Retinoic Acid Receptors Interfere with the TGF-B/Smad Signaling Pathway in a Ligand-Specific Manner

Retinoic acid receptors interfere with the TGF-b/Smad signaling pathway in a ligand-speciﬁc manner

Vale´ rie Pendaries1, Franck Verrecchia1, Serge Michel2w and Alain Mauviel*,1

1INSERM U532, Universite´ Paris VII, Institut de Recherche sur la Peau, Pavillon Bazin, Hoˆpital Saint-Louis, 1, avenue Claude Vellefaux, F-75475 Paris Cedex 10, France; 2Galderma R&D, F-06902 Sophia-Antipolis, France

Transforming growth factor-b (TGF-b) and retinoic acid steroid, vitamin D, and thyroid hormone receptors, the (RA) are important regulators of cell growth and peroxisome proliferator-activated receptor, the insect differentiation. The TGF-b receptors utilize Smad pro- edysteroid receptor, and a number of orphan receptors teins to transduce signals intracellularly and regulate whose ligands are unknown (Chambon, 1996; Jetten transcription of target genes, either directly or in et al., 2001). All these nuclear receptors share several combination with other sequence-specific transcription well-characterized structural domains, including a con- factors. Two classes of nuclear receptors, the retinoic acid served DNA- binding domain, and a ligand-binding receptors (RARs) and the retinoic X receptors, are domain at the carboxyl terminus of the receptor involved in mediating transcriptional responses to RA. (Pemrick et al., 1994; Kumar and Thompson, 1999). Given the known interactions between the TGF-b and Nuclear retinoic acid receptors, the retinoic acid RAR pathways, we have investigated the role played by receptor (RAR), and retinoic X receptor (RXR) classes, RAR ligands in modulating functional interactions are each composed of a, b, and g subtypes with more between Smad3 and RARs. Using transient cell transfec- than one isoform for each receptor subtype (Chambon, tion experiments with an artificial Smad3/Smad4-depen- 1996).RARs associate with RXRs to form RAR–RXR dent reporter construct, we demonstrate that RAR heterodimers which bind target DNA sequences, overexpression enhances Smad-driven transactivation, an called retinoic acid response elements (RAREs), effect that requires both Smad3 and Smad4. We provide consisting of directly repeated hexameric half-sites with evidence that RAR effect on Smad3/Smad4-driven the consensus sequence 50-PuG(G/A)(T/A)CA-30, sepa- transcription is prevented by natural and synthetic RAR rated by one to five nucleotides depending on the target agonists, and potentiated by synthetic RAR antagonists. gene.Members of the RAR family recognize two The activity of two TGF-b-responsive human gene natural stereoisomers of RA, all-trans RA (ATRA) promoter constructs was regulated in a parallel fashion. and 9-cis RA, whereas the RXR family exclusively Using both mammalian two-hybrid and immunoprecipita- recognizes 9-cis RA.It has been suggested that tion/Western methods, we demonstrate a direct interac- individual receptor subtypes may control distinct gene tion between the region DEF of RARc and the MH2 expression patterns important for cell growth and domain of Smad3, inhibited by RAR agonists and differentiation (Chambon, 1996). enhanced by their antagonists. We propose that RARs Cellular signaling from the TGF-b family of growth may function as coactivators of the Smad pathway in the factors (e.g. activins, bone morphogenic proteins, and absence of RAR agonists or in the presence of their the TGF-bs) is initiated by binding of the ligand to antagonists, a phenomenon that contrasts with their transmembrane receptor serine/threonine kinases, TbRI known role as agonist-activated transcriptional regulators and TbRII (Derynck and Feng, 1997).Following ligand of RA-dependent genes. activation, signaling from the receptors to the nucleus Oncogene (2003) 22, 8212–8220. doi:10.1038/sj.onc.1206913 occurs predominantly by phosphorylation of cytoplas- mic mediators belonging to the Smad family (Datto and Keywords: transforming growth factor-b; nuclear recep- Wang, 2000; Attisano and Wrana, 2002).The receptor- tors; RAR; Smad; promoter region; gene expression associated Smads (R-Smads), such as Smad1, Smad2, Smad3, and Smad5, interact directly with, and are phosphorylated by, activated type I receptors of the TGF-b superfamily.Activation of the R-Smads is Introduction ligand-specific, but each then forms, upon phosphorylation, heteromeric complexes with Smad4, a common The retinoid receptors belong to a large superfamily of mediator for all receptor-activated Smad pathways. ligand-inducible transcription factors that include the R-Smad/Smad4 complexes are then translocated into the nucleus where they function as transcription *Correspondence: A Mauviel; E-mail: [email protected] factors, either directly or in association with wDeceased other DNA-binding factors (Massague´ and Wotton, Received 31 March 2003; revised 20 June 2003; accepted 30 June 2003 2000).Activated T bRI may phosphorylate Smad2, Modulation of RAR/Smad interactions V Pendaries et al 8213 which drives activin-like transcriptional responses Results through association with transcription factors of the FAST family, or Smad3, which drives TGF-b-specific RAR ligands alter TGF-b-induced Smad3/4-specific transcriptional responses, directly binding specific cis- transcription elements which most often contain the nucleotidic As a first approach to examine the possible interference sequence CAGACA.Another group of Smad proteins, of RAR signaling with the TGF-b/Smad pathway, we the inhibitory Smads, including Smad6 or Smad7, determined the effect of RAR agonists and antagonists prevent phosphorylation and/or nuclear translocation on TGF-b-induced, Smad-driven, gene transactivation of R-Smad/Smad4 heterocomplexes (Datto and Wang, in WI-26 lung fibroblast cultures transfected with the 2000; Massague´ and Wotton, 2000; Attisano and Wrana, 2002). artificial (CAGA)9-lux reporter construct.The latter, which consists of nine copies of the nucleotidic motif Several recent reports have provided novel insights CAGA known to bind Smad3 and Smad4 with high regarding the interactions of nuclear receptors with the Smad signaling pathway.In the case of vitamin D, affinity, cloned upstream of the adenovirus major late promoter, to control the expression of the luciferase which controls transcription of target genes through reporter gene (Dennler et al., 1998), allows to monitor the vitamin D receptor (VDR), Smad3 acts as a Smad3/Smad4-specific transcription in response to coactivator specific for ligand-induced VDR-dependent TGF-b.As expected, TGF- b alone dramatically ele- transactivation, by forming a complex with a member of vated (CAGA) -lux activity (Figure 1a).ATRA and the steroid receptor coactivator-1 protein family 9 CD2043, a synthetic RAR agonist, had a dose- (SRC-1) in the nucleus (Yanagisawa et al., 1999). dependent inhibitory effect on TGF-b-induced transac- The estrogen receptor (ER) suppresses TGF-b-induced tivation, with maximal repressions of 35 and 55%, Smad3 activity when, at the mean time, ER-mediated respectively, at a concentration of 10À6 m.At each transcription is enhanced by TGF-b signaling À10 À8 À6 m (Matsuda et al., 2001). Direct association of Smad3 concentration tested (10 ,10 , and 10 ), CD2043 was consistently more potent than ATRA to repress with the androgen receptor has also been described, Smad-dependent transcription.Inversely, CD3106, a leading to suppression of Smad3/DNA association and synthetic RAR antagonist, dose-dependently poten- resulting in Smad3-dependent gene transactivation tiated TGF-b-induced, Smad3/Smad4-specific, tran- downstream of TGF-b (Chipuk et al., 2001). Peroxi- scriptional response (Figure 1b), with a maximal some proliferator-activated receptor g has been shown effect, three-fold above levels achieved with TGF-b to inhibit TGF-b-induced connective tissue growth alone, at a concentration of 10À6 m.RAR agonists factor expression by interfering with Smad3 (Fu et al., CD2043 and ATRA (not shown), added simultaneously 2001). with CD3106 at a 1/10 ratio (10À7 m CD2043 or ATRA, Cooperation of TGF-b with RA occurs both in À6 m physiological and pathological situations (Roberts and 10 CD3106), neutralized each other in the Smad transactivation assay (Figure 1c), consistent with RAR Sporn, 1992).An explanation may be that RA increases agonist/antagonist effects exerted at the level of the production of active TGF-b and its receptors by a variety of cell types, a mechanism responsible for the endogenous RARs. growth inhibition of human keratinocytes and U937 et al leukemia cells, for example (Nunes ., 1996; RAR and RXR overexpression enhances TGF-b-induced Defacque et al., 1997; Borger et al., 2000). Most Smad3/Smad4-specific transcription recently, it was shown that ATRA is able to decrease the levels of phosphorylated Smad2/Smad3 in response We next tried to determine whether overexpression of to TGF-b, possibly via a RARa-mediated activation of RARs and RXR would affect TGF-b response.As a nuclear serine–threonine phosphatase that remains to shown in Figure 2a, transfection of expression vectors be identified (Cao et al., 2003). Another report demon- for RARa,RARb, and RARg resulted in the upregula- strated intracellular interactions between the TGF-b and tion of Smad3/Smad4-dependent transactivation of RA pathways, whereby RAR stimulates, and fusion (CAGA)9-lux in response to TGF-b.Reproducibly, proteins of retinoid receptors antagonize, TGF-b- RARg was the most potent among RARs to enhance induced growth inhibition of lung epithelial cells the TGF-b response.Similarly, when Smad3 was over- (La et al., 2003). In the latter case, direct interactions expressed ectopically as an alternative to exogenous between RARa and the fusion protein PLZF-RARa TGF-b, coexpression of RARs resulted in a significant with Smad3 were described, that correlate with in- potentiation of the transactivation of (CAGA)9-lux creased or decreased p15INK4b expression, respectively. exerted by Smad3 (Figure 2b).Similar to what we In this report, we have studied the modulation of observed with exogenous addition of TGF-b,RARg was Smad3/Smad4-dependent gene transcription by RAR the most potent of all RARs in this assay.Interestingly, agonists and antagonists.Our data demonstrate RAR identical results were obtained when experiments were ligand-dependent modulation of TGF-b response, performed in medium complemented with lipid-free whereby RAR agonists repress and RAR antagonists serum (not shown), indicating that the RAR effect is not potentiate Smad-driven transcription via opposite mod- due to RA present in the serum. ulation of RAR–Smad3 interactions.Details are pro- Since RARs function as heterodimers with RXRs to vided herein. drive RARE-dependent transcription, we next examined

Oncogene Modulation of RAR/Smad interactions V Pendaries et al 8214

Figure 2 RAR overexpression potentiates TGF-b/Smad signaling. (a) Subconfluent WI-26 flbroblast cultures were cotransfected with (CAGA)9-lux together with expression vectors for RARa,RARb or RARg, as indicated.At 6 h after transfection, TGF- b (10 ng/ml) was added.( b) Subconfluent WI-26 flbroblast cultures were cotransfected with (CAGA)9-lux together with RAR expression vectors, in the absence (À) or presence ( þ ) of a Smad3 expression vector, as indicated.( c) Subconfluent WI-26 fibroblast cultures were cotransfected with (CAGA)9-lux together with expression vectors for RARa, RARb or RARg, as indicated, in the absence (À) or presence ( þ ) of either Smad3 or RXRa expression vectors, alone or in combination.In each case, incubations were continued for 24 h before reporter gene activity was determined

Figure 1 Effect of RAR ligands on Smad3/Smad4-dependent whether RXRa had similar effects as RARs on Smad3- transcription downstream of TGF-b in WI-26 human lung specific transcription, and whether RXRa would co- fibroblasts.( a) Subconfluent WI-26 flbroblast cultures were operate with RARs to activate Smad3-driven gene transfected with (CAGA)9-lux.Following glycerol shock, DMEM containing 1% FCS was added.After 6 h, several concentrations of expression.As shown in Figure 2c, RXR a overexpres- either ATRA (shaded bars) or the synthetic RAR agonist CD2043 sion potentiated Smad3-driven (CAGA)9-lux transacti- (striped bars) were added, without (À) or with ( þ ) TGF-b (10 ng/ vation.Coexpression of RXR a together with RARa, ml).Incubations were continued for 24 h before reporter gene RARb, and RARg resulted in a strong potentiation of activity was determined.( b) The same protocol as in panel a was RARa effect on Smad3-specific transcription, whereas followed, except that the synthetic RAR antagonist CD3106 was added instead of RAR agonists.( c) Subconfluent WI-26 flbroblast little augmentation of RARg effect was observed, the cultures were transfected with (CAGA)9-lux.After 6 h, CD2043 latter likely masked by the strong effect of RARg alone. (10À7 m) was added, in the absence or presence of CD3106 (10À6 m), Little effect of RARb was observed, without or with without (À) or with ( þ ) TGF-b.Incubations were continued for RXRa. 24 h before reporter gene activity was determined.Results are mean7s.d. of three independent experiments The current model for Smad-dependent transcription in response to TGF-b implies heterodimerization of

Oncogene Modulation of RAR/Smad interactions V Pendaries et al 8215 phosphorylated Smad3 with Smad4, nuclear translocation, and subsequent DNA binding of Smad3/4 heterocomplexes (Datto and Wang, 2000; Massague´ and Wotton, 2000; Attisano and Wrana, 2002).However, recent evidence suggests that nuclear transport of Smad3 in response to TGF-b does not require Smad4, although the latter is essential for Smad3-dependent transcriptional response (Fink et al., 2003). Two distinct approaches were taken to determine whether the effects of RARs described above were strictly dependent on a fully functional Smad3/Smad4 pathway or if RARs could replace one of the normal constituents of the Smad complex bound to its cognate DNA sequence. Firstly, expression of a dominant-negative form of Smad3, Smad3DMH2, not only prevented TGF-b effect on (CAGA)9-lux transactivation in WI26 fibroblasts, but also abolished further activation by RARs (Figure 3a).This result suggests that direct protein– protein interactions with Smad3 bound to DNA via its MH1 domain are necessary for RAR effect.Secondly, in the Smad4-deficient MDA-MB-468 human breast carcinoma cell line transfected with (CAGA)9-lux and a RARg expression vector, there was no activation of the construct by TGF-b, consistent with previous observa- tions (Vindevoghel et al., 1998), whether in the absence or presence of RARg (Figure 3b).In the latter cell type, ectopic expression of Smad4 restored a ‘normal’ TGF-b response, as well as the potentiating effect of RARg on the latter.These data support the notion that proper Smad3/Smad4-driven promoter transactivation in response to TGF-b, that is, concomitant presence of Smad3 and Smad4 in the nucleus bound to their cognate DNA element, is required for RARs to exert their effect. Also, our data indicate that RARg cannot replace Smad4 to drive Smad3-specific gene expression in a context devoid of Smad4.

RARg interacts with the MH2 domain of Smad3 through its DEF region Figure 3 Smad3 and Smad4 are both required for RAR effect on (CAGA) -lux transactivation by TGF-b.(a) Subconfluent WI-26 Next, to better understand the molecular mechanisms 9 fibroblast cultures were cotransfected with (CAGA)9-lux together underlying the functional cooperation of RARs with with RAR expression vectors, in the absence (À) or presence ( þ ) Smad3, protein–protein interactions between RARg and of a dominant-negative Smad3 expression vector (D/N Smad3), as Smad3 were examined by IP/Western approaches.For indicated.At 6 h after transfection, TGF- b was added ( þ ), and incubations continued for 24 h before reporter gene activity was this purpose, COS-7 cells were transfected with Smad3- determined.( b) Subconfluent Smad4-deficient MDA-MB-468 hu- Flag and Gal4BD-RARgDEF expression vectors, to- man breast adenocarcinoma cell cultures were cotransfected with gether with a constitutively active form of TbRI. (CAGA)9-lux together with expression vectors for RARg, in the Immunoprecipitations of the cell lysates were performed absence (À) or presence ( þ ) of a Smad4 expression vector, as with an anti-Gal BD antibody, followed by Western indicated.At 6 h after transfection, TGF- b was added, and 4 incubations continued for 24 h before reporter gene activity was analysis with anti-Flag and anti-Gal4BD antibodies.As determined shown in Figure 4a, full-length Smad3 interacts with RARg in solution, as evidenced by the co-precipitation of Smad3-Flag together with Gal4BD-RARgDEF. Similarly, the Myc-tagged MH2 domain of Smad3 RARgDEF fusion protein, in the absence or presence of coimmunoprecipitated with RARg (Figure 4b), indicat- an expression vector for VP16AD-Smad3.When ing that Smad3-RARg interactions occur through the Gal4BD-RARgDEF is bound to the Gal4-binding sites MH2 domain of Smad3. of the Gal4-lux construct, Smad3-RARg interaction is We next used a mammalian two-hybrid transactiva- expected to drive luciferase activity higher, due to the tion assay aimed at measuring Smad3/RARg interac- strong VP16 transactivation domain fused to Smad3. tions.In such system, the Gal4-lux reporter construct is Results of a representative experiment are presented in cotransfected with an expression vector for the Gal4BD- Figure 4c: Gal4-RARgDEF expression resulted in

Oncogene Modulation of RAR/Smad interactions V Pendaries et al 8216 detectable transactivation of Gal4-lux, likely represen- indicating that the MH2 domain of Smad3 is not tative of the activity of the AF-2 transactivation region dispensable in allowing RARg–Smad3 interaction. within the E module of RARg (Chambon, 1996). These results are consistent with those obtained by IP/ Coexpression of VP16AD-Smad3 strongly enhanced Western and presented in Figure 4b, and with the Gal4BD-RARgDEF-mediated transactivation of Gal4- known role of the MH2 domain of R-Smads in lux, a phenomenon reflecting direct RARg-Smad3 mediating protein–protein interactions, such as with interaction.No effect of either empty pVP16AD or Smad4, activated TbRI, Fast or Jun proteins, for VP16AD-Smad3 alone was observed.When VP16AD- example (Liberati et al., 1999; Nagarajan et al., 1999; Smad3(1–146), lacking the Smad3 MH2 domain, was Shi, 2001; Verrecchia et al., 2001b). expressed instead of the full-length Smad3 fusion protein, no transactivation of Gal4-lux was observed, Ligand-dependent modulation of RARg–Smad3 interactions and of Smad3-dependent transcription To gain insights into the involvement of RARs in the modulation of Smad3/Smad4-dependent transcription, we examined the role of synthetic RAR agonists and antagonists on the functional interactions between Smad3 and RARg.As shown in Figure 5a, the RAR antagonist CD3106 dramatically potentiated the effect of RARg on Smad3-driven transcription (over fourfold above the transactivation level achieved with the RARg/ Smad3 combination), whereas the RAR agonist CD2043 and ATRA abolished the effect of RARg on Smad3-driven (CAGA)9-lux transactivation. We next examined the influence of RAR ligands, agonists and antagonists, on RARg/Smad3 interactions, in IP/Western experiments.COS-7 cells were transfected with Flag-tagged full-length Smad3 and Gal4BD- RARgDEF expression vectors, either alone or in combination, in the absence (Figure 5b, lanes 1–3) or presence or either CD2043 (lanes 4–6) or CD3106 (lanes 7–9).Expression of all proteins from their transfected expression vectors was verified in each sample prior to IP (Figure 5b, lower panels).The lower migrating band in lanes 2, 5, and 8 obtained with the anti-Gal4BD antibody represents the Gal4BD itself, not fused to RARg, properly expressed by the ‘empty’ Gal4BD vector used as a DNA filler in these samples.Following IP with an anti-Gal4BD antibody (upper panel), Western analysis with a Flag antibody revealed the coimmunoprecipitation of Smad3 with RARg in lane 3,

3 - Figure 4 Physical interactions occur between Smad3 and RARg. (a) COS-7 cells were transfected with Flag-tagged Smad3, in the absence or presence of Gal4-tagged RARgDEF, as indicated.Empty Gal4 expression vector (Gal4e) was used as a control.Cell lysates were subjected to immunoprecipitation (IP) with anti-Gal4 antibodies, and co-precipitating Smad3 was detected by Western blotting (WB) with anti-Flag antibodies (top panel).Expression of RARg and Smad3 was confirmed using anti-Flag (middle panel) and anti-Gal4 antibodies (bottom panel).( b) COS-7 cells were transfected with Myc-tagged Smad3 MH2 domain and Gal4-RARgDEF expression vectors, as indicated.After 40 h, cell extracts were immunoprecipitated with an anti-Gal4 antibody, and immunoblotted with anti-Myc antibodies (see legend).Expression of the proteins of interest was verified by WB prior to IP.( c) Confluent WI-26 fibroblast cultures were cotransfected with Gal4-lux reporter construct Gal4-RARgDEF fusion protein expression vector, in the presence or absence of either pVP16Smad3FL of pVP16Smad3(1–146) expression vectors.Empty VP16 and Gal4 vectors were used to maintain equal amounts of transfected DNA in each plate, where indicated.After 40 h, luciferase activity, representative of protein–protein interactions, was measured

Oncogene Modulation of RAR/Smad interactions V Pendaries et al 8217 antagonists, in a fashion compatible with the modulation of Smad3-driven transcription exerted by RAR ligands.Speciﬁcally, the RAR agonist CD2043, which prevents the stimulatory activity of RARs on Smad- dependent transcription, also reduces Smad3–RARg interactions, whereas the RAR antagonist CD3106, which potentiates RARg effect on Smad-dependent transcription, augments Smad3–RARg physical interactions.

RAR ligands modulate TGF-b-induced transcription of natural promoter sequences The effects of RAR ligands on TGF-b-induced transcription were next tested in a genomic DNA context. Specifically, we examined the effects of CD2043 and CD3106 regulation of TGF-b-induced transactivation of the human type VII collagen (COL7A1)and plasminogen activator inhibitor-1 (PAI-1) gene promoters.Both COL7A1 and PAI-1 promoter fragments are defined Smad3/Smad4 targets downstream of TGF-b (Dennler et al., 1998; Vindevoghel et al., 1998). As shown in Figure 6a, CD3106 strongly potentiated, while CD2043 inhibited the transactivation of the COL7A1 promoter induced by TGF-b, in perfect parallel with the data obtained with the artificial, yet highly specific, Smad3/Smad4-dependent construct (CAGA)9-lux (see above Figure 5a).Similar modulation of TGF- b response of the PAI-1 promoter by RAR agonists was observed (Figure 6b).Of note, significant regulation of both COL7A1 and PAI-1 promoter activities by RAR ligands alone was observed, suggesting that specific RAR responsive elements may exist within these promoters, and raising the possibility that, in addition to acting via the Smad response elements as suspected from our current data, RARs may also cooperate with Smad3/4 through the use of distinct regulatory sequences.

Figure 5 RAR ligands modulate the transcriptional outcome or RARg-Smad3 functional interaction and modify RARg–Smad3 interactions.( a) Subconfluent WI-26 fibroblast cultures were Discussion cotransfected with (CAGA)9-lux together with expression vectors for RARg, in the absence (À) or presence ( þ ) of a Smad3 expression vector, as indicated.At 6 h after transfection, the RAR Until recently, although functional interactions between antagonist CD3106 (10À6 m) or the RAR agonist CD2043 (10À7 m) TGF-b and retinoids had been observed in a variety of were added, as indicated.Luciferase activity was determined 24 h settings (Roberts and Sporn, 1992), little was known later.( b) COS-7 cells were transfected with Flag-tagged Smad3, in about the precise mechanisms underlying the intracel- the absence or presence of Gal4-tagged RARgDEF, as indicated. lular interplay between their respective signaling path- Empty Gal4 expression vector (Gal4e) was used as a control.At 6 h after transfection, CD2043 (10À7 m) or CD3106 (10À6 m) were ways.In this report, we have identified a new molecular added.After 40 h, cell lysates were subjected to immunoprecipita- mechanism by which RARs are able to modulate Smad- tion (IP) with anti-Gal4 antibodies, and co-precipitating Smad3 dependent transcription in a ligand-dependent fashion. was detected by immunoblotting (WB) with anti-Flag antibodies We demonstrate that RARs potentiate Smad3/Smad4- (top panel).Expression of RAR g and Smad3 was confirmed using anti-Flag (middle panel) and anti-Gal4 antibodies (bottom panel) dependent transcriptional responses downstream of TGF-b, and that RAR agonists oppose while RAR antagonists favor Smad3–RAR interactions, resulting in either inhibition or potentiation of TGF-b signaling, respectively.Although a role of RAR conformation for as expected from our data presented in Figure 4.The RAR-Smad3 may only be ascertained by X-ray crystal- amount of RARg-bound Smad3 was severely reduced lography, our results are compatible with RAR by CD2043 (lane 6), whereas it was significantly conformation-dependent RAR–Smad3 interactions increased by CD3106 (lane 9), suggesting that RARg– whereby an RAR in the presence of its antagonist, that Smad3 interactions are altered by RAR agonists and is, in a conformation that is transcriptionally inactive in

Oncogene Modulation of RAR/Smad interactions V Pendaries et al 8218 therapeutic implications of such findings remain to be identified. Escape of epithelial cells from TGF-b growth control is a hallmark of many cancers (Gold, 1999; Reiss, 1999; de Caestecker et al., 2000). The capacity of TGF-b signaling to function as a tumor suppressor pathway in early carcinogenesis is best illustrated by the presence of inactivating mutations in genes encoding TGF-b receptors and Smads in human carcinomas.However, contrasting with its tumor-suppressor activities that control epithelial cell growth and maintain genetic stability, TGF-b may become pro-oncogenic at later stages of carcinogenesis.Indeed, not only are high levels of TGF-b expression correlated with the advanced clinical stage of a tumor, but TGF-b is also known to exacerbate the malignant phenotype of transformed and tumor-derived cells.Furthermore, tumor-derived TGF- b could contribute to tumor growth indirectly by suppressing immunosurveillance or by stimulating the production of proangiogenic factors, or by inducing an epithelial to mesenchymal transition in various tumor cells, leading to increased invasiveness and metastatic activity. Given the complexity of TGF-b actions during carcinogenesis, identifying and understanding the mechanisms underlying cellular signaling crosstalks that modify TGF-b responses is critical for the development of efficient therapeutic agents used for chemoprevention or for the treatment of advanced cancers.Moreover, since TGF-b plays opposite roles on malignant progres- sion depending on the stage of carcinogenesis, therapeutic approaches targeting the TGF-b pathway would have to be context-specific.Theoretically, chemoprevention would aim at enhancing TGF-b signaling in order to promote its tumor-suppressor activities, whereas treatment of advanced cancer would require strategies aimed at inactivating TGF-b or its downstream signals. Retinoids are widely used as chemopreventive drugs in the context on nonmelanoma skin cancers or acute promyelocytic leukemia, for example.Interestingly, it has been suggested that their efficacy may result, to some extent, from their ability to induce the production of TGF-b by a variety of cell types (Gold, 1999; Reiss, 1999; de Caestecker et al., 2000). It is, however, not a Figure 6 RAR ligands modulate the activity of Smad3/Smad4- generalized phenomenon as, for example, in a study dependent target gene promoters.Subconfluent WI-26 fibroblast comparing the effect of ATRA on melanoma growth cultures were transfected with the human type VII collagen promoter construct À524COL7A1-lux (a) or the human plasmino- and migration, it was shown that although proliferation gen activator inhibitor-1 promoter construct p800-lux (b).Follow- and chemotaxis were inhibited by ATRA in all the cell ing glycerol shock, DMEM containing 1% FCS was added.After lines tested, cell lines established from metastases were 6 h, the synthetic RAR antagonist CD3106 or the RAR agonist significantly more sensitive with respect to inhibition of CD2043 were added, without (À) or with ( þ ) TGF-b (10 ng/ml). invasion by ATRA.The most striking difference Incubations were continued for 24 h before reporter gene activity was determined between the cell lines was a strong downregulation of TGF-b expression in cell lines derived from metastases when treated with ATRA in contrast to cell lines from the context of RAR responsive cis-elements (RARE), primary melanomas (Jacob et al., 1998). Of note, acts as a coactivator of Smad3/Smad4-specific transcrip- previously we reported constitutive Smad-dependent tion.Inversely, RAR agonists, that render RARs transcription in human melanoma cells, regardless of transcriptionally competent when bound to their specific their proliferative response to exogenous TGF-b DNA recognition sites, diminish both RAR–Smad3 (Rodeck et al., 1999). It will be interesting to determine interactions and Smad3/Smad4-dependent transcription whether the results from Jacob et al.(see above) are due downstream of the TGF-b receptors.Potential to a direct interference of ATRA with Smad signaling,

Oncogene Modulation of RAR/Smad interactions V Pendaries et al 8219 and whether such interference is modified by the stage of Fungizonet).Cells were maintained at 37 1Cina5% the tumor. CO2 – 95% air atmosphere.In some experiments, In another study, Ro 41-5253, an RARa-selective charcoal-stripped FCS was used to alleviate the antagonist that binds RARa but does not induce potential influence of FCS-contained lipids (e.g. reti- transcriptional activation and does not influence RAR/ noids).No difference in the outcome of the experiments RXR heterodimerization and DNA binding, was shown was noted when using lipid-free serum instead of whole to inhibit proliferation and to induce apoptosis in MCF- serum (not shown).Human recombinant TGF- b1, 7 and ZR-75.1 ER-positive breast-carcinoma cells referred to as TGF-b throughout the text, was (Toma et al., 1998). The authors suggested that the purchased from R&D Systems Inc.(Minneapolis, MN, antiproliferative effect was probably mediated by anti- USA).All- trans retinoic acid (all-trans RA) was AP-1 activity, a mechanism known to be implicated in purchased from Sigma (St Louis, MO, USA).Synthetic the action of several retinoids (Allenby, 1995; Chen et al., RA agonist (CD2043) and antagonists (CD3106, 1995).However, the authors reported that induction of CD3581, and CD2848) were from Galderma R&D apoptosis by Ro 41-5253 correlated with an increased (Sophia-Antipolis, France). production of TGF-b1 protein by the tumor cells.In view of our current data, it cannot be excluded that Ro Plasmid constructs 41-5253 may also be potentiating TGF-b antiproliferative action through activation of the Smad pathway, the Human RARa,RARb, RARg, RXRa, and GAL4BD- latter leading to the expression of p15INK4B (Feng et al., RARDEF expression vectors, kind gifts from Drs C 2000; Seoane et al., 2001) and p21WAF1/CIP1 (Datto et al., Rochette-Egly and P Chambon, Strasbourg, France, 1995; Moustakas and Kardassis, 1998), two cyclin- have been described previously (Allenby et al., 1993). dependent kinase inhibitors, known effectors of TGF-b- RARg deletion expression vectors in pGL3 (Promega induced cell cycle arrest.These data, together with our Corp.) were generated by PCR. (CAGA)9-lux, a gift results herein demonstrating the cooperative activity of from Drs S Dennler and J-M Gauthier (Glaxo-Well- antagonist-bound RARs with Smad3/Smad4, require to come, Les Ulis, France, consisting of nine tandem further examine the exact mechanisms for the antitumor repeats of the motif CAGA binding Smad3 and Smad4, activity of RAR antagonists.It is particularly important was used as a Smad3/Smad4-specific reporter (Dennler because it was clearly shown that Ro 41-5253 is a et al., 1998)). (ARE)3-lux, a prototypic Smad2-specific compound that exerts its antitumor activity without reporter construct consisting of three repeats of the inducing the toxic side effects of retinoids, and might activin response element (ARE) of the Xenopus mix.2 therefore be considered as a candidate for cancer gene (Chen et al., 1997), was cotransfected with a Fast-1 therapy (Toma et al., 1998). expression vector (Germain et al., 2000), kind gift from To conclude, we provide novel insights on how Dr CS Hill (Imperial Cancer Research Fund, London, retinoids may affect TGF-b signaling.Specifically, this UK).N-terminally Flag- and Myc-tagged Smad3, work represents the first demonstration for RAR- dominant-negative Smad3DMH2, and Vp16AD-Smad3 ligand-controlled RAR–Smad3 interactions.Most sur- expression vectors, gifts from A Atfi, Paris, France, have prisingly, we have identified RAR antagonists as been described previously (Vindevoghel et al., 1998; activators of Smad3/Smad4-dependent transcription, Verrecchia et al., 2001a). Smad3/Smad4 responsiveness correlated with their capacity to increase RAR–Smad3 of the human COL7A1 and PAI-1 promoter constructs physical interactions.Work is in progress to determine –527COL7A1-lux and p800-lux – has been described whether RAR antagonists are able to restore TGF-b- previously (Dennler et al., 1998; Vindevoghel et al., induced growth inhibition in tumor cells that have 1998). escaped from TGF-b control and do not exhibit mutations in the genes encoding TGF-b signaling Transient cell transfections and reporter assays molecules (e.g. receptors or Smads) but rather, reduced signaling. Transient cell transfections were performed with the calcium phosphate/DNA co-precipitation procedure using a commercial assay kit (Promega Corp., Madison, WI, USA).After glycerol shock, cultures were placed in Material and methods DMEM containing 1% FCS.In some experiments, Cell cultures TGF-b and/or retinoids were added 3 h later.pCMV- b- galactosidase was cotransfected in every experiment to Human lung fibroblasts (WI-26), MDA-MB-468 human monitor transfection efficiencies.Reporter activities breast adenocarcinoma cells, a cell line derived from a were determined with commercial kits (Promega). patient with breast carcinoma and bearing a homo- zygous deletion of the complete SMAD4 coding region Immunoprecipitations and Western blotting (Schutte et al., 1996), and simian COS-7 cells were grown in Dulbecco’s modified Eagle’s medium supple- COS-7 cells were transfected with tagged Smad3 and mented with 10% heat-inactivated fetal calf serum RARg expression vectors, together with pCMV-b- (FCS), 2 mm glutamine and antibiotics (100 U/ml galactosidase, the latter serving as a control for penicillin, 50 mg/ml streptomycin G, and 0.25 mg/ml transfection efficiency.After 24 h, cells were washed

Oncogene Modulation of RAR/Smad interactions V Pendaries et al 8220 twice with cold PBS, scraped, and solubilized in a buffer nescence detection system (ECL þ , Amersham-Pharma- containing 20 mm Tris-HCl, pH 8, 150 mm NaCl, 5 mm cia Biotech). MgCl2, 0, 5% NP-40, 10% glycerol, 1 mm orthovana- date, 1 mm PMSF, 20 mg/ml aprotinine, and 20 mg/ml leupeptine.For Gal4-tagged protein immunoprecipitation, cell lysates were cleared of debris by centrifugation, Abbreviations and incubated with an anti-Gal4 antibody (Santa-Cruz ATRA, all-trans retinoic acid; FCS, fetal calf serum; RA, Biotech, Santa Cruz, CA, USA) overnight at 41C, retinoic acid; RAR, retinoic acid receptor; TGF-b, transforming growth factor-b. followed by incubation with protein G–sepharose beads at 41C for 1 h (Amersham Pharmacia Biotech, Uppsala, Sweden).After five washes with solubilization buffer, Acknowledgements immunoprecipitates were eluted by boiling for 3 min in We are indebted to Drs A Atfi (INSERM U482, Paris, SDS sample buffer (100 mm Tris-HCl, pH 8, 8, 0, 01% France), P Chambon, and C Rochette-Egly (IGBMC, bromophenol blue, 36% glycerol, 4% SDS) and Strasbourg, France), S Dennler and J-M Gauthier (Glaxo- subjected to SDS–polyacrylamide gel electrophoresis. Wellcome, Les Ulis, France), and CS Hill (Imperial Cancer Proteins were then electrotransferred to nitrocellulose Research Fund, London, UK) for providing us with reagents essential for these studies.This work was supported by filters, immunoblotted with anti-Myc-HRP (Roche INSERM and a research grant from Galderma R&D Diagnostics, Indianapolis, IN, USA), anti-Flag-HRP (Sophia-Antipolis, France) to AM.This work is dedicated to (Sigma) or anti-Gal4 (Santa-Cruz Biotech, Santa Cruz, the memory of our colleague and friend Serge. CA, USA) antibodies, and revealed using a chemilumi-

References

Allenby G.(1995). Hum. Exp. Toxicol., 14, 226–230. Jetten AM, Kurebayashi S and Ueda E.(2001). Prog. Nucleic Allenby G, Bocquel MT, Saunders M, Kazmer S, Speck J, Acid Res. Mol. Biol., 69, 205–247. Rosenberger M, Lovey A, Kastner P, Grippo JF Kumar R and Thompson EB.(1999). Steroids, 64, 310–319. Chambon P and Levin AA.(1993). Proc. Natl. Acad. Sci. La P, Morgan TA, Sykes SM, Mao H, Schnepp RW, Petersen USA, 90, 30–34. CD and Hua X.(2003). Oncogene, 22, 198–210. Attisano L and Wrana JL.(2002). Science, 296, 1646–1647. Liberati NT, Datto MB, Frederick JP, Shen X, Wong C, Borger DR, Mi Y, Geslani G, Zyzak LL, Batova A, Engin TS, Rougier-Chapman EM and Wang XF.(1999). Proc. Natl. Pirisi L and Creek KE.(2000). Virology, 270, 397–407. Acad. Sci. USA, 96, 4844–4849. Cao Z, Flanders KC, Bertolette D, Lyakh LA, Wurthner JU, Massague´ J and Wotton D.(2000). EMBO J., 19, 1745–1754. Parks WT, Letterio JJ, Ruscetti FW and Roberts AB. Matsuda T, Yamamoto T, Muraguchi A and Saatcioglu F. (2003). Blood, 101, 498–507. (2001). J. Biol. Chem., 276, 42908–42914. Chambon P.(1996). FASEB J., 10, 940–954. Moustakas A and Kardassis D.(1998). Proc. Natl. Acad. Sci. Chen JY, Penco S, Ostrowski J, Balaguer P, Pons M, Starrett USA, 95, 6733–6738. JE, Reczek P, Chambon P and Gronemeyer H.(1995). Nagarajan RP, Liu J and Chen Y.(1999). J. Biol. Chem., 274, EMBO J., 14, 1187–1197. 31229–31235. Chen X, Weisberg E, Fridmacher V, Watanabe M, Naco G Nunes I, Kojima S and Rifkin DB.(1996). Cancer Res., 56, and Whitman M.(1997). Nature, 389, 85–89. 495–499. Chipuk JE, Cornelius SS, Pultz NJ, Jorgenson JS, Bonham Pemrick SM, Lucas DA and Grippo JF.(1994). Leukemia, 8, MJ, Kim SJ and Danielpour D.(2001). J. Biol. Chem., 13, 13. 1797–1806. Datto MB and Wang XF.(2000). Cytokine Growth Factor Reiss M.(1999). Microbes Infect., 1, 1327–1347. Rev., 11, 37–48. Roberts AB and Sporn MB.(1992). Cancer Surv., 14, 205–220. Datto MB, Yu Y and Wang XF.(1995). J. Biol. Chem., 270, Rodeck U, Nishiyama T and Mauviel A.(1999). Cancer Res., 28623–28628. 59, 547–550. de Caestecker MP, Piek E and Roberts AB.(2000). J. Natl. Schutte M, Hruban RH, Hedrick L, Cho KR, Nadasdy GM, Cancer Inst., 92, 1388–1402. Weinstein CL, Bova GS, Isaacs WB, Cairns P, Nawroz H, Defacque H, Sevilla C, Piquemal D, Rochette-Egly C, Marti J Sidransky D, Casero Jr RA, Meltzer PS, Hahn SA and Kern and Commes T.(1997). Leukemia, 11, 221–227. SE.(1996). Cancer Res., 56, 2527–2530. Dennler S, Itoh S, Vivien D, ten Dijke P, Huet S and Gauthier Seoane J, Pouponnot C, Staller P, Schader M, Eilers M and JM.(1998). EMBO J., 17, 3091–3100. Massague´ J.(2001). Nat. Cell Biol., 3, 400–408. Derynck R and Feng XH.(1997). Biochim. Biophys. Acta, Shi Y.(2001). BioEssays, 23, 223–232. 1333, F105–F150. Toma S, Isnardi L, Raffo P, Riccardi L, Dastoli G, Apfel C, Feng XH, Lin X and Derynck R.(2000). EMBO J., 19, 5178– LeMotte P and Bollag W.(1998). Int. J. Cancer, 78, 86–94. 5793. Verrecchia F, Chu ML and Mauviel A.(2001a). J. Biol. Chem., Fink SP, Mikkola D, Willson JK and Markowitz S.(2003). 276, 17058–17062. Oncogene, 22, 1317–1323. Verrecchia F, Vindevoghel L, Lechleider RJ, Uitto J, Roberts Fu M, Zhang J, Zhu X, Myles DE, Willson TM, Liu X and AB and Mauviel A.(2001b). Oncogene, 20, 3332–3340. Chen YE.(2001). J. Biol. Chem., 276, 45888–45894. Vindevoghel L, Lechleider RJ, Kon A, de Caestecker MP, Germain S, Howell M, Esslemont GM and Hill CS.(2000). Uitto J, Roberts AB and Mauviel A.(1998). Proc. Natl. Genes Dev., 14, 435–451. Acad. Sci. USA, 95, 14769–14774. Gold LI.(1999). Crit. Rev. Oncogen., 10, 303–360. Yanagisawa J, Yanagi Y, Masuhiro Y, Suzawa M, Watanabe Jacob K, Wach F, Holzapfel U, Hein R, Lengyel E, Buettner M, Kashiwagi K, Toriyabe T, Kawabata M, Miyazono K R and Bosserhoff AK.(1998). Melanoma Res., 8, 211–219. and Kato S.(1999). Science, 283, 1317–1321.

Oncogene