Transcription Activation of FLRG and Follistatin by Activin A, Through Smad Proteins, Participates in a Negative Feedback Loop to Modulate Activin a Function

Transcription activation of FLRG and follistatin by activin A, through Smad proteins, participates in a negative feedback loop to modulate activin A function

Laurent Bartholin1,3,VeÂ ronique Maguer-Satta1,3, Sandrine Hayette1,2, Sylvie Martel1, MyleÁ ne Gadoux2, Laura Corbo1, Jean-Pierre Magaud1,2 and Ruth Rimokh*,1

1INSERM U453, Centre LeÂon BeÂrard, 69373 Lyon, France; 2Laboratoire de CytogeÂneÂtique MoleÂculaire, HoÃpital Edouard Herriot, 69437 Lyon, France

Signaling of TGFb family members such as activin is location (Hayette et al., 1998). FLRG is a secreted tightly regulated by soluble binding proteins. Follistatin glycoprotein which is highly homologous to follistatin, binds to activin A with high anity, and prevents activin a protein known to bind to activin, a member of the binding to its own receptors, thereby blocking its transforming growth factor b (TGFb) superfamily signaling. We previously identi®ed FLRG gene from a (Nakamura et al., 1990). Activin and follistatin B-cell leukemia carrying a t(11;19)(q13;p13) transloca- proteins were originally isolated from ovarian ¯uid as tion. We and others have already shown that FLRG, a result of their ability to stimulate and inhibit, which is highly homologous to follistatin, may be respectively, the secretion of follicle-stimulating hor- involved in the regulation of the activin function through mone from the pituitary gland (Phillips and de Kretser, its binding to activin. In this study, we found that, like 1998). Further studies have revealed that activin, a follistatin, FLRG protein inhibited activin A signaling as secreted protein produced by a variety of tissues, has demonstrated by the use of a transcriptional reporter important endocrine and paracrine roles. Activin has assay, and blocked the activin A-induced growth been shown to control fundamental processes such as inhibition of HepG2 cells. We have recently shown that cell proliferation and dierentiation, development and the TGFb-induced expression of FLRG occurs at a reproduction (DePaolo, 1997; Phillips and de Kretser, transcriptional level through the action of Smad proteins. 1998; Ying et al., 1997). Several reports have shown Here we show that activin A increases FLRG and that activin, which is produced by bone marrow follistatin at both the mRNA and protein levels. We stromal cells (Yamashita et al., 1992) such as FLRG found that Smad proteins are involved in the activin A- (Hayette et al., 1998), is also involved in hematopoietic induced transcription activation of FLRG and follistatin. dierentiation, particularly in the erythroid lineage (Yu Finally we demonstrate that FLRG protein regulates its and Dolter, 1997). We have previously shown that the own activin-induced expression. In conclusion, activin A expression of both FLRG and activin A is up-regulated induces FLRG and follistatin expression. This observa- during hematopoietic dierentiation, and in TGFb- tion, in conjunction with the antagonistic eect of FLRG treated human bone marrow stromal cells (Maguer- and follistatin on activin signaling, indicates that these Satta et al., 2001). Several studies have revealed that two proteins participate in a negative feedback loop the biological eects of activin can be modulated by which regulates the activin function. the capability of follistatin to bind to and inactivate Oncogene (2002) 21, 2227 ± 2235. DOI: 10.1038/sj/ activin. Follistatin interacts with activin A with high onc/1205294 anity, and prevents activin binding to its own receptors, thereby blocking activin A-induced signaling Keywords: FLRG; follistatin; activin; TGFb,Smad (Phillips, 2000). Recently we and others have shown that FLRG may be involved in the regulation of the biological activity of activin through its binding to activin (Maguer-Satta et al., 2001; Tsuchida et al., Introduction 2000). The TGFb family includes TGFbs, activins and bone We previously identi®ed a human gene, FLRG morphogenetic proteins (BMP). Like that of other (Follistatin-related gene), from a B-cell chronic lym- TGFb family members, activin signaling operates phocytic leukemia carrying a t(11;19)(q13;p13) trans- through a heteromeric complex of transmembrane serine/threonine kinase receptors: activin binds to a type II receptor, which recruits, phosphorylates and thereby activates an activin type I receptor propagating *Correspondence: R Rimokh; E-mail: [email protected] signals along the Smad pathway. Activin and TGFb 3The ®rst two authors contributed equally to this work Received 17 September 2001; revised 2 January 2002; accepted 8 act through Smad2 and Smad3, whereas BMP acts January 2002 through Smad1 and Smad5. Phosphorylated Smad2 FLRG regulates activin function L Bartholin et al 2228 and Smad3 form heteromeric complex with the common mediator Smad4, translocate into the nucleus, and activate the transcription of target genes in a cell- type-speci®c manner through interaction with DNA and other nuclear factors (Derynck et al., 1998; Massague and Wotton, 2000; ten Dijke et al., 2000; Wrana, 2000). Smad6 and Smad7 has been reported to function as negative regulators of TGFb family. Smad7 has been shown to inhibit TGFb and activin signal transduction by blocking type I receptor phosphoryla- tion of Smad2 and Smad3 (Hayashi et al., 1997; Lebrun et al., 1999; Nakao et al., 1997). Smad6, the other negative regulator of the TGFb family (Imamura et al., 1997), seems to preferentially inhibit BMP signaling. In the present study, we analysed the regulation of activin A signaling by FLRG and follistatin, two activin-binding proteins. We have recently shown that TGFb induces transcriptional activation of FLRG in the human hepatoma cell line HepG2 through Smad proteins (Bartholin et al., 2001). We examined the expression of FLRG and follistatin in HepG2 cells following treatment with activin A, a member of TGFb superfamily. We found that follistatin and FLRG were transcriptionally induced by activin A through Smad proteins, and that FLRG, like follistatin, regulated the biological functions of activin A.

Results Figure 1 Inhibitory eect of FLRG and follistatin proteins on activin A-induced transcription activation. HepG2 cells were cotransfected with the (CAGA)9-MLP-Luc reporter vector (a and Neutralization of activin A-induced transcription b), along with the pRL-CMV internal control vector. Twenty-four activation by FLRG and follistatin proteins hours after transfection, the cells were stimulated by activin A (20 ng/ml), or left unstimulated, in the presence or absence of We have shown that FLRG may be involved in the increasing amounts of FLRG (a) or follistatin (b) proteins, and regulation of the biological activity of activin the luciferase activity was determined 24 h later. This activity through its binding to activin (Maguer-Satta et al., (normalized by reference to the renilla luciferase activity) is given as the mean+s.d. of an experiment performed in triplicate, 2001). In order to analyse the antagonistic eect of representative of three experiments. Fold-induction calculated by the FLRG protein on activin A bioactivity, we ®rst comparing activin-treated cells to untreated control cells is produced a secreted human recombinant FLRG indicated at the top of each black bar protein, using a baculovirus expression system. We carried out a transcriptional reporter assay using an activin-responsive reporter construct in order to analyse the eect of FLRG protein on activin Effect of FLRG protein on HepG2 cell growth signaling. HepG2 cells were transiently transfected with the (CAGA)9-MLP-Luc reporter vector, which Activin A inhibits the growth of primary hepatocytes is known to respond to activin A-induced transcrip- and HepG2 cells in a dose dependent-way, and tion (Dennler et al., 1998). Twenty-four hours after follistatin, an activin-binding protein, has been shown transfection, the cells were treated with activin A in to inhibit activin activity in the liver (Phillips and de the presence or absence of FLRG or follistatin Kretser, 1998). Here we used a ¯uometric growth assay proteins. As shown in Figure 1a, b, activin A to evaluate the eect of exogenous FLRG protein on increased the luciferase activity by more than 30- exponentially growing HepG2 cells. Cells seeded in the fold. The addition of FLRG or follistatin alone did presence of activin A did not exhibit any net increase not aect the background level of the luciferase in cell numbers over a 5-day period, as compared to activity (Figure 1a,b, open bars). The addition of the untreated control culture (Figure 2b). Thus, as FLRG or follistatin proteins dramatically inhibited expected, activin A inhibits the proliferation of HepG2 the activin A-induced transcription activation in a cells, and the eect of activin A on HepG2 prolifera- dose-dependent way (Figure 1a,b, black bars). As tion was abolished by adding follistatin protein (data previously reported (Tsuchida et al., 2000), we not shown). FLRG protein alone had no eect on cell found that FLRG protein, like follistatin, can proliferation (Figure 2a), while in combination with inhibit the activin-induced transcription activation. activin, it neutralized the inhibitory eect of activin A

Oncogene FLRG regulates activin function L Bartholin et al 2229

Figure 2 Antagonistic eect of FLRG protein on activin A-induced growth inhibition of HepG2 cells. (a) HepG2 cells were incubated for 5 days with or without various amounts of FLRG protein, as indicated. (b) HepG2 cells were incubated for 5 days with activin A (Act.A) at 10 ng/ml, and with or without various amounts of FLRG protein, as indicated. At each time point an Alamar Blue assay was performed to quantitatively measure the proliferation of the cells. Data points represent the mean values mean+s.d. of triplicates. The experiment shown is representative of two separate experiments performed in triplicate

on proliferation in a dose-dependent way (Figure 2b). protein level, we performed a Western blot with a These results indicate that the exogenous addition of ®vefold concentrated conditioned medium from cells FLRG protein blocks activin A-induced growth treated or not with activin A for 48 h. Using antibodies inhibition. directed against FLRG and follistatin proteins, we observed immunoreactive bands of the expected molecular size (33 and 31 ± 39 kDa), the known M of Induction of FLRG and follistatin expression by activin A r glycosylated FLRG and follistatin proteins respectively We have previously reported that TGFb is a potent (Figure 3b). These bands were not found in a medium inducer of FLRG gene expression in human bone obtained from untreated control cells. These results marrow stromal cells and the human hepatoma cell line indicate that activin A is a potent inducer of the HepG2, as shown at both the protein and mRNA expression of the activin ligands, FLRG and follistatin, levels (Bartholin et al., 2001; Maguer-Satta et al., in the human hepatoma cell line HepG2. 2001). In this study, we investigated whether activin A, a member of the TGFb superfamily, modulated FLRG Transcriptional regulation of FLRG and follistatin by expression in the HepG2 cell line, which is known to activin A contain functional activin receptors. HepG2 cells were treated for dierent periods of time with activin A, and To ®nd out if the increase in FLRG and follistatin total cellular RNA was analysed by Northern blot expression in the presence of activin A resulted from using FLRG or follistatin cDNA probes. As previously the transcription activation of the FLRG and follistatin shown (Russell et al., 1999; Zhang et al., 1997), activin promoters, we used a luciferase-based construct A increased follistatin mRNA levels. Treatment of containing these promoters. The pGL3-MLP(73293/ HepG2 cells with activin A induced a high level of +6) and pGL3-basic(72165/+2) plasmids containing, time-dependent FLRG mRNA expression. This large respectively, the FLRG and follistatin promoters were increase in FLRG and follistatin transcripts was independently transfected into HepG2 cells and tested observed as early as 3 h after the addition of activin for inducibility by activin A. These two constructs led A, and it continued to rise over time after the addition to a threefold activin A-mediated induction (Figure 4), of cytokine (Figure 3a). which indicated a transcription activation of the FLRG To ®nd out if the up-regulating eect of activin A on and follistatin promoters in response to the addition of FLRG and follistatin mRNA was also re¯ected at the activin A. These results indicate that activin A up-

Oncogene FLRG regulates activin function L Bartholin et al 2230

Figure 4 Transcriptional activation of FLRG and follistatin by activin A. HepG2 cells were transiently cotransfected with the wild type pGL3 luciferase reporter vector (wt) or with the FLRG or follistatin promoter-luciferase constructs, along with the pRL- CMV internal control vector. Twenty-four hours before harvesting, the transfected cells were either stimulated by activin A, or left unstimulated. The relative luciferase activity was normalized with regard to the renilla luciferase activity expressed by the pRL- CMV vector. Fold-induction (mean+s.d. of three experiments performed in triplicate) was calculated by comparing activin- treated cells to untreated control cells

Figure 3 Activin A increases expression of FLRG and follistatin. (a) HepG2 cells were treated for 3, 6, or 15 h with activin A As shown in Figure 5a, the overexpression of Smad3 (20 ng/ml). Ten mg of total RNA from control cells and activin- induced follistatin transcription. Moreover it syner- treated cells were subjected to Northern blot analysis for the gized with Smad4 to stimulate the follistatin promoter, indicated times. The ®lter was sequentially assayed for hybridiza- as shown by the fact that the co-expression of Smad3 tion to FLRG and follistatin cDNA a-32P-labeled probes. Equal amounts of RNA were checked by ethidium bromide (EtBr) and Smad4 increased the promoter activity by more staining. (b) Western blot analysis of conditioned media than 40-fold. concentrated ®vefold from HepG2 cells, either treated or not To analyse the eect of Smad proteins expression on with activin A (20 ng/ml) for 24 or 48 h. Proteins separated by the follistatin mRNA level, we transfected expression SDS-12% polyacrylamide gel electrophoresis were immuno- vectors for Smad2, Smad3 and Smad4, independently detected using anti-FLRG or anti-follistatin antibodies or in combination, into HepG2 cells, and analysed total cellular RNA by Northern blot using follistatin cDNA as a probe. In accordance with the results obtained using the follistatin promoter fused to regulate FLRG and follistatin expression at a tran- luciferase (Figure 5a), a cooperative eect of Smad3 scriptional level in HepG2 cells. and Smad4 proteins was observed in follistatin mRNA expression (Figure 5b). Next we investigated whether Smad proteins are Participation of Smad proteins in the activin A-induced involved in the activin-mediated activation of FLRG expression of FLRG and follistatin and follistatin promoters. For this purpose, we As for other target genes transcriptionally regulated by cotransfected HepG2 cells with FLRG and follistatin TGFb or activin (Moustakas and Kardassis, 1998; promoter-luciferase constructs along with expression Nagarajan et al., 1999; Vindevoghel et al., 1998), we vectors encoding the inhibitory Smad7, or C-terminally have reported that although Smad3 alone was found to truncated forms of Smad3 and Smad4, which are increase FLRG mRNA, the highest transcript level was known to act as speci®c dominant-negative inhibitors observed when the Smad2, Smad3 and Smad4 of Smad signaling (Lagna et al., 1996; Zhang et al., expression vectors were cotransfected (Bartholin et 1996). We found that transfection of Smad7, or Smad3 al., 2001). In the present case, we examined the and Smad4 dominant-negative mutants signi®cantly involvement of Smad proteins in follistatin transcrip- reduced the activin A-induced transcription of the tional activation. For this purpose, we cotransfected FLRG and follistatin promoter reporter constructs the expression vectors for Smad2, Smad3 and Smad4, (Table 1). In the presence of Smad7 we observed an independently or in combination, into HepG2 cells, inhibition of 77.5 and 45.5% of the activin A-induced along with the follistatin promoter-luciferase construct. FLRG and follistatin transcription, respectively. This

Oncogene FLRG regulates activin function L Bartholin et al 2231

Figure 5 Activation of the follistatin promoter by Smad proteins. (a) HepG2 cells were cotransfected with the follistatin promoter- luciferase construct together with the pRL-CMV internal control vector and expression vectors for Smad2, Smad3 and Smad4, either independently or in combination. The luciferase activity, measured 48 h after the transfection, and normalized with renilla luciferase activity, is given as the mean+s.d. of an experiment performed in triplicate. (b) Northern blot analysis with a speci®c follistatin a-32P-labeled probe using 10 mg of total RNA from HepG2 cells cotransfected with the indicated Smad expression vectors. Equal amounts of RNA were checked by ethidium bromide (EtBr) staining

Table 1 Involvement of Smad proteins in activin A-induced FLRG and follistatin transcription FLRG protein down-regulates its own activin-induced Cotransfected vectors % inhibition of expression Promoter Expression activin A-induced Finally, we investigated whether the inhibitory eect of constructs vectors transcription activation FLRG protein on the activin A-induced transcription FLRG Smad7 77.5+5.0 activation was also obtained with the FLRG promoter DN-Smad3 40.0+1.3 reporter construct. For this purpose, HepG2 cells were DN-Smad4 38.0+6.4 transiently transfected with the FLRG promoter Follistatin Smad7 45.5+0.9 reporter vector, which is known to respond to activin DN-Smad3 42.8+11.6 A-induced transcription (Figure 4). Twenty-four h after DN-Smad4 54.6+5.7 transfection, the cells were treated with activin A in the presence or absence of FLRG protein. The addition of HepG2 cells were cotransfected with the FLRG or follistatin promoter-luciferase constructs, along with the pRL-CMV internal FLRG alone did not aect the background level of the control vector and expression vectors for the Smad7 inhibitory luciferase activity (Figure 6, open bars). The addition protein or dominant-negative forms of Smad3 (DN-Smad3) or of FLRG protein inhibited the activin A-induced Smad4 (DN-Smad4) proteins, as indicated. Twenty-four hours before transcription activation in a dose-dependent way harvesting, the transfected cells were either stimulated by activin A or (Figure 6, black bars). This is shown by the fact that left unstimulated. The luciferase activity was normalized with regard to the renilla luciferase activity. Fold-induction was calculated by the 2.6-fold increase in transcriptional response to comparing activin-treated cells to untreated control cells, and data activin A which resulted from using the reporter are expressed as % of inhibition of activin A-induced transcription construct containing the FLRG promoter was signi®- activation. Data are indicated as mean+s.d. of two independent cantly reduced by the addition of FLRG protein experiments, each carried out in triplicate (Figure 6). These results indicate that the exogenous addition of FLRG protein reduced its own promoter activation when mediated by activin A, and that FLRG protein is involved in a negative feedback loop might suggest that the Activin signaling pathways to modulate activin signaling. involved in the transcription regulation of FLRG and follistatin are not identical. Altogether, our results indicate that Smad proteins Discussion are involved in FLRG and follistatin regulation by activin A. FLRG and follistatin genes can thus be TGFb is the prototype of a family of signaling considered as targets of activin transcriptional activa- molecules which also includes activins and BMPs. tion through the action of Smad proteins. Signaling of TGFb family members such as activin is

Oncogene FLRG regulates activin function L Bartholin et al 2232 characterized activin-binding protein, inhibits the eects of activin in HepG2 cells. The neutralizing eect of follistatin on activin bioactivity results from interference with activin binding to the type II receptor (de Winter et al., 1996). Further studies will be needed to determine whether FLRG, like follistatin, forms an inactive complex with activin, and thus neutralizes its activity. Activin and follistatin are expressed in many types of tissue, and participate in a number of physiological and developmental processes such as oogenesis, spermato- genesis, hematopoiesis and embryogenesis, and in the pituitary, liver and other tissues (Phillips and de Kretser, 1998). Similarly, FLRG is expressed in Figure 6 FLRG protein inhibits activin A-induced transcription numerous tissues, and during murine embryogenesis activation on its own promoter. HepG2 cells were cotransfected (Hayette et al., 1998; Tsuchida et al., 2000), suggesting with the FLRG promoter-luciferase construct, along with the that it plays a role in various physiological systems. pRL-CMV internal control vector. Twenty-four hours after The high homology between FLRG and follistatin may transfection, the cells were stimulated by activin A (20 ng/ml), or left unstimulated, in the presence or absence of increasing indicate functional redundancy. However, as we amounts of FLRG proteins, and the luciferase activity was recently demonstrated, FLRG and follistatin expression determined 24 h later. This activity (normalized by reference to patterns are dierent in primary hematopoietic sub- the renilla luciferase activity) is given as the mean+s.d. of an populations, which suggests that their expression may experiment performed in triplicate, representative of three be restricted in time and space, and that their activity experiments. Fold-induction calculated by comparing activin- treated cells to untreated control cells is indicated at the top of may not always be redundant. each black bar We have recently shown that TGFb is a potent inducer of FLRG gene expression through Smad proteins (Bartholin et al., 2001). As activin signaling is also mediated by the Smad proteins, we have analysed the tightly regulated by soluble binding proteins. Follista- regulation of FLRG and follistatin expression, and we tin binds to activin with high anity, preventing observed that activin A increased FLRG and follistatin receptor activation by ligand binding, thus blocking mRNA levels in HepG2 cells. The stimulatory eect of the biological activity of activin. Moreover, follistatin activin A was found to be time-dependent and associated interacts with BMP proteins, and modulates activity of with an increase in FLRG and follistatin proteins. To BMP (Fainsod et al., 1997; Gamer et al., 1999; Iemura investigate the molecular mechanisms underlying the et al., 1998). We and others have already shown that activin-induced expression of FLRG and follistatin, we FLRG protein, which is highly homologous to looked at the promoter activity of the human FLRG and follistatin, associates with activin A, and may partici- follistatin genes. We found that activin A induced the pate in the regulation of its biological activity (Hayette expression of FLRG and follistatin by a transcriptional et al., 1998; Maguer-Satta et al., 2001; Tsuchida et al., mechanism. We have previously demonstrated that 2000). FLRG, like follistatin, binds to BMP2, and Smad proteins, which are critical components of the inhibits BMP-induced transcriptional activation (Tsu- TGFb- and activin-signaling pathways, are involved in chida et al., 2000). the activation of the FLRG promoter by TGFb In the present study, we analysed the regulation of (Bartholin et al., 2001). Here we show that over- activin A bioactivity by FLRG and follistatin, two expression of the Smad3 and Smad4 proteins increases activin-binding proteins. Follistatin inhibits activin the follistatin mRNA level. In addition, overexpression functions in various cell types (Phillips, 2000). To of either the inhibitory Smad7 or the dominant-negative assess the ability of FLRG protein to inhibit activin A Smad3 or Smad4 mutants inhibits the activin-induced signaling, we have used a transcriptional assay with a activation of the FLRG and follistatin promoters. These reporter plasmid known to respond to activin A results indicate that Smad proteins are involved in the (Dennler et al., 1998). A dose-dependent inhibitory activation of FLRG and follistatin promoters by activin eect on the transcription activation induced by activin A. To our knowledge, this is the ®rst demonstration of a A was obtained by the addition of FLRG or follistatin response by the FLRG and follistatin promoters to proteins. It has been reported that in hepatocytes activin A. activin-induced cell growth inhibition is antagonized by Several studies have shown that signaling of the the exogenous addition of follistatin (Russell et al., TGFb superfamily is regulated at various dierent 1999; Schwall et al., 1993; Yasuda et al., 1993). Here, levels. The transcription of Smad7, a negative regulator as expected, we found that activin A inhibited the of TGFb and activin signal transduction, is regulated growth of the human hepatoma cell line HepG2. We by TGFb and activin signaling through Smad proteins found that soluble FLRG, like follistatin, blocked (Brodin et al., 2000; Denissova et al., 2000; Nagarajan activin A-induced growth inhibition in a dose-depen- et al., 1999). These data indicate that Smad7 is dent way. These results indicate that FLRG, a newly- involved in a feedback loop that modulates the eects

Oncogene FLRG regulates activin function L Bartholin et al 2233 of TGFb and activin A (Piek et al., 1999). Other manufacturer's instructions (Pharmingen). FLRG cDNA signaling pathways, such as interferon-g and tumor encoding the human FLRG protein, residues 27 to 263, necrosis factor-a, which operate, respectively, through was subcloned into the pAcGP67 baculovirus transfer vector JAK1/STAT1 and the nuclear factor-kB/RelA, stimu- fused to the sequence coding for the gp67 signal peptide, late the expression of Smad7 to block TGFb signaling under the control of the strong baculovirus polyhedrin promoter. The concentration of the secreted recombinant (Bitzer et al., 2000; Ulloa et al., 1999). Thus Smad7 FLRG protein, which was produced in a serum-free medium, modulates the cellular response to TGFb and activin was estimated on a polyacrylamide gel by silver staining using by using dierent signaling pathways. Smad6, the other de®ned amounts of recombinant 6His-FLRG protein pro- negative regulator of the TGFb family (Imamura et al., duced in a bacterial system. 1997), seems to preferentially inhibit BMP signaling. BMP2 and BMP7 increase the Smad6 mRNA level Plasmid constructs (Takase et al., 1998), indicating that the inhibitory Smad6 protein may play a role in a feedback loop To study the transcriptional regulation of human follistatin which regulates BMP signaling. The production of the gene expression, we subcloned the follistatin promoter inhibitory Smad6 and Smad7 proteins by TGFb family between position 72165 and +2 relative to the major transcription start site (Shimasaki et al., 1988) into the pGL3- members provides an autoinhibitory mechanism in basic luciferase reporter gene vector (lacking promoter and TGFb signaling. An additional level of control of the enhancer sequences; Promega), giving the pGL3-ba- biological activity of the TGFb family is provided by sic(72165/+2). For this purpose, human genomic DNA soluble binding proteins. Small leucine-rich proteogly- extracted from normal peripheral blood lymphocytes was cans such as decorin and biglycan bind to TGFb and used as a template for PCR reaction, with primers selected inhibit its activity (Piek et al., 1999). Noggin, chordin, from genomic sequences which contain the follistatin gene, gremlin, follistatin and FLRG can bind BMP proteins and which are available in the public domain (accession No. and antagonize their bioactivity (Reddi, 2001; Smith, AC008901). To study the transcriptional regulation of human 1999; Tsuchida et al., 2000). Our results and previous FLRG gene expression, we used the pGL3-MLP(73293/+6) reports (Maguer-Satta et al., 2001; Phillips and de plasmid (Bartholin et al., 2001) containing the FLRG promoter elements between nucleotides 72306 and +6, Kretser, 1998; Tsuchida et al., 2000) indicate that which promote the expression of the luciferase reporter gene. follistatin and FLRG reduce the biological eects of All the constructs were sequence-checked. activin A, and may play a role in regulating the Mammalian expression vectors encoding ¯ag-tagged hu- bioavailability of activin. man Smad2, Smad3, Smad4 and Smad7 were kindly provided In conclusion, activin A induces FLRG and by P ten Dijke (Ludwig Institute for Cancer Research, follistatin transcription through the action of Smad Uppsala, Sweden). Mammalian expression vectors encoding proteins. This observation, in conjunction with the the dominant-negative human Smad3 (DN-Smad3) and inhibitory eect of FLRG and follistatin on activin Smad4 (DN-Smad4) proteins were kindly provided by function indicates that these two activin-binding R Derynck (University of California, CA, USA). The proteins participate in a negative feedback loop that pGL3(CAGA)9-MLP-Luc reporter vector with nine copies of the CAGA sequence is known to respond to TGFb- and regulates activin signaling. Moreover, we found that activin A-induced transcription (Dennler et al., 1998), was FLRG protein regulates its own activin-induced kindly provided by JM Gauthier (Glaxo Wellcome labora- expression. The induction of FLRG and follistatin by tory, Les Ullis, France). activin A may protect cells from excessive exposure to activin. Similarly, it has been shown that BMP proteins Transient tranfections and reporter assays induce the expression of gremlin and noggin, two secreted proteins which inhibit BMP activity (Gazzerro For transient transfections, HepG2 cells were seeded in 12- 5 et al., 1998; Pereira et al., 2000). In all, these well plates (1.4610 cells per well) and transfected 48 h later observations suggest that soluble proteins like inhibi- by Exgen 500 (Euromedex, Souelweyersheim, France) with tory Smad proteins play a central role in the negative the indicated reporter constructs (800 ng) and the pRL-CMV internal control vector (25 ng) under the control of the autoregulation of TGFb family signaling. cytomegalovirus promoter (Promega). When Smad expression vectors were cotransfected, the same amount of Smad expression vector was transfected in each case (150 ng for Smad2, Smad3 and Smad4, and 350 ng for DN-Smad3, DN- Materials and methods Smad4 and Smad7), and total DNA was kept constant (1.25 mg) by the addition of an empty vector. Twenty ng/ml Cell cultures of human recombinant activin A (R&D) was added 24 h The human hepatoma cell line HepG2 was obtained by after transfection, and the luciferase activity was quanti®ed in American-Type Culture Collection (ATCC), and cultured in the cell lysates after a further 24 h using the Dual Luciferase Dulbecco's modi®ed Eagle's medium containing 10% fetal Assay (Promega). All the experiments were performed in calf serum (FCS), 0.03% L-glutamine, 100 mg/ml penicillin triplicate, and the luciferase activity was normalized with and 100 mg/ml streptomycin sulfate. regard to the renilla luciferase activity expressed by the pRL- CMV vector (Promega). For Northern blot analysis, HepG2 cells were seeded in FLRG production six-well plates (46105 cells per well) and transfected 24 h Human recombinant FLRG protein was produced in Sf9 later by Exgen 500 with expression vectors for Smad2, cells using a baculovirus expression system according to the Smad3 and Smad4, either independently or in combina-

Oncogene FLRG regulates activin function L Bartholin et al 2234 tion. The same amount of Smad expression vector visualize ribosomal RNA as a way of con®rming equal (650 ng) was transfected in each case, and total DNA RNA loading. was kept constant (2 mg) by the addition of an empty vector. Total RNA from the transfected cells was Western blot analysis extracted 48 h later. A ®vefold concentrated conditioned medium obtained from a low-concentration serum culture (1% FCS) derived from Alamar blue proliferation assay HepG2 cells, either treated or not treated with activin A To quantitatively measure the proliferation of HepG2 cells, (20 ng/ml), was subjected to electrophoresis on a 12% SDS- we used the Alamar Blue assay (BioSource International), polyacrylamide gel in denaturing and reducing conditions. which incorporates a ¯uometric/colorimetric growth indi- The separated proteins were transferred onto a membrane cator based on the detection of metabolic activity by (Millipore) by electroblotting. After blocking with a 5% non- monitoring the reducing environment of the proliferating fat milk solution, the ®lters were incubated with the rabbit cells. HepG2 cells were seeded in 96-well microtiter plates at polyclonal antisera raised against FLRG (Hayette et al., 3000 cells per well in 200 ml of DMEM containing 10% 1998), or with an anti-follistatin monoclonal antibody FCS, then incubated for 5 days with 10 ng/ml of activin A (R&D). Bound primary antibodies were detected with (R&D), follistatin (R&D), and FLRG, either independently horseradish peroxydase-labeled anti-rabbit or anti-mouse or in combination. After various incubation times at 378C, secondary antibodies. Finally, the membranes were washed Alamar Blue was added, and the reduction was monitored and developed with chemiluminescence reagents (ECL by a ¯uorometer 5 h after the addition of the Alamar Blue system, Amersham Pharmacia Biotech). to the culture. Fluorescence measurements were made by exciting at 530 ± 560 nm and measuring emission at 590 nm. After background subtraction, data are expressed in ¯uorescence emission-intensity units as a function of Acknowledgments incubation time. This study was supported by grants from INSERM, the Association pour la Recherche contre le Cancer, the Ligue contre le Cancer (ComiteÂ sduRhoÃneetdelaSaoÃneet Northern blot analysis Loire). L Bartholin holds a doctoral fellowship from the Total cellular RNA was puri®ed by the acid guanidium LiguecontreleCancer,ComiteÂ de la Haute Savoie. We thiocyanate-phenol-chloroform method using the TRIzol would like to thank JM Gauthier for providing the reagent as per the manufacturer's instructions (Gibco ± (CAGA)9-MLP-Luc reporter vector, P ten Dijke for the BRL). For Northern blot analysis, 10 mg of total RNA were mammalian expression vectors encoding the ¯ag-tagged size-fractionated in formaldehyde-1.2% agarose gels and human Smad2, Smad3, Smad4 and Smad7, and R Derynck transferred onto nylon ®lters. The probes were labeled with for the mammalian expression vectors encoding human a-32P-dCTP using the Rediprime labeling kit (Amersham dominant-negative Smad3 and Smad4. We would also like Pharmacia Biotech) and hybridized according to standard tothankMJN'GuyenandCLamadonfortheirtechnical procedures. The gel was stained with ethidium bromide to assistance.

References

Bartholin L, Maguer-Satta V, Hayette S, Martel S, Gadoux Gazzerro E, Gangji V and Canalis E. (1998). J. Clin. Invest., M, Bertrand S, Corbo L, Lamadon C, Morera A, Magaud 102, 2106 ± 2114. J and Rimokh, R. (2001). Oncogene, 20, 5409 ± 5419. Hayashi H, Abdollah S, Qiu Y, Cai J, Xu YY, Grinnell BW, Bitzer M, von Gersdor G, Liang D, Dominguez-Rosales A, Richardson MA, Topper JN, Gimbrone Jr MA, Wrana JL Beg AA, Rojkind M and Bottinger EP. (2000). Genes Dev., and Falb D. (1997). Cell, 89, 1165 ± 1173. 14, 187 ± 197. Hayette S, Gadoux M, Martel S, Bertrand S, Tigaud I, Brodin G, Ahgren A, ten Dijke P, Heldin CH and Heuchel R. Magaud JP and Rimokh R. (1998). Oncogene, 16, 2949 ± (2000). J. Biol. Chem., 275, 29023 ± 29030. 2954. de Winter JP, ten Dijke P, de Vries CJ, van Achterberg TA, Iemura S, Yamamoto TS, Takagi C, Uchiyama H, Natsume Sugino H, de Waele P, Huylebroeck D, Verschueren K T, Shimasaki S, Sugino H and Ueno N. (1998). Proc. Natl. and van den Eijnden-van Raaij AJ. (1996). Mol. Cell. Acad.Sci.USA,95, 9337 ± 9342. Endocrinol., 116, 105 ± 114. Imamura T, Takase M, Nishihara A, Oeda E, Hanai J, Denissova NG, Pouponnot C, Long J, He D and Liu F. Kawabata M and Miyazono K. (1997). Nature, 389, 622 ± (2000). Proc. Natl. Acad. Sci. USA, 97, 6397 ± 6402. 626. Dennler S, Itoh S, Vivien D, ten Dijke P, Huet S and Lagna G, Hata A, Hemmati-Brivanlou A and Massague J. Gauthier JM. (1998). EMBO J., 17, 3091 ± 3100. (1996). Nature, 383, 832 ± 836. DePaolo LV. (1997). Proc.Soc.Exp.Biol.Med.,214, 328 ± Lebrun JJ, Takabe K, Chen Y and Vale W. (1999). Mol. 339. Endocrinol., 13, 15 ± 23. Derynck R, Zhang Y and Feng XH. (1998). Cell, 95, 737 ± Maguer-Satta V, Bartholin L, Jeanpierre S, Gadoux M, 740. Bertrand S, Martel S, Magaud JP and Rimokh R. (2001). Fainsod A, Deissler K, Yelin R, Marom K, Epstein M, Exp. Hematol., 29, 301 ± 308. Pillemer G, Steinbeisser H and Blum M. (1997). Mech. Massague J and Wotton D. (2000). EMBO J., 19, 1745 ± Dev., 63, 39 ± 50. 1754. Gamer LW, Wolfman NM, Celeste AJ, Hattersley G, Moustakas A and Kardassis D. (1998). Proc. Natl. Acad. Sci. Hewick R and Rosen V. (1999). Dev. Biol., 208, 222 ± 232. USA, 95, 6733 ± 6738.

Oncogene FLRG regulates activin function L Bartholin et al 2235 Nagarajan RP, Zhang J, Li W and Chen Y. (1999). J. Biol. ten Dijke P, Miyazono K and Heldin CH. (2000). Trends Chem., 274, 33412 ± 33418. Biochem. Sci., 25, 64 ± 70. Nakamura T, Takio K, Eto Y, Shibai H, Titani K and Tsuchida K, Arai KY, Kuramoto Y, Yamakawa N, Sugino H. (1990). Science, 247, 836 ± 838. Hasegawa Y and Sugino H. (2000). J. Biol. Chem., 275, Nakao A, Afrakhte M, Moren A, Nakayama T, Christian 40788 ± 40796. JL, Heuchel R, Itoh S, Kawabata M, Heldin NE, Heldin Ulloa L, Doody J and Massague J. (1999). Nature, 397, 710 ± CH and ten Dijke P. (1997). Nature, 389, 631 ± 635. 713. Pereira RC, Economides AN and Canalis E. (2000). Vindevoghel L, Lechleider RJ, Kon A, de Caestecker MP, Endocrinology, 141, 4558 ± 4563. Uitto J, Roberts AB and Mauviel A. (1998). Proc. Natl. Phillips DJ. (2000). Bioessays, 22, 689 ± 696. Acad.Sci.USA,95, 14769 ± 14774. Phillips DJ and de Kretser DM. (1998). Front. Neuroendo- Wrana JL. (2000). Cell, 100, 189 ± 192. crinol., 19, 287 ± 322. Yamashita T, Takahashi S and Ogata E. (1992). Blood, 79, Piek E, Heldin CH and Ten Dijke P. (1999). Faseb J., 13, 304 ± 307. 2105 ± 2124. Yasuda H, Mine T, Shibata H, Eto Y, Hasegawa Y, Reddi AH. (2001). Arthritis Res., 3, 1±5. Takeuchi T, Asano S and Kojima I. (1993). J. Clin. Russell CE, Hedger MP, Brauman JN, de Kretser DM and Invest., 92, 1491 ± 1496. Phillips DJ. (1999). Mol. Cell. Endocrinol., 148, 129 ± 136. Ying SY, Zhang Z, Furst B, Batres Y, Huang G and Li G. Schwall RH, Robbins K, Jardieu P, Chang L, Lai C and (1997). Proc.Soc.Exp.Biol.Med.,214, 114 ± 122. Terrell TG. (1993). Hepatology, 18, 347 ± 356. Yu J and Dolter KE. (1997). Cytokines Cell. Mol. Ther., 3, Shimasaki S, Koga M, Esch F, Cooksey K, Mercado M, 169 ± 177. Koba A, Ueno N, Ying SY, Ling N and Guillemin R. Zhang Y, Feng X, We R and Derynck R. (1996). Nature, 383, (1988). Proc. Natl. Acad. Sci. USA, 85, 4218 ± 4222. 168 ± 172. Smith WC. (1999). Trends Genet., 15, 3±5. Zhang YQ, Kanzaki M, Shibata H and Kojima I. (1997). TakaseM,ImamuraT,SampathTK,TakedaK,IchijoH, Biochim. Biophys. Acta, 1354, 204 ± 210. Miyazono K and Kawabata M. (1998). Biochem. Biophys. Res. Commun., 244, 26 ± 29.

Oncogene