This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Copyright Permission Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Hwa, V. Articles by Rosenfeld, R. G. Search for Related Content PubMed PubMed Citation Articles by Hwa, V. Articles by Rosenfeld, R. G.

The Insulin-Like Growth Factor-Binding Protein (IGFBP) Superfamily¹

Department of Pediatrics, Oregon Health Sciences University, Portland, Oregon 97201

	Abstract

Top Abstract I. Introduction II. IGFBP Family III. Mac25 IV. CCN Family V. L56 VI. ESM-1 VII. Structural Relationships... VIII. Functional Relationships... IX. Evolutionary Relationships... X. Summary References

 

I. Introduction

A. Concept of an IGFBP superfamily

B. Superfamily nomenclature

II. IGFBP Family

A. Structure of IGFBPs

B. Correlations between structure and function

C. Biological functions of IGFBPs

D. Genomics of IGFBPs

III. Mac25

A. IGFBP-rP1 (Mac25/TAF/PSF)

IV. CCN Family

A. IGFBP-rP2 (CTGF)

B. IGFBP-rP3 (NovH)

C. IGFBP-rP4 (Cyr61)

D. New members

V. L56

A. IGFBP-rP5 (L56/HtrA)

VI. ESM-1

A. IGFBP-rP6 (endothelial specific molecule-1)

VII. Structural Relationships Within the IGFBP Superfamily

A. Protein domains (modules)

B. Gene structure and correlation with protein domains

VIII. Functional Relationships Within the IGFBP Superfamily

A. IGF binding

B. Insulin binding

C. IGF/insulin-independent actions

IX. Evolutionary Relationships Within the IGFBP Superfamily

A. N-terminal domains

B. C-terminal domains

C. Evolutionary models

X. Summary

	I. Introduction

Top Abstract I. Introduction II. IGFBP Family III. Mac25 IV. CCN Family V. L56 VI. ESM-1 VII. Structural Relationships... VIII. Functional Relationships... IX. Evolutionary Relationships... X. Summary References

 
THE insulin-like growth factor (IGF) system (Fig. 1) is well defined, with profound effects on the growth and differentiation of normal and malignant cells. The established components of the IGF system include IGFs (IGF-I and IGF-II), type I and type II IGF receptors, IGF-binding proteins (IGFBPs), and IGFBP proteases. IGF-I and IGF-II, which are structurally similar to insulin, are two highly homologous small hormone peptides of approximately 7 kDa molecular mass. First identified in 1957, they were originally named sulfation factors (1), nonsuppressible insulin-like activity (2), and multiplication-stimulating activity (3). They were renamed somatomedins (4) and subsequently IGFs (5). IGFs are ubiquitously expressed and are important mitogens that affect cell growth and metabolism. In addition to endocrine effects exerted by circulating IGFs (6, 7), locally produced IGFs exert paracrine, as well as autocrine, effects on cell proliferation (8, 9, 10). The IGFs interact with specific cell surface receptors, designated type I and type II IGF receptors, and can also interact with the insulin receptor. The mitogenic effects of IGF are mediated mainly through interactions with the type I IGF receptor, which, like the insulin receptor, is a receptor with tyrosine kinase activity. The type II IGF receptor is structurally distinct, binds primarily IGF-II, but also serves as a receptor for mannose-6-phosphate-containing ligands (11). The role(s) of the type II receptor in mediating IGF action is less well defined (12).

View larger version (32K): [in this window] [in a new window]   Figure 1. The IGF system. The components of this system include the peptides IGF-I and -II, IGFBP-1 to -6, IGFBP-rPs, IGFBP proteases, type I and type II IGF receptors, and potential IGFBP(s) and IGFBP-rP(s) receptors. M6P, Mannose-6-phosphate.

Of particular interest is the recent discovery of several groups of cysteine-rich proteins with discrete, but striking, structural and functional similarities to the IGFBPs (21, 22, 23, 24). This has led to the proposal of an IGFBP superfamily, comprised of the IGFBPs and these IGFBP-related proteins (IGFBP-rPs) (23). Since several comprehensive reviews on IGFBPs are available (10, 25, 26), the present review will include only the most recent information on IGFBP structure and function and will focus on the IGFBP-rPs and their structural, functional, and evolutionary relationships with the conventional IGFBPs.

A. Concept of an IGFBP superfamily
First coined by Dayhoff in 1978, the term "superfamily" was used to discriminate between closely related and distantly related proteins (27). The relatedness of proteins was based solely on similarities between the primary protein structures, with amino acid similarities set arbitrarily at equal to or greater than 50% for closely related proteins (considered a family), and less than 50% for those more distantly related (considered a superfamily). With the wealth of information now available on protein structures from different organisms, the more acceptable classification of proteins into families and superfamilies is determined not only by amino acid similarities, but also by considering ancillary features such as tertiary structures (conformational similarities), functional similarities, and even tissue specificity (28). Furthermore, establishing relatedness among proteins requires that the evolutionary relationships be considered. A current definition of a superfamily, therefore, is a number of families who share some structural and functional characteristics that have been conserved through evolution. The list of superfamilies of genes is long, and includes the globins, collagens, actins, immunoglobulins, serine proteases, and, more recently, the transforming growth factor-ß and the nuclear receptor superfamilies.

The existence of proteins able to bind IGFs with high affinity had been suspected since the late 1960s [see review (29)]. The first IGFBP to be purified and its cDNA cloned was IGFBP-1 (30, 31, 32, 33, 34, 35). Development of the Western ligand blot techniques, using ¹²⁵I-IGF ligands to probe for proteins immobilized on nitrocellulose filters (36) facilitated elucidation of the IGFBPs. By 1991, six IGFBPs (IGFBP-1 to IGFBP-6) demonstrating high IGF binding affinity had been identified from a variety of biological fluids, mammalian and nonmammalian, and, in many cases, their respective cDNAs and genes had been cloned and characterized. The structural characteristics of the human IGFBPs are summarized in Table 1.

The discovery of cysteine-rich proteins sharing similarities with the IGFBPs led to the proposal of a new superfamily, an IGFBP superfamily (23), an hypothesis consistent with the current definition for a superfamily. The six established IGFBPs were classified as a family based on two key features. First, the IGFBPs are cysteine-rich proteins (16–20 cysteines in the pre-peptides) sharing high similarity in their primary amino acid sequences. Structurally, the cysteines are clustered at the conserved N-terminal third (12 cysteines in IGFBP-1 to -5; 10 in IGFBP-6) and at the conserved C-terminal third (6 cysteines) of the proteins (Fig. 2). The N and C domains are separated by a midregion of little similarity among the IGFBPs. The second key feature of the IGFBPs is their unique ability to bind IGFs with high affinity, presumably as a result of the N and C domains forming the correct tertiary configuration for high-affinity IGF binding. These two criteria, used to distinguish and classify conventional IGFBPs, were recently reevaluated in light of the identification of additional cysteine-rich proteins that share structural similarities with the IGFBPs: they carry the N-terminal domain of IGFBPs, but deviate from the common IGFBP structure in the midregion and C terminus. Functionally, at least four of these proteins are able to bind IGFs in in vitro assays, albeit at 100-fold or lower affinity than that observed with IGFBPs (22, 23, 37, 38, 39). These results, thus, substantiated that these proteins, while not falling into the classical definition of IGFBPs, are certainly related to the IGFBPs.

View larger version (66K): [in this window] [in a new window]   Figure 2. Amino acid sequence alignment of human IGFBP-1 to -6. The signal peptides (number of amino acid residues are indicated in brackets) were not included in the analysis. The numbering system (left of the figure) is that of the mature peptide. Alignment was performed using the Clustral method in the DNA STAR program. Small gaps were introduced to optimize alignment. Consensus amino acid residues are as indicated (shaded solid black); boxed amino acid residues indicate matched residues.

B. Superfamily nomenclature
The various IGFBP-rPs were discovered by a number of different groups and designated accordingly (Table 2). Although both mammalian and nonmammalian IGFBP-rPs have been described, only the human IGFBP-rPs are presented in Table 2. At present, there are four proteins/families that are related to the IGFBPs. Mac25 was originally identified as a cDNA derived from leptomeninges (40); the mac25 cDNA was subsequently expressed in a baculovirus system, and the synthesized protein was shown to bind IGFs and was renamed IGFBP-7 (22). Independently, the same protein has been purified from human diploid fibroblast cells and named prostacyclin-stimulating factor [PSF (41)] and from human bladder carcinoma cells [tumor adhesion factor (TAF) (42)]. The CCN family of proteins consists of a human growth factor-inducible, immediate-early gene [cyr61 (43), connective tissue growth factor (CTGF) (44), and the human nephroblastoma overexpression gene (novH) (45)]. Recently, three new members of the CCN family have been identified in Wnt-1 (cysteine-rich glycosylated signaling proteins that are oncogenic) transformed cells: WISP-2 (46) and its rat counterpart, rCop-1 (47); WISP-1 (46) and the mouse orthologue, Elm-1 (48); and WISP-3 (46). Independently, WISP-2 was identified in primary human osteoblast cells and designated CTGF-L [CTGF-like (39)]. Two other proteins related to the IGFBPs are L56 (49), also named HtrA (50), a potential serine protease of IGFBPs, and endothelial cell-specific molecule, ESM-1 (51).

The IGFBP-rPs will be discussed in depth in Sections III–VI. Their structural and functional relationships with the IGFBPs will be presented in Sections VII and VIII, respectively.

	II. IGFBP Family

Top Abstract I. Introduction II. IGFBP Family III. Mac25 IV. CCN Family V. L56 VI. ESM-1 VII. Structural Relationships... VIII. Functional Relationships... IX. Evolutionary Relationships... X. Summary References

 
Since the advent of the Western ligand blot technique (36), biological fluids from numerous systems have been studied for the presence of IGFBPs. In mammalian species (human, mouse, rat, bovine, and, more recently, water buffalo), IGFBP-1 through -6 have been well documented, their expression and regulation studied, their proteins purified, and their cDNAs and genes cloned and characterized (10, 25, 26, 53). In nonmammalian species, IGFBPs have been described as well, but are less well characterized. The cDNAs for IGFBP-2 and -5 from chicken have been cloned (54, 55), as is a partial Xenopus IGFBP-5 sequence (56). IGFBPs, as detected by Western ligand blot, are present in the serum of reptiles, such as crocodile (Crocodilian johnsonii), lizard (Tiliqua rugosus), snake (Notechis scutatus), tortoise (Chelodina longicollis and Emydura macccquarii), and tuatara (Sphenodon punctatus) (57). Whether these reptilian IGFBPs correlate to any mammalian IGFBPs is unknown, but since they are readily detectable by the Western ligand blot technique, these proteins probably have high affinities for IGFs, and, therefore, are likely to be structurally similar to mammalian IGFBPs (57). IGFBPs with molecular weights similar to, and similarly regulated as, the mammalian counterparts have been detected in bony fish [see review (58)]. Even the serum from the agnathan lamprey contains IGFBPs, as detected by ligand blot (57). Again, the structures of these ancient IGFBPs are unknown, but, like the reptilian IGFBPs, they presumably are structurally similar to the mammalian IGFBPs.

In this section, an analysis of IGFBP structure is summarized, based on information gathered from studies with mammalian IGFBPs. Correlation between IGFBP structure and function at the molecular level has only recently begun to be elucidated. Particular attention has been paid to the unique attributes of the N-terminal domain, alone and in combination with other IGFBP domains, since it is this portion of the molecule that is conserved exclusively in the IGFBP-rPs.

A. Structure of IGFBPs
The primary structures of mammalian IGFBPs appear to contain three distinct domains of roughly equivalent sizes: the conserved N-terminal domain, the highly variable midregion, and the conserved C-terminal domain. Alignment [Clustral method (59), DNA Star program] of the human IGFBPs indicates that, overall, the human IGFBPs share approximately 36% similarity (defined as "the direct comparisons of amino acid sequence without accounting for phylogenetic relationships", DNA Star), although as presented below, alignment of the conserved N and C domains shows significantly higher similarities. Between mammalian species, each IGFBP is highly conserved.

1. N-terminal domain. In the mature IGFBP peptides, the N terminus third of the proteins contains 80–93 amino acid residues after the signal peptides (Fig. 2) and shares approximately 58% similarity. Ten to 12 of the 16–20 cysteines found in the prepeptides are located within this domain. In IGFBP-1 to -5, these 12 cysteines are fully conserved, whereas in IGFBP-6, 10 of the 12 cysteines are invariant. Interestingly, rat IGFBP-6 is missing an additional 2 cysteines (the first 2 cysteines in the primary sequence) in the N-terminal domain (60). The high number of cysteines within such a small domain suggests that this domain is highly structured, with a maximum of 6 disulfide bonds formed (5 in the case of IGFBP-6). The even number of cysteines suggests that intradomain disulfide bond formation is more likely than interdomain disulfide linkages with cysteines in the C-terminal domain. Indeed, two recent studies have provided data supporting the hypothesis that the N-terminal and C-terminal domains are not linked by disulfide linkages (61, 62). Furthermore, analysis of tryptic digested fragments of human IGFBP-6 indicated that the native disulfide linkages in IGFBP-6 occur between cysteines that are close together in the primary sequence, forming sequential subdomains with at least 2 disulfide-linked subdomains in the N-terminal domain (61, 63). In the N-terminal domain, the first 6 cysteines form the first subdomain, and the next 4 cysteines form the second subdomain. Interestingly, in rat IGFBP-3 (64) and human IGFBP-5 (65), it was similarly demonstrated that the last 4 cysteines of the N-terminal domain formed overlapping disulfide linkages (Cys⁵⁶-Cys⁶⁹ and C⁶³-C⁸⁹ for rat IGFBP-3 (64), consistent with the second subdomain structure proposed by Neumann et al. (61). These results have been recently supported by nuclear magnetic resonance (NMR) spectroscopy of this subdomain from human IGFBP-5, which shows a rigid, globular structure stabilized by the two disulfide bridges (65).

Within the N-terminal domain, a local motif (GCGCCxxC) is well conserved among the IGFBPs. The exception is in IGFBP-6, which substitutes a GCAEAEGC sequence, thereby accounting for the two "missing" cysteines in IGFBP-6. The significance of this motif is as yet unknown, but, as indicated below, is also highly conserved in the IGFBP-rPs. One hypothesis is that the motif may be important in interactions with IGFs. However, IGFBP-6 has retained the high affinity for IGFs (in particular IGF-II), although the internal GCC amino acids are replaced with AEA. A search of the protein data bank (BLAST search, SwissProt) revealed several other proteins containing the GCGCCxxC motif, including a cysteine-rich protein found in the eggshell of the silk moth [chorion protein (66)]. Variants of this motif are also found in GP40, a small undefined protein found in Mycobacteriophage 15 (67), in the peplomer protein [a viral spike glycoprotein (68)], and in metallothionein-like protein 1 (69). The role(s) of the GCGCCxxC motif, or its close variants, are not known. Clearly, the GCGCCxxC motif appears to be highly conserved in, but not unique to, IGFBPs.

2. Midregion. For the human IGFBPs, the midprotein segment, ranging in size from 55 amino acid residues to 95 amino acids, separates the N-terminal domain from the C-terminal domain. The amino acid sequence for each midsegment appears to be unique to the protein, with shared similarity less than 15%. The belief is that this region acts structurally as a hinge between the N- and C-terminal domains.

Intriguingly, posttranslational modifications (glycosylation, phosphorylation) of the IGFBPs have been found so far in the midregion, but not in the N- or C-terminal domains. There has been no clear evidence to date that IGFBP-1 or -2 are glycosylated (an early paper suggests that IGFBP-1 was glycosylated (70), whereas IGFBP-3 and -4 are N-glycosylated, and IGFBP-5 and -6 are O-glycosylated. N-glycosylation occurs only on an asparagine that is part of the consensus sequence Asn-X-Ser/Thr, where X is any amino acid except proline. Consistent with this prediction, three N-glycosylation sites in the mature protein, Asn⁸⁹, Asn¹⁰⁹ and Asn¹⁷² (corresponding to prepeptide, Asn¹¹⁶, Asn¹³⁶, and Asn¹⁹⁹) have been characterized in IGFBP-3 (71), and one, Asn¹⁰⁴ (Asn¹²⁵ of prepeptide), in IGFBP-4 (72). Although there is one potential N-glycosylation site in the C terminus of IGFBP-6, this site does not appear to be glycosylated (73). In contrast to the N-glycosylation sites, there are no consensus sequences for predicting O-glycosylation sites. Nevertheless, it has been demonstrated that IGFBP-5 (74) and IGFBP-6 (61, 73) are both O-glycosylated. Very recently, the O-glycosylation sites in human IGFBP-6 were determined to be within the midregion at 5 residues, Thr¹²⁶, Ser¹⁴⁴, Thr¹⁴⁵, Thr¹⁴⁶, and Ser¹⁵² (61). The O-glycosylation states of other mammalian IGFBP-6 indicate that in the rat, mouse (75, 76), and pig (53) IGFBP-6 is glycosylated to a lesser extent than human IGFBP-6 (61); bovine IGFBP-6 may be similarly glycosylated as human IGFBP-6. The ability to bind IGFs with high affinity appears not be influenced by N- or O-glycosylation, although there may be effects on other function(s) of IGFBPs, such as resistance to proteolysis (61).

Three of the six IGFBPs, IGFBP-1, -3, and -5, are posttranslationally modified by phosphorylation (77). Phosphorylation of proteins is an important and critical posttranslational modification mechanism that is used by cells to stringently regulate the activities of numerous intracellular proteins, including proteins involved in the signal transduction pathways, in the cell cycle, and in gene expression. The purpose of phosphorylating secreted proteins like IGFBPs is unclear, but there is evidence that, at least for human IGFBP-1, phosphorylation enhances the affinity of hIGFBP-1 for IGFs by 5-fold (78, 79). Phosphorylation in all three IGFBPs is predominantly at serine residues found in the midregion of the IGFBPs. Phosphorylation of hIGFBP-1, first described by Frost and Tseng (80) and by Jones et al. (78), is at three serine residues of the mature peptide: Ser¹⁰¹ and Ser¹¹⁹, both in the midregion of the protein, and Ser¹⁶⁹ (located at the amino-terminal end of the hIGFBP-1 C-terminal domain) (81). Frost and Tseng (80) demonstrated that, under in vitro conditions, casein kinase II and cAMP-dependent protein kinase are able to phosphorylate IGFBP-1. Recently, phosphorylated rat IGFBP-1 was described (82). Unlike hIGFBP-1, however, only two serine residues were phosphorylated (Ser¹⁰⁷ and Ser¹³² in the nonconserved midregion), and more importantly, phosphorylation did not appear to affect IGF binding.

The phosphorylation status of human IGFBP-3 was analyzed by Hoeck and Mukku (83), who showed that Ser¹¹¹ and Ser¹¹³ (Ser¹³⁸ and Ser¹⁴⁰ of prepeptide), which are within the consensus sequences for protein kinase CKII, were phosphorylated. Both serines are also in close proximity to one of the N-glycosylation sites, Asn¹⁰⁹ (Asn¹³⁶ of prepeptide). Further, they demonstrated that phosphorylation did not appear to affect IGF binding by IGFBP-3. Interestingly, phosphorylation of IGFBP-3 can be up-regulated by IGFs, through a mechanism involving IGF-I-type I IGF receptor interaction (84). The significance of phosphorylated IGFBP-3, and the significance of IGF regulation of phosphorylation, are unclear but may affect IGFBP-3 interactions with acid-labile subunit (ALS) or with the cell surface (84).

Evidence for the phosphorylation of IGFBP-5 at serine and threonine residues is limited to one report (85), although, like all the IGFBPs, there are several potential phosphorylation sites (84). The biological significance of IGFBP-5 phosphorylation is unknown.

3. C-terminal region. The C-termini of IGFBPs, like the N-terminal domain, are highly conserved and, among the human IGFBPs, share a similarity of approximately 34%. The remaining 6 cysteines of the total 16–20 cysteines are found in the C terminus and are strictly conserved (Fig. 2). Evidence from two independent studies indicated that the 6 cysteines are involved in intradomain disulfide bond formation (58, 59, 63). Neumann and Bach (63), in their studies of human IGFBP-6, and Forbes et al. (62), in their studies of bovine IGFBP-2, deduced that the disulfide bonding pattern of the C-terminal region was between adjacent cysteines.

The primary sequence of all members of the IGFBP family surrounding the last 5 cysteines is strikingly similar (40%), implying that the tertiary structure of the C-terminal domain should be almost identical. Interestingly, the amino acid sequences encompassing these last 5 cysteines share 37% similarity with the thyroglobulin-type-I domain (86). The thyroglobulin-type I domain consists of about 65 amino acid residues, which are repeated 10 times in the N-terminal part of thyroglobulin (86). Its function(s) is unknown, but the domain is found in a number of proteins with varying physiological functions in different organisms (87). These include the major histocompatibility complex class II-associated p41 invariant chain fragment (88), nidogen (89), saxiphilin (90, 91), a tumor-associated cell surface antigen known also as GA733 (92), a cysteine protease inhibitor from the egg of Chum salmon (93), equistatin, a new inhibitor of cysteine proteinases from sea anemones (94), and entactin-2, a new basement membrane protein (95). In IGFBPs, the role of this domain has yet to be determined, but is likely to affect binding to IGFs, and perhaps participate in the binding of IGFBPs to cell surfaces and/or to the extracellular matrix (ECM) proteins via heparin-binding sites. Consistent with this is the observation that IGFBP-1 and -2 contain RGD motifs, which are known to be involved in binding integrins (96). Similarly, heparin binding motifs (xBBBxxBx, where B is a basic residue, Arg, Lys, or His, and X is any residue) are found within the C-terminal domains of IGFBP-3, -5, and -6, and, for IGFBP-3 and -5, are involved in binding to cell surface and/or the ECM (16, 97, 98, 99, 100).

B. Correlations between structure and function
1. IGF-IGFBP interactions. The IGFBPs were so designated because of their abilities to bind IGFs with high affinity (K_d 10^-10 M). However, the precise molecular interactions between IGFBPs and IGFs are still unclear. It has also become apparent that IGFBPs can interact with proteins other than IGFs, including the ALS from serum, insulin, components of the cell surface, ECM proteins, and, potentially, intracellular components. These additional interactions may result in biological consequences not directly related to IGF action. Correlations between IGFBP structure and function have recently begun to emerge and will be summarized here.

It is worth noting here that the methods used to detect and study IGF-IGFBP interactions have become increasingly sensitive, permitting better assessment of both high and low IGF-affinity binding. Although the ligand blotting technique is the preferred and most commonly used method for detecting IGF-IGFBP interactions, more sensitive methods include affinity cross-linking assays, charcoal solution binding, solid-phase binding, and, recently, BIACORE analysis. Since the methods used vary among research groups, discrepancies in IGF-binding affinities among different studies are likely to arise, and it is often not obvious whether such differences reflect technical variations or underlying biology.

The high-affinity binding of IGFs by IGFBPs has long been hypothesized to involve interactions between the conserved N-terminal and C-terminal domains. Support for this hypothesis initially came from in vivo observations that in biological fluids, IGFBPs can be proteolysed resulting in diminished affinities for IGFs. Proteolysis was, thus, proposed as a mechanism for modulating IGF bioavailability (101, 102, 103, 104, 105). Proteolysis of IGFBPs was first observed in serum from pregnant women, where it was demonstrated that IGFBP-3 was proteolytically cleaved to yield a predominant 29–30 kDa form that was still capable of binding IGFs, but with reduced affinity (101, 106). Since those observations, proteolysis of IGFBP-2 to -6 has been described in numerous studies of various biological fluids from different organisms, generating IGFBP fragments with decreased or no apparent affinity for IGFs.

In vitro generation of IGFBP fragments by limited proteolysis supports the in vivo data. For example, limited proteolysis of recombinant human IGFBP-3 (nonglycosylated and glycosylated) with the serine protease prostate-specific antigen (PSA) (102, 107) or with plasmin (108, 109) generated a 22/25 kDa fragment with weak affinity for IGFs (residues 1–160) and a 16-kDa fragment (residues 1–95 that includes the N-terminal domain) with no detectable affinity for IGFs by affinity cross-linking assays. Vorwerk et al. similarly generated an approximately 15-kDa plasmin-digested IGFBP-3 fragment, corresponding to the N-terminal domain and part of the midregion (amino acid residues 1–97), capable of weakly binding to IGF-I, and detectable by both Western ligand blot and affinity cross-linking assays (110). Proteolytically modified IGFBP-4 generated a 16-kDa fragment that also could be affinity cross-linked specifically to IGF-I and IGF-II, although with a 20-fold lower affinity compared with intact IGFBP-4 (111). This 16-kDa IGFBP-4 fragment, like the 16-kDa N terminus IGFBP-3 fragment, corresponds to the N-terminal region and a small portion of the midregion (112). Similarly, a 23-kDa IGFBP-5 fragment from osteoblast-like cells that is carboxy truncated (113, 114) and a 10-kDa fragment (residues 1–94 of mature peptide) corresponding to the N-terminal fragment of endoproteinase Asp N-digested IGFBP-5 (65), demonstrated decreased binding affinity for IGF-I and IGF-II. Thus it appears that the N-terminal domain, with perhaps part of the midregion, can bind IGFs, but high-affinity IGF-binding also requires the added presence of the C-terminal domain.

The ability of IGFBP proteolytic fragments to bind IGFs, albeit with reduced affinity, has been further investigated using in vitro generated recombinant IGFBP peptides. Spencer and Chan (115) generated IGFBP-3 fragments that essentially corresponded to the N-terminal half (residues 1–147) and the C-terminal half (residues 151–263) of the IGFBP and showed that each of these fragments bound IGFs, but with less affinity than intact IGFBPs. Only a handful of studies have examined in depth the ability of the N-terminal domain to interact with IGFs, and results have been mixed. Baxter and Firth synthesized IGFBP-3 fragments that correlated to the N-terminal domain, the N-terminal domain plus the midregion, and a mutant IGFBP-3 in which the midregion was deleted (116). None of the fragments, however, was detectable by ligand blotting, although recent data indicate the N-terminal domain bound ¹²⁵I-labeled IGF-II in solution binding assays (117). In contrast, recombinant N-terminal domain fragments generated by Yamanaka et al. (118) and by Vorwerk et al. (110), were detectable by ligand blotting, as well as by the more sensitive affinity cross-linking assay.

Further delineation of specific regions and subdomains of IGFBPs involved in IGF binding has come from limited chimera studies, as well as regional mutagenesis of the N and C domains. In the IGFBP N-terminal domain, the conserved GCGCCxxC motif was thought to be important for interactions with IGFs. The fact that the motif is incompletely conserved in IGFBP-6 (GCAEAxxC), however, would suggest that the role of this motif in the binding of IGFs may be subtle, or that there may be other explanations for its conservation in the IGFBPs and IGFBP-rPs.

In a recent study of human IGFBP-4 deletion mutants, Qin et al. (119) concluded that Leu⁷²-Ser⁹¹ is important for IGF-II binding, as deletion of this region rendered the N-terminal peptide undetectable by ligand blot. Further, within this segment, a structural disruption generated by a His⁷⁴ (a basic amino acid conserved in IGFBP-4 from different species, but not conserved among the IGFBPs) to Pro⁷⁴ point mutation reduced the affinity of full-length IGFBP-4 for IGF-II by 50-fold. The N-terminal domain of rat IGFBP-3 demonstrated a reduction in IGF-II binding to less than 12% relative to full-length IGFBP-3 as determined by a sensitive solid-phase binding assay (64). A smaller fragment of the rat IGFBP-3 corresponding to the last 4 cysteines of the N-terminal domain [i.e., the second N-terminal subdomain described by Neumann et al. (61) and Kalus et al. (65)] dramatically reduced IGF-II binding by four-logs. The same subdomain in recombinant human IGFBP-5 demonstrated 10- to 200-fold reduced affinity for IGFs by BIACORE analysis (65). The remaining N-terminal region, i.e., the N-terminal subdomain encompassing the first 6 (IGFBP-6) or 8 cysteines (IGFBP-1–5), has not been tested for IGF affinity. Interestingly, Hobba et al. (120, 121) showed that in bovine IGFBP-2, Tyr⁶⁰, which is within the second subdomain and highly conserved among the IGFBPs and across species, substitution by Ala⁶⁰ or Phe⁶⁰ reduced, but did not abolish, affinity for IGF-I (4-fold and 8.4-fold, respectively) and for IGF-II (3.5-fold and 4-fold, respectively). These results were consistent by both charcoal binding assays and BIACORE analysis (120, 121). Mutations of adjacent residues, which are well conserved, did not reduce affinity. From these results, it can be deduced that Tyr⁶⁰ is probably one of the many contact points with IGFs. This is supported by NMR studies of IGFBP-5-IGF-II complexes, in which the analogous Tyr (Tyr⁵⁰) is proposed to interact with IGF-II, as are residues Val⁴⁹, Pro⁶², and Lys⁶⁸-Leu⁷⁴ (65). Based on this handful of studies, it can be inferred that the two N-terminal subdomains proposed by Neumann et al. (61) are important for the integrity of the (partial) IGF binding pocket. Further studies are required to elucidate the precise points of contact with IGFs, which may vary from IGFBP to IGFBP.

The C-terminal domain of IGFBPs, without question, is essential for high affinity IGF binding, although more data are available regarding their non-IGF binding properties than their IGF binding characteristics (see below). Chimeras constructed between rat IGFBP-3 and IGFBP-2 indicated, not surprisingly, that the C-terminal domain from IGFBP-3 can be exchanged for the C-terminal domain of IGFBP-2 with no loss of IGF-II binding (64). However, replacement of the IGFBP-3 midregion with the IGFBP-2 midregion reduced the relative affinity of the resultant chimera for IGF-II by at least 37%, suggesting that the midregion of each IGFBP may maximize high-affinity IGF binding by the specific IGFBP.

Mutagenesis of the carboxy end of the IGFBP-1 cDNA (122) showed that deletion of the C-terminal 20 amino acids resulted in loss of IGF binding by ligand blotting. In contrast to IGFBP-1, deletion of a similar region in human IGFBP-4 (C-terminal Lys²¹⁵-Glu²³⁷) had no effect on relative IGF binding (bands in ligand blot assessed by radioactivity), but an additional deletion of 10 amino acids (removal included the highly conserved Cys-Trp-Cys-Val motif) reduced relative binding to less than 15% of wild-type IGFBP-4 (119). Similar sequential C-terminal deletion studies, with recombinant bovine IGFBP-2 and using charcoal-binding assays, suggested that loss of the region spanning the last four cysteines reduced IGF binding (62). In particular, results suggest that residues Lys²²²–Asn²³⁶ may be in close proximity to the N-terminal domain, to allow both domains to interact with IGF. Consistent with this proposal, recent site-specific mutagenesis of the strictly conserved amino acids, Gly²⁰³ or Gln²⁰⁹, within the corresponding region in rat IGFBP-5, reduced IGF-I binding affinity by 8- and 6-fold, respectively (123). In contrast, mutagenesis of adjacent basic amino acid residues in the equivalent region of human IGFBP-5 (amino acids 201–218) did not alter IGF-I binding affinity, although ability to interact with ECM was affected (see below and Ref. 124). The IGF binding properties of the C-terminal domain, itself, have yet to be tested thoroughly, although a recent study indicated that a natural C-terminal fragment of human IGFBP-2 retained partial IGF-binding activity (125). This observation is consistent with an earlier study where it was observed that a proteolyzed rat IGFBP-2 fragment containing half the midregion and the C-terminal domain showed similar reduced IGF binding compared with full-length rat IGFBP-2 (126). In contrast, a synthetic peptide corresponding to half the midregion and C-terminal domain of IGFBP-4 (as defined in this review; see Fig. 2), His¹²¹-Glu²³⁷, did not show detectable IGF binding (radioactivity in the bands from ligand blots were quantitated) (119); neither did an analogous C-terminal peptide, Asp¹³⁵-Phe²⁴⁶, from IGFBP-5 demonstrate IGF binding (65). A comparable region in IGFBP-3, on the other hand, had demonstrable IGF binding capabilities by solution assays (115) and by ligand blot (G. R. Devi, D.-H. Yang, R. G. Rosenfeld, and Y. Oh, unpublished).

The midregion of the IGFBPs does not appear to bind IGFs; its contribution to the high-affinity binding of IGFs is likely to relate to its ability to promote a tertiary structure, which permits optimal relationships between the N-terminal and C-terminal domains.

2. Effect of posttranslational modification on IGF binding. Limited data are available on the effects of posttranslational modification of IGFBPs on IGF binding. Results so far indicate that neither glycosylation nor phosphorylation appear to have much influence on the IGF binding affinities of IGFBPs (83, 84). The exception is the phosphorylation of human IGFBP-1, where it has been shown that phosphorylation enhances IGF binding by at least 5-fold (78). Similar results were not observed with rat IGFBP-1 (82).

3. Other structure-function associations of IGFBPs. The regions of IGFBPs that are involved in functions unrelated to IGF binding appear to be predominantly in the mid- and C-terminal domains. To date, the only function clearly associated with the N-terminal domain is IGF binding, and more recently, insulin binding (110, 118). This does not rule out other potential functions for the N-terminal domain, either alone or in concert with other IGFBP domains. Lalou et al. (108), for example, have reported that IGFBP-3 (residues 1–95) inhibits cell replication. For the midregions, aside from simply acting as a "hinge" between the N-terminal and C-terminal domains, the fact that these regions are posttranslationally modified suggests that specific functions, as yet undefined, may be associated with this region. For IGFBP-3, the midregion appears to be involved in specific membrane association (127). Interestingly, the proteolytic sites for a number of IGFBP proteases are found in the midregion. It is possible that individual characteristics of each IGFBP reside in these nonconserved regions.

In the C-terminal domain, more information is available about functions other than IGF binding. The C-terminal domain of IGFBP-3, irrespective of its ability to bind IGF-II, has been shown to be essential for interactions with the acid-labile subunit (64, 116), most likely through the IGFBP-3 basic region, Lys²²⁸-Arg²³² (116). A recent report suggested that in addition to ALS, IGFBP-3 can interact with other high mol wt proteins found in human serum (128); whether these interactions are through the C-terminal region or midregion is unknown. Interestingly, IGFBP-5 also forms a ternary complex with ALS and IGFs (129). Since the C-terminal domain between IGFBP-3 and -5 is highly similar (54%), particularly in the sequences spanning the basic region (see Fig. 2), IGFBP-5 presumably also interacts wth ALS through this domain.

The other notable motifs in the C-terminal domains are the RGD sequence found in analogous positions in IGFBP-1 (amino acid residues 221–223) and IGFBP-2 and the highly basic heparin-binding sequences found in the thyroglobulin type I domain in IGFBP-3, -5, and -6. The RGD motif in IGFBP-1 was shown by Jones et al. (14) to interact with integrins, which are a large family of heterodimeric cell adhesion receptors involved in both cell-cell and cell-ECM interactions (130). It has been hypothesized that interactions of IGFBPs with the ECM, via the integrins, may allow the IGFBPs to provide a reservoir of IGFs (26). In IGFBP-3, the heparin-binding motif can associate with glycosaminoglycan-containing molecules, like proteoglycans found on cell surfaces and in ECM (116, 131). While the consequences of IGFBP-3 interacting with glycosaminoglycan are unclear, these interactions may enhance localization of IGFBP-3 to the cell suface and, perhaps, the ECM (131). The same motif in IGFBP-5 and IGFBP-6 may have similar functions and, in fact, there is strong evidence that the highly basic region surrounding the heparin-binding motif (Arg²⁰¹-Arg²¹⁸) mediates binding of IGFBP-5 to osteoblast cells (16), to ECM (124, 132, 133), and to mesangial cell surface (134). The highly basic regions from all three IGFBPs (IGFBP-3, -5, and -6) are capable of specifically inhibiting IGFBP-4 degradation, and the inhibition of IGFBP-4 degradation is abrogated by IGFs (98). The mechanisms by which this inhibition is mediated are not understood, but since IGFBP-3, -5, and -6 are not themselves substrates for the IGFBP-4 protease, one hypothesis is that the highly basic region in these IGFBPs may act as a protease inhibitor (98). The physiological ramifications of inhibiting IGFBP-4 degradation are unclear, although IGFBP-4 is known to be a potent inhibitor of IGF actions and proteolysis of IGFBP-4 could, therefore, potentiate IGF actions.

One of the most intriguing observations made within recent years has been evidence for the targeting of IGFBP-3 and IGFBP-5, but not IGFBP-1 or -2, to the nucleus. Although there are no definitive consensus amino acid sequences for nuclear localization signals (NLS) (135), many proteins do contain sequences rich in basic amino acids similar to the NLS (PKKKRKV) of SV40 large T antigen. Potential NLS sequences in IGFBP-3 and in IGFBP-5 were first noted in 1994 by Radulescu (136). It was not until 1997, however, that the evidence supporting nuclear IGFBP-3 was published (137, 138). Not only was endogenous IGFBP-3 clearly found in the nucleus of lung cancer cells (138), but labeled recombinant IGFBP-3 added exogenously to wounded opossum kidney cells was transported into the nucleus, whereas in resting cells, IGFBP-3 was internalized and accumulated in the endosomal compartment (137). Intriguingly, IGFs bound to IGFBP-3 can also localize to the nucleus. In human keratinocytes, nuclear IGFBP-3 was detected in cells undergoing division (139). Recent in vitro studies have demonstrated that both recombinant IGFBP-3 and IGFBP-5, but not IGFBP-1 or -2, can translocate from the extracelluar compartment to the nucleus in rapidly dividing human breast cancer cells (140). Site-specific mutagenesis confirmed that the putative NLS in IGFBP-3 is the predicted basic sequence in the C-terminal domain (Lys²¹⁵-Arg²³²) (139). The biological significance of translocating IGFBPs into the nucleus is unclear at present but is consistent with potential IGF-independent actions of some IGFBPs (see below).

C. Biological functions of IGFBPs
The detailed biological functions of IGFBPs have been well reviewed in recent years (10, 13, 24, 25) and will not be reiterated in this review. The aim in this section is to put into perspective the correlations made between the structure of IGFBPs and their functions in the context of IGF-dependent vs. IGF-independent actions of IGFBPs.

1. IGF-dependent actions of IGFBPs. The term "IGF-dependent" functions of IGFBPs has been used to define functions of IGFBPs, both positive and negative, that are directly linked with IGF bioactivities (10, 13, 25, 26). Since IGFBPs are well established secreted proteins, this inevitably meant that the focus has been on the extracellular sequestration of IGFs by IGFBPs, and the effects this sequestration has on the consequent loss of interactions between IGFs and the type I IGF receptor. There is a plethora of in vivo and in vitro studies describing and supporting this mechanism of IGFBP action. Most recently, the approach taken has been to directly test this hypothesis by generating recombinant mutated forms of IGFBPs with reduced affinities for IGFs, and subsequently testing whether these mutants have effects on IGF bioactivities. This was most clearly demonstrated in the case of IGFBP-4, an IGFBP known to inhibit the mitogenic effect of IGFs on bone cell growth. Mutations in human IGFBP-4 that greatly reduced its affinity for IGF-II resulted in an inability of the mutant IGFBP-4 to inhibit IGF-II-induced human osteoblast proliferation (119).

An extension of the IGF-dependent actions of IGFBPs is investigations into the mechanisms of IGF release from IGFBPs. Reducing affinity for IGFs is an obvious mechanism for the release of IGFs and is achieved by proteolysis of IGFBPs, alteration in phosphorylation status of IGFBP-1 (78), and perhaps also by IGFBP conformational changes, such as via binding of the IGFBPs to ECM and/or to the cell surface. Molecular evidence for the importance of IGFBP proteolysis to IGF-dependent actions was provided recently by site-specific mutagenesis of the proteolytic site, resulting in enhanced IGFBP growth-inhibitory effects (141, 142). In contrast, there is yet to be evidence for release of IGFs by conformational change of IGFBP-IGF interactions, although interactions between IGFBPs and ECM and/or cell surface (see above) would support this postulate. An alternative hypothesis, suggested in a recent study, implicate a physical occlusion effect based on the observation that, in IGFBP-5, the regions involved in IGF and ECM interactions overlap (123).

In contrast to the extracellular effects of IGFBP on IGF-type-I IGF receptor complexes, intracellular effects of IGFBPs, particularly any effects on type-I IGF receptor signaling pathway, have yet to be addressed. In light of very recent data indicating the internalization and nuclear localization of IGFBP-3 and of IGFBP-5, it may be necessary to redefine "IGF-dependent" actions.

2. IGF-independent actions of IGFBPs. Given the classical definition of IGF-dependent actions of IGFBPs, IGF-independent actions of IGFBPs are defined as biological effects exerted by IGFBPs that involved neither binding of IGFs nor activation or inhibition of the type I IGF receptor. There has been a steady accumulation of data supporting the existence of IGF-independent actions for IGFBP-3 and IGFBP-5 (13) and limited data for IGFBP-1 (14). The recent demonstration of IGFBP-3 and -5 translocation to, and localization in, the nucleus support the concept that these two IGFBPs have functions unrelated to direct IGF actions. Presumably, these IGF-independent functions are through the C-terminal domains and perhaps also the midregions of the IGFBPs (see Section II.B.3. above).

As initially demonstrated by Oh et al. (18) in breast cancer cells, the epithelial growth-inhibitory actions of IGFBP-3 are mediated through specific binding of IGFBP-3 to cell surface molecules that are not type I IGF receptors. The purification and cDNA cloning of a specific IGFBP-3 receptor, however, remains elusive, but, using the yeast two-hybrid system, cDNAs encoding IGFBP-3 interacting proteins have been obtained (Y. Oh, unpublished). Interestingly, a recent report suggests that the type V TGFß receptor could be the putative IGFBP-3 receptor (143, 144) and that IGFBP-4 and IGFBP-5 may also interact with this receptor (144). Presence of this receptor, however, has not been convincingly demonstrated in breast cancer cells (Y. Oh, unpublished).

A similar sequence of events has led to the conclusion that IGFBP-5 also has biological actions that are IGF independent (16). Supporting this hypothesis, a novel, putative IGFBP-5 membrane receptor, a 420-kDa membrane protein, was very recently purified from osteoblast cells (145). Although not fully characterized and the cDNA not cloned, it would appear, at least in vitro, that the binding of IGFBP-5, through the basic region in its C-terminal domain, to the receptor stimulated phosphorylation of the receptor.

D. Genomics of IGFBPs
1. Chromosomal locations of IGFBPs. The genomic locations of all human IGFBPs are known and are summarized in Table 1. Interestingly, the genes for IGFBP-1 and IGFBP-3 not only reside on the same chromosome, at the locus 7p14-p12, but are only 20 kb apart, with transcription orientated in a tail-to-tail configuration (146). IGFBP-2 and IGFBP-5 constitute another gene pair, located 20–40 kb apart on chromosome 2q. Based on amino acid similarity analysis, IGFBP-1 is more closely related to IGFBP-2 than to IGFBP-3, which, in turn, is more closely related to IGFBP-5. IGFBP-4, found on chromosome 17q12–21.1, is more closely related to IGFBP-1 and -2, whereas IGFBP-6, located on chromosome 12q13, appears to be the most divergent of the IGFBPs. The similarity in configuration of the human IGFBP genes, especially the gene pairs, is striking, and, together with analysis of the protein sequences, has led to the hypothesis that a tandem gene duplication and inversion occurred early in the evolution of the IGFBPs [one suggestion is that IGFBP-6 is the proto-IGFBP, (58)], and subsequent gene duplications primarily involved partial chromosome duplication (see review in Ref. 58).

An intriguing observation that was made in the analysis of chromosomal locations of the human IGFBP genes is that the genes appear to co-map with genes encoding homeoboxes (HOX) and epidermal growth factor receptors [see review by Reinecke and Collet (58)]. Homeobox is a conserved element of 180 bp that is found in all homeotic (and also nonhomeotic) genes. The importance of homeotic genes is that they are the master control genes that regulate development of higher organisms. Thus, by association, the inference is that IGFBPs are important and fundamental proteins in development. The evolutionary implication is that there may be an association between the evolution of the vertebrate homeobox genes, the epidermal growth factor receptors, and the IGFBPs. Since this area is summarized in a very recent, comprehensive review on IGF phylogeny by Reinecke and Collet (58), readers are referred to that review for more details.

2. IGFBP gene structures. The gene structures of human IGFBPs are highly similar, although the sizes of the genes vary from 5.7 kb for IGFBP-1 to 33 kb for IGFBP-5 (Table 1), due to variations in the sizes of the introns. All of the IGFBPs are encoded by four exons, with the exception of IGFBP-3, which carries an extra exon, exon 5, that is not translated. The corresponding exons among the IGFBPs are equivalent in size, with exon 1 less than 600 bp, exons 2 and 3 both small exons of less than 230 bp, and exon 4 more variable in size. There is a stiking correlation between these exons and the three protein domains of IGFBPs. The N-terminal domain, as defined in Fig. 2, is encoded within exon 1 in all of the IGFBPs, as is the 5'-untranslated region and a few amino acids of the midregion. Exon 2 encodes for the nonconserved midregion. Both exon 3, which ends precisely at the invariant Gln (Q) residue in the thyroglobulin domain, and exon 4 encode for the conserved C-terminal domain. The fact that the N-terminal domain is contained within one exon strongly supports the concept of an IGFBP superfamily, as will be discussed in Section IX.

	III. Mac25

Top Abstract I. Introduction II. IGFBP Family III. Mac25 IV. CCN Family V. L56 VI. ESM-1 VII. Structural Relationships... VIII. Functional Relationships... IX. Evolutionary Relationships... X. Summary References

 
A. IGFBP-rP1 (MAC25/TAF/PSF)
The (human) Mac25 was the first protein proven to be functionally related to the IGFBPs (22, 37). It was, therefore, provisionally named IGFBP-7 (22), and, subsequently, redesignated IGFBP-rP1 (Table 2). The gene for human IGFBP-rP1 has been localized to chromosome 4q12–13 (147). A mouse homolog, sharing 87.5% nucleotide identity and 94.4% similarity with human IGFBP-rP1, has been described (148). Three groups independently identified the human IGFBP-rP1 protein, and each has continued to use its own designations. Without a doubt, however, they are the same protein. The structural relationship to the IGFBPs was initially noted by Murphy et al. (40), who were the first to identify the putative protein, deduced from the cDNA clone, mac25. Akaogi et al. (42) and Yamauchi et al. (41) independently purified and characterized the protein, which they designated TAF and PSF, respectively. Oh et al. (22) synthesized the Mac25 protein in a baculovirus system, demonstrated its ability to bind IGFs, and provisionally named the protein IGFBP-7, later to be redesignated IGFBP-rP1. Structurally, the region of similarity of IGFBP-rP1 (Mac25/TAF/PSF) to IGFBPs is confined to the N-terminal domain (see Section VII). Functionally, the protein appears to have multiple roles, including the ability to bind IGFs and insulin (see Section VIII), but the physiological significance of this protein is still largely unknown. The structural and binding characteristics of IGFBP-rP1 will be discussed in more detail in Sections VII–VIII. Below, a historical perspective of IGFBP-rP1 will be presented.

1. Mac25. Murphy et al. (40) employed subtractive hybridization to search for genes whose expression were altered in meningioma cell lines, compared with normal leptomeningeal cells. The cDNA clone they designated mac25 was found to be preferentially expressed in normal leptomeningeal cells, compared with meningiomas. mac25 Expression in breast carcinomas has also been examined and it was noted that expression may be related to the estrogen receptor status of the cancer cells: that is, the presence of estrogen receptor (ER) mRNA appeared to be negatively correlated to expression of mac25 mRNA. A more extensive examination of mac25 expression between ER+ vs. ER- breast cancer cell lines indicated that some ER- cancer cells also did not express mac25 mRNA (147). mac25 cDNA was identified by differential display, as one of the genes overexpressed in senescent normal human mammary epithelial cells (HMEC) (147), and as one of the genes that was down-regulated in breast carcinomas (149). Furthermore, there appeared to be a significant (5/10 tumor tissues examined) loss of heterozygosity in the mac25 gene in breast tumors (149). Consistent with these observations was a recent in situ hybridization study of IGFBP-rP1 (mac25/TAF/PSF) expression in normal prostate tissue vs. prostate tumors, where a marked decrease in IGFBP-rP1 expression was associated with increasing malignancy (150). Interestingly, a malignant prostatic cell line stably transfected with IGFBP-rP1 cDNA was shown to be poorly tumorigenic in both in vitro and in vivo assays, when compared with cells stably transfected with empty vector, suggesting a potential tumor-suppressive function for IGFBP-rP1 (151).

Expression of IGFBP-rP1 is regulated by growth factors. In midpassage HMEC (147), breast cancer cells Hs578T (Y. Oh, unpublished), and in immortalized prostate epithelial cells (P69) (150), IGFBP-rP1 expression is up-regulated by retinoids. TGFß also up-regulates IGFBP-rP1 expression, both at the mRNA and protein levels, in Hs578T and P69 cells (150). Whether IGFBP-rP1 mediates the epithelial growth-inhibitory effects of TGFß and retinoic acids has yet to be determined.

One recent study indicated that mac25 mRNA expression is higher in dividing mouse myoblasts than in nondividing, undifferentiated myotubes (152), suggesting that IGFBP-rP1 may play a role in differentiation of muscles. IGFBP-rP1 may also play a role in differentiation of rat osteoblast cells, as PTH and the glucocorticoid, cortisol, both increase IGFBP-rP1 mRNA (153, 154).

2. TAF. TAF, a 30-kDa protein isolated from the conditioned media of a human bladder carcinoma cell line, was so named because it was tumor derived and promoted cell adhesion activity. Initial studies of the purified protein showed that it promoted the attachment and spreading of rat liver cells and human endothelial cells, but did not stimulate endothelial cell growth (42). Subsequent structural analysis of the purified protein and its cDNA indicated identity with PSF (Ref. 41 and see below) and close similarity with Mac25 (37, 42). The discrepancies in primary amino acid sequence between PSF and Mac25 will be discussed below. A monoclonal antibody generated against a C-terminal peptide of purified TAF was used to determine the distribution of TAF in various human cancer tissues (155). Results, based solely on immunohistochemical staining of tissues using this monoclonal antibody, indicated that TAF appears to specifically accumulate in new blood vessels in various human cancer tissues, but not in those of normal tissues, and also in capillary tube-like structures of cultured vascular endothelial cells. These observations, in conjunction with an affinity of TAF for type IV collagen, that was inhibitable by heparin, suggested that TAF may be involved in the formation of new capillary vessels by vascular endothelial cells. This led to the suggestion of renaming the protein "angiomodulin" (155). TAF, at high concentrations (1 µg/ml), also appears to be capable of stimulating and enhancing IGF and insulin-mediated fibroblast cell growth (37). The seemingly diverse functions of IGFBP-rP1 (TAF/Mac25) can be reconciled by the fact that expression and function of this protein are most likely cell type specific, but further studies are clearly necessary.

3. PSF. Yamauchi et al. (41) was the third group to purify "IGFBP-rP1." Their interest was in an activity found in plasma that stimulated prostacyclin production in endothelial cells but that was reduced in patients with diabetes mellitus (156, 157). Prostacyclin is a vasodilator and inhibitor of platelet adhesion and aggregation, whose synthesis is stimulated by many factors, including proteases such as thrombin (158). The prostacyclin-stimulating activity in serum was relatively heat stable, acid labile, and nondialyzable (156, 157). A similar activity was detected in the conditioned media of human diploid fibroblast cells (41). Purified PSF was approximately 31 kDa on SDS-PAGE and was able to stimulate prostacyclin production in endothelial cells at a concentration as low as 10 ng/ml (41). PSF was subsequently identified to be the same protein as Mac25 and TAF (159). An antibody generated to a synthetic PSF C-terminal peptide indicated that PSF is expressed in arterial endothelial cells and in smooth muscle cells of human tissues (160, 161).

Although PSF, Mac25, and TAF are the same protein, there are a few discrepancies in the published cDNA nucleotide sequences between PSF and Mac25. Four nucleotides differ in the signal peptide region, resulting in three amino acid substitutions; one nucleotide differs in the N-terminal domain of the molecules resulting in an Arg for Mac25, and Lys for PSF, both basic amino acids. Finally, the major difference is an extra nucleotide found near the C terminus of Mac25 that results in a stop codon within 5 amino acids of the insertion. In PSF, the lack of this one extra nucleotide generated a completely different sequence and extended the sequence by 10 amino acids. Of the two sequences, the PSF sequence is most likely the correct sequence, as it concurs with amino acid sequencing data (41), and with the cloned genomic IGFBP-rP1 gene (V. Hwa and R. G. Rosenfeld, unpublished and Ref. 159). In addition, TAF has an identical sequence with PSF (37). It is quite likely, therefore, that the discrepancy between the Mac25 and PSF cDNAs was the result of errors in the sequencing of mac25.

In summary, IGFBP-rP1 appears to be involved in diverse biological functions, from regulation of epithelial cell growth, to stimulation of fibroblast cell growth, to stimulation of prostacyclin production in endothelial cells. Further, it can associate with type IV collagen (37) and can bind IGFs and insulin (22, 37, 118). Its expression is regulated, not only by specific growth factors such as IGFs, PTH, cortisol, TGFß, and retinoic acid, but by unknown factors involved in the progression of tumorigenesis, in senescing epithelial cells, in diabetes, and in vascular development. IGFBP-rP1 has thus been hypothesized to have a significant biological role in senescence, tumor suppression, and vascular disease; these multiple effects will need to be substantiated.

	IV. CCN Family

Top Abstract I. Introduction II. IGFBP Family III. Mac25 IV. CCN Family V. L56 VI. ESM-1 VII. Structural Relationships... VIII. Functional Relationships... IX. Evolutionary Relationships... X. Summary References

 
A group of highly related, cysteine-rich proteins was recently identified and shown to contain a N-terminal domain that has significant structural similarity with the N terminus of IGFBPs (21). Coined the CCN family by Bork (21), they include human CTGF, a growth factor-inducible immediate-early gene; cyr61 and its chicken ortholog cef10; a potential oncogene, nov (for nephroblastomas overexpressing gene); and, more recently, three genes (WISP-1, WISP-2, and WISP-3) that are up-regulated in Wnt-1-transformed cells and are aberrantly expressed in human colon tumors (46); [see also recent review by Brigstock (162) and Lau and Lam (163)]. The first members of this family to be described were cef10, an immediate-early gene, from chicken, detected after induction by the viral oncogene pp60^v-src (164), and a related protein from the mouse, cyr61 (165). Structurally, the proteins share an overall similarity of 46% and range in sequence length from 349 to 381 amino acids (prepeptides). There are 38 completely conserved cysteines in the mature proteins. Each of the proteins is comprised of four domains: the N-terminal domain of the IGFBP family, the Von Willebrand factor type C repeat (VWC), the thrombospondin type I repeat, and the C-terminal (CT) domain. A central "fifth" domain is variable among the CCN proteins. Detailed structure and function analysis of the N-terminal domain, particularly in relationship to the IGFBP family, will be discussed in Sections VII and VIII. The specific functions of the latter three domains are still unclear, but are believed to be involved in dimerization, oligomerization, and interactions with the ECM through heparin binding regions. The biological roles of the CCN family are under investigation and appear to be important in the regulation of cellular proliferation. Below, a summary of each CCN family member will be described.

A. IGFBP-rP2 (CTGF)
Connective tissue growth factor was the first human protein of the CCN family to be described (44) and the second protein tested and shown to be related to the IGFBPs (23). CTGF (IGFBP-rP2) was purified from media conditioned by human umbilical vein endothelial cells (HUVEC) as the major secreted protein that was immunoreactive with antibody against platelet-derived growth factor (PDGF) (44). Interestingly, it was subsequently shown that the two proteins did not share any regions of amino acid sequence similarity (44). The purified protein is a monomer of 36–38 kDa that demonstrated mitogenic activity and chemotactic activities for fibroblastic cells. Furthermore, the protein could bind to the cell surface of endothelial cells and was competitively displaced by purified PDGF (44). Subsequent studies have demonstrated that a 10-kDa proteolytic fragment of CTGF, corresponding to the CT domain, can associate with the cell surface of fibroblasts and is capable of inducing cell proliferation (166, 167, 168). These activities are inhibitable by heparin and clearly do not involve the PDGF receptor (168). The specific cell surface receptor for CTGF is not known, although a recent study has suggested that, at least for human chondrocytes, CTGF, which participates in endochondral ossification (169), interacts with an uncharacterized 280-kDa membrane protein (170).

The chromosomal location for the CTGF gene has been determined to be 6q23.1 and proximal to the oncogene, c-myb (45). The cDNA for CTGF encodes a 38-kDa protein with two potential glycosylation sites and hybridizes to a single 2.4- kb mRNA species in Northern blots (44). Expression of CTGF is regulated in a manner consistent with an immediate-early gene. In fibroblast cells, it is selectively up-regulated by TGFß, a potent stimulator of fibroblast cell proliferation and a critical factor in cell regeneration and wound repair, within an hour of exposure to TGFß (171). Unlike other immediate-early genes, however, short-term exposure to TGFß induces prolonged CTGF mRNA expression, for up to 36–48 h (171, 172). A novel TGFß response element found in both the human and murine CTGF promoters, but absent in other genes regulated by TGFß, suggests that regulation of CTGF gene expression may function by a mechanism distinct from other TGFß-regulated genes (172). Importantly, some of the biological effects of TGFß on fibroblast and endothelial cells appear to be mediated by the up-regulated CTGF protein (173, 174, 175, 176, 177).

Since the discovery and initial characterization of CTGF, there has been considerable research into the regulation, biology, and clinical implications of this protein, which will be briefly summarized here. The readers are referred to recent reviews for more comprehensive coverage (162, 174, 178, 179, 180). Because of clinical implications in fibrosis and mucosal repair, IGFBP-rP2 (CTGF) research has focused on its role(s) in fibroblast and endothelial systems. However, IGFBP-rP2 may also be important for epithelial growth, as recent data suggest that TGFß, which is inhibitory for epithelial cell proliferation, up-regulates IGFBP-rP2 expression (mRNA and protein) in mammary cells (181). Interestingly, in situ studies of mammary tumors have suggested that IGFBP-rP2 mRNA expression is exclusively in the fibrous stroma (182). The implications are unclear at present.

B. IGFBP-rP3 (NovH)
The gene nov was first discovered in myeloblastosis-associated virus type I-induced avian nephroblastomas (183). Expression of nov was elevated in these nephroblastomas, compared with normal adult avian kidney cells, suggesting that nov may be a protooncogene. Supporting this concept, human novH (184) expression was shown to be elevated in Wilms tumors of the stromal type, which histologically are similar to avian nephroblastomas (45, 184). Of particular interest is that the novH gene maps to chromosome 8q24.1, proximal to c-myc (45), a region often involved in chromosomal abnormalities associated with human tumors, including Wilms tumor. The expression of novH appears to be inversely correlated with the expression of the tumor suppressor gene, WT1 (185, 186), whose inactivation is postulated to participate in the etiology of Wilms tumors. Indeed, recent studies indicate that WT1 does transcriptionally down-regulate novH expression (187).

Aside from its oncogenic potential, novH and nov are involved in other biological processes. For example, the effects of nov on chicken embryo fibroblast (CEF) cells are quite different: overexpression of nov inhibits fibroblast cell growth, although, interestingly, overexpression of an N-terminally truncated form of nov (which deleted the N-terminal domain) induced cellular transformation of the fibroblast cells (183). Consistent with the growth-inhibitory effects observed in CEF cells, it was demonstrated that nov was expressed only in quiescent CEF cells, and that transformation of CEF by p60^v-src oncogene down-regulated expression of the nov gene (188). In humans, novH is associated with the developing kidney, where observations suggest that NovH protein is stably accumulated in embryonic kidney in glomerular podocytes undergoing differentiation, and, after birth, the persistence of high levels of NovH protein may be required for maintenance of podocyte structure and/or for specfic podocytic functions (186).

The structures of mammalian and nonmammalian Nov proteins are similar to that of Cef-10 (164), Cyr61 (165), Fisp-12 (189), and CTGF (44). The human NovH protein, deduced from the cloned cDNA, indicated that the cDNA encodes a putative 39-kDa secreted polypeptide (184, 186). Immunoblots of biological fluids and media conditioned by various cell lines indicate that NovH is at least 44 kDa and is N glycosylated (38, 186). Interestingly, intracellular isoforms of NovH were detected and appeared to be less stable than the extracellular form (186). Like all members of the CCN family, Nov/NovH consists of four domains (21), of which the first domain (after the signal peptide) is an IGFBP N-terminal domain, leading to the redesignation of NovH as IGFBP-rP3. Recently, Burren et al. (38) demonstrated that IGFBP-rP3 could bind IGF with low affinity, similar to that detected for IGFBP-rP1 and -rP2. More structural and functional information will be given in Sections VII and VIII below.

C. IGFBP-rP4 (Cyr61)
The cyr61 gene was originally identified in mouse 3T3 fibroblasts as an immediate-early gene that was rapidly activated by serum, PDGF, fibroblast growth factor, and 12-O-tetradecanoylphorbol-13-acetate (165). Unlike other immediate-early genes, but similar to CTGF, induced cyr61 mRNA persists for a considerable time after induction. The human cyr61 gene, cloned recently, and mapped to chromosome 1p22-p31 (43), is similar to mouse cyr61 in both structure [sharing 85% amino acid similarity (190)] and function. Both cyr61 mRNAs are not detected in quiescent fibroblasts, but are abundant in logarithmically growing cells and serum-stimulated cells (165, 190). Human cyr61 mRNA was also recently shown to be up-regulated by factors important for osteoblast function and differentiation, such as 1,25-dihydroxyvitamin D₃, EGF, tumor necrosis factor (TNF), and interleukin-1 (191).

Cyr61 protein, like the rest of the CCN family, is a secreted protein. However, unlike the other members, it is not readily detected in conditioned media of cell lines examined, apparently because it associates with the ECM and cell surfaces, most likely through its heparin binding regions (192, 193); (see recent review in Ref. 163). In fact, it was recently demonstrated that Cyr61 protein adheres to HUVEC cells through integrin _vß₃ (194). This adherence of Cyr61 to HUVEC cells may be a mechanism by which Cyr61 promotes the attachment and spreading of endothelial cells (193). Support for this hypothesis comes from recent studies in which purified Cyr61 was shown to promote angiogenesis through an _vß₃-dependent pathway (195). Cyr61 also promotes the adhesion of fibroblasts and epithelial cells (193, 196), induces chemotaxis of fibroblasts (193), enhances growth factor-stimulated DNA synthesis in both fibroblast and endothelial cells (176, 193, 196), and plays a role in chondrogenesis (197).

There is recent evidence that Cyr61 may also play a role in tumorigenesis. Stably cyr61-transfected gastric adenocarcinoma cells demonstrated increased tumor growth when tested in a nude mouse model, suggesting that Cyr61 promoted tumor growth (195). However, Cyr61 protein expression was down-regulated in prostate carcinomas (198). Although the role of Cyr61 in cancer is unclear, it is of interest to note that the human cyr61 gene is mapped to chromosome 1p22-p31 (43), as abnormalities of chromosome 1p have been shown to correlate with breast cancer (199), neuroblastoma (200), and pheochromocytoma (201).

The human cyr61 cDNA encodes a 381-amino acid protein rich in cysteine and proline residues (43, 190). There are two distinct mRNA species, a major one at 2.5 kb and a minor one at about 4.0 kb, which are believed to be either alternatively spliced transcripts or transcripts with different polyadenylation signals (43). Cyr61 is structurally consistent with other members of the CCN family, but has yet to be tested for its ability to bind IGFs. It is predicted, however, that it will prove capable of binding IGFs with low affinity, similar to that observed with CTGF and NovH.

D. New members
1. IGFBP-rP7 (rCOP-1/WISP-2/CTGF-L). The gene rCop-1 was very recently identified by differential display from rat embryo fibroblasts (REFs) as one of three genes whose expression was lost specifically upon cell transformation (47). By sequence comparison, it appears that rCop-1 belongs to the CCN family of proteins, but, unlike the other CCN proteins, it only has the first three conserved protein domains and lacks the last domain (the CT domain). The cDNA encodes a unique 250-amino acid protein with a signal peptide and is detectable as a single 1.7 kb transcript, thereby ruling out the possibility that rCop-1 mRNA is a result of alternative splicing of other CCN transcripts. However, until the gene for rCop-1 is fully characterized, this possibility cannot be completely ruled out. Although rCop-1 has a signal peptide, it is not detectable in conditioned media of fibroblast cells, nor is it associated with the ECM, like Cyr61, perhaps due to loss of the CT domain. Rather, it seems to be predominantly cell surface associated. Not only is the structure of rCop-1 distinct from the other CCN proteins, but the pattern of expression of rCop-1 mRNA indicates that its regulation may be through different mechanisms than for the rest of the CCN family. It is not an immediate-early gene, like the CTGF and cyr61 genes; it is not serum inducible, and, in fact, expression is inversely related to that of cyr61 in normal fibroblast cells. Overexpression of rCop-1 in transformed cells reduced tumorigenicity and increased cell death. In primary cultures of rat and mouse fibroblasts, the rCop-1 gene was detected only when cells became senescent during passage in culture.

A human ortholog of the rCop-1 gene, WISP-2 (46) and CTGF-L (39), has been subsequently identified and redesignated IGFBP-rP7 in this review. The gene maps to human chromosome 20q12–20q13 (46) and appears to be linked to tumorigenesis (46) as well as to the modulation of bone turnover (39). WISP-2/IGFBP-rP7 mRNA expression was reduced in human colon tumors and is one of three WISP genes tha