Heparan Sulfate Proteoglycans: A GAGgle of Skeletal-Hematopoietic Regulators

Summary

This review summarizes our current understanding of the presence and function of heparan sulfate proteoglycans (HSPGs) in skeletal development and hematopoiesis. Although proteoglycans (PGs) comprise a large and diverse group of cell surface and matrix molecules, we chose to focus on HSPGs owing to their many proposed functions in skeletogenesis and hematopoiesis. Specifically, we discuss how HSPGs play predominant roles in establishing and regulating niches during skeleto-hematopoietic development by participating in distinct developmental processes such as patterning, compartmentalization, growth, differentiation, and maintenance of tissues. Special emphasis is placed on our novel hypothesis that mechanistically links endochondral skeletogenesis to the establishment of the hematopoietic stem cell (HSC) niche in the marrow. HSPGs may contribute to these developmental processes through their unique abilities to establish and mediate morphogen, growth factor, and cytokine gradients, facilitate signaling, provide structural stability to tissues, and act as molecular filters and barriers.

Tribute to Dr. Elizabeth D. Hay

Dr. Elizabeth Dexter Hay, “Betty” to her friends, would often introduce her seminars on extracellular matrix (ECM) and epithelial-mesenchymal cell transformations by showing pictures of her cats asleep intertwined with a blanket. She would use this analogy to highlight the intimate association that exists between a cell and its ECM during development. She was a pioneer who brought forth the notion that the ECM is not just a static “stuffing between cells”, but has interactive and instructive roles in development.

Betty’s contribution to the ECM field began with her seminal discoveries in the emerging field of electron microscopy. She worked with Keith Porter, George Palade, Don Fawcett, Susumo Ito, Jean-Paul Revel and others, to unravel the intricate microanatomy of the cell and its ECM. Once, when working with Jean-Paul Revel to extend autoradiography to the ultrastructural level, Betty noticed that tritiated proline was incorporated into collagen in the ECM outside of the cartilage cells. Betty went on to demonstrate that epithelial cells, as well as other non-fibroblastic cells, do secrete collagen. Moreover, in her descriptions of the development and structure of the various matrices, Betty brought forth the hypothesis that the ECM interacts with cells, and that through these interactions, the cells are instructed to modify their behavior. This notion was showcased in the ”Epithelial-Mesenchymal Transformation Model” she proposed to describe the dramatic morphological transformations of epithelia cells suspended in collagen gels.

Among her numerous scientific accomplishments, Betty can be considered a founder of the field of cell biology. Moreover, through her personal attributes and interactions with colleagues and her protegees, Betty has set the standard for many a scientist. This review is a tribute to Betty, a friend, colleague, and mentor, and highlights our developing premise that the ECM establishes both static and transient niches during vertebrate embryogenesis.

Introduction

Heparan sulfate proteoglycans (HSPGs) are key extracellular matrix (ECM) components of the skeleton (Koyama et al., 1996; Litwack et al., 1998), where they act as regulators of several dynamic aspects of skeletal development including patterning, differentiation, growth, and homeostasis (Arikawa-Hirasawa et al., 1999; Cano-Gauci et al., 1999; Costell et al., 1999; Rodgers et al., 2007). Moreover, heparan sulfate (HS) within the growth plate and bone marrow is also implicated as a critical component of the hematopoietic stem cell niche (Bruno et al., 1995). Marrow establishment is dependent on endochondral skeletal development, emphasizing the importance of HSPGs as skeletal-hematopoietic regulators. We propose that HSPGs may function by orchestrating and delineating temporal-spatial niches in at least three ways. First, HPSGs act as co-receptors for morphogens (Hufnagel et al., 2006), which determine the location of a developmental niche. Second, HSPGs can sequester growth factors and cytokines to regulate cell differentiation and growth within the niche (Chen et al., 2007a). Third, HSPGs help compose ECM scaffolds that physically separate the niche from cellular and signaling influences of neighboring niches or tissues. As a corollary to this hypothesis, disruption or dysregulation of niche functions resulting from mutations in HSPGs or in their binding partners might result in skeletal and/or hematopoietic disease phenotypes. This review will focus on discussing a novel HSPG niche hypothesis of skeleto-hematopoietic development and disease, with a special emphasis on endochondral ossification (EO) and subsequent hematopoiesis in EO-derived tissues.

Categories of Proteoglycans

Proteoglycans (PGs) consist of specific core proteins to which a variable number of polysaccharide chains, or glycosaminoglycans (GAGs), are covalently attached. Several reviews cover the biosynthesis, modifications, and structural variations of the different GAG chains (Oldberg et al., 1990; Kjellen and Lindahl, 1991; Iozzo, 1998; Esko and Lindahl, 2001; Bulow and Hobert, 2006). The six types of GAGs, which include hyaluronic acid or hyaluronan (HA), keratan sulfate (KS), chondroitin sulfate (CS), dermatan sulfate (DS), HS, and heparin (HN), represent straight chain polymers of disaccharide repeats of d-glucuronic or l-iduronic acid, and either N-acetylglucosamine or N-acetlygalactosamine (Fig 1, colored boxes). With the exception of HA, each GAG is synthesized on a core protein. Due to their high content of sulphate and carboxyl groups, complex patterns of sulfation, and uronic acid epimerizations, GAG chains confer upon PGs the diverse capacities to function as ideal physiological barriers, reservoirs for signaling proteins, and binding partners for structural macromolecules. Thus, while core proteins certainly have important functions, many roles of PGs may be attributed to the chemistry of their GAG chains. The unique combinations of core proteins with specific GAGs impart a remarkable degree of diversity among PGs, and provide a single PG the ability to play diverse roles in a time and tissue-dependant manner (Bulow and Hobert, 2006).

Figure 1. Cartoon representation of the six categories of PGs and the five GAG disaccharide structures.

Representative members of each of the six PG categories detailed in this review together with the molecular structure of five GAG disaccharide units are drawn. Aggrecan, containing both KS and CS GAG chains attached to HA, represents the lectican PG family. Decorin, containing a DS GAG, represents the SLRPs. Perlecan and collagen XVIII, both depicted carrying HS GAGs, represent ECM PGs. A glypican is drawn to represent the GPI-linked family of PGs, with the GPI linkage shown at the plasma membrane (PM). A syndecan, carrying both HS and CS GAGs, represents the membrane-spanning class of PGs. Serglycin, drawn to carry HN and/or HS GAGs (HN is not pictured since it differs from HS only in the extent of sulfation), is drawn within a secretory vesicle (SV) to represent this PG class. PGs are drawn to approximate scale and cellular/extracellular location. Disaccharide units comprising five distinct GAG components of PGs, DS, KS, HA, CS, and HS, are drawn near corresponding GAG chains. GAG chains are represented by long thin colored lines emerging from the core proteins. HSPGs are found in all families except the lecticans and SLRPs. ES, extracellular; IS, intracellular; N, nucleus.

The five GAG components of PGs may be distinguished based on their disaccharide composition and sulfate positions, and can be generally divided into two groups: glucosaminoglycans, which include HN, HS, and KS, and galactosaminoglycans, which include CS and DS. One or more GAG chain(s) from either of the groups is covalently bound to a PG core protein (Table 1), providing an often-used PG classification scheme according to the type of GAG attached. The propensity of some PGs to carry more than one type of GAG makes this manner of classification imprecise. Perlecan is an excellent example of this elegant mode of diversity as its core protein can carry the galactosaminoglycan CS, or the glucosaminoglycans HS, and/or KS (Knox et al., 2005); moreover, these differences are tissue- and time-dependent. Perlecan and all four syndecans can additionally carry HS and CS chains concomitantly, reflecting their complexity in form and function (Ueno et al., 2001; French et al., 2002). The exact chemical composition of HS GAG chains, including the positions of its sulfate, carboxyl, and hydroxyl groups, confers particular properties to HSPGs such as the ability to confer growth factor ligand-receptor specificity in fibroblast growth factor receptor (FGFR) signaling (Guimond and Turnbull, 1999; Harmer, 2006).

Table 1. HSPGs: distribution, mouse models, associated human disorders, and proposed functions.

HSPGs found within the skeletal and/or hematopoietic system are in boldface. NA, data not available.

HSPG	Tissue Distribution	Mouse Model(s)	Human Disorder(s)	Proposed Function
Agrin	Basement membranes of muscle (Lieth et al., 1992), kidney (Groffen et al., 1998), brain (O'Connor et al., 1994); immune system (Khan et al., 2001); chondrocytes (Hausser et al., 2007)	Null mice (Gautam et al., 1996) and agrin-z null mice (Burgess et al., 1999) have NMJ defects and die at birth; NMJ-rescued mice survive with skeletal defects (Hausser et al., 2007)	NA	AChR aggregation, NMJ formation (McMahan, 1990); T-cell signaling (Khan et al., 2001); skeletogenesis (Hausser et al., 2007)
Betaglycan	Mesenchyme, epithelial cells, neurons (Boyd et al., 1990)	Null mice die embryonically of heart and liver defects (Stenvers et al., 2003)	NA	TGF-β signal transduction (Lopez-Casillas et al., 1994)
Collagen XV	Vascular, neuronal, mesenchymal, and some epithelial basement membranes (Myers et al., 1996)	Null mice have mild muscle and capillary defects (Eklund et al., 2001)	NA	Stabilizes skeletal muscle and microvessels (Eklund et al., 2001); *restin, the NC1 domain of collagen XV, may be anti-angiogenic (Ramchandran et al., 1999)
Collagen XVIII	Wide expression in basement membranes (Oh et al., 1994; Halfter et al., 1998)	Null mice have eye abnormalities modeling Knoblock syndrome (Fukai et al., 2002) and basement membrane defects (Utriainen et al., 2004)	Knoblock Syndrome (Suzuki et al., 2002)	Eye blood vessel formation (Strader et al., 2004); basement membrane stability (Utriainen et al., 2004); *endostatin, a C-terminal fragment of collagen XVIII, is an anti-angiogenic factor (O' Reilly et al., 1997)
Epican	Keritinocytes (Haggerty et al., 1992)	NA	NA	Mediates cell-cell adhesion (Milstone et al., 1994)
Glypican-1	Neurons, periosteum, bony trabeculae, bone marrow, epidermis, and kidney (Litwack et al., 1998); myoblasts (Bonneh-Barkay et al., 1997)	NA	NA	Establish morphogen gradients (Hufnagel et al., 2006)
Glypican-2	Dynamic neuronal expression (Stipp et al., 1994)	NA	NA	Neuronal motility (Stipp et al., 1994)
Glypican-3	Cartilage condensations (Pellegrini et al., 1998); embryonic mesoderm	Null mice model SGBS (Cano-Gauci et al., 1999); null mice have altered hematopoiesis and delayed EO (Viviano et al., 2005)	Simpson-Golabi-Behmel syndrome (Pilia et al., 1996); marker for hepatocellular carcinoma (Hsu et al., 1997)	Inhibit cell proliferation; lineage-specific blood cell differentiation; tumor suppressor (Gonzalez et al., 1998)
Glypican-4	Developing brain (Ybot-Gonzalez et al., 2005); kidney (Karihaloo et al., 2004)	NA	Associated with Simpson-Golabi-Behmel syndrome (Veugelers et al., 1998)	Neural patterning (Luxardi et al., 2007)
Glypican-5	Developing brain, CNS, limb, kidney (Saunders et al., 1997)	NA	NA	Growth factor signaling in embryonic mesenchyma (Saunders et al., 1997)
Glypican-6	Wide expression including mesenchyme and blood vessels (Veugelers et al., 1999)	NA	NA	Growth factor signaling in embryonic mesenchyma (Veugelers et al., 1999)
Perlecan	Limb bud mesenchyme (Solursh and Jensen, 1988); growth plate cartilage (Handler et al., 1997); articular cartilage (SundarRaj et al., 1995);bone marrow stroma (Schofield et al., 1999); all basement membranes (Hassell et al., 1980)	Hypomorphic mice model SJS (Rodgers et al., 2007); null mice model DDSH (Arikawa-Hirasawa et al., 1999; Costell et al., 1999)	Schwartz-Jampel syndrome (Nicole et al., 2000); Dyssegmental Dysplasia, Silverman-Handmaker type (Arikawa-Hirasawa et al., 2001)	Growth factor signaling; collagen fibrilogenesis; structural stability; vasculogenesis; *endorepellin, a C-terminal fragment of perlecn is a n anti-angiogenic factor (Mongiat et al., 2003)
Serglycin	Hematopoietic and endothelial cells (Kolset et al., 2004)	Null mice have defective mast cell secretory granule maturation (Abrink et al., 2004) and platelet aggregation (Woulfe et al., 2008)	NA	Granulopoiesis (Abrink et al., 2004); immunity (Niemann et al., 2007); platelet activation (Woulfe et al., 2008)
Syndecan-1	Limb mesenchyme (Solursh et al., 1990); osteoblasts (Molteni et al., 1999); pre-B cells, plasma cells, macrophages, and endothelial cells (Kim et al., 1994)	Null mice resist Wnt-1-induced tumorogenesis (Alexander et al., 2000) and microbial pathogenesis (Park et al., 2001)	NA	Matrix and growth factor receptor (Bernfield and Sanderson, 1990); trans HIV receptor (Bobardt et al., 2003); wnt signalling (Alexander et al., 2000)
Syndecan-2	Lung (Lories et al., 1989); wide mesenchymal expression (David et al., 1993); perichondrium and osteoblasts (Molteni et al., 1999); macrophages; endothelial cells (Kim et al., 1994)	NA	NA	Tissue morphogenesis (David et al., 1993); trans HIV receptor (Bobardt et al., 2003); cell aggregation
Syndecan-3	Mesenchymal condensations (Gould et al., 1992); proliferative chondrocytes (Kosher, 1998)	Null mice have mild hippocampal defects (Kaksonen et al., 2002) and impaired locomotion (Cornelison et al., 2004)	NA	Trans HIV receptor-Bobardt et al., 2003; modulates feeding behavior (Strader et al., 2004); muscular dystrophy and satellite cell function (Cornelison et al., 2004)
Syndecan-4	Perichondrium and osteoblasts (Molteni et al., 1999); articular chondrocytes (Grover and Roughley, 1995); neutrophils (Kaneider et al., 2001); lymphocytes and monocytes (Kaneider et al., 2002)	Null mice have no overt phenotype (Ishiguro et al., 2000)	NA	Proliferation; differentiation; adhesion; migration; trans HIV receptor (Bobardt et al., 2003)
Testican-1	Seminal fluid (Alliel et al., 1993); brain (Bonnet et al., 1996); blood (BaSalamah et al., 2001); cartilage (Hausser et al., 2004)	Null mice have no overt phenotype (Roll et al., 2006)	NA	Neuronal attachment (Marr and Edgell, 2003); MMP activation (Nakada et al., 2001)

A perhaps more accurate mode of classification is to segregate PGs based on properties of their core protein. This approach takes cellular/extracellular localization, size, and amino acid composition into account in addition to the type of GAG carried. PG core proteins may be membrane-spanning (Marynen et al., 1989; Saunders et al., 1989), glycophosphatidyl-inositol (GPI)-linked (Carey and Stahl, 1990; David et al., 1990), secreted, or intracellular (Kolset et al., 2004). The six primary categories of PGs (Fig. 1) using this nomenclature system are 1) Lecticans/hyalectans, which may be hybrid glucosaminoglycans/galactosaminoglycans and have protein core domain homology. Lecticans contain both a hyaluronan-binding domain and a C-type lectin domain. This family includes aggrecan, versican, neurocan, and brevican. All are found extracellularly and a splice form of brevican may also be tethered to the plasma membrane via a GPI-linkage. 2) Small leucine rich proteoglycans (SLRPs) (Iozzo, 1999; McEwan et al., 2006) are found intracellularly, extracellularly (most commonly associated with collagen fibrils), and at the cell surface. SLRP core proteins typically contain sequences of 10–12 tandemly repeated hydrophobic amino acids. This family includes decorin, biglycan, asporin, ECM2, keratocan, PRELP, osteoadherin, lumican, fibromodulin, opticin, epiphycan, osteoglycin, podocan, chondroadherin, and nyctalopin. 3) ECM PGs do not have a significant amount of core protein homology, but can all be found in basement membranes, elsewhere in extracellular matrices, and cell surface-associated. This family includes perlecan (Noonan et al., 1991), agrin (Tsen et al., 1995), collagen XV (Muragaki et al., 1994; Amenta et al., 2005), and collagen XVIII (Muragaki et al., 1994; Halfter et al., 1998). 4) Glycophosphatidyl-inositol (GPI)-linked PGs include glypicans 1–6 and a splice form of brevican. 5) Membrane-spanning PGs include syndecans 1–4, betaglycan, and CD44v3. 6) The best known class of secretory vesicle PGs are the serglycins, which contain either HN, HS, or CS and reside within secretory vesicles until their release from the cell (Kolset and Tveit, 2007).

With the exceptions of the lecticans and SLRPs, HSPGs are found in all of the PG families, emphasizing their diverse biological properties and functions. HSPGs include syndecans 1–4, glypicans 1–6, perlecan, agrin, betaglycan, epican, collagen XV, collagen XVIII, testican, serglycin, and possibly others (Table 1). Many of these molecules exhibit a temporo-spatial localization in skeleto-hematopoietic tissues, and play a role in endochondral skeletal formation and homeostasis, and/or hematopoiesis in the marrow. A number of genitically-modified mice have been developed to study in vivo roles of HSPGs, as listed in Table 1. Individual deletions of perlecan (Arikawa-Hirasawa et al., 1999; Costell et al., 1999), agrin (Gautam et al., 1996), and betaglycan (Stenvers et al., 2003) each result in perinatal lethality due to skeletal defects leading to respiratory distress, defective neuromuscular junctions (NMJ), and heart and liver defects, respectively. Interestingly, agrin NMJ-rescued mice present with altered EO, implying a role for agrin in skeletogenesis (Hausser et al., 2007). Mice with diminished, or hypomorphic, levels of perlecan (Rodgers et al., 2007), and glypican-3 null mice (Cano-Gauci et al., 1999) present with both defective EO and altered hematopoiesis (Viviano et al., 2005), strongly supporting our hypothesis that EO and hematopoiesis are intricately linked (Campbell et al., 2004)(Sweeney et al., manuscript accepted to this Journal issue). No skeletal phenotypes have yet been reported for the other HSPG mutant mice (Table 1). However, syndecan-1 deletion results in resistance to microbial pathogenesis (Park et al., 2001) and serglycin-null mice have defects in mast cells (Abrink et al., 2004) and platelet aggregation (Woulfe et al., 2008), providing evidence that these HSPGs are involved in hematopoiesis and/or immunity. Likewise, syndecan-1 has been shown to mediate B-cell interactions with the ECM including the establishment of a proper microenvironment, or niche (Sanderson et al., 1992). One can conceptualize the regulation of skeletogenesis and hematopoiesis by HSPGs, as they are uniquely suited to carry out these functions by establishing and/or maintaining niches that affect patterning, formation, growth, homeostasis, as well as hematopoietic niche establishment in the marrow, as reviewed here.

HSPGs and Skeletal Development

Overt skeletogenesis proceeds through two distinct processes, intramembranous ossification (IO), and EO. IO is responsible for formation of most craniofacial bones, including the flat bones of the skull, the lateral halves of the clavicles, and all the periosteal bones. This process involves a direct differentiation of mesenchymal cells to osteoblasts, bypassing the transient cartilaginous skeletal anlagen that is the hallmark of EO (Hall and Miyake, 1992). The end result of IO is formation of dense (cortical, compact) bone, comprised of a well-organized Haversian System, lacking a bone marrow cavity (Enlow, 1962). EO is responsible for the formation of the axial and appendicular skeleton, as well as certain cranial bones. Unlike IO, EO entails a cartilaginous template that delineates future skeletal elements, all of which contain a marrow cavity. The end result of EO is the formation of a porous trabecular (cancellous, spongy) bone engulfed by bone marrow (Mackie et al., 2008). Numerous reports have implicated trabecular and endosteal bone surfaces within the marrow as sites for the hematopoietic stem cell (HSC) niche (Nilsson et al., 1997). Data from our laboratory involving collagen X and perlecan mouse models extend the HSC niche to the chondro-osseous junction, where bone marrow interfaces with the cartilaginous growth plate and EO-derived trabecular spicules. The following discussion will focus primarily on how HSPGs help orchestrate processes prior to overt skeletogenesis, including patterning, condensation, differentiation, and growth, as well as their role in EO and HSC niche establishment.

Patterning of the vertebrate limb bud

Pattern formation is the process whereby the general features of the body plan are first specified along the rostral-caudal axis (Johnson and Tabin, 1997). Patterning of the presumptive skeletal elements and other limb features occurs along the anterior-posterior (AP), proximal-distal (PD), and dorsal-ventral (DV) axes. Studies using Drosophila and vertebrate models have shown that different signaling molecules control each patterning axis (Zeller and Duboule, 1997; Martin, 1998; Capdevila and Izpisua Belmonte, 2001) and have demonstrated a role for HSPGs in these signaling pathways (Lin, 2004).

AP axis patterning is thought to be regulated via Sonic hedgehog (Shh) by triggering the release of bone morphogenetic protein 2 (BMP2), which diffuses along the AP axis of the limb bud specifying the positions of skeletal elements and controlling their growth (Niswander et al., 1994). Homeobox (HOX) genes are the target of the Shh pathway (Knezevic et al., 1997), and are proposed to control pattern specification and subsequent development of specific skeletal elements. For example, mice null for both HOXA11 and HOXD11 are normal except for the extremely stunted growth of the radius and ulna (Davis et al., 1995), while humans harboring HOXD13 mutations display synpolydactyly and homeotic transformations of their metacarpal and metatarsal bones (Muragaki et al., 1996). Syndecan-1 has been shown to be transcriptionally regulated by HOX proteins (Bernfield et al., 1993), indicating a role for HSPGs in vertebrate HOX signaling. Similar to the HOX mutants, mice null for Shh lack identifiable AP polarity in regions distal to the forming elbow and knee joints (Chiang et al., 1996; Chiang et al., 2001). While the role of HSPG in the Shh pathway requires further elucidation in vertebrate AP axis patterning, information can be gleaned from fly studies. Improper HS biosynthesis in Drosophila results in altered Hedgehog (Hh; the fly ortholog of Shh) localization within the wing disk, (Bellaiche et al., 1998) and HSPGs have been proposed to protect Hh from degradation and to be involved in Hh mobility (Bornemann et al., 2004; Takei et al., 2004). Analogous roles of morphogen sequestration, distribution, and signaling can be presumed in vertebrates as at least one of their Shhs contain a highly conserved Cardin and Weintraub heparin-binding consensus sequence and bind heparin and HSPGs with high affinity (Rubin et al., 2002). In fact, perlecan has been shown to mediate Shh signaling in vertebrates (Datta et al., 2006a) and Shh has been shown to interact with both the perlecan core protein and its HS chains, leading to increases in both Shh signaling and prostate tumor metastasis (Datta and Datta, 2006; Datta et al., 2006b). Other HSPGs expressed by limb bud mesenchyme (Gould et al., 1992; Dealy et al., 1997; French et al., 1999; Dagoneau et al., 2004) may also contribute to AP patterning.

Limb outgrowth, or PD patterning, is promoted by a component of the ectodermal jacket of the limb bud, the apical ectodermal ridge (AER), which secretes fibroblast growth factors (FGFs) that maintain mesenchymal cell proliferation (Summerbell et al., 1973; Johnson and Tabin, 1997). Several FGFs have been functionally implicated in limb outgrowth, and FGF2, FGF4, and FGF8 are all expressed in various times and spatial distributions in the AER through limb development (Olwin et al., 1994; Dealy et al., 1997; Sun et al., 2002). For proper FGF activity, the growth factor must associate with both a receptor (FGFR) and a HSPG (Yayon et al., 1991), forming a HSPG-FGF-FGFR ternary complex, making, HSPGs mandatory for FGF function (Ornitz et al., 1995). Recent studies emphasize the modulation of FGF signaling by HSPGs in development. For example, improper HS GAG biosynthesis in Drosophila leads to defective FGF signaling during development (Lin et al., 1999) and Dally mutations (the fly homolog of glypican) lead to polarity defects attributable to altered FGF signaling (Lin and Perrimon, 1999). Surprisingly, the only reported human mutations in glypican (glypican-3) lead to Simpson-Golabi-Behmel syndrome, primarily characterized by overgrowth (Pilia et al., 1996). Glypican-3 null mice model this disease (Cano-Gauci et al., 1999) and additionally present with altered hematopoiesis and delayed EO (Table 1) (Viviano et al., 2005). Redundant functions of glypicans during limb outgrowth might explain the lack of patterning defects in these mice and human patients. HSPGs are likely involved in vertebrate patterning defects attributable to FGFR mutations (Coutts and Gallagher, 1995), possibly due to inefficient receptor signaling. Furthermore, lack of proper HS sulfation in chick development results in truncated limb buds (Kobayashi et al., 2007).

Last, DV limb bud patterning relies upon Wnt signaling, and the activities of the HOX genes lmx-1 and engrailed (Johnson and Tabin, 1997). Although the role of HSPGs in these pathways has not been well explored in the vertebrate skeleton, their function in Shh and Wnt signaling has been shown during muscle development (Bernfield et al., 1993; Dhoot et al., 2001) and syndecan-3 null mice present with impaired hind limb locomotion attributed to perturbed satellite cell function (Table 1) (Cornelison et al., 2004). In addition, HSPGs are implicated in Wnt signaling during analogous patterning pathways in Drosophila (Lin, 2004), and involve homologs of perlecan (Lindner et al., 2007) and glypican (Crickmore and Mann, 2007).

Together, these data demonstrate critical roles HSPGs play in limb bud patterning. These include sequestering of morphogens, establishing gradients of signaling molecules, and acting as co-receptors for growth factors. It is further plausible that HSPGs help establish ECM boundaries between niches in which these critical morphogens and signaling factors may function.

Condensation in the vertebrate limb bud

Among the first visible consequences of pattern formation in the limb bud is rapid condensation, or aggregation, of limb bud mesenchymal cells prior to chondrogenesis (Hall and Miyake, 2000). Various mechanisms of condensation have been proposed, including cadherin-mediated aggregation (Oberlender and Tuan, 1994), enzymatic removal of the space filling HA from inter-mesenchymal spaces (Li et al., 2007), and matrix-driven translocation (Newman et al., 1985), where polymerization of matrix molecules drives rapid cellular movements. Cell-surface HSPG interactions have been proposed to play a role in the latter model. For example, fibronectin binds HS (Newman et al., 1987; Jaikaria et al., 1991) and is upregulated at the onset of chondrogenesis at the sites of future skeletal elements (Dessau et al., 1980; Kosher et al., 1982; Tomasek et al., 1982; Kulyk et al., 1989). Furthermore, its deposition corresponds with the precartilage condensation of mesenchymal cells and appears to influence pattern formation (Frenz et al., 1989a; Frenz et al., 1989b; Downie and Newman, 1994). Blocking the 29kDa HS-binding amino terminal domain of fibronectin has been shown to inhibit mesenchymal condensation (Frenz et al., 1989b), implying that the HS binding partner of fibronectin, presumably a cell surface-associated HSPG of precartilage mesenchyme, may be of critical importance. Such HSPGs may include syndecan-3, which is highly expressed in chick limb bud mesenchyme at the onset of condensation (Gould et al., 1992), and perlecan, present throughout the process (French et al., 1999). Lastly, morphogens may also play a role in condensation. For example, condensation of precartilage mesenchyme is proposed to be mediated by Indian hedge hog (Ihh), via parathyroid hormone related peptide (PTHrP) regulation (Vortkamp et al., 1996), which may be directly influenced by HS (Hu et al., 2007).

Limb bud chondrogenesis

Secretion and assembly of cartilage specific ECM components including type II collagen (Kosher et al., 1986), aggregan (Mundlos et al., 1991), HA (Toole, 1972), and link protein (McKeown-Longo et al., 1983), are the first evidence of cartilage differentiation within limb bud mesenchymal condensations. HSPGs may be important at this stage of development as mediators of FGF signaling, which in turn may stimulate expression of Sox-9 (Murakami et al., 2000), a transcription factor crucial to chondrogenesis and sustained chondrocyte differentiation (Wagner et al., 1994; Wright et al., 1995; Bi et al., 1999). Indeed, a primary role of HSPGs during chondrogenic differentiation may be achieved via the activities of HS-binding morphogens, growth factors, and signaling molecules, including parathyroid hormone-like peptide (PTHrP), Indian hedgehog (Ihh), TGF-βs and FGFs; the key growth factors regulating chondrocyte differentiation (Vortkamp, 2001). Both glypican- 5 (Saunders et al., 1997) and the hybrid CS/HS syndecans are found during the early stages of chondrogenesis (Solursh et al., 1990; David et al., 1993) where they may interact with Hh proteins, and/or FGFs, thereby mediating chondrocyte differentiation (Moftah et al., 2002; Ornitz, 2005). Perlecan is another HSPG that may facilitate TGF-β signaling within the developing skeleton. For example, chondrocytes plated on matrices containing perlecan domain I and collagen II showed improved matrix binding and enhanced retention of BMP to support chondrogenic differentiation (Yang et al., 2006). BMPs are members of the TGF-β superfamily and act within the developing skeleton (Pogue and Lyons, 2006). While some specific HSPG/FGF interactions are known, controversy remains regarding their activities. For example, one study shows that syndecans and glypican are inhibitors of FGF (Aviezer et al., 1994), while others show syndecan-1 to bind FGF2 and mediate receptor activation (Salmivirta et al., 1992; Filla et al., 1998). Likewise, glypican-1 has been demonstrated to differentially modulate FGF1 and FGF7 in keratinocytes (Bonneh-Barkay et al., 1997). Taken together, it is likely that different situations call for different HS/FGF solutions, as previously suggested (Chang et al., 2000). These data highlight the importance of controlled and localized expression of these developmental proteins with which HSPGs are widely assumed to be universally involved.

An additional mode of HSPG regulation of chondrogenesis could be at the structural level. For example, HS and HN potently stimulate chondrogenesis in high density cultures of limb bud mesenchyme (San Antonio et al., 1987). These GAGs may bind cartilage matrix components, trap them around the newly emerging cartilage nodules, and thus create a “chondrogenic niche” (San Antonio et al., 1992), thereby stabilizing or promoting their differentiation. Indeed, syndecan-1 and HN bind type II collagen, the predominant cartilage protein (San Antonio et al., 1994). However, it is perlecan, whose expression is upregulated following the expression of collagen II by differentiated chondrocytes (Handler et al., 1997; French et al., 1999), that is required for proper fibrillogenesis of collagen I and collagen II (Kvist et al., 2006). While this action is largely attributed to the degree of CS sulfation of perlecan domain I GAG chains, the HS chains of this domain have also been reported as essential for chondrocyte differentiation (French et al., 2002), possibly via interaction with FGF2 (Knox et al., 2002; Yang et al., 2005). Indeed, perlecan has been shown to directly modulate the signaling of FGF2, (Govindraj et al., 2006), vascular endothelial growth factor (VEGF) (Jiang and Couchman, 2003), and likely members of the transforming growth factor beta (TGF-β) family (Ruppert et al., 1996; Takada et al., 2003; Kirn-Safran et al., 2004; Yang et al., 2006) on chondrocytes and to bind connective tissue growth factor/hypertrophic chondrocyte-specific gene product 24 (CTGF/Hsc24) (Nishida et al., 2003).

Taken together, chondrocyte cell surface HSPGs may help orchestrate the assembly of the elaborate ECM scaffold at the cell surface. In fact, an analogous role has been proposed for syndecans in fibronectin and laminin assembly at cell surfaces (Klass et al., 2000). Last, a role for perlecan in creating a barrier between chondrogenic mesenchyme and the neighboring undifferentiated cells may be proposed. In this scenario, perlecan deposited around the chondrogenic limb bud core could block heterotypic cell-cell interactions disruptive to chondrogenesis and concentrate or retain crucial growth and differentiation factors within the chondrogenic niche. This is plausible as, in addition to fibronectin and collagen, perlecan binds numerous ECM molecules and receptors that may be important for separation of distinct niches. These binding partners include nidogen, collagen IV, laminin (Hopf et al., 1999), α-dystroglycan (Talts et al., 1999), β1 and β3 integrins (Hayashi et al., 1992; Brown et al., 1997), and collagen XVIII (Sasaki et al., 1998), the parent molecule of endostatin, a potent angiogenesis inhibitor (O'Reilly et al., 1997).

Skeletal growth through EO

At a genetically predetermined stage, the cartilaginous skeletal anlagen begin to be replaced by trabecular bone and marrow through EO. A hallmark of this process is maturation of chondrocytes to hypertrophy, evidenced by chondrocyte enlargement and a cessation of proliferation (Godman and Porter, 1960; Chan and Jacenko, 1998). Upon hypertrophy, chondrocytes secrete a new repertoire of differentiation-specific gene products including collagen X (Gibson and Flint, 1985; Schmid and Linsenmayer, 1985) and VEGF (Gerber et al., 1999). VEGF signaling is essential for metaphyseal vascularization (Maes et al., 2002) and is likely mediated by HSPGs as inactivation of the two HN binding VEGF splice forms (Park et al., 1993) in mice leads to delayed blood vessel invasion (Zelzer et al., 2004). Vascular invasion is a critical step in skeletal growth as invading blood vessels import a mix of cells, including chondro/osteoclasts, mesenchymal cells, and hematopoietic precursors, which, together with growth factors, cytokines, and hormones, establish the primary center of ossification and the hematopoietic bone marrow cavity. Thus, proper chondrocyte differentiation, vascular invasion, and the gradual replacement of the cartilaginous anlagen by trabecular bone and marrow through EO, underscore the intricate orchestration of skeleto-hematopoiteic development. As discussed in more detail below, this skeleto-hematopoietic link is strongly supported by several animal models where altered EO leads to hematopoietic defects and includes collagen X mice (Jacenko et al., 2002), glypican-3 null mice (Viviano et al., 2005), and perlecan hypomorphic mice (unpublished observation). Taken together, these animal models provide additional support to the role of HSPGs in both EO and hematopoiesis.

Cartilage flanking the primary ossification center is comprised of distinct populations of chondrocytes representing different stages of maturation. As reviewed, (Lefebvre and Smits, 2005), a population of epiphyseal chondrocytes provides a pool of stem-like cells, some of which undergo hypertrophy and are invaded by blood vessels to form the secondary site of ossification defining the growth plate and articular joint regions. The HS side chains of perlecan may induce FGF2 signaling from within the growth cartilage (Garcia-Ramirez et al., 2000) to help mediate this vascular event (Aviezer et al., 1994). More distal epiphyseal chondrocytes, or resting cells, give rise to the proliferative zone of the growth plate, which consists of stacked columns of proliferative chondrocytes aligned parallel to the longitudinal axis of bone growth. Syndecan-3, primarily secreted by proliferative chondrocytes at this stage, is thought to not only regulate their division (Kosher 1998), but to compartmentalize proliferation to this specific growth plate zone (Pacifici et al., 2005). These roles may be accomplished through Ihh signaling (Shimo et al., 2004), as Ihh is also important at this stage of skeletal development (Long et al., 2001). Recent data also suggest that HSPGs mediate binding interactions between fibronectin and latent TGF-β-binding protein-1 (Chen et al., 2007a), which directly links the HN-binding domain of fibronectin with signaling cascades critical for growth plate function (Pedrozo et al., 1998).

As differentiation continues distally down the growth plate, proliferative chondrocytes hypertrophy and eventually undergo apoptosis or autophagy (Shapiro et al., 2005). Interactions between FGF2 and HS within the growth plate (Chintala et al., 1994) are critical for this terminal differentiation of hypertrophic chondrocytes (Chintala et al., 1995). Again in this circumstance, HSPGs appear to play a role in morphogen clustering and signal transduction. These signaling molecules come from the growth plate, which contains hypertrophic chondrocytes, and/or from the vascular bone marrow cavity. Within the marrow, invading vessels bring with them chondro/osteoclasts, which degrade the ECM surrounding hypertrophic chondrocytes (Vu et al., 1998). Remnants of the hypertrophic cartilage ECM then serve as scaffolds upon which osteoblasts deposit osteoid, thus establishing trabecular bone (Navagiri and Dubey, 1976; Chan and Jacenko, 1998). These trabecular spicules protrude into and are engulfed by the hematopoietic marrow (Jacenko et al., 2002), and likely represent the HSC niche (Nilsson et al., 1997; Nilsson et al., 2001). The continual replacement of hypertrophic chondrocytes by trabecular bone and marrow provides longitudinal skeletal growth and robust hematopoiesis until maturity, when in most non-rodent vertebrates EO ceases and growth plates close (Kilborn et al., 2002).

Altered perlecan expression exemplifies the critical involvement of HSPGs within the developing skeleton. Mice (Arikawa-Hirasawa et al., 1999; Costell et al., 1999; Rodgers et al., 2007) and humans (Nicole et al., 2000; Arikawa-Hirasawa et al., 2001) lacking adequate levels of perlecan present with skeletal dysplasia. Patients with Dysegmental Dysplasia of the Silverman-Handmaker type (DDSH) and perlecan null mice both develop severe chondrodysplasia characterized by a flat face, disorganized growth plate, cleft palate, and die by birth. Activating mutations in FGFR3 mice (Naski et al., 1998) lead to a skeletal phenotype similar to that of the perlecan null mice, and it has been suggested that similar signaling pathways involving Ihh and BMPs are affected in the perlecan null situation (Arikawa-Hirasawa et al., 1999). An altered ECM scaffold has also been proposed (Costell et al., 1999). Altered morphogen gradients, aberrant signaling, and destabilized ECM likely all play into the perlecan null phenotype as this HSPG has a role in all of these mechanisms.

A less severe disease, Schwartz-Jampel syndrome (SJS), is a progressive disorder caused by reduced levels of perlecan (Nicole et al., 2000) resulting in reduced stature, short tubular bones, distinguishable facial features, malformed hip structures, pigeon breast, and other forms of skeletal dysplasia (Aberfeld et al., 1965; Giedion et al., 1997; Spranger et al., 2000). Mice harboring a human genetic alteration described to cause SJS (G4595A) (Nicole et al., 2000) in addition to a neomycin selection cassette (C1532Yneo mice) present with a skeletal disease phenotype characteristic of SJS patients, including transcriptional alterations in HSPG2, the gene encoding perlecan, leading to reduced levels of perlecan secretion into the ECM (Rodgers et al., 2007). These perlecan hypomorphic mice, like SJS patients, are smaller than non-affected controls, as evidenced by comparing whole skeletons stained with Alizarin Red S for mineralized matrix and alcian blue for cartilaginous matrix (Fig. 2 A). This staining also reveals patterning defects in sterna of C1532Yneo mice (Fig. 2 B) correlating with the pigeon breast diagnosis often given to SJS patients, and possibly relating to a defect in patterning of specific skeletal anlagen during development. This idea is supported by the defective patterning of spinal cartilage elements (dyssegmental dysplasia) of DDSH patients and perlecan null mice. While perlecan null mice were reported to have normal cartilage anlagen of all long bones (Costell et al., 1999), it is plausible that this is not true for all EO-derived bones. Likewise, condensation of long bone elements does not appear to be altered in C1532Yneo mice. However, they do present with impaired primary ossification and delayed secondary ossification resulting in misshapen long bones (Rodgers et al., 2007). The mice display a transient expansion of hypertrophic cartilage (Fig. 3 A), disorganized growth plate zones (Fig. 3 B), and a horizontal/radial trabecular bone orientation (Fig. 3 C), consistent with defective EO and impaired skeletal growth. These phenotypes are likely attributable to a direct role of perlecan in collagen fibrilogenesis, growth factor signaling, and vascularization, as perlecan may act to establish morphogen gradients, sequester and mediate growth factor signaling, and possibly separate distinct niches via interactions with multiple ECM and receptor molecules.

Figure 2. Perlecan mutations in humans (SJS syndrome) and mice result in stunted growth and skeletal malformations.

A. Whole-mount skeletal staining using Alizarin red S for mineralized tissue and alcian blue for cartilaginous tissue shows decreased size of C1532Yneo newborn mice compared with wildtype (WT) littermates. B. Skeletal staining additionally shows curvature within newborn C1532Yneo sternum. Figure modified from Human Molecular Genetics, 2007, Vol. 16, No.5: 515–528 with permission from the publisher.

Figure 3. Altered endochondral skeletogenesis in perlecan hypomorphic mice manifests as transient expansion of hypertrophic cartilage, and radially arranged trabecular bone.

Longitudinal tibial sections from newborn wildtype (left) and C1532Yneo (right) mice stained with hematoxlyn and eosin (A,B) for overall tissue architecture or von Kosa/safranin orange (C) to visualize mineralization and proteoglycan content, respectively. A. The hypertrophic zone (outlined in black) in newborn C1532Yneo mice is expanded in comparison to WT. B. Higher magnification of the hypertrophic zones from (A) shows decreased hypertrophic chondrocyte cell number, lack of cellular organization, and an increased cell-to-matrix ratio in C1532Yneo growth plates. C. In the C1532Yneo growth plate, radial trabecular bone (dark brown) orientation and reduced staining for PGs (bright red) is seen, as opposed to the vertical arrangement of trabecular spicules in WT. Figure modified from Human Molecular Genetics, 2007, Vol. 16, No.5: 515–528; figure 6D, E with kind permission of Springer Science and Business Media.

Degenerative joint phenotypes of the C1532Yneo mice further implicate perlecan in both establishment and maintenance of structural niches that provide tissue integrity needed for proper articular cartilage function. While most cartilage of the vertebrate skeleton is eventually replaced by trabecular bone and marrow through EO, articular cartilage remnants remain to line the joint surfaces. Proper remodeling is key to reorganizing the ECM to consist of collagen fibers arranged parallel to the surface of the joint (Gepstein et al., 2002; Archer et al., 2003), imparting strength to withstand compressive forces. C1532Yneo mice have diagnostic signs of shoulder dysplasia reminiscent of osteoarthritis, and hip defects that result in osteonecrosis of the femoral head (Fig. 4 A), mimicking patients with Stuve-Wiedemann syndrome (SWS), formerly categorized as SJS2 (Dagoneau et al., 2004) (Fig. 4 B). SWS has now been linked to mutations in the leukemia inhibitory factor receptor gene (LIFR) (Dagoneau et al., 2004), another modulator of FGF activity. More recent findings reveal that the near-absence of perlecan results in aberrant MMP activity in both skeletal elements and extra-skeletal tissues, such as lung and kidney (Rodgers et al., manuscript in preparation). While extra-skeletal elements of the C1532Yneo mice appear to function normally, these data indicate changes that may lead to weakened structural stability upon challenge. Altered MMP levels and activity indicate disturbed tissue remodeling, a finding that is supported by altered number and activity of osteoclasts in long bones (Rodgers et al., 2007). This aberrant remodeling, likely combined with mechanical stresses due to abnormal bone shape and size, results in early-onset osteoarthritis in all joints examined, including the knee, hip, shoulder, and temporomandibular joints, and implicates the C1532Yneo mice as a model for degenerative joint diseases. In fact, perlecan expression by joint synovial cells has been shown to be upregulated by TGF-β and has thus already been implicated in the pathogenesis of arthritis (Dodge et al., 1995).

Figure 4. Loss of femoral heads in perlecan hypomorphic C1532Yneo mice (A) and Stuve-Wiedemann patients (B).

A. Alizarin red S and alcian blue skeletal staining of week 4 wildtype (WT) and C1532Yneo mice reveals the lack of a femoral neck (arrow) and of the mineralized femoral head in C1532Yneo mice. B. X-ray analysis of a Stuve-Wiedemann Syndrome (SWS) patient at 6.5 years and again six months later reveals the complete radiographic loss of the femoral heads. Figure modified from Eur. J. Pediatr. 156: 220 and Human Molecular Genetics, 2007, Vol. 16, No.5: 515–528 with permission from the publishers and original authors.

HSPGs In The Hematopoietic Stem Cell Niche

Hematopoiesis is the processs by which hematopoietic stem cells (HSCs) generate and replenish progenitors as well as all mature blood cells. During vertebrate ontogeny, hematopoiesis is established sequentially in several different anatomic sites. However, once marrow forms as a result of EO, hematopoiesis occurs almost exclusively within bone (Aguila and Rowe, 2005). Understanding how a marrow environment is established and what makes it preferential toward supporting hematopoiesis, are key to clarifying the nature of the HSC niche. Based on the skeleto-hematopoietic disease phenotype in our collagen X mouse models, which are discussed below, we propose that EO establishes the HSC niche, which resides within the chondro-osseous junction and likely consists of a cytokine and growth factor-rich collagen X-HSPG network (Jacenko et al., 2001; Jacenko et al., 2002) (Sweeney et al, manuscript accepted to this Journal issue).

The idea of a specific HSC niche was first proposed three decades ago by Schofield (Schofield, 1978), and subsequently, niches have been proposed for stem cells of many different origins (Lin, 2002; Fuchs et al., 2004; Ohlstein et al., 2004; Aguila and Rowe, 2005; Taichman, 2005; Adams and Scadden, 2006; Borthakur et al., 2006). An interdependence between EO-derived bone and hematopoiesis has long been apparent, however this view has not included hypertrophic cartilage as a potential contributor in hematopoietic interactions. HSCs have been shown to preferentially localize to the endosteal surfaces of bone where bone matrix, osteoblasts, osteoclasts, and marrow meet (Nilsson et al., 1997; Nilsson et al., 2001). Here, these components, together with stromal cells, vascular cells, and various ECM components, encounter to physically compartmentalize the marrow (Klein, 1995; Taichman and Emerson, 1998; Taichman, 2005).

HSPGs in particular are proposed to orchestrate HSC niches in the marrow by binding cytokines and presenting them to stromal cells and HSCs (Gupta et al., 1998). For example, one study shows that HSPG interaction with growth factors and other ECM proteins enhances progenitor cell localization within the HSC niche (Bruno et al., 1995). Likewise, other data imply that HSPGs are essential for maintenance of long-term culture-initiating cells in ex vivo cultures via interactions with interleukin-3, macrophage inflammatory protein-1alpha, and thrombospondin (Gupta et al., 1996; Gupta et al., 1998), all components critical for hematopoiesis (Long et al., 1992; Verfaillie et al., 1994). Later studies by the same group demonstrate that the source of HSPG is critical for proper hematopoietic function, likely due to the specific type and degree of GAG sulfation (Gupta et al., 1998). Data from our collagen X mouse models implicate an interaction between collagen X and HSPGs to cooperatively enhance the HSC niche environment. Moreover, our findings extend the concept of the HSC niche to the chondro-osseous junction, which comprises the cartilaginous growth plate, trabecular bone with its hypertrophic cartilage core, osteoclasts that continually remodel this matrix, and the marrow and stromal constituents.

Mice in which the function of collagen X is disrupted either through transgenesis leading to dominant interference (transgenic “Tg” mice) (Jacenko et al., 1993; Jacenko et al., 2001), or through gene inactivation (knock-out “KO”) (Rosati et al., 1994; Kwan et al., 1997), develop a skeleto-hematopoietic disease phenotype that varies in severity (Chan and Jacenko, 1998; Gress and Jacenko, 2000; Jacenko et al., 2002). At week 3 after birth, ~25% of Tg and ~10% of the KO mice begin wasting, become lethargic, and die. Survivors show transient dwarfism and normal life spans, but are prone to skeletal deformities, skin ulcers, and lymphosarcomas. In both Tg and KO collagen X mice, disease manifestation involves all EO-derived tissues and marrow, and includes growth plate compressions and diminished chondrocyte hypertrophy in Tg mice (Jacenko et al., 2002), and slightly thinned growth plates with an altered proliferative cartilage zone in KO mice (Gress and Jacenko, 2000). In both murine subsets, trabecular bony spicules are reduced in number and length (Fig. 5 B & F). Hematopoietic changes in all mice manifest as marrow hypoplasia and impaired hematopoiesis, with the most severe skeletal defects, including marrow aplasia (Fig. 5 D), lymphopenia (Fig. 6), and lymphatic organ atrophy, present in the subset of Tg and KO mice displaying early lethality. Surviving mice exhibit altered lymphopoiesis throughout life along with additional subtle changes in hematopoietic lineage profiles in the marrow and lymphatic organs (Fig. 6). Moreover, every mouse with altered collagen X function has an impaired immune response, as observed through both in vitro T cell function assays and in vivo parasite challenge studies (Sweeney et al., manuscript accepted to this Journal issue). These disease manifestations are a direct result of disrupted collagen X function in hypertrophic cartilage undergoing EO, as all other possibilities have been excluded (Jacenko et al., 1993; Campbell et al., 2004)(Sweeney et al., manuscript accepted to this Journal issue).

Figure 5. A skeleto-hematopoeitic disease phenotype manifests in collagen X transgenic mice as growth plate compressions, trabecular bone reductions, and marrow aplasia in EO-derived tissues.

Tibial longitudinal sections from control (A,C,E) and collagen X Tg perinatal-lethal mutant (B,D,F) mice at week 3 after birth. A & B. Alcian blue/H&E staining reveals in B a compressed growth plate, diminished trabecular bone, and a predominance of erythrocytes in marrow C & D. Giemsa staining of bone marrow within parallel sections shown in A & B highlights marrow aplasia in D, manifested as a depletion of leukocytes (stained blue). E & F. Higher magnification view of an H&E-stained chondro-osseous junction within parallel sections shown in A & B. Note in E an intricate network of trabecular bony spicules, comprised of a hypertrophic cartilage core (light purple) with osteoid (pinker matrix) deposited on the surface, surrounded by a hematopoietic marrow. In F, note reduction of trabecular spicules, and a predominance of erythrocytes and depletion of leukocytes in the marrow.

Figure 6. Hematopoietic defects manifest in collagen X mice as altered lymphocyte profiles in marrows and lymphatic organs.

Flow cytometry of marrow aspirates and splenic lymphocytes from week 3 wild type (WT), and collagen X Tg and KO mice with either mild (Tg, KO), or perinatal-lethal (Tg mut; KO mut) disease phenotypes. Notable B cell lineage declines are seen in Tg, Tg mut, and KO mut mice, whereas a transient increase in B cells is observed in KO mice.

To understand how skeletal changes involving hypertrophic cartilage could translate to an alteration of the marrow environment, ultrastructural studies were used to identify the primary morphologic defect in the pericellular matrix of hypertrophic cartilage (Jacenko et al., 2001). Electron microscopy revealed a lattice-like network likely composed of collagen X around hypertrophic chondorcytes in control growth plates (Fig. 7 A & C). In contrast, this pericellular lattice was lacking in Tg mice, while ruthenium hexamine trichloride (RHT)-positive aggregates (RHT precipitates PGs and GAGs) were observed along hypotrophic chondrocyte surfaces (Fig. 6 B & D) as well as throughout other growth plate zones. Comparable observations were made by Cheah and co-workers in the KO mice (Kwan et al., 1997; Chan and Jacenko, 1998). These data indicate that partitioning of the GAGs and PGs may be dislodged due to a disrupted collagen X-containing pericellular network. Indeed, the growth plates of collagen X Tg and KO mice show an altered distribution of HSPG (Jacenko et al., 2001) with strong immunoreactivity for HSPG in the hypertrophic cartilage growth plate zone of control mice (Fig. 8 A & D), but faint or undetected staining in the Tg and KO mice (Fig. 8 B, C, E, & F).

Figure 7. Ultrastructure of hypertrophic cartilage of week 3 tibial growth plates reveals a pericellular matrix defect in collagen X trangenic mice.

In controls, hypertrophic chondrocytes were surrounded by a gray zone corresponding to the pericellular matrix, which consisted of a fine meshwork (A; arrows in boxed-in region); this matrix was reduced or lacking in the Tg mice (B; arrow). Higher magnification revealed a lattice-like array in the pericellular matrix of hypertrophic chondrocytes from controls (C). In Tg mice, no ordered networks were evident; instead, RHT-positive aggregates accumulated near cell surfaces (D). A,B: bar=2 µm; C,D: bar=200 nm. Reprinted from Am J Pathol. 2001, 159: 2257–2269 with permission from the American Society for Investigative Pathology.

Figure 8. Aberrant immunohistochemical localization of HSPG in growth plates of collagen X Tg and KO mice.

Longitudinal sections of week 3 tibiae from wild type controls (A,D), Tg mice (B), KO mice (E), and Tg (C) or KO (F) mice exhibiting perinatal lethality. Note HSPG localization to proliferative and hypertrophic cartilage, and to trabecular bone in A. In B, staining is pericellular in proliferative chondrocytes, and intensity is reduced in hypertrophic cartilage. In C, E, and F, staining is either faint or absent in the growth plate. D: control where heparitinase was omitted. Brackets approximate width of hypertrophic cartilage. bar, 100 µm. Reprinted from Am J Pathol. 2001, 159: 2257–2269 with permission from the American Society for Investigative Pathology.

Combined, these data strongly support an organizational role for collagen X in hypertrophic cartilage by binding HSPGs and thus sequestering hematopoietic cytokines to the chondro-osseous junction. It is noteworthy that immunohistochemistry of collagen X Tg and KO growth plates revealed an altered perlecan distribution pattern similar to that of HSPG staining in Fig. 8. Moreover, analyses of the perlecan hypomorhic mice reveal an altered B lymphocyte profile (unpublished observations) comparable to that reported for the surviving collagen X Tg mice in Fig. 6. These, as well as other preliminary studies from our group, implicate perlecan as a potential binding partner for collagen X and strengthen our hypothesis that in the chondro-osseus junction, collagen X forms a network that is stabilized by HSPGs such as perlecan.

Taken together, these data generate a provocative possibility that links the disruption of a collagen X-containing matrix at the hypertrophic cartilage/marrow interface, to an altered HSPG distribution and a potential locus for hematopoietic failure. Perlecan has been proposed to influence the fate of bone marrow stromal cells (Bi et al., 2005), and is the major ECM PG localized in the marrow (Grassel et al., 1995; Drzeniek et al., 1997; Schofield et al., 1999). Moreover, perlecan is thought to be involved in compartmentalization of the bone marrow microenvironment, partially due to its ability to bind granulocyte/macrophage-colony stimulating factor and present it to hematopoietic progenitor cells (Klein et al., 1995). Induction of this growth factor by progenitor cells has been shown to result in neutrophil-specific gene induction (Berliner et al., 1995), emphasizing the importance of these specific interactions. Members of the glypican and syndecan families are additional HSPG candidates whose localization might be altered in the collagen X mutant mice. Glypican-4 is found in HSCs (Siebertz et al., 1999) and loss of function mutations in glypican-3 result in perturbed hematopoiesis (Viviano et al., 2005). Furthermore, syndecan is a likely mediator of thrombospondin activity (Chen et al., 1997) during development of the HSC niche (Corless et al., 1992), with syndecan-1 and perlecan additionally involved in the mesenchymal stem cell niche (Chen et al., 2007b). Many of these HSPG-modulated activities have recently been reviewed (Nurcombe and Cool, 2007).

We propose that the collagen X/HSPG network sequesters hematopoietic cytokines to the chondro-osseous junction. During EO, continual remodeling of hypertrophic cartilage/trabecular bone to mature secondary bone by osteoclasts may result in liberation of cytokines to the HSCs. This scenario is strongly supported by recent data from Emerson and colleagues, who have demonstrated in vitro and in vivo that osteoblasts are essential for B-cell commitment and maturation in mice (Taichman and Emerson, 1998; Taichman, 2005; Zhu et al., 2007). Likewise, Scadden and colleagues (Adams and Scadden, 2006) have shown that HSCs preferentially home to and are retained at calcium-enriched regions where trabecular bone is being remodeled via a calcium-sensing receptor (CaR). While further data should provide credence to this HSC niche model, these observations of the hematopoietic defects in collagen X mutant and perlecan hypomorphic mice strongly support the role of HSPGs in the orchestration of an HSC niche in the marrow by stabilizing the matrix and binding growth factors and cytokines.

Perspectives

This review provides an overview of different classes of PGs and details the current body of knowledge concerning the localization and roles of HSPGs during EO and hematopoiesis, emphasizing a link between skeletogenesis and blood cell development. Specifically, vertebrate limb bud patterning, condensation, chondrogenesis, and skeletal growth are outlined to highlight the roles of HSPGs during each of these processes. Proposed functions of HSPGs in hematopoiesis are additionally reviewed here. Based on the available literature and our own data, we propose a mechanistic link between endochondral skeletogenesis and establishment of the HSC niche, and implicate HSPGs as predominant regulators of these developmental processes. In particular, HSPGs have distinct properties to regulate developmental niches during development by sequestering morphogens and growth factors, establishing gradients of these signaling molecules, and performing as co-receptors to facilitate cell signaling. In addition, HSPGs provide structural stability to tissues and function as filters and barriers.

While many roles of HSPGs in skeletal patterning and development are inferred from studies involving either in vitro systems or non-mammalian models, availability of mouse models with specific in vivo disruptions of HSPG function would provide direct confirmation of these functions. For example, much of the knowledge concerning HSPGs in patterning and condensation comes from fly wing development. A significant amount of additional information can be gleaned from chicken wing bud development. While there is validity in finding parallels and drawing analogies between these different systems, more solid data concerning HSPGs in mammalian limb development would be a welcome opening to this potentially exciting field. Specifically, it would be helpful to ferret out the details of morphogen sequestration and gradient establishment during limb bud patterning by identifying the PGs involved at each stage, the chemistry of their GAG side chains, and investigation of redundant properties. With this knowledge, it would be possible to specifically define different developmental niches.

An important area of current and emerging focus that would aid these needed developmental studies concerns defining the fine structures of HS isolated from specific tissues at particular developmental stages. For example, HSPG structural heterogeneity is not only provided by the core PG to which HS GAG chains are attached but is also spatially and temporally regulated. The degree of GAG modification, such as sulfation, is also critical in defining the function of PGs and provides further distinction between the classes. In fact, several human skeletal diseases are caused by sulfate transporter gene defects, including diastrophic dysplasia, atelosteogenesis type 2, achondrogenesis type 1B, and McAlister dysplasia (Rossi et al., 1997; Rossi and Superti-Furga, 2001). A recent mouse model for diastrophic dysplasia will certainly provide insight to the critical role of HSPG sulfation (Forlino et al., 2005). Likewise, mouse models for the human skeletal condition Hereditary Multiple Exostoses (HME) syndrome are helping to uncover the role of glycosyltransferase activity is the HS biosynthesis pathway (Lin et al., 2000; Morimoto et al., 2002). Differences in HS GAG biosynthesis and modification may define why one PG over another is used for functions such as binding cytokines within the collagen X scaffold of hypertrophic cartilage to form an HSC niche within the chondro-osseous junction. However, more work is needed to define these differences and to uncover their functions, such as identifying which hormones and cytokines bind to HSPGs of different chemistries, and how these interactions are spatially and temporally regulated.

Thus, As Betty Hay was so well known to advocate, the ECM plays an active role in embryogenesis and tissue homeostasis. Our review builds on this theme, focusing on how the HSPGs, being prominent and diverse components of cell surfaces and the ECM, help establish the location, maintenance, and function of developmental niches throughout skeletal development and homeostasis. As the genetics behind more human diseases are discovered, new animal models are developed, and the biochemical nature of HSPGs are better defined, further valuable insights into the role of HSPGs as critical regulators of skeleto-hematopoietic development will be gained, making this an ever-more exciting field of study.