Smad7 in TGF-β Signalling

(1)

!"#$ % %

&'

(2)

Dissertation for the Degree of Doctor of Philosophy (Faculty of Medicine) in Molecular Cell Biology presented at Uppsala University in 2002, from the Ludwig Institute for Cancer Research

Abstract

Brodin, G. 2002 Smad7 in TGF-β Signalling. Acta Universitatis Upsaliensis.

Comprehensive Summaries of Uppsala Dissertations from the Faculty of Medicine

1135. 57pp. Uppsala. ISBN 91-554-5271-X.

Members of the transforming growth factor-β (TGF-β) superfamily of growth and differentiation factors regulate a vast array of biological functions in the adult, and are of great importance in governing cell fate determination and patterning in the developing embryo. The TGF-β signal is propagated intracellularly by Smad proteins resulting in transcriptional responses. Smad6 and Smad7 are inhibitory Smads known to downregulate the TGF-β signal and thereby possibly modulating the biological response. This thesis describes a functional analysis of the inhibitory Smad7 from an in

vitro and in vivo perspective.

The prostate gland is dependent on androgens for its growth and differentiation. Androgen withdrawal can cause regression and apoptosis in normal and malignant prostate. Previous studies suggest a role for TGF-β in the apoptotic mechanism. We investigated the expression levels of Smad proteins in the rat ventral prostate as well as in an androgen sensitive prostate tumor model (Dunning R3327 PAP) by immunohistochemistry. We observed an increased immunoreactivity for Smad3, Smad4 and phosphorylated Smad2 in the rat ventral prostate epithelial cells after castration, as well as in the prostate tumor cells. Expression of inhibitory Smad6 and Smad7 were also increased in both normal and malignant prostate in response to castration.

Several studies have shown that Smad7 is upregulated in response to TGF-β

stimuli, suggesting a role in a negative feedback loop attenuating the TGF-β response. We investigated the molecular mechanism behind that response by studying the transcriptional regulation of the Smad7 gene. We identified a palindromic Smad binding element (SBE) in the promoter. Point mutations introduced into the SBE abolished transcriptional activation via TGF-β. We also observed that mutating or deleting binding motifs for Sp1 and AP-1, led to an attenuation of the TGF-β mediated transcriptional induction as well as the basal promoter activity.

Gene ablation of Smad proteins has revealed specific physiological and developmental roles. We analysed mice targeted on the Smad7 locus. The mice appeared viable and fertile with a slight reduction in litter size, suggesting a perinatal loss. Biochemical analysis of mouse embryonic fibroblasts (MEFs) showed no major difference between wild type and mutant MEFs.

Greger Brodin, Ludwig Institute for Cancer Research, Biomedical Centre, Box 595, SE-751 24, Uppsala, Sweden

(3)

To

My Parents

&

(4)

This thesis is based on the following papers, referred to in the text by their Roman numerals:

I. Brodin, G., ten Dijke, P., Funa, K., Heldin, C., and Landström, M. (1999) Increased Smad expression and activation are associated with apoptosis in normal and malignant prostate after castration. Cancer Research 59,

2731-2738

II. Brodin, G., Åhgren, A., ten Dijke, P., Heldin, C., and Heuchel, R. (2000) Efficient TGF-β induction of the Smad7 gene requires co-operation between AP-1, Sp1, and Smad proteins on the mouse Smad7 promoter. J. Biol. Chem.

275, 29023-29030

III. Brodin, G., Cheng, A., Åhgren, A., Kulkarni, S., Pawson, T., Heldin, C.H., and Heuchel, R. Targeting of the mouse Smad7 gene. Manuscript.

Reprints were made with permission from the publishers

We shall not cease from exploration And the end of all our exploring Will be to arrive where we started And know the place for the first time.

(5)

ABBREVIATIONS

ActR activin receptor

ALK activin receptor-like kinase

AMH/MIS anti-Müllerian hormone/Müllerian inhibiting substance

ATF2 activating transcription factor 2

AVM arteriovenous malformations

BAMBI BMP and Activin membrane-bound inhibitor

BMP bone morphogenetic protein

CBFA/AML core-binding factor A/acute myeloneus leukaemia EMSA electrophoretic mobility shift assays

FAST forkhead activin signal transducer

GDFs growth and differentiation factors

GDNF glial cell line-derived neurotrophic factor

HAT histone acetyltransferase

HDAC histone deacetylases

JNK Jun N-terminal kinase

LAP latency-associated protein

LEF1/TCF lymphoid enhancer-binding factor 1/T-cell specific factor

LLCs large latent complexes

LTBP latent TGF-β-binding protein

MEF mouse embryonic fibroblasts

MH1/2 Mad homology ½

SAD Smad activation domain

SARA Smad anchor for receptor activation

SBE Smad binding element

Ski Sloan-Kettering Institute proto-oncogene

SLC small latent complex

SMURF Smad ubiquitylation regulatory factor

SnoN Ski-related novel gene N

TβR TGF-β receptor

TAK1 TGF-β-activated kinase

TFE3 transcription factor binding to immunoglobulin heavy constant mu enhancer 3

TGF-β transforming growth factor-β

TGIF TG3-intercating factor

VSMC vascular smooth muscle cells

(7)

INTRODUCTION

We are all wired. The term “we” is used in a rather generous way referring to ourselves and all the multicellular organisms sharing life with us. The fact that prevents us all from being a shapeless pile of individual cells is biochemical communication. This is achieved through an elaborate signalling network relaying a continuous flow of information within cells and between the cells and their environment. Pathological or dysfunctional conditions may arise when the regulation of this information is perturbed, which ultimately can affect the development of an organism or the maintenance of physiological functions in the adult organism. Cells exchange information by means of direct cell to cell interactions, cell contact with extracellular matrix components or by responding to soluble factors such as hormones, neurotransmitters, cytokines, growth factors or small ions. The extracellular signal is converted into an intracellular one by the use of a specific class of proteins, the so-called receptors. The initiated signal cascade will affect cellular functions such as survival, division, death, movement or differentiation.

Transforming growth factor-β (TGF-β) belongs to a superfamily of cytokines that exert a multitude of cellular functions on the organism affecting such diverse things as tissue homeostasis, cell division and multicellular patterning and development. These diverse effects are mediated through the combination and activation of specific type I and type II serine/threonine receptor kinases that transduce the signal from the cellular surface to the nucleus via downstream effector proteins, called Smads. These proteins can be divided into three subclasses, receptor activated (R-Smads), common mediator Smad (Smad4) and the inhibitory Smads (Smad6 and Smad7). The R-Smads (Smad1, 2, 3, 5 and 8) become phosphorylated in their C-termini and thus activated by the type I receptors. They associate with the common mediator Smad4 and subsequently translocate to the nucleus where they trigger transcriptional responses. The inhibitory Smads exert their effect as attenuators and modulators of the TGF-β signal. The question is how different Smads can influence such a vast array of cellular effects and how pathological conditions such as fibrosis, rheumatoid arthritis and carcinogenesis, can arise in response to a deregulation of this TGF-β signal.

The scope of this thesis was to investigate the functional role of the inhibitory Smad7 from an in vivo and an in vitro perspective. This was achieved by assessing the level of Smad expression in a rat prostate carcinoma model, by characterising the transcriptional regulation of the Smad7 promoter and finally, by disrupting the Smad7 gene in mice, in order to identify phenotypic alterations that would suggest different functional roles for Smad7.

(8)

TRANSFORMING GROWTH FACTOR-

β

(TGF-

β

)

The basis for the discovery of TGF-β was a finding by de Larco and Todaro, where they observed that a pool of polypeptide growth factors released from mouse fibroblasts transformed with murine sarcoma virus, was able to transform cells in a soft agar assay (De Larco and Todaro, 1978). Using a similar approach on normal rat kidney fibroblasts it was demonstrated that the growth factor cocktail actually consisted of two distinct polypetide growth factors, coined transforming growth factor (TGF)- α and -β (Anzano et al., 1983; Roberts et al., 1981). Whereas TGF-α displayed mitogenic activity, it later became clear that TGF-β served as a potent growth inhibitor in most other cell types (Roberts and Sporn, 1990), rather suggesting a role as an important suppressor of tumourigenesis. TGF-β superfamily of ligands 30 40 50 60 80 100% Identity MIS/AMH TGF-β Activin Nodal BMP-3 Vgr-2 GDF-1 GDF-5 BMP-7 BMP-2/4 Dpp

Figure 1. The ligands of the TGF-β superfamily. Dendrogram indicating the relative level of amino acid sequence similarity between members.

(9)

TGF-β serves as the prototype for the large and still growing TGF-β superfamily, consisting of more than 30 members which include bone morphogenetic proteins (BMPs), activins, inhibins, anti-müllerian hormone (AMH) and growth and differentiation factors (GDFs) (see Figure 1) (reviewed in Massague, 1998).

TGF-β

TGF-β exists in three highly similar isoforms termed TGF-β1 (Derynck et al., 1985), TGF-β2 (de Martin et al., 1987; Madisen et al., 1988), and TGF-β3 (Derynck et al., 1988; ten Dijke et al., 1988), all encoded by distinct genes. Besides forming homodimers, it has been reported that heterodimers can form between TGF-β1 and TGF-β2, and between TGF-β2 and TGF-β3 (Cheifetz et al., 1987; Ogawa et al., 1992). The functions of TGF-β ligands include cell cycle arrest in epithelial and haematopoietic cells and control of cell proliferation and differentiation in mesenchymal cells. They are also strong inducers of extracellular matrix production and are involved in wound healing and immunosuppression, (Massague, 1990; Roberts and Sporn, 1990; Roberts and Sporn, 1993). The three TGF-β isoforms affect TGF-β signalling in a rather similar and redundant way in vitro, but display different in vivo expression patterns and functions (reviewed in (Roberts and Sporn, 1992). Gene analysis also revealed that each TGF-β isoform is controlled by a unique and differently regulated promoter (reviewed in Roberts et al., 1991).

Activin/Inhibin

Members of the activin subfamily can induce pituitary follicle-stimulating hormone (FSH) production. Other functions include erythroid cell differentiation and induction of mesoderm as shown in Xenopous. The different members of the activins can form homo- or hetero-dimers between different β-subunits: βA, βB, βC and βE (Gaddy-Kurten et al., 1995; Harland, 1994; Vale et al., 1990).

Inhibin consists of a distantly related α-subunit, which can form hetero-dimers with the different activin β-subunits. Inhibin can act as a counterplayer of activin, and inhibit FSH production as well as other effects of activin (Gaddy-Kurten et al., 1995; Vale et al., 1990).

BMP

Bone morphogenetic proteins (BMPs) constitute the largest group within the TGF-β

superfamily of growth and differentiation factors, with over 20 members. These can be further subdivided into different subfamilies based on structural homology and physiological effects.

Spencer et al. observed in 1982 that flies mutant for the Drosophila gene

decaplentaplegic (dpp) displayed a variety of pattern deficiencies and structure

(10)

information within developing epidermal tissue (Spencer et al., 1982). In a search for factors able to induce ectopical bone and cartilage formation, Wozney et al. identified two mammalian homologues of dpp, BMP2 and BMP4 (Wozney et al., 1988). BMP4 and dpp was later shown to be able to functionally substitute for each other in flies and mammals, suggesting a high degree of conservation between structure and function (Padgett et al., 1993; Sampath et al., 1993).

The BMP2 subfamily consists of BMP2, BMP4 and DPP (a BMP2/4 homologue in

Drosophila). This subfamily affects gastrulation, neurogenesis, and interdigital

apoptosis in mammalians. In Xenopus the BMPs influence patterning of the mesoderm and in Drosophila they affect dorsalization and eye and wing development (Harland, 1994; Hogan, 1996; Mehler et al., 1997). The second subfamily, BMP5, includes members like BMP5, 60A (a BMP5 homologue in Drosophila), BMP6/Vgr1, BMP7/OP1, and BMP8/OP2. Together with BMP2 and BMP4 this subfamily is involved in the development of almost every organ and plays many roles in neuronal development (Hogan, 1996; Mehler et al., 1997).

The BMP5 subfamily members BMP5 (Celeste et al., 1990), BMP7/OP1 (Celeste et al., 1990; Özkaynak et al., 1990), and BMP8/OP2 (Özkaynak et al., 1992), were all identified as bone-inducing proteins. In a search for mammalian homologues for Vg1, previously discovered in Xenopus oocytes (Weeks and Melton, 1987), BMP6/Vgr1 was identified (Lyons et al., 1989). Wharton et al. isolated the Drosophila gene 60A while searching for homologues of TGF-β (Wharton et al., 1991).

Other members

Growth and differentiation factors (GDFs) belonging to the GDF5 subfamily have been reported to affect chondrogenesis in the developing limbs (Hogan, 1996; Kingsley, 1994) while Vg1, belonging to the Vg1 subfamily, affects axial mesoderm induction in frog and fish (Kingsley, 1994).

Members of the BMP3 subfamily, such as BMP3/osteogenein and GDF10 have been reported to influence osteogenic differentiation, endochondral bone formation and monocyte chemotaxis (Cunningham et al., 1992).

Nodal, one of the intermediate members has been shown to be involved in axial mesoderm induction and left-right asymmetry (Beddington, 1996; Hogan, 1996). Other intermediate members, i.e. Dorsalin and GDF8, are involved in regulation of cell differentiation within the neural tube (Basler et al., 1993) and inhibition of skeletal muscle growth (McPherron et al., 1997), respectively.

More distant members of the TGF-β superfamily include anti-Müllerian hormone (AMH) also called Müllerian inhibiting substance (MIS) and glial cell line-derived neurotrophic factor (GDNF). AMH induces Müllerian duct regression, resulting in male genital organs (Cate et al., 1986; Josso et al., 1993). GDNF is a factor that promotes dopaminergic neuron survival and differentiation (Lin et al., 1993), but also affects

(11)

kidney development (Massague, 1996). Surprisingly, GDNF has also been reported to signal via the Ret tyrosine kinase receptor (Massague and Weis-Garcia, 1996).

Control of TGF-β production and activation

TGF-βs are synthesised as inactive dimeric precursors that later become proteolytically processed in the Golgi apparatus by furin, a member of the convertase family of endproteases (Dubois et al., 1995). Furin cleaves the precursor at RXXR sites, located 112-114 amino acids from the C-terminal end resulting in a C-terminal TGF-β part and an N-terminal remnant called latency-associated protein (LAP) (Cui et al., 1998; Munger et al., 1997b). The N-terminal and the mature part of TGF-β, form a still inactive, non-covalent complex called the small latent complex (SLC) which is much more stable than the bioactive form of TGF-β. The process continues in the Golgi by the formation of disulphide bonds between the LAP and latent TGF-β-binding protein (LTBP), resulting in large latent complexes (LLCs). LTBPs serve to enhance stability and secretion of the SLC complex, ensure correct folding of TGF-β, and target the latent TGF-β complex either to the cell surface for activation, or to the extracellular matrix of distinct cells and tissues for storage (reviewed in Munger et al., 1997b; Taipale and Keski-Oja, 1997). An additional function of the LTBPs might be to influence TGF-β to activate integrin signalling. It has been shown that LTBP-1, -2, and -4 have RGD sequences which are integrin binding sites (Munger et al., 1998b; Saharinen et al., 1998). Indeed, it has been shown that large latent TGF-β complexes can associate directly to integrin αVβ1 at the cell surface (Munger et al., 1998b). In addition, it has been shown that the epithelial specific integrin αVβ6 can bind the RGD-motif in LAP, suggesting a cytoskeleton-mediated activation of TGF-β (Munger et al., 1998a).

Final activation of TGF-β is physically controlled by the binding of LAP to mannose-6-phosphate receptors, and by proteases such as plasmin and cathepsin that cleave LAP (Munger et al., 1997a; Taipale and Keski-Oja, 1997). Thrombospondin-1 appears to be another important activator of TGF-β in vivo. It induces a conformational change in LAP, which results in TGF-β activation (Crawford et al., 1998). Interestingly, TGF-β

induces PAI-1 (plasminogen activator inhibitor-1), suggesting some level of self-regulation of the activation process. Another activation candidate is matrix metalloproteinase-9 (MMP-9) which can activate latent TGF-β2 and TGF-β3 in vitro (Yu and Stamenkovic, 2000).

TGF-β superfamily binding proteins

Bioactive TGF-β superfamily members can associate with several extracellular proteins that can modify their activity. Extracellular matrix proteoglycans: decorin and biglycan, have been shown to inhibit TGF-β activity (Yamaguchi et al., 1990). In addition, a 60-kDa protein has been reported to inhibit the interaction between the receptor and TGF-β

ligands (Piek et al., 1997). Furthermore, it has been demonstrated that α 2-macroglobulin acts as a clearance factor for circulating TGF-βs and activins in serum (Mather, 1996).

(12)

BMPs are regulated by soluble factors that play important roles during embryonic development. By competing with BMPs for ligand-receptor interactions, noggin (Zimmerman et al., 1996), chordin (Piccolo et al., 1996), short gastrulation (Sog) a

Drosophila homologue of chordin (Francois and Bier, 1995), Dan and gremlin (Hsu et

al., 1998), modulate differentiation of mesoderm and ectoderm. Follistatin can also inhibit receptor-ligand interactions by interacting directly with activin (Nakamura et al., 1990), as well as with BMPs (Yamashita et al., 1995a). In addition, cerberus has been identified as an antagonist not only for BMP and activin, but also for nodal signalling (Hsu et al., 1998).

TGF-β superfamily receptors

Receptors for polypeptide growth factors are transmembrane proteins that are able to transduce the extracellular message across the plasma membrane into an intracellular signal. The TGF-βs and related factors signal through a group of transmembrane protein serine/threonine kinases known as the TGF-β receptor family.

Type I and II receptors

The TGF-β receptor family is divided into two subfamilies, type I and type II receptors, based on their structural and functional characteristics. The vertebrate type I receptor subfamily forms three subgroups based on similarities in kinase domains and signalling activities. In mammals, one group includes TβR-I, ActR-IB, and ALK7, another includes BMPR-IA and -IB, and a third group includes ALK1 and ALK2 (reviewed in (Massague, 1998)). As a result of being cloned simultaneously by different groups, most type I receptors received different names. Initially the neutral ALK (activin receptor-like kinase) nomenclature was used whereas a more descriptive name was designated after the identification of a physiological ligand. As a consequence, the TGF-β type I receptor initially identified as ALK-5 (Franzen et al., 1993) is now called TβR-I (Yamashita et al., 1994). On a similar note, the activin receptor type I previously known as ALK-4 (ten Dijke et al., 1993) is now called ActR-IB (Carcamo et al., 1994), and ALK-3 and ALK-6 are referred to as BMPR-IA and BMPR-IB, respectively (Koenig et al., 1994; Yamashita et al., 1995b). Type I receptors with no known ligands include mammalian ALK-7 (Rydén et al., 1996; Tsuchida et al., 1996) and the related Xenopus XTrR-I (Mahony and Gurdon, 1995). TGF-β can bind to ALK-1 (Attisano et al., 1993) but more weakly than to TβR-I (ten Dijke et al., 1994a). ALK-1 has been reported mediate a TGF-β response (Attisano et al., 1993). ALK-2 is also known as ActR-I since it can bind to activin and mediate certain activin responses (Attisano et al., 1993; Yamashita et al., 1995a). ActR-I is a bit promiscuous since it also can binds BMP2 and BMP4 (Liu et al., 1995; ten Dijke et al., 1994b). In addition, the ActR-I mouse homologue can when overexpressed bind TGF-β (Ebner et al., 1993; ten Dijke et al., 1994a). It has also been proposed that ALK-2/ActR-I can function as an MIS/AMH type I receptor (He et al., 1993).

(13)

The type II receptor subfamily includes in vertebrates TβR-II, BMPR-II, and AMHR, which bind selectively to TGF-β (Lin et al., 1992), BMPs (Liu et al., 1995; Nohno et al., 1995; Rosenzweig et al., 1995) and MIS, respectively (Baarends et al., 1994; di Clemente et al., 1994). ActR-II and -IIB can together with the BMP type I receptors bind BMP2, BMP4 and BMP7 as well as GDF5 (Hoodless et al., 1996; Nishitoh et al., 1996; Yamashita et al., 1995a). In addition, ActR-II and -IIB can also associate with activins, either alone or in concert with activin type I receptors (Attisano et al., 1992; Mathews and Vale, 1991; Mathews et al., 1992).

The type I receptor contains a unique region, known as the GS-domain (Wrana et al., 1994). It has been shown that ligand-induced phosphorylation of TβR-I by the type II receptor takes place on serines and threonines in a TTSGSGSG sequence in the GS-domain. This is important for receptor activation and signalling (Souchelnytskyi et al., 1996; Wieser et al., 1995; Wrana et al., 1994).

Type III/Endoglin receptors

Ligand crosslinking experiments have identified an additional class of accessory receptors, the type III receptors. These consists of two related proteins, betaglycan and endoglin (Gougos and Letarte, 1990; Wang et al., 1991). Since these receptors lack intrinsic signalling activity they are believed to regulate the TGF-β access to the signalling receptors. Betaglycan has been reported to bind to all three TGF-β ligands with high affinity (Cheifetz and Massagué, 1991; Segarini et al., 1989), and to facilitate ligand binding to the type II receptor (Lopez-Casillas et al., 1993; Wang et al., 1991). In contrast, endoglin has been shown to bind to TGF-β1 and -β3, but not to TGF-β2 (Cheifetz et al., 1992).

TGF-β superfamily intracellular signalling

The signal initiated by the TGF-β superfamily ligands is transduced by type I and type II serine/threonine kinase receptors into the intracellular space. The ligands bind type II receptor, forming a heterodimeric complex which can recruit and activate the type I receptor by phoshorylating serine and threonine residues located primarily in the GS-domain (Souchelnytskyi et al., 1996; Wrana et al., 1994). Ligands such as TGF-β1, TGF-β3, and activins have been shown to associate to the receptors through sequential binding (Massague, 1998). In contrast, TGF-β2 (Rodriguez et al., 1995), as well as BMP2 and BMP7, have affinity for both type I and II receptors, and associate with the receptor complex through co-operative binding (Massague, 1998). The activated receptor can recruit downstream signalling molecules, known as Smads (figure 2).

The Smad family of signal transducers

In a genetic screen looking for enhancers of a weak decapentaplegic (dpp) maternal phenotype in Drosophila, a new gene mad (mothers against dpp) was isolated (Raftery et al., 1995; Sekelsky et al., 1995). This was followed by the discovery of three

(14)

Mad-homologues: sma-2, sma-3 and sma-4 in C. Elegans (Savage et al., 1996). Mutations of these sma genes resembled the small body sized phenotype observed in the Daf4 mutants (type II serine/threonine receptor) of C. Elegans. The vertebrate homologues of the mad- and sma-genes are called smads. The proteins derived from these genes can be divided into three different subclasses (figure 2), i) receptor activated Smads (R-Smads), ii) common mediator Smads (Co-(R-Smads), and iii) inhibitory Smads (I-Smads) depending on their diverse roles in signalling. The N-terminal and C-terminal regions of the R-Smads and the Co-Smads display a great deal of homology and have been designated the Mad homology 1 (MH1) and Mad homology 2 (MH2) domains respectively. A proline-rich linker domain exists between the MH1 and MH2 domain. The R-Smads, but not the Co- or I-Smads, contain a C-terminal SSXS motif which becomes phosphorylated upon receptor interaction (reviewed in Heldin et al., 1997).

ActR-IIA ActR-IIB BMPR-II ActR-IIA ActR-IIB TGF-β Activin BMP TβRII TβRI/ALK-5 ALK-1 ? Smad1 Smad5 Smad8 ALK-4/ActR-IB Smad2 Smad3 ALK-2/ActR-I ALK-3/BMPR-IA ALK-6/BPMR-IB Smad2 Smad3 Smad1 Smad5 Smad8

Smad4 Smad4 Smad4 Smad4

Smad6 Smad7 Smad6 Smad7 Smad6 Smad7 Smad6 Smad7 TGF-β superfamily Type-II R Type-I R R-Smads Co-Smads Inhibitory Smads Biological responses Inhibition of mitogenicity Induction of extracellular matrix

Induction of dorsal mesoderm, erythroid differentiation, and FSH-release

Induction of ventral mesoderm, cartilage and bone, and apoptosis

Figure 2. TGF-β superfamily members, their signalling molecules and some biological responses. Adapted from Heldin et al., 1997.

The R-Smads Smad2 and Smad3 mediate signals from TGF-β and activin ligands through the TβR-I/Alk-5 and ActR-IB receptors, respectively (see Figure 2) (Eppert et al., 1996; Macias-Silva et al., 1996; Zhang et al., 1996). BMP signalling is mediated through R-Smads 1, 5 and 8 which become phosphorylated and activated by the ActR-I, BMPR-IA or BMPR-IB receptors (Thomsen, 1996). The determinant of specificity

(15)

between the R-Smads and their interaction with either TGF-β/activin or BMP receptors, is the L3 loop region within the MH2 domain (Lo et al., 1998). However, some reports suggest that Smad1, 5, and 8 might be promiscous towards the TGF-β receptors as well (Lux et al., 1999; Macias-Silva et al., 1998; Oh et al., 2000). The Co-Smad, Smad4 (also known as DPC4, deleted in pancreatic carcinomas), appears to play a critical role in both BMP- and TGF-β/activin-mediated pathways.

The affinity of Smad4 for R-Smads can be increased through phosphorylation of the C-terminal part of R-Smads leading to the formation of a complex (Souchelnytskyi et al., 1997). Moreover, Smad4 has a unique Smad activation domain (SAD) in the linker region, which governs transcriptional activation via the co-activator p300 (de Caestecker et al., 2000). One Co-Smad has been identified in mammals, but there might be others. Two Co-Smads have been found in Xenopus laevis (Howell et al., 1999; Masuyama et al., 1999).

The Smad pathway

Upon receptor activation Smad molecules can be recruited to the receptor complex (figure 3). It has been shown that microtubules play an important role in the guiding of Smads to the plasma membrane (Dong et al., 2000), where a FYVE-domain protein, Smad anchor for receptor activation (SARA), presents the R-Smads to the receptor (Tsukazaki et al., 1998). The FYVE-domain is a protein structure that in other proteins has been associated with anchoring to endosomes (Tsukazaki et al., 1998). This has raised the possibility that the receptor needs to be internalised before it can bind the SARA-sequestered Smad proteins. Besides restraining the movement of Smads it seems as if SARA also masks a region of Smad2 that otherwise mediates nuclear import (Xu et al., 2000). While receptor-mediated phosphorylation and activation of R-Smads increases the affinity for Smad4 it decreases the affinity for SARA at the same time (Xu et al., 2000).

The R-Smads are presented to the receptor complex and become activated by phosphorylation on the C-terminal SSXS motif (Heldin et al., 1997). The L45 loop, a distinct region in the type I receptor containing two β strands (number 4 and 5) flanking the kinase domain, has been reported to determine the association of Smad molecules to the receptor (Feng and Derynck, 1997). To underscore the importance of the L45 loop in determination of specificity, it has been shown that swapping of the L45 loop of TβR-I and BMPR-IB also alters Smad1 and Smad2 recognition, subsequently leading to switched transcriptional response (Chen et al., 1998; Persson et al., 1998). After receptor activation, the R-Smads oligomerise with the Co-Smad, forming a heteromeric complex that is translocated to the nucleus where it regulates transcription of target genes (Heldin et al., 1997). Smad3 and Smad4 have been shown to interact directly with specific DNA sequences via their MH1 domain (Dennler et al., 1998; Vindevoghel et al., 1998; Yingling et al., 1997). In order to fully activate transcription of the target promoters, the Smad complexes must recruit additional factors, like the transcription factor components AP-1 (Liberati et al., 1999), DNA-binding adaptors like FAST-1 (Chen et al., 1996), or co-activators such as CBP/p300 (Feng et al., 1998; Janknecht et

(16)

al., 1998; Nishihara et al., 1998; Shen et al., 1998; Topper et al., 1998). This transactivating role of the Smad proteins has been ascribed to the MH2 domain (reviewed by Massagué and Wotton, 2000).

R-SMAD Co-SMAD Nucleus Ras Erk SNIP1 COREPRESSORS TGIF Ski/SnoN TGF-β CO-ACTIVATORS P300 CBP SIGNALLING RECEPTORS ACCESSORY RECEPTORS II I II I LIGANDS DNA-BINDING COFACTORS Betaglycan Endoglin Crypto LEF1/TCF CBFA1 CBFA3 JUN β-catenin SMAD1 SMAD2 JNK BMP TGF-β TNF-α IFN-γ EGF SMAD1 SMAD2 NFκB STAT Co-SMAD R-SMAD COF A CT OR BAMBI SMURF1 SMAD7 SMAD6 Wnt BMP TGF-β TNF-α TGF-β superfamily

Figure 3. Schematic overview of the transforming growth factor-β (TGF-β) signalling pathway. Adapted from Massagué et al., 2000.

To date, two I-Smads have been identified in mammals, Smad6 and Smad7 (Imamura et al., 1997; Nakao et al., 1997; Topper et al., 1997). They have been characterised as inhibitors of TGF-β/activin and BMP signalling and have been proposed to function in negative feed-back loops, since their expression is induced by TGF-β/activin and BMP-members (Christian and Nakayama, 1999). It has been shown that I-Smads can interact stably with the type I receptor and block further activation of R-Smads (Imamura et al., 1997; Nakao et al., 1997; Souchelnytskyi et al., 1998). For Smad6 an additional mechanism has been suggested, where Smad6 competes with Smad4 for binding to Smad1, thereby preventing the formation of a functional heteromeric Smad1/Smad4 complex (Hata et al., 1998). Smad7 is considered as a general inhibitor of TGF-β

superfamily-induced responses, whereas Smad6 is thought to preferentially block BMP-mediated signalling (Itoh et al., 1998), although this is controversial (Imamura et al., 1997).

(17)

Self-regulation of the TGF-β pathway through negative feedback has been reported to occur through other mechanisms than antagonistic I-Smads. BMP has been shown to exert a negative feedback control via the protein BAMBI (BMP and Activin membrane-bound inhibitor) which is a truncated and kinase-deficient type I receptor that interferes with activation by binding to the BMP receptors (Onichtchouk et al., 1999).

Smads are disposed from the cytoplasm and the nucleus by ubiquitination and proteasomal degradation. The cytoplasmic Smad1 and Smad2 become targeted by proteolysis via the ubiquitin-ligase proteins Smurf1 and Smurf2, respectively (Lin et al., 2000; Zhu et al., 1999). The nuclear Smad2 on the other hand, is degraded via the UbcH5 family of ubiquitin-conjugating proteins (Lo and Massagué, 1999).

Non-Smad pathways

To this date, only Smads are recognised as TGF-β receptor substrates and signal transducers. However, numerous reports indicate that TGF-β and BMP are also able to transduce signals via the mitogen-activated protein kinase (MAPK) pathway (figure 4) (reviewed by Massagué and Wotton, 2000) TGF-β has been reported to activate the Jun N-terminal kinase (JNK) via MKK4 (MAPK kinase 4) in a rapid fashion in a fibrosarcoma cell line (Hocevar et al., 1999). In addition, p38 seems to be activated by TGF-β in lung and kidney epithelial cell lines via MKK3 (Hanafusa et al., 1999) leading to activation of the nuclear target ATF2 (activating transcription factor 2) (Sano et al., 1999). The biochemical link between the TGF-β receptor and the MKKs is to this date still unknown. One candidate could be TAK1 (TGF-β activated kinase), a member of the MAPK kinase kinase (MAPKKK) family, implicated in activation of p38 via MKK3/MKK6 (Moriguchi et al., 1996) and activation of JNKs via MKK4 (Shirakabe et al., 1997). JNK signalling by TAK1 is also reported during Drosophila development (Takatsu et al., 2000). Other factors such as TAB1 and XIAP might be located further upstream in the signalling pathway affecting TAK1.

TAB1 is one of the proposed activators of TAK1 (Shibuya et al., 1996). Also, since injection of TAB1- and TAK1-mRNA into the dorsal marginal zone of Xenopus induced ventral mesoderm, TAB1 and TAK1 are suggested to play a functional role in BMP signalling during early Xenopus development (Shibuya et al., 1998).

XIAP, a positive regulator of BMP signalling, has been proposed as a link between the BMP receptors and TAB1-TAK1. XIAP can associate not only with TAB1, but also with BMP-receptors in mammalian cells. Furthermore injection of XIAP mRNA into dorsal blastomeres enhanced the ventralisation of Xenopus embryos in a TAB1-TAK1 dependent manner (Yamaguchi et al., 1999).

(18)

GRB2 II I Plasma membrane TGF-β TGF-β receptor Growth factor Growth factor receptor P P SOS Nuclear membrane Ubiquitination and degradation R-SMAD Co-SMAD R-SMAD CO F A CT O R

Fos Jun AFT2

TGIF DNA P MKK4 MKK3 JNK p38 RAS MEK P XIAP TAK1 Cytokine receptor Cytokine ERK

Figure 4. Cross-talk between Smads and the MAP kinase pathway. Adapted from

Massagué et al., 2000.

Cross-talk between Smads and the MAP kinase pathway

The Erk-activated Ras pathway has been reported to modify the TGF-β signalling pathways at different levels (figure 4). It has been shown that a hyperactive Ras pathway can downregulate TGF-β receptors in H-Ras transformed rat intestinal epithelial cells (Zhao and Buick, 1995). Moreover, it has been reported that oncogenic Ras can inhibit BMP and TGF-β signalling negatively by decreasing the accumulation of R-Smads in the nucleus. It was observed that EGF and HGF could phosphorylate Smad1, Smad2 and Smad3 on MAPK and Erk consensus sites located in the linker region, which led to a retention of R-Smads in the cytoplasm and reduced transcriptional activity (Kretzschmar et al., 1997a; Kretzschmar et al., 1999; Kretzschmar et al., 1997b). When mutations were introduced into the Erk phosphorylation sites of Smad3, it resulted in a Ras-resistant form that could rescue the growth inhibitory response of TGF-β in Ras-transformed cells. Moreover, EGF induced a less extensive phosphorylation and cytoplasmic retention of Smad2 and Smad3, as compared to oncogenic H-Ras, which might explain the silencing of anti-mitogenic TGF-β functions by hyperactive Ras in cancer cells (Kretzschmar et al., 1999).

The Erk/MAP kinase pathway has also been implicated in regulation of Smad mediated transcription at other levels than Smad retention. It has been reported that the

(19)

transcriptional co-repressor TGIF becomes upregulated in response to Erk signalling (Lo et al., 2001).

Specificity in Smad signalling

How can specificity in Smad signal transduction be achieved? It has been shown that the different branches of TGF-β/activin and BMP signalling can access distinct sets of target genes, via their respective down-stream modulators such as Smad1 and Smad 2, (Chen et al., 1998). An example of the opposing and somewhat complementary roles between Smad1- and Smad2-dependent pathways, is demonstrated in Xenopous laevis embryogenesis. Smad2 can, in response to Nodal related factors (Xnr-1, Xnr-2 and Vg1), activate genes responsible for induction of dorsal mesoderm. In contrast Smad1, in response to BMP4, activates a different set of genes that affect induction of ventral mesoderm and suppression of neural fates. Ectopic expression of Smad1 or Smad2 induces of ventral and dorsal structures, respectively (Baker and Harland, 1996; Graff et al., 1996). In a similar complementary fashion, BMP and Nodal dictate the specification of left-right symmetry in the developing vertebrate embryo (Rodriguez Esteban et al., 1999; Saijoh et al., 2000; Yokouchi et al., 1999).

As transcription factors, R- and Co-Smad complexes are able to recognise and interact with distinct DNA sequences, e.g. CAGAC, GTCTAGAC, and other GC rich sequences. This interaction is rather weak which suggests that the Smads need the assistance of other proteins for correct binding (Shi et al., 1998). Since the pentameric CAGAC DNA-binding motif statistically occurs rather frequently in the genome, a high affinity of the Smads to their recognition sequence would result in a rather general and unspecific association with DNA. Other factors are therefore required for increased specificity and target selection such as co-factors (figure 3). The co-factors are structurally diverse proteins that share the ability to associate with Smad molecules and a neighbouring DNA sequence (reviewed (Massagué and Wotton, 2000)). Cell type specific responses could be explained by the fact that certain co-factor combinations are expressed in distinct cells or tissues (Chen et al., 1997; Hata et al., 2000).

One group of co-factors can be described as adaptors for the DNA-Smad interaction. To achieve DNA binding Smad1 has been shown to bind to OAZ (olf-associated zinc finger) (Hata et al., 2000), while Smad2 binds FAST (forkhead activin signal transducer) (Chen et al., 1996) and Mixer (Germain et al., 2000). The adaptor proteins OAZ, FAST, and Mixer lack intrinsic transcriptional activity. The recognition between the Smad molecule and the adaptor protein requires a bulging alpha-helix 2 region of the MH2 domain in the Smad proteins and a Smad-interaction domain, which is conserved in FAST and Mixer but not in OAZ (Chen et al., 1998; Germain et al., 2000; Hata et al., 2000).

Transcription factors are another group of proteins that can form functional complexes with Smad molecules (figure 3). Examples are JunB (Zhang et al., 1998), TFE3 (transcription factor binding to immunoglobulin heavy constant mu enhancer 3) (Hua et al., 1999), core-binding factor A/acute myeloneus leukaemia (CBFA/AML) proteins

(20)

(Hanai et al., 1999; Pardali et al., 2000; Tsuji et al., 1998) and lymphoid enhancer-binding factor 1/T-cell specific factor (LEF1/TCF) (Labbe et al., 2000; Nishita et al., 2000). LEF1/TCF is the mediator of WNT/β-catenin signalling and is able to co-operate with Smads in the activation of Xtwn (Xenopous twin) in response to Nodal related signals (Nishita et al., 2000).

Smads are able to recruit transcriptional repressors as well as activators (figure 3). Smads can associate with repressors such as TG3-interacting factor (TGIF) (Wotton et al., 1999), Sloan-Kettering Institute proto-oncogene (Ski) (Luo et al., 1999) and Ski-related novel gene N (SnoN) (Sun et al., 1999). The repressors can bind histone deacetylases (HDAC) which are generally implicated in chromatin condensation and transcriptional silencing. The HDAC binding would counteract the effect of histone acetyltransferase (HAT) activity associated with the co-activators (figure 3) CBP and p300, and transcriptional activation (Massagué and Wotton, 2000). The relative levels of repressor versus co-activator could determine the final outcome of transcription. For example, it has been reported that the repressor TGIF can associate with Smad2 and Smad3 in competition with the co-activator p300 (Wotton et al., 1999).

Genetic targeting

The functional importance of TGF-β superfamily members in embryonic patterning and tissue homeostasis has been assessed by gene ablation experiments in mice (Table 1-3). They revealed a plethora of effects that range from defective vasculogenesis, inflammation, autoimmunity, and skeletal abnormalities to early malformations affecting egg cylinder formation, gastrulation and mesoderm formation. Interestingly, Smad3 deficient mice developed metastatic colorectal cancer at 4-6 months of age (Table 3) (Zhu et al., 1998), highlighting the potential role of TGF-β as a potent tumour suppressor. This is a complex issue, since other groups have reported phenotypical variations for the same targeted genes, perhaps revealing the importance of genetic strain background in determining the phenotypic outcome of specific null mutants.

(21)

Table 1. Major defects in TGF-β superfamily ligand-deficient mice. Adapted from Goumans and Mummery, 2000.

Targeted gene Phenotype References

TGF-β1

TGF-β2

TGF-β3

- Defective yolk sac vasculogenesis and haematopoiesis. Embryonic lethal (E9.5-11.5)

- Inflammation and autoimmunity - Cardiac, lung, craniofacial, limb, spinal column, eye, inner ear, urogenital defects. Perinatal lethality.

- Cleft palate, delayed lung maturation. Perinatal lethality.

(Dickson et al., 1995) (Shull et al., 1992) (Kulkarni et al., 1993) (Letterio and Roberts, 1996) (Sanford et al., 1997) (Proetzel et al., 1995) (Kaartinen et al., 1995) BMP-2 BMP-4 BMP-7

- Embryonic lethal (E7.5-10.5) - Failure of proamniotic canal to close, heart malformation.

- Embryonic lethal (E7.5-9.5)

- 1. Arrest at egg cylinder stage, lack of mesoderm.

- 2. Develop until early somite stage w. disorganised/truncated posterior structures and reduced extra-embryonic mesoderm.

- Defects in primordial germ cells and allantois formation.

- Skull, eye and kidney defects. - Perinatal lethality

(Zhang and Bradley, 1996) (Winnier et al., 1995) (Lawson et al., 1999) (Dudley et al., 1995) (Luo et al., 1995) Activin βA Activin βB Activin βA/βB

- Cleft pallet, lack of whiskers and incisors.

- Failure of eyelid fusion. Females show impaired reproductive ability.

- No additional defects.

(Matzuk, 1995) (Schrewe et al., 1994) (Vassalli et al., 1994) (Matzuk, 1995) Nodal - Embryonic lethal (E7.5). Failure in

gastrulation and primitive streak formation

(22)

Table 2. Major defects in TGF-β superfamily receptor-deficient mice. Adapted from Goumans and Mummery, 2000.

ALK-1 ALK-2 ALK-3 ALK-4 ALK-5 ALK-6

- Embryonic lethal (E10.5-11.5). - Defects in angiogenesis and vascular smooth muscle cell differentiation. - Embryonic lethal (E7.5-9.5)

- Failure in primitive streak elongation, delayed mesoderm formation and malformed visceral endoderm. - Embryonic lethal (E7.5-9.5) - Fail to form mesoderm, reduced proliferation of the epiblast. - Embryonic lethal (E7.5-9.5)

- Defect in epiblast and extraembryonic ectoderm organisation and gastrulation. - Embryonic lethal around E 10.5. - Severe defects in vascular development of placenta and yolk sac.

- Intact haematopoietic potential of precursors.

- Defects in endothelial cell proliferation, migration and fibronectin production. - Defects in digit formation and fore- and hindlimb development (Oh et al., 2000) (Gu et al., 1999) (Mishina et al., 1995) (Gu et al., 1998) (Larsson et al., 2001) (Yi et al., 2000) (Baur et al., 2000) TβRII ActR-IIA ActR-IIB ActR-IIA/ ActR-IIb

- Embryonic lethal (E10)

- Defective yolk sac vasculogenesis. - Skeletal and facial abnormalities in percentage of mice.

- Perinatal lethality

- Cardiac defects associated with defects in left-right asymmetry

- Homeotic transformation of the skeleton

- Embryonic lethal

- Defect in primitive streak formation and gastrulation.

(Oshima et al., 1996) (Matzuk, 1995) (Oh and Li, 1997)

(Song et al., 1999) Endoglin -Embryonic lethal (E10.5-11.5).

-Defective yolk sac vasculogenesis, embryonic angiogenesis and vascular smooth muscle cell development. - Cardiac malformations.

(Arthur et al., 1999) (Li et al., 1999) (Bourdeau et al., 1999)

(23)

Table 3. Major defects in Smad-deficient mice. Adapted from Goumans and Mummery,

2000.

Smad1 Smad2 Smad3 Smad4 Smad5 Smad6

-Embryonic lethal (E9.5)

- Failure in establishing chorion-allantoic circulation.

- Embryonic lethal (E7.5-8.5) - Failure in egg cylinder elongation, gastrulation and mesoderm formation. - Metastatic colorectal cancer (4-6 months of age)

- Impaired immunity and chronic infection.

- Accelerated wound healing. - Embryonic lethal (E7.5-8.5) - Growth retardation, no mesoderm formation, abnormal visceral endoderm. - Embryonic lethal (E9.5-10.5)

- Defect in angiogenesis, left/right asymmetry, craniofascial abnormalities and induced mesenchymal apoptosis. - Cardiovascular abnormalities. - Defect in endocardial cushion transformation

(Lechleider et al., 2001)

(Weinstein et al., 1998) (Nomura and Li, 1998) (Waldrip et al., 1998) (Zhu et al., 1998) (Datto et al., 1999) (Yang et al., 1999b) (Ashcroft et al., 1999) (Yang et al., 1998) (Sirard et al., 1998) (Chang et al., 1999),(Chang et al., 2000)

(Yang et al., 1999a) (Galvin et al., 2000)

SMAD7

Two research groups, using two different experimental approaches, discovered Smad7. In a screening of Smad homologues in mouse EST and human cDNA libraries, Nakao et al. identified a protein of 426 amino acids and termed it Smad7 (Nakao et al., 1997). Topper et al. using a differential display approach identified Smad6 and Smad7 as novel genes upregulated in response to laminar vascular flow-stress in human endothelial cells (Topper et al., 1997).

(24)

Smad7 signalling

Smad7 has been proposed to function as an antagonist of TGF-β signalling. Nakao et al. observed that Smad7 could block responses initiated by TGF-β1 in Mv1Lu- and HaCaT cells. They also demonstrated that injection of Smad7 mRNA in a Xenopous embryo blocked the activin/TGF-β signalling. Moreover, ectopic expression of Smad7 resulted in a failure to form head and tail structures and showed a negative effect on the formation of mesodermal derivatives such as muscle and notochord (Nakao et al., 1997). This mimicked the effect of a dominant negative form of the activin receptor (Hemmati-Brivanlou and Melton, 1992), or dominant negative Smad4 (Lagna et al., 1996). The expression of Brachyury, a mesodermal marker gene, was also downregulated by Smad7 injection in the blastomere of two-cell Xenopus embryos (Nakao et al., 1997). Taken together, these data underscored the possible function of Smad7 as a possible attenuator of TGF-β/activin- mediated signalling, both in vivo and

in vitro.

Based on the observations that TGF-β mediated phosphorylation of Smad2 and Smad3 is blocked in Smad7-transfected cells and that overexpressed Smad7 associates with the TGF-β type I receptor in COS cells, a molecular mechanism was proposed stipulating that the inhibitory effect resides in the ability of Smad7 to compete with R-Smads for the type I receptor (Hayashi et al., 1997; Nakao et al., 1997).

An additional Smad7 function was proposed by Kavsak et al. where Smad7 can target the TGF-β receptor for proteasomal and lysosomal degradation, through the interaction with an ubiquitin ligase, Smurf2. In addition, they also observed that the association between Smurf2 and Smad7 induced a translocation of Smurf2 from nucleus to the activated TGF-β receptor. This Smurf2/Smad7-complex formation and TGF-β receptor turnover could be further enhanced by interferon-γ (IFN-γ) (Kavsak et al., 2000). On a similar note, Ebisawa et al. observed that Smurf1, an E3 ubiquitin ligase for BMP-specific Smads, also could bind Smad7 and thereby induce TGF-β receptor degradation (Ebisawa et al., 2001). Moreover, it has been demonstrated that STRAP, a WD40 repeat protein could associate with both TβR-I and TβR-II (Datta et al., 1998). This stabilised the interaction between Smad7 and the activated receptor, thereby assisting in Smad7 inhibition of TGF-β signalling (Datta et al., 1998; Datta and Moses, 2000).

Based on observations that Smad7 mRNA expression could be induced by TGF-β

stimulation, Smad7 has been proposed to take part in a negative feedback loop downregulating the TGF-β signal (Nakao et al., 1997). Additional pathways have been shown to upregulate Smad7 transcription. Bitzer et al. reported that the mRNA expression of Smad7 was upregulated via the NF-κB/RelA pathway by the pro-inflammatory molecules tumour necrosis factor-α and interleukin 1β (Bitzer et al., 2000). In addition, it was demonstrated that IFN-γ also could upregulate Smad7 transcription via Jak1/Stat1 (Ulloa et al., 1999).

(25)

I-Smads have conserved C-terminal Mad homology 2 (MH2) domains, whereas the amino acid sequences of their N-terminal regions (N-domain) are highly divergent, not only between the Smad families but also between the I-Smads of different species. (Christian and Nakayama, 1999; Nakayama et al., 2000). Both the N- and C-terminal MH2-domains are important to obtain full activity of Smad6 and Smad7 in Xenopus (Nakayama et al., 2001). The N-domain and MH2-domain can bind to each other, which facilitates the interaction between Smad7 and TGF-β receptors and thereby enhances the inhibitory effect of Smad7/MH2-complex (Hanyu et al., 2001).

Smad7 has been suggested to exert other functional roles distinct from the antagonistic effect in receptor-mediated Smad activation. It has been reported that overexpressed Smad7 is predominantly located to the nucleus in the absence of ligand and becomes translocated to the cytoplasm in response to TGF-β stimulation (Itoh et al., 1998). Moreover, Pulaski et al., observed that phosphorylation of Smad7 at serine residue 249 could affect Smad7-dependent transcriptional activation on an SV40 minimal promoter (using the GAL4 DNA binding domain fused to Smad7) (Pulaski et al., 2001). In addition, it has been proposed that I-Smads can act as transcriptional co-repressors by recruiting histone deacetylases (HDACs) (Bai and Cao, 2002; Bai et al., 2000).

Smad7 in physiological and pathological conditions

Vasculogenesis

The developing vascular system of the early embryo originates from a small population of mesodermal endothelial precursor cells. These cells join to form a capillary plexus that is gradually remodelled by the development of distinct arterial-venous parts and the recruitment of mural cells such as capillary-associated pericytes and vascular smooth muscle cells (VSMC). These structures constitute the basis for future angiogenesis using existing vessels to form new ones (Daniel and Abrahamson, 2000; Hungerford and Little, 1999; Risau, 1997; Yancopoulos et al., 2000). A multitude of TGF-β

superfamily ligands, receptors and signalling molecules have been implicated in the development of the vascular system (Gatherer et al., 1990; Pelton et al., 1990; Pelton et al., 1989; Roberts and Sporn, 1992; Schmid et al., 1991). Several in vitro studies suggest that TGF-β can regulate the growth of endothelial cells as well as influence their migration and fusion into capillary tubes. In addition, TGF-β can also affect vessel lumen size (Gajdusek et al., 1993; Madri et al., 1988; Merwin et al., 1990; Pepper, 1997; Pepper et al., 1990; Pepper et al., 1993; Roberts and Sporn, 1989). However, the behaviour of the endothelial cell in response to TGF-β stimulation, varies greatly between experimental set ups, making functional assessments difficult (Daniel and Abrahamson, 2000; Klagsbrun and D'Amore, 1991). A different approach used to investigate the function of individual genes in vasculogenesis in vivo is gene targeting. Such studies have revealed that activin receptor-like kinase 1 (ALK1) a TGF-β/activin binding receptor, endoglin (Eng) a TGF-β superfamily co-receptor, and Smad5 are examples of proteins necessary for normal angiogenesis. Null mutants lacking either of these three gene functions display fragile, haemorrhagic and dilated vessels. In addition,

(26)

extracellular matrix remodelling enzymes have abnormal expression levels and the recruitment of vascular smooth muscle cells (VSMC) to the vascular endothelium is impaired (Chang et al., 1999; Li et al., 1999; Oh et al., 2000; Pepper, 1997; Urness et al., 2000; Yang et al., 1999b). Moreover, ALK1 null mutant mice display an arteriovenous malformation phenotype, characterised by the development of shunts between the arterial and venous circulatory systems (Urness et al., 2000). Interestingly, these phenotypes resemble hereditary haemorrhagic telangiectasia, a human pathological condition associated to haploinsufficiency of ALK1 or Eng (Shovlin and Letarte, 1999).

What is the role of TGF-β signalling in the developing vascular system? Recent models suggest that the primary function is to promote VSMC recruitment and differentiation while inhibiting endothelial proliferation through VSMC-endothelial interactions. However, non-TGF-β superfamily (e.g. angiopoeitins, PDGFs) signalling mutants also exhibit vessel dilations, fragility and haemorrhage associated with disturbed VSMC-endothelial interactions (Puri et al., 1995; Sato et al., 1995; Suri et al., 1996; Vikkula et al., 1996). It is therefore not clear to what extent loss of VSMCs, vessel dilation and reduced vessel integrity is a direct consequence of impaired TGF-β signalling or if it is a secondary effect caused by other disturbed TGF-β dependent functions (Folkman and D'Amore, 1996; Li et al., 1999; Oh et al., 2000).

Smad6 and Smad7 were originally identified as genes induced in vascular endothelium in response to steady laminar shear stress, a physiologic biomechanical stimulus. This system mimicking the effect of blood flowing past the endothelial cells, suggested an important role in the maintenance of the vascular endothelium in the adult organism. But could these Smads also play a role in the developing endothelium? Zwijsen et al. investigated the expression of Smad7 mRNA in early mouse development by RT-PCR. They found that Smad7 was indeed upregulated in the developing vascular system of the mouse embryo, especially in endothelial cells of larger blood vessels, possibly as a result of a larger blood flow. In addition, they observed a very early Smad 7 expression during preimplantation and gastrulation. Furthermore, overexpression of Smad7 in the mouse zygote inhibited development at the 2-cell stage (Zwijsen et al., 2000).

In an approach that resembled the inhibition of TGF-β signals modulating angiogenesis and vasculogenesis, Vargesson et al. virally misexpressed Smad7 in the developing chick limb and head. They found that the larger vessels became dilated and frequently developed arteriovenous shunts similar to arteriovenous malformations (AVM). Expression of constitutively active BMP receptor could counteract the effect of Smad7 overexpression, suggesting that a BMP-like pathway in contrast to TGF-β signalling was the target of Smad7 inhibition. Moreover, Smad7 overexpressing vessels were not haemorrhagic and had a normal structure. In addition, the dilation of vessels was independent of VSMCs and the recruitment of VSMCs was not affected. Taken together, these findings suggest that the TGF-β pathway regulates vessel calibre and is necessary for correct vessel connectivity in this experimental setting. They also indicate that vessel dilation not necessarily must lead to vessel rupture and haemorrhages. Furthermore, the findings suggest that a VSMC coat is not a prerequisite for vessel

(27)

maintenance and that TGF-β signalling is not involved in VSMC recruitment and differentiation (Vargesson and Laufer, 2001). Another study suggests that Smad7 overexpression can attenuate the growth inhibitory effects of TGF-β on cultured rat smooth muscle cells (Kato et al., 2001).

Lung morphogenesis

TGF-β is involved in the negative regulation of lung growth and development during early lung organogenesis (Warburton and Lee, 1999). It had been shown that receptor regulated Smad2 and Smad3 were required for the TGF-β mediated inhibition of embryonic lung branching morphogenesis and epithelial cell differentiation (Zhao et al., 1998). The role of Smad7 in lung morphogenesis was investigated by introducing Smad7 antisense oligonucleotides in cellular embryonic lung explants. There it was demonstrated that TGF-β mediated inhibition of lung branching was significantly increased in cells with abrogated Smad7 gene expression. In addition, it was observed by immunohistochemistry that Smad7 together with Smad2 and Smad3 co-localised in distal bronchial epithelial cells (Zhao et al., 2000a). A different approach to assess the role of Smad7 in lung morphogenesis, used adenoviral Smad7 overexpression. The observation was that Smad7, but not Smad6, could abolish TGF-β mediated branching inhibition. In addition, Smad7 also inhibited TGF-β induced down regulation of surfactant protein C, a bronchial epithelial differentiation marker (Zhao et al., 2000b).

Apoptosis

The idea that Smad7 works as a modulator and attenuator of TGF-β mediated signalling has gained acceptance over time. The fact that Smad7 can inhibit TGF-β signals also suggested that Smad7 might inhibit apoptosis initiated by TGF-β. Indeed, studies in mouse B- and T-cells, supported this theory and suggested that Smad7 could act as an inhibitor of activin-induced growth arrest and apoptosis (Ishisaki et al., 1998). Moreover, studies in WEHI 231 B-lymphocytes further corroborated this notion, showing that Smad7 could protect these cells from TGF-β induced apoptosis (Patil et al., 2000). Other reports suggested that the picture is much more complicated. Some studies indicated that Smad7 by itself could work as an inducer of apoptosis. Transgenic mice overexpressing TGF-β1 under the control of the albumin promoter displayed increased apoptosis associated with a depletion of podocytes in progressive glomerulosclerosis (Schiffer et al., 2001). TGF-β1 and Smad7 both seemed able to induce apoptosis, but through different pathways. The TGF-β mediated effect required the activation of p38 MAP kinase and caspase-3, which was not required for Smad7-mediated apoptosis. In contrast to TGF-β, Smad7 was able to inhibit the nuclear translocation and transcriptional activity of the cell survival-promoting factor, NF-kB (Schiffer et al., 2001). Additional evidence in prostatic carcinoma cells (PC-3U) suggested that Smad7 antisense RNA could inhibit TGF-β induced apoptosis (Landström et al., 2000). Moreover, overexpressing Smad7 alone using an inducible promoter also resulted in apoptosis, implicating Smad7 itself as an effector of apoptosis (Landström et al., 2000).

(28)

What signalling pathway could serve as the transducer of this Smad7 mediated apoptosis? One candidate is the c-Jun N-terminal kinase (JNK) cascade. Mazars et al. demonstrated that expression of Smad7 caused a strong and sustained activation of JNK. The use of a dominant-interfering mutant of mitogen-activated protein kinase kinase 4 (MKK4) completely abolished the Smad7-induced activation of JNK. Furthermore, expression of the mutant MKK4 also blocked the ability of Smad7 to promote cell death (Mazars et al., 2001).

Fibrosis

TGF-β is a potent regulator of extracellular matrix (ECM) by inducing proteins such as collagen, fibronectin but also PAI-1 and TIMP. It is believed that a sustained overproduction of TGF-β caused by repeated chemical and or biological injury can result in pathological amounts of ECM being collected in the tissue leading to a functional deterioration (Border and Noble, 1994). TGF-β has been implicated in many fibrotic disorders such as idiopathic pulmonary fibrosis (Broekelmann et al., 1991), autoimmune lung diseases (Deguchi, 1992), and bleomycin-induced lung fibrosis in animal models (Khalil et al., 1989; Khalil et al., 1993; Westergren-Thorsson et al., 1993; Zhang et al., 1995). TGF-β was found to be necessary for the bleomycin-induced tissue fibrosis in rodents (Giri et al., 1993). Based on these findings, Nakao et al. introduced Smad7 and Smad6 by adenoviral transient gene transfer into bleomycin treated mice. They observed that Smad7, but not Smad6, was able to reduce expression of type I precollagen mRNA and abolish the morphological fibrotic responses in transgenic mice. In addition, the phospho-Smad2 immunoreactivity was reduced in bleomycin treated Smad7 transgenic mice, suggesting inhibition of the TGF-β/Smad pathway (Nakao et al., 1999).

Immunological disorders

TGF-β is a powerful modulator of immune responses. It affects the differentiation, proliferation and state of activation of all immune cells. The dysregulation of TGF-β

signalling has been implicated in immune abnormalities related to autoimmunity, opportunistic infections and fibrotic complications (Letterio and Roberts, 1998). Studies in murine models clearly underscore the connection between disrupted TGF-β

signalling and inflammatory disease. Deletion of the TGF-β1 gene in mice results in systemic inflammation and early death (Shull et al., 1992). In addition, overexpression of a dominant negative TGF-β type II receptor gives rise to CD4+ T-cell hyperactivity and autoimmunity (Gorelik and Flavell, 2000). On a similar note, targeted disruption of the Smad3 gene resulted in inflammation of mucosal surfaces (Yang et al., 1999c). Given the strong immunosuppressive role of TGF-β1, what would the effect be if a TGF-β signalling inhibitor like Smad7 was overexpressed in immune cells? Nakao et al. studied the effect of Smad7 overexpression in peripheral T-cells in transgenic mice. The mutant T-cells were no longer growth inhibited by TGF-β but no overt phenotype was observed in the unchallenged transgenic mice. However, when subjected to

(29)

antigen-induced airway inflammation the transgenic mice displayed an increased airway reactivity and inflammation (Nakao et al., 2000). Despite the fact that this was an example of a highly artificial system it can still give us some idea how Smad7 could influence TGF-β signalling in vivo. Interestingly, a human disorder exists where the expression levels of Smad7 are increased. Monteleone et al. (2001) demonstrated that the Smad7 expression was highly upregulated in the colon mucosa and in T-cells purified from the colonic mucosa in patients with inflammatory bowel disease (IBD) and Crohns disease. The diseased tissue displayed reduced immunoreactivity for phospho-Smad3, indicating a reduced level of TGF-β signalling and responsiveness, despite an increased level of TGF-β ligand. When subjecting cells derived from IBD patient tissue to antisense Smad7 oligonucleotides, they showed a reduction in Smad7 protein expression. More importantly, silencing of Smad7 not only restored the ability to respond to exogenous TGF-β, but lead also to reduced mRNA levels of the pro-inflammatory cytokines IFN-γ and TNF-α, the major counterplayers of TGF-β in the regulation of the inflammation status. This resulted in a TGF-β mediated inhibition of IFN-γ and TNF-α production and suppression of inflammation (Monteleone et al., 2001).

Carcinogenesis

Human Smad7 has been mapped to the chromosomal region 18q21, between the Mad-Related-2 (MADR2) and Deleted in Pancreatic Cancer-4 (DPC4) locus. MADR2 and DPC4 encode the Smad2 and Smad4 proteins, respectively (Boulay et al., 2001; Röijer et al., 1998). Deletions in the 18q21 region are the most common genetic variations observed in colorectal carcinoma. An analysis of the gene copy number from colorectal tumour biopsies revealed that Smad2 or Smad4 were more often deleted than Smad7. On the contrary, Smad7 appeared to be more often amplified than the other genes (Boulay et al., 2001). The most common genetic alteration of Smad genes in the tumour samples were simultaneous deletions of Smad2 and Smad4 while Smad7 was often either retained by normal diploidy or amplified (Boulay et al., 2001).

Mutational analyses performed in other tumour types such as hepatocellular carcinoma, ovarian cancers and pancreatic cancer made it seem unlikely that deletion or duplication of the Smad7 gene would influence cancer progression (Jonson et al., 1999; Wang et al., 2000; Kawate et al., 2001). Kleeff et al. (1999) however, claims that Smad7 can enhance tumourigenicity in pancreatic cancer. They observed increased Smad7 mRNA levels in human pancreatic cancer. When they transfected COLO-357 human pancreatic cancer cells with Smad7, the cells lost their growth inhibitory response to TGF-β1. In addition these transfected cells showed increased anchorage-independent growth and accelerated growth in nude mice (Kleeff et al., 1999).

(30)

PRESENT INVESTIGATIONS

Smad expression in normal and malignant prostate (paper I)

The prostate gland is dependent on androgens for its growth and differentiation. It has been observed that androgen ablation (i.e. castration) leads to regression and apoptosis (Kerr et al., 1972) in the rat ventral prostate (Kyprianou and Isaacs, 1988) as well as in PC-82 human prostate cancer (Kyprianou and Isaacs, 1988). Additional studies have implicated TGF-β as an inducer of apoptosis in the normal prostate as well as in malignant prostatic carcinoma epithelial cells (Hsing et al., 1996; Kyprianou and Isaacs, 1989; Landström et al., 1994; Rajah et al., 1997). Moreover, castration and, in particular, castration in combination with estrogen treatment in a rat prostatic adenocarcinoma model caused an increased protein expression of TGF-β1, as well as of TGF-β receptors (Landström et al., 1996). Taken together, these observations suggested a role for TGF-β in the induction of programmed cell death in the normal and androgen-sensitive malignant prostate. However, the molecular mechanism behind such a TGF-β

mediated apoptosis remained poorly understood.

We wanted to investigate a possible functional role for Smad proteins, the transducers of the TGF-β signal, in the normal and malignant prostate after castration. To address this issue, we examined the immunoreactivity of receptor-activated Smads (1, 2, 3 and 5), the common mediator Smad4 and inhibitory Smads (6 and 7), in the rat ventral prostate and in an androgen and estrogen sensitive prostate tumour model (Dunning R3327 PAP) (Isaacs, 1987). In parallel, we performed TUNEL-staining to assess the level and localisation of apoptotic cells within the two prostate models, which allowed us to correlate the Smad expression profile with the presence of apoptotic cells.

We observed increased expression levels of the Smads involved in TGF-β signal transduction (i.e. Smad2, Smad3 and Smad4) in the epithelial cells of the normal prostate after castration, as well as in the Dunning tumour cells. Furthermore, in prostate epithelial cells after castration, we demonstrated an increased activation of Smad2 as detected by antisera specific for phosphorylated Smad2 (Chen et al., 1998; Persson et al., 1998; Piek et al., 1999). These elevated levels of phosphorylated Smad2 were also observed in the Dunning tumour after castration, but were less prominent. Interestingly, the expression of Smad2 was very low in the Dunning tumour cells prior to castration, in contrast to the rat ventral prostate, where the Smad2 levels were elevated before castration and remained high after androgen ablation. Smad3 immunostaining was also elevated in response to castration, both in normal and to a lesser degree in malignant prostate. Interestingly, in the rat ventral prostate Smad4 expression first increased after castration, only to become sharply reduced at later time points. The expression levels of inhibitory Smad6 and Smad7 were low in the Dunning tumour when compared with normal prostate, but increased significantly after treatment.

(31)

There was a significant increase of the number of apoptotic cells in the normal and malignant prostate in response to castration. Estrogen treatment of rats transplanted with the Dunning tumours had an additive effect on castration-induced apoptosis. Interestingly, we were able to correlate areas of apoptotic cells with increased Smad protein expression.

In summary, we observed elevated protein expression levels for Smad2, Smad3, and Smad4, increased activation of Smad2, as well as elevated protein expression levels of inhibitory Smads in normal prostatic epithelial cells and, to some extent also in the malignant prostatic epithelial cells, after castration. These observations lend credence to the notion that there is a correlation between the TGF-β/Smad pathway and apoptosis in

vivo. Deregulated expression or inactivation of components in this pathway may

interfere with TGF-β induced apoptosis and create a favourable milieu for prostatic tumour development.

Transcriptional regulation of the Smad7 gene (paper II)

Smad7 has been proposed to be an inhibitor of TGF-β signalling (Hayashi et al., 1997; Nakao et al., 1997). Previous studies have shown that Smad7 is upregulated in response to TGF-β stimuli (Afrakhte et al., 1998; Nakao et al., 1997), suggesting a possible role in a negative feedback loop modulating the TGF-β signal. We wanted to investigate the molecular mechanism underlying Smad7 activation, by studying the transcriptional regulation of the mouse Smad7 gene promoter by TGF-β.

To obtain the mouse Smad7 promoter, we isolated several overlapping λ-phage clones spanning the Smad7 genomic region. Using the promoterless pGL3 luciferase reporter system we found that a 3 kb (kilo base pair) Xho1 fragment conferred TGF-β inducible luciferase activity in HepG2 cells. Sequential deletions of the 3 kb promoter fragment revealed the presence of a TGF-β-responsive region containing a palindromic GTCTAGAC motif. Point mutations introduced into the palindromic sequence abolished TGF-β mediated transcriptional activation. This palindromic element was initially identified by Zawel and co-workers (Zawel et al., 1998) in a artificial PCR-based screen in search for optimal Smad binding motifs. Indeed we could show by electrophoretic mobility shift assays (EMSA) that Smad2, Smad3, and Smad4 were part of a complex associating with the palindromic Smad binding element (SBE).

In addition we observed the presence of consensus binding sequences for AP-1 and Sp1, which we showed subsequently to bind c-Jun and c-Fos, and Sp1, respectively, in vitro, in electrophoretic mobility shift assays (EMSA). Deletion or mutation of either AP-1 or Sp1 binding motifs led to a dramatic decrease in promoter activity while retaining some TGF-β inducibility. In contrast, the major consequence of deletion or point mutation of the SBE site, was the complete loss of TGF-β inducibility. These findings suggested the importance of co-operation between Smads and general transcription factors, such as AP-1 and Sp1.

Smad7 in TGF-β Signalling

        

       

 !"#$  %  %

&'

To