REVIEW

Proteoglycans of the extracellular environment: clues from the gene and protein side offer novel perspectives in molecular diversity and function

RENATO V. IOZZO’ AND ALAN D. MURDOCH Department of Pathology, Anatomy, and Cell Biology, and the Jefferson Cancer Institute, Thomas Jefferson University, Philadelphia, Pennsylvania 19107, USA

ABSTRACT This review focuses on the extracellu- and multifunctional molecules of the animal kingdom. lar . Special emphasis is placed on the They carry a variety of structural units that vary from structural features of their protein cores, their gene simple linear sugars to the most highly charged, sulfated organization, and their transcriptional control. A polysaccharide in nature: heparin. They provide struc- simplified nomenclature comprising two broad tural constraints, function as growth-supportive or sup- groups of extracellular proteoglycans is offered: the pressive molecules, possess adhesive and anti-adhesive small leucine-rich proteoglycans or SLRPS, pro- properties, act as major biological filters, promote angio- nounced “slurps,” and the modular proteoglycans. genesis, induce neurite outgrowth, and bind, store, and The first group encompasses at least five distinct deliver growth factors to target cells during normal devel- members of a gene family characterized by a central opment and in pathologic states. This review centers on domain composed of leucine-rich repeats flanked by the extracellular proteoglycans, those molecules whose two cysteine-rich regions. The second group consists broad determining roles are to be secreted into the of those proteoglycans whose unifying feature is the pericellular environment and to provide bridging informa- assembly of various protein modules in a relatively tion to the cells. Special emphasis is placed on the struc- elongated and often highly glycosylated structure, ture of the protein cores, the gene organization, the This group is quite heterogeneous and includes a transcriptional control, and the functional implications distinct family of proteoglycans, the “hyalectans,” derived from the design of these molecules. First, we will that bind hyaluronan and contain a C-type lectin offer a simplified nomenclature of these motif that is likely to bind carbohydrates, and a less groupings. Then we will discuss the structure/function re- distinct group that contains structural homologies lationships of some paradigmatic proteoglycans, and ex- but lacks hyaluronan-binding properties or lectin- amine novel aspects of proteoglycan biology derived from like domains.-Iozzo, R. V., Murdoch, A. D. Proteo- recent genetic and structural studies. glycans of the extracellular environment: clues from the gene and protein side offer novel perspectives in A SIMPLIFIED NOMENCLATURE FOR molecular diversity and function. FASEB J. 10, PERICELLULAR PROTEOGLYCANS 598-614 (1996) The growth of the proteoglycan gene family has been staggering in the past decade. To date, more than 25 dis- Key Words: exlracellular matrix ‘ leucine-rich repeats modular genes tinct genes scattered throughout the mammalian genome code for protein cores that carry at least one glycosamino- UNDER PHYSIOLOGICAL CONDITIONS, the adhesive force glycan (GAG)2 chain, the hallmark of proteoglycans. In between two proteoglycans of the marine sponge Micro- addition, different proteoglycans exist as structural van- ciona pro1fera is about 400 pN. This strong bond is me- diated by the homophilic and polyvalent, 1To whom correspondence and requests for reprints should be ad- calci urn-dependent interactions between two adhesive dressed, at: Department of Pathology, Anatomy ,nd Cell Biology, Room 249, Jefferson Alumni Hall, Thomas Jefferson University, 1020 Locust proteoglycans (1). Thus, a single pair of these molecules Street, Philadelphia, PA 19107, USA. can hold the weight of 1600 cells. That’s incredible! If 2Abbreviations: bFGF, basic fibroblast growth factor; CRP, complement one had to consider the billions of proteoglycan mole- regulatory protein; CS, ; DS, ; EGF, cules contained in a gram of cartilage or tendon, for in- epidermal growth factor; FGF, fibroblast growth factor; GAG, glycosami- stance, their importance as a major intercellular glue noglycan; HBR, hyaluronan binding region; HS, ; hyalec- would be better appreciated. More astonishing is the fact tan, hyaluronan- and lectin-binding proteoglycan; Ig, immunoglobulin; KS, ; LDL, low density lipoprotein; LRR, leucine-rich that proteoglycans have evolved during the last 50 mil- repeat; N-CAM, neural cell adhesion molecule; PG, proteoglycan; RI, lion years to reach an unprecedented level of sophistica- ribonuclease inhibitor; SLRP, small leucine-rich proteoglycan; TGF-, tion. Proteoglycans comprise some of the most complex transforming growth factor beta.

598 Vol. 10 April 1996 0892-6638/96/001 0-0598/$01 .50. © FASEB REVIEW ants, further increasing the diversity of this class of mac- alternating hydrophobic and hydrophilic amino acid resi- romolecules. The classification presented in Table 1 is dues that harbor multiple amino acid repeats with con- based on the nature, overall structure, and biological served leucine residues. The latter are generally properties of the protein cores. We have divided the se- multidomain assemblies of protein motifs with a relatively creted penicellular proteoglycans into two broad catego- elongated and often highly glycosylated structure carrying ries: the SLRPs (pronounced “slurps”), an acronym for numerous protein signatures or modules shared with other small leucine-nich proteoglycans; and the modular proteo- proteins involved in the control of cellular growth, differ- glycans. The former are typically compact proteins with entiation, lipid metabolism, and adhesion.

TABLE 1.Structure and properties of secreted pericellular proteoglycans

General features Designation Protein core” Clycosaminoglycan Chromosomal location Tissue distribution

Gene product Gene Size (kDa) Type [Number] Human Mouse

SLRP DCN 36 CS/DS [1] l2q21.3-q23 10 Ubiquitous, collagenous matrices, bone, teeth, mesothelia, floor plate

Small, ubiquitous PGs BGN 38 CS/DS [1-2] Xq28 X Interstitium, and cell enriched in leucine, surfaces with 24 amino acid tandem repeats flanked by cysteine clusters. FMOD 42 KS [4] lq32 Collagenous matrices

Lumican LUM 38 KS [2-3] 12q2l.3-q22 10,distal Cornea, intestine, liver. muscle, cartilage

Epiphycan6 36 CS/DS [2] Epiphyseal cartilage

Modular CSPG2 265-370 CS/DS [10-30] 5ql3.2 13 Blood vessels, brain, skin, Multidomain cartilage proteoglycans with protein modules AGC1 220 CS [ 100] l5q26 7 Cartilage, brain, blood homologous to the Ig vessels superfamily, selectin, EGF, laminin, LDL MNC1 136 CS [3-7] 8 Brain, cartilage receptor, N-CAM and protease inhibitors. 100 CS [1-3] Brain

Perlecan HSPG2 400-467 HS/CS [3-10] lp36 4, distal Basement membranes, cell surfaces, sinusoidal spaces, cartilage

Agrin AGRN 200 HS [3-6] 1p32-pter 4 Synaptic sites of neuromuscular junctions, renal basement membranes

Testican 44 HS/CS[1-2] 21 Seminal fluid

“The size is based on the amino acid sequence deduced from eDNA cloning. In general, however, the size of the individual protein cores is larger when estimated by DS-PAGE due to varying degrees of N- and 0-linked glycosytations. ‘This proteoglycan. originally named PC-Lb (6), has been recently renamed epiphycan (L. Rosenberg and M. Hook, personal communication) to reflect its typical tissue distribution in the epiphyseal cartilage,

PROTEOGLYCANS 599 REVIEW

THE SMALL LEUCINE RICH PROTEOGLYCANS sentially on how many conserved amino acids one re- (SLRPs) quires for a given repeat. For simplicity’s sake, we con- sider true repeats only those flanked by the two clusters General structural features of cysteine residues. Accordingly, the four major mem- bers of the SLRP family contain 10 LRRs, with epiphy- SLRPs encompass a class of secreted proteoglycans that can containing 8 repeats (Fig. 1). To date, these repeats include five structurally related but genetically distinct have been found in more than 40 gene products and are members (Table 1 and Fig. 1): deconin (2), biglycan (3), present as intracellular, transmembrane, and secreted fibromodulin (4), (5), and epiphycan, which was products in mammalian cells, plant cells, yeast, and originally called PG-Lb (6). These proteoglycans share prokaryotes (8). The number of residues in a given LRR the unique feature of being composed primarily of is between 20 and 29, with the most common being the leucine-rich tandem repeats that confer most of the bio- 24 amino acid residues seen in the SLRPs. The consen- logical functions. A close examination of the overall pro- sus sequence derived from all the known LRR proteins tein core structure reveals that it consists of three main contains leucine or other aliphatic residues at positions 2, regions: an amino-terminal region, which contains the 5, 7, 12, 16, 21, and 24, and cysteine/threonine (A-type) negatively charged GAGs or tyrosine sulfate; a central or asparagine (B-type) at position 10. The consensus se- domain with varying numbers of LRRs; and a carboxyl quence for the SLRP family is LxxLxLxxNxIJIxS/TxV/I, a end region of poorly defined function. In all cases, the B-type repeat, followed by a less homologous sequence of central domain is flanked by cysteine-rich clusters (Fig. about 7-10 residues (9). 1). In decorin, biglycan, and epiphycan, the amino-termi- The crystal structure of porcine nibonuclease inhibitor nal region harbors 1-2 GAG chains that can be either (RI), a leucine-rich protein with structural homology to dermatan or chondroitin sulfate. In contrast, fibromodulin the SLRPs, has been recently elucidated; its domain and lumican contain tyrosine sulfate, which could provide structure resolved at the 2.5 A resolution level revealed a an analogous negative charge domain. This region of the novel class of aJ protein foldings (8). The entire RI pro- molecule might be directly involved in interactions with tein is built of 15 tandem homologous repeats of 29 (A- cationic domains of extracellular matrix or cell surface type) and 28 (B-type) LRRs with only two short proteins. nonhomologous flanking sequences at the termini. The Following this control region is a domain with a cluster overall structure resembles a horseshoe and comprises al- of four highly conserved cysteine residues. Among the ternating n-sheet and a-helices in an antiparallel ar- SLRPs, these cysteine residues are similarly spaced in a rangement (8). Specifically, 17 p-strands form a curved stretch of about 20 amino acids with the following con- parallel n-sheet on the inner surface of the horseshoe, sensus sequence: Cx23CxCx68C, where x is any amino and 16 a-helices align on the outer circumference. The acid, and the subscripts denote the number of intervening residues that are conserved all cluster in the n-strand re- amino acid residues. At least in bovine articular carti- gion. The unusual nonglobular shape of the molecule, to- lage, the first and fourth cysteine residue of biglycan are gether with its solvent-exposed, parallel f-sheets and linked by a disulfide bond (7). The majority of the SLRP conformational flexibility, could explain why LRRs are protein cores are made up of tandem 24-residue repeats used to achieve strong protein-protein interactions (8). composed of alternating hydrophobic and hydrophilic Perhaps, the specificity and diversity of the protein-pro- amino acids with highly conserved leucine and tein interactions may have arisen from the nonconsensus asparagine residues. Note that fibromodulin and lumican residues. Is it possible, therefore, that the protein cores of contain N-linked keratan sulfate (KS) chains in the cen- the SLRPs may have structural-functional relationships tral domain. similar to the RI? As recently pointed out for the LRR The carboxyl end contains a ioop that is stabilized by protein chondroadherin (10), the grouping of functionally an intrachain disulfide bond between the two conserved unrelated proteins based on the simple conservation of cysteine residues in a stretch of about 35 amino acids hydrophobic residues may be misleading. Perhaps the (7). The consensus sequence, deduced from the aligned pattern of residues between the conserved amino acids amino acid sequences of the five SLRP members, is may be more important because these residues tend to be FCx1s16Lx2Nx1213C. Among the superfamily of LRR pro- exposed toward the solvent. Furthermore, although it is teins, these flanking cysteine clusters appear to be a probable that most LRR proteins form a turn and an a- property of adhesive proteins and receptors (8). helix, there are significant differences between SLRPs The sequence homologies are most pronounced in the and the RI protein. The RI lacks disulfide bonds at the central domain, which can represent up to 80% of the en- amino and carboxyl termini and the LRR peniodicity in tire SLRP. Because it seems that the central region may RI is 28-29 residues vs. 23-24 in SLRPs. Whereas the - be responsible for most of the functional roles of this sheets are formed by 11 amino acids in both cases, the class of molecules, it is worth analyzing this region in turn and ct-helix are much shorter in SLRPs (7 vs. 11 greater detail. First, the number of the LRRs vary if the residues). Finally, the repeats in RI are formed by alter- stringency is relaxed. For instance, biglycan, fibro- nating A-type (cysteine-containing) and B-type modulin and lumican could have 11-12 repeats based es- (asparagine-containing) repeats, whereas the SLRPs con-

600 Vol. 10 April 1996 The FASEB Journal IOZZO AND MURDOCH REVIEW tam only B-type repeats. Thus, although the 3-dimen- syntenic region on chromosome 10 (13, 14). In contrast, sional structure reported for RI may be prototypical for biglycan and fibromodulin map to the X chromosome (15) LRR proteins, there may be significant deviations in the and chromosome 1q32 (16), respectively. overall arrangement among members of the SLRP family. An analysis of the evolutionary relationships among the SLRP protein cores by multiple sequence alignment re- SLRP genes: structural and evolutionary veals three distinct classes of SLRPs (Fig. 2A). Based on implications amino acid sequence similarity, decorin and biglycan form the first group, whereas fibromodulin and lumican The SLRP genes that have been mapped so far are situ- form the second group. Members of the third group, the ated on at least three different chromosomes (Table 1). least characterized to date, are epiphycan and osteogly- Only decorin and lumican are on the same chromosome; cm, a protein that was originally called osteoinductive in the human these two genes are located between l2q2l factor; however, is not a proteoglycan. Ac- and q23 (11, 12), and in the mouse they are located in a cording to this analysis, the divergence of epiphycan from

Small Leucine-Rich Proteoglycans (SLRPs)

\ Decorin C CIV “.J LLxNL/I V V

V... Biglycan I I VVVVVVV CC CC CC

VffY A Fibromodulin I VVVVVVV “VII CC CC CC

SW A4AAA II VVVVVVWVVWII Lumican CC CC CC

\, Epiphycan CII wVw,,www CCC

- Leucine-rich repeat ‘‘ - N-linked oligosaccharide

C - Cysteine residue ‘F’ - Tyrosine sulfate

- Chondroitin/Dennatan sulfate - Keratan sulfate

Figure 1. The family of the small leucine-rich proteoglycans. The consensus sequence for the LRR is shown in the rectangle. A key to the various structural components is given in the bottom panel.

PROTEOGLYCANS 601 REVIEW the common ancestor of SLRP genes preceded the pre- ations. Thus, even though the interspecies conservation of sumed gene duplication and independent evolution of the the biglycan protein sequence (3) is relatively higher first two groups. than that of decorin (13), it seems that the nucleotide se- The evolutionary relationships depicted in Fig. 24 are quence of decorin is much more stringently maintained in also supported by the pairwise similarities at the gene a given population. structure level (Fig. 2B). The human decorin (11) and biglycan (17) genes are encoded by eight distinct exons. Functional interactions of SLRPs with protein and In this group of SLRP genes, the LRRs are encoded by growth factors six exons (Ill-Vill). The intron/exon junctions are highly conserved with nearly identical exon organization in both After the original observation that decorin inhibits colla- the murine decorin (13) and biglycan (18) genes. The hu- gen fibrillogenesis (20), it has become apparent that the man decorin gene harbors two leader exons (Ia and Ib) SLRPs bind to and interact with numerous vital proteins that are alternatively spliced to exon 11(11), whereas the that are involved not only in matrix assembly, but also in murine gene does not contain exon lb (13). In contrast, the control of cell proliferation and tissue morphogenesis. the second group of SLRP genes has a triexonic configu- As discussed above, a characteristic attribute of SLRPs ration. The human fibromodulin (19) and lumican (12) may be a nonglobular protein core, a feature that is likely genes, as well as the murine lumican gene (S. Chak- involved in protein/protein interactions. This structure al- ravarti, personal communication), are encoded by three lows an increase in the area available for interactions exons, with the entire LRR region being contained within with smaller, globular proteins, thereby enhancing affinity a single, central exon of -1 kb (Fig. 2B). It would be of (8). In Table 2 we have outlined some of the known in- interest to determine the genomic organization of epiphy- teractions of various SLRPs with three types of collagen, can and to establish whether this gene would fall into a Clq, fibronectin, and TGF- (21-29). In addition, deco- separate class of SLRP genes, as predicted from the den- rin can bind thrombospondin and -amyloid (9). The dogram shown in Fig. 2A. Based on the periodicity in the binding to type I collagen appears to be a common fea- primary structure of decorin, it was hypothesized that this ture of all the SLRPs investigated so far. Deconn and fi- protein arose from a series of duplications of an oligonu- bromodulin bind to separate sites on either collagen types cleotmde sequence encoding the 24 amino acid residues of I or II fibrils because bound radiolabeled PGs could be the tandem repeat. However, the intron-exon organization displaced only by the corresponding unlabeled PGs (28). of both the human (11) and murine (13) decorin genes It appears from two independent investigations that have does not offer a clear picture. That is, introns split the used either recombinant biglycan/deconin chimeras (30) middle of the LRRs 2, 5, 7, 8, and 10 without any recog- or recombinant decorin peptides (31) that the central re- nizable pattern. In contrast to other members of the LRR gion of the LRR, located between the fourth and sixth superfamily where it seems evident that exon duplication LRR, is responsible for the collagen binding activity. and shuffling from a single prototypic exon produced all Whether this region is also functioning in other members the LRRs (8), such an evolutionary relationship in the of the SLRP family needs to be elucidated in future stud- SLRP family is not so obvious. In group I, the deco- ies. rin/biglycan group, the introns often localize at similar Elucidation of how collagen fibrils grow during devel- positions, most frequently between residues 1 and 6 of opment is obviously important in understanding how tis- the LRR, and all the intron phases are identical. In group sues repair and react during pathologic processes. During II, the fibromodulin/lumican group, the first intron pre- tendon development, the increase in collagen fibril di- cedes the translation initiation site and the second intron mensions is associated with a significant decline in fibnil- precedes the termination codons (12). A probable mecha- bound decorin. Thus, the original in vitro observation that nism of evolution of LRR-containing proteins would en- decorin delayed fibril formation and induced the genera- tail unequal crossover and duplications of gene fragments tion of thinner fibers has an in vivo corollary: a decrease corresponding to prototypic leucine-rich building blocks in fibril-associated decorin is necessary for fibnil growth (8). One possibility would evoke an ancestral LRR mod- linked to tissue maturation. Other examples of how ule of early evolution that subsequently duplicated to SLRP-collagen interactions may contribute to tissue ho- give rise to individual members of the LRR superfamily, meostasis are offered by fibromodulin and lumican. In which independently developed their unique consensus cartilage, fibromodulin concentration follows a gradient, sequences. Alternatively, duplications might have oc- with the lowest amount around the chondrocytes and the curred separately in members of each protein family, and highest in the interterritonial matrix (32). Fibromodulin under the same evolutionary pressure, they could have may, therefore, represent a factor used by chondrocytes to evolved similar consensus sequences. A search for muta- regulate the assembly of cartilage collagen fibnils. In de- tions in the coding region of decorin, biglycan, and fibro- veloping cornea, during the period of acquisition of con- modulin using single-strand conformation polymorphism neal transparency there is a marked up-regulation of has found that the decorin sequence did not vary at all in lumican mRNA and protein with the concurrent switch eight individual mRNAs from fibroblasts, whereas both from a polylactosamine (nonsulfated KS) to fully sulfated biglycan and fibromodulin showed several sequence van- KS (33). The increased production of the proteoglycan

602 Vol. 10 April 1996 The FASEB journal IOZZO AND MURDOCH REVIEW form of lumican at the onset of corneal transparency sug- teins that act as adhesins mediating attachment of the or- gests that the presence of sulfated lumican is necessary ganisms to dermal collagen via deconin (34). Because for comeal transparency (33). deconin binds and inactivates the complement component Decorin bound to collagen may also be used by patho- Clq (23) as well as TGF- (35), it seems plausible that gens for their entry and colonization of tissues. For exam- deconin is involved not only in the initial colonization of ple, Borrelia burgdorferi, the spirochete that causes Lyme the dermis, but also in the further steps involving cytok- disease, the most commonly reported tick-borne disease me-controlled mediators of inflammation and immu- in the United States, adheres to collagen-associated deco- nological responses. Indeed, TGF-f interacts also with tin but does not adhere directly to types I or III collagen two other members of the SLRP family, namely, biglycan (34). The spirochete contains two decorin-binding pro- and fibromodulin, and affinity measurements suggest a

Deconin A

Biglycan

Fibromodulin

Lumican

Epiphycan

Osteoglycin

B HumanIII DeconnIV VGeneVI VII Ia VIII n

LRRQOQi!

Human Fibromodulin Gene

IH I I II In

Figure 2. Evolutionary and structural relationships of SLRP genes. Panel A shows a sequence- based evolutionary tree (dendrogram) of members of the SLRP gene family. Osteoglycin, however, is not a proteoglycan. Branch points (horizontal lines) are proportional to evolutionary distances. Panel B shows the intron/exon organization of two members of the SLRP family vis-#{225}-visthe leucine-rich repeats. Untranslated exons are represented by yellow rectangles. Introns (not in scale) are shown by thin lines.

PROTEOGLYCANS 603 REVIEW two-site binding model with Kd values of 1-20 nM and homopynimidine stretch of about 150 bp that could adopt 50-200 nM for the high- and low-affinity binding sites an intramolecular hairpin triplex structure and could play (26), respectively (Table 2). a role in the chromatin organization of the deconin gene locus. This region is capable of up-regulating a minimal Regulation of SLRP gene expression promoter in transient cell transfection assays (37), sug- gesting that this homopynimidine loop plays a role in the Studies of the distributioti of the various SLRPs (Table 1) regulation of decorin gene transcription. In contrast to the have shown that the expression of these genes rarely deconin promoter, the biglycan promoter does not contain overlaps, suggesting a tissue-specific expression. The TATA or CAAT boxes and, as often encountered in best-studied SLRPs, decorin and biglycan, show a quite housekeeping genes, is located in a CpG island with an substantial variation in tissue expression, and very fre- average G+C content of -60% and peaks reaching 85% quently their distribution is mutually exclusive, suggest- (18). These features are also present in the human bigly- ing distinct mechanisms. For example, biglycan is can gene 5’ flanking region (17), a region that exhibits primarily associated with the cell surface of epithelial functional promoter activity in human fibroblasts and cells such as keratinocytes and renal tubular epithelial bone cells (M. Young, personal communication). Similar cells, whereas decorin is often distributed in the connec- to the human, the munine biglycan promoter contains two tive tissues such as dermis, tendon, and cornea where SP-1 and several AP-2 motifs. The murine biglycan pro- biglycan is either a minor component or totally absent moter contains five PEA-3 sequences for phorbol ester, (36). In the developing mouse embryo, one of the first ap- epidermal growth factor, and serum response elements. pearances of decorin message at day 11 postconception is Deletion analysis of this 5’ flanking DNA sequence has in a strategic location, the floor plate, the region that cor- indicated that the distal promoter region is required for responds to the unpaired ventral zone forming the floor of full transcriptional activity (18). the neural tube (13). With progressive maturation, deco- Another striking contrast between deconin and biglycan nin expression becomes concentrated in the linings of all is that their transcription is differentially regulated by the major organs including the mesothelial layers of the growth factors and cytokines. For instance, in most cases pleura, pericardium, and peritoneum as well as the renal TGF- down-regulates the expression of deconin while capsule and meninges (13). This striking distribution of up-regulating that of biglycan (40). Similar effects are decorin suggests that this gene product may be involved seen using dexamethasone, which also prevents the TGF- in the control of organ shaping during development. t3-mediated effects on the two proteoglycans (41). Reti- What do we know about the transcriptional machinery noic acid down-regulates decorin mRNA levels without of these SLRP genes? Although several SLRP genes have affecting those of biglycan (41). That tissue-specific fac- been cloned, only two have been characterized insofar as tors do play a significant role is shown by the finding that their functional promoters: the human deconin (37, 38) in bovine chondrocytes, retinoic acid markedly up-regu- and the murine biglycan (18) genes. The differential re- lates decorin mRNA and protein while rapidly reducing sponse to cytokines (9) would predict that the organiza- the biglycan mRNA and protein levels (42). Nuclear run- tion of the control regions of these two genes would be on assays have shown that the retinoic acid-mediated ef- quite different. Indeed, this is the case. The two promot- fects on these two SLRPs is transcriptional for biglycan ers harbor quite different cis-acting elements. The human and post-transcriptional for deconin, perhaps via stabiliza- decorin promoter, the -1 kb region flanking exon lb. can tion of deconin mRNA (42). We have recently shown that be subdivided into two distinct regions: a proximal pro- TNF-a is capable of down-regulating the transcription of moter (from 1 to -140 bp), and a distal promoter (from - the deconin gene in quiescent human fibroblasts, a situ- 141 to -983 bp) (37). The proximal promoter contains two ation in which the gene is markedly upregulated (38). We closely spaced TATA boxes and a CAAT box. In vitro also investigated whether TGF-13 and TNF-a act inde- transcription analyses demonstrated that both TATA pendently of each other or in a synergistic fashion. The boxes are used (37). In addition, the proximal promoter results showed that the two cytokines down-regulate region contains a TN F-a responsive element between - deconin gene expression through independent mecha- 140 and -180 bp, and an additional TNF-a responsive nisms and indicate that TNF-a may be a key modulator element resides within exon lb (38). We have shown that of the biological function of this proteoglycan. in transient cell transfection assays, there is a dose-de- From the studies discussed above, it is clear that the pendent transcriptional repression of deconin by TNF-a diversity in response is related not only to the divergent (38). The distal promoter region harbors an AP-1, an AP- tissue distribution of deconin and biglycan gene products, 5, and two NF-kb motifs, as well as several direct repeats but also to the distinctive transcriptional machineries and a TGF-3 negative element. This TGF-3 negative ele- controlling these two genes. The involvement of inflam- ment has been found in several proteases, such as matory cytokines, together with the unique binding prop- stromelysmn and elastase, which are down-regulated by erties of the SLRPs (Table 2) indicate that there is a this cytokine (39), and could function as a transcriptional finely balanced regulation of the transcriptional activity silencer and suppress deconin expression in vivo. The of the SLRP genes during tissue repair, regeneration, and most distal part of the deconin promoter harbors a long tumor formation.

604 Vol. 10 April 1996 The FASEB journal IOZZO AND MURDOCH REVIEW

TABLE 2. Binding properties of SLRPs

Proteoglycan Collagen type I Collagen type II Collagen Type VI Clq Fibronectin TCF-I1 Reference

Dissociation constant (K,j)

Decorin, high affinity 0.7 nM 12-16 nM 0.3MM 7.6 nM 10-20 nM 1-20 nM (21-26) low affinity 3 nM 110-130 nM 50-200nM

Biglycan, high affinity 87 nM (+)#{176} 1-17 nM (21, 27)

low affinity 50-200 nM

Fibromodulin, high affinity 10-35 nM 35 nM (+)#{176} (+) 17-17 nM (22, 27, 28) low affinity 50-200 nM

Lumican + (+)#{176} (27, 29) ‘These interactions are based on immunology da ta in fetal rabbit corneas (27).

Decorin and the control of cell proliferation deconin in the suppression of the transformed phenotype insofar as the decorin-induced effects were independent The genes that regulate the transition from the prolifera- of TGF-, clonal selection or specific integration site of tive phase of the cell cycle to quiescence are not yet the transfected DNA. The challenge is now to determine clearly understood. Under the appropriate circumstances, the precise mechanism of action and the signal transduc- however, such genes may act as tumor suppressor genes. ing pathway through which this decorin-induced effect is Several lines of evidence indicate that deconin is linked mediated. Perhaps other members of SLRP gene family to growth inhibition of mammalian cells. Decorin overex- may hold analogous growth-suppressive abilities. pression in Chinese hamster ovary cells inhibits cell pro- liferation, a phenomenon that is at least in part due to the blocking of TGF- activity (35). Decorin is markedly up- regulated when cells reach quiescence (38, 43, 44), THE MODULAR PROTEOGLYCANS whereas its expression is suppressed upon SV4O-induced viral transformation (44). Decorin has been identified as Modular proteoglycans are works of art produced by an one of eight genes, named quiescins, that are overex- assembly of protein fragments in unlikely or unexpected pressed by lung fibroblasts during the stationary phase of juxtaposition. It is a sort of protein collage, the major uni- contact inhibition and quiescence (44). Deconin is rarely fying feature of all modular proteoglycans. This class of expressed by transformed cells derived from human tu- macromolecules can be further divided into two subfami- mors, including those derived from colon, breast, lung, lies; the hyalectans, an acronym for hyaluronan- and and prostate carcinomas (our unpublished observation). lectin-binding proteoglycans, which comprise versican, In contrast, deconin is markedly increased at both the aggrecan, neurocan, and brevican (this family is charac- mRNA and proteoglycan level in the stroma of human co- terized by a similar amino terminus that binds hyaluro- lon cancer (45); this may represent a regional biochemi- nan, an extended central domain that harbors the GAG cal response of the stroma to the invading tumor cells. A chains, and a carboxyl terminus that is homologous to the fundamental question is whether decorin is a tumor sup- selectin family); and the nonhyaluronan-binding proteo- pressor synthesized by the stromal cells to counteract the glycans, which comprise and agrin, two large growth of carcinoma cells. To address this question, we PGs that carry primarily HS chains, and testican, an transfected colon carcinoma cells, which do not constitu- HS/CS-carrying proteoglyan isolated from seminal fluid. tively express deconin, with the full-length deconin cDNA driven by the strong cytomegalovirus promoter (46). We discovered that de novo decorin gene expression sup- Hyalectans, the hyaluronan- and lectin-binding presses the malignant phenotype in colon carcinoma modular proteoglycans: structural considerations cells. The cells show a decline in growth rate, do not grow in soft agar, and fail to generate tumors when in- This family comprises four members encoded by distinct jected into scid/scid mice (46). The cells are arrested in genes: versican, aggrecan, neurocan, and brevican (Fig. Cl and can reenter the cell cycle when deconin expres- 3). A trait shared by these proteoglycans is a tridomain sion is abrogated by deconin-specific anti-sense oligode- structure, with two flanking regions that bind hyaluronan oxynucleotides. These data establish a direct role of and contain C-type lectin-like domains, respectively,

PROTEOGLYCANS 605 REVIEW separated by an extended central region that carries most The most recent member added to the hyalectan family of the GAG chains. The largest member of this family is is brevican, another brain-specific CS-PG (57). The most versican, a proteoglycan first cloned from a fibroblast prominent feature of brevican, and hence its name, is the cDNA library (47) that is widely expressed in vascular remarkably short central domain. This nonhomologous re- and avascular connective tissues (48). The amino-termi- gion is enriched in acidic residues, especially glutamic nal globular domain contains the hyaluronan-binding re- acid, which may be used to bind cationic substances and gion that lies within characteristic loops that are minerals (57). As in the case of neurocan, brevican is tandemly repeated. Recombinant forms of this domain found as two major products: a full-length proteoglycan bind to hyaluronan with high affinity (Kd -4 nM), thereby and a GAG-deficient protein core that is generated by establishing a strong functional correlation with aggrecan proteolytic cleavage within the central domain. Because a (49). On this structural basis alone, one could predict cleaved version of versican exists in brain extracts and a that the other members of this family would also bind sizable proportion of aggrecan molecules are C-terminally hyaluronan. The second domain in versican encompasses truncated in cartilage extracts, it is possible that pro- two alternatively spliced exons that harbor the GAG at- teolytic processing of the hyalectans is a general feature tachment regions, which have been designated GAG-a required for their specialized function in tissues. and GAG- (50, 51). These regions are free of cysteine residues, but are enriched in acidic amino acids, and The hyalectan genes: structural and evolutionary carry up to -30 binding sites for GAG chains as well as considerations several binding sites for 0- and N-linked oligosaccha- rides. Four possible variants of versican can exist, and Current view indicates that relatively few protein modules this has been shown by analysis of both mRNA and pro- are exploited in the making of multidomain polypeptides. tein isoforms (51). The carboxyl terminus of versican At the gene level, evolutionary processes have used exon contains a series of structural motifs characteristically shuffling and duplication/amplification of the respective found in the selectin family: two EGF repeats, a C-type modules, thereby enabling a progressive refinement as lectin domain, and a CRP-like motif (Fig. 3). well as the emergence of new hybrid proteins with di- Aggrecan structure is quite similar to versican, with a verse functions. We have recently deciphered the entire few exceptions. First, it contains a second globular do- genomic organization of the human versican gene, desig- main at the amino terminus whose function is not clearly nated CSPG2, established functional activity of its pro- understood insofar as it does not bind hyaluronan. Sec- moter (50), and mapped the human CSPG2 (58) and the ond, it contains a GAG-binding region that is similar in munine Cspg2 (59) genes to chromosome 5q13.2 and 13, size to the GAG- of versican but that harbors many respectively. The rat (60), mouse (61), and human (62) more Ser-Gly consensus sequences. Thus, a fully glycosy- aggrecan genes have also been sequenced, and the gene lated aggrecan would contain more than threefold the AGC1 assigned to human chromosome l5q26 (63) and number of side chains than the largest isoform of versi- mouse 7 (64), respectively, while the munine neurocan can, -100 vs 30 chains, respectively. Third, in the car- Mncl gene maps to chromosome 8 (65). boxyl end of aggrecan, the two EGF repeats can be The disparate chromosomal location of the hyalectan alternatively spliced in a significant portion of the mole- genes indicates an early evolutionary divergence of this cules (52, 53). These structural differences would have gene family, but suggests that a strong evolutionary pres- significant consequences in the functional properties of sure acted toward the conservation of these modular pro- the hyalectans. Recombinant forms of the C-type lectin teoglycans (65). Close analysis of the intron/exon domain from both versican and aggrecan bind various organization of the versican, aggrecan, and neurocan carbohydrates, including galactose and fucose, in a cal- genes reveals a striking similarity in the two regions of cium-dependent manner (54); this further emphasizes the protein homology, the HBR and the selectin regions (Fig. functional relationship between hyalectan domains in dif- 4). For instance, the HBRs are all encoded by four dis- ferent PGs. tinct exons of identical size and intron phasing. The two The third member of this family is neurocan, a major -100 amino acid repeats that bind hyaluronan are en- CS-PG found in early postnatal brain (55). As with other coded by an exon doublet (exons 4 and 5, which are members of the hyalectan family, neurocan harbors amino separated by a small intron). The pattern of intron phas- and carboxyl-terminal domains that share more than 40 ing and exon sizing is interesting when compared with and 60% identity to the hyaluronan binding region and other members of hyaluronan-binding protein family such the selectin region of versican and aggrecan, respectively as the link protein or CD44 (Fig. 4). In link protein, (Fig. 4). The central domain, which contains up to seven there is only one exon encoding the first loop, an exon GAG attachment sites, is quite unique and shows no sig- that corresponds exactly to the summation of exons 4 and nificant homology to other sequenced PGs (56). Neurocan 5 of versican. CD44, which harbors a region similar to is developmentally regulated and it appears that the adult the first repeat of versican, has this region encoded by form, which lacks the amino-terminal domain, is gener- two exons nearly identical to those in other hyalectans ated by an endoproteolytic cleavage of the protein core that are similarly separated by phase II introns. The pres- (55). ence of phase II introns between exons 4 and 5 suggests

606 Vol. 10 April 1996 The FASEB Journal IOZZO AND MURDOCH REVIEW that the two exons cannot be alternatively spliced or du- and -5.3 kb, respectively) or the central region of aggre- plicated independently of each other (62). Another strik- can (-3.9 kb) are also unique features of the hyalectan ing feature of the hyalectan genes is that all of the introns genes. In the vast majority of genes, the average exon size between those protein modules that are considered to fold is -150 bp. That these genes have evolved a way to defy autonomously (65) are phase I introns. This allows alter- the rules of exon definition is intriguing and poses ques- native splicing events to occur without altering the read- tions of how this process is regulated. A proposed func- ing frame, as in the case of the versican GAG-a and tion of introns is to limit the amplification of genomic GAG- domains (51). The colossal sizes of the exons en- sequences that cannot tolerate variations in sizes, as in coding either the two central domains of versican (-3 collagen genes. Therefore, genes that encode proteins in

Modular Proteoglycans

I- 1r______Versican L?,? JL

HBR GAGct GAG3 Selectin -t,iMiii1i(i,4, 11r’r_1r_11__r_1 I LJLJLJ Aggrecan

t’ /1 Neurocan ) \(‘.\

Brevican

111111111111 Perlecan Do44cIIIE V V V

.*eeecs Agrin

(1

Testican

o Ig fold #{149}complement regulatory-like laminin G domain #{149}EGF repeat 0 laminin domain Ill-like I CWCV domain #{149}link protein-like laminin domain IV-like #{149}SPARC-like domain U GAG attachment domain 6 LDL receptor-like GAG chain lectin-like #{149}follistatin repeat (3) SEA module

Figure 3. A schematic diagram of the modular proteoglycans. The hyalectans, the proteoglycans that harbor hyaluronan- and lectin-binding modules at opposite ends, comprise versican, aggrecan, neurocan, and brevican. The nonhyaluronan- binding modular proteoglycans comprise perlecan, agrin, and testican. A key to the various protein modules and structural motifs is provided in the bottom panel. Parentheses indicate the site of alternative splicing.

PROTEOGLYCANS 607 REVIEW which the exact size or number of repeats is not so rigor- domains together would function as a bridge between the ously required for their function are likely to contain surface coats of cells, enriched in hyaluronan, and the large exons. These large exons would grant improved extracellular molecules endowed with carbohydrate RNA processing for the hyalectan genes. chains. Less clear is the function of the central domain The 3’ ends of the hyalectan genes also share striking inasmuch as the central regions of all the hyalectans are similarities. This region, which codes for a protein mod- nonhomologous to each other and contain various num- ule characteristically found in selectins, the family of leu- bers of GAG chains. The potential numbers of GAG side kocyte homing and cell adhesion molecules, comprises chains can vary from 3 in brevican to 100 in aggrecan six exons with functional domains separated by identical (Table 1). Therefore, the various members of the hyalec- phasing introns (Fig. 4). The first two exons code for EGF tan gene family will have a unique potential for introduc- repeats that can be alternatively spliced in aggrecan (53), ing CS chains in a given tissue, thereby influencing the but not in versican (51) or neurocan (55). The second water and ion content of the extracellular matrix as well EGF repeat contains amino acid sequences that could as the volume it occupies. The functional properties of mediate Ca2 binding, whereas the first EGF repeat does aggrecan are thought to reside in two structural features: not. Thus, it is possible that the two repeats may dictate the high concentration of GAG chains and the formation different functions. The second subdomain of the selectin of large supramolecular aggregates with hyaluronan (52). is a C-type lectin motif (54) encoded by three exons. The high charge density of aggrecan results in each Other members of the selectin family, such as P- and E- monomer occupying a large hydrodynamic volume. When selectin, carry a C-type lectin motif encoded by a single subjected to compressive forces, water is displaced from exon, whose size nearly equals the summation of the the individual monomers, thereby increasing swelling po- three exons present in hyalectans (Fig. 4). The strict con- tential, which is dissipated upon removal of the compres- servation of numerous amino acid residues crucial for the sive forces by the water molecules now siphoned back proper folding of selectins and for their carbohydrate- into the tissue (52). It is plausible that some of these binding and interactions, together with the correlation of physicochemical properties of aggrecan could be shared gene structure and protein modules, has led to the group- by versican by providing improved elastic return of the ing of selectins into four major categories, with hyalec- blood vessel walls after cardiac systole. The functional tans belonging to group 1 (54). An implication of these roles of the two brain PGs, neurocan and brevican, could structural-functional correlations is that the lectin motifs also be linked to their modular organization. Physiologi- in hyalectans might have evolved from an ancestral lectin cally relevant features of neurocan include the presence module that duplicated before the exon shuffling/intron of a cell-attachment RGDS sequence and a relatively recombination events took place to generate the various high substitution with 0-linked oligosaccharides (55). progenitors (50). The terminal part of the selectin mod- This makeup would clearly contribute to some of the pro- ule, the so-called complement regulatory protein (CRP), posed functional roles of neurocan, including inhibition is encoded by a single exon of identical size in the of the homophilic interactions of N-CAM. Thus, neurocan hyalectans (183 bp) and very similar to the exon present may act as a repulsive molecule by modulating cell-cell in P- and E-selectins (186 bp). In general, the other and cell-matrix interactions. Finally, neurocan (55) and members of the selectin family contain several CRP mo- brevican (57) can play a direct role in defining a destina- tifs that are thought to bind C3b and C4b and regulate tion for migrating axons that form the cortical plate and the C3/C5 convertases. As with the EGF repeats, this represent active components of astroglial barriers to ax- exon undergoes alternative splicing in aggrecan (66) but onal outgrowth. A recent report also implicates versican not in versican (51) or neurocan (55). Perhaps the EGF in the inhibition of neural crest cell migration and out- and CRP motifs may be important in determining the growth of motor and sensory axons (67). specificity of the lectin-like interactions by contacting the protein moiety of a (50). Nonhyaluronan-binding modular proteoglycans-perlecan, agrin, and testican: Hyalectans: functional implications general structural features

From the structural considerations summarized above, Of the groups described so far, this collection of modular what roles might the hyalectans play in tissue homeosta- proteoglycans, which includes perlecan, agrin, and testi- sis? One general conclusion is that the quantitative dif- can, appears to have the least in common with each other ferences between versican and aggrecan in blood vessels (Table 1). Nevertheless, there is a familiar thread running vs. cartilage, for example, could be based on tissue-spe- through their structures: the repeated use of common pro- cific transcriptional control. The constitutive transcription tein modules (Fig. 3). In addition, these three members of these PG genes and the expression of the various iso- carry primarily heparan sulfate side chains, and at least forms must respect finite biological barriers. The charac- perlecan and agrin can be closely associated with the cell teristic combination of structural domains indicates that surface. Because of their chimeric structural design, the the hyalectans bind hyaluronan at their amino terminus members of this family will likely be involved in a variety and carbohydrates at the carboxyl terminus. These two of biological processes. The shared structural motifs,

608 Vol. 10 April 1996 The FASEB Journal IOZZO AND MURDOCH REVIEW when translated into multimeric functional units, could mology with any previously sequenced molecule (69). be effectively used by different tissues during ontogeny However, a recent database homology search allowed us and remodeling. As amply reviewed before (68), perlecan to delineate perlecan domain I into two separate subdo- is the most structurally complex of these molecules, with mains: a GAG attachment domain and a so-called SEA five separate protein modules that also appear in other module, named for the three molecules-sperm protein, extracellular and cell surface proteins. The amino-termi- enterokinase, and agrin-in which it was first identified nal domain I harbors the GAG attachment region. This (70). A conserved 80 amino acid residue sequence, the region was originally thought to have no significant ho- SEA module is found in a variety of extracellular mole-

Hyaluronan Binding Region Selectin Region

Human Versican 9W 11 12 13 14 -H- H-I 375 175 128 294 EGF EGF L Lectin -I CRP 114 114 159 83 145 183 Human Aggrecan 15 16 17 18 B-I 175 128 294 EGF EGF L...... Lectin .-_..J CRP 114 114 159 83 145 183 Mouse Neurocan

4 5 6 Ti -Hi LI- -OH1-IHI-I-I-- 375 175 128 294 EGF EGF Leclin -. CRP 114 114 159 83 145 183 303

Rat Link Protein Human P-Selectin Hi 372 303 227 Lectin EGF CRPI 387 108 186

Human CD44 Human E-Selectin 23 3 TI ‘U 166 134 Lectin EGF CRPI 384 108 186

Intron Codon Phases 1= U[I o= LI

Figure 4. A schematic diagram of the gene organization of the 5’ and 3’ regions of the modular proteoglycans versican, aggrecan and neurocan vis-#{225}-vis that of homologous genes. A key to the intron codon phases is provided in the bottom panel. Introns that interrupt a codon triplet after the first nucleotide are indicated as I; the introns that interrupt a codon triplet after the second nucleotide are indicated as II, and the introns that do not interrupt a codon triplet are indicated as 0.

PROTEOGLYCANS 609 REVIEW cules as diverse as enterokinase, MUC1 (episialin), a sea human chromosome 21 (77). Testican is largely com- urchin sperm protein, and agrin, and could be conceiv- posed of domains typically found in SPARC/, ably involved in mediating binding to neighboring carbo- a multifunctional extracellular matrix protein that dis- hydrate structures. Indeed, all the proteins that harbor rupts cellular contacts and suppresses cell cycle progres- the SEA module are either highly glycosylated glycopro- sion (78). These domains are related to the follistatin teins or proteoglycans (70). A potential role of the SEA repeat, the third repeat in testican being the most similar module may be as a recognizing signal for proper glyco- to those found in follistatin and agrin (Fig. 3). Down- sylation of the protein core during the multistep modifica- stream of the SPARC-like region lies a 46 amino acid se- tions occurring in the Golgi apparatus. This region is quence enriched in cysteine residues with similarities to followed by domain II, which harbors four repeats of a to a motif originally described in and also structure found in the LDL receptor molecule. Domain III present in nidogen (76). This region contains the rare se- shares homology with the subdomains III and IV of quence Cys-Trp-Cys-Val (the CWCV module in Fig. 3), a laminin chains, particularly those of the laminin a chain. region found in extracellular and cell surface molecules A large block of 21 consecutive Ig fold repeats is next, and thought to have binding properties. The carboxyl ter- comprising a major portion of the perlecan protein core. minus of testican contains the GAG-attachment region These repeats are most similar in sequence to those with two, closely spaced potential binding sites for found in the neural cell adhesion molecule. The most heparan and chondroitin sulfate side chains. The more 5’ carboxyl-terminal domain of perlecan is made up of binding site contains one SGD triplet preceded by acidic seven subdomains. Four EGF-like repeats are inter- amino acid residues, a feature that resembles very closely spersed with globular repeats, with homology and pre- that found in domain I of perlecan where the HS chains sumably similar structural features to the subdomains are attached (79). that form the globular structures at the carboxyl terminus of the laminin a chain. Agrin, a product of motor neurons, is a new arrival in Genes encoding perlecan and agrin the modular family of proteoglycans (71). For years, agrin had been investigated as a major product of the synaptic Of the three nonhyaluronan-binding PGs, only the space necessary for the regeneration of neuromuscular genomic organization of human perlecan and mouse agrin synapses and acetyicholine receptor clustering (72). It is have been described to date. Both allow some interesting now evident that agrin belongs to the proteoglycan gene observations to be made regarding the evolutionary devel- family (71), and that the “active” agrin investigated in opment of modular proteoglycans. The human perlecan past studies is a proteolytic derivative of the parent pro- gene (HSPG2) is a single-copy gene located on the short teoglycan. Not only is agrin similar in structure to the do- arm of human chromosome 1 at lp3#{243}(80) and on a syn- main V of perlecan and the neurexin molecules (73), it tenic region of mouse chromosome 4. The human gene also contains a SEA module and several SGxG repeats spans at least 120 kb of genomic DNA and is composed that can carry HS chains (Fig. 3). However, the precise of 94 exons (81). The gene duplication theory of molecu- location of the HS chains within the agrin has not been lar evolution is well illustrated by the remarkable conser- determined. Furthermore, two antigenically and structur- vation between the perlecan gene and the genes of ally related HSPGs that have been isolated from bovine molecules that are homologous to the various domains. renal tubular basement membrane exhibit high homology For example, the exon sizes of the LDL receptor-like re- to rat agrin (74). Structurally, a significant proportion of peats in perlecan domain II are very similar in size to the the agrin molecule (72) is composed of a nine times re- exons from the ligand binding region of the LDL receptor peated motif belonging to the follistatin module family gene and are split by introns in the same phase. Conser- (75). Between repeats 8 and 9 are two repeats with ho- vation such as this suggests strongly that the perlecan mology to laminin domain III structures. In this case, gene and the genes of homologous molecules diverged these repeated structures are most closely related to do- from common ancestors before acquiring individual func- main III sequences found in laminin and ‘y chains. Fol- tions. Homology scans of the perlecan cDNA sequence lowing this region, is one repeat of the SEA module (70). (79) reveal significant inner homology between repeated The other major repeated domains are found in the car- elements. In the case of the immunoglobulin repeats of boxyl-terminal third of the agrin molecule. As mentioned domain IV, some are almost identical repeats, differing above, this region is similar to the corresponding domain only by 2 or 3 nucleotides out of 300. There is also a of the perlecan molecule, having a set of three laminin-a striking conservation of intron phase between units en- G domain-like repeats interrupted by EGF-like repeats. coding the Ig fold, probably indicating extensive duplica- This structural arrangement is also present in the repeats tion of exons by intronic recombination in the generation of three distinct species of neurexin: Ia, Ila, and lila of the modern perlecan molecule. The promoter region of (68). the human perlecan gene is contained in a CpG island Testican is a modular proteoglycan originally isolated and contains a high number of Spi-binding sites (81) from seminal fluid and presumably synthesized by the with an overall organization typical of housekeeping and testis (76)-hence its name-whose gene is localized on growth factor genes.

610 Vol. 10 April 1996 The FASEB Journal IOZZO AND MURDOCH REVIEW

The human agrin gene (AGRN) has been localized to await future experimentation. Perlecan has been shown to the distal short arm region of chromosome 1 at 1p32-pter, interact with other components of the extracellular ma- relatively close to the perlecan gene and to the same syn- trix, in particular with other basement membrane mole- tenic region of mouse chromosome 4 (82). Once again, cules such as laminin and nidogen (89). Perlecan will the common theme of exon shuffling and duplication also self-associate via its carboxyl-terminal domain V seems to have been involved in the evolution of the agrin which contains modules with homology to laminin G do- molecule. The follistatin repeats 1-7 in the amino-termi- mains and EGF repeats (68). Proteins with similar do- nal third of the structure are all encoded by individual mains, such as the laminin a chain, and the Drosophila exons flanked by phase I introns and could have evolved proteins crumbs, fat, and slit appear to be involved in from a series of unequal crossover events after the dele- processes of differentiation such as epithelial polarization tion of the internal intron found in repeats 8 and 9. In and neuralization, but it is not clear whether perlecan has contrast, in the carboxyl-terminal part of agrin there is no similar functions. As an interesting aside, the Drosophila correlation between exon/intron organization of the gene protein slit is also a large leucine-rich protein, which fur- and the modular organization of the protein. When one ther illustrates the promiscuity of these protein domains. compares the regions of the perlecan and agrin genes that Other functions suggested by sequence homology, such as share domains homologous to laminin G domain and EGF involvement in lipoprotein metabolism via the LDL re- repeats, there is evidence of considerable evolutionary ceptor-like domain II, remain under investigation. A ma- distance between these molecules. There is no consistent jor function of perlecan is attributed to the properties of conservation of exon boundaries at the junctions of the the HS chains carried by the protein core. These chains motifs, nor are there any similarities in the phases of the are able to bind growth factors and cytokines and seques- introns dividing the coding sequences. Thus, whereas at ter them in the extracellular matrix, where they may the amino acid level these two modular proteoglycans are function as a store or reserve to be released under appro- clearly related, genomic analysis has revealed the dis- priate stimuli. Perlecan may be crucially important in the tance of this relationship. In addition, the striking differ- action of bFGF, functioning as a low-affinity coreceptor ence between exon/intron boundaries of the amino- and for this factor, and therefore becoming an important com- carboxyl-ends of the agrin gene suggests that their evolu- ponent in the processes of bFGF-mediated mitogenesis tionary history is different and support the notion that and angiogenesis (90). The perlecan-bound bFGF can be these two regions have diverse biological functions (75). released from the matrix in a presumably active form fol- lowing degradation of either the perlecan protein core or the heparan sulfate chains. It is clear that such protease- Functional properties of nonhyaluronan-binding, or heparitinase-stimulated release of growth factors could modular proteoglycans play an important role in a number of physiological sce- narios, including tumor angiogenesis and neovasculariza- The perlecan molecule is deposited in the extracellular tion as a consequence of platelet degranulation at sites of matrix of the embryo at very early stages of development, injury. Perlecan may also be responsible for the matrix- and is considered to be the principal heparan sulfate pro- associated binding of granulocyte/macrophage-colony teoglycan of basement membranes. First isolated from a stimulating factor, interleukin-3, interferon ‘, and hepato- basement membrane producing murine tumor, and later cyte growth factor/scatter factor (68), all of which are from other basement membranes, its main function ap- known to bind extracellular matrix heparan sulfate. peared to be the anchoring of heparan sulfate chains in Agrin was originally isolated from torpedo ray electric the basement membrane to provide a carpet of negative organ and was found to induce acetylcholine receptor ag- charge in a barrier function (68). Subsequent cDNA clon- gregation in cultured myotubes. Subsequent investigation ing revealed an elaborate composite protein core with a revealed that the molecule was secreted by motor neurons myriad of potential functions. Perlecan has been localized and deposited in the synaptic cleft basement membrane in every basement membrane studied, as well as in ex- where it performed its functions, not only aggregating tracellular locations as diverse as the hepatic sinusoid, acetylcholine receptors but also the acetylcholine connective tissue stroma, and the cartilage matrix (68, esterase that limits the action of acetylcholine (91). Other 83-85). The protein core appears to support the adhesion components of the synaptic cleft basement mem- of cells such as endothelial cells (86) and chondrocytes brane-cell surface integral membrane proteins and cy- (84), but is repellent for some cells of the hematopoietic toskeletal elements-also appear to undergo system (87). Perlecan is constitutively expressed by bone rearrangements after the appearance of agrin, suggesting marrow-derived tumor cells, such as the erythroleukemia a role for agrin as a postsynaptic cleft organizer. Agrin K562 cells (88), thus suggesting a potential role of this may also play a key role in the regeneration of synapses molecule in the biology of this tissue. Even though some during reinnervation of damaged sites by providing a adhesive properties described for mouse perlecan can be template of the location of the original synaptic site. It attributed to integrin binding of an RGD sequence, the has been suggested that the follistatin modules of agrin lack of such a motif in the human perlecan protein core may serve as binding sites for the local sequestration of means that detailed analysis of these functions must growth factors of the PDGF and TGF- families, perhaps

PROTEOGLYCANS 611 REVIEW contributing to motor neuron survival. Localization of teoglycan genes, or a combination of genes, will provide growth factors is an attribute of growing importance in the definitive answers about their specific functions. proteoglycan world, and particularly for the modular pro- We would like to thank Dr. C. C. Clark for critical reading of the teoglycans agrin and perlecan, another resident of the manuscript and the following investigators for providing valuable com- neuromuscular junction. It is clear that the receptor ag- ments and/or unpublished data: S. Chakravarti, L. Fisher, J. Hassell, M. gregating activity resides in the carboxyl-terminal third of Hook, R. Margolis, P. Neame, U. Rauch, L. Rosenberg, P. Roughley, M. Young, and Y. Yamaguchi. The original research was supported by the the molecule, in the region containing EGF and laminin National Institutes of Health grants ROl CA394.81 and ROl CA47282 G domain repeats. This activity may be contained within to R.V.I. We would like to apologize to our colleagues whose original alternatively spliced regions of the agrin structure, al- work was not referenced because of space limitations imposed by the editorial policy of the journal. though this appears to be dependent on the cell lines used in the assay. Alternative splicing gives rise to a number of splice variants of agrin, including ones with differences in the signal peptide region which are pre- REFERENCES dominantly expressed in nonneuronal tissues (92). The 1. Dammer, U., Popescu, 0., Wagner, P., Anselmetti, D., G#{252}ntherodt,H. J., and expression of agrin seems quite widespread; it is found Misevic, C. N. (1995) Binding strength hetween cell adhesion proteoglycans predominantly in the brain, but also in avian smooth and measured by atomic force microscopy. Science 267. 1173-1175 2. Krusius, T., and Ruoslahti, E. (1986) Primary structure of an extracellular cardiac muscles (92) as well as a constituent of mammal- matrix proteoglycan core protein deduced from cloned eDNA. Proc. Nail. ian renal tubular basement membranes (74). The particu- Acad. Sci. USA 83,7683-7687 3. Fisher, L. W., Termine, J. D., and Young, M. F. (1989) Deduced protein lar splice variants and their roles in some of these sequence of bone small proteoglycan I (biglycan) shows homology with locations remains unclear. proteoglycan II (decorin) and several nonconnective tissue proteins in a variety of s,pecies. J. Biol. Chem. 264,4571-4576 As mentioned, the testican core protein contains re- 4. Oldherg, A., Antonsson, P., Lindblom, K., and Heineglrd, D. )1989) A peats with similarity to regions of adhesive (CWCV do- collagen binding 59 kd protein (fibromodulin) is structurally related to the main) and anti-adhesive (SPARC-like domains) small interstitial proteoglycans PG Si and PG S2 (decorin). EMBO J. 8, 2601-2604 molecules; however, its functions are presently unknown. 5. Blochberger, T. C., Vergnes, J. P., Hempel, J., and Hassell, J. H. (1992) It is the parental molecule of a peptide found in seminal eDNA to chick lumican (corneal keratan sulfate proteoglycan) reveals ho- mology to the small interstitial proteoglycan gene family and expression in plasma, which carries both HS and CS side chains (76). muscle and intestine. J. Biol. Chem. 267. 347-352 The carboxyl-terminal 60 amino acids of testican contain 6. Shinomura, 1., and Kimata, K. (1992) Proteoglycan Lh, a small dermatan sulfate proteoglycan expressed in embryonic chick epiphyseal cartilage, is a high content of acidic amino acids, with glutamic acid structurally related to osteoinductive factor. J. Biol. C/tern. 267, 1265-1270 comprising up to 40% of the residues. These acidic 7. Neame. P. J., Choi, H. U.. and Rosenberg, L. C. (1989) The primary structure of the core protein of the small. leucine rich proteoglycan (PG 1) from bovine stretches are also present in versican and brevican, and articular cartilage.). Biol. C/tern. 264, 8653-8661 may be involved in the binding to cationic substances 8. Kohe, B., and Deisenhofer, J. (1994) The leucine rich repeat: a versatile binding motif. Trends Biochern. Sci. 19,415-421 and minerals. Future studies are needed to establish 9. Kresse, H., Hausser, H., and Schonherr, E. (1993) Small proteoglycans. whether testican is expressed by tissues other than the Experientia 49. 403-416 10. Neame, P. J.. Sommarin. Y., Boynlon, R. E., and Heineglrd, D. (1994) The testis and to decipher its functional roles. structure of a 38 kDa leucine rich protein (Chondroadherin) isolated from bovine cartilage. J. Biol. Chem. 269,21547-21554 11. Danielson, K. C., Fazzio, A., Cohen, I., Cannizzaro, 1. A., Eichstetter, 1., and lozzo, R. V. (1993) The human decorin gene: intron exon organization, CONCLUDING REMARKS discovery of two alternatively spliced exons in the 5’ untranslated region, and mapping of the gene to chromosome l2q23. Genomics 15, 146-160 12. Grover, J., Chen, X. N., Korenherg, J. R., and Roughley, P. J. (1995) The Undoubtedly, the world of proteoglycan biology has been human lumican gene. Organization, chromosomal location, and expression very prolific, with new members of this multifunctional in articular cartilage.). Bid. C/tern. 270.21942-21949 13. Scholzen, T., Solursh, M., Suzuki, S., Reiter, R., Morgan, J. L., Buchberg, A. gene family being discovered every year. Now that we M., Siracusa, L. D., and lozzo, R. V. (1994) The murine decorin. Complete have a better conceptual grasp of the structure of various cDNA cloning, genomic organization, chromosomal assignment and expres- sion during organogenesis and tissue differentiation. J. Biol. Chem. 269, proteoglycan family members, we can experiment with 28270-28281 more focus and precision. It is manifest from this synop- 14. Chakravarti, S., and Magnuson, T. (1995) Localization of mouse lumican (keratan sulfate proteoglycan) to distal chromosome 10. Mammalian Genorne sis that two fundamental properties of all secreted 6,367-368 pericellular proteoglycans are to interact with other pro- 15. McBride, 0. W., Fisher, L. W., and Young, M. F. (1990) Localization of PG I (biglycan,BGN) and PCI! (decorin, DCN, PG 40) genes on human chromo- tein molecules and to turn on and off growth factors. somes Xq13 qter and 12q, respectively. Genomics 6,219-225 Some proteoglycans regulate cell growth, others promote 16. Sztrolovics, R., Chen, X., Grover, J., Roughley, P. J., and Korenherg, j. R. (1994) Localization of the human fibromodulin gene (FMOD) to chromosome differentiation and neurite outgrowth, and still others act 1q32 and completion of the cDNA sequence. Genomics 23, 715-717 as biological barriers and repellents. These opposing 17. Fisher, L. W., Heegaard, A. M., Vetter, U., Vogel, W., Just, W., Termine, J. D., and Young, M. F. (1991) Human biglycan gene. Putative promoter, intron functions should not come as a surprise given their multi- exon Junctions, and chromosomal localization. J. Biol. C/tern. 266, valent nature. During development and tissue regenera- 14371-14377 18. Wegrowski, Y., Pillarisetti, J., Danielson, K. C., Suzuki, S.. and lozzo, H. V. tion, these roles are rendered accessible by virtue of their (1995) The murine biglycan: complete eDNA cloning, genomic organization, multipolar conformations. The proteoglycan realm is now promoter function and expression. Genornics 30,8-17 mature enough to produce important biological discover- 19. Antonsson, P., Heineglrd. D., and Oldberg, A.(1993) Structure and deduced amino acid sequence of the human fibromodulin gene. Biochirn. Biophys. ies. The isolation, characterization, and mapping of novel Acia 1174,204-206 20. Vogel, K. C., Paulsson, M., and ileineglrd, D. (1984) Specific inhibition of proteoglycan-encoding genes, together with the utilization type land type II collagen uibrillogenesis by the small proteoglycan of tendon. of mutant animals with targeted disruption of single pro- Biochem. J. 223,587-597

612 Vol. 10 April 1996 The FASEB Journal IOZZO AND MURDOCH REVIEW

21. SchOnherr, E., Witsch Prehm, P., Harrach, B., Robenek, H., Rauterberg, J., epidermis and in association with the elastic network of the dermis.). Cell and Kresse, H. (1995) Interaction of biglycan with type I collagen.). Biol. Biol. 124,817-825 Chem. 270, 2776-2783 49. LeBaron, R. C., Zimmermann, D. R., and Ruoslahti, E. (1992) Hyaluronate 22. Hedbom, E., and HeinegIrd, D. (1989) Interaction of a 59 kDa connective binding properties of versican.). Biol. Chern. 267, 10003-10010 tissue matrix protein with collagen I and collagen II. J. Biol. C/tern. 264, 50. Naso, M. F., Zimmermann, D. R., and lozzo, R. V. (1994) Characterization 6898-6905 of the complete genomic structure of the human versican gene and functional 23. Krumdieck, R., Hook, M., Rosenberg, L. C., and Volanakis, J. E. (1992) The analysis of its promoter. J. Biol. C/tern. 269, 32999-33008 proteoglycan decorin binds Clq and inhibits the activity of the Ci complex. 51. Dours Zimmermann, M. T., and Zimmermann, D. R. (1994) A novel gly- J. Immunol. 149, 3695-3701 cosaminoglycan attachment domain identification in two alternative splice 24. Bidanset, D. J., Guidry, C., Rosenberg, L. C., Choi, H. U., Timpl. R.. and variants of human versican.). Biol. C/tern. 269, 32992-32998 Hook, M. (1992) Binding of the proteoglycan decorin to collagen type IV. J. 52. Roughley. P.J.. and Lee. E. R. (1994) Cartilage proteoglycans: structure and Biol.Chem. 267,5250-5256 potential functions. Micro. Res. Techniq. 28,385-397 25. Schmidt, C., Hausser, H., and Kresse, H. (1991) Interaction of the small 53. Fulop, C., Walcz, E., Valyon, M., and GIant, T. 1. (1993) Expression of proteoglycan decorin with fibronectin. Involvement of the sequence NKISK alternatively spliced epidermal growth factor like domains in of of the core protein. Biochern.). 280,411-414 different species. Evidence for a novel module. ). Biol. C/tern. 268, 26. Hildebrand, A., Romaris, M., Rasmussen, L. M., Heineglrd, D., Twardzik, 17377-17383 D. R., Border, W. A., and Ruoslahti, E. (1994) Interaction of the small 54. Drickamer, K. (1993) Ca2v dependent carbohydrate recognition domains in interstitial proteogycans biglycan, decorin and fibromodulin with transform- animal proteins. Curr. Opin. Struct. Biol. 3,393-400 ing growth factor i. Biochem.). 302, 527-534 55. Rauch, U., Karthikeyan, L., Maurel, P., Margolis, H. U., and Margolis, H. K. 27. Takahashi, T., Cho, H. I., Kublin, C. I., and Cintron, C. (1993) Keratan (1992) Cloning and primary structure of neurocan, a developmentally regu- sulfate and dermatan sulfate proteoglycans associate with type VI collagen lated, aggregating chondroitin sulfate proteoglycan of brain.). Biol. C/tern. in fetal rabbit cornea. I. Histochern. Cytochern. 41, 1447-1457 267, 19536-19547 28. Hedbom, E., and HeinegArd, D. (1993) Binding of fibromodulin and decorin 56. Margolis, H. U., and Margolis, R. K. (1994) Aggrecan versican neurocan to separate sites on llbrillar collagens. J.Biol. Chern. 268,27307-27312 family of proteoglycans. Methods Enzyrnol. 245, 105-126 29. Rada, J. A., Comuet, P. K., and Hassell, J. R. (1993) Regulation of corneal 57. Yamada, H., Watanabe, K., Shimonaka, M., and Yamaguchi, V. (1994) collagen fibrillogenesis in vitro by comeal keratan sulfate proteoglycan Molecular cloning of brevican, a novel brain proteoglycan of the aggre- (lumican) and decorin core proteins. Exp. ETe Res. 56, 635-648 can/versican family.). Biol. Chern. 269, 101 19-10126 30. Svensson, L., HeinegArd, D., and Oldberg, A. (1995) Decorin binding sites 58. lozzo, R. V., Naso, M. F., Cannizzaro, L. A., Wasmuth, J. J., and McPherson, for collagen type I are mainly located in leucine rich repeats 4 5.). Biol. J. D. (1992) Mapping of the versican proteoglycan gene (CSPC2) to the long Chem. 270, 207 12-20716 arm of human chromosome 5 (Sq 12 3q14). Genomics 14,845-851 31. SchOnherr, E., Hausser, H., Beavan, L., and Kresse, H. (1995) Decorin type 59. Naso, M. F., Morgan, J. L., Burchberg, A. M., Siracusa, L. D., and Iozzo, H. I collagen interaction. Presence of separate core protein binding domains.). V. (1995) Expression pattern and mapping of the murine versican gene Biol. C/tern. 270,8877-8883 (Cspg2) to chromosome 13. Genornics 29, 297-300 32. Hedlund, H., Mengarelli Windholm, S., Heineg#{226}rd,D., Reinholt, F. P., and 60. Doege, K. .1., Garrison, K., Coulter, S. N., and Yamada, V. (1994) The Svensson, 0. (1994) Fibromodulin distribution and association with collagen. structure of the rat aggrecan gene and preliminary characterization of its Matrix Biol. 14f;45, 227-232 promoter.). Biol. Chern. 269, 29232-29240 33. Cornuet, P. K., Blochberger, T. C., and Hassell, J. R. (1994) Molecular polymorphism of lumican during corneal development. Invest. Ophihalmol. 61. Watanabe, H., Gao. L., Sugiyama, S., Doege, K., Kimata, K., and Yamada, Vis. Sci. 35,870-877 Y. (1995) Mouse aggrecan, a large cartilage proteoglycan: protein sequence, 34. Cuo, B., Norris, S., Rosenberg, L. C., and Hook, M. (1995) Adherence of gene structure and promoter sequence. Biochern.). 308, 433-440 Borrelia burgdorferi to the proteoglycan decorin. Infect. Irnrnun. 63, 62. Valhmu, W. B., Palmer, C. D., Rivers, P.A., Ebara, S., Cheng, J. F., Fischer, 3467-3472 S., and Ratcliffe, A. (1993) Structure of the human aggrecan gene: exon intron 35. Yamaguchi, V., Mann, D. M., and Ruoslahti, E. (1990) Negative regulation organization and association with the protein domains. Biochern. J. 309, of transforming growth factor by the proteoglycan decorin. Nature (London) 535-542 346,281-284 63. Korenberg, J. H., Chen, X. N., Doege, K., Grover, J., and Roughley, P. J. 36. Bianco, P., Fisher, L. W., Young, M. F., Termine, J. D., and Robey, P. C. (1993) Assignment of the human aggrecan gene (AGC1) to 15q26 using (1990) Expression and localization of the two small proteoglycans biglycan flurorescence in situ hybridization analysis. Genornics 16, 546-548 and decorin in developing human skeletal and non skeletal tissues. ). 64. Watanabe, H., Kimata, K., Line, S., Strong, D., Gao, L. y., Kozak, C. A., and Histochem. Cytochem. 38,1549-1563 Yamada, Y. (1994) Mouse cartilage matrix deficiency (Cmd) caused by a 7 37. Santra, M., Danielson, K. C., and Iozzo, R. V. (1994) Structural and func- bp deletion in the aggrecan gene. Nature Genet. 7, 154-137 tional characterization of the human decorin gene promoter. J. Biol. Chem. 65. Rauch, U., Grimpe, B., Kulbe, C., Arnold Ammer, I., Beier, D. R., and 269,579-587 F#{228}ssler,H. (1995) Structure and chromosomal localization of the mouse 38. Mauviel, A., Santra, M., Chen, Y. Q.,Uitto, J., and lozzo, R. V. (1995) neurocan gene. Genornics 28, 405-410 Transcriptional regulation of decorin gene expression. Induction by quies- 66. Grover, J., and Roughley, P. J. (1993) Versican gene expression in human cence and repression by a. J. Biol. C/tern. 270, articular cartilage and comparison of mRNA splicing variation with aggrecan. 11692-i 1700 Biochem. J. 291,361-367 39. lozzo, H. V., and Cohen, 1.(1993) Altered proteoglycan gene expression and 67. Landolt, R. M., Vaughan, L., Winterhalter, K. H., and Zimmermann, D. H. the tumor stroma. Experientia 49, 447-455 (1995) Versican is selectively expressed in embryonic tissues that act as 40. Khh#{228}ri,V., Lajarava, H., and Uitto, J. (1991) Differential regulation of barriers to neural crest cell migration and axon outgrowth. Development 121, extracellular matrix proteoglycan (PG) gene expression.). Biol. C/tern. 266, 2303-2312 10608-10615 68. lozzo, R. V., Cohen,!. R., CrSssel, S., and Murdoch, A. D. (1994) The biology 41. KOh#{228}ri,V. M., Hakkinen, L., Westermarck, J., and Latjava, H. (1995) of perlecan: the multifaceted heparan sulphate proteoglycan of basement Differential regulation of decorin and biglycan gene expression by dex- membranes and pericellular matrices. Biochem. J. 302, 625-639 amethasone and retinoic acid in cultured human skin fibroblasts.). Invest. 69. lozzo, R. V. (1994) Perlecan: a gem of a proteoglycan. Matrix Biol. 14, Dermasol. 104,503-508 203-208 42. Pearson, D., and Sasse, J. (1992) Differential regulation of biglycan and 70. Bork, P., and Patthy, L. (1995) The SEA module: a new extracellular domain decorin by retinoic acid in bovine chondrocytes. J. Biol. Chern. 267, associated with 0 glycosylation. Protein Sci. 4, 1421-1425 25364-25370 71. Tsen, C., Halfter, W., Kroger, S., and Cole, C. J. (1995) Agrin is a heparan 43. Asundi, V. K., and Dreher, K. L. (1992) Molecular characterization of sulfate proteoglycan. I. Biol. C/rem. 270. 3392-3399 vascular smooth muscle decorin: deduced core protein structure and regula- 72. Rupp, F., Payan, D. C., Magill SoIc, C., Cowan, D. M., and Scheller, R. H. tion of gene expression. Ear. J. Cell Biol. 59, 314-321 (1991) Structure and expression of a rat agrin. Neuron 6,811-823 44. Coppock, D. 1., Kopman, C., Scandalis, S., and Cilleran, S. (1993) Prefer- 73. Rutishauser, U., Acheson, A., Hall, A. K., Mann, D. M., and Sunshine, J. ential gene expression in quiescent human lung fibroblasts. Cell Growth (1988) The neural cell adhesion molecule (NCAM) as a regulator of cell cell Dfferentiation 4,483-493 interactions. Science 240, 53-57 45. Adany, R., Heimer, R., Caterson, B., Sorrell, J. M., and lozzo, R. V. (1990) 74. Hagen, S. C., Michael, A. F., and Butkowski, H. J. (1993) Immunochemical Altered expression of chondroitin sulfate proteoglycan in the stroma of human and biochemical evidence for distinct basement membrane heparan sulfate colon carcinoma. Hypomethylation of PG 40 gene correlates with increased proteoglycans. J. Biol. C/tern. 268, 726 1-7269 PG 40 content and mRNA levels. J. Biol. Chem. 265, 11389-11396 75. Patthy, L., and Nikolics, K. (1994) Agrin like proteins of the neuromuscular 46. Santra, M., Skorski, T., Calabretta, B., Lattime, E. C., and lozzo, R. V. (1995) junction. Neuroc/tern. hit. 24,301-316 De novo decorin gene expression suppresses the malignant phenotype in 76. Alliel, P. M., Perin, J., Jolles, P.. and Bonnet. F. (1993) Testican, a multido- human colon cancer cells. Proc. Nail. Acad. Sci. USA 92,7016-7020 main testicular proteoglycan resembling modulators of cell social behaviour. 47. Zimmermann, D. H., and Ruoslahti, E. (1989) Multiple domains of the large Eur. J. Bioc/tem. 214,347-350 fibroblast proteoglycan. versican. EMBO J. 8,2975-2981 77. Cheng, J., Boyartchuk, V., and Zhu, V. (1994) Isolation and mapping of 48. Zimmermann, D. R., Dours Zimmermann, M. T., Schubert, M., and Bruckner human chromosome 21 cDNA: progress in constructing a chromosome 21 Tuderman, L. (1994) Versican is expressed in the proliferating zone in the expression map. Genornics 23, 73-84

PROTEOGLYCANS 613 REVIEW

78. Lane, 1. F., and Sage, E. H. (1994) The biology of SPARC, a protein that cysteine rich regions of perlecan domain Ill. J. Histochem. Cytochem. 43, modulates cell matrix interactions. FASEB J. 8, 163-173 955-963 79. Murdoch, A. D., Dodge, C. R., Cohen, I., Tuan, R. S., and lozzo, R. V. (1992) 86. Hayashi, K., Madri, J. A., and Yurchenco, P. D. (1992) Endothelial cells Primary structure of the human heparan sulfate proteoglycan from basement interact with the core protein of basement membrane perlecan through Ill membrane (HSPGperlecan). A chimeric molecule with multiple domains and 3 integrins: an adhesion modulated by glycosaminoglycans.J. Cell Biol. homologous to the low density lipoprotein receptor, laminin, neural cell 119,945-939 adhesion molecules, and epidermal growth factor. I. Biol. C/tern. 267, 87. Klein, C., Conzelmann, S., Beck, S., Timpl, R., and Muller, C. A. (1995) 8544-8357 Perlecan in human bone marrow: a growth factor presenting, but anti adhe- 80. Dodge, C. R., Kovalszky, I., Chu, M. L., Hassell, J. H., McBride, 0. W., Yi, sive, extracellular matrix component for hematopoietic cells. Matrix Biol. H. F., and lozzo, R. V. (1991) Heparan sulfate proteoglycan of human colon: 14,457-465 partial molecular cloning, cellular expression, and mapping of the gene 88. Grgssel, S., Cohen, I. R., Murdoch, A. D., Eichstetter, I., and Iozzo, R. V. (HSPC2) to the short arm of human chromosome 1. Genomics 10,673-680 (1995) The proteoglycan perlecan is expressed in the erythmleukemia cell 81. Cohen, I. H., Crgssel, S., Murdoch, A. D., and lozzo, R. V. (1993) Structural line K562 and is upregulated by sodium butyrate and phorbol ester. Mol. characterization of the complete human perlecan gene and its promoter. Proc. Cell. Biochern. 145,61-68 Nail. ,4cad. Sci. USA 90, 10404-10408 89. Battaglia, C., Mayer. U., Aumailley, M., and Timpl, R. (1992) Basement 82. Rupp, F., Ozcelik, 1., Linial, M., Peterson, K., Francke, U., and Scheller, R. membrane heparan sulfate proteoglycan binds to laminin by its heparan (1992) Structure and chromosomal localization of the mammalian agrin gene. sulfate chains and to nidogen by sites in the protein core. Eur. J. Biochern. ). Neurosci. 12, 3535-3544 208,359-366 83. Murdoch, A. D., Liu, B., Schwarting, H., Tuan, R. S., and Iozzo, R. V. (1994) Widespread expression of perlecan proteoglycan in basement membranes 90. Aviezer, D., Hecht, D.. Safran, M., Fisinger, M., David, C., and Yayon. A. and extracellular matrices of human tissues as detected by a novel mono- (1994) Perlecan, basal lamina proteoglycan, promotes basic fibroblast growth clonal antibody against domain Ill and by in situ hybridization.). Histochern. factor receptor binding, mitogenesis, and angiogenesis. Cell 79, 1005-1013 Cytochem. 42,239-249 91. Hoch, W., Campanelli, J. 1., and Scheller, R. H. (1994) Agrin induced 84. SundarRaj, N., Fite, D., Ledbetter. S., Chakravarti, L., and Hassell, J. R. clustering of acetylcholine receptors: a cytoskeletal link. J. Cell Biol. 126, (1995) Perlecan is a component of cartilage matrix and promotes chondrocyte 1-4 attachment.). Cell Sci. 108, 2601-2604 92. Tsen, C., Napier, A., Halfter, W., and Cole, C. J. (1995) Identification of a 83. Couchman, J. R., Ljubimov, A. V., Sthanam, M., Horchar, T., and Hassell, novel alternatively spliced agrin mRNA that is preferentially expressed in J. R. (1995) Antibody mapping and tissue localization of globular and nonneuronal cells. J. Biol. C/tern. 270, 15934-15937

10th Symposium of The Protein Society San Jose, California August 3-7, 1996

In addition to the sessionslisted below, numerous poster sessions are detailed in the Program, which will be mailed to preregistrants prior to the meeting. SATELLITE MEETING WORKSHOPS I. Data Bases-ESTs to Coordinates. J. Sussman, ABRF Symposium and Award Ceremony B. Berger, F. Spencer PLENARY SESSION-Protein Complexes II. Evaluating the Quality of X-ray and NMR H. Huber, S. Yoshikawa Structures

I. Receptor Activation and Signalling. W. STEIN AND MOORE SYMPOSIUM Hendrickson, S. Ealick D. Wiley, T. Terwilliger, J. Collier, D. Eisenberg II. CellAdhesion. R. Hynes, B. Liddington IX. Characterizing Interactions by Mass Ill. Membrane Protein Complexes. E. Gouaux, W. Spectroscopy. D. Hunt, S. Radford Kuhkbrandt X. Novel Cofactors and Enzymes. IV. Pathways for Electron Transport. H. Gray, D. Beratran P. Leadlay, P. Knowles LINSTROM-LANG SYMPOSIUM AWARDS BANQUET J. Scheliman, W. Englander Speaker-R. Doolittle XI. Microscopy: From Single Molecules to Large V. Protein Folding and Biochemical Diversity. G. Rose, J. Wells Complexes. P. Hansma, H. Saibil VI. Biochemical Feedback and Regulation XII. RNA -Protein Complexes. A. Frankel, J. Networks. A. Clarke, H. McAdams Frank VII. Protein Folding and Biochemical Diversity. CLOSING PLENARY SESSION K. Dill, S. Kent. VIII. The Cell Cycle. N. Pavietich, D. Morgan DNA-Repair Systems. G. Verdine, J. Deisenhofer

For further information contact: Dr. R. Newburgh The Protein Society l-800-99AMINO or 301-571-0662

614 Vol. 10 April 1996 The FASEB Journal IOZZO AND MURDOCH