All in the CCN family: essential matricellular signaling modulators emerge from the bunker.

The CCN family is a group of six secreted proteins that specifically associate with the extracellular matrix. Structurally, CCN proteins are modular, containing up to four distinct functional domains. CCN family members are induced by growth factors and cytokines such as TGFβ and endothelin 1 and cellular stress such as hypoxia, and are overexpressed in pathological conditions that affect connective tissues, including scarring, fibrosis and cancer. Although CCN family members were discovered over a decade ago, the precise biological role, mechanism of action and physiological function of these proteins has remained elusive until recently, when several key mechanistic insights into the CCN family emerged. The CCNs have been shown to have key roles as matricellular proteins, serving as adaptor molecules connecting the cell surface and extracellular matrix (ECM). Although they appear not to have specific high-affinity receptors, they signal through integrins and proteoglycans. Furthermore, in addition to having inherent adhesive abilities that modulate focal adhesions and control cell attachment and migration, they execute their functions by modulating the activity of a variety of different growth factors, such as TGFβ. CCN proteins not only regulate crucial biological processes including cell differentiation, proliferation, adhesion, migration, apoptosis, ECM production, chondrogenesis and angiogenesis, but also have more sinister roles promoting conditions such as fibrogenesis.


Introduction
Named after three prototypical members, cysteine-rich protein 61 (Cyr61; also known as CCN1), connective tissue growth factor (CTGF; also known as CCN2) and nephroblastoma overexpressed protein (Nov; also known as CCN3), the CCN family comprises six secreted proteins grouped together on the basis of a similar predicted modular secondary structure ( Fig.  1) (Bork, 1993;Perbal, 2004) 1 . CCN proteins comprise four modules: an insulin-like growth factor binding protein (IGFBP) domain (module I), a Von Willebrand factor domain (module II), a thrombospondin-homology domain (module III), and a cysteine knot, heparin-binding domain (module IV) (Bork, 1993;Perbal, 2004). Befitting secreted proteins, each also possesses a signal sequence (Lechner et al., 2000;Chen, Y. et al., 2001) (Fig. 1). Between modules II and III is the 'hinge region', which is susceptible to proteinase cleavage. Indeed, in biological fluids, CCN2 can be found as fragments, including N-terminal and C-terminal halves cleaved in the 'hinge region', as well as the individual 10-12 kDa C-terminal heparin-binding domain (module IV) (Brigstock et al., 1997). These variations in structure may have a direct effect on CCN function.
The three prototypical members of the CCN family were originally identified ~15 years ago (O'Brien et al., 1990;Bradham et al., 1991;Joliot et al., 1992). Since their discovery, >800 papers on CCN2, ~200 papers on CCN1 and ~100 papers on CCN3 have been published. Substantially fewer papers have examined CCN4, CCN5 and CCN6. In part, progress in this field has been hampered by a lack of unrestricted, readily commercially available reagents, such as 'gold standard' recombinant material and neutralizing antibodies. Consequently in vitro studies have used a variety of different protein sources, purification procedures and antibodies. Indeed, a major difficulty in the CCN field has been the purification of active proteins, presumably because of the presence of repeated cysteine residues, which require the use of mammalian expression systems such as baculovirus or stable cell lines.
Initially, it was believed that these proteins were classical growth factors, and that simple application of recombinant material to cells was sufficient to recapitulate the entire range of CCN-dependent activities. The collective work of many laboratories, and especially the recent development of transgenic and knockout mice, has resulted in a greater appreciation of the range and complexity of CCN action. Indeed, it is now established that the CCN proteins are not growth factors and thus should not be referred to as such. This fact was a principal driving force within the CCN community The CCN family is a group of six secreted proteins that specifically associate with the extracellular matrix. Structurally, CCN proteins are modular, containing up to four distinct functional domains. CCN family members are induced by growth factors and cytokines such as TGF␤ ␤ and endothelin 1 and cellular stress such as hypoxia, and are overexpressed in pathological conditions that affect connective tissues, including scarring, fibrosis and cancer. Although CCN family members were discovered over a decade ago, the precise biological role, mechanism of action and physiological function of these proteins has remained elusive until recently, when several key mechanistic insights into the CCN family emerged. The CCNs have been shown to have key roles as matricellular proteins, serving as adaptor molecules connecting the cell surface and extracellular matrix (ECM). Although they appear not to have specific high-affinity receptors, they signal through integrins and proteoglycans. Furthermore, in addition to having inherent adhesive abilities that modulate focal adhesions and control cell attachment and migration, they execute their functions by modulating the activity of a variety of different growth factors, such as TGF␤ ␤. CCN proteins not only regulate crucial biological processes including cell differentiation, proliferation, adhesion, migration, apoptosis, ECM production, chondrogenesis and angiogenesis, but also have more sinister roles promoting conditions such as fibrogenesis.
to rename each family member. We now appreciate that, although the CCN proteins indeed have some independent activity, they principally modify signaling of other molecules, the identity of which can widely vary, depending on the particular biological system. CCN proteins stimulate mitosis, adhesion, apoptosis, extracellular matrix production, growth arrest and migration of multiple cell types. As a result, they play essential roles in development, wound healing and disease. The reader is referred to a recent comprehensive book on this subject (Perbal and Takigawa, 2005). Here, we summarize current knowledge of role and function of CCN family members in these processes.

CCN receptors
Much effort has been made to identify specific receptors mediating the effects of CCN family members. CCN family members appear not to rely upon a unique, specific signaling receptor. Rather, CCN1, CCN2 and CCN3 all directly bind to specific integrins through discrete and often separate domains, several of which have been mapped by peptide inhibition studies (e.g. Ellis et al., 2003;Schober et al., 2003;Leu et al., 2003;Leu et al., 2004; (Fig. 3). That CCN proteins bind integrins was first shown by Kireeva et al. (Kireeva et al., 1998), who showed CCN1 binds to integrin ␣v␤3. The integrins bound include the principal integrins mediating angiogenesis and matrix attachment, such as ␣v␤3, ␣5␤1 and ␣6␤1 (Chen et al., 2000;Ellis et al., 2003;Leu et al., 2003;Chen, Y. et al., 2004;; however, it is not known whether CCN proteins display preferential binding to ␣ or ␤ integrin subunits. Both CCN1 and CCN2 are ligands of integrins ␣II␤3 in platelets and ␣M␤2 in monocytes (Jedsadayanmata et al., 1999;Schober et al., 2003). CCN1 and CCN2 also interact with heparan-sulfatecontaining proteoglycans (HSPGs), including syndecan 4 and perlecan, through a heparin-binding domain in module IV Nishida et al., 2003;Todorovicc et al., 2005) (Fig. 3). CCN4 binds the HSPGs decorin and biglycan (Desnoyers et al., 2001). As discussed below, integrins and HSPGs are essential for the adhesive and mitogenic function of CCN proteins and should be properly considered to be the Journal of Cell Science 119 (23) functional receptors for this family Nishida et al., 2003;Todorovicc et al., 2005). In addition, syndecan 4 and integrin ␣6␤1 are required for CCN1 to induce apoptosis in fibroblasts (Todorovicc et al., 2005). CCN2 also binds to the low-density lipoprotein (LDL) receptor-related protein 1 (LRP1) , in a heparin-dependent fashion, through module III (Gao and Brigstock, 2003). Furthermore, it can bind to the Wnt co-receptor LDL-receptorrelated protein 6 (LRP6) through the C-terminal (CT) domain (LRP6) (Mercurio et al., 2004). Finally, CCN2 can bind to the tyrosine kinase receptor TrkA (Wahab et al., 2005). Therefore multiple biological activities can be elicited through the interactions of discrete 'functional' domains of the CCN family members and different cell surface receptors.

CCN proteins: role in adhesion, migration and signaling
As might be expected from proteins that bind integrins and proteoglycans, CCN family members are independently active in standard adhesion assays identical to those used for type I collagen and fibronectin. CCN proteins promote adhesion through both integrins and HSPGs, although the identity of the integrins and HSPGs differ, depending on the system and cell type examined (Chen et al., 2000;Ellis et al., 2003;Leu et al., 2003;Chen, Y. et al., 2004). This presumably reflects the relative importance of the particular integrin to the adhesive ability of fibroblast compared with endothelial cells rather than an inherent specificity for CCN family members toward any particular integrins or HSPG. In addition to integrins and proteoglycans, CCN2-dependent adhesion can also involve LRP1 (Gao and Brigstock, 2003).
Fibroblasts plated on CCN2 activate the ERK pathway, which is required for their ability to attach to CCN2 . Unlike cells adhering to fibronectin, fibroblasts adhering to CCN1 or CCN2 need not spread properly or generate actin fiber networks and phosphorylate focal adhesion kinase (FAK) Latinkic et al., 2003). However, in some systems, such mature cell spreading can occur, and FAK phosphorylation occurs in response to CCN1, CCN2 and CCN3 Chen, C. C. et al., 2001). Intriguingly, immobilized recombinant CCN2 regulates migration of mesangial cells (Blom et al., 2001) by promoting dephosphorylation of FAK (Crean et al., 2004). Furthermore, CCN2-deficient mouse embryonic fibroblasts (MEFs) show defects in cell adhesion to fibronectin, including a significant reduction in FAK and ERK phosphorylation and delays in cell spreading and formation of actin stress fibers .
In normal mouse embryo fibroblasts CCN2 is located within a complex composed of fibronectin and the fibronectin receptors, integrin ␣4␤1 and integrin ␣5␤1 and syndecan 4 . Although likely, it is not clear whether other CCN molecules are also present in a complex of matrix and matrix receptors. One important physiological function of CCN proteins therefore appears to be to integrate cellular adhesion responses with the extracellular matrix environment, which is consistent with the proposed crucial role of CCN family members as adapter molecules integrating signaling between extracellular ligands and their receptors (Perbal, 2004) (Fig. 2). Note, however, that although the observations discussed above suggest that CCN proteins have some inherent adhesive ability, their physiological function may be to modify, rather than be the immediate cause of, cellular adhesive responses.

CCNs as co-factors for the ECM, growth factors and cytokines
In addition to their ability to bind integrins and HSPGs, CCN proteins can also bind growth factors and cytokines. For example, CCN2 binds TGF␤, through the N-terminal Von-Willebrand factor domain (module II) of CCN2 (Abreu et al., 2002) (Fig. 2), and this may play an important role in augmenting TGF␤ activity (see below). It also binds to fibronectin, through module IV Hoshijima et al., 2006) (Fig. 2) and to VEGF (Inoki et al., 2002). Similarly, CCN3 binds connexin 43, although the domain of CCN3 mediating this action is unknown (Fu et al., 2004). CCN3 also associates with the epidermal-growthfactor-like repeats of Notch1 through its C-terminal cysteine knot domain (Sakamoto et al., 2002). Consistent with the notion that CCN proteins promote signaling from growth factors is the observation that CCN1 does not possess mitogenic activity on its own, but rather enhances that of FGF and PDGF (Kireeva et al., 1996). Similarly, CCN2 also enhances FGF-or EGF-induced DNA synthesis, but lacks mitogenic activity alone Grotendorst and Duncan, 2005). CCN2 induces differentiation of myofibroblasts and increased collagen synthesis in concert with EGF, insulin-like growth factor 2 or insulin (Grotendorst et al., 2004;Gore-Hyer et al., 2003). Moreover, although not mitogenic on its own, CCN3 enhances basic-FGF-induced DNA synthesis and upregulates matrix metalloproteinase (MMP)-1 and PAI-1 expression (Lafont et al., 2005a;Lin, C. et al., 2005). In some studies, however, CCN2 has been shown to possess independent yet modest proliferative activity (~20% above control) (Asano et al., 2005). The proliferative activity of CCN2 resides in domain IV and acts through Ras/MEK/ERK signaling .
Significantly, CCN3 promotes FGF-and PDGF-mediated proliferation of C2C12 myoblast cells in an integrin-dependent fashion (Lafont et al., 2005a). Through integrins, CCN proteins thus probably modify signaling responses to other proteins and, in principle, CCN molecules could modify a wide range of signal transduction pathways. A corollary of this is that CCN proteins need not necessarily have independent activity; their physiological effects may depend on the action of the particular partner with which the CCN proteins interact.

CCN proteins in tissue repair and fibrosis
Possibly the best insight into the physiological relevance of the interactions between CCN and extracellular ligands comes from observations concerning the contribution of CCN2 to tissue repair following injury and in scarring or fibrotic responses. In adults, CCN1, CCN2 and CCN3 are induced during tissue repair (Igarashi et al., 1993;Latinkic et al., 2001;Lin, C. et al., 2005). Elevated, constitutive CCN2 expression is a hallmark of fibrosis . Transient overexpression of CCN2 results only in a minimal fibrotic response, however (Mori et al., 1999;Bonniaud et al., 2003). Moreover, CCN2-deficient MEFs express ␣-smooth muscle actin (␣-SMA) and type I collagen (Shi-wen et al., 2006a), which indicates that CCN2 is not required for the basal expression of these proteins in embryonic fibroblasts.
What inappropriate overexpression of CCN2 appears to do is to create an environment permissive for other stimuli to induce potent fibrotic responses. For example, overexpression of CCN2 results in fibrosis in mice that are otherwise resistant to developing pulmonary fibrosis in response to bleomycin (Bonniaud et al., 2004), a model that is TGF␤ dependent (Zhao et al., 2002). Simultaneous co-injection of CCN2 and TGF␤ causes sustained fibrotic responses in vivo, in contrast to application of TGF␤ alone, which causes only a transient fibrotic response that depends on the constant injection of ligand (Mori et al., 1999). Indeed, although CCN2 is induced by TGF␤ and has long been hypothesized to be a downstream mediator of at least some of the effects of TGF␤ (Grotendorst, 1997), recent evidence suggests that, in fact, CCN2 is an essential co-factor for and augments TGF␤ activity. In cultured Xenopus cells, CCN2 binds TGF␤ and enhances the ability of TGF␤ to bind TGF␤ receptors at low TGF␤ concentrations and hence indirectly affects Smad-responsive promoters (Abreu et al., 2002).
CCN2 is constitutively expressed in fibrotic and embryonic fibroblasts independently of TGF␤ Holmes et al., 2003;Chen et al., 2006). Experiments using Ccn2 -/-MEFs have shown that loss of CCN2 results in an inability of TGF␤ to induce expression of approximately onethird of those mRNAs induced in Ccn2 +/+ MEFs (Shi-wen et al., 2006a). Consistent with the fact that CCN2 is required only for a subset of TGF␤ responses, Ccn2 -/-MEFs show no impairment of the generic Smad pathway, emphasizing the relative selectivity of CCN2-dependent action (Shi-wen et al., 2006a). In contrast to the lack of effect of loss of CCN2 expression on basal type I collagen and ␣-SMA expression, the ability of TGF␤ to induce these proteins is impaired in Ccn2 -/-MEFs (Shi-wen et al., 2006a). Intriguingly, the ability of TGF␤ to activate adhesive FAK/PI3kinase/Akt signaling is significantly impaired in Ccn2 -/fibroblasts, and this pathway is necessary for optimal induction of CCN2-dependent genes in wild-type MEFs (Shi-wen et al., 2006a) (Fig. 2). Induction of ␣-SMA by TGF␤ has been shown to be FAK-, adhesionand integrin-dependent (Thannickal et al., 2003), which supports the notion that integrins are functional receptors for CCN2. Real-time PCR analysis has revealed that CCN2dependent transcripts require CCN2 even at the extremely early time-points examined, before de novo induction of CCN2 protein (Shi-wen et al., 2006a). Thus, in MEFs, CCN2 appears to be an essential cofactor required for TGF␤ to induce adhesive signaling responses and the correct signals for the formation of myofibroblasts. Indeed, the ability of TGF␤ to induce adhesion to the matrix is impaired in Ccn2 -/-MEFs (Shi-wen et al., 2006a). The results of these in vivo studies are therefore consistent with the in vivo experiments described above and confirm that CCN2 is required for maximal adhesive signaling in fibroblasts undergoing active tissue remodeling, such as in embryogenesis, fibrotic cells or tumor stroma Yang et al., 2005). These results also suggest that CCN proteins act by enhancing signals not only from the extracellular matrix but also from growth factors through integrin-dependent pathways.

CCN proteins play crucial roles in bone formation
Integrins and adhesive signaling are essential for the tissue remodeling necessary for embryonic development (Thiery, 2003). CCN family members might therefore play key roles in this process. Indeed, substantial evidence suggests that CCN proteins are essential for development. In cell culture systems, CCN proteins have long been known to promote differentiation and proliferation of chondrocytes and osteoblasts (for a review, see Takigawa et al., 2003). CCN2 promotes proliferation and differentiation of osteoblast and chondrocyte cell lines Nakanishi et al., 2000). Similarly, CCN1, CCN3 and CCN6 induce expression of chondrogenic markers (Wong et al., 1997;Lafont et al., 2005b;Sen et al., 2004). CCN4 promotes mesenchymal cell proliferation and osteoblastic differentiation while repressing chondrocytic differentiation (French et al., 2004). All CCN family members are expressed in chondrocytes and osteoblasts and are induced during fracture repair (Nakata et al., 2002;French et al., 2004;Schutze et al., 2005;Parisi et al., 2006), and each family member appears to respond differently to stimuli. For example, CCN1 expression and CCN6 expression decrease during chondrogenic differentiation, but only CCN6 RNA expression is reduced during osteogenic differentiation (Schutze et al., 2005). In osteoblasts, TGF␤ enhances CCN1, CCN2 and CCN5 expression but decreases CCN4 expression (Parisi et al., 2006). Intriguingly, CCN1 is activated by Wnt3A, and RNAi-mediated knockdown of CCN1 diminishes Wnt3A-induced osteogenic differentiation (Si et al., 2006). This indicates that CCN1 is an essential mediator of Wnt signaling (Si et al., 2006). Thus, although CCN proteins appear to have similar functions in vitro, each member may have different physiological effects in vivo depending on their different regulation and expression profiles.
Perhaps the most significant recent insights into the specific physiological roles of the CCN family have come from the generation of mutant mice lacking CCN2 (Ivkovic et al., 2003) or CCN1 (Mo et al., 2002). Ccn2 -/mice display severely malformed ribcages and die soon after birth owing to a failure to breathe (Ivkovic et al., 2003). These mice exhibit impaired chondrocyte proliferation and proteoglycan production within the hypertrophic zone (Ivkovic et al., 2003). Excessive chondrocytic hypertrophy and a concomitant reduction in endochondral ossification are also observed (Ivkovic et al., 2003). Further support for the idea that CCN2 regulates bone formation in development comes from studies of transgenic mice that overproduce CCN2 under the control of the mouse type XI collagen promoter. These mice develop normally but show dwarfism within a few months of birth owing to a reduced bone density (Nakanishi et al., 2001). The molecular basis for this deformity has not yet been explored; however, a possible explanation is that CCN2 overexpression results in abnormally premature ossification, before proper chondrocyte maturation. Point mutations in CCN6 have been linked to the autosomal recessive skeletal disease progressive pseudorheumatoid dysplasia (PPD), a human disease (Hurvitz et al., 1999), resulting in juvenile-onset cartilage degeneration. Conversely, mice in which exons III, IV and V of CCN6 were deleted, resulting in absence of CCN6 expression, showed no apparent phenotype (Kutz et al., 2005). In this study CCN6 expression was not detected anywhere in the mouse (Kutz et al., 2005). Although it is possible that complete loss of CCN6 expression in ccn6 -/mice may have resulted in the compensatory overexpression of other CCN family members, it is also plausible that additional factors than CCN6 may be the cause of PPD.

CCN family members contribute to angiogenesis and cancer
CCN proteins probably also have a key role during angiogenesis in development. Recombinant CCN1, CCN2 and CCN3 promote angiogenesis in vivo, when applied alone subcutaneously into corneas or in a chick chorioallantoic membrane assay (Babic et al., 1998;Babic et al., 1999;Shimo et al., 1999;Lin et al., 2003). Angiogenic activity of CCN1 has been confirmed in a rabbit ischemic hindlimb assay (Fataccioli et al., 2002). CCN1 promotes endothelial tubule formation in vitro through integrin ␣v␤3 (Leu et al., 2002). At much higher doses than those that promote angiogenesis, CCN2 can suppress the angiogenic activity of VEGF (Inoki et al., 2002). MMPs can cleave CCN2, neutralizing this effect (Inoki et al., 2002). Individual CCN family members may thus have different effects based on their doses or level of expression. The CCN1-null mutation is embryonic lethal: ~30% of those mice that die exhibit a complete failure in chorioallantoic fusion, whereas the remainder perish as a result of placental vascular insufficiency and compromised vessel integrity (Mo et al., 2002). These observations provide clear evidence for a key physiological role for the protein in angiogenesis. Although CCN2-null mutants do not exhibit defects in vessel integrity or placental vasculature, they display angiogenic deficiency in the growth plates during endochondral bone formation (Ivkovic et al., 2003). These results suggest that, although recombinant CCN1 and CCN2 have similar effects on cells, they have non-overlapping functions in vivo.
For informative, focused reviews on the role of CCNs in cancer, the reader is referred elsewhere (Mendenez et al., 2003;Rachfal and Brigstock, 2005). Confirming a role for CCN1 in cancer, ectopic expression of CCN1 enhances the growth of ovarian cancer cells in liquid culture and increases tumorigenicity in nude mice, whereas inhibition of CCN1 expression decreases proliferation and increases apoptosis in these cells (Gery et al., 2005). Similarly, patients with gastric adenocarcinomas display levels of CCN1 that correlate well with aggressive lymph node metastasis, more advanced tumor stage, histologically diffuse type, and early recurrence (Lin, M. et al., 2005).
Treatment of mice with a CCN2-neutralizing antibody greatly decreases osteolytic bone metastasis, the appearance of microvasculature, and suppresses the growth of subcutaneous tumors (Shimo et al., 2006). Similarly, in a recent study, anti-CCN2 antibody decreased tumor growth and metastasis and attenuated tumor angiogenesis and cancer cell proliferation in vitro and in vivo models of pancreatic cancer (Aikawa et al., 2006). Conversely, CCN5, which lacks the pro-proliferative module IV (Fig. 1), suppresses proliferation and its expression is reduced in cancers (Mason et al., 2004). Intriguingly, expression of CCN3, although it possesses module IV, inversely correlates with tumorigenicity (Gupta et al., 2001); since CCN3 binds connexin, it might promote the formation of gap junctions, which would reduce metastasis (Gellhaus et al., 2004;Fu et al., 2004). RNAi-mediated CCN6 inhibition promotes neoplastic progression, as visualized by increased anchorage-independent growth of human mammary epithelial cells, and elevated responsiveness to the mitogen insulin-like growth factor 1 (Zhang et al., 2005). These results suggest that, although CCN family members share similar structures, their activities differ, presumably because of differences in their amino acid sequences and the proteins with which they interact. Furthermore, they suggest that altering the relative expression levels of individual CCN family members may have profound physiological and pathological effects.

CCN gene regulation
As discussed above, although CCN proteins may have similar activities in vitro, the net effect of their contribution to physiology may be based on their abilities to respond to different stimuli. Of the CCN family, the only member whose gene regulation has been characterized in detail is CCN2. CCN2 is primarily regulated at the level of transcription (Grotendorst et al., 1996;Holmes et al., 2001;Holmes et al., 2003;Chen et al., 2002;. Similarly, faithful expression of a reporter gene in transgenic mice, including in development and in response to wound healing, can be achieved if a 2 kb fragment of the CCN1 promoter is used, which indicates that this gene is also regulated at the level of transcription (Latinkic et al., 2001). Synthesis of CCN2 protein and mRNA is stimulated by specific growth factors, such as endothelin 1 and TGF␤, in addition to environment changes such as hypoxia and biomechanical stimuli (Grotendorst et al., 1996;Holmes et al., 2001;Ott et al., 2003;Shi-wen et al., 2004;Higgins et al., 2004) (Fig. 2). TGF␤ also induces CCN1, CCN4 and CCN5, but not CCN6, and reduces CCN3 expression (Lafont et al., 2002;Sakamoto et al., 2004;Parisi et al., 2006); data obtained regarding the control of CCN2 gene expression by TGF␤ are therefore likely to be applicable for CCN1, CCN4 and CCN5. Indeed, the CCN1 promoter has been shown to contain a 'classical' Smad-binding motif (Bartholin et al., 2006). TGF␤-mediated induction of CCN2 mRNA in fibroblasts occurs within 30 minutes of TGF␤ treatment, without involving de novo protein synthesis (Grotendorst et al., 1996). It is severely impaired in fibroblasts lacking Smad3 . Indeed, a functional Smad element resides within the CCN2 promoter; Smad 3 and 4 potently activates, whereas Smad 7 suppresses, the CCN2 promoter through this motif   (Fig. 3).
The ability of TGF␤ to induce CCN2 also requires protein kinase C and the Ras/MEK/ERK MAP kinase cascade Stratton et al., 2002; (Fig. 3). As in the case of other TGF␤-responsive promoters that do not require the transcription factor AP-1, the induction of CCN2 by TGF␤ is antagonized by hyperactive AP-1 or Jun Nterminal kinase (JNK) , because of the ability of active Jun to bind to Smads off DNA and inhibit Smads from interacting with the target DNA sequences (Verrecchia et al., 2001). Intriguingly, the TGF␤-mediated suppression of CCN3 expression is Smad independent but JNK dependent (Lafont et al., 2002). The Smad element of the CCN2 promoter acts in concert with a tandem repeat of an ETS element . Endogenous Ets-1 transcription factor binds this element and is required for the induction of CCN2 by TGF␤ (van Beek et al., 2006) (Fig.  2). Additional elements that contribute primarily to the basal transcriptional activity of the CCN2 promoter include a BCE-1 (basal control element 1) site and an Sp1 site Holmes et al., 2003;Chen et al., 2002). The BCE-1 site also mediates the Ras/MEK/ERK-dependent response of the CCN2 promoter to endothelin 1 . Because endothelin is induced by TGF␤ (Rodriguez-Pascual et al., 2003;Shi-wen et al., 2006b), BCE-1 is probably indirectly involved in the induction of CCN2 by TGF␤. Befitting a gene associated with angiogenesis, CCN2 is also induced by hypoxia through a hypoxia-inducible factor (HIF)-response element (Higgins et al., 2004) (Fig. 3). In addition, CCN2 mRNA displays a degree of post-transcriptional regulation, primarily concerned with message stability. Thus there is a minimal 84-nucleotide element in the CCN2 mRNA 3Ј-UTR, which acts as a cis-acting element of structure-anchored  Fig. 3. Regulation of the CCN2 promoter and 3Ј untranslated region (3Ј UTR). The CCN2 promoter contains recognition sequences for HIF, Smad, BCE-1, Ets-1 and Sp1, as indicated. The 3Ј UTR of the gene (white rectangle) contains a cis-acting element of structureanchored repression (CAESAR). Hypoxia, TGF␤ and endothelin-1 induce CCN2 as indicated.
repression (CAESAR) in a chondrocyte cell line (Kubota et al., 2000) and also contributes to hypoxia-mediated CCN2 mRNA stabilization (Kondo et al., 2006) (Fig. 3). Post-transcriptional regulatory mechanisms might similarly control the expression of other CCN family members.
CCN2 overexpression is a fairly robust biological marker for fibrotic diseases . An ELISA that detects CCN2 in biological fluids is a very useful marker of the severity of fibrosis in diseases such as scleroderma and diabetic nephropathy (Dziadzio et al., 2005;Nguyen et al., 2006). CCN2 overexpression in the fibrotic disease scleroderma is independent of TGF␤ and Smads but dependent on BCE-1 and Sp1 (Abraham et al., 2000;Holmes et al., 2001;Holmes et al., 2003). It is likely therefore to depend in part on the expression of endothelin 1. Indeed, antagonism of the A and B endothelin receptors significantly reduces the overexpression of CCN2 in scleroderma (X. Shi-wen, A.L. and D.J.A., unpublished results); aberrant endothelin-1-dependent overexpression of CCN2 thus appears to be an important hallmark of fibrogenesis . Whether endothelin induces expression of other CCN family members is not known.

Conclusion
CCN family members provide independent adhesive functions by acting through integrins and HSPGs and facilitating interactions with the ECM. However, a wide body of evidence suggests that alone they are not capable of recapitulating the range of CCN-dependent activities either in vitro or in vivo. Rather, CCN proteins exert their influences by modifying, through integrin-and/or adhesion-mediated signaling pathways, the function of extracellular ligands such as extracellular matrix components and growth factors. It is conceivable that CCN proteins interact with different ligands and thereby have different biological effects. Undoubtedly the entire range of ligands with which they interact has not been elucidated, making establishment of in vitro assays recapitulating the range of CCN actions problematic. Biochemical and two-hybrid analyses are necessary to identify additional ligands of CCN proteins. To gain greater molecular insights into the precise mechanisms of CCN action, genetic analysis of the roles of CCN family members is therefore essential. In particular, conditional knockout or genereplacement strategies, or examination of the effects of loss of protein expression in particular cell types, are warranted. These models could test the effect of ablating CCN action specifically in certain tissues, or the functional equivalence of CCN proteins by replacing one family member with another. The latter approach could also test the extent to which differential control of CCN family gene expression is responsible for the net physiological effect of these proteins. Similarly, although much work has resulted in an appreciation of the control of CCN2 gene expression, a greater understanding of the mechanisms underlying the control of other CCN proteins is warranted.