (25) and p38 (27). The delayed kinetics of both p38 activation and BGN up-regulation and the sensitivity of the TGF- (cid:1) effect on BGN to the protein synthesis inhibitor cycloheximide (25) suggested an indirect effect involv-ing early Smad-mediated transcriptional activation of a gene X, which subsequently activates p38 and finally BGN. These find-ings further predict that signaling intermediates crucial for TGF- (cid:1) activation of p38 are likely to be important for TGF- (cid:1) regulation of BGN, too. As part of a larger project aimed at characterizing the “missing links” in the series of signaling events commencing with the activation of TGF- (cid:1) receptors and culminating in BGN mRNA accumulation, we initially focused on the role of ALK5 and its functional domains, respectively, responsible for transducing the TGF- (cid:1) signal to BGN. We present evidence that induction of both GADD45 (cid:1) and BGN expression as well as activation of p38 by TGF- (cid:1) requires the Smad-activating function of ALK5. We further show that intermittent activation of GADD45 (cid:1) expression is mandatory for TGF- (cid:1) / ALK5 regulation of p38 activation and BGN induction. The results of this study establish the ALK5-GADD45 (cid:1) -p38 pathway to be crucial for TGF- (cid:1) regulation of BGN and provide for the first time data on the regulation of GADD45 (cid:1) expression by the ALK5 kinase.


Ectopic expression of a dominant-negative mutant of the TGF-␤ type II receptor in PANC-1 cells abrogated TGF-␤-induced BGN up-regulation. Similarly, inhibition of the TGF-␤ type I receptor/ALK5 with either SB431542 or by enforced stable expression of a kinasedead mutant greatly attenuated the TGF-␤ effect on both BGN and GADD45␤ expression in PANC-1 and MG-63 cells. The enhancing effect of ALK5 on TGF-␤mediated GADD45␤ and BGN expression and on
GADD45␤ promoter activity was also dependent on its ability to activate Smad signaling, because an ALK5 mutant defective in Smad activation (T␤RImL45) but with an otherwise functional kinase domain failed to mediate these responses. The TGF-␤/ALK5 effect on p38 activation and BGN expression was mimicked by overexpression of GADD45␤ alone (in the absence of TGF-␤ stimulation) and suppressed upon antisense inhibition of GADD45␤ expression. These results show that TGF-␤ induces BGN expression through (the Smad-activating function of) ALK5 and GADD45␤ and suggest that the sensitivity of MyD118 to activation by TGF-␤, which varies between tissues, ultimately determines the strength of the TGF-␤ effect on BGN.
TGF-␤ 1 and its signaling effectors regulate basic cellular functions such as proliferation and apoptosis and act as key determinants of tumor cell behavior (1)(2)(3). Most, if not all, of the cellular activities of TGF-␤ are mediated by specific recep-tor complexes that are assembled upon ligand binding and comprise the TGF-␤ type II receptor (T␤RII) and a type I receptor (4). Whereas only one type II receptor is known so far, there are at least three type I receptors, activin receptor-like kinase (ALK)1/TSR-1 (5), ALK2/Tsk7L (6), and TGF-␤ type I receptor/ALK5 (7) that have been shown to bind TGF-␤ in the presence of T␤RII. ALK5 has an ubiquitous distribution and represents the principle type I receptor that mediates most cellular responses to TGF-␤. The activated ligand-receptor complex activates one or more downstream signaling pathways, the most prominent one being the Smad pathway (1)(2)(3)(4). However, other signaling pathways can be activated by the ALK5 kinase, including mitogen-activated protein kinases (MAPKs), ERK1/2 (8), and p38 MAPK (9). In the case of p38 signaling the activation by ALK5 may be Smad-independent (10) or Smad-dependent (4,11). Recently, transcriptional induction by TGF-␤ of GADD45␤ (encoded by MyD118), a protein involved in the response to genotoxic stress (12), has been shown to be Smad-dependent and to mediate the (delayed) activation of p38 by TGF-␤ (13,14) via binding and activation of the MAPKKK MTK1/MEKK4 (14,15). However, although GADD45␤ has been identified as a positive modulator of TGF-␤-induced apoptosis (Ref. 12 and references therein), its role in other major TGF-␤ responses, e.g. matrix gene expression is less clear.
Biglycan (BGN) belongs to the family of small leucine-rich proteoglycans and is functionally involved in matrix assembly, cellular migration, and the regulation of growth factor, e.g. TGF-␤ activity (reviewed in Refs. 16 and 17). BGN is a major constituent of the bone matrix where it is synthesized by osteoblasts, and studies in BGN-deficient mice indicated a crucial role for this proteoglycan in bone metabolism and mechanical properties (18,19). BGN is markedly up-regulated in fibrotic lesions of lung (20), of hepatic (21) and renal (22) tissues, in corneal scars (23), and in the stroma of solid tumors, e.g. pancreatic carcinoma (24). These compartments, like the bone matrix, are highly enriched in TGF-␤. This correlation is not accidental, because BGN is exceptionally sensitive to induction by TGF-␤ and as such can serve as a marker of fibrotic tissue deposition in the stroma (23). However, BGN-producing cells differ in their BGN response to TGF-␤ stimulation, e.g. osteoblastic MG-63 cells increase BGN expression only moderately by 2-to 3-fold, whereas pancreatic PANC-1 cells respond to TGF-␤ under the same conditions with a 10-fold stronger induction (24,25). Previous studies from our group indicated that in both pancreatic and bone cells the TGF-␤-induced rise in cytoplasmic BGN mRNA involves an as yet undefined nuclear post-transcriptional mechanism rather than an increase in transcriptional activity of the BGN gene promoter (25,26). However, at present it is unknown if the greater inducibility of BGN by TGF-␤ in (pancreatic) epithelial versus mesenchymal cells resides in qualitative and/or quantitative differences in receptor-proximal signaling, or at the transcriptional/posttranscriptional level.
TGF-␤ signaling to BGN has revealed an unexpected degree of complexity. The TGF-␤ effect on BGN is strictly dependent on activation of Smads (25) and p38 (27). The delayed kinetics of both p38 activation and BGN up-regulation and the sensitivity of the TGF-␤ effect on BGN to the protein synthesis inhibitor cycloheximide (25) suggested an indirect effect involving early Smad-mediated transcriptional activation of a gene X, which subsequently activates p38 and finally BGN. These findings further predict that signaling intermediates crucial for TGF-␤ activation of p38 are likely to be important for TGF-␤ regulation of BGN, too. As part of a larger project aimed at characterizing the "missing links" in the series of signaling events commencing with the activation of TGF-␤ receptors and culminating in BGN mRNA accumulation, we initially focused on the role of ALK5 and its functional domains, respectively, responsible for transducing the TGF-␤ signal to BGN. We present evidence that induction of both GADD45␤ and BGN expression as well as activation of p38 by TGF-␤ requires the Smadactivating function of ALK5. We further show that intermittent activation of GADD45␤ expression is mandatory for TGF-␤/ ALK5 regulation of p38 activation and BGN induction. The results of this study establish the ALK5-GADD45␤-p38 pathway to be crucial for TGF-␤ regulation of BGN and provide for the first time data on the regulation of GADD45␤ expression by the ALK5 kinase.
Cell Lines and Cell Culture-Human pancreatic cancer PANC-1 and osteosarcoma MG-63 cells were maintained as described earlier (26,27). Colonic carcinoma SW48 cells were a kind gift of M. Löhr (University of Mannheim) and were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and 2 mM L-glutamine (Invitrogen). Cells stably transduced with various retroviral vectors were cultured in the presence of 700 (PANC-1) or 250 (MG-63) g/ml Geneticin (Invitrogen).
RNA Isolation and RT-PCR Analysis-Total RNA from PANC-1 and MG-63 cells was isolated with peqGOLD RNAPure (Peqlab, Erlangen, Germany). The general RT-PCR protocol, the conditions for competitive quantitative PCR, the PCR primer sequences for BGN and plasminogen activator-inhibitor (PAI-1), as well as the procedure of data analysis were described in detail earlier (25). Sense-strand-specific amplification and quantification of GADD45␤ expression was achieved using primers GADD45␤-forward (AACATGACGCTGGAAGAGCTCG) and GADD-45␤-3Ј-UTR-reverse (ACAGATTCTGCTGCTGGGAAGG), together with a standard fragment lacking nucleotides 120 -187 of the human GADD45␤ cDNA (GenBank TM accession: NM_015675). Amplification of ␤-actin-specific transcripts was carried out with ␤-actin-forward (GA-CGAGGCCCAGAGCAAGAG) and ␤-actin-reverse (ATCTCCTTCTGC-ATCCTGTC) primers. All values for GADD45␤, BGN, and PAI-1 mRNA concentrations were normalized to those for ␤-actin in the same sample to account for small differences in cDNA input.

Construction of Vectors and Retroviral Infection-
The entire open reading frame of human GADD45␤ was amplified by RT-PCR with Turbo Pfu polymerase (Invitrogen) and primers GADD45␤-forward (see above) and GADD45␤-reverse (CTCAGCGTTCCTGAAGAGAGATG, stop codon underlined) and was subcloned in both sense and antisense orientation into the retroviral vector TJBA5bMoLink-neo (TJ-neo) (25). A cDNA for rat FLAG-T␤RImL45(T202D) (10) was excised from pRK5F and inserted into TJ-neo and pcDNA3-HASL. cDNA inserts of hemagglutinin (HA)-tagged versions of dnT␤RII (D404G), caALK5 (T204D), and kdALK5 (K232R), were excised from pcDNA1, pCMV5, or pcDNA3-HASL, respectively, and inserted into TJ-neo. Positive clones (evaluated by PCR, restriction analysis, and sequencing of the plasmid-cDNA junctions) were co-transfected into HEK293T producer cells along with retroviral packaging vectors as described previously (25). Retroviral particles released by 293T cells were used to infect PANC-1 and MG-63 cells. Pools and individual clones of productively infected cells were obtained after selection with Geneticin.
Transient Transfections and Reporter Gene Assays-For transient transfections followed by immunoprecipitation (IP), cells were seeded at a density of 2 ϫ 10 4 cells/cm 2 on day 1 in 10-cm plates, and on day 2 they were co-transfected serum-free with Lipofectamine Plus (Invitrogen) according to the manufacturer's instructions with either FLAG-p38 or FLAG-Smad2 in combination with empty pcDNA3 vector, caALK5, or caT␤RImL45 as indicated in the legend to Fig. 3. Following removal of the transfection solution and a recovery period of 24 h in normal growth medium, cells were stimulated with TGF-␤1 for 1 h. The transfected cells were then lysed in IP buffer (20 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM ␤-glycerolphosphate, 1 mM Na 3 VO 4 , 1 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride) and processed for anti-FLAG and anti-HA immunoprecipitation and immunoblotting (see below). For reporter gene assays, cells were seeded in 96-well plates and were cotransfected on the next day serum-free with Lipofectamine Plus (PANC-1) or Lipofectamine 2000 (Invitrogen) (MG-63) according to the manufacturer's instructions with various cDNAs (in pcDNA3) at an equal molar ratio together with p6SBE-Luc (28), or pG45␤-1 (29), and the Renilla luciferase-encoding vector pRL-TK (Promega). Each well received the same total amount of DNA, and empty vector was added as needed. Following transfection and TGF-␤1 stimulation, luciferase activities were determined with the Dual Luciferase Assay System (Promega). In all reporter gene assays the data were derived from 6 -8 wells processed in parallel and corrected for transfection efficiency with Renilla luciferase activity.
Immunoblot Analysis and Immunoprecipitation-Epitope-tagged proteins were immunoprecipitated from cellular lysates using protein G Plus-Sepharose (Santa Cruz Biotechnology) according to the protocol provided by the supplier and subsequently analyzed by immunoblotting as described in detail earlier (25).

T␤RII Function Is Required for TGF-␤ Induction of BGN
Expression-To assess the role of T␤RII for TGF-␤-induced BGN expression, we measured TGF-␤-stimulated BGN expression in SW48, a colon carcinoma cell line harboring a loss-offunction mutation in T␤RII as a result of a mismatch repair deficiency with microsatellite instability (30). These cells failed to up-regulate BGN mRNA after a 24-h treatment with 5 ng/ml TGF-␤1 (Fig. 1A). Prolonged treatment with or higher concentrations of this growth factor were also ineffective in inducing BGN expression (data not shown). Under the same conditions, TGF-␤-responsive PANC-1 cells responded with a 20-fold induction of BGN mRNA over basal levels (Fig. 1A). To further probe the role of T␤RII in TGF-␤ regulation of BGN, we retrovirally transduced PANC-1 cells in a stable fashion with a T␤RII mutant (T␤RII-D404G). This mutant has been shown to act in a dominant negative fashion by inhibiting the appearance of wild type receptor at the cell surface (31). A clone pool (heterogenous mixture of G418-resistant cells with respect to integration site and expression level of the transgene) displayed strong expression of the mutant as revealed by immunoblotting (Fig. 1B). Notably, TGF-␤-induced expression of BGN and of PAI-1, another TGF-␤ response gene dependent on Smad, but independent of p38 activation (25), was completely suppressed (Fig. 1, C and D). These data indicate that T␤RII is required for BGN induction by TGF-␤ in gastrointestinal epithelial cells.
The ALK5 Kinase Activity Is Required for p38 Activation and BGN Induction-Next, we tested whether ALK5 is involved in TGF-␤ induction of BGN and treated PANC-1 cells with SB431542, a novel imidazole compound with selectivity for the TGF-␤/ALK5-mediated Regulation of Biglycan via GADD45␤ ALK5 group of TGF-␤/BMP type I receptors (ALK5, ALK4, and ALK7) (32) (Fig. 2A). This inhibitor effectively suppressed the BGN response to TGF-␤ in both PANC-1 ( Fig. 2A) and MG-63 cells (data not shown). To more specifically inhibit ALK5, we expressed in PANC-1 and MG-63 cells by stable retroviral transduction a kinase-deficient ALK5 mutant (kdALK5) known to act in a dominant negative fashion. The mutant protein was strongly expressed in the PANC-1 transductants (Fig. 2B, upper panel) and was functional, because TGF-␤induced phosphorylation of Smad2 was markedly attenuated in kdALK5-expressing cells but not in the corresponding vectortransduced control cells (Fig. 2B). Interestingly, ectopic expression of kdALK5 also blocked phosphorylation, and thus activation, of p38 MAPK (Fig. 2B). In agreement with the results from the pharmacological inhibition, BGN mRNA was markedly reduced in both kdALK5-expressing PANC-1 (Fig. 2C) and MG-63 cells (Fig. 2D). The data clearly indicate that the ALK5 kinase activity is crucial for TGF-␤ regulation of BGN in both pancreatic and bone cells.
p38 MAPK Activation and BGN Up-regulation by TGF-␤ Requires the Smad-activating Function of ALK5-Previous data indicated that an early activation of the Smad pathway was required for and preceded the activation of p38 and BGN expression by TGF-␤ (14,25,27), suggesting that, besides the kinase activity, the Smad-activating function of ALK5 is also important. To confirm this directly on the receptor level, we employed an ALK5-derived mutant (T␤RImL45) that has re-tained a functional kinase domain but due to additional mutations in the L45 loop was unable to activate Smads (10). To avoid activation of the endogenous receptors we used the constitutively active (ca) counterparts ALK5-TD (33) and T␤RImL45(TD) (10). As expected, co-transfection of PANC-1 cells with HA-tagged caALK5 or HA-tagged caT␤RImL45 along with FLAG-tagged Smad2 or FLAG-tagged p38 followed by anti-FLAG immunoprecipitation and anti-Smad or anti-p38 immunoblotting, respectively, revealed that caALK5, but not caT␤RImL45, phosphorylated Smad2 and p38 (Fig. 3A). Next, we verified the inability of caT␤RImL45 to activate the Smad pathway in a transcriptional reporter assay using the Smadspecific reporter plasmid p6SBE-Luc. Transcriptional induction of p6SBE-Luc by this mutant was weak in comparison to that of caALK5 (Fig. 3B). Finally, BGN mRNA levels in (unstimulated) MG-63 cells stably expressing caALK5 approached those in TGF-␤-treated control cells, whereas they remained below basal levels in their caT␤RImL45-expressing counterparts (Fig. 3C), despite approximately equal expression of the mutant proteins (not shown). Very similar results with respect to BGN expression were obtained with PANC-1 cells ectopically expressing caALK5 or caT␤RImL45 (data not shown; these cells were also used for analysis of TGF-␤/ALK5 regulation of GADD45␤ expression, see Fig. 4C). These data strongly support the assumption that the Smad-activating function of ALK5 accounts for most, if not all, of p38 activity, which is consistent with the delayed type of p38 activation seen in both

TGF-␤/ALK5-mediated Regulation of Biglycan via GADD45␤
cell types following ligand stimulation (14,27) (Fig. 3D). Given the strict dependence of TGF-␤ regulation of BGN on p38 activation, we hypothesized that the comparatively low induction of BGN expression by TGF-␤ in MG-63 cells (2.5-fold versus 20-fold in PANC-1) is the result of a lower activation of p38. To this end, TGF-␤ induced p38 phosphorylation in MG-63 with similar (delayed) kinetics, but the extent was much weaker (Fig. 3D) than in PANC-1 cells (see vector control in Fig. 2B). This observation further suggests that signaling events upstream of p38 activation ultimately determine the magnitude of the BGN response to TGF-␤.
TGF-␤ Induces GADD45␤ Expression via ALK5-The Smadmediated induction of GADD45␤ expression has been shown to trigger activation of the p38 pathway by TGF-␤ in PANC-1 and other cell types (14). Consistent with this, TGF-␤ induced a robust increase of GADD45␤ mRNA in PANC-1 cells within 1 h of stimulation (Fig. 4A). Induction of GADD45␤ by TGF-␤ was also observed in MG-63 cells, albeit to a lesser extent and corresponded with a lower overall increase (compared with PANC-1) in both p38 activation (see Fig. 3D) and BGN expression (see Fig. 3C). Next, we tested the assumption that GADD45␤ induction by TGF-␤ was mediated by ALK5. As expected, PANC-1 cells treated with SB431542 failed to activate GADD45␤ in response to TGF-␤ (Fig. 4B). We then evaluated the possibility that the inability of caT␤RImL45 to activate p38 and to induce BGN mRNA (see Fig. 3) was due to its failure to induce transcription from MyD118. To this end, GADD45␤ mRNA levels were elevated strongly in PANC-1 cells stably expressing caALK5, but not in PANC-1 cells stably expressing caT␤RImL45 (Fig. 4C). To test in a more direct fashion whether ALK5 can transcriptionally activate MyD118, we monitored caALK5-mediated activation of G45␤-1, a transcriptional reporter containing ϳ1500 bp of 5Ј-flanking region from the human GADD45␤ gene in front of the luciferase gene (29), in MG-63 cells. Notably, caALK5, but not caT␤RImL45, induced G45␤-1 186.7 Ϯ 15.6% over vector controls (Fig. 4D). It should be mentioned that under similar conditions PANC-1 cells were unresponsive to this promoter construct (see "Discussion"). Nevertheless, the data suggest that TGF-␤ transcriptionally activates MyD118 via the Smad activating function of ALK5 in both pancreatic and bone cells.
GADD45␤ Activates p38 MAPK and Enhances BGN Expression-Having demonstrated that GADD45␤ expression is induced by TGF-␤ via ALK5, we next investigated whether ectopic overexpression of GADD45␤ could activate p38 and upregulate BGN in the absence of TGF-␤ stimulation. To test this prediction, PANC-1 cells were stably transduced with a GADD45␤-encoding retrovirus or empty retroviral vector used as control. In GADD45␤-expressing cells (clones #7 and #8, detected by anti-GADD45␤ antibody, Fig. 5A) p38 was activated in the absence of TGF-␤ treatment (Fig. 5B). The enhanced levels of phosphorylated p38 reflected increased p38 kinase activity as determined with an in vitro kinase assay using ATF-2 as substrate (data not shown). To explore whether the ability of GADD45␤ to activate p38 is sufficient to enhance BGN expression, we measured BGN mRNA levels in the

TGF-␤/ALK5-mediated Regulation of Biglycan via GADD45␤
GADD45␤ transductants in the absence of TGF-␤ stimulation. Strikingly, BGN expression was dramatically increased in both clones (Fig. 5C), while expression of PAI-1 remained unaffected (Fig. 5D), indicating the specificity of the effect for BGN. Another consequence of ectopic expression of GADD45␤ and subsequent p38 activation in some cell types may be induction of apoptosis (13). However, we did not observe a higher rate of apoptotic cells in the PANC-1 cultures overexpressing GADD45␤. These results show that ectopic overexpression of GADD45␤ can specifically mimic the TGF-␤ effect on BGN.
Antisense-mediated Inhibition of GADD45␤ Blocks TGF-␤ Induction of Both p38 Activation and BGN Expression-To address this question of whether endogenous GADD45␤ induced by TGF-␤ is responsible for p38 activation and BGN up-regulation, PANC-1 cells were stably infected with antisense GADD45␤ cDNA thereby blocking the expression of the endogenous MyD118. Several individual clones were assayed by quantitative RT-PCR before and after TGF-␤ stimulation.
Antisense GADD45␤ effectively suppressed endogenous GADD45␤ expression albeit to a varying degree. For further analysis we chose two clones, one with moderate and one with high suppression (Fig. 6A). We then measured the activation status of endogenous p38 in these cells using anti-phospho-p38 immunoblotting. Notably, antisense-mediated inhibition of GADD45␤ induction correlated with a marked and dose-dependent reduction in the amount of phosphorylated p38 in the TGF-␤-treated cells (Fig. 6B) indicating a requirement for GADD45␤ expression in TGF-␤-induced p38 activation. To test whether inhibition of endogenous GADD45␤ also affected the BGN response to TGF-␤, we measured BGN expression in these clones by means of RT-PCR. As shown in Fig. 6C, PANC-1 cells stably expressing antisense GADD45␤ exhibited a strong and dose-dependent decrease in their BGN mRNA response to TGF-␤, which correlated well with the phosphorylation status of p38 in these clones (Fig. 6B). Again, PAI-1 expression in the same samples was essentially unaffected by

FIG. 3. The Smad activating function of ALK5 is required for TGF-␤-induced activation of p38 and p38-dependent gene expression.
A, the ability of ALK5 to phosphorylate Smad2 and p38 depends on its Smad-activating function. PANC-1 cells were transiently transfected with either 10 g of pFLAG-Smad2 or pFLAG-p38 together with empty vector, or HA-tagged versions of caALK5, or caT␤RImL45. After 24 h empty vector-transfected cells were stimulated, or not, with TGF-␤ for 2 h followed by lysis and immunoprecipitation (IP) of FLAG-tagged and HA-tagged proteins. Immunoprecipitates were subjected to SDS-PAGE and immunoblotting (IB) with antibodies recognizing phosphorylated forms of Smad2 (p-Smad2) and p38 (p-p38), Smad2 (t-Smad2) and p38 (t-p38), or ALK5. B, caT␤RImL45 is unable to activate a Smad-responsive reporter. PANC-1 cells were transiently transfected with empty vector, caALK5, or caT␤RImL45 along with p6SBE-Luc and pRL-TK as described under "Experimental Procedures" and treated, or not, with TGF-␤. Reporter gene activity was measured 24 h after transfection. The data shown are from one representative experiment out of three experiments performed in total each with very similar results. Data represent the mean Ϯ S.D. of six wells processed in parallel and are expressed relative to the value in untreated vector-transfected cells set arbitrarily at 1. C, caT␤RImL45 is unable to mimic the TGF-␤ effect on BGN in MG-63 cells. Wild type MG-63 cells and polyclonal cultures of MG-63 cells stably expressing empty retroviral vector, caALK5, or caT␤RImL45 were treated with TGF-␤ for 24 h and subjected to quantitative RT-PCR for BGN. D, TGF-␤ weakly activates p38 in MG-63 cells. Confluent cultures of MG-63 cells were serum-starved for 16 h and stimulated with TGF-␤1 for the indicated times or for 1 h with anisomycin as control (Co). Cellular lysates were assayed by immunoblotting for phospho-p38 and total p38.

TGF-␤/ALK5-mediated Regulation of Biglycan via GADD45␤
the antisense GADD45␤ expression (Fig. 6D). These data clearly indicate that TGF-␤/ALK5-dependent expression of endogenous GADD45␤ is necessary and specific for TGF-␤-mediated induction of BGN. DISCUSSION In this study we have analyzed the role of ALK5 in TGF-␤ regulation of GADD45␤ and BGN in PANC-1 and MG-63 cells. Using a combination of kinase-deficient and kinase-active mutants as well as specific pharmacological inhibitors, we present evidence that both the kinase activity and the Smad-activating function of ALK5 are crucial for up-regulation of both GADD45␤ and BGN expression. We go on to show that GADD45␤ acts as a signal transducer in the signaling pathway leading from activated ALK5 to BGN via intermittent activation of p38 MAPK. This conclusion was drawn from the observation that ectopic expression of GADD45␤ in PANC-1 cells mimicked the TGF-␤ effect on p38 and BGN, whereas antisense-mediated suppression of TGF-␤-induced GADD45␤ expression dose-dependently inhibited p38 activation and BGN up-regulation by this growth factor. Fig. 7 integrates GADD45␤ into the signaling pathway involved in TGF-␤ regulation of BGN expression.
We have previously reported that TGF-␤ regulation of BGN required both functional Smad4 expression (25) and activation of p38 (27) and have shown that p38 activation occurred downstream of Smad activation (27). We sought to confirm the Smad dependence of p38 and BGN directly at the receptor level by investigating, if the ability of activated ALK5 to induce p38 and BGN was due to direct Smad activation by the ALK5 kinase. For this purpose we expressed an ALK5-derived mutant with an intact kinase domain but deficient in its ability to activate R-Smads by phosphorylation (10). This mutant was unable to activate p38 and p38-dependent gene expression, e.g. BGN, in transiently transfected and stably transduced PANC-1 and MG-63 cells, respectively. The data obtained with this mutant together with the Smad4 reconstitution experiments reported earlier (27) clearly indicate that the Smad-activating function of ALK5 is required for p38 activation and BGN induction and that p38 activation is Smad-dependent in these cells rather than being Smad-independent as described in NMuMG cells FIG. 4. TGF-␤ induces GADD45␤ expression via ALK5. A, TGF-␤ induces GADD45␤ expression. PANC-1 and MG-63 cells were incubated for various times as indicated with or without TGF-␤1 and subjected to semiquantitative RT-PCR analysis of GADD45␤ expression. B, TGF-␤-induced GADD45␤ expression is sensitive to SB431542 inhibition. PANC-1 cells were stimulated for 24 h with TGF-␤ in the absence or presence of the indicated concentrations of SB431542. At the end of the incubation period cells were lysed and processed for RNA isolation and quantitative RT-PCR for GADD45␤. C, the TGF-␤ effect on GADD45␤ expression is mimicked by caALK5, but not caT␤RImL45. PANC-1 cells stably expressing empty retroviral vector (vector), caALK5 (clone #58), or caT␤RImL45 (clone #27) were subjected to quantitative RT-PCR for GADD45␤. Inset, both ALK5 mutants are expressed at approximately equal intensity as verified by semiquantitative RT-PCR. D, caALK5, but not caT␤RImL45 activates the GADD45␤ gene promoter in the absence of TGF-␤ stimulation. MG-63 cells were transiently transfected with G45␤-1, a construct containing the 5Ј-flanking region of the human GADD45␤ gene fused to the luciferase reporter, together with empty expression vector (pcDNA3), caALK5, or caT␤RImL45. 46 h after transfection cells were processed for measurement of luciferase activities using the Dual Luciferase Assay System. Data represent the normalized mean Ϯ S.D. of six wells processed in parallel. One representative experiment is shown out of a series of three experiments performed in total.
Earlier studies using cycloheximide indicated that de novo protein synthesis is required for TGF-␤ regulation of BGN in PANC-1 cells (25). Given the crucial role of intermittent p38 activation for BGN induction and the recent discovery that TGF-␤/ALK5-mediated activation of p38 is accomplished through Smad-regulated transcriptional activation of MyD118 (13,14), we hypothesized that suppressing TGF-␤-stimulated endogenous GADD45␤ expression would result in inhibition of both p38 activation and BGN induction. To this end, antisensemediated suppression of GADD45␤ expression in PANC-1 cells prevented p38 activation and blocked the TGF-␤ effect on BGN in a dose-dependent fashion. Having shown that ALK5 mediates the TGF-␤ effect on p38 and BGN, we analyzed whether ALK5 was also able to activate MyD118. Notably, the GADD45␤ gene remained silent in TGF-␤-treated PANC-1 cells upon SB431542-mediated inhibition of the ALK5 kinase activity and in PANC-1 cells stably expressing caT␤RImL45, but was activated in caALK5-expressing cells. The differential responsiveness of the GADD45␤ gene was confirmed by promoter-reporter gene assays in MG-63 cells, which showed that kinase-active ALK5, but not kinase-active T␤RImL45 directly activated the upstream GADD45␤ promoter. Under similar conditions, PANC-1 cells were unresponsive to this construct (data not shown), the reason of which is not clear but may reflect tissue-specific differences in promoter usage (see below). In a parallel approach, we observed that stable overexpression of GADD45␤ in PANC-1 cells was able to rescue activation of p38 and BGN induction. GADD45␤ has also been implicated in TGF-␤-induced apoptosis (10,13), and Yoo et al. (13) demonstrated that ectopic expression of GADD45␤ is sufficient to activate p38 and to trigger apoptosis.
In addition to elucidating the molecular pathway acting upstream of p38 and BGN, we have revealed histogenetic differences in the signaling pathways involved in TGF-␤ regulation of BGN. In PANC-1 cells strong activation of MyD118 transcription by TGF-␤ correlated with greater induction of p38 activation and a dramatic rise of BGN expression within 24 h, whereas in MG-63 cells a comparatively small increase in GADD45␤ mRNA corresponded with a lower level of p38 activation and only a moderate up-regulation of BGN, further lending support to the ALK5-GADD45␤-p38-BGN connection. It thus appears that the powerful induction of BGN mRNA by TGF-␤ in PANC-1 cells is the consequence of a stronger activation of MyD118. Hence, the crucial question is what determines the sensitivity of the GADD45␤ gene to induction by TGF-␤, an issue that may be resolved by defining the characteristics of MyD118 transcriptional regulation. Major and Jones (29) have found that the GADD45␤ 5Ј-promoter sequence (which is contained in the G45␤-1 construct) responds to TGF-␤ with an ϳ2-fold increase in transcriptional activity in HaCaT and Mv1Lu cells. In addition, these authors identified a second TGF-␤-responsive module encompassing the highly conserved third intron of the GADD45␤ gene, which conferred 3-fold greater transcriptional induction by TGF-␤ than the 5Ј-promoter sequence (29). In MG-63 cells, caALK5-induced activity of the 5Ј-promoter sequence and TGF-␤-stimulated GADD45␤ FIG . 5. GADD45␤ expression activates p38 and induces BGN expression. A, detection of ectopic GADD45␤ in individual clones (S#7 and S#8) of PANC-1 cells stably infected with a GADD45␤ encoding retrovirus by immunoblot analysis. Basic endogenous levels of GADD45␤ in unstimulated empty vector-transduced controls (left lane) and wild type cells (data not shown) were hardly detectable, but were estimated by quantitative RT-PCR to be 10-times lower than in vector-transduced cells treated for 1.5 h with TGF-␤ (compare Fig. 6B). B, ectopic expression of GADD45␤ activates p38 in the absence of TGF-␤ treatment in PANC-1 cells. As control, vector-transduced cells were stimulated with TGF-␤ for 1.5 h. The phosphorylated p-38 (p-p38) and total p38 (t-p38) are shown in the upper and lower panels, respectively. C and D, ectopic expression of GADD45␤ induces BGN, but not PAI-1 mRNA. Unstimulated PANC-1 cells stably transduced with either an empty retrovirus (vector) or a GADD45␤ encoding retrovirus (clones #7 and #8) were subjected to quantitative RT-PCR for BGN and PAI-1.

TGF-␤/ALK5-mediated Regulation of Biglycan via GADD45␤
FIG. 6. Antisense-mediated inhibition of GADD45␤ blocks TGF-␤-mediated p38 activation and BGN expression. A, expression of GADD45␤ antisense RNA suppresses TGF-␤-induced GADD45␤ mRNA accumulation. Individual clones of PANC-1 cells stably transduced with an empty retroviral vector (vector) or an GADD45␤ antisense encoding retrovirus were screened for GADD45␤ mRNA expression. Following a 1-h incubation of the cells with or without TGF-␤, total RNA was prepared and subjected to both semiquantitative (left panel) and quantitative (right panel) RT-PCR using primers specific for the sense GADD45␤ transcript. B, expression of GADD45␤ antisense RNA suppresses TGF-␤-induced activation of p38. Phosphorylation of endogenous p38 in total cell extracts prepared from the same cells in A, untreated or treated with TGF-␤ for 1.5 h, was analyzed by immunoblotting. The phosphorylated p-38 (p-p38) and total p38 (t-p38) are shown in the upper and lower panels, respectively. C and D, expression of GADD45␤ antisense RNA suppresses TGF-␤ induction of BGN, but not PAI-1 expression. PANC-1 cells stably transduced with either an empty retrovirus (vector) or a GADD45␤ encoding retrovirus (clones #7 and #8) were treated with or without TGF-␤ for 24 h and subjected to quantitative RT-PCR for BGN and PAI-1.

FIG. 7. Current model of the signal flow involved in TGF-␤-induction of BGN expression.
This scheme summarizes available data on the signaling events involved in conveying the TGF-␤ signal from the receptor complex to the nucleus, ultimately resulting in the accumulation of BGN mRNA. Upon activation of T␤RII and ALK5, the Smad-activating function of ALK5 residing in the L45 loop (L45) phosphorylates an R-Smad, which subsequently forms a complex with Smad4, and this complex translocates to the nucleus to induce expression of GADD45␤. GADD45␤ then activates p38 MAPK via activation of a MAPKKK, possibly MTK1/MEKK4. Active p38 is shuttled to the nucleus and induces BGN mRNA accumulation through nuclear mRNA transcript processing, stability, and/or export.

TGF-␤/ALK5-mediated Regulation of Biglycan via GADD45␤
mRNA levels were both increased ϳ2-fold, suggesting that bone cells utilize this 5Ј-promoter. In contrast, PANC-1 cells are more likely to utilize the intronic enhancer, since endogenous GADD45␤ transcript levels were induced 10-fold by TGF-␤. Experiments are in progress to clarify if the greater induction of GADD45␤ in PANC-1 cells is indeed caused by activation of this intronic promoter. Interestingly, transcription from this enhancer was strongly dependent on Smad4, which is in close agreement with the observation that re-expression of Smad4 in Smad4-deficient pancreatic cancer cell lines restored not only TGF-␤-stimulated GADD45␤ expression (14) but also TGF-␤-mediated p38 activation (27) and BGN expression (25).
Our observation that the enforced expression of GADD45␤ alone can stimulate the BGN response opens the real possibility that BGN is up-regulated independently from TGF-␤ in physiological situations in which GADD45␤ is induced, e.g. during terminal differentiation (12), environmental stress-induced apoptosis (12), acute hypertonicity (34), or during an inflammatory response to invading microorganisms (35,36). Of clinical relevance for the treatment of pancreatic carcinoma is the observation that GADD45 family proteins are rapidly induced by genotoxic agents (12,37). Their use in the course of an anticancer therapy may thus promote BGN synthesis through GADD45␤-p38 interactions and thereby enhance the fibrogenic response of the tumor tissue. Given the antiproliferative effect of BGN on tumor cells (24), GADD45␤-p38 signaling may thus play an important role in tumor inhibition. Taken together, the results presented in this study establish a central role for ALK5-induced GADD45␤ in TGF-␤ regulation of BGN.