Introduction

Matricellular proteins are defined as extracellular matrix (ECM)-associated proteins that do not contribute structurally to the ECM, such as the classical ECM proteins, collagens and laminins, but instead modulate cell interaction with the ECM (Bornstein and Sage 2002). SPARC (secreted protein, acidic and rich in cysteine; also known as osteonectin), is a prototypic matricellular protein composed of three modular domains (Brekken and Sage 2001). That SPARC is conserved in a wide variety of evolutionarily diverse organisms (e.g., Caenorhabditis elegans, Drosophila melanogaster, brine shrimp, zebra fish, chicken, mice, and humans), suggests a basic function of this matricellular protein in multicellular biology (Bassuk et al. 1993; Schwarzbauer and Spencer 1993; Bradshaw and Sage 2001; Tanaka et al. 2001; Martinek et al. 2002; Rotllant et al. 2008).

Although the expression of SPARC is associated with many different types of tumors, the function of SPARC in tumorigenesis and metastasis is not clearly defined. A number of excellent reviews summarizing SPARC expression in specific types of tumors and its association with either enhanced or diminished tumor progression are available (Framson and Sage 2004; Clark and Sage 2008; Podhajcer et al. 2008). SPARC activity has proven to be contextual, and thus, seemingly contradictory functions of SPARC in either promoting or inhibiting different types of cancer have emerged. In general, SPARC expression correlates with invasion and progression of gliomas and melanomas (Rempel et al. 1999; Schultz et al. 2002; Prada et al. 2007; Shi et al. 2007; Haber et al. 2008; Yunker et al. 2008). In contrast, many epithelial cancers (e.g., lung, colon, prostate, pancreatic, and endometrial) hypermethylate the SPARC promoter, thus, reducing the amount of SPARC produced by the tumor cells (Sato et al. 2003; Suzuki et al. 2005; Wang et al. 2005b; Sova et al. 2006; Rodriguez-Jimenez et al. 2007; Yang et al. 2007). Targeted promoter demethylation by the nonsteroidal anti-inflammatory drug NS398 in human lung cancer cells restored SPARC expression and reversed the inhibition of cell invasion mediated by SPARC (Pan et al. 2008). However, given the wide-range of activities attributed to SPARC, global targeting of SPARC function has the potential to introduce detrimental off-target effects. A clear description of the molecular mechanisms of SPARC action is needed to understand its divergent effects on human cancers and thus develop effective strategies to manipulate SPARC activity that might be useful in the treatment of cancer growth and metastasis.

This chapter focuses on activities associated with SPARC and proposed cellular mechanisms by which SPARC mediates these activities with primary focus on cell-ECM interaction. We seek to provide some insight into the disparate influences of SPARC and the potential of this matricellular protein to guide either tumor and/or stromal cell interaction with ECM and thereby impact tumor progression and dissemination.

SPARC Structure/Function

The SPARC gene encodes a protein with a predicted molecular weight of 32 kD (Mason et al. 1986). Tissue-specific glycosylation of mammalian SPARC decreases the mobility of this glycoprotein which frequently migrates at ∼40–43 kD under reducing SDS-PAGE conditions (Hughes et al. 1987; Kaufmann et al. 2004). The N-terminal region of SPARC contains a low-affinity, high-capacity Ca2+-binding domain (Maurer et al. 1992; Brekken and Sage 2001). The central portion of the protein includes a region with homology to follistatin that includes a Cu2+-binding site, whereas the C-terminal domain (E-C domain) contains two high-affinity Ca2+-binding EF hands (Hohenester et al. 1996, 1997; Sasaki et al. 1998). The E-C domain of SPARC contains the cell-binding domain as well as the collagen-binding region and is the most conserved among SPARC homologs expressed in C. elegans, Drosophila, and mammals (Sasaki et al. 1998). The capacity of SPARC to bind fibrillar collagens such as types I, III, and V in addition to type IV, is dependent upon the triple helical conformation of collagen and suggests that SPARC might influence ECM composition in both connective tissue (rich in fibrillar collagens I, III, and V) and basal membranes (where collagen IV is a prominent component) (Mayer et al. 1991; Bradshaw and Sage 2001).

The affinity of SPARC binding to collagen is in the 10−7 M range but can differ according to the cellular source of SPARC due to differential post-translational modification (Sasaki et al. 1997). The collagen binding sites of SPARC to collagen types I, II, and III have been mapped by rotary shadowing and to collagen I by atomic force microscopy (Wang et al. 2005a; Giudici et al. 2008). Rotary shadowing, used by Giudici et al. (2008), mapped the major SPARC binding site on procollagen I to a location approximately 180 nm from the C-terminus of types I, II, and III. A lesser site located near the mammalian collagenase cleavage site in types I and II was mapped to a region 60–100 nm from the C-terminus. In contrast, the prominent site mapped by Wang et al. (Wang et al. 2005a) using atomic force microscopy was 87.5–125 nm from the C-terminus of procollagen I with a lesser site located 237.5–262.5 nm from the C-terminus (Giudici et al. 2008). The discrepancy in the two studies might arise from the sources of recombinant SPARC (rSPARC) used to perform the binding assays. rSPARC produced by mammalian cells was used to perform rotary shadowing analysis whereas rSPARC produced in insect cells was used in the atomic force microscopy studies. Differential glycosylation performed by insect versus mammalian cells is one possible explanation for differences in SPARC binding to collagen in the two studies. Similarly, the source of procollagen I differed as Wang et al. used a recombinant homotrimer of procollagen I (3 subunits of collagen α1(I)) whereas Guidici et al. performed rotary shadowing with heterotrimeric procollagen I [2 collagen α(1)I and 1 collagen α(2)I] produced by human dermal fibroblasts.

Guidici et al. went on to show that a synthetic triple-helical peptide from collagen III, GPOGPSGPRGQOGVMGFOGPKGNDGAO (O, 4-hydroxyproline) bound to SPARC with an affinity comparable to that of recombinant procollagen III (Giudici et al. 2008). Interestingly, the region of collagen shown to be bound by SPARC using rotary shadowing overlapped previously mapped binding domains of the collagen receptor DDR2, in addition to von Willebrand Factor (Agarwal et al. 2002; Konitsiotis et al. 2008). Alternately, the sites mapped by atomic force microscopy overlap a subset of those proposed to bind the collagen specific integrins α1β1, α2β1, and α11β1 (Xu et al. 2000; Zhang et al. 2003). Hence, SPARC bound to collagens might limit cell surface receptor interaction, either mediated by DDR2 and/or integrins, with fibrillar collagens in the pericellular milieu.

Digestion of SPARC by several matrix metalloproteinases (MMP) increased the affinity of SPARC for collagen types (Sasaki et al. 1997). Removal of helix αC in the E-C domain increased SPARC affinity for collagens 7–20 fold (Sasaki et al. 1998). In addition, differential glycosylation of SPARC affected collagen-binding affinity. SPARC appears to have a single N-glycosylation site at Asn99 that is highly conserved (Brekken and Sage 2001). SPARC purified from platelets migrated more slowly in SDS-PAGE analysis than SPARC from bone due to increased complexity of the platelet oligosaccharide modification versus that of bone-derived SPARC (Hughes et al. 1987; Kelm and Mann 1991). The difference in glycosylation influenced SPARC binding to collagen as bone SPARC bound with higher affinity to collagen I, III, and V than did SPARC from platelets. Removal of oligosaccharides from both forms of SPARC increased collagen binding to collagen V and abrogated the differences in collagen binding (Kelm and Mann 1991). Therefore, the binding of SPARC to collagen in the extracellular environment is subject to modulation by cell type-specific post-translational modification and by MMP activity.

The expression of SPARC is robust during development and differentiation of most mammalian tissues (Bradshaw and Sage 2001). However, the expression of SPARC declines in most organs as organisms mature. Bones, gut epithelia, and other tissues with high ECM turnover, retain SPARC expression into adulthood (Bradshaw and Sage 2001). Increased SPARC production is associated with adult tissue remodeling events such as wound healing and those that involve fibrotic deposition of collagen such as in liver cirrhosis and in individuals with scleroderma (Reed et al. 1993; Frizell et al. 1995; Zhou et al. 2003). As stated previously, SPARC expression in transformed cells or in stromal cells adjacent to tumors is dependent upon the type of tumor and its tissue of origin (Framson and Sage 2004).

The capacity of SPARC to bind to a number of different types of collagen, including fibrillar collagens I, III, V, and to basal membrane collagen IV, and its high expression in tissues undergoing active remodeling implicates SPARC in the process of ECM assembly and turnover.

ECM Assembly

Evidence that SPARC contributes to the formation and/or stability of a functional basal lamina is suggested from studies in C. elegans and Drosophila in which disruption of SPARC expression gave rise to lethal mutations (Fitzgerald and Schwarzbauer 1998; Martinek et al. 2002, 2008). Whereas fibrillar collagens homologous to mammalian collagen types I-III are not expressed in worms and flies, collagen IV, a primary constituent of basal membranes, is present throughout each organism. In C. elegans, exogenously expressed SPARC-GFP was localized to basal lamina in several tissues (Fitzgerald and Schwarzbauer 1998). In Drosophila, localization of SPARC protein closely followed that of collagen IV (Martinek et al. 2002). In flies with mutated expression of collagen IV, SPARC protein was significantly decreased in basal laminae of certain internal organs. Abrogation of SPARC expression in Drosophila gave rise to disrupted basal laminae similar to that exhibited by collagen IV mutants (Martinek et al. 2008). The function of SPARC in invertebrates appears to be essential for patent basal laminae formation and is implicated in either production and/or assembly of collagen IV into extracellular structures.

Although abrogation of SPARC expression in mice does not give rise to embryonic lethality, SPARC-null mice display a range of phenotypes, the basis of which also appear to reside in disruption of ECM organization (Bradshaw and Sage 2001). The existence of a family of SPARC-related proteins, including hevin, SMOC-1 and 2, and testican, might provide some compensation for the absence of SPARC in mice (Soderling et al. 1997; Roll et al. 2006; Liu et al. 2008). One of the first phenotypes described in SPARC-null mice was premature cataractogenesis. In two independently generated SPARC-null mice, cataract formation in mice of 4 weeks and younger was observed (Gilmour et al. 1998; Norose et al. 1998). Yan et al., went on to show that the basement membrane surrounding the lens epithelial cells exhibited disorganized collagen IV and laminin compared with those of WT mice (Yan et al. 2002). Whereas the plasma membrane of wild type lens epithelial cells formed a sharp demarcation between cells and ECM, the plasma membrane produced by SPARC-null lens epithelial cells was invaginated with integrin β1-positive protrusions extending into the disorganized basement membrane. In the absence of SPARC, lens epithelial cells were not able to deposit and correctly assemble a patent basement membrane so that fluid balance across this ECM was not maintained (Yan et al. 2003). Yan et al. hypothesized that the increased porosity of the SPARC-null ECM, demonstrated by toluidene blue penetrance, gave rise to cataract formation in the SPARC-null mice. SPARC-null lens epithelial cells also demonstrated changes in adhesion and integrin expression versus wild type cells (Weaver et al. 2006).

In addition to aberrations in lens basement membrane, SPARC-null mice also exhibited deficiencies in connective tissue. Reduced levels of collagen I were reported in skin, adipose, heart, and bones of SPARC-null mice (Bradshaw et al. 2003a, b; Delany et al. 2003). The collagen fibrils in the skin of the null mice displayed significant decreases in diameter and a uniformity of size in comparison to those of wild type mice. The decrease in collagen content has been linked to reduced tensile strength of the skin and to accelerated closure of full-thickness wounds (Bradshaw et al. 2003b). Improved healing of dermal wounds was attributed to an increase in skin contractility brought about by decreases in the collagenous ECM generated in the absence of SPARC (Bradshaw et al. 2002). As collagen gels of lesser collagen concentration were contracted by fibroblasts more quickly than those of higher collagen concentration, an extrapolation was made that the dermis of SPARC-null mice, with lesser amounts of collagen than wild type, was more susceptible to contraction by cells in the wound.

Foreign materials, when implanted into mice, are encapsulated by resident cells to “wall off” the exogenous material. A similar event has been observed in some solid tumors. Formation of a collagen capsule is accompanied by increased expression of SPARC. In SPARC-null mice, a decrease in the dimensions of the collagen capsule synthesized in response to such an implanted foreign material was evident versus that formed in WT mice (Puolakkainen et al. 2003). The collagen surrounding the foreign material in SPARC-null animals exhibited more immature fibers that were smaller and more uniform in diameter than those that were seen in samples from WT mice. Collagen fibrils formed by adult dermal fibroblasts in response to an implant retained the phenotypic changes noted in collagen fibrils formed during development in SPARC-null skin (Puolakkainen et al. 2003). Hence, SPARC most likely serves a basic function in the regulation of collagen fibril assembly at least by dermal fibroblasts.

A decrease in collagen deposition was also reported in bleomycin-induced injury in the lungs of SPARC-null mice and, in an animal model of diabetic nephropathy, diminished fibrosis in the kidney of SPARC-null mice treated with streptozocin was observed (Strandjord et al. 1999; Taneda et al. 2003). An increase in collagen production and SPARC expression was associated with both bleomycin-induced injury and in diabetic nephropathy; hence, the absence of SPARC was shown to have significant effects on collagen deposition in response to injury in adult lungs and kidney (Pichler et al. 1996).

SPARC production in some types of tumors was associated with changes in collagen I deposition as well. Lewis lung carcinoma cells injected subcutaneously into SPARC-null mice formed substantially larger tumors in comparison to wild type mice, with a reduction in the collagenous capsule surrounding the tumors in the SPARC-null mice (Brekken et al. 2003). Although the carcinoma cells expressed SPARC in vitro and in vivo, the host response to tumor progression was clearly influenced by the absence of SPARC expression by stromal cells. In addition, a decrease in the infiltration of macrophages was observed in tumors from SPARC-null mice in comparison to wild type mice (Brekken et al. 2003).

We observed that murine pancreatic cancer cells (Pan02, aka Panc02) injected into SPARC-null mice formed larger tumors versus those injected into wild type animals (Puolakkainen et al. 2004). The tumors from SPARC-null mice had decreases in associated ECM (Fig. 8.1) and decreases in macrophage recruitment and invasion, results similar to those found upon injection of Lewis Lung carcinoma cells into SPARC-null and wild type mice (Brekken et al. 2003). Furthermore, when Pan02 cells were implanted orthotopically into the pancreas of SPARC-null and wild type animals, the number of metastatic events was also increased in SPARC-null mice (Arnold et al. 2008). Interestingly, when Pan02 cells engineered to over-express matrix metalloproteinase (MMP)-9 were injected, growth of tumors continued to be enhanced in SPARC-null animals; however, the metastatic burden was decreased in both SPARC-null and wild type mice. Microvessel density was diminished in tumors formed in SPARC-null versus WT mice whereas forced expression of MMP-9 by tumor cells reversed the angiogenic decrease in SPARC-null mice (Arnold et al. 2008). These results suggest a complex interaction between MMP-9 and SPARC, which is a substrate for MMP cleavage. This interaction between a protease prominent in the tumor microenvironment (MMP-9) and an extracellular adaptor protein (SPARC) impacts many of the hallmarks of cancer including invasion and angiogenesis.

Fig. 8.1
figure 1_8

Decreased ECM Deposition in Pan02 tumors grown in SPARC-null animals. Upper panels: Pan02 tumors grown in pancreas of WT (+/+) or SPARC-null (−/−) mice were harvested, snap frozen, and analyzed for collagen IV by immunohistochemistry. Lower panels: TEM analysis of orthotopic Pan02 tumors in WT (+/+) and SPARC-null (−/−) animals. Endothelial cells (*), red blood cells (RBC), and lumens of blood vessels are labeled. Note the reduction of ECM and collagen deposition under endothelium in tumors from SPARC −/− mice (white arrows) but prominent deposition of collagen fibrils in WT mice (black arrows). Total magnification and scale bar (1 μm) are shown

Full size image

Glioblastomas are heterogeneous tumors that can exhibit diverse cellular regions including those involved in proliferation, angiogenesis, apoptosis, and invasion. SPARC is expressed highly in gliomas and promotes invasion of this cell type while inhibiting overall tumor growth. Yunker et al. demonstrated that over-expression of SPARC in transplanted glioma cells gave rise to increased collagen deposition associated with the SPARC-expressing tumors versus control (Yunker et al. 2008). The author postulated that SPARC expression reduced tumor growth through an increase in ECM deposition as well as a reduction in VEGF-induced angiogenesis. Here, increased SPARC expression by tumor cells was sufficient to drive increased collagen production and incorporation into the ECM and clearly influenced tumor progression.

Hence, there is convincing evidence from a variety of studies that SPARC is a critical factor in the synthesis, deposition, and/or stabilization of collagen I in the ECM of connective tissue and might contribute to inhibition of some tumors through increased production and deposition of a stromal or tumor ECM rich in collagen I.

Cellular Mechanisms of SPARC in ECM Assembly and Cell Signaling

The molecular basis of SPARC to influence collagen I deposition and fibrillogenesis has been investigated in primary fibroblasts from SPARC-null mice. In Rentz et al., SPARC-null dermal fibroblasts were shown to exhibit increased cell-associated procollagen I in comparison to wild type cells (Rentz et al. 2007). Procollagen I is processed to mature collagen I by removal of the C and N-terminal propeptides. Typically, procollagen I associated with fibroblast cell layers exists in four forms: procollagen I with N and C terminal propeptides, pC collagen I (C-propeptide attached, N-propeptide removed), pN collagen I (N-propeptide attached, C-propeptide removed), and collagen I (both propeptides removed). In the absence of SPARC, an increase in the proportion of total collagen I in the cell layer was present as enzymatically cleaved collagen I, with propeptides removed. As the propeptides of procollagen I are generally considered to be inhibitory to collagen fibril incorporation, these results suggested that SPARC influenced procollagen processing and enhanced production of fibril-forming collagen I. However, SPARC-null fibroblasts were inefficient in the incorporation of collagen I into a detergent-insoluble ECM. We proposed that SPARC bound to collagen I might diminish cell surface receptor interaction and promote assembly of collagen I into insoluble ECM. Without SPARC, increased collagen engagement by receptors might enhance collagen turnover either through phagocytosis and/or pericellular degradation pathways (McCulloch 2004; Lee et al. 2006).

One class of collagen receptors is the integrin family of ECM receptors (White et al. 2004). Interestingly, a function of SPARC in the regulation of integrin linked kinase (ILK), a down-stream component of the integrin signal transduction pathway, is emerging. Primary lung fibroblasts from SPARC-null mice exhibit reduced fibronectin-induced ILK activation. Associated with the decrease in ILK activity in cells lacking SPARC, a diminished capacity to generate stress-fibers on fibronectin - a critical step in fibronectin assembly - was observed (Barker et al. 2005). Expression of exogenous SPARC in SPARC-null cells restored the capacity of the lung fibroblasts to form fibronectin-induced stress fibers and fibronectin-dependent activation of ILK. As fibronectin is required for collagen I ECM assembly in vitro, inefficient fibronectin assembly in vivo, predicted from the absence of SPARC, might impair collagen I ECM deposition (Velling et al. 2002).

The capacity of SPARC to enhance ILK activation in lens epithelial cells promoted cell survival in vitro. Weaver et al. further showed that SPARC bound to ILK through an integrin β1 complex (Weaver et al. 2008). The copper-binding region of SPARC located in the modular domain with follistatin homology was implicated in the interaction of SPARC with β1 integrin and ILK.

In glioma cells, two studies examining SPARC and ILK activity have been reported. In one case, inhibition of SPARC expression by short interfering RNA (siRNA) reduced ILK activity coincident with reduced Akt and Focal Adhesion Kinase (FAK) activation (Shi et al. 2007). Golembieski et al. reported that in glioma cells expressing SPARC tagged with green fluorescent protein (GFP), FAK and Akt activity were not changed in response to fibronectin versus control cells whereas total levels of ILK were increased in SPARC-expressing cells (Golembieski et al. 2008). In the latter study, heat shock protein (HSP) 27 was shown to be a major downstream effector of SPARC activity. HSP27 is a protein implicated in actin polymerization, cell contraction, and migration and, as such, is postulated to have potent effects on cell behavior.

In addition to the function of SPARC in the regulation of ECM-cell interaction, an active role of SPARC in the regulation of collagen fibrillogenesis is suggested from in vitro and in vivo results. In vitro, SPARC was shown to inhibit collagen fibrillogenesis using recombinant SPARC and purified collagen (Giudici et al. 2008). Addition of SPARC following initiation of fibrillogenesis had little effect on collagen fibril formation whereas near complete inhibition of fibril assembly occurred when SPARC was added during the fibril nucleation phase. Collagen fibrils in SPARC-null dermis, as mentioned above, displayed a distinct morphology versus those of WT and suggested that SPARC participates in collagen fibril assembly in vivo (Bradshaw et al. 2003b).

As extracellular SPARC is difficult to detect in tissues, presumably SPARC is not retained in collagen fibrils incorporated into insoluble ECM. The spatial and temporal regulation of procollagen processing is believed to be essential for collagen deposition. One possibility is that SPARC bound to procollagen I serves to reduce collagen binding to receptors in the pericellular environment following procollagen secretion while inhibiting fibril nucleation events. A number of collagen-binding proteoglycans including decorin, lumican, fibromodulin, and dermatopontin, influence collagen fibril diameter as well as incorporation of collagens type III and V into collagen I fibrils (Danielson et al. 1997; Ezura et al. 2000; Takeda et al. 2002). SPARC might promote appropriate processing of propeptides and perhaps association with other proteins incorporated into fibrils so that assembly of collagen I into tissue-specific ECMs is accomplished. In such a scenario, SPARC could be considered a type of extracellular chaperone for collagen that is released following initiation of fibrillogenesis.

Alternatively, a function for SPARC as an intracellular chaperone for collagen, similar to perhaps HSP47, has been proposed (Tasab et al. 2000; Martinek et al. 2007). Whereas HSP47 is required in the endoplasmic reticulum to assemble procollagen molecules, Martinek et al. put forth evidence supporting a function of SPARC to facilitate post-endoplasmic reticulum events in procollagen maturation that influence collagen fibrillogenesis (Martinek et al. 2008). Along these lines, SPARC exhibited classic chaperone activity in thermal aggregation assays carried out in vitro (Emerson et al. 2006). A distinct possibility is that SPARC has both intra and extracellular activities that influence collagen ECM assembly.

With regard to cell-ECM interaction, it is noteworthy that SPARC was shown to be a substrate of transglutaminase (Hohenadl et al. 1995). Transglutaminase cross-links ECM components and has been implicated in fibronectin assembly via interaction with integrin receptors at the plasma membrane (Telci et al. 2008). The expression of SPARC is therefore predicted to influence transglutaminase-dependent events in the pericellular environment.

Cell Motility

Unlike many other matricellular proteins, SPARC does not contain a classical cell attachment, integrin-binding, RGD sequence. In fact, purified SPARC protein induces rounding when added to a number of different types of cell cultures, most notably endothelial cells (Lane and Sage 1990). Consequently, SPARC has been designated a counter-adhesive protein. Expression of SPARC has been suggested therefore to enhance migration of certain cell types that must disengage from existing ECM ties to initiate movement. In prostrate cancer cells, SPARC supported migration of metastatic cells to bone (De et al. 2003). The increase in migration generated in response to SPARC was dependent upon activation of αvβ3 and αvβ5, RGD-binding integrins, and SPARC therefore did not mediate the migration directly. Furthermore, the SPARC-induced migration was supported by an autocrine vascular endothelial growth factor (VEGF)/VEGF receptor 2 pathway on the prostrate cancer cells.

HSP27, as described above, has been implicated in SPARC-mediated effects in glioma. As HSP27 mediates various cellular activities shown to be affected by SPARC expression such as motility and stress fiber formation, HSP27 as a down-stream target of SPARC activity is plausible. SPARC expression by melanoma cells was associated with aggressive invasion whereas inhibition of SPARC expression diminished tumorgenicity of melanoma cells. Proteomic analysis of proteins affected by inhibition of SPARC expression in melanoma cells revealed an increase in HSP27 in cells with diminished SPARC expression (Sosa et al. 2007). N-Cadherin, a cell adhesion molecule, and clusterin, in contrast, were decreased in response to decreased SPARC expression in the melanoma cells.

Robert et al. reported that over-expression of SPARC in normal melanocytes resulted in a phenotypic shift to a fibroblast-like morphology (Robert et al. 2006). A decrease in two cell adhesion molecules, E-cadherin and P-cadherin expression, was associated with the mesenchymal transition induced by over-expression of SPARC. As loss of E-cadherin contributes to melanoma cell growth and invasion, SPARC might represent an important regulator of E-cadherin expression in melanoma cells.

Growth Factors and Cytokines

The capacity of SPARC to modulate the activity of several growth factors including basic fibroblast growth factor (bFGF), VEGF, platelet derived growth factor (PDGF), and transforming growth factor (TGF)-β1, has been established (Hasselaar and Sage 1992; Raines et al. 1992; Kupprion et al. 1998; Francki et al. 2004). In the case of bFGF, PDGF, and VEGF, SPARC inhibited the action of these growth factors. SPARC was shown to bind directly to PDGF to diminish PDGF receptor activation whereas in the case of bFGF, SPARC was inhibitory for bFGF-induced migration of endothelial cells but did not bind directly to bFGF (Hasselaar and Sage 1992; Raines et al. 1992).

Similar to PDGF, SPARC was shown to bind directly to VEGF and diminished VEGF interaction with receptors on microvascular endothelial cells. As SPARC had previously been shown to decrease proliferation of endothelial cells in response to mitogenic stimuli, the function of SPARC as a negative regulator of angiogenesis was proposed (Funk and Sage 1991). In an in vivo model of angiogenesis, sponges implanted into the sub-dermal space of SPARC-null mice demonstrated an increased fibrovascular invasion in comparison to that of WT mice (Bradshaw et al. 2001). An increase in VEGF expression was noted in SPARC-null sponges and by SPARC-null dermal fibroblasts versus expression levels in WT mice and levels of VEGF produced by WT cells. The capacity of SPARC to reduce VEGF activity provided additional evidence to support an anti-angiogenic function of SPARC (Nozaki et al. 2006).

In glioblastomas, increased expression of SPARC decreased VEGF expression in part due to reduced levels of mRNA encoding VEGF 165 (Yunker et al. 2008). Likewise, purified SPARC protein inhibited angiogenesis in neuroblastoma xenografts inoculated to athymic nude mice (Chlenski et al. 2004). Chlenski et al. mapped the domain of SPARC responsible for inhibition of bFGF-induced migration of endothelial cells using a matrigel plug containing neuroblastoma cells delivered to athymic nude mice. Cysteine-linked peptides associated with distinct regions of SPARC were used to isolate anti-angiogenic activity. Peptide FS-E (representing amino acids 55–76) within the follistatin domain of SPARC was found to confer significant and specific inhibition of microvessel density (Chlenski et al. 2004), although within the follistatin domain, FS-E represents a distinct site from that found to be responsible for ILK activation mapped by Weaver et al. (Weaver et al. 2008).

In a mouse model of ovarian cancer, host-derived SPARC was shown to be an important contributor to cancer dissemination and lethality (Said and Motamed 2005). SPARC-null mice injected with syngeneic ID8 ovarian cancer cells developed greater peritoneal nodule dissemination and increased lethality in comparison to WT mice (Said and Motamed 2005). An increase in levels of VEGF was detected in SPARC-null ascitic fluid and was proposed to contribute to the increased invasion of the ID8 cells. SPARC was also shown to diminish basal and VEGF-induced activation of integrins in ID8 cells. Said et al. demonstrated that SPARC substantially reduced integrin activation and clustering, two critical aspects of integrin receptors required for cell movement and signal transduction events in ovarian cancer cells (Said et al. 2007).

SPARC activity has also been implicated in the regulation of TGF-β1. Mesangial cells isolated from SPARC-null mice were found to synthesize decreased amounts of collagen I and TGF-β1 in vitro (Francki et al. 1999). Addition of rSPARC restored collagen I and TGF-β1 expression to that approximating the level produced by WT cells. Francki et al. showed that SPARC appeared to influence TGF-β1 activity through an interaction with the TGF-β1/receptor II (TGFβRII) complex that was dependent upon TGF-β1 bound to receptor II (Francki et al. 2004). Schiemann et al. also found SPARC to influence TGF-β1 signaling pathways in epithelial cells (Schiemann et al. 2003).

Adenocarcinoma of the pancreas is a highly desmoplastic disease and is also frequently associated with mutations that effect the TGFβ pathway (Truty and Urrutia 2007). For instance, deletion of SMAD4 or mutation of TGFβRII in tumor cells occurs in greater than 50% of cases of pancreatic adenocarcinoma. We found recently that there is a significant increase in the level of active TGFβ1 in pancreatic tumors (Pan02) grown in SPARC-null mice, which corresponded to a more aggressive phenotype of these tumors in the absence of host-derived SPARC. The change in phenotype of the Pan02 tumors, which was SMAD4 null, might reflect an increase in TGF-β1 driven epithelial to mesenchymal transition or exacerbate known immune-suppressive effects of TGF-β1 in the tumor.

Immune System

SPARC activity has been shown to contribute significantly to inflammatory mediators particularly in animal models of tumor growth and invasion. A familiar theme with SPARC activity in cancer progression also held true with regard to immune response; the source of SPARC expression either by host stromal cells or by transformed tumor cells seemed to be an important contributor to the outcome (Prada et al. 2007; Haber et al. 2008).

For example, melanoma cells with suppressed SPARC expression injected into nude mice resulted in increased polymorphonuclear leukocyte (PMN) recruitment and abrogated tumor growth in comparison to tumors generated from melanoma cells with high SPARC expression (Alvarez et al. 2005). In vitro, melanoma cells with reduced SPARC expression induced PMN migration and antimelanoma cytotoxic activity whereas addition of rSPARC counteracted these effects. Seemingly, SPARC expression by melanoma cells decreases PMN recruitment, a first-line of defense in the immune surveillance against cancer, so that an inhibition of SPARC expression in melanoma cells enhanced the capacity of PMNs to combat tumor growth.

Conversely, SPARC expression by leukocytes might be an important factor in the recruitment of leukocytes from the vasculature. SPARC was identified as a counter ligand for the cell adhesion molecule vascular cell adhesion molecule-1 (VCAM-1) expressed on endothelial cells (Kelly et al. 2007). SPARC expressed by leukocytes interacted with VCAM-1 to initiate actin rearrangement in endothelial cells which led to the generation of intercellular gaps that allowed leukocyte transmigration through the endothelial monolayers, a process referred to as diapedesis. SPARC-null mice were shown to exhibit abnormalities in leukocyte recruitment to inflamed peritoneum (Kelly et al. 2007). Whereas SPARC expression by melanoma cells was inhibitory to PMN recruitment, SPARC expression by leukocytes was a critical step in inflammatory cell recruitment. These results highlight the complexity and potential pitfalls of targeting SPARC activity in cancer treatment.

SPARC-null mice have now been shown to exhibit aberrant splenic morphology - a demonstration that SPARC is a critical factor in the development of a competent immune system at least in mice. The spleens of SPARC-null mice were larger and had increased amounts of white pulp, hyperproliferative B cells in germinal centers, and marginal zones that were decreased compared with those of WT mice (Rempel et al. 2007). SPARC-null mice failed to generate an immune response after administration of lipopolysaccharide to the footpad whereas WT mice treated identically exhibited significant swelling. Although Rempel et al. stated that an increase in infections was observed in mouse colonies under their care - particularly in older mice - in >10 years of maintaining a SPARC-null colony, no differences in infection rates or progression to infection of superficial wounds between SPARC-null and WT mice have been noted at other sites. In addition, significant differences in life span of SPARC-null versus WT mice have not been reported.

In agreement with results that demonstrated an impaired immune response in SPARC-null mice, tumors induced with either Lewis lung carcinoma cells or pancreatic Pan02 adenocarcinoma cells demonstrated reduced macrophage recruitment in SPARC-null versus WT mice (Brekken et al. 2003; Puolakkainen et al. 2004). Therefore, SPARC has been proposed as an important mediator of macrophage recruitment. In the event that SPARC actively recruits macrophages, a process for regulating extracellular SPARC by macrophages might serve as an effective feed-back control mechanism. Stabilin-1 is a scavenger receptor expressed on alternatively activated macrophages and sinusoidal endothelial cells known to internalize and degrade acetylated low density lipid. Stabilin-1 was also shown to be a receptor for SPARC (Kzhyshkowska et al. 2006). Upon endocytosis of SPARC through stabilin-1, SPARC was targeted for lysosomal degradation. Hence, expression of stabilin-1 by macrophages enabled these inflammatory cells to clear SPARC from the extracellular milieu and thus reduced SPARC concentration and perhaps further macrophage recruitment. That SPARC is important for immune cell function is supported in part by two studies from Sangaletti et al. (Sangaletti et al. 2003; Sangaletti et al. 2005). In the first report, the authors demonstrated that SPARC produced by infiltrating leukocytes was instrumental in appropriate deposition of collagen IV in tumors from mammary carcinoma (Sangaletti et al. 2003). The second study found that dendritic cell migration and T-cell priming was enhanced in the absence of host SPARC (Sangaletti et al. 2005).

Conclusions

Significant changes in levels of mRNA encoding SPARC or SPARC protein are frequently revealed in studies that analyze expression profiles of tumors and transformed cell lines versus non-cancerous tissue and cells. SPARC has been shown to act as a modulator of cell adhesion, proliferation, survival, growth factor activity, and ECM assembly. As integrin receptors have also been implicated in each of the cellular processes mentioned above, the concept that SPARC regulates cell-ECM interaction through modulation of integrin binding is provocative. Integrin signaling is complex and contextual. For example, in endothelial cells, trans-inhibition of RGD binding integrins was observed upon engagement of collagen-binding integrins (Orr et al. 2006). Hence, one type of integrin receptor has the capacity to regulate function and down-stream signaling pathways of other types of integrin receptors within the same cell. In view of the fact that SPARC is generally classified as a counter-adhesive protein and integrins are best known as mediators of adhesion, one might anticipate that SPARC bound to a β1 integrin complex decreases integrin activity. However, a scenario in which SPARC influences a specific subset of integrin receptors might invoke a layer of complexity to cell adhesion/ECM assembly pathways that are yet to be fully characterized.

Convincing results generated from SPARC-null mice and tumor studies have established SPARC as a significant participant in collagen deposition and assembly in the ECM. As tumor biologists have long appreciated that different types of tumors possess different ECM signatures, perhaps that SPARC has diverse roles in different tumors is not surprising. Future experiments that define the molecular mechanisms and binding partners of SPARC in tumors will contribute enormously to future strategies to exploit the promise of manipulating host response to control tumor progression and invasion.